diff options
| author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-16 19:19:13 +0000 |
|---|---|---|
| committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2024-04-16 19:19:13 +0000 |
| commit | ccd992355df7192993c666236047820244914598 (patch) | |
| tree | f00fea65147227b7743083c6148396f74cd66935 /src/internal | |
| parent | Initial commit. (diff) | |
| download | golang-1.21-ccd992355df7192993c666236047820244914598.tar.xz golang-1.21-ccd992355df7192993c666236047820244914598.zip | |
Adding upstream version 1.21.8.upstream/1.21.8
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'src/internal')
837 files changed, 74776 insertions, 0 deletions
diff --git a/src/internal/abi/abi.go b/src/internal/abi/abi.go new file mode 100644 index 0000000..e1c8adc --- /dev/null +++ b/src/internal/abi/abi.go @@ -0,0 +1,102 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +import ( + "internal/goarch" + "unsafe" +) + +// RegArgs is a struct that has space for each argument +// and return value register on the current architecture. +// +// Assembly code knows the layout of the first two fields +// of RegArgs. +// +// RegArgs also contains additional space to hold pointers +// when it may not be safe to keep them only in the integer +// register space otherwise. +type RegArgs struct { + // Values in these slots should be precisely the bit-by-bit + // representation of how they would appear in a register. + // + // This means that on big endian arches, integer values should + // be in the top bits of the slot. Floats are usually just + // directly represented, but some architectures treat narrow + // width floating point values specially (e.g. they're promoted + // first, or they need to be NaN-boxed). + Ints [IntArgRegs]uintptr // untyped integer registers + Floats [FloatArgRegs]uint64 // untyped float registers + + // Fields above this point are known to assembly. + + // Ptrs is a space that duplicates Ints but with pointer type, + // used to make pointers passed or returned in registers + // visible to the GC by making the type unsafe.Pointer. + Ptrs [IntArgRegs]unsafe.Pointer + + // ReturnIsPtr is a bitmap that indicates which registers + // contain or will contain pointers on the return path from + // a reflectcall. The i'th bit indicates whether the i'th + // register contains or will contain a valid Go pointer. + ReturnIsPtr IntArgRegBitmap +} + +func (r *RegArgs) Dump() { + print("Ints:") + for _, x := range r.Ints { + print(" ", x) + } + println() + print("Floats:") + for _, x := range r.Floats { + print(" ", x) + } + println() + print("Ptrs:") + for _, x := range r.Ptrs { + print(" ", x) + } + println() +} + +// IntRegArgAddr returns a pointer inside of r.Ints[reg] that is appropriately +// offset for an argument of size argSize. +// +// argSize must be non-zero, fit in a register, and a power-of-two. +// +// This method is a helper for dealing with the endianness of different CPU +// architectures, since sub-word-sized arguments in big endian architectures +// need to be "aligned" to the upper edge of the register to be interpreted +// by the CPU correctly. +func (r *RegArgs) IntRegArgAddr(reg int, argSize uintptr) unsafe.Pointer { + if argSize > goarch.PtrSize || argSize == 0 || argSize&(argSize-1) != 0 { + panic("invalid argSize") + } + offset := uintptr(0) + if goarch.BigEndian { + offset = goarch.PtrSize - argSize + } + return unsafe.Pointer(uintptr(unsafe.Pointer(&r.Ints[reg])) + offset) +} + +// IntArgRegBitmap is a bitmap large enough to hold one bit per +// integer argument/return register. +type IntArgRegBitmap [(IntArgRegs + 7) / 8]uint8 + +// Set sets the i'th bit of the bitmap to 1. +func (b *IntArgRegBitmap) Set(i int) { + b[i/8] |= uint8(1) << (i % 8) +} + +// Get returns whether the i'th bit of the bitmap is set. +// +// nosplit because it's called in extremely sensitive contexts, like +// on the reflectcall return path. +// +//go:nosplit +func (b *IntArgRegBitmap) Get(i int) bool { + return b[i/8]&(uint8(1)<<(i%8)) != 0 +} diff --git a/src/internal/abi/abi_amd64.go b/src/internal/abi/abi_amd64.go new file mode 100644 index 0000000..d3c5678 --- /dev/null +++ b/src/internal/abi/abi_amd64.go @@ -0,0 +1,18 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +const ( + // See abi_generic.go. + + // RAX, RBX, RCX, RDI, RSI, R8, R9, R10, R11. + IntArgRegs = 9 + + // X0 -> X14. + FloatArgRegs = 15 + + // We use SSE2 registers which support 64-bit float operations. + EffectiveFloatRegSize = 8 +) diff --git a/src/internal/abi/abi_arm64.go b/src/internal/abi/abi_arm64.go new file mode 100644 index 0000000..4dc5143 --- /dev/null +++ b/src/internal/abi/abi_arm64.go @@ -0,0 +1,17 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +const ( + // See abi_generic.go. + + // R0 - R15. + IntArgRegs = 16 + + // F0 - F15. + FloatArgRegs = 16 + + EffectiveFloatRegSize = 8 +) diff --git a/src/internal/abi/abi_generic.go b/src/internal/abi/abi_generic.go new file mode 100644 index 0000000..76ef2e2 --- /dev/null +++ b/src/internal/abi/abi_generic.go @@ -0,0 +1,38 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !goexperiment.regabiargs && !amd64 && !arm64 && !ppc64 && !ppc64le && !riscv64 + +package abi + +const ( + // ABI-related constants. + // + // In the generic case, these are all zero + // which lets them gracefully degrade to ABI0. + + // IntArgRegs is the number of registers dedicated + // to passing integer argument values. Result registers are identical + // to argument registers, so this number is used for those too. + IntArgRegs = 0 + + // FloatArgRegs is the number of registers dedicated + // to passing floating-point argument values. Result registers are + // identical to argument registers, so this number is used for + // those too. + FloatArgRegs = 0 + + // EffectiveFloatRegSize describes the width of floating point + // registers on the current platform from the ABI's perspective. + // + // Since Go only supports 32-bit and 64-bit floating point primitives, + // this number should be either 0, 4, or 8. 0 indicates no floating + // point registers for the ABI or that floating point values will be + // passed via the softfloat ABI. + // + // For platforms that support larger floating point register widths, + // such as x87's 80-bit "registers" (not that we support x87 currently), + // use 8. + EffectiveFloatRegSize = 0 +) diff --git a/src/internal/abi/abi_ppc64x.go b/src/internal/abi/abi_ppc64x.go new file mode 100644 index 0000000..73416d7 --- /dev/null +++ b/src/internal/abi/abi_ppc64x.go @@ -0,0 +1,19 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package abi + +const ( + // See abi_generic.go. + + // R3 - R10, R14 - R17. + IntArgRegs = 12 + + // F1 - F12. + FloatArgRegs = 12 + + EffectiveFloatRegSize = 8 +) diff --git a/src/internal/abi/abi_riscv64.go b/src/internal/abi/abi_riscv64.go new file mode 100644 index 0000000..2bcd9d6 --- /dev/null +++ b/src/internal/abi/abi_riscv64.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +const ( + // See abi_generic.go. + + // X8 - X23 + IntArgRegs = 16 + + // F8 - F23. + FloatArgRegs = 16 + + EffectiveFloatRegSize = 8 +) diff --git a/src/internal/abi/abi_test.go b/src/internal/abi/abi_test.go new file mode 100644 index 0000000..44b9e78 --- /dev/null +++ b/src/internal/abi/abi_test.go @@ -0,0 +1,79 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi_test + +import ( + "internal/abi" + "internal/testenv" + "path/filepath" + "strings" + "testing" +) + +func TestFuncPC(t *testing.T) { + // Test that FuncPC* can get correct function PC. + pcFromAsm := abi.FuncPCTestFnAddr + + // Test FuncPC for locally defined function + pcFromGo := abi.FuncPCTest() + if pcFromGo != pcFromAsm { + t.Errorf("FuncPC returns wrong PC, want %x, got %x", pcFromAsm, pcFromGo) + } + + // Test FuncPC for imported function + pcFromGo = abi.FuncPCABI0(abi.FuncPCTestFn) + if pcFromGo != pcFromAsm { + t.Errorf("FuncPC returns wrong PC, want %x, got %x", pcFromAsm, pcFromGo) + } +} + +func TestFuncPCCompileError(t *testing.T) { + // Test that FuncPC* on a function of a mismatched ABI is rejected. + testenv.MustHaveGoBuild(t) + + // We want to test internal package, which we cannot normally import. + // Run the assembler and compiler manually. + tmpdir := t.TempDir() + asmSrc := filepath.Join("testdata", "x.s") + goSrc := filepath.Join("testdata", "x.go") + symabi := filepath.Join(tmpdir, "symabi") + obj := filepath.Join(tmpdir, "x.o") + + // Write an importcfg file for the dependencies of the package. + importcfgfile := filepath.Join(tmpdir, "hello.importcfg") + testenv.WriteImportcfg(t, importcfgfile, nil, "internal/abi") + + // parse assembly code for symabi. + cmd := testenv.Command(t, testenv.GoToolPath(t), "tool", "asm", "-gensymabis", "-o", symabi, asmSrc) + out, err := cmd.CombinedOutput() + if err != nil { + t.Fatalf("go tool asm -gensymabis failed: %v\n%s", err, out) + } + + // compile go code. + cmd = testenv.Command(t, testenv.GoToolPath(t), "tool", "compile", "-importcfg="+importcfgfile, "-p=p", "-symabis", symabi, "-o", obj, goSrc) + out, err = cmd.CombinedOutput() + if err == nil { + t.Fatalf("go tool compile did not fail") + } + + // Expect errors in line 17, 18, 20, no errors on other lines. + want := []string{"x.go:17", "x.go:18", "x.go:20"} + got := strings.Split(string(out), "\n") + if got[len(got)-1] == "" { + got = got[:len(got)-1] // remove last empty line + } + for i, s := range got { + if !strings.Contains(s, want[i]) { + t.Errorf("did not error on line %s", want[i]) + } + } + if len(got) != len(want) { + t.Errorf("unexpected number of errors, want %d, got %d", len(want), len(got)) + } + if t.Failed() { + t.Logf("output:\n%s", string(out)) + } +} diff --git a/src/internal/abi/abi_test.s b/src/internal/abi/abi_test.s new file mode 100644 index 0000000..93ace3e --- /dev/null +++ b/src/internal/abi/abi_test.s @@ -0,0 +1,27 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +#ifdef GOARCH_386 +#define PTRSIZE 4 +#endif +#ifdef GOARCH_arm +#define PTRSIZE 4 +#endif +#ifdef GOARCH_mips +#define PTRSIZE 4 +#endif +#ifdef GOARCH_mipsle +#define PTRSIZE 4 +#endif +#ifndef PTRSIZE +#define PTRSIZE 8 +#endif + +TEXT internal∕abi·FuncPCTestFn(SB),NOSPLIT,$0-0 + RET + +GLOBL internal∕abi·FuncPCTestFnAddr(SB), NOPTR, $PTRSIZE +DATA internal∕abi·FuncPCTestFnAddr(SB)/PTRSIZE, $internal∕abi·FuncPCTestFn(SB) diff --git a/src/internal/abi/compiletype.go b/src/internal/abi/compiletype.go new file mode 100644 index 0000000..d92adde --- /dev/null +++ b/src/internal/abi/compiletype.go @@ -0,0 +1,167 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +// These functions are the build-time version of the Go type data structures. + +// Their contents must be kept in sync with their definitions. +// Because the host and target type sizes can differ, the compiler and +// linker cannot use the host information that they might get from +// either unsafe.Sizeof and Alignof, nor runtime, reflect, or reflectlite. + +// CommonSize returns sizeof(Type) for a compilation target with a given ptrSize +func CommonSize(ptrSize int) int { return 4*ptrSize + 8 + 8 } + +// StructFieldSize returns sizeof(StructField) for a compilation target with a given ptrSize +func StructFieldSize(ptrSize int) int { return 3 * ptrSize } + +// UncommonSize returns sizeof(UncommonType). This currently does not depend on ptrSize. +// This exported function is in an internal package, so it may change to depend on ptrSize in the future. +func UncommonSize() uint64 { return 4 + 2 + 2 + 4 + 4 } + +// IMethodSize returns sizeof(IMethod) for a compilation target with a given ptrSize +func IMethodSize(ptrSize int) int { return 4 + 4 } + +// KindOff returns the offset of Type.Kind_ for a compilation target with a given ptrSize +func KindOff(ptrSize int) int { return 2*ptrSize + 7 } + +// SizeOff returns the offset of Type.Size_ for a compilation target with a given ptrSize +func SizeOff(ptrSize int) int { return 0 } + +// PtrBytes returns the offset of Type.PtrBytes for a compilation target with a given ptrSize +func PtrBytesOff(ptrSize int) int { return ptrSize } + +// TFlagOff returns the offset of Type.TFlag for a compilation target with a given ptrSize +func TFlagOff(ptrSize int) int { return 2*ptrSize + 4 } + +// Offset is for computing offsets of type data structures at compile/link time; +// the target platform may not be the host platform. Its state includes the +// current offset, necessary alignment for the sequence of types, and the size +// of pointers and alignment of slices, interfaces, and strings (this is for tearing- +// resistant access to these types, if/when that is supported). +type Offset struct { + off uint64 // the current offset + align uint8 // the required alignmentof the container + ptrSize uint8 // the size of a pointer in bytes + sliceAlign uint8 // the alignment of slices (and interfaces and strings) +} + +// NewOffset returns a new Offset with offset 0 and alignment 1. +func NewOffset(ptrSize uint8, twoWordAlignSlices bool) Offset { + if twoWordAlignSlices { + return Offset{off: 0, align: 1, ptrSize: ptrSize, sliceAlign: 2 * ptrSize} + } + return Offset{off: 0, align: 1, ptrSize: ptrSize, sliceAlign: ptrSize} +} + +func assertIsAPowerOfTwo(x uint8) { + if x == 0 { + panic("Zero is not a power of two") + } + if x&-x == x { + return + } + panic("Not a power of two") +} + +// InitializedOffset returns a new Offset with specified offset, alignment, pointer size, and slice alignment. +func InitializedOffset(off int, align uint8, ptrSize uint8, twoWordAlignSlices bool) Offset { + assertIsAPowerOfTwo(align) + o0 := NewOffset(ptrSize, twoWordAlignSlices) + o0.off = uint64(off) + o0.align = align + return o0 +} + +func (o Offset) align_(a uint8) Offset { + o.off = (o.off + uint64(a) - 1) & ^(uint64(a) - 1) + if o.align < a { + o.align = a + } + return o +} + +// Align returns the offset obtained by aligning offset to a multiple of a. +// a must be a power of two. +func (o Offset) Align(a uint8) Offset { + assertIsAPowerOfTwo(a) + return o.align_(a) +} + +// plus returns the offset obtained by appending a power-of-2-sized-and-aligned object to o. +func (o Offset) plus(x uint64) Offset { + o = o.align_(uint8(x)) + o.off += x + return o +} + +// D8 returns the offset obtained by appending an 8-bit field to o. +func (o Offset) D8() Offset { + return o.plus(1) +} + +// D16 returns the offset obtained by appending a 16-bit field to o. +func (o Offset) D16() Offset { + return o.plus(2) +} + +// D32 returns the offset obtained by appending a 32-bit field to o. +func (o Offset) D32() Offset { + return o.plus(4) +} + +// D64 returns the offset obtained by appending a 64-bit field to o. +func (o Offset) D64() Offset { + return o.plus(8) +} + +// D64 returns the offset obtained by appending a pointer field to o. +func (o Offset) P() Offset { + if o.ptrSize == 0 { + panic("This offset has no defined pointer size") + } + return o.plus(uint64(o.ptrSize)) +} + +// Slice returns the offset obtained by appending a slice field to o. +func (o Offset) Slice() Offset { + o = o.align_(o.sliceAlign) + o.off += 3 * uint64(o.ptrSize) + // There's been discussion of whether slices should be 2-word aligned to allow + // use of aligned 2-word load/store to prevent tearing, this is future proofing. + // In general, for purposes of struct layout (and very likely default C layout + // compatibility) the "size" of a Go type is rounded up to its alignment. + return o.Align(o.sliceAlign) +} + +// String returns the offset obtained by appending a string field to o. +func (o Offset) String() Offset { + o = o.align_(o.sliceAlign) + o.off += 2 * uint64(o.ptrSize) + return o // We "know" it needs no further alignment +} + +// Interface returns the offset obtained by appending an interface field to o. +func (o Offset) Interface() Offset { + o = o.align_(o.sliceAlign) + o.off += 2 * uint64(o.ptrSize) + return o // We "know" it needs no further alignment +} + +// Offset returns the struct-aligned offset (size) of o. +// This is at least as large as the current internal offset; it may be larger. +func (o Offset) Offset() uint64 { + return o.Align(o.align).off +} + +func (o Offset) PlusUncommon() Offset { + o.off += UncommonSize() + return o +} + +// CommonOffset returns the Offset to the data after the common portion of type data structures. +func CommonOffset(ptrSize int, twoWordAlignSlices bool) Offset { + return InitializedOffset(CommonSize(ptrSize), uint8(ptrSize), uint8(ptrSize), twoWordAlignSlices) +} diff --git a/src/internal/abi/export_test.go b/src/internal/abi/export_test.go new file mode 100644 index 0000000..2a87e9d --- /dev/null +++ b/src/internal/abi/export_test.go @@ -0,0 +1,14 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +func FuncPCTestFn() + +var FuncPCTestFnAddr uintptr // address of FuncPCTestFn, directly retrieved from assembly + +//go:noinline +func FuncPCTest() uintptr { + return FuncPCABI0(FuncPCTestFn) +} diff --git a/src/internal/abi/funcpc.go b/src/internal/abi/funcpc.go new file mode 100644 index 0000000..e038d36 --- /dev/null +++ b/src/internal/abi/funcpc.go @@ -0,0 +1,31 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !gccgo + +package abi + +// FuncPC* intrinsics. +// +// CAREFUL: In programs with plugins, FuncPC* can return different values +// for the same function (because there are actually multiple copies of +// the same function in the address space). To be safe, don't use the +// results of this function in any == expression. It is only safe to +// use the result as an address at which to start executing code. + +// FuncPCABI0 returns the entry PC of the function f, which must be a +// direct reference of a function defined as ABI0. Otherwise it is a +// compile-time error. +// +// Implemented as a compile intrinsic. +func FuncPCABI0(f interface{}) uintptr + +// FuncPCABIInternal returns the entry PC of the function f. If f is a +// direct reference of a function, it must be defined as ABIInternal. +// Otherwise it is a compile-time error. If f is not a direct reference +// of a defined function, it assumes that f is a func value. Otherwise +// the behavior is undefined. +// +// Implemented as a compile intrinsic. +func FuncPCABIInternal(f interface{}) uintptr diff --git a/src/internal/abi/funcpc_gccgo.go b/src/internal/abi/funcpc_gccgo.go new file mode 100644 index 0000000..ad5fa52 --- /dev/null +++ b/src/internal/abi/funcpc_gccgo.go @@ -0,0 +1,21 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// For bootstrapping with gccgo. + +//go:build gccgo + +package abi + +import "unsafe" + +func FuncPCABI0(f interface{}) uintptr { + words := (*[2]unsafe.Pointer)(unsafe.Pointer(&f)) + return *(*uintptr)(unsafe.Pointer(words[1])) +} + +func FuncPCABIInternal(f interface{}) uintptr { + words := (*[2]unsafe.Pointer)(unsafe.Pointer(&f)) + return *(*uintptr)(unsafe.Pointer(words[1])) +} diff --git a/src/internal/abi/map.go b/src/internal/abi/map.go new file mode 100644 index 0000000..e5b0a0b --- /dev/null +++ b/src/internal/abi/map.go @@ -0,0 +1,14 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +// Map constants common to several packages +// runtime/runtime-gdb.py:MapTypePrinter contains its own copy +const ( + MapBucketCountBits = 3 // log2 of number of elements in a bucket. + MapBucketCount = 1 << MapBucketCountBits + MapMaxKeyBytes = 128 // Must fit in a uint8. + MapMaxElemBytes = 128 // Must fit in a uint8. +) diff --git a/src/internal/abi/stack.go b/src/internal/abi/stack.go new file mode 100644 index 0000000..8e3327e --- /dev/null +++ b/src/internal/abi/stack.go @@ -0,0 +1,33 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +const ( + // StackNosplitBase is the base maximum number of bytes that a chain of + // NOSPLIT functions can use. + // + // This value must be multiplied by the stack guard multiplier, so do not + // use it directly. See runtime/stack.go:stackNosplit and + // cmd/internal/objabi/stack.go:StackNosplit. + StackNosplitBase = 800 + + // We have three different sequences for stack bounds checks, depending on + // whether the stack frame of a function is small, big, or huge. + + // After a stack split check the SP is allowed to be StackSmall bytes below + // the stack guard. + // + // Functions that need frames <= StackSmall can perform the stack check + // using a single comparison directly between the stack guard and the SP + // because we ensure that StackSmall bytes of stack space are available + // beyond the stack guard. + StackSmall = 128 + + // Functions that need frames <= StackBig can assume that neither + // SP-framesize nor stackGuard-StackSmall will underflow, and thus use a + // more efficient check. In order to ensure this, StackBig must be <= the + // size of the unmapped space at zero. + StackBig = 4096 +) diff --git a/src/internal/abi/stub.s b/src/internal/abi/stub.s new file mode 100644 index 0000000..5bad98d --- /dev/null +++ b/src/internal/abi/stub.s @@ -0,0 +1,7 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file silences errors about body-less functions +// that are provided by intrinsics in the latest version of the compiler, +// but may not be known to the bootstrap compiler. diff --git a/src/internal/abi/symtab.go b/src/internal/abi/symtab.go new file mode 100644 index 0000000..bf6ea82 --- /dev/null +++ b/src/internal/abi/symtab.go @@ -0,0 +1,106 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +// A FuncFlag records bits about a function, passed to the runtime. +type FuncFlag uint8 + +const ( + // FuncFlagTopFrame indicates a function that appears at the top of its stack. + // The traceback routine stop at such a function and consider that a + // successful, complete traversal of the stack. + // Examples of TopFrame functions include goexit, which appears + // at the top of a user goroutine stack, and mstart, which appears + // at the top of a system goroutine stack. + FuncFlagTopFrame FuncFlag = 1 << iota + + // FuncFlagSPWrite indicates a function that writes an arbitrary value to SP + // (any write other than adding or subtracting a constant amount). + // The traceback routines cannot encode such changes into the + // pcsp tables, so the function traceback cannot safely unwind past + // SPWrite functions. Stopping at an SPWrite function is considered + // to be an incomplete unwinding of the stack. In certain contexts + // (in particular garbage collector stack scans) that is a fatal error. + FuncFlagSPWrite + + // FuncFlagAsm indicates that a function was implemented in assembly. + FuncFlagAsm +) + +// A FuncID identifies particular functions that need to be treated +// specially by the runtime. +// Note that in some situations involving plugins, there may be multiple +// copies of a particular special runtime function. +type FuncID uint8 + +const ( + // If you add a FuncID, you probably also want to add an entry to the map in + // ../../cmd/internal/objabi/funcid.go + + FuncIDNormal FuncID = iota // not a special function + FuncID_abort + FuncID_asmcgocall + FuncID_asyncPreempt + FuncID_cgocallback + FuncID_debugCallV2 + FuncID_gcBgMarkWorker + FuncID_goexit + FuncID_gogo + FuncID_gopanic + FuncID_handleAsyncEvent + FuncID_mcall + FuncID_morestack + FuncID_mstart + FuncID_panicwrap + FuncID_rt0_go + FuncID_runfinq + FuncID_runtime_main + FuncID_sigpanic + FuncID_systemstack + FuncID_systemstack_switch + FuncIDWrapper // any autogenerated code (hash/eq algorithms, method wrappers, etc.) +) + +// ArgsSizeUnknown is set in Func.argsize to mark all functions +// whose argument size is unknown (C vararg functions, and +// assembly code without an explicit specification). +// This value is generated by the compiler, assembler, or linker. +const ArgsSizeUnknown = -0x80000000 + +// IDs for PCDATA and FUNCDATA tables in Go binaries. +// +// These must agree with ../../../runtime/funcdata.h. +const ( + PCDATA_UnsafePoint = 0 + PCDATA_StackMapIndex = 1 + PCDATA_InlTreeIndex = 2 + PCDATA_ArgLiveIndex = 3 + + FUNCDATA_ArgsPointerMaps = 0 + FUNCDATA_LocalsPointerMaps = 1 + FUNCDATA_StackObjects = 2 + FUNCDATA_InlTree = 3 + FUNCDATA_OpenCodedDeferInfo = 4 + FUNCDATA_ArgInfo = 5 + FUNCDATA_ArgLiveInfo = 6 + FUNCDATA_WrapInfo = 7 +) + +// Special values for the PCDATA_UnsafePoint table. +const ( + UnsafePointSafe = -1 // Safe for async preemption + UnsafePointUnsafe = -2 // Unsafe for async preemption + + // UnsafePointRestart1(2) apply on a sequence of instructions, within + // which if an async preemption happens, we should back off the PC + // to the start of the sequence when resuming. + // We need two so we can distinguish the start/end of the sequence + // in case that two sequences are next to each other. + UnsafePointRestart1 = -3 + UnsafePointRestart2 = -4 + + // Like UnsafePointRestart1, but back to function entry if async preempted. + UnsafePointRestartAtEntry = -5 +) diff --git a/src/internal/abi/testdata/x.go b/src/internal/abi/testdata/x.go new file mode 100644 index 0000000..cae103d --- /dev/null +++ b/src/internal/abi/testdata/x.go @@ -0,0 +1,22 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package x + +import "internal/abi" + +func Fn0() // defined in assembly + +func Fn1() {} + +var FnExpr func() + +func test() { + _ = abi.FuncPCABI0(Fn0) // line 16, no error + _ = abi.FuncPCABIInternal(Fn0) // line 17, error + _ = abi.FuncPCABI0(Fn1) // line 18, error + _ = abi.FuncPCABIInternal(Fn1) // line 19, no error + _ = abi.FuncPCABI0(FnExpr) // line 20, error + _ = abi.FuncPCABIInternal(FnExpr) // line 21, no error +} diff --git a/src/internal/abi/testdata/x.s b/src/internal/abi/testdata/x.s new file mode 100644 index 0000000..63c1385 --- /dev/null +++ b/src/internal/abi/testdata/x.s @@ -0,0 +1,6 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +TEXT ·Fn0(SB), 0, $0-0 + RET diff --git a/src/internal/abi/type.go b/src/internal/abi/type.go new file mode 100644 index 0000000..4794f5a --- /dev/null +++ b/src/internal/abi/type.go @@ -0,0 +1,712 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package abi + +import ( + "unsafe" +) + +// Type is the runtime representation of a Go type. +// +// Type is also referenced implicitly +// (in the form of expressions involving constants and arch.PtrSize) +// in cmd/compile/internal/reflectdata/reflect.go +// and cmd/link/internal/ld/decodesym.go +// (e.g. data[2*arch.PtrSize+4] references the TFlag field) +// unsafe.OffsetOf(Type{}.TFlag) cannot be used directly in those +// places because it varies with cross compilation and experiments. +type Type struct { + Size_ uintptr + PtrBytes uintptr // number of (prefix) bytes in the type that can contain pointers + Hash uint32 // hash of type; avoids computation in hash tables + TFlag TFlag // extra type information flags + Align_ uint8 // alignment of variable with this type + FieldAlign_ uint8 // alignment of struct field with this type + Kind_ uint8 // enumeration for C + // function for comparing objects of this type + // (ptr to object A, ptr to object B) -> ==? + Equal func(unsafe.Pointer, unsafe.Pointer) bool + // GCData stores the GC type data for the garbage collector. + // If the KindGCProg bit is set in kind, GCData is a GC program. + // Otherwise it is a ptrmask bitmap. See mbitmap.go for details. + GCData *byte + Str NameOff // string form + PtrToThis TypeOff // type for pointer to this type, may be zero +} + +// A Kind represents the specific kind of type that a Type represents. +// The zero Kind is not a valid kind. +type Kind uint + +const ( + Invalid Kind = iota + Bool + Int + Int8 + Int16 + Int32 + Int64 + Uint + Uint8 + Uint16 + Uint32 + Uint64 + Uintptr + Float32 + Float64 + Complex64 + Complex128 + Array + Chan + Func + Interface + Map + Pointer + Slice + String + Struct + UnsafePointer +) + +const ( + // TODO (khr, drchase) why aren't these in TFlag? Investigate, fix if possible. + KindDirectIface = 1 << 5 + KindGCProg = 1 << 6 // Type.gc points to GC program + KindMask = (1 << 5) - 1 +) + +// TFlag is used by a Type to signal what extra type information is +// available in the memory directly following the Type value. +type TFlag uint8 + +const ( + // TFlagUncommon means that there is a data with a type, UncommonType, + // just beyond the shared-per-type common data. That is, the data + // for struct types will store their UncommonType at one offset, the + // data for interface types will store their UncommonType at a different + // offset. UncommonType is always accessed via a pointer that is computed + // using trust-us-we-are-the-implementors pointer arithmetic. + // + // For example, if t.Kind() == Struct and t.tflag&TFlagUncommon != 0, + // then t has UncommonType data and it can be accessed as: + // + // type structTypeUncommon struct { + // structType + // u UncommonType + // } + // u := &(*structTypeUncommon)(unsafe.Pointer(t)).u + TFlagUncommon TFlag = 1 << 0 + + // TFlagExtraStar means the name in the str field has an + // extraneous '*' prefix. This is because for most types T in + // a program, the type *T also exists and reusing the str data + // saves binary size. + TFlagExtraStar TFlag = 1 << 1 + + // TFlagNamed means the type has a name. + TFlagNamed TFlag = 1 << 2 + + // TFlagRegularMemory means that equal and hash functions can treat + // this type as a single region of t.size bytes. + TFlagRegularMemory TFlag = 1 << 3 +) + +// NameOff is the offset to a name from moduledata.types. See resolveNameOff in runtime. +type NameOff int32 + +// TypeOff is the offset to a type from moduledata.types. See resolveTypeOff in runtime. +type TypeOff int32 + +// TextOff is an offset from the top of a text section. See (rtype).textOff in runtime. +type TextOff int32 + +// String returns the name of k. +func (k Kind) String() string { + if int(k) < len(kindNames) { + return kindNames[k] + } + return kindNames[0] +} + +var kindNames = []string{ + Invalid: "invalid", + Bool: "bool", + Int: "int", + Int8: "int8", + Int16: "int16", + Int32: "int32", + Int64: "int64", + Uint: "uint", + Uint8: "uint8", + Uint16: "uint16", + Uint32: "uint32", + Uint64: "uint64", + Uintptr: "uintptr", + Float32: "float32", + Float64: "float64", + Complex64: "complex64", + Complex128: "complex128", + Array: "array", + Chan: "chan", + Func: "func", + Interface: "interface", + Map: "map", + Pointer: "ptr", + Slice: "slice", + String: "string", + Struct: "struct", + UnsafePointer: "unsafe.Pointer", +} + +func (t *Type) Kind() Kind { return Kind(t.Kind_ & KindMask) } + +func (t *Type) HasName() bool { + return t.TFlag&TFlagNamed != 0 +} + +func (t *Type) Pointers() bool { return t.PtrBytes != 0 } + +// IfaceIndir reports whether t is stored indirectly in an interface value. +func (t *Type) IfaceIndir() bool { + return t.Kind_&KindDirectIface == 0 +} + +// isDirectIface reports whether t is stored directly in an interface value. +func (t *Type) IsDirectIface() bool { + return t.Kind_&KindDirectIface != 0 +} + +func (t *Type) GcSlice(begin, end uintptr) []byte { + return unsafeSliceFor(t.GCData, int(end))[begin:] +} + +// Method on non-interface type +type Method struct { + Name NameOff // name of method + Mtyp TypeOff // method type (without receiver) + Ifn TextOff // fn used in interface call (one-word receiver) + Tfn TextOff // fn used for normal method call +} + +// UncommonType is present only for defined types or types with methods +// (if T is a defined type, the uncommonTypes for T and *T have methods). +// Using a pointer to this struct reduces the overall size required +// to describe a non-defined type with no methods. +type UncommonType struct { + PkgPath NameOff // import path; empty for built-in types like int, string + Mcount uint16 // number of methods + Xcount uint16 // number of exported methods + Moff uint32 // offset from this uncommontype to [mcount]Method + _ uint32 // unused +} + +func (t *UncommonType) Methods() []Method { + if t.Mcount == 0 { + return nil + } + return (*[1 << 16]Method)(addChecked(unsafe.Pointer(t), uintptr(t.Moff), "t.mcount > 0"))[:t.Mcount:t.Mcount] +} + +func (t *UncommonType) ExportedMethods() []Method { + if t.Xcount == 0 { + return nil + } + return (*[1 << 16]Method)(addChecked(unsafe.Pointer(t), uintptr(t.Moff), "t.xcount > 0"))[:t.Xcount:t.Xcount] +} + +// addChecked returns p+x. +// +// The whySafe string is ignored, so that the function still inlines +// as efficiently as p+x, but all call sites should use the string to +// record why the addition is safe, which is to say why the addition +// does not cause x to advance to the very end of p's allocation +// and therefore point incorrectly at the next block in memory. +func addChecked(p unsafe.Pointer, x uintptr, whySafe string) unsafe.Pointer { + return unsafe.Pointer(uintptr(p) + x) +} + +// Imethod represents a method on an interface type +type Imethod struct { + Name NameOff // name of method + Typ TypeOff // .(*FuncType) underneath +} + +// ArrayType represents a fixed array type. +type ArrayType struct { + Type + Elem *Type // array element type + Slice *Type // slice type + Len uintptr +} + +// Len returns the length of t if t is an array type, otherwise 0 +func (t *Type) Len() int { + if t.Kind() == Array { + return int((*ArrayType)(unsafe.Pointer(t)).Len) + } + return 0 +} + +func (t *Type) Common() *Type { + return t +} + +type ChanDir int + +const ( + RecvDir ChanDir = 1 << iota // <-chan + SendDir // chan<- + BothDir = RecvDir | SendDir // chan + InvalidDir ChanDir = 0 +) + +// ChanType represents a channel type +type ChanType struct { + Type + Elem *Type + Dir ChanDir +} + +type structTypeUncommon struct { + StructType + u UncommonType +} + +// ChanDir returns the direction of t if t is a channel type, otherwise InvalidDir (0). +func (t *Type) ChanDir() ChanDir { + if t.Kind() == Chan { + ch := (*ChanType)(unsafe.Pointer(t)) + return ch.Dir + } + return InvalidDir +} + +// Uncommon returns a pointer to T's "uncommon" data if there is any, otherwise nil +func (t *Type) Uncommon() *UncommonType { + if t.TFlag&TFlagUncommon == 0 { + return nil + } + switch t.Kind() { + case Struct: + return &(*structTypeUncommon)(unsafe.Pointer(t)).u + case Pointer: + type u struct { + PtrType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Func: + type u struct { + FuncType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Slice: + type u struct { + SliceType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Array: + type u struct { + ArrayType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Chan: + type u struct { + ChanType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Map: + type u struct { + MapType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + case Interface: + type u struct { + InterfaceType + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + default: + type u struct { + Type + u UncommonType + } + return &(*u)(unsafe.Pointer(t)).u + } +} + +// Elem returns the element type for t if t is an array, channel, map, pointer, or slice, otherwise nil. +func (t *Type) Elem() *Type { + switch t.Kind() { + case Array: + tt := (*ArrayType)(unsafe.Pointer(t)) + return tt.Elem + case Chan: + tt := (*ChanType)(unsafe.Pointer(t)) + return tt.Elem + case Map: + tt := (*MapType)(unsafe.Pointer(t)) + return tt.Elem + case Pointer: + tt := (*PtrType)(unsafe.Pointer(t)) + return tt.Elem + case Slice: + tt := (*SliceType)(unsafe.Pointer(t)) + return tt.Elem + } + return nil +} + +// StructType returns t cast to a *StructType, or nil if its tag does not match. +func (t *Type) StructType() *StructType { + if t.Kind() != Struct { + return nil + } + return (*StructType)(unsafe.Pointer(t)) +} + +// MapType returns t cast to a *MapType, or nil if its tag does not match. +func (t *Type) MapType() *MapType { + if t.Kind() != Map { + return nil + } + return (*MapType)(unsafe.Pointer(t)) +} + +// ArrayType returns t cast to a *ArrayType, or nil if its tag does not match. +func (t *Type) ArrayType() *ArrayType { + if t.Kind() != Array { + return nil + } + return (*ArrayType)(unsafe.Pointer(t)) +} + +// FuncType returns t cast to a *FuncType, or nil if its tag does not match. +func (t *Type) FuncType() *FuncType { + if t.Kind() != Func { + return nil + } + return (*FuncType)(unsafe.Pointer(t)) +} + +// InterfaceType returns t cast to a *InterfaceType, or nil if its tag does not match. +func (t *Type) InterfaceType() *InterfaceType { + if t.Kind() != Interface { + return nil + } + return (*InterfaceType)(unsafe.Pointer(t)) +} + +// Size returns the size of data with type t. +func (t *Type) Size() uintptr { return t.Size_ } + +// Align returns the alignment of data with type t. +func (t *Type) Align() int { return int(t.Align_) } + +func (t *Type) FieldAlign() int { return int(t.FieldAlign_) } + +type InterfaceType struct { + Type + PkgPath Name // import path + Methods []Imethod // sorted by hash +} + +func (t *Type) ExportedMethods() []Method { + ut := t.Uncommon() + if ut == nil { + return nil + } + return ut.ExportedMethods() +} + +func (t *Type) NumMethod() int { + if t.Kind() == Interface { + tt := (*InterfaceType)(unsafe.Pointer(t)) + return tt.NumMethod() + } + return len(t.ExportedMethods()) +} + +// NumMethod returns the number of interface methods in the type's method set. +func (t *InterfaceType) NumMethod() int { return len(t.Methods) } + +type MapType struct { + Type + Key *Type + Elem *Type + Bucket *Type // internal type representing a hash bucket + // function for hashing keys (ptr to key, seed) -> hash + Hasher func(unsafe.Pointer, uintptr) uintptr + KeySize uint8 // size of key slot + ValueSize uint8 // size of elem slot + BucketSize uint16 // size of bucket + Flags uint32 +} + +// Note: flag values must match those used in the TMAP case +// in ../cmd/compile/internal/reflectdata/reflect.go:writeType. +func (mt *MapType) IndirectKey() bool { // store ptr to key instead of key itself + return mt.Flags&1 != 0 +} +func (mt *MapType) IndirectElem() bool { // store ptr to elem instead of elem itself + return mt.Flags&2 != 0 +} +func (mt *MapType) ReflexiveKey() bool { // true if k==k for all keys + return mt.Flags&4 != 0 +} +func (mt *MapType) NeedKeyUpdate() bool { // true if we need to update key on an overwrite + return mt.Flags&8 != 0 +} +func (mt *MapType) HashMightPanic() bool { // true if hash function might panic + return mt.Flags&16 != 0 +} + +func (t *Type) Key() *Type { + if t.Kind() == Map { + return (*MapType)(unsafe.Pointer(t)).Key + } + return nil +} + +type SliceType struct { + Type + Elem *Type // slice element type +} + +// funcType represents a function type. +// +// A *Type for each in and out parameter is stored in an array that +// directly follows the funcType (and possibly its uncommonType). So +// a function type with one method, one input, and one output is: +// +// struct { +// funcType +// uncommonType +// [2]*rtype // [0] is in, [1] is out +// } +type FuncType struct { + Type + InCount uint16 + OutCount uint16 // top bit is set if last input parameter is ... +} + +func (t *FuncType) In(i int) *Type { + return t.InSlice()[i] +} + +func (t *FuncType) NumIn() int { + return int(t.InCount) +} + +func (t *FuncType) NumOut() int { + return int(t.OutCount & (1<<15 - 1)) +} + +func (t *FuncType) Out(i int) *Type { + return (t.OutSlice()[i]) +} + +func (t *FuncType) InSlice() []*Type { + uadd := unsafe.Sizeof(*t) + if t.TFlag&TFlagUncommon != 0 { + uadd += unsafe.Sizeof(UncommonType{}) + } + if t.InCount == 0 { + return nil + } + return (*[1 << 16]*Type)(addChecked(unsafe.Pointer(t), uadd, "t.inCount > 0"))[:t.InCount:t.InCount] +} +func (t *FuncType) OutSlice() []*Type { + outCount := uint16(t.NumOut()) + if outCount == 0 { + return nil + } + uadd := unsafe.Sizeof(*t) + if t.TFlag&TFlagUncommon != 0 { + uadd += unsafe.Sizeof(UncommonType{}) + } + return (*[1 << 17]*Type)(addChecked(unsafe.Pointer(t), uadd, "outCount > 0"))[t.InCount : t.InCount+outCount : t.InCount+outCount] +} + +func (t *FuncType) IsVariadic() bool { + return t.OutCount&(1<<15) != 0 +} + +type PtrType struct { + Type + Elem *Type // pointer element (pointed at) type +} + +type StructField struct { + Name Name // name is always non-empty + Typ *Type // type of field + Offset uintptr // byte offset of field +} + +func (f *StructField) Embedded() bool { + return f.Name.IsEmbedded() +} + +type StructType struct { + Type + PkgPath Name + Fields []StructField +} + +// Name is an encoded type Name with optional extra data. +// +// The first byte is a bit field containing: +// +// 1<<0 the name is exported +// 1<<1 tag data follows the name +// 1<<2 pkgPath nameOff follows the name and tag +// 1<<3 the name is of an embedded (a.k.a. anonymous) field +// +// Following that, there is a varint-encoded length of the name, +// followed by the name itself. +// +// If tag data is present, it also has a varint-encoded length +// followed by the tag itself. +// +// If the import path follows, then 4 bytes at the end of +// the data form a nameOff. The import path is only set for concrete +// methods that are defined in a different package than their type. +// +// If a name starts with "*", then the exported bit represents +// whether the pointed to type is exported. +// +// Note: this encoding must match here and in: +// cmd/compile/internal/reflectdata/reflect.go +// cmd/link/internal/ld/decodesym.go + +type Name struct { + Bytes *byte +} + +// DataChecked does pointer arithmetic on n's Bytes, and that arithmetic is asserted to +// be safe for the reason in whySafe (which can appear in a backtrace, etc.) +func (n Name) DataChecked(off int, whySafe string) *byte { + return (*byte)(addChecked(unsafe.Pointer(n.Bytes), uintptr(off), whySafe)) +} + +// Data does pointer arithmetic on n's Bytes, and that arithmetic is asserted to +// be safe because the runtime made the call (other packages use DataChecked) +func (n Name) Data(off int) *byte { + return (*byte)(addChecked(unsafe.Pointer(n.Bytes), uintptr(off), "the runtime doesn't need to give you a reason")) +} + +// IsExported returns "is n exported?" +func (n Name) IsExported() bool { + return (*n.Bytes)&(1<<0) != 0 +} + +// HasTag returns true iff there is tag data following this name +func (n Name) HasTag() bool { + return (*n.Bytes)&(1<<1) != 0 +} + +// IsEmbedded returns true iff n is embedded (an anonymous field). +func (n Name) IsEmbedded() bool { + return (*n.Bytes)&(1<<3) != 0 +} + +// ReadVarint parses a varint as encoded by encoding/binary. +// It returns the number of encoded bytes and the encoded value. +func (n Name) ReadVarint(off int) (int, int) { + v := 0 + for i := 0; ; i++ { + x := *n.DataChecked(off+i, "read varint") + v += int(x&0x7f) << (7 * i) + if x&0x80 == 0 { + return i + 1, v + } + } +} + +// IsBlank indicates whether n is "_". +func (n Name) IsBlank() bool { + if n.Bytes == nil { + return false + } + _, l := n.ReadVarint(1) + return l == 1 && *n.Data(2) == '_' +} + +// writeVarint writes n to buf in varint form. Returns the +// number of bytes written. n must be nonnegative. +// Writes at most 10 bytes. +func writeVarint(buf []byte, n int) int { + for i := 0; ; i++ { + b := byte(n & 0x7f) + n >>= 7 + if n == 0 { + buf[i] = b + return i + 1 + } + buf[i] = b | 0x80 + } +} + +// Name returns the tag string for n, or empty if there is none. +func (n Name) Name() string { + if n.Bytes == nil { + return "" + } + i, l := n.ReadVarint(1) + return unsafeStringFor(n.DataChecked(1+i, "non-empty string"), l) +} + +// Tag returns the tag string for n, or empty if there is none. +func (n Name) Tag() string { + if !n.HasTag() { + return "" + } + i, l := n.ReadVarint(1) + i2, l2 := n.ReadVarint(1 + i + l) + return unsafeStringFor(n.DataChecked(1+i+l+i2, "non-empty string"), l2) +} + +func NewName(n, tag string, exported, embedded bool) Name { + if len(n) >= 1<<29 { + panic("abi.NewName: name too long: " + n[:1024] + "...") + } + if len(tag) >= 1<<29 { + panic("abi.NewName: tag too long: " + tag[:1024] + "...") + } + var nameLen [10]byte + var tagLen [10]byte + nameLenLen := writeVarint(nameLen[:], len(n)) + tagLenLen := writeVarint(tagLen[:], len(tag)) + + var bits byte + l := 1 + nameLenLen + len(n) + if exported { + bits |= 1 << 0 + } + if len(tag) > 0 { + l += tagLenLen + len(tag) + bits |= 1 << 1 + } + if embedded { + bits |= 1 << 3 + } + + b := make([]byte, l) + b[0] = bits + copy(b[1:], nameLen[:nameLenLen]) + copy(b[1+nameLenLen:], n) + if len(tag) > 0 { + tb := b[1+nameLenLen+len(n):] + copy(tb, tagLen[:tagLenLen]) + copy(tb[tagLenLen:], tag) + } + + return Name{Bytes: &b[0]} +} diff --git a/src/internal/abi/unsafestring_go119.go b/src/internal/abi/unsafestring_go119.go new file mode 100644 index 0000000..a710384 --- /dev/null +++ b/src/internal/abi/unsafestring_go119.go @@ -0,0 +1,32 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !go1.20 +// +build !go1.20 + +package abi + +import "unsafe" + +type ( + stringHeader struct { + Data *byte + Len int + } + sliceHeader struct { + Data *byte + Len int + Cap int + } +) + +func unsafeStringFor(b *byte, l int) string { + h := stringHeader{Data: b, Len: l} + return *(*string)(unsafe.Pointer(&h)) +} + +func unsafeSliceFor(b *byte, l int) []byte { + h := sliceHeader{Data: b, Len: l, Cap: l} + return *(*[]byte)(unsafe.Pointer(&h)) +} diff --git a/src/internal/abi/unsafestring_go120.go b/src/internal/abi/unsafestring_go120.go new file mode 100644 index 0000000..93ff8ea --- /dev/null +++ b/src/internal/abi/unsafestring_go120.go @@ -0,0 +1,18 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build go1.20 +// +build go1.20 + +package abi + +import "unsafe" + +func unsafeStringFor(b *byte, l int) string { + return unsafe.String(b, l) +} + +func unsafeSliceFor(b *byte, l int) []byte { + return unsafe.Slice(b, l) +} diff --git a/src/internal/bisect/bisect.go b/src/internal/bisect/bisect.go new file mode 100644 index 0000000..48c796e --- /dev/null +++ b/src/internal/bisect/bisect.go @@ -0,0 +1,795 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package bisect can be used by compilers and other programs +// to serve as a target for the bisect debugging tool. +// See [golang.org/x/tools/cmd/bisect] for details about using the tool. +// +// To be a bisect target, allowing bisect to help determine which of a set of independent +// changes provokes a failure, a program needs to: +// +// 1. Define a way to accept a change pattern on its command line or in its environment. +// The most common mechanism is a command-line flag. +// The pattern can be passed to [New] to create a [Matcher], the compiled form of a pattern. +// +// 2. Assign each change a unique ID. One possibility is to use a sequence number, +// but the most common mechanism is to hash some kind of identifying information +// like the file and line number where the change might be applied. +// [Hash] hashes its arguments to compute an ID. +// +// 3. Enable each change that the pattern says should be enabled. +// The [Matcher.ShouldEnable] method answers this question for a given change ID. +// +// 4. Print a report identifying each change that the pattern says should be printed. +// The [Matcher.ShouldPrint] method answers this question for a given change ID. +// The report consists of one more lines on standard error or standard output +// that contain a “match marker”. [Marker] returns the match marker for a given ID. +// When bisect reports a change as causing the failure, it identifies the change +// by printing the report lines with the match marker removed. +// +// # Example Usage +// +// A program starts by defining how it receives the pattern. In this example, we will assume a flag. +// The next step is to compile the pattern: +// +// m, err := bisect.New(patternFlag) +// if err != nil { +// log.Fatal(err) +// } +// +// Then, each time a potential change is considered, the program computes +// a change ID by hashing identifying information (source file and line, in this case) +// and then calls m.ShouldPrint and m.ShouldEnable to decide whether to +// print and enable the change, respectively. The two can return different values +// depending on whether bisect is trying to find a minimal set of changes to +// disable or to enable to provoke the failure. +// +// It is usually helpful to write a helper function that accepts the identifying information +// and then takes care of hashing, printing, and reporting whether the identified change +// should be enabled. For example, a helper for changes identified by a file and line number +// would be: +// +// func ShouldEnable(file string, line int) { +// h := bisect.Hash(file, line) +// if m.ShouldPrint(h) { +// fmt.Fprintf(os.Stderr, "%v %s:%d\n", bisect.Marker(h), file, line) +// } +// return m.ShouldEnable(h) +// } +// +// Finally, note that New returns a nil Matcher when there is no pattern, +// meaning that the target is not running under bisect at all, +// so all changes should be enabled and none should be printed. +// In that common case, the computation of the hash can be avoided entirely +// by checking for m == nil first: +// +// func ShouldEnable(file string, line int) bool { +// if m == nil { +// return false +// } +// h := bisect.Hash(file, line) +// if m.ShouldPrint(h) { +// fmt.Fprintf(os.Stderr, "%v %s:%d\n", bisect.Marker(h), file, line) +// } +// return m.ShouldEnable(h) +// } +// +// When the identifying information is expensive to format, this code can call +// [Matcher.MarkerOnly] to find out whether short report lines containing only the +// marker are permitted for a given run. (Bisect permits such lines when it is +// still exploring the space of possible changes and will not be showing the +// output to the user.) If so, the client can choose to print only the marker: +// +// func ShouldEnable(file string, line int) bool { +// if m == nil { +// return false +// } +// h := bisect.Hash(file, line) +// if m.ShouldPrint(h) { +// if m.MarkerOnly() { +// bisect.PrintMarker(os.Stderr) +// } else { +// fmt.Fprintf(os.Stderr, "%v %s:%d\n", bisect.Marker(h), file, line) +// } +// } +// return m.ShouldEnable(h) +// } +// +// This specific helper – deciding whether to enable a change identified by +// file and line number and printing about the change when necessary – is +// provided by the [Matcher.FileLine] method. +// +// Another common usage is deciding whether to make a change in a function +// based on the caller's stack, to identify the specific calling contexts that the +// change breaks. The [Matcher.Stack] method takes care of obtaining the stack, +// printing it when necessary, and reporting whether to enable the change +// based on that stack. +// +// # Pattern Syntax +// +// Patterns are generated by the bisect tool and interpreted by [New]. +// Users should not have to understand the patterns except when +// debugging a target's bisect support or debugging the bisect tool itself. +// +// The pattern syntax selecting a change is a sequence of bit strings +// separated by + and - operators. Each bit string denotes the set of +// changes with IDs ending in those bits, + is set addition, - is set subtraction, +// and the expression is evaluated in the usual left-to-right order. +// The special binary number “y” denotes the set of all changes, +// standing in for the empty bit string. +// In the expression, all the + operators must appear before all the - operators. +// A leading + adds to an empty set. A leading - subtracts from the set of all +// possible suffixes. +// +// For example: +// +// - “01+10” and “+01+10” both denote the set of changes +// with IDs ending with the bits 01 or 10. +// +// - “01+10-1001” denotes the set of changes with IDs +// ending with the bits 01 or 10, but excluding those ending in 1001. +// +// - “-01-1000” and “y-01-1000 both denote the set of all changes +// with IDs not ending in 01 nor 1000. +// +// - “0+1-01+001” is not a valid pattern, because all the + operators do not +// appear before all the - operators. +// +// In the syntaxes described so far, the pattern specifies the changes to +// enable and report. If a pattern is prefixed by a “!”, the meaning +// changes: the pattern specifies the changes to DISABLE and report. This +// mode of operation is needed when a program passes with all changes +// enabled but fails with no changes enabled. In this case, bisect +// searches for minimal sets of changes to disable. +// Put another way, the leading “!” inverts the result from [Matcher.ShouldEnable] +// but does not invert the result from [Matcher.ShouldPrint]. +// +// As a convenience for manual debugging, “n” is an alias for “!y”, +// meaning to disable and report all changes. +// +// Finally, a leading “v” in the pattern indicates that the reports will be shown +// to the user of bisect to describe the changes involved in a failure. +// At the API level, the leading “v” causes [Matcher.Visible] to return true. +// See the next section for details. +// +// # Match Reports +// +// The target program must enable only those changed matched +// by the pattern, and it must print a match report for each such change. +// A match report consists of one or more lines of text that will be +// printed by the bisect tool to describe a change implicated in causing +// a failure. Each line in the report for a given change must contain a +// match marker with that change ID, as returned by [Marker]. +// The markers are elided when displaying the lines to the user. +// +// A match marker has the form “[bisect-match 0x1234]” where +// 0x1234 is the change ID in hexadecimal. +// An alternate form is “[bisect-match 010101]”, giving the change ID in binary. +// +// When [Matcher.Visible] returns false, the match reports are only +// being processed by bisect to learn the set of enabled changes, +// not shown to the user, meaning that each report can be a match +// marker on a line by itself, eliding the usual textual description. +// When the textual description is expensive to compute, +// checking [Matcher.Visible] can help the avoid that expense +// in most runs. +package bisect + +import ( + "runtime" + "sync" + "sync/atomic" + "unsafe" +) + +// New creates and returns a new Matcher implementing the given pattern. +// The pattern syntax is defined in the package doc comment. +// +// In addition to the pattern syntax syntax, New("") returns nil, nil. +// The nil *Matcher is valid for use: it returns true from ShouldEnable +// and false from ShouldPrint for all changes. Callers can avoid calling +// [Hash], [Matcher.ShouldEnable], and [Matcher.ShouldPrint] entirely +// when they recognize the nil Matcher. +func New(pattern string) (*Matcher, error) { + if pattern == "" { + return nil, nil + } + + m := new(Matcher) + + p := pattern + // Special case for leading 'q' so that 'qn' quietly disables, e.g. fmahash=qn to disable fma + // Any instance of 'v' disables 'q'. + if len(p) > 0 && p[0] == 'q' { + m.quiet = true + p = p[1:] + if p == "" { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + } + // Allow multiple v, so that “bisect cmd vPATTERN” can force verbose all the time. + for len(p) > 0 && p[0] == 'v' { + m.verbose = true + m.quiet = false + p = p[1:] + if p == "" { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + } + + // Allow multiple !, each negating the last, so that “bisect cmd !PATTERN” works + // even when bisect chooses to add its own !. + m.enable = true + for len(p) > 0 && p[0] == '!' { + m.enable = !m.enable + p = p[1:] + if p == "" { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + } + + if p == "n" { + // n is an alias for !y. + m.enable = !m.enable + p = "y" + } + + // Parse actual pattern syntax. + result := true + bits := uint64(0) + start := 0 + wid := 1 // 1-bit (binary); sometimes 4-bit (hex) + for i := 0; i <= len(p); i++ { + // Imagine a trailing - at the end of the pattern to flush final suffix + c := byte('-') + if i < len(p) { + c = p[i] + } + if i == start && wid == 1 && c == 'x' { // leading x for hex + start = i + 1 + wid = 4 + continue + } + switch c { + default: + return nil, &parseError{"invalid pattern syntax: " + pattern} + case '2', '3', '4', '5', '6', '7', '8', '9': + if wid != 4 { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + fallthrough + case '0', '1': + bits <<= wid + bits |= uint64(c - '0') + case 'a', 'b', 'c', 'd', 'e', 'f', 'A', 'B', 'C', 'D', 'E', 'F': + if wid != 4 { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + bits <<= 4 + bits |= uint64(c&^0x20 - 'A' + 10) + case 'y': + if i+1 < len(p) && (p[i+1] == '0' || p[i+1] == '1') { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + bits = 0 + case '+', '-': + if c == '+' && result == false { + // Have already seen a -. Should be - from here on. + return nil, &parseError{"invalid pattern syntax (+ after -): " + pattern} + } + if i > 0 { + n := (i - start) * wid + if n > 64 { + return nil, &parseError{"pattern bits too long: " + pattern} + } + if n <= 0 { + return nil, &parseError{"invalid pattern syntax: " + pattern} + } + if p[start] == 'y' { + n = 0 + } + mask := uint64(1)<<n - 1 + m.list = append(m.list, cond{mask, bits, result}) + } else if c == '-' { + // leading - subtracts from complete set + m.list = append(m.list, cond{0, 0, true}) + } + bits = 0 + result = c == '+' + start = i + 1 + wid = 1 + } + } + return m, nil +} + +// A Matcher is the parsed, compiled form of a PATTERN string. +// The nil *Matcher is valid: it has all changes enabled but none reported. +type Matcher struct { + verbose bool // annotate reporting with human-helpful information + quiet bool // disables all reporting. reset if verbose is true. use case is -d=fmahash=qn + enable bool // when true, list is for “enable and report” (when false, “disable and report”) + list []cond // conditions; later ones win over earlier ones + dedup atomicPointerDedup +} + +// atomicPointerDedup is an atomic.Pointer[dedup], +// but we are avoiding using Go 1.19's atomic.Pointer +// until the bootstrap toolchain can be relied upon to have it. +type atomicPointerDedup struct { + p unsafe.Pointer +} + +func (p *atomicPointerDedup) Load() *dedup { + return (*dedup)(atomic.LoadPointer(&p.p)) +} + +func (p *atomicPointerDedup) CompareAndSwap(old, new *dedup) bool { + return atomic.CompareAndSwapPointer(&p.p, unsafe.Pointer(old), unsafe.Pointer(new)) +} + +// A cond is a single condition in the matcher. +// Given an input id, if id&mask == bits, return the result. +type cond struct { + mask uint64 + bits uint64 + result bool +} + +// MarkerOnly reports whether it is okay to print only the marker for +// a given change, omitting the identifying information. +// MarkerOnly returns true when bisect is using the printed reports +// only for an intermediate search step, not for showing to users. +func (m *Matcher) MarkerOnly() bool { + return !m.verbose +} + +// ShouldEnable reports whether the change with the given id should be enabled. +func (m *Matcher) ShouldEnable(id uint64) bool { + if m == nil { + return true + } + return m.matchResult(id) == m.enable +} + +// ShouldPrint reports whether to print identifying information about the change with the given id. +func (m *Matcher) ShouldPrint(id uint64) bool { + if m == nil || m.quiet { + return false + } + return m.matchResult(id) +} + +// matchResult returns the result from the first condition that matches id. +func (m *Matcher) matchResult(id uint64) bool { + for i := len(m.list) - 1; i >= 0; i-- { + c := &m.list[i] + if id&c.mask == c.bits { + return c.result + } + } + return false +} + +// FileLine reports whether the change identified by file and line should be enabled. +// If the change should be printed, FileLine prints a one-line report to w. +func (m *Matcher) FileLine(w Writer, file string, line int) bool { + if m == nil { + return true + } + return m.fileLine(w, file, line) +} + +// fileLine does the real work for FileLine. +// This lets FileLine's body handle m == nil and potentially be inlined. +func (m *Matcher) fileLine(w Writer, file string, line int) bool { + h := Hash(file, line) + if m.ShouldPrint(h) { + if m.MarkerOnly() { + PrintMarker(w, h) + } else { + printFileLine(w, h, file, line) + } + } + return m.ShouldEnable(h) +} + +// printFileLine prints a non-marker-only report for file:line to w. +func printFileLine(w Writer, h uint64, file string, line int) error { + const markerLen = 40 // overestimate + b := make([]byte, 0, markerLen+len(file)+24) + b = AppendMarker(b, h) + b = appendFileLine(b, file, line) + b = append(b, '\n') + _, err := w.Write(b) + return err +} + +// appendFileLine appends file:line to dst, returning the extended slice. +func appendFileLine(dst []byte, file string, line int) []byte { + dst = append(dst, file...) + dst = append(dst, ':') + u := uint(line) + if line < 0 { + dst = append(dst, '-') + u = -u + } + var buf [24]byte + i := len(buf) + for i == len(buf) || u > 0 { + i-- + buf[i] = '0' + byte(u%10) + u /= 10 + } + dst = append(dst, buf[i:]...) + return dst +} + +// MatchStack assigns the current call stack a change ID. +// If the stack should be printed, MatchStack prints it. +// Then MatchStack reports whether a change at the current call stack should be enabled. +func (m *Matcher) Stack(w Writer) bool { + if m == nil { + return true + } + return m.stack(w) +} + +// stack does the real work for Stack. +// This lets stack's body handle m == nil and potentially be inlined. +func (m *Matcher) stack(w Writer) bool { + const maxStack = 16 + var stk [maxStack]uintptr + n := runtime.Callers(2, stk[:]) + // caller #2 is not for printing; need it to normalize PCs if ASLR. + if n <= 1 { + return false + } + + base := stk[0] + // normalize PCs + for i := range stk[:n] { + stk[i] -= base + } + + h := Hash(stk[:n]) + if m.ShouldPrint(h) { + var d *dedup + for { + d = m.dedup.Load() + if d != nil { + break + } + d = new(dedup) + if m.dedup.CompareAndSwap(nil, d) { + break + } + } + + if m.MarkerOnly() { + if !d.seenLossy(h) { + PrintMarker(w, h) + } + } else { + if !d.seen(h) { + // Restore PCs in stack for printing + for i := range stk[:n] { + stk[i] += base + } + printStack(w, h, stk[1:n]) + } + } + } + return m.ShouldEnable(h) + +} + +// Writer is the same interface as io.Writer. +// It is duplicated here to avoid importing io. +type Writer interface { + Write([]byte) (int, error) +} + +// PrintMarker prints to w a one-line report containing only the marker for h. +// It is appropriate to use when [Matcher.ShouldPrint] and [Matcher.MarkerOnly] both return true. +func PrintMarker(w Writer, h uint64) error { + var buf [50]byte + b := AppendMarker(buf[:], h) + b = append(b, '\n') + _, err := w.Write(b) + return err +} + +// printStack prints to w a multi-line report containing a formatting of the call stack stk, +// with each line preceded by the marker for h. +func printStack(w Writer, h uint64, stk []uintptr) error { + buf := make([]byte, 0, 2048) + + var prefixBuf [100]byte + prefix := AppendMarker(prefixBuf[:0], h) + + frames := runtime.CallersFrames(stk) + for { + f, more := frames.Next() + buf = append(buf, prefix...) + buf = append(buf, f.Func.Name()...) + buf = append(buf, "()\n"...) + buf = append(buf, prefix...) + buf = append(buf, '\t') + buf = appendFileLine(buf, f.File, f.Line) + buf = append(buf, '\n') + if !more { + break + } + } + buf = append(buf, prefix...) + buf = append(buf, '\n') + _, err := w.Write(buf) + return err +} + +// Marker returns the match marker text to use on any line reporting details +// about a match of the given ID. +// It always returns the hexadecimal format. +func Marker(id uint64) string { + return string(AppendMarker(nil, id)) +} + +// AppendMarker is like [Marker] but appends the marker to dst. +func AppendMarker(dst []byte, id uint64) []byte { + const prefix = "[bisect-match 0x" + var buf [len(prefix) + 16 + 1]byte + copy(buf[:], prefix) + for i := 0; i < 16; i++ { + buf[len(prefix)+i] = "0123456789abcdef"[id>>60] + id <<= 4 + } + buf[len(prefix)+16] = ']' + return append(dst, buf[:]...) +} + +// CutMarker finds the first match marker in line and removes it, +// returning the shortened line (with the marker removed), +// the ID from the match marker, +// and whether a marker was found at all. +// If there is no marker, CutMarker returns line, 0, false. +func CutMarker(line string) (short string, id uint64, ok bool) { + // Find first instance of prefix. + prefix := "[bisect-match " + i := 0 + for ; ; i++ { + if i >= len(line)-len(prefix) { + return line, 0, false + } + if line[i] == '[' && line[i:i+len(prefix)] == prefix { + break + } + } + + // Scan to ]. + j := i + len(prefix) + for j < len(line) && line[j] != ']' { + j++ + } + if j >= len(line) { + return line, 0, false + } + + // Parse id. + idstr := line[i+len(prefix) : j] + if len(idstr) >= 3 && idstr[:2] == "0x" { + // parse hex + if len(idstr) > 2+16 { // max 0x + 16 digits + return line, 0, false + } + for i := 2; i < len(idstr); i++ { + id <<= 4 + switch c := idstr[i]; { + case '0' <= c && c <= '9': + id |= uint64(c - '0') + case 'a' <= c && c <= 'f': + id |= uint64(c - 'a' + 10) + case 'A' <= c && c <= 'F': + id |= uint64(c - 'A' + 10) + } + } + } else { + if idstr == "" || len(idstr) > 64 { // min 1 digit, max 64 digits + return line, 0, false + } + // parse binary + for i := 0; i < len(idstr); i++ { + id <<= 1 + switch c := idstr[i]; c { + default: + return line, 0, false + case '0', '1': + id |= uint64(c - '0') + } + } + } + + // Construct shortened line. + // Remove at most one space from around the marker, + // so that "foo [marker] bar" shortens to "foo bar". + j++ // skip ] + if i > 0 && line[i-1] == ' ' { + i-- + } else if j < len(line) && line[j] == ' ' { + j++ + } + short = line[:i] + line[j:] + return short, id, true +} + +// Hash computes a hash of the data arguments, +// each of which must be of type string, byte, int, uint, int32, uint32, int64, uint64, uintptr, or a slice of one of those types. +func Hash(data ...any) uint64 { + h := offset64 + for _, v := range data { + switch v := v.(type) { + default: + // Note: Not printing the type, because reflect.ValueOf(v) + // would make the interfaces prepared by the caller escape + // and therefore allocate. This way, Hash(file, line) runs + // without any allocation. It should be clear from the + // source code calling Hash what the bad argument was. + panic("bisect.Hash: unexpected argument type") + case string: + h = fnvString(h, v) + case byte: + h = fnv(h, v) + case int: + h = fnvUint64(h, uint64(v)) + case uint: + h = fnvUint64(h, uint64(v)) + case int32: + h = fnvUint32(h, uint32(v)) + case uint32: + h = fnvUint32(h, v) + case int64: + h = fnvUint64(h, uint64(v)) + case uint64: + h = fnvUint64(h, v) + case uintptr: + h = fnvUint64(h, uint64(v)) + case []string: + for _, x := range v { + h = fnvString(h, x) + } + case []byte: + for _, x := range v { + h = fnv(h, x) + } + case []int: + for _, x := range v { + h = fnvUint64(h, uint64(x)) + } + case []uint: + for _, x := range v { + h = fnvUint64(h, uint64(x)) + } + case []int32: + for _, x := range v { + h = fnvUint32(h, uint32(x)) + } + case []uint32: + for _, x := range v { + h = fnvUint32(h, x) + } + case []int64: + for _, x := range v { + h = fnvUint64(h, uint64(x)) + } + case []uint64: + for _, x := range v { + h = fnvUint64(h, x) + } + case []uintptr: + for _, x := range v { + h = fnvUint64(h, uint64(x)) + } + } + } + return h +} + +// Trivial error implementation, here to avoid importing errors. + +// parseError is a trivial error implementation, +// defined here to avoid importing errors. +type parseError struct{ text string } + +func (e *parseError) Error() string { return e.text } + +// FNV-1a implementation. See Go's hash/fnv/fnv.go. +// Copied here for simplicity (can handle integers more directly) +// and to avoid importing hash/fnv. + +const ( + offset64 uint64 = 14695981039346656037 + prime64 uint64 = 1099511628211 +) + +func fnv(h uint64, x byte) uint64 { + h ^= uint64(x) + h *= prime64 + return h +} + +func fnvString(h uint64, x string) uint64 { + for i := 0; i < len(x); i++ { + h ^= uint64(x[i]) + h *= prime64 + } + return h +} + +func fnvUint64(h uint64, x uint64) uint64 { + for i := 0; i < 8; i++ { + h ^= uint64(x & 0xFF) + x >>= 8 + h *= prime64 + } + return h +} + +func fnvUint32(h uint64, x uint32) uint64 { + for i := 0; i < 4; i++ { + h ^= uint64(x & 0xFF) + x >>= 8 + h *= prime64 + } + return h +} + +// A dedup is a deduplicator for call stacks, so that we only print +// a report for new call stacks, not for call stacks we've already +// reported. +// +// It has two modes: an approximate but lock-free mode that +// may still emit some duplicates, and a precise mode that uses +// a lock and never emits duplicates. +type dedup struct { + // 128-entry 4-way, lossy cache for seenLossy + recent [128][4]uint64 + + // complete history for seen + mu sync.Mutex + m map[uint64]bool +} + +// seen records that h has now been seen and reports whether it was seen before. +// When seen returns false, the caller is expected to print a report for h. +func (d *dedup) seen(h uint64) bool { + d.mu.Lock() + if d.m == nil { + d.m = make(map[uint64]bool) + } + seen := d.m[h] + d.m[h] = true + d.mu.Unlock() + return seen +} + +// seenLossy is a variant of seen that avoids a lock by using a cache of recently seen hashes. +// Each cache entry is N-way set-associative: h can appear in any of the slots. +// If h does not appear in any of them, then it is inserted into a random slot, +// overwriting whatever was there before. +func (d *dedup) seenLossy(h uint64) bool { + cache := &d.recent[uint(h)%uint(len(d.recent))] + for i := 0; i < len(cache); i++ { + if atomic.LoadUint64(&cache[i]) == h { + return true + } + } + + // Compute index in set to evict as hash of current set. + ch := offset64 + for _, x := range cache { + ch = fnvUint64(ch, x) + } + atomic.StoreUint64(&cache[uint(ch)%uint(len(cache))], h) + return false +} diff --git a/src/internal/buildcfg/cfg.go b/src/internal/buildcfg/cfg.go new file mode 100644 index 0000000..b97b9c1 --- /dev/null +++ b/src/internal/buildcfg/cfg.go @@ -0,0 +1,235 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package buildcfg provides access to the build configuration +// described by the current environment. It is for use by build tools +// such as cmd/go or cmd/compile and for setting up go/build's Default context. +// +// Note that it does NOT provide access to the build configuration used to +// build the currently-running binary. For that, use runtime.GOOS etc +// as well as internal/goexperiment. +package buildcfg + +import ( + "fmt" + "os" + "path/filepath" + "runtime" + "strconv" + "strings" +) + +var ( + GOROOT = runtime.GOROOT() // cached for efficiency + GOARCH = envOr("GOARCH", defaultGOARCH) + GOOS = envOr("GOOS", defaultGOOS) + GO386 = envOr("GO386", defaultGO386) + GOAMD64 = goamd64() + GOARM = goarm() + GOMIPS = gomips() + GOMIPS64 = gomips64() + GOPPC64 = goppc64() + GOWASM = gowasm() + ToolTags = toolTags() + GO_LDSO = defaultGO_LDSO + Version = version +) + +// Error is one of the errors found (if any) in the build configuration. +var Error error + +// Check exits the program with a fatal error if Error is non-nil. +func Check() { + if Error != nil { + fmt.Fprintf(os.Stderr, "%s: %v\n", filepath.Base(os.Args[0]), Error) + os.Exit(2) + } +} + +func envOr(key, value string) string { + if x := os.Getenv(key); x != "" { + return x + } + return value +} + +func goamd64() int { + switch v := envOr("GOAMD64", defaultGOAMD64); v { + case "v1": + return 1 + case "v2": + return 2 + case "v3": + return 3 + case "v4": + return 4 + } + Error = fmt.Errorf("invalid GOAMD64: must be v1, v2, v3, v4") + return int(defaultGOAMD64[len("v")] - '0') +} + +func goarm() int { + def := defaultGOARM + if GOOS == "android" && GOARCH == "arm" { + // Android arm devices always support GOARM=7. + def = "7" + } + switch v := envOr("GOARM", def); v { + case "5": + return 5 + case "6": + return 6 + case "7": + return 7 + } + Error = fmt.Errorf("invalid GOARM: must be 5, 6, 7") + return int(def[0] - '0') +} + +func gomips() string { + switch v := envOr("GOMIPS", defaultGOMIPS); v { + case "hardfloat", "softfloat": + return v + } + Error = fmt.Errorf("invalid GOMIPS: must be hardfloat, softfloat") + return defaultGOMIPS +} + +func gomips64() string { + switch v := envOr("GOMIPS64", defaultGOMIPS64); v { + case "hardfloat", "softfloat": + return v + } + Error = fmt.Errorf("invalid GOMIPS64: must be hardfloat, softfloat") + return defaultGOMIPS64 +} + +func goppc64() int { + switch v := envOr("GOPPC64", defaultGOPPC64); v { + case "power8": + return 8 + case "power9": + return 9 + case "power10": + return 10 + } + Error = fmt.Errorf("invalid GOPPC64: must be power8, power9, power10") + return int(defaultGOPPC64[len("power")] - '0') +} + +type gowasmFeatures struct { + SatConv bool + SignExt bool +} + +func (f gowasmFeatures) String() string { + var flags []string + if f.SatConv { + flags = append(flags, "satconv") + } + if f.SignExt { + flags = append(flags, "signext") + } + return strings.Join(flags, ",") +} + +func gowasm() (f gowasmFeatures) { + for _, opt := range strings.Split(envOr("GOWASM", ""), ",") { + switch opt { + case "satconv": + f.SatConv = true + case "signext": + f.SignExt = true + case "": + // ignore + default: + Error = fmt.Errorf("invalid GOWASM: no such feature %q", opt) + } + } + return +} + +func Getgoextlinkenabled() string { + return envOr("GO_EXTLINK_ENABLED", defaultGO_EXTLINK_ENABLED) +} + +func toolTags() []string { + tags := experimentTags() + tags = append(tags, gogoarchTags()...) + return tags +} + +func experimentTags() []string { + var list []string + // For each experiment that has been enabled in the toolchain, define a + // build tag with the same name but prefixed by "goexperiment." which can be + // used for compiling alternative files for the experiment. This allows + // changes for the experiment, like extra struct fields in the runtime, + // without affecting the base non-experiment code at all. + for _, exp := range Experiment.Enabled() { + list = append(list, "goexperiment."+exp) + } + return list +} + +// GOGOARCH returns the name and value of the GO$GOARCH setting. +// For example, if GOARCH is "amd64" it might return "GOAMD64", "v2". +func GOGOARCH() (name, value string) { + switch GOARCH { + case "386": + return "GO386", GO386 + case "amd64": + return "GOAMD64", fmt.Sprintf("v%d", GOAMD64) + case "arm": + return "GOARM", strconv.Itoa(GOARM) + case "mips", "mipsle": + return "GOMIPS", GOMIPS + case "mips64", "mips64le": + return "GOMIPS64", GOMIPS64 + case "ppc64", "ppc64le": + return "GOPPC64", fmt.Sprintf("power%d", GOPPC64) + case "wasm": + return "GOWASM", GOWASM.String() + } + return "", "" +} + +func gogoarchTags() []string { + switch GOARCH { + case "386": + return []string{GOARCH + "." + GO386} + case "amd64": + var list []string + for i := 1; i <= GOAMD64; i++ { + list = append(list, fmt.Sprintf("%s.v%d", GOARCH, i)) + } + return list + case "arm": + var list []string + for i := 5; i <= GOARM; i++ { + list = append(list, fmt.Sprintf("%s.%d", GOARCH, i)) + } + return list + case "mips", "mipsle": + return []string{GOARCH + "." + GOMIPS} + case "mips64", "mips64le": + return []string{GOARCH + "." + GOMIPS64} + case "ppc64", "ppc64le": + var list []string + for i := 8; i <= GOPPC64; i++ { + list = append(list, fmt.Sprintf("%s.power%d", GOARCH, i)) + } + return list + case "wasm": + var list []string + if GOWASM.SatConv { + list = append(list, GOARCH+".satconv") + } + if GOWASM.SignExt { + list = append(list, GOARCH+".signext") + } + return list + } + return nil +} diff --git a/src/internal/buildcfg/cfg_test.go b/src/internal/buildcfg/cfg_test.go new file mode 100644 index 0000000..0123593 --- /dev/null +++ b/src/internal/buildcfg/cfg_test.go @@ -0,0 +1,26 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package buildcfg + +import ( + "os" + "testing" +) + +func TestConfigFlags(t *testing.T) { + os.Setenv("GOAMD64", "v1") + if goamd64() != 1 { + t.Errorf("Wrong parsing of GOAMD64=v1") + } + os.Setenv("GOAMD64", "v4") + if goamd64() != 4 { + t.Errorf("Wrong parsing of GOAMD64=v4") + } + Error = nil + os.Setenv("GOAMD64", "1") + if goamd64(); Error == nil { + t.Errorf("Wrong parsing of GOAMD64=1") + } +} diff --git a/src/internal/buildcfg/exp.go b/src/internal/buildcfg/exp.go new file mode 100644 index 0000000..513070c --- /dev/null +++ b/src/internal/buildcfg/exp.go @@ -0,0 +1,190 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package buildcfg + +import ( + "fmt" + "reflect" + "strings" + + "internal/goexperiment" +) + +// ExperimentFlags represents a set of GOEXPERIMENT flags relative to a baseline +// (platform-default) experiment configuration. +type ExperimentFlags struct { + goexperiment.Flags + baseline goexperiment.Flags +} + +// Experiment contains the toolchain experiments enabled for the +// current build. +// +// (This is not necessarily the set of experiments the compiler itself +// was built with.) +// +// experimentBaseline specifies the experiment flags that are enabled by +// default in the current toolchain. This is, in effect, the "control" +// configuration and any variation from this is an experiment. +var Experiment ExperimentFlags = func() ExperimentFlags { + flags, err := ParseGOEXPERIMENT(GOOS, GOARCH, envOr("GOEXPERIMENT", defaultGOEXPERIMENT)) + if err != nil { + Error = err + return ExperimentFlags{} + } + return *flags +}() + +// DefaultGOEXPERIMENT is the embedded default GOEXPERIMENT string. +// It is not guaranteed to be canonical. +const DefaultGOEXPERIMENT = defaultGOEXPERIMENT + +// FramePointerEnabled enables the use of platform conventions for +// saving frame pointers. +// +// This used to be an experiment, but now it's always enabled on +// platforms that support it. +// +// Note: must agree with runtime.framepointer_enabled. +var FramePointerEnabled = GOARCH == "amd64" || GOARCH == "arm64" + +// ParseGOEXPERIMENT parses a (GOOS, GOARCH, GOEXPERIMENT) +// configuration tuple and returns the enabled and baseline experiment +// flag sets. +// +// TODO(mdempsky): Move to internal/goexperiment. +func ParseGOEXPERIMENT(goos, goarch, goexp string) (*ExperimentFlags, error) { + // regabiSupported is set to true on platforms where register ABI is + // supported and enabled by default. + // regabiAlwaysOn is set to true on platforms where register ABI is + // always on. + var regabiSupported, regabiAlwaysOn bool + switch goarch { + case "amd64", "arm64", "ppc64le", "ppc64", "riscv64": + regabiAlwaysOn = true + regabiSupported = true + } + + baseline := goexperiment.Flags{ + RegabiWrappers: regabiSupported, + RegabiArgs: regabiSupported, + CoverageRedesign: true, + } + + // Start with the statically enabled set of experiments. + flags := &ExperimentFlags{ + Flags: baseline, + baseline: baseline, + } + + // Pick up any changes to the baseline configuration from the + // GOEXPERIMENT environment. This can be set at make.bash time + // and overridden at build time. + if goexp != "" { + // Create a map of known experiment names. + names := make(map[string]func(bool)) + rv := reflect.ValueOf(&flags.Flags).Elem() + rt := rv.Type() + for i := 0; i < rt.NumField(); i++ { + field := rv.Field(i) + names[strings.ToLower(rt.Field(i).Name)] = field.SetBool + } + + // "regabi" is an alias for all working regabi + // subexperiments, and not an experiment itself. Doing + // this as an alias make both "regabi" and "noregabi" + // do the right thing. + names["regabi"] = func(v bool) { + flags.RegabiWrappers = v + flags.RegabiArgs = v + } + + // Parse names. + for _, f := range strings.Split(goexp, ",") { + if f == "" { + continue + } + if f == "none" { + // GOEXPERIMENT=none disables all experiment flags. + // This is used by cmd/dist, which doesn't know how + // to build with any experiment flags. + flags.Flags = goexperiment.Flags{} + continue + } + val := true + if strings.HasPrefix(f, "no") { + f, val = f[2:], false + } + set, ok := names[f] + if !ok { + return nil, fmt.Errorf("unknown GOEXPERIMENT %s", f) + } + set(val) + } + } + + if regabiAlwaysOn { + flags.RegabiWrappers = true + flags.RegabiArgs = true + } + // regabi is only supported on amd64, arm64, riscv64, ppc64 and ppc64le. + if !regabiSupported { + flags.RegabiWrappers = false + flags.RegabiArgs = false + } + // Check regabi dependencies. + if flags.RegabiArgs && !flags.RegabiWrappers { + return nil, fmt.Errorf("GOEXPERIMENT regabiargs requires regabiwrappers") + } + return flags, nil +} + +// String returns the canonical GOEXPERIMENT string to enable this experiment +// configuration. (Experiments in the same state as in the baseline are elided.) +func (exp *ExperimentFlags) String() string { + return strings.Join(expList(&exp.Flags, &exp.baseline, false), ",") +} + +// expList returns the list of lower-cased experiment names for +// experiments that differ from base. base may be nil to indicate no +// experiments. If all is true, then include all experiment flags, +// regardless of base. +func expList(exp, base *goexperiment.Flags, all bool) []string { + var list []string + rv := reflect.ValueOf(exp).Elem() + var rBase reflect.Value + if base != nil { + rBase = reflect.ValueOf(base).Elem() + } + rt := rv.Type() + for i := 0; i < rt.NumField(); i++ { + name := strings.ToLower(rt.Field(i).Name) + val := rv.Field(i).Bool() + baseVal := false + if base != nil { + baseVal = rBase.Field(i).Bool() + } + if all || val != baseVal { + if val { + list = append(list, name) + } else { + list = append(list, "no"+name) + } + } + } + return list +} + +// Enabled returns a list of enabled experiments, as +// lower-cased experiment names. +func (exp *ExperimentFlags) Enabled() []string { + return expList(&exp.Flags, nil, false) +} + +// All returns a list of all experiment settings. +// Disabled experiments appear in the list prefixed by "no". +func (exp *ExperimentFlags) All() []string { + return expList(&exp.Flags, nil, true) +} diff --git a/src/internal/bytealg/bytealg.go b/src/internal/bytealg/bytealg.go new file mode 100644 index 0000000..28f2742 --- /dev/null +++ b/src/internal/bytealg/bytealg.go @@ -0,0 +1,155 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +import ( + "internal/cpu" + "unsafe" +) + +// Offsets into internal/cpu records for use in assembly. +const ( + offsetX86HasSSE42 = unsafe.Offsetof(cpu.X86.HasSSE42) + offsetX86HasAVX2 = unsafe.Offsetof(cpu.X86.HasAVX2) + offsetX86HasPOPCNT = unsafe.Offsetof(cpu.X86.HasPOPCNT) + + offsetS390xHasVX = unsafe.Offsetof(cpu.S390X.HasVX) + + offsetPPC64HasPOWER9 = unsafe.Offsetof(cpu.PPC64.IsPOWER9) +) + +// MaxLen is the maximum length of the string to be searched for (argument b) in Index. +// If MaxLen is not 0, make sure MaxLen >= 4. +var MaxLen int + +// FIXME: the logic of HashStrBytes, HashStrRevBytes, IndexRabinKarpBytes and HashStr, HashStrRev, +// IndexRabinKarp are exactly the same, except that the types are different. Can we eliminate +// three of them without causing allocation? + +// PrimeRK is the prime base used in Rabin-Karp algorithm. +const PrimeRK = 16777619 + +// HashStrBytes returns the hash and the appropriate multiplicative +// factor for use in Rabin-Karp algorithm. +func HashStrBytes(sep []byte) (uint32, uint32) { + hash := uint32(0) + for i := 0; i < len(sep); i++ { + hash = hash*PrimeRK + uint32(sep[i]) + } + var pow, sq uint32 = 1, PrimeRK + for i := len(sep); i > 0; i >>= 1 { + if i&1 != 0 { + pow *= sq + } + sq *= sq + } + return hash, pow +} + +// HashStr returns the hash and the appropriate multiplicative +// factor for use in Rabin-Karp algorithm. +func HashStr(sep string) (uint32, uint32) { + hash := uint32(0) + for i := 0; i < len(sep); i++ { + hash = hash*PrimeRK + uint32(sep[i]) + } + var pow, sq uint32 = 1, PrimeRK + for i := len(sep); i > 0; i >>= 1 { + if i&1 != 0 { + pow *= sq + } + sq *= sq + } + return hash, pow +} + +// HashStrRevBytes returns the hash of the reverse of sep and the +// appropriate multiplicative factor for use in Rabin-Karp algorithm. +func HashStrRevBytes(sep []byte) (uint32, uint32) { + hash := uint32(0) + for i := len(sep) - 1; i >= 0; i-- { + hash = hash*PrimeRK + uint32(sep[i]) + } + var pow, sq uint32 = 1, PrimeRK + for i := len(sep); i > 0; i >>= 1 { + if i&1 != 0 { + pow *= sq + } + sq *= sq + } + return hash, pow +} + +// HashStrRev returns the hash of the reverse of sep and the +// appropriate multiplicative factor for use in Rabin-Karp algorithm. +func HashStrRev(sep string) (uint32, uint32) { + hash := uint32(0) + for i := len(sep) - 1; i >= 0; i-- { + hash = hash*PrimeRK + uint32(sep[i]) + } + var pow, sq uint32 = 1, PrimeRK + for i := len(sep); i > 0; i >>= 1 { + if i&1 != 0 { + pow *= sq + } + sq *= sq + } + return hash, pow +} + +// IndexRabinKarpBytes uses the Rabin-Karp search algorithm to return the index of the +// first occurrence of substr in s, or -1 if not present. +func IndexRabinKarpBytes(s, sep []byte) int { + // Rabin-Karp search + hashsep, pow := HashStrBytes(sep) + n := len(sep) + var h uint32 + for i := 0; i < n; i++ { + h = h*PrimeRK + uint32(s[i]) + } + if h == hashsep && Equal(s[:n], sep) { + return 0 + } + for i := n; i < len(s); { + h *= PrimeRK + h += uint32(s[i]) + h -= pow * uint32(s[i-n]) + i++ + if h == hashsep && Equal(s[i-n:i], sep) { + return i - n + } + } + return -1 +} + +// IndexRabinKarp uses the Rabin-Karp search algorithm to return the index of the +// first occurrence of substr in s, or -1 if not present. +func IndexRabinKarp(s, substr string) int { + // Rabin-Karp search + hashss, pow := HashStr(substr) + n := len(substr) + var h uint32 + for i := 0; i < n; i++ { + h = h*PrimeRK + uint32(s[i]) + } + if h == hashss && s[:n] == substr { + return 0 + } + for i := n; i < len(s); { + h *= PrimeRK + h += uint32(s[i]) + h -= pow * uint32(s[i-n]) + i++ + if h == hashss && s[i-n:i] == substr { + return i - n + } + } + return -1 +} + +// MakeNoZero makes a slice of length and capacity n without zeroing the bytes. +// It is the caller's responsibility to ensure uninitialized bytes +// do not leak to the end user. +func MakeNoZero(n int) []byte diff --git a/src/internal/bytealg/compare_386.s b/src/internal/bytealg/compare_386.s new file mode 100644 index 0000000..27b660c --- /dev/null +++ b/src/internal/bytealg/compare_386.s @@ -0,0 +1,144 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT,$0-28 + MOVL a_base+0(FP), SI + MOVL a_len+4(FP), BX + MOVL b_base+12(FP), DI + MOVL b_len+16(FP), DX + LEAL ret+24(FP), AX + JMP cmpbody<>(SB) + +TEXT runtime·cmpstring(SB),NOSPLIT,$0-20 + MOVL a_base+0(FP), SI + MOVL a_len+4(FP), BX + MOVL b_base+8(FP), DI + MOVL b_len+12(FP), DX + LEAL ret+16(FP), AX + JMP cmpbody<>(SB) + +// input: +// SI = a +// DI = b +// BX = alen +// DX = blen +// AX = address of return word (set to 1/0/-1) +TEXT cmpbody<>(SB),NOSPLIT,$0-0 + MOVL DX, BP + SUBL BX, DX // DX = blen-alen + JLE 2(PC) + MOVL BX, BP // BP = min(alen, blen) + CMPL SI, DI + JEQ allsame + CMPL BP, $4 + JB small +#ifdef GO386_softfloat + JMP mediumloop +#endif +largeloop: + CMPL BP, $16 + JB mediumloop + MOVOU (SI), X0 + MOVOU (DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, BX + XORL $0xffff, BX // convert EQ to NE + JNE diff16 // branch if at least one byte is not equal + ADDL $16, SI + ADDL $16, DI + SUBL $16, BP + JMP largeloop + +diff16: + BSFL BX, BX // index of first byte that differs + XORL DX, DX + MOVB (SI)(BX*1), CX + CMPB CX, (DI)(BX*1) + SETHI DX + LEAL -1(DX*2), DX // convert 1/0 to +1/-1 + MOVL DX, (AX) + RET + +mediumloop: + CMPL BP, $4 + JBE _0through4 + MOVL (SI), BX + MOVL (DI), CX + CMPL BX, CX + JNE diff4 + ADDL $4, SI + ADDL $4, DI + SUBL $4, BP + JMP mediumloop + +_0through4: + MOVL -4(SI)(BP*1), BX + MOVL -4(DI)(BP*1), CX + CMPL BX, CX + JEQ allsame + +diff4: + BSWAPL BX // reverse order of bytes + BSWAPL CX + XORL BX, CX // find bit differences + BSRL CX, CX // index of highest bit difference + SHRL CX, BX // move a's bit to bottom + ANDL $1, BX // mask bit + LEAL -1(BX*2), BX // 1/0 => +1/-1 + MOVL BX, (AX) + RET + + // 0-3 bytes in common +small: + LEAL (BP*8), CX + NEGL CX + JEQ allsame + + // load si + CMPB SI, $0xfc + JA si_high + MOVL (SI), SI + JMP si_finish +si_high: + MOVL -4(SI)(BP*1), SI + SHRL CX, SI +si_finish: + SHLL CX, SI + + // same for di + CMPB DI, $0xfc + JA di_high + MOVL (DI), DI + JMP di_finish +di_high: + MOVL -4(DI)(BP*1), DI + SHRL CX, DI +di_finish: + SHLL CX, DI + + BSWAPL SI // reverse order of bytes + BSWAPL DI + XORL SI, DI // find bit differences + JEQ allsame + BSRL DI, CX // index of highest bit difference + SHRL CX, SI // move a's bit to bottom + ANDL $1, SI // mask bit + LEAL -1(SI*2), BX // 1/0 => +1/-1 + MOVL BX, (AX) + RET + + // all the bytes in common are the same, so we just need + // to compare the lengths. +allsame: + XORL BX, BX + XORL CX, CX + TESTL DX, DX + SETLT BX // 1 if alen > blen + SETEQ CX // 1 if alen == blen + LEAL -1(CX)(BX*2), BX // 1,0,-1 result + MOVL BX, (AX) + RET diff --git a/src/internal/bytealg/compare_amd64.s b/src/internal/bytealg/compare_amd64.s new file mode 100644 index 0000000..fdd015f --- /dev/null +++ b/src/internal/bytealg/compare_amd64.s @@ -0,0 +1,237 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "asm_amd64.h" +#include "textflag.h" + +TEXT ·Compare<ABIInternal>(SB),NOSPLIT,$0-56 + // AX = a_base (want in SI) + // BX = a_len (want in BX) + // CX = a_cap (unused) + // DI = b_base (want in DI) + // SI = b_len (want in DX) + // R8 = b_cap (unused) + MOVQ SI, DX + MOVQ AX, SI + JMP cmpbody<>(SB) + +TEXT runtime·cmpstring<ABIInternal>(SB),NOSPLIT,$0-40 + // AX = a_base (want in SI) + // BX = a_len (want in BX) + // CX = b_base (want in DI) + // DI = b_len (want in DX) + MOVQ AX, SI + MOVQ DI, DX + MOVQ CX, DI + JMP cmpbody<>(SB) + +// input: +// SI = a +// DI = b +// BX = alen +// DX = blen +// output: +// AX = output (-1/0/1) +TEXT cmpbody<>(SB),NOSPLIT,$0-0 + CMPQ SI, DI + JEQ allsame + CMPQ BX, DX + MOVQ DX, R8 + CMOVQLT BX, R8 // R8 = min(alen, blen) = # of bytes to compare + CMPQ R8, $8 + JB small + + CMPQ R8, $63 + JBE loop +#ifndef hasAVX2 + CMPB internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1 + JEQ big_loop_avx2 + JMP big_loop +#else + JMP big_loop_avx2 +#endif +loop: + CMPQ R8, $16 + JBE _0through16 + MOVOU (SI), X0 + MOVOU (DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, AX + XORQ $0xffff, AX // convert EQ to NE + JNE diff16 // branch if at least one byte is not equal + ADDQ $16, SI + ADDQ $16, DI + SUBQ $16, R8 + JMP loop + +diff64: + ADDQ $48, SI + ADDQ $48, DI + JMP diff16 +diff48: + ADDQ $32, SI + ADDQ $32, DI + JMP diff16 +diff32: + ADDQ $16, SI + ADDQ $16, DI + // AX = bit mask of differences +diff16: + BSFQ AX, BX // index of first byte that differs + XORQ AX, AX + MOVB (SI)(BX*1), CX + CMPB CX, (DI)(BX*1) + SETHI AX + LEAQ -1(AX*2), AX // convert 1/0 to +1/-1 + RET + + // 0 through 16 bytes left, alen>=8, blen>=8 +_0through16: + CMPQ R8, $8 + JBE _0through8 + MOVQ (SI), AX + MOVQ (DI), CX + CMPQ AX, CX + JNE diff8 +_0through8: + MOVQ -8(SI)(R8*1), AX + MOVQ -8(DI)(R8*1), CX + CMPQ AX, CX + JEQ allsame + + // AX and CX contain parts of a and b that differ. +diff8: + BSWAPQ AX // reverse order of bytes + BSWAPQ CX + XORQ AX, CX + BSRQ CX, CX // index of highest bit difference + SHRQ CX, AX // move a's bit to bottom + ANDQ $1, AX // mask bit + LEAQ -1(AX*2), AX // 1/0 => +1/-1 + RET + + // 0-7 bytes in common +small: + LEAQ (R8*8), CX // bytes left -> bits left + NEGQ CX // - bits lift (== 64 - bits left mod 64) + JEQ allsame + + // load bytes of a into high bytes of AX + CMPB SI, $0xf8 + JA si_high + MOVQ (SI), SI + JMP si_finish +si_high: + MOVQ -8(SI)(R8*1), SI + SHRQ CX, SI +si_finish: + SHLQ CX, SI + + // load bytes of b in to high bytes of BX + CMPB DI, $0xf8 + JA di_high + MOVQ (DI), DI + JMP di_finish +di_high: + MOVQ -8(DI)(R8*1), DI + SHRQ CX, DI +di_finish: + SHLQ CX, DI + + BSWAPQ SI // reverse order of bytes + BSWAPQ DI + XORQ SI, DI // find bit differences + JEQ allsame + BSRQ DI, CX // index of highest bit difference + SHRQ CX, SI // move a's bit to bottom + ANDQ $1, SI // mask bit + LEAQ -1(SI*2), AX // 1/0 => +1/-1 + RET + +allsame: + XORQ AX, AX + XORQ CX, CX + CMPQ BX, DX + SETGT AX // 1 if alen > blen + SETEQ CX // 1 if alen == blen + LEAQ -1(CX)(AX*2), AX // 1,0,-1 result + RET + + // this works for >= 64 bytes of data. +#ifndef hasAVX2 +big_loop: + MOVOU (SI), X0 + MOVOU (DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, AX + XORQ $0xffff, AX + JNE diff16 + + MOVOU 16(SI), X0 + MOVOU 16(DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, AX + XORQ $0xffff, AX + JNE diff32 + + MOVOU 32(SI), X0 + MOVOU 32(DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, AX + XORQ $0xffff, AX + JNE diff48 + + MOVOU 48(SI), X0 + MOVOU 48(DI), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, AX + XORQ $0xffff, AX + JNE diff64 + + ADDQ $64, SI + ADDQ $64, DI + SUBQ $64, R8 + CMPQ R8, $64 + JBE loop + JMP big_loop +#endif + + // Compare 64-bytes per loop iteration. + // Loop is unrolled and uses AVX2. +big_loop_avx2: + VMOVDQU (SI), Y2 + VMOVDQU (DI), Y3 + VMOVDQU 32(SI), Y4 + VMOVDQU 32(DI), Y5 + VPCMPEQB Y2, Y3, Y0 + VPMOVMSKB Y0, AX + XORL $0xffffffff, AX + JNE diff32_avx2 + VPCMPEQB Y4, Y5, Y6 + VPMOVMSKB Y6, AX + XORL $0xffffffff, AX + JNE diff64_avx2 + + ADDQ $64, SI + ADDQ $64, DI + SUBQ $64, R8 + CMPQ R8, $64 + JB big_loop_avx2_exit + JMP big_loop_avx2 + + // Avoid AVX->SSE transition penalty and search first 32 bytes of 64 byte chunk. +diff32_avx2: + VZEROUPPER + JMP diff16 + + // Same as diff32_avx2, but for last 32 bytes. +diff64_avx2: + VZEROUPPER + JMP diff48 + + // For <64 bytes remainder jump to normal loop. +big_loop_avx2_exit: + VZEROUPPER + JMP loop diff --git a/src/internal/bytealg/compare_arm.s b/src/internal/bytealg/compare_arm.s new file mode 100644 index 0000000..80d01a2 --- /dev/null +++ b/src/internal/bytealg/compare_arm.s @@ -0,0 +1,86 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT|NOFRAME,$0-28 + MOVW a_base+0(FP), R2 + MOVW a_len+4(FP), R0 + MOVW b_base+12(FP), R3 + MOVW b_len+16(FP), R1 + ADD $28, R13, R7 + B cmpbody<>(SB) + +TEXT runtime·cmpstring(SB),NOSPLIT|NOFRAME,$0-20 + MOVW a_base+0(FP), R2 + MOVW a_len+4(FP), R0 + MOVW b_base+8(FP), R3 + MOVW b_len+12(FP), R1 + ADD $20, R13, R7 + B cmpbody<>(SB) + +// On entry: +// R0 is the length of a +// R1 is the length of b +// R2 points to the start of a +// R3 points to the start of b +// R7 points to return value (-1/0/1 will be written here) +// +// On exit: +// R4, R5, R6 and R8 are clobbered +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMP R2, R3 + BEQ samebytes + CMP R0, R1 + MOVW R0, R6 + MOVW.LT R1, R6 // R6 is min(R0, R1) + + CMP $0, R6 + BEQ samebytes + CMP $4, R6 + ADD R2, R6 // R2 is current byte in a, R6 is the end of the range to compare + BLT byte_loop // length < 4 + AND $3, R2, R8 + CMP $0, R8 + BNE byte_loop // unaligned a, use byte-wise compare (TODO: try to align a) +aligned_a: + AND $3, R3, R8 + CMP $0, R8 + BNE byte_loop // unaligned b, use byte-wise compare + AND $0xfffffffc, R6, R8 + // length >= 4 +chunk4_loop: + MOVW.P 4(R2), R4 + MOVW.P 4(R3), R5 + CMP R4, R5 + BNE cmp + CMP R2, R8 + BNE chunk4_loop + CMP R2, R6 + BEQ samebytes // all compared bytes were the same; compare lengths +byte_loop: + MOVBU.P 1(R2), R4 + MOVBU.P 1(R3), R5 + CMP R4, R5 + BNE ret + CMP R2, R6 + BNE byte_loop +samebytes: + CMP R0, R1 + MOVW.LT $1, R0 + MOVW.GT $-1, R0 + MOVW.EQ $0, R0 + MOVW R0, (R7) + RET +ret: + // bytes differed + MOVW.LT $1, R0 + MOVW.GT $-1, R0 + MOVW R0, (R7) + RET +cmp: + SUB $4, R2, R2 + SUB $4, R3, R3 + B byte_loop diff --git a/src/internal/bytealg/compare_arm64.s b/src/internal/bytealg/compare_arm64.s new file mode 100644 index 0000000..cc02c46 --- /dev/null +++ b/src/internal/bytealg/compare_arm64.s @@ -0,0 +1,125 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-56 + // R0 = a_base (want in R0) + // R1 = a_len (want in R1) + // R2 = a_cap (unused) + // R3 = b_base (want in R2) + // R4 = b_len (want in R3) + // R5 = b_cap (unused) + MOVD R3, R2 + MOVD R4, R3 + B cmpbody<>(SB) + +TEXT runtime·cmpstring<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // R0 = a_base + // R1 = a_len + // R2 = b_base + // R3 = b_len + B cmpbody<>(SB) + +// On entry: +// R0 points to the start of a +// R1 is the length of a +// R2 points to the start of b +// R3 is the length of b +// +// On exit: +// R0 is the result +// R4, R5, R6, R8, R9 and R10 are clobbered +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMP R0, R2 + BEQ samebytes // same starting pointers; compare lengths + CMP R1, R3 + CSEL LT, R3, R1, R6 // R6 is min(R1, R3) + + CBZ R6, samebytes + BIC $0xf, R6, R10 + CBZ R10, small // length < 16 + ADD R0, R10 // end of chunk16 + // length >= 16 +chunk16_loop: + LDP.P 16(R0), (R4, R8) + LDP.P 16(R2), (R5, R9) + CMP R4, R5 + BNE cmp + CMP R8, R9 + BNE cmpnext + CMP R10, R0 + BNE chunk16_loop + AND $0xf, R6, R6 + CBZ R6, samebytes + SUBS $8, R6 + BLT tail + // the length of tail > 8 bytes + MOVD.P 8(R0), R4 + MOVD.P 8(R2), R5 + CMP R4, R5 + BNE cmp + SUB $8, R6 + // compare last 8 bytes +tail: + MOVD (R0)(R6), R4 + MOVD (R2)(R6), R5 + CMP R4, R5 + BEQ samebytes +cmp: + REV R4, R4 + REV R5, R5 + CMP R4, R5 +ret: + MOVD $1, R0 + CNEG HI, R0, R0 + RET +small: + TBZ $3, R6, lt_8 + MOVD (R0), R4 + MOVD (R2), R5 + CMP R4, R5 + BNE cmp + SUBS $8, R6 + BEQ samebytes + ADD $8, R0 + ADD $8, R2 + SUB $8, R6 + B tail +lt_8: + TBZ $2, R6, lt_4 + MOVWU (R0), R4 + MOVWU (R2), R5 + CMPW R4, R5 + BNE cmp + SUBS $4, R6 + BEQ samebytes + ADD $4, R0 + ADD $4, R2 +lt_4: + TBZ $1, R6, lt_2 + MOVHU (R0), R4 + MOVHU (R2), R5 + CMPW R4, R5 + BNE cmp + ADD $2, R0 + ADD $2, R2 +lt_2: + TBZ $0, R6, samebytes +one: + MOVBU (R0), R4 + MOVBU (R2), R5 + CMPW R4, R5 + BNE ret +samebytes: + CMP R3, R1 + CSET NE, R0 + CNEG LO, R0, R0 + RET +cmpnext: + REV R8, R4 + REV R9, R5 + CMP R4, R5 + B ret diff --git a/src/internal/bytealg/compare_generic.go b/src/internal/bytealg/compare_generic.go new file mode 100644 index 0000000..b04e275 --- /dev/null +++ b/src/internal/bytealg/compare_generic.go @@ -0,0 +1,60 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !386 && !amd64 && !s390x && !arm && !arm64 && !loong64 && !ppc64 && !ppc64le && !mips && !mipsle && !wasm && !mips64 && !mips64le && !riscv64 + +package bytealg + +import _ "unsafe" // for go:linkname + +func Compare(a, b []byte) int { + l := len(a) + if len(b) < l { + l = len(b) + } + if l == 0 || &a[0] == &b[0] { + goto samebytes + } + for i := 0; i < l; i++ { + c1, c2 := a[i], b[i] + if c1 < c2 { + return -1 + } + if c1 > c2 { + return +1 + } + } +samebytes: + if len(a) < len(b) { + return -1 + } + if len(a) > len(b) { + return +1 + } + return 0 +} + +//go:linkname runtime_cmpstring runtime.cmpstring +func runtime_cmpstring(a, b string) int { + l := len(a) + if len(b) < l { + l = len(b) + } + for i := 0; i < l; i++ { + c1, c2 := a[i], b[i] + if c1 < c2 { + return -1 + } + if c1 > c2 { + return +1 + } + } + if len(a) < len(b) { + return -1 + } + if len(a) > len(b) { + return +1 + } + return 0 +} diff --git a/src/internal/bytealg/compare_loong64.s b/src/internal/bytealg/compare_loong64.s new file mode 100644 index 0000000..c89c5a9 --- /dev/null +++ b/src/internal/bytealg/compare_loong64.s @@ -0,0 +1,87 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT,$0-56 + MOVV a_base+0(FP), R6 + MOVV b_base+24(FP), R7 + MOVV a_len+8(FP), R4 + MOVV b_len+32(FP), R5 + MOVV $ret+48(FP), R13 + JMP cmpbody<>(SB) + +TEXT runtime·cmpstring(SB),NOSPLIT,$0-40 + MOVV a_base+0(FP), R6 + MOVV b_base+16(FP), R7 + MOVV a_len+8(FP), R4 + MOVV b_len+24(FP), R5 + MOVV $ret+32(FP), R13 + JMP cmpbody<>(SB) + +// On entry: +// R4 length of a +// R5 length of b +// R6 points to the start of a +// R7 points to the start of b +// R13 points to the return value (-1/0/1) +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0 + BEQ R6, R7, samebytes // same start of a and b + + SGTU R4, R5, R9 + BNE R0, R9, r2_lt_r1 + MOVV R4, R14 + JMP entry +r2_lt_r1: + MOVV R5, R14 // R14 is min(R4, R5) +entry: + ADDV R6, R14, R12 // R6 start of a, R14 end of a + BEQ R6, R12, samebytes // length is 0 + + SRLV $4, R14 // R14 is number of chunks + BEQ R0, R14, byte_loop + + // make sure both a and b are aligned. + OR R6, R7, R15 + AND $7, R15 + BNE R0, R15, byte_loop + + PCALIGN $16 +chunk16_loop: + BEQ R0, R14, byte_loop + MOVV (R6), R8 + MOVV (R7), R9 + BNE R8, R9, byte_loop + MOVV 8(R6), R16 + MOVV 8(R7), R17 + ADDV $16, R6 + ADDV $16, R7 + SUBVU $1, R14 + BEQ R16, R17, chunk16_loop + SUBV $8, R6 + SUBV $8, R7 + +byte_loop: + BEQ R6, R12, samebytes + MOVBU (R6), R8 + ADDVU $1, R6 + MOVBU (R7), R9 + ADDVU $1, R7 + BEQ R8, R9, byte_loop + +byte_cmp: + SGTU R8, R9, R12 // R12 = 1 if (R8 > R9) + BNE R0, R12, ret + MOVV $-1, R12 + JMP ret + +samebytes: + SGTU R4, R5, R8 + SGTU R5, R4, R9 + SUBV R9, R8, R12 + +ret: + MOVV R12, (R13) + RET diff --git a/src/internal/bytealg/compare_mips64x.s b/src/internal/bytealg/compare_mips64x.s new file mode 100644 index 0000000..117a9ef --- /dev/null +++ b/src/internal/bytealg/compare_mips64x.s @@ -0,0 +1,88 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips64 || mips64le + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT,$0-56 + MOVV a_base+0(FP), R3 + MOVV b_base+24(FP), R4 + MOVV a_len+8(FP), R1 + MOVV b_len+32(FP), R2 + MOVV $ret+48(FP), R9 + JMP cmpbody<>(SB) + +TEXT runtime·cmpstring(SB),NOSPLIT,$0-40 + MOVV a_base+0(FP), R3 + MOVV b_base+16(FP), R4 + MOVV a_len+8(FP), R1 + MOVV b_len+24(FP), R2 + MOVV $ret+32(FP), R9 + JMP cmpbody<>(SB) + +// On entry: +// R1 length of a +// R2 length of b +// R3 points to the start of a +// R4 points to the start of b +// R9 points to the return value (-1/0/1) +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0 + BEQ R3, R4, samebytes // same start of a and b + + SGTU R1, R2, R7 + BNE R0, R7, r2_lt_r1 + MOVV R1, R10 + JMP entry +r2_lt_r1: + MOVV R2, R10 // R10 is min(R1, R2) +entry: + ADDV R3, R10, R8 // R3 start of a, R8 end of a + BEQ R3, R8, samebytes // length is 0 + + SRLV $4, R10 // R10 is number of chunks + BEQ R0, R10, byte_loop + + // make sure both a and b are aligned. + OR R3, R4, R11 + AND $7, R11 + BNE R0, R11, byte_loop + +chunk16_loop: + BEQ R0, R10, byte_loop + MOVV (R3), R6 + MOVV (R4), R7 + BNE R6, R7, byte_loop + MOVV 8(R3), R13 + MOVV 8(R4), R14 + ADDV $16, R3 + ADDV $16, R4 + SUBVU $1, R10 + BEQ R13, R14, chunk16_loop + SUBV $8, R3 + SUBV $8, R4 + +byte_loop: + BEQ R3, R8, samebytes + MOVBU (R3), R6 + ADDVU $1, R3 + MOVBU (R4), R7 + ADDVU $1, R4 + BEQ R6, R7, byte_loop + +byte_cmp: + SGTU R6, R7, R8 // R8 = 1 if (R6 > R7) + BNE R0, R8, ret + MOVV $-1, R8 + JMP ret + +samebytes: + SGTU R1, R2, R6 + SGTU R2, R1, R7 + SUBV R7, R6, R8 + +ret: + MOVV R8, (R9) + RET diff --git a/src/internal/bytealg/compare_mipsx.s b/src/internal/bytealg/compare_mipsx.s new file mode 100644 index 0000000..857ac13 --- /dev/null +++ b/src/internal/bytealg/compare_mipsx.s @@ -0,0 +1,72 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips || mipsle + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT,$0-28 + MOVW a_base+0(FP), R3 + MOVW b_base+12(FP), R4 + MOVW a_len+4(FP), R1 + MOVW b_len+16(FP), R2 + BEQ R3, R4, samebytes + SGTU R1, R2, R7 + MOVW R1, R8 + CMOVN R7, R2, R8 // R8 is min(R1, R2) + + ADDU R3, R8 // R3 is current byte in a, R8 is last byte in a to compare +loop: + BEQ R3, R8, samebytes + + MOVBU (R3), R6 + ADDU $1, R3 + MOVBU (R4), R7 + ADDU $1, R4 + BEQ R6, R7 , loop + + SGTU R6, R7, R8 + MOVW $-1, R6 + CMOVZ R8, R6, R8 + JMP cmp_ret +samebytes: + SGTU R1, R2, R6 + SGTU R2, R1, R7 + SUBU R7, R6, R8 +cmp_ret: + MOVW R8, ret+24(FP) + RET + +TEXT runtime·cmpstring(SB),NOSPLIT,$0-20 + MOVW a_base+0(FP), R3 + MOVW a_len+4(FP), R1 + MOVW b_base+8(FP), R4 + MOVW b_len+12(FP), R2 + BEQ R3, R4, samebytes + SGTU R1, R2, R7 + MOVW R1, R8 + CMOVN R7, R2, R8 // R8 is min(R1, R2) + + ADDU R3, R8 // R3 is current byte in a, R8 is last byte in a to compare +loop: + BEQ R3, R8, samebytes // all compared bytes were the same; compare lengths + + MOVBU (R3), R6 + ADDU $1, R3 + MOVBU (R4), R7 + ADDU $1, R4 + BEQ R6, R7 , loop + // bytes differed + SGTU R6, R7, R8 + MOVW $-1, R6 + CMOVZ R8, R6, R8 + JMP cmp_ret +samebytes: + SGTU R1, R2, R6 + SGTU R2, R1, R7 + SUBU R7, R6, R8 +cmp_ret: + MOVW R8, ret+16(FP) + RET diff --git a/src/internal/bytealg/compare_native.go b/src/internal/bytealg/compare_native.go new file mode 100644 index 0000000..34964e2 --- /dev/null +++ b/src/internal/bytealg/compare_native.go @@ -0,0 +1,19 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 || s390x || arm || arm64 || loong64 || ppc64 || ppc64le || mips || mipsle || wasm || mips64 || mips64le || riscv64 + +package bytealg + +import _ "unsafe" // For go:linkname + +//go:noescape +func Compare(a, b []byte) int + +// The declaration below generates ABI wrappers for functions +// implemented in assembly in this package but declared in another +// package. + +//go:linkname abigen_runtime_cmpstring runtime.cmpstring +func abigen_runtime_cmpstring(a, b string) int diff --git a/src/internal/bytealg/compare_ppc64x.s b/src/internal/bytealg/compare_ppc64x.s new file mode 100644 index 0000000..63c33ee --- /dev/null +++ b/src/internal/bytealg/compare_ppc64x.s @@ -0,0 +1,332 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +#include "go_asm.h" +#include "textflag.h" + +// Helper names for x-form loads in BE ordering. +#ifdef GOARCH_ppc64le +#define _LDBEX MOVDBR +#define _LWBEX MOVWBR +#define _LHBEX MOVHBR +#else +#define _LDBEX MOVD +#define _LWBEX MOVW +#define _LHBEX MOVH +#endif + +#ifdef GOPPC64_power9 +#define SETB_CR0(rout) SETB CR0, rout +#define SETB_CR1(rout) SETB CR1, rout +#define SETB_INIT() +#define SETB_CR0_NE(rout) SETB_CR0(rout) +#else +// A helper macro to emulate SETB on P8. This assumes +// -1 is in R20, and 1 is in R21. crxlt and crxeq must +// also be the same CR field. +#define _SETB(crxlt, crxeq, rout) \ + ISEL crxeq,R0,R21,rout \ + ISEL crxlt,R20,rout,rout + +// A special case when it is know the comparison +// will always be not equal. The result must be -1 or 1. +#define SETB_CR0_NE(rout) \ + ISEL CR0LT,R20,R21,rout + +#define SETB_CR0(rout) _SETB(CR0LT, CR0EQ, rout) +#define SETB_CR1(rout) _SETB(CR1LT, CR1EQ, rout) +#define SETB_INIT() \ + MOVD $-1,R20 \ + MOVD $1,R21 +#endif + +TEXT ·Compare<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-56 + // incoming: + // R3 a addr + // R4 a len + // R6 b addr + // R7 b len + // + // on entry to cmpbody: + // R3 return value if len(a) == len(b) + // R5 a addr + // R6 b addr + // R9 min(len(a),len(b)) + SETB_INIT() + MOVD R3,R5 + CMP R4,R7,CR0 + CMP R3,R6,CR7 + ISEL CR0LT,R4,R7,R9 + SETB_CR0(R3) + BC $12,30,LR // beqlr cr7 + BR cmpbody<>(SB) + +TEXT runtime·cmpstring<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // incoming: + // R3 a addr -> R5 + // R4 a len -> R3 + // R5 b addr -> R6 + // R6 b len -> R4 + // + // on entry to cmpbody: + // R3 compare value if compared length is same. + // R5 a addr + // R6 b addr + // R9 min(len(a),len(b)) + SETB_INIT() + CMP R4,R6,CR0 + CMP R3,R5,CR7 + ISEL CR0LT,R4,R6,R9 + MOVD R5,R6 + MOVD R3,R5 + SETB_CR0(R3) + BC $12,30,LR // beqlr cr7 + BR cmpbody<>(SB) + +#ifdef GOARCH_ppc64le +DATA byteswap<>+0(SB)/8, $0x0706050403020100 +DATA byteswap<>+8(SB)/8, $0x0f0e0d0c0b0a0908 +GLOBL byteswap<>+0(SB), RODATA, $16 +#define SWAP V21 +#endif + +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0-0 +start: + CMP R9,$16,CR0 + CMP R9,$32,CR1 + CMP R9,$64,CR2 + MOVD $16,R10 + BLT cmp8 + BLT CR1,cmp16 + BLT CR2,cmp32 + +cmp64: // >= 64B + DCBT (R5) // optimize for size>=64 + DCBT (R6) // cache hint + + SRD $6,R9,R14 // There is at least one iteration. + MOVD R14,CTR + ANDCC $63,R9,R9 + CMP R9,$16,CR1 // Do setup for tail check early on. + CMP R9,$32,CR2 + CMP R9,$48,CR3 + ADD $-16,R9,R9 + + MOVD $32,R11 // set offsets to load into vector + MOVD $48,R12 // set offsets to load into vector + + PCALIGN $16 +cmp64_loop: + LXVD2X (R5)(R0),V3 // load bytes of A at offset 0 into vector + LXVD2X (R6)(R0),V4 // load bytes of B at offset 0 into vector + VCMPEQUDCC V3,V4,V1 + BGE CR6,different // jump out if its different + + LXVD2X (R5)(R10),V3 // load bytes of A at offset 16 into vector + LXVD2X (R6)(R10),V4 // load bytes of B at offset 16 into vector + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R5)(R11),V3 // load bytes of A at offset 32 into vector + LXVD2X (R6)(R11),V4 // load bytes of B at offset 32 into vector + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R5)(R12),V3 // load bytes of A at offset 64 into vector + LXVD2X (R6)(R12),V4 // load bytes of B at offset 64 into vector + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + ADD $64,R5,R5 // increment to next 64 bytes of A + ADD $64,R6,R6 // increment to next 64 bytes of B + BDNZ cmp64_loop + BC $12,2,LR // beqlr + + // Finish out tail with minimal overlapped checking. + // Note, 0 tail is handled by beqlr above. + BLE CR1,cmp64_tail_gt0 + BLE CR2,cmp64_tail_gt16 + BLE CR3,cmp64_tail_gt32 + +cmp64_tail_gt48: // 49 - 63 B + LXVD2X (R0)(R5),V3 + LXVD2X (R0)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R5)(R10),V3 + LXVD2X (R6)(R10),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R5)(R11),V3 + LXVD2X (R6)(R11),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + BR cmp64_tail_gt0 + + PCALIGN $16 +cmp64_tail_gt32: // 33 - 48B + LXVD2X (R0)(R5),V3 + LXVD2X (R0)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R5)(R10),V3 + LXVD2X (R6)(R10),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + BR cmp64_tail_gt0 + + PCALIGN $16 +cmp64_tail_gt16: // 17 - 32B + LXVD2X (R0)(R5),V3 + LXVD2X (R0)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + BR cmp64_tail_gt0 + + PCALIGN $16 +cmp64_tail_gt0: // 1 - 16B + LXVD2X (R5)(R9),V3 + LXVD2X (R6)(R9),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + RET + + PCALIGN $16 +cmp32: // 32 - 63B + ANDCC $31,R9,R9 + + LXVD2X (R0)(R5),V3 + LXVD2X (R0)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R10)(R5),V3 + LXVD2X (R10)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + BC $12,2,LR // beqlr + ADD R9,R10,R10 + + LXVD2X (R9)(R5),V3 + LXVD2X (R9)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + + LXVD2X (R10)(R5),V3 + LXVD2X (R10)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + RET + + PCALIGN $16 +cmp16: // 16 - 31B + ANDCC $15,R9,R9 + LXVD2X (R0)(R5),V3 + LXVD2X (R0)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + BC $12,2,LR // beqlr + + LXVD2X (R9)(R5),V3 + LXVD2X (R9)(R6),V4 + VCMPEQUDCC V3,V4,V1 + BGE CR6,different + RET + + PCALIGN $16 +different: +#ifdef GOARCH_ppc64le + MOVD $byteswap<>+00(SB),R16 + LXVD2X (R16)(R0),SWAP // Set up swap string + + VPERM V3,V3,SWAP,V3 + VPERM V4,V4,SWAP,V4 +#endif + + MFVSRD VS35,R16 // move upper doublewords of A and B into GPR for comparison + MFVSRD VS36,R10 + + CMPU R16,R10 + BEQ lower + SETB_CR0_NE(R3) + RET + + PCALIGN $16 +lower: + VSLDOI $8,V3,V3,V3 // move lower doublewords of A and B into GPR for comparison + MFVSRD VS35,R16 + VSLDOI $8,V4,V4,V4 + MFVSRD VS36,R10 + + CMPU R16,R10 + SETB_CR0_NE(R3) + RET + + PCALIGN $16 +cmp8: // 8 - 15B + CMP R9,$8 + BLT cmp4 + ANDCC $7,R9,R9 + _LDBEX (R0)(R5),R10 + _LDBEX (R0)(R6),R11 + _LDBEX (R9)(R5),R12 + _LDBEX (R9)(R6),R14 + CMPU R10,R11,CR0 + SETB_CR0(R5) + CMPU R12,R14,CR1 + SETB_CR1(R6) + CRAND CR0EQ,CR1EQ,CR1EQ // If both equal, length determines return value. + ISEL CR0EQ,R6,R5,R4 + ISEL CR1EQ,R3,R4,R3 + RET + + PCALIGN $16 +cmp4: // 4 - 7B + CMP R9,$4 + BLT cmp2 + ANDCC $3,R9,R9 + _LWBEX (R0)(R5),R10 + _LWBEX (R0)(R6),R11 + _LWBEX (R9)(R5),R12 + _LWBEX (R9)(R6),R14 + RLDIMI $32,R10,$0,R12 + RLDIMI $32,R11,$0,R14 + CMPU R12,R14 + BR cmp0 + + PCALIGN $16 +cmp2: // 2 - 3B + CMP R9,$2 + BLT cmp1 + ANDCC $1,R9,R9 + _LHBEX (R0)(R5),R10 + _LHBEX (R0)(R6),R11 + _LHBEX (R9)(R5),R12 + _LHBEX (R9)(R6),R14 + RLDIMI $32,R10,$0,R12 + RLDIMI $32,R11,$0,R14 + CMPU R12,R14 + BR cmp0 + + PCALIGN $16 +cmp1: + CMP R9,$0 + BEQ cmp0 + MOVBZ (R5),R10 + MOVBZ (R6),R11 + CMPU R10,R11 +cmp0: + SETB_CR0(R6) + ISEL CR0EQ,R3,R6,R3 + RET diff --git a/src/internal/bytealg/compare_riscv64.s b/src/internal/bytealg/compare_riscv64.s new file mode 100644 index 0000000..a4164a2 --- /dev/null +++ b/src/internal/bytealg/compare_riscv64.s @@ -0,0 +1,222 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-56 + // X10 = a_base + // X11 = a_len + // X12 = a_cap (unused) + // X13 = b_base (want in X12) + // X14 = b_len (want in X13) + // X15 = b_cap (unused) + MOV X13, X12 + MOV X14, X13 + JMP compare<>(SB) + +TEXT runtime·cmpstring<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // X10 = a_base + // X11 = a_len + // X12 = b_base + // X13 = b_len + JMP compare<>(SB) + +// On entry: +// X10 points to start of a +// X11 length of a +// X12 points to start of b +// X13 length of b +// for non-regabi X14 points to the address to store the return value (-1/0/1) +// for regabi the return value in X10 +TEXT compare<>(SB),NOSPLIT|NOFRAME,$0 + BEQ X10, X12, cmp_len + + MOV X11, X5 + BGE X13, X5, use_a_len // X5 = min(len(a), len(b)) + MOV X13, X5 +use_a_len: + BEQZ X5, cmp_len + + MOV $32, X6 + BLT X5, X6, check8_unaligned + + // Check alignment - if alignment differs we have to do one byte at a time. + AND $7, X10, X7 + AND $7, X12, X8 + BNE X7, X8, check8_unaligned + BEQZ X7, compare32 + + // Check one byte at a time until we reach 8 byte alignment. + SUB X7, X0, X7 + ADD $8, X7, X7 + SUB X7, X5, X5 +align: + ADD $-1, X7 + MOVBU 0(X10), X8 + MOVBU 0(X12), X9 + BNE X8, X9, cmp + ADD $1, X10 + ADD $1, X12 + BNEZ X7, align + +check32: + // X6 contains $32 + BLT X5, X6, compare16 +compare32: + MOV 0(X10), X15 + MOV 0(X12), X16 + MOV 8(X10), X17 + MOV 8(X12), X18 + BNE X15, X16, cmp8a + BNE X17, X18, cmp8b + MOV 16(X10), X15 + MOV 16(X12), X16 + MOV 24(X10), X17 + MOV 24(X12), X18 + BNE X15, X16, cmp8a + BNE X17, X18, cmp8b + ADD $32, X10 + ADD $32, X12 + ADD $-32, X5 + BGE X5, X6, compare32 + BEQZ X5, cmp_len + +check16: + MOV $16, X6 + BLT X5, X6, check8_unaligned +compare16: + MOV 0(X10), X15 + MOV 0(X12), X16 + MOV 8(X10), X17 + MOV 8(X12), X18 + BNE X15, X16, cmp8a + BNE X17, X18, cmp8b + ADD $16, X10 + ADD $16, X12 + ADD $-16, X5 + BEQZ X5, cmp_len + +check8_unaligned: + MOV $8, X6 + BLT X5, X6, check4_unaligned +compare8_unaligned: + MOVBU 0(X10), X8 + MOVBU 1(X10), X15 + MOVBU 2(X10), X17 + MOVBU 3(X10), X19 + MOVBU 4(X10), X21 + MOVBU 5(X10), X23 + MOVBU 6(X10), X25 + MOVBU 7(X10), X29 + MOVBU 0(X12), X9 + MOVBU 1(X12), X16 + MOVBU 2(X12), X18 + MOVBU 3(X12), X20 + MOVBU 4(X12), X22 + MOVBU 5(X12), X24 + MOVBU 6(X12), X28 + MOVBU 7(X12), X30 + BNE X8, X9, cmp1a + BNE X15, X16, cmp1b + BNE X17, X18, cmp1c + BNE X19, X20, cmp1d + BNE X21, X22, cmp1e + BNE X23, X24, cmp1f + BNE X25, X28, cmp1g + BNE X29, X30, cmp1h + ADD $8, X10 + ADD $8, X12 + ADD $-8, X5 + BGE X5, X6, compare8_unaligned + BEQZ X5, cmp_len + +check4_unaligned: + MOV $4, X6 + BLT X5, X6, compare1 +compare4_unaligned: + MOVBU 0(X10), X8 + MOVBU 1(X10), X15 + MOVBU 2(X10), X17 + MOVBU 3(X10), X19 + MOVBU 0(X12), X9 + MOVBU 1(X12), X16 + MOVBU 2(X12), X18 + MOVBU 3(X12), X20 + BNE X8, X9, cmp1a + BNE X15, X16, cmp1b + BNE X17, X18, cmp1c + BNE X19, X20, cmp1d + ADD $4, X10 + ADD $4, X12 + ADD $-4, X5 + BGE X5, X6, compare4_unaligned + +compare1: + BEQZ X5, cmp_len + MOVBU 0(X10), X8 + MOVBU 0(X12), X9 + BNE X8, X9, cmp + ADD $1, X10 + ADD $1, X12 + ADD $-1, X5 + JMP compare1 + + // Compare 8 bytes of memory in X15/X16 that are known to differ. +cmp8a: + MOV X15, X17 + MOV X16, X18 + + // Compare 8 bytes of memory in X17/X18 that are known to differ. +cmp8b: + MOV $0xff, X19 +cmp8_loop: + AND X17, X19, X8 + AND X18, X19, X9 + BNE X8, X9, cmp + SLLI $8, X19 + JMP cmp8_loop + +cmp1a: + SLTU X9, X8, X5 + SLTU X8, X9, X6 + JMP cmp_ret +cmp1b: + SLTU X16, X15, X5 + SLTU X15, X16, X6 + JMP cmp_ret +cmp1c: + SLTU X18, X17, X5 + SLTU X17, X18, X6 + JMP cmp_ret +cmp1d: + SLTU X20, X19, X5 + SLTU X19, X20, X6 + JMP cmp_ret +cmp1e: + SLTU X22, X21, X5 + SLTU X21, X22, X6 + JMP cmp_ret +cmp1f: + SLTU X24, X23, X5 + SLTU X23, X24, X6 + JMP cmp_ret +cmp1g: + SLTU X28, X25, X5 + SLTU X25, X28, X6 + JMP cmp_ret +cmp1h: + SLTU X30, X29, X5 + SLTU X29, X30, X6 + JMP cmp_ret + +cmp_len: + MOV X11, X8 + MOV X13, X9 +cmp: + SLTU X9, X8, X5 + SLTU X8, X9, X6 +cmp_ret: + SUB X5, X6, X10 + RET diff --git a/src/internal/bytealg/compare_s390x.s b/src/internal/bytealg/compare_s390x.s new file mode 100644 index 0000000..5394548 --- /dev/null +++ b/src/internal/bytealg/compare_s390x.s @@ -0,0 +1,69 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB),NOSPLIT|NOFRAME,$0-56 + MOVD a_base+0(FP), R3 + MOVD a_len+8(FP), R4 + MOVD b_base+24(FP), R5 + MOVD b_len+32(FP), R6 + LA ret+48(FP), R7 + BR cmpbody<>(SB) + +TEXT runtime·cmpstring(SB),NOSPLIT|NOFRAME,$0-40 + MOVD a_base+0(FP), R3 + MOVD a_len+8(FP), R4 + MOVD b_base+16(FP), R5 + MOVD b_len+24(FP), R6 + LA ret+32(FP), R7 + BR cmpbody<>(SB) + +// input: +// R3 = a +// R4 = alen +// R5 = b +// R6 = blen +// R7 = address of output word (stores -1/0/1 here) +TEXT cmpbody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMPBEQ R3, R5, cmplengths + MOVD R4, R8 + CMPBLE R4, R6, amin + MOVD R6, R8 +amin: + CMPBEQ R8, $0, cmplengths + CMP R8, $256 + BLE tail +loop: + CLC $256, 0(R3), 0(R5) + BGT gt + BLT lt + SUB $256, R8 + MOVD $256(R3), R3 + MOVD $256(R5), R5 + CMP R8, $256 + BGT loop +tail: + SUB $1, R8 + EXRL $cmpbodyclc<>(SB), R8 + BGT gt + BLT lt +cmplengths: + CMP R4, R6 + BEQ eq + BLT lt +gt: + MOVD $1, 0(R7) + RET +lt: + MOVD $-1, 0(R7) + RET +eq: + MOVD $0, 0(R7) + RET + +TEXT cmpbodyclc<>(SB),NOSPLIT|NOFRAME,$0-0 + CLC $1, 0(R3), 0(R5) + RET diff --git a/src/internal/bytealg/compare_wasm.s b/src/internal/bytealg/compare_wasm.s new file mode 100644 index 0000000..dc8fb33 --- /dev/null +++ b/src/internal/bytealg/compare_wasm.s @@ -0,0 +1,115 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Compare(SB), NOSPLIT, $0-56 + Get SP + I64Load a_base+0(FP) + I64Load a_len+8(FP) + I64Load b_base+24(FP) + I64Load b_len+32(FP) + Call cmpbody<>(SB) + I64Store ret+48(FP) + RET + +TEXT runtime·cmpstring(SB), NOSPLIT, $0-40 + Get SP + I64Load a_base+0(FP) + I64Load a_len+8(FP) + I64Load b_base+16(FP) + I64Load b_len+24(FP) + Call cmpbody<>(SB) + I64Store ret+32(FP) + RET + +// params: a, alen, b, blen +// ret: -1/0/1 +TEXT cmpbody<>(SB), NOSPLIT, $0-0 + // len = min(alen, blen) + Get R1 + Get R3 + Get R1 + Get R3 + I64LtU + Select + Set R4 + + Get R0 + I32WrapI64 + Get R2 + I32WrapI64 + Get R4 + I32WrapI64 + Call memcmp<>(SB) + I64ExtendI32S + Tee R5 + + I64Eqz + If + // check length + Get R1 + Get R3 + I64Sub + Set R5 + End + + I64Const $0 + I64Const $-1 + I64Const $1 + Get R5 + I64Const $0 + I64LtS + Select + Get R5 + I64Eqz + Select + Return + +// compiled with emscripten +// params: a, b, len +// ret: <0/0/>0 +TEXT memcmp<>(SB), NOSPLIT, $0-0 + Get R2 + If $1 + Loop + Get R0 + I32Load8S $0 + Tee R3 + Get R1 + I32Load8S $0 + Tee R4 + I32Eq + If + Get R0 + I32Const $1 + I32Add + Set R0 + Get R1 + I32Const $1 + I32Add + Set R1 + I32Const $0 + Get R2 + I32Const $-1 + I32Add + Tee R2 + I32Eqz + BrIf $3 + Drop + Br $1 + End + End + Get R3 + I32Const $255 + I32And + Get R4 + I32Const $255 + I32And + I32Sub + Else + I32Const $0 + End + Return diff --git a/src/internal/bytealg/count_amd64.s b/src/internal/bytealg/count_amd64.s new file mode 100644 index 0000000..efb17f8 --- /dev/null +++ b/src/internal/bytealg/count_amd64.s @@ -0,0 +1,208 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "asm_amd64.h" +#include "textflag.h" + +TEXT ·Count(SB),NOSPLIT,$0-40 +#ifndef hasPOPCNT + CMPB internal∕cpu·X86+const_offsetX86HasPOPCNT(SB), $1 + JEQ 2(PC) + JMP ·countGeneric(SB) +#endif + MOVQ b_base+0(FP), SI + MOVQ b_len+8(FP), BX + MOVB c+24(FP), AL + LEAQ ret+32(FP), R8 + JMP countbody<>(SB) + +TEXT ·CountString(SB),NOSPLIT,$0-32 +#ifndef hasPOPCNT + CMPB internal∕cpu·X86+const_offsetX86HasPOPCNT(SB), $1 + JEQ 2(PC) + JMP ·countGenericString(SB) +#endif + MOVQ s_base+0(FP), SI + MOVQ s_len+8(FP), BX + MOVB c+16(FP), AL + LEAQ ret+24(FP), R8 + JMP countbody<>(SB) + +// input: +// SI: data +// BX: data len +// AL: byte sought +// R8: address to put result +// This function requires the POPCNT instruction. +TEXT countbody<>(SB),NOSPLIT,$0 + // Shuffle X0 around so that each byte contains + // the character we're looking for. + MOVD AX, X0 + PUNPCKLBW X0, X0 + PUNPCKLBW X0, X0 + PSHUFL $0, X0, X0 + + CMPQ BX, $16 + JLT small + + MOVQ $0, R12 // Accumulator + + MOVQ SI, DI + + CMPQ BX, $32 + JA avx2 +sse: + LEAQ -16(SI)(BX*1), AX // AX = address of last 16 bytes + JMP sseloopentry + +sseloop: + // Move the next 16-byte chunk of the data into X1. + MOVOU (DI), X1 + // Compare bytes in X0 to X1. + PCMPEQB X0, X1 + // Take the top bit of each byte in X1 and put the result in DX. + PMOVMSKB X1, DX + // Count number of matching bytes + POPCNTL DX, DX + // Accumulate into R12 + ADDQ DX, R12 + // Advance to next block. + ADDQ $16, DI +sseloopentry: + CMPQ DI, AX + JBE sseloop + + // Get the number of bytes to consider in the last 16 bytes + ANDQ $15, BX + JZ end + + // Create mask to ignore overlap between previous 16 byte block + // and the next. + MOVQ $16,CX + SUBQ BX, CX + MOVQ $0xFFFF, R10 + SARQ CL, R10 + SALQ CL, R10 + + // Process the last 16-byte chunk. This chunk may overlap with the + // chunks we've already searched so we need to mask part of it. + MOVOU (AX), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, DX + // Apply mask + ANDQ R10, DX + POPCNTL DX, DX + ADDQ DX, R12 +end: + MOVQ R12, (R8) + RET + +// handle for lengths < 16 +small: + TESTQ BX, BX + JEQ endzero + + // Check if we'll load across a page boundary. + LEAQ 16(SI), AX + TESTW $0xff0, AX + JEQ endofpage + + // We must ignore high bytes as they aren't part of our slice. + // Create mask. + MOVB BX, CX + MOVQ $1, R10 + SALQ CL, R10 + SUBQ $1, R10 + + // Load data + MOVOU (SI), X1 + // Compare target byte with each byte in data. + PCMPEQB X0, X1 + // Move result bits to integer register. + PMOVMSKB X1, DX + // Apply mask + ANDQ R10, DX + POPCNTL DX, DX + // Directly return DX, we don't need to accumulate + // since we have <16 bytes. + MOVQ DX, (R8) + RET +endzero: + MOVQ $0, (R8) + RET + +endofpage: + // We must ignore low bytes as they aren't part of our slice. + MOVQ $16,CX + SUBQ BX, CX + MOVQ $0xFFFF, R10 + SARQ CL, R10 + SALQ CL, R10 + + // Load data into the high end of X1. + MOVOU -16(SI)(BX*1), X1 + // Compare target byte with each byte in data. + PCMPEQB X0, X1 + // Move result bits to integer register. + PMOVMSKB X1, DX + // Apply mask + ANDQ R10, DX + // Directly return DX, we don't need to accumulate + // since we have <16 bytes. + POPCNTL DX, DX + MOVQ DX, (R8) + RET + +avx2: +#ifndef hasAVX2 + CMPB internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1 + JNE sse +#endif + MOVD AX, X0 + LEAQ -32(SI)(BX*1), R11 + VPBROADCASTB X0, Y1 +avx2_loop: + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPMOVMSKB Y3, DX + POPCNTL DX, DX + ADDQ DX, R12 + ADDQ $32, DI + CMPQ DI, R11 + JLE avx2_loop + + // If last block is already processed, + // skip to the end. + CMPQ DI, R11 + JEQ endavx + + // Load address of the last 32 bytes. + // There is an overlap with the previous block. + MOVQ R11, DI + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPMOVMSKB Y3, DX + // Exit AVX mode. + VZEROUPPER + + // Create mask to ignore overlap between previous 32 byte block + // and the next. + ANDQ $31, BX + MOVQ $32,CX + SUBQ BX, CX + MOVQ $0xFFFFFFFF, R10 + SARQ CL, R10 + SALQ CL, R10 + // Apply mask + ANDQ R10, DX + POPCNTL DX, DX + ADDQ DX, R12 + MOVQ R12, (R8) + RET +endavx: + // Exit AVX mode. + VZEROUPPER + MOVQ R12, (R8) + RET diff --git a/src/internal/bytealg/count_arm.s b/src/internal/bytealg/count_arm.s new file mode 100644 index 0000000..f704ea0 --- /dev/null +++ b/src/internal/bytealg/count_arm.s @@ -0,0 +1,43 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Count(SB),NOSPLIT,$0-20 + MOVW b_base+0(FP), R0 + MOVW b_len+4(FP), R1 + MOVBU c+12(FP), R2 + MOVW $ret+16(FP), R7 + B countbytebody<>(SB) + +TEXT ·CountString(SB),NOSPLIT,$0-16 + MOVW s_base+0(FP), R0 + MOVW s_len+4(FP), R1 + MOVBU c+8(FP), R2 + MOVW $ret+12(FP), R7 + B countbytebody<>(SB) + +// Input: +// R0: data +// R1: data length +// R2: byte to find +// R7: address to put result +// +// On exit: +// R4 and R8 are clobbered +TEXT countbytebody<>(SB),NOSPLIT,$0 + MOVW $0, R8 // R8 = count of byte to search + CMP $0, R1 + B.EQ done // short path to handle 0-byte case + ADD R0, R1 // R1 is the end of the range +byte_loop: + MOVBU.P 1(R0), R4 + CMP R4, R2 + ADD.EQ $1, R8 + CMP R0, R1 + B.NE byte_loop +done: + MOVW R8, (R7) + RET diff --git a/src/internal/bytealg/count_arm64.s b/src/internal/bytealg/count_arm64.s new file mode 100644 index 0000000..8cd703d --- /dev/null +++ b/src/internal/bytealg/count_arm64.s @@ -0,0 +1,90 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Count(SB),NOSPLIT,$0-40 + MOVD b_base+0(FP), R0 + MOVD b_len+8(FP), R2 + MOVBU c+24(FP), R1 + MOVD $ret+32(FP), R8 + B countbytebody<>(SB) + +TEXT ·CountString(SB),NOSPLIT,$0-32 + MOVD s_base+0(FP), R0 + MOVD s_len+8(FP), R2 + MOVBU c+16(FP), R1 + MOVD $ret+24(FP), R8 + B countbytebody<>(SB) + +// input: +// R0: data +// R2: data len +// R1: byte to find +// R8: address to put result +TEXT countbytebody<>(SB),NOSPLIT,$0 + // R11 = count of byte to search + MOVD $0, R11 + // short path to handle 0-byte case + CBZ R2, done + CMP $0x20, R2 + // jump directly to tail if length < 32 + BLO tail + ANDS $0x1f, R0, R9 + BEQ chunk + // Work with not 32-byte aligned head + BIC $0x1f, R0, R3 + ADD $0x20, R3 +head_loop: + MOVBU.P 1(R0), R5 + CMP R5, R1 + CINC EQ, R11, R11 + SUB $1, R2, R2 + CMP R0, R3 + BNE head_loop + // Work with 32-byte aligned chunks +chunk: + BIC $0x1f, R2, R9 + // The first chunk can also be the last + CBZ R9, tail + // R3 = end of 32-byte chunks + ADD R0, R9, R3 + MOVD $1, R5 + VMOV R5, V5.B16 + // R2 = length of tail + SUB R9, R2, R2 + // Duplicate R1 (byte to search) to 16 1-byte elements of V0 + VMOV R1, V0.B16 + // Clear the low 64-bit element of V7 and V8 + VEOR V7.B8, V7.B8, V7.B8 + VEOR V8.B8, V8.B8, V8.B8 + // Count the target byte in 32-byte chunk +chunk_loop: + VLD1.P (R0), [V1.B16, V2.B16] + CMP R0, R3 + VCMEQ V0.B16, V1.B16, V3.B16 + VCMEQ V0.B16, V2.B16, V4.B16 + // Clear the higher 7 bits + VAND V5.B16, V3.B16, V3.B16 + VAND V5.B16, V4.B16, V4.B16 + // Count lanes match the requested byte + VADDP V4.B16, V3.B16, V6.B16 // 32B->16B + VUADDLV V6.B16, V7 + // Accumulate the count in low 64-bit element of V8 when inside the loop + VADD V7, V8 + BNE chunk_loop + VMOV V8.D[0], R6 + ADD R6, R11, R11 + CBZ R2, done +tail: + // Work with tail shorter than 32 bytes + MOVBU.P 1(R0), R5 + SUB $1, R2, R2 + CMP R5, R1 + CINC EQ, R11, R11 + CBNZ R2, tail +done: + MOVD R11, (R8) + RET diff --git a/src/internal/bytealg/count_generic.go b/src/internal/bytealg/count_generic.go new file mode 100644 index 0000000..932a7c5 --- /dev/null +++ b/src/internal/bytealg/count_generic.go @@ -0,0 +1,27 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !amd64 && !arm && !arm64 && !ppc64le && !ppc64 && !riscv64 && !s390x + +package bytealg + +func Count(b []byte, c byte) int { + n := 0 + for _, x := range b { + if x == c { + n++ + } + } + return n +} + +func CountString(s string, c byte) int { + n := 0 + for i := 0; i < len(s); i++ { + if s[i] == c { + n++ + } + } + return n +} diff --git a/src/internal/bytealg/count_native.go b/src/internal/bytealg/count_native.go new file mode 100644 index 0000000..90189c9 --- /dev/null +++ b/src/internal/bytealg/count_native.go @@ -0,0 +1,33 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build amd64 || arm || arm64 || ppc64le || ppc64 || riscv64 || s390x + +package bytealg + +//go:noescape +func Count(b []byte, c byte) int + +//go:noescape +func CountString(s string, c byte) int + +// A backup implementation to use by assembly. +func countGeneric(b []byte, c byte) int { + n := 0 + for _, x := range b { + if x == c { + n++ + } + } + return n +} +func countGenericString(s string, c byte) int { + n := 0 + for i := 0; i < len(s); i++ { + if s[i] == c { + n++ + } + } + return n +} diff --git a/src/internal/bytealg/count_ppc64x.s b/src/internal/bytealg/count_ppc64x.s new file mode 100644 index 0000000..2d2490b --- /dev/null +++ b/src/internal/bytealg/count_ppc64x.s @@ -0,0 +1,96 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64le || ppc64 + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Count<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // R3 = byte array pointer + // R4 = length + MOVBZ R6, R5 // R5 = byte + BR countbytebody<>(SB) + +TEXT ·CountString<ABIInternal>(SB), NOSPLIT|NOFRAME, $0-32 + // R3 = byte array pointer + // R4 = length + MOVBZ R5, R5 // R5 = byte + BR countbytebody<>(SB) + +// R3: addr of string +// R4: len of string +// R5: byte to count +// On exit: +// R3: return value +// endianness shouldn't matter since we are just counting and order +// is irrelevant +TEXT countbytebody<>(SB), NOSPLIT|NOFRAME, $0-0 + DCBT (R3) // Prepare cache line. + MOVD R0, R18 // byte count + MOVD R3, R19 // Save base address for calculating the index later. + MOVD R4, R16 + + MOVD R5, R6 + RLDIMI $8, R6, $48, R6 + RLDIMI $16, R6, $32, R6 + RLDIMI $32, R6, $0, R6 // fill reg with the byte to count + + VSPLTISW $3, V4 // used for shift + MTVRD R6, V1 // move compare byte + VSPLTB $7, V1, V1 // replicate byte across V1 + + CMPU R4, $32 // Check if it's a small string (<32 bytes) + BLT tail // Jump to the small string case + XXLXOR VS37, VS37, VS37 // clear V5 (aka VS37) to use as accumulator + +cmploop: + LXVW4X (R3), VS32 // load bytes from string + + // when the bytes match, the corresponding byte contains all 1s + VCMPEQUB V1, V0, V2 // compare bytes + VPOPCNTD V2, V3 // each double word contains its count + VADDUDM V3, V5, V5 // accumulate bit count in each double word + ADD $16, R3, R3 // increment pointer + SUB $16, R16, R16 // remaining bytes + CMP R16, $16 // at least 16 remaining? + BGE cmploop + VSRD V5, V4, V5 // shift by 3 to convert bits to bytes + VSLDOI $8, V5, V5, V6 // get the double word values from vector + MFVSRD V5, R9 + MFVSRD V6, R10 + ADD R9, R10, R9 + ADD R9, R18, R18 + +tail: + CMP R16, $8 // 8 bytes left? + BLT small + + MOVD (R3), R12 // load 8 bytes + CMPB R12, R6, R17 // compare bytes + POPCNTD R17, R15 // bit count + SRD $3, R15, R15 // byte count + ADD R15, R18, R18 // add to byte count + +next1: + ADD $8, R3, R3 + SUB $8, R16, R16 // remaining bytes + BR tail + +small: + CMP $0, R16 // any remaining + BEQ done + MOVBZ (R3), R12 // check each remaining byte + CMP R12, R5 + BNE next2 + ADD $1, R18 + +next2: + SUB $1, R16 + ADD $1, R3 // inc address + BR small + +done: + MOVD R18, R3 // return count + RET diff --git a/src/internal/bytealg/count_riscv64.s b/src/internal/bytealg/count_riscv64.s new file mode 100644 index 0000000..d123cbd --- /dev/null +++ b/src/internal/bytealg/count_riscv64.s @@ -0,0 +1,47 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Count<ABIInternal>(SB),NOSPLIT,$0-40 + // X10 = b_base + // X11 = b_len + // X12 = b_cap (unused) + // X13 = byte to count (want in X12) + AND $0xff, X13, X12 + MOV ZERO, X14 // count + ADD X10, X11 // end + +loop: + BEQ X10, X11, done + MOVBU (X10), X15 + ADD $1, X10 + BNE X12, X15, loop + ADD $1, X14 + JMP loop + +done: + MOV X14, X10 + RET + +TEXT ·CountString<ABIInternal>(SB),NOSPLIT,$0-32 + // X10 = s_base + // X11 = s_len + // X12 = byte to count + AND $0xff, X12 + MOV ZERO, X14 // count + ADD X10, X11 // end + +loop: + BEQ X10, X11, done + MOVBU (X10), X15 + ADD $1, X10 + BNE X12, X15, loop + ADD $1, X14 + JMP loop + +done: + MOV X14, X10 + RET diff --git a/src/internal/bytealg/count_s390x.s b/src/internal/bytealg/count_s390x.s new file mode 100644 index 0000000..2a3b5c0 --- /dev/null +++ b/src/internal/bytealg/count_s390x.s @@ -0,0 +1,169 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// condition code masks +#define EQ 8 +#define NE 7 + +// register assignments +#define R_ZERO R0 +#define R_VAL R1 +#define R_TMP R2 +#define R_PTR R3 +#define R_LEN R4 +#define R_CHAR R5 +#define R_RET R6 +#define R_ITER R7 +#define R_CNT R8 +#define R_MPTR R9 + +// vector register assignments +#define V_ZERO V0 +#define V_CHAR V1 +#define V_MASK V2 +#define V_VAL V3 +#define V_CNT V4 + +// mask for trailing bytes in vector implementation +GLOBL countbytemask<>(SB), RODATA, $16 +DATA countbytemask<>+0(SB)/8, $0x0101010101010101 +DATA countbytemask<>+8(SB)/8, $0x0101010101010101 + +// func Count(b []byte, c byte) int +TEXT ·Count(SB), NOSPLIT|NOFRAME, $0-40 + LMG b+0(FP), R_PTR, R_LEN + MOVBZ c+24(FP), R_CHAR + MOVD $ret+32(FP), R_RET + BR countbytebody<>(SB) + +// func CountString(s string, c byte) int +TEXT ·CountString(SB), NOSPLIT|NOFRAME, $0-32 + LMG s+0(FP), R_PTR, R_LEN + MOVBZ c+16(FP), R_CHAR + MOVD $ret+24(FP), R_RET + BR countbytebody<>(SB) + +// input: +// R_PTR = address of array of bytes +// R_LEN = number of bytes in array +// R_CHAR = byte value to count zero (extended to register width) +// R_RET = address of return value +TEXT countbytebody<>(SB), NOSPLIT|NOFRAME, $0-0 + MOVD $internal∕cpu·S390X+const_offsetS390xHasVX(SB), R_TMP + MOVD $countbytemask<>(SB), R_MPTR + CGIJ $EQ, R_LEN, $0, ret0 // return if length is 0. + SRD $4, R_LEN, R_ITER // R_ITER is the number of 16-byte chunks + MOVBZ (R_TMP), R_TMP // load bool indicating support for vector facility + CGIJ $EQ, R_TMP, $0, novx // jump to scalar code if the vector facility is not available + + // Start of vector code (have vector facility). + // + // Set R_LEN to be the length mod 16 minus 1 to use as an index for + // vector 'load with length' (VLL). It will be in the range [-1,14]. + // Also replicate c across a 16-byte vector and initialize V_ZERO. + ANDW $0xf, R_LEN + VLVGB $0, R_CHAR, V_CHAR // V_CHAR = [16]byte{c, 0, ..., 0, 0} + VZERO V_ZERO // V_ZERO = [1]uint128{0} + ADDW $-1, R_LEN + VREPB $0, V_CHAR, V_CHAR // V_CHAR = [16]byte{c, c, ..., c, c} + + // Jump to loop if we have more than 15 bytes to process. + CGIJ $NE, R_ITER, $0, vxchunks + + // Load 1-15 bytes and corresponding mask. + // Note: only the low 32-bits of R_LEN are used for the index. + VLL R_LEN, (R_PTR), V_VAL + VLL R_LEN, (R_MPTR), V_MASK + + // Compare each byte in input chunk against byte to be counted. + // Each byte element will be set to either 0 (no match) or 1 (match). + VCEQB V_CHAR, V_VAL, V_VAL // each byte will be either 0xff or 0x00 + VN V_MASK, V_VAL, V_VAL // mask out most significant 7 bits + + // Accumulate matched byte count in 128-bit integer value. + VSUMB V_VAL, V_ZERO, V_VAL // [16]byte{x0, x1, ..., x14, x15} → [4]uint32{x0+x1+x2+x3, ..., x12+x13+x14+x15} + VSUMQF V_VAL, V_ZERO, V_CNT // [4]uint32{x0, x1, x2, x3} → [1]uint128{x0+x1+x2+x3} + + // Return rightmost (lowest) 64-bit part of accumulator. + VSTEG $1, V_CNT, (R_RET) + RET + +vxchunks: + // Load 0x01 into every byte element in the 16-byte mask vector. + VREPIB $1, V_MASK // V_MASK = [16]byte{1, 1, ..., 1, 1} + VZERO V_CNT // initial uint128 count of 0 + +vxloop: + // Load input bytes in 16-byte chunks. + VL (R_PTR), V_VAL + + // Compare each byte in input chunk against byte to be counted. + // Each byte element will be set to either 0 (no match) or 1 (match). + VCEQB V_CHAR, V_VAL, V_VAL // each byte will be either 0xff or 0x00 + VN V_MASK, V_VAL, V_VAL // mask out most significant 7 bits + + // Increment input string address. + MOVD $16(R_PTR), R_PTR + + // Accumulate matched byte count in 128-bit integer value. + VSUMB V_VAL, V_ZERO, V_VAL // [16]byte{x0, x1, ..., x14, x15} → [4]uint32{x0+x1+x2+x3, ..., x12+x13+x14+x15} + VSUMQF V_VAL, V_ZERO, V_VAL // [4]uint32{x0, x1, x2, x3} → [1]uint128{x0+x1+x2+x3} + VAQ V_VAL, V_CNT, V_CNT // accumulate + + // Repeat until all 16-byte chunks are done. + BRCTG R_ITER, vxloop + + // Skip to end if there are no trailing bytes. + CIJ $EQ, R_LEN, $-1, vxret + + // Load 1-15 bytes and corresponding mask. + // Note: only the low 32-bits of R_LEN are used for the index. + VLL R_LEN, (R_PTR), V_VAL + VLL R_LEN, (R_MPTR), V_MASK + + // Compare each byte in input chunk against byte to be counted. + // Each byte element will be set to either 0 (no match) or 1 (match). + VCEQB V_CHAR, V_VAL, V_VAL + VN V_MASK, V_VAL, V_VAL + + // Accumulate matched byte count in 128-bit integer value. + VSUMB V_VAL, V_ZERO, V_VAL // [16]byte{x0, x1, ..., x14, x15} → [4]uint32{x0+x1+x2+x3, ..., x12+x13+x14+x15} + VSUMQF V_VAL, V_ZERO, V_VAL // [4]uint32{x0, x1, x2, x3} → [1]uint128{x0+x1+x2+x3} + VAQ V_VAL, V_CNT, V_CNT // accumulate + +vxret: + // Return rightmost (lowest) 64-bit part of accumulator. + VSTEG $1, V_CNT, (R_RET) + RET + +novx: + // Start of non-vector code (the vector facility not available). + // + // Initialise counter and constant zero. + MOVD $0, R_CNT + MOVD $0, R_ZERO + +loop: + // Read 1-byte from input and compare. + // Note: avoid putting LOCGR in critical path. + MOVBZ (R_PTR), R_VAL + MOVD $1, R_TMP + MOVD $1(R_PTR), R_PTR + CMPW R_VAL, R_CHAR + LOCGR $NE, R_ZERO, R_TMP // select 0 if no match (1 if there is a match) + ADD R_TMP, R_CNT // accumulate 64-bit result + + // Repeat until all bytes have been checked. + BRCTG R_LEN, loop + +ret: + MOVD R_CNT, (R_RET) + RET + +ret0: + MOVD $0, (R_RET) + RET diff --git a/src/internal/bytealg/equal_386.s b/src/internal/bytealg/equal_386.s new file mode 100644 index 0000000..58b3cbe --- /dev/null +++ b/src/internal/bytealg/equal_386.s @@ -0,0 +1,130 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT,$0-13 + MOVL a+0(FP), SI + MOVL b+4(FP), DI + CMPL SI, DI + JEQ eq + MOVL size+8(FP), BX + LEAL ret+12(FP), AX + JMP memeqbody<>(SB) +eq: + MOVB $1, ret+12(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT,$0-9 + MOVL a+0(FP), SI + MOVL b+4(FP), DI + CMPL SI, DI + JEQ eq + MOVL 4(DX), BX // compiler stores size at offset 4 in the closure + LEAL ret+8(FP), AX + JMP memeqbody<>(SB) +eq: + MOVB $1, ret+8(FP) + RET + +// a in SI +// b in DI +// count in BX +// address of result byte in AX +TEXT memeqbody<>(SB),NOSPLIT,$0-0 + CMPL BX, $4 + JB small + + // 64 bytes at a time using xmm registers +hugeloop: + CMPL BX, $64 + JB bigloop +#ifdef GO386_softfloat + JMP bigloop +#endif + MOVOU (SI), X0 + MOVOU (DI), X1 + MOVOU 16(SI), X2 + MOVOU 16(DI), X3 + MOVOU 32(SI), X4 + MOVOU 32(DI), X5 + MOVOU 48(SI), X6 + MOVOU 48(DI), X7 + PCMPEQB X1, X0 + PCMPEQB X3, X2 + PCMPEQB X5, X4 + PCMPEQB X7, X6 + PAND X2, X0 + PAND X6, X4 + PAND X4, X0 + PMOVMSKB X0, DX + ADDL $64, SI + ADDL $64, DI + SUBL $64, BX + CMPL DX, $0xffff + JEQ hugeloop + MOVB $0, (AX) + RET + + // 4 bytes at a time using 32-bit register +bigloop: + CMPL BX, $4 + JBE leftover + MOVL (SI), CX + MOVL (DI), DX + ADDL $4, SI + ADDL $4, DI + SUBL $4, BX + CMPL CX, DX + JEQ bigloop + MOVB $0, (AX) + RET + + // remaining 0-4 bytes +leftover: + MOVL -4(SI)(BX*1), CX + MOVL -4(DI)(BX*1), DX + CMPL CX, DX + SETEQ (AX) + RET + +small: + CMPL BX, $0 + JEQ equal + + LEAL 0(BX*8), CX + NEGL CX + + MOVL SI, DX + CMPB DX, $0xfc + JA si_high + + // load at SI won't cross a page boundary. + MOVL (SI), SI + JMP si_finish +si_high: + // address ends in 111111xx. Load up to bytes we want, move to correct position. + MOVL -4(SI)(BX*1), SI + SHRL CX, SI +si_finish: + + // same for DI. + MOVL DI, DX + CMPB DX, $0xfc + JA di_high + MOVL (DI), DI + JMP di_finish +di_high: + MOVL -4(DI)(BX*1), DI + SHRL CX, DI +di_finish: + + SUBL SI, DI + SHLL CX, DI +equal: + SETEQ (AX) + RET diff --git a/src/internal/bytealg/equal_amd64.s b/src/internal/bytealg/equal_amd64.s new file mode 100644 index 0000000..d178a33 --- /dev/null +++ b/src/internal/bytealg/equal_amd64.s @@ -0,0 +1,162 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "asm_amd64.h" +#include "textflag.h" + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal<ABIInternal>(SB),NOSPLIT,$0-25 + // AX = a (want in SI) + // BX = b (want in DI) + // CX = size (want in BX) + CMPQ AX, BX + JNE neq + MOVQ $1, AX // return 1 + RET +neq: + MOVQ AX, SI + MOVQ BX, DI + MOVQ CX, BX + JMP memeqbody<>(SB) + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT,$0-17 + // AX = a (want in SI) + // BX = b (want in DI) + // 8(DX) = size (want in BX) + CMPQ AX, BX + JNE neq + MOVQ $1, AX // return 1 + RET +neq: + MOVQ AX, SI + MOVQ BX, DI + MOVQ 8(DX), BX // compiler stores size at offset 8 in the closure + JMP memeqbody<>(SB) + +// Input: +// a in SI +// b in DI +// count in BX +// Output: +// result in AX +TEXT memeqbody<>(SB),NOSPLIT,$0-0 + CMPQ BX, $8 + JB small + CMPQ BX, $64 + JB bigloop +#ifndef hasAVX2 + CMPB internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1 + JE hugeloop_avx2 + + // 64 bytes at a time using xmm registers +hugeloop: + CMPQ BX, $64 + JB bigloop + MOVOU (SI), X0 + MOVOU (DI), X1 + MOVOU 16(SI), X2 + MOVOU 16(DI), X3 + MOVOU 32(SI), X4 + MOVOU 32(DI), X5 + MOVOU 48(SI), X6 + MOVOU 48(DI), X7 + PCMPEQB X1, X0 + PCMPEQB X3, X2 + PCMPEQB X5, X4 + PCMPEQB X7, X6 + PAND X2, X0 + PAND X6, X4 + PAND X4, X0 + PMOVMSKB X0, DX + ADDQ $64, SI + ADDQ $64, DI + SUBQ $64, BX + CMPL DX, $0xffff + JEQ hugeloop + XORQ AX, AX // return 0 + RET +#endif + + // 64 bytes at a time using ymm registers +hugeloop_avx2: + CMPQ BX, $64 + JB bigloop_avx2 + VMOVDQU (SI), Y0 + VMOVDQU (DI), Y1 + VMOVDQU 32(SI), Y2 + VMOVDQU 32(DI), Y3 + VPCMPEQB Y1, Y0, Y4 + VPCMPEQB Y2, Y3, Y5 + VPAND Y4, Y5, Y6 + VPMOVMSKB Y6, DX + ADDQ $64, SI + ADDQ $64, DI + SUBQ $64, BX + CMPL DX, $0xffffffff + JEQ hugeloop_avx2 + VZEROUPPER + XORQ AX, AX // return 0 + RET + +bigloop_avx2: + VZEROUPPER + + // 8 bytes at a time using 64-bit register +bigloop: + CMPQ BX, $8 + JBE leftover + MOVQ (SI), CX + MOVQ (DI), DX + ADDQ $8, SI + ADDQ $8, DI + SUBQ $8, BX + CMPQ CX, DX + JEQ bigloop + XORQ AX, AX // return 0 + RET + + // remaining 0-8 bytes +leftover: + MOVQ -8(SI)(BX*1), CX + MOVQ -8(DI)(BX*1), DX + CMPQ CX, DX + SETEQ AX + RET + +small: + CMPQ BX, $0 + JEQ equal + + LEAQ 0(BX*8), CX + NEGQ CX + + CMPB SI, $0xf8 + JA si_high + + // load at SI won't cross a page boundary. + MOVQ (SI), SI + JMP si_finish +si_high: + // address ends in 11111xxx. Load up to bytes we want, move to correct position. + MOVQ -8(SI)(BX*1), SI + SHRQ CX, SI +si_finish: + + // same for DI. + CMPB DI, $0xf8 + JA di_high + MOVQ (DI), DI + JMP di_finish +di_high: + MOVQ -8(DI)(BX*1), DI + SHRQ CX, DI +di_finish: + + SUBQ SI, DI + SHLQ CX, DI +equal: + SETEQ AX + RET diff --git a/src/internal/bytealg/equal_arm.s b/src/internal/bytealg/equal_arm.s new file mode 100644 index 0000000..a6c4369 --- /dev/null +++ b/src/internal/bytealg/equal_arm.s @@ -0,0 +1,91 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT|NOFRAME,$0-13 + MOVW a+0(FP), R0 + MOVW b+4(FP), R2 + CMP R0, R2 + B.EQ eq + MOVW size+8(FP), R1 + CMP $0, R1 + B.EQ eq // short path to handle 0-byte case + MOVW $ret+12(FP), R7 + B memeqbody<>(SB) +eq: + MOVW $1, R0 + MOVB R0, ret+12(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT|NOFRAME,$0-9 + MOVW a+0(FP), R0 + MOVW b+4(FP), R2 + CMP R0, R2 + B.EQ eq + MOVW 4(R7), R1 // compiler stores size at offset 4 in the closure + CMP $0, R1 + B.EQ eq // short path to handle 0-byte case + MOVW $ret+8(FP), R7 + B memeqbody<>(SB) +eq: + MOVW $1, R0 + MOVB R0, ret+8(FP) + RET + +// Input: +// R0: data of a +// R1: length +// R2: data of b +// R7: points to return value +// +// On exit: +// R4, R5 and R6 are clobbered +TEXT memeqbody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMP $1, R1 + B.EQ one // 1-byte special case for better performance + + CMP $4, R1 + ADD R0, R1 // R1 is the end of the range to compare + B.LT byte_loop // length < 4 + AND $3, R0, R6 + CMP $0, R6 + B.NE byte_loop // unaligned a, use byte-wise compare (TODO: try to align a) + AND $3, R2, R6 + CMP $0, R6 + B.NE byte_loop // unaligned b, use byte-wise compare + AND $0xfffffffc, R1, R6 + // length >= 4 +chunk4_loop: + MOVW.P 4(R0), R4 + MOVW.P 4(R2), R5 + CMP R4, R5 + B.NE notequal + CMP R0, R6 + B.NE chunk4_loop + CMP R0, R1 + B.EQ equal // reached the end +byte_loop: + MOVBU.P 1(R0), R4 + MOVBU.P 1(R2), R5 + CMP R4, R5 + B.NE notequal + CMP R0, R1 + B.NE byte_loop +equal: + MOVW $1, R0 + MOVB R0, (R7) + RET +one: + MOVBU (R0), R4 + MOVBU (R2), R5 + CMP R4, R5 + B.EQ equal +notequal: + MOVW $0, R0 + MOVB R0, (R7) + RET diff --git a/src/internal/bytealg/equal_arm64.s b/src/internal/bytealg/equal_arm64.s new file mode 100644 index 0000000..d3aabba --- /dev/null +++ b/src/internal/bytealg/equal_arm64.s @@ -0,0 +1,121 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-25 + // short path to handle 0-byte case + CBZ R2, equal + B memeqbody<>(SB) +equal: + MOVD $1, R0 + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT,$0-17 + CMP R0, R1 + BEQ eq + MOVD 8(R26), R2 // compiler stores size at offset 8 in the closure + CBZ R2, eq + B memeqbody<>(SB) +eq: + MOVD $1, R0 + RET + +// input: +// R0: pointer a +// R1: pointer b +// R2: data len +// at return: result in R0 +TEXT memeqbody<>(SB),NOSPLIT,$0 + CMP $1, R2 + // handle 1-byte special case for better performance + BEQ one + CMP $16, R2 + // handle specially if length < 16 + BLO tail + BIC $0x3f, R2, R3 + CBZ R3, chunk16 + // work with 64-byte chunks + ADD R3, R0, R6 // end of chunks +chunk64_loop: + VLD1.P (R0), [V0.D2, V1.D2, V2.D2, V3.D2] + VLD1.P (R1), [V4.D2, V5.D2, V6.D2, V7.D2] + VCMEQ V0.D2, V4.D2, V8.D2 + VCMEQ V1.D2, V5.D2, V9.D2 + VCMEQ V2.D2, V6.D2, V10.D2 + VCMEQ V3.D2, V7.D2, V11.D2 + VAND V8.B16, V9.B16, V8.B16 + VAND V8.B16, V10.B16, V8.B16 + VAND V8.B16, V11.B16, V8.B16 + CMP R0, R6 + VMOV V8.D[0], R4 + VMOV V8.D[1], R5 + CBZ R4, not_equal + CBZ R5, not_equal + BNE chunk64_loop + AND $0x3f, R2, R2 + CBZ R2, equal +chunk16: + // work with 16-byte chunks + BIC $0xf, R2, R3 + CBZ R3, tail + ADD R3, R0, R6 // end of chunks +chunk16_loop: + LDP.P 16(R0), (R4, R5) + LDP.P 16(R1), (R7, R9) + EOR R4, R7 + CBNZ R7, not_equal + EOR R5, R9 + CBNZ R9, not_equal + CMP R0, R6 + BNE chunk16_loop + AND $0xf, R2, R2 + CBZ R2, equal +tail: + // special compare of tail with length < 16 + TBZ $3, R2, lt_8 + MOVD (R0), R4 + MOVD (R1), R5 + EOR R4, R5 + CBNZ R5, not_equal + SUB $8, R2, R6 // offset of the last 8 bytes + MOVD (R0)(R6), R4 + MOVD (R1)(R6), R5 + EOR R4, R5 + CBNZ R5, not_equal + B equal +lt_8: + TBZ $2, R2, lt_4 + MOVWU (R0), R4 + MOVWU (R1), R5 + EOR R4, R5 + CBNZ R5, not_equal + SUB $4, R2, R6 // offset of the last 4 bytes + MOVWU (R0)(R6), R4 + MOVWU (R1)(R6), R5 + EOR R4, R5 + CBNZ R5, not_equal + B equal +lt_4: + TBZ $1, R2, lt_2 + MOVHU.P 2(R0), R4 + MOVHU.P 2(R1), R5 + CMP R4, R5 + BNE not_equal +lt_2: + TBZ $0, R2, equal +one: + MOVBU (R0), R4 + MOVBU (R1), R5 + CMP R4, R5 + BNE not_equal +equal: + MOVD $1, R0 + RET +not_equal: + MOVB ZR, R0 + RET diff --git a/src/internal/bytealg/equal_generic.go b/src/internal/bytealg/equal_generic.go new file mode 100644 index 0000000..59bdf8f --- /dev/null +++ b/src/internal/bytealg/equal_generic.go @@ -0,0 +1,18 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +// Equal reports whether a and b +// are the same length and contain the same bytes. +// A nil argument is equivalent to an empty slice. +// +// Equal is equivalent to bytes.Equal. +// It is provided here for convenience, +// because some packages cannot depend on bytes. +func Equal(a, b []byte) bool { + // Neither cmd/compile nor gccgo allocates for these string conversions. + // There is a test for this in package bytes. + return string(a) == string(b) +} diff --git a/src/internal/bytealg/equal_loong64.s b/src/internal/bytealg/equal_loong64.s new file mode 100644 index 0000000..ba2a557 --- /dev/null +++ b/src/internal/bytealg/equal_loong64.s @@ -0,0 +1,53 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +#define REGCTXT R29 + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT|NOFRAME,$0-25 + MOVV a+0(FP), R4 + MOVV b+8(FP), R5 + BEQ R4, R5, eq + MOVV size+16(FP), R6 + ADDV R4, R6, R7 + PCALIGN $16 +loop: + BNE R4, R7, test + MOVV $1, R4 + MOVB R4, ret+24(FP) + RET +test: + MOVBU (R4), R9 + ADDV $1, R4 + MOVBU (R5), R10 + ADDV $1, R5 + BEQ R9, R10, loop + + MOVB R0, ret+24(FP) + RET +eq: + MOVV $1, R4 + MOVB R4, ret+24(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT,$40-17 + MOVV a+0(FP), R4 + MOVV b+8(FP), R5 + BEQ R4, R5, eq + MOVV 8(REGCTXT), R6 // compiler stores size at offset 8 in the closure + MOVV R4, 8(R3) + MOVV R5, 16(R3) + MOVV R6, 24(R3) + JAL runtime·memequal(SB) + MOVBU 32(R3), R4 + MOVB R4, ret+16(FP) + RET +eq: + MOVV $1, R4 + MOVB R4, ret+16(FP) + RET diff --git a/src/internal/bytealg/equal_mips64x.s b/src/internal/bytealg/equal_mips64x.s new file mode 100644 index 0000000..d92f225 --- /dev/null +++ b/src/internal/bytealg/equal_mips64x.s @@ -0,0 +1,118 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips64 || mips64le + +#include "go_asm.h" +#include "textflag.h" + +#define REGCTXT R22 + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT|NOFRAME,$0-25 + MOVV a+0(FP), R1 + MOVV b+8(FP), R2 + BEQ R1, R2, eq + MOVV size+16(FP), R3 + ADDV R1, R3, R4 + + // chunk size is 16 + SGTU $16, R3, R8 + BEQ R0, R8, chunk_entry + +byte_loop: + BNE R1, R4, byte_test + MOVV $1, R1 + MOVB R1, ret+24(FP) + RET +byte_test: + MOVBU (R1), R6 + ADDV $1, R1 + MOVBU (R2), R7 + ADDV $1, R2 + BEQ R6, R7, byte_loop + JMP not_eq + +chunk_entry: + // make sure both a and b are aligned + OR R1, R2, R9 + AND $0x7, R9 + BNE R0, R9, byte_loop + JMP chunk_loop_1 + +chunk_loop: + // chunk size is 16 + SGTU $16, R3, R8 + BNE R0, R8, chunk_tail_8 +chunk_loop_1: + MOVV (R1), R6 + MOVV (R2), R7 + BNE R6, R7, not_eq + MOVV 8(R1), R12 + MOVV 8(R2), R13 + ADDV $16, R1 + ADDV $16, R2 + SUBV $16, R3 + BEQ R12, R13, chunk_loop + JMP not_eq + +chunk_tail_8: + AND $8, R3, R14 + BEQ R0, R14, chunk_tail_4 + MOVV (R1), R6 + MOVV (R2), R7 + BNE R6, R7, not_eq + ADDV $8, R1 + ADDV $8, R2 + +chunk_tail_4: + AND $4, R3, R14 + BEQ R0, R14, chunk_tail_2 + MOVWU (R1), R6 + MOVWU (R2), R7 + BNE R6, R7, not_eq + ADDV $4, R1 + ADDV $4, R2 + +chunk_tail_2: + AND $2, R3, R14 + BEQ R0, R14, chunk_tail_1 + MOVHU (R1), R6 + MOVHU (R2), R7 + BNE R6, R7, not_eq + ADDV $2, R1 + ADDV $2, R2 + +chunk_tail_1: + AND $1, R3, R14 + BEQ R0, R14, eq + MOVBU (R1), R6 + MOVBU (R2), R7 + BEQ R6, R7, eq + +not_eq: + MOVB R0, ret+24(FP) + RET +eq: + MOVV $1, R1 + MOVB R1, ret+24(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT,$40-17 + MOVV a+0(FP), R1 + MOVV b+8(FP), R2 + BEQ R1, R2, eq + MOVV 8(REGCTXT), R3 // compiler stores size at offset 8 in the closure + MOVV R1, 8(R29) + MOVV R2, 16(R29) + MOVV R3, 24(R29) + JAL runtime·memequal(SB) + MOVBU 32(R29), R1 + MOVB R1, ret+16(FP) + RET +eq: + MOVV $1, R1 + MOVB R1, ret+16(FP) + RET diff --git a/src/internal/bytealg/equal_mipsx.s b/src/internal/bytealg/equal_mipsx.s new file mode 100644 index 0000000..4c46dd4 --- /dev/null +++ b/src/internal/bytealg/equal_mipsx.s @@ -0,0 +1,62 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips || mipsle + +#include "go_asm.h" +#include "textflag.h" + +#define REGCTXT R22 + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT,$0-13 + MOVW a+0(FP), R1 + MOVW b+4(FP), R2 + BEQ R1, R2, eq + MOVW size+8(FP), R3 + ADDU R1, R3, R4 +loop: + BNE R1, R4, test + MOVW $1, R1 + MOVB R1, ret+12(FP) + RET +test: + MOVBU (R1), R6 + ADDU $1, R1 + MOVBU (R2), R7 + ADDU $1, R2 + BEQ R6, R7, loop + + MOVB R0, ret+12(FP) + RET +eq: + MOVW $1, R1 + MOVB R1, ret+12(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT,$0-9 + MOVW a+0(FP), R1 + MOVW b+4(FP), R2 + BEQ R1, R2, eq + MOVW 4(REGCTXT), R3 // compiler stores size at offset 4 in the closure + ADDU R1, R3, R4 +loop: + BNE R1, R4, test + MOVW $1, R1 + MOVB R1, ret+8(FP) + RET +test: + MOVBU (R1), R6 + ADDU $1, R1 + MOVBU (R2), R7 + ADDU $1, R2 + BEQ R6, R7, loop + + MOVB R0, ret+8(FP) + RET +eq: + MOVW $1, R1 + MOVB R1, ret+8(FP) + RET diff --git a/src/internal/bytealg/equal_native.go b/src/internal/bytealg/equal_native.go new file mode 100644 index 0000000..cf3a245 --- /dev/null +++ b/src/internal/bytealg/equal_native.go @@ -0,0 +1,21 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +import "unsafe" + +// The declarations below generate ABI wrappers for functions +// implemented in assembly in this package but declared in another +// package. + +// The compiler generates calls to runtime.memequal and runtime.memequal_varlen. +// In addition, the runtime calls runtime.memequal explicitly. +// Those functions are implemented in this package. + +//go:linkname abigen_runtime_memequal runtime.memequal +func abigen_runtime_memequal(a, b unsafe.Pointer, size uintptr) bool + +//go:linkname abigen_runtime_memequal_varlen runtime.memequal_varlen +func abigen_runtime_memequal_varlen(a, b unsafe.Pointer) bool diff --git a/src/internal/bytealg/equal_ppc64x.s b/src/internal/bytealg/equal_ppc64x.s new file mode 100644 index 0000000..07dce80 --- /dev/null +++ b/src/internal/bytealg/equal_ppc64x.s @@ -0,0 +1,207 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +#include "go_asm.h" +#include "textflag.h" + +// 4K (smallest case) page size offset mask for PPC64. +#define PAGE_OFFSET 4095 + +// Likewise, the BC opcode is hard to read, and no extended +// mnemonics are offered for these forms. +#define BGELR_CR6 BC 4, CR6LT, (LR) +#define BEQLR BC 12, CR0EQ, (LR) + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-25 + // R3 = a + // R4 = b + // R5 = size + BR memeqbody<>(SB) + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-17 + // R3 = a + // R4 = b + CMP R3, R4 + BEQ eq + MOVD 8(R11), R5 // compiler stores size at offset 8 in the closure + BR memeqbody<>(SB) +eq: + MOVD $1, R3 + RET + +// Do an efficient memequal for ppc64 +// R3 = s1 +// R4 = s2 +// R5 = len +// On exit: +// R3 = return value +TEXT memeqbody<>(SB),NOSPLIT|NOFRAME,$0-0 + MOVD R3, R8 // Move s1 into R8 + ADD R5, R3, R9 // &s1[len(s1)] + ADD R5, R4, R10 // &s2[len(s2)] + MOVD $1, R11 + CMP R5, $16 // Use GPR checks for check for len <= 16 + BLE check0_16 + MOVD $0, R3 // Assume no-match in case BGELR CR6 returns + CMP R5, $32 // Use overlapping VSX loads for len <= 32 + BLE check17_32 // Do a pair of overlapping VSR compares + CMP R5, $64 + BLE check33_64 // Hybrid check + overlap compare. + +setup64: + SRD $6, R5, R6 // number of 64 byte chunks to compare + MOVD R6, CTR + MOVD $16, R14 // index for VSX loads and stores + MOVD $32, R15 + MOVD $48, R16 + ANDCC $0x3F, R5, R5 // len%64==0? + + PCALIGN $16 +loop64: + LXVD2X (R8+R0), V0 + LXVD2X (R4+R0), V1 + VCMPEQUBCC V0, V1, V2 // compare, setting CR6 + BGELR_CR6 + LXVD2X (R8+R14), V0 + LXVD2X (R4+R14), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + LXVD2X (R8+R15), V0 + LXVD2X (R4+R15), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + LXVD2X (R8+R16), V0 + LXVD2X (R4+R16), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + ADD $64,R8 // bump up to next 64 + ADD $64,R4 + BDNZ loop64 + + ISEL CR0EQ, R11, R3, R3 // If no tail, return 1, otherwise R3 remains 0. + BEQLR // return if no tail. + + ADD $-64, R9, R8 + ADD $-64, R10, R4 + LXVD2X (R8+R0), V0 + LXVD2X (R4+R0), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + LXVD2X (R8+R14), V0 + LXVD2X (R4+R14), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + LXVD2X (R8+R15), V0 + LXVD2X (R4+R15), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + LXVD2X (R8+R16), V0 + LXVD2X (R4+R16), V1 + VCMPEQUBCC V0, V1, V2 + ISEL CR6LT, R11, R0, R3 + RET + +check33_64: + // Bytes 0-15 + LXVD2X (R8+R0), V0 + LXVD2X (R4+R0), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + ADD $16, R8 + ADD $16, R4 + + // Bytes 16-31 + LXVD2X (R8+R0), V0 + LXVD2X (R4+R0), V1 + VCMPEQUBCC V0, V1, V2 + BGELR_CR6 + + // A little tricky, but point R4,R8 to &sx[len-32], + // and reuse check17_32 to check the next 1-31 bytes (with some overlap) + ADD $-32, R9, R8 + ADD $-32, R10, R4 + // Fallthrough + +check17_32: + LXVD2X (R8+R0), V0 + LXVD2X (R4+R0), V1 + VCMPEQUBCC V0, V1, V2 + ISEL CR6LT, R11, R0, R5 + + // Load sX[len(sX)-16:len(sX)] and compare. + ADD $-16, R9 + ADD $-16, R10 + LXVD2X (R9+R0), V0 + LXVD2X (R10+R0), V1 + VCMPEQUBCC V0, V1, V2 + ISEL CR6LT, R5, R0, R3 + RET + +check0_16: +#ifdef GOPPC64_power10 + SLD $56, R5, R7 + LXVL R8, R7, V0 + LXVL R4, R7, V1 + VCMPEQUDCC V0, V1, V2 + ISEL CR6LT, R11, R0, R3 + RET +#else + CMP R5, $8 + BLT check0_7 + // Load sX[0:7] and compare. + MOVD (R8), R6 + MOVD (R4), R7 + CMP R6, R7 + ISEL CR0EQ, R11, R0, R5 + // Load sX[len(sX)-8:len(sX)] and compare. + MOVD -8(R9), R6 + MOVD -8(R10), R7 + CMP R6, R7 + ISEL CR0EQ, R5, R0, R3 + RET + +check0_7: + CMP R5,$0 + MOVD $1, R3 + BEQLR // return if len == 0 + + // Check < 8B loads with a single compare, but select the load address + // such that it cannot cross a page boundary. Load a few bytes from the + // lower address if that does not cross the lower page. Or, load a few + // extra bytes from the higher addresses. And align those values + // consistently in register as either address may have differing + // alignment requirements. + ANDCC $PAGE_OFFSET, R8, R6 // &sX & PAGE_OFFSET + ANDCC $PAGE_OFFSET, R4, R9 + SUBC R5, $8, R12 // 8-len + SLD $3, R12, R14 // (8-len)*8 + CMPU R6, R12, CR1 // Enough bytes lower in the page to load lower? + CMPU R9, R12, CR0 + SUB R12, R8, R6 // compute lower load address + SUB R12, R4, R9 + ISEL CR1LT, R8, R6, R8 // R8 = R6 < 0 ? R8 (&s1) : R6 (&s1 - (8-len)) + ISEL CR0LT, R4, R9, R4 // Similar for s2 + MOVD (R8), R15 + MOVD (R4), R16 + SLD R14, R15, R7 + SLD R14, R16, R17 + SRD R14, R7, R7 // Clear the upper (8-len) bytes (with 2 shifts) + SRD R14, R17, R17 + SRD R14, R15, R6 // Clear the lower (8-len) bytes + SRD R14, R16, R9 +#ifdef GOARCH_ppc64le + ISEL CR1LT, R7, R6, R8 // Choose the correct len bytes to compare based on alignment + ISEL CR0LT, R17, R9, R4 +#else + ISEL CR1LT, R6, R7, R8 + ISEL CR0LT, R9, R17, R4 +#endif + CMP R4, R8 + ISEL CR0EQ, R11, R0, R3 + RET +#endif // tail processing if !defined(GOPPC64_power10) diff --git a/src/internal/bytealg/equal_riscv64.s b/src/internal/bytealg/equal_riscv64.s new file mode 100644 index 0000000..503aac5 --- /dev/null +++ b/src/internal/bytealg/equal_riscv64.s @@ -0,0 +1,126 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +#define CTXT S10 + +// func memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-25 + // X10 = a_base + // X11 = b_base + // X12 = size + JMP memequal<>(SB) + +// func memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-17 + MOV 8(CTXT), X12 // compiler stores size at offset 8 in the closure + // X10 = a_base + // X11 = b_base + JMP memequal<>(SB) + +// On entry X10 and X11 contain pointers, X12 contains length. +// For non-regabi X13 contains address for return value. +// For regabi return value in X10. +TEXT memequal<>(SB),NOSPLIT|NOFRAME,$0 + BEQ X10, X11, eq + + MOV $32, X23 + BLT X12, X23, loop4_check + + // Check alignment - if alignment differs we have to do one byte at a time. + AND $7, X10, X9 + AND $7, X11, X19 + BNE X9, X19, loop4_check + BEQZ X9, loop32_check + + // Check one byte at a time until we reach 8 byte alignment. + SUB X9, X0, X9 + ADD $8, X9, X9 + SUB X9, X12, X12 +align: + ADD $-1, X9 + MOVBU 0(X10), X19 + MOVBU 0(X11), X20 + BNE X19, X20, not_eq + ADD $1, X10 + ADD $1, X11 + BNEZ X9, align + +loop32_check: + MOV $32, X9 + BLT X12, X9, loop16_check +loop32: + MOV 0(X10), X19 + MOV 0(X11), X20 + MOV 8(X10), X21 + MOV 8(X11), X22 + BNE X19, X20, not_eq + BNE X21, X22, not_eq + MOV 16(X10), X14 + MOV 16(X11), X15 + MOV 24(X10), X16 + MOV 24(X11), X17 + BNE X14, X15, not_eq + BNE X16, X17, not_eq + ADD $32, X10 + ADD $32, X11 + ADD $-32, X12 + BGE X12, X9, loop32 + BEQZ X12, eq + +loop16_check: + MOV $16, X23 + BLT X12, X23, loop4_check +loop16: + MOV 0(X10), X19 + MOV 0(X11), X20 + MOV 8(X10), X21 + MOV 8(X11), X22 + BNE X19, X20, not_eq + BNE X21, X22, not_eq + ADD $16, X10 + ADD $16, X11 + ADD $-16, X12 + BGE X12, X23, loop16 + BEQZ X12, eq + +loop4_check: + MOV $4, X23 + BLT X12, X23, loop1 +loop4: + MOVBU 0(X10), X19 + MOVBU 0(X11), X20 + MOVBU 1(X10), X21 + MOVBU 1(X11), X22 + BNE X19, X20, not_eq + BNE X21, X22, not_eq + MOVBU 2(X10), X14 + MOVBU 2(X11), X15 + MOVBU 3(X10), X16 + MOVBU 3(X11), X17 + BNE X14, X15, not_eq + BNE X16, X17, not_eq + ADD $4, X10 + ADD $4, X11 + ADD $-4, X12 + BGE X12, X23, loop4 + +loop1: + BEQZ X12, eq + MOVBU 0(X10), X19 + MOVBU 0(X11), X20 + BNE X19, X20, not_eq + ADD $1, X10 + ADD $1, X11 + ADD $-1, X12 + JMP loop1 + +not_eq: + MOVB ZERO, X10 + RET +eq: + MOV $1, X10 + RET diff --git a/src/internal/bytealg/equal_s390x.s b/src/internal/bytealg/equal_s390x.s new file mode 100644 index 0000000..67f814d --- /dev/null +++ b/src/internal/bytealg/equal_s390x.s @@ -0,0 +1,92 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// memequal(a, b unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB),NOSPLIT|NOFRAME,$0-25 + MOVD a+0(FP), R3 + MOVD b+8(FP), R5 + MOVD size+16(FP), R6 + LA ret+24(FP), R7 + BR memeqbody<>(SB) + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB),NOSPLIT|NOFRAME,$0-17 + MOVD a+0(FP), R3 + MOVD b+8(FP), R5 + MOVD 8(R12), R6 // compiler stores size at offset 8 in the closure + LA ret+16(FP), R7 + BR memeqbody<>(SB) + +// input: +// R3 = a +// R5 = b +// R6 = len +// R7 = address of output byte (stores 0 or 1 here) +// a and b have the same length +TEXT memeqbody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMPBEQ R3, R5, equal +loop: + CMPBEQ R6, $0, equal + CMPBLT R6, $32, tiny + CMP R6, $256 + BLT tail + CLC $256, 0(R3), 0(R5) + BNE notequal + SUB $256, R6 + LA 256(R3), R3 + LA 256(R5), R5 + BR loop +tail: + SUB $1, R6, R8 + EXRL $memeqbodyclc<>(SB), R8 + BEQ equal +notequal: + MOVB $0, 0(R7) + RET +equal: + MOVB $1, 0(R7) + RET +tiny: + MOVD $0, R2 + CMPBLT R6, $16, lt16 + MOVD 0(R3), R8 + MOVD 0(R5), R9 + CMPBNE R8, R9, notequal + MOVD 8(R3), R8 + MOVD 8(R5), R9 + CMPBNE R8, R9, notequal + LA 16(R2), R2 + SUB $16, R6 +lt16: + CMPBLT R6, $8, lt8 + MOVD 0(R3)(R2*1), R8 + MOVD 0(R5)(R2*1), R9 + CMPBNE R8, R9, notequal + LA 8(R2), R2 + SUB $8, R6 +lt8: + CMPBLT R6, $4, lt4 + MOVWZ 0(R3)(R2*1), R8 + MOVWZ 0(R5)(R2*1), R9 + CMPBNE R8, R9, notequal + LA 4(R2), R2 + SUB $4, R6 +lt4: +#define CHECK(n) \ + CMPBEQ R6, $n, equal \ + MOVB n(R3)(R2*1), R8 \ + MOVB n(R5)(R2*1), R9 \ + CMPBNE R8, R9, notequal + CHECK(0) + CHECK(1) + CHECK(2) + CHECK(3) + BR equal + +TEXT memeqbodyclc<>(SB),NOSPLIT|NOFRAME,$0-0 + CLC $1, 0(R3), 0(R5) + RET diff --git a/src/internal/bytealg/equal_wasm.s b/src/internal/bytealg/equal_wasm.s new file mode 100644 index 0000000..a2b76c1 --- /dev/null +++ b/src/internal/bytealg/equal_wasm.s @@ -0,0 +1,77 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// memequal(p, q unsafe.Pointer, size uintptr) bool +TEXT runtime·memequal(SB), NOSPLIT, $0-25 + Get SP + I64Load a+0(FP) + I64Load b+8(FP) + I64Load size+16(FP) + Call memeqbody<>(SB) + I64Store8 ret+24(FP) + RET + +// memequal_varlen(a, b unsafe.Pointer) bool +TEXT runtime·memequal_varlen(SB), NOSPLIT, $0-17 + Get SP + I64Load a+0(FP) + I64Load b+8(FP) + I64Load 8(CTXT) // compiler stores size at offset 8 in the closure + Call memeqbody<>(SB) + I64Store8 ret+16(FP) + RET + +// params: a, b, len +// ret: 0/1 +TEXT memeqbody<>(SB), NOSPLIT, $0-0 + Get R0 + Get R1 + I64Eq + If + I64Const $1 + Return + End + +loop: + Loop + Get R2 + I64Eqz + If + I64Const $1 + Return + End + + Get R0 + I32WrapI64 + I64Load8U $0 + Get R1 + I32WrapI64 + I64Load8U $0 + I64Ne + If + I64Const $0 + Return + End + + Get R0 + I64Const $1 + I64Add + Set R0 + + Get R1 + I64Const $1 + I64Add + Set R1 + + Get R2 + I64Const $1 + I64Sub + Set R2 + + Br loop + End + UNDEF diff --git a/src/internal/bytealg/index_amd64.go b/src/internal/bytealg/index_amd64.go new file mode 100644 index 0000000..c7a1941 --- /dev/null +++ b/src/internal/bytealg/index_amd64.go @@ -0,0 +1,26 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +import "internal/cpu" + +const MaxBruteForce = 64 + +func init() { + if cpu.X86.HasAVX2 { + MaxLen = 63 + } else { + MaxLen = 31 + } +} + +// Cutover reports the number of failures of IndexByte we should tolerate +// before switching over to Index. +// n is the number of bytes processed so far. +// See the bytes.Index implementation for details. +func Cutover(n int) int { + // 1 error per 8 characters, plus a few slop to start. + return (n + 16) / 8 +} diff --git a/src/internal/bytealg/index_amd64.s b/src/internal/bytealg/index_amd64.s new file mode 100644 index 0000000..0431491 --- /dev/null +++ b/src/internal/bytealg/index_amd64.s @@ -0,0 +1,276 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Index(SB),NOSPLIT,$0-56 + MOVQ a_base+0(FP), DI + MOVQ a_len+8(FP), DX + MOVQ b_base+24(FP), R8 + MOVQ b_len+32(FP), AX + MOVQ DI, R10 + LEAQ ret+48(FP), R11 + JMP indexbody<>(SB) + +TEXT ·IndexString(SB),NOSPLIT,$0-40 + MOVQ a_base+0(FP), DI + MOVQ a_len+8(FP), DX + MOVQ b_base+16(FP), R8 + MOVQ b_len+24(FP), AX + MOVQ DI, R10 + LEAQ ret+32(FP), R11 + JMP indexbody<>(SB) + +// AX: length of string, that we are searching for +// DX: length of string, in which we are searching +// DI: pointer to string, in which we are searching +// R8: pointer to string, that we are searching for +// R11: address, where to put return value +// Note: We want len in DX and AX, because PCMPESTRI implicitly consumes them +TEXT indexbody<>(SB),NOSPLIT,$0 + CMPQ AX, DX + JA fail + CMPQ DX, $16 + JAE sse42 +no_sse42: + CMPQ AX, $2 + JA _3_or_more + MOVW (R8), R8 + LEAQ -1(DI)(DX*1), DX +loop2: + MOVW (DI), SI + CMPW SI,R8 + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop2 + JMP fail +_3_or_more: + CMPQ AX, $3 + JA _4_or_more + MOVW 1(R8), BX + MOVW (R8), R8 + LEAQ -2(DI)(DX*1), DX +loop3: + MOVW (DI), SI + CMPW SI,R8 + JZ partial_success3 + ADDQ $1,DI + CMPQ DI,DX + JB loop3 + JMP fail +partial_success3: + MOVW 1(DI), SI + CMPW SI,BX + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop3 + JMP fail +_4_or_more: + CMPQ AX, $4 + JA _5_or_more + MOVL (R8), R8 + LEAQ -3(DI)(DX*1), DX +loop4: + MOVL (DI), SI + CMPL SI,R8 + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop4 + JMP fail +_5_or_more: + CMPQ AX, $7 + JA _8_or_more + LEAQ 1(DI)(DX*1), DX + SUBQ AX, DX + MOVL -4(R8)(AX*1), BX + MOVL (R8), R8 +loop5to7: + MOVL (DI), SI + CMPL SI,R8 + JZ partial_success5to7 + ADDQ $1,DI + CMPQ DI,DX + JB loop5to7 + JMP fail +partial_success5to7: + MOVL -4(AX)(DI*1), SI + CMPL SI,BX + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop5to7 + JMP fail +_8_or_more: + CMPQ AX, $8 + JA _9_or_more + MOVQ (R8), R8 + LEAQ -7(DI)(DX*1), DX +loop8: + MOVQ (DI), SI + CMPQ SI,R8 + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop8 + JMP fail +_9_or_more: + CMPQ AX, $15 + JA _16_or_more + LEAQ 1(DI)(DX*1), DX + SUBQ AX, DX + MOVQ -8(R8)(AX*1), BX + MOVQ (R8), R8 +loop9to15: + MOVQ (DI), SI + CMPQ SI,R8 + JZ partial_success9to15 + ADDQ $1,DI + CMPQ DI,DX + JB loop9to15 + JMP fail +partial_success9to15: + MOVQ -8(AX)(DI*1), SI + CMPQ SI,BX + JZ success + ADDQ $1,DI + CMPQ DI,DX + JB loop9to15 + JMP fail +_16_or_more: + CMPQ AX, $16 + JA _17_or_more + MOVOU (R8), X1 + LEAQ -15(DI)(DX*1), DX +loop16: + MOVOU (DI), X2 + PCMPEQB X1, X2 + PMOVMSKB X2, SI + CMPQ SI, $0xffff + JE success + ADDQ $1,DI + CMPQ DI,DX + JB loop16 + JMP fail +_17_or_more: + CMPQ AX, $31 + JA _32_or_more + LEAQ 1(DI)(DX*1), DX + SUBQ AX, DX + MOVOU -16(R8)(AX*1), X0 + MOVOU (R8), X1 +loop17to31: + MOVOU (DI), X2 + PCMPEQB X1,X2 + PMOVMSKB X2, SI + CMPQ SI, $0xffff + JE partial_success17to31 + ADDQ $1,DI + CMPQ DI,DX + JB loop17to31 + JMP fail +partial_success17to31: + MOVOU -16(AX)(DI*1), X3 + PCMPEQB X0, X3 + PMOVMSKB X3, SI + CMPQ SI, $0xffff + JE success + ADDQ $1,DI + CMPQ DI,DX + JB loop17to31 + JMP fail +// We can get here only when AVX2 is enabled and cutoff for indexShortStr is set to 63 +// So no need to check cpuid +_32_or_more: + CMPQ AX, $32 + JA _33_to_63 + VMOVDQU (R8), Y1 + LEAQ -31(DI)(DX*1), DX +loop32: + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPMOVMSKB Y3, SI + CMPL SI, $0xffffffff + JE success_avx2 + ADDQ $1,DI + CMPQ DI,DX + JB loop32 + JMP fail_avx2 +_33_to_63: + LEAQ 1(DI)(DX*1), DX + SUBQ AX, DX + VMOVDQU -32(R8)(AX*1), Y0 + VMOVDQU (R8), Y1 +loop33to63: + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPMOVMSKB Y3, SI + CMPL SI, $0xffffffff + JE partial_success33to63 + ADDQ $1,DI + CMPQ DI,DX + JB loop33to63 + JMP fail_avx2 +partial_success33to63: + VMOVDQU -32(AX)(DI*1), Y3 + VPCMPEQB Y0, Y3, Y4 + VPMOVMSKB Y4, SI + CMPL SI, $0xffffffff + JE success_avx2 + ADDQ $1,DI + CMPQ DI,DX + JB loop33to63 +fail_avx2: + VZEROUPPER +fail: + MOVQ $-1, (R11) + RET +success_avx2: + VZEROUPPER + JMP success +sse42: +#ifndef hasSSE42 + CMPB internal∕cpu·X86+const_offsetX86HasSSE42(SB), $1 + JNE no_sse42 +#endif + CMPQ AX, $12 + // PCMPESTRI is slower than normal compare, + // so using it makes sense only if we advance 4+ bytes per compare + // This value was determined experimentally and is the ~same + // on Nehalem (first with SSE42) and Haswell. + JAE _9_or_more + LEAQ 16(R8), SI + TESTW $0xff0, SI + JEQ no_sse42 + MOVOU (R8), X1 + LEAQ -15(DI)(DX*1), SI + MOVQ $16, R9 + SUBQ AX, R9 // We advance by 16-len(sep) each iteration, so precalculate it into R9 +loop_sse42: + // 0x0c means: unsigned byte compare (bits 0,1 are 00) + // for equality (bits 2,3 are 11) + // result is not masked or inverted (bits 4,5 are 00) + // and corresponds to first matching byte (bit 6 is 0) + PCMPESTRI $0x0c, (DI), X1 + // CX == 16 means no match, + // CX > R9 means partial match at the end of the string, + // otherwise sep is at offset CX from X1 start + CMPQ CX, R9 + JBE sse42_success + ADDQ R9, DI + CMPQ DI, SI + JB loop_sse42 + PCMPESTRI $0x0c, -1(SI), X1 + CMPQ CX, R9 + JA fail + LEAQ -1(SI), DI +sse42_success: + ADDQ CX, DI +success: + SUBQ R10, DI + MOVQ DI, (R11) + RET diff --git a/src/internal/bytealg/index_arm64.go b/src/internal/bytealg/index_arm64.go new file mode 100644 index 0000000..e87c109 --- /dev/null +++ b/src/internal/bytealg/index_arm64.go @@ -0,0 +1,23 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +// Empirical data shows that using Index can get better +// performance when len(s) <= 16. +const MaxBruteForce = 16 + +func init() { + // Optimize cases where the length of the substring is less than 32 bytes + MaxLen = 32 +} + +// Cutover reports the number of failures of IndexByte we should tolerate +// before switching over to Index. +// n is the number of bytes processed so far. +// See the bytes.Index implementation for details. +func Cutover(n int) int { + // 1 error per 16 characters, plus a few slop to start. + return 4 + n>>4 +} diff --git a/src/internal/bytealg/index_arm64.s b/src/internal/bytealg/index_arm64.s new file mode 100644 index 0000000..3a551a7 --- /dev/null +++ b/src/internal/bytealg/index_arm64.s @@ -0,0 +1,206 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·Index(SB),NOSPLIT,$0-56 + MOVD a_base+0(FP), R0 + MOVD a_len+8(FP), R1 + MOVD b_base+24(FP), R2 + MOVD b_len+32(FP), R3 + MOVD $ret+48(FP), R9 + B indexbody<>(SB) + +TEXT ·IndexString(SB),NOSPLIT,$0-40 + MOVD a_base+0(FP), R0 + MOVD a_len+8(FP), R1 + MOVD b_base+16(FP), R2 + MOVD b_len+24(FP), R3 + MOVD $ret+32(FP), R9 + B indexbody<>(SB) + +// input: +// R0: haystack +// R1: length of haystack +// R2: needle +// R3: length of needle (2 <= len <= 32) +// R9: address to put result +TEXT indexbody<>(SB),NOSPLIT,$0-56 + // main idea is to load 'sep' into separate register(s) + // to avoid repeatedly re-load it again and again + // for sebsequent substring comparisons + SUB R3, R1, R4 + // R4 contains the start of last substring for comparison + ADD R0, R4, R4 + ADD $1, R0, R8 + + CMP $8, R3 + BHI greater_8 + TBZ $3, R3, len_2_7 +len_8: + // R5 contains 8-byte of sep + MOVD (R2), R5 +loop_8: + // R6 contains substring for comparison + CMP R4, R0 + BHI not_found + MOVD.P 1(R0), R6 + CMP R5, R6 + BNE loop_8 + B found +len_2_7: + TBZ $2, R3, len_2_3 + TBZ $1, R3, len_4_5 + TBZ $0, R3, len_6 +len_7: + // R5 and R6 contain 7-byte of sep + MOVWU (R2), R5 + // 1-byte overlap with R5 + MOVWU 3(R2), R6 +loop_7: + CMP R4, R0 + BHI not_found + MOVWU.P 1(R0), R3 + CMP R5, R3 + BNE loop_7 + MOVWU 2(R0), R3 + CMP R6, R3 + BNE loop_7 + B found +len_6: + // R5 and R6 contain 6-byte of sep + MOVWU (R2), R5 + MOVHU 4(R2), R6 +loop_6: + CMP R4, R0 + BHI not_found + MOVWU.P 1(R0), R3 + CMP R5, R3 + BNE loop_6 + MOVHU 3(R0), R3 + CMP R6, R3 + BNE loop_6 + B found +len_4_5: + TBZ $0, R3, len_4 +len_5: + // R5 and R7 contain 5-byte of sep + MOVWU (R2), R5 + MOVBU 4(R2), R7 +loop_5: + CMP R4, R0 + BHI not_found + MOVWU.P 1(R0), R3 + CMP R5, R3 + BNE loop_5 + MOVBU 3(R0), R3 + CMP R7, R3 + BNE loop_5 + B found +len_4: + // R5 contains 4-byte of sep + MOVWU (R2), R5 +loop_4: + CMP R4, R0 + BHI not_found + MOVWU.P 1(R0), R6 + CMP R5, R6 + BNE loop_4 + B found +len_2_3: + TBZ $0, R3, len_2 +len_3: + // R6 and R7 contain 3-byte of sep + MOVHU (R2), R6 + MOVBU 2(R2), R7 +loop_3: + CMP R4, R0 + BHI not_found + MOVHU.P 1(R0), R3 + CMP R6, R3 + BNE loop_3 + MOVBU 1(R0), R3 + CMP R7, R3 + BNE loop_3 + B found +len_2: + // R5 contains 2-byte of sep + MOVHU (R2), R5 +loop_2: + CMP R4, R0 + BHI not_found + MOVHU.P 1(R0), R6 + CMP R5, R6 + BNE loop_2 +found: + SUB R8, R0, R0 + MOVD R0, (R9) + RET +not_found: + MOVD $-1, R0 + MOVD R0, (R9) + RET +greater_8: + SUB $9, R3, R11 // len(sep) - 9, offset of R0 for last 8 bytes + CMP $16, R3 + BHI greater_16 +len_9_16: + MOVD.P 8(R2), R5 // R5 contains the first 8-byte of sep + SUB $16, R3, R7 // len(sep) - 16, offset of R2 for last 8 bytes + MOVD (R2)(R7), R6 // R6 contains the last 8-byte of sep +loop_9_16: + // search the first 8 bytes first + CMP R4, R0 + BHI not_found + MOVD.P 1(R0), R7 + CMP R5, R7 + BNE loop_9_16 + MOVD (R0)(R11), R7 + CMP R6, R7 // compare the last 8 bytes + BNE loop_9_16 + B found +greater_16: + CMP $24, R3 + BHI len_25_32 +len_17_24: + LDP.P 16(R2), (R5, R6) // R5 and R6 contain the first 16-byte of sep + SUB $24, R3, R10 // len(sep) - 24 + MOVD (R2)(R10), R7 // R7 contains the last 8-byte of sep +loop_17_24: + // search the first 16 bytes first + CMP R4, R0 + BHI not_found + MOVD.P 1(R0), R10 + CMP R5, R10 + BNE loop_17_24 + MOVD 7(R0), R10 + CMP R6, R10 + BNE loop_17_24 + MOVD (R0)(R11), R10 + CMP R7, R10 // compare the last 8 bytes + BNE loop_17_24 + B found +len_25_32: + LDP.P 16(R2), (R5, R6) + MOVD.P 8(R2), R7 // R5, R6 and R7 contain the first 24-byte of sep + SUB $32, R3, R12 // len(sep) - 32 + MOVD (R2)(R12), R10 // R10 contains the last 8-byte of sep +loop_25_32: + // search the first 24 bytes first + CMP R4, R0 + BHI not_found + MOVD.P 1(R0), R12 + CMP R5, R12 + BNE loop_25_32 + MOVD 7(R0), R12 + CMP R6, R12 + BNE loop_25_32 + MOVD 15(R0), R12 + CMP R7, R12 + BNE loop_25_32 + MOVD (R0)(R11), R12 + CMP R10, R12 // compare the last 8 bytes + BNE loop_25_32 + B found diff --git a/src/internal/bytealg/index_generic.go b/src/internal/bytealg/index_generic.go new file mode 100644 index 0000000..a59e329 --- /dev/null +++ b/src/internal/bytealg/index_generic.go @@ -0,0 +1,29 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !amd64 && !arm64 && !s390x && !ppc64le && !ppc64 + +package bytealg + +const MaxBruteForce = 0 + +// Index returns the index of the first instance of b in a, or -1 if b is not present in a. +// Requires 2 <= len(b) <= MaxLen. +func Index(a, b []byte) int { + panic("unimplemented") +} + +// IndexString returns the index of the first instance of b in a, or -1 if b is not present in a. +// Requires 2 <= len(b) <= MaxLen. +func IndexString(a, b string) int { + panic("unimplemented") +} + +// Cutover reports the number of failures of IndexByte we should tolerate +// before switching over to Index. +// n is the number of bytes processed so far. +// See the bytes.Index implementation for details. +func Cutover(n int) int { + panic("unimplemented") +} diff --git a/src/internal/bytealg/index_native.go b/src/internal/bytealg/index_native.go new file mode 100644 index 0000000..59c93f9 --- /dev/null +++ b/src/internal/bytealg/index_native.go @@ -0,0 +1,19 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build amd64 || arm64 || s390x || ppc64le || ppc64 + +package bytealg + +// Index returns the index of the first instance of b in a, or -1 if b is not present in a. +// Requires 2 <= len(b) <= MaxLen. +// +//go:noescape +func Index(a, b []byte) int + +// IndexString returns the index of the first instance of b in a, or -1 if b is not present in a. +// Requires 2 <= len(b) <= MaxLen. +// +//go:noescape +func IndexString(a, b string) int diff --git a/src/internal/bytealg/index_ppc64x.go b/src/internal/bytealg/index_ppc64x.go new file mode 100644 index 0000000..720d517 --- /dev/null +++ b/src/internal/bytealg/index_ppc64x.go @@ -0,0 +1,26 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package bytealg + +import "internal/cpu" + +const MaxBruteForce = 16 + +var SupportsPower9 = cpu.PPC64.IsPOWER9 + +func init() { + MaxLen = 32 +} + +// Cutover reports the number of failures of IndexByte we should tolerate +// before switching over to Index. +// n is the number of bytes processed so far. +// See the bytes.Index implementation for details. +func Cutover(n int) int { + // 1 error per 8 characters, plus a few slop to start. + return (n + 16) / 8 +} diff --git a/src/internal/bytealg/index_ppc64x.s b/src/internal/bytealg/index_ppc64x.s new file mode 100644 index 0000000..80a1f85 --- /dev/null +++ b/src/internal/bytealg/index_ppc64x.s @@ -0,0 +1,841 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This is an implementation based on the s390x +// implementation. + +// Find a separator with 2 <= len <= 32 within a string. +// Separators with lengths of 2, 3 or 4 are handled +// specially. + +// This works on power8 and above. The loads and +// compares are done in big endian order +// since that allows the used of VCLZD, and allows +// the same implementation to work on big and little +// endian platforms with minimal conditional changes. + +// NOTE: There is a power9 implementation that +// improves performance by 10-15% on little +// endian for some of the benchmarks. +// Unrolled index2to16 loop by 4 on ppc64le/power9 +// Work is still needed for a big endian +// implementation on power9. + +//go:build ppc64 || ppc64le + +#include "go_asm.h" +#include "textflag.h" + +// Needed to swap LXVD2X loads to the correct +// byte order to work on POWER8. + +#ifdef GOARCH_ppc64 +DATA byteswap<>+0(SB)/8, $0x0001020304050607 +DATA byteswap<>+8(SB)/8, $0x08090a0b0c0d0e0f +#else +DATA byteswap<>+0(SB)/8, $0x0706050403020100 +DATA byteswap<>+8(SB)/8, $0x0f0e0d0c0b0a0908 +#endif + +// Load bytes in big endian order. Address +// alignment does not need checking. +#define VLOADSWAP(base, index, vreg, vsreg) \ + LXVD2X (base)(index), vsreg; \ + VPERM vreg, vreg, SWAP, vreg + +GLOBL byteswap<>+0(SB), RODATA, $16 + +TEXT ·Index<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-56 + // R3 = byte array pointer + // R4 = length + MOVD R6, R5 // R5 = separator pointer + MOVD R7, R6 // R6 = separator length + +#ifdef GOARCH_ppc64le + MOVBZ internal∕cpu·PPC64+const_offsetPPC64HasPOWER9(SB), R7 + CMP R7, $1 + BNE power8 + BR indexbodyp9<>(SB) +#endif +power8: + BR indexbody<>(SB) + +TEXT ·IndexString<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // R3 = string + // R4 = length + // R5 = separator pointer + // R6 = separator length + +#ifdef GOARCH_ppc64le + MOVBZ internal∕cpu·PPC64+const_offsetPPC64HasPOWER9(SB), R7 + CMP R7, $1 + BNE power8 + BR indexbodyp9<>(SB) + +#endif +power8: + BR indexbody<>(SB) + + // s: string we are searching + // sep: string to search for + // R3=&s[0], R4=len(s) + // R5=&sep[0], R6=len(sep) + // R14=&ret (index where sep found) + // R7=working addr of string + // R16=index value 16 + // R17=index value 17 + // R18=index value 18 + // R19=index value 1 + // R26=LASTBYTE of string + // R27=LASTSTR last start byte to compare with sep + // R8, R9 scratch + // V0=sep left justified zero fill + // CR4=sep length >= 16 + +#define SEPMASK V17 +#define LASTBYTE R26 +#define LASTSTR R27 +#define ONES V20 +#define SWAP V21 +#define SWAP_ VS53 +TEXT indexbody<>(SB), NOSPLIT|NOFRAME, $0 + CMP R6, R4 // Compare lengths + BGT notfound // If sep len is > string, notfound + ADD R4, R3, LASTBYTE // find last byte addr + SUB R6, LASTBYTE, LASTSTR // LAST=&s[len(s)-len(sep)] (last valid start index) + CMP R6, $0 // Check sep len + BEQ notfound // sep len 0 -- not found + MOVD R3, R7 // Copy of string addr + MOVD $16, R16 // Index value 16 + MOVD $17, R17 // Index value 17 + MOVD $18, R18 // Index value 18 + MOVD $1, R19 // Index value 1 + MOVD $byteswap<>+00(SB), R8 + VSPLTISB $0xFF, ONES // splat all 1s + LXVD2X (R8)(R0), SWAP_ // Set up swap string + + CMP R6, $16, CR4 // CR4 for len(sep) >= 16 + VOR ONES, ONES, SEPMASK // Set up full SEPMASK + BGE CR4, loadge16 // Load for len(sep) >= 16 + SUB R6, R16, R9 // 16-len of sep + SLD $3, R9 // Set up for VSLO + MTVSRD R9, V9 // Set up for VSLO + VSLDOI $8, V9, V9, V9 // Set up for VSLO + VSLO ONES, V9, SEPMASK // Mask for separator len(sep) < 16 + +loadge16: + ANDCC $15, R5, R9 // Find byte offset of sep + ADD R9, R6, R10 // Add sep len + CMP R10, $16 // Check if sep len+offset > 16 + BGT sepcross16 // Sep crosses 16 byte boundary + + RLDICR $0, R5, $59, R8 // Adjust addr to 16 byte container + VLOADSWAP(R8, R0, V0, V0) // Load 16 bytes @R8 into V0 + SLD $3, R9 // Set up shift count for VSLO + MTVSRD R9, V8 // Set up shift count for VSLO + VSLDOI $8, V8, V8, V8 + VSLO V0, V8, V0 // Shift by start byte + + VAND V0, SEPMASK, V0 // Mask separator (< 16) + BR index2plus + +sepcross16: + VLOADSWAP(R5, R0, V0, V0) // Load 16 bytes @R5 into V0 + + VAND V0, SEPMASK, V0 // mask out separator + BLE CR4, index2to16 + BR index17plus // Handle sep > 16 + +index2plus: + CMP R6, $2 // Check length of sep + BNE index3plus // If not 2, check for 3 + ADD $16, R7, R9 // Check if next 16 bytes past last + CMP R9, LASTBYTE // compare with last + BGE index2to16 // 2 <= len(string) <= 16 + MOVD $0xff00, R21 // Mask for later + MTVSRD R21, V25 // Move to Vreg + VSPLTH $3, V25, V31 // Splat mask + VSPLTH $0, V0, V1 // Splat 1st 2 bytes of sep + VSPLTISB $0, V10 // Clear V10 + + // First case: 2 byte separator + // V1: 2 byte separator splatted + // V2: 16 bytes at addr + // V4: 16 bytes at addr+1 + // Compare 2 byte separator at start + // and at start+1. Use VSEL to combine + // those results to find the first + // matching start byte, returning + // that value when found. Loop as + // long as len(string) > 16 +index2loop2: + VLOADSWAP(R7, R19, V3, V3) // Load 16 bytes @R7+1 into V3 + +index2loop: + VLOADSWAP(R7, R0, V2, V2) // Load 16 bytes @R7 into V2 + VCMPEQUH V1, V2, V5 // Search for sep + VCMPEQUH V1, V3, V6 // Search for sep offset by 1 + VSEL V6, V5, V31, V7 // merge even and odd indices + VCLZD V7, V18 // find index of first match + MFVSRD V18, R25 // get first value + CMP R25, $64 // Found if < 64 + BLT foundR25 // Return byte index where found + VSLDOI $8, V18, V18, V18 // Adjust 2nd value + MFVSRD V18, R25 // get second value + CMP R25, $64 // Found if < 64 + ADD $64, R25 // Update byte offset + BLT foundR25 // Return value + ADD $16, R7 // R7+=16 Update string pointer + ADD $17, R7, R9 // R9=F7+17 since loop unrolled + CMP R9, LASTBYTE // Compare addr+17 against last byte + BLT index2loop2 // If < last, continue loop + CMP R7, LASTBYTE // Compare addr+16 against last byte + BLT index2to16 // If < 16 handle specially + VLOADSWAP(R7, R0, V3, V3) // Load 16 bytes @R7 into V3 + VSLDOI $1, V3, V10, V3 // Shift left by 1 byte + BR index2loop + +index3plus: + CMP R6, $3 // Check if sep == 3 + BNE index4plus // If not check larger + ADD $19, R7, R9 // Find bytes for use in this loop + CMP R9, LASTBYTE // Compare against last byte + BGE index2to16 // Remaining string 2<=len<=16 + MOVD $0xff00, R21 // Set up mask for upcoming loop + MTVSRD R21, V25 // Move mask to Vreg + VSPLTH $3, V25, V31 // Splat mask + VSPLTH $0, V0, V1 // Splat 1st two bytes of sep + VSPLTB $2, V0, V8 // Splat 3rd byte of sep + + // Loop to process 3 byte separator. + // string[0:16] is in V2 + // string[2:18] is in V3 + // sep[0:2] splatted in V1 + // sec[3] splatted in v8 + // Load vectors at string, string+1 + // and string+2. Compare string, string+1 + // against first 2 bytes of separator + // splatted, and string+2 against 3rd + // byte splatted. Merge the results with + // VSEL to find the first byte of a match. + + // Special handling for last 16 bytes if the + // string fits in 16 byte multiple. +index3loop2: + MOVD $2, R21 // Set up index for 2 + VSPLTISB $0, V10 // Clear V10 + VLOADSWAP(R7, R21, V3, V3)// Load 16 bytes @R7+2 into V3 + VSLDOI $14, V3, V10, V3 // Left justify next 2 bytes + +index3loop: + VLOADSWAP(R7, R0, V2, V2) // Load with correct order + VSLDOI $1, V2, V3, V4 // string[1:17] + VSLDOI $2, V2, V3, V9 // string[2:18] + VCMPEQUH V1, V2, V5 // compare hw even indices + VCMPEQUH V1, V4, V6 // compare hw odd indices + VCMPEQUB V8, V9, V10 // compare 3rd to last byte + VSEL V6, V5, V31, V7 // Find 1st matching byte using mask + VAND V7, V10, V7 // AND matched bytes with matched 3rd byte + VCLZD V7, V18 // Find first nonzero indexes + MFVSRD V18, R25 // Move 1st doubleword + CMP R25, $64 // If < 64 found + BLT foundR25 // Return matching index + VSLDOI $8, V18, V18, V18 // Move value + MFVSRD V18, R25 // Move 2nd doubleword + CMP R25, $64 // If < 64 found + ADD $64, R25 // Update byte index + BLT foundR25 // Return matching index + ADD $16, R7 // R7+=16 string ptr + ADD $19, R7, R9 // Number of string bytes for loop + CMP R9, LASTBYTE // Compare against last byte of string + BLT index3loop2 // If within, continue this loop + CMP R7, LASTSTR // Compare against last start byte + BLT index2to16 // Process remainder + VSPLTISB $0, V3 // Special case for last 16 bytes + BR index3loop // Continue this loop + + // Loop to process 4 byte separator + // string[0:16] in V2 + // string[3:16] in V3 + // sep[0:4] splatted in V1 + // Set up vectors with strings at offsets + // 0, 1, 2, 3 and compare against the 4 byte + // separator also splatted. Use VSEL with the + // compare results to find the first byte where + // a separator match is found. +index4plus: + CMP R6, $4 // Check if 4 byte separator + BNE index5plus // If not next higher + ADD $20, R7, R9 // Check string size to load + CMP R9, LASTBYTE // Verify string length + BGE index2to16 // If not large enough, process remaining + MOVD $2, R15 // Set up index + + // Set up masks for use with VSEL + MOVD $0xff, R21 // Set up mask 0xff000000ff000000... + SLD $24, R21 + MTVSRD R21, V10 + VSPLTW $1, V10, V29 + VSLDOI $2, V29, V29, V30 // Mask 0x0000ff000000ff00... + MOVD $0xffff, R21 + SLD $16, R21 + MTVSRD R21, V10 + VSPLTW $1, V10, V31 // Mask 0xffff0000ffff0000... + VSPLTW $0, V0, V1 // Splat 1st word of separator + +index4loop: + VLOADSWAP(R7, R0, V2, V2) // Load 16 bytes @R7 into V2 + +next4: + VSPLTISB $0, V10 // Clear + MOVD $3, R9 // Number of bytes beyond 16 + VLOADSWAP(R7, R9, V3, V3) // Load 16 bytes @R7+3 into V3 + VSLDOI $13, V3, V10, V3 // Shift left last 3 bytes + VSLDOI $1, V2, V3, V4 // V4=(V2:V3)<<1 + VSLDOI $2, V2, V3, V9 // V9=(V2:V3)<<2 + VSLDOI $3, V2, V3, V10 // V10=(V2:v3)<<3 + VCMPEQUW V1, V2, V5 // compare index 0, 4, ... with sep + VCMPEQUW V1, V4, V6 // compare index 1, 5, ... with sep + VCMPEQUW V1, V9, V11 // compare index 2, 6, ... with sep + VCMPEQUW V1, V10, V12 // compare index 3, 7, ... with sep + VSEL V6, V5, V29, V13 // merge index 0, 1, 4, 5, using mask + VSEL V12, V11, V30, V14 // merge index 2, 3, 6, 7, using mask + VSEL V14, V13, V31, V7 // final merge + VCLZD V7, V18 // Find first index for each half + MFVSRD V18, R25 // Isolate value + CMP R25, $64 // If < 64, found + BLT foundR25 // Return found index + VSLDOI $8, V18, V18, V18 // Move for MFVSRD + MFVSRD V18, R25 // Isolate other value + CMP R25, $64 // If < 64, found + ADD $64, R25 // Update index for high doubleword + BLT foundR25 // Return found index + ADD $16, R7 // R7+=16 for next string + ADD $20, R7, R9 // R+20 for all bytes to load + CMP R9, LASTBYTE // Past end? Maybe check for extra? + BLT index4loop // If not, continue loop + CMP R7, LASTSTR // Check remainder + BLE index2to16 // Process remainder + BR notfound // Not found + +index5plus: + CMP R6, $16 // Check for sep > 16 + BGT index17plus // Handle large sep + + // Assumption is that the separator is smaller than the string at this point +index2to16: + CMP R7, LASTSTR // Compare last start byte + BGT notfound // last takes len(sep) into account + + ADD $16, R7, R9 // Check for last byte of string + CMP R9, LASTBYTE + BGT index2to16tail + + // At least 16 bytes of string left + // Mask the number of bytes in sep +index2to16loop: + VLOADSWAP(R7, R0, V1, V1) // Load 16 bytes @R7 into V1 + +compare: + VAND V1, SEPMASK, V2 // Mask out sep size + VCMPEQUBCC V0, V2, V3 // Compare masked string + BLT CR6, found // All equal + ADD $1, R7 // Update ptr to next byte + CMP R7, LASTSTR // Still less than last start byte + BGT notfound // Not found + ADD $16, R7, R9 // Verify remaining bytes + CMP R9, LASTBYTE // At least 16 + BLT index2to16loop // Try again + + // Less than 16 bytes remaining in string + // Separator >= 2 +index2to16tail: + ADD R3, R4, R9 // End of string + SUB R7, R9, R9 // Number of bytes left + ANDCC $15, R7, R10 // 16 byte offset + ADD R10, R9, R11 // offset + len + CMP R11, $16 // >= 16? + BLE short // Does not cross 16 bytes + VLOADSWAP(R7, R0, V1, V1) // Load 16 bytes @R7 into V1 + BR index2to16next // Continue on + +short: + RLDICR $0, R7, $59, R9 // Adjust addr to 16 byte container + VLOADSWAP(R9, R0, V1, V1)// Load 16 bytes @R9 into V1 + SLD $3, R10 // Set up shift + MTVSRD R10, V8 // Set up shift + VSLDOI $8, V8, V8, V8 + VSLO V1, V8, V1 // Shift by start byte + VSPLTISB $0, V25 // Clear for later use + +index2to16next: + VAND V1, SEPMASK, V2 // Just compare size of sep + VCMPEQUBCC V0, V2, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BGT notfound // If at end, then not found + VSLDOI $1, V1, V25, V1 // Shift string left by 1 byte + BR index2to16next // Check the next partial string + +index17plus: + CMP R6, $32 // Check if 17 < len(sep) <= 32 + BGT index33plus + SUB $16, R6, R9 // Extra > 16 + SLD $56, R9, R10 // Shift to use in VSLO + MTVSRD R10, V9 // Set up for VSLO + VLOADSWAP(R5, R9, V1, V1)// Load 16 bytes @R5+R9 into V1 + VSLO V1, V9, V1 // Shift left + VSPLTISB $0xff, V7 // Splat 1s + VSPLTISB $0, V27 // Splat 0 + +index17to32loop: + VLOADSWAP(R7, R0, V2, V2) // Load 16 bytes @R7 into V2 + +next17: + VLOADSWAP(R7, R9, V3, V3) // Load 16 bytes @R7+R9 into V3 + VSLO V3, V9, V3 // Shift left + VCMPEQUB V0, V2, V4 // Compare first 16 bytes + VCMPEQUB V1, V3, V5 // Compare extra over 16 bytes + VAND V4, V5, V6 // Check if both equal + VCMPEQUBCC V6, V7, V8 // All equal? + BLT CR6, found // Yes + ADD $1, R7 // On to next byte + CMP R7, LASTSTR // Check if last start byte + BGT notfound // If too high, not found + BR index17to32loop // Continue + +notfound: + MOVD $-1, R3 // Return -1 if not found + RET + +index33plus: + MOVD $0, (R0) // Case not implemented + RET // Crash before return + +foundR25: + SRD $3, R25 // Convert from bits to bytes + ADD R25, R7 // Add to current string address + SUB R3, R7 // Subtract from start of string + MOVD R7, R3 // Return byte where found + RET + +found: + SUB R3, R7 // Return byte where found + MOVD R7, R3 + RET + +TEXT indexbodyp9<>(SB), NOSPLIT|NOFRAME, $0 + CMP R6, R4 // Compare lengths + BGT notfound // If sep len is > string, notfound + ADD R4, R3, LASTBYTE // find last byte addr + SUB R6, LASTBYTE, LASTSTR // LAST=&s[len(s)-len(sep)] (last valid start index) + CMP R6, $0 // Check sep len + BEQ notfound // sep len 0 -- not found + MOVD R3, R7 // Copy of string addr +#ifndef GOPPC64_power10 + MOVD $16, R16 // Index value 16 + MOVD $17, R17 // Index value 17 + MOVD $18, R18 // Index value 18 + VSPLTISB $0xFF, ONES // splat all 1s + VOR ONES, ONES, SEPMASK // Set up full SEPMASK +#else + SLD $56, R6, R14 // Set up separator length for LXVLL +#endif + MOVD $1, R19 // Index value 1 + CMP R6, $16, CR4 // CR4 for len(sep) >= 16 + BGE CR4, loadge16 // Load for len(sep) >= 16 +#ifndef GOPPC64_power10 + SUB R6, R16, R9 // 16-len of sep + SLD $3, R9 // Set up for VSLO + MTVSRD R9, V9 // Set up for VSLO + VSLDOI $8, V9, V9, V9 // Set up for VSLO + VSLO ONES, V9, SEPMASK // Mask for separator len(sep) < 16 +#endif +loadge16: + ANDCC $15, R5, R9 // Find byte offset of sep + ADD R9, R6, R10 // Add sep len + CMP R10, $16 // Check if sep len+offset > 16 + BGT sepcross16 // Sep crosses 16 byte boundary +#ifdef GOPPC64_power10 + LXVLL R5, R14, V0 // Load separator +#else + RLDICR $0, R5, $59, R8 // Adjust addr to 16 byte container + LXVB16X (R8)(R0), V0 // Load 16 bytes @R8 into V0 + SLD $3, R9 // Set up shift count for VSLO + MTVSRD R9, V8 // Set up shift count for VSLO + VSLDOI $8, V8, V8, V8 + VSLO V0, V8, V0 // Shift by start byte + VAND V0, SEPMASK, V0 // Mask separator (< 16) +#endif + BR index2plus +sepcross16: +#ifdef GOPPC64_power10 + LXVLL R5, R14, V0 // Load separator +#else + LXVB16X (R5)(R0), V0 // Load 16 bytes @R5 into V0\ + VAND V0, SEPMASK, V0 // mask out separator +#endif + BLE CR4, index2to16 + BR index17plus // Handle sep > 16 + +index2plus: + CMP R6, $2 // Check length of sep + BNE index3plus // If not 2, check for 3 + ADD $16, R7, R9 // Check if next 16 bytes past last + CMP R9, LASTBYTE // compare with last + BGE index2to16 // 2 <= len(string) <= 16 + MOVD $0xff00, R21 // Mask for later + MTVSRD R21, V25 // Move to Vreg + VSPLTH $3, V25, V31 // Splat mask + VSPLTH $0, V0, V1 // Splat 1st 2 bytes of sep + VSPLTISB $0, V10 // Clear V10 + + // First case: 2 byte separator + // V1: 2 byte separator splatted + // V2: 16 bytes at addr + // V4: 16 bytes at addr+1 + // Compare 2 byte separator at start + // and at start+1. Use VSEL to combine + // those results to find the first + // matching start byte, returning + // that value when found. Loop as + // long as len(string) > 16 +index2loop2: + LXVB16X (R7)(R19), V3 // Load 16 bytes @R7+1 into V3 + +index2loop: + LXVB16X (R7)(R0), V2 // Load 16 bytes @R7 into V2 + VCMPEQUH V1, V2, V5 // Search for sep + VCMPEQUH V1, V3, V6 // Search for sep offset by 1 + VSEL V6, V5, V31, V7 // merge even and odd indices + VCLZD V7, V18 // find index of first match + MFVSRD V18, R25 // get first value + CMP R25, $64 // Found if < 64 + BLT foundR25 // Return byte index where found + + MFVSRLD V18, R25 // get second value + CMP R25, $64 // Found if < 64 + ADD $64, R25 // Update byte offset + BLT foundR25 // Return value + ADD $16, R7 // R7+=16 Update string pointer + ADD $17, R7, R9 // R9=F7+17 since loop unrolled + CMP R9, LASTBYTE // Compare addr+17 against last byte + BLT index2loop2 // If < last, continue loop + CMP R7, LASTBYTE // Compare addr+16 against last byte + BLT index2to16 // If < 16 handle specially + LXVB16X (R7)(R0), V3 // Load 16 bytes @R7 into V3 + VSLDOI $1, V3, V10, V3 // Shift left by 1 byte + BR index2loop + +index3plus: + CMP R6, $3 // Check if sep == 3 + BNE index4plus // If not check larger + ADD $19, R7, R9 // Find bytes for use in this loop + CMP R9, LASTBYTE // Compare against last byte + BGE index2to16 // Remaining string 2<=len<=16 + MOVD $0xff00, R21 // Set up mask for upcoming loop + MTVSRD R21, V25 // Move mask to Vreg + VSPLTH $3, V25, V31 // Splat mask + VSPLTH $0, V0, V1 // Splat 1st two bytes of sep + VSPLTB $2, V0, V8 // Splat 3rd byte of sep + + // Loop to process 3 byte separator. + // string[0:16] is in V2 + // string[2:18] is in V3 + // sep[0:2] splatted in V1 + // sec[3] splatted in v8 + // Load vectors at string, string+1 + // and string+2. Compare string, string+1 + // against first 2 bytes of separator + // splatted, and string+2 against 3rd + // byte splatted. Merge the results with + // VSEL to find the first byte of a match. + + // Special handling for last 16 bytes if the + // string fits in 16 byte multiple. +index3loop2: + MOVD $2, R21 // Set up index for 2 + VSPLTISB $0, V10 // Clear V10 + LXVB16X (R7)(R21), V3 // Load 16 bytes @R7+2 into V3 + VSLDOI $14, V3, V10, V3 // Left justify next 2 bytes + +index3loop: + LXVB16X (R7)(R0), V2 // Load 16 bytes @R7 + VSLDOI $1, V2, V3, V4 // string[1:17] + VSLDOI $2, V2, V3, V9 // string[2:18] + VCMPEQUH V1, V2, V5 // compare hw even indices + VCMPEQUH V1, V4, V6 // compare hw odd indices + VCMPEQUB V8, V9, V10 // compare 3rd to last byte + VSEL V6, V5, V31, V7 // Find 1st matching byte using mask + VAND V7, V10, V7 // AND matched bytes with matched 3rd byte + VCLZD V7, V18 // Find first nonzero indexes + MFVSRD V18, R25 // Move 1st doubleword + CMP R25, $64 // If < 64 found + BLT foundR25 // Return matching index + + MFVSRLD V18, R25 // Move 2nd doubleword + CMP R25, $64 // If < 64 found + ADD $64, R25 // Update byte index + BLT foundR25 // Return matching index + ADD $16, R7 // R7+=16 string ptr + ADD $19, R7, R9 // Number of string bytes for loop + CMP R9, LASTBYTE // Compare against last byte of string + BLT index3loop2 // If within, continue this loop + CMP R7, LASTSTR // Compare against last start byte + BLT index2to16 // Process remainder + VSPLTISB $0, V3 // Special case for last 16 bytes + BR index3loop // Continue this loop + + // Loop to process 4 byte separator + // string[0:16] in V2 + // string[3:16] in V3 + // sep[0:4] splatted in V1 + // Set up vectors with strings at offsets + // 0, 1, 2, 3 and compare against the 4 byte + // separator also splatted. Use VSEL with the + // compare results to find the first byte where + // a separator match is found. +index4plus: + CMP R6, $4 // Check if 4 byte separator + BNE index5plus // If not next higher + ADD $20, R7, R9 // Check string size to load + CMP R9, LASTBYTE // Verify string length + BGE index2to16 // If not large enough, process remaining + + // Set up masks for use with VSEL + MOVD $0xff, R21 // Set up mask 0xff000000ff000000... + SLD $24, R21 + MTVSRWS R21, V29 + + VSLDOI $2, V29, V29, V30 // Mask 0x0000ff000000ff00... + MOVD $0xffff, R21 + SLD $16, R21 + MTVSRWS R21, V31 + + VSPLTW $0, V0, V1 // Splat 1st word of separator + +index4loop: + LXVB16X (R7)(R0), V2 // Load 16 bytes @R7 into V2 + +next4: + VSPLTISB $0, V10 // Clear + MOVD $3, R9 // Number of bytes beyond 16 + LXVB16X (R7)(R9), V3 // Load 16 bytes @R7 into V3 + VSLDOI $13, V3, V10, V3 // Shift left last 3 bytes + VSLDOI $1, V2, V3, V4 // V4=(V2:V3)<<1 + VSLDOI $2, V2, V3, V9 // V9=(V2:V3)<<2 + VSLDOI $3, V2, V3, V10 // V10=(V2:v3)<<3 + VCMPEQUW V1, V2, V5 // compare index 0, 4, ... with sep + VCMPEQUW V1, V4, V6 // compare index 1, 5, ... with sep + VCMPEQUW V1, V9, V11 // compare index 2, 6, ... with sep + VCMPEQUW V1, V10, V12 // compare index 3, 7, ... with sep + VSEL V6, V5, V29, V13 // merge index 0, 1, 4, 5, using mask + VSEL V12, V11, V30, V14 // merge index 2, 3, 6, 7, using mask + VSEL V14, V13, V31, V7 // final merge + VCLZD V7, V18 // Find first index for each half + MFVSRD V18, R25 // Isolate value + CMP R25, $64 // If < 64, found + BLT foundR25 // Return found index + + MFVSRLD V18, R25 // Isolate other value + CMP R25, $64 // If < 64, found + ADD $64, R25 // Update index for high doubleword + BLT foundR25 // Return found index + ADD $16, R7 // R7+=16 for next string + ADD $20, R7, R9 // R+20 for all bytes to load + CMP R9, LASTBYTE // Past end? Maybe check for extra? + BLT index4loop // If not, continue loop + CMP R7, LASTSTR // Check remainder + BLE index2to16 // Process remainder + BR notfound // Not found + +index5plus: + CMP R6, $16 // Check for sep > 16 + BGT index17plus // Handle large sep + + // Assumption is that the separator is smaller than the string at this point +index2to16: + CMP R7, LASTSTR // Compare last start byte + BGT notfound // last takes len(sep) into account + + ADD $19, R7, R9 // To check 4 indices per iteration, need at least 16+3 bytes + CMP R9, LASTBYTE + // At least 16 bytes of string left + // Mask the number of bytes in sep + VSPLTISB $0, V10 // Clear + BGT index2to16tail + +#ifdef GOPPC64_power10 + ADD $3,R7, R17 // Base+3 + ADD $2,R7, R8 // Base+2 + ADD $1,R7, R10 // Base+1 +#else + MOVD $3, R17 // Number of bytes beyond 16 +#endif + PCALIGN $16 + +index2to16loop: + +#ifdef GOPPC64_power10 + LXVLL R7, R14, V8 // Load next 16 bytes of string from Base + LXVLL R10, R14, V9 // Load next 16 bytes of string from Base+1 + LXVLL R8, R14, V11 // Load next 16 bytes of string from Base+2 + LXVLL R17,R14, V12 // Load next 16 bytes of string from Base+3 +#else + LXVB16X (R7)(R0), V1 // Load next 16 bytes of string into V1 from R7 + LXVB16X (R7)(R17), V5 // Load next 16 bytes of string into V5 from R7+3 + + VSLDOI $13, V5, V10, V2 // Shift left last 3 bytes + VSLDOI $1, V1, V2, V3 // V3=(V1:V2)<<1 + VSLDOI $2, V1, V2, V4 // V4=(V1:V2)<<2 + VAND V1, SEPMASK, V8 // Mask out sep size 0th index + VAND V3, SEPMASK, V9 // Mask out sep size 1st index + VAND V4, SEPMASK, V11 // Mask out sep size 2nd index + VAND V5, SEPMASK, V12 // Mask out sep size 3rd index +#endif + VCMPEQUBCC V0, V8, V8 // compare masked string + BLT CR6, found // All equal while comparing 0th index + VCMPEQUBCC V0, V9, V9 // compare masked string + BLT CR6, found2 // All equal while comparing 1st index + VCMPEQUBCC V0, V11, V11 // compare masked string + BLT CR6, found3 // All equal while comparing 2nd index + VCMPEQUBCC V0, V12, V12 // compare masked string + BLT CR6, found4 // All equal while comparing 3rd index + + ADD $4, R7 // Update ptr to next 4 bytes +#ifdef GOPPC64_power10 + ADD $4, R17 // Update ptr to next 4 bytes + ADD $4, R8 // Update ptr to next 4 bytes + ADD $4, R10 // Update ptr to next 4 bytes +#endif + CMP R7, LASTSTR // Still less than last start byte + BGT notfound // Not found + ADD $19, R7, R9 // Verify remaining bytes + CMP R9, LASTBYTE // length of string at least 19 + BLE index2to16loop // Try again, else do post processing and jump to index2to16next + PCALIGN $32 + // <19 bytes left, post process the remaining string +index2to16tail: +#ifdef GOPPC64_power10 +index2to16next_p10: + LXVLL R7,R14, V1 // Load 16 bytes @R7 into V1 + VCMPEQUBCC V1, V0, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BLE index2to16next_p10 // If at end, then not found + BR notfound // go to remainder loop +#else + ADD R3, R4, R9 // End of string + SUB R7, R9, R9 // Number of bytes left + ANDCC $15, R7, R10 // 16 byte offset + ADD R10, R9, R11 // offset + len + CMP R11, $16 // >= 16? + BLE short // Does not cross 16 bytes + LXVB16X (R7)(R0), V1 // Load 16 bytes @R7 into V1 + CMP R9, $16 // Post-processing of unrolled loop + BLE index2to16next // continue to index2to16next if <= 16 bytes + SUB R16, R9, R10 // R9 should be 18 or 17 hence R10 is 1 or 2 + LXVB16X (R7)(R10), V9 + CMP R10, $1 // string length is 17, compare 1 more byte + BNE extra2 // string length is 18, compare 2 more bytes + VSLDOI $15, V9, V10, V25 + VAND V1, SEPMASK, V2 // Just compare size of sep + VCMPEQUBCC V0, V2, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BGT notfound // If at end, then not found + VSLDOI $1, V1, V25, V1 // Shift string left by 1 byte + BR index2to16next // go to remainder loop +extra2: + VSLDOI $14, V9, V10, V25 + VAND V1, SEPMASK, V2 // Just compare size of sep + VCMPEQUBCC V0, V2, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BGT notfound // If at end, then not found + VOR V1, V1, V4 // save remaining string + VSLDOI $1, V1, V25, V1 // Shift string left by 1 byte for 17th byte + VAND V1, SEPMASK, V2 // Just compare size of sep + VCMPEQUBCC V0, V2, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BGT notfound // If at end, then not found + VSLDOI $2, V4, V25, V1 // Shift saved string left by 2 bytes for 18th byte + BR index2to16next // Check the remaining partial string in index2to16next + +short: + RLDICR $0, R7, $59, R9 // Adjust addr to 16 byte container + LXVB16X (R9)(R0), V1 // Load 16 bytes @R9 into V1 + SLD $3, R10 // Set up shift + MTVSRD R10, V8 // Set up shift + VSLDOI $8, V8, V8, V8 + VSLO V1, V8, V1 // Shift by start byte + PCALIGN $16 +index2to16next: + VAND V1, SEPMASK, V2 // Just compare size of sep + VCMPEQUBCC V0, V2, V3 // Compare sep and partial string + BLT CR6, found // Found + ADD $1, R7 // Not found, try next partial string + CMP R7, LASTSTR // Check for end of string + BGT notfound // If at end, then not found + VSLDOI $1, V1, V10, V1 // Shift string left by 1 byte + BR index2to16next // Check the next partial string +#endif // Tail processing if GOPPC64!=power10 + +index17plus: + CMP R6, $32 // Check if 17 < len(sep) <= 32 + BGT index33plus + SUB $16, R6, R9 // Extra > 16 + SLD $56, R9, R10 // Shift to use in VSLO + MTVSRD R10, V9 // Set up for VSLO + LXVB16X (R5)(R9), V1 // Load 16 bytes @R5+R9 into V1 + VSLO V1, V9, V1 // Shift left + VSPLTISB $0xff, V7 // Splat 1s + VSPLTISB $0, V27 // Splat 0 + +index17to32loop: + LXVB16X (R7)(R0), V2 // Load 16 bytes @R7 into V2 + +next17: + LXVB16X (R7)(R9), V3 // Load 16 bytes @R7+R9 into V3 + VSLO V3, V9, V3 // Shift left + VCMPEQUB V0, V2, V4 // Compare first 16 bytes + VCMPEQUB V1, V3, V5 // Compare extra over 16 bytes + VAND V4, V5, V6 // Check if both equal + VCMPEQUBCC V6, V7, V8 // All equal? + BLT CR6, found // Yes + ADD $1, R7 // On to next byte + CMP R7, LASTSTR // Check if last start byte + BGT notfound // If too high, not found + BR index17to32loop // Continue + +notfound: + MOVD $-1, R3 // Return -1 if not found + RET + +index33plus: + MOVD $0, (R0) // Case not implemented + RET // Crash before return + +foundR25: + SRD $3, R25 // Convert from bits to bytes + ADD R25, R7 // Add to current string address + SUB R3, R7 // Subtract from start of string + MOVD R7, R3 // Return byte where found + RET +found4: + ADD $1, R7 // found from unrolled loop at index 3 +found3: + ADD $1, R7 // found from unrolled loop at index 2 +found2: + ADD $1, R7 // found from unrolled loop at index 1 +found: // found at index 0 + SUB R3, R7 // Return byte where found + MOVD R7, R3 + RET diff --git a/src/internal/bytealg/index_s390x.go b/src/internal/bytealg/index_s390x.go new file mode 100644 index 0000000..9340cf1 --- /dev/null +++ b/src/internal/bytealg/index_s390x.go @@ -0,0 +1,31 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package bytealg + +import "internal/cpu" + +const MaxBruteForce = 64 + +func init() { + // Note: we're kind of lucky that this flag is available at this point. + // The runtime sets HasVX when processing auxv records, and that happens + // to happen *before* running the init functions of packages that + // the runtime depends on. + // TODO: it would really be nicer for internal/cpu to figure out this + // flag by itself. Then we wouldn't need to depend on quirks of + // early startup initialization order. + if cpu.S390X.HasVX { + MaxLen = 64 + } +} + +// Cutover reports the number of failures of IndexByte we should tolerate +// before switching over to Index. +// n is the number of bytes processed so far. +// See the bytes.Index implementation for details. +func Cutover(n int) int { + // 1 error per 8 characters, plus a few slop to start. + return (n + 16) / 8 +} diff --git a/src/internal/bytealg/index_s390x.s b/src/internal/bytealg/index_s390x.s new file mode 100644 index 0000000..491d5bc --- /dev/null +++ b/src/internal/bytealg/index_s390x.s @@ -0,0 +1,216 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +// Caller must confirm availability of vx facility before calling. +TEXT ·Index(SB),NOSPLIT|NOFRAME,$0-56 + LMG a_base+0(FP), R1, R2 // R1=&s[0], R2=len(s) + LMG b_base+24(FP), R3, R4 // R3=&sep[0], R4=len(sep) + MOVD $ret+48(FP), R5 + BR indexbody<>(SB) + +// Caller must confirm availability of vx facility before calling. +TEXT ·IndexString(SB),NOSPLIT|NOFRAME,$0-40 + LMG a_base+0(FP), R1, R2 // R1=&s[0], R2=len(s) + LMG b_base+16(FP), R3, R4 // R3=&sep[0], R4=len(sep) + MOVD $ret+32(FP), R5 + BR indexbody<>(SB) + +// s: string we are searching +// sep: string to search for +// R1=&s[0], R2=len(s) +// R3=&sep[0], R4=len(sep) +// R5=&ret (int) +// Caller must confirm availability of vx facility before calling. +TEXT indexbody<>(SB),NOSPLIT|NOFRAME,$0 + CMPBGT R4, R2, notfound + ADD R1, R2 + SUB R4, R2 // R2=&s[len(s)-len(sep)] (last valid index) + CMPBEQ R4, $0, notfound + SUB $1, R4 // R4=len(sep)-1 for use as VLL index + VLL R4, (R3), V0 // contains first 16 bytes of sep + MOVD R1, R7 +index2plus: + CMPBNE R4, $1, index3plus + MOVD $15(R7), R9 + CMPBGE R9, R2, index2to16 + VGBM $0xaaaa, V31 // 0xff00ff00ff00ff00... + VONE V16 + VREPH $0, V0, V1 + CMPBGE R9, R2, index2to16 +index2loop: + VL 0(R7), V2 // 16 bytes, even indices + VL 1(R7), V4 // 16 bytes, odd indices + VCEQH V1, V2, V5 // compare even indices + VCEQH V1, V4, V6 // compare odd indices + VSEL V5, V6, V31, V7 // merge even and odd indices + VFEEBS V16, V7, V17 // find leftmost index, set condition to 1 if found + BLT foundV17 + MOVD $16(R7), R7 // R7+=16 + ADD $15, R7, R9 + CMPBLE R9, R2, index2loop // continue if (R7+15) <= R2 (last index to search) + CMPBLE R7, R2, index2to16 + BR notfound + +index3plus: + CMPBNE R4, $2, index4plus + ADD $15, R7, R9 + CMPBGE R9, R2, index2to16 + MOVD $1, R0 + VGBM $0xaaaa, V31 // 0xff00ff00ff00ff00... + VONE V16 + VREPH $0, V0, V1 + VREPB $2, V0, V8 +index3loop: + VL (R7), V2 // load 16-bytes into V2 + VLL R0, 16(R7), V3 // load 2-bytes into V3 + VSLDB $1, V2, V3, V4 // V4=(V2:V3)<<1 + VSLDB $2, V2, V3, V9 // V9=(V2:V3)<<2 + VCEQH V1, V2, V5 // compare 2-byte even indices + VCEQH V1, V4, V6 // compare 2-byte odd indices + VCEQB V8, V9, V10 // compare last bytes + VSEL V5, V6, V31, V7 // merge even and odd indices + VN V7, V10, V7 // AND indices with last byte + VFEEBS V16, V7, V17 // find leftmost index, set condition to 1 if found + BLT foundV17 + MOVD $16(R7), R7 // R7+=16 + ADD $15, R7, R9 + CMPBLE R9, R2, index3loop // continue if (R7+15) <= R2 (last index to search) + CMPBLE R7, R2, index2to16 + BR notfound + +index4plus: + CMPBNE R4, $3, index5plus + ADD $15, R7, R9 + CMPBGE R9, R2, index2to16 + MOVD $2, R0 + VGBM $0x8888, V29 // 0xff000000ff000000... + VGBM $0x2222, V30 // 0x0000ff000000ff00... + VGBM $0xcccc, V31 // 0xffff0000ffff0000... + VONE V16 + VREPF $0, V0, V1 +index4loop: + VL (R7), V2 // load 16-bytes into V2 + VLL R0, 16(R7), V3 // load 3-bytes into V3 + VSLDB $1, V2, V3, V4 // V4=(V2:V3)<<1 + VSLDB $2, V2, V3, V9 // V9=(V2:V3)<<1 + VSLDB $3, V2, V3, V10 // V10=(V2:V3)<<1 + VCEQF V1, V2, V5 // compare index 0, 4, ... + VCEQF V1, V4, V6 // compare index 1, 5, ... + VCEQF V1, V9, V11 // compare index 2, 6, ... + VCEQF V1, V10, V12 // compare index 3, 7, ... + VSEL V5, V6, V29, V13 // merge index 0, 1, 4, 5, ... + VSEL V11, V12, V30, V14 // merge index 2, 3, 6, 7, ... + VSEL V13, V14, V31, V7 // final merge + VFEEBS V16, V7, V17 // find leftmost index, set condition to 1 if found + BLT foundV17 + MOVD $16(R7), R7 // R7+=16 + ADD $15, R7, R9 + CMPBLE R9, R2, index4loop // continue if (R7+15) <= R2 (last index to search) + CMPBLE R7, R2, index2to16 + BR notfound + +index5plus: + CMPBGT R4, $15, index17plus +index2to16: + CMPBGT R7, R2, notfound + MOVD $1(R7), R8 + CMPBGT R8, R2, index2to16tail +index2to16loop: + // unrolled 2x + VLL R4, (R7), V1 + VLL R4, 1(R7), V2 + VCEQGS V0, V1, V3 + BEQ found + MOVD $1(R7), R7 + VCEQGS V0, V2, V4 + BEQ found + MOVD $1(R7), R7 + CMPBLT R7, R2, index2to16loop + CMPBGT R7, R2, notfound +index2to16tail: + VLL R4, (R7), V1 + VCEQGS V0, V1, V2 + BEQ found + BR notfound + +index17plus: + CMPBGT R4, $31, index33plus + SUB $16, R4, R0 + VLL R0, 16(R3), V1 + VONE V7 +index17to32loop: + VL (R7), V2 + VLL R0, 16(R7), V3 + VCEQG V0, V2, V4 + VCEQG V1, V3, V5 + VN V4, V5, V6 + VCEQGS V6, V7, V8 + BEQ found + MOVD $1(R7), R7 + CMPBLE R7, R2, index17to32loop + BR notfound + +index33plus: + CMPBGT R4, $47, index49plus + SUB $32, R4, R0 + VL 16(R3), V1 + VLL R0, 32(R3), V2 + VONE V11 +index33to48loop: + VL (R7), V3 + VL 16(R7), V4 + VLL R0, 32(R7), V5 + VCEQG V0, V3, V6 + VCEQG V1, V4, V7 + VCEQG V2, V5, V8 + VN V6, V7, V9 + VN V8, V9, V10 + VCEQGS V10, V11, V12 + BEQ found + MOVD $1(R7), R7 + CMPBLE R7, R2, index33to48loop + BR notfound + +index49plus: + CMPBGT R4, $63, index65plus + SUB $48, R4, R0 + VL 16(R3), V1 + VL 32(R3), V2 + VLL R0, 48(R3), V3 + VONE V15 +index49to64loop: + VL (R7), V4 + VL 16(R7), V5 + VL 32(R7), V6 + VLL R0, 48(R7), V7 + VCEQG V0, V4, V8 + VCEQG V1, V5, V9 + VCEQG V2, V6, V10 + VCEQG V3, V7, V11 + VN V8, V9, V12 + VN V10, V11, V13 + VN V12, V13, V14 + VCEQGS V14, V15, V16 + BEQ found + MOVD $1(R7), R7 + CMPBLE R7, R2, index49to64loop +notfound: + MOVD $-1, (R5) + RET + +index65plus: + // not implemented + MOVD $0, (R0) + RET + +foundV17: // index is in doubleword V17[0] + VLGVG $0, V17, R8 + ADD R8, R7 +found: + SUB R1, R7 + MOVD R7, (R5) + RET diff --git a/src/internal/bytealg/indexbyte_386.s b/src/internal/bytealg/indexbyte_386.s new file mode 100644 index 0000000..8a03054 --- /dev/null +++ b/src/internal/bytealg/indexbyte_386.s @@ -0,0 +1,34 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-20 + MOVL b_base+0(FP), SI + MOVL b_len+4(FP), CX + MOVB c+12(FP), AL + MOVL SI, DI + CLD; REPN; SCASB + JZ 3(PC) + MOVL $-1, ret+16(FP) + RET + SUBL SI, DI + SUBL $1, DI + MOVL DI, ret+16(FP) + RET + +TEXT ·IndexByteString(SB),NOSPLIT,$0-16 + MOVL s_base+0(FP), SI + MOVL s_len+4(FP), CX + MOVB c+8(FP), AL + MOVL SI, DI + CLD; REPN; SCASB + JZ 3(PC) + MOVL $-1, ret+12(FP) + RET + SUBL SI, DI + SUBL $1, DI + MOVL DI, ret+12(FP) + RET diff --git a/src/internal/bytealg/indexbyte_amd64.s b/src/internal/bytealg/indexbyte_amd64.s new file mode 100644 index 0000000..1ca70e3 --- /dev/null +++ b/src/internal/bytealg/indexbyte_amd64.s @@ -0,0 +1,149 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB), NOSPLIT, $0-40 + MOVQ b_base+0(FP), SI + MOVQ b_len+8(FP), BX + MOVB c+24(FP), AL + LEAQ ret+32(FP), R8 + JMP indexbytebody<>(SB) + +TEXT ·IndexByteString(SB), NOSPLIT, $0-32 + MOVQ s_base+0(FP), SI + MOVQ s_len+8(FP), BX + MOVB c+16(FP), AL + LEAQ ret+24(FP), R8 + JMP indexbytebody<>(SB) + +// input: +// SI: data +// BX: data len +// AL: byte sought +// R8: address to put result +TEXT indexbytebody<>(SB), NOSPLIT, $0 + // Shuffle X0 around so that each byte contains + // the character we're looking for. + MOVD AX, X0 + PUNPCKLBW X0, X0 + PUNPCKLBW X0, X0 + PSHUFL $0, X0, X0 + + CMPQ BX, $16 + JLT small + + MOVQ SI, DI + + CMPQ BX, $32 + JA avx2 +sse: + LEAQ -16(SI)(BX*1), AX // AX = address of last 16 bytes + JMP sseloopentry + +sseloop: + // Move the next 16-byte chunk of the data into X1. + MOVOU (DI), X1 + // Compare bytes in X0 to X1. + PCMPEQB X0, X1 + // Take the top bit of each byte in X1 and put the result in DX. + PMOVMSKB X1, DX + // Find first set bit, if any. + BSFL DX, DX + JNZ ssesuccess + // Advance to next block. + ADDQ $16, DI +sseloopentry: + CMPQ DI, AX + JB sseloop + + // Search the last 16-byte chunk. This chunk may overlap with the + // chunks we've already searched, but that's ok. + MOVQ AX, DI + MOVOU (AX), X1 + PCMPEQB X0, X1 + PMOVMSKB X1, DX + BSFL DX, DX + JNZ ssesuccess + +failure: + MOVQ $-1, (R8) + RET + +// We've found a chunk containing the byte. +// The chunk was loaded from DI. +// The index of the matching byte in the chunk is DX. +// The start of the data is SI. +ssesuccess: + SUBQ SI, DI // Compute offset of chunk within data. + ADDQ DX, DI // Add offset of byte within chunk. + MOVQ DI, (R8) + RET + +// handle for lengths < 16 +small: + TESTQ BX, BX + JEQ failure + + // Check if we'll load across a page boundary. + LEAQ 16(SI), AX + TESTW $0xff0, AX + JEQ endofpage + + MOVOU (SI), X1 // Load data + PCMPEQB X0, X1 // Compare target byte with each byte in data. + PMOVMSKB X1, DX // Move result bits to integer register. + BSFL DX, DX // Find first set bit. + JZ failure // No set bit, failure. + CMPL DX, BX + JAE failure // Match is past end of data. + MOVQ DX, (R8) + RET + +endofpage: + MOVOU -16(SI)(BX*1), X1 // Load data into the high end of X1. + PCMPEQB X0, X1 // Compare target byte with each byte in data. + PMOVMSKB X1, DX // Move result bits to integer register. + MOVL BX, CX + SHLL CX, DX + SHRL $16, DX // Shift desired bits down to bottom of register. + BSFL DX, DX // Find first set bit. + JZ failure // No set bit, failure. + MOVQ DX, (R8) + RET + +avx2: +#ifndef hasAVX2 + CMPB internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1 + JNE sse +#endif + MOVD AX, X0 + LEAQ -32(SI)(BX*1), R11 + VPBROADCASTB X0, Y1 +avx2_loop: + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPTEST Y3, Y3 + JNZ avx2success + ADDQ $32, DI + CMPQ DI, R11 + JLT avx2_loop + MOVQ R11, DI + VMOVDQU (DI), Y2 + VPCMPEQB Y1, Y2, Y3 + VPTEST Y3, Y3 + JNZ avx2success + VZEROUPPER + MOVQ $-1, (R8) + RET + +avx2success: + VPMOVMSKB Y3, DX + BSFL DX, DX + SUBQ SI, DI + ADDQ DI, DX + MOVQ DX, (R8) + VZEROUPPER + RET diff --git a/src/internal/bytealg/indexbyte_arm.s b/src/internal/bytealg/indexbyte_arm.s new file mode 100644 index 0000000..faf9797 --- /dev/null +++ b/src/internal/bytealg/indexbyte_arm.s @@ -0,0 +1,46 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-20 + MOVW b_base+0(FP), R0 + MOVW b_len+4(FP), R1 + MOVBU c+12(FP), R2 // byte to find + MOVW $ret+16(FP), R5 + B indexbytebody<>(SB) + +TEXT ·IndexByteString(SB),NOSPLIT,$0-16 + MOVW s_base+0(FP), R0 + MOVW s_len+4(FP), R1 + MOVBU c+8(FP), R2 // byte to find + MOVW $ret+12(FP), R5 + B indexbytebody<>(SB) + +// input: +// R0: data +// R1: data length +// R2: byte to find +// R5: address to put result +TEXT indexbytebody<>(SB),NOSPLIT,$0-0 + MOVW R0, R4 // store base for later + ADD R0, R1 // end + +loop: + CMP R0, R1 + B.EQ notfound + MOVBU.P 1(R0), R3 + CMP R2, R3 + B.NE loop + + SUB $1, R0 // R0 will be one beyond the position we want + SUB R4, R0 // remove base + MOVW R0, (R5) + RET + +notfound: + MOVW $-1, R0 + MOVW R0, (R5) + RET diff --git a/src/internal/bytealg/indexbyte_arm64.s b/src/internal/bytealg/indexbyte_arm64.s new file mode 100644 index 0000000..40843fb --- /dev/null +++ b/src/internal/bytealg/indexbyte_arm64.s @@ -0,0 +1,126 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-40 + MOVD b_base+0(FP), R0 + MOVD b_len+8(FP), R2 + MOVBU c+24(FP), R1 + MOVD $ret+32(FP), R8 + B indexbytebody<>(SB) + +TEXT ·IndexByteString(SB),NOSPLIT,$0-32 + MOVD s_base+0(FP), R0 + MOVD s_len+8(FP), R2 + MOVBU c+16(FP), R1 + MOVD $ret+24(FP), R8 + B indexbytebody<>(SB) + +// input: +// R0: data +// R1: byte to search +// R2: data len +// R8: address to put result +TEXT indexbytebody<>(SB),NOSPLIT,$0 + // Core algorithm: + // For each 32-byte chunk we calculate a 64-bit syndrome value, + // with two bits per byte. For each tuple, bit 0 is set if the + // relevant byte matched the requested character and bit 1 is + // not used (faster than using a 32bit syndrome). Since the bits + // in the syndrome reflect exactly the order in which things occur + // in the original string, counting trailing zeros allows to + // identify exactly which byte has matched. + + CBZ R2, fail + MOVD R0, R11 + // Magic constant 0x40100401 allows us to identify + // which lane matches the requested byte. + // 0x40100401 = ((1<<0) + (4<<8) + (16<<16) + (64<<24)) + // Different bytes have different bit masks (i.e: 1, 4, 16, 64) + MOVD $0x40100401, R5 + VMOV R1, V0.B16 + // Work with aligned 32-byte chunks + BIC $0x1f, R0, R3 + VMOV R5, V5.S4 + ANDS $0x1f, R0, R9 + AND $0x1f, R2, R10 + BEQ loop + + // Input string is not 32-byte aligned. We calculate the + // syndrome value for the aligned 32 bytes block containing + // the first bytes and mask off the irrelevant part. + VLD1.P (R3), [V1.B16, V2.B16] + SUB $0x20, R9, R4 + ADDS R4, R2, R2 + VCMEQ V0.B16, V1.B16, V3.B16 + VCMEQ V0.B16, V2.B16, V4.B16 + VAND V5.B16, V3.B16, V3.B16 + VAND V5.B16, V4.B16, V4.B16 + VADDP V4.B16, V3.B16, V6.B16 // 256->128 + VADDP V6.B16, V6.B16, V6.B16 // 128->64 + VMOV V6.D[0], R6 + // Clear the irrelevant lower bits + LSL $1, R9, R4 + LSR R4, R6, R6 + LSL R4, R6, R6 + // The first block can also be the last + BLS masklast + // Have we found something already? + CBNZ R6, tail + +loop: + VLD1.P (R3), [V1.B16, V2.B16] + SUBS $0x20, R2, R2 + VCMEQ V0.B16, V1.B16, V3.B16 + VCMEQ V0.B16, V2.B16, V4.B16 + // If we're out of data we finish regardless of the result + BLS end + // Use a fast check for the termination condition + VORR V4.B16, V3.B16, V6.B16 + VADDP V6.D2, V6.D2, V6.D2 + VMOV V6.D[0], R6 + // We're not out of data, loop if we haven't found the character + CBZ R6, loop + +end: + // Termination condition found, let's calculate the syndrome value + VAND V5.B16, V3.B16, V3.B16 + VAND V5.B16, V4.B16, V4.B16 + VADDP V4.B16, V3.B16, V6.B16 + VADDP V6.B16, V6.B16, V6.B16 + VMOV V6.D[0], R6 + // Only do the clear for the last possible block with less than 32 bytes + // Condition flags come from SUBS in the loop + BHS tail + +masklast: + // Clear the irrelevant upper bits + ADD R9, R10, R4 + AND $0x1f, R4, R4 + SUB $0x20, R4, R4 + NEG R4<<1, R4 + LSL R4, R6, R6 + LSR R4, R6, R6 + +tail: + // Check that we have found a character + CBZ R6, fail + // Count the trailing zeros using bit reversing + RBIT R6, R6 + // Compensate the last post-increment + SUB $0x20, R3, R3 + // And count the leading zeros + CLZ R6, R6 + // R6 is twice the offset into the fragment + ADD R6>>1, R3, R0 + // Compute the offset result + SUB R11, R0, R0 + MOVD R0, (R8) + RET + +fail: + MOVD $-1, R0 + MOVD R0, (R8) + RET diff --git a/src/internal/bytealg/indexbyte_generic.go b/src/internal/bytealg/indexbyte_generic.go new file mode 100644 index 0000000..b89d34f --- /dev/null +++ b/src/internal/bytealg/indexbyte_generic.go @@ -0,0 +1,25 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !386 && !amd64 && !s390x && !arm && !arm64 && !loong64 && !ppc64 && !ppc64le && !mips && !mipsle && !mips64 && !mips64le && !riscv64 && !wasm + +package bytealg + +func IndexByte(b []byte, c byte) int { + for i, x := range b { + if x == c { + return i + } + } + return -1 +} + +func IndexByteString(s string, c byte) int { + for i := 0; i < len(s); i++ { + if s[i] == c { + return i + } + } + return -1 +} diff --git a/src/internal/bytealg/indexbyte_loong64.s b/src/internal/bytealg/indexbyte_loong64.s new file mode 100644 index 0000000..6049705 --- /dev/null +++ b/src/internal/bytealg/indexbyte_loong64.s @@ -0,0 +1,54 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-40 + MOVV b_base+0(FP), R4 + MOVV b_len+8(FP), R5 + MOVBU c+24(FP), R6 // byte to find + MOVV R4, R7 // store base for later + ADDV R4, R5 // end + ADDV $-1, R4 + + PCALIGN $16 +loop: + ADDV $1, R4 + BEQ R4, R5, notfound + MOVBU (R4), R8 + BNE R6, R8, loop + + SUBV R7, R4 // remove base + MOVV R4, ret+32(FP) + RET + +notfound: + MOVV $-1, R4 + MOVV R4, ret+32(FP) + RET + +TEXT ·IndexByteString(SB),NOSPLIT,$0-32 + MOVV s_base+0(FP), R4 + MOVV s_len+8(FP), R5 + MOVBU c+16(FP), R6 // byte to find + MOVV R4, R7 // store base for later + ADDV R4, R5 // end + ADDV $-1, R4 + + PCALIGN $16 +loop: + ADDV $1, R4 + BEQ R4, R5, notfound + MOVBU (R4), R8 + BNE R6, R8, loop + + SUBV R7, R4 // remove base + MOVV R4, ret+24(FP) + RET + +notfound: + MOVV $-1, R4 + MOVV R4, ret+24(FP) + RET diff --git a/src/internal/bytealg/indexbyte_mips64x.s b/src/internal/bytealg/indexbyte_mips64x.s new file mode 100644 index 0000000..5689f84 --- /dev/null +++ b/src/internal/bytealg/indexbyte_mips64x.s @@ -0,0 +1,54 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips64 || mips64le + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-40 + MOVV b_base+0(FP), R1 + MOVV b_len+8(FP), R2 + MOVBU c+24(FP), R3 // byte to find + MOVV R1, R4 // store base for later + ADDV R1, R2 // end + ADDV $-1, R1 + +loop: + ADDV $1, R1 + BEQ R1, R2, notfound + MOVBU (R1), R5 + BNE R3, R5, loop + + SUBV R4, R1 // remove base + MOVV R1, ret+32(FP) + RET + +notfound: + MOVV $-1, R1 + MOVV R1, ret+32(FP) + RET + +TEXT ·IndexByteString(SB),NOSPLIT,$0-32 + MOVV s_base+0(FP), R1 + MOVV s_len+8(FP), R2 + MOVBU c+16(FP), R3 // byte to find + MOVV R1, R4 // store base for later + ADDV R1, R2 // end + ADDV $-1, R1 + +loop: + ADDV $1, R1 + BEQ R1, R2, notfound + MOVBU (R1), R5 + BNE R3, R5, loop + + SUBV R4, R1 // remove base + MOVV R1, ret+24(FP) + RET + +notfound: + MOVV $-1, R1 + MOVV R1, ret+24(FP) + RET diff --git a/src/internal/bytealg/indexbyte_mipsx.s b/src/internal/bytealg/indexbyte_mipsx.s new file mode 100644 index 0000000..1c2b104 --- /dev/null +++ b/src/internal/bytealg/indexbyte_mipsx.s @@ -0,0 +1,52 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips || mipsle + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT,$0-20 + MOVW b_base+0(FP), R1 + MOVW b_len+4(FP), R2 + MOVBU c+12(FP), R3 // byte to find + ADDU $1, R1, R4 // store base+1 for later + ADDU R1, R2 // end + +loop: + BEQ R1, R2, notfound + MOVBU (R1), R5 + ADDU $1, R1 + BNE R3, R5, loop + + SUBU R4, R1 // R1 will be one beyond the position we want so remove (base+1) + MOVW R1, ret+16(FP) + RET + +notfound: + MOVW $-1, R1 + MOVW R1, ret+16(FP) + RET + +TEXT ·IndexByteString(SB),NOSPLIT,$0-16 + MOVW s_base+0(FP), R1 + MOVW s_len+4(FP), R2 + MOVBU c+8(FP), R3 // byte to find + ADDU $1, R1, R4 // store base+1 for later + ADDU R1, R2 // end + +loop: + BEQ R1, R2, notfound + MOVBU (R1), R5 + ADDU $1, R1 + BNE R3, R5, loop + + SUBU R4, R1 // remove (base+1) + MOVW R1, ret+12(FP) + RET + +notfound: + MOVW $-1, R1 + MOVW R1, ret+12(FP) + RET diff --git a/src/internal/bytealg/indexbyte_native.go b/src/internal/bytealg/indexbyte_native.go new file mode 100644 index 0000000..c5bb2df --- /dev/null +++ b/src/internal/bytealg/indexbyte_native.go @@ -0,0 +1,13 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 || s390x || arm || arm64 || loong64 || ppc64 || ppc64le || mips || mipsle || mips64 || mips64le || riscv64 || wasm + +package bytealg + +//go:noescape +func IndexByte(b []byte, c byte) int + +//go:noescape +func IndexByteString(s string, c byte) int diff --git a/src/internal/bytealg/indexbyte_ppc64x.s b/src/internal/bytealg/indexbyte_ppc64x.s new file mode 100644 index 0000000..b6714f4 --- /dev/null +++ b/src/internal/bytealg/indexbyte_ppc64x.s @@ -0,0 +1,314 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-40 + // R3 = byte array pointer + // R4 = length + MOVD R6, R5 // R5 = byte + BR indexbytebody<>(SB) + +TEXT ·IndexByteString<ABIInternal>(SB),NOSPLIT|NOFRAME,$0-32 + // R3 = string + // R4 = length + // R5 = byte + BR indexbytebody<>(SB) + +#ifndef GOPPC64_power9 +#ifdef GOARCH_ppc64le +DATA indexbytevbperm<>+0(SB)/8, $0x3830282018100800 +DATA indexbytevbperm<>+8(SB)/8, $0x7870686058504840 +#else +DATA indexbytevbperm<>+0(SB)/8, $0x0008101820283038 +DATA indexbytevbperm<>+8(SB)/8, $0x4048505860687078 +#endif +GLOBL indexbytevbperm<>+0(SB), RODATA, $16 +#endif + +// Some operations are endian specific, choose the correct opcode base on GOARCH. +// Note, _VCZBEBB is only available on power9 and newer. +#ifdef GOARCH_ppc64le +#define _LDBEX MOVDBR +#define _LWBEX MOVWBR +#define _LHBEX MOVHBR +#define _VCZBEBB VCTZLSBB +#else +#define _LDBEX MOVD +#define _LWBEX MOVW +#define _LHBEX MOVH +#define _VCZBEBB VCLZLSBB +#endif + +// R3 = addr of string +// R4 = len of string +// R5 = byte to find +// On exit: +// R3 = return value +TEXT indexbytebody<>(SB),NOSPLIT|NOFRAME,$0-0 + CMPU R4,$32 + +#ifndef GOPPC64_power9 + // Load VBPERMQ constant to reduce compare into an ordered bit mask. + MOVD $indexbytevbperm<>+00(SB),R16 + LXVD2X (R16),V0 // Set up swap string +#endif + + MTVRD R5,V1 + VSPLTB $7,V1,V1 // Replicate byte across V1 + + BLT cmp16 // Jump to the small string case if it's <32 bytes. + + CMP R4,$64,CR1 + MOVD $16,R11 + MOVD R3,R8 + BLT CR1,cmp32 // Special case for length 32 - 63 + MOVD $32,R12 + MOVD $48,R6 + + RLDICR $0,R4,$63-6,R9 // R9 = len &^ 63 + ADD R3,R9,R9 // R9 = &s[len &^ 63] + ANDCC $63,R4 // (len &= 63) cmp 0. + + PCALIGN $16 +loop64: + LXVD2X (R0)(R8),V2 // Scan 64 bytes at a time, starting at &s[0] + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 // Match found at R8, jump out + + LXVD2X (R11)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat1 // Match found at R8+16 bytes, jump out + + LXVD2X (R12)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat2 // Match found at R8+32 bytes, jump out + + LXVD2X (R6)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat3 // Match found at R8+48 bytes, jump out + + ADD $64,R8 + CMPU R8,R9,CR1 + BNE CR1,loop64 // R8 != &s[len &^ 63]? + + PCALIGN $32 + BEQ notfound // Is tail length 0? CR0 is set before entering loop64. + + CMP R4,$32 // Tail length >= 32, use cmp32 path. + CMP R4,$16,CR1 + BGE cmp32 + + ADD R8,R4,R9 + ADD $-16,R9 + BLE CR1,cmp64_tail_gt0 + +cmp64_tail_gt16: // Tail length 17 - 32 + LXVD2X (R0)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 + +cmp64_tail_gt0: // Tail length 1 - 16 + MOVD R9,R8 + LXVD2X (R0)(R9),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 + + BR notfound + +cmp32: // Length 32 - 63 + + // Bytes 0 - 15 + LXVD2X (R0)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 + + // Bytes 16 - 31 + LXVD2X (R8)(R11),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat1 // Match found at R8+16 bytes, jump out + + BEQ notfound // Is length <= 32? (CR0 holds this comparison on entry to cmp32) + CMP R4,$48 + + ADD R4,R8,R9 // Compute &s[len(s)-16] + ADD $32,R8,R8 + ADD $-16,R9,R9 + ISEL CR0GT,R8,R9,R8 // R8 = len(s) <= 48 ? R9 : R8 + + // Bytes 33 - 47 + LXVD2X (R0)(R8),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 // match found at R8+32 bytes, jump out + + BLE notfound + + // Bytes 48 - 63 + MOVD R9,R8 // R9 holds the final check. + LXVD2X (R0)(R9),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 // Match found at R8+48 bytes, jump out + + BR notfound + +// If ISA 3.0 instructions are unavailable, we need to account for the extra 16 added by CNTLZW. +#ifndef GOPPC64_power9 +#define ADJUST_FOR_CNTLZW -16 +#else +#define ADJUST_FOR_CNTLZW 0 +#endif + +// Now, find the index of the 16B vector the match was discovered in. If CNTLZW is used +// to determine the offset into the 16B vector, it will overcount by 16. Account for it here. +foundat3: + SUB R3,R8,R3 + ADD $48+ADJUST_FOR_CNTLZW,R3 + BR vfound +foundat2: + SUB R3,R8,R3 + ADD $32+ADJUST_FOR_CNTLZW,R3 + BR vfound +foundat1: + SUB R3,R8,R3 + ADD $16+ADJUST_FOR_CNTLZW,R3 + BR vfound +foundat0: + SUB R3,R8,R3 + ADD $0+ADJUST_FOR_CNTLZW,R3 +vfound: + // Map equal values into a 16 bit value with earlier matches setting higher bits. +#ifndef GOPPC64_power9 + VBPERMQ V6,V0,V6 + MFVRD V6,R4 + CNTLZW R4,R4 +#else +#ifdef GOARCH_ppc64le + // Put the value back into LE ordering by swapping doublewords. + XXPERMDI V6,V6,$2,V6 +#endif + _VCZBEBB V6,R4 +#endif + ADD R3,R4,R3 + RET + +cmp16: // Length 16 - 31 + CMPU R4,$16 + ADD R4,R3,R9 + BLT cmp8 + + ADD $-16,R9,R9 // &s[len(s)-16] + + // Bytes 0 - 15 + LXVD2X (R0)(R3),V2 + VCMPEQUBCC V2,V1,V6 + MOVD R3,R8 + BNE CR6,foundat0 // Match found at R8+32 bytes, jump out + + BEQ notfound + + // Bytes 16 - 30 + MOVD R9,R8 // R9 holds the final check. + LXVD2X (R0)(R9),V2 + VCMPEQUBCC V2,V1,V6 + BNE CR6,foundat0 // Match found at R8+48 bytes, jump out + + BR notfound + + +cmp8: // Length 8 - 15 +#ifdef GOPPC64_power10 + // Load all the bytes into a single VSR in BE order. + SLD $56,R4,R5 + LXVLL R3,R5,V2 + // Compare and count the number which don't match. + VCMPEQUB V2,V1,V6 + VCLZLSBB V6,R3 + // If count is the number of bytes, or more. No matches are found. + CMPU R3,R4 + MOVD $-1,R5 + // Otherwise, the count is the index of the first match. + ISEL CR0LT,R3,R5,R3 + RET +#else + RLDIMI $8,R5,$48,R5 // Replicating the byte across the register. + RLDIMI $16,R5,$32,R5 + RLDIMI $32,R5,$0,R5 + CMPU R4,$8 + BLT cmp4 + MOVD $-8,R11 + ADD $-8,R4,R4 + + _LDBEX (R0)(R3),R10 + _LDBEX (R11)(R9),R11 + CMPB R10,R5,R10 + CMPB R11,R5,R11 + CMPU R10,$0 + CMPU R11,$0,CR1 + CNTLZD R10,R10 + CNTLZD R11,R11 + SRD $3,R10,R3 + SRD $3,R11,R11 + BNE found + + ADD R4,R11,R4 + MOVD $-1,R3 + ISEL CR1EQ,R3,R4,R3 + RET + +cmp4: // Length 4 - 7 + CMPU R4,$4 + BLT cmp2 + MOVD $-4,R11 + ADD $-4,R4,R4 + + _LWBEX (R0)(R3),R10 + _LWBEX (R11)(R9),R11 + CMPB R10,R5,R10 + CMPB R11,R5,R11 + CNTLZW R10,R10 + CNTLZW R11,R11 + CMPU R10,$32 + CMPU R11,$32,CR1 + SRD $3,R10,R3 + SRD $3,R11,R11 + BNE found + + ADD R4,R11,R4 + MOVD $-1,R3 + ISEL CR1EQ,R3,R4,R3 + RET + +cmp2: // Length 2 - 3 + CMPU R4,$2 + BLT cmp1 + + _LHBEX (R0)(R3),R10 + CMPB R10,R5,R10 + SLDCC $48,R10,R10 + CNTLZD R10,R10 + SRD $3,R10,R3 + BNE found + +cmp1: // Length 1 + MOVD $-1,R3 + ANDCC $1,R4,R31 + BEQ found + + MOVBZ -1(R9),R10 + CMPB R10,R5,R10 + ANDCC $1,R10 + ADD $-1,R4 + ISEL CR0EQ,R3,R4,R3 + +found: + RET +#endif + +notfound: + MOVD $-1,R3 + RET + diff --git a/src/internal/bytealg/indexbyte_riscv64.s b/src/internal/bytealg/indexbyte_riscv64.s new file mode 100644 index 0000000..8be78ed --- /dev/null +++ b/src/internal/bytealg/indexbyte_riscv64.s @@ -0,0 +1,51 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte<ABIInternal>(SB),NOSPLIT,$0-40 + // X10 = b_base + // X11 = b_len + // X12 = b_cap (unused) + // X13 = byte to find + AND $0xff, X13 + MOV X10, X12 // store base for later + ADD X10, X11 // end + ADD $-1, X10 + +loop: + ADD $1, X10 + BEQ X10, X11, notfound + MOVBU (X10), X14 + BNE X13, X14, loop + + SUB X12, X10 // remove base + RET + +notfound: + MOV $-1, X10 + RET + +TEXT ·IndexByteString<ABIInternal>(SB),NOSPLIT,$0-32 + // X10 = b_base + // X11 = b_len + // X12 = byte to find + AND $0xff, X12 + MOV X10, X13 // store base for later + ADD X10, X11 // end + ADD $-1, X10 + +loop: + ADD $1, X10 + BEQ X10, X11, notfound + MOVBU (X10), X14 + BNE X12, X14, loop + + SUB X13, X10 // remove base + RET + +notfound: + MOV $-1, X10 + RET diff --git a/src/internal/bytealg/indexbyte_s390x.s b/src/internal/bytealg/indexbyte_s390x.s new file mode 100644 index 0000000..cf88d92 --- /dev/null +++ b/src/internal/bytealg/indexbyte_s390x.s @@ -0,0 +1,108 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB),NOSPLIT|NOFRAME,$0-40 + MOVD b_base+0(FP), R3// b_base => R3 + MOVD b_len+8(FP), R4 // b_len => R4 + MOVBZ c+24(FP), R5 // c => R5 + MOVD $ret+32(FP), R2 // &ret => R9 + BR indexbytebody<>(SB) + +TEXT ·IndexByteString(SB),NOSPLIT|NOFRAME,$0-32 + MOVD s_base+0(FP), R3// s_base => R3 + MOVD s_len+8(FP), R4 // s_len => R4 + MOVBZ c+16(FP), R5 // c => R5 + MOVD $ret+24(FP), R2 // &ret => R9 + BR indexbytebody<>(SB) + +// input: +// R3: s +// R4: s_len +// R5: c -- byte sought +// R2: &ret -- address to put index into +TEXT indexbytebody<>(SB),NOSPLIT|NOFRAME,$0 + CMPBEQ R4, $0, notfound + MOVD R3, R6 // store base for later + ADD R3, R4, R8 // the address after the end of the string + //if the length is small, use loop; otherwise, use vector or srst search + CMPBGE R4, $16, large + +residual: + CMPBEQ R3, R8, notfound + MOVBZ 0(R3), R7 + LA 1(R3), R3 + CMPBNE R7, R5, residual + +found: + SUB R6, R3 + SUB $1, R3 + MOVD R3, 0(R2) + RET + +notfound: + MOVD $-1, 0(R2) + RET + +large: + MOVBZ internal∕cpu·S390X+const_offsetS390xHasVX(SB), R1 + CMPBNE R1, $0, vectorimpl + +srstimpl: // no vector facility + MOVBZ R5, R0 // c needs to be in R0, leave until last minute as currently R0 is expected to be 0 +srstloop: + WORD $0xB25E0083 // srst %r8, %r3 (search the range [R3, R8)) + BVS srstloop // interrupted - continue + BGT notfoundr0 +foundr0: + XOR R0, R0 // reset R0 + SUB R6, R8 // remove base + MOVD R8, 0(R2) + RET +notfoundr0: + XOR R0, R0 // reset R0 + MOVD $-1, 0(R2) + RET + +vectorimpl: + //if the address is not 16byte aligned, use loop for the header + MOVD R3, R8 + AND $15, R8 + CMPBGT R8, $0, notaligned + +aligned: + ADD R6, R4, R8 + MOVD R8, R7 + AND $-16, R7 + // replicate c across V17 + VLVGB $0, R5, V19 + VREPB $0, V19, V17 + +vectorloop: + CMPBGE R3, R7, residual + VL 0(R3), V16 // load string to be searched into V16 + ADD $16, R3 + VFEEBS V16, V17, V18 // search V17 in V16 and set conditional code accordingly + BVS vectorloop + + // when vector search found c in the string + VLGVB $7, V18, R7 // load 7th element of V18 containing index into R7 + SUB $16, R3 + SUB R6, R3 + ADD R3, R7 + MOVD R7, 0(R2) + RET + +notaligned: + MOVD R3, R8 + AND $-16, R8 + ADD $16, R8 +notalignedloop: + CMPBEQ R3, R8, aligned + MOVBZ 0(R3), R7 + LA 1(R3), R3 + CMPBNE R7, R5, notalignedloop + BR found diff --git a/src/internal/bytealg/indexbyte_wasm.s b/src/internal/bytealg/indexbyte_wasm.s new file mode 100644 index 0000000..ef4bd93 --- /dev/null +++ b/src/internal/bytealg/indexbyte_wasm.s @@ -0,0 +1,195 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "go_asm.h" +#include "textflag.h" + +TEXT ·IndexByte(SB), NOSPLIT, $0-40 + I64Load b_base+0(FP) + I32WrapI64 + I32Load8U c+24(FP) + I64Load b_len+8(FP) + I32WrapI64 + Call memchr<>(SB) + I64ExtendI32S + Set R0 + + Get SP + I64Const $-1 + Get R0 + I64Load b_base+0(FP) + I64Sub + Get R0 + I64Eqz $0 + Select + I64Store ret+32(FP) + + RET + +TEXT ·IndexByteString(SB), NOSPLIT, $0-32 + Get SP + I64Load s_base+0(FP) + I32WrapI64 + I32Load8U c+16(FP) + I64Load s_len+8(FP) + I32WrapI64 + Call memchr<>(SB) + I64ExtendI32S + Set R0 + + I64Const $-1 + Get R0 + I64Load s_base+0(FP) + I64Sub + Get R0 + I64Eqz $0 + Select + I64Store ret+24(FP) + + RET + +// initially compiled with emscripten and then modified over time. +// params: +// R0: s +// R1: c +// R2: len +// ret: index +TEXT memchr<>(SB), NOSPLIT, $0 + Get R1 + Set R4 + Block + Block + Get R2 + I32Const $0 + I32Ne + Tee R3 + Get R0 + I32Const $3 + I32And + I32Const $0 + I32Ne + I32And + If + Loop + Get R0 + I32Load8U $0 + Get R1 + I32Eq + BrIf $2 + Get R2 + I32Const $-1 + I32Add + Tee R2 + I32Const $0 + I32Ne + Tee R3 + Get R0 + I32Const $1 + I32Add + Tee R0 + I32Const $3 + I32And + I32Const $0 + I32Ne + I32And + BrIf $0 + End + End + Get R3 + BrIf $0 + I32Const $0 + Set R1 + Br $1 + End + Get R0 + I32Load8U $0 + Get R4 + Tee R3 + I32Eq + If + Get R2 + Set R1 + Else + Get R4 + I32Const $16843009 + I32Mul + Set R4 + Block + Block + Get R2 + I32Const $3 + I32GtU + If + Get R2 + Set R1 + Loop + Get R0 + I32Load $0 + Get R4 + I32Xor + Tee R2 + I32Const $-2139062144 + I32And + I32Const $-2139062144 + I32Xor + Get R2 + I32Const $-16843009 + I32Add + I32And + I32Eqz + If + Get R0 + I32Const $4 + I32Add + Set R0 + Get R1 + I32Const $-4 + I32Add + Tee R1 + I32Const $3 + I32GtU + BrIf $1 + Br $3 + End + End + Else + Get R2 + Set R1 + Br $1 + End + Br $1 + End + Get R1 + I32Eqz + If + I32Const $0 + Set R1 + Br $3 + End + End + Loop + Get R0 + I32Load8U $0 + Get R3 + I32Eq + BrIf $2 + Get R0 + I32Const $1 + I32Add + Set R0 + Get R1 + I32Const $-1 + I32Add + Tee R1 + BrIf $0 + I32Const $0 + Set R1 + End + End + End + Get R0 + I32Const $0 + Get R1 + Select + Return diff --git a/src/internal/cfg/cfg.go b/src/internal/cfg/cfg.go new file mode 100644 index 0000000..2af0ec7 --- /dev/null +++ b/src/internal/cfg/cfg.go @@ -0,0 +1,70 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package cfg holds configuration shared by the Go command and internal/testenv. +// Definitions that don't need to be exposed outside of cmd/go should be in +// cmd/go/internal/cfg instead of this package. +package cfg + +// KnownEnv is a list of environment variables that affect the operation +// of the Go command. +const KnownEnv = ` + AR + CC + CGO_CFLAGS + CGO_CFLAGS_ALLOW + CGO_CFLAGS_DISALLOW + CGO_CPPFLAGS + CGO_CPPFLAGS_ALLOW + CGO_CPPFLAGS_DISALLOW + CGO_CXXFLAGS + CGO_CXXFLAGS_ALLOW + CGO_CXXFLAGS_DISALLOW + CGO_ENABLED + CGO_FFLAGS + CGO_FFLAGS_ALLOW + CGO_FFLAGS_DISALLOW + CGO_LDFLAGS + CGO_LDFLAGS_ALLOW + CGO_LDFLAGS_DISALLOW + CXX + FC + GCCGO + GO111MODULE + GO386 + GOAMD64 + GOARCH + GOARM + GOBIN + GOCACHE + GOCACHEPROG + GOENV + GOEXE + GOEXPERIMENT + GOFLAGS + GOGCCFLAGS + GOHOSTARCH + GOHOSTOS + GOINSECURE + GOMIPS + GOMIPS64 + GOMODCACHE + GONOPROXY + GONOSUMDB + GOOS + GOPATH + GOPPC64 + GOPRIVATE + GOPROXY + GOROOT + GOSUMDB + GOTMPDIR + GOTOOLCHAIN + GOTOOLDIR + GOVCS + GOWASM + GOWORK + GO_EXTLINK_ENABLED + PKG_CONFIG +` diff --git a/src/internal/coverage/calloc/batchcounteralloc.go b/src/internal/coverage/calloc/batchcounteralloc.go new file mode 100644 index 0000000..2b6495d --- /dev/null +++ b/src/internal/coverage/calloc/batchcounteralloc.go @@ -0,0 +1,29 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package calloc + +// This package contains a simple "batch" allocator for allocating +// coverage counters (slices of uint32 basically), for working with +// coverage data files. Collections of counter arrays tend to all be +// live/dead over the same time period, so a good fit for batch +// allocation. + +type BatchCounterAlloc struct { + pool []uint32 +} + +func (ca *BatchCounterAlloc) AllocateCounters(n int) []uint32 { + const chunk = 8192 + if n > cap(ca.pool) { + siz := chunk + if n > chunk { + siz = n + } + ca.pool = make([]uint32, siz) + } + rv := ca.pool[:n] + ca.pool = ca.pool[n:] + return rv +} diff --git a/src/internal/coverage/cformat/fmt_test.go b/src/internal/coverage/cformat/fmt_test.go new file mode 100644 index 0000000..f5ed01b --- /dev/null +++ b/src/internal/coverage/cformat/fmt_test.go @@ -0,0 +1,155 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cformat_test + +import ( + "internal/coverage" + "internal/coverage/cformat" + "slices" + "strings" + "testing" +) + +func TestBasics(t *testing.T) { + fm := cformat.NewFormatter(coverage.CtrModeAtomic) + + mku := func(stl, enl, nx uint32) coverage.CoverableUnit { + return coverage.CoverableUnit{ + StLine: stl, + EnLine: enl, + NxStmts: nx, + } + } + fn1units := []coverage.CoverableUnit{ + mku(10, 11, 2), + mku(15, 11, 1), + } + fn2units := []coverage.CoverableUnit{ + mku(20, 25, 3), + mku(30, 31, 2), + mku(33, 40, 7), + } + fn3units := []coverage.CoverableUnit{ + mku(99, 100, 1), + } + fm.SetPackage("my/pack1") + for k, u := range fn1units { + fm.AddUnit("p.go", "f1", false, u, uint32(k)) + } + for k, u := range fn2units { + fm.AddUnit("q.go", "f2", false, u, 0) + fm.AddUnit("q.go", "f2", false, u, uint32(k)) + } + fm.SetPackage("my/pack2") + for _, u := range fn3units { + fm.AddUnit("lit.go", "f3", true, u, 0) + } + + var b1, b2, b3, b4 strings.Builder + if err := fm.EmitTextual(&b1); err != nil { + t.Fatalf("EmitTextual returned %v", err) + } + wantText := strings.TrimSpace(` +mode: atomic +p.go:10.0,11.0 2 0 +p.go:15.0,11.0 1 1 +q.go:20.0,25.0 3 0 +q.go:30.0,31.0 2 1 +q.go:33.0,40.0 7 2 +lit.go:99.0,100.0 1 0`) + gotText := strings.TrimSpace(b1.String()) + if wantText != gotText { + t.Errorf("emit text: got:\n%s\nwant:\n%s\n", gotText, wantText) + } + + // Percent output with no aggregation. + noCoverPkg := "" + if err := fm.EmitPercent(&b2, noCoverPkg, false, false); err != nil { + t.Fatalf("EmitPercent returned %v", err) + } + wantPercent := strings.Fields(` + my/pack1 coverage: 66.7% of statements + my/pack2 coverage: 0.0% of statements +`) + gotPercent := strings.Fields(b2.String()) + if !slices.Equal(wantPercent, gotPercent) { + t.Errorf("emit percent: got:\n%+v\nwant:\n%+v\n", + gotPercent, wantPercent) + } + + // Percent mode with aggregation. + withCoverPkg := " in ./..." + if err := fm.EmitPercent(&b3, withCoverPkg, false, true); err != nil { + t.Fatalf("EmitPercent returned %v", err) + } + wantPercent = strings.Fields(` + coverage: 62.5% of statements in ./... +`) + gotPercent = strings.Fields(b3.String()) + if !slices.Equal(wantPercent, gotPercent) { + t.Errorf("emit percent: got:\n%+v\nwant:\n%+v\n", + gotPercent, wantPercent) + } + + if err := fm.EmitFuncs(&b4); err != nil { + t.Fatalf("EmitFuncs returned %v", err) + } + wantFuncs := strings.TrimSpace(` +p.go:10: f1 33.3% +q.go:20: f2 75.0% +total (statements) 62.5%`) + gotFuncs := strings.TrimSpace(b4.String()) + if wantFuncs != gotFuncs { + t.Errorf("emit funcs: got:\n%s\nwant:\n%s\n", gotFuncs, wantFuncs) + } + if false { + t.Logf("text is %s\n", b1.String()) + t.Logf("perc is %s\n", b2.String()) + t.Logf("perc2 is %s\n", b3.String()) + t.Logf("funcs is %s\n", b4.String()) + } +} + +func TestEmptyPackages(t *testing.T) { + + fm := cformat.NewFormatter(coverage.CtrModeAtomic) + fm.SetPackage("my/pack1") + fm.SetPackage("my/pack2") + + // No aggregation. + { + var b strings.Builder + noCoverPkg := "" + if err := fm.EmitPercent(&b, noCoverPkg, true, false); err != nil { + t.Fatalf("EmitPercent returned %v", err) + } + wantPercent := strings.Fields(` + my/pack1 coverage: [no statements] + my/pack2 coverage: [no statements] +`) + gotPercent := strings.Fields(b.String()) + if !slices.Equal(wantPercent, gotPercent) { + t.Errorf("emit percent: got:\n%+v\nwant:\n%+v\n", + gotPercent, wantPercent) + } + } + + // With aggregation. + { + var b strings.Builder + noCoverPkg := "" + if err := fm.EmitPercent(&b, noCoverPkg, true, true); err != nil { + t.Fatalf("EmitPercent returned %v", err) + } + wantPercent := strings.Fields(` + coverage: [no statements] +`) + gotPercent := strings.Fields(b.String()) + if !slices.Equal(wantPercent, gotPercent) { + t.Errorf("emit percent: got:\n%+v\nwant:\n%+v\n", + gotPercent, wantPercent) + } + } +} diff --git a/src/internal/coverage/cformat/format.go b/src/internal/coverage/cformat/format.go new file mode 100644 index 0000000..7e7a277 --- /dev/null +++ b/src/internal/coverage/cformat/format.go @@ -0,0 +1,352 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cformat + +// This package provides apis for producing human-readable summaries +// of coverage data (e.g. a coverage percentage for a given package or +// set of packages) and for writing data in the legacy test format +// emitted by "go test -coverprofile=<outfile>". +// +// The model for using these apis is to create a Formatter object, +// then make a series of calls to SetPackage and AddUnit passing in +// data read from coverage meta-data and counter-data files. E.g. +// +// myformatter := cformat.NewFormatter() +// ... +// for each package P in meta-data file: { +// myformatter.SetPackage(P) +// for each function F in P: { +// for each coverable unit U in F: { +// myformatter.AddUnit(U) +// } +// } +// } +// myformatter.EmitPercent(os.Stdout, "", true, true) +// myformatter.EmitTextual(somefile) +// +// These apis are linked into tests that are built with "-cover", and +// called at the end of test execution to produce text output or +// emit coverage percentages. + +import ( + "fmt" + "internal/coverage" + "internal/coverage/cmerge" + "io" + "sort" + "text/tabwriter" +) + +type Formatter struct { + // Maps import path to package state. + pm map[string]*pstate + // Records current package being visited. + pkg string + // Pointer to current package state. + p *pstate + // Counter mode. + cm coverage.CounterMode +} + +// pstate records package-level coverage data state: +// - a table of functions (file/fname/literal) +// - a map recording the index/ID of each func encountered so far +// - a table storing execution count for the coverable units in each func +type pstate struct { + // slice of unique functions + funcs []fnfile + // maps function to index in slice above (index acts as function ID) + funcTable map[fnfile]uint32 + + // A table storing coverage counts for each coverable unit. + unitTable map[extcu]uint32 +} + +// extcu encapsulates a coverable unit within some function. +type extcu struct { + fnfid uint32 // index into p.funcs slice + coverage.CoverableUnit +} + +// fnfile is a function-name/file-name tuple. +type fnfile struct { + file string + fname string + lit bool +} + +func NewFormatter(cm coverage.CounterMode) *Formatter { + return &Formatter{ + pm: make(map[string]*pstate), + cm: cm, + } +} + +// SetPackage tells the formatter that we're about to visit the +// coverage data for the package with the specified import path. +// Note that it's OK to call SetPackage more than once with the +// same import path; counter data values will be accumulated. +func (fm *Formatter) SetPackage(importpath string) { + if importpath == fm.pkg { + return + } + fm.pkg = importpath + ps, ok := fm.pm[importpath] + if !ok { + ps = new(pstate) + fm.pm[importpath] = ps + ps.unitTable = make(map[extcu]uint32) + ps.funcTable = make(map[fnfile]uint32) + } + fm.p = ps +} + +// AddUnit passes info on a single coverable unit (file, funcname, +// literal flag, range of lines, and counter value) to the formatter. +// Counter values will be accumulated where appropriate. +func (fm *Formatter) AddUnit(file string, fname string, isfnlit bool, unit coverage.CoverableUnit, count uint32) { + if fm.p == nil { + panic("AddUnit invoked before SetPackage") + } + fkey := fnfile{file: file, fname: fname, lit: isfnlit} + idx, ok := fm.p.funcTable[fkey] + if !ok { + idx = uint32(len(fm.p.funcs)) + fm.p.funcs = append(fm.p.funcs, fkey) + fm.p.funcTable[fkey] = idx + } + ukey := extcu{fnfid: idx, CoverableUnit: unit} + pcount := fm.p.unitTable[ukey] + var result uint32 + if fm.cm == coverage.CtrModeSet { + if count != 0 || pcount != 0 { + result = 1 + } + } else { + // Use saturating arithmetic. + result, _ = cmerge.SaturatingAdd(pcount, count) + } + fm.p.unitTable[ukey] = result +} + +// sortUnits sorts a slice of extcu objects in a package according to +// source position information (e.g. file and line). Note that we don't +// include function name as part of the sorting criteria, the thinking +// being that is better to provide things in the original source order. +func (p *pstate) sortUnits(units []extcu) { + sort.Slice(units, func(i, j int) bool { + ui := units[i] + uj := units[j] + ifile := p.funcs[ui.fnfid].file + jfile := p.funcs[uj.fnfid].file + if ifile != jfile { + return ifile < jfile + } + // NB: not taking function literal flag into account here (no + // need, since other fields are guaranteed to be distinct). + if units[i].StLine != units[j].StLine { + return units[i].StLine < units[j].StLine + } + if units[i].EnLine != units[j].EnLine { + return units[i].EnLine < units[j].EnLine + } + if units[i].StCol != units[j].StCol { + return units[i].StCol < units[j].StCol + } + if units[i].EnCol != units[j].EnCol { + return units[i].EnCol < units[j].EnCol + } + return units[i].NxStmts < units[j].NxStmts + }) +} + +// EmitTextual writes the accumulated coverage data in the legacy +// cmd/cover text format to the writer 'w'. We sort the data items by +// importpath, source file, and line number before emitting (this sorting +// is not explicitly mandated by the format, but seems like a good idea +// for repeatable/deterministic dumps). +func (fm *Formatter) EmitTextual(w io.Writer) error { + if fm.cm == coverage.CtrModeInvalid { + panic("internal error, counter mode unset") + } + if _, err := fmt.Fprintf(w, "mode: %s\n", fm.cm.String()); err != nil { + return err + } + pkgs := make([]string, 0, len(fm.pm)) + for importpath := range fm.pm { + pkgs = append(pkgs, importpath) + } + sort.Strings(pkgs) + for _, importpath := range pkgs { + p := fm.pm[importpath] + units := make([]extcu, 0, len(p.unitTable)) + for u := range p.unitTable { + units = append(units, u) + } + p.sortUnits(units) + for _, u := range units { + count := p.unitTable[u] + file := p.funcs[u.fnfid].file + if _, err := fmt.Fprintf(w, "%s:%d.%d,%d.%d %d %d\n", + file, u.StLine, u.StCol, + u.EnLine, u.EnCol, u.NxStmts, count); err != nil { + return err + } + } + } + return nil +} + +// EmitPercent writes out a "percentage covered" string to the writer 'w'. +func (fm *Formatter) EmitPercent(w io.Writer, covpkgs string, noteEmpty bool, aggregate bool) error { + pkgs := make([]string, 0, len(fm.pm)) + for importpath := range fm.pm { + pkgs = append(pkgs, importpath) + } + + rep := func(cov, tot uint64) error { + if tot != 0 { + if _, err := fmt.Fprintf(w, "coverage: %.1f%% of statements%s\n", + 100.0*float64(cov)/float64(tot), covpkgs); err != nil { + return err + } + } else if noteEmpty { + if _, err := fmt.Fprintf(w, "coverage: [no statements]\n"); err != nil { + return err + } + } + return nil + } + + sort.Strings(pkgs) + var totalStmts, coveredStmts uint64 + for _, importpath := range pkgs { + p := fm.pm[importpath] + if !aggregate { + totalStmts, coveredStmts = 0, 0 + } + for unit, count := range p.unitTable { + nx := uint64(unit.NxStmts) + totalStmts += nx + if count != 0 { + coveredStmts += nx + } + } + if !aggregate { + if _, err := fmt.Fprintf(w, "\t%s\t\t", importpath); err != nil { + return err + } + if err := rep(coveredStmts, totalStmts); err != nil { + return err + } + } + } + if aggregate { + if err := rep(coveredStmts, totalStmts); err != nil { + return err + } + } + + return nil +} + +// EmitFuncs writes out a function-level summary to the writer 'w'. A +// note on handling function literals: although we collect coverage +// data for unnamed literals, it probably does not make sense to +// include them in the function summary since there isn't any good way +// to name them (this is also consistent with the legacy cmd/cover +// implementation). We do want to include their counts in the overall +// summary however. +func (fm *Formatter) EmitFuncs(w io.Writer) error { + if fm.cm == coverage.CtrModeInvalid { + panic("internal error, counter mode unset") + } + perc := func(covered, total uint64) float64 { + if total == 0 { + total = 1 + } + return 100.0 * float64(covered) / float64(total) + } + tabber := tabwriter.NewWriter(w, 1, 8, 1, '\t', 0) + defer tabber.Flush() + allStmts := uint64(0) + covStmts := uint64(0) + + pkgs := make([]string, 0, len(fm.pm)) + for importpath := range fm.pm { + pkgs = append(pkgs, importpath) + } + sort.Strings(pkgs) + + // Emit functions for each package, sorted by import path. + for _, importpath := range pkgs { + p := fm.pm[importpath] + if len(p.unitTable) == 0 { + continue + } + units := make([]extcu, 0, len(p.unitTable)) + for u := range p.unitTable { + units = append(units, u) + } + + // Within a package, sort the units, then walk through the + // sorted array. Each time we hit a new function, emit the + // summary entry for the previous function, then make one last + // emit call at the end of the loop. + p.sortUnits(units) + fname := "" + ffile := "" + flit := false + var fline uint32 + var cstmts, tstmts uint64 + captureFuncStart := func(u extcu) { + fname = p.funcs[u.fnfid].fname + ffile = p.funcs[u.fnfid].file + flit = p.funcs[u.fnfid].lit + fline = u.StLine + } + emitFunc := func(u extcu) error { + // Don't emit entries for function literals (see discussion + // in function header comment above). + if !flit { + if _, err := fmt.Fprintf(tabber, "%s:%d:\t%s\t%.1f%%\n", + ffile, fline, fname, perc(cstmts, tstmts)); err != nil { + return err + } + } + captureFuncStart(u) + allStmts += tstmts + covStmts += cstmts + tstmts = 0 + cstmts = 0 + return nil + } + for k, u := range units { + if k == 0 { + captureFuncStart(u) + } else { + if fname != p.funcs[u.fnfid].fname { + // New function; emit entry for previous one. + if err := emitFunc(u); err != nil { + return err + } + } + } + tstmts += uint64(u.NxStmts) + count := p.unitTable[u] + if count != 0 { + cstmts += uint64(u.NxStmts) + } + } + if err := emitFunc(extcu{}); err != nil { + return err + } + } + if _, err := fmt.Fprintf(tabber, "%s\t%s\t%.1f%%\n", + "total", "(statements)", perc(covStmts, allStmts)); err != nil { + return err + } + return nil +} diff --git a/src/internal/coverage/cmddefs.go b/src/internal/coverage/cmddefs.go new file mode 100644 index 0000000..49376a4 --- /dev/null +++ b/src/internal/coverage/cmddefs.go @@ -0,0 +1,87 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package coverage + +// CoverPkgConfig is a bundle of information passed from the Go +// command to the cover command during "go build -cover" runs. The +// Go command creates and fills in a struct as below, then passes +// file containing the encoded JSON for the struct to the "cover" +// tool when instrumenting the source files in a Go package. +type CoverPkgConfig struct { + // File into which cmd/cover should emit summary info + // when instrumentation is complete. + OutConfig string + + // Import path for the package being instrumented. + PkgPath string + + // Package name. + PkgName string + + // Instrumentation granularity: one of "perfunc" or "perblock" (default) + Granularity string + + // Module path for this package (empty if no go.mod in use) + ModulePath string + + // Local mode indicates we're doing a coverage build or test of a + // package selected via local import path, e.g. "./..." or + // "./foo/bar" as opposed to a non-relative import path. See the + // corresponding field in cmd/go's PackageInternal struct for more + // info. + Local bool +} + +// CoverFixupConfig contains annotations/notes generated by the +// cmd/cover tool (during instrumentation) to be passed on to the +// compiler when the instrumented code is compiled. The cmd/cover tool +// creates a struct of this type, JSON-encodes it, and emits the +// result to a file, which the Go command then passes to the compiler +// when the instrumented package is built. +type CoverFixupConfig struct { + // Name of the variable (created by cmd/cover) containing the + // encoded meta-data for the package. + MetaVar string + + // Length of the meta-data. + MetaLen int + + // Hash computed by cmd/cover of the meta-data. + MetaHash string + + // Instrumentation strategy. For now this is always set to + // "normal", but in the future we may add new values (for example, + // if panic paths are instrumented, or if the instrumenter + // eliminates redundant counters). + Strategy string + + // Prefix assigned to the names of counter variables generated + // during instrumentation by cmd/cover. + CounterPrefix string + + // Name chosen for the package ID variable generated during + // instrumentation. + PkgIdVar string + + // Counter mode (e.g. set/count/atomic) + CounterMode string + + // Counter granularity (perblock or perfunc). + CounterGranularity string +} + +// MetaFilePaths contains information generated by the Go command and +// the read in by coverage test support functions within an executing +// "go test -cover" binary. +type MetaFileCollection struct { + ImportPaths []string + MetaFileFragments []string +} + +// Name of file within the "go test -cover" temp coverdir directory +// containing a list of meta-data files for packages being tested +// in a "go test -coverpkg=... ..." run. This constant is shared +// by the Go command and by the coverage runtime. +const MetaFilesFileName = "metafiles.txt" diff --git a/src/internal/coverage/cmerge/merge.go b/src/internal/coverage/cmerge/merge.go new file mode 100644 index 0000000..1339803 --- /dev/null +++ b/src/internal/coverage/cmerge/merge.go @@ -0,0 +1,127 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cmerge + +// package cmerge provides a few small utility APIs for helping +// with merging of counter data for a given function. + +import ( + "fmt" + "internal/coverage" + "math" +) + +type ModeMergePolicy uint8 + +const ( + ModeMergeStrict ModeMergePolicy = iota + ModeMergeRelaxed +) + +// Merger provides state and methods to help manage the process of +// merging together coverage counter data for a given function, for +// tools that need to implicitly merge counter as they read multiple +// coverage counter data files. +type Merger struct { + cmode coverage.CounterMode + cgran coverage.CounterGranularity + policy ModeMergePolicy + overflow bool +} + +func (cm *Merger) SetModeMergePolicy(policy ModeMergePolicy) { + cm.policy = policy +} + +// MergeCounters takes the counter values in 'src' and merges them +// into 'dst' according to the correct counter mode. +func (m *Merger) MergeCounters(dst, src []uint32) (error, bool) { + if len(src) != len(dst) { + return fmt.Errorf("merging counters: len(dst)=%d len(src)=%d", len(dst), len(src)), false + } + if m.cmode == coverage.CtrModeSet { + for i := 0; i < len(src); i++ { + if src[i] != 0 { + dst[i] = 1 + } + } + } else { + for i := 0; i < len(src); i++ { + dst[i] = m.SaturatingAdd(dst[i], src[i]) + } + } + ovf := m.overflow + m.overflow = false + return nil, ovf +} + +// Saturating add does a saturating addition of 'dst' and 'src', +// returning added value or math.MaxUint32 if there is an overflow. +// Overflows are recorded in case the client needs to track them. +func (m *Merger) SaturatingAdd(dst, src uint32) uint32 { + result, overflow := SaturatingAdd(dst, src) + if overflow { + m.overflow = true + } + return result +} + +// Saturating add does a saturating addition of 'dst' and 'src', +// returning added value or math.MaxUint32 plus an overflow flag. +func SaturatingAdd(dst, src uint32) (uint32, bool) { + d, s := uint64(dst), uint64(src) + sum := d + s + overflow := false + if uint64(uint32(sum)) != sum { + overflow = true + sum = math.MaxUint32 + } + return uint32(sum), overflow +} + +// SetModeAndGranularity records the counter mode and granularity for +// the current merge. In the specific case of merging across coverage +// data files from different binaries, where we're combining data from +// more than one meta-data file, we need to check for and resolve +// mode/granularity clashes. +func (cm *Merger) SetModeAndGranularity(mdf string, cmode coverage.CounterMode, cgran coverage.CounterGranularity) error { + if cm.cmode == coverage.CtrModeInvalid { + // Set merger mode based on what we're seeing here. + cm.cmode = cmode + cm.cgran = cgran + } else { + // Granularity clashes are always errors. + if cm.cgran != cgran { + return fmt.Errorf("counter granularity clash while reading meta-data file %s: previous file had %s, new file has %s", mdf, cm.cgran.String(), cgran.String()) + } + // Mode clashes are treated as errors if we're using the + // default strict policy. + if cm.cmode != cmode { + if cm.policy == ModeMergeStrict { + return fmt.Errorf("counter mode clash while reading meta-data file %s: previous file had %s, new file has %s", mdf, cm.cmode.String(), cmode.String()) + } + // In the case of a relaxed mode merge policy, upgrade + // mode if needed. + if cm.cmode < cmode { + cm.cmode = cmode + } + } + } + return nil +} + +func (cm *Merger) ResetModeAndGranularity() { + cm.cmode = coverage.CtrModeInvalid + cm.cgran = coverage.CtrGranularityInvalid + cm.overflow = false +} + +func (cm *Merger) Mode() coverage.CounterMode { + return cm.cmode +} + +func (cm *Merger) Granularity() coverage.CounterGranularity { + return cm.cgran +} diff --git a/src/internal/coverage/cmerge/merge_test.go b/src/internal/coverage/cmerge/merge_test.go new file mode 100644 index 0000000..0e6112a --- /dev/null +++ b/src/internal/coverage/cmerge/merge_test.go @@ -0,0 +1,118 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cmerge_test + +import ( + "fmt" + "internal/coverage" + "internal/coverage/cmerge" + "testing" +) + +func TestClash(t *testing.T) { + m := &cmerge.Merger{} + err := m.SetModeAndGranularity("mdf1.data", coverage.CtrModeSet, coverage.CtrGranularityPerBlock) + if err != nil { + t.Fatalf("unexpected clash: %v", err) + } + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeSet, coverage.CtrGranularityPerBlock) + if err != nil { + t.Fatalf("unexpected clash: %v", err) + } + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeCount, coverage.CtrGranularityPerBlock) + if err == nil { + t.Fatalf("expected mode clash, not found") + } + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeSet, coverage.CtrGranularityPerFunc) + if err == nil { + t.Fatalf("expected granularity clash, not found") + } + m.SetModeMergePolicy(cmerge.ModeMergeRelaxed) + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeCount, coverage.CtrGranularityPerBlock) + if err != nil { + t.Fatalf("unexpected clash: %v", err) + } + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeSet, coverage.CtrGranularityPerBlock) + if err != nil { + t.Fatalf("unexpected clash: %v", err) + } + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeAtomic, coverage.CtrGranularityPerBlock) + if err != nil { + t.Fatalf("unexpected clash: %v", err) + } + m.ResetModeAndGranularity() + err = m.SetModeAndGranularity("mdf1.data", coverage.CtrModeCount, coverage.CtrGranularityPerFunc) + if err != nil { + t.Fatalf("unexpected clash after reset: %v", err) + } +} + +func TestBasic(t *testing.T) { + scenarios := []struct { + cmode coverage.CounterMode + cgran coverage.CounterGranularity + src, dst, res []uint32 + iters int + merr bool + overflow bool + }{ + { + cmode: coverage.CtrModeSet, + cgran: coverage.CtrGranularityPerBlock, + src: []uint32{1, 0, 1}, + dst: []uint32{1, 1, 0}, + res: []uint32{1, 1, 1}, + iters: 2, + overflow: false, + }, + { + cmode: coverage.CtrModeCount, + cgran: coverage.CtrGranularityPerBlock, + src: []uint32{1, 0, 3}, + dst: []uint32{5, 7, 0}, + res: []uint32{6, 7, 3}, + iters: 1, + overflow: false, + }, + { + cmode: coverage.CtrModeCount, + cgran: coverage.CtrGranularityPerBlock, + src: []uint32{4294967200, 0, 3}, + dst: []uint32{4294967001, 7, 0}, + res: []uint32{4294967295, 7, 3}, + iters: 1, + overflow: true, + }, + } + + for k, scenario := range scenarios { + var err error + var ovf bool + m := &cmerge.Merger{} + mdf := fmt.Sprintf("file%d", k) + err = m.SetModeAndGranularity(mdf, scenario.cmode, scenario.cgran) + if err != nil { + t.Fatalf("case %d SetModeAndGranularity failed: %v", k, err) + } + for i := 0; i < scenario.iters; i++ { + err, ovf = m.MergeCounters(scenario.dst, scenario.src) + if ovf != scenario.overflow { + t.Fatalf("case %d overflow mismatch: got %v want %v", k, ovf, scenario.overflow) + } + if !scenario.merr && err != nil { + t.Fatalf("case %d unexpected err %v", k, err) + } + if scenario.merr && err == nil { + t.Fatalf("case %d expected err, not received", k) + } + for i := range scenario.dst { + if scenario.dst[i] != scenario.res[i] { + t.Fatalf("case %d: bad merge at %d got %d want %d", + k, i, scenario.dst[i], scenario.res[i]) + } + } + } + } +} diff --git a/src/internal/coverage/decodecounter/decodecounterfile.go b/src/internal/coverage/decodecounter/decodecounterfile.go new file mode 100644 index 0000000..83934fe --- /dev/null +++ b/src/internal/coverage/decodecounter/decodecounterfile.go @@ -0,0 +1,373 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package decodecounter + +import ( + "encoding/binary" + "fmt" + "internal/coverage" + "internal/coverage/slicereader" + "internal/coverage/stringtab" + "io" + "os" + "strconv" + "unsafe" +) + +// This file contains helpers for reading counter data files created +// during the executions of a coverage-instrumented binary. + +type CounterDataReader struct { + stab *stringtab.Reader + args map[string]string + osargs []string + goarch string // GOARCH setting from run that produced counter data + goos string // GOOS setting from run that produced counter data + mr io.ReadSeeker + hdr coverage.CounterFileHeader + ftr coverage.CounterFileFooter + shdr coverage.CounterSegmentHeader + u32b []byte + u8b []byte + fcnCount uint32 + segCount uint32 + debug bool +} + +func NewCounterDataReader(fn string, rs io.ReadSeeker) (*CounterDataReader, error) { + cdr := &CounterDataReader{ + mr: rs, + u32b: make([]byte, 4), + u8b: make([]byte, 1), + } + // Read header + if err := binary.Read(rs, binary.LittleEndian, &cdr.hdr); err != nil { + return nil, err + } + if cdr.debug { + fmt.Fprintf(os.Stderr, "=-= counter file header: %+v\n", cdr.hdr) + } + if !checkMagic(cdr.hdr.Magic) { + return nil, fmt.Errorf("invalid magic string: not a counter data file") + } + if cdr.hdr.Version > coverage.CounterFileVersion { + return nil, fmt.Errorf("version data incompatibility: reader is %d data is %d", coverage.CounterFileVersion, cdr.hdr.Version) + } + + // Read footer. + if err := cdr.readFooter(); err != nil { + return nil, err + } + // Seek back to just past the file header. + hsz := int64(unsafe.Sizeof(cdr.hdr)) + if _, err := cdr.mr.Seek(hsz, io.SeekStart); err != nil { + return nil, err + } + // Read preamble for first segment. + if err := cdr.readSegmentPreamble(); err != nil { + return nil, err + } + return cdr, nil +} + +func checkMagic(v [4]byte) bool { + g := coverage.CovCounterMagic + return v[0] == g[0] && v[1] == g[1] && v[2] == g[2] && v[3] == g[3] +} + +func (cdr *CounterDataReader) readFooter() error { + ftrSize := int64(unsafe.Sizeof(cdr.ftr)) + if _, err := cdr.mr.Seek(-ftrSize, io.SeekEnd); err != nil { + return err + } + if err := binary.Read(cdr.mr, binary.LittleEndian, &cdr.ftr); err != nil { + return err + } + if !checkMagic(cdr.ftr.Magic) { + return fmt.Errorf("invalid magic string (not a counter data file)") + } + if cdr.ftr.NumSegments == 0 { + return fmt.Errorf("invalid counter data file (no segments)") + } + return nil +} + +// readSegmentPreamble reads and consumes the segment header, segment string +// table, and segment args table. +func (cdr *CounterDataReader) readSegmentPreamble() error { + // Read segment header. + if err := binary.Read(cdr.mr, binary.LittleEndian, &cdr.shdr); err != nil { + return err + } + if cdr.debug { + fmt.Fprintf(os.Stderr, "=-= read counter segment header: %+v", cdr.shdr) + fmt.Fprintf(os.Stderr, " FcnEntries=0x%x StrTabLen=0x%x ArgsLen=0x%x\n", + cdr.shdr.FcnEntries, cdr.shdr.StrTabLen, cdr.shdr.ArgsLen) + } + + // Read string table and args. + if err := cdr.readStringTable(); err != nil { + return err + } + if err := cdr.readArgs(); err != nil { + return err + } + // Seek past any padding to bring us up to a 4-byte boundary. + if of, err := cdr.mr.Seek(0, io.SeekCurrent); err != nil { + return err + } else { + rem := of % 4 + if rem != 0 { + pad := 4 - rem + if _, err := cdr.mr.Seek(pad, io.SeekCurrent); err != nil { + return err + } + } + } + return nil +} + +func (cdr *CounterDataReader) readStringTable() error { + b := make([]byte, cdr.shdr.StrTabLen) + nr, err := cdr.mr.Read(b) + if err != nil { + return err + } + if nr != int(cdr.shdr.StrTabLen) { + return fmt.Errorf("error: short read on string table") + } + slr := slicereader.NewReader(b, false /* not readonly */) + cdr.stab = stringtab.NewReader(slr) + cdr.stab.Read() + return nil +} + +func (cdr *CounterDataReader) readArgs() error { + b := make([]byte, cdr.shdr.ArgsLen) + nr, err := cdr.mr.Read(b) + if err != nil { + return err + } + if nr != int(cdr.shdr.ArgsLen) { + return fmt.Errorf("error: short read on args table") + } + slr := slicereader.NewReader(b, false /* not readonly */) + sget := func() (string, error) { + kidx := slr.ReadULEB128() + if int(kidx) >= cdr.stab.Entries() { + return "", fmt.Errorf("malformed string table ref") + } + return cdr.stab.Get(uint32(kidx)), nil + } + nents := slr.ReadULEB128() + cdr.args = make(map[string]string, int(nents)) + for i := uint64(0); i < nents; i++ { + k, errk := sget() + if errk != nil { + return errk + } + v, errv := sget() + if errv != nil { + return errv + } + if _, ok := cdr.args[k]; ok { + return fmt.Errorf("malformed args table") + } + cdr.args[k] = v + } + if argcs, ok := cdr.args["argc"]; ok { + argc, err := strconv.Atoi(argcs) + if err != nil { + return fmt.Errorf("malformed argc in counter data file args section") + } + cdr.osargs = make([]string, 0, argc) + for i := 0; i < argc; i++ { + arg := cdr.args[fmt.Sprintf("argv%d", i)] + cdr.osargs = append(cdr.osargs, arg) + } + } + if goos, ok := cdr.args["GOOS"]; ok { + cdr.goos = goos + } + if goarch, ok := cdr.args["GOARCH"]; ok { + cdr.goarch = goarch + } + return nil +} + +// OsArgs returns the program arguments (saved from os.Args during +// the run of the instrumented binary) read from the counter +// data file. Not all coverage data files will have os.Args values; +// for example, if a data file is produced by merging coverage +// data from two distinct runs, no os args will be available (an +// empty list is returned). +func (cdr *CounterDataReader) OsArgs() []string { + return cdr.osargs +} + +// Goos returns the GOOS setting in effect for the "-cover" binary +// that produced this counter data file. The GOOS value may be +// empty in the case where the counter data file was produced +// from a merge in which more than one GOOS value was present. +func (cdr *CounterDataReader) Goos() string { + return cdr.goos +} + +// Goarch returns the GOARCH setting in effect for the "-cover" binary +// that produced this counter data file. The GOARCH value may be +// empty in the case where the counter data file was produced +// from a merge in which more than one GOARCH value was present. +func (cdr *CounterDataReader) Goarch() string { + return cdr.goarch +} + +// FuncPayload encapsulates the counter data payload for a single +// function as read from a counter data file. +type FuncPayload struct { + PkgIdx uint32 + FuncIdx uint32 + Counters []uint32 +} + +// NumSegments returns the number of execution segments in the file. +func (cdr *CounterDataReader) NumSegments() uint32 { + return cdr.ftr.NumSegments +} + +// BeginNextSegment sets up the reader to read the next segment, +// returning TRUE if we do have another segment to read, or FALSE +// if we're done with all the segments (also an error if +// something went wrong). +func (cdr *CounterDataReader) BeginNextSegment() (bool, error) { + if cdr.segCount >= cdr.ftr.NumSegments { + return false, nil + } + cdr.segCount++ + cdr.fcnCount = 0 + // Seek past footer from last segment. + ftrSize := int64(unsafe.Sizeof(cdr.ftr)) + if _, err := cdr.mr.Seek(ftrSize, io.SeekCurrent); err != nil { + return false, err + } + // Read preamble for this segment. + if err := cdr.readSegmentPreamble(); err != nil { + return false, err + } + return true, nil +} + +// NumFunctionsInSegment returns the number of live functions +// in the currently selected segment. +func (cdr *CounterDataReader) NumFunctionsInSegment() uint32 { + return uint32(cdr.shdr.FcnEntries) +} + +const supportDeadFunctionsInCounterData = false + +// NextFunc reads data for the next function in this current segment +// into "p", returning TRUE if the read was successful or FALSE +// if we've read all the functions already (also an error if +// something went wrong with the read or we hit a premature +// EOF). +func (cdr *CounterDataReader) NextFunc(p *FuncPayload) (bool, error) { + if cdr.fcnCount >= uint32(cdr.shdr.FcnEntries) { + return false, nil + } + cdr.fcnCount++ + var rdu32 func() (uint32, error) + if cdr.hdr.CFlavor == coverage.CtrULeb128 { + rdu32 = func() (uint32, error) { + var shift uint + var value uint64 + for { + _, err := cdr.mr.Read(cdr.u8b) + if err != nil { + return 0, err + } + b := cdr.u8b[0] + value |= (uint64(b&0x7F) << shift) + if b&0x80 == 0 { + break + } + shift += 7 + } + return uint32(value), nil + } + } else if cdr.hdr.CFlavor == coverage.CtrRaw { + if cdr.hdr.BigEndian { + rdu32 = func() (uint32, error) { + n, err := cdr.mr.Read(cdr.u32b) + if err != nil { + return 0, err + } + if n != 4 { + return 0, io.EOF + } + return binary.BigEndian.Uint32(cdr.u32b), nil + } + } else { + rdu32 = func() (uint32, error) { + n, err := cdr.mr.Read(cdr.u32b) + if err != nil { + return 0, err + } + if n != 4 { + return 0, io.EOF + } + return binary.LittleEndian.Uint32(cdr.u32b), nil + } + } + } else { + panic("internal error: unknown counter flavor") + } + + // Alternative/experimental path: one way we could handling writing + // out counter data would be to just memcpy the counter segment + // out to a file, meaning that a region in the counter memory + // corresponding to a dead (never-executed) function would just be + // zeroes. The code path below handles this case. + var nc uint32 + var err error + if supportDeadFunctionsInCounterData { + for { + nc, err = rdu32() + if err == io.EOF { + return false, io.EOF + } else if err != nil { + break + } + if nc != 0 { + break + } + } + } else { + nc, err = rdu32() + } + if err != nil { + return false, err + } + + // Read package and func indices. + p.PkgIdx, err = rdu32() + if err != nil { + return false, err + } + p.FuncIdx, err = rdu32() + if err != nil { + return false, err + } + if cap(p.Counters) < 1024 { + p.Counters = make([]uint32, 0, 1024) + } + p.Counters = p.Counters[:0] + for i := uint32(0); i < nc; i++ { + v, err := rdu32() + if err != nil { + return false, err + } + p.Counters = append(p.Counters, v) + } + return true, nil +} diff --git a/src/internal/coverage/decodemeta/decode.go b/src/internal/coverage/decodemeta/decode.go new file mode 100644 index 0000000..fa047c7 --- /dev/null +++ b/src/internal/coverage/decodemeta/decode.go @@ -0,0 +1,136 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package decodemeta + +// This package contains APIs and helpers for decoding a single package's +// meta data "blob" emitted by the compiler when coverage instrumentation +// is turned on. + +import ( + "encoding/binary" + "fmt" + "internal/coverage" + "internal/coverage/slicereader" + "internal/coverage/stringtab" + "io" + "os" +) + +// See comments in the encodecovmeta package for details on the format. + +type CoverageMetaDataDecoder struct { + r *slicereader.Reader + hdr coverage.MetaSymbolHeader + strtab *stringtab.Reader + tmp []byte + debug bool +} + +func NewCoverageMetaDataDecoder(b []byte, readonly bool) (*CoverageMetaDataDecoder, error) { + slr := slicereader.NewReader(b, readonly) + x := &CoverageMetaDataDecoder{ + r: slr, + tmp: make([]byte, 0, 256), + } + if err := x.readHeader(); err != nil { + return nil, err + } + if err := x.readStringTable(); err != nil { + return nil, err + } + return x, nil +} + +func (d *CoverageMetaDataDecoder) readHeader() error { + if err := binary.Read(d.r, binary.LittleEndian, &d.hdr); err != nil { + return err + } + if d.debug { + fmt.Fprintf(os.Stderr, "=-= after readHeader: %+v\n", d.hdr) + } + return nil +} + +func (d *CoverageMetaDataDecoder) readStringTable() error { + // Seek to the correct location to read the string table. + stringTableLocation := int64(coverage.CovMetaHeaderSize + 4*d.hdr.NumFuncs) + if _, err := d.r.Seek(stringTableLocation, io.SeekStart); err != nil { + return err + } + + // Read the table itself. + d.strtab = stringtab.NewReader(d.r) + d.strtab.Read() + return nil +} + +func (d *CoverageMetaDataDecoder) PackagePath() string { + return d.strtab.Get(d.hdr.PkgPath) +} + +func (d *CoverageMetaDataDecoder) PackageName() string { + return d.strtab.Get(d.hdr.PkgName) +} + +func (d *CoverageMetaDataDecoder) ModulePath() string { + return d.strtab.Get(d.hdr.ModulePath) +} + +func (d *CoverageMetaDataDecoder) NumFuncs() uint32 { + return d.hdr.NumFuncs +} + +// ReadFunc reads the coverage meta-data for the function with index +// 'findex', filling it into the FuncDesc pointed to by 'f'. +func (d *CoverageMetaDataDecoder) ReadFunc(fidx uint32, f *coverage.FuncDesc) error { + if fidx >= d.hdr.NumFuncs { + return fmt.Errorf("illegal function index") + } + + // Seek to the correct location to read the function offset and read it. + funcOffsetLocation := int64(coverage.CovMetaHeaderSize + 4*fidx) + if _, err := d.r.Seek(funcOffsetLocation, io.SeekStart); err != nil { + return err + } + foff := d.r.ReadUint32() + + // Check assumptions + if foff < uint32(funcOffsetLocation) || foff > d.hdr.Length { + return fmt.Errorf("malformed func offset %d", foff) + } + + // Seek to the correct location to read the function. + floc := int64(foff) + if _, err := d.r.Seek(floc, io.SeekStart); err != nil { + return err + } + + // Preamble containing number of units, file, and function. + numUnits := uint32(d.r.ReadULEB128()) + fnameidx := uint32(d.r.ReadULEB128()) + fileidx := uint32(d.r.ReadULEB128()) + + f.Srcfile = d.strtab.Get(fileidx) + f.Funcname = d.strtab.Get(fnameidx) + + // Now the units + f.Units = f.Units[:0] + if cap(f.Units) < int(numUnits) { + f.Units = make([]coverage.CoverableUnit, 0, numUnits) + } + for k := uint32(0); k < numUnits; k++ { + f.Units = append(f.Units, + coverage.CoverableUnit{ + StLine: uint32(d.r.ReadULEB128()), + StCol: uint32(d.r.ReadULEB128()), + EnLine: uint32(d.r.ReadULEB128()), + EnCol: uint32(d.r.ReadULEB128()), + NxStmts: uint32(d.r.ReadULEB128()), + }) + } + lit := d.r.ReadULEB128() + f.Lit = lit != 0 + return nil +} diff --git a/src/internal/coverage/decodemeta/decodefile.go b/src/internal/coverage/decodemeta/decodefile.go new file mode 100644 index 0000000..6580dd5 --- /dev/null +++ b/src/internal/coverage/decodemeta/decodefile.go @@ -0,0 +1,223 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package decodemeta + +// This package contains APIs and helpers for reading and decoding +// meta-data output files emitted by the runtime when a +// coverage-instrumented binary executes. A meta-data file contains +// top-level info (counter mode, number of packages) and then a +// separate self-contained meta-data section for each Go package. + +import ( + "bufio" + "crypto/md5" + "encoding/binary" + "fmt" + "internal/coverage" + "internal/coverage/slicereader" + "internal/coverage/stringtab" + "io" + "os" +) + +// CoverageMetaFileReader provides state and methods for reading +// a meta-data file from a code coverage run. +type CoverageMetaFileReader struct { + f *os.File + hdr coverage.MetaFileHeader + tmp []byte + pkgOffsets []uint64 + pkgLengths []uint64 + strtab *stringtab.Reader + fileRdr *bufio.Reader + fileView []byte + debug bool +} + +// NewCoverageMetaFileReader returns a new helper object for reading +// the coverage meta-data output file 'f'. The param 'fileView' is a +// read-only slice containing the contents of 'f' obtained by mmap'ing +// the file read-only; 'fileView' may be nil, in which case the helper +// will read the contents of the file using regular file Read +// operations. +func NewCoverageMetaFileReader(f *os.File, fileView []byte) (*CoverageMetaFileReader, error) { + r := &CoverageMetaFileReader{ + f: f, + fileView: fileView, + tmp: make([]byte, 256), + } + + if err := r.readFileHeader(); err != nil { + return nil, err + } + return r, nil +} + +func (r *CoverageMetaFileReader) readFileHeader() error { + var err error + + r.fileRdr = bufio.NewReader(r.f) + + // Read file header. + if err := binary.Read(r.fileRdr, binary.LittleEndian, &r.hdr); err != nil { + return err + } + + // Verify magic string + m := r.hdr.Magic + g := coverage.CovMetaMagic + if m[0] != g[0] || m[1] != g[1] || m[2] != g[2] || m[3] != g[3] { + return fmt.Errorf("invalid meta-data file magic string") + } + + // Vet the version. If this is a meta-data file from the future, + // we won't be able to read it. + if r.hdr.Version > coverage.MetaFileVersion { + return fmt.Errorf("meta-data file withn unknown version %d (expected %d)", r.hdr.Version, coverage.MetaFileVersion) + } + + // Read package offsets for good measure + r.pkgOffsets = make([]uint64, r.hdr.Entries) + for i := uint64(0); i < r.hdr.Entries; i++ { + if r.pkgOffsets[i], err = r.rdUint64(); err != nil { + return err + } + if r.pkgOffsets[i] > r.hdr.TotalLength { + return fmt.Errorf("insane pkg offset %d: %d > totlen %d", + i, r.pkgOffsets[i], r.hdr.TotalLength) + } + } + r.pkgLengths = make([]uint64, r.hdr.Entries) + for i := uint64(0); i < r.hdr.Entries; i++ { + if r.pkgLengths[i], err = r.rdUint64(); err != nil { + return err + } + if r.pkgLengths[i] > r.hdr.TotalLength { + return fmt.Errorf("insane pkg length %d: %d > totlen %d", + i, r.pkgLengths[i], r.hdr.TotalLength) + } + } + + // Read string table. + b := make([]byte, r.hdr.StrTabLength) + nr, err := r.fileRdr.Read(b) + if err != nil { + return err + } + if nr != int(r.hdr.StrTabLength) { + return fmt.Errorf("error: short read on string table") + } + slr := slicereader.NewReader(b, false /* not readonly */) + r.strtab = stringtab.NewReader(slr) + r.strtab.Read() + + if r.debug { + fmt.Fprintf(os.Stderr, "=-= read-in header is: %+v\n", *r) + } + + return nil +} + +func (r *CoverageMetaFileReader) rdUint64() (uint64, error) { + r.tmp = r.tmp[:0] + r.tmp = append(r.tmp, make([]byte, 8)...) + n, err := r.fileRdr.Read(r.tmp) + if err != nil { + return 0, err + } + if n != 8 { + return 0, fmt.Errorf("premature end of file on read") + } + v := binary.LittleEndian.Uint64(r.tmp) + return v, nil +} + +// NumPackages returns the number of packages for which this file +// contains meta-data. +func (r *CoverageMetaFileReader) NumPackages() uint64 { + return r.hdr.Entries +} + +// CounterMode returns the counter mode (set, count, atomic) used +// when building for coverage for the program that produce this +// meta-data file. +func (r *CoverageMetaFileReader) CounterMode() coverage.CounterMode { + return r.hdr.CMode +} + +// CounterMode returns the counter granularity (single counter per +// function, or counter per block) selected when building for coverage +// for the program that produce this meta-data file. +func (r *CoverageMetaFileReader) CounterGranularity() coverage.CounterGranularity { + return r.hdr.CGranularity +} + +// FileHash returns the hash computed for all of the package meta-data +// blobs. Coverage counter data files refer to this hash, and the +// hash will be encoded into the meta-data file name. +func (r *CoverageMetaFileReader) FileHash() [16]byte { + return r.hdr.MetaFileHash +} + +// GetPackageDecoder requests a decoder object for the package within +// the meta-data file whose index is 'pkIdx'. If the +// CoverageMetaFileReader was set up with a read-only file view, a +// pointer into that file view will be returned, otherwise the buffer +// 'payloadbuf' will be written to (or if it is not of sufficient +// size, a new buffer will be allocated). Return value is the decoder, +// a byte slice with the encoded meta-data, and an error. +func (r *CoverageMetaFileReader) GetPackageDecoder(pkIdx uint32, payloadbuf []byte) (*CoverageMetaDataDecoder, []byte, error) { + pp, err := r.GetPackagePayload(pkIdx, payloadbuf) + if r.debug { + fmt.Fprintf(os.Stderr, "=-= pkidx=%d payload length is %d hash=%s\n", + pkIdx, len(pp), fmt.Sprintf("%x", md5.Sum(pp))) + } + if err != nil { + return nil, nil, err + } + mdd, err := NewCoverageMetaDataDecoder(pp, r.fileView != nil) + if err != nil { + return nil, nil, err + } + return mdd, pp, nil +} + +// GetPackagePayload returns the raw (encoded) meta-data payload for the +// package with index 'pkIdx'. As with GetPackageDecoder, if the +// CoverageMetaFileReader was set up with a read-only file view, a +// pointer into that file view will be returned, otherwise the buffer +// 'payloadbuf' will be written to (or if it is not of sufficient +// size, a new buffer will be allocated). Return value is the decoder, +// a byte slice with the encoded meta-data, and an error. +func (r *CoverageMetaFileReader) GetPackagePayload(pkIdx uint32, payloadbuf []byte) ([]byte, error) { + + // Determine correct offset/length. + if uint64(pkIdx) >= r.hdr.Entries { + return nil, fmt.Errorf("GetPackagePayload: illegal pkg index %d", pkIdx) + } + off := r.pkgOffsets[pkIdx] + len := r.pkgLengths[pkIdx] + + if r.debug { + fmt.Fprintf(os.Stderr, "=-= for pk %d, off=%d len=%d\n", pkIdx, off, len) + } + + if r.fileView != nil { + return r.fileView[off : off+len], nil + } + + payload := payloadbuf[:0] + if cap(payload) < int(len) { + payload = make([]byte, 0, len) + } + payload = append(payload, make([]byte, len)...) + if _, err := r.f.Seek(int64(off), io.SeekStart); err != nil { + return nil, err + } + if _, err := io.ReadFull(r.f, payload); err != nil { + return nil, err + } + return payload, nil +} diff --git a/src/internal/coverage/defs.go b/src/internal/coverage/defs.go new file mode 100644 index 0000000..8751b9f --- /dev/null +++ b/src/internal/coverage/defs.go @@ -0,0 +1,374 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package coverage + +// Types and constants related to the output files written +// by code coverage tooling. When a coverage-instrumented binary +// is run, it emits two output files: a meta-data output file, and +// a counter data output file. + +//..................................................................... +// +// Meta-data definitions: +// +// The meta-data file is composed of a file header, a series of +// meta-data blobs/sections (one per instrumented package), and an offsets +// area storing the offsets of each section. Format of the meta-data +// file looks like: +// +// --header---------- +// | magic: [4]byte magic string +// | version +// | total length of meta-data file in bytes +// | numPkgs: number of package entries in file +// | hash: [16]byte hash of entire meta-data payload +// | offset to string table section +// | length of string table +// | number of entries in string table +// | counter mode +// | counter granularity +// --package offsets table------ +// <offset to pkg 0> +// <offset to pkg 1> +// ... +// --package lengths table------ +// <length of pkg 0> +// <length of pkg 1> +// ... +// --string table------ +// <uleb128 len> 8 +// <data> "somestring" +// ... +// --package payloads------ +// <meta-symbol for pkg 0> +// <meta-symbol for pkg 1> +// ... +// +// Each package payload is a stand-alone blob emitted by the compiler, +// and does not depend on anything else in the meta-data file. In +// particular, each blob has it's own string table. Note that the +// file-level string table is expected to be very short (most strings +// will be in the meta-data blobs themselves). + +// CovMetaMagic holds the magic string for a meta-data file. +var CovMetaMagic = [4]byte{'\x00', '\x63', '\x76', '\x6d'} + +// MetaFilePref is a prefix used when emitting meta-data files; these +// files are of the form "covmeta.<hash>", where hash is a hash +// computed from the hashes of all the package meta-data symbols in +// the program. +const MetaFilePref = "covmeta" + +// MetaFileVersion contains the current (most recent) meta-data file version. +const MetaFileVersion = 1 + +// MetaFileHeader stores file header information for a meta-data file. +type MetaFileHeader struct { + Magic [4]byte + Version uint32 + TotalLength uint64 + Entries uint64 + MetaFileHash [16]byte + StrTabOffset uint32 + StrTabLength uint32 + CMode CounterMode + CGranularity CounterGranularity + _ [6]byte // padding +} + +// MetaSymbolHeader stores header information for a single +// meta-data blob, e.g. the coverage meta-data payload +// computed for a given Go package. +type MetaSymbolHeader struct { + Length uint32 // size of meta-symbol payload in bytes + PkgName uint32 // string table index + PkgPath uint32 // string table index + ModulePath uint32 // string table index + MetaHash [16]byte + _ byte // currently unused + _ [3]byte // padding + NumFiles uint32 + NumFuncs uint32 +} + +const CovMetaHeaderSize = 16 + 4 + 4 + 4 + 4 + 4 + 4 + 4 // keep in sync with above + +// As an example, consider the following Go package: +// +// 01: package p +// 02: +// 03: var v, w, z int +// 04: +// 05: func small(x, y int) int { +// 06: v++ +// 07: // comment +// 08: if y == 0 { +// 09: return x +// 10: } +// 11: return (x << 1) ^ (9 / y) +// 12: } +// 13: +// 14: func Medium(q, r int) int { +// 15: s1 := small(q, r) +// 16: z += s1 +// 17: s2 := small(r, q) +// 18: w -= s2 +// 19: return w + z +// 20: } +// +// The meta-data blob for the single package above might look like the +// following: +// +// -- MetaSymbolHeader header---------- +// | size: size of this blob in bytes +// | packagepath: <path to p> +// | modulepath: <modpath for p> +// | nfiles: 1 +// | nfunctions: 2 +// --func offsets table------ +// <offset to func 0> +// <offset to func 1> +// --string table (contains all files and functions)------ +// | <uleb128 len> 4 +// | <data> "p.go" +// | <uleb128 len> 5 +// | <data> "small" +// | <uleb128 len> 6 +// | <data> "Medium" +// --func 0------ +// | <uleb128> num units: 3 +// | <uleb128> func name: S1 (index into string table) +// | <uleb128> file: S0 (index into string table) +// | <unit 0>: S0 L6 L8 2 +// | <unit 1>: S0 L9 L9 1 +// | <unit 2>: S0 L11 L11 1 +// --func 1------ +// | <uleb128> num units: 1 +// | <uleb128> func name: S2 (index into string table) +// | <uleb128> file: S0 (index into string table) +// | <unit 0>: S0 L15 L19 5 +// ---end----------- + +// The following types and constants used by the meta-data encoder/decoder. + +// FuncDesc encapsulates the meta-data definitions for a single Go function. +// This version assumes that we're looking at a function before inlining; +// if we want to capture a post-inlining view of the world, the +// representations of source positions would need to be a good deal more +// complicated. +type FuncDesc struct { + Funcname string + Srcfile string + Units []CoverableUnit + Lit bool // true if this is a function literal +} + +// CoverableUnit describes the source characteristics of a single +// program unit for which we want to gather coverage info. Coverable +// units are either "simple" or "intraline"; a "simple" coverable unit +// corresponds to a basic block (region of straight-line code with no +// jumps or control transfers). An "intraline" unit corresponds to a +// logical clause nested within some other simple unit. A simple unit +// will have a zero Parent value; for an intraline unit NxStmts will +// be zero and Parent will be set to 1 plus the index of the +// containing simple statement. Example: +// +// L7: q := 1 +// L8: x := (y == 101 || launch() == false) +// L9: r := x * 2 +// +// For the code above we would have three simple units (one for each +// line), then an intraline unit describing the "launch() == false" +// clause in line 8, with Parent pointing to the index of the line 8 +// unit in the units array. +// +// Note: in the initial version of the coverage revamp, only simple +// units will be in use. +type CoverableUnit struct { + StLine, StCol uint32 + EnLine, EnCol uint32 + NxStmts uint32 + Parent uint32 +} + +// CounterMode tracks the "flavor" of the coverage counters being +// used in a given coverage-instrumented program. +type CounterMode uint8 + +const ( + CtrModeInvalid CounterMode = iota + CtrModeSet // "set" mode + CtrModeCount // "count" mode + CtrModeAtomic // "atomic" mode + CtrModeRegOnly // registration-only pseudo-mode + CtrModeTestMain // testmain pseudo-mode +) + +func (cm CounterMode) String() string { + switch cm { + case CtrModeSet: + return "set" + case CtrModeCount: + return "count" + case CtrModeAtomic: + return "atomic" + case CtrModeRegOnly: + return "regonly" + case CtrModeTestMain: + return "testmain" + } + return "<invalid>" +} + +func ParseCounterMode(mode string) CounterMode { + var cm CounterMode + switch mode { + case "set": + cm = CtrModeSet + case "count": + cm = CtrModeCount + case "atomic": + cm = CtrModeAtomic + case "regonly": + cm = CtrModeRegOnly + case "testmain": + cm = CtrModeTestMain + default: + cm = CtrModeInvalid + } + return cm +} + +// CounterGranularity tracks the granularity of the coverage counters being +// used in a given coverage-instrumented program. +type CounterGranularity uint8 + +const ( + CtrGranularityInvalid CounterGranularity = iota + CtrGranularityPerBlock + CtrGranularityPerFunc +) + +func (cm CounterGranularity) String() string { + switch cm { + case CtrGranularityPerBlock: + return "perblock" + case CtrGranularityPerFunc: + return "perfunc" + } + return "<invalid>" +} + +//..................................................................... +// +// Counter data definitions: +// + +// A counter data file is composed of a file header followed by one or +// more "segments" (each segment representing a given run or partial +// run of a give binary) followed by a footer. + +// CovCounterMagic holds the magic string for a coverage counter-data file. +var CovCounterMagic = [4]byte{'\x00', '\x63', '\x77', '\x6d'} + +// CounterFileVersion stores the most recent counter data file version. +const CounterFileVersion = 1 + +// CounterFileHeader stores files header information for a counter-data file. +type CounterFileHeader struct { + Magic [4]byte + Version uint32 + MetaHash [16]byte + CFlavor CounterFlavor + BigEndian bool + _ [6]byte // padding +} + +// CounterSegmentHeader encapsulates information about a specific +// segment in a counter data file, which at the moment contains +// counters data from a single execution of a coverage-instrumented +// program. Following the segment header will be the string table and +// args table, and then (possibly) padding bytes to bring the byte +// size of the preamble up to a multiple of 4. Immediately following +// that will be the counter payloads. +// +// The "args" section of a segment is used to store annotations +// describing where the counter data came from; this section is +// basically a series of key-value pairs (can be thought of as an +// encoded 'map[string]string'). At the moment we only write os.Args() +// data to this section, using pairs of the form "argc=<integer>", +// "argv0=<os.Args[0]>", "argv1=<os.Args[1]>", and so on. In the +// future the args table may also include things like GOOS/GOARCH +// values, and/or tags indicating which tests were run to generate the +// counter data. +type CounterSegmentHeader struct { + FcnEntries uint64 + StrTabLen uint32 + ArgsLen uint32 +} + +// CounterFileFooter appears at the tail end of a counter data file, +// and stores the number of segments it contains. +type CounterFileFooter struct { + Magic [4]byte + _ [4]byte // padding + NumSegments uint32 + _ [4]byte // padding +} + +// CounterFilePref is the file prefix used when emitting coverage data +// output files. CounterFileTemplate describes the format of the file +// name: prefix followed by meta-file hash followed by process ID +// followed by emit UnixNanoTime. +const CounterFilePref = "covcounters" +const CounterFileTempl = "%s.%x.%d.%d" +const CounterFileRegexp = `^%s\.(\S+)\.(\d+)\.(\d+)+$` + +// CounterFlavor describes how function and counters are +// stored/represented in the counter section of the file. +type CounterFlavor uint8 + +const ( + // "Raw" representation: all values (pkg ID, func ID, num counters, + // and counters themselves) are stored as uint32's. + CtrRaw CounterFlavor = iota + 1 + + // "ULeb" representation: all values (pkg ID, func ID, num counters, + // and counters themselves) are stored with ULEB128 encoding. + CtrULeb128 +) + +func Round4(x int) int { + return (x + 3) &^ 3 +} + +//..................................................................... +// +// Runtime counter data definitions. +// + +// At runtime within a coverage-instrumented program, the "counters" +// object we associated with instrumented function can be thought of +// as a struct of the following form: +// +// struct { +// numCtrs uint32 +// pkgid uint32 +// funcid uint32 +// counterArray [numBlocks]uint32 +// } +// +// where "numCtrs" is the number of blocks / coverable units within the +// function, "pkgid" is the unique index assigned to this package by +// the runtime, "funcid" is the index of this function within its containing +// package, and "counterArray" stores the actual counters. +// +// The counter variable itself is created not as a struct but as a flat +// array of uint32's; we then use the offsets below to index into it. + +const NumCtrsOffset = 0 +const PkgIdOffset = 1 +const FuncIdOffset = 2 +const FirstCtrOffset = 3 diff --git a/src/internal/coverage/encodecounter/encode.go b/src/internal/coverage/encodecounter/encode.go new file mode 100644 index 0000000..5958673 --- /dev/null +++ b/src/internal/coverage/encodecounter/encode.go @@ -0,0 +1,297 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package encodecounter + +import ( + "bufio" + "encoding/binary" + "fmt" + "internal/coverage" + "internal/coverage/slicewriter" + "internal/coverage/stringtab" + "internal/coverage/uleb128" + "io" + "os" + "sort" +) + +// This package contains APIs and helpers for encoding initial portions +// of the counter data files emitted at runtime when coverage instrumentation +// is enabled. Counter data files may contain multiple segments; the file +// header and first segment are written via the "Write" method below, and +// additional segments can then be added using "AddSegment". + +type CoverageDataWriter struct { + stab *stringtab.Writer + w *bufio.Writer + csh coverage.CounterSegmentHeader + tmp []byte + cflavor coverage.CounterFlavor + segs uint32 + debug bool +} + +func NewCoverageDataWriter(w io.Writer, flav coverage.CounterFlavor) *CoverageDataWriter { + r := &CoverageDataWriter{ + stab: &stringtab.Writer{}, + w: bufio.NewWriter(w), + + tmp: make([]byte, 64), + cflavor: flav, + } + r.stab.InitWriter() + r.stab.Lookup("") + return r +} + +// CounterVisitor describes a helper object used during counter file +// writing; when writing counter data files, clients pass a +// CounterVisitor to the write/emit routines, then the expectation is +// that the VisitFuncs method will then invoke the callback "f" with +// data for each function to emit to the file. +type CounterVisitor interface { + VisitFuncs(f CounterVisitorFn) error +} + +// CounterVisitorFn describes a callback function invoked when writing +// coverage counter data. +type CounterVisitorFn func(pkid uint32, funcid uint32, counters []uint32) error + +// Write writes the contents of the count-data file to the writer +// previously supplied to NewCoverageDataWriter. Returns an error +// if something went wrong somewhere with the write. +func (cfw *CoverageDataWriter) Write(metaFileHash [16]byte, args map[string]string, visitor CounterVisitor) error { + if err := cfw.writeHeader(metaFileHash); err != nil { + return err + } + return cfw.AppendSegment(args, visitor) +} + +func padToFourByteBoundary(ws *slicewriter.WriteSeeker) error { + sz := len(ws.BytesWritten()) + zeros := []byte{0, 0, 0, 0} + rem := uint32(sz) % 4 + if rem != 0 { + pad := zeros[:(4 - rem)] + if nw, err := ws.Write(pad); err != nil { + return err + } else if nw != len(pad) { + return fmt.Errorf("error: short write") + } + } + return nil +} + +func (cfw *CoverageDataWriter) patchSegmentHeader(ws *slicewriter.WriteSeeker) error { + // record position + off, err := ws.Seek(0, io.SeekCurrent) + if err != nil { + return fmt.Errorf("error seeking in patchSegmentHeader: %v", err) + } + // seek back to start so that we can update the segment header + if _, err := ws.Seek(0, io.SeekStart); err != nil { + return fmt.Errorf("error seeking in patchSegmentHeader: %v", err) + } + if cfw.debug { + fmt.Fprintf(os.Stderr, "=-= writing counter segment header: %+v", cfw.csh) + } + if err := binary.Write(ws, binary.LittleEndian, cfw.csh); err != nil { + return err + } + // ... and finally return to the original offset. + if _, err := ws.Seek(off, io.SeekStart); err != nil { + return fmt.Errorf("error seeking in patchSegmentHeader: %v", err) + } + return nil +} + +func (cfw *CoverageDataWriter) writeSegmentPreamble(args map[string]string, ws *slicewriter.WriteSeeker) error { + if err := binary.Write(ws, binary.LittleEndian, cfw.csh); err != nil { + return err + } + hdrsz := uint32(len(ws.BytesWritten())) + + // Write string table and args to a byte slice (since we need + // to capture offsets at various points), then emit the slice + // once we are done. + cfw.stab.Freeze() + if err := cfw.stab.Write(ws); err != nil { + return err + } + cfw.csh.StrTabLen = uint32(len(ws.BytesWritten())) - hdrsz + + akeys := make([]string, 0, len(args)) + for k := range args { + akeys = append(akeys, k) + } + sort.Strings(akeys) + + wrULEB128 := func(v uint) error { + cfw.tmp = cfw.tmp[:0] + cfw.tmp = uleb128.AppendUleb128(cfw.tmp, v) + if _, err := ws.Write(cfw.tmp); err != nil { + return err + } + return nil + } + + // Count of arg pairs. + if err := wrULEB128(uint(len(args))); err != nil { + return err + } + // Arg pairs themselves. + for _, k := range akeys { + ki := uint(cfw.stab.Lookup(k)) + if err := wrULEB128(ki); err != nil { + return err + } + v := args[k] + vi := uint(cfw.stab.Lookup(v)) + if err := wrULEB128(vi); err != nil { + return err + } + } + if err := padToFourByteBoundary(ws); err != nil { + return err + } + cfw.csh.ArgsLen = uint32(len(ws.BytesWritten())) - (cfw.csh.StrTabLen + hdrsz) + + return nil +} + +// AppendSegment appends a new segment to a counter data, with a new +// args section followed by a payload of counter data clauses. +func (cfw *CoverageDataWriter) AppendSegment(args map[string]string, visitor CounterVisitor) error { + cfw.stab = &stringtab.Writer{} + cfw.stab.InitWriter() + cfw.stab.Lookup("") + + var err error + for k, v := range args { + cfw.stab.Lookup(k) + cfw.stab.Lookup(v) + } + + ws := &slicewriter.WriteSeeker{} + if err = cfw.writeSegmentPreamble(args, ws); err != nil { + return err + } + if err = cfw.writeCounters(visitor, ws); err != nil { + return err + } + if err = cfw.patchSegmentHeader(ws); err != nil { + return err + } + if err := cfw.writeBytes(ws.BytesWritten()); err != nil { + return err + } + if err = cfw.writeFooter(); err != nil { + return err + } + if err := cfw.w.Flush(); err != nil { + return fmt.Errorf("write error: %v", err) + } + cfw.stab = nil + return nil +} + +func (cfw *CoverageDataWriter) writeHeader(metaFileHash [16]byte) error { + // Emit file header. + ch := coverage.CounterFileHeader{ + Magic: coverage.CovCounterMagic, + Version: coverage.CounterFileVersion, + MetaHash: metaFileHash, + CFlavor: cfw.cflavor, + BigEndian: false, + } + if err := binary.Write(cfw.w, binary.LittleEndian, ch); err != nil { + return err + } + return nil +} + +func (cfw *CoverageDataWriter) writeBytes(b []byte) error { + if len(b) == 0 { + return nil + } + nw, err := cfw.w.Write(b) + if err != nil { + return fmt.Errorf("error writing counter data: %v", err) + } + if len(b) != nw { + return fmt.Errorf("error writing counter data: short write") + } + return nil +} + +func (cfw *CoverageDataWriter) writeCounters(visitor CounterVisitor, ws *slicewriter.WriteSeeker) error { + // Notes: + // - this version writes everything little-endian, which means + // a call is needed to encode every value (expensive) + // - we may want to move to a model in which we just blast out + // all counters, or possibly mmap the file and do the write + // implicitly. + ctrb := make([]byte, 4) + wrval := func(val uint32) error { + var buf []byte + var towr int + if cfw.cflavor == coverage.CtrRaw { + binary.LittleEndian.PutUint32(ctrb, val) + buf = ctrb + towr = 4 + } else if cfw.cflavor == coverage.CtrULeb128 { + cfw.tmp = cfw.tmp[:0] + cfw.tmp = uleb128.AppendUleb128(cfw.tmp, uint(val)) + buf = cfw.tmp + towr = len(buf) + } else { + panic("internal error: bad counter flavor") + } + if sz, err := ws.Write(buf); err != nil { + return err + } else if sz != towr { + return fmt.Errorf("writing counters: short write") + } + return nil + } + + // Write out entries for each live function. + emitter := func(pkid uint32, funcid uint32, counters []uint32) error { + cfw.csh.FcnEntries++ + if err := wrval(uint32(len(counters))); err != nil { + return err + } + + if err := wrval(pkid); err != nil { + return err + } + + if err := wrval(funcid); err != nil { + return err + } + for _, val := range counters { + if err := wrval(val); err != nil { + return err + } + } + return nil + } + if err := visitor.VisitFuncs(emitter); err != nil { + return err + } + return nil +} + +func (cfw *CoverageDataWriter) writeFooter() error { + cfw.segs++ + cf := coverage.CounterFileFooter{ + Magic: coverage.CovCounterMagic, + NumSegments: cfw.segs, + } + if err := binary.Write(cfw.w, binary.LittleEndian, cf); err != nil { + return err + } + return nil +} diff --git a/src/internal/coverage/encodemeta/encode.go b/src/internal/coverage/encodemeta/encode.go new file mode 100644 index 0000000..d211c7c --- /dev/null +++ b/src/internal/coverage/encodemeta/encode.go @@ -0,0 +1,215 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package encodemeta + +// This package contains APIs and helpers for encoding the meta-data +// "blob" for a single Go package, created when coverage +// instrumentation is turned on. + +import ( + "bytes" + "crypto/md5" + "encoding/binary" + "fmt" + "hash" + "internal/coverage" + "internal/coverage/stringtab" + "internal/coverage/uleb128" + "io" + "os" +) + +type CoverageMetaDataBuilder struct { + stab stringtab.Writer + funcs []funcDesc + tmp []byte // temp work slice + h hash.Hash + pkgpath uint32 + pkgname uint32 + modpath uint32 + debug bool + werr error +} + +func NewCoverageMetaDataBuilder(pkgpath string, pkgname string, modulepath string) (*CoverageMetaDataBuilder, error) { + if pkgpath == "" { + return nil, fmt.Errorf("invalid empty package path") + } + x := &CoverageMetaDataBuilder{ + tmp: make([]byte, 0, 256), + h: md5.New(), + } + x.stab.InitWriter() + x.stab.Lookup("") + x.pkgpath = x.stab.Lookup(pkgpath) + x.pkgname = x.stab.Lookup(pkgname) + x.modpath = x.stab.Lookup(modulepath) + io.WriteString(x.h, pkgpath) + io.WriteString(x.h, pkgname) + io.WriteString(x.h, modulepath) + return x, nil +} + +func h32(x uint32, h hash.Hash, tmp []byte) { + tmp = tmp[:0] + tmp = append(tmp, []byte{0, 0, 0, 0}...) + binary.LittleEndian.PutUint32(tmp, x) + h.Write(tmp) +} + +type funcDesc struct { + encoded []byte +} + +// AddFunc registers a new function with the meta data builder. +func (b *CoverageMetaDataBuilder) AddFunc(f coverage.FuncDesc) uint { + hashFuncDesc(b.h, &f, b.tmp) + fd := funcDesc{} + b.tmp = b.tmp[:0] + b.tmp = uleb128.AppendUleb128(b.tmp, uint(len(f.Units))) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(b.stab.Lookup(f.Funcname))) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(b.stab.Lookup(f.Srcfile))) + for _, u := range f.Units { + b.tmp = uleb128.AppendUleb128(b.tmp, uint(u.StLine)) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(u.StCol)) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(u.EnLine)) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(u.EnCol)) + b.tmp = uleb128.AppendUleb128(b.tmp, uint(u.NxStmts)) + } + lit := uint(0) + if f.Lit { + lit = 1 + } + b.tmp = uleb128.AppendUleb128(b.tmp, lit) + fd.encoded = bytes.Clone(b.tmp) + rv := uint(len(b.funcs)) + b.funcs = append(b.funcs, fd) + return rv +} + +func (b *CoverageMetaDataBuilder) emitFuncOffsets(w io.WriteSeeker, off int64) int64 { + nFuncs := len(b.funcs) + var foff int64 = coverage.CovMetaHeaderSize + int64(b.stab.Size()) + int64(nFuncs)*4 + for idx := 0; idx < nFuncs; idx++ { + b.wrUint32(w, uint32(foff)) + foff += int64(len(b.funcs[idx].encoded)) + } + return off + (int64(len(b.funcs)) * 4) +} + +func (b *CoverageMetaDataBuilder) emitFunc(w io.WriteSeeker, off int64, f funcDesc) (int64, error) { + ew := len(f.encoded) + if nw, err := w.Write(f.encoded); err != nil { + return 0, err + } else if ew != nw { + return 0, fmt.Errorf("short write emitting coverage meta-data") + } + return off + int64(ew), nil +} + +func (b *CoverageMetaDataBuilder) reportWriteError(err error) { + if b.werr != nil { + b.werr = err + } +} + +func (b *CoverageMetaDataBuilder) wrUint32(w io.WriteSeeker, v uint32) { + b.tmp = b.tmp[:0] + b.tmp = append(b.tmp, []byte{0, 0, 0, 0}...) + binary.LittleEndian.PutUint32(b.tmp, v) + if nw, err := w.Write(b.tmp); err != nil { + b.reportWriteError(err) + } else if nw != 4 { + b.reportWriteError(fmt.Errorf("short write")) + } +} + +// Emit writes the meta-data accumulated so far in this builder to 'w'. +// Returns a hash of the meta-data payload and an error. +func (b *CoverageMetaDataBuilder) Emit(w io.WriteSeeker) ([16]byte, error) { + // Emit header. Length will initially be zero, we'll + // back-patch it later. + var digest [16]byte + copy(digest[:], b.h.Sum(nil)) + mh := coverage.MetaSymbolHeader{ + // hash and length initially zero, will be back-patched + PkgPath: uint32(b.pkgpath), + PkgName: uint32(b.pkgname), + ModulePath: uint32(b.modpath), + NumFiles: uint32(b.stab.Nentries()), + NumFuncs: uint32(len(b.funcs)), + MetaHash: digest, + } + if b.debug { + fmt.Fprintf(os.Stderr, "=-= writing header: %+v\n", mh) + } + if err := binary.Write(w, binary.LittleEndian, mh); err != nil { + return digest, fmt.Errorf("error writing meta-file header: %v", err) + } + off := int64(coverage.CovMetaHeaderSize) + + // Write function offsets section + off = b.emitFuncOffsets(w, off) + + // Check for any errors up to this point. + if b.werr != nil { + return digest, b.werr + } + + // Write string table. + if err := b.stab.Write(w); err != nil { + return digest, err + } + off += int64(b.stab.Size()) + + // Write functions + for _, f := range b.funcs { + var err error + off, err = b.emitFunc(w, off, f) + if err != nil { + return digest, err + } + } + + // Back-patch the length. + totalLength := uint32(off) + if _, err := w.Seek(0, io.SeekStart); err != nil { + return digest, err + } + b.wrUint32(w, totalLength) + if b.werr != nil { + return digest, b.werr + } + return digest, nil +} + +// HashFuncDesc computes an md5 sum of a coverage.FuncDesc and returns +// a digest for it. +func HashFuncDesc(f *coverage.FuncDesc) [16]byte { + h := md5.New() + tmp := make([]byte, 0, 32) + hashFuncDesc(h, f, tmp) + var r [16]byte + copy(r[:], h.Sum(nil)) + return r +} + +// hashFuncDesc incorporates a given function 'f' into the hash 'h'. +func hashFuncDesc(h hash.Hash, f *coverage.FuncDesc, tmp []byte) { + io.WriteString(h, f.Funcname) + io.WriteString(h, f.Srcfile) + for _, u := range f.Units { + h32(u.StLine, h, tmp) + h32(u.StCol, h, tmp) + h32(u.EnLine, h, tmp) + h32(u.EnCol, h, tmp) + h32(u.NxStmts, h, tmp) + } + lit := uint32(0) + if f.Lit { + lit = 1 + } + h32(lit, h, tmp) +} diff --git a/src/internal/coverage/encodemeta/encodefile.go b/src/internal/coverage/encodemeta/encodefile.go new file mode 100644 index 0000000..38ae46e --- /dev/null +++ b/src/internal/coverage/encodemeta/encodefile.go @@ -0,0 +1,132 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package encodemeta + +import ( + "bufio" + "crypto/md5" + "encoding/binary" + "fmt" + "internal/coverage" + "internal/coverage/stringtab" + "io" + "os" + "unsafe" +) + +// This package contains APIs and helpers for writing out a meta-data +// file (composed of a file header, offsets/lengths, and then a series of +// meta-data blobs emitted by the compiler, one per Go package). + +type CoverageMetaFileWriter struct { + stab stringtab.Writer + mfname string + w *bufio.Writer + tmp []byte + debug bool +} + +func NewCoverageMetaFileWriter(mfname string, w io.Writer) *CoverageMetaFileWriter { + r := &CoverageMetaFileWriter{ + mfname: mfname, + w: bufio.NewWriter(w), + tmp: make([]byte, 64), + } + r.stab.InitWriter() + r.stab.Lookup("") + return r +} + +func (m *CoverageMetaFileWriter) Write(finalHash [16]byte, blobs [][]byte, mode coverage.CounterMode, granularity coverage.CounterGranularity) error { + mhsz := uint64(unsafe.Sizeof(coverage.MetaFileHeader{})) + stSize := m.stab.Size() + stOffset := mhsz + uint64(16*len(blobs)) + preambleLength := stOffset + uint64(stSize) + + if m.debug { + fmt.Fprintf(os.Stderr, "=+= sizeof(MetaFileHeader)=%d\n", mhsz) + fmt.Fprintf(os.Stderr, "=+= preambleLength=%d stSize=%d\n", preambleLength, stSize) + } + + // Compute total size + tlen := preambleLength + for i := 0; i < len(blobs); i++ { + tlen += uint64(len(blobs[i])) + } + + // Emit header + mh := coverage.MetaFileHeader{ + Magic: coverage.CovMetaMagic, + Version: coverage.MetaFileVersion, + TotalLength: tlen, + Entries: uint64(len(blobs)), + MetaFileHash: finalHash, + StrTabOffset: uint32(stOffset), + StrTabLength: stSize, + CMode: mode, + CGranularity: granularity, + } + var err error + if err = binary.Write(m.w, binary.LittleEndian, mh); err != nil { + return fmt.Errorf("error writing %s: %v", m.mfname, err) + } + + if m.debug { + fmt.Fprintf(os.Stderr, "=+= len(blobs) is %d\n", mh.Entries) + } + + // Emit package offsets section followed by package lengths section. + off := preambleLength + off2 := mhsz + buf := make([]byte, 8) + for _, blob := range blobs { + binary.LittleEndian.PutUint64(buf, off) + if _, err = m.w.Write(buf); err != nil { + return fmt.Errorf("error writing %s: %v", m.mfname, err) + } + if m.debug { + fmt.Fprintf(os.Stderr, "=+= pkg offset %d 0x%x\n", off, off) + } + off += uint64(len(blob)) + off2 += 8 + } + for _, blob := range blobs { + bl := uint64(len(blob)) + binary.LittleEndian.PutUint64(buf, bl) + if _, err = m.w.Write(buf); err != nil { + return fmt.Errorf("error writing %s: %v", m.mfname, err) + } + if m.debug { + fmt.Fprintf(os.Stderr, "=+= pkg len %d 0x%x\n", bl, bl) + } + off2 += 8 + } + + // Emit string table + if err = m.stab.Write(m.w); err != nil { + return err + } + + // Now emit blobs themselves. + for k, blob := range blobs { + if m.debug { + fmt.Fprintf(os.Stderr, "=+= writing blob %d len %d at off=%d hash %s\n", k, len(blob), off2, fmt.Sprintf("%x", md5.Sum(blob))) + } + if _, err = m.w.Write(blob); err != nil { + return fmt.Errorf("error writing %s: %v", m.mfname, err) + } + if m.debug { + fmt.Fprintf(os.Stderr, "=+= wrote package payload of %d bytes\n", + len(blob)) + } + off2 += uint64(len(blob)) + } + + // Flush writer, and we're done. + if err = m.w.Flush(); err != nil { + return fmt.Errorf("error writing %s: %v", m.mfname, err) + } + return nil +} diff --git a/src/internal/coverage/pkid.go b/src/internal/coverage/pkid.go new file mode 100644 index 0000000..8ddd44d --- /dev/null +++ b/src/internal/coverage/pkid.go @@ -0,0 +1,80 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package coverage + +// Building the runtime package with coverage instrumentation enabled +// is tricky. For all other packages, you can be guaranteed that +// the package init function is run before any functions are executed, +// but this invariant is not maintained for packages such as "runtime", +// "internal/cpu", etc. To handle this, hard-code the package ID for +// the set of packages whose functions may be running before the +// init function of the package is complete. +// +// Hardcoding is unfortunate because it means that the tool that does +// coverage instrumentation has to keep a list of runtime packages, +// meaning that if someone makes changes to the pkg "runtime" +// dependencies, unexpected behavior will result for coverage builds. +// The coverage runtime will detect and report the unexpected +// behavior; look for an error of this form: +// +// internal error in coverage meta-data tracking: +// list of hard-coded runtime package IDs needs revising. +// registered list: +// slot: 0 path='internal/cpu' hard-coded id: 1 +// slot: 1 path='internal/goarch' hard-coded id: 2 +// slot: 2 path='runtime/internal/atomic' hard-coded id: 3 +// slot: 3 path='internal/goos' +// slot: 4 path='runtime/internal/sys' hard-coded id: 5 +// slot: 5 path='internal/abi' hard-coded id: 4 +// slot: 6 path='runtime/internal/math' hard-coded id: 6 +// slot: 7 path='internal/bytealg' hard-coded id: 7 +// slot: 8 path='internal/goexperiment' +// slot: 9 path='runtime/internal/syscall' hard-coded id: 8 +// slot: 10 path='runtime' hard-coded id: 9 +// fatal error: runtime.addCovMeta +// +// For the error above, the hard-coded list is missing "internal/goos" +// and "internal/goexperiment" ; the developer in question will need +// to copy the list above into "rtPkgs" below. +// +// Note: this strategy assumes that the list of dependencies of +// package runtime is fixed, and doesn't vary depending on OS/arch. If +// this were to be the case, we would need a table of some sort below +// as opposed to a fixed list. + +var rtPkgs = [...]string{ + "internal/cpu", + "internal/goarch", + "runtime/internal/atomic", + "internal/goos", + "runtime/internal/sys", + "internal/abi", + "runtime/internal/math", + "internal/bytealg", + "internal/goexperiment", + "runtime/internal/syscall", + "runtime", +} + +// Scoping note: the constants and apis in this file are internal +// only, not expected to ever be exposed outside of the runtime (unlike +// other coverage file formats and APIs, which will likely be shared +// at some point). + +// NotHardCoded is a package pseudo-ID indicating that a given package +// is not part of the runtime and doesn't require a hard-coded ID. +const NotHardCoded = -1 + +// HardCodedPkgID returns the hard-coded ID for the specified package +// path, or -1 if we don't use a hard-coded ID. Hard-coded IDs start +// at -2 and decrease as we go down the list. +func HardCodedPkgID(pkgpath string) int { + for k, p := range rtPkgs { + if p == pkgpath { + return (0 - k) - 2 + } + } + return NotHardCoded +} diff --git a/src/internal/coverage/pods/pods.go b/src/internal/coverage/pods/pods.go new file mode 100644 index 0000000..e08f82e --- /dev/null +++ b/src/internal/coverage/pods/pods.go @@ -0,0 +1,197 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pods + +import ( + "fmt" + "internal/coverage" + "os" + "path/filepath" + "regexp" + "sort" + "strconv" +) + +// Pod encapsulates a set of files emitted during the executions of a +// coverage-instrumented binary. Each pod contains a single meta-data +// file, and then 0 or more counter data files that refer to that +// meta-data file. Pods are intended to simplify processing of +// coverage output files in the case where we have several coverage +// output directories containing output files derived from more +// than one instrumented executable. In the case where the files that +// make up a pod are spread out across multiple directories, each +// element of the "Origins" field below will be populated with the +// index of the originating directory for the corresponding counter +// data file (within the slice of input dirs handed to CollectPods). +// The ProcessIDs field will be populated with the process ID of each +// data file in the CounterDataFiles slice. +type Pod struct { + MetaFile string + CounterDataFiles []string + Origins []int + ProcessIDs []int +} + +// CollectPods visits the files contained within the directories in +// the list 'dirs', collects any coverage-related files, partitions +// them into pods, and returns a list of the pods to the caller, along +// with an error if something went wrong during directory/file +// reading. +// +// CollectPods skips over any file that is not related to coverage +// (e.g. avoids looking at things that are not meta-data files or +// counter-data files). CollectPods also skips over 'orphaned' counter +// data files (e.g. counter data files for which we can't find the +// corresponding meta-data file). If "warn" is true, CollectPods will +// issue warnings to stderr when it encounters non-fatal problems (for +// orphans or a directory with no meta-data files). +func CollectPods(dirs []string, warn bool) ([]Pod, error) { + files := []string{} + dirIndices := []int{} + for k, dir := range dirs { + dents, err := os.ReadDir(dir) + if err != nil { + return nil, err + } + for _, e := range dents { + if e.IsDir() { + continue + } + files = append(files, filepath.Join(dir, e.Name())) + dirIndices = append(dirIndices, k) + } + } + return collectPodsImpl(files, dirIndices, warn), nil +} + +// CollectPodsFromFiles functions the same as "CollectPods" but +// operates on an explicit list of files instead of a directory. +func CollectPodsFromFiles(files []string, warn bool) []Pod { + return collectPodsImpl(files, nil, warn) +} + +type fileWithAnnotations struct { + file string + origin int + pid int +} + +type protoPod struct { + mf string + elements []fileWithAnnotations +} + +// collectPodsImpl examines the specified list of files and picks out +// subsets that correspond to coverage pods. The first stage in this +// process is collecting a set { M1, M2, ... MN } where each M_k is a +// distinct coverage meta-data file. We then create a single pod for +// each meta-data file M_k, then find all of the counter data files +// that refer to that meta-data file (recall that the counter data +// file name incorporates the meta-data hash), and add the counter +// data file to the appropriate pod. +// +// This process is complicated by the fact that we need to keep track +// of directory indices for counter data files. Here is an example to +// motivate: +// +// directory 1: +// +// M1 covmeta.9bbf1777f47b3fcacb05c38b035512d6 +// C1 covcounters.9bbf1777f47b3fcacb05c38b035512d6.1677673.1662138360208416486 +// C2 covcounters.9bbf1777f47b3fcacb05c38b035512d6.1677637.1662138359974441782 +// +// directory 2: +// +// M2 covmeta.9bbf1777f47b3fcacb05c38b035512d6 +// C3 covcounters.9bbf1777f47b3fcacb05c38b035512d6.1677445.1662138360208416480 +// C4 covcounters.9bbf1777f47b3fcacb05c38b035512d6.1677677.1662138359974441781 +// M3 covmeta.a723844208cea2ae80c63482c78b2245 +// C5 covcounters.a723844208cea2ae80c63482c78b2245.3677445.1662138360208416480 +// C6 covcounters.a723844208cea2ae80c63482c78b2245.1877677.1662138359974441781 +// +// In these two directories we have three meta-data files, but only +// two are distinct, meaning that we'll wind up with two pods. The +// first pod (with meta-file M1) will have four counter data files +// (C1, C2, C3, C4) and the second pod will have two counter data files +// (C5, C6). +func collectPodsImpl(files []string, dirIndices []int, warn bool) []Pod { + metaRE := regexp.MustCompile(fmt.Sprintf(`^%s\.(\S+)$`, coverage.MetaFilePref)) + mm := make(map[string]protoPod) + for _, f := range files { + base := filepath.Base(f) + if m := metaRE.FindStringSubmatch(base); m != nil { + tag := m[1] + // We need to allow for the possibility of duplicate + // meta-data files. If we hit this case, use the + // first encountered as the canonical version. + if _, ok := mm[tag]; !ok { + mm[tag] = protoPod{mf: f} + } + // FIXME: should probably check file length and hash here for + // the duplicate. + } + } + counterRE := regexp.MustCompile(fmt.Sprintf(coverage.CounterFileRegexp, coverage.CounterFilePref)) + for k, f := range files { + base := filepath.Base(f) + if m := counterRE.FindStringSubmatch(base); m != nil { + tag := m[1] // meta hash + pid, err := strconv.Atoi(m[2]) + if err != nil { + continue + } + if v, ok := mm[tag]; ok { + idx := -1 + if dirIndices != nil { + idx = dirIndices[k] + } + fo := fileWithAnnotations{file: f, origin: idx, pid: pid} + v.elements = append(v.elements, fo) + mm[tag] = v + } else { + if warn { + warning("skipping orphaned counter file: %s", f) + } + } + } + } + if len(mm) == 0 { + if warn { + warning("no coverage data files found") + } + return nil + } + pods := make([]Pod, 0, len(mm)) + for _, p := range mm { + sort.Slice(p.elements, func(i, j int) bool { + if p.elements[i].origin != p.elements[j].origin { + return p.elements[i].origin < p.elements[j].origin + } + return p.elements[i].file < p.elements[j].file + }) + pod := Pod{ + MetaFile: p.mf, + CounterDataFiles: make([]string, 0, len(p.elements)), + Origins: make([]int, 0, len(p.elements)), + ProcessIDs: make([]int, 0, len(p.elements)), + } + for _, e := range p.elements { + pod.CounterDataFiles = append(pod.CounterDataFiles, e.file) + pod.Origins = append(pod.Origins, e.origin) + pod.ProcessIDs = append(pod.ProcessIDs, e.pid) + } + pods = append(pods, pod) + } + sort.Slice(pods, func(i, j int) bool { + return pods[i].MetaFile < pods[j].MetaFile + }) + return pods +} + +func warning(s string, a ...interface{}) { + fmt.Fprintf(os.Stderr, "warning: ") + fmt.Fprintf(os.Stderr, s, a...) + fmt.Fprintf(os.Stderr, "\n") +} diff --git a/src/internal/coverage/pods/pods_test.go b/src/internal/coverage/pods/pods_test.go new file mode 100644 index 0000000..69c16e0 --- /dev/null +++ b/src/internal/coverage/pods/pods_test.go @@ -0,0 +1,142 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pods_test + +import ( + "crypto/md5" + "fmt" + "internal/coverage" + "internal/coverage/pods" + "os" + "path/filepath" + "runtime" + "testing" +) + +func TestPodCollection(t *testing.T) { + //testenv.MustHaveGoBuild(t) + + mkdir := func(d string, perm os.FileMode) string { + dp := filepath.Join(t.TempDir(), d) + if err := os.Mkdir(dp, perm); err != nil { + t.Fatal(err) + } + return dp + } + + mkfile := func(d string, fn string) string { + fp := filepath.Join(d, fn) + if err := os.WriteFile(fp, []byte("foo"), 0666); err != nil { + t.Fatal(err) + } + return fp + } + + mkmeta := func(dir string, tag string) string { + hash := md5.Sum([]byte(tag)) + fn := fmt.Sprintf("%s.%x", coverage.MetaFilePref, hash) + return mkfile(dir, fn) + } + + mkcounter := func(dir string, tag string, nt int, pid int) string { + hash := md5.Sum([]byte(tag)) + fn := fmt.Sprintf(coverage.CounterFileTempl, coverage.CounterFilePref, hash, pid, nt) + return mkfile(dir, fn) + } + + trim := func(path string) string { + b := filepath.Base(path) + d := filepath.Dir(path) + db := filepath.Base(d) + return db + "/" + b + } + + podToString := func(p pods.Pod) string { + rv := trim(p.MetaFile) + " [\n" + for k, df := range p.CounterDataFiles { + rv += trim(df) + if p.Origins != nil { + rv += fmt.Sprintf(" o:%d", p.Origins[k]) + } + rv += "\n" + } + return rv + "]" + } + + // Create a couple of directories. + o1 := mkdir("o1", 0777) + o2 := mkdir("o2", 0777) + + // Add some random files (not coverage related) + mkfile(o1, "blah.txt") + mkfile(o1, "something.exe") + + // Add a meta-data file with two counter files to first dir. + mkmeta(o1, "m1") + mkcounter(o1, "m1", 1, 42) + mkcounter(o1, "m1", 2, 41) + mkcounter(o1, "m1", 2, 40) + + // Add a counter file with no associated meta file. + mkcounter(o1, "orphan", 9, 39) + + // Add a meta-data file with three counter files to second dir. + mkmeta(o2, "m2") + mkcounter(o2, "m2", 1, 38) + mkcounter(o2, "m2", 2, 37) + mkcounter(o2, "m2", 3, 36) + + // Add a duplicate of the first meta-file and a corresponding + // counter file to the second dir. This is intended to capture + // the scenario where we have two different runs of the same + // coverage-instrumented binary, but with the output files + // sent to separate directories. + mkmeta(o2, "m1") + mkcounter(o2, "m1", 11, 35) + + // Collect pods. + podlist, err := pods.CollectPods([]string{o1, o2}, true) + if err != nil { + t.Fatal(err) + } + + // Verify pods + if len(podlist) != 2 { + t.Fatalf("expected 2 pods got %d pods", len(podlist)) + } + + for k, p := range podlist { + t.Logf("%d: mf=%s\n", k, p.MetaFile) + } + + expected := []string{ + `o1/covmeta.ae7be26cdaa742ca148068d5ac90eaca [ +o1/covcounters.ae7be26cdaa742ca148068d5ac90eaca.40.2 o:0 +o1/covcounters.ae7be26cdaa742ca148068d5ac90eaca.41.2 o:0 +o1/covcounters.ae7be26cdaa742ca148068d5ac90eaca.42.1 o:0 +o2/covcounters.ae7be26cdaa742ca148068d5ac90eaca.35.11 o:1 +]`, + `o2/covmeta.aaf2f89992379705dac844c0a2a1d45f [ +o2/covcounters.aaf2f89992379705dac844c0a2a1d45f.36.3 o:1 +o2/covcounters.aaf2f89992379705dac844c0a2a1d45f.37.2 o:1 +o2/covcounters.aaf2f89992379705dac844c0a2a1d45f.38.1 o:1 +]`, + } + for k, exp := range expected { + got := podToString(podlist[k]) + if exp != got { + t.Errorf("pod %d: expected:\n%s\ngot:\n%s", k, exp, got) + } + } + + // Check handling of bad/unreadable dir. + if runtime.GOOS == "linux" { + dbad := "/dev/null" + _, err = pods.CollectPods([]string{dbad}, true) + if err == nil { + t.Errorf("executed error due to unreadable dir") + } + } +} diff --git a/src/internal/coverage/rtcov/rtcov.go b/src/internal/coverage/rtcov/rtcov.go new file mode 100644 index 0000000..bbb93ac --- /dev/null +++ b/src/internal/coverage/rtcov/rtcov.go @@ -0,0 +1,34 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package rtcov + +// This package contains types whose structure is shared between +// the runtime package and the "runtime/coverage" package. + +// CovMetaBlob is a container for holding the meta-data symbol (an +// RODATA variable) for an instrumented Go package. Here "p" points to +// the symbol itself, "len" is the length of the sym in bytes, and +// "hash" is an md5sum for the sym computed by the compiler. When +// the init function for a coverage-instrumented package executes, it +// will make a call into the runtime which will create a covMetaBlob +// object for the package and chain it onto a global list. +type CovMetaBlob struct { + P *byte + Len uint32 + Hash [16]byte + PkgPath string + PkgID int + CounterMode uint8 // coverage.CounterMode + CounterGranularity uint8 // coverage.CounterGranularity +} + +// CovCounterBlob is a container for encapsulating a counter section +// (BSS variable) for an instrumented Go module. Here "counters" +// points to the counter payload and "len" is the number of uint32 +// entries in the section. +type CovCounterBlob struct { + Counters *uint32 + Len uint64 +} diff --git a/src/internal/coverage/slicereader/slicereader.go b/src/internal/coverage/slicereader/slicereader.go new file mode 100644 index 0000000..d9f2a7e --- /dev/null +++ b/src/internal/coverage/slicereader/slicereader.go @@ -0,0 +1,123 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package slicereader + +import ( + "encoding/binary" + "fmt" + "io" + "unsafe" +) + +// This file contains the helper "SliceReader", a utility for +// reading values from a byte slice that may or may not be backed +// by a read-only mmap'd region. + +type Reader struct { + b []byte + readonly bool + off int64 +} + +func NewReader(b []byte, readonly bool) *Reader { + r := Reader{ + b: b, + readonly: readonly, + } + return &r +} + +func (r *Reader) Read(b []byte) (int, error) { + amt := len(b) + toread := r.b[r.off:] + if len(toread) < amt { + amt = len(toread) + } + copy(b, toread) + r.off += int64(amt) + return amt, nil +} + +func (r *Reader) Seek(offset int64, whence int) (ret int64, err error) { + switch whence { + case io.SeekStart: + if offset < 0 || offset > int64(len(r.b)) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", offset, len(r.b)) + } + r.off = offset + return offset, nil + case io.SeekCurrent: + newoff := r.off + offset + if newoff < 0 || newoff > int64(len(r.b)) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", newoff, len(r.b)) + } + r.off = newoff + return r.off, nil + case io.SeekEnd: + newoff := int64(len(r.b)) + offset + if newoff < 0 || newoff > int64(len(r.b)) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", newoff, len(r.b)) + } + r.off = newoff + return r.off, nil + } + // other modes are not supported + return 0, fmt.Errorf("unsupported seek mode %d", whence) +} + +func (r *Reader) Offset() int64 { + return r.off +} + +func (r *Reader) ReadUint8() uint8 { + rv := uint8(r.b[int(r.off)]) + r.off += 1 + return rv +} + +func (r *Reader) ReadUint32() uint32 { + end := int(r.off) + 4 + rv := binary.LittleEndian.Uint32(r.b[int(r.off):end:end]) + r.off += 4 + return rv +} + +func (r *Reader) ReadUint64() uint64 { + end := int(r.off) + 8 + rv := binary.LittleEndian.Uint64(r.b[int(r.off):end:end]) + r.off += 8 + return rv +} + +func (r *Reader) ReadULEB128() (value uint64) { + var shift uint + + for { + b := r.b[r.off] + r.off++ + value |= (uint64(b&0x7F) << shift) + if b&0x80 == 0 { + break + } + shift += 7 + } + return +} + +func (r *Reader) ReadString(len int64) string { + b := r.b[r.off : r.off+len] + r.off += len + if r.readonly { + return toString(b) // backed by RO memory, ok to make unsafe string + } + return string(b) +} + +func toString(b []byte) string { + if len(b) == 0 { + return "" + } + return unsafe.String(&b[0], len(b)) +} diff --git a/src/internal/coverage/slicereader/slr_test.go b/src/internal/coverage/slicereader/slr_test.go new file mode 100644 index 0000000..461436d --- /dev/null +++ b/src/internal/coverage/slicereader/slr_test.go @@ -0,0 +1,95 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package slicereader + +import ( + "encoding/binary" + "io" + "testing" +) + +func TestSliceReader(t *testing.T) { + b := []byte{} + + bt := make([]byte, 4) + e32 := uint32(1030507) + binary.LittleEndian.PutUint32(bt, e32) + b = append(b, bt...) + + bt = make([]byte, 8) + e64 := uint64(907050301) + binary.LittleEndian.PutUint64(bt, e64) + b = append(b, bt...) + + b = appendUleb128(b, uint(e32)) + b = appendUleb128(b, uint(e64)) + b = appendUleb128(b, 6) + s1 := "foobar" + s1b := []byte(s1) + b = append(b, s1b...) + b = appendUleb128(b, 9) + s2 := "bazbasher" + s2b := []byte(s2) + b = append(b, s2b...) + + readStr := func(slr *Reader) string { + len := slr.ReadULEB128() + return slr.ReadString(int64(len)) + } + + for i := 0; i < 2; i++ { + slr := NewReader(b, i == 0) + g32 := slr.ReadUint32() + if g32 != e32 { + t.Fatalf("slr.ReadUint32() got %d want %d", g32, e32) + } + g64 := slr.ReadUint64() + if g64 != e64 { + t.Fatalf("slr.ReadUint64() got %d want %d", g64, e64) + } + g32 = uint32(slr.ReadULEB128()) + if g32 != e32 { + t.Fatalf("slr.ReadULEB128() got %d want %d", g32, e32) + } + g64 = slr.ReadULEB128() + if g64 != e64 { + t.Fatalf("slr.ReadULEB128() got %d want %d", g64, e64) + } + gs1 := readStr(slr) + if gs1 != s1 { + t.Fatalf("readStr got %s want %s", gs1, s1) + } + gs2 := readStr(slr) + if gs2 != s2 { + t.Fatalf("readStr got %s want %s", gs2, s2) + } + if _, err := slr.Seek(4, io.SeekStart); err != nil { + t.Fatal(err) + } + off := slr.Offset() + if off != 4 { + t.Fatalf("Offset() returned %d wanted 4", off) + } + g64 = slr.ReadUint64() + if g64 != e64 { + t.Fatalf("post-seek slr.ReadUint64() got %d want %d", g64, e64) + } + } +} + +func appendUleb128(b []byte, v uint) []byte { + for { + c := uint8(v & 0x7f) + v >>= 7 + if v != 0 { + c |= 0x80 + } + b = append(b, c) + if c&0x80 == 0 { + break + } + } + return b +} diff --git a/src/internal/coverage/slicewriter/slicewriter.go b/src/internal/coverage/slicewriter/slicewriter.go new file mode 100644 index 0000000..460e9dc --- /dev/null +++ b/src/internal/coverage/slicewriter/slicewriter.go @@ -0,0 +1,80 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package slicewriter + +import ( + "fmt" + "io" +) + +// WriteSeeker is a helper object that implements the io.WriteSeeker +// interface. Clients can create a WriteSeeker, make a series of Write +// calls to add data to it (and possibly Seek calls to update +// previously written portions), then finally invoke BytesWritten() to +// get a pointer to the constructed byte slice. +type WriteSeeker struct { + payload []byte + off int64 +} + +func (sws *WriteSeeker) Write(p []byte) (n int, err error) { + amt := len(p) + towrite := sws.payload[sws.off:] + if len(towrite) < amt { + sws.payload = append(sws.payload, make([]byte, amt-len(towrite))...) + towrite = sws.payload[sws.off:] + } + copy(towrite, p) + sws.off += int64(amt) + return amt, nil +} + +// Seek repositions the read/write position of the WriteSeeker within +// its internally maintained slice. Note that it is not possible to +// expand the size of the slice using SEEK_SET; trying to seek outside +// the slice will result in an error. +func (sws *WriteSeeker) Seek(offset int64, whence int) (int64, error) { + switch whence { + case io.SeekStart: + if sws.off != offset && (offset < 0 || offset > int64(len(sws.payload))) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", offset, len(sws.payload)) + } + sws.off = offset + return offset, nil + case io.SeekCurrent: + newoff := sws.off + offset + if newoff != sws.off && (newoff < 0 || newoff > int64(len(sws.payload))) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", newoff, len(sws.payload)) + } + sws.off += offset + return sws.off, nil + case io.SeekEnd: + newoff := int64(len(sws.payload)) + offset + if newoff != sws.off && (newoff < 0 || newoff > int64(len(sws.payload))) { + return 0, fmt.Errorf("invalid seek: new offset %d (out of range [0 %d]", newoff, len(sws.payload)) + } + sws.off = newoff + return sws.off, nil + } + // other modes not supported + return 0, fmt.Errorf("unsupported seek mode %d", whence) +} + +// BytesWritten returns the underlying byte slice for the WriteSeeker, +// containing the data written to it via Write/Seek calls. +func (sws *WriteSeeker) BytesWritten() []byte { + return sws.payload +} + +func (sws *WriteSeeker) Read(p []byte) (n int, err error) { + amt := len(p) + toread := sws.payload[sws.off:] + if len(toread) < amt { + amt = len(toread) + } + copy(p, toread) + sws.off += int64(amt) + return amt, nil +} diff --git a/src/internal/coverage/slicewriter/slw_test.go b/src/internal/coverage/slicewriter/slw_test.go new file mode 100644 index 0000000..9e26767 --- /dev/null +++ b/src/internal/coverage/slicewriter/slw_test.go @@ -0,0 +1,134 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package slicewriter + +import ( + "io" + "testing" +) + +func TestSliceWriter(t *testing.T) { + + sleq := func(t *testing.T, got []byte, want []byte) { + t.Helper() + if len(got) != len(want) { + t.Fatalf("bad length got %d want %d", len(got), len(want)) + } + for i := range got { + if got[i] != want[i] { + t.Fatalf("bad read at %d got %d want %d", i, got[i], want[i]) + } + } + } + + wf := func(t *testing.T, ws *WriteSeeker, p []byte) { + t.Helper() + nw, werr := ws.Write(p) + if werr != nil { + t.Fatalf("unexpected write error: %v", werr) + } + if nw != len(p) { + t.Fatalf("wrong amount written want %d got %d", len(p), nw) + } + } + + rf := func(t *testing.T, ws *WriteSeeker, p []byte) { + t.Helper() + b := make([]byte, len(p)) + nr, rerr := ws.Read(b) + if rerr != nil { + t.Fatalf("unexpected read error: %v", rerr) + } + if nr != len(p) { + t.Fatalf("wrong amount read want %d got %d", len(p), nr) + } + sleq(t, b, p) + } + + sk := func(t *testing.T, ws *WriteSeeker, offset int64, whence int) int64 { + t.Helper() + off, err := ws.Seek(offset, whence) + if err != nil { + t.Fatalf("unexpected seek error: %v", err) + } + return off + } + + wp1 := []byte{1, 2} + ws := &WriteSeeker{} + + // write some stuff + wf(t, ws, wp1) + // check that BytesWritten returns what we wrote. + sleq(t, ws.BytesWritten(), wp1) + // offset is at end of slice, so reading should return zero bytes. + rf(t, ws, []byte{}) + + // write some more stuff + wp2 := []byte{7, 8, 9} + wf(t, ws, wp2) + // check that BytesWritten returns what we expect. + wpex := []byte{1, 2, 7, 8, 9} + sleq(t, ws.BytesWritten(), wpex) + rf(t, ws, []byte{}) + + // seeks and reads. + sk(t, ws, 1, io.SeekStart) + rf(t, ws, []byte{2, 7}) + sk(t, ws, -2, io.SeekCurrent) + rf(t, ws, []byte{2, 7}) + sk(t, ws, -4, io.SeekEnd) + rf(t, ws, []byte{2, 7}) + off := sk(t, ws, 0, io.SeekEnd) + sk(t, ws, off, io.SeekStart) + + // seek back and overwrite + sk(t, ws, 1, io.SeekStart) + wf(t, ws, []byte{9, 11}) + wpex = []byte{1, 9, 11, 8, 9} + sleq(t, ws.BytesWritten(), wpex) + + // seeks on empty writer. + ws2 := &WriteSeeker{} + sk(t, ws2, 0, io.SeekStart) + sk(t, ws2, 0, io.SeekCurrent) + sk(t, ws2, 0, io.SeekEnd) + + // check for seek errors. + _, err := ws.Seek(-1, io.SeekStart) + if err == nil { + t.Fatalf("expected error on invalid -1 seek") + } + _, err = ws.Seek(int64(len(ws.BytesWritten())+1), io.SeekStart) + if err == nil { + t.Fatalf("expected error on invalid %d seek", len(ws.BytesWritten())) + } + + ws.Seek(0, io.SeekStart) + _, err = ws.Seek(-1, io.SeekCurrent) + if err == nil { + t.Fatalf("expected error on invalid -1 seek") + } + _, err = ws.Seek(int64(len(ws.BytesWritten())+1), io.SeekCurrent) + if err == nil { + t.Fatalf("expected error on invalid %d seek", len(ws.BytesWritten())) + } + + _, err = ws.Seek(1, io.SeekEnd) + if err == nil { + t.Fatalf("expected error on invalid 1 seek") + } + bsamt := int64(-1*len(ws.BytesWritten()) - 1) + _, err = ws.Seek(bsamt, io.SeekEnd) + if err == nil { + t.Fatalf("expected error on invalid %d seek", bsamt) + } + + // bad seek mode + _, err = ws.Seek(-1, io.SeekStart+9) + if err == nil { + t.Fatalf("expected error on invalid seek mode") + } +} diff --git a/src/internal/coverage/stringtab/stringtab.go b/src/internal/coverage/stringtab/stringtab.go new file mode 100644 index 0000000..156c8ad --- /dev/null +++ b/src/internal/coverage/stringtab/stringtab.go @@ -0,0 +1,139 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package stringtab + +import ( + "fmt" + "internal/coverage/slicereader" + "internal/coverage/uleb128" + "io" +) + +// This package implements string table writer and reader utilities, +// for use in emitting and reading/decoding coverage meta-data and +// counter-data files. + +// Writer implements a string table writing utility. +type Writer struct { + stab map[string]uint32 + strs []string + tmp []byte + frozen bool +} + +// InitWriter initializes a stringtab.Writer. +func (stw *Writer) InitWriter() { + stw.stab = make(map[string]uint32) + stw.tmp = make([]byte, 64) +} + +// Nentries returns the number of strings interned so far. +func (stw *Writer) Nentries() uint32 { + return uint32(len(stw.strs)) +} + +// Lookup looks up string 's' in the writer's table, adding +// a new entry if need be, and returning an index into the table. +func (stw *Writer) Lookup(s string) uint32 { + if idx, ok := stw.stab[s]; ok { + return idx + } + if stw.frozen { + panic("internal error: string table previously frozen") + } + idx := uint32(len(stw.strs)) + stw.stab[s] = idx + stw.strs = append(stw.strs, s) + return idx +} + +// Size computes the memory in bytes needed for the serialized +// version of a stringtab.Writer. +func (stw *Writer) Size() uint32 { + rval := uint32(0) + stw.tmp = stw.tmp[:0] + stw.tmp = uleb128.AppendUleb128(stw.tmp, uint(len(stw.strs))) + rval += uint32(len(stw.tmp)) + for _, s := range stw.strs { + stw.tmp = stw.tmp[:0] + slen := uint(len(s)) + stw.tmp = uleb128.AppendUleb128(stw.tmp, slen) + rval += uint32(len(stw.tmp)) + uint32(slen) + } + return rval +} + +// Write writes the string table in serialized form to the specified +// io.Writer. +func (stw *Writer) Write(w io.Writer) error { + wr128 := func(v uint) error { + stw.tmp = stw.tmp[:0] + stw.tmp = uleb128.AppendUleb128(stw.tmp, v) + if nw, err := w.Write(stw.tmp); err != nil { + return fmt.Errorf("writing string table: %v", err) + } else if nw != len(stw.tmp) { + return fmt.Errorf("short write emitting stringtab uleb") + } + return nil + } + if err := wr128(uint(len(stw.strs))); err != nil { + return err + } + for _, s := range stw.strs { + if err := wr128(uint(len(s))); err != nil { + return err + } + if nw, err := w.Write([]byte(s)); err != nil { + return fmt.Errorf("writing string table: %v", err) + } else if nw != len([]byte(s)) { + return fmt.Errorf("short write emitting stringtab") + } + } + return nil +} + +// Freeze sends a signal to the writer that no more additions are +// allowed, only lookups of existing strings (if a lookup triggers +// addition, a panic will result). Useful as a mechanism for +// "finalizing" a string table prior to writing it out. +func (stw *Writer) Freeze() { + stw.frozen = true +} + +// Reader is a helper for reading a string table previously +// serialized by a Writer.Write call. +type Reader struct { + r *slicereader.Reader + strs []string +} + +// NewReader creates a stringtab.Reader to read the contents +// of a string table from 'r'. +func NewReader(r *slicereader.Reader) *Reader { + str := &Reader{ + r: r, + } + return str +} + +// Read reads/decodes a string table using the reader provided. +func (str *Reader) Read() { + numEntries := int(str.r.ReadULEB128()) + str.strs = make([]string, 0, numEntries) + for idx := 0; idx < numEntries; idx++ { + slen := str.r.ReadULEB128() + str.strs = append(str.strs, str.r.ReadString(int64(slen))) + } +} + +// Entries returns the number of decoded entries in a string table. +func (str *Reader) Entries() int { + return len(str.strs) +} + +// Get returns string 'idx' within the string table. +func (str *Reader) Get(idx uint32) string { + return str.strs[idx] +} diff --git a/src/internal/coverage/test/counter_test.go b/src/internal/coverage/test/counter_test.go new file mode 100644 index 0000000..e29baed --- /dev/null +++ b/src/internal/coverage/test/counter_test.go @@ -0,0 +1,237 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package test + +import ( + "fmt" + "internal/coverage" + "internal/coverage/decodecounter" + "internal/coverage/encodecounter" + "io" + "os" + "path/filepath" + "testing" +) + +type ctrVis struct { + funcs []decodecounter.FuncPayload +} + +func (v *ctrVis) VisitFuncs(f encodecounter.CounterVisitorFn) error { + for _, fn := range v.funcs { + if err := f(fn.PkgIdx, fn.FuncIdx, fn.Counters); err != nil { + return err + } + } + return nil +} + +func mkfunc(p uint32, f uint32, c []uint32) decodecounter.FuncPayload { + return decodecounter.FuncPayload{ + PkgIdx: p, + FuncIdx: f, + Counters: c, + } +} + +func TestCounterDataWriterReader(t *testing.T) { + flavors := []coverage.CounterFlavor{ + coverage.CtrRaw, + coverage.CtrULeb128, + } + + isDead := func(fp decodecounter.FuncPayload) bool { + for _, v := range fp.Counters { + if v != 0 { + return false + } + } + return true + } + + funcs := []decodecounter.FuncPayload{ + mkfunc(0, 0, []uint32{13, 14, 15}), + mkfunc(0, 1, []uint32{16, 17}), + mkfunc(1, 0, []uint32{18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 976543, 7}), + } + writeVisitor := &ctrVis{funcs: funcs} + + for kf, flav := range flavors { + + t.Logf("testing flavor %d\n", flav) + + // Open a counter data file in preparation for emitting data. + d := t.TempDir() + cfpath := filepath.Join(d, fmt.Sprintf("covcounters.hash.0.%d", kf)) + of, err := os.OpenFile(cfpath, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0666) + if err != nil { + t.Fatalf("opening covcounters: %v", err) + } + + // Perform the encode and write. + cdfw := encodecounter.NewCoverageDataWriter(of, flav) + if cdfw == nil { + t.Fatalf("NewCoverageDataWriter failed") + } + finalHash := [16]byte{1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0} + args := map[string]string{"argc": "3", "argv0": "arg0", "argv1": "arg1", "argv2": "arg_________2"} + if err := cdfw.Write(finalHash, args, writeVisitor); err != nil { + t.Fatalf("counter file Write failed: %v", err) + } + if err := of.Close(); err != nil { + t.Fatalf("closing covcounters: %v", err) + } + cdfw = nil + + // Decode the same file. + var cdr *decodecounter.CounterDataReader + inf, err := os.Open(cfpath) + defer func() { + if err := inf.Close(); err != nil { + t.Fatalf("close failed with: %v", err) + } + }() + + if err != nil { + t.Fatalf("reopening covcounters file: %v", err) + } + if cdr, err = decodecounter.NewCounterDataReader(cfpath, inf); err != nil { + t.Fatalf("opening covcounters for read: %v", err) + } + decodedArgs := cdr.OsArgs() + aWant := "[arg0 arg1 arg_________2]" + aGot := fmt.Sprintf("%+v", decodedArgs) + if aWant != aGot { + t.Errorf("reading decoded args, got %s want %s", aGot, aWant) + } + for i := range funcs { + if isDead(funcs[i]) { + continue + } + var fp decodecounter.FuncPayload + if ok, err := cdr.NextFunc(&fp); err != nil { + t.Fatalf("reading func %d: %v", i, err) + } else if !ok { + t.Fatalf("reading func %d: bad return", i) + } + got := fmt.Sprintf("%+v", fp) + want := fmt.Sprintf("%+v", funcs[i]) + if got != want { + t.Errorf("cdr.NextFunc iter %d\ngot %+v\nwant %+v", i, got, want) + } + } + var dummy decodecounter.FuncPayload + if ok, err := cdr.NextFunc(&dummy); err != nil { + t.Fatalf("reading func after loop: %v", err) + } else if ok { + t.Fatalf("reading func after loop: expected EOF") + } + } +} + +func TestCounterDataAppendSegment(t *testing.T) { + d := t.TempDir() + cfpath := filepath.Join(d, "covcounters.hash2.0") + of, err := os.OpenFile(cfpath, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0666) + if err != nil { + t.Fatalf("opening covcounters: %v", err) + } + + const numSegments = 2 + + // Write a counter with with multiple segments. + args := map[string]string{"argc": "1", "argv0": "prog.exe"} + allfuncs := [][]decodecounter.FuncPayload{} + ctrs := []uint32{} + q := uint32(0) + var cdfw *encodecounter.CoverageDataWriter + for idx := 0; idx < numSegments; idx++ { + args[fmt.Sprintf("seg%d", idx)] = "x" + q += 7 + ctrs = append(ctrs, q) + funcs := []decodecounter.FuncPayload{} + for k := 0; k < idx+1; k++ { + c := make([]uint32, len(ctrs)) + copy(c, ctrs) + funcs = append(funcs, mkfunc(uint32(idx), uint32(k), c)) + } + allfuncs = append(allfuncs, funcs) + + writeVisitor := &ctrVis{funcs: funcs} + + if idx == 0 { + // Perform the encode and write. + cdfw = encodecounter.NewCoverageDataWriter(of, coverage.CtrRaw) + if cdfw == nil { + t.Fatalf("NewCoverageDataWriter failed") + } + finalHash := [16]byte{1, 2} + if err := cdfw.Write(finalHash, args, writeVisitor); err != nil { + t.Fatalf("counter file Write failed: %v", err) + } + } else { + if err := cdfw.AppendSegment(args, writeVisitor); err != nil { + t.Fatalf("counter file AppendSegment failed: %v", err) + } + } + } + if err := of.Close(); err != nil { + t.Fatalf("closing covcounters: %v", err) + } + + // Read the result file. + var cdr *decodecounter.CounterDataReader + inf, err := os.Open(cfpath) + defer func() { + if err := inf.Close(); err != nil { + t.Fatalf("close failed with: %v", err) + } + }() + + if err != nil { + t.Fatalf("reopening covcounters file: %v", err) + } + if cdr, err = decodecounter.NewCounterDataReader(cfpath, inf); err != nil { + t.Fatalf("opening covcounters for read: %v", err) + } + ns := cdr.NumSegments() + if ns != numSegments { + t.Fatalf("got %d segments want %d", ns, numSegments) + } + if len(allfuncs) != numSegments { + t.Fatalf("expected %d got %d", numSegments, len(allfuncs)) + } + + for sidx := 0; sidx < int(ns); sidx++ { + if off, err := inf.Seek(0, io.SeekCurrent); err != nil { + t.Fatalf("Seek failed: %v", err) + } else { + t.Logf("sidx=%d off=%d\n", sidx, off) + } + + if sidx != 0 { + if ok, err := cdr.BeginNextSegment(); err != nil { + t.Fatalf("BeginNextSegment failed: %v", err) + } else if !ok { + t.Fatalf("BeginNextSegment return %v on iter %d", + ok, sidx) + } + } + funcs := allfuncs[sidx] + for i := range funcs { + var fp decodecounter.FuncPayload + if ok, err := cdr.NextFunc(&fp); err != nil { + t.Fatalf("reading func %d: %v", i, err) + } else if !ok { + t.Fatalf("reading func %d: bad return", i) + } + got := fmt.Sprintf("%+v", fp) + want := fmt.Sprintf("%+v", funcs[i]) + if got != want { + t.Errorf("cdr.NextFunc iter %d\ngot %+v\nwant %+v", i, got, want) + } + } + } +} diff --git a/src/internal/coverage/test/roundtrip_test.go b/src/internal/coverage/test/roundtrip_test.go new file mode 100644 index 0000000..614f56e --- /dev/null +++ b/src/internal/coverage/test/roundtrip_test.go @@ -0,0 +1,331 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package test + +import ( + "fmt" + "internal/coverage" + "internal/coverage/decodemeta" + "internal/coverage/encodemeta" + "internal/coverage/slicewriter" + "io" + "os" + "path/filepath" + "testing" +) + +func cmpFuncDesc(want, got coverage.FuncDesc) string { + swant := fmt.Sprintf("%+v", want) + sgot := fmt.Sprintf("%+v", got) + if swant == sgot { + return "" + } + return fmt.Sprintf("wanted %q got %q", swant, sgot) +} + +func TestMetaDataEmptyPackage(t *testing.T) { + // Make sure that encoding/decoding works properly with packages + // that don't actually have any functions. + p := "empty/package" + pn := "package" + mp := "m" + b, err := encodemeta.NewCoverageMetaDataBuilder(p, pn, mp) + if err != nil { + t.Fatalf("making builder: %v", err) + } + drws := &slicewriter.WriteSeeker{} + b.Emit(drws) + drws.Seek(0, io.SeekStart) + dec, err := decodemeta.NewCoverageMetaDataDecoder(drws.BytesWritten(), false) + if err != nil { + t.Fatalf("making decoder: %v", err) + } + nf := dec.NumFuncs() + if nf != 0 { + t.Errorf("dec.NumFuncs(): got %d want %d", nf, 0) + } + pp := dec.PackagePath() + if pp != p { + t.Errorf("dec.PackagePath(): got %s want %s", pp, p) + } + ppn := dec.PackageName() + if ppn != pn { + t.Errorf("dec.PackageName(): got %s want %s", ppn, pn) + } + pmp := dec.ModulePath() + if pmp != mp { + t.Errorf("dec.ModulePath(): got %s want %s", pmp, mp) + } +} + +func TestMetaDataEncoderDecoder(t *testing.T) { + // Test encode path. + pp := "foo/bar/pkg" + pn := "pkg" + mp := "barmod" + b, err := encodemeta.NewCoverageMetaDataBuilder(pp, pn, mp) + if err != nil { + t.Fatalf("making builder: %v", err) + } + f1 := coverage.FuncDesc{ + Funcname: "func", + Srcfile: "foo.go", + Units: []coverage.CoverableUnit{ + coverage.CoverableUnit{StLine: 1, StCol: 2, EnLine: 3, EnCol: 4, NxStmts: 5}, + coverage.CoverableUnit{StLine: 6, StCol: 7, EnLine: 8, EnCol: 9, NxStmts: 10}, + }, + } + idx := b.AddFunc(f1) + if idx != 0 { + t.Errorf("b.AddFunc(f1) got %d want %d", idx, 0) + } + + f2 := coverage.FuncDesc{ + Funcname: "xfunc", + Srcfile: "bar.go", + Units: []coverage.CoverableUnit{ + coverage.CoverableUnit{StLine: 1, StCol: 2, EnLine: 3, EnCol: 4, NxStmts: 5}, + coverage.CoverableUnit{StLine: 6, StCol: 7, EnLine: 8, EnCol: 9, NxStmts: 10}, + coverage.CoverableUnit{StLine: 11, StCol: 12, EnLine: 13, EnCol: 14, NxStmts: 15}, + }, + } + idx = b.AddFunc(f2) + if idx != 1 { + t.Errorf("b.AddFunc(f2) got %d want %d", idx, 0) + } + + // Emit into a writer. + drws := &slicewriter.WriteSeeker{} + b.Emit(drws) + + // Test decode path. + drws.Seek(0, io.SeekStart) + dec, err := decodemeta.NewCoverageMetaDataDecoder(drws.BytesWritten(), false) + if err != nil { + t.Fatalf("NewCoverageMetaDataDecoder error: %v", err) + } + nf := dec.NumFuncs() + if nf != 2 { + t.Errorf("dec.NumFuncs(): got %d want %d", nf, 2) + } + + gotpp := dec.PackagePath() + if gotpp != pp { + t.Errorf("packagepath: got %s want %s", gotpp, pp) + } + gotpn := dec.PackageName() + if gotpn != pn { + t.Errorf("packagename: got %s want %s", gotpn, pn) + } + + cases := []coverage.FuncDesc{f1, f2} + for i := uint32(0); i < uint32(len(cases)); i++ { + var fn coverage.FuncDesc + if err := dec.ReadFunc(i, &fn); err != nil { + t.Fatalf("err reading function %d: %v", i, err) + } + res := cmpFuncDesc(cases[i], fn) + if res != "" { + t.Errorf("ReadFunc(%d): %s", i, res) + } + } +} + +func createFuncs(i int) []coverage.FuncDesc { + res := []coverage.FuncDesc{} + lc := uint32(1) + for fi := 0; fi < i+1; fi++ { + units := []coverage.CoverableUnit{} + for ui := 0; ui < (fi+1)*(i+1); ui++ { + units = append(units, + coverage.CoverableUnit{StLine: lc, StCol: lc + 1, + EnLine: lc + 2, EnCol: lc + 3, NxStmts: lc + 4, + }) + lc += 5 + } + f := coverage.FuncDesc{ + Funcname: fmt.Sprintf("func_%d_%d", i, fi), + Srcfile: fmt.Sprintf("foo_%d.go", i), + Units: units, + } + res = append(res, f) + } + return res +} + +func createBlob(t *testing.T, i int) []byte { + nomodule := "" + b, err := encodemeta.NewCoverageMetaDataBuilder("foo/pkg", "pkg", nomodule) + if err != nil { + t.Fatalf("making builder: %v", err) + } + + funcs := createFuncs(i) + for _, f := range funcs { + b.AddFunc(f) + } + drws := &slicewriter.WriteSeeker{} + b.Emit(drws) + return drws.BytesWritten() +} + +func createMetaDataBlobs(t *testing.T, nb int) [][]byte { + res := [][]byte{} + for i := 0; i < nb; i++ { + res = append(res, createBlob(t, i)) + } + return res +} + +func TestMetaDataWriterReader(t *testing.T) { + d := t.TempDir() + + // Emit a meta-file... + mfpath := filepath.Join(d, "covmeta.hash.0") + of, err := os.OpenFile(mfpath, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0666) + if err != nil { + t.Fatalf("opening covmeta: %v", err) + } + //t.Logf("meta-file path is %s", mfpath) + blobs := createMetaDataBlobs(t, 7) + gran := coverage.CtrGranularityPerBlock + mfw := encodemeta.NewCoverageMetaFileWriter(mfpath, of) + finalHash := [16]byte{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15} + err = mfw.Write(finalHash, blobs, coverage.CtrModeAtomic, gran) + if err != nil { + t.Fatalf("writing meta-file: %v", err) + } + if err = of.Close(); err != nil { + t.Fatalf("closing meta-file: %v", err) + } + + // ... then read it back in, first time without setting fileView, + // second time setting it. + for k := 0; k < 2; k++ { + var fileView []byte + + inf, err := os.Open(mfpath) + if err != nil { + t.Fatalf("open() on meta-file: %v", err) + } + + if k != 0 { + // Use fileview to exercise different paths in reader. + fi, err := os.Stat(mfpath) + if err != nil { + t.Fatalf("stat() on meta-file: %v", err) + } + fileView = make([]byte, fi.Size()) + if _, err := inf.Read(fileView); err != nil { + t.Fatalf("read() on meta-file: %v", err) + } + if _, err := inf.Seek(int64(0), io.SeekStart); err != nil { + t.Fatalf("seek() on meta-file: %v", err) + } + } + + mfr, err := decodemeta.NewCoverageMetaFileReader(inf, fileView) + if err != nil { + t.Fatalf("k=%d NewCoverageMetaFileReader failed with: %v", k, err) + } + np := mfr.NumPackages() + if np != 7 { + t.Fatalf("k=%d wanted 7 packages got %d", k, np) + } + md := mfr.CounterMode() + wmd := coverage.CtrModeAtomic + if md != wmd { + t.Fatalf("k=%d wanted mode %d got %d", k, wmd, md) + } + gran := mfr.CounterGranularity() + wgran := coverage.CtrGranularityPerBlock + if gran != wgran { + t.Fatalf("k=%d wanted gran %d got %d", k, wgran, gran) + } + + payload := []byte{} + for pi := 0; pi < int(np); pi++ { + var pd *decodemeta.CoverageMetaDataDecoder + var err error + pd, payload, err = mfr.GetPackageDecoder(uint32(pi), payload) + if err != nil { + t.Fatalf("GetPackageDecoder(%d) failed with: %v", pi, err) + } + efuncs := createFuncs(pi) + nf := pd.NumFuncs() + if len(efuncs) != int(nf) { + t.Fatalf("decoding pk %d wanted %d funcs got %d", + pi, len(efuncs), nf) + } + var f coverage.FuncDesc + for fi := 0; fi < int(nf); fi++ { + if err := pd.ReadFunc(uint32(fi), &f); err != nil { + t.Fatalf("ReadFunc(%d) pk %d got error %v", + fi, pi, err) + } + res := cmpFuncDesc(efuncs[fi], f) + if res != "" { + t.Errorf("ReadFunc(%d) pk %d: %s", fi, pi, res) + } + } + } + inf.Close() + } +} + +func TestMetaDataDecodeLitFlagIssue57942(t *testing.T) { + + // Encode a package with a few functions. The funcs alternate + // between regular functions and function literals. + pp := "foo/bar/pkg" + pn := "pkg" + mp := "barmod" + b, err := encodemeta.NewCoverageMetaDataBuilder(pp, pn, mp) + if err != nil { + t.Fatalf("making builder: %v", err) + } + const NF = 6 + const NCU = 1 + ln := uint32(10) + wantfds := []coverage.FuncDesc{} + for fi := uint32(0); fi < NF; fi++ { + fis := fmt.Sprintf("%d", fi) + fd := coverage.FuncDesc{ + Funcname: "func" + fis, + Srcfile: "foo" + fis + ".go", + Units: []coverage.CoverableUnit{ + coverage.CoverableUnit{StLine: ln + 1, StCol: 2, EnLine: ln + 3, EnCol: 4, NxStmts: fi + 2}, + }, + Lit: (fi % 2) == 0, + } + wantfds = append(wantfds, fd) + b.AddFunc(fd) + } + + // Emit into a writer. + drws := &slicewriter.WriteSeeker{} + b.Emit(drws) + + // Decode the result. + drws.Seek(0, io.SeekStart) + dec, err := decodemeta.NewCoverageMetaDataDecoder(drws.BytesWritten(), false) + if err != nil { + t.Fatalf("making decoder: %v", err) + } + nf := dec.NumFuncs() + if nf != NF { + t.Fatalf("decoder number of functions: got %d want %d", nf, NF) + } + var fn coverage.FuncDesc + for i := uint32(0); i < uint32(NF); i++ { + if err := dec.ReadFunc(i, &fn); err != nil { + t.Fatalf("err reading function %d: %v", i, err) + } + res := cmpFuncDesc(wantfds[i], fn) + if res != "" { + t.Errorf("ReadFunc(%d): %s", i, res) + } + } +} diff --git a/src/internal/coverage/uleb128/uleb128.go b/src/internal/coverage/uleb128/uleb128.go new file mode 100644 index 0000000..e5cd92a --- /dev/null +++ b/src/internal/coverage/uleb128/uleb128.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package uleb128 + +func AppendUleb128(b []byte, v uint) []byte { + for { + c := uint8(v & 0x7f) + v >>= 7 + if v != 0 { + c |= 0x80 + } + b = append(b, c) + if c&0x80 == 0 { + break + } + } + return b +} diff --git a/src/internal/cpu/cpu.go b/src/internal/cpu/cpu.go new file mode 100644 index 0000000..1352810 --- /dev/null +++ b/src/internal/cpu/cpu.go @@ -0,0 +1,222 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package cpu implements processor feature detection +// used by the Go standard library. +package cpu + +// DebugOptions is set to true by the runtime if the OS supports reading +// GODEBUG early in runtime startup. +// This should not be changed after it is initialized. +var DebugOptions bool + +// CacheLinePad is used to pad structs to avoid false sharing. +type CacheLinePad struct{ _ [CacheLinePadSize]byte } + +// CacheLineSize is the CPU's assumed cache line size. +// There is currently no runtime detection of the real cache line size +// so we use the constant per GOARCH CacheLinePadSize as an approximation. +var CacheLineSize uintptr = CacheLinePadSize + +// The booleans in X86 contain the correspondingly named cpuid feature bit. +// HasAVX and HasAVX2 are only set if the OS does support XMM and YMM registers +// in addition to the cpuid feature bit being set. +// The struct is padded to avoid false sharing. +var X86 struct { + _ CacheLinePad + HasAES bool + HasADX bool + HasAVX bool + HasAVX2 bool + HasBMI1 bool + HasBMI2 bool + HasERMS bool + HasFMA bool + HasOSXSAVE bool + HasPCLMULQDQ bool + HasPOPCNT bool + HasRDTSCP bool + HasSHA bool + HasSSE3 bool + HasSSSE3 bool + HasSSE41 bool + HasSSE42 bool + _ CacheLinePad +} + +// The booleans in ARM contain the correspondingly named cpu feature bit. +// The struct is padded to avoid false sharing. +var ARM struct { + _ CacheLinePad + HasVFPv4 bool + HasIDIVA bool + _ CacheLinePad +} + +// The booleans in ARM64 contain the correspondingly named cpu feature bit. +// The struct is padded to avoid false sharing. +var ARM64 struct { + _ CacheLinePad + HasAES bool + HasPMULL bool + HasSHA1 bool + HasSHA2 bool + HasSHA512 bool + HasCRC32 bool + HasATOMICS bool + HasCPUID bool + IsNeoverse bool + _ CacheLinePad +} + +var MIPS64X struct { + _ CacheLinePad + HasMSA bool // MIPS SIMD architecture + _ CacheLinePad +} + +// For ppc64(le), it is safe to check only for ISA level starting on ISA v3.00, +// since there are no optional categories. There are some exceptions that also +// require kernel support to work (darn, scv), so there are feature bits for +// those as well. The minimum processor requirement is POWER8 (ISA 2.07). +// The struct is padded to avoid false sharing. +var PPC64 struct { + _ CacheLinePad + HasDARN bool // Hardware random number generator (requires kernel enablement) + HasSCV bool // Syscall vectored (requires kernel enablement) + IsPOWER8 bool // ISA v2.07 (POWER8) + IsPOWER9 bool // ISA v3.00 (POWER9) + IsPOWER10 bool // ISA v3.1 (POWER10) + _ CacheLinePad +} + +var S390X struct { + _ CacheLinePad + HasZARCH bool // z architecture mode is active [mandatory] + HasSTFLE bool // store facility list extended [mandatory] + HasLDISP bool // long (20-bit) displacements [mandatory] + HasEIMM bool // 32-bit immediates [mandatory] + HasDFP bool // decimal floating point + HasETF3EH bool // ETF-3 enhanced + HasMSA bool // message security assist (CPACF) + HasAES bool // KM-AES{128,192,256} functions + HasAESCBC bool // KMC-AES{128,192,256} functions + HasAESCTR bool // KMCTR-AES{128,192,256} functions + HasAESGCM bool // KMA-GCM-AES{128,192,256} functions + HasGHASH bool // KIMD-GHASH function + HasSHA1 bool // K{I,L}MD-SHA-1 functions + HasSHA256 bool // K{I,L}MD-SHA-256 functions + HasSHA512 bool // K{I,L}MD-SHA-512 functions + HasSHA3 bool // K{I,L}MD-SHA3-{224,256,384,512} and K{I,L}MD-SHAKE-{128,256} functions + HasVX bool // vector facility. Note: the runtime sets this when it processes auxv records. + HasVXE bool // vector-enhancements facility 1 + HasKDSA bool // elliptic curve functions + HasECDSA bool // NIST curves + HasEDDSA bool // Edwards curves + _ CacheLinePad +} + +// Initialize examines the processor and sets the relevant variables above. +// This is called by the runtime package early in program initialization, +// before normal init functions are run. env is set by runtime if the OS supports +// cpu feature options in GODEBUG. +func Initialize(env string) { + doinit() + processOptions(env) +} + +// options contains the cpu debug options that can be used in GODEBUG. +// Options are arch dependent and are added by the arch specific doinit functions. +// Features that are mandatory for the specific GOARCH should not be added to options +// (e.g. SSE2 on amd64). +var options []option + +// Option names should be lower case. e.g. avx instead of AVX. +type option struct { + Name string + Feature *bool + Specified bool // whether feature value was specified in GODEBUG + Enable bool // whether feature should be enabled +} + +// processOptions enables or disables CPU feature values based on the parsed env string. +// The env string is expected to be of the form cpu.feature1=value1,cpu.feature2=value2... +// where feature names is one of the architecture specific list stored in the +// cpu packages options variable and values are either 'on' or 'off'. +// If env contains cpu.all=off then all cpu features referenced through the options +// variable are disabled. Other feature names and values result in warning messages. +func processOptions(env string) { +field: + for env != "" { + field := "" + i := indexByte(env, ',') + if i < 0 { + field, env = env, "" + } else { + field, env = env[:i], env[i+1:] + } + if len(field) < 4 || field[:4] != "cpu." { + continue + } + i = indexByte(field, '=') + if i < 0 { + print("GODEBUG: no value specified for \"", field, "\"\n") + continue + } + key, value := field[4:i], field[i+1:] // e.g. "SSE2", "on" + + var enable bool + switch value { + case "on": + enable = true + case "off": + enable = false + default: + print("GODEBUG: value \"", value, "\" not supported for cpu option \"", key, "\"\n") + continue field + } + + if key == "all" { + for i := range options { + options[i].Specified = true + options[i].Enable = enable + } + continue field + } + + for i := range options { + if options[i].Name == key { + options[i].Specified = true + options[i].Enable = enable + continue field + } + } + + print("GODEBUG: unknown cpu feature \"", key, "\"\n") + } + + for _, o := range options { + if !o.Specified { + continue + } + + if o.Enable && !*o.Feature { + print("GODEBUG: can not enable \"", o.Name, "\", missing CPU support\n") + continue + } + + *o.Feature = o.Enable + } +} + +// indexByte returns the index of the first instance of c in s, +// or -1 if c is not present in s. +func indexByte(s string, c byte) int { + for i := 0; i < len(s); i++ { + if s[i] == c { + return i + } + } + return -1 +} diff --git a/src/internal/cpu/cpu.s b/src/internal/cpu/cpu.s new file mode 100644 index 0000000..3c770c1 --- /dev/null +++ b/src/internal/cpu/cpu.s @@ -0,0 +1,6 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This assembly file exists to allow internal/cpu to call +// non-exported runtime functions that use "go:linkname".
\ No newline at end of file diff --git a/src/internal/cpu/cpu_arm.go b/src/internal/cpu/cpu_arm.go new file mode 100644 index 0000000..b624526 --- /dev/null +++ b/src/internal/cpu/cpu_arm.go @@ -0,0 +1,34 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 32 + +// arm doesn't have a 'cpuid' equivalent, so we rely on HWCAP/HWCAP2. +// These are initialized by archauxv() and should not be changed after they are +// initialized. +var HWCap uint +var HWCap2 uint + +// HWCAP/HWCAP2 bits. These are exposed by Linux and FreeBSD. +const ( + hwcap_VFPv4 = 1 << 16 + hwcap_IDIVA = 1 << 17 +) + +func doinit() { + options = []option{ + {Name: "vfpv4", Feature: &ARM.HasVFPv4}, + {Name: "idiva", Feature: &ARM.HasIDIVA}, + } + + // HWCAP feature bits + ARM.HasVFPv4 = isSet(HWCap, hwcap_VFPv4) + ARM.HasIDIVA = isSet(HWCap, hwcap_IDIVA) +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} diff --git a/src/internal/cpu/cpu_arm64.go b/src/internal/cpu/cpu_arm64.go new file mode 100644 index 0000000..4a302f2 --- /dev/null +++ b/src/internal/cpu/cpu_arm64.go @@ -0,0 +1,69 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +// CacheLinePadSize is used to prevent false sharing of cache lines. +// We choose 128 because Apple Silicon, a.k.a. M1, has 128-byte cache line size. +// It doesn't cost much and is much more future-proof. +const CacheLinePadSize = 128 + +func doinit() { + options = []option{ + {Name: "aes", Feature: &ARM64.HasAES}, + {Name: "pmull", Feature: &ARM64.HasPMULL}, + {Name: "sha1", Feature: &ARM64.HasSHA1}, + {Name: "sha2", Feature: &ARM64.HasSHA2}, + {Name: "sha512", Feature: &ARM64.HasSHA512}, + {Name: "crc32", Feature: &ARM64.HasCRC32}, + {Name: "atomics", Feature: &ARM64.HasATOMICS}, + {Name: "cpuid", Feature: &ARM64.HasCPUID}, + {Name: "isNeoverse", Feature: &ARM64.IsNeoverse}, + } + + // arm64 uses different ways to detect CPU features at runtime depending on the operating system. + osInit() +} + +func getisar0() uint64 + +func getMIDR() uint64 + +func extractBits(data uint64, start, end uint) uint { + return (uint)(data>>start) & ((1 << (end - start + 1)) - 1) +} + +func parseARM64SystemRegisters(isar0 uint64) { + // ID_AA64ISAR0_EL1 + switch extractBits(isar0, 4, 7) { + case 1: + ARM64.HasAES = true + case 2: + ARM64.HasAES = true + ARM64.HasPMULL = true + } + + switch extractBits(isar0, 8, 11) { + case 1: + ARM64.HasSHA1 = true + } + + switch extractBits(isar0, 12, 15) { + case 1: + ARM64.HasSHA2 = true + case 2: + ARM64.HasSHA2 = true + ARM64.HasSHA512 = true + } + + switch extractBits(isar0, 16, 19) { + case 1: + ARM64.HasCRC32 = true + } + + switch extractBits(isar0, 20, 23) { + case 2: + ARM64.HasATOMICS = true + } +} diff --git a/src/internal/cpu/cpu_arm64.s b/src/internal/cpu/cpu_arm64.s new file mode 100644 index 0000000..d6e7f44 --- /dev/null +++ b/src/internal/cpu/cpu_arm64.s @@ -0,0 +1,18 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +// func getisar0() uint64 +TEXT ·getisar0(SB),NOSPLIT,$0 + // get Instruction Set Attributes 0 into R0 + MRS ID_AA64ISAR0_EL1, R0 + MOVD R0, ret+0(FP) + RET + +// func getMIDR() uint64 +TEXT ·getMIDR(SB), NOSPLIT, $0-8 + MRS MIDR_EL1, R0 + MOVD R0, ret+0(FP) + RET diff --git a/src/internal/cpu/cpu_arm64_android.go b/src/internal/cpu/cpu_arm64_android.go new file mode 100644 index 0000000..fbdf7ba --- /dev/null +++ b/src/internal/cpu/cpu_arm64_android.go @@ -0,0 +1,11 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 + +package cpu + +func osInit() { + hwcapInit("android") +} diff --git a/src/internal/cpu/cpu_arm64_darwin.go b/src/internal/cpu/cpu_arm64_darwin.go new file mode 100644 index 0000000..60beadd --- /dev/null +++ b/src/internal/cpu/cpu_arm64_darwin.go @@ -0,0 +1,33 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 && darwin && !ios + +package cpu + +func osInit() { + ARM64.HasATOMICS = sysctlEnabled([]byte("hw.optional.armv8_1_atomics\x00")) + ARM64.HasCRC32 = sysctlEnabled([]byte("hw.optional.armv8_crc32\x00")) + ARM64.HasSHA512 = sysctlEnabled([]byte("hw.optional.armv8_2_sha512\x00")) + + // There are no hw.optional sysctl values for the below features on Mac OS 11.0 + // to detect their supported state dynamically. Assume the CPU features that + // Apple Silicon M1 supports to be available as a minimal set of features + // to all Go programs running on darwin/arm64. + ARM64.HasAES = true + ARM64.HasPMULL = true + ARM64.HasSHA1 = true + ARM64.HasSHA2 = true +} + +//go:noescape +func getsysctlbyname(name []byte) (int32, int32) + +func sysctlEnabled(name []byte) bool { + ret, value := getsysctlbyname(name) + if ret < 0 { + return false + } + return value > 0 +} diff --git a/src/internal/cpu/cpu_arm64_freebsd.go b/src/internal/cpu/cpu_arm64_freebsd.go new file mode 100644 index 0000000..96ed359 --- /dev/null +++ b/src/internal/cpu/cpu_arm64_freebsd.go @@ -0,0 +1,14 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 + +package cpu + +func osInit() { + // Retrieve info from system register ID_AA64ISAR0_EL1. + isar0 := getisar0() + + parseARM64SystemRegisters(isar0) +} diff --git a/src/internal/cpu/cpu_arm64_hwcap.go b/src/internal/cpu/cpu_arm64_hwcap.go new file mode 100644 index 0000000..2fabbb6 --- /dev/null +++ b/src/internal/cpu/cpu_arm64_hwcap.go @@ -0,0 +1,66 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 && linux + +package cpu + +// HWCap may be initialized by archauxv and +// should not be changed after it was initialized. +var HWCap uint + +// HWCAP bits. These are exposed by Linux. +const ( + hwcap_AES = 1 << 3 + hwcap_PMULL = 1 << 4 + hwcap_SHA1 = 1 << 5 + hwcap_SHA2 = 1 << 6 + hwcap_CRC32 = 1 << 7 + hwcap_ATOMICS = 1 << 8 + hwcap_CPUID = 1 << 11 + hwcap_SHA512 = 1 << 21 +) + +func hwcapInit(os string) { + // HWCap was populated by the runtime from the auxiliary vector. + // Use HWCap information since reading aarch64 system registers + // is not supported in user space on older linux kernels. + ARM64.HasAES = isSet(HWCap, hwcap_AES) + ARM64.HasPMULL = isSet(HWCap, hwcap_PMULL) + ARM64.HasSHA1 = isSet(HWCap, hwcap_SHA1) + ARM64.HasSHA2 = isSet(HWCap, hwcap_SHA2) + ARM64.HasCRC32 = isSet(HWCap, hwcap_CRC32) + ARM64.HasCPUID = isSet(HWCap, hwcap_CPUID) + ARM64.HasSHA512 = isSet(HWCap, hwcap_SHA512) + + // The Samsung S9+ kernel reports support for atomics, but not all cores + // actually support them, resulting in SIGILL. See issue #28431. + // TODO(elias.naur): Only disable the optimization on bad chipsets on android. + ARM64.HasATOMICS = isSet(HWCap, hwcap_ATOMICS) && os != "android" + + // Check to see if executing on a Neoverse core and in order to do that, + // check the AUXV for the CPUID bit. The getMIDR function executes an + // instruction which would normally be an illegal instruction, but it's + // trapped by the kernel, the value sanitized and then returned. + // Without the CPUID bit the kernel will not trap the instruction and the + // process will be terminated with SIGILL. + if ARM64.HasCPUID { + midr := getMIDR() + part_num := uint16((midr >> 4) & 0xfff) + implementor := byte((midr >> 24) & 0xff) + + // d0c - NeoverseN1 + // d40 - NeoverseV1 + // d49 - NeoverseN2 + // d4f - NeoverseV2 + if implementor == 'A' && (part_num == 0xd0c || part_num == 0xd40 || + part_num == 0xd49 || part_num == 0xd4f) { + ARM64.IsNeoverse = true + } + } +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} diff --git a/src/internal/cpu/cpu_arm64_linux.go b/src/internal/cpu/cpu_arm64_linux.go new file mode 100644 index 0000000..d746bdb --- /dev/null +++ b/src/internal/cpu/cpu_arm64_linux.go @@ -0,0 +1,11 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 && linux && !android + +package cpu + +func osInit() { + hwcapInit("linux") +} diff --git a/src/internal/cpu/cpu_arm64_openbsd.go b/src/internal/cpu/cpu_arm64_openbsd.go new file mode 100644 index 0000000..1259309 --- /dev/null +++ b/src/internal/cpu/cpu_arm64_openbsd.go @@ -0,0 +1,28 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 + +package cpu + +const ( + // From OpenBSD's sys/sysctl.h. + _CTL_MACHDEP = 7 + + // From OpenBSD's machine/cpu.h. + _CPU_ID_AA64ISAR0 = 2 + _CPU_ID_AA64ISAR1 = 3 +) + +//go:noescape +func sysctlUint64(mib []uint32) (uint64, bool) + +func osInit() { + // Get ID_AA64ISAR0 from sysctl. + isar0, ok := sysctlUint64([]uint32{_CTL_MACHDEP, _CPU_ID_AA64ISAR0}) + if !ok { + return + } + parseARM64SystemRegisters(isar0) +} diff --git a/src/internal/cpu/cpu_arm64_other.go b/src/internal/cpu/cpu_arm64_other.go new file mode 100644 index 0000000..44592cf --- /dev/null +++ b/src/internal/cpu/cpu_arm64_other.go @@ -0,0 +1,13 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 && !linux && !freebsd && !android && (!darwin || ios) && !openbsd + +package cpu + +func osInit() { + // Other operating systems do not support reading HWCap from auxiliary vector, + // reading privileged aarch64 system registers or sysctl in user space to detect + // CPU features at runtime. +} diff --git a/src/internal/cpu/cpu_loong64.go b/src/internal/cpu/cpu_loong64.go new file mode 100644 index 0000000..1c90c24 --- /dev/null +++ b/src/internal/cpu/cpu_loong64.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build loong64 + +package cpu + +// CacheLinePadSize is used to prevent false sharing of cache lines. +// We choose 64 because Loongson 3A5000 the L1 Dcache is 4-way 256-line 64-byte-per-line. +const CacheLinePadSize = 64 + +func doinit() {} diff --git a/src/internal/cpu/cpu_mips.go b/src/internal/cpu/cpu_mips.go new file mode 100644 index 0000000..14a9c97 --- /dev/null +++ b/src/internal/cpu/cpu_mips.go @@ -0,0 +1,10 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 32 + +func doinit() { +} diff --git a/src/internal/cpu/cpu_mips64x.go b/src/internal/cpu/cpu_mips64x.go new file mode 100644 index 0000000..c452ffd --- /dev/null +++ b/src/internal/cpu/cpu_mips64x.go @@ -0,0 +1,32 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips64 || mips64le + +package cpu + +const CacheLinePadSize = 32 + +// This is initialized by archauxv and should not be changed after it is +// initialized. +var HWCap uint + +// HWCAP bits. These are exposed by the Linux kernel 5.4. +const ( + // CPU features + hwcap_MIPS_MSA = 1 << 1 +) + +func doinit() { + options = []option{ + {Name: "msa", Feature: &MIPS64X.HasMSA}, + } + + // HWCAP feature bits + MIPS64X.HasMSA = isSet(HWCap, hwcap_MIPS_MSA) +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} diff --git a/src/internal/cpu/cpu_mipsle.go b/src/internal/cpu/cpu_mipsle.go new file mode 100644 index 0000000..14a9c97 --- /dev/null +++ b/src/internal/cpu/cpu_mipsle.go @@ -0,0 +1,10 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 32 + +func doinit() { +} diff --git a/src/internal/cpu/cpu_no_name.go b/src/internal/cpu/cpu_no_name.go new file mode 100644 index 0000000..2adfa1b --- /dev/null +++ b/src/internal/cpu/cpu_no_name.go @@ -0,0 +1,18 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !386 && !amd64 && !ppc64 && !ppc64le + +package cpu + +// Name returns the CPU name given by the vendor +// if it can be read directly from memory or by CPU instructions. +// If the CPU name can not be determined an empty string is returned. +// +// Implementations that use the Operating System (e.g. sysctl or /sys/) +// to gather CPU information for display should be placed in internal/sysinfo. +func Name() string { + // "A CPU has no name". + return "" +} diff --git a/src/internal/cpu/cpu_ppc64x.go b/src/internal/cpu/cpu_ppc64x.go new file mode 100644 index 0000000..c4a08fe --- /dev/null +++ b/src/internal/cpu/cpu_ppc64x.go @@ -0,0 +1,35 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package cpu + +const CacheLinePadSize = 128 + +func doinit() { + options = []option{ + {Name: "darn", Feature: &PPC64.HasDARN}, + {Name: "scv", Feature: &PPC64.HasSCV}, + {Name: "power9", Feature: &PPC64.IsPOWER9}, + } + + osinit() +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} + +func Name() string { + switch { + case PPC64.IsPOWER10: + return "POWER10" + case PPC64.IsPOWER9: + return "POWER9" + case PPC64.IsPOWER8: + return "POWER8" + } + return "" +} diff --git a/src/internal/cpu/cpu_ppc64x_aix.go b/src/internal/cpu/cpu_ppc64x_aix.go new file mode 100644 index 0000000..f05ed6f --- /dev/null +++ b/src/internal/cpu/cpu_ppc64x_aix.go @@ -0,0 +1,25 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package cpu + +const ( + // getsystemcfg constants + _SC_IMPL = 2 + _IMPL_POWER8 = 0x10000 + _IMPL_POWER9 = 0x20000 + _IMPL_POWER10 = 0x40000 +) + +func osinit() { + impl := getsystemcfg(_SC_IMPL) + PPC64.IsPOWER8 = isSet(impl, _IMPL_POWER8) + PPC64.IsPOWER9 = isSet(impl, _IMPL_POWER9) + PPC64.IsPOWER10 = isSet(impl, _IMPL_POWER10) +} + +// getsystemcfg is defined in runtime/os2_aix.go +func getsystemcfg(label uint) uint diff --git a/src/internal/cpu/cpu_ppc64x_linux.go b/src/internal/cpu/cpu_ppc64x_linux.go new file mode 100644 index 0000000..9df82ca --- /dev/null +++ b/src/internal/cpu/cpu_ppc64x_linux.go @@ -0,0 +1,33 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package cpu + +// ppc64 doesn't have a 'cpuid' equivalent, so we rely on HWCAP/HWCAP2. +// These are initialized by archauxv and should not be changed after they are +// initialized. +var HWCap uint +var HWCap2 uint + +// HWCAP bits. These are exposed by Linux. +const ( + // ISA Level + hwcap2_ARCH_2_07 = 0x80000000 + hwcap2_ARCH_3_00 = 0x00800000 + hwcap2_ARCH_3_1 = 0x00040000 + + // CPU features + hwcap2_DARN = 0x00200000 + hwcap2_SCV = 0x00100000 +) + +func osinit() { + PPC64.IsPOWER8 = isSet(HWCap2, hwcap2_ARCH_2_07) + PPC64.IsPOWER9 = isSet(HWCap2, hwcap2_ARCH_3_00) + PPC64.IsPOWER10 = isSet(HWCap2, hwcap2_ARCH_3_1) + PPC64.HasDARN = isSet(HWCap2, hwcap2_DARN) + PPC64.HasSCV = isSet(HWCap2, hwcap2_SCV) +} diff --git a/src/internal/cpu/cpu_ppc64x_other.go b/src/internal/cpu/cpu_ppc64x_other.go new file mode 100644 index 0000000..d5b629d --- /dev/null +++ b/src/internal/cpu/cpu_ppc64x_other.go @@ -0,0 +1,13 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build (ppc64 || ppc64le) && !aix && !linux + +package cpu + +func osinit() { + // Other operating systems do not support reading HWCap from auxiliary vector, + // reading privileged system registers or sysctl in user space to detect CPU + // features at runtime. +} diff --git a/src/internal/cpu/cpu_riscv64.go b/src/internal/cpu/cpu_riscv64.go new file mode 100644 index 0000000..54b8c33 --- /dev/null +++ b/src/internal/cpu/cpu_riscv64.go @@ -0,0 +1,10 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 32 + +func doinit() { +} diff --git a/src/internal/cpu/cpu_s390x.go b/src/internal/cpu/cpu_s390x.go new file mode 100644 index 0000000..45d8ed2 --- /dev/null +++ b/src/internal/cpu/cpu_s390x.go @@ -0,0 +1,205 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 256 + +var HWCap uint + +// bitIsSet reports whether the bit at index is set. The bit index +// is in big endian order, so bit index 0 is the leftmost bit. +func bitIsSet(bits []uint64, index uint) bool { + return bits[index/64]&((1<<63)>>(index%64)) != 0 +} + +// function is the function code for the named function. +type function uint8 + +const ( + // KM{,A,C,CTR} function codes + aes128 function = 18 // AES-128 + aes192 function = 19 // AES-192 + aes256 function = 20 // AES-256 + + // K{I,L}MD function codes + sha1 function = 1 // SHA-1 + sha256 function = 2 // SHA-256 + sha512 function = 3 // SHA-512 + sha3_224 function = 32 // SHA3-224 + sha3_256 function = 33 // SHA3-256 + sha3_384 function = 34 // SHA3-384 + sha3_512 function = 35 // SHA3-512 + shake128 function = 36 // SHAKE-128 + shake256 function = 37 // SHAKE-256 + + // KLMD function codes + ghash function = 65 // GHASH +) + +const ( + // KDSA function codes + ecdsaVerifyP256 function = 1 // NIST P256 + ecdsaVerifyP384 function = 2 // NIST P384 + ecdsaVerifyP521 function = 3 // NIST P521 + ecdsaSignP256 function = 9 // NIST P256 + ecdsaSignP384 function = 10 // NIST P384 + ecdsaSignP521 function = 11 // NIST P521 + eddsaVerifyEd25519 function = 32 // Curve25519 + eddsaVerifyEd448 function = 36 // Curve448 + eddsaSignEd25519 function = 40 // Curve25519 + eddsaSignEd448 function = 44 // Curve448 +) + +// queryResult contains the result of a Query function +// call. Bits are numbered in big endian order so the +// leftmost bit (the MSB) is at index 0. +type queryResult struct { + bits [2]uint64 +} + +// Has reports whether the given functions are present. +func (q *queryResult) Has(fns ...function) bool { + if len(fns) == 0 { + panic("no function codes provided") + } + for _, f := range fns { + if !bitIsSet(q.bits[:], uint(f)) { + return false + } + } + return true +} + +// facility is a bit index for the named facility. +type facility uint8 + +const ( + // mandatory facilities + zarch facility = 1 // z architecture mode is active + stflef facility = 7 // store-facility-list-extended + ldisp facility = 18 // long-displacement + eimm facility = 21 // extended-immediate + + // miscellaneous facilities + dfp facility = 42 // decimal-floating-point + etf3eh facility = 30 // extended-translation 3 enhancement + + // cryptography facilities + msa facility = 17 // message-security-assist + msa3 facility = 76 // message-security-assist extension 3 + msa4 facility = 77 // message-security-assist extension 4 + msa5 facility = 57 // message-security-assist extension 5 + msa8 facility = 146 // message-security-assist extension 8 + msa9 facility = 155 // message-security-assist extension 9 + + // vector facilities + vxe facility = 135 // vector-enhancements 1 + + // Note: vx requires kernel support + // and so must be fetched from HWCAP. + + hwcap_VX = 1 << 11 // vector facility +) + +// facilityList contains the result of an STFLE call. +// Bits are numbered in big endian order so the +// leftmost bit (the MSB) is at index 0. +type facilityList struct { + bits [4]uint64 +} + +// Has reports whether the given facilities are present. +func (s *facilityList) Has(fs ...facility) bool { + if len(fs) == 0 { + panic("no facility bits provided") + } + for _, f := range fs { + if !bitIsSet(s.bits[:], uint(f)) { + return false + } + } + return true +} + +// The following feature detection functions are defined in cpu_s390x.s. +// They are likely to be expensive to call so the results should be cached. +func stfle() facilityList +func kmQuery() queryResult +func kmcQuery() queryResult +func kmctrQuery() queryResult +func kmaQuery() queryResult +func kimdQuery() queryResult +func klmdQuery() queryResult +func kdsaQuery() queryResult + +func doinit() { + options = []option{ + {Name: "zarch", Feature: &S390X.HasZARCH}, + {Name: "stfle", Feature: &S390X.HasSTFLE}, + {Name: "ldisp", Feature: &S390X.HasLDISP}, + {Name: "msa", Feature: &S390X.HasMSA}, + {Name: "eimm", Feature: &S390X.HasEIMM}, + {Name: "dfp", Feature: &S390X.HasDFP}, + {Name: "etf3eh", Feature: &S390X.HasETF3EH}, + {Name: "vx", Feature: &S390X.HasVX}, + {Name: "vxe", Feature: &S390X.HasVXE}, + {Name: "kdsa", Feature: &S390X.HasKDSA}, + } + + aes := []function{aes128, aes192, aes256} + facilities := stfle() + + S390X.HasZARCH = facilities.Has(zarch) + S390X.HasSTFLE = facilities.Has(stflef) + S390X.HasLDISP = facilities.Has(ldisp) + S390X.HasEIMM = facilities.Has(eimm) + S390X.HasDFP = facilities.Has(dfp) + S390X.HasETF3EH = facilities.Has(etf3eh) + S390X.HasMSA = facilities.Has(msa) + + if S390X.HasMSA { + // cipher message + km, kmc := kmQuery(), kmcQuery() + S390X.HasAES = km.Has(aes...) + S390X.HasAESCBC = kmc.Has(aes...) + if facilities.Has(msa4) { + kmctr := kmctrQuery() + S390X.HasAESCTR = kmctr.Has(aes...) + } + if facilities.Has(msa8) { + kma := kmaQuery() + S390X.HasAESGCM = kma.Has(aes...) + } + + // compute message digest + kimd := kimdQuery() // intermediate (no padding) + klmd := klmdQuery() // last (padding) + S390X.HasSHA1 = kimd.Has(sha1) && klmd.Has(sha1) + S390X.HasSHA256 = kimd.Has(sha256) && klmd.Has(sha256) + S390X.HasSHA512 = kimd.Has(sha512) && klmd.Has(sha512) + S390X.HasGHASH = kimd.Has(ghash) // KLMD-GHASH does not exist + sha3 := []function{ + sha3_224, sha3_256, sha3_384, sha3_512, + shake128, shake256, + } + S390X.HasSHA3 = kimd.Has(sha3...) && klmd.Has(sha3...) + S390X.HasKDSA = facilities.Has(msa9) // elliptic curves + if S390X.HasKDSA { + kdsa := kdsaQuery() + S390X.HasECDSA = kdsa.Has(ecdsaVerifyP256, ecdsaSignP256, ecdsaVerifyP384, ecdsaSignP384, ecdsaVerifyP521, ecdsaSignP521) + S390X.HasEDDSA = kdsa.Has(eddsaVerifyEd25519, eddsaSignEd25519, eddsaVerifyEd448, eddsaSignEd448) + } + } + + S390X.HasVX = isSet(HWCap, hwcap_VX) + + if S390X.HasVX { + S390X.HasVXE = facilities.Has(vxe) + } +} + +func isSet(hwc uint, value uint) bool { + return hwc&value != 0 +} diff --git a/src/internal/cpu/cpu_s390x.s b/src/internal/cpu/cpu_s390x.s new file mode 100644 index 0000000..a1243aa --- /dev/null +++ b/src/internal/cpu/cpu_s390x.s @@ -0,0 +1,63 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +// func stfle() facilityList +TEXT ·stfle(SB), NOSPLIT|NOFRAME, $0-32 + MOVD $ret+0(FP), R1 + MOVD $3, R0 // last doubleword index to store + XC $32, (R1), (R1) // clear 4 doublewords (32 bytes) + WORD $0xb2b01000 // store facility list extended (STFLE) + RET + +// func kmQuery() queryResult +TEXT ·kmQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KM-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB92E0024 // cipher message (KM) + RET + +// func kmcQuery() queryResult +TEXT ·kmcQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KMC-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB92F0024 // cipher message with chaining (KMC) + RET + +// func kmctrQuery() queryResult +TEXT ·kmctrQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KMCTR-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB92D4024 // cipher message with counter (KMCTR) + RET + +// func kmaQuery() queryResult +TEXT ·kmaQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KMA-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xb9296024 // cipher message with authentication (KMA) + RET + +// func kimdQuery() queryResult +TEXT ·kimdQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KIMD-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB93E0024 // compute intermediate message digest (KIMD) + RET + +// func klmdQuery() queryResult +TEXT ·klmdQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KLMD-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB93F0024 // compute last message digest (KLMD) + RET + +// func kdsaQuery() queryResult +TEXT ·kdsaQuery(SB), NOSPLIT|NOFRAME, $0-16 + MOVD $0, R0 // set function code to 0 (KLMD-Query) + MOVD $ret+0(FP), R1 // address of 16-byte return value + WORD $0xB93A0008 // compute digital signature authentication + RET + diff --git a/src/internal/cpu/cpu_s390x_test.go b/src/internal/cpu/cpu_s390x_test.go new file mode 100644 index 0000000..ad86858 --- /dev/null +++ b/src/internal/cpu/cpu_s390x_test.go @@ -0,0 +1,63 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu_test + +import ( + "errors" + . "internal/cpu" + "os" + "regexp" + "testing" +) + +func getFeatureList() ([]string, error) { + cpuinfo, err := os.ReadFile("/proc/cpuinfo") + if err != nil { + return nil, err + } + r := regexp.MustCompile("features\\s*:\\s*(.*)") + b := r.FindSubmatch(cpuinfo) + if len(b) < 2 { + return nil, errors.New("no feature list in /proc/cpuinfo") + } + return regexp.MustCompile("\\s+").Split(string(b[1]), -1), nil +} + +func TestS390XAgainstCPUInfo(t *testing.T) { + // mapping of linux feature strings to S390X fields + mapping := make(map[string]*bool) + for _, option := range Options { + mapping[option.Name] = option.Feature + } + + // these must be true on the machines Go supports + mandatory := make(map[string]bool) + mandatory["zarch"] = false + mandatory["eimm"] = false + mandatory["ldisp"] = false + mandatory["stfle"] = false + + features, err := getFeatureList() + if err != nil { + t.Error(err) + } + for _, feature := range features { + if _, ok := mandatory[feature]; ok { + mandatory[feature] = true + } + if flag, ok := mapping[feature]; ok { + if !*flag { + t.Errorf("feature '%v' not detected", feature) + } + } else { + t.Logf("no entry for '%v'", feature) + } + } + for k, v := range mandatory { + if !v { + t.Errorf("mandatory feature '%v' not detected", k) + } + } +} diff --git a/src/internal/cpu/cpu_test.go b/src/internal/cpu/cpu_test.go new file mode 100644 index 0000000..b8c74f2 --- /dev/null +++ b/src/internal/cpu/cpu_test.go @@ -0,0 +1,61 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu_test + +import ( + . "internal/cpu" + "internal/godebug" + "internal/testenv" + "os" + "os/exec" + "testing" +) + +func MustHaveDebugOptionsSupport(t *testing.T) { + if !DebugOptions { + t.Skipf("skipping test: cpu feature options not supported by OS") + } +} + +func MustSupportFeatureDetection(t *testing.T) { + // TODO: add platforms that do not have CPU feature detection support. +} + +func runDebugOptionsTest(t *testing.T, test string, options string) { + MustHaveDebugOptionsSupport(t) + + testenv.MustHaveExec(t) + + env := "GODEBUG=" + options + + cmd := exec.Command(os.Args[0], "-test.run="+test) + cmd.Env = append(cmd.Env, env) + + output, err := cmd.CombinedOutput() + if err != nil { + t.Fatalf("%s with %s: run failed: %v output:\n%s\n", + test, env, err, string(output)) + } +} + +func TestDisableAllCapabilities(t *testing.T) { + MustSupportFeatureDetection(t) + runDebugOptionsTest(t, "TestAllCapabilitiesDisabled", "cpu.all=off") +} + +func TestAllCapabilitiesDisabled(t *testing.T) { + MustHaveDebugOptionsSupport(t) + + if godebug.New("#cpu.all").Value() != "off" { + t.Skipf("skipping test: GODEBUG=cpu.all=off not set") + } + + for _, o := range Options { + want := false + if got := *o.Feature; got != want { + t.Errorf("%v: expected %v, got %v", o.Name, want, got) + } + } +} diff --git a/src/internal/cpu/cpu_wasm.go b/src/internal/cpu/cpu_wasm.go new file mode 100644 index 0000000..2310ad6 --- /dev/null +++ b/src/internal/cpu/cpu_wasm.go @@ -0,0 +1,10 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +const CacheLinePadSize = 64 + +func doinit() { +} diff --git a/src/internal/cpu/cpu_x86.go b/src/internal/cpu/cpu_x86.go new file mode 100644 index 0000000..96b8ef9 --- /dev/null +++ b/src/internal/cpu/cpu_x86.go @@ -0,0 +1,190 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 + +package cpu + +const CacheLinePadSize = 64 + +// cpuid is implemented in cpu_x86.s. +func cpuid(eaxArg, ecxArg uint32) (eax, ebx, ecx, edx uint32) + +// xgetbv with ecx = 0 is implemented in cpu_x86.s. +func xgetbv() (eax, edx uint32) + +// getGOAMD64level is implemented in cpu_x86.s. Returns number in [1,4]. +func getGOAMD64level() int32 + +const ( + // edx bits + cpuid_SSE2 = 1 << 26 + + // ecx bits + cpuid_SSE3 = 1 << 0 + cpuid_PCLMULQDQ = 1 << 1 + cpuid_SSSE3 = 1 << 9 + cpuid_FMA = 1 << 12 + cpuid_SSE41 = 1 << 19 + cpuid_SSE42 = 1 << 20 + cpuid_POPCNT = 1 << 23 + cpuid_AES = 1 << 25 + cpuid_OSXSAVE = 1 << 27 + cpuid_AVX = 1 << 28 + + // ebx bits + cpuid_BMI1 = 1 << 3 + cpuid_AVX2 = 1 << 5 + cpuid_BMI2 = 1 << 8 + cpuid_ERMS = 1 << 9 + cpuid_ADX = 1 << 19 + cpuid_SHA = 1 << 29 + + // edx bits for CPUID 0x80000001 + cpuid_RDTSCP = 1 << 27 +) + +var maxExtendedFunctionInformation uint32 + +func doinit() { + options = []option{ + {Name: "adx", Feature: &X86.HasADX}, + {Name: "aes", Feature: &X86.HasAES}, + {Name: "erms", Feature: &X86.HasERMS}, + {Name: "pclmulqdq", Feature: &X86.HasPCLMULQDQ}, + {Name: "rdtscp", Feature: &X86.HasRDTSCP}, + {Name: "sha", Feature: &X86.HasSHA}, + } + level := getGOAMD64level() + if level < 2 { + // These options are required at level 2. At lower levels + // they can be turned off. + options = append(options, + option{Name: "popcnt", Feature: &X86.HasPOPCNT}, + option{Name: "sse3", Feature: &X86.HasSSE3}, + option{Name: "sse41", Feature: &X86.HasSSE41}, + option{Name: "sse42", Feature: &X86.HasSSE42}, + option{Name: "ssse3", Feature: &X86.HasSSSE3}) + } + if level < 3 { + // These options are required at level 3. At lower levels + // they can be turned off. + options = append(options, + option{Name: "avx", Feature: &X86.HasAVX}, + option{Name: "avx2", Feature: &X86.HasAVX2}, + option{Name: "bmi1", Feature: &X86.HasBMI1}, + option{Name: "bmi2", Feature: &X86.HasBMI2}, + option{Name: "fma", Feature: &X86.HasFMA}) + } + + maxID, _, _, _ := cpuid(0, 0) + + if maxID < 1 { + return + } + + maxExtendedFunctionInformation, _, _, _ = cpuid(0x80000000, 0) + + _, _, ecx1, _ := cpuid(1, 0) + + X86.HasSSE3 = isSet(ecx1, cpuid_SSE3) + X86.HasPCLMULQDQ = isSet(ecx1, cpuid_PCLMULQDQ) + X86.HasSSSE3 = isSet(ecx1, cpuid_SSSE3) + X86.HasSSE41 = isSet(ecx1, cpuid_SSE41) + X86.HasSSE42 = isSet(ecx1, cpuid_SSE42) + X86.HasPOPCNT = isSet(ecx1, cpuid_POPCNT) + X86.HasAES = isSet(ecx1, cpuid_AES) + + // OSXSAVE can be false when using older Operating Systems + // or when explicitly disabled on newer Operating Systems by + // e.g. setting the xsavedisable boot option on Windows 10. + X86.HasOSXSAVE = isSet(ecx1, cpuid_OSXSAVE) + + // The FMA instruction set extension only has VEX prefixed instructions. + // VEX prefixed instructions require OSXSAVE to be enabled. + // See Intel 64 and IA-32 Architecture Software Developer’s Manual Volume 2 + // Section 2.4 "AVX and SSE Instruction Exception Specification" + X86.HasFMA = isSet(ecx1, cpuid_FMA) && X86.HasOSXSAVE + + osSupportsAVX := false + // For XGETBV, OSXSAVE bit is required and sufficient. + if X86.HasOSXSAVE { + eax, _ := xgetbv() + // Check if XMM and YMM registers have OS support. + osSupportsAVX = isSet(eax, 1<<1) && isSet(eax, 1<<2) + } + + X86.HasAVX = isSet(ecx1, cpuid_AVX) && osSupportsAVX + + if maxID < 7 { + return + } + + _, ebx7, _, _ := cpuid(7, 0) + X86.HasBMI1 = isSet(ebx7, cpuid_BMI1) + X86.HasAVX2 = isSet(ebx7, cpuid_AVX2) && osSupportsAVX + X86.HasBMI2 = isSet(ebx7, cpuid_BMI2) + X86.HasERMS = isSet(ebx7, cpuid_ERMS) + X86.HasADX = isSet(ebx7, cpuid_ADX) + X86.HasSHA = isSet(ebx7, cpuid_SHA) + + var maxExtendedInformation uint32 + maxExtendedInformation, _, _, _ = cpuid(0x80000000, 0) + + if maxExtendedInformation < 0x80000001 { + return + } + + _, _, _, edxExt1 := cpuid(0x80000001, 0) + X86.HasRDTSCP = isSet(edxExt1, cpuid_RDTSCP) +} + +func isSet(hwc uint32, value uint32) bool { + return hwc&value != 0 +} + +// Name returns the CPU name given by the vendor. +// If the CPU name can not be determined an +// empty string is returned. +func Name() string { + if maxExtendedFunctionInformation < 0x80000004 { + return "" + } + + data := make([]byte, 0, 3*4*4) + + var eax, ebx, ecx, edx uint32 + eax, ebx, ecx, edx = cpuid(0x80000002, 0) + data = appendBytes(data, eax, ebx, ecx, edx) + eax, ebx, ecx, edx = cpuid(0x80000003, 0) + data = appendBytes(data, eax, ebx, ecx, edx) + eax, ebx, ecx, edx = cpuid(0x80000004, 0) + data = appendBytes(data, eax, ebx, ecx, edx) + + // Trim leading spaces. + for len(data) > 0 && data[0] == ' ' { + data = data[1:] + } + + // Trim tail after and including the first null byte. + for i, c := range data { + if c == '\x00' { + data = data[:i] + break + } + } + + return string(data) +} + +func appendBytes(b []byte, args ...uint32) []byte { + for _, arg := range args { + b = append(b, + byte((arg >> 0)), + byte((arg >> 8)), + byte((arg >> 16)), + byte((arg >> 24))) + } + return b +} diff --git a/src/internal/cpu/cpu_x86.s b/src/internal/cpu/cpu_x86.s new file mode 100644 index 0000000..2ee8eca --- /dev/null +++ b/src/internal/cpu/cpu_x86.s @@ -0,0 +1,43 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 + +#include "textflag.h" + +// func cpuid(eaxArg, ecxArg uint32) (eax, ebx, ecx, edx uint32) +TEXT ·cpuid(SB), NOSPLIT, $0-24 + MOVL eaxArg+0(FP), AX + MOVL ecxArg+4(FP), CX + CPUID + MOVL AX, eax+8(FP) + MOVL BX, ebx+12(FP) + MOVL CX, ecx+16(FP) + MOVL DX, edx+20(FP) + RET + +// func xgetbv() (eax, edx uint32) +TEXT ·xgetbv(SB),NOSPLIT,$0-8 + MOVL $0, CX + XGETBV + MOVL AX, eax+0(FP) + MOVL DX, edx+4(FP) + RET + +// func getGOAMD64level() int32 +TEXT ·getGOAMD64level(SB),NOSPLIT,$0-4 +#ifdef GOAMD64_v4 + MOVL $4, ret+0(FP) +#else +#ifdef GOAMD64_v3 + MOVL $3, ret+0(FP) +#else +#ifdef GOAMD64_v2 + MOVL $2, ret+0(FP) +#else + MOVL $1, ret+0(FP) +#endif +#endif +#endif + RET diff --git a/src/internal/cpu/cpu_x86_test.go b/src/internal/cpu/cpu_x86_test.go new file mode 100644 index 0000000..8564ccc --- /dev/null +++ b/src/internal/cpu/cpu_x86_test.go @@ -0,0 +1,39 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 + +package cpu_test + +import ( + . "internal/cpu" + "internal/godebug" + "testing" +) + +func TestX86ifAVX2hasAVX(t *testing.T) { + if X86.HasAVX2 && !X86.HasAVX { + t.Fatalf("HasAVX expected true when HasAVX2 is true, got false") + } +} + +func TestDisableSSE3(t *testing.T) { + if GetGOAMD64level() > 1 { + t.Skip("skipping test: can't run on GOAMD64>v1 machines") + } + runDebugOptionsTest(t, "TestSSE3DebugOption", "cpu.sse3=off") +} + +func TestSSE3DebugOption(t *testing.T) { + MustHaveDebugOptionsSupport(t) + + if godebug.New("#cpu.sse3").Value() != "off" { + t.Skipf("skipping test: GODEBUG=cpu.sse3=off not set") + } + + want := false + if got := X86.HasSSE3; got != want { + t.Errorf("X86.HasSSE3 expected %v, got %v", want, got) + } +} diff --git a/src/internal/cpu/export_test.go b/src/internal/cpu/export_test.go new file mode 100644 index 0000000..91bfc1b --- /dev/null +++ b/src/internal/cpu/export_test.go @@ -0,0 +1,9 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cpu + +var ( + Options = options +) diff --git a/src/internal/cpu/export_x86_test.go b/src/internal/cpu/export_x86_test.go new file mode 100644 index 0000000..a12b6f2 --- /dev/null +++ b/src/internal/cpu/export_x86_test.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build 386 || amd64 + +package cpu + +var ( + GetGOAMD64level = getGOAMD64level +) diff --git a/src/internal/dag/alg.go b/src/internal/dag/alg.go new file mode 100644 index 0000000..8800279 --- /dev/null +++ b/src/internal/dag/alg.go @@ -0,0 +1,63 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package dag + +// Transpose reverses all edges in g. +func (g *Graph) Transpose() { + old := g.edges + + g.edges = make(map[string]map[string]bool) + for _, n := range g.Nodes { + g.edges[n] = make(map[string]bool) + } + + for from, tos := range old { + for to := range tos { + g.edges[to][from] = true + } + } +} + +// Topo returns a topological sort of g. This function is deterministic. +func (g *Graph) Topo() []string { + topo := make([]string, 0, len(g.Nodes)) + marks := make(map[string]bool) + + var visit func(n string) + visit = func(n string) { + if marks[n] { + return + } + for _, to := range g.Edges(n) { + visit(to) + } + marks[n] = true + topo = append(topo, n) + } + for _, root := range g.Nodes { + visit(root) + } + for i, j := 0, len(topo)-1; i < j; i, j = i+1, j-1 { + topo[i], topo[j] = topo[j], topo[i] + } + return topo +} + +// TransitiveReduction removes edges from g that are transitively +// reachable. g must be transitively closed. +func (g *Graph) TransitiveReduction() { + // For i -> j -> k, if i -> k exists, delete it. + for _, i := range g.Nodes { + for _, j := range g.Nodes { + if g.HasEdge(i, j) { + for _, k := range g.Nodes { + if g.HasEdge(j, k) { + g.DelEdge(i, k) + } + } + } + } + } +} diff --git a/src/internal/dag/alg_test.go b/src/internal/dag/alg_test.go new file mode 100644 index 0000000..e5ea8b6 --- /dev/null +++ b/src/internal/dag/alg_test.go @@ -0,0 +1,46 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package dag + +import ( + "reflect" + "strings" + "testing" +) + +func TestTranspose(t *testing.T) { + g := mustParse(t, diamond) + g.Transpose() + wantEdges(t, g, "a->b a->c a->d b->d c->d") +} + +func TestTopo(t *testing.T) { + g := mustParse(t, diamond) + got := g.Topo() + // "d" is the root, so it's first. + // + // "c" and "b" could be in either order, but Topo is + // deterministic in reverse node definition order. + // + // "a" is a leaf. + wantNodes := strings.Fields("d c b a") + if !reflect.DeepEqual(wantNodes, got) { + t.Fatalf("want topo sort %v, got %v", wantNodes, got) + } +} + +func TestTransitiveReduction(t *testing.T) { + t.Run("diamond", func(t *testing.T) { + g := mustParse(t, diamond) + g.TransitiveReduction() + wantEdges(t, g, "b->a c->a d->b d->c") + }) + t.Run("chain", func(t *testing.T) { + const chain = `NONE < a < b < c < d; a, d < e;` + g := mustParse(t, chain) + g.TransitiveReduction() + wantEdges(t, g, "e->d d->c c->b b->a") + }) +} diff --git a/src/internal/dag/parse.go b/src/internal/dag/parse.go new file mode 100644 index 0000000..9d5b918 --- /dev/null +++ b/src/internal/dag/parse.go @@ -0,0 +1,314 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package dag implements a language for expressing directed acyclic +// graphs. +// +// The general syntax of a rule is: +// +// a, b < c, d; +// +// which means c and d come after a and b in the partial order +// (that is, there are edges from c and d to a and b), +// but doesn't provide a relative order between a vs b or c vs d. +// +// The rules can chain together, as in: +// +// e < f, g < h; +// +// which is equivalent to +// +// e < f, g; +// f, g < h; +// +// Except for the special bottom element "NONE", each name +// must appear exactly once on the right-hand side of any rule. +// That rule serves as the definition of the allowed successor +// for that name. The definition must appear before any uses +// of the name on the left-hand side of a rule. (That is, the +// rules themselves must be ordered according to the partial +// order, for easier reading by people.) +// +// Negative assertions double-check the partial order: +// +// i !< j +// +// means that it must NOT be the case that i < j. +// Negative assertions may appear anywhere in the rules, +// even before i and j have been defined. +// +// Comments begin with #. +package dag + +import ( + "fmt" + "sort" + "strings" +) + +type Graph struct { + Nodes []string + byLabel map[string]int + edges map[string]map[string]bool +} + +func newGraph() *Graph { + return &Graph{byLabel: map[string]int{}, edges: map[string]map[string]bool{}} +} + +func (g *Graph) addNode(label string) bool { + if _, ok := g.byLabel[label]; ok { + return false + } + g.byLabel[label] = len(g.Nodes) + g.Nodes = append(g.Nodes, label) + g.edges[label] = map[string]bool{} + return true +} + +func (g *Graph) AddEdge(from, to string) { + g.edges[from][to] = true +} + +func (g *Graph) DelEdge(from, to string) { + delete(g.edges[from], to) +} + +func (g *Graph) HasEdge(from, to string) bool { + return g.edges[from] != nil && g.edges[from][to] +} + +func (g *Graph) Edges(from string) []string { + edges := make([]string, 0, 16) + for k := range g.edges[from] { + edges = append(edges, k) + } + sort.Slice(edges, func(i, j int) bool { return g.byLabel[edges[i]] < g.byLabel[edges[j]] }) + return edges +} + +// Parse parses the DAG language and returns the transitive closure of +// the described graph. In the returned graph, there is an edge from "b" +// to "a" if b < a (or a > b) in the partial order. +func Parse(dag string) (*Graph, error) { + g := newGraph() + disallowed := []rule{} + + rules, err := parseRules(dag) + if err != nil { + return nil, err + } + + // TODO: Add line numbers to errors. + var errors []string + errorf := func(format string, a ...any) { + errors = append(errors, fmt.Sprintf(format, a...)) + } + for _, r := range rules { + if r.op == "!<" { + disallowed = append(disallowed, r) + continue + } + for _, def := range r.def { + if def == "NONE" { + errorf("NONE cannot be a predecessor") + continue + } + if !g.addNode(def) { + errorf("multiple definitions for %s", def) + } + for _, less := range r.less { + if less == "NONE" { + continue + } + if _, ok := g.byLabel[less]; !ok { + errorf("use of %s before its definition", less) + } else { + g.AddEdge(def, less) + } + } + } + } + + // Check for missing definition. + for _, tos := range g.edges { + for to := range tos { + if g.edges[to] == nil { + errorf("missing definition for %s", to) + } + } + } + + // Complete transitive closure. + for _, k := range g.Nodes { + for _, i := range g.Nodes { + for _, j := range g.Nodes { + if i != k && k != j && g.HasEdge(i, k) && g.HasEdge(k, j) { + if i == j { + // Can only happen along with a "use of X before deps" error above, + // but this error is more specific - it makes clear that reordering the + // rules will not be enough to fix the problem. + errorf("graph cycle: %s < %s < %s", j, k, i) + } + g.AddEdge(i, j) + } + } + } + } + + // Check negative assertions against completed allowed graph. + for _, bad := range disallowed { + for _, less := range bad.less { + for _, def := range bad.def { + if g.HasEdge(def, less) { + errorf("graph edge assertion failed: %s !< %s", less, def) + } + } + } + } + + if len(errors) > 0 { + return nil, fmt.Errorf("%s", strings.Join(errors, "\n")) + } + + return g, nil +} + +// A rule is a line in the DAG language where "less < def" or "less !< def". +type rule struct { + less []string + op string // Either "<" or "!<" + def []string +} + +type syntaxError string + +func (e syntaxError) Error() string { + return string(e) +} + +// parseRules parses the rules of a DAG. +func parseRules(rules string) (out []rule, err error) { + defer func() { + e := recover() + switch e := e.(type) { + case nil: + return + case syntaxError: + err = e + default: + panic(e) + } + }() + p := &rulesParser{lineno: 1, text: rules} + + var prev []string + var op string + for { + list, tok := p.nextList() + if tok == "" { + if prev == nil { + break + } + p.syntaxError("unexpected EOF") + } + if prev != nil { + out = append(out, rule{prev, op, list}) + } + prev = list + if tok == ";" { + prev = nil + op = "" + continue + } + if tok != "<" && tok != "!<" { + p.syntaxError("missing <") + } + op = tok + } + + return out, err +} + +// A rulesParser parses the depsRules syntax described above. +type rulesParser struct { + lineno int + lastWord string + text string +} + +// syntaxError reports a parsing error. +func (p *rulesParser) syntaxError(msg string) { + panic(syntaxError(fmt.Sprintf("parsing graph: line %d: syntax error: %s near %s", p.lineno, msg, p.lastWord))) +} + +// nextList parses and returns a comma-separated list of names. +func (p *rulesParser) nextList() (list []string, token string) { + for { + tok := p.nextToken() + switch tok { + case "": + if len(list) == 0 { + return nil, "" + } + fallthrough + case ",", "<", "!<", ";": + p.syntaxError("bad list syntax") + } + list = append(list, tok) + + tok = p.nextToken() + if tok != "," { + return list, tok + } + } +} + +// nextToken returns the next token in the deps rules, +// one of ";" "," "<" "!<" or a name. +func (p *rulesParser) nextToken() string { + for { + if p.text == "" { + return "" + } + switch p.text[0] { + case ';', ',', '<': + t := p.text[:1] + p.text = p.text[1:] + return t + + case '!': + if len(p.text) < 2 || p.text[1] != '<' { + p.syntaxError("unexpected token !") + } + p.text = p.text[2:] + return "!<" + + case '#': + i := strings.Index(p.text, "\n") + if i < 0 { + i = len(p.text) + } + p.text = p.text[i:] + continue + + case '\n': + p.lineno++ + fallthrough + case ' ', '\t': + p.text = p.text[1:] + continue + + default: + i := strings.IndexAny(p.text, "!;,<#\n \t") + if i < 0 { + i = len(p.text) + } + t := p.text[:i] + p.text = p.text[i:] + p.lastWord = t + return t + } + } +} diff --git a/src/internal/dag/parse_test.go b/src/internal/dag/parse_test.go new file mode 100644 index 0000000..b2520c3 --- /dev/null +++ b/src/internal/dag/parse_test.go @@ -0,0 +1,61 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package dag + +import ( + "reflect" + "strings" + "testing" +) + +const diamond = ` +NONE < a < b, c < d; +` + +func mustParse(t *testing.T, dag string) *Graph { + t.Helper() + g, err := Parse(dag) + if err != nil { + t.Fatal(err) + } + return g +} + +func wantEdges(t *testing.T, g *Graph, edges string) { + t.Helper() + + wantEdges := strings.Fields(edges) + wantEdgeMap := make(map[string]bool) + for _, e := range wantEdges { + wantEdgeMap[e] = true + } + + for _, n1 := range g.Nodes { + for _, n2 := range g.Nodes { + got := g.HasEdge(n1, n2) + want := wantEdgeMap[n1+"->"+n2] + if got && want { + t.Logf("%s->%s", n1, n2) + } else if got && !want { + t.Errorf("%s->%s present but not expected", n1, n2) + } else if want && !got { + t.Errorf("%s->%s missing but expected", n1, n2) + } + } + } +} + +func TestParse(t *testing.T) { + // Basic smoke test for graph parsing. + g := mustParse(t, diamond) + + wantNodes := strings.Fields("a b c d") + if !reflect.DeepEqual(wantNodes, g.Nodes) { + t.Fatalf("want nodes %v, got %v", wantNodes, g.Nodes) + } + + // Parse returns the transitive closure, so it adds d->a. + wantEdges(t, g, "b->a c->a d->a d->b d->c") +} diff --git a/src/internal/diff/diff.go b/src/internal/diff/diff.go new file mode 100644 index 0000000..0aeeb75 --- /dev/null +++ b/src/internal/diff/diff.go @@ -0,0 +1,261 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package diff + +import ( + "bytes" + "fmt" + "sort" + "strings" +) + +// A pair is a pair of values tracked for both the x and y side of a diff. +// It is typically a pair of line indexes. +type pair struct{ x, y int } + +// Diff returns an anchored diff of the two texts old and new +// in the “unified diff” format. If old and new are identical, +// Diff returns a nil slice (no output). +// +// Unix diff implementations typically look for a diff with +// the smallest number of lines inserted and removed, +// which can in the worst case take time quadratic in the +// number of lines in the texts. As a result, many implementations +// either can be made to run for a long time or cut off the search +// after a predetermined amount of work. +// +// In contrast, this implementation looks for a diff with the +// smallest number of “unique” lines inserted and removed, +// where unique means a line that appears just once in both old and new. +// We call this an “anchored diff” because the unique lines anchor +// the chosen matching regions. An anchored diff is usually clearer +// than a standard diff, because the algorithm does not try to +// reuse unrelated blank lines or closing braces. +// The algorithm also guarantees to run in O(n log n) time +// instead of the standard O(n²) time. +// +// Some systems call this approach a “patience diff,” named for +// the “patience sorting” algorithm, itself named for a solitaire card game. +// We avoid that name for two reasons. First, the name has been used +// for a few different variants of the algorithm, so it is imprecise. +// Second, the name is frequently interpreted as meaning that you have +// to wait longer (to be patient) for the diff, meaning that it is a slower algorithm, +// when in fact the algorithm is faster than the standard one. +func Diff(oldName string, old []byte, newName string, new []byte) []byte { + if bytes.Equal(old, new) { + return nil + } + x := lines(old) + y := lines(new) + + // Print diff header. + var out bytes.Buffer + fmt.Fprintf(&out, "diff %s %s\n", oldName, newName) + fmt.Fprintf(&out, "--- %s\n", oldName) + fmt.Fprintf(&out, "+++ %s\n", newName) + + // Loop over matches to consider, + // expanding each match to include surrounding lines, + // and then printing diff chunks. + // To avoid setup/teardown cases outside the loop, + // tgs returns a leading {0,0} and trailing {len(x), len(y)} pair + // in the sequence of matches. + var ( + done pair // printed up to x[:done.x] and y[:done.y] + chunk pair // start lines of current chunk + count pair // number of lines from each side in current chunk + ctext []string // lines for current chunk + ) + for _, m := range tgs(x, y) { + if m.x < done.x { + // Already handled scanning forward from earlier match. + continue + } + + // Expand matching lines as far possible, + // establishing that x[start.x:end.x] == y[start.y:end.y]. + // Note that on the first (or last) iteration we may (or definitely do) + // have an empty match: start.x==end.x and start.y==end.y. + start := m + for start.x > done.x && start.y > done.y && x[start.x-1] == y[start.y-1] { + start.x-- + start.y-- + } + end := m + for end.x < len(x) && end.y < len(y) && x[end.x] == y[end.y] { + end.x++ + end.y++ + } + + // Emit the mismatched lines before start into this chunk. + // (No effect on first sentinel iteration, when start = {0,0}.) + for _, s := range x[done.x:start.x] { + ctext = append(ctext, "-"+s) + count.x++ + } + for _, s := range y[done.y:start.y] { + ctext = append(ctext, "+"+s) + count.y++ + } + + // If we're not at EOF and have too few common lines, + // the chunk includes all the common lines and continues. + const C = 3 // number of context lines + if (end.x < len(x) || end.y < len(y)) && + (end.x-start.x < C || (len(ctext) > 0 && end.x-start.x < 2*C)) { + for _, s := range x[start.x:end.x] { + ctext = append(ctext, " "+s) + count.x++ + count.y++ + } + done = end + continue + } + + // End chunk with common lines for context. + if len(ctext) > 0 { + n := end.x - start.x + if n > C { + n = C + } + for _, s := range x[start.x : start.x+n] { + ctext = append(ctext, " "+s) + count.x++ + count.y++ + } + done = pair{start.x + n, start.y + n} + + // Format and emit chunk. + // Convert line numbers to 1-indexed. + // Special case: empty file shows up as 0,0 not 1,0. + if count.x > 0 { + chunk.x++ + } + if count.y > 0 { + chunk.y++ + } + fmt.Fprintf(&out, "@@ -%d,%d +%d,%d @@\n", chunk.x, count.x, chunk.y, count.y) + for _, s := range ctext { + out.WriteString(s) + } + count.x = 0 + count.y = 0 + ctext = ctext[:0] + } + + // If we reached EOF, we're done. + if end.x >= len(x) && end.y >= len(y) { + break + } + + // Otherwise start a new chunk. + chunk = pair{end.x - C, end.y - C} + for _, s := range x[chunk.x:end.x] { + ctext = append(ctext, " "+s) + count.x++ + count.y++ + } + done = end + } + + return out.Bytes() +} + +// lines returns the lines in the file x, including newlines. +// If the file does not end in a newline, one is supplied +// along with a warning about the missing newline. +func lines(x []byte) []string { + l := strings.SplitAfter(string(x), "\n") + if l[len(l)-1] == "" { + l = l[:len(l)-1] + } else { + // Treat last line as having a message about the missing newline attached, + // using the same text as BSD/GNU diff (including the leading backslash). + l[len(l)-1] += "\n\\ No newline at end of file\n" + } + return l +} + +// tgs returns the pairs of indexes of the longest common subsequence +// of unique lines in x and y, where a unique line is one that appears +// once in x and once in y. +// +// The longest common subsequence algorithm is as described in +// Thomas G. Szymanski, “A Special Case of the Maximal Common +// Subsequence Problem,” Princeton TR #170 (January 1975), +// available at https://research.swtch.com/tgs170.pdf. +func tgs(x, y []string) []pair { + // Count the number of times each string appears in a and b. + // We only care about 0, 1, many, counted as 0, -1, -2 + // for the x side and 0, -4, -8 for the y side. + // Using negative numbers now lets us distinguish positive line numbers later. + m := make(map[string]int) + for _, s := range x { + if c := m[s]; c > -2 { + m[s] = c - 1 + } + } + for _, s := range y { + if c := m[s]; c > -8 { + m[s] = c - 4 + } + } + + // Now unique strings can be identified by m[s] = -1+-4. + // + // Gather the indexes of those strings in x and y, building: + // xi[i] = increasing indexes of unique strings in x. + // yi[i] = increasing indexes of unique strings in y. + // inv[i] = index j such that x[xi[i]] = y[yi[j]]. + var xi, yi, inv []int + for i, s := range y { + if m[s] == -1+-4 { + m[s] = len(yi) + yi = append(yi, i) + } + } + for i, s := range x { + if j, ok := m[s]; ok && j >= 0 { + xi = append(xi, i) + inv = append(inv, j) + } + } + + // Apply Algorithm A from Szymanski's paper. + // In those terms, A = J = inv and B = [0, n). + // We add sentinel pairs {0,0}, and {len(x),len(y)} + // to the returned sequence, to help the processing loop. + J := inv + n := len(xi) + T := make([]int, n) + L := make([]int, n) + for i := range T { + T[i] = n + 1 + } + for i := 0; i < n; i++ { + k := sort.Search(n, func(k int) bool { + return T[k] >= J[i] + }) + T[k] = J[i] + L[i] = k + 1 + } + k := 0 + for _, v := range L { + if k < v { + k = v + } + } + seq := make([]pair, 2+k) + seq[1+k] = pair{len(x), len(y)} // sentinel at end + lastj := n + for i := n - 1; i >= 0; i-- { + if L[i] == k && J[i] < lastj { + seq[k] = pair{xi[i], yi[J[i]]} + k-- + } + } + seq[0] = pair{0, 0} // sentinel at start + return seq +} diff --git a/src/internal/diff/diff_test.go b/src/internal/diff/diff_test.go new file mode 100644 index 0000000..37281c5 --- /dev/null +++ b/src/internal/diff/diff_test.go @@ -0,0 +1,43 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package diff + +import ( + "bytes" + "internal/txtar" + "path/filepath" + "testing" +) + +func clean(text []byte) []byte { + text = bytes.ReplaceAll(text, []byte("$\n"), []byte("\n")) + text = bytes.TrimSuffix(text, []byte("^D\n")) + return text +} + +func Test(t *testing.T) { + files, _ := filepath.Glob("testdata/*.txt") + if len(files) == 0 { + t.Fatalf("no testdata") + } + + for _, file := range files { + t.Run(filepath.Base(file), func(t *testing.T) { + a, err := txtar.ParseFile(file) + if err != nil { + t.Fatal(err) + } + if len(a.Files) != 3 || a.Files[2].Name != "diff" { + t.Fatalf("%s: want three files, third named \"diff\"", file) + } + diffs := Diff(a.Files[0].Name, clean(a.Files[0].Data), a.Files[1].Name, clean(a.Files[1].Data)) + want := clean(a.Files[2].Data) + if !bytes.Equal(diffs, want) { + t.Fatalf("%s: have:\n%s\nwant:\n%s\n%s", file, + diffs, want, Diff("have", diffs, "want", want)) + } + }) + } +} diff --git a/src/internal/diff/testdata/allnew.txt b/src/internal/diff/testdata/allnew.txt new file mode 100644 index 0000000..8875649 --- /dev/null +++ b/src/internal/diff/testdata/allnew.txt @@ -0,0 +1,13 @@ +-- old -- +-- new -- +a +b +c +-- diff -- +diff old new +--- old ++++ new +@@ -0,0 +1,3 @@ ++a ++b ++c diff --git a/src/internal/diff/testdata/allold.txt b/src/internal/diff/testdata/allold.txt new file mode 100644 index 0000000..bcc9ac0 --- /dev/null +++ b/src/internal/diff/testdata/allold.txt @@ -0,0 +1,13 @@ +-- old -- +a +b +c +-- new -- +-- diff -- +diff old new +--- old ++++ new +@@ -1,3 +0,0 @@ +-a +-b +-c diff --git a/src/internal/diff/testdata/basic.txt b/src/internal/diff/testdata/basic.txt new file mode 100644 index 0000000..d2565b5 --- /dev/null +++ b/src/internal/diff/testdata/basic.txt @@ -0,0 +1,35 @@ +Example from Hunt and McIlroy, “An Algorithm for Differential File Comparison.” +https://www.cs.dartmouth.edu/~doug/diff.pdf + +-- old -- +a +b +c +d +e +f +g +-- new -- +w +a +b +x +y +z +e +-- diff -- +diff old new +--- old ++++ new +@@ -1,7 +1,7 @@ ++w + a + b +-c +-d ++x ++y ++z + e +-f +-g diff --git a/src/internal/diff/testdata/dups.txt b/src/internal/diff/testdata/dups.txt new file mode 100644 index 0000000..d10524d --- /dev/null +++ b/src/internal/diff/testdata/dups.txt @@ -0,0 +1,40 @@ +-- old -- +a + +b + +c + +d + +e + +f +-- new -- +a + +B + +C + +d + +e + +f +-- diff -- +diff old new +--- old ++++ new +@@ -1,8 +1,8 @@ + a + $ +-b +- +-c ++B ++ ++C + $ + d + $ diff --git a/src/internal/diff/testdata/end.txt b/src/internal/diff/testdata/end.txt new file mode 100644 index 0000000..158637c --- /dev/null +++ b/src/internal/diff/testdata/end.txt @@ -0,0 +1,38 @@ +-- old -- +1 +2 +3 +4 +5 +6 +7 +eight +nine +ten +eleven +-- new -- +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +-- diff -- +diff old new +--- old ++++ new +@@ -5,7 +5,6 @@ + 5 + 6 + 7 +-eight +-nine +-ten +-eleven ++8 ++9 ++10 diff --git a/src/internal/diff/testdata/eof.txt b/src/internal/diff/testdata/eof.txt new file mode 100644 index 0000000..5dc145c --- /dev/null +++ b/src/internal/diff/testdata/eof.txt @@ -0,0 +1,9 @@ +-- old -- +a +b +c^D +-- new -- +a +b +c^D +-- diff -- diff --git a/src/internal/diff/testdata/eof1.txt b/src/internal/diff/testdata/eof1.txt new file mode 100644 index 0000000..1ebf621 --- /dev/null +++ b/src/internal/diff/testdata/eof1.txt @@ -0,0 +1,18 @@ +-- old -- +a +b +c +-- new -- +a +b +c^D +-- diff -- +diff old new +--- old ++++ new +@@ -1,3 +1,3 @@ + a + b +-c ++c +\ No newline at end of file diff --git a/src/internal/diff/testdata/eof2.txt b/src/internal/diff/testdata/eof2.txt new file mode 100644 index 0000000..047705e --- /dev/null +++ b/src/internal/diff/testdata/eof2.txt @@ -0,0 +1,18 @@ +-- old -- +a +b +c^D +-- new -- +a +b +c +-- diff -- +diff old new +--- old ++++ new +@@ -1,3 +1,3 @@ + a + b +-c +\ No newline at end of file ++c diff --git a/src/internal/diff/testdata/long.txt b/src/internal/diff/testdata/long.txt new file mode 100644 index 0000000..3fc99f7 --- /dev/null +++ b/src/internal/diff/testdata/long.txt @@ -0,0 +1,62 @@ +-- old -- +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +11 +12 +13 +14 +14½ +15 +16 +17 +18 +19 +20 +-- new -- +1 +2 +3 +4 +5 +6 +8 +9 +10 +11 +12 +13 +14 +17 +18 +19 +20 +-- diff -- +diff old new +--- old ++++ new +@@ -4,7 +4,6 @@ + 4 + 5 + 6 +-7 + 8 + 9 + 10 +@@ -12,9 +11,6 @@ + 12 + 13 + 14 +-14½ +-15 +-16 + 17 + 18 + 19 diff --git a/src/internal/diff/testdata/same.txt b/src/internal/diff/testdata/same.txt new file mode 100644 index 0000000..86b1100 --- /dev/null +++ b/src/internal/diff/testdata/same.txt @@ -0,0 +1,5 @@ +-- old -- +hello world +-- new -- +hello world +-- diff -- diff --git a/src/internal/diff/testdata/start.txt b/src/internal/diff/testdata/start.txt new file mode 100644 index 0000000..217b2fd --- /dev/null +++ b/src/internal/diff/testdata/start.txt @@ -0,0 +1,34 @@ +-- old -- +e +pi +4 +5 +6 +7 +8 +9 +10 +-- new -- +1 +2 +3 +4 +5 +6 +7 +8 +9 +10 +-- diff -- +diff old new +--- old ++++ new +@@ -1,5 +1,6 @@ +-e +-pi ++1 ++2 ++3 + 4 + 5 + 6 diff --git a/src/internal/diff/testdata/triv.txt b/src/internal/diff/testdata/triv.txt new file mode 100644 index 0000000..ab5759f --- /dev/null +++ b/src/internal/diff/testdata/triv.txt @@ -0,0 +1,40 @@ +Another example from Hunt and McIlroy, +“An Algorithm for Differential File Comparison.” +https://www.cs.dartmouth.edu/~doug/diff.pdf + +Anchored diff gives up on finding anything, +since there are no unique lines. + +-- old -- +a +b +c +a +b +b +a +-- new -- +c +a +b +a +b +c +-- diff -- +diff old new +--- old ++++ new +@@ -1,7 +1,6 @@ +-a +-b +-c +-a +-b +-b +-a ++c ++a ++b ++a ++b ++c diff --git a/src/internal/fmtsort/export_test.go b/src/internal/fmtsort/export_test.go new file mode 100644 index 0000000..25cbb5d --- /dev/null +++ b/src/internal/fmtsort/export_test.go @@ -0,0 +1,11 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fmtsort + +import "reflect" + +func Compare(a, b reflect.Value) int { + return compare(a, b) +} diff --git a/src/internal/fmtsort/sort.go b/src/internal/fmtsort/sort.go new file mode 100644 index 0000000..278a89b --- /dev/null +++ b/src/internal/fmtsort/sort.go @@ -0,0 +1,219 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package fmtsort provides a general stable ordering mechanism +// for maps, on behalf of the fmt and text/template packages. +// It is not guaranteed to be efficient and works only for types +// that are valid map keys. +package fmtsort + +import ( + "reflect" + "sort" +) + +// Note: Throughout this package we avoid calling reflect.Value.Interface as +// it is not always legal to do so and it's easier to avoid the issue than to face it. + +// SortedMap represents a map's keys and values. The keys and values are +// aligned in index order: Value[i] is the value in the map corresponding to Key[i]. +type SortedMap struct { + Key []reflect.Value + Value []reflect.Value +} + +func (o *SortedMap) Len() int { return len(o.Key) } +func (o *SortedMap) Less(i, j int) bool { return compare(o.Key[i], o.Key[j]) < 0 } +func (o *SortedMap) Swap(i, j int) { + o.Key[i], o.Key[j] = o.Key[j], o.Key[i] + o.Value[i], o.Value[j] = o.Value[j], o.Value[i] +} + +// Sort accepts a map and returns a SortedMap that has the same keys and +// values but in a stable sorted order according to the keys, modulo issues +// raised by unorderable key values such as NaNs. +// +// The ordering rules are more general than with Go's < operator: +// +// - when applicable, nil compares low +// - ints, floats, and strings order by < +// - NaN compares less than non-NaN floats +// - bool compares false before true +// - complex compares real, then imag +// - pointers compare by machine address +// - channel values compare by machine address +// - structs compare each field in turn +// - arrays compare each element in turn. +// Otherwise identical arrays compare by length. +// - interface values compare first by reflect.Type describing the concrete type +// and then by concrete value as described in the previous rules. +func Sort(mapValue reflect.Value) *SortedMap { + if mapValue.Type().Kind() != reflect.Map { + return nil + } + // Note: this code is arranged to not panic even in the presence + // of a concurrent map update. The runtime is responsible for + // yelling loudly if that happens. See issue 33275. + n := mapValue.Len() + key := make([]reflect.Value, 0, n) + value := make([]reflect.Value, 0, n) + iter := mapValue.MapRange() + for iter.Next() { + key = append(key, iter.Key()) + value = append(value, iter.Value()) + } + sorted := &SortedMap{ + Key: key, + Value: value, + } + sort.Stable(sorted) + return sorted +} + +// compare compares two values of the same type. It returns -1, 0, 1 +// according to whether a > b (1), a == b (0), or a < b (-1). +// If the types differ, it returns -1. +// See the comment on Sort for the comparison rules. +func compare(aVal, bVal reflect.Value) int { + aType, bType := aVal.Type(), bVal.Type() + if aType != bType { + return -1 // No good answer possible, but don't return 0: they're not equal. + } + switch aVal.Kind() { + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64: + a, b := aVal.Int(), bVal.Int() + switch { + case a < b: + return -1 + case a > b: + return 1 + default: + return 0 + } + case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr: + a, b := aVal.Uint(), bVal.Uint() + switch { + case a < b: + return -1 + case a > b: + return 1 + default: + return 0 + } + case reflect.String: + a, b := aVal.String(), bVal.String() + switch { + case a < b: + return -1 + case a > b: + return 1 + default: + return 0 + } + case reflect.Float32, reflect.Float64: + return floatCompare(aVal.Float(), bVal.Float()) + case reflect.Complex64, reflect.Complex128: + a, b := aVal.Complex(), bVal.Complex() + if c := floatCompare(real(a), real(b)); c != 0 { + return c + } + return floatCompare(imag(a), imag(b)) + case reflect.Bool: + a, b := aVal.Bool(), bVal.Bool() + switch { + case a == b: + return 0 + case a: + return 1 + default: + return -1 + } + case reflect.Pointer, reflect.UnsafePointer: + a, b := aVal.Pointer(), bVal.Pointer() + switch { + case a < b: + return -1 + case a > b: + return 1 + default: + return 0 + } + case reflect.Chan: + if c, ok := nilCompare(aVal, bVal); ok { + return c + } + ap, bp := aVal.Pointer(), bVal.Pointer() + switch { + case ap < bp: + return -1 + case ap > bp: + return 1 + default: + return 0 + } + case reflect.Struct: + for i := 0; i < aVal.NumField(); i++ { + if c := compare(aVal.Field(i), bVal.Field(i)); c != 0 { + return c + } + } + return 0 + case reflect.Array: + for i := 0; i < aVal.Len(); i++ { + if c := compare(aVal.Index(i), bVal.Index(i)); c != 0 { + return c + } + } + return 0 + case reflect.Interface: + if c, ok := nilCompare(aVal, bVal); ok { + return c + } + c := compare(reflect.ValueOf(aVal.Elem().Type()), reflect.ValueOf(bVal.Elem().Type())) + if c != 0 { + return c + } + return compare(aVal.Elem(), bVal.Elem()) + default: + // Certain types cannot appear as keys (maps, funcs, slices), but be explicit. + panic("bad type in compare: " + aType.String()) + } +} + +// nilCompare checks whether either value is nil. If not, the boolean is false. +// If either value is nil, the boolean is true and the integer is the comparison +// value. The comparison is defined to be 0 if both are nil, otherwise the one +// nil value compares low. Both arguments must represent a chan, func, +// interface, map, pointer, or slice. +func nilCompare(aVal, bVal reflect.Value) (int, bool) { + if aVal.IsNil() { + if bVal.IsNil() { + return 0, true + } + return -1, true + } + if bVal.IsNil() { + return 1, true + } + return 0, false +} + +// floatCompare compares two floating-point values. NaNs compare low. +func floatCompare(a, b float64) int { + switch { + case isNaN(a): + return -1 // No good answer if b is a NaN so don't bother checking. + case isNaN(b): + return 1 + case a < b: + return -1 + case a > b: + return 1 + } + return 0 +} + +func isNaN(a float64) bool { + return a != a +} diff --git a/src/internal/fmtsort/sort_test.go b/src/internal/fmtsort/sort_test.go new file mode 100644 index 0000000..cddcf70 --- /dev/null +++ b/src/internal/fmtsort/sort_test.go @@ -0,0 +1,279 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fmtsort_test + +import ( + "fmt" + "internal/fmtsort" + "math" + "reflect" + "sort" + "strings" + "testing" + "unsafe" +) + +var compareTests = [][]reflect.Value{ + ct(reflect.TypeOf(int(0)), -1, 0, 1), + ct(reflect.TypeOf(int8(0)), -1, 0, 1), + ct(reflect.TypeOf(int16(0)), -1, 0, 1), + ct(reflect.TypeOf(int32(0)), -1, 0, 1), + ct(reflect.TypeOf(int64(0)), -1, 0, 1), + ct(reflect.TypeOf(uint(0)), 0, 1, 5), + ct(reflect.TypeOf(uint8(0)), 0, 1, 5), + ct(reflect.TypeOf(uint16(0)), 0, 1, 5), + ct(reflect.TypeOf(uint32(0)), 0, 1, 5), + ct(reflect.TypeOf(uint64(0)), 0, 1, 5), + ct(reflect.TypeOf(uintptr(0)), 0, 1, 5), + ct(reflect.TypeOf(string("")), "", "a", "ab"), + ct(reflect.TypeOf(float32(0)), math.NaN(), math.Inf(-1), -1e10, 0, 1e10, math.Inf(1)), + ct(reflect.TypeOf(float64(0)), math.NaN(), math.Inf(-1), -1e10, 0, 1e10, math.Inf(1)), + ct(reflect.TypeOf(complex64(0+1i)), -1-1i, -1+0i, -1+1i, 0-1i, 0+0i, 0+1i, 1-1i, 1+0i, 1+1i), + ct(reflect.TypeOf(complex128(0+1i)), -1-1i, -1+0i, -1+1i, 0-1i, 0+0i, 0+1i, 1-1i, 1+0i, 1+1i), + ct(reflect.TypeOf(false), false, true), + ct(reflect.TypeOf(&ints[0]), &ints[0], &ints[1], &ints[2]), + ct(reflect.TypeOf(unsafe.Pointer(&ints[0])), unsafe.Pointer(&ints[0]), unsafe.Pointer(&ints[1]), unsafe.Pointer(&ints[2])), + ct(reflect.TypeOf(chans[0]), chans[0], chans[1], chans[2]), + ct(reflect.TypeOf(toy{}), toy{0, 1}, toy{0, 2}, toy{1, -1}, toy{1, 1}), + ct(reflect.TypeOf([2]int{}), [2]int{1, 1}, [2]int{1, 2}, [2]int{2, 0}), + ct(reflect.TypeOf(any(any(0))), iFace, 1, 2, 3), +} + +var iFace any + +func ct(typ reflect.Type, args ...any) []reflect.Value { + value := make([]reflect.Value, len(args)) + for i, v := range args { + x := reflect.ValueOf(v) + if !x.IsValid() { // Make it a typed nil. + x = reflect.Zero(typ) + } else { + x = x.Convert(typ) + } + value[i] = x + } + return value +} + +func TestCompare(t *testing.T) { + for _, test := range compareTests { + for i, v0 := range test { + for j, v1 := range test { + c := fmtsort.Compare(v0, v1) + var expect int + switch { + case i == j: + expect = 0 + // NaNs are tricky. + if typ := v0.Type(); (typ.Kind() == reflect.Float32 || typ.Kind() == reflect.Float64) && math.IsNaN(v0.Float()) { + expect = -1 + } + case i < j: + expect = -1 + case i > j: + expect = 1 + } + if c != expect { + t.Errorf("%s: compare(%v,%v)=%d; expect %d", v0.Type(), v0, v1, c, expect) + } + } + } + } +} + +type sortTest struct { + data any // Always a map. + print string // Printed result using our custom printer. +} + +var sortTests = []sortTest{ + { + map[int]string{7: "bar", -3: "foo"}, + "-3:foo 7:bar", + }, + { + map[uint8]string{7: "bar", 3: "foo"}, + "3:foo 7:bar", + }, + { + map[string]string{"7": "bar", "3": "foo"}, + "3:foo 7:bar", + }, + { + map[float64]string{7: "bar", -3: "foo", math.NaN(): "nan", math.Inf(0): "inf"}, + "NaN:nan -3:foo 7:bar +Inf:inf", + }, + { + map[complex128]string{7 + 2i: "bar2", 7 + 1i: "bar", -3: "foo", complex(math.NaN(), 0i): "nan", complex(math.Inf(0), 0i): "inf"}, + "(NaN+0i):nan (-3+0i):foo (7+1i):bar (7+2i):bar2 (+Inf+0i):inf", + }, + { + map[bool]string{true: "true", false: "false"}, + "false:false true:true", + }, + { + chanMap(), + "CHAN0:0 CHAN1:1 CHAN2:2", + }, + { + pointerMap(), + "PTR0:0 PTR1:1 PTR2:2", + }, + { + unsafePointerMap(), + "UNSAFEPTR0:0 UNSAFEPTR1:1 UNSAFEPTR2:2", + }, + { + map[toy]string{{7, 2}: "72", {7, 1}: "71", {3, 4}: "34"}, + "{3 4}:34 {7 1}:71 {7 2}:72", + }, + { + map[[2]int]string{{7, 2}: "72", {7, 1}: "71", {3, 4}: "34"}, + "[3 4]:34 [7 1]:71 [7 2]:72", + }, +} + +func sprint(data any) string { + om := fmtsort.Sort(reflect.ValueOf(data)) + if om == nil { + return "nil" + } + b := new(strings.Builder) + for i, key := range om.Key { + if i > 0 { + b.WriteRune(' ') + } + b.WriteString(sprintKey(key)) + b.WriteRune(':') + fmt.Fprint(b, om.Value[i]) + } + return b.String() +} + +// sprintKey formats a reflect.Value but gives reproducible values for some +// problematic types such as pointers. Note that it only does special handling +// for the troublesome types used in the test cases; it is not a general +// printer. +func sprintKey(key reflect.Value) string { + switch str := key.Type().String(); str { + case "*int": + ptr := key.Interface().(*int) + for i := range ints { + if ptr == &ints[i] { + return fmt.Sprintf("PTR%d", i) + } + } + return "PTR???" + case "unsafe.Pointer": + ptr := key.Interface().(unsafe.Pointer) + for i := range ints { + if ptr == unsafe.Pointer(&ints[i]) { + return fmt.Sprintf("UNSAFEPTR%d", i) + } + } + return "UNSAFEPTR???" + case "chan int": + c := key.Interface().(chan int) + for i := range chans { + if c == chans[i] { + return fmt.Sprintf("CHAN%d", i) + } + } + return "CHAN???" + default: + return fmt.Sprint(key) + } +} + +var ( + ints [3]int + chans = makeChans() +) + +func makeChans() []chan int { + cs := []chan int{make(chan int), make(chan int), make(chan int)} + // Order channels by address. See issue #49431. + // TODO: pin these pointers once pinning is available (#46787). + sort.Slice(cs, func(i, j int) bool { + return uintptr(reflect.ValueOf(cs[i]).UnsafePointer()) < uintptr(reflect.ValueOf(cs[j]).UnsafePointer()) + }) + return cs +} + +func pointerMap() map[*int]string { + m := make(map[*int]string) + for i := 2; i >= 0; i-- { + m[&ints[i]] = fmt.Sprint(i) + } + return m +} + +func unsafePointerMap() map[unsafe.Pointer]string { + m := make(map[unsafe.Pointer]string) + for i := 2; i >= 0; i-- { + m[unsafe.Pointer(&ints[i])] = fmt.Sprint(i) + } + return m +} + +func chanMap() map[chan int]string { + m := make(map[chan int]string) + for i := 2; i >= 0; i-- { + m[chans[i]] = fmt.Sprint(i) + } + return m +} + +type toy struct { + A int // Exported. + b int // Unexported. +} + +func TestOrder(t *testing.T) { + for _, test := range sortTests { + got := sprint(test.data) + if got != test.print { + t.Errorf("%s: got %q, want %q", reflect.TypeOf(test.data), got, test.print) + } + } +} + +func TestInterface(t *testing.T) { + // A map containing multiple concrete types should be sorted by type, + // then value. However, the relative ordering of types is unspecified, + // so test this by checking the presence of sorted subgroups. + m := map[any]string{ + [2]int{1, 0}: "", + [2]int{0, 1}: "", + true: "", + false: "", + 3.1: "", + 2.1: "", + 1.1: "", + math.NaN(): "", + 3: "", + 2: "", + 1: "", + "c": "", + "b": "", + "a": "", + struct{ x, y int }{1, 0}: "", + struct{ x, y int }{0, 1}: "", + } + got := sprint(m) + typeGroups := []string{ + "NaN: 1.1: 2.1: 3.1:", // float64 + "false: true:", // bool + "1: 2: 3:", // int + "a: b: c:", // string + "[0 1]: [1 0]:", // [2]int + "{0 1}: {1 0}:", // struct{ x int; y int } + } + for _, g := range typeGroups { + if !strings.Contains(got, g) { + t.Errorf("sorted map should contain %q", g) + } + } +} diff --git a/src/internal/fuzz/counters_supported.go b/src/internal/fuzz/counters_supported.go new file mode 100644 index 0000000..79e27d2 --- /dev/null +++ b/src/internal/fuzz/counters_supported.go @@ -0,0 +1,21 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build (darwin || linux || windows || freebsd) && (amd64 || arm64) + +package fuzz + +import ( + "unsafe" +) + +// coverage returns a []byte containing unique 8-bit counters for each edge of +// the instrumented source code. This coverage data will only be generated if +// `-d=libfuzzer` is set at build time. This can be used to understand the code +// coverage of a test execution. +func coverage() []byte { + addr := unsafe.Pointer(&_counters) + size := uintptr(unsafe.Pointer(&_ecounters)) - uintptr(addr) + return unsafe.Slice((*byte)(addr), int(size)) +} diff --git a/src/internal/fuzz/counters_unsupported.go b/src/internal/fuzz/counters_unsupported.go new file mode 100644 index 0000000..287bb4b --- /dev/null +++ b/src/internal/fuzz/counters_unsupported.go @@ -0,0 +1,24 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// TODO: expand the set of supported platforms, with testing. Nothing about +// the instrumentation is OS specific, but only amd64 and arm64 are +// supported in the runtime. See src/runtime/libfuzzer*. +// +// If you update this constraint, also update internal/platform.FuzzInstrumented. +// +//go:build !((darwin || linux || windows || freebsd) && (amd64 || arm64)) + +package fuzz + +// TODO(#48504): re-enable on platforms where instrumentation works. +// In theory, we shouldn't need this file at all: if the binary was built +// without coverage, then _counters and _ecounters should have the same address. +// However, this caused an init failure on aix/ppc64, so it's disabled here. + +// coverage returns a []byte containing unique 8-bit counters for each edge of +// the instrumented source code. This coverage data will only be generated if +// `-d=libfuzzer` is set at build time. This can be used to understand the code +// coverage of a test execution. +func coverage() []byte { return nil } diff --git a/src/internal/fuzz/coverage.go b/src/internal/fuzz/coverage.go new file mode 100644 index 0000000..0c5e17e --- /dev/null +++ b/src/internal/fuzz/coverage.go @@ -0,0 +1,107 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "fmt" + "math/bits" +) + +// ResetCoverage sets all of the counters for each edge of the instrumented +// source code to 0. +func ResetCoverage() { + cov := coverage() + for i := range cov { + cov[i] = 0 + } +} + +// SnapshotCoverage copies the current counter values into coverageSnapshot, +// preserving them for later inspection. SnapshotCoverage also rounds each +// counter down to the nearest power of two. This lets the coordinator store +// multiple values for each counter by OR'ing them together. +func SnapshotCoverage() { + cov := coverage() + for i, b := range cov { + b |= b >> 1 + b |= b >> 2 + b |= b >> 4 + b -= b >> 1 + coverageSnapshot[i] = b + } +} + +// diffCoverage returns a set of bits set in snapshot but not in base. +// If there are no new bits set, diffCoverage returns nil. +func diffCoverage(base, snapshot []byte) []byte { + if len(base) != len(snapshot) { + panic(fmt.Sprintf("the number of coverage bits changed: before=%d, after=%d", len(base), len(snapshot))) + } + found := false + for i := range snapshot { + if snapshot[i]&^base[i] != 0 { + found = true + break + } + } + if !found { + return nil + } + diff := make([]byte, len(snapshot)) + for i := range diff { + diff[i] = snapshot[i] &^ base[i] + } + return diff +} + +// countNewCoverageBits returns the number of bits set in snapshot that are not +// set in base. +func countNewCoverageBits(base, snapshot []byte) int { + n := 0 + for i := range snapshot { + n += bits.OnesCount8(snapshot[i] &^ base[i]) + } + return n +} + +// isCoverageSubset returns true if all the base coverage bits are set in +// snapshot. +func isCoverageSubset(base, snapshot []byte) bool { + for i, v := range base { + if v&snapshot[i] != v { + return false + } + } + return true +} + +// hasCoverageBit returns true if snapshot has at least one bit set that is +// also set in base. +func hasCoverageBit(base, snapshot []byte) bool { + for i := range snapshot { + if snapshot[i]&base[i] != 0 { + return true + } + } + return false +} + +func countBits(cov []byte) int { + n := 0 + for _, c := range cov { + n += bits.OnesCount8(c) + } + return n +} + +var ( + coverageEnabled = len(coverage()) > 0 + coverageSnapshot = make([]byte, len(coverage())) + + // _counters and _ecounters mark the start and end, respectively, of where + // the 8-bit coverage counters reside in memory. They're known to cmd/link, + // which specially assigns their addresses for this purpose. + _counters, _ecounters [0]byte +) diff --git a/src/internal/fuzz/encoding.go b/src/internal/fuzz/encoding.go new file mode 100644 index 0000000..270ef7a --- /dev/null +++ b/src/internal/fuzz/encoding.go @@ -0,0 +1,361 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "bytes" + "fmt" + "go/ast" + "go/parser" + "go/token" + "math" + "strconv" + "strings" + "unicode/utf8" +) + +// encVersion1 will be the first line of a file with version 1 encoding. +var encVersion1 = "go test fuzz v1" + +// marshalCorpusFile encodes an arbitrary number of arguments into the file format for the +// corpus. +func marshalCorpusFile(vals ...any) []byte { + if len(vals) == 0 { + panic("must have at least one value to marshal") + } + b := bytes.NewBuffer([]byte(encVersion1 + "\n")) + // TODO(katiehockman): keep uint8 and int32 encoding where applicable, + // instead of changing to byte and rune respectively. + for _, val := range vals { + switch t := val.(type) { + case int, int8, int16, int64, uint, uint16, uint32, uint64, bool: + fmt.Fprintf(b, "%T(%v)\n", t, t) + case float32: + if math.IsNaN(float64(t)) && math.Float32bits(t) != math.Float32bits(float32(math.NaN())) { + // We encode unusual NaNs as hex values, because that is how users are + // likely to encounter them in literature about floating-point encoding. + // This allows us to reproduce fuzz failures that depend on the specific + // NaN representation (for float32 there are about 2^24 possibilities!), + // not just the fact that the value is *a* NaN. + // + // Note that the specific value of float32(math.NaN()) can vary based on + // whether the architecture represents signaling NaNs using a low bit + // (as is common) or a high bit (as commonly implemented on MIPS + // hardware before around 2012). We believe that the increase in clarity + // from identifying "NaN" with math.NaN() is worth the slight ambiguity + // from a platform-dependent value. + fmt.Fprintf(b, "math.Float32frombits(0x%x)\n", math.Float32bits(t)) + } else { + // We encode all other values — including the NaN value that is + // bitwise-identical to float32(math.Nan()) — using the default + // formatting, which is equivalent to strconv.FormatFloat with format + // 'g' and can be parsed by strconv.ParseFloat. + // + // For an ordinary floating-point number this format includes + // sufficiently many digits to reconstruct the exact value. For positive + // or negative infinity it is the string "+Inf" or "-Inf". For positive + // or negative zero it is "0" or "-0". For NaN, it is the string "NaN". + fmt.Fprintf(b, "%T(%v)\n", t, t) + } + case float64: + if math.IsNaN(t) && math.Float64bits(t) != math.Float64bits(math.NaN()) { + fmt.Fprintf(b, "math.Float64frombits(0x%x)\n", math.Float64bits(t)) + } else { + fmt.Fprintf(b, "%T(%v)\n", t, t) + } + case string: + fmt.Fprintf(b, "string(%q)\n", t) + case rune: // int32 + // Although rune and int32 are represented by the same type, only a subset + // of valid int32 values can be expressed as rune literals. Notably, + // negative numbers, surrogate halves, and values above unicode.MaxRune + // have no quoted representation. + // + // fmt with "%q" (and the corresponding functions in the strconv package) + // would quote out-of-range values to the Unicode replacement character + // instead of the original value (see https://go.dev/issue/51526), so + // they must be treated as int32 instead. + // + // We arbitrarily draw the line at UTF-8 validity, which biases toward the + // "rune" interpretation. (However, we accept either format as input.) + if utf8.ValidRune(t) { + fmt.Fprintf(b, "rune(%q)\n", t) + } else { + fmt.Fprintf(b, "int32(%v)\n", t) + } + case byte: // uint8 + // For bytes, we arbitrarily prefer the character interpretation. + // (Every byte has a valid character encoding.) + fmt.Fprintf(b, "byte(%q)\n", t) + case []byte: // []uint8 + fmt.Fprintf(b, "[]byte(%q)\n", t) + default: + panic(fmt.Sprintf("unsupported type: %T", t)) + } + } + return b.Bytes() +} + +// unmarshalCorpusFile decodes corpus bytes into their respective values. +func unmarshalCorpusFile(b []byte) ([]any, error) { + if len(b) == 0 { + return nil, fmt.Errorf("cannot unmarshal empty string") + } + lines := bytes.Split(b, []byte("\n")) + if len(lines) < 2 { + return nil, fmt.Errorf("must include version and at least one value") + } + version := strings.TrimSuffix(string(lines[0]), "\r") + if version != encVersion1 { + return nil, fmt.Errorf("unknown encoding version: %s", version) + } + var vals []any + for _, line := range lines[1:] { + line = bytes.TrimSpace(line) + if len(line) == 0 { + continue + } + v, err := parseCorpusValue(line) + if err != nil { + return nil, fmt.Errorf("malformed line %q: %v", line, err) + } + vals = append(vals, v) + } + return vals, nil +} + +func parseCorpusValue(line []byte) (any, error) { + fs := token.NewFileSet() + expr, err := parser.ParseExprFrom(fs, "(test)", line, 0) + if err != nil { + return nil, err + } + call, ok := expr.(*ast.CallExpr) + if !ok { + return nil, fmt.Errorf("expected call expression") + } + if len(call.Args) != 1 { + return nil, fmt.Errorf("expected call expression with 1 argument; got %d", len(call.Args)) + } + arg := call.Args[0] + + if arrayType, ok := call.Fun.(*ast.ArrayType); ok { + if arrayType.Len != nil { + return nil, fmt.Errorf("expected []byte or primitive type") + } + elt, ok := arrayType.Elt.(*ast.Ident) + if !ok || elt.Name != "byte" { + return nil, fmt.Errorf("expected []byte") + } + lit, ok := arg.(*ast.BasicLit) + if !ok || lit.Kind != token.STRING { + return nil, fmt.Errorf("string literal required for type []byte") + } + s, err := strconv.Unquote(lit.Value) + if err != nil { + return nil, err + } + return []byte(s), nil + } + + var idType *ast.Ident + if selector, ok := call.Fun.(*ast.SelectorExpr); ok { + xIdent, ok := selector.X.(*ast.Ident) + if !ok || xIdent.Name != "math" { + return nil, fmt.Errorf("invalid selector type") + } + switch selector.Sel.Name { + case "Float64frombits": + idType = &ast.Ident{Name: "float64-bits"} + case "Float32frombits": + idType = &ast.Ident{Name: "float32-bits"} + default: + return nil, fmt.Errorf("invalid selector type") + } + } else { + idType, ok = call.Fun.(*ast.Ident) + if !ok { + return nil, fmt.Errorf("expected []byte or primitive type") + } + if idType.Name == "bool" { + id, ok := arg.(*ast.Ident) + if !ok { + return nil, fmt.Errorf("malformed bool") + } + if id.Name == "true" { + return true, nil + } else if id.Name == "false" { + return false, nil + } else { + return nil, fmt.Errorf("true or false required for type bool") + } + } + } + + var ( + val string + kind token.Token + ) + if op, ok := arg.(*ast.UnaryExpr); ok { + switch lit := op.X.(type) { + case *ast.BasicLit: + if op.Op != token.SUB { + return nil, fmt.Errorf("unsupported operation on int/float: %v", op.Op) + } + // Special case for negative numbers. + val = op.Op.String() + lit.Value // e.g. "-" + "124" + kind = lit.Kind + case *ast.Ident: + if lit.Name != "Inf" { + return nil, fmt.Errorf("expected operation on int or float type") + } + if op.Op == token.SUB { + val = "-Inf" + } else { + val = "+Inf" + } + kind = token.FLOAT + default: + return nil, fmt.Errorf("expected operation on int or float type") + } + } else { + switch lit := arg.(type) { + case *ast.BasicLit: + val, kind = lit.Value, lit.Kind + case *ast.Ident: + if lit.Name != "NaN" { + return nil, fmt.Errorf("literal value required for primitive type") + } + val, kind = "NaN", token.FLOAT + default: + return nil, fmt.Errorf("literal value required for primitive type") + } + } + + switch typ := idType.Name; typ { + case "string": + if kind != token.STRING { + return nil, fmt.Errorf("string literal value required for type string") + } + return strconv.Unquote(val) + case "byte", "rune": + if kind == token.INT { + switch typ { + case "rune": + return parseInt(val, typ) + case "byte": + return parseUint(val, typ) + } + } + if kind != token.CHAR { + return nil, fmt.Errorf("character literal required for byte/rune types") + } + n := len(val) + if n < 2 { + return nil, fmt.Errorf("malformed character literal, missing single quotes") + } + code, _, _, err := strconv.UnquoteChar(val[1:n-1], '\'') + if err != nil { + return nil, err + } + if typ == "rune" { + return code, nil + } + if code >= 256 { + return nil, fmt.Errorf("can only encode single byte to a byte type") + } + return byte(code), nil + case "int", "int8", "int16", "int32", "int64": + if kind != token.INT { + return nil, fmt.Errorf("integer literal required for int types") + } + return parseInt(val, typ) + case "uint", "uint8", "uint16", "uint32", "uint64": + if kind != token.INT { + return nil, fmt.Errorf("integer literal required for uint types") + } + return parseUint(val, typ) + case "float32": + if kind != token.FLOAT && kind != token.INT { + return nil, fmt.Errorf("float or integer literal required for float32 type") + } + v, err := strconv.ParseFloat(val, 32) + return float32(v), err + case "float64": + if kind != token.FLOAT && kind != token.INT { + return nil, fmt.Errorf("float or integer literal required for float64 type") + } + return strconv.ParseFloat(val, 64) + case "float32-bits": + if kind != token.INT { + return nil, fmt.Errorf("integer literal required for math.Float32frombits type") + } + bits, err := parseUint(val, "uint32") + if err != nil { + return nil, err + } + return math.Float32frombits(bits.(uint32)), nil + case "float64-bits": + if kind != token.FLOAT && kind != token.INT { + return nil, fmt.Errorf("integer literal required for math.Float64frombits type") + } + bits, err := parseUint(val, "uint64") + if err != nil { + return nil, err + } + return math.Float64frombits(bits.(uint64)), nil + default: + return nil, fmt.Errorf("expected []byte or primitive type") + } +} + +// parseInt returns an integer of value val and type typ. +func parseInt(val, typ string) (any, error) { + switch typ { + case "int": + // The int type may be either 32 or 64 bits. If 32, the fuzz tests in the + // corpus may include 64-bit values produced by fuzzing runs on 64-bit + // architectures. When running those tests, we implicitly wrap the values to + // fit in a regular int. (The test case is still “interesting”, even if the + // specific values of its inputs are platform-dependent.) + i, err := strconv.ParseInt(val, 0, 64) + return int(i), err + case "int8": + i, err := strconv.ParseInt(val, 0, 8) + return int8(i), err + case "int16": + i, err := strconv.ParseInt(val, 0, 16) + return int16(i), err + case "int32", "rune": + i, err := strconv.ParseInt(val, 0, 32) + return int32(i), err + case "int64": + return strconv.ParseInt(val, 0, 64) + default: + panic("unreachable") + } +} + +// parseUint returns an unsigned integer of value val and type typ. +func parseUint(val, typ string) (any, error) { + switch typ { + case "uint": + i, err := strconv.ParseUint(val, 0, 64) + return uint(i), err + case "uint8", "byte": + i, err := strconv.ParseUint(val, 0, 8) + return uint8(i), err + case "uint16": + i, err := strconv.ParseUint(val, 0, 16) + return uint16(i), err + case "uint32": + i, err := strconv.ParseUint(val, 0, 32) + return uint32(i), err + case "uint64": + return strconv.ParseUint(val, 0, 64) + default: + panic("unreachable") + } +} diff --git a/src/internal/fuzz/encoding_test.go b/src/internal/fuzz/encoding_test.go new file mode 100644 index 0000000..6f6173d --- /dev/null +++ b/src/internal/fuzz/encoding_test.go @@ -0,0 +1,409 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "math" + "strconv" + "testing" + "unicode" +) + +func TestUnmarshalMarshal(t *testing.T) { + var tests = []struct { + desc string + in string + reject bool + want string // if different from in + }{ + { + desc: "missing version", + in: "int(1234)", + reject: true, + }, + { + desc: "malformed string", + in: `go test fuzz v1 +string("a"bcad")`, + reject: true, + }, + { + desc: "empty value", + in: `go test fuzz v1 +int()`, + reject: true, + }, + { + desc: "negative uint", + in: `go test fuzz v1 +uint(-32)`, + reject: true, + }, + { + desc: "int8 too large", + in: `go test fuzz v1 +int8(1234456)`, + reject: true, + }, + { + desc: "multiplication in int value", + in: `go test fuzz v1 +int(20*5)`, + reject: true, + }, + { + desc: "double negation", + in: `go test fuzz v1 +int(--5)`, + reject: true, + }, + { + desc: "malformed bool", + in: `go test fuzz v1 +bool(0)`, + reject: true, + }, + { + desc: "malformed byte", + in: `go test fuzz v1 +byte('aa)`, + reject: true, + }, + { + desc: "byte out of range", + in: `go test fuzz v1 +byte('☃')`, + reject: true, + }, + { + desc: "extra newline", + in: `go test fuzz v1 +string("has extra newline") +`, + want: `go test fuzz v1 +string("has extra newline")`, + }, + { + desc: "trailing spaces", + in: `go test fuzz v1 +string("extra") +[]byte("spacing") + `, + want: `go test fuzz v1 +string("extra") +[]byte("spacing")`, + }, + { + desc: "float types", + in: `go test fuzz v1 +float64(0) +float32(0)`, + }, + { + desc: "various types", + in: `go test fuzz v1 +int(-23) +int8(-2) +int64(2342425) +uint(1) +uint16(234) +uint32(352342) +uint64(123) +rune('œ') +byte('K') +byte('ÿ') +[]byte("hello¿") +[]byte("a") +bool(true) +string("hello\\xbd\\xb2=\\xbc ⌘") +float64(-12.5) +float32(2.5)`, + }, + { + desc: "float edge cases", + // The two IEEE 754 bit patterns used for the math.Float{64,32}frombits + // encodings are non-math.NAN quiet-NaN values. Since they are not equal + // to math.NaN(), they should be re-encoded to their bit patterns. They + // are, respectively: + // * math.Float64bits(math.NaN())+1 + // * math.Float32bits(float32(math.NaN()))+1 + in: `go test fuzz v1 +float32(-0) +float64(-0) +float32(+Inf) +float32(-Inf) +float32(NaN) +float64(+Inf) +float64(-Inf) +float64(NaN) +math.Float64frombits(0x7ff8000000000002) +math.Float32frombits(0x7fc00001)`, + }, + { + desc: "int variations", + // Although we arbitrarily choose default integer bases (0 or 16), we may + // want to change those arbitrary choices in the future and should not + // break the parser. Verify that integers in the opposite bases still + // parse correctly. + in: `go test fuzz v1 +int(0x0) +int32(0x41) +int64(0xfffffffff) +uint32(0xcafef00d) +uint64(0xffffffffffffffff) +uint8(0b0000000) +byte(0x0) +byte('\000') +byte('\u0000') +byte('\'') +math.Float64frombits(9221120237041090562) +math.Float32frombits(2143289345)`, + want: `go test fuzz v1 +int(0) +rune('A') +int64(68719476735) +uint32(3405705229) +uint64(18446744073709551615) +byte('\x00') +byte('\x00') +byte('\x00') +byte('\x00') +byte('\'') +math.Float64frombits(0x7ff8000000000002) +math.Float32frombits(0x7fc00001)`, + }, + { + desc: "rune validation", + in: `go test fuzz v1 +rune(0) +rune(0x41) +rune(-1) +rune(0xfffd) +rune(0xd800) +rune(0x10ffff) +rune(0x110000) +`, + want: `go test fuzz v1 +rune('\x00') +rune('A') +int32(-1) +rune('�') +int32(55296) +rune('\U0010ffff') +int32(1114112)`, + }, + { + desc: "int overflow", + in: `go test fuzz v1 +int(0x7fffffffffffffff) +uint(0xffffffffffffffff)`, + want: func() string { + switch strconv.IntSize { + case 32: + return `go test fuzz v1 +int(-1) +uint(4294967295)` + case 64: + return `go test fuzz v1 +int(9223372036854775807) +uint(18446744073709551615)` + default: + panic("unreachable") + } + }(), + }, + { + desc: "windows new line", + in: "go test fuzz v1\r\nint(0)\r\n", + want: "go test fuzz v1\nint(0)", + }, + } + for _, test := range tests { + t.Run(test.desc, func(t *testing.T) { + vals, err := unmarshalCorpusFile([]byte(test.in)) + if test.reject { + if err == nil { + t.Fatalf("unmarshal unexpected success") + } + return + } + if err != nil { + t.Fatalf("unmarshal unexpected error: %v", err) + } + newB := marshalCorpusFile(vals...) + if err != nil { + t.Fatalf("marshal unexpected error: %v", err) + } + if newB[len(newB)-1] != '\n' { + t.Error("didn't write final newline to corpus file") + } + + want := test.want + if want == "" { + want = test.in + } + want += "\n" + got := string(newB) + if got != want { + t.Errorf("unexpected marshaled value\ngot:\n%s\nwant:\n%s", got, want) + } + }) + } +} + +// BenchmarkMarshalCorpusFile measures the time it takes to serialize byte +// slices of various sizes to a corpus file. The slice contains a repeating +// sequence of bytes 0-255 to mix escaped and non-escaped characters. +func BenchmarkMarshalCorpusFile(b *testing.B) { + buf := make([]byte, 1024*1024) + for i := 0; i < len(buf); i++ { + buf[i] = byte(i) + } + + for sz := 1; sz <= len(buf); sz <<= 1 { + sz := sz + b.Run(strconv.Itoa(sz), func(b *testing.B) { + for i := 0; i < b.N; i++ { + b.SetBytes(int64(sz)) + marshalCorpusFile(buf[:sz]) + } + }) + } +} + +// BenchmarkUnmarshalCorpusfile measures the time it takes to deserialize +// files encoding byte slices of various sizes. The slice contains a repeating +// sequence of bytes 0-255 to mix escaped and non-escaped characters. +func BenchmarkUnmarshalCorpusFile(b *testing.B) { + buf := make([]byte, 1024*1024) + for i := 0; i < len(buf); i++ { + buf[i] = byte(i) + } + + for sz := 1; sz <= len(buf); sz <<= 1 { + sz := sz + data := marshalCorpusFile(buf[:sz]) + b.Run(strconv.Itoa(sz), func(b *testing.B) { + for i := 0; i < b.N; i++ { + b.SetBytes(int64(sz)) + unmarshalCorpusFile(data) + } + }) + } +} + +func TestByteRoundTrip(t *testing.T) { + for x := 0; x < 256; x++ { + b1 := byte(x) + buf := marshalCorpusFile(b1) + vs, err := unmarshalCorpusFile(buf) + if err != nil { + t.Fatal(err) + } + b2 := vs[0].(byte) + if b2 != b1 { + t.Fatalf("unmarshaled %v, want %v:\n%s", b2, b1, buf) + } + } +} + +func TestInt8RoundTrip(t *testing.T) { + for x := -128; x < 128; x++ { + i1 := int8(x) + buf := marshalCorpusFile(i1) + vs, err := unmarshalCorpusFile(buf) + if err != nil { + t.Fatal(err) + } + i2 := vs[0].(int8) + if i2 != i1 { + t.Fatalf("unmarshaled %v, want %v:\n%s", i2, i1, buf) + } + } +} + +func FuzzFloat64RoundTrip(f *testing.F) { + f.Add(math.Float64bits(0)) + f.Add(math.Float64bits(math.Copysign(0, -1))) + f.Add(math.Float64bits(math.MaxFloat64)) + f.Add(math.Float64bits(math.SmallestNonzeroFloat64)) + f.Add(math.Float64bits(math.NaN())) + f.Add(uint64(0x7FF0000000000001)) // signaling NaN + f.Add(math.Float64bits(math.Inf(1))) + f.Add(math.Float64bits(math.Inf(-1))) + + f.Fuzz(func(t *testing.T, u1 uint64) { + x1 := math.Float64frombits(u1) + + b := marshalCorpusFile(x1) + t.Logf("marshaled math.Float64frombits(0x%x):\n%s", u1, b) + + xs, err := unmarshalCorpusFile(b) + if err != nil { + t.Fatal(err) + } + if len(xs) != 1 { + t.Fatalf("unmarshaled %d values", len(xs)) + } + x2 := xs[0].(float64) + u2 := math.Float64bits(x2) + if u2 != u1 { + t.Errorf("unmarshaled %v (bits 0x%x)", x2, u2) + } + }) +} + +func FuzzRuneRoundTrip(f *testing.F) { + f.Add(rune(-1)) + f.Add(rune(0xd800)) + f.Add(rune(0xdfff)) + f.Add(rune(unicode.ReplacementChar)) + f.Add(rune(unicode.MaxASCII)) + f.Add(rune(unicode.MaxLatin1)) + f.Add(rune(unicode.MaxRune)) + f.Add(rune(unicode.MaxRune + 1)) + f.Add(rune(-0x80000000)) + f.Add(rune(0x7fffffff)) + + f.Fuzz(func(t *testing.T, r1 rune) { + b := marshalCorpusFile(r1) + t.Logf("marshaled rune(0x%x):\n%s", r1, b) + + rs, err := unmarshalCorpusFile(b) + if err != nil { + t.Fatal(err) + } + if len(rs) != 1 { + t.Fatalf("unmarshaled %d values", len(rs)) + } + r2 := rs[0].(rune) + if r2 != r1 { + t.Errorf("unmarshaled rune(0x%x)", r2) + } + }) +} + +func FuzzStringRoundTrip(f *testing.F) { + f.Add("") + f.Add("\x00") + f.Add(string([]rune{unicode.ReplacementChar})) + + f.Fuzz(func(t *testing.T, s1 string) { + b := marshalCorpusFile(s1) + t.Logf("marshaled %q:\n%s", s1, b) + + rs, err := unmarshalCorpusFile(b) + if err != nil { + t.Fatal(err) + } + if len(rs) != 1 { + t.Fatalf("unmarshaled %d values", len(rs)) + } + s2 := rs[0].(string) + if s2 != s1 { + t.Errorf("unmarshaled %q", s2) + } + }) +} diff --git a/src/internal/fuzz/fuzz.go b/src/internal/fuzz/fuzz.go new file mode 100644 index 0000000..fb4e1d3 --- /dev/null +++ b/src/internal/fuzz/fuzz.go @@ -0,0 +1,1102 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package fuzz provides common fuzzing functionality for tests built with +// "go test" and for programs that use fuzzing functionality in the testing +// package. +package fuzz + +import ( + "bytes" + "context" + "crypto/sha256" + "errors" + "fmt" + "internal/godebug" + "io" + "math/bits" + "os" + "path/filepath" + "reflect" + "runtime" + "strings" + "time" +) + +// CoordinateFuzzingOpts is a set of arguments for CoordinateFuzzing. +// The zero value is valid for each field unless specified otherwise. +type CoordinateFuzzingOpts struct { + // Log is a writer for logging progress messages and warnings. + // If nil, io.Discard will be used instead. + Log io.Writer + + // Timeout is the amount of wall clock time to spend fuzzing after the corpus + // has loaded. If zero, there will be no time limit. + Timeout time.Duration + + // Limit is the number of random values to generate and test. If zero, + // there will be no limit on the number of generated values. + Limit int64 + + // MinimizeTimeout is the amount of wall clock time to spend minimizing + // after discovering a crasher. If zero, there will be no time limit. If + // MinimizeTimeout and MinimizeLimit are both zero, then minimization will + // be disabled. + MinimizeTimeout time.Duration + + // MinimizeLimit is the maximum number of calls to the fuzz function to be + // made while minimizing after finding a crash. If zero, there will be no + // limit. Calls to the fuzz function made when minimizing also count toward + // Limit. If MinimizeTimeout and MinimizeLimit are both zero, then + // minimization will be disabled. + MinimizeLimit int64 + + // parallel is the number of worker processes to run in parallel. If zero, + // CoordinateFuzzing will run GOMAXPROCS workers. + Parallel int + + // Seed is a list of seed values added by the fuzz target with testing.F.Add + // and in testdata. + Seed []CorpusEntry + + // Types is the list of types which make up a corpus entry. + // Types must be set and must match values in Seed. + Types []reflect.Type + + // CorpusDir is a directory where files containing values that crash the + // code being tested may be written. CorpusDir must be set. + CorpusDir string + + // CacheDir is a directory containing additional "interesting" values. + // The fuzzer may derive new values from these, and may write new values here. + CacheDir string +} + +// CoordinateFuzzing creates several worker processes and communicates with +// them to test random inputs that could trigger crashes and expose bugs. +// The worker processes run the same binary in the same directory with the +// same environment variables as the coordinator process. Workers also run +// with the same arguments as the coordinator, except with the -test.fuzzworker +// flag prepended to the argument list. +// +// If a crash occurs, the function will return an error containing information +// about the crash, which can be reported to the user. +func CoordinateFuzzing(ctx context.Context, opts CoordinateFuzzingOpts) (err error) { + if err := ctx.Err(); err != nil { + return err + } + if opts.Log == nil { + opts.Log = io.Discard + } + if opts.Parallel == 0 { + opts.Parallel = runtime.GOMAXPROCS(0) + } + if opts.Limit > 0 && int64(opts.Parallel) > opts.Limit { + // Don't start more workers than we need. + opts.Parallel = int(opts.Limit) + } + + c, err := newCoordinator(opts) + if err != nil { + return err + } + + if opts.Timeout > 0 { + var cancel func() + ctx, cancel = context.WithTimeout(ctx, opts.Timeout) + defer cancel() + } + + // fuzzCtx is used to stop workers, for example, after finding a crasher. + fuzzCtx, cancelWorkers := context.WithCancel(ctx) + defer cancelWorkers() + doneC := ctx.Done() + + // stop is called when a worker encounters a fatal error. + var fuzzErr error + stopping := false + stop := func(err error) { + if shouldPrintDebugInfo() { + _, file, line, ok := runtime.Caller(1) + if ok { + c.debugLogf("stop called at %s:%d. stopping: %t", file, line, stopping) + } else { + c.debugLogf("stop called at unknown. stopping: %t", stopping) + } + } + + if err == fuzzCtx.Err() || isInterruptError(err) { + // Suppress cancellation errors and terminations due to SIGINT. + // The messages are not helpful since either the user triggered the error + // (with ^C) or another more helpful message will be printed (a crasher). + err = nil + } + if err != nil && (fuzzErr == nil || fuzzErr == ctx.Err()) { + fuzzErr = err + } + if stopping { + return + } + stopping = true + cancelWorkers() + doneC = nil + } + + // Ensure that any crash we find is written to the corpus, even if an error + // or interruption occurs while minimizing it. + crashWritten := false + defer func() { + if c.crashMinimizing == nil || crashWritten { + return + } + werr := writeToCorpus(&c.crashMinimizing.entry, opts.CorpusDir) + if werr != nil { + err = fmt.Errorf("%w\n%v", err, werr) + return + } + if err == nil { + err = &crashError{ + path: c.crashMinimizing.entry.Path, + err: errors.New(c.crashMinimizing.crasherMsg), + } + } + }() + + // Start workers. + // TODO(jayconrod): do we want to support fuzzing different binaries? + dir := "" // same as self + binPath := os.Args[0] + args := append([]string{"-test.fuzzworker"}, os.Args[1:]...) + env := os.Environ() // same as self + + errC := make(chan error) + workers := make([]*worker, opts.Parallel) + for i := range workers { + var err error + workers[i], err = newWorker(c, dir, binPath, args, env) + if err != nil { + return err + } + } + for i := range workers { + w := workers[i] + go func() { + err := w.coordinate(fuzzCtx) + if fuzzCtx.Err() != nil || isInterruptError(err) { + err = nil + } + cleanErr := w.cleanup() + if err == nil { + err = cleanErr + } + errC <- err + }() + } + + // Main event loop. + // Do not return until all workers have terminated. We avoid a deadlock by + // receiving messages from workers even after ctx is cancelled. + activeWorkers := len(workers) + statTicker := time.NewTicker(3 * time.Second) + defer statTicker.Stop() + defer c.logStats() + + c.logStats() + for { + // If there is an execution limit, and we've reached it, stop. + if c.opts.Limit > 0 && c.count >= c.opts.Limit { + stop(nil) + } + + var inputC chan fuzzInput + input, ok := c.peekInput() + if ok && c.crashMinimizing == nil && !stopping { + inputC = c.inputC + } + + var minimizeC chan fuzzMinimizeInput + minimizeInput, ok := c.peekMinimizeInput() + if ok && !stopping { + minimizeC = c.minimizeC + } + + select { + case <-doneC: + // Interrupted, cancelled, or timed out. + // stop sets doneC to nil so we don't busy wait here. + stop(ctx.Err()) + + case err := <-errC: + // A worker terminated, possibly after encountering a fatal error. + stop(err) + activeWorkers-- + if activeWorkers == 0 { + return fuzzErr + } + + case result := <-c.resultC: + // Received response from worker. + if stopping { + break + } + c.updateStats(result) + + if result.crasherMsg != "" { + if c.warmupRun() && result.entry.IsSeed { + target := filepath.Base(c.opts.CorpusDir) + fmt.Fprintf(c.opts.Log, "failure while testing seed corpus entry: %s/%s\n", target, testName(result.entry.Parent)) + stop(errors.New(result.crasherMsg)) + break + } + if c.canMinimize() && result.canMinimize { + if c.crashMinimizing != nil { + // This crash is not minimized, and another crash is being minimized. + // Ignore this one and wait for the other one to finish. + if shouldPrintDebugInfo() { + c.debugLogf("found unminimized crasher, skipping in favor of minimizable crasher") + } + break + } + // Found a crasher but haven't yet attempted to minimize it. + // Send it back to a worker for minimization. Disable inputC so + // other workers don't continue fuzzing. + c.crashMinimizing = &result + fmt.Fprintf(c.opts.Log, "fuzz: minimizing %d-byte failing input file\n", len(result.entry.Data)) + c.queueForMinimization(result, nil) + } else if !crashWritten { + // Found a crasher that's either minimized or not minimizable. + // Write to corpus and stop. + err := writeToCorpus(&result.entry, opts.CorpusDir) + if err == nil { + crashWritten = true + err = &crashError{ + path: result.entry.Path, + err: errors.New(result.crasherMsg), + } + } + if shouldPrintDebugInfo() { + c.debugLogf( + "found crasher, id: %s, parent: %s, gen: %d, size: %d, exec time: %s", + result.entry.Path, + result.entry.Parent, + result.entry.Generation, + len(result.entry.Data), + result.entryDuration, + ) + } + stop(err) + } + } else if result.coverageData != nil { + if c.warmupRun() { + if shouldPrintDebugInfo() { + c.debugLogf( + "processed an initial input, id: %s, new bits: %d, size: %d, exec time: %s", + result.entry.Parent, + countBits(diffCoverage(c.coverageMask, result.coverageData)), + len(result.entry.Data), + result.entryDuration, + ) + } + c.updateCoverage(result.coverageData) + c.warmupInputLeft-- + if c.warmupInputLeft == 0 { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, gathering baseline coverage: %d/%d completed, now fuzzing with %d workers\n", c.elapsed(), c.warmupInputCount, c.warmupInputCount, c.opts.Parallel) + if shouldPrintDebugInfo() { + c.debugLogf( + "finished processing input corpus, entries: %d, initial coverage bits: %d", + len(c.corpus.entries), + countBits(c.coverageMask), + ) + } + } + } else if keepCoverage := diffCoverage(c.coverageMask, result.coverageData); keepCoverage != nil { + // Found a value that expanded coverage. + // It's not a crasher, but we may want to add it to the on-disk + // corpus and prioritize it for future fuzzing. + // TODO(jayconrod, katiehockman): Prioritize fuzzing these + // values which expanded coverage, perhaps based on the + // number of new edges that this result expanded. + // TODO(jayconrod, katiehockman): Don't write a value that's already + // in the corpus. + if c.canMinimize() && result.canMinimize && c.crashMinimizing == nil { + // Send back to workers to find a smaller value that preserves + // at least one new coverage bit. + c.queueForMinimization(result, keepCoverage) + } else { + // Update the coordinator's coverage mask and save the value. + inputSize := len(result.entry.Data) + entryNew, err := c.addCorpusEntries(true, result.entry) + if err != nil { + stop(err) + break + } + if !entryNew { + if shouldPrintDebugInfo() { + c.debugLogf( + "ignoring duplicate input which increased coverage, id: %s", + result.entry.Path, + ) + } + break + } + c.updateCoverage(keepCoverage) + c.inputQueue.enqueue(result.entry) + c.interestingCount++ + if shouldPrintDebugInfo() { + c.debugLogf( + "new interesting input, id: %s, parent: %s, gen: %d, new bits: %d, total bits: %d, size: %d, exec time: %s", + result.entry.Path, + result.entry.Parent, + result.entry.Generation, + countBits(keepCoverage), + countBits(c.coverageMask), + inputSize, + result.entryDuration, + ) + } + } + } else { + if shouldPrintDebugInfo() { + c.debugLogf( + "worker reported interesting input that doesn't expand coverage, id: %s, parent: %s, canMinimize: %t", + result.entry.Path, + result.entry.Parent, + result.canMinimize, + ) + } + } + } else if c.warmupRun() { + // No error or coverage data was reported for this input during + // warmup, so continue processing results. + c.warmupInputLeft-- + if c.warmupInputLeft == 0 { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, testing seed corpus: %d/%d completed, now fuzzing with %d workers\n", c.elapsed(), c.warmupInputCount, c.warmupInputCount, c.opts.Parallel) + if shouldPrintDebugInfo() { + c.debugLogf( + "finished testing-only phase, entries: %d", + len(c.corpus.entries), + ) + } + } + } + + case inputC <- input: + // Sent the next input to a worker. + c.sentInput(input) + + case minimizeC <- minimizeInput: + // Sent the next input for minimization to a worker. + c.sentMinimizeInput(minimizeInput) + + case <-statTicker.C: + c.logStats() + } + } + + // TODO(jayconrod,katiehockman): if a crasher can't be written to the corpus, + // write to the cache instead. +} + +// crashError wraps a crasher written to the seed corpus. It saves the name +// of the file where the input causing the crasher was saved. The testing +// framework uses this to report a command to re-run that specific input. +type crashError struct { + path string + err error +} + +func (e *crashError) Error() string { + return e.err.Error() +} + +func (e *crashError) Unwrap() error { + return e.err +} + +func (e *crashError) CrashPath() string { + return e.path +} + +type corpus struct { + entries []CorpusEntry + hashes map[[sha256.Size]byte]bool +} + +// addCorpusEntries adds entries to the corpus, and optionally writes the entries +// to the cache directory. If an entry is already in the corpus it is skipped. If +// all of the entries are unique, addCorpusEntries returns true and a nil error, +// if at least one of the entries was a duplicate, it returns false and a nil error. +func (c *coordinator) addCorpusEntries(addToCache bool, entries ...CorpusEntry) (bool, error) { + noDupes := true + for _, e := range entries { + data, err := corpusEntryData(e) + if err != nil { + return false, err + } + h := sha256.Sum256(data) + if c.corpus.hashes[h] { + noDupes = false + continue + } + if addToCache { + if err := writeToCorpus(&e, c.opts.CacheDir); err != nil { + return false, err + } + // For entries written to disk, we don't hold onto the bytes, + // since the corpus would consume a significant amount of + // memory. + e.Data = nil + } + c.corpus.hashes[h] = true + c.corpus.entries = append(c.corpus.entries, e) + } + return noDupes, nil +} + +// CorpusEntry represents an individual input for fuzzing. +// +// We must use an equivalent type in the testing and testing/internal/testdeps +// packages, but testing can't import this package directly, and we don't want +// to export this type from testing. Instead, we use the same struct type and +// use a type alias (not a defined type) for convenience. +type CorpusEntry = struct { + Parent string + + // Path is the path of the corpus file, if the entry was loaded from disk. + // For other entries, including seed values provided by f.Add, Path is the + // name of the test, e.g. seed#0 or its hash. + Path string + + // Data is the raw input data. Data should only be populated for seed + // values. For on-disk corpus files, Data will be nil, as it will be loaded + // from disk using Path. + Data []byte + + // Values is the unmarshaled values from a corpus file. + Values []any + + Generation int + + // IsSeed indicates whether this entry is part of the seed corpus. + IsSeed bool +} + +// corpusEntryData returns the raw input bytes, either from the data struct +// field, or from disk. +func corpusEntryData(ce CorpusEntry) ([]byte, error) { + if ce.Data != nil { + return ce.Data, nil + } + + return os.ReadFile(ce.Path) +} + +type fuzzInput struct { + // entry is the value to test initially. The worker will randomly mutate + // values from this starting point. + entry CorpusEntry + + // timeout is the time to spend fuzzing variations of this input, + // not including starting or cleaning up. + timeout time.Duration + + // limit is the maximum number of calls to the fuzz function the worker may + // make. The worker may make fewer calls, for example, if it finds an + // error early. If limit is zero, there is no limit on calls to the + // fuzz function. + limit int64 + + // warmup indicates whether this is a warmup input before fuzzing begins. If + // true, the input should not be fuzzed. + warmup bool + + // coverageData reflects the coordinator's current coverageMask. + coverageData []byte +} + +type fuzzResult struct { + // entry is an interesting value or a crasher. + entry CorpusEntry + + // crasherMsg is an error message from a crash. It's "" if no crash was found. + crasherMsg string + + // canMinimize is true if the worker should attempt to minimize this result. + // It may be false because an attempt has already been made. + canMinimize bool + + // coverageData is set if the worker found new coverage. + coverageData []byte + + // limit is the number of values the coordinator asked the worker + // to test. 0 if there was no limit. + limit int64 + + // count is the number of values the worker actually tested. + count int64 + + // totalDuration is the time the worker spent testing inputs. + totalDuration time.Duration + + // entryDuration is the time the worker spent execution an interesting result + entryDuration time.Duration +} + +type fuzzMinimizeInput struct { + // entry is an interesting value or crasher to minimize. + entry CorpusEntry + + // crasherMsg is an error message from a crash. It's "" if no crash was found. + // If set, the worker will attempt to find a smaller input that also produces + // an error, though not necessarily the same error. + crasherMsg string + + // limit is the maximum number of calls to the fuzz function the worker may + // make. The worker may make fewer calls, for example, if it can't reproduce + // an error. If limit is zero, there is no limit on calls to the fuzz function. + limit int64 + + // timeout is the time to spend minimizing this input. + // A zero timeout means no limit. + timeout time.Duration + + // keepCoverage is a set of coverage bits that entry found that were not in + // the coordinator's combined set. When minimizing, the worker should find an + // input that preserves at least one of these bits. keepCoverage is nil for + // crashing inputs. + keepCoverage []byte +} + +// coordinator holds channels that workers can use to communicate with +// the coordinator. +type coordinator struct { + opts CoordinateFuzzingOpts + + // startTime is the time we started the workers after loading the corpus. + // Used for logging. + startTime time.Time + + // inputC is sent values to fuzz by the coordinator. Any worker may receive + // values from this channel. Workers send results to resultC. + inputC chan fuzzInput + + // minimizeC is sent values to minimize by the coordinator. Any worker may + // receive values from this channel. Workers send results to resultC. + minimizeC chan fuzzMinimizeInput + + // resultC is sent results of fuzzing by workers. The coordinator + // receives these. Multiple types of messages are allowed. + resultC chan fuzzResult + + // count is the number of values fuzzed so far. + count int64 + + // countLastLog is the number of values fuzzed when the output was last + // logged. + countLastLog int64 + + // timeLastLog is the time at which the output was last logged. + timeLastLog time.Time + + // interestingCount is the number of unique interesting values which have + // been found this execution. + interestingCount int + + // warmupInputCount is the count of all entries in the corpus which will + // need to be received from workers to run once during warmup, but not fuzz. + // This could be for coverage data, or only for the purposes of verifying + // that the seed corpus doesn't have any crashers. See warmupRun. + warmupInputCount int + + // warmupInputLeft is the number of entries in the corpus which still need + // to be received from workers to run once during warmup, but not fuzz. + // See warmupInputLeft. + warmupInputLeft int + + // duration is the time spent fuzzing inside workers, not counting time + // starting up or tearing down. + duration time.Duration + + // countWaiting is the number of fuzzing executions the coordinator is + // waiting on workers to complete. + countWaiting int64 + + // corpus is a set of interesting values, including the seed corpus and + // generated values that workers reported as interesting. + corpus corpus + + // minimizationAllowed is true if one or more of the types of fuzz + // function's parameters can be minimized. + minimizationAllowed bool + + // inputQueue is a queue of inputs that workers should try fuzzing. This is + // initially populated from the seed corpus and cached inputs. More inputs + // may be added as new coverage is discovered. + inputQueue queue + + // minimizeQueue is a queue of inputs that caused errors or exposed new + // coverage. Workers should attempt to find smaller inputs that do the + // same thing. + minimizeQueue queue + + // crashMinimizing is the crash that is currently being minimized. + crashMinimizing *fuzzResult + + // coverageMask aggregates coverage that was found for all inputs in the + // corpus. Each byte represents a single basic execution block. Each set bit + // within the byte indicates that an input has triggered that block at least + // 1 << n times, where n is the position of the bit in the byte. For example, a + // value of 12 indicates that separate inputs have triggered this block + // between 4-7 times and 8-15 times. + coverageMask []byte +} + +func newCoordinator(opts CoordinateFuzzingOpts) (*coordinator, error) { + // Make sure all of the seed corpus has marshalled data. + for i := range opts.Seed { + if opts.Seed[i].Data == nil && opts.Seed[i].Values != nil { + opts.Seed[i].Data = marshalCorpusFile(opts.Seed[i].Values...) + } + } + c := &coordinator{ + opts: opts, + startTime: time.Now(), + inputC: make(chan fuzzInput), + minimizeC: make(chan fuzzMinimizeInput), + resultC: make(chan fuzzResult), + timeLastLog: time.Now(), + corpus: corpus{hashes: make(map[[sha256.Size]byte]bool)}, + } + if err := c.readCache(); err != nil { + return nil, err + } + if opts.MinimizeLimit > 0 || opts.MinimizeTimeout > 0 { + for _, t := range opts.Types { + if isMinimizable(t) { + c.minimizationAllowed = true + break + } + } + } + + covSize := len(coverage()) + if covSize == 0 { + fmt.Fprintf(c.opts.Log, "warning: the test binary was not built with coverage instrumentation, so fuzzing will run without coverage guidance and may be inefficient\n") + // Even though a coverage-only run won't occur, we should still run all + // of the seed corpus to make sure there are no existing failures before + // we start fuzzing. + c.warmupInputCount = len(c.opts.Seed) + for _, e := range c.opts.Seed { + c.inputQueue.enqueue(e) + } + } else { + c.warmupInputCount = len(c.corpus.entries) + for _, e := range c.corpus.entries { + c.inputQueue.enqueue(e) + } + // Set c.coverageMask to a clean []byte full of zeros. + c.coverageMask = make([]byte, covSize) + } + c.warmupInputLeft = c.warmupInputCount + + if len(c.corpus.entries) == 0 { + fmt.Fprintf(c.opts.Log, "warning: starting with empty corpus\n") + var vals []any + for _, t := range opts.Types { + vals = append(vals, zeroValue(t)) + } + data := marshalCorpusFile(vals...) + h := sha256.Sum256(data) + name := fmt.Sprintf("%x", h[:4]) + c.addCorpusEntries(false, CorpusEntry{Path: name, Data: data}) + } + + return c, nil +} + +func (c *coordinator) updateStats(result fuzzResult) { + c.count += result.count + c.countWaiting -= result.limit + c.duration += result.totalDuration +} + +func (c *coordinator) logStats() { + now := time.Now() + if c.warmupRun() { + runSoFar := c.warmupInputCount - c.warmupInputLeft + if coverageEnabled { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, gathering baseline coverage: %d/%d completed\n", c.elapsed(), runSoFar, c.warmupInputCount) + } else { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, testing seed corpus: %d/%d completed\n", c.elapsed(), runSoFar, c.warmupInputCount) + } + } else if c.crashMinimizing != nil { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, minimizing\n", c.elapsed()) + } else { + rate := float64(c.count-c.countLastLog) / now.Sub(c.timeLastLog).Seconds() + if coverageEnabled { + total := c.warmupInputCount + c.interestingCount + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, execs: %d (%.0f/sec), new interesting: %d (total: %d)\n", c.elapsed(), c.count, rate, c.interestingCount, total) + } else { + fmt.Fprintf(c.opts.Log, "fuzz: elapsed: %s, execs: %d (%.0f/sec)\n", c.elapsed(), c.count, rate) + } + } + c.countLastLog = c.count + c.timeLastLog = now +} + +// peekInput returns the next value that should be sent to workers. +// If the number of executions is limited, the returned value includes +// a limit for one worker. If there are no executions left, peekInput returns +// a zero value and false. +// +// peekInput doesn't actually remove the input from the queue. The caller +// must call sentInput after sending the input. +// +// If the input queue is empty and the coverage/testing-only run has completed, +// queue refills it from the corpus. +func (c *coordinator) peekInput() (fuzzInput, bool) { + if c.opts.Limit > 0 && c.count+c.countWaiting >= c.opts.Limit { + // Already making the maximum number of calls to the fuzz function. + // Don't send more inputs right now. + return fuzzInput{}, false + } + if c.inputQueue.len == 0 { + if c.warmupRun() { + // Wait for coverage/testing-only run to finish before sending more + // inputs. + return fuzzInput{}, false + } + c.refillInputQueue() + } + + entry, ok := c.inputQueue.peek() + if !ok { + panic("input queue empty after refill") + } + input := fuzzInput{ + entry: entry.(CorpusEntry), + timeout: workerFuzzDuration, + warmup: c.warmupRun(), + } + if c.coverageMask != nil { + input.coverageData = bytes.Clone(c.coverageMask) + } + if input.warmup { + // No fuzzing will occur, but it should count toward the limit set by + // -fuzztime. + input.limit = 1 + return input, true + } + + if c.opts.Limit > 0 { + input.limit = c.opts.Limit / int64(c.opts.Parallel) + if c.opts.Limit%int64(c.opts.Parallel) > 0 { + input.limit++ + } + remaining := c.opts.Limit - c.count - c.countWaiting + if input.limit > remaining { + input.limit = remaining + } + } + return input, true +} + +// sentInput updates internal counters after an input is sent to c.inputC. +func (c *coordinator) sentInput(input fuzzInput) { + c.inputQueue.dequeue() + c.countWaiting += input.limit +} + +// refillInputQueue refills the input queue from the corpus after it becomes +// empty. +func (c *coordinator) refillInputQueue() { + for _, e := range c.corpus.entries { + c.inputQueue.enqueue(e) + } +} + +// queueForMinimization creates a fuzzMinimizeInput from result and adds it +// to the minimization queue to be sent to workers. +func (c *coordinator) queueForMinimization(result fuzzResult, keepCoverage []byte) { + if shouldPrintDebugInfo() { + c.debugLogf( + "queueing input for minimization, id: %s, parent: %s, keepCoverage: %t, crasher: %t", + result.entry.Path, + result.entry.Parent, + keepCoverage != nil, + result.crasherMsg != "", + ) + } + if result.crasherMsg != "" { + c.minimizeQueue.clear() + } + + input := fuzzMinimizeInput{ + entry: result.entry, + crasherMsg: result.crasherMsg, + keepCoverage: keepCoverage, + } + c.minimizeQueue.enqueue(input) +} + +// peekMinimizeInput returns the next input that should be sent to workers for +// minimization. +func (c *coordinator) peekMinimizeInput() (fuzzMinimizeInput, bool) { + if !c.canMinimize() { + // Already making the maximum number of calls to the fuzz function. + // Don't send more inputs right now. + return fuzzMinimizeInput{}, false + } + v, ok := c.minimizeQueue.peek() + if !ok { + return fuzzMinimizeInput{}, false + } + input := v.(fuzzMinimizeInput) + + if c.opts.MinimizeTimeout > 0 { + input.timeout = c.opts.MinimizeTimeout + } + if c.opts.MinimizeLimit > 0 { + input.limit = c.opts.MinimizeLimit + } else if c.opts.Limit > 0 { + if input.crasherMsg != "" { + input.limit = c.opts.Limit + } else { + input.limit = c.opts.Limit / int64(c.opts.Parallel) + if c.opts.Limit%int64(c.opts.Parallel) > 0 { + input.limit++ + } + } + } + if c.opts.Limit > 0 { + remaining := c.opts.Limit - c.count - c.countWaiting + if input.limit > remaining { + input.limit = remaining + } + } + return input, true +} + +// sentMinimizeInput removes an input from the minimization queue after it's +// sent to minimizeC. +func (c *coordinator) sentMinimizeInput(input fuzzMinimizeInput) { + c.minimizeQueue.dequeue() + c.countWaiting += input.limit +} + +// warmupRun returns true while the coordinator is running inputs without +// mutating them as a warmup before fuzzing. This could be to gather baseline +// coverage data for entries in the corpus, or to test all of the seed corpus +// for errors before fuzzing begins. +// +// The coordinator doesn't store coverage data in the cache with each input +// because that data would be invalid when counter offsets in the test binary +// change. +// +// When gathering coverage, the coordinator sends each entry to a worker to +// gather coverage for that entry only, without fuzzing or minimizing. This +// phase ends when all workers have finished, and the coordinator has a combined +// coverage map. +func (c *coordinator) warmupRun() bool { + return c.warmupInputLeft > 0 +} + +// updateCoverage sets bits in c.coverageMask that are set in newCoverage. +// updateCoverage returns the number of newly set bits. See the comment on +// coverageMask for the format. +func (c *coordinator) updateCoverage(newCoverage []byte) int { + if len(newCoverage) != len(c.coverageMask) { + panic(fmt.Sprintf("number of coverage counters changed at runtime: %d, expected %d", len(newCoverage), len(c.coverageMask))) + } + newBitCount := 0 + for i := range newCoverage { + diff := newCoverage[i] &^ c.coverageMask[i] + newBitCount += bits.OnesCount8(diff) + c.coverageMask[i] |= newCoverage[i] + } + return newBitCount +} + +// canMinimize returns whether the coordinator should attempt to find smaller +// inputs that reproduce a crash or new coverage. +func (c *coordinator) canMinimize() bool { + return c.minimizationAllowed && + (c.opts.Limit == 0 || c.count+c.countWaiting < c.opts.Limit) +} + +func (c *coordinator) elapsed() time.Duration { + return time.Since(c.startTime).Round(1 * time.Second) +} + +// readCache creates a combined corpus from seed values and values in the cache +// (in GOCACHE/fuzz). +// +// TODO(fuzzing): need a mechanism that can remove values that +// aren't useful anymore, for example, because they have the wrong type. +func (c *coordinator) readCache() error { + if _, err := c.addCorpusEntries(false, c.opts.Seed...); err != nil { + return err + } + entries, err := ReadCorpus(c.opts.CacheDir, c.opts.Types) + if err != nil { + if _, ok := err.(*MalformedCorpusError); !ok { + // It's okay if some files in the cache directory are malformed and + // are not included in the corpus, but fail if it's an I/O error. + return err + } + // TODO(jayconrod,katiehockman): consider printing some kind of warning + // indicating the number of files which were skipped because they are + // malformed. + } + if _, err := c.addCorpusEntries(false, entries...); err != nil { + return err + } + return nil +} + +// MalformedCorpusError is an error found while reading the corpus from the +// filesystem. All of the errors are stored in the errs list. The testing +// framework uses this to report malformed files in testdata. +type MalformedCorpusError struct { + errs []error +} + +func (e *MalformedCorpusError) Error() string { + var msgs []string + for _, s := range e.errs { + msgs = append(msgs, s.Error()) + } + return strings.Join(msgs, "\n") +} + +// ReadCorpus reads the corpus from the provided dir. The returned corpus +// entries are guaranteed to match the given types. Any malformed files will +// be saved in a MalformedCorpusError and returned, along with the most recent +// error. +func ReadCorpus(dir string, types []reflect.Type) ([]CorpusEntry, error) { + files, err := os.ReadDir(dir) + if os.IsNotExist(err) { + return nil, nil // No corpus to read + } else if err != nil { + return nil, fmt.Errorf("reading seed corpus from testdata: %v", err) + } + var corpus []CorpusEntry + var errs []error + for _, file := range files { + // TODO(jayconrod,katiehockman): determine when a file is a fuzzing input + // based on its name. We should only read files created by writeToCorpus. + // If we read ALL files, we won't be able to change the file format by + // changing the extension. We also won't be able to add files like + // README.txt explaining why the directory exists. + if file.IsDir() { + continue + } + filename := filepath.Join(dir, file.Name()) + data, err := os.ReadFile(filename) + if err != nil { + return nil, fmt.Errorf("failed to read corpus file: %v", err) + } + var vals []any + vals, err = readCorpusData(data, types) + if err != nil { + errs = append(errs, fmt.Errorf("%q: %v", filename, err)) + continue + } + corpus = append(corpus, CorpusEntry{Path: filename, Values: vals}) + } + if len(errs) > 0 { + return corpus, &MalformedCorpusError{errs: errs} + } + return corpus, nil +} + +func readCorpusData(data []byte, types []reflect.Type) ([]any, error) { + vals, err := unmarshalCorpusFile(data) + if err != nil { + return nil, fmt.Errorf("unmarshal: %v", err) + } + if err = CheckCorpus(vals, types); err != nil { + return nil, err + } + return vals, nil +} + +// CheckCorpus verifies that the types in vals match the expected types +// provided. +func CheckCorpus(vals []any, types []reflect.Type) error { + if len(vals) != len(types) { + return fmt.Errorf("wrong number of values in corpus entry: %d, want %d", len(vals), len(types)) + } + valsT := make([]reflect.Type, len(vals)) + for valsI, v := range vals { + valsT[valsI] = reflect.TypeOf(v) + } + for i := range types { + if valsT[i] != types[i] { + return fmt.Errorf("mismatched types in corpus entry: %v, want %v", valsT, types) + } + } + return nil +} + +// writeToCorpus atomically writes the given bytes to a new file in testdata. If +// the directory does not exist, it will create one. If the file already exists, +// writeToCorpus will not rewrite it. writeToCorpus sets entry.Path to the new +// file that was just written or an error if it failed. +func writeToCorpus(entry *CorpusEntry, dir string) (err error) { + sum := fmt.Sprintf("%x", sha256.Sum256(entry.Data))[:16] + entry.Path = filepath.Join(dir, sum) + if err := os.MkdirAll(dir, 0777); err != nil { + return err + } + if err := os.WriteFile(entry.Path, entry.Data, 0666); err != nil { + os.Remove(entry.Path) // remove partially written file + return err + } + return nil +} + +func testName(path string) string { + return filepath.Base(path) +} + +func zeroValue(t reflect.Type) any { + for _, v := range zeroVals { + if reflect.TypeOf(v) == t { + return v + } + } + panic(fmt.Sprintf("unsupported type: %v", t)) +} + +var zeroVals []any = []any{ + []byte(""), + string(""), + false, + byte(0), + rune(0), + float32(0), + float64(0), + int(0), + int8(0), + int16(0), + int32(0), + int64(0), + uint(0), + uint8(0), + uint16(0), + uint32(0), + uint64(0), +} + +var debugInfo = godebug.New("#fuzzdebug").Value() == "1" + +func shouldPrintDebugInfo() bool { + return debugInfo +} + +func (c *coordinator) debugLogf(format string, args ...any) { + t := time.Now().Format("2006-01-02 15:04:05.999999999") + fmt.Fprintf(c.opts.Log, t+" DEBUG "+format+"\n", args...) +} diff --git a/src/internal/fuzz/mem.go b/src/internal/fuzz/mem.go new file mode 100644 index 0000000..4155e4e --- /dev/null +++ b/src/internal/fuzz/mem.go @@ -0,0 +1,138 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "bytes" + "fmt" + "os" + "unsafe" +) + +// sharedMem manages access to a region of virtual memory mapped from a file, +// shared between multiple processes. The region includes space for a header and +// a value of variable length. +// +// When fuzzing, the coordinator creates a sharedMem from a temporary file for +// each worker. This buffer is used to pass values to fuzz between processes. +// Care must be taken to manage access to shared memory across processes; +// sharedMem provides no synchronization on its own. See workerComm for an +// explanation. +type sharedMem struct { + // f is the file mapped into memory. + f *os.File + + // region is the mapped region of virtual memory for f. The content of f may + // be read or written through this slice. + region []byte + + // removeOnClose is true if the file should be deleted by Close. + removeOnClose bool + + // sys contains OS-specific information. + sys sharedMemSys +} + +// sharedMemHeader stores metadata in shared memory. +type sharedMemHeader struct { + // count is the number of times the worker has called the fuzz function. + // May be reset by coordinator. + count int64 + + // valueLen is the number of bytes in region which should be read. + valueLen int + + // randState and randInc hold the state of a pseudo-random number generator. + randState, randInc uint64 + + // rawInMem is true if the region holds raw bytes, which occurs during + // minimization. If true after the worker fails during minimization, this + // indicates that an unrecoverable error occurred, and the region can be + // used to retrieve the raw bytes that caused the error. + rawInMem bool +} + +// sharedMemSize returns the size needed for a shared memory buffer that can +// contain values of the given size. +func sharedMemSize(valueSize int) int { + // TODO(jayconrod): set a reasonable maximum size per platform. + return int(unsafe.Sizeof(sharedMemHeader{})) + valueSize +} + +// sharedMemTempFile creates a new temporary file of the given size, then maps +// it into memory. The file will be removed when the Close method is called. +func sharedMemTempFile(size int) (m *sharedMem, err error) { + // Create a temporary file. + f, err := os.CreateTemp("", "fuzz-*") + if err != nil { + return nil, err + } + defer func() { + if err != nil { + f.Close() + os.Remove(f.Name()) + } + }() + + // Resize it to the correct size. + totalSize := sharedMemSize(size) + if err := f.Truncate(int64(totalSize)); err != nil { + return nil, err + } + + // Map the file into memory. + removeOnClose := true + return sharedMemMapFile(f, totalSize, removeOnClose) +} + +// header returns a pointer to metadata within the shared memory region. +func (m *sharedMem) header() *sharedMemHeader { + return (*sharedMemHeader)(unsafe.Pointer(&m.region[0])) +} + +// valueRef returns the value currently stored in shared memory. The returned +// slice points to shared memory; it is not a copy. +func (m *sharedMem) valueRef() []byte { + length := m.header().valueLen + valueOffset := int(unsafe.Sizeof(sharedMemHeader{})) + return m.region[valueOffset : valueOffset+length] +} + +// valueCopy returns a copy of the value stored in shared memory. +func (m *sharedMem) valueCopy() []byte { + ref := m.valueRef() + return bytes.Clone(ref) +} + +// setValue copies the data in b into the shared memory buffer and sets +// the length. len(b) must be less than or equal to the capacity of the buffer +// (as returned by cap(m.value())). +func (m *sharedMem) setValue(b []byte) { + v := m.valueRef() + if len(b) > cap(v) { + panic(fmt.Sprintf("value length %d larger than shared memory capacity %d", len(b), cap(v))) + } + m.header().valueLen = len(b) + copy(v[:cap(v)], b) +} + +// setValueLen sets the length of the shared memory buffer returned by valueRef +// to n, which may be at most the cap of that slice. +// +// Note that we can only store the length in the shared memory header. The full +// slice header contains a pointer, which is likely only valid for one process, +// since each process can map shared memory at a different virtual address. +func (m *sharedMem) setValueLen(n int) { + v := m.valueRef() + if n > cap(v) { + panic(fmt.Sprintf("length %d larger than shared memory capacity %d", n, cap(v))) + } + m.header().valueLen = n +} + +// TODO(jayconrod): add method to resize the buffer. We'll need that when the +// mutator can increase input length. Only the coordinator will be able to +// do it, since we'll need to send a message to the worker telling it to +// remap the file. diff --git a/src/internal/fuzz/minimize.go b/src/internal/fuzz/minimize.go new file mode 100644 index 0000000..0e410fb --- /dev/null +++ b/src/internal/fuzz/minimize.go @@ -0,0 +1,95 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "reflect" +) + +func isMinimizable(t reflect.Type) bool { + return t == reflect.TypeOf("") || t == reflect.TypeOf([]byte(nil)) +} + +func minimizeBytes(v []byte, try func([]byte) bool, shouldStop func() bool) { + tmp := make([]byte, len(v)) + // If minimization was successful at any point during minimizeBytes, + // then the vals slice in (*workerServer).minimizeInput will point to + // tmp. Since tmp is altered while making new candidates, we need to + // make sure that it is equal to the correct value, v, before exiting + // this function. + defer copy(tmp, v) + + // First, try to cut the tail. + for n := 1024; n != 0; n /= 2 { + for len(v) > n { + if shouldStop() { + return + } + candidate := v[:len(v)-n] + if !try(candidate) { + break + } + // Set v to the new value to continue iterating. + v = candidate + } + } + + // Then, try to remove each individual byte. + for i := 0; i < len(v)-1; i++ { + if shouldStop() { + return + } + candidate := tmp[:len(v)-1] + copy(candidate[:i], v[:i]) + copy(candidate[i:], v[i+1:]) + if !try(candidate) { + continue + } + // Update v to delete the value at index i. + copy(v[i:], v[i+1:]) + v = v[:len(candidate)] + // v[i] is now different, so decrement i to redo this iteration + // of the loop with the new value. + i-- + } + + // Then, try to remove each possible subset of bytes. + for i := 0; i < len(v)-1; i++ { + copy(tmp, v[:i]) + for j := len(v); j > i+1; j-- { + if shouldStop() { + return + } + candidate := tmp[:len(v)-j+i] + copy(candidate[i:], v[j:]) + if !try(candidate) { + continue + } + // Update v and reset the loop with the new length. + copy(v[i:], v[j:]) + v = v[:len(candidate)] + j = len(v) + } + } + + // Then, try to make it more simplified and human-readable by trying to replace each + // byte with a printable character. + printableChars := []byte("012789ABCXYZabcxyz !\"#$%&'()*+,.") + for i, b := range v { + if shouldStop() { + return + } + + for _, pc := range printableChars { + v[i] = pc + if try(v) { + // Successful. Move on to the next byte in v. + break + } + // Unsuccessful. Revert v[i] back to original value. + v[i] = b + } + } +} diff --git a/src/internal/fuzz/minimize_test.go b/src/internal/fuzz/minimize_test.go new file mode 100644 index 0000000..2db2633 --- /dev/null +++ b/src/internal/fuzz/minimize_test.go @@ -0,0 +1,182 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build darwin || freebsd || linux || windows + +package fuzz + +import ( + "bytes" + "context" + "errors" + "fmt" + "reflect" + "testing" + "time" + "unicode" + "unicode/utf8" +) + +func TestMinimizeInput(t *testing.T) { + type testcase struct { + name string + fn func(CorpusEntry) error + input []any + expected []any + } + cases := []testcase{ + { + name: "ones_byte", + fn: func(e CorpusEntry) error { + b := e.Values[0].([]byte) + ones := 0 + for _, v := range b { + if v == 1 { + ones++ + } + } + if ones == 3 { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{[]byte{0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}}, + expected: []any{[]byte{1, 1, 1}}, + }, + { + name: "single_bytes", + fn: func(e CorpusEntry) error { + b := e.Values[0].([]byte) + if len(b) < 2 { + return nil + } + if len(b) == 2 && b[0] == 1 && b[1] == 2 { + return nil + } + return fmt.Errorf("bad %v", e.Values[0]) + }, + input: []any{[]byte{1, 2, 3, 4, 5}}, + expected: []any{[]byte("00")}, + }, + { + name: "set_of_bytes", + fn: func(e CorpusEntry) error { + b := e.Values[0].([]byte) + if len(b) < 3 { + return nil + } + if bytes.Equal(b, []byte{0, 1, 2, 3, 4, 5}) || bytes.Equal(b, []byte{0, 4, 5}) { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{[]byte{0, 1, 2, 3, 4, 5}}, + expected: []any{[]byte{0, 4, 5}}, + }, + { + name: "non_ascii_bytes", + fn: func(e CorpusEntry) error { + b := e.Values[0].([]byte) + if len(b) == 3 { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{[]byte("ท")}, // ท is 3 bytes + expected: []any{[]byte("000")}, + }, + { + name: "ones_string", + fn: func(e CorpusEntry) error { + b := e.Values[0].(string) + ones := 0 + for _, v := range b { + if v == '1' { + ones++ + } + } + if ones == 3 { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{"001010001000000000000000000"}, + expected: []any{"111"}, + }, + { + name: "string_length", + fn: func(e CorpusEntry) error { + b := e.Values[0].(string) + if len(b) == 5 { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{"zzzzz"}, + expected: []any{"00000"}, + }, + { + name: "string_with_letter", + fn: func(e CorpusEntry) error { + b := e.Values[0].(string) + r, _ := utf8.DecodeRune([]byte(b)) + if unicode.IsLetter(r) { + return fmt.Errorf("bad %v", e.Values[0]) + } + return nil + }, + input: []any{"ZZZZZ"}, + expected: []any{"A"}, + }, + } + + for _, tc := range cases { + tc := tc + t.Run(tc.name, func(t *testing.T) { + t.Parallel() + ws := &workerServer{ + fuzzFn: func(e CorpusEntry) (time.Duration, error) { + return time.Second, tc.fn(e) + }, + } + mem := &sharedMem{region: make([]byte, 100)} // big enough to hold value and header + vals := tc.input + success, err := ws.minimizeInput(context.Background(), vals, mem, minimizeArgs{}) + if !success { + t.Errorf("minimizeInput did not succeed") + } + if err == nil { + t.Fatal("minimizeInput didn't provide an error") + } + if expected := fmt.Sprintf("bad %v", tc.expected[0]); err.Error() != expected { + t.Errorf("unexpected error: got %q, want %q", err, expected) + } + if !reflect.DeepEqual(vals, tc.expected) { + t.Errorf("unexpected results: got %v, want %v", vals, tc.expected) + } + }) + } +} + +// TestMinimizeFlaky checks that if we're minimizing an interesting +// input and a flaky failure occurs, that minimization was not indicated +// to be successful, and the error isn't returned (since it's flaky). +func TestMinimizeFlaky(t *testing.T) { + ws := &workerServer{fuzzFn: func(e CorpusEntry) (time.Duration, error) { + return time.Second, errors.New("ohno") + }} + mem := &sharedMem{region: make([]byte, 100)} // big enough to hold value and header + vals := []any{[]byte(nil)} + args := minimizeArgs{KeepCoverage: make([]byte, len(coverageSnapshot))} + success, err := ws.minimizeInput(context.Background(), vals, mem, args) + if success { + t.Error("unexpected success") + } + if err != nil { + t.Errorf("unexpected error: %v", err) + } + if count := mem.header().count; count != 1 { + t.Errorf("count: got %d, want 1", count) + } +} diff --git a/src/internal/fuzz/mutator.go b/src/internal/fuzz/mutator.go new file mode 100644 index 0000000..bb96066 --- /dev/null +++ b/src/internal/fuzz/mutator.go @@ -0,0 +1,300 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "encoding/binary" + "fmt" + "math" + "unsafe" +) + +type mutator struct { + r mutatorRand + scratch []byte // scratch slice to avoid additional allocations +} + +func newMutator() *mutator { + return &mutator{r: newPcgRand()} +} + +func (m *mutator) rand(n int) int { + return m.r.intn(n) +} + +func (m *mutator) randByteOrder() binary.ByteOrder { + if m.r.bool() { + return binary.LittleEndian + } + return binary.BigEndian +} + +// chooseLen chooses length of range mutation in range [1,n]. It gives +// preference to shorter ranges. +func (m *mutator) chooseLen(n int) int { + switch x := m.rand(100); { + case x < 90: + return m.rand(min(8, n)) + 1 + case x < 99: + return m.rand(min(32, n)) + 1 + default: + return m.rand(n) + 1 + } +} + +func min(a, b int) int { + if a < b { + return a + } + return b +} + +// mutate performs several mutations on the provided values. +func (m *mutator) mutate(vals []any, maxBytes int) { + // TODO(katiehockman): pull some of these functions into helper methods and + // test that each case is working as expected. + // TODO(katiehockman): perform more types of mutations for []byte. + + // maxPerVal will represent the maximum number of bytes that each value be + // allowed after mutating, giving an equal amount of capacity to each line. + // Allow a little wiggle room for the encoding. + maxPerVal := maxBytes/len(vals) - 100 + + // Pick a random value to mutate. + // TODO: consider mutating more than one value at a time. + i := m.rand(len(vals)) + switch v := vals[i].(type) { + case int: + vals[i] = int(m.mutateInt(int64(v), maxInt)) + case int8: + vals[i] = int8(m.mutateInt(int64(v), math.MaxInt8)) + case int16: + vals[i] = int16(m.mutateInt(int64(v), math.MaxInt16)) + case int64: + vals[i] = m.mutateInt(v, maxInt) + case uint: + vals[i] = uint(m.mutateUInt(uint64(v), maxUint)) + case uint16: + vals[i] = uint16(m.mutateUInt(uint64(v), math.MaxUint16)) + case uint32: + vals[i] = uint32(m.mutateUInt(uint64(v), math.MaxUint32)) + case uint64: + vals[i] = m.mutateUInt(uint64(v), maxUint) + case float32: + vals[i] = float32(m.mutateFloat(float64(v), math.MaxFloat32)) + case float64: + vals[i] = m.mutateFloat(v, math.MaxFloat64) + case bool: + if m.rand(2) == 1 { + vals[i] = !v // 50% chance of flipping the bool + } + case rune: // int32 + vals[i] = rune(m.mutateInt(int64(v), math.MaxInt32)) + case byte: // uint8 + vals[i] = byte(m.mutateUInt(uint64(v), math.MaxUint8)) + case string: + if len(v) > maxPerVal { + panic(fmt.Sprintf("cannot mutate bytes of length %d", len(v))) + } + if cap(m.scratch) < maxPerVal { + m.scratch = append(make([]byte, 0, maxPerVal), v...) + } else { + m.scratch = m.scratch[:len(v)] + copy(m.scratch, v) + } + m.mutateBytes(&m.scratch) + vals[i] = string(m.scratch) + case []byte: + if len(v) > maxPerVal { + panic(fmt.Sprintf("cannot mutate bytes of length %d", len(v))) + } + if cap(m.scratch) < maxPerVal { + m.scratch = append(make([]byte, 0, maxPerVal), v...) + } else { + m.scratch = m.scratch[:len(v)] + copy(m.scratch, v) + } + m.mutateBytes(&m.scratch) + vals[i] = m.scratch + default: + panic(fmt.Sprintf("type not supported for mutating: %T", vals[i])) + } +} + +func (m *mutator) mutateInt(v, maxValue int64) int64 { + var max int64 + for { + max = 100 + switch m.rand(2) { + case 0: + // Add a random number + if v >= maxValue { + continue + } + if v > 0 && maxValue-v < max { + // Don't let v exceed maxValue + max = maxValue - v + } + v += int64(1 + m.rand(int(max))) + return v + case 1: + // Subtract a random number + if v <= -maxValue { + continue + } + if v < 0 && maxValue+v < max { + // Don't let v drop below -maxValue + max = maxValue + v + } + v -= int64(1 + m.rand(int(max))) + return v + } + } +} + +func (m *mutator) mutateUInt(v, maxValue uint64) uint64 { + var max uint64 + for { + max = 100 + switch m.rand(2) { + case 0: + // Add a random number + if v >= maxValue { + continue + } + if v > 0 && maxValue-v < max { + // Don't let v exceed maxValue + max = maxValue - v + } + + v += uint64(1 + m.rand(int(max))) + return v + case 1: + // Subtract a random number + if v <= 0 { + continue + } + if v < max { + // Don't let v drop below 0 + max = v + } + v -= uint64(1 + m.rand(int(max))) + return v + } + } +} + +func (m *mutator) mutateFloat(v, maxValue float64) float64 { + var max float64 + for { + switch m.rand(4) { + case 0: + // Add a random number + if v >= maxValue { + continue + } + max = 100 + if v > 0 && maxValue-v < max { + // Don't let v exceed maxValue + max = maxValue - v + } + v += float64(1 + m.rand(int(max))) + return v + case 1: + // Subtract a random number + if v <= -maxValue { + continue + } + max = 100 + if v < 0 && maxValue+v < max { + // Don't let v drop below -maxValue + max = maxValue + v + } + v -= float64(1 + m.rand(int(max))) + return v + case 2: + // Multiply by a random number + absV := math.Abs(v) + if v == 0 || absV >= maxValue { + continue + } + max = 10 + if maxValue/absV < max { + // Don't let v go beyond the minimum or maximum value + max = maxValue / absV + } + v *= float64(1 + m.rand(int(max))) + return v + case 3: + // Divide by a random number + if v == 0 { + continue + } + v /= float64(1 + m.rand(10)) + return v + } + } +} + +type byteSliceMutator func(*mutator, []byte) []byte + +var byteSliceMutators = []byteSliceMutator{ + byteSliceRemoveBytes, + byteSliceInsertRandomBytes, + byteSliceDuplicateBytes, + byteSliceOverwriteBytes, + byteSliceBitFlip, + byteSliceXORByte, + byteSliceSwapByte, + byteSliceArithmeticUint8, + byteSliceArithmeticUint16, + byteSliceArithmeticUint32, + byteSliceArithmeticUint64, + byteSliceOverwriteInterestingUint8, + byteSliceOverwriteInterestingUint16, + byteSliceOverwriteInterestingUint32, + byteSliceInsertConstantBytes, + byteSliceOverwriteConstantBytes, + byteSliceShuffleBytes, + byteSliceSwapBytes, +} + +func (m *mutator) mutateBytes(ptrB *[]byte) { + b := *ptrB + defer func() { + if unsafe.SliceData(*ptrB) != unsafe.SliceData(b) { + panic("data moved to new address") + } + *ptrB = b + }() + + for { + mut := byteSliceMutators[m.rand(len(byteSliceMutators))] + if mutated := mut(m, b); mutated != nil { + b = mutated + return + } + } +} + +var ( + interesting8 = []int8{-128, -1, 0, 1, 16, 32, 64, 100, 127} + interesting16 = []int16{-32768, -129, 128, 255, 256, 512, 1000, 1024, 4096, 32767} + interesting32 = []int32{-2147483648, -100663046, -32769, 32768, 65535, 65536, 100663045, 2147483647} +) + +const ( + maxUint = uint64(^uint(0)) + maxInt = int64(maxUint >> 1) +) + +func init() { + for _, v := range interesting8 { + interesting16 = append(interesting16, int16(v)) + } + for _, v := range interesting16 { + interesting32 = append(interesting32, int32(v)) + } +} diff --git a/src/internal/fuzz/mutator_test.go b/src/internal/fuzz/mutator_test.go new file mode 100644 index 0000000..cea7e2e --- /dev/null +++ b/src/internal/fuzz/mutator_test.go @@ -0,0 +1,117 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "bytes" + "fmt" + "os" + "strconv" + "testing" +) + +func BenchmarkMutatorBytes(b *testing.B) { + origEnv := os.Getenv("GODEBUG") + defer func() { os.Setenv("GODEBUG", origEnv) }() + os.Setenv("GODEBUG", fmt.Sprintf("%s,fuzzseed=123", origEnv)) + m := newMutator() + + for _, size := range []int{ + 1, + 10, + 100, + 1000, + 10000, + 100000, + } { + b.Run(strconv.Itoa(size), func(b *testing.B) { + buf := make([]byte, size) + b.ResetTimer() + + for i := 0; i < b.N; i++ { + // resize buffer to the correct shape and reset the PCG + buf = buf[0:size] + m.r = newPcgRand() + m.mutate([]any{buf}, workerSharedMemSize) + } + }) + } +} + +func BenchmarkMutatorString(b *testing.B) { + origEnv := os.Getenv("GODEBUG") + defer func() { os.Setenv("GODEBUG", origEnv) }() + os.Setenv("GODEBUG", fmt.Sprintf("%s,fuzzseed=123", origEnv)) + m := newMutator() + + for _, size := range []int{ + 1, + 10, + 100, + 1000, + 10000, + 100000, + } { + b.Run(strconv.Itoa(size), func(b *testing.B) { + buf := make([]byte, size) + b.ResetTimer() + + for i := 0; i < b.N; i++ { + // resize buffer to the correct shape and reset the PCG + buf = buf[0:size] + m.r = newPcgRand() + m.mutate([]any{string(buf)}, workerSharedMemSize) + } + }) + } +} + +func BenchmarkMutatorAllBasicTypes(b *testing.B) { + origEnv := os.Getenv("GODEBUG") + defer func() { os.Setenv("GODEBUG", origEnv) }() + os.Setenv("GODEBUG", fmt.Sprintf("%s,fuzzseed=123", origEnv)) + m := newMutator() + + types := []any{ + []byte(""), + string(""), + false, + float32(0), + float64(0), + int(0), + int8(0), + int16(0), + int32(0), + int64(0), + uint8(0), + uint16(0), + uint32(0), + uint64(0), + } + + for _, t := range types { + b.Run(fmt.Sprintf("%T", t), func(b *testing.B) { + for i := 0; i < b.N; i++ { + m.r = newPcgRand() + m.mutate([]any{t}, workerSharedMemSize) + } + }) + } +} + +func TestStringImmutability(t *testing.T) { + v := []any{"hello"} + m := newMutator() + m.mutate(v, 1024) + original := v[0].(string) + originalCopy := make([]byte, len(original)) + copy(originalCopy, []byte(original)) + for i := 0; i < 25; i++ { + m.mutate(v, 1024) + } + if !bytes.Equal([]byte(original), originalCopy) { + t.Fatalf("string was mutated: got %x, want %x", []byte(original), originalCopy) + } +} diff --git a/src/internal/fuzz/mutators_byteslice.go b/src/internal/fuzz/mutators_byteslice.go new file mode 100644 index 0000000..d9dab1d --- /dev/null +++ b/src/internal/fuzz/mutators_byteslice.go @@ -0,0 +1,313 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +// byteSliceRemoveBytes removes a random chunk of bytes from b. +func byteSliceRemoveBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + pos0 := m.rand(len(b)) + pos1 := pos0 + m.chooseLen(len(b)-pos0) + copy(b[pos0:], b[pos1:]) + b = b[:len(b)-(pos1-pos0)] + return b +} + +// byteSliceInsertRandomBytes inserts a chunk of random bytes into b at a random +// position. +func byteSliceInsertRandomBytes(m *mutator, b []byte) []byte { + pos := m.rand(len(b) + 1) + n := m.chooseLen(1024) + if len(b)+n >= cap(b) { + return nil + } + b = b[:len(b)+n] + copy(b[pos+n:], b[pos:]) + for i := 0; i < n; i++ { + b[pos+i] = byte(m.rand(256)) + } + return b +} + +// byteSliceDuplicateBytes duplicates a chunk of bytes in b and inserts it into +// a random position. +func byteSliceDuplicateBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + src := m.rand(len(b)) + dst := m.rand(len(b)) + for dst == src { + dst = m.rand(len(b)) + } + n := m.chooseLen(len(b) - src) + // Use the end of the slice as scratch space to avoid doing an + // allocation. If the slice is too small abort and try something + // else. + if len(b)+(n*2) >= cap(b) { + return nil + } + end := len(b) + // Increase the size of b to fit the duplicated block as well as + // some extra working space + b = b[:end+(n*2)] + // Copy the block of bytes we want to duplicate to the end of the + // slice + copy(b[end+n:], b[src:src+n]) + // Shift the bytes after the splice point n positions to the right + // to make room for the new block + copy(b[dst+n:end+n], b[dst:end]) + // Insert the duplicate block into the splice point + copy(b[dst:], b[end+n:]) + b = b[:end+n] + return b +} + +// byteSliceOverwriteBytes overwrites a chunk of b with another chunk of b. +func byteSliceOverwriteBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + src := m.rand(len(b)) + dst := m.rand(len(b)) + for dst == src { + dst = m.rand(len(b)) + } + n := m.chooseLen(len(b) - src - 1) + copy(b[dst:], b[src:src+n]) + return b +} + +// byteSliceBitFlip flips a random bit in a random byte in b. +func byteSliceBitFlip(m *mutator, b []byte) []byte { + if len(b) == 0 { + return nil + } + pos := m.rand(len(b)) + b[pos] ^= 1 << uint(m.rand(8)) + return b +} + +// byteSliceXORByte XORs a random byte in b with a random value. +func byteSliceXORByte(m *mutator, b []byte) []byte { + if len(b) == 0 { + return nil + } + pos := m.rand(len(b)) + // In order to avoid a no-op (where the random value matches + // the existing value), use XOR instead of just setting to + // the random value. + b[pos] ^= byte(1 + m.rand(255)) + return b +} + +// byteSliceSwapByte swaps two random bytes in b. +func byteSliceSwapByte(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + src := m.rand(len(b)) + dst := m.rand(len(b)) + for dst == src { + dst = m.rand(len(b)) + } + b[src], b[dst] = b[dst], b[src] + return b +} + +// byteSliceArithmeticUint8 adds/subtracts from a random byte in b. +func byteSliceArithmeticUint8(m *mutator, b []byte) []byte { + if len(b) == 0 { + return nil + } + pos := m.rand(len(b)) + v := byte(m.rand(35) + 1) + if m.r.bool() { + b[pos] += v + } else { + b[pos] -= v + } + return b +} + +// byteSliceArithmeticUint16 adds/subtracts from a random uint16 in b. +func byteSliceArithmeticUint16(m *mutator, b []byte) []byte { + if len(b) < 2 { + return nil + } + v := uint16(m.rand(35) + 1) + if m.r.bool() { + v = 0 - v + } + pos := m.rand(len(b) - 1) + enc := m.randByteOrder() + enc.PutUint16(b[pos:], enc.Uint16(b[pos:])+v) + return b +} + +// byteSliceArithmeticUint32 adds/subtracts from a random uint32 in b. +func byteSliceArithmeticUint32(m *mutator, b []byte) []byte { + if len(b) < 4 { + return nil + } + v := uint32(m.rand(35) + 1) + if m.r.bool() { + v = 0 - v + } + pos := m.rand(len(b) - 3) + enc := m.randByteOrder() + enc.PutUint32(b[pos:], enc.Uint32(b[pos:])+v) + return b +} + +// byteSliceArithmeticUint64 adds/subtracts from a random uint64 in b. +func byteSliceArithmeticUint64(m *mutator, b []byte) []byte { + if len(b) < 8 { + return nil + } + v := uint64(m.rand(35) + 1) + if m.r.bool() { + v = 0 - v + } + pos := m.rand(len(b) - 7) + enc := m.randByteOrder() + enc.PutUint64(b[pos:], enc.Uint64(b[pos:])+v) + return b +} + +// byteSliceOverwriteInterestingUint8 overwrites a random byte in b with an interesting +// value. +func byteSliceOverwriteInterestingUint8(m *mutator, b []byte) []byte { + if len(b) == 0 { + return nil + } + pos := m.rand(len(b)) + b[pos] = byte(interesting8[m.rand(len(interesting8))]) + return b +} + +// byteSliceOverwriteInterestingUint16 overwrites a random uint16 in b with an interesting +// value. +func byteSliceOverwriteInterestingUint16(m *mutator, b []byte) []byte { + if len(b) < 2 { + return nil + } + pos := m.rand(len(b) - 1) + v := uint16(interesting16[m.rand(len(interesting16))]) + m.randByteOrder().PutUint16(b[pos:], v) + return b +} + +// byteSliceOverwriteInterestingUint32 overwrites a random uint16 in b with an interesting +// value. +func byteSliceOverwriteInterestingUint32(m *mutator, b []byte) []byte { + if len(b) < 4 { + return nil + } + pos := m.rand(len(b) - 3) + v := uint32(interesting32[m.rand(len(interesting32))]) + m.randByteOrder().PutUint32(b[pos:], v) + return b +} + +// byteSliceInsertConstantBytes inserts a chunk of constant bytes into a random position in b. +func byteSliceInsertConstantBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + dst := m.rand(len(b)) + // TODO(rolandshoemaker,katiehockman): 4096 was mainly picked + // randomly. We may want to either pick a much larger value + // (AFL uses 32768, paired with a similar impl to chooseLen + // which biases towards smaller lengths that grow over time), + // or set the max based on characteristics of the corpus + // (libFuzzer sets a min/max based on the min/max size of + // entries in the corpus and then picks uniformly from + // that range). + n := m.chooseLen(4096) + if len(b)+n >= cap(b) { + return nil + } + b = b[:len(b)+n] + copy(b[dst+n:], b[dst:]) + rb := byte(m.rand(256)) + for i := dst; i < dst+n; i++ { + b[i] = rb + } + return b +} + +// byteSliceOverwriteConstantBytes overwrites a chunk of b with constant bytes. +func byteSliceOverwriteConstantBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + dst := m.rand(len(b)) + n := m.chooseLen(len(b) - dst) + rb := byte(m.rand(256)) + for i := dst; i < dst+n; i++ { + b[i] = rb + } + return b +} + +// byteSliceShuffleBytes shuffles a chunk of bytes in b. +func byteSliceShuffleBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + dst := m.rand(len(b)) + n := m.chooseLen(len(b) - dst) + if n <= 2 { + return nil + } + // Start at the end of the range, and iterate backwards + // to dst, swapping each element with another element in + // dst:dst+n (Fisher-Yates shuffle). + for i := n - 1; i > 0; i-- { + j := m.rand(i + 1) + b[dst+i], b[dst+j] = b[dst+j], b[dst+i] + } + return b +} + +// byteSliceSwapBytes swaps two chunks of bytes in b. +func byteSliceSwapBytes(m *mutator, b []byte) []byte { + if len(b) <= 1 { + return nil + } + src := m.rand(len(b)) + dst := m.rand(len(b)) + for dst == src { + dst = m.rand(len(b)) + } + // Choose the random length as len(b) - max(src, dst) + // so that we don't attempt to swap a chunk that extends + // beyond the end of the slice + max := dst + if src > max { + max = src + } + n := m.chooseLen(len(b) - max - 1) + // Check that neither chunk intersect, so that we don't end up + // duplicating parts of the input, rather than swapping them + if src > dst && dst+n >= src || dst > src && src+n >= dst { + return nil + } + // Use the end of the slice as scratch space to avoid doing an + // allocation. If the slice is too small abort and try something + // else. + if len(b)+n >= cap(b) { + return nil + } + end := len(b) + b = b[:end+n] + copy(b[end:], b[dst:dst+n]) + copy(b[dst:], b[src:src+n]) + copy(b[src:], b[end:]) + b = b[:end] + return b +} diff --git a/src/internal/fuzz/mutators_byteslice_test.go b/src/internal/fuzz/mutators_byteslice_test.go new file mode 100644 index 0000000..7886967 --- /dev/null +++ b/src/internal/fuzz/mutators_byteslice_test.go @@ -0,0 +1,186 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "bytes" + "testing" +) + +type mockRand struct { + values []int + counter int + b bool +} + +func (mr *mockRand) uint32() uint32 { + c := mr.values[mr.counter] + mr.counter++ + return uint32(c) +} + +func (mr *mockRand) intn(n int) int { + c := mr.values[mr.counter] + mr.counter++ + return c % n +} + +func (mr *mockRand) uint32n(n uint32) uint32 { + c := mr.values[mr.counter] + mr.counter++ + return uint32(c) % n +} + +func (mr *mockRand) exp2() int { + c := mr.values[mr.counter] + mr.counter++ + return c +} + +func (mr *mockRand) bool() bool { + b := mr.b + mr.b = !mr.b + return b +} + +func (mr *mockRand) save(*uint64, *uint64) { + panic("unimplemented") +} + +func (mr *mockRand) restore(uint64, uint64) { + panic("unimplemented") +} + +func TestByteSliceMutators(t *testing.T) { + for _, tc := range []struct { + name string + mutator func(*mutator, []byte) []byte + randVals []int + input []byte + expected []byte + }{ + { + name: "byteSliceRemoveBytes", + mutator: byteSliceRemoveBytes, + input: []byte{1, 2, 3, 4}, + expected: []byte{4}, + }, + { + name: "byteSliceInsertRandomBytes", + mutator: byteSliceInsertRandomBytes, + input: make([]byte, 4, 8), + expected: []byte{3, 4, 5, 0, 0, 0, 0}, + }, + { + name: "byteSliceDuplicateBytes", + mutator: byteSliceDuplicateBytes, + input: append(make([]byte, 0, 13), []byte{1, 2, 3, 4}...), + expected: []byte{1, 1, 2, 3, 4, 2, 3, 4}, + }, + { + name: "byteSliceOverwriteBytes", + mutator: byteSliceOverwriteBytes, + input: []byte{1, 2, 3, 4}, + expected: []byte{1, 1, 3, 4}, + }, + { + name: "byteSliceBitFlip", + mutator: byteSliceBitFlip, + input: []byte{1, 2, 3, 4}, + expected: []byte{3, 2, 3, 4}, + }, + { + name: "byteSliceXORByte", + mutator: byteSliceXORByte, + input: []byte{1, 2, 3, 4}, + expected: []byte{3, 2, 3, 4}, + }, + { + name: "byteSliceSwapByte", + mutator: byteSliceSwapByte, + input: []byte{1, 2, 3, 4}, + expected: []byte{2, 1, 3, 4}, + }, + { + name: "byteSliceArithmeticUint8", + mutator: byteSliceArithmeticUint8, + input: []byte{1, 2, 3, 4}, + expected: []byte{255, 2, 3, 4}, + }, + { + name: "byteSliceArithmeticUint16", + mutator: byteSliceArithmeticUint16, + input: []byte{1, 2, 3, 4}, + expected: []byte{1, 3, 3, 4}, + }, + { + name: "byteSliceArithmeticUint32", + mutator: byteSliceArithmeticUint32, + input: []byte{1, 2, 3, 4}, + expected: []byte{2, 2, 3, 4}, + }, + { + name: "byteSliceArithmeticUint64", + mutator: byteSliceArithmeticUint64, + input: []byte{1, 2, 3, 4, 5, 6, 7, 8}, + expected: []byte{2, 2, 3, 4, 5, 6, 7, 8}, + }, + { + name: "byteSliceOverwriteInterestingUint8", + mutator: byteSliceOverwriteInterestingUint8, + input: []byte{1, 2, 3, 4}, + expected: []byte{255, 2, 3, 4}, + }, + { + name: "byteSliceOverwriteInterestingUint16", + mutator: byteSliceOverwriteInterestingUint16, + input: []byte{1, 2, 3, 4}, + expected: []byte{255, 127, 3, 4}, + }, + { + name: "byteSliceOverwriteInterestingUint32", + mutator: byteSliceOverwriteInterestingUint32, + input: []byte{1, 2, 3, 4}, + expected: []byte{250, 0, 0, 250}, + }, + { + name: "byteSliceInsertConstantBytes", + mutator: byteSliceInsertConstantBytes, + input: append(make([]byte, 0, 8), []byte{1, 2, 3, 4}...), + expected: []byte{3, 3, 3, 1, 2, 3, 4}, + }, + { + name: "byteSliceOverwriteConstantBytes", + mutator: byteSliceOverwriteConstantBytes, + input: []byte{1, 2, 3, 4}, + expected: []byte{3, 3, 3, 4}, + }, + { + name: "byteSliceShuffleBytes", + mutator: byteSliceShuffleBytes, + input: []byte{1, 2, 3, 4}, + expected: []byte{2, 3, 1, 4}, + }, + { + name: "byteSliceSwapBytes", + mutator: byteSliceSwapBytes, + randVals: []int{0, 2, 0, 2}, + input: append(make([]byte, 0, 9), []byte{1, 2, 3, 4}...), + expected: []byte{3, 2, 1, 4}, + }, + } { + t.Run(tc.name, func(t *testing.T) { + r := &mockRand{values: []int{0, 1, 2, 3, 4, 5}} + if tc.randVals != nil { + r.values = tc.randVals + } + m := &mutator{r: r} + b := tc.mutator(m, tc.input) + if !bytes.Equal(b, tc.expected) { + t.Errorf("got %x, want %x", b, tc.expected) + } + }) + } +} diff --git a/src/internal/fuzz/pcg.go b/src/internal/fuzz/pcg.go new file mode 100644 index 0000000..c9ea0af --- /dev/null +++ b/src/internal/fuzz/pcg.go @@ -0,0 +1,145 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "math/bits" + "os" + "strconv" + "strings" + "sync/atomic" + "time" +) + +type mutatorRand interface { + uint32() uint32 + intn(int) int + uint32n(uint32) uint32 + exp2() int + bool() bool + + save(randState, randInc *uint64) + restore(randState, randInc uint64) +} + +// The functions in pcg implement a 32 bit PRNG with a 64 bit period: pcg xsh rr +// 64 32. See https://www.pcg-random.org/ for more information. This +// implementation is geared specifically towards the needs of fuzzing: Simple +// creation and use, no reproducibility, no concurrency safety, just the +// necessary methods, optimized for speed. + +var globalInc uint64 // PCG stream + +const multiplier uint64 = 6364136223846793005 + +// pcgRand is a PRNG. It should not be copied or shared. No Rand methods are +// concurrency safe. +type pcgRand struct { + noCopy noCopy // help avoid mistakes: ask vet to ensure that we don't make a copy + state uint64 + inc uint64 +} + +func godebugSeed() *int { + debug := strings.Split(os.Getenv("GODEBUG"), ",") + for _, f := range debug { + if strings.HasPrefix(f, "fuzzseed=") { + seed, err := strconv.Atoi(strings.TrimPrefix(f, "fuzzseed=")) + if err != nil { + panic("malformed fuzzseed") + } + return &seed + } + } + return nil +} + +// newPcgRand generates a new, seeded Rand, ready for use. +func newPcgRand() *pcgRand { + r := new(pcgRand) + now := uint64(time.Now().UnixNano()) + if seed := godebugSeed(); seed != nil { + now = uint64(*seed) + } + inc := atomic.AddUint64(&globalInc, 1) + r.state = now + r.inc = (inc << 1) | 1 + r.step() + r.state += now + r.step() + return r +} + +func (r *pcgRand) step() { + r.state *= multiplier + r.state += r.inc +} + +func (r *pcgRand) save(randState, randInc *uint64) { + *randState = r.state + *randInc = r.inc +} + +func (r *pcgRand) restore(randState, randInc uint64) { + r.state = randState + r.inc = randInc +} + +// uint32 returns a pseudo-random uint32. +func (r *pcgRand) uint32() uint32 { + x := r.state + r.step() + return bits.RotateLeft32(uint32(((x>>18)^x)>>27), -int(x>>59)) +} + +// intn returns a pseudo-random number in [0, n). +// n must fit in a uint32. +func (r *pcgRand) intn(n int) int { + if int(uint32(n)) != n { + panic("large Intn") + } + return int(r.uint32n(uint32(n))) +} + +// uint32n returns a pseudo-random number in [0, n). +// +// For implementation details, see: +// https://lemire.me/blog/2016/06/27/a-fast-alternative-to-the-modulo-reduction +// https://lemire.me/blog/2016/06/30/fast-random-shuffling +func (r *pcgRand) uint32n(n uint32) uint32 { + v := r.uint32() + prod := uint64(v) * uint64(n) + low := uint32(prod) + if low < n { + thresh := uint32(-int32(n)) % n + for low < thresh { + v = r.uint32() + prod = uint64(v) * uint64(n) + low = uint32(prod) + } + } + return uint32(prod >> 32) +} + +// exp2 generates n with probability 1/2^(n+1). +func (r *pcgRand) exp2() int { + return bits.TrailingZeros32(r.uint32()) +} + +// bool generates a random bool. +func (r *pcgRand) bool() bool { + return r.uint32()&1 == 0 +} + +// noCopy may be embedded into structs which must not be copied +// after the first use. +// +// See https://golang.org/issues/8005#issuecomment-190753527 +// for details. +type noCopy struct{} + +// lock is a no-op used by -copylocks checker from `go vet`. +func (*noCopy) lock() {} +func (*noCopy) unlock() {} diff --git a/src/internal/fuzz/queue.go b/src/internal/fuzz/queue.go new file mode 100644 index 0000000..195d6eb --- /dev/null +++ b/src/internal/fuzz/queue.go @@ -0,0 +1,71 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +// queue holds a growable sequence of inputs for fuzzing and minimization. +// +// For now, this is a simple ring buffer +// (https://en.wikipedia.org/wiki/Circular_buffer). +// +// TODO(golang.org/issue/46224): use a prioritization algorithm based on input +// size, previous duration, coverage, and any other metrics that seem useful. +type queue struct { + // elems holds a ring buffer. + // The queue is empty when begin = end. + // The queue is full (until grow is called) when end = begin + N - 1 (mod N) + // where N = cap(elems). + elems []any + head, len int +} + +func (q *queue) cap() int { + return len(q.elems) +} + +func (q *queue) grow() { + oldCap := q.cap() + newCap := oldCap * 2 + if newCap == 0 { + newCap = 8 + } + newElems := make([]any, newCap) + oldLen := q.len + for i := 0; i < oldLen; i++ { + newElems[i] = q.elems[(q.head+i)%oldCap] + } + q.elems = newElems + q.head = 0 +} + +func (q *queue) enqueue(e any) { + if q.len+1 > q.cap() { + q.grow() + } + i := (q.head + q.len) % q.cap() + q.elems[i] = e + q.len++ +} + +func (q *queue) dequeue() (any, bool) { + if q.len == 0 { + return nil, false + } + e := q.elems[q.head] + q.elems[q.head] = nil + q.head = (q.head + 1) % q.cap() + q.len-- + return e, true +} + +func (q *queue) peek() (any, bool) { + if q.len == 0 { + return nil, false + } + return q.elems[q.head], true +} + +func (q *queue) clear() { + *q = queue{} +} diff --git a/src/internal/fuzz/queue_test.go b/src/internal/fuzz/queue_test.go new file mode 100644 index 0000000..3b179af --- /dev/null +++ b/src/internal/fuzz/queue_test.go @@ -0,0 +1,58 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import "testing" + +func TestQueue(t *testing.T) { + // Zero valued queue should have 0 length and capacity. + var q queue + if n := q.len; n != 0 { + t.Fatalf("empty queue has len %d; want 0", n) + } + if n := q.cap(); n != 0 { + t.Fatalf("empty queue has cap %d; want 0", n) + } + + // As we add elements, len should grow. + N := 32 + for i := 0; i < N; i++ { + q.enqueue(i) + if n := q.len; n != i+1 { + t.Fatalf("after adding %d elements, queue has len %d", i, n) + } + if v, ok := q.peek(); !ok { + t.Fatalf("couldn't peek after adding %d elements", i) + } else if v.(int) != 0 { + t.Fatalf("after adding %d elements, peek is %d; want 0", i, v) + } + } + + // As we remove and add elements, len should shrink and grow. + // We should also remove elements in the same order they were added. + want := 0 + for _, r := range []int{1, 2, 3, 5, 8, 13, 21} { + s := make([]int, 0, r) + for i := 0; i < r; i++ { + if got, ok := q.dequeue(); !ok { + t.Fatalf("after removing %d of %d elements, could not dequeue", i+1, r) + } else if got != want { + t.Fatalf("after removing %d of %d elements, got %d; want %d", i+1, r, got, want) + } else { + s = append(s, got.(int)) + } + want = (want + 1) % N + if n := q.len; n != N-i-1 { + t.Fatalf("after removing %d of %d elements, len is %d; want %d", i+1, r, n, N-i-1) + } + } + for i, v := range s { + q.enqueue(v) + if n := q.len; n != N-r+i+1 { + t.Fatalf("after adding back %d of %d elements, len is %d; want %d", i+1, r, n, n-r+i+1) + } + } + } +} diff --git a/src/internal/fuzz/sys_posix.go b/src/internal/fuzz/sys_posix.go new file mode 100644 index 0000000..fec6054 --- /dev/null +++ b/src/internal/fuzz/sys_posix.go @@ -0,0 +1,130 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build darwin || freebsd || linux + +package fuzz + +import ( + "fmt" + "os" + "os/exec" + "syscall" +) + +type sharedMemSys struct{} + +func sharedMemMapFile(f *os.File, size int, removeOnClose bool) (*sharedMem, error) { + prot := syscall.PROT_READ | syscall.PROT_WRITE + flags := syscall.MAP_FILE | syscall.MAP_SHARED + region, err := syscall.Mmap(int(f.Fd()), 0, size, prot, flags) + if err != nil { + return nil, err + } + + return &sharedMem{f: f, region: region, removeOnClose: removeOnClose}, nil +} + +// Close unmaps the shared memory and closes the temporary file. If this +// sharedMem was created with sharedMemTempFile, Close also removes the file. +func (m *sharedMem) Close() error { + // Attempt all operations, even if we get an error for an earlier operation. + // os.File.Close may fail due to I/O errors, but we still want to delete + // the temporary file. + var errs []error + errs = append(errs, + syscall.Munmap(m.region), + m.f.Close()) + if m.removeOnClose { + errs = append(errs, os.Remove(m.f.Name())) + } + for _, err := range errs { + if err != nil { + return err + } + } + return nil +} + +// setWorkerComm configures communication channels on the cmd that will +// run a worker process. +func setWorkerComm(cmd *exec.Cmd, comm workerComm) { + mem := <-comm.memMu + memFile := mem.f + comm.memMu <- mem + cmd.ExtraFiles = []*os.File{comm.fuzzIn, comm.fuzzOut, memFile} +} + +// getWorkerComm returns communication channels in the worker process. +func getWorkerComm() (comm workerComm, err error) { + fuzzIn := os.NewFile(3, "fuzz_in") + fuzzOut := os.NewFile(4, "fuzz_out") + memFile := os.NewFile(5, "fuzz_mem") + fi, err := memFile.Stat() + if err != nil { + return workerComm{}, err + } + size := int(fi.Size()) + if int64(size) != fi.Size() { + return workerComm{}, fmt.Errorf("fuzz temp file exceeds maximum size") + } + removeOnClose := false + mem, err := sharedMemMapFile(memFile, size, removeOnClose) + if err != nil { + return workerComm{}, err + } + memMu := make(chan *sharedMem, 1) + memMu <- mem + return workerComm{fuzzIn: fuzzIn, fuzzOut: fuzzOut, memMu: memMu}, nil +} + +// isInterruptError returns whether an error was returned by a process that +// was terminated by an interrupt signal (SIGINT). +func isInterruptError(err error) bool { + exitErr, ok := err.(*exec.ExitError) + if !ok || exitErr.ExitCode() >= 0 { + return false + } + status := exitErr.Sys().(syscall.WaitStatus) + return status.Signal() == syscall.SIGINT +} + +// terminationSignal checks if err is an exec.ExitError with a signal status. +// If it is, terminationSignal returns the signal and true. +// If not, -1 and false. +func terminationSignal(err error) (os.Signal, bool) { + exitErr, ok := err.(*exec.ExitError) + if !ok || exitErr.ExitCode() >= 0 { + return syscall.Signal(-1), false + } + status := exitErr.Sys().(syscall.WaitStatus) + return status.Signal(), status.Signaled() +} + +// isCrashSignal returns whether a signal was likely to have been caused by an +// error in the program that received it, triggered by a fuzz input. For +// example, SIGSEGV would be received after a nil pointer dereference. +// Other signals like SIGKILL or SIGHUP are more likely to have been sent by +// another process, and we shouldn't record a crasher if the worker process +// receives one of these. +// +// Note that Go installs its own signal handlers on startup, so some of these +// signals may only be received if signal handlers are changed. For example, +// SIGSEGV is normally transformed into a panic that causes the process to exit +// with status 2 if not recovered, which we handle as a crash. +func isCrashSignal(signal os.Signal) bool { + switch signal { + case + syscall.SIGILL, // illegal instruction + syscall.SIGTRAP, // breakpoint + syscall.SIGABRT, // abort() called + syscall.SIGBUS, // invalid memory access (e.g., misaligned address) + syscall.SIGFPE, // math error, e.g., integer divide by zero + syscall.SIGSEGV, // invalid memory access (e.g., write to read-only) + syscall.SIGPIPE: // sent data to closed pipe or socket + return true + default: + return false + } +} diff --git a/src/internal/fuzz/sys_unimplemented.go b/src/internal/fuzz/sys_unimplemented.go new file mode 100644 index 0000000..8687c1f --- /dev/null +++ b/src/internal/fuzz/sys_unimplemented.go @@ -0,0 +1,44 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// If you update this constraint, also update internal/platform.FuzzSupported. +// +//go:build !darwin && !freebsd && !linux && !windows + +package fuzz + +import ( + "os" + "os/exec" +) + +type sharedMemSys struct{} + +func sharedMemMapFile(f *os.File, size int, removeOnClose bool) (*sharedMem, error) { + panic("not implemented") +} + +func (m *sharedMem) Close() error { + panic("not implemented") +} + +func setWorkerComm(cmd *exec.Cmd, comm workerComm) { + panic("not implemented") +} + +func getWorkerComm() (comm workerComm, err error) { + panic("not implemented") +} + +func isInterruptError(err error) bool { + panic("not implemented") +} + +func terminationSignal(err error) (os.Signal, bool) { + panic("not implemented") +} + +func isCrashSignal(signal os.Signal) bool { + panic("not implemented") +} diff --git a/src/internal/fuzz/sys_windows.go b/src/internal/fuzz/sys_windows.go new file mode 100644 index 0000000..aa85be7 --- /dev/null +++ b/src/internal/fuzz/sys_windows.go @@ -0,0 +1,147 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "fmt" + "os" + "os/exec" + "syscall" + "unsafe" +) + +type sharedMemSys struct { + mapObj syscall.Handle +} + +func sharedMemMapFile(f *os.File, size int, removeOnClose bool) (mem *sharedMem, err error) { + defer func() { + if err != nil { + err = fmt.Errorf("mapping temporary file %s: %w", f.Name(), err) + } + }() + + // Create a file mapping object. The object itself is not shared. + mapObj, err := syscall.CreateFileMapping( + syscall.Handle(f.Fd()), // fhandle + nil, // sa + syscall.PAGE_READWRITE, // prot + 0, // maxSizeHigh + 0, // maxSizeLow + nil, // name + ) + if err != nil { + return nil, err + } + + // Create a view from the file mapping object. + access := uint32(syscall.FILE_MAP_READ | syscall.FILE_MAP_WRITE) + addr, err := syscall.MapViewOfFile( + mapObj, // handle + access, // access + 0, // offsetHigh + 0, // offsetLow + uintptr(size), // length + ) + if err != nil { + syscall.CloseHandle(mapObj) + return nil, err + } + + region := unsafe.Slice((*byte)(unsafe.Pointer(addr)), size) + return &sharedMem{ + f: f, + region: region, + removeOnClose: removeOnClose, + sys: sharedMemSys{mapObj: mapObj}, + }, nil +} + +// Close unmaps the shared memory and closes the temporary file. If this +// sharedMem was created with sharedMemTempFile, Close also removes the file. +func (m *sharedMem) Close() error { + // Attempt all operations, even if we get an error for an earlier operation. + // os.File.Close may fail due to I/O errors, but we still want to delete + // the temporary file. + var errs []error + errs = append(errs, + syscall.UnmapViewOfFile(uintptr(unsafe.Pointer(&m.region[0]))), + syscall.CloseHandle(m.sys.mapObj), + m.f.Close()) + if m.removeOnClose { + errs = append(errs, os.Remove(m.f.Name())) + } + for _, err := range errs { + if err != nil { + return err + } + } + return nil +} + +// setWorkerComm configures communication channels on the cmd that will +// run a worker process. +func setWorkerComm(cmd *exec.Cmd, comm workerComm) { + mem := <-comm.memMu + memName := mem.f.Name() + comm.memMu <- mem + syscall.SetHandleInformation(syscall.Handle(comm.fuzzIn.Fd()), syscall.HANDLE_FLAG_INHERIT, 1) + syscall.SetHandleInformation(syscall.Handle(comm.fuzzOut.Fd()), syscall.HANDLE_FLAG_INHERIT, 1) + cmd.Env = append(cmd.Env, fmt.Sprintf("GO_TEST_FUZZ_WORKER_HANDLES=%x,%x,%q", comm.fuzzIn.Fd(), comm.fuzzOut.Fd(), memName)) + cmd.SysProcAttr = &syscall.SysProcAttr{AdditionalInheritedHandles: []syscall.Handle{syscall.Handle(comm.fuzzIn.Fd()), syscall.Handle(comm.fuzzOut.Fd())}} +} + +// getWorkerComm returns communication channels in the worker process. +func getWorkerComm() (comm workerComm, err error) { + v := os.Getenv("GO_TEST_FUZZ_WORKER_HANDLES") + if v == "" { + return workerComm{}, fmt.Errorf("GO_TEST_FUZZ_WORKER_HANDLES not set") + } + var fuzzInFD, fuzzOutFD uintptr + var memName string + if _, err := fmt.Sscanf(v, "%x,%x,%q", &fuzzInFD, &fuzzOutFD, &memName); err != nil { + return workerComm{}, fmt.Errorf("parsing GO_TEST_FUZZ_WORKER_HANDLES=%s: %v", v, err) + } + + fuzzIn := os.NewFile(fuzzInFD, "fuzz_in") + fuzzOut := os.NewFile(fuzzOutFD, "fuzz_out") + tmpFile, err := os.OpenFile(memName, os.O_RDWR, 0) + if err != nil { + return workerComm{}, fmt.Errorf("worker opening temp file: %w", err) + } + fi, err := tmpFile.Stat() + if err != nil { + return workerComm{}, fmt.Errorf("worker checking temp file size: %w", err) + } + size := int(fi.Size()) + if int64(size) != fi.Size() { + return workerComm{}, fmt.Errorf("fuzz temp file exceeds maximum size") + } + removeOnClose := false + mem, err := sharedMemMapFile(tmpFile, size, removeOnClose) + if err != nil { + return workerComm{}, err + } + memMu := make(chan *sharedMem, 1) + memMu <- mem + + return workerComm{fuzzIn: fuzzIn, fuzzOut: fuzzOut, memMu: memMu}, nil +} + +func isInterruptError(err error) bool { + // On Windows, we can't tell whether the process was interrupted by the error + // returned by Wait. It looks like an ExitError with status 1. + return false +} + +// terminationSignal returns -1 and false because Windows doesn't have signals. +func terminationSignal(err error) (os.Signal, bool) { + return syscall.Signal(-1), false +} + +// isCrashSignal is not implemented because Windows doesn't have signals. +func isCrashSignal(signal os.Signal) bool { + panic("not implemented: no signals on windows") +} diff --git a/src/internal/fuzz/trace.go b/src/internal/fuzz/trace.go new file mode 100644 index 0000000..a15c370 --- /dev/null +++ b/src/internal/fuzz/trace.go @@ -0,0 +1,35 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !libfuzzer + +package fuzz + +import _ "unsafe" // for go:linkname + +//go:linkname libfuzzerTraceCmp1 runtime.libfuzzerTraceCmp1 +//go:linkname libfuzzerTraceCmp2 runtime.libfuzzerTraceCmp2 +//go:linkname libfuzzerTraceCmp4 runtime.libfuzzerTraceCmp4 +//go:linkname libfuzzerTraceCmp8 runtime.libfuzzerTraceCmp8 + +//go:linkname libfuzzerTraceConstCmp1 runtime.libfuzzerTraceConstCmp1 +//go:linkname libfuzzerTraceConstCmp2 runtime.libfuzzerTraceConstCmp2 +//go:linkname libfuzzerTraceConstCmp4 runtime.libfuzzerTraceConstCmp4 +//go:linkname libfuzzerTraceConstCmp8 runtime.libfuzzerTraceConstCmp8 + +//go:linkname libfuzzerHookStrCmp runtime.libfuzzerHookStrCmp +//go:linkname libfuzzerHookEqualFold runtime.libfuzzerHookEqualFold + +func libfuzzerTraceCmp1(arg0, arg1 uint8, fakePC uint) {} +func libfuzzerTraceCmp2(arg0, arg1 uint16, fakePC uint) {} +func libfuzzerTraceCmp4(arg0, arg1 uint32, fakePC uint) {} +func libfuzzerTraceCmp8(arg0, arg1 uint64, fakePC uint) {} + +func libfuzzerTraceConstCmp1(arg0, arg1 uint8, fakePC uint) {} +func libfuzzerTraceConstCmp2(arg0, arg1 uint16, fakePC uint) {} +func libfuzzerTraceConstCmp4(arg0, arg1 uint32, fakePC uint) {} +func libfuzzerTraceConstCmp8(arg0, arg1 uint64, fakePC uint) {} + +func libfuzzerHookStrCmp(arg0, arg1 string, fakePC uint) {} +func libfuzzerHookEqualFold(arg0, arg1 string, fakePC uint) {} diff --git a/src/internal/fuzz/worker.go b/src/internal/fuzz/worker.go new file mode 100644 index 0000000..c952670 --- /dev/null +++ b/src/internal/fuzz/worker.go @@ -0,0 +1,1195 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "bytes" + "context" + "crypto/sha256" + "encoding/json" + "errors" + "fmt" + "io" + "os" + "os/exec" + "reflect" + "runtime" + "sync" + "time" +) + +const ( + // workerFuzzDuration is the amount of time a worker can spend testing random + // variations of an input given by the coordinator. + workerFuzzDuration = 100 * time.Millisecond + + // workerTimeoutDuration is the amount of time a worker can go without + // responding to the coordinator before being stopped. + workerTimeoutDuration = 1 * time.Second + + // workerExitCode is used as an exit code by fuzz worker processes after an internal error. + // This distinguishes internal errors from uncontrolled panics and other crashes. + // Keep in sync with internal/fuzz.workerExitCode. + workerExitCode = 70 + + // workerSharedMemSize is the maximum size of the shared memory file used to + // communicate with workers. This limits the size of fuzz inputs. + workerSharedMemSize = 100 << 20 // 100 MB +) + +// worker manages a worker process running a test binary. The worker object +// exists only in the coordinator (the process started by 'go test -fuzz'). +// workerClient is used by the coordinator to send RPCs to the worker process, +// which handles them with workerServer. +type worker struct { + dir string // working directory, same as package directory + binPath string // path to test executable + args []string // arguments for test executable + env []string // environment for test executable + + coordinator *coordinator + + memMu chan *sharedMem // mutex guarding shared memory with worker; persists across processes. + + cmd *exec.Cmd // current worker process + client *workerClient // used to communicate with worker process + waitErr error // last error returned by wait, set before termC is closed. + interrupted bool // true after stop interrupts a running worker. + termC chan struct{} // closed by wait when worker process terminates +} + +func newWorker(c *coordinator, dir, binPath string, args, env []string) (*worker, error) { + mem, err := sharedMemTempFile(workerSharedMemSize) + if err != nil { + return nil, err + } + memMu := make(chan *sharedMem, 1) + memMu <- mem + return &worker{ + dir: dir, + binPath: binPath, + args: args, + env: env[:len(env):len(env)], // copy on append to ensure workers don't overwrite each other. + coordinator: c, + memMu: memMu, + }, nil +} + +// cleanup releases persistent resources associated with the worker. +func (w *worker) cleanup() error { + mem := <-w.memMu + if mem == nil { + return nil + } + close(w.memMu) + return mem.Close() +} + +// coordinate runs the test binary to perform fuzzing. +// +// coordinate loops until ctx is cancelled or a fatal error is encountered. +// If a test process terminates unexpectedly while fuzzing, coordinate will +// attempt to restart and continue unless the termination can be attributed +// to an interruption (from a timer or the user). +// +// While looping, coordinate receives inputs from the coordinator, passes +// those inputs to the worker process, then passes the results back to +// the coordinator. +func (w *worker) coordinate(ctx context.Context) error { + // Main event loop. + for { + // Start or restart the worker if it's not running. + if !w.isRunning() { + if err := w.startAndPing(ctx); err != nil { + return err + } + } + + select { + case <-ctx.Done(): + // Worker was told to stop. + err := w.stop() + if err != nil && !w.interrupted && !isInterruptError(err) { + return err + } + return ctx.Err() + + case <-w.termC: + // Worker process terminated unexpectedly while waiting for input. + err := w.stop() + if w.interrupted { + panic("worker interrupted after unexpected termination") + } + if err == nil || isInterruptError(err) { + // Worker stopped, either by exiting with status 0 or after being + // interrupted with a signal that was not sent by the coordinator. + // + // When the user presses ^C, on POSIX platforms, SIGINT is delivered to + // all processes in the group concurrently, and the worker may see it + // before the coordinator. The worker should exit 0 gracefully (in + // theory). + // + // This condition is probably intended by the user, so suppress + // the error. + return nil + } + if exitErr, ok := err.(*exec.ExitError); ok && exitErr.ExitCode() == workerExitCode { + // Worker exited with a code indicating F.Fuzz was not called correctly, + // for example, F.Fail was called first. + return fmt.Errorf("fuzzing process exited unexpectedly due to an internal failure: %w", err) + } + // Worker exited non-zero or was terminated by a non-interrupt + // signal (for example, SIGSEGV) while fuzzing. + return fmt.Errorf("fuzzing process hung or terminated unexpectedly: %w", err) + // TODO(jayconrod,katiehockman): if -keepfuzzing, restart worker. + + case input := <-w.coordinator.inputC: + // Received input from coordinator. + args := fuzzArgs{ + Limit: input.limit, + Timeout: input.timeout, + Warmup: input.warmup, + CoverageData: input.coverageData, + } + entry, resp, isInternalError, err := w.client.fuzz(ctx, input.entry, args) + canMinimize := true + if err != nil { + // Error communicating with worker. + w.stop() + if ctx.Err() != nil { + // Timeout or interruption. + return ctx.Err() + } + if w.interrupted { + // Communication error before we stopped the worker. + // Report an error, but don't record a crasher. + return fmt.Errorf("communicating with fuzzing process: %v", err) + } + if sig, ok := terminationSignal(w.waitErr); ok && !isCrashSignal(sig) { + // Worker terminated by a signal that probably wasn't caused by a + // specific input to the fuzz function. For example, on Linux, + // the kernel (OOM killer) may send SIGKILL to a process using a lot + // of memory. Or the shell might send SIGHUP when the terminal + // is closed. Don't record a crasher. + return fmt.Errorf("fuzzing process terminated by unexpected signal; no crash will be recorded: %v", w.waitErr) + } + if isInternalError { + // An internal error occurred which shouldn't be considered + // a crash. + return err + } + // Unexpected termination. Set error message and fall through. + // We'll restart the worker on the next iteration. + // Don't attempt to minimize this since it crashed the worker. + resp.Err = fmt.Sprintf("fuzzing process hung or terminated unexpectedly: %v", w.waitErr) + canMinimize = false + } + result := fuzzResult{ + limit: input.limit, + count: resp.Count, + totalDuration: resp.TotalDuration, + entryDuration: resp.InterestingDuration, + entry: entry, + crasherMsg: resp.Err, + coverageData: resp.CoverageData, + canMinimize: canMinimize, + } + w.coordinator.resultC <- result + + case input := <-w.coordinator.minimizeC: + // Received input to minimize from coordinator. + result, err := w.minimize(ctx, input) + if err != nil { + // Error minimizing. Send back the original input. If it didn't cause + // an error before, report it as causing an error now. + // TODO: double-check this is handled correctly when + // implementing -keepfuzzing. + result = fuzzResult{ + entry: input.entry, + crasherMsg: input.crasherMsg, + canMinimize: false, + limit: input.limit, + } + if result.crasherMsg == "" { + result.crasherMsg = err.Error() + } + } + if shouldPrintDebugInfo() { + w.coordinator.debugLogf( + "input minimized, id: %s, original id: %s, crasher: %t, originally crasher: %t, minimizing took: %s", + result.entry.Path, + input.entry.Path, + result.crasherMsg != "", + input.crasherMsg != "", + result.totalDuration, + ) + } + w.coordinator.resultC <- result + } + } +} + +// minimize tells a worker process to attempt to find a smaller value that +// either causes an error (if we started minimizing because we found an input +// that causes an error) or preserves new coverage (if we started minimizing +// because we found an input that expands coverage). +func (w *worker) minimize(ctx context.Context, input fuzzMinimizeInput) (min fuzzResult, err error) { + if w.coordinator.opts.MinimizeTimeout != 0 { + var cancel func() + ctx, cancel = context.WithTimeout(ctx, w.coordinator.opts.MinimizeTimeout) + defer cancel() + } + + args := minimizeArgs{ + Limit: input.limit, + Timeout: input.timeout, + KeepCoverage: input.keepCoverage, + } + entry, resp, err := w.client.minimize(ctx, input.entry, args) + if err != nil { + // Error communicating with worker. + w.stop() + if ctx.Err() != nil || w.interrupted || isInterruptError(w.waitErr) { + // Worker was interrupted, possibly by the user pressing ^C. + // Normally, workers can handle interrupts and timeouts gracefully and + // will return without error. An error here indicates the worker + // may not have been in a good state, but the error won't be meaningful + // to the user. Just return the original crasher without logging anything. + return fuzzResult{ + entry: input.entry, + crasherMsg: input.crasherMsg, + coverageData: input.keepCoverage, + canMinimize: false, + limit: input.limit, + }, nil + } + return fuzzResult{ + entry: entry, + crasherMsg: fmt.Sprintf("fuzzing process hung or terminated unexpectedly while minimizing: %v", err), + canMinimize: false, + limit: input.limit, + count: resp.Count, + totalDuration: resp.Duration, + }, nil + } + + if input.crasherMsg != "" && resp.Err == "" { + return fuzzResult{}, fmt.Errorf("attempted to minimize a crash but could not reproduce") + } + + return fuzzResult{ + entry: entry, + crasherMsg: resp.Err, + coverageData: resp.CoverageData, + canMinimize: false, + limit: input.limit, + count: resp.Count, + totalDuration: resp.Duration, + }, nil +} + +func (w *worker) isRunning() bool { + return w.cmd != nil +} + +// startAndPing starts the worker process and sends it a message to make sure it +// can communicate. +// +// startAndPing returns an error if any part of this didn't work, including if +// the context is expired or the worker process was interrupted before it +// responded. Errors that happen after start but before the ping response +// likely indicate that the worker did not call F.Fuzz or called F.Fail first. +// We don't record crashers for these errors. +func (w *worker) startAndPing(ctx context.Context) error { + if ctx.Err() != nil { + return ctx.Err() + } + if err := w.start(); err != nil { + return err + } + if err := w.client.ping(ctx); err != nil { + w.stop() + if ctx.Err() != nil { + return ctx.Err() + } + if isInterruptError(err) { + // User may have pressed ^C before worker responded. + return err + } + // TODO: record and return stderr. + return fmt.Errorf("fuzzing process terminated without fuzzing: %w", err) + } + return nil +} + +// start runs a new worker process. +// +// If the process couldn't be started, start returns an error. Start won't +// return later termination errors from the process if they occur. +// +// If the process starts successfully, start returns nil. stop must be called +// once later to clean up, even if the process terminates on its own. +// +// When the process terminates, w.waitErr is set to the error (if any), and +// w.termC is closed. +func (w *worker) start() (err error) { + if w.isRunning() { + panic("worker already started") + } + w.waitErr = nil + w.interrupted = false + w.termC = nil + + cmd := exec.Command(w.binPath, w.args...) + cmd.Dir = w.dir + cmd.Env = w.env[:len(w.env):len(w.env)] // copy on append to ensure workers don't overwrite each other. + + // Create the "fuzz_in" and "fuzz_out" pipes so we can communicate with + // the worker. We don't use stdin and stdout, since the test binary may + // do something else with those. + // + // Each pipe has a reader and a writer. The coordinator writes to fuzzInW + // and reads from fuzzOutR. The worker inherits fuzzInR and fuzzOutW. + // The coordinator closes fuzzInR and fuzzOutW after starting the worker, + // since we have no further need of them. + fuzzInR, fuzzInW, err := os.Pipe() + if err != nil { + return err + } + defer fuzzInR.Close() + fuzzOutR, fuzzOutW, err := os.Pipe() + if err != nil { + fuzzInW.Close() + return err + } + defer fuzzOutW.Close() + setWorkerComm(cmd, workerComm{fuzzIn: fuzzInR, fuzzOut: fuzzOutW, memMu: w.memMu}) + + // Start the worker process. + if err := cmd.Start(); err != nil { + fuzzInW.Close() + fuzzOutR.Close() + return err + } + + // Worker started successfully. + // After this, w.client owns fuzzInW and fuzzOutR, so w.client.Close must be + // called later by stop. + w.cmd = cmd + w.termC = make(chan struct{}) + comm := workerComm{fuzzIn: fuzzInW, fuzzOut: fuzzOutR, memMu: w.memMu} + m := newMutator() + w.client = newWorkerClient(comm, m) + + go func() { + w.waitErr = w.cmd.Wait() + close(w.termC) + }() + + return nil +} + +// stop tells the worker process to exit by closing w.client, then blocks until +// it terminates. If the worker doesn't terminate after a short time, stop +// signals it with os.Interrupt (where supported), then os.Kill. +// +// stop returns the error the process terminated with, if any (same as +// w.waitErr). +// +// stop must be called at least once after start returns successfully, even if +// the worker process terminates unexpectedly. +func (w *worker) stop() error { + if w.termC == nil { + panic("worker was not started successfully") + } + select { + case <-w.termC: + // Worker already terminated. + if w.client == nil { + // stop already called. + return w.waitErr + } + // Possible unexpected termination. + w.client.Close() + w.cmd = nil + w.client = nil + return w.waitErr + default: + // Worker still running. + } + + // Tell the worker to stop by closing fuzz_in. It won't actually stop until it + // finishes with earlier calls. + closeC := make(chan struct{}) + go func() { + w.client.Close() + close(closeC) + }() + + sig := os.Interrupt + if runtime.GOOS == "windows" { + // Per https://golang.org/pkg/os/#Signal, “Interrupt is not implemented on + // Windows; using it with os.Process.Signal will return an error.” + // Fall back to Kill instead. + sig = os.Kill + } + + t := time.NewTimer(workerTimeoutDuration) + for { + select { + case <-w.termC: + // Worker terminated. + t.Stop() + <-closeC + w.cmd = nil + w.client = nil + return w.waitErr + + case <-t.C: + // Timer fired before worker terminated. + w.interrupted = true + switch sig { + case os.Interrupt: + // Try to stop the worker with SIGINT and wait a little longer. + w.cmd.Process.Signal(sig) + sig = os.Kill + t.Reset(workerTimeoutDuration) + + case os.Kill: + // Try to stop the worker with SIGKILL and keep waiting. + w.cmd.Process.Signal(sig) + sig = nil + t.Reset(workerTimeoutDuration) + + case nil: + // Still waiting. Print a message to let the user know why. + fmt.Fprintf(w.coordinator.opts.Log, "waiting for fuzzing process to terminate...\n") + } + } + } +} + +// RunFuzzWorker is called in a worker process to communicate with the +// coordinator process in order to fuzz random inputs. RunFuzzWorker loops +// until the coordinator tells it to stop. +// +// fn is a wrapper on the fuzz function. It may return an error to indicate +// a given input "crashed". The coordinator will also record a crasher if +// the function times out or terminates the process. +// +// RunFuzzWorker returns an error if it could not communicate with the +// coordinator process. +func RunFuzzWorker(ctx context.Context, fn func(CorpusEntry) error) error { + comm, err := getWorkerComm() + if err != nil { + return err + } + srv := &workerServer{ + workerComm: comm, + fuzzFn: func(e CorpusEntry) (time.Duration, error) { + timer := time.AfterFunc(10*time.Second, func() { + panic("deadlocked!") // this error message won't be printed + }) + defer timer.Stop() + start := time.Now() + err := fn(e) + return time.Since(start), err + }, + m: newMutator(), + } + return srv.serve(ctx) +} + +// call is serialized and sent from the coordinator on fuzz_in. It acts as +// a minimalist RPC mechanism. Exactly one of its fields must be set to indicate +// which method to call. +type call struct { + Ping *pingArgs + Fuzz *fuzzArgs + Minimize *minimizeArgs +} + +// minimizeArgs contains arguments to workerServer.minimize. The value to +// minimize is already in shared memory. +type minimizeArgs struct { + // Timeout is the time to spend minimizing. This may include time to start up, + // especially if the input causes the worker process to terminated, requiring + // repeated restarts. + Timeout time.Duration + + // Limit is the maximum number of values to test, without spending more time + // than Duration. 0 indicates no limit. + Limit int64 + + // KeepCoverage is a set of coverage counters the worker should attempt to + // keep in minimized values. When provided, the worker will reject inputs that + // don't cause at least one of these bits to be set. + KeepCoverage []byte + + // Index is the index of the fuzz target parameter to be minimized. + Index int +} + +// minimizeResponse contains results from workerServer.minimize. +type minimizeResponse struct { + // WroteToMem is true if the worker found a smaller input and wrote it to + // shared memory. If minimizeArgs.KeepCoverage was set, the minimized input + // preserved at least one coverage bit and did not cause an error. + // Otherwise, the minimized input caused some error, recorded in Err. + WroteToMem bool + + // Err is the error string caused by the value in shared memory, if any. + Err string + + // CoverageData is the set of coverage bits activated by the minimized value + // in shared memory. When set, it contains at least one bit from KeepCoverage. + // CoverageData will be nil if Err is set or if minimization failed. + CoverageData []byte + + // Duration is the time spent minimizing, not including starting or cleaning up. + Duration time.Duration + + // Count is the number of values tested. + Count int64 +} + +// fuzzArgs contains arguments to workerServer.fuzz. The value to fuzz is +// passed in shared memory. +type fuzzArgs struct { + // Timeout is the time to spend fuzzing, not including starting or + // cleaning up. + Timeout time.Duration + + // Limit is the maximum number of values to test, without spending more time + // than Duration. 0 indicates no limit. + Limit int64 + + // Warmup indicates whether this is part of a warmup run, meaning that + // fuzzing should not occur. If coverageEnabled is true, then coverage data + // should be reported. + Warmup bool + + // CoverageData is the coverage data. If set, the worker should update its + // local coverage data prior to fuzzing. + CoverageData []byte +} + +// fuzzResponse contains results from workerServer.fuzz. +type fuzzResponse struct { + // Duration is the time spent fuzzing, not including starting or cleaning up. + TotalDuration time.Duration + InterestingDuration time.Duration + + // Count is the number of values tested. + Count int64 + + // CoverageData is set if the value in shared memory expands coverage + // and therefore may be interesting to the coordinator. + CoverageData []byte + + // Err is the error string caused by the value in shared memory, which is + // non-empty if the value in shared memory caused a crash. + Err string + + // InternalErr is the error string caused by an internal error in the + // worker. This shouldn't be considered a crasher. + InternalErr string +} + +// pingArgs contains arguments to workerServer.ping. +type pingArgs struct{} + +// pingResponse contains results from workerServer.ping. +type pingResponse struct{} + +// workerComm holds pipes and shared memory used for communication +// between the coordinator process (client) and a worker process (server). +// These values are unique to each worker; they are shared only with the +// coordinator, not with other workers. +// +// Access to shared memory is synchronized implicitly over the RPC protocol +// implemented in workerServer and workerClient. During a call, the client +// (worker) has exclusive access to shared memory; at other times, the server +// (coordinator) has exclusive access. +type workerComm struct { + fuzzIn, fuzzOut *os.File + memMu chan *sharedMem // mutex guarding shared memory +} + +// workerServer is a minimalist RPC server, run by fuzz worker processes. +// It allows the coordinator process (using workerClient) to call methods in a +// worker process. This system allows the coordinator to run multiple worker +// processes in parallel and to collect inputs that caused crashes from shared +// memory after a worker process terminates unexpectedly. +type workerServer struct { + workerComm + m *mutator + + // coverageMask is the local coverage data for the worker. It is + // periodically updated to reflect the data in the coordinator when new + // coverage is found. + coverageMask []byte + + // fuzzFn runs the worker's fuzz target on the given input and returns an + // error if it finds a crasher (the process may also exit or crash), and the + // time it took to run the input. It sets a deadline of 10 seconds, at which + // point it will panic with the assumption that the process is hanging or + // deadlocked. + fuzzFn func(CorpusEntry) (time.Duration, error) +} + +// serve reads serialized RPC messages on fuzzIn. When serve receives a message, +// it calls the corresponding method, then sends the serialized result back +// on fuzzOut. +// +// serve handles RPC calls synchronously; it will not attempt to read a message +// until the previous call has finished. +// +// serve returns errors that occurred when communicating over pipes. serve +// does not return errors from method calls; those are passed through serialized +// responses. +func (ws *workerServer) serve(ctx context.Context) error { + enc := json.NewEncoder(ws.fuzzOut) + dec := json.NewDecoder(&contextReader{ctx: ctx, r: ws.fuzzIn}) + for { + var c call + if err := dec.Decode(&c); err != nil { + if err == io.EOF || err == ctx.Err() { + return nil + } else { + return err + } + } + + var resp any + switch { + case c.Fuzz != nil: + resp = ws.fuzz(ctx, *c.Fuzz) + case c.Minimize != nil: + resp = ws.minimize(ctx, *c.Minimize) + case c.Ping != nil: + resp = ws.ping(ctx, *c.Ping) + default: + return errors.New("no arguments provided for any call") + } + + if err := enc.Encode(resp); err != nil { + return err + } + } +} + +// chainedMutations is how many mutations are applied before the worker +// resets the input to it's original state. +// NOTE: this number was picked without much thought. It is low enough that +// it seems to create a significant diversity in mutated inputs. We may want +// to consider looking into this more closely once we have a proper performance +// testing framework. Another option is to randomly pick the number of chained +// mutations on each invocation of the workerServer.fuzz method (this appears to +// be what libFuzzer does, although there seems to be no documentation which +// explains why this choice was made.) +const chainedMutations = 5 + +// fuzz runs the test function on random variations of the input value in shared +// memory for a limited duration or number of iterations. +// +// fuzz returns early if it finds an input that crashes the fuzz function (with +// fuzzResponse.Err set) or an input that expands coverage (with +// fuzzResponse.InterestingDuration set). +// +// fuzz does not modify the input in shared memory. Instead, it saves the +// initial PRNG state in shared memory and increments a counter in shared +// memory before each call to the test function. The caller may reconstruct +// the crashing input with this information, since the PRNG is deterministic. +func (ws *workerServer) fuzz(ctx context.Context, args fuzzArgs) (resp fuzzResponse) { + if args.CoverageData != nil { + if ws.coverageMask != nil && len(args.CoverageData) != len(ws.coverageMask) { + resp.InternalErr = fmt.Sprintf("unexpected size for CoverageData: got %d, expected %d", len(args.CoverageData), len(ws.coverageMask)) + return resp + } + ws.coverageMask = args.CoverageData + } + start := time.Now() + defer func() { resp.TotalDuration = time.Since(start) }() + + if args.Timeout != 0 { + var cancel func() + ctx, cancel = context.WithTimeout(ctx, args.Timeout) + defer cancel() + } + mem := <-ws.memMu + ws.m.r.save(&mem.header().randState, &mem.header().randInc) + defer func() { + resp.Count = mem.header().count + ws.memMu <- mem + }() + if args.Limit > 0 && mem.header().count >= args.Limit { + resp.InternalErr = fmt.Sprintf("mem.header().count %d already exceeds args.Limit %d", mem.header().count, args.Limit) + return resp + } + + originalVals, err := unmarshalCorpusFile(mem.valueCopy()) + if err != nil { + resp.InternalErr = err.Error() + return resp + } + vals := make([]any, len(originalVals)) + copy(vals, originalVals) + + shouldStop := func() bool { + return args.Limit > 0 && mem.header().count >= args.Limit + } + fuzzOnce := func(entry CorpusEntry) (dur time.Duration, cov []byte, errMsg string) { + mem.header().count++ + var err error + dur, err = ws.fuzzFn(entry) + if err != nil { + errMsg = err.Error() + if errMsg == "" { + errMsg = "fuzz function failed with no input" + } + return dur, nil, errMsg + } + if ws.coverageMask != nil && countNewCoverageBits(ws.coverageMask, coverageSnapshot) > 0 { + return dur, coverageSnapshot, "" + } + return dur, nil, "" + } + + if args.Warmup { + dur, _, errMsg := fuzzOnce(CorpusEntry{Values: vals}) + if errMsg != "" { + resp.Err = errMsg + return resp + } + resp.InterestingDuration = dur + if coverageEnabled { + resp.CoverageData = coverageSnapshot + } + return resp + } + + for { + select { + case <-ctx.Done(): + return resp + default: + if mem.header().count%chainedMutations == 0 { + copy(vals, originalVals) + ws.m.r.save(&mem.header().randState, &mem.header().randInc) + } + ws.m.mutate(vals, cap(mem.valueRef())) + + entry := CorpusEntry{Values: vals} + dur, cov, errMsg := fuzzOnce(entry) + if errMsg != "" { + resp.Err = errMsg + return resp + } + if cov != nil { + resp.CoverageData = cov + resp.InterestingDuration = dur + return resp + } + if shouldStop() { + return resp + } + } + } +} + +func (ws *workerServer) minimize(ctx context.Context, args minimizeArgs) (resp minimizeResponse) { + start := time.Now() + defer func() { resp.Duration = time.Since(start) }() + mem := <-ws.memMu + defer func() { ws.memMu <- mem }() + vals, err := unmarshalCorpusFile(mem.valueCopy()) + if err != nil { + panic(err) + } + inpHash := sha256.Sum256(mem.valueCopy()) + if args.Timeout != 0 { + var cancel func() + ctx, cancel = context.WithTimeout(ctx, args.Timeout) + defer cancel() + } + + // Minimize the values in vals, then write to shared memory. We only write + // to shared memory after completing minimization. + success, err := ws.minimizeInput(ctx, vals, mem, args) + if success { + writeToMem(vals, mem) + outHash := sha256.Sum256(mem.valueCopy()) + mem.header().rawInMem = false + resp.WroteToMem = true + if err != nil { + resp.Err = err.Error() + } else { + // If the values didn't change during minimization then coverageSnapshot is likely + // a dirty snapshot which represents the very last step of minimization, not the + // coverage for the initial input. In that case just return the coverage we were + // given initially, since it more accurately represents the coverage map for the + // input we are returning. + if outHash != inpHash { + resp.CoverageData = coverageSnapshot + } else { + resp.CoverageData = args.KeepCoverage + } + } + } + return resp +} + +// minimizeInput applies a series of minimizing transformations on the provided +// vals, ensuring that each minimization still causes an error, or keeps +// coverage, in fuzzFn. It uses the context to determine how long to run, +// stopping once closed. It returns a bool indicating whether minimization was +// successful and an error if one was found. +func (ws *workerServer) minimizeInput(ctx context.Context, vals []any, mem *sharedMem, args minimizeArgs) (success bool, retErr error) { + keepCoverage := args.KeepCoverage + memBytes := mem.valueRef() + bPtr := &memBytes + count := &mem.header().count + shouldStop := func() bool { + return ctx.Err() != nil || + (args.Limit > 0 && *count >= args.Limit) + } + if shouldStop() { + return false, nil + } + + // Check that the original value preserves coverage or causes an error. + // If not, then whatever caused us to think the value was interesting may + // have been a flake, and we can't minimize it. + *count++ + _, retErr = ws.fuzzFn(CorpusEntry{Values: vals}) + if keepCoverage != nil { + if !hasCoverageBit(keepCoverage, coverageSnapshot) || retErr != nil { + return false, nil + } + } else if retErr == nil { + return false, nil + } + mem.header().rawInMem = true + + // tryMinimized runs the fuzz function with candidate replacing the value + // at index valI. tryMinimized returns whether the input with candidate is + // interesting for the same reason as the original input: it returns + // an error if one was expected, or it preserves coverage. + tryMinimized := func(candidate []byte) bool { + prev := vals[args.Index] + switch prev.(type) { + case []byte: + vals[args.Index] = candidate + case string: + vals[args.Index] = string(candidate) + default: + panic("impossible") + } + copy(*bPtr, candidate) + *bPtr = (*bPtr)[:len(candidate)] + mem.setValueLen(len(candidate)) + *count++ + _, err := ws.fuzzFn(CorpusEntry{Values: vals}) + if err != nil { + retErr = err + if keepCoverage != nil { + // Now that we've found a crash, that's more important than any + // minimization of interesting inputs that was being done. Clear out + // keepCoverage to only minimize the crash going forward. + keepCoverage = nil + } + return true + } + // Minimization should preserve coverage bits. + if keepCoverage != nil && isCoverageSubset(keepCoverage, coverageSnapshot) { + return true + } + vals[args.Index] = prev + return false + } + switch v := vals[args.Index].(type) { + case string: + minimizeBytes([]byte(v), tryMinimized, shouldStop) + case []byte: + minimizeBytes(v, tryMinimized, shouldStop) + default: + panic("impossible") + } + return true, retErr +} + +func writeToMem(vals []any, mem *sharedMem) { + b := marshalCorpusFile(vals...) + mem.setValue(b) +} + +// ping does nothing. The coordinator calls this method to ensure the worker +// has called F.Fuzz and can communicate. +func (ws *workerServer) ping(ctx context.Context, args pingArgs) pingResponse { + return pingResponse{} +} + +// workerClient is a minimalist RPC client. The coordinator process uses a +// workerClient to call methods in each worker process (handled by +// workerServer). +type workerClient struct { + workerComm + m *mutator + + // mu is the mutex protecting the workerComm.fuzzIn pipe. This must be + // locked before making calls to the workerServer. It prevents + // workerClient.Close from closing fuzzIn while workerClient methods are + // writing to it concurrently, and prevents multiple callers from writing to + // fuzzIn concurrently. + mu sync.Mutex +} + +func newWorkerClient(comm workerComm, m *mutator) *workerClient { + return &workerClient{workerComm: comm, m: m} +} + +// Close shuts down the connection to the RPC server (the worker process) by +// closing fuzz_in. Close drains fuzz_out (avoiding a SIGPIPE in the worker), +// and closes it after the worker process closes the other end. +func (wc *workerClient) Close() error { + wc.mu.Lock() + defer wc.mu.Unlock() + + // Close fuzzIn. This signals to the server that there are no more calls, + // and it should exit. + if err := wc.fuzzIn.Close(); err != nil { + wc.fuzzOut.Close() + return err + } + + // Drain fuzzOut and close it. When the server exits, the kernel will close + // its end of fuzzOut, and we'll get EOF. + if _, err := io.Copy(io.Discard, wc.fuzzOut); err != nil { + wc.fuzzOut.Close() + return err + } + return wc.fuzzOut.Close() +} + +// errSharedMemClosed is returned by workerClient methods that cannot access +// shared memory because it was closed and unmapped by another goroutine. That +// can happen when worker.cleanup is called in the worker goroutine while a +// workerClient.fuzz call runs concurrently. +// +// This error should not be reported. It indicates the operation was +// interrupted. +var errSharedMemClosed = errors.New("internal error: shared memory was closed and unmapped") + +// minimize tells the worker to call the minimize method. See +// workerServer.minimize. +func (wc *workerClient) minimize(ctx context.Context, entryIn CorpusEntry, args minimizeArgs) (entryOut CorpusEntry, resp minimizeResponse, retErr error) { + wc.mu.Lock() + defer wc.mu.Unlock() + + mem, ok := <-wc.memMu + if !ok { + return CorpusEntry{}, minimizeResponse{}, errSharedMemClosed + } + defer func() { wc.memMu <- mem }() + mem.header().count = 0 + inp, err := corpusEntryData(entryIn) + if err != nil { + return CorpusEntry{}, minimizeResponse{}, err + } + mem.setValue(inp) + entryOut = entryIn + entryOut.Values, err = unmarshalCorpusFile(inp) + if err != nil { + return CorpusEntry{}, minimizeResponse{}, fmt.Errorf("workerClient.minimize unmarshaling provided value: %v", err) + } + for i, v := range entryOut.Values { + if !isMinimizable(reflect.TypeOf(v)) { + continue + } + + wc.memMu <- mem + args.Index = i + c := call{Minimize: &args} + callErr := wc.callLocked(ctx, c, &resp) + mem, ok = <-wc.memMu + if !ok { + return CorpusEntry{}, minimizeResponse{}, errSharedMemClosed + } + + if callErr != nil { + retErr = callErr + if !mem.header().rawInMem { + // An unrecoverable error occurred before minimization began. + return entryIn, minimizeResponse{}, retErr + } + // An unrecoverable error occurred during minimization. mem now + // holds the raw, unmarshalled bytes of entryIn.Values[i] that + // caused the error. + switch entryOut.Values[i].(type) { + case string: + entryOut.Values[i] = string(mem.valueCopy()) + case []byte: + entryOut.Values[i] = mem.valueCopy() + default: + panic("impossible") + } + entryOut.Data = marshalCorpusFile(entryOut.Values...) + // Stop minimizing; another unrecoverable error is likely to occur. + break + } + + if resp.WroteToMem { + // Minimization succeeded, and mem holds the marshaled data. + entryOut.Data = mem.valueCopy() + entryOut.Values, err = unmarshalCorpusFile(entryOut.Data) + if err != nil { + return CorpusEntry{}, minimizeResponse{}, fmt.Errorf("workerClient.minimize unmarshaling minimized value: %v", err) + } + } + + // Prepare for next iteration of the loop. + if args.Timeout != 0 { + args.Timeout -= resp.Duration + if args.Timeout <= 0 { + break + } + } + if args.Limit != 0 { + args.Limit -= mem.header().count + if args.Limit <= 0 { + break + } + } + } + resp.Count = mem.header().count + h := sha256.Sum256(entryOut.Data) + entryOut.Path = fmt.Sprintf("%x", h[:4]) + return entryOut, resp, retErr +} + +// fuzz tells the worker to call the fuzz method. See workerServer.fuzz. +func (wc *workerClient) fuzz(ctx context.Context, entryIn CorpusEntry, args fuzzArgs) (entryOut CorpusEntry, resp fuzzResponse, isInternalError bool, err error) { + wc.mu.Lock() + defer wc.mu.Unlock() + + mem, ok := <-wc.memMu + if !ok { + return CorpusEntry{}, fuzzResponse{}, true, errSharedMemClosed + } + mem.header().count = 0 + inp, err := corpusEntryData(entryIn) + if err != nil { + wc.memMu <- mem + return CorpusEntry{}, fuzzResponse{}, true, err + } + mem.setValue(inp) + wc.memMu <- mem + + c := call{Fuzz: &args} + callErr := wc.callLocked(ctx, c, &resp) + if resp.InternalErr != "" { + return CorpusEntry{}, fuzzResponse{}, true, errors.New(resp.InternalErr) + } + mem, ok = <-wc.memMu + if !ok { + return CorpusEntry{}, fuzzResponse{}, true, errSharedMemClosed + } + defer func() { wc.memMu <- mem }() + resp.Count = mem.header().count + + if !bytes.Equal(inp, mem.valueRef()) { + return CorpusEntry{}, fuzzResponse{}, true, errors.New("workerServer.fuzz modified input") + } + needEntryOut := callErr != nil || resp.Err != "" || + (!args.Warmup && resp.CoverageData != nil) + if needEntryOut { + valuesOut, err := unmarshalCorpusFile(inp) + if err != nil { + return CorpusEntry{}, fuzzResponse{}, true, fmt.Errorf("unmarshaling fuzz input value after call: %v", err) + } + wc.m.r.restore(mem.header().randState, mem.header().randInc) + if !args.Warmup { + // Only mutate the valuesOut if fuzzing actually occurred. + numMutations := ((resp.Count - 1) % chainedMutations) + 1 + for i := int64(0); i < numMutations; i++ { + wc.m.mutate(valuesOut, cap(mem.valueRef())) + } + } + dataOut := marshalCorpusFile(valuesOut...) + + h := sha256.Sum256(dataOut) + name := fmt.Sprintf("%x", h[:4]) + entryOut = CorpusEntry{ + Parent: entryIn.Path, + Path: name, + Data: dataOut, + Generation: entryIn.Generation + 1, + } + if args.Warmup { + // The bytes weren't mutated, so if entryIn was a seed corpus value, + // then entryOut is too. + entryOut.IsSeed = entryIn.IsSeed + } + } + + return entryOut, resp, false, callErr +} + +// ping tells the worker to call the ping method. See workerServer.ping. +func (wc *workerClient) ping(ctx context.Context) error { + wc.mu.Lock() + defer wc.mu.Unlock() + c := call{Ping: &pingArgs{}} + var resp pingResponse + return wc.callLocked(ctx, c, &resp) +} + +// callLocked sends an RPC from the coordinator to the worker process and waits +// for the response. The callLocked may be cancelled with ctx. +func (wc *workerClient) callLocked(ctx context.Context, c call, resp any) (err error) { + enc := json.NewEncoder(wc.fuzzIn) + dec := json.NewDecoder(&contextReader{ctx: ctx, r: wc.fuzzOut}) + if err := enc.Encode(c); err != nil { + return err + } + return dec.Decode(resp) +} + +// contextReader wraps a Reader with a Context. If the context is cancelled +// while the underlying reader is blocked, Read returns immediately. +// +// This is useful for reading from a pipe. Closing a pipe file descriptor does +// not unblock pending Reads on that file descriptor. All copies of the pipe's +// other file descriptor (the write end) must be closed in all processes that +// inherit it. This is difficult to do correctly in the situation we care about +// (process group termination). +type contextReader struct { + ctx context.Context + r io.Reader +} + +func (cr *contextReader) Read(b []byte) (int, error) { + if ctxErr := cr.ctx.Err(); ctxErr != nil { + return 0, ctxErr + } + done := make(chan struct{}) + + // This goroutine may stay blocked after Read returns because the underlying + // read is blocked. + var n int + var err error + go func() { + n, err = cr.r.Read(b) + close(done) + }() + + select { + case <-cr.ctx.Done(): + return 0, cr.ctx.Err() + case <-done: + return n, err + } +} diff --git a/src/internal/fuzz/worker_test.go b/src/internal/fuzz/worker_test.go new file mode 100644 index 0000000..d0b21da --- /dev/null +++ b/src/internal/fuzz/worker_test.go @@ -0,0 +1,206 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package fuzz + +import ( + "context" + "errors" + "flag" + "fmt" + "internal/race" + "io" + "os" + "os/signal" + "reflect" + "strconv" + "testing" + "time" +) + +var benchmarkWorkerFlag = flag.Bool("benchmarkworker", false, "") + +func TestMain(m *testing.M) { + flag.Parse() + if *benchmarkWorkerFlag { + runBenchmarkWorker() + return + } + os.Exit(m.Run()) +} + +func BenchmarkWorkerFuzzOverhead(b *testing.B) { + if race.Enabled { + b.Skip("TODO(48504): fix and re-enable") + } + origEnv := os.Getenv("GODEBUG") + defer func() { os.Setenv("GODEBUG", origEnv) }() + os.Setenv("GODEBUG", fmt.Sprintf("%s,fuzzseed=123", origEnv)) + + ws := &workerServer{ + fuzzFn: func(_ CorpusEntry) (time.Duration, error) { return time.Second, nil }, + workerComm: workerComm{memMu: make(chan *sharedMem, 1)}, + } + + mem, err := sharedMemTempFile(workerSharedMemSize) + if err != nil { + b.Fatalf("failed to create temporary shared memory file: %s", err) + } + defer func() { + if err := mem.Close(); err != nil { + b.Error(err) + } + }() + + initialVal := []any{make([]byte, 32)} + encodedVals := marshalCorpusFile(initialVal...) + mem.setValue(encodedVals) + + ws.memMu <- mem + + b.ResetTimer() + for i := 0; i < b.N; i++ { + ws.m = newMutator() + mem.setValue(encodedVals) + mem.header().count = 0 + + ws.fuzz(context.Background(), fuzzArgs{Limit: 1}) + } +} + +// BenchmarkWorkerPing acts as the coordinator and measures the time it takes +// a worker to respond to N pings. This is a rough measure of our RPC latency. +func BenchmarkWorkerPing(b *testing.B) { + if race.Enabled { + b.Skip("TODO(48504): fix and re-enable") + } + b.SetParallelism(1) + w := newWorkerForTest(b) + for i := 0; i < b.N; i++ { + if err := w.client.ping(context.Background()); err != nil { + b.Fatal(err) + } + } +} + +// BenchmarkWorkerFuzz acts as the coordinator and measures the time it takes +// a worker to mutate a given input and call a trivial fuzz function N times. +func BenchmarkWorkerFuzz(b *testing.B) { + if race.Enabled { + b.Skip("TODO(48504): fix and re-enable") + } + b.SetParallelism(1) + w := newWorkerForTest(b) + entry := CorpusEntry{Values: []any{[]byte(nil)}} + entry.Data = marshalCorpusFile(entry.Values...) + for i := int64(0); i < int64(b.N); { + args := fuzzArgs{ + Limit: int64(b.N) - i, + Timeout: workerFuzzDuration, + } + _, resp, _, err := w.client.fuzz(context.Background(), entry, args) + if err != nil { + b.Fatal(err) + } + if resp.Err != "" { + b.Fatal(resp.Err) + } + if resp.Count == 0 { + b.Fatal("worker did not make progress") + } + i += resp.Count + } +} + +// newWorkerForTest creates and starts a worker process for testing or +// benchmarking. The worker process calls RunFuzzWorker, which responds to +// RPC messages until it's stopped. The process is stopped and cleaned up +// automatically when the test is done. +func newWorkerForTest(tb testing.TB) *worker { + tb.Helper() + c, err := newCoordinator(CoordinateFuzzingOpts{ + Types: []reflect.Type{reflect.TypeOf([]byte(nil))}, + Log: io.Discard, + }) + if err != nil { + tb.Fatal(err) + } + dir := "" // same as self + binPath := os.Args[0] // same as self + args := append(os.Args[1:], "-benchmarkworker") + env := os.Environ() // same as self + w, err := newWorker(c, dir, binPath, args, env) + if err != nil { + tb.Fatal(err) + } + tb.Cleanup(func() { + if err := w.cleanup(); err != nil { + tb.Error(err) + } + }) + if err := w.startAndPing(context.Background()); err != nil { + tb.Fatal(err) + } + tb.Cleanup(func() { + if err := w.stop(); err != nil { + tb.Error(err) + } + }) + return w +} + +func runBenchmarkWorker() { + ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt) + defer cancel() + fn := func(CorpusEntry) error { return nil } + if err := RunFuzzWorker(ctx, fn); err != nil && err != ctx.Err() { + panic(err) + } +} + +func BenchmarkWorkerMinimize(b *testing.B) { + if race.Enabled { + b.Skip("TODO(48504): fix and re-enable") + } + + ws := &workerServer{ + workerComm: workerComm{memMu: make(chan *sharedMem, 1)}, + } + + mem, err := sharedMemTempFile(workerSharedMemSize) + if err != nil { + b.Fatalf("failed to create temporary shared memory file: %s", err) + } + defer func() { + if err := mem.Close(); err != nil { + b.Error(err) + } + }() + ws.memMu <- mem + + bytes := make([]byte, 1024) + ctx := context.Background() + for sz := 1; sz <= len(bytes); sz <<= 1 { + sz := sz + input := []any{bytes[:sz]} + encodedVals := marshalCorpusFile(input...) + mem = <-ws.memMu + mem.setValue(encodedVals) + ws.memMu <- mem + b.Run(strconv.Itoa(sz), func(b *testing.B) { + i := 0 + ws.fuzzFn = func(_ CorpusEntry) (time.Duration, error) { + if i == 0 { + i++ + return time.Second, errors.New("initial failure for deflake") + } + return time.Second, nil + } + for i := 0; i < b.N; i++ { + b.SetBytes(int64(sz)) + ws.minimize(ctx, minimizeArgs{}) + } + }) + } +} diff --git a/src/internal/goarch/gengoarch.go b/src/internal/goarch/gengoarch.go new file mode 100644 index 0000000..0b0be5c --- /dev/null +++ b/src/internal/goarch/gengoarch.go @@ -0,0 +1,60 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ignore + +package main + +import ( + "bytes" + "fmt" + "log" + "os" + "strings" +) + +var goarches []string + +func main() { + data, err := os.ReadFile("../../go/build/syslist.go") + if err != nil { + log.Fatal(err) + } + const goarchPrefix = `var knownArch = map[string]bool{` + inGOARCH := false + for _, line := range strings.Split(string(data), "\n") { + if strings.HasPrefix(line, goarchPrefix) { + inGOARCH = true + } else if inGOARCH && strings.HasPrefix(line, "}") { + break + } else if inGOARCH { + goarch := strings.Fields(line)[0] + goarch = strings.TrimPrefix(goarch, `"`) + goarch = strings.TrimSuffix(goarch, `":`) + goarches = append(goarches, goarch) + } + } + + for _, target := range goarches { + if target == "amd64p32" { + continue + } + var buf bytes.Buffer + fmt.Fprintf(&buf, "// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT.\n\n") + fmt.Fprintf(&buf, "//go:build %s\n\n", target) // must explicitly include target for bootstrapping purposes + fmt.Fprintf(&buf, "package goarch\n\n") + fmt.Fprintf(&buf, "const GOARCH = `%s`\n\n", target) + for _, goarch := range goarches { + value := 0 + if goarch == target { + value = 1 + } + fmt.Fprintf(&buf, "const Is%s = %d\n", strings.Title(goarch), value) + } + err := os.WriteFile("zgoarch_"+target+".go", buf.Bytes(), 0666) + if err != nil { + log.Fatal(err) + } + } +} diff --git a/src/internal/goarch/goarch.go b/src/internal/goarch/goarch.go new file mode 100644 index 0000000..3dda62f --- /dev/null +++ b/src/internal/goarch/goarch.go @@ -0,0 +1,60 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// package goarch contains GOARCH-specific constants. +package goarch + +// The next line makes 'go generate' write the zgoarch*.go files with +// per-arch information, including constants named $GOARCH for every +// GOARCH. The constant is 1 on the current system, 0 otherwise; multiplying +// by them is useful for defining GOARCH-specific constants. +// +//go:generate go run gengoarch.go + +type ArchFamilyType int + +const ( + AMD64 ArchFamilyType = iota + ARM + ARM64 + I386 + LOONG64 + MIPS + MIPS64 + PPC64 + RISCV64 + S390X + WASM +) + +// PtrSize is the size of a pointer in bytes - unsafe.Sizeof(uintptr(0)) but as an ideal constant. +// It is also the size of the machine's native word size (that is, 4 on 32-bit systems, 8 on 64-bit). +const PtrSize = 4 << (^uintptr(0) >> 63) + +// ArchFamily is the architecture family (AMD64, ARM, ...) +const ArchFamily ArchFamilyType = _ArchFamily + +// BigEndian reports whether the architecture is big-endian. +const BigEndian = IsArmbe|IsArm64be|IsMips|IsMips64|IsPpc|IsPpc64|IsS390|IsS390x|IsSparc|IsSparc64 == 1 + +// DefaultPhysPageSize is the default physical page size. +const DefaultPhysPageSize = _DefaultPhysPageSize + +// PCQuantum is the minimal unit for a program counter (1 on x86, 4 on most other systems). +// The various PC tables record PC deltas pre-divided by PCQuantum. +const PCQuantum = _PCQuantum + +// Int64Align is the required alignment for a 64-bit integer (4 on 32-bit systems, 8 on 64-bit). +const Int64Align = PtrSize + +// MinFrameSize is the size of the system-reserved words at the bottom +// of a frame (just above the architectural stack pointer). +// It is zero on x86 and PtrSize on most non-x86 (LR-based) systems. +// On PowerPC it is larger, to cover three more reserved words: +// the compiler word, the link editor word, and the TOC save word. +const MinFrameSize = _MinFrameSize + +// StackAlign is the required alignment of the SP register. +// The stack must be at least word aligned, but some architectures require more. +const StackAlign = _StackAlign diff --git a/src/internal/goarch/goarch_386.go b/src/internal/goarch/goarch_386.go new file mode 100644 index 0000000..c621421 --- /dev/null +++ b/src/internal/goarch/goarch_386.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = I386 + _DefaultPhysPageSize = 4096 + _PCQuantum = 1 + _MinFrameSize = 0 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_amd64.go b/src/internal/goarch/goarch_amd64.go new file mode 100644 index 0000000..911e3e7 --- /dev/null +++ b/src/internal/goarch/goarch_amd64.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = AMD64 + _DefaultPhysPageSize = 4096 + _PCQuantum = 1 + _MinFrameSize = 0 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_arm.go b/src/internal/goarch/goarch_arm.go new file mode 100644 index 0000000..a659171 --- /dev/null +++ b/src/internal/goarch/goarch_arm.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = ARM + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 4 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_arm64.go b/src/internal/goarch/goarch_arm64.go new file mode 100644 index 0000000..85d0b47 --- /dev/null +++ b/src/internal/goarch/goarch_arm64.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = ARM64 + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 8 + _StackAlign = 16 +) diff --git a/src/internal/goarch/goarch_loong64.go b/src/internal/goarch/goarch_loong64.go new file mode 100644 index 0000000..dae1f4d --- /dev/null +++ b/src/internal/goarch/goarch_loong64.go @@ -0,0 +1,15 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build loong64 + +package goarch + +const ( + _ArchFamily = LOONG64 + _DefaultPhysPageSize = 16384 + _PCQuantum = 4 + _MinFrameSize = 8 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_mips.go b/src/internal/goarch/goarch_mips.go new file mode 100644 index 0000000..59f3995 --- /dev/null +++ b/src/internal/goarch/goarch_mips.go @@ -0,0 +1,13 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = MIPS + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 4 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_mips64.go b/src/internal/goarch/goarch_mips64.go new file mode 100644 index 0000000..9e4f827 --- /dev/null +++ b/src/internal/goarch/goarch_mips64.go @@ -0,0 +1,13 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = MIPS64 + _DefaultPhysPageSize = 16384 + _PCQuantum = 4 + _MinFrameSize = 8 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_mips64le.go b/src/internal/goarch/goarch_mips64le.go new file mode 100644 index 0000000..9e4f827 --- /dev/null +++ b/src/internal/goarch/goarch_mips64le.go @@ -0,0 +1,13 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = MIPS64 + _DefaultPhysPageSize = 16384 + _PCQuantum = 4 + _MinFrameSize = 8 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_mipsle.go b/src/internal/goarch/goarch_mipsle.go new file mode 100644 index 0000000..3e6642b --- /dev/null +++ b/src/internal/goarch/goarch_mipsle.go @@ -0,0 +1,13 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = MIPS + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 4 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_ppc64.go b/src/internal/goarch/goarch_ppc64.go new file mode 100644 index 0000000..60cc846 --- /dev/null +++ b/src/internal/goarch/goarch_ppc64.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = PPC64 + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 32 + _StackAlign = 16 +) diff --git a/src/internal/goarch/goarch_ppc64le.go b/src/internal/goarch/goarch_ppc64le.go new file mode 100644 index 0000000..60cc846 --- /dev/null +++ b/src/internal/goarch/goarch_ppc64le.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = PPC64 + _DefaultPhysPageSize = 65536 + _PCQuantum = 4 + _MinFrameSize = 32 + _StackAlign = 16 +) diff --git a/src/internal/goarch/goarch_riscv64.go b/src/internal/goarch/goarch_riscv64.go new file mode 100644 index 0000000..3b6da1e --- /dev/null +++ b/src/internal/goarch/goarch_riscv64.go @@ -0,0 +1,13 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = RISCV64 + _DefaultPhysPageSize = 4096 + _PCQuantum = 4 + _MinFrameSize = 8 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_s390x.go b/src/internal/goarch/goarch_s390x.go new file mode 100644 index 0000000..20c5705 --- /dev/null +++ b/src/internal/goarch/goarch_s390x.go @@ -0,0 +1,13 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = S390X + _DefaultPhysPageSize = 4096 + _PCQuantum = 2 + _MinFrameSize = 8 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/goarch_wasm.go b/src/internal/goarch/goarch_wasm.go new file mode 100644 index 0000000..98618d6 --- /dev/null +++ b/src/internal/goarch/goarch_wasm.go @@ -0,0 +1,13 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goarch + +const ( + _ArchFamily = WASM + _DefaultPhysPageSize = 65536 + _PCQuantum = 1 + _MinFrameSize = 0 + _StackAlign = PtrSize +) diff --git a/src/internal/goarch/zgoarch_386.go b/src/internal/goarch/zgoarch_386.go new file mode 100644 index 0000000..4a9b0e6 --- /dev/null +++ b/src/internal/goarch/zgoarch_386.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build 386 + +package goarch + +const GOARCH = `386` + +const Is386 = 1 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_amd64.go b/src/internal/goarch/zgoarch_amd64.go new file mode 100644 index 0000000..7926392 --- /dev/null +++ b/src/internal/goarch/zgoarch_amd64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build amd64 + +package goarch + +const GOARCH = `amd64` + +const Is386 = 0 +const IsAmd64 = 1 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_arm.go b/src/internal/goarch/zgoarch_arm.go new file mode 100644 index 0000000..6c03b8b --- /dev/null +++ b/src/internal/goarch/zgoarch_arm.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build arm + +package goarch + +const GOARCH = `arm` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 1 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_arm64.go b/src/internal/goarch/zgoarch_arm64.go new file mode 100644 index 0000000..ad342d7 --- /dev/null +++ b/src/internal/goarch/zgoarch_arm64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build arm64 + +package goarch + +const GOARCH = `arm64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 1 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_arm64be.go b/src/internal/goarch/zgoarch_arm64be.go new file mode 100644 index 0000000..0f26003 --- /dev/null +++ b/src/internal/goarch/zgoarch_arm64be.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build arm64be + +package goarch + +const GOARCH = `arm64be` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 1 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_armbe.go b/src/internal/goarch/zgoarch_armbe.go new file mode 100644 index 0000000..6092fee --- /dev/null +++ b/src/internal/goarch/zgoarch_armbe.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build armbe + +package goarch + +const GOARCH = `armbe` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 1 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_loong64.go b/src/internal/goarch/zgoarch_loong64.go new file mode 100644 index 0000000..21c67e1 --- /dev/null +++ b/src/internal/goarch/zgoarch_loong64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build loong64 + +package goarch + +const GOARCH = `loong64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 1 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mips.go b/src/internal/goarch/zgoarch_mips.go new file mode 100644 index 0000000..0db1974 --- /dev/null +++ b/src/internal/goarch/zgoarch_mips.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mips + +package goarch + +const GOARCH = `mips` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 1 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mips64.go b/src/internal/goarch/zgoarch_mips64.go new file mode 100644 index 0000000..738806f --- /dev/null +++ b/src/internal/goarch/zgoarch_mips64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mips64 + +package goarch + +const GOARCH = `mips64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 1 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mips64le.go b/src/internal/goarch/zgoarch_mips64le.go new file mode 100644 index 0000000..8de5beb --- /dev/null +++ b/src/internal/goarch/zgoarch_mips64le.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mips64le + +package goarch + +const GOARCH = `mips64le` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 1 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mips64p32.go b/src/internal/goarch/zgoarch_mips64p32.go new file mode 100644 index 0000000..ea461be --- /dev/null +++ b/src/internal/goarch/zgoarch_mips64p32.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mips64p32 + +package goarch + +const GOARCH = `mips64p32` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 1 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mips64p32le.go b/src/internal/goarch/zgoarch_mips64p32le.go new file mode 100644 index 0000000..15473ce --- /dev/null +++ b/src/internal/goarch/zgoarch_mips64p32le.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mips64p32le + +package goarch + +const GOARCH = `mips64p32le` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 1 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_mipsle.go b/src/internal/goarch/zgoarch_mipsle.go new file mode 100644 index 0000000..4955142 --- /dev/null +++ b/src/internal/goarch/zgoarch_mipsle.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build mipsle + +package goarch + +const GOARCH = `mipsle` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 1 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_ppc.go b/src/internal/goarch/zgoarch_ppc.go new file mode 100644 index 0000000..ec01763 --- /dev/null +++ b/src/internal/goarch/zgoarch_ppc.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build ppc + +package goarch + +const GOARCH = `ppc` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 1 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_ppc64.go b/src/internal/goarch/zgoarch_ppc64.go new file mode 100644 index 0000000..39be392 --- /dev/null +++ b/src/internal/goarch/zgoarch_ppc64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build ppc64 + +package goarch + +const GOARCH = `ppc64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 1 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_ppc64le.go b/src/internal/goarch/zgoarch_ppc64le.go new file mode 100644 index 0000000..5f959e0 --- /dev/null +++ b/src/internal/goarch/zgoarch_ppc64le.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build ppc64le + +package goarch + +const GOARCH = `ppc64le` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 1 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_riscv.go b/src/internal/goarch/zgoarch_riscv.go new file mode 100644 index 0000000..8d81a14 --- /dev/null +++ b/src/internal/goarch/zgoarch_riscv.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build riscv + +package goarch + +const GOARCH = `riscv` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 1 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_riscv64.go b/src/internal/goarch/zgoarch_riscv64.go new file mode 100644 index 0000000..1df989c --- /dev/null +++ b/src/internal/goarch/zgoarch_riscv64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build riscv64 + +package goarch + +const GOARCH = `riscv64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 1 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_s390.go b/src/internal/goarch/zgoarch_s390.go new file mode 100644 index 0000000..56815b9 --- /dev/null +++ b/src/internal/goarch/zgoarch_s390.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build s390 + +package goarch + +const GOARCH = `s390` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 1 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_s390x.go b/src/internal/goarch/zgoarch_s390x.go new file mode 100644 index 0000000..e61e9bd --- /dev/null +++ b/src/internal/goarch/zgoarch_s390x.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build s390x + +package goarch + +const GOARCH = `s390x` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 1 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_sparc.go b/src/internal/goarch/zgoarch_sparc.go new file mode 100644 index 0000000..ee5b746 --- /dev/null +++ b/src/internal/goarch/zgoarch_sparc.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build sparc + +package goarch + +const GOARCH = `sparc` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 1 +const IsSparc64 = 0 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_sparc64.go b/src/internal/goarch/zgoarch_sparc64.go new file mode 100644 index 0000000..519aaa1 --- /dev/null +++ b/src/internal/goarch/zgoarch_sparc64.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build sparc64 + +package goarch + +const GOARCH = `sparc64` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 1 +const IsWasm = 0 diff --git a/src/internal/goarch/zgoarch_wasm.go b/src/internal/goarch/zgoarch_wasm.go new file mode 100644 index 0000000..25567a1 --- /dev/null +++ b/src/internal/goarch/zgoarch_wasm.go @@ -0,0 +1,32 @@ +// Code generated by gengoarch.go using 'go generate'. DO NOT EDIT. + +//go:build wasm + +package goarch + +const GOARCH = `wasm` + +const Is386 = 0 +const IsAmd64 = 0 +const IsAmd64p32 = 0 +const IsArm = 0 +const IsArmbe = 0 +const IsArm64 = 0 +const IsArm64be = 0 +const IsLoong64 = 0 +const IsMips = 0 +const IsMipsle = 0 +const IsMips64 = 0 +const IsMips64le = 0 +const IsMips64p32 = 0 +const IsMips64p32le = 0 +const IsPpc = 0 +const IsPpc64 = 0 +const IsPpc64le = 0 +const IsRiscv = 0 +const IsRiscv64 = 0 +const IsS390 = 0 +const IsS390x = 0 +const IsSparc = 0 +const IsSparc64 = 0 +const IsWasm = 1 diff --git a/src/internal/godebug/godebug.go b/src/internal/godebug/godebug.go new file mode 100644 index 0000000..36bfeac --- /dev/null +++ b/src/internal/godebug/godebug.go @@ -0,0 +1,290 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package godebug makes the settings in the $GODEBUG environment variable +// available to other packages. These settings are often used for compatibility +// tweaks, when we need to change a default behavior but want to let users +// opt back in to the original. For example GODEBUG=http2server=0 disables +// HTTP/2 support in the net/http server. +// +// In typical usage, code should declare a Setting as a global +// and then call Value each time the current setting value is needed: +// +// var http2server = godebug.New("http2server") +// +// func ServeConn(c net.Conn) { +// if http2server.Value() == "0" { +// disallow HTTP/2 +// ... +// } +// ... +// } +// +// Each time a non-default setting causes a change in program behavior, +// code should call [Setting.IncNonDefault] to increment a counter that can +// be reported by [runtime/metrics.Read]. +// Note that counters used with IncNonDefault must be added to +// various tables in other packages. See the [Setting.IncNonDefault] +// documentation for details. +package godebug + +// Note: Be careful about new imports here. Any package +// that internal/godebug imports cannot itself import internal/godebug, +// meaning it cannot introduce a GODEBUG setting of its own. +// We keep imports to the absolute bare minimum. +import ( + "internal/bisect" + "internal/godebugs" + "sync" + "sync/atomic" + "unsafe" + _ "unsafe" // go:linkname +) + +// A Setting is a single setting in the $GODEBUG environment variable. +type Setting struct { + name string + once sync.Once + *setting +} + +type setting struct { + value atomic.Pointer[value] + nonDefaultOnce sync.Once + nonDefault atomic.Uint64 + info *godebugs.Info +} + +type value struct { + text string + bisect *bisect.Matcher +} + +// New returns a new Setting for the $GODEBUG setting with the given name. +// +// GODEBUGs meant for use by end users must be listed in ../godebugs/table.go, +// which is used for generating and checking various documentation. +// If the name is not listed in that table, New will succeed but calling Value +// on the returned Setting will panic. +// To disable that panic for access to an undocumented setting, +// prefix the name with a #, as in godebug.New("#gofsystrace"). +// The # is a signal to New but not part of the key used in $GODEBUG. +func New(name string) *Setting { + return &Setting{name: name} +} + +// Name returns the name of the setting. +func (s *Setting) Name() string { + if s.name != "" && s.name[0] == '#' { + return s.name[1:] + } + return s.name +} + +// Undocumented reports whether this is an undocumented setting. +func (s *Setting) Undocumented() bool { + return s.name != "" && s.name[0] == '#' +} + +// String returns a printable form for the setting: name=value. +func (s *Setting) String() string { + return s.Name() + "=" + s.Value() +} + +// IncNonDefault increments the non-default behavior counter +// associated with the given setting. +// This counter is exposed in the runtime/metrics value +// /godebug/non-default-behavior/<name>:events. +// +// Note that Value must be called at least once before IncNonDefault. +func (s *Setting) IncNonDefault() { + s.nonDefaultOnce.Do(s.register) + s.nonDefault.Add(1) +} + +func (s *Setting) register() { + if s.info == nil || s.info.Opaque { + panic("godebug: unexpected IncNonDefault of " + s.name) + } + registerMetric("/godebug/non-default-behavior/"+s.Name()+":events", s.nonDefault.Load) +} + +// cache is a cache of all the GODEBUG settings, +// a locked map[string]*atomic.Pointer[string]. +// +// All Settings with the same name share a single +// *atomic.Pointer[string], so that when GODEBUG +// changes only that single atomic string pointer +// needs to be updated. +// +// A name appears in the values map either if it is the +// name of a Setting for which Value has been called +// at least once, or if the name has ever appeared in +// a name=value pair in the $GODEBUG environment variable. +// Once entered into the map, the name is never removed. +var cache sync.Map // name string -> value *atomic.Pointer[string] + +var empty value + +// Value returns the current value for the GODEBUG setting s. +// +// Value maintains an internal cache that is synchronized +// with changes to the $GODEBUG environment variable, +// making Value efficient to call as frequently as needed. +// Clients should therefore typically not attempt their own +// caching of Value's result. +func (s *Setting) Value() string { + s.once.Do(func() { + s.setting = lookup(s.Name()) + if s.info == nil && !s.Undocumented() { + panic("godebug: Value of name not listed in godebugs.All: " + s.name) + } + }) + v := *s.value.Load() + if v.bisect != nil && !v.bisect.Stack(&stderr) { + return "" + } + return v.text +} + +// lookup returns the unique *setting value for the given name. +func lookup(name string) *setting { + if v, ok := cache.Load(name); ok { + return v.(*setting) + } + s := new(setting) + s.info = godebugs.Lookup(name) + s.value.Store(&empty) + if v, loaded := cache.LoadOrStore(name, s); loaded { + // Lost race: someone else created it. Use theirs. + return v.(*setting) + } + + return s +} + +// setUpdate is provided by package runtime. +// It calls update(def, env), where def is the default GODEBUG setting +// and env is the current value of the $GODEBUG environment variable. +// After that first call, the runtime calls update(def, env) +// again each time the environment variable changes +// (due to use of os.Setenv, for example). +// +//go:linkname setUpdate +func setUpdate(update func(string, string)) + +// registerMetric is provided by package runtime. +// It forwards registrations to runtime/metrics. +// +//go:linkname registerMetric +func registerMetric(name string, read func() uint64) + +// setNewIncNonDefault is provided by package runtime. +// The runtime can do +// +// inc := newNonDefaultInc(name) +// +// instead of +// +// inc := godebug.New(name).IncNonDefault +// +// since it cannot import godebug. +// +//go:linkname setNewIncNonDefault +func setNewIncNonDefault(newIncNonDefault func(string) func()) + +func init() { + setUpdate(update) + setNewIncNonDefault(newIncNonDefault) +} + +func newIncNonDefault(name string) func() { + s := New(name) + s.Value() + return s.IncNonDefault +} + +var updateMu sync.Mutex + +// update records an updated GODEBUG setting. +// def is the default GODEBUG setting for the running binary, +// and env is the current value of the $GODEBUG environment variable. +func update(def, env string) { + updateMu.Lock() + defer updateMu.Unlock() + + // Update all the cached values, creating new ones as needed. + // We parse the environment variable first, so that any settings it has + // are already locked in place (did[name] = true) before we consider + // the defaults. + did := make(map[string]bool) + parse(did, env) + parse(did, def) + + // Clear any cached values that are no longer present. + cache.Range(func(name, s any) bool { + if !did[name.(string)] { + s.(*setting).value.Store(&empty) + } + return true + }) +} + +// parse parses the GODEBUG setting string s, +// which has the form k=v,k2=v2,k3=v3. +// Later settings override earlier ones. +// Parse only updates settings k=v for which did[k] = false. +// It also sets did[k] = true for settings that it updates. +// Each value v can also have the form v#pattern, +// in which case the GODEBUG is only enabled for call stacks +// matching pattern, for use with golang.org/x/tools/cmd/bisect. +func parse(did map[string]bool, s string) { + // Scan the string backward so that later settings are used + // and earlier settings are ignored. + // Note that a forward scan would cause cached values + // to temporarily use the ignored value before being + // updated to the "correct" one. + end := len(s) + eq := -1 + for i := end - 1; i >= -1; i-- { + if i == -1 || s[i] == ',' { + if eq >= 0 { + name, arg := s[i+1:eq], s[eq+1:end] + if !did[name] { + did[name] = true + v := &value{text: arg} + for j := 0; j < len(arg); j++ { + if arg[j] == '#' { + v.text = arg[:j] + v.bisect, _ = bisect.New(arg[j+1:]) + break + } + } + lookup(name).value.Store(v) + } + } + eq = -1 + end = i + } else if s[i] == '=' { + eq = i + } + } +} + +type runtimeStderr struct{} + +var stderr runtimeStderr + +func (*runtimeStderr) Write(b []byte) (int, error) { + if len(b) > 0 { + write(2, unsafe.Pointer(&b[0]), int32(len(b))) + } + return len(b), nil +} + +// Since we cannot import os or syscall, use the runtime's write function +// to print to standard error. +// +//go:linkname write runtime.write +func write(fd uintptr, p unsafe.Pointer, n int32) int32 diff --git a/src/internal/godebug/godebug_test.go b/src/internal/godebug/godebug_test.go new file mode 100644 index 0000000..65dd256 --- /dev/null +++ b/src/internal/godebug/godebug_test.go @@ -0,0 +1,162 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package godebug_test + +import ( + "fmt" + . "internal/godebug" + "internal/race" + "internal/testenv" + "os" + "os/exec" + "reflect" + "runtime/metrics" + "sort" + "strings" + "testing" +) + +func TestGet(t *testing.T) { + foo := New("#foo") + tests := []struct { + godebug string + setting *Setting + want string + }{ + {"", New("#"), ""}, + {"", foo, ""}, + {"foo=bar", foo, "bar"}, + {"foo=bar,after=x", foo, "bar"}, + {"before=x,foo=bar,after=x", foo, "bar"}, + {"before=x,foo=bar", foo, "bar"}, + {",,,foo=bar,,,", foo, "bar"}, + {"foodecoy=wrong,foo=bar", foo, "bar"}, + {"foo=", foo, ""}, + {"foo", foo, ""}, + {",foo", foo, ""}, + {"foo=bar,baz", New("#loooooooong"), ""}, + } + for _, tt := range tests { + t.Setenv("GODEBUG", tt.godebug) + got := tt.setting.Value() + if got != tt.want { + t.Errorf("get(%q, %q) = %q; want %q", tt.godebug, tt.setting.Name(), got, tt.want) + } + } +} + +func TestMetrics(t *testing.T) { + const name = "http2client" // must be a real name so runtime will accept it + + var m [1]metrics.Sample + m[0].Name = "/godebug/non-default-behavior/" + name + ":events" + metrics.Read(m[:]) + if kind := m[0].Value.Kind(); kind != metrics.KindUint64 { + t.Fatalf("NonDefault kind = %v, want uint64", kind) + } + + s := New(name) + s.Value() + s.IncNonDefault() + s.IncNonDefault() + s.IncNonDefault() + metrics.Read(m[:]) + if kind := m[0].Value.Kind(); kind != metrics.KindUint64 { + t.Fatalf("NonDefault kind = %v, want uint64", kind) + } + if count := m[0].Value.Uint64(); count != 3 { + t.Fatalf("NonDefault value = %d, want 3", count) + } +} + +// TestPanicNilRace checks for a race in the runtime caused by use of runtime +// atomics (not visible to usual race detection) to install the counter for +// non-default panic(nil) semantics. For #64649. +func TestPanicNilRace(t *testing.T) { + if !race.Enabled { + t.Skip("Skipping test intended for use with -race.") + } + if os.Getenv("GODEBUG") != "panicnil=1" { + cmd := testenv.CleanCmdEnv(testenv.Command(t, os.Args[0], "-test.run=^TestPanicNilRace$", "-test.v", "-test.parallel=2", "-test.count=1")) + cmd.Env = append(cmd.Env, "GODEBUG=panicnil=1") + out, err := cmd.CombinedOutput() + t.Logf("output:\n%s", out) + + if err != nil { + t.Errorf("Was not expecting a crash") + } + return + } + + test := func(t *testing.T) { + t.Parallel() + defer func() { + recover() + }() + panic(nil) + } + t.Run("One", test) + t.Run("Two", test) +} + +func TestCmdBisect(t *testing.T) { + testenv.MustHaveGoBuild(t) + out, err := exec.Command("go", "run", "cmd/vendor/golang.org/x/tools/cmd/bisect", "GODEBUG=buggy=1#PATTERN", os.Args[0], "-test.run=BisectTestCase").CombinedOutput() + if err != nil { + t.Fatalf("exec bisect: %v\n%s", err, out) + } + + var want []string + src, err := os.ReadFile("godebug_test.go") + for i, line := range strings.Split(string(src), "\n") { + if strings.Contains(line, "BISECT"+" "+"BUG") { + want = append(want, fmt.Sprintf("godebug_test.go:%d", i+1)) + } + } + sort.Strings(want) + + var have []string + for _, line := range strings.Split(string(out), "\n") { + if strings.Contains(line, "godebug_test.go:") { + have = append(have, line[strings.LastIndex(line, "godebug_test.go:"):]) + } + } + sort.Strings(have) + + if !reflect.DeepEqual(have, want) { + t.Errorf("bad bisect output:\nhave %v\nwant %v\ncomplete output:\n%s", have, want, string(out)) + } +} + +// This test does nothing by itself, but you can run +// +// bisect 'GODEBUG=buggy=1#PATTERN' go test -run=BisectTestCase +// +// to see that the GODEBUG bisect support is working. +// TestCmdBisect above does exactly that. +func TestBisectTestCase(t *testing.T) { + s := New("#buggy") + for i := 0; i < 10; i++ { + a := s.Value() == "1" + b := s.Value() == "1" + c := s.Value() == "1" // BISECT BUG + d := s.Value() == "1" // BISECT BUG + e := s.Value() == "1" // BISECT BUG + + if a { + t.Log("ok") + } + if b { + t.Log("ok") + } + if c { + t.Error("bug") + } + if d && + e { + t.Error("bug") + } + } +} diff --git a/src/internal/godebugs/godebugs_test.go b/src/internal/godebugs/godebugs_test.go new file mode 100644 index 0000000..a1cb8d4 --- /dev/null +++ b/src/internal/godebugs/godebugs_test.go @@ -0,0 +1,46 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package godebugs_test + +import ( + "internal/godebugs" + "internal/testenv" + "os" + "runtime" + "strings" + "testing" +) + +func TestAll(t *testing.T) { + data, err := os.ReadFile("../../../doc/godebug.md") + if err != nil { + if os.IsNotExist(err) && (testenv.Builder() == "" || runtime.GOOS != "linux") { + t.Skip(err) + } + t.Fatal(err) + } + doc := string(data) + + last := "" + for _, info := range godebugs.All { + if info.Name <= last { + t.Errorf("All not sorted: %s then %s", last, info.Name) + } + last = info.Name + + if info.Package == "" { + t.Errorf("Name=%s missing Package", info.Name) + } + if info.Changed != 0 && info.Old == "" { + t.Errorf("Name=%s has Changed, missing Old", info.Name) + } + if info.Old != "" && info.Changed == 0 { + t.Errorf("Name=%s has Old, missing Changed", info.Name) + } + if !strings.Contains(doc, "`"+info.Name+"`") { + t.Errorf("Name=%s not documented in doc/godebug.md", info.Name) + } + } +} diff --git a/src/internal/godebugs/table.go b/src/internal/godebugs/table.go new file mode 100644 index 0000000..b1711d9 --- /dev/null +++ b/src/internal/godebugs/table.go @@ -0,0 +1,69 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package godebugs provides a table of known GODEBUG settings, +// for use by a variety of other packages, including internal/godebug, +// runtime, runtime/metrics, and cmd/go/internal/load. +package godebugs + +// An Info describes a single known GODEBUG setting. +type Info struct { + Name string // name of the setting ("panicnil") + Package string // package that uses the setting ("runtime") + Changed int // minor version when default changed, if any; 21 means Go 1.21 + Old string // value that restores behavior prior to Changed + Opaque bool // setting does not export information to runtime/metrics using [internal/godebug.Setting.IncNonDefault] +} + +// All is the table of known settings, sorted by Name. +// +// Note: After adding entries to this table, run 'go generate runtime/metrics' +// to update the runtime/metrics doc comment. +// (Otherwise the runtime/metrics test will fail.) +// +// Note: After adding entries to this table, update the list in doc/godebug.md as well. +// (Otherwise the test in this package will fail.) +var All = []Info{ + {Name: "execerrdot", Package: "os/exec"}, + {Name: "gocachehash", Package: "cmd/go"}, + {Name: "gocachetest", Package: "cmd/go"}, + {Name: "gocacheverify", Package: "cmd/go"}, + {Name: "http2client", Package: "net/http"}, + {Name: "http2debug", Package: "net/http", Opaque: true}, + {Name: "http2server", Package: "net/http"}, + {Name: "installgoroot", Package: "go/build"}, + {Name: "jstmpllitinterp", Package: "html/template"}, + //{Name: "multipartfiles", Package: "mime/multipart"}, + {Name: "multipartmaxheaders", Package: "mime/multipart"}, + {Name: "multipartmaxparts", Package: "mime/multipart"}, + {Name: "multipathtcp", Package: "net"}, + {Name: "netdns", Package: "net", Opaque: true}, + {Name: "panicnil", Package: "runtime", Changed: 21, Old: "1"}, + {Name: "randautoseed", Package: "math/rand"}, + {Name: "tarinsecurepath", Package: "archive/tar"}, + {Name: "tlsmaxrsasize", Package: "crypto/tls"}, + {Name: "x509sha1", Package: "crypto/x509"}, + {Name: "x509usefallbackroots", Package: "crypto/x509"}, + {Name: "zipinsecurepath", Package: "archive/zip"}, +} + +// Lookup returns the Info with the given name. +func Lookup(name string) *Info { + // binary search, avoiding import of sort. + lo := 0 + hi := len(All) + for lo < hi { + m := lo + (hi-lo)>>1 + mid := All[m].Name + if name == mid { + return &All[m] + } + if name < mid { + hi = m + } else { + lo = m + 1 + } + } + return nil +} diff --git a/src/internal/goexperiment/exp_arenas_off.go b/src/internal/goexperiment/exp_arenas_off.go new file mode 100644 index 0000000..9e40ebc --- /dev/null +++ b/src/internal/goexperiment/exp_arenas_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.arenas +// +build !goexperiment.arenas + +package goexperiment + +const Arenas = false +const ArenasInt = 0 diff --git a/src/internal/goexperiment/exp_arenas_on.go b/src/internal/goexperiment/exp_arenas_on.go new file mode 100644 index 0000000..92ef748 --- /dev/null +++ b/src/internal/goexperiment/exp_arenas_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.arenas +// +build goexperiment.arenas + +package goexperiment + +const Arenas = true +const ArenasInt = 1 diff --git a/src/internal/goexperiment/exp_boringcrypto_off.go b/src/internal/goexperiment/exp_boringcrypto_off.go new file mode 100644 index 0000000..020c75b --- /dev/null +++ b/src/internal/goexperiment/exp_boringcrypto_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.boringcrypto +// +build !goexperiment.boringcrypto + +package goexperiment + +const BoringCrypto = false +const BoringCryptoInt = 0 diff --git a/src/internal/goexperiment/exp_boringcrypto_on.go b/src/internal/goexperiment/exp_boringcrypto_on.go new file mode 100644 index 0000000..1454329 --- /dev/null +++ b/src/internal/goexperiment/exp_boringcrypto_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.boringcrypto +// +build goexperiment.boringcrypto + +package goexperiment + +const BoringCrypto = true +const BoringCryptoInt = 1 diff --git a/src/internal/goexperiment/exp_cacheprog_off.go b/src/internal/goexperiment/exp_cacheprog_off.go new file mode 100644 index 0000000..29aa869 --- /dev/null +++ b/src/internal/goexperiment/exp_cacheprog_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.cacheprog +// +build !goexperiment.cacheprog + +package goexperiment + +const CacheProg = false +const CacheProgInt = 0 diff --git a/src/internal/goexperiment/exp_cacheprog_on.go b/src/internal/goexperiment/exp_cacheprog_on.go new file mode 100644 index 0000000..121b299 --- /dev/null +++ b/src/internal/goexperiment/exp_cacheprog_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.cacheprog +// +build goexperiment.cacheprog + +package goexperiment + +const CacheProg = true +const CacheProgInt = 1 diff --git a/src/internal/goexperiment/exp_cgocheck2_off.go b/src/internal/goexperiment/exp_cgocheck2_off.go new file mode 100644 index 0000000..77aa538 --- /dev/null +++ b/src/internal/goexperiment/exp_cgocheck2_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.cgocheck2 +// +build !goexperiment.cgocheck2 + +package goexperiment + +const CgoCheck2 = false +const CgoCheck2Int = 0 diff --git a/src/internal/goexperiment/exp_cgocheck2_on.go b/src/internal/goexperiment/exp_cgocheck2_on.go new file mode 100644 index 0000000..6201249 --- /dev/null +++ b/src/internal/goexperiment/exp_cgocheck2_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.cgocheck2 +// +build goexperiment.cgocheck2 + +package goexperiment + +const CgoCheck2 = true +const CgoCheck2Int = 1 diff --git a/src/internal/goexperiment/exp_coverageredesign_off.go b/src/internal/goexperiment/exp_coverageredesign_off.go new file mode 100644 index 0000000..95d3a6c --- /dev/null +++ b/src/internal/goexperiment/exp_coverageredesign_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.coverageredesign +// +build !goexperiment.coverageredesign + +package goexperiment + +const CoverageRedesign = false +const CoverageRedesignInt = 0 diff --git a/src/internal/goexperiment/exp_coverageredesign_on.go b/src/internal/goexperiment/exp_coverageredesign_on.go new file mode 100644 index 0000000..330a234 --- /dev/null +++ b/src/internal/goexperiment/exp_coverageredesign_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.coverageredesign +// +build goexperiment.coverageredesign + +package goexperiment + +const CoverageRedesign = true +const CoverageRedesignInt = 1 diff --git a/src/internal/goexperiment/exp_fieldtrack_off.go b/src/internal/goexperiment/exp_fieldtrack_off.go new file mode 100644 index 0000000..e5e1326 --- /dev/null +++ b/src/internal/goexperiment/exp_fieldtrack_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.fieldtrack +// +build !goexperiment.fieldtrack + +package goexperiment + +const FieldTrack = false +const FieldTrackInt = 0 diff --git a/src/internal/goexperiment/exp_fieldtrack_on.go b/src/internal/goexperiment/exp_fieldtrack_on.go new file mode 100644 index 0000000..0d8c447 --- /dev/null +++ b/src/internal/goexperiment/exp_fieldtrack_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.fieldtrack +// +build goexperiment.fieldtrack + +package goexperiment + +const FieldTrack = true +const FieldTrackInt = 1 diff --git a/src/internal/goexperiment/exp_heapminimum512kib_off.go b/src/internal/goexperiment/exp_heapminimum512kib_off.go new file mode 100644 index 0000000..09da431 --- /dev/null +++ b/src/internal/goexperiment/exp_heapminimum512kib_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.heapminimum512kib +// +build !goexperiment.heapminimum512kib + +package goexperiment + +const HeapMinimum512KiB = false +const HeapMinimum512KiBInt = 0 diff --git a/src/internal/goexperiment/exp_heapminimum512kib_on.go b/src/internal/goexperiment/exp_heapminimum512kib_on.go new file mode 100644 index 0000000..bab684b --- /dev/null +++ b/src/internal/goexperiment/exp_heapminimum512kib_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.heapminimum512kib +// +build goexperiment.heapminimum512kib + +package goexperiment + +const HeapMinimum512KiB = true +const HeapMinimum512KiBInt = 1 diff --git a/src/internal/goexperiment/exp_loopvar_off.go b/src/internal/goexperiment/exp_loopvar_off.go new file mode 100644 index 0000000..1cd7ee1 --- /dev/null +++ b/src/internal/goexperiment/exp_loopvar_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.loopvar +// +build !goexperiment.loopvar + +package goexperiment + +const LoopVar = false +const LoopVarInt = 0 diff --git a/src/internal/goexperiment/exp_loopvar_on.go b/src/internal/goexperiment/exp_loopvar_on.go new file mode 100644 index 0000000..e3c8980 --- /dev/null +++ b/src/internal/goexperiment/exp_loopvar_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.loopvar +// +build goexperiment.loopvar + +package goexperiment + +const LoopVar = true +const LoopVarInt = 1 diff --git a/src/internal/goexperiment/exp_pagetrace_off.go b/src/internal/goexperiment/exp_pagetrace_off.go new file mode 100644 index 0000000..789e883 --- /dev/null +++ b/src/internal/goexperiment/exp_pagetrace_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.pagetrace +// +build !goexperiment.pagetrace + +package goexperiment + +const PageTrace = false +const PageTraceInt = 0 diff --git a/src/internal/goexperiment/exp_pagetrace_on.go b/src/internal/goexperiment/exp_pagetrace_on.go new file mode 100644 index 0000000..ea72b54 --- /dev/null +++ b/src/internal/goexperiment/exp_pagetrace_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.pagetrace +// +build goexperiment.pagetrace + +package goexperiment + +const PageTrace = true +const PageTraceInt = 1 diff --git a/src/internal/goexperiment/exp_preemptibleloops_off.go b/src/internal/goexperiment/exp_preemptibleloops_off.go new file mode 100644 index 0000000..7a26088 --- /dev/null +++ b/src/internal/goexperiment/exp_preemptibleloops_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.preemptibleloops +// +build !goexperiment.preemptibleloops + +package goexperiment + +const PreemptibleLoops = false +const PreemptibleLoopsInt = 0 diff --git a/src/internal/goexperiment/exp_preemptibleloops_on.go b/src/internal/goexperiment/exp_preemptibleloops_on.go new file mode 100644 index 0000000..a9ca28c --- /dev/null +++ b/src/internal/goexperiment/exp_preemptibleloops_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.preemptibleloops +// +build goexperiment.preemptibleloops + +package goexperiment + +const PreemptibleLoops = true +const PreemptibleLoopsInt = 1 diff --git a/src/internal/goexperiment/exp_regabiargs_off.go b/src/internal/goexperiment/exp_regabiargs_off.go new file mode 100644 index 0000000..31a139b --- /dev/null +++ b/src/internal/goexperiment/exp_regabiargs_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.regabiargs +// +build !goexperiment.regabiargs + +package goexperiment + +const RegabiArgs = false +const RegabiArgsInt = 0 diff --git a/src/internal/goexperiment/exp_regabiargs_on.go b/src/internal/goexperiment/exp_regabiargs_on.go new file mode 100644 index 0000000..9b26f3c --- /dev/null +++ b/src/internal/goexperiment/exp_regabiargs_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.regabiargs +// +build goexperiment.regabiargs + +package goexperiment + +const RegabiArgs = true +const RegabiArgsInt = 1 diff --git a/src/internal/goexperiment/exp_regabiwrappers_off.go b/src/internal/goexperiment/exp_regabiwrappers_off.go new file mode 100644 index 0000000..bfa0fa3 --- /dev/null +++ b/src/internal/goexperiment/exp_regabiwrappers_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.regabiwrappers +// +build !goexperiment.regabiwrappers + +package goexperiment + +const RegabiWrappers = false +const RegabiWrappersInt = 0 diff --git a/src/internal/goexperiment/exp_regabiwrappers_on.go b/src/internal/goexperiment/exp_regabiwrappers_on.go new file mode 100644 index 0000000..11ffffb --- /dev/null +++ b/src/internal/goexperiment/exp_regabiwrappers_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.regabiwrappers +// +build goexperiment.regabiwrappers + +package goexperiment + +const RegabiWrappers = true +const RegabiWrappersInt = 1 diff --git a/src/internal/goexperiment/exp_staticlockranking_off.go b/src/internal/goexperiment/exp_staticlockranking_off.go new file mode 100644 index 0000000..3d546c0 --- /dev/null +++ b/src/internal/goexperiment/exp_staticlockranking_off.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build !goexperiment.staticlockranking +// +build !goexperiment.staticlockranking + +package goexperiment + +const StaticLockRanking = false +const StaticLockRankingInt = 0 diff --git a/src/internal/goexperiment/exp_staticlockranking_on.go b/src/internal/goexperiment/exp_staticlockranking_on.go new file mode 100644 index 0000000..78188fb --- /dev/null +++ b/src/internal/goexperiment/exp_staticlockranking_on.go @@ -0,0 +1,9 @@ +// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build goexperiment.staticlockranking +// +build goexperiment.staticlockranking + +package goexperiment + +const StaticLockRanking = true +const StaticLockRankingInt = 1 diff --git a/src/internal/goexperiment/flags.go b/src/internal/goexperiment/flags.go new file mode 100644 index 0000000..ae3cbaf --- /dev/null +++ b/src/internal/goexperiment/flags.go @@ -0,0 +1,112 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package goexperiment implements support for toolchain experiments. +// +// Toolchain experiments are controlled by the GOEXPERIMENT +// environment variable. GOEXPERIMENT is a comma-separated list of +// experiment names. GOEXPERIMENT can be set at make.bash time, which +// sets the default experiments for binaries built with the tool +// chain; or it can be set at build time. GOEXPERIMENT can also be set +// to "none", which disables any experiments that were enabled at +// make.bash time. +// +// Experiments are exposed to the build in the following ways: +// +// - Build tag goexperiment.x is set if experiment x (lower case) is +// enabled. +// +// - For each experiment x (in camel case), this package contains a +// boolean constant x and an integer constant xInt. +// +// - In runtime assembly, the macro GOEXPERIMENT_x is defined if +// experiment x (lower case) is enabled. +// +// In the toolchain, the set of experiments enabled for the current +// build should be accessed via objabi.Experiment. +// +// The set of experiments is included in the output of runtime.Version() +// and "go version <binary>" if it differs from the default experiments. +// +// For the set of experiments supported by the current toolchain, see +// "go doc goexperiment.Flags". +// +// Note that this package defines the set of experiments (in Flags) +// and records the experiments that were enabled when the package +// was compiled (as boolean and integer constants). +// +// Note especially that this package does not itself change behavior +// at run time based on the GOEXPERIMENT variable. +// The code used in builds to interpret the GOEXPERIMENT variable +// is in the separate package internal/buildcfg. +package goexperiment + +//go:generate go run mkconsts.go + +// Flags is the set of experiments that can be enabled or disabled in +// the current toolchain. +// +// When specified in the GOEXPERIMENT environment variable or as build +// tags, experiments use the strings.ToLower of their field name. +// +// For the baseline experimental configuration, see +// objabi.experimentBaseline. +// +// If you change this struct definition, run "go generate". +type Flags struct { + FieldTrack bool + PreemptibleLoops bool + StaticLockRanking bool + BoringCrypto bool + + // Regabi is split into several sub-experiments that can be + // enabled individually. Not all combinations work. + // The "regabi" GOEXPERIMENT is an alias for all "working" + // subexperiments. + + // RegabiWrappers enables ABI wrappers for calling between + // ABI0 and ABIInternal functions. Without this, the ABIs are + // assumed to be identical so cross-ABI calls are direct. + RegabiWrappers bool + // RegabiArgs enables register arguments/results in all + // compiled Go functions. + // + // Requires wrappers (to do ABI translation), and reflect (so + // reflection calls use registers). + RegabiArgs bool + + // HeapMinimum512KiB reduces the minimum heap size to 512 KiB. + // + // This was originally reduced as part of PacerRedesign, but + // has been broken out to its own experiment that is disabled + // by default. + HeapMinimum512KiB bool + + // CoverageRedesign enables the new compiler-based code coverage + // tooling. + CoverageRedesign bool + + // Arenas causes the "arena" standard library package to be visible + // to the outside world. + Arenas bool + + // PageTrace enables GODEBUG=pagetrace=/path/to/result. This feature + // is a GOEXPERIMENT due to a security risk with setuid binaries: + // this compels the Go runtime to write to some arbitrary file, which + // may be exploited. + PageTrace bool + + // CgoCheck2 enables an expensive cgo rule checker. + // When this experiment is enabled, cgo rule checks occur regardless + // of the GODEBUG=cgocheck setting provided at runtime. + CgoCheck2 bool + + // LoopVar changes loop semantics so that each iteration gets its own + // copy of the iteration variable. + LoopVar bool + + // CacheProg adds support to cmd/go to use a child process to implement + // the build cache; see https://github.com/golang/go/issues/59719. + CacheProg bool +} diff --git a/src/internal/goexperiment/mkconsts.go b/src/internal/goexperiment/mkconsts.go new file mode 100644 index 0000000..204ca9d --- /dev/null +++ b/src/internal/goexperiment/mkconsts.go @@ -0,0 +1,74 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ignore +// +build ignore + +// mkconsts generates const definition files for each GOEXPERIMENT. +package main + +import ( + "bytes" + "fmt" + "internal/goexperiment" + "log" + "os" + "reflect" + "strings" +) + +func main() { + // Delete existing experiment constant files. + ents, err := os.ReadDir(".") + if err != nil { + log.Fatal(err) + } + for _, ent := range ents { + name := ent.Name() + if !strings.HasPrefix(name, "exp_") { + continue + } + // Check that this is definitely a generated file. + data, err := os.ReadFile(name) + if err != nil { + log.Fatalf("reading %s: %v", name, err) + } + if !bytes.Contains(data, []byte("Code generated by mkconsts")) { + log.Fatalf("%s: expected generated file", name) + } + if err := os.Remove(name); err != nil { + log.Fatal(err) + } + } + + // Generate new experiment constant files. + rt := reflect.TypeOf(&goexperiment.Flags{}).Elem() + for i := 0; i < rt.NumField(); i++ { + f := rt.Field(i).Name + buildTag := "goexperiment." + strings.ToLower(f) + for _, val := range []bool{false, true} { + name := fmt.Sprintf("exp_%s_%s.go", strings.ToLower(f), pick(val, "off", "on")) + data := fmt.Sprintf(`// Code generated by mkconsts.go. DO NOT EDIT. + +//go:build %s%s +// +build %s%s + +package goexperiment + +const %s = %v +const %sInt = %s +`, pick(val, "!", ""), buildTag, pick(val, "!", ""), buildTag, f, val, f, pick(val, "0", "1")) + if err := os.WriteFile(name, []byte(data), 0666); err != nil { + log.Fatalf("writing %s: %v", name, err) + } + } + } +} + +func pick(v bool, f, t string) string { + if v { + return t + } + return f +} diff --git a/src/internal/goos/gengoos.go b/src/internal/goos/gengoos.go new file mode 100644 index 0000000..37d9706 --- /dev/null +++ b/src/internal/goos/gengoos.go @@ -0,0 +1,71 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ignore + +package main + +import ( + "bytes" + "fmt" + "log" + "os" + "strings" +) + +var gooses []string + +func main() { + data, err := os.ReadFile("../../go/build/syslist.go") + if err != nil { + log.Fatal(err) + } + const goosPrefix = `var knownOS = map[string]bool{` + inGOOS := false + for _, line := range strings.Split(string(data), "\n") { + if strings.HasPrefix(line, goosPrefix) { + inGOOS = true + } else if inGOOS && strings.HasPrefix(line, "}") { + break + } else if inGOOS { + goos := strings.Fields(line)[0] + goos = strings.TrimPrefix(goos, `"`) + goos = strings.TrimSuffix(goos, `":`) + gooses = append(gooses, goos) + } + } + + for _, target := range gooses { + if target == "nacl" { + continue + } + var tags []string + if target == "linux" { + tags = append(tags, "!android") // must explicitly exclude android for linux + } + if target == "solaris" { + tags = append(tags, "!illumos") // must explicitly exclude illumos for solaris + } + if target == "darwin" { + tags = append(tags, "!ios") // must explicitly exclude ios for darwin + } + tags = append(tags, target) // must explicitly include target for bootstrapping purposes + var buf bytes.Buffer + fmt.Fprintf(&buf, "// Code generated by gengoos.go using 'go generate'. DO NOT EDIT.\n\n") + fmt.Fprintf(&buf, "//go:build %s\n\n", strings.Join(tags, " && ")) + fmt.Fprintf(&buf, "package goos\n\n") + fmt.Fprintf(&buf, "const GOOS = `%s`\n\n", target) + for _, goos := range gooses { + value := 0 + if goos == target { + value = 1 + } + fmt.Fprintf(&buf, "const Is%s = %d\n", strings.Title(goos), value) + } + err := os.WriteFile("zgoos_"+target+".go", buf.Bytes(), 0666) + if err != nil { + log.Fatal(err) + } + } +} diff --git a/src/internal/goos/goos.go b/src/internal/goos/goos.go new file mode 100644 index 0000000..02dc968 --- /dev/null +++ b/src/internal/goos/goos.go @@ -0,0 +1,13 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// package goos contains GOOS-specific constants. +package goos + +// The next line makes 'go generate' write the zgoos*.go files with +// per-OS information, including constants named Is$GOOS for every +// known GOOS. The constant is 1 on the current system, 0 otherwise; +// multiplying by them is useful for defining GOOS-specific constants. +// +//go:generate go run gengoos.go diff --git a/src/internal/goos/nonunix.go b/src/internal/goos/nonunix.go new file mode 100644 index 0000000..2ba5c85 --- /dev/null +++ b/src/internal/goos/nonunix.go @@ -0,0 +1,9 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !unix + +package goos + +const IsUnix = false diff --git a/src/internal/goos/unix.go b/src/internal/goos/unix.go new file mode 100644 index 0000000..6cfd5ef --- /dev/null +++ b/src/internal/goos/unix.go @@ -0,0 +1,9 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package goos + +const IsUnix = true diff --git a/src/internal/goos/zgoos_aix.go b/src/internal/goos/zgoos_aix.go new file mode 100644 index 0000000..24e05c9 --- /dev/null +++ b/src/internal/goos/zgoos_aix.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build aix + +package goos + +const GOOS = `aix` + +const IsAix = 1 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_android.go b/src/internal/goos/zgoos_android.go new file mode 100644 index 0000000..3c4a318 --- /dev/null +++ b/src/internal/goos/zgoos_android.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build android + +package goos + +const GOOS = `android` + +const IsAix = 0 +const IsAndroid = 1 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_darwin.go b/src/internal/goos/zgoos_darwin.go new file mode 100644 index 0000000..10b1499 --- /dev/null +++ b/src/internal/goos/zgoos_darwin.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build !ios && darwin + +package goos + +const GOOS = `darwin` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 1 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_dragonfly.go b/src/internal/goos/zgoos_dragonfly.go new file mode 100644 index 0000000..b92d126 --- /dev/null +++ b/src/internal/goos/zgoos_dragonfly.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build dragonfly + +package goos + +const GOOS = `dragonfly` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 1 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_freebsd.go b/src/internal/goos/zgoos_freebsd.go new file mode 100644 index 0000000..f547591 --- /dev/null +++ b/src/internal/goos/zgoos_freebsd.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build freebsd + +package goos + +const GOOS = `freebsd` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 1 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_hurd.go b/src/internal/goos/zgoos_hurd.go new file mode 100644 index 0000000..1189d65 --- /dev/null +++ b/src/internal/goos/zgoos_hurd.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build hurd + +package goos + +const GOOS = `hurd` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 1 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_illumos.go b/src/internal/goos/zgoos_illumos.go new file mode 100644 index 0000000..4f02540 --- /dev/null +++ b/src/internal/goos/zgoos_illumos.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build illumos + +package goos + +const GOOS = `illumos` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 1 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_ios.go b/src/internal/goos/zgoos_ios.go new file mode 100644 index 0000000..02f3586 --- /dev/null +++ b/src/internal/goos/zgoos_ios.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build ios + +package goos + +const GOOS = `ios` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 1 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_js.go b/src/internal/goos/zgoos_js.go new file mode 100644 index 0000000..4818741 --- /dev/null +++ b/src/internal/goos/zgoos_js.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build js + +package goos + +const GOOS = `js` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 1 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_linux.go b/src/internal/goos/zgoos_linux.go new file mode 100644 index 0000000..6f4d4e0 --- /dev/null +++ b/src/internal/goos/zgoos_linux.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build !android && linux + +package goos + +const GOOS = `linux` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 1 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_netbsd.go b/src/internal/goos/zgoos_netbsd.go new file mode 100644 index 0000000..948603d --- /dev/null +++ b/src/internal/goos/zgoos_netbsd.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build netbsd + +package goos + +const GOOS = `netbsd` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 1 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_openbsd.go b/src/internal/goos/zgoos_openbsd.go new file mode 100644 index 0000000..f4b2014 --- /dev/null +++ b/src/internal/goos/zgoos_openbsd.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build openbsd + +package goos + +const GOOS = `openbsd` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 1 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_plan9.go b/src/internal/goos/zgoos_plan9.go new file mode 100644 index 0000000..95572df --- /dev/null +++ b/src/internal/goos/zgoos_plan9.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build plan9 + +package goos + +const GOOS = `plan9` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 1 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_solaris.go b/src/internal/goos/zgoos_solaris.go new file mode 100644 index 0000000..c705826 --- /dev/null +++ b/src/internal/goos/zgoos_solaris.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build !illumos && solaris + +package goos + +const GOOS = `solaris` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 1 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_wasip1.go b/src/internal/goos/zgoos_wasip1.go new file mode 100644 index 0000000..ae35eeb --- /dev/null +++ b/src/internal/goos/zgoos_wasip1.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build wasip1 + +package goos + +const GOOS = `wasip1` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 1 +const IsWindows = 0 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_windows.go b/src/internal/goos/zgoos_windows.go new file mode 100644 index 0000000..f89f4cf --- /dev/null +++ b/src/internal/goos/zgoos_windows.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build windows + +package goos + +const GOOS = `windows` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 1 +const IsZos = 0 diff --git a/src/internal/goos/zgoos_zos.go b/src/internal/goos/zgoos_zos.go new file mode 100644 index 0000000..29fb0f8 --- /dev/null +++ b/src/internal/goos/zgoos_zos.go @@ -0,0 +1,26 @@ +// Code generated by gengoos.go using 'go generate'. DO NOT EDIT. + +//go:build zos + +package goos + +const GOOS = `zos` + +const IsAix = 0 +const IsAndroid = 0 +const IsDarwin = 0 +const IsDragonfly = 0 +const IsFreebsd = 0 +const IsHurd = 0 +const IsIllumos = 0 +const IsIos = 0 +const IsJs = 0 +const IsLinux = 0 +const IsNacl = 0 +const IsNetbsd = 0 +const IsOpenbsd = 0 +const IsPlan9 = 0 +const IsSolaris = 0 +const IsWasip1 = 0 +const IsWindows = 0 +const IsZos = 1 diff --git a/src/internal/goroot/gc.go b/src/internal/goroot/gc.go new file mode 100644 index 0000000..c0216f4 --- /dev/null +++ b/src/internal/goroot/gc.go @@ -0,0 +1,131 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build gc + +package goroot + +import ( + "os" + "os/exec" + "path/filepath" + "strings" + "sync" +) + +// IsStandardPackage reports whether path is a standard package, +// given goroot and compiler. +func IsStandardPackage(goroot, compiler, path string) bool { + switch compiler { + case "gc": + dir := filepath.Join(goroot, "src", path) + info, err := os.Stat(dir) + return err == nil && info.IsDir() + case "gccgo": + return gccgoSearch.isStandard(path) + default: + panic("unknown compiler " + compiler) + } +} + +// gccgoSearch holds the gccgo search directories. +type gccgoDirs struct { + once sync.Once + dirs []string +} + +// gccgoSearch is used to check whether a gccgo package exists in the +// standard library. +var gccgoSearch gccgoDirs + +// init finds the gccgo search directories. If this fails it leaves dirs == nil. +func (gd *gccgoDirs) init() { + gccgo := os.Getenv("GCCGO") + if gccgo == "" { + gccgo = "gccgo" + } + bin, err := exec.LookPath(gccgo) + if err != nil { + return + } + + allDirs, err := exec.Command(bin, "-print-search-dirs").Output() + if err != nil { + return + } + versionB, err := exec.Command(bin, "-dumpversion").Output() + if err != nil { + return + } + version := strings.TrimSpace(string(versionB)) + machineB, err := exec.Command(bin, "-dumpmachine").Output() + if err != nil { + return + } + machine := strings.TrimSpace(string(machineB)) + + dirsEntries := strings.Split(string(allDirs), "\n") + const prefix = "libraries: =" + var dirs []string + for _, dirEntry := range dirsEntries { + if strings.HasPrefix(dirEntry, prefix) { + dirs = filepath.SplitList(strings.TrimPrefix(dirEntry, prefix)) + break + } + } + if len(dirs) == 0 { + return + } + + var lastDirs []string + for _, dir := range dirs { + goDir := filepath.Join(dir, "go", version) + if fi, err := os.Stat(goDir); err == nil && fi.IsDir() { + gd.dirs = append(gd.dirs, goDir) + goDir = filepath.Join(goDir, machine) + if fi, err = os.Stat(goDir); err == nil && fi.IsDir() { + gd.dirs = append(gd.dirs, goDir) + } + } + if fi, err := os.Stat(dir); err == nil && fi.IsDir() { + lastDirs = append(lastDirs, dir) + } + } + gd.dirs = append(gd.dirs, lastDirs...) +} + +// isStandard reports whether path is a standard library for gccgo. +func (gd *gccgoDirs) isStandard(path string) bool { + // Quick check: if the first path component has a '.', it's not + // in the standard library. This skips most GOPATH directories. + i := strings.Index(path, "/") + if i < 0 { + i = len(path) + } + if strings.Contains(path[:i], ".") { + return false + } + + if path == "unsafe" { + // Special case. + return true + } + + gd.once.Do(gd.init) + if gd.dirs == nil { + // We couldn't find the gccgo search directories. + // Best guess, since the first component did not contain + // '.', is that this is a standard library package. + return true + } + + for _, dir := range gd.dirs { + full := filepath.Join(dir, path) + ".gox" + if fi, err := os.Stat(full); err == nil && !fi.IsDir() { + return true + } + } + + return false +} diff --git a/src/internal/goroot/gccgo.go b/src/internal/goroot/gccgo.go new file mode 100644 index 0000000..6284122 --- /dev/null +++ b/src/internal/goroot/gccgo.go @@ -0,0 +1,27 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build gccgo + +package goroot + +import ( + "os" + "path/filepath" +) + +// IsStandardPackage reports whether path is a standard package, +// given goroot and compiler. +func IsStandardPackage(goroot, compiler, path string) bool { + switch compiler { + case "gc": + dir := filepath.Join(goroot, "src", path) + _, err := os.Stat(dir) + return err == nil + case "gccgo": + return stdpkg[path] + default: + panic("unknown compiler " + compiler) + } +} diff --git a/src/internal/goversion/goversion.go b/src/internal/goversion/goversion.go new file mode 100644 index 0000000..5a52f9e --- /dev/null +++ b/src/internal/goversion/goversion.go @@ -0,0 +1,12 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package goversion + +// Version is the Go 1.x version which is currently +// in development and will eventually get released. +// +// It should be updated at the start of each development cycle to be +// the version of the next Go 1.x release. See golang.org/issue/40705. +const Version = 21 diff --git a/src/internal/intern/intern.go b/src/internal/intern/intern.go new file mode 100644 index 0000000..2f97c2e --- /dev/null +++ b/src/internal/intern/intern.go @@ -0,0 +1,181 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package intern lets you make smaller comparable values by boxing +// a larger comparable value (such as a 16 byte string header) down +// into a globally unique 8 byte pointer. +// +// The globally unique pointers are garbage collected with weak +// references and finalizers. This package hides that. +package intern + +import ( + "internal/godebug" + "runtime" + "sync" + "unsafe" +) + +// A Value pointer is the handle to an underlying comparable value. +// See func Get for how Value pointers may be used. +type Value struct { + _ [0]func() // prevent people from accidentally using value type as comparable + cmpVal any + // resurrected is guarded by mu (for all instances of Value). + // It is set true whenever v is synthesized from a uintptr. + resurrected bool +} + +// Get returns the comparable value passed to the Get func +// that returned v. +func (v *Value) Get() any { return v.cmpVal } + +// key is a key in our global value map. +// It contains type-specialized fields to avoid allocations +// when converting common types to empty interfaces. +type key struct { + s string + cmpVal any + // isString reports whether key contains a string. + // Without it, the zero value of key is ambiguous. + isString bool +} + +// keyFor returns a key to use with cmpVal. +func keyFor(cmpVal any) key { + if s, ok := cmpVal.(string); ok { + return key{s: s, isString: true} + } + return key{cmpVal: cmpVal} +} + +// Value returns a *Value built from k. +func (k key) Value() *Value { + if k.isString { + return &Value{cmpVal: k.s} + } + return &Value{cmpVal: k.cmpVal} +} + +var ( + // mu guards valMap, a weakref map of *Value by underlying value. + // It also guards the resurrected field of all *Values. + mu sync.Mutex + valMap = map[key]uintptr{} // to uintptr(*Value) + valSafe = safeMap() // non-nil in safe+leaky mode +) + +var intern = godebug.New("#intern") + +// safeMap returns a non-nil map if we're in safe-but-leaky mode, +// as controlled by GODEBUG=intern=leaky +func safeMap() map[key]*Value { + if intern.Value() == "leaky" { + return map[key]*Value{} + } + return nil +} + +// Get returns a pointer representing the comparable value cmpVal. +// +// The returned pointer will be the same for Get(v) and Get(v2) +// if and only if v == v2, and can be used as a map key. +func Get(cmpVal any) *Value { + return get(keyFor(cmpVal)) +} + +// GetByString is identical to Get, except that it is specialized for strings. +// This avoids an allocation from putting a string into an interface{} +// to pass as an argument to Get. +func GetByString(s string) *Value { + return get(key{s: s, isString: true}) +} + +// We play unsafe games that violate Go's rules (and assume a non-moving +// collector). So we quiet Go here. +// See the comment below Get for more implementation details. +// +//go:nocheckptr +func get(k key) *Value { + mu.Lock() + defer mu.Unlock() + + var v *Value + if valSafe != nil { + v = valSafe[k] + } else if addr, ok := valMap[k]; ok { + v = (*Value)(unsafe.Pointer(addr)) + v.resurrected = true + } + if v != nil { + return v + } + v = k.Value() + if valSafe != nil { + valSafe[k] = v + } else { + // SetFinalizer before uintptr conversion (theoretical concern; + // see https://github.com/go4org/intern/issues/13) + runtime.SetFinalizer(v, finalize) + valMap[k] = uintptr(unsafe.Pointer(v)) + } + return v +} + +func finalize(v *Value) { + mu.Lock() + defer mu.Unlock() + if v.resurrected { + // We lost the race. Somebody resurrected it while we + // were about to finalize it. Try again next round. + v.resurrected = false + runtime.SetFinalizer(v, finalize) + return + } + delete(valMap, keyFor(v.cmpVal)) +} + +// Interning is simple if you don't require that unused values be +// garbage collectable. But we do require that; we don't want to be +// DOS vector. We do this by using a uintptr to hide the pointer from +// the garbage collector, and using a finalizer to eliminate the +// pointer when no other code is using it. +// +// The obvious implementation of this is to use a +// map[interface{}]uintptr-of-*interface{}, and set up a finalizer to +// delete from the map. Unfortunately, this is racy. Because pointers +// are being created in violation of Go's unsafety rules, it's +// possible to create a pointer to a value concurrently with the GC +// concluding that the value can be collected. There are other races +// that break the equality invariant as well, but the use-after-free +// will cause a runtime crash. +// +// To make this work, the finalizer needs to know that no references +// have been unsafely created since the finalizer was set up. To do +// this, values carry a "resurrected" sentinel, which gets set +// whenever a pointer is unsafely created. If the finalizer encounters +// the sentinel, it clears the sentinel and delays collection for one +// additional GC cycle, by re-installing itself as finalizer. This +// ensures that the unsafely created pointer is visible to the GC, and +// will correctly prevent collection. +// +// This technique does mean that interned values that get reused take +// at least 3 GC cycles to fully collect (1 to clear the sentinel, 1 +// to clean up the unsafe map, 1 to be actually deleted). +// +// @ianlancetaylor commented in +// https://github.com/golang/go/issues/41303#issuecomment-717401656 +// that it is possible to implement weak references in terms of +// finalizers without unsafe. Unfortunately, the approach he outlined +// does not work here, for two reasons. First, there is no way to +// construct a strong pointer out of a weak pointer; our map stores +// weak pointers, but we must return strong pointers to callers. +// Second, and more fundamentally, we must return not just _a_ strong +// pointer to callers, but _the same_ strong pointer to callers. In +// order to return _the same_ strong pointer to callers, we must track +// it, which is exactly what we cannot do with strong pointers. +// +// See https://github.com/inetaf/netaddr/issues/53 for more +// discussion, and https://github.com/go4org/intern/issues/2 for an +// illustration of the subtleties at play. diff --git a/src/internal/intern/intern_test.go b/src/internal/intern/intern_test.go new file mode 100644 index 0000000..d1e409e --- /dev/null +++ b/src/internal/intern/intern_test.go @@ -0,0 +1,199 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package intern + +import ( + "fmt" + "runtime" + "testing" +) + +func TestBasics(t *testing.T) { + clearMap() + foo := Get("foo") + bar := Get("bar") + empty := Get("") + nilEface := Get(nil) + i := Get(0x7777777) + foo2 := Get("foo") + bar2 := Get("bar") + empty2 := Get("") + nilEface2 := Get(nil) + i2 := Get(0x7777777) + foo3 := GetByString("foo") + empty3 := GetByString("") + + if foo.Get() != foo2.Get() { + t.Error("foo/foo2 values differ") + } + if foo.Get() != foo3.Get() { + t.Error("foo/foo3 values differ") + } + if foo.Get() != "foo" { + t.Error("foo.Get not foo") + } + if foo != foo2 { + t.Error("foo/foo2 pointers differ") + } + if foo != foo3 { + t.Error("foo/foo3 pointers differ") + } + + if bar.Get() != bar2.Get() { + t.Error("bar values differ") + } + if bar.Get() != "bar" { + t.Error("bar.Get not bar") + } + if bar != bar2 { + t.Error("bar pointers differ") + } + + if i.Get() != i.Get() { + t.Error("i values differ") + } + if i.Get() != 0x7777777 { + t.Error("i.Get not 0x7777777") + } + if i != i2 { + t.Error("i pointers differ") + } + + if empty.Get() != empty2.Get() { + t.Error("empty/empty2 values differ") + } + if empty.Get() != empty.Get() { + t.Error("empty/empty3 values differ") + } + if empty.Get() != "" { + t.Error("empty.Get not empty string") + } + if empty != empty2 { + t.Error("empty/empty2 pointers differ") + } + if empty != empty3 { + t.Error("empty/empty3 pointers differ") + } + + if nilEface.Get() != nilEface2.Get() { + t.Error("nilEface values differ") + } + if nilEface.Get() != nil { + t.Error("nilEface.Get not nil") + } + if nilEface != nilEface2 { + t.Error("nilEface pointers differ") + } + + if n := mapLen(); n != 5 { + t.Errorf("map len = %d; want 4", n) + } + + wantEmpty(t) +} + +func wantEmpty(t testing.TB) { + t.Helper() + const gcTries = 5000 + for try := 0; try < gcTries; try++ { + runtime.GC() + n := mapLen() + if n == 0 { + break + } + if try == gcTries-1 { + t.Errorf("map len = %d after (%d GC tries); want 0, contents: %v", n, gcTries, mapKeys()) + } + } +} + +func TestStress(t *testing.T) { + iters := 10000 + if testing.Short() { + iters = 1000 + } + var sink []byte + for i := 0; i < iters; i++ { + _ = Get("foo") + sink = make([]byte, 1<<20) + } + _ = sink +} + +func BenchmarkStress(b *testing.B) { + done := make(chan struct{}) + defer close(done) + go func() { + for { + select { + case <-done: + return + default: + } + runtime.GC() + } + }() + + clearMap() + v1 := Get("foo") + b.ReportAllocs() + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + v2 := Get("foo") + if v1 != v2 { + b.Fatal("wrong value") + } + // And also a key we don't retain: + _ = Get("bar") + } + }) + runtime.GC() + wantEmpty(b) +} + +func mapLen() int { + mu.Lock() + defer mu.Unlock() + return len(valMap) +} + +func mapKeys() (keys []string) { + mu.Lock() + defer mu.Unlock() + for k := range valMap { + keys = append(keys, fmt.Sprint(k)) + } + return keys +} + +func clearMap() { + mu.Lock() + defer mu.Unlock() + for k := range valMap { + delete(valMap, k) + } +} + +var ( + globalString = "not a constant" + sink string +) + +func TestGetByStringAllocs(t *testing.T) { + allocs := int(testing.AllocsPerRun(100, func() { + GetByString(globalString) + })) + if allocs != 0 { + t.Errorf("GetString allocated %d objects, want 0", allocs) + } +} + +func BenchmarkGetByString(b *testing.B) { + b.ReportAllocs() + for i := 0; i < b.N; i++ { + v := GetByString(globalString) + sink = v.Get().(string) + } +} diff --git a/src/internal/itoa/itoa.go b/src/internal/itoa/itoa.go new file mode 100644 index 0000000..c6062d9 --- /dev/null +++ b/src/internal/itoa/itoa.go @@ -0,0 +1,33 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Simple conversions to avoid depending on strconv. + +package itoa + +// Itoa converts val to a decimal string. +func Itoa(val int) string { + if val < 0 { + return "-" + Uitoa(uint(-val)) + } + return Uitoa(uint(val)) +} + +// Uitoa converts val to a decimal string. +func Uitoa(val uint) string { + if val == 0 { // avoid string allocation + return "0" + } + var buf [20]byte // big enough for 64bit value base 10 + i := len(buf) - 1 + for val >= 10 { + q := val / 10 + buf[i] = byte('0' + val - q*10) + i-- + val = q + } + // val < 10 + buf[i] = byte('0' + val) + return string(buf[i:]) +} diff --git a/src/internal/itoa/itoa_test.go b/src/internal/itoa/itoa_test.go new file mode 100644 index 0000000..71931c1 --- /dev/null +++ b/src/internal/itoa/itoa_test.go @@ -0,0 +1,40 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package itoa_test + +import ( + "fmt" + "internal/itoa" + "math" + "testing" +) + +var ( + minInt64 int64 = math.MinInt64 + maxInt64 int64 = math.MaxInt64 + maxUint64 uint64 = math.MaxUint64 +) + +func TestItoa(t *testing.T) { + tests := []int{int(minInt64), math.MinInt32, -999, -100, -1, 0, 1, 100, 999, math.MaxInt32, int(maxInt64)} + for _, tt := range tests { + got := itoa.Itoa(tt) + want := fmt.Sprint(tt) + if want != got { + t.Fatalf("Itoa(%d) = %s, want %s", tt, got, want) + } + } +} + +func TestUitoa(t *testing.T) { + tests := []uint{0, 1, 100, 999, math.MaxUint32, uint(maxUint64)} + for _, tt := range tests { + got := itoa.Uitoa(tt) + want := fmt.Sprint(tt) + if want != got { + t.Fatalf("Uitoa(%d) = %s, want %s", tt, got, want) + } + } +} diff --git a/src/internal/lazyregexp/lazyre.go b/src/internal/lazyregexp/lazyre.go new file mode 100644 index 0000000..2681af3 --- /dev/null +++ b/src/internal/lazyregexp/lazyre.go @@ -0,0 +1,78 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package lazyregexp is a thin wrapper over regexp, allowing the use of global +// regexp variables without forcing them to be compiled at init. +package lazyregexp + +import ( + "os" + "regexp" + "strings" + "sync" +) + +// Regexp is a wrapper around regexp.Regexp, where the underlying regexp will be +// compiled the first time it is needed. +type Regexp struct { + str string + once sync.Once + rx *regexp.Regexp +} + +func (r *Regexp) re() *regexp.Regexp { + r.once.Do(r.build) + return r.rx +} + +func (r *Regexp) build() { + r.rx = regexp.MustCompile(r.str) + r.str = "" +} + +func (r *Regexp) FindSubmatch(s []byte) [][]byte { + return r.re().FindSubmatch(s) +} + +func (r *Regexp) FindStringSubmatch(s string) []string { + return r.re().FindStringSubmatch(s) +} + +func (r *Regexp) FindStringSubmatchIndex(s string) []int { + return r.re().FindStringSubmatchIndex(s) +} + +func (r *Regexp) ReplaceAllString(src, repl string) string { + return r.re().ReplaceAllString(src, repl) +} + +func (r *Regexp) FindString(s string) string { + return r.re().FindString(s) +} + +func (r *Regexp) FindAllString(s string, n int) []string { + return r.re().FindAllString(s, n) +} + +func (r *Regexp) MatchString(s string) bool { + return r.re().MatchString(s) +} + +func (r *Regexp) SubexpNames() []string { + return r.re().SubexpNames() +} + +var inTest = len(os.Args) > 0 && strings.HasSuffix(strings.TrimSuffix(os.Args[0], ".exe"), ".test") + +// New creates a new lazy regexp, delaying the compiling work until it is first +// needed. If the code is being run as part of tests, the regexp compiling will +// happen immediately. +func New(str string) *Regexp { + lr := &Regexp{str: str} + if inTest { + // In tests, always compile the regexps early. + lr.re() + } + return lr +} diff --git a/src/internal/lazytemplate/lazytemplate.go b/src/internal/lazytemplate/lazytemplate.go new file mode 100644 index 0000000..8eeed5a --- /dev/null +++ b/src/internal/lazytemplate/lazytemplate.go @@ -0,0 +1,52 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package lazytemplate is a thin wrapper over text/template, allowing the use +// of global template variables without forcing them to be parsed at init. +package lazytemplate + +import ( + "io" + "os" + "strings" + "sync" + "text/template" +) + +// Template is a wrapper around text/template.Template, where the underlying +// template will be parsed the first time it is needed. +type Template struct { + name, text string + + once sync.Once + tmpl *template.Template +} + +func (r *Template) tp() *template.Template { + r.once.Do(r.build) + return r.tmpl +} + +func (r *Template) build() { + r.tmpl = template.Must(template.New(r.name).Parse(r.text)) + r.name, r.text = "", "" +} + +func (r *Template) Execute(w io.Writer, data any) error { + return r.tp().Execute(w, data) +} + +var inTest = len(os.Args) > 0 && strings.HasSuffix(strings.TrimSuffix(os.Args[0], ".exe"), ".test") + +// New creates a new lazy template, delaying the parsing work until it is first +// needed. If the code is being run as part of tests, the template parsing will +// happen immediately. +func New(name, text string) *Template { + lt := &Template{name: name, text: text} + if inTest { + // In tests, always parse the templates early. + lt.tp() + } + return lt +} diff --git a/src/internal/nettrace/nettrace.go b/src/internal/nettrace/nettrace.go new file mode 100644 index 0000000..0a2bf92 --- /dev/null +++ b/src/internal/nettrace/nettrace.go @@ -0,0 +1,46 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package nettrace contains internal hooks for tracing activity in +// the net package. This package is purely internal for use by the +// net/http/httptrace package and has no stable API exposed to end +// users. +package nettrace + +// TraceKey is a context.Context Value key. Its associated value should +// be a *Trace struct. +type TraceKey struct{} + +// LookupIPAltResolverKey is a context.Context Value key used by tests to +// specify an alternate resolver func. +// It is not exposed to outsider users. (But see issue 12503) +// The value should be the same type as lookupIP: +// +// func lookupIP(ctx context.Context, host string) ([]IPAddr, error) +type LookupIPAltResolverKey struct{} + +// Trace contains a set of hooks for tracing events within +// the net package. Any specific hook may be nil. +type Trace struct { + // DNSStart is called with the hostname of a DNS lookup + // before it begins. + DNSStart func(name string) + + // DNSDone is called after a DNS lookup completes (or fails). + // The coalesced parameter is whether singleflight de-duped + // the call. The addrs are of type net.IPAddr but can't + // actually be for circular dependency reasons. + DNSDone func(netIPs []any, coalesced bool, err error) + + // ConnectStart is called before a Dial, excluding Dials made + // during DNS lookups. In the case of DualStack (Happy Eyeballs) + // dialing, this may be called multiple times, from multiple + // goroutines. + ConnectStart func(network, addr string) + + // ConnectStart is called after a Dial with the results, excluding + // Dials made during DNS lookups. It may also be called multiple + // times, like ConnectStart. + ConnectDone func(network, addr string, err error) +} diff --git a/src/internal/obscuretestdata/obscuretestdata.go b/src/internal/obscuretestdata/obscuretestdata.go new file mode 100644 index 0000000..d54d3f6 --- /dev/null +++ b/src/internal/obscuretestdata/obscuretestdata.go @@ -0,0 +1,65 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package obscuretestdata contains functionality used by tests to more easily +// work with testdata that must be obscured primarily due to +// golang.org/issue/34986. +package obscuretestdata + +import ( + "encoding/base64" + "io" + "os" +) + +// Rot13 returns the rot13 encoding or decoding of its input. +func Rot13(data []byte) []byte { + out := make([]byte, len(data)) + copy(out, data) + for i, c := range out { + switch { + case 'A' <= c && c <= 'M' || 'a' <= c && c <= 'm': + out[i] = c + 13 + case 'N' <= c && c <= 'Z' || 'n' <= c && c <= 'z': + out[i] = c - 13 + } + } + return out +} + +// DecodeToTempFile decodes the named file to a temporary location. +// If successful, it returns the path of the decoded file. +// The caller is responsible for ensuring that the temporary file is removed. +func DecodeToTempFile(name string) (path string, err error) { + f, err := os.Open(name) + if err != nil { + return "", err + } + defer f.Close() + + tmp, err := os.CreateTemp("", "obscuretestdata-decoded-") + if err != nil { + return "", err + } + if _, err := io.Copy(tmp, base64.NewDecoder(base64.StdEncoding, f)); err != nil { + tmp.Close() + os.Remove(tmp.Name()) + return "", err + } + if err := tmp.Close(); err != nil { + os.Remove(tmp.Name()) + return "", err + } + return tmp.Name(), nil +} + +// ReadFile reads the named file and returns its decoded contents. +func ReadFile(name string) ([]byte, error) { + f, err := os.Open(name) + if err != nil { + return nil, err + } + defer f.Close() + return io.ReadAll(base64.NewDecoder(base64.StdEncoding, f)) +} diff --git a/src/internal/oserror/errors.go b/src/internal/oserror/errors.go new file mode 100644 index 0000000..28a1ab3 --- /dev/null +++ b/src/internal/oserror/errors.go @@ -0,0 +1,18 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package oserror defines errors values used in the os package. +// +// These types are defined here to permit the syscall package to reference them. +package oserror + +import "errors" + +var ( + ErrInvalid = errors.New("invalid argument") + ErrPermission = errors.New("permission denied") + ErrExist = errors.New("file already exists") + ErrNotExist = errors.New("file does not exist") + ErrClosed = errors.New("file already closed") +) diff --git a/src/internal/pkgbits/codes.go b/src/internal/pkgbits/codes.go new file mode 100644 index 0000000..f0cabde --- /dev/null +++ b/src/internal/pkgbits/codes.go @@ -0,0 +1,77 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +// A Code is an enum value that can be encoded into bitstreams. +// +// Code types are preferable for enum types, because they allow +// Decoder to detect desyncs. +type Code interface { + // Marker returns the SyncMarker for the Code's dynamic type. + Marker() SyncMarker + + // Value returns the Code's ordinal value. + Value() int +} + +// A CodeVal distinguishes among go/constant.Value encodings. +type CodeVal int + +func (c CodeVal) Marker() SyncMarker { return SyncVal } +func (c CodeVal) Value() int { return int(c) } + +// Note: These values are public and cannot be changed without +// updating the go/types importers. + +const ( + ValBool CodeVal = iota + ValString + ValInt64 + ValBigInt + ValBigRat + ValBigFloat +) + +// A CodeType distinguishes among go/types.Type encodings. +type CodeType int + +func (c CodeType) Marker() SyncMarker { return SyncType } +func (c CodeType) Value() int { return int(c) } + +// Note: These values are public and cannot be changed without +// updating the go/types importers. + +const ( + TypeBasic CodeType = iota + TypeNamed + TypePointer + TypeSlice + TypeArray + TypeChan + TypeMap + TypeSignature + TypeStruct + TypeInterface + TypeUnion + TypeTypeParam +) + +// A CodeObj distinguishes among go/types.Object encodings. +type CodeObj int + +func (c CodeObj) Marker() SyncMarker { return SyncCodeObj } +func (c CodeObj) Value() int { return int(c) } + +// Note: These values are public and cannot be changed without +// updating the go/types importers. + +const ( + ObjAlias CodeObj = iota + ObjConst + ObjType + ObjFunc + ObjVar + ObjStub +) diff --git a/src/internal/pkgbits/decoder.go b/src/internal/pkgbits/decoder.go new file mode 100644 index 0000000..4fe024d --- /dev/null +++ b/src/internal/pkgbits/decoder.go @@ -0,0 +1,515 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +import ( + "encoding/binary" + "errors" + "fmt" + "go/constant" + "go/token" + "io" + "math/big" + "os" + "runtime" + "strings" +) + +// A PkgDecoder provides methods for decoding a package's Unified IR +// export data. +type PkgDecoder struct { + // version is the file format version. + version uint32 + + // sync indicates whether the file uses sync markers. + sync bool + + // pkgPath is the package path for the package to be decoded. + // + // TODO(mdempsky): Remove; unneeded since CL 391014. + pkgPath string + + // elemData is the full data payload of the encoded package. + // Elements are densely and contiguously packed together. + // + // The last 8 bytes of elemData are the package fingerprint. + elemData string + + // elemEnds stores the byte-offset end positions of element + // bitstreams within elemData. + // + // For example, element I's bitstream data starts at elemEnds[I-1] + // (or 0, if I==0) and ends at elemEnds[I]. + // + // Note: elemEnds is indexed by absolute indices, not + // section-relative indices. + elemEnds []uint32 + + // elemEndsEnds stores the index-offset end positions of relocation + // sections within elemEnds. + // + // For example, section K's end positions start at elemEndsEnds[K-1] + // (or 0, if K==0) and end at elemEndsEnds[K]. + elemEndsEnds [numRelocs]uint32 + + scratchRelocEnt []RelocEnt +} + +// PkgPath returns the package path for the package +// +// TODO(mdempsky): Remove; unneeded since CL 391014. +func (pr *PkgDecoder) PkgPath() string { return pr.pkgPath } + +// SyncMarkers reports whether pr uses sync markers. +func (pr *PkgDecoder) SyncMarkers() bool { return pr.sync } + +// NewPkgDecoder returns a PkgDecoder initialized to read the Unified +// IR export data from input. pkgPath is the package path for the +// compilation unit that produced the export data. +// +// TODO(mdempsky): Remove pkgPath parameter; unneeded since CL 391014. +func NewPkgDecoder(pkgPath, input string) PkgDecoder { + pr := PkgDecoder{ + pkgPath: pkgPath, + } + + // TODO(mdempsky): Implement direct indexing of input string to + // avoid copying the position information. + + r := strings.NewReader(input) + + assert(binary.Read(r, binary.LittleEndian, &pr.version) == nil) + + switch pr.version { + default: + panic(fmt.Errorf("unsupported version: %v", pr.version)) + case 0: + // no flags + case 1: + var flags uint32 + assert(binary.Read(r, binary.LittleEndian, &flags) == nil) + pr.sync = flags&flagSyncMarkers != 0 + } + + assert(binary.Read(r, binary.LittleEndian, pr.elemEndsEnds[:]) == nil) + + pr.elemEnds = make([]uint32, pr.elemEndsEnds[len(pr.elemEndsEnds)-1]) + assert(binary.Read(r, binary.LittleEndian, pr.elemEnds[:]) == nil) + + pos, err := r.Seek(0, io.SeekCurrent) + assert(err == nil) + + pr.elemData = input[pos:] + assert(len(pr.elemData)-8 == int(pr.elemEnds[len(pr.elemEnds)-1])) + + return pr +} + +// NumElems returns the number of elements in section k. +func (pr *PkgDecoder) NumElems(k RelocKind) int { + count := int(pr.elemEndsEnds[k]) + if k > 0 { + count -= int(pr.elemEndsEnds[k-1]) + } + return count +} + +// TotalElems returns the total number of elements across all sections. +func (pr *PkgDecoder) TotalElems() int { + return len(pr.elemEnds) +} + +// Fingerprint returns the package fingerprint. +func (pr *PkgDecoder) Fingerprint() [8]byte { + var fp [8]byte + copy(fp[:], pr.elemData[len(pr.elemData)-8:]) + return fp +} + +// AbsIdx returns the absolute index for the given (section, index) +// pair. +func (pr *PkgDecoder) AbsIdx(k RelocKind, idx Index) int { + absIdx := int(idx) + if k > 0 { + absIdx += int(pr.elemEndsEnds[k-1]) + } + if absIdx >= int(pr.elemEndsEnds[k]) { + errorf("%v:%v is out of bounds; %v", k, idx, pr.elemEndsEnds) + } + return absIdx +} + +// DataIdx returns the raw element bitstream for the given (section, +// index) pair. +func (pr *PkgDecoder) DataIdx(k RelocKind, idx Index) string { + absIdx := pr.AbsIdx(k, idx) + + var start uint32 + if absIdx > 0 { + start = pr.elemEnds[absIdx-1] + } + end := pr.elemEnds[absIdx] + + return pr.elemData[start:end] +} + +// StringIdx returns the string value for the given string index. +func (pr *PkgDecoder) StringIdx(idx Index) string { + return pr.DataIdx(RelocString, idx) +} + +// NewDecoder returns a Decoder for the given (section, index) pair, +// and decodes the given SyncMarker from the element bitstream. +func (pr *PkgDecoder) NewDecoder(k RelocKind, idx Index, marker SyncMarker) Decoder { + r := pr.NewDecoderRaw(k, idx) + r.Sync(marker) + return r +} + +// TempDecoder returns a Decoder for the given (section, index) pair, +// and decodes the given SyncMarker from the element bitstream. +// If possible the Decoder should be RetireDecoder'd when it is no longer +// needed, this will avoid heap allocations. +func (pr *PkgDecoder) TempDecoder(k RelocKind, idx Index, marker SyncMarker) Decoder { + r := pr.TempDecoderRaw(k, idx) + r.Sync(marker) + return r +} + +func (pr *PkgDecoder) RetireDecoder(d *Decoder) { + pr.scratchRelocEnt = d.Relocs + d.Relocs = nil +} + +// NewDecoderRaw returns a Decoder for the given (section, index) pair. +// +// Most callers should use NewDecoder instead. +func (pr *PkgDecoder) NewDecoderRaw(k RelocKind, idx Index) Decoder { + r := Decoder{ + common: pr, + k: k, + Idx: idx, + } + + r.Data.Reset(pr.DataIdx(k, idx)) + r.Sync(SyncRelocs) + r.Relocs = make([]RelocEnt, r.Len()) + for i := range r.Relocs { + r.Sync(SyncReloc) + r.Relocs[i] = RelocEnt{RelocKind(r.Len()), Index(r.Len())} + } + + return r +} + +func (pr *PkgDecoder) TempDecoderRaw(k RelocKind, idx Index) Decoder { + r := Decoder{ + common: pr, + k: k, + Idx: idx, + } + + r.Data.Reset(pr.DataIdx(k, idx)) + r.Sync(SyncRelocs) + l := r.Len() + if cap(pr.scratchRelocEnt) >= l { + r.Relocs = pr.scratchRelocEnt[:l] + pr.scratchRelocEnt = nil + } else { + r.Relocs = make([]RelocEnt, l) + } + for i := range r.Relocs { + r.Sync(SyncReloc) + r.Relocs[i] = RelocEnt{RelocKind(r.Len()), Index(r.Len())} + } + + return r +} + +// A Decoder provides methods for decoding an individual element's +// bitstream data. +type Decoder struct { + common *PkgDecoder + + Relocs []RelocEnt + Data strings.Reader + + k RelocKind + Idx Index +} + +func (r *Decoder) checkErr(err error) { + if err != nil { + errorf("unexpected decoding error: %w", err) + } +} + +func (r *Decoder) rawUvarint() uint64 { + x, err := readUvarint(&r.Data) + r.checkErr(err) + return x +} + +// readUvarint is a type-specialized copy of encoding/binary.ReadUvarint. +// This avoids the interface conversion and thus has better escape properties, +// which flows up the stack. +func readUvarint(r *strings.Reader) (uint64, error) { + var x uint64 + var s uint + for i := 0; i < binary.MaxVarintLen64; i++ { + b, err := r.ReadByte() + if err != nil { + if i > 0 && err == io.EOF { + err = io.ErrUnexpectedEOF + } + return x, err + } + if b < 0x80 { + if i == binary.MaxVarintLen64-1 && b > 1 { + return x, overflow + } + return x | uint64(b)<<s, nil + } + x |= uint64(b&0x7f) << s + s += 7 + } + return x, overflow +} + +var overflow = errors.New("pkgbits: readUvarint overflows a 64-bit integer") + +func (r *Decoder) rawVarint() int64 { + ux := r.rawUvarint() + + // Zig-zag decode. + x := int64(ux >> 1) + if ux&1 != 0 { + x = ^x + } + return x +} + +func (r *Decoder) rawReloc(k RelocKind, idx int) Index { + e := r.Relocs[idx] + assert(e.Kind == k) + return e.Idx +} + +// Sync decodes a sync marker from the element bitstream and asserts +// that it matches the expected marker. +// +// If EnableSync is false, then Sync is a no-op. +func (r *Decoder) Sync(mWant SyncMarker) { + if !r.common.sync { + return + } + + pos, _ := r.Data.Seek(0, io.SeekCurrent) + mHave := SyncMarker(r.rawUvarint()) + writerPCs := make([]int, r.rawUvarint()) + for i := range writerPCs { + writerPCs[i] = int(r.rawUvarint()) + } + + if mHave == mWant { + return + } + + // There's some tension here between printing: + // + // (1) full file paths that tools can recognize (e.g., so emacs + // hyperlinks the "file:line" text for easy navigation), or + // + // (2) short file paths that are easier for humans to read (e.g., by + // omitting redundant or irrelevant details, so it's easier to + // focus on the useful bits that remain). + // + // The current formatting favors the former, as it seems more + // helpful in practice. But perhaps the formatting could be improved + // to better address both concerns. For example, use relative file + // paths if they would be shorter, or rewrite file paths to contain + // "$GOROOT" (like objabi.AbsFile does) if tools can be taught how + // to reliably expand that again. + + fmt.Printf("export data desync: package %q, section %v, index %v, offset %v\n", r.common.pkgPath, r.k, r.Idx, pos) + + fmt.Printf("\nfound %v, written at:\n", mHave) + if len(writerPCs) == 0 { + fmt.Printf("\t[stack trace unavailable; recompile package %q with -d=syncframes]\n", r.common.pkgPath) + } + for _, pc := range writerPCs { + fmt.Printf("\t%s\n", r.common.StringIdx(r.rawReloc(RelocString, pc))) + } + + fmt.Printf("\nexpected %v, reading at:\n", mWant) + var readerPCs [32]uintptr // TODO(mdempsky): Dynamically size? + n := runtime.Callers(2, readerPCs[:]) + for _, pc := range fmtFrames(readerPCs[:n]...) { + fmt.Printf("\t%s\n", pc) + } + + // We already printed a stack trace for the reader, so now we can + // simply exit. Printing a second one with panic or base.Fatalf + // would just be noise. + os.Exit(1) +} + +// Bool decodes and returns a bool value from the element bitstream. +func (r *Decoder) Bool() bool { + r.Sync(SyncBool) + x, err := r.Data.ReadByte() + r.checkErr(err) + assert(x < 2) + return x != 0 +} + +// Int64 decodes and returns an int64 value from the element bitstream. +func (r *Decoder) Int64() int64 { + r.Sync(SyncInt64) + return r.rawVarint() +} + +// Int64 decodes and returns a uint64 value from the element bitstream. +func (r *Decoder) Uint64() uint64 { + r.Sync(SyncUint64) + return r.rawUvarint() +} + +// Len decodes and returns a non-negative int value from the element bitstream. +func (r *Decoder) Len() int { x := r.Uint64(); v := int(x); assert(uint64(v) == x); return v } + +// Int decodes and returns an int value from the element bitstream. +func (r *Decoder) Int() int { x := r.Int64(); v := int(x); assert(int64(v) == x); return v } + +// Uint decodes and returns a uint value from the element bitstream. +func (r *Decoder) Uint() uint { x := r.Uint64(); v := uint(x); assert(uint64(v) == x); return v } + +// Code decodes a Code value from the element bitstream and returns +// its ordinal value. It's the caller's responsibility to convert the +// result to an appropriate Code type. +// +// TODO(mdempsky): Ideally this method would have signature "Code[T +// Code] T" instead, but we don't allow generic methods and the +// compiler can't depend on generics yet anyway. +func (r *Decoder) Code(mark SyncMarker) int { + r.Sync(mark) + return r.Len() +} + +// Reloc decodes a relocation of expected section k from the element +// bitstream and returns an index to the referenced element. +func (r *Decoder) Reloc(k RelocKind) Index { + r.Sync(SyncUseReloc) + return r.rawReloc(k, r.Len()) +} + +// String decodes and returns a string value from the element +// bitstream. +func (r *Decoder) String() string { + r.Sync(SyncString) + return r.common.StringIdx(r.Reloc(RelocString)) +} + +// Strings decodes and returns a variable-length slice of strings from +// the element bitstream. +func (r *Decoder) Strings() []string { + res := make([]string, r.Len()) + for i := range res { + res[i] = r.String() + } + return res +} + +// Value decodes and returns a constant.Value from the element +// bitstream. +func (r *Decoder) Value() constant.Value { + r.Sync(SyncValue) + isComplex := r.Bool() + val := r.scalar() + if isComplex { + val = constant.BinaryOp(val, token.ADD, constant.MakeImag(r.scalar())) + } + return val +} + +func (r *Decoder) scalar() constant.Value { + switch tag := CodeVal(r.Code(SyncVal)); tag { + default: + panic(fmt.Errorf("unexpected scalar tag: %v", tag)) + + case ValBool: + return constant.MakeBool(r.Bool()) + case ValString: + return constant.MakeString(r.String()) + case ValInt64: + return constant.MakeInt64(r.Int64()) + case ValBigInt: + return constant.Make(r.bigInt()) + case ValBigRat: + num := r.bigInt() + denom := r.bigInt() + return constant.Make(new(big.Rat).SetFrac(num, denom)) + case ValBigFloat: + return constant.Make(r.bigFloat()) + } +} + +func (r *Decoder) bigInt() *big.Int { + v := new(big.Int).SetBytes([]byte(r.String())) + if r.Bool() { + v.Neg(v) + } + return v +} + +func (r *Decoder) bigFloat() *big.Float { + v := new(big.Float).SetPrec(512) + assert(v.UnmarshalText([]byte(r.String())) == nil) + return v +} + +// @@@ Helpers + +// TODO(mdempsky): These should probably be removed. I think they're a +// smell that the export data format is not yet quite right. + +// PeekPkgPath returns the package path for the specified package +// index. +func (pr *PkgDecoder) PeekPkgPath(idx Index) string { + var path string + { + r := pr.TempDecoder(RelocPkg, idx, SyncPkgDef) + path = r.String() + pr.RetireDecoder(&r) + } + if path == "" { + path = pr.pkgPath + } + return path +} + +// PeekObj returns the package path, object name, and CodeObj for the +// specified object index. +func (pr *PkgDecoder) PeekObj(idx Index) (string, string, CodeObj) { + var ridx Index + var name string + var rcode int + { + r := pr.TempDecoder(RelocName, idx, SyncObject1) + r.Sync(SyncSym) + r.Sync(SyncPkg) + ridx = r.Reloc(RelocPkg) + name = r.String() + rcode = r.Code(SyncCodeObj) + pr.RetireDecoder(&r) + } + + path := pr.PeekPkgPath(ridx) + assert(name != "") + + tag := CodeObj(rcode) + + return path, name, tag +} diff --git a/src/internal/pkgbits/doc.go b/src/internal/pkgbits/doc.go new file mode 100644 index 0000000..4862e39 --- /dev/null +++ b/src/internal/pkgbits/doc.go @@ -0,0 +1,30 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package pkgbits implements low-level coding abstractions for +// Unified IR's export data format. +// +// At a low-level, a package is a collection of bitstream elements. +// Each element has a "kind" and a dense, non-negative index. +// Elements can be randomly accessed given their kind and index. +// +// Individual elements are sequences of variable-length values (e.g., +// integers, booleans, strings, go/constant values, cross-references +// to other elements). Package pkgbits provides APIs for encoding and +// decoding these low-level values, but the details of mapping +// higher-level Go constructs into elements is left to higher-level +// abstractions. +// +// Elements may cross-reference each other with "relocations." For +// example, an element representing a pointer type has a relocation +// referring to the element type. +// +// Go constructs may be composed as a constellation of multiple +// elements. For example, a declared function may have one element to +// describe the object (e.g., its name, type, position), and a +// separate element to describe its function body. This allows readers +// some flexibility in efficiently seeking or re-reading data (e.g., +// inlining requires re-reading the function body for each inlined +// call, without needing to re-read the object-level details). +package pkgbits diff --git a/src/internal/pkgbits/encoder.go b/src/internal/pkgbits/encoder.go new file mode 100644 index 0000000..70a2cba --- /dev/null +++ b/src/internal/pkgbits/encoder.go @@ -0,0 +1,394 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +import ( + "bytes" + "crypto/md5" + "encoding/binary" + "go/constant" + "io" + "math/big" + "runtime" + "strings" +) + +// currentVersion is the current version number. +// +// - v0: initial prototype +// +// - v1: adds the flags uint32 word +// +// TODO(mdempsky): For the next version bump: +// - remove the legacy "has init" bool from the public root +// - remove obj's "derived func instance" bool +const currentVersion uint32 = 1 + +// A PkgEncoder provides methods for encoding a package's Unified IR +// export data. +type PkgEncoder struct { + // elems holds the bitstream for previously encoded elements. + elems [numRelocs][]string + + // stringsIdx maps previously encoded strings to their index within + // the RelocString section, to allow deduplication. That is, + // elems[RelocString][stringsIdx[s]] == s (if present). + stringsIdx map[string]Index + + // syncFrames is the number of frames to write at each sync + // marker. A negative value means sync markers are omitted. + syncFrames int +} + +// SyncMarkers reports whether pw uses sync markers. +func (pw *PkgEncoder) SyncMarkers() bool { return pw.syncFrames >= 0 } + +// NewPkgEncoder returns an initialized PkgEncoder. +// +// syncFrames is the number of caller frames that should be serialized +// at Sync points. Serializing additional frames results in larger +// export data files, but can help diagnosing desync errors in +// higher-level Unified IR reader/writer code. If syncFrames is +// negative, then sync markers are omitted entirely. +func NewPkgEncoder(syncFrames int) PkgEncoder { + return PkgEncoder{ + stringsIdx: make(map[string]Index), + syncFrames: syncFrames, + } +} + +// DumpTo writes the package's encoded data to out0 and returns the +// package fingerprint. +func (pw *PkgEncoder) DumpTo(out0 io.Writer) (fingerprint [8]byte) { + h := md5.New() + out := io.MultiWriter(out0, h) + + writeUint32 := func(x uint32) { + assert(binary.Write(out, binary.LittleEndian, x) == nil) + } + + writeUint32(currentVersion) + + var flags uint32 + if pw.SyncMarkers() { + flags |= flagSyncMarkers + } + writeUint32(flags) + + // Write elemEndsEnds. + var sum uint32 + for _, elems := range &pw.elems { + sum += uint32(len(elems)) + writeUint32(sum) + } + + // Write elemEnds. + sum = 0 + for _, elems := range &pw.elems { + for _, elem := range elems { + sum += uint32(len(elem)) + writeUint32(sum) + } + } + + // Write elemData. + for _, elems := range &pw.elems { + for _, elem := range elems { + _, err := io.WriteString(out, elem) + assert(err == nil) + } + } + + // Write fingerprint. + copy(fingerprint[:], h.Sum(nil)) + _, err := out0.Write(fingerprint[:]) + assert(err == nil) + + return +} + +// StringIdx adds a string value to the strings section, if not +// already present, and returns its index. +func (pw *PkgEncoder) StringIdx(s string) Index { + if idx, ok := pw.stringsIdx[s]; ok { + assert(pw.elems[RelocString][idx] == s) + return idx + } + + idx := Index(len(pw.elems[RelocString])) + pw.elems[RelocString] = append(pw.elems[RelocString], s) + pw.stringsIdx[s] = idx + return idx +} + +// NewEncoder returns an Encoder for a new element within the given +// section, and encodes the given SyncMarker as the start of the +// element bitstream. +func (pw *PkgEncoder) NewEncoder(k RelocKind, marker SyncMarker) Encoder { + e := pw.NewEncoderRaw(k) + e.Sync(marker) + return e +} + +// NewEncoderRaw returns an Encoder for a new element within the given +// section. +// +// Most callers should use NewEncoder instead. +func (pw *PkgEncoder) NewEncoderRaw(k RelocKind) Encoder { + idx := Index(len(pw.elems[k])) + pw.elems[k] = append(pw.elems[k], "") // placeholder + + return Encoder{ + p: pw, + k: k, + Idx: idx, + } +} + +// An Encoder provides methods for encoding an individual element's +// bitstream data. +type Encoder struct { + p *PkgEncoder + + Relocs []RelocEnt + RelocMap map[RelocEnt]uint32 + Data bytes.Buffer // accumulated element bitstream data + + encodingRelocHeader bool + + k RelocKind + Idx Index // index within relocation section +} + +// Flush finalizes the element's bitstream and returns its Index. +func (w *Encoder) Flush() Index { + var sb strings.Builder + + // Backup the data so we write the relocations at the front. + var tmp bytes.Buffer + io.Copy(&tmp, &w.Data) + + // TODO(mdempsky): Consider writing these out separately so they're + // easier to strip, along with function bodies, so that we can prune + // down to just the data that's relevant to go/types. + if w.encodingRelocHeader { + panic("encodingRelocHeader already true; recursive flush?") + } + w.encodingRelocHeader = true + w.Sync(SyncRelocs) + w.Len(len(w.Relocs)) + for _, rEnt := range w.Relocs { + w.Sync(SyncReloc) + w.Len(int(rEnt.Kind)) + w.Len(int(rEnt.Idx)) + } + + io.Copy(&sb, &w.Data) + io.Copy(&sb, &tmp) + w.p.elems[w.k][w.Idx] = sb.String() + + return w.Idx +} + +func (w *Encoder) checkErr(err error) { + if err != nil { + errorf("unexpected encoding error: %v", err) + } +} + +func (w *Encoder) rawUvarint(x uint64) { + var buf [binary.MaxVarintLen64]byte + n := binary.PutUvarint(buf[:], x) + _, err := w.Data.Write(buf[:n]) + w.checkErr(err) +} + +func (w *Encoder) rawVarint(x int64) { + // Zig-zag encode. + ux := uint64(x) << 1 + if x < 0 { + ux = ^ux + } + + w.rawUvarint(ux) +} + +func (w *Encoder) rawReloc(r RelocKind, idx Index) int { + e := RelocEnt{r, idx} + if w.RelocMap != nil { + if i, ok := w.RelocMap[e]; ok { + return int(i) + } + } else { + w.RelocMap = make(map[RelocEnt]uint32) + } + + i := len(w.Relocs) + w.RelocMap[e] = uint32(i) + w.Relocs = append(w.Relocs, e) + return i +} + +func (w *Encoder) Sync(m SyncMarker) { + if !w.p.SyncMarkers() { + return + } + + // Writing out stack frame string references requires working + // relocations, but writing out the relocations themselves involves + // sync markers. To prevent infinite recursion, we simply trim the + // stack frame for sync markers within the relocation header. + var frames []string + if !w.encodingRelocHeader && w.p.syncFrames > 0 { + pcs := make([]uintptr, w.p.syncFrames) + n := runtime.Callers(2, pcs) + frames = fmtFrames(pcs[:n]...) + } + + // TODO(mdempsky): Save space by writing out stack frames as a + // linked list so we can share common stack frames. + w.rawUvarint(uint64(m)) + w.rawUvarint(uint64(len(frames))) + for _, frame := range frames { + w.rawUvarint(uint64(w.rawReloc(RelocString, w.p.StringIdx(frame)))) + } +} + +// Bool encodes and writes a bool value into the element bitstream, +// and then returns the bool value. +// +// For simple, 2-alternative encodings, the idiomatic way to call Bool +// is something like: +// +// if w.Bool(x != 0) { +// // alternative #1 +// } else { +// // alternative #2 +// } +// +// For multi-alternative encodings, use Code instead. +func (w *Encoder) Bool(b bool) bool { + w.Sync(SyncBool) + var x byte + if b { + x = 1 + } + err := w.Data.WriteByte(x) + w.checkErr(err) + return b +} + +// Int64 encodes and writes an int64 value into the element bitstream. +func (w *Encoder) Int64(x int64) { + w.Sync(SyncInt64) + w.rawVarint(x) +} + +// Uint64 encodes and writes a uint64 value into the element bitstream. +func (w *Encoder) Uint64(x uint64) { + w.Sync(SyncUint64) + w.rawUvarint(x) +} + +// Len encodes and writes a non-negative int value into the element bitstream. +func (w *Encoder) Len(x int) { assert(x >= 0); w.Uint64(uint64(x)) } + +// Int encodes and writes an int value into the element bitstream. +func (w *Encoder) Int(x int) { w.Int64(int64(x)) } + +// Len encodes and writes a uint value into the element bitstream. +func (w *Encoder) Uint(x uint) { w.Uint64(uint64(x)) } + +// Reloc encodes and writes a relocation for the given (section, +// index) pair into the element bitstream. +// +// Note: Only the index is formally written into the element +// bitstream, so bitstream decoders must know from context which +// section an encoded relocation refers to. +func (w *Encoder) Reloc(r RelocKind, idx Index) { + w.Sync(SyncUseReloc) + w.Len(w.rawReloc(r, idx)) +} + +// Code encodes and writes a Code value into the element bitstream. +func (w *Encoder) Code(c Code) { + w.Sync(c.Marker()) + w.Len(c.Value()) +} + +// String encodes and writes a string value into the element +// bitstream. +// +// Internally, strings are deduplicated by adding them to the strings +// section (if not already present), and then writing a relocation +// into the element bitstream. +func (w *Encoder) String(s string) { + w.StringRef(w.p.StringIdx(s)) +} + +// StringRef writes a reference to the given index, which must be a +// previously encoded string value. +func (w *Encoder) StringRef(idx Index) { + w.Sync(SyncString) + w.Reloc(RelocString, idx) +} + +// Strings encodes and writes a variable-length slice of strings into +// the element bitstream. +func (w *Encoder) Strings(ss []string) { + w.Len(len(ss)) + for _, s := range ss { + w.String(s) + } +} + +// Value encodes and writes a constant.Value into the element +// bitstream. +func (w *Encoder) Value(val constant.Value) { + w.Sync(SyncValue) + if w.Bool(val.Kind() == constant.Complex) { + w.scalar(constant.Real(val)) + w.scalar(constant.Imag(val)) + } else { + w.scalar(val) + } +} + +func (w *Encoder) scalar(val constant.Value) { + switch v := constant.Val(val).(type) { + default: + errorf("unhandled %v (%v)", val, val.Kind()) + case bool: + w.Code(ValBool) + w.Bool(v) + case string: + w.Code(ValString) + w.String(v) + case int64: + w.Code(ValInt64) + w.Int64(v) + case *big.Int: + w.Code(ValBigInt) + w.bigInt(v) + case *big.Rat: + w.Code(ValBigRat) + w.bigInt(v.Num()) + w.bigInt(v.Denom()) + case *big.Float: + w.Code(ValBigFloat) + w.bigFloat(v) + } +} + +func (w *Encoder) bigInt(v *big.Int) { + b := v.Bytes() + w.String(string(b)) // TODO: More efficient encoding. + w.Bool(v.Sign() < 0) +} + +func (w *Encoder) bigFloat(v *big.Float) { + b := v.Append(nil, 'p', -1) + w.String(string(b)) // TODO: More efficient encoding. +} diff --git a/src/internal/pkgbits/flags.go b/src/internal/pkgbits/flags.go new file mode 100644 index 0000000..6542227 --- /dev/null +++ b/src/internal/pkgbits/flags.go @@ -0,0 +1,9 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +const ( + flagSyncMarkers = 1 << iota // file format contains sync markers +) diff --git a/src/internal/pkgbits/reloc.go b/src/internal/pkgbits/reloc.go new file mode 100644 index 0000000..fcdfb97 --- /dev/null +++ b/src/internal/pkgbits/reloc.go @@ -0,0 +1,42 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +// A RelocKind indicates a particular section within a unified IR export. +type RelocKind int32 + +// An Index represents a bitstream element index within a particular +// section. +type Index int32 + +// A relocEnt (relocation entry) is an entry in an element's local +// reference table. +// +// TODO(mdempsky): Rename this too. +type RelocEnt struct { + Kind RelocKind + Idx Index +} + +// Reserved indices within the meta relocation section. +const ( + PublicRootIdx Index = 0 + PrivateRootIdx Index = 1 +) + +const ( + RelocString RelocKind = iota + RelocMeta + RelocPosBase + RelocPkg + RelocName + RelocType + RelocObj + RelocObjExt + RelocObjDict + RelocBody + + numRelocs = iota +) diff --git a/src/internal/pkgbits/support.go b/src/internal/pkgbits/support.go new file mode 100644 index 0000000..f7579df --- /dev/null +++ b/src/internal/pkgbits/support.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +import "fmt" + +func assert(b bool) { + if !b { + panic("assertion failed") + } +} + +func errorf(format string, args ...any) { + panic(fmt.Errorf(format, args...)) +} diff --git a/src/internal/pkgbits/sync.go b/src/internal/pkgbits/sync.go new file mode 100644 index 0000000..1520b73 --- /dev/null +++ b/src/internal/pkgbits/sync.go @@ -0,0 +1,136 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkgbits + +import ( + "fmt" + "runtime" + "strings" +) + +// fmtFrames formats a backtrace for reporting reader/writer desyncs. +func fmtFrames(pcs ...uintptr) []string { + res := make([]string, 0, len(pcs)) + walkFrames(pcs, func(file string, line int, name string, offset uintptr) { + // Trim package from function name. It's just redundant noise. + name = strings.TrimPrefix(name, "cmd/compile/internal/noder.") + + res = append(res, fmt.Sprintf("%s:%v: %s +0x%v", file, line, name, offset)) + }) + return res +} + +type frameVisitor func(file string, line int, name string, offset uintptr) + +// walkFrames calls visit for each call frame represented by pcs. +// +// pcs should be a slice of PCs, as returned by runtime.Callers. +func walkFrames(pcs []uintptr, visit frameVisitor) { + if len(pcs) == 0 { + return + } + + frames := runtime.CallersFrames(pcs) + for { + frame, more := frames.Next() + visit(frame.File, frame.Line, frame.Function, frame.PC-frame.Entry) + if !more { + return + } + } +} + +// SyncMarker is an enum type that represents markers that may be +// written to export data to ensure the reader and writer stay +// synchronized. +type SyncMarker int + +//go:generate stringer -type=SyncMarker -trimprefix=Sync + +const ( + _ SyncMarker = iota + + // Public markers (known to go/types importers). + + // Low-level coding markers. + SyncEOF + SyncBool + SyncInt64 + SyncUint64 + SyncString + SyncValue + SyncVal + SyncRelocs + SyncReloc + SyncUseReloc + + // Higher-level object and type markers. + SyncPublic + SyncPos + SyncPosBase + SyncObject + SyncObject1 + SyncPkg + SyncPkgDef + SyncMethod + SyncType + SyncTypeIdx + SyncTypeParamNames + SyncSignature + SyncParams + SyncParam + SyncCodeObj + SyncSym + SyncLocalIdent + SyncSelector + + // Private markers (only known to cmd/compile). + SyncPrivate + + SyncFuncExt + SyncVarExt + SyncTypeExt + SyncPragma + + SyncExprList + SyncExprs + SyncExpr + SyncExprType + SyncAssign + SyncOp + SyncFuncLit + SyncCompLit + + SyncDecl + SyncFuncBody + SyncOpenScope + SyncCloseScope + SyncCloseAnotherScope + SyncDeclNames + SyncDeclName + + SyncStmts + SyncBlockStmt + SyncIfStmt + SyncForStmt + SyncSwitchStmt + SyncRangeStmt + SyncCaseClause + SyncCommClause + SyncSelectStmt + SyncDecls + SyncLabeledStmt + SyncUseObjLocal + SyncAddLocal + SyncLinkname + SyncStmt1 + SyncStmtsEnd + SyncLabel + SyncOptLabel + + SyncMultiExpr + SyncRType + SyncConvRTTI +) diff --git a/src/internal/pkgbits/syncmarker_string.go b/src/internal/pkgbits/syncmarker_string.go new file mode 100644 index 0000000..582ad56 --- /dev/null +++ b/src/internal/pkgbits/syncmarker_string.go @@ -0,0 +1,92 @@ +// Code generated by "stringer -type=SyncMarker -trimprefix=Sync"; DO NOT EDIT. + +package pkgbits + +import "strconv" + +func _() { + // An "invalid array index" compiler error signifies that the constant values have changed. + // Re-run the stringer command to generate them again. + var x [1]struct{} + _ = x[SyncEOF-1] + _ = x[SyncBool-2] + _ = x[SyncInt64-3] + _ = x[SyncUint64-4] + _ = x[SyncString-5] + _ = x[SyncValue-6] + _ = x[SyncVal-7] + _ = x[SyncRelocs-8] + _ = x[SyncReloc-9] + _ = x[SyncUseReloc-10] + _ = x[SyncPublic-11] + _ = x[SyncPos-12] + _ = x[SyncPosBase-13] + _ = x[SyncObject-14] + _ = x[SyncObject1-15] + _ = x[SyncPkg-16] + _ = x[SyncPkgDef-17] + _ = x[SyncMethod-18] + _ = x[SyncType-19] + _ = x[SyncTypeIdx-20] + _ = x[SyncTypeParamNames-21] + _ = x[SyncSignature-22] + _ = x[SyncParams-23] + _ = x[SyncParam-24] + _ = x[SyncCodeObj-25] + _ = x[SyncSym-26] + _ = x[SyncLocalIdent-27] + _ = x[SyncSelector-28] + _ = x[SyncPrivate-29] + _ = x[SyncFuncExt-30] + _ = x[SyncVarExt-31] + _ = x[SyncTypeExt-32] + _ = x[SyncPragma-33] + _ = x[SyncExprList-34] + _ = x[SyncExprs-35] + _ = x[SyncExpr-36] + _ = x[SyncExprType-37] + _ = x[SyncAssign-38] + _ = x[SyncOp-39] + _ = x[SyncFuncLit-40] + _ = x[SyncCompLit-41] + _ = x[SyncDecl-42] + _ = x[SyncFuncBody-43] + _ = x[SyncOpenScope-44] + _ = x[SyncCloseScope-45] + _ = x[SyncCloseAnotherScope-46] + _ = x[SyncDeclNames-47] + _ = x[SyncDeclName-48] + _ = x[SyncStmts-49] + _ = x[SyncBlockStmt-50] + _ = x[SyncIfStmt-51] + _ = x[SyncForStmt-52] + _ = x[SyncSwitchStmt-53] + _ = x[SyncRangeStmt-54] + _ = x[SyncCaseClause-55] + _ = x[SyncCommClause-56] + _ = x[SyncSelectStmt-57] + _ = x[SyncDecls-58] + _ = x[SyncLabeledStmt-59] + _ = x[SyncUseObjLocal-60] + _ = x[SyncAddLocal-61] + _ = x[SyncLinkname-62] + _ = x[SyncStmt1-63] + _ = x[SyncStmtsEnd-64] + _ = x[SyncLabel-65] + _ = x[SyncOptLabel-66] + _ = x[SyncMultiExpr-67] + _ = x[SyncRType-68] + _ = x[SyncConvRTTI-69] +} + +const _SyncMarker_name = "EOFBoolInt64Uint64StringValueValRelocsRelocUseRelocPublicPosPosBaseObjectObject1PkgPkgDefMethodTypeTypeIdxTypeParamNamesSignatureParamsParamCodeObjSymLocalIdentSelectorPrivateFuncExtVarExtTypeExtPragmaExprListExprsExprExprTypeAssignOpFuncLitCompLitDeclFuncBodyOpenScopeCloseScopeCloseAnotherScopeDeclNamesDeclNameStmtsBlockStmtIfStmtForStmtSwitchStmtRangeStmtCaseClauseCommClauseSelectStmtDeclsLabeledStmtUseObjLocalAddLocalLinknameStmt1StmtsEndLabelOptLabelMultiExprRTypeConvRTTI" + +var _SyncMarker_index = [...]uint16{0, 3, 7, 12, 18, 24, 29, 32, 38, 43, 51, 57, 60, 67, 73, 80, 83, 89, 95, 99, 106, 120, 129, 135, 140, 147, 150, 160, 168, 175, 182, 188, 195, 201, 209, 214, 218, 226, 232, 234, 241, 248, 252, 260, 269, 279, 296, 305, 313, 318, 327, 333, 340, 350, 359, 369, 379, 389, 394, 405, 416, 424, 432, 437, 445, 450, 458, 467, 472, 480} + +func (i SyncMarker) String() string { + i -= 1 + if i < 0 || i >= SyncMarker(len(_SyncMarker_index)-1) { + return "SyncMarker(" + strconv.FormatInt(int64(i+1), 10) + ")" + } + return _SyncMarker_name[_SyncMarker_index[i]:_SyncMarker_index[i+1]] +} diff --git a/src/internal/platform/supported.go b/src/internal/platform/supported.go new file mode 100644 index 0000000..230a952 --- /dev/null +++ b/src/internal/platform/supported.go @@ -0,0 +1,286 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:generate go test . -run=TestGenerated -fix + +package platform + +// An OSArch is a pair of GOOS and GOARCH values indicating a platform. +type OSArch struct { + GOOS, GOARCH string +} + +func (p OSArch) String() string { + return p.GOOS + "/" + p.GOARCH +} + +// RaceDetectorSupported reports whether goos/goarch supports the race +// detector. There is a copy of this function in cmd/dist/test.go. +// Race detector only supports 48-bit VMA on arm64. But it will always +// return true for arm64, because we don't have VMA size information during +// the compile time. +func RaceDetectorSupported(goos, goarch string) bool { + switch goos { + case "linux": + return goarch == "amd64" || goarch == "ppc64le" || goarch == "arm64" || goarch == "s390x" + case "darwin": + return goarch == "amd64" || goarch == "arm64" + case "freebsd", "netbsd", "openbsd", "windows": + return goarch == "amd64" + default: + return false + } +} + +// MSanSupported reports whether goos/goarch supports the memory +// sanitizer option. +func MSanSupported(goos, goarch string) bool { + switch goos { + case "linux": + return goarch == "amd64" || goarch == "arm64" + case "freebsd": + return goarch == "amd64" + default: + return false + } +} + +// ASanSupported reports whether goos/goarch supports the address +// sanitizer option. +func ASanSupported(goos, goarch string) bool { + switch goos { + case "linux": + return goarch == "arm64" || goarch == "amd64" || goarch == "riscv64" || goarch == "ppc64le" + default: + return false + } +} + +// FuzzSupported reports whether goos/goarch supports fuzzing +// ('go test -fuzz=.'). +func FuzzSupported(goos, goarch string) bool { + switch goos { + case "darwin", "freebsd", "linux", "windows": + return true + default: + return false + } +} + +// FuzzInstrumented reports whether fuzzing on goos/goarch uses coverage +// instrumentation. (FuzzInstrumented implies FuzzSupported.) +func FuzzInstrumented(goos, goarch string) bool { + switch goarch { + case "amd64", "arm64": + // TODO(#14565): support more architectures. + return FuzzSupported(goos, goarch) + default: + return false + } +} + +// MustLinkExternal reports whether goos/goarch requires external linking +// with or without cgo dependencies. +func MustLinkExternal(goos, goarch string, withCgo bool) bool { + if withCgo { + switch goarch { + case "loong64", + "mips", "mipsle", "mips64", "mips64le", + "riscv64": + // Internally linking cgo is incomplete on some architectures. + // https://go.dev/issue/14449 + return true + case "arm64": + if goos == "windows" { + // windows/arm64 internal linking is not implemented. + return true + } + case "ppc64": + // Big Endian PPC64 cgo internal linking is not implemented for aix or linux. + // https://go.dev/issue/8912 + return true + } + + switch goos { + case "android": + return true + case "dragonfly": + // It seems that on Dragonfly thread local storage is + // set up by the dynamic linker, so internal cgo linking + // doesn't work. Test case is "go test runtime/cgo". + return true + } + } + + switch goos { + case "android": + if goarch != "arm64" { + return true + } + case "ios": + if goarch == "arm64" { + return true + } + } + return false +} + +// BuildModeSupported reports whether goos/goarch supports the given build mode +// using the given compiler. +// There is a copy of this function in cmd/dist/test.go. +func BuildModeSupported(compiler, buildmode, goos, goarch string) bool { + if compiler == "gccgo" { + return true + } + + if _, ok := distInfo[OSArch{goos, goarch}]; !ok { + return false // platform unrecognized + } + + platform := goos + "/" + goarch + switch buildmode { + case "archive": + return true + + case "c-archive": + switch goos { + case "aix", "darwin", "ios", "windows": + return true + case "linux": + switch goarch { + case "386", "amd64", "arm", "armbe", "arm64", "arm64be", "loong64", "ppc64le", "riscv64", "s390x": + // linux/ppc64 not supported because it does + // not support external linking mode yet. + return true + default: + // Other targets do not support -shared, + // per ParseFlags in + // cmd/compile/internal/base/flag.go. + // For c-archive the Go tool passes -shared, + // so that the result is suitable for inclusion + // in a PIE or shared library. + return false + } + case "freebsd": + return goarch == "amd64" + } + return false + + case "c-shared": + switch platform { + case "linux/amd64", "linux/arm", "linux/arm64", "linux/loong64", "linux/386", "linux/ppc64le", "linux/riscv64", "linux/s390x", + "android/amd64", "android/arm", "android/arm64", "android/386", + "freebsd/amd64", + "darwin/amd64", "darwin/arm64", + "windows/amd64", "windows/386", "windows/arm64": + return true + } + return false + + case "default": + return true + + case "exe": + return true + + case "pie": + switch platform { + case "linux/386", "linux/amd64", "linux/arm", "linux/arm64", "linux/loong64", "linux/ppc64le", "linux/riscv64", "linux/s390x", + "android/amd64", "android/arm", "android/arm64", "android/386", + "freebsd/amd64", + "darwin/amd64", "darwin/arm64", + "ios/amd64", "ios/arm64", + "aix/ppc64", + "windows/386", "windows/amd64", "windows/arm", "windows/arm64": + return true + } + return false + + case "shared": + switch platform { + case "linux/386", "linux/amd64", "linux/arm", "linux/arm64", "linux/ppc64le", "linux/s390x": + return true + } + return false + + case "plugin": + switch platform { + case "linux/amd64", "linux/arm", "linux/arm64", "linux/386", "linux/s390x", "linux/ppc64le", + "android/amd64", "android/386", + "darwin/amd64", "darwin/arm64", + "freebsd/amd64": + return true + } + return false + + default: + return false + } +} + +func InternalLinkPIESupported(goos, goarch string) bool { + switch goos + "/" + goarch { + case "android/arm64", + "darwin/amd64", "darwin/arm64", + "linux/amd64", "linux/arm64", "linux/ppc64le", + "windows/386", "windows/amd64", "windows/arm", "windows/arm64": + return true + } + return false +} + +// DefaultPIE reports whether goos/goarch produces a PIE binary when using the +// "default" buildmode. On Windows this is affected by -race, +// so force the caller to pass that in to centralize that choice. +func DefaultPIE(goos, goarch string, isRace bool) bool { + switch goos { + case "android", "ios": + return true + case "windows": + if isRace { + // PIE is not supported with -race on windows; + // see https://go.dev/cl/416174. + return false + } + return true + case "darwin": + return goarch == "arm64" + } + return false +} + +// ExecutableHasDWARF reports whether the linked executable includes DWARF +// symbols on goos/goarch. +func ExecutableHasDWARF(goos, goarch string) bool { + switch goos { + case "plan9", "ios": + return false + } + return true +} + +// osArchInfo describes information about an OSArch extracted from cmd/dist and +// stored in the generated distInfo map. +type osArchInfo struct { + CgoSupported bool + FirstClass bool + Broken bool +} + +// CgoSupported reports whether goos/goarch supports cgo. +func CgoSupported(goos, goarch string) bool { + return distInfo[OSArch{goos, goarch}].CgoSupported +} + +// FirstClass reports whether goos/goarch is considered a “first class” port. +// (See https://go.dev/wiki/PortingPolicy#first-class-ports.) +func FirstClass(goos, goarch string) bool { + return distInfo[OSArch{goos, goarch}].FirstClass +} + +// Broken reportsr whether goos/goarch is considered a broken port. +// (See https://go.dev/wiki/PortingPolicy#broken-ports.) +func Broken(goos, goarch string) bool { + return distInfo[OSArch{goos, goarch}].Broken +} diff --git a/src/internal/platform/zosarch.go b/src/internal/platform/zosarch.go new file mode 100644 index 0000000..7f5a290 --- /dev/null +++ b/src/internal/platform/zosarch.go @@ -0,0 +1,114 @@ +// Code generated by go test internal/platform -fix. DO NOT EDIT. + +// To change the information in this file, edit the cgoEnabled and/or firstClass +// maps in cmd/dist/build.go, then run 'go generate internal/platform'. + +package platform + +// List is the list of all valid GOOS/GOARCH combinations, +// including known-broken ports. +var List = []OSArch{ + {"aix", "ppc64"}, + {"android", "386"}, + {"android", "amd64"}, + {"android", "arm"}, + {"android", "arm64"}, + {"darwin", "amd64"}, + {"darwin", "arm64"}, + {"dragonfly", "amd64"}, + {"freebsd", "386"}, + {"freebsd", "amd64"}, + {"freebsd", "arm"}, + {"freebsd", "arm64"}, + {"freebsd", "riscv64"}, + {"illumos", "amd64"}, + {"ios", "amd64"}, + {"ios", "arm64"}, + {"js", "wasm"}, + {"linux", "386"}, + {"linux", "amd64"}, + {"linux", "arm"}, + {"linux", "arm64"}, + {"linux", "loong64"}, + {"linux", "mips"}, + {"linux", "mips64"}, + {"linux", "mips64le"}, + {"linux", "mipsle"}, + {"linux", "ppc64"}, + {"linux", "ppc64le"}, + {"linux", "riscv64"}, + {"linux", "s390x"}, + {"linux", "sparc64"}, + {"netbsd", "386"}, + {"netbsd", "amd64"}, + {"netbsd", "arm"}, + {"netbsd", "arm64"}, + {"openbsd", "386"}, + {"openbsd", "amd64"}, + {"openbsd", "arm"}, + {"openbsd", "arm64"}, + {"openbsd", "mips64"}, + {"openbsd", "ppc64"}, + {"plan9", "386"}, + {"plan9", "amd64"}, + {"plan9", "arm"}, + {"solaris", "amd64"}, + {"wasip1", "wasm"}, + {"windows", "386"}, + {"windows", "amd64"}, + {"windows", "arm"}, + {"windows", "arm64"}, +} + +var distInfo = map[OSArch]osArchInfo{ + {"aix", "ppc64"}: {CgoSupported: true}, + {"android", "386"}: {CgoSupported: true}, + {"android", "amd64"}: {CgoSupported: true}, + {"android", "arm"}: {CgoSupported: true}, + {"android", "arm64"}: {CgoSupported: true}, + {"darwin", "amd64"}: {CgoSupported: true, FirstClass: true}, + {"darwin", "arm64"}: {CgoSupported: true, FirstClass: true}, + {"dragonfly", "amd64"}: {CgoSupported: true}, + {"freebsd", "386"}: {CgoSupported: true}, + {"freebsd", "amd64"}: {CgoSupported: true}, + {"freebsd", "arm"}: {CgoSupported: true}, + {"freebsd", "arm64"}: {CgoSupported: true}, + {"freebsd", "riscv64"}: {CgoSupported: true}, + {"illumos", "amd64"}: {CgoSupported: true}, + {"ios", "amd64"}: {CgoSupported: true}, + {"ios", "arm64"}: {CgoSupported: true}, + {"js", "wasm"}: {}, + {"linux", "386"}: {CgoSupported: true, FirstClass: true}, + {"linux", "amd64"}: {CgoSupported: true, FirstClass: true}, + {"linux", "arm"}: {CgoSupported: true, FirstClass: true}, + {"linux", "arm64"}: {CgoSupported: true, FirstClass: true}, + {"linux", "loong64"}: {CgoSupported: true}, + {"linux", "mips"}: {CgoSupported: true}, + {"linux", "mips64"}: {CgoSupported: true}, + {"linux", "mips64le"}: {CgoSupported: true}, + {"linux", "mipsle"}: {CgoSupported: true}, + {"linux", "ppc64"}: {}, + {"linux", "ppc64le"}: {CgoSupported: true}, + {"linux", "riscv64"}: {CgoSupported: true}, + {"linux", "s390x"}: {CgoSupported: true}, + {"linux", "sparc64"}: {CgoSupported: true, Broken: true}, + {"netbsd", "386"}: {CgoSupported: true}, + {"netbsd", "amd64"}: {CgoSupported: true}, + {"netbsd", "arm"}: {CgoSupported: true}, + {"netbsd", "arm64"}: {CgoSupported: true}, + {"openbsd", "386"}: {CgoSupported: true}, + {"openbsd", "amd64"}: {CgoSupported: true}, + {"openbsd", "arm"}: {CgoSupported: true}, + {"openbsd", "arm64"}: {CgoSupported: true}, + {"openbsd", "mips64"}: {CgoSupported: true, Broken: true}, + {"openbsd", "ppc64"}: {Broken: true}, + {"plan9", "386"}: {}, + {"plan9", "amd64"}: {}, + {"plan9", "arm"}: {}, + {"solaris", "amd64"}: {CgoSupported: true}, + {"wasip1", "wasm"}: {}, + {"windows", "386"}: {CgoSupported: true, FirstClass: true}, + {"windows", "amd64"}: {CgoSupported: true, FirstClass: true}, + {"windows", "arm"}: {}, + {"windows", "arm64"}: {CgoSupported: true}, +} diff --git a/src/internal/platform/zosarch_test.go b/src/internal/platform/zosarch_test.go new file mode 100644 index 0000000..e8ffe9e --- /dev/null +++ b/src/internal/platform/zosarch_test.go @@ -0,0 +1,109 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package platform_test + +import ( + "bytes" + "encoding/json" + "flag" + "internal/diff" + "internal/testenv" + "os" + "os/exec" + "testing" + "text/template" +) + +var flagFix = flag.Bool("fix", false, "if true, fix out-of-date generated files") + +// TestGenerated verifies that zosarch.go is up to date, +// or regenerates it if the -fix flag is set. +func TestGenerated(t *testing.T) { + testenv.MustHaveGoRun(t) + + // Here we use 'go run cmd/dist' instead of 'go tool dist' in case the + // installed cmd/dist is stale or missing. We don't want to miss a + // skew in the data due to a stale binary. + cmd := testenv.Command(t, "go", "run", "cmd/dist", "list", "-json", "-broken") + + // cmd/dist requires GOROOT to be set explicitly in the environment. + cmd.Env = append(cmd.Environ(), "GOROOT="+testenv.GOROOT(t)) + + out, err := cmd.Output() + if err != nil { + if ee, ok := err.(*exec.ExitError); ok && len(ee.Stderr) > 0 { + t.Logf("stderr:\n%s", ee.Stderr) + } + t.Fatalf("%v: %v", cmd, err) + } + + type listEntry struct { + GOOS, GOARCH string + CgoSupported bool + FirstClass bool + Broken bool + } + var entries []listEntry + if err := json.Unmarshal(out, &entries); err != nil { + t.Fatal(err) + } + + tmplOut := new(bytes.Buffer) + tmpl := template.Must(template.New("zosarch").Parse(zosarchTmpl)) + err = tmpl.Execute(tmplOut, entries) + if err != nil { + t.Fatal(err) + } + + cmd = testenv.Command(t, "gofmt") + cmd.Stdin = bytes.NewReader(tmplOut.Bytes()) + want, err := cmd.Output() + if err != nil { + t.Logf("stdin:\n%s", tmplOut.Bytes()) + if ee, ok := err.(*exec.ExitError); ok && len(ee.Stderr) > 0 { + t.Logf("stderr:\n%s", ee.Stderr) + } + t.Fatalf("%v: %v", cmd, err) + } + + got, err := os.ReadFile("zosarch.go") + if err == nil && bytes.Equal(got, want) { + return + } + + if !*flagFix { + if err != nil { + t.Log(err) + } else { + t.Logf("diff:\n%s", diff.Diff("zosarch.go", got, "want", want)) + } + t.Fatalf("zosarch.go is missing or out of date; to regenerate, run\ngo generate internal/platform") + } + + if err := os.WriteFile("zosarch.go", want, 0666); err != nil { + t.Fatal(err) + } +} + +const zosarchTmpl = `// Code generated by go test internal/platform -fix. DO NOT EDIT. + +// To change the information in this file, edit the cgoEnabled and/or firstClass +// maps in cmd/dist/build.go, then run 'go generate internal/platform'. + +package platform + +// List is the list of all valid GOOS/GOARCH combinations, +// including known-broken ports. +var List = []OSArch{ +{{range .}} { {{ printf "%q" .GOOS }}, {{ printf "%q" .GOARCH }} }, +{{end}} +} + +var distInfo = map[OSArch]osArchInfo { +{{range .}} { {{ printf "%q" .GOOS }}, {{ printf "%q" .GOARCH }} }: +{ {{if .CgoSupported}}CgoSupported: true, {{end}}{{if .FirstClass}}FirstClass: true, {{end}}{{if .Broken}} Broken: true, {{end}} }, +{{end}} +} +` diff --git a/src/internal/poll/copy_file_range_linux.go b/src/internal/poll/copy_file_range_linux.go new file mode 100644 index 0000000..ba33f51 --- /dev/null +++ b/src/internal/poll/copy_file_range_linux.go @@ -0,0 +1,128 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "internal/syscall/unix" + "sync" + "syscall" +) + +var ( + kernelVersion53Once sync.Once + kernelVersion53 bool +) + +const maxCopyFileRangeRound = 1 << 30 + +// CopyFileRange copies at most remain bytes of data from src to dst, using +// the copy_file_range system call. dst and src must refer to regular files. +func CopyFileRange(dst, src *FD, remain int64) (written int64, handled bool, err error) { + kernelVersion53Once.Do(func() { + major, minor := unix.KernelVersion() + // copy_file_range(2) is broken in various ways on kernels older than 5.3, + // see issue #42400 and + // https://man7.org/linux/man-pages/man2/copy_file_range.2.html#VERSIONS + if major > 5 || (major == 5 && minor >= 3) { + kernelVersion53 = true + } + }) + + if !kernelVersion53 { + return 0, false, nil + } + + for remain > 0 { + max := remain + if max > maxCopyFileRangeRound { + max = maxCopyFileRangeRound + } + n, err := copyFileRange(dst, src, int(max)) + switch err { + case syscall.ENOSYS: + // copy_file_range(2) was introduced in Linux 4.5. + // Go supports Linux >= 2.6.33, so the system call + // may not be present. + // + // If we see ENOSYS, we have certainly not transferred + // any data, so we can tell the caller that we + // couldn't handle the transfer and let them fall + // back to more generic code. + return 0, false, nil + case syscall.EXDEV, syscall.EINVAL, syscall.EIO, syscall.EOPNOTSUPP, syscall.EPERM: + // Prior to Linux 5.3, it was not possible to + // copy_file_range across file systems. Similarly to + // the ENOSYS case above, if we see EXDEV, we have + // not transferred any data, and we can let the caller + // fall back to generic code. + // + // As for EINVAL, that is what we see if, for example, + // dst or src refer to a pipe rather than a regular + // file. This is another case where no data has been + // transferred, so we consider it unhandled. + // + // If src and dst are on CIFS, we can see EIO. + // See issue #42334. + // + // If the file is on NFS, we can see EOPNOTSUPP. + // See issue #40731. + // + // If the process is running inside a Docker container, + // we might see EPERM instead of ENOSYS. See issue + // #40893. Since EPERM might also be a legitimate error, + // don't mark copy_file_range(2) as unsupported. + return 0, false, nil + case nil: + if n == 0 { + // If we did not read any bytes at all, + // then this file may be in a file system + // where copy_file_range silently fails. + // https://lore.kernel.org/linux-fsdevel/20210126233840.GG4626@dread.disaster.area/T/#m05753578c7f7882f6e9ffe01f981bc223edef2b0 + if written == 0 { + return 0, false, nil + } + // Otherwise src is at EOF, which means + // we are done. + return written, true, nil + } + remain -= n + written += n + default: + return written, true, err + } + } + return written, true, nil +} + +// copyFileRange performs one round of copy_file_range(2). +func copyFileRange(dst, src *FD, max int) (written int64, err error) { + // The signature of copy_file_range(2) is: + // + // ssize_t copy_file_range(int fd_in, loff_t *off_in, + // int fd_out, loff_t *off_out, + // size_t len, unsigned int flags); + // + // Note that in the call to unix.CopyFileRange below, we use nil + // values for off_in and off_out. For the system call, this means + // "use and update the file offsets". That is why we must acquire + // locks for both file descriptors (and why this whole machinery is + // in the internal/poll package to begin with). + if err := dst.writeLock(); err != nil { + return 0, err + } + defer dst.writeUnlock() + if err := src.readLock(); err != nil { + return 0, err + } + defer src.readUnlock() + var n int + for { + n, err = unix.CopyFileRange(src.Sysfd, nil, dst.Sysfd, nil, max, 0) + if err != syscall.EINTR { + break + } + } + return int64(n), err +} diff --git a/src/internal/poll/errno_unix.go b/src/internal/poll/errno_unix.go new file mode 100644 index 0000000..d1a18ab --- /dev/null +++ b/src/internal/poll/errno_unix.go @@ -0,0 +1,33 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || wasip1 + +package poll + +import "syscall" + +// Do the interface allocations only once for common +// Errno values. +var ( + errEAGAIN error = syscall.EAGAIN + errEINVAL error = syscall.EINVAL + errENOENT error = syscall.ENOENT +) + +// errnoErr returns common boxed Errno values, to prevent +// allocations at runtime. +func errnoErr(e syscall.Errno) error { + switch e { + case 0: + return nil + case syscall.EAGAIN: + return errEAGAIN + case syscall.EINVAL: + return errEINVAL + case syscall.ENOENT: + return errENOENT + } + return e +} diff --git a/src/internal/poll/errno_windows.go b/src/internal/poll/errno_windows.go new file mode 100644 index 0000000..6381479 --- /dev/null +++ b/src/internal/poll/errno_windows.go @@ -0,0 +1,31 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package poll + +import "syscall" + +// Do the interface allocations only once for common +// Errno values. + +var ( + errERROR_IO_PENDING error = syscall.Errno(syscall.ERROR_IO_PENDING) +) + +// errnoErr returns common boxed Errno values, to prevent +// allocations at runtime. +func errnoErr(e syscall.Errno) error { + switch e { + case 0: + return nil + case syscall.ERROR_IO_PENDING: + return errERROR_IO_PENDING + } + // TODO: add more here, after collecting data on the common + // error values see on Windows. (perhaps when running + // all.bat?) + return e +} diff --git a/src/internal/poll/error_linux_test.go b/src/internal/poll/error_linux_test.go new file mode 100644 index 0000000..059fb8e --- /dev/null +++ b/src/internal/poll/error_linux_test.go @@ -0,0 +1,31 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "errors" + "internal/poll" + "os" + "syscall" +) + +func badStateFile() (*os.File, error) { + if os.Getuid() != 0 { + return nil, errors.New("must be root") + } + // Using OpenFile for a device file is an easy way to make a + // file attached to the runtime-integrated network poller and + // configured in halfway. + return os.OpenFile("/dev/net/tun", os.O_RDWR, 0) +} + +func isBadStateFileError(err error) (string, bool) { + switch err { + case poll.ErrNotPollable, syscall.EBADFD: + return "", true + default: + return "not pollable or file in bad state error", false + } +} diff --git a/src/internal/poll/error_stub_test.go b/src/internal/poll/error_stub_test.go new file mode 100644 index 0000000..48e0952 --- /dev/null +++ b/src/internal/poll/error_stub_test.go @@ -0,0 +1,21 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !linux + +package poll_test + +import ( + "errors" + "os" + "runtime" +) + +func badStateFile() (*os.File, error) { + return nil, errors.New("not supported on " + runtime.GOOS) +} + +func isBadStateFileError(err error) (string, bool) { + return "", false +} diff --git a/src/internal/poll/error_test.go b/src/internal/poll/error_test.go new file mode 100644 index 0000000..abc8b16 --- /dev/null +++ b/src/internal/poll/error_test.go @@ -0,0 +1,51 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "fmt" + "io/fs" + "net" + "os" + "testing" + "time" +) + +func TestReadError(t *testing.T) { + t.Run("ErrNotPollable", func(t *testing.T) { + f, err := badStateFile() + if err != nil { + t.Skip(err) + } + defer f.Close() + + // Give scheduler a chance to have two separated + // goroutines: an event poller and an event waiter. + time.Sleep(100 * time.Millisecond) + + var b [1]byte + _, err = f.Read(b[:]) + if perr := parseReadError(err, isBadStateFileError); perr != nil { + t.Fatal(perr) + } + }) +} + +func parseReadError(nestedErr error, verify func(error) (string, bool)) error { + err := nestedErr + if nerr, ok := err.(*net.OpError); ok { + err = nerr.Err + } + if nerr, ok := err.(*fs.PathError); ok { + err = nerr.Err + } + if nerr, ok := err.(*os.SyscallError); ok { + err = nerr.Err + } + if s, ok := verify(err); !ok { + return fmt.Errorf("got %v; want %s", nestedErr, s) + } + return nil +} diff --git a/src/internal/poll/export_linux_test.go b/src/internal/poll/export_linux_test.go new file mode 100644 index 0000000..7fba793 --- /dev/null +++ b/src/internal/poll/export_linux_test.go @@ -0,0 +1,22 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Export guts for testing on linux. +// Since testing imports os and os imports internal/poll, +// the internal/poll tests can not be in package poll. + +package poll + +var ( + GetPipe = getPipe + PutPipe = putPipe + NewPipe = newPipe + DestroyPipe = destroyPipe +) + +func GetPipeFds(p *SplicePipe) (int, int) { + return p.rfd, p.wfd +} + +type SplicePipe = splicePipe diff --git a/src/internal/poll/export_posix_test.go b/src/internal/poll/export_posix_test.go new file mode 100644 index 0000000..3415ab3 --- /dev/null +++ b/src/internal/poll/export_posix_test.go @@ -0,0 +1,15 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || windows + +// Export guts for testing on posix. +// Since testing imports os and os imports internal/poll, +// the internal/poll tests can not be in package poll. + +package poll + +func (fd *FD) EOFError(n int, err error) error { + return fd.eofError(n, err) +} diff --git a/src/internal/poll/export_test.go b/src/internal/poll/export_test.go new file mode 100644 index 0000000..66d7c32 --- /dev/null +++ b/src/internal/poll/export_test.go @@ -0,0 +1,35 @@ +// Copyright 2010 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Export guts for testing. +// Since testing imports os and os imports internal/poll, +// the internal/poll tests can not be in package poll. + +package poll + +var Consume = consume + +type XFDMutex struct { + fdMutex +} + +func (mu *XFDMutex) Incref() bool { + return mu.incref() +} + +func (mu *XFDMutex) IncrefAndClose() bool { + return mu.increfAndClose() +} + +func (mu *XFDMutex) Decref() bool { + return mu.decref() +} + +func (mu *XFDMutex) RWLock(read bool) bool { + return mu.rwlock(read) +} + +func (mu *XFDMutex) RWUnlock(read bool) bool { + return mu.rwunlock(read) +} diff --git a/src/internal/poll/export_windows_test.go b/src/internal/poll/export_windows_test.go new file mode 100644 index 0000000..88ed71a --- /dev/null +++ b/src/internal/poll/export_windows_test.go @@ -0,0 +1,17 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Export guts for testing on windows. +// Since testing imports os and os imports internal/poll, +// the internal/poll tests can not be in package poll. + +package poll + +var ( + LogInitFD = &logInitFD +) + +func (fd *FD) IsPartOfNetpoll() bool { + return fd.pd.runtimeCtx != 0 +} diff --git a/src/internal/poll/fd.go b/src/internal/poll/fd.go new file mode 100644 index 0000000..ef61d0c --- /dev/null +++ b/src/internal/poll/fd.go @@ -0,0 +1,83 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package poll supports non-blocking I/O on file descriptors with polling. +// This supports I/O operations that block only a goroutine, not a thread. +// This is used by the net and os packages. +// It uses a poller built into the runtime, with support from the +// runtime scheduler. +package poll + +import ( + "errors" +) + +// errNetClosing is the type of the variable ErrNetClosing. +// This is used to implement the net.Error interface. +type errNetClosing struct{} + +// Error returns the error message for ErrNetClosing. +// Keep this string consistent because of issue #4373: +// since historically programs have not been able to detect +// this error, they look for the string. +func (e errNetClosing) Error() string { return "use of closed network connection" } + +func (e errNetClosing) Timeout() bool { return false } +func (e errNetClosing) Temporary() bool { return false } + +// ErrNetClosing is returned when a network descriptor is used after +// it has been closed. +var ErrNetClosing = errNetClosing{} + +// ErrFileClosing is returned when a file descriptor is used after it +// has been closed. +var ErrFileClosing = errors.New("use of closed file") + +// ErrNoDeadline is returned when a request is made to set a deadline +// on a file type that does not use the poller. +var ErrNoDeadline = errors.New("file type does not support deadline") + +// Return the appropriate closing error based on isFile. +func errClosing(isFile bool) error { + if isFile { + return ErrFileClosing + } + return ErrNetClosing +} + +// ErrDeadlineExceeded is returned for an expired deadline. +// This is exported by the os package as os.ErrDeadlineExceeded. +var ErrDeadlineExceeded error = &DeadlineExceededError{} + +// DeadlineExceededError is returned for an expired deadline. +type DeadlineExceededError struct{} + +// Implement the net.Error interface. +// The string is "i/o timeout" because that is what was returned +// by earlier Go versions. Changing it may break programs that +// match on error strings. +func (e *DeadlineExceededError) Error() string { return "i/o timeout" } +func (e *DeadlineExceededError) Timeout() bool { return true } +func (e *DeadlineExceededError) Temporary() bool { return true } + +// ErrNotPollable is returned when the file or socket is not suitable +// for event notification. +var ErrNotPollable = errors.New("not pollable") + +// consume removes data from a slice of byte slices, for writev. +func consume(v *[][]byte, n int64) { + for len(*v) > 0 { + ln0 := int64(len((*v)[0])) + if ln0 > n { + (*v)[0] = (*v)[0][n:] + return + } + n -= ln0 + (*v)[0] = nil + *v = (*v)[1:] + } +} + +// TestHookDidWritev is a hook for testing writev. +var TestHookDidWritev = func(wrote int) {} diff --git a/src/internal/poll/fd_fsync_darwin.go b/src/internal/poll/fd_fsync_darwin.go new file mode 100644 index 0000000..731b7fd --- /dev/null +++ b/src/internal/poll/fd_fsync_darwin.go @@ -0,0 +1,24 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "internal/syscall/unix" + "syscall" +) + +// Fsync invokes SYS_FCNTL with SYS_FULLFSYNC because +// on OS X, SYS_FSYNC doesn't fully flush contents to disk. +// See Issue #26650 as well as the man page for fsync on OS X. +func (fd *FD) Fsync() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + _, err := unix.Fcntl(fd.Sysfd, syscall.F_FULLFSYNC, 0) + return err + }) +} diff --git a/src/internal/poll/fd_fsync_posix.go b/src/internal/poll/fd_fsync_posix.go new file mode 100644 index 0000000..469ca75 --- /dev/null +++ b/src/internal/poll/fd_fsync_posix.go @@ -0,0 +1,20 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build aix || dragonfly || freebsd || (js && wasm) || linux || netbsd || openbsd || solaris || wasip1 + +package poll + +import "syscall" + +// Fsync wraps syscall.Fsync. +func (fd *FD) Fsync() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + return syscall.Fsync(fd.Sysfd) + }) +} diff --git a/src/internal/poll/fd_fsync_windows.go b/src/internal/poll/fd_fsync_windows.go new file mode 100644 index 0000000..fb12119 --- /dev/null +++ b/src/internal/poll/fd_fsync_windows.go @@ -0,0 +1,16 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// Fsync wraps syscall.Fsync. +func (fd *FD) Fsync() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Fsync(fd.Sysfd) +} diff --git a/src/internal/poll/fd_io_plan9.go b/src/internal/poll/fd_io_plan9.go new file mode 100644 index 0000000..3205ac8 --- /dev/null +++ b/src/internal/poll/fd_io_plan9.go @@ -0,0 +1,92 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "internal/itoa" + "runtime" + "sync" + "syscall" +) + +// asyncIO implements asynchronous cancelable I/O. +// An asyncIO represents a single asynchronous Read or Write +// operation. The result is returned on the result channel. +// The undergoing I/O system call can either complete or be +// interrupted by a note. +type asyncIO struct { + res chan result + + // mu guards the pid field. + mu sync.Mutex + + // pid holds the process id of + // the process running the IO operation. + pid int +} + +// result is the return value of a Read or Write operation. +type result struct { + n int + err error +} + +// newAsyncIO returns a new asyncIO that performs an I/O +// operation by calling fn, which must do one and only one +// interruptible system call. +func newAsyncIO(fn func([]byte) (int, error), b []byte) *asyncIO { + aio := &asyncIO{ + res: make(chan result, 0), + } + aio.mu.Lock() + go func() { + // Lock the current goroutine to its process + // and store the pid in io so that Cancel can + // interrupt it. We ignore the "hangup" signal, + // so the signal does not take down the entire + // Go runtime. + runtime.LockOSThread() + runtime_ignoreHangup() + aio.pid = syscall.Getpid() + aio.mu.Unlock() + + n, err := fn(b) + + aio.mu.Lock() + aio.pid = -1 + runtime_unignoreHangup() + aio.mu.Unlock() + + aio.res <- result{n, err} + }() + return aio +} + +// Cancel interrupts the I/O operation, causing +// the Wait function to return. +func (aio *asyncIO) Cancel() { + aio.mu.Lock() + defer aio.mu.Unlock() + if aio.pid == -1 { + return + } + f, e := syscall.Open("/proc/"+itoa.Itoa(aio.pid)+"/note", syscall.O_WRONLY) + if e != nil { + return + } + syscall.Write(f, []byte("hangup")) + syscall.Close(f) +} + +// Wait for the I/O operation to complete. +func (aio *asyncIO) Wait() (int, error) { + res := <-aio.res + return res.n, res.err +} + +// The following functions, provided by the runtime, are used to +// ignore and unignore the "hangup" signal received by the process. +func runtime_ignoreHangup() +func runtime_unignoreHangup() diff --git a/src/internal/poll/fd_mutex.go b/src/internal/poll/fd_mutex.go new file mode 100644 index 0000000..0a8ee6f --- /dev/null +++ b/src/internal/poll/fd_mutex.go @@ -0,0 +1,252 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "sync/atomic" + +// fdMutex is a specialized synchronization primitive that manages +// lifetime of an fd and serializes access to Read, Write and Close +// methods on FD. +type fdMutex struct { + state uint64 + rsema uint32 + wsema uint32 +} + +// fdMutex.state is organized as follows: +// 1 bit - whether FD is closed, if set all subsequent lock operations will fail. +// 1 bit - lock for read operations. +// 1 bit - lock for write operations. +// 20 bits - total number of references (read+write+misc). +// 20 bits - number of outstanding read waiters. +// 20 bits - number of outstanding write waiters. +const ( + mutexClosed = 1 << 0 + mutexRLock = 1 << 1 + mutexWLock = 1 << 2 + mutexRef = 1 << 3 + mutexRefMask = (1<<20 - 1) << 3 + mutexRWait = 1 << 23 + mutexRMask = (1<<20 - 1) << 23 + mutexWWait = 1 << 43 + mutexWMask = (1<<20 - 1) << 43 +) + +const overflowMsg = "too many concurrent operations on a single file or socket (max 1048575)" + +// Read operations must do rwlock(true)/rwunlock(true). +// +// Write operations must do rwlock(false)/rwunlock(false). +// +// Misc operations must do incref/decref. +// Misc operations include functions like setsockopt and setDeadline. +// They need to use incref/decref to ensure that they operate on the +// correct fd in presence of a concurrent close call (otherwise fd can +// be closed under their feet). +// +// Close operations must do increfAndClose/decref. + +// incref adds a reference to mu. +// It reports whether mu is available for reading or writing. +func (mu *fdMutex) incref() bool { + for { + old := atomic.LoadUint64(&mu.state) + if old&mutexClosed != 0 { + return false + } + new := old + mutexRef + if new&mutexRefMask == 0 { + panic(overflowMsg) + } + if atomic.CompareAndSwapUint64(&mu.state, old, new) { + return true + } + } +} + +// increfAndClose sets the state of mu to closed. +// It returns false if the file was already closed. +func (mu *fdMutex) increfAndClose() bool { + for { + old := atomic.LoadUint64(&mu.state) + if old&mutexClosed != 0 { + return false + } + // Mark as closed and acquire a reference. + new := (old | mutexClosed) + mutexRef + if new&mutexRefMask == 0 { + panic(overflowMsg) + } + // Remove all read and write waiters. + new &^= mutexRMask | mutexWMask + if atomic.CompareAndSwapUint64(&mu.state, old, new) { + // Wake all read and write waiters, + // they will observe closed flag after wakeup. + for old&mutexRMask != 0 { + old -= mutexRWait + runtime_Semrelease(&mu.rsema) + } + for old&mutexWMask != 0 { + old -= mutexWWait + runtime_Semrelease(&mu.wsema) + } + return true + } + } +} + +// decref removes a reference from mu. +// It reports whether there is no remaining reference. +func (mu *fdMutex) decref() bool { + for { + old := atomic.LoadUint64(&mu.state) + if old&mutexRefMask == 0 { + panic("inconsistent poll.fdMutex") + } + new := old - mutexRef + if atomic.CompareAndSwapUint64(&mu.state, old, new) { + return new&(mutexClosed|mutexRefMask) == mutexClosed + } + } +} + +// lock adds a reference to mu and locks mu. +// It reports whether mu is available for reading or writing. +func (mu *fdMutex) rwlock(read bool) bool { + var mutexBit, mutexWait, mutexMask uint64 + var mutexSema *uint32 + if read { + mutexBit = mutexRLock + mutexWait = mutexRWait + mutexMask = mutexRMask + mutexSema = &mu.rsema + } else { + mutexBit = mutexWLock + mutexWait = mutexWWait + mutexMask = mutexWMask + mutexSema = &mu.wsema + } + for { + old := atomic.LoadUint64(&mu.state) + if old&mutexClosed != 0 { + return false + } + var new uint64 + if old&mutexBit == 0 { + // Lock is free, acquire it. + new = (old | mutexBit) + mutexRef + if new&mutexRefMask == 0 { + panic(overflowMsg) + } + } else { + // Wait for lock. + new = old + mutexWait + if new&mutexMask == 0 { + panic(overflowMsg) + } + } + if atomic.CompareAndSwapUint64(&mu.state, old, new) { + if old&mutexBit == 0 { + return true + } + runtime_Semacquire(mutexSema) + // The signaller has subtracted mutexWait. + } + } +} + +// unlock removes a reference from mu and unlocks mu. +// It reports whether there is no remaining reference. +func (mu *fdMutex) rwunlock(read bool) bool { + var mutexBit, mutexWait, mutexMask uint64 + var mutexSema *uint32 + if read { + mutexBit = mutexRLock + mutexWait = mutexRWait + mutexMask = mutexRMask + mutexSema = &mu.rsema + } else { + mutexBit = mutexWLock + mutexWait = mutexWWait + mutexMask = mutexWMask + mutexSema = &mu.wsema + } + for { + old := atomic.LoadUint64(&mu.state) + if old&mutexBit == 0 || old&mutexRefMask == 0 { + panic("inconsistent poll.fdMutex") + } + // Drop lock, drop reference and wake read waiter if present. + new := (old &^ mutexBit) - mutexRef + if old&mutexMask != 0 { + new -= mutexWait + } + if atomic.CompareAndSwapUint64(&mu.state, old, new) { + if old&mutexMask != 0 { + runtime_Semrelease(mutexSema) + } + return new&(mutexClosed|mutexRefMask) == mutexClosed + } + } +} + +// Implemented in runtime package. +func runtime_Semacquire(sema *uint32) +func runtime_Semrelease(sema *uint32) + +// incref adds a reference to fd. +// It returns an error when fd cannot be used. +func (fd *FD) incref() error { + if !fd.fdmu.incref() { + return errClosing(fd.isFile) + } + return nil +} + +// decref removes a reference from fd. +// It also closes fd when the state of fd is set to closed and there +// is no remaining reference. +func (fd *FD) decref() error { + if fd.fdmu.decref() { + return fd.destroy() + } + return nil +} + +// readLock adds a reference to fd and locks fd for reading. +// It returns an error when fd cannot be used for reading. +func (fd *FD) readLock() error { + if !fd.fdmu.rwlock(true) { + return errClosing(fd.isFile) + } + return nil +} + +// readUnlock removes a reference from fd and unlocks fd for reading. +// It also closes fd when the state of fd is set to closed and there +// is no remaining reference. +func (fd *FD) readUnlock() { + if fd.fdmu.rwunlock(true) { + fd.destroy() + } +} + +// writeLock adds a reference to fd and locks fd for writing. +// It returns an error when fd cannot be used for writing. +func (fd *FD) writeLock() error { + if !fd.fdmu.rwlock(false) { + return errClosing(fd.isFile) + } + return nil +} + +// writeUnlock removes a reference from fd and unlocks fd for writing. +// It also closes fd when the state of fd is set to closed and there +// is no remaining reference. +func (fd *FD) writeUnlock() { + if fd.fdmu.rwunlock(false) { + fd.destroy() + } +} diff --git a/src/internal/poll/fd_mutex_test.go b/src/internal/poll/fd_mutex_test.go new file mode 100644 index 0000000..62f9531 --- /dev/null +++ b/src/internal/poll/fd_mutex_test.go @@ -0,0 +1,222 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + . "internal/poll" + "math/rand" + "runtime" + "strings" + "testing" + "time" +) + +func TestMutexLock(t *testing.T) { + var mu XFDMutex + + if !mu.Incref() { + t.Fatal("broken") + } + if mu.Decref() { + t.Fatal("broken") + } + + if !mu.RWLock(true) { + t.Fatal("broken") + } + if mu.RWUnlock(true) { + t.Fatal("broken") + } + + if !mu.RWLock(false) { + t.Fatal("broken") + } + if mu.RWUnlock(false) { + t.Fatal("broken") + } +} + +func TestMutexClose(t *testing.T) { + var mu XFDMutex + if !mu.IncrefAndClose() { + t.Fatal("broken") + } + + if mu.Incref() { + t.Fatal("broken") + } + if mu.RWLock(true) { + t.Fatal("broken") + } + if mu.RWLock(false) { + t.Fatal("broken") + } + if mu.IncrefAndClose() { + t.Fatal("broken") + } +} + +func TestMutexCloseUnblock(t *testing.T) { + c := make(chan bool, 4) + var mu XFDMutex + mu.RWLock(true) + for i := 0; i < 4; i++ { + go func() { + if mu.RWLock(true) { + t.Error("broken") + return + } + c <- true + }() + } + // Concurrent goroutines must not be able to read lock the mutex. + time.Sleep(time.Millisecond) + select { + case <-c: + t.Fatal("broken") + default: + } + mu.IncrefAndClose() // Must unblock the readers. + for i := 0; i < 4; i++ { + select { + case <-c: + case <-time.After(10 * time.Second): + t.Fatal("broken") + } + } + if mu.Decref() { + t.Fatal("broken") + } + if !mu.RWUnlock(true) { + t.Fatal("broken") + } +} + +func TestMutexPanic(t *testing.T) { + ensurePanics := func(f func()) { + defer func() { + if recover() == nil { + t.Fatal("does not panic") + } + }() + f() + } + + var mu XFDMutex + ensurePanics(func() { mu.Decref() }) + ensurePanics(func() { mu.RWUnlock(true) }) + ensurePanics(func() { mu.RWUnlock(false) }) + + ensurePanics(func() { mu.Incref(); mu.Decref(); mu.Decref() }) + ensurePanics(func() { mu.RWLock(true); mu.RWUnlock(true); mu.RWUnlock(true) }) + ensurePanics(func() { mu.RWLock(false); mu.RWUnlock(false); mu.RWUnlock(false) }) + + // ensure that it's still not broken + mu.Incref() + mu.Decref() + mu.RWLock(true) + mu.RWUnlock(true) + mu.RWLock(false) + mu.RWUnlock(false) +} + +func TestMutexOverflowPanic(t *testing.T) { + defer func() { + r := recover() + if r == nil { + t.Fatal("did not panic") + } + msg, ok := r.(string) + if !ok { + t.Fatalf("unexpected panic type %T", r) + } + if !strings.Contains(msg, "too many") || strings.Contains(msg, "inconsistent") { + t.Fatalf("wrong panic message %q", msg) + } + }() + + var mu1 XFDMutex + for i := 0; i < 1<<21; i++ { + mu1.Incref() + } +} + +func TestMutexStress(t *testing.T) { + P := 8 + N := int(1e6) + if testing.Short() { + P = 4 + N = 1e4 + } + defer runtime.GOMAXPROCS(runtime.GOMAXPROCS(P)) + done := make(chan bool, P) + var mu XFDMutex + var readState [2]uint64 + var writeState [2]uint64 + for p := 0; p < P; p++ { + go func() { + defer func() { + done <- !t.Failed() + }() + r := rand.New(rand.NewSource(rand.Int63())) + for i := 0; i < N; i++ { + switch r.Intn(3) { + case 0: + if !mu.Incref() { + t.Error("broken") + return + } + if mu.Decref() { + t.Error("broken") + return + } + case 1: + if !mu.RWLock(true) { + t.Error("broken") + return + } + // Ensure that it provides mutual exclusion for readers. + if readState[0] != readState[1] { + t.Error("broken") + return + } + readState[0]++ + readState[1]++ + if mu.RWUnlock(true) { + t.Error("broken") + return + } + case 2: + if !mu.RWLock(false) { + t.Error("broken") + return + } + // Ensure that it provides mutual exclusion for writers. + if writeState[0] != writeState[1] { + t.Error("broken") + return + } + writeState[0]++ + writeState[1]++ + if mu.RWUnlock(false) { + t.Error("broken") + return + } + } + } + }() + } + for p := 0; p < P; p++ { + if !<-done { + t.FailNow() + } + } + if !mu.IncrefAndClose() { + t.Fatal("broken") + } + if !mu.Decref() { + t.Fatal("broken") + } +} diff --git a/src/internal/poll/fd_opendir_darwin.go b/src/internal/poll/fd_opendir_darwin.go new file mode 100644 index 0000000..3ae2dc8 --- /dev/null +++ b/src/internal/poll/fd_opendir_darwin.go @@ -0,0 +1,39 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "syscall" + _ "unsafe" // for go:linkname +) + +// OpenDir returns a pointer to a DIR structure suitable for +// ReadDir. In case of an error, the name of the failed +// syscall is returned along with a syscall.Errno. +func (fd *FD) OpenDir() (uintptr, string, error) { + // fdopendir(3) takes control of the file descriptor, + // so use a dup. + fd2, call, err := fd.Dup() + if err != nil { + return 0, call, err + } + var dir uintptr + for { + dir, err = fdopendir(fd2) + if err != syscall.EINTR { + break + } + } + if err != nil { + syscall.Close(fd2) + return 0, "fdopendir", err + } + return dir, "", nil +} + +// Implemented in syscall/syscall_darwin.go. +// +//go:linkname fdopendir syscall.fdopendir +func fdopendir(fd int) (dir uintptr, err error) diff --git a/src/internal/poll/fd_plan9.go b/src/internal/poll/fd_plan9.go new file mode 100644 index 0000000..7cc178a --- /dev/null +++ b/src/internal/poll/fd_plan9.go @@ -0,0 +1,232 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "errors" + "io" + "sync" + "time" +) + +type FD struct { + // Lock sysfd and serialize access to Read and Write methods. + fdmu fdMutex + + Destroy func() + + // deadlines + rmu sync.Mutex + wmu sync.Mutex + raio *asyncIO + waio *asyncIO + rtimer *time.Timer + wtimer *time.Timer + rtimedout bool // set true when read deadline has been reached + wtimedout bool // set true when write deadline has been reached + + // Whether this is a normal file. + // On Plan 9 we do not use this package for ordinary files, + // so this is always false, but the field is present because + // shared code in fd_mutex.go checks it. + isFile bool +} + +// We need this to close out a file descriptor when it is unlocked, +// but the real implementation has to live in the net package because +// it uses os.File's. +func (fd *FD) destroy() error { + if fd.Destroy != nil { + fd.Destroy() + } + return nil +} + +// Close handles the locking for closing an FD. The real operation +// is in the net package. +func (fd *FD) Close() error { + if !fd.fdmu.increfAndClose() { + return errClosing(fd.isFile) + } + return nil +} + +// Read implements io.Reader. +func (fd *FD) Read(fn func([]byte) (int, error), b []byte) (int, error) { + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + if len(b) == 0 { + return 0, nil + } + fd.rmu.Lock() + if fd.rtimedout { + fd.rmu.Unlock() + return 0, ErrDeadlineExceeded + } + fd.raio = newAsyncIO(fn, b) + fd.rmu.Unlock() + n, err := fd.raio.Wait() + fd.raio = nil + if isHangup(err) { + err = io.EOF + } + if isInterrupted(err) { + err = ErrDeadlineExceeded + } + return n, err +} + +// Write implements io.Writer. +func (fd *FD) Write(fn func([]byte) (int, error), b []byte) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + fd.wmu.Lock() + if fd.wtimedout { + fd.wmu.Unlock() + return 0, ErrDeadlineExceeded + } + fd.waio = newAsyncIO(fn, b) + fd.wmu.Unlock() + n, err := fd.waio.Wait() + fd.waio = nil + if isInterrupted(err) { + err = ErrDeadlineExceeded + } + return n, err +} + +// SetDeadline sets the read and write deadlines associated with fd. +func (fd *FD) SetDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r'+'w') +} + +// SetReadDeadline sets the read deadline associated with fd. +func (fd *FD) SetReadDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r') +} + +// SetWriteDeadline sets the write deadline associated with fd. +func (fd *FD) SetWriteDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'w') +} + +func setDeadlineImpl(fd *FD, t time.Time, mode int) error { + d := t.Sub(time.Now()) + if mode == 'r' || mode == 'r'+'w' { + fd.rmu.Lock() + defer fd.rmu.Unlock() + if fd.rtimer != nil { + fd.rtimer.Stop() + fd.rtimer = nil + } + fd.rtimedout = false + } + if mode == 'w' || mode == 'r'+'w' { + fd.wmu.Lock() + defer fd.wmu.Unlock() + if fd.wtimer != nil { + fd.wtimer.Stop() + fd.wtimer = nil + } + fd.wtimedout = false + } + if !t.IsZero() && d > 0 { + // Interrupt I/O operation once timer has expired + if mode == 'r' || mode == 'r'+'w' { + var timer *time.Timer + timer = time.AfterFunc(d, func() { + fd.rmu.Lock() + defer fd.rmu.Unlock() + if fd.rtimer != timer { + // deadline was changed + return + } + fd.rtimedout = true + if fd.raio != nil { + fd.raio.Cancel() + } + }) + fd.rtimer = timer + } + if mode == 'w' || mode == 'r'+'w' { + var timer *time.Timer + timer = time.AfterFunc(d, func() { + fd.wmu.Lock() + defer fd.wmu.Unlock() + if fd.wtimer != timer { + // deadline was changed + return + } + fd.wtimedout = true + if fd.waio != nil { + fd.waio.Cancel() + } + }) + fd.wtimer = timer + } + } + if !t.IsZero() && d <= 0 { + // Interrupt current I/O operation + if mode == 'r' || mode == 'r'+'w' { + fd.rtimedout = true + if fd.raio != nil { + fd.raio.Cancel() + } + } + if mode == 'w' || mode == 'r'+'w' { + fd.wtimedout = true + if fd.waio != nil { + fd.waio.Cancel() + } + } + } + return nil +} + +// On Plan 9 only, expose the locking for the net code. + +// ReadLock wraps FD.readLock. +func (fd *FD) ReadLock() error { + return fd.readLock() +} + +// ReadUnlock wraps FD.readUnlock. +func (fd *FD) ReadUnlock() { + fd.readUnlock() +} + +func isHangup(err error) bool { + return err != nil && stringsHasSuffix(err.Error(), "Hangup") +} + +func isInterrupted(err error) bool { + return err != nil && stringsHasSuffix(err.Error(), "interrupted") +} + +// IsPollDescriptor reports whether fd is the descriptor being used by the poller. +// This is only used for testing. +func IsPollDescriptor(fd uintptr) bool { + return false +} + +// RawControl invokes the user-defined function f for a non-IO +// operation. +func (fd *FD) RawControl(f func(uintptr)) error { + return errors.New("not implemented") +} + +// RawRead invokes the user-defined function f for a read operation. +func (fd *FD) RawRead(f func(uintptr) bool) error { + return errors.New("not implemented") +} + +// RawWrite invokes the user-defined function f for a write operation. +func (fd *FD) RawWrite(f func(uintptr) bool) error { + return errors.New("not implemented") +} diff --git a/src/internal/poll/fd_poll_js.go b/src/internal/poll/fd_poll_js.go new file mode 100644 index 0000000..fe5e73a --- /dev/null +++ b/src/internal/poll/fd_poll_js.go @@ -0,0 +1,99 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build js && wasm + +package poll + +import ( + "syscall" + "time" +) + +type pollDesc struct { + fd *FD + closing bool +} + +func (pd *pollDesc) init(fd *FD) error { pd.fd = fd; return nil } + +func (pd *pollDesc) close() {} + +func (pd *pollDesc) evict() { + pd.closing = true + if pd.fd != nil { + syscall.StopIO(pd.fd.Sysfd) + } +} + +func (pd *pollDesc) prepare(mode int, isFile bool) error { + if pd.closing { + return errClosing(isFile) + } + return nil +} + +func (pd *pollDesc) prepareRead(isFile bool) error { return pd.prepare('r', isFile) } + +func (pd *pollDesc) prepareWrite(isFile bool) error { return pd.prepare('w', isFile) } + +func (pd *pollDesc) wait(mode int, isFile bool) error { + if pd.closing { + return errClosing(isFile) + } + if isFile { // TODO(neelance): js/wasm: Use callbacks from JS to block until the read/write finished. + return nil + } + return ErrDeadlineExceeded +} + +func (pd *pollDesc) waitRead(isFile bool) error { return pd.wait('r', isFile) } + +func (pd *pollDesc) waitWrite(isFile bool) error { return pd.wait('w', isFile) } + +func (pd *pollDesc) waitCanceled(mode int) {} + +func (pd *pollDesc) pollable() bool { return true } + +// SetDeadline sets the read and write deadlines associated with fd. +func (fd *FD) SetDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r'+'w') +} + +// SetReadDeadline sets the read deadline associated with fd. +func (fd *FD) SetReadDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r') +} + +// SetWriteDeadline sets the write deadline associated with fd. +func (fd *FD) SetWriteDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'w') +} + +func setDeadlineImpl(fd *FD, t time.Time, mode int) error { + d := t.UnixNano() + if t.IsZero() { + d = 0 + } + if err := fd.incref(); err != nil { + return err + } + switch mode { + case 'r': + syscall.SetReadDeadline(fd.Sysfd, d) + case 'w': + syscall.SetWriteDeadline(fd.Sysfd, d) + case 'r' + 'w': + syscall.SetReadDeadline(fd.Sysfd, d) + syscall.SetWriteDeadline(fd.Sysfd, d) + } + fd.decref() + return nil +} + +// IsPollDescriptor reports whether fd is the descriptor being used by the poller. +// This is only used for testing. +func IsPollDescriptor(fd uintptr) bool { + return false +} diff --git a/src/internal/poll/fd_poll_runtime.go b/src/internal/poll/fd_poll_runtime.go new file mode 100644 index 0000000..b51535e --- /dev/null +++ b/src/internal/poll/fd_poll_runtime.go @@ -0,0 +1,169 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || windows || wasip1 + +package poll + +import ( + "errors" + "sync" + "syscall" + "time" + _ "unsafe" // for go:linkname +) + +// runtimeNano returns the current value of the runtime clock in nanoseconds. +// +//go:linkname runtimeNano runtime.nanotime +func runtimeNano() int64 + +func runtime_pollServerInit() +func runtime_pollOpen(fd uintptr) (uintptr, int) +func runtime_pollClose(ctx uintptr) +func runtime_pollWait(ctx uintptr, mode int) int +func runtime_pollWaitCanceled(ctx uintptr, mode int) +func runtime_pollReset(ctx uintptr, mode int) int +func runtime_pollSetDeadline(ctx uintptr, d int64, mode int) +func runtime_pollUnblock(ctx uintptr) +func runtime_isPollServerDescriptor(fd uintptr) bool + +type pollDesc struct { + runtimeCtx uintptr +} + +var serverInit sync.Once + +func (pd *pollDesc) init(fd *FD) error { + serverInit.Do(runtime_pollServerInit) + ctx, errno := runtime_pollOpen(uintptr(fd.Sysfd)) + if errno != 0 { + return errnoErr(syscall.Errno(errno)) + } + pd.runtimeCtx = ctx + return nil +} + +func (pd *pollDesc) close() { + if pd.runtimeCtx == 0 { + return + } + runtime_pollClose(pd.runtimeCtx) + pd.runtimeCtx = 0 +} + +// Evict evicts fd from the pending list, unblocking any I/O running on fd. +func (pd *pollDesc) evict() { + if pd.runtimeCtx == 0 { + return + } + runtime_pollUnblock(pd.runtimeCtx) +} + +func (pd *pollDesc) prepare(mode int, isFile bool) error { + if pd.runtimeCtx == 0 { + return nil + } + res := runtime_pollReset(pd.runtimeCtx, mode) + return convertErr(res, isFile) +} + +func (pd *pollDesc) prepareRead(isFile bool) error { + return pd.prepare('r', isFile) +} + +func (pd *pollDesc) prepareWrite(isFile bool) error { + return pd.prepare('w', isFile) +} + +func (pd *pollDesc) wait(mode int, isFile bool) error { + if pd.runtimeCtx == 0 { + return errors.New("waiting for unsupported file type") + } + res := runtime_pollWait(pd.runtimeCtx, mode) + return convertErr(res, isFile) +} + +func (pd *pollDesc) waitRead(isFile bool) error { + return pd.wait('r', isFile) +} + +func (pd *pollDesc) waitWrite(isFile bool) error { + return pd.wait('w', isFile) +} + +func (pd *pollDesc) waitCanceled(mode int) { + if pd.runtimeCtx == 0 { + return + } + runtime_pollWaitCanceled(pd.runtimeCtx, mode) +} + +func (pd *pollDesc) pollable() bool { + return pd.runtimeCtx != 0 +} + +// Error values returned by runtime_pollReset and runtime_pollWait. +// These must match the values in runtime/netpoll.go. +const ( + pollNoError = 0 + pollErrClosing = 1 + pollErrTimeout = 2 + pollErrNotPollable = 3 +) + +func convertErr(res int, isFile bool) error { + switch res { + case pollNoError: + return nil + case pollErrClosing: + return errClosing(isFile) + case pollErrTimeout: + return ErrDeadlineExceeded + case pollErrNotPollable: + return ErrNotPollable + } + println("unreachable: ", res) + panic("unreachable") +} + +// SetDeadline sets the read and write deadlines associated with fd. +func (fd *FD) SetDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r'+'w') +} + +// SetReadDeadline sets the read deadline associated with fd. +func (fd *FD) SetReadDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'r') +} + +// SetWriteDeadline sets the write deadline associated with fd. +func (fd *FD) SetWriteDeadline(t time.Time) error { + return setDeadlineImpl(fd, t, 'w') +} + +func setDeadlineImpl(fd *FD, t time.Time, mode int) error { + var d int64 + if !t.IsZero() { + d = int64(time.Until(t)) + if d == 0 { + d = -1 // don't confuse deadline right now with no deadline + } + } + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + if fd.pd.runtimeCtx == 0 { + return ErrNoDeadline + } + runtime_pollSetDeadline(fd.pd.runtimeCtx, d, mode) + return nil +} + +// IsPollDescriptor reports whether fd is the descriptor being used by the poller. +// This is only used for testing. +func IsPollDescriptor(fd uintptr) bool { + return runtime_isPollServerDescriptor(fd) +} diff --git a/src/internal/poll/fd_posix.go b/src/internal/poll/fd_posix.go new file mode 100644 index 0000000..5bd333b --- /dev/null +++ b/src/internal/poll/fd_posix.go @@ -0,0 +1,79 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || (js && wasm) || wasip1 || windows + +package poll + +import ( + "io" + "syscall" +) + +// eofError returns io.EOF when fd is available for reading end of +// file. +func (fd *FD) eofError(n int, err error) error { + if n == 0 && err == nil && fd.ZeroReadIsEOF { + return io.EOF + } + return err +} + +// Shutdown wraps syscall.Shutdown. +func (fd *FD) Shutdown(how int) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Shutdown(fd.Sysfd, how) +} + +// Fchown wraps syscall.Fchown. +func (fd *FD) Fchown(uid, gid int) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + return syscall.Fchown(fd.Sysfd, uid, gid) + }) +} + +// Ftruncate wraps syscall.Ftruncate. +func (fd *FD) Ftruncate(size int64) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + return syscall.Ftruncate(fd.Sysfd, size) + }) +} + +// RawControl invokes the user-defined function f for a non-IO +// operation. +func (fd *FD) RawControl(f func(uintptr)) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + f(uintptr(fd.Sysfd)) + return nil +} + +// ignoringEINTR makes a function call and repeats it if it returns +// an EINTR error. This appears to be required even though we install all +// signal handlers with SA_RESTART: see #22838, #38033, #38836, #40846. +// Also #20400 and #36644 are issues in which a signal handler is +// installed without setting SA_RESTART. None of these are the common case, +// but there are enough of them that it seems that we can't avoid +// an EINTR loop. +func ignoringEINTR(fn func() error) error { + for { + err := fn() + if err != syscall.EINTR { + return err + } + } +} diff --git a/src/internal/poll/fd_posix_test.go b/src/internal/poll/fd_posix_test.go new file mode 100644 index 0000000..b97e465 --- /dev/null +++ b/src/internal/poll/fd_posix_test.go @@ -0,0 +1,43 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || windows + +package poll_test + +import ( + . "internal/poll" + "io" + "testing" +) + +var eofErrorTests = []struct { + n int + err error + fd *FD + expected error +}{ + {100, nil, &FD{ZeroReadIsEOF: true}, nil}, + {100, io.EOF, &FD{ZeroReadIsEOF: true}, io.EOF}, + {100, ErrNetClosing, &FD{ZeroReadIsEOF: true}, ErrNetClosing}, + {0, nil, &FD{ZeroReadIsEOF: true}, io.EOF}, + {0, io.EOF, &FD{ZeroReadIsEOF: true}, io.EOF}, + {0, ErrNetClosing, &FD{ZeroReadIsEOF: true}, ErrNetClosing}, + + {100, nil, &FD{ZeroReadIsEOF: false}, nil}, + {100, io.EOF, &FD{ZeroReadIsEOF: false}, io.EOF}, + {100, ErrNetClosing, &FD{ZeroReadIsEOF: false}, ErrNetClosing}, + {0, nil, &FD{ZeroReadIsEOF: false}, nil}, + {0, io.EOF, &FD{ZeroReadIsEOF: false}, io.EOF}, + {0, ErrNetClosing, &FD{ZeroReadIsEOF: false}, ErrNetClosing}, +} + +func TestEOFError(t *testing.T) { + for _, tt := range eofErrorTests { + actual := tt.fd.EOFError(tt.n, tt.err) + if actual != tt.expected { + t.Errorf("eofError(%v, %v, %v): expected %v, actual %v", tt.n, tt.err, tt.fd.ZeroReadIsEOF, tt.expected, actual) + } + } +} diff --git a/src/internal/poll/fd_unix.go b/src/internal/poll/fd_unix.go new file mode 100644 index 0000000..61c2338 --- /dev/null +++ b/src/internal/poll/fd_unix.go @@ -0,0 +1,741 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || (js && wasm) || wasip1 + +package poll + +import ( + "internal/syscall/unix" + "io" + "sync/atomic" + "syscall" +) + +// FD is a file descriptor. The net and os packages use this type as a +// field of a larger type representing a network connection or OS file. +type FD struct { + // Lock sysfd and serialize access to Read and Write methods. + fdmu fdMutex + + // System file descriptor. Immutable until Close. + Sysfd int + + // Platform dependent state of the file descriptor. + SysFile + + // I/O poller. + pd pollDesc + + // Semaphore signaled when file is closed. + csema uint32 + + // Non-zero if this file has been set to blocking mode. + isBlocking uint32 + + // Whether this is a streaming descriptor, as opposed to a + // packet-based descriptor like a UDP socket. Immutable. + IsStream bool + + // Whether a zero byte read indicates EOF. This is false for a + // message based socket connection. + ZeroReadIsEOF bool + + // Whether this is a file rather than a network socket. + isFile bool +} + +// Init initializes the FD. The Sysfd field should already be set. +// This can be called multiple times on a single FD. +// The net argument is a network name from the net package (e.g., "tcp"), +// or "file". +// Set pollable to true if fd should be managed by runtime netpoll. +func (fd *FD) Init(net string, pollable bool) error { + fd.SysFile.init() + + // We don't actually care about the various network types. + if net == "file" { + fd.isFile = true + } + if !pollable { + fd.isBlocking = 1 + return nil + } + err := fd.pd.init(fd) + if err != nil { + // If we could not initialize the runtime poller, + // assume we are using blocking mode. + fd.isBlocking = 1 + } + return err +} + +// Destroy closes the file descriptor. This is called when there are +// no remaining references. +func (fd *FD) destroy() error { + // Poller may want to unregister fd in readiness notification mechanism, + // so this must be executed before CloseFunc. + fd.pd.close() + + err := fd.SysFile.destroy(fd.Sysfd) + + fd.Sysfd = -1 + runtime_Semrelease(&fd.csema) + return err +} + +// Close closes the FD. The underlying file descriptor is closed by the +// destroy method when there are no remaining references. +func (fd *FD) Close() error { + if !fd.fdmu.increfAndClose() { + return errClosing(fd.isFile) + } + + // Unblock any I/O. Once it all unblocks and returns, + // so that it cannot be referring to fd.sysfd anymore, + // the final decref will close fd.sysfd. This should happen + // fairly quickly, since all the I/O is non-blocking, and any + // attempts to block in the pollDesc will return errClosing(fd.isFile). + fd.pd.evict() + + // The call to decref will call destroy if there are no other + // references. + err := fd.decref() + + // Wait until the descriptor is closed. If this was the only + // reference, it is already closed. Only wait if the file has + // not been set to blocking mode, as otherwise any current I/O + // may be blocking, and that would block the Close. + // No need for an atomic read of isBlocking, increfAndClose means + // we have exclusive access to fd. + if fd.isBlocking == 0 { + runtime_Semacquire(&fd.csema) + } + + return err +} + +// SetBlocking puts the file into blocking mode. +func (fd *FD) SetBlocking() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + // Atomic store so that concurrent calls to SetBlocking + // do not cause a race condition. isBlocking only ever goes + // from 0 to 1 so there is no real race here. + atomic.StoreUint32(&fd.isBlocking, 1) + return syscall.SetNonblock(fd.Sysfd, false) +} + +// Darwin and FreeBSD can't read or write 2GB+ files at a time, +// even on 64-bit systems. +// The same is true of socket implementations on many systems. +// See golang.org/issue/7812 and golang.org/issue/16266. +// Use 1GB instead of, say, 2GB-1, to keep subsequent reads aligned. +const maxRW = 1 << 30 + +// Read implements io.Reader. +func (fd *FD) Read(p []byte) (int, error) { + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + if len(p) == 0 { + // If the caller wanted a zero byte read, return immediately + // without trying (but after acquiring the readLock). + // Otherwise syscall.Read returns 0, nil which looks like + // io.EOF. + // TODO(bradfitz): make it wait for readability? (Issue 15735) + return 0, nil + } + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, err + } + if fd.IsStream && len(p) > maxRW { + p = p[:maxRW] + } + for { + n, err := ignoringEINTRIO(syscall.Read, fd.Sysfd, p) + if err != nil { + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, err + } +} + +// Pread wraps the pread system call. +func (fd *FD) Pread(p []byte, off int64) (int, error) { + // Call incref, not readLock, because since pread specifies the + // offset it is independent from other reads. + // Similarly, using the poller doesn't make sense for pread. + if err := fd.incref(); err != nil { + return 0, err + } + if fd.IsStream && len(p) > maxRW { + p = p[:maxRW] + } + var ( + n int + err error + ) + for { + n, err = syscall.Pread(fd.Sysfd, p, off) + if err != syscall.EINTR { + break + } + } + if err != nil { + n = 0 + } + fd.decref() + err = fd.eofError(n, err) + return n, err +} + +// ReadFrom wraps the recvfrom network call. +func (fd *FD) ReadFrom(p []byte) (int, syscall.Sockaddr, error) { + if err := fd.readLock(); err != nil { + return 0, nil, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, nil, err + } + for { + n, sa, err := syscall.Recvfrom(fd.Sysfd, p, 0) + if err != nil { + if err == syscall.EINTR { + continue + } + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, sa, err + } +} + +// ReadFromInet4 wraps the recvfrom network call for IPv4. +func (fd *FD) ReadFromInet4(p []byte, from *syscall.SockaddrInet4) (int, error) { + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, err + } + for { + n, err := unix.RecvfromInet4(fd.Sysfd, p, 0, from) + if err != nil { + if err == syscall.EINTR { + continue + } + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, err + } +} + +// ReadFromInet6 wraps the recvfrom network call for IPv6. +func (fd *FD) ReadFromInet6(p []byte, from *syscall.SockaddrInet6) (int, error) { + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, err + } + for { + n, err := unix.RecvfromInet6(fd.Sysfd, p, 0, from) + if err != nil { + if err == syscall.EINTR { + continue + } + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, err + } +} + +// ReadMsg wraps the recvmsg network call. +func (fd *FD) ReadMsg(p []byte, oob []byte, flags int) (int, int, int, syscall.Sockaddr, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, nil, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, 0, 0, nil, err + } + for { + n, oobn, sysflags, sa, err := syscall.Recvmsg(fd.Sysfd, p, oob, flags) + if err != nil { + if err == syscall.EINTR { + continue + } + // TODO(dfc) should n and oobn be set to 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, oobn, sysflags, sa, err + } +} + +// ReadMsgInet4 is ReadMsg, but specialized for syscall.SockaddrInet4. +func (fd *FD) ReadMsgInet4(p []byte, oob []byte, flags int, sa4 *syscall.SockaddrInet4) (int, int, int, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, 0, 0, err + } + for { + n, oobn, sysflags, err := unix.RecvmsgInet4(fd.Sysfd, p, oob, flags, sa4) + if err != nil { + if err == syscall.EINTR { + continue + } + // TODO(dfc) should n and oobn be set to 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, oobn, sysflags, err + } +} + +// ReadMsgInet6 is ReadMsg, but specialized for syscall.SockaddrInet6. +func (fd *FD) ReadMsgInet6(p []byte, oob []byte, flags int, sa6 *syscall.SockaddrInet6) (int, int, int, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return 0, 0, 0, err + } + for { + n, oobn, sysflags, err := unix.RecvmsgInet6(fd.Sysfd, p, oob, flags, sa6) + if err != nil { + if err == syscall.EINTR { + continue + } + // TODO(dfc) should n and oobn be set to 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + err = fd.eofError(n, err) + return n, oobn, sysflags, err + } +} + +// Write implements io.Writer. +func (fd *FD) Write(p []byte) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, err + } + var nn int + for { + max := len(p) + if fd.IsStream && max-nn > maxRW { + max = nn + maxRW + } + n, err := ignoringEINTRIO(syscall.Write, fd.Sysfd, p[nn:max]) + if n > 0 { + nn += n + } + if nn == len(p) { + return nn, err + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return nn, err + } + if n == 0 { + return nn, io.ErrUnexpectedEOF + } + } +} + +// Pwrite wraps the pwrite system call. +func (fd *FD) Pwrite(p []byte, off int64) (int, error) { + // Call incref, not writeLock, because since pwrite specifies the + // offset it is independent from other writes. + // Similarly, using the poller doesn't make sense for pwrite. + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + var nn int + for { + max := len(p) + if fd.IsStream && max-nn > maxRW { + max = nn + maxRW + } + n, err := syscall.Pwrite(fd.Sysfd, p[nn:max], off+int64(nn)) + if err == syscall.EINTR { + continue + } + if n > 0 { + nn += n + } + if nn == len(p) { + return nn, err + } + if err != nil { + return nn, err + } + if n == 0 { + return nn, io.ErrUnexpectedEOF + } + } +} + +// WriteToInet4 wraps the sendto network call for IPv4 addresses. +func (fd *FD) WriteToInet4(p []byte, sa *syscall.SockaddrInet4) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, err + } + for { + err := unix.SendtoInet4(fd.Sysfd, p, 0, sa) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return 0, err + } + return len(p), nil + } +} + +// WriteToInet6 wraps the sendto network call for IPv6 addresses. +func (fd *FD) WriteToInet6(p []byte, sa *syscall.SockaddrInet6) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, err + } + for { + err := unix.SendtoInet6(fd.Sysfd, p, 0, sa) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return 0, err + } + return len(p), nil + } +} + +// WriteTo wraps the sendto network call. +func (fd *FD) WriteTo(p []byte, sa syscall.Sockaddr) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, err + } + for { + err := syscall.Sendto(fd.Sysfd, p, 0, sa) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return 0, err + } + return len(p), nil + } +} + +// WriteMsg wraps the sendmsg network call. +func (fd *FD) WriteMsg(p []byte, oob []byte, sa syscall.Sockaddr) (int, int, error) { + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, 0, err + } + for { + n, err := syscall.SendmsgN(fd.Sysfd, p, oob, sa, 0) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return n, 0, err + } + return n, len(oob), err + } +} + +// WriteMsgInet4 is WriteMsg specialized for syscall.SockaddrInet4. +func (fd *FD) WriteMsgInet4(p []byte, oob []byte, sa *syscall.SockaddrInet4) (int, int, error) { + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, 0, err + } + for { + n, err := unix.SendmsgNInet4(fd.Sysfd, p, oob, sa, 0) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return n, 0, err + } + return n, len(oob), err + } +} + +// WriteMsgInet6 is WriteMsg specialized for syscall.SockaddrInet6. +func (fd *FD) WriteMsgInet6(p []byte, oob []byte, sa *syscall.SockaddrInet6) (int, int, error) { + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, 0, err + } + for { + n, err := unix.SendmsgNInet6(fd.Sysfd, p, oob, sa, 0) + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + if err != nil { + return n, 0, err + } + return n, len(oob), err + } +} + +// Accept wraps the accept network call. +func (fd *FD) Accept() (int, syscall.Sockaddr, string, error) { + if err := fd.readLock(); err != nil { + return -1, nil, "", err + } + defer fd.readUnlock() + + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return -1, nil, "", err + } + for { + s, rsa, errcall, err := accept(fd.Sysfd) + if err == nil { + return s, rsa, "", err + } + switch err { + case syscall.EINTR: + continue + case syscall.EAGAIN: + if fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + case syscall.ECONNABORTED: + // This means that a socket on the listen + // queue was closed before we Accept()ed it; + // it's a silly error, so try again. + continue + } + return -1, nil, errcall, err + } +} + +// Fchmod wraps syscall.Fchmod. +func (fd *FD) Fchmod(mode uint32) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + return syscall.Fchmod(fd.Sysfd, mode) + }) +} + +// Fstat wraps syscall.Fstat +func (fd *FD) Fstat(s *syscall.Stat_t) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return ignoringEINTR(func() error { + return syscall.Fstat(fd.Sysfd, s) + }) +} + +// dupCloexecUnsupported indicates whether F_DUPFD_CLOEXEC is supported by the kernel. +var dupCloexecUnsupported atomic.Bool + +// DupCloseOnExec dups fd and marks it close-on-exec. +func DupCloseOnExec(fd int) (int, string, error) { + if syscall.F_DUPFD_CLOEXEC != 0 && !dupCloexecUnsupported.Load() { + r0, err := unix.Fcntl(fd, syscall.F_DUPFD_CLOEXEC, 0) + if err == nil { + return r0, "", nil + } + switch err { + case syscall.EINVAL, syscall.ENOSYS: + // Old kernel, or js/wasm (which returns + // ENOSYS). Fall back to the portable way from + // now on. + dupCloexecUnsupported.Store(true) + default: + return -1, "fcntl", err + } + } + return dupCloseOnExecOld(fd) +} + +// Dup duplicates the file descriptor. +func (fd *FD) Dup() (int, string, error) { + if err := fd.incref(); err != nil { + return -1, "", err + } + defer fd.decref() + return DupCloseOnExec(fd.Sysfd) +} + +// On Unix variants only, expose the IO event for the net code. + +// WaitWrite waits until data can be read from fd. +func (fd *FD) WaitWrite() error { + return fd.pd.waitWrite(fd.isFile) +} + +// WriteOnce is for testing only. It makes a single write call. +func (fd *FD) WriteOnce(p []byte) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + return ignoringEINTRIO(syscall.Write, fd.Sysfd, p) +} + +// RawRead invokes the user-defined function f for a read operation. +func (fd *FD) RawRead(f func(uintptr) bool) error { + if err := fd.readLock(); err != nil { + return err + } + defer fd.readUnlock() + if err := fd.pd.prepareRead(fd.isFile); err != nil { + return err + } + for { + if f(uintptr(fd.Sysfd)) { + return nil + } + if err := fd.pd.waitRead(fd.isFile); err != nil { + return err + } + } +} + +// RawWrite invokes the user-defined function f for a write operation. +func (fd *FD) RawWrite(f func(uintptr) bool) error { + if err := fd.writeLock(); err != nil { + return err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return err + } + for { + if f(uintptr(fd.Sysfd)) { + return nil + } + if err := fd.pd.waitWrite(fd.isFile); err != nil { + return err + } + } +} + +// ignoringEINTRIO is like ignoringEINTR, but just for IO calls. +func ignoringEINTRIO(fn func(fd int, p []byte) (int, error), fd int, p []byte) (int, error) { + for { + n, err := fn(fd, p) + if err != syscall.EINTR { + return n, err + } + } +} diff --git a/src/internal/poll/fd_unixjs.go b/src/internal/poll/fd_unixjs.go new file mode 100644 index 0000000..090974d --- /dev/null +++ b/src/internal/poll/fd_unixjs.go @@ -0,0 +1,79 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || (js && wasm) + +package poll + +import "syscall" + +type SysFile struct { + // Writev cache. + iovecs *[]syscall.Iovec +} + +func (s *SysFile) init() {} + +func (s *SysFile) destroy(fd int) error { + // We don't use ignoringEINTR here because POSIX does not define + // whether the descriptor is closed if close returns EINTR. + // If the descriptor is indeed closed, using a loop would race + // with some other goroutine opening a new descriptor. + // (The Linux kernel guarantees that it is closed on an EINTR error.) + return CloseFunc(fd) +} + +// dupCloseOnExecOld is the traditional way to dup an fd and +// set its O_CLOEXEC bit, using two system calls. +func dupCloseOnExecOld(fd int) (int, string, error) { + syscall.ForkLock.RLock() + defer syscall.ForkLock.RUnlock() + newfd, err := syscall.Dup(fd) + if err != nil { + return -1, "dup", err + } + syscall.CloseOnExec(newfd) + return newfd, "", nil +} + +// Fchdir wraps syscall.Fchdir. +func (fd *FD) Fchdir() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Fchdir(fd.Sysfd) +} + +// ReadDirent wraps syscall.ReadDirent. +// We treat this like an ordinary system call rather than a call +// that tries to fill the buffer. +func (fd *FD) ReadDirent(buf []byte) (int, error) { + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + for { + n, err := ignoringEINTRIO(syscall.ReadDirent, fd.Sysfd, buf) + if err != nil { + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + // Do not call eofError; caller does not expect to see io.EOF. + return n, err + } +} + +// Seek wraps syscall.Seek. +func (fd *FD) Seek(offset int64, whence int) (int64, error) { + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + return syscall.Seek(fd.Sysfd, offset, whence) +} diff --git a/src/internal/poll/fd_wasip1.go b/src/internal/poll/fd_wasip1.go new file mode 100644 index 0000000..aecd896 --- /dev/null +++ b/src/internal/poll/fd_wasip1.go @@ -0,0 +1,239 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "sync/atomic" + "syscall" + "unsafe" +) + +type SysFile struct { + // RefCountPtr is a pointer to the reference count of Sysfd. + // + // WASI preview 1 lacks a dup(2) system call. When the os and net packages + // need to share a file/socket, instead of duplicating the underlying file + // descriptor, we instead provide a way to copy FD instances and manage the + // underlying file descriptor with reference counting. + RefCountPtr *int32 + + // RefCount is the reference count of Sysfd. When a copy of an FD is made, + // it points to the reference count of the original FD instance. + RefCount int32 + + // Cache for the file type, lazily initialized when Seek is called. + Filetype uint32 + + // If the file represents a directory, this field contains the current + // readdir position. It is reset to zero if the program calls Seek(0, 0). + Dircookie uint64 + + // Absolute path of the file, as returned by syscall.PathOpen; + // this is used by Fchdir to emulate setting the current directory + // to an open file descriptor. + Path string + + // TODO(achille): it could be meaningful to move isFile from FD to a method + // on this struct type, and expose it as `IsFile() bool` which derives the + // result from the Filetype field. We would need to ensure that Filetype is + // always set instead of being lazily initialized. +} + +func (s *SysFile) init() { + if s.RefCountPtr == nil { + s.RefCount = 1 + s.RefCountPtr = &s.RefCount + } +} + +func (s *SysFile) ref() SysFile { + atomic.AddInt32(s.RefCountPtr, +1) + return SysFile{RefCountPtr: s.RefCountPtr} +} + +func (s *SysFile) destroy(fd int) error { + if s.RefCountPtr != nil && atomic.AddInt32(s.RefCountPtr, -1) > 0 { + return nil + } + + // We don't use ignoringEINTR here because POSIX does not define + // whether the descriptor is closed if close returns EINTR. + // If the descriptor is indeed closed, using a loop would race + // with some other goroutine opening a new descriptor. + // (The Linux kernel guarantees that it is closed on an EINTR error.) + return CloseFunc(fd) +} + +// Copy creates a copy of the FD. +// +// The FD instance points to the same underlying file descriptor. The file +// descriptor isn't closed until all FD instances that refer to it have been +// closed/destroyed. +func (fd *FD) Copy() FD { + return FD{ + Sysfd: fd.Sysfd, + SysFile: fd.SysFile.ref(), + IsStream: fd.IsStream, + ZeroReadIsEOF: fd.ZeroReadIsEOF, + isBlocking: fd.isBlocking, + isFile: fd.isFile, + } +} + +// dupCloseOnExecOld always errors on wasip1 because there is no mechanism to +// duplicate file descriptors. +func dupCloseOnExecOld(fd int) (int, string, error) { + return -1, "dup", syscall.ENOSYS +} + +// Fchdir wraps syscall.Fchdir. +func (fd *FD) Fchdir() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Chdir(fd.Path) +} + +// ReadDir wraps syscall.ReadDir. +// We treat this like an ordinary system call rather than a call +// that tries to fill the buffer. +func (fd *FD) ReadDir(buf []byte, cookie syscall.Dircookie) (int, error) { + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + for { + n, err := syscall.ReadDir(fd.Sysfd, buf, cookie) + if err != nil { + n = 0 + if err == syscall.EAGAIN && fd.pd.pollable() { + if err = fd.pd.waitRead(fd.isFile); err == nil { + continue + } + } + } + // Do not call eofError; caller does not expect to see io.EOF. + return n, err + } +} + +func (fd *FD) ReadDirent(buf []byte) (int, error) { + n, err := fd.ReadDir(buf, fd.Dircookie) + if err != nil { + return 0, err + } + if n <= 0 { + return n, nil // EOF + } + + // We assume that the caller of ReadDirent will consume the entire buffer + // up to the last full entry, so we scan through the buffer looking for the + // value of the last next cookie. + b := buf[:n] + + for len(b) > 0 { + next, ok := direntNext(b) + if !ok { + break + } + size, ok := direntReclen(b) + if !ok { + break + } + if size > uint64(len(b)) { + break + } + fd.Dircookie = syscall.Dircookie(next) + b = b[size:] + } + + // Trim a potentially incomplete trailing entry; this is necessary because + // the code in src/os/dir_unix.go does not deal well with partial values in + // calls to direntReclen, etc... and ends up causing an early EOF before all + // directory entries were consumed. ReadDirent is called with a large enough + // buffer (8 KiB) that at least one entry should always fit, tho this seems + // a bit brittle but cannot be addressed without a large change of the + // algorithm in the os.(*File).readdir method. + return n - len(b), nil +} + +// Seek wraps syscall.Seek. +func (fd *FD) Seek(offset int64, whence int) (int64, error) { + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + // syscall.Filetype is a uint8 but we store it as a uint32 in SysFile in + // order to use atomic load/store on the field, which is why we have to + // perform this type conversion. + fileType := syscall.Filetype(atomic.LoadUint32(&fd.Filetype)) + + if fileType == syscall.FILETYPE_UNKNOWN { + var stat syscall.Stat_t + if err := fd.Fstat(&stat); err != nil { + return 0, err + } + fileType = stat.Filetype + atomic.StoreUint32(&fd.Filetype, uint32(fileType)) + } + + if fileType == syscall.FILETYPE_DIRECTORY { + // If the file descriptor is opened on a directory, we reset the readdir + // cookie when seeking back to the beginning to allow reusing the file + // descriptor to scan the directory again. + if offset == 0 && whence == 0 { + fd.Dircookie = 0 + return 0, nil + } else { + return 0, syscall.EINVAL + } + } + + return syscall.Seek(fd.Sysfd, offset, whence) +} + +// https://github.com/WebAssembly/WASI/blob/main/legacy/preview1/docs.md#-dirent-record +const sizeOfDirent = 24 + +func direntReclen(buf []byte) (uint64, bool) { + namelen, ok := direntNamlen(buf) + return sizeOfDirent + namelen, ok +} + +func direntNamlen(buf []byte) (uint64, bool) { + return readInt(buf, unsafe.Offsetof(syscall.Dirent{}.Namlen), unsafe.Sizeof(syscall.Dirent{}.Namlen)) +} + +func direntNext(buf []byte) (uint64, bool) { + return readInt(buf, unsafe.Offsetof(syscall.Dirent{}.Next), unsafe.Sizeof(syscall.Dirent{}.Next)) +} + +// readInt returns the size-bytes unsigned integer in native byte order at offset off. +func readInt(b []byte, off, size uintptr) (u uint64, ok bool) { + if len(b) < int(off+size) { + return 0, false + } + return readIntLE(b[off:], size), true +} + +func readIntLE(b []byte, size uintptr) uint64 { + switch size { + case 1: + return uint64(b[0]) + case 2: + _ = b[1] // bounds check hint to compiler; see golang.org/issue/14808 + return uint64(b[0]) | uint64(b[1])<<8 + case 4: + _ = b[3] // bounds check hint to compiler; see golang.org/issue/14808 + return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 + case 8: + _ = b[7] // bounds check hint to compiler; see golang.org/issue/14808 + return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 | + uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56 + default: + panic("internal/poll: readInt with unsupported size") + } +} diff --git a/src/internal/poll/fd_windows.go b/src/internal/poll/fd_windows.go new file mode 100644 index 0000000..2095a6a --- /dev/null +++ b/src/internal/poll/fd_windows.go @@ -0,0 +1,1331 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "errors" + "internal/race" + "internal/syscall/windows" + "io" + "sync" + "syscall" + "unicode/utf16" + "unicode/utf8" + "unsafe" +) + +var ( + initErr error + ioSync uint64 +) + +// This package uses the SetFileCompletionNotificationModes Windows +// API to skip calling GetQueuedCompletionStatus if an IO operation +// completes synchronously. There is a known bug where +// SetFileCompletionNotificationModes crashes on some systems (see +// https://support.microsoft.com/kb/2568167 for details). + +var useSetFileCompletionNotificationModes bool // determines is SetFileCompletionNotificationModes is present and safe to use + +// checkSetFileCompletionNotificationModes verifies that +// SetFileCompletionNotificationModes Windows API is present +// on the system and is safe to use. +// See https://support.microsoft.com/kb/2568167 for details. +func checkSetFileCompletionNotificationModes() { + err := syscall.LoadSetFileCompletionNotificationModes() + if err != nil { + return + } + protos := [2]int32{syscall.IPPROTO_TCP, 0} + var buf [32]syscall.WSAProtocolInfo + len := uint32(unsafe.Sizeof(buf)) + n, err := syscall.WSAEnumProtocols(&protos[0], &buf[0], &len) + if err != nil { + return + } + for i := int32(0); i < n; i++ { + if buf[i].ServiceFlags1&syscall.XP1_IFS_HANDLES == 0 { + return + } + } + useSetFileCompletionNotificationModes = true +} + +func init() { + var d syscall.WSAData + e := syscall.WSAStartup(uint32(0x202), &d) + if e != nil { + initErr = e + } + checkSetFileCompletionNotificationModes() +} + +// operation contains superset of data necessary to perform all async IO. +type operation struct { + // Used by IOCP interface, it must be first field + // of the struct, as our code rely on it. + o syscall.Overlapped + + // fields used by runtime.netpoll + runtimeCtx uintptr + mode int32 + errno int32 + qty uint32 + + // fields used only by net package + fd *FD + buf syscall.WSABuf + msg windows.WSAMsg + sa syscall.Sockaddr + rsa *syscall.RawSockaddrAny + rsan int32 + handle syscall.Handle + flags uint32 + bufs []syscall.WSABuf +} + +func (o *operation) InitBuf(buf []byte) { + o.buf.Len = uint32(len(buf)) + o.buf.Buf = nil + if len(buf) != 0 { + o.buf.Buf = &buf[0] + } +} + +func (o *operation) InitBufs(buf *[][]byte) { + if o.bufs == nil { + o.bufs = make([]syscall.WSABuf, 0, len(*buf)) + } else { + o.bufs = o.bufs[:0] + } + for _, b := range *buf { + if len(b) == 0 { + o.bufs = append(o.bufs, syscall.WSABuf{}) + continue + } + for len(b) > maxRW { + o.bufs = append(o.bufs, syscall.WSABuf{Len: maxRW, Buf: &b[0]}) + b = b[maxRW:] + } + if len(b) > 0 { + o.bufs = append(o.bufs, syscall.WSABuf{Len: uint32(len(b)), Buf: &b[0]}) + } + } +} + +// ClearBufs clears all pointers to Buffers parameter captured +// by InitBufs, so it can be released by garbage collector. +func (o *operation) ClearBufs() { + for i := range o.bufs { + o.bufs[i].Buf = nil + } + o.bufs = o.bufs[:0] +} + +func (o *operation) InitMsg(p []byte, oob []byte) { + o.InitBuf(p) + o.msg.Buffers = &o.buf + o.msg.BufferCount = 1 + + o.msg.Name = nil + o.msg.Namelen = 0 + + o.msg.Flags = 0 + o.msg.Control.Len = uint32(len(oob)) + o.msg.Control.Buf = nil + if len(oob) != 0 { + o.msg.Control.Buf = &oob[0] + } +} + +// execIO executes a single IO operation o. It submits and cancels +// IO in the current thread for systems where Windows CancelIoEx API +// is available. Alternatively, it passes the request onto +// runtime netpoll and waits for completion or cancels request. +func execIO(o *operation, submit func(o *operation) error) (int, error) { + if o.fd.pd.runtimeCtx == 0 { + return 0, errors.New("internal error: polling on unsupported descriptor type") + } + + fd := o.fd + // Notify runtime netpoll about starting IO. + err := fd.pd.prepare(int(o.mode), fd.isFile) + if err != nil { + return 0, err + } + // Start IO. + err = submit(o) + switch err { + case nil: + // IO completed immediately + if o.fd.skipSyncNotif { + // No completion message will follow, so return immediately. + return int(o.qty), nil + } + // Need to get our completion message anyway. + case syscall.ERROR_IO_PENDING: + // IO started, and we have to wait for its completion. + err = nil + default: + return 0, err + } + // Wait for our request to complete. + err = fd.pd.wait(int(o.mode), fd.isFile) + if err == nil { + // All is good. Extract our IO results and return. + if o.errno != 0 { + err = syscall.Errno(o.errno) + // More data available. Return back the size of received data. + if err == syscall.ERROR_MORE_DATA || err == windows.WSAEMSGSIZE { + return int(o.qty), err + } + return 0, err + } + return int(o.qty), nil + } + // IO is interrupted by "close" or "timeout" + netpollErr := err + switch netpollErr { + case ErrNetClosing, ErrFileClosing, ErrDeadlineExceeded: + // will deal with those. + default: + panic("unexpected runtime.netpoll error: " + netpollErr.Error()) + } + // Cancel our request. + err = syscall.CancelIoEx(fd.Sysfd, &o.o) + // Assuming ERROR_NOT_FOUND is returned, if IO is completed. + if err != nil && err != syscall.ERROR_NOT_FOUND { + // TODO(brainman): maybe do something else, but panic. + panic(err) + } + // Wait for cancellation to complete. + fd.pd.waitCanceled(int(o.mode)) + if o.errno != 0 { + err = syscall.Errno(o.errno) + if err == syscall.ERROR_OPERATION_ABORTED { // IO Canceled + err = netpollErr + } + return 0, err + } + // We issued a cancellation request. But, it seems, IO operation succeeded + // before the cancellation request run. We need to treat the IO operation as + // succeeded (the bytes are actually sent/recv from network). + return int(o.qty), nil +} + +// FD is a file descriptor. The net and os packages embed this type in +// a larger type representing a network connection or OS file. +type FD struct { + // Lock sysfd and serialize access to Read and Write methods. + fdmu fdMutex + + // System file descriptor. Immutable until Close. + Sysfd syscall.Handle + + // Read operation. + rop operation + // Write operation. + wop operation + + // I/O poller. + pd pollDesc + + // Used to implement pread/pwrite. + l sync.Mutex + + // For console I/O. + lastbits []byte // first few bytes of the last incomplete rune in last write + readuint16 []uint16 // buffer to hold uint16s obtained with ReadConsole + readbyte []byte // buffer to hold decoding of readuint16 from utf16 to utf8 + readbyteOffset int // readbyte[readOffset:] is yet to be consumed with file.Read + + // Semaphore signaled when file is closed. + csema uint32 + + skipSyncNotif bool + + // Whether this is a streaming descriptor, as opposed to a + // packet-based descriptor like a UDP socket. + IsStream bool + + // Whether a zero byte read indicates EOF. This is false for a + // message based socket connection. + ZeroReadIsEOF bool + + // Whether this is a file rather than a network socket. + isFile bool + + // The kind of this file. + kind fileKind +} + +// fileKind describes the kind of file. +type fileKind byte + +const ( + kindNet fileKind = iota + kindFile + kindConsole + kindPipe +) + +// logInitFD is set by tests to enable file descriptor initialization logging. +var logInitFD func(net string, fd *FD, err error) + +// Init initializes the FD. The Sysfd field should already be set. +// This can be called multiple times on a single FD. +// The net argument is a network name from the net package (e.g., "tcp"), +// or "file" or "console" or "dir". +// Set pollable to true if fd should be managed by runtime netpoll. +func (fd *FD) Init(net string, pollable bool) (string, error) { + if initErr != nil { + return "", initErr + } + + switch net { + case "file", "dir": + fd.kind = kindFile + case "console": + fd.kind = kindConsole + case "pipe": + fd.kind = kindPipe + case "tcp", "tcp4", "tcp6", + "udp", "udp4", "udp6", + "ip", "ip4", "ip6", + "unix", "unixgram", "unixpacket": + fd.kind = kindNet + default: + return "", errors.New("internal error: unknown network type " + net) + } + fd.isFile = fd.kind != kindNet + + var err error + if pollable { + // Only call init for a network socket. + // This means that we don't add files to the runtime poller. + // Adding files to the runtime poller can confuse matters + // if the user is doing their own overlapped I/O. + // See issue #21172. + // + // In general the code below avoids calling the execIO + // function for non-network sockets. If some method does + // somehow call execIO, then execIO, and therefore the + // calling method, will return an error, because + // fd.pd.runtimeCtx will be 0. + err = fd.pd.init(fd) + } + if logInitFD != nil { + logInitFD(net, fd, err) + } + if err != nil { + return "", err + } + if pollable && useSetFileCompletionNotificationModes { + // We do not use events, so we can skip them always. + flags := uint8(syscall.FILE_SKIP_SET_EVENT_ON_HANDLE) + switch net { + case "tcp", "tcp4", "tcp6", + "udp", "udp4", "udp6": + flags |= syscall.FILE_SKIP_COMPLETION_PORT_ON_SUCCESS + } + err := syscall.SetFileCompletionNotificationModes(fd.Sysfd, flags) + if err == nil && flags&syscall.FILE_SKIP_COMPLETION_PORT_ON_SUCCESS != 0 { + fd.skipSyncNotif = true + } + } + // Disable SIO_UDP_CONNRESET behavior. + // http://support.microsoft.com/kb/263823 + switch net { + case "udp", "udp4", "udp6": + ret := uint32(0) + flag := uint32(0) + size := uint32(unsafe.Sizeof(flag)) + err := syscall.WSAIoctl(fd.Sysfd, syscall.SIO_UDP_CONNRESET, (*byte)(unsafe.Pointer(&flag)), size, nil, 0, &ret, nil, 0) + if err != nil { + return "wsaioctl", err + } + } + fd.rop.mode = 'r' + fd.wop.mode = 'w' + fd.rop.fd = fd + fd.wop.fd = fd + fd.rop.runtimeCtx = fd.pd.runtimeCtx + fd.wop.runtimeCtx = fd.pd.runtimeCtx + return "", nil +} + +func (fd *FD) destroy() error { + if fd.Sysfd == syscall.InvalidHandle { + return syscall.EINVAL + } + // Poller may want to unregister fd in readiness notification mechanism, + // so this must be executed before fd.CloseFunc. + fd.pd.close() + var err error + switch fd.kind { + case kindNet: + // The net package uses the CloseFunc variable for testing. + err = CloseFunc(fd.Sysfd) + default: + err = syscall.CloseHandle(fd.Sysfd) + } + fd.Sysfd = syscall.InvalidHandle + runtime_Semrelease(&fd.csema) + return err +} + +// Close closes the FD. The underlying file descriptor is closed by +// the destroy method when there are no remaining references. +func (fd *FD) Close() error { + if !fd.fdmu.increfAndClose() { + return errClosing(fd.isFile) + } + if fd.kind == kindPipe { + syscall.CancelIoEx(fd.Sysfd, nil) + } + // unblock pending reader and writer + fd.pd.evict() + err := fd.decref() + // Wait until the descriptor is closed. If this was the only + // reference, it is already closed. + runtime_Semacquire(&fd.csema) + return err +} + +// Windows ReadFile and WSARecv use DWORD (uint32) parameter to pass buffer length. +// This prevents us reading blocks larger than 4GB. +// See golang.org/issue/26923. +const maxRW = 1 << 30 // 1GB is large enough and keeps subsequent reads aligned + +// Read implements io.Reader. +func (fd *FD) Read(buf []byte) (int, error) { + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + + if len(buf) > maxRW { + buf = buf[:maxRW] + } + + var n int + var err error + if fd.isFile { + fd.l.Lock() + defer fd.l.Unlock() + switch fd.kind { + case kindConsole: + n, err = fd.readConsole(buf) + default: + n, err = syscall.Read(fd.Sysfd, buf) + if fd.kind == kindPipe && err == syscall.ERROR_OPERATION_ABORTED { + // Close uses CancelIoEx to interrupt concurrent I/O for pipes. + // If the fd is a pipe and the Read was interrupted by CancelIoEx, + // we assume it is interrupted by Close. + err = ErrFileClosing + } + } + if err != nil { + n = 0 + } + } else { + o := &fd.rop + o.InitBuf(buf) + n, err = execIO(o, func(o *operation) error { + return syscall.WSARecv(o.fd.Sysfd, &o.buf, 1, &o.qty, &o.flags, &o.o, nil) + }) + if race.Enabled { + race.Acquire(unsafe.Pointer(&ioSync)) + } + } + if len(buf) != 0 { + err = fd.eofError(n, err) + } + return n, err +} + +var ReadConsole = syscall.ReadConsole // changed for testing + +// readConsole reads utf16 characters from console File, +// encodes them into utf8 and stores them in buffer b. +// It returns the number of utf8 bytes read and an error, if any. +func (fd *FD) readConsole(b []byte) (int, error) { + if len(b) == 0 { + return 0, nil + } + + if fd.readuint16 == nil { + // Note: syscall.ReadConsole fails for very large buffers. + // The limit is somewhere around (but not exactly) 16384. + // Stay well below. + fd.readuint16 = make([]uint16, 0, 10000) + fd.readbyte = make([]byte, 0, 4*cap(fd.readuint16)) + } + + for fd.readbyteOffset >= len(fd.readbyte) { + n := cap(fd.readuint16) - len(fd.readuint16) + if n > len(b) { + n = len(b) + } + var nw uint32 + err := ReadConsole(fd.Sysfd, &fd.readuint16[:len(fd.readuint16)+1][len(fd.readuint16)], uint32(n), &nw, nil) + if err != nil { + return 0, err + } + uint16s := fd.readuint16[:len(fd.readuint16)+int(nw)] + fd.readuint16 = fd.readuint16[:0] + buf := fd.readbyte[:0] + for i := 0; i < len(uint16s); i++ { + r := rune(uint16s[i]) + if utf16.IsSurrogate(r) { + if i+1 == len(uint16s) { + if nw > 0 { + // Save half surrogate pair for next time. + fd.readuint16 = fd.readuint16[:1] + fd.readuint16[0] = uint16(r) + break + } + r = utf8.RuneError + } else { + r = utf16.DecodeRune(r, rune(uint16s[i+1])) + if r != utf8.RuneError { + i++ + } + } + } + buf = utf8.AppendRune(buf, r) + } + fd.readbyte = buf + fd.readbyteOffset = 0 + if nw == 0 { + break + } + } + + src := fd.readbyte[fd.readbyteOffset:] + var i int + for i = 0; i < len(src) && i < len(b); i++ { + x := src[i] + if x == 0x1A { // Ctrl-Z + if i == 0 { + fd.readbyteOffset++ + } + break + } + b[i] = x + } + fd.readbyteOffset += i + return i, nil +} + +// Pread emulates the Unix pread system call. +func (fd *FD) Pread(b []byte, off int64) (int, error) { + if fd.kind == kindPipe { + // Pread does not work with pipes + return 0, syscall.ESPIPE + } + // Call incref, not readLock, because since pread specifies the + // offset it is independent from other reads. + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + + if len(b) > maxRW { + b = b[:maxRW] + } + + fd.l.Lock() + defer fd.l.Unlock() + curoffset, e := syscall.Seek(fd.Sysfd, 0, io.SeekCurrent) + if e != nil { + return 0, e + } + defer syscall.Seek(fd.Sysfd, curoffset, io.SeekStart) + o := syscall.Overlapped{ + OffsetHigh: uint32(off >> 32), + Offset: uint32(off), + } + var done uint32 + e = syscall.ReadFile(fd.Sysfd, b, &done, &o) + if e != nil { + done = 0 + if e == syscall.ERROR_HANDLE_EOF { + e = io.EOF + } + } + if len(b) != 0 { + e = fd.eofError(int(done), e) + } + return int(done), e +} + +// ReadFrom wraps the recvfrom network call. +func (fd *FD) ReadFrom(buf []byte) (int, syscall.Sockaddr, error) { + if len(buf) == 0 { + return 0, nil, nil + } + if len(buf) > maxRW { + buf = buf[:maxRW] + } + if err := fd.readLock(); err != nil { + return 0, nil, err + } + defer fd.readUnlock() + o := &fd.rop + o.InitBuf(buf) + n, err := execIO(o, func(o *operation) error { + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.rsan = int32(unsafe.Sizeof(*o.rsa)) + return syscall.WSARecvFrom(o.fd.Sysfd, &o.buf, 1, &o.qty, &o.flags, o.rsa, &o.rsan, &o.o, nil) + }) + err = fd.eofError(n, err) + if err != nil { + return n, nil, err + } + sa, _ := o.rsa.Sockaddr() + return n, sa, nil +} + +// ReadFromInet4 wraps the recvfrom network call for IPv4. +func (fd *FD) ReadFromInet4(buf []byte, sa4 *syscall.SockaddrInet4) (int, error) { + if len(buf) == 0 { + return 0, nil + } + if len(buf) > maxRW { + buf = buf[:maxRW] + } + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + o := &fd.rop + o.InitBuf(buf) + n, err := execIO(o, func(o *operation) error { + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.rsan = int32(unsafe.Sizeof(*o.rsa)) + return syscall.WSARecvFrom(o.fd.Sysfd, &o.buf, 1, &o.qty, &o.flags, o.rsa, &o.rsan, &o.o, nil) + }) + err = fd.eofError(n, err) + if err != nil { + return n, err + } + rawToSockaddrInet4(o.rsa, sa4) + return n, err +} + +// ReadFromInet6 wraps the recvfrom network call for IPv6. +func (fd *FD) ReadFromInet6(buf []byte, sa6 *syscall.SockaddrInet6) (int, error) { + if len(buf) == 0 { + return 0, nil + } + if len(buf) > maxRW { + buf = buf[:maxRW] + } + if err := fd.readLock(); err != nil { + return 0, err + } + defer fd.readUnlock() + o := &fd.rop + o.InitBuf(buf) + n, err := execIO(o, func(o *operation) error { + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.rsan = int32(unsafe.Sizeof(*o.rsa)) + return syscall.WSARecvFrom(o.fd.Sysfd, &o.buf, 1, &o.qty, &o.flags, o.rsa, &o.rsan, &o.o, nil) + }) + err = fd.eofError(n, err) + if err != nil { + return n, err + } + rawToSockaddrInet6(o.rsa, sa6) + return n, err +} + +// Write implements io.Writer. +func (fd *FD) Write(buf []byte) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if fd.isFile { + fd.l.Lock() + defer fd.l.Unlock() + } + + ntotal := 0 + for len(buf) > 0 { + b := buf + if len(b) > maxRW { + b = b[:maxRW] + } + var n int + var err error + if fd.isFile { + switch fd.kind { + case kindConsole: + n, err = fd.writeConsole(b) + default: + n, err = syscall.Write(fd.Sysfd, b) + if fd.kind == kindPipe && err == syscall.ERROR_OPERATION_ABORTED { + // Close uses CancelIoEx to interrupt concurrent I/O for pipes. + // If the fd is a pipe and the Write was interrupted by CancelIoEx, + // we assume it is interrupted by Close. + err = ErrFileClosing + } + } + if err != nil { + n = 0 + } + } else { + if race.Enabled { + race.ReleaseMerge(unsafe.Pointer(&ioSync)) + } + o := &fd.wop + o.InitBuf(b) + n, err = execIO(o, func(o *operation) error { + return syscall.WSASend(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, &o.o, nil) + }) + } + ntotal += n + if err != nil { + return ntotal, err + } + buf = buf[n:] + } + return ntotal, nil +} + +// writeConsole writes len(b) bytes to the console File. +// It returns the number of bytes written and an error, if any. +func (fd *FD) writeConsole(b []byte) (int, error) { + n := len(b) + runes := make([]rune, 0, 256) + if len(fd.lastbits) > 0 { + b = append(fd.lastbits, b...) + fd.lastbits = nil + + } + for len(b) >= utf8.UTFMax || utf8.FullRune(b) { + r, l := utf8.DecodeRune(b) + runes = append(runes, r) + b = b[l:] + } + if len(b) > 0 { + fd.lastbits = make([]byte, len(b)) + copy(fd.lastbits, b) + } + // syscall.WriteConsole seems to fail, if given large buffer. + // So limit the buffer to 16000 characters. This number was + // discovered by experimenting with syscall.WriteConsole. + const maxWrite = 16000 + for len(runes) > 0 { + m := len(runes) + if m > maxWrite { + m = maxWrite + } + chunk := runes[:m] + runes = runes[m:] + uint16s := utf16.Encode(chunk) + for len(uint16s) > 0 { + var written uint32 + err := syscall.WriteConsole(fd.Sysfd, &uint16s[0], uint32(len(uint16s)), &written, nil) + if err != nil { + return 0, err + } + uint16s = uint16s[written:] + } + } + return n, nil +} + +// Pwrite emulates the Unix pwrite system call. +func (fd *FD) Pwrite(buf []byte, off int64) (int, error) { + if fd.kind == kindPipe { + // Pwrite does not work with pipes + return 0, syscall.ESPIPE + } + // Call incref, not writeLock, because since pwrite specifies the + // offset it is independent from other writes. + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + + fd.l.Lock() + defer fd.l.Unlock() + curoffset, e := syscall.Seek(fd.Sysfd, 0, io.SeekCurrent) + if e != nil { + return 0, e + } + defer syscall.Seek(fd.Sysfd, curoffset, io.SeekStart) + + ntotal := 0 + for len(buf) > 0 { + b := buf + if len(b) > maxRW { + b = b[:maxRW] + } + var n uint32 + o := syscall.Overlapped{ + OffsetHigh: uint32(off >> 32), + Offset: uint32(off), + } + e = syscall.WriteFile(fd.Sysfd, b, &n, &o) + ntotal += int(n) + if e != nil { + return ntotal, e + } + buf = buf[n:] + off += int64(n) + } + return ntotal, nil +} + +// Writev emulates the Unix writev system call. +func (fd *FD) Writev(buf *[][]byte) (int64, error) { + if len(*buf) == 0 { + return 0, nil + } + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if race.Enabled { + race.ReleaseMerge(unsafe.Pointer(&ioSync)) + } + o := &fd.wop + o.InitBufs(buf) + n, err := execIO(o, func(o *operation) error { + return syscall.WSASend(o.fd.Sysfd, &o.bufs[0], uint32(len(o.bufs)), &o.qty, 0, &o.o, nil) + }) + o.ClearBufs() + TestHookDidWritev(n) + consume(buf, int64(n)) + return int64(n), err +} + +// WriteTo wraps the sendto network call. +func (fd *FD) WriteTo(buf []byte, sa syscall.Sockaddr) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + + if len(buf) == 0 { + // handle zero-byte payload + o := &fd.wop + o.InitBuf(buf) + o.sa = sa + n, err := execIO(o, func(o *operation) error { + return syscall.WSASendto(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, o.sa, &o.o, nil) + }) + return n, err + } + + ntotal := 0 + for len(buf) > 0 { + b := buf + if len(b) > maxRW { + b = b[:maxRW] + } + o := &fd.wop + o.InitBuf(b) + o.sa = sa + n, err := execIO(o, func(o *operation) error { + return syscall.WSASendto(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, o.sa, &o.o, nil) + }) + ntotal += int(n) + if err != nil { + return ntotal, err + } + buf = buf[n:] + } + return ntotal, nil +} + +// WriteToInet4 is WriteTo, specialized for syscall.SockaddrInet4. +func (fd *FD) WriteToInet4(buf []byte, sa4 *syscall.SockaddrInet4) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + + if len(buf) == 0 { + // handle zero-byte payload + o := &fd.wop + o.InitBuf(buf) + n, err := execIO(o, func(o *operation) error { + return windows.WSASendtoInet4(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, sa4, &o.o, nil) + }) + return n, err + } + + ntotal := 0 + for len(buf) > 0 { + b := buf + if len(b) > maxRW { + b = b[:maxRW] + } + o := &fd.wop + o.InitBuf(b) + n, err := execIO(o, func(o *operation) error { + return windows.WSASendtoInet4(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, sa4, &o.o, nil) + }) + ntotal += int(n) + if err != nil { + return ntotal, err + } + buf = buf[n:] + } + return ntotal, nil +} + +// WriteToInet6 is WriteTo, specialized for syscall.SockaddrInet6. +func (fd *FD) WriteToInet6(buf []byte, sa6 *syscall.SockaddrInet6) (int, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + + if len(buf) == 0 { + // handle zero-byte payload + o := &fd.wop + o.InitBuf(buf) + n, err := execIO(o, func(o *operation) error { + return windows.WSASendtoInet6(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, sa6, &o.o, nil) + }) + return n, err + } + + ntotal := 0 + for len(buf) > 0 { + b := buf + if len(b) > maxRW { + b = b[:maxRW] + } + o := &fd.wop + o.InitBuf(b) + n, err := execIO(o, func(o *operation) error { + return windows.WSASendtoInet6(o.fd.Sysfd, &o.buf, 1, &o.qty, 0, sa6, &o.o, nil) + }) + ntotal += int(n) + if err != nil { + return ntotal, err + } + buf = buf[n:] + } + return ntotal, nil +} + +// Call ConnectEx. This doesn't need any locking, since it is only +// called when the descriptor is first created. This is here rather +// than in the net package so that it can use fd.wop. +func (fd *FD) ConnectEx(ra syscall.Sockaddr) error { + o := &fd.wop + o.sa = ra + _, err := execIO(o, func(o *operation) error { + return ConnectExFunc(o.fd.Sysfd, o.sa, nil, 0, nil, &o.o) + }) + return err +} + +func (fd *FD) acceptOne(s syscall.Handle, rawsa []syscall.RawSockaddrAny, o *operation) (string, error) { + // Submit accept request. + o.handle = s + o.rsan = int32(unsafe.Sizeof(rawsa[0])) + _, err := execIO(o, func(o *operation) error { + return AcceptFunc(o.fd.Sysfd, o.handle, (*byte)(unsafe.Pointer(&rawsa[0])), 0, uint32(o.rsan), uint32(o.rsan), &o.qty, &o.o) + }) + if err != nil { + CloseFunc(s) + return "acceptex", err + } + + // Inherit properties of the listening socket. + err = syscall.Setsockopt(s, syscall.SOL_SOCKET, syscall.SO_UPDATE_ACCEPT_CONTEXT, (*byte)(unsafe.Pointer(&fd.Sysfd)), int32(unsafe.Sizeof(fd.Sysfd))) + if err != nil { + CloseFunc(s) + return "setsockopt", err + } + + return "", nil +} + +// Accept handles accepting a socket. The sysSocket parameter is used +// to allocate the net socket. +func (fd *FD) Accept(sysSocket func() (syscall.Handle, error)) (syscall.Handle, []syscall.RawSockaddrAny, uint32, string, error) { + if err := fd.readLock(); err != nil { + return syscall.InvalidHandle, nil, 0, "", err + } + defer fd.readUnlock() + + o := &fd.rop + var rawsa [2]syscall.RawSockaddrAny + for { + s, err := sysSocket() + if err != nil { + return syscall.InvalidHandle, nil, 0, "", err + } + + errcall, err := fd.acceptOne(s, rawsa[:], o) + if err == nil { + return s, rawsa[:], uint32(o.rsan), "", nil + } + + // Sometimes we see WSAECONNRESET and ERROR_NETNAME_DELETED is + // returned here. These happen if connection reset is received + // before AcceptEx could complete. These errors relate to new + // connection, not to AcceptEx, so ignore broken connection and + // try AcceptEx again for more connections. + errno, ok := err.(syscall.Errno) + if !ok { + return syscall.InvalidHandle, nil, 0, errcall, err + } + switch errno { + case syscall.ERROR_NETNAME_DELETED, syscall.WSAECONNRESET: + // ignore these and try again + default: + return syscall.InvalidHandle, nil, 0, errcall, err + } + } +} + +// Seek wraps syscall.Seek. +func (fd *FD) Seek(offset int64, whence int) (int64, error) { + if fd.kind == kindPipe { + return 0, syscall.ESPIPE + } + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + + fd.l.Lock() + defer fd.l.Unlock() + + return syscall.Seek(fd.Sysfd, offset, whence) +} + +// Fchmod updates syscall.ByHandleFileInformation.Fileattributes when needed. +func (fd *FD) Fchmod(mode uint32) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + + var d syscall.ByHandleFileInformation + if err := syscall.GetFileInformationByHandle(fd.Sysfd, &d); err != nil { + return err + } + attrs := d.FileAttributes + if mode&syscall.S_IWRITE != 0 { + attrs &^= syscall.FILE_ATTRIBUTE_READONLY + } else { + attrs |= syscall.FILE_ATTRIBUTE_READONLY + } + if attrs == d.FileAttributes { + return nil + } + + var du windows.FILE_BASIC_INFO + du.FileAttributes = attrs + return windows.SetFileInformationByHandle(fd.Sysfd, windows.FileBasicInfo, unsafe.Pointer(&du), uint32(unsafe.Sizeof(du))) +} + +// Fchdir wraps syscall.Fchdir. +func (fd *FD) Fchdir() error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Fchdir(fd.Sysfd) +} + +// GetFileType wraps syscall.GetFileType. +func (fd *FD) GetFileType() (uint32, error) { + if err := fd.incref(); err != nil { + return 0, err + } + defer fd.decref() + return syscall.GetFileType(fd.Sysfd) +} + +// GetFileInformationByHandle wraps GetFileInformationByHandle. +func (fd *FD) GetFileInformationByHandle(data *syscall.ByHandleFileInformation) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.GetFileInformationByHandle(fd.Sysfd, data) +} + +// RawRead invokes the user-defined function f for a read operation. +func (fd *FD) RawRead(f func(uintptr) bool) error { + if err := fd.readLock(); err != nil { + return err + } + defer fd.readUnlock() + for { + if f(uintptr(fd.Sysfd)) { + return nil + } + + // Use a zero-byte read as a way to get notified when this + // socket is readable. h/t https://stackoverflow.com/a/42019668/332798 + o := &fd.rop + o.InitBuf(nil) + if !fd.IsStream { + o.flags |= windows.MSG_PEEK + } + _, err := execIO(o, func(o *operation) error { + return syscall.WSARecv(o.fd.Sysfd, &o.buf, 1, &o.qty, &o.flags, &o.o, nil) + }) + if err == windows.WSAEMSGSIZE { + // expected with a 0-byte peek, ignore. + } else if err != nil { + return err + } + } +} + +// RawWrite invokes the user-defined function f for a write operation. +func (fd *FD) RawWrite(f func(uintptr) bool) error { + if err := fd.writeLock(); err != nil { + return err + } + defer fd.writeUnlock() + + if f(uintptr(fd.Sysfd)) { + return nil + } + + // TODO(tmm1): find a way to detect socket writability + return syscall.EWINDOWS +} + +func sockaddrInet4ToRaw(rsa *syscall.RawSockaddrAny, sa *syscall.SockaddrInet4) int32 { + *rsa = syscall.RawSockaddrAny{} + raw := (*syscall.RawSockaddrInet4)(unsafe.Pointer(rsa)) + raw.Family = syscall.AF_INET + p := (*[2]byte)(unsafe.Pointer(&raw.Port)) + p[0] = byte(sa.Port >> 8) + p[1] = byte(sa.Port) + raw.Addr = sa.Addr + return int32(unsafe.Sizeof(*raw)) +} + +func sockaddrInet6ToRaw(rsa *syscall.RawSockaddrAny, sa *syscall.SockaddrInet6) int32 { + *rsa = syscall.RawSockaddrAny{} + raw := (*syscall.RawSockaddrInet6)(unsafe.Pointer(rsa)) + raw.Family = syscall.AF_INET6 + p := (*[2]byte)(unsafe.Pointer(&raw.Port)) + p[0] = byte(sa.Port >> 8) + p[1] = byte(sa.Port) + raw.Scope_id = sa.ZoneId + raw.Addr = sa.Addr + return int32(unsafe.Sizeof(*raw)) +} + +func rawToSockaddrInet4(rsa *syscall.RawSockaddrAny, sa *syscall.SockaddrInet4) { + pp := (*syscall.RawSockaddrInet4)(unsafe.Pointer(rsa)) + p := (*[2]byte)(unsafe.Pointer(&pp.Port)) + sa.Port = int(p[0])<<8 + int(p[1]) + sa.Addr = pp.Addr +} + +func rawToSockaddrInet6(rsa *syscall.RawSockaddrAny, sa *syscall.SockaddrInet6) { + pp := (*syscall.RawSockaddrInet6)(unsafe.Pointer(rsa)) + p := (*[2]byte)(unsafe.Pointer(&pp.Port)) + sa.Port = int(p[0])<<8 + int(p[1]) + sa.ZoneId = pp.Scope_id + sa.Addr = pp.Addr +} + +func sockaddrToRaw(rsa *syscall.RawSockaddrAny, sa syscall.Sockaddr) (int32, error) { + switch sa := sa.(type) { + case *syscall.SockaddrInet4: + sz := sockaddrInet4ToRaw(rsa, sa) + return sz, nil + case *syscall.SockaddrInet6: + sz := sockaddrInet6ToRaw(rsa, sa) + return sz, nil + default: + return 0, syscall.EWINDOWS + } +} + +// ReadMsg wraps the WSARecvMsg network call. +func (fd *FD) ReadMsg(p []byte, oob []byte, flags int) (int, int, int, syscall.Sockaddr, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, nil, err + } + defer fd.readUnlock() + + if len(p) > maxRW { + p = p[:maxRW] + } + + o := &fd.rop + o.InitMsg(p, oob) + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = int32(unsafe.Sizeof(*o.rsa)) + o.msg.Flags = uint32(flags) + n, err := execIO(o, func(o *operation) error { + return windows.WSARecvMsg(o.fd.Sysfd, &o.msg, &o.qty, &o.o, nil) + }) + err = fd.eofError(n, err) + var sa syscall.Sockaddr + if err == nil { + sa, err = o.rsa.Sockaddr() + } + return n, int(o.msg.Control.Len), int(o.msg.Flags), sa, err +} + +// ReadMsgInet4 is ReadMsg, but specialized to return a syscall.SockaddrInet4. +func (fd *FD) ReadMsgInet4(p []byte, oob []byte, flags int, sa4 *syscall.SockaddrInet4) (int, int, int, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, err + } + defer fd.readUnlock() + + if len(p) > maxRW { + p = p[:maxRW] + } + + o := &fd.rop + o.InitMsg(p, oob) + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = int32(unsafe.Sizeof(*o.rsa)) + o.msg.Flags = uint32(flags) + n, err := execIO(o, func(o *operation) error { + return windows.WSARecvMsg(o.fd.Sysfd, &o.msg, &o.qty, &o.o, nil) + }) + err = fd.eofError(n, err) + if err == nil { + rawToSockaddrInet4(o.rsa, sa4) + } + return n, int(o.msg.Control.Len), int(o.msg.Flags), err +} + +// ReadMsgInet6 is ReadMsg, but specialized to return a syscall.SockaddrInet6. +func (fd *FD) ReadMsgInet6(p []byte, oob []byte, flags int, sa6 *syscall.SockaddrInet6) (int, int, int, error) { + if err := fd.readLock(); err != nil { + return 0, 0, 0, err + } + defer fd.readUnlock() + + if len(p) > maxRW { + p = p[:maxRW] + } + + o := &fd.rop + o.InitMsg(p, oob) + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = int32(unsafe.Sizeof(*o.rsa)) + o.msg.Flags = uint32(flags) + n, err := execIO(o, func(o *operation) error { + return windows.WSARecvMsg(o.fd.Sysfd, &o.msg, &o.qty, &o.o, nil) + }) + err = fd.eofError(n, err) + if err == nil { + rawToSockaddrInet6(o.rsa, sa6) + } + return n, int(o.msg.Control.Len), int(o.msg.Flags), err +} + +// WriteMsg wraps the WSASendMsg network call. +func (fd *FD) WriteMsg(p []byte, oob []byte, sa syscall.Sockaddr) (int, int, error) { + if len(p) > maxRW { + return 0, 0, errors.New("packet is too large (only 1GB is allowed)") + } + + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + + o := &fd.wop + o.InitMsg(p, oob) + if sa != nil { + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + len, err := sockaddrToRaw(o.rsa, sa) + if err != nil { + return 0, 0, err + } + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = len + } + n, err := execIO(o, func(o *operation) error { + return windows.WSASendMsg(o.fd.Sysfd, &o.msg, 0, &o.qty, &o.o, nil) + }) + return n, int(o.msg.Control.Len), err +} + +// WriteMsgInet4 is WriteMsg specialized for syscall.SockaddrInet4. +func (fd *FD) WriteMsgInet4(p []byte, oob []byte, sa *syscall.SockaddrInet4) (int, int, error) { + if len(p) > maxRW { + return 0, 0, errors.New("packet is too large (only 1GB is allowed)") + } + + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + + o := &fd.wop + o.InitMsg(p, oob) + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + len := sockaddrInet4ToRaw(o.rsa, sa) + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = len + n, err := execIO(o, func(o *operation) error { + return windows.WSASendMsg(o.fd.Sysfd, &o.msg, 0, &o.qty, &o.o, nil) + }) + return n, int(o.msg.Control.Len), err +} + +// WriteMsgInet6 is WriteMsg specialized for syscall.SockaddrInet6. +func (fd *FD) WriteMsgInet6(p []byte, oob []byte, sa *syscall.SockaddrInet6) (int, int, error) { + if len(p) > maxRW { + return 0, 0, errors.New("packet is too large (only 1GB is allowed)") + } + + if err := fd.writeLock(); err != nil { + return 0, 0, err + } + defer fd.writeUnlock() + + o := &fd.wop + o.InitMsg(p, oob) + if o.rsa == nil { + o.rsa = new(syscall.RawSockaddrAny) + } + len := sockaddrInet6ToRaw(o.rsa, sa) + o.msg.Name = (syscall.Pointer)(unsafe.Pointer(o.rsa)) + o.msg.Namelen = len + n, err := execIO(o, func(o *operation) error { + return windows.WSASendMsg(o.fd.Sysfd, &o.msg, 0, &o.qty, &o.o, nil) + }) + return n, int(o.msg.Control.Len), err +} diff --git a/src/internal/poll/fd_windows_test.go b/src/internal/poll/fd_windows_test.go new file mode 100644 index 0000000..f0697a0 --- /dev/null +++ b/src/internal/poll/fd_windows_test.go @@ -0,0 +1,198 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "errors" + "fmt" + "internal/poll" + "internal/syscall/windows" + "os" + "sync" + "syscall" + "testing" + "unsafe" +) + +type loggedFD struct { + Net string + FD *poll.FD + Err error +} + +var ( + logMu sync.Mutex + loggedFDs map[syscall.Handle]*loggedFD +) + +func logFD(net string, fd *poll.FD, err error) { + logMu.Lock() + defer logMu.Unlock() + + loggedFDs[fd.Sysfd] = &loggedFD{ + Net: net, + FD: fd, + Err: err, + } +} + +func init() { + loggedFDs = make(map[syscall.Handle]*loggedFD) + *poll.LogInitFD = logFD +} + +func findLoggedFD(h syscall.Handle) (lfd *loggedFD, found bool) { + logMu.Lock() + defer logMu.Unlock() + + lfd, found = loggedFDs[h] + return lfd, found +} + +// checkFileIsNotPartOfNetpoll verifies that f is not managed by netpoll. +// It returns error, if check fails. +func checkFileIsNotPartOfNetpoll(f *os.File) error { + lfd, found := findLoggedFD(syscall.Handle(f.Fd())) + if !found { + return fmt.Errorf("%v fd=%v: is not found in the log", f.Name(), f.Fd()) + } + if lfd.FD.IsPartOfNetpoll() { + return fmt.Errorf("%v fd=%v: is part of netpoll, but should not be (logged: net=%v err=%v)", f.Name(), f.Fd(), lfd.Net, lfd.Err) + } + return nil +} + +func TestFileFdsAreInitialised(t *testing.T) { + exe, err := os.Executable() + if err != nil { + t.Fatal(err) + } + f, err := os.Open(exe) + if err != nil { + t.Fatal(err) + } + defer f.Close() + + err = checkFileIsNotPartOfNetpoll(f) + if err != nil { + t.Fatal(err) + } +} + +func TestSerialFdsAreInitialised(t *testing.T) { + for _, name := range []string{"COM1", "COM2", "COM3", "COM4"} { + t.Run(name, func(t *testing.T) { + h, err := syscall.CreateFile(syscall.StringToUTF16Ptr(name), + syscall.GENERIC_READ|syscall.GENERIC_WRITE, + 0, + nil, + syscall.OPEN_EXISTING, + syscall.FILE_ATTRIBUTE_NORMAL|syscall.FILE_FLAG_OVERLAPPED, + 0) + if err != nil { + if errno, ok := err.(syscall.Errno); ok { + switch errno { + case syscall.ERROR_FILE_NOT_FOUND, + syscall.ERROR_ACCESS_DENIED: + t.Log("Skipping: ", err) + return + } + } + t.Fatal(err) + } + f := os.NewFile(uintptr(h), name) + defer f.Close() + + err = checkFileIsNotPartOfNetpoll(f) + if err != nil { + t.Fatal(err) + } + }) + } +} + +func TestWSASocketConflict(t *testing.T) { + s, err := windows.WSASocket(syscall.AF_INET, syscall.SOCK_STREAM, syscall.IPPROTO_TCP, nil, 0, windows.WSA_FLAG_OVERLAPPED) + if err != nil { + t.Fatal(err) + } + fd := poll.FD{Sysfd: s, IsStream: true, ZeroReadIsEOF: true} + _, err = fd.Init("tcp", true) + if err != nil { + syscall.CloseHandle(s) + t.Fatal(err) + } + defer fd.Close() + + const SIO_TCP_INFO = syscall.IOC_INOUT | syscall.IOC_VENDOR | 39 + inbuf := uint32(0) + var outbuf _TCP_INFO_v0 + cbbr := uint32(0) + + var ovs []syscall.Overlapped = make([]syscall.Overlapped, 2) + // Attempt to exercise behavior where a user-owned syscall.Overlapped + // induces an invalid pointer dereference in the Windows-specific version + // of runtime.netpoll. + ovs[1].Internal -= 1 + + // Create an event so that we can efficiently wait for completion + // of a requested overlapped I/O operation. + ovs[0].HEvent, _ = windows.CreateEvent(nil, 0, 0, nil) + if ovs[0].HEvent == 0 { + t.Fatalf("could not create the event!") + } + + // Set the low bit of the Event Handle so that the completion + // of the overlapped I/O event will not trigger a completion event + // on any I/O completion port associated with the handle. + ovs[0].HEvent |= 0x1 + + if err = fd.WSAIoctl( + SIO_TCP_INFO, + (*byte)(unsafe.Pointer(&inbuf)), + uint32(unsafe.Sizeof(inbuf)), + (*byte)(unsafe.Pointer(&outbuf)), + uint32(unsafe.Sizeof(outbuf)), + &cbbr, + &ovs[0], + 0, + ); err != nil && !errors.Is(err, syscall.ERROR_IO_PENDING) { + t.Fatalf("could not perform the WSAIoctl: %v", err) + } + + if err != nil && errors.Is(err, syscall.ERROR_IO_PENDING) { + // It is possible that the overlapped I/O operation completed + // immediately so there is no need to wait for it to complete. + if res, err := syscall.WaitForSingleObject(ovs[0].HEvent, syscall.INFINITE); res != 0 { + t.Fatalf("waiting for the completion of the overlapped IO failed: %v", err) + } + } + + if err = syscall.CloseHandle(ovs[0].HEvent); err != nil { + t.Fatalf("could not close the event handle: %v", err) + } +} + +type _TCP_INFO_v0 struct { + State uint32 + Mss uint32 + ConnectionTimeMs uint64 + TimestampsEnabled bool + RttUs uint32 + MinRttUs uint32 + BytesInFlight uint32 + Cwnd uint32 + SndWnd uint32 + RcvWnd uint32 + RcvBuf uint32 + BytesOut uint64 + BytesIn uint64 + BytesReordered uint32 + BytesRetrans uint32 + FastRetrans uint32 + DupAcksIn uint32 + TimeoutEpisodes uint32 + SynRetrans uint8 +} diff --git a/src/internal/poll/fd_writev_libc.go b/src/internal/poll/fd_writev_libc.go new file mode 100644 index 0000000..0a60473 --- /dev/null +++ b/src/internal/poll/fd_writev_libc.go @@ -0,0 +1,15 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build aix || darwin || (openbsd && !mips64) || solaris + +package poll + +import ( + "syscall" + _ "unsafe" // for go:linkname +) + +//go:linkname writev syscall.writev +func writev(fd int, iovecs []syscall.Iovec) (uintptr, error) diff --git a/src/internal/poll/fd_writev_unix.go b/src/internal/poll/fd_writev_unix.go new file mode 100644 index 0000000..005638b --- /dev/null +++ b/src/internal/poll/fd_writev_unix.go @@ -0,0 +1,29 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build dragonfly || freebsd || linux || netbsd || (openbsd && mips64) + +package poll + +import ( + "syscall" + "unsafe" +) + +func writev(fd int, iovecs []syscall.Iovec) (uintptr, error) { + var ( + r uintptr + e syscall.Errno + ) + for { + r, _, e = syscall.Syscall(syscall.SYS_WRITEV, uintptr(fd), uintptr(unsafe.Pointer(&iovecs[0])), uintptr(len(iovecs))) + if e != syscall.EINTR { + break + } + } + if e != 0 { + return r, e + } + return r, nil +} diff --git a/src/internal/poll/file_plan9.go b/src/internal/poll/file_plan9.go new file mode 100644 index 0000000..57dc0c6 --- /dev/null +++ b/src/internal/poll/file_plan9.go @@ -0,0 +1,42 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +// Expose fdMutex for use by the os package on Plan 9. +// On Plan 9 we don't want to use async I/O for file operations, +// but we still want the locking semantics that fdMutex provides. + +// FDMutex is an exported fdMutex, only for Plan 9. +type FDMutex struct { + fdmu fdMutex +} + +func (fdmu *FDMutex) Incref() bool { + return fdmu.fdmu.incref() +} + +func (fdmu *FDMutex) Decref() bool { + return fdmu.fdmu.decref() +} + +func (fdmu *FDMutex) IncrefAndClose() bool { + return fdmu.fdmu.increfAndClose() +} + +func (fdmu *FDMutex) ReadLock() bool { + return fdmu.fdmu.rwlock(true) +} + +func (fdmu *FDMutex) ReadUnlock() bool { + return fdmu.fdmu.rwunlock(true) +} + +func (fdmu *FDMutex) WriteLock() bool { + return fdmu.fdmu.rwlock(false) +} + +func (fdmu *FDMutex) WriteUnlock() bool { + return fdmu.fdmu.rwunlock(false) +} diff --git a/src/internal/poll/hook_cloexec.go b/src/internal/poll/hook_cloexec.go new file mode 100644 index 0000000..5b3cdce --- /dev/null +++ b/src/internal/poll/hook_cloexec.go @@ -0,0 +1,12 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build dragonfly || freebsd || linux || netbsd || openbsd || solaris + +package poll + +import "syscall" + +// Accept4Func is used to hook the accept4 call. +var Accept4Func func(int, int) (int, syscall.Sockaddr, error) = syscall.Accept4 diff --git a/src/internal/poll/hook_unix.go b/src/internal/poll/hook_unix.go new file mode 100644 index 0000000..b3f4f9e --- /dev/null +++ b/src/internal/poll/hook_unix.go @@ -0,0 +1,15 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || (js && wasm) || wasip1 + +package poll + +import "syscall" + +// CloseFunc is used to hook the close call. +var CloseFunc func(int) error = syscall.Close + +// AcceptFunc is used to hook the accept call. +var AcceptFunc func(int) (int, syscall.Sockaddr, error) = syscall.Accept diff --git a/src/internal/poll/hook_windows.go b/src/internal/poll/hook_windows.go new file mode 100644 index 0000000..0bd950e --- /dev/null +++ b/src/internal/poll/hook_windows.go @@ -0,0 +1,16 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// CloseFunc is used to hook the close call. +var CloseFunc func(syscall.Handle) error = syscall.Closesocket + +// AcceptFunc is used to hook the accept call. +var AcceptFunc func(syscall.Handle, syscall.Handle, *byte, uint32, uint32, uint32, *uint32, *syscall.Overlapped) error = syscall.AcceptEx + +// ConnectExFunc is used to hook the ConnectEx call. +var ConnectExFunc func(syscall.Handle, syscall.Sockaddr, *byte, uint32, *uint32, *syscall.Overlapped) error = syscall.ConnectEx diff --git a/src/internal/poll/iovec_solaris.go b/src/internal/poll/iovec_solaris.go new file mode 100644 index 0000000..e68f833 --- /dev/null +++ b/src/internal/poll/iovec_solaris.go @@ -0,0 +1,14 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "syscall" + "unsafe" +) + +func newIovecWithBase(base *byte) syscall.Iovec { + return syscall.Iovec{Base: (*int8)(unsafe.Pointer(base))} +} diff --git a/src/internal/poll/iovec_unix.go b/src/internal/poll/iovec_unix.go new file mode 100644 index 0000000..3f2833e --- /dev/null +++ b/src/internal/poll/iovec_unix.go @@ -0,0 +1,13 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build aix || darwin || dragonfly || freebsd || linux || netbsd || openbsd + +package poll + +import "syscall" + +func newIovecWithBase(base *byte) syscall.Iovec { + return syscall.Iovec{Base: base} +} diff --git a/src/internal/poll/read_test.go b/src/internal/poll/read_test.go new file mode 100644 index 0000000..598a52e --- /dev/null +++ b/src/internal/poll/read_test.go @@ -0,0 +1,61 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "os" + "runtime" + "sync" + "testing" + "time" +) + +func TestRead(t *testing.T) { + t.Run("SpecialFile", func(t *testing.T) { + var wg sync.WaitGroup + for _, p := range specialFiles() { + for i := 0; i < 4; i++ { + wg.Add(1) + go func(p string) { + defer wg.Done() + for i := 0; i < 100; i++ { + if _, err := os.ReadFile(p); err != nil { + t.Error(err) + return + } + time.Sleep(time.Nanosecond) + } + }(p) + } + } + wg.Wait() + }) +} + +func specialFiles() []string { + var ps []string + switch runtime.GOOS { + case "darwin", "ios", "dragonfly", "freebsd", "netbsd", "openbsd": + ps = []string{ + "/dev/null", + } + case "linux": + ps = []string{ + "/dev/null", + "/proc/stat", + "/sys/devices/system/cpu/online", + } + } + nps := ps[:0] + for _, p := range ps { + f, err := os.Open(p) + if err != nil { + continue + } + f.Close() + nps = append(nps, p) + } + return nps +} diff --git a/src/internal/poll/sendfile_bsd.go b/src/internal/poll/sendfile_bsd.go new file mode 100644 index 0000000..89315a8 --- /dev/null +++ b/src/internal/poll/sendfile_bsd.go @@ -0,0 +1,59 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build darwin || dragonfly || freebsd + +package poll + +import "syscall" + +// maxSendfileSize is the largest chunk size we ask the kernel to copy +// at a time. +const maxSendfileSize int = 4 << 20 + +// SendFile wraps the sendfile system call. +func SendFile(dstFD *FD, src int, pos, remain int64) (int64, error) { + if err := dstFD.writeLock(); err != nil { + return 0, err + } + defer dstFD.writeUnlock() + if err := dstFD.pd.prepareWrite(dstFD.isFile); err != nil { + return 0, err + } + + dst := dstFD.Sysfd + var written int64 + var err error + for remain > 0 { + n := maxSendfileSize + if int64(n) > remain { + n = int(remain) + } + pos1 := pos + n, err1 := syscall.Sendfile(dst, src, &pos1, n) + if n > 0 { + pos += int64(n) + written += int64(n) + remain -= int64(n) + } else if n == 0 && err1 == nil { + break + } + if err1 == syscall.EINTR { + continue + } + if err1 == syscall.EAGAIN { + if err1 = dstFD.pd.waitWrite(dstFD.isFile); err1 == nil { + continue + } + } + if err1 != nil { + // This includes syscall.ENOSYS (no kernel + // support) and syscall.EINVAL (fd types which + // don't implement sendfile) + err = err1 + break + } + } + return written, err +} diff --git a/src/internal/poll/sendfile_linux.go b/src/internal/poll/sendfile_linux.go new file mode 100644 index 0000000..cc31969 --- /dev/null +++ b/src/internal/poll/sendfile_linux.go @@ -0,0 +1,59 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// maxSendfileSize is the largest chunk size we ask the kernel to copy +// at a time. +const maxSendfileSize int = 4 << 20 + +// SendFile wraps the sendfile system call. +func SendFile(dstFD *FD, src int, remain int64) (int64, error, bool) { + if err := dstFD.writeLock(); err != nil { + return 0, err, false + } + defer dstFD.writeUnlock() + if err := dstFD.pd.prepareWrite(dstFD.isFile); err != nil { + return 0, err, false + } + + dst := dstFD.Sysfd + var ( + written int64 + err error + handled = true + ) + for remain > 0 { + n := maxSendfileSize + if int64(n) > remain { + n = int(remain) + } + n, err1 := syscall.Sendfile(dst, src, nil, n) + if n > 0 { + written += int64(n) + remain -= int64(n) + } else if n == 0 && err1 == nil { + break + } + if err1 == syscall.EINTR { + continue + } + if err1 == syscall.EAGAIN { + if err1 = dstFD.pd.waitWrite(dstFD.isFile); err1 == nil { + continue + } + } + if err1 != nil { + // This includes syscall.ENOSYS (no kernel + // support) and syscall.EINVAL (fd types which + // don't implement sendfile) + err = err1 + handled = false + break + } + } + return written, err, handled +} diff --git a/src/internal/poll/sendfile_solaris.go b/src/internal/poll/sendfile_solaris.go new file mode 100644 index 0000000..7ae18f4 --- /dev/null +++ b/src/internal/poll/sendfile_solaris.go @@ -0,0 +1,66 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// Not strictly needed, but very helpful for debugging, see issue #10221. +// +//go:cgo_import_dynamic _ _ "libsendfile.so" +//go:cgo_import_dynamic _ _ "libsocket.so" + +// maxSendfileSize is the largest chunk size we ask the kernel to copy +// at a time. +const maxSendfileSize int = 4 << 20 + +// SendFile wraps the sendfile system call. +func SendFile(dstFD *FD, src int, pos, remain int64) (int64, error) { + if err := dstFD.writeLock(); err != nil { + return 0, err + } + defer dstFD.writeUnlock() + if err := dstFD.pd.prepareWrite(dstFD.isFile); err != nil { + return 0, err + } + + dst := dstFD.Sysfd + var written int64 + var err error + for remain > 0 { + n := maxSendfileSize + if int64(n) > remain { + n = int(remain) + } + pos1 := pos + n, err1 := syscall.Sendfile(dst, src, &pos1, n) + if err1 == syscall.EAGAIN || err1 == syscall.EINTR { + // partial write may have occurred + n = int(pos1 - pos) + } + if n > 0 { + pos += int64(n) + written += int64(n) + remain -= int64(n) + } else if n == 0 && err1 == nil { + break + } + if err1 == syscall.EAGAIN { + if err1 = dstFD.pd.waitWrite(dstFD.isFile); err1 == nil { + continue + } + } + if err1 == syscall.EINTR { + continue + } + if err1 != nil { + // This includes syscall.ENOSYS (no kernel + // support) and syscall.EINVAL (fd types which + // don't implement sendfile) + err = err1 + break + } + } + return written, err +} diff --git a/src/internal/poll/sendfile_windows.go b/src/internal/poll/sendfile_windows.go new file mode 100644 index 0000000..8c3353b --- /dev/null +++ b/src/internal/poll/sendfile_windows.go @@ -0,0 +1,84 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "io" + "syscall" +) + +// SendFile wraps the TransmitFile call. +func SendFile(fd *FD, src syscall.Handle, n int64) (written int64, err error) { + if fd.kind == kindPipe { + // TransmitFile does not work with pipes + return 0, syscall.ESPIPE + } + if ft, _ := syscall.GetFileType(src); ft == syscall.FILE_TYPE_PIPE { + return 0, syscall.ESPIPE + } + + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + + o := &fd.wop + o.handle = src + + // TODO(brainman): skip calling syscall.Seek if OS allows it + curpos, err := syscall.Seek(o.handle, 0, io.SeekCurrent) + if err != nil { + return 0, err + } + + if n <= 0 { // We don't know the size of the file so infer it. + // Find the number of bytes offset from curpos until the end of the file. + n, err = syscall.Seek(o.handle, -curpos, io.SeekEnd) + if err != nil { + return + } + // Now seek back to the original position. + if _, err = syscall.Seek(o.handle, curpos, io.SeekStart); err != nil { + return + } + } + + // TransmitFile can be invoked in one call with at most + // 2,147,483,646 bytes: the maximum value for a 32-bit integer minus 1. + // See https://docs.microsoft.com/en-us/windows/win32/api/mswsock/nf-mswsock-transmitfile + const maxChunkSizePerCall = int64(0x7fffffff - 1) + + for n > 0 { + chunkSize := maxChunkSizePerCall + if chunkSize > n { + chunkSize = n + } + + o.qty = uint32(chunkSize) + o.o.Offset = uint32(curpos) + o.o.OffsetHigh = uint32(curpos >> 32) + + nw, err := execIO(o, func(o *operation) error { + return syscall.TransmitFile(o.fd.Sysfd, o.handle, o.qty, 0, &o.o, nil, syscall.TF_WRITE_BEHIND) + }) + if err != nil { + return written, err + } + + curpos += int64(nw) + + // Some versions of Windows (Windows 10 1803) do not set + // file position after TransmitFile completes. + // So just use Seek to set file position. + if _, err = syscall.Seek(o.handle, curpos, io.SeekStart); err != nil { + return written, err + } + + n -= int64(nw) + written += int64(nw) + } + + return +} diff --git a/src/internal/poll/sock_cloexec.go b/src/internal/poll/sock_cloexec.go new file mode 100644 index 0000000..361c11b --- /dev/null +++ b/src/internal/poll/sock_cloexec.go @@ -0,0 +1,49 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file implements accept for platforms that provide a fast path for +// setting SetNonblock and CloseOnExec. + +//go:build dragonfly || freebsd || (linux && !arm) || netbsd || openbsd || solaris + +package poll + +import "syscall" + +// Wrapper around the accept system call that marks the returned file +// descriptor as nonblocking and close-on-exec. +func accept(s int) (int, syscall.Sockaddr, string, error) { + ns, sa, err := Accept4Func(s, syscall.SOCK_NONBLOCK|syscall.SOCK_CLOEXEC) + // TODO: We can remove the fallback on Linux and *BSD, + // as currently supported versions all support accept4 + // with SOCK_CLOEXEC, but Solaris does not. See issue #59359. + switch err { + case nil: + return ns, sa, "", nil + default: // errors other than the ones listed + return -1, sa, "accept4", err + case syscall.ENOSYS: // syscall missing + case syscall.EINVAL: // some Linux use this instead of ENOSYS + case syscall.EACCES: // some Linux use this instead of ENOSYS + case syscall.EFAULT: // some Linux use this instead of ENOSYS + } + + // See ../syscall/exec_unix.go for description of ForkLock. + // It is probably okay to hold the lock across syscall.Accept + // because we have put fd.sysfd into non-blocking mode. + // However, a call to the File method will put it back into + // blocking mode. We can't take that risk, so no use of ForkLock here. + ns, sa, err = AcceptFunc(s) + if err == nil { + syscall.CloseOnExec(ns) + } + if err != nil { + return -1, nil, "accept", err + } + if err = syscall.SetNonblock(ns, true); err != nil { + CloseFunc(ns) + return -1, nil, "setnonblock", err + } + return ns, sa, "", nil +} diff --git a/src/internal/poll/sock_cloexec_accept.go b/src/internal/poll/sock_cloexec_accept.go new file mode 100644 index 0000000..4b86de5 --- /dev/null +++ b/src/internal/poll/sock_cloexec_accept.go @@ -0,0 +1,51 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file implements accept for platforms that provide a fast path for +// setting SetNonblock and CloseOnExec, but don't necessarily have accept4. +// This is the code we used for accept in Go 1.17 and earlier. +// On Linux the accept4 system call was introduced in 2.6.28 kernel, +// and our minimum requirement is 2.6.32, so we simplified the function. +// Unfortunately, on ARM accept4 wasn't added until 2.6.36, so for ARM +// only we continue using the older code. + +//go:build linux && arm + +package poll + +import "syscall" + +// Wrapper around the accept system call that marks the returned file +// descriptor as nonblocking and close-on-exec. +func accept(s int) (int, syscall.Sockaddr, string, error) { + ns, sa, err := Accept4Func(s, syscall.SOCK_NONBLOCK|syscall.SOCK_CLOEXEC) + switch err { + case nil: + return ns, sa, "", nil + default: // errors other than the ones listed + return -1, sa, "accept4", err + case syscall.ENOSYS: // syscall missing + case syscall.EINVAL: // some Linux use this instead of ENOSYS + case syscall.EACCES: // some Linux use this instead of ENOSYS + case syscall.EFAULT: // some Linux use this instead of ENOSYS + } + + // See ../syscall/exec_unix.go for description of ForkLock. + // It is probably okay to hold the lock across syscall.Accept + // because we have put fd.sysfd into non-blocking mode. + // However, a call to the File method will put it back into + // blocking mode. We can't take that risk, so no use of ForkLock here. + ns, sa, err = AcceptFunc(s) + if err == nil { + syscall.CloseOnExec(ns) + } + if err != nil { + return -1, nil, "accept", err + } + if err = syscall.SetNonblock(ns, true); err != nil { + CloseFunc(ns) + return -1, nil, "setnonblock", err + } + return ns, sa, "", nil +} diff --git a/src/internal/poll/sockopt.go b/src/internal/poll/sockopt.go new file mode 100644 index 0000000..a87a9e6 --- /dev/null +++ b/src/internal/poll/sockopt.go @@ -0,0 +1,45 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || windows + +package poll + +import "syscall" + +// SetsockoptInt wraps the setsockopt network call with an int argument. +func (fd *FD) SetsockoptInt(level, name, arg int) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptInt(fd.Sysfd, level, name, arg) +} + +// SetsockoptInet4Addr wraps the setsockopt network call with an IPv4 address. +func (fd *FD) SetsockoptInet4Addr(level, name int, arg [4]byte) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptInet4Addr(fd.Sysfd, level, name, arg) +} + +// SetsockoptLinger wraps the setsockopt network call with a Linger argument. +func (fd *FD) SetsockoptLinger(level, name int, l *syscall.Linger) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptLinger(fd.Sysfd, level, name, l) +} + +// GetsockoptInt wraps the getsockopt network call with an int argument. +func (fd *FD) GetsockoptInt(level, name int) (int, error) { + if err := fd.incref(); err != nil { + return -1, err + } + defer fd.decref() + return syscall.GetsockoptInt(fd.Sysfd, level, name) +} diff --git a/src/internal/poll/sockopt_linux.go b/src/internal/poll/sockopt_linux.go new file mode 100644 index 0000000..bc79c35 --- /dev/null +++ b/src/internal/poll/sockopt_linux.go @@ -0,0 +1,16 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// SetsockoptIPMreqn wraps the setsockopt network call with an IPMreqn argument. +func (fd *FD) SetsockoptIPMreqn(level, name int, mreq *syscall.IPMreqn) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptIPMreqn(fd.Sysfd, level, name, mreq) +} diff --git a/src/internal/poll/sockopt_unix.go b/src/internal/poll/sockopt_unix.go new file mode 100644 index 0000000..9cba44d --- /dev/null +++ b/src/internal/poll/sockopt_unix.go @@ -0,0 +1,18 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package poll + +import "syscall" + +// SetsockoptByte wraps the setsockopt network call with a byte argument. +func (fd *FD) SetsockoptByte(level, name int, arg byte) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptByte(fd.Sysfd, level, name, arg) +} diff --git a/src/internal/poll/sockopt_windows.go b/src/internal/poll/sockopt_windows.go new file mode 100644 index 0000000..dd5fb70 --- /dev/null +++ b/src/internal/poll/sockopt_windows.go @@ -0,0 +1,25 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import "syscall" + +// Setsockopt wraps the setsockopt network call. +func (fd *FD) Setsockopt(level, optname int32, optval *byte, optlen int32) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.Setsockopt(fd.Sysfd, level, optname, optval, optlen) +} + +// WSAIoctl wraps the WSAIoctl network call. +func (fd *FD) WSAIoctl(iocc uint32, inbuf *byte, cbif uint32, outbuf *byte, cbob uint32, cbbr *uint32, overlapped *syscall.Overlapped, completionRoutine uintptr) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.WSAIoctl(fd.Sysfd, iocc, inbuf, cbif, outbuf, cbob, cbbr, overlapped, completionRoutine) +} diff --git a/src/internal/poll/sockoptip.go b/src/internal/poll/sockoptip.go new file mode 100644 index 0000000..41955e1 --- /dev/null +++ b/src/internal/poll/sockoptip.go @@ -0,0 +1,27 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix || windows + +package poll + +import "syscall" + +// SetsockoptIPMreq wraps the setsockopt network call with an IPMreq argument. +func (fd *FD) SetsockoptIPMreq(level, name int, mreq *syscall.IPMreq) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptIPMreq(fd.Sysfd, level, name, mreq) +} + +// SetsockoptIPv6Mreq wraps the setsockopt network call with an IPv6Mreq argument. +func (fd *FD) SetsockoptIPv6Mreq(level, name int, mreq *syscall.IPv6Mreq) error { + if err := fd.incref(); err != nil { + return err + } + defer fd.decref() + return syscall.SetsockoptIPv6Mreq(fd.Sysfd, level, name, mreq) +} diff --git a/src/internal/poll/splice_linux.go b/src/internal/poll/splice_linux.go new file mode 100644 index 0000000..72cca34 --- /dev/null +++ b/src/internal/poll/splice_linux.go @@ -0,0 +1,250 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll + +import ( + "internal/syscall/unix" + "runtime" + "sync" + "syscall" + "unsafe" +) + +const ( + // spliceNonblock doesn't make the splice itself necessarily nonblocking + // (because the actual file descriptors that are spliced from/to may block + // unless they have the O_NONBLOCK flag set), but it makes the splice pipe + // operations nonblocking. + spliceNonblock = 0x2 + + // maxSpliceSize is the maximum amount of data Splice asks + // the kernel to move in a single call to splice(2). + // We use 1MB as Splice writes data through a pipe, and 1MB is the default maximum pipe buffer size, + // which is determined by /proc/sys/fs/pipe-max-size. + maxSpliceSize = 1 << 20 +) + +// Splice transfers at most remain bytes of data from src to dst, using the +// splice system call to minimize copies of data from and to userspace. +// +// Splice gets a pipe buffer from the pool or creates a new one if needed, to serve as a buffer for the data transfer. +// src and dst must both be stream-oriented sockets. +// +// If err != nil, sc is the system call which caused the error. +func Splice(dst, src *FD, remain int64) (written int64, handled bool, sc string, err error) { + p, sc, err := getPipe() + if err != nil { + return 0, false, sc, err + } + defer putPipe(p) + var inPipe, n int + for err == nil && remain > 0 { + max := maxSpliceSize + if int64(max) > remain { + max = int(remain) + } + inPipe, err = spliceDrain(p.wfd, src, max) + // The operation is considered handled if splice returns no + // error, or an error other than EINVAL. An EINVAL means the + // kernel does not support splice for the socket type of src. + // The failed syscall does not consume any data so it is safe + // to fall back to a generic copy. + // + // spliceDrain should never return EAGAIN, so if err != nil, + // Splice cannot continue. + // + // If inPipe == 0 && err == nil, src is at EOF, and the + // transfer is complete. + handled = handled || (err != syscall.EINVAL) + if err != nil || inPipe == 0 { + break + } + p.data += inPipe + + n, err = splicePump(dst, p.rfd, inPipe) + if n > 0 { + written += int64(n) + remain -= int64(n) + p.data -= n + } + } + if err != nil { + return written, handled, "splice", err + } + return written, true, "", nil +} + +// spliceDrain moves data from a socket to a pipe. +// +// Invariant: when entering spliceDrain, the pipe is empty. It is either in its +// initial state, or splicePump has emptied it previously. +// +// Given this, spliceDrain can reasonably assume that the pipe is ready for +// writing, so if splice returns EAGAIN, it must be because the socket is not +// ready for reading. +// +// If spliceDrain returns (0, nil), src is at EOF. +func spliceDrain(pipefd int, sock *FD, max int) (int, error) { + if err := sock.readLock(); err != nil { + return 0, err + } + defer sock.readUnlock() + if err := sock.pd.prepareRead(sock.isFile); err != nil { + return 0, err + } + for { + // In theory calling splice(2) with SPLICE_F_NONBLOCK could end up an infinite loop here, + // because it could return EAGAIN ceaselessly when the write end of the pipe is full, + // but this shouldn't be a concern here, since the pipe buffer must be sufficient for + // this data transmission on the basis of the workflow in Splice. + n, err := splice(pipefd, sock.Sysfd, max, spliceNonblock) + if err == syscall.EINTR { + continue + } + if err != syscall.EAGAIN { + return n, err + } + if sock.pd.pollable() { + if err := sock.pd.waitRead(sock.isFile); err != nil { + return n, err + } + } + } +} + +// splicePump moves all the buffered data from a pipe to a socket. +// +// Invariant: when entering splicePump, there are exactly inPipe +// bytes of data in the pipe, from a previous call to spliceDrain. +// +// By analogy to the condition from spliceDrain, splicePump +// only needs to poll the socket for readiness, if splice returns +// EAGAIN. +// +// If splicePump cannot move all the data in a single call to +// splice(2), it loops over the buffered data until it has written +// all of it to the socket. This behavior is similar to the Write +// step of an io.Copy in userspace. +func splicePump(sock *FD, pipefd int, inPipe int) (int, error) { + if err := sock.writeLock(); err != nil { + return 0, err + } + defer sock.writeUnlock() + if err := sock.pd.prepareWrite(sock.isFile); err != nil { + return 0, err + } + written := 0 + for inPipe > 0 { + // In theory calling splice(2) with SPLICE_F_NONBLOCK could end up an infinite loop here, + // because it could return EAGAIN ceaselessly when the read end of the pipe is empty, + // but this shouldn't be a concern here, since the pipe buffer must contain inPipe size of + // data on the basis of the workflow in Splice. + n, err := splice(sock.Sysfd, pipefd, inPipe, spliceNonblock) + if err == syscall.EINTR { + continue + } + // Here, the condition n == 0 && err == nil should never be + // observed, since Splice controls the write side of the pipe. + if n > 0 { + inPipe -= n + written += n + continue + } + if err != syscall.EAGAIN { + return written, err + } + if sock.pd.pollable() { + if err := sock.pd.waitWrite(sock.isFile); err != nil { + return written, err + } + } + } + return written, nil +} + +// splice wraps the splice system call. Since the current implementation +// only uses splice on sockets and pipes, the offset arguments are unused. +// splice returns int instead of int64, because callers never ask it to +// move more data in a single call than can fit in an int32. +func splice(out int, in int, max int, flags int) (int, error) { + n, err := syscall.Splice(in, nil, out, nil, max, flags) + return int(n), err +} + +type splicePipeFields struct { + rfd int + wfd int + data int +} + +type splicePipe struct { + splicePipeFields + + // We want to use a finalizer, so ensure that the size is + // large enough to not use the tiny allocator. + _ [24 - unsafe.Sizeof(splicePipeFields{})%24]byte +} + +// splicePipePool caches pipes to avoid high-frequency construction and destruction of pipe buffers. +// The garbage collector will free all pipes in the sync.Pool periodically, thus we need to set up +// a finalizer for each pipe to close its file descriptors before the actual GC. +var splicePipePool = sync.Pool{New: newPoolPipe} + +func newPoolPipe() any { + // Discard the error which occurred during the creation of pipe buffer, + // redirecting the data transmission to the conventional way utilizing read() + write() as a fallback. + p := newPipe() + if p == nil { + return nil + } + runtime.SetFinalizer(p, destroyPipe) + return p +} + +// getPipe tries to acquire a pipe buffer from the pool or create a new one with newPipe() if it gets nil from the cache. +// +// Note that it may fail to create a new pipe buffer by newPipe(), in which case getPipe() will return a generic error +// and system call name splice in a string as the indication. +func getPipe() (*splicePipe, string, error) { + v := splicePipePool.Get() + if v == nil { + return nil, "splice", syscall.EINVAL + } + return v.(*splicePipe), "", nil +} + +func putPipe(p *splicePipe) { + // If there is still data left in the pipe, + // then close and discard it instead of putting it back into the pool. + if p.data != 0 { + runtime.SetFinalizer(p, nil) + destroyPipe(p) + return + } + splicePipePool.Put(p) +} + +// newPipe sets up a pipe for a splice operation. +func newPipe() *splicePipe { + var fds [2]int + if err := syscall.Pipe2(fds[:], syscall.O_CLOEXEC|syscall.O_NONBLOCK); err != nil { + return nil + } + + // Splice will loop writing maxSpliceSize bytes from the source to the pipe, + // and then write those bytes from the pipe to the destination. + // Set the pipe buffer size to maxSpliceSize to optimize that. + // Ignore errors here, as a smaller buffer size will work, + // although it will require more system calls. + unix.Fcntl(fds[0], syscall.F_SETPIPE_SZ, maxSpliceSize) + + return &splicePipe{splicePipeFields: splicePipeFields{rfd: fds[0], wfd: fds[1]}} +} + +// destroyPipe destroys a pipe. +func destroyPipe(p *splicePipe) { + CloseFunc(p.rfd) + CloseFunc(p.wfd) +} diff --git a/src/internal/poll/splice_linux_test.go b/src/internal/poll/splice_linux_test.go new file mode 100644 index 0000000..29bcaab --- /dev/null +++ b/src/internal/poll/splice_linux_test.go @@ -0,0 +1,136 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "internal/poll" + "runtime" + "sync" + "sync/atomic" + "testing" + "time" +) + +var closeHook atomic.Value // func(fd int) + +func init() { + closeFunc := poll.CloseFunc + poll.CloseFunc = func(fd int) (err error) { + if v := closeHook.Load(); v != nil { + if hook := v.(func(int)); hook != nil { + hook(fd) + } + } + return closeFunc(fd) + } +} + +func TestSplicePipePool(t *testing.T) { + const N = 64 + var ( + p *poll.SplicePipe + ps []*poll.SplicePipe + allFDs []int + pendingFDs sync.Map // fd → struct{}{} + err error + ) + + closeHook.Store(func(fd int) { pendingFDs.Delete(fd) }) + t.Cleanup(func() { closeHook.Store((func(int))(nil)) }) + + for i := 0; i < N; i++ { + p, _, err = poll.GetPipe() + if err != nil { + t.Skipf("failed to create pipe due to error(%v), skip this test", err) + } + _, pwfd := poll.GetPipeFds(p) + allFDs = append(allFDs, pwfd) + pendingFDs.Store(pwfd, struct{}{}) + ps = append(ps, p) + } + for _, p = range ps { + poll.PutPipe(p) + } + ps = nil + p = nil + + // Exploit the timeout of "go test" as a timer for the subsequent verification. + timeout := 5 * time.Minute + if deadline, ok := t.Deadline(); ok { + timeout = deadline.Sub(time.Now()) + timeout -= timeout / 10 // Leave 10% headroom for cleanup. + } + expiredTime := time.NewTimer(timeout) + defer expiredTime.Stop() + + // Trigger garbage collection repeatedly, waiting for all pipes in sync.Pool + // to either be deallocated and closed, or to time out. + for { + runtime.GC() + time.Sleep(10 * time.Millisecond) + + // Detect whether all pipes are closed properly. + var leakedFDs []int + pendingFDs.Range(func(k, v any) bool { + leakedFDs = append(leakedFDs, k.(int)) + return true + }) + if len(leakedFDs) == 0 { + break + } + + select { + case <-expiredTime.C: + t.Logf("all descriptors: %v", allFDs) + t.Fatalf("leaked descriptors: %v", leakedFDs) + default: + } + } +} + +func BenchmarkSplicePipe(b *testing.B) { + b.Run("SplicePipeWithPool", func(b *testing.B) { + for i := 0; i < b.N; i++ { + p, _, err := poll.GetPipe() + if err != nil { + continue + } + poll.PutPipe(p) + } + }) + b.Run("SplicePipeWithoutPool", func(b *testing.B) { + for i := 0; i < b.N; i++ { + p := poll.NewPipe() + if p == nil { + b.Skip("newPipe returned nil") + } + poll.DestroyPipe(p) + } + }) +} + +func BenchmarkSplicePipePoolParallel(b *testing.B) { + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + p, _, err := poll.GetPipe() + if err != nil { + continue + } + poll.PutPipe(p) + } + }) +} + +func BenchmarkSplicePipeNativeParallel(b *testing.B) { + b.RunParallel(func(pb *testing.PB) { + for pb.Next() { + p := poll.NewPipe() + if p == nil { + b.Skip("newPipe returned nil") + } + poll.DestroyPipe(p) + } + }) +} diff --git a/src/internal/poll/strconv.go b/src/internal/poll/strconv.go new file mode 100644 index 0000000..2b052fa --- /dev/null +++ b/src/internal/poll/strconv.go @@ -0,0 +1,13 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build plan9 + +package poll + +// stringsHasSuffix is strings.HasSuffix. It reports whether s ends in +// suffix. +func stringsHasSuffix(s, suffix string) bool { + return len(s) >= len(suffix) && s[len(s)-len(suffix):] == suffix +} diff --git a/src/internal/poll/sys_cloexec.go b/src/internal/poll/sys_cloexec.go new file mode 100644 index 0000000..2c5da7d --- /dev/null +++ b/src/internal/poll/sys_cloexec.go @@ -0,0 +1,36 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file implements accept for platforms that do not provide a fast path for +// setting SetNonblock and CloseOnExec. + +//go:build aix || darwin || (js && wasm) || wasip1 + +package poll + +import ( + "syscall" +) + +// Wrapper around the accept system call that marks the returned file +// descriptor as nonblocking and close-on-exec. +func accept(s int) (int, syscall.Sockaddr, string, error) { + // See ../syscall/exec_unix.go for description of ForkLock. + // It is probably okay to hold the lock across syscall.Accept + // because we have put fd.sysfd into non-blocking mode. + // However, a call to the File method will put it back into + // blocking mode. We can't take that risk, so no use of ForkLock here. + ns, sa, err := AcceptFunc(s) + if err == nil { + syscall.CloseOnExec(ns) + } + if err != nil { + return -1, nil, "accept", err + } + if err = syscall.SetNonblock(ns, true); err != nil { + CloseFunc(ns) + return -1, nil, "setnonblock", err + } + return ns, sa, "", nil +} diff --git a/src/internal/poll/writev.go b/src/internal/poll/writev.go new file mode 100644 index 0000000..75c8b64 --- /dev/null +++ b/src/internal/poll/writev.go @@ -0,0 +1,92 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package poll + +import ( + "io" + "runtime" + "syscall" +) + +// Writev wraps the writev system call. +func (fd *FD) Writev(v *[][]byte) (int64, error) { + if err := fd.writeLock(); err != nil { + return 0, err + } + defer fd.writeUnlock() + if err := fd.pd.prepareWrite(fd.isFile); err != nil { + return 0, err + } + + var iovecs []syscall.Iovec + if fd.iovecs != nil { + iovecs = *fd.iovecs + } + // TODO: read from sysconf(_SC_IOV_MAX)? The Linux default is + // 1024 and this seems conservative enough for now. Darwin's + // UIO_MAXIOV also seems to be 1024. + maxVec := 1024 + if runtime.GOOS == "aix" || runtime.GOOS == "solaris" { + // IOV_MAX is set to XOPEN_IOV_MAX on AIX and Solaris. + maxVec = 16 + } + + var n int64 + var err error + for len(*v) > 0 { + iovecs = iovecs[:0] + for _, chunk := range *v { + if len(chunk) == 0 { + continue + } + iovecs = append(iovecs, newIovecWithBase(&chunk[0])) + if fd.IsStream && len(chunk) > 1<<30 { + iovecs[len(iovecs)-1].SetLen(1 << 30) + break // continue chunk on next writev + } + iovecs[len(iovecs)-1].SetLen(len(chunk)) + if len(iovecs) == maxVec { + break + } + } + if len(iovecs) == 0 { + break + } + if fd.iovecs == nil { + fd.iovecs = new([]syscall.Iovec) + } + *fd.iovecs = iovecs // cache + + var wrote uintptr + wrote, err = writev(fd.Sysfd, iovecs) + if wrote == ^uintptr(0) { + wrote = 0 + } + TestHookDidWritev(int(wrote)) + n += int64(wrote) + consume(v, int64(wrote)) + for i := range iovecs { + iovecs[i] = syscall.Iovec{} + } + if err != nil { + if err == syscall.EINTR { + continue + } + if err == syscall.EAGAIN { + if err = fd.pd.waitWrite(fd.isFile); err == nil { + continue + } + } + break + } + if n == 0 { + err = io.ErrUnexpectedEOF + break + } + } + return n, err +} diff --git a/src/internal/poll/writev_test.go b/src/internal/poll/writev_test.go new file mode 100644 index 0000000..b46657c --- /dev/null +++ b/src/internal/poll/writev_test.go @@ -0,0 +1,62 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package poll_test + +import ( + "internal/poll" + "reflect" + "testing" +) + +func TestConsume(t *testing.T) { + tests := []struct { + in [][]byte + consume int64 + want [][]byte + }{ + { + in: [][]byte{[]byte("foo"), []byte("bar")}, + consume: 0, + want: [][]byte{[]byte("foo"), []byte("bar")}, + }, + { + in: [][]byte{[]byte("foo"), []byte("bar")}, + consume: 2, + want: [][]byte{[]byte("o"), []byte("bar")}, + }, + { + in: [][]byte{[]byte("foo"), []byte("bar")}, + consume: 3, + want: [][]byte{[]byte("bar")}, + }, + { + in: [][]byte{[]byte("foo"), []byte("bar")}, + consume: 4, + want: [][]byte{[]byte("ar")}, + }, + { + in: [][]byte{nil, nil, nil, []byte("bar")}, + consume: 1, + want: [][]byte{[]byte("ar")}, + }, + { + in: [][]byte{nil, nil, nil, []byte("foo")}, + consume: 0, + want: [][]byte{[]byte("foo")}, + }, + { + in: [][]byte{nil, nil, nil}, + consume: 0, + want: [][]byte{}, + }, + } + for i, tt := range tests { + in := tt.in + poll.Consume(&in, tt.consume) + if !reflect.DeepEqual(in, tt.want) { + t.Errorf("%d. after consume(%d) = %+v, want %+v", i, tt.consume, in, tt.want) + } + } +} diff --git a/src/internal/profile/encode.go b/src/internal/profile/encode.go new file mode 100644 index 0000000..77d77f1 --- /dev/null +++ b/src/internal/profile/encode.go @@ -0,0 +1,482 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package profile + +import ( + "errors" + "fmt" + "sort" +) + +func (p *Profile) decoder() []decoder { + return profileDecoder +} + +// preEncode populates the unexported fields to be used by encode +// (with suffix X) from the corresponding exported fields. The +// exported fields are cleared up to facilitate testing. +func (p *Profile) preEncode() { + strings := make(map[string]int) + addString(strings, "") + + for _, st := range p.SampleType { + st.typeX = addString(strings, st.Type) + st.unitX = addString(strings, st.Unit) + } + + for _, s := range p.Sample { + s.labelX = nil + var keys []string + for k := range s.Label { + keys = append(keys, k) + } + sort.Strings(keys) + for _, k := range keys { + vs := s.Label[k] + for _, v := range vs { + s.labelX = append(s.labelX, + Label{ + keyX: addString(strings, k), + strX: addString(strings, v), + }, + ) + } + } + var numKeys []string + for k := range s.NumLabel { + numKeys = append(numKeys, k) + } + sort.Strings(numKeys) + for _, k := range numKeys { + vs := s.NumLabel[k] + for _, v := range vs { + s.labelX = append(s.labelX, + Label{ + keyX: addString(strings, k), + numX: v, + }, + ) + } + } + s.locationIDX = nil + for _, l := range s.Location { + s.locationIDX = append(s.locationIDX, l.ID) + } + } + + for _, m := range p.Mapping { + m.fileX = addString(strings, m.File) + m.buildIDX = addString(strings, m.BuildID) + } + + for _, l := range p.Location { + for i, ln := range l.Line { + if ln.Function != nil { + l.Line[i].functionIDX = ln.Function.ID + } else { + l.Line[i].functionIDX = 0 + } + } + if l.Mapping != nil { + l.mappingIDX = l.Mapping.ID + } else { + l.mappingIDX = 0 + } + } + for _, f := range p.Function { + f.nameX = addString(strings, f.Name) + f.systemNameX = addString(strings, f.SystemName) + f.filenameX = addString(strings, f.Filename) + } + + p.dropFramesX = addString(strings, p.DropFrames) + p.keepFramesX = addString(strings, p.KeepFrames) + + if pt := p.PeriodType; pt != nil { + pt.typeX = addString(strings, pt.Type) + pt.unitX = addString(strings, pt.Unit) + } + + p.stringTable = make([]string, len(strings)) + for s, i := range strings { + p.stringTable[i] = s + } +} + +func (p *Profile) encode(b *buffer) { + for _, x := range p.SampleType { + encodeMessage(b, 1, x) + } + for _, x := range p.Sample { + encodeMessage(b, 2, x) + } + for _, x := range p.Mapping { + encodeMessage(b, 3, x) + } + for _, x := range p.Location { + encodeMessage(b, 4, x) + } + for _, x := range p.Function { + encodeMessage(b, 5, x) + } + encodeStrings(b, 6, p.stringTable) + encodeInt64Opt(b, 7, p.dropFramesX) + encodeInt64Opt(b, 8, p.keepFramesX) + encodeInt64Opt(b, 9, p.TimeNanos) + encodeInt64Opt(b, 10, p.DurationNanos) + if pt := p.PeriodType; pt != nil && (pt.typeX != 0 || pt.unitX != 0) { + encodeMessage(b, 11, p.PeriodType) + } + encodeInt64Opt(b, 12, p.Period) +} + +var profileDecoder = []decoder{ + nil, // 0 + // repeated ValueType sample_type = 1 + func(b *buffer, m message) error { + x := new(ValueType) + pp := m.(*Profile) + pp.SampleType = append(pp.SampleType, x) + return decodeMessage(b, x) + }, + // repeated Sample sample = 2 + func(b *buffer, m message) error { + x := new(Sample) + pp := m.(*Profile) + pp.Sample = append(pp.Sample, x) + return decodeMessage(b, x) + }, + // repeated Mapping mapping = 3 + func(b *buffer, m message) error { + x := new(Mapping) + pp := m.(*Profile) + pp.Mapping = append(pp.Mapping, x) + return decodeMessage(b, x) + }, + // repeated Location location = 4 + func(b *buffer, m message) error { + x := new(Location) + pp := m.(*Profile) + pp.Location = append(pp.Location, x) + return decodeMessage(b, x) + }, + // repeated Function function = 5 + func(b *buffer, m message) error { + x := new(Function) + pp := m.(*Profile) + pp.Function = append(pp.Function, x) + return decodeMessage(b, x) + }, + // repeated string string_table = 6 + func(b *buffer, m message) error { + err := decodeStrings(b, &m.(*Profile).stringTable) + if err != nil { + return err + } + if m.(*Profile).stringTable[0] != "" { + return errors.New("string_table[0] must be ''") + } + return nil + }, + // repeated int64 drop_frames = 7 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).dropFramesX) }, + // repeated int64 keep_frames = 8 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).keepFramesX) }, + // repeated int64 time_nanos = 9 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).TimeNanos) }, + // repeated int64 duration_nanos = 10 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).DurationNanos) }, + // optional string period_type = 11 + func(b *buffer, m message) error { + x := new(ValueType) + pp := m.(*Profile) + pp.PeriodType = x + return decodeMessage(b, x) + }, + // repeated int64 period = 12 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).Period) }, + // repeated int64 comment = 13 + func(b *buffer, m message) error { return decodeInt64s(b, &m.(*Profile).commentX) }, + // int64 defaultSampleType = 14 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Profile).defaultSampleTypeX) }, +} + +// postDecode takes the unexported fields populated by decode (with +// suffix X) and populates the corresponding exported fields. +// The unexported fields are cleared up to facilitate testing. +func (p *Profile) postDecode() error { + var err error + + mappings := make(map[uint64]*Mapping) + for _, m := range p.Mapping { + m.File, err = getString(p.stringTable, &m.fileX, err) + m.BuildID, err = getString(p.stringTable, &m.buildIDX, err) + mappings[m.ID] = m + } + + functions := make(map[uint64]*Function) + for _, f := range p.Function { + f.Name, err = getString(p.stringTable, &f.nameX, err) + f.SystemName, err = getString(p.stringTable, &f.systemNameX, err) + f.Filename, err = getString(p.stringTable, &f.filenameX, err) + functions[f.ID] = f + } + + locations := make(map[uint64]*Location) + for _, l := range p.Location { + l.Mapping = mappings[l.mappingIDX] + l.mappingIDX = 0 + for i, ln := range l.Line { + if id := ln.functionIDX; id != 0 { + l.Line[i].Function = functions[id] + if l.Line[i].Function == nil { + return fmt.Errorf("Function ID %d not found", id) + } + l.Line[i].functionIDX = 0 + } + } + locations[l.ID] = l + } + + for _, st := range p.SampleType { + st.Type, err = getString(p.stringTable, &st.typeX, err) + st.Unit, err = getString(p.stringTable, &st.unitX, err) + } + + for _, s := range p.Sample { + labels := make(map[string][]string) + numLabels := make(map[string][]int64) + for _, l := range s.labelX { + var key, value string + key, err = getString(p.stringTable, &l.keyX, err) + if l.strX != 0 { + value, err = getString(p.stringTable, &l.strX, err) + labels[key] = append(labels[key], value) + } else { + numLabels[key] = append(numLabels[key], l.numX) + } + } + if len(labels) > 0 { + s.Label = labels + } + if len(numLabels) > 0 { + s.NumLabel = numLabels + } + s.Location = nil + for _, lid := range s.locationIDX { + s.Location = append(s.Location, locations[lid]) + } + s.locationIDX = nil + } + + p.DropFrames, err = getString(p.stringTable, &p.dropFramesX, err) + p.KeepFrames, err = getString(p.stringTable, &p.keepFramesX, err) + + if pt := p.PeriodType; pt == nil { + p.PeriodType = &ValueType{} + } + + if pt := p.PeriodType; pt != nil { + pt.Type, err = getString(p.stringTable, &pt.typeX, err) + pt.Unit, err = getString(p.stringTable, &pt.unitX, err) + } + for _, i := range p.commentX { + var c string + c, err = getString(p.stringTable, &i, err) + p.Comments = append(p.Comments, c) + } + + p.commentX = nil + p.DefaultSampleType, err = getString(p.stringTable, &p.defaultSampleTypeX, err) + p.stringTable = nil + return nil +} + +func (p *ValueType) decoder() []decoder { + return valueTypeDecoder +} + +func (p *ValueType) encode(b *buffer) { + encodeInt64Opt(b, 1, p.typeX) + encodeInt64Opt(b, 2, p.unitX) +} + +var valueTypeDecoder = []decoder{ + nil, // 0 + // optional int64 type = 1 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*ValueType).typeX) }, + // optional int64 unit = 2 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*ValueType).unitX) }, +} + +func (p *Sample) decoder() []decoder { + return sampleDecoder +} + +func (p *Sample) encode(b *buffer) { + encodeUint64s(b, 1, p.locationIDX) + for _, x := range p.Value { + encodeInt64(b, 2, x) + } + for _, x := range p.labelX { + encodeMessage(b, 3, x) + } +} + +var sampleDecoder = []decoder{ + nil, // 0 + // repeated uint64 location = 1 + func(b *buffer, m message) error { return decodeUint64s(b, &m.(*Sample).locationIDX) }, + // repeated int64 value = 2 + func(b *buffer, m message) error { return decodeInt64s(b, &m.(*Sample).Value) }, + // repeated Label label = 3 + func(b *buffer, m message) error { + s := m.(*Sample) + n := len(s.labelX) + s.labelX = append(s.labelX, Label{}) + return decodeMessage(b, &s.labelX[n]) + }, +} + +func (p Label) decoder() []decoder { + return labelDecoder +} + +func (p Label) encode(b *buffer) { + encodeInt64Opt(b, 1, p.keyX) + encodeInt64Opt(b, 2, p.strX) + encodeInt64Opt(b, 3, p.numX) +} + +var labelDecoder = []decoder{ + nil, // 0 + // optional int64 key = 1 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Label).keyX) }, + // optional int64 str = 2 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Label).strX) }, + // optional int64 num = 3 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Label).numX) }, +} + +func (p *Mapping) decoder() []decoder { + return mappingDecoder +} + +func (p *Mapping) encode(b *buffer) { + encodeUint64Opt(b, 1, p.ID) + encodeUint64Opt(b, 2, p.Start) + encodeUint64Opt(b, 3, p.Limit) + encodeUint64Opt(b, 4, p.Offset) + encodeInt64Opt(b, 5, p.fileX) + encodeInt64Opt(b, 6, p.buildIDX) + encodeBoolOpt(b, 7, p.HasFunctions) + encodeBoolOpt(b, 8, p.HasFilenames) + encodeBoolOpt(b, 9, p.HasLineNumbers) + encodeBoolOpt(b, 10, p.HasInlineFrames) +} + +var mappingDecoder = []decoder{ + nil, // 0 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Mapping).ID) }, // optional uint64 id = 1 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Mapping).Start) }, // optional uint64 memory_offset = 2 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Mapping).Limit) }, // optional uint64 memory_limit = 3 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Mapping).Offset) }, // optional uint64 file_offset = 4 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Mapping).fileX) }, // optional int64 filename = 5 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Mapping).buildIDX) }, // optional int64 build_id = 6 + func(b *buffer, m message) error { return decodeBool(b, &m.(*Mapping).HasFunctions) }, // optional bool has_functions = 7 + func(b *buffer, m message) error { return decodeBool(b, &m.(*Mapping).HasFilenames) }, // optional bool has_filenames = 8 + func(b *buffer, m message) error { return decodeBool(b, &m.(*Mapping).HasLineNumbers) }, // optional bool has_line_numbers = 9 + func(b *buffer, m message) error { return decodeBool(b, &m.(*Mapping).HasInlineFrames) }, // optional bool has_inline_frames = 10 +} + +func (p *Location) decoder() []decoder { + return locationDecoder +} + +func (p *Location) encode(b *buffer) { + encodeUint64Opt(b, 1, p.ID) + encodeUint64Opt(b, 2, p.mappingIDX) + encodeUint64Opt(b, 3, p.Address) + for i := range p.Line { + encodeMessage(b, 4, &p.Line[i]) + } +} + +var locationDecoder = []decoder{ + nil, // 0 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Location).ID) }, // optional uint64 id = 1; + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Location).mappingIDX) }, // optional uint64 mapping_id = 2; + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Location).Address) }, // optional uint64 address = 3; + func(b *buffer, m message) error { // repeated Line line = 4 + pp := m.(*Location) + n := len(pp.Line) + pp.Line = append(pp.Line, Line{}) + return decodeMessage(b, &pp.Line[n]) + }, +} + +func (p *Line) decoder() []decoder { + return lineDecoder +} + +func (p *Line) encode(b *buffer) { + encodeUint64Opt(b, 1, p.functionIDX) + encodeInt64Opt(b, 2, p.Line) +} + +var lineDecoder = []decoder{ + nil, // 0 + // optional uint64 function_id = 1 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Line).functionIDX) }, + // optional int64 line = 2 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Line).Line) }, +} + +func (p *Function) decoder() []decoder { + return functionDecoder +} + +func (p *Function) encode(b *buffer) { + encodeUint64Opt(b, 1, p.ID) + encodeInt64Opt(b, 2, p.nameX) + encodeInt64Opt(b, 3, p.systemNameX) + encodeInt64Opt(b, 4, p.filenameX) + encodeInt64Opt(b, 5, p.StartLine) +} + +var functionDecoder = []decoder{ + nil, // 0 + // optional uint64 id = 1 + func(b *buffer, m message) error { return decodeUint64(b, &m.(*Function).ID) }, + // optional int64 function_name = 2 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Function).nameX) }, + // optional int64 function_system_name = 3 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Function).systemNameX) }, + // repeated int64 filename = 4 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Function).filenameX) }, + // optional int64 start_line = 5 + func(b *buffer, m message) error { return decodeInt64(b, &m.(*Function).StartLine) }, +} + +func addString(strings map[string]int, s string) int64 { + i, ok := strings[s] + if !ok { + i = len(strings) + strings[s] = i + } + return int64(i) +} + +func getString(strings []string, strng *int64, err error) (string, error) { + if err != nil { + return "", err + } + s := int(*strng) + if s < 0 || s >= len(strings) { + return "", errMalformed + } + *strng = 0 + return strings[s], nil +} diff --git a/src/internal/profile/filter.go b/src/internal/profile/filter.go new file mode 100644 index 0000000..141dd1f --- /dev/null +++ b/src/internal/profile/filter.go @@ -0,0 +1,158 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Implements methods to filter samples from profiles. + +package profile + +import "regexp" + +// FilterSamplesByName filters the samples in a profile and only keeps +// samples where at least one frame matches focus but none match ignore. +// Returns true is the corresponding regexp matched at least one sample. +func (p *Profile) FilterSamplesByName(focus, ignore, hide *regexp.Regexp) (fm, im, hm bool) { + focusOrIgnore := make(map[uint64]bool) + hidden := make(map[uint64]bool) + for _, l := range p.Location { + if ignore != nil && l.matchesName(ignore) { + im = true + focusOrIgnore[l.ID] = false + } else if focus == nil || l.matchesName(focus) { + fm = true + focusOrIgnore[l.ID] = true + } + if hide != nil && l.matchesName(hide) { + hm = true + l.Line = l.unmatchedLines(hide) + if len(l.Line) == 0 { + hidden[l.ID] = true + } + } + } + + s := make([]*Sample, 0, len(p.Sample)) + for _, sample := range p.Sample { + if focusedAndNotIgnored(sample.Location, focusOrIgnore) { + if len(hidden) > 0 { + var locs []*Location + for _, loc := range sample.Location { + if !hidden[loc.ID] { + locs = append(locs, loc) + } + } + if len(locs) == 0 { + // Remove sample with no locations (by not adding it to s). + continue + } + sample.Location = locs + } + s = append(s, sample) + } + } + p.Sample = s + + return +} + +// matchesName reports whether the function name or file in the +// location matches the regular expression. +func (loc *Location) matchesName(re *regexp.Regexp) bool { + for _, ln := range loc.Line { + if fn := ln.Function; fn != nil { + if re.MatchString(fn.Name) { + return true + } + if re.MatchString(fn.Filename) { + return true + } + } + } + return false +} + +// unmatchedLines returns the lines in the location that do not match +// the regular expression. +func (loc *Location) unmatchedLines(re *regexp.Regexp) []Line { + var lines []Line + for _, ln := range loc.Line { + if fn := ln.Function; fn != nil { + if re.MatchString(fn.Name) { + continue + } + if re.MatchString(fn.Filename) { + continue + } + } + lines = append(lines, ln) + } + return lines +} + +// focusedAndNotIgnored looks up a slice of ids against a map of +// focused/ignored locations. The map only contains locations that are +// explicitly focused or ignored. Returns whether there is at least +// one focused location but no ignored locations. +func focusedAndNotIgnored(locs []*Location, m map[uint64]bool) bool { + var f bool + for _, loc := range locs { + if focus, focusOrIgnore := m[loc.ID]; focusOrIgnore { + if focus { + // Found focused location. Must keep searching in case there + // is an ignored one as well. + f = true + } else { + // Found ignored location. Can return false right away. + return false + } + } + } + return f +} + +// TagMatch selects tags for filtering +type TagMatch func(key, val string, nval int64) bool + +// FilterSamplesByTag removes all samples from the profile, except +// those that match focus and do not match the ignore regular +// expression. +func (p *Profile) FilterSamplesByTag(focus, ignore TagMatch) (fm, im bool) { + samples := make([]*Sample, 0, len(p.Sample)) + for _, s := range p.Sample { + focused, ignored := focusedSample(s, focus, ignore) + fm = fm || focused + im = im || ignored + if focused && !ignored { + samples = append(samples, s) + } + } + p.Sample = samples + return +} + +// focusedSample checks a sample against focus and ignore regexps. +// Returns whether the focus/ignore regexps match any tags. +func focusedSample(s *Sample, focus, ignore TagMatch) (fm, im bool) { + fm = focus == nil + for key, vals := range s.Label { + for _, val := range vals { + if ignore != nil && ignore(key, val, 0) { + im = true + } + if !fm && focus(key, val, 0) { + fm = true + } + } + } + for key, vals := range s.NumLabel { + for _, val := range vals { + if ignore != nil && ignore(key, "", val) { + im = true + } + if !fm && focus(key, "", val) { + fm = true + } + } + } + return fm, im +} diff --git a/src/internal/profile/legacy_profile.go b/src/internal/profile/legacy_profile.go new file mode 100644 index 0000000..373a6c0 --- /dev/null +++ b/src/internal/profile/legacy_profile.go @@ -0,0 +1,1268 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file implements parsers to convert legacy profiles into the +// profile.proto format. + +package profile + +import ( + "bufio" + "bytes" + "fmt" + "internal/lazyregexp" + "io" + "math" + "strconv" + "strings" +) + +var ( + countStartRE = lazyregexp.New(`\A(\w+) profile: total \d+\n\z`) + countRE = lazyregexp.New(`\A(\d+) @(( 0x[0-9a-f]+)+)\n\z`) + + heapHeaderRE = lazyregexp.New(`heap profile: *(\d+): *(\d+) *\[ *(\d+): *(\d+) *\] *@ *(heap[_a-z0-9]*)/?(\d*)`) + heapSampleRE = lazyregexp.New(`(-?\d+): *(-?\d+) *\[ *(\d+): *(\d+) *] @([ x0-9a-f]*)`) + + contentionSampleRE = lazyregexp.New(`(\d+) *(\d+) @([ x0-9a-f]*)`) + + hexNumberRE = lazyregexp.New(`0x[0-9a-f]+`) + + growthHeaderRE = lazyregexp.New(`heap profile: *(\d+): *(\d+) *\[ *(\d+): *(\d+) *\] @ growthz`) + + fragmentationHeaderRE = lazyregexp.New(`heap profile: *(\d+): *(\d+) *\[ *(\d+): *(\d+) *\] @ fragmentationz`) + + threadzStartRE = lazyregexp.New(`--- threadz \d+ ---`) + threadStartRE = lazyregexp.New(`--- Thread ([[:xdigit:]]+) \(name: (.*)/(\d+)\) stack: ---`) + + procMapsRE = lazyregexp.New(`([[:xdigit:]]+)-([[:xdigit:]]+)\s+([-rwxp]+)\s+([[:xdigit:]]+)\s+([[:xdigit:]]+):([[:xdigit:]]+)\s+([[:digit:]]+)\s*(\S+)?`) + + briefMapsRE = lazyregexp.New(`\s*([[:xdigit:]]+)-([[:xdigit:]]+):\s*(\S+)(\s.*@)?([[:xdigit:]]+)?`) + + // LegacyHeapAllocated instructs the heapz parsers to use the + // allocated memory stats instead of the default in-use memory. Note + // that tcmalloc doesn't provide all allocated memory, only in-use + // stats. + LegacyHeapAllocated bool +) + +func isSpaceOrComment(line string) bool { + trimmed := strings.TrimSpace(line) + return len(trimmed) == 0 || trimmed[0] == '#' +} + +// parseGoCount parses a Go count profile (e.g., threadcreate or +// goroutine) and returns a new Profile. +func parseGoCount(b []byte) (*Profile, error) { + r := bytes.NewBuffer(b) + + var line string + var err error + for { + // Skip past comments and empty lines seeking a real header. + line, err = r.ReadString('\n') + if err != nil { + return nil, err + } + if !isSpaceOrComment(line) { + break + } + } + + m := countStartRE.FindStringSubmatch(line) + if m == nil { + return nil, errUnrecognized + } + profileType := m[1] + p := &Profile{ + PeriodType: &ValueType{Type: profileType, Unit: "count"}, + Period: 1, + SampleType: []*ValueType{{Type: profileType, Unit: "count"}}, + } + locations := make(map[uint64]*Location) + for { + line, err = r.ReadString('\n') + if err != nil { + if err == io.EOF { + break + } + return nil, err + } + if isSpaceOrComment(line) { + continue + } + if strings.HasPrefix(line, "---") { + break + } + m := countRE.FindStringSubmatch(line) + if m == nil { + return nil, errMalformed + } + n, err := strconv.ParseInt(m[1], 0, 64) + if err != nil { + return nil, errMalformed + } + fields := strings.Fields(m[2]) + locs := make([]*Location, 0, len(fields)) + for _, stk := range fields { + addr, err := strconv.ParseUint(stk, 0, 64) + if err != nil { + return nil, errMalformed + } + // Adjust all frames by -1 to land on the call instruction. + addr-- + loc := locations[addr] + if loc == nil { + loc = &Location{ + Address: addr, + } + locations[addr] = loc + p.Location = append(p.Location, loc) + } + locs = append(locs, loc) + } + p.Sample = append(p.Sample, &Sample{ + Location: locs, + Value: []int64{n}, + }) + } + + if err = parseAdditionalSections(strings.TrimSpace(line), r, p); err != nil { + return nil, err + } + return p, nil +} + +// remapLocationIDs ensures there is a location for each address +// referenced by a sample, and remaps the samples to point to the new +// location ids. +func (p *Profile) remapLocationIDs() { + seen := make(map[*Location]bool, len(p.Location)) + var locs []*Location + + for _, s := range p.Sample { + for _, l := range s.Location { + if seen[l] { + continue + } + l.ID = uint64(len(locs) + 1) + locs = append(locs, l) + seen[l] = true + } + } + p.Location = locs +} + +func (p *Profile) remapFunctionIDs() { + seen := make(map[*Function]bool, len(p.Function)) + var fns []*Function + + for _, l := range p.Location { + for _, ln := range l.Line { + fn := ln.Function + if fn == nil || seen[fn] { + continue + } + fn.ID = uint64(len(fns) + 1) + fns = append(fns, fn) + seen[fn] = true + } + } + p.Function = fns +} + +// remapMappingIDs matches location addresses with existing mappings +// and updates them appropriately. This is O(N*M), if this ever shows +// up as a bottleneck, evaluate sorting the mappings and doing a +// binary search, which would make it O(N*log(M)). +func (p *Profile) remapMappingIDs() { + if len(p.Mapping) == 0 { + return + } + + // Some profile handlers will incorrectly set regions for the main + // executable if its section is remapped. Fix them through heuristics. + + // Remove the initial mapping if named '/anon_hugepage' and has a + // consecutive adjacent mapping. + if m := p.Mapping[0]; strings.HasPrefix(m.File, "/anon_hugepage") { + if len(p.Mapping) > 1 && m.Limit == p.Mapping[1].Start { + p.Mapping = p.Mapping[1:] + } + } + + for _, l := range p.Location { + if a := l.Address; a != 0 { + for _, m := range p.Mapping { + if m.Start <= a && a < m.Limit { + l.Mapping = m + break + } + } + } + } + + // Reset all mapping IDs. + for i, m := range p.Mapping { + m.ID = uint64(i + 1) + } +} + +var cpuInts = []func([]byte) (uint64, []byte){ + get32l, + get32b, + get64l, + get64b, +} + +func get32l(b []byte) (uint64, []byte) { + if len(b) < 4 { + return 0, nil + } + return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24, b[4:] +} + +func get32b(b []byte) (uint64, []byte) { + if len(b) < 4 { + return 0, nil + } + return uint64(b[3]) | uint64(b[2])<<8 | uint64(b[1])<<16 | uint64(b[0])<<24, b[4:] +} + +func get64l(b []byte) (uint64, []byte) { + if len(b) < 8 { + return 0, nil + } + return uint64(b[0]) | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 | uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56, b[8:] +} + +func get64b(b []byte) (uint64, []byte) { + if len(b) < 8 { + return 0, nil + } + return uint64(b[7]) | uint64(b[6])<<8 | uint64(b[5])<<16 | uint64(b[4])<<24 | uint64(b[3])<<32 | uint64(b[2])<<40 | uint64(b[1])<<48 | uint64(b[0])<<56, b[8:] +} + +// ParseTracebacks parses a set of tracebacks and returns a newly +// populated profile. It will accept any text file and generate a +// Profile out of it with any hex addresses it can identify, including +// a process map if it can recognize one. Each sample will include a +// tag "source" with the addresses recognized in string format. +func ParseTracebacks(b []byte) (*Profile, error) { + r := bytes.NewBuffer(b) + + p := &Profile{ + PeriodType: &ValueType{Type: "trace", Unit: "count"}, + Period: 1, + SampleType: []*ValueType{ + {Type: "trace", Unit: "count"}, + }, + } + + var sources []string + var sloc []*Location + + locs := make(map[uint64]*Location) + for { + l, err := r.ReadString('\n') + if err != nil { + if err != io.EOF { + return nil, err + } + if l == "" { + break + } + } + if sectionTrigger(l) == memoryMapSection { + break + } + if s, addrs := extractHexAddresses(l); len(s) > 0 { + for _, addr := range addrs { + // Addresses from stack traces point to the next instruction after + // each call. Adjust by -1 to land somewhere on the actual call. + addr-- + loc := locs[addr] + if locs[addr] == nil { + loc = &Location{ + Address: addr, + } + p.Location = append(p.Location, loc) + locs[addr] = loc + } + sloc = append(sloc, loc) + } + + sources = append(sources, s...) + } else { + if len(sources) > 0 || len(sloc) > 0 { + addTracebackSample(sloc, sources, p) + sloc, sources = nil, nil + } + } + } + + // Add final sample to save any leftover data. + if len(sources) > 0 || len(sloc) > 0 { + addTracebackSample(sloc, sources, p) + } + + if err := p.ParseMemoryMap(r); err != nil { + return nil, err + } + return p, nil +} + +func addTracebackSample(l []*Location, s []string, p *Profile) { + p.Sample = append(p.Sample, + &Sample{ + Value: []int64{1}, + Location: l, + Label: map[string][]string{"source": s}, + }) +} + +// parseCPU parses a profilez legacy profile and returns a newly +// populated Profile. +// +// The general format for profilez samples is a sequence of words in +// binary format. The first words are a header with the following data: +// +// 1st word -- 0 +// 2nd word -- 3 +// 3rd word -- 0 if a c++ application, 1 if a java application. +// 4th word -- Sampling period (in microseconds). +// 5th word -- Padding. +func parseCPU(b []byte) (*Profile, error) { + var parse func([]byte) (uint64, []byte) + var n1, n2, n3, n4, n5 uint64 + for _, parse = range cpuInts { + var tmp []byte + n1, tmp = parse(b) + n2, tmp = parse(tmp) + n3, tmp = parse(tmp) + n4, tmp = parse(tmp) + n5, tmp = parse(tmp) + + if tmp != nil && n1 == 0 && n2 == 3 && n3 == 0 && n4 > 0 && n5 == 0 { + b = tmp + return cpuProfile(b, int64(n4), parse) + } + } + return nil, errUnrecognized +} + +// cpuProfile returns a new Profile from C++ profilez data. +// b is the profile bytes after the header, period is the profiling +// period, and parse is a function to parse 8-byte chunks from the +// profile in its native endianness. +func cpuProfile(b []byte, period int64, parse func(b []byte) (uint64, []byte)) (*Profile, error) { + p := &Profile{ + Period: period * 1000, + PeriodType: &ValueType{Type: "cpu", Unit: "nanoseconds"}, + SampleType: []*ValueType{ + {Type: "samples", Unit: "count"}, + {Type: "cpu", Unit: "nanoseconds"}, + }, + } + var err error + if b, _, err = parseCPUSamples(b, parse, true, p); err != nil { + return nil, err + } + + // If all samples have the same second-to-the-bottom frame, it + // strongly suggests that it is an uninteresting artifact of + // measurement -- a stack frame pushed by the signal handler. The + // bottom frame is always correct as it is picked up from the signal + // structure, not the stack. Check if this is the case and if so, + // remove. + if len(p.Sample) > 1 && len(p.Sample[0].Location) > 1 { + allSame := true + id1 := p.Sample[0].Location[1].Address + for _, s := range p.Sample { + if len(s.Location) < 2 || id1 != s.Location[1].Address { + allSame = false + break + } + } + if allSame { + for _, s := range p.Sample { + s.Location = append(s.Location[:1], s.Location[2:]...) + } + } + } + + if err := p.ParseMemoryMap(bytes.NewBuffer(b)); err != nil { + return nil, err + } + return p, nil +} + +// parseCPUSamples parses a collection of profilez samples from a +// profile. +// +// profilez samples are a repeated sequence of stack frames of the +// form: +// +// 1st word -- The number of times this stack was encountered. +// 2nd word -- The size of the stack (StackSize). +// 3rd word -- The first address on the stack. +// ... +// StackSize + 2 -- The last address on the stack +// +// The last stack trace is of the form: +// +// 1st word -- 0 +// 2nd word -- 1 +// 3rd word -- 0 +// +// Addresses from stack traces may point to the next instruction after +// each call. Optionally adjust by -1 to land somewhere on the actual +// call (except for the leaf, which is not a call). +func parseCPUSamples(b []byte, parse func(b []byte) (uint64, []byte), adjust bool, p *Profile) ([]byte, map[uint64]*Location, error) { + locs := make(map[uint64]*Location) + for len(b) > 0 { + var count, nstk uint64 + count, b = parse(b) + nstk, b = parse(b) + if b == nil || nstk > uint64(len(b)/4) { + return nil, nil, errUnrecognized + } + var sloc []*Location + addrs := make([]uint64, nstk) + for i := 0; i < int(nstk); i++ { + addrs[i], b = parse(b) + } + + if count == 0 && nstk == 1 && addrs[0] == 0 { + // End of data marker + break + } + for i, addr := range addrs { + if adjust && i > 0 { + addr-- + } + loc := locs[addr] + if loc == nil { + loc = &Location{ + Address: addr, + } + locs[addr] = loc + p.Location = append(p.Location, loc) + } + sloc = append(sloc, loc) + } + p.Sample = append(p.Sample, + &Sample{ + Value: []int64{int64(count), int64(count) * p.Period}, + Location: sloc, + }) + } + // Reached the end without finding the EOD marker. + return b, locs, nil +} + +// parseHeap parses a heapz legacy or a growthz profile and +// returns a newly populated Profile. +func parseHeap(b []byte) (p *Profile, err error) { + r := bytes.NewBuffer(b) + l, err := r.ReadString('\n') + if err != nil { + return nil, errUnrecognized + } + + sampling := "" + + if header := heapHeaderRE.FindStringSubmatch(l); header != nil { + p = &Profile{ + SampleType: []*ValueType{ + {Type: "objects", Unit: "count"}, + {Type: "space", Unit: "bytes"}, + }, + PeriodType: &ValueType{Type: "objects", Unit: "bytes"}, + } + + var period int64 + if len(header[6]) > 0 { + if period, err = strconv.ParseInt(header[6], 10, 64); err != nil { + return nil, errUnrecognized + } + } + + switch header[5] { + case "heapz_v2", "heap_v2": + sampling, p.Period = "v2", period + case "heapprofile": + sampling, p.Period = "", 1 + case "heap": + sampling, p.Period = "v2", period/2 + default: + return nil, errUnrecognized + } + } else if header = growthHeaderRE.FindStringSubmatch(l); header != nil { + p = &Profile{ + SampleType: []*ValueType{ + {Type: "objects", Unit: "count"}, + {Type: "space", Unit: "bytes"}, + }, + PeriodType: &ValueType{Type: "heapgrowth", Unit: "count"}, + Period: 1, + } + } else if header = fragmentationHeaderRE.FindStringSubmatch(l); header != nil { + p = &Profile{ + SampleType: []*ValueType{ + {Type: "objects", Unit: "count"}, + {Type: "space", Unit: "bytes"}, + }, + PeriodType: &ValueType{Type: "allocations", Unit: "count"}, + Period: 1, + } + } else { + return nil, errUnrecognized + } + + if LegacyHeapAllocated { + for _, st := range p.SampleType { + st.Type = "alloc_" + st.Type + } + } else { + for _, st := range p.SampleType { + st.Type = "inuse_" + st.Type + } + } + + locs := make(map[uint64]*Location) + for { + l, err = r.ReadString('\n') + if err != nil { + if err != io.EOF { + return nil, err + } + + if l == "" { + break + } + } + + if isSpaceOrComment(l) { + continue + } + l = strings.TrimSpace(l) + + if sectionTrigger(l) != unrecognizedSection { + break + } + + value, blocksize, addrs, err := parseHeapSample(l, p.Period, sampling) + if err != nil { + return nil, err + } + var sloc []*Location + for _, addr := range addrs { + // Addresses from stack traces point to the next instruction after + // each call. Adjust by -1 to land somewhere on the actual call. + addr-- + loc := locs[addr] + if locs[addr] == nil { + loc = &Location{ + Address: addr, + } + p.Location = append(p.Location, loc) + locs[addr] = loc + } + sloc = append(sloc, loc) + } + + p.Sample = append(p.Sample, &Sample{ + Value: value, + Location: sloc, + NumLabel: map[string][]int64{"bytes": {blocksize}}, + }) + } + + if err = parseAdditionalSections(l, r, p); err != nil { + return nil, err + } + return p, nil +} + +// parseHeapSample parses a single row from a heap profile into a new Sample. +func parseHeapSample(line string, rate int64, sampling string) (value []int64, blocksize int64, addrs []uint64, err error) { + sampleData := heapSampleRE.FindStringSubmatch(line) + if len(sampleData) != 6 { + return value, blocksize, addrs, fmt.Errorf("unexpected number of sample values: got %d, want 6", len(sampleData)) + } + + // Use first two values by default; tcmalloc sampling generates the + // same value for both, only the older heap-profile collect separate + // stats for in-use and allocated objects. + valueIndex := 1 + if LegacyHeapAllocated { + valueIndex = 3 + } + + var v1, v2 int64 + if v1, err = strconv.ParseInt(sampleData[valueIndex], 10, 64); err != nil { + return value, blocksize, addrs, fmt.Errorf("malformed sample: %s: %v", line, err) + } + if v2, err = strconv.ParseInt(sampleData[valueIndex+1], 10, 64); err != nil { + return value, blocksize, addrs, fmt.Errorf("malformed sample: %s: %v", line, err) + } + + if v1 == 0 { + if v2 != 0 { + return value, blocksize, addrs, fmt.Errorf("allocation count was 0 but allocation bytes was %d", v2) + } + } else { + blocksize = v2 / v1 + if sampling == "v2" { + v1, v2 = scaleHeapSample(v1, v2, rate) + } + } + + value = []int64{v1, v2} + addrs = parseHexAddresses(sampleData[5]) + + return value, blocksize, addrs, nil +} + +// extractHexAddresses extracts hex numbers from a string and returns +// them, together with their numeric value, in a slice. +func extractHexAddresses(s string) ([]string, []uint64) { + hexStrings := hexNumberRE.FindAllString(s, -1) + var ids []uint64 + for _, s := range hexStrings { + if id, err := strconv.ParseUint(s, 0, 64); err == nil { + ids = append(ids, id) + } else { + // Do not expect any parsing failures due to the regexp matching. + panic("failed to parse hex value:" + s) + } + } + return hexStrings, ids +} + +// parseHexAddresses parses hex numbers from a string and returns them +// in a slice. +func parseHexAddresses(s string) []uint64 { + _, ids := extractHexAddresses(s) + return ids +} + +// scaleHeapSample adjusts the data from a heapz Sample to +// account for its probability of appearing in the collected +// data. heapz profiles are a sampling of the memory allocations +// requests in a program. We estimate the unsampled value by dividing +// each collected sample by its probability of appearing in the +// profile. heapz v2 profiles rely on a poisson process to determine +// which samples to collect, based on the desired average collection +// rate R. The probability of a sample of size S to appear in that +// profile is 1-exp(-S/R). +func scaleHeapSample(count, size, rate int64) (int64, int64) { + if count == 0 || size == 0 { + return 0, 0 + } + + if rate <= 1 { + // if rate==1 all samples were collected so no adjustment is needed. + // if rate<1 treat as unknown and skip scaling. + return count, size + } + + avgSize := float64(size) / float64(count) + scale := 1 / (1 - math.Exp(-avgSize/float64(rate))) + + return int64(float64(count) * scale), int64(float64(size) * scale) +} + +// parseContention parses a mutex or contention profile. There are 2 cases: +// "--- contentionz " for legacy C++ profiles (and backwards compatibility) +// "--- mutex:" or "--- contention:" for profiles generated by the Go runtime. +// This code converts the text output from runtime into a *Profile. (In the future +// the runtime might write a serialized Profile directly making this unnecessary.) +func parseContention(b []byte) (*Profile, error) { + r := bytes.NewBuffer(b) + var l string + var err error + for { + // Skip past comments and empty lines seeking a real header. + l, err = r.ReadString('\n') + if err != nil { + return nil, err + } + if !isSpaceOrComment(l) { + break + } + } + + if strings.HasPrefix(l, "--- contentionz ") { + return parseCppContention(r) + } else if strings.HasPrefix(l, "--- mutex:") { + return parseCppContention(r) + } else if strings.HasPrefix(l, "--- contention:") { + return parseCppContention(r) + } + return nil, errUnrecognized +} + +// parseCppContention parses the output from synchronization_profiling.cc +// for backward compatibility, and the compatible (non-debug) block profile +// output from the Go runtime. +func parseCppContention(r *bytes.Buffer) (*Profile, error) { + p := &Profile{ + PeriodType: &ValueType{Type: "contentions", Unit: "count"}, + Period: 1, + SampleType: []*ValueType{ + {Type: "contentions", Unit: "count"}, + {Type: "delay", Unit: "nanoseconds"}, + }, + } + + var cpuHz int64 + var l string + var err error + // Parse text of the form "attribute = value" before the samples. + const delimiter = '=' + for { + l, err = r.ReadString('\n') + if err != nil { + if err != io.EOF { + return nil, err + } + + if l == "" { + break + } + } + if isSpaceOrComment(l) { + continue + } + + if l = strings.TrimSpace(l); l == "" { + continue + } + + if strings.HasPrefix(l, "---") { + break + } + + index := strings.IndexByte(l, delimiter) + if index < 0 { + break + } + key := l[:index] + val := l[index+1:] + + key, val = strings.TrimSpace(key), strings.TrimSpace(val) + var err error + switch key { + case "cycles/second": + if cpuHz, err = strconv.ParseInt(val, 0, 64); err != nil { + return nil, errUnrecognized + } + case "sampling period": + if p.Period, err = strconv.ParseInt(val, 0, 64); err != nil { + return nil, errUnrecognized + } + case "ms since reset": + ms, err := strconv.ParseInt(val, 0, 64) + if err != nil { + return nil, errUnrecognized + } + p.DurationNanos = ms * 1000 * 1000 + case "format": + // CPP contentionz profiles don't have format. + return nil, errUnrecognized + case "resolution": + // CPP contentionz profiles don't have resolution. + return nil, errUnrecognized + case "discarded samples": + default: + return nil, errUnrecognized + } + } + + locs := make(map[uint64]*Location) + for { + if !isSpaceOrComment(l) { + if l = strings.TrimSpace(l); strings.HasPrefix(l, "---") { + break + } + value, addrs, err := parseContentionSample(l, p.Period, cpuHz) + if err != nil { + return nil, err + } + var sloc []*Location + for _, addr := range addrs { + // Addresses from stack traces point to the next instruction after + // each call. Adjust by -1 to land somewhere on the actual call. + addr-- + loc := locs[addr] + if locs[addr] == nil { + loc = &Location{ + Address: addr, + } + p.Location = append(p.Location, loc) + locs[addr] = loc + } + sloc = append(sloc, loc) + } + p.Sample = append(p.Sample, &Sample{ + Value: value, + Location: sloc, + }) + } + + if l, err = r.ReadString('\n'); err != nil { + if err != io.EOF { + return nil, err + } + if l == "" { + break + } + } + } + + if err = parseAdditionalSections(l, r, p); err != nil { + return nil, err + } + + return p, nil +} + +// parseContentionSample parses a single row from a contention profile +// into a new Sample. +func parseContentionSample(line string, period, cpuHz int64) (value []int64, addrs []uint64, err error) { + sampleData := contentionSampleRE.FindStringSubmatch(line) + if sampleData == nil { + return value, addrs, errUnrecognized + } + + v1, err := strconv.ParseInt(sampleData[1], 10, 64) + if err != nil { + return value, addrs, fmt.Errorf("malformed sample: %s: %v", line, err) + } + v2, err := strconv.ParseInt(sampleData[2], 10, 64) + if err != nil { + return value, addrs, fmt.Errorf("malformed sample: %s: %v", line, err) + } + + // Unsample values if period and cpuHz are available. + // - Delays are scaled to cycles and then to nanoseconds. + // - Contentions are scaled to cycles. + if period > 0 { + if cpuHz > 0 { + cpuGHz := float64(cpuHz) / 1e9 + v1 = int64(float64(v1) * float64(period) / cpuGHz) + } + v2 = v2 * period + } + + value = []int64{v2, v1} + addrs = parseHexAddresses(sampleData[3]) + + return value, addrs, nil +} + +// parseThread parses a Threadz profile and returns a new Profile. +func parseThread(b []byte) (*Profile, error) { + r := bytes.NewBuffer(b) + + var line string + var err error + for { + // Skip past comments and empty lines seeking a real header. + line, err = r.ReadString('\n') + if err != nil { + return nil, err + } + if !isSpaceOrComment(line) { + break + } + } + + if m := threadzStartRE.FindStringSubmatch(line); m != nil { + // Advance over initial comments until first stack trace. + for { + line, err = r.ReadString('\n') + if err != nil { + if err != io.EOF { + return nil, err + } + + if line == "" { + break + } + } + if sectionTrigger(line) != unrecognizedSection || line[0] == '-' { + break + } + } + } else if t := threadStartRE.FindStringSubmatch(line); len(t) != 4 { + return nil, errUnrecognized + } + + p := &Profile{ + SampleType: []*ValueType{{Type: "thread", Unit: "count"}}, + PeriodType: &ValueType{Type: "thread", Unit: "count"}, + Period: 1, + } + + locs := make(map[uint64]*Location) + // Recognize each thread and populate profile samples. + for sectionTrigger(line) == unrecognizedSection { + if strings.HasPrefix(line, "---- no stack trace for") { + line = "" + break + } + if t := threadStartRE.FindStringSubmatch(line); len(t) != 4 { + return nil, errUnrecognized + } + + var addrs []uint64 + line, addrs, err = parseThreadSample(r) + if err != nil { + return nil, errUnrecognized + } + if len(addrs) == 0 { + // We got a --same as previous threads--. Bump counters. + if len(p.Sample) > 0 { + s := p.Sample[len(p.Sample)-1] + s.Value[0]++ + } + continue + } + + var sloc []*Location + for _, addr := range addrs { + // Addresses from stack traces point to the next instruction after + // each call. Adjust by -1 to land somewhere on the actual call. + addr-- + loc := locs[addr] + if locs[addr] == nil { + loc = &Location{ + Address: addr, + } + p.Location = append(p.Location, loc) + locs[addr] = loc + } + sloc = append(sloc, loc) + } + + p.Sample = append(p.Sample, &Sample{ + Value: []int64{1}, + Location: sloc, + }) + } + + if err = parseAdditionalSections(line, r, p); err != nil { + return nil, err + } + + return p, nil +} + +// parseThreadSample parses a symbolized or unsymbolized stack trace. +// Returns the first line after the traceback, the sample (or nil if +// it hits a 'same-as-previous' marker) and an error. +func parseThreadSample(b *bytes.Buffer) (nextl string, addrs []uint64, err error) { + var l string + sameAsPrevious := false + for { + if l, err = b.ReadString('\n'); err != nil { + if err != io.EOF { + return "", nil, err + } + if l == "" { + break + } + } + if l = strings.TrimSpace(l); l == "" { + continue + } + + if strings.HasPrefix(l, "---") { + break + } + if strings.Contains(l, "same as previous thread") { + sameAsPrevious = true + continue + } + + addrs = append(addrs, parseHexAddresses(l)...) + } + + if sameAsPrevious { + return l, nil, nil + } + return l, addrs, nil +} + +// parseAdditionalSections parses any additional sections in the +// profile, ignoring any unrecognized sections. +func parseAdditionalSections(l string, b *bytes.Buffer, p *Profile) (err error) { + for { + if sectionTrigger(l) == memoryMapSection { + break + } + // Ignore any unrecognized sections. + if l, err := b.ReadString('\n'); err != nil { + if err != io.EOF { + return err + } + if l == "" { + break + } + } + } + return p.ParseMemoryMap(b) +} + +// ParseMemoryMap parses a memory map in the format of +// /proc/self/maps, and overrides the mappings in the current profile. +// It renumbers the samples and locations in the profile correspondingly. +func (p *Profile) ParseMemoryMap(rd io.Reader) error { + b := bufio.NewReader(rd) + + var attrs []string + var r *strings.Replacer + const delimiter = '=' + for { + l, err := b.ReadString('\n') + if err != nil { + if err != io.EOF { + return err + } + if l == "" { + break + } + } + if l = strings.TrimSpace(l); l == "" { + continue + } + + if r != nil { + l = r.Replace(l) + } + m, err := parseMappingEntry(l) + if err != nil { + if err == errUnrecognized { + // Recognize assignments of the form: attr=value, and replace + // $attr with value on subsequent mappings. + idx := strings.IndexByte(l, delimiter) + if idx >= 0 { + attr := l[:idx] + value := l[idx+1:] + attrs = append(attrs, "$"+strings.TrimSpace(attr), strings.TrimSpace(value)) + r = strings.NewReplacer(attrs...) + } + // Ignore any unrecognized entries + continue + } + return err + } + if m == nil || (m.File == "" && len(p.Mapping) != 0) { + // In some cases the first entry may include the address range + // but not the name of the file. It should be followed by + // another entry with the name. + continue + } + if len(p.Mapping) == 1 && p.Mapping[0].File == "" { + // Update the name if this is the entry following that empty one. + p.Mapping[0].File = m.File + continue + } + p.Mapping = append(p.Mapping, m) + } + p.remapLocationIDs() + p.remapFunctionIDs() + p.remapMappingIDs() + return nil +} + +func parseMappingEntry(l string) (*Mapping, error) { + mapping := &Mapping{} + var err error + if me := procMapsRE.FindStringSubmatch(l); len(me) == 9 { + if !strings.Contains(me[3], "x") { + // Skip non-executable entries. + return nil, nil + } + if mapping.Start, err = strconv.ParseUint(me[1], 16, 64); err != nil { + return nil, errUnrecognized + } + if mapping.Limit, err = strconv.ParseUint(me[2], 16, 64); err != nil { + return nil, errUnrecognized + } + if me[4] != "" { + if mapping.Offset, err = strconv.ParseUint(me[4], 16, 64); err != nil { + return nil, errUnrecognized + } + } + mapping.File = me[8] + return mapping, nil + } + + if me := briefMapsRE.FindStringSubmatch(l); len(me) == 6 { + if mapping.Start, err = strconv.ParseUint(me[1], 16, 64); err != nil { + return nil, errUnrecognized + } + if mapping.Limit, err = strconv.ParseUint(me[2], 16, 64); err != nil { + return nil, errUnrecognized + } + mapping.File = me[3] + if me[5] != "" { + if mapping.Offset, err = strconv.ParseUint(me[5], 16, 64); err != nil { + return nil, errUnrecognized + } + } + return mapping, nil + } + + return nil, errUnrecognized +} + +type sectionType int + +const ( + unrecognizedSection sectionType = iota + memoryMapSection +) + +var memoryMapTriggers = []string{ + "--- Memory map: ---", + "MAPPED_LIBRARIES:", +} + +func sectionTrigger(line string) sectionType { + for _, trigger := range memoryMapTriggers { + if strings.Contains(line, trigger) { + return memoryMapSection + } + } + return unrecognizedSection +} + +func (p *Profile) addLegacyFrameInfo() { + switch { + case isProfileType(p, heapzSampleTypes) || + isProfileType(p, heapzInUseSampleTypes) || + isProfileType(p, heapzAllocSampleTypes): + p.DropFrames, p.KeepFrames = allocRxStr, allocSkipRxStr + case isProfileType(p, contentionzSampleTypes): + p.DropFrames, p.KeepFrames = lockRxStr, "" + default: + p.DropFrames, p.KeepFrames = cpuProfilerRxStr, "" + } +} + +var heapzSampleTypes = []string{"allocations", "size"} // early Go pprof profiles +var heapzInUseSampleTypes = []string{"inuse_objects", "inuse_space"} +var heapzAllocSampleTypes = []string{"alloc_objects", "alloc_space"} +var contentionzSampleTypes = []string{"contentions", "delay"} + +func isProfileType(p *Profile, t []string) bool { + st := p.SampleType + if len(st) != len(t) { + return false + } + + for i := range st { + if st[i].Type != t[i] { + return false + } + } + return true +} + +var allocRxStr = strings.Join([]string{ + // POSIX entry points. + `calloc`, + `cfree`, + `malloc`, + `free`, + `memalign`, + `do_memalign`, + `(__)?posix_memalign`, + `pvalloc`, + `valloc`, + `realloc`, + + // TC malloc. + `tcmalloc::.*`, + `tc_calloc`, + `tc_cfree`, + `tc_malloc`, + `tc_free`, + `tc_memalign`, + `tc_posix_memalign`, + `tc_pvalloc`, + `tc_valloc`, + `tc_realloc`, + `tc_new`, + `tc_delete`, + `tc_newarray`, + `tc_deletearray`, + `tc_new_nothrow`, + `tc_newarray_nothrow`, + + // Memory-allocation routines on OS X. + `malloc_zone_malloc`, + `malloc_zone_calloc`, + `malloc_zone_valloc`, + `malloc_zone_realloc`, + `malloc_zone_memalign`, + `malloc_zone_free`, + + // Go runtime + `runtime\..*`, + + // Other misc. memory allocation routines + `BaseArena::.*`, + `(::)?do_malloc_no_errno`, + `(::)?do_malloc_pages`, + `(::)?do_malloc`, + `DoSampledAllocation`, + `MallocedMemBlock::MallocedMemBlock`, + `_M_allocate`, + `__builtin_(vec_)?delete`, + `__builtin_(vec_)?new`, + `__gnu_cxx::new_allocator::allocate`, + `__libc_malloc`, + `__malloc_alloc_template::allocate`, + `allocate`, + `cpp_alloc`, + `operator new(\[\])?`, + `simple_alloc::allocate`, +}, `|`) + +var allocSkipRxStr = strings.Join([]string{ + // Preserve Go runtime frames that appear in the middle/bottom of + // the stack. + `runtime\.panic`, + `runtime\.reflectcall`, + `runtime\.call[0-9]*`, +}, `|`) + +var cpuProfilerRxStr = strings.Join([]string{ + `ProfileData::Add`, + `ProfileData::prof_handler`, + `CpuProfiler::prof_handler`, + `__pthread_sighandler`, + `__restore`, +}, `|`) + +var lockRxStr = strings.Join([]string{ + `RecordLockProfileData`, + `(base::)?RecordLockProfileData.*`, + `(base::)?SubmitMutexProfileData.*`, + `(base::)?SubmitSpinLockProfileData.*`, + `(Mutex::)?AwaitCommon.*`, + `(Mutex::)?Unlock.*`, + `(Mutex::)?UnlockSlow.*`, + `(Mutex::)?ReaderUnlock.*`, + `(MutexLock::)?~MutexLock.*`, + `(SpinLock::)?Unlock.*`, + `(SpinLock::)?SlowUnlock.*`, + `(SpinLockHolder::)?~SpinLockHolder.*`, +}, `|`) diff --git a/src/internal/profile/merge.go b/src/internal/profile/merge.go new file mode 100644 index 0000000..3ea7d4c --- /dev/null +++ b/src/internal/profile/merge.go @@ -0,0 +1,461 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package profile + +import ( + "fmt" + "sort" + "strconv" + "strings" +) + +// Merge merges all the profiles in profs into a single Profile. +// Returns a new profile independent of the input profiles. The merged +// profile is compacted to eliminate unused samples, locations, +// functions and mappings. Profiles must have identical profile sample +// and period types or the merge will fail. profile.Period of the +// resulting profile will be the maximum of all profiles, and +// profile.TimeNanos will be the earliest nonzero one. +func Merge(srcs []*Profile) (*Profile, error) { + if len(srcs) == 0 { + return nil, fmt.Errorf("no profiles to merge") + } + p, err := combineHeaders(srcs) + if err != nil { + return nil, err + } + + pm := &profileMerger{ + p: p, + samples: make(map[sampleKey]*Sample, len(srcs[0].Sample)), + locations: make(map[locationKey]*Location, len(srcs[0].Location)), + functions: make(map[functionKey]*Function, len(srcs[0].Function)), + mappings: make(map[mappingKey]*Mapping, len(srcs[0].Mapping)), + } + + for _, src := range srcs { + // Clear the profile-specific hash tables + pm.locationsByID = make(map[uint64]*Location, len(src.Location)) + pm.functionsByID = make(map[uint64]*Function, len(src.Function)) + pm.mappingsByID = make(map[uint64]mapInfo, len(src.Mapping)) + + if len(pm.mappings) == 0 && len(src.Mapping) > 0 { + // The Mapping list has the property that the first mapping + // represents the main binary. Take the first Mapping we see, + // otherwise the operations below will add mappings in an + // arbitrary order. + pm.mapMapping(src.Mapping[0]) + } + + for _, s := range src.Sample { + if !isZeroSample(s) { + pm.mapSample(s) + } + } + } + + for _, s := range p.Sample { + if isZeroSample(s) { + // If there are any zero samples, re-merge the profile to GC + // them. + return Merge([]*Profile{p}) + } + } + + return p, nil +} + +// Normalize normalizes the source profile by multiplying each value in profile by the +// ratio of the sum of the base profile's values of that sample type to the sum of the +// source profile's value of that sample type. +func (p *Profile) Normalize(pb *Profile) error { + + if err := p.compatible(pb); err != nil { + return err + } + + baseVals := make([]int64, len(p.SampleType)) + for _, s := range pb.Sample { + for i, v := range s.Value { + baseVals[i] += v + } + } + + srcVals := make([]int64, len(p.SampleType)) + for _, s := range p.Sample { + for i, v := range s.Value { + srcVals[i] += v + } + } + + normScale := make([]float64, len(baseVals)) + for i := range baseVals { + if srcVals[i] == 0 { + normScale[i] = 0.0 + } else { + normScale[i] = float64(baseVals[i]) / float64(srcVals[i]) + } + } + p.ScaleN(normScale) + return nil +} + +func isZeroSample(s *Sample) bool { + for _, v := range s.Value { + if v != 0 { + return false + } + } + return true +} + +type profileMerger struct { + p *Profile + + // Memoization tables within a profile. + locationsByID map[uint64]*Location + functionsByID map[uint64]*Function + mappingsByID map[uint64]mapInfo + + // Memoization tables for profile entities. + samples map[sampleKey]*Sample + locations map[locationKey]*Location + functions map[functionKey]*Function + mappings map[mappingKey]*Mapping +} + +type mapInfo struct { + m *Mapping + offset int64 +} + +func (pm *profileMerger) mapSample(src *Sample) *Sample { + s := &Sample{ + Location: make([]*Location, len(src.Location)), + Value: make([]int64, len(src.Value)), + Label: make(map[string][]string, len(src.Label)), + NumLabel: make(map[string][]int64, len(src.NumLabel)), + NumUnit: make(map[string][]string, len(src.NumLabel)), + } + for i, l := range src.Location { + s.Location[i] = pm.mapLocation(l) + } + for k, v := range src.Label { + vv := make([]string, len(v)) + copy(vv, v) + s.Label[k] = vv + } + for k, v := range src.NumLabel { + u := src.NumUnit[k] + vv := make([]int64, len(v)) + uu := make([]string, len(u)) + copy(vv, v) + copy(uu, u) + s.NumLabel[k] = vv + s.NumUnit[k] = uu + } + // Check memoization table. Must be done on the remapped location to + // account for the remapped mapping. Add current values to the + // existing sample. + k := s.key() + if ss, ok := pm.samples[k]; ok { + for i, v := range src.Value { + ss.Value[i] += v + } + return ss + } + copy(s.Value, src.Value) + pm.samples[k] = s + pm.p.Sample = append(pm.p.Sample, s) + return s +} + +// key generates sampleKey to be used as a key for maps. +func (sample *Sample) key() sampleKey { + ids := make([]string, len(sample.Location)) + for i, l := range sample.Location { + ids[i] = strconv.FormatUint(l.ID, 16) + } + + labels := make([]string, 0, len(sample.Label)) + for k, v := range sample.Label { + labels = append(labels, fmt.Sprintf("%q%q", k, v)) + } + sort.Strings(labels) + + numlabels := make([]string, 0, len(sample.NumLabel)) + for k, v := range sample.NumLabel { + numlabels = append(numlabels, fmt.Sprintf("%q%x%x", k, v, sample.NumUnit[k])) + } + sort.Strings(numlabels) + + return sampleKey{ + strings.Join(ids, "|"), + strings.Join(labels, ""), + strings.Join(numlabels, ""), + } +} + +type sampleKey struct { + locations string + labels string + numlabels string +} + +func (pm *profileMerger) mapLocation(src *Location) *Location { + if src == nil { + return nil + } + + if l, ok := pm.locationsByID[src.ID]; ok { + pm.locationsByID[src.ID] = l + return l + } + + mi := pm.mapMapping(src.Mapping) + l := &Location{ + ID: uint64(len(pm.p.Location) + 1), + Mapping: mi.m, + Address: uint64(int64(src.Address) + mi.offset), + Line: make([]Line, len(src.Line)), + IsFolded: src.IsFolded, + } + for i, ln := range src.Line { + l.Line[i] = pm.mapLine(ln) + } + // Check memoization table. Must be done on the remapped location to + // account for the remapped mapping ID. + k := l.key() + if ll, ok := pm.locations[k]; ok { + pm.locationsByID[src.ID] = ll + return ll + } + pm.locationsByID[src.ID] = l + pm.locations[k] = l + pm.p.Location = append(pm.p.Location, l) + return l +} + +// key generates locationKey to be used as a key for maps. +func (l *Location) key() locationKey { + key := locationKey{ + addr: l.Address, + isFolded: l.IsFolded, + } + if l.Mapping != nil { + // Normalizes address to handle address space randomization. + key.addr -= l.Mapping.Start + key.mappingID = l.Mapping.ID + } + lines := make([]string, len(l.Line)*2) + for i, line := range l.Line { + if line.Function != nil { + lines[i*2] = strconv.FormatUint(line.Function.ID, 16) + } + lines[i*2+1] = strconv.FormatInt(line.Line, 16) + } + key.lines = strings.Join(lines, "|") + return key +} + +type locationKey struct { + addr, mappingID uint64 + lines string + isFolded bool +} + +func (pm *profileMerger) mapMapping(src *Mapping) mapInfo { + if src == nil { + return mapInfo{} + } + + if mi, ok := pm.mappingsByID[src.ID]; ok { + return mi + } + + // Check memoization tables. + mk := src.key() + if m, ok := pm.mappings[mk]; ok { + mi := mapInfo{m, int64(m.Start) - int64(src.Start)} + pm.mappingsByID[src.ID] = mi + return mi + } + m := &Mapping{ + ID: uint64(len(pm.p.Mapping) + 1), + Start: src.Start, + Limit: src.Limit, + Offset: src.Offset, + File: src.File, + BuildID: src.BuildID, + HasFunctions: src.HasFunctions, + HasFilenames: src.HasFilenames, + HasLineNumbers: src.HasLineNumbers, + HasInlineFrames: src.HasInlineFrames, + } + pm.p.Mapping = append(pm.p.Mapping, m) + + // Update memoization tables. + pm.mappings[mk] = m + mi := mapInfo{m, 0} + pm.mappingsByID[src.ID] = mi + return mi +} + +// key generates encoded strings of Mapping to be used as a key for +// maps. +func (m *Mapping) key() mappingKey { + // Normalize addresses to handle address space randomization. + // Round up to next 4K boundary to avoid minor discrepancies. + const mapsizeRounding = 0x1000 + + size := m.Limit - m.Start + size = size + mapsizeRounding - 1 + size = size - (size % mapsizeRounding) + key := mappingKey{ + size: size, + offset: m.Offset, + } + + switch { + case m.BuildID != "": + key.buildIDOrFile = m.BuildID + case m.File != "": + key.buildIDOrFile = m.File + default: + // A mapping containing neither build ID nor file name is a fake mapping. A + // key with empty buildIDOrFile is used for fake mappings so that they are + // treated as the same mapping during merging. + } + return key +} + +type mappingKey struct { + size, offset uint64 + buildIDOrFile string +} + +func (pm *profileMerger) mapLine(src Line) Line { + ln := Line{ + Function: pm.mapFunction(src.Function), + Line: src.Line, + } + return ln +} + +func (pm *profileMerger) mapFunction(src *Function) *Function { + if src == nil { + return nil + } + if f, ok := pm.functionsByID[src.ID]; ok { + return f + } + k := src.key() + if f, ok := pm.functions[k]; ok { + pm.functionsByID[src.ID] = f + return f + } + f := &Function{ + ID: uint64(len(pm.p.Function) + 1), + Name: src.Name, + SystemName: src.SystemName, + Filename: src.Filename, + StartLine: src.StartLine, + } + pm.functions[k] = f + pm.functionsByID[src.ID] = f + pm.p.Function = append(pm.p.Function, f) + return f +} + +// key generates a struct to be used as a key for maps. +func (f *Function) key() functionKey { + return functionKey{ + f.StartLine, + f.Name, + f.SystemName, + f.Filename, + } +} + +type functionKey struct { + startLine int64 + name, systemName, fileName string +} + +// combineHeaders checks that all profiles can be merged and returns +// their combined profile. +func combineHeaders(srcs []*Profile) (*Profile, error) { + for _, s := range srcs[1:] { + if err := srcs[0].compatible(s); err != nil { + return nil, err + } + } + + var timeNanos, durationNanos, period int64 + var comments []string + seenComments := map[string]bool{} + var defaultSampleType string + for _, s := range srcs { + if timeNanos == 0 || s.TimeNanos < timeNanos { + timeNanos = s.TimeNanos + } + durationNanos += s.DurationNanos + if period == 0 || period < s.Period { + period = s.Period + } + for _, c := range s.Comments { + if seen := seenComments[c]; !seen { + comments = append(comments, c) + seenComments[c] = true + } + } + if defaultSampleType == "" { + defaultSampleType = s.DefaultSampleType + } + } + + p := &Profile{ + SampleType: make([]*ValueType, len(srcs[0].SampleType)), + + DropFrames: srcs[0].DropFrames, + KeepFrames: srcs[0].KeepFrames, + + TimeNanos: timeNanos, + DurationNanos: durationNanos, + PeriodType: srcs[0].PeriodType, + Period: period, + + Comments: comments, + DefaultSampleType: defaultSampleType, + } + copy(p.SampleType, srcs[0].SampleType) + return p, nil +} + +// compatible determines if two profiles can be compared/merged. +// returns nil if the profiles are compatible; otherwise an error with +// details on the incompatibility. +func (p *Profile) compatible(pb *Profile) error { + if !equalValueType(p.PeriodType, pb.PeriodType) { + return fmt.Errorf("incompatible period types %v and %v", p.PeriodType, pb.PeriodType) + } + + if len(p.SampleType) != len(pb.SampleType) { + return fmt.Errorf("incompatible sample types %v and %v", p.SampleType, pb.SampleType) + } + + for i := range p.SampleType { + if !equalValueType(p.SampleType[i], pb.SampleType[i]) { + return fmt.Errorf("incompatible sample types %v and %v", p.SampleType, pb.SampleType) + } + } + return nil +} + +// equalValueType returns true if the two value types are semantically +// equal. It ignores the internal fields used during encode/decode. +func equalValueType(st1, st2 *ValueType) bool { + return st1.Type == st2.Type && st1.Unit == st2.Unit +} diff --git a/src/internal/profile/profile.go b/src/internal/profile/profile.go new file mode 100644 index 0000000..c779bb2 --- /dev/null +++ b/src/internal/profile/profile.go @@ -0,0 +1,613 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package profile provides a representation of +// github.com/google/pprof/proto/profile.proto and +// methods to encode/decode/merge profiles in this format. +package profile + +import ( + "bytes" + "compress/gzip" + "fmt" + "internal/lazyregexp" + "io" + "strings" + "time" +) + +// Profile is an in-memory representation of profile.proto. +type Profile struct { + SampleType []*ValueType + DefaultSampleType string + Sample []*Sample + Mapping []*Mapping + Location []*Location + Function []*Function + Comments []string + + DropFrames string + KeepFrames string + + TimeNanos int64 + DurationNanos int64 + PeriodType *ValueType + Period int64 + + commentX []int64 + dropFramesX int64 + keepFramesX int64 + stringTable []string + defaultSampleTypeX int64 +} + +// ValueType corresponds to Profile.ValueType +type ValueType struct { + Type string // cpu, wall, inuse_space, etc + Unit string // seconds, nanoseconds, bytes, etc + + typeX int64 + unitX int64 +} + +// Sample corresponds to Profile.Sample +type Sample struct { + Location []*Location + Value []int64 + Label map[string][]string + NumLabel map[string][]int64 + NumUnit map[string][]string + + locationIDX []uint64 + labelX []Label +} + +// Label corresponds to Profile.Label +type Label struct { + keyX int64 + // Exactly one of the two following values must be set + strX int64 + numX int64 // Integer value for this label +} + +// Mapping corresponds to Profile.Mapping +type Mapping struct { + ID uint64 + Start uint64 + Limit uint64 + Offset uint64 + File string + BuildID string + HasFunctions bool + HasFilenames bool + HasLineNumbers bool + HasInlineFrames bool + + fileX int64 + buildIDX int64 +} + +// Location corresponds to Profile.Location +type Location struct { + ID uint64 + Mapping *Mapping + Address uint64 + Line []Line + IsFolded bool + + mappingIDX uint64 +} + +// Line corresponds to Profile.Line +type Line struct { + Function *Function + Line int64 + + functionIDX uint64 +} + +// Function corresponds to Profile.Function +type Function struct { + ID uint64 + Name string + SystemName string + Filename string + StartLine int64 + + nameX int64 + systemNameX int64 + filenameX int64 +} + +// Parse parses a profile and checks for its validity. The input +// may be a gzip-compressed encoded protobuf or one of many legacy +// profile formats which may be unsupported in the future. +func Parse(r io.Reader) (*Profile, error) { + orig, err := io.ReadAll(r) + if err != nil { + return nil, err + } + + var p *Profile + if len(orig) >= 2 && orig[0] == 0x1f && orig[1] == 0x8b { + gz, err := gzip.NewReader(bytes.NewBuffer(orig)) + if err != nil { + return nil, fmt.Errorf("decompressing profile: %v", err) + } + data, err := io.ReadAll(gz) + if err != nil { + return nil, fmt.Errorf("decompressing profile: %v", err) + } + orig = data + } + if p, err = parseUncompressed(orig); err != nil { + if p, err = parseLegacy(orig); err != nil { + return nil, fmt.Errorf("parsing profile: %v", err) + } + } + + if err := p.CheckValid(); err != nil { + return nil, fmt.Errorf("malformed profile: %v", err) + } + return p, nil +} + +var errUnrecognized = fmt.Errorf("unrecognized profile format") +var errMalformed = fmt.Errorf("malformed profile format") + +func parseLegacy(data []byte) (*Profile, error) { + parsers := []func([]byte) (*Profile, error){ + parseCPU, + parseHeap, + parseGoCount, // goroutine, threadcreate + parseThread, + parseContention, + } + + for _, parser := range parsers { + p, err := parser(data) + if err == nil { + p.setMain() + p.addLegacyFrameInfo() + return p, nil + } + if err != errUnrecognized { + return nil, err + } + } + return nil, errUnrecognized +} + +func parseUncompressed(data []byte) (*Profile, error) { + p := &Profile{} + if err := unmarshal(data, p); err != nil { + return nil, err + } + + if err := p.postDecode(); err != nil { + return nil, err + } + + return p, nil +} + +var libRx = lazyregexp.New(`([.]so$|[.]so[._][0-9]+)`) + +// setMain scans Mapping entries and guesses which entry is main +// because legacy profiles don't obey the convention of putting main +// first. +func (p *Profile) setMain() { + for i := 0; i < len(p.Mapping); i++ { + file := strings.TrimSpace(strings.ReplaceAll(p.Mapping[i].File, "(deleted)", "")) + if len(file) == 0 { + continue + } + if len(libRx.FindStringSubmatch(file)) > 0 { + continue + } + if strings.HasPrefix(file, "[") { + continue + } + // Swap what we guess is main to position 0. + p.Mapping[i], p.Mapping[0] = p.Mapping[0], p.Mapping[i] + break + } +} + +// Write writes the profile as a gzip-compressed marshaled protobuf. +func (p *Profile) Write(w io.Writer) error { + p.preEncode() + b := marshal(p) + zw := gzip.NewWriter(w) + defer zw.Close() + _, err := zw.Write(b) + return err +} + +// CheckValid tests whether the profile is valid. Checks include, but are +// not limited to: +// - len(Profile.Sample[n].value) == len(Profile.value_unit) +// - Sample.id has a corresponding Profile.Location +func (p *Profile) CheckValid() error { + // Check that sample values are consistent + sampleLen := len(p.SampleType) + if sampleLen == 0 && len(p.Sample) != 0 { + return fmt.Errorf("missing sample type information") + } + for _, s := range p.Sample { + if len(s.Value) != sampleLen { + return fmt.Errorf("mismatch: sample has: %d values vs. %d types", len(s.Value), len(p.SampleType)) + } + } + + // Check that all mappings/locations/functions are in the tables + // Check that there are no duplicate ids + mappings := make(map[uint64]*Mapping, len(p.Mapping)) + for _, m := range p.Mapping { + if m.ID == 0 { + return fmt.Errorf("found mapping with reserved ID=0") + } + if mappings[m.ID] != nil { + return fmt.Errorf("multiple mappings with same id: %d", m.ID) + } + mappings[m.ID] = m + } + functions := make(map[uint64]*Function, len(p.Function)) + for _, f := range p.Function { + if f.ID == 0 { + return fmt.Errorf("found function with reserved ID=0") + } + if functions[f.ID] != nil { + return fmt.Errorf("multiple functions with same id: %d", f.ID) + } + functions[f.ID] = f + } + locations := make(map[uint64]*Location, len(p.Location)) + for _, l := range p.Location { + if l.ID == 0 { + return fmt.Errorf("found location with reserved id=0") + } + if locations[l.ID] != nil { + return fmt.Errorf("multiple locations with same id: %d", l.ID) + } + locations[l.ID] = l + if m := l.Mapping; m != nil { + if m.ID == 0 || mappings[m.ID] != m { + return fmt.Errorf("inconsistent mapping %p: %d", m, m.ID) + } + } + for _, ln := range l.Line { + if f := ln.Function; f != nil { + if f.ID == 0 || functions[f.ID] != f { + return fmt.Errorf("inconsistent function %p: %d", f, f.ID) + } + } + } + } + return nil +} + +// Aggregate merges the locations in the profile into equivalence +// classes preserving the request attributes. It also updates the +// samples to point to the merged locations. +func (p *Profile) Aggregate(inlineFrame, function, filename, linenumber, address bool) error { + for _, m := range p.Mapping { + m.HasInlineFrames = m.HasInlineFrames && inlineFrame + m.HasFunctions = m.HasFunctions && function + m.HasFilenames = m.HasFilenames && filename + m.HasLineNumbers = m.HasLineNumbers && linenumber + } + + // Aggregate functions + if !function || !filename { + for _, f := range p.Function { + if !function { + f.Name = "" + f.SystemName = "" + } + if !filename { + f.Filename = "" + } + } + } + + // Aggregate locations + if !inlineFrame || !address || !linenumber { + for _, l := range p.Location { + if !inlineFrame && len(l.Line) > 1 { + l.Line = l.Line[len(l.Line)-1:] + } + if !linenumber { + for i := range l.Line { + l.Line[i].Line = 0 + } + } + if !address { + l.Address = 0 + } + } + } + + return p.CheckValid() +} + +// Print dumps a text representation of a profile. Intended mainly +// for debugging purposes. +func (p *Profile) String() string { + + ss := make([]string, 0, len(p.Sample)+len(p.Mapping)+len(p.Location)) + if pt := p.PeriodType; pt != nil { + ss = append(ss, fmt.Sprintf("PeriodType: %s %s", pt.Type, pt.Unit)) + } + ss = append(ss, fmt.Sprintf("Period: %d", p.Period)) + if p.TimeNanos != 0 { + ss = append(ss, fmt.Sprintf("Time: %v", time.Unix(0, p.TimeNanos))) + } + if p.DurationNanos != 0 { + ss = append(ss, fmt.Sprintf("Duration: %v", time.Duration(p.DurationNanos))) + } + + ss = append(ss, "Samples:") + var sh1 string + for _, s := range p.SampleType { + sh1 = sh1 + fmt.Sprintf("%s/%s ", s.Type, s.Unit) + } + ss = append(ss, strings.TrimSpace(sh1)) + for _, s := range p.Sample { + var sv string + for _, v := range s.Value { + sv = fmt.Sprintf("%s %10d", sv, v) + } + sv = sv + ": " + for _, l := range s.Location { + sv = sv + fmt.Sprintf("%d ", l.ID) + } + ss = append(ss, sv) + const labelHeader = " " + if len(s.Label) > 0 { + ls := labelHeader + for k, v := range s.Label { + ls = ls + fmt.Sprintf("%s:%v ", k, v) + } + ss = append(ss, ls) + } + if len(s.NumLabel) > 0 { + ls := labelHeader + for k, v := range s.NumLabel { + ls = ls + fmt.Sprintf("%s:%v ", k, v) + } + ss = append(ss, ls) + } + } + + ss = append(ss, "Locations") + for _, l := range p.Location { + locStr := fmt.Sprintf("%6d: %#x ", l.ID, l.Address) + if m := l.Mapping; m != nil { + locStr = locStr + fmt.Sprintf("M=%d ", m.ID) + } + if len(l.Line) == 0 { + ss = append(ss, locStr) + } + for li := range l.Line { + lnStr := "??" + if fn := l.Line[li].Function; fn != nil { + lnStr = fmt.Sprintf("%s %s:%d s=%d", + fn.Name, + fn.Filename, + l.Line[li].Line, + fn.StartLine) + if fn.Name != fn.SystemName { + lnStr = lnStr + "(" + fn.SystemName + ")" + } + } + ss = append(ss, locStr+lnStr) + // Do not print location details past the first line + locStr = " " + } + } + + ss = append(ss, "Mappings") + for _, m := range p.Mapping { + bits := "" + if m.HasFunctions { + bits += "[FN]" + } + if m.HasFilenames { + bits += "[FL]" + } + if m.HasLineNumbers { + bits += "[LN]" + } + if m.HasInlineFrames { + bits += "[IN]" + } + ss = append(ss, fmt.Sprintf("%d: %#x/%#x/%#x %s %s %s", + m.ID, + m.Start, m.Limit, m.Offset, + m.File, + m.BuildID, + bits)) + } + + return strings.Join(ss, "\n") + "\n" +} + +// Merge adds profile p adjusted by ratio r into profile p. Profiles +// must be compatible (same Type and SampleType). +// TODO(rsilvera): consider normalizing the profiles based on the +// total samples collected. +func (p *Profile) Merge(pb *Profile, r float64) error { + if err := p.Compatible(pb); err != nil { + return err + } + + pb = pb.Copy() + + // Keep the largest of the two periods. + if pb.Period > p.Period { + p.Period = pb.Period + } + + p.DurationNanos += pb.DurationNanos + + p.Mapping = append(p.Mapping, pb.Mapping...) + for i, m := range p.Mapping { + m.ID = uint64(i + 1) + } + p.Location = append(p.Location, pb.Location...) + for i, l := range p.Location { + l.ID = uint64(i + 1) + } + p.Function = append(p.Function, pb.Function...) + for i, f := range p.Function { + f.ID = uint64(i + 1) + } + + if r != 1.0 { + for _, s := range pb.Sample { + for i, v := range s.Value { + s.Value[i] = int64((float64(v) * r)) + } + } + } + p.Sample = append(p.Sample, pb.Sample...) + return p.CheckValid() +} + +// Compatible determines if two profiles can be compared/merged. +// returns nil if the profiles are compatible; otherwise an error with +// details on the incompatibility. +func (p *Profile) Compatible(pb *Profile) error { + if !compatibleValueTypes(p.PeriodType, pb.PeriodType) { + return fmt.Errorf("incompatible period types %v and %v", p.PeriodType, pb.PeriodType) + } + + if len(p.SampleType) != len(pb.SampleType) { + return fmt.Errorf("incompatible sample types %v and %v", p.SampleType, pb.SampleType) + } + + for i := range p.SampleType { + if !compatibleValueTypes(p.SampleType[i], pb.SampleType[i]) { + return fmt.Errorf("incompatible sample types %v and %v", p.SampleType, pb.SampleType) + } + } + + return nil +} + +// HasFunctions determines if all locations in this profile have +// symbolized function information. +func (p *Profile) HasFunctions() bool { + for _, l := range p.Location { + if l.Mapping == nil || !l.Mapping.HasFunctions { + return false + } + } + return true +} + +// HasFileLines determines if all locations in this profile have +// symbolized file and line number information. +func (p *Profile) HasFileLines() bool { + for _, l := range p.Location { + if l.Mapping == nil || (!l.Mapping.HasFilenames || !l.Mapping.HasLineNumbers) { + return false + } + } + return true +} + +func compatibleValueTypes(v1, v2 *ValueType) bool { + if v1 == nil || v2 == nil { + return true // No grounds to disqualify. + } + return v1.Type == v2.Type && v1.Unit == v2.Unit +} + +// Copy makes a fully independent copy of a profile. +func (p *Profile) Copy() *Profile { + p.preEncode() + b := marshal(p) + + pp := &Profile{} + if err := unmarshal(b, pp); err != nil { + panic(err) + } + if err := pp.postDecode(); err != nil { + panic(err) + } + + return pp +} + +// Demangler maps symbol names to a human-readable form. This may +// include C++ demangling and additional simplification. Names that +// are not demangled may be missing from the resulting map. +type Demangler func(name []string) (map[string]string, error) + +// Demangle attempts to demangle and optionally simplify any function +// names referenced in the profile. It works on a best-effort basis: +// it will silently preserve the original names in case of any errors. +func (p *Profile) Demangle(d Demangler) error { + // Collect names to demangle. + var names []string + for _, fn := range p.Function { + names = append(names, fn.SystemName) + } + + // Update profile with demangled names. + demangled, err := d(names) + if err != nil { + return err + } + for _, fn := range p.Function { + if dd, ok := demangled[fn.SystemName]; ok { + fn.Name = dd + } + } + return nil +} + +// Empty reports whether the profile contains no samples. +func (p *Profile) Empty() bool { + return len(p.Sample) == 0 +} + +// Scale multiplies all sample values in a profile by a constant. +func (p *Profile) Scale(ratio float64) { + if ratio == 1 { + return + } + ratios := make([]float64, len(p.SampleType)) + for i := range p.SampleType { + ratios[i] = ratio + } + p.ScaleN(ratios) +} + +// ScaleN multiplies each sample values in a sample by a different amount. +func (p *Profile) ScaleN(ratios []float64) error { + if len(p.SampleType) != len(ratios) { + return fmt.Errorf("mismatched scale ratios, got %d, want %d", len(ratios), len(p.SampleType)) + } + allOnes := true + for _, r := range ratios { + if r != 1 { + allOnes = false + break + } + } + if allOnes { + return nil + } + for _, s := range p.Sample { + for i, v := range s.Value { + if ratios[i] != 1 { + s.Value[i] = int64(float64(v) * ratios[i]) + } + } + } + return nil +} diff --git a/src/internal/profile/profile_test.go b/src/internal/profile/profile_test.go new file mode 100644 index 0000000..e1963f3 --- /dev/null +++ b/src/internal/profile/profile_test.go @@ -0,0 +1,79 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package profile + +import ( + "bytes" + "testing" +) + +func TestEmptyProfile(t *testing.T) { + var buf bytes.Buffer + p, err := Parse(&buf) + if err != nil { + t.Error("Want no error, got", err) + } + if p == nil { + t.Fatal("Want a valid profile, got <nil>") + } + if !p.Empty() { + t.Errorf("Profile should be empty, got %#v", p) + } +} + +func TestParseContention(t *testing.T) { + tests := []struct { + name string + in string + wantErr bool + }{ + { + name: "valid", + in: `--- mutex: +cycles/second=3491920901 +sampling period=1 +43227965305 1659640 @ 0x45e851 0x45f764 0x4a2be1 0x44ea31 +34035731690 15760 @ 0x45e851 0x45f764 0x4a2b17 0x44ea31 +`, + }, + { + name: "valid with comment", + in: `--- mutex: +cycles/second=3491920901 +sampling period=1 +43227965305 1659640 @ 0x45e851 0x45f764 0x4a2be1 0x44ea31 +# 0x45e850 sync.(*Mutex).Unlock+0x80 /go/src/sync/mutex.go:126 +# 0x45f763 sync.(*RWMutex).Unlock+0x83 /go/src/sync/rwmutex.go:125 +# 0x4a2be0 main.main.func3+0x70 /go/src/internal/pprof/profile/a_binary.go:58 + +34035731690 15760 @ 0x45e851 0x45f764 0x4a2b17 0x44ea31 +# 0x45e850 sync.(*Mutex).Unlock+0x80 /go/src/sync/mutex.go:126 +# 0x45f763 sync.(*RWMutex).Unlock+0x83 /go/src/sync/rwmutex.go:125 +# 0x4a2b16 main.main.func2+0xd6 /go/src/internal/pprof/profile/a_binary.go:48 +`, + }, + { + name: "empty", + in: `--- mutex:`, + wantErr: true, + }, + { + name: "invalid header", + in: `--- channel: +43227965305 1659640 @ 0x45e851 0x45f764 0x4a2be1 0x44ea31`, + wantErr: true, + }, + } + for _, tc := range tests { + _, err := parseContention([]byte(tc.in)) + if tc.wantErr && err == nil { + t.Errorf("parseContention(%q) succeeded unexpectedly", tc.name) + } + if !tc.wantErr && err != nil { + t.Errorf("parseContention(%q) failed unexpectedly: %v", tc.name, err) + } + } + +} diff --git a/src/internal/profile/proto.go b/src/internal/profile/proto.go new file mode 100644 index 0000000..58ff0ad --- /dev/null +++ b/src/internal/profile/proto.go @@ -0,0 +1,356 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is a simple protocol buffer encoder and decoder. +// +// A protocol message must implement the message interface: +// decoder() []decoder +// encode(*buffer) +// +// The decode method returns a slice indexed by field number that gives the +// function to decode that field. +// The encode method encodes its receiver into the given buffer. +// +// The two methods are simple enough to be implemented by hand rather than +// by using a protocol compiler. +// +// See profile.go for examples of messages implementing this interface. +// +// There is no support for groups, message sets, or "has" bits. + +package profile + +import ( + "errors" + "fmt" +) + +type buffer struct { + field int + typ int + u64 uint64 + data []byte + tmp [16]byte +} + +type decoder func(*buffer, message) error + +type message interface { + decoder() []decoder + encode(*buffer) +} + +func marshal(m message) []byte { + var b buffer + m.encode(&b) + return b.data +} + +func encodeVarint(b *buffer, x uint64) { + for x >= 128 { + b.data = append(b.data, byte(x)|0x80) + x >>= 7 + } + b.data = append(b.data, byte(x)) +} + +func encodeLength(b *buffer, tag int, len int) { + encodeVarint(b, uint64(tag)<<3|2) + encodeVarint(b, uint64(len)) +} + +func encodeUint64(b *buffer, tag int, x uint64) { + // append varint to b.data + encodeVarint(b, uint64(tag)<<3|0) + encodeVarint(b, x) +} + +func encodeUint64s(b *buffer, tag int, x []uint64) { + if len(x) > 2 { + // Use packed encoding + n1 := len(b.data) + for _, u := range x { + encodeVarint(b, u) + } + n2 := len(b.data) + encodeLength(b, tag, n2-n1) + n3 := len(b.data) + copy(b.tmp[:], b.data[n2:n3]) + copy(b.data[n1+(n3-n2):], b.data[n1:n2]) + copy(b.data[n1:], b.tmp[:n3-n2]) + return + } + for _, u := range x { + encodeUint64(b, tag, u) + } +} + +func encodeUint64Opt(b *buffer, tag int, x uint64) { + if x == 0 { + return + } + encodeUint64(b, tag, x) +} + +func encodeInt64(b *buffer, tag int, x int64) { + u := uint64(x) + encodeUint64(b, tag, u) +} + +func encodeInt64Opt(b *buffer, tag int, x int64) { + if x == 0 { + return + } + encodeInt64(b, tag, x) +} + +func encodeInt64s(b *buffer, tag int, x []int64) { + if len(x) > 2 { + // Use packed encoding + n1 := len(b.data) + for _, u := range x { + encodeVarint(b, uint64(u)) + } + n2 := len(b.data) + encodeLength(b, tag, n2-n1) + n3 := len(b.data) + copy(b.tmp[:], b.data[n2:n3]) + copy(b.data[n1+(n3-n2):], b.data[n1:n2]) + copy(b.data[n1:], b.tmp[:n3-n2]) + return + } + for _, u := range x { + encodeInt64(b, tag, u) + } +} + +func encodeString(b *buffer, tag int, x string) { + encodeLength(b, tag, len(x)) + b.data = append(b.data, x...) +} + +func encodeStrings(b *buffer, tag int, x []string) { + for _, s := range x { + encodeString(b, tag, s) + } +} + +func encodeBool(b *buffer, tag int, x bool) { + if x { + encodeUint64(b, tag, 1) + } else { + encodeUint64(b, tag, 0) + } +} + +func encodeBoolOpt(b *buffer, tag int, x bool) { + if !x { + return + } + encodeBool(b, tag, x) +} + +func encodeMessage(b *buffer, tag int, m message) { + n1 := len(b.data) + m.encode(b) + n2 := len(b.data) + encodeLength(b, tag, n2-n1) + n3 := len(b.data) + copy(b.tmp[:], b.data[n2:n3]) + copy(b.data[n1+(n3-n2):], b.data[n1:n2]) + copy(b.data[n1:], b.tmp[:n3-n2]) +} + +func unmarshal(data []byte, m message) (err error) { + b := buffer{data: data, typ: 2} + return decodeMessage(&b, m) +} + +func le64(p []byte) uint64 { + return uint64(p[0]) | uint64(p[1])<<8 | uint64(p[2])<<16 | uint64(p[3])<<24 | uint64(p[4])<<32 | uint64(p[5])<<40 | uint64(p[6])<<48 | uint64(p[7])<<56 +} + +func le32(p []byte) uint32 { + return uint32(p[0]) | uint32(p[1])<<8 | uint32(p[2])<<16 | uint32(p[3])<<24 +} + +func decodeVarint(data []byte) (uint64, []byte, error) { + var i int + var u uint64 + for i = 0; ; i++ { + if i >= 10 || i >= len(data) { + return 0, nil, errors.New("bad varint") + } + u |= uint64(data[i]&0x7F) << uint(7*i) + if data[i]&0x80 == 0 { + return u, data[i+1:], nil + } + } +} + +func decodeField(b *buffer, data []byte) ([]byte, error) { + x, data, err := decodeVarint(data) + if err != nil { + return nil, err + } + b.field = int(x >> 3) + b.typ = int(x & 7) + b.data = nil + b.u64 = 0 + switch b.typ { + case 0: + b.u64, data, err = decodeVarint(data) + if err != nil { + return nil, err + } + case 1: + if len(data) < 8 { + return nil, errors.New("not enough data") + } + b.u64 = le64(data[:8]) + data = data[8:] + case 2: + var n uint64 + n, data, err = decodeVarint(data) + if err != nil { + return nil, err + } + if n > uint64(len(data)) { + return nil, errors.New("too much data") + } + b.data = data[:n] + data = data[n:] + case 5: + if len(data) < 4 { + return nil, errors.New("not enough data") + } + b.u64 = uint64(le32(data[:4])) + data = data[4:] + default: + return nil, fmt.Errorf("unknown wire type: %d", b.typ) + } + + return data, nil +} + +func checkType(b *buffer, typ int) error { + if b.typ != typ { + return errors.New("type mismatch") + } + return nil +} + +func decodeMessage(b *buffer, m message) error { + if err := checkType(b, 2); err != nil { + return err + } + dec := m.decoder() + data := b.data + for len(data) > 0 { + // pull varint field# + type + var err error + data, err = decodeField(b, data) + if err != nil { + return err + } + if b.field >= len(dec) || dec[b.field] == nil { + continue + } + if err := dec[b.field](b, m); err != nil { + return err + } + } + return nil +} + +func decodeInt64(b *buffer, x *int64) error { + if err := checkType(b, 0); err != nil { + return err + } + *x = int64(b.u64) + return nil +} + +func decodeInt64s(b *buffer, x *[]int64) error { + if b.typ == 2 { + // Packed encoding + data := b.data + for len(data) > 0 { + var u uint64 + var err error + + if u, data, err = decodeVarint(data); err != nil { + return err + } + *x = append(*x, int64(u)) + } + return nil + } + var i int64 + if err := decodeInt64(b, &i); err != nil { + return err + } + *x = append(*x, i) + return nil +} + +func decodeUint64(b *buffer, x *uint64) error { + if err := checkType(b, 0); err != nil { + return err + } + *x = b.u64 + return nil +} + +func decodeUint64s(b *buffer, x *[]uint64) error { + if b.typ == 2 { + data := b.data + // Packed encoding + for len(data) > 0 { + var u uint64 + var err error + + if u, data, err = decodeVarint(data); err != nil { + return err + } + *x = append(*x, u) + } + return nil + } + var u uint64 + if err := decodeUint64(b, &u); err != nil { + return err + } + *x = append(*x, u) + return nil +} + +func decodeString(b *buffer, x *string) error { + if err := checkType(b, 2); err != nil { + return err + } + *x = string(b.data) + return nil +} + +func decodeStrings(b *buffer, x *[]string) error { + var s string + if err := decodeString(b, &s); err != nil { + return err + } + *x = append(*x, s) + return nil +} + +func decodeBool(b *buffer, x *bool) error { + if err := checkType(b, 0); err != nil { + return err + } + if int64(b.u64) == 0 { + *x = false + } else { + *x = true + } + return nil +} diff --git a/src/internal/profile/proto_test.go b/src/internal/profile/proto_test.go new file mode 100644 index 0000000..46c6d83 --- /dev/null +++ b/src/internal/profile/proto_test.go @@ -0,0 +1,71 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package profile + +import ( + "reflect" + "testing" +) + +func TestPackedEncoding(t *testing.T) { + + type testcase struct { + uint64s []uint64 + int64s []int64 + encoded []byte + } + for i, tc := range []testcase{ + { + []uint64{0, 1, 10, 100, 1000, 10000}, + []int64{1000, 0, 1000}, + []byte{10, 8, 0, 1, 10, 100, 232, 7, 144, 78, 18, 5, 232, 7, 0, 232, 7}, + }, + { + []uint64{10000}, + nil, + []byte{8, 144, 78}, + }, + { + nil, + []int64{-10000}, + []byte{16, 240, 177, 255, 255, 255, 255, 255, 255, 255, 1}, + }, + } { + source := &packedInts{tc.uint64s, tc.int64s} + if got, want := marshal(source), tc.encoded; !reflect.DeepEqual(got, want) { + t.Errorf("failed encode %d, got %v, want %v", i, got, want) + } + + dest := new(packedInts) + if err := unmarshal(tc.encoded, dest); err != nil { + t.Errorf("failed decode %d: %v", i, err) + continue + } + if got, want := dest.uint64s, tc.uint64s; !reflect.DeepEqual(got, want) { + t.Errorf("failed decode uint64s %d, got %v, want %v", i, got, want) + } + if got, want := dest.int64s, tc.int64s; !reflect.DeepEqual(got, want) { + t.Errorf("failed decode int64s %d, got %v, want %v", i, got, want) + } + } +} + +type packedInts struct { + uint64s []uint64 + int64s []int64 +} + +func (u *packedInts) decoder() []decoder { + return []decoder{ + nil, + func(b *buffer, m message) error { return decodeUint64s(b, &m.(*packedInts).uint64s) }, + func(b *buffer, m message) error { return decodeInt64s(b, &m.(*packedInts).int64s) }, + } +} + +func (u *packedInts) encode(b *buffer) { + encodeUint64s(b, 1, u.uint64s) + encodeInt64s(b, 2, u.int64s) +} diff --git a/src/internal/profile/prune.go b/src/internal/profile/prune.go new file mode 100644 index 0000000..1924fad --- /dev/null +++ b/src/internal/profile/prune.go @@ -0,0 +1,97 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Implements methods to remove frames from profiles. + +package profile + +import ( + "fmt" + "regexp" +) + +// Prune removes all nodes beneath a node matching dropRx, and not +// matching keepRx. If the root node of a Sample matches, the sample +// will have an empty stack. +func (p *Profile) Prune(dropRx, keepRx *regexp.Regexp) { + prune := make(map[uint64]bool) + pruneBeneath := make(map[uint64]bool) + + for _, loc := range p.Location { + var i int + for i = len(loc.Line) - 1; i >= 0; i-- { + if fn := loc.Line[i].Function; fn != nil && fn.Name != "" { + funcName := fn.Name + // Account for leading '.' on the PPC ELF v1 ABI. + if funcName[0] == '.' { + funcName = funcName[1:] + } + if dropRx.MatchString(funcName) { + if keepRx == nil || !keepRx.MatchString(funcName) { + break + } + } + } + } + + if i >= 0 { + // Found matching entry to prune. + pruneBeneath[loc.ID] = true + + // Remove the matching location. + if i == len(loc.Line)-1 { + // Matched the top entry: prune the whole location. + prune[loc.ID] = true + } else { + loc.Line = loc.Line[i+1:] + } + } + } + + // Prune locs from each Sample + for _, sample := range p.Sample { + // Scan from the root to the leaves to find the prune location. + // Do not prune frames before the first user frame, to avoid + // pruning everything. + foundUser := false + for i := len(sample.Location) - 1; i >= 0; i-- { + id := sample.Location[i].ID + if !prune[id] && !pruneBeneath[id] { + foundUser = true + continue + } + if !foundUser { + continue + } + if prune[id] { + sample.Location = sample.Location[i+1:] + break + } + if pruneBeneath[id] { + sample.Location = sample.Location[i:] + break + } + } + } +} + +// RemoveUninteresting prunes and elides profiles using built-in +// tables of uninteresting function names. +func (p *Profile) RemoveUninteresting() error { + var keep, drop *regexp.Regexp + var err error + + if p.DropFrames != "" { + if drop, err = regexp.Compile("^(" + p.DropFrames + ")$"); err != nil { + return fmt.Errorf("failed to compile regexp %s: %v", p.DropFrames, err) + } + if p.KeepFrames != "" { + if keep, err = regexp.Compile("^(" + p.KeepFrames + ")$"); err != nil { + return fmt.Errorf("failed to compile regexp %s: %v", p.KeepFrames, err) + } + } + p.Prune(drop, keep) + } + return nil +} diff --git a/src/internal/race/doc.go b/src/internal/race/doc.go new file mode 100644 index 0000000..8fa44ce --- /dev/null +++ b/src/internal/race/doc.go @@ -0,0 +1,11 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +/* +Package race contains helper functions for manually instrumenting code for the race detector. + +The runtime package intentionally exports these functions only in the race build; +this package exports them unconditionally but without the "race" build tag they are no-ops. +*/ +package race diff --git a/src/internal/race/norace.go b/src/internal/race/norace.go new file mode 100644 index 0000000..67b1305 --- /dev/null +++ b/src/internal/race/norace.go @@ -0,0 +1,43 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !race +// +build !race + +package race + +import ( + "unsafe" +) + +const Enabled = false + +func Acquire(addr unsafe.Pointer) { +} + +func Release(addr unsafe.Pointer) { +} + +func ReleaseMerge(addr unsafe.Pointer) { +} + +func Disable() { +} + +func Enable() { +} + +func Read(addr unsafe.Pointer) { +} + +func Write(addr unsafe.Pointer) { +} + +func ReadRange(addr unsafe.Pointer, len int) { +} + +func WriteRange(addr unsafe.Pointer, len int) { +} + +func Errors() int { return 0 } diff --git a/src/internal/race/race.go b/src/internal/race/race.go new file mode 100644 index 0000000..40f2c99 --- /dev/null +++ b/src/internal/race/race.go @@ -0,0 +1,55 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build race +// +build race + +package race + +import ( + "runtime" + "unsafe" +) + +const Enabled = true + +func Acquire(addr unsafe.Pointer) { + runtime.RaceAcquire(addr) +} + +func Release(addr unsafe.Pointer) { + runtime.RaceRelease(addr) +} + +func ReleaseMerge(addr unsafe.Pointer) { + runtime.RaceReleaseMerge(addr) +} + +func Disable() { + runtime.RaceDisable() +} + +func Enable() { + runtime.RaceEnable() +} + +func Read(addr unsafe.Pointer) { + runtime.RaceRead(addr) +} + +func Write(addr unsafe.Pointer) { + runtime.RaceWrite(addr) +} + +func ReadRange(addr unsafe.Pointer, len int) { + runtime.RaceReadRange(addr, len) +} + +func WriteRange(addr unsafe.Pointer, len int) { + runtime.RaceWriteRange(addr, len) +} + +func Errors() int { + return runtime.RaceErrors() +} diff --git a/src/internal/reflectlite/all_test.go b/src/internal/reflectlite/all_test.go new file mode 100644 index 0000000..820b4ae --- /dev/null +++ b/src/internal/reflectlite/all_test.go @@ -0,0 +1,1039 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite_test + +import ( + "encoding/base64" + "fmt" + "internal/abi" + . "internal/reflectlite" + "math" + "reflect" + "runtime" + "testing" + "unsafe" +) + +func ToValue(v Value) reflect.Value { + return reflect.ValueOf(ToInterface(v)) +} + +func TypeString(t Type) string { + return fmt.Sprintf("%T", ToInterface(Zero(t))) +} + +type integer int +type T struct { + a int + b float64 + c string + d *int +} + +type pair struct { + i any + s string +} + +func assert(t *testing.T, s, want string) { + t.Helper() + if s != want { + t.Errorf("have %#q want %#q", s, want) + } +} + +var typeTests = []pair{ + {struct{ x int }{}, "int"}, + {struct{ x int8 }{}, "int8"}, + {struct{ x int16 }{}, "int16"}, + {struct{ x int32 }{}, "int32"}, + {struct{ x int64 }{}, "int64"}, + {struct{ x uint }{}, "uint"}, + {struct{ x uint8 }{}, "uint8"}, + {struct{ x uint16 }{}, "uint16"}, + {struct{ x uint32 }{}, "uint32"}, + {struct{ x uint64 }{}, "uint64"}, + {struct{ x float32 }{}, "float32"}, + {struct{ x float64 }{}, "float64"}, + {struct{ x int8 }{}, "int8"}, + {struct{ x (**int8) }{}, "**int8"}, + {struct{ x (**integer) }{}, "**reflectlite_test.integer"}, + {struct{ x ([32]int32) }{}, "[32]int32"}, + {struct{ x ([]int8) }{}, "[]int8"}, + {struct{ x (map[string]int32) }{}, "map[string]int32"}, + {struct{ x (chan<- string) }{}, "chan<- string"}, + {struct { + x struct { + c chan *int32 + d float32 + } + }{}, + "struct { c chan *int32; d float32 }", + }, + {struct{ x (func(a int8, b int32)) }{}, "func(int8, int32)"}, + {struct { + x struct { + c func(chan *integer, *int8) + } + }{}, + "struct { c func(chan *reflectlite_test.integer, *int8) }", + }, + {struct { + x struct { + a int8 + b int32 + } + }{}, + "struct { a int8; b int32 }", + }, + {struct { + x struct { + a int8 + b int8 + c int32 + } + }{}, + "struct { a int8; b int8; c int32 }", + }, + {struct { + x struct { + a int8 + b int8 + c int8 + d int32 + } + }{}, + "struct { a int8; b int8; c int8; d int32 }", + }, + {struct { + x struct { + a int8 + b int8 + c int8 + d int8 + e int32 + } + }{}, + "struct { a int8; b int8; c int8; d int8; e int32 }", + }, + {struct { + x struct { + a int8 + b int8 + c int8 + d int8 + e int8 + f int32 + } + }{}, + "struct { a int8; b int8; c int8; d int8; e int8; f int32 }", + }, + {struct { + x struct { + a int8 `reflect:"hi there"` + } + }{}, + `struct { a int8 "reflect:\"hi there\"" }`, + }, + {struct { + x struct { + a int8 `reflect:"hi \x00there\t\n\"\\"` + } + }{}, + `struct { a int8 "reflect:\"hi \\x00there\\t\\n\\\"\\\\\"" }`, + }, + {struct { + x struct { + f func(args ...int) + } + }{}, + "struct { f func(...int) }", + }, + // {struct { + // x (interface { + // a(func(func(int) int) func(func(int)) int) + // b() + // }) + // }{}, + // "interface { reflectlite_test.a(func(func(int) int) func(func(int)) int); reflectlite_test.b() }", + // }, + {struct { + x struct { + int32 + int64 + } + }{}, + "struct { int32; int64 }", + }, +} + +var valueTests = []pair{ + {new(int), "132"}, + {new(int8), "8"}, + {new(int16), "16"}, + {new(int32), "32"}, + {new(int64), "64"}, + {new(uint), "132"}, + {new(uint8), "8"}, + {new(uint16), "16"}, + {new(uint32), "32"}, + {new(uint64), "64"}, + {new(float32), "256.25"}, + {new(float64), "512.125"}, + {new(complex64), "532.125+10i"}, + {new(complex128), "564.25+1i"}, + {new(string), "stringy cheese"}, + {new(bool), "true"}, + {new(*int8), "*int8(0)"}, + {new(**int8), "**int8(0)"}, + {new([5]int32), "[5]int32{0, 0, 0, 0, 0}"}, + {new(**integer), "**reflectlite_test.integer(0)"}, + {new(map[string]int32), "map[string]int32{<can't iterate on maps>}"}, + {new(chan<- string), "chan<- string"}, + {new(func(a int8, b int32)), "func(int8, int32)(arg)"}, + {new(struct { + c chan *int32 + d float32 + }), + "struct { c chan *int32; d float32 }{chan *int32, 0}", + }, + {new(struct{ c func(chan *integer, *int8) }), + "struct { c func(chan *reflectlite_test.integer, *int8) }{func(chan *reflectlite_test.integer, *int8)(arg)}", + }, + {new(struct { + a int8 + b int32 + }), + "struct { a int8; b int32 }{0, 0}", + }, + {new(struct { + a int8 + b int8 + c int32 + }), + "struct { a int8; b int8; c int32 }{0, 0, 0}", + }, +} + +func testType(t *testing.T, i int, typ Type, want string) { + s := TypeString(typ) + if s != want { + t.Errorf("#%d: have %#q, want %#q", i, s, want) + } +} + +func testReflectType(t *testing.T, i int, typ Type, want string) { + s := TypeString(typ) + if s != want { + t.Errorf("#%d: have %#q, want %#q", i, s, want) + } +} + +func TestTypes(t *testing.T) { + for i, tt := range typeTests { + testReflectType(t, i, Field(ValueOf(tt.i), 0).Type(), tt.s) + } +} + +func TestSetValue(t *testing.T) { + for i, tt := range valueTests { + v := ValueOf(tt.i).Elem() + switch v.Kind() { + case abi.Int: + v.Set(ValueOf(int(132))) + case abi.Int8: + v.Set(ValueOf(int8(8))) + case abi.Int16: + v.Set(ValueOf(int16(16))) + case abi.Int32: + v.Set(ValueOf(int32(32))) + case abi.Int64: + v.Set(ValueOf(int64(64))) + case abi.Uint: + v.Set(ValueOf(uint(132))) + case abi.Uint8: + v.Set(ValueOf(uint8(8))) + case abi.Uint16: + v.Set(ValueOf(uint16(16))) + case abi.Uint32: + v.Set(ValueOf(uint32(32))) + case abi.Uint64: + v.Set(ValueOf(uint64(64))) + case abi.Float32: + v.Set(ValueOf(float32(256.25))) + case abi.Float64: + v.Set(ValueOf(512.125)) + case abi.Complex64: + v.Set(ValueOf(complex64(532.125 + 10i))) + case abi.Complex128: + v.Set(ValueOf(complex128(564.25 + 1i))) + case abi.String: + v.Set(ValueOf("stringy cheese")) + case abi.Bool: + v.Set(ValueOf(true)) + } + s := valueToString(v) + if s != tt.s { + t.Errorf("#%d: have %#q, want %#q", i, s, tt.s) + } + } +} + +func TestCanSetField(t *testing.T) { + type embed struct{ x, X int } + type Embed struct{ x, X int } + type S1 struct { + embed + x, X int + } + type S2 struct { + *embed + x, X int + } + type S3 struct { + Embed + x, X int + } + type S4 struct { + *Embed + x, X int + } + + type testCase struct { + index []int + canSet bool + } + tests := []struct { + val Value + cases []testCase + }{{ + val: ValueOf(&S1{}), + cases: []testCase{ + {[]int{0}, false}, + {[]int{0, 0}, false}, + {[]int{0, 1}, true}, + {[]int{1}, false}, + {[]int{2}, true}, + }, + }, { + val: ValueOf(&S2{embed: &embed{}}), + cases: []testCase{ + {[]int{0}, false}, + {[]int{0, 0}, false}, + {[]int{0, 1}, true}, + {[]int{1}, false}, + {[]int{2}, true}, + }, + }, { + val: ValueOf(&S3{}), + cases: []testCase{ + {[]int{0}, true}, + {[]int{0, 0}, false}, + {[]int{0, 1}, true}, + {[]int{1}, false}, + {[]int{2}, true}, + }, + }, { + val: ValueOf(&S4{Embed: &Embed{}}), + cases: []testCase{ + {[]int{0}, true}, + {[]int{0, 0}, false}, + {[]int{0, 1}, true}, + {[]int{1}, false}, + {[]int{2}, true}, + }, + }} + + for _, tt := range tests { + t.Run(tt.val.Type().Name(), func(t *testing.T) { + for _, tc := range tt.cases { + f := tt.val + for _, i := range tc.index { + if f.Kind() == Ptr { + f = f.Elem() + } + f = Field(f, i) + } + if got := f.CanSet(); got != tc.canSet { + t.Errorf("CanSet() = %v, want %v", got, tc.canSet) + } + } + }) + } +} + +var _i = 7 + +var valueToStringTests = []pair{ + {123, "123"}, + {123.5, "123.5"}, + {byte(123), "123"}, + {"abc", "abc"}, + {T{123, 456.75, "hello", &_i}, "reflectlite_test.T{123, 456.75, hello, *int(&7)}"}, + {new(chan *T), "*chan *reflectlite_test.T(&chan *reflectlite_test.T)"}, + {[10]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, "[10]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}"}, + {&[10]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, "*[10]int(&[10]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10})"}, + {[]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, "[]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}"}, + {&[]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10}, "*[]int(&[]int{1, 2, 3, 4, 5, 6, 7, 8, 9, 10})"}, +} + +func TestValueToString(t *testing.T) { + for i, test := range valueToStringTests { + s := valueToString(ValueOf(test.i)) + if s != test.s { + t.Errorf("#%d: have %#q, want %#q", i, s, test.s) + } + } +} + +func TestPtrSetNil(t *testing.T) { + var i int32 = 1234 + ip := &i + vip := ValueOf(&ip) + vip.Elem().Set(Zero(vip.Elem().Type())) + if ip != nil { + t.Errorf("got non-nil (%d), want nil", *ip) + } +} + +func TestMapSetNil(t *testing.T) { + m := make(map[string]int) + vm := ValueOf(&m) + vm.Elem().Set(Zero(vm.Elem().Type())) + if m != nil { + t.Errorf("got non-nil (%p), want nil", m) + } +} + +func TestAll(t *testing.T) { + testType(t, 1, TypeOf((int8)(0)), "int8") + testType(t, 2, TypeOf((*int8)(nil)).Elem(), "int8") + + typ := TypeOf((*struct { + c chan *int32 + d float32 + })(nil)) + testType(t, 3, typ, "*struct { c chan *int32; d float32 }") + etyp := typ.Elem() + testType(t, 4, etyp, "struct { c chan *int32; d float32 }") +} + +func TestInterfaceValue(t *testing.T) { + var inter struct { + E any + } + inter.E = 123.456 + v1 := ValueOf(&inter) + v2 := Field(v1.Elem(), 0) + // assert(t, TypeString(v2.Type()), "interface {}") + v3 := v2.Elem() + assert(t, TypeString(v3.Type()), "float64") + + i3 := ToInterface(v2) + if _, ok := i3.(float64); !ok { + t.Error("v2.Interface() did not return float64, got ", TypeOf(i3)) + } +} + +func TestFunctionValue(t *testing.T) { + var x any = func() {} + v := ValueOf(x) + if fmt.Sprint(ToInterface(v)) != fmt.Sprint(x) { + t.Fatalf("TestFunction returned wrong pointer") + } + assert(t, TypeString(v.Type()), "func()") +} + +var appendTests = []struct { + orig, extra []int +}{ + {make([]int, 2, 4), []int{22}}, + {make([]int, 2, 4), []int{22, 33, 44}}, +} + +func sameInts(x, y []int) bool { + if len(x) != len(y) { + return false + } + for i, xx := range x { + if xx != y[i] { + return false + } + } + return true +} + +func TestBigUnnamedStruct(t *testing.T) { + b := struct{ a, b, c, d int64 }{1, 2, 3, 4} + v := ValueOf(b) + b1 := ToInterface(v).(struct { + a, b, c, d int64 + }) + if b1.a != b.a || b1.b != b.b || b1.c != b.c || b1.d != b.d { + t.Errorf("ValueOf(%v).Interface().(*Big) = %v", b, b1) + } +} + +type big struct { + a, b, c, d, e int64 +} + +func TestBigStruct(t *testing.T) { + b := big{1, 2, 3, 4, 5} + v := ValueOf(b) + b1 := ToInterface(v).(big) + if b1.a != b.a || b1.b != b.b || b1.c != b.c || b1.d != b.d || b1.e != b.e { + t.Errorf("ValueOf(%v).Interface().(big) = %v", b, b1) + } +} + +type Basic struct { + x int + y float32 +} + +type NotBasic Basic + +type DeepEqualTest struct { + a, b any + eq bool +} + +// Simple functions for DeepEqual tests. +var ( + fn1 func() // nil. + fn2 func() // nil. + fn3 = func() { fn1() } // Not nil. +) + +type self struct{} + +type Loop *Loop +type Loopy any + +var loop1, loop2 Loop +var loopy1, loopy2 Loopy + +func init() { + loop1 = &loop2 + loop2 = &loop1 + + loopy1 = &loopy2 + loopy2 = &loopy1 +} + +var typeOfTests = []DeepEqualTest{ + // Equalities + {nil, nil, true}, + {1, 1, true}, + {int32(1), int32(1), true}, + {0.5, 0.5, true}, + {float32(0.5), float32(0.5), true}, + {"hello", "hello", true}, + {make([]int, 10), make([]int, 10), true}, + {&[3]int{1, 2, 3}, &[3]int{1, 2, 3}, true}, + {Basic{1, 0.5}, Basic{1, 0.5}, true}, + {error(nil), error(nil), true}, + {map[int]string{1: "one", 2: "two"}, map[int]string{2: "two", 1: "one"}, true}, + {fn1, fn2, true}, + + // Inequalities + {1, 2, false}, + {int32(1), int32(2), false}, + {0.5, 0.6, false}, + {float32(0.5), float32(0.6), false}, + {"hello", "hey", false}, + {make([]int, 10), make([]int, 11), false}, + {&[3]int{1, 2, 3}, &[3]int{1, 2, 4}, false}, + {Basic{1, 0.5}, Basic{1, 0.6}, false}, + {Basic{1, 0}, Basic{2, 0}, false}, + {map[int]string{1: "one", 3: "two"}, map[int]string{2: "two", 1: "one"}, false}, + {map[int]string{1: "one", 2: "txo"}, map[int]string{2: "two", 1: "one"}, false}, + {map[int]string{1: "one"}, map[int]string{2: "two", 1: "one"}, false}, + {map[int]string{2: "two", 1: "one"}, map[int]string{1: "one"}, false}, + {nil, 1, false}, + {1, nil, false}, + {fn1, fn3, false}, + {fn3, fn3, false}, + {[][]int{{1}}, [][]int{{2}}, false}, + {math.NaN(), math.NaN(), false}, + {&[1]float64{math.NaN()}, &[1]float64{math.NaN()}, false}, + {&[1]float64{math.NaN()}, self{}, true}, + {[]float64{math.NaN()}, []float64{math.NaN()}, false}, + {[]float64{math.NaN()}, self{}, true}, + {map[float64]float64{math.NaN(): 1}, map[float64]float64{1: 2}, false}, + {map[float64]float64{math.NaN(): 1}, self{}, true}, + + // Nil vs empty: not the same. + {[]int{}, []int(nil), false}, + {[]int{}, []int{}, true}, + {[]int(nil), []int(nil), true}, + {map[int]int{}, map[int]int(nil), false}, + {map[int]int{}, map[int]int{}, true}, + {map[int]int(nil), map[int]int(nil), true}, + + // Mismatched types + {1, 1.0, false}, + {int32(1), int64(1), false}, + {0.5, "hello", false}, + {[]int{1, 2, 3}, [3]int{1, 2, 3}, false}, + {&[3]any{1, 2, 4}, &[3]any{1, 2, "s"}, false}, + {Basic{1, 0.5}, NotBasic{1, 0.5}, false}, + {map[uint]string{1: "one", 2: "two"}, map[int]string{2: "two", 1: "one"}, false}, + + // Possible loops. + {&loop1, &loop1, true}, + {&loop1, &loop2, true}, + {&loopy1, &loopy1, true}, + {&loopy1, &loopy2, true}, +} + +func TestTypeOf(t *testing.T) { + // Special case for nil + if typ := TypeOf(nil); typ != nil { + t.Errorf("expected nil type for nil value; got %v", typ) + } + for _, test := range typeOfTests { + v := ValueOf(test.a) + if !v.IsValid() { + continue + } + typ := TypeOf(test.a) + if typ != v.Type() { + t.Errorf("TypeOf(%v) = %v, but ValueOf(%v).Type() = %v", test.a, typ, test.a, v.Type()) + } + } +} + +func Nil(a any, t *testing.T) { + n := Field(ValueOf(a), 0) + if !n.IsNil() { + t.Errorf("%v should be nil", a) + } +} + +func NotNil(a any, t *testing.T) { + n := Field(ValueOf(a), 0) + if n.IsNil() { + t.Errorf("value of type %v should not be nil", TypeString(ValueOf(a).Type())) + } +} + +func TestIsNil(t *testing.T) { + // These implement IsNil. + // Wrap in extra struct to hide interface type. + doNil := []any{ + struct{ x *int }{}, + struct{ x any }{}, + struct{ x map[string]int }{}, + struct{ x func() bool }{}, + struct{ x chan int }{}, + struct{ x []string }{}, + struct{ x unsafe.Pointer }{}, + } + for _, ts := range doNil { + ty := TField(TypeOf(ts), 0) + v := Zero(ty) + v.IsNil() // panics if not okay to call + } + + // Check the implementations + var pi struct { + x *int + } + Nil(pi, t) + pi.x = new(int) + NotNil(pi, t) + + var si struct { + x []int + } + Nil(si, t) + si.x = make([]int, 10) + NotNil(si, t) + + var ci struct { + x chan int + } + Nil(ci, t) + ci.x = make(chan int) + NotNil(ci, t) + + var mi struct { + x map[int]int + } + Nil(mi, t) + mi.x = make(map[int]int) + NotNil(mi, t) + + var ii struct { + x any + } + Nil(ii, t) + ii.x = 2 + NotNil(ii, t) + + var fi struct { + x func(t *testing.T) + } + Nil(fi, t) + fi.x = TestIsNil + NotNil(fi, t) +} + +// Indirect returns the value that v points to. +// If v is a nil pointer, Indirect returns a zero Value. +// If v is not a pointer, Indirect returns v. +func Indirect(v Value) Value { + if v.Kind() != Ptr { + return v + } + return v.Elem() +} + +func TestNilPtrValueSub(t *testing.T) { + var pi *int + if pv := ValueOf(pi); pv.Elem().IsValid() { + t.Error("ValueOf((*int)(nil)).Elem().IsValid()") + } +} + +type Point struct { + x, y int +} + +// This will be index 0. +func (p Point) AnotherMethod(scale int) int { + return -1 +} + +// This will be index 1. +func (p Point) Dist(scale int) int { + //println("Point.Dist", p.x, p.y, scale) + return p.x*p.x*scale + p.y*p.y*scale +} + +// This will be index 2. +func (p Point) GCMethod(k int) int { + runtime.GC() + return k + p.x +} + +// This will be index 3. +func (p Point) NoArgs() { + // Exercise no-argument/no-result paths. +} + +// This will be index 4. +func (p Point) TotalDist(points ...Point) int { + tot := 0 + for _, q := range points { + dx := q.x - p.x + dy := q.y - p.y + tot += dx*dx + dy*dy // Should call Sqrt, but it's just a test. + + } + return tot +} + +type D1 struct { + d int +} +type D2 struct { + d int +} + +func TestImportPath(t *testing.T) { + tests := []struct { + t Type + path string + }{ + {TypeOf(&base64.Encoding{}).Elem(), "encoding/base64"}, + {TypeOf(int(0)), ""}, + {TypeOf(int8(0)), ""}, + {TypeOf(int16(0)), ""}, + {TypeOf(int32(0)), ""}, + {TypeOf(int64(0)), ""}, + {TypeOf(uint(0)), ""}, + {TypeOf(uint8(0)), ""}, + {TypeOf(uint16(0)), ""}, + {TypeOf(uint32(0)), ""}, + {TypeOf(uint64(0)), ""}, + {TypeOf(uintptr(0)), ""}, + {TypeOf(float32(0)), ""}, + {TypeOf(float64(0)), ""}, + {TypeOf(complex64(0)), ""}, + {TypeOf(complex128(0)), ""}, + {TypeOf(byte(0)), ""}, + {TypeOf(rune(0)), ""}, + {TypeOf([]byte(nil)), ""}, + {TypeOf([]rune(nil)), ""}, + {TypeOf(string("")), ""}, + {TypeOf((*any)(nil)).Elem(), ""}, + {TypeOf((*byte)(nil)), ""}, + {TypeOf((*rune)(nil)), ""}, + {TypeOf((*int64)(nil)), ""}, + {TypeOf(map[string]int{}), ""}, + {TypeOf((*error)(nil)).Elem(), ""}, + {TypeOf((*Point)(nil)), ""}, + {TypeOf((*Point)(nil)).Elem(), "internal/reflectlite_test"}, + } + for _, test := range tests { + if path := test.t.PkgPath(); path != test.path { + t.Errorf("%v.PkgPath() = %q, want %q", test.t, path, test.path) + } + } +} + +func noAlloc(t *testing.T, n int, f func(int)) { + if testing.Short() { + t.Skip("skipping malloc count in short mode") + } + if runtime.GOMAXPROCS(0) > 1 { + t.Skip("skipping; GOMAXPROCS>1") + } + i := -1 + allocs := testing.AllocsPerRun(n, func() { + f(i) + i++ + }) + if allocs > 0 { + t.Errorf("%d iterations: got %v mallocs, want 0", n, allocs) + } +} + +func TestAllocations(t *testing.T) { + noAlloc(t, 100, func(j int) { + var i any + var v Value + + // We can uncomment this when compiler escape analysis + // is good enough to see that the integer assigned to i + // does not escape and therefore need not be allocated. + // + // i = 42 + j + // v = ValueOf(i) + // if int(v.Int()) != 42+j { + // panic("wrong int") + // } + + i = func(j int) int { return j } + v = ValueOf(i) + if ToInterface(v).(func(int) int)(j) != j { + panic("wrong result") + } + }) +} + +func TestSetPanic(t *testing.T) { + ok := func(f func()) { f() } + bad := shouldPanic + clear := func(v Value) { v.Set(Zero(v.Type())) } + + type t0 struct { + W int + } + + type t1 struct { + Y int + t0 + } + + type T2 struct { + Z int + namedT0 t0 + } + + type T struct { + X int + t1 + T2 + NamedT1 t1 + NamedT2 T2 + namedT1 t1 + namedT2 T2 + } + + // not addressable + v := ValueOf(T{}) + bad(func() { clear(Field(v, 0)) }) // .X + bad(func() { clear(Field(v, 1)) }) // .t1 + bad(func() { clear(Field(Field(v, 1), 0)) }) // .t1.Y + bad(func() { clear(Field(Field(v, 1), 1)) }) // .t1.t0 + bad(func() { clear(Field(Field(Field(v, 1), 1), 0)) }) // .t1.t0.W + bad(func() { clear(Field(v, 2)) }) // .T2 + bad(func() { clear(Field(Field(v, 2), 0)) }) // .T2.Z + bad(func() { clear(Field(Field(v, 2), 1)) }) // .T2.namedT0 + bad(func() { clear(Field(Field(Field(v, 2), 1), 0)) }) // .T2.namedT0.W + bad(func() { clear(Field(v, 3)) }) // .NamedT1 + bad(func() { clear(Field(Field(v, 3), 0)) }) // .NamedT1.Y + bad(func() { clear(Field(Field(v, 3), 1)) }) // .NamedT1.t0 + bad(func() { clear(Field(Field(Field(v, 3), 1), 0)) }) // .NamedT1.t0.W + bad(func() { clear(Field(v, 4)) }) // .NamedT2 + bad(func() { clear(Field(Field(v, 4), 0)) }) // .NamedT2.Z + bad(func() { clear(Field(Field(v, 4), 1)) }) // .NamedT2.namedT0 + bad(func() { clear(Field(Field(Field(v, 4), 1), 0)) }) // .NamedT2.namedT0.W + bad(func() { clear(Field(v, 5)) }) // .namedT1 + bad(func() { clear(Field(Field(v, 5), 0)) }) // .namedT1.Y + bad(func() { clear(Field(Field(v, 5), 1)) }) // .namedT1.t0 + bad(func() { clear(Field(Field(Field(v, 5), 1), 0)) }) // .namedT1.t0.W + bad(func() { clear(Field(v, 6)) }) // .namedT2 + bad(func() { clear(Field(Field(v, 6), 0)) }) // .namedT2.Z + bad(func() { clear(Field(Field(v, 6), 1)) }) // .namedT2.namedT0 + bad(func() { clear(Field(Field(Field(v, 6), 1), 0)) }) // .namedT2.namedT0.W + + // addressable + v = ValueOf(&T{}).Elem() + ok(func() { clear(Field(v, 0)) }) // .X + bad(func() { clear(Field(v, 1)) }) // .t1 + ok(func() { clear(Field(Field(v, 1), 0)) }) // .t1.Y + bad(func() { clear(Field(Field(v, 1), 1)) }) // .t1.t0 + ok(func() { clear(Field(Field(Field(v, 1), 1), 0)) }) // .t1.t0.W + ok(func() { clear(Field(v, 2)) }) // .T2 + ok(func() { clear(Field(Field(v, 2), 0)) }) // .T2.Z + bad(func() { clear(Field(Field(v, 2), 1)) }) // .T2.namedT0 + bad(func() { clear(Field(Field(Field(v, 2), 1), 0)) }) // .T2.namedT0.W + ok(func() { clear(Field(v, 3)) }) // .NamedT1 + ok(func() { clear(Field(Field(v, 3), 0)) }) // .NamedT1.Y + bad(func() { clear(Field(Field(v, 3), 1)) }) // .NamedT1.t0 + ok(func() { clear(Field(Field(Field(v, 3), 1), 0)) }) // .NamedT1.t0.W + ok(func() { clear(Field(v, 4)) }) // .NamedT2 + ok(func() { clear(Field(Field(v, 4), 0)) }) // .NamedT2.Z + bad(func() { clear(Field(Field(v, 4), 1)) }) // .NamedT2.namedT0 + bad(func() { clear(Field(Field(Field(v, 4), 1), 0)) }) // .NamedT2.namedT0.W + bad(func() { clear(Field(v, 5)) }) // .namedT1 + bad(func() { clear(Field(Field(v, 5), 0)) }) // .namedT1.Y + bad(func() { clear(Field(Field(v, 5), 1)) }) // .namedT1.t0 + bad(func() { clear(Field(Field(Field(v, 5), 1), 0)) }) // .namedT1.t0.W + bad(func() { clear(Field(v, 6)) }) // .namedT2 + bad(func() { clear(Field(Field(v, 6), 0)) }) // .namedT2.Z + bad(func() { clear(Field(Field(v, 6), 1)) }) // .namedT2.namedT0 + bad(func() { clear(Field(Field(Field(v, 6), 1), 0)) }) // .namedT2.namedT0.W +} + +func shouldPanic(f func()) { + defer func() { + if recover() == nil { + panic("did not panic") + } + }() + f() +} + +type S struct { + i1 int64 + i2 int64 +} + +func TestBigZero(t *testing.T) { + const size = 1 << 10 + var v [size]byte + z := ToInterface(Zero(ValueOf(v).Type())).([size]byte) + for i := 0; i < size; i++ { + if z[i] != 0 { + t.Fatalf("Zero object not all zero, index %d", i) + } + } +} + +func TestInvalid(t *testing.T) { + // Used to have inconsistency between IsValid() and Kind() != Invalid. + type T struct{ v any } + + v := Field(ValueOf(T{}), 0) + if v.IsValid() != true || v.Kind() != Interface { + t.Errorf("field: IsValid=%v, Kind=%v, want true, Interface", v.IsValid(), v.Kind()) + } + v = v.Elem() + if v.IsValid() != false || v.Kind() != abi.Invalid { + t.Errorf("field elem: IsValid=%v, Kind=%v, want false, Invalid", v.IsValid(), v.Kind()) + } +} + +type TheNameOfThisTypeIsExactly255BytesLongSoWhenTheCompilerPrependsTheReflectTestPackageNameAndExtraStarTheLinkerRuntimeAndReflectPackagesWillHaveToCorrectlyDecodeTheSecondLengthByte0123456789_0123456789_0123456789_0123456789_0123456789_012345678 int + +type nameTest struct { + v any + want string +} + +type A struct{} +type B[T any] struct{} + +var nameTests = []nameTest{ + {(*int32)(nil), "int32"}, + {(*D1)(nil), "D1"}, + {(*[]D1)(nil), ""}, + {(*chan D1)(nil), ""}, + {(*func() D1)(nil), ""}, + {(*<-chan D1)(nil), ""}, + {(*chan<- D1)(nil), ""}, + {(*any)(nil), ""}, + {(*interface { + F() + })(nil), ""}, + {(*TheNameOfThisTypeIsExactly255BytesLongSoWhenTheCompilerPrependsTheReflectTestPackageNameAndExtraStarTheLinkerRuntimeAndReflectPackagesWillHaveToCorrectlyDecodeTheSecondLengthByte0123456789_0123456789_0123456789_0123456789_0123456789_012345678)(nil), "TheNameOfThisTypeIsExactly255BytesLongSoWhenTheCompilerPrependsTheReflectTestPackageNameAndExtraStarTheLinkerRuntimeAndReflectPackagesWillHaveToCorrectlyDecodeTheSecondLengthByte0123456789_0123456789_0123456789_0123456789_0123456789_012345678"}, + {(*B[A])(nil), "B[internal/reflectlite_test.A]"}, + {(*B[B[A]])(nil), "B[internal/reflectlite_test.B[internal/reflectlite_test.A]]"}, +} + +func TestNames(t *testing.T) { + for _, test := range nameTests { + typ := TypeOf(test.v).Elem() + if got := typ.Name(); got != test.want { + t.Errorf("%v Name()=%q, want %q", typ, got, test.want) + } + } +} + +// TestUnaddressableField tests that the reflect package will not allow +// a type from another package to be used as a named type with an +// unexported field. +// +// This ensures that unexported fields cannot be modified by other packages. +func TestUnaddressableField(t *testing.T) { + var b Buffer // type defined in reflect, a different package + var localBuffer struct { + buf []byte + } + lv := ValueOf(&localBuffer).Elem() + rv := ValueOf(b) + shouldPanic(func() { + lv.Set(rv) + }) +} + +type Tint int + +type Tint2 = Tint + +type Talias1 struct { + byte + uint8 + int + int32 + rune +} + +type Talias2 struct { + Tint + Tint2 +} + +func TestAliasNames(t *testing.T) { + t1 := Talias1{byte: 1, uint8: 2, int: 3, int32: 4, rune: 5} + out := fmt.Sprintf("%#v", t1) + want := "reflectlite_test.Talias1{byte:0x1, uint8:0x2, int:3, int32:4, rune:5}" + if out != want { + t.Errorf("Talias1 print:\nhave: %s\nwant: %s", out, want) + } + + t2 := Talias2{Tint: 1, Tint2: 2} + out = fmt.Sprintf("%#v", t2) + want = "reflectlite_test.Talias2{Tint:1, Tint2:2}" + if out != want { + t.Errorf("Talias2 print:\nhave: %s\nwant: %s", out, want) + } +} diff --git a/src/internal/reflectlite/asm.s b/src/internal/reflectlite/asm.s new file mode 100644 index 0000000..a7b69b6 --- /dev/null +++ b/src/internal/reflectlite/asm.s @@ -0,0 +1,5 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Trigger build without complete flag.
\ No newline at end of file diff --git a/src/internal/reflectlite/export_test.go b/src/internal/reflectlite/export_test.go new file mode 100644 index 0000000..88be6e2 --- /dev/null +++ b/src/internal/reflectlite/export_test.go @@ -0,0 +1,117 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite + +import ( + "unsafe" +) + +// Field returns the i'th field of the struct v. +// It panics if v's Kind is not Struct or i is out of range. +func Field(v Value, i int) Value { + if v.kind() != Struct { + panic(&ValueError{"reflect.Value.Field", v.kind()}) + } + tt := (*structType)(unsafe.Pointer(v.typ)) + if uint(i) >= uint(len(tt.Fields)) { + panic("reflect: Field index out of range") + } + field := &tt.Fields[i] + typ := field.Typ + + // Inherit permission bits from v, but clear flagEmbedRO. + fl := v.flag&(flagStickyRO|flagIndir|flagAddr) | flag(typ.Kind()) + // Using an unexported field forces flagRO. + if !field.Name.IsExported() { + if field.Embedded() { + fl |= flagEmbedRO + } else { + fl |= flagStickyRO + } + } + // Either flagIndir is set and v.ptr points at struct, + // or flagIndir is not set and v.ptr is the actual struct data. + // In the former case, we want v.ptr + offset. + // In the latter case, we must have field.offset = 0, + // so v.ptr + field.offset is still the correct address. + ptr := add(v.ptr, field.Offset, "same as non-reflect &v.field") + return Value{typ, ptr, fl} +} + +func TField(typ Type, i int) Type { + t := typ.(rtype) + if t.Kind() != Struct { + panic("reflect: Field of non-struct type") + } + tt := (*structType)(unsafe.Pointer(t.Type)) + + return StructFieldType(tt, i) +} + +// Field returns the i'th struct field. +func StructFieldType(t *structType, i int) Type { + if i < 0 || i >= len(t.Fields) { + panic("reflect: Field index out of bounds") + } + p := &t.Fields[i] + return toType(p.Typ) +} + +// Zero returns a Value representing the zero value for the specified type. +// The result is different from the zero value of the Value struct, +// which represents no value at all. +// For example, Zero(TypeOf(42)) returns a Value with Kind Int and value 0. +// The returned value is neither addressable nor settable. +func Zero(typ Type) Value { + if typ == nil { + panic("reflect: Zero(nil)") + } + t := typ.common() + fl := flag(t.Kind()) + if ifaceIndir(t) { + return Value{t, unsafe_New(t), fl | flagIndir} + } + return Value{t, nil, fl} +} + +// ToInterface returns v's current value as an interface{}. +// It is equivalent to: +// +// var i interface{} = (v's underlying value) +// +// It panics if the Value was obtained by accessing +// unexported struct fields. +func ToInterface(v Value) (i any) { + return valueInterface(v) +} + +type EmbedWithUnexpMeth struct{} + +func (EmbedWithUnexpMeth) f() {} + +type pinUnexpMeth interface { + f() +} + +var pinUnexpMethI = pinUnexpMeth(EmbedWithUnexpMeth{}) + +func FirstMethodNameBytes(t Type) *byte { + _ = pinUnexpMethI + + ut := t.uncommon() + if ut == nil { + panic("type has no methods") + } + m := ut.Methods()[0] + mname := t.(rtype).nameOff(m.Name) + if *mname.DataChecked(0, "name flag field")&(1<<2) == 0 { + panic("method name does not have pkgPath *string") + } + return mname.Bytes +} + +type Buffer struct { + buf []byte +} diff --git a/src/internal/reflectlite/reflect_mirror_test.go b/src/internal/reflectlite/reflect_mirror_test.go new file mode 100644 index 0000000..0fd004e --- /dev/null +++ b/src/internal/reflectlite/reflect_mirror_test.go @@ -0,0 +1,133 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite_test + +import ( + "fmt" + "go/ast" + "go/parser" + "go/token" + "io/fs" + "os" + "path/filepath" + "runtime" + "strings" + "sync" + "testing" +) + +var typeNames = []string{ + "uncommonType", + "arrayType", + "chanType", + "funcType", + "interfaceType", + "mapType", + "ptrType", + "sliceType", + "structType", +} + +type visitor struct { + m map[string]map[string]bool +} + +func newVisitor() visitor { + v := visitor{} + v.m = make(map[string]map[string]bool) + + return v +} +func (v visitor) filter(name string) bool { + for _, typeName := range typeNames { + if typeName == name { + return true + } + } + return false +} + +func (v visitor) Visit(n ast.Node) ast.Visitor { + switch x := n.(type) { + case *ast.TypeSpec: + if v.filter(x.Name.String()) { + if st, ok := x.Type.(*ast.StructType); ok { + v.m[x.Name.String()] = make(map[string]bool) + for _, field := range st.Fields.List { + k := fmt.Sprintf("%s", field.Type) + if len(field.Names) > 0 { + k = field.Names[0].Name + } + v.m[x.Name.String()][k] = true + } + } + } + } + return v +} + +func loadTypes(path, pkgName string, v visitor) { + fset := token.NewFileSet() + + filter := func(fi fs.FileInfo) bool { + return strings.HasSuffix(fi.Name(), ".go") + } + pkgs, err := parser.ParseDir(fset, path, filter, 0) + if err != nil { + panic(err) + } + + pkg := pkgs[pkgName] + + for _, f := range pkg.Files { + ast.Walk(v, f) + } +} + +func TestMirrorWithReflect(t *testing.T) { + // TODO when the dust clears, figure out what this should actually test. + t.Skipf("reflect and reflectlite are out of sync for now") + reflectDir := filepath.Join(runtime.GOROOT(), "src", "reflect") + if _, err := os.Stat(reflectDir); os.IsNotExist(err) { + // On some mobile builders, the test binary executes on a machine without a + // complete GOROOT source tree. + t.Skipf("GOROOT source not present") + } + + var wg sync.WaitGroup + rl, r := newVisitor(), newVisitor() + + for _, tc := range []struct { + path, pkg string + v visitor + }{ + {".", "reflectlite", rl}, + {reflectDir, "reflect", r}, + } { + tc := tc + wg.Add(1) + go func() { + defer wg.Done() + loadTypes(tc.path, tc.pkg, tc.v) + }() + } + wg.Wait() + + if len(rl.m) != len(r.m) { + t.Fatalf("number of types mismatch, reflect: %d, reflectlite: %d (%+v, %+v)", len(r.m), len(rl.m), r.m, rl.m) + } + + for typName := range r.m { + if len(r.m[typName]) != len(rl.m[typName]) { + t.Errorf("type %s number of fields mismatch, reflect: %d, reflectlite: %d", typName, len(r.m[typName]), len(rl.m[typName])) + continue + } + for field := range r.m[typName] { + if _, ok := rl.m[typName][field]; !ok { + t.Errorf(`Field mismatch, reflect have "%s", relectlite does not.`, field) + } + } + } +} diff --git a/src/internal/reflectlite/set_test.go b/src/internal/reflectlite/set_test.go new file mode 100644 index 0000000..ca7ea9b --- /dev/null +++ b/src/internal/reflectlite/set_test.go @@ -0,0 +1,101 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite_test + +import ( + "bytes" + "go/ast" + "go/token" + . "internal/reflectlite" + "io" + "testing" +) + +func TestImplicitSetConversion(t *testing.T) { + // Assume TestImplicitMapConversion covered the basics. + // Just make sure conversions are being applied at all. + var r io.Reader + b := new(bytes.Buffer) + rv := ValueOf(&r).Elem() + rv.Set(ValueOf(b)) + if r != b { + t.Errorf("after Set: r=%T(%v)", r, r) + } +} + +var implementsTests = []struct { + x any + t any + b bool +}{ + {new(*bytes.Buffer), new(io.Reader), true}, + {new(bytes.Buffer), new(io.Reader), false}, + {new(*bytes.Buffer), new(io.ReaderAt), false}, + {new(*ast.Ident), new(ast.Expr), true}, + {new(*notAnExpr), new(ast.Expr), false}, + {new(*ast.Ident), new(notASTExpr), false}, + {new(notASTExpr), new(ast.Expr), false}, + {new(ast.Expr), new(notASTExpr), false}, + {new(*notAnExpr), new(notASTExpr), true}, + {new(mapError), new(error), true}, + {new(*mapError), new(error), true}, +} + +type notAnExpr struct{} + +func (notAnExpr) Pos() token.Pos { return token.NoPos } +func (notAnExpr) End() token.Pos { return token.NoPos } +func (notAnExpr) exprNode() {} + +type notASTExpr interface { + Pos() token.Pos + End() token.Pos + exprNode() +} + +type mapError map[string]string + +func (mapError) Error() string { return "mapError" } + +var _ error = mapError{} +var _ error = new(mapError) + +func TestImplements(t *testing.T) { + for _, tt := range implementsTests { + xv := TypeOf(tt.x).Elem() + xt := TypeOf(tt.t).Elem() + if b := xv.Implements(xt); b != tt.b { + t.Errorf("(%s).Implements(%s) = %v, want %v", TypeString(xv), TypeString(xt), b, tt.b) + } + } +} + +var assignableTests = []struct { + x any + t any + b bool +}{ + {new(chan int), new(<-chan int), true}, + {new(<-chan int), new(chan int), false}, + {new(*int), new(IntPtr), true}, + {new(IntPtr), new(*int), true}, + {new(IntPtr), new(IntPtr1), false}, + {new(Ch), new(<-chan any), true}, + // test runs implementsTests too +} + +type IntPtr *int +type IntPtr1 *int +type Ch <-chan any + +func TestAssignableTo(t *testing.T) { + for i, tt := range append(assignableTests, implementsTests...) { + xv := TypeOf(tt.x).Elem() + xt := TypeOf(tt.t).Elem() + if b := xv.AssignableTo(xt); b != tt.b { + t.Errorf("%d:AssignableTo: got %v, want %v", i, b, tt.b) + } + } +} diff --git a/src/internal/reflectlite/swapper.go b/src/internal/reflectlite/swapper.go new file mode 100644 index 0000000..ac17d9b --- /dev/null +++ b/src/internal/reflectlite/swapper.go @@ -0,0 +1,78 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite + +import ( + "internal/goarch" + "internal/unsafeheader" + "unsafe" +) + +// Swapper returns a function that swaps the elements in the provided +// slice. +// +// Swapper panics if the provided interface is not a slice. +func Swapper(slice any) func(i, j int) { + v := ValueOf(slice) + if v.Kind() != Slice { + panic(&ValueError{Method: "Swapper", Kind: v.Kind()}) + } + // Fast path for slices of size 0 and 1. Nothing to swap. + switch v.Len() { + case 0: + return func(i, j int) { panic("reflect: slice index out of range") } + case 1: + return func(i, j int) { + if i != 0 || j != 0 { + panic("reflect: slice index out of range") + } + } + } + + typ := v.Type().Elem().common() + size := typ.Size() + hasPtr := typ.PtrBytes != 0 + + // Some common & small cases, without using memmove: + if hasPtr { + if size == goarch.PtrSize { + ps := *(*[]unsafe.Pointer)(v.ptr) + return func(i, j int) { ps[i], ps[j] = ps[j], ps[i] } + } + if typ.Kind() == String { + ss := *(*[]string)(v.ptr) + return func(i, j int) { ss[i], ss[j] = ss[j], ss[i] } + } + } else { + switch size { + case 8: + is := *(*[]int64)(v.ptr) + return func(i, j int) { is[i], is[j] = is[j], is[i] } + case 4: + is := *(*[]int32)(v.ptr) + return func(i, j int) { is[i], is[j] = is[j], is[i] } + case 2: + is := *(*[]int16)(v.ptr) + return func(i, j int) { is[i], is[j] = is[j], is[i] } + case 1: + is := *(*[]int8)(v.ptr) + return func(i, j int) { is[i], is[j] = is[j], is[i] } + } + } + + s := (*unsafeheader.Slice)(v.ptr) + tmp := unsafe_New(typ) // swap scratch space + + return func(i, j int) { + if uint(i) >= uint(s.Len) || uint(j) >= uint(s.Len) { + panic("reflect: slice index out of range") + } + val1 := arrayAt(s.Data, i, size, "i < s.Len") + val2 := arrayAt(s.Data, j, size, "j < s.Len") + typedmemmove(typ, tmp, val1) + typedmemmove(typ, val1, val2) + typedmemmove(typ, val2, tmp) + } +} diff --git a/src/internal/reflectlite/tostring_test.go b/src/internal/reflectlite/tostring_test.go new file mode 100644 index 0000000..966b0bd --- /dev/null +++ b/src/internal/reflectlite/tostring_test.go @@ -0,0 +1,98 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Formatting of reflection types and values for debugging. +// Not defined as methods so they do not need to be linked into most binaries; +// the functions are not used by the library itself, only in tests. + +package reflectlite_test + +import ( + . "internal/reflectlite" + "reflect" + "strconv" +) + +// valueToString returns a textual representation of the reflection value val. +// For debugging only. +func valueToString(v Value) string { + return valueToStringImpl(reflect.ValueOf(ToInterface(v))) +} + +func valueToStringImpl(val reflect.Value) string { + var str string + if !val.IsValid() { + return "<zero Value>" + } + typ := val.Type() + switch val.Kind() { + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64: + return strconv.FormatInt(val.Int(), 10) + case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64, reflect.Uintptr: + return strconv.FormatUint(val.Uint(), 10) + case reflect.Float32, reflect.Float64: + return strconv.FormatFloat(val.Float(), 'g', -1, 64) + case reflect.Complex64, reflect.Complex128: + c := val.Complex() + return strconv.FormatFloat(real(c), 'g', -1, 64) + "+" + strconv.FormatFloat(imag(c), 'g', -1, 64) + "i" + case reflect.String: + return val.String() + case reflect.Bool: + if val.Bool() { + return "true" + } else { + return "false" + } + case reflect.Pointer: + v := val + str = typ.String() + "(" + if v.IsNil() { + str += "0" + } else { + str += "&" + valueToStringImpl(v.Elem()) + } + str += ")" + return str + case reflect.Array, reflect.Slice: + v := val + str += typ.String() + str += "{" + for i := 0; i < v.Len(); i++ { + if i > 0 { + str += ", " + } + str += valueToStringImpl(v.Index(i)) + } + str += "}" + return str + case reflect.Map: + str += typ.String() + str += "{" + str += "<can't iterate on maps>" + str += "}" + return str + case reflect.Chan: + str = typ.String() + return str + case reflect.Struct: + t := typ + v := val + str += t.String() + str += "{" + for i, n := 0, v.NumField(); i < n; i++ { + if i > 0 { + str += ", " + } + str += valueToStringImpl(v.Field(i)) + } + str += "}" + return str + case reflect.Interface: + return typ.String() + "(" + valueToStringImpl(val.Elem()) + ")" + case reflect.Func: + return typ.String() + "(arg)" + default: + panic("valueToString: can't print type " + typ.String()) + } +} diff --git a/src/internal/reflectlite/type.go b/src/internal/reflectlite/type.go new file mode 100644 index 0000000..f13ce8f --- /dev/null +++ b/src/internal/reflectlite/type.go @@ -0,0 +1,659 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package reflectlite implements lightweight version of reflect, not using +// any package except for "runtime", "unsafe", and "internal/abi" +package reflectlite + +import ( + "internal/abi" + "unsafe" +) + +// Type is the representation of a Go type. +// +// Not all methods apply to all kinds of types. Restrictions, +// if any, are noted in the documentation for each method. +// Use the Kind method to find out the kind of type before +// calling kind-specific methods. Calling a method +// inappropriate to the kind of type causes a run-time panic. +// +// Type values are comparable, such as with the == operator, +// so they can be used as map keys. +// Two Type values are equal if they represent identical types. +type Type interface { + // Methods applicable to all types. + + // Name returns the type's name within its package for a defined type. + // For other (non-defined) types it returns the empty string. + Name() string + + // PkgPath returns a defined type's package path, that is, the import path + // that uniquely identifies the package, such as "encoding/base64". + // If the type was predeclared (string, error) or not defined (*T, struct{}, + // []int, or A where A is an alias for a non-defined type), the package path + // will be the empty string. + PkgPath() string + + // Size returns the number of bytes needed to store + // a value of the given type; it is analogous to unsafe.Sizeof. + Size() uintptr + + // Kind returns the specific kind of this type. + Kind() Kind + + // Implements reports whether the type implements the interface type u. + Implements(u Type) bool + + // AssignableTo reports whether a value of the type is assignable to type u. + AssignableTo(u Type) bool + + // Comparable reports whether values of this type are comparable. + Comparable() bool + + // String returns a string representation of the type. + // The string representation may use shortened package names + // (e.g., base64 instead of "encoding/base64") and is not + // guaranteed to be unique among types. To test for type identity, + // compare the Types directly. + String() string + + // Elem returns a type's element type. + // It panics if the type's Kind is not Ptr. + Elem() Type + + common() *abi.Type + uncommon() *uncommonType +} + +/* + * These data structures are known to the compiler (../../cmd/internal/reflectdata/reflect.go). + * A few are known to ../runtime/type.go to convey to debuggers. + * They are also known to ../runtime/type.go. + */ + +// A Kind represents the specific kind of type that a Type represents. +// The zero Kind is not a valid kind. +type Kind = abi.Kind + +const Ptr = abi.Pointer + +const ( + // Import-and-export these constants as necessary + Interface = abi.Interface + Slice = abi.Slice + String = abi.String + Struct = abi.Struct +) + +type nameOff = abi.NameOff +type typeOff = abi.TypeOff +type textOff = abi.TextOff + +type rtype struct { + *abi.Type +} + +// uncommonType is present only for defined types or types with methods +// (if T is a defined type, the uncommonTypes for T and *T have methods). +// Using a pointer to this struct reduces the overall size required +// to describe a non-defined type with no methods. +type uncommonType = abi.UncommonType + +// arrayType represents a fixed array type. +type arrayType = abi.ArrayType + +// chanType represents a channel type. +type chanType = abi.ChanType + +type funcType = abi.FuncType + +type interfaceType = abi.InterfaceType + +// mapType represents a map type. +type mapType struct { + rtype + Key *abi.Type // map key type + Elem *abi.Type // map element (value) type + Bucket *abi.Type // internal bucket structure + // function for hashing keys (ptr to key, seed) -> hash + Hasher func(unsafe.Pointer, uintptr) uintptr + KeySize uint8 // size of key slot + ValueSize uint8 // size of value slot + BucketSize uint16 // size of bucket + Flags uint32 +} + +// ptrType represents a pointer type. +type ptrType = abi.PtrType + +// sliceType represents a slice type. +type sliceType = abi.SliceType + +// structType represents a struct type. +type structType = abi.StructType + +// name is an encoded type name with optional extra data. +// +// The first byte is a bit field containing: +// +// 1<<0 the name is exported +// 1<<1 tag data follows the name +// 1<<2 pkgPath nameOff follows the name and tag +// +// The next two bytes are the data length: +// +// l := uint16(data[1])<<8 | uint16(data[2]) +// +// Bytes [3:3+l] are the string data. +// +// If tag data follows then bytes 3+l and 3+l+1 are the tag length, +// with the data following. +// +// If the import path follows, then 4 bytes at the end of +// the data form a nameOff. The import path is only set for concrete +// methods that are defined in a different package than their type. +// +// If a name starts with "*", then the exported bit represents +// whether the pointed to type is exported. +type name struct { + bytes *byte +} + +func (n name) data(off int, whySafe string) *byte { + return (*byte)(add(unsafe.Pointer(n.bytes), uintptr(off), whySafe)) +} + +func (n name) isExported() bool { + return (*n.bytes)&(1<<0) != 0 +} + +func (n name) hasTag() bool { + return (*n.bytes)&(1<<1) != 0 +} + +func (n name) embedded() bool { + return (*n.bytes)&(1<<3) != 0 +} + +// readVarint parses a varint as encoded by encoding/binary. +// It returns the number of encoded bytes and the encoded value. +func (n name) readVarint(off int) (int, int) { + v := 0 + for i := 0; ; i++ { + x := *n.data(off+i, "read varint") + v += int(x&0x7f) << (7 * i) + if x&0x80 == 0 { + return i + 1, v + } + } +} + +func (n name) name() string { + if n.bytes == nil { + return "" + } + i, l := n.readVarint(1) + return unsafe.String(n.data(1+i, "non-empty string"), l) +} + +func (n name) tag() string { + if !n.hasTag() { + return "" + } + i, l := n.readVarint(1) + i2, l2 := n.readVarint(1 + i + l) + return unsafe.String(n.data(1+i+l+i2, "non-empty string"), l2) +} + +func pkgPath(n abi.Name) string { + if n.Bytes == nil || *n.DataChecked(0, "name flag field")&(1<<2) == 0 { + return "" + } + i, l := n.ReadVarint(1) + off := 1 + i + l + if n.HasTag() { + i2, l2 := n.ReadVarint(off) + off += i2 + l2 + } + var nameOff int32 + // Note that this field may not be aligned in memory, + // so we cannot use a direct int32 assignment here. + copy((*[4]byte)(unsafe.Pointer(&nameOff))[:], (*[4]byte)(unsafe.Pointer(n.DataChecked(off, "name offset field")))[:]) + pkgPathName := name{(*byte)(resolveTypeOff(unsafe.Pointer(n.Bytes), nameOff))} + return pkgPathName.name() +} + +/* + * The compiler knows the exact layout of all the data structures above. + * The compiler does not know about the data structures and methods below. + */ + +// resolveNameOff resolves a name offset from a base pointer. +// The (*rtype).nameOff method is a convenience wrapper for this function. +// Implemented in the runtime package. +func resolveNameOff(ptrInModule unsafe.Pointer, off int32) unsafe.Pointer + +// resolveTypeOff resolves an *rtype offset from a base type. +// The (*rtype).typeOff method is a convenience wrapper for this function. +// Implemented in the runtime package. +func resolveTypeOff(rtype unsafe.Pointer, off int32) unsafe.Pointer + +func (t rtype) nameOff(off nameOff) abi.Name { + return abi.Name{Bytes: (*byte)(resolveNameOff(unsafe.Pointer(t.Type), int32(off)))} +} + +func (t rtype) typeOff(off typeOff) *abi.Type { + return (*abi.Type)(resolveTypeOff(unsafe.Pointer(t.Type), int32(off))) +} + +func (t rtype) uncommon() *uncommonType { + return t.Uncommon() +} + +func (t rtype) String() string { + s := t.nameOff(t.Str).Name() + if t.TFlag&abi.TFlagExtraStar != 0 { + return s[1:] + } + return s +} + +func (t rtype) common() *abi.Type { return t.Type } + +func (t rtype) exportedMethods() []abi.Method { + ut := t.uncommon() + if ut == nil { + return nil + } + return ut.ExportedMethods() +} + +func (t rtype) NumMethod() int { + tt := t.Type.InterfaceType() + if tt != nil { + return tt.NumMethod() + } + return len(t.exportedMethods()) +} + +func (t rtype) PkgPath() string { + if t.TFlag&abi.TFlagNamed == 0 { + return "" + } + ut := t.uncommon() + if ut == nil { + return "" + } + return t.nameOff(ut.PkgPath).Name() +} + +func (t rtype) Name() string { + if !t.HasName() { + return "" + } + s := t.String() + i := len(s) - 1 + sqBrackets := 0 + for i >= 0 && (s[i] != '.' || sqBrackets != 0) { + switch s[i] { + case ']': + sqBrackets++ + case '[': + sqBrackets-- + } + i-- + } + return s[i+1:] +} + +func toRType(t *abi.Type) rtype { + return rtype{t} +} + +func elem(t *abi.Type) *abi.Type { + et := t.Elem() + if et != nil { + return et + } + panic("reflect: Elem of invalid type " + toRType(t).String()) +} + +func (t rtype) Elem() Type { + return toType(elem(t.common())) +} + +func (t rtype) In(i int) Type { + tt := t.Type.FuncType() + if tt == nil { + panic("reflect: In of non-func type") + } + return toType(tt.InSlice()[i]) +} + +func (t rtype) Key() Type { + tt := t.Type.MapType() + if tt == nil { + panic("reflect: Key of non-map type") + } + return toType(tt.Key) +} + +func (t rtype) Len() int { + tt := t.Type.ArrayType() + if tt == nil { + panic("reflect: Len of non-array type") + } + return int(tt.Len) +} + +func (t rtype) NumField() int { + tt := t.Type.StructType() + if tt == nil { + panic("reflect: NumField of non-struct type") + } + return len(tt.Fields) +} + +func (t rtype) NumIn() int { + tt := t.Type.FuncType() + if tt == nil { + panic("reflect: NumIn of non-func type") + } + return int(tt.InCount) +} + +func (t rtype) NumOut() int { + tt := t.Type.FuncType() + if tt == nil { + panic("reflect: NumOut of non-func type") + } + return tt.NumOut() +} + +func (t rtype) Out(i int) Type { + tt := t.Type.FuncType() + if tt == nil { + panic("reflect: Out of non-func type") + } + return toType(tt.OutSlice()[i]) +} + +// add returns p+x. +// +// The whySafe string is ignored, so that the function still inlines +// as efficiently as p+x, but all call sites should use the string to +// record why the addition is safe, which is to say why the addition +// does not cause x to advance to the very end of p's allocation +// and therefore point incorrectly at the next block in memory. +func add(p unsafe.Pointer, x uintptr, whySafe string) unsafe.Pointer { + return unsafe.Pointer(uintptr(p) + x) +} + +// TypeOf returns the reflection Type that represents the dynamic type of i. +// If i is a nil interface value, TypeOf returns nil. +func TypeOf(i any) Type { + eface := *(*emptyInterface)(unsafe.Pointer(&i)) + return toType(eface.typ) +} + +func (t rtype) Implements(u Type) bool { + if u == nil { + panic("reflect: nil type passed to Type.Implements") + } + if u.Kind() != Interface { + panic("reflect: non-interface type passed to Type.Implements") + } + return implements(u.common(), t.common()) +} + +func (t rtype) AssignableTo(u Type) bool { + if u == nil { + panic("reflect: nil type passed to Type.AssignableTo") + } + uu := u.common() + tt := t.common() + return directlyAssignable(uu, tt) || implements(uu, tt) +} + +func (t rtype) Comparable() bool { + return t.Equal != nil +} + +// implements reports whether the type V implements the interface type T. +func implements(T, V *abi.Type) bool { + t := T.InterfaceType() + if t == nil { + return false + } + if len(t.Methods) == 0 { + return true + } + rT := toRType(T) + rV := toRType(V) + + // The same algorithm applies in both cases, but the + // method tables for an interface type and a concrete type + // are different, so the code is duplicated. + // In both cases the algorithm is a linear scan over the two + // lists - T's methods and V's methods - simultaneously. + // Since method tables are stored in a unique sorted order + // (alphabetical, with no duplicate method names), the scan + // through V's methods must hit a match for each of T's + // methods along the way, or else V does not implement T. + // This lets us run the scan in overall linear time instead of + // the quadratic time a naive search would require. + // See also ../runtime/iface.go. + if V.Kind() == Interface { + v := (*interfaceType)(unsafe.Pointer(V)) + i := 0 + for j := 0; j < len(v.Methods); j++ { + tm := &t.Methods[i] + tmName := rT.nameOff(tm.Name) + vm := &v.Methods[j] + vmName := rV.nameOff(vm.Name) + if vmName.Name() == tmName.Name() && rV.typeOff(vm.Typ) == rT.typeOff(tm.Typ) { + if !tmName.IsExported() { + tmPkgPath := pkgPath(tmName) + if tmPkgPath == "" { + tmPkgPath = t.PkgPath.Name() + } + vmPkgPath := pkgPath(vmName) + if vmPkgPath == "" { + vmPkgPath = v.PkgPath.Name() + } + if tmPkgPath != vmPkgPath { + continue + } + } + if i++; i >= len(t.Methods) { + return true + } + } + } + return false + } + + v := V.Uncommon() + if v == nil { + return false + } + i := 0 + vmethods := v.Methods() + for j := 0; j < int(v.Mcount); j++ { + tm := &t.Methods[i] + tmName := rT.nameOff(tm.Name) + vm := vmethods[j] + vmName := rV.nameOff(vm.Name) + if vmName.Name() == tmName.Name() && rV.typeOff(vm.Mtyp) == rT.typeOff(tm.Typ) { + if !tmName.IsExported() { + tmPkgPath := pkgPath(tmName) + if tmPkgPath == "" { + tmPkgPath = t.PkgPath.Name() + } + vmPkgPath := pkgPath(vmName) + if vmPkgPath == "" { + vmPkgPath = rV.nameOff(v.PkgPath).Name() + } + if tmPkgPath != vmPkgPath { + continue + } + } + if i++; i >= len(t.Methods) { + return true + } + } + } + return false +} + +// directlyAssignable reports whether a value x of type V can be directly +// assigned (using memmove) to a value of type T. +// https://golang.org/doc/go_spec.html#Assignability +// Ignoring the interface rules (implemented elsewhere) +// and the ideal constant rules (no ideal constants at run time). +func directlyAssignable(T, V *abi.Type) bool { + // x's type V is identical to T? + if T == V { + return true + } + + // Otherwise at least one of T and V must not be defined + // and they must have the same kind. + if T.HasName() && V.HasName() || T.Kind() != V.Kind() { + return false + } + + // x's type T and V must have identical underlying types. + return haveIdenticalUnderlyingType(T, V, true) +} + +func haveIdenticalType(T, V *abi.Type, cmpTags bool) bool { + if cmpTags { + return T == V + } + + if toRType(T).Name() != toRType(V).Name() || T.Kind() != V.Kind() { + return false + } + + return haveIdenticalUnderlyingType(T, V, false) +} + +func haveIdenticalUnderlyingType(T, V *abi.Type, cmpTags bool) bool { + if T == V { + return true + } + + kind := T.Kind() + if kind != V.Kind() { + return false + } + + // Non-composite types of equal kind have same underlying type + // (the predefined instance of the type). + if abi.Bool <= kind && kind <= abi.Complex128 || kind == abi.String || kind == abi.UnsafePointer { + return true + } + + // Composite types. + switch kind { + case abi.Array: + return T.Len() == V.Len() && haveIdenticalType(T.Elem(), V.Elem(), cmpTags) + + case abi.Chan: + // Special case: + // x is a bidirectional channel value, T is a channel type, + // and x's type V and T have identical element types. + if V.ChanDir() == abi.BothDir && haveIdenticalType(T.Elem(), V.Elem(), cmpTags) { + return true + } + + // Otherwise continue test for identical underlying type. + return V.ChanDir() == T.ChanDir() && haveIdenticalType(T.Elem(), V.Elem(), cmpTags) + + case abi.Func: + t := (*funcType)(unsafe.Pointer(T)) + v := (*funcType)(unsafe.Pointer(V)) + if t.OutCount != v.OutCount || t.InCount != v.InCount { + return false + } + for i := 0; i < t.NumIn(); i++ { + if !haveIdenticalType(t.In(i), v.In(i), cmpTags) { + return false + } + } + for i := 0; i < t.NumOut(); i++ { + if !haveIdenticalType(t.Out(i), v.Out(i), cmpTags) { + return false + } + } + return true + + case Interface: + t := (*interfaceType)(unsafe.Pointer(T)) + v := (*interfaceType)(unsafe.Pointer(V)) + if len(t.Methods) == 0 && len(v.Methods) == 0 { + return true + } + // Might have the same methods but still + // need a run time conversion. + return false + + case abi.Map: + return haveIdenticalType(T.Key(), V.Key(), cmpTags) && haveIdenticalType(T.Elem(), V.Elem(), cmpTags) + + case Ptr, abi.Slice: + return haveIdenticalType(T.Elem(), V.Elem(), cmpTags) + + case abi.Struct: + t := (*structType)(unsafe.Pointer(T)) + v := (*structType)(unsafe.Pointer(V)) + if len(t.Fields) != len(v.Fields) { + return false + } + if t.PkgPath.Name() != v.PkgPath.Name() { + return false + } + for i := range t.Fields { + tf := &t.Fields[i] + vf := &v.Fields[i] + if tf.Name.Name() != vf.Name.Name() { + return false + } + if !haveIdenticalType(tf.Typ, vf.Typ, cmpTags) { + return false + } + if cmpTags && tf.Name.Tag() != vf.Name.Tag() { + return false + } + if tf.Offset != vf.Offset { + return false + } + if tf.Embedded() != vf.Embedded() { + return false + } + } + return true + } + + return false +} + +// toType converts from a *rtype to a Type that can be returned +// to the client of package reflect. In gc, the only concern is that +// a nil *rtype must be replaced by a nil Type, but in gccgo this +// function takes care of ensuring that multiple *rtype for the same +// type are coalesced into a single Type. +func toType(t *abi.Type) Type { + if t == nil { + return nil + } + return toRType(t) +} + +// ifaceIndir reports whether t is stored indirectly in an interface value. +func ifaceIndir(t *abi.Type) bool { + return t.Kind_&abi.KindDirectIface == 0 +} diff --git a/src/internal/reflectlite/value.go b/src/internal/reflectlite/value.go new file mode 100644 index 0000000..eb79894 --- /dev/null +++ b/src/internal/reflectlite/value.go @@ -0,0 +1,478 @@ +// Copyright 2009 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package reflectlite + +import ( + "internal/abi" + "internal/goarch" + "internal/unsafeheader" + "runtime" + "unsafe" +) + +// Value is the reflection interface to a Go value. +// +// Not all methods apply to all kinds of values. Restrictions, +// if any, are noted in the documentation for each method. +// Use the Kind method to find out the kind of value before +// calling kind-specific methods. Calling a method +// inappropriate to the kind of type causes a run time panic. +// +// The zero Value represents no value. +// Its IsValid method returns false, its Kind method returns Invalid, +// its String method returns "<invalid Value>", and all other methods panic. +// Most functions and methods never return an invalid value. +// If one does, its documentation states the conditions explicitly. +// +// A Value can be used concurrently by multiple goroutines provided that +// the underlying Go value can be used concurrently for the equivalent +// direct operations. +// +// To compare two Values, compare the results of the Interface method. +// Using == on two Values does not compare the underlying values +// they represent. +type Value struct { + // typ holds the type of the value represented by a Value. + typ *abi.Type + + // Pointer-valued data or, if flagIndir is set, pointer to data. + // Valid when either flagIndir is set or typ.pointers() is true. + ptr unsafe.Pointer + + // flag holds metadata about the value. + // The lowest bits are flag bits: + // - flagStickyRO: obtained via unexported not embedded field, so read-only + // - flagEmbedRO: obtained via unexported embedded field, so read-only + // - flagIndir: val holds a pointer to the data + // - flagAddr: v.CanAddr is true (implies flagIndir) + // Value cannot represent method values. + // The next five bits give the Kind of the value. + // This repeats typ.Kind() except for method values. + // The remaining 23+ bits give a method number for method values. + // If flag.kind() != Func, code can assume that flagMethod is unset. + // If ifaceIndir(typ), code can assume that flagIndir is set. + flag + + // A method value represents a curried method invocation + // like r.Read for some receiver r. The typ+val+flag bits describe + // the receiver r, but the flag's Kind bits say Func (methods are + // functions), and the top bits of the flag give the method number + // in r's type's method table. +} + +type flag uintptr + +const ( + flagKindWidth = 5 // there are 27 kinds + flagKindMask flag = 1<<flagKindWidth - 1 + flagStickyRO flag = 1 << 5 + flagEmbedRO flag = 1 << 6 + flagIndir flag = 1 << 7 + flagAddr flag = 1 << 8 + flagMethod flag = 1 << 9 + flagMethodShift = 10 + flagRO flag = flagStickyRO | flagEmbedRO +) + +func (f flag) kind() Kind { + return Kind(f & flagKindMask) +} + +func (f flag) ro() flag { + if f&flagRO != 0 { + return flagStickyRO + } + return 0 +} + +// pointer returns the underlying pointer represented by v. +// v.Kind() must be Pointer, Map, Chan, Func, or UnsafePointer +func (v Value) pointer() unsafe.Pointer { + if v.typ.Size() != goarch.PtrSize || !v.typ.Pointers() { + panic("can't call pointer on a non-pointer Value") + } + if v.flag&flagIndir != 0 { + return *(*unsafe.Pointer)(v.ptr) + } + return v.ptr +} + +// packEface converts v to the empty interface. +func packEface(v Value) any { + t := v.typ + var i any + e := (*emptyInterface)(unsafe.Pointer(&i)) + // First, fill in the data portion of the interface. + switch { + case ifaceIndir(t): + if v.flag&flagIndir == 0 { + panic("bad indir") + } + // Value is indirect, and so is the interface we're making. + ptr := v.ptr + if v.flag&flagAddr != 0 { + // TODO: pass safe boolean from valueInterface so + // we don't need to copy if safe==true? + c := unsafe_New(t) + typedmemmove(t, c, ptr) + ptr = c + } + e.word = ptr + case v.flag&flagIndir != 0: + // Value is indirect, but interface is direct. We need + // to load the data at v.ptr into the interface data word. + e.word = *(*unsafe.Pointer)(v.ptr) + default: + // Value is direct, and so is the interface. + e.word = v.ptr + } + // Now, fill in the type portion. We're very careful here not + // to have any operation between the e.word and e.typ assignments + // that would let the garbage collector observe the partially-built + // interface value. + e.typ = t + return i +} + +// unpackEface converts the empty interface i to a Value. +func unpackEface(i any) Value { + e := (*emptyInterface)(unsafe.Pointer(&i)) + // NOTE: don't read e.word until we know whether it is really a pointer or not. + t := e.typ + if t == nil { + return Value{} + } + f := flag(t.Kind()) + if ifaceIndir(t) { + f |= flagIndir + } + return Value{t, e.word, f} +} + +// A ValueError occurs when a Value method is invoked on +// a Value that does not support it. Such cases are documented +// in the description of each method. +type ValueError struct { + Method string + Kind Kind +} + +func (e *ValueError) Error() string { + if e.Kind == 0 { + return "reflect: call of " + e.Method + " on zero Value" + } + return "reflect: call of " + e.Method + " on " + e.Kind.String() + " Value" +} + +// methodName returns the name of the calling method, +// assumed to be two stack frames above. +func methodName() string { + pc, _, _, _ := runtime.Caller(2) + f := runtime.FuncForPC(pc) + if f == nil { + return "unknown method" + } + return f.Name() +} + +// emptyInterface is the header for an interface{} value. +type emptyInterface struct { + typ *abi.Type + word unsafe.Pointer +} + +// mustBeExported panics if f records that the value was obtained using +// an unexported field. +func (f flag) mustBeExported() { + if f == 0 { + panic(&ValueError{methodName(), 0}) + } + if f&flagRO != 0 { + panic("reflect: " + methodName() + " using value obtained using unexported field") + } +} + +// mustBeAssignable panics if f records that the value is not assignable, +// which is to say that either it was obtained using an unexported field +// or it is not addressable. +func (f flag) mustBeAssignable() { + if f == 0 { + panic(&ValueError{methodName(), abi.Invalid}) + } + // Assignable if addressable and not read-only. + if f&flagRO != 0 { + panic("reflect: " + methodName() + " using value obtained using unexported field") + } + if f&flagAddr == 0 { + panic("reflect: " + methodName() + " using unaddressable value") + } +} + +// CanSet reports whether the value of v can be changed. +// A Value can be changed only if it is addressable and was not +// obtained by the use of unexported struct fields. +// If CanSet returns false, calling Set or any type-specific +// setter (e.g., SetBool, SetInt) will panic. +func (v Value) CanSet() bool { + return v.flag&(flagAddr|flagRO) == flagAddr +} + +// Elem returns the value that the interface v contains +// or that the pointer v points to. +// It panics if v's Kind is not Interface or Pointer. +// It returns the zero Value if v is nil. +func (v Value) Elem() Value { + k := v.kind() + switch k { + case abi.Interface: + var eface any + if v.typ.NumMethod() == 0 { + eface = *(*any)(v.ptr) + } else { + eface = (any)(*(*interface { + M() + })(v.ptr)) + } + x := unpackEface(eface) + if x.flag != 0 { + x.flag |= v.flag.ro() + } + return x + case abi.Pointer: + ptr := v.ptr + if v.flag&flagIndir != 0 { + ptr = *(*unsafe.Pointer)(ptr) + } + // The returned value's address is v's value. + if ptr == nil { + return Value{} + } + tt := (*ptrType)(unsafe.Pointer(v.typ)) + typ := tt.Elem + fl := v.flag&flagRO | flagIndir | flagAddr + fl |= flag(typ.Kind()) + return Value{typ, ptr, fl} + } + panic(&ValueError{"reflectlite.Value.Elem", v.kind()}) +} + +func valueInterface(v Value) any { + if v.flag == 0 { + panic(&ValueError{"reflectlite.Value.Interface", 0}) + } + + if v.kind() == abi.Interface { + // Special case: return the element inside the interface. + // Empty interface has one layout, all interfaces with + // methods have a second layout. + if v.numMethod() == 0 { + return *(*any)(v.ptr) + } + return *(*interface { + M() + })(v.ptr) + } + + // TODO: pass safe to packEface so we don't need to copy if safe==true? + return packEface(v) +} + +// IsNil reports whether its argument v is nil. The argument must be +// a chan, func, interface, map, pointer, or slice value; if it is +// not, IsNil panics. Note that IsNil is not always equivalent to a +// regular comparison with nil in Go. For example, if v was created +// by calling ValueOf with an uninitialized interface variable i, +// i==nil will be true but v.IsNil will panic as v will be the zero +// Value. +func (v Value) IsNil() bool { + k := v.kind() + switch k { + case abi.Chan, abi.Func, abi.Map, abi.Pointer, abi.UnsafePointer: + // if v.flag&flagMethod != 0 { + // return false + // } + ptr := v.ptr + if v.flag&flagIndir != 0 { + ptr = *(*unsafe.Pointer)(ptr) + } + return ptr == nil + case abi.Interface, abi.Slice: + // Both interface and slice are nil if first word is 0. + // Both are always bigger than a word; assume flagIndir. + return *(*unsafe.Pointer)(v.ptr) == nil + } + panic(&ValueError{"reflectlite.Value.IsNil", v.kind()}) +} + +// IsValid reports whether v represents a value. +// It returns false if v is the zero Value. +// If IsValid returns false, all other methods except String panic. +// Most functions and methods never return an invalid Value. +// If one does, its documentation states the conditions explicitly. +func (v Value) IsValid() bool { + return v.flag != 0 +} + +// Kind returns v's Kind. +// If v is the zero Value (IsValid returns false), Kind returns Invalid. +func (v Value) Kind() Kind { + return v.kind() +} + +// implemented in runtime: +func chanlen(unsafe.Pointer) int +func maplen(unsafe.Pointer) int + +// Len returns v's length. +// It panics if v's Kind is not Array, Chan, Map, Slice, or String. +func (v Value) Len() int { + k := v.kind() + switch k { + case abi.Array: + tt := (*arrayType)(unsafe.Pointer(v.typ)) + return int(tt.Len) + case abi.Chan: + return chanlen(v.pointer()) + case abi.Map: + return maplen(v.pointer()) + case abi.Slice: + // Slice is bigger than a word; assume flagIndir. + return (*unsafeheader.Slice)(v.ptr).Len + case abi.String: + // String is bigger than a word; assume flagIndir. + return (*unsafeheader.String)(v.ptr).Len + } + panic(&ValueError{"reflect.Value.Len", v.kind()}) +} + +// NumMethod returns the number of exported methods in the value's method set. +func (v Value) numMethod() int { + if v.typ == nil { + panic(&ValueError{"reflectlite.Value.NumMethod", abi.Invalid}) + } + return v.typ.NumMethod() +} + +// Set assigns x to the value v. +// It panics if CanSet returns false. +// As in Go, x's value must be assignable to v's type. +func (v Value) Set(x Value) { + v.mustBeAssignable() + x.mustBeExported() // do not let unexported x leak + var target unsafe.Pointer + if v.kind() == abi.Interface { + target = v.ptr + } + x = x.assignTo("reflectlite.Set", v.typ, target) + if x.flag&flagIndir != 0 { + typedmemmove(v.typ, v.ptr, x.ptr) + } else { + *(*unsafe.Pointer)(v.ptr) = x.ptr + } +} + +// Type returns v's type. +func (v Value) Type() Type { + f := v.flag + if f == 0 { + panic(&ValueError{"reflectlite.Value.Type", abi.Invalid}) + } + // Method values not supported. + return toRType(v.typ) +} + +/* + * constructors + */ + +// implemented in package runtime +func unsafe_New(*abi.Type) unsafe.Pointer + +// ValueOf returns a new Value initialized to the concrete value +// stored in the interface i. ValueOf(nil) returns the zero Value. +func ValueOf(i any) Value { + if i == nil { + return Value{} + } + + // TODO: Maybe allow contents of a Value to live on the stack. + // For now we make the contents always escape to the heap. It + // makes life easier in a few places (see chanrecv/mapassign + // comment below). + escapes(i) + + return unpackEface(i) +} + +// assignTo returns a value v that can be assigned directly to typ. +// It panics if v is not assignable to typ. +// For a conversion to an interface type, target is a suggested scratch space to use. +func (v Value) assignTo(context string, dst *abi.Type, target unsafe.Pointer) Value { + // if v.flag&flagMethod != 0 { + // v = makeMethodValue(context, v) + // } + + switch { + case directlyAssignable(dst, v.typ): + // Overwrite type so that they match. + // Same memory layout, so no harm done. + fl := v.flag&(flagAddr|flagIndir) | v.flag.ro() + fl |= flag(dst.Kind()) + return Value{dst, v.ptr, fl} + + case implements(dst, v.typ): + if target == nil { + target = unsafe_New(dst) + } + if v.Kind() == abi.Interface && v.IsNil() { + // A nil ReadWriter passed to nil Reader is OK, + // but using ifaceE2I below will panic. + // Avoid the panic by returning a nil dst (e.g., Reader) explicitly. + return Value{dst, nil, flag(abi.Interface)} + } + x := valueInterface(v) + if dst.NumMethod() == 0 { + *(*any)(target) = x + } else { + ifaceE2I(dst, x, target) + } + return Value{dst, target, flagIndir | flag(abi.Interface)} + } + + // Failed. + panic(context + ": value of type " + toRType(v.typ).String() + " is not assignable to type " + toRType(dst).String()) +} + +// arrayAt returns the i-th element of p, +// an array whose elements are eltSize bytes wide. +// The array pointed at by p must have at least i+1 elements: +// it is invalid (but impossible to check here) to pass i >= len, +// because then the result will point outside the array. +// whySafe must explain why i < len. (Passing "i < len" is fine; +// the benefit is to surface this assumption at the call site.) +func arrayAt(p unsafe.Pointer, i int, eltSize uintptr, whySafe string) unsafe.Pointer { + return add(p, uintptr(i)*eltSize, "i < len") +} + +func ifaceE2I(t *abi.Type, src any, dst unsafe.Pointer) + +// typedmemmove copies a value of type t to dst from src. +// +//go:noescape +func typedmemmove(t *abi.Type, dst, src unsafe.Pointer) + +// Dummy annotation marking that the value x escapes, +// for use in cases where the reflect code is so clever that +// the compiler cannot follow. +func escapes(x any) { + if dummy.b { + dummy.x = x + } +} + +var dummy struct { + b bool + x any +} diff --git a/src/internal/safefilepath/path.go b/src/internal/safefilepath/path.go new file mode 100644 index 0000000..0f0a270 --- /dev/null +++ b/src/internal/safefilepath/path.go @@ -0,0 +1,21 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package safefilepath manipulates operating-system file paths. +package safefilepath + +import ( + "errors" +) + +var errInvalidPath = errors.New("invalid path") + +// FromFS converts a slash-separated path into an operating-system path. +// +// FromFS returns an error if the path cannot be represented by the operating +// system. For example, paths containing '\' and ':' characters are rejected +// on Windows. +func FromFS(path string) (string, error) { + return fromFS(path) +} diff --git a/src/internal/safefilepath/path_other.go b/src/internal/safefilepath/path_other.go new file mode 100644 index 0000000..974e775 --- /dev/null +++ b/src/internal/safefilepath/path_other.go @@ -0,0 +1,23 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !windows + +package safefilepath + +import "runtime" + +func fromFS(path string) (string, error) { + if runtime.GOOS == "plan9" { + if len(path) > 0 && path[0] == '#' { + return "", errInvalidPath + } + } + for i := range path { + if path[i] == 0 { + return "", errInvalidPath + } + } + return path, nil +} diff --git a/src/internal/safefilepath/path_test.go b/src/internal/safefilepath/path_test.go new file mode 100644 index 0000000..dc662c1 --- /dev/null +++ b/src/internal/safefilepath/path_test.go @@ -0,0 +1,88 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package safefilepath_test + +import ( + "internal/safefilepath" + "os" + "path/filepath" + "runtime" + "testing" +) + +type PathTest struct { + path, result string +} + +const invalid = "" + +var fspathtests = []PathTest{ + {".", "."}, + {"/a/b/c", "/a/b/c"}, + {"a\x00b", invalid}, +} + +var winreservedpathtests = []PathTest{ + {`a\b`, `a\b`}, + {`a:b`, `a:b`}, + {`a/b:c`, `a/b:c`}, + {`NUL`, `NUL`}, + {`./com1`, `./com1`}, + {`a/nul/b`, `a/nul/b`}, +} + +// Whether a reserved name with an extension is reserved or not varies by +// Windows version. +var winreservedextpathtests = []PathTest{ + {"nul.txt", "nul.txt"}, + {"a/nul.txt/b", "a/nul.txt/b"}, +} + +var plan9reservedpathtests = []PathTest{ + {`#c`, `#c`}, +} + +func TestFromFS(t *testing.T) { + switch runtime.GOOS { + case "windows": + if canWriteFile(t, "NUL") { + t.Errorf("can unexpectedly write a file named NUL on Windows") + } + if canWriteFile(t, "nul.txt") { + fspathtests = append(fspathtests, winreservedextpathtests...) + } else { + winreservedpathtests = append(winreservedpathtests, winreservedextpathtests...) + } + for i := range winreservedpathtests { + winreservedpathtests[i].result = invalid + } + for i := range fspathtests { + fspathtests[i].result = filepath.FromSlash(fspathtests[i].result) + } + case "plan9": + for i := range plan9reservedpathtests { + plan9reservedpathtests[i].result = invalid + } + } + tests := fspathtests + tests = append(tests, winreservedpathtests...) + tests = append(tests, plan9reservedpathtests...) + for _, test := range tests { + got, err := safefilepath.FromFS(test.path) + if (got == "") != (err != nil) { + t.Errorf(`FromFS(%q) = %q, %v; want "" only if err != nil`, test.path, got, err) + } + if got != test.result { + t.Errorf("FromFS(%q) = %q, %v; want %q", test.path, got, err, test.result) + } + } +} + +func canWriteFile(t *testing.T, name string) bool { + path := filepath.Join(t.TempDir(), name) + os.WriteFile(path, []byte("ok"), 0666) + b, _ := os.ReadFile(path) + return string(b) == "ok" +} diff --git a/src/internal/safefilepath/path_windows.go b/src/internal/safefilepath/path_windows.go new file mode 100644 index 0000000..7cfd6ce --- /dev/null +++ b/src/internal/safefilepath/path_windows.go @@ -0,0 +1,141 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package safefilepath + +import ( + "syscall" + "unicode/utf8" +) + +func fromFS(path string) (string, error) { + if !utf8.ValidString(path) { + return "", errInvalidPath + } + for len(path) > 1 && path[0] == '/' && path[1] == '/' { + path = path[1:] + } + containsSlash := false + for p := path; p != ""; { + // Find the next path element. + i := 0 + for i < len(p) && p[i] != '/' { + switch p[i] { + case 0, '\\', ':': + return "", errInvalidPath + } + i++ + } + part := p[:i] + if i < len(p) { + containsSlash = true + p = p[i+1:] + } else { + p = "" + } + if IsReservedName(part) { + return "", errInvalidPath + } + } + if containsSlash { + // We can't depend on strings, so substitute \ for / manually. + buf := []byte(path) + for i, b := range buf { + if b == '/' { + buf[i] = '\\' + } + } + path = string(buf) + } + return path, nil +} + +// IsReservedName reports if name is a Windows reserved device name. +// It does not detect names with an extension, which are also reserved on some Windows versions. +// +// For details, search for PRN in +// https://docs.microsoft.com/en-us/windows/desktop/fileio/naming-a-file. +func IsReservedName(name string) bool { + // Device names can have arbitrary trailing characters following a dot or colon. + base := name + for i := 0; i < len(base); i++ { + switch base[i] { + case ':', '.': + base = base[:i] + } + } + // Trailing spaces in the last path element are ignored. + for len(base) > 0 && base[len(base)-1] == ' ' { + base = base[:len(base)-1] + } + if !isReservedBaseName(base) { + return false + } + if len(base) == len(name) { + return true + } + // The path element is a reserved name with an extension. + // Some Windows versions consider this a reserved name, + // while others do not. Use FullPath to see if the name is + // reserved. + if p, _ := syscall.FullPath(name); len(p) >= 4 && p[:4] == `\\.\` { + return true + } + return false +} + +func isReservedBaseName(name string) bool { + if len(name) == 3 { + switch string([]byte{toUpper(name[0]), toUpper(name[1]), toUpper(name[2])}) { + case "CON", "PRN", "AUX", "NUL": + return true + } + } + if len(name) >= 4 { + switch string([]byte{toUpper(name[0]), toUpper(name[1]), toUpper(name[2])}) { + case "COM", "LPT": + if len(name) == 4 && '1' <= name[3] && name[3] <= '9' { + return true + } + // Superscript ¹, ², and ³ are considered numbers as well. + switch name[3:] { + case "\u00b2", "\u00b3", "\u00b9": + return true + } + return false + } + } + + // Passing CONIN$ or CONOUT$ to CreateFile opens a console handle. + // https://learn.microsoft.com/en-us/windows/win32/api/fileapi/nf-fileapi-createfilea#consoles + // + // While CONIN$ and CONOUT$ aren't documented as being files, + // they behave the same as CON. For example, ./CONIN$ also opens the console input. + if len(name) == 6 && name[5] == '$' && equalFold(name, "CONIN$") { + return true + } + if len(name) == 7 && name[6] == '$' && equalFold(name, "CONOUT$") { + return true + } + return false +} + +func equalFold(a, b string) bool { + if len(a) != len(b) { + return false + } + for i := 0; i < len(a); i++ { + if toUpper(a[i]) != toUpper(b[i]) { + return false + } + } + return true +} + +func toUpper(c byte) byte { + if 'a' <= c && c <= 'z' { + return c - ('a' - 'A') + } + return c +} diff --git a/src/internal/saferio/io.go b/src/internal/saferio/io.go new file mode 100644 index 0000000..66cc044 --- /dev/null +++ b/src/internal/saferio/io.go @@ -0,0 +1,135 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package saferio provides I/O functions that avoid allocating large +// amounts of memory unnecessarily. This is intended for packages that +// read data from an [io.Reader] where the size is part of the input +// data but the input may be corrupt, or may be provided by an +// untrustworthy attacker. +package saferio + +import ( + "io" + "reflect" +) + +// chunk is an arbitrary limit on how much memory we are willing +// to allocate without concern. +const chunk = 10 << 20 // 10M + +// ReadData reads n bytes from the input stream, but avoids allocating +// all n bytes if n is large. This avoids crashing the program by +// allocating all n bytes in cases where n is incorrect. +// +// The error is io.EOF only if no bytes were read. +// If an io.EOF happens after reading some but not all the bytes, +// ReadData returns io.ErrUnexpectedEOF. +func ReadData(r io.Reader, n uint64) ([]byte, error) { + if int64(n) < 0 || n != uint64(int(n)) { + // n is too large to fit in int, so we can't allocate + // a buffer large enough. Treat this as a read failure. + return nil, io.ErrUnexpectedEOF + } + + if n < chunk { + buf := make([]byte, n) + _, err := io.ReadFull(r, buf) + if err != nil { + return nil, err + } + return buf, nil + } + + var buf []byte + buf1 := make([]byte, chunk) + for n > 0 { + next := n + if next > chunk { + next = chunk + } + _, err := io.ReadFull(r, buf1[:next]) + if err != nil { + if len(buf) > 0 && err == io.EOF { + err = io.ErrUnexpectedEOF + } + return nil, err + } + buf = append(buf, buf1[:next]...) + n -= next + } + return buf, nil +} + +// ReadDataAt reads n bytes from the input stream at off, but avoids +// allocating all n bytes if n is large. This avoids crashing the program +// by allocating all n bytes in cases where n is incorrect. +func ReadDataAt(r io.ReaderAt, n uint64, off int64) ([]byte, error) { + if int64(n) < 0 || n != uint64(int(n)) { + // n is too large to fit in int, so we can't allocate + // a buffer large enough. Treat this as a read failure. + return nil, io.ErrUnexpectedEOF + } + + if n < chunk { + buf := make([]byte, n) + _, err := r.ReadAt(buf, off) + if err != nil { + // io.SectionReader can return EOF for n == 0, + // but for our purposes that is a success. + if err != io.EOF || n > 0 { + return nil, err + } + } + return buf, nil + } + + var buf []byte + buf1 := make([]byte, chunk) + for n > 0 { + next := n + if next > chunk { + next = chunk + } + _, err := r.ReadAt(buf1[:next], off) + if err != nil { + return nil, err + } + buf = append(buf, buf1[:next]...) + n -= next + off += int64(next) + } + return buf, nil +} + +// SliceCap returns the capacity to use when allocating a slice. +// After the slice is allocated with the capacity, it should be +// built using append. This will avoid allocating too much memory +// if the capacity is large and incorrect. +// +// A negative result means that the value is always too big. +// +// The element type is described by passing a pointer to a value of that type. +// This would ideally use generics, but this code is built with +// the bootstrap compiler which need not support generics. +// We use a pointer so that we can handle slices of interface type. +func SliceCap(v any, c uint64) int { + if int64(c) < 0 || c != uint64(int(c)) { + return -1 + } + typ := reflect.TypeOf(v) + if typ.Kind() != reflect.Ptr { + panic("SliceCap called with non-pointer type") + } + size := uint64(typ.Elem().Size()) + if size > 0 && c > (1<<64-1)/size { + return -1 + } + if c*size > chunk { + c = uint64(chunk / size) + if c == 0 { + c = 1 + } + } + return int(c) +} diff --git a/src/internal/saferio/io_test.go b/src/internal/saferio/io_test.go new file mode 100644 index 0000000..356c9eb --- /dev/null +++ b/src/internal/saferio/io_test.go @@ -0,0 +1,136 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package saferio + +import ( + "bytes" + "io" + "testing" +) + +func TestReadData(t *testing.T) { + const count = 100 + input := bytes.Repeat([]byte{'a'}, count) + + t.Run("small", func(t *testing.T) { + got, err := ReadData(bytes.NewReader(input), count) + if err != nil { + t.Fatal(err) + } + if !bytes.Equal(got, input) { + t.Errorf("got %v, want %v", got, input) + } + }) + + t.Run("large", func(t *testing.T) { + _, err := ReadData(bytes.NewReader(input), 10<<30) + if err == nil { + t.Error("large read succeeded unexpectedly") + } + }) + + t.Run("maxint", func(t *testing.T) { + _, err := ReadData(bytes.NewReader(input), 1<<62) + if err == nil { + t.Error("large read succeeded unexpectedly") + } + }) + + t.Run("small-EOF", func(t *testing.T) { + _, err := ReadData(bytes.NewReader(nil), chunk-1) + if err != io.EOF { + t.Errorf("ReadData = %v, want io.EOF", err) + } + }) + + t.Run("large-EOF", func(t *testing.T) { + _, err := ReadData(bytes.NewReader(nil), chunk+1) + if err != io.EOF { + t.Errorf("ReadData = %v, want io.EOF", err) + } + }) + + t.Run("large-UnexpectedEOF", func(t *testing.T) { + _, err := ReadData(bytes.NewReader(make([]byte, chunk)), chunk+1) + if err != io.ErrUnexpectedEOF { + t.Errorf("ReadData = %v, want io.ErrUnexpectedEOF", err) + } + }) +} + +func TestReadDataAt(t *testing.T) { + const count = 100 + input := bytes.Repeat([]byte{'a'}, count) + + t.Run("small", func(t *testing.T) { + got, err := ReadDataAt(bytes.NewReader(input), count, 0) + if err != nil { + t.Fatal(err) + } + if !bytes.Equal(got, input) { + t.Errorf("got %v, want %v", got, input) + } + }) + + t.Run("large", func(t *testing.T) { + _, err := ReadDataAt(bytes.NewReader(input), 10<<30, 0) + if err == nil { + t.Error("large read succeeded unexpectedly") + } + }) + + t.Run("maxint", func(t *testing.T) { + _, err := ReadDataAt(bytes.NewReader(input), 1<<62, 0) + if err == nil { + t.Error("large read succeeded unexpectedly") + } + }) + + t.Run("SectionReader", func(t *testing.T) { + // Reading 0 bytes from an io.SectionReader at the end + // of the section will return EOF, but ReadDataAt + // should succeed and return 0 bytes. + sr := io.NewSectionReader(bytes.NewReader(input), 0, 0) + got, err := ReadDataAt(sr, 0, 0) + if err != nil { + t.Fatal(err) + } + if len(got) > 0 { + t.Errorf("got %d bytes, expected 0", len(got)) + } + }) +} + +func TestSliceCap(t *testing.T) { + t.Run("small", func(t *testing.T) { + c := SliceCap((*int)(nil), 10) + if c != 10 { + t.Errorf("got capacity %d, want %d", c, 10) + } + }) + + t.Run("large", func(t *testing.T) { + c := SliceCap((*byte)(nil), 1<<30) + if c < 0 { + t.Error("SliceCap failed unexpectedly") + } else if c == 1<<30 { + t.Errorf("got capacity %d which is too high", c) + } + }) + + t.Run("maxint", func(t *testing.T) { + c := SliceCap((*byte)(nil), 1<<63) + if c >= 0 { + t.Errorf("SliceCap returned %d, expected failure", c) + } + }) + + t.Run("overflow", func(t *testing.T) { + c := SliceCap((*int64)(nil), 1<<62) + if c >= 0 { + t.Errorf("SliceCap returned %d, expected failure", c) + } + }) +} diff --git a/src/internal/singleflight/singleflight.go b/src/internal/singleflight/singleflight.go new file mode 100644 index 0000000..d0e6d2f --- /dev/null +++ b/src/internal/singleflight/singleflight.go @@ -0,0 +1,123 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package singleflight provides a duplicate function call suppression +// mechanism. +package singleflight + +import "sync" + +// call is an in-flight or completed singleflight.Do call +type call struct { + wg sync.WaitGroup + + // These fields are written once before the WaitGroup is done + // and are only read after the WaitGroup is done. + val any + err error + + // These fields are read and written with the singleflight + // mutex held before the WaitGroup is done, and are read but + // not written after the WaitGroup is done. + dups int + chans []chan<- Result +} + +// Group represents a class of work and forms a namespace in +// which units of work can be executed with duplicate suppression. +type Group struct { + mu sync.Mutex // protects m + m map[string]*call // lazily initialized +} + +// Result holds the results of Do, so they can be passed +// on a channel. +type Result struct { + Val any + Err error + Shared bool +} + +// Do executes and returns the results of the given function, making +// sure that only one execution is in-flight for a given key at a +// time. If a duplicate comes in, the duplicate caller waits for the +// original to complete and receives the same results. +// The return value shared indicates whether v was given to multiple callers. +func (g *Group) Do(key string, fn func() (any, error)) (v any, err error, shared bool) { + g.mu.Lock() + if g.m == nil { + g.m = make(map[string]*call) + } + if c, ok := g.m[key]; ok { + c.dups++ + g.mu.Unlock() + c.wg.Wait() + return c.val, c.err, true + } + c := new(call) + c.wg.Add(1) + g.m[key] = c + g.mu.Unlock() + + g.doCall(c, key, fn) + return c.val, c.err, c.dups > 0 +} + +// DoChan is like Do but returns a channel that will receive the +// results when they are ready. +func (g *Group) DoChan(key string, fn func() (any, error)) <-chan Result { + ch := make(chan Result, 1) + g.mu.Lock() + if g.m == nil { + g.m = make(map[string]*call) + } + if c, ok := g.m[key]; ok { + c.dups++ + c.chans = append(c.chans, ch) + g.mu.Unlock() + return ch + } + c := &call{chans: []chan<- Result{ch}} + c.wg.Add(1) + g.m[key] = c + g.mu.Unlock() + + go g.doCall(c, key, fn) + + return ch +} + +// doCall handles the single call for a key. +func (g *Group) doCall(c *call, key string, fn func() (any, error)) { + c.val, c.err = fn() + + g.mu.Lock() + c.wg.Done() + if g.m[key] == c { + delete(g.m, key) + } + for _, ch := range c.chans { + ch <- Result{c.val, c.err, c.dups > 0} + } + g.mu.Unlock() +} + +// ForgetUnshared tells the singleflight to forget about a key if it is not +// shared with any other goroutines. Future calls to Do for a forgotten key +// will call the function rather than waiting for an earlier call to complete. +// Returns whether the key was forgotten or unknown--that is, whether no +// other goroutines are waiting for the result. +func (g *Group) ForgetUnshared(key string) bool { + g.mu.Lock() + defer g.mu.Unlock() + c, ok := g.m[key] + if !ok { + return true + } + if c.dups == 0 { + delete(g.m, key) + return true + } + return false +} diff --git a/src/internal/singleflight/singleflight_test.go b/src/internal/singleflight/singleflight_test.go new file mode 100644 index 0000000..279e1be --- /dev/null +++ b/src/internal/singleflight/singleflight_test.go @@ -0,0 +1,186 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package singleflight + +import ( + "errors" + "fmt" + "sync" + "sync/atomic" + "testing" + "time" +) + +func TestDo(t *testing.T) { + var g Group + v, err, _ := g.Do("key", func() (any, error) { + return "bar", nil + }) + if got, want := fmt.Sprintf("%v (%T)", v, v), "bar (string)"; got != want { + t.Errorf("Do = %v; want %v", got, want) + } + if err != nil { + t.Errorf("Do error = %v", err) + } +} + +func TestDoErr(t *testing.T) { + var g Group + someErr := errors.New("some error") + v, err, _ := g.Do("key", func() (any, error) { + return nil, someErr + }) + if err != someErr { + t.Errorf("Do error = %v; want someErr %v", err, someErr) + } + if v != nil { + t.Errorf("unexpected non-nil value %#v", v) + } +} + +func TestDoDupSuppress(t *testing.T) { + var g Group + var wg1, wg2 sync.WaitGroup + c := make(chan string, 1) + var calls atomic.Int32 + fn := func() (any, error) { + if calls.Add(1) == 1 { + // First invocation. + wg1.Done() + } + v := <-c + c <- v // pump; make available for any future calls + + time.Sleep(10 * time.Millisecond) // let more goroutines enter Do + + return v, nil + } + + const n = 10 + wg1.Add(1) + for i := 0; i < n; i++ { + wg1.Add(1) + wg2.Add(1) + go func() { + defer wg2.Done() + wg1.Done() + v, err, _ := g.Do("key", fn) + if err != nil { + t.Errorf("Do error: %v", err) + return + } + if s, _ := v.(string); s != "bar" { + t.Errorf("Do = %T %v; want %q", v, v, "bar") + } + }() + } + wg1.Wait() + // At least one goroutine is in fn now and all of them have at + // least reached the line before the Do. + c <- "bar" + wg2.Wait() + if got := calls.Load(); got <= 0 || got >= n { + t.Errorf("number of calls = %d; want over 0 and less than %d", got, n) + } +} + +func TestForgetUnshared(t *testing.T) { + var g Group + + var firstStarted, firstFinished sync.WaitGroup + + firstStarted.Add(1) + firstFinished.Add(1) + + key := "key" + firstCh := make(chan struct{}) + go func() { + g.Do(key, func() (i interface{}, e error) { + firstStarted.Done() + <-firstCh + return + }) + firstFinished.Done() + }() + + firstStarted.Wait() + g.ForgetUnshared(key) // from this point no two function using same key should be executed concurrently + + secondCh := make(chan struct{}) + go func() { + g.Do(key, func() (i interface{}, e error) { + // Notify that we started + secondCh <- struct{}{} + <-secondCh + return 2, nil + }) + }() + + <-secondCh + + resultCh := g.DoChan(key, func() (i interface{}, e error) { + panic("third must not be started") + }) + + if g.ForgetUnshared(key) { + t.Errorf("Before first goroutine finished, key %q is shared, should return false", key) + } + + close(firstCh) + firstFinished.Wait() + + if g.ForgetUnshared(key) { + t.Errorf("After first goroutine finished, key %q is still shared, should return false", key) + } + + secondCh <- struct{}{} + + if result := <-resultCh; result.Val != 2 { + t.Errorf("We should receive result produced by second call, expected: 2, got %d", result.Val) + } +} + +func TestDoAndForgetUnsharedRace(t *testing.T) { + t.Parallel() + + var g Group + key := "key" + d := time.Millisecond + for { + var calls, shared atomic.Int64 + const n = 1000 + var wg sync.WaitGroup + wg.Add(n) + for i := 0; i < n; i++ { + go func() { + g.Do(key, func() (interface{}, error) { + time.Sleep(d) + return calls.Add(1), nil + }) + if !g.ForgetUnshared(key) { + shared.Add(1) + } + wg.Done() + }() + } + wg.Wait() + + if calls.Load() != 1 { + // The goroutines didn't park in g.Do in time, + // so the key was re-added and may have been shared after the call. + // Try again with more time to park. + d *= 2 + continue + } + + // All of the Do calls ended up sharing the first + // invocation, so the key should have been unused + // (and therefore unshared) when they returned. + if shared.Load() > 0 { + t.Errorf("after a single shared Do, ForgetUnshared returned false %d times", shared.Load()) + } + break + } +} diff --git a/src/internal/syscall/execenv/execenv_default.go b/src/internal/syscall/execenv/execenv_default.go new file mode 100644 index 0000000..335647c --- /dev/null +++ b/src/internal/syscall/execenv/execenv_default.go @@ -0,0 +1,19 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !windows + +package execenv + +import "syscall" + +// Default will return the default environment +// variables based on the process attributes +// provided. +// +// Defaults to syscall.Environ() on all platforms +// other than Windows. +func Default(sys *syscall.SysProcAttr) ([]string, error) { + return syscall.Environ(), nil +} diff --git a/src/internal/syscall/execenv/execenv_windows.go b/src/internal/syscall/execenv/execenv_windows.go new file mode 100644 index 0000000..2a89ed1 --- /dev/null +++ b/src/internal/syscall/execenv/execenv_windows.go @@ -0,0 +1,47 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package execenv + +import ( + "internal/syscall/windows" + "syscall" + "unsafe" +) + +// Default will return the default environment +// variables based on the process attributes +// provided. +// +// If the process attributes contain a token, then +// the environment variables will be sourced from +// the defaults for that user token, otherwise they +// will be sourced from syscall.Environ(). +func Default(sys *syscall.SysProcAttr) (env []string, err error) { + if sys == nil || sys.Token == 0 { + return syscall.Environ(), nil + } + var blockp *uint16 + err = windows.CreateEnvironmentBlock(&blockp, sys.Token, false) + if err != nil { + return nil, err + } + defer windows.DestroyEnvironmentBlock(blockp) + + const size = unsafe.Sizeof(*blockp) + for *blockp != 0 { // environment block ends with empty string + // find NUL terminator + end := unsafe.Add(unsafe.Pointer(blockp), size) + for *(*uint16)(end) != 0 { + end = unsafe.Add(end, size) + } + + entry := unsafe.Slice(blockp, (uintptr(end)-uintptr(unsafe.Pointer(blockp)))/2) + env = append(env, syscall.UTF16ToString(entry)) + blockp = (*uint16)(unsafe.Add(end, size)) + } + return +} diff --git a/src/internal/syscall/unix/asm_aix_ppc64.s b/src/internal/syscall/unix/asm_aix_ppc64.s new file mode 100644 index 0000000..9e82e3e --- /dev/null +++ b/src/internal/syscall/unix/asm_aix_ppc64.s @@ -0,0 +1,12 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +// +// System calls for aix/ppc64 are implemented in syscall/syscall_aix.go +// + +TEXT ·syscall6(SB),NOSPLIT,$0 + JMP syscall·syscall6(SB) diff --git a/src/internal/syscall/unix/asm_darwin.s b/src/internal/syscall/unix/asm_darwin.s new file mode 100644 index 0000000..8662c28 --- /dev/null +++ b/src/internal/syscall/unix/asm_darwin.s @@ -0,0 +1,24 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +TEXT ·libc_getentropy_trampoline(SB),NOSPLIT,$0-0; JMP libc_getentropy(SB) +TEXT ·libc_getaddrinfo_trampoline(SB),NOSPLIT,$0-0; JMP libc_getaddrinfo(SB) +TEXT ·libc_freeaddrinfo_trampoline(SB),NOSPLIT,$0-0; JMP libc_freeaddrinfo(SB) +TEXT ·libc_getnameinfo_trampoline(SB),NOSPLIT,$0-0; JMP libc_getnameinfo(SB) +TEXT ·libc_gai_strerror_trampoline(SB),NOSPLIT,$0-0; JMP libc_gai_strerror(SB) +TEXT ·libresolv_res_9_ninit_trampoline(SB),NOSPLIT,$0-0; JMP libresolv_res_9_ninit(SB) +TEXT ·libresolv_res_9_nclose_trampoline(SB),NOSPLIT,$0-0; JMP libresolv_res_9_nclose(SB) +TEXT ·libresolv_res_9_nsearch_trampoline(SB),NOSPLIT,$0-0; JMP libresolv_res_9_nsearch(SB) +TEXT ·libc_grantpt_trampoline(SB),NOSPLIT,$0-0; JMP libc_grantpt(SB) +TEXT ·libc_unlockpt_trampoline(SB),NOSPLIT,$0-0; JMP libc_unlockpt(SB) +TEXT ·libc_ptsname_r_trampoline(SB),NOSPLIT,$0-0; JMP libc_ptsname_r(SB) +TEXT ·libc_posix_openpt_trampoline(SB),NOSPLIT,$0-0; JMP libc_posix_openpt(SB) +TEXT ·libc_getgrouplist_trampoline(SB),NOSPLIT,$0-0; JMP libc_getgrouplist(SB) +TEXT ·libc_getpwnam_r_trampoline(SB),NOSPLIT,$0-0; JMP libc_getpwnam_r(SB) +TEXT ·libc_getpwuid_r_trampoline(SB),NOSPLIT,$0-0; JMP libc_getpwuid_r(SB) +TEXT ·libc_getgrnam_r_trampoline(SB),NOSPLIT,$0-0; JMP libc_getgrnam_r(SB) +TEXT ·libc_getgrgid_r_trampoline(SB),NOSPLIT,$0-0; JMP libc_getgrgid_r(SB) +TEXT ·libc_sysconf_trampoline(SB),NOSPLIT,$0-0; JMP libc_sysconf(SB) diff --git a/src/internal/syscall/unix/asm_solaris.s b/src/internal/syscall/unix/asm_solaris.s new file mode 100644 index 0000000..2057338 --- /dev/null +++ b/src/internal/syscall/unix/asm_solaris.s @@ -0,0 +1,10 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +#include "textflag.h" + +// System calls for Solaris are implemented in runtime/syscall_solaris.go + +TEXT ·syscall6(SB),NOSPLIT,$0-88 + JMP syscall·sysvicall6(SB) diff --git a/src/internal/syscall/unix/at.go b/src/internal/syscall/unix/at.go new file mode 100644 index 0000000..cfb6e41 --- /dev/null +++ b/src/internal/syscall/unix/at.go @@ -0,0 +1,40 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build dragonfly || freebsd || linux || netbsd || (openbsd && mips64) + +package unix + +import ( + "syscall" + "unsafe" +) + +func Unlinkat(dirfd int, path string, flags int) error { + p, err := syscall.BytePtrFromString(path) + if err != nil { + return err + } + + _, _, errno := syscall.Syscall(unlinkatTrap, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(flags)) + if errno != 0 { + return errno + } + + return nil +} + +func Openat(dirfd int, path string, flags int, perm uint32) (int, error) { + p, err := syscall.BytePtrFromString(path) + if err != nil { + return 0, err + } + + fd, _, errno := syscall.Syscall6(openatTrap, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(flags), uintptr(perm), 0, 0) + if errno != 0 { + return 0, errno + } + + return int(fd), nil +} diff --git a/src/internal/syscall/unix/at_aix.go b/src/internal/syscall/unix/at_aix.go new file mode 100644 index 0000000..3fe3285 --- /dev/null +++ b/src/internal/syscall/unix/at_aix.go @@ -0,0 +1,15 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +//go:cgo_import_dynamic libc_fstatat fstatat "libc.a/shr_64.o" +//go:cgo_import_dynamic libc_openat openat "libc.a/shr_64.o" +//go:cgo_import_dynamic libc_unlinkat unlinkat "libc.a/shr_64.o" + +const ( + AT_REMOVEDIR = 0x1 + AT_SYMLINK_NOFOLLOW = 0x1 + UTIME_OMIT = -0x3 +) diff --git a/src/internal/syscall/unix/at_fstatat.go b/src/internal/syscall/unix/at_fstatat.go new file mode 100644 index 0000000..8f25fe9 --- /dev/null +++ b/src/internal/syscall/unix/at_fstatat.go @@ -0,0 +1,28 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build dragonfly || (linux && !loong64) || netbsd || (openbsd && mips64) + +package unix + +import ( + "syscall" + "unsafe" +) + +func Fstatat(dirfd int, path string, stat *syscall.Stat_t, flags int) error { + var p *byte + p, err := syscall.BytePtrFromString(path) + if err != nil { + return err + } + + _, _, errno := syscall.Syscall6(fstatatTrap, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(unsafe.Pointer(stat)), uintptr(flags), 0, 0) + if errno != 0 { + return errno + } + + return nil + +} diff --git a/src/internal/syscall/unix/at_fstatat2.go b/src/internal/syscall/unix/at_fstatat2.go new file mode 100644 index 0000000..8d20e1a --- /dev/null +++ b/src/internal/syscall/unix/at_fstatat2.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build freebsd || (linux && loong64) + +package unix + +import "syscall" + +func Fstatat(dirfd int, path string, stat *syscall.Stat_t, flags int) error { + return syscall.Fstatat(dirfd, path, stat, flags) +} diff --git a/src/internal/syscall/unix/at_js.go b/src/internal/syscall/unix/at_js.go new file mode 100644 index 0000000..d05ccce --- /dev/null +++ b/src/internal/syscall/unix/at_js.go @@ -0,0 +1,13 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + // UTIME_OMIT is the sentinel value to indicate that a time value should not + // be changed. It is useful for example to indicate for example with UtimesNano + // to avoid changing AccessTime or ModifiedTime. + // Its value must match syscall/fs_js.go + UTIME_OMIT = -0x2 +) diff --git a/src/internal/syscall/unix/at_libc.go b/src/internal/syscall/unix/at_libc.go new file mode 100644 index 0000000..f48d379 --- /dev/null +++ b/src/internal/syscall/unix/at_libc.go @@ -0,0 +1,64 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build aix || solaris + +package unix + +import ( + "syscall" + "unsafe" +) + +//go:linkname procFstatat libc_fstatat +//go:linkname procOpenat libc_openat +//go:linkname procUnlinkat libc_unlinkat + +var ( + procFstatat, + procOpenat, + procUnlinkat uintptr +) + +func Unlinkat(dirfd int, path string, flags int) error { + p, err := syscall.BytePtrFromString(path) + if err != nil { + return err + } + + _, _, errno := syscall6(uintptr(unsafe.Pointer(&procUnlinkat)), 3, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(flags), 0, 0, 0) + if errno != 0 { + return errno + } + + return nil +} + +func Openat(dirfd int, path string, flags int, perm uint32) (int, error) { + p, err := syscall.BytePtrFromString(path) + if err != nil { + return 0, err + } + + fd, _, errno := syscall6(uintptr(unsafe.Pointer(&procOpenat)), 4, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(flags), uintptr(perm), 0, 0) + if errno != 0 { + return 0, errno + } + + return int(fd), nil +} + +func Fstatat(dirfd int, path string, stat *syscall.Stat_t, flags int) error { + p, err := syscall.BytePtrFromString(path) + if err != nil { + return err + } + + _, _, errno := syscall6(uintptr(unsafe.Pointer(&procFstatat)), 4, uintptr(dirfd), uintptr(unsafe.Pointer(p)), uintptr(unsafe.Pointer(stat)), uintptr(flags), 0, 0) + if errno != 0 { + return errno + } + + return nil +} diff --git a/src/internal/syscall/unix/at_libc2.go b/src/internal/syscall/unix/at_libc2.go new file mode 100644 index 0000000..93d0cf4 --- /dev/null +++ b/src/internal/syscall/unix/at_libc2.go @@ -0,0 +1,33 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build darwin || (openbsd && !mips64) + +package unix + +import ( + "syscall" + _ "unsafe" // for linkname +) + +func Unlinkat(dirfd int, path string, flags int) error { + return unlinkat(dirfd, path, flags) +} + +func Openat(dirfd int, path string, flags int, perm uint32) (int, error) { + return openat(dirfd, path, flags, perm) +} + +func Fstatat(dirfd int, path string, stat *syscall.Stat_t, flags int) error { + return fstatat(dirfd, path, stat, flags) +} + +//go:linkname unlinkat syscall.unlinkat +func unlinkat(dirfd int, path string, flags int) error + +//go:linkname openat syscall.openat +func openat(dirfd int, path string, flags int, perm uint32) (int, error) + +//go:linkname fstatat syscall.fstatat +func fstatat(dirfd int, path string, stat *syscall.Stat_t, flags int) error diff --git a/src/internal/syscall/unix/at_solaris.go b/src/internal/syscall/unix/at_solaris.go new file mode 100644 index 0000000..4ab224d --- /dev/null +++ b/src/internal/syscall/unix/at_solaris.go @@ -0,0 +1,21 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +// Implemented as sysvicall6 in runtime/syscall_solaris.go. +func syscall6(trap, nargs, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2 uintptr, err syscall.Errno) + +//go:cgo_import_dynamic libc_fstatat fstatat "libc.so" +//go:cgo_import_dynamic libc_openat openat "libc.so" +//go:cgo_import_dynamic libc_unlinkat unlinkat "libc.so" + +const ( + AT_REMOVEDIR = 0x1 + AT_SYMLINK_NOFOLLOW = 0x1000 + + UTIME_OMIT = -0x2 +) diff --git a/src/internal/syscall/unix/at_sysnum_darwin.go b/src/internal/syscall/unix/at_sysnum_darwin.go new file mode 100644 index 0000000..208ff34 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_darwin.go @@ -0,0 +1,10 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const AT_REMOVEDIR = 0x80 +const AT_SYMLINK_NOFOLLOW = 0x0020 + +const UTIME_OMIT = -0x2 diff --git a/src/internal/syscall/unix/at_sysnum_dragonfly.go b/src/internal/syscall/unix/at_sysnum_dragonfly.go new file mode 100644 index 0000000..b7ed3f7 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_dragonfly.go @@ -0,0 +1,16 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +const unlinkatTrap uintptr = syscall.SYS_UNLINKAT +const openatTrap uintptr = syscall.SYS_OPENAT +const fstatatTrap uintptr = syscall.SYS_FSTATAT + +const AT_REMOVEDIR = 0x2 +const AT_SYMLINK_NOFOLLOW = 0x1 + +const UTIME_OMIT = -0x1 diff --git a/src/internal/syscall/unix/at_sysnum_freebsd.go b/src/internal/syscall/unix/at_sysnum_freebsd.go new file mode 100644 index 0000000..9cd5da6 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_freebsd.go @@ -0,0 +1,18 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +const ( + AT_REMOVEDIR = 0x800 + AT_SYMLINK_NOFOLLOW = 0x200 + + UTIME_OMIT = -0x2 + + unlinkatTrap uintptr = syscall.SYS_UNLINKAT + openatTrap uintptr = syscall.SYS_OPENAT + posixFallocateTrap uintptr = syscall.SYS_POSIX_FALLOCATE +) diff --git a/src/internal/syscall/unix/at_sysnum_fstatat64_linux.go b/src/internal/syscall/unix/at_sysnum_fstatat64_linux.go new file mode 100644 index 0000000..445b0c3 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_fstatat64_linux.go @@ -0,0 +1,11 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm || mips || mipsle || 386 + +package unix + +import "syscall" + +const fstatatTrap uintptr = syscall.SYS_FSTATAT64 diff --git a/src/internal/syscall/unix/at_sysnum_fstatat_linux.go b/src/internal/syscall/unix/at_sysnum_fstatat_linux.go new file mode 100644 index 0000000..73a3da5 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_fstatat_linux.go @@ -0,0 +1,11 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build arm64 || riscv64 + +package unix + +import "syscall" + +const fstatatTrap uintptr = syscall.SYS_FSTATAT diff --git a/src/internal/syscall/unix/at_sysnum_linux.go b/src/internal/syscall/unix/at_sysnum_linux.go new file mode 100644 index 0000000..7c3b15c --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_linux.go @@ -0,0 +1,19 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +const unlinkatTrap uintptr = syscall.SYS_UNLINKAT +const openatTrap uintptr = syscall.SYS_OPENAT + +const ( + AT_EACCESS = 0x200 + AT_FDCWD = -0x64 + AT_REMOVEDIR = 0x200 + AT_SYMLINK_NOFOLLOW = 0x100 + + UTIME_OMIT = 0x3ffffffe +) diff --git a/src/internal/syscall/unix/at_sysnum_netbsd.go b/src/internal/syscall/unix/at_sysnum_netbsd.go new file mode 100644 index 0000000..becc1bd --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_netbsd.go @@ -0,0 +1,16 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +const unlinkatTrap uintptr = syscall.SYS_UNLINKAT +const openatTrap uintptr = syscall.SYS_OPENAT +const fstatatTrap uintptr = syscall.SYS_FSTATAT + +const AT_REMOVEDIR = 0x800 +const AT_SYMLINK_NOFOLLOW = 0x200 + +const UTIME_OMIT = (1 << 30) - 2 diff --git a/src/internal/syscall/unix/at_sysnum_newfstatat_linux.go b/src/internal/syscall/unix/at_sysnum_newfstatat_linux.go new file mode 100644 index 0000000..76edf67 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_newfstatat_linux.go @@ -0,0 +1,11 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build amd64 || mips64 || mips64le || ppc64 || ppc64le || s390x + +package unix + +import "syscall" + +const fstatatTrap uintptr = syscall.SYS_NEWFSTATAT diff --git a/src/internal/syscall/unix/at_sysnum_openbsd.go b/src/internal/syscall/unix/at_sysnum_openbsd.go new file mode 100644 index 0000000..fd38947 --- /dev/null +++ b/src/internal/syscall/unix/at_sysnum_openbsd.go @@ -0,0 +1,16 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +const unlinkatTrap uintptr = syscall.SYS_UNLINKAT +const openatTrap uintptr = syscall.SYS_OPENAT +const fstatatTrap uintptr = syscall.SYS_FSTATAT + +const AT_REMOVEDIR = 0x08 +const AT_SYMLINK_NOFOLLOW = 0x02 + +const UTIME_OMIT = -0x1 diff --git a/src/internal/syscall/unix/at_wasip1.go b/src/internal/syscall/unix/at_wasip1.go new file mode 100644 index 0000000..3d47d7e --- /dev/null +++ b/src/internal/syscall/unix/at_wasip1.go @@ -0,0 +1,13 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + // UTIME_OMIT is the sentinel value to indicate that a time value should not + // be changed. It is useful for example to indicate for example with UtimesNano + // to avoid changing AccessTime or ModifiedTime. + // Its value must match syscall/fs_wasip1.go + UTIME_OMIT = -0x2 +) diff --git a/src/internal/syscall/unix/constants.go b/src/internal/syscall/unix/constants.go new file mode 100644 index 0000000..e324589 --- /dev/null +++ b/src/internal/syscall/unix/constants.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package unix + +const ( + R_OK = 0x4 + W_OK = 0x2 + X_OK = 0x1 +) diff --git a/src/internal/syscall/unix/copy_file_range_linux.go b/src/internal/syscall/unix/copy_file_range_linux.go new file mode 100644 index 0000000..cf0a279 --- /dev/null +++ b/src/internal/syscall/unix/copy_file_range_linux.go @@ -0,0 +1,26 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "syscall" + "unsafe" +) + +func CopyFileRange(rfd int, roff *int64, wfd int, woff *int64, len int, flags int) (n int, err error) { + r1, _, errno := syscall.Syscall6(copyFileRangeTrap, + uintptr(rfd), + uintptr(unsafe.Pointer(roff)), + uintptr(wfd), + uintptr(unsafe.Pointer(woff)), + uintptr(len), + uintptr(flags), + ) + n = int(r1) + if errno != 0 { + err = errno + } + return +} diff --git a/src/internal/syscall/unix/eaccess_linux.go b/src/internal/syscall/unix/eaccess_linux.go new file mode 100644 index 0000000..5695a5e --- /dev/null +++ b/src/internal/syscall/unix/eaccess_linux.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +func Eaccess(path string, mode uint32) error { + return syscall.Faccessat(AT_FDCWD, path, mode, AT_EACCESS) +} diff --git a/src/internal/syscall/unix/eaccess_other.go b/src/internal/syscall/unix/eaccess_other.go new file mode 100644 index 0000000..23be118 --- /dev/null +++ b/src/internal/syscall/unix/eaccess_other.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix && !linux + +package unix + +import "syscall" + +func Eaccess(path string, mode uint32) error { + return syscall.ENOSYS +} diff --git a/src/internal/syscall/unix/fallocate_freebsd_386.go b/src/internal/syscall/unix/fallocate_freebsd_386.go new file mode 100644 index 0000000..535b23d --- /dev/null +++ b/src/internal/syscall/unix/fallocate_freebsd_386.go @@ -0,0 +1,17 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +func PosixFallocate(fd int, off int64, size int64) error { + // If successful, posix_fallocate() returns zero. It returns an error on failure, without + // setting errno. See https://man.freebsd.org/cgi/man.cgi?query=posix_fallocate&sektion=2&n=1 + r1, _, _ := syscall.Syscall6(posixFallocateTrap, uintptr(fd), uintptr(off), uintptr(off>>32), uintptr(size), uintptr(size>>32), 0) + if r1 != 0 { + return syscall.Errno(r1) + } + return nil +} diff --git a/src/internal/syscall/unix/fallocate_freebsd_64bit.go b/src/internal/syscall/unix/fallocate_freebsd_64bit.go new file mode 100644 index 0000000..a9d5228 --- /dev/null +++ b/src/internal/syscall/unix/fallocate_freebsd_64bit.go @@ -0,0 +1,19 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build freebsd && (amd64 || arm64 || riscv64) + +package unix + +import "syscall" + +func PosixFallocate(fd int, off int64, size int64) error { + // If successful, posix_fallocate() returns zero. It returns an error on failure, without + // setting errno. See https://man.freebsd.org/cgi/man.cgi?query=posix_fallocate&sektion=2&n=1 + r1, _, _ := syscall.Syscall(posixFallocateTrap, uintptr(fd), uintptr(off), uintptr(size)) + if r1 != 0 { + return syscall.Errno(r1) + } + return nil +} diff --git a/src/internal/syscall/unix/fallocate_freebsd_arm.go b/src/internal/syscall/unix/fallocate_freebsd_arm.go new file mode 100644 index 0000000..1ded50f --- /dev/null +++ b/src/internal/syscall/unix/fallocate_freebsd_arm.go @@ -0,0 +1,22 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import "syscall" + +func PosixFallocate(fd int, off int64, size int64) error { + // If successful, posix_fallocate() returns zero. It returns an error on failure, without + // setting errno. See https://man.freebsd.org/cgi/man.cgi?query=posix_fallocate&sektion=2&n=1 + // + // The padding 0 argument is needed because the ARM calling convention requires that if an + // argument (off in this case) needs double-word alignment (8-byte), the NCRN (next core + // register number) is rounded up to the next even register number. + // See https://github.com/ARM-software/abi-aa/blob/2bcab1e3b22d55170c563c3c7940134089176746/aapcs32/aapcs32.rst#parameter-passing + r1, _, _ := syscall.Syscall6(posixFallocateTrap, uintptr(fd), 0, uintptr(off), uintptr(off>>32), uintptr(size), uintptr(size>>32)) + if r1 != 0 { + return syscall.Errno(r1) + } + return nil +} diff --git a/src/internal/syscall/unix/fcntl_js.go b/src/internal/syscall/unix/fcntl_js.go new file mode 100644 index 0000000..bdfb8e0 --- /dev/null +++ b/src/internal/syscall/unix/fcntl_js.go @@ -0,0 +1,13 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build js && wasm + +package unix + +import "syscall" + +func Fcntl(fd int, cmd int, arg int) (int, error) { + return 0, syscall.ENOSYS +} diff --git a/src/internal/syscall/unix/fcntl_unix.go b/src/internal/syscall/unix/fcntl_unix.go new file mode 100644 index 0000000..6f9e124 --- /dev/null +++ b/src/internal/syscall/unix/fcntl_unix.go @@ -0,0 +1,25 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package unix + +import ( + "syscall" + _ "unsafe" // for go:linkname +) + +// Implemented in the runtime package. +// +//go:linkname fcntl runtime.fcntl +func fcntl(fd int32, cmd int32, arg int32) (int32, int32) + +func Fcntl(fd int, cmd int, arg int) (int, error) { + val, errno := fcntl(int32(fd), int32(cmd), int32(arg)) + if val == -1 { + return int(val), syscall.Errno(errno) + } + return int(val), nil +} diff --git a/src/internal/syscall/unix/fcntl_wasip1.go b/src/internal/syscall/unix/fcntl_wasip1.go new file mode 100644 index 0000000..e70cd74 --- /dev/null +++ b/src/internal/syscall/unix/fcntl_wasip1.go @@ -0,0 +1,17 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build wasip1 + +package unix + +import "syscall" + +func Fcntl(fd int, cmd int, arg int) (int, error) { + if cmd == syscall.F_GETFL { + flags, err := fd_fdstat_get_flags(fd) + return int(flags), err + } + return 0, syscall.ENOSYS +} diff --git a/src/internal/syscall/unix/getentropy_darwin.go b/src/internal/syscall/unix/getentropy_darwin.go new file mode 100644 index 0000000..834099f --- /dev/null +++ b/src/internal/syscall/unix/getentropy_darwin.go @@ -0,0 +1,28 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build darwin && !ios + +package unix + +import ( + "internal/abi" + "unsafe" +) + +//go:cgo_import_dynamic libc_getentropy getentropy "/usr/lib/libSystem.B.dylib" + +func libc_getentropy_trampoline() + +// GetEntropy calls the macOS getentropy system call. +func GetEntropy(p []byte) error { + _, _, errno := syscall_syscall(abi.FuncPCABI0(libc_getentropy_trampoline), + uintptr(unsafe.Pointer(&p[0])), + uintptr(len(p)), + 0) + if errno != 0 { + return errno + } + return nil +} diff --git a/src/internal/syscall/unix/getentropy_openbsd.go b/src/internal/syscall/unix/getentropy_openbsd.go new file mode 100644 index 0000000..ad0914d --- /dev/null +++ b/src/internal/syscall/unix/getentropy_openbsd.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build openbsd && !mips64 + +package unix + +import _ "unsafe" // for linkname + +// GetEntropy calls the OpenBSD getentropy system call. +func GetEntropy(p []byte) error { + return getentropy(p) +} + +//go:linkname getentropy syscall.getentropy +func getentropy(p []byte) error diff --git a/src/internal/syscall/unix/getentropy_openbsd_mips64.go b/src/internal/syscall/unix/getentropy_openbsd_mips64.go new file mode 100644 index 0000000..d5caa80 --- /dev/null +++ b/src/internal/syscall/unix/getentropy_openbsd_mips64.go @@ -0,0 +1,25 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "syscall" + "unsafe" +) + +// getentropy(2)'s syscall number, from /usr/src/sys/kern/syscalls.master +const entropyTrap uintptr = 7 + +// GetEntropy calls the OpenBSD getentropy system call. +func GetEntropy(p []byte) error { + _, _, errno := syscall.Syscall(entropyTrap, + uintptr(unsafe.Pointer(&p[0])), + uintptr(len(p)), + 0) + if errno != 0 { + return errno + } + return nil +} diff --git a/src/internal/syscall/unix/getrandom.go b/src/internal/syscall/unix/getrandom.go new file mode 100644 index 0000000..e83f0cd --- /dev/null +++ b/src/internal/syscall/unix/getrandom.go @@ -0,0 +1,39 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build dragonfly || freebsd || linux + +package unix + +import ( + "sync/atomic" + "syscall" + "unsafe" +) + +var getrandomUnsupported atomic.Bool + +// GetRandomFlag is a flag supported by the getrandom system call. +type GetRandomFlag uintptr + +// GetRandom calls the getrandom system call. +func GetRandom(p []byte, flags GetRandomFlag) (n int, err error) { + if len(p) == 0 { + return 0, nil + } + if getrandomUnsupported.Load() { + return 0, syscall.ENOSYS + } + r1, _, errno := syscall.Syscall(getrandomTrap, + uintptr(unsafe.Pointer(&p[0])), + uintptr(len(p)), + uintptr(flags)) + if errno != 0 { + if errno == syscall.ENOSYS { + getrandomUnsupported.Store(true) + } + return 0, errno + } + return int(r1), nil +} diff --git a/src/internal/syscall/unix/getrandom_dragonfly.go b/src/internal/syscall/unix/getrandom_dragonfly.go new file mode 100644 index 0000000..fbf78f9 --- /dev/null +++ b/src/internal/syscall/unix/getrandom_dragonfly.go @@ -0,0 +1,16 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +// DragonFlyBSD getrandom system call number. +const getrandomTrap uintptr = 550 + +const ( + // GRND_RANDOM is only set for portability purpose, no-op on DragonFlyBSD. + GRND_RANDOM GetRandomFlag = 0x0001 + + // GRND_NONBLOCK means return EAGAIN rather than blocking. + GRND_NONBLOCK GetRandomFlag = 0x0002 +) diff --git a/src/internal/syscall/unix/getrandom_freebsd.go b/src/internal/syscall/unix/getrandom_freebsd.go new file mode 100644 index 0000000..8c4f3df --- /dev/null +++ b/src/internal/syscall/unix/getrandom_freebsd.go @@ -0,0 +1,16 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +// FreeBSD getrandom system call number. +const getrandomTrap uintptr = 563 + +const ( + // GRND_NONBLOCK means return EAGAIN rather than blocking. + GRND_NONBLOCK GetRandomFlag = 0x0001 + + // GRND_RANDOM is only set for portability purpose, no-op on FreeBSD. + GRND_RANDOM GetRandomFlag = 0x0002 +) diff --git a/src/internal/syscall/unix/getrandom_linux.go b/src/internal/syscall/unix/getrandom_linux.go new file mode 100644 index 0000000..8ccd8d3 --- /dev/null +++ b/src/internal/syscall/unix/getrandom_linux.go @@ -0,0 +1,13 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + // GRND_NONBLOCK means return EAGAIN rather than blocking. + GRND_NONBLOCK GetRandomFlag = 0x0001 + + // GRND_RANDOM means use the /dev/random pool instead of /dev/urandom. + GRND_RANDOM GetRandomFlag = 0x0002 +) diff --git a/src/internal/syscall/unix/getrandom_netbsd.go b/src/internal/syscall/unix/getrandom_netbsd.go new file mode 100644 index 0000000..c83e3b2 --- /dev/null +++ b/src/internal/syscall/unix/getrandom_netbsd.go @@ -0,0 +1,56 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "sync" + "sync/atomic" + "syscall" + "unsafe" +) + +// NetBSD getrandom system call number. +const getrandomTrap uintptr = 91 + +var getrandomUnsupported atomic.Bool + +// GetRandomFlag is a flag supported by the getrandom system call. +type GetRandomFlag uintptr + +// GetRandom calls the getrandom system call. +func GetRandom(p []byte, flags GetRandomFlag) (n int, err error) { + if len(p) == 0 { + return 0, nil + } + if getrandomUnsupported.Load() { + return 0, syscall.ENOSYS + } + // getrandom(2) was added in NetBSD 10.0 + if getOSRevision() < 1000000000 { + getrandomUnsupported.Store(true) + return 0, syscall.ENOSYS + } + r1, _, errno := syscall.Syscall(getrandomTrap, + uintptr(unsafe.Pointer(&p[0])), + uintptr(len(p)), + uintptr(flags)) + if errno != 0 { + if errno == syscall.ENOSYS { + getrandomUnsupported.Store(true) + } + return 0, errno + } + return int(r1), nil +} + +var ( + osrevisionOnce sync.Once + osrevision uint32 +) + +func getOSRevision() uint32 { + osrevisionOnce.Do(func() { osrevision, _ = syscall.SysctlUint32("kern.osrevision") }) + return osrevision +} diff --git a/src/internal/syscall/unix/getrandom_solaris.go b/src/internal/syscall/unix/getrandom_solaris.go new file mode 100644 index 0000000..cf4f35a --- /dev/null +++ b/src/internal/syscall/unix/getrandom_solaris.go @@ -0,0 +1,53 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "sync/atomic" + "syscall" + "unsafe" +) + +//go:cgo_import_dynamic libc_getrandom getrandom "libc.so" + +//go:linkname procGetrandom libc_getrandom + +var procGetrandom uintptr + +var getrandomUnsupported atomic.Bool + +// GetRandomFlag is a flag supported by the getrandom system call. +type GetRandomFlag uintptr + +const ( + // GRND_NONBLOCK means return EAGAIN rather than blocking. + GRND_NONBLOCK GetRandomFlag = 0x0001 + + // GRND_RANDOM means use the /dev/random pool instead of /dev/urandom. + GRND_RANDOM GetRandomFlag = 0x0002 +) + +// GetRandom calls the getrandom system call. +func GetRandom(p []byte, flags GetRandomFlag) (n int, err error) { + if len(p) == 0 { + return 0, nil + } + if getrandomUnsupported.Load() { + return 0, syscall.ENOSYS + } + r1, _, errno := syscall6(uintptr(unsafe.Pointer(&procGetrandom)), + 3, + uintptr(unsafe.Pointer(&p[0])), + uintptr(len(p)), + uintptr(flags), + 0, 0, 0) + if errno != 0 { + if errno == syscall.ENOSYS { + getrandomUnsupported.Store(true) + } + return 0, errno + } + return int(r1), nil +} diff --git a/src/internal/syscall/unix/ioctl_aix.go b/src/internal/syscall/unix/ioctl_aix.go new file mode 100644 index 0000000..d361533 --- /dev/null +++ b/src/internal/syscall/unix/ioctl_aix.go @@ -0,0 +1,25 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "syscall" + "unsafe" +) + +//go:cgo_import_dynamic libc_ioctl ioctl "libc.a/shr_64.o" +//go:linkname libc_ioctl libc_ioctl +var libc_ioctl uintptr + +// Implemented in syscall/syscall_aix.go. +func syscall6(trap, nargs, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2 uintptr, err syscall.Errno) + +func Ioctl(fd int, cmd int, args unsafe.Pointer) (err error) { + _, _, e1 := syscall6(uintptr(unsafe.Pointer(&libc_ioctl)), 3, uintptr(fd), uintptr(cmd), uintptr(args), 0, 0, 0) + if e1 != 0 { + err = e1 + } + return +} diff --git a/src/internal/syscall/unix/kernel_version_linux.go b/src/internal/syscall/unix/kernel_version_linux.go new file mode 100644 index 0000000..71e8aa4 --- /dev/null +++ b/src/internal/syscall/unix/kernel_version_linux.go @@ -0,0 +1,42 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "syscall" +) + +// KernelVersion returns major and minor kernel version numbers, parsed from +// the syscall.Uname's Release field, or 0, 0 if the version can't be obtained +// or parsed. +// +// Currently only implemented for Linux. +func KernelVersion() (major, minor int) { + var uname syscall.Utsname + if err := syscall.Uname(&uname); err != nil { + return + } + + var ( + values [2]int + value, vi int + ) + for _, c := range uname.Release { + if '0' <= c && c <= '9' { + value = (value * 10) + int(c-'0') + } else { + // Note that we're assuming N.N.N here. + // If we see anything else, we are likely to mis-parse it. + values[vi] = value + vi++ + if vi >= len(values) { + break + } + value = 0 + } + } + + return values[0], values[1] +} diff --git a/src/internal/syscall/unix/kernel_version_other.go b/src/internal/syscall/unix/kernel_version_other.go new file mode 100644 index 0000000..00af9f2 --- /dev/null +++ b/src/internal/syscall/unix/kernel_version_other.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !linux + +package unix + +func KernelVersion() (major int, minor int) { + return 0, 0 +} diff --git a/src/internal/syscall/unix/net.go b/src/internal/syscall/unix/net.go new file mode 100644 index 0000000..5618d40 --- /dev/null +++ b/src/internal/syscall/unix/net.go @@ -0,0 +1,44 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package unix + +import ( + "syscall" + _ "unsafe" +) + +//go:linkname RecvfromInet4 syscall.recvfromInet4 +//go:noescape +func RecvfromInet4(fd int, p []byte, flags int, from *syscall.SockaddrInet4) (int, error) + +//go:linkname RecvfromInet6 syscall.recvfromInet6 +//go:noescape +func RecvfromInet6(fd int, p []byte, flags int, from *syscall.SockaddrInet6) (n int, err error) + +//go:linkname SendtoInet4 syscall.sendtoInet4 +//go:noescape +func SendtoInet4(fd int, p []byte, flags int, to *syscall.SockaddrInet4) (err error) + +//go:linkname SendtoInet6 syscall.sendtoInet6 +//go:noescape +func SendtoInet6(fd int, p []byte, flags int, to *syscall.SockaddrInet6) (err error) + +//go:linkname SendmsgNInet4 syscall.sendmsgNInet4 +//go:noescape +func SendmsgNInet4(fd int, p, oob []byte, to *syscall.SockaddrInet4, flags int) (n int, err error) + +//go:linkname SendmsgNInet6 syscall.sendmsgNInet6 +//go:noescape +func SendmsgNInet6(fd int, p, oob []byte, to *syscall.SockaddrInet6, flags int) (n int, err error) + +//go:linkname RecvmsgInet4 syscall.recvmsgInet4 +//go:noescape +func RecvmsgInet4(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet4) (n, oobn int, recvflags int, err error) + +//go:linkname RecvmsgInet6 syscall.recvmsgInet6 +//go:noescape +func RecvmsgInet6(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet6) (n, oobn int, recvflags int, err error) diff --git a/src/internal/syscall/unix/net_darwin.go b/src/internal/syscall/unix/net_darwin.go new file mode 100644 index 0000000..5601b49 --- /dev/null +++ b/src/internal/syscall/unix/net_darwin.go @@ -0,0 +1,162 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "internal/abi" + "syscall" + "unsafe" +) + +const ( + AI_CANONNAME = 0x2 + AI_ALL = 0x100 + AI_V4MAPPED = 0x800 + AI_MASK = 0x1407 + + EAI_AGAIN = 2 + EAI_NODATA = 7 + EAI_NONAME = 8 + EAI_SYSTEM = 11 + EAI_OVERFLOW = 14 + + NI_NAMEREQD = 4 +) + +type Addrinfo struct { + Flags int32 + Family int32 + Socktype int32 + Protocol int32 + Addrlen uint32 + Canonname *byte + Addr *syscall.RawSockaddr + Next *Addrinfo +} + +//go:cgo_ldflag "-lresolv" + +//go:cgo_import_dynamic libc_getaddrinfo getaddrinfo "/usr/lib/libSystem.B.dylib" +func libc_getaddrinfo_trampoline() + +func Getaddrinfo(hostname, servname *byte, hints *Addrinfo, res **Addrinfo) (int, error) { + gerrno, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_getaddrinfo_trampoline), + uintptr(unsafe.Pointer(hostname)), + uintptr(unsafe.Pointer(servname)), + uintptr(unsafe.Pointer(hints)), + uintptr(unsafe.Pointer(res)), + 0, + 0) + var err error + if errno != 0 { + err = errno + } + return int(gerrno), err +} + +//go:cgo_import_dynamic libc_freeaddrinfo freeaddrinfo "/usr/lib/libSystem.B.dylib" +func libc_freeaddrinfo_trampoline() + +func Freeaddrinfo(ai *Addrinfo) { + syscall_syscall6(abi.FuncPCABI0(libc_freeaddrinfo_trampoline), + uintptr(unsafe.Pointer(ai)), + 0, 0, 0, 0, 0) +} + +//go:cgo_import_dynamic libc_getnameinfo getnameinfo "/usr/lib/libSystem.B.dylib" +func libc_getnameinfo_trampoline() + +func Getnameinfo(sa *syscall.RawSockaddr, salen int, host *byte, hostlen int, serv *byte, servlen int, flags int) (int, error) { + gerrno, _, errno := syscall_syscall9(abi.FuncPCABI0(libc_getnameinfo_trampoline), + uintptr(unsafe.Pointer(sa)), + uintptr(salen), + uintptr(unsafe.Pointer(host)), + uintptr(hostlen), + uintptr(unsafe.Pointer(serv)), + uintptr(servlen), + uintptr(flags), + 0, + 0) + var err error + if errno != 0 { + err = errno + } + return int(gerrno), err +} + +//go:cgo_import_dynamic libc_gai_strerror gai_strerror "/usr/lib/libSystem.B.dylib" +func libc_gai_strerror_trampoline() + +func GaiStrerror(ecode int) string { + r1, _, _ := syscall_syscall(abi.FuncPCABI0(libc_gai_strerror_trampoline), + uintptr(ecode), + 0, 0) + return GoString((*byte)(unsafe.Pointer(r1))) +} + +// Implemented in the runtime package. +func gostring(*byte) string + +func GoString(p *byte) string { + return gostring(p) +} + +//go:linkname syscall_syscall syscall.syscall +func syscall_syscall(fn, a1, a2, a3 uintptr) (r1, r2 uintptr, err syscall.Errno) + +//go:linkname syscall_syscallPtr syscall.syscallPtr +func syscall_syscallPtr(fn, a1, a2, a3 uintptr) (r1, r2 uintptr, err syscall.Errno) + +//go:linkname syscall_syscall6 syscall.syscall6 +func syscall_syscall6(fn, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2 uintptr, err syscall.Errno) + +//go:linkname syscall_syscall6X syscall.syscall6X +func syscall_syscall6X(fn, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2 uintptr, err syscall.Errno) + +//go:linkname syscall_syscall9 syscall.syscall9 +func syscall_syscall9(fn, a1, a2, a3, a4, a5, a6, a7, a8, a9 uintptr) (r1, r2 uintptr, err syscall.Errno) + +type ResState struct { + unexported [69]uintptr +} + +//go:cgo_import_dynamic libresolv_res_9_ninit res_9_ninit "/usr/lib/libresolv.9.dylib" +func libresolv_res_9_ninit_trampoline() + +func ResNinit(state *ResState) error { + _, _, errno := syscall_syscall(abi.FuncPCABI0(libresolv_res_9_ninit_trampoline), + uintptr(unsafe.Pointer(state)), + 0, 0) + if errno != 0 { + return errno + } + return nil +} + +//go:cgo_import_dynamic libresolv_res_9_nclose res_9_nclose "/usr/lib/libresolv.9.dylib" +func libresolv_res_9_nclose_trampoline() + +func ResNclose(state *ResState) { + syscall_syscall(abi.FuncPCABI0(libresolv_res_9_nclose_trampoline), + uintptr(unsafe.Pointer(state)), + 0, 0) +} + +//go:cgo_import_dynamic libresolv_res_9_nsearch res_9_nsearch "/usr/lib/libresolv.9.dylib" +func libresolv_res_9_nsearch_trampoline() + +func ResNsearch(state *ResState, dname *byte, class, typ int, ans *byte, anslen int) (int, error) { + r1, _, errno := syscall_syscall6(abi.FuncPCABI0(libresolv_res_9_nsearch_trampoline), + uintptr(unsafe.Pointer(state)), + uintptr(unsafe.Pointer(dname)), + uintptr(class), + uintptr(typ), + uintptr(unsafe.Pointer(ans)), + uintptr(anslen)) + if errno != 0 { + return 0, errno + } + return int(int32(r1)), nil +} diff --git a/src/internal/syscall/unix/net_js.go b/src/internal/syscall/unix/net_js.go new file mode 100644 index 0000000..622fc8e --- /dev/null +++ b/src/internal/syscall/unix/net_js.go @@ -0,0 +1,44 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build js + +package unix + +import ( + "syscall" + _ "unsafe" +) + +func RecvfromInet4(fd int, p []byte, flags int, from *syscall.SockaddrInet4) (int, error) { + return 0, syscall.ENOSYS +} + +func RecvfromInet6(fd int, p []byte, flags int, from *syscall.SockaddrInet6) (n int, err error) { + return 0, syscall.ENOSYS +} + +func SendtoInet4(fd int, p []byte, flags int, to *syscall.SockaddrInet4) (err error) { + return syscall.ENOSYS +} + +func SendtoInet6(fd int, p []byte, flags int, to *syscall.SockaddrInet6) (err error) { + return syscall.ENOSYS +} + +func SendmsgNInet4(fd int, p, oob []byte, to *syscall.SockaddrInet4, flags int) (n int, err error) { + return 0, syscall.ENOSYS +} + +func SendmsgNInet6(fd int, p, oob []byte, to *syscall.SockaddrInet6, flags int) (n int, err error) { + return 0, syscall.ENOSYS +} + +func RecvmsgInet4(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet4) (n, oobn int, recvflags int, err error) { + return 0, 0, 0, syscall.ENOSYS +} + +func RecvmsgInet6(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet6) (n, oobn int, recvflags int, err error) { + return 0, 0, 0, syscall.ENOSYS +} diff --git a/src/internal/syscall/unix/net_wasip1.go b/src/internal/syscall/unix/net_wasip1.go new file mode 100644 index 0000000..8a60e8f --- /dev/null +++ b/src/internal/syscall/unix/net_wasip1.go @@ -0,0 +1,44 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build wasip1 + +package unix + +import ( + "syscall" + _ "unsafe" +) + +func RecvfromInet4(fd int, p []byte, flags int, from *syscall.SockaddrInet4) (int, error) { + return 0, syscall.ENOSYS +} + +func RecvfromInet6(fd int, p []byte, flags int, from *syscall.SockaddrInet6) (n int, err error) { + return 0, syscall.ENOSYS +} + +func SendtoInet4(fd int, p []byte, flags int, to *syscall.SockaddrInet4) (err error) { + return syscall.ENOSYS +} + +func SendtoInet6(fd int, p []byte, flags int, to *syscall.SockaddrInet6) (err error) { + return syscall.ENOSYS +} + +func SendmsgNInet4(fd int, p, oob []byte, to *syscall.SockaddrInet4, flags int) (n int, err error) { + return 0, syscall.ENOSYS +} + +func SendmsgNInet6(fd int, p, oob []byte, to *syscall.SockaddrInet6, flags int) (n int, err error) { + return 0, syscall.ENOSYS +} + +func RecvmsgInet4(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet4) (n, oobn int, recvflags int, err error) { + return 0, 0, 0, syscall.ENOSYS +} + +func RecvmsgInet6(fd int, p, oob []byte, flags int, from *syscall.SockaddrInet6) (n, oobn int, recvflags int, err error) { + return 0, 0, 0, syscall.ENOSYS +} diff --git a/src/internal/syscall/unix/nonblocking_js.go b/src/internal/syscall/unix/nonblocking_js.go new file mode 100644 index 0000000..cfe78c5 --- /dev/null +++ b/src/internal/syscall/unix/nonblocking_js.go @@ -0,0 +1,15 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build js && wasm + +package unix + +func IsNonblock(fd int) (nonblocking bool, err error) { + return false, nil +} + +func HasNonblockFlag(flag int) bool { + return false +} diff --git a/src/internal/syscall/unix/nonblocking_unix.go b/src/internal/syscall/unix/nonblocking_unix.go new file mode 100644 index 0000000..fc0bc27 --- /dev/null +++ b/src/internal/syscall/unix/nonblocking_unix.go @@ -0,0 +1,21 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package unix + +import "syscall" + +func IsNonblock(fd int) (nonblocking bool, err error) { + flag, e1 := Fcntl(fd, syscall.F_GETFL, 0) + if e1 != nil { + return false, e1 + } + return flag&syscall.O_NONBLOCK != 0, nil +} + +func HasNonblockFlag(flag int) bool { + return flag&syscall.O_NONBLOCK != 0 +} diff --git a/src/internal/syscall/unix/nonblocking_wasip1.go b/src/internal/syscall/unix/nonblocking_wasip1.go new file mode 100644 index 0000000..5b2b53b --- /dev/null +++ b/src/internal/syscall/unix/nonblocking_wasip1.go @@ -0,0 +1,31 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build wasip1 + +package unix + +import ( + "syscall" + _ "unsafe" // for go:linkname +) + +func IsNonblock(fd int) (nonblocking bool, err error) { + flags, e1 := fd_fdstat_get_flags(fd) + if e1 != nil { + return false, e1 + } + return flags&syscall.FDFLAG_NONBLOCK != 0, nil +} + +func HasNonblockFlag(flag int) bool { + return flag&syscall.FDFLAG_NONBLOCK != 0 +} + +// This helper is implemented in the syscall package. It means we don't have +// to redefine the fd_fdstat_get host import or the fdstat struct it +// populates. +// +//go:linkname fd_fdstat_get_flags syscall.fd_fdstat_get_flags +func fd_fdstat_get_flags(fd int) (uint32, error) diff --git a/src/internal/syscall/unix/pty_darwin.go b/src/internal/syscall/unix/pty_darwin.go new file mode 100644 index 0000000..b43321a --- /dev/null +++ b/src/internal/syscall/unix/pty_darwin.go @@ -0,0 +1,65 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "internal/abi" + "unsafe" +) + +//go:cgo_import_dynamic libc_grantpt grantpt "/usr/lib/libSystem.B.dylib" +func libc_grantpt_trampoline() + +func Grantpt(fd int) error { + _, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_grantpt_trampoline), uintptr(fd), 0, 0, 0, 0, 0) + if errno != 0 { + return errno + } + return nil +} + +//go:cgo_import_dynamic libc_unlockpt unlockpt "/usr/lib/libSystem.B.dylib" +func libc_unlockpt_trampoline() + +func Unlockpt(fd int) error { + _, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_unlockpt_trampoline), uintptr(fd), 0, 0, 0, 0, 0) + if errno != 0 { + return errno + } + return nil +} + +//go:cgo_import_dynamic libc_ptsname_r ptsname_r "/usr/lib/libSystem.B.dylib" +func libc_ptsname_r_trampoline() + +func Ptsname(fd int) (string, error) { + buf := make([]byte, 256) + _, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_ptsname_r_trampoline), + uintptr(fd), + uintptr(unsafe.Pointer(&buf[0])), + uintptr(len(buf)-1), + 0, 0, 0) + if errno != 0 { + return "", errno + } + for i, c := range buf { + if c == 0 { + buf = buf[:i] + break + } + } + return string(buf), nil +} + +//go:cgo_import_dynamic libc_posix_openpt posix_openpt "/usr/lib/libSystem.B.dylib" +func libc_posix_openpt_trampoline() + +func PosixOpenpt(flag int) (fd int, err error) { + ufd, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_posix_openpt_trampoline), uintptr(flag), 0, 0, 0, 0, 0) + if errno != 0 { + return -1, errno + } + return int(ufd), nil +} diff --git a/src/internal/syscall/unix/sysnum_linux_386.go b/src/internal/syscall/unix/sysnum_linux_386.go new file mode 100644 index 0000000..2bda08c --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_386.go @@ -0,0 +1,10 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + getrandomTrap uintptr = 355 + copyFileRangeTrap uintptr = 377 +) diff --git a/src/internal/syscall/unix/sysnum_linux_amd64.go b/src/internal/syscall/unix/sysnum_linux_amd64.go new file mode 100644 index 0000000..ae5239e --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_amd64.go @@ -0,0 +1,10 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + getrandomTrap uintptr = 318 + copyFileRangeTrap uintptr = 326 +) diff --git a/src/internal/syscall/unix/sysnum_linux_arm.go b/src/internal/syscall/unix/sysnum_linux_arm.go new file mode 100644 index 0000000..acaec05 --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_arm.go @@ -0,0 +1,10 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + getrandomTrap uintptr = 384 + copyFileRangeTrap uintptr = 391 +) diff --git a/src/internal/syscall/unix/sysnum_linux_generic.go b/src/internal/syscall/unix/sysnum_linux_generic.go new file mode 100644 index 0000000..8c132c6 --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_generic.go @@ -0,0 +1,16 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build linux && (arm64 || loong64 || riscv64) + +package unix + +// This file is named "generic" because at a certain point Linux started +// standardizing on system call numbers across architectures. So far this +// means only arm64 loong64 and riscv64 use the standard numbers. + +const ( + getrandomTrap uintptr = 278 + copyFileRangeTrap uintptr = 285 +) diff --git a/src/internal/syscall/unix/sysnum_linux_mips64x.go b/src/internal/syscall/unix/sysnum_linux_mips64x.go new file mode 100644 index 0000000..bca526d --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_mips64x.go @@ -0,0 +1,12 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips64 || mips64le + +package unix + +const ( + getrandomTrap uintptr = 5313 + copyFileRangeTrap uintptr = 5320 +) diff --git a/src/internal/syscall/unix/sysnum_linux_mipsx.go b/src/internal/syscall/unix/sysnum_linux_mipsx.go new file mode 100644 index 0000000..c86195e --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_mipsx.go @@ -0,0 +1,12 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build mips || mipsle + +package unix + +const ( + getrandomTrap uintptr = 4353 + copyFileRangeTrap uintptr = 4360 +) diff --git a/src/internal/syscall/unix/sysnum_linux_ppc64x.go b/src/internal/syscall/unix/sysnum_linux_ppc64x.go new file mode 100644 index 0000000..a4dcf2b --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_ppc64x.go @@ -0,0 +1,12 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build ppc64 || ppc64le + +package unix + +const ( + getrandomTrap uintptr = 359 + copyFileRangeTrap uintptr = 379 +) diff --git a/src/internal/syscall/unix/sysnum_linux_s390x.go b/src/internal/syscall/unix/sysnum_linux_s390x.go new file mode 100644 index 0000000..bf2c01e --- /dev/null +++ b/src/internal/syscall/unix/sysnum_linux_s390x.go @@ -0,0 +1,10 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +const ( + getrandomTrap uintptr = 349 + copyFileRangeTrap uintptr = 375 +) diff --git a/src/internal/syscall/unix/user_darwin.go b/src/internal/syscall/unix/user_darwin.go new file mode 100644 index 0000000..d05acda --- /dev/null +++ b/src/internal/syscall/unix/user_darwin.go @@ -0,0 +1,121 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unix + +import ( + "internal/abi" + "syscall" + "unsafe" +) + +//go:cgo_import_dynamic libc_getgrouplist getgrouplist "/usr/lib/libSystem.B.dylib" +func libc_getgrouplist_trampoline() + +func Getgrouplist(name *byte, gid uint32, gids *uint32, n *int32) error { + _, _, errno := syscall_syscall6(abi.FuncPCABI0(libc_getgrouplist_trampoline), + uintptr(unsafe.Pointer(name)), uintptr(gid), uintptr(unsafe.Pointer(gids)), + uintptr(unsafe.Pointer(n)), 0, 0) + if errno != 0 { + return errno + } + return nil +} + +const ( + SC_GETGR_R_SIZE_MAX = 0x46 + SC_GETPW_R_SIZE_MAX = 0x47 +) + +type Passwd struct { + Name *byte + Passwd *byte + Uid uint32 // uid_t + Gid uint32 // gid_t + Change int64 // time_t + Class *byte + Gecos *byte + Dir *byte + Shell *byte + Expire int64 // time_t +} + +type Group struct { + Name *byte + Passwd *byte + Gid uint32 // gid_t + Mem **byte +} + +//go:cgo_import_dynamic libc_getpwnam_r getpwnam_r "/usr/lib/libSystem.B.dylib" +func libc_getpwnam_r_trampoline() + +func Getpwnam(name *byte, pwd *Passwd, buf *byte, size uintptr, result **Passwd) syscall.Errno { + // Note: Returns an errno as its actual result, not in global errno. + errno, _, _ := syscall_syscall6(abi.FuncPCABI0(libc_getpwnam_r_trampoline), + uintptr(unsafe.Pointer(name)), + uintptr(unsafe.Pointer(pwd)), + uintptr(unsafe.Pointer(buf)), + size, + uintptr(unsafe.Pointer(result)), + 0) + return syscall.Errno(errno) +} + +//go:cgo_import_dynamic libc_getpwuid_r getpwuid_r "/usr/lib/libSystem.B.dylib" +func libc_getpwuid_r_trampoline() + +func Getpwuid(uid uint32, pwd *Passwd, buf *byte, size uintptr, result **Passwd) syscall.Errno { + // Note: Returns an errno as its actual result, not in global errno. + errno, _, _ := syscall_syscall6(abi.FuncPCABI0(libc_getpwuid_r_trampoline), + uintptr(uid), + uintptr(unsafe.Pointer(pwd)), + uintptr(unsafe.Pointer(buf)), + size, + uintptr(unsafe.Pointer(result)), + 0) + return syscall.Errno(errno) +} + +//go:cgo_import_dynamic libc_getgrnam_r getgrnam_r "/usr/lib/libSystem.B.dylib" +func libc_getgrnam_r_trampoline() + +func Getgrnam(name *byte, grp *Group, buf *byte, size uintptr, result **Group) syscall.Errno { + // Note: Returns an errno as its actual result, not in global errno. + errno, _, _ := syscall_syscall6(abi.FuncPCABI0(libc_getgrnam_r_trampoline), + uintptr(unsafe.Pointer(name)), + uintptr(unsafe.Pointer(grp)), + uintptr(unsafe.Pointer(buf)), + size, + uintptr(unsafe.Pointer(result)), + 0) + return syscall.Errno(errno) +} + +//go:cgo_import_dynamic libc_getgrgid_r getgrgid_r "/usr/lib/libSystem.B.dylib" +func libc_getgrgid_r_trampoline() + +func Getgrgid(gid uint32, grp *Group, buf *byte, size uintptr, result **Group) syscall.Errno { + // Note: Returns an errno as its actual result, not in global errno. + errno, _, _ := syscall_syscall6(abi.FuncPCABI0(libc_getgrgid_r_trampoline), + uintptr(gid), + uintptr(unsafe.Pointer(grp)), + uintptr(unsafe.Pointer(buf)), + size, + uintptr(unsafe.Pointer(result)), + 0) + return syscall.Errno(errno) +} + +//go:cgo_import_dynamic libc_sysconf sysconf "/usr/lib/libSystem.B.dylib" +func libc_sysconf_trampoline() + +func Sysconf(key int32) int64 { + val, _, errno := syscall_syscall6X(abi.FuncPCABI0(libc_sysconf_trampoline), + uintptr(key), 0, 0, 0, 0, 0) + if errno != 0 { + return -1 + } + return int64(val) +} diff --git a/src/internal/syscall/windows/exec_windows_test.go b/src/internal/syscall/windows/exec_windows_test.go new file mode 100644 index 0000000..3311da5 --- /dev/null +++ b/src/internal/syscall/windows/exec_windows_test.go @@ -0,0 +1,139 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package windows_test + +import ( + "fmt" + "internal/syscall/windows" + "os" + "os/exec" + "syscall" + "testing" + "unsafe" +) + +func TestRunAtLowIntegrity(t *testing.T) { + if os.Getenv("GO_WANT_HELPER_PROCESS") == "1" { + wil, err := getProcessIntegrityLevel() + if err != nil { + fmt.Fprintf(os.Stderr, "error: %s\n", err.Error()) + os.Exit(9) + return + } + fmt.Printf("%s", wil) + os.Exit(0) + return + } + + cmd := exec.Command(os.Args[0], "-test.run=TestRunAtLowIntegrity", "--") + cmd.Env = []string{"GO_WANT_HELPER_PROCESS=1"} + + token, err := getIntegrityLevelToken(sidWilLow) + if err != nil { + t.Fatal(err) + } + defer token.Close() + + cmd.SysProcAttr = &syscall.SysProcAttr{ + Token: token, + } + + out, err := cmd.CombinedOutput() + if err != nil { + t.Fatal(err) + } + + if string(out) != sidWilLow { + t.Fatalf("Child process did not run as low integrity level: %s", string(out)) + } +} + +const ( + sidWilLow = `S-1-16-4096` +) + +func getProcessIntegrityLevel() (string, error) { + procToken, err := syscall.OpenCurrentProcessToken() + if err != nil { + return "", err + } + defer procToken.Close() + + p, err := tokenGetInfo(procToken, syscall.TokenIntegrityLevel, 64) + if err != nil { + return "", err + } + + tml := (*windows.TOKEN_MANDATORY_LABEL)(p) + + sid := (*syscall.SID)(unsafe.Pointer(tml.Label.Sid)) + + return sid.String() +} + +func tokenGetInfo(t syscall.Token, class uint32, initSize int) (unsafe.Pointer, error) { + n := uint32(initSize) + for { + b := make([]byte, n) + e := syscall.GetTokenInformation(t, class, &b[0], uint32(len(b)), &n) + if e == nil { + return unsafe.Pointer(&b[0]), nil + } + if e != syscall.ERROR_INSUFFICIENT_BUFFER { + return nil, e + } + if n <= uint32(len(b)) { + return nil, e + } + } +} + +func getIntegrityLevelToken(wns string) (syscall.Token, error) { + var procToken, token syscall.Token + + proc, err := syscall.GetCurrentProcess() + if err != nil { + return 0, err + } + defer syscall.CloseHandle(proc) + + err = syscall.OpenProcessToken(proc, + syscall.TOKEN_DUPLICATE| + syscall.TOKEN_ADJUST_DEFAULT| + syscall.TOKEN_QUERY| + syscall.TOKEN_ASSIGN_PRIMARY, + &procToken) + if err != nil { + return 0, err + } + defer procToken.Close() + + sid, err := syscall.StringToSid(wns) + if err != nil { + return 0, err + } + + tml := &windows.TOKEN_MANDATORY_LABEL{} + tml.Label.Attributes = windows.SE_GROUP_INTEGRITY + tml.Label.Sid = sid + + err = windows.DuplicateTokenEx(procToken, 0, nil, windows.SecurityImpersonation, + windows.TokenPrimary, &token) + if err != nil { + return 0, err + } + + err = windows.SetTokenInformation(token, + syscall.TokenIntegrityLevel, + uintptr(unsafe.Pointer(tml)), + tml.Size()) + if err != nil { + token.Close() + return 0, err + } + return token, nil +} diff --git a/src/internal/syscall/windows/memory_windows.go b/src/internal/syscall/windows/memory_windows.go new file mode 100644 index 0000000..8fb34cf --- /dev/null +++ b/src/internal/syscall/windows/memory_windows.go @@ -0,0 +1,24 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +type MemoryBasicInformation struct { + // A pointer to the base address of the region of pages. + BaseAddress uintptr + // A pointer to the base address of a range of pages allocated by the VirtualAlloc function. + // The page pointed to by the BaseAddress member is contained within this allocation range. + AllocationBase uintptr + // The memory protection option when the region was initially allocated + AllocationProtect uint32 + PartitionId uint16 + // The size of the region beginning at the base address in which all pages have identical attributes, in bytes. + RegionSize uintptr + // The state of the pages in the region. + State uint32 + // The access protection of the pages in the region. + Protect uint32 + // The type of pages in the region. + Type uint32 +} diff --git a/src/internal/syscall/windows/mksyscall.go b/src/internal/syscall/windows/mksyscall.go new file mode 100644 index 0000000..81f08c6 --- /dev/null +++ b/src/internal/syscall/windows/mksyscall.go @@ -0,0 +1,9 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build generate + +package windows + +//go:generate go run ../../../syscall/mksyscall_windows.go -output zsyscall_windows.go syscall_windows.go security_windows.go psapi_windows.go symlink_windows.go diff --git a/src/internal/syscall/windows/net_windows.go b/src/internal/syscall/windows/net_windows.go new file mode 100644 index 0000000..42c600c --- /dev/null +++ b/src/internal/syscall/windows/net_windows.go @@ -0,0 +1,40 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +import ( + "sync" + "syscall" + _ "unsafe" +) + +//go:linkname WSASendtoInet4 syscall.wsaSendtoInet4 +//go:noescape +func WSASendtoInet4(s syscall.Handle, bufs *syscall.WSABuf, bufcnt uint32, sent *uint32, flags uint32, to *syscall.SockaddrInet4, overlapped *syscall.Overlapped, croutine *byte) (err error) + +//go:linkname WSASendtoInet6 syscall.wsaSendtoInet6 +//go:noescape +func WSASendtoInet6(s syscall.Handle, bufs *syscall.WSABuf, bufcnt uint32, sent *uint32, flags uint32, to *syscall.SockaddrInet6, overlapped *syscall.Overlapped, croutine *byte) (err error) + +const ( + SIO_TCP_INITIAL_RTO = syscall.IOC_IN | syscall.IOC_VENDOR | 17 + TCP_INITIAL_RTO_UNSPECIFIED_RTT = ^uint16(0) + TCP_INITIAL_RTO_NO_SYN_RETRANSMISSIONS = ^uint8(1) +) + +type TCP_INITIAL_RTO_PARAMETERS struct { + Rtt uint16 + MaxSynRetransmissions uint8 +} + +var Support_TCP_INITIAL_RTO_NO_SYN_RETRANSMISSIONS = sync.OnceValue(func() bool { + var maj, min, build uint32 + rtlGetNtVersionNumbers(&maj, &min, &build) + return maj >= 10 && build&0xffff >= 16299 +}) + +//go:linkname rtlGetNtVersionNumbers syscall.rtlGetNtVersionNumbers +//go:noescape +func rtlGetNtVersionNumbers(majorVersion *uint32, minorVersion *uint32, buildNumber *uint32) diff --git a/src/internal/syscall/windows/psapi_windows.go b/src/internal/syscall/windows/psapi_windows.go new file mode 100644 index 0000000..b138e65 --- /dev/null +++ b/src/internal/syscall/windows/psapi_windows.go @@ -0,0 +1,20 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +type PROCESS_MEMORY_COUNTERS struct { + CB uint32 + PageFaultCount uint32 + PeakWorkingSetSize uintptr + WorkingSetSize uintptr + QuotaPeakPagedPoolUsage uintptr + QuotaPagedPoolUsage uintptr + QuotaPeakNonPagedPoolUsage uintptr + QuotaNonPagedPoolUsage uintptr + PagefileUsage uintptr + PeakPagefileUsage uintptr +} + +//sys GetProcessMemoryInfo(handle syscall.Handle, memCounters *PROCESS_MEMORY_COUNTERS, cb uint32) (err error) = psapi.GetProcessMemoryInfo diff --git a/src/internal/syscall/windows/registry/export_test.go b/src/internal/syscall/windows/registry/export_test.go new file mode 100644 index 0000000..7f1ac70 --- /dev/null +++ b/src/internal/syscall/windows/registry/export_test.go @@ -0,0 +1,11 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package registry + +func (k Key) SetValue(name string, valtype uint32, data []byte) error { + return k.setValue(name, valtype, data) +} diff --git a/src/internal/syscall/windows/registry/key.go b/src/internal/syscall/windows/registry/key.go new file mode 100644 index 0000000..ce6397f --- /dev/null +++ b/src/internal/syscall/windows/registry/key.go @@ -0,0 +1,168 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +// Package registry provides access to the Windows registry. +// +// Here is a simple example, opening a registry key and reading a string value from it. +// +// k, err := registry.OpenKey(registry.LOCAL_MACHINE, `SOFTWARE\Microsoft\Windows NT\CurrentVersion`, registry.QUERY_VALUE) +// if err != nil { +// log.Fatal(err) +// } +// defer k.Close() +// +// s, _, err := k.GetStringValue("SystemRoot") +// if err != nil { +// log.Fatal(err) +// } +// fmt.Printf("Windows system root is %q\n", s) +// +// NOTE: This package is a copy of golang.org/x/sys/windows/registry +// with KeyInfo.ModTime removed to prevent dependency cycles. +package registry + +import ( + "runtime" + "syscall" +) + +const ( + // Registry key security and access rights. + // See https://msdn.microsoft.com/en-us/library/windows/desktop/ms724878.aspx + // for details. + ALL_ACCESS = 0xf003f + CREATE_LINK = 0x00020 + CREATE_SUB_KEY = 0x00004 + ENUMERATE_SUB_KEYS = 0x00008 + EXECUTE = 0x20019 + NOTIFY = 0x00010 + QUERY_VALUE = 0x00001 + READ = 0x20019 + SET_VALUE = 0x00002 + WOW64_32KEY = 0x00200 + WOW64_64KEY = 0x00100 + WRITE = 0x20006 +) + +// Key is a handle to an open Windows registry key. +// Keys can be obtained by calling OpenKey; there are +// also some predefined root keys such as CURRENT_USER. +// Keys can be used directly in the Windows API. +type Key syscall.Handle + +const ( + // Windows defines some predefined root keys that are always open. + // An application can use these keys as entry points to the registry. + // Normally these keys are used in OpenKey to open new keys, + // but they can also be used anywhere a Key is required. + CLASSES_ROOT = Key(syscall.HKEY_CLASSES_ROOT) + CURRENT_USER = Key(syscall.HKEY_CURRENT_USER) + LOCAL_MACHINE = Key(syscall.HKEY_LOCAL_MACHINE) + USERS = Key(syscall.HKEY_USERS) + CURRENT_CONFIG = Key(syscall.HKEY_CURRENT_CONFIG) +) + +// Close closes open key k. +func (k Key) Close() error { + return syscall.RegCloseKey(syscall.Handle(k)) +} + +// OpenKey opens a new key with path name relative to key k. +// It accepts any open key, including CURRENT_USER and others, +// and returns the new key and an error. +// The access parameter specifies desired access rights to the +// key to be opened. +func OpenKey(k Key, path string, access uint32) (Key, error) { + p, err := syscall.UTF16PtrFromString(path) + if err != nil { + return 0, err + } + var subkey syscall.Handle + err = syscall.RegOpenKeyEx(syscall.Handle(k), p, 0, access, &subkey) + if err != nil { + return 0, err + } + return Key(subkey), nil +} + +// ReadSubKeyNames returns the names of subkeys of key k. +func (k Key) ReadSubKeyNames() ([]string, error) { + // RegEnumKeyEx must be called repeatedly and to completion. + // During this time, this goroutine cannot migrate away from + // its current thread. See #49320. + runtime.LockOSThread() + defer runtime.UnlockOSThread() + + names := make([]string, 0) + // Registry key size limit is 255 bytes and described there: + // https://msdn.microsoft.com/library/windows/desktop/ms724872.aspx + buf := make([]uint16, 256) //plus extra room for terminating zero byte +loopItems: + for i := uint32(0); ; i++ { + l := uint32(len(buf)) + for { + err := syscall.RegEnumKeyEx(syscall.Handle(k), i, &buf[0], &l, nil, nil, nil, nil) + if err == nil { + break + } + if err == syscall.ERROR_MORE_DATA { + // Double buffer size and try again. + l = uint32(2 * len(buf)) + buf = make([]uint16, l) + continue + } + if err == _ERROR_NO_MORE_ITEMS { + break loopItems + } + return names, err + } + names = append(names, syscall.UTF16ToString(buf[:l])) + } + return names, nil +} + +// CreateKey creates a key named path under open key k. +// CreateKey returns the new key and a boolean flag that reports +// whether the key already existed. +// The access parameter specifies the access rights for the key +// to be created. +func CreateKey(k Key, path string, access uint32) (newk Key, openedExisting bool, err error) { + var h syscall.Handle + var d uint32 + err = regCreateKeyEx(syscall.Handle(k), syscall.StringToUTF16Ptr(path), + 0, nil, _REG_OPTION_NON_VOLATILE, access, nil, &h, &d) + if err != nil { + return 0, false, err + } + return Key(h), d == _REG_OPENED_EXISTING_KEY, nil +} + +// DeleteKey deletes the subkey path of key k and its values. +func DeleteKey(k Key, path string) error { + return regDeleteKey(syscall.Handle(k), syscall.StringToUTF16Ptr(path)) +} + +// A KeyInfo describes the statistics of a key. It is returned by Stat. +type KeyInfo struct { + SubKeyCount uint32 + MaxSubKeyLen uint32 // size of the key's subkey with the longest name, in Unicode characters, not including the terminating zero byte + ValueCount uint32 + MaxValueNameLen uint32 // size of the key's longest value name, in Unicode characters, not including the terminating zero byte + MaxValueLen uint32 // longest data component among the key's values, in bytes + lastWriteTime syscall.Filetime +} + +// Stat retrieves information about the open key k. +func (k Key) Stat() (*KeyInfo, error) { + var ki KeyInfo + err := syscall.RegQueryInfoKey(syscall.Handle(k), nil, nil, nil, + &ki.SubKeyCount, &ki.MaxSubKeyLen, nil, &ki.ValueCount, + &ki.MaxValueNameLen, &ki.MaxValueLen, nil, &ki.lastWriteTime) + if err != nil { + return nil, err + } + return &ki, nil +} diff --git a/src/internal/syscall/windows/registry/mksyscall.go b/src/internal/syscall/windows/registry/mksyscall.go new file mode 100644 index 0000000..0e0b421 --- /dev/null +++ b/src/internal/syscall/windows/registry/mksyscall.go @@ -0,0 +1,9 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build generate + +package registry + +//go:generate go run ../../../../syscall/mksyscall_windows.go -output zsyscall_windows.go syscall.go diff --git a/src/internal/syscall/windows/registry/registry_test.go b/src/internal/syscall/windows/registry/registry_test.go new file mode 100644 index 0000000..278b0b4 --- /dev/null +++ b/src/internal/syscall/windows/registry/registry_test.go @@ -0,0 +1,672 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package registry_test + +import ( + "bytes" + "crypto/rand" + "os" + "syscall" + "testing" + "unsafe" + + "internal/syscall/windows/registry" +) + +func randKeyName(prefix string) string { + const numbers = "0123456789" + buf := make([]byte, 10) + rand.Read(buf) + for i, b := range buf { + buf[i] = numbers[b%byte(len(numbers))] + } + return prefix + string(buf) +} + +func TestReadSubKeyNames(t *testing.T) { + k, err := registry.OpenKey(registry.CLASSES_ROOT, "TypeLib", registry.ENUMERATE_SUB_KEYS) + if err != nil { + t.Fatal(err) + } + defer k.Close() + + names, err := k.ReadSubKeyNames() + if err != nil { + t.Fatal(err) + } + var foundStdOle bool + for _, name := range names { + // Every PC has "stdole 2.0 OLE Automation" library installed. + if name == "{00020430-0000-0000-C000-000000000046}" { + foundStdOle = true + } + } + if !foundStdOle { + t.Fatal("could not find stdole 2.0 OLE Automation") + } +} + +func TestCreateOpenDeleteKey(t *testing.T) { + k, err := registry.OpenKey(registry.CURRENT_USER, "Software", registry.QUERY_VALUE) + if err != nil { + t.Fatal(err) + } + defer k.Close() + + testKName := randKeyName("TestCreateOpenDeleteKey_") + + testK, exist, err := registry.CreateKey(k, testKName, registry.CREATE_SUB_KEY) + if err != nil { + t.Fatal(err) + } + defer testK.Close() + + if exist { + t.Fatalf("key %q already exists", testKName) + } + + testKAgain, exist, err := registry.CreateKey(k, testKName, registry.CREATE_SUB_KEY) + if err != nil { + t.Fatal(err) + } + defer testKAgain.Close() + + if !exist { + t.Fatalf("key %q should already exist", testKName) + } + + testKOpened, err := registry.OpenKey(k, testKName, registry.ENUMERATE_SUB_KEYS) + if err != nil { + t.Fatal(err) + } + defer testKOpened.Close() + + err = registry.DeleteKey(k, testKName) + if err != nil { + t.Fatal(err) + } + + testKOpenedAgain, err := registry.OpenKey(k, testKName, registry.ENUMERATE_SUB_KEYS) + if err == nil { + defer testKOpenedAgain.Close() + t.Fatalf("key %q should already been deleted", testKName) + } + if err != registry.ErrNotExist { + t.Fatalf(`unexpected error ("not exist" expected): %v`, err) + } +} + +func equalStringSlice(a, b []string) bool { + if len(a) != len(b) { + return false + } + if a == nil { + return true + } + for i := range a { + if a[i] != b[i] { + return false + } + } + return true +} + +type ValueTest struct { + Type uint32 + Name string + Value any + WillFail bool +} + +var ValueTests = []ValueTest{ + {Type: registry.SZ, Name: "String1", Value: ""}, + {Type: registry.SZ, Name: "String2", Value: "\000", WillFail: true}, + {Type: registry.SZ, Name: "String3", Value: "Hello World"}, + {Type: registry.SZ, Name: "String4", Value: "Hello World\000", WillFail: true}, + {Type: registry.EXPAND_SZ, Name: "ExpString1", Value: ""}, + {Type: registry.EXPAND_SZ, Name: "ExpString2", Value: "\000", WillFail: true}, + {Type: registry.EXPAND_SZ, Name: "ExpString3", Value: "Hello World"}, + {Type: registry.EXPAND_SZ, Name: "ExpString4", Value: "Hello\000World", WillFail: true}, + {Type: registry.EXPAND_SZ, Name: "ExpString5", Value: "%PATH%"}, + {Type: registry.EXPAND_SZ, Name: "ExpString6", Value: "%NO_SUCH_VARIABLE%"}, + {Type: registry.EXPAND_SZ, Name: "ExpString7", Value: "%PATH%;."}, + {Type: registry.BINARY, Name: "Binary1", Value: []byte{}}, + {Type: registry.BINARY, Name: "Binary2", Value: []byte{1, 2, 3}}, + {Type: registry.BINARY, Name: "Binary3", Value: []byte{3, 2, 1, 0, 1, 2, 3}}, + {Type: registry.DWORD, Name: "Dword1", Value: uint64(0)}, + {Type: registry.DWORD, Name: "Dword2", Value: uint64(1)}, + {Type: registry.DWORD, Name: "Dword3", Value: uint64(0xff)}, + {Type: registry.DWORD, Name: "Dword4", Value: uint64(0xffff)}, + {Type: registry.QWORD, Name: "Qword1", Value: uint64(0)}, + {Type: registry.QWORD, Name: "Qword2", Value: uint64(1)}, + {Type: registry.QWORD, Name: "Qword3", Value: uint64(0xff)}, + {Type: registry.QWORD, Name: "Qword4", Value: uint64(0xffff)}, + {Type: registry.QWORD, Name: "Qword5", Value: uint64(0xffffff)}, + {Type: registry.QWORD, Name: "Qword6", Value: uint64(0xffffffff)}, + {Type: registry.MULTI_SZ, Name: "MultiString1", Value: []string{"a", "b", "c"}}, + {Type: registry.MULTI_SZ, Name: "MultiString2", Value: []string{"abc", "", "cba"}}, + {Type: registry.MULTI_SZ, Name: "MultiString3", Value: []string{""}}, + {Type: registry.MULTI_SZ, Name: "MultiString4", Value: []string{"abcdef"}}, + {Type: registry.MULTI_SZ, Name: "MultiString5", Value: []string{"\000"}, WillFail: true}, + {Type: registry.MULTI_SZ, Name: "MultiString6", Value: []string{"a\000b"}, WillFail: true}, + {Type: registry.MULTI_SZ, Name: "MultiString7", Value: []string{"ab", "\000", "cd"}, WillFail: true}, + {Type: registry.MULTI_SZ, Name: "MultiString8", Value: []string{"\000", "cd"}, WillFail: true}, + {Type: registry.MULTI_SZ, Name: "MultiString9", Value: []string{"ab", "\000"}, WillFail: true}, +} + +func setValues(t *testing.T, k registry.Key) { + for _, test := range ValueTests { + var err error + switch test.Type { + case registry.SZ: + err = k.SetStringValue(test.Name, test.Value.(string)) + case registry.EXPAND_SZ: + err = k.SetExpandStringValue(test.Name, test.Value.(string)) + case registry.MULTI_SZ: + err = k.SetStringsValue(test.Name, test.Value.([]string)) + case registry.BINARY: + err = k.SetBinaryValue(test.Name, test.Value.([]byte)) + case registry.DWORD: + err = k.SetDWordValue(test.Name, uint32(test.Value.(uint64))) + case registry.QWORD: + err = k.SetQWordValue(test.Name, test.Value.(uint64)) + default: + t.Fatalf("unsupported type %d for %s value", test.Type, test.Name) + } + if test.WillFail { + if err == nil { + t.Fatalf("setting %s value %q should fail, but succeeded", test.Name, test.Value) + } + } else { + if err != nil { + t.Fatal(err) + } + } + } +} + +func enumerateValues(t *testing.T, k registry.Key) { + names, err := k.ReadValueNames() + if err != nil { + t.Error(err) + return + } + haveNames := make(map[string]bool) + for _, n := range names { + haveNames[n] = false + } + for _, test := range ValueTests { + wantFound := !test.WillFail + _, haveFound := haveNames[test.Name] + if wantFound && !haveFound { + t.Errorf("value %s is not found while enumerating", test.Name) + } + if haveFound && !wantFound { + t.Errorf("value %s is found while enumerating, but expected to fail", test.Name) + } + if haveFound { + delete(haveNames, test.Name) + } + } + for n, v := range haveNames { + t.Errorf("value %s (%v) is found while enumerating, but has not been created", n, v) + } +} + +func testErrNotExist(t *testing.T, name string, err error) { + if err == nil { + t.Errorf("%s value should not exist", name) + return + } + if err != registry.ErrNotExist { + t.Errorf("reading %s value should return 'not exist' error, but got: %s", name, err) + return + } +} + +func testErrUnexpectedType(t *testing.T, test ValueTest, gottype uint32, err error) { + if err == nil { + t.Errorf("GetXValue(%q) should not succeed", test.Name) + return + } + if err != registry.ErrUnexpectedType { + t.Errorf("reading %s value should return 'unexpected key value type' error, but got: %s", test.Name, err) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } +} + +func testGetStringValue(t *testing.T, k registry.Key, test ValueTest) { + got, gottype, err := k.GetStringValue(test.Name) + if err != nil { + t.Errorf("GetStringValue(%s) failed: %v", test.Name, err) + return + } + if got != test.Value { + t.Errorf("want %s value %q, got %q", test.Name, test.Value, got) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } + if gottype == registry.EXPAND_SZ { + _, err = registry.ExpandString(got) + if err != nil { + t.Errorf("ExpandString(%s) failed: %v", got, err) + return + } + } +} + +func testGetIntegerValue(t *testing.T, k registry.Key, test ValueTest) { + got, gottype, err := k.GetIntegerValue(test.Name) + if err != nil { + t.Errorf("GetIntegerValue(%s) failed: %v", test.Name, err) + return + } + if got != test.Value.(uint64) { + t.Errorf("want %s value %v, got %v", test.Name, test.Value, got) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } +} + +func testGetBinaryValue(t *testing.T, k registry.Key, test ValueTest) { + got, gottype, err := k.GetBinaryValue(test.Name) + if err != nil { + t.Errorf("GetBinaryValue(%s) failed: %v", test.Name, err) + return + } + if !bytes.Equal(got, test.Value.([]byte)) { + t.Errorf("want %s value %v, got %v", test.Name, test.Value, got) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } +} + +func testGetStringsValue(t *testing.T, k registry.Key, test ValueTest) { + got, gottype, err := k.GetStringsValue(test.Name) + if err != nil { + t.Errorf("GetStringsValue(%s) failed: %v", test.Name, err) + return + } + if !equalStringSlice(got, test.Value.([]string)) { + t.Errorf("want %s value %#v, got %#v", test.Name, test.Value, got) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } +} + +func testGetValue(t *testing.T, k registry.Key, test ValueTest, size int) { + if size <= 0 { + return + } + // read data with no buffer + gotsize, gottype, err := k.GetValue(test.Name, nil) + if err != nil { + t.Errorf("GetValue(%s, [%d]byte) failed: %v", test.Name, size, err) + return + } + if gotsize != size { + t.Errorf("want %s value size of %d, got %v", test.Name, size, gotsize) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } + // read data with short buffer + gotsize, gottype, err = k.GetValue(test.Name, make([]byte, size-1)) + if err == nil { + t.Errorf("GetValue(%s, [%d]byte) should fail, but succeeded", test.Name, size-1) + return + } + if err != registry.ErrShortBuffer { + t.Errorf("reading %s value should return 'short buffer' error, but got: %s", test.Name, err) + return + } + if gotsize != size { + t.Errorf("want %s value size of %d, got %v", test.Name, size, gotsize) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } + // read full data + gotsize, gottype, err = k.GetValue(test.Name, make([]byte, size)) + if err != nil { + t.Errorf("GetValue(%s, [%d]byte) failed: %v", test.Name, size, err) + return + } + if gotsize != size { + t.Errorf("want %s value size of %d, got %v", test.Name, size, gotsize) + return + } + if gottype != test.Type { + t.Errorf("want %s value type %v, got %v", test.Name, test.Type, gottype) + return + } + // check GetValue returns ErrNotExist as required + _, _, err = k.GetValue(test.Name+"_not_there", make([]byte, size)) + if err == nil { + t.Errorf("GetValue(%q) should not succeed", test.Name) + return + } + if err != registry.ErrNotExist { + t.Errorf("GetValue(%q) should return 'not exist' error, but got: %s", test.Name, err) + return + } +} + +func testValues(t *testing.T, k registry.Key) { + for _, test := range ValueTests { + switch test.Type { + case registry.SZ, registry.EXPAND_SZ: + if test.WillFail { + _, _, err := k.GetStringValue(test.Name) + testErrNotExist(t, test.Name, err) + } else { + testGetStringValue(t, k, test) + _, gottype, err := k.GetIntegerValue(test.Name) + testErrUnexpectedType(t, test, gottype, err) + // Size of utf16 string in bytes is not perfect, + // but correct for current test values. + // Size also includes terminating 0. + testGetValue(t, k, test, (len(test.Value.(string))+1)*2) + } + _, _, err := k.GetStringValue(test.Name + "_string_not_created") + testErrNotExist(t, test.Name+"_string_not_created", err) + case registry.DWORD, registry.QWORD: + testGetIntegerValue(t, k, test) + _, gottype, err := k.GetBinaryValue(test.Name) + testErrUnexpectedType(t, test, gottype, err) + _, _, err = k.GetIntegerValue(test.Name + "_int_not_created") + testErrNotExist(t, test.Name+"_int_not_created", err) + size := 8 + if test.Type == registry.DWORD { + size = 4 + } + testGetValue(t, k, test, size) + case registry.BINARY: + testGetBinaryValue(t, k, test) + _, gottype, err := k.GetStringsValue(test.Name) + testErrUnexpectedType(t, test, gottype, err) + _, _, err = k.GetBinaryValue(test.Name + "_byte_not_created") + testErrNotExist(t, test.Name+"_byte_not_created", err) + testGetValue(t, k, test, len(test.Value.([]byte))) + case registry.MULTI_SZ: + if test.WillFail { + _, _, err := k.GetStringsValue(test.Name) + testErrNotExist(t, test.Name, err) + } else { + testGetStringsValue(t, k, test) + _, gottype, err := k.GetStringValue(test.Name) + testErrUnexpectedType(t, test, gottype, err) + size := 0 + for _, s := range test.Value.([]string) { + size += len(s) + 1 // nil terminated + } + size += 1 // extra nil at the end + size *= 2 // count bytes, not uint16 + testGetValue(t, k, test, size) + } + _, _, err := k.GetStringsValue(test.Name + "_strings_not_created") + testErrNotExist(t, test.Name+"_strings_not_created", err) + default: + t.Errorf("unsupported type %d for %s value", test.Type, test.Name) + continue + } + } +} + +func testStat(t *testing.T, k registry.Key) { + subk, _, err := registry.CreateKey(k, "subkey", registry.CREATE_SUB_KEY) + if err != nil { + t.Error(err) + return + } + defer subk.Close() + + defer registry.DeleteKey(k, "subkey") + + ki, err := k.Stat() + if err != nil { + t.Error(err) + return + } + if ki.SubKeyCount != 1 { + t.Error("key must have 1 subkey") + } + if ki.MaxSubKeyLen != 6 { + t.Error("key max subkey name length must be 6") + } + if ki.ValueCount != 24 { + t.Errorf("key must have 24 values, but is %d", ki.ValueCount) + } + if ki.MaxValueNameLen != 12 { + t.Errorf("key max value name length must be 10, but is %d", ki.MaxValueNameLen) + } + if ki.MaxValueLen != 38 { + t.Errorf("key max value length must be 38, but is %d", ki.MaxValueLen) + } +} + +func deleteValues(t *testing.T, k registry.Key) { + for _, test := range ValueTests { + if test.WillFail { + continue + } + err := k.DeleteValue(test.Name) + if err != nil { + t.Error(err) + continue + } + } + names, err := k.ReadValueNames() + if err != nil { + t.Error(err) + return + } + if len(names) != 0 { + t.Errorf("some values remain after deletion: %v", names) + } +} + +func TestValues(t *testing.T) { + softwareK, err := registry.OpenKey(registry.CURRENT_USER, "Software", registry.QUERY_VALUE) + if err != nil { + t.Fatal(err) + } + defer softwareK.Close() + + testKName := randKeyName("TestValues_") + + k, exist, err := registry.CreateKey(softwareK, testKName, registry.CREATE_SUB_KEY|registry.QUERY_VALUE|registry.SET_VALUE) + if err != nil { + t.Fatal(err) + } + defer k.Close() + + if exist { + t.Fatalf("key %q already exists", testKName) + } + + defer registry.DeleteKey(softwareK, testKName) + + setValues(t, k) + + enumerateValues(t, k) + + testValues(t, k) + + testStat(t, k) + + deleteValues(t, k) +} + +func TestExpandString(t *testing.T) { + got, err := registry.ExpandString("%PATH%") + if err != nil { + t.Fatal(err) + } + want := os.Getenv("PATH") + if got != want { + t.Errorf("want %q string expanded, got %q", want, got) + } +} + +func TestInvalidValues(t *testing.T) { + softwareK, err := registry.OpenKey(registry.CURRENT_USER, "Software", registry.QUERY_VALUE) + if err != nil { + t.Fatal(err) + } + defer softwareK.Close() + + testKName := randKeyName("TestInvalidValues_") + + k, exist, err := registry.CreateKey(softwareK, testKName, registry.CREATE_SUB_KEY|registry.QUERY_VALUE|registry.SET_VALUE) + if err != nil { + t.Fatal(err) + } + defer k.Close() + + if exist { + t.Fatalf("key %q already exists", testKName) + } + + defer registry.DeleteKey(softwareK, testKName) + + var tests = []struct { + Type uint32 + Name string + Data []byte + }{ + {registry.DWORD, "Dword1", nil}, + {registry.DWORD, "Dword2", []byte{1, 2, 3}}, + {registry.QWORD, "Qword1", nil}, + {registry.QWORD, "Qword2", []byte{1, 2, 3}}, + {registry.QWORD, "Qword3", []byte{1, 2, 3, 4, 5, 6, 7}}, + {registry.MULTI_SZ, "MultiString1", nil}, + {registry.MULTI_SZ, "MultiString2", []byte{0}}, + {registry.MULTI_SZ, "MultiString3", []byte{'a', 'b', 0}}, + {registry.MULTI_SZ, "MultiString4", []byte{'a', 0, 0, 'b', 0}}, + {registry.MULTI_SZ, "MultiString5", []byte{'a', 0, 0}}, + } + + for _, test := range tests { + err := k.SetValue(test.Name, test.Type, test.Data) + if err != nil { + t.Fatalf("SetValue for %q failed: %v", test.Name, err) + } + } + + for _, test := range tests { + switch test.Type { + case registry.DWORD, registry.QWORD: + value, valType, err := k.GetIntegerValue(test.Name) + if err == nil { + t.Errorf("GetIntegerValue(%q) succeeded. Returns type=%d value=%v", test.Name, valType, value) + } + case registry.MULTI_SZ: + value, valType, err := k.GetStringsValue(test.Name) + if err == nil { + if len(value) != 0 { + t.Errorf("GetStringsValue(%q) succeeded. Returns type=%d value=%v", test.Name, valType, value) + } + } + default: + t.Errorf("unsupported type %d for %s value", test.Type, test.Name) + } + } +} + +func TestGetMUIStringValue(t *testing.T) { + if err := registry.LoadRegLoadMUIString(); err != nil { + t.Skip("regLoadMUIString not supported; skipping") + } + if err := procGetDynamicTimeZoneInformation.Find(); err != nil { + t.Skipf("%s not supported; skipping", procGetDynamicTimeZoneInformation.Name) + } + var dtzi DynamicTimezoneinformation + if _, err := GetDynamicTimeZoneInformation(&dtzi); err != nil { + t.Fatal(err) + } + tzKeyName := syscall.UTF16ToString(dtzi.TimeZoneKeyName[:]) + timezoneK, err := registry.OpenKey(registry.LOCAL_MACHINE, + `SOFTWARE\Microsoft\Windows NT\CurrentVersion\Time Zones\`+tzKeyName, registry.READ) + if err != nil { + t.Fatal(err) + } + defer timezoneK.Close() + + type testType struct { + name string + want string + } + var tests = []testType{ + {"MUI_Std", syscall.UTF16ToString(dtzi.StandardName[:])}, + } + if dtzi.DynamicDaylightTimeDisabled == 0 { + tests = append(tests, testType{"MUI_Dlt", syscall.UTF16ToString(dtzi.DaylightName[:])}) + } + + for _, test := range tests { + got, err := timezoneK.GetMUIStringValue(test.name) + if err != nil { + t.Error("GetMUIStringValue:", err) + } + + if got != test.want { + t.Errorf("GetMUIStringValue: %s: Got %q, want %q", test.name, got, test.want) + } + } +} + +type DynamicTimezoneinformation struct { + Bias int32 + StandardName [32]uint16 + StandardDate syscall.Systemtime + StandardBias int32 + DaylightName [32]uint16 + DaylightDate syscall.Systemtime + DaylightBias int32 + TimeZoneKeyName [128]uint16 + DynamicDaylightTimeDisabled uint8 +} + +var ( + kernel32DLL = syscall.NewLazyDLL("kernel32") + + procGetDynamicTimeZoneInformation = kernel32DLL.NewProc("GetDynamicTimeZoneInformation") +) + +func GetDynamicTimeZoneInformation(dtzi *DynamicTimezoneinformation) (rc uint32, err error) { + r0, _, e1 := syscall.Syscall(procGetDynamicTimeZoneInformation.Addr(), 1, uintptr(unsafe.Pointer(dtzi)), 0, 0) + rc = uint32(r0) + if rc == 0xffffffff { + if e1 != 0 { + err = error(e1) + } else { + err = syscall.EINVAL + } + } + return +} diff --git a/src/internal/syscall/windows/registry/syscall.go b/src/internal/syscall/windows/registry/syscall.go new file mode 100644 index 0000000..cb315ad --- /dev/null +++ b/src/internal/syscall/windows/registry/syscall.go @@ -0,0 +1,31 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package registry + +import "syscall" + +const ( + _REG_OPTION_NON_VOLATILE = 0 + + _REG_CREATED_NEW_KEY = 1 + _REG_OPENED_EXISTING_KEY = 2 + + _ERROR_NO_MORE_ITEMS syscall.Errno = 259 +) + +func LoadRegLoadMUIString() error { + return procRegLoadMUIStringW.Find() +} + +//sys regCreateKeyEx(key syscall.Handle, subkey *uint16, reserved uint32, class *uint16, options uint32, desired uint32, sa *syscall.SecurityAttributes, result *syscall.Handle, disposition *uint32) (regerrno error) = advapi32.RegCreateKeyExW +//sys regDeleteKey(key syscall.Handle, subkey *uint16) (regerrno error) = advapi32.RegDeleteKeyW +//sys regSetValueEx(key syscall.Handle, valueName *uint16, reserved uint32, vtype uint32, buf *byte, bufsize uint32) (regerrno error) = advapi32.RegSetValueExW +//sys regEnumValue(key syscall.Handle, index uint32, name *uint16, nameLen *uint32, reserved *uint32, valtype *uint32, buf *byte, buflen *uint32) (regerrno error) = advapi32.RegEnumValueW +//sys regDeleteValue(key syscall.Handle, name *uint16) (regerrno error) = advapi32.RegDeleteValueW +//sys regLoadMUIString(key syscall.Handle, name *uint16, buf *uint16, buflen uint32, buflenCopied *uint32, flags uint32, dir *uint16) (regerrno error) = advapi32.RegLoadMUIStringW + +//sys expandEnvironmentStrings(src *uint16, dst *uint16, size uint32) (n uint32, err error) = kernel32.ExpandEnvironmentStringsW diff --git a/src/internal/syscall/windows/registry/value.go b/src/internal/syscall/windows/registry/value.go new file mode 100644 index 0000000..7dfee03 --- /dev/null +++ b/src/internal/syscall/windows/registry/value.go @@ -0,0 +1,372 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +package registry + +import ( + "errors" + "syscall" + "unicode/utf16" + "unsafe" +) + +const ( + // Registry value types. + NONE = 0 + SZ = 1 + EXPAND_SZ = 2 + BINARY = 3 + DWORD = 4 + DWORD_BIG_ENDIAN = 5 + LINK = 6 + MULTI_SZ = 7 + RESOURCE_LIST = 8 + FULL_RESOURCE_DESCRIPTOR = 9 + RESOURCE_REQUIREMENTS_LIST = 10 + QWORD = 11 +) + +var ( + // ErrShortBuffer is returned when the buffer was too short for the operation. + ErrShortBuffer = syscall.ERROR_MORE_DATA + + // ErrNotExist is returned when a registry key or value does not exist. + ErrNotExist = syscall.ERROR_FILE_NOT_FOUND + + // ErrUnexpectedType is returned by Get*Value when the value's type was unexpected. + ErrUnexpectedType = errors.New("unexpected key value type") +) + +// GetValue retrieves the type and data for the specified value associated +// with an open key k. It fills up buffer buf and returns the retrieved +// byte count n. If buf is too small to fit the stored value it returns +// ErrShortBuffer error along with the required buffer size n. +// If no buffer is provided, it returns true and actual buffer size n. +// If no buffer is provided, GetValue returns the value's type only. +// If the value does not exist, the error returned is ErrNotExist. +// +// GetValue is a low level function. If value's type is known, use the appropriate +// Get*Value function instead. +func (k Key) GetValue(name string, buf []byte) (n int, valtype uint32, err error) { + pname, err := syscall.UTF16PtrFromString(name) + if err != nil { + return 0, 0, err + } + var pbuf *byte + if len(buf) > 0 { + pbuf = (*byte)(unsafe.Pointer(&buf[0])) + } + l := uint32(len(buf)) + err = syscall.RegQueryValueEx(syscall.Handle(k), pname, nil, &valtype, pbuf, &l) + if err != nil { + return int(l), valtype, err + } + return int(l), valtype, nil +} + +func (k Key) getValue(name string, buf []byte) (date []byte, valtype uint32, err error) { + p, err := syscall.UTF16PtrFromString(name) + if err != nil { + return nil, 0, err + } + var t uint32 + n := uint32(len(buf)) + for { + err = syscall.RegQueryValueEx(syscall.Handle(k), p, nil, &t, (*byte)(unsafe.Pointer(&buf[0])), &n) + if err == nil { + return buf[:n], t, nil + } + if err != syscall.ERROR_MORE_DATA { + return nil, 0, err + } + if n <= uint32(len(buf)) { + return nil, 0, err + } + buf = make([]byte, n) + } +} + +// GetStringValue retrieves the string value for the specified +// value name associated with an open key k. It also returns the value's type. +// If value does not exist, GetStringValue returns ErrNotExist. +// If value is not SZ or EXPAND_SZ, it will return the correct value +// type and ErrUnexpectedType. +func (k Key) GetStringValue(name string) (val string, valtype uint32, err error) { + data, typ, err2 := k.getValue(name, make([]byte, 64)) + if err2 != nil { + return "", typ, err2 + } + switch typ { + case SZ, EXPAND_SZ: + default: + return "", typ, ErrUnexpectedType + } + if len(data) == 0 { + return "", typ, nil + } + u := (*[1 << 29]uint16)(unsafe.Pointer(&data[0]))[: len(data)/2 : len(data)/2] + return syscall.UTF16ToString(u), typ, nil +} + +// GetMUIStringValue retrieves the localized string value for +// the specified value name associated with an open key k. +// If the value name doesn't exist or the localized string value +// can't be resolved, GetMUIStringValue returns ErrNotExist. +// GetMUIStringValue panics if the system doesn't support +// regLoadMUIString; use LoadRegLoadMUIString to check if +// regLoadMUIString is supported before calling this function. +func (k Key) GetMUIStringValue(name string) (string, error) { + pname, err := syscall.UTF16PtrFromString(name) + if err != nil { + return "", err + } + + buf := make([]uint16, 1024) + var buflen uint32 + var pdir *uint16 + + err = regLoadMUIString(syscall.Handle(k), pname, &buf[0], uint32(len(buf)), &buflen, 0, pdir) + if err == syscall.ERROR_FILE_NOT_FOUND { // Try fallback path + + // Try to resolve the string value using the system directory as + // a DLL search path; this assumes the string value is of the form + // @[path]\dllname,-strID but with no path given, e.g. @tzres.dll,-320. + + // This approach works with tzres.dll but may have to be revised + // in the future to allow callers to provide custom search paths. + + var s string + s, err = ExpandString("%SystemRoot%\\system32\\") + if err != nil { + return "", err + } + pdir, err = syscall.UTF16PtrFromString(s) + if err != nil { + return "", err + } + + err = regLoadMUIString(syscall.Handle(k), pname, &buf[0], uint32(len(buf)), &buflen, 0, pdir) + } + + for err == syscall.ERROR_MORE_DATA { // Grow buffer if needed + if buflen <= uint32(len(buf)) { + break // Buffer not growing, assume race; break + } + buf = make([]uint16, buflen) + err = regLoadMUIString(syscall.Handle(k), pname, &buf[0], uint32(len(buf)), &buflen, 0, pdir) + } + + if err != nil { + return "", err + } + + return syscall.UTF16ToString(buf), nil +} + +// ExpandString expands environment-variable strings and replaces +// them with the values defined for the current user. +// Use ExpandString to expand EXPAND_SZ strings. +func ExpandString(value string) (string, error) { + if value == "" { + return "", nil + } + p, err := syscall.UTF16PtrFromString(value) + if err != nil { + return "", err + } + r := make([]uint16, 100) + for { + n, err := expandEnvironmentStrings(p, &r[0], uint32(len(r))) + if err != nil { + return "", err + } + if n <= uint32(len(r)) { + return syscall.UTF16ToString(r[:n]), nil + } + r = make([]uint16, n) + } +} + +// GetStringsValue retrieves the []string value for the specified +// value name associated with an open key k. It also returns the value's type. +// If value does not exist, GetStringsValue returns ErrNotExist. +// If value is not MULTI_SZ, it will return the correct value +// type and ErrUnexpectedType. +func (k Key) GetStringsValue(name string) (val []string, valtype uint32, err error) { + data, typ, err2 := k.getValue(name, make([]byte, 64)) + if err2 != nil { + return nil, typ, err2 + } + if typ != MULTI_SZ { + return nil, typ, ErrUnexpectedType + } + if len(data) == 0 { + return nil, typ, nil + } + p := (*[1 << 29]uint16)(unsafe.Pointer(&data[0]))[: len(data)/2 : len(data)/2] + if len(p) == 0 { + return nil, typ, nil + } + if p[len(p)-1] == 0 { + p = p[:len(p)-1] // remove terminating null + } + val = make([]string, 0, 5) + from := 0 + for i, c := range p { + if c == 0 { + val = append(val, syscall.UTF16ToString(p[from:i])) + from = i + 1 + } + } + return val, typ, nil +} + +// GetIntegerValue retrieves the integer value for the specified +// value name associated with an open key k. It also returns the value's type. +// If value does not exist, GetIntegerValue returns ErrNotExist. +// If value is not DWORD or QWORD, it will return the correct value +// type and ErrUnexpectedType. +func (k Key) GetIntegerValue(name string) (val uint64, valtype uint32, err error) { + data, typ, err2 := k.getValue(name, make([]byte, 8)) + if err2 != nil { + return 0, typ, err2 + } + switch typ { + case DWORD: + if len(data) != 4 { + return 0, typ, errors.New("DWORD value is not 4 bytes long") + } + return uint64(*(*uint32)(unsafe.Pointer(&data[0]))), DWORD, nil + case QWORD: + if len(data) != 8 { + return 0, typ, errors.New("QWORD value is not 8 bytes long") + } + return uint64(*(*uint64)(unsafe.Pointer(&data[0]))), QWORD, nil + default: + return 0, typ, ErrUnexpectedType + } +} + +// GetBinaryValue retrieves the binary value for the specified +// value name associated with an open key k. It also returns the value's type. +// If value does not exist, GetBinaryValue returns ErrNotExist. +// If value is not BINARY, it will return the correct value +// type and ErrUnexpectedType. +func (k Key) GetBinaryValue(name string) (val []byte, valtype uint32, err error) { + data, typ, err2 := k.getValue(name, make([]byte, 64)) + if err2 != nil { + return nil, typ, err2 + } + if typ != BINARY { + return nil, typ, ErrUnexpectedType + } + return data, typ, nil +} + +func (k Key) setValue(name string, valtype uint32, data []byte) error { + p, err := syscall.UTF16PtrFromString(name) + if err != nil { + return err + } + if len(data) == 0 { + return regSetValueEx(syscall.Handle(k), p, 0, valtype, nil, 0) + } + return regSetValueEx(syscall.Handle(k), p, 0, valtype, &data[0], uint32(len(data))) +} + +// SetDWordValue sets the data and type of a name value +// under key k to value and DWORD. +func (k Key) SetDWordValue(name string, value uint32) error { + return k.setValue(name, DWORD, (*[4]byte)(unsafe.Pointer(&value))[:]) +} + +// SetQWordValue sets the data and type of a name value +// under key k to value and QWORD. +func (k Key) SetQWordValue(name string, value uint64) error { + return k.setValue(name, QWORD, (*[8]byte)(unsafe.Pointer(&value))[:]) +} + +func (k Key) setStringValue(name string, valtype uint32, value string) error { + v, err := syscall.UTF16FromString(value) + if err != nil { + return err + } + buf := (*[1 << 29]byte)(unsafe.Pointer(&v[0]))[: len(v)*2 : len(v)*2] + return k.setValue(name, valtype, buf) +} + +// SetStringValue sets the data and type of a name value +// under key k to value and SZ. The value must not contain a zero byte. +func (k Key) SetStringValue(name, value string) error { + return k.setStringValue(name, SZ, value) +} + +// SetExpandStringValue sets the data and type of a name value +// under key k to value and EXPAND_SZ. The value must not contain a zero byte. +func (k Key) SetExpandStringValue(name, value string) error { + return k.setStringValue(name, EXPAND_SZ, value) +} + +// SetStringsValue sets the data and type of a name value +// under key k to value and MULTI_SZ. The value strings +// must not contain a zero byte. +func (k Key) SetStringsValue(name string, value []string) error { + ss := "" + for _, s := range value { + for i := 0; i < len(s); i++ { + if s[i] == 0 { + return errors.New("string cannot have 0 inside") + } + } + ss += s + "\x00" + } + v := utf16.Encode([]rune(ss + "\x00")) + buf := (*[1 << 29]byte)(unsafe.Pointer(&v[0]))[: len(v)*2 : len(v)*2] + return k.setValue(name, MULTI_SZ, buf) +} + +// SetBinaryValue sets the data and type of a name value +// under key k to value and BINARY. +func (k Key) SetBinaryValue(name string, value []byte) error { + return k.setValue(name, BINARY, value) +} + +// DeleteValue removes a named value from the key k. +func (k Key) DeleteValue(name string) error { + return regDeleteValue(syscall.Handle(k), syscall.StringToUTF16Ptr(name)) +} + +// ReadValueNames returns the value names of key k. +func (k Key) ReadValueNames() ([]string, error) { + ki, err := k.Stat() + if err != nil { + return nil, err + } + names := make([]string, 0, ki.ValueCount) + buf := make([]uint16, ki.MaxValueNameLen+1) // extra room for terminating null character +loopItems: + for i := uint32(0); ; i++ { + l := uint32(len(buf)) + for { + err := regEnumValue(syscall.Handle(k), i, &buf[0], &l, nil, nil, nil, nil) + if err == nil { + break + } + if err == syscall.ERROR_MORE_DATA { + // Double buffer size and try again. + l = uint32(2 * len(buf)) + buf = make([]uint16, l) + continue + } + if err == _ERROR_NO_MORE_ITEMS { + break loopItems + } + return names, err + } + names = append(names, syscall.UTF16ToString(buf[:l])) + } + return names, nil +} diff --git a/src/internal/syscall/windows/registry/zsyscall_windows.go b/src/internal/syscall/windows/registry/zsyscall_windows.go new file mode 100644 index 0000000..cab1319 --- /dev/null +++ b/src/internal/syscall/windows/registry/zsyscall_windows.go @@ -0,0 +1,107 @@ +// Code generated by 'go generate'; DO NOT EDIT. + +package registry + +import ( + "internal/syscall/windows/sysdll" + "syscall" + "unsafe" +) + +var _ unsafe.Pointer + +// Do the interface allocations only once for common +// Errno values. +const ( + errnoERROR_IO_PENDING = 997 +) + +var ( + errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING) + errERROR_EINVAL error = syscall.EINVAL +) + +// errnoErr returns common boxed Errno values, to prevent +// allocations at runtime. +func errnoErr(e syscall.Errno) error { + switch e { + case 0: + return errERROR_EINVAL + case errnoERROR_IO_PENDING: + return errERROR_IO_PENDING + } + // TODO: add more here, after collecting data on the common + // error values see on Windows. (perhaps when running + // all.bat?) + return e +} + +var ( + modadvapi32 = syscall.NewLazyDLL(sysdll.Add("advapi32.dll")) + modkernel32 = syscall.NewLazyDLL(sysdll.Add("kernel32.dll")) + + procRegCreateKeyExW = modadvapi32.NewProc("RegCreateKeyExW") + procRegDeleteKeyW = modadvapi32.NewProc("RegDeleteKeyW") + procRegDeleteValueW = modadvapi32.NewProc("RegDeleteValueW") + procRegEnumValueW = modadvapi32.NewProc("RegEnumValueW") + procRegLoadMUIStringW = modadvapi32.NewProc("RegLoadMUIStringW") + procRegSetValueExW = modadvapi32.NewProc("RegSetValueExW") + procExpandEnvironmentStringsW = modkernel32.NewProc("ExpandEnvironmentStringsW") +) + +func regCreateKeyEx(key syscall.Handle, subkey *uint16, reserved uint32, class *uint16, options uint32, desired uint32, sa *syscall.SecurityAttributes, result *syscall.Handle, disposition *uint32) (regerrno error) { + r0, _, _ := syscall.Syscall9(procRegCreateKeyExW.Addr(), 9, uintptr(key), uintptr(unsafe.Pointer(subkey)), uintptr(reserved), uintptr(unsafe.Pointer(class)), uintptr(options), uintptr(desired), uintptr(unsafe.Pointer(sa)), uintptr(unsafe.Pointer(result)), uintptr(unsafe.Pointer(disposition))) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func regDeleteKey(key syscall.Handle, subkey *uint16) (regerrno error) { + r0, _, _ := syscall.Syscall(procRegDeleteKeyW.Addr(), 2, uintptr(key), uintptr(unsafe.Pointer(subkey)), 0) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func regDeleteValue(key syscall.Handle, name *uint16) (regerrno error) { + r0, _, _ := syscall.Syscall(procRegDeleteValueW.Addr(), 2, uintptr(key), uintptr(unsafe.Pointer(name)), 0) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func regEnumValue(key syscall.Handle, index uint32, name *uint16, nameLen *uint32, reserved *uint32, valtype *uint32, buf *byte, buflen *uint32) (regerrno error) { + r0, _, _ := syscall.Syscall9(procRegEnumValueW.Addr(), 8, uintptr(key), uintptr(index), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(nameLen)), uintptr(unsafe.Pointer(reserved)), uintptr(unsafe.Pointer(valtype)), uintptr(unsafe.Pointer(buf)), uintptr(unsafe.Pointer(buflen)), 0) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func regLoadMUIString(key syscall.Handle, name *uint16, buf *uint16, buflen uint32, buflenCopied *uint32, flags uint32, dir *uint16) (regerrno error) { + r0, _, _ := syscall.Syscall9(procRegLoadMUIStringW.Addr(), 7, uintptr(key), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(buf)), uintptr(buflen), uintptr(unsafe.Pointer(buflenCopied)), uintptr(flags), uintptr(unsafe.Pointer(dir)), 0, 0) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func regSetValueEx(key syscall.Handle, valueName *uint16, reserved uint32, vtype uint32, buf *byte, bufsize uint32) (regerrno error) { + r0, _, _ := syscall.Syscall6(procRegSetValueExW.Addr(), 6, uintptr(key), uintptr(unsafe.Pointer(valueName)), uintptr(reserved), uintptr(vtype), uintptr(unsafe.Pointer(buf)), uintptr(bufsize)) + if r0 != 0 { + regerrno = syscall.Errno(r0) + } + return +} + +func expandEnvironmentStrings(src *uint16, dst *uint16, size uint32) (n uint32, err error) { + r0, _, e1 := syscall.Syscall(procExpandEnvironmentStringsW.Addr(), 3, uintptr(unsafe.Pointer(src)), uintptr(unsafe.Pointer(dst)), uintptr(size)) + n = uint32(r0) + if n == 0 { + err = errnoErr(e1) + } + return +} diff --git a/src/internal/syscall/windows/reparse_windows.go b/src/internal/syscall/windows/reparse_windows.go new file mode 100644 index 0000000..6caf47e --- /dev/null +++ b/src/internal/syscall/windows/reparse_windows.go @@ -0,0 +1,91 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +import ( + "syscall" + "unsafe" +) + +const ( + FSCTL_SET_REPARSE_POINT = 0x000900A4 + IO_REPARSE_TAG_MOUNT_POINT = 0xA0000003 + IO_REPARSE_TAG_DEDUP = 0x80000013 + + SYMLINK_FLAG_RELATIVE = 1 +) + +// These structures are described +// in https://msdn.microsoft.com/en-us/library/cc232007.aspx +// and https://msdn.microsoft.com/en-us/library/cc232006.aspx. + +type REPARSE_DATA_BUFFER struct { + ReparseTag uint32 + ReparseDataLength uint16 + Reserved uint16 + DUMMYUNIONNAME byte +} + +// REPARSE_DATA_BUFFER_HEADER is a common part of REPARSE_DATA_BUFFER structure. +type REPARSE_DATA_BUFFER_HEADER struct { + ReparseTag uint32 + // The size, in bytes, of the reparse data that follows + // the common portion of the REPARSE_DATA_BUFFER element. + // This value is the length of the data starting at the + // SubstituteNameOffset field. + ReparseDataLength uint16 + Reserved uint16 +} + +type SymbolicLinkReparseBuffer struct { + // The integer that contains the offset, in bytes, + // of the substitute name string in the PathBuffer array, + // computed as an offset from byte 0 of PathBuffer. Note that + // this offset must be divided by 2 to get the array index. + SubstituteNameOffset uint16 + // The integer that contains the length, in bytes, of the + // substitute name string. If this string is null-terminated, + // SubstituteNameLength does not include the Unicode null character. + SubstituteNameLength uint16 + // PrintNameOffset is similar to SubstituteNameOffset. + PrintNameOffset uint16 + // PrintNameLength is similar to SubstituteNameLength. + PrintNameLength uint16 + // Flags specifies whether the substitute name is a full path name or + // a path name relative to the directory containing the symbolic link. + Flags uint32 + PathBuffer [1]uint16 +} + +// Path returns path stored in rb. +func (rb *SymbolicLinkReparseBuffer) Path() string { + n1 := rb.SubstituteNameOffset / 2 + n2 := (rb.SubstituteNameOffset + rb.SubstituteNameLength) / 2 + return syscall.UTF16ToString((*[0xffff]uint16)(unsafe.Pointer(&rb.PathBuffer[0]))[n1:n2:n2]) +} + +type MountPointReparseBuffer struct { + // The integer that contains the offset, in bytes, + // of the substitute name string in the PathBuffer array, + // computed as an offset from byte 0 of PathBuffer. Note that + // this offset must be divided by 2 to get the array index. + SubstituteNameOffset uint16 + // The integer that contains the length, in bytes, of the + // substitute name string. If this string is null-terminated, + // SubstituteNameLength does not include the Unicode null character. + SubstituteNameLength uint16 + // PrintNameOffset is similar to SubstituteNameOffset. + PrintNameOffset uint16 + // PrintNameLength is similar to SubstituteNameLength. + PrintNameLength uint16 + PathBuffer [1]uint16 +} + +// Path returns path stored in rb. +func (rb *MountPointReparseBuffer) Path() string { + n1 := rb.SubstituteNameOffset / 2 + n2 := (rb.SubstituteNameOffset + rb.SubstituteNameLength) / 2 + return syscall.UTF16ToString((*[0xffff]uint16)(unsafe.Pointer(&rb.PathBuffer[0]))[n1:n2:n2]) +} diff --git a/src/internal/syscall/windows/security_windows.go b/src/internal/syscall/windows/security_windows.go new file mode 100644 index 0000000..4a2dfc0 --- /dev/null +++ b/src/internal/syscall/windows/security_windows.go @@ -0,0 +1,128 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +import ( + "syscall" + "unsafe" +) + +const ( + SecurityAnonymous = 0 + SecurityIdentification = 1 + SecurityImpersonation = 2 + SecurityDelegation = 3 +) + +//sys ImpersonateSelf(impersonationlevel uint32) (err error) = advapi32.ImpersonateSelf +//sys RevertToSelf() (err error) = advapi32.RevertToSelf + +const ( + TOKEN_ADJUST_PRIVILEGES = 0x0020 + SE_PRIVILEGE_ENABLED = 0x00000002 +) + +type LUID struct { + LowPart uint32 + HighPart int32 +} + +type LUID_AND_ATTRIBUTES struct { + Luid LUID + Attributes uint32 +} + +type TOKEN_PRIVILEGES struct { + PrivilegeCount uint32 + Privileges [1]LUID_AND_ATTRIBUTES +} + +//sys OpenThreadToken(h syscall.Handle, access uint32, openasself bool, token *syscall.Token) (err error) = advapi32.OpenThreadToken +//sys LookupPrivilegeValue(systemname *uint16, name *uint16, luid *LUID) (err error) = advapi32.LookupPrivilegeValueW +//sys adjustTokenPrivileges(token syscall.Token, disableAllPrivileges bool, newstate *TOKEN_PRIVILEGES, buflen uint32, prevstate *TOKEN_PRIVILEGES, returnlen *uint32) (ret uint32, err error) [true] = advapi32.AdjustTokenPrivileges + +func AdjustTokenPrivileges(token syscall.Token, disableAllPrivileges bool, newstate *TOKEN_PRIVILEGES, buflen uint32, prevstate *TOKEN_PRIVILEGES, returnlen *uint32) error { + ret, err := adjustTokenPrivileges(token, disableAllPrivileges, newstate, buflen, prevstate, returnlen) + if ret == 0 { + // AdjustTokenPrivileges call failed + return err + } + // AdjustTokenPrivileges call succeeded + if err == syscall.EINVAL { + // GetLastError returned ERROR_SUCCESS + return nil + } + return err +} + +//sys DuplicateTokenEx(hExistingToken syscall.Token, dwDesiredAccess uint32, lpTokenAttributes *syscall.SecurityAttributes, impersonationLevel uint32, tokenType TokenType, phNewToken *syscall.Token) (err error) = advapi32.DuplicateTokenEx +//sys SetTokenInformation(tokenHandle syscall.Token, tokenInformationClass uint32, tokenInformation uintptr, tokenInformationLength uint32) (err error) = advapi32.SetTokenInformation + +type SID_AND_ATTRIBUTES struct { + Sid *syscall.SID + Attributes uint32 +} + +type TOKEN_MANDATORY_LABEL struct { + Label SID_AND_ATTRIBUTES +} + +func (tml *TOKEN_MANDATORY_LABEL) Size() uint32 { + return uint32(unsafe.Sizeof(TOKEN_MANDATORY_LABEL{})) + syscall.GetLengthSid(tml.Label.Sid) +} + +const SE_GROUP_INTEGRITY = 0x00000020 + +type TokenType uint32 + +const ( + TokenPrimary TokenType = 1 + TokenImpersonation TokenType = 2 +) + +//sys GetProfilesDirectory(dir *uint16, dirLen *uint32) (err error) = userenv.GetProfilesDirectoryW + +const ( + LG_INCLUDE_INDIRECT = 0x1 + MAX_PREFERRED_LENGTH = 0xFFFFFFFF +) + +type LocalGroupUserInfo0 struct { + Name *uint16 +} + +type UserInfo4 struct { + Name *uint16 + Password *uint16 + PasswordAge uint32 + Priv uint32 + HomeDir *uint16 + Comment *uint16 + Flags uint32 + ScriptPath *uint16 + AuthFlags uint32 + FullName *uint16 + UsrComment *uint16 + Parms *uint16 + Workstations *uint16 + LastLogon uint32 + LastLogoff uint32 + AcctExpires uint32 + MaxStorage uint32 + UnitsPerWeek uint32 + LogonHours *byte + BadPwCount uint32 + NumLogons uint32 + LogonServer *uint16 + CountryCode uint32 + CodePage uint32 + UserSid *syscall.SID + PrimaryGroupID uint32 + Profile *uint16 + HomeDirDrive *uint16 + PasswordExpired uint32 +} + +//sys NetUserGetLocalGroups(serverName *uint16, userName *uint16, level uint32, flags uint32, buf **byte, prefMaxLen uint32, entriesRead *uint32, totalEntries *uint32) (neterr error) = netapi32.NetUserGetLocalGroups diff --git a/src/internal/syscall/windows/symlink_windows.go b/src/internal/syscall/windows/symlink_windows.go new file mode 100644 index 0000000..b64d058 --- /dev/null +++ b/src/internal/syscall/windows/symlink_windows.go @@ -0,0 +1,39 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +import "syscall" + +const ( + ERROR_INVALID_PARAMETER syscall.Errno = 87 + + // symlink support for CreateSymbolicLink() starting with Windows 10 (1703, v10.0.14972) + SYMBOLIC_LINK_FLAG_ALLOW_UNPRIVILEGED_CREATE = 0x2 + + // FileInformationClass values + FileBasicInfo = 0 // FILE_BASIC_INFO + FileStandardInfo = 1 // FILE_STANDARD_INFO + FileNameInfo = 2 // FILE_NAME_INFO + FileStreamInfo = 7 // FILE_STREAM_INFO + FileCompressionInfo = 8 // FILE_COMPRESSION_INFO + FileAttributeTagInfo = 9 // FILE_ATTRIBUTE_TAG_INFO + FileIdBothDirectoryInfo = 0xa // FILE_ID_BOTH_DIR_INFO + FileIdBothDirectoryRestartInfo = 0xb // FILE_ID_BOTH_DIR_INFO + FileRemoteProtocolInfo = 0xd // FILE_REMOTE_PROTOCOL_INFO + FileFullDirectoryInfo = 0xe // FILE_FULL_DIR_INFO + FileFullDirectoryRestartInfo = 0xf // FILE_FULL_DIR_INFO + FileStorageInfo = 0x10 // FILE_STORAGE_INFO + FileAlignmentInfo = 0x11 // FILE_ALIGNMENT_INFO + FileIdInfo = 0x12 // FILE_ID_INFO + FileIdExtdDirectoryInfo = 0x13 // FILE_ID_EXTD_DIR_INFO + FileIdExtdDirectoryRestartInfo = 0x14 // FILE_ID_EXTD_DIR_INFO +) + +type FILE_ATTRIBUTE_TAG_INFO struct { + FileAttributes uint32 + ReparseTag uint32 +} + +//sys GetFileInformationByHandleEx(handle syscall.Handle, class uint32, info *byte, bufsize uint32) (err error) diff --git a/src/internal/syscall/windows/syscall_windows.go b/src/internal/syscall/windows/syscall_windows.go new file mode 100644 index 0000000..ab2f9a1 --- /dev/null +++ b/src/internal/syscall/windows/syscall_windows.go @@ -0,0 +1,390 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package windows + +import ( + "sync" + "syscall" + "unsafe" +) + +// UTF16PtrToString is like UTF16ToString, but takes *uint16 +// as a parameter instead of []uint16. +func UTF16PtrToString(p *uint16) string { + if p == nil { + return "" + } + end := unsafe.Pointer(p) + n := 0 + for *(*uint16)(end) != 0 { + end = unsafe.Pointer(uintptr(end) + unsafe.Sizeof(*p)) + n++ + } + return syscall.UTF16ToString(unsafe.Slice(p, n)) +} + +const ( + ERROR_BAD_LENGTH syscall.Errno = 24 + ERROR_SHARING_VIOLATION syscall.Errno = 32 + ERROR_LOCK_VIOLATION syscall.Errno = 33 + ERROR_NOT_SUPPORTED syscall.Errno = 50 + ERROR_CALL_NOT_IMPLEMENTED syscall.Errno = 120 + ERROR_INVALID_NAME syscall.Errno = 123 + ERROR_LOCK_FAILED syscall.Errno = 167 + ERROR_NO_UNICODE_TRANSLATION syscall.Errno = 1113 +) + +const GAA_FLAG_INCLUDE_PREFIX = 0x00000010 + +const ( + IF_TYPE_OTHER = 1 + IF_TYPE_ETHERNET_CSMACD = 6 + IF_TYPE_ISO88025_TOKENRING = 9 + IF_TYPE_PPP = 23 + IF_TYPE_SOFTWARE_LOOPBACK = 24 + IF_TYPE_ATM = 37 + IF_TYPE_IEEE80211 = 71 + IF_TYPE_TUNNEL = 131 + IF_TYPE_IEEE1394 = 144 +) + +type SocketAddress struct { + Sockaddr *syscall.RawSockaddrAny + SockaddrLength int32 +} + +type IpAdapterUnicastAddress struct { + Length uint32 + Flags uint32 + Next *IpAdapterUnicastAddress + Address SocketAddress + PrefixOrigin int32 + SuffixOrigin int32 + DadState int32 + ValidLifetime uint32 + PreferredLifetime uint32 + LeaseLifetime uint32 + OnLinkPrefixLength uint8 +} + +type IpAdapterAnycastAddress struct { + Length uint32 + Flags uint32 + Next *IpAdapterAnycastAddress + Address SocketAddress +} + +type IpAdapterMulticastAddress struct { + Length uint32 + Flags uint32 + Next *IpAdapterMulticastAddress + Address SocketAddress +} + +type IpAdapterDnsServerAdapter struct { + Length uint32 + Reserved uint32 + Next *IpAdapterDnsServerAdapter + Address SocketAddress +} + +type IpAdapterPrefix struct { + Length uint32 + Flags uint32 + Next *IpAdapterPrefix + Address SocketAddress + PrefixLength uint32 +} + +type IpAdapterAddresses struct { + Length uint32 + IfIndex uint32 + Next *IpAdapterAddresses + AdapterName *byte + FirstUnicastAddress *IpAdapterUnicastAddress + FirstAnycastAddress *IpAdapterAnycastAddress + FirstMulticastAddress *IpAdapterMulticastAddress + FirstDnsServerAddress *IpAdapterDnsServerAdapter + DnsSuffix *uint16 + Description *uint16 + FriendlyName *uint16 + PhysicalAddress [syscall.MAX_ADAPTER_ADDRESS_LENGTH]byte + PhysicalAddressLength uint32 + Flags uint32 + Mtu uint32 + IfType uint32 + OperStatus uint32 + Ipv6IfIndex uint32 + ZoneIndices [16]uint32 + FirstPrefix *IpAdapterPrefix + /* more fields might be present here. */ +} + +type SecurityAttributes struct { + Length uint16 + SecurityDescriptor uintptr + InheritHandle bool +} + +type FILE_BASIC_INFO struct { + CreationTime int64 + LastAccessTime int64 + LastWriteTime int64 + ChangedTime int64 + FileAttributes uint32 + + // Pad out to 8-byte alignment. + // + // Without this padding, TestChmod fails due to an argument validation error + // in SetFileInformationByHandle on windows/386. + // + // https://learn.microsoft.com/en-us/cpp/build/reference/zp-struct-member-alignment?view=msvc-170 + // says that “The C/C++ headers in the Windows SDK assume the platform's + // default alignment is used.” What we see here is padding rather than + // alignment, but maybe it is related. + _ uint32 +} + +const ( + IfOperStatusUp = 1 + IfOperStatusDown = 2 + IfOperStatusTesting = 3 + IfOperStatusUnknown = 4 + IfOperStatusDormant = 5 + IfOperStatusNotPresent = 6 + IfOperStatusLowerLayerDown = 7 +) + +//sys GetAdaptersAddresses(family uint32, flags uint32, reserved uintptr, adapterAddresses *IpAdapterAddresses, sizePointer *uint32) (errcode error) = iphlpapi.GetAdaptersAddresses +//sys GetComputerNameEx(nameformat uint32, buf *uint16, n *uint32) (err error) = GetComputerNameExW +//sys MoveFileEx(from *uint16, to *uint16, flags uint32) (err error) = MoveFileExW +//sys GetModuleFileName(module syscall.Handle, fn *uint16, len uint32) (n uint32, err error) = kernel32.GetModuleFileNameW +//sys SetFileInformationByHandle(handle syscall.Handle, fileInformationClass uint32, buf unsafe.Pointer, bufsize uint32) (err error) = kernel32.SetFileInformationByHandle +//sys VirtualQuery(address uintptr, buffer *MemoryBasicInformation, length uintptr) (err error) = kernel32.VirtualQuery +//sys GetTempPath2(buflen uint32, buf *uint16) (n uint32, err error) = GetTempPath2W + +const ( + // flags for CreateToolhelp32Snapshot + TH32CS_SNAPMODULE = 0x08 + TH32CS_SNAPMODULE32 = 0x10 +) + +const MAX_MODULE_NAME32 = 255 + +type ModuleEntry32 struct { + Size uint32 + ModuleID uint32 + ProcessID uint32 + GlblcntUsage uint32 + ProccntUsage uint32 + ModBaseAddr uintptr + ModBaseSize uint32 + ModuleHandle syscall.Handle + Module [MAX_MODULE_NAME32 + 1]uint16 + ExePath [syscall.MAX_PATH]uint16 +} + +const SizeofModuleEntry32 = unsafe.Sizeof(ModuleEntry32{}) + +//sys Module32First(snapshot syscall.Handle, moduleEntry *ModuleEntry32) (err error) = kernel32.Module32FirstW +//sys Module32Next(snapshot syscall.Handle, moduleEntry *ModuleEntry32) (err error) = kernel32.Module32NextW + +const ( + WSA_FLAG_OVERLAPPED = 0x01 + WSA_FLAG_NO_HANDLE_INHERIT = 0x80 + + WSAEMSGSIZE syscall.Errno = 10040 + + MSG_PEEK = 0x2 + MSG_TRUNC = 0x0100 + MSG_CTRUNC = 0x0200 + + socket_error = uintptr(^uint32(0)) +) + +var WSAID_WSASENDMSG = syscall.GUID{ + Data1: 0xa441e712, + Data2: 0x754f, + Data3: 0x43ca, + Data4: [8]byte{0x84, 0xa7, 0x0d, 0xee, 0x44, 0xcf, 0x60, 0x6d}, +} + +var WSAID_WSARECVMSG = syscall.GUID{ + Data1: 0xf689d7c8, + Data2: 0x6f1f, + Data3: 0x436b, + Data4: [8]byte{0x8a, 0x53, 0xe5, 0x4f, 0xe3, 0x51, 0xc3, 0x22}, +} + +var sendRecvMsgFunc struct { + once sync.Once + sendAddr uintptr + recvAddr uintptr + err error +} + +type WSAMsg struct { + Name syscall.Pointer + Namelen int32 + Buffers *syscall.WSABuf + BufferCount uint32 + Control syscall.WSABuf + Flags uint32 +} + +//sys WSASocket(af int32, typ int32, protocol int32, protinfo *syscall.WSAProtocolInfo, group uint32, flags uint32) (handle syscall.Handle, err error) [failretval==syscall.InvalidHandle] = ws2_32.WSASocketW + +func loadWSASendRecvMsg() error { + sendRecvMsgFunc.once.Do(func() { + var s syscall.Handle + s, sendRecvMsgFunc.err = syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_UDP) + if sendRecvMsgFunc.err != nil { + return + } + defer syscall.CloseHandle(s) + var n uint32 + sendRecvMsgFunc.err = syscall.WSAIoctl(s, + syscall.SIO_GET_EXTENSION_FUNCTION_POINTER, + (*byte)(unsafe.Pointer(&WSAID_WSARECVMSG)), + uint32(unsafe.Sizeof(WSAID_WSARECVMSG)), + (*byte)(unsafe.Pointer(&sendRecvMsgFunc.recvAddr)), + uint32(unsafe.Sizeof(sendRecvMsgFunc.recvAddr)), + &n, nil, 0) + if sendRecvMsgFunc.err != nil { + return + } + sendRecvMsgFunc.err = syscall.WSAIoctl(s, + syscall.SIO_GET_EXTENSION_FUNCTION_POINTER, + (*byte)(unsafe.Pointer(&WSAID_WSASENDMSG)), + uint32(unsafe.Sizeof(WSAID_WSASENDMSG)), + (*byte)(unsafe.Pointer(&sendRecvMsgFunc.sendAddr)), + uint32(unsafe.Sizeof(sendRecvMsgFunc.sendAddr)), + &n, nil, 0) + }) + return sendRecvMsgFunc.err +} + +func WSASendMsg(fd syscall.Handle, msg *WSAMsg, flags uint32, bytesSent *uint32, overlapped *syscall.Overlapped, croutine *byte) error { + err := loadWSASendRecvMsg() + if err != nil { + return err + } + r1, _, e1 := syscall.Syscall6(sendRecvMsgFunc.sendAddr, 6, uintptr(fd), uintptr(unsafe.Pointer(msg)), uintptr(flags), uintptr(unsafe.Pointer(bytesSent)), uintptr(unsafe.Pointer(overlapped)), uintptr(unsafe.Pointer(croutine))) + if r1 == socket_error { + if e1 != 0 { + err = errnoErr(e1) + } else { + err = syscall.EINVAL + } + } + return err +} + +func WSARecvMsg(fd syscall.Handle, msg *WSAMsg, bytesReceived *uint32, overlapped *syscall.Overlapped, croutine *byte) error { + err := loadWSASendRecvMsg() + if err != nil { + return err + } + r1, _, e1 := syscall.Syscall6(sendRecvMsgFunc.recvAddr, 5, uintptr(fd), uintptr(unsafe.Pointer(msg)), uintptr(unsafe.Pointer(bytesReceived)), uintptr(unsafe.Pointer(overlapped)), uintptr(unsafe.Pointer(croutine)), 0) + if r1 == socket_error { + if e1 != 0 { + err = errnoErr(e1) + } else { + err = syscall.EINVAL + } + } + return err +} + +const ( + ComputerNameNetBIOS = 0 + ComputerNameDnsHostname = 1 + ComputerNameDnsDomain = 2 + ComputerNameDnsFullyQualified = 3 + ComputerNamePhysicalNetBIOS = 4 + ComputerNamePhysicalDnsHostname = 5 + ComputerNamePhysicalDnsDomain = 6 + ComputerNamePhysicalDnsFullyQualified = 7 + ComputerNameMax = 8 + + MOVEFILE_REPLACE_EXISTING = 0x1 + MOVEFILE_COPY_ALLOWED = 0x2 + MOVEFILE_DELAY_UNTIL_REBOOT = 0x4 + MOVEFILE_WRITE_THROUGH = 0x8 + MOVEFILE_CREATE_HARDLINK = 0x10 + MOVEFILE_FAIL_IF_NOT_TRACKABLE = 0x20 +) + +func Rename(oldpath, newpath string) error { + from, err := syscall.UTF16PtrFromString(oldpath) + if err != nil { + return err + } + to, err := syscall.UTF16PtrFromString(newpath) + if err != nil { + return err + } + return MoveFileEx(from, to, MOVEFILE_REPLACE_EXISTING) +} + +//sys LockFileEx(file syscall.Handle, flags uint32, reserved uint32, bytesLow uint32, bytesHigh uint32, overlapped *syscall.Overlapped) (err error) = kernel32.LockFileEx +//sys UnlockFileEx(file syscall.Handle, reserved uint32, bytesLow uint32, bytesHigh uint32, overlapped *syscall.Overlapped) (err error) = kernel32.UnlockFileEx + +const ( + LOCKFILE_FAIL_IMMEDIATELY = 0x00000001 + LOCKFILE_EXCLUSIVE_LOCK = 0x00000002 +) + +const MB_ERR_INVALID_CHARS = 8 + +//sys GetACP() (acp uint32) = kernel32.GetACP +//sys GetConsoleCP() (ccp uint32) = kernel32.GetConsoleCP +//sys MultiByteToWideChar(codePage uint32, dwFlags uint32, str *byte, nstr int32, wchar *uint16, nwchar int32) (nwrite int32, err error) = kernel32.MultiByteToWideChar +//sys GetCurrentThread() (pseudoHandle syscall.Handle, err error) = kernel32.GetCurrentThread + +const STYPE_DISKTREE = 0x00 + +type SHARE_INFO_2 struct { + Netname *uint16 + Type uint32 + Remark *uint16 + Permissions uint32 + MaxUses uint32 + CurrentUses uint32 + Path *uint16 + Passwd *uint16 +} + +//sys NetShareAdd(serverName *uint16, level uint32, buf *byte, parmErr *uint16) (neterr error) = netapi32.NetShareAdd +//sys NetShareDel(serverName *uint16, netName *uint16, reserved uint32) (neterr error) = netapi32.NetShareDel + +const ( + FILE_NAME_NORMALIZED = 0x0 + FILE_NAME_OPENED = 0x8 + + VOLUME_NAME_DOS = 0x0 + VOLUME_NAME_GUID = 0x1 + VOLUME_NAME_NONE = 0x4 + VOLUME_NAME_NT = 0x2 +) + +//sys GetFinalPathNameByHandle(file syscall.Handle, filePath *uint16, filePathSize uint32, flags uint32) (n uint32, err error) = kernel32.GetFinalPathNameByHandleW + +func LoadGetFinalPathNameByHandle() error { + return procGetFinalPathNameByHandleW.Find() +} + +func ErrorLoadingGetTempPath2() error { + return procGetTempPath2W.Find() +} + +//sys CreateEnvironmentBlock(block **uint16, token syscall.Token, inheritExisting bool) (err error) = userenv.CreateEnvironmentBlock +//sys DestroyEnvironmentBlock(block *uint16) (err error) = userenv.DestroyEnvironmentBlock +//sys CreateEvent(eventAttrs *SecurityAttributes, manualReset uint32, initialState uint32, name *uint16) (handle syscall.Handle, err error) = kernel32.CreateEventW + +//sys ProcessPrng(buf []byte) (err error) = bcryptprimitives.ProcessPrng + +//sys RtlLookupFunctionEntry(pc uintptr, baseAddress *uintptr, table *byte) (ret uintptr) = kernel32.RtlLookupFunctionEntry +//sys RtlVirtualUnwind(handlerType uint32, baseAddress uintptr, pc uintptr, entry uintptr, ctxt uintptr, data *uintptr, frame *uintptr, ctxptrs *byte) (ret uintptr) = kernel32.RtlVirtualUnwind diff --git a/src/internal/syscall/windows/sysdll/sysdll.go b/src/internal/syscall/windows/sysdll/sysdll.go new file mode 100644 index 0000000..e79fd19 --- /dev/null +++ b/src/internal/syscall/windows/sysdll/sysdll.go @@ -0,0 +1,30 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows + +// Package sysdll is an internal leaf package that records and reports +// which Windows DLL names are used by Go itself. These DLLs are then +// only loaded from the System32 directory. See Issue 14959. +package sysdll + +// IsSystemDLL reports whether the named dll key (a base name, like +// "foo.dll") is a system DLL which should only be loaded from the +// Windows SYSTEM32 directory. +// +// Filenames are case sensitive, but that doesn't matter because +// the case registered with Add is also the same case used with +// LoadDLL later. +// +// It has no associated mutex and should only be mutated serially +// (currently: during init), and not concurrent with DLL loading. +var IsSystemDLL = map[string]bool{} + +// Add notes that dll is a system32 DLL which should only be loaded +// from the Windows SYSTEM32 directory. It returns its argument back, +// for ease of use in generated code. +func Add(dll string) string { + IsSystemDLL[dll] = true + return dll +} diff --git a/src/internal/syscall/windows/zsyscall_windows.go b/src/internal/syscall/windows/zsyscall_windows.go new file mode 100644 index 0000000..6be7aa4 --- /dev/null +++ b/src/internal/syscall/windows/zsyscall_windows.go @@ -0,0 +1,389 @@ +// Code generated by 'go generate'; DO NOT EDIT. + +package windows + +import ( + "internal/syscall/windows/sysdll" + "syscall" + "unsafe" +) + +var _ unsafe.Pointer + +// Do the interface allocations only once for common +// Errno values. +const ( + errnoERROR_IO_PENDING = 997 +) + +var ( + errERROR_IO_PENDING error = syscall.Errno(errnoERROR_IO_PENDING) + errERROR_EINVAL error = syscall.EINVAL +) + +// errnoErr returns common boxed Errno values, to prevent +// allocations at runtime. +func errnoErr(e syscall.Errno) error { + switch e { + case 0: + return errERROR_EINVAL + case errnoERROR_IO_PENDING: + return errERROR_IO_PENDING + } + // TODO: add more here, after collecting data on the common + // error values see on Windows. (perhaps when running + // all.bat?) + return e +} + +var ( + modadvapi32 = syscall.NewLazyDLL(sysdll.Add("advapi32.dll")) + modbcryptprimitives = syscall.NewLazyDLL(sysdll.Add("bcryptprimitives.dll")) + modiphlpapi = syscall.NewLazyDLL(sysdll.Add("iphlpapi.dll")) + modkernel32 = syscall.NewLazyDLL(sysdll.Add("kernel32.dll")) + modnetapi32 = syscall.NewLazyDLL(sysdll.Add("netapi32.dll")) + modpsapi = syscall.NewLazyDLL(sysdll.Add("psapi.dll")) + moduserenv = syscall.NewLazyDLL(sysdll.Add("userenv.dll")) + modws2_32 = syscall.NewLazyDLL(sysdll.Add("ws2_32.dll")) + + procAdjustTokenPrivileges = modadvapi32.NewProc("AdjustTokenPrivileges") + procDuplicateTokenEx = modadvapi32.NewProc("DuplicateTokenEx") + procImpersonateSelf = modadvapi32.NewProc("ImpersonateSelf") + procLookupPrivilegeValueW = modadvapi32.NewProc("LookupPrivilegeValueW") + procOpenThreadToken = modadvapi32.NewProc("OpenThreadToken") + procRevertToSelf = modadvapi32.NewProc("RevertToSelf") + procSetTokenInformation = modadvapi32.NewProc("SetTokenInformation") + procProcessPrng = modbcryptprimitives.NewProc("ProcessPrng") + procGetAdaptersAddresses = modiphlpapi.NewProc("GetAdaptersAddresses") + procCreateEventW = modkernel32.NewProc("CreateEventW") + procGetACP = modkernel32.NewProc("GetACP") + procGetComputerNameExW = modkernel32.NewProc("GetComputerNameExW") + procGetConsoleCP = modkernel32.NewProc("GetConsoleCP") + procGetCurrentThread = modkernel32.NewProc("GetCurrentThread") + procGetFileInformationByHandleEx = modkernel32.NewProc("GetFileInformationByHandleEx") + procGetFinalPathNameByHandleW = modkernel32.NewProc("GetFinalPathNameByHandleW") + procGetModuleFileNameW = modkernel32.NewProc("GetModuleFileNameW") + procGetTempPath2W = modkernel32.NewProc("GetTempPath2W") + procLockFileEx = modkernel32.NewProc("LockFileEx") + procModule32FirstW = modkernel32.NewProc("Module32FirstW") + procModule32NextW = modkernel32.NewProc("Module32NextW") + procMoveFileExW = modkernel32.NewProc("MoveFileExW") + procMultiByteToWideChar = modkernel32.NewProc("MultiByteToWideChar") + procRtlLookupFunctionEntry = modkernel32.NewProc("RtlLookupFunctionEntry") + procRtlVirtualUnwind = modkernel32.NewProc("RtlVirtualUnwind") + procSetFileInformationByHandle = modkernel32.NewProc("SetFileInformationByHandle") + procUnlockFileEx = modkernel32.NewProc("UnlockFileEx") + procVirtualQuery = modkernel32.NewProc("VirtualQuery") + procNetShareAdd = modnetapi32.NewProc("NetShareAdd") + procNetShareDel = modnetapi32.NewProc("NetShareDel") + procNetUserGetLocalGroups = modnetapi32.NewProc("NetUserGetLocalGroups") + procGetProcessMemoryInfo = modpsapi.NewProc("GetProcessMemoryInfo") + procCreateEnvironmentBlock = moduserenv.NewProc("CreateEnvironmentBlock") + procDestroyEnvironmentBlock = moduserenv.NewProc("DestroyEnvironmentBlock") + procGetProfilesDirectoryW = moduserenv.NewProc("GetProfilesDirectoryW") + procWSASocketW = modws2_32.NewProc("WSASocketW") +) + +func adjustTokenPrivileges(token syscall.Token, disableAllPrivileges bool, newstate *TOKEN_PRIVILEGES, buflen uint32, prevstate *TOKEN_PRIVILEGES, returnlen *uint32) (ret uint32, err error) { + var _p0 uint32 + if disableAllPrivileges { + _p0 = 1 + } + r0, _, e1 := syscall.Syscall6(procAdjustTokenPrivileges.Addr(), 6, uintptr(token), uintptr(_p0), uintptr(unsafe.Pointer(newstate)), uintptr(buflen), uintptr(unsafe.Pointer(prevstate)), uintptr(unsafe.Pointer(returnlen))) + ret = uint32(r0) + if true { + err = errnoErr(e1) + } + return +} + +func DuplicateTokenEx(hExistingToken syscall.Token, dwDesiredAccess uint32, lpTokenAttributes *syscall.SecurityAttributes, impersonationLevel uint32, tokenType TokenType, phNewToken *syscall.Token) (err error) { + r1, _, e1 := syscall.Syscall6(procDuplicateTokenEx.Addr(), 6, uintptr(hExistingToken), uintptr(dwDesiredAccess), uintptr(unsafe.Pointer(lpTokenAttributes)), uintptr(impersonationLevel), uintptr(tokenType), uintptr(unsafe.Pointer(phNewToken))) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func ImpersonateSelf(impersonationlevel uint32) (err error) { + r1, _, e1 := syscall.Syscall(procImpersonateSelf.Addr(), 1, uintptr(impersonationlevel), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func LookupPrivilegeValue(systemname *uint16, name *uint16, luid *LUID) (err error) { + r1, _, e1 := syscall.Syscall(procLookupPrivilegeValueW.Addr(), 3, uintptr(unsafe.Pointer(systemname)), uintptr(unsafe.Pointer(name)), uintptr(unsafe.Pointer(luid))) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func OpenThreadToken(h syscall.Handle, access uint32, openasself bool, token *syscall.Token) (err error) { + var _p0 uint32 + if openasself { + _p0 = 1 + } + r1, _, e1 := syscall.Syscall6(procOpenThreadToken.Addr(), 4, uintptr(h), uintptr(access), uintptr(_p0), uintptr(unsafe.Pointer(token)), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func RevertToSelf() (err error) { + r1, _, e1 := syscall.Syscall(procRevertToSelf.Addr(), 0, 0, 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func SetTokenInformation(tokenHandle syscall.Token, tokenInformationClass uint32, tokenInformation uintptr, tokenInformationLength uint32) (err error) { + r1, _, e1 := syscall.Syscall6(procSetTokenInformation.Addr(), 4, uintptr(tokenHandle), uintptr(tokenInformationClass), uintptr(tokenInformation), uintptr(tokenInformationLength), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func ProcessPrng(buf []byte) (err error) { + var _p0 *byte + if len(buf) > 0 { + _p0 = &buf[0] + } + r1, _, e1 := syscall.Syscall(procProcessPrng.Addr(), 2, uintptr(unsafe.Pointer(_p0)), uintptr(len(buf)), 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func GetAdaptersAddresses(family uint32, flags uint32, reserved uintptr, adapterAddresses *IpAdapterAddresses, sizePointer *uint32) (errcode error) { + r0, _, _ := syscall.Syscall6(procGetAdaptersAddresses.Addr(), 5, uintptr(family), uintptr(flags), uintptr(reserved), uintptr(unsafe.Pointer(adapterAddresses)), uintptr(unsafe.Pointer(sizePointer)), 0) + if r0 != 0 { + errcode = syscall.Errno(r0) + } + return +} + +func CreateEvent(eventAttrs *SecurityAttributes, manualReset uint32, initialState uint32, name *uint16) (handle syscall.Handle, err error) { + r0, _, e1 := syscall.Syscall6(procCreateEventW.Addr(), 4, uintptr(unsafe.Pointer(eventAttrs)), uintptr(manualReset), uintptr(initialState), uintptr(unsafe.Pointer(name)), 0, 0) + handle = syscall.Handle(r0) + if handle == 0 { + err = errnoErr(e1) + } + return +} + +func GetACP() (acp uint32) { + r0, _, _ := syscall.Syscall(procGetACP.Addr(), 0, 0, 0, 0) + acp = uint32(r0) + return +} + +func GetComputerNameEx(nameformat uint32, buf *uint16, n *uint32) (err error) { + r1, _, e1 := syscall.Syscall(procGetComputerNameExW.Addr(), 3, uintptr(nameformat), uintptr(unsafe.Pointer(buf)), uintptr(unsafe.Pointer(n))) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func GetConsoleCP() (ccp uint32) { + r0, _, _ := syscall.Syscall(procGetConsoleCP.Addr(), 0, 0, 0, 0) + ccp = uint32(r0) + return +} + +func GetCurrentThread() (pseudoHandle syscall.Handle, err error) { + r0, _, e1 := syscall.Syscall(procGetCurrentThread.Addr(), 0, 0, 0, 0) + pseudoHandle = syscall.Handle(r0) + if pseudoHandle == 0 { + err = errnoErr(e1) + } + return +} + +func GetFileInformationByHandleEx(handle syscall.Handle, class uint32, info *byte, bufsize uint32) (err error) { + r1, _, e1 := syscall.Syscall6(procGetFileInformationByHandleEx.Addr(), 4, uintptr(handle), uintptr(class), uintptr(unsafe.Pointer(info)), uintptr(bufsize), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func GetFinalPathNameByHandle(file syscall.Handle, filePath *uint16, filePathSize uint32, flags uint32) (n uint32, err error) { + r0, _, e1 := syscall.Syscall6(procGetFinalPathNameByHandleW.Addr(), 4, uintptr(file), uintptr(unsafe.Pointer(filePath)), uintptr(filePathSize), uintptr(flags), 0, 0) + n = uint32(r0) + if n == 0 { + err = errnoErr(e1) + } + return +} + +func GetModuleFileName(module syscall.Handle, fn *uint16, len uint32) (n uint32, err error) { + r0, _, e1 := syscall.Syscall(procGetModuleFileNameW.Addr(), 3, uintptr(module), uintptr(unsafe.Pointer(fn)), uintptr(len)) + n = uint32(r0) + if n == 0 { + err = errnoErr(e1) + } + return +} + +func GetTempPath2(buflen uint32, buf *uint16) (n uint32, err error) { + r0, _, e1 := syscall.Syscall(procGetTempPath2W.Addr(), 2, uintptr(buflen), uintptr(unsafe.Pointer(buf)), 0) + n = uint32(r0) + if n == 0 { + err = errnoErr(e1) + } + return +} + +func LockFileEx(file syscall.Handle, flags uint32, reserved uint32, bytesLow uint32, bytesHigh uint32, overlapped *syscall.Overlapped) (err error) { + r1, _, e1 := syscall.Syscall6(procLockFileEx.Addr(), 6, uintptr(file), uintptr(flags), uintptr(reserved), uintptr(bytesLow), uintptr(bytesHigh), uintptr(unsafe.Pointer(overlapped))) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func Module32First(snapshot syscall.Handle, moduleEntry *ModuleEntry32) (err error) { + r1, _, e1 := syscall.Syscall(procModule32FirstW.Addr(), 2, uintptr(snapshot), uintptr(unsafe.Pointer(moduleEntry)), 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func Module32Next(snapshot syscall.Handle, moduleEntry *ModuleEntry32) (err error) { + r1, _, e1 := syscall.Syscall(procModule32NextW.Addr(), 2, uintptr(snapshot), uintptr(unsafe.Pointer(moduleEntry)), 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func MoveFileEx(from *uint16, to *uint16, flags uint32) (err error) { + r1, _, e1 := syscall.Syscall(procMoveFileExW.Addr(), 3, uintptr(unsafe.Pointer(from)), uintptr(unsafe.Pointer(to)), uintptr(flags)) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func MultiByteToWideChar(codePage uint32, dwFlags uint32, str *byte, nstr int32, wchar *uint16, nwchar int32) (nwrite int32, err error) { + r0, _, e1 := syscall.Syscall6(procMultiByteToWideChar.Addr(), 6, uintptr(codePage), uintptr(dwFlags), uintptr(unsafe.Pointer(str)), uintptr(nstr), uintptr(unsafe.Pointer(wchar)), uintptr(nwchar)) + nwrite = int32(r0) + if nwrite == 0 { + err = errnoErr(e1) + } + return +} + +func RtlLookupFunctionEntry(pc uintptr, baseAddress *uintptr, table *byte) (ret uintptr) { + r0, _, _ := syscall.Syscall(procRtlLookupFunctionEntry.Addr(), 3, uintptr(pc), uintptr(unsafe.Pointer(baseAddress)), uintptr(unsafe.Pointer(table))) + ret = uintptr(r0) + return +} + +func RtlVirtualUnwind(handlerType uint32, baseAddress uintptr, pc uintptr, entry uintptr, ctxt uintptr, data *uintptr, frame *uintptr, ctxptrs *byte) (ret uintptr) { + r0, _, _ := syscall.Syscall9(procRtlVirtualUnwind.Addr(), 8, uintptr(handlerType), uintptr(baseAddress), uintptr(pc), uintptr(entry), uintptr(ctxt), uintptr(unsafe.Pointer(data)), uintptr(unsafe.Pointer(frame)), uintptr(unsafe.Pointer(ctxptrs)), 0) + ret = uintptr(r0) + return +} + +func SetFileInformationByHandle(handle syscall.Handle, fileInformationClass uint32, buf unsafe.Pointer, bufsize uint32) (err error) { + r1, _, e1 := syscall.Syscall6(procSetFileInformationByHandle.Addr(), 4, uintptr(handle), uintptr(fileInformationClass), uintptr(buf), uintptr(bufsize), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func UnlockFileEx(file syscall.Handle, reserved uint32, bytesLow uint32, bytesHigh uint32, overlapped *syscall.Overlapped) (err error) { + r1, _, e1 := syscall.Syscall6(procUnlockFileEx.Addr(), 5, uintptr(file), uintptr(reserved), uintptr(bytesLow), uintptr(bytesHigh), uintptr(unsafe.Pointer(overlapped)), 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func VirtualQuery(address uintptr, buffer *MemoryBasicInformation, length uintptr) (err error) { + r1, _, e1 := syscall.Syscall(procVirtualQuery.Addr(), 3, uintptr(address), uintptr(unsafe.Pointer(buffer)), uintptr(length)) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func NetShareAdd(serverName *uint16, level uint32, buf *byte, parmErr *uint16) (neterr error) { + r0, _, _ := syscall.Syscall6(procNetShareAdd.Addr(), 4, uintptr(unsafe.Pointer(serverName)), uintptr(level), uintptr(unsafe.Pointer(buf)), uintptr(unsafe.Pointer(parmErr)), 0, 0) + if r0 != 0 { + neterr = syscall.Errno(r0) + } + return +} + +func NetShareDel(serverName *uint16, netName *uint16, reserved uint32) (neterr error) { + r0, _, _ := syscall.Syscall(procNetShareDel.Addr(), 3, uintptr(unsafe.Pointer(serverName)), uintptr(unsafe.Pointer(netName)), uintptr(reserved)) + if r0 != 0 { + neterr = syscall.Errno(r0) + } + return +} + +func NetUserGetLocalGroups(serverName *uint16, userName *uint16, level uint32, flags uint32, buf **byte, prefMaxLen uint32, entriesRead *uint32, totalEntries *uint32) (neterr error) { + r0, _, _ := syscall.Syscall9(procNetUserGetLocalGroups.Addr(), 8, uintptr(unsafe.Pointer(serverName)), uintptr(unsafe.Pointer(userName)), uintptr(level), uintptr(flags), uintptr(unsafe.Pointer(buf)), uintptr(prefMaxLen), uintptr(unsafe.Pointer(entriesRead)), uintptr(unsafe.Pointer(totalEntries)), 0) + if r0 != 0 { + neterr = syscall.Errno(r0) + } + return +} + +func GetProcessMemoryInfo(handle syscall.Handle, memCounters *PROCESS_MEMORY_COUNTERS, cb uint32) (err error) { + r1, _, e1 := syscall.Syscall(procGetProcessMemoryInfo.Addr(), 3, uintptr(handle), uintptr(unsafe.Pointer(memCounters)), uintptr(cb)) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func CreateEnvironmentBlock(block **uint16, token syscall.Token, inheritExisting bool) (err error) { + var _p0 uint32 + if inheritExisting { + _p0 = 1 + } + r1, _, e1 := syscall.Syscall(procCreateEnvironmentBlock.Addr(), 3, uintptr(unsafe.Pointer(block)), uintptr(token), uintptr(_p0)) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func DestroyEnvironmentBlock(block *uint16) (err error) { + r1, _, e1 := syscall.Syscall(procDestroyEnvironmentBlock.Addr(), 1, uintptr(unsafe.Pointer(block)), 0, 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func GetProfilesDirectory(dir *uint16, dirLen *uint32) (err error) { + r1, _, e1 := syscall.Syscall(procGetProfilesDirectoryW.Addr(), 2, uintptr(unsafe.Pointer(dir)), uintptr(unsafe.Pointer(dirLen)), 0) + if r1 == 0 { + err = errnoErr(e1) + } + return +} + +func WSASocket(af int32, typ int32, protocol int32, protinfo *syscall.WSAProtocolInfo, group uint32, flags uint32) (handle syscall.Handle, err error) { + r0, _, e1 := syscall.Syscall6(procWSASocketW.Addr(), 6, uintptr(af), uintptr(typ), uintptr(protocol), uintptr(unsafe.Pointer(protinfo)), uintptr(group), uintptr(flags)) + handle = syscall.Handle(r0) + if handle == syscall.InvalidHandle { + err = errnoErr(e1) + } + return +} diff --git a/src/internal/sysinfo/sysinfo.go b/src/internal/sysinfo/sysinfo.go new file mode 100644 index 0000000..961be7a --- /dev/null +++ b/src/internal/sysinfo/sysinfo.go @@ -0,0 +1,31 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package sysinfo implements high level hardware information gathering +// that can be used for debugging or information purposes. +package sysinfo + +import ( + internalcpu "internal/cpu" + "sync" +) + +type cpuInfo struct { + once sync.Once + name string +} + +var CPU cpuInfo + +func (cpu *cpuInfo) Name() string { + cpu.once.Do(func() { + // Try to get the information from internal/cpu. + if name := internalcpu.Name(); name != "" { + cpu.name = name + return + } + // TODO(martisch): use /proc/cpuinfo and /sys/devices/system/cpu/ on Linux as fallback. + }) + return cpu.name +} diff --git a/src/internal/testenv/exec.go b/src/internal/testenv/exec.go new file mode 100644 index 0000000..50d3b0d --- /dev/null +++ b/src/internal/testenv/exec.go @@ -0,0 +1,219 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package testenv + +import ( + "context" + "errors" + "fmt" + "os" + "os/exec" + "runtime" + "strconv" + "strings" + "sync" + "testing" + "time" +) + +// MustHaveExec checks that the current system can start new processes +// using os.StartProcess or (more commonly) exec.Command. +// If not, MustHaveExec calls t.Skip with an explanation. +// +// On some platforms MustHaveExec checks for exec support by re-executing the +// current executable, which must be a binary built by 'go test'. +// We intentionally do not provide a HasExec function because of the risk of +// inappropriate recursion in TestMain functions. +// +// To check for exec support outside of a test, just try to exec the command. +// If exec is not supported, testenv.SyscallIsNotSupported will return true +// for the resulting error. +func MustHaveExec(t testing.TB) { + tryExecOnce.Do(func() { + tryExecErr = tryExec() + }) + if tryExecErr != nil { + t.Skipf("skipping test: cannot exec subprocess on %s/%s: %v", runtime.GOOS, runtime.GOARCH, tryExecErr) + } +} + +var ( + tryExecOnce sync.Once + tryExecErr error +) + +func tryExec() error { + switch runtime.GOOS { + case "wasip1", "js", "ios": + default: + // Assume that exec always works on non-mobile platforms and Android. + return nil + } + + // ios has an exec syscall but on real iOS devices it might return a + // permission error. In an emulated environment (such as a Corellium host) + // it might succeed, so if we need to exec we'll just have to try it and + // find out. + // + // As of 2023-04-19 wasip1 and js don't have exec syscalls at all, but we + // may as well use the same path so that this branch can be tested without + // an ios environment. + + if !testing.Testing() { + // This isn't a standard 'go test' binary, so we don't know how to + // self-exec in a way that should succeed without side effects. + // Just forget it. + return errors.New("can't probe for exec support with a non-test executable") + } + + // We know that this is a test executable. We should be able to run it with a + // no-op flag to check for overall exec support. + exe, err := os.Executable() + if err != nil { + return fmt.Errorf("can't probe for exec support: %w", err) + } + cmd := exec.Command(exe, "-test.list=^$") + cmd.Env = origEnv + return cmd.Run() +} + +var execPaths sync.Map // path -> error + +// MustHaveExecPath checks that the current system can start the named executable +// using os.StartProcess or (more commonly) exec.Command. +// If not, MustHaveExecPath calls t.Skip with an explanation. +func MustHaveExecPath(t testing.TB, path string) { + MustHaveExec(t) + + err, found := execPaths.Load(path) + if !found { + _, err = exec.LookPath(path) + err, _ = execPaths.LoadOrStore(path, err) + } + if err != nil { + t.Skipf("skipping test: %s: %s", path, err) + } +} + +// CleanCmdEnv will fill cmd.Env with the environment, excluding certain +// variables that could modify the behavior of the Go tools such as +// GODEBUG and GOTRACEBACK. +func CleanCmdEnv(cmd *exec.Cmd) *exec.Cmd { + if cmd.Env != nil { + panic("environment already set") + } + for _, env := range os.Environ() { + // Exclude GODEBUG from the environment to prevent its output + // from breaking tests that are trying to parse other command output. + if strings.HasPrefix(env, "GODEBUG=") { + continue + } + // Exclude GOTRACEBACK for the same reason. + if strings.HasPrefix(env, "GOTRACEBACK=") { + continue + } + cmd.Env = append(cmd.Env, env) + } + return cmd +} + +// CommandContext is like exec.CommandContext, but: +// - skips t if the platform does not support os/exec, +// - sends SIGQUIT (if supported by the platform) instead of SIGKILL +// in its Cancel function +// - if the test has a deadline, adds a Context timeout and WaitDelay +// for an arbitrary grace period before the test's deadline expires, +// - fails the test if the command does not complete before the test's deadline, and +// - sets a Cleanup function that verifies that the test did not leak a subprocess. +func CommandContext(t testing.TB, ctx context.Context, name string, args ...string) *exec.Cmd { + t.Helper() + MustHaveExec(t) + + var ( + cancelCtx context.CancelFunc + gracePeriod time.Duration // unlimited unless the test has a deadline (to allow for interactive debugging) + ) + + if t, ok := t.(interface { + testing.TB + Deadline() (time.Time, bool) + }); ok { + if td, ok := t.Deadline(); ok { + // Start with a minimum grace period, just long enough to consume the + // output of a reasonable program after it terminates. + gracePeriod = 100 * time.Millisecond + if s := os.Getenv("GO_TEST_TIMEOUT_SCALE"); s != "" { + scale, err := strconv.Atoi(s) + if err != nil { + t.Fatalf("invalid GO_TEST_TIMEOUT_SCALE: %v", err) + } + gracePeriod *= time.Duration(scale) + } + + // If time allows, increase the termination grace period to 5% of the + // test's remaining time. + testTimeout := time.Until(td) + if gp := testTimeout / 20; gp > gracePeriod { + gracePeriod = gp + } + + // When we run commands that execute subprocesses, we want to reserve two + // grace periods to clean up: one for the delay between the first + // termination signal being sent (via the Cancel callback when the Context + // expires) and the process being forcibly terminated (via the WaitDelay + // field), and a second one for the delay between the process being + // terminated and the test logging its output for debugging. + // + // (We want to ensure that the test process itself has enough time to + // log the output before it is also terminated.) + cmdTimeout := testTimeout - 2*gracePeriod + + if cd, ok := ctx.Deadline(); !ok || time.Until(cd) > cmdTimeout { + // Either ctx doesn't have a deadline, or its deadline would expire + // after (or too close before) the test has already timed out. + // Add a shorter timeout so that the test will produce useful output. + ctx, cancelCtx = context.WithTimeout(ctx, cmdTimeout) + } + } + } + + cmd := exec.CommandContext(ctx, name, args...) + cmd.Cancel = func() error { + if cancelCtx != nil && ctx.Err() == context.DeadlineExceeded { + // The command timed out due to running too close to the test's deadline. + // There is no way the test did that intentionally — it's too close to the + // wire! — so mark it as a test failure. That way, if the test expects the + // command to fail for some other reason, it doesn't have to distinguish + // between that reason and a timeout. + t.Errorf("test timed out while running command: %v", cmd) + } else { + // The command is being terminated due to ctx being canceled, but + // apparently not due to an explicit test deadline that we added. + // Log that information in case it is useful for diagnosing a failure, + // but don't actually fail the test because of it. + t.Logf("%v: terminating command: %v", ctx.Err(), cmd) + } + return cmd.Process.Signal(Sigquit) + } + cmd.WaitDelay = gracePeriod + + t.Cleanup(func() { + if cancelCtx != nil { + cancelCtx() + } + if cmd.Process != nil && cmd.ProcessState == nil { + t.Errorf("command was started, but test did not wait for it to complete: %v", cmd) + } + }) + + return cmd +} + +// Command is like exec.Command, but applies the same changes as +// testenv.CommandContext (with a default Context). +func Command(t testing.TB, name string, args ...string) *exec.Cmd { + t.Helper() + return CommandContext(t, context.Background(), name, args...) +} diff --git a/src/internal/testenv/noopt.go b/src/internal/testenv/noopt.go new file mode 100644 index 0000000..ae2a3d0 --- /dev/null +++ b/src/internal/testenv/noopt.go @@ -0,0 +1,12 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build noopt + +package testenv + +// OptimizationOff reports whether optimization is disabled. +func OptimizationOff() bool { + return true +} diff --git a/src/internal/testenv/opt.go b/src/internal/testenv/opt.go new file mode 100644 index 0000000..1bb96f7 --- /dev/null +++ b/src/internal/testenv/opt.go @@ -0,0 +1,12 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !noopt + +package testenv + +// OptimizationOff reports whether optimization is disabled. +func OptimizationOff() bool { + return false +} diff --git a/src/internal/testenv/testenv.go b/src/internal/testenv/testenv.go new file mode 100644 index 0000000..31b58dd --- /dev/null +++ b/src/internal/testenv/testenv.go @@ -0,0 +1,506 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package testenv provides information about what functionality +// is available in different testing environments run by the Go team. +// +// It is an internal package because these details are specific +// to the Go team's test setup (on build.golang.org) and not +// fundamental to tests in general. +package testenv + +import ( + "bytes" + "errors" + "flag" + "fmt" + "internal/cfg" + "internal/platform" + "os" + "os/exec" + "path/filepath" + "runtime" + "strconv" + "strings" + "sync" + "testing" +) + +// Save the original environment during init for use in checks. A test +// binary may modify its environment before calling HasExec to change its +// behavior (such as mimicking a command-line tool), and that modified +// environment might cause environment checks to behave erratically. +var origEnv = os.Environ() + +// Builder reports the name of the builder running this test +// (for example, "linux-amd64" or "windows-386-gce"). +// If the test is not running on the build infrastructure, +// Builder returns the empty string. +func Builder() string { + return os.Getenv("GO_BUILDER_NAME") +} + +// HasGoBuild reports whether the current system can build programs with “go build” +// and then run them with os.StartProcess or exec.Command. +func HasGoBuild() bool { + if os.Getenv("GO_GCFLAGS") != "" { + // It's too much work to require every caller of the go command + // to pass along "-gcflags="+os.Getenv("GO_GCFLAGS"). + // For now, if $GO_GCFLAGS is set, report that we simply can't + // run go build. + return false + } + + goBuildOnce.Do(func() { + // To run 'go build', we need to be able to exec a 'go' command. + // We somewhat arbitrarily choose to exec 'go tool -n compile' because that + // also confirms that cmd/go can find the compiler. (Before CL 472096, + // we sometimes ended up with cmd/go installed in the test environment + // without a cmd/compile it could use to actually build things.) + cmd := exec.Command("go", "tool", "-n", "compile") + cmd.Env = origEnv + out, err := cmd.Output() + if err != nil { + goBuildErr = fmt.Errorf("%v: %w", cmd, err) + return + } + out = bytes.TrimSpace(out) + if len(out) == 0 { + goBuildErr = fmt.Errorf("%v: no tool reported", cmd) + return + } + if _, err := exec.LookPath(string(out)); err != nil { + goBuildErr = err + return + } + + if platform.MustLinkExternal(runtime.GOOS, runtime.GOARCH, false) { + // We can assume that we always have a complete Go toolchain available. + // However, this platform requires a C linker to build even pure Go + // programs, including tests. Do we have one in the test environment? + // (On Android, for example, the device running the test might not have a + // C toolchain installed.) + // + // If CC is set explicitly, assume that we do. Otherwise, use 'go env CC' + // to determine which toolchain it would use by default. + if os.Getenv("CC") == "" { + cmd := exec.Command("go", "env", "CC") + cmd.Env = origEnv + out, err := cmd.Output() + if err != nil { + goBuildErr = fmt.Errorf("%v: %w", cmd, err) + return + } + out = bytes.TrimSpace(out) + if len(out) == 0 { + goBuildErr = fmt.Errorf("%v: no CC reported", cmd) + return + } + _, goBuildErr = exec.LookPath(string(out)) + } + } + }) + + return goBuildErr == nil +} + +var ( + goBuildOnce sync.Once + goBuildErr error +) + +// MustHaveGoBuild checks that the current system can build programs with “go build” +// and then run them with os.StartProcess or exec.Command. +// If not, MustHaveGoBuild calls t.Skip with an explanation. +func MustHaveGoBuild(t testing.TB) { + if os.Getenv("GO_GCFLAGS") != "" { + t.Helper() + t.Skipf("skipping test: 'go build' not compatible with setting $GO_GCFLAGS") + } + if !HasGoBuild() { + t.Helper() + t.Skipf("skipping test: 'go build' unavailable: %v", goBuildErr) + } +} + +// HasGoRun reports whether the current system can run programs with “go run.” +func HasGoRun() bool { + // For now, having go run and having go build are the same. + return HasGoBuild() +} + +// MustHaveGoRun checks that the current system can run programs with “go run.” +// If not, MustHaveGoRun calls t.Skip with an explanation. +func MustHaveGoRun(t testing.TB) { + if !HasGoRun() { + t.Skipf("skipping test: 'go run' not available on %s/%s", runtime.GOOS, runtime.GOARCH) + } +} + +// HasParallelism reports whether the current system can execute multiple +// threads in parallel. +// There is a copy of this function in cmd/dist/test.go. +func HasParallelism() bool { + switch runtime.GOOS { + case "js", "wasip1": + return false + } + return true +} + +// MustHaveParallelism checks that the current system can execute multiple +// threads in parallel. If not, MustHaveParallelism calls t.Skip with an explanation. +func MustHaveParallelism(t testing.TB) { + if !HasParallelism() { + t.Skipf("skipping test: no parallelism available on %s/%s", runtime.GOOS, runtime.GOARCH) + } +} + +// GoToolPath reports the path to the Go tool. +// It is a convenience wrapper around GoTool. +// If the tool is unavailable GoToolPath calls t.Skip. +// If the tool should be available and isn't, GoToolPath calls t.Fatal. +func GoToolPath(t testing.TB) string { + MustHaveGoBuild(t) + path, err := GoTool() + if err != nil { + t.Fatal(err) + } + // Add all environment variables that affect the Go command to test metadata. + // Cached test results will be invalidate when these variables change. + // See golang.org/issue/32285. + for _, envVar := range strings.Fields(cfg.KnownEnv) { + os.Getenv(envVar) + } + return path +} + +var ( + gorootOnce sync.Once + gorootPath string + gorootErr error +) + +func findGOROOT() (string, error) { + gorootOnce.Do(func() { + gorootPath = runtime.GOROOT() + if gorootPath != "" { + // If runtime.GOROOT() is non-empty, assume that it is valid. + // + // (It might not be: for example, the user may have explicitly set GOROOT + // to the wrong directory, or explicitly set GOROOT_FINAL but not GOROOT + // and hasn't moved the tree to GOROOT_FINAL yet. But those cases are + // rare, and if that happens the user can fix what they broke.) + return + } + + // runtime.GOROOT doesn't know where GOROOT is (perhaps because the test + // binary was built with -trimpath, or perhaps because GOROOT_FINAL was set + // without GOROOT and the tree hasn't been moved there yet). + // + // Since this is internal/testenv, we can cheat and assume that the caller + // is a test of some package in a subdirectory of GOROOT/src. ('go test' + // runs the test in the directory containing the packaged under test.) That + // means that if we start walking up the tree, we should eventually find + // GOROOT/src/go.mod, and we can report the parent directory of that. + // + // Notably, this works even if we can't run 'go env GOROOT' as a + // subprocess. + + cwd, err := os.Getwd() + if err != nil { + gorootErr = fmt.Errorf("finding GOROOT: %w", err) + return + } + + dir := cwd + for { + parent := filepath.Dir(dir) + if parent == dir { + // dir is either "." or only a volume name. + gorootErr = fmt.Errorf("failed to locate GOROOT/src in any parent directory") + return + } + + if base := filepath.Base(dir); base != "src" { + dir = parent + continue // dir cannot be GOROOT/src if it doesn't end in "src". + } + + b, err := os.ReadFile(filepath.Join(dir, "go.mod")) + if err != nil { + if os.IsNotExist(err) { + dir = parent + continue + } + gorootErr = fmt.Errorf("finding GOROOT: %w", err) + return + } + goMod := string(b) + + for goMod != "" { + var line string + line, goMod, _ = strings.Cut(goMod, "\n") + fields := strings.Fields(line) + if len(fields) >= 2 && fields[0] == "module" && fields[1] == "std" { + // Found "module std", which is the module declaration in GOROOT/src! + gorootPath = parent + return + } + } + } + }) + + return gorootPath, gorootErr +} + +// GOROOT reports the path to the directory containing the root of the Go +// project source tree. This is normally equivalent to runtime.GOROOT, but +// works even if the test binary was built with -trimpath and cannot exec +// 'go env GOROOT'. +// +// If GOROOT cannot be found, GOROOT skips t if t is non-nil, +// or panics otherwise. +func GOROOT(t testing.TB) string { + path, err := findGOROOT() + if err != nil { + if t == nil { + panic(err) + } + t.Helper() + t.Skip(err) + } + return path +} + +// GoTool reports the path to the Go tool. +func GoTool() (string, error) { + if !HasGoBuild() { + return "", errors.New("platform cannot run go tool") + } + goToolOnce.Do(func() { + goToolPath, goToolErr = exec.LookPath("go") + }) + return goToolPath, goToolErr +} + +var ( + goToolOnce sync.Once + goToolPath string + goToolErr error +) + +// HasSrc reports whether the entire source tree is available under GOROOT. +func HasSrc() bool { + switch runtime.GOOS { + case "ios": + return false + } + return true +} + +// HasExternalNetwork reports whether the current system can use +// external (non-localhost) networks. +func HasExternalNetwork() bool { + return !testing.Short() && runtime.GOOS != "js" && runtime.GOOS != "wasip1" +} + +// MustHaveExternalNetwork checks that the current system can use +// external (non-localhost) networks. +// If not, MustHaveExternalNetwork calls t.Skip with an explanation. +func MustHaveExternalNetwork(t testing.TB) { + if runtime.GOOS == "js" || runtime.GOOS == "wasip1" { + t.Helper() + t.Skipf("skipping test: no external network on %s", runtime.GOOS) + } + if testing.Short() { + t.Helper() + t.Skipf("skipping test: no external network in -short mode") + } +} + +// HasCGO reports whether the current system can use cgo. +func HasCGO() bool { + hasCgoOnce.Do(func() { + goTool, err := GoTool() + if err != nil { + return + } + cmd := exec.Command(goTool, "env", "CGO_ENABLED") + cmd.Env = origEnv + out, err := cmd.Output() + if err != nil { + panic(fmt.Sprintf("%v: %v", cmd, out)) + } + hasCgo, err = strconv.ParseBool(string(bytes.TrimSpace(out))) + if err != nil { + panic(fmt.Sprintf("%v: non-boolean output %q", cmd, out)) + } + }) + return hasCgo +} + +var ( + hasCgoOnce sync.Once + hasCgo bool +) + +// MustHaveCGO calls t.Skip if cgo is not available. +func MustHaveCGO(t testing.TB) { + if !HasCGO() { + t.Skipf("skipping test: no cgo") + } +} + +// CanInternalLink reports whether the current system can link programs with +// internal linking. +func CanInternalLink(withCgo bool) bool { + return !platform.MustLinkExternal(runtime.GOOS, runtime.GOARCH, withCgo) +} + +// MustInternalLink checks that the current system can link programs with internal +// linking. +// If not, MustInternalLink calls t.Skip with an explanation. +func MustInternalLink(t testing.TB, withCgo bool) { + if !CanInternalLink(withCgo) { + if withCgo && CanInternalLink(false) { + t.Skipf("skipping test: internal linking on %s/%s is not supported with cgo", runtime.GOOS, runtime.GOARCH) + } + t.Skipf("skipping test: internal linking on %s/%s is not supported", runtime.GOOS, runtime.GOARCH) + } +} + +// MustHaveBuildMode reports whether the current system can build programs in +// the given build mode. +// If not, MustHaveBuildMode calls t.Skip with an explanation. +func MustHaveBuildMode(t testing.TB, buildmode string) { + if !platform.BuildModeSupported(runtime.Compiler, buildmode, runtime.GOOS, runtime.GOARCH) { + t.Skipf("skipping test: build mode %s on %s/%s is not supported by the %s compiler", buildmode, runtime.GOOS, runtime.GOARCH, runtime.Compiler) + } +} + +// HasSymlink reports whether the current system can use os.Symlink. +func HasSymlink() bool { + ok, _ := hasSymlink() + return ok +} + +// MustHaveSymlink reports whether the current system can use os.Symlink. +// If not, MustHaveSymlink calls t.Skip with an explanation. +func MustHaveSymlink(t testing.TB) { + ok, reason := hasSymlink() + if !ok { + t.Skipf("skipping test: cannot make symlinks on %s/%s: %s", runtime.GOOS, runtime.GOARCH, reason) + } +} + +// HasLink reports whether the current system can use os.Link. +func HasLink() bool { + // From Android release M (Marshmallow), hard linking files is blocked + // and an attempt to call link() on a file will return EACCES. + // - https://code.google.com/p/android-developer-preview/issues/detail?id=3150 + return runtime.GOOS != "plan9" && runtime.GOOS != "android" +} + +// MustHaveLink reports whether the current system can use os.Link. +// If not, MustHaveLink calls t.Skip with an explanation. +func MustHaveLink(t testing.TB) { + if !HasLink() { + t.Skipf("skipping test: hardlinks are not supported on %s/%s", runtime.GOOS, runtime.GOARCH) + } +} + +var flaky = flag.Bool("flaky", false, "run known-flaky tests too") + +func SkipFlaky(t testing.TB, issue int) { + t.Helper() + if !*flaky { + t.Skipf("skipping known flaky test without the -flaky flag; see golang.org/issue/%d", issue) + } +} + +func SkipFlakyNet(t testing.TB) { + t.Helper() + if v, _ := strconv.ParseBool(os.Getenv("GO_BUILDER_FLAKY_NET")); v { + t.Skip("skipping test on builder known to have frequent network failures") + } +} + +// CPUIsSlow reports whether the CPU running the test is suspected to be slow. +func CPUIsSlow() bool { + switch runtime.GOARCH { + case "arm", "mips", "mipsle", "mips64", "mips64le", "wasm": + return true + } + return false +} + +// SkipIfShortAndSlow skips t if -short is set and the CPU running the test is +// suspected to be slow. +// +// (This is useful for CPU-intensive tests that otherwise complete quickly.) +func SkipIfShortAndSlow(t testing.TB) { + if testing.Short() && CPUIsSlow() { + t.Helper() + t.Skipf("skipping test in -short mode on %s", runtime.GOARCH) + } +} + +// SkipIfOptimizationOff skips t if optimization is disabled. +func SkipIfOptimizationOff(t testing.TB) { + if OptimizationOff() { + t.Helper() + t.Skip("skipping test with optimization disabled") + } +} + +// WriteImportcfg writes an importcfg file used by the compiler or linker to +// dstPath containing entries for the file mappings in packageFiles, as well +// as for the packages transitively imported by the package(s) in pkgs. +// +// pkgs may include any package pattern that is valid to pass to 'go list', +// so it may also be a list of Go source files all in the same directory. +func WriteImportcfg(t testing.TB, dstPath string, packageFiles map[string]string, pkgs ...string) { + t.Helper() + + icfg := new(bytes.Buffer) + icfg.WriteString("# import config\n") + for k, v := range packageFiles { + fmt.Fprintf(icfg, "packagefile %s=%s\n", k, v) + } + + if len(pkgs) > 0 { + // Use 'go list' to resolve any missing packages and rewrite the import map. + cmd := Command(t, GoToolPath(t), "list", "-export", "-deps", "-f", `{{if ne .ImportPath "command-line-arguments"}}{{if .Export}}{{.ImportPath}}={{.Export}}{{end}}{{end}}`) + cmd.Args = append(cmd.Args, pkgs...) + cmd.Stderr = new(strings.Builder) + out, err := cmd.Output() + if err != nil { + t.Fatalf("%v: %v\n%s", cmd, err, cmd.Stderr) + } + + for _, line := range strings.Split(string(out), "\n") { + if line == "" { + continue + } + importPath, export, ok := strings.Cut(line, "=") + if !ok { + t.Fatalf("invalid line in output from %v:\n%s", cmd, line) + } + if packageFiles[importPath] == "" { + fmt.Fprintf(icfg, "packagefile %s=%s\n", importPath, export) + } + } + } + + if err := os.WriteFile(dstPath, icfg.Bytes(), 0666); err != nil { + t.Fatal(err) + } +} + +// SyscallIsNotSupported reports whether err may indicate that a system call is +// not supported by the current platform or execution environment. +func SyscallIsNotSupported(err error) bool { + return syscallIsNotSupported(err) +} diff --git a/src/internal/testenv/testenv_notunix.go b/src/internal/testenv/testenv_notunix.go new file mode 100644 index 0000000..a7df5f5 --- /dev/null +++ b/src/internal/testenv/testenv_notunix.go @@ -0,0 +1,21 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build windows || plan9 || (js && wasm) || wasip1 + +package testenv + +import ( + "errors" + "io/fs" + "os" +) + +// Sigquit is the signal to send to kill a hanging subprocess. +// On Unix we send SIGQUIT, but on non-Unix we only have os.Kill. +var Sigquit = os.Kill + +func syscallIsNotSupported(err error) bool { + return errors.Is(err, fs.ErrPermission) || errors.Is(err, errors.ErrUnsupported) +} diff --git a/src/internal/testenv/testenv_notwin.go b/src/internal/testenv/testenv_notwin.go new file mode 100644 index 0000000..30e159a --- /dev/null +++ b/src/internal/testenv/testenv_notwin.go @@ -0,0 +1,46 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !windows + +package testenv + +import ( + "fmt" + "os" + "path/filepath" + "runtime" +) + +func hasSymlink() (ok bool, reason string) { + switch runtime.GOOS { + case "plan9": + return false, "" + case "android", "wasip1": + // For wasip1, some runtimes forbid absolute symlinks, + // or symlinks that escape the current working directory. + // Perform a simple test to see whether the runtime + // supports symlinks or not. If we get a permission + // error, the runtime does not support symlinks. + dir, err := os.MkdirTemp("", "") + if err != nil { + return false, "" + } + defer func() { + _ = os.RemoveAll(dir) + }() + fpath := filepath.Join(dir, "testfile.txt") + if err := os.WriteFile(fpath, nil, 0644); err != nil { + return false, "" + } + if err := os.Symlink(fpath, filepath.Join(dir, "testlink")); err != nil { + if SyscallIsNotSupported(err) { + return false, fmt.Sprintf("symlinks unsupported: %s", err.Error()) + } + return false, "" + } + } + + return true, "" +} diff --git a/src/internal/testenv/testenv_test.go b/src/internal/testenv/testenv_test.go new file mode 100644 index 0000000..e4ef3bc --- /dev/null +++ b/src/internal/testenv/testenv_test.go @@ -0,0 +1,185 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package testenv_test + +import ( + "internal/platform" + "internal/testenv" + "os" + "path/filepath" + "runtime" + "strings" + "testing" +) + +func TestGoToolLocation(t *testing.T) { + testenv.MustHaveGoBuild(t) + + var exeSuffix string + if runtime.GOOS == "windows" { + exeSuffix = ".exe" + } + + // Tests are defined to run within their package source directory, + // and this package's source directory is $GOROOT/src/internal/testenv. + // The 'go' command is installed at $GOROOT/bin/go, so if the environment + // is correct then testenv.GoTool() should be identical to ../../../bin/go. + + relWant := "../../../bin/go" + exeSuffix + absWant, err := filepath.Abs(relWant) + if err != nil { + t.Fatal(err) + } + + wantInfo, err := os.Stat(absWant) + if err != nil { + t.Fatal(err) + } + t.Logf("found go tool at %q (%q)", relWant, absWant) + + goTool, err := testenv.GoTool() + if err != nil { + t.Fatalf("testenv.GoTool(): %v", err) + } + t.Logf("testenv.GoTool() = %q", goTool) + + gotInfo, err := os.Stat(goTool) + if err != nil { + t.Fatal(err) + } + if !os.SameFile(wantInfo, gotInfo) { + t.Fatalf("%q is not the same file as %q", absWant, goTool) + } +} + +func TestHasGoBuild(t *testing.T) { + if !testenv.HasGoBuild() { + switch runtime.GOOS { + case "js", "wasip1": + // No exec syscall, so these shouldn't be able to 'go build'. + t.Logf("HasGoBuild is false on %s", runtime.GOOS) + return + } + + b := testenv.Builder() + if b == "" { + // We shouldn't make assumptions about what kind of sandbox or build + // environment external Go users may be running in. + t.Skipf("skipping: 'go build' unavailable") + } + + // Since we control the Go builders, we know which ones ought + // to be able to run 'go build'. Check that they can. + // + // (Note that we don't verify that any builders *can't* run 'go build'. + // If a builder starts running 'go build' tests when it shouldn't, + // we will presumably find out about it when those tests fail.) + switch runtime.GOOS { + case "ios": + if isCorelliumBuilder(b) { + // The corellium environment is self-hosting, so it should be able + // to build even though real "ios" devices can't exec. + } else { + // The usual iOS sandbox does not allow the app to start another + // process. If we add builders on stock iOS devices, they presumably + // will not be able to exec, so we may as well allow that now. + t.Logf("HasGoBuild is false on %s", b) + return + } + case "android": + if isEmulatedBuilder(b) && platform.MustLinkExternal(runtime.GOOS, runtime.GOARCH, false) { + // As of 2023-05-02, the test environment on the emulated builders is + // missing a C linker. + t.Logf("HasGoBuild is false on %s", b) + return + } + } + + if strings.Contains(b, "-noopt") { + // The -noopt builder sets GO_GCFLAGS, which causes tests of 'go build' to + // be skipped. + t.Logf("HasGoBuild is false on %s", b) + return + } + + t.Fatalf("HasGoBuild unexpectedly false on %s", b) + } + + t.Logf("HasGoBuild is true; checking consistency with other functions") + + hasExec := false + hasExecGo := false + t.Run("MustHaveExec", func(t *testing.T) { + testenv.MustHaveExec(t) + hasExec = true + }) + t.Run("MustHaveExecPath", func(t *testing.T) { + testenv.MustHaveExecPath(t, "go") + hasExecGo = true + }) + if !hasExec { + t.Errorf(`MustHaveExec(t) skipped unexpectedly`) + } + if !hasExecGo { + t.Errorf(`MustHaveExecPath(t, "go") skipped unexpectedly`) + } + + dir := t.TempDir() + mainGo := filepath.Join(dir, "main.go") + if err := os.WriteFile(mainGo, []byte("package main\nfunc main() {}\n"), 0644); err != nil { + t.Fatal(err) + } + cmd := testenv.Command(t, "go", "build", "-o", os.DevNull, mainGo) + out, err := cmd.CombinedOutput() + if err != nil { + t.Fatalf("%v: %v\n%s", cmd, err, out) + } +} + +func TestMustHaveExec(t *testing.T) { + hasExec := false + t.Run("MustHaveExec", func(t *testing.T) { + testenv.MustHaveExec(t) + t.Logf("MustHaveExec did not skip") + hasExec = true + }) + + switch runtime.GOOS { + case "js", "wasip1": + if hasExec { + // js and wasip1 lack an “exec” syscall. + t.Errorf("expected MustHaveExec to skip on %v", runtime.GOOS) + } + case "ios": + if b := testenv.Builder(); isCorelliumBuilder(b) && !hasExec { + // Most ios environments can't exec, but the corellium builder can. + t.Errorf("expected MustHaveExec not to skip on %v", b) + } + default: + if b := testenv.Builder(); b != "" && !hasExec { + t.Errorf("expected MustHaveExec not to skip on %v", b) + } + } +} + +func isCorelliumBuilder(builderName string) bool { + // Support both the old infra's builder names and the LUCI builder names. + // The former's names are ad-hoc so we could maintain this invariant on + // the builder side. The latter's names are structured, and "corellium" will + // appear as a "host" suffix after the GOOS and GOARCH, which always begin + // with an underscore. + return strings.HasSuffix(builderName, "-corellium") || strings.Contains(builderName, "_corellium") +} + +func isEmulatedBuilder(builderName string) bool { + // Support both the old infra's builder names and the LUCI builder names. + // The former's names are ad-hoc so we could maintain this invariant on + // the builder side. The latter's names are structured, and the signifier + // of emulation "emu" will appear as a "host" suffix after the GOOS and + // GOARCH because it modifies the run environment in such a way that it + // the target GOOS and GOARCH may not match the host. This suffix always + // begins with an underscore. + return strings.HasSuffix(builderName, "-emu") || strings.Contains(builderName, "_emu") +}
\ No newline at end of file diff --git a/src/internal/testenv/testenv_unix.go b/src/internal/testenv/testenv_unix.go new file mode 100644 index 0000000..a629078 --- /dev/null +++ b/src/internal/testenv/testenv_unix.go @@ -0,0 +1,43 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build unix + +package testenv + +import ( + "errors" + "io/fs" + "syscall" +) + +// Sigquit is the signal to send to kill a hanging subprocess. +// Send SIGQUIT to get a stack trace. +var Sigquit = syscall.SIGQUIT + +func syscallIsNotSupported(err error) bool { + if err == nil { + return false + } + + var errno syscall.Errno + if errors.As(err, &errno) { + switch errno { + case syscall.EPERM, syscall.EROFS: + // User lacks permission: either the call requires root permission and the + // user is not root, or the call is denied by a container security policy. + return true + case syscall.EINVAL: + // Some containers return EINVAL instead of EPERM if a system call is + // denied by security policy. + return true + } + } + + if errors.Is(err, fs.ErrPermission) || errors.Is(err, errors.ErrUnsupported) { + return true + } + + return false +} diff --git a/src/internal/testenv/testenv_windows.go b/src/internal/testenv/testenv_windows.go new file mode 100644 index 0000000..4802b13 --- /dev/null +++ b/src/internal/testenv/testenv_windows.go @@ -0,0 +1,47 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package testenv + +import ( + "os" + "path/filepath" + "sync" + "syscall" +) + +var symlinkOnce sync.Once +var winSymlinkErr error + +func initWinHasSymlink() { + tmpdir, err := os.MkdirTemp("", "symtest") + if err != nil { + panic("failed to create temp directory: " + err.Error()) + } + defer os.RemoveAll(tmpdir) + + err = os.Symlink("target", filepath.Join(tmpdir, "symlink")) + if err != nil { + err = err.(*os.LinkError).Err + switch err { + case syscall.EWINDOWS, syscall.ERROR_PRIVILEGE_NOT_HELD: + winSymlinkErr = err + } + } +} + +func hasSymlink() (ok bool, reason string) { + symlinkOnce.Do(initWinHasSymlink) + + switch winSymlinkErr { + case nil: + return true, "" + case syscall.EWINDOWS: + return false, ": symlinks are not supported on your version of Windows" + case syscall.ERROR_PRIVILEGE_NOT_HELD: + return false, ": you don't have enough privileges to create symlinks" + } + + return false, "" +} diff --git a/src/internal/testlog/exit.go b/src/internal/testlog/exit.go new file mode 100644 index 0000000..e15defd --- /dev/null +++ b/src/internal/testlog/exit.go @@ -0,0 +1,33 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package testlog + +import "sync" + +// PanicOnExit0 reports whether to panic on a call to os.Exit(0). +// This is in the testlog package because, like other definitions in +// package testlog, it is a hook between the testing package and the +// os package. This is used to ensure that an early call to os.Exit(0) +// does not cause a test to pass. +func PanicOnExit0() bool { + panicOnExit0.mu.Lock() + defer panicOnExit0.mu.Unlock() + return panicOnExit0.val +} + +// panicOnExit0 is the flag used for PanicOnExit0. This uses a lock +// because the value can be cleared via a timer call that may race +// with calls to os.Exit +var panicOnExit0 struct { + mu sync.Mutex + val bool +} + +// SetPanicOnExit0 sets panicOnExit0 to v. +func SetPanicOnExit0(v bool) { + panicOnExit0.mu.Lock() + defer panicOnExit0.mu.Unlock() + panicOnExit0.val = v +} diff --git a/src/internal/testlog/log.go b/src/internal/testlog/log.go new file mode 100644 index 0000000..3c5f780 --- /dev/null +++ b/src/internal/testlog/log.go @@ -0,0 +1,69 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package testlog provides a back-channel communication path +// between tests and package os, so that cmd/go can see which +// environment variables and files a test consults. +package testlog + +import "sync/atomic" + +// Interface is the interface required of test loggers. +// The os package will invoke the interface's methods to indicate that +// it is inspecting the given environment variables or files. +// Multiple goroutines may call these methods simultaneously. +type Interface interface { + Getenv(key string) + Stat(file string) + Open(file string) + Chdir(dir string) +} + +// logger is the current logger Interface. +// We use an atomic.Value in case test startup +// is racing with goroutines started during init. +// That must not cause a race detector failure, +// although it will still result in limited visibility +// into exactly what those goroutines do. +var logger atomic.Value + +// SetLogger sets the test logger implementation for the current process. +// It must be called only once, at process startup. +func SetLogger(impl Interface) { + if logger.Load() != nil { + panic("testlog: SetLogger must be called only once") + } + logger.Store(&impl) +} + +// Logger returns the current test logger implementation. +// It returns nil if there is no logger. +func Logger() Interface { + impl := logger.Load() + if impl == nil { + return nil + } + return *impl.(*Interface) +} + +// Getenv calls Logger().Getenv, if a logger has been set. +func Getenv(name string) { + if log := Logger(); log != nil { + log.Getenv(name) + } +} + +// Open calls Logger().Open, if a logger has been set. +func Open(name string) { + if log := Logger(); log != nil { + log.Open(name) + } +} + +// Stat calls Logger().Stat, if a logger has been set. +func Stat(name string) { + if log := Logger(); log != nil { + log.Stat(name) + } +} diff --git a/src/internal/testpty/pty.go b/src/internal/testpty/pty.go new file mode 100644 index 0000000..f0b2a33 --- /dev/null +++ b/src/internal/testpty/pty.go @@ -0,0 +1,38 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package testpty is a simple pseudo-terminal package for Unix systems, +// implemented by calling C functions via cgo. +package testpty + +import ( + "errors" + "fmt" + "os" +) + +type PtyError struct { + FuncName string + ErrorString string + Errno error +} + +func ptyError(name string, err error) *PtyError { + return &PtyError{name, err.Error(), err} +} + +func (e *PtyError) Error() string { + return fmt.Sprintf("%s: %s", e.FuncName, e.ErrorString) +} + +func (e *PtyError) Unwrap() error { return e.Errno } + +var ErrNotSupported = errors.New("testpty.Open not implemented on this platform") + +// Open returns a control pty and the name of the linked process tty. +// +// If Open is not implemented on this platform, it returns ErrNotSupported. +func Open() (pty *os.File, processTTY string, err error) { + return open() +} diff --git a/src/internal/testpty/pty_cgo.go b/src/internal/testpty/pty_cgo.go new file mode 100644 index 0000000..442fbcf --- /dev/null +++ b/src/internal/testpty/pty_cgo.go @@ -0,0 +1,34 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build cgo && (aix || dragonfly || freebsd || (linux && !android) || netbsd || openbsd) + +package testpty + +/* +#define _XOPEN_SOURCE 600 +#include <fcntl.h> +#include <stdlib.h> +#include <unistd.h> +*/ +import "C" + +import "os" + +func open() (pty *os.File, processTTY string, err error) { + m, err := C.posix_openpt(C.O_RDWR) + if m < 0 { + return nil, "", ptyError("posix_openpt", err) + } + if res, err := C.grantpt(m); res < 0 { + C.close(m) + return nil, "", ptyError("grantpt", err) + } + if res, err := C.unlockpt(m); res < 0 { + C.close(m) + return nil, "", ptyError("unlockpt", err) + } + processTTY = C.GoString(C.ptsname(m)) + return os.NewFile(uintptr(m), "pty"), processTTY, nil +} diff --git a/src/internal/testpty/pty_darwin.go b/src/internal/testpty/pty_darwin.go new file mode 100644 index 0000000..f29517c --- /dev/null +++ b/src/internal/testpty/pty_darwin.go @@ -0,0 +1,32 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package testpty + +import ( + "internal/syscall/unix" + "os" + "syscall" +) + +func open() (pty *os.File, processTTY string, err error) { + m, err := unix.PosixOpenpt(syscall.O_RDWR) + if err != nil { + return nil, "", ptyError("posix_openpt", err) + } + if err := unix.Grantpt(m); err != nil { + syscall.Close(m) + return nil, "", ptyError("grantpt", err) + } + if err := unix.Unlockpt(m); err != nil { + syscall.Close(m) + return nil, "", ptyError("unlockpt", err) + } + processTTY, err = unix.Ptsname(m) + if err != nil { + syscall.Close(m) + return nil, "", ptyError("ptsname", err) + } + return os.NewFile(uintptr(m), "pty"), processTTY, nil +} diff --git a/src/internal/testpty/pty_none.go b/src/internal/testpty/pty_none.go new file mode 100644 index 0000000..4f9e2b7 --- /dev/null +++ b/src/internal/testpty/pty_none.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +//go:build !(cgo && (aix || dragonfly || freebsd || (linux && !android) || netbsd || openbsd)) && !darwin + +package testpty + +import "os" + +func open() (pty *os.File, processTTY string, err error) { + return nil, "", ErrNotSupported +} diff --git a/src/internal/trace/gc.go b/src/internal/trace/gc.go new file mode 100644 index 0000000..3bd284e --- /dev/null +++ b/src/internal/trace/gc.go @@ -0,0 +1,826 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "container/heap" + "math" + "sort" + "strings" + "time" +) + +// MutatorUtil is a change in mutator utilization at a particular +// time. Mutator utilization functions are represented as a +// time-ordered []MutatorUtil. +type MutatorUtil struct { + Time int64 + // Util is the mean mutator utilization starting at Time. This + // is in the range [0, 1]. + Util float64 +} + +// UtilFlags controls the behavior of MutatorUtilization. +type UtilFlags int + +const ( + // UtilSTW means utilization should account for STW events. + // This includes non-GC STW events, which are typically user-requested. + UtilSTW UtilFlags = 1 << iota + // UtilBackground means utilization should account for + // background mark workers. + UtilBackground + // UtilAssist means utilization should account for mark + // assists. + UtilAssist + // UtilSweep means utilization should account for sweeping. + UtilSweep + + // UtilPerProc means each P should be given a separate + // utilization function. Otherwise, there is a single function + // and each P is given a fraction of the utilization. + UtilPerProc +) + +// MutatorUtilization returns a set of mutator utilization functions +// for the given trace. Each function will always end with 0 +// utilization. The bounds of each function are implicit in the first +// and last event; outside of these bounds each function is undefined. +// +// If the UtilPerProc flag is not given, this always returns a single +// utilization function. Otherwise, it returns one function per P. +func MutatorUtilization(events []*Event, flags UtilFlags) [][]MutatorUtil { + if len(events) == 0 { + return nil + } + + type perP struct { + // gc > 0 indicates that GC is active on this P. + gc int + // series the logical series number for this P. This + // is necessary because Ps may be removed and then + // re-added, and then the new P needs a new series. + series int + } + ps := []perP{} + stw := 0 + + out := [][]MutatorUtil{} + assists := map[uint64]bool{} + block := map[uint64]*Event{} + bgMark := map[uint64]bool{} + + for _, ev := range events { + switch ev.Type { + case EvGomaxprocs: + gomaxprocs := int(ev.Args[0]) + if len(ps) > gomaxprocs { + if flags&UtilPerProc != 0 { + // End each P's series. + for _, p := range ps[gomaxprocs:] { + out[p.series] = addUtil(out[p.series], MutatorUtil{ev.Ts, 0}) + } + } + ps = ps[:gomaxprocs] + } + for len(ps) < gomaxprocs { + // Start new P's series. + series := 0 + if flags&UtilPerProc != 0 || len(out) == 0 { + series = len(out) + out = append(out, []MutatorUtil{{ev.Ts, 1}}) + } + ps = append(ps, perP{series: series}) + } + case EvSTWStart: + if flags&UtilSTW != 0 { + stw++ + } + case EvSTWDone: + if flags&UtilSTW != 0 { + stw-- + } + case EvGCMarkAssistStart: + if flags&UtilAssist != 0 { + ps[ev.P].gc++ + assists[ev.G] = true + } + case EvGCMarkAssistDone: + if flags&UtilAssist != 0 { + ps[ev.P].gc-- + delete(assists, ev.G) + } + case EvGCSweepStart: + if flags&UtilSweep != 0 { + ps[ev.P].gc++ + } + case EvGCSweepDone: + if flags&UtilSweep != 0 { + ps[ev.P].gc-- + } + case EvGoStartLabel: + if flags&UtilBackground != 0 && strings.HasPrefix(ev.SArgs[0], "GC ") && ev.SArgs[0] != "GC (idle)" { + // Background mark worker. + // + // If we're in per-proc mode, we don't + // count dedicated workers because + // they kick all of the goroutines off + // that P, so don't directly + // contribute to goroutine latency. + if !(flags&UtilPerProc != 0 && ev.SArgs[0] == "GC (dedicated)") { + bgMark[ev.G] = true + ps[ev.P].gc++ + } + } + fallthrough + case EvGoStart: + if assists[ev.G] { + // Unblocked during assist. + ps[ev.P].gc++ + } + block[ev.G] = ev.Link + default: + if ev != block[ev.G] { + continue + } + + if assists[ev.G] { + // Blocked during assist. + ps[ev.P].gc-- + } + if bgMark[ev.G] { + // Background mark worker done. + ps[ev.P].gc-- + delete(bgMark, ev.G) + } + delete(block, ev.G) + } + + if flags&UtilPerProc == 0 { + // Compute the current average utilization. + if len(ps) == 0 { + continue + } + gcPs := 0 + if stw > 0 { + gcPs = len(ps) + } else { + for i := range ps { + if ps[i].gc > 0 { + gcPs++ + } + } + } + mu := MutatorUtil{ev.Ts, 1 - float64(gcPs)/float64(len(ps))} + + // Record the utilization change. (Since + // len(ps) == len(out), we know len(out) > 0.) + out[0] = addUtil(out[0], mu) + } else { + // Check for per-P utilization changes. + for i := range ps { + p := &ps[i] + util := 1.0 + if stw > 0 || p.gc > 0 { + util = 0.0 + } + out[p.series] = addUtil(out[p.series], MutatorUtil{ev.Ts, util}) + } + } + } + + // Add final 0 utilization event to any remaining series. This + // is important to mark the end of the trace. The exact value + // shouldn't matter since no window should extend beyond this, + // but using 0 is symmetric with the start of the trace. + mu := MutatorUtil{events[len(events)-1].Ts, 0} + for i := range ps { + out[ps[i].series] = addUtil(out[ps[i].series], mu) + } + return out +} + +func addUtil(util []MutatorUtil, mu MutatorUtil) []MutatorUtil { + if len(util) > 0 { + if mu.Util == util[len(util)-1].Util { + // No change. + return util + } + if mu.Time == util[len(util)-1].Time { + // Take the lowest utilization at a time stamp. + if mu.Util < util[len(util)-1].Util { + util[len(util)-1] = mu + } + return util + } + } + return append(util, mu) +} + +// totalUtil is total utilization, measured in nanoseconds. This is a +// separate type primarily to distinguish it from mean utilization, +// which is also a float64. +type totalUtil float64 + +func totalUtilOf(meanUtil float64, dur int64) totalUtil { + return totalUtil(meanUtil * float64(dur)) +} + +// mean returns the mean utilization over dur. +func (u totalUtil) mean(dur time.Duration) float64 { + return float64(u) / float64(dur) +} + +// An MMUCurve is the minimum mutator utilization curve across +// multiple window sizes. +type MMUCurve struct { + series []mmuSeries +} + +type mmuSeries struct { + util []MutatorUtil + // sums[j] is the cumulative sum of util[:j]. + sums []totalUtil + // bands summarizes util in non-overlapping bands of duration + // bandDur. + bands []mmuBand + // bandDur is the duration of each band. + bandDur int64 +} + +type mmuBand struct { + // minUtil is the minimum instantaneous mutator utilization in + // this band. + minUtil float64 + // cumUtil is the cumulative total mutator utilization between + // time 0 and the left edge of this band. + cumUtil totalUtil + + // integrator is the integrator for the left edge of this + // band. + integrator integrator +} + +// NewMMUCurve returns an MMU curve for the given mutator utilization +// function. +func NewMMUCurve(utils [][]MutatorUtil) *MMUCurve { + series := make([]mmuSeries, len(utils)) + for i, util := range utils { + series[i] = newMMUSeries(util) + } + return &MMUCurve{series} +} + +// bandsPerSeries is the number of bands to divide each series into. +// This is only changed by tests. +var bandsPerSeries = 1000 + +func newMMUSeries(util []MutatorUtil) mmuSeries { + // Compute cumulative sum. + sums := make([]totalUtil, len(util)) + var prev MutatorUtil + var sum totalUtil + for j, u := range util { + sum += totalUtilOf(prev.Util, u.Time-prev.Time) + sums[j] = sum + prev = u + } + + // Divide the utilization curve up into equal size + // non-overlapping "bands" and compute a summary for each of + // these bands. + // + // Compute the duration of each band. + numBands := bandsPerSeries + if numBands > len(util) { + // There's no point in having lots of bands if there + // aren't many events. + numBands = len(util) + } + dur := util[len(util)-1].Time - util[0].Time + bandDur := (dur + int64(numBands) - 1) / int64(numBands) + if bandDur < 1 { + bandDur = 1 + } + // Compute the bands. There are numBands+1 bands in order to + // record the final cumulative sum. + bands := make([]mmuBand, numBands+1) + s := mmuSeries{util, sums, bands, bandDur} + leftSum := integrator{&s, 0} + for i := range bands { + startTime, endTime := s.bandTime(i) + cumUtil := leftSum.advance(startTime) + predIdx := leftSum.pos + minUtil := 1.0 + for i := predIdx; i < len(util) && util[i].Time < endTime; i++ { + minUtil = math.Min(minUtil, util[i].Util) + } + bands[i] = mmuBand{minUtil, cumUtil, leftSum} + } + + return s +} + +func (s *mmuSeries) bandTime(i int) (start, end int64) { + start = int64(i)*s.bandDur + s.util[0].Time + end = start + s.bandDur + return +} + +type bandUtil struct { + // Utilization series index + series int + // Band index + i int + // Lower bound of mutator utilization for all windows + // with a left edge in this band. + utilBound float64 +} + +type bandUtilHeap []bandUtil + +func (h bandUtilHeap) Len() int { + return len(h) +} + +func (h bandUtilHeap) Less(i, j int) bool { + return h[i].utilBound < h[j].utilBound +} + +func (h bandUtilHeap) Swap(i, j int) { + h[i], h[j] = h[j], h[i] +} + +func (h *bandUtilHeap) Push(x any) { + *h = append(*h, x.(bandUtil)) +} + +func (h *bandUtilHeap) Pop() any { + x := (*h)[len(*h)-1] + *h = (*h)[:len(*h)-1] + return x +} + +// UtilWindow is a specific window at Time. +type UtilWindow struct { + Time int64 + // MutatorUtil is the mean mutator utilization in this window. + MutatorUtil float64 +} + +type utilHeap []UtilWindow + +func (h utilHeap) Len() int { + return len(h) +} + +func (h utilHeap) Less(i, j int) bool { + if h[i].MutatorUtil != h[j].MutatorUtil { + return h[i].MutatorUtil > h[j].MutatorUtil + } + return h[i].Time > h[j].Time +} + +func (h utilHeap) Swap(i, j int) { + h[i], h[j] = h[j], h[i] +} + +func (h *utilHeap) Push(x any) { + *h = append(*h, x.(UtilWindow)) +} + +func (h *utilHeap) Pop() any { + x := (*h)[len(*h)-1] + *h = (*h)[:len(*h)-1] + return x +} + +// An accumulator takes a windowed mutator utilization function and +// tracks various statistics for that function. +type accumulator struct { + mmu float64 + + // bound is the mutator utilization bound where adding any + // mutator utilization above this bound cannot affect the + // accumulated statistics. + bound float64 + + // Worst N window tracking + nWorst int + wHeap utilHeap + + // Mutator utilization distribution tracking + mud *mud + // preciseMass is the distribution mass that must be precise + // before accumulation is stopped. + preciseMass float64 + // lastTime and lastMU are the previous point added to the + // windowed mutator utilization function. + lastTime int64 + lastMU float64 +} + +// resetTime declares a discontinuity in the windowed mutator +// utilization function by resetting the current time. +func (acc *accumulator) resetTime() { + // This only matters for distribution collection, since that's + // the only thing that depends on the progression of the + // windowed mutator utilization function. + acc.lastTime = math.MaxInt64 +} + +// addMU adds a point to the windowed mutator utilization function at +// (time, mu). This must be called for monotonically increasing values +// of time. +// +// It returns true if further calls to addMU would be pointless. +func (acc *accumulator) addMU(time int64, mu float64, window time.Duration) bool { + if mu < acc.mmu { + acc.mmu = mu + } + acc.bound = acc.mmu + + if acc.nWorst == 0 { + // If the minimum has reached zero, it can't go any + // lower, so we can stop early. + return mu == 0 + } + + // Consider adding this window to the n worst. + if len(acc.wHeap) < acc.nWorst || mu < acc.wHeap[0].MutatorUtil { + // This window is lower than the K'th worst window. + // + // Check if there's any overlapping window + // already in the heap and keep whichever is + // worse. + for i, ui := range acc.wHeap { + if time+int64(window) > ui.Time && ui.Time+int64(window) > time { + if ui.MutatorUtil <= mu { + // Keep the first window. + goto keep + } else { + // Replace it with this window. + heap.Remove(&acc.wHeap, i) + break + } + } + } + + heap.Push(&acc.wHeap, UtilWindow{time, mu}) + if len(acc.wHeap) > acc.nWorst { + heap.Pop(&acc.wHeap) + } + keep: + } + + if len(acc.wHeap) < acc.nWorst { + // We don't have N windows yet, so keep accumulating. + acc.bound = 1.0 + } else { + // Anything above the least worst window has no effect. + acc.bound = math.Max(acc.bound, acc.wHeap[0].MutatorUtil) + } + + if acc.mud != nil { + if acc.lastTime != math.MaxInt64 { + // Update distribution. + acc.mud.add(acc.lastMU, mu, float64(time-acc.lastTime)) + } + acc.lastTime, acc.lastMU = time, mu + if _, mudBound, ok := acc.mud.approxInvCumulativeSum(); ok { + acc.bound = math.Max(acc.bound, mudBound) + } else { + // We haven't accumulated enough total precise + // mass yet to even reach our goal, so keep + // accumulating. + acc.bound = 1 + } + // It's not worth checking percentiles every time, so + // just keep accumulating this band. + return false + } + + // If we've found enough 0 utilizations, we can stop immediately. + return len(acc.wHeap) == acc.nWorst && acc.wHeap[0].MutatorUtil == 0 +} + +// MMU returns the minimum mutator utilization for the given time +// window. This is the minimum utilization for all windows of this +// duration across the execution. The returned value is in the range +// [0, 1]. +func (c *MMUCurve) MMU(window time.Duration) (mmu float64) { + acc := accumulator{mmu: 1.0, bound: 1.0} + c.mmu(window, &acc) + return acc.mmu +} + +// Examples returns n specific examples of the lowest mutator +// utilization for the given window size. The returned windows will be +// disjoint (otherwise there would be a huge number of +// mostly-overlapping windows at the single lowest point). There are +// no guarantees on which set of disjoint windows this returns. +func (c *MMUCurve) Examples(window time.Duration, n int) (worst []UtilWindow) { + acc := accumulator{mmu: 1.0, bound: 1.0, nWorst: n} + c.mmu(window, &acc) + sort.Sort(sort.Reverse(acc.wHeap)) + return ([]UtilWindow)(acc.wHeap) +} + +// MUD returns mutator utilization distribution quantiles for the +// given window size. +// +// The mutator utilization distribution is the distribution of mean +// mutator utilization across all windows of the given window size in +// the trace. +// +// The minimum mutator utilization is the minimum (0th percentile) of +// this distribution. (However, if only the minimum is desired, it's +// more efficient to use the MMU method.) +func (c *MMUCurve) MUD(window time.Duration, quantiles []float64) []float64 { + if len(quantiles) == 0 { + return []float64{} + } + + // Each unrefined band contributes a known total mass to the + // distribution (bandDur except at the end), but in an unknown + // way. However, we know that all the mass it contributes must + // be at or above its worst-case mean mutator utilization. + // + // Hence, we refine bands until the highest desired + // distribution quantile is less than the next worst-case mean + // mutator utilization. At this point, all further + // contributions to the distribution must be beyond the + // desired quantile and hence cannot affect it. + // + // First, find the highest desired distribution quantile. + maxQ := quantiles[0] + for _, q := range quantiles { + if q > maxQ { + maxQ = q + } + } + // The distribution's mass is in units of time (it's not + // normalized because this would make it more annoying to + // account for future contributions of unrefined bands). The + // total final mass will be the duration of the trace itself + // minus the window size. Using this, we can compute the mass + // corresponding to quantile maxQ. + var duration int64 + for _, s := range c.series { + duration1 := s.util[len(s.util)-1].Time - s.util[0].Time + if duration1 >= int64(window) { + duration += duration1 - int64(window) + } + } + qMass := float64(duration) * maxQ + + // Accumulate the MUD until we have precise information for + // everything to the left of qMass. + acc := accumulator{mmu: 1.0, bound: 1.0, preciseMass: qMass, mud: new(mud)} + acc.mud.setTrackMass(qMass) + c.mmu(window, &acc) + + // Evaluate the quantiles on the accumulated MUD. + out := make([]float64, len(quantiles)) + for i := range out { + mu, _ := acc.mud.invCumulativeSum(float64(duration) * quantiles[i]) + if math.IsNaN(mu) { + // There are a few legitimate ways this can + // happen: + // + // 1. If the window is the full trace + // duration, then the windowed MU function is + // only defined at a single point, so the MU + // distribution is not well-defined. + // + // 2. If there are no events, then the MU + // distribution has no mass. + // + // Either way, all of the quantiles will have + // converged toward the MMU at this point. + mu = acc.mmu + } + out[i] = mu + } + return out +} + +func (c *MMUCurve) mmu(window time.Duration, acc *accumulator) { + if window <= 0 { + acc.mmu = 0 + return + } + + var bandU bandUtilHeap + windows := make([]time.Duration, len(c.series)) + for i, s := range c.series { + windows[i] = window + if max := time.Duration(s.util[len(s.util)-1].Time - s.util[0].Time); window > max { + windows[i] = max + } + + bandU1 := bandUtilHeap(s.mkBandUtil(i, windows[i])) + if bandU == nil { + bandU = bandU1 + } else { + bandU = append(bandU, bandU1...) + } + } + + // Process bands from lowest utilization bound to highest. + heap.Init(&bandU) + + // Refine each band into a precise window and MMU until + // refining the next lowest band can no longer affect the MMU + // or windows. + for len(bandU) > 0 && bandU[0].utilBound < acc.bound { + i := bandU[0].series + c.series[i].bandMMU(bandU[0].i, windows[i], acc) + heap.Pop(&bandU) + } +} + +func (c *mmuSeries) mkBandUtil(series int, window time.Duration) []bandUtil { + // For each band, compute the worst-possible total mutator + // utilization for all windows that start in that band. + + // minBands is the minimum number of bands a window can span + // and maxBands is the maximum number of bands a window can + // span in any alignment. + minBands := int((int64(window) + c.bandDur - 1) / c.bandDur) + maxBands := int((int64(window) + 2*(c.bandDur-1)) / c.bandDur) + if window > 1 && maxBands < 2 { + panic("maxBands < 2") + } + tailDur := int64(window) % c.bandDur + nUtil := len(c.bands) - maxBands + 1 + if nUtil < 0 { + nUtil = 0 + } + bandU := make([]bandUtil, nUtil) + for i := range bandU { + // To compute the worst-case MU, we assume the minimum + // for any bands that are only partially overlapped by + // some window and the mean for any bands that are + // completely covered by all windows. + var util totalUtil + + // Find the lowest and second lowest of the partial + // bands. + l := c.bands[i].minUtil + r1 := c.bands[i+minBands-1].minUtil + r2 := c.bands[i+maxBands-1].minUtil + minBand := math.Min(l, math.Min(r1, r2)) + // Assume the worst window maximally overlaps the + // worst minimum and then the rest overlaps the second + // worst minimum. + if minBands == 1 { + util += totalUtilOf(minBand, int64(window)) + } else { + util += totalUtilOf(minBand, c.bandDur) + midBand := 0.0 + switch { + case minBand == l: + midBand = math.Min(r1, r2) + case minBand == r1: + midBand = math.Min(l, r2) + case minBand == r2: + midBand = math.Min(l, r1) + } + util += totalUtilOf(midBand, tailDur) + } + + // Add the total mean MU of bands that are completely + // overlapped by all windows. + if minBands > 2 { + util += c.bands[i+minBands-1].cumUtil - c.bands[i+1].cumUtil + } + + bandU[i] = bandUtil{series, i, util.mean(window)} + } + + return bandU +} + +// bandMMU computes the precise minimum mutator utilization for +// windows with a left edge in band bandIdx. +func (c *mmuSeries) bandMMU(bandIdx int, window time.Duration, acc *accumulator) { + util := c.util + + // We think of the mutator utilization over time as the + // box-filtered utilization function, which we call the + // "windowed mutator utilization function". The resulting + // function is continuous and piecewise linear (unless + // window==0, which we handle elsewhere), where the boundaries + // between segments occur when either edge of the window + // encounters a change in the instantaneous mutator + // utilization function. Hence, the minimum of this function + // will always occur when one of the edges of the window + // aligns with a utilization change, so these are the only + // points we need to consider. + // + // We compute the mutator utilization function incrementally + // by tracking the integral from t=0 to the left edge of the + // window and to the right edge of the window. + left := c.bands[bandIdx].integrator + right := left + time, endTime := c.bandTime(bandIdx) + if utilEnd := util[len(util)-1].Time - int64(window); utilEnd < endTime { + endTime = utilEnd + } + acc.resetTime() + for { + // Advance edges to time and time+window. + mu := (right.advance(time+int64(window)) - left.advance(time)).mean(window) + if acc.addMU(time, mu, window) { + break + } + if time == endTime { + break + } + + // The maximum slope of the windowed mutator + // utilization function is 1/window, so we can always + // advance the time by at least (mu - mmu) * window + // without dropping below mmu. + minTime := time + int64((mu-acc.bound)*float64(window)) + + // Advance the window to the next time where either + // the left or right edge of the window encounters a + // change in the utilization curve. + if t1, t2 := left.next(time), right.next(time+int64(window))-int64(window); t1 < t2 { + time = t1 + } else { + time = t2 + } + if time < minTime { + time = minTime + } + if time >= endTime { + // For MMUs we could stop here, but for MUDs + // it's important that we span the entire + // band. + time = endTime + } + } +} + +// An integrator tracks a position in a utilization function and +// integrates it. +type integrator struct { + u *mmuSeries + // pos is the index in u.util of the current time's non-strict + // predecessor. + pos int +} + +// advance returns the integral of the utilization function from 0 to +// time. advance must be called on monotonically increasing values of +// times. +func (in *integrator) advance(time int64) totalUtil { + util, pos := in.u.util, in.pos + // Advance pos until pos+1 is time's strict successor (making + // pos time's non-strict predecessor). + // + // Very often, this will be nearby, so we optimize that case, + // but it may be arbitrarily far away, so we handled that + // efficiently, too. + const maxSeq = 8 + if pos+maxSeq < len(util) && util[pos+maxSeq].Time > time { + // Nearby. Use a linear scan. + for pos+1 < len(util) && util[pos+1].Time <= time { + pos++ + } + } else { + // Far. Binary search for time's strict successor. + l, r := pos, len(util) + for l < r { + h := int(uint(l+r) >> 1) + if util[h].Time <= time { + l = h + 1 + } else { + r = h + } + } + pos = l - 1 // Non-strict predecessor. + } + in.pos = pos + var partial totalUtil + if time != util[pos].Time { + partial = totalUtilOf(util[pos].Util, time-util[pos].Time) + } + return in.u.sums[pos] + partial +} + +// next returns the smallest time t' > time of a change in the +// utilization function. +func (in *integrator) next(time int64) int64 { + for _, u := range in.u.util[in.pos:] { + if u.Time > time { + return u.Time + } + } + return 1<<63 - 1 +} diff --git a/src/internal/trace/gc_test.go b/src/internal/trace/gc_test.go new file mode 100644 index 0000000..9b9771e --- /dev/null +++ b/src/internal/trace/gc_test.go @@ -0,0 +1,202 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "bytes" + "math" + "os" + "testing" + "time" +) + +// aeq returns true if x and y are equal up to 8 digits (1 part in 100 +// million). +func aeq(x, y float64) bool { + if x < 0 && y < 0 { + x, y = -x, -y + } + const digits = 8 + factor := 1 - math.Pow(10, -digits+1) + return x*factor <= y && y*factor <= x +} + +func TestMMU(t *testing.T) { + t.Parallel() + + // MU + // 1.0 ***** ***** ***** + // 0.5 * * * * + // 0.0 ***** ***** + // 0 1 2 3 4 5 + util := [][]MutatorUtil{{ + {0e9, 1}, + {1e9, 0}, + {2e9, 1}, + {3e9, 0}, + {4e9, 1}, + {5e9, 0}, + }} + mmuCurve := NewMMUCurve(util) + + for _, test := range []struct { + window time.Duration + want float64 + worst []float64 + }{ + {0, 0, []float64{}}, + {time.Millisecond, 0, []float64{0, 0}}, + {time.Second, 0, []float64{0, 0}}, + {2 * time.Second, 0.5, []float64{0.5, 0.5}}, + {3 * time.Second, 1 / 3.0, []float64{1 / 3.0}}, + {4 * time.Second, 0.5, []float64{0.5}}, + {5 * time.Second, 3 / 5.0, []float64{3 / 5.0}}, + {6 * time.Second, 3 / 5.0, []float64{3 / 5.0}}, + } { + if got := mmuCurve.MMU(test.window); !aeq(test.want, got) { + t.Errorf("for %s window, want mu = %f, got %f", test.window, test.want, got) + } + worst := mmuCurve.Examples(test.window, 2) + // Which exact windows are returned is unspecified + // (and depends on the exact banding), so we just + // check that we got the right number with the right + // utilizations. + if len(worst) != len(test.worst) { + t.Errorf("for %s window, want worst %v, got %v", test.window, test.worst, worst) + } else { + for i := range worst { + if worst[i].MutatorUtil != test.worst[i] { + t.Errorf("for %s window, want worst %v, got %v", test.window, test.worst, worst) + break + } + } + } + } +} + +func TestMMUTrace(t *testing.T) { + // Can't be t.Parallel() because it modifies the + // testingOneBand package variable. + if testing.Short() { + // test input too big for all.bash + t.Skip("skipping in -short mode") + } + + data, err := os.ReadFile("testdata/stress_1_10_good") + if err != nil { + t.Fatalf("failed to read input file: %v", err) + } + _, events, err := parse(bytes.NewReader(data), "") + if err != nil { + t.Fatalf("failed to parse trace: %s", err) + } + mu := MutatorUtilization(events.Events, UtilSTW|UtilBackground|UtilAssist) + mmuCurve := NewMMUCurve(mu) + + // Test the optimized implementation against the "obviously + // correct" implementation. + for window := time.Nanosecond; window < 10*time.Second; window *= 10 { + want := mmuSlow(mu[0], window) + got := mmuCurve.MMU(window) + if !aeq(want, got) { + t.Errorf("want %f, got %f mutator utilization in window %s", want, got, window) + } + } + + // Test MUD with band optimization against MUD without band + // optimization. We don't have a simple testing implementation + // of MUDs (the simplest implementation is still quite + // complex), but this is still a pretty good test. + defer func(old int) { bandsPerSeries = old }(bandsPerSeries) + bandsPerSeries = 1 + mmuCurve2 := NewMMUCurve(mu) + quantiles := []float64{0, 1 - .999, 1 - .99} + for window := time.Microsecond; window < time.Second; window *= 10 { + mud1 := mmuCurve.MUD(window, quantiles) + mud2 := mmuCurve2.MUD(window, quantiles) + for i := range mud1 { + if !aeq(mud1[i], mud2[i]) { + t.Errorf("for quantiles %v at window %v, want %v, got %v", quantiles, window, mud2, mud1) + break + } + } + } +} + +func BenchmarkMMU(b *testing.B) { + data, err := os.ReadFile("testdata/stress_1_10_good") + if err != nil { + b.Fatalf("failed to read input file: %v", err) + } + _, events, err := parse(bytes.NewReader(data), "") + if err != nil { + b.Fatalf("failed to parse trace: %s", err) + } + mu := MutatorUtilization(events.Events, UtilSTW|UtilBackground|UtilAssist|UtilSweep) + b.ResetTimer() + + for i := 0; i < b.N; i++ { + mmuCurve := NewMMUCurve(mu) + xMin, xMax := time.Microsecond, time.Second + logMin, logMax := math.Log(float64(xMin)), math.Log(float64(xMax)) + const samples = 100 + for i := 0; i < samples; i++ { + window := time.Duration(math.Exp(float64(i)/(samples-1)*(logMax-logMin) + logMin)) + mmuCurve.MMU(window) + } + } +} + +func mmuSlow(util []MutatorUtil, window time.Duration) (mmu float64) { + if max := time.Duration(util[len(util)-1].Time - util[0].Time); window > max { + window = max + } + + mmu = 1.0 + + // muInWindow returns the mean mutator utilization between + // util[0].Time and end. + muInWindow := func(util []MutatorUtil, end int64) float64 { + total := 0.0 + var prevU MutatorUtil + for _, u := range util { + if u.Time > end { + total += prevU.Util * float64(end-prevU.Time) + break + } + total += prevU.Util * float64(u.Time-prevU.Time) + prevU = u + } + return total / float64(end-util[0].Time) + } + update := func() { + for i, u := range util { + if u.Time+int64(window) > util[len(util)-1].Time { + break + } + mmu = math.Min(mmu, muInWindow(util[i:], u.Time+int64(window))) + } + } + + // Consider all left-aligned windows. + update() + // Reverse the trace. Slightly subtle because each MutatorUtil + // is a *change*. + rutil := make([]MutatorUtil, len(util)) + if util[len(util)-1].Util != 0 { + panic("irreversible trace") + } + for i, u := range util { + util1 := 0.0 + if i != 0 { + util1 = util[i-1].Util + } + rutil[len(rutil)-i-1] = MutatorUtil{Time: -u.Time, Util: util1} + } + util = rutil + // Consider all right-aligned windows. + update() + return +} diff --git a/src/internal/trace/goroutines.go b/src/internal/trace/goroutines.go new file mode 100644 index 0000000..4b4f13d --- /dev/null +++ b/src/internal/trace/goroutines.go @@ -0,0 +1,358 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "sort" + "strings" +) + +// GDesc contains statistics and execution details of a single goroutine. +type GDesc struct { + ID uint64 + Name string + PC uint64 + CreationTime int64 + StartTime int64 + EndTime int64 + + // List of regions in the goroutine, sorted based on the start time. + Regions []*UserRegionDesc + + // Statistics of execution time during the goroutine execution. + GExecutionStat + + *gdesc // private part. +} + +// UserRegionDesc represents a region and goroutine execution stats +// while the region was active. +type UserRegionDesc struct { + TaskID uint64 + Name string + + // Region start event. Normally EvUserRegion start event or nil, + // but can be EvGoCreate event if the region is a synthetic + // region representing task inheritance from the parent goroutine. + Start *Event + + // Region end event. Normally EvUserRegion end event or nil, + // but can be EvGoStop or EvGoEnd event if the goroutine + // terminated without explicitly ending the region. + End *Event + + GExecutionStat +} + +// GExecutionStat contains statistics about a goroutine's execution +// during a period of time. +type GExecutionStat struct { + ExecTime int64 + SchedWaitTime int64 + IOTime int64 + BlockTime int64 + SyscallTime int64 + GCTime int64 + SweepTime int64 + TotalTime int64 +} + +// sub returns the stats v-s. +func (s GExecutionStat) sub(v GExecutionStat) (r GExecutionStat) { + r = s + r.ExecTime -= v.ExecTime + r.SchedWaitTime -= v.SchedWaitTime + r.IOTime -= v.IOTime + r.BlockTime -= v.BlockTime + r.SyscallTime -= v.SyscallTime + r.GCTime -= v.GCTime + r.SweepTime -= v.SweepTime + r.TotalTime -= v.TotalTime + return r +} + +// snapshotStat returns the snapshot of the goroutine execution statistics. +// This is called as we process the ordered trace event stream. lastTs and +// activeGCStartTime are used to process pending statistics if this is called +// before any goroutine end event. +func (g *GDesc) snapshotStat(lastTs, activeGCStartTime int64) (ret GExecutionStat) { + ret = g.GExecutionStat + + if g.gdesc == nil { + return ret // finalized GDesc. No pending state. + } + + if activeGCStartTime != 0 { // terminating while GC is active + if g.CreationTime < activeGCStartTime { + ret.GCTime += lastTs - activeGCStartTime + } else { + // The goroutine's lifetime completely overlaps + // with a GC. + ret.GCTime += lastTs - g.CreationTime + } + } + + if g.TotalTime == 0 { + ret.TotalTime = lastTs - g.CreationTime + } + + if g.lastStartTime != 0 { + ret.ExecTime += lastTs - g.lastStartTime + } + if g.blockNetTime != 0 { + ret.IOTime += lastTs - g.blockNetTime + } + if g.blockSyncTime != 0 { + ret.BlockTime += lastTs - g.blockSyncTime + } + if g.blockSyscallTime != 0 { + ret.SyscallTime += lastTs - g.blockSyscallTime + } + if g.blockSchedTime != 0 { + ret.SchedWaitTime += lastTs - g.blockSchedTime + } + if g.blockSweepTime != 0 { + ret.SweepTime += lastTs - g.blockSweepTime + } + return ret +} + +// finalize is called when processing a goroutine end event or at +// the end of trace processing. This finalizes the execution stat +// and any active regions in the goroutine, in which case trigger is nil. +func (g *GDesc) finalize(lastTs, activeGCStartTime int64, trigger *Event) { + if trigger != nil { + g.EndTime = trigger.Ts + } + finalStat := g.snapshotStat(lastTs, activeGCStartTime) + + g.GExecutionStat = finalStat + + // System goroutines are never part of regions, even though they + // "inherit" a task due to creation (EvGoCreate) from within a region. + // This may happen e.g. if the first GC is triggered within a region, + // starting the GC worker goroutines. + if !IsSystemGoroutine(g.Name) { + for _, s := range g.activeRegions { + s.End = trigger + s.GExecutionStat = finalStat.sub(s.GExecutionStat) + g.Regions = append(g.Regions, s) + } + } + *(g.gdesc) = gdesc{} +} + +// gdesc is a private part of GDesc that is required only during analysis. +type gdesc struct { + lastStartTime int64 + blockNetTime int64 + blockSyncTime int64 + blockSyscallTime int64 + blockSweepTime int64 + blockGCTime int64 + blockSchedTime int64 + + activeRegions []*UserRegionDesc // stack of active regions +} + +// GoroutineStats generates statistics for all goroutines in the trace. +func GoroutineStats(events []*Event) map[uint64]*GDesc { + gs := make(map[uint64]*GDesc) + var lastTs int64 + var gcStartTime int64 // gcStartTime == 0 indicates gc is inactive. + for _, ev := range events { + lastTs = ev.Ts + switch ev.Type { + case EvGoCreate: + g := &GDesc{ID: ev.Args[0], CreationTime: ev.Ts, gdesc: new(gdesc)} + g.blockSchedTime = ev.Ts + // When a goroutine is newly created, inherit the task + // of the active region. For ease handling of this + // case, we create a fake region description with the + // task id. This isn't strictly necessary as this + // goroutine may not be associated with the task, but + // it can be convenient to see all children created + // during a region. + if creatorG := gs[ev.G]; creatorG != nil && len(creatorG.gdesc.activeRegions) > 0 { + regions := creatorG.gdesc.activeRegions + s := regions[len(regions)-1] + if s.TaskID != 0 { + g.gdesc.activeRegions = []*UserRegionDesc{ + {TaskID: s.TaskID, Start: ev}, + } + } + } + gs[g.ID] = g + case EvGoStart, EvGoStartLabel: + g := gs[ev.G] + if g.PC == 0 && len(ev.Stk) > 0 { + g.PC = ev.Stk[0].PC + g.Name = ev.Stk[0].Fn + } + g.lastStartTime = ev.Ts + if g.StartTime == 0 { + g.StartTime = ev.Ts + } + if g.blockSchedTime != 0 { + g.SchedWaitTime += ev.Ts - g.blockSchedTime + g.blockSchedTime = 0 + } + case EvGoEnd, EvGoStop: + g := gs[ev.G] + g.finalize(ev.Ts, gcStartTime, ev) + case EvGoBlockSend, EvGoBlockRecv, EvGoBlockSelect, + EvGoBlockSync, EvGoBlockCond: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + g.blockSyncTime = ev.Ts + case EvGoSched, EvGoPreempt: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + g.blockSchedTime = ev.Ts + case EvGoSleep, EvGoBlock: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + case EvGoBlockNet: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + g.blockNetTime = ev.Ts + case EvGoBlockGC: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + g.blockGCTime = ev.Ts + case EvGoUnblock: + g := gs[ev.Args[0]] + if g.blockNetTime != 0 { + g.IOTime += ev.Ts - g.blockNetTime + g.blockNetTime = 0 + } + if g.blockSyncTime != 0 { + g.BlockTime += ev.Ts - g.blockSyncTime + g.blockSyncTime = 0 + } + g.blockSchedTime = ev.Ts + case EvGoSysBlock: + g := gs[ev.G] + g.ExecTime += ev.Ts - g.lastStartTime + g.lastStartTime = 0 + g.blockSyscallTime = ev.Ts + case EvGoSysExit: + g := gs[ev.G] + if g.blockSyscallTime != 0 { + g.SyscallTime += ev.Ts - g.blockSyscallTime + g.blockSyscallTime = 0 + } + g.blockSchedTime = ev.Ts + case EvGCSweepStart: + g := gs[ev.G] + if g != nil { + // Sweep can happen during GC on system goroutine. + g.blockSweepTime = ev.Ts + } + case EvGCSweepDone: + g := gs[ev.G] + if g != nil && g.blockSweepTime != 0 { + g.SweepTime += ev.Ts - g.blockSweepTime + g.blockSweepTime = 0 + } + case EvGCStart: + gcStartTime = ev.Ts + case EvGCDone: + for _, g := range gs { + if g.EndTime != 0 { + continue + } + if gcStartTime < g.CreationTime { + g.GCTime += ev.Ts - g.CreationTime + } else { + g.GCTime += ev.Ts - gcStartTime + } + } + gcStartTime = 0 // indicates gc is inactive. + case EvUserRegion: + g := gs[ev.G] + switch mode := ev.Args[1]; mode { + case 0: // region start + g.activeRegions = append(g.activeRegions, &UserRegionDesc{ + Name: ev.SArgs[0], + TaskID: ev.Args[0], + Start: ev, + GExecutionStat: g.snapshotStat(lastTs, gcStartTime), + }) + case 1: // region end + var sd *UserRegionDesc + if regionStk := g.activeRegions; len(regionStk) > 0 { + n := len(regionStk) + sd = regionStk[n-1] + regionStk = regionStk[:n-1] // pop + g.activeRegions = regionStk + } else { + sd = &UserRegionDesc{ + Name: ev.SArgs[0], + TaskID: ev.Args[0], + } + } + sd.GExecutionStat = g.snapshotStat(lastTs, gcStartTime).sub(sd.GExecutionStat) + sd.End = ev + g.Regions = append(g.Regions, sd) + } + } + } + + for _, g := range gs { + g.finalize(lastTs, gcStartTime, nil) + + // sort based on region start time + sort.Slice(g.Regions, func(i, j int) bool { + x := g.Regions[i].Start + y := g.Regions[j].Start + if x == nil { + return true + } + if y == nil { + return false + } + return x.Ts < y.Ts + }) + + g.gdesc = nil + } + + return gs +} + +// RelatedGoroutines finds a set of goroutines related to goroutine goid. +func RelatedGoroutines(events []*Event, goid uint64) map[uint64]bool { + // BFS of depth 2 over "unblock" edges + // (what goroutines unblock goroutine goid?). + gmap := make(map[uint64]bool) + gmap[goid] = true + for i := 0; i < 2; i++ { + gmap1 := make(map[uint64]bool) + for g := range gmap { + gmap1[g] = true + } + for _, ev := range events { + if ev.Type == EvGoUnblock && gmap[ev.Args[0]] { + gmap1[ev.G] = true + } + } + gmap = gmap1 + } + gmap[0] = true // for GC events + return gmap +} + +func IsSystemGoroutine(entryFn string) bool { + // This mimics runtime.isSystemGoroutine as closely as + // possible. + // Also, locked g in extra M (with empty entryFn) is system goroutine. + return entryFn == "" || entryFn != "runtime.main" && strings.HasPrefix(entryFn, "runtime.") +} diff --git a/src/internal/trace/mkcanned.bash b/src/internal/trace/mkcanned.bash new file mode 100755 index 0000000..879cf1c --- /dev/null +++ b/src/internal/trace/mkcanned.bash @@ -0,0 +1,20 @@ +#!/usr/bin/env bash +# Copyright 2016 The Go Authors. All rights reserved. +# Use of this source code is governed by a BSD-style +# license that can be found in the LICENSE file. + +# mkcanned.bash creates canned traces for the trace test suite using +# the current Go version. + +set -e + +if [ $# != 1 ]; then + echo "usage: $0 <label>" >&2 + exit 1 +fi + +go test -run '^$' -bench ClientServerParallel4 -benchtime 10x -trace "testdata/http_$1_good" net/http +go test -run 'TraceStress$|TraceStressStartStop$|TestUserTaskRegion$' runtime/trace -savetraces +mv ../../runtime/trace/TestTraceStress.trace "testdata/stress_$1_good" +mv ../../runtime/trace/TestTraceStressStartStop.trace "testdata/stress_start_stop_$1_good" +mv ../../runtime/trace/TestUserTaskRegion.trace "testdata/user_task_region_$1_good" diff --git a/src/internal/trace/mud.go b/src/internal/trace/mud.go new file mode 100644 index 0000000..8826306 --- /dev/null +++ b/src/internal/trace/mud.go @@ -0,0 +1,223 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "math" + "sort" +) + +// mud is an updatable mutator utilization distribution. +// +// This is a continuous distribution of duration over mutator +// utilization. For example, the integral from mutator utilization a +// to b is the total duration during which the mutator utilization was +// in the range [a, b]. +// +// This distribution is *not* normalized (it is not a probability +// distribution). This makes it easier to work with as it's being +// updated. +// +// It is represented as the sum of scaled uniform distribution +// functions and Dirac delta functions (which are treated as +// degenerate uniform distributions). +type mud struct { + sorted, unsorted []edge + + // trackMass is the inverse cumulative sum to track as the + // distribution is updated. + trackMass float64 + // trackBucket is the bucket in which trackMass falls. If the + // total mass of the distribution is < trackMass, this is + // len(hist). + trackBucket int + // trackSum is the cumulative sum of hist[:trackBucket]. Once + // trackSum >= trackMass, trackBucket must be recomputed. + trackSum float64 + + // hist is a hierarchical histogram of distribution mass. + hist [mudDegree]float64 +} + +const ( + // mudDegree is the number of buckets in the MUD summary + // histogram. + mudDegree = 1024 +) + +type edge struct { + // At x, the function increases by y. + x, delta float64 + // Additionally at x is a Dirac delta function with area dirac. + dirac float64 +} + +// add adds a uniform function over [l, r] scaled so the total weight +// of the uniform is area. If l==r, this adds a Dirac delta function. +func (d *mud) add(l, r, area float64) { + if area == 0 { + return + } + + if r < l { + l, r = r, l + } + + // Add the edges. + if l == r { + d.unsorted = append(d.unsorted, edge{l, 0, area}) + } else { + delta := area / (r - l) + d.unsorted = append(d.unsorted, edge{l, delta, 0}, edge{r, -delta, 0}) + } + + // Update the histogram. + h := &d.hist + lbFloat, lf := math.Modf(l * mudDegree) + lb := int(lbFloat) + if lb >= mudDegree { + lb, lf = mudDegree-1, 1 + } + if l == r { + h[lb] += area + } else { + rbFloat, rf := math.Modf(r * mudDegree) + rb := int(rbFloat) + if rb >= mudDegree { + rb, rf = mudDegree-1, 1 + } + if lb == rb { + h[lb] += area + } else { + perBucket := area / (r - l) / mudDegree + h[lb] += perBucket * (1 - lf) + h[rb] += perBucket * rf + for i := lb + 1; i < rb; i++ { + h[i] += perBucket + } + } + } + + // Update mass tracking. + if thresh := float64(d.trackBucket) / mudDegree; l < thresh { + if r < thresh { + d.trackSum += area + } else { + d.trackSum += area * (thresh - l) / (r - l) + } + if d.trackSum >= d.trackMass { + // The tracked mass now falls in a different + // bucket. Recompute the inverse cumulative sum. + d.setTrackMass(d.trackMass) + } + } +} + +// setTrackMass sets the mass to track the inverse cumulative sum for. +// +// Specifically, mass is a cumulative duration, and the mutator +// utilization bounds for this duration can be queried using +// approxInvCumulativeSum. +func (d *mud) setTrackMass(mass float64) { + d.trackMass = mass + + // Find the bucket currently containing trackMass by computing + // the cumulative sum. + sum := 0.0 + for i, val := range d.hist[:] { + newSum := sum + val + if newSum > mass { + // mass falls in bucket i. + d.trackBucket = i + d.trackSum = sum + return + } + sum = newSum + } + d.trackBucket = len(d.hist) + d.trackSum = sum +} + +// approxInvCumulativeSum is like invCumulativeSum, but specifically +// operates on the tracked mass and returns an upper and lower bound +// approximation of the inverse cumulative sum. +// +// The true inverse cumulative sum will be in the range [lower, upper). +func (d *mud) approxInvCumulativeSum() (float64, float64, bool) { + if d.trackBucket == len(d.hist) { + return math.NaN(), math.NaN(), false + } + return float64(d.trackBucket) / mudDegree, float64(d.trackBucket+1) / mudDegree, true +} + +// invCumulativeSum returns x such that the integral of d from -∞ to x +// is y. If the total weight of d is less than y, it returns the +// maximum of the distribution and false. +// +// Specifically, y is a cumulative duration, and invCumulativeSum +// returns the mutator utilization x such that at least y time has +// been spent with mutator utilization <= x. +func (d *mud) invCumulativeSum(y float64) (float64, bool) { + if len(d.sorted) == 0 && len(d.unsorted) == 0 { + return math.NaN(), false + } + + // Sort edges. + edges := d.unsorted + sort.Slice(edges, func(i, j int) bool { + return edges[i].x < edges[j].x + }) + // Merge with sorted edges. + d.unsorted = nil + if d.sorted == nil { + d.sorted = edges + } else { + oldSorted := d.sorted + newSorted := make([]edge, len(oldSorted)+len(edges)) + i, j := 0, 0 + for o := range newSorted { + if i >= len(oldSorted) { + copy(newSorted[o:], edges[j:]) + break + } else if j >= len(edges) { + copy(newSorted[o:], oldSorted[i:]) + break + } else if oldSorted[i].x < edges[j].x { + newSorted[o] = oldSorted[i] + i++ + } else { + newSorted[o] = edges[j] + j++ + } + } + d.sorted = newSorted + } + + // Traverse edges in order computing a cumulative sum. + csum, rate, prevX := 0.0, 0.0, 0.0 + for _, e := range d.sorted { + newCsum := csum + (e.x-prevX)*rate + if newCsum >= y { + // y was exceeded between the previous edge + // and this one. + if rate == 0 { + // Anywhere between prevX and + // e.x will do. We return e.x + // because that takes care of + // the y==0 case naturally. + return e.x, true + } + return (y-csum)/rate + prevX, true + } + newCsum += e.dirac + if newCsum >= y { + // y was exceeded by the Dirac delta at e.x. + return e.x, true + } + csum, prevX = newCsum, e.x + rate += e.delta + } + return prevX, false +} diff --git a/src/internal/trace/mud_test.go b/src/internal/trace/mud_test.go new file mode 100644 index 0000000..b3d74dc --- /dev/null +++ b/src/internal/trace/mud_test.go @@ -0,0 +1,87 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "math/rand" + "testing" +) + +func TestMUD(t *testing.T) { + // Insert random uniforms and check histogram mass and + // cumulative sum approximations. + rnd := rand.New(rand.NewSource(42)) + mass := 0.0 + var mud mud + for i := 0; i < 100; i++ { + area, l, r := rnd.Float64(), rnd.Float64(), rnd.Float64() + if rnd.Intn(10) == 0 { + r = l + } + t.Log(l, r, area) + mud.add(l, r, area) + mass += area + + // Check total histogram weight. + hmass := 0.0 + for _, val := range mud.hist { + hmass += val + } + if !aeq(mass, hmass) { + t.Fatalf("want mass %g, got %g", mass, hmass) + } + + // Check inverse cumulative sum approximations. + for j := 0.0; j < mass; j += mass * 0.099 { + mud.setTrackMass(j) + l, u, ok := mud.approxInvCumulativeSum() + inv, ok2 := mud.invCumulativeSum(j) + if !ok || !ok2 { + t.Fatalf("inverse cumulative sum failed: approx %v, exact %v", ok, ok2) + } + if !(l <= inv && inv < u) { + t.Fatalf("inverse(%g) = %g, not ∈ [%g, %g)", j, inv, l, u) + } + } + } +} + +func TestMUDTracking(t *testing.T) { + // Test that the tracked mass is tracked correctly across + // updates. + rnd := rand.New(rand.NewSource(42)) + const uniforms = 100 + for trackMass := 0.0; trackMass < uniforms; trackMass += uniforms / 50 { + var mud mud + mass := 0.0 + mud.setTrackMass(trackMass) + for i := 0; i < uniforms; i++ { + area, l, r := rnd.Float64(), rnd.Float64(), rnd.Float64() + mud.add(l, r, area) + mass += area + l, u, ok := mud.approxInvCumulativeSum() + inv, ok2 := mud.invCumulativeSum(trackMass) + + if mass < trackMass { + if ok { + t.Errorf("approx(%g) = [%g, %g), but mass = %g", trackMass, l, u, mass) + } + if ok2 { + t.Errorf("exact(%g) = %g, but mass = %g", trackMass, inv, mass) + } + } else { + if !ok { + t.Errorf("approx(%g) failed, but mass = %g", trackMass, mass) + } + if !ok2 { + t.Errorf("exact(%g) failed, but mass = %g", trackMass, mass) + } + if ok && ok2 && !(l <= inv && inv < u) { + t.Errorf("inverse(%g) = %g, not ∈ [%g, %g)", trackMass, inv, l, u) + } + } + } + } +} diff --git a/src/internal/trace/order.go b/src/internal/trace/order.go new file mode 100644 index 0000000..07a6e13 --- /dev/null +++ b/src/internal/trace/order.go @@ -0,0 +1,285 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "fmt" + "sort" +) + +type eventBatch struct { + events []*Event + selected bool +} + +type orderEvent struct { + ev *Event + batch int + g uint64 + init gState + next gState +} + +type gStatus int + +type gState struct { + seq uint64 + status gStatus +} + +const ( + gDead gStatus = iota + gRunnable + gRunning + gWaiting + + unordered = ^uint64(0) + garbage = ^uint64(0) - 1 + noseq = ^uint64(0) + seqinc = ^uint64(0) - 1 +) + +// order1007 merges a set of per-P event batches into a single, consistent stream. +// The high level idea is as follows. Events within an individual batch are in +// correct order, because they are emitted by a single P. So we need to produce +// a correct interleaving of the batches. To do this we take first unmerged event +// from each batch (frontier). Then choose subset that is "ready" to be merged, +// that is, events for which all dependencies are already merged. Then we choose +// event with the lowest timestamp from the subset, merge it and repeat. +// This approach ensures that we form a consistent stream even if timestamps are +// incorrect (condition observed on some machines). +func order1007(m map[int][]*Event) (events []*Event, err error) { + pending := 0 + // The ordering of CPU profile sample events in the data stream is based on + // when each run of the signal handler was able to acquire the spinlock, + // with original timestamps corresponding to when ReadTrace pulled the data + // off of the profBuf queue. Re-sort them by the timestamp we captured + // inside the signal handler. + sort.Stable(eventList(m[ProfileP])) + var batches []*eventBatch + for _, v := range m { + pending += len(v) + batches = append(batches, &eventBatch{v, false}) + } + gs := make(map[uint64]gState) + var frontier []orderEvent + for ; pending != 0; pending-- { + for i, b := range batches { + if b.selected || len(b.events) == 0 { + continue + } + ev := b.events[0] + g, init, next := stateTransition(ev) + if !transitionReady(g, gs[g], init) { + continue + } + frontier = append(frontier, orderEvent{ev, i, g, init, next}) + b.events = b.events[1:] + b.selected = true + // Get rid of "Local" events, they are intended merely for ordering. + switch ev.Type { + case EvGoStartLocal: + ev.Type = EvGoStart + case EvGoUnblockLocal: + ev.Type = EvGoUnblock + case EvGoSysExitLocal: + ev.Type = EvGoSysExit + } + } + if len(frontier) == 0 { + return nil, fmt.Errorf("no consistent ordering of events possible") + } + sort.Sort(orderEventList(frontier)) + f := frontier[0] + frontier[0] = frontier[len(frontier)-1] + frontier = frontier[:len(frontier)-1] + events = append(events, f.ev) + transition(gs, f.g, f.init, f.next) + if !batches[f.batch].selected { + panic("frontier batch is not selected") + } + batches[f.batch].selected = false + } + + // At this point we have a consistent stream of events. + // Make sure time stamps respect the ordering. + // The tests will skip (not fail) the test case if they see this error. + if !sort.IsSorted(eventList(events)) { + return nil, ErrTimeOrder + } + + // The last part is giving correct timestamps to EvGoSysExit events. + // The problem with EvGoSysExit is that actual syscall exit timestamp (ev.Args[2]) + // is potentially acquired long before event emission. So far we've used + // timestamp of event emission (ev.Ts). + // We could not set ev.Ts = ev.Args[2] earlier, because it would produce + // seemingly broken timestamps (misplaced event). + // We also can't simply update the timestamp and resort events, because + // if timestamps are broken we will misplace the event and later report + // logically broken trace (instead of reporting broken timestamps). + lastSysBlock := make(map[uint64]int64) + for _, ev := range events { + switch ev.Type { + case EvGoSysBlock, EvGoInSyscall: + lastSysBlock[ev.G] = ev.Ts + case EvGoSysExit: + ts := int64(ev.Args[2]) + if ts == 0 { + continue + } + block := lastSysBlock[ev.G] + if block == 0 { + return nil, fmt.Errorf("stray syscall exit") + } + if ts < block { + return nil, ErrTimeOrder + } + ev.Ts = ts + } + } + sort.Stable(eventList(events)) + + return +} + +// stateTransition returns goroutine state (sequence and status) when the event +// becomes ready for merging (init) and the goroutine state after the event (next). +func stateTransition(ev *Event) (g uint64, init, next gState) { + switch ev.Type { + case EvGoCreate: + g = ev.Args[0] + init = gState{0, gDead} + next = gState{1, gRunnable} + case EvGoWaiting, EvGoInSyscall: + g = ev.G + init = gState{1, gRunnable} + next = gState{2, gWaiting} + case EvGoStart, EvGoStartLabel: + g = ev.G + init = gState{ev.Args[1], gRunnable} + next = gState{ev.Args[1] + 1, gRunning} + case EvGoStartLocal: + // noseq means that this event is ready for merging as soon as + // frontier reaches it (EvGoStartLocal is emitted on the same P + // as the corresponding EvGoCreate/EvGoUnblock, and thus the latter + // is already merged). + // seqinc is a stub for cases when event increments g sequence, + // but since we don't know current seq we also don't know next seq. + g = ev.G + init = gState{noseq, gRunnable} + next = gState{seqinc, gRunning} + case EvGoBlock, EvGoBlockSend, EvGoBlockRecv, EvGoBlockSelect, + EvGoBlockSync, EvGoBlockCond, EvGoBlockNet, EvGoSleep, + EvGoSysBlock, EvGoBlockGC: + g = ev.G + init = gState{noseq, gRunning} + next = gState{noseq, gWaiting} + case EvGoSched, EvGoPreempt: + g = ev.G + init = gState{noseq, gRunning} + next = gState{noseq, gRunnable} + case EvGoUnblock, EvGoSysExit: + g = ev.Args[0] + init = gState{ev.Args[1], gWaiting} + next = gState{ev.Args[1] + 1, gRunnable} + case EvGoUnblockLocal, EvGoSysExitLocal: + g = ev.Args[0] + init = gState{noseq, gWaiting} + next = gState{seqinc, gRunnable} + case EvGCStart: + g = garbage + init = gState{ev.Args[0], gDead} + next = gState{ev.Args[0] + 1, gDead} + default: + // no ordering requirements + g = unordered + } + return +} + +func transitionReady(g uint64, curr, init gState) bool { + return g == unordered || (init.seq == noseq || init.seq == curr.seq) && init.status == curr.status +} + +func transition(gs map[uint64]gState, g uint64, init, next gState) { + if g == unordered { + return + } + curr := gs[g] + if !transitionReady(g, curr, init) { + panic("event sequences are broken") + } + switch next.seq { + case noseq: + next.seq = curr.seq + case seqinc: + next.seq = curr.seq + 1 + } + gs[g] = next +} + +// order1005 merges a set of per-P event batches into a single, consistent stream. +func order1005(m map[int][]*Event) (events []*Event, err error) { + for _, batch := range m { + events = append(events, batch...) + } + for _, ev := range events { + if ev.Type == EvGoSysExit { + // EvGoSysExit emission is delayed until the thread has a P. + // Give it the real sequence number and time stamp. + ev.seq = int64(ev.Args[1]) + if ev.Args[2] != 0 { + ev.Ts = int64(ev.Args[2]) + } + } + } + sort.Sort(eventSeqList(events)) + if !sort.IsSorted(eventList(events)) { + return nil, ErrTimeOrder + } + return +} + +type orderEventList []orderEvent + +func (l orderEventList) Len() int { + return len(l) +} + +func (l orderEventList) Less(i, j int) bool { + return l[i].ev.Ts < l[j].ev.Ts +} + +func (l orderEventList) Swap(i, j int) { + l[i], l[j] = l[j], l[i] +} + +type eventList []*Event + +func (l eventList) Len() int { + return len(l) +} + +func (l eventList) Less(i, j int) bool { + return l[i].Ts < l[j].Ts +} + +func (l eventList) Swap(i, j int) { + l[i], l[j] = l[j], l[i] +} + +type eventSeqList []*Event + +func (l eventSeqList) Len() int { + return len(l) +} + +func (l eventSeqList) Less(i, j int) bool { + return l[i].seq < l[j].seq +} + +func (l eventSeqList) Swap(i, j int) { + l[i], l[j] = l[j], l[i] +} diff --git a/src/internal/trace/parser.go b/src/internal/trace/parser.go new file mode 100644 index 0000000..67fa60b --- /dev/null +++ b/src/internal/trace/parser.go @@ -0,0 +1,1174 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "bufio" + "bytes" + "fmt" + "io" + "math/rand" + "os" + "os/exec" + "path/filepath" + "runtime" + "strconv" + "strings" + _ "unsafe" +) + +func goCmd() string { + var exeSuffix string + if runtime.GOOS == "windows" { + exeSuffix = ".exe" + } + path := filepath.Join(runtime.GOROOT(), "bin", "go"+exeSuffix) + if _, err := os.Stat(path); err == nil { + return path + } + return "go" +} + +// Event describes one event in the trace. +type Event struct { + Off int // offset in input file (for debugging and error reporting) + Type byte // one of Ev* + seq int64 // sequence number + Ts int64 // timestamp in nanoseconds + P int // P on which the event happened (can be one of TimerP, NetpollP, SyscallP) + G uint64 // G on which the event happened + StkID uint64 // unique stack ID + Stk []*Frame // stack trace (can be empty) + Args [3]uint64 // event-type-specific arguments + SArgs []string // event-type-specific string args + // linked event (can be nil), depends on event type: + // for GCStart: the GCStop + // for GCSTWStart: the GCSTWDone + // for GCSweepStart: the GCSweepDone + // for GoCreate: first GoStart of the created goroutine + // for GoStart/GoStartLabel: the associated GoEnd, GoBlock or other blocking event + // for GoSched/GoPreempt: the next GoStart + // for GoBlock and other blocking events: the unblock event + // for GoUnblock: the associated GoStart + // for blocking GoSysCall: the associated GoSysExit + // for GoSysExit: the next GoStart + // for GCMarkAssistStart: the associated GCMarkAssistDone + // for UserTaskCreate: the UserTaskEnd + // for UserRegion: if the start region, the corresponding UserRegion end event + Link *Event +} + +// Frame is a frame in stack traces. +type Frame struct { + PC uint64 + Fn string + File string + Line int +} + +const ( + // Special P identifiers: + FakeP = 1000000 + iota + TimerP // depicts timer unblocks + NetpollP // depicts network unblocks + SyscallP // depicts returns from syscalls + GCP // depicts GC state + ProfileP // depicts recording of CPU profile samples +) + +// ParseResult is the result of Parse. +type ParseResult struct { + // Events is the sorted list of Events in the trace. + Events []*Event + // Stacks is the stack traces keyed by stack IDs from the trace. + Stacks map[uint64][]*Frame +} + +// Parse parses, post-processes and verifies the trace. +func Parse(r io.Reader, bin string) (ParseResult, error) { + ver, res, err := parse(r, bin) + if err != nil { + return ParseResult{}, err + } + if ver < 1007 && bin == "" { + return ParseResult{}, fmt.Errorf("for traces produced by go 1.6 or below, the binary argument must be provided") + } + return res, nil +} + +// parse parses, post-processes and verifies the trace. It returns the +// trace version and the list of events. +func parse(r io.Reader, bin string) (int, ParseResult, error) { + ver, rawEvents, strings, err := readTrace(r) + if err != nil { + return 0, ParseResult{}, err + } + events, stacks, err := parseEvents(ver, rawEvents, strings) + if err != nil { + return 0, ParseResult{}, err + } + events = removeFutile(events) + err = postProcessTrace(ver, events) + if err != nil { + return 0, ParseResult{}, err + } + // Attach stack traces. + for _, ev := range events { + if ev.StkID != 0 { + ev.Stk = stacks[ev.StkID] + } + } + if ver < 1007 && bin != "" { + if err := symbolize(events, bin); err != nil { + return 0, ParseResult{}, err + } + } + return ver, ParseResult{Events: events, Stacks: stacks}, nil +} + +// rawEvent is a helper type used during parsing. +type rawEvent struct { + off int + typ byte + args []uint64 + sargs []string +} + +// readTrace does wire-format parsing and verification. +// It does not care about specific event types and argument meaning. +func readTrace(r io.Reader) (ver int, events []rawEvent, strings map[uint64]string, err error) { + // Read and validate trace header. + var buf [16]byte + off, err := io.ReadFull(r, buf[:]) + if err != nil { + err = fmt.Errorf("failed to read header: read %v, err %v", off, err) + return + } + ver, err = parseHeader(buf[:]) + if err != nil { + return + } + switch ver { + case 1005, 1007, 1008, 1009, 1010, 1011, 1019, 1021: + // Note: When adding a new version, confirm that canned traces from the + // old version are part of the test suite. Add them using mkcanned.bash. + break + default: + err = fmt.Errorf("unsupported trace file version %v.%v (update Go toolchain) %v", ver/1000, ver%1000, ver) + return + } + + // Read events. + strings = make(map[uint64]string) + for { + // Read event type and number of arguments (1 byte). + off0 := off + var n int + n, err = r.Read(buf[:1]) + if err == io.EOF { + err = nil + break + } + if err != nil || n != 1 { + err = fmt.Errorf("failed to read trace at offset 0x%x: n=%v err=%v", off0, n, err) + return + } + off += n + typ := buf[0] << 2 >> 2 + narg := buf[0]>>6 + 1 + inlineArgs := byte(4) + if ver < 1007 { + narg++ + inlineArgs++ + } + if typ == EvNone || typ >= EvCount || EventDescriptions[typ].minVersion > ver { + err = fmt.Errorf("unknown event type %v at offset 0x%x", typ, off0) + return + } + if typ == EvString { + // String dictionary entry [ID, length, string]. + var id uint64 + id, off, err = readVal(r, off) + if err != nil { + return + } + if id == 0 { + err = fmt.Errorf("string at offset %d has invalid id 0", off) + return + } + if strings[id] != "" { + err = fmt.Errorf("string at offset %d has duplicate id %v", off, id) + return + } + var ln uint64 + ln, off, err = readVal(r, off) + if err != nil { + return + } + if ln == 0 { + err = fmt.Errorf("string at offset %d has invalid length 0", off) + return + } + if ln > 1e6 { + err = fmt.Errorf("string at offset %d has too large length %v", off, ln) + return + } + buf := make([]byte, ln) + var n int + n, err = io.ReadFull(r, buf) + if err != nil { + err = fmt.Errorf("failed to read trace at offset %d: read %v, want %v, error %v", off, n, ln, err) + return + } + off += n + strings[id] = string(buf) + continue + } + ev := rawEvent{typ: typ, off: off0} + if narg < inlineArgs { + for i := 0; i < int(narg); i++ { + var v uint64 + v, off, err = readVal(r, off) + if err != nil { + err = fmt.Errorf("failed to read event %v argument at offset %v (%v)", typ, off, err) + return + } + ev.args = append(ev.args, v) + } + } else { + // More than inlineArgs args, the first value is length of the event in bytes. + var v uint64 + v, off, err = readVal(r, off) + if err != nil { + err = fmt.Errorf("failed to read event %v argument at offset %v (%v)", typ, off, err) + return + } + evLen := v + off1 := off + for evLen > uint64(off-off1) { + v, off, err = readVal(r, off) + if err != nil { + err = fmt.Errorf("failed to read event %v argument at offset %v (%v)", typ, off, err) + return + } + ev.args = append(ev.args, v) + } + if evLen != uint64(off-off1) { + err = fmt.Errorf("event has wrong length at offset 0x%x: want %v, got %v", off0, evLen, off-off1) + return + } + } + switch ev.typ { + case EvUserLog: // EvUserLog records are followed by a value string of length ev.args[len(ev.args)-1] + var s string + s, off, err = readStr(r, off) + ev.sargs = append(ev.sargs, s) + } + events = append(events, ev) + } + return +} + +func readStr(r io.Reader, off0 int) (s string, off int, err error) { + var sz uint64 + sz, off, err = readVal(r, off0) + if err != nil || sz == 0 { + return "", off, err + } + if sz > 1e6 { + return "", off, fmt.Errorf("string at offset %d is too large (len=%d)", off, sz) + } + buf := make([]byte, sz) + n, err := io.ReadFull(r, buf) + if err != nil || sz != uint64(n) { + return "", off + n, fmt.Errorf("failed to read trace at offset %d: read %v, want %v, error %v", off, n, sz, err) + } + return string(buf), off + n, nil +} + +// parseHeader parses trace header of the form "go 1.7 trace\x00\x00\x00\x00" +// and returns parsed version as 1007. +func parseHeader(buf []byte) (int, error) { + if len(buf) != 16 { + return 0, fmt.Errorf("bad header length") + } + if buf[0] != 'g' || buf[1] != 'o' || buf[2] != ' ' || + buf[3] < '1' || buf[3] > '9' || + buf[4] != '.' || + buf[5] < '1' || buf[5] > '9' { + return 0, fmt.Errorf("not a trace file") + } + ver := int(buf[5] - '0') + i := 0 + for ; buf[6+i] >= '0' && buf[6+i] <= '9' && i < 2; i++ { + ver = ver*10 + int(buf[6+i]-'0') + } + ver += int(buf[3]-'0') * 1000 + if !bytes.Equal(buf[6+i:], []byte(" trace\x00\x00\x00\x00")[:10-i]) { + return 0, fmt.Errorf("not a trace file") + } + return ver, nil +} + +// Parse events transforms raw events into events. +// It does analyze and verify per-event-type arguments. +func parseEvents(ver int, rawEvents []rawEvent, strings map[uint64]string) (events []*Event, stacks map[uint64][]*Frame, err error) { + var ticksPerSec, lastSeq, lastTs int64 + var lastG uint64 + var lastP int + timerGoids := make(map[uint64]bool) + lastGs := make(map[int]uint64) // last goroutine running on P + stacks = make(map[uint64][]*Frame) + batches := make(map[int][]*Event) // events by P + for _, raw := range rawEvents { + desc := EventDescriptions[raw.typ] + if desc.Name == "" { + err = fmt.Errorf("missing description for event type %v", raw.typ) + return + } + narg := argNum(raw, ver) + if len(raw.args) != narg { + err = fmt.Errorf("%v has wrong number of arguments at offset 0x%x: want %v, got %v", + desc.Name, raw.off, narg, len(raw.args)) + return + } + switch raw.typ { + case EvBatch: + lastGs[lastP] = lastG + lastP = int(raw.args[0]) + lastG = lastGs[lastP] + if ver < 1007 { + lastSeq = int64(raw.args[1]) + lastTs = int64(raw.args[2]) + } else { + lastTs = int64(raw.args[1]) + } + case EvFrequency: + ticksPerSec = int64(raw.args[0]) + if ticksPerSec <= 0 { + // The most likely cause for this is tick skew on different CPUs. + // For example, solaris/amd64 seems to have wildly different + // ticks on different CPUs. + err = ErrTimeOrder + return + } + case EvTimerGoroutine: + timerGoids[raw.args[0]] = true + case EvStack: + if len(raw.args) < 2 { + err = fmt.Errorf("EvStack has wrong number of arguments at offset 0x%x: want at least 2, got %v", + raw.off, len(raw.args)) + return + } + size := raw.args[1] + if size > 1000 { + err = fmt.Errorf("EvStack has bad number of frames at offset 0x%x: %v", + raw.off, size) + return + } + want := 2 + 4*size + if ver < 1007 { + want = 2 + size + } + if uint64(len(raw.args)) != want { + err = fmt.Errorf("EvStack has wrong number of arguments at offset 0x%x: want %v, got %v", + raw.off, want, len(raw.args)) + return + } + id := raw.args[0] + if id != 0 && size > 0 { + stk := make([]*Frame, size) + for i := 0; i < int(size); i++ { + if ver < 1007 { + stk[i] = &Frame{PC: raw.args[2+i]} + } else { + pc := raw.args[2+i*4+0] + fn := raw.args[2+i*4+1] + file := raw.args[2+i*4+2] + line := raw.args[2+i*4+3] + stk[i] = &Frame{PC: pc, Fn: strings[fn], File: strings[file], Line: int(line)} + } + } + stacks[id] = stk + } + default: + e := &Event{Off: raw.off, Type: raw.typ, P: lastP, G: lastG} + var argOffset int + if ver < 1007 { + e.seq = lastSeq + int64(raw.args[0]) + e.Ts = lastTs + int64(raw.args[1]) + lastSeq = e.seq + argOffset = 2 + } else { + e.Ts = lastTs + int64(raw.args[0]) + argOffset = 1 + } + lastTs = e.Ts + for i := argOffset; i < narg; i++ { + if i == narg-1 && desc.Stack { + e.StkID = raw.args[i] + } else { + e.Args[i-argOffset] = raw.args[i] + } + } + switch raw.typ { + case EvGoStart, EvGoStartLocal, EvGoStartLabel: + lastG = e.Args[0] + e.G = lastG + if raw.typ == EvGoStartLabel { + e.SArgs = []string{strings[e.Args[2]]} + } + case EvSTWStart: + e.G = 0 + if ver < 1021 { + switch e.Args[0] { + case 0: + e.SArgs = []string{"mark termination"} + case 1: + e.SArgs = []string{"sweep termination"} + default: + err = fmt.Errorf("unknown STW kind %d", e.Args[0]) + return + } + } else if ver == 1021 { + if kind := e.Args[0]; kind < uint64(len(stwReasonStringsGo121)) { + e.SArgs = []string{stwReasonStringsGo121[kind]} + } else { + e.SArgs = []string{"unknown"} + } + } else { + // Can't make any assumptions. + e.SArgs = []string{"unknown"} + } + case EvGCStart, EvGCDone, EvSTWDone: + e.G = 0 + case EvGoEnd, EvGoStop, EvGoSched, EvGoPreempt, + EvGoSleep, EvGoBlock, EvGoBlockSend, EvGoBlockRecv, + EvGoBlockSelect, EvGoBlockSync, EvGoBlockCond, EvGoBlockNet, + EvGoSysBlock, EvGoBlockGC: + lastG = 0 + case EvGoSysExit, EvGoWaiting, EvGoInSyscall: + e.G = e.Args[0] + case EvUserTaskCreate: + // e.Args 0: taskID, 1:parentID, 2:nameID + e.SArgs = []string{strings[e.Args[2]]} + case EvUserRegion: + // e.Args 0: taskID, 1: mode, 2:nameID + e.SArgs = []string{strings[e.Args[2]]} + case EvUserLog: + // e.Args 0: taskID, 1:keyID, 2: stackID + e.SArgs = []string{strings[e.Args[1]], raw.sargs[0]} + case EvCPUSample: + e.Ts = int64(e.Args[0]) + e.P = int(e.Args[1]) + e.G = e.Args[2] + e.Args[0] = 0 + } + switch raw.typ { + default: + batches[lastP] = append(batches[lastP], e) + case EvCPUSample: + // Most events are written out by the active P at the exact + // moment they describe. CPU profile samples are different + // because they're written to the tracing log after some delay, + // by a separate worker goroutine, into a separate buffer. + // + // We keep these in their own batch until all of the batches are + // merged in timestamp order. We also (right before the merge) + // re-sort these events by the timestamp captured in the + // profiling signal handler. + batches[ProfileP] = append(batches[ProfileP], e) + } + } + } + if len(batches) == 0 { + err = fmt.Errorf("trace is empty") + return + } + if ticksPerSec == 0 { + err = fmt.Errorf("no EvFrequency event") + return + } + if BreakTimestampsForTesting { + var batchArr [][]*Event + for _, batch := range batches { + batchArr = append(batchArr, batch) + } + for i := 0; i < 5; i++ { + batch := batchArr[rand.Intn(len(batchArr))] + batch[rand.Intn(len(batch))].Ts += int64(rand.Intn(2000) - 1000) + } + } + if ver < 1007 { + events, err = order1005(batches) + } else { + events, err = order1007(batches) + } + if err != nil { + return + } + + // Translate cpu ticks to real time. + minTs := events[0].Ts + // Use floating point to avoid integer overflows. + freq := 1e9 / float64(ticksPerSec) + for _, ev := range events { + ev.Ts = int64(float64(ev.Ts-minTs) * freq) + // Move timers and syscalls to separate fake Ps. + if timerGoids[ev.G] && ev.Type == EvGoUnblock { + ev.P = TimerP + } + if ev.Type == EvGoSysExit { + ev.P = SyscallP + } + } + + return +} + +// removeFutile removes all constituents of futile wakeups (block, unblock, start). +// For example, a goroutine was unblocked on a mutex, but another goroutine got +// ahead and acquired the mutex before the first goroutine is scheduled, +// so the first goroutine has to block again. Such wakeups happen on buffered +// channels and sync.Mutex, but are generally not interesting for end user. +func removeFutile(events []*Event) []*Event { + // Two non-trivial aspects: + // 1. A goroutine can be preempted during a futile wakeup and migrate to another P. + // We want to remove all of that. + // 2. Tracing can start in the middle of a futile wakeup. + // That is, we can see a futile wakeup event w/o the actual wakeup before it. + // postProcessTrace runs after us and ensures that we leave the trace in a consistent state. + + // Phase 1: determine futile wakeup sequences. + type G struct { + futile bool + wakeup []*Event // wakeup sequence (subject for removal) + } + gs := make(map[uint64]G) + futile := make(map[*Event]bool) + for _, ev := range events { + switch ev.Type { + case EvGoUnblock: + g := gs[ev.Args[0]] + g.wakeup = []*Event{ev} + gs[ev.Args[0]] = g + case EvGoStart, EvGoPreempt, EvFutileWakeup: + g := gs[ev.G] + g.wakeup = append(g.wakeup, ev) + if ev.Type == EvFutileWakeup { + g.futile = true + } + gs[ev.G] = g + case EvGoBlock, EvGoBlockSend, EvGoBlockRecv, EvGoBlockSelect, EvGoBlockSync, EvGoBlockCond: + g := gs[ev.G] + if g.futile { + futile[ev] = true + for _, ev1 := range g.wakeup { + futile[ev1] = true + } + } + delete(gs, ev.G) + } + } + + // Phase 2: remove futile wakeup sequences. + newEvents := events[:0] // overwrite the original slice + for _, ev := range events { + if !futile[ev] { + newEvents = append(newEvents, ev) + } + } + return newEvents +} + +// ErrTimeOrder is returned by Parse when the trace contains +// time stamps that do not respect actual event ordering. +var ErrTimeOrder = fmt.Errorf("time stamps out of order") + +// postProcessTrace does inter-event verification and information restoration. +// The resulting trace is guaranteed to be consistent +// (for example, a P does not run two Gs at the same time, or a G is indeed +// blocked before an unblock event). +func postProcessTrace(ver int, events []*Event) error { + const ( + gDead = iota + gRunnable + gRunning + gWaiting + ) + type gdesc struct { + state int + ev *Event + evStart *Event + evCreate *Event + evMarkAssist *Event + } + type pdesc struct { + running bool + g uint64 + evSTW *Event + evSweep *Event + } + + gs := make(map[uint64]gdesc) + ps := make(map[int]pdesc) + tasks := make(map[uint64]*Event) // task id to task creation events + activeRegions := make(map[uint64][]*Event) // goroutine id to stack of regions + gs[0] = gdesc{state: gRunning} + var evGC, evSTW *Event + + checkRunning := func(p pdesc, g gdesc, ev *Event, allowG0 bool) error { + name := EventDescriptions[ev.Type].Name + if g.state != gRunning { + return fmt.Errorf("g %v is not running while %v (offset %v, time %v)", ev.G, name, ev.Off, ev.Ts) + } + if p.g != ev.G { + return fmt.Errorf("p %v is not running g %v while %v (offset %v, time %v)", ev.P, ev.G, name, ev.Off, ev.Ts) + } + if !allowG0 && ev.G == 0 { + return fmt.Errorf("g 0 did %v (offset %v, time %v)", EventDescriptions[ev.Type].Name, ev.Off, ev.Ts) + } + return nil + } + + for _, ev := range events { + g := gs[ev.G] + p := ps[ev.P] + + switch ev.Type { + case EvProcStart: + if p.running { + return fmt.Errorf("p %v is running before start (offset %v, time %v)", ev.P, ev.Off, ev.Ts) + } + p.running = true + case EvProcStop: + if !p.running { + return fmt.Errorf("p %v is not running before stop (offset %v, time %v)", ev.P, ev.Off, ev.Ts) + } + if p.g != 0 { + return fmt.Errorf("p %v is running a goroutine %v during stop (offset %v, time %v)", ev.P, p.g, ev.Off, ev.Ts) + } + p.running = false + case EvGCStart: + if evGC != nil { + return fmt.Errorf("previous GC is not ended before a new one (offset %v, time %v)", ev.Off, ev.Ts) + } + evGC = ev + // Attribute this to the global GC state. + ev.P = GCP + case EvGCDone: + if evGC == nil { + return fmt.Errorf("bogus GC end (offset %v, time %v)", ev.Off, ev.Ts) + } + evGC.Link = ev + evGC = nil + case EvSTWStart: + evp := &evSTW + if ver < 1010 { + // Before 1.10, EvSTWStart was per-P. + evp = &p.evSTW + } + if *evp != nil { + return fmt.Errorf("previous STW is not ended before a new one (offset %v, time %v)", ev.Off, ev.Ts) + } + *evp = ev + case EvSTWDone: + evp := &evSTW + if ver < 1010 { + // Before 1.10, EvSTWDone was per-P. + evp = &p.evSTW + } + if *evp == nil { + return fmt.Errorf("bogus STW end (offset %v, time %v)", ev.Off, ev.Ts) + } + (*evp).Link = ev + *evp = nil + case EvGCSweepStart: + if p.evSweep != nil { + return fmt.Errorf("previous sweeping is not ended before a new one (offset %v, time %v)", ev.Off, ev.Ts) + } + p.evSweep = ev + case EvGCMarkAssistStart: + if g.evMarkAssist != nil { + return fmt.Errorf("previous mark assist is not ended before a new one (offset %v, time %v)", ev.Off, ev.Ts) + } + g.evMarkAssist = ev + case EvGCMarkAssistDone: + // Unlike most events, mark assists can be in progress when a + // goroutine starts tracing, so we can't report an error here. + if g.evMarkAssist != nil { + g.evMarkAssist.Link = ev + g.evMarkAssist = nil + } + case EvGCSweepDone: + if p.evSweep == nil { + return fmt.Errorf("bogus sweeping end (offset %v, time %v)", ev.Off, ev.Ts) + } + p.evSweep.Link = ev + p.evSweep = nil + case EvGoWaiting: + if g.state != gRunnable { + return fmt.Errorf("g %v is not runnable before EvGoWaiting (offset %v, time %v)", ev.G, ev.Off, ev.Ts) + } + g.state = gWaiting + g.ev = ev + case EvGoInSyscall: + if g.state != gRunnable { + return fmt.Errorf("g %v is not runnable before EvGoInSyscall (offset %v, time %v)", ev.G, ev.Off, ev.Ts) + } + g.state = gWaiting + g.ev = ev + case EvGoCreate: + if err := checkRunning(p, g, ev, true); err != nil { + return err + } + if _, ok := gs[ev.Args[0]]; ok { + return fmt.Errorf("g %v already exists (offset %v, time %v)", ev.Args[0], ev.Off, ev.Ts) + } + gs[ev.Args[0]] = gdesc{state: gRunnable, ev: ev, evCreate: ev} + case EvGoStart, EvGoStartLabel: + if g.state != gRunnable { + return fmt.Errorf("g %v is not runnable before start (offset %v, time %v)", ev.G, ev.Off, ev.Ts) + } + if p.g != 0 { + return fmt.Errorf("p %v is already running g %v while start g %v (offset %v, time %v)", ev.P, p.g, ev.G, ev.Off, ev.Ts) + } + g.state = gRunning + g.evStart = ev + p.g = ev.G + if g.evCreate != nil { + if ver < 1007 { + // +1 because symbolizer expects return pc. + ev.Stk = []*Frame{{PC: g.evCreate.Args[1] + 1}} + } else { + ev.StkID = g.evCreate.Args[1] + } + g.evCreate = nil + } + + if g.ev != nil { + g.ev.Link = ev + g.ev = nil + } + case EvGoEnd, EvGoStop: + if err := checkRunning(p, g, ev, false); err != nil { + return err + } + g.evStart.Link = ev + g.evStart = nil + g.state = gDead + p.g = 0 + + if ev.Type == EvGoEnd { // flush all active regions + regions := activeRegions[ev.G] + for _, s := range regions { + s.Link = ev + } + delete(activeRegions, ev.G) + } + + case EvGoSched, EvGoPreempt: + if err := checkRunning(p, g, ev, false); err != nil { + return err + } + g.state = gRunnable + g.evStart.Link = ev + g.evStart = nil + p.g = 0 + g.ev = ev + case EvGoUnblock: + if g.state != gRunning { + return fmt.Errorf("g %v is not running while unpark (offset %v, time %v)", ev.G, ev.Off, ev.Ts) + } + if ev.P != TimerP && p.g != ev.G { + return fmt.Errorf("p %v is not running g %v while unpark (offset %v, time %v)", ev.P, ev.G, ev.Off, ev.Ts) + } + g1 := gs[ev.Args[0]] + if g1.state != gWaiting { + return fmt.Errorf("g %v is not waiting before unpark (offset %v, time %v)", ev.Args[0], ev.Off, ev.Ts) + } + if g1.ev != nil && g1.ev.Type == EvGoBlockNet && ev.P != TimerP { + ev.P = NetpollP + } + if g1.ev != nil { + g1.ev.Link = ev + } + g1.state = gRunnable + g1.ev = ev + gs[ev.Args[0]] = g1 + case EvGoSysCall: + if err := checkRunning(p, g, ev, false); err != nil { + return err + } + g.ev = ev + case EvGoSysBlock: + if err := checkRunning(p, g, ev, false); err != nil { + return err + } + g.state = gWaiting + g.evStart.Link = ev + g.evStart = nil + p.g = 0 + case EvGoSysExit: + if g.state != gWaiting { + return fmt.Errorf("g %v is not waiting during syscall exit (offset %v, time %v)", ev.G, ev.Off, ev.Ts) + } + if g.ev != nil && g.ev.Type == EvGoSysCall { + g.ev.Link = ev + } + g.state = gRunnable + g.ev = ev + case EvGoSleep, EvGoBlock, EvGoBlockSend, EvGoBlockRecv, + EvGoBlockSelect, EvGoBlockSync, EvGoBlockCond, EvGoBlockNet, EvGoBlockGC: + if err := checkRunning(p, g, ev, false); err != nil { + return err + } + g.state = gWaiting + g.ev = ev + g.evStart.Link = ev + g.evStart = nil + p.g = 0 + case EvUserTaskCreate: + taskid := ev.Args[0] + if prevEv, ok := tasks[taskid]; ok { + return fmt.Errorf("task id conflicts (id:%d), %q vs %q", taskid, ev, prevEv) + } + tasks[ev.Args[0]] = ev + case EvUserTaskEnd: + taskid := ev.Args[0] + if taskCreateEv, ok := tasks[taskid]; ok { + taskCreateEv.Link = ev + delete(tasks, taskid) + } + case EvUserRegion: + mode := ev.Args[1] + regions := activeRegions[ev.G] + if mode == 0 { // region start + activeRegions[ev.G] = append(regions, ev) // push + } else if mode == 1 { // region end + n := len(regions) + if n > 0 { // matching region start event is in the trace. + s := regions[n-1] + if s.Args[0] != ev.Args[0] || s.SArgs[0] != ev.SArgs[0] { // task id, region name mismatch + return fmt.Errorf("misuse of region in goroutine %d: span end %q when the inner-most active span start event is %q", ev.G, ev, s) + } + // Link region start event with span end event + s.Link = ev + + if n > 1 { + activeRegions[ev.G] = regions[:n-1] + } else { + delete(activeRegions, ev.G) + } + } + } else { + return fmt.Errorf("invalid user region mode: %q", ev) + } + } + + gs[ev.G] = g + ps[ev.P] = p + } + + // TODO(dvyukov): restore stacks for EvGoStart events. + // TODO(dvyukov): test that all EvGoStart events has non-nil Link. + + return nil +} + +// symbolize attaches func/file/line info to stack traces. +func symbolize(events []*Event, bin string) error { + // First, collect and dedup all pcs. + pcs := make(map[uint64]*Frame) + for _, ev := range events { + for _, f := range ev.Stk { + pcs[f.PC] = nil + } + } + + // Start addr2line. + cmd := exec.Command(goCmd(), "tool", "addr2line", bin) + in, err := cmd.StdinPipe() + if err != nil { + return fmt.Errorf("failed to pipe addr2line stdin: %v", err) + } + cmd.Stderr = os.Stderr + out, err := cmd.StdoutPipe() + if err != nil { + return fmt.Errorf("failed to pipe addr2line stdout: %v", err) + } + err = cmd.Start() + if err != nil { + return fmt.Errorf("failed to start addr2line: %v", err) + } + outb := bufio.NewReader(out) + + // Write all pcs to addr2line. + // Need to copy pcs to an array, because map iteration order is non-deterministic. + var pcArray []uint64 + for pc := range pcs { + pcArray = append(pcArray, pc) + _, err := fmt.Fprintf(in, "0x%x\n", pc-1) + if err != nil { + return fmt.Errorf("failed to write to addr2line: %v", err) + } + } + in.Close() + + // Read in answers. + for _, pc := range pcArray { + fn, err := outb.ReadString('\n') + if err != nil { + return fmt.Errorf("failed to read from addr2line: %v", err) + } + file, err := outb.ReadString('\n') + if err != nil { + return fmt.Errorf("failed to read from addr2line: %v", err) + } + f := &Frame{PC: pc} + f.Fn = fn[:len(fn)-1] + f.File = file[:len(file)-1] + if colon := strings.LastIndex(f.File, ":"); colon != -1 { + ln, err := strconv.Atoi(f.File[colon+1:]) + if err == nil { + f.File = f.File[:colon] + f.Line = ln + } + } + pcs[pc] = f + } + cmd.Wait() + + // Replace frames in events array. + for _, ev := range events { + for i, f := range ev.Stk { + ev.Stk[i] = pcs[f.PC] + } + } + + return nil +} + +// readVal reads unsigned base-128 value from r. +func readVal(r io.Reader, off0 int) (v uint64, off int, err error) { + off = off0 + for i := 0; i < 10; i++ { + var buf [1]byte + var n int + n, err = r.Read(buf[:]) + if err != nil || n != 1 { + return 0, 0, fmt.Errorf("failed to read trace at offset %d: read %v, error %v", off0, n, err) + } + off++ + v |= uint64(buf[0]&0x7f) << (uint(i) * 7) + if buf[0]&0x80 == 0 { + return + } + } + return 0, 0, fmt.Errorf("bad value at offset 0x%x", off0) +} + +// Print dumps events to stdout. For debugging. +func Print(events []*Event) { + for _, ev := range events { + PrintEvent(ev) + } +} + +// PrintEvent dumps the event to stdout. For debugging. +func PrintEvent(ev *Event) { + fmt.Printf("%s\n", ev) +} + +func (ev *Event) String() string { + desc := EventDescriptions[ev.Type] + w := new(strings.Builder) + fmt.Fprintf(w, "%v %v p=%v g=%v off=%v", ev.Ts, desc.Name, ev.P, ev.G, ev.Off) + for i, a := range desc.Args { + fmt.Fprintf(w, " %v=%v", a, ev.Args[i]) + } + for i, a := range desc.SArgs { + fmt.Fprintf(w, " %v=%v", a, ev.SArgs[i]) + } + return w.String() +} + +// argNum returns total number of args for the event accounting for timestamps, +// sequence numbers and differences between trace format versions. +func argNum(raw rawEvent, ver int) int { + desc := EventDescriptions[raw.typ] + if raw.typ == EvStack { + return len(raw.args) + } + narg := len(desc.Args) + if desc.Stack { + narg++ + } + switch raw.typ { + case EvBatch, EvFrequency, EvTimerGoroutine: + if ver < 1007 { + narg++ // there was an unused arg before 1.7 + } + return narg + } + narg++ // timestamp + if ver < 1007 { + narg++ // sequence + } + switch raw.typ { + case EvGCSweepDone: + if ver < 1009 { + narg -= 2 // 1.9 added two arguments + } + case EvGCStart, EvGoStart, EvGoUnblock: + if ver < 1007 { + narg-- // 1.7 added an additional seq arg + } + case EvSTWStart: + if ver < 1010 { + narg-- // 1.10 added an argument + } + } + return narg +} + +// BreakTimestampsForTesting causes the parser to randomly alter timestamps (for testing of broken cputicks). +var BreakTimestampsForTesting bool + +// Event types in the trace. +// Verbatim copy from src/runtime/trace.go with the "trace" prefix removed. +const ( + EvNone = 0 // unused + EvBatch = 1 // start of per-P batch of events [pid, timestamp] + EvFrequency = 2 // contains tracer timer frequency [frequency (ticks per second)] + EvStack = 3 // stack [stack id, number of PCs, array of {PC, func string ID, file string ID, line}] + EvGomaxprocs = 4 // current value of GOMAXPROCS [timestamp, GOMAXPROCS, stack id] + EvProcStart = 5 // start of P [timestamp, thread id] + EvProcStop = 6 // stop of P [timestamp] + EvGCStart = 7 // GC start [timestamp, seq, stack id] + EvGCDone = 8 // GC done [timestamp] + EvSTWStart = 9 // GC mark termination start [timestamp, kind] + EvSTWDone = 10 // GC mark termination done [timestamp] + EvGCSweepStart = 11 // GC sweep start [timestamp, stack id] + EvGCSweepDone = 12 // GC sweep done [timestamp, swept, reclaimed] + EvGoCreate = 13 // goroutine creation [timestamp, new goroutine id, new stack id, stack id] + EvGoStart = 14 // goroutine starts running [timestamp, goroutine id, seq] + EvGoEnd = 15 // goroutine ends [timestamp] + EvGoStop = 16 // goroutine stops (like in select{}) [timestamp, stack] + EvGoSched = 17 // goroutine calls Gosched [timestamp, stack] + EvGoPreempt = 18 // goroutine is preempted [timestamp, stack] + EvGoSleep = 19 // goroutine calls Sleep [timestamp, stack] + EvGoBlock = 20 // goroutine blocks [timestamp, stack] + EvGoUnblock = 21 // goroutine is unblocked [timestamp, goroutine id, seq, stack] + EvGoBlockSend = 22 // goroutine blocks on chan send [timestamp, stack] + EvGoBlockRecv = 23 // goroutine blocks on chan recv [timestamp, stack] + EvGoBlockSelect = 24 // goroutine blocks on select [timestamp, stack] + EvGoBlockSync = 25 // goroutine blocks on Mutex/RWMutex [timestamp, stack] + EvGoBlockCond = 26 // goroutine blocks on Cond [timestamp, stack] + EvGoBlockNet = 27 // goroutine blocks on network [timestamp, stack] + EvGoSysCall = 28 // syscall enter [timestamp, stack] + EvGoSysExit = 29 // syscall exit [timestamp, goroutine id, seq, real timestamp] + EvGoSysBlock = 30 // syscall blocks [timestamp] + EvGoWaiting = 31 // denotes that goroutine is blocked when tracing starts [timestamp, goroutine id] + EvGoInSyscall = 32 // denotes that goroutine is in syscall when tracing starts [timestamp, goroutine id] + EvHeapAlloc = 33 // gcController.heapLive change [timestamp, heap live bytes] + EvHeapGoal = 34 // gcController.heapGoal change [timestamp, heap goal bytes] + EvTimerGoroutine = 35 // denotes timer goroutine [timer goroutine id] + EvFutileWakeup = 36 // denotes that the previous wakeup of this goroutine was futile [timestamp] + EvString = 37 // string dictionary entry [ID, length, string] + EvGoStartLocal = 38 // goroutine starts running on the same P as the last event [timestamp, goroutine id] + EvGoUnblockLocal = 39 // goroutine is unblocked on the same P as the last event [timestamp, goroutine id, stack] + EvGoSysExitLocal = 40 // syscall exit on the same P as the last event [timestamp, goroutine id, real timestamp] + EvGoStartLabel = 41 // goroutine starts running with label [timestamp, goroutine id, seq, label string id] + EvGoBlockGC = 42 // goroutine blocks on GC assist [timestamp, stack] + EvGCMarkAssistStart = 43 // GC mark assist start [timestamp, stack] + EvGCMarkAssistDone = 44 // GC mark assist done [timestamp] + EvUserTaskCreate = 45 // trace.NewTask [timestamp, internal task id, internal parent id, name string, stack] + EvUserTaskEnd = 46 // end of task [timestamp, internal task id, stack] + EvUserRegion = 47 // trace.WithRegion [timestamp, internal task id, mode(0:start, 1:end), name string, stack] + EvUserLog = 48 // trace.Log [timestamp, internal id, key string id, stack, value string] + EvCPUSample = 49 // CPU profiling sample [timestamp, real timestamp, real P id (-1 when absent), goroutine id, stack] + EvCount = 50 +) + +var EventDescriptions = [EvCount]struct { + Name string + minVersion int + Stack bool + Args []string + SArgs []string // string arguments +}{ + EvNone: {"None", 1005, false, []string{}, nil}, + EvBatch: {"Batch", 1005, false, []string{"p", "ticks"}, nil}, // in 1.5 format it was {"p", "seq", "ticks"} + EvFrequency: {"Frequency", 1005, false, []string{"freq"}, nil}, // in 1.5 format it was {"freq", "unused"} + EvStack: {"Stack", 1005, false, []string{"id", "siz"}, nil}, + EvGomaxprocs: {"Gomaxprocs", 1005, true, []string{"procs"}, nil}, + EvProcStart: {"ProcStart", 1005, false, []string{"thread"}, nil}, + EvProcStop: {"ProcStop", 1005, false, []string{}, nil}, + EvGCStart: {"GCStart", 1005, true, []string{"seq"}, nil}, // in 1.5 format it was {} + EvGCDone: {"GCDone", 1005, false, []string{}, nil}, + EvSTWStart: {"STWStart", 1005, false, []string{"kindid"}, []string{"kind"}}, // <= 1.9, args was {} (implicitly {0}) + EvSTWDone: {"STWDone", 1005, false, []string{}, nil}, + EvGCSweepStart: {"GCSweepStart", 1005, true, []string{}, nil}, + EvGCSweepDone: {"GCSweepDone", 1005, false, []string{"swept", "reclaimed"}, nil}, // before 1.9, format was {} + EvGoCreate: {"GoCreate", 1005, true, []string{"g", "stack"}, nil}, + EvGoStart: {"GoStart", 1005, false, []string{"g", "seq"}, nil}, // in 1.5 format it was {"g"} + EvGoEnd: {"GoEnd", 1005, false, []string{}, nil}, + EvGoStop: {"GoStop", 1005, true, []string{}, nil}, + EvGoSched: {"GoSched", 1005, true, []string{}, nil}, + EvGoPreempt: {"GoPreempt", 1005, true, []string{}, nil}, + EvGoSleep: {"GoSleep", 1005, true, []string{}, nil}, + EvGoBlock: {"GoBlock", 1005, true, []string{}, nil}, + EvGoUnblock: {"GoUnblock", 1005, true, []string{"g", "seq"}, nil}, // in 1.5 format it was {"g"} + EvGoBlockSend: {"GoBlockSend", 1005, true, []string{}, nil}, + EvGoBlockRecv: {"GoBlockRecv", 1005, true, []string{}, nil}, + EvGoBlockSelect: {"GoBlockSelect", 1005, true, []string{}, nil}, + EvGoBlockSync: {"GoBlockSync", 1005, true, []string{}, nil}, + EvGoBlockCond: {"GoBlockCond", 1005, true, []string{}, nil}, + EvGoBlockNet: {"GoBlockNet", 1005, true, []string{}, nil}, + EvGoSysCall: {"GoSysCall", 1005, true, []string{}, nil}, + EvGoSysExit: {"GoSysExit", 1005, false, []string{"g", "seq", "ts"}, nil}, + EvGoSysBlock: {"GoSysBlock", 1005, false, []string{}, nil}, + EvGoWaiting: {"GoWaiting", 1005, false, []string{"g"}, nil}, + EvGoInSyscall: {"GoInSyscall", 1005, false, []string{"g"}, nil}, + EvHeapAlloc: {"HeapAlloc", 1005, false, []string{"mem"}, nil}, + EvHeapGoal: {"HeapGoal", 1005, false, []string{"mem"}, nil}, + EvTimerGoroutine: {"TimerGoroutine", 1005, false, []string{"g"}, nil}, // in 1.5 format it was {"g", "unused"} + EvFutileWakeup: {"FutileWakeup", 1005, false, []string{}, nil}, + EvString: {"String", 1007, false, []string{}, nil}, + EvGoStartLocal: {"GoStartLocal", 1007, false, []string{"g"}, nil}, + EvGoUnblockLocal: {"GoUnblockLocal", 1007, true, []string{"g"}, nil}, + EvGoSysExitLocal: {"GoSysExitLocal", 1007, false, []string{"g", "ts"}, nil}, + EvGoStartLabel: {"GoStartLabel", 1008, false, []string{"g", "seq", "labelid"}, []string{"label"}}, + EvGoBlockGC: {"GoBlockGC", 1008, true, []string{}, nil}, + EvGCMarkAssistStart: {"GCMarkAssistStart", 1009, true, []string{}, nil}, + EvGCMarkAssistDone: {"GCMarkAssistDone", 1009, false, []string{}, nil}, + EvUserTaskCreate: {"UserTaskCreate", 1011, true, []string{"taskid", "pid", "typeid"}, []string{"name"}}, + EvUserTaskEnd: {"UserTaskEnd", 1011, true, []string{"taskid"}, nil}, + EvUserRegion: {"UserRegion", 1011, true, []string{"taskid", "mode", "typeid"}, []string{"name"}}, + EvUserLog: {"UserLog", 1011, true, []string{"id", "keyid"}, []string{"category", "message"}}, + EvCPUSample: {"CPUSample", 1019, true, []string{"ts", "p", "g"}, nil}, +} + +// Copied from src/runtime/proc.go:stwReasonStrings in Go 1.21. +var stwReasonStringsGo121 = [...]string{ + "unknown", + "GC mark termination", + "GC sweep termination", + "write heap dump", + "goroutine profile", + "goroutine profile cleanup", + "all goroutines stack trace", + "read mem stats", + "AllThreadsSyscall", + "GOMAXPROCS", + "start trace", + "stop trace", + "CountPagesInUse (test)", + "ReadMetricsSlow (test)", + "ReadMemStatsSlow (test)", + "PageCachePagesLeaked (test)", + "ResetDebugLog (test)", +} diff --git a/src/internal/trace/parser_test.go b/src/internal/trace/parser_test.go new file mode 100644 index 0000000..fce660c --- /dev/null +++ b/src/internal/trace/parser_test.go @@ -0,0 +1,123 @@ +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import ( + "bytes" + "os" + "path/filepath" + "strings" + "testing" +) + +func TestCorruptedInputs(t *testing.T) { + // These inputs crashed parser previously. + tests := []string{ + "gotrace\x00\x020", + "gotrace\x00Q00\x020", + "gotrace\x00T00\x020", + "gotrace\x00\xc3\x0200", + "go 1.5 trace\x00\x00\x00\x00\x020", + "go 1.5 trace\x00\x00\x00\x00Q00\x020", + "go 1.5 trace\x00\x00\x00\x00T00\x020", + "go 1.5 trace\x00\x00\x00\x00\xc3\x0200", + } + for _, data := range tests { + res, err := Parse(strings.NewReader(data), "") + if err == nil || res.Events != nil || res.Stacks != nil { + t.Fatalf("no error on input: %q", data) + } + } +} + +func TestParseCanned(t *testing.T) { + files, err := os.ReadDir("./testdata") + if err != nil { + t.Fatalf("failed to read ./testdata: %v", err) + } + for _, f := range files { + info, err := f.Info() + if err != nil { + t.Fatal(err) + } + if testing.Short() && info.Size() > 10000 { + continue + } + name := filepath.Join("./testdata", f.Name()) + data, err := os.ReadFile(name) + if err != nil { + t.Fatal(err) + } + // Instead of Parse that requires a proper binary name for old traces, + // we use 'parse' that omits symbol lookup if an empty string is given. + ver, res, err := parse(bytes.NewReader(data), "") + switch { + case strings.HasSuffix(f.Name(), "_good"): + if err != nil { + t.Errorf("failed to parse good trace %v: %v", f.Name(), err) + } + checkTrace(t, ver, res) + case strings.HasSuffix(f.Name(), "_unordered"): + if err != ErrTimeOrder { + t.Errorf("unordered trace is not detected %v: %v", f.Name(), err) + } + default: + t.Errorf("unknown input file suffix: %v", f.Name()) + } + } +} + +// checkTrace walks over a good trace and makes a bunch of additional checks +// that may not cause the parser to outright fail. +func checkTrace(t *testing.T, ver int, res ParseResult) { + for _, ev := range res.Events { + if ver >= 1021 { + if ev.Type == EvSTWStart && ev.SArgs[0] == "unknown" { + t.Errorf("found unknown STW event; update stwReasonStrings?") + } + } + } +} + +func TestParseVersion(t *testing.T) { + tests := map[string]int{ + "go 1.5 trace\x00\x00\x00\x00": 1005, + "go 1.7 trace\x00\x00\x00\x00": 1007, + "go 1.10 trace\x00\x00\x00": 1010, + "go 1.25 trace\x00\x00\x00": 1025, + "go 1.234 trace\x00\x00": 1234, + "go 1.2345 trace\x00": -1, + "go 0.0 trace\x00\x00\x00\x00": -1, + "go a.b trace\x00\x00\x00\x00": -1, + } + for header, ver := range tests { + ver1, err := parseHeader([]byte(header)) + if ver == -1 { + if err == nil { + t.Fatalf("no error on input: %q, version %v", header, ver1) + } + } else { + if err != nil { + t.Fatalf("failed to parse: %q (%v)", header, err) + } + if ver != ver1 { + t.Fatalf("wrong version: %v, want %v, input: %q", ver1, ver, header) + } + } + } +} + +func TestTimestampOverflow(t *testing.T) { + // Test that parser correctly handles large timestamps (long tracing). + w := NewWriter() + w.Emit(EvBatch, 0, 0) + w.Emit(EvFrequency, 1e9) + for ts := uint64(1); ts < 1e16; ts *= 2 { + w.Emit(EvGoCreate, ts, ts, 0, 0) + } + if _, err := Parse(w, ""); err != nil { + t.Fatalf("failed to parse: %v", err) + } +} diff --git a/src/internal/trace/testdata/http_1_10_good b/src/internal/trace/testdata/http_1_10_good Binary files differnew file mode 100644 index 0000000..a4f2ed8 --- /dev/null +++ b/src/internal/trace/testdata/http_1_10_good diff --git a/src/internal/trace/testdata/http_1_11_good b/src/internal/trace/testdata/http_1_11_good Binary files differnew file mode 100644 index 0000000..0efcc6f --- /dev/null +++ b/src/internal/trace/testdata/http_1_11_good diff --git a/src/internal/trace/testdata/http_1_19_good b/src/internal/trace/testdata/http_1_19_good Binary files differnew file mode 100644 index 0000000..c1d519e --- /dev/null +++ b/src/internal/trace/testdata/http_1_19_good diff --git a/src/internal/trace/testdata/http_1_21_good b/src/internal/trace/testdata/http_1_21_good Binary files differnew file mode 100644 index 0000000..b3295f9 --- /dev/null +++ b/src/internal/trace/testdata/http_1_21_good diff --git a/src/internal/trace/testdata/http_1_5_good b/src/internal/trace/testdata/http_1_5_good Binary files differnew file mode 100644 index 0000000..0736cae --- /dev/null +++ b/src/internal/trace/testdata/http_1_5_good diff --git a/src/internal/trace/testdata/http_1_7_good b/src/internal/trace/testdata/http_1_7_good Binary files differnew file mode 100644 index 0000000..b0e318e --- /dev/null +++ b/src/internal/trace/testdata/http_1_7_good diff --git a/src/internal/trace/testdata/http_1_9_good b/src/internal/trace/testdata/http_1_9_good Binary files differnew file mode 100644 index 0000000..ca89278 --- /dev/null +++ b/src/internal/trace/testdata/http_1_9_good diff --git a/src/internal/trace/testdata/stress_1_10_good b/src/internal/trace/testdata/stress_1_10_good Binary files differnew file mode 100644 index 0000000..19778b0 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_10_good diff --git a/src/internal/trace/testdata/stress_1_11_good b/src/internal/trace/testdata/stress_1_11_good Binary files differnew file mode 100644 index 0000000..6468d89 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_11_good diff --git a/src/internal/trace/testdata/stress_1_19_good b/src/internal/trace/testdata/stress_1_19_good Binary files differnew file mode 100644 index 0000000..13f5926 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_19_good diff --git a/src/internal/trace/testdata/stress_1_21_good b/src/internal/trace/testdata/stress_1_21_good Binary files differnew file mode 100644 index 0000000..1ade5e0 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_21_good diff --git a/src/internal/trace/testdata/stress_1_5_good b/src/internal/trace/testdata/stress_1_5_good Binary files differnew file mode 100644 index 0000000..c5055eb --- /dev/null +++ b/src/internal/trace/testdata/stress_1_5_good diff --git a/src/internal/trace/testdata/stress_1_5_unordered b/src/internal/trace/testdata/stress_1_5_unordered Binary files differnew file mode 100644 index 0000000..11f7d74 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_5_unordered diff --git a/src/internal/trace/testdata/stress_1_7_good b/src/internal/trace/testdata/stress_1_7_good Binary files differnew file mode 100644 index 0000000..b4d927d --- /dev/null +++ b/src/internal/trace/testdata/stress_1_7_good diff --git a/src/internal/trace/testdata/stress_1_9_good b/src/internal/trace/testdata/stress_1_9_good Binary files differnew file mode 100644 index 0000000..dcf17f1 --- /dev/null +++ b/src/internal/trace/testdata/stress_1_9_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_10_good b/src/internal/trace/testdata/stress_start_stop_1_10_good Binary files differnew file mode 100644 index 0000000..b908e10 --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_10_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_11_good b/src/internal/trace/testdata/stress_start_stop_1_11_good Binary files differnew file mode 100644 index 0000000..457f01a --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_11_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_19_good b/src/internal/trace/testdata/stress_start_stop_1_19_good Binary files differnew file mode 100644 index 0000000..92d9278 --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_19_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_21_good b/src/internal/trace/testdata/stress_start_stop_1_21_good Binary files differnew file mode 100644 index 0000000..fff46a9 --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_21_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_5_good b/src/internal/trace/testdata/stress_start_stop_1_5_good Binary files differnew file mode 100644 index 0000000..72a887b --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_5_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_7_good b/src/internal/trace/testdata/stress_start_stop_1_7_good Binary files differnew file mode 100644 index 0000000..c23ed7d --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_7_good diff --git a/src/internal/trace/testdata/stress_start_stop_1_9_good b/src/internal/trace/testdata/stress_start_stop_1_9_good Binary files differnew file mode 100644 index 0000000..f00f190 --- /dev/null +++ b/src/internal/trace/testdata/stress_start_stop_1_9_good diff --git a/src/internal/trace/testdata/user_task_region_1_11_good b/src/internal/trace/testdata/user_task_region_1_11_good Binary files differnew file mode 100644 index 0000000..f4edb67 --- /dev/null +++ b/src/internal/trace/testdata/user_task_region_1_11_good diff --git a/src/internal/trace/testdata/user_task_region_1_19_good b/src/internal/trace/testdata/user_task_region_1_19_good Binary files differnew file mode 100644 index 0000000..1daa3b2 --- /dev/null +++ b/src/internal/trace/testdata/user_task_region_1_19_good diff --git a/src/internal/trace/testdata/user_task_region_1_21_good b/src/internal/trace/testdata/user_task_region_1_21_good Binary files differnew file mode 100644 index 0000000..5c01a64 --- /dev/null +++ b/src/internal/trace/testdata/user_task_region_1_21_good diff --git a/src/internal/trace/writer.go b/src/internal/trace/writer.go new file mode 100644 index 0000000..dd0b9f1 --- /dev/null +++ b/src/internal/trace/writer.go @@ -0,0 +1,49 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package trace + +import "bytes" + +// Writer is a test trace writer. +type Writer struct { + bytes.Buffer +} + +func NewWriter() *Writer { + w := new(Writer) + w.Write([]byte("go 1.9 trace\x00\x00\x00\x00")) + return w +} + +// Emit writes an event record to the trace. +// See Event types for valid types and required arguments. +func (w *Writer) Emit(typ byte, args ...uint64) { + nargs := byte(len(args)) - 1 + if nargs > 3 { + nargs = 3 + } + buf := []byte{typ | nargs<<6} + if nargs == 3 { + buf = append(buf, 0) + } + for _, a := range args { + buf = appendVarint(buf, a) + } + if nargs == 3 { + buf[1] = byte(len(buf) - 2) + } + n, err := w.Write(buf) + if n != len(buf) || err != nil { + panic("failed to write") + } +} + +func appendVarint(buf []byte, v uint64) []byte { + for ; v >= 0x80; v >>= 7 { + buf = append(buf, 0x80|byte(v)) + } + buf = append(buf, byte(v)) + return buf +} diff --git a/src/internal/txtar/archive.go b/src/internal/txtar/archive.go new file mode 100644 index 0000000..fd95f1e --- /dev/null +++ b/src/internal/txtar/archive.go @@ -0,0 +1,140 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package txtar implements a trivial text-based file archive format. +// +// The goals for the format are: +// +// - be trivial enough to create and edit by hand. +// - be able to store trees of text files describing go command test cases. +// - diff nicely in git history and code reviews. +// +// Non-goals include being a completely general archive format, +// storing binary data, storing file modes, storing special files like +// symbolic links, and so on. +// +// # Txtar format +// +// A txtar archive is zero or more comment lines and then a sequence of file entries. +// Each file entry begins with a file marker line of the form "-- FILENAME --" +// and is followed by zero or more file content lines making up the file data. +// The comment or file content ends at the next file marker line. +// The file marker line must begin with the three-byte sequence "-- " +// and end with the three-byte sequence " --", but the enclosed +// file name can be surrounding by additional white space, +// all of which is stripped. +// +// If the txtar file is missing a trailing newline on the final line, +// parsers should consider a final newline to be present anyway. +// +// There are no possible syntax errors in a txtar archive. +package txtar + +import ( + "bytes" + "fmt" + "os" + "strings" +) + +// An Archive is a collection of files. +type Archive struct { + Comment []byte + Files []File +} + +// A File is a single file in an archive. +type File struct { + Name string // name of file ("foo/bar.txt") + Data []byte // text content of file +} + +// Format returns the serialized form of an Archive. +// It is assumed that the Archive data structure is well-formed: +// a.Comment and all a.File[i].Data contain no file marker lines, +// and all a.File[i].Name is non-empty. +func Format(a *Archive) []byte { + var buf bytes.Buffer + buf.Write(fixNL(a.Comment)) + for _, f := range a.Files { + fmt.Fprintf(&buf, "-- %s --\n", f.Name) + buf.Write(fixNL(f.Data)) + } + return buf.Bytes() +} + +// ParseFile parses the named file as an archive. +func ParseFile(file string) (*Archive, error) { + data, err := os.ReadFile(file) + if err != nil { + return nil, err + } + return Parse(data), nil +} + +// Parse parses the serialized form of an Archive. +// The returned Archive holds slices of data. +func Parse(data []byte) *Archive { + a := new(Archive) + var name string + a.Comment, name, data = findFileMarker(data) + for name != "" { + f := File{name, nil} + f.Data, name, data = findFileMarker(data) + a.Files = append(a.Files, f) + } + return a +} + +var ( + newlineMarker = []byte("\n-- ") + marker = []byte("-- ") + markerEnd = []byte(" --") +) + +// findFileMarker finds the next file marker in data, +// extracts the file name, and returns the data before the marker, +// the file name, and the data after the marker. +// If there is no next marker, findFileMarker returns before = fixNL(data), name = "", after = nil. +func findFileMarker(data []byte) (before []byte, name string, after []byte) { + var i int + for { + if name, after = isMarker(data[i:]); name != "" { + return data[:i], name, after + } + j := bytes.Index(data[i:], newlineMarker) + if j < 0 { + return fixNL(data), "", nil + } + i += j + 1 // positioned at start of new possible marker + } +} + +// isMarker checks whether data begins with a file marker line. +// If so, it returns the name from the line and the data after the line. +// Otherwise it returns name == "" with an unspecified after. +func isMarker(data []byte) (name string, after []byte) { + if !bytes.HasPrefix(data, marker) { + return "", nil + } + if i := bytes.IndexByte(data, '\n'); i >= 0 { + data, after = data[:i], data[i+1:] + } + if !(bytes.HasSuffix(data, markerEnd) && len(data) >= len(marker)+len(markerEnd)) { + return "", nil + } + return strings.TrimSpace(string(data[len(marker) : len(data)-len(markerEnd)])), after +} + +// If data is empty or ends in \n, fixNL returns data. +// Otherwise fixNL returns a new slice consisting of data with a final \n added. +func fixNL(data []byte) []byte { + if len(data) == 0 || data[len(data)-1] == '\n' { + return data + } + d := make([]byte, len(data)+1) + copy(d, data) + d[len(data)] = '\n' + return d +} diff --git a/src/internal/types/errors/code_string.go b/src/internal/types/errors/code_string.go new file mode 100644 index 0000000..719fc73 --- /dev/null +++ b/src/internal/types/errors/code_string.go @@ -0,0 +1,199 @@ +// Code generated by "stringer -type Code codes.go"; DO NOT EDIT. + +package errors + +import "strconv" + +func _() { + // An "invalid array index" compiler error signifies that the constant values have changed. + // Re-run the stringer command to generate them again. + var x [1]struct{} + _ = x[InvalidSyntaxTree - -1] + _ = x[Test-1] + _ = x[BlankPkgName-2] + _ = x[MismatchedPkgName-3] + _ = x[InvalidPkgUse-4] + _ = x[BadImportPath-5] + _ = x[BrokenImport-6] + _ = x[ImportCRenamed-7] + _ = x[UnusedImport-8] + _ = x[InvalidInitCycle-9] + _ = x[DuplicateDecl-10] + _ = x[InvalidDeclCycle-11] + _ = x[InvalidTypeCycle-12] + _ = x[InvalidConstInit-13] + _ = x[InvalidConstVal-14] + _ = x[InvalidConstType-15] + _ = x[UntypedNilUse-16] + _ = x[WrongAssignCount-17] + _ = x[UnassignableOperand-18] + _ = x[NoNewVar-19] + _ = x[MultiValAssignOp-20] + _ = x[InvalidIfaceAssign-21] + _ = x[InvalidChanAssign-22] + _ = x[IncompatibleAssign-23] + _ = x[UnaddressableFieldAssign-24] + _ = x[NotAType-25] + _ = x[InvalidArrayLen-26] + _ = x[BlankIfaceMethod-27] + _ = x[IncomparableMapKey-28] + _ = x[InvalidPtrEmbed-30] + _ = x[BadRecv-31] + _ = x[InvalidRecv-32] + _ = x[DuplicateFieldAndMethod-33] + _ = x[DuplicateMethod-34] + _ = x[InvalidBlank-35] + _ = x[InvalidIota-36] + _ = x[MissingInitBody-37] + _ = x[InvalidInitSig-38] + _ = x[InvalidInitDecl-39] + _ = x[InvalidMainDecl-40] + _ = x[TooManyValues-41] + _ = x[NotAnExpr-42] + _ = x[TruncatedFloat-43] + _ = x[NumericOverflow-44] + _ = x[UndefinedOp-45] + _ = x[MismatchedTypes-46] + _ = x[DivByZero-47] + _ = x[NonNumericIncDec-48] + _ = x[UnaddressableOperand-49] + _ = x[InvalidIndirection-50] + _ = x[NonIndexableOperand-51] + _ = x[InvalidIndex-52] + _ = x[SwappedSliceIndices-53] + _ = x[NonSliceableOperand-54] + _ = x[InvalidSliceExpr-55] + _ = x[InvalidShiftCount-56] + _ = x[InvalidShiftOperand-57] + _ = x[InvalidReceive-58] + _ = x[InvalidSend-59] + _ = x[DuplicateLitKey-60] + _ = x[MissingLitKey-61] + _ = x[InvalidLitIndex-62] + _ = x[OversizeArrayLit-63] + _ = x[MixedStructLit-64] + _ = x[InvalidStructLit-65] + _ = x[MissingLitField-66] + _ = x[DuplicateLitField-67] + _ = x[UnexportedLitField-68] + _ = x[InvalidLitField-69] + _ = x[UntypedLit-70] + _ = x[InvalidLit-71] + _ = x[AmbiguousSelector-72] + _ = x[UndeclaredImportedName-73] + _ = x[UnexportedName-74] + _ = x[UndeclaredName-75] + _ = x[MissingFieldOrMethod-76] + _ = x[BadDotDotDotSyntax-77] + _ = x[NonVariadicDotDotDot-78] + _ = x[MisplacedDotDotDot-79] + _ = x[InvalidDotDotDot-81] + _ = x[UncalledBuiltin-82] + _ = x[InvalidAppend-83] + _ = x[InvalidCap-84] + _ = x[InvalidClose-85] + _ = x[InvalidCopy-86] + _ = x[InvalidComplex-87] + _ = x[InvalidDelete-88] + _ = x[InvalidImag-89] + _ = x[InvalidLen-90] + _ = x[SwappedMakeArgs-91] + _ = x[InvalidMake-92] + _ = x[InvalidReal-93] + _ = x[InvalidAssert-94] + _ = x[ImpossibleAssert-95] + _ = x[InvalidConversion-96] + _ = x[InvalidUntypedConversion-97] + _ = x[BadOffsetofSyntax-98] + _ = x[InvalidOffsetof-99] + _ = x[UnusedExpr-100] + _ = x[UnusedVar-101] + _ = x[MissingReturn-102] + _ = x[WrongResultCount-103] + _ = x[OutOfScopeResult-104] + _ = x[InvalidCond-105] + _ = x[InvalidPostDecl-106] + _ = x[InvalidIterVar-108] + _ = x[InvalidRangeExpr-109] + _ = x[MisplacedBreak-110] + _ = x[MisplacedContinue-111] + _ = x[MisplacedFallthrough-112] + _ = x[DuplicateCase-113] + _ = x[DuplicateDefault-114] + _ = x[BadTypeKeyword-115] + _ = x[InvalidTypeSwitch-116] + _ = x[InvalidExprSwitch-117] + _ = x[InvalidSelectCase-118] + _ = x[UndeclaredLabel-119] + _ = x[DuplicateLabel-120] + _ = x[MisplacedLabel-121] + _ = x[UnusedLabel-122] + _ = x[JumpOverDecl-123] + _ = x[JumpIntoBlock-124] + _ = x[InvalidMethodExpr-125] + _ = x[WrongArgCount-126] + _ = x[InvalidCall-127] + _ = x[UnusedResults-128] + _ = x[InvalidDefer-129] + _ = x[InvalidGo-130] + _ = x[BadDecl-131] + _ = x[RepeatedDecl-132] + _ = x[InvalidUnsafeAdd-133] + _ = x[InvalidUnsafeSlice-134] + _ = x[UnsupportedFeature-135] + _ = x[NotAGenericType-136] + _ = x[WrongTypeArgCount-137] + _ = x[CannotInferTypeArgs-138] + _ = x[InvalidTypeArg-139] + _ = x[InvalidInstanceCycle-140] + _ = x[InvalidUnion-141] + _ = x[MisplacedConstraintIface-142] + _ = x[InvalidMethodTypeParams-143] + _ = x[MisplacedTypeParam-144] + _ = x[InvalidUnsafeSliceData-145] + _ = x[InvalidUnsafeString-146] + _ = x[InvalidClear-148] + _ = x[TypeTooLarge-149] + _ = x[InvalidMinMaxOperand-150] +} + +const ( + _Code_name_0 = "InvalidSyntaxTree" + _Code_name_1 = "TestBlankPkgNameMismatchedPkgNameInvalidPkgUseBadImportPathBrokenImportImportCRenamedUnusedImportInvalidInitCycleDuplicateDeclInvalidDeclCycleInvalidTypeCycleInvalidConstInitInvalidConstValInvalidConstTypeUntypedNilUseWrongAssignCountUnassignableOperandNoNewVarMultiValAssignOpInvalidIfaceAssignInvalidChanAssignIncompatibleAssignUnaddressableFieldAssignNotATypeInvalidArrayLenBlankIfaceMethodIncomparableMapKey" + _Code_name_2 = "InvalidPtrEmbedBadRecvInvalidRecvDuplicateFieldAndMethodDuplicateMethodInvalidBlankInvalidIotaMissingInitBodyInvalidInitSigInvalidInitDeclInvalidMainDeclTooManyValuesNotAnExprTruncatedFloatNumericOverflowUndefinedOpMismatchedTypesDivByZeroNonNumericIncDecUnaddressableOperandInvalidIndirectionNonIndexableOperandInvalidIndexSwappedSliceIndicesNonSliceableOperandInvalidSliceExprInvalidShiftCountInvalidShiftOperandInvalidReceiveInvalidSendDuplicateLitKeyMissingLitKeyInvalidLitIndexOversizeArrayLitMixedStructLitInvalidStructLitMissingLitFieldDuplicateLitFieldUnexportedLitFieldInvalidLitFieldUntypedLitInvalidLitAmbiguousSelectorUndeclaredImportedNameUnexportedNameUndeclaredNameMissingFieldOrMethodBadDotDotDotSyntaxNonVariadicDotDotDotMisplacedDotDotDot" + _Code_name_3 = "InvalidDotDotDotUncalledBuiltinInvalidAppendInvalidCapInvalidCloseInvalidCopyInvalidComplexInvalidDeleteInvalidImagInvalidLenSwappedMakeArgsInvalidMakeInvalidRealInvalidAssertImpossibleAssertInvalidConversionInvalidUntypedConversionBadOffsetofSyntaxInvalidOffsetofUnusedExprUnusedVarMissingReturnWrongResultCountOutOfScopeResultInvalidCondInvalidPostDecl" + _Code_name_4 = "InvalidIterVarInvalidRangeExprMisplacedBreakMisplacedContinueMisplacedFallthroughDuplicateCaseDuplicateDefaultBadTypeKeywordInvalidTypeSwitchInvalidExprSwitchInvalidSelectCaseUndeclaredLabelDuplicateLabelMisplacedLabelUnusedLabelJumpOverDeclJumpIntoBlockInvalidMethodExprWrongArgCountInvalidCallUnusedResultsInvalidDeferInvalidGoBadDeclRepeatedDeclInvalidUnsafeAddInvalidUnsafeSliceUnsupportedFeatureNotAGenericTypeWrongTypeArgCountCannotInferTypeArgsInvalidTypeArgInvalidInstanceCycleInvalidUnionMisplacedConstraintIfaceInvalidMethodTypeParamsMisplacedTypeParamInvalidUnsafeSliceDataInvalidUnsafeString" + _Code_name_5 = "InvalidClearTypeTooLargeInvalidMinMaxOperand" +) + +var ( + _Code_index_1 = [...]uint16{0, 4, 16, 33, 46, 59, 71, 85, 97, 113, 126, 142, 158, 174, 189, 205, 218, 234, 253, 261, 277, 295, 312, 330, 354, 362, 377, 393, 411} + _Code_index_2 = [...]uint16{0, 15, 22, 33, 56, 71, 83, 94, 109, 123, 138, 153, 166, 175, 189, 204, 215, 230, 239, 255, 275, 293, 312, 324, 343, 362, 378, 395, 414, 428, 439, 454, 467, 482, 498, 512, 528, 543, 560, 578, 593, 603, 613, 630, 652, 666, 680, 700, 718, 738, 756} + _Code_index_3 = [...]uint16{0, 16, 31, 44, 54, 66, 77, 91, 104, 115, 125, 140, 151, 162, 175, 191, 208, 232, 249, 264, 274, 283, 296, 312, 328, 339, 354} + _Code_index_4 = [...]uint16{0, 14, 30, 44, 61, 81, 94, 110, 124, 141, 158, 175, 190, 204, 218, 229, 241, 254, 271, 284, 295, 308, 320, 329, 336, 348, 364, 382, 400, 415, 432, 451, 465, 485, 497, 521, 544, 562, 584, 603} + _Code_index_5 = [...]uint8{0, 12, 24, 44} +) + +func (i Code) String() string { + switch { + case i == -1: + return _Code_name_0 + case 1 <= i && i <= 28: + i -= 1 + return _Code_name_1[_Code_index_1[i]:_Code_index_1[i+1]] + case 30 <= i && i <= 79: + i -= 30 + return _Code_name_2[_Code_index_2[i]:_Code_index_2[i+1]] + case 81 <= i && i <= 106: + i -= 81 + return _Code_name_3[_Code_index_3[i]:_Code_index_3[i+1]] + case 108 <= i && i <= 146: + i -= 108 + return _Code_name_4[_Code_index_4[i]:_Code_index_4[i+1]] + case 148 <= i && i <= 150: + i -= 148 + return _Code_name_5[_Code_index_5[i]:_Code_index_5[i+1]] + default: + return "Code(" + strconv.FormatInt(int64(i), 10) + ")" + } +} diff --git a/src/internal/types/errors/codes.go b/src/internal/types/errors/codes.go new file mode 100644 index 0000000..62358c7 --- /dev/null +++ b/src/internal/types/errors/codes.go @@ -0,0 +1,1477 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package errors + +//go:generate stringer -type Code codes.go + +type Code int + +// This file defines the error codes that can be produced during type-checking. +// Collectively, these codes provide an identifier that may be used to +// implement special handling for certain types of errors. +// +// Error code values should not be changed: add new codes at the end. +// +// Error codes should be fine-grained enough that the exact nature of the error +// can be easily determined, but coarse enough that they are not an +// implementation detail of the type checking algorithm. As a rule-of-thumb, +// errors should be considered equivalent if there is a theoretical refactoring +// of the type checker in which they are emitted in exactly one place. For +// example, the type checker emits different error messages for "too many +// arguments" and "too few arguments", but one can imagine an alternative type +// checker where this check instead just emits a single "wrong number of +// arguments", so these errors should have the same code. +// +// Error code names should be as brief as possible while retaining accuracy and +// distinctiveness. In most cases names should start with an adjective +// describing the nature of the error (e.g. "invalid", "unused", "misplaced"), +// and end with a noun identifying the relevant language object. For example, +// "_DuplicateDecl" or "_InvalidSliceExpr". For brevity, naming follows the +// convention that "bad" implies a problem with syntax, and "invalid" implies a +// problem with types. + +const ( + // InvalidSyntaxTree occurs if an invalid syntax tree is provided + // to the type checker. It should never happen. + InvalidSyntaxTree Code = -1 +) + +const ( + // The zero Code value indicates an unset (invalid) error code. + _ Code = iota + + // Test is reserved for errors that only apply while in self-test mode. + Test + + // BlankPkgName occurs when a package name is the blank identifier "_". + // + // Per the spec: + // "The PackageName must not be the blank identifier." + // + // Example: + // package _ + BlankPkgName + + // MismatchedPkgName occurs when a file's package name doesn't match the + // package name already established by other files. + MismatchedPkgName + + // InvalidPkgUse occurs when a package identifier is used outside of a + // selector expression. + // + // Example: + // import "fmt" + // + // var _ = fmt + InvalidPkgUse + + // BadImportPath occurs when an import path is not valid. + BadImportPath + + // BrokenImport occurs when importing a package fails. + // + // Example: + // import "amissingpackage" + BrokenImport + + // ImportCRenamed occurs when the special import "C" is renamed. "C" is a + // pseudo-package, and must not be renamed. + // + // Example: + // import _ "C" + ImportCRenamed + + // UnusedImport occurs when an import is unused. + // + // Example: + // import "fmt" + // + // func main() {} + UnusedImport + + // InvalidInitCycle occurs when an invalid cycle is detected within the + // initialization graph. + // + // Example: + // var x int = f() + // + // func f() int { return x } + InvalidInitCycle + + // DuplicateDecl occurs when an identifier is declared multiple times. + // + // Example: + // var x = 1 + // var x = 2 + DuplicateDecl + + // InvalidDeclCycle occurs when a declaration cycle is not valid. + // + // Example: + // type S struct { + // S + // } + // + InvalidDeclCycle + + // InvalidTypeCycle occurs when a cycle in type definitions results in a + // type that is not well-defined. + // + // Example: + // import "unsafe" + // + // type T [unsafe.Sizeof(T{})]int + InvalidTypeCycle + + // InvalidConstInit occurs when a const declaration has a non-constant + // initializer. + // + // Example: + // var x int + // const _ = x + InvalidConstInit + + // InvalidConstVal occurs when a const value cannot be converted to its + // target type. + // + // TODO(findleyr): this error code and example are not very clear. Consider + // removing it. + // + // Example: + // const _ = 1 << "hello" + InvalidConstVal + + // InvalidConstType occurs when the underlying type in a const declaration + // is not a valid constant type. + // + // Example: + // const c *int = 4 + InvalidConstType + + // UntypedNilUse occurs when the predeclared (untyped) value nil is used to + // initialize a variable declared without an explicit type. + // + // Example: + // var x = nil + UntypedNilUse + + // WrongAssignCount occurs when the number of values on the right-hand side + // of an assignment or initialization expression does not match the number + // of variables on the left-hand side. + // + // Example: + // var x = 1, 2 + WrongAssignCount + + // UnassignableOperand occurs when the left-hand side of an assignment is + // not assignable. + // + // Example: + // func f() { + // const c = 1 + // c = 2 + // } + UnassignableOperand + + // NoNewVar occurs when a short variable declaration (':=') does not declare + // new variables. + // + // Example: + // func f() { + // x := 1 + // x := 2 + // } + NoNewVar + + // MultiValAssignOp occurs when an assignment operation (+=, *=, etc) does + // not have single-valued left-hand or right-hand side. + // + // Per the spec: + // "In assignment operations, both the left- and right-hand expression lists + // must contain exactly one single-valued expression" + // + // Example: + // func f() int { + // x, y := 1, 2 + // x, y += 1 + // return x + y + // } + MultiValAssignOp + + // InvalidIfaceAssign occurs when a value of type T is used as an + // interface, but T does not implement a method of the expected interface. + // + // Example: + // type I interface { + // f() + // } + // + // type T int + // + // var x I = T(1) + InvalidIfaceAssign + + // InvalidChanAssign occurs when a chan assignment is invalid. + // + // Per the spec, a value x is assignable to a channel type T if: + // "x is a bidirectional channel value, T is a channel type, x's type V and + // T have identical element types, and at least one of V or T is not a + // defined type." + // + // Example: + // type T1 chan int + // type T2 chan int + // + // var x T1 + // // Invalid assignment because both types are named + // var _ T2 = x + InvalidChanAssign + + // IncompatibleAssign occurs when the type of the right-hand side expression + // in an assignment cannot be assigned to the type of the variable being + // assigned. + // + // Example: + // var x []int + // var _ int = x + IncompatibleAssign + + // UnaddressableFieldAssign occurs when trying to assign to a struct field + // in a map value. + // + // Example: + // func f() { + // m := make(map[string]struct{i int}) + // m["foo"].i = 42 + // } + UnaddressableFieldAssign + + // NotAType occurs when the identifier used as the underlying type in a type + // declaration or the right-hand side of a type alias does not denote a type. + // + // Example: + // var S = 2 + // + // type T S + NotAType + + // InvalidArrayLen occurs when an array length is not a constant value. + // + // Example: + // var n = 3 + // var _ = [n]int{} + InvalidArrayLen + + // BlankIfaceMethod occurs when a method name is '_'. + // + // Per the spec: + // "The name of each explicitly specified method must be unique and not + // blank." + // + // Example: + // type T interface { + // _(int) + // } + BlankIfaceMethod + + // IncomparableMapKey occurs when a map key type does not support the == and + // != operators. + // + // Per the spec: + // "The comparison operators == and != must be fully defined for operands of + // the key type; thus the key type must not be a function, map, or slice." + // + // Example: + // var x map[T]int + // + // type T []int + IncomparableMapKey + + // InvalidIfaceEmbed occurs when a non-interface type is embedded in an + // interface (for go 1.17 or earlier). + _ // not used anymore + + // InvalidPtrEmbed occurs when an embedded field is of the pointer form *T, + // and T itself is itself a pointer, an unsafe.Pointer, or an interface. + // + // Per the spec: + // "An embedded field must be specified as a type name T or as a pointer to + // a non-interface type name *T, and T itself may not be a pointer type." + // + // Example: + // type T *int + // + // type S struct { + // *T + // } + InvalidPtrEmbed + + // BadRecv occurs when a method declaration does not have exactly one + // receiver parameter. + // + // Example: + // func () _() {} + BadRecv + + // InvalidRecv occurs when a receiver type expression is not of the form T + // or *T, or T is a pointer type. + // + // Example: + // type T struct {} + // + // func (**T) m() {} + InvalidRecv + + // DuplicateFieldAndMethod occurs when an identifier appears as both a field + // and method name. + // + // Example: + // type T struct { + // m int + // } + // + // func (T) m() {} + DuplicateFieldAndMethod + + // DuplicateMethod occurs when two methods on the same receiver type have + // the same name. + // + // Example: + // type T struct {} + // func (T) m() {} + // func (T) m(i int) int { return i } + DuplicateMethod + + // InvalidBlank occurs when a blank identifier is used as a value or type. + // + // Per the spec: + // "The blank identifier may appear as an operand only on the left-hand side + // of an assignment." + // + // Example: + // var x = _ + InvalidBlank + + // InvalidIota occurs when the predeclared identifier iota is used outside + // of a constant declaration. + // + // Example: + // var x = iota + InvalidIota + + // MissingInitBody occurs when an init function is missing its body. + // + // Example: + // func init() + MissingInitBody + + // InvalidInitSig occurs when an init function declares parameters or + // results. + // + // Deprecated: no longer emitted by the type checker. _InvalidInitDecl is + // used instead. + InvalidInitSig + + // InvalidInitDecl occurs when init is declared as anything other than a + // function. + // + // Example: + // var init = 1 + // + // Example: + // func init() int { return 1 } + InvalidInitDecl + + // InvalidMainDecl occurs when main is declared as anything other than a + // function, in a main package. + InvalidMainDecl + + // TooManyValues occurs when a function returns too many values for the + // expression context in which it is used. + // + // Example: + // func ReturnTwo() (int, int) { + // return 1, 2 + // } + // + // var x = ReturnTwo() + TooManyValues + + // NotAnExpr occurs when a type expression is used where a value expression + // is expected. + // + // Example: + // type T struct {} + // + // func f() { + // T + // } + NotAnExpr + + // TruncatedFloat occurs when a float constant is truncated to an integer + // value. + // + // Example: + // var _ int = 98.6 + TruncatedFloat + + // NumericOverflow occurs when a numeric constant overflows its target type. + // + // Example: + // var x int8 = 1000 + NumericOverflow + + // UndefinedOp occurs when an operator is not defined for the type(s) used + // in an operation. + // + // Example: + // var c = "a" - "b" + UndefinedOp + + // MismatchedTypes occurs when operand types are incompatible in a binary + // operation. + // + // Example: + // var a = "hello" + // var b = 1 + // var c = a - b + MismatchedTypes + + // DivByZero occurs when a division operation is provable at compile + // time to be a division by zero. + // + // Example: + // const divisor = 0 + // var x int = 1/divisor + DivByZero + + // NonNumericIncDec occurs when an increment or decrement operator is + // applied to a non-numeric value. + // + // Example: + // func f() { + // var c = "c" + // c++ + // } + NonNumericIncDec + + // UnaddressableOperand occurs when the & operator is applied to an + // unaddressable expression. + // + // Example: + // var x = &1 + UnaddressableOperand + + // InvalidIndirection occurs when a non-pointer value is indirected via the + // '*' operator. + // + // Example: + // var x int + // var y = *x + InvalidIndirection + + // NonIndexableOperand occurs when an index operation is applied to a value + // that cannot be indexed. + // + // Example: + // var x = 1 + // var y = x[1] + NonIndexableOperand + + // InvalidIndex occurs when an index argument is not of integer type, + // negative, or out-of-bounds. + // + // Example: + // var s = [...]int{1,2,3} + // var x = s[5] + // + // Example: + // var s = []int{1,2,3} + // var _ = s[-1] + // + // Example: + // var s = []int{1,2,3} + // var i string + // var _ = s[i] + InvalidIndex + + // SwappedSliceIndices occurs when constant indices in a slice expression + // are decreasing in value. + // + // Example: + // var _ = []int{1,2,3}[2:1] + SwappedSliceIndices + + // NonSliceableOperand occurs when a slice operation is applied to a value + // whose type is not sliceable, or is unaddressable. + // + // Example: + // var x = [...]int{1, 2, 3}[:1] + // + // Example: + // var x = 1 + // var y = 1[:1] + NonSliceableOperand + + // InvalidSliceExpr occurs when a three-index slice expression (a[x:y:z]) is + // applied to a string. + // + // Example: + // var s = "hello" + // var x = s[1:2:3] + InvalidSliceExpr + + // InvalidShiftCount occurs when the right-hand side of a shift operation is + // either non-integer, negative, or too large. + // + // Example: + // var ( + // x string + // y int = 1 << x + // ) + InvalidShiftCount + + // InvalidShiftOperand occurs when the shifted operand is not an integer. + // + // Example: + // var s = "hello" + // var x = s << 2 + InvalidShiftOperand + + // InvalidReceive occurs when there is a channel receive from a value that + // is either not a channel, or is a send-only channel. + // + // Example: + // func f() { + // var x = 1 + // <-x + // } + InvalidReceive + + // InvalidSend occurs when there is a channel send to a value that is not a + // channel, or is a receive-only channel. + // + // Example: + // func f() { + // var x = 1 + // x <- "hello!" + // } + InvalidSend + + // DuplicateLitKey occurs when an index is duplicated in a slice, array, or + // map literal. + // + // Example: + // var _ = []int{0:1, 0:2} + // + // Example: + // var _ = map[string]int{"a": 1, "a": 2} + DuplicateLitKey + + // MissingLitKey occurs when a map literal is missing a key expression. + // + // Example: + // var _ = map[string]int{1} + MissingLitKey + + // InvalidLitIndex occurs when the key in a key-value element of a slice or + // array literal is not an integer constant. + // + // Example: + // var i = 0 + // var x = []string{i: "world"} + InvalidLitIndex + + // OversizeArrayLit occurs when an array literal exceeds its length. + // + // Example: + // var _ = [2]int{1,2,3} + OversizeArrayLit + + // MixedStructLit occurs when a struct literal contains a mix of positional + // and named elements. + // + // Example: + // var _ = struct{i, j int}{i: 1, 2} + MixedStructLit + + // InvalidStructLit occurs when a positional struct literal has an incorrect + // number of values. + // + // Example: + // var _ = struct{i, j int}{1,2,3} + InvalidStructLit + + // MissingLitField occurs when a struct literal refers to a field that does + // not exist on the struct type. + // + // Example: + // var _ = struct{i int}{j: 2} + MissingLitField + + // DuplicateLitField occurs when a struct literal contains duplicated + // fields. + // + // Example: + // var _ = struct{i int}{i: 1, i: 2} + DuplicateLitField + + // UnexportedLitField occurs when a positional struct literal implicitly + // assigns an unexported field of an imported type. + UnexportedLitField + + // InvalidLitField occurs when a field name is not a valid identifier. + // + // Example: + // var _ = struct{i int}{1: 1} + InvalidLitField + + // UntypedLit occurs when a composite literal omits a required type + // identifier. + // + // Example: + // type outer struct{ + // inner struct { i int } + // } + // + // var _ = outer{inner: {1}} + UntypedLit + + // InvalidLit occurs when a composite literal expression does not match its + // type. + // + // Example: + // type P *struct{ + // x int + // } + // var _ = P {} + InvalidLit + + // AmbiguousSelector occurs when a selector is ambiguous. + // + // Example: + // type E1 struct { i int } + // type E2 struct { i int } + // type T struct { E1; E2 } + // + // var x T + // var _ = x.i + AmbiguousSelector + + // UndeclaredImportedName occurs when a package-qualified identifier is + // undeclared by the imported package. + // + // Example: + // import "go/types" + // + // var _ = types.NotAnActualIdentifier + UndeclaredImportedName + + // UnexportedName occurs when a selector refers to an unexported identifier + // of an imported package. + // + // Example: + // import "reflect" + // + // type _ reflect.flag + UnexportedName + + // UndeclaredName occurs when an identifier is not declared in the current + // scope. + // + // Example: + // var x T + UndeclaredName + + // MissingFieldOrMethod occurs when a selector references a field or method + // that does not exist. + // + // Example: + // type T struct {} + // + // var x = T{}.f + MissingFieldOrMethod + + // BadDotDotDotSyntax occurs when a "..." occurs in a context where it is + // not valid. + // + // Example: + // var _ = map[int][...]int{0: {}} + BadDotDotDotSyntax + + // NonVariadicDotDotDot occurs when a "..." is used on the final argument to + // a non-variadic function. + // + // Example: + // func printArgs(s []string) { + // for _, a := range s { + // println(a) + // } + // } + // + // func f() { + // s := []string{"a", "b", "c"} + // printArgs(s...) + // } + NonVariadicDotDotDot + + // MisplacedDotDotDot occurs when a "..." is used somewhere other than the + // final argument in a function declaration. + // + // Example: + // func f(...int, int) + MisplacedDotDotDot + + _ // InvalidDotDotDotOperand was removed. + + // InvalidDotDotDot occurs when a "..." is used in a non-variadic built-in + // function. + // + // Example: + // var s = []int{1, 2, 3} + // var l = len(s...) + InvalidDotDotDot + + // UncalledBuiltin occurs when a built-in function is used as a + // function-valued expression, instead of being called. + // + // Per the spec: + // "The built-in functions do not have standard Go types, so they can only + // appear in call expressions; they cannot be used as function values." + // + // Example: + // var _ = copy + UncalledBuiltin + + // InvalidAppend occurs when append is called with a first argument that is + // not a slice. + // + // Example: + // var _ = append(1, 2) + InvalidAppend + + // InvalidCap occurs when an argument to the cap built-in function is not of + // supported type. + // + // See https://golang.org/ref/spec#Length_and_capacity for information on + // which underlying types are supported as arguments to cap and len. + // + // Example: + // var s = 2 + // var x = cap(s) + InvalidCap + + // InvalidClose occurs when close(...) is called with an argument that is + // not of channel type, or that is a receive-only channel. + // + // Example: + // func f() { + // var x int + // close(x) + // } + InvalidClose + + // InvalidCopy occurs when the arguments are not of slice type or do not + // have compatible type. + // + // See https://golang.org/ref/spec#Appending_and_copying_slices for more + // information on the type requirements for the copy built-in. + // + // Example: + // func f() { + // var x []int + // y := []int64{1,2,3} + // copy(x, y) + // } + InvalidCopy + + // InvalidComplex occurs when the complex built-in function is called with + // arguments with incompatible types. + // + // Example: + // var _ = complex(float32(1), float64(2)) + InvalidComplex + + // InvalidDelete occurs when the delete built-in function is called with a + // first argument that is not a map. + // + // Example: + // func f() { + // m := "hello" + // delete(m, "e") + // } + InvalidDelete + + // InvalidImag occurs when the imag built-in function is called with an + // argument that does not have complex type. + // + // Example: + // var _ = imag(int(1)) + InvalidImag + + // InvalidLen occurs when an argument to the len built-in function is not of + // supported type. + // + // See https://golang.org/ref/spec#Length_and_capacity for information on + // which underlying types are supported as arguments to cap and len. + // + // Example: + // var s = 2 + // var x = len(s) + InvalidLen + + // SwappedMakeArgs occurs when make is called with three arguments, and its + // length argument is larger than its capacity argument. + // + // Example: + // var x = make([]int, 3, 2) + SwappedMakeArgs + + // InvalidMake occurs when make is called with an unsupported type argument. + // + // See https://golang.org/ref/spec#Making_slices_maps_and_channels for + // information on the types that may be created using make. + // + // Example: + // var x = make(int) + InvalidMake + + // InvalidReal occurs when the real built-in function is called with an + // argument that does not have complex type. + // + // Example: + // var _ = real(int(1)) + InvalidReal + + // InvalidAssert occurs when a type assertion is applied to a + // value that is not of interface type. + // + // Example: + // var x = 1 + // var _ = x.(float64) + InvalidAssert + + // ImpossibleAssert occurs for a type assertion x.(T) when the value x of + // interface cannot have dynamic type T, due to a missing or mismatching + // method on T. + // + // Example: + // type T int + // + // func (t *T) m() int { return int(*t) } + // + // type I interface { m() int } + // + // var x I + // var _ = x.(T) + ImpossibleAssert + + // InvalidConversion occurs when the argument type cannot be converted to the + // target. + // + // See https://golang.org/ref/spec#Conversions for the rules of + // convertibility. + // + // Example: + // var x float64 + // var _ = string(x) + InvalidConversion + + // InvalidUntypedConversion occurs when there is no valid implicit + // conversion from an untyped value satisfying the type constraints of the + // context in which it is used. + // + // Example: + // var _ = 1 + []int{} + InvalidUntypedConversion + + // BadOffsetofSyntax occurs when unsafe.Offsetof is called with an argument + // that is not a selector expression. + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.Offsetof(x) + BadOffsetofSyntax + + // InvalidOffsetof occurs when unsafe.Offsetof is called with a method + // selector, rather than a field selector, or when the field is embedded via + // a pointer. + // + // Per the spec: + // + // "If f is an embedded field, it must be reachable without pointer + // indirections through fields of the struct. " + // + // Example: + // import "unsafe" + // + // type T struct { f int } + // type S struct { *T } + // var s S + // var _ = unsafe.Offsetof(s.f) + // + // Example: + // import "unsafe" + // + // type S struct{} + // + // func (S) m() {} + // + // var s S + // var _ = unsafe.Offsetof(s.m) + InvalidOffsetof + + // UnusedExpr occurs when a side-effect free expression is used as a + // statement. Such a statement has no effect. + // + // Example: + // func f(i int) { + // i*i + // } + UnusedExpr + + // UnusedVar occurs when a variable is declared but unused. + // + // Example: + // func f() { + // x := 1 + // } + UnusedVar + + // MissingReturn occurs when a function with results is missing a return + // statement. + // + // Example: + // func f() int {} + MissingReturn + + // WrongResultCount occurs when a return statement returns an incorrect + // number of values. + // + // Example: + // func ReturnOne() int { + // return 1, 2 + // } + WrongResultCount + + // OutOfScopeResult occurs when the name of a value implicitly returned by + // an empty return statement is shadowed in a nested scope. + // + // Example: + // func factor(n int) (i int) { + // for i := 2; i < n; i++ { + // if n%i == 0 { + // return + // } + // } + // return 0 + // } + OutOfScopeResult + + // InvalidCond occurs when an if condition is not a boolean expression. + // + // Example: + // func checkReturn(i int) { + // if i { + // panic("non-zero return") + // } + // } + InvalidCond + + // InvalidPostDecl occurs when there is a declaration in a for-loop post + // statement. + // + // Example: + // func f() { + // for i := 0; i < 10; j := 0 {} + // } + InvalidPostDecl + + _ // InvalidChanRange was removed. + + // InvalidIterVar occurs when two iteration variables are used while ranging + // over a channel. + // + // Example: + // func f(c chan int) { + // for k, v := range c { + // println(k, v) + // } + // } + InvalidIterVar + + // InvalidRangeExpr occurs when the type of a range expression is not array, + // slice, string, map, or channel. + // + // Example: + // func f(i int) { + // for j := range i { + // println(j) + // } + // } + InvalidRangeExpr + + // MisplacedBreak occurs when a break statement is not within a for, switch, + // or select statement of the innermost function definition. + // + // Example: + // func f() { + // break + // } + MisplacedBreak + + // MisplacedContinue occurs when a continue statement is not within a for + // loop of the innermost function definition. + // + // Example: + // func sumeven(n int) int { + // proceed := func() { + // continue + // } + // sum := 0 + // for i := 1; i <= n; i++ { + // if i % 2 != 0 { + // proceed() + // } + // sum += i + // } + // return sum + // } + MisplacedContinue + + // MisplacedFallthrough occurs when a fallthrough statement is not within an + // expression switch. + // + // Example: + // func typename(i interface{}) string { + // switch i.(type) { + // case int64: + // fallthrough + // case int: + // return "int" + // } + // return "unsupported" + // } + MisplacedFallthrough + + // DuplicateCase occurs when a type or expression switch has duplicate + // cases. + // + // Example: + // func printInt(i int) { + // switch i { + // case 1: + // println("one") + // case 1: + // println("One") + // } + // } + DuplicateCase + + // DuplicateDefault occurs when a type or expression switch has multiple + // default clauses. + // + // Example: + // func printInt(i int) { + // switch i { + // case 1: + // println("one") + // default: + // println("One") + // default: + // println("1") + // } + // } + DuplicateDefault + + // BadTypeKeyword occurs when a .(type) expression is used anywhere other + // than a type switch. + // + // Example: + // type I interface { + // m() + // } + // var t I + // var _ = t.(type) + BadTypeKeyword + + // InvalidTypeSwitch occurs when .(type) is used on an expression that is + // not of interface type. + // + // Example: + // func f(i int) { + // switch x := i.(type) {} + // } + InvalidTypeSwitch + + // InvalidExprSwitch occurs when a switch expression is not comparable. + // + // Example: + // func _() { + // var a struct{ _ func() } + // switch a /* ERROR cannot switch on a */ { + // } + // } + InvalidExprSwitch + + // InvalidSelectCase occurs when a select case is not a channel send or + // receive. + // + // Example: + // func checkChan(c <-chan int) bool { + // select { + // case c: + // return true + // default: + // return false + // } + // } + InvalidSelectCase + + // UndeclaredLabel occurs when an undeclared label is jumped to. + // + // Example: + // func f() { + // goto L + // } + UndeclaredLabel + + // DuplicateLabel occurs when a label is declared more than once. + // + // Example: + // func f() int { + // L: + // L: + // return 1 + // } + DuplicateLabel + + // MisplacedLabel occurs when a break or continue label is not on a for, + // switch, or select statement. + // + // Example: + // func f() { + // L: + // a := []int{1,2,3} + // for _, e := range a { + // if e > 10 { + // break L + // } + // println(a) + // } + // } + MisplacedLabel + + // UnusedLabel occurs when a label is declared and not used. + // + // Example: + // func f() { + // L: + // } + UnusedLabel + + // JumpOverDecl occurs when a label jumps over a variable declaration. + // + // Example: + // func f() int { + // goto L + // x := 2 + // L: + // x++ + // return x + // } + JumpOverDecl + + // JumpIntoBlock occurs when a forward jump goes to a label inside a nested + // block. + // + // Example: + // func f(x int) { + // goto L + // if x > 0 { + // L: + // print("inside block") + // } + // } + JumpIntoBlock + + // InvalidMethodExpr occurs when a pointer method is called but the argument + // is not addressable. + // + // Example: + // type T struct {} + // + // func (*T) m() int { return 1 } + // + // var _ = T.m(T{}) + InvalidMethodExpr + + // WrongArgCount occurs when too few or too many arguments are passed by a + // function call. + // + // Example: + // func f(i int) {} + // var x = f() + WrongArgCount + + // InvalidCall occurs when an expression is called that is not of function + // type. + // + // Example: + // var x = "x" + // var y = x() + InvalidCall + + // UnusedResults occurs when a restricted expression-only built-in function + // is suspended via go or defer. Such a suspension discards the results of + // these side-effect free built-in functions, and therefore is ineffectual. + // + // Example: + // func f(a []int) int { + // defer len(a) + // return i + // } + UnusedResults + + // InvalidDefer occurs when a deferred expression is not a function call, + // for example if the expression is a type conversion. + // + // Example: + // func f(i int) int { + // defer int32(i) + // return i + // } + InvalidDefer + + // InvalidGo occurs when a go expression is not a function call, for example + // if the expression is a type conversion. + // + // Example: + // func f(i int) int { + // go int32(i) + // return i + // } + InvalidGo + + // All codes below were added in Go 1.17. + + // BadDecl occurs when a declaration has invalid syntax. + BadDecl + + // RepeatedDecl occurs when an identifier occurs more than once on the left + // hand side of a short variable declaration. + // + // Example: + // func _() { + // x, y, y := 1, 2, 3 + // } + RepeatedDecl + + // InvalidUnsafeAdd occurs when unsafe.Add is called with a + // length argument that is not of integer type. + // It also occurs if it is used in a package compiled for a + // language version before go1.17. + // + // Example: + // import "unsafe" + // + // var p unsafe.Pointer + // var _ = unsafe.Add(p, float64(1)) + InvalidUnsafeAdd + + // InvalidUnsafeSlice occurs when unsafe.Slice is called with a + // pointer argument that is not of pointer type or a length argument + // that is not of integer type, negative, or out of bounds. + // It also occurs if it is used in a package compiled for a language + // version before go1.17. + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.Slice(x, 1) + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.Slice(&x, float64(1)) + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.Slice(&x, -1) + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.Slice(&x, uint64(1) << 63) + InvalidUnsafeSlice + + // All codes below were added in Go 1.18. + + // UnsupportedFeature occurs when a language feature is used that is not + // supported at this Go version. + UnsupportedFeature + + // NotAGenericType occurs when a non-generic type is used where a generic + // type is expected: in type or function instantiation. + // + // Example: + // type T int + // + // var _ T[int] + NotAGenericType + + // WrongTypeArgCount occurs when a type or function is instantiated with an + // incorrect number of type arguments, including when a generic type or + // function is used without instantiation. + // + // Errors involving failed type inference are assigned other error codes. + // + // Example: + // type T[p any] int + // + // var _ T[int, string] + // + // Example: + // func f[T any]() {} + // + // var x = f + WrongTypeArgCount + + // CannotInferTypeArgs occurs when type or function type argument inference + // fails to infer all type arguments. + // + // Example: + // func f[T any]() {} + // + // func _() { + // f() + // } + CannotInferTypeArgs + + // InvalidTypeArg occurs when a type argument does not satisfy its + // corresponding type parameter constraints. + // + // Example: + // type T[P ~int] struct{} + // + // var _ T[string] + InvalidTypeArg // arguments? InferenceFailed + + // InvalidInstanceCycle occurs when an invalid cycle is detected + // within the instantiation graph. + // + // Example: + // func f[T any]() { f[*T]() } + InvalidInstanceCycle + + // InvalidUnion occurs when an embedded union or approximation element is + // not valid. + // + // Example: + // type _ interface { + // ~int | interface{ m() } + // } + InvalidUnion + + // MisplacedConstraintIface occurs when a constraint-type interface is used + // outside of constraint position. + // + // Example: + // type I interface { ~int } + // + // var _ I + MisplacedConstraintIface + + // InvalidMethodTypeParams occurs when methods have type parameters. + // + // It cannot be encountered with an AST parsed using go/parser. + InvalidMethodTypeParams + + // MisplacedTypeParam occurs when a type parameter is used in a place where + // it is not permitted. + // + // Example: + // type T[P any] P + // + // Example: + // type T[P any] struct{ *P } + MisplacedTypeParam + + // InvalidUnsafeSliceData occurs when unsafe.SliceData is called with + // an argument that is not of slice type. It also occurs if it is used + // in a package compiled for a language version before go1.20. + // + // Example: + // import "unsafe" + // + // var x int + // var _ = unsafe.SliceData(x) + InvalidUnsafeSliceData + + // InvalidUnsafeString occurs when unsafe.String is called with + // a length argument that is not of integer type, negative, or + // out of bounds. It also occurs if it is used in a package + // compiled for a language version before go1.20. + // + // Example: + // import "unsafe" + // + // var b [10]byte + // var _ = unsafe.String(&b[0], -1) + InvalidUnsafeString + + // InvalidUnsafeStringData occurs if it is used in a package + // compiled for a language version before go1.20. + _ // not used anymore + + // InvalidClear occurs when clear is called with an argument + // that is not of map or slice type. + // + // Example: + // func _(x int) { + // clear(x) + // } + InvalidClear + + // TypeTooLarge occurs if unsafe.Sizeof or unsafe.Offsetof is + // called with an expression whose type is too large. + // + // Example: + // import "unsafe" + // + // type E [1 << 31 - 1]int + // var a [1 << 31]E + // var _ = unsafe.Sizeof(a) + // + // Example: + // import "unsafe" + // + // type E [1 << 31 - 1]int + // var s struct { + // _ [1 << 31]E + // x int + // } + // var _ = unsafe.Offsetof(s.x) + TypeTooLarge + + // InvalidMinMaxOperand occurs if min or max is called + // with an operand that cannot be ordered because it + // does not support the < operator. + // + // Example: + // const _ = min(true) + // + // Example: + // var s, t []byte + // var _ = max(s, t) + InvalidMinMaxOperand +) diff --git a/src/internal/types/errors/codes_test.go b/src/internal/types/errors/codes_test.go new file mode 100644 index 0000000..2490ade --- /dev/null +++ b/src/internal/types/errors/codes_test.go @@ -0,0 +1,197 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package errors_test + +import ( + "fmt" + "go/ast" + "go/constant" + "go/importer" + "go/parser" + "go/token" + "internal/testenv" + "reflect" + "strings" + "testing" + + . "go/types" +) + +func TestErrorCodeExamples(t *testing.T) { + testenv.MustHaveGoBuild(t) // go command needed to resolve std .a files for importer.Default(). + + walkCodes(t, func(name string, value int, spec *ast.ValueSpec) { + t.Run(name, func(t *testing.T) { + doc := spec.Doc.Text() + examples := strings.Split(doc, "Example:") + for i := 1; i < len(examples); i++ { + example := strings.TrimSpace(examples[i]) + err := checkExample(t, example) + if err == nil { + t.Fatalf("no error in example #%d", i) + } + typerr, ok := err.(Error) + if !ok { + t.Fatalf("not a types.Error: %v", err) + } + if got := readCode(typerr); got != value { + t.Errorf("%s: example #%d returned code %d (%s), want %d", name, i, got, err, value) + } + } + }) + }) +} + +func walkCodes(t *testing.T, f func(string, int, *ast.ValueSpec)) { + t.Helper() + fset := token.NewFileSet() + file, err := parser.ParseFile(fset, "codes.go", nil, parser.ParseComments) + if err != nil { + t.Fatal(err) + } + conf := Config{Importer: importer.Default()} + info := &Info{ + Types: make(map[ast.Expr]TypeAndValue), + Defs: make(map[*ast.Ident]Object), + Uses: make(map[*ast.Ident]Object), + } + _, err = conf.Check("types", fset, []*ast.File{file}, info) + if err != nil { + t.Fatal(err) + } + for _, decl := range file.Decls { + decl, ok := decl.(*ast.GenDecl) + if !ok || decl.Tok != token.CONST { + continue + } + for _, spec := range decl.Specs { + spec, ok := spec.(*ast.ValueSpec) + if !ok || len(spec.Names) == 0 { + continue + } + obj := info.ObjectOf(spec.Names[0]) + if named, ok := obj.Type().(*Named); ok && named.Obj().Name() == "Code" { + if len(spec.Names) != 1 { + t.Fatalf("bad Code declaration for %q: got %d names, want exactly 1", spec.Names[0].Name, len(spec.Names)) + } + codename := spec.Names[0].Name + value := int(constant.Val(obj.(*Const).Val()).(int64)) + f(codename, value, spec) + } + } + } +} + +func readCode(err Error) int { + v := reflect.ValueOf(err) + return int(v.FieldByName("go116code").Int()) +} + +func checkExample(t *testing.T, example string) error { + t.Helper() + fset := token.NewFileSet() + if !strings.HasPrefix(example, "package") { + example = "package p\n\n" + example + } + file, err := parser.ParseFile(fset, "example.go", example, 0) + if err != nil { + t.Fatal(err) + } + conf := Config{ + FakeImportC: true, + Importer: importer.Default(), + } + _, err = conf.Check("example", fset, []*ast.File{file}, nil) + return err +} + +func TestErrorCodeStyle(t *testing.T) { + // The set of error codes is large and intended to be self-documenting, so + // this test enforces some style conventions. + forbiddenInIdent := []string{ + // use invalid instead + "illegal", + // words with a common short-form + "argument", + "assertion", + "assignment", + "boolean", + "channel", + "condition", + "declaration", + "expression", + "function", + "initial", // use init for initializer, initialization, etc. + "integer", + "interface", + "iterat", // use iter for iterator, iteration, etc. + "literal", + "operation", + "package", + "pointer", + "receiver", + "signature", + "statement", + "variable", + } + forbiddenInComment := []string{ + // lhs and rhs should be spelled-out. + "lhs", "rhs", + // builtin should be hyphenated. + "builtin", + // Use dot-dot-dot. + "ellipsis", + } + nameHist := make(map[int]int) + longestName := "" + maxValue := 0 + + walkCodes(t, func(name string, value int, spec *ast.ValueSpec) { + if name == "_" { + return + } + nameHist[len(name)]++ + if value > maxValue { + maxValue = value + } + if len(name) > len(longestName) { + longestName = name + } + if !token.IsExported(name) { + t.Errorf("%q is not exported", name) + } + lower := strings.ToLower(name) + for _, bad := range forbiddenInIdent { + if strings.Contains(lower, bad) { + t.Errorf("%q contains forbidden word %q", name, bad) + } + } + doc := spec.Doc.Text() + if doc == "" { + t.Errorf("%q is undocumented", name) + } else if !strings.HasPrefix(doc, name) { + t.Errorf("doc for %q does not start with the error code name", name) + } + lowerComment := strings.ToLower(strings.TrimPrefix(doc, name)) + for _, bad := range forbiddenInComment { + if strings.Contains(lowerComment, bad) { + t.Errorf("doc for %q contains forbidden word %q", name, bad) + } + } + }) + + if testing.Verbose() { + var totChars, totCount int + for chars, count := range nameHist { + totChars += chars * count + totCount += count + } + avg := float64(totChars) / float64(totCount) + fmt.Println() + fmt.Printf("%d error codes\n", totCount) + fmt.Printf("average length: %.2f chars\n", avg) + fmt.Printf("max length: %d (%s)\n", len(longestName), longestName) + } +} diff --git a/src/internal/types/errors/generrordocs.go b/src/internal/types/errors/generrordocs.go new file mode 100644 index 0000000..46343be --- /dev/null +++ b/src/internal/types/errors/generrordocs.go @@ -0,0 +1,117 @@ +//go:build ignore + +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// generrordocs creates a Markdown file for each (compiler) error code +// and its associated documentation. +// Note: this program must be run in this directory. +// go run generrordocs.go <dir> + +//go:generate go run generrordocs.go errors_markdown + +package main + +import ( + "bytes" + "fmt" + "go/ast" + "go/importer" + "go/parser" + "go/token" + "log" + "os" + "path" + "strings" + "text/template" + + . "go/types" +) + +func main() { + if len(os.Args) != 2 { + log.Fatal("missing argument: generrordocs <dir>") + } + outDir := os.Args[1] + if err := os.MkdirAll(outDir, 0755); err != nil { + log.Fatal("unable to create output directory: %s", err) + } + walkCodes(func(name string, vs *ast.ValueSpec) { + // ignore unused errors + if name == "_" { + return + } + // Ensure that < are represented correctly when its included in code + // blocks. The goldmark Markdown parser converts them to &lt; + // when not escaped. It is the only known string with this issue. + desc := strings.ReplaceAll(vs.Doc.Text(), "<", `{{raw "<"}}`) + e := struct { + Name string + Description string + }{ + Name: name, + Description: fmt.Sprintf("```\n%s```\n", desyc), + } + var buf bytes.Buffer + err := template.Must(template.New("eachError").Parse(markdownTemplate)).Execute(&buf, e) + if err != nil { + log.Fatalf("template.Must: %s", err) + } + if err := os.WriteFile(path.Join(outDir, name+".md"), buf.Bytes(), 0660); err != nil { + log.Fatalf("os.WriteFile: %s\n", err) + } + }) + log.Printf("output directory: %s\n", outDir) +} + +func walkCodes(f func(string, *ast.ValueSpec)) { + fset := token.NewFileSet() + file, err := parser.ParseFile(fset, "codes.go", nil, parser.ParseComments) + if err != nil { + log.Fatalf("ParseFile failed: %s", err) + } + conf := Config{Importer: importer.Default()} + info := &Info{ + Types: make(map[ast.Expr]TypeAndValue), + Defs: make(map[*ast.Ident]Object), + Uses: make(map[*ast.Ident]Object), + } + _, err = conf.Check("types", fset, []*ast.File{file}, info) + if err != nil { + log.Fatalf("Check failed: %s", err) + } + for _, decl := range file.Decls { + decl, ok := decl.(*ast.GenDecl) + if !ok || decl.Tok != token.CONST { + continue + } + for _, spec := range decl.Specs { + spec, ok := spec.(*ast.ValueSpec) + if !ok || len(spec.Names) == 0 { + continue + } + obj := info.ObjectOf(spec.Names[0]) + if named, ok := obj.Type().(*Named); ok && named.Obj().Name() == "Code" { + if len(spec.Names) != 1 { + log.Fatalf("bad Code declaration for %q: got %d names, want exactly 1", spec.Names[0].Name, len(spec.Names)) + } + codename := spec.Names[0].Name + f(codename, spec) + } + } + } +} + +const markdownTemplate = `--- +title: {{.Name}} +layout: article +--- +<!-- Copyright 2023 The Go Authors. All rights reserved. + Use of this source code is governed by a BSD-style + license that can be found in the LICENSE file. --> + +<!-- Code generated by generrordocs.go; DO NOT EDIT. --> + +{{.Description}} +` diff --git a/src/internal/types/testdata/check/blank.go b/src/internal/types/testdata/check/blank.go new file mode 100644 index 0000000..2bea11f --- /dev/null +++ b/src/internal/types/testdata/check/blank.go @@ -0,0 +1,5 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package _ /* ERROR "invalid package name" */ diff --git a/src/internal/types/testdata/check/builtins0.go b/src/internal/types/testdata/check/builtins0.go new file mode 100644 index 0000000..ed4769e --- /dev/null +++ b/src/internal/types/testdata/check/builtins0.go @@ -0,0 +1,1075 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// builtin calls + +package builtins + +import "unsafe" + +func f0() {} + +func append1() { + var b byte + var x int + var s []byte + _ = append() // ERROR "not enough arguments" + _ = append("foo" /* ERROR "must be a slice" */ ) + _ = append(nil /* ERROR "must be a slice" */ , s) + _ = append(x /* ERROR "must be a slice" */ , s) + _ = append(s) + _ = append(s, nil...) + append /* ERROR "not used" */ (s) + + _ = append(s, b) + _ = append(s, x /* ERROR "cannot use x" */ ) + _ = append(s, s /* ERROR "cannot use s" */ ) + _ = append(s...) /* ERROR "not enough arguments" */ + _ = append(s, b, s /* ERROR "too many arguments" */ ...) + _ = append(s, 1, 2, 3) + _ = append(s, 1, 2, 3, x /* ERROR "cannot use x" */ , 5, 6, 6) + _ = append(s, 1, 2 /* ERROR "too many arguments" */, s...) + _ = append([]interface{}(nil), 1, 2, "foo", x, 3.1425, false) + + type S []byte + type T string + var t T + _ = append(s, "foo" /* ERRORx `cannot use .* in argument to append` */ ) + _ = append(s, "foo"...) + _ = append(S(s), "foo" /* ERRORx `cannot use .* in argument to append` */ ) + _ = append(S(s), "foo"...) + _ = append(s, t /* ERROR "cannot use t" */ ) + _ = append(s, t...) + _ = append(s, T("foo")...) + _ = append(S(s), t /* ERROR "cannot use t" */ ) + _ = append(S(s), t...) + _ = append(S(s), T("foo")...) + _ = append([]string{}, t /* ERROR "cannot use t" */ , "foo") + _ = append([]T{}, t, "foo") +} + +// from the spec +func append2() { + s0 := []int{0, 0} + s1 := append(s0, 2) // append a single element s1 == []int{0, 0, 2} + s2 := append(s1, 3, 5, 7) // append multiple elements s2 == []int{0, 0, 2, 3, 5, 7} + s3 := append(s2, s0...) // append a slice s3 == []int{0, 0, 2, 3, 5, 7, 0, 0} + s4 := append(s3[3:6], s3[2:]...) // append overlapping slice s4 == []int{3, 5, 7, 2, 3, 5, 7, 0, 0} + + var t []interface{} + t = append(t, 42, 3.1415, "foo") // t == []interface{}{42, 3.1415, "foo"} + + var b []byte + b = append(b, "bar"...) // append string contents b == []byte{'b', 'a', 'r' } + + _ = s4 +} + +func append3() { + f1 := func() (s []int) { return } + f2 := func() (s []int, x int) { return } + f3 := func() (s []int, x, y int) { return } + f5 := func() (s []interface{}, x int, y float32, z string, b bool) { return } + ff := func() (int, float32) { return 0, 0 } + _ = append(f0 /* ERROR "used as value" */ ()) + _ = append(f1()) + _ = append(f2()) + _ = append(f3()) + _ = append(f5()) + _ = append(ff /* ERROR "must be a slice" */ ()) // TODO(gri) better error message +} + +func cap1() { + var a [10]bool + var p *[20]int + var c chan string + _ = cap() // ERROR "not enough arguments" + _ = cap(1, 2) // ERROR "too many arguments" + _ = cap(42 /* ERROR "invalid" */) + const _3 = cap(a) + assert(_3 == 10) + const _4 = cap(p) + assert(_4 == 20) + _ = cap(c) + cap /* ERROR "not used" */ (c) + + // issue 4744 + type T struct{ a [10]int } + const _ = cap(((*T)(nil)).a) + + var s [][]byte + _ = cap(s) + _ = cap(s... /* ERROR "invalid use of ..." */ ) +} + +func cap2() { + f1a := func() (a [10]int) { return } + f1s := func() (s []int) { return } + f2 := func() (s []int, x int) { return } + _ = cap(f0 /* ERROR "used as value" */ ()) + _ = cap(f1a()) + _ = cap(f1s()) + _ = cap(f2()) // ERROR "too many arguments" +} + +// test cases for issue 7387 +func cap3() { + var f = func() int { return 0 } + var x = f() + const ( + _ = cap([4]int{}) + _ = cap([4]int{x}) + _ = cap /* ERROR "not constant" */ ([4]int{f()}) + _ = cap /* ERROR "not constant" */ ([4]int{cap([]int{})}) + _ = cap([4]int{cap([4]int{})}) + ) + var y float64 + var z complex128 + const ( + _ = cap([4]float64{}) + _ = cap([4]float64{y}) + _ = cap([4]float64{real(2i)}) + _ = cap /* ERROR "not constant" */ ([4]float64{real(z)}) + ) + var ch chan [10]int + const ( + _ = cap /* ERROR "not constant" */ (<-ch) + _ = cap /* ERROR "not constant" */ ([4]int{(<-ch)[0]}) + ) +} + +func clear1() { + var a [10]int + var m map[float64]string + var s []byte + clear(a /* ERROR "cannot clear a" */) + clear(&/* ERROR "cannot clear &a" */a) + clear(m) + clear(s) + clear([]int{}) +} + +func close1() { + var c chan int + var r <-chan int + close() // ERROR "not enough arguments" + close(1, 2) // ERROR "too many arguments" + close(42 /* ERROR "cannot close non-channel" */) + close(r /* ERROR "receive-only channel" */) + close(c) + _ = close /* ERROR "used as value" */ (c) + + var s []chan int + close(s... /* ERROR "invalid use of ..." */ ) +} + +func close2() { + f1 := func() (ch chan int) { return } + f2 := func() (ch chan int, x int) { return } + close(f0 /* ERROR "used as value" */ ()) + close(f1()) + close(f2()) // ERROR "too many arguments" +} + +func complex1() { + var i32 int32 + var f32 float32 + var f64 float64 + var c64 complex64 + var c128 complex128 + _ = complex() // ERROR "not enough arguments" + _ = complex(1) // ERROR "not enough arguments" + _ = complex(true /* ERROR "mismatched types" */ , 0) + _ = complex(i32 /* ERROR "expected floating-point" */ , 0) + _ = complex("foo" /* ERROR "mismatched types" */ , 0) + _ = complex(c64 /* ERROR "expected floating-point" */ , 0) + _ = complex(0 /* ERROR "mismatched types" */ , true) + _ = complex(0 /* ERROR "expected floating-point" */ , i32) + _ = complex(0 /* ERROR "mismatched types" */ , "foo") + _ = complex(0 /* ERROR "expected floating-point" */ , c64) + _ = complex(f32, f32) + _ = complex(f32, 1) + _ = complex(f32, 1.0) + _ = complex(f32, 'a') + _ = complex(f64, f64) + _ = complex(f64, 1) + _ = complex(f64, 1.0) + _ = complex(f64, 'a') + _ = complex(f32 /* ERROR "mismatched types" */ , f64) + _ = complex(f64 /* ERROR "mismatched types" */ , f32) + _ = complex(1, 1) + _ = complex(1, 1.1) + _ = complex(1, 'a') + complex /* ERROR "not used" */ (1, 2) + + var _ complex64 = complex(f32, f32) + var _ complex64 = complex /* ERRORx `cannot use .* in variable declaration` */ (f64, f64) + + var _ complex128 = complex /* ERRORx `cannot use .* in variable declaration` */ (f32, f32) + var _ complex128 = complex(f64, f64) + + // untyped constants + const _ int = complex(1, 0) + const _ float32 = complex(1, 0) + const _ complex64 = complex(1, 0) + const _ complex128 = complex(1, 0) + const _ = complex(0i, 0i) + const _ = complex(0i, 0) + const _ int = 1.0 + complex(1, 0i) + + const _ int = complex /* ERROR "int" */ (1.1, 0) + const _ float32 = complex /* ERROR "float32" */ (1, 2) + + // untyped values + var s uint + _ = complex(1 /* ERROR "integer" */ <<s, 0) + const _ = complex /* ERROR "not constant" */ (1 /* ERROR "integer" */ <<s, 0) + var _ int = complex /* ERRORx `cannot use .* in variable declaration` */ (1 /* ERROR "integer" */ <<s, 0) + + // floating-point argument types must be identical + type F32 float32 + type F64 float64 + var x32 F32 + var x64 F64 + c64 = complex(x32, x32) + _ = complex(x32 /* ERROR "mismatched types" */ , f32) + _ = complex(f32 /* ERROR "mismatched types" */ , x32) + c128 = complex(x64, x64) + _ = c128 + _ = complex(x64 /* ERROR "mismatched types" */ , f64) + _ = complex(f64 /* ERROR "mismatched types" */ , x64) + + var t []float32 + _ = complex(t... /* ERROR "invalid use of ..." */ ) +} + +func complex2() { + f1 := func() (x float32) { return } + f2 := func() (x, y float32) { return } + f3 := func() (x, y, z float32) { return } + _ = complex(f0 /* ERROR "used as value" */ ()) + _ = complex(f1()) // ERROR "not enough arguments" + _ = complex(f2()) + _ = complex(f3()) // ERROR "too many arguments" +} + +func copy1() { + copy() // ERROR "not enough arguments" + copy("foo") // ERROR "not enough arguments" + copy([ /* ERROR "copy expects slice arguments" */ ...]int{}, []int{}) + copy([ /* ERROR "copy expects slice arguments" */ ]int{}, [...]int{}) + copy([ /* ERROR "different element types" */ ]int8{}, "foo") + + // spec examples + var a = [...]int{0, 1, 2, 3, 4, 5, 6, 7} + var s = make([]int, 6) + var b = make([]byte, 5) + n1 := copy(s, a[0:]) // n1 == 6, s == []int{0, 1, 2, 3, 4, 5} + n2 := copy(s, s[2:]) // n2 == 4, s == []int{2, 3, 4, 5, 4, 5} + n3 := copy(b, "Hello, World!") // n3 == 5, b == []byte("Hello") + _, _, _ = n1, n2, n3 + + var t [][]int + copy(t, t) + copy(t /* ERROR "copy expects slice arguments" */ , nil) + copy(nil /* ERROR "copy expects slice arguments" */ , t) + copy(nil /* ERROR "copy expects slice arguments" */ , nil) + copy(t... /* ERROR "invalid use of ..." */ ) +} + +func copy2() { + f1 := func() (a []int) { return } + f2 := func() (a, b []int) { return } + f3 := func() (a, b, c []int) { return } + copy(f0 /* ERROR "used as value" */ ()) + copy(f1()) // ERROR "not enough arguments" + copy(f2()) + copy(f3()) // ERROR "too many arguments" +} + +func delete1() { + var m map[string]int + var s string + delete() // ERROR "not enough arguments" + delete(1) // ERROR "not enough arguments" + delete(1, 2, 3) // ERROR "too many arguments" + delete(m, 0 /* ERROR "cannot use" */) + delete(m, s) + _ = delete /* ERROR "used as value" */ (m, s) + + var t []map[string]string + delete(t... /* ERROR "invalid use of ..." */ ) +} + +func delete2() { + f1 := func() (m map[string]int) { return } + f2 := func() (m map[string]int, k string) { return } + f3 := func() (m map[string]int, k string, x float32) { return } + delete(f0 /* ERROR "used as value" */ ()) + delete(f1()) // ERROR "not enough arguments" + delete(f2()) + delete(f3()) // ERROR "too many arguments" +} + +func imag1() { + var f32 float32 + var f64 float64 + var c64 complex64 + var c128 complex128 + _ = imag() // ERROR "not enough arguments" + _ = imag(1, 2) // ERROR "too many arguments" + _ = imag(10) + _ = imag(2.7182818) + _ = imag("foo" /* ERROR "expected complex" */) + _ = imag('a') + const _5 = imag(1 + 2i) + assert(_5 == 2) + f32 = _5 + f64 = _5 + const _6 = imag(0i) + assert(_6 == 0) + f32 = imag(c64) + f64 = imag(c128) + f32 = imag /* ERRORx `cannot use .* in assignment` */ (c128) + f64 = imag /* ERRORx `cannot use .* in assignment` */ (c64) + imag /* ERROR "not used" */ (c64) + _, _ = f32, f64 + + // complex type may not be predeclared + type C64 complex64 + type C128 complex128 + var x64 C64 + var x128 C128 + f32 = imag(x64) + f64 = imag(x128) + + var a []complex64 + _ = imag(a... /* ERROR "invalid use of ..." */ ) + + // if argument is untyped, result is untyped + const _ byte = imag(1.2 + 3i) + const _ complex128 = imag(1.2 + 3i) + + // lhs constant shift operands are typed as complex128 + var s uint + _ = imag(1 /* ERROR "must be integer" */ << s) +} + +func imag2() { + f1 := func() (x complex128) { return } + f2 := func() (x, y complex128) { return } + _ = imag(f0 /* ERROR "used as value" */ ()) + _ = imag(f1()) + _ = imag(f2()) // ERROR "too many arguments" +} + +func len1() { + const c = "foobar" + var a [10]bool + var p *[20]int + var m map[string]complex128 + _ = len() // ERROR "not enough arguments" + _ = len(1, 2) // ERROR "too many arguments" + _ = len(42 /* ERROR "invalid" */) + const _3 = len(c) + assert(_3 == 6) + const _4 = len(a) + assert(_4 == 10) + const _5 = len(p) + assert(_5 == 20) + _ = len(m) + len /* ERROR "not used" */ (c) + + // esoteric case + var t string + var hash map[interface{}][]*[10]int + const n = len /* ERROR "not constant" */ (hash[recover()][len(t)]) + assert(n == 10) // ok because n has unknown value and no error is reported + var ch <-chan int + const nn = len /* ERROR "not constant" */ (hash[<-ch][len(t)]) + + // issue 4744 + type T struct{ a [10]int } + const _ = len(((*T)(nil)).a) + + var s [][]byte + _ = len(s) + _ = len(s... /* ERROR "invalid use of ..." */ ) +} + +func len2() { + f1 := func() (x []int) { return } + f2 := func() (x, y []int) { return } + _ = len(f0 /* ERROR "used as value" */ ()) + _ = len(f1()) + _ = len(f2()) // ERROR "too many arguments" +} + +// test cases for issue 7387 +func len3() { + var f = func() int { return 0 } + var x = f() + const ( + _ = len([4]int{}) + _ = len([4]int{x}) + _ = len /* ERROR "not constant" */ ([4]int{f()}) + _ = len /* ERROR "not constant" */ ([4]int{len([]int{})}) + _ = len([4]int{len([4]int{})}) + ) + var y float64 + var z complex128 + const ( + _ = len([4]float64{}) + _ = len([4]float64{y}) + _ = len([4]float64{real(2i)}) + _ = len /* ERROR "not constant" */ ([4]float64{real(z)}) + ) + var ch chan [10]int + const ( + _ = len /* ERROR "not constant" */ (<-ch) + _ = len /* ERROR "not constant" */ ([4]int{(<-ch)[0]}) + ) +} + +func make1() { + var n int + var m float32 + var s uint + + _ = make() // ERROR "not enough arguments" + _ = make(1 /* ERROR "not a type" */) + _ = make(int /* ERROR "cannot make" */) + + // slices + _ = make/* ERROR "arguments" */ ([]int) + _ = make/* ERROR "arguments" */ ([]int, 2, 3, 4) + _ = make([]int, int /* ERROR "not an expression" */) + _ = make([]int, 10, float32 /* ERROR "not an expression" */) + _ = make([]int, "foo" /* ERROR "cannot convert" */) + _ = make([]int, 10, 2.3 /* ERROR "truncated" */) + _ = make([]int, 5, 10.0) + _ = make([]int, 0i) + _ = make([]int, 1.0) + _ = make([]int, 1.0<<s) + _ = make([]int, 1.1 /* ERROR "int" */ <<s) + _ = make([]int, - /* ERROR "must not be negative" */ 1, 10) + _ = make([]int, 0, - /* ERROR "must not be negative" */ 1) + _ = make([]int, - /* ERROR "must not be negative" */ 1, - /* ERROR "must not be negative" */ 1) + _ = make([]int, 1 /* ERROR "overflows" */ <<100, 1 /* ERROR "overflows" */ <<100) + _ = make([]int, 10 /* ERROR "length and capacity swapped" */ , 9) + _ = make([]int, 1 /* ERROR "overflows" */ <<100, 12345) + _ = make([]int, m /* ERROR "must be integer" */ ) + _ = &make /* ERROR "cannot take address" */ ([]int, 0) + + // maps + _ = make /* ERROR "arguments" */ (map[int]string, 10, 20) + _ = make(map[int]float32, int /* ERROR "not an expression" */) + _ = make(map[int]float32, "foo" /* ERROR "cannot convert" */) + _ = make(map[int]float32, 10) + _ = make(map[int]float32, n) + _ = make(map[int]float32, int64(n)) + _ = make(map[string]bool, 10.0) + _ = make(map[string]bool, 10.0<<s) + _ = &make /* ERROR "cannot take address" */ (map[string]bool) + + // channels + _ = make /* ERROR "arguments" */ (chan int, 10, 20) + _ = make(chan int, int /* ERROR "not an expression" */) + _ = make(chan<- int, "foo" /* ERROR "cannot convert" */) + _ = make(chan int, - /* ERROR "must not be negative" */ 10) + _ = make(<-chan float64, 10) + _ = make(chan chan int, n) + _ = make(chan string, int64(n)) + _ = make(chan bool, 10.0) + _ = make(chan bool, 10.0<<s) + _ = &make /* ERROR "cannot take address" */ (chan bool) + + make /* ERROR "not used" */ ([]int, 10) + + var t []int + _ = make([]int, t[0], t[1]) + _ = make([]int, t... /* ERROR "invalid use of ..." */ ) +} + +func make2() { + f1 := func() (x []int) { return } + _ = make(f0 /* ERROR "not a type" */ ()) + _ = make(f1 /* ERROR "not a type" */ ()) +} + +func max1() { + var b bool + var c complex128 + var x int + var s string + type myint int + var m myint + _ = max() /* ERROR "not enough arguments" */ + _ = max(b /* ERROR "cannot be ordered" */ ) + _ = max(c /* ERROR "cannot be ordered" */ ) + _ = max(x) + _ = max(s) + _ = max(x, x) + _ = max(x, x, x, x, x) + var _ int = max /* ERROR "cannot use max(m) (value of type myint) as int value" */ (m) + _ = max(x, m /* ERROR "invalid argument: mismatched types int (previous argument) and myint (type of m)" */ , x) + + _ = max(1, x) + _ = max(1.0, x) + _ = max(1.2 /* ERROR "1.2 (untyped float constant) truncated to int" */ , x) + _ = max(-10, 1.0, c /* ERROR "cannot be ordered" */ ) + + const ( + _ = max /* ERROR "max(x) (value of type int) is not constant" */ (x) + _ = max(true /* ERROR "invalid argument: true (untyped bool constant) cannot be ordered" */ ) + _ = max(1) + _ = max(1, 2.3, 'a') + _ = max(1, "foo" /* ERROR "mismatched types" */ ) + _ = max(1, 0i /* ERROR "cannot be ordered" */ ) + _ = max(1, 2 /* ERROR "cannot be ordered" */ + 3i ) + ) +} + +func max2() { + _ = assert(max(0) == 0) + _ = assert(max(0, 1) == 1) + _ = assert(max(0, -10, 123456789) == 123456789) + _ = assert(max(-12345678901234567890, 0) == 0) + + _ = assert(max(1, 2.3) == 2.3) + _ = assert(max(1, 2.3, 'a') == 'a') + + _ = assert(max("", "a") == "a") + _ = assert(max("abcde", "xyz", "foo", "bar") == "xyz") + + const ( + _ int = max(1.0) + _ float32 = max(1, 2) + _ int = max /* ERROR "cannot use max(1, 2.3) (untyped float constant 2.3) as int value" */ (1, 2.3) + _ int = max(1.2, 3) // ok! + _ byte = max(1, 'a') + ) +} + +func min1() { + var b bool + var c complex128 + var x int + var s string + type myint int + var m myint + _ = min() /* ERROR "not enough arguments" */ + _ = min(b /* ERROR "cannot be ordered" */ ) + _ = min(c /* ERROR "cannot be ordered" */ ) + _ = min(x) + _ = min(s) + _ = min(x, x) + _ = min(x, x, x, x, x) + var _ int = min /* ERROR "cannot use min(m) (value of type myint) as int value" */ (m) + _ = min(x, m /* ERROR "invalid argument: mismatched types int (previous argument) and myint (type of m)" */ , x) + + _ = min(1, x) + _ = min(1.0, x) + _ = min(1.2 /* ERROR "1.2 (untyped float constant) truncated to int" */ , x) + _ = min(-10, 1.0, c /* ERROR "cannot be ordered" */ ) + + const ( + _ = min /* ERROR "min(x) (value of type int) is not constant" */ (x) + _ = min(true /* ERROR "invalid argument: true (untyped bool constant) cannot be ordered" */ ) + _ = min(1) + _ = min(1, 2.3, 'a') + _ = min(1, "foo" /* ERROR "mismatched types" */ ) + _ = min(1, 0i /* ERROR "cannot be ordered" */ ) + _ = min(1, 2 /* ERROR "cannot be ordered" */ + 3i ) + ) +} + +func min2() { + _ = assert(min(0) == 0) + _ = assert(min(0, 1) == 0) + _ = assert(min(0, -10, 123456789) == -10) + _ = assert(min(-12345678901234567890, 0) == -12345678901234567890) + + _ = assert(min(1, 2.3) == 1) + _ = assert(min(1, 2.3, 'a') == 1) + + _ = assert(min("", "a") == "") + _ = assert(min("abcde", "xyz", "foo", "bar") == "abcde") + + const ( + _ int = min(1.0) + _ float32 = min(1, 2) + _ int = min(1, 2.3) // ok! + _ int = min /* ERROR "cannot use min(1.2, 3) (untyped float constant 1.2) as int value" */ (1.2, 3) + _ byte = min(1, 'a') + ) +} + +func new1() { + _ = new() // ERROR "not enough arguments" + _ = new(1, 2) // ERROR "too many arguments" + _ = new("foo" /* ERROR "not a type" */) + p := new(float64) + _ = new(struct{ x, y int }) + q := new(*float64) + _ = *p == **q + new /* ERROR "not used" */ (int) + _ = &new /* ERROR "cannot take address" */ (int) + + _ = new(int... /* ERROR "invalid use of ..." */ ) +} + +func new2() { + f1 := func() (x []int) { return } + _ = new(f0 /* ERROR "not a type" */ ()) + _ = new(f1 /* ERROR "not a type" */ ()) +} + +func panic1() { + panic() // ERROR "not enough arguments" + panic(1, 2) // ERROR "too many arguments" + panic(0) + panic("foo") + panic(false) + panic(1<<10) + panic(1 << /* ERROR "constant shift overflow" */ 1000) + _ = panic /* ERROR "used as value" */ (0) + + var s []byte + panic(s) + panic(s... /* ERROR "invalid use of ..." */ ) +} + +func panic2() { + f1 := func() (x int) { return } + f2 := func() (x, y int) { return } + panic(f0 /* ERROR "used as value" */ ()) + panic(f1()) + panic(f2()) // ERROR "too many arguments" +} + +func print1() { + print() + print(1) + print(1, 2) + print("foo") + print(2.718281828) + print(false) + print(1<<10) + print(1 << /* ERROR "constant shift overflow" */ 1000) + println(nil /* ERROR "untyped nil" */ ) + + var s []int + print(s... /* ERROR "invalid use of ..." */ ) + _ = print /* ERROR "used as value" */ () +} + +func print2() { + f1 := func() (x int) { return } + f2 := func() (x, y int) { return } + f3 := func() (x int, y float32, z string) { return } + print(f0 /* ERROR "used as value" */ ()) + print(f1()) + print(f2()) + print(f3()) +} + +func println1() { + println() + println(1) + println(1, 2) + println("foo") + println(2.718281828) + println(false) + println(1<<10) + println(1 << /* ERROR "constant shift overflow" */ 1000) + println(nil /* ERROR "untyped nil" */ ) + + var s []int + println(s... /* ERROR "invalid use of ..." */ ) + _ = println /* ERROR "used as value" */ () +} + +func println2() { + f1 := func() (x int) { return } + f2 := func() (x, y int) { return } + f3 := func() (x int, y float32, z string) { return } + println(f0 /* ERROR "used as value" */ ()) + println(f1()) + println(f2()) + println(f3()) +} + +func real1() { + var f32 float32 + var f64 float64 + var c64 complex64 + var c128 complex128 + _ = real() // ERROR "not enough arguments" + _ = real(1, 2) // ERROR "too many arguments" + _ = real(10) + _ = real(2.7182818) + _ = real("foo" /* ERROR "expected complex" */) + const _5 = real(1 + 2i) + assert(_5 == 1) + f32 = _5 + f64 = _5 + const _6 = real(0i) + assert(_6 == 0) + f32 = real(c64) + f64 = real(c128) + f32 = real /* ERRORx `cannot use .* in assignment` */ (c128) + f64 = real /* ERRORx `cannot use .* in assignment` */ (c64) + real /* ERROR "not used" */ (c64) + + // complex type may not be predeclared + type C64 complex64 + type C128 complex128 + var x64 C64 + var x128 C128 + f32 = imag(x64) + f64 = imag(x128) + _, _ = f32, f64 + + var a []complex64 + _ = real(a... /* ERROR "invalid use of ..." */ ) + + // if argument is untyped, result is untyped + const _ byte = real(1 + 2.3i) + const _ complex128 = real(1 + 2.3i) + + // lhs constant shift operands are typed as complex128 + var s uint + _ = real(1 /* ERROR "must be integer" */ << s) +} + +func real2() { + f1 := func() (x complex128) { return } + f2 := func() (x, y complex128) { return } + _ = real(f0 /* ERROR "used as value" */ ()) + _ = real(f1()) + _ = real(f2()) // ERROR "too many arguments" +} + +func recover1() { + _ = recover() + _ = recover(10) // ERROR "too many arguments" + recover() + + var s []int + recover(s... /* ERROR "invalid use of ..." */ ) +} + +func recover2() { + f1 := func() (x int) { return } + f2 := func() (x, y int) { return } + _ = recover(f0 /* ERROR "used as value" */ ()) + _ = recover(f1()) // ERROR "too many arguments" + _ = recover(f2()) // ERROR "too many arguments" +} + +// assuming types.DefaultPtrSize == 8 +type S0 struct{ // offset + a bool // 0 + b rune // 4 + c *int // 8 + d bool // 16 + e complex128 // 24 +} // 40 + +type S1 struct{ // offset + x float32 // 0 + y string // 8 + z *S1 // 24 + S0 // 32 +} // 72 + +type S2 struct{ // offset + *S1 // 0 +} // 8 + +type S3 struct { // offset + a int64 // 0 + b int32 // 8 +} // 12 + +type S4 struct { // offset + S3 // 0 + int32 // 12 +} // 16 + +type S5 struct { // offset + a [3]int32 // 0 + b int32 // 12 +} // 16 + +func (S2) m() {} + +func Alignof1() { + var x int + _ = unsafe.Alignof() // ERROR "not enough arguments" + _ = unsafe.Alignof(1, 2) // ERROR "too many arguments" + _ = unsafe.Alignof(int /* ERROR "not an expression" */) + _ = unsafe.Alignof(42) + _ = unsafe.Alignof(new(struct{})) + _ = unsafe.Alignof(1<<10) + _ = unsafe.Alignof(1 << /* ERROR "constant shift overflow" */ 1000) + _ = unsafe.Alignof(nil /* ERROR "untyped nil" */ ) + unsafe /* ERROR "not used" */ .Alignof(x) + + var y S0 + assert(unsafe.Alignof(y.a) == 1) + assert(unsafe.Alignof(y.b) == 4) + assert(unsafe.Alignof(y.c) == 8) + assert(unsafe.Alignof(y.d) == 1) + assert(unsafe.Alignof(y.e) == 8) + + var s []byte + _ = unsafe.Alignof(s) + _ = unsafe.Alignof(s... /* ERROR "invalid use of ..." */ ) +} + +func Alignof2() { + f1 := func() (x int32) { return } + f2 := func() (x, y int32) { return } + _ = unsafe.Alignof(f0 /* ERROR "used as value" */ ()) + assert(unsafe.Alignof(f1()) == 4) + _ = unsafe.Alignof(f2()) // ERROR "too many arguments" +} + +func Offsetof1() { + var x struct{ f int } + _ = unsafe.Offsetof() // ERROR "not enough arguments" + _ = unsafe.Offsetof(1, 2) // ERROR "too many arguments" + _ = unsafe.Offsetof(int /* ERROR "not a selector expression" */ ) + _ = unsafe.Offsetof(x /* ERROR "not a selector expression" */ ) + _ = unsafe.Offsetof(nil /* ERROR "not a selector expression" */ ) + _ = unsafe.Offsetof(x.f) + _ = unsafe.Offsetof((x.f)) + _ = unsafe.Offsetof((((((((x))).f))))) + unsafe /* ERROR "not used" */ .Offsetof(x.f) + + var y0 S0 + assert(unsafe.Offsetof(y0.a) == 0) + assert(unsafe.Offsetof(y0.b) == 4) + assert(unsafe.Offsetof(y0.c) == 8) + assert(unsafe.Offsetof(y0.d) == 16) + assert(unsafe.Offsetof(y0.e) == 24) + + var y1 S1 + assert(unsafe.Offsetof(y1.x) == 0) + assert(unsafe.Offsetof(y1.y) == 8) + assert(unsafe.Offsetof(y1.z) == 24) + assert(unsafe.Offsetof(y1.S0) == 32) + + assert(unsafe.Offsetof(y1.S0.a) == 0) // relative to S0 + assert(unsafe.Offsetof(y1.a) == 32) // relative to S1 + assert(unsafe.Offsetof(y1.b) == 36) // relative to S1 + assert(unsafe.Offsetof(y1.c) == 40) // relative to S1 + assert(unsafe.Offsetof(y1.d) == 48) // relative to S1 + assert(unsafe.Offsetof(y1.e) == 56) // relative to S1 + + var y1p *S1 + assert(unsafe.Offsetof(y1p.S0) == 32) + + type P *S1 + var p P = y1p + assert(unsafe.Offsetof(p.S0) == 32) + + var y2 S2 + assert(unsafe.Offsetof(y2.S1) == 0) + _ = unsafe.Offsetof(y2 /* ERROR "embedded via a pointer" */ .x) + _ = unsafe.Offsetof(y2 /* ERROR "method value" */ .m) + + var s []byte + _ = unsafe.Offsetof(s... /* ERROR "invalid use of ..." */ ) +} + +func Offsetof2() { + f1 := func() (x int32) { return } + f2 := func() (x, y int32) { return } + _ = unsafe.Offsetof(f0 /* ERROR "not a selector expression" */ ()) + _ = unsafe.Offsetof(f1 /* ERROR "not a selector expression" */ ()) + _ = unsafe.Offsetof(f2 /* ERROR "not a selector expression" */ ()) +} + +func Sizeof1() { + var x int + _ = unsafe.Sizeof() // ERROR "not enough arguments" + _ = unsafe.Sizeof(1, 2) // ERROR "too many arguments" + _ = unsafe.Sizeof(int /* ERROR "not an expression" */) + _ = unsafe.Sizeof(42) + _ = unsafe.Sizeof(new(complex128)) + _ = unsafe.Sizeof(1<<10) + _ = unsafe.Sizeof(1 << /* ERROR "constant shift overflow" */ 1000) + _ = unsafe.Sizeof(nil /* ERROR "untyped nil" */ ) + unsafe /* ERROR "not used" */ .Sizeof(x) + + // basic types have size guarantees + assert(unsafe.Sizeof(byte(0)) == 1) + assert(unsafe.Sizeof(uint8(0)) == 1) + assert(unsafe.Sizeof(int8(0)) == 1) + assert(unsafe.Sizeof(uint16(0)) == 2) + assert(unsafe.Sizeof(int16(0)) == 2) + assert(unsafe.Sizeof(uint32(0)) == 4) + assert(unsafe.Sizeof(int32(0)) == 4) + assert(unsafe.Sizeof(float32(0)) == 4) + assert(unsafe.Sizeof(uint64(0)) == 8) + assert(unsafe.Sizeof(int64(0)) == 8) + assert(unsafe.Sizeof(float64(0)) == 8) + assert(unsafe.Sizeof(complex64(0)) == 8) + assert(unsafe.Sizeof(complex128(0)) == 16) + + var y0 S0 + assert(unsafe.Sizeof(y0.a) == 1) + assert(unsafe.Sizeof(y0.b) == 4) + assert(unsafe.Sizeof(y0.c) == 8) + assert(unsafe.Sizeof(y0.d) == 1) + assert(unsafe.Sizeof(y0.e) == 16) + assert(unsafe.Sizeof(y0) == 40) + + var y1 S1 + assert(unsafe.Sizeof(y1) == 72) + + var y2 S2 + assert(unsafe.Sizeof(y2) == 8) + + var y3 S3 + assert(unsafe.Sizeof(y3) == 12) + + var y4 S4 + assert(unsafe.Sizeof(y4) == 16) + + var y5 S5 + assert(unsafe.Sizeof(y5) == 16) + + var a3 [10]S3 + assert(unsafe.Sizeof(a3) == 156) + + // test case for issue 5670 + type T struct { + a int32 + _ int32 + c int32 + } + assert(unsafe.Sizeof(T{}) == 12) + + var s []byte + _ = unsafe.Sizeof(s) + _ = unsafe.Sizeof(s... /* ERROR "invalid use of ..." */ ) +} + +func Sizeof2() { + f1 := func() (x int64) { return } + f2 := func() (x, y int64) { return } + _ = unsafe.Sizeof(f0 /* ERROR "used as value" */ ()) + assert(unsafe.Sizeof(f1()) == 8) + _ = unsafe.Sizeof(f2()) // ERROR "too many arguments" +} + +func Slice1() { + var x int + unsafe.Slice() // ERROR "not enough arguments" + unsafe.Slice(1, 2, 3) // ERROR "too many arguments" + unsafe.Slice(1 /* ERROR "is not a pointer" */ , 2) + unsafe.Slice(nil /* ERROR "nil is not a pointer" */ , 0) + unsafe.Slice(&x, "foo" /* ERRORx `cannot convert .* to type int` */ ) + unsafe.Slice(&x, 1.2 /* ERROR "truncated to int" */ ) + unsafe.Slice(&x, - /* ERROR "must not be negative" */ 1) + unsafe /* ERROR "not used" */ .Slice(&x, 0) + var _ []byte = unsafe /* ERROR "value of type []int" */ .Slice(&x, 0) + + var _ []int = unsafe.Slice(&x, 0) + _ = unsafe.Slice(&x, 1.0) + _ = unsafe.Slice((*int)(nil), 0) +} + +func SliceData1() { + var s []int + unsafe.SliceData(0 /* ERROR "not a slice" */) + unsafe /* ERROR "not used" */ .SliceData(s) + + type S []int + _ = unsafe.SliceData(s) + _ = unsafe.SliceData(S{}) +} + +func String1() { + var b byte + unsafe.String() // ERROR "not enough arguments" + unsafe.String(1, 2, 3) // ERROR "too many arguments" + unsafe.String(1 /* ERROR "cannot use 1" */ , 2) + unsafe.String(&b, "foo" /* ERRORx `cannot convert .* to type int` */ ) + unsafe.String(&b, 1.2 /* ERROR "truncated to int" */ ) + unsafe.String(&b, - /* ERROR "must not be negative" */ 1) + unsafe /* ERROR "not used" */ .String(&b, 0) + var _ []byte = unsafe /* ERROR "value of type string" */ .String(&b, 0) + + var _ string = unsafe.String(&b, 0) + _ = unsafe.String(&b, 1.0) + _ = unsafe.String(nil, 0) // here we allow nil as ptr argument (in contrast to unsafe.Slice) +} + +func StringData1() { + var s string + type S string + unsafe.StringData(0 /* ERROR "cannot use 0" */) + unsafe.StringData(S /* ERROR "cannot use S" */ ("foo")) + unsafe /* ERROR "not used" */ .StringData(s) + + _ = unsafe.StringData(s) + _ = unsafe.StringData("foo") +} + +// self-testing only +func assert1() { + var x int + assert() /* ERROR "not enough arguments" */ + assert(1, 2) /* ERROR "too many arguments" */ + assert("foo" /* ERROR "boolean constant" */ ) + assert(x /* ERROR "boolean constant" */) + assert(true) + assert /* ERROR "failed" */ (false) + _ = assert(true) + + var s []byte + assert(s... /* ERROR "invalid use of ..." */ ) +} + +func assert2() { + f1 := func() (x bool) { return } + f2 := func() (x bool) { return } + assert(f0 /* ERROR "used as value" */ ()) + assert(f1 /* ERROR "boolean constant" */ ()) + assert(f2 /* ERROR "boolean constant" */ ()) +} + +// self-testing only +func trace1() { + // Uncomment the code below to test trace - will produce console output + // _ = trace /* ERROR "no value" */ () + // _ = trace(1) + // _ = trace(true, 1.2, '\'', "foo", 42i, "foo" <= "bar") + + var s []byte + trace(s... /* ERROR "invalid use of ..." */ ) +} + +func trace2() { + f1 := func() (x int) { return } + f2 := func() (x int, y string) { return } + f3 := func() (x int, y string, z []int) { return } + _ = f1 + _ = f2 + _ = f3 + // Uncomment the code below to test trace - will produce console output + // trace(f0()) + // trace(f1()) + // trace(f2()) + // trace(f3()) + // trace(f0(), 1) + // trace(f1(), 1, 2) + // trace(f2(), 1, 2, 3) + // trace(f3(), 1, 2, 3, 4) +} diff --git a/src/internal/types/testdata/check/builtins1.go b/src/internal/types/testdata/check/builtins1.go new file mode 100644 index 0000000..f7ac72d --- /dev/null +++ b/src/internal/types/testdata/check/builtins1.go @@ -0,0 +1,330 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file tests built-in calls on generic types. + +package builtins + +import "unsafe" + +// clear + +func _[T any](x T) { + clear(x /* ERROR "cannot clear x" */) +} + +func _[T ~map[int]string | ~[]byte](x T) { + clear(x) +} + +func _[T ~map[int]string | ~[]byte | ~*[10]int | string](x T) { + clear(x /* ERROR "cannot clear x" */) +} + +// close + +type C0 interface{ int } +type C1 interface{ chan int } +type C2 interface{ chan int | <-chan int } +type C3 interface{ chan int | chan float32 } +type C4 interface{ chan int | chan<- int } +type C5[T any] interface{ ~chan T | chan<- T } + +func _[T any](ch T) { + close(ch /* ERROR "cannot close non-channel" */) +} + +func _[T C0](ch T) { + close(ch /* ERROR "cannot close non-channel" */) +} + +func _[T C1](ch T) { + close(ch) +} + +func _[T C2](ch T) { + close(ch /* ERROR "cannot close receive-only channel" */) +} + +func _[T C3](ch T) { + close(ch) +} + +func _[T C4](ch T) { + close(ch) +} + +func _[T C5[X], X any](ch T) { + close(ch) +} + +// copy + +func _[T any](x, y T) { + copy(x /* ERROR "copy expects slice arguments" */ , y) +} + +func _[T ~[]byte](x, y T) { + copy(x, y) + copy(x, "foo") + copy("foo" /* ERROR "expects slice arguments" */ , y) + + var x2 []byte + copy(x2, y) // element types are identical + copy(y, x2) // element types are identical + + type myByte byte + var x3 []myByte + copy(x3 /* ERROR "different element types" */ , y) + copy(y /* ERROR "different element types" */ , x3) +} + +func _[T ~[]E, E any](x T, y []E) { + copy(x, y) + copy(x /* ERROR "different element types" */ , "foo") +} + +func _[T ~string](x []byte, y T) { + copy(x, y) + copy(y /* ERROR "expects slice arguments" */ , x) +} + +func _[T ~[]byte|~string](x T, y []byte) { + copy(x /* ERROR "expects slice arguments" */ , y) + copy(y, x) +} + +type L0 []int +type L1 []int + +func _[T L0 | L1](x, y T) { + copy(x, y) +} + +// delete + +type M0 interface{ int } +type M1 interface{ map[string]int } +type M2 interface { map[string]int | map[string]float64 } +type M3 interface{ map[string]int | map[rune]int } +type M4[K comparable, V any] interface{ map[K]V | map[rune]V } + +func _[T any](m T) { + delete(m /* ERROR "not a map" */, "foo") +} + +func _[T M0](m T) { + delete(m /* ERROR "not a map" */, "foo") +} + +func _[T M1](m T) { + delete(m, "foo") +} + +func _[T M2](m T) { + delete(m, "foo") + delete(m, 0 /* ERRORx `cannot use .* as string` */) +} + +func _[T M3](m T) { + delete(m /* ERROR "must have identical key types" */, "foo") +} + +func _[T M4[rune, V], V any](m T) { + delete(m, 'k') +} + +func _[T M4[K, V], K comparable, V any](m T) { + delete(m /* ERROR "must have identical key types" */, "foo") +} + +// make + +type myChan chan int + +func _[ + S1 ~[]int, + S2 ~[]int | ~chan int, + + M1 ~map[string]int, + M2 ~map[string]int | ~chan int, + + C1 ~chan int, + C2 ~chan int | ~chan string, + C3 chan int | myChan, // single underlying type +]() { + type S0 []int + _ = make([]int, 10) + _ = make(S0, 10) + _ = make(S1, 10) + _ = make() /* ERROR "not enough arguments" */ + _ = make /* ERROR "expects 2 or 3 arguments" */ (S1) + _ = make(S1, 10, 20) + _ = make /* ERROR "expects 2 or 3 arguments" */ (S1, 10, 20, 30) + _ = make(S2 /* ERROR "cannot make S2: no core type" */ , 10) + + type M0 map[string]int + _ = make(map[string]int) + _ = make(M0) + _ = make(M1) + _ = make(M1, 10) + _ = make/* ERROR "expects 1 or 2 arguments" */(M1, 10, 20) + _ = make(M2 /* ERROR "cannot make M2: no core type" */ ) + + type C0 chan int + _ = make(chan int) + _ = make(C0) + _ = make(C1) + _ = make(C1, 10) + _ = make/* ERROR "expects 1 or 2 arguments" */(C1, 10, 20) + _ = make(C2 /* ERROR "cannot make C2: no core type" */ ) + _ = make(C3) +} + +// max + +func _[ + P1 ~int|~float64, + P2 ~int|~string|~uint, + P3 ~int|bool, +]() { + var x1 P1 + _ = max(x1) + _ = max(x1, x1) + _ = max(1, x1, 2) + const _ = max /* ERROR "max(1, x1, 2) (value of type P1 constrained by ~int | ~float64) is not constant" */ (1, x1, 2) + + var x2 P2 + _ = max(x2) + _ = max(x2, x2) + _ = max(1, 2 /* ERROR "cannot convert 2 (untyped int constant) to type P2" */, x2) // error at 2 because max is 2 + + _ = max(x1, x2 /* ERROR "mismatched types P1 (previous argument) and P2 (type of x2)" */ ) +} + +// min + +func _[ + P1 ~int|~float64, + P2 ~int|~string|~uint, + P3 ~int|bool, +]() { + var x1 P1 + _ = min(x1) + _ = min(x1, x1) + _ = min(1, x1, 2) + const _ = min /* ERROR "min(1, x1, 2) (value of type P1 constrained by ~int | ~float64) is not constant" */ (1, x1, 2) + + var x2 P2 + _ = min(x2) + _ = min(x2, x2) + _ = min(1 /* ERROR "cannot convert 1 (untyped int constant) to type P2" */ , 2, x2) // error at 1 because min is 1 + + _ = min(x1, x2 /* ERROR "mismatched types P1 (previous argument) and P2 (type of x2)" */ ) +} + +// unsafe.Alignof + +func _[T comparable]() { + var ( + b int64 + a [10]T + s struct{ f T } + p *T + l []T + f func(T) + i interface{ m() T } + c chan T + m map[T]T + t T + ) + + const bb = unsafe.Alignof(b) + assert(bb == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(a) + const _ = unsafe /* ERROR "not constant" */ .Alignof(s) + const pp = unsafe.Alignof(p) + assert(pp == 8) + const ll = unsafe.Alignof(l) + assert(ll == 8) + const ff = unsafe.Alignof(f) + assert(ff == 8) + const ii = unsafe.Alignof(i) + assert(ii == 8) + const cc = unsafe.Alignof(c) + assert(cc == 8) + const mm = unsafe.Alignof(m) + assert(mm == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(t) +} + +// unsafe.Offsetof + +func _[T comparable]() { + var ( + b struct{ _, f int64 } + a struct{ _, f [10]T } + s struct{ _, f struct{ f T } } + p struct{ _, f *T } + l struct{ _, f []T } + f struct{ _, f func(T) } + i struct{ _, f interface{ m() T } } + c struct{ _, f chan T } + m struct{ _, f map[T]T } + t struct{ _, f T } + ) + + const bb = unsafe.Offsetof(b.f) + assert(bb == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(a) + const _ = unsafe /* ERROR "not constant" */ .Alignof(s) + const pp = unsafe.Offsetof(p.f) + assert(pp == 8) + const ll = unsafe.Offsetof(l.f) + assert(ll == 24) + const ff = unsafe.Offsetof(f.f) + assert(ff == 8) + const ii = unsafe.Offsetof(i.f) + assert(ii == 16) + const cc = unsafe.Offsetof(c.f) + assert(cc == 8) + const mm = unsafe.Offsetof(m.f) + assert(mm == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(t) +} + +// unsafe.Sizeof + +func _[T comparable]() { + var ( + b int64 + a [10]T + s struct{ f T } + p *T + l []T + f func(T) + i interface{ m() T } + c chan T + m map[T]T + t T + ) + + const bb = unsafe.Sizeof(b) + assert(bb == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(a) + const _ = unsafe /* ERROR "not constant" */ .Alignof(s) + const pp = unsafe.Sizeof(p) + assert(pp == 8) + const ll = unsafe.Sizeof(l) + assert(ll == 24) + const ff = unsafe.Sizeof(f) + assert(ff == 8) + const ii = unsafe.Sizeof(i) + assert(ii == 16) + const cc = unsafe.Sizeof(c) + assert(cc == 8) + const mm = unsafe.Sizeof(m) + assert(mm == 8) + const _ = unsafe /* ERROR "not constant" */ .Alignof(t) +} diff --git a/src/internal/types/testdata/check/chans.go b/src/internal/types/testdata/check/chans.go new file mode 100644 index 0000000..fad2bce --- /dev/null +++ b/src/internal/types/testdata/check/chans.go @@ -0,0 +1,62 @@ +package chans + +import "runtime" + +// Ranger returns a Sender and a Receiver. The Receiver provides a +// Next method to retrieve values. The Sender provides a Send method +// to send values and a Close method to stop sending values. The Next +// method indicates when the Sender has been closed, and the Send +// method indicates when the Receiver has been freed. +// +// This is a convenient way to exit a goroutine sending values when +// the receiver stops reading them. +func Ranger[T any]() (*Sender[T], *Receiver[T]) { + c := make(chan T) + d := make(chan bool) + s := &Sender[T]{values: c, done: d} + r := &Receiver[T]{values: c, done: d} + runtime.SetFinalizer(r, r.finalize) + return s, r +} + +// A sender is used to send values to a Receiver. +type Sender[T any] struct { + values chan<- T + done <-chan bool +} + +// Send sends a value to the receiver. It returns whether any more +// values may be sent; if it returns false the value was not sent. +func (s *Sender[T]) Send(v T) bool { + select { + case s.values <- v: + return true + case <-s.done: + return false + } +} + +// Close tells the receiver that no more values will arrive. +// After Close is called, the Sender may no longer be used. +func (s *Sender[T]) Close() { + close(s.values) +} + +// A Receiver receives values from a Sender. +type Receiver[T any] struct { + values <-chan T + done chan<- bool +} + +// Next returns the next value from the channel. The bool result +// indicates whether the value is valid, or whether the Sender has +// been closed and no more values will be received. +func (r *Receiver[T]) Next() (T, bool) { + v, ok := <-r.values + return v, ok +} + +// finalize is a finalizer for the receiver. +func (r *Receiver[T]) finalize() { + close(r.done) +} diff --git a/src/internal/types/testdata/check/compliterals.go b/src/internal/types/testdata/check/compliterals.go new file mode 100644 index 0000000..60eac97 --- /dev/null +++ b/src/internal/types/testdata/check/compliterals.go @@ -0,0 +1,22 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Composite literals with parameterized types + +package comp_literals + +type myStruct struct { + f int +} + +type slice[E any] []E + +func struct_literals[S struct{f int}|myStruct]() { + _ = S{} + _ = S{0} + _ = S{f: 0} + + _ = slice[int]{1, 2, 3} + _ = slice[S]{{}, {0}, {f:0}} +} diff --git a/src/internal/types/testdata/check/const0.go b/src/internal/types/testdata/check/const0.go new file mode 100644 index 0000000..49c62d6 --- /dev/null +++ b/src/internal/types/testdata/check/const0.go @@ -0,0 +1,382 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// constant declarations + +package const0 + +import "unsafe" + +// constants declarations must be initialized by constants +var x = 0 +const c0 = x /* ERROR "not constant" */ + +// typed constants must have constant types +const _ interface /* ERROR "invalid constant type" */ {} = 0 + +func _ () { + const _ interface /* ERROR "invalid constant type" */ {} = 0 + for i := 0; i < 10; i++ {} // don't crash with non-nil iota here +} + +// untyped constants +const ( + // boolean values + ub0 = false + ub1 = true + ub2 = 2 < 1 + ub3 = ui1 == uf1 + ub4 = true == 0 /* ERROR "mismatched types untyped bool and untyped int" */ + + // integer values + ui0 = 0 + ui1 = 1 + ui2 = 42 + ui3 = 3141592653589793238462643383279502884197169399375105820974944592307816406286 + ui4 = -10 + + ui5 = ui0 + ui1 + ui6 = ui1 - ui1 + ui7 = ui2 * ui1 + ui8 = ui3 / ui3 + ui9 = ui3 % ui3 + + ui10 = 1 / 0 /* ERROR "division by zero" */ + ui11 = ui1 / 0 /* ERROR "division by zero" */ + ui12 = ui3 / ui0 /* ERROR "division by zero" */ + ui13 = 1 % 0 /* ERROR "division by zero" */ + ui14 = ui1 % 0 /* ERROR "division by zero" */ + ui15 = ui3 % ui0 /* ERROR "division by zero" */ + + ui16 = ui2 & ui3 + ui17 = ui2 | ui3 + ui18 = ui2 ^ ui3 + ui19 = 1 /* ERROR "invalid operation" */ % 1.0 + + // floating point values + uf0 = 0. + uf1 = 1. + uf2 = 4.2e1 + uf3 = 3.141592653589793238462643383279502884197169399375105820974944592307816406286 + uf4 = 1e-1 + + uf5 = uf0 + uf1 + uf6 = uf1 - uf1 + uf7 = uf2 * uf1 + uf8 = uf3 / uf3 + uf9 = uf3 /* ERROR "not defined" */ % uf3 + + uf10 = 1 / 0 /* ERROR "division by zero" */ + uf11 = uf1 / 0 /* ERROR "division by zero" */ + uf12 = uf3 / uf0 /* ERROR "division by zero" */ + + uf16 = uf2 /* ERROR "not defined" */ & uf3 + uf17 = uf2 /* ERROR "not defined" */ | uf3 + uf18 = uf2 /* ERROR "not defined" */ ^ uf3 + + // complex values + uc0 = 0.i + uc1 = 1.i + uc2 = 4.2e1i + uc3 = 3.141592653589793238462643383279502884197169399375105820974944592307816406286i + uc4 = 1e-1i + + uc5 = uc0 + uc1 + uc6 = uc1 - uc1 + uc7 = uc2 * uc1 + uc8 = uc3 / uc3 + uc9 = uc3 /* ERROR "not defined" */ % uc3 + + uc10 = 1 / 0 /* ERROR "division by zero" */ + uc11 = uc1 / 0 /* ERROR "division by zero" */ + uc12 = uc3 / uc0 /* ERROR "division by zero" */ + + uc16 = uc2 /* ERROR "not defined" */ & uc3 + uc17 = uc2 /* ERROR "not defined" */ | uc3 + uc18 = uc2 /* ERROR "not defined" */ ^ uc3 +) + +type ( + mybool bool + myint int + myfloat float64 + mycomplex complex128 +) + +// typed constants +const ( + // boolean values + tb0 bool = false + tb1 bool = true + tb2 mybool = 2 < 1 + tb3 mybool = ti1 == tf1 /* ERROR "mismatched types" */ + + // integer values + ti0 int8 = ui0 + ti1 int32 = ui1 + ti2 int64 = ui2 + ti3 myint = ui3 /* ERROR "overflows" */ + ti4 myint = ui4 + + ti5 = ti0 /* ERROR "mismatched types" */ + ti1 + ti6 = ti1 - ti1 + ti7 = ti2 /* ERROR "mismatched types" */ * ti1 + ti8 = ti3 / ti3 + ti9 = ti3 % ti3 + + ti10 = 1 / 0 /* ERROR "division by zero" */ + ti11 = ti1 / 0 /* ERROR "division by zero" */ + ti12 = ti3 /* ERROR "mismatched types" */ / ti0 + ti13 = 1 % 0 /* ERROR "division by zero" */ + ti14 = ti1 % 0 /* ERROR "division by zero" */ + ti15 = ti3 /* ERROR "mismatched types" */ % ti0 + + ti16 = ti2 /* ERROR "mismatched types" */ & ti3 + ti17 = ti2 /* ERROR "mismatched types" */ | ti4 + ti18 = ti2 ^ ti5 // no mismatched types error because the type of ti5 is unknown + + // floating point values + tf0 float32 = 0. + tf1 float32 = 1. + tf2 float64 = 4.2e1 + tf3 myfloat = 3.141592653589793238462643383279502884197169399375105820974944592307816406286 + tf4 myfloat = 1e-1 + + tf5 = tf0 + tf1 + tf6 = tf1 - tf1 + tf7 = tf2 /* ERROR "mismatched types" */ * tf1 + tf8 = tf3 / tf3 + tf9 = tf3 /* ERROR "not defined" */ % tf3 + + tf10 = 1 / 0 /* ERROR "division by zero" */ + tf11 = tf1 / 0 /* ERROR "division by zero" */ + tf12 = tf3 /* ERROR "mismatched types" */ / tf0 + + tf16 = tf2 /* ERROR "mismatched types" */ & tf3 + tf17 = tf2 /* ERROR "mismatched types" */ | tf3 + tf18 = tf2 /* ERROR "mismatched types" */ ^ tf3 + + // complex values + tc0 = 0.i + tc1 = 1.i + tc2 = 4.2e1i + tc3 = 3.141592653589793238462643383279502884197169399375105820974944592307816406286i + tc4 = 1e-1i + + tc5 = tc0 + tc1 + tc6 = tc1 - tc1 + tc7 = tc2 * tc1 + tc8 = tc3 / tc3 + tc9 = tc3 /* ERROR "not defined" */ % tc3 + + tc10 = 1 / 0 /* ERROR "division by zero" */ + tc11 = tc1 / 0 /* ERROR "division by zero" */ + tc12 = tc3 / tc0 /* ERROR "division by zero" */ + + tc16 = tc2 /* ERROR "not defined" */ & tc3 + tc17 = tc2 /* ERROR "not defined" */ | tc3 + tc18 = tc2 /* ERROR "not defined" */ ^ tc3 +) + +// initialization cycles +const ( + a /* ERROR "initialization cycle" */ = a + b /* ERROR "initialization cycle" */ , c /* ERROR "initialization cycle" */, d, e = e, d, c, b // TODO(gri) should only have one cycle error + f float64 = d +) + +// multiple initialization +const ( + a1, a2, a3 = 7, 3.1415926, "foo" + b1, b2, b3 = b3, b1, 42 + c1, c2, c3 /* ERROR "missing init expr for c3" */ = 1, 2 + d1, d2, d3 = 1, 2, 3, 4 /* ERROR "extra init expr 4" */ + _p0 = assert(a1 == 7) + _p1 = assert(a2 == 3.1415926) + _p2 = assert(a3 == "foo") + _p3 = assert(b1 == 42) + _p4 = assert(b2 == 42) + _p5 = assert(b3 == 42) +) + +func _() { + const ( + a1, a2, a3 = 7, 3.1415926, "foo" + b1, b2, b3 = b3, b1, 42 + c1, c2, c3 /* ERROR "missing init expr for c3" */ = 1, 2 + d1, d2, d3 = 1, 2, 3, 4 /* ERROR "extra init expr 4" */ + _p0 = assert(a1 == 7) + _p1 = assert(a2 == 3.1415926) + _p2 = assert(a3 == "foo") + _p3 = assert(b1 == 42) + _p4 = assert(b2 == 42) + _p5 = assert(b3 == 42) + ) +} + +// iota +const ( + iota0 = iota + iota1 = iota + iota2 = iota*2 + _a0 = assert(iota0 == 0) + _a1 = assert(iota1 == 1) + _a2 = assert(iota2 == 4) + iota6 = iota*3 + + iota7 + iota8 + _a3 = assert(iota7 == 21) + _a4 = assert(iota8 == 24) +) + +const ( + _b0 = iota + _b1 = assert(iota + iota2 == 5) + _b2 = len([iota]int{}) // iota may appear in a type! + _b3 = assert(_b2 == 2) + _b4 = len(A{}) +) + +type A [iota /* ERROR "cannot use iota" */ ]int + +// constant expressions with operands across different +// constant declarations must use the right iota values +const ( + _c0 = iota + _c1 + _c2 + _x = _c2 + _d1 + _e0 // 3 +) + +const ( + _d0 = iota + _d1 +) + +const ( + _e0 = iota +) + +var _ = assert(_x == 3) + +// special cases +const ( + _n0 = nil /* ERROR "not constant" */ + _n1 = [ /* ERROR "not constant" */ ]int{} +) + +// iotas must not be usable in expressions outside constant declarations +type _ [iota /* ERROR "iota outside constant decl" */ ]byte +var _ = iota /* ERROR "iota outside constant decl" */ +func _() { + _ = iota /* ERROR "iota outside constant decl" */ + const _ = iota + _ = iota /* ERROR "iota outside constant decl" */ +} + +func _() { + iota := 123 + const x = iota /* ERROR "is not constant" */ + var y = iota + _ = y +} + +// iotas are usable inside closures in constant declarations (#22345) +const ( + _ = iota + _ = len([iota]byte{}) + _ = unsafe.Sizeof(iota) + _ = unsafe.Sizeof(func() { _ = iota }) + _ = unsafe.Sizeof(func() { var _ = iota }) + _ = unsafe.Sizeof(func() { const _ = iota }) + _ = unsafe.Sizeof(func() { type _ [iota]byte }) + _ = unsafe.Sizeof(func() { func() int { return iota }() }) +) + +// verify inner and outer const declarations have distinct iotas +const ( + zero = iota + one = iota + _ = unsafe.Sizeof(func() { + var x [iota]int // [2]int + const ( + Zero = iota + One + Two + _ = unsafe.Sizeof([iota-1]int{} == x) // assert types are equal + _ = unsafe.Sizeof([Two]int{} == x) // assert types are equal + ) + var z [iota]int // [2]int + _ = unsafe.Sizeof([2]int{} == z) // assert types are equal + }) + three = iota // the sequence continues +) +var _ [three]int = [3]int{} // assert 'three' has correct value + +var ( + _ = iota /* ERROR "iota outside constant decl" */ + _ = unsafe.Sizeof(iota /* ERROR "iota outside constant decl" */ ) + _ = unsafe.Sizeof(func() { _ = iota /* ERROR "iota outside constant decl" */ }) + _ = unsafe.Sizeof(func() { var _ = iota /* ERROR "iota outside constant decl" */ }) + _ = unsafe.Sizeof(func() { type _ [iota /* ERROR "iota outside constant decl" */ ]byte }) + _ = unsafe.Sizeof(func() { func() int { return iota /* ERROR "iota outside constant decl" */ }() }) +) + +// constant arithmetic precision and rounding must lead to expected (integer) results +var _ = []int64{ + 0.0005 * 1e9, + 0.001 * 1e9, + 0.005 * 1e9, + 0.01 * 1e9, + 0.05 * 1e9, + 0.1 * 1e9, + 0.5 * 1e9, + 1 * 1e9, + 5 * 1e9, +} + +const _ = unsafe.Sizeof(func() { + const _ = 0 + _ = iota + + const ( + zero = iota + one + ) + assert(one == 1) + assert(iota == 0) +}) + +// issue #52438 +const i1 = iota +const i2 = iota +const i3 = iota + +func _() { + assert(i1 == 0) + assert(i2 == 0) + assert(i3 == 0) + + const i4 = iota + const i5 = iota + const i6 = iota + + assert(i4 == 0) + assert(i5 == 0) + assert(i6 == 0) +} + +// untyped constants must not get arbitrarily large +const prec = 512 // internal maximum precision for integers +const maxInt = (1<<(prec/2) - 1) * (1<<(prec/2) + 1) // == 1<<prec - 1 + +const _ = maxInt + /* ERROR "constant addition overflow" */ 1 +const _ = -maxInt - /* ERROR "constant subtraction overflow" */ 1 +const _ = maxInt ^ /* ERROR "constant bitwise XOR overflow" */ -1 +const _ = maxInt * /* ERROR "constant multiplication overflow" */ 2 +const _ = maxInt << /* ERROR "constant shift overflow" */ 2 +const _ = 1 << /* ERROR "constant shift overflow" */ prec + +const _ = ^ /* ERROR "constant bitwise complement overflow" */ maxInt diff --git a/src/internal/types/testdata/check/const1.go b/src/internal/types/testdata/check/const1.go new file mode 100644 index 0000000..c912801 --- /dev/null +++ b/src/internal/types/testdata/check/const1.go @@ -0,0 +1,334 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// constant conversions + +package const1 + +import "math" + +const( + mi = ^int(0) + mu = ^uint(0) + mp = ^uintptr(0) + + logSizeofInt = uint(mi>>8&1 + mi>>16&1 + mi>>32&1) + logSizeofUint = uint(mu>>8&1 + mu>>16&1 + mu>>32&1) + logSizeofUintptr = uint(mp>>8&1 + mp>>16&1 + mp>>32&1) +) + +const ( + minInt8 = -1<<(8<<iota - 1) + minInt16 + minInt32 + minInt64 + minInt = -1<<(8<<logSizeofInt - 1) +) + +const ( + maxInt8 = 1<<(8<<iota - 1) - 1 + maxInt16 + maxInt32 + maxInt64 + maxInt = 1<<(8<<logSizeofInt - 1) - 1 +) + +const ( + maxUint8 = 1<<(8<<iota) - 1 + maxUint16 + maxUint32 + maxUint64 + maxUint = 1<<(8<<logSizeofUint) - 1 + maxUintptr = 1<<(8<<logSizeofUintptr) - 1 +) + +const ( + smallestFloat32 = 1.0 / (1<<(127 - 1 + 23)) + // TODO(gri) The compiler limits integers to 512 bit and thus + // we cannot compute the value (1<<(1023 - 1 + 52)) + // without overflow. For now we match the compiler. + // See also issue #44057. + // smallestFloat64 = 1.0 / (1<<(1023 - 1 + 52)) + smallestFloat64 = math.SmallestNonzeroFloat64 +) + +const ( + _ = assert(smallestFloat32 > 0) + _ = assert(smallestFloat64 > 0) +) + +const ( + maxFloat32 = 1<<127 * (1<<24 - 1) / (1.0<<23) + // TODO(gri) The compiler limits integers to 512 bit and thus + // we cannot compute the value 1<<1023 + // without overflow. For now we match the compiler. + // See also issue #44057. + // maxFloat64 = 1<<1023 * (1<<53 - 1) / (1.0<<52) + maxFloat64 = math.MaxFloat64 +) + +const ( + _ int8 = minInt8 /* ERROR "overflows" */ - 1 + _ int8 = minInt8 + _ int8 = maxInt8 + _ int8 = maxInt8 /* ERROR "overflows" */ + 1 + _ int8 = smallestFloat64 /* ERROR "truncated" */ + + _ = int8(minInt8 /* ERROR "cannot convert" */ - 1) + _ = int8(minInt8) + _ = int8(maxInt8) + _ = int8(maxInt8 /* ERROR "cannot convert" */ + 1) + _ = int8(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ int16 = minInt16 /* ERROR "overflows" */ - 1 + _ int16 = minInt16 + _ int16 = maxInt16 + _ int16 = maxInt16 /* ERROR "overflows" */ + 1 + _ int16 = smallestFloat64 /* ERROR "truncated" */ + + _ = int16(minInt16 /* ERROR "cannot convert" */ - 1) + _ = int16(minInt16) + _ = int16(maxInt16) + _ = int16(maxInt16 /* ERROR "cannot convert" */ + 1) + _ = int16(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ int32 = minInt32 /* ERROR "overflows" */ - 1 + _ int32 = minInt32 + _ int32 = maxInt32 + _ int32 = maxInt32 /* ERROR "overflows" */ + 1 + _ int32 = smallestFloat64 /* ERROR "truncated" */ + + _ = int32(minInt32 /* ERROR "cannot convert" */ - 1) + _ = int32(minInt32) + _ = int32(maxInt32) + _ = int32(maxInt32 /* ERROR "cannot convert" */ + 1) + _ = int32(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ int64 = minInt64 /* ERROR "overflows" */ - 1 + _ int64 = minInt64 + _ int64 = maxInt64 + _ int64 = maxInt64 /* ERROR "overflows" */ + 1 + _ int64 = smallestFloat64 /* ERROR "truncated" */ + + _ = int64(minInt64 /* ERROR "cannot convert" */ - 1) + _ = int64(minInt64) + _ = int64(maxInt64) + _ = int64(maxInt64 /* ERROR "cannot convert" */ + 1) + _ = int64(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ int = minInt /* ERROR "overflows" */ - 1 + _ int = minInt + _ int = maxInt + _ int = maxInt /* ERROR "overflows" */ + 1 + _ int = smallestFloat64 /* ERROR "truncated" */ + + _ = int(minInt /* ERROR "cannot convert" */ - 1) + _ = int(minInt) + _ = int(maxInt) + _ = int(maxInt /* ERROR "cannot convert" */ + 1) + _ = int(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uint8 = 0 /* ERROR "overflows" */ - 1 + _ uint8 = 0 + _ uint8 = maxUint8 + _ uint8 = maxUint8 /* ERROR "overflows" */ + 1 + _ uint8 = smallestFloat64 /* ERROR "truncated" */ + + _ = uint8(0 /* ERROR "cannot convert" */ - 1) + _ = uint8(0) + _ = uint8(maxUint8) + _ = uint8(maxUint8 /* ERROR "cannot convert" */ + 1) + _ = uint8(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uint16 = 0 /* ERROR "overflows" */ - 1 + _ uint16 = 0 + _ uint16 = maxUint16 + _ uint16 = maxUint16 /* ERROR "overflows" */ + 1 + _ uint16 = smallestFloat64 /* ERROR "truncated" */ + + _ = uint16(0 /* ERROR "cannot convert" */ - 1) + _ = uint16(0) + _ = uint16(maxUint16) + _ = uint16(maxUint16 /* ERROR "cannot convert" */ + 1) + _ = uint16(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uint32 = 0 /* ERROR "overflows" */ - 1 + _ uint32 = 0 + _ uint32 = maxUint32 + _ uint32 = maxUint32 /* ERROR "overflows" */ + 1 + _ uint32 = smallestFloat64 /* ERROR "truncated" */ + + _ = uint32(0 /* ERROR "cannot convert" */ - 1) + _ = uint32(0) + _ = uint32(maxUint32) + _ = uint32(maxUint32 /* ERROR "cannot convert" */ + 1) + _ = uint32(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uint64 = 0 /* ERROR "overflows" */ - 1 + _ uint64 = 0 + _ uint64 = maxUint64 + _ uint64 = maxUint64 /* ERROR "overflows" */ + 1 + _ uint64 = smallestFloat64 /* ERROR "truncated" */ + + _ = uint64(0 /* ERROR "cannot convert" */ - 1) + _ = uint64(0) + _ = uint64(maxUint64) + _ = uint64(maxUint64 /* ERROR "cannot convert" */ + 1) + _ = uint64(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uint = 0 /* ERROR "overflows" */ - 1 + _ uint = 0 + _ uint = maxUint + _ uint = maxUint /* ERROR "overflows" */ + 1 + _ uint = smallestFloat64 /* ERROR "truncated" */ + + _ = uint(0 /* ERROR "cannot convert" */ - 1) + _ = uint(0) + _ = uint(maxUint) + _ = uint(maxUint /* ERROR "cannot convert" */ + 1) + _ = uint(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ uintptr = 0 /* ERROR "overflows" */ - 1 + _ uintptr = 0 + _ uintptr = maxUintptr + _ uintptr = maxUintptr /* ERROR "overflows" */ + 1 + _ uintptr = smallestFloat64 /* ERROR "truncated" */ + + _ = uintptr(0 /* ERROR "cannot convert" */ - 1) + _ = uintptr(0) + _ = uintptr(maxUintptr) + _ = uintptr(maxUintptr /* ERROR "cannot convert" */ + 1) + _ = uintptr(smallestFloat64 /* ERROR "cannot convert" */) +) + +const ( + _ float32 = minInt64 + _ float64 = minInt64 + _ complex64 = minInt64 + _ complex128 = minInt64 + + _ = float32(minInt64) + _ = float64(minInt64) + _ = complex64(minInt64) + _ = complex128(minInt64) +) + +const ( + _ float32 = maxUint64 + _ float64 = maxUint64 + _ complex64 = maxUint64 + _ complex128 = maxUint64 + + _ = float32(maxUint64) + _ = float64(maxUint64) + _ = complex64(maxUint64) + _ = complex128(maxUint64) +) + +// TODO(gri) find smaller deltas below + +const delta32 = maxFloat32/(1 << 23) + +const ( + _ float32 = - /* ERROR "overflow" */ (maxFloat32 + delta32) + _ float32 = -maxFloat32 + _ float32 = maxFloat32 + _ float32 = maxFloat32 /* ERROR "overflow" */ + delta32 + + _ = float32(- /* ERROR "cannot convert" */ (maxFloat32 + delta32)) + _ = float32(-maxFloat32) + _ = float32(maxFloat32) + _ = float32(maxFloat32 /* ERROR "cannot convert" */ + delta32) + + _ = assert(float32(smallestFloat32) == smallestFloat32) + _ = assert(float32(smallestFloat32/2) == 0) + _ = assert(float32(smallestFloat64) == 0) + _ = assert(float32(smallestFloat64/2) == 0) +) + +const delta64 = maxFloat64/(1 << 52) + +const ( + _ float64 = - /* ERROR "overflow" */ (maxFloat64 + delta64) + _ float64 = -maxFloat64 + _ float64 = maxFloat64 + _ float64 = maxFloat64 /* ERROR "overflow" */ + delta64 + + _ = float64(- /* ERROR "cannot convert" */ (maxFloat64 + delta64)) + _ = float64(-maxFloat64) + _ = float64(maxFloat64) + _ = float64(maxFloat64 /* ERROR "cannot convert" */ + delta64) + + _ = assert(float64(smallestFloat32) == smallestFloat32) + _ = assert(float64(smallestFloat32/2) == smallestFloat32/2) + _ = assert(float64(smallestFloat64) == smallestFloat64) + _ = assert(float64(smallestFloat64/2) == 0) +) + +const ( + _ complex64 = - /* ERROR "overflow" */ (maxFloat32 + delta32) + _ complex64 = -maxFloat32 + _ complex64 = maxFloat32 + _ complex64 = maxFloat32 /* ERROR "overflow" */ + delta32 + + _ = complex64(- /* ERROR "cannot convert" */ (maxFloat32 + delta32)) + _ = complex64(-maxFloat32) + _ = complex64(maxFloat32) + _ = complex64(maxFloat32 /* ERROR "cannot convert" */ + delta32) +) + +const ( + _ complex128 = - /* ERROR "overflow" */ (maxFloat64 + delta64) + _ complex128 = -maxFloat64 + _ complex128 = maxFloat64 + _ complex128 = maxFloat64 /* ERROR "overflow" */ + delta64 + + _ = complex128(- /* ERROR "cannot convert" */ (maxFloat64 + delta64)) + _ = complex128(-maxFloat64) + _ = complex128(maxFloat64) + _ = complex128(maxFloat64 /* ERROR "cannot convert" */ + delta64) +) + +// Initialization of typed constant and conversion are the same: +const ( + f32 = 1 + smallestFloat32 + x32 float32 = f32 + y32 = float32(f32) + _ = assert(x32 - y32 == 0) +) + +const ( + f64 = 1 + smallestFloat64 + x64 float64 = f64 + y64 = float64(f64) + _ = assert(x64 - y64 == 0) +) + +const ( + _ = int8(-1) << 7 + _ = int8 /* ERROR "overflows" */ (-1) << 8 + + _ = uint32(1) << 31 + _ = uint32 /* ERROR "overflows" */ (1) << 32 +) diff --git a/src/internal/types/testdata/check/constdecl.go b/src/internal/types/testdata/check/constdecl.go new file mode 100644 index 0000000..e7b871b --- /dev/null +++ b/src/internal/types/testdata/check/constdecl.go @@ -0,0 +1,160 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package constdecl + +import "math" +import "unsafe" + +var v int + +// Const decls must be initialized by constants. +const _ = v /* ERROR "not constant" */ +const _ = math /* ERROR "not constant" */ .Sin(0) +const _ = int /* ERROR "not an expression" */ + +func _() { + const _ = v /* ERROR "not constant" */ + const _ = math /* ERROR "not constant" */ .Sin(0) + const _ = int /* ERROR "not an expression" */ +} + +// Identifier and expression arity must match. +const _ /* ERROR "missing init expr for _" */ +const _ = 1, 2 /* ERROR "extra init expr 2" */ + +const _ /* ERROR "missing init expr for _" */ int +const _ int = 1, 2 /* ERROR "extra init expr 2" */ + +const ( + _ /* ERROR "missing init expr for _" */ + _ = 1, 2 /* ERROR "extra init expr 2" */ + + _ /* ERROR "missing init expr for _" */ int + _ int = 1, 2 /* ERROR "extra init expr 2" */ +) + +const ( + _ = 1 + _ + _, _ /* ERROR "missing init expr for _" */ + _ +) + +const ( + _, _ = 1, 2 + _, _ + _ /* ERROR "extra init expr at" */ + _, _ + _, _, _ /* ERROR "missing init expr for _" */ + _, _ +) + +func _() { + const _ /* ERROR "missing init expr for _" */ + const _ = 1, 2 /* ERROR "extra init expr 2" */ + + const _ /* ERROR "missing init expr for _" */ int + const _ int = 1, 2 /* ERROR "extra init expr 2" */ + + const ( + _ /* ERROR "missing init expr for _" */ + _ = 1, 2 /* ERROR "extra init expr 2" */ + + _ /* ERROR "missing init expr for _" */ int + _ int = 1, 2 /* ERROR "extra init expr 2" */ + ) + + const ( + _ = 1 + _ + _, _ /* ERROR "missing init expr for _" */ + _ + ) + + const ( + _, _ = 1, 2 + _, _ + _ /* ERROR "extra init expr at" */ + _, _ + _, _, _ /* ERROR "missing init expr for _" */ + _, _ + ) +} + +// Test case for constant with invalid initialization. +// Caused panic because the constant value was not set up (gri - 7/8/2014). +func _() { + const ( + x string = missing /* ERROR "undefined" */ + y = x + "" + ) +} + +// Test case for constants depending on function literals (see also #22992). +const A /* ERROR "initialization cycle" */ = unsafe.Sizeof(func() { _ = A }) + +func _() { + // The function literal below must not see a. + const a = unsafe.Sizeof(func() { _ = a /* ERROR "undefined" */ }) + const b = unsafe.Sizeof(func() { _ = a }) + + // The function literal below must not see x, y, or z. + const x, y, z = 0, 1, unsafe.Sizeof(func() { _ = x /* ERROR "undefined" */ + y /* ERROR "undefined" */ + z /* ERROR "undefined" */ }) +} + +// Test cases for errors in inherited constant initialization expressions. +// Errors related to inherited initialization expressions must appear at +// the constant identifier being declared, not at the original expression +// (issues #42991, #42992). +const ( + _ byte = 255 + iota + /* some gap */ + _ // ERROR "overflows" + /* some gap */ + /* some gap */ _ /* ERROR "overflows" */; _ /* ERROR "overflows" */ + /* some gap */ + _ = 255 + iota + _ = byte /* ERROR "overflows" */ (255) + iota + _ /* ERROR "overflows" */ +) + +// Test cases from issue. +const ( + ok = byte(iota + 253) + bad + barn + bard // ERROR "cannot convert" +) + +const ( + c = len([1 - iota]int{}) + d + e // ERROR "invalid array length" + f // ERROR "invalid array length" +) + +// Test that identifiers in implicit (omitted) RHS +// expressions of constant declarations are resolved +// in the correct context; see issues #49157, #53585. +const X = 2 + +func _() { + const ( + A = iota // 0 + iota = iota // 1 + B // 1 (iota is declared locally on prev. line) + C // 1 + ) + assert(A == 0 && B == 1 && C == 1) + + const ( + X = X + X + Y + Z = iota + ) + assert(X == 4 && Y == 8 && Z == 1) +} + +// TODO(gri) move extra tests from testdata/const0.src into here diff --git a/src/internal/types/testdata/check/conversions0.go b/src/internal/types/testdata/check/conversions0.go new file mode 100644 index 0000000..e1336c0 --- /dev/null +++ b/src/internal/types/testdata/check/conversions0.go @@ -0,0 +1,93 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// conversions + +package conversions + +import "unsafe" + +// argument count +var ( + _ = int() /* ERROR "missing argument" */ + _ = int(1, 2 /* ERROR "too many arguments" */ ) +) + +// numeric constant conversions are in const1.src. + +func string_conversions() { + const A = string(65) + assert(A == "A") + const E = string(-1) + assert(E == "\uFFFD") + assert(E == string(1234567890)) + + type myint int + assert(A == string(myint(65))) + + type mystring string + const _ mystring = mystring("foo") + + const _ = string(true /* ERROR "cannot convert" */ ) + const _ = string(1.2 /* ERROR "cannot convert" */ ) + const _ = string(nil /* ERROR "cannot convert" */ ) + + // issues 11357, 11353: argument must be of integer type + _ = string(0.0 /* ERROR "cannot convert" */ ) + _ = string(0i /* ERROR "cannot convert" */ ) + _ = string(1 /* ERROR "cannot convert" */ + 2i) +} + +func interface_conversions() { + type E interface{} + + type I1 interface{ + m1() + } + + type I2 interface{ + m1() + m2(x int) + } + + type I3 interface{ + m1() + m2() int + } + + var e E + var i1 I1 + var i2 I2 + var i3 I3 + + _ = E(0) + _ = E(nil) + _ = E(e) + _ = E(i1) + _ = E(i2) + + _ = I1(0 /* ERROR "cannot convert" */ ) + _ = I1(nil) + _ = I1(i1) + _ = I1(e /* ERROR "cannot convert" */ ) + _ = I1(i2) + + _ = I2(nil) + _ = I2(i1 /* ERROR "cannot convert" */ ) + _ = I2(i2) + _ = I2(i3 /* ERROR "cannot convert" */ ) + + _ = I3(nil) + _ = I3(i1 /* ERROR "cannot convert" */ ) + _ = I3(i2 /* ERROR "cannot convert" */ ) + _ = I3(i3) + + // TODO(gri) add more tests, improve error message +} + +func issue6326() { + type T unsafe.Pointer + var x T + _ = uintptr(x) // see issue 6326 +} diff --git a/src/internal/types/testdata/check/conversions1.go b/src/internal/types/testdata/check/conversions1.go new file mode 100644 index 0000000..65aabde --- /dev/null +++ b/src/internal/types/testdata/check/conversions1.go @@ -0,0 +1,313 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Test various valid and invalid struct assignments and conversions. +// Does not compile. + +package conversions2 + +type I interface { + m() +} + +// conversions between structs + +func _() { + type S struct{} + type T struct{} + var s S + var t T + var u struct{} + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u + s = S(s) + s = S(t) + s = S(u) + t = u + t = T(u) +} + +func _() { + type S struct{ x int } + type T struct { + x int "foo" + } + var s S + var t T + var u struct { + x int "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = S(s) + s = S(t) + s = S(u) + t = u // ERRORx `cannot use .* in assignment` + t = T(u) +} + +func _() { + type E struct{ x int } + type S struct{ x E } + type T struct { + x E "foo" + } + var s S + var t T + var u struct { + x E "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = S(s) + s = S(t) + s = S(u) + t = u // ERRORx `cannot use .* in assignment` + t = T(u) +} + +func _() { + type S struct { + x struct { + x int "foo" + } + } + type T struct { + x struct { + x int "bar" + } "foo" + } + var s S + var t T + var u struct { + x struct { + x int "bar" + } "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = S(s) + s = S(t) + s = S(u) + t = u // ERRORx `cannot use .* in assignment` + t = T(u) +} + +func _() { + type E1 struct { + x int "foo" + } + type E2 struct { + x int "bar" + } + type S struct{ x E1 } + type T struct { + x E2 "foo" + } + var s S + var t T + var u struct { + x E2 "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = S(s) + s = S(t /* ERROR "cannot convert" */ ) + s = S(u /* ERROR "cannot convert" */ ) + t = u // ERRORx `cannot use .* in assignment` + t = T(u) +} + +func _() { + type E struct{ x int } + type S struct { + f func(struct { + x int "foo" + }) + } + type T struct { + f func(struct { + x int "bar" + }) + } + var s S + var t T + var u struct{ f func(E) } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = S(s) + s = S(t) + s = S(u /* ERROR "cannot convert" */ ) + t = u // ERRORx `cannot use .* in assignment` + t = T(u /* ERROR "cannot convert" */ ) +} + +// conversions between pointers to structs + +func _() { + type S struct{} + type T struct{} + var s *S + var t *T + var u *struct{} + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u) +} + +func _() { + type S struct{ x int } + type T struct { + x int "foo" + } + var s *S + var t *T + var u *struct { + x int "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u) +} + +func _() { + type E struct{ x int } + type S struct{ x E } + type T struct { + x E "foo" + } + var s *S + var t *T + var u *struct { + x E "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u) +} + +func _() { + type S struct { + x struct { + x int "foo" + } + } + type T struct { + x struct { + x int "bar" + } "foo" + } + var s *S + var t *T + var u *struct { + x struct { + x int "bar" + } "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u) +} + +func _() { + type E1 struct { + x int "foo" + } + type E2 struct { + x int "bar" + } + type S struct{ x E1 } + type T struct { + x E2 "foo" + } + var s *S + var t *T + var u *struct { + x E2 "bar" + } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t /* ERROR "cannot convert" */ ) + s = (*S)(u /* ERROR "cannot convert" */ ) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u) +} + +func _() { + type E struct{ x int } + type S struct { + f func(struct { + x int "foo" + }) + } + type T struct { + f func(struct { + x int "bar" + }) + } + var s *S + var t *T + var u *struct{ f func(E) } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u /* ERROR "cannot convert" */ ) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u /* ERROR "cannot convert" */ ) +} + +func _() { + type E struct{ x int } + type S struct { + f func(*struct { + x int "foo" + }) + } + type T struct { + f func(*struct { + x int "bar" + }) + } + var s *S + var t *T + var u *struct{ f func(E) } + s = s + s = t // ERRORx `cannot use .* in assignment` + s = u // ERRORx `cannot use .* in assignment` + s = (*S)(s) + s = (*S)(t) + s = (*S)(u /* ERROR "cannot convert" */ ) + t = u // ERRORx `cannot use .* in assignment` + t = (*T)(u /* ERROR "cannot convert" */ ) +} diff --git a/src/internal/types/testdata/check/cycles0.go b/src/internal/types/testdata/check/cycles0.go new file mode 100644 index 0000000..8ad7877 --- /dev/null +++ b/src/internal/types/testdata/check/cycles0.go @@ -0,0 +1,175 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package cycles + +import "unsafe" + +type ( + T0 int + T1 /* ERROR "invalid recursive type: T1 refers to itself" */ T1 + T2 *T2 + + T3 /* ERROR "invalid recursive type" */ T4 + T4 T5 + T5 T3 + + T6 T7 + T7 *T8 + T8 T6 + + // arrays + A0 /* ERROR "invalid recursive type" */ [10]A0 + A1 [10]*A1 + + A2 /* ERROR "invalid recursive type" */ [10]A3 + A3 [10]A4 + A4 A2 + + A5 [10]A6 + A6 *A5 + + // slices + L0 []L0 + + // structs + S0 /* ERROR "invalid recursive type: S0 refers to itself" */ struct{ _ S0 } + S1 /* ERROR "invalid recursive type: S1 refers to itself" */ struct{ S1 } + S2 struct{ _ *S2 } + S3 struct{ *S3 } + + S4 /* ERROR "invalid recursive type" */ struct{ S5 } + S5 struct{ S6 } + S6 S4 + + // pointers + P0 *P0 + PP *struct{ PP.f /* ERROR "PP.f is not a type" */ } + + // functions + F0 func(F0) + F1 func() F1 + F2 func(F2) F2 + + // interfaces + I0 /* ERROR "invalid recursive type: I0 refers to itself" */ interface{ I0 } + + I1 /* ERROR "invalid recursive type" */ interface{ I2 } + I2 interface{ I3 } + I3 interface{ I1 } + + I4 interface{ f(I4) } + + // testcase for issue 5090 + I5 interface{ f(I6) } + I6 interface{ I5 } + + // maps + M0 map[M0 /* ERROR "invalid map key" */ ]M0 + + // channels + C0 chan C0 +) + +// test case for issue #34771 +type ( + AA /* ERROR "invalid recursive type" */ B + B C + C [10]D + D E + E AA +) + +func _() { + type ( + t1 /* ERROR "invalid recursive type: t1 refers to itself" */ t1 + t2 *t2 + + t3 t4 /* ERROR "undefined" */ + t4 t5 /* ERROR "undefined" */ + t5 t3 + + // arrays + a0 /* ERROR "invalid recursive type: a0 refers to itself" */ [10]a0 + a1 [10]*a1 + + // slices + l0 []l0 + + // structs + s0 /* ERROR "invalid recursive type: s0 refers to itself" */ struct{ _ s0 } + s1 /* ERROR "invalid recursive type: s1 refers to itself" */ struct{ s1 } + s2 struct{ _ *s2 } + s3 struct{ *s3 } + + // pointers + p0 *p0 + + // functions + f0 func(f0) + f1 func() f1 + f2 func(f2) f2 + + // interfaces + i0 /* ERROR "invalid recursive type: i0 refers to itself" */ interface{ i0 } + + // maps + m0 map[m0 /* ERROR "invalid map key" */ ]m0 + + // channels + c0 chan c0 + ) +} + +// test cases for issue 6667 + +type A [10]map[A /* ERROR "invalid map key" */ ]bool + +type S struct { + m map[S /* ERROR "invalid map key" */ ]bool +} + +// test cases for issue 7236 +// (cycle detection must not be dependent on starting point of resolution) + +type ( + P1 *T9 + T9 /* ERROR "invalid recursive type: T9 refers to itself" */ T9 + + T10 /* ERROR "invalid recursive type: T10 refers to itself" */ T10 + P2 *T10 +) + +func (T11) m() {} + +type T11 /* ERROR "invalid recursive type: T11 refers to itself" */ struct{ T11 } + +type T12 /* ERROR "invalid recursive type: T12 refers to itself" */ struct{ T12 } + +func (*T12) m() {} + +type ( + P3 *T13 + T13 /* ERROR "invalid recursive type" */ T13 +) + +// test cases for issue 18643 +// (type cycle detection when non-type expressions are involved) +type ( + T14 [len(T14 /* ERROR "invalid recursive type" */ {})]int + T15 [][len(T15 /* ERROR "invalid recursive type" */ {})]int + T16 map[[len(T16 /* ERROR "invalid recursive type" */ {1:2})]int]int + T17 map[int][len(T17 /* ERROR "invalid recursive type" */ {1:2})]int +) + +// Test case for types depending on function literals (see also #22992). +type T20 chan [unsafe.Sizeof(func(ch T20){ _ = <-ch })]byte +type T22 = chan [unsafe.Sizeof(func(ch T20){ _ = <-ch })]byte + +func _() { + type T0 func(T0) + type T1 /* ERROR "invalid recursive type" */ = func(T1) + type T2 chan [unsafe.Sizeof(func(ch T2){ _ = <-ch })]byte + type T3 /* ERROR "invalid recursive type" */ = chan [unsafe.Sizeof(func(ch T3){ _ = <-ch })]byte +} diff --git a/src/internal/types/testdata/check/cycles1.go b/src/internal/types/testdata/check/cycles1.go new file mode 100644 index 0000000..ae2b38e --- /dev/null +++ b/src/internal/types/testdata/check/cycles1.go @@ -0,0 +1,77 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + A interface { + a() interface { + ABC1 + } + } + B interface { + b() interface { + ABC2 + } + } + C interface { + c() interface { + ABC3 + } + } + + AB interface { + A + B + } + BC interface { + B + C + } + + ABC1 interface { + A + B + C + } + ABC2 interface { + AB + C + } + ABC3 interface { + A + BC + } +) + +var ( + x1 ABC1 + x2 ABC2 + x3 ABC3 +) + +func _() { + // all types have the same method set + x1 = x2 + x2 = x1 + + x1 = x3 + x3 = x1 + + x2 = x3 + x3 = x2 + + // all methods return the same type again + x1 = x1.a() + x1 = x1.b() + x1 = x1.c() + + x2 = x2.a() + x2 = x2.b() + x2 = x2.c() + + x3 = x3.a() + x3 = x3.b() + x3 = x3.c() +} diff --git a/src/internal/types/testdata/check/cycles2.go b/src/internal/types/testdata/check/cycles2.go new file mode 100644 index 0000000..a932d28 --- /dev/null +++ b/src/internal/types/testdata/check/cycles2.go @@ -0,0 +1,98 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +// Test case for issue 5090 + +type t interface { + f(u) +} + +type u interface { + t +} + +func _() { + var t t + var u u + + t.f(t) + t.f(u) + + u.f(t) + u.f(u) +} + + +// Test case for issues #6589, #33656. + +type A interface { + a() interface { + AB + } +} + +type B interface { + b() interface { + AB + } +} + +type AB interface { + a() interface { + A + B + } + b() interface { + A + B + } +} + +var x AB +var y interface { + A + B +} + +var _ = x == y + + +// Test case for issue 6638. + +type T interface { + m() [T(nil).m /* ERROR "undefined" */ ()[0]]int +} + +// Variations of this test case. + +type T1 /* ERROR "invalid recursive type" */ interface { + m() [x1.m()[0]]int +} + +var x1 T1 + +type T2 /* ERROR "invalid recursive type" */ interface { + m() [len(x2.m())]int +} + +var x2 T2 + +type T3 /* ERROR "invalid recursive type" */ interface { + m() [unsafe.Sizeof(x3.m)]int +} + +var x3 T3 + +type T4 /* ERROR "invalid recursive type" */ interface { + m() [unsafe.Sizeof(cast4(x4.m))]int // cast is invalid but we have a cycle, so all bets are off +} + +var x4 T4 +var _ = cast4(x4.m) + +type cast4 func() diff --git a/src/internal/types/testdata/check/cycles3.go b/src/internal/types/testdata/check/cycles3.go new file mode 100644 index 0000000..3ed999c --- /dev/null +++ b/src/internal/types/testdata/check/cycles3.go @@ -0,0 +1,60 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +var ( + _ A = A(nil).a().b().c().d().e().f() + _ A = A(nil).b().c().d().e().f() + _ A = A(nil).c().d().e().f() + _ A = A(nil).d().e().f() + _ A = A(nil).e().f() + _ A = A(nil).f() + _ A = A(nil) +) + +type ( + A interface { + a() B + B + } + + B interface { + b() C + C + } + + C interface { + c() D + D + } + + D interface { + d() E + E + } + + E interface { + e() F + F + } + + F interface { + f() A + } +) + +type ( + U /* ERROR "invalid recursive type" */ interface { + V + } + + V interface { + v() [unsafe.Sizeof(u)]int + } +) + +var u U diff --git a/src/internal/types/testdata/check/cycles4.go b/src/internal/types/testdata/check/cycles4.go new file mode 100644 index 0000000..e823001 --- /dev/null +++ b/src/internal/types/testdata/check/cycles4.go @@ -0,0 +1,121 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +// Check that all methods of T are collected before +// determining the result type of m (which embeds +// all methods of T). + +type T interface { + m() interface {T} + E +} + +var _ int = T.m(nil).m().e() + +type E interface { + e() int +} + +// Check that unresolved forward chains are followed +// (see also comment in resolver.go, checker.typeDecl). + +var _ int = C.m(nil).m().e() + +type A B + +type B interface { + m() interface{C} + E +} + +type C A + +// Check that interface type comparison for identity +// does not recur endlessly. + +type T1 interface { + m() interface{T1} +} + +type T2 interface { + m() interface{T2} +} + +func _(x T1, y T2) { + // Checking for assignability of interfaces must check + // if all methods of x are present in y, and that they + // have identical signatures. The signatures recur via + // the result type, which is an interface that embeds + // a single method m that refers to the very interface + // that contains it. This requires cycle detection in + // identity checks for interface types. + x = y +} + +type T3 interface { + m() interface{T4} +} + +type T4 interface { + m() interface{T3} +} + +func _(x T1, y T3) { + x = y +} + +// Check that interfaces are type-checked in order of +// (embedded interface) dependencies (was issue 7158). + +var x1 T5 = T7(nil) + +type T5 interface { + T6 +} + +type T6 interface { + m() T7 +} +type T7 interface { + T5 +} + +// Actual test case from issue 7158. + +func wrapNode() Node { + return wrapElement() +} + +func wrapElement() Element { + return nil +} + +type EventTarget interface { + AddEventListener(Event) +} + +type Node interface { + EventTarget +} + +type Element interface { + Node +} + +type Event interface { + Target() Element +} + +// Check that accessing an interface method too early doesn't lead +// to follow-on errors due to an incorrectly computed type set. + +type T8 interface { + m() [unsafe.Sizeof(T8.m /* ERROR "undefined" */ )]int +} + +var _ = T8.m // no error expected here diff --git a/src/internal/types/testdata/check/cycles5.go b/src/internal/types/testdata/check/cycles5.go new file mode 100644 index 0000000..a863aa8 --- /dev/null +++ b/src/internal/types/testdata/check/cycles5.go @@ -0,0 +1,200 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +// test case from issue #18395 + +type ( + A interface { B } + B interface { C } + C interface { D; F() A } + D interface { G() B } +) + +var _ = A(nil).G // G must be found + + +// test case from issue #21804 + +type sourceBridge interface { + listVersions() ([]Version, error) +} + +type Constraint interface { + copyTo(*ConstraintMsg) +} + +type ConstraintMsg struct{} + +func (m *ConstraintMsg) asUnpairedVersion() UnpairedVersion { + return nil +} + +type Version interface { + Constraint +} + +type UnpairedVersion interface { + Version +} + +var _ Constraint = UnpairedVersion(nil) + + +// derived test case from issue #21804 + +type ( + _ interface{ m(B1) } + A1 interface{ a(D1) } + B1 interface{ A1 } + C1 interface{ B1 } + D1 interface{ C1 } +) + +var _ A1 = C1(nil) + + +// derived test case from issue #22701 + +func F(x I4) interface{} { + return x.Method() +} + +type Unused interface { + RefersToI1(a I1) +} + +type I1 interface { + I2 + I3 +} + +type I2 interface { + RefersToI4() I4 +} + +type I3 interface { + Method() interface{} +} + +type I4 interface { + I1 +} + + +// check embedding of error interface + +type Error interface{ error } + +var err Error +var _ = err.Error() + + +// more esoteric cases + +type ( + T1 interface { T2 } + T2 /* ERROR "invalid recursive type" */ T2 +) + +type ( + T3 interface { T4 } + T4 /* ERROR "invalid recursive type" */ T5 + T5 = T6 + T6 = T7 + T7 = T4 +) + + +// arbitrary code may appear inside an interface + +const n = unsafe.Sizeof(func(){}) + +type I interface { + m([unsafe.Sizeof(func() { I.m(nil, [n]byte{}) })]byte) +} + + +// test cases for varias alias cycles + +type T10 /* ERROR "invalid recursive type" */ = *T10 // issue #25141 +type T11 /* ERROR "invalid recursive type" */ = interface{ f(T11) } // issue #23139 + +// issue #18640 +type ( + aa = bb + bb struct { + *aa + } +) + +type ( + a struct{ *b } + b = c + c struct{ *b /* ERROR "invalid use of type alias" */ } +) + +// issue #24939 +type ( + _ interface { + M(P) + } + + M interface { + F() P // ERROR "invalid use of type alias" + } + + P = interface { + I() M + } +) + +// issue #8699 +type T12 /* ERROR "invalid recursive type" */ [len(a12)]int +var a12 = makeArray() +func makeArray() (res T12) { return } + +// issue #20770 +var r /* ERROR "invalid cycle in declaration of r" */ = newReader() +func newReader() r + +// variations of the theme of #8699 and #20770 +var arr /* ERROR "cycle" */ = f() +func f() [len(arr)]int + +// issue #25790 +func ff(ff /* ERROR "not a type" */ ) +func gg((gg /* ERROR "not a type" */ )) + +type T13 /* ERROR "invalid recursive type T13" */ [len(b13)]int +var b13 T13 + +func g1() [unsafe.Sizeof(g1)]int +func g2() [unsafe.Sizeof(x2)]int +var x2 = g2 + +// verify that we get the correct sizes for the functions above +// (note: assert is statically evaluated in go/types test mode) +func init() { + assert(unsafe.Sizeof(g1) == 8) + assert(unsafe.Sizeof(x2) == 8) +} + +func h() [h /* ERROR "no value" */ ()[0]]int { panic(0) } + +var c14 /* ERROR "cycle" */ T14 +type T14 [uintptr(unsafe.Sizeof(&c14))]byte + +// issue #34333 +type T15 /* ERROR "invalid recursive type T15" */ struct { + f func() T16 + b T16 +} + +type T16 struct { + T15 +}
\ No newline at end of file diff --git a/src/internal/types/testdata/check/decls0.go b/src/internal/types/testdata/check/decls0.go new file mode 100644 index 0000000..0b99faa --- /dev/null +++ b/src/internal/types/testdata/check/decls0.go @@ -0,0 +1,210 @@ +// -lang=go1.17 + +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// type declarations + +package p // don't permit non-interface elements in interfaces + +import "unsafe" + +const pi = 3.1415 + +type ( + N undefined /* ERROR "undefined" */ + B bool + I int32 + A [10]P + T struct { + x, y P + } + P *T + R (*R) + F func(A) I + Y interface { + f(A) I + } + S [](((P))) + M map[I]F + C chan<- I + + // blank types must be typechecked + _ pi /* ERROR "not a type" */ + _ struct{} + _ struct{ pi /* ERROR "not a type" */ } +) + + +// declarations of init +const _, init /* ERROR "cannot declare init" */ , _ = 0, 1, 2 +type init /* ERROR "cannot declare init" */ struct{} +var _, init /* ERROR "cannot declare init" */ int + +func init() {} +func init /* ERROR "missing function body" */ () + +func _() { const init = 0 } +func _() { type init int } +func _() { var init int; _ = init } + +// invalid array types +type ( + iA0 [... /* ERROR "invalid use of [...] array" */ ]byte + // The error message below could be better. At the moment + // we believe an integer that is too large is not an integer. + // But at least we get an error. + iA1 [1 /* ERROR "invalid array length" */ <<100]int + iA2 [- /* ERROR "invalid array length" */ 1]complex128 + iA3 ["foo" /* ERROR "must be integer" */ ]string + iA4 [float64 /* ERROR "must be integer" */ (0)]int +) + + +type ( + p1 pi.foo /* ERROR "pi.foo is not a type" */ + p2 unsafe.Pointer +) + + +type ( + Pi pi /* ERROR "not a type" */ + + a /* ERROR "invalid recursive type" */ a + a /* ERROR "redeclared" */ int + + b /* ERROR "invalid recursive type" */ c + c d + d e + e b + + t *t + + U V + V *W + W U + + P1 *S2 + P2 P1 + + S0 struct { + } + S1 struct { + a, b, c int + u, v, a /* ERROR "redeclared" */ float32 + } + S2 struct { + S0 // embedded field + S0 /* ERROR "redeclared" */ int + } + S3 struct { + x S2 + } + S4/* ERROR "invalid recursive type" */ struct { + S4 + } + S5 /* ERROR "invalid recursive type" */ struct { + S6 + } + S6 struct { + field S7 + } + S7 struct { + S5 + } + + L1 []L1 + L2 []int + + A1 [10.0]int + A2 /* ERROR "invalid recursive type" */ [10]A2 + A3 /* ERROR "invalid recursive type" */ [10]struct { + x A4 + } + A4 [10]A3 + + F1 func() + F2 func(x, y, z float32) + F3 func(x, y, x /* ERROR "redeclared" */ float32) + F4 func() (x, y, x /* ERROR "redeclared" */ float32) + F5 func(x int) (x /* ERROR "redeclared" */ float32) + F6 func(x ...int) + + I1 interface{} + I2 interface { + m1() + } + I3 interface { + m1() + m1 /* ERROR "duplicate method" */ () + } + I4 interface { + m1(x, y, x /* ERROR "redeclared" */ float32) + m2() (x, y, x /* ERROR "redeclared" */ float32) + m3(x int) (x /* ERROR "redeclared" */ float32) + } + I5 interface { + m1(I5) + } + I6 interface { + S0 /* ERROR "non-interface type S0" */ + } + I7 interface { + I1 + I1 + } + I8 /* ERROR "invalid recursive type" */ interface { + I8 + } + I9 /* ERROR "invalid recursive type" */ interface { + I10 + } + I10 interface { + I11 + } + I11 interface { + I9 + } + + C1 chan int + C2 <-chan int + C3 chan<- C3 + C4 chan C5 + C5 chan C6 + C6 chan C4 + + M1 map[Last]string + M2 map[string]M2 + + Last int +) + +// cycles in function/method declarations +// (test cases for issues #5217, #25790 and variants) +func f1(x f1 /* ERROR "not a type" */ ) {} +func f2(x *f2 /* ERROR "not a type" */ ) {} +func f3() (x f3 /* ERROR "not a type" */ ) { return } +func f4() (x *f4 /* ERROR "not a type" */ ) { return } +// TODO(#43215) this should be detected as a cycle error +func f5([unsafe.Sizeof(f5)]int) {} + +func (S0) m1 (x S0.m1 /* ERROR "S0.m1 is not a type" */ ) {} +func (S0) m2 (x *S0.m2 /* ERROR "S0.m2 is not a type" */ ) {} +func (S0) m3 () (x S0.m3 /* ERROR "S0.m3 is not a type" */ ) { return } +func (S0) m4 () (x *S0.m4 /* ERROR "S0.m4 is not a type" */ ) { return } + +// interfaces may not have any blank methods +type BlankI interface { + _ /* ERROR "methods must have a unique non-blank name" */ () + _ /* ERROR "methods must have a unique non-blank name" */ (float32) int + m() +} + +// non-interface types may have multiple blank methods +type BlankT struct{} + +func (BlankT) _() {} +func (BlankT) _(int) {} +func (BlankT) _() int { return 0 } +func (BlankT) _(int) int { return 0} diff --git a/src/internal/types/testdata/check/decls1.go b/src/internal/types/testdata/check/decls1.go new file mode 100644 index 0000000..06f3b2e --- /dev/null +++ b/src/internal/types/testdata/check/decls1.go @@ -0,0 +1,146 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// variable declarations + +package decls1 + +import ( + "math" +) + +// Global variables without initialization +var ( + a, b bool + c byte + d uint8 + r rune + i int + j, k, l int + x, y float32 + xx, yy float64 + u, v complex64 + uu, vv complex128 + s, t string + array []byte + iface interface{} + + blank _ /* ERROR "cannot use _" */ +) + +// Global variables with initialization +var ( + s1 = i + j + s2 = i /* ERROR "mismatched types" */ + x + s3 = c + d + s4 = s + t + s5 = s /* ERROR "invalid operation" */ / t + s6 = array[t1] + s7 = array[x /* ERROR "integer" */] + s8 = &a + s10 = &42 /* ERROR "cannot take address" */ + s11 = &v + s12 = -(u + *t11) / *&v + s13 = a /* ERROR "shifted operand" */ << d + s14 = i << j + s18 = math.Pi * 10.0 + s19 = s1 /* ERROR "cannot call" */ () + s20 = f0 /* ERROR "no value" */ () + s21 = f6(1, s1, i) + s22 = f6(1, s1, uu /* ERRORx `cannot use .* in argument` */ ) + + t1 int = i + j + t2 int = i /* ERROR "mismatched types" */ + x + t3 int = c /* ERRORx `cannot use .* variable declaration` */ + d + t4 string = s + t + t5 string = s /* ERROR "invalid operation" */ / t + t6 byte = array[t1] + t7 byte = array[x /* ERROR "must be integer" */] + t8 *int = & /* ERRORx `cannot use .* variable declaration` */ a + t10 *int = &42 /* ERROR "cannot take address" */ + t11 *complex64 = &v + t12 complex64 = -(u + *t11) / *&v + t13 int = a /* ERROR "shifted operand" */ << d + t14 int = i << j + t15 math /* ERROR "not in selector" */ + t16 math.xxx /* ERROR "undefined" */ + t17 math /* ERROR "not a type" */ .Pi + t18 float64 = math.Pi * 10.0 + t19 int = t1 /* ERROR "cannot call" */ () + t20 int = f0 /* ERROR "no value" */ () + t21 int = a /* ERRORx `cannot use .* variable declaration` */ +) + +// Various more complex expressions +var ( + u1 = x /* ERROR "not an interface" */ .(int) + u2 = iface.([]int) + u3 = iface.(a /* ERROR "not a type" */ ) + u4, ok = iface.(int) + u5, ok2, ok3 = iface /* ERROR "assignment mismatch" */ .(int) +) + +// Constant expression initializations +var ( + v1 = 1 /* ERROR "mismatched types untyped int and untyped string" */ + "foo" + v2 = c + 255 + v3 = c + 256 /* ERROR "overflows" */ + v4 = r + 2147483647 + v5 = r + 2147483648 /* ERROR "overflows" */ + v6 = 42 + v7 = v6 + 9223372036854775807 + v8 = v6 + 9223372036854775808 /* ERROR "overflows" */ + v9 = i + 1 << 10 + v10 byte = 1024 /* ERROR "overflows" */ + v11 = xx/yy*yy - xx + v12 = true && false + v13 = nil /* ERROR "use of untyped nil" */ + v14 string = 257 // ERRORx `cannot use 257 .* as string value in variable declaration$` + v15 int8 = 257 // ERRORx `cannot use 257 .* as int8 value in variable declaration .*overflows` +) + +// Multiple assignment expressions +var ( + m1a, m1b = 1, 2 + m2a, m2b, m2c /* ERROR "missing init expr for m2c" */ = 1, 2 + m3a, m3b = 1, 2, 3 /* ERROR "extra init expr 3" */ +) + +func _() { + var ( + m1a, m1b = 1, 2 + m2a, m2b, m2c /* ERROR "missing init expr for m2c" */ = 1, 2 + m3a, m3b = 1, 2, 3 /* ERROR "extra init expr 3" */ + ) + + _, _ = m1a, m1b + _, _, _ = m2a, m2b, m2c + _, _ = m3a, m3b +} + +// Declaration of parameters and results +func f0() {} +func f1(a /* ERROR "not a type" */) {} +func f2(a, b, c d /* ERROR "not a type" */) {} + +func f3() int { return 0 } +func f4() a /* ERROR "not a type" */ { return 0 } +func f5() (a, b, c d /* ERROR "not a type" */) { return } + +func f6(a, b, c int) complex128 { return 0 } + +// Declaration of receivers +type T struct{} + +func (T) m0() {} +func (*T) m1() {} +func (x T) m2() {} +func (x *T) m3() {} + +// Initialization functions +func init() {} +func init /* ERROR "no arguments and no return values" */ (int) {} +func init /* ERROR "no arguments and no return values" */ () int { return 0 } +func init /* ERROR "no arguments and no return values" */ (int) int { return 0 } +func (T) init(int) int { return 0 } diff --git a/src/internal/types/testdata/check/decls2/decls2a.go b/src/internal/types/testdata/check/decls2/decls2a.go new file mode 100644 index 0000000..c2fb421 --- /dev/null +++ b/src/internal/types/testdata/check/decls2/decls2a.go @@ -0,0 +1,111 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// method declarations + +package decls2 + +import "time" +import "unsafe" + +// T1 declared before its methods. +type T1 struct{ + f int +} + +func (T1) m() {} +func (T1) m /* ERROR "already declared" */ () {} +func (x *T1) f /* ERROR "field and method" */ () {} + +// Conflict between embedded field and method name, +// with the embedded field being a basic type. +type T1b struct { + int +} + +func (T1b) int /* ERROR "field and method" */ () {} + +type T1c struct { + time.Time +} + +func (T1c) Time /* ERROR "field and method" */ () int { return 0 } + +// Disabled for now: LookupFieldOrMethod will find Pointer even though +// it's double-declared (it would cost extra in the common case to verify +// this). But the MethodSet computation will not find it due to the name +// collision caused by the double-declaration, leading to an internal +// inconsistency while we are verifying one computation against the other. +// var _ = T1c{}.Pointer + +// T2's method declared before the type. +func (*T2) f /* ERROR "field and method" */ () {} + +type T2 struct { + f int +} + +// Methods declared without a declared type. +func (undefined /* ERROR "undefined" */) m() {} +func (x *undefined /* ERROR "undefined" */) m() {} + +func (pi /* ERROR "not a type" */) m1() {} +func (x pi /* ERROR "not a type" */) m2() {} +func (x *pi /* ERROR "not a type" */ ) m3() {} + +// Blank types. +type _ struct { m int } +type _ struct { m int } + +func (_ /* ERROR "cannot use _" */) m() {} +func m(_ /* ERROR "cannot use _" */) {} + +// Methods with receiver base type declared in another file. +func (T3) m1() {} +func (*T3) m2() {} +func (x T3) m3() {} +func (x *T3) f /* ERROR "field and method" */ () {} + +// Methods of non-struct type. +type T4 func() + +func (self T4) m() func() { return self } + +// Methods associated with an interface. +type T5 interface { + m() int +} + +func (T5 /* ERROR "invalid receiver" */ ) m1() {} +func (T5 /* ERROR "invalid receiver" */ ) m2() {} + +// Methods associated with a named pointer type. +type ptr *int +func (ptr /* ERROR "invalid receiver" */ ) _() {} +func (* /* ERROR "invalid receiver" */ ptr) _() {} + +// Methods with zero or multiple receivers. +func ( /* ERROR "method has no receiver" */ ) _() {} +func (T3, * /* ERROR "method has multiple receivers" */ T3) _() {} +func (T3, T3, T3 /* ERROR "method has multiple receivers" */ ) _() {} +func (a, b /* ERROR "method has multiple receivers" */ T3) _() {} +func (a, b, c /* ERROR "method has multiple receivers" */ T3) _() {} + +// Methods associated with non-local or unnamed types. +func (int /* ERROR "cannot define new methods on non-local type int" */ ) m() {} +func ([ /* ERROR "invalid receiver" */ ]int) m() {} +func (time /* ERROR "cannot define new methods on non-local type time.Time" */ .Time) m() {} +func (* /* ERROR "cannot define new methods on non-local type time.Time" */ time.Time) m() {} +func (x /* ERROR "invalid receiver" */ interface{}) m() {} + +// Unsafe.Pointer is treated like a pointer when used as receiver type. +type UP unsafe.Pointer +func (UP /* ERROR "invalid" */ ) m1() {} +func (* /* ERROR "invalid" */ UP) m2() {} + +// Double declarations across package files +const c_double = 0 +type t_double int +var v_double int +func f_double() {} diff --git a/src/internal/types/testdata/check/decls2/decls2b.go b/src/internal/types/testdata/check/decls2/decls2b.go new file mode 100644 index 0000000..dd6cd44 --- /dev/null +++ b/src/internal/types/testdata/check/decls2/decls2b.go @@ -0,0 +1,75 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// method declarations + +package decls2 + +import "io" + +const pi = 3.1415 + +func (T1) m /* ERROR "already declared" */ () {} +func (T2) m(io.Writer) {} + +type T3 struct { + f *T3 +} + +type T6 struct { + x int +} + +func (t *T6) m1() int { + return t.x +} + +func f() { + var t *T6 + t.m1() +} + +// Double declarations across package files +const c_double /* ERROR "redeclared" */ = 0 +type t_double /* ERROR "redeclared" */ int +var v_double /* ERROR "redeclared" */ int +func f_double /* ERROR "redeclared" */ () {} + +// Blank methods need to be type-checked. +// Verify by checking that errors are reported. +func (T /* ERROR "undefined" */ ) _() {} +func (T1) _(undefined /* ERROR "undefined" */ ) {} +func (T1) _() int { return "foo" /* ERRORx "cannot use .* in return statement" */ } + +// Methods with undefined receiver type can still be checked. +// Verify by checking that errors are reported. +func (Foo /* ERROR "undefined" */ ) m() {} +func (Foo /* ERROR "undefined" */ ) m(undefined /* ERROR "undefined" */ ) {} +func (Foo /* ERRORx `undefined` */ ) m() int { return "foo" /* ERRORx "cannot use .* in return statement" */ } + +func (Foo /* ERROR "undefined" */ ) _() {} +func (Foo /* ERROR "undefined" */ ) _(undefined /* ERROR "undefined" */ ) {} +func (Foo /* ERROR "undefined" */ ) _() int { return "foo" /* ERRORx "cannot use .* in return statement" */ } + +// Receiver declarations are regular parameter lists; +// receiver types may use parentheses, and the list +// may have a trailing comma. +type T7 struct {} + +func (T7) m1() {} +func ((T7)) m2() {} +func ((*T7)) m3() {} +func (x *(T7),) m4() {} +func (x (*(T7)),) m5() {} +func (x ((*((T7)))),) m6() {} + +// Check that methods with parenthesized receiver are actually present (issue #23130). +var ( + _ = T7.m1 + _ = T7.m2 + _ = (*T7).m3 + _ = (*T7).m4 + _ = (*T7).m5 + _ = (*T7).m6 +) diff --git a/src/internal/types/testdata/check/decls3.go b/src/internal/types/testdata/check/decls3.go new file mode 100644 index 0000000..3d00a58 --- /dev/null +++ b/src/internal/types/testdata/check/decls3.go @@ -0,0 +1,309 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// embedded types + +package decls3 + +import "unsafe" +import "fmt" + +// fields with the same name at the same level cancel each other out + +func _() { + type ( + T1 struct { X int } + T2 struct { X int } + T3 struct { T1; T2 } // X is embedded twice at the same level via T1->X, T2->X + ) + + var t T3 + _ = t.X /* ERROR "ambiguous selector t.X" */ +} + +func _() { + type ( + T1 struct { X int } + T2 struct { T1 } + T3 struct { T1 } + T4 struct { T2; T3 } // X is embedded twice at the same level via T2->T1->X, T3->T1->X + ) + + var t T4 + _ = t.X /* ERROR "ambiguous selector t.X" */ +} + +func issue4355() { + type ( + T1 struct {X int} + T2 struct {T1} + T3 struct {T2} + T4 struct {T2} + T5 struct {T3; T4} // X is embedded twice at the same level via T3->T2->T1->X, T4->T2->T1->X + ) + + var t T5 + _ = t.X /* ERROR "ambiguous selector t.X" */ +} + +func _() { + type State int + type A struct{ State } + type B struct{ fmt.State } + type T struct{ A; B } + + var t T + _ = t.State /* ERROR "ambiguous selector t.State" */ +} + +// Embedded fields can be predeclared types. + +func _() { + type T0 struct{ + int + float32 + f int + } + var x T0 + _ = x.int + _ = x.float32 + _ = x.f + + type T1 struct{ + T0 + } + var y T1 + _ = y.int + _ = y.float32 + _ = y.f +} + +// Restrictions on embedded field types. + +func _() { + type I1 interface{} + type I2 interface{} + type P1 *int + type P2 *int + type UP unsafe.Pointer + + type T1 struct { + I1 + * /* ERROR "cannot be a pointer to an interface" */ I2 + * /* ERROR "cannot be a pointer to an interface" */ error + P1 /* ERROR "cannot be a pointer" */ + * /* ERROR "cannot be a pointer" */ P2 + } + + // unsafe.Pointers are treated like regular pointers when embedded + type T2 struct { + unsafe /* ERROR "cannot be unsafe.Pointer" */ .Pointer + */* ERROR "cannot be unsafe.Pointer" */ unsafe.Pointer /* ERROR "Pointer redeclared" */ + UP /* ERROR "cannot be unsafe.Pointer" */ + * /* ERROR "cannot be unsafe.Pointer" */ UP /* ERROR "UP redeclared" */ + } +} + +// Named types that are pointers. + +type S struct{ x int } +func (*S) m() {} +type P *S + +func _() { + var s *S + _ = s.x + _ = s.m + + var p P + _ = p.x + _ = p.m /* ERROR "no field or method" */ + _ = P.m /* ERROR "no field or method" */ +} + +// Borrowed from the FieldByName test cases in reflect/all_test.go. + +type D1 struct { + d int +} +type D2 struct { + d int +} + +type S0 struct { + A, B, C int + D1 + D2 +} + +type S1 struct { + B int + S0 +} + +type S2 struct { + A int + *S1 +} + +type S1x struct { + S1 +} + +type S1y struct { + S1 +} + +type S3 struct { + S1x + S2 + D, E int + *S1y +} + +type S4 struct { + *S4 + A int +} + +// The X in S6 and S7 annihilate, but they also block the X in S8.S9. +type S5 struct { + S6 + S7 + S8 +} + +type S6 struct { + X int +} + +type S7 S6 + +type S8 struct { + S9 +} + +type S9 struct { + X int + Y int +} + +// The X in S11.S6 and S12.S6 annihilate, but they also block the X in S13.S8.S9. +type S10 struct { + S11 + S12 + S13 +} + +type S11 struct { + S6 +} + +type S12 struct { + S6 +} + +type S13 struct { + S8 +} + +func _() { + _ = struct{}{}.Foo /* ERROR "no field or method" */ + _ = S0{}.A + _ = S0{}.D /* ERROR "no field or method" */ + _ = S1{}.A + _ = S1{}.B + _ = S1{}.S0 + _ = S1{}.C + _ = S2{}.A + _ = S2{}.S1 + _ = S2{}.B + _ = S2{}.C + _ = S2{}.D /* ERROR "no field or method" */ + _ = S3{}.S1 /* ERROR "ambiguous selector S3{}.S1" */ + _ = S3{}.A + _ = S3{}.B /* ERROR "ambiguous selector S3{}.B" */ + _ = S3{}.D + _ = S3{}.E + _ = S4{}.A + _ = S4{}.B /* ERROR "no field or method" */ + _ = S5{}.X /* ERROR "ambiguous selector S5{}.X" */ + _ = S5{}.Y + _ = S10{}.X /* ERROR "ambiguous selector S10{}.X" */ + _ = S10{}.Y +} + +// Borrowed from the FieldByName benchmark in reflect/all_test.go. + +type R0 struct { + *R1 + *R2 + *R3 + *R4 +} + +type R1 struct { + *R5 + *R6 + *R7 + *R8 +} + +type R2 R1 +type R3 R1 +type R4 R1 + +type R5 struct { + *R9 + *R10 + *R11 + *R12 +} + +type R6 R5 +type R7 R5 +type R8 R5 + +type R9 struct { + *R13 + *R14 + *R15 + *R16 +} + +type R10 R9 +type R11 R9 +type R12 R9 + +type R13 struct { + *R17 + *R18 + *R19 + *R20 +} + +type R14 R13 +type R15 R13 +type R16 R13 + +type R17 struct { + *R21 + *R22 + *R23 + *R24 +} + +type R18 R17 +type R19 R17 +type R20 R17 + +type R21 struct { + X int +} + +type R22 R21 +type R23 R21 +type R24 R21 + +var _ = R0{}.X /* ERROR "ambiguous selector R0{}.X" */
\ No newline at end of file diff --git a/src/internal/types/testdata/check/decls4.go b/src/internal/types/testdata/check/decls4.go new file mode 100644 index 0000000..c47a68d --- /dev/null +++ b/src/internal/types/testdata/check/decls4.go @@ -0,0 +1,199 @@ +// Copyright 2016 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// type aliases + +package decls4 + +type ( + T0 [10]int + T1 []byte + T2 struct { + x int + } + T3 interface{ + m() T2 + } + T4 func(int, T0) chan T2 +) + +type ( + Ai = int + A0 = T0 + A1 = T1 + A2 = T2 + A3 = T3 + A4 = T4 + + A10 = [10]int + A11 = []byte + A12 = struct { + x int + } + A13 = interface{ + m() A2 + } + A14 = func(int, A0) chan A2 +) + +// check assignment compatibility due to equality of types +var ( + xi_ int + ai Ai = xi_ + + x0 T0 + a0 A0 = x0 + + x1 T1 + a1 A1 = x1 + + x2 T2 + a2 A2 = x2 + + x3 T3 + a3 A3 = x3 + + x4 T4 + a4 A4 = x4 +) + +// alias receiver types +func (Ai /* ERROR "cannot define new methods on non-local type int" */) m1() {} +func (T0) m1() {} +func (A0) m1 /* ERROR "already declared" */ () {} +func (A0) m2 () {} +func (A3 /* ERROR "invalid receiver" */ ) m1 () {} +func (A10 /* ERROR "invalid receiver" */ ) m1() {} + +// x0 has methods m1, m2 declared via receiver type names T0 and A0 +var _ interface{ m1(); m2() } = x0 + +// alias receiver types (test case for issue #23042) +type T struct{} + +var ( + _ = T.m + _ = T{}.m + _ interface{m()} = T{} +) + +var ( + _ = T.n + _ = T{}.n + _ interface{m(); n()} = T{} +) + +type U = T +func (U) m() {} + +// alias receiver types (long type declaration chains) +type ( + V0 = V1 + V1 = (V2) + V2 = ((V3)) + V3 = T +) + +func (V0) m /* ERROR "already declared" */ () {} +func (V1) n() {} + +// alias receiver types (invalid due to cycles) +type ( + W0 /* ERROR "invalid recursive type" */ = W1 + W1 = (W2) + W2 = ((W0)) +) + +func (W0) m() {} // no error expected (due to above cycle error) +func (W1) n() {} + +// alias receiver types (invalid due to builtin underlying type) +type ( + B0 = B1 + B1 = B2 + B2 = int +) + +func (B0 /* ERROR "cannot define new methods on non-local type int" */ ) m() {} +func (B1 /* ERROR "cannot define new methods on non-local type int" */ ) n() {} + +// cycles +type ( + C2 /* ERROR "invalid recursive type" */ = C2 + C3 /* ERROR "invalid recursive type" */ = C4 + C4 = C3 + C5 struct { + f *C6 + } + C6 = C5 + C7 /* ERROR "invalid recursive type" */ struct { + f C8 + } + C8 = C7 +) + +// embedded fields +var ( + s0 struct { T0 } + s1 struct { A0 } = s0 /* ERROR "cannot use" */ // embedded field names are different +) + +// embedding and lookup of fields and methods +func _(s struct{A0}) { s.A0 = x0 } + +type eX struct{xf int} + +func (eX) xm() + +type eY = struct{eX} // field/method set of eY includes xf, xm + +type eZ = *struct{eX} // field/method set of eZ includes xf, xm + +type eA struct { + eX // eX contributes xf, xm to eA +} + +type eA2 struct { + *eX // *eX contributes xf, xm to eA +} + +type eB struct { + eY // eY contributes xf, xm to eB +} + +type eB2 struct { + *eY // *eY contributes xf, xm to eB +} + +type eC struct { + eZ // eZ contributes xf, xm to eC +} + +var ( + _ = eA{}.xf + _ = eA{}.xm + _ = eA2{}.xf + _ = eA2{}.xm + _ = eB{}.xf + _ = eB{}.xm + _ = eB2{}.xf + _ = eB2{}.xm + _ = eC{}.xf + _ = eC{}.xm +) + +// ambiguous selectors due to embedding via type aliases +type eD struct { + eY + eZ +} + +var ( + _ = eD{}.xf /* ERROR "ambiguous selector eD{}.xf" */ + _ = eD{}.xm /* ERROR "ambiguous selector eD{}.xm" */ +) + +var ( + _ interface{ xm() } = eD /* ERROR "ambiguous selector eD.xm" */ {} +)
\ No newline at end of file diff --git a/src/internal/types/testdata/check/decls5.go b/src/internal/types/testdata/check/decls5.go new file mode 100644 index 0000000..88d3194 --- /dev/null +++ b/src/internal/types/testdata/check/decls5.go @@ -0,0 +1,10 @@ +// Copyright 2017 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +// declarations of main +const _, main /* ERROR "cannot declare main" */ , _ = 0, 1, 2 +type main /* ERROR "cannot declare main" */ struct{} +var _, main /* ERROR "cannot declare main" */ int diff --git a/src/internal/types/testdata/check/errors.go b/src/internal/types/testdata/check/errors.go new file mode 100644 index 0000000..10b6a22 --- /dev/null +++ b/src/internal/types/testdata/check/errors.go @@ -0,0 +1,66 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package errors + +// Testing precise operand formatting in error messages +// (matching messages are regular expressions, hence the \'s). +func f(x int, m map[string]int) { + // no values + _ = f /* ERROR "f(0, m) (no value) used as value" */ (0, m) + + // built-ins + _ = println // ERROR "println (built-in) must be called" + + // types + _ = complex128 // ERROR "complex128 (type) is not an expression" + + // constants + const c1 = 991 + const c2 float32 = 0.5 + const c3 = "foo" + 0 // ERROR "0 (untyped int constant) is not used" + 0.5 // ERROR "0.5 (untyped float constant) is not used" + "foo" // ERROR `"foo" (untyped string constant) is not used` + c1 // ERROR "c1 (untyped int constant 991) is not used" + c2 // ERROR "c2 (constant 0.5 of type float32) is not used" + c1 /* ERROR "c1 + c2 (constant 991.5 of type float32) is not used" */ + c2 + c3 // ERROR `c3 (untyped string constant "foo") is not used` + + // variables + x // ERROR "x (variable of type int) is not used" + + // values + nil // ERROR "nil is not used" + ( /* ERROR "(*int)(nil) (value of type *int) is not used" */ *int)(nil) + x /* ERROR "x != x (untyped bool value) is not used" */ != x + x /* ERROR "x + x (value of type int) is not used" */ + x + + // value, ok's + const s = "foo" + m /* ERROR "m[s] (map index expression of type int) is not used" */ [s] +} + +// Valid ERROR comments can have a variety of forms. +func _() { + 0 /* ERRORx "0 .* is not used" */ + 0 /* ERRORx "0 .* is not used" */ + 0 // ERRORx "0 .* is not used" + 0 // ERRORx "0 .* is not used" +} + +// Don't report spurious errors as a consequence of earlier errors. +// Add more tests as needed. +func _() { + if err := foo /* ERROR "undefined" */ (); err != nil /* "no error here" */ {} +} + +// Use unqualified names for package-local objects. +type T struct{} +var _ int = T /* ERROR "value of type T" */ {} // use T in error message rather than errors.T + +// Don't report errors containing "invalid type" (issue #24182). +func _(x *missing /* ERROR "undefined: missing" */ ) { + x.m() // there shouldn't be an error here referring to *invalid type +} diff --git a/src/internal/types/testdata/check/expr0.go b/src/internal/types/testdata/check/expr0.go new file mode 100644 index 0000000..eba991e --- /dev/null +++ b/src/internal/types/testdata/check/expr0.go @@ -0,0 +1,196 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// unary expressions + +package expr0 + +type mybool bool + +var ( + // bool + b0 = true + b1 bool = b0 + b2 = !true + b3 = !b1 + b4 bool = !true + b5 bool = !b4 + b6 = +b0 /* ERROR "not defined" */ + b7 = -b0 /* ERROR "not defined" */ + b8 = ^b0 /* ERROR "not defined" */ + b9 = *b0 /* ERROR "cannot indirect" */ + b10 = &true /* ERROR "cannot take address" */ + b11 = &b0 + b12 = <-b0 /* ERROR "cannot receive" */ + b13 = & & /* ERROR "cannot take address" */ b0 + b14 = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ b0 + + // byte + _ = byte(0) + _ = byte(- /* ERROR "cannot convert" */ 1) + _ = - /* ERROR "-byte(1) (constant -1 of type byte) overflows byte" */ byte(1) // test for issue 11367 + _ = byte /* ERROR "overflows byte" */ (0) - byte(1) + _ = ~ /* ERROR "cannot use ~ outside of interface or type constraint (use ^ for bitwise complement)" */ byte(0) + + // int + i0 = 1 + i1 int = i0 + i2 = +1 + i3 = +i0 + i4 int = +1 + i5 int = +i4 + i6 = -1 + i7 = -i0 + i8 int = -1 + i9 int = -i4 + i10 = !i0 /* ERROR "not defined" */ + i11 = ^1 + i12 = ^i0 + i13 int = ^1 + i14 int = ^i4 + i15 = *i0 /* ERROR "cannot indirect" */ + i16 = &i0 + i17 = *i16 + i18 = <-i16 /* ERROR "cannot receive" */ + i19 = ~ /* ERROR "cannot use ~ outside of interface or type constraint (use ^ for bitwise complement)" */ i0 + + // uint + u0 = uint(1) + u1 uint = u0 + u2 = +1 + u3 = +u0 + u4 uint = +1 + u5 uint = +u4 + u6 = -1 + u7 = -u0 + u8 uint = - /* ERROR "overflows" */ 1 + u9 uint = -u4 + u10 = !u0 /* ERROR "not defined" */ + u11 = ^1 + u12 = ^i0 + u13 uint = ^ /* ERROR "overflows" */ 1 + u14 uint = ^u4 + u15 = *u0 /* ERROR "cannot indirect" */ + u16 = &u0 + u17 = *u16 + u18 = <-u16 /* ERROR "cannot receive" */ + u19 = ^uint(0) + u20 = ~ /* ERROR "cannot use ~ outside of interface or type constraint (use ^ for bitwise complement)" */ u0 + + // float64 + f0 = float64(1) + f1 float64 = f0 + f2 = +1 + f3 = +f0 + f4 float64 = +1 + f5 float64 = +f4 + f6 = -1 + f7 = -f0 + f8 float64 = -1 + f9 float64 = -f4 + f10 = !f0 /* ERROR "not defined" */ + f11 = ^1 + f12 = ^i0 + f13 float64 = ^1 + f14 float64 = ^f4 /* ERROR "not defined" */ + f15 = *f0 /* ERROR "cannot indirect" */ + f16 = &f0 + f17 = *u16 + f18 = <-u16 /* ERROR "cannot receive" */ + f19 = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ f0 + + // complex128 + c0 = complex128(1) + c1 complex128 = c0 + c2 = +1 + c3 = +c0 + c4 complex128 = +1 + c5 complex128 = +c4 + c6 = -1 + c7 = -c0 + c8 complex128 = -1 + c9 complex128 = -c4 + c10 = !c0 /* ERROR "not defined" */ + c11 = ^1 + c12 = ^i0 + c13 complex128 = ^1 + c14 complex128 = ^c4 /* ERROR "not defined" */ + c15 = *c0 /* ERROR "cannot indirect" */ + c16 = &c0 + c17 = *u16 + c18 = <-u16 /* ERROR "cannot receive" */ + c19 = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ c0 + + // string + s0 = "foo" + s1 = +"foo" /* ERROR "not defined" */ + s2 = -s0 /* ERROR "not defined" */ + s3 = !s0 /* ERROR "not defined" */ + s4 = ^s0 /* ERROR "not defined" */ + s5 = *s4 + s6 = &s4 + s7 = *s6 + s8 = <-s7 + s9 = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ s0 + + // channel + ch chan int + rc <-chan float64 + sc chan <- string + ch0 = +ch /* ERROR "not defined" */ + ch1 = -ch /* ERROR "not defined" */ + ch2 = !ch /* ERROR "not defined" */ + ch3 = ^ch /* ERROR "not defined" */ + ch4 = *ch /* ERROR "cannot indirect" */ + ch5 = &ch + ch6 = *ch5 + ch7 = <-ch + ch8 = <-rc + ch9 = <-sc /* ERROR "cannot receive" */ + ch10, ok = <-ch + // ok is of type bool + ch11, myok = <-ch + _ mybool = myok /* ERRORx `cannot use .* in variable declaration` */ + ch12 = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ ch + +) + +// address of composite literals +type T struct{x, y int} + +func f() T { return T{} } + +var ( + _ = &T{1, 2} + _ = &[...]int{} + _ = &[]int{} + _ = &[]int{} + _ = &map[string]T{} + _ = &(T{1, 2}) + _ = &((((T{1, 2})))) + _ = &f /* ERROR "cannot take address" */ () +) + +// recursive pointer types +type P *P + +var ( + p1 P = new(P) + p2 P = *p1 + p3 P = &p2 +) + +func g() (a, b int) { return } + +func _() { + _ = -g /* ERROR "multiple-value g" */ () + _ = <-g /* ERROR "multiple-value g" */ () +} + +// ~ is accepted as unary operator only permitted in interface type elements +var ( + _ = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ 0 + _ = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ "foo" + _ = ~ /* ERROR "cannot use ~ outside of interface or type constraint" */ i0 +) diff --git a/src/internal/types/testdata/check/expr1.go b/src/internal/types/testdata/check/expr1.go new file mode 100644 index 0000000..1c04c8f --- /dev/null +++ b/src/internal/types/testdata/check/expr1.go @@ -0,0 +1,127 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// binary expressions + +package expr1 + +type mybool bool + +func _(x, y bool, z mybool) { + x = x || y + x = x || true + x = x || false + x = x && y + x = x && true + x = x && false + + z = z /* ERROR "mismatched types" */ || y + z = z || true + z = z || false + z = z /* ERROR "mismatched types" */ && y + z = z && true + z = z && false +} + +type myint int + +func _(x, y int, z myint) { + x = x + 1 + x = x + 1.0 + x = x + 1.1 // ERROR "truncated to int" + x = x + y + x = x - y + x = x * y + x = x / y + x = x % y + x = x << y + x = x >> y + + z = z + 1 + z = z + 1.0 + z = z + 1.1 // ERROR "truncated to int" + z = z /* ERROR "mismatched types" */ + y + z = z /* ERROR "mismatched types" */ - y + z = z /* ERROR "mismatched types" */ * y + z = z /* ERROR "mismatched types" */ / y + z = z /* ERROR "mismatched types" */ % y + z = z << y + z = z >> y +} + +type myuint uint + +func _(x, y uint, z myuint) { + x = x + 1 + x = x + - /* ERROR "overflows uint" */ 1 + x = x + 1.0 + x = x + 1.1 // ERROR "truncated to uint" + x = x + y + x = x - y + x = x * y + x = x / y + x = x % y + x = x << y + x = x >> y + + z = z + 1 + z = x + - /* ERROR "overflows uint" */ 1 + z = z + 1.0 + z = z + 1.1 // ERROR "truncated to uint" + z = z /* ERROR "mismatched types" */ + y + z = z /* ERROR "mismatched types" */ - y + z = z /* ERROR "mismatched types" */ * y + z = z /* ERROR "mismatched types" */ / y + z = z /* ERROR "mismatched types" */ % y + z = z << y + z = z >> y +} + +type myfloat64 float64 + +func _(x, y float64, z myfloat64) { + x = x + 1 + x = x + -1 + x = x + 1.0 + x = x + 1.1 + x = x + y + x = x - y + x = x * y + x = x / y + x = x /* ERROR "not defined" */ % y + x = x /* ERRORx `operand x .* must be integer` */ << y + x = x /* ERRORx `operand x .* must be integer` */ >> y + + z = z + 1 + z = z + -1 + z = z + 1.0 + z = z + 1.1 + z = z /* ERROR "mismatched types" */ + y + z = z /* ERROR "mismatched types" */ - y + z = z /* ERROR "mismatched types" */ * y + z = z /* ERROR "mismatched types" */ / y + z = z /* ERROR "mismatched types" */ % y + z = z /* ERRORx `operand z .* must be integer` */ << y + z = z /* ERRORx `operand z .* must be integer` */ >> y +} + +type mystring string + +func _(x, y string, z mystring) { + x = x + "foo" + x = x /* ERROR "not defined" */ - "foo" + x = x /* ERROR "mismatched types string and untyped int" */ + 1 + x = x + y + x = x /* ERROR "not defined" */ - y + x = x /* ERROR "mismatched types string and untyped int" */* 10 +} + +func f() (a, b int) { return } + +func _(x int) { + _ = f /* ERROR "multiple-value f" */ () + 1 + _ = x + f /* ERROR "multiple-value f" */ () + _ = f /* ERROR "multiple-value f" */ () + f + _ = f /* ERROR "multiple-value f" */ () + f /* ERROR "multiple-value f" */ () +} diff --git a/src/internal/types/testdata/check/expr2.go b/src/internal/types/testdata/check/expr2.go new file mode 100644 index 0000000..ebb85eb --- /dev/null +++ b/src/internal/types/testdata/check/expr2.go @@ -0,0 +1,260 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// comparisons + +package expr2 + +func _bool() { + const t = true == true + const f = true == false + _ = t /* ERRORx `operator .* not defined` */ < f + _ = 0 == t /* ERROR "mismatched types untyped int and untyped bool" */ + var b bool + var x, y float32 + b = x < y + _ = b + _ = struct{b bool}{x < y} +} + +// corner cases +var ( + v0 = nil == nil // ERROR "operator == not defined on untyped nil" +) + +func arrays() { + // basics + var a, b [10]int + _ = a == b + _ = a != b + _ = a /* ERROR "< not defined" */ < b + _ = a == nil /* ERROR "mismatched types" */ + + type C [10]int + var c C + _ = a == c + + type D [10]int + var d D + _ = c == d /* ERROR "mismatched types" */ + + var e [10]func() int + _ = e /* ERROR "[10]func() int cannot be compared" */ == e +} + +func structs() { + // basics + var s, t struct { + x int + a [10]float32 + _ bool + } + _ = s == t + _ = s != t + _ = s /* ERROR "< not defined" */ < t + _ = s == nil /* ERROR "mismatched types" */ + + type S struct { + x int + a [10]float32 + _ bool + } + type T struct { + x int + a [10]float32 + _ bool + } + var ss S + var tt T + _ = s == ss + _ = ss == tt /* ERROR "mismatched types" */ + + var u struct { + x int + a [10]map[string]int + } + _ = u /* ERROR "cannot be compared" */ == u +} + +func pointers() { + // nil + _ = nil == nil // ERROR "operator == not defined on untyped nil" + _ = nil != nil // ERROR "operator != not defined on untyped nil" + _ = nil /* ERROR "< not defined" */ < nil + _ = nil /* ERROR "<= not defined" */ <= nil + _ = nil /* ERROR "> not defined" */ > nil + _ = nil /* ERROR ">= not defined" */ >= nil + + // basics + var p, q *int + _ = p == q + _ = p != q + + _ = p == nil + _ = p != nil + _ = nil == q + _ = nil != q + + _ = p /* ERROR "< not defined" */ < q + _ = p /* ERROR "<= not defined" */ <= q + _ = p /* ERROR "> not defined" */ > q + _ = p /* ERROR ">= not defined" */ >= q + + // various element types + type ( + S1 struct{} + S2 struct{} + P1 *S1 + P2 *S2 + ) + var ( + ps1 *S1 + ps2 *S2 + p1 P1 + p2 P2 + ) + _ = ps1 == ps1 + _ = ps1 == ps2 /* ERROR "mismatched types" */ + _ = ps2 == ps1 /* ERROR "mismatched types" */ + + _ = p1 == p1 + _ = p1 == p2 /* ERROR "mismatched types" */ + + _ = p1 == ps1 +} + +func channels() { + // basics + var c, d chan int + _ = c == d + _ = c != d + _ = c == nil + _ = c /* ERROR "< not defined" */ < d + + // various element types (named types) + type ( + C1 chan int + C1r <-chan int + C1s chan<- int + C2 chan float32 + ) + var ( + c1 C1 + c1r C1r + c1s C1s + c1a chan int + c2 C2 + ) + _ = c1 == c1 + _ = c1 == c1r /* ERROR "mismatched types" */ + _ = c1 == c1s /* ERROR "mismatched types" */ + _ = c1r == c1s /* ERROR "mismatched types" */ + _ = c1 == c1a + _ = c1a == c1 + _ = c1 == c2 /* ERROR "mismatched types" */ + _ = c1a == c2 /* ERROR "mismatched types" */ + + // various element types (unnamed types) + var ( + d1 chan int + d1r <-chan int + d1s chan<- int + d1a chan<- int + d2 chan float32 + ) + _ = d1 == d1 + _ = d1 == d1r + _ = d1 == d1s + _ = d1r == d1s /* ERROR "mismatched types" */ + _ = d1 == d1a + _ = d1a == d1 + _ = d1 == d2 /* ERROR "mismatched types" */ + _ = d1a == d2 /* ERROR "mismatched types" */ +} + +// for interfaces test +type S1 struct{} +type S11 struct{} +type S2 struct{} +func (*S1) m() int +func (*S11) m() int +func (*S11) n() +func (*S2) m() float32 + +func interfaces() { + // basics + var i, j interface{ m() int } + _ = i == j + _ = i != j + _ = i == nil + _ = i /* ERROR "< not defined" */ < j + + // various interfaces + var ii interface { m() int; n() } + var k interface { m() float32 } + _ = i == ii + _ = i == k /* ERROR "mismatched types" */ + + // interfaces vs values + var s1 S1 + var s11 S11 + var s2 S2 + + _ = i == 0 /* ERROR "cannot convert" */ + _ = i == s1 /* ERROR "mismatched types" */ + _ = i == &s1 + _ = i == &s11 + + _ = i == s2 /* ERROR "mismatched types" */ + _ = i == & /* ERROR "mismatched types" */ s2 + + // issue #28164 + // testcase from issue + _ = interface{}(nil) == [ /* ERROR "slice can only be compared to nil" */ ]int(nil) + + // related cases + var e interface{} + var s []int + var x int + _ = e == s // ERROR "slice can only be compared to nil" + _ = s /* ERROR "slice can only be compared to nil" */ == e + _ = e /* ERROR "operator < not defined on interface" */ < x + _ = x < e // ERROR "operator < not defined on interface" +} + +func slices() { + // basics + var s []int + _ = s == nil + _ = s != nil + _ = s /* ERROR "< not defined" */ < nil + + // slices are not otherwise comparable + _ = s /* ERROR "slice can only be compared to nil" */ == s + _ = s /* ERROR "< not defined" */ < s +} + +func maps() { + // basics + var m map[string]int + _ = m == nil + _ = m != nil + _ = m /* ERROR "< not defined" */ < nil + + // maps are not otherwise comparable + _ = m /* ERROR "map can only be compared to nil" */ == m + _ = m /* ERROR "< not defined" */ < m +} + +func funcs() { + // basics + var f func(int) float32 + _ = f == nil + _ = f != nil + _ = f /* ERROR "< not defined" */ < nil + + // funcs are not otherwise comparable + _ = f /* ERROR "func can only be compared to nil" */ == f + _ = f /* ERROR "< not defined" */ < f +} diff --git a/src/internal/types/testdata/check/expr3.go b/src/internal/types/testdata/check/expr3.go new file mode 100644 index 0000000..91534cd --- /dev/null +++ b/src/internal/types/testdata/check/expr3.go @@ -0,0 +1,564 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package expr3 + +import "time" + +func indexes() { + _ = 1 /* ERROR "cannot index" */ [0] + _ = indexes /* ERROR "cannot index" */ [0] + _ = ( /* ERROR "cannot slice" */ 12 + 3)[1:2] + + var a [10]int + _ = a[true /* ERROR "cannot convert" */ ] + _ = a["foo" /* ERROR "cannot convert" */ ] + _ = a[1.1 /* ERROR "truncated" */ ] + _ = a[1.0] + _ = a[- /* ERROR "negative" */ 1] + _ = a[- /* ERROR "negative" */ 1 :] + _ = a[: - /* ERROR "negative" */ 1] + _ = a[: /* ERROR "middle index required" */ : /* ERROR "final index required" */ ] + _ = a[0: /* ERROR "middle index required" */ : /* ERROR "final index required" */ ] + _ = a[0: /* ERROR "middle index required" */ :10] + _ = a[:10:10] + + var a0 int + a0 = a[0] + _ = a0 + var a1 int32 + a1 = a /* ERRORx `cannot use .* in assignment` */ [1] + _ = a1 + + _ = a[9] + _ = a[10 /* ERRORx `index .* out of bounds` */ ] + _ = a[1 /* ERROR "overflows" */ <<100] + _ = a[1<< /* ERROR "constant shift overflow" */ 1000] // no out-of-bounds follow-on error + _ = a[10:] + _ = a[:10] + _ = a[10:10] + _ = a[11 /* ERRORx `index .* out of bounds` */ :] + _ = a[: 11 /* ERRORx `index .* out of bounds` */ ] + _ = a[: 1 /* ERROR "overflows" */ <<100] + _ = a[:10:10] + _ = a[:11 /* ERRORx `index .* out of bounds` */ :10] + _ = a[:10:11 /* ERRORx `index .* out of bounds` */ ] + _ = a[10:0 /* ERROR "invalid slice indices" */ :10] + _ = a[0:10:0 /* ERROR "invalid slice indices" */ ] + _ = a[10:0 /* ERROR "invalid slice indices" */:0] + _ = &a /* ERROR "cannot take address" */ [:10] + + pa := &a + _ = pa[9] + _ = pa[10 /* ERRORx `index .* out of bounds` */ ] + _ = pa[1 /* ERROR "overflows" */ <<100] + _ = pa[10:] + _ = pa[:10] + _ = pa[10:10] + _ = pa[11 /* ERRORx `index .* out of bounds` */ :] + _ = pa[: 11 /* ERRORx `index .* out of bounds` */ ] + _ = pa[: 1 /* ERROR "overflows" */ <<100] + _ = pa[:10:10] + _ = pa[:11 /* ERRORx `index .* out of bounds` */ :10] + _ = pa[:10:11 /* ERRORx `index .* out of bounds` */ ] + _ = pa[10:0 /* ERROR "invalid slice indices" */ :10] + _ = pa[0:10:0 /* ERROR "invalid slice indices" */ ] + _ = pa[10:0 /* ERROR "invalid slice indices" */ :0] + _ = &pa /* ERROR "cannot take address" */ [:10] + + var b [0]int + _ = b[0 /* ERRORx `index .* out of bounds` */ ] + _ = b[:] + _ = b[0:] + _ = b[:0] + _ = b[0:0] + _ = b[0:0:0] + _ = b[1 /* ERRORx `index .* out of bounds` */ :0:0] + + var s []int + _ = s[- /* ERROR "negative" */ 1] + _ = s[- /* ERROR "negative" */ 1 :] + _ = s[: - /* ERROR "negative" */ 1] + _ = s[0] + _ = s[1:2] + _ = s[2:1 /* ERROR "invalid slice indices" */ ] + _ = s[2:] + _ = s[: 1 /* ERROR "overflows" */ <<100] + _ = s[1 /* ERROR "overflows" */ <<100 :] + _ = s[1 /* ERROR "overflows" */ <<100 : 1 /* ERROR "overflows" */ <<100] + _ = s[: /* ERROR "middle index required" */ : /* ERROR "final index required" */ ] + _ = s[:10:10] + _ = s[10:0 /* ERROR "invalid slice indices" */ :10] + _ = s[0:10:0 /* ERROR "invalid slice indices" */ ] + _ = s[10:0 /* ERROR "invalid slice indices" */ :0] + _ = &s /* ERROR "cannot take address" */ [:10] + + var m map[string]int + _ = m[0 /* ERRORx `cannot use .* in map index` */ ] + _ = m /* ERROR "cannot slice" */ ["foo" : "bar"] + _ = m["foo"] + // ok is of type bool + type mybool bool + var ok mybool + _, ok = m["bar"] + _ = ok + _ = m/* ERROR "mismatched types int and untyped string" */[0 /* ERROR "cannot use 0" */ ] + "foo" + + var t string + _ = t[- /* ERROR "negative" */ 1] + _ = t[- /* ERROR "negative" */ 1 :] + _ = t[: - /* ERROR "negative" */ 1] + _ = t[1:2:3 /* ERROR "3-index slice of string" */ ] + _ = "foo"[1:2:3 /* ERROR "3-index slice of string" */ ] + var t0 byte + t0 = t[0] + _ = t0 + var t1 rune + t1 = t /* ERRORx `cannot use .* in assignment` */ [2] + _ = t1 + _ = ("foo" + "bar")[5] + _ = ("foo" + "bar")[6 /* ERRORx `index .* out of bounds` */ ] + + const c = "foo" + _ = c[- /* ERROR "negative" */ 1] + _ = c[- /* ERROR "negative" */ 1 :] + _ = c[: - /* ERROR "negative" */ 1] + var c0 byte + c0 = c[0] + _ = c0 + var c2 float32 + c2 = c /* ERRORx `cannot use .* in assignment` */ [2] + _ = c[3 /* ERRORx `index .* out of bounds` */ ] + _ = ""[0 /* ERRORx `index .* out of bounds` */ ] + _ = c2 + + _ = s[1<<30] // no compile-time error here + + // issue 4913 + type mystring string + var ss string + var ms mystring + var i, j int + ss = "foo"[1:2] + ss = "foo"[i:j] + ms = "foo" /* ERRORx `cannot use .* in assignment` */ [1:2] + ms = "foo" /* ERRORx `cannot use .* in assignment` */ [i:j] + _, _ = ss, ms +} + +type T struct { + x int + y func() +} + +func (*T) m() {} + +func method_expressions() { + _ = T.a /* ERROR "no field or method" */ + _ = T.x /* ERROR "has no method" */ + _ = T.m /* ERROR "invalid method expression T.m (needs pointer receiver (*T).m)" */ + _ = (*T).m + + var f func(*T) = T.m /* ERROR "invalid method expression T.m (needs pointer receiver (*T).m)" */ + var g func(*T) = (*T).m + _, _ = f, g + + _ = T.y /* ERROR "has no method" */ + _ = (*T).y /* ERROR "has no method" */ +} + +func struct_literals() { + type T0 struct { + a, b, c int + } + + type T1 struct { + T0 + a, b int + u float64 + s string + } + + // keyed elements + _ = T1{} + _ = T1{a: 0, 1 /* ERRORx `mixture of .* elements` */ } + _ = T1{aa /* ERROR "unknown field" */ : 0} + _ = T1{1 /* ERROR "invalid field name" */ : 0} + _ = T1{a: 0, s: "foo", u: 0, a /* ERROR "duplicate field" */: 10} + _ = T1{a: "foo" /* ERRORx `cannot use .* in struct literal` */ } + _ = T1{c /* ERROR "unknown field" */ : 0} + _ = T1{T0: { /* ERROR "missing type" */ }} // struct literal element type may not be elided + _ = T1{T0: T0{}} + _ = T1{T0 /* ERROR "invalid field name" */ .a: 0} + + // unkeyed elements + _ = T0{1, 2, 3} + _ = T0{1, b /* ERROR "mixture" */ : 2, 3} + _ = T0{1, 2} /* ERROR "too few values" */ + _ = T0{1, 2, 3, 4 /* ERROR "too many values" */ } + _ = T0{1, "foo" /* ERRORx `cannot use .* in struct literal` */, 3.4 /* ERRORx `cannot use .*\(truncated\)` */} + + // invalid type + type P *struct{ + x int + } + _ = P /* ERROR "invalid composite literal type" */ {} + + // unexported fields + _ = time.Time{} + _ = time.Time{sec /* ERROR "unknown field" */ : 0} + _ = time.Time{ + 0 /* ERROR "implicit assignment to unexported field wall in struct literal" */, + 0 /* ERROR "implicit assignment" */ , + nil /* ERROR "implicit assignment" */ , + } +} + +func array_literals() { + type A0 [0]int + _ = A0{} + _ = A0{0 /* ERRORx `index .* out of bounds` */} + _ = A0{0 /* ERRORx `index .* out of bounds` */ : 0} + + type A1 [10]int + _ = A1{} + _ = A1{0, 1, 2} + _ = A1{0, 1, 2, 3, 4, 5, 6, 7, 8, 9} + _ = A1{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 /* ERRORx `index .* out of bounds` */ } + _ = A1{- /* ERROR "negative" */ 1: 0} + _ = A1{8: 8, 9} + _ = A1{8: 8, 9, 10 /* ERRORx `index .* out of bounds` */ } + _ = A1{0, 1, 2, 0 /* ERROR "duplicate index" */ : 0, 3: 3, 4} + _ = A1{5: 5, 6, 7, 3: 3, 4} + _ = A1{5: 5, 6, 7, 3: 3, 4, 5 /* ERROR "duplicate index" */ } + _ = A1{10 /* ERRORx `index .* out of bounds` */ : 10, 10 /* ERRORx `index .* out of bounds` */ : 10} + _ = A1{5: 5, 6, 7, 3: 3, 1 /* ERROR "overflows" */ <<100: 4, 5 /* ERROR "duplicate index" */ } + _ = A1{5: 5, 6, 7, 4: 4, 1 /* ERROR "overflows" */ <<100: 4} + _ = A1{2.0} + _ = A1{2.1 /* ERROR "truncated" */ } + _ = A1{"foo" /* ERRORx `cannot use .* in array or slice literal` */ } + + // indices must be integer constants + i := 1 + const f = 2.1 + const s = "foo" + _ = A1{i /* ERROR "index i must be integer constant" */ : 0} + _ = A1{f /* ERROR "truncated" */ : 0} + _ = A1{s /* ERROR "cannot convert" */ : 0} + + a0 := [...]int{} + assert(len(a0) == 0) + + a1 := [...]int{0, 1, 2} + assert(len(a1) == 3) + var a13 [3]int + var a14 [4]int + a13 = a1 + a14 = a1 /* ERRORx `cannot use .* in assignment` */ + _, _ = a13, a14 + + a2 := [...]int{- /* ERROR "negative" */ 1: 0} + _ = a2 + + a3 := [...]int{0, 1, 2, 0 /* ERROR "duplicate index" */ : 0, 3: 3, 4} + assert(len(a3) == 5) // somewhat arbitrary + + a4 := [...]complex128{0, 1, 2, 1<<10-2: -1i, 1i, 400: 10, 12, 14} + assert(len(a4) == 1024) + + // composite literal element types may be elided + type T []int + _ = [10]T{T{}, {}, 5: T{1, 2, 3}, 7: {1, 2, 3}} + a6 := [...]T{T{}, {}, 5: T{1, 2, 3}, 7: {1, 2, 3}} + assert(len(a6) == 8) + + // recursively so + _ = [10][10]T{{}, [10]T{{}}, {{1, 2, 3}}} + + // from the spec + type Point struct { x, y float32 } + _ = [...]Point{Point{1.5, -3.5}, Point{0, 0}} + _ = [...]Point{{1.5, -3.5}, {0, 0}} + _ = [][]int{[]int{1, 2, 3}, []int{4, 5}} + _ = [][]int{{1, 2, 3}, {4, 5}} + _ = [...]*Point{&Point{1.5, -3.5}, &Point{0, 0}} + _ = [...]*Point{{1.5, -3.5}, {0, 0}} +} + +func slice_literals() { + type S0 []int + _ = S0{} + _ = S0{0, 1, 2} + _ = S0{0, 1, 2, 3, 4, 5, 6, 7, 8, 9} + _ = S0{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10} + _ = S0{- /* ERROR "negative" */ 1: 0} + _ = S0{8: 8, 9} + _ = S0{8: 8, 9, 10} + _ = S0{0, 1, 2, 0 /* ERROR "duplicate index" */ : 0, 3: 3, 4} + _ = S0{5: 5, 6, 7, 3: 3, 4} + _ = S0{5: 5, 6, 7, 3: 3, 4, 5 /* ERROR "duplicate index" */ } + _ = S0{10: 10, 10 /* ERROR "duplicate index" */ : 10} + _ = S0{5: 5, 6, 7, 3: 3, 1 /* ERROR "overflows" */ <<100: 4, 5 /* ERROR "duplicate index" */ } + _ = S0{5: 5, 6, 7, 4: 4, 1 /* ERROR "overflows" */ <<100: 4} + _ = S0{2.0} + _ = S0{2.1 /* ERROR "truncated" */ } + _ = S0{"foo" /* ERRORx `cannot use .* in array or slice literal` */ } + + // indices must be resolved correctly + const index1 = 1 + _ = S0{index1: 1} + _ = S0{index2: 2} + _ = S0{index3 /* ERROR "undefined" */ : 3} + + // indices must be integer constants + i := 1 + const f = 2.1 + const s = "foo" + _ = S0{i /* ERROR "index i must be integer constant" */ : 0} + _ = S0{f /* ERROR "truncated" */ : 0} + _ = S0{s /* ERROR "cannot convert" */ : 0} + + // composite literal element types may be elided + type T []int + _ = []T{T{}, {}, 5: T{1, 2, 3}, 7: {1, 2, 3}} + _ = [][]int{{1, 2, 3}, {4, 5}} + + // recursively so + _ = [][]T{{}, []T{{}}, {{1, 2, 3}}} + + // issue 17954 + type T0 *struct { s string } + _ = []T0{{}} + _ = []T0{{"foo"}} + + type T1 *struct{ int } + _ = []T1{} + _ = []T1{{0}, {1}, {2}} + + type T2 T1 + _ = []T2{} + _ = []T2{{0}, {1}, {2}} + + _ = map[T0]T2{} + _ = map[T0]T2{{}: {}} +} + +const index2 int = 2 + +type N int +func (N) f() {} + +func map_literals() { + type M0 map[string]int + type M1 map[bool]int + type M2 map[*int]int + + _ = M0{} + _ = M0{1 /* ERROR "missing key" */ } + _ = M0{1 /* ERRORx `cannot use .* in map literal` */ : 2} + _ = M0{"foo": "bar" /* ERRORx `cannot use .* in map literal` */ } + _ = M0{"foo": 1, "bar": 2, "foo" /* ERROR "duplicate key" */ : 3 } + + _ = map[interface{}]int{2: 1, 2 /* ERROR "duplicate key" */ : 1} + _ = map[interface{}]int{int(2): 1, int16(2): 1} + _ = map[interface{}]int{int16(2): 1, int16 /* ERROR "duplicate key" */ (2): 1} + + type S string + + _ = map[interface{}]int{"a": 1, "a" /* ERROR "duplicate key" */ : 1} + _ = map[interface{}]int{"a": 1, S("a"): 1} + _ = map[interface{}]int{S("a"): 1, S /* ERROR "duplicate key" */ ("a"): 1} + _ = map[interface{}]int{1.0: 1, 1.0 /* ERROR "duplicate key" */: 1} + _ = map[interface{}]int{int64(-1): 1, int64 /* ERROR "duplicate key" */ (-1) : 1} + _ = map[interface{}]int{^uint64(0): 1, ^ /* ERROR "duplicate key" */ uint64(0): 1} + _ = map[interface{}]int{complex(1,2): 1, complex /* ERROR "duplicate key" */ (1,2) : 1} + + type I interface { + f() + } + + _ = map[I]int{N(0): 1, N(2): 1} + _ = map[I]int{N(2): 1, N /* ERROR "duplicate key" */ (2): 1} + + // map keys must be resolved correctly + key1 := "foo" + _ = M0{key1: 1} + _ = M0{key2: 2} + _ = M0{key3 /* ERROR "undefined" */ : 2} + + var value int + _ = M1{true: 1, false: 0} + _ = M2{nil: 0, &value: 1} + + // composite literal element types may be elided + type T [2]int + _ = map[int]T{0: T{3, 4}, 1: {5, 6}} + + // recursively so + _ = map[int][]T{0: {}, 1: {{}, T{1, 2}}} + + // composite literal key types may be elided + _ = map[T]int{T{3, 4}: 0, {5, 6}: 1} + + // recursively so + _ = map[[2]T]int{{}: 0, {{}}: 1, [2]T{{}}: 2, {T{1, 2}}: 3} + + // composite literal element and key types may be elided + _ = map[T]T{{}: {}, {1, 2}: T{3, 4}, T{4, 5}: {}} + _ = map[T]M0{{} : {}, T{1, 2}: M0{"foo": 0}, {1, 3}: {"foo": 1}} + + // recursively so + _ = map[[2]T][]T{{}: {}, {{}}: {{}, T{1, 2}}, [2]T{{}}: nil, {T{1, 2}}: {{}, {}}} + + // from the spec + type Point struct { x, y float32 } + _ = map[string]Point{"orig": {0, 0}} + _ = map[*Point]string{{0, 0}: "orig"} + + // issue 17954 + type T0 *struct{ s string } + type T1 *struct{ int } + type T2 T1 + + _ = map[T0]T2{} + _ = map[T0]T2{{}: {}} +} + +var key2 string = "bar" + +type I interface { + m() +} + +type I2 interface { + m(int) +} + +type T1 struct{} +type T2 struct{} + +func (T2) m(int) {} + +type mybool bool + +func type_asserts() { + var x int + _ = x /* ERROR "not an interface" */ .(int) + + var e interface{} + var ok bool + x, ok = e.(int) + _ = ok + + // ok value is of type bool + var myok mybool + _, myok = e.(int) + _ = myok + + var t I + _ = t /* ERRORx `use of .* outside type switch` */ .(type) + _ = t /* ERROR "m has pointer receiver" */ .(T) + _ = t.(*T) + _ = t /* ERROR "missing method m" */ .(T1) + _ = t /* ERROR "wrong type for method m" */ .(T2) + _ = t /* STRICT "wrong type for method m" */ .(I2) // only an error in strict mode (issue 8561) + + // e doesn't statically have an m, but may have one dynamically. + _ = e.(I2) +} + +func f0() {} +func f1(x int) {} +func f2(u float32, s string) {} +func fs(s []byte) {} +func fv(x ...int) {} +func fi(x ... interface{}) {} +func (T) fm(x ...int) + +func g0() {} +func g1() int { return 0} +func g2() (u float32, s string) { return } +func gs() []byte { return nil } + +func _calls() { + var x int + var y float32 + var s []int + + f0() + _ = f0 /* ERROR "used as value" */ () + f0(g0 /* ERROR "too many arguments" */ ) + + f1(0) + f1(x) + f1(10.0) + f1() /* ERROR "not enough arguments in call to f1\n\thave ()\n\twant (int)" */ + f1(x, y /* ERROR "too many arguments in call to f1\n\thave (int, float32)\n\twant (int)" */ ) + f1(s /* ERRORx `cannot use .* in argument` */ ) + f1(x ... /* ERROR "cannot use ..." */ ) + f1(g0 /* ERROR "used as value" */ ()) + f1(g1()) + f1(g2 /* ERROR "too many arguments in call to f1\n\thave (float32, string)\n\twant (int)" */ ()) + + f2() /* ERROR "not enough arguments in call to f2\n\thave ()\n\twant (float32, string)" */ + f2(3.14) /* ERROR "not enough arguments in call to f2\n\thave (number)\n\twant (float32, string)" */ + f2(3.14, "foo") + f2(x /* ERRORx `cannot use .* in argument` */ , "foo") + f2(g0 /* ERROR "used as value" */ ()) /* ERROR "not enough arguments in call to f2\n\thave (func())\n\twant (float32, string)" */ + f2(g1()) /* ERROR "not enough arguments in call to f2\n\thave (int)\n\twant (float32, string)" */ + f2(g2()) + + fs() /* ERROR "not enough arguments" */ + fs(g0 /* ERROR "used as value" */ ()) + fs(g1 /* ERRORx `cannot use .* in argument` */ ()) + fs(g2 /* ERROR "too many arguments" */ ()) + fs(gs()) + + fv() + fv(1, 2.0, x) + fv(s /* ERRORx `cannot use .* in argument` */ ) + fv(s...) + fv(x /* ERROR "cannot use" */ ...) + fv(1, s /* ERROR "too many arguments" */ ...) + fv(gs /* ERRORx `cannot use .* in argument` */ ()) + fv(gs /* ERRORx `cannot use .* in argument` */ ()...) + + var t T + t.fm() + t.fm(1, 2.0, x) + t.fm(s /* ERRORx `cannot use .* in argument` */ ) + t.fm(g1()) + t.fm(1, s /* ERROR "too many arguments" */ ...) + t.fm(gs /* ERRORx `cannot use .* in argument` */ ()) + t.fm(gs /* ERRORx `cannot use .* in argument` */ ()...) + + T.fm(t, ) + T.fm(t, 1, 2.0, x) + T.fm(t, s /* ERRORx `cannot use .* in argument` */ ) + T.fm(t, g1()) + T.fm(t, 1, s /* ERROR "too many arguments" */ ...) + T.fm(t, gs /* ERRORx `cannot use .* in argument` */ ()) + T.fm(t, gs /* ERRORx `cannot use .* in argument` */ ()...) + + var i interface{ fm(x ...int) } = t + i.fm() + i.fm(1, 2.0, x) + i.fm(s /* ERRORx `cannot use .* in argument` */ ) + i.fm(g1()) + i.fm(1, s /* ERROR "too many arguments" */ ...) + i.fm(gs /* ERRORx `cannot use .* in argument` */ ()) + i.fm(gs /* ERRORx `cannot use .* in argument` */ ()...) + + fi() + fi(1, 2.0, x, 3.14, "foo") + fi(g2()) + fi(0, g2) + fi(0, g2 /* ERROR "multiple-value g2" */ ()) +} + +func issue6344() { + type T []interface{} + var x T + fi(x...) // ... applies also to named slices +} diff --git a/src/internal/types/testdata/check/funcinference.go b/src/internal/types/testdata/check/funcinference.go new file mode 100644 index 0000000..e0e978f --- /dev/null +++ b/src/internal/types/testdata/check/funcinference.go @@ -0,0 +1,112 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package funcInference + +import "strconv" + +type any interface{} + +func f0[A any, B interface{*C}, C interface{*D}, D interface{*A}](A, B, C, D) {} +func _() { + f := f0[string] + f("a", nil, nil, nil) + f0("a", nil, nil, nil) +} + +func f1[A any, B interface{*A}](A, B) {} +func _() { + f := f1[int] + f(int(0), new(int)) + f1(int(0), new(int)) +} + +func f2[A any, B interface{[]A}](A, B) {} +func _() { + f := f2[byte] + f(byte(0), []byte{}) + f2(byte(0), []byte{}) +} + +// Embedding stand-alone type parameters is not permitted for now. Disabled. +// func f3[A any, B interface{~C}, C interface{~*A}](A, B, C) +// func _() { +// f := f3[int] +// var x int +// f(x, &x, &x) +// f3(x, &x, &x) +// } + +func f4[A any, B interface{[]C}, C interface{*A}](A, B, C) {} +func _() { + f := f4[int] + var x int + f(x, []*int{}, &x) + f4(x, []*int{}, &x) +} + +func f5[A interface{struct{b B; c C}}, B any, C interface{*B}](x B) A { panic(0) } +func _() { + x := f5(1.2) + var _ float64 = x.b + var _ float64 = *x.c +} + +func f6[A any, B interface{~struct{f []A}}](B) A { panic(0) } +func _() { + x := f6(struct{f []string}{}) + var _ string = x +} + +func f7[A interface{*B}, B interface{~*A}]() {} + +// More realistic examples + +func Double[S interface{ ~[]E }, E interface{ ~int | ~int8 | ~int16 | ~int32 | ~int64 }](s S) S { + r := make(S, len(s)) + for i, v := range s { + r[i] = v + v + } + return r +} + +type MySlice []int + +var _ = Double(MySlice{1}) + +// From the draft design. + +type Setter[B any] interface { + Set(string) + *B +} + +func FromStrings[T interface{}, PT Setter[T]](s []string) []T { + result := make([]T, len(s)) + for i, v := range s { + // The type of &result[i] is *T which is in the type set + // of Setter, so we can convert it to PT. + p := PT(&result[i]) + // PT has a Set method. + p.Set(v) + } + return result +} + +type Settable int + +func (p *Settable) Set(s string) { + i, _ := strconv.Atoi(s) // real code should not ignore the error + *p = Settable(i) +} + +var _ = FromStrings[Settable]([]string{"1", "2"}) + +// Suitable error message when the type parameter is provided (rather than inferred). + +func f8[P, Q any](P, Q) {} + +func _(s string) { + f8[int](s /* ERROR "cannot use s (variable of type string) as int value in argument to f8[int]" */ , s) +} diff --git a/src/internal/types/testdata/check/go1_12.go b/src/internal/types/testdata/check/go1_12.go new file mode 100644 index 0000000..b47d3de --- /dev/null +++ b/src/internal/types/testdata/check/go1_12.go @@ -0,0 +1,36 @@ +// -lang=go1.12 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +package p + +// numeric literals +const ( + _ = 1_000 // ERROR "underscores in numeric literals requires go1.13 or later" + _ = 0b111 // ERROR "binary literals requires go1.13 or later" + _ = 0o567 // ERROR "0o/0O-style octal literals requires go1.13 or later" + _ = 0xabc // ok + _ = 0x0p1 // ERROR "hexadecimal floating-point literals requires go1.13 or later" + + _ = 0B111 // ERROR "binary" + _ = 0O567 // ERROR "octal" + _ = 0Xabc // ok + _ = 0X0P1 // ERROR "hexadecimal floating-point" + + _ = 1_000i // ERROR "underscores" + _ = 0b111i // ERROR "binary" + _ = 0o567i // ERROR "octal" + _ = 0xabci // ERROR "hexadecimal floating-point" + _ = 0x0p1i // ERROR "hexadecimal floating-point" +) + +// signed shift counts +var ( + s int + _ = 1 << s // ERROR "invalid operation: signed shift count s (variable of type int) requires go1.13 or later" + _ = 1 >> s // ERROR "signed shift count" +) diff --git a/src/internal/types/testdata/check/go1_13.go b/src/internal/types/testdata/check/go1_13.go new file mode 100644 index 0000000..cc7861d --- /dev/null +++ b/src/internal/types/testdata/check/go1_13.go @@ -0,0 +1,23 @@ +// -lang=go1.13 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +package p + +// interface embedding + +type I interface { m() } + +type _ interface { + m() + I // ERROR "duplicate method m" +} + +type _ interface { + I + I // ERROR "duplicate method m" +} diff --git a/src/internal/types/testdata/check/go1_16.go b/src/internal/types/testdata/check/go1_16.go new file mode 100644 index 0000000..9675b29 --- /dev/null +++ b/src/internal/types/testdata/check/go1_16.go @@ -0,0 +1,15 @@ +// -lang=go1.16 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (*Array)(s /* ERROR "requires go1.17 or later" */ ) diff --git a/src/internal/types/testdata/check/go1_19.go b/src/internal/types/testdata/check/go1_19.go new file mode 100644 index 0000000..b6cff4f --- /dev/null +++ b/src/internal/types/testdata/check/go1_19.go @@ -0,0 +1,15 @@ +// -lang=go1.19 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (Array)(s /* ERROR "requires go1.20 or later" */) diff --git a/src/internal/types/testdata/check/go1_19_20.go b/src/internal/types/testdata/check/go1_19_20.go new file mode 100644 index 0000000..52e5dfd --- /dev/null +++ b/src/internal/types/testdata/check/go1_19_20.go @@ -0,0 +1,17 @@ +// -lang=go1.19 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +//go:build go1.20 + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (Array)(s /* ok */) diff --git a/src/internal/types/testdata/check/go1_20_19.go b/src/internal/types/testdata/check/go1_20_19.go new file mode 100644 index 0000000..08365a7 --- /dev/null +++ b/src/internal/types/testdata/check/go1_20_19.go @@ -0,0 +1,17 @@ +// -lang=go1.20 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +//go:build go1.19 + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (Array)(s /* ok because Go 1.20 ignored the //go:build go1.19 */) diff --git a/src/internal/types/testdata/check/go1_21_19.go b/src/internal/types/testdata/check/go1_21_19.go new file mode 100644 index 0000000..2acd258 --- /dev/null +++ b/src/internal/types/testdata/check/go1_21_19.go @@ -0,0 +1,17 @@ +// -lang=go1.21 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +//go:build go1.19 + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (Array)(s /* ERROR "requires go1.20 or later" */) diff --git a/src/internal/types/testdata/check/go1_8.go b/src/internal/types/testdata/check/go1_8.go new file mode 100644 index 0000000..6a7e639 --- /dev/null +++ b/src/internal/types/testdata/check/go1_8.go @@ -0,0 +1,12 @@ +// -lang=go1.8 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +package p + +// type alias declarations +type any = /* ERROR "type aliases requires go1.9 or later" */ interface{} diff --git a/src/internal/types/testdata/check/go1_xx_19.go b/src/internal/types/testdata/check/go1_xx_19.go new file mode 100644 index 0000000..01f6b7d --- /dev/null +++ b/src/internal/types/testdata/check/go1_xx_19.go @@ -0,0 +1,15 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check Go language version-specific errors. + +//go:build go1.19 + +package p + +type Slice []byte +type Array [8]byte + +var s Slice +var p = (Array)(s /* ok because Go 1.X prior to Go 1.21 ignored the //go:build go1.19 */) diff --git a/src/internal/types/testdata/check/gotos.go b/src/internal/types/testdata/check/gotos.go new file mode 100644 index 0000000..069a94b --- /dev/null +++ b/src/internal/types/testdata/check/gotos.go @@ -0,0 +1,560 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is a modified copy of $GOROOT/test/goto.go. + +package gotos + +var ( + i, n int + x []int + c chan int + m map[int]int + s string +) + +// goto after declaration okay +func _() { + x := 1 + goto L +L: + _ = x +} + +// goto before declaration okay +func _() { + goto L +L: + x := 1 + _ = x +} + +// goto across declaration not okay +func _() { + goto L /* ERROR "goto L jumps over variable declaration at line 36" */ + x := 1 + _ = x +L: +} + +// goto across declaration in inner scope okay +func _() { + goto L + { + x := 1 + _ = x + } +L: +} + +// goto across declaration after inner scope not okay +func _() { + goto L /* ERROR "goto L jumps over variable declaration at line 58" */ + { + x := 1 + _ = x + } + x := 1 + _ = x +L: +} + +// goto across declaration in reverse okay +func _() { +L: + x := 1 + _ = x + goto L +} + +func _() { +L: L1: + x := 1 + _ = x + goto L + goto L1 +} + +// error shows first offending variable +func _() { + goto L /* ERROR "goto L jumps over variable declaration at line 84" */ + x := 1 + _ = x + y := 1 + _ = y +L: +} + +// goto not okay even if code path is dead +func _() { + goto L /* ERROR "goto L jumps over variable declaration" */ + x := 1 + _ = x + y := 1 + _ = y + return +L: +} + +// goto into outer block okay +func _() { + { + goto L + } +L: +} + +func _() { + { + goto L + goto L1 + } +L: L1: +} + +// goto backward into outer block okay +func _() { +L: + { + goto L + } +} + +func _() { +L: L1: + { + goto L + goto L1 + } +} + +// goto into inner block not okay +func _() { + goto L /* ERROR "goto L jumps into block" */ + { + L: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + goto L1 /* ERROR "goto L1 jumps into block" */ + { + L: L1: + } +} + +// goto backward into inner block still not okay +func _() { + { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + { + L: L1: + } + goto L /* ERROR "goto L jumps into block" */ + goto L1 /* ERROR "goto L1 jumps into block" */ +} + +// error shows first (outermost) offending block +func _() { + goto L /* ERROR "goto L jumps into block" */ + { + { + { + L: + } + } + } +} + +// error prefers block diagnostic over declaration diagnostic +func _() { + goto L /* ERROR "goto L jumps into block" */ + x := 1 + _ = x + { + L: + } +} + +// many kinds of blocks, all invalid to jump into or among, +// but valid to jump out of + +// if + +func _() { +L: + if true { + goto L + } +} + +func _() { +L: + if true { + goto L + } else { + } +} + +func _() { +L: + if false { + } else { + goto L + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + if true { + L: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + if true { + L: + } else { + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + if true { + } else { + L: + } +} + +func _() { + if false { + L: + } else { + goto L /* ERROR "goto L jumps into block" */ + } +} + +func _() { + if true { + goto L /* ERROR "goto L jumps into block" */ + } else { + L: + } +} + +func _() { + if true { + goto L /* ERROR "goto L jumps into block" */ + } else if false { + L: + } +} + +func _() { + if true { + goto L /* ERROR "goto L jumps into block" */ + } else if false { + L: + } else { + } +} + +func _() { + if true { + goto L /* ERROR "goto L jumps into block" */ + } else if false { + } else { + L: + } +} + +func _() { + if true { + goto L /* ERROR "goto L jumps into block" */ + } else { + L: + } +} + +func _() { + if true { + L: + } else { + goto L /* ERROR "goto L jumps into block" */ + } +} + +// for + +func _() { + for { + goto L + } +L: +} + +func _() { + for { + goto L + L: + } +} + +func _() { + for { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for { + goto L + L1: + } +L: + goto L1 /* ERROR "goto L1 jumps into block" */ +} + +func _() { + for i < n { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for i = 0; i < n; i++ { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for i = range x { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for i = range c { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for i = range m { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +func _() { + for i = range s { + L: + } + goto L /* ERROR "goto L jumps into block" */ +} + +// switch + +func _() { +L: + switch i { + case 0: + goto L + } +} + +func _() { +L: + switch i { + case 0: + + default: + goto L + } +} + +func _() { + switch i { + case 0: + + default: + L: + goto L + } +} + +func _() { + switch i { + case 0: + + default: + goto L + L: + } +} + +func _() { + switch i { + case 0: + goto L + L: + ; + default: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + switch i { + case 0: + L: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + switch i { + case 0: + L: + ; + default: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + switch i { + case 0: + default: + L: + } +} + +func _() { + switch i { + default: + goto L /* ERROR "goto L jumps into block" */ + case 0: + L: + } +} + +func _() { + switch i { + case 0: + L: + ; + default: + goto L /* ERROR "goto L jumps into block" */ + } +} + +// select +// different from switch. the statement has no implicit block around it. + +func _() { +L: + select { + case <-c: + goto L + } +} + +func _() { +L: + select { + case c <- 1: + + default: + goto L + } +} + +func _() { + select { + case <-c: + + default: + L: + goto L + } +} + +func _() { + select { + case c <- 1: + + default: + goto L + L: + } +} + +func _() { + select { + case <-c: + goto L + L: + ; + default: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + select { + case c <- 1: + L: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + select { + case c <- 1: + L: + ; + default: + } +} + +func _() { + goto L /* ERROR "goto L jumps into block" */ + select { + case <-c: + default: + L: + } +} + +func _() { + select { + default: + goto L /* ERROR "goto L jumps into block" */ + case <-c: + L: + } +} + +func _() { + select { + case <-c: + L: + ; + default: + goto L /* ERROR "goto L jumps into block" */ + } +} diff --git a/src/internal/types/testdata/check/importC.go b/src/internal/types/testdata/check/importC.go new file mode 100644 index 0000000..2cdf383 --- /dev/null +++ b/src/internal/types/testdata/check/importC.go @@ -0,0 +1,56 @@ +// -fakeImportC + +// Copyright 2015 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package importC + +import "C" +import _ /* ERROR `cannot rename import "C"` */ "C" +import foo /* ERROR `cannot rename import "C"` */ "C" +import . /* ERROR `cannot rename import "C"` */ "C" + +// Test cases extracted from issue #22090. + +import "unsafe" + +const _ C.int = 0xff // no error due to invalid constant type + +type T struct { + Name string + Ordinal int +} + +func _(args []T) { + var s string + for i, v := range args { + cname := C.CString(v.Name) + args[i].Ordinal = int(C.sqlite3_bind_parameter_index(s, cname)) // no error due to i not being "used" + C.free(unsafe.Pointer(cname)) + } +} + +type CType C.Type + +const _ CType = C.X // no error due to invalid constant type +const _ = C.X + +// Test cases extracted from issue #23712. + +func _() { + var a [C.ArrayLength]byte + _ = a[0] // no index out of bounds error here +} + +// Additional tests to verify fix for #23712. + +func _() { + var a [C.ArrayLength1]byte + _ = 1 / len(a) // no division by zero error here and below + _ = 1 / cap(a) + _ = uint(unsafe.Sizeof(a)) // must not be negative + + var b [C.ArrayLength2]byte + a = b // should be valid +} diff --git a/src/internal/types/testdata/check/importdecl0/importdecl0a.go b/src/internal/types/testdata/check/importdecl0/importdecl0a.go new file mode 100644 index 0000000..d514ae4 --- /dev/null +++ b/src/internal/types/testdata/check/importdecl0/importdecl0a.go @@ -0,0 +1,53 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package importdecl0 + +import () + +import ( + // we can have multiple blank imports (was bug) + _ "math" + _ "net/rpc" + init /* ERROR "cannot import package as init" */ "fmt" + // reflect defines a type "flag" which shows up in the gc export data + "reflect" + . /* ERROR "imported and not used" */ "reflect" +) + +import "math" /* ERROR "imported and not used" */ +import m /* ERROR "imported as m and not used" */ "math" +import _ "math" + +import ( + "math/big" /* ERROR "imported and not used" */ + b /* ERROR "imported as b and not used" */ "math/big" + _ "math/big" +) + +import "fmt" +import f1 "fmt" +import f2 "fmt" + +// reflect.flag must not be visible in this package +type flag int +type _ reflect.flag /* ERROR "not exported" */ + +// imported package name may conflict with local objects +type reflect /* ERROR "reflect already declared" */ int + +// dot-imported exported objects may conflict with local objects +type Value /* ERROR "Value already declared through dot-import of package reflect" */ struct{} + +var _ = fmt.Println // use "fmt" + +func _() { + f1.Println() // use "fmt" +} + +func _() { + _ = func() { + f2.Println() // use "fmt" + } +} diff --git a/src/internal/types/testdata/check/importdecl0/importdecl0b.go b/src/internal/types/testdata/check/importdecl0/importdecl0b.go new file mode 100644 index 0000000..99e1d1e --- /dev/null +++ b/src/internal/types/testdata/check/importdecl0/importdecl0b.go @@ -0,0 +1,30 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package importdecl0 + +import "math" +import m "math" + +import . "testing" // declares T in file scope +import . /* ERRORx `.unsafe. imported and not used` */ "unsafe" +import . "fmt" // declares Println in file scope + +import ( + "" /* ERROR "invalid import path" */ + "a!b" /* ERROR "invalid import path" */ + "abc\xffdef" /* ERROR "invalid import path" */ +) + +// using "math" in this file doesn't affect its use in other files +const Pi0 = math.Pi +const Pi1 = m.Pi + +type _ T // use "testing" + +func _() func() interface{} { + return func() interface{} { + return Println // use "fmt" + } +} diff --git a/src/internal/types/testdata/check/importdecl1/importdecl1a.go b/src/internal/types/testdata/check/importdecl1/importdecl1a.go new file mode 100644 index 0000000..d377c01 --- /dev/null +++ b/src/internal/types/testdata/check/importdecl1/importdecl1a.go @@ -0,0 +1,22 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Test case for issue 8969. + +package importdecl1 + +import "go/ast" +import . "unsafe" + +var _ Pointer // use dot-imported package unsafe + +// Test cases for issue 23914. + +type A interface { + // Methods m1, m2 must be type-checked in this file scope + // even when embedded in an interface in a different + // file of the same package. + m1() ast.Node + m2() Pointer +} diff --git a/src/internal/types/testdata/check/importdecl1/importdecl1b.go b/src/internal/types/testdata/check/importdecl1/importdecl1b.go new file mode 100644 index 0000000..49ac2d5 --- /dev/null +++ b/src/internal/types/testdata/check/importdecl1/importdecl1b.go @@ -0,0 +1,11 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package importdecl1 + +import . /* ERRORx ".unsafe. imported and not used" */ "unsafe" + +type B interface { + A +} diff --git a/src/internal/types/testdata/check/init0.go b/src/internal/types/testdata/check/init0.go new file mode 100644 index 0000000..ee2175e --- /dev/null +++ b/src/internal/types/testdata/check/init0.go @@ -0,0 +1,106 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// initialization cycles + +package init0 + +// initialization cycles (we don't know the types) +const ( + s0 /* ERROR "initialization cycle: s0 refers to itself" */ = s0 + + x0 /* ERROR "initialization cycle for x0" */ = y0 + y0 = x0 + + a0 = b0 + b0 /* ERROR "initialization cycle for b0" */ = c0 + c0 = d0 + d0 = b0 +) + +var ( + s1 /* ERROR "initialization cycle: s1 refers to itself" */ = s1 + + x1 /* ERROR "initialization cycle for x1" */ = y1 + y1 = x1 + + a1 = b1 + b1 /* ERROR "initialization cycle for b1" */ = c1 + c1 = d1 + d1 = b1 +) + +// initialization cycles (we know the types) +const ( + s2 /* ERROR "initialization cycle: s2 refers to itself" */ int = s2 + + x2 /* ERROR "initialization cycle for x2" */ int = y2 + y2 = x2 + + a2 = b2 + b2 /* ERROR "initialization cycle for b2" */ int = c2 + c2 = d2 + d2 = b2 +) + +var ( + s3 /* ERROR "initialization cycle: s3 refers to itself" */ int = s3 + + x3 /* ERROR "initialization cycle for x3" */ int = y3 + y3 = x3 + + a3 = b3 + b3 /* ERROR "initialization cycle for b3" */ int = c3 + c3 = d3 + d3 = b3 +) + +// cycles via struct fields + +type S1 struct { + f int +} +const cx3 S1 /* ERROR "invalid constant type" */ = S1{cx3.f} +var vx3 /* ERROR "initialization cycle: vx3 refers to itself" */ S1 = S1{vx3.f} + +// cycles via functions + +var x4 = x5 +var x5 /* ERROR "initialization cycle for x5" */ = f1() +func f1() int { return x5*10 } + +var x6, x7 /* ERROR "initialization cycle" */ = f2() +var x8 = x7 +func f2() (int, int) { return f3() + f3(), 0 } +func f3() int { return x8 } + +// cycles via function literals + +var x9 /* ERROR "initialization cycle: x9 refers to itself" */ = func() int { return x9 }() + +var x10 /* ERROR "initialization cycle for x10" */ = f4() + +func f4() int { + _ = func() { + _ = x10 + } + return 0 +} + +// cycles via method expressions + +type T1 struct{} + +func (T1) m() bool { _ = x11; return false } + +var x11 /* ERROR "initialization cycle for x11" */ = T1.m(T1{}) + +// cycles via method values + +type T2 struct{} + +func (T2) m() bool { _ = x12; return false } + +var t1 T2 +var x12 /* ERROR "initialization cycle for x12" */ = t1.m diff --git a/src/internal/types/testdata/check/init1.go b/src/internal/types/testdata/check/init1.go new file mode 100644 index 0000000..c89032a --- /dev/null +++ b/src/internal/types/testdata/check/init1.go @@ -0,0 +1,97 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// initialization cycles + +package init1 + +// issue 6683 (marked as WorkingAsIntended) + +type T0 struct{} + +func (T0) m() int { return y0 } + +var x0 = T0{} + +var y0 /* ERROR "initialization cycle" */ = x0.m() + +type T1 struct{} + +func (T1) m() int { return y1 } + +var x1 interface { + m() int +} = T1{} + +var y1 = x1.m() // no cycle reported, x1 is of interface type + +// issue 6703 (modified) + +var x2 /* ERROR "initialization cycle" */ = T2.m + +var y2 = x2 + +type T2 struct{} + +func (T2) m() int { + _ = y2 + return 0 +} + +var x3 /* ERROR "initialization cycle" */ = T3.m(T3{}) // <<<< added (T3{}) + +var y3 = x3 + +type T3 struct{} + +func (T3) m() int { + _ = y3 + return 0 +} + +var x4 /* ERROR "initialization cycle" */ = T4{}.m // <<<< added {} + +var y4 = x4 + +type T4 struct{} + +func (T4) m() int { + _ = y4 + return 0 +} + +var x5 /* ERROR "initialization cycle" */ = T5{}.m() // <<<< added () + +var y5 = x5 + +type T5 struct{} + +func (T5) m() int { + _ = y5 + return 0 +} + +// issue 4847 +// simplified test case + +var x6 = f6 +var y6 /* ERROR "initialization cycle" */ = f6 +func f6() { _ = y6 } + +// full test case + +type ( + E int + S int +) + +type matcher func(s *S) E + +func matchList(s *S) E { return matcher(matchAnyFn)(s) } + +var foo = matcher(matchList) + +var matchAny /* ERROR "initialization cycle" */ = matcher(matchList) + +func matchAnyFn(s *S) (err E) { return matchAny(s) }
\ No newline at end of file diff --git a/src/internal/types/testdata/check/init2.go b/src/internal/types/testdata/check/init2.go new file mode 100644 index 0000000..24e9277 --- /dev/null +++ b/src/internal/types/testdata/check/init2.go @@ -0,0 +1,139 @@ +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// initialization cycles + +package init2 + +// cycles through functions + +func f1() int { _ = x1; return 0 } +var x1 /* ERROR "initialization cycle" */ = f1 + +func f2() int { _ = x2; return 0 } +var x2 /* ERROR "initialization cycle" */ = f2() + +// cycles through method expressions + +type T3 int +func (T3) m() int { _ = x3; return 0 } +var x3 /* ERROR "initialization cycle" */ = T3.m + +type T4 int +func (T4) m() int { _ = x4; return 0 } +var x4 /* ERROR "initialization cycle" */ = T4.m(0) + +type T3p int +func (*T3p) m() int { _ = x3p; return 0 } +var x3p /* ERROR "initialization cycle" */ = (*T3p).m + +type T4p int +func (*T4p) m() int { _ = x4p; return 0 } +var x4p /* ERROR "initialization cycle" */ = (*T4p).m(nil) + +// cycles through method expressions of embedded methods + +type T5 struct { E5 } +type E5 int +func (E5) m() int { _ = x5; return 0 } +var x5 /* ERROR "initialization cycle" */ = T5.m + +type T6 struct { E6 } +type E6 int +func (E6) m() int { _ = x6; return 0 } +var x6 /* ERROR "initialization cycle" */ = T6.m(T6{0}) + +type T5p struct { E5p } +type E5p int +func (*E5p) m() int { _ = x5p; return 0 } +var x5p /* ERROR "initialization cycle" */ = (*T5p).m + +type T6p struct { E6p } +type E6p int +func (*E6p) m() int { _ = x6p; return 0 } +var x6p /* ERROR "initialization cycle" */ = (*T6p).m(nil) + +// cycles through method values + +type T7 int +func (T7) m() int { _ = x7; return 0 } +var x7 /* ERROR "initialization cycle" */ = T7(0).m + +type T8 int +func (T8) m() int { _ = x8; return 0 } +var x8 /* ERROR "initialization cycle" */ = T8(0).m() + +type T7p int +func (*T7p) m() int { _ = x7p; return 0 } +var x7p /* ERROR "initialization cycle" */ = new(T7p).m + +type T8p int +func (*T8p) m() int { _ = x8p; return 0 } +var x8p /* ERROR "initialization cycle" */ = new(T8p).m() + +type T7v int +func (T7v) m() int { _ = x7v; return 0 } +var x7var T7v +var x7v /* ERROR "initialization cycle" */ = x7var.m + +type T8v int +func (T8v) m() int { _ = x8v; return 0 } +var x8var T8v +var x8v /* ERROR "initialization cycle" */ = x8var.m() + +type T7pv int +func (*T7pv) m() int { _ = x7pv; return 0 } +var x7pvar *T7pv +var x7pv /* ERROR "initialization cycle" */ = x7pvar.m + +type T8pv int +func (*T8pv) m() int { _ = x8pv; return 0 } +var x8pvar *T8pv +var x8pv /* ERROR "initialization cycle" */ = x8pvar.m() + +// cycles through method values of embedded methods + +type T9 struct { E9 } +type E9 int +func (E9) m() int { _ = x9; return 0 } +var x9 /* ERROR "initialization cycle" */ = T9{0}.m + +type T10 struct { E10 } +type E10 int +func (E10) m() int { _ = x10; return 0 } +var x10 /* ERROR "initialization cycle" */ = T10{0}.m() + +type T9p struct { E9p } +type E9p int +func (*E9p) m() int { _ = x9p; return 0 } +var x9p /* ERROR "initialization cycle" */ = new(T9p).m + +type T10p struct { E10p } +type E10p int +func (*E10p) m() int { _ = x10p; return 0 } +var x10p /* ERROR "initialization cycle" */ = new(T10p).m() + +type T9v struct { E9v } +type E9v int +func (E9v) m() int { _ = x9v; return 0 } +var x9var T9v +var x9v /* ERROR "initialization cycle" */ = x9var.m + +type T10v struct { E10v } +type E10v int +func (E10v) m() int { _ = x10v; return 0 } +var x10var T10v +var x10v /* ERROR "initialization cycle" */ = x10var.m() + +type T9pv struct { E9pv } +type E9pv int +func (*E9pv) m() int { _ = x9pv; return 0 } +var x9pvar *T9pv +var x9pv /* ERROR "initialization cycle" */ = x9pvar.m + +type T10pv struct { E10pv } +type E10pv int +func (*E10pv) m() int { _ = x10pv; return 0 } +var x10pvar *T10pv +var x10pv /* ERROR "initialization cycle" */ = x10pvar.m() diff --git a/src/internal/types/testdata/check/issue25008/issue25008a.go b/src/internal/types/testdata/check/issue25008/issue25008a.go new file mode 100644 index 0000000..cf71ca1 --- /dev/null +++ b/src/internal/types/testdata/check/issue25008/issue25008a.go @@ -0,0 +1,15 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "io" + +type A interface { + io.Reader +} + +func f(a A) { + a.Read(nil) +} diff --git a/src/internal/types/testdata/check/issue25008/issue25008b.go b/src/internal/types/testdata/check/issue25008/issue25008b.go new file mode 100644 index 0000000..f132b7f --- /dev/null +++ b/src/internal/types/testdata/check/issue25008/issue25008b.go @@ -0,0 +1,9 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type B interface { + A +} diff --git a/src/internal/types/testdata/check/issues0.go b/src/internal/types/testdata/check/issues0.go new file mode 100644 index 0000000..6039df9 --- /dev/null +++ b/src/internal/types/testdata/check/issues0.go @@ -0,0 +1,373 @@ +// -lang=go1.17 + +// Copyright 2014 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p // don't permit non-interface elements in interfaces + +import ( + "fmt" + syn "regexp/syntax" + t1 "text/template" + t2 "html/template" +) + +func issue7035() { + type T struct{ X int } + _ = func() { + fmt.Println() // must refer to imported fmt rather than the fmt below + } + fmt := new(T) + _ = fmt.X +} + +func issue8066() { + const ( + _ = float32(340282356779733661637539395458142568447) + _ = float32(340282356779733661637539395458142568448 /* ERROR "cannot convert" */ ) + ) +} + +// Check that a missing identifier doesn't lead to a spurious error cascade. +func issue8799a() { + x, ok := missing /* ERROR "undefined" */ () + _ = !ok + _ = x +} + +func issue8799b(x int, ok bool) { + x, ok = missing /* ERROR "undefined" */ () + _ = !ok + _ = x +} + +func issue9182() { + type Point C /* ERROR "undefined" */ .Point + // no error for composite literal based on unknown type + _ = Point{x: 1, y: 2} +} + +func f0() (a []int) { return } +func f1() (a []int, b int) { return } +func f2() (a, b []int) { return } + +func append_([]int, ...int) {} + +func issue9473(a []int, b ...int) { + // variadic builtin function + _ = append(f0()) + _ = append(f0(), f0()...) + _ = append(f1()) + _ = append(f2 /* ERRORx `cannot use .* in argument` */ ()) + _ = append(f2()... /* ERROR "cannot use ..." */ ) + _ = append(f0(), f1 /* ERROR "multiple-value f1" */ ()) + _ = append(f0(), f2 /* ERROR "multiple-value f2" */ ()) + _ = append(f0(), f1 /* ERROR "multiple-value f1" */ ()...) + _ = append(f0(), f2 /* ERROR "multiple-value f2" */ ()...) + + // variadic user-defined function + append_(f0()) + append_(f0(), f0()...) + append_(f1()) + append_(f2 /* ERRORx `cannot use .* in argument` */ ()) + append_(f2()... /* ERROR "cannot use ..." */ ) + append_(f0(), f1 /* ERROR "multiple-value f1" */ ()) + append_(f0(), f2 /* ERROR "multiple-value f2" */ ()) + append_(f0(), f1 /* ERROR "multiple-value f1" */ ()...) + append_(f0(), f2 /* ERROR "multiple-value f2" */ ()...) +} + +// Check that embedding a non-interface type in an interface results in a good error message. +func issue10979() { + type _ interface { + int /* ERROR "non-interface type int" */ + } + type T struct{} + type _ interface { + T /* ERROR "non-interface type T" */ + } + type _ interface { + nosuchtype /* ERROR "undefined: nosuchtype" */ + } + type _ interface { + fmt.Nosuchtype /* ERROR "undefined: fmt.Nosuchtype" */ + } + type _ interface { + nosuchpkg /* ERROR "undefined: nosuchpkg" */ .Nosuchtype + } + type I interface { + I.m /* ERROR "I.m is not a type" */ + m() + } +} + +// issue11347 +// These should not crash. +var a1, b1 /* ERROR "cycle" */ , c1 /* ERROR "cycle" */ b1 = 0 > 0<<""[""[c1]]>c1 +var a2, b2 /* ERROR "cycle" */ = 0 /* ERROR "assignment mismatch" */ /* ERROR "assignment mismatch" */ > 0<<""[b2] +var a3, b3 /* ERROR "cycle" */ = int /* ERROR "assignment mismatch" */ /* ERROR "assignment mismatch" */ (1<<""[b3]) + +// issue10260 +// Check that error messages explain reason for interface assignment failures. +type ( + I0 interface{} + I1 interface{ foo() } + I2 interface{ foo(x int) } + T0 struct{} + T1 struct{} + T2 struct{} +) + +func (*T1) foo() {} +func (*T2) foo(x int) {} + +func issue10260() { + var ( + i0 I0 + i1 I1 + i2 I2 + t0 *T0 + t1 *T1 + t2 *T2 + ) + + var x I1 + x = T1 /* ERRORx `cannot use T1{} .* as I1 value in assignment: T1 does not implement I1 \(method foo has pointer receiver\)` */ {} + _ = x /* ERROR "impossible type assertion: x.(T1)\n\tT1 does not implement I1 (method foo has pointer receiver)" */ .(T1) + + T1{}.foo /* ERROR "cannot call pointer method foo on T1" */ () + x.Foo /* ERROR "x.Foo undefined (type I1 has no field or method Foo, but does have foo)" */ () + + _ = i2 /* ERROR "impossible type assertion: i2.(*T1)\n\t*T1 does not implement I2 (wrong type for method foo)\n\t\thave foo()\n\t\twant foo(int)" */ .(*T1) + + i1 = i0 /* ERRORx `cannot use i0 .* as I1 value in assignment: I0 does not implement I1 \(missing method foo\)` */ + i1 = t0 /* ERRORx `.* t0 .* as I1 .*: \*T0 does not implement I1 \(missing method foo\)` */ + i1 = i2 /* ERRORx `.* i2 .* as I1 .*: I2 does not implement I1 \(wrong type for method foo\)\n\t\thave foo\(int\)\n\t\twant foo\(\)` */ + i1 = t2 /* ERRORx `.* t2 .* as I1 .*: \*T2 does not implement I1 \(wrong type for method foo\)\n\t\thave foo\(int\)\n\t\twant foo\(\)` */ + i2 = i1 /* ERRORx `.* i1 .* as I2 .*: I1 does not implement I2 \(wrong type for method foo\)\n\t\thave foo\(\)\n\t\twant foo\(int\)` */ + i2 = t1 /* ERRORx `.* t1 .* as I2 .*: \*T1 does not implement I2 \(wrong type for method foo\)\n\t\thave foo\(\)\n\t\twant foo\(int\)` */ + + _ = func() I1 { return i0 /* ERRORx `cannot use i0 .* as I1 value in return statement: I0 does not implement I1 \(missing method foo\)` */ } + _ = func() I1 { return t0 /* ERRORx `.* t0 .* as I1 .*: \*T0 does not implement I1 \(missing method foo\)` */ } + _ = func() I1 { return i2 /* ERRORx `.* i2 .* as I1 .*: I2 does not implement I1 \(wrong type for method foo\)\n\t\thave foo\(int\)\n\t\twant foo\(\)` */ } + _ = func() I1 { return t2 /* ERRORx `.* t2 .* as I1 .*: \*T2 does not implement I1 \(wrong type for method foo\)\n\t\thave foo\(int\)\n\t\twant foo\(\)` */ } + _ = func() I2 { return i1 /* ERRORx `.* i1 .* as I2 .*: I1 does not implement I2 \(wrong type for method foo\)\n\t\thave foo\(\)\n\t\twant foo\(int\)` */ } + _ = func() I2 { return t1 /* ERRORx `.* t1 .* as I2 .*: \*T1 does not implement I2 \(wrong type for method foo\)\n\t\thave foo\(\)\n\t\twant foo\(int\)` */ } + + // a few more - less exhaustive now + + f := func(I1, I2){} + f(i0 /* ERROR "missing method foo" */ , i1 /* ERROR "wrong type for method foo" */ ) + + _ = [...]I1{i0 /* ERRORx `cannot use i0 .* as I1 value in array or slice literal: I0 does not implement I1 \(missing method foo\)` */ } + _ = [...]I1{i2 /* ERRORx `cannot use i2 .* as I1 value in array or slice literal: I2 does not implement I1 \(wrong type for method foo\)\n\t\thave foo\(int\)\n\t\twant foo\(\)` */ } + _ = []I1{i0 /* ERROR "missing method foo" */ } + _ = []I1{i2 /* ERROR "wrong type for method foo" */ } + _ = map[int]I1{0: i0 /* ERROR "missing method foo" */ } + _ = map[int]I1{0: i2 /* ERROR "wrong type for method foo" */ } + + make(chan I1) <- i0 /* ERROR "missing method foo" */ + make(chan I1) <- i2 /* ERROR "wrong type for method foo" */ +} + +// Check that constants representable as integers are in integer form +// before being used in operations that are only defined on integers. +func issue14229() { + // from the issue + const _ = int64(-1<<63) % 1e6 + + // related + const ( + a int = 3 + b = 4.0 + _ = a / b + _ = a % b + _ = b / a + _ = b % a + ) +} + +// Check that in a n:1 variable declaration with type and initialization +// expression the type is distributed to all variables of the lhs before +// the initialization expression assignment is checked. +func issue15755() { + // from issue + var i interface{} + type b bool + var x, y b = i.(b) + _ = x == y + + // related: we should see an error since the result of f1 is ([]int, int) + var u, v []int = f1 /* ERROR "cannot use f1" */ () + _ = u + _ = v +} + +// Test that we don't get "declared and not used" +// errors in the context of invalid/C objects. +func issue20358() { + var F C /* ERROR "undefined" */ .F + var A C /* ERROR "undefined" */ .A + var S C /* ERROR "undefined" */ .S + type T C /* ERROR "undefined" */ .T + type P C /* ERROR "undefined" */ .P + + // these variables must be "used" even though + // the LHS expressions/types below in which + // context they are used are unknown/invalid + var f, a, s1, s2, s3, t, p int + + _ = F(f) + _ = A[a] + _ = S[s1:s2:s3] + _ = T{t} + _ = P{f: p} +} + +// Test that we don't declare lhs variables in short variable +// declarations before we type-check function literals on the +// rhs. +func issue24026() { + f := func() int { f(0) /* must refer to outer f */; return 0 } + _ = f + + _ = func() { + f := func() { _ = f() /* must refer to outer f */ } + _ = f + } + + // b and c must not be visible inside function literal + a := 0 + a, b, c := func() (int, int, int) { + return a, b /* ERROR "undefined" */ , c /* ERROR "undefined" */ + }() + _, _ = b, c +} + +func f(int) {} // for issue24026 + +// Test that we don't report a "missing return statement" error +// (due to incorrect context when type-checking interfaces). +func issue24140(x interface{}) int { + switch x.(type) { + case interface{}: + return 0 + default: + panic(0) + } +} + +// Test that we don't crash when the 'if' condition is missing. +func issue25438() { + if { /* ERROR "missing condition" */ } + if x := 0; /* ERROR "missing condition" */ { _ = x } + if + { /* ERROR "missing condition" */ } +} + +// Test that we can embed alias type names in interfaces. +type issue25301 interface { + E +} + +type E = interface { + m() +} + +// Test case from issue. +// cmd/compile reports a cycle as well. +type issue25301b /* ERROR "invalid recursive type" */ = interface { + m() interface{ issue25301b } +} + +type issue25301c interface { + notE // ERROR "non-interface type struct{}" +} + +type notE = struct{} + +// Test that method declarations don't introduce artificial cycles +// (issue #26124). +const CC TT = 1 +type TT int +func (TT) MM() [CC]TT + +// Reduced test case from issue #26124. +const preloadLimit LNumber = 128 +type LNumber float64 +func (LNumber) assertFunction() *LFunction +type LFunction struct { + GFunction LGFunction +} +type LGFunction func(*LState) +type LState struct { + reg *registry +} +type registry struct { + alloc *allocator +} +type allocator struct { + _ [int(preloadLimit)]int +} + +// Test that we don't crash when type-checking composite literals +// containing errors in the type. +var issue27346 = [][n /* ERROR "undefined" */ ]int{ + 0: {}, +} + +var issue22467 = map[int][... /* ERROR "invalid use of [...] array" */ ]int{0: {}} + +// Test that invalid use of ... in parameter lists is recognized +// (issue #28281). +func issue28281a(int, int, ...int) +func issue28281b(a, b int, c ...int) +func issue28281c(a, b, c ... /* ERROR "can only use ... with final parameter" */ int) +func issue28281d(... /* ERROR "can only use ... with final parameter" */ int, int) +func issue28281e(a, b, c ... /* ERROR "can only use ... with final parameter" */ int, d int) +func issue28281f(... /* ERROR "can only use ... with final parameter" */ int, ... /* ERROR "can only use ... with final parameter" */ int, int) +func (... /* ERROR "can only use ... with final parameter" */ TT) f() +func issue28281g() (... /* ERROR "can only use ... with final parameter" */ TT) + +// Issue #26234: Make various field/method lookup errors easier to read by matching cmd/compile's output +func issue26234a(f *syn.Prog) { + // The error message below should refer to the actual package name (syntax) + // not the local package name (syn). + f.foo /* ERROR "f.foo undefined (type *syntax.Prog has no field or method foo)" */ +} + +type T struct { + x int + E1 + E2 +} + +type E1 struct{ f int } +type E2 struct{ f int } + +func issue26234b(x T) { + _ = x.f /* ERROR "ambiguous selector x.f" */ +} + +func issue26234c() { + T.x /* ERROR "T.x undefined (type T has no method x)" */ () +} + +func issue35895() { + // T is defined in this package, don't qualify its name with the package name. + var _ T = 0 // ERROR "cannot use 0 (untyped int constant) as T" + + // There is only one package with name syntax imported, only use the (global) package name in error messages. + var _ *syn.Prog = 0 // ERROR "cannot use 0 (untyped int constant) as *syntax.Prog" + + // Because both t1 and t2 have the same global package name (template), + // qualify packages with full path name in this case. + var _ t1.Template = t2 /* ERRORx `cannot use .* \(value of type .html/template.\.Template\) as .text/template.\.Template` */ .Template{} +} + +func issue42989(s uint) { + var m map[int]string + delete(m, 1<<s) + delete(m, 1.<<s) +} diff --git a/src/internal/types/testdata/check/issues1.go b/src/internal/types/testdata/check/issues1.go new file mode 100644 index 0000000..72c6cf7 --- /dev/null +++ b/src/internal/types/testdata/check/issues1.go @@ -0,0 +1,250 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file contains regression tests for bugs found. + +package p + +import "io" +import "context" + +func eql[T comparable](x, y T) bool { + return x == y +} + +func _[X comparable, Y interface{comparable; m()}]() { + var x X + var y Y + eql(x, y /* ERROR "does not match" */ ) // interfaces of different types + eql(x, x) + eql(y, y) + eql(y, nil /* ERROR "cannot use nil as Y value in argument to eql" */ ) + eql[io.Reader](nil, nil) +} + +// If we have a receiver of pointer to type parameter type (below: *T) +// we don't have any methods, like for interfaces. +type C[T any] interface { + m() +} + +// using type bound C +func _[T C[T]](x *T) { + x.m /* ERROR "x.m undefined" */ () +} + +// using an interface literal as bound +func _[T interface{ m() }](x *T) { + x.m /* ERROR "x.m undefined" */ () +} + +func f2[_ interface{ m1(); m2() }]() {} + +type T struct{} +func (T) m1() +func (*T) m2() + +func _() { + f2[T /* ERROR "m2 has pointer receiver" */ ]() + f2[*T]() +} + +// When a type parameter is used as an argument to instantiate a parameterized +// type, the type argument's type set must be a subset of the instantiated type +// parameter's type set. +type T1[P interface{~uint}] struct{} + +func _[P any]() { + _ = T1[P /* ERROR "P does not satisfy interface{~uint}" */ ]{} +} + +// This is the original (simplified) program causing the same issue. +type Unsigned interface { + ~uint +} + +type T2[U Unsigned] struct { + s U +} + +func (u T2[U]) Add1() U { + return u.s + 1 +} + +func NewT2[U any]() T2[U /* ERROR "U does not satisfy Unsigned" */ ] { + return T2[U /* ERROR "U does not satisfy Unsigned" */ ]{} +} + +func _() { + u := NewT2[string]() + _ = u.Add1() +} + +// When we encounter an instantiated type such as Elem[T] we must +// not "expand" the instantiation when the type to be instantiated +// (Elem in this case) is not yet fully set up. +type Elem[T any] struct { + next *Elem[T] + list *List[T] +} + +type List[T any] struct { + root Elem[T] +} + +func (l *List[T]) Init() { + l.root.next = &l.root +} + +// This is the original program causing the same issue. +type Element2[TElem any] struct { + next, prev *Element2[TElem] + list *List2[TElem] + Value TElem +} + +type List2[TElem any] struct { + root Element2[TElem] + len int +} + +func (l *List2[TElem]) Init() *List2[TElem] { + l.root.next = &l.root + l.root.prev = &l.root + l.len = 0 + return l +} + +// Self-recursive instantiations must work correctly. +type A[P any] struct { _ *A[P] } + +type AB[P any] struct { _ *BA[P] } +type BA[P any] struct { _ *AB[P] } + +// And a variation that also caused a problem with an +// unresolved underlying type. +type Element3[TElem any] struct { + next, prev *Element3[TElem] + list *List3[TElem] + Value TElem +} + +func (e *Element3[TElem]) Next() *Element3[TElem] { + if p := e.next; e.list != nil && p != &e.list.root { + return p + } + return nil +} + +type List3[TElem any] struct { + root Element3[TElem] + len int +} + +// Infinite generic type declarations must lead to an error. +type inf1[T any] struct{ _ inf1 /* ERROR "invalid recursive type" */ [T] } +type inf2[T any] struct{ inf2 /* ERROR "invalid recursive type" */ [T] } + +// The implementation of conversions T(x) between integers and floating-point +// numbers checks that both T and x have either integer or floating-point +// type. When the type of T or x is a type parameter, the respective simple +// predicate disjunction in the implementation was wrong because if a type set +// contains both an integer and a floating-point type, the type parameter is +// neither an integer or a floating-point number. +func convert[T1, T2 interface{~int | ~uint | ~float32}](v T1) T2 { + return T2(v) +} + +func _() { + convert[int, uint](5) +} + +// When testing binary operators, for +, the operand types must either be +// both numeric, or both strings. The implementation had the same problem +// with this check as the conversion issue above (issue #39623). + +func issue39623[T interface{~int | ~string}](x, y T) T { + return x + y +} + +// Simplified, from https://go2goplay.golang.org/p/efS6x6s-9NI: +func Sum[T interface{~int | ~string}](s []T) (sum T) { + for _, v := range s { + sum += v + } + return +} + +// Assignability of an unnamed pointer type to a type parameter that +// has a matching underlying type. +func _[T interface{}, PT interface{~*T}] (x T) PT { + return &x +} + +// Indexing of type parameters containing type parameters in their constraint terms: +func at[T interface{ ~[]E }, E interface{}](x T, i int) E { + return x[i] +} + +// Conversion of a local type to a type parameter. +func _[T interface{~int}](x T) { + type myint int + var _ int = int(x) + var _ T = 42 + var _ T = T(myint(42)) +} + +// Indexing a type parameter with an array type bound checks length. +// (Example by mdempsky@.) +func _[T interface { ~[10]int }](x T) { + _ = x[9] // ok + _ = x[20 /* ERROR "out of bounds" */ ] +} + +// Pointer indirection of a type parameter. +func _[T interface{ ~*int }](p T) int { + return *p +} + +// Channel sends and receives on type parameters. +func _[T interface{ ~chan int }](ch T) int { + ch <- 0 + return <- ch +} + +// Calling of a generic variable. +func _[T interface{ ~func() }](f T) { + f() + go f() +} + +type F1 func() +type F2 func() +func _[T interface{ func()|F1|F2 }](f T) { + f() + go f() +} + +// We must compare against the (possibly underlying) types of term list +// elements when checking if a constraint is satisfied by a type. +// The underlying type of each term must be computed after the +// interface has been instantiated as its constraint may contain +// a type parameter that was substituted with a defined type. +// Test case from an (originally) failing example. + +type sliceOf[E any] interface{ ~[]E } + +func append[T interface{}, S sliceOf[T], T2 interface{}](s S, t ...T2) S { panic(0) } + +var f func() +var cancelSlice []context.CancelFunc +var _ = append[context.CancelFunc, []context.CancelFunc, context.CancelFunc](cancelSlice, f) + +// A generic function must be instantiated with a type, not a value. + +func g[T any](T) T { panic(0) } + +var _ = g[int] +var _ = g[nil /* ERROR "is not a type" */ ] +var _ = g(0) diff --git a/src/internal/types/testdata/check/labels.go b/src/internal/types/testdata/check/labels.go new file mode 100644 index 0000000..5948952 --- /dev/null +++ b/src/internal/types/testdata/check/labels.go @@ -0,0 +1,207 @@ +// Copyright 2011 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is a modified concatenation of the files +// $GOROOT/test/label.go and $GOROOT/test/label1.go. + +package labels + +var x int + +func f0() { +L1 /* ERROR "label L1 declared and not used" */ : + for { + } +L2 /* ERROR "label L2 declared and not used" */ : + select { + } +L3 /* ERROR "label L3 declared and not used" */ : + switch { + } +L4 /* ERROR "label L4 declared and not used" */ : + if true { + } +L5 /* ERROR "label L5 declared and not used" */ : + f0() +L6: + f0() +L6 /* ERROR "label L6 already declared" */ : + f0() + if x == 20 { + goto L6 + } + +L7: + for { + break L7 + break L8 /* ERROR "invalid break label L8" */ + } + +// A label must be directly associated with a switch, select, or +// for statement; it cannot be the label of a labeled statement. + +L7a /* ERROR "declared and not used" */ : L7b: + for { + break L7a /* ERROR "invalid break label L7a" */ + continue L7a /* ERROR "invalid continue label L7a" */ + continue L7b + } + +L8: + for { + if x == 21 { + continue L8 + continue L7 /* ERROR "invalid continue label L7" */ + } + } + +L9: + switch { + case true: + break L9 + defalt /* ERROR "label defalt declared and not used" */ : + } + +L10: + select { + default: + break L10 + break L9 /* ERROR "invalid break label L9" */ + } + + goto L10a +L10a: L10b: + select { + default: + break L10a /* ERROR "invalid break label L10a" */ + break L10b + continue L10b /* ERROR "invalid continue label L10b" */ + } +} + +func f1() { +L1: + for { + if x == 0 { + break L1 + } + if x == 1 { + continue L1 + } + goto L1 + } + +L2: + select { + default: + if x == 0 { + break L2 + } + if x == 1 { + continue L2 /* ERROR "invalid continue label L2" */ + } + goto L2 + } + +L3: + switch { + case x > 10: + if x == 11 { + break L3 + } + if x == 12 { + continue L3 /* ERROR "invalid continue label L3" */ + } + goto L3 + } + +L4: + if true { + if x == 13 { + break L4 /* ERROR "invalid break label L4" */ + } + if x == 14 { + continue L4 /* ERROR "invalid continue label L4" */ + } + if x == 15 { + goto L4 + } + } + +L5: + f1() + if x == 16 { + break L5 /* ERROR "invalid break label L5" */ + } + if x == 17 { + continue L5 /* ERROR "invalid continue label L5" */ + } + if x == 18 { + goto L5 + } + + for { + if x == 19 { + break L1 /* ERROR "invalid break label L1" */ + } + if x == 20 { + continue L1 /* ERROR "invalid continue label L1" */ + } + if x == 21 { + goto L1 + } + } +} + +// Additional tests not in the original files. + +func f2() { +L1 /* ERROR "label L1 declared and not used" */ : + if x == 0 { + for { + continue L1 /* ERROR "invalid continue label L1" */ + } + } +} + +func f3() { +L1: +L2: +L3: + for { + break L1 /* ERROR "invalid break label L1" */ + break L2 /* ERROR "invalid break label L2" */ + break L3 + continue L1 /* ERROR "invalid continue label L1" */ + continue L2 /* ERROR "invalid continue label L2" */ + continue L3 + goto L1 + goto L2 + goto L3 + } +} + +// Blank labels are never declared. + +func f4() { +_: +_: // multiple blank labels are ok + goto _ /* ERROR "label _ not declared" */ +} + +func f5() { +_: + for { + break _ /* ERROR "invalid break label _" */ + continue _ /* ERROR "invalid continue label _" */ + } +} + +func f6() { +_: + switch { + default: + break _ /* ERROR "invalid break label _" */ + } +} diff --git a/src/internal/types/testdata/check/linalg.go b/src/internal/types/testdata/check/linalg.go new file mode 100644 index 0000000..f02e773 --- /dev/null +++ b/src/internal/types/testdata/check/linalg.go @@ -0,0 +1,82 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package linalg + +// Numeric is type bound that matches any numeric type. +// It would likely be in a constraints package in the standard library. +type Numeric interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 | + ~uint | ~uint8 | ~uint16 | ~uint32 | ~uint64 | ~uintptr | + ~float32 | ~float64 | + ~complex64 | ~complex128 +} + +func DotProduct[T Numeric](s1, s2 []T) T { + if len(s1) != len(s2) { + panic("DotProduct: slices of unequal length") + } + var r T + for i := range s1 { + r += s1[i] * s2[i] + } + return r +} + +// NumericAbs matches numeric types with an Abs method. +type NumericAbs[T any] interface { + Numeric + + Abs() T +} + +// AbsDifference computes the absolute value of the difference of +// a and b, where the absolute value is determined by the Abs method. +func AbsDifference[T NumericAbs[T]](a, b T) T { + d := a - b + return d.Abs() +} + +// OrderedNumeric is a type bound that matches numeric types that support the < operator. +type OrderedNumeric interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 | + ~uint | ~uint8 | ~uint16 | ~uint32 | ~uint64 | ~uintptr | + ~float32 | ~float64 +} + +// Complex is a type bound that matches the two complex types, which do not have a < operator. +type Complex interface { + ~complex64 | ~complex128 +} + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// // OrderedAbs is a helper type that defines an Abs method for +// // ordered numeric types. +// type OrderedAbs[T OrderedNumeric] T +// +// func (a OrderedAbs[T]) Abs() OrderedAbs[T] { +// if a < 0 { +// return -a +// } +// return a +// } +// +// // ComplexAbs is a helper type that defines an Abs method for +// // complex types. +// type ComplexAbs[T Complex] T +// +// func (a ComplexAbs[T]) Abs() ComplexAbs[T] { +// r := float64(real(a)) +// i := float64(imag(a)) +// d := math.Sqrt(r * r + i * i) +// return ComplexAbs[T](complex(d, 0)) +// } +// +// func OrderedAbsDifference[T OrderedNumeric](a, b T) T { +// return T(AbsDifference(OrderedAbs[T](a), OrderedAbs[T](b))) +// } +// +// func ComplexAbsDifference[T Complex](a, b T) T { +// return T(AbsDifference(ComplexAbs[T](a), ComplexAbs[T](b))) +// } diff --git a/src/internal/types/testdata/check/literals.go b/src/internal/types/testdata/check/literals.go new file mode 100644 index 0000000..494a465 --- /dev/null +++ b/src/internal/types/testdata/check/literals.go @@ -0,0 +1,111 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file tests various representations of literals +// and compares them with literals or constant expressions +// of equal values. + +package literals + +func _() { + // 0-octals + assert(0_123 == 0123) + assert(0123_456 == 0123456) + + // decimals + assert(1_234 == 1234) + assert(1_234_567 == 1234567) + + // hexadecimals + assert(0X_0 == 0) + assert(0X_1234 == 0x1234) + assert(0X_CAFE_f00d == 0xcafef00d) + + // octals + assert(0o0 == 0) + assert(0o1234 == 01234) + assert(0o01234567 == 01234567) + + assert(0O0 == 0) + assert(0O1234 == 01234) + assert(0O01234567 == 01234567) + + assert(0o_0 == 0) + assert(0o_1234 == 01234) + assert(0o0123_4567 == 01234567) + + assert(0O_0 == 0) + assert(0O_1234 == 01234) + assert(0O0123_4567 == 01234567) + + // binaries + assert(0b0 == 0) + assert(0b1011 == 0xb) + assert(0b00101101 == 0x2d) + + assert(0B0 == 0) + assert(0B1011 == 0xb) + assert(0B00101101 == 0x2d) + + assert(0b_0 == 0) + assert(0b10_11 == 0xb) + assert(0b_0010_1101 == 0x2d) + + // decimal floats + assert(1_2_3. == 123.) + assert(0_123. == 123.) + + assert(0_0e0 == 0.) + assert(1_2_3e0 == 123.) + assert(0_123e0 == 123.) + + assert(0e-0_0 == 0.) + assert(1_2_3E+0 == 123.) + assert(0123E1_2_3 == 123e123) + + assert(0.e+1 == 0.) + assert(123.E-1_0 == 123e-10) + assert(01_23.e123 == 123e123) + + assert(.0e-1 == .0) + assert(.123E+10 == .123e10) + assert(.0123E123 == .0123e123) + + assert(1_2_3.123 == 123.123) + assert(0123.01_23 == 123.0123) + + // hexadecimal floats + assert(0x0.p+0 == 0.) + assert(0Xdeadcafe.p-10 == 0xdeadcafe/1024.0) + assert(0x1234.P84 == 0x1234000000000000000000000) + + assert(0x.1p-0 == 1./16) + assert(0X.deadcafep4 == 1.0*0xdeadcafe/0x10000000) + assert(0x.1234P+12 == 1.0*0x1234/0x10) + + assert(0x0p0 == 0.) + assert(0Xdeadcafep+1 == 0x1bd5b95fc) + assert(0x1234P-10 == 0x1234/1024.0) + + assert(0x0.0p0 == 0.) + assert(0Xdead.cafep+1 == 1.0*0x1bd5b95fc/0x10000) + assert(0x12.34P-10 == 1.0*0x1234/0x40000) + + assert(0Xdead_cafep+1 == 0xdeadcafep+1) + assert(0x_1234P-10 == 0x1234p-10) + + assert(0X_dead_cafe.p-10 == 0xdeadcafe.p-10) + assert(0x12_34.P1_2_3 == 0x1234.p123) + + assert(1_234i == 1234i) + assert(1_234_567i == 1234567i) + + assert(0.i == 0i) + assert(123.i == 123i) + assert(0123.i == 123i) + + assert(0.e+1i == 0i) + assert(123.E-1_0i == 123e-10i) + assert(01_23.e123i == 123e123i) +} diff --git a/src/internal/types/testdata/check/main0.go b/src/internal/types/testdata/check/main0.go new file mode 100644 index 0000000..95a8ed1 --- /dev/null +++ b/src/internal/types/testdata/check/main0.go @@ -0,0 +1,9 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +func main() +func main /* ERROR "no arguments and no return values" */ /* ERROR "redeclared" */ (int) +func main /* ERROR "no arguments and no return values" */ /* ERROR "redeclared" */ () int diff --git a/src/internal/types/testdata/check/main1.go b/src/internal/types/testdata/check/main1.go new file mode 100644 index 0000000..fb567a0 --- /dev/null +++ b/src/internal/types/testdata/check/main1.go @@ -0,0 +1,7 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +func main[T /* ERROR "func main must have no type parameters" */ any]() {} diff --git a/src/internal/types/testdata/check/map0.go b/src/internal/types/testdata/check/map0.go new file mode 100644 index 0000000..21c989c --- /dev/null +++ b/src/internal/types/testdata/check/map0.go @@ -0,0 +1,113 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package orderedmap provides an ordered map, implemented as a binary tree. +package orderedmap + +// TODO(gri) fix imports for tests +import "chans" // ERROR "could not import" + +// Map is an ordered map. +type Map[K, V any] struct { + root *node[K, V] + compare func(K, K) int +} + +// node is the type of a node in the binary tree. +type node[K, V any] struct { + key K + val V + left, right *node[K, V] +} + +// New returns a new map. +func New[K, V any](compare func(K, K) int) *Map[K, V] { + return &Map[K, V]{compare: compare} +} + +// find looks up key in the map, and returns either a pointer +// to the node holding key, or a pointer to the location where +// such a node would go. +func (m *Map[K, V]) find(key K) **node[K, V] { + pn := &m.root + for *pn != nil { + switch cmp := m.compare(key, (*pn).key); { + case cmp < 0: + pn = &(*pn).left + case cmp > 0: + pn = &(*pn).right + default: + return pn + } + } + return pn +} + +// Insert inserts a new key/value into the map. +// If the key is already present, the value is replaced. +// Returns true if this is a new key, false if already present. +func (m *Map[K, V]) Insert(key K, val V) bool { + pn := m.find(key) + if *pn != nil { + (*pn).val = val + return false + } + *pn = &node[K, V]{key: key, val: val} + return true +} + +// Find returns the value associated with a key, or zero if not present. +// The found result reports whether the key was found. +func (m *Map[K, V]) Find(key K) (V, bool) { + pn := m.find(key) + if *pn == nil { + var zero V // see the discussion of zero values, above + return zero, false + } + return (*pn).val, true +} + +// keyValue is a pair of key and value used when iterating. +type keyValue[K, V any] struct { + key K + val V +} + +// InOrder returns an iterator that does an in-order traversal of the map. +func (m *Map[K, V]) InOrder() *Iterator[K, V] { + sender, receiver := chans.Ranger[keyValue[K, V]]() + var f func(*node[K, V]) bool + f = func(n *node[K, V]) bool { + if n == nil { + return true + } + // Stop sending values if sender.Send returns false, + // meaning that nothing is listening at the receiver end. + return f(n.left) && + sender.Send(keyValue[K, V]{n.key, n.val}) && + f(n.right) + } + go func() { + f(m.root) + sender.Close() + }() + return &Iterator[K, V]{receiver} +} + +// Iterator is used to iterate over the map. +type Iterator[K, V any] struct { + r *chans.Receiver[keyValue[K, V]] +} + +// Next returns the next key and value pair, and a boolean indicating +// whether they are valid or whether we have reached the end. +func (it *Iterator[K, V]) Next() (K, V, bool) { + keyval, ok := it.r.Next() + if !ok { + var zerok K + var zerov V + return zerok, zerov, false + } + return keyval.key, keyval.val, true +} diff --git a/src/internal/types/testdata/check/map1.go b/src/internal/types/testdata/check/map1.go new file mode 100644 index 0000000..e13bf33 --- /dev/null +++ b/src/internal/types/testdata/check/map1.go @@ -0,0 +1,146 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is like map.go2, but instead if importing chans, it contains +// the necessary functionality at the end of the file. + +// Package orderedmap provides an ordered map, implemented as a binary tree. +package orderedmap + +// Map is an ordered map. +type Map[K, V any] struct { + root *node[K, V] + compare func(K, K) int +} + +// node is the type of a node in the binary tree. +type node[K, V any] struct { + key K + val V + left, right *node[K, V] +} + +// New returns a new map. +func New[K, V any](compare func(K, K) int) *Map[K, V] { + return &Map[K, V]{compare: compare} +} + +// find looks up key in the map, and returns either a pointer +// to the node holding key, or a pointer to the location where +// such a node would go. +func (m *Map[K, V]) find(key K) **node[K, V] { + pn := &m.root + for *pn != nil { + switch cmp := m.compare(key, (*pn).key); { + case cmp < 0: + pn = &(*pn).left + case cmp > 0: + pn = &(*pn).right + default: + return pn + } + } + return pn +} + +// Insert inserts a new key/value into the map. +// If the key is already present, the value is replaced. +// Returns true if this is a new key, false if already present. +func (m *Map[K, V]) Insert(key K, val V) bool { + pn := m.find(key) + if *pn != nil { + (*pn).val = val + return false + } + *pn = &node[K, V]{key: key, val: val} + return true +} + +// Find returns the value associated with a key, or zero if not present. +// The found result reports whether the key was found. +func (m *Map[K, V]) Find(key K) (V, bool) { + pn := m.find(key) + if *pn == nil { + var zero V // see the discussion of zero values, above + return zero, false + } + return (*pn).val, true +} + +// keyValue is a pair of key and value used when iterating. +type keyValue[K, V any] struct { + key K + val V +} + +// InOrder returns an iterator that does an in-order traversal of the map. +func (m *Map[K, V]) InOrder() *Iterator[K, V] { + sender, receiver := chans_Ranger[keyValue[K, V]]() + var f func(*node[K, V]) bool + f = func(n *node[K, V]) bool { + if n == nil { + return true + } + // Stop sending values if sender.Send returns false, + // meaning that nothing is listening at the receiver end. + return f(n.left) && + sender.Send(keyValue[K, V]{n.key, n.val}) && + f(n.right) + } + go func() { + f(m.root) + sender.Close() + }() + return &Iterator[K, V]{receiver} +} + +// Iterator is used to iterate over the map. +type Iterator[K, V any] struct { + r *chans_Receiver[keyValue[K, V]] +} + +// Next returns the next key and value pair, and a boolean indicating +// whether they are valid or whether we have reached the end. +func (it *Iterator[K, V]) Next() (K, V, bool) { + keyval, ok := it.r.Next() + if !ok { + var zerok K + var zerov V + return zerok, zerov, false + } + return keyval.key, keyval.val, true +} + +// chans + +func chans_Ranger[T any]() (*chans_Sender[T], *chans_Receiver[T]) { panic(0) } + +// A sender is used to send values to a Receiver. +type chans_Sender[T any] struct { + values chan<- T + done <-chan bool +} + +func (s *chans_Sender[T]) Send(v T) bool { + select { + case s.values <- v: + return true + case <-s.done: + return false + } +} + +func (s *chans_Sender[T]) Close() { + close(s.values) +} + +type chans_Receiver[T any] struct { + values <-chan T + done chan<- bool +} + +func (r *chans_Receiver[T]) Next() (T, bool) { + v, ok := <-r.values + return v, ok +} diff --git a/src/internal/types/testdata/check/methodsets.go b/src/internal/types/testdata/check/methodsets.go new file mode 100644 index 0000000..5b3e4a2 --- /dev/null +++ b/src/internal/types/testdata/check/methodsets.go @@ -0,0 +1,214 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package methodsets + +type T0 struct {} + +func (T0) v0() {} +func (*T0) p0() {} + +type T1 struct {} // like T0 with different method names + +func (T1) v1() {} +func (*T1) p1() {} + +type T2 interface { + v2() + p2() +} + +type T3 struct { + T0 + *T1 + T2 +} + +// Method expressions +func _() { + var ( + _ func(T0) = T0.v0 + _ = T0.p0 /* ERROR "invalid method expression T0.p0 (needs pointer receiver (*T0).p0)" */ + + _ func (*T0) = (*T0).v0 + _ func (*T0) = (*T0).p0 + + // T1 is like T0 + + _ func(T2) = T2.v2 + _ func(T2) = T2.p2 + + _ func(T3) = T3.v0 + _ func(T3) = T3.p0 /* ERROR "invalid method expression T3.p0 (needs pointer receiver (*T3).p0)" */ + _ func(T3) = T3.v1 + _ func(T3) = T3.p1 + _ func(T3) = T3.v2 + _ func(T3) = T3.p2 + + _ func(*T3) = (*T3).v0 + _ func(*T3) = (*T3).p0 + _ func(*T3) = (*T3).v1 + _ func(*T3) = (*T3).p1 + _ func(*T3) = (*T3).v2 + _ func(*T3) = (*T3).p2 + ) +} + +// Method values with addressable receivers +func _() { + var ( + v0 T0 + _ func() = v0.v0 + _ func() = v0.p0 + ) + + var ( + p0 *T0 + _ func() = p0.v0 + _ func() = p0.p0 + ) + + // T1 is like T0 + + var ( + v2 T2 + _ func() = v2.v2 + _ func() = v2.p2 + ) + + var ( + v4 T3 + _ func() = v4.v0 + _ func() = v4.p0 + _ func() = v4.v1 + _ func() = v4.p1 + _ func() = v4.v2 + _ func() = v4.p2 + ) + + var ( + p4 *T3 + _ func() = p4.v0 + _ func() = p4.p0 + _ func() = p4.v1 + _ func() = p4.p1 + _ func() = p4.v2 + _ func() = p4.p2 + ) +} + +// Method calls with addressable receivers +func _() { + var v0 T0 + v0.v0() + v0.p0() + + var p0 *T0 + p0.v0() + p0.p0() + + // T1 is like T0 + + var v2 T2 + v2.v2() + v2.p2() + + var v4 T3 + v4.v0() + v4.p0() + v4.v1() + v4.p1() + v4.v2() + v4.p2() + + var p4 *T3 + p4.v0() + p4.p0() + p4.v1() + p4.p1() + p4.v2() + p4.p2() +} + +// Method values with value receivers +func _() { + var ( + _ func() = T0{}.v0 + _ func() = T0{}.p0 /* ERROR "cannot call pointer method p0 on T0" */ + + _ func() = (&T0{}).v0 + _ func() = (&T0{}).p0 + + // T1 is like T0 + + // no values for T2 + + _ func() = T3{}.v0 + _ func() = T3{}.p0 /* ERROR "cannot call pointer method p0 on T3" */ + _ func() = T3{}.v1 + _ func() = T3{}.p1 + _ func() = T3{}.v2 + _ func() = T3{}.p2 + + _ func() = (&T3{}).v0 + _ func() = (&T3{}).p0 + _ func() = (&T3{}).v1 + _ func() = (&T3{}).p1 + _ func() = (&T3{}).v2 + _ func() = (&T3{}).p2 + ) +} + +// Method calls with value receivers +func _() { + T0{}.v0() + T0{}.p0 /* ERROR "cannot call pointer method p0 on T0" */ () + + (&T0{}).v0() + (&T0{}).p0() + + // T1 is like T0 + + // no values for T2 + + T3{}.v0() + T3{}.p0 /* ERROR "cannot call pointer method p0 on T3" */ () + T3{}.v1() + T3{}.p1() + T3{}.v2() + T3{}.p2() + + (&T3{}).v0() + (&T3{}).p0() + (&T3{}).v1() + (&T3{}).p1() + (&T3{}).v2() + (&T3{}).p2() +} + +// *T has no methods if T is an interface type +func issue5918() { + var ( + err error + _ = err.Error() + _ func() string = err.Error + _ func(error) string = error.Error + + perr = &err + _ = perr.Error /* ERROR "type *error is pointer to interface, not interface" */ () + _ func() string = perr.Error /* ERROR "type *error is pointer to interface, not interface" */ + _ func(*error) string = (*error).Error /* ERROR "type *error is pointer to interface, not interface" */ + ) + + type T *interface{ m() int } + var ( + x T + _ = (*x).m() + _ = (*x).m + + _ = x.m /* ERROR "type T is pointer to interface, not interface" */ () + _ = x.m /* ERROR "type T is pointer to interface, not interface" */ + _ = T.m /* ERROR "type T is pointer to interface, not interface" */ + ) +} diff --git a/src/internal/types/testdata/check/shifts.go b/src/internal/types/testdata/check/shifts.go new file mode 100644 index 0000000..6ae3985 --- /dev/null +++ b/src/internal/types/testdata/check/shifts.go @@ -0,0 +1,399 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package shifts + +func shifts0() { + // basic constant shifts + const ( + s = 10 + _ = 0<<0 + _ = 1<<s + _ = 1<<- /* ERROR "negative shift count" */ 1 + // For the test below we may decide to convert to int + // rather than uint and then report a negative shift + // count instead, which might be a better error. The + // (minor) difference is that this would restrict the + // shift count range by half (from all uint values to + // the positive int values). + // This depends on the exact spec wording which is not + // done yet. + // TODO(gri) revisit and adjust when spec change is done + _ = 1<<- /* ERROR "negative shift count" */ 1.0 + _ = 1<<1075 /* ERROR "invalid shift" */ + _ = 2.0<<1 + _ = 1<<1.0 + _ = 1<<(1+0i) + + _ int = 2<<s + _ float32 = 2<<s + _ complex64 = 2<<s + + _ int = 2.0<<s + _ float32 = 2.0<<s + _ complex64 = 2.0<<s + + _ int = 'a'<<s + _ float32 = 'a'<<s + _ complex64 = 'a'<<s + ) +} + +func shifts1() { + // basic non-constant shifts + var ( + i int + u uint + + _ = 1<<0 + _ = 1<<i + _ = 1<<u + _ = 1<<"foo" /* ERROR "cannot convert" */ + _ = i<<0 + _ = i<<- /* ERROR "negative shift count" */ 1 + _ = i<<1.0 + _ = 1<<(1+0i) + _ = 1 /* ERROR "overflows" */ <<100 + + _ uint = 1 << 0 + _ uint = 1 << u + _ float32 = 1 /* ERROR "must be integer" */ << u + + // issue #14822 + _ = 1<<( /* ERROR "overflows uint" */ 1<<64) + _ = 1<<( /* ERROR "invalid shift count" */ 1<<64-1) + + // issue #43697 + _ = u<<( /* ERROR "overflows uint" */ 1<<64) + _ = u<<(1<<64-1) + ) +} + +func shifts2() { + // from the spec + var ( + s uint = 33 + i = 1<<s // 1 has type int + j int32 = 1<<s // 1 has type int32; j == 0 + k = uint64(1<<s) // 1 has type uint64; k == 1<<33 + m int = 1.0<<s // 1.0 has type int + n = 1.0<<s != i // 1.0 has type int; n == false if ints are 32bits in size + o = 1<<s == 2<<s // 1 and 2 have type int; o == true if ints are 32bits in size + p = 1<<s == 1<<33 // illegal if ints are 32bits in size: 1 has type int, but 1<<33 overflows int + u = 1.0 /* ERROR "must be integer" */ <<s // illegal: 1.0 has type float64, cannot shift + u1 = 1.0 /* ERROR "must be integer" */ <<s != 0 // illegal: 1.0 has type float64, cannot shift + u2 = 1 /* ERROR "must be integer" */ <<s != 1.0 // illegal: 1 has type float64, cannot shift + v float32 = 1 /* ERROR "must be integer" */ <<s // illegal: 1 has type float32, cannot shift + w int64 = 1.0<<33 // 1.0<<33 is a constant shift expression + ) + _, _, _, _, _, _, _, _, _, _, _, _ = i, j, k, m, n, o, p, u, u1, u2, v, w +} + +func shifts3(a int16, b float32) { + // random tests + var ( + s uint = 11 + u = 1 /* ERROR "must be integer" */ <<s + 1.0 + v complex128 = 1 /* ERROR "must be integer" */ << s + 1.0 /* ERROR "must be integer" */ << s + 1 + ) + x := 1.0 /* ERROR "must be integer" */ <<s + 1 + shifts3(1.0 << s, 1 /* ERROR "must be integer" */ >> s) + _, _, _ = u, v, x +} + +func shifts4() { + // shifts in comparisons w/ untyped operands + var s uint + + _ = 1<<s == 1 + _ = 1 /* ERROR "integer" */ <<s == 1. + _ = 1. /* ERROR "integer" */ <<s == 1 + _ = 1. /* ERROR "integer" */ <<s == 1. + + _ = 1<<s + 1 == 1 + _ = 1 /* ERROR "integer" */ <<s + 1 == 1. + _ = 1 /* ERROR "integer" */ <<s + 1. == 1 + _ = 1 /* ERROR "integer" */ <<s + 1. == 1. + _ = 1. /* ERROR "integer" */ <<s + 1 == 1 + _ = 1. /* ERROR "integer" */ <<s + 1 == 1. + _ = 1. /* ERROR "integer" */ <<s + 1. == 1 + _ = 1. /* ERROR "integer" */ <<s + 1. == 1. + + _ = 1<<s == 1<<s + _ = 1 /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + + _ = 1<<s + 1<<s == 1 + _ = 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1 + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. + + _ = 1<<s + 1<<s == 1<<s + 1<<s + _ = 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1 /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1 /* ERROR "integer" */ <<s + _ = 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s == 1. /* ERROR "integer" */ <<s + 1. /* ERROR "integer" */ <<s +} + +func shifts5() { + // shifts in comparisons w/ typed operands + var s uint + var x int + + _ = 1<<s == x + _ = 1.<<s == x + _ = 1.1 /* ERROR "int" */ <<s == x + + _ = 1<<s + x == 1 + _ = 1<<s + x == 1. + _ = 1<<s + x == 1.1 /* ERROR "int" */ + _ = 1.<<s + x == 1 + _ = 1.<<s + x == 1. + _ = 1.<<s + x == 1.1 /* ERROR "int" */ + _ = 1.1 /* ERROR "int" */ <<s + x == 1 + _ = 1.1 /* ERROR "int" */ <<s + x == 1. + _ = 1.1 /* ERROR "int" */ <<s + x == 1.1 + + _ = 1<<s == x<<s + _ = 1.<<s == x<<s + _ = 1.1 /* ERROR "int" */ <<s == x<<s +} + +func shifts6() { + // shifts as operands in non-arithmetic operations and as arguments + var a [10]int + var s uint + + _ = a[1<<s] + _ = a[1.0] + _ = a[1.0<<s] + + _ = make([]int, 1.0) + _ = make([]int, 1.0<<s) + _ = make([]int, 1.1 /* ERROR "must be integer" */ <<s) + + _ = float32(1) + _ = float32(1 /* ERROR "must be integer" */ <<s) + _ = float32(1.0) + _ = float32(1.0 /* ERROR "must be integer" */ <<s) + _ = float32(1.1 /* ERROR "must be integer" */ <<s) + + // TODO(gri) Re-enable these tests once types2 has the go/types fixes. + // Issue #52080. + // _ = int32(0x80000000 /* ERROR "overflows int32" */ << s) + // TODO(rfindley) Eliminate the redundant error here. + // _ = int32(( /* ERROR "truncated to int32" */ 0x80000000 /* ERROR "truncated to int32" */ + 0i) << s) + + _ = int(1+0i<<0) + // _ = int((1+0i)<<s) + // _ = int(1.0<<s) + // _ = int(complex(1, 0)<<s) + _ = int(float32/* ERROR "must be integer" */(1.0) <<s) + _ = int(1.1 /* ERROR "must be integer" */ <<s) + _ = int(( /* ERROR "must be integer" */ 1+1i) <<s) + + _ = complex(1 /* ERROR "must be integer" */ <<s, 0) + + var b []int + _ = append(b, 1<<s) + _ = append(b, 1.0<<s) + _ = append(b, (1+0i)<<s) + _ = append(b, 1.1 /* ERROR "must be integer" */ <<s) + _ = append(b, (1 + 0i) <<s) + _ = append(b, ( /* ERROR "must be integer" */ 1 + 1i) <<s) + + _ = complex(1.0 /* ERROR "must be integer" */ <<s, 0) + _ = complex(1.1 /* ERROR "must be integer" */ <<s, 0) + _ = complex(0, 1.0 /* ERROR "must be integer" */ <<s) + _ = complex(0, 1.1 /* ERROR "must be integer" */ <<s) + + // TODO(gri) The delete below is not type-checked correctly yet. + // var m1 map[int]string + // delete(m1, 1<<s) +} + +func shifts7() { + // shifts of shifts + var s uint + var x int + _ = x + + _ = 1<<(1<<s) + _ = 1<<(1.<<s) + _ = 1. /* ERROR "integer" */ <<(1<<s) + _ = 1. /* ERROR "integer" */ <<(1.<<s) + + x = 1<<(1<<s) + x = 1<<(1.<<s) + x = 1.<<(1<<s) + x = 1.<<(1.<<s) + + _ = (1<<s)<<(1<<s) + _ = (1<<s)<<(1.<<s) + _ = ( /* ERROR "integer" */ 1.<<s)<<(1<<s) + _ = ( /* ERROR "integer" */ 1.<<s)<<(1.<<s) + + x = (1<<s)<<(1<<s) + x = (1<<s)<<(1.<<s) + x = ( /* ERROR "integer" */ 1.<<s)<<(1<<s) + x = ( /* ERROR "integer" */ 1.<<s)<<(1.<<s) +} + +func shifts8() { + // shift examples from shift discussion: better error messages + var s uint + _ = 1.0 /* ERROR "shifted operand 1.0 (type float64) must be integer" */ <<s == 1 + _ = 1.0 /* ERROR "shifted operand 1.0 (type float64) must be integer" */ <<s == 1.0 + _ = 1 /* ERROR "shifted operand 1 (type float64) must be integer" */ <<s == 1.0 + _ = 1 /* ERROR "shifted operand 1 (type float64) must be integer" */ <<s + 1.0 == 1 + _ = 1 /* ERROR "shifted operand 1 (type float64) must be integer" */ <<s + 1.1 == 1 + _ = 1 /* ERROR "shifted operand 1 (type float64) must be integer" */ <<s + 1 == 1.0 + + // additional cases + _ = complex(1.0 /* ERROR "shifted operand 1.0 (type float64) must be integer" */ <<s, 1) + _ = complex(1.0, 1 /* ERROR "shifted operand 1 (type float64) must be integer" */ <<s) + + _ = int(1.<<s) + _ = int(1.1 /* ERRORx `shifted operand .* must be integer` */ <<s) + _ = float32(1 /* ERRORx `shifted operand .* must be integer` */ <<s) + _ = float32(1. /* ERRORx `shifted operand .* must be integer` */ <<s) + _ = float32(1.1 /* ERRORx `shifted operand .* must be integer` */ <<s) + // TODO(gri) the error messages for these two are incorrect - disabled for now + // _ = complex64(1<<s) + // _ = complex64(1.<<s) + _ = complex64(1.1 /* ERRORx `shifted operand .* must be integer` */ <<s) +} + +func shifts9() { + // various originally failing snippets of code from the std library + // from src/compress/lzw/reader.go:90 + { + var d struct { + bits uint32 + width uint + } + _ = uint16(d.bits & (1<<d.width - 1)) + } + + // from src/debug/dwarf/buf.go:116 + { + var ux uint64 + var bits uint + x := int64(ux) + if x&(1<<(bits-1)) != 0 {} + } + + // from src/encoding/asn1/asn1.go:160 + { + var bytes []byte + if bytes[len(bytes)-1]&((1<<bytes[0])-1) != 0 {} + } + + // from src/math/big/rat.go:140 + { + var exp int + var mantissa uint64 + shift := uint64(-1022 - (exp - 1)) // [1..53) + _ = mantissa & (1<<shift - 1) + } + + // from src/net/interface.go:51 + { + type Flags uint + var f Flags + var i int + if f&(1<<uint(i)) != 0 {} + } + + // from src/runtime/softfloat64.go:234 + { + var gm uint64 + var shift uint + _ = gm & (1<<shift - 1) + } + + // from src/strconv/atof.go:326 + { + var mant uint64 + var mantbits uint + if mant == 2<<mantbits {} + } + + // from src/route_bsd.go:82 + { + var Addrs int32 + const rtaRtMask = 1 + var i uint + if Addrs&rtaRtMask&(1<<i) == 0 {} + } + + // from src/text/scanner/scanner.go:540 + { + var s struct { Whitespace uint64 } + var ch rune + for s.Whitespace&(1<<uint(ch)) != 0 {} + } +} + +func issue5895() { + var x = 'a' << 1 // type of x must be rune + var _ rune = x +} + +func issue11325() { + var _ = 0 >> 1.1 /* ERROR "truncated to uint" */ // example from issue 11325 + _ = 0 >> 1.1 /* ERROR "truncated to uint" */ + _ = 0 << 1.1 /* ERROR "truncated to uint" */ + _ = 0 >> 1. + _ = 1 >> 1.1 /* ERROR "truncated to uint" */ + _ = 1 >> 1. + _ = 1. >> 1 + _ = 1. >> 1. + _ = 1.1 /* ERROR "must be integer" */ >> 1 +} + +func issue11594() { + var _ = complex64 /* ERROR "must be integer" */ (1) << 2 // example from issue 11594 + _ = float32 /* ERROR "must be integer" */ (0) << 1 + _ = float64 /* ERROR "must be integer" */ (0) >> 2 + _ = complex64 /* ERROR "must be integer" */ (0) << 3 + _ = complex64 /* ERROR "must be integer" */ (0) >> 4 +} + +func issue21727() { + var s uint + var a = make([]int, 1<<s + 1.2 /* ERROR "truncated to int" */ ) + var _ = a[1<<s - 2.3 /* ERROR "truncated to int" */ ] + var _ int = 1<<s + 3.4 /* ERROR "truncated to int" */ + var _ = string(1 /* ERRORx `shifted operand 1 .* must be integer` */ << s) + var _ = string(1.0 /* ERROR "cannot convert" */ << s) +} + +func issue22969() { + var s uint + var a []byte + _ = a[0xffffffffffffffff /* ERROR "overflows int" */ <<s] // example from issue 22969 + _ = make([]int, 0xffffffffffffffff /* ERROR "overflows int" */ << s) + _ = make([]int, 0, 0xffffffffffffffff /* ERROR "overflows int" */ << s) + var _ byte = 0x100 /* ERROR "overflows byte" */ << s + var _ int8 = 0xff /* ERROR "overflows int8" */ << s + var _ int16 = 0xffff /* ERROR "overflows int16" */ << s + var _ int32 = 0x80000000 /* ERROR "overflows int32" */ << s +} diff --git a/src/internal/types/testdata/check/slices.go b/src/internal/types/testdata/check/slices.go new file mode 100644 index 0000000..2c33518 --- /dev/null +++ b/src/internal/types/testdata/check/slices.go @@ -0,0 +1,68 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package slices implements various slice algorithms. +package slices + +// Map turns a []T1 to a []T2 using a mapping function. +func Map[T1, T2 any](s []T1, f func(T1) T2) []T2 { + r := make([]T2, len(s)) + for i, v := range s { + r[i] = f(v) + } + return r +} + +// Reduce reduces a []T1 to a single value using a reduction function. +func Reduce[T1, T2 any](s []T1, initializer T2, f func(T2, T1) T2) T2 { + r := initializer + for _, v := range s { + r = f(r, v) + } + return r +} + +// Filter filters values from a slice using a filter function. +func Filter[T any](s []T, f func(T) bool) []T { + var r []T + for _, v := range s { + if f(v) { + r = append(r, v) + } + } + return r +} + +// Example uses + +func limiter(x int) byte { + switch { + case x < 0: + return 0 + default: + return byte(x) + case x > 255: + return 255 + } +} + +var input = []int{-4, 68954, 7, 44, 0, -555, 6945} +var limited1 = Map[int, byte](input, limiter) +var limited2 = Map(input, limiter) // using type inference + +func reducer(x float64, y int) float64 { + return x + float64(y) +} + +var reduced1 = Reduce[int, float64](input, 0, reducer) +var reduced2 = Reduce(input, 1i /* ERROR "overflows" */, reducer) // using type inference +var reduced3 = Reduce(input, 1, reducer) // using type inference + +func filter(x int) bool { + return x&1 != 0 +} + +var filtered1 = Filter[int](input, filter) +var filtered2 = Filter(input, filter) // using type inference + diff --git a/src/internal/types/testdata/check/stmt0.go b/src/internal/types/testdata/check/stmt0.go new file mode 100644 index 0000000..5232285 --- /dev/null +++ b/src/internal/types/testdata/check/stmt0.go @@ -0,0 +1,994 @@ +// Copyright 2012 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// statements + +package stmt0 + +func assignments0() (int, int) { + var a, b, c int + var ch chan int + f0 := func() {} + f1 := func() int { return 1 } + f2 := func() (int, int) { return 1, 2 } + f3 := func() (int, int, int) { return 1, 2, 3 } + + a, b, c = 1, 2, 3 + a, b, c = 1 /* ERROR "assignment mismatch: 3 variables but 2 values" */ , 2 + a, b, c = 1 /* ERROR "assignment mismatch: 3 variables but 4 values" */ , 2, 3, 4 + _, _, _ = a, b, c + + a = f0 /* ERROR "used as value" */ () + a = f1() + a = f2 /* ERROR "assignment mismatch: 1 variable but f2 returns 2 values" */ () + a, b = f2() + a, b, c = f2 /* ERROR "assignment mismatch: 3 variables but f2 returns 2 values" */ () + a, b, c = f3() + a, b = f3 /* ERROR "assignment mismatch: 2 variables but f3 returns 3 values" */ () + + a, b, c = <- /* ERROR "assignment mismatch: 3 variables but 1 value" */ ch + + return /* ERROR "not enough return values\n\thave ()\n\twant (int, int)" */ + return 1 /* ERROR "not enough return values\n\thave (number)\n\twant (int, int)" */ + return 1, 2 + return 1, 2, 3 /* ERROR "too many return values\n\thave (number, number, number)\n\twant (int, int)" */ +} + +func assignments1() { + b, i, f, c, s := false, 1, 1.0, 1i, "foo" + b = i /* ERRORx `cannot use .* in assignment` */ + i = f /* ERRORx `cannot use .* in assignment` */ + f = c /* ERRORx `cannot use .* in assignment` */ + c = s /* ERRORx `cannot use .* in assignment` */ + s = b /* ERRORx `cannot use .* in assignment` */ + + v0, v1, v2 := 1 /* ERROR "assignment mismatch" */ , 2, 3, 4 + _, _, _ = v0, v1, v2 + + b = true + + i += 1 + i /* ERROR "mismatched types int and untyped string" */+= "foo" + + f -= 1 + f /= 0 + f = float32(0)/0 /* ERROR "division by zero" */ + f /* ERROR "mismatched types float64 and untyped string" */-= "foo" + + c *= 1 + c /= 0 + + s += "bar" + s /* ERROR "mismatched types string and untyped int" */+= 1 + + var u64 uint64 + u64 += 1<<u64 + + undefined /* ERROR "undefined" */ = 991 + + // test cases for issue 5800 + var ( + _ int = nil /* ERROR "cannot use nil as int value in variable declaration" */ + _ [10]int = nil /* ERROR "cannot use nil as [10]int value in variable declaration" */ + _ []byte = nil + _ struct{} = nil /* ERROR "cannot use nil as struct{} value in variable declaration" */ + _ func() = nil + _ map[int]string = nil + _ chan int = nil + ) + + // test cases for issue 5500 + _ = func() (int, bool) { + var m map[int]int + return m /* ERROR "not enough return values" */ [0] + } + + g := func(int, bool){} + var m map[int]int + g(m[0]) /* ERROR "not enough arguments" */ + + // assignments to _ + _ = nil /* ERROR "use of untyped nil" */ + _ = 1 << /* ERROR "constant shift overflow" */ 1000 + (_) = 0 +} + +func assignments2() { + type mybool bool + var m map[string][]bool + var s []bool + var b bool + var d mybool + _ = s + _ = b + _ = d + + // assignments to map index expressions are ok + s, b = m["foo"] + _, d = m["bar"] + m["foo"] = nil + m["foo"] = nil /* ERROR "assignment mismatch: 1 variable but 2 values" */ , false + _ = append(m["foo"]) + _ = append(m["foo"], true) + + var c chan int + _, b = <-c + _, d = <-c + <- /* ERROR "cannot assign" */ c = 0 + <-c = 0 /* ERROR "assignment mismatch: 1 variable but 2 values" */ , false + + var x interface{} + _, b = x.(int) + x /* ERROR "cannot assign" */ .(int) = 0 + x.(int) = 0 /* ERROR "assignment mismatch: 1 variable but 2 values" */ , false + + assignments2 /* ERROR "used as value" */ () = nil + int /* ERROR "not an expression" */ = 0 +} + +func issue6487() { + type S struct{x int} + _ = &S /* ERROR "cannot take address" */ {}.x + _ = &( /* ERROR "cannot take address" */ S{}.x) + _ = (&S{}).x + S /* ERROR "cannot assign" */ {}.x = 0 + (&S{}).x = 0 + + type M map[string]S + var m M + m /* ERROR "cannot assign to struct field" */ ["foo"].x = 0 + _ = &( /* ERROR "cannot take address" */ m["foo"].x) + _ = &m /* ERROR "cannot take address" */ ["foo"].x +} + +func issue6766a() { + a, a /* ERROR "a repeated on left side of :=" */ := 1, 2 + _ = a + a, b, b /* ERROR "b repeated on left side of :=" */ := 1, 2, 3 + _ = b + c, c /* ERROR "c repeated on left side of :=" */, b := 1, 2, 3 + _ = c + a, b := /* ERROR "no new variables" */ 1, 2 +} + +func shortVarDecls1() { + const c = 0 + type d int + a, b, c /* ERROR "cannot assign" */ , d /* ERROR "cannot assign" */ := 1, "zwei", 3.0, 4 + var _ int = a // a is of type int + var _ string = b // b is of type string +} + +func incdecs() { + const c = 3.14 + c /* ERROR "cannot assign" */ ++ + s := "foo" + s /* ERROR "invalid operation" */ -- + 3.14 /* ERROR "cannot assign" */ ++ + var ( + x int + y float32 + z complex128 + ) + x++ + y-- + z++ +} + +func sends() { + var ch chan int + var rch <-chan int + var x int + x <- /* ERROR "cannot send" */ x + rch <- /* ERROR "cannot send" */ x + ch <- "foo" /* ERRORx `cannot use .* in send` */ + ch <- x +} + +func selects() { + select {} + var ( + ch chan int + sc chan <- bool + ) + select { + case <-ch: + case (<-ch): + case t := <-ch: + _ = t + case t := (<-ch): + _ = t + case t, ok := <-ch: + _, _ = t, ok + case t, ok := (<-ch): + _, _ = t, ok + case <-sc /* ERROR "cannot receive from send-only channel" */ : + } + select { + default: + default /* ERROR "multiple defaults" */ : + } + select { + case a, b := <-ch: + _, b = a, b + case x /* ERROR "send or receive" */ : + case a /* ERROR "send or receive" */ := ch: + } + + // test for issue 9570: ch2 in second case falsely resolved to + // ch2 declared in body of first case + ch1 := make(chan int) + ch2 := make(chan int) + select { + case <-ch1: + var ch2 /* ERROR "ch2 declared and not used" */ chan bool + case i := <-ch2: + print(i + 1) + } +} + +func gos() { + go 1; /* ERROR "must be function call" */ + go int /* ERROR "go requires function call, not conversion" */ (0) + go ( /* ERROR "expression in go must not be parenthesized" */ gos()) + go gos() + var c chan int + go close(c) + go len /* ERROR "go discards result" */ (c) +} + +func defers() { + defer 1; /* ERROR "must be function call" */ + defer int /* ERROR "defer requires function call, not conversion" */ (0) + defer ( /* ERROR "expression in defer must not be parenthesized" */ defers()) + defer defers() + var c chan int + defer close(c) + defer len /* ERROR "defer discards result" */ (c) +} + +func breaks() { + var x, y int + + break /* ERROR "break" */ + { + break /* ERROR "break" */ + } + if x < y { + break /* ERROR "break" */ + } + + switch x { + case 0: + break + case 1: + if x == y { + break + } + default: + break + break + } + + var z interface{} + switch z.(type) { + case int: + break + } + + for { + break + } + + var a []int + for _ = range a { + break + } + + for { + if x == y { + break + } + } + + var ch chan int + select { + case <-ch: + break + } + + select { + case <-ch: + if x == y { + break + } + default: + break + } +} + +func continues() { + var x, y int + + continue /* ERROR "continue" */ + { + continue /* ERROR "continue" */ + } + + if x < y { + continue /* ERROR "continue" */ + } + + switch x { + case 0: + continue /* ERROR "continue" */ + } + + var z interface{} + switch z.(type) { + case int: + continue /* ERROR "continue" */ + } + + var ch chan int + select { + case <-ch: + continue /* ERROR "continue" */ + } + + for i := 0; i < 10; i++ { + continue + if x < y { + continue + break + } + switch x { + case y: + continue + default: + break + } + select { + case <-ch: + continue + } + } + + var a []int + for _ = range a { + continue + if x < y { + continue + break + } + switch x { + case y: + continue + default: + break + } + select { + case <-ch: + continue + } + } +} + +func returns0() { + return + return 0 /* ERROR "too many return values" */ +} + +func returns1(x float64) (int, *float64) { + return 0, &x + return /* ERROR "not enough return values" */ + return "foo" /* ERRORx `cannot .* in return statement` */, x /* ERRORx `cannot use .* in return statement` */ + return 0, &x, 1 /* ERROR "too many return values" */ +} + +func returns2() (a, b int) { + return + return 1, "foo" /* ERRORx `cannot use .* in return statement` */ + return 1, 2, 3 /* ERROR "too many return values" */ + { + type a int + return 1, 2 + return /* ERROR "a not in scope at return" */ + } +} + +func returns3() (_ int) { + return + { + var _ int // blank (_) identifiers never shadow since they are in no scope + return + } +} + +func switches0() { + var x int + + switch x { + } + + switch x { + default: + default /* ERROR "multiple defaults" */ : + } + + switch { + case 1 /* ERROR "cannot convert" */ : + } + + true := "false" + _ = true + // A tagless switch is equivalent to the bool + // constant true, not the identifier 'true'. + switch { + case "false" /* ERROR "cannot convert" */: + } + + switch int32(x) { + case 1, 2: + case x /* ERROR "invalid case x in switch on int32(x) (mismatched types int and int32)" */ : + } + + switch x { + case 1 /* ERROR "overflows" */ << 100: + } + + switch x { + case 1: + case 1 /* ERROR "duplicate case" */ : + case ( /* ERROR "duplicate case" */ 1): + case 2, 3, 4: + case 5, 1 /* ERROR "duplicate case" */ : + } + + switch uint64(x) { + case 1<<64 - 1: + case 1 /* ERROR "duplicate case" */ <<64 - 1: + case 2, 3, 4: + case 5, 1 /* ERROR "duplicate case" */ <<64 - 1: + } + + var y32 float32 + switch y32 { + case 1.1: + case 11/10: // integer division! + case 11. /* ERROR "duplicate case" */ /10: + case 2, 3.0, 4.1: + case 5.2, 1.10 /* ERROR "duplicate case" */ : + } + + var y64 float64 + switch y64 { + case 1.1: + case 11/10: // integer division! + case 11. /* ERROR "duplicate case" */ /10: + case 2, 3.0, 4.1: + case 5.2, 1.10 /* ERROR "duplicate case" */ : + } + + var s string + switch s { + case "foo": + case "foo" /* ERROR "duplicate case" */ : + case "f" /* ERROR "duplicate case" */ + "oo": + case "abc", "def", "ghi": + case "jkl", "foo" /* ERROR "duplicate case" */ : + } + + type T int + type F float64 + type S string + type B bool + var i interface{} + switch i { + case nil: + case nil: // no duplicate detection + case (*int)(nil): + case (*int)(nil): // do duplicate detection + case 1: + case byte(1): + case int /* ERROR "duplicate case" */ (1): + case T(1): + case 1.0: + case F(1.0): + case F /* ERROR "duplicate case" */ (1.0): + case "hello": + case S("hello"): + case S /* ERROR "duplicate case" */ ("hello"): + case 1==1, B(false): + case false, B(2==2): + } + + // switch on array + var a [3]int + switch a { + case [3]int{1, 2, 3}: + case [3]int{1, 2, 3}: // no duplicate detection + case [ /* ERROR "mismatched types" */ 4]int{4, 5, 6}: + } + + // switch on channel + var c1, c2 chan int + switch c1 { + case nil: + case c1: + case c2: + case c1, c2: // no duplicate detection + } +} + +func switches1() { + fallthrough /* ERROR "fallthrough statement out of place" */ + + var x int + switch x { + case 0: + fallthrough /* ERROR "fallthrough statement out of place" */ + break + case 1: + fallthrough + case 2: + fallthrough; ; ; // trailing empty statements are ok + case 3: + default: + fallthrough; ; + case 4: + fallthrough /* ERROR "cannot fallthrough final case in switch" */ + } + + var y interface{} + switch y.(type) { + case int: + fallthrough /* ERROR "cannot fallthrough in type switch" */ ; ; ; + default: + } + + switch x { + case 0: + if x == 0 { + fallthrough /* ERROR "fallthrough statement out of place" */ + } + } + + switch x { + case 0: + goto L1 + L1: fallthrough; ; + case 1: + goto L2 + goto L3 + goto L4 + L2: L3: L4: fallthrough + default: + } + + switch x { + case 0: + goto L5 + L5: fallthrough + default: + goto L6 + goto L7 + goto L8 + L6: L7: L8: fallthrough /* ERROR "cannot fallthrough final case in switch" */ + } + + switch x { + case 0: + fallthrough; ; + case 1: + { + fallthrough /* ERROR "fallthrough statement out of place" */ + } + case 2: + fallthrough + case 3: + fallthrough /* ERROR "fallthrough statement out of place" */ + { /* empty block is not an empty statement */ }; ; + default: + fallthrough /* ERROR "cannot fallthrough final case in switch" */ + } + + switch x { + case 0: + { + fallthrough /* ERROR "fallthrough statement out of place" */ + } + } +} + +func switches2() { + // untyped nil is not permitted as switch expression + switch nil /* ERROR "use of untyped nil" */ { + case 1, 2, "foo": // don't report additional errors here + } + + // untyped constants are converted to default types + switch 1<<63-1 { + } + switch 1 /* ERRORx `cannot use .* as int value.*\(overflows\)` */ << 63 { + } + var x int + switch 1.0 { + case 1.0, 2.0, x /* ERROR "mismatched types int and float64" */ : + } + switch x { + case 1.0: + } + + // untyped bools become of type bool + type B bool + var b B = true + switch x == x { + case b /* ERROR "mismatched types B and bool" */ : + } + switch { + case b /* ERROR "mismatched types B and bool" */ : + } +} + +func issue11667() { + switch 9223372036854775808 /* ERRORx `cannot use .* as int value.*\(overflows\)` */ { + } + switch 9223372036854775808 /* ERRORx `cannot use .* as int value.*\(overflows\)` */ { + case 9223372036854775808: + } + var x int + switch x { + case 9223372036854775808 /* ERROR "overflows int" */ : + } + var y float64 + switch y { + case 9223372036854775808: + } +} + +func issue11687() { + f := func() (_, _ int) { return } + switch f /* ERROR "multiple-value f" */ () { + } + var x int + switch f /* ERROR "multiple-value f" */ () { + case x: + } + switch x { + case f /* ERROR "multiple-value f" */ (): + } +} + +type I interface { + m() +} + +type I2 interface { + m(int) +} + +type T struct{} +type T1 struct{} +type T2 struct{} + +func (T) m() {} +func (T2) m(int) {} + +func typeswitches() { + var i int + var x interface{} + + switch x.(type) {} + switch (x /* ERROR "outside type switch" */ .(type)) {} + + switch x.(type) { + default: + default /* ERROR "multiple defaults" */ : + } + + switch x /* ERROR "declared and not used" */ := x.(type) {} + switch _ /* ERROR "no new variable on left side of :=" */ := x.(type) {} + + switch x := x.(type) { + case int: + var y int = x + _ = y + } + + switch x /* ERROR "x declared and not used" */ := i /* ERROR "not an interface" */ .(type) {} + + switch t := x.(type) { + case nil: + var v bool = t /* ERRORx `cannot use .* in variable declaration` */ + _ = v + case int: + var v int = t + _ = v + case float32, complex64: + var v float32 = t /* ERRORx `cannot use .* in variable declaration` */ + _ = v + default: + var v float32 = t /* ERRORx `cannot use .* in variable declaration` */ + _ = v + } + + var t I + switch t.(type) { + case T: + case T1 /* ERROR "missing method m" */ : + case T2 /* ERROR "wrong type for method m" */ : + case I2 /* STRICT "wrong type for method m" */ : // only an error in strict mode (issue 8561) + } + + + { + x := 1 + v := 2 + switch v /* ERROR "v (variable of type int) is not an interface" */ .(type) { + case int: + println(x) + println(x / 0 /* ERROR "invalid operation: division by zero" */) + case 1 /* ERROR "1 is not a type" */: + } + } +} + +// Test that each case clause uses the correct type of the variable +// declared by the type switch (issue 5504). +func typeswitch0() { + switch y := interface{}(nil).(type) { + case int: + func() int { return y + 0 }() + case float32: + func() float32 { return y }() + } +} + +// Test correct scope setup. +// (no redeclaration errors expected in the type switch) +func typeswitch1() { + var t I + switch t := t; t := t.(type) { + case nil: + var _ I = t + case T: + var _ T = t + default: + var _ I = t + } +} + +// Test correct typeswitch against interface types. +type A interface { a() } +type B interface { b() } +type C interface { a(int) } + +func typeswitch2() { + switch A(nil).(type) { + case A: + case B: + case C /* STRICT "cannot have dynamic type" */: // only an error in strict mode (issue 8561) + } +} + +func typeswitch3(x interface{}) { + switch x.(type) { + case int: + case float64: + case int /* ERROR "duplicate case" */ : + } + + switch x.(type) { + case nil: + case int: + case nil /* ERROR "duplicate case" */ , nil /* ERROR "duplicate case" */ : + } + + type F func(int) + switch x.(type) { + case nil: + case int, func(int): + case float32, func /* ERROR "duplicate case" */ (x int): + case F: + } +} + +func fors1() { + for {} + var i string + _ = i + for i := 0; i < 10; i++ {} + for i := 0; i < 10; j /* ERROR "cannot declare" */ := 0 {} +} + +func rangeloops1() { + var ( + x int + a [10]float32 + b []string + p *[10]complex128 + pp **[10]complex128 + s string + m map[int]bool + c chan int + sc chan<- int + rc <-chan int + ) + + for range x /* ERROR "cannot range over" */ {} + for _ = range x /* ERROR "cannot range over" */ {} + for i := range x /* ERROR "cannot range over" */ {} + + for range a {} + for i := range a { + var ii int + ii = i + _ = ii + } + for i, x := range a { + var ii int + ii = i + _ = ii + var xx float64 + xx = x /* ERRORx `cannot use .* in assignment` */ + _ = xx + } + var ii int + var xx float32 + for ii, xx = range a {} + _, _ = ii, xx + + for range b {} + for i := range b { + var ii int + ii = i + _ = ii + } + for i, x := range b { + var ii int + ii = i + _ = ii + var xx string + xx = x + _ = xx + } + + for range s {} + for i := range s { + var ii int + ii = i + _ = ii + } + for i, x := range s { + var ii int + ii = i + _ = ii + var xx rune + xx = x + _ = xx + } + + for range p {} + for _, x := range p { + var xx complex128 + xx = x + _ = xx + } + + for range pp /* ERROR "cannot range over" */ {} + for _, x := range pp /* ERROR "cannot range over" */ {} + + for range m {} + for k := range m { + var kk int32 + kk = k /* ERRORx `cannot use .* in assignment` */ + _ = kk + } + for k, v := range m { + var kk int + kk = k + _ = kk + if v {} + } + + for range c {} + for _, _ /* ERROR "only one iteration variable" */ = range c {} + for e := range c { + var ee int + ee = e + _ = ee + } + for _ = range sc /* ERROR "cannot range over" */ {} + for _ = range rc {} + + // constant strings + const cs = "foo" + for range cs {} + for range "" {} + for i, x := range cs { _, _ = i, x } + for i, x := range "" { + var ii int + ii = i + _ = ii + var xx rune + xx = x + _ = xx + } +} + +func rangeloops2() { + type I int + type R rune + + var a [10]int + var i I + _ = i + for i /* ERRORx `cannot use .* in assignment` */ = range a {} + for i /* ERRORx `cannot use .* in assignment` */ = range &a {} + for i /* ERRORx `cannot use .* in assignment` */ = range a[:] {} + + var s string + var r R + _ = r + for i /* ERRORx `cannot use .* in assignment` */ = range s {} + for i /* ERRORx `cannot use .* in assignment` */ = range "foo" {} + for _, r /* ERRORx `cannot use .* in assignment` */ = range s {} + for _, r /* ERRORx `cannot use .* in assignment` */ = range "foo" {} +} + +func issue6766b() { + for _ := /* ERROR "no new variables" */ range "" {} + for a, a /* ERROR "redeclared" */ := range "" { _ = a } + var a int + _ = a + for a, a /* ERROR "redeclared" */ := range []int{1, 2, 3} { _ = a } +} + +// Test that despite errors in the range clause, +// the loop body is still type-checked (and thus +// errors reported). +func issue10148() { + for y /* ERROR "declared and not used" */ := range "" { + _ = "" /* ERROR "mismatched types untyped string and untyped int" */ + 1 + } + for range 1 /* ERROR "cannot range over 1" */ { + _ = "" /* ERROR "mismatched types untyped string and untyped int" */ + 1 + } + for y := range 1 /* ERROR "cannot range over 1" */ { + _ = "" /* ERROR "mismatched types untyped string and untyped int" */ + 1 + } +} + +func labels0() { + goto L0 + goto L1 + L0: + L1: + L1 /* ERROR "already declared" */ : + if true { + goto L2 + L2: + L0 /* ERROR "already declared" */ : + } + _ = func() { + goto L0 + goto L1 + goto L2 + L0: + L1: + L2: + } +} + +func expression_statements(ch chan int) { + expression_statements(ch) + <-ch + println() + + 0 /* ERROR "not used" */ + 1 /* ERROR "not used" */ +2 + cap /* ERROR "not used" */ (ch) + println /* ERROR "must be called" */ +} diff --git a/src/internal/types/testdata/check/stmt1.go b/src/internal/types/testdata/check/stmt1.go new file mode 100644 index 0000000..f79f920 --- /dev/null +++ b/src/internal/types/testdata/check/stmt1.go @@ -0,0 +1,259 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// terminating statements + +package stmt1 + +func _() {} + +func _() int {} /* ERROR "missing return" */ + +func _() int { panic(0) } +func _() int { (panic(0)) } + +// block statements +func _(x, y int) (z int) { + { + return + } +} + +func _(x, y int) (z int) { + { + return; ; ; // trailing empty statements are ok + } + ; ; ; +} + +func _(x, y int) (z int) { + { + } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + { + ; ; ; + } + ; ; ; +} /* ERROR "missing return" */ + +// if statements +func _(x, y int) (z int) { + if x < y { return } + return 1 +} + +func _(x, y int) (z int) { + if x < y { return; ; ; ; } + return 1 +} + +func _(x, y int) (z int) { + if x < y { return } + return 1; ; +} + +func _(x, y int) (z int) { + if x < y { return } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + if x < y { + } else { return 1 + } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + if x < y { return + } else { return + } +} + +// for statements +func _(x, y int) (z int) { + for x < y { + return + } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + for { + return + } +} + +func _(x, y int) (z int) { + for { + return; ; ; ; + } +} + +func _(x, y int) (z int) { + for { + return + break + } + ; ; ; +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + for { + for { break } + return + } +} + +func _(x, y int) (z int) { + for { + for { break } + return ; ; + } + ; +} + +func _(x, y int) (z int) { +L: for { + for { break L } + return + } +} /* ERROR "missing return" */ + +// switch statements +func _(x, y int) (z int) { + switch x { + case 0: return + default: return + } +} + +func _(x, y int) (z int) { + switch x { + case 0: return; + default: return; ; ; + } +} + +func _(x, y int) (z int) { + switch x { + case 0: return + } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + switch x { + case 0: return + case 1: break + } +} /* ERROR "missing return" */ + +func _(x, y int) (z int) { + switch x { + case 0: return + default: + switch y { + case 0: break + } + panic(0) + } +} + +func _(x, y int) (z int) { + switch x { + case 0: return + default: + switch y { + case 0: break + } + panic(0); ; ; + } + ; +} + +func _(x, y int) (z int) { +L: switch x { + case 0: return + default: + switch y { + case 0: break L + } + panic(0) + } +} /* ERROR "missing return" */ + +// select statements +func _(ch chan int) (z int) { + select {} +} // nice! + +func _(ch chan int) (z int) { + select {} + ; ; +} + +func _(ch chan int) (z int) { + select { + default: break + } +} /* ERROR "missing return" */ + +func _(ch chan int) (z int) { + select { + case <-ch: return + default: break + } +} /* ERROR "missing return" */ + +func _(ch chan int) (z int) { + select { + case <-ch: return + default: + for i := 0; i < 10; i++ { + break + } + return + } +} + +func _(ch chan int) (z int) { + select { + case <-ch: return; ; ; + default: + for i := 0; i < 10; i++ { + break + } + return; ; ; + } + ; ; ; +} + +func _(ch chan int) (z int) { +L: select { + case <-ch: return + default: + for i := 0; i < 10; i++ { + break L + } + return + } + ; ; ; +} /* ERROR "missing return" */ + +func parenPanic() int { + ((((((panic)))(0)))) +} + +func issue23218a() int { + { + panic := func(interface{}){} + panic(0) + } +} /* ERROR "missing return" */ + +func issue23218b() int { + { + panic := func(interface{}){} + ((((panic))))(0) + } +} /* ERROR "missing return" */ diff --git a/src/internal/types/testdata/check/typeinference.go b/src/internal/types/testdata/check/typeinference.go new file mode 100644 index 0000000..0478d93 --- /dev/null +++ b/src/internal/types/testdata/check/typeinference.go @@ -0,0 +1,49 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package typeInference + +// As of issue #51527, type-type inference has been disabled. + +// basic inference +type Tb[P ~*Q, Q any] int +func _() { + var x Tb /* ERROR "got 1 arguments" */ [*int] + var y Tb[*int, int] + x = y /* ERRORx `cannot use y .* in assignment` */ + _ = x +} + +// recursive inference +type Tr[A any, B *C, C *D, D *A] int +func _() { + var x Tr /* ERROR "got 1 arguments" */ [string] + var y Tr[string, ***string, **string, *string] + var z Tr[int, ***int, **int, *int] + x = y /* ERRORx `cannot use y .* in assignment` */ + x = z // ERRORx `cannot use z .* as Tr` + _ = x +} + +// other patterns of inference +type To0[A any, B []A] int +type To1[A any, B struct{a A}] int +type To2[A any, B [][]A] int +type To3[A any, B [3]*A] int +type To4[A any, B any, C struct{a A; b B}] int +func _() { + var _ To0 /* ERROR "got 1 arguments" */ [int] + var _ To1 /* ERROR "got 1 arguments" */ [int] + var _ To2 /* ERROR "got 1 arguments" */ [int] + var _ To3 /* ERROR "got 1 arguments" */ [int] + var _ To4 /* ERROR "got 2 arguments" */ [int, string] +} + +// failed inference +type Tf0[A, B any] int +type Tf1[A any, B ~struct{a A; c C}, C any] int +func _() { + var _ Tf0 /* ERROR "got 1 arguments but 2 type parameters" */ [int] + var _ Tf1 /* ERROR "got 1 arguments but 3 type parameters" */ [int] +} diff --git a/src/internal/types/testdata/check/typeinst0.go b/src/internal/types/testdata/check/typeinst0.go new file mode 100644 index 0000000..bbcdaec --- /dev/null +++ b/src/internal/types/testdata/check/typeinst0.go @@ -0,0 +1,62 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type myInt int + +// Parameterized type declarations + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +type T1[P any] P // ERROR "cannot use a type parameter as RHS in type declaration" + +type T2[P any] struct { + f P + g int // int should still be in scope chain +} + +type List[P any] []P + +// Alias type declarations cannot have type parameters. +// Issue #46477 proposes to change that. +type A1[P any] = /* ERROR "cannot be alias" */ struct{} + +// Pending clarification of #46477 we disallow aliases +// of generic types. +type A2 = List // ERROR "cannot use generic type" +var _ A2[int] +var _ A2 + +type A3 = List[int] +var _ A3 + +// Parameterized type instantiations + +var x int +type _ x /* ERROR "not a type" */ [int] + +type _ int /* ERROR "not a generic type" */ [] // ERROR "expected type argument list" +type _ myInt /* ERROR "not a generic type" */ [] // ERROR "expected type argument list" + +// TODO(gri) better error messages +type _ T1[] // ERROR "expected type argument list" +type _ T1[x /* ERROR "not a type" */ ] +type _ T1 /* ERROR "got 2 arguments but 1 type parameters" */ [int, float32] + +var _ T2[int] = T2[int]{} + +var _ List[int] = []int{1, 2, 3} +var _ List[[]int] = [][]int{{1, 2, 3}} +var _ List[List[List[int]]] + +// Parameterized types containing parameterized types + +type T3[P any] List[P] + +var _ T3[int] = T3[int](List[int]{1, 2, 3}) + +// Self-recursive generic types are not permitted + +type self1[P any] self1 /* ERROR "invalid recursive type" */ [P] +type self2[P any] *self2[P] // this is ok diff --git a/src/internal/types/testdata/check/typeinst1.go b/src/internal/types/testdata/check/typeinst1.go new file mode 100644 index 0000000..0e09e70 --- /dev/null +++ b/src/internal/types/testdata/check/typeinst1.go @@ -0,0 +1,282 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type List[E any] []E +var _ List[List[List[int]]] +var _ List[List[List[int]]] = []List[List[int]]{} + +type ( + T1[P1 any] struct { + f1 T2[P1, float32] + } + + T2[P2, P3 any] struct { + f2 P2 + f3 P3 + } +) + +func _() { + var x1 T1[int] + var x2 T2[int, float32] + + x1.f1.f2 = 0 + x1.f1 = x2 +} + +type T3[P any] T1[T2[P, P]] + +func _() { + var x1 T3[int] + var x2 T2[int, int] + x1.f1.f2 = x2 +} + +func f[P any] (x P) List[P] { + return List[P]{x} +} + +var ( + _ []int = f(0) + _ []float32 = f[float32](10) + _ List[complex128] = f(1i) + _ []List[int] = f(List[int]{}) + _ List[List[int]] = []List[int]{} + _ = []List[int]{} +) + +// Parameterized types with methods + +func (l List[E]) Head() (_ E, _ bool) { + if len(l) > 0 { + return l[0], true + } + return +} + +// A test case for instantiating types with other types (extracted from map.go2) + +type Pair[K any] struct { + key K +} + +type Receiver[T any] struct { + values T +} + +type Iterator[K any] struct { + r Receiver[Pair[K]] +} + +func Values [T any] (r Receiver[T]) T { + return r.values +} + +func (it Iterator[K]) Next() K { + return Values[Pair[K]](it.r).key +} + +// A more complex test case testing type bounds (extracted from linalg.go2 and reduced to essence) + +type NumericAbs[T any] interface { + Abs() T +} + +func AbsDifference[T NumericAbs[T]](x T) { panic(0) } + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type OrderedAbs[T any] T +// +// func (a OrderedAbs[T]) Abs() OrderedAbs[T] +// +// func OrderedAbsDifference[T any](x T) { +// AbsDifference(OrderedAbs[T](x)) +// } + +// same code, reduced to essence + +func g[P interface{ m() P }](x P) { panic(0) } + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type T4[P any] P +// +// func (_ T4[P]) m() T4[P] +// +// func _[Q any](x Q) { +// g(T4[Q](x)) +// } + +// Another test case that caused problems in the past + +type T5[_ interface { a() }, _ interface{}] struct{} + +type A[P any] struct{ x P } + +func (_ A[P]) a() {} + +var _ T5[A[int], int] + +// Invoking methods with parameterized receiver types uses +// type inference to determine the actual type arguments matching +// the receiver type parameters from the actual receiver argument. +// Go does implicit address-taking and dereferenciation depending +// on the actual receiver and the method's receiver type. To make +// type inference work, the type-checker matches "pointer-ness" +// of the actual receiver and the method's receiver type. +// The following code tests this mechanism. + +type R1[A any] struct{} +func (_ R1[A]) vm() +func (_ *R1[A]) pm() + +func _[T any](r R1[T], p *R1[T]) { + r.vm() + r.pm() + p.vm() + p.pm() +} + +type R2[A, B any] struct{} +func (_ R2[A, B]) vm() +func (_ *R2[A, B]) pm() + +func _[T any](r R2[T, int], p *R2[string, T]) { + r.vm() + r.pm() + p.vm() + p.pm() +} + +// It is ok to have multiple embedded unions. +type _ interface { + m0() + ~int | ~string | ~bool + ~float32 | ~float64 + m1() + m2() + ~complex64 | ~complex128 + ~rune +} + +// Type sets may contain each type at most once. +type _ interface { + ~int|~ /* ERROR "overlapping terms ~int" */ int + ~int|int /* ERROR "overlapping terms int" */ + int|int /* ERROR "overlapping terms int" */ +} + +type _ interface { + ~struct{f int} | ~struct{g int} | ~ /* ERROR "overlapping terms" */ struct{f int} +} + +// Interface term lists can contain any type, incl. *Named types. +// Verify that we use the underlying type(s) of the type(s) in the +// term list when determining if an operation is permitted. + +type MyInt int +func add1[T interface{MyInt}](x T) T { + return x + 1 +} + +type MyString string +func double[T interface{MyInt|MyString}](x T) T { + return x + x +} + +// Embedding of interfaces with term lists leads to interfaces +// with term lists that are the intersection of the embedded +// term lists. + +type E0 interface { + ~int | ~bool | ~string +} + +type E1 interface { + ~int | ~float64 | ~string +} + +type E2 interface { + ~float64 +} + +type I0 interface { + E0 +} + +func f0[T I0]() {} +var _ = f0[int] +var _ = f0[bool] +var _ = f0[string] +var _ = f0[float64 /* ERROR "does not satisfy I0" */ ] + +type I01 interface { + E0 + E1 +} + +func f01[T I01]() {} +var _ = f01[int] +var _ = f01[bool /* ERROR "does not satisfy I0" */ ] +var _ = f01[string] +var _ = f01[float64 /* ERROR "does not satisfy I0" */ ] + +type I012 interface { + E0 + E1 + E2 +} + +func f012[T I012]() {} +var _ = f012[int /* ERRORx `cannot satisfy I012.*empty type set` */ ] +var _ = f012[bool /* ERRORx `cannot satisfy I012.*empty type set` */ ] +var _ = f012[string /* ERRORx `cannot satisfy I012.*empty type set` */ ] +var _ = f012[float64 /* ERRORx `cannot satisfy I012.*empty type set` */ ] + +type I12 interface { + E1 + E2 +} + +func f12[T I12]() {} +var _ = f12[int /* ERROR "does not satisfy I12" */ ] +var _ = f12[bool /* ERROR "does not satisfy I12" */ ] +var _ = f12[string /* ERROR "does not satisfy I12" */ ] +var _ = f12[float64] + +type I0_ interface { + E0 + ~int +} + +func f0_[T I0_]() {} +var _ = f0_[int] +var _ = f0_[bool /* ERROR "does not satisfy I0_" */ ] +var _ = f0_[string /* ERROR "does not satisfy I0_" */ ] +var _ = f0_[float64 /* ERROR "does not satisfy I0_" */ ] + +// Using a function instance as a type is an error. +var _ f0 // ERROR "not a type" +var _ f0 /* ERROR "not a type" */ [int] + +// Empty type sets can only be satisfied by empty type sets. +type none interface { + // force an empty type set + int + string +} + +func ff[T none]() {} +func gg[T any]() {} +func hh[T ~int]() {} + +func _[T none]() { + _ = ff[int /* ERROR "cannot satisfy none (empty type set)" */ ] + _ = ff[T] // pathological but ok because T's type set is empty, too + _ = gg[int] + _ = gg[T] + _ = hh[int] + _ = hh[T] +} diff --git a/src/internal/types/testdata/check/typeinstcycles.go b/src/internal/types/testdata/check/typeinstcycles.go new file mode 100644 index 0000000..74fe191 --- /dev/null +++ b/src/internal/types/testdata/check/typeinstcycles.go @@ -0,0 +1,11 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +func F1[T any](_ [unsafe.Sizeof(F1[int])]T) (res T) { return } +func F2[T any](_ T) (res [unsafe.Sizeof(F2[string])]int) { return } +func F3[T any](_ [unsafe.Sizeof(F1[string])]int) {} diff --git a/src/internal/types/testdata/check/typeparams.go b/src/internal/types/testdata/check/typeparams.go new file mode 100644 index 0000000..b002377 --- /dev/null +++ b/src/internal/types/testdata/check/typeparams.go @@ -0,0 +1,508 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// import "io" // for type assertion tests + +var _ any // ok to use any anywhere +func _[_ any, _ interface{any}](any) { + var _ any +} + +func identity[T any](x T) T { return x } + +func _[_ any](x int) int { panic(0) } +func _[T any](T /* ERROR "redeclared" */ T)() {} +func _[T, T /* ERROR "redeclared" */ any]() {} + +// Constraints (incl. any) may be parenthesized. +func _[_ (any)]() {} +func _[_ (interface{})]() {} + +func reverse[T any](list []T) []T { + rlist := make([]T, len(list)) + i := len(list) + for _, x := range list { + i-- + rlist[i] = x + } + return rlist +} + +var _ = reverse /* ERROR "cannot use generic function reverse" */ +var _ = reverse[int, float32 /* ERROR "got 2 type arguments" */ ] ([]int{1, 2, 3}) +var _ = reverse[int]([ /* ERROR "cannot use" */ ]float32{1, 2, 3}) +var f = reverse[chan int] +var _ = f(0 /* ERRORx `cannot use 0 .* as \[\]chan int` */ ) + +func swap[A, B any](a A, b B) (B, A) { return b, a } + +var _ = swap /* ERROR "multiple-value" */ [int, float32](1, 2) +var f32, i = swap[int, float32](swap[float32, int](1, 2)) +var _ float32 = f32 +var _ int = i + +func swapswap[A, B any](a A, b B) (A, B) { + return swap[B, A](b, a) +} + +type F[A, B any] func(A, B) (B, A) + +func min[T interface{ ~int }](x, y T) T { + if x < y { + return x + } + return y +} + +func _[T interface{~int | ~float32}](x, y T) bool { return x < y } +func _[T any](x, y T) bool { return x /* ERROR "type parameter T is not comparable" */ < y } +func _[T interface{~int | ~float32 | ~bool}](x, y T) bool { return x /* ERROR "type parameter T is not comparable" */ < y } + +func _[T C1[T]](x, y T) bool { return x /* ERROR "type parameter T is not comparable" */ < y } +func _[T C2[T]](x, y T) bool { return x < y } + +type C1[T any] interface{} +type C2[T any] interface{ ~int | ~float32 } + +func new[T any]() *T { + var x T + return &x +} + +var _ = new /* ERROR "cannot use generic function new" */ +var _ *int = new[int]() + +func _[T any](map[T /* ERROR "invalid map key type T (missing comparable constraint)" */]int) {} // w/o constraint we don't know if T is comparable + +func f1[T1 any](struct{T1 /* ERRORx `cannot be a .* type parameter` */ }) int { panic(0) } +var _ = f1[int](struct{T1}{}) +type T1 = int + +func f2[t1 any](struct{t1 /* ERRORx `cannot be a .* type parameter` */ ; x float32}) int { panic(0) } +var _ = f2[t1](struct{t1; x float32}{}) +type t1 = int + + +func f3[A, B, C any](A, struct{x B}, func(A, struct{x B}, *C)) int { panic(0) } + +var _ = f3[int, rune, bool](1, struct{x rune}{}, nil) + +// indexing + +func _[T any] (x T, i int) { _ = x /* ERROR "cannot index" */ [i] } +func _[T interface{ ~int }] (x T, i int) { _ = x /* ERROR "cannot index" */ [i] } +func _[T interface{ ~string }] (x T, i int) { _ = x[i] } +func _[T interface{ ~[]int }] (x T, i int) { _ = x[i] } +func _[T interface{ ~[10]int | ~*[20]int | ~map[int]int }] (x T, i int) { _ = x /* ERROR "cannot index" */ [i] } // map and non-map types +func _[T interface{ ~string | ~[]byte }] (x T, i int) { _ = x[i] } +func _[T interface{ ~[]int | ~[1]rune }] (x T, i int) { _ = x /* ERROR "cannot index" */ [i] } +func _[T interface{ ~string | ~[]rune }] (x T, i int) { _ = x /* ERROR "cannot index" */ [i] } + +// indexing with various combinations of map types in type sets (see issue #42616) +func _[T interface{ ~[]E | ~map[int]E }, E any](x T, i int) { _ = x /* ERROR "cannot index" */ [i] } // map and non-map types +func _[T interface{ ~[]E }, E any](x T, i int) { _ = &x[i] } +func _[T interface{ ~map[int]E }, E any](x T, i int) { _, _ = x[i] } // comma-ok permitted +func _[T interface{ ~map[int]E }, E any](x T, i int) { _ = &x /* ERROR "cannot take address" */ [i] } +func _[T interface{ ~map[int]E | ~map[uint]E }, E any](x T, i int) { _ = x /* ERROR "cannot index" */ [i] } // different map element types +func _[T interface{ ~[]E | ~map[string]E }, E any](x T, i int) { _ = x /* ERROR "cannot index" */ [i] } // map and non-map types + +// indexing with various combinations of array and other types in type sets +func _[T interface{ [10]int }](x T, i int) { _ = x[i]; _ = x[9]; _ = x[10 /* ERROR "out of bounds" */ ] } +func _[T interface{ [10]byte | string }](x T, i int) { _ = x[i]; _ = x[9]; _ = x[10 /* ERROR "out of bounds" */ ] } +func _[T interface{ [10]int | *[20]int | []int }](x T, i int) { _ = x[i]; _ = x[9]; _ = x[10 /* ERROR "out of bounds" */ ] } + +// indexing with strings and non-variable arrays (assignment not permitted) +func _[T string](x T) { _ = x[0]; x /* ERROR "cannot assign" */ [0] = 0 } +func _[T []byte | string](x T) { x /* ERROR "cannot assign" */ [0] = 0 } +func _[T [10]byte]() { f := func() (x T) { return }; f /* ERROR "cannot assign" */ ()[0] = 0 } +func _[T [10]byte]() { f := func() (x *T) { return }; f /* ERROR "cannot index" */ ()[0] = 0 } +func _[T [10]byte]() { f := func() (x *T) { return }; (*f())[0] = 0 } +func _[T *[10]byte]() { f := func() (x T) { return }; f()[0] = 0 } + +// slicing + +func _[T interface{ ~[10]E }, E any] (x T, i, j, k int) { var _ []E = x[i:j] } +func _[T interface{ ~[10]E }, E any] (x T, i, j, k int) { var _ []E = x[i:j:k] } +func _[T interface{ ~[]byte }] (x T, i, j, k int) { var _ T = x[i:j] } +func _[T interface{ ~[]byte }] (x T, i, j, k int) { var _ T = x[i:j:k] } +func _[T interface{ ~string }] (x T, i, j, k int) { var _ T = x[i:j] } +func _[T interface{ ~string }] (x T, i, j, k int) { var _ T = x[i:j:k /* ERROR "3-index slice of string" */ ] } + +type myByte1 []byte +type myByte2 []byte +func _[T interface{ []byte | myByte1 | myByte2 }] (x T, i, j, k int) { var _ T = x[i:j:k] } +func _[T interface{ []byte | myByte1 | []int }] (x T, i, j, k int) { var _ T = x /* ERROR "no core type" */ [i:j:k] } + +func _[T interface{ []byte | myByte1 | myByte2 | string }] (x T, i, j, k int) { var _ T = x[i:j] } +func _[T interface{ []byte | myByte1 | myByte2 | string }] (x T, i, j, k int) { var _ T = x[i:j:k /* ERROR "3-index slice of string" */ ] } +func _[T interface{ []byte | myByte1 | []int | string }] (x T, i, j, k int) { var _ T = x /* ERROR "no core type" */ [i:j] } + +// len/cap built-ins + +func _[T any](x T) { _ = len(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~int }](x T) { _ = len(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~string | ~[]byte | ~int }](x T) { _ = len(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~string }](x T) { _ = len(x) } +func _[T interface{ ~[10]int }](x T) { _ = len(x) } +func _[T interface{ ~[]byte }](x T) { _ = len(x) } +func _[T interface{ ~map[int]int }](x T) { _ = len(x) } +func _[T interface{ ~chan int }](x T) { _ = len(x) } +func _[T interface{ ~string | ~[]byte | ~chan int }](x T) { _ = len(x) } + +func _[T any](x T) { _ = cap(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~int }](x T) { _ = cap(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~string | ~[]byte | ~int }](x T) { _ = cap(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~string }](x T) { _ = cap(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~[10]int }](x T) { _ = cap(x) } +func _[T interface{ ~[]byte }](x T) { _ = cap(x) } +func _[T interface{ ~map[int]int }](x T) { _ = cap(x /* ERROR "invalid argument" */ ) } +func _[T interface{ ~chan int }](x T) { _ = cap(x) } +func _[T interface{ ~[]byte | ~chan int }](x T) { _ = cap(x) } + +// range iteration + +func _[T interface{}](x T) { + for range x /* ERROR "cannot range" */ {} +} + +type myString string + +func _[ + B1 interface{ string }, + B2 interface{ string | myString }, + + C1 interface{ chan int }, + C2 interface{ chan int | <-chan int }, + C3 interface{ chan<- int }, + + S1 interface{ []int }, + S2 interface{ []int | [10]int }, + + A1 interface{ [10]int }, + A2 interface{ [10]int | []int }, + + P1 interface{ *[10]int }, + P2 interface{ *[10]int | *[]int }, + + M1 interface{ map[string]int }, + M2 interface{ map[string]int | map[string]string }, +]() { + var b0 string + for range b0 {} + for _ = range b0 {} + for _, _ = range b0 {} + + var b1 B1 + for range b1 {} + for _ = range b1 {} + for _, _ = range b1 {} + + var b2 B2 + for range b2 {} + + var c0 chan int + for range c0 {} + for _ = range c0 {} + for _, _ /* ERROR "permits only one iteration variable" */ = range c0 {} + + var c1 C1 + for range c1 {} + for _ = range c1 {} + for _, _ /* ERROR "permits only one iteration variable" */ = range c1 {} + + var c2 C2 + for range c2 {} + + var c3 C3 + for range c3 /* ERROR "receive from send-only channel" */ {} + + var s0 []int + for range s0 {} + for _ = range s0 {} + for _, _ = range s0 {} + + var s1 S1 + for range s1 {} + for _ = range s1 {} + for _, _ = range s1 {} + + var s2 S2 + for range s2 /* ERRORx `cannot range over s2.*no core type` */ {} + + var a0 []int + for range a0 {} + for _ = range a0 {} + for _, _ = range a0 {} + + var a1 A1 + for range a1 {} + for _ = range a1 {} + for _, _ = range a1 {} + + var a2 A2 + for range a2 /* ERRORx `cannot range over a2.*no core type` */ {} + + var p0 *[10]int + for range p0 {} + for _ = range p0 {} + for _, _ = range p0 {} + + var p1 P1 + for range p1 {} + for _ = range p1 {} + for _, _ = range p1 {} + + var p2 P2 + for range p2 /* ERRORx `cannot range over p2.*no core type` */ {} + + var m0 map[string]int + for range m0 {} + for _ = range m0 {} + for _, _ = range m0 {} + + var m1 M1 + for range m1 {} + for _ = range m1 {} + for _, _ = range m1 {} + + var m2 M2 + for range m2 /* ERRORx `cannot range over m2.*no core type` */ {} +} + +// type inference checks + +var _ = new /* ERROR "cannot infer T" */ () + +func f4[A, B, C any](A, B) C { panic(0) } + +var _ = f4 /* ERROR "cannot infer C" */ (1, 2) +var _ = f4[int, float32, complex128](1, 2) + +func f5[A, B, C any](A, []*B, struct{f []C}) int { panic(0) } + +var _ = f5[int, float32, complex128](0, nil, struct{f []complex128}{}) +var _ = f5 /* ERROR "cannot infer" */ (0, nil, struct{f []complex128}{}) +var _ = f5(0, []*float32{new[float32]()}, struct{f []complex128}{}) + +func f6[A any](A, []A) int { panic(0) } + +var _ = f6(0, nil) + +func f6nil[A any](A) int { panic(0) } + +var _ = f6nil /* ERROR "cannot infer" */ (nil) + +// type inference with variadic functions + +func f7[T any](...T) T { panic(0) } + +var _ int = f7 /* ERROR "cannot infer T" */ () +var _ int = f7(1) +var _ int = f7(1, 2) +var _ int = f7([]int{}...) +var _ int = f7 /* ERROR "cannot use" */ ([]float64{}...) +var _ float64 = f7([]float64{}...) +var _ = f7[float64](1, 2.3) +var _ = f7(float64(1), 2.3) +var _ = f7(1, 2.3) +var _ = f7(1.2, 3) + +func f8[A, B any](A, B, ...B) int { panic(0) } + +var _ = f8(1) /* ERROR "not enough arguments" */ +var _ = f8(1, 2.3) +var _ = f8(1, 2.3, 3.4, 4.5) +var _ = f8(1, 2.3, 3.4, 4) +var _ = f8[int, float64](1, 2.3, 3.4, 4) + +var _ = f8[int, float64](0, 0, nil...) // test case for #18268 + +// init functions cannot have type parameters + +func init() {} +func init[_ /* ERROR "func init must have no type parameters" */ any]() {} +func init[P /* ERROR "func init must have no type parameters" */ any]() {} + +type T struct {} + +func (T) m1() {} +func (T) m2[ /* ERROR "method must have no type parameters" */ _ any]() {} +func (T) m3[ /* ERROR "method must have no type parameters" */ P any]() {} + +// type inference across parameterized types + +type S1[P any] struct { f P } + +func f9[P any](x S1[P]) {} + +func _() { + f9[int](S1[int]{42}) + f9(S1[int]{42}) +} + +type S2[A, B, C any] struct{} + +func f10[X, Y, Z any](a S2[X, int, Z], b S2[X, Y, bool]) {} + +func _[P any]() { + f10[int, float32, string](S2[int, int, string]{}, S2[int, float32, bool]{}) + f10(S2[int, int, string]{}, S2[int, float32, bool]{}) + f10(S2[P, int, P]{}, S2[P, float32, bool]{}) +} + +// corner case for type inference +// (was bug: after instantiating f11, the type-checker didn't mark f11 as non-generic) + +func f11[T any]() {} + +func _() { + f11[int]() +} + +// the previous example was extracted from + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// func f12[T interface{m() T}]() {} +// +// type A[T any] T +// +// func (a A[T]) m() A[T] +// +// func _[T any]() { +// f12[A[T]]() +// } + +// method expressions + +func (_ S1[P]) m() + +func _() { + m := S1[int].m + m(struct { f int }{42}) +} + +func _[T any] (x T) { + m := S1[T].m + m(S1[T]{x}) +} + +type I1[A any] interface { + m1(A) +} + +var _ I1[int] = r1[int]{} + +type r1[T any] struct{} + +func (_ r1[T]) m1(T) + +type I2[A, B any] interface { + m1(A) + m2(A) B +} + +var _ I2[int, float32] = R2[int, float32]{} + +type R2[P, Q any] struct{} + +func (_ R2[X, Y]) m1(X) +func (_ R2[X, Y]) m2(X) Y + +// type assertions and type switches over generic types +// NOTE: These are currently disabled because it's unclear what the correct +// approach is, and one can always work around by assigning the variable to +// an interface first. + +// // ReadByte1 corresponds to the ReadByte example in the draft design. +// func ReadByte1[T io.Reader](r T) (byte, error) { +// if br, ok := r.(io.ByteReader); ok { +// return br.ReadByte() +// } +// var b [1]byte +// _, err := r.Read(b[:]) +// return b[0], err +// } +// +// // ReadBytes2 is like ReadByte1 but uses a type switch instead. +// func ReadByte2[T io.Reader](r T) (byte, error) { +// switch br := r.(type) { +// case io.ByteReader: +// return br.ReadByte() +// } +// var b [1]byte +// _, err := r.Read(b[:]) +// return b[0], err +// } +// +// // type assertions and type switches over generic types are strict +// type I3 interface { +// m(int) +// } +// +// type I4 interface { +// m() int // different signature from I3.m +// } +// +// func _[T I3](x I3, p T) { +// // type assertions and type switches over interfaces are not strict +// _ = x.(I4) +// switch x.(type) { +// case I4: +// } +// +// // type assertions and type switches over generic types are strict +// _ = p /* ERROR "cannot have dynamic type I4" */.(I4) +// switch p.(type) { +// case I4 /* ERROR "cannot have dynamic type I4" */ : +// } +// } + +// type assertions and type switches over generic types lead to errors for now + +func _[T any](x T) { + _ = x /* ERROR "cannot use type assertion" */ .(int) + switch x /* ERROR "cannot use type switch" */ .(type) { + } + + // work-around + var t interface{} = x + _ = t.(int) + switch t.(type) { + } +} + +func _[T interface{~int}](x T) { + _ = x /* ERROR "cannot use type assertion" */ .(int) + switch x /* ERROR "cannot use type switch" */ .(type) { + } + + // work-around + var t interface{} = x + _ = t.(int) + switch t.(type) { + } +} + +// error messages related to type bounds mention those bounds +type C[P any] interface{} + +func _[P C[P]] (x P) { + x.m /* ERROR "x.m undefined" */ () +} + +type I interface {} + +func _[P I] (x P) { + x.m /* ERROR "type P has no field or method m" */ () +} + +func _[P interface{}] (x P) { + x.m /* ERROR "type P has no field or method m" */ () +} + +func _[P any] (x P) { + x.m /* ERROR "type P has no field or method m" */ () +} diff --git a/src/internal/types/testdata/check/unions.go b/src/internal/types/testdata/check/unions.go new file mode 100644 index 0000000..5a3a9ed --- /dev/null +++ b/src/internal/types/testdata/check/unions.go @@ -0,0 +1,66 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check that overlong unions don't bog down type checking. +// Disallow them for now. + +package p + +type t int + +type ( + t00 t; t01 t; t02 t; t03 t; t04 t; t05 t; t06 t; t07 t; t08 t; t09 t + t10 t; t11 t; t12 t; t13 t; t14 t; t15 t; t16 t; t17 t; t18 t; t19 t + t20 t; t21 t; t22 t; t23 t; t24 t; t25 t; t26 t; t27 t; t28 t; t29 t + t30 t; t31 t; t32 t; t33 t; t34 t; t35 t; t36 t; t37 t; t38 t; t39 t + t40 t; t41 t; t42 t; t43 t; t44 t; t45 t; t46 t; t47 t; t48 t; t49 t + t50 t; t51 t; t52 t; t53 t; t54 t; t55 t; t56 t; t57 t; t58 t; t59 t + t60 t; t61 t; t62 t; t63 t; t64 t; t65 t; t66 t; t67 t; t68 t; t69 t + t70 t; t71 t; t72 t; t73 t; t74 t; t75 t; t76 t; t77 t; t78 t; t79 t + t80 t; t81 t; t82 t; t83 t; t84 t; t85 t; t86 t; t87 t; t88 t; t89 t + t90 t; t91 t; t92 t; t93 t; t94 t; t95 t; t96 t; t97 t; t98 t; t99 t +) + +type u99 interface { + t00|t01|t02|t03|t04|t05|t06|t07|t08|t09| + t10|t11|t12|t13|t14|t15|t16|t17|t18|t19| + t20|t21|t22|t23|t24|t25|t26|t27|t28|t29| + t30|t31|t32|t33|t34|t35|t36|t37|t38|t39| + t40|t41|t42|t43|t44|t45|t46|t47|t48|t49| + t50|t51|t52|t53|t54|t55|t56|t57|t58|t59| + t60|t61|t62|t63|t64|t65|t66|t67|t68|t69| + t70|t71|t72|t73|t74|t75|t76|t77|t78|t79| + t80|t81|t82|t83|t84|t85|t86|t87|t88|t89| + t90|t91|t92|t93|t94|t95|t96|t97|t98 +} + +type u100a interface { + u99|float32 +} + +type u100b interface { + u99|float64 +} + +type u101 interface { + t00|t01|t02|t03|t04|t05|t06|t07|t08|t09| + t10|t11|t12|t13|t14|t15|t16|t17|t18|t19| + t20|t21|t22|t23|t24|t25|t26|t27|t28|t29| + t30|t31|t32|t33|t34|t35|t36|t37|t38|t39| + t40|t41|t42|t43|t44|t45|t46|t47|t48|t49| + t50|t51|t52|t53|t54|t55|t56|t57|t58|t59| + t60|t61|t62|t63|t64|t65|t66|t67|t68|t69| + t70|t71|t72|t73|t74|t75|t76|t77|t78|t79| + t80|t81|t82|t83|t84|t85|t86|t87|t88|t89| + t90|t91|t92|t93|t94|t95|t96|t97|t98|t99| + int // ERROR "cannot handle more than 100 union terms" +} + +type u102 interface { + int /* ERROR "cannot handle more than 100 union terms" */ |string|u100a +} + +type u200 interface { + u100a /* ERROR "cannot handle more than 100 union terms" */ |u100b +} diff --git a/src/internal/types/testdata/check/vardecl.go b/src/internal/types/testdata/check/vardecl.go new file mode 100644 index 0000000..726b619 --- /dev/null +++ b/src/internal/types/testdata/check/vardecl.go @@ -0,0 +1,215 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package vardecl + +// Prerequisites. +import "math" +func f() {} +func g() (x, y int) { return } +var m map[string]int + +// Var decls must have a type or an initializer. +var _ int +var _, _ int + +var _; /* ERROR "expected type" */ +var _, _; /* ERROR "expected type" */ +var _, _, _; /* ERROR "expected type" */ + +// The initializer must be an expression. +var _ = int /* ERROR "not an expression" */ +var _ = f /* ERROR "used as value" */ () + +// Identifier and expression arity must match. +var _, _ = 1, 2 +var _ = 1, 2 /* ERROR "extra init expr 2" */ +var _, _ = 1 /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ +var _, _, _ /* ERROR "missing init expr for _" */ = 1, 2 + +var _ = g /* ERROR "multiple-value g" */ () +var _, _ = g() +var _, _, _ = g /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ () + +var _ = m["foo"] +var _, _ = m["foo"] +var _, _, _ = m /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ ["foo"] + +var _, _ int = 1, 2 +var _ int = 1, 2 /* ERROR "extra init expr 2" */ +var _, _ int = 1 /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ +var _, _, _ /* ERROR "missing init expr for _" */ int = 1, 2 + +var ( + _, _ = 1, 2 + _ = 1, 2 /* ERROR "extra init expr 2" */ + _, _ = 1 /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ + _, _, _ /* ERROR "missing init expr for _" */ = 1, 2 + + _ = g /* ERROR "multiple-value g" */ () + _, _ = g() + _, _, _ = g /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ () + + _ = m["foo"] + _, _ = m["foo"] + _, _, _ = m /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ ["foo"] + + _, _ int = 1, 2 + _ int = 1, 2 /* ERROR "extra init expr 2" */ + _, _ int = 1 /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ + _, _, _ /* ERROR "missing init expr for _" */ int = 1, 2 +) + +// Variables declared in function bodies must be 'used'. +type T struct{} +func (r T) _(a, b, c int) (u, v, w int) { + var x1 /* ERROR "declared and not used" */ int + var x2 /* ERROR "declared and not used" */ int + x1 = 1 + (x2) = 2 + + y1 /* ERROR "declared and not used" */ := 1 + y2 /* ERROR "declared and not used" */ := 2 + y1 = 1 + (y1) = 2 + + { + var x1 /* ERROR "declared and not used" */ int + var x2 /* ERROR "declared and not used" */ int + x1 = 1 + (x2) = 2 + + y1 /* ERROR "declared and not used" */ := 1 + y2 /* ERROR "declared and not used" */ := 2 + y1 = 1 + (y1) = 2 + } + + if x /* ERROR "declared and not used" */ := 0; a < b {} + + switch x /* ERROR "declared and not used" */, y := 0, 1; a { + case 0: + _ = y + case 1: + x /* ERROR "declared and not used" */ := 0 + } + + var t interface{} + switch t /* ERROR "declared and not used" */ := t.(type) {} + + switch t /* ERROR "declared and not used" */ := t.(type) { + case int: + } + + switch t /* ERROR "declared and not used" */ := t.(type) { + case int: + case float32, complex64: + t = nil + } + + switch t := t.(type) { + case int: + case float32, complex64: + _ = t + } + + switch t := t.(type) { + case int: + case float32: + case string: + _ = func() string { + return t + } + } + + switch t := t; t /* ERROR "declared and not used" */ := t.(type) {} + + var z1 /* ERROR "declared and not used" */ int + var z2 int + _ = func(a, b, c int) (u, v, w int) { + z1 = a + (z1) = b + a = z2 + return + } + + var s []int + var i /* ERROR "declared and not used" */ , j int + for i, j = range s { + _ = j + } + + for i, j /* ERROR "declared and not used" */ := range s { + _ = func() int { + return i + } + } + return +} + +// Unused variables in function literals must lead to only one error (issue #22524). +func _() { + _ = func() { + var x /* ERROR "declared and not used" */ int + } +} + +// Invalid variable declarations must not lead to "declared and not used errors". +// TODO(gri) enable these tests once go/types follows types2 logic for declared and not used variables +// func _() { +// var a x // DISABLED_ERROR undefined: x +// var b = x // DISABLED_ERROR undefined: x +// var c int = x // DISABLED_ERROR undefined: x +// var d, e, f x /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ +// var g, h, i = x, x, x /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ +// var j, k, l float32 = x, x, x /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ /* DISABLED_ERROR x */ +// // but no "declared and not used" errors +// } + +// Invalid (unused) expressions must not lead to spurious "declared and not used errors". +func _() { + var a, b, c int + var x, y int + x, y = a /* ERRORx `assignment mismatch: [1-9]+ variables but.*[1-9]+ value(s)?` */ , b, c + _ = x + _ = y +} + +func _() { + var x int + return x /* ERROR "too many return values" */ + return math /* ERROR "too many return values" */ .Sin(0) +} + +func _() int { + var x, y int + return x, y /* ERROR "too many return values" */ +} + +// Short variable declarations must declare at least one new non-blank variable. +func _() { + _ := /* ERROR "no new variables" */ 0 + _, a := 0, 1 + _, a := /* ERROR "no new variables" */ 0, 1 + _, a, b := 0, 1, 2 + _, _, _ := /* ERROR "no new variables" */ 0, 1, 2 + + _ = a + _ = b +} + +// Test case for variables depending on function literals (see also #22992). +var A /* ERROR "initialization cycle" */ = func() int { return A }() + +func _() { + // The function literal below must not see a. + var a = func() int { return a /* ERROR "undefined" */ }() + var _ = func() int { return a }() + + // The function literal below must not see x, y, or z. + var x, y, z = 0, 1, func() int { return x /* ERROR "undefined" */ + y /* ERROR "undefined" */ + z /* ERROR "undefined" */ }() + _, _, _ = x, y, z +} + +// TODO(gri) consolidate other var decl checks in this file
\ No newline at end of file diff --git a/src/internal/types/testdata/examples/constraints.go b/src/internal/types/testdata/examples/constraints.go new file mode 100644 index 0000000..4c97a40 --- /dev/null +++ b/src/internal/types/testdata/examples/constraints.go @@ -0,0 +1,80 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of generic constraint interfaces. + +package p + +type MyInt int + +type ( + // Arbitrary types may be embedded like interfaces. + _ interface{int} + _ interface{~int} + + // Types may be combined into a union. + union interface{int|~string} + + // Union terms must describe disjoint (non-overlapping) type sets. + _ interface{int|int /* ERROR "overlapping terms int" */ } + _ interface{int|~ /* ERROR "overlapping terms ~int" */ int } + _ interface{~int|~ /* ERROR "overlapping terms ~int" */ int } + _ interface{~int|MyInt /* ERROR "overlapping terms p.MyInt and ~int" */ } + _ interface{int|any} + _ interface{int|~string|union} + _ interface{int|~string|interface{int}} + _ interface{union|int} // interfaces (here: union) are ignored when checking for overlap + _ interface{union|union} // ditto + + // For now we do not permit interfaces with methods in unions. + _ interface{~ /* ERROR "invalid use of ~" */ any} + _ interface{int|interface /* ERRORx `cannot use .* in union` */ { m() }} +) + +type ( + // Tilde is not permitted on defined types or interfaces. + foo int + bar any + _ interface{foo} + _ interface{~ /* ERROR "invalid use of ~" */ foo } + _ interface{~ /* ERROR "invalid use of ~" */ bar } +) + +// Stand-alone type parameters are not permitted as elements or terms in unions. +type ( + _[T interface{ *T } ] struct{} // ok + _[T interface{ int | *T } ] struct{} // ok + _[T interface{ T /* ERROR "term cannot be a type parameter" */ } ] struct{} + _[T interface{ ~T /* ERROR "type in term ~T cannot be a type parameter" */ } ] struct{} + _[T interface{ int|T /* ERROR "term cannot be a type parameter" */ }] struct{} +) + +// Multiple embedded union elements are intersected. The order in which they +// appear in the interface doesn't matter since intersection is a symmetric +// operation. + +type myInt1 int +type myInt2 int + +func _[T interface{ myInt1|myInt2; ~int }]() T { return T(0) } +func _[T interface{ ~int; myInt1|myInt2 }]() T { return T(0) } + +// Here the intersections are empty - there's no type that's in the type set of T. +func _[T interface{ myInt1|myInt2; int }]() T { return T(0 /* ERROR "cannot convert" */ ) } +func _[T interface{ int; myInt1|myInt2 }]() T { return T(0 /* ERROR "cannot convert" */ ) } + +// Union elements may be interfaces as long as they don't define +// any methods or embed comparable. + +type ( + Integer interface{ ~int|~int8|~int16|~int32|~int64 } + Unsigned interface{ ~uint|~uint8|~uint16|~uint32|~uint64 } + Floats interface{ ~float32|~float64 } + Complex interface{ ~complex64|~complex128 } + Number interface{ Integer|Unsigned|Floats|Complex } + Ordered interface{ Integer|Unsigned|Floats|~string } + + _ interface{ Number | error /* ERROR "cannot use error in union" */ } + _ interface{ Ordered | comparable /* ERROR "cannot use comparable in union" */ } +) diff --git a/src/internal/types/testdata/examples/functions.go b/src/internal/types/testdata/examples/functions.go new file mode 100644 index 0000000..fdc67e7 --- /dev/null +++ b/src/internal/types/testdata/examples/functions.go @@ -0,0 +1,219 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of type-parameterized functions. + +package p + +// Reverse is a generic function that takes a []T argument and +// reverses that slice in place. +func Reverse[T any](list []T) { + i := 0 + j := len(list)-1 + for i < j { + list[i], list[j] = list[j], list[i] + i++ + j-- + } +} + +func _() { + // Reverse can be called with an explicit type argument. + Reverse[int](nil) + Reverse[string]([]string{"foo", "bar"}) + Reverse[struct{x, y int}]([]struct{x, y int}{{1, 2}, {2, 3}, {3, 4}}) + + // Since the type parameter is used for an incoming argument, + // it can be inferred from the provided argument's type. + Reverse([]string{"foo", "bar"}) + Reverse([]struct{x, y int}{{1, 2}, {2, 3}, {3, 4}}) + + // But the incoming argument must have a type, even if it's a + // default type. An untyped nil won't work. + // Reverse(nil) // this won't type-check + + // A typed nil will work, though. + Reverse([]int(nil)) +} + +// Certain functions, such as the built-in `new` could be written using +// type parameters. +func new[T any]() *T { + var x T + return &x +} + +// When calling our own `new`, we need to pass the type parameter +// explicitly since there is no (value) argument from which the +// result type could be inferred. We don't try to infer the +// result type from the assignment to keep things simple and +// easy to understand. +var _ = new[int]() +var _ *float64 = new[float64]() // the result type is indeed *float64 + +// A function may have multiple type parameters, of course. +func foo[A, B, C any](a A, b []B, c *C) B { + // do something here + return b[0] +} + +// As before, we can pass type parameters explicitly. +var s = foo[int, string, float64](1, []string{"first"}, new[float64]()) + +// Or we can use type inference. +var _ float64 = foo(42, []float64{1.0}, &s) + +// Type inference works in a straight-forward manner even +// for variadic functions. +func variadic[A, B any](A, B, ...B) int { panic(0) } + +// var _ = variadic(1) // ERROR "not enough arguments" +var _ = variadic(1, 2.3) +var _ = variadic(1, 2.3, 3.4, 4.5) +var _ = variadic[int, float64](1, 2.3, 3.4, 4) + +// Type inference also works in recursive function calls where +// the inferred type is the type parameter of the caller. +func f1[T any](x T) { + f1(x) +} + +func f2a[T any](x, y T) { + f2a(x, y) +} + +func f2b[T any](x, y T) { + f2b(y, x) +} + +func g2a[P, Q any](x P, y Q) { + g2a(x, y) +} + +func g2b[P, Q any](x P, y Q) { + g2b(y, x) +} + +// Here's an example of a recursive function call with variadic +// arguments and type inference inferring the type parameter of +// the caller (i.e., itself). +func max[T interface{ ~int }](x ...T) T { + var x0 T + if len(x) > 0 { + x0 = x[0] + } + if len(x) > 1 { + x1 := max(x[1:]...) + if x1 > x0 { + return x1 + } + } + return x0 +} + +// When inferring channel types, the channel direction is ignored +// for the purpose of type inference. Once the type has been in- +// fered, the usual parameter passing rules are applied. +// Thus even if a type can be inferred successfully, the function +// call may not be valid. + +func fboth[T any](chan T) {} +func frecv[T any](<-chan T) {} +func fsend[T any](chan<- T) {} + +func _() { + var both chan int + var recv <-chan int + var send chan<-int + + fboth(both) + fboth(recv /* ERROR "cannot use" */ ) + fboth(send /* ERROR "cannot use" */ ) + + frecv(both) + frecv(recv) + frecv(send /* ERROR "cannot use" */ ) + + fsend(both) + fsend(recv /* ERROR "cannot use" */) + fsend(send) +} + +func ffboth[T any](func(chan T)) {} +func ffrecv[T any](func(<-chan T)) {} +func ffsend[T any](func(chan<- T)) {} + +func _() { + var both func(chan int) + var recv func(<-chan int) + var send func(chan<- int) + + ffboth(both) + ffboth(recv /* ERROR "does not match" */ ) + ffboth(send /* ERROR "does not match" */ ) + + ffrecv(both /* ERROR "does not match" */ ) + ffrecv(recv) + ffrecv(send /* ERROR "does not match" */ ) + + ffsend(both /* ERROR "does not match" */ ) + ffsend(recv /* ERROR "does not match" */ ) + ffsend(send) +} + +// When inferring elements of unnamed composite parameter types, +// if the arguments are defined types, use their underlying types. +// Even though the matching types are not exactly structurally the +// same (one is a type literal, the other a named type), because +// assignment is permitted, parameter passing is permitted as well, +// so type inference should be able to handle these cases well. + +func g1[T any]([]T) {} +func g2[T any]([]T, T) {} +func g3[T any](*T, ...T) {} + +func _() { + type intSlice []int + g1([]int{}) + g1(intSlice{}) + g2(nil, 0) + + type myString string + var s1 string + g3(nil, "1", myString("2"), "3") + g3(& /* ERROR "cannot use &s1 (value of type *string) as *myString value in argument to g3" */ s1, "1", myString("2"), "3") + _ = s1 + + type myStruct struct{x int} + var s2 myStruct + g3(nil, struct{x int}{}, myStruct{}) + g3(&s2, struct{x int}{}, myStruct{}) + g3(nil, myStruct{}, struct{x int}{}) + g3(&s2, myStruct{}, struct{x int}{}) +} + +// Here's a realistic example. + +func append[T any](s []T, t ...T) []T { panic(0) } + +func _() { + var f func() + type Funcs []func() + var funcs Funcs + _ = append(funcs, f) +} + +// Generic type declarations cannot have empty type parameter lists +// (that would indicate a slice type). Thus, generic functions cannot +// have empty type parameter lists, either. This is a syntax error. + +func h[] /* ERROR "empty type parameter list" */ () {} + +func _() { + h /* ERROR "cannot index" */ [] /* ERROR "operand" */ () +} + +// Generic functions must have a function body. + +func _ /* ERROR "generic function is missing function body" */ [P any]() diff --git a/src/internal/types/testdata/examples/inference.go b/src/internal/types/testdata/examples/inference.go new file mode 100644 index 0000000..0aaaa82 --- /dev/null +++ b/src/internal/types/testdata/examples/inference.go @@ -0,0 +1,163 @@ +// -lang=go1.20 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of type inference. + +package p + +type Ordered interface { + ~int | ~float64 | ~string +} + +func min[T Ordered](x, y T) T { panic(0) } + +func _() { + // min can be called with explicit instantiation. + _ = min[int](1, 2) + + // Alternatively, the type argument can be inferred from + // one of the arguments. Untyped arguments will be considered + // last. + var x int + _ = min(x, x) + _ = min(x, 1) + _ = min(x, 1.0) + _ = min(1, 2) + _ = min(1, 2.3) + + var y float64 + _ = min(1, y) + _ = min(1.2, y) + _ = min(1.2, 3.4) + _ = min(1.2, 3) + + var s string + _ = min(s, "foo") + _ = min("foo", "bar") +} + +func mixed[T1, T2, T3 any](T1, T2, T3) {} + +func _() { + // mixed can be called with explicit instantiation. + mixed[int, string, bool](0, "", false) + + // Alternatively, partial type arguments may be provided + // (from left to right), and the other may be inferred. + mixed[int, string](0, "", false) + mixed[int](0, "", false) + mixed(0, "", false) + + // Provided type arguments always take precedence over + // inferred types. + mixed[int, string](1.1 /* ERROR "cannot use 1.1" */, "", false) +} + +func related1[Slice interface{ ~[]Elem }, Elem any](s Slice, e Elem) {} + +func _() { + // related1 can be called with explicit instantiation. + var si []int + related1[[]int, int](si, 0) + + // Alternatively, the 2nd type argument can be inferred + // from the first one through constraint type inference. + var ss []string + _ = related1[[]string] + related1[[]string](ss, "foo") + + // A type argument inferred from another explicitly provided + // type argument overrides whatever value argument type is given. + related1[[]string](ss, 0 /* ERROR "cannot use 0" */) + + // A type argument may be inferred from a value argument + // and then help infer another type argument via constraint + // type inference. + related1(si, 0) + related1(si, "foo" /* ERROR `cannot use "foo"` */) +} + +func related2[Elem any, Slice interface{ []Elem }](e Elem, s Slice) {} + +func _() { + // related2 can be called with explicit instantiation. + var si []int + related2[int, []int](0, si) + + // Alternatively, the 2nd type argument can be inferred + // from the first one through constraint type inference. + var ss []string + _ = related2[string] + related2[string]("foo", ss) + + // A type argument may be inferred from a value argument + // and then help infer another type argument via constraint + // type inference. Untyped arguments are always considered + // last. + related2(1.2, []float64{}) + related2(1.0, []int{}) + related2 /* ERROR "Slice (type []int) does not satisfy interface{[]Elem}" */ (float64(1.0), []int{}) // TODO(gri) better error message +} + +type List[P any] []P + +func related3[Elem any, Slice []Elem | List[Elem]]() Slice { return nil } + +func _() { + // related3 can be instantiated explicitly + related3[int, []int]() + related3[byte, List[byte]]() + + // The 2nd type argument cannot be inferred from the first + // one because there's two possible choices: []Elem and + // List[Elem]. + related3 /* ERROR "cannot infer Slice" */ [int]() +} + +func wantsMethods[P interface { + m1(Q) + m2() R +}, Q, R any](P) { +} + +type hasMethods1 struct{} + +func (hasMethods1) m1(int) +func (hasMethods1) m2() string + +type hasMethods2 struct{} + +func (*hasMethods2) m1(int) +func (*hasMethods2) m2() string + +type hasMethods3 interface { + m1(float64) + m2() complex128 +} + +type hasMethods4 interface { + m1() +} + +func _() { + // wantsMethod can be called with arguments that have the relevant methods + // and wantsMethod's type arguments are inferred from those types' method + // signatures. + wantsMethods(hasMethods1{}) + wantsMethods(&hasMethods1{}) + wantsMethods /* ERROR "P (type hasMethods2) does not satisfy interface{m1(Q); m2() R} (method m1 has pointer receiver)" */ (hasMethods2{}) + wantsMethods(&hasMethods2{}) + wantsMethods(hasMethods3(nil)) + wantsMethods /* ERROR "P (type any) does not satisfy interface{m1(Q); m2() R} (missing method m1)" */ (any(nil)) + wantsMethods /* ERROR "P (type hasMethods4) does not satisfy interface{m1(Q); m2() R} (wrong type for method m1)" */ (hasMethods4(nil)) +} + +// "Reverse" type inference is not yet permitted. + +func f[P any](P) {} + +// This must not crash. +var _ func(int) = f // ERROR "implicitly instantiated function in assignment requires go1.21 or later" diff --git a/src/internal/types/testdata/examples/inference2.go b/src/internal/types/testdata/examples/inference2.go new file mode 100644 index 0000000..6097c2b --- /dev/null +++ b/src/internal/types/testdata/examples/inference2.go @@ -0,0 +1,100 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of "reverse" type inference +// where the type arguments for generic functions are determined +// from assigning the functions. + +package p + +func f1[P any](P) {} +func f2[P any]() P { var x P; return x } +func f3[P, Q any](P) Q { var x Q; return x } +func f4[P any](P, P) {} +func f5[P any](P) []P { return nil } +func f6[P any](int) P { var x P; return x } +func f7[P any](P) string { return "" } + +// initialization expressions +var ( + v1 = f1 // ERROR "cannot use generic function f1 without instantiation" + v2 func(int) = f2 // ERROR "cannot infer P" + + v3 func(int) = f1 + v4 func() int = f2 + v5 func(int) int = f3 + _ func(int) int = f3[int] + + v6 func(int, int) = f4 + v7 func(int, string) = f4 // ERROR "type func(int, string) of variable in assignment does not match inferred type func(int, int) for func(P, P)" + v8 func(int) []int = f5 + v9 func(string) []int = f5 // ERROR "type func(string) []int of variable in assignment does not match inferred type func(string) []string for func(P) []P" + + _, _ func(int) = f1, f1 + _, _ func(int) = f1, f2 // ERROR "cannot infer P" +) + +// Regular assignments +func _() { + v1 = f1 // no error here because v1 is invalid (we don't know its type) due to the error above + var v1_ func() int + _ = v1_ + v1_ = f1 // ERROR "cannot infer P" + v2 = f2 // ERROR "cannot infer P" + + v3 = f1 + v4 = f2 + v5 = f3 + v5 = f3[int] + + v6 = f4 + v7 = f4 // ERROR "type func(int, string) of variable in assignment does not match inferred type func(int, int) for func(P, P)" + v8 = f5 + v9 = f5 // ERROR "type func(string) []int of variable in assignment does not match inferred type func(string) []string for func(P) []P" +} + +// Return statements +func _() func(int) { return f1 } +func _() func() int { return f2 } +func _() func(int) int { return f3 } +func _() func(int) int { return f3[int] } + +func _() func(int, int) { return f4 } +func _() func(int, string) { + return f4 /* ERROR "type func(int, string) of variable in assignment does not match inferred type func(int, int) for func(P, P)" */ +} +func _() func(int) []int { return f5 } +func _() func(string) []int { + return f5 /* ERROR "type func(string) []int of variable in assignment does not match inferred type func(string) []string for func(P) []P" */ +} + +func _() (_, _ func(int)) { return f1, f1 } +func _() (_, _ func(int)) { return f1, f2 /* ERROR "cannot infer P" */ } + +// Argument passing +func g1(func(int)) {} +func g2(func(int, int)) {} +func g3(func(int) string) {} +func g4[P any](func(P) string) {} +func g5[P, Q any](func(P) string, func(P) Q) {} +func g6(func(int), func(string)) {} + +func _() { + g1(f1) + g1(f2 /* ERROR "cannot infer P" */) + g2(f4) + g4(f6) + g5(f6, f7) + g6(f1, f1) +} + +// Argument passing of partially instantiated functions +func h(func(int, string), func(string, int)) {} + +func p[P, Q any](P, Q) {} + +func _() { + h(p, p) + h(p[int], p[string]) +} diff --git a/src/internal/types/testdata/examples/methods.go b/src/internal/types/testdata/examples/methods.go new file mode 100644 index 0000000..e92dc50 --- /dev/null +++ b/src/internal/types/testdata/examples/methods.go @@ -0,0 +1,112 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of methods on type-parameterized types. + +package p + +// Parameterized types may have methods. +type T1[A any] struct{ a A } + +// When declaring a method for a parameterized type, the "instantiated" +// receiver type acts as an implicit declaration of the type parameters +// for the receiver type. In the example below, method m1 on type T1 has +// the receiver type T1[A] which declares the type parameter A for use +// with this method. That is, within the method m1, A stands for the +// actual type argument provided to an instantiated T1. +func (t T1[A]) m1() A { return t.a } + +// For instance, if T1 is instantiated with the type int, the type +// parameter A in m1 assumes that type (int) as well and we can write +// code like this: +var x T1[int] +var _ int = x.m1() + +// Because the type parameter provided to a parameterized receiver type +// is declared through that receiver declaration, it must be an identifier. +// It cannot possibly be some other type because the receiver type is not +// instantiated with concrete types, it is standing for the parameterized +// receiver type. +func (t T1[[ /* ERROR "must be an identifier" */ ]int]) m2() {} + +// Note that using what looks like a predeclared identifier, say int, +// as type parameter in this situation is deceptive and considered bad +// style. In m3 below, int is the name of the local receiver type parameter +// and it shadows the predeclared identifier int which then cannot be used +// anymore as expected. +// This is no different from locally re-declaring a predeclared identifier +// and usually should be avoided. There are some notable exceptions; e.g., +// sometimes it makes sense to use the identifier "copy" which happens to +// also be the name of a predeclared built-in function. +func (t T1[int]) m3() { var _ int = 42 /* ERRORx `cannot use 42 .* as int` */ } + +// The names of the type parameters used in a parameterized receiver +// type don't have to match the type parameter names in the declaration +// of the type used for the receiver. In our example, even though T1 is +// declared with type parameter named A, methods using that receiver type +// are free to use their own name for that type parameter. That is, the +// name of type parameters is always local to the declaration where they +// are introduced. In our example we can write a method m2 and use the +// name X instead of A for the type parameter w/o any difference. +func (t T1[X]) m4() X { return t.a } + +// If the receiver type is parameterized, type parameters must always be +// provided: this simply follows from the general rule that a parameterized +// type must be instantiated before it can be used. A method receiver +// declaration using a parameterized receiver type is no exception. It is +// simply that such receiver type expressions perform two tasks simultaneously: +// they declare the (local) type parameters and then use them to instantiate +// the receiver type. Forgetting to provide a type parameter leads to an error. +func (t T1 /* ERRORx `generic type .* without instantiation` */ ) m5() {} + +// However, sometimes we don't need the type parameter, and thus it is +// inconvenient to have to choose a name. Since the receiver type expression +// serves as a declaration for its type parameters, we are free to choose the +// blank identifier: +func (t T1[_]) m6() {} + +// Naturally, these rules apply to any number of type parameters on the receiver +// type. Here are some more complex examples. +type T2[A, B, C any] struct { + a A + b B + c C +} + +// Naming of the type parameters is local and has no semantic impact: +func (t T2[A, B, C]) m1() (A, B, C) { return t.a, t.b, t.c } +func (t T2[C, B, A]) m2() (C, B, A) { return t.a, t.b, t.c } +func (t T2[X, Y, Z]) m3() (X, Y, Z) { return t.a, t.b, t.c } + +// Type parameters may be left blank if they are not needed: +func (t T2[A, _, C]) m4() (A, C) { return t.a, t.c } +func (t T2[_, _, X]) m5() X { return t.c } +func (t T2[_, _, _]) m6() {} + +// As usual, blank names may be used for any object which we don't care about +// using later. For instance, we may write an unnamed method with a receiver +// that cannot be accessed: +func (_ T2[_, _, _]) _() int { return 42 } + +// Because a receiver parameter list is simply a parameter list, we can +// leave the receiver argument away for receiver types. +type T0 struct{} +func (T0) _() {} +func (T1[A]) _() {} + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// // A generic receiver type may constrain its type parameter such +// // that it must be a pointer type. Such receiver types are not +// // permitted. +// type T3a[P interface{ ~int | ~string | ~float64 }] P +// +// func (T3a[_]) m() {} // this is ok +// +// type T3b[P interface{ ~unsafe.Pointer }] P +// +// func (T3b /* ERROR "invalid receiver" */ [_]) m() {} +// +// type T3c[P interface{ *int | *string }] P +// +// func (T3c /* ERROR "invalid receiver" */ [_]) m() {} diff --git a/src/internal/types/testdata/examples/operations.go b/src/internal/types/testdata/examples/operations.go new file mode 100644 index 0000000..9fb95d0 --- /dev/null +++ b/src/internal/types/testdata/examples/operations.go @@ -0,0 +1,29 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// indirection + +func _[P any](p P) { + _ = *p // ERROR "cannot indirect p" +} + +func _[P interface{ int }](p P) { + _ = *p // ERROR "cannot indirect p" +} + +func _[P interface{ *int }](p P) { + _ = *p +} + +func _[P interface{ *int | *string }](p P) { + _ = *p // ERROR "must have identical base types" +} + +type intPtr *int + +func _[P interface{ *int | intPtr } ](p P) { + var _ int = *p +} diff --git a/src/internal/types/testdata/examples/types.go b/src/internal/types/testdata/examples/types.go new file mode 100644 index 0000000..67f1534 --- /dev/null +++ b/src/internal/types/testdata/examples/types.go @@ -0,0 +1,315 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of generic types. + +package p + +// List is just what it says - a slice of E elements. +type List[E any] []E + +// A generic (parameterized) type must always be instantiated +// before it can be used to designate the type of a variable +// (including a struct field, or function parameter); though +// for the latter cases, the provided type may be another type +// parameter. So: +var _ List[byte] = []byte{} + +// A generic binary tree might be declared as follows. +type Tree[E any] struct { + left, right *Tree[E] + payload E +} + +// A simple instantiation of Tree: +var root1 Tree[int] + +// The actual type parameter provided may be a generic type itself: +var root2 Tree[List[int]] + +// A couple of more complex examples. +// We don't need extra parentheses around the element type of the slices on +// the right (unlike when we use ()'s rather than []'s for type parameters). +var _ List[List[int]] = []List[int]{} +var _ List[List[List[Tree[int]]]] = []List[List[Tree[int]]]{} + +// Type parameters act like type aliases when used in generic types +// in the sense that we can "emulate" a specific type instantiation +// with type aliases. +type T1[P any] struct { + f P +} + +type T2[P any] struct { + f struct { + g P + } +} + +var x1 T1[struct{ g int }] +var x2 T2[int] + +func _() { + // This assignment is invalid because the types of x1, x2 are T1(...) + // and T2(...) respectively, which are two different defined types. + x1 = x2 // ERROR "assignment" + + // This assignment is valid because the types of x1.f and x2.f are + // both struct { g int }; the type parameters act like type aliases + // and their actual names don't come into play here. + x1.f = x2.f +} + +// We can verify this behavior using type aliases instead: +type T1a struct { + f A1 +} +type A1 = struct { g int } + +type T2a struct { + f struct { + g A2 + } +} +type A2 = int + +var x1a T1a +var x2a T2a + +func _() { + x1a = x2a // ERROR "assignment" + x1a.f = x2a.f +} + +// Another interesting corner case are generic types that don't use +// their type arguments. For instance: +type T[P any] struct{} + +var xint T[int] +var xbool T[bool] + +// Are these two variables of the same type? After all, their underlying +// types are identical. We consider them to be different because each type +// instantiation creates a new named type, in this case T<int> and T<bool> +// even if their underlying types are identical. This is sensible because +// we might still have methods that have different signatures or behave +// differently depending on the type arguments, and thus we can't possibly +// consider such types identical. Consequently: +func _() { + xint = xbool // ERROR "assignment" +} + +// Generic types cannot be used without instantiation. +var _ T // ERROR "cannot use generic type T" +var _ = T /* ERROR "cannot use generic type T" */ (0) + +// In type context, generic (parameterized) types cannot be parenthesized before +// being instantiated. See also NOTES entry from 12/4/2019. +var _ (T /* ERROR "cannot use generic type T" */ )[ /* ERRORx `unexpected \[|expected ';'` */ int] + +// All types may be parameterized, including interfaces. +type I1[T any] interface{ + m1(T) +} + +// There is no such thing as a variadic generic type. +type _[T ... /* ERROR "invalid use of '...'" */ any] struct{} + +// Generic interfaces may be embedded as one would expect. +type I2 interface { + I1(int) // method! + I1[string] // embedded I1 +} + +func _() { + var x I2 + x.I1(0) + x.m1("foo") +} + +type I0 interface { + m0() +} + +type I3 interface { + I0 + I1[bool] + m(string) +} + +func _() { + var x I3 + x.m0() + x.m1(true) + x.m("foo") +} + +type _ struct { + ( /* ERROR "cannot parenthesize" */ int8) + ( /* ERROR "cannot parenthesize" */ *int16) + *( /* ERROR "cannot parenthesize" */ int32) + List[int] + + int8 /* ERROR "int8 redeclared" */ + *int16 /* ERROR "int16 redeclared" */ + List /* ERROR "List redeclared" */ [int] +} + +// Issue #45639: We don't allow this anymore. Keep this code +// in case we decide to revisit this decision. +// +// It's possible to declare local types whose underlying types +// are type parameters. As with ordinary type definitions, the +// types underlying properties are "inherited" but the methods +// are not. +// func _[T interface{ m(); ~int }]() { +// type L T +// var x L +// +// // m is not defined on L (it is not "inherited" from +// // its underlying type). +// x.m /* ERROR "x.m undefined" */ () +// +// // But the properties of T, such that as that it supports +// // the operations of the types given by its type bound, +// // are also the properties of L. +// x++ +// _ = x - x +// +// // On the other hand, if we define a local alias for T, +// // that alias stands for T as expected. +// type A = T +// var y A +// y.m() +// _ = y < 0 +// } + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// // It is not permitted to declare a local type whose underlying +// // type is a type parameter not declared by that type declaration. +// func _[T any]() { +// type _ T // ERROR "cannot use function type parameter T as RHS in type declaration" +// type _ [_ any] T // ERROR "cannot use function type parameter T as RHS in type declaration" +// } + +// As a special case, an explicit type argument may be omitted +// from a type parameter bound if the type bound expects exactly +// one type argument. In that case, the type argument is the +// respective type parameter to which the type bound applies. +// Note: We may not permit this syntactic sugar at first. +// Note: This is now disabled. All examples below are adjusted. +type Adder[T any] interface { + Add(T) T +} + +// We don't need to explicitly instantiate the Adder bound +// if we have exactly one type parameter. +func Sum[T Adder[T]](list []T) T { + var sum T + for _, x := range list { + sum = sum.Add(x) + } + return sum +} + +// Valid and invalid variations. +type B0 any +type B1[_ any] any +type B2[_, _ any] any + +func _[T1 B0]() {} +func _[T1 B1[T1]]() {} +func _[T1 B2 /* ERRORx `cannot use generic type .* without instantiation` */ ]() {} + +func _[T1, T2 B0]() {} +func _[T1 B1[T1], T2 B1[T2]]() {} +func _[T1, T2 B2 /* ERRORx `cannot use generic type .* without instantiation` */ ]() {} + +func _[T1 B0, T2 B1[T2]]() {} // here B1 applies to T2 + +// When the type argument is left away, the type bound is +// instantiated for each type parameter with that type +// parameter. +// Note: We may not permit this syntactic sugar at first. +func _[A Adder[A], B Adder[B], C Adder[A]]() { + var a A // A's type bound is Adder[A] + a = a.Add(a) + var b B // B's type bound is Adder[B] + b = b.Add(b) + var c C // C's type bound is Adder[A] + a = c.Add(a) +} + +// The type of variables (incl. parameters and return values) cannot +// be an interface with type constraints or be/embed comparable. +type I interface { + ~int +} + +var ( + _ interface /* ERROR "contains type constraints" */ {~int} + _ I /* ERROR "contains type constraints" */ +) + +func _(I /* ERROR "contains type constraints" */ ) +func _(x, y, z I /* ERROR "contains type constraints" */ ) +func _() I /* ERROR "contains type constraints" */ + +func _() { + var _ I /* ERROR "contains type constraints" */ +} + +type C interface { + comparable +} + +var _ comparable /* ERROR "comparable" */ +var _ C /* ERROR "comparable" */ + +func _(_ comparable /* ERROR "comparable" */ , _ C /* ERROR "comparable" */ ) + +func _() { + var _ comparable /* ERROR "comparable" */ + var _ C /* ERROR "comparable" */ +} + +// Type parameters are never const types, i.e., it's +// not possible to declare a constant of type parameter type. +// (If a type set contains just a single const type, we could +// allow it, but such type sets don't make much sense in the +// first place.) +func _[T interface{~int|~float64}]() { + // not valid + const _ = T /* ERROR "not constant" */ (0) + const _ T /* ERROR "invalid constant type T" */ = 1 + + // valid + var _ = T(0) + var _ T = 1 + _ = T(0) +} + +// It is possible to create composite literals of type parameter +// type as long as it's possible to create a composite literal +// of the core type of the type parameter's constraint. +func _[P interface{ ~[]int }]() P { + return P{} + return P{1, 2, 3} +} + +func _[P interface{ ~[]E }, E interface{ map[string]P } ]() P { + x := P{} + return P{{}} + return P{E{}} + return P{E{"foo": x}} + return P{{"foo": x}, {}} +} + +// This is a degenerate case with a singleton type set, but we can create +// composite literals even if the core type is a defined type. +type MyInts []int + +func _[P MyInts]() P { + return P{} +} diff --git a/src/internal/types/testdata/examples/typesets.go b/src/internal/types/testdata/examples/typesets.go new file mode 100644 index 0000000..93eada7 --- /dev/null +++ b/src/internal/types/testdata/examples/typesets.go @@ -0,0 +1,59 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file shows some examples of constraint literals with elided interfaces. +// These examples are permitted if proposal issue #48424 is accepted. + +package p + +// Constraint type sets of the form T, ~T, or A|B may omit the interface. +type ( + _[T int] struct{} + _[T ~int] struct{} + _[T int | string] struct{} + _[T ~int | ~string] struct{} +) + +func min[T int | string](x, y T) T { + if x < y { + return x + } + return y +} + +func lookup[M ~map[K]V, K comparable, V any](m M, k K) V { + return m[k] +} + +func deref[P ~*E, E any](p P) E { + return *p +} + +func _() int { + p := new(int) + return deref(p) +} + +func addrOfCopy[V any, P *V](v V) P { + return &v +} + +func _() *int { + return addrOfCopy(0) +} + +// A type parameter may not be embedded in an interface; +// so it can also not be used as a constraint. +func _[A any, B A /* ERROR "cannot use a type parameter as constraint" */]() {} +func _[A any, B, C A /* ERROR "cannot use a type parameter as constraint" */]() {} + +// Error messages refer to the type constraint as it appears in the source. +// (No implicit interface should be exposed.) +func _[T string](x T) T { + return x /* ERROR "constrained by string" */ * x +} + +func _[T int | string](x T) T { + return x /* ERROR "constrained by int | string" */ * x +} diff --git a/src/internal/types/testdata/fixedbugs/issue20583.go b/src/internal/types/testdata/fixedbugs/issue20583.go new file mode 100644 index 0000000..55cbd94 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue20583.go @@ -0,0 +1,14 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue20583 + +const ( + _ = 6e886451608 /* ERROR "malformed constant" */ /2 + _ = 6e886451608i /* ERROR "malformed constant" */ /2 + _ = 0 * 1e+1000000000 // ERROR "malformed constant" + + x = 1e100000000 + _ = x*x*x*x*x*x* /* ERROR "not representable" */ x +) diff --git a/src/internal/types/testdata/fixedbugs/issue23203a.go b/src/internal/types/testdata/fixedbugs/issue23203a.go new file mode 100644 index 0000000..48cb588 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue23203a.go @@ -0,0 +1,14 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +import "unsafe" + +type T struct{} + +func (T) m1() {} +func (T) m2([unsafe.Sizeof(T.m1)]int) {} + +func main() {} diff --git a/src/internal/types/testdata/fixedbugs/issue23203b.go b/src/internal/types/testdata/fixedbugs/issue23203b.go new file mode 100644 index 0000000..638ec6c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue23203b.go @@ -0,0 +1,14 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +import "unsafe" + +type T struct{} + +func (T) m2([unsafe.Sizeof(T.m1)]int) {} +func (T) m1() {} + +func main() {} diff --git a/src/internal/types/testdata/fixedbugs/issue25838.go b/src/internal/types/testdata/fixedbugs/issue25838.go new file mode 100644 index 0000000..adbd138 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue25838.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// examples from the issue + +type ( + e = f + f = g + g = []h + h i + i = j + j = e +) + +type ( + e1 = []h1 + h1 e1 +) + +type ( + P = *T + T P +) diff --git a/src/internal/types/testdata/fixedbugs/issue26390.go b/src/internal/types/testdata/fixedbugs/issue26390.go new file mode 100644 index 0000000..9e0101f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue26390.go @@ -0,0 +1,13 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// stand-alone test to ensure case is triggered + +package issue26390 + +type A = T + +func (t *T) m() *A { return t } + +type T struct{} diff --git a/src/internal/types/testdata/fixedbugs/issue28251.go b/src/internal/types/testdata/fixedbugs/issue28251.go new file mode 100644 index 0000000..77fd369 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue28251.go @@ -0,0 +1,65 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file contains test cases for various forms of +// method receiver declarations, per the spec clarification +// https://golang.org/cl/142757. + +package issue28251 + +// test case from issue28251 +type T struct{} + +type T0 = *T + +func (T0) m() {} + +func _() { (&T{}).m() } + +// various alternative forms +type ( + T1 = (((T))) +) + +func ((*(T1))) m1() {} +func _() { (T{}).m2() } +func _() { (&T{}).m2() } + +type ( + T2 = (((T3))) + T3 = T +) + +func (T2) m2() {} +func _() { (T{}).m2() } +func _() { (&T{}).m2() } + +type ( + T4 = ((*(T5))) + T5 = T +) + +func (T4) m4() {} +func _() { (T{}).m4 /* ERROR "cannot call pointer method m4 on T" */ () } +func _() { (&T{}).m4() } + +type ( + T6 = (((T7))) + T7 = (*(T8)) + T8 = T +) + +func (T6) m6() {} +func _() { (T{}).m6 /* ERROR "cannot call pointer method m6 on T" */ () } +func _() { (&T{}).m6() } + +type ( + T9 = *T10 + T10 = *T11 + T11 = T +) + +func (T9 /* ERROR "invalid receiver type **T" */ ) m9() {} +func _() { (T{}).m9 /* ERROR "has no field or method m9" */ () } +func _() { (&T{}).m9 /* ERROR "has no field or method m9" */ () } diff --git a/src/internal/types/testdata/fixedbugs/issue3117.go b/src/internal/types/testdata/fixedbugs/issue3117.go new file mode 100644 index 0000000..16c0afc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue3117.go @@ -0,0 +1,13 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S struct { + a [1]int +} + +func _(m map[int]S, key int) { + m /* ERROR "cannot assign to m[key].a[0] (neither addressable nor a map index expression)" */ [key].a[0] = 0 +} diff --git a/src/internal/types/testdata/fixedbugs/issue39634.go b/src/internal/types/testdata/fixedbugs/issue39634.go new file mode 100644 index 0000000..591b00e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39634.go @@ -0,0 +1,90 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Examples adjusted to match new [T any] syntax for type parameters. +// Also, previously permitted empty type parameter lists and instantiations +// are now syntax errors. + +package p + +// crash 1 +type nt1[_ any]interface{g /* ERROR "undefined" */ } +type ph1[e nt1[e],g(d /* ERROR "undefined" */ )]s /* ERROR "undefined" */ +func(*ph1[e,e /* ERROR "redeclared" */ ])h(d /* ERROR "undefined" */ ) + +// crash 2 +// Disabled: empty []'s are now syntax errors. This example leads to too many follow-on errors. +// type Numeric2 interface{t2 /* ERROR "not a type" */ } +// func t2[T Numeric2](s[]T){0 /* ERROR "not a type */ []{s /* ERROR cannot index" */ [0][0]}} + +// crash 3 +type t3 *interface{ t3.p /* ERROR "t3.p is not a type" */ } + +// crash 4 +type Numeric4 interface{t4 /* ERROR "not a type" */ } +func t4[T Numeric4](s[]T){if( /* ERROR "non-boolean" */ 0){*s /* ERROR "cannot indirect" */ [0]}} + +// crash 7 +type foo7 interface { bar() } +type x7[A any] struct{ foo7 } +func main7() { var _ foo7 = x7[int]{} } + +// crash 8 +type foo8[A any] interface { ~A /* ERROR "cannot be a type parameter" */ } +func bar8[A foo8[A]](a A) {} + +// crash 9 +type foo9[A any] interface { foo9 /* ERROR "invalid recursive type" */ [A] } +func _() { var _ = new(foo9[int]) } + +// crash 12 +var u /* ERROR "cycle" */ , i [func /* ERROR "used as value" */ /* ERROR "used as value" */ (u, c /* ERROR "undefined" */ /* ERROR "undefined" */ ) {}(0, len /* ERROR "must be called" */ /* ERROR "must be called" */ )]c /* ERROR "undefined" */ /* ERROR "undefined" */ + +// crash 15 +func y15() { var a /* ERROR "declared and not used" */ interface{ p() } = G15[string]{} } +type G15[X any] s /* ERROR "undefined" */ +func (G15 /* ERRORx `generic type .* without instantiation` */ ) p() + +// crash 16 +type Foo16[T any] r16 /* ERROR "not a type" */ +func r16[T any]() Foo16[Foo16[T]] { panic(0) } + +// crash 17 +type Y17 interface{ c() } +type Z17 interface { + c() Y17 + Y17 /* ERROR "duplicate method" */ +} +func F17[T Z17](T) {} + +// crash 18 +type o18[T any] []func(_ o18[[]_ /* ERROR "cannot use _" */ ]) + +// crash 19 +type Z19 [][[]Z19{}[0][0]]c19 /* ERROR "undefined" */ + +// crash 20 +type Z20 /* ERROR "invalid recursive type" */ interface{ Z20 } +func F20[t Z20]() { F20(t /* ERROR "invalid composite literal type" */ /* ERROR "too many arguments in call to F20\n\thave (unknown type)\n\twant ()" */ {}) } + +// crash 21 +type Z21 /* ERROR "invalid recursive type" */ interface{ Z21 } +func F21[T Z21]() { ( /* ERROR "not used" */ F21[Z21]) } + +// crash 24 +type T24[P any] P // ERROR "cannot use a type parameter as RHS in type declaration" +func (r T24[P]) m() { T24 /* ERROR "without instantiation" */ .m() } + +// crash 25 +type T25[A any] int +func (t T25[A]) m1() {} +var x T25 /* ERROR "without instantiation" */ .m1 + +// crash 26 +type T26 = interface{ F26[ /* ERROR "interface method must have no type parameters" */ Z any]() } +func F26[Z any]() T26 { return F26[] /* ERROR "operand" */ } + +// crash 27 +func e27[T any]() interface{ x27 /* ERROR "not a type" */ } { panic(0) } +func x27() { e27 /* ERROR "cannot infer T" */ () } diff --git a/src/internal/types/testdata/fixedbugs/issue39664.go b/src/internal/types/testdata/fixedbugs/issue39664.go new file mode 100644 index 0000000..a8148c6 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39664.go @@ -0,0 +1,15 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[_ any] struct {} + +func (T /* ERROR "instantiation" */ ) m() + +func _() { + var x interface { m() } + x = T[int]{} + _ = x +} diff --git a/src/internal/types/testdata/fixedbugs/issue39680.go b/src/internal/types/testdata/fixedbugs/issue39680.go new file mode 100644 index 0000000..e56bc35 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39680.go @@ -0,0 +1,31 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Embedding stand-alone type parameters is not permitted for now. Disabled. + +/* +import "fmt" + +// Minimal test case. +func _[T interface{~T}](x T) T{ + return x +} + +// Test case from issue. +type constr[T any] interface { + ~T +} + +func Print[T constr[T]](s []T) { + for _, v := range s { + fmt.Print(v) + } +} + +func f() { + Print([]string{"Hello, ", "playground\n"}) +} +*/ diff --git a/src/internal/types/testdata/fixedbugs/issue39693.go b/src/internal/types/testdata/fixedbugs/issue39693.go new file mode 100644 index 0000000..03f2789 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39693.go @@ -0,0 +1,23 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Number1 interface { + // embedding non-interface types is permitted + int + float64 +} + +func Add1[T Number1](a, b T) T { + return a /* ERROR "not defined" */ + b +} + +type Number2 interface { + int | float64 +} + +func Add2[T Number2](a, b T) T { + return a + b +} diff --git a/src/internal/types/testdata/fixedbugs/issue39699.go b/src/internal/types/testdata/fixedbugs/issue39699.go new file mode 100644 index 0000000..73ba0c4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39699.go @@ -0,0 +1,29 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T0 interface{ +} + +type T1 interface{ + ~int +} + +type T2 interface{ + comparable +} + +type T3 interface { + T0 + T1 + T2 +} + +func _() { + _ = T0(0) + _ = T1 /* ERROR "cannot use interface T1 in conversion" */ (1) + _ = T2 /* ERROR "cannot use interface T2 in conversion" */ (2) + _ = T3 /* ERROR "cannot use interface T3 in conversion" */ (3) +} diff --git a/src/internal/types/testdata/fixedbugs/issue39711.go b/src/internal/types/testdata/fixedbugs/issue39711.go new file mode 100644 index 0000000..d85fa03 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39711.go @@ -0,0 +1,13 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Do not report a duplicate type error for this term list. +// (Check types after interfaces have been completed.) +type _ interface { + // TODO(rfindley) Once we have full type sets we can enable this again. + // Fow now we don't permit interfaces in term lists. + // type interface{ Error() string }, interface{ String() string } +} diff --git a/src/internal/types/testdata/fixedbugs/issue39723.go b/src/internal/types/testdata/fixedbugs/issue39723.go new file mode 100644 index 0000000..19e5e80 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39723.go @@ -0,0 +1,9 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// A constraint must be an interface; it cannot +// be a type parameter, for instance. +func _[A interface{ ~int }, B A /* ERROR "cannot use a type parameter as constraint" */ ]() {} diff --git a/src/internal/types/testdata/fixedbugs/issue39725.go b/src/internal/types/testdata/fixedbugs/issue39725.go new file mode 100644 index 0000000..0145667 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39725.go @@ -0,0 +1,16 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[T1, T2 any](T1, T2, struct{a T1; b T2}) {} +func _() { + f1(42, string("foo"), struct /* ERROR "does not match inferred type struct{a int; b string}" */ {a, b int}{}) +} + +// simplified test case from issue +func f2[T any](_ []T, _ func(T)) {} +func _() { + f2([]string{}, func /* ERROR "does not match inferred type func(string)" */ (f []byte) {}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue39754.go b/src/internal/types/testdata/fixedbugs/issue39754.go new file mode 100644 index 0000000..a1bd5ba --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39754.go @@ -0,0 +1,21 @@ +// Copyright 2020 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package p
+
+type Optional[T any] struct {}
+
+func (_ Optional[T]) Val() (T, bool)
+
+type Box[T any] interface {
+ Val() (T, bool)
+}
+
+func f[V interface{}, A, B Box[V]]() {}
+
+func _() {
+ f[int, Optional[int], Optional[int]]()
+ _ = f[int, Optional[int], Optional /* ERROR "does not satisfy Box" */ [string]]
+ _ = f[int, Optional[int], Optional /* ERRORx "Optional.* does not satisfy Box.*" */ [string]]
+}
diff --git a/src/internal/types/testdata/fixedbugs/issue39755.go b/src/internal/types/testdata/fixedbugs/issue39755.go new file mode 100644 index 0000000..257b73a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39755.go @@ -0,0 +1,23 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[T interface{~map[string]int}](x T) { + _ = x == nil +} + +// simplified test case from issue + +type PathParamsConstraint interface { + ~map[string]string | ~[]struct{key, value string} +} + +type PathParams[T PathParamsConstraint] struct { + t T +} + +func (pp *PathParams[T]) IsNil() bool { + return pp.t == nil // this must succeed +} diff --git a/src/internal/types/testdata/fixedbugs/issue39768.go b/src/internal/types/testdata/fixedbugs/issue39768.go new file mode 100644 index 0000000..51a4177 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39768.go @@ -0,0 +1,21 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type T[P any] P +// type A = T // ERROR "cannot use generic type" +// var x A[int] +// var _ A +// +// type B = T[int] +// var y B = x +// var _ B /* ERROR "not a generic type" */ [int] + +// test case from issue + +type Vector[T any] []T +type VectorAlias = Vector // ERROR "cannot use generic type" +var v Vector[int] diff --git a/src/internal/types/testdata/fixedbugs/issue39938.go b/src/internal/types/testdata/fixedbugs/issue39938.go new file mode 100644 index 0000000..bd5bdca --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39938.go @@ -0,0 +1,54 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// All but E2 and E5 provide an "indirection" and break infinite expansion of a type. +type E0[P any] []P +type E1[P any] *P +type E2[P any] struct{ _ P } +type E3[P any] struct{ _ *P } +type E5[P any] struct{ _ [10]P } + +type T0 struct { + _ E0[T0] +} + +type T0_ struct { + E0[T0_] +} + +type T1 struct { + _ E1[T1] +} + +type T2 /* ERROR "invalid recursive type" */ struct { + _ E2[T2] +} + +type T3 struct { + _ E3[T3] +} + +type T4 /* ERROR "invalid recursive type" */ [10]E5[T4] + +type T5 struct { + _ E0[E2[T5]] +} + +type T6 struct { + _ E0[E2[E0[E1[E2[[10]T6]]]]] +} + +type T7 struct { + _ E0[[10]E2[E0[E2[E2[T7]]]]] +} + +type T8 struct { + _ E0[[]E2[E0[E2[E2[T8]]]]] +} + +type T9 /* ERROR "invalid recursive type" */ [10]E2[E5[E2[T9]]] + +type T10 [10]E2[E5[E2[func(T10)]]] diff --git a/src/internal/types/testdata/fixedbugs/issue39948.go b/src/internal/types/testdata/fixedbugs/issue39948.go new file mode 100644 index 0000000..e4430cf --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39948.go @@ -0,0 +1,9 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] interface{ + P // ERROR "term cannot be a type parameter" +} diff --git a/src/internal/types/testdata/fixedbugs/issue39976.go b/src/internal/types/testdata/fixedbugs/issue39976.go new file mode 100644 index 0000000..b622cd9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39976.go @@ -0,0 +1,16 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type policy[K, V any] interface{} +type LRU[K, V any] struct{} + +func NewCache[K, V any](p policy[K, V]) {} + +func _() { + var lru LRU[int, string] + NewCache[int, string](&lru) + NewCache /* ERROR "cannot infer K" */ (&lru) +} diff --git a/src/internal/types/testdata/fixedbugs/issue39982.go b/src/internal/types/testdata/fixedbugs/issue39982.go new file mode 100644 index 0000000..9810b63 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue39982.go @@ -0,0 +1,36 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + T[_ any] struct{} + S[_ any] struct { + data T[*T[int]] + } +) + +func _() { + _ = S[int]{ + data: T[*T[int]]{}, + } +} + +// full test case from issue + +type ( + Element[TElem any] struct{} + + entry[K comparable] struct{} + + Cache[K comparable] struct { + data map[K]*Element[*entry[K]] + } +) + +func _() { + _ = Cache[int]{ + data: make(map[int](*Element[*entry[int]])), + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue40038.go b/src/internal/types/testdata/fixedbugs/issue40038.go new file mode 100644 index 0000000..5f81fcb --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40038.go @@ -0,0 +1,15 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type A[T any] int + +func (A[T]) m(A[T]) + +func f[P interface{m(P)}]() {} + +func _() { + _ = f[A[int]] +} diff --git a/src/internal/types/testdata/fixedbugs/issue40056.go b/src/internal/types/testdata/fixedbugs/issue40056.go new file mode 100644 index 0000000..ce882e7 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40056.go @@ -0,0 +1,15 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + NewS /* ERROR "cannot infer T" */ ().M() +} + +type S struct {} + +func NewS[T any]() *S { panic(0) } + +func (_ *S /* ERROR "S is not a generic type" */ [T]) M() diff --git a/src/internal/types/testdata/fixedbugs/issue40057.go b/src/internal/types/testdata/fixedbugs/issue40057.go new file mode 100644 index 0000000..2996d39 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40057.go @@ -0,0 +1,17 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + var x interface{} + switch t := x.(type) { + case S /* ERROR "cannot use generic type" */ : + t.m() + } +} + +type S[T any] struct {} + +func (_ S[T]) m() diff --git a/src/internal/types/testdata/fixedbugs/issue40301.go b/src/internal/types/testdata/fixedbugs/issue40301.go new file mode 100644 index 0000000..c78f9a1 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40301.go @@ -0,0 +1,12 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +func _[T any](x T) { + _ = unsafe.Alignof(x) + _ = unsafe.Sizeof(x) +} diff --git a/src/internal/types/testdata/fixedbugs/issue40350.go b/src/internal/types/testdata/fixedbugs/issue40350.go new file mode 100644 index 0000000..b7ceb33 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40350.go @@ -0,0 +1,16 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type number interface { + ~float64 | ~int | ~int32 + float64 | ~int32 +} + +func f[T number]() {} + +func _() { + _ = f[int /* ERROR "int does not satisfy number (number mentions int, but int is not in the type set of number)" */] +} diff --git a/src/internal/types/testdata/fixedbugs/issue40684.go b/src/internal/types/testdata/fixedbugs/issue40684.go new file mode 100644 index 0000000..4805184 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40684.go @@ -0,0 +1,15 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[_ any] int + +func f[_ any]() {} +func g[_, _ any]() {} + +func _() { + _ = f[T /* ERROR "without instantiation" */ ] + _ = g[T /* ERROR "without instantiation" */ , T /* ERROR "without instantiation" */ ] +} diff --git a/src/internal/types/testdata/fixedbugs/issue40789.go b/src/internal/types/testdata/fixedbugs/issue40789.go new file mode 100644 index 0000000..9eea4ad --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue40789.go @@ -0,0 +1,37 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +import "fmt" + +func main() { + m := map[string]int{ + "a": 6, + "b": 7, + } + fmt.Println(copyMap[map[string]int, string, int](m)) +} + +type Map[K comparable, V any] interface { + map[K] V +} + +func copyMap[M Map[K, V], K comparable, V any](m M) M { + m1 := make(M) + for k, v := range m { + m1[k] = v + } + return m1 +} + +// simpler test case from the same issue + +type A[X comparable] interface { + []X +} + +func f[B A[X], X comparable]() B { + return nil +} diff --git a/src/internal/types/testdata/fixedbugs/issue41124.go b/src/internal/types/testdata/fixedbugs/issue41124.go new file mode 100644 index 0000000..0f828dc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue41124.go @@ -0,0 +1,91 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Test case from issue. + +type Nat /* ERROR "invalid recursive type" */ interface { + Zero|Succ +} + +type Zero struct{} +type Succ struct{ + Nat // Nat contains type constraints but is invalid, so no error +} + +// Struct tests. + +type I1 interface { + comparable +} + +type I2 interface { + ~int +} + +type I3 interface { + I1 + I2 +} + +type _ struct { + f I1 // ERRORx `interface is .* comparable` +} + +type _ struct { + comparable // ERRORx `interface is .* comparable` +} + +type _ struct{ + I1 // ERRORx `interface is .* comparable` +} + +type _ struct{ + I2 // ERROR "interface contains type constraints" +} + +type _ struct{ + I3 // ERROR "interface contains type constraints" +} + +// General composite types. + +type ( + _ [10]I1 // ERRORx `interface is .* comparable` + _ [10]I2 // ERROR "interface contains type constraints" + + _ []I1 // ERRORx `interface is .* comparable` + _ []I2 // ERROR "interface contains type constraints" + + _ *I3 // ERROR "interface contains type constraints" + _ map[I1 /* ERRORx `interface is .* comparable` */ ]I2 // ERROR "interface contains type constraints" + _ chan I3 // ERROR "interface contains type constraints" + _ func(I1 /* ERRORx `interface is .* comparable` */ ) + _ func() I2 // ERROR "interface contains type constraints" +) + +// Other cases. + +var _ = [...]I3 /* ERROR "interface contains type constraints" */ {} + +func _(x interface{}) { + _ = x.(I3 /* ERROR "interface contains type constraints" */ ) +} + +type T1[_ any] struct{} +type T3[_, _, _ any] struct{} +var _ T1[I2 /* ERROR "interface contains type constraints" */ ] +var _ T3[int, I2 /* ERROR "interface contains type constraints" */ , float32] + +func f1[_ any]() int { panic(0) } +var _ = f1[I2 /* ERROR "interface contains type constraints" */ ]() +func f3[_, _, _ any]() int { panic(0) } +var _ = f3[int, I2 /* ERROR "interface contains type constraints" */ , float32]() + +func _(x interface{}) { + switch x.(type) { + case I2 /* ERROR "interface contains type constraints" */ : + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue41176.go b/src/internal/types/testdata/fixedbugs/issue41176.go new file mode 100644 index 0000000..755e83a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue41176.go @@ -0,0 +1,21 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S struct{} + +func (S) M() byte { + return 0 +} + +type I[T any] interface { + M() T +} + +func f[T any](x I[T]) {} + +func _() { + f(S{}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue42695.go b/src/internal/types/testdata/fixedbugs/issue42695.go new file mode 100644 index 0000000..4551e9f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue42695.go @@ -0,0 +1,17 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue42695 + +const _ = 6e5518446744 // ERROR "malformed constant" +const _ uint8 = 6e5518446744 // ERROR "malformed constant" + +var _ = 6e5518446744 // ERROR "malformed constant" +var _ uint8 = 6e5518446744 // ERROR "malformed constant" + +func f(x int) int { + return x + 6e5518446744 // ERROR "malformed constant" +} + +var _ = f(6e5518446744 /* ERROR "malformed constant" */ ) diff --git a/src/internal/types/testdata/fixedbugs/issue42758.go b/src/internal/types/testdata/fixedbugs/issue42758.go new file mode 100644 index 0000000..4e1df34 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue42758.go @@ -0,0 +1,33 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[T any](x interface{}){ + switch x.(type) { + case T: // ok to use a type parameter + case int: + } + + switch x.(type) { + case T: + case T /* ERROR "duplicate case" */ : + } +} + +type constraint interface { + ~int +} + +func _[T constraint](x interface{}){ + switch x.(type) { + case T: // ok to use a type parameter even if type set contains int + case int: + } +} + +func _(x constraint /* ERROR "contains type constraints" */ ) { + switch x.(type) { // no need to report another error + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue42881.go b/src/internal/types/testdata/fixedbugs/issue42881.go new file mode 100644 index 0000000..b766b5e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue42881.go @@ -0,0 +1,16 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + T1 interface{ comparable } + T2 interface{ int } +) + +var ( + _ comparable // ERROR "cannot use type comparable outside a type constraint: interface is (or embeds) comparable" + _ T1 // ERROR "cannot use type T1 outside a type constraint: interface is (or embeds) comparable" + _ T2 // ERROR "cannot use type T2 outside a type constraint: interface contains type constraints" +) diff --git a/src/internal/types/testdata/fixedbugs/issue42987.go b/src/internal/types/testdata/fixedbugs/issue42987.go new file mode 100644 index 0000000..21c14c1 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue42987.go @@ -0,0 +1,8 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Check that there is only one error (no follow-on errors). + +package p +var _ = [ ... /* ERROR "invalid use of [...] array" */ ]byte("foo")
\ No newline at end of file diff --git a/src/internal/types/testdata/fixedbugs/issue43056.go b/src/internal/types/testdata/fixedbugs/issue43056.go new file mode 100644 index 0000000..8ff4e7f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43056.go @@ -0,0 +1,31 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// simplified example +func f[T ~func(T)](a, b T) {} + +type F func(F) + +func _() { + var i F + var j func(F) + + f(i, j) + f(j, i) +} + +// example from issue +func g[T interface{ Equal(T) bool }](a, b T) {} + +type I interface{ Equal(I) bool } + +func _() { + var i I + var j interface{ Equal(I) bool } + + g(i, j) + g(j, i) +} diff --git a/src/internal/types/testdata/fixedbugs/issue43087.go b/src/internal/types/testdata/fixedbugs/issue43087.go new file mode 100644 index 0000000..222fac8 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43087.go @@ -0,0 +1,43 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + a, b, b /* ERROR "b repeated on left side of :=" */ := 1, 2, 3 + _ = a + _ = b +} + +func _() { + a, _, _ := 1, 2, 3 // multiple _'s ok + _ = a +} + +func _() { + var b int + a, b, b /* ERROR "b repeated on left side of :=" */ := 1, 2, 3 + _ = a + _ = b +} + +func _() { + var a []int + a /* ERRORx `non-name .* on left side of :=` */ [0], b := 1, 2 + _ = a + _ = b +} + +func _() { + var a int + a, a /* ERROR "a repeated on left side of :=" */ := 1, 2 + _ = a +} + +func _() { + var a, b int + a, b := /* ERROR "no new variables on left side of :=" */ 1, 2 + _ = a + _ = b +} diff --git a/src/internal/types/testdata/fixedbugs/issue43109.go b/src/internal/types/testdata/fixedbugs/issue43109.go new file mode 100644 index 0000000..5d21a56 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43109.go @@ -0,0 +1,10 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Ensure there is no "imported and not used" error +// if a package wasn't imported in the first place. + +package p + +import . "/foo" // ERROR "could not import /foo" diff --git a/src/internal/types/testdata/fixedbugs/issue43110.go b/src/internal/types/testdata/fixedbugs/issue43110.go new file mode 100644 index 0000000..1e85022 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43110.go @@ -0,0 +1,43 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type P *struct{} + +func _() { + // want an error even if the switch is empty + var a struct{ _ func() } + switch a /* ERROR "cannot switch on a" */ { + } + + switch a /* ERROR "cannot switch on a" */ { + case a: // no follow-on error here + } + + // this is ok because f can be compared to nil + var f func() + switch f { + } + + switch f { + case nil: + } + + switch (func())(nil) { + case nil: + } + + switch (func())(nil) { + case f /* ERRORx `invalid case f in switch on .* \(func can only be compared to nil\)` */ : + } + + switch nil /* ERROR "use of untyped nil in switch expression" */ { + } + + // this is ok + switch P(nil) { + case P(nil): + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue43124.go b/src/internal/types/testdata/fixedbugs/issue43124.go new file mode 100644 index 0000000..ce26ae1 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43124.go @@ -0,0 +1,16 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var _ = int(0 /* ERROR "invalid use of ... in conversion to int" */ ...) + +// test case from issue + +type M []string + +var ( + x = []string{"a", "b"} + _ = M(x /* ERROR "invalid use of ... in conversion to M" */ ...) +) diff --git a/src/internal/types/testdata/fixedbugs/issue43125.go b/src/internal/types/testdata/fixedbugs/issue43125.go new file mode 100644 index 0000000..d0d6feb --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43125.go @@ -0,0 +1,8 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var _ = new(- /* ERROR "not a type" */ 1) +var _ = new(1 /* ERROR "not a type" */ + 1) diff --git a/src/internal/types/testdata/fixedbugs/issue43190.go b/src/internal/types/testdata/fixedbugs/issue43190.go new file mode 100644 index 0000000..b4d1fa4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43190.go @@ -0,0 +1,31 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// The errors below are produced by the parser, but we check +// them here for consistency with the types2 tests. + +package p + +import ; // ERROR "missing import path" +import "" // ERROR "invalid import path (empty string)" +import +var /* ERROR "missing import path" */ _ int +import .; // ERROR "missing import path" +import 'x' // ERROR "import path must be a string" +var _ int +import /* ERROR "imports must appear before other declarations" */ _ "math" + +// Don't repeat previous error for each immediately following import ... +import () +import (.) // ERROR "missing import path" +import ( + "fmt" + . +) // ERROR "missing import path" + +// ... but remind with error again if we start a new import section after +// other declarations +var _ = fmt.Println +import /* ERROR "imports must appear before other declarations" */ _ "math" +import _ "math" diff --git a/src/internal/types/testdata/fixedbugs/issue43527.go b/src/internal/types/testdata/fixedbugs/issue43527.go new file mode 100644 index 0000000..473ab96 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43527.go @@ -0,0 +1,16 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +const L = 10 + +type ( + _ [L]struct{} + _ [A /* ERROR "undefined array length A or missing type constraint" */ ]struct{} + _ [B /* ERROR "invalid array length B" */ ]struct{} + _[A any] struct{} + + B int +) diff --git a/src/internal/types/testdata/fixedbugs/issue43671.go b/src/internal/types/testdata/fixedbugs/issue43671.go new file mode 100644 index 0000000..be4c9ee --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue43671.go @@ -0,0 +1,58 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type C0 interface{ int } +type C1 interface{ chan int } +type C2 interface{ chan int | <-chan int } +type C3 interface{ chan int | chan float32 } +type C4 interface{ chan int | chan<- int } +type C5[T any] interface{ ~chan T | <-chan T } + +func _[T any](ch T) { + <-ch // ERRORx `cannot receive from ch .* \(no core type\)` +} + +func _[T C0](ch T) { + <-ch // ERROR "cannot receive from non-channel ch" +} + +func _[T C1](ch T) { + <-ch +} + +func _[T C2](ch T) { + <-ch +} + +func _[T C3](ch T) { + <-ch // ERRORx `cannot receive from ch .* \(no core type\)` +} + +func _[T C4](ch T) { + <-ch // ERROR "cannot receive from send-only channel" +} + +func _[T C5[X], X any](ch T, x X) { + x = <-ch +} + +// test case from issue, slightly modified +type RecvChan[T any] interface { + ~chan T | ~<-chan T +} + +func _[T any, C RecvChan[T]](ch C) T { + return <-ch +} + +func f[T any, C interface{ chan T }](ch C) T { + return <-ch +} + +func _(ch chan int) { + var x int = f(ch) // test constraint type inference for this case + _ = x +} diff --git a/src/internal/types/testdata/fixedbugs/issue44688.go b/src/internal/types/testdata/fixedbugs/issue44688.go new file mode 100644 index 0000000..512bfcc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue44688.go @@ -0,0 +1,83 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package P + +type A1[T any] struct{} + +func (*A1[T]) m1(T) {} + +type A2[T any] interface { + m2(T) +} + +type B1[T any] struct { + filler int + *A1[T] + A2[T] +} + +type B2[T any] interface { + A2[T] +} + +type C[T any] struct { + filler1 int + filler2 int + B1[T] +} + +type D[T any] struct { + filler1 int + filler2 int + filler3 int + C[T] +} + +func _() { + // calling embedded methods + var b1 B1[string] + + b1.A1.m1("") + b1.m1("") + + b1.A2.m2("") + b1.m2("") + + var b2 B2[string] + b2.m2("") + + // a deeper nesting + var d D[string] + d.m1("") + d.m2("") + + // calling method expressions + m1x := B1[string].m1 + m1x(b1, "") + m2x := B2[string].m2 + m2x(b2, "") + + // calling method values + m1v := b1.m1 + m1v("") + m2v := b1.m2 + m2v("") + b2v := b2.m2 + b2v("") +} + +// actual test case from issue + +type A[T any] struct{} + +func (*A[T]) f(T) {} + +type B[T any] struct{ A[T] } + +func _() { + var b B[string] + b.A.f("") + b.f("") +} diff --git a/src/internal/types/testdata/fixedbugs/issue44799.go b/src/internal/types/testdata/fixedbugs/issue44799.go new file mode 100644 index 0000000..9e528a7 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue44799.go @@ -0,0 +1,19 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +func Map[F, T any](s []F, f func(F) T) []T { return nil } + +func Reduce[Elem1, Elem2 any](s []Elem1, initializer Elem2, f func(Elem2, Elem1) Elem2) Elem2 { var x Elem2; return x } + +func main() { + var s []int + var f1 func(int) float64 + var f2 func(float64, int) float64 + _ = Map[int](s, f1) + _ = Map(s, f1) + _ = Reduce[int](s, 0, f2) + _ = Reduce(s, 0, f2) +} diff --git a/src/internal/types/testdata/fixedbugs/issue45114.go b/src/internal/types/testdata/fixedbugs/issue45114.go new file mode 100644 index 0000000..e51b3f7 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45114.go @@ -0,0 +1,8 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var s uint +var _ = string(1 /* ERRORx `shifted operand 1 .* must be integer` */ << s) diff --git a/src/internal/types/testdata/fixedbugs/issue45548.go b/src/internal/types/testdata/fixedbugs/issue45548.go new file mode 100644 index 0000000..01c9672 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45548.go @@ -0,0 +1,13 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[F interface{*Q}, G interface{*R}, Q, R any](q Q, r R) {} + +func _() { + f[*float64, *int](1, 2) + f[*float64](1, 2) + f(1, 2) +} diff --git a/src/internal/types/testdata/fixedbugs/issue45550.go b/src/internal/types/testdata/fixedbugs/issue45550.go new file mode 100644 index 0000000..2ea4ffe --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45550.go @@ -0,0 +1,10 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Builder /* ERROR "invalid recursive type" */ [T interface{ struct{ Builder[T] } }] struct{} +type myBuilder struct { + Builder[myBuilder] +} diff --git a/src/internal/types/testdata/fixedbugs/issue45635.go b/src/internal/types/testdata/fixedbugs/issue45635.go new file mode 100644 index 0000000..b83d477 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45635.go @@ -0,0 +1,31 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +func main() { + some /* ERROR "undefined" */ [int, int]() +} + +type N[T any] struct{} + +var _ N [] // ERROR "expected type argument list" + +type I interface { + ~[]int +} + +func _[T I](i, j int) { + var m map[int]int + _ = m[i, j /* ERROR "more than one index" */ ] + + var a [3]int + _ = a[i, j /* ERROR "more than one index" */ ] + + var s []int + _ = s[i, j /* ERROR "more than one index" */ ] + + var t T + _ = t[i, j /* ERROR "more than one index" */ ] +} diff --git a/src/internal/types/testdata/fixedbugs/issue45639.go b/src/internal/types/testdata/fixedbugs/issue45639.go new file mode 100644 index 0000000..a224aed --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45639.go @@ -0,0 +1,13 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package P + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// // It is not permitted to declare a local type whose underlying +// // type is a type parameters not declared by that type declaration. +// func _[T any]() { +// type _ T // ERROR "cannot use function type parameter T as RHS in type declaration" +// type _ [_ any] T // ERROR "cannot use function type parameter T as RHS in type declaration" +// } diff --git a/src/internal/types/testdata/fixedbugs/issue45920.go b/src/internal/types/testdata/fixedbugs/issue45920.go new file mode 100644 index 0000000..716abb1 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45920.go @@ -0,0 +1,17 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[T any, C chan T | <-chan T](ch C) {} + +func _(ch chan int) { f1(ch) } +func _(ch <-chan int) { f1(ch) } +func _(ch chan<- int) { f1 /* ERROR "chan<- int does not satisfy chan int | <-chan int" */ (ch) } + +func f2[T any, C chan T | chan<- T](ch C) {} + +func _(ch chan int) { f2(ch) } +func _(ch <-chan int) { f2 /* ERROR "<-chan int does not satisfy chan int | chan<- int" */ (ch) } +func _(ch chan<- int) { f2(ch) } diff --git a/src/internal/types/testdata/fixedbugs/issue45985.go b/src/internal/types/testdata/fixedbugs/issue45985.go new file mode 100644 index 0000000..c486150 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue45985.go @@ -0,0 +1,13 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue45985 + +func app[S interface{ ~[]T }, T any](s S, e T) S { + return append(s, e) +} + +func _() { + _ = app /* ERROR "S (type int) does not satisfy interface{~[]T}" */ [int] // TODO(gri) better error message +} diff --git a/src/internal/types/testdata/fixedbugs/issue46090.go b/src/internal/types/testdata/fixedbugs/issue46090.go new file mode 100644 index 0000000..59670da --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46090.go @@ -0,0 +1,11 @@ +// -lang=go1.17 + +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// The predeclared type comparable is not visible before Go 1.18. + +package p + +type _ comparable // ERROR "predeclared comparable" diff --git a/src/internal/types/testdata/fixedbugs/issue46275.go b/src/internal/types/testdata/fixedbugs/issue46275.go new file mode 100644 index 0000000..0862d5b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46275.go @@ -0,0 +1,26 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue46275 + +type N[T any] struct { + *N[T] + t T +} + +func (n *N[T]) Elem() T { + return n.t +} + +type I interface { + Elem() string +} + +func _() { + var n1 *N[string] + var _ I = n1 + type NS N[string] + var n2 *NS + var _ I = n2 +} diff --git a/src/internal/types/testdata/fixedbugs/issue46403.go b/src/internal/types/testdata/fixedbugs/issue46403.go new file mode 100644 index 0000000..fc60340 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46403.go @@ -0,0 +1,11 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue46403 + +func _() { + // a should be used, despite the parser error below. + var a []int + var _ = a[] // ERROR "expected operand" +} diff --git a/src/internal/types/testdata/fixedbugs/issue46404.go b/src/internal/types/testdata/fixedbugs/issue46404.go new file mode 100644 index 0000000..e3c93f6 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46404.go @@ -0,0 +1,10 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue46404 + +// TODO(gri) re-enable this test with matching errors +// between go/types and types2 +// Check that we don't type check t[_] as an instantiation. +// type t [t /* type parameters must be named */ /* not a generic type */ [_]]_ // cannot use diff --git a/src/internal/types/testdata/fixedbugs/issue46461.go b/src/internal/types/testdata/fixedbugs/issue46461.go new file mode 100644 index 0000000..ae70048 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46461.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// test case 1 +type T /* ERROR "invalid recursive type" */ [U interface{ M() T[U] }] int + +type X int + +func (X) M() T[X] { return 0 } + +// test case 2 +type A /* ERROR "invalid recursive type" */ [T interface{ A[T] }] interface{} + +// test case 3 +type A2 /* ERROR "invalid recursive type" */ [U interface{ A2[U] }] interface{ M() A2[U] } + +type I interface{ A2[I]; M() A2[I] } diff --git a/src/internal/types/testdata/fixedbugs/issue46583.go b/src/internal/types/testdata/fixedbugs/issue46583.go new file mode 100644 index 0000000..1901bff --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue46583.go @@ -0,0 +1,28 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T1 struct{} +func (t T1) m(int) {} +var f1 func(T1) + +type T2 struct{} +func (t T2) m(x int) {} +var f2 func(T2) + +type T3 struct{} +func (T3) m(int) {} +var f3 func(T3) + +type T4 struct{} +func (T4) m(x int) {} +var f4 func(T4) + +func _() { + f1 = T1 /* ERROR "func(T1, int)" */ .m + f2 = T2 /* ERROR "func(t T2, x int)" */ .m + f3 = T3 /* ERROR "func(T3, int)" */ .m + f4 = T4 /* ERROR "func(_ T4, x int)" */ .m +} diff --git a/src/internal/types/testdata/fixedbugs/issue47031.go b/src/internal/types/testdata/fixedbugs/issue47031.go new file mode 100644 index 0000000..23a9c55 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47031.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Mer interface { M() } + +func F[T Mer](p *T) { + p.M /* ERROR "p.M undefined" */ () +} + +type MyMer int + +func (MyMer) M() {} + +func _() { + F(new(MyMer)) + F[Mer](nil) +} diff --git a/src/internal/types/testdata/fixedbugs/issue47115.go b/src/internal/types/testdata/fixedbugs/issue47115.go new file mode 100644 index 0000000..2d2be34 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47115.go @@ -0,0 +1,40 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type C0 interface{ int } +type C1 interface{ chan int } +type C2 interface{ chan int | <-chan int } +type C3 interface{ chan int | chan float32 } +type C4 interface{ chan int | chan<- int } +type C5[T any] interface{ ~chan T | chan<- T } + +func _[T any](ch T) { + ch <- /* ERRORx `cannot send to ch .* no core type` */ 0 +} + +func _[T C0](ch T) { + ch <- /* ERROR "cannot send to non-channel" */ 0 +} + +func _[T C1](ch T) { + ch <- 0 +} + +func _[T C2](ch T) { + ch <-/* ERROR "cannot send to receive-only channel" */ 0 +} + +func _[T C3](ch T) { + ch <- /* ERRORx `cannot send to ch .* no core type` */ 0 +} + +func _[T C4](ch T) { + ch <- 0 +} + +func _[T C5[X], X any](ch T, x X) { + ch <- x +} diff --git a/src/internal/types/testdata/fixedbugs/issue47127.go b/src/internal/types/testdata/fixedbugs/issue47127.go new file mode 100644 index 0000000..b663938 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47127.go @@ -0,0 +1,37 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Embedding of stand-alone type parameters is not permitted. + +package p + +type ( + _[P any] interface{ *P | []P | chan P | map[string]P } + _[P any] interface{ P /* ERROR "term cannot be a type parameter" */ } + _[P any] interface{ ~P /* ERROR "type in term ~P cannot be a type parameter" */ } + _[P any] interface{ int | P /* ERROR "term cannot be a type parameter" */ } + _[P any] interface{ int | ~P /* ERROR "type in term ~P cannot be a type parameter" */ } +) + +func _[P any]() { + type ( + _[P any] interface{ *P | []P | chan P | map[string]P } + _[P any] interface{ P /* ERROR "term cannot be a type parameter" */ } + _[P any] interface{ ~P /* ERROR "type in term ~P cannot be a type parameter" */ } + _[P any] interface{ int | P /* ERROR "term cannot be a type parameter" */ } + _[P any] interface{ int | ~P /* ERROR "type in term ~P cannot be a type parameter" */ } + + _ interface{ *P | []P | chan P | map[string]P } + _ interface{ P /* ERROR "term cannot be a type parameter" */ } + _ interface{ ~P /* ERROR "type in term ~P cannot be a type parameter" */ } + _ interface{ int | P /* ERROR "term cannot be a type parameter" */ } + _ interface{ int | ~P /* ERROR "type in term ~P cannot be a type parameter" */ } + ) +} + +func _[P any, Q interface{ *P | []P | chan P | map[string]P }]() {} +func _[P any, Q interface{ P /* ERROR "term cannot be a type parameter" */ }]() {} +func _[P any, Q interface{ ~P /* ERROR "type in term ~P cannot be a type parameter" */ }]() {} +func _[P any, Q interface{ int | P /* ERROR "term cannot be a type parameter" */ }]() {} +func _[P any, Q interface{ int | ~P /* ERROR "type in term ~P cannot be a type parameter" */ }]() {} diff --git a/src/internal/types/testdata/fixedbugs/issue47411.go b/src/internal/types/testdata/fixedbugs/issue47411.go new file mode 100644 index 0000000..97b5942 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47411.go @@ -0,0 +1,26 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[_ comparable]() {} +func g[_ interface{interface{comparable; ~int|~string}}]() {} + +func _[P comparable, + Q interface{ comparable; ~int|~string }, + R any, // not comparable + S interface{ comparable; ~func() }, // not comparable +]() { + _ = f[int] + _ = f[P] + _ = f[Q] + _ = f[func /* ERROR "does not satisfy comparable" */ ()] + _ = f[R /* ERROR "R does not satisfy comparable" */ ] + + _ = g[int] + _ = g[P /* ERROR "P does not satisfy interface{interface{comparable; ~int | ~string}" */ ] + _ = g[Q] + _ = g[func /* ERROR "func() does not satisfy interface{interface{comparable; ~int | ~string}}" */ ()] + _ = g[R /* ERROR "R does not satisfy interface{interface{comparable; ~int | ~string}" */ ] +} diff --git a/src/internal/types/testdata/fixedbugs/issue47747.go b/src/internal/types/testdata/fixedbugs/issue47747.go new file mode 100644 index 0000000..34c78d3 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47747.go @@ -0,0 +1,71 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type T1[P any] P +// +// func (T1[_]) m() {} +// +// func _[P any](x *T1[P]) { +// // x.m exists because x is of type *T1 where T1 is a defined type +// // (even though under(T1) is a type parameter) +// x.m() +// } + + +func _[P interface{ m() }](x P) { + x.m() + // (&x).m doesn't exist because &x is of type *P + // and pointers to type parameters don't have methods + (&x).m /* ERROR "type *P is pointer to type parameter, not type parameter" */ () +} + + +type T2 interface{ m() } + +func _(x *T2) { + // x.m doesn't exists because x is of type *T2 + // and pointers to interfaces don't have methods + x.m /* ERROR "type *T2 is pointer to interface, not interface" */() +} + +// Test case 1 from issue + +type Fooer1[t any] interface { + Foo(Barer[t]) +} +type Barer[t any] interface { + Bar(t) +} + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type Foo1[t any] t +// type Bar[t any] t +// +// func (l Foo1[t]) Foo(v Barer[t]) { v.Bar(t(l)) } +// func (b *Bar[t]) Bar(l t) { *b = Bar[t](l) } +// +// func _[t any](f Fooer1[t]) t { +// var b Bar[t] +// f.Foo(&b) +// return t(b) +// } + +// Test case 2 from issue + +// For now, a lone type parameter is not permitted as RHS in a type declaration (issue #45639). +// type Fooer2[t any] interface { +// Foo() +// } +// +// type Foo2[t any] t +// +// func (f *Foo2[t]) Foo() {} +// +// func _[t any](v t) { +// var f = Foo2[t](v) +// _ = Fooer2[t](&f) +// } diff --git a/src/internal/types/testdata/fixedbugs/issue47796.go b/src/internal/types/testdata/fixedbugs/issue47796.go new file mode 100644 index 0000000..7f719ff --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47796.go @@ -0,0 +1,33 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// parameterized types with self-recursive constraints +type ( + T1 /* ERROR "invalid recursive type" */ [P T1[P]] interface{} + T2 /* ERROR "invalid recursive type" */ [P, Q T2[P, Q]] interface{} + T3[P T2[P, Q], Q interface{ ~string }] interface{} + + T4a /* ERROR "invalid recursive type" */ [P T4a[P]] interface{ ~int } + T4b /* ERROR "invalid recursive type" */ [P T4b[int]] interface{ ~int } + T4c /* ERROR "invalid recursive type" */ [P T4c[string]] interface{ ~int } + + // mutually recursive constraints + T5 /* ERROR "invalid recursive type" */ [P T6[P]] interface{ int } + T6[P T5[P]] interface{ int } +) + +// verify that constraints are checked as expected +var ( + _ T1[int] + _ T2[int, string] + _ T3[int, string] +) + +// test case from issue + +type Eq /* ERROR "invalid recursive type" */ [a Eq[a]] interface { + Equal(that a) bool +} diff --git a/src/internal/types/testdata/fixedbugs/issue47818.go b/src/internal/types/testdata/fixedbugs/issue47818.go new file mode 100644 index 0000000..21c8539 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47818.go @@ -0,0 +1,61 @@ +// -lang=go1.17 + +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Parser accepts type parameters but the type checker +// needs to report any operations that are not permitted +// before Go 1.18. + +package p + +type T[P /* ERROR "type parameter requires go1.18 or later" */ any /* ERROR "predeclared any requires go1.18 or later" */] struct{} + +// for init (and main, but we're not in package main) we should only get one error +func init[P /* ERROR "func init must have no type parameters" */ any /* ERROR "predeclared any requires go1.18 or later" */]() { +} +func main[P /* ERROR "type parameter requires go1.18 or later" */ any /* ERROR "predeclared any requires go1.18 or later" */]() { +} + +func f[P /* ERROR "type parameter requires go1.18 or later" */ any /* ERROR "predeclared any requires go1.18 or later" */](x P) { + var _ T[ /* ERROR "type instantiation requires go1.18 or later" */ int] + var _ (T[ /* ERROR "type instantiation requires go1.18 or later" */ int]) + _ = T[ /* ERROR "type instantiation requires go1.18 or later" */ int]{} + _ = T[ /* ERROR "type instantiation requires go1.18 or later" */ int](struct{}{}) +} + +func (T[ /* ERROR "type instantiation requires go1.18 or later" */ P]) g(x int) { + f[ /* ERROR "function instantiation requires go1.18 or later" */ int](0) // explicit instantiation + (f[ /* ERROR "function instantiation requires go1.18 or later" */ int])(0) // parentheses (different code path) + f( /* ERROR "implicit function instantiation requires go1.18 or later" */ x) // implicit instantiation +} + +type C1 interface { + comparable // ERROR "predeclared comparable requires go1.18 or later" +} + +type C2 interface { + comparable // ERROR "predeclared comparable requires go1.18 or later" + int // ERROR "embedding non-interface type int requires go1.18 or later" + ~ /* ERROR "embedding interface element ~int requires go1.18 or later" */ int + int /* ERROR "embedding interface element int | ~string requires go1.18 or later" */ | ~string +} + +type _ interface { + // errors for these were reported with their declaration + C1 + C2 +} + +type ( + _ comparable // ERROR "predeclared comparable requires go1.18 or later" + // errors for these were reported with their declaration + _ C1 + _ C2 + + _ = comparable // ERROR "predeclared comparable requires go1.18 or later" + // errors for these were reported with their declaration + _ = C1 + _ = C2 +) diff --git a/src/internal/types/testdata/fixedbugs/issue47887.go b/src/internal/types/testdata/fixedbugs/issue47887.go new file mode 100644 index 0000000..4c4fc2f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47887.go @@ -0,0 +1,28 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Fooer[t any] interface { + foo(Barer[t]) +} +type Barer[t any] interface { + bar(Bazer[t]) +} +type Bazer[t any] interface { + Fooer[t] + baz(t) +} + +type Int int + +func (n Int) baz(int) {} +func (n Int) foo(b Barer[int]) { b.bar(n) } + +type F[t any] interface { f(G[t]) } +type G[t any] interface { g(H[t]) } +type H[t any] interface { F[t] } + +type T struct{} +func (n T) f(b G[T]) { b.g(n) } diff --git a/src/internal/types/testdata/fixedbugs/issue47968.go b/src/internal/types/testdata/fixedbugs/issue47968.go new file mode 100644 index 0000000..c516eee --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue47968.go @@ -0,0 +1,21 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] struct{} + +func (T[P]) m1() + +type A1 = T // ERROR "cannot use generic type" + +func (A1[P]) m2() {} + +type A2 = T[int] + +func (A2 /* ERROR "cannot define new methods on instantiated type T[int]" */) m3() {} +func (_ /* ERROR "cannot define new methods on instantiated type T[int]" */ A2) m4() {} + +func (T[int]) m5() {} // int is the type parameter name, not an instantiation +func (T[* /* ERROR "must be an identifier" */ int]) m6() {} // syntax error diff --git a/src/internal/types/testdata/fixedbugs/issue48008.go b/src/internal/types/testdata/fixedbugs/issue48008.go new file mode 100644 index 0000000..8d0c640 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48008.go @@ -0,0 +1,60 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] struct{} + +func _(x interface{}) { + switch x.(type) { + case nil: + case int: + + case T[int]: + case []T[int]: + case [10]T[int]: + case struct{T[int]}: + case *T[int]: + case func(T[int]): + case interface{m(T[int])}: + case map[T[int]] string: + case chan T[int]: + + case T /* ERROR "cannot use generic type T[P any] without instantiation" */ : + case []T /* ERROR "cannot use generic type" */ : + case [10]T /* ERROR "cannot use generic type" */ : + case struct{T /* ERROR "cannot use generic type" */ }: + case *T /* ERROR "cannot use generic type" */ : + case func(T /* ERROR "cannot use generic type" */ ): + case interface{m(T /* ERROR "cannot use generic type" */ )}: + case map[T /* ERROR "cannot use generic type" */ ] string: + case chan T /* ERROR "cannot use generic type" */ : + + case T /* ERROR "cannot use generic type" */ , *T /* ERROR "cannot use generic type" */ : + } +} + +// Make sure a parenthesized nil is ok. + +func _(x interface{}) { + switch x.(type) { + case ((nil)), int: + } +} + +// Make sure we look for the predeclared nil. + +func _(x interface{}) { + type nil int + switch x.(type) { + case nil: // ok - this is the type nil + } +} + +func _(x interface{}) { + var nil int + switch x.(type) { + case nil /* ERROR "not a type" */ : // not ok - this is the variable nil + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue48018.go b/src/internal/types/testdata/fixedbugs/issue48018.go new file mode 100644 index 0000000..3df908a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48018.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package main + +type Box[A any] struct { + value A +} + +func Nest[A /* ERROR "instantiation cycle" */ any](b Box[A], n int) interface{} { + if n == 0 { + return b + } + return Nest(Box[Box[A]]{b}, n-1) +} + +func main() { + Nest(Box[int]{0}, 10) +} diff --git a/src/internal/types/testdata/fixedbugs/issue48048.go b/src/internal/types/testdata/fixedbugs/issue48048.go new file mode 100644 index 0000000..98a03ea --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48048.go @@ -0,0 +1,15 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] struct{} + +func (T[_]) A() {} + +var _ = (T[int]).A +var _ = (*T[int]).A + +var _ = (T /* ERROR "cannot use generic type" */).A +var _ = (*T /* ERROR "cannot use generic type" */).A diff --git a/src/internal/types/testdata/fixedbugs/issue48082.go b/src/internal/types/testdata/fixedbugs/issue48082.go new file mode 100644 index 0000000..648c512 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48082.go @@ -0,0 +1,7 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package issue48082 + +import "init" /* ERROR "init must be a func" */ /* ERROR "could not import init" */ diff --git a/src/internal/types/testdata/fixedbugs/issue48083.go b/src/internal/types/testdata/fixedbugs/issue48083.go new file mode 100644 index 0000000..15e9b70 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48083.go @@ -0,0 +1,9 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] struct{} + +type _ interface{ int | T /* ERROR "cannot use generic type" */ }
\ No newline at end of file diff --git a/src/internal/types/testdata/fixedbugs/issue48136.go b/src/internal/types/testdata/fixedbugs/issue48136.go new file mode 100644 index 0000000..b76322e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48136.go @@ -0,0 +1,36 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[P interface{ *P }]() {} +func f2[P interface{ func(P) }]() {} +func f3[P, Q interface{ func(Q) P }]() {} +func f4[P interface{ *Q }, Q interface{ func(P) }]() {} +func f5[P interface{ func(P) }]() {} +func f6[P interface { *Tree[P] }, Q any ]() {} + +func _() { + f1 /* ERROR "cannot infer P" */ () + f2 /* ERROR "cannot infer P" */ () + f3 /* ERROR "cannot infer P" */ () + f4 /* ERROR "cannot infer P" */ () + f5 /* ERROR "cannot infer P" */ () + f6 /* ERROR "cannot infer P" */ () +} + +type Tree[P any] struct { + left, right *Tree[P] + data P +} + +// test case from issue + +func foo[Src interface { func() Src }]() Src { + return foo[Src] +} + +func _() { + foo /* ERROR "cannot infer Src" */ () +} diff --git a/src/internal/types/testdata/fixedbugs/issue48234.go b/src/internal/types/testdata/fixedbugs/issue48234.go new file mode 100644 index 0000000..e069930 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48234.go @@ -0,0 +1,10 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var _ = interface{ + m() + m /* ERROR "duplicate method" */ () +}(nil) diff --git a/src/internal/types/testdata/fixedbugs/issue48312.go b/src/internal/types/testdata/fixedbugs/issue48312.go new file mode 100644 index 0000000..708201b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48312.go @@ -0,0 +1,20 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T interface{ m() } +type P *T + +func _(p *T) { + p.m /* ERROR "type *T is pointer to interface, not interface" */ () +} + +func _(p P) { + p.m /* ERROR "type P is pointer to interface, not interface" */ () +} + +func _[P T](p *P) { + p.m /* ERROR "type *P is pointer to type parameter, not type parameter" */ () +} diff --git a/src/internal/types/testdata/fixedbugs/issue48472.go b/src/internal/types/testdata/fixedbugs/issue48472.go new file mode 100644 index 0000000..169ab0d --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48472.go @@ -0,0 +1,16 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func g() { + var s string + var i int + _ = s /* ERROR "invalid operation: s + i (mismatched types string and int)" */ + i +} + +func f(i int) int { + i /* ERROR `invalid operation: i += "1" (mismatched types int and untyped string)` */ += "1" + return i +} diff --git a/src/internal/types/testdata/fixedbugs/issue48529.go b/src/internal/types/testdata/fixedbugs/issue48529.go new file mode 100644 index 0000000..bcc5e35 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48529.go @@ -0,0 +1,11 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T /* ERROR "invalid recursive type" */ [U interface{ M() T[U, int] }] int + +type X int + +func (X) M() T[X] { return 0 } diff --git a/src/internal/types/testdata/fixedbugs/issue48582.go b/src/internal/types/testdata/fixedbugs/issue48582.go new file mode 100644 index 0000000..8ffcd5a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48582.go @@ -0,0 +1,29 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type N /* ERROR "invalid recursive type" */ interface { + int | N +} + +type A /* ERROR "invalid recursive type" */ interface { + int | B +} + +type B interface { + int | A +} + +type S /* ERROR "invalid recursive type" */ struct { + I // ERROR "interface contains type constraints" +} + +type I interface { + int | S +} + +type P interface { + *P // ERROR "interface contains type constraints" +} diff --git a/src/internal/types/testdata/fixedbugs/issue48619.go b/src/internal/types/testdata/fixedbugs/issue48619.go new file mode 100644 index 0000000..fc5dce0 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48619.go @@ -0,0 +1,22 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P any](a, _ P) { + var x int + // TODO(gri) these error messages, while correct, could be better + f(a, x /* ERROR "type int of x does not match inferred type P for P" */) + f(x, a /* ERROR "type P of a does not match inferred type int for P" */) +} + +func g[P any](a, b P) { + g(a, b) + g(&a, &b) + g([]P{}, []P{}) + + // work-around: provide type argument explicitly + g[*P](&a, &b) + g[[]P]([]P{}, []P{}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue48656.go b/src/internal/types/testdata/fixedbugs/issue48656.go new file mode 100644 index 0000000..f77e08a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48656.go @@ -0,0 +1,13 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P *Q, Q any](P, Q) { + _ = f[P] +} + +func f2[P /* ERROR "instantiation cycle" */ *Q, Q any](P, Q) { + _ = f2[*P] +} diff --git a/src/internal/types/testdata/fixedbugs/issue48695.go b/src/internal/types/testdata/fixedbugs/issue48695.go new file mode 100644 index 0000000..9f4a768 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48695.go @@ -0,0 +1,14 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func g[P ~func(T) P, T any](P) {} + +func _() { + type F func(int) F + var f F + g(f) + _ = g[F] +} diff --git a/src/internal/types/testdata/fixedbugs/issue48703.go b/src/internal/types/testdata/fixedbugs/issue48703.go new file mode 100644 index 0000000..89c667b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48703.go @@ -0,0 +1,27 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +// The actual example from the issue. +type List[P any] struct{} + +func (_ List[P]) m() (_ List[List[P]]) { return } + +// Other types of recursion through methods. +type R[P any] int + +func (*R[R /* ERROR "must be an identifier" */ [int]]) m0() {} +func (R[P]) m1(R[R[P]]) {} +func (R[P]) m2(R[*P]) {} +func (R[P]) m3([unsafe.Sizeof(new(R[P]))]int) {} +func (R[P]) m4([unsafe.Sizeof(new(R[R[P]]))]int) {} + +// Mutual recursion +type M[P any] int + +func (R[P]) m5(M[M[P]]) {} +func (M[P]) m(R[R[P]]) {} diff --git a/src/internal/types/testdata/fixedbugs/issue48712.go b/src/internal/types/testdata/fixedbugs/issue48712.go new file mode 100644 index 0000000..76ad16c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48712.go @@ -0,0 +1,41 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P comparable](x, y P) { + _ = x == x + _ = x == y + _ = y == x + _ = y == y + + _ = x /* ERROR "type parameter P is not comparable with <" */ < y +} + +func _[P comparable](x P, y any) { + _ = x == x + _ = x == y + _ = y == x + _ = y == y + + _ = x /* ERROR "type parameter P is not comparable with <" */ < y +} + +func _[P any](x, y P) { + _ = x /* ERROR "incomparable types in type set" */ == x + _ = x /* ERROR "incomparable types in type set" */ == y + _ = y /* ERROR "incomparable types in type set" */ == x + _ = y /* ERROR "incomparable types in type set" */ == y + + _ = x /* ERROR "type parameter P is not comparable with <" */ < y +} + +func _[P any](x P, y any) { + _ = x /* ERROR "incomparable types in type set" */ == x + _ = x /* ERROR "incomparable types in type set" */ == y + _ = y == x // ERROR "incomparable types in type set" + _ = y == y + + _ = x /* ERROR "type parameter P is not comparable with <" */ < y +} diff --git a/src/internal/types/testdata/fixedbugs/issue48819.go b/src/internal/types/testdata/fixedbugs/issue48819.go new file mode 100644 index 0000000..916faaf --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48819.go @@ -0,0 +1,15 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type T /* ERROR "invalid recursive type: T refers to itself" */ struct { + T +} + +func _(t T) { + _ = unsafe.Sizeof(t) // should not go into infinite recursion here +} diff --git a/src/internal/types/testdata/fixedbugs/issue48827.go b/src/internal/types/testdata/fixedbugs/issue48827.go new file mode 100644 index 0000000..bd08835 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48827.go @@ -0,0 +1,19 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type G[P any] int + +type ( + _ G[int] + _ G[G /* ERRORx `cannot use.*without instantiation` */] + _ bool /* ERROR "invalid operation: bool[int] (bool is not a generic type)" */ [int] + _ bool /* ERROR "invalid operation: bool[G] (bool is not a generic type)" */[G] +) + +// The example from the issue. +func _() { + _ = &([10]bool /* ERRORx `invalid operation.*bool is not a generic type` */ [1 /* ERROR "expected type" */ ]{}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue48951.go b/src/internal/types/testdata/fixedbugs/issue48951.go new file mode 100644 index 0000000..8d6f850 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48951.go @@ -0,0 +1,21 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + A1[P any] [10]A1 /* ERROR "invalid recursive type" */ [P] + A2[P any] [10]A2 /* ERROR "invalid recursive type" */ [*P] + A3[P any] [10]*A3[P] + + L1[P any] []L1[P] + + S1[P any] struct{ f S1 /* ERROR "invalid recursive type" */ [P] } + S2[P any] struct{ f S2 /* ERROR "invalid recursive type" */ [*P] } // like example in issue + S3[P any] struct{ f *S3[P] } + + I1[P any] interface{ I1 /* ERROR "invalid recursive type" */ [P] } + I2[P any] interface{ I2 /* ERROR "invalid recursive type" */ [*P] } + I3[P any] interface{ *I3 /* ERROR "interface contains type constraints" */ [P] } +) diff --git a/src/internal/types/testdata/fixedbugs/issue48962.go b/src/internal/types/testdata/fixedbugs/issue48962.go new file mode 100644 index 0000000..4294cf0 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48962.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T0[P any] struct { + f P +} + +type T1 /* ERROR "invalid recursive type" */ struct { + _ T0[T1] +} diff --git a/src/internal/types/testdata/fixedbugs/issue48974.go b/src/internal/types/testdata/fixedbugs/issue48974.go new file mode 100644 index 0000000..08d8656 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue48974.go @@ -0,0 +1,22 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Fooer interface { + Foo() +} + +type Fooable[F /* ERROR "instantiation cycle" */ Fooer] struct { + ptr F +} + +func (f *Fooable[F]) Adapter() *Fooable[*FooerImpl[F]] { + return &Fooable[*FooerImpl[F]]{&FooerImpl[F]{}} +} + +type FooerImpl[F Fooer] struct { +} + +func (fi *FooerImpl[F]) Foo() {} diff --git a/src/internal/types/testdata/fixedbugs/issue49003.go b/src/internal/types/testdata/fixedbugs/issue49003.go new file mode 100644 index 0000000..bf2e6c4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49003.go @@ -0,0 +1,10 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f(s string) int { + for range s { + } +} // ERROR "missing return" diff --git a/src/internal/types/testdata/fixedbugs/issue49005.go b/src/internal/types/testdata/fixedbugs/issue49005.go new file mode 100644 index 0000000..d91c207 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49005.go @@ -0,0 +1,31 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T1 interface{ M() } + +func F1() T1 + +var _ = F1().(*X1 /* ERROR "undefined: X1" */) + +func _() { + switch F1().(type) { + case *X1 /* ERROR "undefined: X1" */ : + } +} + +type T2 interface{ M() } + +func F2() T2 + +var _ = F2 /* ERROR "impossible type assertion: F2().(*X2)\n\t*X2 does not implement T2 (missing method M)" */ ().(*X2) + +type X2 struct{} + +func _() { + switch F2().(type) { + case * /* ERROR "impossible type switch case: *X2\n\tF2() (value of type T2) cannot have dynamic type *X2 (missing method M)" */ X2: + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue49043.go b/src/internal/types/testdata/fixedbugs/issue49043.go new file mode 100644 index 0000000..7594b32 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49043.go @@ -0,0 +1,24 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// The example from the issue. +type ( + N[P any] M /* ERROR "invalid recursive type" */ [P] + M[P any] N[P] +) + +// A slightly more complicated case. +type ( + A[P any] B /* ERROR "invalid recursive type" */ [P] + B[P any] C[P] + C[P any] A[P] +) + +// Confusing but valid (note that `type T *T` is valid). +type ( + N1[P any] *M1[P] + M1[P any] *N1[P] +) diff --git a/src/internal/types/testdata/fixedbugs/issue49112.go b/src/internal/types/testdata/fixedbugs/issue49112.go new file mode 100644 index 0000000..e87d1c0 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49112.go @@ -0,0 +1,15 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P int](P) {} + +func _() { + _ = f[int] + _ = f[[ /* ERROR "[]int does not satisfy int ([]int missing in int)" */ ]int] + + f(0) + f /* ERROR "P (type []int) does not satisfy int" */ ([]int{}) // TODO(gri) better error message +} diff --git a/src/internal/types/testdata/fixedbugs/issue49179.go b/src/internal/types/testdata/fixedbugs/issue49179.go new file mode 100644 index 0000000..1f8da29 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49179.go @@ -0,0 +1,37 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[P int | string]() {} +func f2[P ~int | string | float64]() {} +func f3[P int](x P) {} + +type myInt int +type myFloat float64 + +func _() { + _ = f1[int] + _ = f1[myInt /* ERROR "possibly missing ~ for int in int | string" */] + _ = f2[myInt] + _ = f2[myFloat /* ERROR "possibly missing ~ for float64 in ~int | string | float64" */] + var x myInt + f3 /* ERROR "myInt does not satisfy int (possibly missing ~ for int in int)" */ (x) +} + +// test case from the issue + +type SliceConstraint[T any] interface { + []T +} + +func Map[S SliceConstraint[E], E any](s S, f func(E) E) S { + return s +} + +type MySlice []int + +func f(s MySlice) { + Map[MySlice /* ERROR "MySlice does not satisfy SliceConstraint[int] (possibly missing ~ for []int in SliceConstraint[int])" */, int](s, nil) +} diff --git a/src/internal/types/testdata/fixedbugs/issue49242.go b/src/internal/types/testdata/fixedbugs/issue49242.go new file mode 100644 index 0000000..0415bf6 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49242.go @@ -0,0 +1,27 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P int](x P) int { + return x // ERRORx `cannot use x .* as int value in return statement` +} + +func _[P int]() int { + return P /* ERRORx `cannot use P\(1\) .* as int value in return statement` */ (1) +} + +func _[P int](x int) P { + return x // ERRORx `cannot use x .* as P value in return statement` +} + +func _[P, Q any](x P) Q { + return x // ERRORx `cannot use x .* as Q value in return statement` +} + +// test case from issue +func F[G interface{ uint }]() int { + f := func(uint) int { return 0 } + return f(G /* ERRORx `cannot use G\(1\) .* as uint value in argument to f` */ (1)) +} diff --git a/src/internal/types/testdata/fixedbugs/issue49247.go b/src/internal/types/testdata/fixedbugs/issue49247.go new file mode 100644 index 0000000..0ad2e29 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49247.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type integer interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 | + ~uint | ~uint8 | ~uint16 | ~uint32 | ~uint64 | ~uintptr +} + +func Add1024[T integer](s []T) { + for i, v := range s { + s[i] = v + 1024 // ERROR "cannot convert 1024 (untyped int constant) to type T" + } +} + +func f[T interface{ int8 }]() { + println(T(1024 /* ERROR "cannot convert 1024 (untyped int value) to type T" */)) +} diff --git a/src/internal/types/testdata/fixedbugs/issue49276.go b/src/internal/types/testdata/fixedbugs/issue49276.go new file mode 100644 index 0000000..bdfb42f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49276.go @@ -0,0 +1,46 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type S /* ERROR "invalid recursive type S" */ struct { + _ [unsafe.Sizeof(s)]byte +} + +var s S + +// Since f is a pointer, this case could be valid. +// But it's pathological and not worth the expense. +type T struct { + f *[unsafe.Sizeof(T /* ERROR "invalid recursive type" */ {})]int +} + +// a mutually recursive case using unsafe.Sizeof +type ( + A1 struct { + _ [unsafe.Sizeof(B1{})]int + } + + B1 struct { + _ [unsafe.Sizeof(A1 /* ERROR "invalid recursive type" */ {})]int + } +) + +// a mutually recursive case using len +type ( + A2 struct { + f [len(B2{}.f)]int + } + + B2 struct { + f [len(A2 /* ERROR "invalid recursive type" */ {}.f)]int + } +) + +// test case from issue +type a struct { + _ [42 - unsafe.Sizeof(a /* ERROR "invalid recursive type" */ {})]byte +} diff --git a/src/internal/types/testdata/fixedbugs/issue49296.go b/src/internal/types/testdata/fixedbugs/issue49296.go new file mode 100644 index 0000000..c8c5208 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49296.go @@ -0,0 +1,20 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[ + T0 any, + T1 []int, + T2 ~float64 | ~complex128 | chan int, +]() { + _ = T0(nil /* ERROR "cannot convert nil to type T0" */ ) + _ = T1(1 /* ERRORx `cannot convert 1 .* to type T1` */ ) + _ = T2(2 /* ERRORx `cannot convert 2 .* to type T2` */ ) +} + +// test case from issue +func f[T interface{[]int}]() { + _ = T(1 /* ERROR "cannot convert" */ ) +} diff --git a/src/internal/types/testdata/fixedbugs/issue49439.go b/src/internal/types/testdata/fixedbugs/issue49439.go new file mode 100644 index 0000000..3852f16 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49439.go @@ -0,0 +1,26 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type T0 /* ERROR "invalid recursive type" */ [P T0[P]] struct{} + +type T1 /* ERROR "invalid recursive type" */ [P T2[P]] struct{} +type T2[P T1[P]] struct{} + +type T3 /* ERROR "invalid recursive type" */ [P interface{ ~struct{ f T3[int] } }] struct{} + +// valid cycle in M +type N[P M[P]] struct{} +type M[Q any] struct { F *M[Q] } + +// "crazy" case +type TC[P [unsafe.Sizeof(func() { + type T [P [unsafe.Sizeof(func(){})]byte] struct{} +})]byte] struct{} + +// test case from issue +type X /* ERROR "invalid recursive type" */ [T any, PT X[T]] interface{} diff --git a/src/internal/types/testdata/fixedbugs/issue49482.go b/src/internal/types/testdata/fixedbugs/issue49482.go new file mode 100644 index 0000000..7139bae --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49482.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// The following is OK, per the special handling for type literals discussed in issue #49482. +type _[P *struct{}] struct{} +type _[P *int,] int +type _[P (*int),] int + +const P = 2 // declare P to avoid noisy 'undefined' errors below. + +// The following parse as invalid array types due to parsing ambiguitiues. +type _ [P *int /* ERROR "int (type) is not an expression" */ ]int +type _ [P /* ERROR "non-function P" */ (*int)]int + +// Adding a trailing comma or an enclosing interface resolves the ambiguity. +type _[P *int,] int +type _[P (*int),] int +type _[P interface{*int}] int +type _[P interface{(*int)}] int + +// The following parse correctly as valid generic types. +type _[P *struct{} | int] struct{} +type _[P *struct{} | ~int] struct{} diff --git a/src/internal/types/testdata/fixedbugs/issue49541.go b/src/internal/types/testdata/fixedbugs/issue49541.go new file mode 100644 index 0000000..da37311 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49541.go @@ -0,0 +1,45 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S[A, B any] struct { + f int +} + +func (S[A, B]) m() {} + +// TODO(gri): with type-type inference enabled we should only report one error +// below. See issue #50588. + +func _[A any](s S /* ERROR "got 1 arguments but 2 type parameters" */ [A]) { + // we should see no follow-on errors below + s.f = 1 + s.m() +} + +// another test case from the issue + +func _() { + X /* ERROR "cannot infer Q" */ (Interface[*F /* ERROR "got 1 arguments but 2 type parameters" */ [string]](Impl{})) +} + +func X[Q Qer](fs Interface[Q]) { +} + +type Impl struct{} + +func (Impl) M() {} + +type Interface[Q Qer] interface { + M() +} + +type Qer interface { + Q() +} + +type F[A, B any] struct{} + +func (f *F[A, B]) Q() {} diff --git a/src/internal/types/testdata/fixedbugs/issue49579.go b/src/internal/types/testdata/fixedbugs/issue49579.go new file mode 100644 index 0000000..780859c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49579.go @@ -0,0 +1,17 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type I[F any] interface { + Q(*F) +} + +func G[F any]() I[any] { + return g /* ERRORx `cannot use g\[F\]{} .* as I\[any\] value in return statement: g\[F\] does not implement I\[any\] \(method Q has pointer receiver\)` */ [F]{} +} + +type g[F any] struct{} + +func (*g[F]) Q(*any) {} diff --git a/src/internal/types/testdata/fixedbugs/issue49592.go b/src/internal/types/testdata/fixedbugs/issue49592.go new file mode 100644 index 0000000..846deaa --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49592.go @@ -0,0 +1,11 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + var x *interface{} + var y interface{} + _ = x == y +} diff --git a/src/internal/types/testdata/fixedbugs/issue49602.go b/src/internal/types/testdata/fixedbugs/issue49602.go new file mode 100644 index 0000000..09cc969 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49602.go @@ -0,0 +1,19 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type M interface { + m() +} + +type C interface { + comparable +} + +type _ interface { + int | M // ERROR "cannot use p.M in union (p.M contains methods)" + int | comparable // ERROR "cannot use comparable in union" + int | C // ERROR "cannot use p.C in union (p.C embeds comparable)" +} diff --git a/src/internal/types/testdata/fixedbugs/issue49705.go b/src/internal/types/testdata/fixedbugs/issue49705.go new file mode 100644 index 0000000..5b5fba2 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49705.go @@ -0,0 +1,14 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Integer interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 | + ~uint | ~uint8 | ~uint16 | ~uint32 | ~uint64 | ~uintptr +} + +func shl[I Integer](n int) I { + return 1 << n +} diff --git a/src/internal/types/testdata/fixedbugs/issue49735.go b/src/internal/types/testdata/fixedbugs/issue49735.go new file mode 100644 index 0000000..0fcc778 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49735.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P1 any, P2 ~byte](s1 P1, s2 P2) { + _ = append(nil /* ERROR "first argument to append must be a slice; have untyped nil" */ , 0) + _ = append(s1 /* ERRORx `s1 .* has no core type` */ , 0) + _ = append(s2 /* ERRORx `s2 .* has core type byte` */ , 0) +} diff --git a/src/internal/types/testdata/fixedbugs/issue49739.go b/src/internal/types/testdata/fixedbugs/issue49739.go new file mode 100644 index 0000000..73825f4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49739.go @@ -0,0 +1,23 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Verify that we get an empty type set (not just an error) +// when using an invalid ~A. + +package p + +type A int +type C interface { + ~ /* ERROR "invalid use of ~" */ A +} + +func f[_ C]() {} +func g[_ interface{ C }]() {} +func h[_ C | int]() {} + +func _() { + _ = f[int /* ERROR "cannot satisfy C (empty type set)" */] + _ = g[int /* ERROR "cannot satisfy interface{C} (empty type set)" */] + _ = h[int] +} diff --git a/src/internal/types/testdata/fixedbugs/issue49864.go b/src/internal/types/testdata/fixedbugs/issue49864.go new file mode 100644 index 0000000..8ccd77c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue49864.go @@ -0,0 +1,9 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P ~int, Q any](p P) { + _ = Q(p /* ERROR "cannot convert" */ ) +} diff --git a/src/internal/types/testdata/fixedbugs/issue50259.go b/src/internal/types/testdata/fixedbugs/issue50259.go new file mode 100644 index 0000000..6df8c64 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50259.go @@ -0,0 +1,18 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var x T[B] + +type T[_ any] struct{} +type A T[B] +type B = T[A] + +// test case from issue + +var v Box[Step] +type Box[T any] struct{} +type Step = Box[StepBox] +type StepBox Box[Step] diff --git a/src/internal/types/testdata/fixedbugs/issue50276.go b/src/internal/types/testdata/fixedbugs/issue50276.go new file mode 100644 index 0000000..97e477e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50276.go @@ -0,0 +1,39 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// simplified test case + +type transform[T any] struct{} +type pair[S any] struct {} + +var _ transform[step] + +type box transform[step] +type step = pair[box] + +// test case from issue + +type Transform[T any] struct{ hold T } +type Pair[S, T any] struct { + First S + Second T +} + +var first Transform[Step] + +// This line doesn't use the Step alias, and it compiles fine if you uncomment it. +var second Transform[Pair[Box, interface{}]] + +type Box *Transform[Step] + +// This line is the same as the `first` line, but it comes after the Box declaration and +// does not break the compile. +var third Transform[Step] + +type Step = Pair[Box, interface{}] + +// This line also does not break the compile +var fourth Transform[Step] diff --git a/src/internal/types/testdata/fixedbugs/issue50281.go b/src/internal/types/testdata/fixedbugs/issue50281.go new file mode 100644 index 0000000..f333e81 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50281.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[S string | []byte](s S) { + var buf []byte + _ = append(buf, s...) +} + +func _[S ~string | ~[]byte](s S) { + var buf []byte + _ = append(buf, s...) +} + +// test case from issue + +type byteseq interface { + string | []byte +} + +// This should allow to eliminate the two functions above. +func AppendByteString[source byteseq](buf []byte, s source) []byte { + return append(buf, s[1:6]...) +} diff --git a/src/internal/types/testdata/fixedbugs/issue50321.go b/src/internal/types/testdata/fixedbugs/issue50321.go new file mode 100644 index 0000000..ab2a31b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50321.go @@ -0,0 +1,8 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func Ln[A A /* ERROR "cannot use a type parameter as constraint" */ ](p A) { +} diff --git a/src/internal/types/testdata/fixedbugs/issue50372.go b/src/internal/types/testdata/fixedbugs/issue50372.go new file mode 100644 index 0000000..10d2a24 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50372.go @@ -0,0 +1,27 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(s []int) { + var i, j, k, l int + _, _, _, _ = i, j, k, l + + for range s {} + for i = range s {} + for i, j = range s {} + for i, j, k /* ERRORx "range clause permits at most two iteration variables|at most 2 expressions" */ = range s {} + for i, j, k, l /* ERRORx "range clause permits at most two iteration variables|at most 2 expressions" */ = range s {} +} + +func _(s chan int) { + var i, j, k, l int + _, _, _, _ = i, j, k, l + + for range s {} + for i = range s {} + for i, j /* ERRORx `range over .* permits only one iteration variable` */ = range s {} + for i, j, k /* ERRORx `range over .* permits only one iteration variable|at most 2 expressions` */ = range s {} + for i, j, k, l /* ERRORx `range over .* permits only one iteration variable|at most 2 expressions` */ = range s {} +} diff --git a/src/internal/types/testdata/fixedbugs/issue50417.go b/src/internal/types/testdata/fixedbugs/issue50417.go new file mode 100644 index 0000000..c70898e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50417.go @@ -0,0 +1,68 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Field accesses through type parameters are disabled +// until we have a more thorough understanding of the +// implications on the spec. See issue #51576. + +package p + +type Sf struct { + f int +} + +func f0[P Sf](p P) { + _ = p.f // ERROR "p.f undefined" + p.f /* ERROR "p.f undefined" */ = 0 +} + +func f0t[P ~struct{f int}](p P) { + _ = p.f // ERROR "p.f undefined" + p.f /* ERROR "p.f undefined" */ = 0 +} + +var _ = f0[Sf] +var _ = f0t[Sf] + +var _ = f0[Sm /* ERROR "does not satisfy" */ ] +var _ = f0t[Sm /* ERROR "does not satisfy" */ ] + +func f1[P interface{ Sf; m() }](p P) { + _ = p.f // ERROR "p.f undefined" + p.f /* ERROR "p.f undefined" */ = 0 + p.m() +} + +var _ = f1[Sf /* ERROR "missing method m" */ ] +var _ = f1[Sm /* ERROR "does not satisfy" */ ] + +type Sm struct {} + +func (Sm) m() {} + +type Sfm struct { + f int +} + +func (Sfm) m() {} + +func f2[P interface{ Sfm; m() }](p P) { + _ = p.f // ERROR "p.f undefined" + p.f /* ERROR "p.f undefined" */ = 0 + p.m() +} + +var _ = f2[Sfm] + +// special case: core type is a named pointer type + +type PSfm *Sfm + +func f3[P interface{ PSfm }](p P) { + _ = p.f // ERROR "p.f undefined" + p.f /* ERROR "p.f undefined" */ = 0 + p.m /* ERROR "type P has no field or method m" */ () +} + +var _ = f3[PSfm] diff --git a/src/internal/types/testdata/fixedbugs/issue50426.go b/src/internal/types/testdata/fixedbugs/issue50426.go new file mode 100644 index 0000000..17ec0ce --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50426.go @@ -0,0 +1,44 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type A1 [2]uint64 +type A2 [2]uint64 + +func (a A1) m() A1 { return a } +func (a A2) m() A2 { return a } + +func f[B any, T interface { + A1 | A2 + m() T +}](v T) { +} + +func _() { + var v A2 + // Use function type inference to infer type A2 for T. + // Don't use constraint type inference before function + // type inference for typed arguments, otherwise it would + // infer type [2]uint64 for T which doesn't have method m + // (was the bug). + f[int](v) +} + +// Keep using constraint type inference before function type +// inference for untyped arguments so we infer type float64 +// for E below, and not int (which would not work). +func g[S ~[]E, E any](S, E) {} + +func _() { + var s []float64 + g[[]float64](s, 0) +} + +// Keep using constraint type inference after function +// type inference for untyped arguments so we infer +// missing type arguments for which we only have the +// untyped arguments as starting point. +func h[E any, R []E](v E) R { return R{v} } +func _() []int { return h(0) } diff --git a/src/internal/types/testdata/fixedbugs/issue50427.go b/src/internal/types/testdata/fixedbugs/issue50427.go new file mode 100644 index 0000000..d89d63e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50427.go @@ -0,0 +1,23 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// The parser no longer parses type parameters for methods. +// In the past, type checking the code below led to a crash (#50427). + +type T interface{ m[ /* ERROR "must have no type parameters" */ P any]() } + +func _(t T) { + var _ interface{ m[ /* ERROR "must have no type parameters" */ P any](); n() } = t /* ERROR "does not implement" */ +} + +type S struct{} + +func (S) m[ /* ERROR "must have no type parameters" */ P any]() {} + +func _(s S) { + var _ interface{ m[ /* ERROR "must have no type parameters" */ P any](); n() } = s /* ERROR "does not implement" */ + +} diff --git a/src/internal/types/testdata/fixedbugs/issue50450.go b/src/internal/types/testdata/fixedbugs/issue50450.go new file mode 100644 index 0000000..bae3111 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50450.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S struct{} + +func f[P S]() {} + +var _ = f[S] diff --git a/src/internal/types/testdata/fixedbugs/issue50516.go b/src/internal/types/testdata/fixedbugs/issue50516.go new file mode 100644 index 0000000..fcaefed --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50516.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P struct{ f int }](x P) { + _ = x.g // ERROR "type P has no field or method g" +} + +func _[P struct{ f int } | struct{ g int }](x P) { + _ = x.g // ERROR "type P has no field or method g" +} diff --git a/src/internal/types/testdata/fixedbugs/issue50646.go b/src/internal/types/testdata/fixedbugs/issue50646.go new file mode 100644 index 0000000..2c16cfc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50646.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[_ comparable]() {} +func f2[_ interface{ comparable }]() {} + +type T interface{ m() } + +func _[P comparable, Q ~int, R any]() { + _ = f1[int] + _ = f1[T] + _ = f1[any] + _ = f1[P] + _ = f1[Q] + _ = f1[R /* ERROR "R does not satisfy comparable" */] + + _ = f2[int] + _ = f2[T] + _ = f2[any] + _ = f2[P] + _ = f2[Q] + _ = f2[R /* ERROR "R does not satisfy comparable" */] +} diff --git a/src/internal/types/testdata/fixedbugs/issue50729.go b/src/internal/types/testdata/fixedbugs/issue50729.go new file mode 100644 index 0000000..fe19fdf --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50729.go @@ -0,0 +1,19 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// version 1 +var x1 T1[B1] + +type T1[_ any] struct{} +type A1 T1[B1] +type B1 = T1[A1] + +// version 2 +type T2[_ any] struct{} +type A2 T2[B2] +type B2 = T2[A2] + +var x2 T2[B2] diff --git a/src/internal/types/testdata/fixedbugs/issue50755.go b/src/internal/types/testdata/fixedbugs/issue50755.go new file mode 100644 index 0000000..afc7b24 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50755.go @@ -0,0 +1,47 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// The core type of M2 unifies with the type of m1 +// during function argument type inference. +// M2's constraint is unnamed. +func f1[K1 comparable, E1 any](m1 map[K1]E1) {} + +func f2[M2 map[string]int](m2 M2) { + f1(m2) +} + +// The core type of M3 unifies with the type of m1 +// during function argument type inference. +// M3's constraint is named. +type Map3 map[string]int + +func f3[M3 Map3](m3 M3) { + f1(m3) +} + +// The core type of M5 unifies with the core type of M4 +// during constraint type inference. +func f4[M4 map[K4]int, K4 comparable](m4 M4) {} + +func f5[M5 map[K5]int, K5 comparable](m5 M5) { + f4(m5) +} + +// test case from issue + +func Copy[MC ~map[KC]VC, KC comparable, VC any](dst, src MC) { + for k, v := range src { + dst[k] = v + } +} + +func Merge[MM ~map[KM]VM, KM comparable, VM any](ms ...MM) MM { + result := MM{} + for _, m := range ms { + Copy(result, m) + } + return result +} diff --git a/src/internal/types/testdata/fixedbugs/issue50779.go b/src/internal/types/testdata/fixedbugs/issue50779.go new file mode 100644 index 0000000..09ddf53 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50779.go @@ -0,0 +1,23 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type AC interface { + C +} + +type ST []int + +type R[S any, P any] struct{} + +type SR = R[SS, ST] + +type SS interface { + NSR(any) *SR // ERROR "invalid use of type alias SR in recursive type" +} + +type C interface { + NSR(any) *SR +} diff --git a/src/internal/types/testdata/fixedbugs/issue50782.go b/src/internal/types/testdata/fixedbugs/issue50782.go new file mode 100644 index 0000000..97e8f6c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50782.go @@ -0,0 +1,47 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Field accesses through type parameters are disabled +// until we have a more thorough understanding of the +// implications on the spec. See issue #51576. + +package p + +// The first example from the issue. +type Numeric interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 +} + +// numericAbs matches numeric types with an Abs method. +type numericAbs[T Numeric] interface { + ~struct{ Value T } + Abs() T +} + +// AbsDifference computes the absolute value of the difference of +// a and b, where the absolute value is determined by the Abs method. +func absDifference[T numericAbs[T /* ERROR "T does not satisfy Numeric" */]](a, b T) T { + // Field accesses are not permitted for now. Keep an error so + // we can find and fix this code once the situation changes. + return a.Value // ERROR "a.Value undefined" + // TODO: The error below should probably be positioned on the '-'. + // d := a /* ERROR "invalid operation: operator - not defined" */ .Value - b.Value + // return d.Abs() +} + +// The second example from the issue. +type T[P int] struct{ f P } + +func _[P T[P /* ERROR "P does not satisfy int" */ ]]() {} + +// Additional tests +func _[P T[T /* ERROR "T[P] does not satisfy int" */ [P /* ERROR "P does not satisfy int" */ ]]]() {} +func _[P T[Q /* ERROR "Q does not satisfy int" */ ], Q T[P /* ERROR "P does not satisfy int" */ ]]() {} +func _[P T[Q], Q int]() {} + +type C[P comparable] struct{ f P } +func _[P C[C[P]]]() {} +func _[P C[C /* ERROR "C[Q] does not satisfy comparable" */ [Q /* ERROR "Q does not satisfy comparable" */]], Q func()]() {} +func _[P [10]C[P]]() {} +func _[P struct{ f C[C[P]]}]() {} diff --git a/src/internal/types/testdata/fixedbugs/issue50816.go b/src/internal/types/testdata/fixedbugs/issue50816.go new file mode 100644 index 0000000..b7c28cd --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50816.go @@ -0,0 +1,23 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package pkg + +type I interface { + Foo() +} + +type T1 struct{} + +func (T1) foo() {} + +type T2 struct{} + +func (T2) foo() string { return "" } + +func _() { + var i I + _ = i /* ERROR "impossible type assertion: i.(T1)\n\tT1 does not implement I (missing method Foo)\n\t\thave foo()\n\t\twant Foo()" */ .(T1) + _ = i /* ERROR "impossible type assertion: i.(T2)\n\tT2 does not implement I (missing method Foo)\n\t\thave foo() string\n\t\twant Foo()" */ .(T2) +} diff --git a/src/internal/types/testdata/fixedbugs/issue50833.go b/src/internal/types/testdata/fixedbugs/issue50833.go new file mode 100644 index 0000000..e912e4d --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50833.go @@ -0,0 +1,16 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + S struct{ f int } + PS *S +) + +func a() []*S { return []*S{{f: 1}} } +func b() []PS { return []PS{{f: 1}} } + +func c[P *S]() []P { return []P{{f: 1}} } +func d[P PS]() []P { return []P{{f: 1}} } diff --git a/src/internal/types/testdata/fixedbugs/issue50912.go b/src/internal/types/testdata/fixedbugs/issue50912.go new file mode 100644 index 0000000..a99fa7b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50912.go @@ -0,0 +1,19 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func Real[P ~complex128](x P) { + _ = real(x /* ERROR "not supported" */ ) +} + +func Imag[P ~complex128](x P) { + _ = imag(x /* ERROR "not supported" */ ) +} + +func Complex[P ~float64](x P) { + _ = complex(x /* ERROR "not supported" */ , 0) + _ = complex(0 /* ERROR "not supported" */ , x) + _ = complex(x /* ERROR "not supported" */ , x) +} diff --git a/src/internal/types/testdata/fixedbugs/issue50918.go b/src/internal/types/testdata/fixedbugs/issue50918.go new file mode 100644 index 0000000..5744fa8 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50918.go @@ -0,0 +1,21 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type thing1 struct { + things []string +} + +type thing2 struct { + things []thing1 +} + +func _() { + var a1, b1 thing1 + _ = a1 /* ERROR "struct containing []string cannot be compared" */ == b1 + + var a2, b2 thing2 + _ = a2 /* ERROR "struct containing []thing1 cannot be compared" */ == b2 +} diff --git a/src/internal/types/testdata/fixedbugs/issue50929.go b/src/internal/types/testdata/fixedbugs/issue50929.go new file mode 100644 index 0000000..64c7cd6 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50929.go @@ -0,0 +1,68 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is tested when running "go test -run Manual" +// without source arguments. Use for one-off debugging. + +package p + +import "fmt" + +type F[A, B any] int + +func G[A, B any](F[A, B]) { +} + +func _() { + // TODO(gri) only report one error below (issue #50932) + var x F /* ERROR "got 1 arguments but 2 type parameters" */ [int] + G(x /* ERROR "does not match" */) +} + +// test case from issue +// (lots of errors but doesn't crash anymore) + +type RC[G any, RG any] interface { + ~[]RG +} + +type RG[G any] struct{} + +type RSC[G any] []*RG[G] + +type M[Rc RC[G, RG], G any, RG any] struct { + Fn func(Rc) +} + +type NFn[Rc RC[G, RG], G any, RG any] func(Rc) + +func NC[Rc RC[G, RG], G any, RG any](nFn NFn[Rc, G, RG]) { + var empty Rc + nFn(empty) +} + +func NSG[G any](c RSC[G]) { + fmt.Println(c) +} + +func MMD[Rc RC /* ERROR "got 1 arguments" */ [RG], RG any, G any]() M /* ERROR "got 2 arguments" */ [Rc, RG] { + + var nFn NFn /* ERROR "got 2 arguments" */ [Rc, RG] + + var empty Rc + switch any(empty).(type) { + case BC /* ERROR "undefined: BC" */ : + + case RSC[G]: + nFn = NSG /* ERROR "cannot use NSG[G]" */ [G] + } + + return M /* ERROR "got 2 arguments" */ [Rc, RG]{ + Fn: func(rc Rc) { + NC(nFn /* ERROR "does not match" */ ) + }, + } + + return M /* ERROR "got 2 arguments" */ [Rc, RG]{} +} diff --git a/src/internal/types/testdata/fixedbugs/issue50965.go b/src/internal/types/testdata/fixedbugs/issue50965.go new file mode 100644 index 0000000..79059e9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue50965.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(x int, c string) { + switch x { + case c /* ERROR "invalid case c in switch on x (mismatched types string and int)" */ : + } +} + +func _(x, c []int) { + switch x { + case c /* ERROR "invalid case c in switch on x (slice can only be compared to nil)" */ : + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue51025.go b/src/internal/types/testdata/fixedbugs/issue51025.go new file mode 100644 index 0000000..caaabd5 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51025.go @@ -0,0 +1,38 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var _ interface{ m() } = struct /* ERROR "m is a field, not a method" */ { + m func() +}{} + +var _ interface{ m() } = & /* ERROR "m is a field, not a method" */ struct { + m func() +}{} + +var _ interface{ M() } = struct /* ERROR "missing method M" */ { + m func() +}{} + +var _ interface{ M() } = & /* ERROR "missing method M" */ struct { + m func() +}{} + +// test case from issue +type I interface{ m() } +type T struct{ m func() } +type M struct{} + +func (M) m() {} + +func _() { + var t T + var m M + var i I + + i = m + i = t // ERROR "m is a field, not a method" + _ = i +} diff --git a/src/internal/types/testdata/fixedbugs/issue51048.go b/src/internal/types/testdata/fixedbugs/issue51048.go new file mode 100644 index 0000000..5830837 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51048.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P int]() { + _ = f[P] +} + +func f[T int]() {} diff --git a/src/internal/types/testdata/fixedbugs/issue51139.go b/src/internal/types/testdata/fixedbugs/issue51139.go new file mode 100644 index 0000000..4c460d4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51139.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[S []T, T any](S, T) {} + +func _() { + type L chan int + f([]L{}, make(chan int)) + f([]L{}, make(L)) + f([]chan int{}, make(chan int)) + f /* ERROR "[]chan int does not satisfy []L ([]chan int missing in []p.L)" */ ([]chan int{}, make(L)) +} + +// test case from issue + +func Append[S ~[]T, T any](s S, x ...T) S { /* implementation of append */ return s } + +func _() { + type MyPtr *int + var x []MyPtr + _ = append(x, new(int)) + _ = Append(x, new(int)) +} diff --git a/src/internal/types/testdata/fixedbugs/issue51145.go b/src/internal/types/testdata/fixedbugs/issue51145.go new file mode 100644 index 0000000..1f970d9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51145.go @@ -0,0 +1,18 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "fmt" + +type ( + _ [fmt /* ERROR "invalid array length fmt" */ ]int + _ [float64 /* ERROR "invalid array length float64" */ ]int + _ [f /* ERROR "invalid array length f" */ ]int + _ [nil /* ERROR "invalid array length nil" */ ]int +) + +func f() + +var _ fmt.Stringer // use fmt diff --git a/src/internal/types/testdata/fixedbugs/issue51158.go b/src/internal/types/testdata/fixedbugs/issue51158.go new file mode 100644 index 0000000..3edc505 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51158.go @@ -0,0 +1,18 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Type checking the following code should not cause an infinite recursion. +func f[M map[K]int, K comparable](m M) { + f(m) +} + +// Equivalent code using mutual recursion. +func f1[M map[K]int, K comparable](m M) { + f2(m) +} +func f2[M map[K]int, K comparable](m M) { + f1(m) +} diff --git a/src/internal/types/testdata/fixedbugs/issue51229.go b/src/internal/types/testdata/fixedbugs/issue51229.go new file mode 100644 index 0000000..22a9113 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51229.go @@ -0,0 +1,164 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Constraint type inference should be independent of the +// ordering of the type parameter declarations. Try all +// permutations in the test case below. +// Permutations produced by https://go.dev/play/p/PHcZNGJTEBZ. + +func f00[S1 ~[]E1, S2 ~[]E2, E1 ~byte, E2 ~byte](S1, S2) {} +func f01[S2 ~[]E2, S1 ~[]E1, E1 ~byte, E2 ~byte](S1, S2) {} +func f02[E1 ~byte, S1 ~[]E1, S2 ~[]E2, E2 ~byte](S1, S2) {} +func f03[S1 ~[]E1, E1 ~byte, S2 ~[]E2, E2 ~byte](S1, S2) {} +func f04[S2 ~[]E2, E1 ~byte, S1 ~[]E1, E2 ~byte](S1, S2) {} +func f05[E1 ~byte, S2 ~[]E2, S1 ~[]E1, E2 ~byte](S1, S2) {} +func f06[E2 ~byte, S2 ~[]E2, S1 ~[]E1, E1 ~byte](S1, S2) {} +func f07[S2 ~[]E2, E2 ~byte, S1 ~[]E1, E1 ~byte](S1, S2) {} +func f08[S1 ~[]E1, E2 ~byte, S2 ~[]E2, E1 ~byte](S1, S2) {} +func f09[E2 ~byte, S1 ~[]E1, S2 ~[]E2, E1 ~byte](S1, S2) {} +func f10[S2 ~[]E2, S1 ~[]E1, E2 ~byte, E1 ~byte](S1, S2) {} +func f11[S1 ~[]E1, S2 ~[]E2, E2 ~byte, E1 ~byte](S1, S2) {} +func f12[S1 ~[]E1, E1 ~byte, E2 ~byte, S2 ~[]E2](S1, S2) {} +func f13[E1 ~byte, S1 ~[]E1, E2 ~byte, S2 ~[]E2](S1, S2) {} +func f14[E2 ~byte, S1 ~[]E1, E1 ~byte, S2 ~[]E2](S1, S2) {} +func f15[S1 ~[]E1, E2 ~byte, E1 ~byte, S2 ~[]E2](S1, S2) {} +func f16[E1 ~byte, E2 ~byte, S1 ~[]E1, S2 ~[]E2](S1, S2) {} +func f17[E2 ~byte, E1 ~byte, S1 ~[]E1, S2 ~[]E2](S1, S2) {} +func f18[E2 ~byte, E1 ~byte, S2 ~[]E2, S1 ~[]E1](S1, S2) {} +func f19[E1 ~byte, E2 ~byte, S2 ~[]E2, S1 ~[]E1](S1, S2) {} +func f20[S2 ~[]E2, E2 ~byte, E1 ~byte, S1 ~[]E1](S1, S2) {} +func f21[E2 ~byte, S2 ~[]E2, E1 ~byte, S1 ~[]E1](S1, S2) {} +func f22[E1 ~byte, S2 ~[]E2, E2 ~byte, S1 ~[]E1](S1, S2) {} +func f23[S2 ~[]E2, E1 ~byte, E2 ~byte, S1 ~[]E1](S1, S2) {} + +type myByte byte + +func _(a []byte, b []myByte) { + f00(a, b) + f01(a, b) + f02(a, b) + f03(a, b) + f04(a, b) + f05(a, b) + f06(a, b) + f07(a, b) + f08(a, b) + f09(a, b) + f10(a, b) + f11(a, b) + f12(a, b) + f13(a, b) + f14(a, b) + f15(a, b) + f16(a, b) + f17(a, b) + f18(a, b) + f19(a, b) + f20(a, b) + f21(a, b) + f22(a, b) + f23(a, b) +} + +// Constraint type inference may have to iterate. +// Again, the order of the type parameters shouldn't matter. + +func g0[S ~[]E, M ~map[string]S, E any](m M) {} +func g1[M ~map[string]S, S ~[]E, E any](m M) {} +func g2[E any, S ~[]E, M ~map[string]S](m M) {} +func g3[S ~[]E, E any, M ~map[string]S](m M) {} +func g4[M ~map[string]S, E any, S ~[]E](m M) {} +func g5[E any, M ~map[string]S, S ~[]E](m M) {} + +func _(m map[string][]byte) { + g0(m) + g1(m) + g2(m) + g3(m) + g4(m) + g5(m) +} + +// Worst-case scenario. +// There are 10 unknown type parameters. In each iteration of +// constraint type inference we infer one more, from right to left. +// Each iteration looks repeatedly at all 11 type parameters, +// requiring a total of 10*11 = 110 iterations with the current +// implementation. Pathological case. + +func h[K any, J ~*K, I ~*J, H ~*I, G ~*H, F ~*G, E ~*F, D ~*E, C ~*D, B ~*C, A ~*B](x A) {} + +func _(x **********int) { + h(x) +} + +// Examples with channel constraints and tilde. + +func ch1[P chan<- int]() (_ P) { return } // core(P) == chan<- int (single type, no tilde) +func ch2[P ~chan int]() { return } // core(P) == ~chan<- int (tilde) +func ch3[P chan E, E any](E) { return } // core(P) == chan<- E (single type, no tilde) +func ch4[P chan E | ~chan<- E, E any](E) { return } // core(P) == ~chan<- E (tilde) +func ch5[P chan int | chan<- int]() { return } // core(P) == chan<- int (not a single type) + +func _() { + // P can be inferred as there's a single specific type and no tilde. + var _ chan int = ch1 /* ERRORx `cannot use ch1.*value of type chan<- int` */ () + var _ chan<- int = ch1() + + // P cannot be inferred as there's a tilde. + ch2 /* ERROR "cannot infer P" */ () + type myChan chan int + ch2[myChan]() + + // P can be inferred as there's a single specific type and no tilde. + var e int + ch3(e) + + // P cannot be inferred as there's more than one specific type and a tilde. + ch4 /* ERROR "cannot infer P" */ (e) + _ = ch4[chan int] + + // P cannot be inferred as there's more than one specific type. + ch5 /* ERROR "cannot infer P" */ () + ch5[chan<- int]() +} + +// test case from issue + +func equal[M1 ~map[K1]V1, M2 ~map[K2]V2, K1, K2 ~uint32, V1, V2 ~string](m1 M1, m2 M2) bool { + if len(m1) != len(m2) { + return false + } + for k, v1 := range m1 { + if v2, ok := m2[K2(k)]; !ok || V2(v1) != v2 { + return false + } + } + return true +} + +func equalFixed[K1, K2 ~uint32, V1, V2 ~string](m1 map[K1]V1, m2 map[K2]V2) bool { + if len(m1) != len(m2) { + return false + } + for k, v1 := range m1 { + if v2, ok := m2[K2(k)]; !ok || v1 != V1(v2) { + return false + } + } + return true +} + +type ( + someNumericID uint32 + someStringID string +) + +func _() { + foo := map[uint32]string{10: "bar"} + bar := map[someNumericID]someStringID{10: "bar"} + equal(foo, bar) +} diff --git a/src/internal/types/testdata/fixedbugs/issue51232.go b/src/internal/types/testdata/fixedbugs/issue51232.go new file mode 100644 index 0000000..27693a3 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51232.go @@ -0,0 +1,30 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type RC[RG any] interface { + ~[]RG +} + +type Fn[RCT RC[RG], RG any] func(RCT) + +type F[RCT RC[RG], RG any] interface { + Fn() Fn /* ERROR "got 1 arguments" */ [RCT] +} + +type concreteF[RCT RC[RG], RG any] struct { + makeFn func() Fn /* ERROR "got 1 arguments" */ [RCT] +} + +func (c *concreteF[RCT, RG]) Fn() Fn /* ERROR "got 1 arguments" */ [RCT] { + return c.makeFn() +} + +func NewConcrete[RCT RC[RG], RG any](Rc RCT) F /* ERROR "got 1 arguments" */ [RCT] { + // TODO(rfindley): eliminate the duplicate error below. + return & /* ERRORx `cannot use .* as F\[RCT\]` */ concreteF /* ERROR "got 1 arguments" */ [RCT]{ + makeFn: nil, + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue51233.go b/src/internal/types/testdata/fixedbugs/issue51233.go new file mode 100644 index 0000000..e2f97fc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51233.go @@ -0,0 +1,27 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// As of issue #51527, type-type inference has been disabled. + +type RC[RG any] interface { + ~[]RG +} + +type Fn[RCT RC[RG], RG any] func(RCT) + +type FFn[RCT RC[RG], RG any] func() Fn /* ERROR "got 1 arguments" */ [RCT] + +type F[RCT RC[RG], RG any] interface { + Fn() Fn /* ERROR "got 1 arguments" */ [RCT] +} + +type concreteF[RCT RC[RG], RG any] struct { + makeFn FFn /* ERROR "got 1 arguments" */ [RCT] +} + +func (c *concreteF[RCT, RG]) Fn() Fn /* ERROR "got 1 arguments" */ [RCT] { + return c.makeFn() +} diff --git a/src/internal/types/testdata/fixedbugs/issue51257.go b/src/internal/types/testdata/fixedbugs/issue51257.go new file mode 100644 index 0000000..828612b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51257.go @@ -0,0 +1,46 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[_ comparable]() {} + +type S1 struct{ x int } +type S2 struct{ x any } +type S3 struct{ x [10]interface{ m() } } + +func _[P1 comparable, P2 S2]() { + _ = f[S1] + _ = f[S2] + _ = f[S3] + + type L1 struct { x P1 } + type L2 struct { x P2 } + _ = f[L1] + _ = f[L2 /* ERROR "L2 does not satisfy comparable" */ ] +} + + +// example from issue + +type Set[T comparable] map[T]struct{} + +func NewSetFromSlice[T comparable](items []T) *Set[T] { + s := Set[T]{} + + for _, item := range items { + s[item] = struct{}{} + } + + return &s +} + +type T struct{ x any } + +func main() { + NewSetFromSlice([]T{ + {"foo"}, + {5}, + }) +} diff --git a/src/internal/types/testdata/fixedbugs/issue51335.go b/src/internal/types/testdata/fixedbugs/issue51335.go new file mode 100644 index 0000000..04dc04e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51335.go @@ -0,0 +1,16 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S1 struct{} +type S2 struct{} + +func _[P *S1|*S2]() { + _= []P{{ /* ERROR "invalid composite literal element type P (no core type)" */ }} +} + +func _[P *S1|S1]() { + _= []P{{ /* ERROR "invalid composite literal element type P (no core type)" */ }} +} diff --git a/src/internal/types/testdata/fixedbugs/issue51339.go b/src/internal/types/testdata/fixedbugs/issue51339.go new file mode 100644 index 0000000..65c2134 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51339.go @@ -0,0 +1,18 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This file is tested when running "go test -run Manual" +// without source arguments. Use for one-off debugging. + +package p + +type T[P any, B *P] struct{} + +func (T /* ERROR "cannot use generic type" */ ) m0() {} + +// TODO(rfindley): eliminate the duplicate errors here. +func (/* ERROR "got 1 type parameter, but receiver base type declares 2" */ T /* ERROR "got 1 arguments but 2 type parameters" */ [_]) m1() {} +func (T[_, _]) m2() {} +// TODO(gri) this error is unfortunate (issue #51343) +func (T /* ERROR "got 3 arguments but 2 type parameters" */ [_, _, _]) m3() {} diff --git a/src/internal/types/testdata/fixedbugs/issue51360.go b/src/internal/types/testdata/fixedbugs/issue51360.go new file mode 100644 index 0000000..1b9c45a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51360.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + len.Println /* ERROR "cannot select on len" */ + len.Println /* ERROR "cannot select on len" */ () + _ = len.Println /* ERROR "cannot select on len" */ + _ = len /* ERROR "cannot index len" */ [0] + _ = *len /* ERROR "cannot indirect len" */ +} diff --git a/src/internal/types/testdata/fixedbugs/issue51376.go b/src/internal/types/testdata/fixedbugs/issue51376.go new file mode 100644 index 0000000..0f2ab8e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51376.go @@ -0,0 +1,24 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Map map[string]int + +func f[M ~map[K]V, K comparable, V any](M) {} +func g[M map[K]V, K comparable, V any](M) {} + +func _[M1 ~map[K]V, M2 map[K]V, K comparable, V any]() { + var m1 M1 + f(m1) + g /* ERROR "M1 does not satisfy map[K]V" */ (m1) // M1 has tilde + + var m2 M2 + f(m2) + g(m2) // M1 does not have tilde + + var m3 Map + f(m3) + g /* ERROR "Map does not satisfy map[string]int" */ (m3) // M in g does not have tilde +} diff --git a/src/internal/types/testdata/fixedbugs/issue51386.go b/src/internal/types/testdata/fixedbugs/issue51386.go new file mode 100644 index 0000000..ef62239 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51386.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type myString string + +func _[P ~string | ~[]byte | ~[]rune]() { + _ = P("") + const s myString = "" + _ = P(s) +} + +func _[P myString]() { + _ = P("") +} diff --git a/src/internal/types/testdata/fixedbugs/issue51437.go b/src/internal/types/testdata/fixedbugs/issue51437.go new file mode 100644 index 0000000..3762615 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51437.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T struct{} + +func (T) m() []int { return nil } + +func f(x T) { + for _, x := range func() []int { + return x.m() // x declared in parameter list of f + }() { + _ = x // x declared by range clause + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue51472.go b/src/internal/types/testdata/fixedbugs/issue51472.go new file mode 100644 index 0000000..6dfff05 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51472.go @@ -0,0 +1,54 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[T comparable](x T) { + _ = x == x +} + +func _[T interface{interface{comparable}}](x T) { + _ = x == x +} + +func _[T interface{comparable; interface{comparable}}](x T) { + _ = x == x +} + +func _[T interface{comparable; ~int}](x T) { + _ = x == x +} + +func _[T interface{comparable; ~[]byte}](x T) { + _ = x /* ERROR "empty type set" */ == x +} + +// TODO(gri) The error message here should be better. See issue #51525. +func _[T interface{comparable; ~int; ~string}](x T) { + _ = x /* ERROR "empty type set" */ == x +} + +// TODO(gri) The error message here should be better. See issue #51525. +func _[T interface{~int; ~string}](x T) { + _ = x /* ERROR "empty type set" */ == x +} + +func _[T interface{comparable; interface{~int}; interface{int|float64}}](x T) { + _ = x == x +} + +func _[T interface{interface{comparable; ~int}; interface{~float64; comparable; m()}}](x T) { + _ = x /* ERROR "empty type set" */ == x +} + +// test case from issue + +func f[T interface{comparable; []byte|string}](x T) { + _ = x == x +} + +func _(s []byte) { + f /* ERROR "T (type []byte) does not satisfy interface{comparable; []byte | string}" */ (s) // TODO(gri) better error message (T's type set only contains string!) + _ = f[[ /* ERROR "does not satisfy" */ ]byte] +} diff --git a/src/internal/types/testdata/fixedbugs/issue51509.go b/src/internal/types/testdata/fixedbugs/issue51509.go new file mode 100644 index 0000000..4737a82 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51509.go @@ -0,0 +1,7 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T /* ERROR "invalid recursive type" */ T.x diff --git a/src/internal/types/testdata/fixedbugs/issue51525.go b/src/internal/types/testdata/fixedbugs/issue51525.go new file mode 100644 index 0000000..af569c8 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51525.go @@ -0,0 +1,20 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[T interface { + int + string +}](x T) { + _ = x /* ERROR "empty type set" */ == x + _ = x /* ERROR "empty type set" */ + x + <-x /* ERROR "empty type set" */ + x <- /* ERROR "empty type set" */ 0 + close(x /* ERROR "empty type set" */) +} + +func _[T interface{ int | []byte }](x T) { + _ = x /* ERROR "incomparable types in type set" */ == x +} diff --git a/src/internal/types/testdata/fixedbugs/issue51533.go b/src/internal/types/testdata/fixedbugs/issue51533.go new file mode 100644 index 0000000..166e5fe --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51533.go @@ -0,0 +1,20 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(x any) { + switch x { + case 0: + fallthrough // ERROR "fallthrough statement out of place" + _ = x + default: + } + + switch x.(type) { + case int: + fallthrough // ERROR "cannot fallthrough in type switch" + default: + } +} diff --git a/src/internal/types/testdata/fixedbugs/issue51578.go b/src/internal/types/testdata/fixedbugs/issue51578.go new file mode 100644 index 0000000..d2e8a28 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51578.go @@ -0,0 +1,17 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var _ = (*interface /* ERROR "interface contains type constraints" */ {int})(nil) + +// abbreviated test case from issue + +type TypeSet interface{ int | string } + +func _() { + f((*TypeSet /* ERROR "interface contains type constraints" */)(nil)) +} + +func f(any) {}
\ No newline at end of file diff --git a/src/internal/types/testdata/fixedbugs/issue51593.go b/src/internal/types/testdata/fixedbugs/issue51593.go new file mode 100644 index 0000000..62b0a56 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51593.go @@ -0,0 +1,13 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P interface{ m(R) }, R any]() {} + +type T = interface { m(int) } + +func _() { + _ = f[T] +} diff --git a/src/internal/types/testdata/fixedbugs/issue51607.go b/src/internal/types/testdata/fixedbugs/issue51607.go new file mode 100644 index 0000000..298adad --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51607.go @@ -0,0 +1,65 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Interface types must be ignored during overlap test. + +type ( + T1 interface{int} + T2 interface{~int} + T3 interface{T1 | bool | string} + T4 interface{T2 | ~bool | ~string} +) + +type ( + // overlap errors for non-interface terms + // (like the interface terms, but explicitly inlined) + _ interface{int | int /* ERROR "overlapping terms int and int" */ } + _ interface{int | ~ /* ERROR "overlapping terms ~int and int" */ int} + _ interface{~int | int /* ERROR "overlapping terms int and ~int" */ } + _ interface{~int | ~ /* ERROR "overlapping terms ~int and ~int" */ int} + + _ interface{T1 | bool | string | T1 | bool /* ERROR "overlapping terms bool and bool" */ | string /* ERROR "overlapping terms string and string" */ } + _ interface{T1 | bool | string | T2 | ~ /* ERROR "overlapping terms ~bool and bool" */ bool | ~ /* ERROR "overlapping terms ~string and string" */ string} + + // no errors for interface terms + _ interface{T1 | T1} + _ interface{T1 | T2} + _ interface{T2 | T1} + _ interface{T2 | T2} + + _ interface{T3 | T3 | int} + _ interface{T3 | T4 | bool } + _ interface{T4 | T3 | string } + _ interface{T4 | T4 | float64 } +) + +func _[_ T1 | bool | string | T1 | bool /* ERROR "overlapping terms" */ ]() {} +func _[_ T1 | bool | string | T2 | ~ /* ERROR "overlapping terms" */ bool ]() {} +func _[_ T2 | ~bool | ~string | T1 | bool /* ERROR "overlapping terms" */ ]() {} +func _[_ T2 | ~bool | ~string | T2 | ~ /* ERROR "overlapping terms" */ bool ]() {} + +func _[_ T3 | T3 | int]() {} +func _[_ T3 | T4 | bool]() {} +func _[_ T4 | T3 | string]() {} +func _[_ T4 | T4 | float64]() {} + +// test cases from issue + +type _ interface { + interface {bool | int} | interface {bool | string} +} + +type _ interface { + interface {bool | int} ; interface {bool | string} +} + +type _ interface { + interface {bool; int} ; interface {bool; string} +} + +type _ interface { + interface {bool; int} | interface {bool; string} +}
\ No newline at end of file diff --git a/src/internal/types/testdata/fixedbugs/issue51610.go b/src/internal/types/testdata/fixedbugs/issue51610.go new file mode 100644 index 0000000..d0bc1ac --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51610.go @@ -0,0 +1,9 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _[P int | float64 | complex128]() { + _ = map[P]int{1: 1, 1.0 /* ERROR "duplicate key 1" */ : 2, 1 /* ERROR "duplicate key (1 + 0i)" */ + 0i: 3} +} diff --git a/src/internal/types/testdata/fixedbugs/issue51616.go b/src/internal/types/testdata/fixedbugs/issue51616.go new file mode 100644 index 0000000..e0efc9e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51616.go @@ -0,0 +1,19 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + C[T any] interface{~int; M() T} + + _ C[bool] + _ comparable + _ interface {~[]byte | ~string} + + // Alias type declarations may refer to "constraint" types + // like ordinary type declarations. + _ = C[bool] + _ = comparable + _ = interface {~[]byte | ~string} +) diff --git a/src/internal/types/testdata/fixedbugs/issue51658.go b/src/internal/types/testdata/fixedbugs/issue51658.go new file mode 100644 index 0000000..36e2fdd --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51658.go @@ -0,0 +1,43 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// This test checks syntax errors which differ between +// go/parser and the syntax package. +// TODO: consolidate eventually + +package p + +type F { // ERRORx "expected type|type declaration" + float64 +} // ERRORx "expected declaration|non-declaration statement" + +func _[T F | int](x T) { + _ = x == 0 // don't crash when recording type of 0 +} + +// test case from issue + +type FloatType { // ERRORx "expected type|type declaration" + float32 | float64 +} // ERRORx "expected declaration|non-declaration statement" + +type IntegerType interface { + int8 | int16 | int32 | int64 | int | + uint8 | uint16 | uint32 | uint64 | uint +} + +type ComplexType interface { + complex64 | complex128 +} + +type Number interface { + FloatType | IntegerType | ComplexType +} + +func GetDefaultNumber[T Number](value, defaultValue T) T { + if value == 0 { + return defaultValue + } + return value +} diff --git a/src/internal/types/testdata/fixedbugs/issue51877.go b/src/internal/types/testdata/fixedbugs/issue51877.go new file mode 100644 index 0000000..4b0e5bc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue51877.go @@ -0,0 +1,18 @@ +// Copyright 2013 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S struct { + f1 int + f2 bool +} + +var ( + _ = S{0} /* ERROR "too few values in struct literal" */ + _ = struct{ f1, f2 int }{0} /* ERROR "too few values in struct literal" */ + + _ = S{0, true, "foo" /* ERROR "too many values in struct literal" */} + _ = struct{ f1, f2 int }{0, 1, 2 /* ERROR "too many values in struct literal" */} +) diff --git a/src/internal/types/testdata/fixedbugs/issue52031.go b/src/internal/types/testdata/fixedbugs/issue52031.go new file mode 100644 index 0000000..5b75b75 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue52031.go @@ -0,0 +1,33 @@ +// -lang=go1.12 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type resultFlags uint + +// Example from #52031. +// +// The following shifts should not produce errors on Go < 1.13, as their +// untyped constant operands are representable by type uint. +const ( + _ resultFlags = (1 << iota) / 2 + + reportEqual + reportUnequal + reportByIgnore + reportByMethod + reportByFunc + reportByCycle +) + +// Invalid cases. +var x int = 1 +var _ = (8 << x /* ERRORx `signed shift count .* requires go1.13 or later` */) + +const _ = (1 << 1.2 /* ERROR "truncated to uint" */) + +var y float64 +var _ = (1 << y /* ERROR "must be integer" */) diff --git a/src/internal/types/testdata/fixedbugs/issue52401.go b/src/internal/types/testdata/fixedbugs/issue52401.go new file mode 100644 index 0000000..ccc32d3 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue52401.go @@ -0,0 +1,11 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + const x = 0 + x /* ERROR "cannot assign to x" */ += 1 + x /* ERROR "cannot assign to x" */ ++ +} diff --git a/src/internal/types/testdata/fixedbugs/issue52529.go b/src/internal/types/testdata/fixedbugs/issue52529.go new file mode 100644 index 0000000..de7b296 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue52529.go @@ -0,0 +1,15 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Foo[P any] struct { + _ *Bar[P] +} + +type Bar[Q any] Foo[Q] + +func (v *Bar[R]) M() { + _ = (*Foo[R])(v) +} diff --git a/src/internal/types/testdata/fixedbugs/issue52698.go b/src/internal/types/testdata/fixedbugs/issue52698.go new file mode 100644 index 0000000..20f24f0 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue52698.go @@ -0,0 +1,62 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// correctness check: ensure that cycles through generic instantiations are detected +type T[P any] struct { + _ P +} + +type S /* ERROR "invalid recursive type" */ struct { + _ T[S] +} + +// simplified test 1 + +var _ A1[A1[string]] + +type A1[P any] struct { + _ B1[P] +} + +type B1[P any] struct { + _ P +} + +// simplified test 2 +var _ B2[A2] + +type A2 struct { + _ B2[string] +} + +type B2[P any] struct { + _ C2[P] +} + +type C2[P any] struct { + _ P +} + +// test case from issue +type T23 interface { + ~struct { + Field0 T13[T15] + } +} + +type T1[P1 interface { +}] struct { + Field2 P1 +} + +type T13[P2 interface { +}] struct { + Field2 T1[P2] +} + +type T15 struct { + Field0 T13[string] +} diff --git a/src/internal/types/testdata/fixedbugs/issue52915.go b/src/internal/types/testdata/fixedbugs/issue52915.go new file mode 100644 index 0000000..70dc664 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue52915.go @@ -0,0 +1,23 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type T[P any] struct { + T /* ERROR "invalid recursive type" */ [P] +} + +func _[P any]() { + _ = unsafe.Sizeof(T[int]{}) + _ = unsafe.Sizeof(struct{ T[int] }{}) + + _ = unsafe.Sizeof(T[P]{}) + _ = unsafe.Sizeof(struct{ T[P] }{}) +} + +// TODO(gri) This is a follow-on error due to T[int] being invalid. +// We should try to avoid it. +const _ = unsafe /* ERROR "not constant" */ .Sizeof(T[int]{}) diff --git a/src/internal/types/testdata/fixedbugs/issue53358.go b/src/internal/types/testdata/fixedbugs/issue53358.go new file mode 100644 index 0000000..67c095c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue53358.go @@ -0,0 +1,19 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type A struct{} + +func (*A) m() int { return 0 } + +var _ = A.m /* ERROR "invalid method expression A.m (needs pointer receiver (*A).m)" */ () +var _ = (*A).m(nil) + +type B struct{ A } + +var _ = B.m // ERROR "invalid method expression B.m (needs pointer receiver (*B).m)" +var _ = (*B).m + +var _ = struct{ A }.m // ERROR "invalid method expression struct{A}.m (needs pointer receiver (*struct{A}).m)" diff --git a/src/internal/types/testdata/fixedbugs/issue53650.go b/src/internal/types/testdata/fixedbugs/issue53650.go new file mode 100644 index 0000000..4bba59e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue53650.go @@ -0,0 +1,59 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import ( + "reflect" + "testing" +) + +type T1 int +type T2 int + +func f[P T1 | T2, _ []P]() {} + +var _ = f[T1] + +// test case from issue + +type BaseT interface { + Type1 | Type2 +} +type BaseType int +type Type1 BaseType +type Type2 BaseType // float64 + +type ValueT[T BaseT] struct { + A1 T +} + +func NewType1() *ValueT[Type1] { + r := NewT[Type1]() + return r +} +func NewType2() *ValueT[Type2] { + r := NewT[Type2]() + return r +} + +func NewT[TBase BaseT, TVal ValueT[TBase]]() *TVal { + ret := TVal{} + return &ret +} +func TestGoType(t *testing.T) { + r1 := NewType1() + r2 := NewType2() + t.Log(r1, r2) + t.Log(reflect.TypeOf(r1), reflect.TypeOf(r2)) + fooT1(r1.A1) + fooT2(r2.A1) +} + +func fooT1(t1 Type1) { + +} +func fooT2(t2 Type2) { + +} diff --git a/src/internal/types/testdata/fixedbugs/issue53692.go b/src/internal/types/testdata/fixedbugs/issue53692.go new file mode 100644 index 0000000..dc1a76c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue53692.go @@ -0,0 +1,15 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Cache[K comparable, V any] interface{} + +type LRU[K comparable, V any] struct{} + +func WithLocking2[K comparable, V any](Cache[K, V]) {} + +func _() { + WithLocking2 /* ERROR "cannot infer V" */ [string](LRU[string, int]{}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue54280.go b/src/internal/types/testdata/fixedbugs/issue54280.go new file mode 100644 index 0000000..9465894 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue54280.go @@ -0,0 +1,7 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +const C = 912_345_678_901_234_567_890_123_456_789_012_345_678_901_234_567_890_912_345_678_901_234_567_890_123_456_789_012_345_678_901_234_567_890_912_345_678_901_234_567_890_123_456_789_012_345_678_901_234_567_890_912 // ERROR "constant overflow" diff --git a/src/internal/types/testdata/fixedbugs/issue54405.go b/src/internal/types/testdata/fixedbugs/issue54405.go new file mode 100644 index 0000000..688fee1 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue54405.go @@ -0,0 +1,16 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Test that we don't see spurious errors for == +// for values with invalid types due to prior errors. + +package p + +var x struct { + f *NotAType /* ERROR "undefined" */ +} +var _ = x.f == nil // no error expected here + +var y *NotAType /* ERROR "undefined" */ +var _ = y == nil // no error expected here diff --git a/src/internal/types/testdata/fixedbugs/issue54424.go b/src/internal/types/testdata/fixedbugs/issue54424.go new file mode 100644 index 0000000..ebfb83d --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue54424.go @@ -0,0 +1,12 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P ~*T, T any]() { + var p P + var tp *T + tp = p // this assignment is valid + _ = tp +} diff --git a/src/internal/types/testdata/fixedbugs/issue54942.go b/src/internal/types/testdata/fixedbugs/issue54942.go new file mode 100644 index 0000000..b9a5cce --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue54942.go @@ -0,0 +1,38 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import ( + "context" + "database/sql" +) + +type I interface { + m(int, int, *int, int) +} + +type T struct{} + +func (_ *T) m(a, b, c, d int) {} + +var _ I = new /* ERROR "have m(int, int, int, int)\n\t\twant m(int, int, *int, int)" */ (T) + +// (slightly modified) test case from issue + +type Result struct { + Value string +} + +type Executor interface { + Execute(context.Context, sql.Stmt, int, []sql.NamedArg, int) (Result, error) +} + +type myExecutor struct{} + +func (_ *myExecutor) Execute(ctx context.Context, stmt sql.Stmt, maxrows int, args []sql.NamedArg, urgency int) (*Result, error) { + return &Result{}, nil +} + +var ex Executor = new /* ERROR "have Execute(context.Context, sql.Stmt, int, []sql.NamedArg, int) (*Result, error)\n\t\twant Execute(context.Context, sql.Stmt, int, []sql.NamedArg, int) (Result, error)" */ (myExecutor) diff --git a/src/internal/types/testdata/fixedbugs/issue56351.go b/src/internal/types/testdata/fixedbugs/issue56351.go new file mode 100644 index 0000000..eee142c --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue56351.go @@ -0,0 +1,11 @@ +// -lang=go1.20 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(s []int) { + clear /* ERROR "clear requires go1.21 or later" */ (s) +} diff --git a/src/internal/types/testdata/fixedbugs/issue56425.go b/src/internal/types/testdata/fixedbugs/issue56425.go new file mode 100644 index 0000000..d85733f --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue56425.go @@ -0,0 +1,8 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +const s float32 = 0 +var _ = 0 << s diff --git a/src/internal/types/testdata/fixedbugs/issue56665.go b/src/internal/types/testdata/fixedbugs/issue56665.go new file mode 100644 index 0000000..1f787d0 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue56665.go @@ -0,0 +1,30 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Example from the issue: +type A[T any] interface { + *T +} + +type B[T any] interface { + B /* ERROR "invalid recursive type" */ [*T] +} + +type C[T any, U B[U]] interface { + *T +} + +// Simplified reproducer: +type X[T any] interface { + X /* ERROR "invalid recursive type" */ [*T] +} + +var _ X[int] + +// A related example that doesn't go through interfaces. +type A2[P any] [10]A2 /* ERROR "invalid recursive type" */ [*P] + +var _ A2[int] diff --git a/src/internal/types/testdata/fixedbugs/issue57155.go b/src/internal/types/testdata/fixedbugs/issue57155.go new file mode 100644 index 0000000..ec9fb2b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57155.go @@ -0,0 +1,14 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P *Q, Q any](p P, q Q) { + func() { + _ = f[P] + f(p, q) + f[P](p, q) + f[P, Q](p, q) + }() +} diff --git a/src/internal/types/testdata/fixedbugs/issue57160.go b/src/internal/types/testdata/fixedbugs/issue57160.go new file mode 100644 index 0000000..446d019 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57160.go @@ -0,0 +1,10 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(x *int) { + _ = 0 < x // ERROR "invalid operation" + _ = x < 0 // ERROR "invalid operation" +} diff --git a/src/internal/types/testdata/fixedbugs/issue57192.go b/src/internal/types/testdata/fixedbugs/issue57192.go new file mode 100644 index 0000000..6c7894a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57192.go @@ -0,0 +1,22 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type I1[T any] interface { + m1(T) +} +type I2[T any] interface { + I1[T] + m2(T) +} + +var V1 I1[int] +var V2 I2[int] + +func g[T any](I1[T]) {} +func _() { + g(V1) + g(V2) +} diff --git a/src/internal/types/testdata/fixedbugs/issue57352.go b/src/internal/types/testdata/fixedbugs/issue57352.go new file mode 100644 index 0000000..2b31700 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57352.go @@ -0,0 +1,21 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type A interface { + a() +} + +type AB interface { + A + b() +} + +type AAB struct { + A + AB +} + +var _ AB = AAB /* ERROR "ambiguous selector AAB.a" */ {} diff --git a/src/internal/types/testdata/fixedbugs/issue57486.go b/src/internal/types/testdata/fixedbugs/issue57486.go new file mode 100644 index 0000000..933eeb4 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57486.go @@ -0,0 +1,29 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type C1 interface { + comparable +} + +type C2 interface { + comparable + [2]any | int +} + +func G1[T C1](t T) { _ = t == t } +func G2[T C2](t T) { _ = t == t } + +func F1[V [2]any](v V) { + _ = G1[V /* ERROR "V does not satisfy comparable" */] + _ = G1[[2]any] + _ = G1[int] +} + +func F2[V [2]any](v V) { + _ = G2[V /* ERROR "V does not satisfy C2" */] + _ = G2[[ /* ERROR "[2]any does not satisfy C2 (C2 mentions [2]any, but [2]any is not in the type set of C2)" */ 2]any] + _ = G2[int] +} diff --git a/src/internal/types/testdata/fixedbugs/issue57500.go b/src/internal/types/testdata/fixedbugs/issue57500.go new file mode 100644 index 0000000..4a90d47 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57500.go @@ -0,0 +1,16 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type C interface { + comparable + [2]any | int +} + +func f[T C]() {} + +func _() { + _ = f[[ /* ERROR "[2]any does not satisfy C (C mentions [2]any, but [2]any is not in the type set of C)" */ 2]any] +} diff --git a/src/internal/types/testdata/fixedbugs/issue57522.go b/src/internal/types/testdata/fixedbugs/issue57522.go new file mode 100644 index 0000000..d83e5b2 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue57522.go @@ -0,0 +1,24 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// A simplified version of the code in the original report. +type S[T any] struct{} +var V = S[any]{} +func (fs *S[T]) M(V.M /* ERROR "V.M is not a type" */) {} + +// Other minimal reproducers. +type S1[T any] V1.M /* ERROR "V1.M is not a type" */ +type V1 = S1[any] + +type S2[T any] struct{} +type V2 = S2[any] +func (fs *S2[T]) M(x V2.M /* ERROR "V2.M is not a type" */ ) {} + +// The following still panics, as the selector is reached from check.expr +// rather than check.typexpr. TODO(rfindley): fix this. +// type X[T any] int +// func (X[T]) M(x [X[int].M]int) {} + diff --git a/src/internal/types/testdata/fixedbugs/issue58611.go b/src/internal/types/testdata/fixedbugs/issue58611.go new file mode 100644 index 0000000..1ff30f7 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue58611.go @@ -0,0 +1,27 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import ( + "sort" + "strings" +) + +func f[int any](x int) { + x = 0 /* ERRORx "cannot use 0.*(as int.*with int declared at|type parameter)" */ +} + +// test case from issue + +type Set[T comparable] map[T]struct{} + +func (s *Set[string]) String() string { + keys := make([]string, 0, len(*s)) + for k := range *s { + keys = append(keys, k) + } + sort.Strings(keys /* ERRORx "cannot use keys.*with string declared at.*|type parameter" */ ) + return strings /* ERROR "cannot use strings.Join" */ .Join(keys /* ERRORx "cannot use keys.*with string declared at.*|type parameter" */ , ",") +} diff --git a/src/internal/types/testdata/fixedbugs/issue58612.go b/src/internal/types/testdata/fixedbugs/issue58612.go new file mode 100644 index 0000000..db6a62d --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue58612.go @@ -0,0 +1,14 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + var x = new(T) + f[x /* ERROR "not a type" */ /* ERROR "use of .(type) outside type switch" */ .(type)]() +} + +type T struct{} + +func f[_ any]() {} diff --git a/src/internal/types/testdata/fixedbugs/issue58671.go b/src/internal/types/testdata/fixedbugs/issue58671.go new file mode 100644 index 0000000..fa964aa --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue58671.go @@ -0,0 +1,20 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func g[P any](...P) P { var x P; return x } + +func _() { + var ( + _ int = g(1, 2) + _ rune = g(1, 'a') + _ float64 = g(1, 'a', 2.3) + _ float64 = g('a', 2.3) + _ complex128 = g(2.3, 'a', 1i) + ) + g(true, 'a' /* ERROR "mismatched types untyped bool and untyped rune (cannot infer P)" */) + g(1, "foo" /* ERROR "mismatched types untyped int and untyped string (cannot infer P)" */) + g(1, 2.3, "bar" /* ERROR "mismatched types untyped float and untyped string (cannot infer P)" */) +} diff --git a/src/internal/types/testdata/fixedbugs/issue58742.go b/src/internal/types/testdata/fixedbugs/issue58742.go new file mode 100644 index 0000000..b649a49 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue58742.go @@ -0,0 +1,18 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() (int, UndefinedType /* ERROR "undefined: UndefinedType" */ , string) { + return 0 // ERROR "not enough return values\n\thave (number)\n\twant (int, unknown type, string)" +} + +func _() (int, UndefinedType /* ERROR "undefined: UndefinedType" */ ) { + return 0, 1, 2 // ERROR "too many return values\n\thave (number, number, number)\n\twant (int, unknown type)" +} + +// test case from issue +func _() UndefinedType /* ERROR "undefined: UndefinedType" */ { + return // ERROR "not enough return values\n\thave ()\n\twant (unknown type)" +} diff --git a/src/internal/types/testdata/fixedbugs/issue59190.go b/src/internal/types/testdata/fixedbugs/issue59190.go new file mode 100644 index 0000000..fd08303 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59190.go @@ -0,0 +1,36 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type E [1 << 30]complex128 +var a [1 << 30]E +var _ = unsafe.Sizeof(a /* ERROR "too large" */ ) + +var s struct { + _ [1 << 30]E + x int +} +var _ = unsafe.Offsetof(s /* ERROR "too large" */ .x) + +// Test case from issue (modified so it also triggers on 32-bit platforms). + +type A [1]int +type S struct { + x A + y [1 << 30]A + z [1 << 30]struct{} +} +type T [1 << 30][1 << 30]S + +func _() { + var a A + var s S + var t T + _ = unsafe.Sizeof(a) + _ = unsafe.Sizeof(s) + _ = unsafe.Sizeof(t /* ERROR "too large" */ ) +} diff --git a/src/internal/types/testdata/fixedbugs/issue59207.go b/src/internal/types/testdata/fixedbugs/issue59207.go new file mode 100644 index 0000000..59b36e2 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59207.go @@ -0,0 +1,12 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "unsafe" + +type E [1 << 32]byte + +var a [1 << 32]E // size of a must not overflow to 0 +var _ = unsafe.Sizeof(a /* ERROR "too large" */ ) diff --git a/src/internal/types/testdata/fixedbugs/issue59209.go b/src/internal/types/testdata/fixedbugs/issue59209.go new file mode 100644 index 0000000..870ae52 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59209.go @@ -0,0 +1,11 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type ( + _ [1 /* ERROR "invalid array length" */ << 100]int + _ [1.0]int + _ [1.1 /* ERROR "must be integer" */ ]int +) diff --git a/src/internal/types/testdata/fixedbugs/issue59338a.go b/src/internal/types/testdata/fixedbugs/issue59338a.go new file mode 100644 index 0000000..34864dc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59338a.go @@ -0,0 +1,21 @@ +// -lang=go1.20 + +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func g[P any](P) {} +func h[P, Q any](P) Q { panic(0) } + +var _ func(int) = g /* ERROR "implicitly instantiated function in assignment requires go1.21 or later" */ +var _ func(int) string = h[ /* ERROR "partially instantiated function in assignment requires go1.21 or later" */ int] + +func f1(func(int)) {} +func f2(int, func(int)) {} + +func _() { + f1(g /* ERROR "implicitly instantiated function as argument requires go1.21 or later" */) + f2(0, g /* ERROR "implicitly instantiated function as argument requires go1.21 or later" */) +} diff --git a/src/internal/types/testdata/fixedbugs/issue59338b.go b/src/internal/types/testdata/fixedbugs/issue59338b.go new file mode 100644 index 0000000..1a5530a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59338b.go @@ -0,0 +1,19 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func g[P any](P) {} +func h[P, Q any](P) Q { panic(0) } + +var _ func(int) = g +var _ func(int) string = h[int] + +func f1(func(int)) {} +func f2(int, func(int)) {} + +func _() { + f1(g) + f2(0, g) +} diff --git a/src/internal/types/testdata/fixedbugs/issue59371.go b/src/internal/types/testdata/fixedbugs/issue59371.go new file mode 100644 index 0000000..d5b4db6 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59371.go @@ -0,0 +1,17 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +var m map[int]int + +func _() { + _, ok /* ERROR "undefined: ok" */ = m[0] // must not crash +} + +func _() { + var ok = undef /* ERROR "undefined: undef" */ + x, ok := m[0] // must not crash + _, _ = x, ok +} diff --git a/src/internal/types/testdata/fixedbugs/issue59639.go b/src/internal/types/testdata/fixedbugs/issue59639.go new file mode 100644 index 0000000..1117668 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59639.go @@ -0,0 +1,11 @@ +// -lang=go1.17 + +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[P /* ERROR "requires go1.18" */ interface{}](P) {} + +var v func(int) = f /* ERROR "requires go1.18" */ diff --git a/src/internal/types/testdata/fixedbugs/issue59740.go b/src/internal/types/testdata/fixedbugs/issue59740.go new file mode 100644 index 0000000..31cd03b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59740.go @@ -0,0 +1,25 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type F[T any] func(func(F[T])) + +func f(F[int]) {} +func g[T any](F[T]) {} + +func _() { + g(f /* ERROR "type func(F[int]) of f does not match F[T] (cannot infer T)" */) // type inference/unification must not panic +} + +// original test case from issue + +type List[T any] func(T, func(T, List[T]) T) T + +func nil[T any](n T, _ List[T]) T { return n } +func cons[T any](h T, t List[T]) List[T] { return func(n T, f func(T, List[T]) T) T { return f(h, t) } } + +func nums[T any](t T) List[T] { + return cons(t, cons(t, nil /* ERROR "type func(n T, _ List[T]) T of nil[T] does not match inferred type List[T] for List[T]" */ [T])) +} diff --git a/src/internal/types/testdata/fixedbugs/issue59848.go b/src/internal/types/testdata/fixedbugs/issue59848.go new file mode 100644 index 0000000..51da747 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59848.go @@ -0,0 +1,10 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T struct{} +type I interface{ M() } +var _ I = T /* ERROR "missing method M" */ {} // must not crash +func (T) m() {} diff --git a/src/internal/types/testdata/fixedbugs/issue59890.go b/src/internal/types/testdata/fixedbugs/issue59890.go new file mode 100644 index 0000000..ed7afd9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59890.go @@ -0,0 +1,17 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { g /* ERROR "cannot infer T" */ () } + +func g[T any]() (_ /* ERROR "cannot use _ as value or type" */, int) { panic(0) } + +// test case from issue + +var _ = append(f /* ERROR "cannot infer T" */ ()()) + +func f[T any]() (_ /* ERROR "cannot use _" */, _ /* ERROR "cannot use _" */, int) { + panic("not implemented") +} diff --git a/src/internal/types/testdata/fixedbugs/issue59953.go b/src/internal/types/testdata/fixedbugs/issue59953.go new file mode 100644 index 0000000..b10ced7 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59953.go @@ -0,0 +1,9 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { f(g) } +func f[P any](P) {} +func g[Q int](Q) {} diff --git a/src/internal/types/testdata/fixedbugs/issue59956.go b/src/internal/types/testdata/fixedbugs/issue59956.go new file mode 100644 index 0000000..646b50e --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59956.go @@ -0,0 +1,47 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1(func(int)) +func f2(func(int), func(string)) +func f3(func(int), func(string), func(float32)) + +func g1[P any](P) {} + +func _() { + f1(g1) + f2(g1, g1) + f3(g1, g1, g1) +} + +// More complex examples + +func g2[P any](P, P) {} +func h3[P any](func(P), func(P), func() P) {} +func h4[P, Q any](func(P), func(P, Q), func() Q, func(P, Q)) {} + +func r1() int { return 0 } + +func _() { + h3(g1, g1, r1) + h4(g1, g2, r1, g2) +} + +// Variadic cases + +func f(func(int)) +func g[P any](P) {} + +func d[P any](...func(P)) {} + +func _() { + d /* ERROR "cannot infer P" */ () + d(f) + d(f, g) + d(f, g, g) + d /* ERROR "cannot infer P" */ (g, g, g) + d(g, g, f) + d(g, f, g, f) +} diff --git a/src/internal/types/testdata/fixedbugs/issue59958.go b/src/internal/types/testdata/fixedbugs/issue59958.go new file mode 100644 index 0000000..4a4b4dc --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue59958.go @@ -0,0 +1,22 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f(func(int) string) {} + +func g2[P, Q any](P) Q { var q Q; return q } +func g3[P, Q, R any](P) R { var r R; return r } + +func _() { + f(g2) + f(g2[int]) + f(g2[int, string]) + + f(g3[int, bool]) + f(g3[int, bool, string]) + + var _ func(int) string = g2 + var _ func(int) string = g2[int] +} diff --git a/src/internal/types/testdata/fixedbugs/issue60346.go b/src/internal/types/testdata/fixedbugs/issue60346.go new file mode 100644 index 0000000..6dc057b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60346.go @@ -0,0 +1,17 @@ +// -lang=go1.20 + +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func F[P any, Q *P](p P) {} + +var _ = F[int] + +func G[R any](func(R)) {} + +func _() { + G(F[int]) +} diff --git a/src/internal/types/testdata/fixedbugs/issue60377.go b/src/internal/types/testdata/fixedbugs/issue60377.go new file mode 100644 index 0000000..b754f89 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60377.go @@ -0,0 +1,88 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// The type parameter P is not used in interface T1. +// T1 is a defined parameterized interface type which +// can be assigned to any other interface with the same +// methods. We cannot infer a type argument in this case +// because any type would do. + +type T1[P any] interface{ m() } + +func g[P any](T1[P]) {} + +func _() { + var x T1[int] + g /* ERROR "cannot infer P" */ (x) + g[int](x) // int is ok for P + g[string](x) // string is also ok for P! +} + +// This is analogous to the above example, +// but uses two interface types of the same structure. + +type T2[P any] interface{ m() } + +func _() { + var x T2[int] + g /* ERROR "cannot infer P" */ (x) + g[int](x) // int is ok for P + g[string](x) // string is also ok for P! +} + +// Analogous to the T2 example but using an unparameterized interface T3. + +type T3 interface{ m() } + +func _() { + var x T3 + g /* ERROR "cannot infer P" */ (x) + g[int](x) // int is ok for P + g[string](x) // string is also ok for P! +} + +// The type parameter P is not used in struct S. +// S is a defined parameterized (non-interface) type which can only +// be assigned to another type S with the same type argument. +// Therefore we can infer a type argument in this case. + +type S[P any] struct{} + +func g4[P any](S[P]) {} + +func _() { + var x S[int] + g4(x) // we can infer int for P + g4[int](x) // int is the correct type argument + g4[string](x /* ERROR "cannot use x (variable of type S[int]) as S[string] value in argument to g4[string]" */) +} + +// This is similar to the first example but here T1 is a component +// of a func type. In this case types must match exactly: P must +// match int. + +func g5[P any](func(T1[P])) {} + +func _() { + var f func(T1[int]) + g5(f) + g5[int](f) + g5[string](f /* ERROR "cannot use f (variable of type func(T1[int])) as func(T1[string]) value in argument to g5[string]" */) +} + +// This example would fail if we were to infer the type argument int for P +// exactly because any type argument would be ok for the first argument. +// Choosing the wrong type would cause the second argument to not match. + +type T[P any] interface{} + +func g6[P any](T[P], P) {} + +func _() { + var x T[int] + g6(x, 1.2) + g6(x, "") +} diff --git a/src/internal/types/testdata/fixedbugs/issue60460.go b/src/internal/types/testdata/fixedbugs/issue60460.go new file mode 100644 index 0000000..a9cb3d9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60460.go @@ -0,0 +1,88 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +// Simplified (representative) test case. + +func _() { + f(R1{}) +} + +func f[T any](R[T]) {} + +type R[T any] interface { + m(R[T]) +} + +type R1 struct{} + +func (R1) m(R[int]) {} + +// Test case from issue. + +func _() { + r := newTestRules() + NewSet(r) + r2 := newTestRules2() + NewSet(r2) +} + +type Set[T any] struct { + rules Rules[T] +} + +func NewSet[T any](rules Rules[T]) Set[T] { + return Set[T]{ + rules: rules, + } +} + +func (s Set[T]) Copy() Set[T] { + return NewSet(s.rules) +} + +type Rules[T any] interface { + Hash(T) int + Equivalent(T, T) bool + SameRules(Rules[T]) bool +} + +type testRules struct{} + +func newTestRules() Rules[int] { + return testRules{} +} + +func (r testRules) Hash(val int) int { + return val % 16 +} + +func (r testRules) Equivalent(val1 int, val2 int) bool { + return val1 == val2 +} + +func (r testRules) SameRules(other Rules[int]) bool { + _, ok := other.(testRules) + return ok +} + +type testRules2 struct{} + +func newTestRules2() Rules[string] { + return testRules2{} +} + +func (r testRules2) Hash(val string) int { + return 16 +} + +func (r testRules2) Equivalent(val1 string, val2 string) bool { + return val1 == val2 +} + +func (r testRules2) SameRules(other Rules[string]) bool { + _, ok := other.(testRules2) + return ok +} diff --git a/src/internal/types/testdata/fixedbugs/issue60500.go b/src/internal/types/testdata/fixedbugs/issue60500.go new file mode 100644 index 0000000..be8ccaf --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60500.go @@ -0,0 +1,11 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + log("This is a test %v" /* ERROR "cannot use \"This is a test %v\" (untyped string constant) as bool value in argument to log" */, "foo") +} + +func log(enabled bool, format string, args ...any) diff --git a/src/internal/types/testdata/fixedbugs/issue60542.go b/src/internal/types/testdata/fixedbugs/issue60542.go new file mode 100644 index 0000000..b536ddb --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60542.go @@ -0,0 +1,12 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func Clip[S ~[]E, E any](s S) S { + return s +} + +var versions func() +var _ = Clip /* ERROR "S (type func()) does not satisfy ~[]E" */ (versions) diff --git a/src/internal/types/testdata/fixedbugs/issue60556.go b/src/internal/types/testdata/fixedbugs/issue60556.go new file mode 100644 index 0000000..77e5034 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60556.go @@ -0,0 +1,19 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type I[T any] interface { + m(I[T]) +} + +type S[T any] struct{} + +func (S[T]) m(I[T]) {} + +func f[T I[E], E any](T) {} + +func _() { + f(S[int]{}) +} diff --git a/src/internal/types/testdata/fixedbugs/issue60562.go b/src/internal/types/testdata/fixedbugs/issue60562.go new file mode 100644 index 0000000..c08bbf3 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60562.go @@ -0,0 +1,61 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type S[T any] struct{} + +func (S[T]) m(T) {} + +func f0[T any](chan S[T]) {} + +func _() { + var x chan interface{ m(int) } + f0(x /* ERROR "type chan interface{m(int)} of x does not match chan S[T] (cannot infer T)" */) +} + +// variants of the theme + +func f1[T any]([]S[T]) {} + +func _() { + var x []interface{ m(int) } + f1(x /* ERROR "type []interface{m(int)} of x does not match []S[T] (cannot infer T)" */) +} + +type I[T any] interface { + m(T) +} + +func f2[T any](func(I[T])) {} + +func _() { + var x func(interface{ m(int) }) + f2(x /* ERROR "type func(interface{m(int)}) of x does not match func(I[T]) (cannot infer T)" */) +} + +func f3[T any](func(I[T])) {} + +func _() { + var x func(I[int]) + f3(x) // but this is correct: I[T] and I[int] can be made identical with T == int +} + +func f4[T any]([10]I[T]) {} + +func _() { + var x [10]interface{ I[int] } + f4(x /* ERROR "type [10]interface{I[int]} of x does not match [10]I[T] (cannot infer T)" */) +} + +func f5[T any](I[T]) {} + +func _() { + var x interface { + m(int) + n() + } + f5(x) + f5[int](x) // ok +} diff --git a/src/internal/types/testdata/fixedbugs/issue60688.go b/src/internal/types/testdata/fixedbugs/issue60688.go new file mode 100644 index 0000000..38d90ee --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60688.go @@ -0,0 +1,16 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type String string + +func g[P any](P, string) {} + +// String and string are not identical and thus must not unify +// (they are element types of the func type and therefore must +// be identical to match). +// The result is an error from type inference, rather than an +// error from an assignment mismatch. +var f func(int, String) = g // ERROR "type func(int, String) of variable in assignment does not match inferred type func(int, string) for func(P, string)" diff --git a/src/internal/types/testdata/fixedbugs/issue60906.go b/src/internal/types/testdata/fixedbugs/issue60906.go new file mode 100644 index 0000000..2744e89 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60906.go @@ -0,0 +1,11 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _() { + var x int + var f func() []int + _ = f /* ERROR "cannot index f" */ [x] +} diff --git a/src/internal/types/testdata/fixedbugs/issue60933.go b/src/internal/types/testdata/fixedbugs/issue60933.go new file mode 100644 index 0000000..9b10237 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60933.go @@ -0,0 +1,67 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import ( + "io" + "os" +) + +func g[T any](...T) {} + +// Interface and non-interface types do not match. +func _() { + var file *os.File + g(file, io /* ERROR "type io.Writer of io.Discard does not match inferred type *os.File for T" */ .Discard) + g(file, os.Stdout) +} + +func _() { + var a *os.File + var b any + g(a, a) + g(a, b /* ERROR "type any of b does not match inferred type *os.File for T" */) +} + +var writer interface { + Write(p []byte) (n int, err error) +} + +func _() { + var file *os.File + g(file, writer /* ERROR "type interface{Write(p []byte) (n int, err error)} of writer does not match inferred type *os.File for T" */) + g(writer, file /* ERROR "type *os.File of file does not match inferred type interface{Write(p []byte) (n int, err error)} for T" */) +} + +// Different named interface types do not match. +func _() { + g(io.ReadWriter(nil), io.ReadWriter(nil)) + g(io.ReadWriter(nil), io /* ERROR "does not match" */ .Writer(nil)) + g(io.Writer(nil), io /* ERROR "does not match" */ .ReadWriter(nil)) +} + +// Named and unnamed interface types match if they have the same methods. +func _() { + g(io.Writer(nil), writer) + g(io.ReadWriter(nil), writer /* ERROR "does not match" */ ) +} + +// There must be no order dependency for named and unnamed interfaces. +func f[T interface{ m(T) }](a, b T) {} + +type F interface { + m(F) +} + +func _() { + var i F + var j interface { + m(F) + } + + // order doesn't matter + f(i, j) + f(j, i) +}
\ No newline at end of file diff --git a/src/internal/types/testdata/fixedbugs/issue60946.go b/src/internal/types/testdata/fixedbugs/issue60946.go new file mode 100644 index 0000000..a66254b --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue60946.go @@ -0,0 +1,38 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type Tn interface{ m() } +type T1 struct{} +type T2 struct{} + +func (*T1) m() {} +func (*T2) m() {} + +func g[P any](...P) {} + +func _() { + var t interface{ m() } + var tn Tn + var t1 *T1 + var t2 *T2 + + // these are ok (interface types only) + g(t, t) + g(t, tn) + g(tn, t) + g(tn, tn) + + // these are not ok (interface and non-interface types) + g(t, t1 /* ERROR "does not match" */) + g(t1, t /* ERROR "does not match" */) + g(tn, t1 /* ERROR "does not match" */) + g(t1, tn /* ERROR "does not match" */) + + g(t, t1 /* ERROR "does not match" */, t2) + g(t1, t2 /* ERROR "does not match" */, t) + g(tn, t1 /* ERROR "does not match" */, t2) + g(t1, t2 /* ERROR "does not match" */, tn) +} diff --git a/src/internal/types/testdata/fixedbugs/issue61486.go b/src/internal/types/testdata/fixedbugs/issue61486.go new file mode 100644 index 0000000..b12a800 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue61486.go @@ -0,0 +1,9 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func _(s uint) { + _ = min(1 << s) +} diff --git a/src/internal/types/testdata/fixedbugs/issue61879.go b/src/internal/types/testdata/fixedbugs/issue61879.go new file mode 100644 index 0000000..542bc2d --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue61879.go @@ -0,0 +1,57 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "fmt" + +type Interface[T any] interface { + m(Interface[T]) +} + +func f[S []Interface[T], T any](S) {} + +func _() { + var s []Interface[int] + f(s) // panic here +} + +// Larger example from issue + +type InterfaceA[T comparable] interface { + setData(string) InterfaceA[T] +} + +type ImplA[T comparable] struct { + data string + args []any +} + +func NewInterfaceA[T comparable](args ...any) InterfaceA[T] { + return &ImplA[T]{ + data: fmt.Sprintf("%v", args...), + args: args, + } +} + +func (k *ImplA[T]) setData(data string) InterfaceA[T] { + k.data = data + return k +} + +func Foo[M ~map[InterfaceA[T]]V, T comparable, V any](m M) { + // DO SOMETHING HERE + return +} + +func Bar() { + keys := make([]InterfaceA[int], 0, 10) + m := make(map[InterfaceA[int]]int) + for i := 0; i < 10; i++ { + keys = append(keys, NewInterfaceA[int](i)) + m[keys[i]] = i + } + + Foo(m) // panic here +} diff --git a/src/internal/types/testdata/fixedbugs/issue61903.go b/src/internal/types/testdata/fixedbugs/issue61903.go new file mode 100644 index 0000000..8a6fcd9 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue61903.go @@ -0,0 +1,20 @@ +// -lang=go1.20 + +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +type T[P any] interface{} + +func f1[P any](T[P]) {} +func f2[P any](T[P], P) {} + +func _() { + var t T[int] + f1(t) + + var s string + f2(t, s /* ERROR "type string of s does not match inferred type int for P" */) +} diff --git a/src/internal/types/testdata/fixedbugs/issue62157.go b/src/internal/types/testdata/fixedbugs/issue62157.go new file mode 100644 index 0000000..c44f921 --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue62157.go @@ -0,0 +1,128 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f[T any](...T) T { var x T; return x } + +// Test case 1 + +func _() { + var a chan string + var b <-chan string + f(a, b) + f(b, a) +} + +// Test case 2 + +type F[T any] func(T) bool + +func g[T any](T) F[<-chan T] { return nil } + +func f1[T any](T, F[T]) {} +func f2[T any](F[T], T) {} + +func _() { + var ch chan string + f1(ch, g("")) + f2(g(""), ch) +} + +// Test case 3: named and directional types combined + +func _() { + type namedA chan int + type namedB chan<- int + + var a chan int + var A namedA + var b chan<- int + var B namedB + + // Defined types win over channel types irrespective of channel direction. + f(A, b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */) + f(b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */, A) + + f(a, b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */, A) + f(a, A, b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */) + f(b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */, A, a) + f(b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */, a, A) + f(A, a, b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */) + f(A, b /* ERROR "cannot use b (variable of type chan<- int) as namedA value in argument to f" */, a) + + // Unnamed directed channels win over bidirectional channels. + b = f(a, b) + b = f(b, a) + + // Defined directed channels win over defined bidirectional channels. + A = f(A, a) + A = f(a, A) + B = f(B, b) + B = f(b, B) + + f(a, b, B) + f(a, B, b) + f(b, B, a) + f(b, a, B) + f(B, a, b) + f(B, b, a) + + // Differently named channel types conflict irrespective of channel direction. + f(A, B /* ERROR "type namedB of B does not match inferred type namedA for T" */) + f(B, A /* ERROR "type namedA of A does not match inferred type namedB for T" */) + + // Ensure that all combinations of directional and + // bidirectional channels with a named directional + // channel lead to the correct (named) directional + // channel. + B = f(a, b) + B = f(a, B) + B = f(b, a) + B = f(B, a) + + B = f(a, b, B) + B = f(a, B, b) + B = f(b, B, a) + B = f(b, a, B) + B = f(B, a, b) + B = f(B, b, a) + + // verify type error + A = f /* ERROR "cannot use f(B, b, a) (value of type namedB) as namedA value in assignment" */ (B, b, a) +} + +// Test case 4: some more combinations + +func _() { + type A chan int + type B chan int + type C = chan int + type D = chan<- int + + var a A + var b B + var c C + var d D + + f(a, b /* ERROR "type B of b does not match inferred type A for T" */, c) + f(c, a, b /* ERROR "type B of b does not match inferred type A for T" */) + f(a, b /* ERROR "type B of b does not match inferred type A for T" */, d) + f(d, a, b /* ERROR "type B of b does not match inferred type A for T" */) +} + +// Simplified test case from issue + +type Matcher[T any] func(T) bool + +func Produces[T any](T) Matcher[<-chan T] { return nil } + +func Assert1[T any](Matcher[T], T) {} +func Assert2[T any](T, Matcher[T]) {} + +func _() { + var ch chan string + Assert1(Produces(""), ch) + Assert2(ch, Produces("")) +} diff --git a/src/internal/types/testdata/fixedbugs/issue6977.go b/src/internal/types/testdata/fixedbugs/issue6977.go new file mode 100644 index 0000000..c455d3a --- /dev/null +++ b/src/internal/types/testdata/fixedbugs/issue6977.go @@ -0,0 +1,82 @@ +// Copyright 2019 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +import "io" + +// Alan's initial report. + +type I interface { f(); String() string } +type J interface { g(); String() string } + +type IJ1 = interface { I; J } +type IJ2 = interface { f(); g(); String() string } + +var _ = (*IJ1)(nil) == (*IJ2)(nil) // static assert that IJ1 and IJ2 are identical types + +// The canonical example. + +type ReadWriteCloser interface { io.ReadCloser; io.WriteCloser } + +// Some more cases. + +type M interface { m() } +type M32 interface { m() int32 } +type M64 interface { m() int64 } + +type U1 interface { m() } +type U2 interface { m(); M } +type U3 interface { M; m() } +type U4 interface { M; M; M } +type U5 interface { U1; U2; U3; U4 } + +type U6 interface { m(); m /* ERROR "duplicate method" */ () } +type U7 interface { M32 /* ERROR "duplicate method" */ ; m() } +type U8 interface { m(); M32 /* ERROR "duplicate method" */ } +type U9 interface { M32; M64 /* ERROR "duplicate method" */ } + +// Verify that repeated embedding of the same interface(s) +// eliminates duplicate methods early (rather than at the +// end) to prevent exponential memory and time use. +// Without early elimination, computing T29 may take dozens +// of minutes. +type ( + T0 interface { m() } + T1 interface { T0; T0 } + T2 interface { T1; T1 } + T3 interface { T2; T2 } + T4 interface { T3; T3 } + T5 interface { T4; T4 } + T6 interface { T5; T5 } + T7 interface { T6; T6 } + T8 interface { T7; T7 } + T9 interface { T8; T8 } + + T10 interface { T9; T9 } + T11 interface { T10; T10 } + T12 interface { T11; T11 } + T13 interface { T12; T12 } + T14 interface { T13; T13 } + T15 interface { T14; T14 } + T16 interface { T15; T15 } + T17 interface { T16; T16 } + T18 interface { T17; T17 } + T19 interface { T18; T18 } + + T20 interface { T19; T19 } + T21 interface { T20; T20 } + T22 interface { T21; T21 } + T23 interface { T22; T22 } + T24 interface { T23; T23 } + T25 interface { T24; T24 } + T26 interface { T25; T25 } + T27 interface { T26; T26 } + T28 interface { T27; T27 } + T29 interface { T28; T28 } +) + +// Verify that m is present. +var x T29 +var _ = x.m diff --git a/src/internal/types/testdata/spec/assignability.go b/src/internal/types/testdata/spec/assignability.go new file mode 100644 index 0000000..6670870 --- /dev/null +++ b/src/internal/types/testdata/spec/assignability.go @@ -0,0 +1,264 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package assignability + +// See the end of this package for the declarations +// of the types and variables used in these tests. + +// "x's type is identical to T" +func _[TP any](X TP) { + b = b + a = a + l = l + s = s + p = p + f = f + i = i + m = m + c = c + d = d + + B = B + A = A + L = L + S = S + P = P + F = F + I = I + M = M + C = C + D = D + X = X +} + +// "x's type V and T have identical underlying types +// and at least one of V or T is not a named type." +// (here a named type is a type with a name) +func _[TP1, TP2 Interface](X1 TP1, X2 TP2) { + b = B // ERRORx `cannot use B .* as int value` + a = A + l = L + s = S + p = P + f = F + i = I + m = M + c = C + d = D + + B = b // ERRORx `cannot use b .* as Basic value` + A = a + L = l + S = s + P = p + F = f + I = i + M = m + C = c + D = d + X1 = i // ERRORx `cannot use i .* as TP1 value` + X1 = X2 // ERRORx `cannot use X2 .* as TP1 value` +} + +// "T is an interface type and x implements T and T is not a type parameter" +func _[TP Interface](X TP) { + i = d // ERROR "missing method m" + i = D + i = X + X = i // ERRORx `cannot use i .* as TP value` +} + +// "x is a bidirectional channel value, T is a channel type, x's type V and T have identical element types, and at least one of V or T is not a named type" +// (here a named type is a type with a name) +type ( + _SendChan = chan<- int + _RecvChan = <-chan int + + SendChan _SendChan + RecvChan _RecvChan +) + +func _[ + _CC ~_Chan, + _SC ~_SendChan, + _RC ~_RecvChan, + + CC Chan, + SC SendChan, + RC RecvChan, +]() { + var ( + _ _SendChan = c + _ _RecvChan = c + _ _Chan = c + + _ _SendChan = C + _ _RecvChan = C + _ _Chan = C + + _ SendChan = c + _ RecvChan = c + _ Chan = c + + _ SendChan = C // ERRORx `cannot use C .* as SendChan value` + _ RecvChan = C // ERRORx `cannot use C .* as RecvChan value` + _ Chan = C + _ Chan = make /* ERRORx `cannot use make\(chan Basic\) .* as Chan value` */ (chan Basic) + ) + + var ( + _ _CC = C // ERRORx `cannot use C .* as _CC value` + _ _SC = C // ERRORx `cannot use C .* as _SC value` + _ _RC = C // ERRORx `cannot use C .* as _RC value` + + _ CC = _CC /* ERRORx `cannot use _CC\(nil\) .* as CC value` */ (nil) + _ SC = _CC /* ERRORx `cannot use _CC\(nil\) .* as SC value` */ (nil) + _ RC = _CC /* ERRORx `cannot use _CC\(nil\) .* as RC value` */ (nil) + + _ CC = C // ERRORx `cannot use C .* as CC value` + _ SC = C // ERRORx `cannot use C .* as SC value` + _ RC = C // ERRORx `cannot use C .* as RC value` + ) +} + +// "x's type V is not a named type and T is a type parameter, and x is assignable to each specific type in T's type set." +func _[ + TP0 any, + TP1 ~_Chan, + TP2 ~chan int | ~chan byte, +]() { + var ( + _ TP0 = c // ERRORx `cannot use c .* as TP0 value` + _ TP0 = C // ERRORx `cannot use C .* as TP0 value` + _ TP1 = c + _ TP1 = C // ERRORx `cannot use C .* as TP1 value` + _ TP2 = c // ERRORx `.* cannot assign chan int to chan byte` + ) +} + +// "x's type V is a type parameter and T is not a named type, and values x' of each specific type in V's type set are assignable to T." +func _[ + TP0 Interface, + TP1 ~_Chan, + TP2 ~chan int | ~chan byte, +](X0 TP0, X1 TP1, X2 TP2) { + i = X0 + I = X0 + c = X1 + C = X1 // ERRORx `cannot use X1 .* as Chan value` + c = X2 // ERRORx `.* cannot assign chan byte \(in TP2\) to chan int` +} + +// "x is the predeclared identifier nil and T is a pointer, function, slice, map, channel, or interface type" +func _[TP Interface](X TP) { + b = nil // ERROR "cannot use nil" + a = nil // ERROR "cannot use nil" + l = nil + s = nil // ERROR "cannot use nil" + p = nil + f = nil + i = nil + m = nil + c = nil + d = nil // ERROR "cannot use nil" + + B = nil // ERROR "cannot use nil" + A = nil // ERROR "cannot use nil" + L = nil + S = nil // ERROR "cannot use nil" + P = nil + F = nil + I = nil + M = nil + C = nil + D = nil // ERROR "cannot use nil" + X = nil // ERROR "cannot use nil" +} + +// "x is an untyped constant representable by a value of type T" +func _[ + Int8 ~int8, + Int16 ~int16, + Int32 ~int32, + Int64 ~int64, + Int8_16 ~int8 | ~int16, +]( + i8 Int8, + i16 Int16, + i32 Int32, + i64 Int64, + i8_16 Int8_16, +) { + b = 42 + b = 42.0 + // etc. + + i8 = -1 << 7 + i8 = 1<<7 - 1 + i16 = -1 << 15 + i16 = 1<<15 - 1 + i32 = -1 << 31 + i32 = 1<<31 - 1 + i64 = -1 << 63 + i64 = 1<<63 - 1 + + i8_16 = -1 << 7 + i8_16 = 1<<7 - 1 + i8_16 = - /* ERRORx `cannot use .* as Int8_16` */ 1 << 15 + i8_16 = 1 /* ERRORx `cannot use .* as Int8_16` */ <<15 - 1 +} + +// proto-types for tests + +type ( + _Basic = int + _Array = [10]int + _Slice = []int + _Struct = struct{ f int } + _Pointer = *int + _Func = func(x int) string + _Interface = interface{ m() int } + _Map = map[string]int + _Chan = chan int + + Basic _Basic + Array _Array + Slice _Slice + Struct _Struct + Pointer _Pointer + Func _Func + Interface _Interface + Map _Map + Chan _Chan + Defined _Struct +) + +func (Defined) m() int + +// proto-variables for tests + +var ( + b _Basic + a _Array + l _Slice + s _Struct + p _Pointer + f _Func + i _Interface + m _Map + c _Chan + d _Struct + + B Basic + A Array + L Slice + S Struct + P Pointer + F Func + I Interface + M Map + C Chan + D Defined +) diff --git a/src/internal/types/testdata/spec/comparable.go b/src/internal/types/testdata/spec/comparable.go new file mode 100644 index 0000000..211fa11 --- /dev/null +++ b/src/internal/types/testdata/spec/comparable.go @@ -0,0 +1,26 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[_ comparable]() {} +func f2[_ interface{ comparable }]() {} + +type T interface{ m() } + +func _[P comparable, Q ~int, R any]() { + _ = f1[int] + _ = f1[T /* T does satisfy comparable */] + _ = f1[any /* any does satisfy comparable */] + _ = f1[P] + _ = f1[Q] + _ = f1[R /* ERROR "R does not satisfy comparable" */] + + _ = f2[int] + _ = f2[T /* T does satisfy comparable */] + _ = f2[any /* any does satisfy comparable */] + _ = f2[P] + _ = f2[Q] + _ = f2[R /* ERROR "R does not satisfy comparable" */] +} diff --git a/src/internal/types/testdata/spec/comparable1.19.go b/src/internal/types/testdata/spec/comparable1.19.go new file mode 100644 index 0000000..7a4b2a0 --- /dev/null +++ b/src/internal/types/testdata/spec/comparable1.19.go @@ -0,0 +1,28 @@ +// -lang=go1.19 + +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package p + +func f1[_ comparable]() {} +func f2[_ interface{ comparable }]() {} + +type T interface{ m() } + +func _[P comparable, Q ~int, R any]() { + _ = f1[int] + _ = f1[T /* ERROR "T to satisfy comparable requires go1.20 or later" */] + _ = f1[any /* ERROR "any to satisfy comparable requires go1.20 or later" */] + _ = f1[P] + _ = f1[Q] + _ = f1[R /* ERROR "R does not satisfy comparable" */] + + _ = f2[int] + _ = f2[T /* ERROR "T to satisfy comparable requires go1.20 or later" */] + _ = f2[any /* ERROR "any to satisfy comparable requires go1.20 or later" */] + _ = f2[P] + _ = f2[Q] + _ = f2[R /* ERROR "R does not satisfy comparable" */] +} diff --git a/src/internal/types/testdata/spec/comparisons.go b/src/internal/types/testdata/spec/comparisons.go new file mode 100644 index 0000000..492890e --- /dev/null +++ b/src/internal/types/testdata/spec/comparisons.go @@ -0,0 +1,120 @@ +// Copyright 2022 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package comparisons + +type ( + B int // basic type representative + A [10]func() + L []byte + S struct{ f []byte } + P *S + F func() + I interface{} + M map[string]int + C chan int +) + +var ( + b B + a A + l L + s S + p P + f F + i I + m M + c C +) + +func _() { + _ = nil == nil // ERROR "operator == not defined on untyped nil" + _ = b == b + _ = a /* ERROR "[10]func() cannot be compared" */ == a + _ = l /* ERROR "slice can only be compared to nil" */ == l + _ = s /* ERROR "struct containing []byte cannot be compared" */ == s + _ = p == p + _ = f /* ERROR "func can only be compared to nil" */ == f + _ = i == i + _ = m /* ERROR "map can only be compared to nil" */ == m + _ = c == c + + _ = b == nil /* ERROR "mismatched types" */ + _ = a == nil /* ERROR "mismatched types" */ + _ = l == nil + _ = s == nil /* ERROR "mismatched types" */ + _ = p == nil + _ = f == nil + _ = i == nil + _ = m == nil + _ = c == nil + + _ = nil /* ERROR "operator < not defined on untyped nil" */ < nil + _ = b < b + _ = a /* ERROR "operator < not defined on array" */ < a + _ = l /* ERROR "operator < not defined on slice" */ < l + _ = s /* ERROR "operator < not defined on struct" */ < s + _ = p /* ERROR "operator < not defined on pointer" */ < p + _ = f /* ERROR "operator < not defined on func" */ < f + _ = i /* ERROR "operator < not defined on interface" */ < i + _ = m /* ERROR "operator < not defined on map" */ < m + _ = c /* ERROR "operator < not defined on chan" */ < c +} + +func _[ + B int, + A [10]func(), + L []byte, + S struct{ f []byte }, + P *S, + F func(), + I interface{}, + J comparable, + M map[string]int, + C chan int, +]( + b B, + a A, + l L, + s S, + p P, + f F, + i I, + j J, + m M, + c C, +) { + _ = b == b + _ = a /* ERROR "incomparable types in type set" */ == a + _ = l /* ERROR "incomparable types in type set" */ == l + _ = s /* ERROR "incomparable types in type set" */ == s + _ = p == p + _ = f /* ERROR "incomparable types in type set" */ == f + _ = i /* ERROR "incomparable types in type set" */ == i + _ = j == j + _ = m /* ERROR "incomparable types in type set" */ == m + _ = c == c + + _ = b == nil /* ERROR "mismatched types" */ + _ = a == nil /* ERROR "mismatched types" */ + _ = l == nil + _ = s == nil /* ERROR "mismatched types" */ + _ = p == nil + _ = f == nil + _ = i == nil /* ERROR "mismatched types" */ + _ = j == nil /* ERROR "mismatched types" */ + _ = m == nil + _ = c == nil + + _ = b < b + _ = a /* ERROR "type parameter A is not comparable with <" */ < a + _ = l /* ERROR "type parameter L is not comparable with <" */ < l + _ = s /* ERROR "type parameter S is not comparable with <" */ < s + _ = p /* ERROR "type parameter P is not comparable with <" */ < p + _ = f /* ERROR "type parameter F is not comparable with <" */ < f + _ = i /* ERROR "type parameter I is not comparable with <" */ < i + _ = j /* ERROR "type parameter J is not comparable with <" */ < j + _ = m /* ERROR "type parameter M is not comparable with <" */ < m + _ = c /* ERROR "type parameter C is not comparable with <" */ < c +} diff --git a/src/internal/types/testdata/spec/conversions.go b/src/internal/types/testdata/spec/conversions.go new file mode 100644 index 0000000..081439e --- /dev/null +++ b/src/internal/types/testdata/spec/conversions.go @@ -0,0 +1,208 @@ +// Copyright 2021 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package conversions + +import "unsafe" + +// constant conversions + +func _[T ~byte]() T { return 255 } +func _[T ~byte]() T { return 256 /* ERRORx `cannot use 256 .* as T value` */ } + +func _[T ~byte]() { + const _ = T /* ERRORx `T\(0\) .* is not constant` */ (0) + var _ T = 255 + var _ T = 256 // ERRORx `cannot use 256 .* as T value` +} + +func _[T ~string]() T { return T('a') } +func _[T ~int | ~string]() T { return T('a') } +func _[T ~byte | ~int | ~string]() T { return T(256 /* ERRORx `cannot convert 256 .* to type T` */) } + +// implicit conversions never convert to string +func _[T ~string]() { + var _ string = 0 // ERRORx `cannot use .* as string value` + var _ T = 0 // ERRORx `cannot use .* as T value` +} + +// failing const conversions of constants to type parameters report a cause +func _[ + T1 any, + T2 interface{ m() }, + T3 ~int | ~float64 | ~bool, + T4 ~int | ~string, +]() { + _ = T1(0 /* ERRORx `cannot convert 0 .* to type T1: T1 does not contain specific types` */) + _ = T2(1 /* ERRORx `cannot convert 1 .* to type T2: T2 does not contain specific types` */) + _ = T3(2 /* ERRORx `cannot convert 2 .* to type T3: cannot convert 2 .* to type bool \(in T3\)` */) + _ = T4(3.14 /* ERRORx `cannot convert 3.14 .* to type T4: cannot convert 3.14 .* to type int \(in T4\)` */) +} + +// "x is assignable to T" +// - tested via assignability tests + +// "x's type and T have identical underlying types if tags are ignored" + +func _[X ~int, T ~int](x X) T { return T(x) } +func _[X struct { + f int "foo" +}, T struct { + f int "bar" +}](x X) T { + return T(x) +} + +type Foo struct { + f int "foo" +} +type Bar struct { + f int "bar" +} +type Far struct{ f float64 } + +func _[X Foo, T Bar](x X) T { return T(x) } +func _[X Foo | Bar, T Bar](x X) T { return T(x) } +func _[X Foo, T Foo | Bar](x X) T { return T(x) } +func _[X Foo, T Far](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by Foo) to type T: cannot convert Foo (in X) to type Far (in T)" */) +} + +// "x's type and T are unnamed pointer types and their pointer base types +// have identical underlying types if tags are ignored" + +func _[X ~*Foo, T ~*Bar](x X) T { return T(x) } +func _[X ~*Foo | ~*Bar, T ~*Bar](x X) T { return T(x) } +func _[X ~*Foo, T ~*Foo | ~*Bar](x X) T { return T(x) } +func _[X ~*Foo, T ~*Far](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by ~*Foo) to type T: cannot convert *Foo (in X) to type *Far (in T)" */) +} + +// Verify that the defined types in constraints are considered for the rule above. + +type ( + B int + C int + X0 *B + T0 *C +) + +func _(x X0) T0 { return T0(x /* ERROR "cannot convert" */) } // non-generic reference +func _[X X0, T T0](x X) T { return T(x /* ERROR "cannot convert" */) } +func _[T T0](x X0) T { return T(x /* ERROR "cannot convert" */) } +func _[X X0](x X) T0 { return T0(x /* ERROR "cannot convert" */) } + +// "x's type and T are both integer or floating point types" + +func _[X Integer, T Integer](x X) T { return T(x) } +func _[X Unsigned, T Integer](x X) T { return T(x) } +func _[X Float, T Integer](x X) T { return T(x) } + +func _[X Integer, T Unsigned](x X) T { return T(x) } +func _[X Unsigned, T Unsigned](x X) T { return T(x) } +func _[X Float, T Unsigned](x X) T { return T(x) } + +func _[X Integer, T Float](x X) T { return T(x) } +func _[X Unsigned, T Float](x X) T { return T(x) } +func _[X Float, T Float](x X) T { return T(x) } + +func _[X, T Integer | Unsigned | Float](x X) T { return T(x) } +func _[X, T Integer | ~string](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by Integer | ~string) to type T: cannot convert string (in X) to type int (in T)" */) +} + +// "x's type and T are both complex types" + +func _[X, T Complex](x X) T { return T(x) } +func _[X, T Float | Complex](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by Float | Complex) to type T: cannot convert float32 (in X) to type complex64 (in T)" */) +} + +// "x is an integer or a slice of bytes or runes and T is a string type" + +type myInt int +type myString string + +func _[T ~string](x int) T { return T(x) } +func _[T ~string](x myInt) T { return T(x) } +func _[X Integer](x X) string { return string(x) } +func _[X Integer](x X) myString { return myString(x) } +func _[X Integer](x X) *string { + return (*string)(x /* ERROR "cannot convert x (variable of type X constrained by Integer) to type *string: cannot convert int (in X) to type *string" */) +} + +func _[T ~string](x []byte) T { return T(x) } +func _[T ~string](x []rune) T { return T(x) } +func _[X ~[]byte, T ~string](x X) T { return T(x) } +func _[X ~[]rune, T ~string](x X) T { return T(x) } +func _[X Integer | ~[]byte | ~[]rune, T ~string](x X) T { return T(x) } +func _[X Integer | ~[]byte | ~[]rune, T ~*string](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by Integer | ~[]byte | ~[]rune) to type T: cannot convert int (in X) to type *string (in T)" */) +} + +// "x is a string and T is a slice of bytes or runes" + +func _[T ~[]byte](x string) T { return T(x) } +func _[T ~[]rune](x string) T { return T(x) } +func _[T ~[]rune](x *string) T { + return T(x /* ERROR "cannot convert x (variable of type *string) to type T: cannot convert *string to type []rune (in T)" */) +} + +func _[X ~string, T ~[]byte](x X) T { return T(x) } +func _[X ~string, T ~[]rune](x X) T { return T(x) } +func _[X ~string, T ~[]byte | ~[]rune](x X) T { return T(x) } +func _[X ~*string, T ~[]byte | ~[]rune](x X) T { + return T(x /* ERROR "cannot convert x (variable of type X constrained by ~*string) to type T: cannot convert *string (in X) to type []byte (in T)" */) +} + +// package unsafe: +// "any pointer or value of underlying type uintptr can be converted into a unsafe.Pointer" + +type myUintptr uintptr + +func _[X ~uintptr](x X) unsafe.Pointer { return unsafe.Pointer(x) } +func _[T unsafe.Pointer](x myUintptr) T { return T(x) } +func _[T unsafe.Pointer](x int64) T { + return T(x /* ERROR "cannot convert x (variable of type int64) to type T: cannot convert int64 to type unsafe.Pointer (in T)" */) +} + +// "and vice versa" + +func _[T ~uintptr](x unsafe.Pointer) T { return T(x) } +func _[X unsafe.Pointer](x X) uintptr { return uintptr(x) } +func _[X unsafe.Pointer](x X) myUintptr { return myUintptr(x) } +func _[X unsafe.Pointer](x X) int64 { + return int64(x /* ERROR "cannot convert x (variable of type X constrained by unsafe.Pointer) to type int64: cannot convert unsafe.Pointer (in X) to type int64" */) +} + +// "x is a slice, T is an array or pointer-to-array type, +// and the slice and array types have identical element types." + +func _[X ~[]E, T ~[10]E, E any](x X) T { return T(x) } +func _[X ~[]E, T ~*[10]E, E any](x X) T { return T(x) } + +// ---------------------------------------------------------------------------- +// The following declarations can be replaced by the exported types of the +// constraints package once all builders support importing interfaces with +// type constraints. + +type Signed interface { + ~int | ~int8 | ~int16 | ~int32 | ~int64 +} + +type Unsigned interface { + ~uint | ~uint8 | ~uint16 | ~uint32 | ~uint64 | ~uintptr +} + +type Integer interface { + Signed | Unsigned +} + +type Float interface { + ~float32 | ~float64 +} + +type Complex interface { + ~complex64 | ~complex128 +} diff --git a/src/internal/unsafeheader/unsafeheader.go b/src/internal/unsafeheader/unsafeheader.go new file mode 100644 index 0000000..6d092c6 --- /dev/null +++ b/src/internal/unsafeheader/unsafeheader.go @@ -0,0 +1,37 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package unsafeheader contains header declarations for the Go runtime's slice +// and string implementations. +// +// This package allows packages that cannot import "reflect" to use types that +// are tested to be equivalent to reflect.SliceHeader and reflect.StringHeader. +package unsafeheader + +import ( + "unsafe" +) + +// Slice is the runtime representation of a slice. +// It cannot be used safely or portably and its representation may +// change in a later release. +// +// Unlike reflect.SliceHeader, its Data field is sufficient to guarantee the +// data it references will not be garbage collected. +type Slice struct { + Data unsafe.Pointer + Len int + Cap int +} + +// String is the runtime representation of a string. +// It cannot be used safely or portably and its representation may +// change in a later release. +// +// Unlike reflect.StringHeader, its Data field is sufficient to guarantee the +// data it references will not be garbage collected. +type String struct { + Data unsafe.Pointer + Len int +} diff --git a/src/internal/unsafeheader/unsafeheader_test.go b/src/internal/unsafeheader/unsafeheader_test.go new file mode 100644 index 0000000..f3d1a9b --- /dev/null +++ b/src/internal/unsafeheader/unsafeheader_test.go @@ -0,0 +1,100 @@ +// Copyright 2020 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package unsafeheader_test + +import ( + "bytes" + "internal/unsafeheader" + "reflect" + "testing" + "unsafe" +) + +// TestTypeMatchesReflectType ensures that the name and layout of the +// unsafeheader types matches the corresponding Header types in the reflect +// package. +func TestTypeMatchesReflectType(t *testing.T) { + t.Run("Slice", func(t *testing.T) { + testHeaderMatchesReflect(t, unsafeheader.Slice{}, reflect.SliceHeader{}) + }) + + t.Run("String", func(t *testing.T) { + testHeaderMatchesReflect(t, unsafeheader.String{}, reflect.StringHeader{}) + }) +} + +func testHeaderMatchesReflect(t *testing.T, header, reflectHeader any) { + h := reflect.TypeOf(header) + rh := reflect.TypeOf(reflectHeader) + + for i := 0; i < h.NumField(); i++ { + f := h.Field(i) + rf, ok := rh.FieldByName(f.Name) + if !ok { + t.Errorf("Field %d of %v is named %s, but no such field exists in %v", i, h, f.Name, rh) + continue + } + if !typeCompatible(f.Type, rf.Type) { + t.Errorf("%v.%s has type %v, but %v.%s has type %v", h, f.Name, f.Type, rh, rf.Name, rf.Type) + } + if f.Offset != rf.Offset { + t.Errorf("%v.%s has offset %d, but %v.%s has offset %d", h, f.Name, f.Offset, rh, rf.Name, rf.Offset) + } + } + + if h.NumField() != rh.NumField() { + t.Errorf("%v has %d fields, but %v has %d", h, h.NumField(), rh, rh.NumField()) + } + if h.Align() != rh.Align() { + t.Errorf("%v has alignment %d, but %v has alignment %d", h, h.Align(), rh, rh.Align()) + } +} + +var ( + unsafePointerType = reflect.TypeOf(unsafe.Pointer(nil)) + uintptrType = reflect.TypeOf(uintptr(0)) +) + +func typeCompatible(t, rt reflect.Type) bool { + return t == rt || (t == unsafePointerType && rt == uintptrType) +} + +// TestWriteThroughHeader ensures that the headers in the unsafeheader package +// can successfully mutate variables of the corresponding built-in types. +// +// This test is expected to fail under -race (which implicitly enables +// -d=checkptr) if the runtime views the header types as incompatible with the +// underlying built-in types. +func TestWriteThroughHeader(t *testing.T) { + t.Run("Slice", func(t *testing.T) { + s := []byte("Hello, checkptr!")[:5] + + var alias []byte + hdr := (*unsafeheader.Slice)(unsafe.Pointer(&alias)) + hdr.Data = unsafe.Pointer(&s[0]) + hdr.Cap = cap(s) + hdr.Len = len(s) + + if !bytes.Equal(alias, s) { + t.Errorf("alias of %T(%q) constructed via Slice = %T(%q)", s, s, alias, alias) + } + if cap(alias) != cap(s) { + t.Errorf("alias of %T with cap %d has cap %d", s, cap(s), cap(alias)) + } + }) + + t.Run("String", func(t *testing.T) { + s := "Hello, checkptr!" + + var alias string + hdr := (*unsafeheader.String)(unsafe.Pointer(&alias)) + hdr.Data = (*unsafeheader.String)(unsafe.Pointer(&s)).Data + hdr.Len = len(s) + + if alias != s { + t.Errorf("alias of %q constructed via String = %q", s, alias) + } + }) +} diff --git a/src/internal/xcoff/ar.go b/src/internal/xcoff/ar.go new file mode 100644 index 0000000..9cbd50d --- /dev/null +++ b/src/internal/xcoff/ar.go @@ -0,0 +1,226 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package xcoff + +import ( + "encoding/binary" + "fmt" + "io" + "os" + "strconv" + "strings" +) + +const ( + SAIAMAG = 0x8 + AIAFMAG = "`\n" + AIAMAG = "<aiaff>\n" + AIAMAGBIG = "<bigaf>\n" + + // Sizeof + FL_HSZ_BIG = 0x80 + AR_HSZ_BIG = 0x70 +) + +type bigarFileHeader struct { + Flmagic [SAIAMAG]byte // Archive magic string + Flmemoff [20]byte // Member table offset + Flgstoff [20]byte // 32-bits global symtab offset + Flgst64off [20]byte // 64-bits global symtab offset + Flfstmoff [20]byte // First member offset + Fllstmoff [20]byte // Last member offset + Flfreeoff [20]byte // First member on free list offset +} + +type bigarMemberHeader struct { + Arsize [20]byte // File member size + Arnxtmem [20]byte // Next member pointer + Arprvmem [20]byte // Previous member pointer + Ardate [12]byte // File member date + Aruid [12]byte // File member uid + Argid [12]byte // File member gid + Armode [12]byte // File member mode (octal) + Arnamlen [4]byte // File member name length + // _ar_nam is removed because it's easier to get name without it. +} + +// Archive represents an open AIX big archive. +type Archive struct { + ArchiveHeader + Members []*Member + + closer io.Closer +} + +// MemberHeader holds information about a big archive file header +type ArchiveHeader struct { + magic string +} + +// Member represents a member of an AIX big archive. +type Member struct { + MemberHeader + sr *io.SectionReader +} + +// MemberHeader holds information about a big archive member +type MemberHeader struct { + Name string + Size uint64 +} + +// OpenArchive opens the named archive using os.Open and prepares it for use +// as an AIX big archive. +func OpenArchive(name string) (*Archive, error) { + f, err := os.Open(name) + if err != nil { + return nil, err + } + arch, err := NewArchive(f) + if err != nil { + f.Close() + return nil, err + } + arch.closer = f + return arch, nil +} + +// Close closes the Archive. +// If the Archive was created using NewArchive directly instead of OpenArchive, +// Close has no effect. +func (a *Archive) Close() error { + var err error + if a.closer != nil { + err = a.closer.Close() + a.closer = nil + } + return err +} + +// NewArchive creates a new Archive for accessing an AIX big archive in an underlying reader. +func NewArchive(r io.ReaderAt) (*Archive, error) { + parseDecimalBytes := func(b []byte) (int64, error) { + return strconv.ParseInt(strings.TrimSpace(string(b)), 10, 64) + } + sr := io.NewSectionReader(r, 0, 1<<63-1) + + // Read File Header + var magic [SAIAMAG]byte + if _, err := sr.ReadAt(magic[:], 0); err != nil { + return nil, err + } + + arch := new(Archive) + switch string(magic[:]) { + case AIAMAGBIG: + arch.magic = string(magic[:]) + case AIAMAG: + return nil, fmt.Errorf("small AIX archive not supported") + default: + return nil, fmt.Errorf("unrecognised archive magic: 0x%x", magic) + } + + var fhdr bigarFileHeader + if _, err := sr.Seek(0, io.SeekStart); err != nil { + return nil, err + } + if err := binary.Read(sr, binary.BigEndian, &fhdr); err != nil { + return nil, err + } + + off, err := parseDecimalBytes(fhdr.Flfstmoff[:]) + if err != nil { + return nil, fmt.Errorf("error parsing offset of first member in archive header(%q); %v", fhdr, err) + } + + if off == 0 { + // Occurs if the archive is empty. + return arch, nil + } + + lastoff, err := parseDecimalBytes(fhdr.Fllstmoff[:]) + if err != nil { + return nil, fmt.Errorf("error parsing offset of first member in archive header(%q); %v", fhdr, err) + } + + // Read members + for { + // Read Member Header + // The member header is normally 2 bytes larger. But it's easier + // to read the name if the header is read without _ar_nam. + // However, AIAFMAG must be read afterward. + if _, err := sr.Seek(off, io.SeekStart); err != nil { + return nil, err + } + + var mhdr bigarMemberHeader + if err := binary.Read(sr, binary.BigEndian, &mhdr); err != nil { + return nil, err + } + + member := new(Member) + arch.Members = append(arch.Members, member) + + size, err := parseDecimalBytes(mhdr.Arsize[:]) + if err != nil { + return nil, fmt.Errorf("error parsing size in member header(%q); %v", mhdr, err) + } + member.Size = uint64(size) + + // Read name + namlen, err := parseDecimalBytes(mhdr.Arnamlen[:]) + if err != nil { + return nil, fmt.Errorf("error parsing name length in member header(%q); %v", mhdr, err) + } + name := make([]byte, namlen) + if err := binary.Read(sr, binary.BigEndian, name); err != nil { + return nil, err + } + member.Name = string(name) + + fileoff := off + AR_HSZ_BIG + namlen + if fileoff&1 != 0 { + fileoff++ + if _, err := sr.Seek(1, io.SeekCurrent); err != nil { + return nil, err + } + } + + // Read AIAFMAG string + var fmag [2]byte + if err := binary.Read(sr, binary.BigEndian, &fmag); err != nil { + return nil, err + } + if string(fmag[:]) != AIAFMAG { + return nil, fmt.Errorf("AIAFMAG not found after member header") + } + + fileoff += 2 // Add the two bytes of AIAFMAG + member.sr = io.NewSectionReader(sr, fileoff, size) + + if off == lastoff { + break + } + off, err = parseDecimalBytes(mhdr.Arnxtmem[:]) + if err != nil { + return nil, fmt.Errorf("error parsing offset of first member in archive header(%q); %v", fhdr, err) + } + + } + + return arch, nil +} + +// GetFile returns the XCOFF file defined by member name. +// FIXME: This doesn't work if an archive has two members with the same +// name which can occur if an archive has both 32-bits and 64-bits files. +func (arch *Archive) GetFile(name string) (*File, error) { + for _, mem := range arch.Members { + if mem.Name == name { + return NewFile(mem.sr) + } + } + return nil, fmt.Errorf("unknown member %s in archive", name) +} diff --git a/src/internal/xcoff/ar_test.go b/src/internal/xcoff/ar_test.go new file mode 100644 index 0000000..83333d6 --- /dev/null +++ b/src/internal/xcoff/ar_test.go @@ -0,0 +1,79 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package xcoff + +import ( + "reflect" + "testing" +) + +type archiveTest struct { + file string + hdr ArchiveHeader + members []*MemberHeader + membersFileHeader []FileHeader +} + +var archTest = []archiveTest{ + { + "testdata/bigar-ppc64", + ArchiveHeader{AIAMAGBIG}, + []*MemberHeader{ + {"printbye.o", 836}, + {"printhello.o", 860}, + }, + []FileHeader{ + {U64_TOCMAGIC}, + {U64_TOCMAGIC}, + }, + }, + { + "testdata/bigar-empty", + ArchiveHeader{AIAMAGBIG}, + []*MemberHeader{}, + []FileHeader{}, + }, +} + +func TestOpenArchive(t *testing.T) { + for i := range archTest { + tt := &archTest[i] + arch, err := OpenArchive(tt.file) + if err != nil { + t.Error(err) + continue + } + if !reflect.DeepEqual(arch.ArchiveHeader, tt.hdr) { + t.Errorf("open archive %s:\n\thave %#v\n\twant %#v\n", tt.file, arch.ArchiveHeader, tt.hdr) + continue + } + + for i, mem := range arch.Members { + if i >= len(tt.members) { + break + } + have := &mem.MemberHeader + want := tt.members[i] + if !reflect.DeepEqual(have, want) { + t.Errorf("open %s, member %d:\n\thave %#v\n\twant %#v\n", tt.file, i, have, want) + } + + f, err := arch.GetFile(mem.Name) + if err != nil { + t.Error(err) + continue + } + if !reflect.DeepEqual(f.FileHeader, tt.membersFileHeader[i]) { + t.Errorf("open %s, member file header %d:\n\thave %#v\n\twant %#v\n", tt.file, i, f.FileHeader, tt.membersFileHeader[i]) + } + } + tn := len(tt.members) + an := len(arch.Members) + if tn != an { + t.Errorf("open %s: len(Members) = %d, want %d", tt.file, an, tn) + } + + } +} diff --git a/src/internal/xcoff/file.go b/src/internal/xcoff/file.go new file mode 100644 index 0000000..9135822 --- /dev/null +++ b/src/internal/xcoff/file.go @@ -0,0 +1,697 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package xcoff implements access to XCOFF (Extended Common Object File Format) files. +package xcoff + +import ( + "debug/dwarf" + "encoding/binary" + "fmt" + "internal/saferio" + "io" + "os" + "strings" +) + +// SectionHeader holds information about an XCOFF section header. +type SectionHeader struct { + Name string + VirtualAddress uint64 + Size uint64 + Type uint32 + Relptr uint64 + Nreloc uint32 +} + +type Section struct { + SectionHeader + Relocs []Reloc + io.ReaderAt + sr *io.SectionReader +} + +// AuxiliaryCSect holds information about an XCOFF symbol in an AUX_CSECT entry. +type AuxiliaryCSect struct { + Length int64 + StorageMappingClass int + SymbolType int +} + +// AuxiliaryFcn holds information about an XCOFF symbol in an AUX_FCN entry. +type AuxiliaryFcn struct { + Size int64 +} + +type Symbol struct { + Name string + Value uint64 + SectionNumber int + StorageClass int + AuxFcn AuxiliaryFcn + AuxCSect AuxiliaryCSect +} + +type Reloc struct { + VirtualAddress uint64 + Symbol *Symbol + Signed bool + InstructionFixed bool + Length uint8 + Type uint8 +} + +// ImportedSymbol holds information about an imported XCOFF symbol. +type ImportedSymbol struct { + Name string + Library string +} + +// FileHeader holds information about an XCOFF file header. +type FileHeader struct { + TargetMachine uint16 +} + +// A File represents an open XCOFF file. +type File struct { + FileHeader + Sections []*Section + Symbols []*Symbol + StringTable []byte + LibraryPaths []string + + closer io.Closer +} + +// Open opens the named file using os.Open and prepares it for use as an XCOFF binary. +func Open(name string) (*File, error) { + f, err := os.Open(name) + if err != nil { + return nil, err + } + ff, err := NewFile(f) + if err != nil { + f.Close() + return nil, err + } + ff.closer = f + return ff, nil +} + +// Close closes the File. +// If the File was created using NewFile directly instead of Open, +// Close has no effect. +func (f *File) Close() error { + var err error + if f.closer != nil { + err = f.closer.Close() + f.closer = nil + } + return err +} + +// Section returns the first section with the given name, or nil if no such +// section exists. +// Xcoff have section's name limited to 8 bytes. Some sections like .gosymtab +// can be trunked but this method will still find them. +func (f *File) Section(name string) *Section { + for _, s := range f.Sections { + if s.Name == name || (len(name) > 8 && s.Name == name[:8]) { + return s + } + } + return nil +} + +// SectionByType returns the first section in f with the +// given type, or nil if there is no such section. +func (f *File) SectionByType(typ uint32) *Section { + for _, s := range f.Sections { + if s.Type == typ { + return s + } + } + return nil +} + +// cstring converts ASCII byte sequence b to string. +// It stops once it finds 0 or reaches end of b. +func cstring(b []byte) string { + var i int + for i = 0; i < len(b) && b[i] != 0; i++ { + } + return string(b[:i]) +} + +// getString extracts a string from an XCOFF string table. +func getString(st []byte, offset uint32) (string, bool) { + if offset < 4 || int(offset) >= len(st) { + return "", false + } + return cstring(st[offset:]), true +} + +// NewFile creates a new File for accessing an XCOFF binary in an underlying reader. +func NewFile(r io.ReaderAt) (*File, error) { + sr := io.NewSectionReader(r, 0, 1<<63-1) + // Read XCOFF target machine + var magic uint16 + if err := binary.Read(sr, binary.BigEndian, &magic); err != nil { + return nil, err + } + if magic != U802TOCMAGIC && magic != U64_TOCMAGIC { + return nil, fmt.Errorf("unrecognised XCOFF magic: 0x%x", magic) + } + + f := new(File) + f.TargetMachine = magic + + // Read XCOFF file header + if _, err := sr.Seek(0, io.SeekStart); err != nil { + return nil, err + } + var nscns uint16 + var symptr uint64 + var nsyms uint32 + var opthdr uint16 + var hdrsz int + switch f.TargetMachine { + case U802TOCMAGIC: + fhdr := new(FileHeader32) + if err := binary.Read(sr, binary.BigEndian, fhdr); err != nil { + return nil, err + } + nscns = fhdr.Fnscns + symptr = uint64(fhdr.Fsymptr) + nsyms = fhdr.Fnsyms + opthdr = fhdr.Fopthdr + hdrsz = FILHSZ_32 + case U64_TOCMAGIC: + fhdr := new(FileHeader64) + if err := binary.Read(sr, binary.BigEndian, fhdr); err != nil { + return nil, err + } + nscns = fhdr.Fnscns + symptr = fhdr.Fsymptr + nsyms = fhdr.Fnsyms + opthdr = fhdr.Fopthdr + hdrsz = FILHSZ_64 + } + + if symptr == 0 || nsyms <= 0 { + return nil, fmt.Errorf("no symbol table") + } + + // Read string table (located right after symbol table). + offset := symptr + uint64(nsyms)*SYMESZ + if _, err := sr.Seek(int64(offset), io.SeekStart); err != nil { + return nil, err + } + // The first 4 bytes contain the length (in bytes). + var l uint32 + if err := binary.Read(sr, binary.BigEndian, &l); err != nil { + return nil, err + } + if l > 4 { + st, err := saferio.ReadDataAt(sr, uint64(l), int64(offset)) + if err != nil { + return nil, err + } + f.StringTable = st + } + + // Read section headers + if _, err := sr.Seek(int64(hdrsz)+int64(opthdr), io.SeekStart); err != nil { + return nil, err + } + c := saferio.SliceCap((**Section)(nil), uint64(nscns)) + if c < 0 { + return nil, fmt.Errorf("too many XCOFF sections (%d)", nscns) + } + f.Sections = make([]*Section, 0, c) + for i := 0; i < int(nscns); i++ { + var scnptr uint64 + s := new(Section) + switch f.TargetMachine { + case U802TOCMAGIC: + shdr := new(SectionHeader32) + if err := binary.Read(sr, binary.BigEndian, shdr); err != nil { + return nil, err + } + s.Name = cstring(shdr.Sname[:]) + s.VirtualAddress = uint64(shdr.Svaddr) + s.Size = uint64(shdr.Ssize) + scnptr = uint64(shdr.Sscnptr) + s.Type = shdr.Sflags + s.Relptr = uint64(shdr.Srelptr) + s.Nreloc = uint32(shdr.Snreloc) + case U64_TOCMAGIC: + shdr := new(SectionHeader64) + if err := binary.Read(sr, binary.BigEndian, shdr); err != nil { + return nil, err + } + s.Name = cstring(shdr.Sname[:]) + s.VirtualAddress = shdr.Svaddr + s.Size = shdr.Ssize + scnptr = shdr.Sscnptr + s.Type = shdr.Sflags + s.Relptr = shdr.Srelptr + s.Nreloc = shdr.Snreloc + } + r2 := r + if scnptr == 0 { // .bss must have all 0s + r2 = zeroReaderAt{} + } + s.sr = io.NewSectionReader(r2, int64(scnptr), int64(s.Size)) + s.ReaderAt = s.sr + f.Sections = append(f.Sections, s) + } + + // Symbol map needed by relocation + var idxToSym = make(map[int]*Symbol) + + // Read symbol table + if _, err := sr.Seek(int64(symptr), io.SeekStart); err != nil { + return nil, err + } + f.Symbols = make([]*Symbol, 0) + for i := 0; i < int(nsyms); i++ { + var numaux int + var ok, needAuxFcn bool + sym := new(Symbol) + switch f.TargetMachine { + case U802TOCMAGIC: + se := new(SymEnt32) + if err := binary.Read(sr, binary.BigEndian, se); err != nil { + return nil, err + } + numaux = int(se.Nnumaux) + sym.SectionNumber = int(se.Nscnum) + sym.StorageClass = int(se.Nsclass) + sym.Value = uint64(se.Nvalue) + needAuxFcn = se.Ntype&SYM_TYPE_FUNC != 0 && numaux > 1 + zeroes := binary.BigEndian.Uint32(se.Nname[:4]) + if zeroes != 0 { + sym.Name = cstring(se.Nname[:]) + } else { + offset := binary.BigEndian.Uint32(se.Nname[4:]) + sym.Name, ok = getString(f.StringTable, offset) + if !ok { + goto skip + } + } + case U64_TOCMAGIC: + se := new(SymEnt64) + if err := binary.Read(sr, binary.BigEndian, se); err != nil { + return nil, err + } + numaux = int(se.Nnumaux) + sym.SectionNumber = int(se.Nscnum) + sym.StorageClass = int(se.Nsclass) + sym.Value = se.Nvalue + needAuxFcn = se.Ntype&SYM_TYPE_FUNC != 0 && numaux > 1 + sym.Name, ok = getString(f.StringTable, se.Noffset) + if !ok { + goto skip + } + } + if sym.StorageClass != C_EXT && sym.StorageClass != C_WEAKEXT && sym.StorageClass != C_HIDEXT { + goto skip + } + // Must have at least one csect auxiliary entry. + if numaux < 1 || i+numaux >= int(nsyms) { + goto skip + } + + if sym.SectionNumber > int(nscns) { + goto skip + } + if sym.SectionNumber == 0 { + sym.Value = 0 + } else { + sym.Value -= f.Sections[sym.SectionNumber-1].VirtualAddress + } + + idxToSym[i] = sym + + // If this symbol is a function, it must retrieve its size from + // its AUX_FCN entry. + // It can happen that a function symbol doesn't have any AUX_FCN. + // In this case, needAuxFcn is false and their size will be set to 0. + if needAuxFcn { + switch f.TargetMachine { + case U802TOCMAGIC: + aux := new(AuxFcn32) + if err := binary.Read(sr, binary.BigEndian, aux); err != nil { + return nil, err + } + sym.AuxFcn.Size = int64(aux.Xfsize) + case U64_TOCMAGIC: + aux := new(AuxFcn64) + if err := binary.Read(sr, binary.BigEndian, aux); err != nil { + return nil, err + } + sym.AuxFcn.Size = int64(aux.Xfsize) + } + } + + // Read csect auxiliary entry (by convention, it is the last). + if !needAuxFcn { + if _, err := sr.Seek(int64(numaux-1)*SYMESZ, io.SeekCurrent); err != nil { + return nil, err + } + } + i += numaux + numaux = 0 + switch f.TargetMachine { + case U802TOCMAGIC: + aux := new(AuxCSect32) + if err := binary.Read(sr, binary.BigEndian, aux); err != nil { + return nil, err + } + sym.AuxCSect.SymbolType = int(aux.Xsmtyp & 0x7) + sym.AuxCSect.StorageMappingClass = int(aux.Xsmclas) + sym.AuxCSect.Length = int64(aux.Xscnlen) + case U64_TOCMAGIC: + aux := new(AuxCSect64) + if err := binary.Read(sr, binary.BigEndian, aux); err != nil { + return nil, err + } + sym.AuxCSect.SymbolType = int(aux.Xsmtyp & 0x7) + sym.AuxCSect.StorageMappingClass = int(aux.Xsmclas) + sym.AuxCSect.Length = int64(aux.Xscnlenhi)<<32 | int64(aux.Xscnlenlo) + } + f.Symbols = append(f.Symbols, sym) + skip: + i += numaux // Skip auxiliary entries + if _, err := sr.Seek(int64(numaux)*SYMESZ, io.SeekCurrent); err != nil { + return nil, err + } + } + + // Read relocations + // Only for .data or .text section + for sectNum, sect := range f.Sections { + if sect.Type != STYP_TEXT && sect.Type != STYP_DATA { + continue + } + if sect.Relptr == 0 { + continue + } + c := saferio.SliceCap((*Reloc)(nil), uint64(sect.Nreloc)) + if c < 0 { + return nil, fmt.Errorf("too many relocs (%d) for section %d", sect.Nreloc, sectNum) + } + sect.Relocs = make([]Reloc, 0, c) + if _, err := sr.Seek(int64(sect.Relptr), io.SeekStart); err != nil { + return nil, err + } + for i := uint32(0); i < sect.Nreloc; i++ { + var reloc Reloc + switch f.TargetMachine { + case U802TOCMAGIC: + rel := new(Reloc32) + if err := binary.Read(sr, binary.BigEndian, rel); err != nil { + return nil, err + } + reloc.VirtualAddress = uint64(rel.Rvaddr) + reloc.Symbol = idxToSym[int(rel.Rsymndx)] + reloc.Type = rel.Rtype + reloc.Length = rel.Rsize&0x3F + 1 + + if rel.Rsize&0x80 != 0 { + reloc.Signed = true + } + if rel.Rsize&0x40 != 0 { + reloc.InstructionFixed = true + } + + case U64_TOCMAGIC: + rel := new(Reloc64) + if err := binary.Read(sr, binary.BigEndian, rel); err != nil { + return nil, err + } + reloc.VirtualAddress = rel.Rvaddr + reloc.Symbol = idxToSym[int(rel.Rsymndx)] + reloc.Type = rel.Rtype + reloc.Length = rel.Rsize&0x3F + 1 + if rel.Rsize&0x80 != 0 { + reloc.Signed = true + } + if rel.Rsize&0x40 != 0 { + reloc.InstructionFixed = true + } + } + + sect.Relocs = append(sect.Relocs, reloc) + } + } + + return f, nil +} + +// zeroReaderAt is ReaderAt that reads 0s. +type zeroReaderAt struct{} + +// ReadAt writes len(p) 0s into p. +func (w zeroReaderAt) ReadAt(p []byte, off int64) (n int, err error) { + for i := range p { + p[i] = 0 + } + return len(p), nil +} + +// Data reads and returns the contents of the XCOFF section s. +func (s *Section) Data() ([]byte, error) { + dat := make([]byte, s.sr.Size()) + n, err := s.sr.ReadAt(dat, 0) + if n == len(dat) { + err = nil + } + return dat[:n], err +} + +// CSect reads and returns the contents of a csect. +func (f *File) CSect(name string) []byte { + for _, sym := range f.Symbols { + if sym.Name == name && sym.AuxCSect.SymbolType == XTY_SD { + if i := sym.SectionNumber - 1; 0 <= i && i < len(f.Sections) { + s := f.Sections[i] + if sym.Value+uint64(sym.AuxCSect.Length) <= s.Size { + dat := make([]byte, sym.AuxCSect.Length) + _, err := s.sr.ReadAt(dat, int64(sym.Value)) + if err != nil { + return nil + } + return dat + } + } + break + } + } + return nil +} + +func (f *File) DWARF() (*dwarf.Data, error) { + // There are many other DWARF sections, but these + // are the ones the debug/dwarf package uses. + // Don't bother loading others. + var subtypes = [...]uint32{SSUBTYP_DWABREV, SSUBTYP_DWINFO, SSUBTYP_DWLINE, SSUBTYP_DWRNGES, SSUBTYP_DWSTR} + var dat [len(subtypes)][]byte + for i, subtype := range subtypes { + s := f.SectionByType(STYP_DWARF | subtype) + if s != nil { + b, err := s.Data() + if err != nil && uint64(len(b)) < s.Size { + return nil, err + } + dat[i] = b + } + } + + abbrev, info, line, ranges, str := dat[0], dat[1], dat[2], dat[3], dat[4] + return dwarf.New(abbrev, nil, nil, info, line, nil, ranges, str) +} + +// readImportID returns the import file IDs stored inside the .loader section. +// Library name pattern is either path/base/member or base/member +func (f *File) readImportIDs(s *Section) ([]string, error) { + // Read loader header + if _, err := s.sr.Seek(0, io.SeekStart); err != nil { + return nil, err + } + var istlen uint32 + var nimpid uint32 + var impoff uint64 + switch f.TargetMachine { + case U802TOCMAGIC: + lhdr := new(LoaderHeader32) + if err := binary.Read(s.sr, binary.BigEndian, lhdr); err != nil { + return nil, err + } + istlen = lhdr.Listlen + nimpid = lhdr.Lnimpid + impoff = uint64(lhdr.Limpoff) + case U64_TOCMAGIC: + lhdr := new(LoaderHeader64) + if err := binary.Read(s.sr, binary.BigEndian, lhdr); err != nil { + return nil, err + } + istlen = lhdr.Listlen + nimpid = lhdr.Lnimpid + impoff = lhdr.Limpoff + } + + // Read loader import file ID table + if _, err := s.sr.Seek(int64(impoff), io.SeekStart); err != nil { + return nil, err + } + table := make([]byte, istlen) + if _, err := io.ReadFull(s.sr, table); err != nil { + return nil, err + } + + offset := 0 + // First import file ID is the default LIBPATH value + libpath := cstring(table[offset:]) + f.LibraryPaths = strings.Split(libpath, ":") + offset += len(libpath) + 3 // 3 null bytes + all := make([]string, 0) + for i := 1; i < int(nimpid); i++ { + impidpath := cstring(table[offset:]) + offset += len(impidpath) + 1 + impidbase := cstring(table[offset:]) + offset += len(impidbase) + 1 + impidmem := cstring(table[offset:]) + offset += len(impidmem) + 1 + var path string + if len(impidpath) > 0 { + path = impidpath + "/" + impidbase + "/" + impidmem + } else { + path = impidbase + "/" + impidmem + } + all = append(all, path) + } + + return all, nil +} + +// ImportedSymbols returns the names of all symbols +// referred to by the binary f that are expected to be +// satisfied by other libraries at dynamic load time. +// It does not return weak symbols. +func (f *File) ImportedSymbols() ([]ImportedSymbol, error) { + s := f.SectionByType(STYP_LOADER) + if s == nil { + return nil, nil + } + // Read loader header + if _, err := s.sr.Seek(0, io.SeekStart); err != nil { + return nil, err + } + var stlen uint32 + var stoff uint64 + var nsyms uint32 + var symoff uint64 + switch f.TargetMachine { + case U802TOCMAGIC: + lhdr := new(LoaderHeader32) + if err := binary.Read(s.sr, binary.BigEndian, lhdr); err != nil { + return nil, err + } + stlen = lhdr.Lstlen + stoff = uint64(lhdr.Lstoff) + nsyms = lhdr.Lnsyms + symoff = LDHDRSZ_32 + case U64_TOCMAGIC: + lhdr := new(LoaderHeader64) + if err := binary.Read(s.sr, binary.BigEndian, lhdr); err != nil { + return nil, err + } + stlen = lhdr.Lstlen + stoff = lhdr.Lstoff + nsyms = lhdr.Lnsyms + symoff = lhdr.Lsymoff + } + + // Read loader section string table + if _, err := s.sr.Seek(int64(stoff), io.SeekStart); err != nil { + return nil, err + } + st := make([]byte, stlen) + if _, err := io.ReadFull(s.sr, st); err != nil { + return nil, err + } + + // Read imported libraries + libs, err := f.readImportIDs(s) + if err != nil { + return nil, err + } + + // Read loader symbol table + if _, err := s.sr.Seek(int64(symoff), io.SeekStart); err != nil { + return nil, err + } + all := make([]ImportedSymbol, 0) + for i := 0; i < int(nsyms); i++ { + var name string + var ifile uint32 + var ok bool + switch f.TargetMachine { + case U802TOCMAGIC: + ldsym := new(LoaderSymbol32) + if err := binary.Read(s.sr, binary.BigEndian, ldsym); err != nil { + return nil, err + } + if ldsym.Lsmtype&0x40 == 0 { + continue // Imported symbols only + } + zeroes := binary.BigEndian.Uint32(ldsym.Lname[:4]) + if zeroes != 0 { + name = cstring(ldsym.Lname[:]) + } else { + offset := binary.BigEndian.Uint32(ldsym.Lname[4:]) + name, ok = getString(st, offset) + if !ok { + continue + } + } + ifile = ldsym.Lifile + case U64_TOCMAGIC: + ldsym := new(LoaderSymbol64) + if err := binary.Read(s.sr, binary.BigEndian, ldsym); err != nil { + return nil, err + } + if ldsym.Lsmtype&0x40 == 0 { + continue // Imported symbols only + } + name, ok = getString(st, ldsym.Loffset) + if !ok { + continue + } + ifile = ldsym.Lifile + } + var sym ImportedSymbol + sym.Name = name + if ifile >= 1 && int(ifile) <= len(libs) { + sym.Library = libs[ifile-1] + } + all = append(all, sym) + } + + return all, nil +} + +// ImportedLibraries returns the names of all libraries +// referred to by the binary f that are expected to be +// linked with the binary at dynamic link time. +func (f *File) ImportedLibraries() ([]string, error) { + s := f.SectionByType(STYP_LOADER) + if s == nil { + return nil, nil + } + all, err := f.readImportIDs(s) + return all, err +} diff --git a/src/internal/xcoff/file_test.go b/src/internal/xcoff/file_test.go new file mode 100644 index 0000000..a6722e9 --- /dev/null +++ b/src/internal/xcoff/file_test.go @@ -0,0 +1,102 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package xcoff + +import ( + "reflect" + "testing" +) + +type fileTest struct { + file string + hdr FileHeader + sections []*SectionHeader + needed []string +} + +var fileTests = []fileTest{ + { + "testdata/gcc-ppc32-aix-dwarf2-exec", + FileHeader{U802TOCMAGIC}, + []*SectionHeader{ + {".text", 0x10000290, 0x00000bbd, STYP_TEXT, 0x7ae6, 0x36}, + {".data", 0x20000e4d, 0x00000437, STYP_DATA, 0x7d02, 0x2b}, + {".bss", 0x20001284, 0x0000021c, STYP_BSS, 0, 0}, + {".loader", 0x00000000, 0x000004b3, STYP_LOADER, 0, 0}, + {".dwline", 0x00000000, 0x000000df, STYP_DWARF | SSUBTYP_DWLINE, 0x7eb0, 0x7}, + {".dwinfo", 0x00000000, 0x00000314, STYP_DWARF | SSUBTYP_DWINFO, 0x7ef6, 0xa}, + {".dwabrev", 0x00000000, 0x000000d6, STYP_DWARF | SSUBTYP_DWABREV, 0, 0}, + {".dwarnge", 0x00000000, 0x00000020, STYP_DWARF | SSUBTYP_DWARNGE, 0x7f5a, 0x2}, + {".dwloc", 0x00000000, 0x00000074, STYP_DWARF | SSUBTYP_DWLOC, 0, 0}, + {".debug", 0x00000000, 0x00005e4f, STYP_DEBUG, 0, 0}, + }, + []string{"libc.a/shr.o"}, + }, + { + "testdata/gcc-ppc64-aix-dwarf2-exec", + FileHeader{U64_TOCMAGIC}, + []*SectionHeader{ + {".text", 0x10000480, 0x00000afd, STYP_TEXT, 0x8322, 0x34}, + {".data", 0x20000f7d, 0x000002f3, STYP_DATA, 0x85fa, 0x25}, + {".bss", 0x20001270, 0x00000428, STYP_BSS, 0, 0}, + {".loader", 0x00000000, 0x00000535, STYP_LOADER, 0, 0}, + {".dwline", 0x00000000, 0x000000b4, STYP_DWARF | SSUBTYP_DWLINE, 0x8800, 0x4}, + {".dwinfo", 0x00000000, 0x0000036a, STYP_DWARF | SSUBTYP_DWINFO, 0x8838, 0x7}, + {".dwabrev", 0x00000000, 0x000000b5, STYP_DWARF | SSUBTYP_DWABREV, 0, 0}, + {".dwarnge", 0x00000000, 0x00000040, STYP_DWARF | SSUBTYP_DWARNGE, 0x889a, 0x2}, + {".dwloc", 0x00000000, 0x00000062, STYP_DWARF | SSUBTYP_DWLOC, 0, 0}, + {".debug", 0x00000000, 0x00006605, STYP_DEBUG, 0, 0}, + }, + []string{"libc.a/shr_64.o"}, + }, +} + +func TestOpen(t *testing.T) { + for i := range fileTests { + tt := &fileTests[i] + + f, err := Open(tt.file) + if err != nil { + t.Error(err) + continue + } + if !reflect.DeepEqual(f.FileHeader, tt.hdr) { + t.Errorf("open %s:\n\thave %#v\n\twant %#v\n", tt.file, f.FileHeader, tt.hdr) + continue + } + + for i, sh := range f.Sections { + if i >= len(tt.sections) { + break + } + have := &sh.SectionHeader + want := tt.sections[i] + if !reflect.DeepEqual(have, want) { + t.Errorf("open %s, section %d:\n\thave %#v\n\twant %#v\n", tt.file, i, have, want) + } + } + tn := len(tt.sections) + fn := len(f.Sections) + if tn != fn { + t.Errorf("open %s: len(Sections) = %d, want %d", tt.file, fn, tn) + } + tl := tt.needed + fl, err := f.ImportedLibraries() + if err != nil { + t.Error(err) + } + if !reflect.DeepEqual(tl, fl) { + t.Errorf("open %s: loader import = %v, want %v", tt.file, tl, fl) + } + } +} + +func TestOpenFailure(t *testing.T) { + filename := "file.go" // not an XCOFF object file + _, err := Open(filename) // don't crash + if err == nil { + t.Errorf("open %s: succeeded unexpectedly", filename) + } +} diff --git a/src/internal/xcoff/testdata/bigar-empty b/src/internal/xcoff/testdata/bigar-empty new file mode 100644 index 0000000..851ccc5 --- /dev/null +++ b/src/internal/xcoff/testdata/bigar-empty @@ -0,0 +1,2 @@ +<bigaf> +0 0 0 0 0 0
\ No newline at end of file diff --git a/src/internal/xcoff/testdata/bigar-ppc64 b/src/internal/xcoff/testdata/bigar-ppc64 Binary files differnew file mode 100644 index 0000000..a8d4979 --- /dev/null +++ b/src/internal/xcoff/testdata/bigar-ppc64 diff --git a/src/internal/xcoff/testdata/gcc-ppc32-aix-dwarf2-exec b/src/internal/xcoff/testdata/gcc-ppc32-aix-dwarf2-exec Binary files differnew file mode 100644 index 0000000..810e21a --- /dev/null +++ b/src/internal/xcoff/testdata/gcc-ppc32-aix-dwarf2-exec diff --git a/src/internal/xcoff/testdata/gcc-ppc64-aix-dwarf2-exec b/src/internal/xcoff/testdata/gcc-ppc64-aix-dwarf2-exec Binary files differnew file mode 100644 index 0000000..707d01e --- /dev/null +++ b/src/internal/xcoff/testdata/gcc-ppc64-aix-dwarf2-exec diff --git a/src/internal/xcoff/testdata/hello.c b/src/internal/xcoff/testdata/hello.c new file mode 100644 index 0000000..34d9ee7 --- /dev/null +++ b/src/internal/xcoff/testdata/hello.c @@ -0,0 +1,7 @@ +#include <stdio.h> + +void +main(int argc, char *argv[]) +{ + printf("hello, world\n"); +} diff --git a/src/internal/xcoff/testdata/printbye.c b/src/internal/xcoff/testdata/printbye.c new file mode 100644 index 0000000..9045079 --- /dev/null +++ b/src/internal/xcoff/testdata/printbye.c @@ -0,0 +1,5 @@ +#include <stdio.h> + +void printbye(){ + printf("Goodbye\n"); +} diff --git a/src/internal/xcoff/testdata/printhello.c b/src/internal/xcoff/testdata/printhello.c new file mode 100644 index 0000000..182aa09 --- /dev/null +++ b/src/internal/xcoff/testdata/printhello.c @@ -0,0 +1,5 @@ +#include <stdio.h> + +void printhello(){ + printf("Helloworld\n"); +} diff --git a/src/internal/xcoff/xcoff.go b/src/internal/xcoff/xcoff.go new file mode 100644 index 0000000..db81542 --- /dev/null +++ b/src/internal/xcoff/xcoff.go @@ -0,0 +1,367 @@ +// Copyright 2018 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package xcoff + +// File Header. +type FileHeader32 struct { + Fmagic uint16 // Target machine + Fnscns uint16 // Number of sections + Ftimedat uint32 // Time and date of file creation + Fsymptr uint32 // Byte offset to symbol table start + Fnsyms uint32 // Number of entries in symbol table + Fopthdr uint16 // Number of bytes in optional header + Fflags uint16 // Flags +} + +type FileHeader64 struct { + Fmagic uint16 // Target machine + Fnscns uint16 // Number of sections + Ftimedat uint32 // Time and date of file creation + Fsymptr uint64 // Byte offset to symbol table start + Fopthdr uint16 // Number of bytes in optional header + Fflags uint16 // Flags + Fnsyms uint32 // Number of entries in symbol table +} + +const ( + FILHSZ_32 = 20 + FILHSZ_64 = 24 +) +const ( + U802TOCMAGIC = 0737 // AIX 32-bit XCOFF + U64_TOCMAGIC = 0767 // AIX 64-bit XCOFF +) + +// Flags that describe the type of the object file. +const ( + F_RELFLG = 0x0001 + F_EXEC = 0x0002 + F_LNNO = 0x0004 + F_FDPR_PROF = 0x0010 + F_FDPR_OPTI = 0x0020 + F_DSA = 0x0040 + F_VARPG = 0x0100 + F_DYNLOAD = 0x1000 + F_SHROBJ = 0x2000 + F_LOADONLY = 0x4000 +) + +// Section Header. +type SectionHeader32 struct { + Sname [8]byte // Section name + Spaddr uint32 // Physical address + Svaddr uint32 // Virtual address + Ssize uint32 // Section size + Sscnptr uint32 // Offset in file to raw data for section + Srelptr uint32 // Offset in file to relocation entries for section + Slnnoptr uint32 // Offset in file to line number entries for section + Snreloc uint16 // Number of relocation entries + Snlnno uint16 // Number of line number entries + Sflags uint32 // Flags to define the section type +} + +type SectionHeader64 struct { + Sname [8]byte // Section name + Spaddr uint64 // Physical address + Svaddr uint64 // Virtual address + Ssize uint64 // Section size + Sscnptr uint64 // Offset in file to raw data for section + Srelptr uint64 // Offset in file to relocation entries for section + Slnnoptr uint64 // Offset in file to line number entries for section + Snreloc uint32 // Number of relocation entries + Snlnno uint32 // Number of line number entries + Sflags uint32 // Flags to define the section type + Spad uint32 // Needs to be 72 bytes long +} + +// Flags defining the section type. +const ( + STYP_DWARF = 0x0010 + STYP_TEXT = 0x0020 + STYP_DATA = 0x0040 + STYP_BSS = 0x0080 + STYP_EXCEPT = 0x0100 + STYP_INFO = 0x0200 + STYP_TDATA = 0x0400 + STYP_TBSS = 0x0800 + STYP_LOADER = 0x1000 + STYP_DEBUG = 0x2000 + STYP_TYPCHK = 0x4000 + STYP_OVRFLO = 0x8000 +) +const ( + SSUBTYP_DWINFO = 0x10000 // DWARF info section + SSUBTYP_DWLINE = 0x20000 // DWARF line-number section + SSUBTYP_DWPBNMS = 0x30000 // DWARF public names section + SSUBTYP_DWPBTYP = 0x40000 // DWARF public types section + SSUBTYP_DWARNGE = 0x50000 // DWARF aranges section + SSUBTYP_DWABREV = 0x60000 // DWARF abbreviation section + SSUBTYP_DWSTR = 0x70000 // DWARF strings section + SSUBTYP_DWRNGES = 0x80000 // DWARF ranges section + SSUBTYP_DWLOC = 0x90000 // DWARF location lists section + SSUBTYP_DWFRAME = 0xA0000 // DWARF frames section + SSUBTYP_DWMAC = 0xB0000 // DWARF macros section +) + +// Symbol Table Entry. +type SymEnt32 struct { + Nname [8]byte // Symbol name + Nvalue uint32 // Symbol value + Nscnum uint16 // Section number of symbol + Ntype uint16 // Basic and derived type specification + Nsclass uint8 // Storage class of symbol + Nnumaux uint8 // Number of auxiliary entries +} + +type SymEnt64 struct { + Nvalue uint64 // Symbol value + Noffset uint32 // Offset of the name in string table or .debug section + Nscnum uint16 // Section number of symbol + Ntype uint16 // Basic and derived type specification + Nsclass uint8 // Storage class of symbol + Nnumaux uint8 // Number of auxiliary entries +} + +const SYMESZ = 18 + +const ( + // Nscnum + N_DEBUG = -2 + N_ABS = -1 + N_UNDEF = 0 + + //Ntype + SYM_V_INTERNAL = 0x1000 + SYM_V_HIDDEN = 0x2000 + SYM_V_PROTECTED = 0x3000 + SYM_V_EXPORTED = 0x4000 + SYM_TYPE_FUNC = 0x0020 // is function +) + +// Storage Class. +const ( + C_NULL = 0 // Symbol table entry marked for deletion + C_EXT = 2 // External symbol + C_STAT = 3 // Static symbol + C_BLOCK = 100 // Beginning or end of inner block + C_FCN = 101 // Beginning or end of function + C_FILE = 103 // Source file name and compiler information + C_HIDEXT = 107 // Unnamed external symbol + C_BINCL = 108 // Beginning of include file + C_EINCL = 109 // End of include file + C_WEAKEXT = 111 // Weak external symbol + C_DWARF = 112 // DWARF symbol + C_GSYM = 128 // Global variable + C_LSYM = 129 // Automatic variable allocated on stack + C_PSYM = 130 // Argument to subroutine allocated on stack + C_RSYM = 131 // Register variable + C_RPSYM = 132 // Argument to function or procedure stored in register + C_STSYM = 133 // Statically allocated symbol + C_BCOMM = 135 // Beginning of common block + C_ECOML = 136 // Local member of common block + C_ECOMM = 137 // End of common block + C_DECL = 140 // Declaration of object + C_ENTRY = 141 // Alternate entry + C_FUN = 142 // Function or procedure + C_BSTAT = 143 // Beginning of static block + C_ESTAT = 144 // End of static block + C_GTLS = 145 // Global thread-local variable + C_STTLS = 146 // Static thread-local variable +) + +// File Auxiliary Entry +type AuxFile64 struct { + Xfname [8]byte // Name or offset inside string table + Xftype uint8 // Source file string type + Xauxtype uint8 // Type of auxiliary entry +} + +// Function Auxiliary Entry +type AuxFcn32 struct { + Xexptr uint32 // File offset to exception table entry + Xfsize uint32 // Size of function in bytes + Xlnnoptr uint32 // File pointer to line number + Xendndx uint32 // Symbol table index of next entry + Xpad uint16 // Unused +} +type AuxFcn64 struct { + Xlnnoptr uint64 // File pointer to line number + Xfsize uint32 // Size of function in bytes + Xendndx uint32 // Symbol table index of next entry + Xpad uint8 // Unused + Xauxtype uint8 // Type of auxiliary entry +} + +type AuxSect64 struct { + Xscnlen uint64 // section length + Xnreloc uint64 // Num RLDs + pad uint8 + Xauxtype uint8 // Type of auxiliary entry +} + +// csect Auxiliary Entry. +type AuxCSect32 struct { + Xscnlen uint32 // Length or symbol table index + Xparmhash uint32 // Offset of parameter type-check string + Xsnhash uint16 // .typchk section number + Xsmtyp uint8 // Symbol alignment and type + Xsmclas uint8 // Storage-mapping class + Xstab uint32 // Reserved + Xsnstab uint16 // Reserved +} + +type AuxCSect64 struct { + Xscnlenlo uint32 // Lower 4 bytes of length or symbol table index + Xparmhash uint32 // Offset of parameter type-check string + Xsnhash uint16 // .typchk section number + Xsmtyp uint8 // Symbol alignment and type + Xsmclas uint8 // Storage-mapping class + Xscnlenhi uint32 // Upper 4 bytes of length or symbol table index + Xpad uint8 // Unused + Xauxtype uint8 // Type of auxiliary entry +} + +// Auxiliary type +const ( + _AUX_EXCEPT = 255 + _AUX_FCN = 254 + _AUX_SYM = 253 + _AUX_FILE = 252 + _AUX_CSECT = 251 + _AUX_SECT = 250 +) + +// Symbol type field. +const ( + XTY_ER = 0 // External reference + XTY_SD = 1 // Section definition + XTY_LD = 2 // Label definition + XTY_CM = 3 // Common csect definition +) + +// Defines for File auxiliary definitions: x_ftype field of x_file +const ( + XFT_FN = 0 // Source File Name + XFT_CT = 1 // Compile Time Stamp + XFT_CV = 2 // Compiler Version Number + XFT_CD = 128 // Compiler Defined Information +) + +// Storage-mapping class. +const ( + XMC_PR = 0 // Program code + XMC_RO = 1 // Read-only constant + XMC_DB = 2 // Debug dictionary table + XMC_TC = 3 // TOC entry + XMC_UA = 4 // Unclassified + XMC_RW = 5 // Read/Write data + XMC_GL = 6 // Global linkage + XMC_XO = 7 // Extended operation + XMC_SV = 8 // 32-bit supervisor call descriptor + XMC_BS = 9 // BSS class + XMC_DS = 10 // Function descriptor + XMC_UC = 11 // Unnamed FORTRAN common + XMC_TC0 = 15 // TOC anchor + XMC_TD = 16 // Scalar data entry in the TOC + XMC_SV64 = 17 // 64-bit supervisor call descriptor + XMC_SV3264 = 18 // Supervisor call descriptor for both 32-bit and 64-bit + XMC_TL = 20 // Read/Write thread-local data + XMC_UL = 21 // Read/Write thread-local data (.tbss) + XMC_TE = 22 // TOC entry +) + +// Loader Header. +type LoaderHeader32 struct { + Lversion uint32 // Loader section version number + Lnsyms uint32 // Number of symbol table entries + Lnreloc uint32 // Number of relocation table entries + Listlen uint32 // Length of import file ID string table + Lnimpid uint32 // Number of import file IDs + Limpoff uint32 // Offset to start of import file IDs + Lstlen uint32 // Length of string table + Lstoff uint32 // Offset to start of string table +} + +type LoaderHeader64 struct { + Lversion uint32 // Loader section version number + Lnsyms uint32 // Number of symbol table entries + Lnreloc uint32 // Number of relocation table entries + Listlen uint32 // Length of import file ID string table + Lnimpid uint32 // Number of import file IDs + Lstlen uint32 // Length of string table + Limpoff uint64 // Offset to start of import file IDs + Lstoff uint64 // Offset to start of string table + Lsymoff uint64 // Offset to start of symbol table + Lrldoff uint64 // Offset to start of relocation entries +} + +const ( + LDHDRSZ_32 = 32 + LDHDRSZ_64 = 56 +) + +// Loader Symbol. +type LoaderSymbol32 struct { + Lname [8]byte // Symbol name or byte offset into string table + Lvalue uint32 // Address field + Lscnum uint16 // Section number containing symbol + Lsmtype uint8 // Symbol type, export, import flags + Lsmclas uint8 // Symbol storage class + Lifile uint32 // Import file ID; ordinal of import file IDs + Lparm uint32 // Parameter type-check field +} + +type LoaderSymbol64 struct { + Lvalue uint64 // Address field + Loffset uint32 // Byte offset into string table of symbol name + Lscnum uint16 // Section number containing symbol + Lsmtype uint8 // Symbol type, export, import flags + Lsmclas uint8 // Symbol storage class + Lifile uint32 // Import file ID; ordinal of import file IDs + Lparm uint32 // Parameter type-check field +} + +type Reloc32 struct { + Rvaddr uint32 // (virtual) address of reference + Rsymndx uint32 // Index into symbol table + Rsize uint8 // Sign and reloc bit len + Rtype uint8 // Toc relocation type +} + +type Reloc64 struct { + Rvaddr uint64 // (virtual) address of reference + Rsymndx uint32 // Index into symbol table + Rsize uint8 // Sign and reloc bit len + Rtype uint8 // Toc relocation type +} + +const ( + R_POS = 0x00 // A(sym) Positive Relocation + R_NEG = 0x01 // -A(sym) Negative Relocation + R_REL = 0x02 // A(sym-*) Relative to self + R_TOC = 0x03 // A(sym-TOC) Relative to TOC + R_TRL = 0x12 // A(sym-TOC) TOC Relative indirect load. + + R_TRLA = 0x13 // A(sym-TOC) TOC Rel load address. modifiable inst + R_GL = 0x05 // A(external TOC of sym) Global Linkage + R_TCL = 0x06 // A(local TOC of sym) Local object TOC address + R_RL = 0x0C // A(sym) Pos indirect load. modifiable instruction + R_RLA = 0x0D // A(sym) Pos Load Address. modifiable instruction + R_REF = 0x0F // AL0(sym) Non relocating ref. No garbage collect + R_BA = 0x08 // A(sym) Branch absolute. Cannot modify instruction + R_RBA = 0x18 // A(sym) Branch absolute. modifiable instruction + R_BR = 0x0A // A(sym-*) Branch rel to self. non modifiable + R_RBR = 0x1A // A(sym-*) Branch rel to self. modifiable instr + + R_TLS = 0x20 // General-dynamic reference to TLS symbol + R_TLS_IE = 0x21 // Initial-exec reference to TLS symbol + R_TLS_LD = 0x22 // Local-dynamic reference to TLS symbol + R_TLS_LE = 0x23 // Local-exec reference to TLS symbol + R_TLSM = 0x24 // Module reference to TLS symbol + R_TLSML = 0x25 // Module reference to local (own) module + + R_TOCU = 0x30 // Relative to TOC - high order bits + R_TOCL = 0x31 // Relative to TOC - low order bits +) diff --git a/src/internal/zstd/bits.go b/src/internal/zstd/bits.go new file mode 100644 index 0000000..c9a2f70 --- /dev/null +++ b/src/internal/zstd/bits.go @@ -0,0 +1,130 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "math/bits" +) + +// block is the data for a single compressed block. +// The data starts immediately after the 3 byte block header, +// and is Block_Size bytes long. +type block []byte + +// bitReader reads a bit stream going forward. +type bitReader struct { + r *Reader // for error reporting + data block // the bits to read + off uint32 // current offset into data + bits uint32 // bits ready to be returned + cnt uint32 // number of valid bits in the bits field +} + +// makeBitReader makes a bit reader starting at off. +func (r *Reader) makeBitReader(data block, off int) bitReader { + return bitReader{ + r: r, + data: data, + off: uint32(off), + } +} + +// moreBits is called to read more bits. +// This ensures that at least 16 bits are available. +func (br *bitReader) moreBits() error { + for br.cnt < 16 { + if br.off >= uint32(len(br.data)) { + return br.r.makeEOFError(int(br.off)) + } + c := br.data[br.off] + br.off++ + br.bits |= uint32(c) << br.cnt + br.cnt += 8 + } + return nil +} + +// val is called to fetch a value of b bits. +func (br *bitReader) val(b uint8) uint32 { + r := br.bits & ((1 << b) - 1) + br.bits >>= b + br.cnt -= uint32(b) + return r +} + +// backup steps back to the last byte we used. +func (br *bitReader) backup() { + for br.cnt >= 8 { + br.off-- + br.cnt -= 8 + } +} + +// makeError returns an error at the current offset wrapping a string. +func (br *bitReader) makeError(msg string) error { + return br.r.makeError(int(br.off), msg) +} + +// reverseBitReader reads a bit stream in reverse. +type reverseBitReader struct { + r *Reader // for error reporting + data block // the bits to read + off uint32 // current offset into data + start uint32 // start in data; we read backward to start + bits uint32 // bits ready to be returned + cnt uint32 // number of valid bits in bits field +} + +// makeReverseBitReader makes a reverseBitReader reading backward +// from off to start. The bitstream starts with a 1 bit in the last +// byte, at off. +func (r *Reader) makeReverseBitReader(data block, off, start int) (reverseBitReader, error) { + streamStart := data[off] + if streamStart == 0 { + return reverseBitReader{}, r.makeError(off, "zero byte at reverse bit stream start") + } + rbr := reverseBitReader{ + r: r, + data: data, + off: uint32(off), + start: uint32(start), + bits: uint32(streamStart), + cnt: uint32(7 - bits.LeadingZeros8(streamStart)), + } + return rbr, nil +} + +// val is called to fetch a value of b bits. +func (rbr *reverseBitReader) val(b uint8) (uint32, error) { + if !rbr.fetch(b) { + return 0, rbr.r.makeEOFError(int(rbr.off)) + } + + rbr.cnt -= uint32(b) + v := (rbr.bits >> rbr.cnt) & ((1 << b) - 1) + return v, nil +} + +// fetch is called to ensure that at least b bits are available. +// It reports false if this can't be done, +// in which case only rbr.cnt bits are available. +func (rbr *reverseBitReader) fetch(b uint8) bool { + for rbr.cnt < uint32(b) { + if rbr.off <= rbr.start { + return false + } + rbr.off-- + c := rbr.data[rbr.off] + rbr.bits <<= 8 + rbr.bits |= uint32(c) + rbr.cnt += 8 + } + return true +} + +// makeError returns an error at the current offset wrapping a string. +func (rbr *reverseBitReader) makeError(msg string) error { + return rbr.r.makeError(int(rbr.off), msg) +} diff --git a/src/internal/zstd/block.go b/src/internal/zstd/block.go new file mode 100644 index 0000000..bd3040c --- /dev/null +++ b/src/internal/zstd/block.go @@ -0,0 +1,436 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "io" +) + +// debug can be set in the source to print debug info using println. +const debug = false + +// compressedBlock decompresses a compressed block, storing the decompressed +// data in r.buffer. The blockSize argument is the compressed size. +// RFC 3.1.1.3. +func (r *Reader) compressedBlock(blockSize int) error { + if len(r.compressedBuf) >= blockSize { + r.compressedBuf = r.compressedBuf[:blockSize] + } else { + // We know that blockSize <= 128K, + // so this won't allocate an enormous amount. + need := blockSize - len(r.compressedBuf) + r.compressedBuf = append(r.compressedBuf, make([]byte, need)...) + } + + if _, err := io.ReadFull(r.r, r.compressedBuf); err != nil { + return r.wrapNonEOFError(0, err) + } + + data := block(r.compressedBuf) + off := 0 + r.buffer = r.buffer[:0] + + litoff, litbuf, err := r.readLiterals(data, off, r.literals[:0]) + if err != nil { + return err + } + r.literals = litbuf + + off = litoff + + seqCount, off, err := r.initSeqs(data, off) + if err != nil { + return err + } + + if seqCount == 0 { + // No sequences, just literals. + if off < len(data) { + return r.makeError(off, "extraneous data after no sequences") + } + if len(litbuf) == 0 { + return r.makeError(off, "no sequences and no literals") + } + r.buffer = append(r.buffer, litbuf...) + return nil + } + + return r.execSeqs(data, off, litbuf, seqCount) +} + +// seqCode is the kind of sequence codes we have to handle. +type seqCode int + +const ( + seqLiteral seqCode = iota + seqOffset + seqMatch +) + +// seqCodeInfoData is the information needed to set up seqTables and +// seqTableBits for a particular kind of sequence code. +type seqCodeInfoData struct { + predefTable []fseBaselineEntry // predefined FSE + predefTableBits int // number of bits in predefTable + maxSym int // max symbol value in FSE + maxBits int // max bits for FSE + + // toBaseline converts from an FSE table to an FSE baseline table. + toBaseline func(*Reader, int, []fseEntry, []fseBaselineEntry) error +} + +// seqCodeInfo is the seqCodeInfoData for each kind of sequence code. +var seqCodeInfo = [3]seqCodeInfoData{ + seqLiteral: { + predefTable: predefinedLiteralTable[:], + predefTableBits: 6, + maxSym: 35, + maxBits: 9, + toBaseline: (*Reader).makeLiteralBaselineFSE, + }, + seqOffset: { + predefTable: predefinedOffsetTable[:], + predefTableBits: 5, + maxSym: 31, + maxBits: 8, + toBaseline: (*Reader).makeOffsetBaselineFSE, + }, + seqMatch: { + predefTable: predefinedMatchTable[:], + predefTableBits: 6, + maxSym: 52, + maxBits: 9, + toBaseline: (*Reader).makeMatchBaselineFSE, + }, +} + +// initSeqs reads the Sequences_Section_Header and sets up the FSE +// tables used to read the sequence codes. It returns the number of +// sequences and the new offset. RFC 3.1.1.3.2.1. +func (r *Reader) initSeqs(data block, off int) (int, int, error) { + if off >= len(data) { + return 0, 0, r.makeEOFError(off) + } + + seqHdr := data[off] + off++ + if seqHdr == 0 { + return 0, off, nil + } + + var seqCount int + if seqHdr < 128 { + seqCount = int(seqHdr) + } else if seqHdr < 255 { + if off >= len(data) { + return 0, 0, r.makeEOFError(off) + } + seqCount = ((int(seqHdr) - 128) << 8) + int(data[off]) + off++ + } else { + if off+1 >= len(data) { + return 0, 0, r.makeEOFError(off) + } + seqCount = int(data[off]) + (int(data[off+1]) << 8) + 0x7f00 + off += 2 + } + + // Read the Symbol_Compression_Modes byte. + + if off >= len(data) { + return 0, 0, r.makeEOFError(off) + } + symMode := data[off] + if symMode&3 != 0 { + return 0, 0, r.makeError(off, "invalid symbol compression mode") + } + off++ + + // Set up the FSE tables used to decode the sequence codes. + + var err error + off, err = r.setSeqTable(data, off, seqLiteral, (symMode>>6)&3) + if err != nil { + return 0, 0, err + } + + off, err = r.setSeqTable(data, off, seqOffset, (symMode>>4)&3) + if err != nil { + return 0, 0, err + } + + off, err = r.setSeqTable(data, off, seqMatch, (symMode>>2)&3) + if err != nil { + return 0, 0, err + } + + return seqCount, off, nil +} + +// setSeqTable uses the Compression_Mode in mode to set up r.seqTables and +// r.seqTableBits for kind. We store these in the Reader because one of +// the modes simply reuses the value from the last block in the frame. +func (r *Reader) setSeqTable(data block, off int, kind seqCode, mode byte) (int, error) { + info := &seqCodeInfo[kind] + switch mode { + case 0: + // Predefined_Mode + r.seqTables[kind] = info.predefTable + r.seqTableBits[kind] = uint8(info.predefTableBits) + return off, nil + + case 1: + // RLE_Mode + if off >= len(data) { + return 0, r.makeEOFError(off) + } + rle := data[off] + off++ + + // Build a simple baseline table that always returns rle. + + entry := []fseEntry{ + { + sym: rle, + bits: 0, + base: 0, + }, + } + if cap(r.seqTableBuffers[kind]) == 0 { + r.seqTableBuffers[kind] = make([]fseBaselineEntry, 1<<info.maxBits) + } + r.seqTableBuffers[kind] = r.seqTableBuffers[kind][:1] + if err := info.toBaseline(r, off, entry, r.seqTableBuffers[kind]); err != nil { + return 0, err + } + + r.seqTables[kind] = r.seqTableBuffers[kind] + r.seqTableBits[kind] = 0 + return off, nil + + case 2: + // FSE_Compressed_Mode + if cap(r.fseScratch) < 1<<info.maxBits { + r.fseScratch = make([]fseEntry, 1<<info.maxBits) + } + r.fseScratch = r.fseScratch[:1<<info.maxBits] + + tableBits, roff, err := r.readFSE(data, off, info.maxSym, info.maxBits, r.fseScratch) + if err != nil { + return 0, err + } + r.fseScratch = r.fseScratch[:1<<tableBits] + + if cap(r.seqTableBuffers[kind]) == 0 { + r.seqTableBuffers[kind] = make([]fseBaselineEntry, 1<<info.maxBits) + } + r.seqTableBuffers[kind] = r.seqTableBuffers[kind][:1<<tableBits] + + if err := info.toBaseline(r, roff, r.fseScratch, r.seqTableBuffers[kind]); err != nil { + return 0, err + } + + r.seqTables[kind] = r.seqTableBuffers[kind] + r.seqTableBits[kind] = uint8(tableBits) + return roff, nil + + case 3: + // Repeat_Mode + if len(r.seqTables[kind]) == 0 { + return 0, r.makeError(off, "missing repeat sequence FSE table") + } + return off, nil + } + panic("unreachable") +} + +// execSeqs reads and executes the sequences. RFC 3.1.1.3.2.1.2. +func (r *Reader) execSeqs(data block, off int, litbuf []byte, seqCount int) error { + // Set up the initial states for the sequence code readers. + + rbr, err := r.makeReverseBitReader(data, len(data)-1, off) + if err != nil { + return err + } + + literalState, err := rbr.val(r.seqTableBits[seqLiteral]) + if err != nil { + return err + } + + offsetState, err := rbr.val(r.seqTableBits[seqOffset]) + if err != nil { + return err + } + + matchState, err := rbr.val(r.seqTableBits[seqMatch]) + if err != nil { + return err + } + + // Read and perform all the sequences. RFC 3.1.1.4. + + seq := 0 + for seq < seqCount { + if len(r.buffer)+len(litbuf) > 128<<10 { + return rbr.makeError("uncompressed size too big") + } + + ptoffset := &r.seqTables[seqOffset][offsetState] + ptmatch := &r.seqTables[seqMatch][matchState] + ptliteral := &r.seqTables[seqLiteral][literalState] + + add, err := rbr.val(ptoffset.basebits) + if err != nil { + return err + } + offset := ptoffset.baseline + add + + add, err = rbr.val(ptmatch.basebits) + if err != nil { + return err + } + match := ptmatch.baseline + add + + add, err = rbr.val(ptliteral.basebits) + if err != nil { + return err + } + literal := ptliteral.baseline + add + + // Handle repeat offsets. RFC 3.1.1.5. + // See the comment in makeOffsetBaselineFSE. + if ptoffset.basebits > 1 { + r.repeatedOffset3 = r.repeatedOffset2 + r.repeatedOffset2 = r.repeatedOffset1 + r.repeatedOffset1 = offset + } else { + if literal == 0 { + offset++ + } + switch offset { + case 1: + offset = r.repeatedOffset1 + case 2: + offset = r.repeatedOffset2 + r.repeatedOffset2 = r.repeatedOffset1 + r.repeatedOffset1 = offset + case 3: + offset = r.repeatedOffset3 + r.repeatedOffset3 = r.repeatedOffset2 + r.repeatedOffset2 = r.repeatedOffset1 + r.repeatedOffset1 = offset + case 4: + offset = r.repeatedOffset1 - 1 + r.repeatedOffset3 = r.repeatedOffset2 + r.repeatedOffset2 = r.repeatedOffset1 + r.repeatedOffset1 = offset + } + } + + seq++ + if seq < seqCount { + // Update the states. + add, err = rbr.val(ptliteral.bits) + if err != nil { + return err + } + literalState = uint32(ptliteral.base) + add + + add, err = rbr.val(ptmatch.bits) + if err != nil { + return err + } + matchState = uint32(ptmatch.base) + add + + add, err = rbr.val(ptoffset.bits) + if err != nil { + return err + } + offsetState = uint32(ptoffset.base) + add + } + + // The next sequence is now in literal, offset, match. + + if debug { + println("literal", literal, "offset", offset, "match", match) + } + + // Copy literal bytes from litbuf. + if literal > uint32(len(litbuf)) { + return rbr.makeError("literal byte overflow") + } + if literal > 0 { + r.buffer = append(r.buffer, litbuf[:literal]...) + litbuf = litbuf[literal:] + } + + if match > 0 { + if err := r.copyFromWindow(&rbr, offset, match); err != nil { + return err + } + } + } + + if len(litbuf) > 0 { + r.buffer = append(r.buffer, litbuf...) + } + + if rbr.cnt != 0 { + return r.makeError(off, "extraneous data after sequences") + } + + return nil +} + +// Copy match bytes from the decoded output, or the window, at offset. +func (r *Reader) copyFromWindow(rbr *reverseBitReader, offset, match uint32) error { + if offset == 0 { + return rbr.makeError("invalid zero offset") + } + + lenBlock := uint32(len(r.buffer)) + if lenBlock < offset { + lenWindow := uint32(len(r.window)) + windowOffset := offset - lenBlock + if windowOffset > lenWindow { + return rbr.makeError("offset past window") + } + from := lenWindow - windowOffset + if from+match <= lenWindow { + r.buffer = append(r.buffer, r.window[from:from+match]...) + return nil + } + r.buffer = append(r.buffer, r.window[from:]...) + copied := lenWindow - from + offset -= copied + match -= copied + + if offset == 0 && match > 0 { + return rbr.makeError("invalid offset") + } + } + + from := lenBlock - offset + if offset >= match { + r.buffer = append(r.buffer, r.buffer[from:from+match]...) + return nil + } + + // We are being asked to copy data that we are adding to the + // buffer in the same copy. + for match > 0 { + var copy uint32 + if offset >= match { + copy = match + } else { + copy = offset + } + r.buffer = append(r.buffer, r.buffer[from:from+copy]...) + match -= copy + from += copy + } + return nil +} diff --git a/src/internal/zstd/fse.go b/src/internal/zstd/fse.go new file mode 100644 index 0000000..ea661d4 --- /dev/null +++ b/src/internal/zstd/fse.go @@ -0,0 +1,437 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "math/bits" +) + +// fseEntry is one entry in an FSE table. +type fseEntry struct { + sym uint8 // value that this entry records + bits uint8 // number of bits to read to determine next state + base uint16 // add those bits to this state to get the next state +} + +// readFSE reads an FSE table from data starting at off. +// maxSym is the maximum symbol value. +// maxBits is the maximum number of bits permitted for symbols in the table. +// The FSE is written into table, which must be at least 1<<maxBits in size. +// This returns the number of bits in the FSE table and the new offset. +// RFC 4.1.1. +func (r *Reader) readFSE(data block, off, maxSym, maxBits int, table []fseEntry) (tableBits, roff int, err error) { + br := r.makeBitReader(data, off) + if err := br.moreBits(); err != nil { + return 0, 0, err + } + + accuracyLog := int(br.val(4)) + 5 + if accuracyLog > maxBits { + return 0, 0, br.makeError("FSE accuracy log too large") + } + + // The number of remaining probabilities, plus 1. + // This determines the number of bits to be read for the next value. + remaining := (1 << accuracyLog) + 1 + + // The current difference between small and large values, + // which depends on the number of remaining values. + // Small values use 1 less bit. + threshold := 1 << accuracyLog + + // The number of bits needed to compute threshold. + bitsNeeded := accuracyLog + 1 + + // The next character value. + sym := 0 + + // Whether the last count was 0. + prev0 := false + + var norm [256]int16 + + for remaining > 1 && sym <= maxSym { + if err := br.moreBits(); err != nil { + return 0, 0, err + } + + if prev0 { + // Previous count was 0, so there is a 2-bit + // repeat flag. If the 2-bit flag is 0b11, + // it adds 3 and then there is another repeat flag. + zsym := sym + for (br.bits & 0xfff) == 0xfff { + zsym += 3 * 6 + br.bits >>= 12 + br.cnt -= 12 + if err := br.moreBits(); err != nil { + return 0, 0, err + } + } + for (br.bits & 3) == 3 { + zsym += 3 + br.bits >>= 2 + br.cnt -= 2 + if err := br.moreBits(); err != nil { + return 0, 0, err + } + } + + // We have at least 14 bits here, + // no need to call moreBits + + zsym += int(br.val(2)) + + if zsym > maxSym { + return 0, 0, br.makeError("FSE symbol index overflow") + } + + for ; sym < zsym; sym++ { + norm[uint8(sym)] = 0 + } + + prev0 = false + continue + } + + max := (2*threshold - 1) - remaining + var count int + if int(br.bits&uint32(threshold-1)) < max { + // A small value. + count = int(br.bits & uint32((threshold - 1))) + br.bits >>= bitsNeeded - 1 + br.cnt -= uint32(bitsNeeded - 1) + } else { + // A large value. + count = int(br.bits & uint32((2*threshold - 1))) + if count >= threshold { + count -= max + } + br.bits >>= bitsNeeded + br.cnt -= uint32(bitsNeeded) + } + + count-- + if count >= 0 { + remaining -= count + } else { + remaining-- + } + if sym >= 256 { + return 0, 0, br.makeError("FSE sym overflow") + } + norm[uint8(sym)] = int16(count) + sym++ + + prev0 = count == 0 + + for remaining < threshold { + bitsNeeded-- + threshold >>= 1 + } + } + + if remaining != 1 { + return 0, 0, br.makeError("too many symbols in FSE table") + } + + for ; sym <= maxSym; sym++ { + norm[uint8(sym)] = 0 + } + + br.backup() + + if err := r.buildFSE(off, norm[:maxSym+1], table, accuracyLog); err != nil { + return 0, 0, err + } + + return accuracyLog, int(br.off), nil +} + +// buildFSE builds an FSE decoding table from a list of probabilities. +// The probabilities are in norm. next is scratch space. The number of bits +// in the table is tableBits. +func (r *Reader) buildFSE(off int, norm []int16, table []fseEntry, tableBits int) error { + tableSize := 1 << tableBits + highThreshold := tableSize - 1 + + var next [256]uint16 + + for i, n := range norm { + if n >= 0 { + next[uint8(i)] = uint16(n) + } else { + table[highThreshold].sym = uint8(i) + highThreshold-- + next[uint8(i)] = 1 + } + } + + pos := 0 + step := (tableSize >> 1) + (tableSize >> 3) + 3 + mask := tableSize - 1 + for i, n := range norm { + for j := 0; j < int(n); j++ { + table[pos].sym = uint8(i) + pos = (pos + step) & mask + for pos > highThreshold { + pos = (pos + step) & mask + } + } + } + if pos != 0 { + return r.makeError(off, "FSE count error") + } + + for i := 0; i < tableSize; i++ { + sym := table[i].sym + nextState := next[sym] + next[sym]++ + + if nextState == 0 { + return r.makeError(off, "FSE state error") + } + + highBit := 15 - bits.LeadingZeros16(nextState) + + bits := tableBits - highBit + table[i].bits = uint8(bits) + table[i].base = (nextState << bits) - uint16(tableSize) + } + + return nil +} + +// fseBaselineEntry is an entry in an FSE baseline table. +// We use these for literal/match/length values. +// Those require mapping the symbol to a baseline value, +// and then reading zero or more bits and adding the value to the baseline. +// Rather than looking thees up in separate tables, +// we convert the FSE table to an FSE baseline table. +type fseBaselineEntry struct { + baseline uint32 // baseline for value that this entry represents + basebits uint8 // number of bits to read to add to baseline + bits uint8 // number of bits to read to determine next state + base uint16 // add the bits to this base to get the next state +} + +// Given a literal length code, we need to read a number of bits and +// add that to a baseline. For states 0 to 15 the baseline is the +// state and the number of bits is zero. RFC 3.1.1.3.2.1.1. + +const literalLengthOffset = 16 + +var literalLengthBase = []uint32{ + 16 | (1 << 24), + 18 | (1 << 24), + 20 | (1 << 24), + 22 | (1 << 24), + 24 | (2 << 24), + 28 | (2 << 24), + 32 | (3 << 24), + 40 | (3 << 24), + 48 | (4 << 24), + 64 | (6 << 24), + 128 | (7 << 24), + 256 | (8 << 24), + 512 | (9 << 24), + 1024 | (10 << 24), + 2048 | (11 << 24), + 4096 | (12 << 24), + 8192 | (13 << 24), + 16384 | (14 << 24), + 32768 | (15 << 24), + 65536 | (16 << 24), +} + +// makeLiteralBaselineFSE converts the literal length fseTable to baselineTable. +func (r *Reader) makeLiteralBaselineFSE(off int, fseTable []fseEntry, baselineTable []fseBaselineEntry) error { + for i, e := range fseTable { + be := fseBaselineEntry{ + bits: e.bits, + base: e.base, + } + if e.sym < literalLengthOffset { + be.baseline = uint32(e.sym) + be.basebits = 0 + } else { + if e.sym > 35 { + return r.makeError(off, "FSE baseline symbol overflow") + } + idx := e.sym - literalLengthOffset + basebits := literalLengthBase[idx] + be.baseline = basebits & 0xffffff + be.basebits = uint8(basebits >> 24) + } + baselineTable[i] = be + } + return nil +} + +// makeOffsetBaselineFSE converts the offset length fseTable to baselineTable. +func (r *Reader) makeOffsetBaselineFSE(off int, fseTable []fseEntry, baselineTable []fseBaselineEntry) error { + for i, e := range fseTable { + be := fseBaselineEntry{ + bits: e.bits, + base: e.base, + } + if e.sym > 31 { + return r.makeError(off, "FSE offset symbol overflow") + } + + // The simple way to write this is + // be.baseline = 1 << e.sym + // be.basebits = e.sym + // That would give us an offset value that corresponds to + // the one described in the RFC. However, for offsets > 3 + // we have to subtract 3. And for offset values 1, 2, 3 + // we use a repeated offset. + // + // The baseline is always a power of 2, and is never 0, + // so for those low values we will see one entry that is + // baseline 1, basebits 0, and one entry that is baseline 2, + // basebits 1. All other entries will have baseline >= 4 + // basebits >= 2. + // + // So we can check for RFC offset <= 3 by checking for + // basebits <= 1. That means that we can subtract 3 here + // and not worry about doing it in the hot loop. + + be.baseline = 1 << e.sym + if e.sym >= 2 { + be.baseline -= 3 + } + be.basebits = e.sym + baselineTable[i] = be + } + return nil +} + +// Given a match length code, we need to read a number of bits and add +// that to a baseline. For states 0 to 31 the baseline is state+3 and +// the number of bits is zero. RFC 3.1.1.3.2.1.1. + +const matchLengthOffset = 32 + +var matchLengthBase = []uint32{ + 35 | (1 << 24), + 37 | (1 << 24), + 39 | (1 << 24), + 41 | (1 << 24), + 43 | (2 << 24), + 47 | (2 << 24), + 51 | (3 << 24), + 59 | (3 << 24), + 67 | (4 << 24), + 83 | (4 << 24), + 99 | (5 << 24), + 131 | (7 << 24), + 259 | (8 << 24), + 515 | (9 << 24), + 1027 | (10 << 24), + 2051 | (11 << 24), + 4099 | (12 << 24), + 8195 | (13 << 24), + 16387 | (14 << 24), + 32771 | (15 << 24), + 65539 | (16 << 24), +} + +// makeMatchBaselineFSE converts the match length fseTable to baselineTable. +func (r *Reader) makeMatchBaselineFSE(off int, fseTable []fseEntry, baselineTable []fseBaselineEntry) error { + for i, e := range fseTable { + be := fseBaselineEntry{ + bits: e.bits, + base: e.base, + } + if e.sym < matchLengthOffset { + be.baseline = uint32(e.sym) + 3 + be.basebits = 0 + } else { + if e.sym > 52 { + return r.makeError(off, "FSE baseline symbol overflow") + } + idx := e.sym - matchLengthOffset + basebits := matchLengthBase[idx] + be.baseline = basebits & 0xffffff + be.basebits = uint8(basebits >> 24) + } + baselineTable[i] = be + } + return nil +} + +// predefinedLiteralTable is the predefined table to use for literal lengths. +// Generated from table in RFC 3.1.1.3.2.2.1. +// Checked by TestPredefinedTables. +var predefinedLiteralTable = [...]fseBaselineEntry{ + {0, 0, 4, 0}, {0, 0, 4, 16}, {1, 0, 5, 32}, + {3, 0, 5, 0}, {4, 0, 5, 0}, {6, 0, 5, 0}, + {7, 0, 5, 0}, {9, 0, 5, 0}, {10, 0, 5, 0}, + {12, 0, 5, 0}, {14, 0, 6, 0}, {16, 1, 5, 0}, + {20, 1, 5, 0}, {22, 1, 5, 0}, {28, 2, 5, 0}, + {32, 3, 5, 0}, {48, 4, 5, 0}, {64, 6, 5, 32}, + {128, 7, 5, 0}, {256, 8, 6, 0}, {1024, 10, 6, 0}, + {4096, 12, 6, 0}, {0, 0, 4, 32}, {1, 0, 4, 0}, + {2, 0, 5, 0}, {4, 0, 5, 32}, {5, 0, 5, 0}, + {7, 0, 5, 32}, {8, 0, 5, 0}, {10, 0, 5, 32}, + {11, 0, 5, 0}, {13, 0, 6, 0}, {16, 1, 5, 32}, + {18, 1, 5, 0}, {22, 1, 5, 32}, {24, 2, 5, 0}, + {32, 3, 5, 32}, {40, 3, 5, 0}, {64, 6, 4, 0}, + {64, 6, 4, 16}, {128, 7, 5, 32}, {512, 9, 6, 0}, + {2048, 11, 6, 0}, {0, 0, 4, 48}, {1, 0, 4, 16}, + {2, 0, 5, 32}, {3, 0, 5, 32}, {5, 0, 5, 32}, + {6, 0, 5, 32}, {8, 0, 5, 32}, {9, 0, 5, 32}, + {11, 0, 5, 32}, {12, 0, 5, 32}, {15, 0, 6, 0}, + {18, 1, 5, 32}, {20, 1, 5, 32}, {24, 2, 5, 32}, + {28, 2, 5, 32}, {40, 3, 5, 32}, {48, 4, 5, 32}, + {65536, 16, 6, 0}, {32768, 15, 6, 0}, {16384, 14, 6, 0}, + {8192, 13, 6, 0}, +} + +// predefinedOffsetTable is the predefined table to use for offsets. +// Generated from table in RFC 3.1.1.3.2.2.3. +// Checked by TestPredefinedTables. +var predefinedOffsetTable = [...]fseBaselineEntry{ + {1, 0, 5, 0}, {61, 6, 4, 0}, {509, 9, 5, 0}, + {32765, 15, 5, 0}, {2097149, 21, 5, 0}, {5, 3, 5, 0}, + {125, 7, 4, 0}, {4093, 12, 5, 0}, {262141, 18, 5, 0}, + {8388605, 23, 5, 0}, {29, 5, 5, 0}, {253, 8, 4, 0}, + {16381, 14, 5, 0}, {1048573, 20, 5, 0}, {1, 2, 5, 0}, + {125, 7, 4, 16}, {2045, 11, 5, 0}, {131069, 17, 5, 0}, + {4194301, 22, 5, 0}, {13, 4, 5, 0}, {253, 8, 4, 16}, + {8189, 13, 5, 0}, {524285, 19, 5, 0}, {2, 1, 5, 0}, + {61, 6, 4, 16}, {1021, 10, 5, 0}, {65533, 16, 5, 0}, + {268435453, 28, 5, 0}, {134217725, 27, 5, 0}, {67108861, 26, 5, 0}, + {33554429, 25, 5, 0}, {16777213, 24, 5, 0}, +} + +// predefinedMatchTable is the predefined table to use for match lengths. +// Generated from table in RFC 3.1.1.3.2.2.2. +// Checked by TestPredefinedTables. +var predefinedMatchTable = [...]fseBaselineEntry{ + {3, 0, 6, 0}, {4, 0, 4, 0}, {5, 0, 5, 32}, + {6, 0, 5, 0}, {8, 0, 5, 0}, {9, 0, 5, 0}, + {11, 0, 5, 0}, {13, 0, 6, 0}, {16, 0, 6, 0}, + {19, 0, 6, 0}, {22, 0, 6, 0}, {25, 0, 6, 0}, + {28, 0, 6, 0}, {31, 0, 6, 0}, {34, 0, 6, 0}, + {37, 1, 6, 0}, {41, 1, 6, 0}, {47, 2, 6, 0}, + {59, 3, 6, 0}, {83, 4, 6, 0}, {131, 7, 6, 0}, + {515, 9, 6, 0}, {4, 0, 4, 16}, {5, 0, 4, 0}, + {6, 0, 5, 32}, {7, 0, 5, 0}, {9, 0, 5, 32}, + {10, 0, 5, 0}, {12, 0, 6, 0}, {15, 0, 6, 0}, + {18, 0, 6, 0}, {21, 0, 6, 0}, {24, 0, 6, 0}, + {27, 0, 6, 0}, {30, 0, 6, 0}, {33, 0, 6, 0}, + {35, 1, 6, 0}, {39, 1, 6, 0}, {43, 2, 6, 0}, + {51, 3, 6, 0}, {67, 4, 6, 0}, {99, 5, 6, 0}, + {259, 8, 6, 0}, {4, 0, 4, 32}, {4, 0, 4, 48}, + {5, 0, 4, 16}, {7, 0, 5, 32}, {8, 0, 5, 32}, + {10, 0, 5, 32}, {11, 0, 5, 32}, {14, 0, 6, 0}, + {17, 0, 6, 0}, {20, 0, 6, 0}, {23, 0, 6, 0}, + {26, 0, 6, 0}, {29, 0, 6, 0}, {32, 0, 6, 0}, + {65539, 16, 6, 0}, {32771, 15, 6, 0}, {16387, 14, 6, 0}, + {8195, 13, 6, 0}, {4099, 12, 6, 0}, {2051, 11, 6, 0}, + {1027, 10, 6, 0}, +} diff --git a/src/internal/zstd/fse_test.go b/src/internal/zstd/fse_test.go new file mode 100644 index 0000000..6f106b6 --- /dev/null +++ b/src/internal/zstd/fse_test.go @@ -0,0 +1,89 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "slices" + "testing" +) + +// literalPredefinedDistribution is the predefined distribution table +// for literal lengths. RFC 3.1.1.3.2.2.1. +var literalPredefinedDistribution = []int16{ + 4, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1, + 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 2, 1, 1, 1, 1, 1, + -1, -1, -1, -1, +} + +// offsetPredefinedDistribution is the predefined distribution table +// for offsets. RFC 3.1.1.3.2.2.3. +var offsetPredefinedDistribution = []int16{ + 1, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, -1, -1, -1, -1, -1, +} + +// matchPredefinedDistribution is the predefined distribution table +// for match lengths. RFC 3.1.1.3.2.2.2. +var matchPredefinedDistribution = []int16{ + 1, 4, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, -1, -1, + -1, -1, -1, -1, -1, +} + +// TestPredefinedTables verifies that we can generate the predefined +// literal/offset/match tables from the input data in RFC 8878. +// This serves as a test of the predefined tables, and also of buildFSE +// and the functions that make baseline FSE tables. +func TestPredefinedTables(t *testing.T) { + tests := []struct { + name string + distribution []int16 + tableBits int + toBaseline func(*Reader, int, []fseEntry, []fseBaselineEntry) error + predef []fseBaselineEntry + }{ + { + name: "literal", + distribution: literalPredefinedDistribution, + tableBits: 6, + toBaseline: (*Reader).makeLiteralBaselineFSE, + predef: predefinedLiteralTable[:], + }, + { + name: "offset", + distribution: offsetPredefinedDistribution, + tableBits: 5, + toBaseline: (*Reader).makeOffsetBaselineFSE, + predef: predefinedOffsetTable[:], + }, + { + name: "match", + distribution: matchPredefinedDistribution, + tableBits: 6, + toBaseline: (*Reader).makeMatchBaselineFSE, + predef: predefinedMatchTable[:], + }, + } + for _, test := range tests { + test := test + t.Run(test.name, func(t *testing.T) { + var r Reader + table := make([]fseEntry, 1<<test.tableBits) + if err := r.buildFSE(0, test.distribution, table, test.tableBits); err != nil { + t.Fatal(err) + } + + baselineTable := make([]fseBaselineEntry, len(table)) + if err := test.toBaseline(&r, 0, table, baselineTable); err != nil { + t.Fatal(err) + } + + if !slices.Equal(baselineTable, test.predef) { + t.Errorf("got %v, want %v", baselineTable, test.predef) + } + }) + } +} diff --git a/src/internal/zstd/fuzz_test.go b/src/internal/zstd/fuzz_test.go new file mode 100644 index 0000000..bb6f0a9 --- /dev/null +++ b/src/internal/zstd/fuzz_test.go @@ -0,0 +1,140 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "bytes" + "io" + "os" + "os/exec" + "testing" +) + +// badStrings is some inputs that FuzzReader failed on earlier. +var badStrings = []string{ + "(\xb5/\xfdd00,\x05\x00\xc4\x0400000000000000000000000000000000000000000000000000000000000000000000000000000 \xa07100000000000000000000000000000000000000000000000000000000000000000000000000aM\x8a2y0B\b", + "(\xb5/\xfd00$\x05\x0020 00X70000a70000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", + "(\xb5/\xfd00$\x05\x0020 00B00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", + "(\xb5/\xfd00}\x00\x0020\x00\x9000000000000", + "(\xb5/\xfd00}\x00\x00&0\x02\x830!000000000", + "(\xb5/\xfd\x1002000$\x05\x0010\xcc0\xa8100000000100000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", + "(\xb5/\xfd\x1002000$\x05\x0000\xcc0\xa8100d\x0000001000000000000000000000000000000000000000000000000000000000000000000000000\x000000000000000000000000000000000000000000000000000000000000000000000000000000", + "(\xb5/\xfd001\x00\x0000000000000000000", +} + +// This is a simple fuzzer to see if the decompressor panics. +func FuzzReader(f *testing.F) { + for _, test := range tests { + f.Add([]byte(test.compressed)) + } + for _, s := range badStrings { + f.Add([]byte(s)) + } + f.Fuzz(func(t *testing.T, b []byte) { + r := NewReader(bytes.NewReader(b)) + io.Copy(io.Discard, r) + }) +} + +// Fuzz test to verify that what we decompress is what we compress. +// This isn't a great fuzz test because the fuzzer can't efficiently +// explore the space of decompressor behavior, since it can't see +// what the compressor is doing. But it's better than nothing. +func FuzzDecompressor(f *testing.F) { + if _, err := os.Stat("/usr/bin/zstd"); err != nil { + f.Skip("skipping because /usr/bin/zstd does not exist") + } + + for _, test := range tests { + f.Add([]byte(test.uncompressed)) + } + + // Add some larger data, as that has more interesting compression. + f.Add(bytes.Repeat([]byte("abcdefghijklmnop"), 256)) + var buf bytes.Buffer + for i := 0; i < 256; i++ { + buf.WriteByte(byte(i)) + } + f.Add(bytes.Repeat(buf.Bytes(), 64)) + f.Add(bigData(f)) + + f.Fuzz(func(t *testing.T, b []byte) { + cmd := exec.Command("/usr/bin/zstd", "-z") + cmd.Stdin = bytes.NewReader(b) + var compressed bytes.Buffer + cmd.Stdout = &compressed + cmd.Stderr = os.Stderr + if err := cmd.Run(); err != nil { + t.Errorf("running zstd failed: %v", err) + } + + r := NewReader(bytes.NewReader(compressed.Bytes())) + got, err := io.ReadAll(r) + if err != nil { + t.Fatal(err) + } + if !bytes.Equal(got, b) { + showDiffs(t, got, b) + } + }) +} + +// Fuzz test to check that if we can decompress some data, +// so can zstd, and that we get the same result. +func FuzzReverse(f *testing.F) { + if _, err := os.Stat("/usr/bin/zstd"); err != nil { + f.Skip("skipping because /usr/bin/zstd does not exist") + } + + for _, test := range tests { + f.Add([]byte(test.compressed)) + } + + // Set a hook to reject some cases where we don't match zstd. + fuzzing = true + defer func() { fuzzing = false }() + + f.Fuzz(func(t *testing.T, b []byte) { + r := NewReader(bytes.NewReader(b)) + goExp, goErr := io.ReadAll(r) + + cmd := exec.Command("/usr/bin/zstd", "-d") + cmd.Stdin = bytes.NewReader(b) + var uncompressed bytes.Buffer + cmd.Stdout = &uncompressed + cmd.Stderr = os.Stderr + zstdErr := cmd.Run() + zstdExp := uncompressed.Bytes() + + if goErr == nil && zstdErr == nil { + if !bytes.Equal(zstdExp, goExp) { + showDiffs(t, zstdExp, goExp) + } + } else { + // Ideally we should check that this package and + // the zstd program both fail or both succeed, + // and that if they both fail one byte sequence + // is an exact prefix of the other. + // Actually trying this proved to be frustrating, + // as the zstd program appears to accept invalid + // byte sequences using rules that are difficult + // to determine. + // So we just check the prefix. + + c := len(goExp) + if c > len(zstdExp) { + c = len(zstdExp) + } + goExp = goExp[:c] + zstdExp = zstdExp[:c] + if !bytes.Equal(goExp, zstdExp) { + t.Error("byte mismatch after error") + t.Logf("Go error: %v\n", goErr) + t.Logf("zstd error: %v\n", zstdErr) + showDiffs(t, zstdExp, goExp) + } + } + }) +} diff --git a/src/internal/zstd/huff.go b/src/internal/zstd/huff.go new file mode 100644 index 0000000..452e24b --- /dev/null +++ b/src/internal/zstd/huff.go @@ -0,0 +1,204 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "io" + "math/bits" +) + +// maxHuffmanBits is the largest possible Huffman table bits. +const maxHuffmanBits = 11 + +// readHuff reads Huffman table from data starting at off into table. +// Each entry in a Huffman table is a pair of bytes. +// The high byte is the encoded value. The low byte is the number +// of bits used to encode that value. We index into the table +// with a value of size tableBits. A value that requires fewer bits +// appear in the table multiple times. +// This returns the number of bits in the Huffman table and the new offset. +// RFC 4.2.1. +func (r *Reader) readHuff(data block, off int, table []uint16) (tableBits, roff int, err error) { + if off >= len(data) { + return 0, 0, r.makeEOFError(off) + } + + hdr := data[off] + off++ + + var weights [256]uint8 + var count int + if hdr < 128 { + // The table is compressed using an FSE. RFC 4.2.1.2. + if len(r.fseScratch) < 1<<6 { + r.fseScratch = make([]fseEntry, 1<<6) + } + fseBits, noff, err := r.readFSE(data, off, 255, 6, r.fseScratch) + if err != nil { + return 0, 0, err + } + fseTable := r.fseScratch + + if off+int(hdr) > len(data) { + return 0, 0, r.makeEOFError(off) + } + + rbr, err := r.makeReverseBitReader(data, off+int(hdr)-1, noff) + if err != nil { + return 0, 0, err + } + + state1, err := rbr.val(uint8(fseBits)) + if err != nil { + return 0, 0, err + } + + state2, err := rbr.val(uint8(fseBits)) + if err != nil { + return 0, 0, err + } + + // There are two independent FSE streams, tracked by + // state1 and state2. We decode them alternately. + + for { + pt := &fseTable[state1] + if !rbr.fetch(pt.bits) { + if count >= 254 { + return 0, 0, rbr.makeError("Huffman count overflow") + } + weights[count] = pt.sym + weights[count+1] = fseTable[state2].sym + count += 2 + break + } + + v, err := rbr.val(pt.bits) + if err != nil { + return 0, 0, err + } + state1 = uint32(pt.base) + v + + if count >= 255 { + return 0, 0, rbr.makeError("Huffman count overflow") + } + + weights[count] = pt.sym + count++ + + pt = &fseTable[state2] + + if !rbr.fetch(pt.bits) { + if count >= 254 { + return 0, 0, rbr.makeError("Huffman count overflow") + } + weights[count] = pt.sym + weights[count+1] = fseTable[state1].sym + count += 2 + break + } + + v, err = rbr.val(pt.bits) + if err != nil { + return 0, 0, err + } + state2 = uint32(pt.base) + v + + if count >= 255 { + return 0, 0, rbr.makeError("Huffman count overflow") + } + + weights[count] = pt.sym + count++ + } + + off += int(hdr) + } else { + // The table is not compressed. Each weight is 4 bits. + + count = int(hdr) - 127 + if off+((count+1)/2) >= len(data) { + return 0, 0, io.ErrUnexpectedEOF + } + for i := 0; i < count; i += 2 { + b := data[off] + off++ + weights[i] = b >> 4 + weights[i+1] = b & 0xf + } + } + + // RFC 4.2.1.3. + + var weightMark [13]uint32 + weightMask := uint32(0) + for _, w := range weights[:count] { + if w > 12 { + return 0, 0, r.makeError(off, "Huffman weight overflow") + } + weightMark[w]++ + if w > 0 { + weightMask += 1 << (w - 1) + } + } + if weightMask == 0 { + return 0, 0, r.makeError(off, "bad Huffman weights") + } + + tableBits = 32 - bits.LeadingZeros32(weightMask) + if tableBits > maxHuffmanBits { + return 0, 0, r.makeError(off, "bad Huffman weights") + } + + if len(table) < 1<<tableBits { + return 0, 0, r.makeError(off, "Huffman table too small") + } + + // Work out the last weight value, which is omitted because + // the weights must sum to a power of two. + left := (uint32(1) << tableBits) - weightMask + if left == 0 { + return 0, 0, r.makeError(off, "bad Huffman weights") + } + highBit := 31 - bits.LeadingZeros32(left) + if uint32(1)<<highBit != left { + return 0, 0, r.makeError(off, "bad Huffman weights") + } + if count >= 256 { + return 0, 0, r.makeError(off, "Huffman weight overflow") + } + weights[count] = uint8(highBit + 1) + count++ + weightMark[highBit+1]++ + + if weightMark[1] < 2 || weightMark[1]&1 != 0 { + return 0, 0, r.makeError(off, "bad Huffman weights") + } + + // Change weightMark from a count of weights to the index of + // the first symbol for that weight. We shift the indexes to + // also store how many we have seen so far, + next := uint32(0) + for i := 0; i < tableBits; i++ { + cur := next + next += weightMark[i+1] << i + weightMark[i+1] = cur + } + + for i, w := range weights[:count] { + if w == 0 { + continue + } + length := uint32(1) << (w - 1) + tval := uint16(i)<<8 | (uint16(tableBits) + 1 - uint16(w)) + start := weightMark[w] + for j := uint32(0); j < length; j++ { + table[start+j] = tval + } + weightMark[w] += length + } + + return tableBits, off, nil +} diff --git a/src/internal/zstd/literals.go b/src/internal/zstd/literals.go new file mode 100644 index 0000000..b46d668 --- /dev/null +++ b/src/internal/zstd/literals.go @@ -0,0 +1,330 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "encoding/binary" +) + +// readLiterals reads and decompresses the literals from data at off. +// The literals are appended to outbuf, which is returned. +// Also returns the new input offset. RFC 3.1.1.3.1. +func (r *Reader) readLiterals(data block, off int, outbuf []byte) (int, []byte, error) { + if off >= len(data) { + return 0, nil, r.makeEOFError(off) + } + + // Literals section header. RFC 3.1.1.3.1.1. + hdr := data[off] + off++ + + if (hdr&3) == 0 || (hdr&3) == 1 { + return r.readRawRLELiterals(data, off, hdr, outbuf) + } else { + return r.readHuffLiterals(data, off, hdr, outbuf) + } +} + +// readRawRLELiterals reads and decompresses a Raw_Literals_Block or +// a RLE_Literals_Block. RFC 3.1.1.3.1.1. +func (r *Reader) readRawRLELiterals(data block, off int, hdr byte, outbuf []byte) (int, []byte, error) { + raw := (hdr & 3) == 0 + + var regeneratedSize int + switch (hdr >> 2) & 3 { + case 0, 2: + regeneratedSize = int(hdr >> 3) + case 1: + if off >= len(data) { + return 0, nil, r.makeEOFError(off) + } + regeneratedSize = int(hdr>>4) + (int(data[off]) << 4) + off++ + case 3: + if off+1 >= len(data) { + return 0, nil, r.makeEOFError(off) + } + regeneratedSize = int(hdr>>4) + (int(data[off]) << 4) + (int(data[off+1]) << 12) + off += 2 + } + + // We are going to use the entire literal block in the output. + // The maximum size of one decompressed block is 128K, + // so we can't have more literals than that. + if regeneratedSize > 128<<10 { + return 0, nil, r.makeError(off, "literal size too large") + } + + if raw { + // RFC 3.1.1.3.1.2. + if off+regeneratedSize > len(data) { + return 0, nil, r.makeError(off, "raw literal size too large") + } + outbuf = append(outbuf, data[off:off+regeneratedSize]...) + off += regeneratedSize + } else { + // RFC 3.1.1.3.1.3. + if off >= len(data) { + return 0, nil, r.makeError(off, "RLE literal missing") + } + rle := data[off] + off++ + for i := 0; i < regeneratedSize; i++ { + outbuf = append(outbuf, rle) + } + } + + return off, outbuf, nil +} + +// readHuffLiterals reads and decompresses a Compressed_Literals_Block or +// a Treeless_Literals_Block. RFC 3.1.1.3.1.4. +func (r *Reader) readHuffLiterals(data block, off int, hdr byte, outbuf []byte) (int, []byte, error) { + var ( + regeneratedSize int + compressedSize int + streams int + ) + switch (hdr >> 2) & 3 { + case 0, 1: + if off+1 >= len(data) { + return 0, nil, r.makeEOFError(off) + } + regeneratedSize = (int(hdr) >> 4) | ((int(data[off]) & 0x3f) << 4) + compressedSize = (int(data[off]) >> 6) | (int(data[off+1]) << 2) + off += 2 + if ((hdr >> 2) & 3) == 0 { + streams = 1 + } else { + streams = 4 + } + case 2: + if off+2 >= len(data) { + return 0, nil, r.makeEOFError(off) + } + regeneratedSize = (int(hdr) >> 4) | (int(data[off]) << 4) | ((int(data[off+1]) & 3) << 12) + compressedSize = (int(data[off+1]) >> 2) | (int(data[off+2]) << 6) + off += 3 + streams = 4 + case 3: + if off+3 >= len(data) { + return 0, nil, r.makeEOFError(off) + } + regeneratedSize = (int(hdr) >> 4) | (int(data[off]) << 4) | ((int(data[off+1]) & 0x3f) << 12) + compressedSize = (int(data[off+1]) >> 6) | (int(data[off+2]) << 2) | (int(data[off+3]) << 10) + off += 4 + streams = 4 + } + + // We are going to use the entire literal block in the output. + // The maximum size of one decompressed block is 128K, + // so we can't have more literals than that. + if regeneratedSize > 128<<10 { + return 0, nil, r.makeError(off, "literal size too large") + } + + roff := off + compressedSize + if roff > len(data) || roff < 0 { + return 0, nil, r.makeEOFError(off) + } + + totalStreamsSize := compressedSize + if (hdr & 3) == 2 { + // Compressed_Literals_Block. + // Read new huffman tree. + + if len(r.huffmanTable) < 1<<maxHuffmanBits { + r.huffmanTable = make([]uint16, 1<<maxHuffmanBits) + } + + huffmanTableBits, hoff, err := r.readHuff(data, off, r.huffmanTable) + if err != nil { + return 0, nil, err + } + r.huffmanTableBits = huffmanTableBits + + if totalStreamsSize < hoff-off { + return 0, nil, r.makeError(off, "Huffman table too big") + } + totalStreamsSize -= hoff - off + off = hoff + } else { + // Treeless_Literals_Block + // Reuse previous Huffman tree. + if r.huffmanTableBits == 0 { + return 0, nil, r.makeError(off, "missing literals Huffman tree") + } + } + + // Decompress compressedSize bytes of data at off using the + // Huffman tree. + + var err error + if streams == 1 { + outbuf, err = r.readLiteralsOneStream(data, off, totalStreamsSize, regeneratedSize, outbuf) + } else { + outbuf, err = r.readLiteralsFourStreams(data, off, totalStreamsSize, regeneratedSize, outbuf) + } + + if err != nil { + return 0, nil, err + } + + return roff, outbuf, nil +} + +// readLiteralsOneStream reads a single stream of compressed literals. +func (r *Reader) readLiteralsOneStream(data block, off, compressedSize, regeneratedSize int, outbuf []byte) ([]byte, error) { + // We let the reverse bit reader read earlier bytes, + // because the Huffman table ignores bits that it doesn't need. + rbr, err := r.makeReverseBitReader(data, off+compressedSize-1, off-2) + if err != nil { + return nil, err + } + + huffTable := r.huffmanTable + huffBits := uint32(r.huffmanTableBits) + huffMask := (uint32(1) << huffBits) - 1 + + for i := 0; i < regeneratedSize; i++ { + if !rbr.fetch(uint8(huffBits)) { + return nil, rbr.makeError("literals Huffman stream out of bits") + } + + var t uint16 + idx := (rbr.bits >> (rbr.cnt - huffBits)) & huffMask + t = huffTable[idx] + outbuf = append(outbuf, byte(t>>8)) + rbr.cnt -= uint32(t & 0xff) + } + + return outbuf, nil +} + +// readLiteralsFourStreams reads four interleaved streams of +// compressed literals. +func (r *Reader) readLiteralsFourStreams(data block, off, totalStreamsSize, regeneratedSize int, outbuf []byte) ([]byte, error) { + // Read the jump table to find out where the streams are. + // RFC 3.1.1.3.1.6. + if off+5 >= len(data) { + return nil, r.makeEOFError(off) + } + if totalStreamsSize < 6 { + return nil, r.makeError(off, "total streams size too small for jump table") + } + + streamSize1 := binary.LittleEndian.Uint16(data[off:]) + streamSize2 := binary.LittleEndian.Uint16(data[off+2:]) + streamSize3 := binary.LittleEndian.Uint16(data[off+4:]) + off += 6 + + tot := uint64(streamSize1) + uint64(streamSize2) + uint64(streamSize3) + if tot > uint64(totalStreamsSize)-6 { + return nil, r.makeEOFError(off) + } + streamSize4 := uint32(totalStreamsSize) - 6 - uint32(tot) + + off-- + off1 := off + int(streamSize1) + start1 := off + 1 + + off2 := off1 + int(streamSize2) + start2 := off1 + 1 + + off3 := off2 + int(streamSize3) + start3 := off2 + 1 + + off4 := off3 + int(streamSize4) + start4 := off3 + 1 + + // We let the reverse bit readers read earlier bytes, + // because the Huffman tables ignore bits that they don't need. + + rbr1, err := r.makeReverseBitReader(data, off1, start1-2) + if err != nil { + return nil, err + } + + rbr2, err := r.makeReverseBitReader(data, off2, start2-2) + if err != nil { + return nil, err + } + + rbr3, err := r.makeReverseBitReader(data, off3, start3-2) + if err != nil { + return nil, err + } + + rbr4, err := r.makeReverseBitReader(data, off4, start4-2) + if err != nil { + return nil, err + } + + regeneratedStreamSize := (regeneratedSize + 3) / 4 + + out1 := len(outbuf) + out2 := out1 + regeneratedStreamSize + out3 := out2 + regeneratedStreamSize + out4 := out3 + regeneratedStreamSize + + regeneratedStreamSize4 := regeneratedSize - regeneratedStreamSize*3 + + outbuf = append(outbuf, make([]byte, regeneratedSize)...) + + huffTable := r.huffmanTable + huffBits := uint32(r.huffmanTableBits) + huffMask := (uint32(1) << huffBits) - 1 + + for i := 0; i < regeneratedStreamSize; i++ { + use4 := i < regeneratedStreamSize4 + + fetchHuff := func(rbr *reverseBitReader) (uint16, error) { + if !rbr.fetch(uint8(huffBits)) { + return 0, rbr.makeError("literals Huffman stream out of bits") + } + idx := (rbr.bits >> (rbr.cnt - huffBits)) & huffMask + return huffTable[idx], nil + } + + t1, err := fetchHuff(&rbr1) + if err != nil { + return nil, err + } + + t2, err := fetchHuff(&rbr2) + if err != nil { + return nil, err + } + + t3, err := fetchHuff(&rbr3) + if err != nil { + return nil, err + } + + if use4 { + t4, err := fetchHuff(&rbr4) + if err != nil { + return nil, err + } + outbuf[out4] = byte(t4 >> 8) + out4++ + rbr4.cnt -= uint32(t4 & 0xff) + } + + outbuf[out1] = byte(t1 >> 8) + out1++ + rbr1.cnt -= uint32(t1 & 0xff) + + outbuf[out2] = byte(t2 >> 8) + out2++ + rbr2.cnt -= uint32(t2 & 0xff) + + outbuf[out3] = byte(t3 >> 8) + out3++ + rbr3.cnt -= uint32(t3 & 0xff) + } + + return outbuf, nil +} diff --git a/src/internal/zstd/xxhash.go b/src/internal/zstd/xxhash.go new file mode 100644 index 0000000..4d579ee --- /dev/null +++ b/src/internal/zstd/xxhash.go @@ -0,0 +1,148 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "encoding/binary" + "math/bits" +) + +const ( + xxhPrime64c1 = 0x9e3779b185ebca87 + xxhPrime64c2 = 0xc2b2ae3d27d4eb4f + xxhPrime64c3 = 0x165667b19e3779f9 + xxhPrime64c4 = 0x85ebca77c2b2ae63 + xxhPrime64c5 = 0x27d4eb2f165667c5 +) + +// xxhash64 is the state of a xxHash-64 checksum. +type xxhash64 struct { + len uint64 // total length hashed + v [4]uint64 // accumulators + buf [32]byte // buffer + cnt int // number of bytes in buffer +} + +// reset discards the current state and prepares to compute a new hash. +// We assume a seed of 0 since that is what zstd uses. +func (xh *xxhash64) reset() { + xh.len = 0 + + // Separate addition for awkward constant overflow. + xh.v[0] = xxhPrime64c1 + xh.v[0] += xxhPrime64c2 + + xh.v[1] = xxhPrime64c2 + xh.v[2] = 0 + + // Separate negation for awkward constant overflow. + xh.v[3] = xxhPrime64c1 + xh.v[3] = -xh.v[3] + + for i := range xh.buf { + xh.buf[i] = 0 + } + xh.cnt = 0 +} + +// update adds a buffer to the has. +func (xh *xxhash64) update(b []byte) { + xh.len += uint64(len(b)) + + if xh.cnt+len(b) < len(xh.buf) { + copy(xh.buf[xh.cnt:], b) + xh.cnt += len(b) + return + } + + if xh.cnt > 0 { + n := copy(xh.buf[xh.cnt:], b) + b = b[n:] + xh.v[0] = xh.round(xh.v[0], binary.LittleEndian.Uint64(xh.buf[:])) + xh.v[1] = xh.round(xh.v[1], binary.LittleEndian.Uint64(xh.buf[8:])) + xh.v[2] = xh.round(xh.v[2], binary.LittleEndian.Uint64(xh.buf[16:])) + xh.v[3] = xh.round(xh.v[3], binary.LittleEndian.Uint64(xh.buf[24:])) + xh.cnt = 0 + } + + for len(b) >= 32 { + xh.v[0] = xh.round(xh.v[0], binary.LittleEndian.Uint64(b)) + xh.v[1] = xh.round(xh.v[1], binary.LittleEndian.Uint64(b[8:])) + xh.v[2] = xh.round(xh.v[2], binary.LittleEndian.Uint64(b[16:])) + xh.v[3] = xh.round(xh.v[3], binary.LittleEndian.Uint64(b[24:])) + b = b[32:] + } + + if len(b) > 0 { + copy(xh.buf[:], b) + xh.cnt = len(b) + } +} + +// digest returns the final hash value. +func (xh *xxhash64) digest() uint64 { + var h64 uint64 + if xh.len < 32 { + h64 = xh.v[2] + xxhPrime64c5 + } else { + h64 = bits.RotateLeft64(xh.v[0], 1) + + bits.RotateLeft64(xh.v[1], 7) + + bits.RotateLeft64(xh.v[2], 12) + + bits.RotateLeft64(xh.v[3], 18) + h64 = xh.mergeRound(h64, xh.v[0]) + h64 = xh.mergeRound(h64, xh.v[1]) + h64 = xh.mergeRound(h64, xh.v[2]) + h64 = xh.mergeRound(h64, xh.v[3]) + } + + h64 += xh.len + + len := xh.len + len &= 31 + buf := xh.buf[:] + for len >= 8 { + k1 := xh.round(0, binary.LittleEndian.Uint64(buf)) + buf = buf[8:] + h64 ^= k1 + h64 = bits.RotateLeft64(h64, 27)*xxhPrime64c1 + xxhPrime64c4 + len -= 8 + } + if len >= 4 { + h64 ^= uint64(binary.LittleEndian.Uint32(buf)) * xxhPrime64c1 + buf = buf[4:] + h64 = bits.RotateLeft64(h64, 23)*xxhPrime64c2 + xxhPrime64c3 + len -= 4 + } + for len > 0 { + h64 ^= uint64(buf[0]) * xxhPrime64c5 + buf = buf[1:] + h64 = bits.RotateLeft64(h64, 11) * xxhPrime64c1 + len-- + } + + h64 ^= h64 >> 33 + h64 *= xxhPrime64c2 + h64 ^= h64 >> 29 + h64 *= xxhPrime64c3 + h64 ^= h64 >> 32 + + return h64 +} + +// round updates a value. +func (xh *xxhash64) round(v, n uint64) uint64 { + v += n * xxhPrime64c2 + v = bits.RotateLeft64(v, 31) + v *= xxhPrime64c1 + return v +} + +// mergeRound updates a value in the final round. +func (xh *xxhash64) mergeRound(v, n uint64) uint64 { + n = xh.round(0, n) + v ^= n + v = v*xxhPrime64c1 + xxhPrime64c4 + return v +} diff --git a/src/internal/zstd/xxhash_test.go b/src/internal/zstd/xxhash_test.go new file mode 100644 index 0000000..646cee8 --- /dev/null +++ b/src/internal/zstd/xxhash_test.go @@ -0,0 +1,105 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "bytes" + "os" + "os/exec" + "strconv" + "testing" +) + +var xxHashTests = []struct { + data string + hash uint64 +}{ + { + "hello, world", + 0xb33a384e6d1b1242, + }, + { + "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789$", + 0x1032d841e824f998, + }, +} + +func TestXXHash(t *testing.T) { + var xh xxhash64 + for i, test := range xxHashTests { + xh.reset() + xh.update([]byte(test.data)) + if got := xh.digest(); got != test.hash { + t.Errorf("#%d: got %#x want %#x", i, got, test.hash) + } + } +} + +func TestLargeXXHash(t *testing.T) { + if testing.Short() { + t.Skip("skipping expensive test in short mode") + } + + data := bigData(t) + var xh xxhash64 + xh.reset() + i := 0 + for i < len(data) { + // Write varying amounts to test buffering. + c := i%4094 + 1 + if i+c > len(data) { + c = len(data) - i + } + xh.update(data[i : i+c]) + i += c + } + + got := xh.digest() + want := uint64(0xf0dd39fd7e063f82) + if got != want { + t.Errorf("got %#x want %#x", got, want) + } +} + +func FuzzXXHash(f *testing.F) { + if _, err := os.Stat("/usr/bin/xxhsum"); err != nil { + f.Skip("skipping because /usr/bin/xxhsum does not exist") + } + + for _, test := range xxHashTests { + f.Add([]byte(test.data)) + } + f.Add(bytes.Repeat([]byte("abcdefghijklmnop"), 256)) + var buf bytes.Buffer + for i := 0; i < 256; i++ { + buf.WriteByte(byte(i)) + } + f.Add(bytes.Repeat(buf.Bytes(), 64)) + f.Add(bigData(f)) + + f.Fuzz(func(t *testing.T, b []byte) { + cmd := exec.Command("/usr/bin/xxhsum", "-H64") + cmd.Stdin = bytes.NewReader(b) + var hhsumHash bytes.Buffer + cmd.Stdout = &hhsumHash + if err := cmd.Run(); err != nil { + t.Fatalf("running hhsum failed: %v", err) + } + hhHashBytes := bytes.Fields(bytes.TrimSpace(hhsumHash.Bytes()))[0] + hhHash, err := strconv.ParseUint(string(hhHashBytes), 16, 64) + if err != nil { + t.Fatalf("could not parse hash %q: %v", hhHashBytes, err) + } + + var xh xxhash64 + xh.reset() + xh.update(b) + goHash := xh.digest() + + if goHash != hhHash { + t.Errorf("Go hash %#x != xxhsum hash %#x", goHash, hhHash) + } + }) +} diff --git a/src/internal/zstd/zstd.go b/src/internal/zstd/zstd.go new file mode 100644 index 0000000..a860789 --- /dev/null +++ b/src/internal/zstd/zstd.go @@ -0,0 +1,508 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +// Package zstd provides a decompressor for zstd streams, +// described in RFC 8878. It does not support dictionaries. +package zstd + +import ( + "encoding/binary" + "errors" + "fmt" + "io" +) + +// fuzzing is a fuzzer hook set to true when fuzzing. +// This is used to reject cases where we don't match zstd. +var fuzzing = false + +// Reader implements [io.Reader] to read a zstd compressed stream. +type Reader struct { + // The underlying Reader. + r io.Reader + + // Whether we have read the frame header. + // This is of interest when buffer is empty. + // If true we expect to see a new block. + sawFrameHeader bool + + // Whether the current frame expects a checksum. + hasChecksum bool + + // Whether we have read at least one frame. + readOneFrame bool + + // True if the frame size is not known. + frameSizeUnknown bool + + // The number of uncompressed bytes remaining in the current frame. + // If frameSizeUnknown is true, this is not valid. + remainingFrameSize uint64 + + // The number of bytes read from r up to the start of the current + // block, for error reporting. + blockOffset int64 + + // Buffered decompressed data. + buffer []byte + // Current read offset in buffer. + off int + + // The current repeated offsets. + repeatedOffset1 uint32 + repeatedOffset2 uint32 + repeatedOffset3 uint32 + + // The current Huffman tree used for compressing literals. + huffmanTable []uint16 + huffmanTableBits int + + // The window for back references. + windowSize int // maximum required window size + window []byte // window data + + // A buffer available to hold a compressed block. + compressedBuf []byte + + // A buffer for literals. + literals []byte + + // Sequence decode FSE tables. + seqTables [3][]fseBaselineEntry + seqTableBits [3]uint8 + + // Buffers for sequence decode FSE tables. + seqTableBuffers [3][]fseBaselineEntry + + // Scratch space used for small reads, to avoid allocation. + scratch [16]byte + + // A scratch table for reading an FSE. Only temporarily valid. + fseScratch []fseEntry + + // For checksum computation. + checksum xxhash64 +} + +// NewReader creates a new Reader that decompresses data from the given reader. +func NewReader(input io.Reader) *Reader { + r := new(Reader) + r.Reset(input) + return r +} + +// Reset discards the current state and starts reading a new stream from r. +// This permits reusing a Reader rather than allocating a new one. +func (r *Reader) Reset(input io.Reader) { + r.r = input + + // Several fields are preserved to avoid allocation. + // Others are always set before they are used. + r.sawFrameHeader = false + r.hasChecksum = false + r.readOneFrame = false + r.frameSizeUnknown = false + r.remainingFrameSize = 0 + r.blockOffset = 0 + // buffer + r.off = 0 + // repeatedOffset1 + // repeatedOffset2 + // repeatedOffset3 + // huffmanTable + // huffmanTableBits + // windowSize + // window + // compressedBuf + // literals + // seqTables + // seqTableBits + // seqTableBuffers + // scratch + // fseScratch +} + +// Read implements [io.Reader]. +func (r *Reader) Read(p []byte) (int, error) { + if err := r.refillIfNeeded(); err != nil { + return 0, err + } + n := copy(p, r.buffer[r.off:]) + r.off += n + return n, nil +} + +// ReadByte implements [io.ByteReader]. +func (r *Reader) ReadByte() (byte, error) { + if err := r.refillIfNeeded(); err != nil { + return 0, err + } + ret := r.buffer[r.off] + r.off++ + return ret, nil +} + +// refillIfNeeded reads the next block if necessary. +func (r *Reader) refillIfNeeded() error { + for r.off >= len(r.buffer) { + if err := r.refill(); err != nil { + return err + } + r.off = 0 + } + return nil +} + +// refill reads and decompresses the next block. +func (r *Reader) refill() error { + if !r.sawFrameHeader { + if err := r.readFrameHeader(); err != nil { + return err + } + } + return r.readBlock() +} + +// readFrameHeader reads the frame header and prepares to read a block. +func (r *Reader) readFrameHeader() error { +retry: + relativeOffset := 0 + + // Read magic number. RFC 3.1.1. + if _, err := io.ReadFull(r.r, r.scratch[:4]); err != nil { + // We require that the stream contain at least one frame. + if err == io.EOF && !r.readOneFrame { + err = io.ErrUnexpectedEOF + } + return r.wrapError(relativeOffset, err) + } + + if magic := binary.LittleEndian.Uint32(r.scratch[:4]); magic != 0xfd2fb528 { + if magic >= 0x184d2a50 && magic <= 0x184d2a5f { + // This is a skippable frame. + r.blockOffset += int64(relativeOffset) + 4 + if err := r.skipFrame(); err != nil { + return err + } + goto retry + } + + return r.makeError(relativeOffset, "invalid magic number") + } + + relativeOffset += 4 + + // Read Frame_Header_Descriptor. RFC 3.1.1.1.1. + if _, err := io.ReadFull(r.r, r.scratch[:1]); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + descriptor := r.scratch[0] + + singleSegment := descriptor&(1<<5) != 0 + + fcsFieldSize := 1 << (descriptor >> 6) + if fcsFieldSize == 1 && !singleSegment { + fcsFieldSize = 0 + } + + var windowDescriptorSize int + if singleSegment { + windowDescriptorSize = 0 + } else { + windowDescriptorSize = 1 + } + + if descriptor&(1<<3) != 0 { + return r.makeError(relativeOffset, "reserved bit set in frame header descriptor") + } + + r.hasChecksum = descriptor&(1<<2) != 0 + if r.hasChecksum { + r.checksum.reset() + } + + if descriptor&3 != 0 { + return r.makeError(relativeOffset, "dictionaries are not supported") + } + + relativeOffset++ + + headerSize := windowDescriptorSize + fcsFieldSize + + if _, err := io.ReadFull(r.r, r.scratch[:headerSize]); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + + // Figure out the maximum amount of data we need to retain + // for backreferences. + + if singleSegment { + // No window required, as all the data is in a single buffer. + r.windowSize = 0 + } else { + // Window descriptor. RFC 3.1.1.1.2. + windowDescriptor := r.scratch[0] + exponent := uint64(windowDescriptor >> 3) + mantissa := uint64(windowDescriptor & 7) + windowLog := exponent + 10 + windowBase := uint64(1) << windowLog + windowAdd := (windowBase / 8) * mantissa + windowSize := windowBase + windowAdd + + // Default zstd sets limits on the window size. + if fuzzing && (windowLog > 31 || windowSize > 1<<27) { + return r.makeError(relativeOffset, "windowSize too large") + } + + // RFC 8878 permits us to set an 8M max on window size. + if windowSize > 8<<20 { + windowSize = 8 << 20 + } + + r.windowSize = int(windowSize) + } + + // Frame_Content_Size. RFC 3.1.1.4. + r.frameSizeUnknown = false + r.remainingFrameSize = 0 + fb := r.scratch[windowDescriptorSize:] + switch fcsFieldSize { + case 0: + r.frameSizeUnknown = true + case 1: + r.remainingFrameSize = uint64(fb[0]) + case 2: + r.remainingFrameSize = 256 + uint64(binary.LittleEndian.Uint16(fb)) + case 4: + r.remainingFrameSize = uint64(binary.LittleEndian.Uint32(fb)) + case 8: + r.remainingFrameSize = binary.LittleEndian.Uint64(fb) + default: + panic("unreachable") + } + + relativeOffset += headerSize + + r.sawFrameHeader = true + r.readOneFrame = true + r.blockOffset += int64(relativeOffset) + + // Prepare to read blocks from the frame. + r.repeatedOffset1 = 1 + r.repeatedOffset2 = 4 + r.repeatedOffset3 = 8 + r.huffmanTableBits = 0 + r.window = r.window[:0] + r.seqTables[0] = nil + r.seqTables[1] = nil + r.seqTables[2] = nil + + return nil +} + +// skipFrame skips a skippable frame. RFC 3.1.2. +func (r *Reader) skipFrame() error { + relativeOffset := 0 + + if _, err := io.ReadFull(r.r, r.scratch[:4]); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + + relativeOffset += 4 + + size := binary.LittleEndian.Uint32(r.scratch[:4]) + + if seeker, ok := r.r.(io.Seeker); ok { + if _, err := seeker.Seek(int64(size), io.SeekCurrent); err != nil { + return err + } + r.blockOffset += int64(relativeOffset) + int64(size) + return nil + } + + var skip []byte + const chunk = 1 << 20 // 1M + for size >= chunk { + if len(skip) == 0 { + skip = make([]byte, chunk) + } + if _, err := io.ReadFull(r.r, skip); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + relativeOffset += chunk + size -= chunk + } + if size > 0 { + if len(skip) == 0 { + skip = make([]byte, size) + } + if _, err := io.ReadFull(r.r, skip); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + relativeOffset += int(size) + } + + r.blockOffset += int64(relativeOffset) + + return nil +} + +// readBlock reads the next block from a frame. +func (r *Reader) readBlock() error { + relativeOffset := 0 + + // Read Block_Header. RFC 3.1.1.2. + if _, err := io.ReadFull(r.r, r.scratch[:3]); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + + relativeOffset += 3 + + header := uint32(r.scratch[0]) | (uint32(r.scratch[1]) << 8) | (uint32(r.scratch[2]) << 16) + + lastBlock := header&1 != 0 + blockType := (header >> 1) & 3 + blockSize := int(header >> 3) + + // Maximum block size is smaller of window size and 128K. + // We don't record the window size for a single segment frame, + // so just use 128K. RFC 3.1.1.2.3, 3.1.1.2.4. + if blockSize > 128<<10 || (r.windowSize > 0 && blockSize > r.windowSize) { + return r.makeError(relativeOffset, "block size too large") + } + + // Handle different block types. RFC 3.1.1.2.2. + switch blockType { + case 0: + r.setBufferSize(blockSize) + if _, err := io.ReadFull(r.r, r.buffer); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + relativeOffset += blockSize + r.blockOffset += int64(relativeOffset) + case 1: + r.setBufferSize(blockSize) + if _, err := io.ReadFull(r.r, r.scratch[:1]); err != nil { + return r.wrapNonEOFError(relativeOffset, err) + } + relativeOffset++ + v := r.scratch[0] + for i := range r.buffer { + r.buffer[i] = v + } + r.blockOffset += int64(relativeOffset) + case 2: + r.blockOffset += int64(relativeOffset) + if err := r.compressedBlock(blockSize); err != nil { + return err + } + r.blockOffset += int64(blockSize) + case 3: + return r.makeError(relativeOffset, "invalid block type") + } + + if !r.frameSizeUnknown { + if uint64(len(r.buffer)) > r.remainingFrameSize { + return r.makeError(relativeOffset, "too many uncompressed bytes in frame") + } + r.remainingFrameSize -= uint64(len(r.buffer)) + } + + if r.hasChecksum { + r.checksum.update(r.buffer) + } + + if !lastBlock { + r.saveWindow(r.buffer) + } else { + if !r.frameSizeUnknown && r.remainingFrameSize != 0 { + return r.makeError(relativeOffset, "not enough uncompressed bytes for frame") + } + // Check for checksum at end of frame. RFC 3.1.1. + if r.hasChecksum { + if _, err := io.ReadFull(r.r, r.scratch[:4]); err != nil { + return r.wrapNonEOFError(0, err) + } + + inputChecksum := binary.LittleEndian.Uint32(r.scratch[:4]) + dataChecksum := uint32(r.checksum.digest()) + if inputChecksum != dataChecksum { + return r.wrapError(0, fmt.Errorf("invalid checksum: got %#x want %#x", dataChecksum, inputChecksum)) + } + + r.blockOffset += 4 + } + r.sawFrameHeader = false + } + + return nil +} + +// setBufferSize sets the decompressed buffer size. +// When this is called the buffer is empty. +func (r *Reader) setBufferSize(size int) { + if cap(r.buffer) < size { + need := size - cap(r.buffer) + r.buffer = append(r.buffer[:cap(r.buffer)], make([]byte, need)...) + } + r.buffer = r.buffer[:size] +} + +// saveWindow saves bytes in the backreference window. +// TODO: use a circular buffer for less data movement. +func (r *Reader) saveWindow(buf []byte) { + if r.windowSize == 0 { + return + } + + if len(buf) >= r.windowSize { + from := len(buf) - r.windowSize + r.window = append(r.window[:0], buf[from:]...) + return + } + + keep := r.windowSize - len(buf) // must be positive + if keep < len(r.window) { + remove := len(r.window) - keep + copy(r.window[:], r.window[remove:]) + } + + r.window = append(r.window, buf...) +} + +// zstdError is an error while decompressing. +type zstdError struct { + offset int64 + err error +} + +func (ze *zstdError) Error() string { + return fmt.Sprintf("zstd decompression error at %d: %v", ze.offset, ze.err) +} + +func (ze *zstdError) Unwrap() error { + return ze.err +} + +func (r *Reader) makeEOFError(off int) error { + return r.wrapError(off, io.ErrUnexpectedEOF) +} + +func (r *Reader) wrapNonEOFError(off int, err error) error { + if err == io.EOF { + err = io.ErrUnexpectedEOF + } + return r.wrapError(off, err) +} + +func (r *Reader) makeError(off int, msg string) error { + return r.wrapError(off, errors.New(msg)) +} + +func (r *Reader) wrapError(off int, err error) error { + if err == io.EOF { + return err + } + return &zstdError{r.blockOffset + int64(off), err} +} diff --git a/src/internal/zstd/zstd_test.go b/src/internal/zstd/zstd_test.go new file mode 100644 index 0000000..bc75e0f --- /dev/null +++ b/src/internal/zstd/zstd_test.go @@ -0,0 +1,249 @@ +// Copyright 2023 The Go Authors. All rights reserved. +// Use of this source code is governed by a BSD-style +// license that can be found in the LICENSE file. + +package zstd + +import ( + "bytes" + "fmt" + "internal/race" + "internal/testenv" + "io" + "os" + "os/exec" + "strings" + "sync" + "testing" +) + +// tests holds some simple test cases, including some found by fuzzing. +var tests = []struct { + name, uncompressed, compressed string +}{ + { + "hello", + "hello, world\n", + "\x28\xb5\x2f\xfd\x24\x0d\x69\x00\x00\x68\x65\x6c\x6c\x6f\x2c\x20\x77\x6f\x72\x6c\x64\x0a\x4c\x1f\xf9\xf1", + }, + { + // a small compressed .debug_ranges section. + "ranges", + "\xcc\x11\x00\x00\x00\x00\x00\x00\xd5\x13\x00\x00\x00\x00\x00\x00" + + "\x1c\x14\x00\x00\x00\x00\x00\x00\x72\x14\x00\x00\x00\x00\x00\x00" + + "\x9d\x14\x00\x00\x00\x00\x00\x00\xd5\x14\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\xfb\x12\x00\x00\x00\x00\x00\x00\x09\x13\x00\x00\x00\x00\x00\x00" + + "\x0c\x13\x00\x00\x00\x00\x00\x00\xcb\x13\x00\x00\x00\x00\x00\x00" + + "\x29\x14\x00\x00\x00\x00\x00\x00\x4e\x14\x00\x00\x00\x00\x00\x00" + + "\x9d\x14\x00\x00\x00\x00\x00\x00\xd5\x14\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\xfb\x12\x00\x00\x00\x00\x00\x00\x09\x13\x00\x00\x00\x00\x00\x00" + + "\x67\x13\x00\x00\x00\x00\x00\x00\xcb\x13\x00\x00\x00\x00\x00\x00" + + "\x9d\x14\x00\x00\x00\x00\x00\x00\xd5\x14\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\x5f\x0b\x00\x00\x00\x00\x00\x00\x6c\x0b\x00\x00\x00\x00\x00\x00" + + "\x7d\x0b\x00\x00\x00\x00\x00\x00\x7e\x0c\x00\x00\x00\x00\x00\x00" + + "\x38\x0f\x00\x00\x00\x00\x00\x00\x5c\x0f\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\x83\x0c\x00\x00\x00\x00\x00\x00\xfa\x0c\x00\x00\x00\x00\x00\x00" + + "\xfd\x0d\x00\x00\x00\x00\x00\x00\xef\x0e\x00\x00\x00\x00\x00\x00" + + "\x14\x0f\x00\x00\x00\x00\x00\x00\x38\x0f\x00\x00\x00\x00\x00\x00" + + "\x9f\x0f\x00\x00\x00\x00\x00\x00\xac\x0f\x00\x00\x00\x00\x00\x00" + + "\xdb\x0f\x00\x00\x00\x00\x00\x00\xff\x0f\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\xfd\x0d\x00\x00\x00\x00\x00\x00\xd8\x0e\x00\x00\x00\x00\x00\x00" + + "\x9f\x0f\x00\x00\x00\x00\x00\x00\xac\x0f\x00\x00\x00\x00\x00\x00" + + "\xdb\x0f\x00\x00\x00\x00\x00\x00\xff\x0f\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\xfa\x0c\x00\x00\x00\x00\x00\x00\xea\x0d\x00\x00\x00\x00\x00\x00" + + "\xef\x0e\x00\x00\x00\x00\x00\x00\x14\x0f\x00\x00\x00\x00\x00\x00" + + "\x5c\x0f\x00\x00\x00\x00\x00\x00\x9f\x0f\x00\x00\x00\x00\x00\x00" + + "\xac\x0f\x00\x00\x00\x00\x00\x00\xdb\x0f\x00\x00\x00\x00\x00\x00" + + "\xff\x0f\x00\x00\x00\x00\x00\x00\x2c\x10\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\x60\x11\x00\x00\x00\x00\x00\x00\xd1\x16\x00\x00\x00\x00\x00\x00" + + "\x40\x0b\x00\x00\x00\x00\x00\x00\x2c\x10\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\x7a\x00\x00\x00\x00\x00\x00\x00\xb6\x00\x00\x00\x00\x00\x00\x00" + + "\x9f\x01\x00\x00\x00\x00\x00\x00\xa7\x01\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" + + "\x7a\x00\x00\x00\x00\x00\x00\x00\xa9\x00\x00\x00\x00\x00\x00\x00" + + "\x9f\x01\x00\x00\x00\x00\x00\x00\xa7\x01\x00\x00\x00\x00\x00\x00" + + "\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00", + + "\x28\xb5\x2f\xfd\x64\xa0\x01\x2d\x05\x00\xc4\x04\xcc\x11\x00\xd5" + + "\x13\x00\x1c\x14\x00\x72\x9d\xd5\xfb\x12\x00\x09\x0c\x13\xcb\x13" + + "\x29\x4e\x67\x5f\x0b\x6c\x0b\x7d\x0b\x7e\x0c\x38\x0f\x5c\x0f\x83" + + "\x0c\xfa\x0c\xfd\x0d\xef\x0e\x14\x38\x9f\x0f\xac\x0f\xdb\x0f\xff" + + "\x0f\xd8\x9f\xac\xdb\xff\xea\x5c\x2c\x10\x60\xd1\x16\x40\x0b\x7a" + + "\x00\xb6\x00\x9f\x01\xa7\x01\xa9\x36\x20\xa0\x83\x14\x34\x63\x4a" + + "\x21\x70\x8c\x07\x46\x03\x4e\x10\x62\x3c\x06\x4e\xc8\x8c\xb0\x32" + + "\x2a\x59\xad\xb2\xf1\x02\x82\x7c\x33\xcb\x92\x6f\x32\x4f\x9b\xb0" + + "\xa2\x30\xf0\xc0\x06\x1e\x98\x99\x2c\x06\x1e\xd8\xc0\x03\x56\xd8" + + "\xc0\x03\x0f\x6c\xe0\x01\xf1\xf0\xee\x9a\xc6\xc8\x97\x99\xd1\x6c" + + "\xb4\x21\x45\x3b\x10\xe4\x7b\x99\x4d\x8a\x36\x64\x5c\x77\x08\x02" + + "\xcb\xe0\xce", + }, + { + "fuzz1", + "0\x00\x00\x00\x00\x000\x00\x00\x00\x00\x001\x00\x00\x00\x00\x000000", + "(\xb5/\xfd\x04X\x8d\x00\x00P0\x000\x001\x000000\x03T\x02\x00\x01\x01m\xf9\xb7G", + }, +} + +func TestSamples(t *testing.T) { + for _, test := range tests { + test := test + t.Run(test.name, func(t *testing.T) { + r := NewReader(strings.NewReader(test.compressed)) + got, err := io.ReadAll(r) + if err != nil { + t.Fatal(err) + } + gotstr := string(got) + if gotstr != test.uncompressed { + t.Errorf("got %q want %q", gotstr, test.uncompressed) + } + }) + } +} + +var ( + bigDataOnce sync.Once + bigDataBytes []byte + bigDataErr error +) + +// bigData returns the contents of our large test file. +func bigData(t testing.TB) []byte { + bigDataOnce.Do(func() { + bigDataBytes, bigDataErr = os.ReadFile("../../testdata/Isaac.Newton-Opticks.txt") + }) + if bigDataErr != nil { + t.Fatal(bigDataErr) + } + return bigDataBytes +} + +var ( + zstdBigOnce sync.Once + zstdBigBytes []byte + zstdBigSkip bool + zstdBigErr error +) + +// zstdBigData returns the compressed contents of our large test file. +// This will only run on Unix systems with zstd installed. +// That's OK as the package is GOOS-independent. +func zstdBigData(t testing.TB) []byte { + input := bigData(t) + + zstdBigOnce.Do(func() { + if _, err := os.Stat("/usr/bin/zstd"); err != nil { + zstdBigSkip = true + return + } + + cmd := exec.Command("/usr/bin/zstd", "-z") + cmd.Stdin = bytes.NewReader(input) + var compressed bytes.Buffer + cmd.Stdout = &compressed + cmd.Stderr = os.Stderr + if err := cmd.Run(); err != nil { + zstdBigErr = fmt.Errorf("running zstd failed: %v", err) + return + } + + zstdBigBytes = compressed.Bytes() + }) + if zstdBigSkip { + t.Skip("skipping because /usr/bin/zstd does not exist") + } + if zstdBigErr != nil { + t.Fatal(zstdBigErr) + } + return zstdBigBytes +} + +// Test decompressing a large file. We don't have a compressor, +// so this test only runs on systems with zstd installed. +func TestLarge(t *testing.T) { + if testing.Short() { + t.Skip("skipping expensive test in short mode") + } + + data := bigData(t) + compressed := zstdBigData(t) + + t.Logf("/usr/bin/zstd compressed %d bytes to %d", len(data), len(compressed)) + + r := NewReader(bytes.NewReader(compressed)) + got, err := io.ReadAll(r) + if err != nil { + t.Fatal(err) + } + + if !bytes.Equal(got, data) { + showDiffs(t, got, data) + } +} + +// showDiffs reports the first few differences in two []byte. +func showDiffs(t *testing.T, got, want []byte) { + t.Error("data mismatch") + if len(got) != len(want) { + t.Errorf("got data length %d, want %d", len(got), len(want)) + } + diffs := 0 + for i, b := range got { + if i >= len(want) { + break + } + if b != want[i] { + diffs++ + if diffs > 20 { + break + } + t.Logf("%d: %#x != %#x", i, b, want[i]) + } + } +} + +func TestAlloc(t *testing.T) { + testenv.SkipIfOptimizationOff(t) + if race.Enabled { + t.Skip("skipping allocation test under race detector") + } + + compressed := zstdBigData(t) + input := bytes.NewReader(compressed) + r := NewReader(input) + c := testing.AllocsPerRun(10, func() { + input.Reset(compressed) + r.Reset(input) + io.Copy(io.Discard, r) + }) + if c != 0 { + t.Errorf("got %v allocs, want 0", c) + } +} + +func BenchmarkLarge(b *testing.B) { + b.StopTimer() + b.ReportAllocs() + + compressed := zstdBigData(b) + + b.SetBytes(int64(len(compressed))) + + input := bytes.NewReader(compressed) + r := NewReader(input) + + b.StartTimer() + for i := 0; i < b.N; i++ { + input.Reset(compressed) + r.Reset(input) + io.Copy(io.Discard, r) + } +} |