From 218caa410aa38c29984be31a5229b9fa717560ee Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Wed, 17 Apr 2024 14:19:13 +0200 Subject: Merging upstream version 1.68.2+dfsg1. Signed-off-by: Daniel Baumann --- library/alloc/Cargo.toml | 4 +- library/alloc/benches/slice.rs | 4 +- library/alloc/src/alloc.rs | 30 +- library/alloc/src/boxed.rs | 34 +- library/alloc/src/boxed/thin.rs | 47 +- library/alloc/src/collections/binary_heap.rs | 1721 ------------------- library/alloc/src/collections/binary_heap/mod.rs | 1766 ++++++++++++++++++++ library/alloc/src/collections/binary_heap/tests.rs | 121 +- library/alloc/src/collections/btree/map/tests.rs | 6 +- library/alloc/src/collections/btree/mod.rs | 3 - library/alloc/src/collections/btree/node.rs | 28 +- library/alloc/src/collections/btree/set/tests.rs | 4 +- .../src/collections/btree/testing/crash_test.rs | 119 -- library/alloc/src/collections/btree/testing/mod.rs | 3 - .../src/collections/btree/testing/ord_chaos.rs | 81 - library/alloc/src/collections/btree/testing/rng.rs | 28 - library/alloc/src/collections/linked_list/tests.rs | 69 +- .../alloc/src/collections/vec_deque/into_iter.rs | 4 + library/alloc/src/collections/vec_deque/mod.rs | 186 ++- .../src/collections/vec_deque/spec_from_iter.rs | 33 + library/alloc/src/collections/vec_deque/tests.rs | 42 + library/alloc/src/fmt.rs | 4 +- library/alloc/src/lib.rs | 30 +- library/alloc/src/rc.rs | 8 +- library/alloc/src/slice.rs | 353 +--- library/alloc/src/slice/tests.rs | 359 ++++ library/alloc/src/str.rs | 14 +- library/alloc/src/string.rs | 13 +- library/alloc/src/sync.rs | 14 +- library/alloc/src/testing/crash_test.rs | 119 ++ library/alloc/src/testing/mod.rs | 3 + library/alloc/src/testing/ord_chaos.rs | 81 + library/alloc/src/testing/rng.rs | 28 + library/alloc/src/vec/drain.rs | 6 +- library/alloc/src/vec/into_iter.rs | 52 +- library/alloc/src/vec/is_zero.rs | 22 +- library/alloc/src/vec/mod.rs | 15 +- library/alloc/src/vec/splice.rs | 6 + library/alloc/tests/autotraits.rs | 2 +- library/alloc/tests/lib.rs | 1 - library/alloc/tests/slice.rs | 349 +--- library/alloc/tests/vec.rs | 31 +- library/alloc/tests/vec_deque.rs | 50 + library/core/Cargo.toml | 4 +- library/core/benches/num/int_log/mod.rs | 4 +- library/core/src/any.rs | 5 +- library/core/src/array/iter.rs | 6 +- library/core/src/array/mod.rs | 4 +- library/core/src/borrow.rs | 2 +- library/core/src/cell.rs | 35 +- library/core/src/cell/lazy.rs | 8 +- library/core/src/cell/once.rs | 13 +- library/core/src/char/decode.rs | 2 +- library/core/src/cmp.rs | 82 +- library/core/src/const_closure.rs | 29 - library/core/src/convert/num.rs | 20 + library/core/src/ffi/mod.rs | 55 +- library/core/src/fmt/mod.rs | 13 +- library/core/src/future/future.rs | 1 + library/core/src/future/mod.rs | 9 +- library/core/src/hash/mod.rs | 4 +- library/core/src/hint.rs | 69 + library/core/src/intrinsics.rs | 69 +- library/core/src/intrinsics/mir.rs | 302 +++- library/core/src/iter/range.rs | 2 +- library/core/src/iter/sources/empty.rs | 7 +- library/core/src/iter/sources/from_generator.rs | 23 +- library/core/src/iter/sources/once_with.rs | 14 +- library/core/src/iter/sources/repeat_n.rs | 2 +- library/core/src/iter/sources/repeat_with.rs | 10 +- library/core/src/iter/traits/accum.rs | 8 + library/core/src/iter/traits/iterator.rs | 43 +- library/core/src/lib.rs | 30 +- library/core/src/macros/mod.rs | 46 +- library/core/src/marker.rs | 20 +- library/core/src/mem/mod.rs | 3 +- library/core/src/num/dec2flt/fpu.rs | 2 +- library/core/src/num/f32.rs | 2 +- library/core/src/num/f64.rs | 2 +- library/core/src/num/int_macros.rs | 111 +- library/core/src/ops/function.rs | 319 ---- library/core/src/ops/index.rs | 2 +- library/core/src/ops/mod.rs | 10 +- library/core/src/ops/unsize.rs | 20 +- library/core/src/option.rs | 57 +- library/core/src/panic.rs | 8 +- library/core/src/panic/panic_info.rs | 2 +- library/core/src/panicking.rs | 39 +- library/core/src/pin.rs | 50 +- library/core/src/prelude/mod.rs | 8 +- library/core/src/prelude/v1.rs | 9 +- library/core/src/ptr/alignment.rs | 12 +- library/core/src/ptr/const_ptr.rs | 110 +- library/core/src/ptr/metadata.rs | 2 +- library/core/src/ptr/mod.rs | 54 +- library/core/src/ptr/mut_ptr.rs | 110 +- library/core/src/ptr/non_null.rs | 17 +- library/core/src/result.rs | 7 +- library/core/src/slice/iter.rs | 12 +- library/core/src/slice/iter/macros.rs | 22 +- library/core/src/slice/mod.rs | 61 +- library/core/src/slice/sort.rs | 544 +++++- library/core/src/str/iter.rs | 185 +- library/core/src/str/mod.rs | 8 +- library/core/src/sync/atomic.rs | 40 +- library/core/src/sync/exclusive.rs | 4 +- library/core/src/task/poll.rs | 2 +- library/core/src/task/wake.rs | 10 +- library/core/src/unicode/mod.rs | 2 +- library/core/tests/any.rs | 1 - library/core/tests/char.rs | 4 + library/core/tests/lazy.rs | 6 + library/core/tests/lib.rs | 14 +- library/core/tests/mem.rs | 1 - library/core/tests/num/flt2dec/random.rs | 6 +- library/core/tests/ptr.rs | 20 +- library/core/tests/slice.rs | 11 +- library/core/tests/str.rs | 2 +- library/core/tests/task.rs | 8 +- library/core/tests/time.rs | 26 + library/panic_abort/src/lib.rs | 6 +- library/panic_unwind/src/lib.rs | 4 +- library/proc_macro/src/bridge/client.rs | 2 +- library/proc_macro/src/bridge/fxhash.rs | 6 +- library/proc_macro/src/bridge/server.rs | 2 +- library/proc_macro/src/lib.rs | 2 +- library/rustc-std-workspace-alloc/lib.rs | 2 +- library/std/Cargo.toml | 5 +- library/std/src/alloc.rs | 2 +- library/std/src/backtrace.rs | 8 +- library/std/src/collections/hash/map/tests.rs | 9 +- library/std/src/collections/hash/set.rs | 2 +- library/std/src/env.rs | 9 +- library/std/src/fs.rs | 12 +- library/std/src/fs/tests.rs | 50 +- library/std/src/io/buffered/tests.rs | 4 +- library/std/src/io/error/repr_bitpacked.rs | 16 +- library/std/src/io/mod.rs | 10 +- library/std/src/io/stdio.rs | 3 +- library/std/src/lib.rs | 37 +- library/std/src/macros.rs | 1 + library/std/src/net/ip_addr.rs | 9 + library/std/src/os/fd/mod.rs | 6 +- library/std/src/os/fd/owned.rs | 2 +- library/std/src/os/net/linux_ext/addr.rs | 2 +- library/std/src/os/unix/net/addr.rs | 2 +- library/std/src/panic.rs | 3 + library/std/src/panicking.rs | 12 +- library/std/src/path.rs | 120 +- library/std/src/personality/dwarf/eh.rs | 31 +- library/std/src/prelude/v1.rs | 9 +- library/std/src/process.rs | 36 +- library/std/src/process/tests.rs | 94 ++ library/std/src/rt.rs | 6 +- library/std/src/sync/lazy_lock.rs | 10 +- library/std/src/sync/lazy_lock/tests.rs | 6 + library/std/src/sync/mod.rs | 3 + library/std/src/sync/mutex/tests.rs | 2 +- library/std/src/sync/once_lock.rs | 17 +- library/std/src/sync/remutex.rs | 178 ++ library/std/src/sync/remutex/tests.rs | 60 + library/std/src/sync/rwlock/tests.rs | 4 +- library/std/src/sys/itron/thread.rs | 27 +- library/std/src/sys/sgx/mod.rs | 1 + library/std/src/sys/sgx/thread.rs | 21 +- library/std/src/sys/sgx/thread_parking.rs | 23 + library/std/src/sys/unix/android.rs | 4 +- library/std/src/sys/unix/fs.rs | 154 +- library/std/src/sys/unix/kernel_copy.rs | 8 +- library/std/src/sys/unix/locks/pthread_condvar.rs | 93 +- library/std/src/sys/unix/mod.rs | 23 +- library/std/src/sys/unix/net.rs | 2 +- library/std/src/sys/unix/pipe.rs | 4 + library/std/src/sys/unix/process/process_common.rs | 67 +- .../std/src/sys/unix/process/process_fuchsia.rs | 15 +- library/std/src/sys/unix/process/process_unix.rs | 20 +- .../std/src/sys/unix/process/process_unix/tests.rs | 8 +- .../src/sys/unix/process/process_unsupported.rs | 4 + .../std/src/sys/unix/process/process_vxworks.rs | 11 +- library/std/src/sys/unix/stack_overflow.rs | 7 +- library/std/src/sys/unix/thread.rs | 46 +- library/std/src/sys/unix/thread_parker/darwin.rs | 131 -- library/std/src/sys/unix/thread_parker/mod.rs | 32 - library/std/src/sys/unix/thread_parker/netbsd.rs | 113 -- library/std/src/sys/unix/thread_parker/pthread.rs | 271 --- library/std/src/sys/unix/thread_parking/darwin.rs | 131 ++ library/std/src/sys/unix/thread_parking/mod.rs | 32 + library/std/src/sys/unix/thread_parking/netbsd.rs | 52 + library/std/src/sys/unix/thread_parking/pthread.rs | 271 +++ library/std/src/sys/unix/time.rs | 3 + library/std/src/sys/unix/weak.rs | 53 +- library/std/src/sys/unsupported/mod.rs | 1 + library/std/src/sys/unsupported/once.rs | 89 + library/std/src/sys/unsupported/pipe.rs | 4 + library/std/src/sys/unsupported/process.rs | 4 + library/std/src/sys/wasi/mod.rs | 2 + library/std/src/sys/wasm/mod.rs | 2 + library/std/src/sys/windows/c.rs | 13 +- library/std/src/sys/windows/mod.rs | 2 +- library/std/src/sys/windows/os.rs | 2 +- library/std/src/sys/windows/pipe.rs | 6 +- library/std/src/sys/windows/process.rs | 5 + library/std/src/sys/windows/rand.rs | 121 +- library/std/src/sys/windows/thread.rs | 2 +- library/std/src/sys/windows/thread_parker.rs | 253 --- library/std/src/sys/windows/thread_parking.rs | 253 +++ library/std/src/sys_common/backtrace.rs | 6 +- library/std/src/sys_common/io.rs | 3 +- library/std/src/sys_common/mod.rs | 3 +- library/std/src/sys_common/once/generic.rs | 283 ---- library/std/src/sys_common/once/mod.rs | 27 +- library/std/src/sys_common/once/queue.rs | 283 ++++ library/std/src/sys_common/process.rs | 42 +- library/std/src/sys_common/remutex.rs | 178 -- library/std/src/sys_common/remutex/tests.rs | 60 - library/std/src/sys_common/thread_local_key.rs | 25 +- library/std/src/sys_common/thread_parker/futex.rs | 97 -- .../std/src/sys_common/thread_parker/generic.rs | 125 -- library/std/src/sys_common/thread_parker/mod.rs | 23 - .../std/src/sys_common/thread_parker/wait_flag.rs | 102 -- library/std/src/sys_common/thread_parking/futex.rs | 97 ++ .../std/src/sys_common/thread_parking/generic.rs | 125 ++ library/std/src/sys_common/thread_parking/id.rs | 108 ++ library/std/src/sys_common/thread_parking/mod.rs | 29 + .../std/src/sys_common/thread_parking/wait_flag.rs | 102 ++ library/std/src/thread/local.rs | 23 +- library/std/src/thread/local/tests.rs | 22 +- library/std/src/thread/mod.rs | 26 +- library/std/tests/env.rs | 20 +- library/std/tests/run-time-detect.rs | 3 +- library/test/src/cli.rs | 12 +- library/test/src/console.rs | 6 +- library/test/src/formatters/json.rs | 23 +- library/test/src/formatters/junit.rs | 16 +- library/test/src/formatters/mod.rs | 2 +- library/test/src/formatters/pretty.rs | 18 +- library/test/src/formatters/terse.rs | 20 +- library/test/src/lib.rs | 9 +- library/test/src/term/terminfo/mod.rs | 6 +- library/test/src/term/terminfo/parm.rs | 8 +- library/test/src/term/terminfo/searcher.rs | 6 +- library/test/src/test_result.rs | 9 +- library/test/src/time.rs | 6 +- library/test/src/types.rs | 2 +- library/unwind/src/libunwind.rs | 2 +- 245 files changed, 7674 insertions(+), 6252 deletions(-) delete mode 100644 library/alloc/src/collections/binary_heap.rs create mode 100644 library/alloc/src/collections/binary_heap/mod.rs delete mode 100644 library/alloc/src/collections/btree/testing/crash_test.rs delete mode 100644 library/alloc/src/collections/btree/testing/mod.rs delete mode 100644 library/alloc/src/collections/btree/testing/ord_chaos.rs delete mode 100644 library/alloc/src/collections/btree/testing/rng.rs create mode 100644 library/alloc/src/collections/vec_deque/spec_from_iter.rs create mode 100644 library/alloc/src/slice/tests.rs create mode 100644 library/alloc/src/testing/crash_test.rs create mode 100644 library/alloc/src/testing/mod.rs create mode 100644 library/alloc/src/testing/ord_chaos.rs create mode 100644 library/alloc/src/testing/rng.rs create mode 100644 library/std/src/sync/remutex.rs create mode 100644 library/std/src/sync/remutex/tests.rs create mode 100644 library/std/src/sys/sgx/thread_parking.rs delete mode 100644 library/std/src/sys/unix/thread_parker/darwin.rs delete mode 100644 library/std/src/sys/unix/thread_parker/mod.rs delete mode 100644 library/std/src/sys/unix/thread_parker/netbsd.rs delete mode 100644 library/std/src/sys/unix/thread_parker/pthread.rs create mode 100644 library/std/src/sys/unix/thread_parking/darwin.rs create mode 100644 library/std/src/sys/unix/thread_parking/mod.rs create mode 100644 library/std/src/sys/unix/thread_parking/netbsd.rs create mode 100644 library/std/src/sys/unix/thread_parking/pthread.rs create mode 100644 library/std/src/sys/unsupported/once.rs delete mode 100644 library/std/src/sys/windows/thread_parker.rs create mode 100644 library/std/src/sys/windows/thread_parking.rs delete mode 100644 library/std/src/sys_common/once/generic.rs create mode 100644 library/std/src/sys_common/once/queue.rs delete mode 100644 library/std/src/sys_common/remutex.rs delete mode 100644 library/std/src/sys_common/remutex/tests.rs delete mode 100644 library/std/src/sys_common/thread_parker/futex.rs delete mode 100644 library/std/src/sys_common/thread_parker/generic.rs delete mode 100644 library/std/src/sys_common/thread_parker/mod.rs delete mode 100644 library/std/src/sys_common/thread_parker/wait_flag.rs create mode 100644 library/std/src/sys_common/thread_parking/futex.rs create mode 100644 library/std/src/sys_common/thread_parking/generic.rs create mode 100644 library/std/src/sys_common/thread_parking/id.rs create mode 100644 library/std/src/sys_common/thread_parking/mod.rs create mode 100644 library/std/src/sys_common/thread_parking/wait_flag.rs (limited to 'library') diff --git a/library/alloc/Cargo.toml b/library/alloc/Cargo.toml index 265020209..95c07abf7 100644 --- a/library/alloc/Cargo.toml +++ b/library/alloc/Cargo.toml @@ -13,8 +13,8 @@ core = { path = "../core" } compiler_builtins = { version = "0.1.40", features = ['rustc-dep-of-std'] } [dev-dependencies] -rand = "0.7" -rand_xorshift = "0.2" +rand = { version = "0.8.5", default-features = false, features = ["alloc"] } +rand_xorshift = "0.3.0" [[test]] name = "collectionstests" diff --git a/library/alloc/benches/slice.rs b/library/alloc/benches/slice.rs index bd6f38f2f..b62be9d39 100644 --- a/library/alloc/benches/slice.rs +++ b/library/alloc/benches/slice.rs @@ -1,6 +1,6 @@ use std::{mem, ptr}; -use rand::distributions::{Alphanumeric, Standard}; +use rand::distributions::{Alphanumeric, DistString, Standard}; use rand::Rng; use test::{black_box, Bencher}; @@ -218,7 +218,7 @@ fn gen_strings(len: usize) -> Vec { let mut v = vec![]; for _ in 0..len { let n = rng.gen::() % 20 + 1; - v.push((&mut rng).sample_iter(&Alphanumeric).take(n).collect()); + v.push(Alphanumeric.sample_string(&mut rng, n)); } v } diff --git a/library/alloc/src/alloc.rs b/library/alloc/src/alloc.rs index e5fbfc557..3a797bd5e 100644 --- a/library/alloc/src/alloc.rs +++ b/library/alloc/src/alloc.rs @@ -20,10 +20,10 @@ use core::marker::Destruct; mod tests; extern "Rust" { - // These are the magic symbols to call the global allocator. rustc generates + // These are the magic symbols to call the global allocator. rustc generates // them to call `__rg_alloc` etc. if there is a `#[global_allocator]` attribute // (the code expanding that attribute macro generates those functions), or to call - // the default implementations in libstd (`__rdl_alloc` etc. in `library/std/src/alloc.rs`) + // the default implementations in std (`__rdl_alloc` etc. in `library/std/src/alloc.rs`) // otherwise. // The rustc fork of LLVM 14 and earlier also special-cases these function names to be able to optimize them // like `malloc`, `realloc`, and `free`, respectively. @@ -353,7 +353,7 @@ pub(crate) const unsafe fn box_free !; @@ -402,20 +402,20 @@ pub mod __alloc_error_handler { // `#[alloc_error_handler]`. #[rustc_std_internal_symbol] pub unsafe fn __rdl_oom(size: usize, _align: usize) -> ! { - panic!("memory allocation of {size} bytes failed") - } - - #[cfg(bootstrap)] - #[rustc_std_internal_symbol] - pub unsafe fn __rg_oom(size: usize, align: usize) -> ! { - use crate::alloc::Layout; - - let layout = unsafe { Layout::from_size_align_unchecked(size, align) }; extern "Rust" { - #[lang = "oom"] - fn oom_impl(layout: Layout) -> !; + // This symbol is emitted by rustc next to __rust_alloc_error_handler. + // Its value depends on the -Zoom={panic,abort} compiler option. + static __rust_alloc_error_handler_should_panic: u8; + } + + #[allow(unused_unsafe)] + if unsafe { __rust_alloc_error_handler_should_panic != 0 } { + panic!("memory allocation of {size} bytes failed") + } else { + core::panicking::panic_nounwind_fmt(format_args!( + "memory allocation of {size} bytes failed" + )) } - unsafe { oom_impl(layout) } } } diff --git a/library/alloc/src/boxed.rs b/library/alloc/src/boxed.rs index e5f6b0c0c..a563b2587 100644 --- a/library/alloc/src/boxed.rs +++ b/library/alloc/src/boxed.rs @@ -158,7 +158,6 @@ use core::hash::{Hash, Hasher}; #[cfg(not(no_global_oom_handling))] use core::iter::FromIterator; use core::iter::{FusedIterator, Iterator}; -#[cfg(not(bootstrap))] use core::marker::Tuple; use core::marker::{Destruct, Unpin, Unsize}; use core::mem; @@ -954,7 +953,7 @@ impl Box { /// [`Layout`]: crate::Layout #[stable(feature = "box_raw", since = "1.4.0")] #[inline] - #[must_use = "call `drop(from_raw(ptr))` if you intend to drop the `Box`"] + #[must_use = "call `drop(Box::from_raw(ptr))` if you intend to drop the `Box`"] pub unsafe fn from_raw(raw: *mut T) -> Self { unsafe { Self::from_raw_in(raw, Global) } } @@ -1981,17 +1980,6 @@ impl ExactSizeIterator for Box FusedIterator for Box {} -#[cfg(bootstrap)] -#[stable(feature = "boxed_closure_impls", since = "1.35.0")] -impl + ?Sized, A: Allocator> FnOnce for Box { - type Output = >::Output; - - extern "rust-call" fn call_once(self, args: Args) -> Self::Output { - >::call_once(*self, args) - } -} - -#[cfg(not(bootstrap))] #[stable(feature = "boxed_closure_impls", since = "1.35.0")] impl + ?Sized, A: Allocator> FnOnce for Box { type Output = >::Output; @@ -2001,15 +1989,6 @@ impl + ?Sized, A: Allocator> FnOnce for Box + ?Sized, A: Allocator> FnMut for Box { - extern "rust-call" fn call_mut(&mut self, args: Args) -> Self::Output { - >::call_mut(self, args) - } -} - -#[cfg(not(bootstrap))] #[stable(feature = "boxed_closure_impls", since = "1.35.0")] impl + ?Sized, A: Allocator> FnMut for Box { extern "rust-call" fn call_mut(&mut self, args: Args) -> Self::Output { @@ -2017,15 +1996,6 @@ impl + ?Sized, A: Allocator> FnMut for Box + ?Sized, A: Allocator> Fn for Box { - extern "rust-call" fn call(&self, args: Args) -> Self::Output { - >::call(self, args) - } -} - -#[cfg(not(bootstrap))] #[stable(feature = "boxed_closure_impls", since = "1.35.0")] impl + ?Sized, A: Allocator> Fn for Box { extern "rust-call" fn call(&self, args: Args) -> Self::Output { @@ -2033,7 +2003,7 @@ impl + ?Sized, A: Allocator> Fn for Box { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized, A: Allocator> CoerceUnsized> for Box {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] diff --git a/library/alloc/src/boxed/thin.rs b/library/alloc/src/boxed/thin.rs index c477c4490..c1a82e452 100644 --- a/library/alloc/src/boxed/thin.rs +++ b/library/alloc/src/boxed/thin.rs @@ -226,24 +226,45 @@ impl WithHeader { // - Assumes that either `value` can be dereferenced, or is the // `NonNull::dangling()` we use when both `T` and `H` are ZSTs. unsafe fn drop(&self, value: *mut T) { + struct DropGuard { + ptr: NonNull, + value_layout: Layout, + _marker: PhantomData, + } + + impl Drop for DropGuard { + fn drop(&mut self) { + unsafe { + // SAFETY: Layout must have been computable if we're in drop + let (layout, value_offset) = + WithHeader::::alloc_layout(self.value_layout).unwrap_unchecked(); + + // Note: Don't deallocate if the layout size is zero, because the pointer + // didn't come from the allocator. + if layout.size() != 0 { + alloc::dealloc(self.ptr.as_ptr().sub(value_offset), layout); + } else { + debug_assert!( + value_offset == 0 + && mem::size_of::() == 0 + && self.value_layout.size() == 0 + ); + } + } + } + } + unsafe { - let value_layout = Layout::for_value_raw(value); - // SAFETY: Layout must have been computable if we're in drop - let (layout, value_offset) = Self::alloc_layout(value_layout).unwrap_unchecked(); + // `_guard` will deallocate the memory when dropped, even if `drop_in_place` unwinds. + let _guard = DropGuard { + ptr: self.0, + value_layout: Layout::for_value_raw(value), + _marker: PhantomData::, + }; // We only drop the value because the Pointee trait requires that the metadata is copy // aka trivially droppable. ptr::drop_in_place::(value); - - // Note: Don't deallocate if the layout size is zero, because the pointer - // didn't come from the allocator. - if layout.size() != 0 { - alloc::dealloc(self.0.as_ptr().sub(value_offset), layout); - } else { - debug_assert!( - value_offset == 0 && mem::size_of::() == 0 && value_layout.size() == 0 - ); - } } } diff --git a/library/alloc/src/collections/binary_heap.rs b/library/alloc/src/collections/binary_heap.rs deleted file mode 100644 index 4583bc9a1..000000000 --- a/library/alloc/src/collections/binary_heap.rs +++ /dev/null @@ -1,1721 +0,0 @@ -//! A priority queue implemented with a binary heap. -//! -//! Insertion and popping the largest element have *O*(log(*n*)) time complexity. -//! Checking the largest element is *O*(1). Converting a vector to a binary heap -//! can be done in-place, and has *O*(*n*) complexity. A binary heap can also be -//! converted to a sorted vector in-place, allowing it to be used for an *O*(*n* * log(*n*)) -//! in-place heapsort. -//! -//! # Examples -//! -//! This is a larger example that implements [Dijkstra's algorithm][dijkstra] -//! to solve the [shortest path problem][sssp] on a [directed graph][dir_graph]. -//! It shows how to use [`BinaryHeap`] with custom types. -//! -//! [dijkstra]: https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm -//! [sssp]: https://en.wikipedia.org/wiki/Shortest_path_problem -//! [dir_graph]: https://en.wikipedia.org/wiki/Directed_graph -//! -//! ``` -//! use std::cmp::Ordering; -//! use std::collections::BinaryHeap; -//! -//! #[derive(Copy, Clone, Eq, PartialEq)] -//! struct State { -//! cost: usize, -//! position: usize, -//! } -//! -//! // The priority queue depends on `Ord`. -//! // Explicitly implement the trait so the queue becomes a min-heap -//! // instead of a max-heap. -//! impl Ord for State { -//! fn cmp(&self, other: &Self) -> Ordering { -//! // Notice that the we flip the ordering on costs. -//! // In case of a tie we compare positions - this step is necessary -//! // to make implementations of `PartialEq` and `Ord` consistent. -//! other.cost.cmp(&self.cost) -//! .then_with(|| self.position.cmp(&other.position)) -//! } -//! } -//! -//! // `PartialOrd` needs to be implemented as well. -//! impl PartialOrd for State { -//! fn partial_cmp(&self, other: &Self) -> Option { -//! Some(self.cmp(other)) -//! } -//! } -//! -//! // Each node is represented as a `usize`, for a shorter implementation. -//! struct Edge { -//! node: usize, -//! cost: usize, -//! } -//! -//! // Dijkstra's shortest path algorithm. -//! -//! // Start at `start` and use `dist` to track the current shortest distance -//! // to each node. This implementation isn't memory-efficient as it may leave duplicate -//! // nodes in the queue. It also uses `usize::MAX` as a sentinel value, -//! // for a simpler implementation. -//! fn shortest_path(adj_list: &Vec>, start: usize, goal: usize) -> Option { -//! // dist[node] = current shortest distance from `start` to `node` -//! let mut dist: Vec<_> = (0..adj_list.len()).map(|_| usize::MAX).collect(); -//! -//! let mut heap = BinaryHeap::new(); -//! -//! // We're at `start`, with a zero cost -//! dist[start] = 0; -//! heap.push(State { cost: 0, position: start }); -//! -//! // Examine the frontier with lower cost nodes first (min-heap) -//! while let Some(State { cost, position }) = heap.pop() { -//! // Alternatively we could have continued to find all shortest paths -//! if position == goal { return Some(cost); } -//! -//! // Important as we may have already found a better way -//! if cost > dist[position] { continue; } -//! -//! // For each node we can reach, see if we can find a way with -//! // a lower cost going through this node -//! for edge in &adj_list[position] { -//! let next = State { cost: cost + edge.cost, position: edge.node }; -//! -//! // If so, add it to the frontier and continue -//! if next.cost < dist[next.position] { -//! heap.push(next); -//! // Relaxation, we have now found a better way -//! dist[next.position] = next.cost; -//! } -//! } -//! } -//! -//! // Goal not reachable -//! None -//! } -//! -//! fn main() { -//! // This is the directed graph we're going to use. -//! // The node numbers correspond to the different states, -//! // and the edge weights symbolize the cost of moving -//! // from one node to another. -//! // Note that the edges are one-way. -//! // -//! // 7 -//! // +-----------------+ -//! // | | -//! // v 1 2 | 2 -//! // 0 -----> 1 -----> 3 ---> 4 -//! // | ^ ^ ^ -//! // | | 1 | | -//! // | | | 3 | 1 -//! // +------> 2 -------+ | -//! // 10 | | -//! // +---------------+ -//! // -//! // The graph is represented as an adjacency list where each index, -//! // corresponding to a node value, has a list of outgoing edges. -//! // Chosen for its efficiency. -//! let graph = vec![ -//! // Node 0 -//! vec![Edge { node: 2, cost: 10 }, -//! Edge { node: 1, cost: 1 }], -//! // Node 1 -//! vec![Edge { node: 3, cost: 2 }], -//! // Node 2 -//! vec![Edge { node: 1, cost: 1 }, -//! Edge { node: 3, cost: 3 }, -//! Edge { node: 4, cost: 1 }], -//! // Node 3 -//! vec![Edge { node: 0, cost: 7 }, -//! Edge { node: 4, cost: 2 }], -//! // Node 4 -//! vec![]]; -//! -//! assert_eq!(shortest_path(&graph, 0, 1), Some(1)); -//! assert_eq!(shortest_path(&graph, 0, 3), Some(3)); -//! assert_eq!(shortest_path(&graph, 3, 0), Some(7)); -//! assert_eq!(shortest_path(&graph, 0, 4), Some(5)); -//! assert_eq!(shortest_path(&graph, 4, 0), None); -//! } -//! ``` - -#![allow(missing_docs)] -#![stable(feature = "rust1", since = "1.0.0")] - -use core::fmt; -use core::iter::{FromIterator, FusedIterator, InPlaceIterable, SourceIter, TrustedLen}; -use core::mem::{self, swap, ManuallyDrop}; -use core::ops::{Deref, DerefMut}; -use core::ptr; - -use crate::collections::TryReserveError; -use crate::slice; -use crate::vec::{self, AsVecIntoIter, Vec}; - -use super::SpecExtend; - -#[cfg(test)] -mod tests; - -/// A priority queue implemented with a binary heap. -/// -/// This will be a max-heap. -/// -/// It is a logic error for an item to be modified in such a way that the -/// item's ordering relative to any other item, as determined by the [`Ord`] -/// trait, changes while it is in the heap. This is normally only possible -/// through [`Cell`], [`RefCell`], global state, I/O, or unsafe code. The -/// behavior resulting from such a logic error is not specified, but will -/// be encapsulated to the `BinaryHeap` that observed the logic error and not -/// result in undefined behavior. This could include panics, incorrect results, -/// aborts, memory leaks, and non-termination. -/// -/// # Examples -/// -/// ``` -/// use std::collections::BinaryHeap; -/// -/// // Type inference lets us omit an explicit type signature (which -/// // would be `BinaryHeap` in this example). -/// let mut heap = BinaryHeap::new(); -/// -/// // We can use peek to look at the next item in the heap. In this case, -/// // there's no items in there yet so we get None. -/// assert_eq!(heap.peek(), None); -/// -/// // Let's add some scores... -/// heap.push(1); -/// heap.push(5); -/// heap.push(2); -/// -/// // Now peek shows the most important item in the heap. -/// assert_eq!(heap.peek(), Some(&5)); -/// -/// // We can check the length of a heap. -/// assert_eq!(heap.len(), 3); -/// -/// // We can iterate over the items in the heap, although they are returned in -/// // a random order. -/// for x in &heap { -/// println!("{x}"); -/// } -/// -/// // If we instead pop these scores, they should come back in order. -/// assert_eq!(heap.pop(), Some(5)); -/// assert_eq!(heap.pop(), Some(2)); -/// assert_eq!(heap.pop(), Some(1)); -/// assert_eq!(heap.pop(), None); -/// -/// // We can clear the heap of any remaining items. -/// heap.clear(); -/// -/// // The heap should now be empty. -/// assert!(heap.is_empty()) -/// ``` -/// -/// A `BinaryHeap` with a known list of items can be initialized from an array: -/// -/// ``` -/// use std::collections::BinaryHeap; -/// -/// let heap = BinaryHeap::from([1, 5, 2]); -/// ``` -/// -/// ## Min-heap -/// -/// Either [`core::cmp::Reverse`] or a custom [`Ord`] implementation can be used to -/// make `BinaryHeap` a min-heap. This makes `heap.pop()` return the smallest -/// value instead of the greatest one. -/// -/// ``` -/// use std::collections::BinaryHeap; -/// use std::cmp::Reverse; -/// -/// let mut heap = BinaryHeap::new(); -/// -/// // Wrap values in `Reverse` -/// heap.push(Reverse(1)); -/// heap.push(Reverse(5)); -/// heap.push(Reverse(2)); -/// -/// // If we pop these scores now, they should come back in the reverse order. -/// assert_eq!(heap.pop(), Some(Reverse(1))); -/// assert_eq!(heap.pop(), Some(Reverse(2))); -/// assert_eq!(heap.pop(), Some(Reverse(5))); -/// assert_eq!(heap.pop(), None); -/// ``` -/// -/// # Time complexity -/// -/// | [push] | [pop] | [peek]/[peek\_mut] | -/// |---------|---------------|--------------------| -/// | *O*(1)~ | *O*(log(*n*)) | *O*(1) | -/// -/// The value for `push` is an expected cost; the method documentation gives a -/// more detailed analysis. -/// -/// [`core::cmp::Reverse`]: core::cmp::Reverse -/// [`Ord`]: core::cmp::Ord -/// [`Cell`]: core::cell::Cell -/// [`RefCell`]: core::cell::RefCell -/// [push]: BinaryHeap::push -/// [pop]: BinaryHeap::pop -/// [peek]: BinaryHeap::peek -/// [peek\_mut]: BinaryHeap::peek_mut -#[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(not(test), rustc_diagnostic_item = "BinaryHeap")] -pub struct BinaryHeap { - data: Vec, -} - -/// Structure wrapping a mutable reference to the greatest item on a -/// `BinaryHeap`. -/// -/// This `struct` is created by the [`peek_mut`] method on [`BinaryHeap`]. See -/// its documentation for more. -/// -/// [`peek_mut`]: BinaryHeap::peek_mut -#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] -pub struct PeekMut<'a, T: 'a + Ord> { - heap: &'a mut BinaryHeap, - sift: bool, -} - -#[stable(feature = "collection_debug", since = "1.17.0")] -impl fmt::Debug for PeekMut<'_, T> { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - f.debug_tuple("PeekMut").field(&self.heap.data[0]).finish() - } -} - -#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] -impl Drop for PeekMut<'_, T> { - fn drop(&mut self) { - if self.sift { - // SAFETY: PeekMut is only instantiated for non-empty heaps. - unsafe { self.heap.sift_down(0) }; - } - } -} - -#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] -impl Deref for PeekMut<'_, T> { - type Target = T; - fn deref(&self) -> &T { - debug_assert!(!self.heap.is_empty()); - // SAFE: PeekMut is only instantiated for non-empty heaps - unsafe { self.heap.data.get_unchecked(0) } - } -} - -#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] -impl DerefMut for PeekMut<'_, T> { - fn deref_mut(&mut self) -> &mut T { - debug_assert!(!self.heap.is_empty()); - self.sift = true; - // SAFE: PeekMut is only instantiated for non-empty heaps - unsafe { self.heap.data.get_unchecked_mut(0) } - } -} - -impl<'a, T: Ord> PeekMut<'a, T> { - /// Removes the peeked value from the heap and returns it. - #[stable(feature = "binary_heap_peek_mut_pop", since = "1.18.0")] - pub fn pop(mut this: PeekMut<'a, T>) -> T { - let value = this.heap.pop().unwrap(); - this.sift = false; - value - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl Clone for BinaryHeap { - fn clone(&self) -> Self { - BinaryHeap { data: self.data.clone() } - } - - fn clone_from(&mut self, source: &Self) { - self.data.clone_from(&source.data); - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl Default for BinaryHeap { - /// Creates an empty `BinaryHeap`. - #[inline] - fn default() -> BinaryHeap { - BinaryHeap::new() - } -} - -#[stable(feature = "binaryheap_debug", since = "1.4.0")] -impl fmt::Debug for BinaryHeap { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - f.debug_list().entries(self.iter()).finish() - } -} - -impl BinaryHeap { - /// Creates an empty `BinaryHeap` as a max-heap. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// heap.push(4); - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - #[must_use] - pub fn new() -> BinaryHeap { - BinaryHeap { data: vec![] } - } - - /// Creates an empty `BinaryHeap` with at least the specified capacity. - /// - /// The binary heap will be able to hold at least `capacity` elements without - /// reallocating. This method is allowed to allocate for more elements than - /// `capacity`. If `capacity` is 0, the binary heap will not allocate. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::with_capacity(10); - /// heap.push(4); - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - #[must_use] - pub fn with_capacity(capacity: usize) -> BinaryHeap { - BinaryHeap { data: Vec::with_capacity(capacity) } - } - - /// Returns a mutable reference to the greatest item in the binary heap, or - /// `None` if it is empty. - /// - /// Note: If the `PeekMut` value is leaked, the heap may be in an - /// inconsistent state. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// assert!(heap.peek_mut().is_none()); - /// - /// heap.push(1); - /// heap.push(5); - /// heap.push(2); - /// { - /// let mut val = heap.peek_mut().unwrap(); - /// *val = 0; - /// } - /// assert_eq!(heap.peek(), Some(&2)); - /// ``` - /// - /// # Time complexity - /// - /// If the item is modified then the worst case time complexity is *O*(log(*n*)), - /// otherwise it's *O*(1). - #[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] - pub fn peek_mut(&mut self) -> Option> { - if self.is_empty() { None } else { Some(PeekMut { heap: self, sift: false }) } - } - - /// Removes the greatest item from the binary heap and returns it, or `None` if it - /// is empty. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::from([1, 3]); - /// - /// assert_eq!(heap.pop(), Some(3)); - /// assert_eq!(heap.pop(), Some(1)); - /// assert_eq!(heap.pop(), None); - /// ``` - /// - /// # Time complexity - /// - /// The worst case cost of `pop` on a heap containing *n* elements is *O*(log(*n*)). - #[stable(feature = "rust1", since = "1.0.0")] - pub fn pop(&mut self) -> Option { - self.data.pop().map(|mut item| { - if !self.is_empty() { - swap(&mut item, &mut self.data[0]); - // SAFETY: !self.is_empty() means that self.len() > 0 - unsafe { self.sift_down_to_bottom(0) }; - } - item - }) - } - - /// Pushes an item onto the binary heap. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// heap.push(3); - /// heap.push(5); - /// heap.push(1); - /// - /// assert_eq!(heap.len(), 3); - /// assert_eq!(heap.peek(), Some(&5)); - /// ``` - /// - /// # Time complexity - /// - /// The expected cost of `push`, averaged over every possible ordering of - /// the elements being pushed, and over a sufficiently large number of - /// pushes, is *O*(1). This is the most meaningful cost metric when pushing - /// elements that are *not* already in any sorted pattern. - /// - /// The time complexity degrades if elements are pushed in predominantly - /// ascending order. In the worst case, elements are pushed in ascending - /// sorted order and the amortized cost per push is *O*(log(*n*)) against a heap - /// containing *n* elements. - /// - /// The worst case cost of a *single* call to `push` is *O*(*n*). The worst case - /// occurs when capacity is exhausted and needs a resize. The resize cost - /// has been amortized in the previous figures. - #[stable(feature = "rust1", since = "1.0.0")] - pub fn push(&mut self, item: T) { - let old_len = self.len(); - self.data.push(item); - // SAFETY: Since we pushed a new item it means that - // old_len = self.len() - 1 < self.len() - unsafe { self.sift_up(0, old_len) }; - } - - /// Consumes the `BinaryHeap` and returns a vector in sorted - /// (ascending) order. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// - /// let mut heap = BinaryHeap::from([1, 2, 4, 5, 7]); - /// heap.push(6); - /// heap.push(3); - /// - /// let vec = heap.into_sorted_vec(); - /// assert_eq!(vec, [1, 2, 3, 4, 5, 6, 7]); - /// ``` - #[must_use = "`self` will be dropped if the result is not used"] - #[stable(feature = "binary_heap_extras_15", since = "1.5.0")] - pub fn into_sorted_vec(mut self) -> Vec { - let mut end = self.len(); - while end > 1 { - end -= 1; - // SAFETY: `end` goes from `self.len() - 1` to 1 (both included), - // so it's always a valid index to access. - // It is safe to access index 0 (i.e. `ptr`), because - // 1 <= end < self.len(), which means self.len() >= 2. - unsafe { - let ptr = self.data.as_mut_ptr(); - ptr::swap(ptr, ptr.add(end)); - } - // SAFETY: `end` goes from `self.len() - 1` to 1 (both included) so: - // 0 < 1 <= end <= self.len() - 1 < self.len() - // Which means 0 < end and end < self.len(). - unsafe { self.sift_down_range(0, end) }; - } - self.into_vec() - } - - // The implementations of sift_up and sift_down use unsafe blocks in - // order to move an element out of the vector (leaving behind a - // hole), shift along the others and move the removed element back into the - // vector at the final location of the hole. - // The `Hole` type is used to represent this, and make sure - // the hole is filled back at the end of its scope, even on panic. - // Using a hole reduces the constant factor compared to using swaps, - // which involves twice as many moves. - - /// # Safety - /// - /// The caller must guarantee that `pos < self.len()`. - unsafe fn sift_up(&mut self, start: usize, pos: usize) -> usize { - // Take out the value at `pos` and create a hole. - // SAFETY: The caller guarantees that pos < self.len() - let mut hole = unsafe { Hole::new(&mut self.data, pos) }; - - while hole.pos() > start { - let parent = (hole.pos() - 1) / 2; - - // SAFETY: hole.pos() > start >= 0, which means hole.pos() > 0 - // and so hole.pos() - 1 can't underflow. - // This guarantees that parent < hole.pos() so - // it's a valid index and also != hole.pos(). - if hole.element() <= unsafe { hole.get(parent) } { - break; - } - - // SAFETY: Same as above - unsafe { hole.move_to(parent) }; - } - - hole.pos() - } - - /// Take an element at `pos` and move it down the heap, - /// while its children are larger. - /// - /// # Safety - /// - /// The caller must guarantee that `pos < end <= self.len()`. - unsafe fn sift_down_range(&mut self, pos: usize, end: usize) { - // SAFETY: The caller guarantees that pos < end <= self.len(). - let mut hole = unsafe { Hole::new(&mut self.data, pos) }; - let mut child = 2 * hole.pos() + 1; - - // Loop invariant: child == 2 * hole.pos() + 1. - while child <= end.saturating_sub(2) { - // compare with the greater of the two children - // SAFETY: child < end - 1 < self.len() and - // child + 1 < end <= self.len(), so they're valid indexes. - // child == 2 * hole.pos() + 1 != hole.pos() and - // child + 1 == 2 * hole.pos() + 2 != hole.pos(). - // FIXME: 2 * hole.pos() + 1 or 2 * hole.pos() + 2 could overflow - // if T is a ZST - child += unsafe { hole.get(child) <= hole.get(child + 1) } as usize; - - // if we are already in order, stop. - // SAFETY: child is now either the old child or the old child+1 - // We already proven that both are < self.len() and != hole.pos() - if hole.element() >= unsafe { hole.get(child) } { - return; - } - - // SAFETY: same as above. - unsafe { hole.move_to(child) }; - child = 2 * hole.pos() + 1; - } - - // SAFETY: && short circuit, which means that in the - // second condition it's already true that child == end - 1 < self.len(). - if child == end - 1 && hole.element() < unsafe { hole.get(child) } { - // SAFETY: child is already proven to be a valid index and - // child == 2 * hole.pos() + 1 != hole.pos(). - unsafe { hole.move_to(child) }; - } - } - - /// # Safety - /// - /// The caller must guarantee that `pos < self.len()`. - unsafe fn sift_down(&mut self, pos: usize) { - let len = self.len(); - // SAFETY: pos < len is guaranteed by the caller and - // obviously len = self.len() <= self.len(). - unsafe { self.sift_down_range(pos, len) }; - } - - /// Take an element at `pos` and move it all the way down the heap, - /// then sift it up to its position. - /// - /// Note: This is faster when the element is known to be large / should - /// be closer to the bottom. - /// - /// # Safety - /// - /// The caller must guarantee that `pos < self.len()`. - unsafe fn sift_down_to_bottom(&mut self, mut pos: usize) { - let end = self.len(); - let start = pos; - - // SAFETY: The caller guarantees that pos < self.len(). - let mut hole = unsafe { Hole::new(&mut self.data, pos) }; - let mut child = 2 * hole.pos() + 1; - - // Loop invariant: child == 2 * hole.pos() + 1. - while child <= end.saturating_sub(2) { - // SAFETY: child < end - 1 < self.len() and - // child + 1 < end <= self.len(), so they're valid indexes. - // child == 2 * hole.pos() + 1 != hole.pos() and - // child + 1 == 2 * hole.pos() + 2 != hole.pos(). - // FIXME: 2 * hole.pos() + 1 or 2 * hole.pos() + 2 could overflow - // if T is a ZST - child += unsafe { hole.get(child) <= hole.get(child + 1) } as usize; - - // SAFETY: Same as above - unsafe { hole.move_to(child) }; - child = 2 * hole.pos() + 1; - } - - if child == end - 1 { - // SAFETY: child == end - 1 < self.len(), so it's a valid index - // and child == 2 * hole.pos() + 1 != hole.pos(). - unsafe { hole.move_to(child) }; - } - pos = hole.pos(); - drop(hole); - - // SAFETY: pos is the position in the hole and was already proven - // to be a valid index. - unsafe { self.sift_up(start, pos) }; - } - - /// Rebuild assuming data[0..start] is still a proper heap. - fn rebuild_tail(&mut self, start: usize) { - if start == self.len() { - return; - } - - let tail_len = self.len() - start; - - #[inline(always)] - fn log2_fast(x: usize) -> usize { - (usize::BITS - x.leading_zeros() - 1) as usize - } - - // `rebuild` takes O(self.len()) operations - // and about 2 * self.len() comparisons in the worst case - // while repeating `sift_up` takes O(tail_len * log(start)) operations - // and about 1 * tail_len * log_2(start) comparisons in the worst case, - // assuming start >= tail_len. For larger heaps, the crossover point - // no longer follows this reasoning and was determined empirically. - let better_to_rebuild = if start < tail_len { - true - } else if self.len() <= 2048 { - 2 * self.len() < tail_len * log2_fast(start) - } else { - 2 * self.len() < tail_len * 11 - }; - - if better_to_rebuild { - self.rebuild(); - } else { - for i in start..self.len() { - // SAFETY: The index `i` is always less than self.len(). - unsafe { self.sift_up(0, i) }; - } - } - } - - fn rebuild(&mut self) { - let mut n = self.len() / 2; - while n > 0 { - n -= 1; - // SAFETY: n starts from self.len() / 2 and goes down to 0. - // The only case when !(n < self.len()) is if - // self.len() == 0, but it's ruled out by the loop condition. - unsafe { self.sift_down(n) }; - } - } - - /// Moves all the elements of `other` into `self`, leaving `other` empty. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// - /// let mut a = BinaryHeap::from([-10, 1, 2, 3, 3]); - /// let mut b = BinaryHeap::from([-20, 5, 43]); - /// - /// a.append(&mut b); - /// - /// assert_eq!(a.into_sorted_vec(), [-20, -10, 1, 2, 3, 3, 5, 43]); - /// assert!(b.is_empty()); - /// ``` - #[stable(feature = "binary_heap_append", since = "1.11.0")] - pub fn append(&mut self, other: &mut Self) { - if self.len() < other.len() { - swap(self, other); - } - - let start = self.data.len(); - - self.data.append(&mut other.data); - - self.rebuild_tail(start); - } - - /// Clears the binary heap, returning an iterator over the removed elements - /// in heap order. If the iterator is dropped before being fully consumed, - /// it drops the remaining elements in heap order. - /// - /// The returned iterator keeps a mutable borrow on the heap to optimize - /// its implementation. - /// - /// Note: - /// * `.drain_sorted()` is *O*(*n* \* log(*n*)); much slower than `.drain()`. - /// You should use the latter for most cases. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// #![feature(binary_heap_drain_sorted)] - /// use std::collections::BinaryHeap; - /// - /// let mut heap = BinaryHeap::from([1, 2, 3, 4, 5]); - /// assert_eq!(heap.len(), 5); - /// - /// drop(heap.drain_sorted()); // removes all elements in heap order - /// assert_eq!(heap.len(), 0); - /// ``` - #[inline] - #[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] - pub fn drain_sorted(&mut self) -> DrainSorted<'_, T> { - DrainSorted { inner: self } - } - - /// Retains only the elements specified by the predicate. - /// - /// In other words, remove all elements `e` for which `f(&e)` returns - /// `false`. The elements are visited in unsorted (and unspecified) order. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// #![feature(binary_heap_retain)] - /// use std::collections::BinaryHeap; - /// - /// let mut heap = BinaryHeap::from([-10, -5, 1, 2, 4, 13]); - /// - /// heap.retain(|x| x % 2 == 0); // only keep even numbers - /// - /// assert_eq!(heap.into_sorted_vec(), [-10, 2, 4]) - /// ``` - #[unstable(feature = "binary_heap_retain", issue = "71503")] - pub fn retain(&mut self, mut f: F) - where - F: FnMut(&T) -> bool, - { - let mut first_removed = self.len(); - let mut i = 0; - self.data.retain(|e| { - let keep = f(e); - if !keep && i < first_removed { - first_removed = i; - } - i += 1; - keep - }); - // data[0..first_removed] is untouched, so we only need to rebuild the tail: - self.rebuild_tail(first_removed); - } -} - -impl BinaryHeap { - /// Returns an iterator visiting all values in the underlying vector, in - /// arbitrary order. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let heap = BinaryHeap::from([1, 2, 3, 4]); - /// - /// // Print 1, 2, 3, 4 in arbitrary order - /// for x in heap.iter() { - /// println!("{x}"); - /// } - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - pub fn iter(&self) -> Iter<'_, T> { - Iter { iter: self.data.iter() } - } - - /// Returns an iterator which retrieves elements in heap order. - /// This method consumes the original heap. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// #![feature(binary_heap_into_iter_sorted)] - /// use std::collections::BinaryHeap; - /// let heap = BinaryHeap::from([1, 2, 3, 4, 5]); - /// - /// assert_eq!(heap.into_iter_sorted().take(2).collect::>(), [5, 4]); - /// ``` - #[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] - pub fn into_iter_sorted(self) -> IntoIterSorted { - IntoIterSorted { inner: self } - } - - /// Returns the greatest item in the binary heap, or `None` if it is empty. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// assert_eq!(heap.peek(), None); - /// - /// heap.push(1); - /// heap.push(5); - /// heap.push(2); - /// assert_eq!(heap.peek(), Some(&5)); - /// - /// ``` - /// - /// # Time complexity - /// - /// Cost is *O*(1) in the worst case. - #[must_use] - #[stable(feature = "rust1", since = "1.0.0")] - pub fn peek(&self) -> Option<&T> { - self.data.get(0) - } - - /// Returns the number of elements the binary heap can hold without reallocating. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::with_capacity(100); - /// assert!(heap.capacity() >= 100); - /// heap.push(4); - /// ``` - #[must_use] - #[stable(feature = "rust1", since = "1.0.0")] - pub fn capacity(&self) -> usize { - self.data.capacity() - } - - /// Reserves the minimum capacity for at least `additional` elements more than - /// the current length. Unlike [`reserve`], this will not - /// deliberately over-allocate to speculatively avoid frequent allocations. - /// After calling `reserve_exact`, capacity will be greater than or equal to - /// `self.len() + additional`. Does nothing if the capacity is already - /// sufficient. - /// - /// [`reserve`]: BinaryHeap::reserve - /// - /// # Panics - /// - /// Panics if the new capacity overflows [`usize`]. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// heap.reserve_exact(100); - /// assert!(heap.capacity() >= 100); - /// heap.push(4); - /// ``` - /// - /// [`reserve`]: BinaryHeap::reserve - #[stable(feature = "rust1", since = "1.0.0")] - pub fn reserve_exact(&mut self, additional: usize) { - self.data.reserve_exact(additional); - } - - /// Reserves capacity for at least `additional` elements more than the - /// current length. The allocator may reserve more space to speculatively - /// avoid frequent allocations. After calling `reserve`, - /// capacity will be greater than or equal to `self.len() + additional`. - /// Does nothing if capacity is already sufficient. - /// - /// # Panics - /// - /// Panics if the new capacity overflows [`usize`]. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// heap.reserve(100); - /// assert!(heap.capacity() >= 100); - /// heap.push(4); - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - pub fn reserve(&mut self, additional: usize) { - self.data.reserve(additional); - } - - /// Tries to reserve the minimum capacity for at least `additional` elements - /// more than the current length. Unlike [`try_reserve`], this will not - /// deliberately over-allocate to speculatively avoid frequent allocations. - /// After calling `try_reserve_exact`, capacity will be greater than or - /// equal to `self.len() + additional` if it returns `Ok(())`. - /// Does nothing if the capacity is already sufficient. - /// - /// Note that the allocator may give the collection more space than it - /// requests. Therefore, capacity can not be relied upon to be precisely - /// minimal. Prefer [`try_reserve`] if future insertions are expected. - /// - /// [`try_reserve`]: BinaryHeap::try_reserve - /// - /// # Errors - /// - /// If the capacity overflows, or the allocator reports a failure, then an error - /// is returned. - /// - /// # Examples - /// - /// ``` - /// use std::collections::BinaryHeap; - /// use std::collections::TryReserveError; - /// - /// fn find_max_slow(data: &[u32]) -> Result, TryReserveError> { - /// let mut heap = BinaryHeap::new(); - /// - /// // Pre-reserve the memory, exiting if we can't - /// heap.try_reserve_exact(data.len())?; - /// - /// // Now we know this can't OOM in the middle of our complex work - /// heap.extend(data.iter()); - /// - /// Ok(heap.pop()) - /// } - /// # find_max_slow(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?"); - /// ``` - #[stable(feature = "try_reserve_2", since = "1.63.0")] - pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> { - self.data.try_reserve_exact(additional) - } - - /// Tries to reserve capacity for at least `additional` elements more than the - /// current length. The allocator may reserve more space to speculatively - /// avoid frequent allocations. After calling `try_reserve`, capacity will be - /// greater than or equal to `self.len() + additional` if it returns - /// `Ok(())`. Does nothing if capacity is already sufficient. This method - /// preserves the contents even if an error occurs. - /// - /// # Errors - /// - /// If the capacity overflows, or the allocator reports a failure, then an error - /// is returned. - /// - /// # Examples - /// - /// ``` - /// use std::collections::BinaryHeap; - /// use std::collections::TryReserveError; - /// - /// fn find_max_slow(data: &[u32]) -> Result, TryReserveError> { - /// let mut heap = BinaryHeap::new(); - /// - /// // Pre-reserve the memory, exiting if we can't - /// heap.try_reserve(data.len())?; - /// - /// // Now we know this can't OOM in the middle of our complex work - /// heap.extend(data.iter()); - /// - /// Ok(heap.pop()) - /// } - /// # find_max_slow(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?"); - /// ``` - #[stable(feature = "try_reserve_2", since = "1.63.0")] - pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> { - self.data.try_reserve(additional) - } - - /// Discards as much additional capacity as possible. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap: BinaryHeap = BinaryHeap::with_capacity(100); - /// - /// assert!(heap.capacity() >= 100); - /// heap.shrink_to_fit(); - /// assert!(heap.capacity() == 0); - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - pub fn shrink_to_fit(&mut self) { - self.data.shrink_to_fit(); - } - - /// Discards capacity with a lower bound. - /// - /// The capacity will remain at least as large as both the length - /// and the supplied value. - /// - /// If the current capacity is less than the lower limit, this is a no-op. - /// - /// # Examples - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap: BinaryHeap = BinaryHeap::with_capacity(100); - /// - /// assert!(heap.capacity() >= 100); - /// heap.shrink_to(10); - /// assert!(heap.capacity() >= 10); - /// ``` - #[inline] - #[stable(feature = "shrink_to", since = "1.56.0")] - pub fn shrink_to(&mut self, min_capacity: usize) { - self.data.shrink_to(min_capacity) - } - - /// Returns a slice of all values in the underlying vector, in arbitrary - /// order. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// #![feature(binary_heap_as_slice)] - /// use std::collections::BinaryHeap; - /// use std::io::{self, Write}; - /// - /// let heap = BinaryHeap::from([1, 2, 3, 4, 5, 6, 7]); - /// - /// io::sink().write(heap.as_slice()).unwrap(); - /// ``` - #[must_use] - #[unstable(feature = "binary_heap_as_slice", issue = "83659")] - pub fn as_slice(&self) -> &[T] { - self.data.as_slice() - } - - /// Consumes the `BinaryHeap` and returns the underlying vector - /// in arbitrary order. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let heap = BinaryHeap::from([1, 2, 3, 4, 5, 6, 7]); - /// let vec = heap.into_vec(); - /// - /// // Will print in some order - /// for x in vec { - /// println!("{x}"); - /// } - /// ``` - #[must_use = "`self` will be dropped if the result is not used"] - #[stable(feature = "binary_heap_extras_15", since = "1.5.0")] - pub fn into_vec(self) -> Vec { - self.into() - } - - /// Returns the length of the binary heap. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let heap = BinaryHeap::from([1, 3]); - /// - /// assert_eq!(heap.len(), 2); - /// ``` - #[must_use] - #[stable(feature = "rust1", since = "1.0.0")] - pub fn len(&self) -> usize { - self.data.len() - } - - /// Checks if the binary heap is empty. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::new(); - /// - /// assert!(heap.is_empty()); - /// - /// heap.push(3); - /// heap.push(5); - /// heap.push(1); - /// - /// assert!(!heap.is_empty()); - /// ``` - #[must_use] - #[stable(feature = "rust1", since = "1.0.0")] - pub fn is_empty(&self) -> bool { - self.len() == 0 - } - - /// Clears the binary heap, returning an iterator over the removed elements - /// in arbitrary order. If the iterator is dropped before being fully - /// consumed, it drops the remaining elements in arbitrary order. - /// - /// The returned iterator keeps a mutable borrow on the heap to optimize - /// its implementation. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::from([1, 3]); - /// - /// assert!(!heap.is_empty()); - /// - /// for x in heap.drain() { - /// println!("{x}"); - /// } - /// - /// assert!(heap.is_empty()); - /// ``` - #[inline] - #[stable(feature = "drain", since = "1.6.0")] - pub fn drain(&mut self) -> Drain<'_, T> { - Drain { iter: self.data.drain(..) } - } - - /// Drops all items from the binary heap. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let mut heap = BinaryHeap::from([1, 3]); - /// - /// assert!(!heap.is_empty()); - /// - /// heap.clear(); - /// - /// assert!(heap.is_empty()); - /// ``` - #[stable(feature = "rust1", since = "1.0.0")] - pub fn clear(&mut self) { - self.drain(); - } -} - -/// Hole represents a hole in a slice i.e., an index without valid value -/// (because it was moved from or duplicated). -/// In drop, `Hole` will restore the slice by filling the hole -/// position with the value that was originally removed. -struct Hole<'a, T: 'a> { - data: &'a mut [T], - elt: ManuallyDrop, - pos: usize, -} - -impl<'a, T> Hole<'a, T> { - /// Create a new `Hole` at index `pos`. - /// - /// Unsafe because pos must be within the data slice. - #[inline] - unsafe fn new(data: &'a mut [T], pos: usize) -> Self { - debug_assert!(pos < data.len()); - // SAFE: pos should be inside the slice - let elt = unsafe { ptr::read(data.get_unchecked(pos)) }; - Hole { data, elt: ManuallyDrop::new(elt), pos } - } - - #[inline] - fn pos(&self) -> usize { - self.pos - } - - /// Returns a reference to the element removed. - #[inline] - fn element(&self) -> &T { - &self.elt - } - - /// Returns a reference to the element at `index`. - /// - /// Unsafe because index must be within the data slice and not equal to pos. - #[inline] - unsafe fn get(&self, index: usize) -> &T { - debug_assert!(index != self.pos); - debug_assert!(index < self.data.len()); - unsafe { self.data.get_unchecked(index) } - } - - /// Move hole to new location - /// - /// Unsafe because index must be within the data slice and not equal to pos. - #[inline] - unsafe fn move_to(&mut self, index: usize) { - debug_assert!(index != self.pos); - debug_assert!(index < self.data.len()); - unsafe { - let ptr = self.data.as_mut_ptr(); - let index_ptr: *const _ = ptr.add(index); - let hole_ptr = ptr.add(self.pos); - ptr::copy_nonoverlapping(index_ptr, hole_ptr, 1); - } - self.pos = index; - } -} - -impl Drop for Hole<'_, T> { - #[inline] - fn drop(&mut self) { - // fill the hole again - unsafe { - let pos = self.pos; - ptr::copy_nonoverlapping(&*self.elt, self.data.get_unchecked_mut(pos), 1); - } - } -} - -/// An iterator over the elements of a `BinaryHeap`. -/// -/// This `struct` is created by [`BinaryHeap::iter()`]. See its -/// documentation for more. -/// -/// [`iter`]: BinaryHeap::iter -#[must_use = "iterators are lazy and do nothing unless consumed"] -#[stable(feature = "rust1", since = "1.0.0")] -pub struct Iter<'a, T: 'a> { - iter: slice::Iter<'a, T>, -} - -#[stable(feature = "collection_debug", since = "1.17.0")] -impl fmt::Debug for Iter<'_, T> { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - f.debug_tuple("Iter").field(&self.iter.as_slice()).finish() - } -} - -// FIXME(#26925) Remove in favor of `#[derive(Clone)]` -#[stable(feature = "rust1", since = "1.0.0")] -impl Clone for Iter<'_, T> { - fn clone(&self) -> Self { - Iter { iter: self.iter.clone() } - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl<'a, T> Iterator for Iter<'a, T> { - type Item = &'a T; - - #[inline] - fn next(&mut self) -> Option<&'a T> { - self.iter.next() - } - - #[inline] - fn size_hint(&self) -> (usize, Option) { - self.iter.size_hint() - } - - #[inline] - fn last(self) -> Option<&'a T> { - self.iter.last() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl<'a, T> DoubleEndedIterator for Iter<'a, T> { - #[inline] - fn next_back(&mut self) -> Option<&'a T> { - self.iter.next_back() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl ExactSizeIterator for Iter<'_, T> { - fn is_empty(&self) -> bool { - self.iter.is_empty() - } -} - -#[stable(feature = "fused", since = "1.26.0")] -impl FusedIterator for Iter<'_, T> {} - -/// An owning iterator over the elements of a `BinaryHeap`. -/// -/// This `struct` is created by [`BinaryHeap::into_iter()`] -/// (provided by the [`IntoIterator`] trait). See its documentation for more. -/// -/// [`into_iter`]: BinaryHeap::into_iter -/// [`IntoIterator`]: core::iter::IntoIterator -#[stable(feature = "rust1", since = "1.0.0")] -#[derive(Clone)] -pub struct IntoIter { - iter: vec::IntoIter, -} - -#[stable(feature = "collection_debug", since = "1.17.0")] -impl fmt::Debug for IntoIter { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - f.debug_tuple("IntoIter").field(&self.iter.as_slice()).finish() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl Iterator for IntoIter { - type Item = T; - - #[inline] - fn next(&mut self) -> Option { - self.iter.next() - } - - #[inline] - fn size_hint(&self) -> (usize, Option) { - self.iter.size_hint() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl DoubleEndedIterator for IntoIter { - #[inline] - fn next_back(&mut self) -> Option { - self.iter.next_back() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl ExactSizeIterator for IntoIter { - fn is_empty(&self) -> bool { - self.iter.is_empty() - } -} - -#[stable(feature = "fused", since = "1.26.0")] -impl FusedIterator for IntoIter {} - -// In addition to the SAFETY invariants of the following three unsafe traits -// also refer to the vec::in_place_collect module documentation to get an overview -#[unstable(issue = "none", feature = "inplace_iteration")] -#[doc(hidden)] -unsafe impl SourceIter for IntoIter { - type Source = IntoIter; - - #[inline] - unsafe fn as_inner(&mut self) -> &mut Self::Source { - self - } -} - -#[unstable(issue = "none", feature = "inplace_iteration")] -#[doc(hidden)] -unsafe impl InPlaceIterable for IntoIter {} - -unsafe impl AsVecIntoIter for IntoIter { - type Item = I; - - fn as_into_iter(&mut self) -> &mut vec::IntoIter { - &mut self.iter - } -} - -#[must_use = "iterators are lazy and do nothing unless consumed"] -#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] -#[derive(Clone, Debug)] -pub struct IntoIterSorted { - inner: BinaryHeap, -} - -#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] -impl Iterator for IntoIterSorted { - type Item = T; - - #[inline] - fn next(&mut self) -> Option { - self.inner.pop() - } - - #[inline] - fn size_hint(&self) -> (usize, Option) { - let exact = self.inner.len(); - (exact, Some(exact)) - } -} - -#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] -impl ExactSizeIterator for IntoIterSorted {} - -#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] -impl FusedIterator for IntoIterSorted {} - -#[unstable(feature = "trusted_len", issue = "37572")] -unsafe impl TrustedLen for IntoIterSorted {} - -/// A draining iterator over the elements of a `BinaryHeap`. -/// -/// This `struct` is created by [`BinaryHeap::drain()`]. See its -/// documentation for more. -/// -/// [`drain`]: BinaryHeap::drain -#[stable(feature = "drain", since = "1.6.0")] -#[derive(Debug)] -pub struct Drain<'a, T: 'a> { - iter: vec::Drain<'a, T>, -} - -#[stable(feature = "drain", since = "1.6.0")] -impl Iterator for Drain<'_, T> { - type Item = T; - - #[inline] - fn next(&mut self) -> Option { - self.iter.next() - } - - #[inline] - fn size_hint(&self) -> (usize, Option) { - self.iter.size_hint() - } -} - -#[stable(feature = "drain", since = "1.6.0")] -impl DoubleEndedIterator for Drain<'_, T> { - #[inline] - fn next_back(&mut self) -> Option { - self.iter.next_back() - } -} - -#[stable(feature = "drain", since = "1.6.0")] -impl ExactSizeIterator for Drain<'_, T> { - fn is_empty(&self) -> bool { - self.iter.is_empty() - } -} - -#[stable(feature = "fused", since = "1.26.0")] -impl FusedIterator for Drain<'_, T> {} - -/// A draining iterator over the elements of a `BinaryHeap`. -/// -/// This `struct` is created by [`BinaryHeap::drain_sorted()`]. See its -/// documentation for more. -/// -/// [`drain_sorted`]: BinaryHeap::drain_sorted -#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] -#[derive(Debug)] -pub struct DrainSorted<'a, T: Ord> { - inner: &'a mut BinaryHeap, -} - -#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] -impl<'a, T: Ord> Drop for DrainSorted<'a, T> { - /// Removes heap elements in heap order. - fn drop(&mut self) { - struct DropGuard<'r, 'a, T: Ord>(&'r mut DrainSorted<'a, T>); - - impl<'r, 'a, T: Ord> Drop for DropGuard<'r, 'a, T> { - fn drop(&mut self) { - while self.0.inner.pop().is_some() {} - } - } - - while let Some(item) = self.inner.pop() { - let guard = DropGuard(self); - drop(item); - mem::forget(guard); - } - } -} - -#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] -impl Iterator for DrainSorted<'_, T> { - type Item = T; - - #[inline] - fn next(&mut self) -> Option { - self.inner.pop() - } - - #[inline] - fn size_hint(&self) -> (usize, Option) { - let exact = self.inner.len(); - (exact, Some(exact)) - } -} - -#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] -impl ExactSizeIterator for DrainSorted<'_, T> {} - -#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] -impl FusedIterator for DrainSorted<'_, T> {} - -#[unstable(feature = "trusted_len", issue = "37572")] -unsafe impl TrustedLen for DrainSorted<'_, T> {} - -#[stable(feature = "binary_heap_extras_15", since = "1.5.0")] -impl From> for BinaryHeap { - /// Converts a `Vec` into a `BinaryHeap`. - /// - /// This conversion happens in-place, and has *O*(*n*) time complexity. - fn from(vec: Vec) -> BinaryHeap { - let mut heap = BinaryHeap { data: vec }; - heap.rebuild(); - heap - } -} - -#[stable(feature = "std_collections_from_array", since = "1.56.0")] -impl From<[T; N]> for BinaryHeap { - /// ``` - /// use std::collections::BinaryHeap; - /// - /// let mut h1 = BinaryHeap::from([1, 4, 2, 3]); - /// let mut h2: BinaryHeap<_> = [1, 4, 2, 3].into(); - /// while let Some((a, b)) = h1.pop().zip(h2.pop()) { - /// assert_eq!(a, b); - /// } - /// ``` - fn from(arr: [T; N]) -> Self { - Self::from_iter(arr) - } -} - -#[stable(feature = "binary_heap_extras_15", since = "1.5.0")] -impl From> for Vec { - /// Converts a `BinaryHeap` into a `Vec`. - /// - /// This conversion requires no data movement or allocation, and has - /// constant time complexity. - fn from(heap: BinaryHeap) -> Vec { - heap.data - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl FromIterator for BinaryHeap { - fn from_iter>(iter: I) -> BinaryHeap { - BinaryHeap::from(iter.into_iter().collect::>()) - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl IntoIterator for BinaryHeap { - type Item = T; - type IntoIter = IntoIter; - - /// Creates a consuming iterator, that is, one that moves each value out of - /// the binary heap in arbitrary order. The binary heap cannot be used - /// after calling this. - /// - /// # Examples - /// - /// Basic usage: - /// - /// ``` - /// use std::collections::BinaryHeap; - /// let heap = BinaryHeap::from([1, 2, 3, 4]); - /// - /// // Print 1, 2, 3, 4 in arbitrary order - /// for x in heap.into_iter() { - /// // x has type i32, not &i32 - /// println!("{x}"); - /// } - /// ``` - fn into_iter(self) -> IntoIter { - IntoIter { iter: self.data.into_iter() } - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl<'a, T> IntoIterator for &'a BinaryHeap { - type Item = &'a T; - type IntoIter = Iter<'a, T>; - - fn into_iter(self) -> Iter<'a, T> { - self.iter() - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -impl Extend for BinaryHeap { - #[inline] - fn extend>(&mut self, iter: I) { - >::spec_extend(self, iter); - } - - #[inline] - fn extend_one(&mut self, item: T) { - self.push(item); - } - - #[inline] - fn extend_reserve(&mut self, additional: usize) { - self.reserve(additional); - } -} - -impl> SpecExtend for BinaryHeap { - default fn spec_extend(&mut self, iter: I) { - self.extend_desugared(iter.into_iter()); - } -} - -impl SpecExtend> for BinaryHeap { - fn spec_extend(&mut self, ref mut other: Vec) { - let start = self.data.len(); - self.data.append(other); - self.rebuild_tail(start); - } -} - -impl SpecExtend> for BinaryHeap { - fn spec_extend(&mut self, ref mut other: BinaryHeap) { - self.append(other); - } -} - -impl BinaryHeap { - fn extend_desugared>(&mut self, iter: I) { - let iterator = iter.into_iter(); - let (lower, _) = iterator.size_hint(); - - self.reserve(lower); - - iterator.for_each(move |elem| self.push(elem)); - } -} - -#[stable(feature = "extend_ref", since = "1.2.0")] -impl<'a, T: 'a + Ord + Copy> Extend<&'a T> for BinaryHeap { - fn extend>(&mut self, iter: I) { - self.extend(iter.into_iter().cloned()); - } - - #[inline] - fn extend_one(&mut self, &item: &'a T) { - self.push(item); - } - - #[inline] - fn extend_reserve(&mut self, additional: usize) { - self.reserve(additional); - } -} diff --git a/library/alloc/src/collections/binary_heap/mod.rs b/library/alloc/src/collections/binary_heap/mod.rs new file mode 100644 index 000000000..0b73b1af4 --- /dev/null +++ b/library/alloc/src/collections/binary_heap/mod.rs @@ -0,0 +1,1766 @@ +//! A priority queue implemented with a binary heap. +//! +//! Insertion and popping the largest element have *O*(log(*n*)) time complexity. +//! Checking the largest element is *O*(1). Converting a vector to a binary heap +//! can be done in-place, and has *O*(*n*) complexity. A binary heap can also be +//! converted to a sorted vector in-place, allowing it to be used for an *O*(*n* * log(*n*)) +//! in-place heapsort. +//! +//! # Examples +//! +//! This is a larger example that implements [Dijkstra's algorithm][dijkstra] +//! to solve the [shortest path problem][sssp] on a [directed graph][dir_graph]. +//! It shows how to use [`BinaryHeap`] with custom types. +//! +//! [dijkstra]: https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm +//! [sssp]: https://en.wikipedia.org/wiki/Shortest_path_problem +//! [dir_graph]: https://en.wikipedia.org/wiki/Directed_graph +//! +//! ``` +//! use std::cmp::Ordering; +//! use std::collections::BinaryHeap; +//! +//! #[derive(Copy, Clone, Eq, PartialEq)] +//! struct State { +//! cost: usize, +//! position: usize, +//! } +//! +//! // The priority queue depends on `Ord`. +//! // Explicitly implement the trait so the queue becomes a min-heap +//! // instead of a max-heap. +//! impl Ord for State { +//! fn cmp(&self, other: &Self) -> Ordering { +//! // Notice that the we flip the ordering on costs. +//! // In case of a tie we compare positions - this step is necessary +//! // to make implementations of `PartialEq` and `Ord` consistent. +//! other.cost.cmp(&self.cost) +//! .then_with(|| self.position.cmp(&other.position)) +//! } +//! } +//! +//! // `PartialOrd` needs to be implemented as well. +//! impl PartialOrd for State { +//! fn partial_cmp(&self, other: &Self) -> Option { +//! Some(self.cmp(other)) +//! } +//! } +//! +//! // Each node is represented as a `usize`, for a shorter implementation. +//! struct Edge { +//! node: usize, +//! cost: usize, +//! } +//! +//! // Dijkstra's shortest path algorithm. +//! +//! // Start at `start` and use `dist` to track the current shortest distance +//! // to each node. This implementation isn't memory-efficient as it may leave duplicate +//! // nodes in the queue. It also uses `usize::MAX` as a sentinel value, +//! // for a simpler implementation. +//! fn shortest_path(adj_list: &Vec>, start: usize, goal: usize) -> Option { +//! // dist[node] = current shortest distance from `start` to `node` +//! let mut dist: Vec<_> = (0..adj_list.len()).map(|_| usize::MAX).collect(); +//! +//! let mut heap = BinaryHeap::new(); +//! +//! // We're at `start`, with a zero cost +//! dist[start] = 0; +//! heap.push(State { cost: 0, position: start }); +//! +//! // Examine the frontier with lower cost nodes first (min-heap) +//! while let Some(State { cost, position }) = heap.pop() { +//! // Alternatively we could have continued to find all shortest paths +//! if position == goal { return Some(cost); } +//! +//! // Important as we may have already found a better way +//! if cost > dist[position] { continue; } +//! +//! // For each node we can reach, see if we can find a way with +//! // a lower cost going through this node +//! for edge in &adj_list[position] { +//! let next = State { cost: cost + edge.cost, position: edge.node }; +//! +//! // If so, add it to the frontier and continue +//! if next.cost < dist[next.position] { +//! heap.push(next); +//! // Relaxation, we have now found a better way +//! dist[next.position] = next.cost; +//! } +//! } +//! } +//! +//! // Goal not reachable +//! None +//! } +//! +//! fn main() { +//! // This is the directed graph we're going to use. +//! // The node numbers correspond to the different states, +//! // and the edge weights symbolize the cost of moving +//! // from one node to another. +//! // Note that the edges are one-way. +//! // +//! // 7 +//! // +-----------------+ +//! // | | +//! // v 1 2 | 2 +//! // 0 -----> 1 -----> 3 ---> 4 +//! // | ^ ^ ^ +//! // | | 1 | | +//! // | | | 3 | 1 +//! // +------> 2 -------+ | +//! // 10 | | +//! // +---------------+ +//! // +//! // The graph is represented as an adjacency list where each index, +//! // corresponding to a node value, has a list of outgoing edges. +//! // Chosen for its efficiency. +//! let graph = vec![ +//! // Node 0 +//! vec![Edge { node: 2, cost: 10 }, +//! Edge { node: 1, cost: 1 }], +//! // Node 1 +//! vec![Edge { node: 3, cost: 2 }], +//! // Node 2 +//! vec![Edge { node: 1, cost: 1 }, +//! Edge { node: 3, cost: 3 }, +//! Edge { node: 4, cost: 1 }], +//! // Node 3 +//! vec![Edge { node: 0, cost: 7 }, +//! Edge { node: 4, cost: 2 }], +//! // Node 4 +//! vec![]]; +//! +//! assert_eq!(shortest_path(&graph, 0, 1), Some(1)); +//! assert_eq!(shortest_path(&graph, 0, 3), Some(3)); +//! assert_eq!(shortest_path(&graph, 3, 0), Some(7)); +//! assert_eq!(shortest_path(&graph, 0, 4), Some(5)); +//! assert_eq!(shortest_path(&graph, 4, 0), None); +//! } +//! ``` + +#![allow(missing_docs)] +#![stable(feature = "rust1", since = "1.0.0")] + +use core::fmt; +use core::iter::{FromIterator, FusedIterator, InPlaceIterable, SourceIter, TrustedLen}; +use core::mem::{self, swap, ManuallyDrop}; +use core::num::NonZeroUsize; +use core::ops::{Deref, DerefMut}; +use core::ptr; + +use crate::collections::TryReserveError; +use crate::slice; +use crate::vec::{self, AsVecIntoIter, Vec}; + +use super::SpecExtend; + +#[cfg(test)] +mod tests; + +/// A priority queue implemented with a binary heap. +/// +/// This will be a max-heap. +/// +/// It is a logic error for an item to be modified in such a way that the +/// item's ordering relative to any other item, as determined by the [`Ord`] +/// trait, changes while it is in the heap. This is normally only possible +/// through interior mutability, global state, I/O, or unsafe code. The +/// behavior resulting from such a logic error is not specified, but will +/// be encapsulated to the `BinaryHeap` that observed the logic error and not +/// result in undefined behavior. This could include panics, incorrect results, +/// aborts, memory leaks, and non-termination. +/// +/// As long as no elements change their relative order while being in the heap +/// as described above, the API of `BinaryHeap` guarantees that the heap +/// invariant remains intact i.e. its methods all behave as documented. For +/// example if a method is documented as iterating in sorted order, that's +/// guaranteed to work as long as elements in the heap have not changed order, +/// even in the presence of closures getting unwinded out of, iterators getting +/// leaked, and similar foolishness. +/// +/// # Examples +/// +/// ``` +/// use std::collections::BinaryHeap; +/// +/// // Type inference lets us omit an explicit type signature (which +/// // would be `BinaryHeap` in this example). +/// let mut heap = BinaryHeap::new(); +/// +/// // We can use peek to look at the next item in the heap. In this case, +/// // there's no items in there yet so we get None. +/// assert_eq!(heap.peek(), None); +/// +/// // Let's add some scores... +/// heap.push(1); +/// heap.push(5); +/// heap.push(2); +/// +/// // Now peek shows the most important item in the heap. +/// assert_eq!(heap.peek(), Some(&5)); +/// +/// // We can check the length of a heap. +/// assert_eq!(heap.len(), 3); +/// +/// // We can iterate over the items in the heap, although they are returned in +/// // a random order. +/// for x in &heap { +/// println!("{x}"); +/// } +/// +/// // If we instead pop these scores, they should come back in order. +/// assert_eq!(heap.pop(), Some(5)); +/// assert_eq!(heap.pop(), Some(2)); +/// assert_eq!(heap.pop(), Some(1)); +/// assert_eq!(heap.pop(), None); +/// +/// // We can clear the heap of any remaining items. +/// heap.clear(); +/// +/// // The heap should now be empty. +/// assert!(heap.is_empty()) +/// ``` +/// +/// A `BinaryHeap` with a known list of items can be initialized from an array: +/// +/// ``` +/// use std::collections::BinaryHeap; +/// +/// let heap = BinaryHeap::from([1, 5, 2]); +/// ``` +/// +/// ## Min-heap +/// +/// Either [`core::cmp::Reverse`] or a custom [`Ord`] implementation can be used to +/// make `BinaryHeap` a min-heap. This makes `heap.pop()` return the smallest +/// value instead of the greatest one. +/// +/// ``` +/// use std::collections::BinaryHeap; +/// use std::cmp::Reverse; +/// +/// let mut heap = BinaryHeap::new(); +/// +/// // Wrap values in `Reverse` +/// heap.push(Reverse(1)); +/// heap.push(Reverse(5)); +/// heap.push(Reverse(2)); +/// +/// // If we pop these scores now, they should come back in the reverse order. +/// assert_eq!(heap.pop(), Some(Reverse(1))); +/// assert_eq!(heap.pop(), Some(Reverse(2))); +/// assert_eq!(heap.pop(), Some(Reverse(5))); +/// assert_eq!(heap.pop(), None); +/// ``` +/// +/// # Time complexity +/// +/// | [push] | [pop] | [peek]/[peek\_mut] | +/// |---------|---------------|--------------------| +/// | *O*(1)~ | *O*(log(*n*)) | *O*(1) | +/// +/// The value for `push` is an expected cost; the method documentation gives a +/// more detailed analysis. +/// +/// [`core::cmp::Reverse`]: core::cmp::Reverse +/// [`Ord`]: core::cmp::Ord +/// [`Cell`]: core::cell::Cell +/// [`RefCell`]: core::cell::RefCell +/// [push]: BinaryHeap::push +/// [pop]: BinaryHeap::pop +/// [peek]: BinaryHeap::peek +/// [peek\_mut]: BinaryHeap::peek_mut +#[stable(feature = "rust1", since = "1.0.0")] +#[cfg_attr(not(test), rustc_diagnostic_item = "BinaryHeap")] +pub struct BinaryHeap { + data: Vec, +} + +/// Structure wrapping a mutable reference to the greatest item on a +/// `BinaryHeap`. +/// +/// This `struct` is created by the [`peek_mut`] method on [`BinaryHeap`]. See +/// its documentation for more. +/// +/// [`peek_mut`]: BinaryHeap::peek_mut +#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] +pub struct PeekMut<'a, T: 'a + Ord> { + heap: &'a mut BinaryHeap, + // If a set_len + sift_down are required, this is Some. If a &mut T has not + // yet been exposed to peek_mut()'s caller, it's None. + original_len: Option, +} + +#[stable(feature = "collection_debug", since = "1.17.0")] +impl fmt::Debug for PeekMut<'_, T> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_tuple("PeekMut").field(&self.heap.data[0]).finish() + } +} + +#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] +impl Drop for PeekMut<'_, T> { + fn drop(&mut self) { + if let Some(original_len) = self.original_len { + // SAFETY: That's how many elements were in the Vec at the time of + // the PeekMut::deref_mut call, and therefore also at the time of + // the BinaryHeap::peek_mut call. Since the PeekMut did not end up + // getting leaked, we are now undoing the leak amplification that + // the DerefMut prepared for. + unsafe { self.heap.data.set_len(original_len.get()) }; + + // SAFETY: PeekMut is only instantiated for non-empty heaps. + unsafe { self.heap.sift_down(0) }; + } + } +} + +#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] +impl Deref for PeekMut<'_, T> { + type Target = T; + fn deref(&self) -> &T { + debug_assert!(!self.heap.is_empty()); + // SAFE: PeekMut is only instantiated for non-empty heaps + unsafe { self.heap.data.get_unchecked(0) } + } +} + +#[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] +impl DerefMut for PeekMut<'_, T> { + fn deref_mut(&mut self) -> &mut T { + debug_assert!(!self.heap.is_empty()); + + let len = self.heap.len(); + if len > 1 { + // Here we preemptively leak all the rest of the underlying vector + // after the currently max element. If the caller mutates the &mut T + // we're about to give them, and then leaks the PeekMut, all these + // elements will remain leaked. If they don't leak the PeekMut, then + // either Drop or PeekMut::pop will un-leak the vector elements. + // + // This is technique is described throughout several other places in + // the standard library as "leak amplification". + unsafe { + // SAFETY: len > 1 so len != 0. + self.original_len = Some(NonZeroUsize::new_unchecked(len)); + // SAFETY: len > 1 so all this does for now is leak elements, + // which is safe. + self.heap.data.set_len(1); + } + } + + // SAFE: PeekMut is only instantiated for non-empty heaps + unsafe { self.heap.data.get_unchecked_mut(0) } + } +} + +impl<'a, T: Ord> PeekMut<'a, T> { + /// Removes the peeked value from the heap and returns it. + #[stable(feature = "binary_heap_peek_mut_pop", since = "1.18.0")] + pub fn pop(mut this: PeekMut<'a, T>) -> T { + if let Some(original_len) = this.original_len.take() { + // SAFETY: This is how many elements were in the Vec at the time of + // the BinaryHeap::peek_mut call. + unsafe { this.heap.data.set_len(original_len.get()) }; + + // Unlike in Drop, here we don't also need to do a sift_down even if + // the caller could've mutated the element. It is removed from the + // heap on the next line and pop() is not sensitive to its value. + } + this.heap.pop().unwrap() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Clone for BinaryHeap { + fn clone(&self) -> Self { + BinaryHeap { data: self.data.clone() } + } + + fn clone_from(&mut self, source: &Self) { + self.data.clone_from(&source.data); + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Default for BinaryHeap { + /// Creates an empty `BinaryHeap`. + #[inline] + fn default() -> BinaryHeap { + BinaryHeap::new() + } +} + +#[stable(feature = "binaryheap_debug", since = "1.4.0")] +impl fmt::Debug for BinaryHeap { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_list().entries(self.iter()).finish() + } +} + +impl BinaryHeap { + /// Creates an empty `BinaryHeap` as a max-heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// heap.push(4); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + pub fn new() -> BinaryHeap { + BinaryHeap { data: vec![] } + } + + /// Creates an empty `BinaryHeap` with at least the specified capacity. + /// + /// The binary heap will be able to hold at least `capacity` elements without + /// reallocating. This method is allowed to allocate for more elements than + /// `capacity`. If `capacity` is 0, the binary heap will not allocate. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::with_capacity(10); + /// heap.push(4); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + #[must_use] + pub fn with_capacity(capacity: usize) -> BinaryHeap { + BinaryHeap { data: Vec::with_capacity(capacity) } + } + + /// Returns a mutable reference to the greatest item in the binary heap, or + /// `None` if it is empty. + /// + /// Note: If the `PeekMut` value is leaked, some heap elements might get + /// leaked along with it, but the remaining elements will remain a valid + /// heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// assert!(heap.peek_mut().is_none()); + /// + /// heap.push(1); + /// heap.push(5); + /// heap.push(2); + /// { + /// let mut val = heap.peek_mut().unwrap(); + /// *val = 0; + /// } + /// assert_eq!(heap.peek(), Some(&2)); + /// ``` + /// + /// # Time complexity + /// + /// If the item is modified then the worst case time complexity is *O*(log(*n*)), + /// otherwise it's *O*(1). + #[stable(feature = "binary_heap_peek_mut", since = "1.12.0")] + pub fn peek_mut(&mut self) -> Option> { + if self.is_empty() { None } else { Some(PeekMut { heap: self, original_len: None }) } + } + + /// Removes the greatest item from the binary heap and returns it, or `None` if it + /// is empty. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::from([1, 3]); + /// + /// assert_eq!(heap.pop(), Some(3)); + /// assert_eq!(heap.pop(), Some(1)); + /// assert_eq!(heap.pop(), None); + /// ``` + /// + /// # Time complexity + /// + /// The worst case cost of `pop` on a heap containing *n* elements is *O*(log(*n*)). + #[stable(feature = "rust1", since = "1.0.0")] + pub fn pop(&mut self) -> Option { + self.data.pop().map(|mut item| { + if !self.is_empty() { + swap(&mut item, &mut self.data[0]); + // SAFETY: !self.is_empty() means that self.len() > 0 + unsafe { self.sift_down_to_bottom(0) }; + } + item + }) + } + + /// Pushes an item onto the binary heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// heap.push(3); + /// heap.push(5); + /// heap.push(1); + /// + /// assert_eq!(heap.len(), 3); + /// assert_eq!(heap.peek(), Some(&5)); + /// ``` + /// + /// # Time complexity + /// + /// The expected cost of `push`, averaged over every possible ordering of + /// the elements being pushed, and over a sufficiently large number of + /// pushes, is *O*(1). This is the most meaningful cost metric when pushing + /// elements that are *not* already in any sorted pattern. + /// + /// The time complexity degrades if elements are pushed in predominantly + /// ascending order. In the worst case, elements are pushed in ascending + /// sorted order and the amortized cost per push is *O*(log(*n*)) against a heap + /// containing *n* elements. + /// + /// The worst case cost of a *single* call to `push` is *O*(*n*). The worst case + /// occurs when capacity is exhausted and needs a resize. The resize cost + /// has been amortized in the previous figures. + #[stable(feature = "rust1", since = "1.0.0")] + pub fn push(&mut self, item: T) { + let old_len = self.len(); + self.data.push(item); + // SAFETY: Since we pushed a new item it means that + // old_len = self.len() - 1 < self.len() + unsafe { self.sift_up(0, old_len) }; + } + + /// Consumes the `BinaryHeap` and returns a vector in sorted + /// (ascending) order. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// + /// let mut heap = BinaryHeap::from([1, 2, 4, 5, 7]); + /// heap.push(6); + /// heap.push(3); + /// + /// let vec = heap.into_sorted_vec(); + /// assert_eq!(vec, [1, 2, 3, 4, 5, 6, 7]); + /// ``` + #[must_use = "`self` will be dropped if the result is not used"] + #[stable(feature = "binary_heap_extras_15", since = "1.5.0")] + pub fn into_sorted_vec(mut self) -> Vec { + let mut end = self.len(); + while end > 1 { + end -= 1; + // SAFETY: `end` goes from `self.len() - 1` to 1 (both included), + // so it's always a valid index to access. + // It is safe to access index 0 (i.e. `ptr`), because + // 1 <= end < self.len(), which means self.len() >= 2. + unsafe { + let ptr = self.data.as_mut_ptr(); + ptr::swap(ptr, ptr.add(end)); + } + // SAFETY: `end` goes from `self.len() - 1` to 1 (both included) so: + // 0 < 1 <= end <= self.len() - 1 < self.len() + // Which means 0 < end and end < self.len(). + unsafe { self.sift_down_range(0, end) }; + } + self.into_vec() + } + + // The implementations of sift_up and sift_down use unsafe blocks in + // order to move an element out of the vector (leaving behind a + // hole), shift along the others and move the removed element back into the + // vector at the final location of the hole. + // The `Hole` type is used to represent this, and make sure + // the hole is filled back at the end of its scope, even on panic. + // Using a hole reduces the constant factor compared to using swaps, + // which involves twice as many moves. + + /// # Safety + /// + /// The caller must guarantee that `pos < self.len()`. + unsafe fn sift_up(&mut self, start: usize, pos: usize) -> usize { + // Take out the value at `pos` and create a hole. + // SAFETY: The caller guarantees that pos < self.len() + let mut hole = unsafe { Hole::new(&mut self.data, pos) }; + + while hole.pos() > start { + let parent = (hole.pos() - 1) / 2; + + // SAFETY: hole.pos() > start >= 0, which means hole.pos() > 0 + // and so hole.pos() - 1 can't underflow. + // This guarantees that parent < hole.pos() so + // it's a valid index and also != hole.pos(). + if hole.element() <= unsafe { hole.get(parent) } { + break; + } + + // SAFETY: Same as above + unsafe { hole.move_to(parent) }; + } + + hole.pos() + } + + /// Take an element at `pos` and move it down the heap, + /// while its children are larger. + /// + /// # Safety + /// + /// The caller must guarantee that `pos < end <= self.len()`. + unsafe fn sift_down_range(&mut self, pos: usize, end: usize) { + // SAFETY: The caller guarantees that pos < end <= self.len(). + let mut hole = unsafe { Hole::new(&mut self.data, pos) }; + let mut child = 2 * hole.pos() + 1; + + // Loop invariant: child == 2 * hole.pos() + 1. + while child <= end.saturating_sub(2) { + // compare with the greater of the two children + // SAFETY: child < end - 1 < self.len() and + // child + 1 < end <= self.len(), so they're valid indexes. + // child == 2 * hole.pos() + 1 != hole.pos() and + // child + 1 == 2 * hole.pos() + 2 != hole.pos(). + // FIXME: 2 * hole.pos() + 1 or 2 * hole.pos() + 2 could overflow + // if T is a ZST + child += unsafe { hole.get(child) <= hole.get(child + 1) } as usize; + + // if we are already in order, stop. + // SAFETY: child is now either the old child or the old child+1 + // We already proven that both are < self.len() and != hole.pos() + if hole.element() >= unsafe { hole.get(child) } { + return; + } + + // SAFETY: same as above. + unsafe { hole.move_to(child) }; + child = 2 * hole.pos() + 1; + } + + // SAFETY: && short circuit, which means that in the + // second condition it's already true that child == end - 1 < self.len(). + if child == end - 1 && hole.element() < unsafe { hole.get(child) } { + // SAFETY: child is already proven to be a valid index and + // child == 2 * hole.pos() + 1 != hole.pos(). + unsafe { hole.move_to(child) }; + } + } + + /// # Safety + /// + /// The caller must guarantee that `pos < self.len()`. + unsafe fn sift_down(&mut self, pos: usize) { + let len = self.len(); + // SAFETY: pos < len is guaranteed by the caller and + // obviously len = self.len() <= self.len(). + unsafe { self.sift_down_range(pos, len) }; + } + + /// Take an element at `pos` and move it all the way down the heap, + /// then sift it up to its position. + /// + /// Note: This is faster when the element is known to be large / should + /// be closer to the bottom. + /// + /// # Safety + /// + /// The caller must guarantee that `pos < self.len()`. + unsafe fn sift_down_to_bottom(&mut self, mut pos: usize) { + let end = self.len(); + let start = pos; + + // SAFETY: The caller guarantees that pos < self.len(). + let mut hole = unsafe { Hole::new(&mut self.data, pos) }; + let mut child = 2 * hole.pos() + 1; + + // Loop invariant: child == 2 * hole.pos() + 1. + while child <= end.saturating_sub(2) { + // SAFETY: child < end - 1 < self.len() and + // child + 1 < end <= self.len(), so they're valid indexes. + // child == 2 * hole.pos() + 1 != hole.pos() and + // child + 1 == 2 * hole.pos() + 2 != hole.pos(). + // FIXME: 2 * hole.pos() + 1 or 2 * hole.pos() + 2 could overflow + // if T is a ZST + child += unsafe { hole.get(child) <= hole.get(child + 1) } as usize; + + // SAFETY: Same as above + unsafe { hole.move_to(child) }; + child = 2 * hole.pos() + 1; + } + + if child == end - 1 { + // SAFETY: child == end - 1 < self.len(), so it's a valid index + // and child == 2 * hole.pos() + 1 != hole.pos(). + unsafe { hole.move_to(child) }; + } + pos = hole.pos(); + drop(hole); + + // SAFETY: pos is the position in the hole and was already proven + // to be a valid index. + unsafe { self.sift_up(start, pos) }; + } + + /// Rebuild assuming data[0..start] is still a proper heap. + fn rebuild_tail(&mut self, start: usize) { + if start == self.len() { + return; + } + + let tail_len = self.len() - start; + + #[inline(always)] + fn log2_fast(x: usize) -> usize { + (usize::BITS - x.leading_zeros() - 1) as usize + } + + // `rebuild` takes O(self.len()) operations + // and about 2 * self.len() comparisons in the worst case + // while repeating `sift_up` takes O(tail_len * log(start)) operations + // and about 1 * tail_len * log_2(start) comparisons in the worst case, + // assuming start >= tail_len. For larger heaps, the crossover point + // no longer follows this reasoning and was determined empirically. + let better_to_rebuild = if start < tail_len { + true + } else if self.len() <= 2048 { + 2 * self.len() < tail_len * log2_fast(start) + } else { + 2 * self.len() < tail_len * 11 + }; + + if better_to_rebuild { + self.rebuild(); + } else { + for i in start..self.len() { + // SAFETY: The index `i` is always less than self.len(). + unsafe { self.sift_up(0, i) }; + } + } + } + + fn rebuild(&mut self) { + let mut n = self.len() / 2; + while n > 0 { + n -= 1; + // SAFETY: n starts from self.len() / 2 and goes down to 0. + // The only case when !(n < self.len()) is if + // self.len() == 0, but it's ruled out by the loop condition. + unsafe { self.sift_down(n) }; + } + } + + /// Moves all the elements of `other` into `self`, leaving `other` empty. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// + /// let mut a = BinaryHeap::from([-10, 1, 2, 3, 3]); + /// let mut b = BinaryHeap::from([-20, 5, 43]); + /// + /// a.append(&mut b); + /// + /// assert_eq!(a.into_sorted_vec(), [-20, -10, 1, 2, 3, 3, 5, 43]); + /// assert!(b.is_empty()); + /// ``` + #[stable(feature = "binary_heap_append", since = "1.11.0")] + pub fn append(&mut self, other: &mut Self) { + if self.len() < other.len() { + swap(self, other); + } + + let start = self.data.len(); + + self.data.append(&mut other.data); + + self.rebuild_tail(start); + } + + /// Clears the binary heap, returning an iterator over the removed elements + /// in heap order. If the iterator is dropped before being fully consumed, + /// it drops the remaining elements in heap order. + /// + /// The returned iterator keeps a mutable borrow on the heap to optimize + /// its implementation. + /// + /// Note: + /// * `.drain_sorted()` is *O*(*n* \* log(*n*)); much slower than `.drain()`. + /// You should use the latter for most cases. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// #![feature(binary_heap_drain_sorted)] + /// use std::collections::BinaryHeap; + /// + /// let mut heap = BinaryHeap::from([1, 2, 3, 4, 5]); + /// assert_eq!(heap.len(), 5); + /// + /// drop(heap.drain_sorted()); // removes all elements in heap order + /// assert_eq!(heap.len(), 0); + /// ``` + #[inline] + #[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] + pub fn drain_sorted(&mut self) -> DrainSorted<'_, T> { + DrainSorted { inner: self } + } + + /// Retains only the elements specified by the predicate. + /// + /// In other words, remove all elements `e` for which `f(&e)` returns + /// `false`. The elements are visited in unsorted (and unspecified) order. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// #![feature(binary_heap_retain)] + /// use std::collections::BinaryHeap; + /// + /// let mut heap = BinaryHeap::from([-10, -5, 1, 2, 4, 13]); + /// + /// heap.retain(|x| x % 2 == 0); // only keep even numbers + /// + /// assert_eq!(heap.into_sorted_vec(), [-10, 2, 4]) + /// ``` + #[unstable(feature = "binary_heap_retain", issue = "71503")] + pub fn retain(&mut self, mut f: F) + where + F: FnMut(&T) -> bool, + { + let mut first_removed = self.len(); + let mut i = 0; + self.data.retain(|e| { + let keep = f(e); + if !keep && i < first_removed { + first_removed = i; + } + i += 1; + keep + }); + // data[0..first_removed] is untouched, so we only need to rebuild the tail: + self.rebuild_tail(first_removed); + } +} + +impl BinaryHeap { + /// Returns an iterator visiting all values in the underlying vector, in + /// arbitrary order. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let heap = BinaryHeap::from([1, 2, 3, 4]); + /// + /// // Print 1, 2, 3, 4 in arbitrary order + /// for x in heap.iter() { + /// println!("{x}"); + /// } + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn iter(&self) -> Iter<'_, T> { + Iter { iter: self.data.iter() } + } + + /// Returns an iterator which retrieves elements in heap order. + /// This method consumes the original heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// #![feature(binary_heap_into_iter_sorted)] + /// use std::collections::BinaryHeap; + /// let heap = BinaryHeap::from([1, 2, 3, 4, 5]); + /// + /// assert_eq!(heap.into_iter_sorted().take(2).collect::>(), [5, 4]); + /// ``` + #[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] + pub fn into_iter_sorted(self) -> IntoIterSorted { + IntoIterSorted { inner: self } + } + + /// Returns the greatest item in the binary heap, or `None` if it is empty. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// assert_eq!(heap.peek(), None); + /// + /// heap.push(1); + /// heap.push(5); + /// heap.push(2); + /// assert_eq!(heap.peek(), Some(&5)); + /// + /// ``` + /// + /// # Time complexity + /// + /// Cost is *O*(1) in the worst case. + #[must_use] + #[stable(feature = "rust1", since = "1.0.0")] + pub fn peek(&self) -> Option<&T> { + self.data.get(0) + } + + /// Returns the number of elements the binary heap can hold without reallocating. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::with_capacity(100); + /// assert!(heap.capacity() >= 100); + /// heap.push(4); + /// ``` + #[must_use] + #[stable(feature = "rust1", since = "1.0.0")] + pub fn capacity(&self) -> usize { + self.data.capacity() + } + + /// Reserves the minimum capacity for at least `additional` elements more than + /// the current length. Unlike [`reserve`], this will not + /// deliberately over-allocate to speculatively avoid frequent allocations. + /// After calling `reserve_exact`, capacity will be greater than or equal to + /// `self.len() + additional`. Does nothing if the capacity is already + /// sufficient. + /// + /// [`reserve`]: BinaryHeap::reserve + /// + /// # Panics + /// + /// Panics if the new capacity overflows [`usize`]. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// heap.reserve_exact(100); + /// assert!(heap.capacity() >= 100); + /// heap.push(4); + /// ``` + /// + /// [`reserve`]: BinaryHeap::reserve + #[stable(feature = "rust1", since = "1.0.0")] + pub fn reserve_exact(&mut self, additional: usize) { + self.data.reserve_exact(additional); + } + + /// Reserves capacity for at least `additional` elements more than the + /// current length. The allocator may reserve more space to speculatively + /// avoid frequent allocations. After calling `reserve`, + /// capacity will be greater than or equal to `self.len() + additional`. + /// Does nothing if capacity is already sufficient. + /// + /// # Panics + /// + /// Panics if the new capacity overflows [`usize`]. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// heap.reserve(100); + /// assert!(heap.capacity() >= 100); + /// heap.push(4); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn reserve(&mut self, additional: usize) { + self.data.reserve(additional); + } + + /// Tries to reserve the minimum capacity for at least `additional` elements + /// more than the current length. Unlike [`try_reserve`], this will not + /// deliberately over-allocate to speculatively avoid frequent allocations. + /// After calling `try_reserve_exact`, capacity will be greater than or + /// equal to `self.len() + additional` if it returns `Ok(())`. + /// Does nothing if the capacity is already sufficient. + /// + /// Note that the allocator may give the collection more space than it + /// requests. Therefore, capacity can not be relied upon to be precisely + /// minimal. Prefer [`try_reserve`] if future insertions are expected. + /// + /// [`try_reserve`]: BinaryHeap::try_reserve + /// + /// # Errors + /// + /// If the capacity overflows, or the allocator reports a failure, then an error + /// is returned. + /// + /// # Examples + /// + /// ``` + /// use std::collections::BinaryHeap; + /// use std::collections::TryReserveError; + /// + /// fn find_max_slow(data: &[u32]) -> Result, TryReserveError> { + /// let mut heap = BinaryHeap::new(); + /// + /// // Pre-reserve the memory, exiting if we can't + /// heap.try_reserve_exact(data.len())?; + /// + /// // Now we know this can't OOM in the middle of our complex work + /// heap.extend(data.iter()); + /// + /// Ok(heap.pop()) + /// } + /// # find_max_slow(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?"); + /// ``` + #[stable(feature = "try_reserve_2", since = "1.63.0")] + pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> { + self.data.try_reserve_exact(additional) + } + + /// Tries to reserve capacity for at least `additional` elements more than the + /// current length. The allocator may reserve more space to speculatively + /// avoid frequent allocations. After calling `try_reserve`, capacity will be + /// greater than or equal to `self.len() + additional` if it returns + /// `Ok(())`. Does nothing if capacity is already sufficient. This method + /// preserves the contents even if an error occurs. + /// + /// # Errors + /// + /// If the capacity overflows, or the allocator reports a failure, then an error + /// is returned. + /// + /// # Examples + /// + /// ``` + /// use std::collections::BinaryHeap; + /// use std::collections::TryReserveError; + /// + /// fn find_max_slow(data: &[u32]) -> Result, TryReserveError> { + /// let mut heap = BinaryHeap::new(); + /// + /// // Pre-reserve the memory, exiting if we can't + /// heap.try_reserve(data.len())?; + /// + /// // Now we know this can't OOM in the middle of our complex work + /// heap.extend(data.iter()); + /// + /// Ok(heap.pop()) + /// } + /// # find_max_slow(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?"); + /// ``` + #[stable(feature = "try_reserve_2", since = "1.63.0")] + pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> { + self.data.try_reserve(additional) + } + + /// Discards as much additional capacity as possible. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap: BinaryHeap = BinaryHeap::with_capacity(100); + /// + /// assert!(heap.capacity() >= 100); + /// heap.shrink_to_fit(); + /// assert!(heap.capacity() == 0); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn shrink_to_fit(&mut self) { + self.data.shrink_to_fit(); + } + + /// Discards capacity with a lower bound. + /// + /// The capacity will remain at least as large as both the length + /// and the supplied value. + /// + /// If the current capacity is less than the lower limit, this is a no-op. + /// + /// # Examples + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap: BinaryHeap = BinaryHeap::with_capacity(100); + /// + /// assert!(heap.capacity() >= 100); + /// heap.shrink_to(10); + /// assert!(heap.capacity() >= 10); + /// ``` + #[inline] + #[stable(feature = "shrink_to", since = "1.56.0")] + pub fn shrink_to(&mut self, min_capacity: usize) { + self.data.shrink_to(min_capacity) + } + + /// Returns a slice of all values in the underlying vector, in arbitrary + /// order. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// #![feature(binary_heap_as_slice)] + /// use std::collections::BinaryHeap; + /// use std::io::{self, Write}; + /// + /// let heap = BinaryHeap::from([1, 2, 3, 4, 5, 6, 7]); + /// + /// io::sink().write(heap.as_slice()).unwrap(); + /// ``` + #[must_use] + #[unstable(feature = "binary_heap_as_slice", issue = "83659")] + pub fn as_slice(&self) -> &[T] { + self.data.as_slice() + } + + /// Consumes the `BinaryHeap` and returns the underlying vector + /// in arbitrary order. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let heap = BinaryHeap::from([1, 2, 3, 4, 5, 6, 7]); + /// let vec = heap.into_vec(); + /// + /// // Will print in some order + /// for x in vec { + /// println!("{x}"); + /// } + /// ``` + #[must_use = "`self` will be dropped if the result is not used"] + #[stable(feature = "binary_heap_extras_15", since = "1.5.0")] + pub fn into_vec(self) -> Vec { + self.into() + } + + /// Returns the length of the binary heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let heap = BinaryHeap::from([1, 3]); + /// + /// assert_eq!(heap.len(), 2); + /// ``` + #[must_use] + #[stable(feature = "rust1", since = "1.0.0")] + pub fn len(&self) -> usize { + self.data.len() + } + + /// Checks if the binary heap is empty. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::new(); + /// + /// assert!(heap.is_empty()); + /// + /// heap.push(3); + /// heap.push(5); + /// heap.push(1); + /// + /// assert!(!heap.is_empty()); + /// ``` + #[must_use] + #[stable(feature = "rust1", since = "1.0.0")] + pub fn is_empty(&self) -> bool { + self.len() == 0 + } + + /// Clears the binary heap, returning an iterator over the removed elements + /// in arbitrary order. If the iterator is dropped before being fully + /// consumed, it drops the remaining elements in arbitrary order. + /// + /// The returned iterator keeps a mutable borrow on the heap to optimize + /// its implementation. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::from([1, 3]); + /// + /// assert!(!heap.is_empty()); + /// + /// for x in heap.drain() { + /// println!("{x}"); + /// } + /// + /// assert!(heap.is_empty()); + /// ``` + #[inline] + #[stable(feature = "drain", since = "1.6.0")] + pub fn drain(&mut self) -> Drain<'_, T> { + Drain { iter: self.data.drain(..) } + } + + /// Drops all items from the binary heap. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let mut heap = BinaryHeap::from([1, 3]); + /// + /// assert!(!heap.is_empty()); + /// + /// heap.clear(); + /// + /// assert!(heap.is_empty()); + /// ``` + #[stable(feature = "rust1", since = "1.0.0")] + pub fn clear(&mut self) { + self.drain(); + } +} + +/// Hole represents a hole in a slice i.e., an index without valid value +/// (because it was moved from or duplicated). +/// In drop, `Hole` will restore the slice by filling the hole +/// position with the value that was originally removed. +struct Hole<'a, T: 'a> { + data: &'a mut [T], + elt: ManuallyDrop, + pos: usize, +} + +impl<'a, T> Hole<'a, T> { + /// Create a new `Hole` at index `pos`. + /// + /// Unsafe because pos must be within the data slice. + #[inline] + unsafe fn new(data: &'a mut [T], pos: usize) -> Self { + debug_assert!(pos < data.len()); + // SAFE: pos should be inside the slice + let elt = unsafe { ptr::read(data.get_unchecked(pos)) }; + Hole { data, elt: ManuallyDrop::new(elt), pos } + } + + #[inline] + fn pos(&self) -> usize { + self.pos + } + + /// Returns a reference to the element removed. + #[inline] + fn element(&self) -> &T { + &self.elt + } + + /// Returns a reference to the element at `index`. + /// + /// Unsafe because index must be within the data slice and not equal to pos. + #[inline] + unsafe fn get(&self, index: usize) -> &T { + debug_assert!(index != self.pos); + debug_assert!(index < self.data.len()); + unsafe { self.data.get_unchecked(index) } + } + + /// Move hole to new location + /// + /// Unsafe because index must be within the data slice and not equal to pos. + #[inline] + unsafe fn move_to(&mut self, index: usize) { + debug_assert!(index != self.pos); + debug_assert!(index < self.data.len()); + unsafe { + let ptr = self.data.as_mut_ptr(); + let index_ptr: *const _ = ptr.add(index); + let hole_ptr = ptr.add(self.pos); + ptr::copy_nonoverlapping(index_ptr, hole_ptr, 1); + } + self.pos = index; + } +} + +impl Drop for Hole<'_, T> { + #[inline] + fn drop(&mut self) { + // fill the hole again + unsafe { + let pos = self.pos; + ptr::copy_nonoverlapping(&*self.elt, self.data.get_unchecked_mut(pos), 1); + } + } +} + +/// An iterator over the elements of a `BinaryHeap`. +/// +/// This `struct` is created by [`BinaryHeap::iter()`]. See its +/// documentation for more. +/// +/// [`iter`]: BinaryHeap::iter +#[must_use = "iterators are lazy and do nothing unless consumed"] +#[stable(feature = "rust1", since = "1.0.0")] +pub struct Iter<'a, T: 'a> { + iter: slice::Iter<'a, T>, +} + +#[stable(feature = "collection_debug", since = "1.17.0")] +impl fmt::Debug for Iter<'_, T> { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_tuple("Iter").field(&self.iter.as_slice()).finish() + } +} + +// FIXME(#26925) Remove in favor of `#[derive(Clone)]` +#[stable(feature = "rust1", since = "1.0.0")] +impl Clone for Iter<'_, T> { + fn clone(&self) -> Self { + Iter { iter: self.iter.clone() } + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<'a, T> Iterator for Iter<'a, T> { + type Item = &'a T; + + #[inline] + fn next(&mut self) -> Option<&'a T> { + self.iter.next() + } + + #[inline] + fn size_hint(&self) -> (usize, Option) { + self.iter.size_hint() + } + + #[inline] + fn last(self) -> Option<&'a T> { + self.iter.last() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<'a, T> DoubleEndedIterator for Iter<'a, T> { + #[inline] + fn next_back(&mut self) -> Option<&'a T> { + self.iter.next_back() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl ExactSizeIterator for Iter<'_, T> { + fn is_empty(&self) -> bool { + self.iter.is_empty() + } +} + +#[stable(feature = "fused", since = "1.26.0")] +impl FusedIterator for Iter<'_, T> {} + +/// An owning iterator over the elements of a `BinaryHeap`. +/// +/// This `struct` is created by [`BinaryHeap::into_iter()`] +/// (provided by the [`IntoIterator`] trait). See its documentation for more. +/// +/// [`into_iter`]: BinaryHeap::into_iter +/// [`IntoIterator`]: core::iter::IntoIterator +#[stable(feature = "rust1", since = "1.0.0")] +#[derive(Clone)] +pub struct IntoIter { + iter: vec::IntoIter, +} + +#[stable(feature = "collection_debug", since = "1.17.0")] +impl fmt::Debug for IntoIter { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_tuple("IntoIter").field(&self.iter.as_slice()).finish() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Iterator for IntoIter { + type Item = T; + + #[inline] + fn next(&mut self) -> Option { + self.iter.next() + } + + #[inline] + fn size_hint(&self) -> (usize, Option) { + self.iter.size_hint() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl DoubleEndedIterator for IntoIter { + #[inline] + fn next_back(&mut self) -> Option { + self.iter.next_back() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl ExactSizeIterator for IntoIter { + fn is_empty(&self) -> bool { + self.iter.is_empty() + } +} + +#[stable(feature = "fused", since = "1.26.0")] +impl FusedIterator for IntoIter {} + +// In addition to the SAFETY invariants of the following three unsafe traits +// also refer to the vec::in_place_collect module documentation to get an overview +#[unstable(issue = "none", feature = "inplace_iteration")] +#[doc(hidden)] +unsafe impl SourceIter for IntoIter { + type Source = IntoIter; + + #[inline] + unsafe fn as_inner(&mut self) -> &mut Self::Source { + self + } +} + +#[unstable(issue = "none", feature = "inplace_iteration")] +#[doc(hidden)] +unsafe impl InPlaceIterable for IntoIter {} + +unsafe impl AsVecIntoIter for IntoIter { + type Item = I; + + fn as_into_iter(&mut self) -> &mut vec::IntoIter { + &mut self.iter + } +} + +#[must_use = "iterators are lazy and do nothing unless consumed"] +#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] +#[derive(Clone, Debug)] +pub struct IntoIterSorted { + inner: BinaryHeap, +} + +#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] +impl Iterator for IntoIterSorted { + type Item = T; + + #[inline] + fn next(&mut self) -> Option { + self.inner.pop() + } + + #[inline] + fn size_hint(&self) -> (usize, Option) { + let exact = self.inner.len(); + (exact, Some(exact)) + } +} + +#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] +impl ExactSizeIterator for IntoIterSorted {} + +#[unstable(feature = "binary_heap_into_iter_sorted", issue = "59278")] +impl FusedIterator for IntoIterSorted {} + +#[unstable(feature = "trusted_len", issue = "37572")] +unsafe impl TrustedLen for IntoIterSorted {} + +/// A draining iterator over the elements of a `BinaryHeap`. +/// +/// This `struct` is created by [`BinaryHeap::drain()`]. See its +/// documentation for more. +/// +/// [`drain`]: BinaryHeap::drain +#[stable(feature = "drain", since = "1.6.0")] +#[derive(Debug)] +pub struct Drain<'a, T: 'a> { + iter: vec::Drain<'a, T>, +} + +#[stable(feature = "drain", since = "1.6.0")] +impl Iterator for Drain<'_, T> { + type Item = T; + + #[inline] + fn next(&mut self) -> Option { + self.iter.next() + } + + #[inline] + fn size_hint(&self) -> (usize, Option) { + self.iter.size_hint() + } +} + +#[stable(feature = "drain", since = "1.6.0")] +impl DoubleEndedIterator for Drain<'_, T> { + #[inline] + fn next_back(&mut self) -> Option { + self.iter.next_back() + } +} + +#[stable(feature = "drain", since = "1.6.0")] +impl ExactSizeIterator for Drain<'_, T> { + fn is_empty(&self) -> bool { + self.iter.is_empty() + } +} + +#[stable(feature = "fused", since = "1.26.0")] +impl FusedIterator for Drain<'_, T> {} + +/// A draining iterator over the elements of a `BinaryHeap`. +/// +/// This `struct` is created by [`BinaryHeap::drain_sorted()`]. See its +/// documentation for more. +/// +/// [`drain_sorted`]: BinaryHeap::drain_sorted +#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] +#[derive(Debug)] +pub struct DrainSorted<'a, T: Ord> { + inner: &'a mut BinaryHeap, +} + +#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] +impl<'a, T: Ord> Drop for DrainSorted<'a, T> { + /// Removes heap elements in heap order. + fn drop(&mut self) { + struct DropGuard<'r, 'a, T: Ord>(&'r mut DrainSorted<'a, T>); + + impl<'r, 'a, T: Ord> Drop for DropGuard<'r, 'a, T> { + fn drop(&mut self) { + while self.0.inner.pop().is_some() {} + } + } + + while let Some(item) = self.inner.pop() { + let guard = DropGuard(self); + drop(item); + mem::forget(guard); + } + } +} + +#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] +impl Iterator for DrainSorted<'_, T> { + type Item = T; + + #[inline] + fn next(&mut self) -> Option { + self.inner.pop() + } + + #[inline] + fn size_hint(&self) -> (usize, Option) { + let exact = self.inner.len(); + (exact, Some(exact)) + } +} + +#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] +impl ExactSizeIterator for DrainSorted<'_, T> {} + +#[unstable(feature = "binary_heap_drain_sorted", issue = "59278")] +impl FusedIterator for DrainSorted<'_, T> {} + +#[unstable(feature = "trusted_len", issue = "37572")] +unsafe impl TrustedLen for DrainSorted<'_, T> {} + +#[stable(feature = "binary_heap_extras_15", since = "1.5.0")] +impl From> for BinaryHeap { + /// Converts a `Vec` into a `BinaryHeap`. + /// + /// This conversion happens in-place, and has *O*(*n*) time complexity. + fn from(vec: Vec) -> BinaryHeap { + let mut heap = BinaryHeap { data: vec }; + heap.rebuild(); + heap + } +} + +#[stable(feature = "std_collections_from_array", since = "1.56.0")] +impl From<[T; N]> for BinaryHeap { + /// ``` + /// use std::collections::BinaryHeap; + /// + /// let mut h1 = BinaryHeap::from([1, 4, 2, 3]); + /// let mut h2: BinaryHeap<_> = [1, 4, 2, 3].into(); + /// while let Some((a, b)) = h1.pop().zip(h2.pop()) { + /// assert_eq!(a, b); + /// } + /// ``` + fn from(arr: [T; N]) -> Self { + Self::from_iter(arr) + } +} + +#[stable(feature = "binary_heap_extras_15", since = "1.5.0")] +impl From> for Vec { + /// Converts a `BinaryHeap` into a `Vec`. + /// + /// This conversion requires no data movement or allocation, and has + /// constant time complexity. + fn from(heap: BinaryHeap) -> Vec { + heap.data + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl FromIterator for BinaryHeap { + fn from_iter>(iter: I) -> BinaryHeap { + BinaryHeap::from(iter.into_iter().collect::>()) + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl IntoIterator for BinaryHeap { + type Item = T; + type IntoIter = IntoIter; + + /// Creates a consuming iterator, that is, one that moves each value out of + /// the binary heap in arbitrary order. The binary heap cannot be used + /// after calling this. + /// + /// # Examples + /// + /// Basic usage: + /// + /// ``` + /// use std::collections::BinaryHeap; + /// let heap = BinaryHeap::from([1, 2, 3, 4]); + /// + /// // Print 1, 2, 3, 4 in arbitrary order + /// for x in heap.into_iter() { + /// // x has type i32, not &i32 + /// println!("{x}"); + /// } + /// ``` + fn into_iter(self) -> IntoIter { + IntoIter { iter: self.data.into_iter() } + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl<'a, T> IntoIterator for &'a BinaryHeap { + type Item = &'a T; + type IntoIter = Iter<'a, T>; + + fn into_iter(self) -> Iter<'a, T> { + self.iter() + } +} + +#[stable(feature = "rust1", since = "1.0.0")] +impl Extend for BinaryHeap { + #[inline] + fn extend>(&mut self, iter: I) { + >::spec_extend(self, iter); + } + + #[inline] + fn extend_one(&mut self, item: T) { + self.push(item); + } + + #[inline] + fn extend_reserve(&mut self, additional: usize) { + self.reserve(additional); + } +} + +impl> SpecExtend for BinaryHeap { + default fn spec_extend(&mut self, iter: I) { + self.extend_desugared(iter.into_iter()); + } +} + +impl SpecExtend> for BinaryHeap { + fn spec_extend(&mut self, ref mut other: Vec) { + let start = self.data.len(); + self.data.append(other); + self.rebuild_tail(start); + } +} + +impl SpecExtend> for BinaryHeap { + fn spec_extend(&mut self, ref mut other: BinaryHeap) { + self.append(other); + } +} + +impl BinaryHeap { + fn extend_desugared>(&mut self, iter: I) { + let iterator = iter.into_iter(); + let (lower, _) = iterator.size_hint(); + + self.reserve(lower); + + iterator.for_each(move |elem| self.push(elem)); + } +} + +#[stable(feature = "extend_ref", since = "1.2.0")] +impl<'a, T: 'a + Ord + Copy> Extend<&'a T> for BinaryHeap { + fn extend>(&mut self, iter: I) { + self.extend(iter.into_iter().cloned()); + } + + #[inline] + fn extend_one(&mut self, &item: &'a T) { + self.push(item); + } + + #[inline] + fn extend_reserve(&mut self, additional: usize) { + self.reserve(additional); + } +} diff --git a/library/alloc/src/collections/binary_heap/tests.rs b/library/alloc/src/collections/binary_heap/tests.rs index 5a05215ae..ffbb6c80a 100644 --- a/library/alloc/src/collections/binary_heap/tests.rs +++ b/library/alloc/src/collections/binary_heap/tests.rs @@ -1,8 +1,9 @@ use super::*; use crate::boxed::Box; +use crate::testing::crash_test::{CrashTestDummy, Panic}; +use core::mem; use std::iter::TrustedLen; use std::panic::{catch_unwind, AssertUnwindSafe}; -use std::sync::atomic::{AtomicU32, Ordering}; #[test] fn test_iterator() { @@ -146,6 +147,24 @@ fn test_peek_mut() { assert_eq!(heap.peek(), Some(&9)); } +#[test] +fn test_peek_mut_leek() { + let data = vec![4, 2, 7]; + let mut heap = BinaryHeap::from(data); + let mut max = heap.peek_mut().unwrap(); + *max = -1; + + // The PeekMut object's Drop impl would have been responsible for moving the + // -1 out of the max position of the BinaryHeap, but we don't run it. + mem::forget(max); + + // Absent some mitigation like leak amplification, the -1 would incorrectly + // end up in the last position of the returned Vec, with the rest of the + // heap's original contents in front of it in sorted order. + let sorted_vec = heap.into_sorted_vec(); + assert!(sorted_vec.is_sorted(), "{:?}", sorted_vec); +} + #[test] fn test_peek_mut_pop() { let data = vec![2, 4, 6, 2, 1, 8, 10, 3, 5, 7, 0, 9, 1]; @@ -291,33 +310,83 @@ fn test_drain_sorted() { #[test] fn test_drain_sorted_leak() { - static DROPS: AtomicU32 = AtomicU32::new(0); - - #[derive(Clone, PartialEq, Eq, PartialOrd, Ord)] - struct D(u32, bool); - - impl Drop for D { - fn drop(&mut self) { - DROPS.fetch_add(1, Ordering::SeqCst); - - if self.1 { - panic!("panic in `drop`"); - } - } - } - + let d0 = CrashTestDummy::new(0); + let d1 = CrashTestDummy::new(1); + let d2 = CrashTestDummy::new(2); + let d3 = CrashTestDummy::new(3); + let d4 = CrashTestDummy::new(4); + let d5 = CrashTestDummy::new(5); let mut q = BinaryHeap::from(vec![ - D(0, false), - D(1, false), - D(2, false), - D(3, true), - D(4, false), - D(5, false), + d0.spawn(Panic::Never), + d1.spawn(Panic::Never), + d2.spawn(Panic::Never), + d3.spawn(Panic::InDrop), + d4.spawn(Panic::Never), + d5.spawn(Panic::Never), ]); - catch_unwind(AssertUnwindSafe(|| drop(q.drain_sorted()))).ok(); + catch_unwind(AssertUnwindSafe(|| drop(q.drain_sorted()))).unwrap_err(); + + assert_eq!(d0.dropped(), 1); + assert_eq!(d1.dropped(), 1); + assert_eq!(d2.dropped(), 1); + assert_eq!(d3.dropped(), 1); + assert_eq!(d4.dropped(), 1); + assert_eq!(d5.dropped(), 1); + assert!(q.is_empty()); +} + +#[test] +fn test_drain_forget() { + let a = CrashTestDummy::new(0); + let b = CrashTestDummy::new(1); + let c = CrashTestDummy::new(2); + let mut q = + BinaryHeap::from(vec![a.spawn(Panic::Never), b.spawn(Panic::Never), c.spawn(Panic::Never)]); + + catch_unwind(AssertUnwindSafe(|| { + let mut it = q.drain(); + it.next(); + mem::forget(it); + })) + .unwrap(); + // Behaviour after leaking is explicitly unspecified and order is arbitrary, + // so it's fine if these start failing, but probably worth knowing. + assert!(q.is_empty()); + assert_eq!(a.dropped() + b.dropped() + c.dropped(), 1); + assert_eq!(a.dropped(), 0); + assert_eq!(b.dropped(), 0); + assert_eq!(c.dropped(), 1); + drop(q); + assert_eq!(a.dropped(), 0); + assert_eq!(b.dropped(), 0); + assert_eq!(c.dropped(), 1); +} - assert_eq!(DROPS.load(Ordering::SeqCst), 6); +#[test] +fn test_drain_sorted_forget() { + let a = CrashTestDummy::new(0); + let b = CrashTestDummy::new(1); + let c = CrashTestDummy::new(2); + let mut q = + BinaryHeap::from(vec![a.spawn(Panic::Never), b.spawn(Panic::Never), c.spawn(Panic::Never)]); + + catch_unwind(AssertUnwindSafe(|| { + let mut it = q.drain_sorted(); + it.next(); + mem::forget(it); + })) + .unwrap(); + // Behaviour after leaking is explicitly unspecified, + // so it's fine if these start failing, but probably worth knowing. + assert_eq!(q.len(), 2); + assert_eq!(a.dropped(), 0); + assert_eq!(b.dropped(), 0); + assert_eq!(c.dropped(), 1); + drop(q); + assert_eq!(a.dropped(), 1); + assert_eq!(b.dropped(), 1); + assert_eq!(c.dropped(), 1); } #[test] @@ -415,7 +484,7 @@ fn test_retain() { #[test] #[cfg(not(target_os = "emscripten"))] fn panic_safe() { - use rand::{seq::SliceRandom, thread_rng}; + use rand::seq::SliceRandom; use std::cmp; use std::panic::{self, AssertUnwindSafe}; use std::sync::atomic::{AtomicUsize, Ordering}; @@ -440,7 +509,7 @@ fn panic_safe() { self.0.partial_cmp(&other.0) } } - let mut rng = thread_rng(); + let mut rng = crate::test_helpers::test_rng(); const DATASZ: usize = 32; // Miri is too slow let ntest = if cfg!(miri) { 1 } else { 10 }; diff --git a/library/alloc/src/collections/btree/map/tests.rs b/library/alloc/src/collections/btree/map/tests.rs index 4c372b1d6..700b1463b 100644 --- a/library/alloc/src/collections/btree/map/tests.rs +++ b/library/alloc/src/collections/btree/map/tests.rs @@ -1,12 +1,12 @@ -use super::super::testing::crash_test::{CrashTestDummy, Panic}; -use super::super::testing::ord_chaos::{Cyclic3, Governed, Governor}; -use super::super::testing::rng::DeterministicRng; use super::Entry::{Occupied, Vacant}; use super::*; use crate::boxed::Box; use crate::fmt::Debug; use crate::rc::Rc; use crate::string::{String, ToString}; +use crate::testing::crash_test::{CrashTestDummy, Panic}; +use crate::testing::ord_chaos::{Cyclic3, Governed, Governor}; +use crate::testing::rng::DeterministicRng; use crate::vec::Vec; use std::cmp::Ordering; use std::convert::TryFrom; diff --git a/library/alloc/src/collections/btree/mod.rs b/library/alloc/src/collections/btree/mod.rs index 9d43ac5c5..7552f2fc0 100644 --- a/library/alloc/src/collections/btree/mod.rs +++ b/library/alloc/src/collections/btree/mod.rs @@ -21,6 +21,3 @@ trait Recover { fn take(&mut self, key: &Q) -> Option; fn replace(&mut self, key: Self::Key) -> Option; } - -#[cfg(test)] -mod testing; diff --git a/library/alloc/src/collections/btree/node.rs b/library/alloc/src/collections/btree/node.rs index da766b67a..691246644 100644 --- a/library/alloc/src/collections/btree/node.rs +++ b/library/alloc/src/collections/btree/node.rs @@ -318,7 +318,10 @@ impl NodeRef pub fn ascend( self, ) -> Result, marker::Edge>, Self> { - let _ = BorrowType::TRAVERSAL_PERMIT; + const { + assert!(BorrowType::TRAVERSAL_PERMIT); + } + // We need to use raw pointers to nodes because, if BorrowType is marker::ValMut, // there might be outstanding mutable references to values that we must not invalidate. let leaf_ptr: *const _ = Self::as_leaf_ptr(&self); @@ -1003,7 +1006,10 @@ impl /// `edge.descend().ascend().unwrap()` and `node.ascend().unwrap().descend()` should /// both, upon success, do nothing. pub fn descend(self) -> NodeRef { - let _ = BorrowType::TRAVERSAL_PERMIT; + const { + assert!(BorrowType::TRAVERSAL_PERMIT); + } + // We need to use raw pointers to nodes because, if BorrowType is // marker::ValMut, there might be outstanding mutable references to // values that we must not invalidate. There's no worry accessing the @@ -1666,17 +1672,17 @@ pub mod marker { pub struct ValMut<'a>(PhantomData<&'a mut ()>); pub trait BorrowType { - // If node references of this borrow type allow traversing to other - // nodes in the tree, this constant can be evaluated. Thus reading it - // serves as a compile-time assertion. - const TRAVERSAL_PERMIT: () = (); + /// If node references of this borrow type allow traversing to other + /// nodes in the tree, this constant is set to `true`. It can be used + /// for a compile-time assertion. + const TRAVERSAL_PERMIT: bool = true; } impl BorrowType for Owned { - // Reject evaluation, because traversal isn't needed. Instead traversal - // happens using the result of `borrow_mut`. - // By disabling traversal, and only creating new references to roots, - // we know that every reference of the `Owned` type is to a root node. - const TRAVERSAL_PERMIT: () = panic!(); + /// Reject traversal, because it isn't needed. Instead traversal + /// happens using the result of `borrow_mut`. + /// By disabling traversal, and only creating new references to roots, + /// we know that every reference of the `Owned` type is to a root node. + const TRAVERSAL_PERMIT: bool = false; } impl BorrowType for Dying {} impl<'a> BorrowType for Immut<'a> {} diff --git a/library/alloc/src/collections/btree/set/tests.rs b/library/alloc/src/collections/btree/set/tests.rs index 502d3e1d1..7b8d41a60 100644 --- a/library/alloc/src/collections/btree/set/tests.rs +++ b/library/alloc/src/collections/btree/set/tests.rs @@ -1,6 +1,6 @@ -use super::super::testing::crash_test::{CrashTestDummy, Panic}; -use super::super::testing::rng::DeterministicRng; use super::*; +use crate::testing::crash_test::{CrashTestDummy, Panic}; +use crate::testing::rng::DeterministicRng; use crate::vec::Vec; use std::cmp::Ordering; use std::hash::{Hash, Hasher}; diff --git a/library/alloc/src/collections/btree/testing/crash_test.rs b/library/alloc/src/collections/btree/testing/crash_test.rs deleted file mode 100644 index bcf5f5f72..000000000 --- a/library/alloc/src/collections/btree/testing/crash_test.rs +++ /dev/null @@ -1,119 +0,0 @@ -// We avoid relying on anything else in the crate, apart from the `Debug` trait. -use crate::fmt::Debug; -use std::cmp::Ordering; -use std::sync::atomic::{AtomicUsize, Ordering::SeqCst}; - -/// A blueprint for crash test dummy instances that monitor particular events. -/// Some instances may be configured to panic at some point. -/// Events are `clone`, `drop` or some anonymous `query`. -/// -/// Crash test dummies are identified and ordered by an id, so they can be used -/// as keys in a BTreeMap. -#[derive(Debug)] -pub struct CrashTestDummy { - pub id: usize, - cloned: AtomicUsize, - dropped: AtomicUsize, - queried: AtomicUsize, -} - -impl CrashTestDummy { - /// Creates a crash test dummy design. The `id` determines order and equality of instances. - pub fn new(id: usize) -> CrashTestDummy { - CrashTestDummy { - id, - cloned: AtomicUsize::new(0), - dropped: AtomicUsize::new(0), - queried: AtomicUsize::new(0), - } - } - - /// Creates an instance of a crash test dummy that records what events it experiences - /// and optionally panics. - pub fn spawn(&self, panic: Panic) -> Instance<'_> { - Instance { origin: self, panic } - } - - /// Returns how many times instances of the dummy have been cloned. - pub fn cloned(&self) -> usize { - self.cloned.load(SeqCst) - } - - /// Returns how many times instances of the dummy have been dropped. - pub fn dropped(&self) -> usize { - self.dropped.load(SeqCst) - } - - /// Returns how many times instances of the dummy have had their `query` member invoked. - pub fn queried(&self) -> usize { - self.queried.load(SeqCst) - } -} - -#[derive(Debug)] -pub struct Instance<'a> { - origin: &'a CrashTestDummy, - panic: Panic, -} - -#[derive(Copy, Clone, Debug, PartialEq, Eq)] -pub enum Panic { - Never, - InClone, - InDrop, - InQuery, -} - -impl Instance<'_> { - pub fn id(&self) -> usize { - self.origin.id - } - - /// Some anonymous query, the result of which is already given. - pub fn query(&self, result: R) -> R { - self.origin.queried.fetch_add(1, SeqCst); - if self.panic == Panic::InQuery { - panic!("panic in `query`"); - } - result - } -} - -impl Clone for Instance<'_> { - fn clone(&self) -> Self { - self.origin.cloned.fetch_add(1, SeqCst); - if self.panic == Panic::InClone { - panic!("panic in `clone`"); - } - Self { origin: self.origin, panic: Panic::Never } - } -} - -impl Drop for Instance<'_> { - fn drop(&mut self) { - self.origin.dropped.fetch_add(1, SeqCst); - if self.panic == Panic::InDrop { - panic!("panic in `drop`"); - } - } -} - -impl PartialOrd for Instance<'_> { - fn partial_cmp(&self, other: &Self) -> Option { - self.id().partial_cmp(&other.id()) - } -} - -impl Ord for Instance<'_> { - fn cmp(&self, other: &Self) -> Ordering { - self.id().cmp(&other.id()) - } -} - -impl PartialEq for Instance<'_> { - fn eq(&self, other: &Self) -> bool { - self.id().eq(&other.id()) - } -} - -impl Eq for Instance<'_> {} diff --git a/library/alloc/src/collections/btree/testing/mod.rs b/library/alloc/src/collections/btree/testing/mod.rs deleted file mode 100644 index 7a094f8a5..000000000 --- a/library/alloc/src/collections/btree/testing/mod.rs +++ /dev/null @@ -1,3 +0,0 @@ -pub mod crash_test; -pub mod ord_chaos; -pub mod rng; diff --git a/library/alloc/src/collections/btree/testing/ord_chaos.rs b/library/alloc/src/collections/btree/testing/ord_chaos.rs deleted file mode 100644 index 96ce7c157..000000000 --- a/library/alloc/src/collections/btree/testing/ord_chaos.rs +++ /dev/null @@ -1,81 +0,0 @@ -use std::cell::Cell; -use std::cmp::Ordering::{self, *}; -use std::ptr; - -// Minimal type with an `Ord` implementation violating transitivity. -#[derive(Debug)] -pub enum Cyclic3 { - A, - B, - C, -} -use Cyclic3::*; - -impl PartialOrd for Cyclic3 { - fn partial_cmp(&self, other: &Self) -> Option { - Some(self.cmp(other)) - } -} - -impl Ord for Cyclic3 { - fn cmp(&self, other: &Self) -> Ordering { - match (self, other) { - (A, A) | (B, B) | (C, C) => Equal, - (A, B) | (B, C) | (C, A) => Less, - (A, C) | (B, A) | (C, B) => Greater, - } - } -} - -impl PartialEq for Cyclic3 { - fn eq(&self, other: &Self) -> bool { - self.cmp(&other) == Equal - } -} - -impl Eq for Cyclic3 {} - -// Controls the ordering of values wrapped by `Governed`. -#[derive(Debug)] -pub struct Governor { - flipped: Cell, -} - -impl Governor { - pub fn new() -> Self { - Governor { flipped: Cell::new(false) } - } - - pub fn flip(&self) { - self.flipped.set(!self.flipped.get()); - } -} - -// Type with an `Ord` implementation that forms a total order at any moment -// (assuming that `T` respects total order), but can suddenly be made to invert -// that total order. -#[derive(Debug)] -pub struct Governed<'a, T>(pub T, pub &'a Governor); - -impl PartialOrd for Governed<'_, T> { - fn partial_cmp(&self, other: &Self) -> Option { - Some(self.cmp(other)) - } -} - -impl Ord for Governed<'_, T> { - fn cmp(&self, other: &Self) -> Ordering { - assert!(ptr::eq(self.1, other.1)); - let ord = self.0.cmp(&other.0); - if self.1.flipped.get() { ord.reverse() } else { ord } - } -} - -impl PartialEq for Governed<'_, T> { - fn eq(&self, other: &Self) -> bool { - assert!(ptr::eq(self.1, other.1)); - self.0.eq(&other.0) - } -} - -impl Eq for Governed<'_, T> {} diff --git a/library/alloc/src/collections/btree/testing/rng.rs b/library/alloc/src/collections/btree/testing/rng.rs deleted file mode 100644 index ecf543bee..000000000 --- a/library/alloc/src/collections/btree/testing/rng.rs +++ /dev/null @@ -1,28 +0,0 @@ -/// XorShiftRng -pub struct DeterministicRng { - count: usize, - x: u32, - y: u32, - z: u32, - w: u32, -} - -impl DeterministicRng { - pub fn new() -> Self { - DeterministicRng { count: 0, x: 0x193a6754, y: 0xa8a7d469, z: 0x97830e05, w: 0x113ba7bb } - } - - /// Guarantees that each returned number is unique. - pub fn next(&mut self) -> u32 { - self.count += 1; - assert!(self.count <= 70029); - let x = self.x; - let t = x ^ (x << 11); - self.x = self.y; - self.y = self.z; - self.z = self.w; - let w_ = self.w; - self.w = w_ ^ (w_ >> 19) ^ (t ^ (t >> 8)); - self.w - } -} diff --git a/library/alloc/src/collections/linked_list/tests.rs b/library/alloc/src/collections/linked_list/tests.rs index f8fbfa1bf..04594d55b 100644 --- a/library/alloc/src/collections/linked_list/tests.rs +++ b/library/alloc/src/collections/linked_list/tests.rs @@ -1,10 +1,11 @@ use super::*; +use crate::testing::crash_test::{CrashTestDummy, Panic}; use crate::vec::Vec; use std::panic::{catch_unwind, AssertUnwindSafe}; use std::thread; -use rand::{thread_rng, RngCore}; +use rand::RngCore; #[test] fn test_basic() { @@ -480,12 +481,12 @@ fn test_split_off_2() { } } -fn fuzz_test(sz: i32) { +fn fuzz_test(sz: i32, rng: &mut impl RngCore) { let mut m: LinkedList<_> = LinkedList::new(); let mut v = vec![]; for i in 0..sz { check_links(&m); - let r: u8 = thread_rng().next_u32() as u8; + let r: u8 = rng.next_u32() as u8; match r % 6 { 0 => { m.pop_back(); @@ -520,11 +521,12 @@ fn fuzz_test(sz: i32) { #[test] fn test_fuzz() { + let mut rng = crate::test_helpers::test_rng(); for _ in 0..25 { - fuzz_test(3); - fuzz_test(16); + fuzz_test(3, &mut rng); + fuzz_test(16, &mut rng); #[cfg(not(miri))] // Miri is too slow - fuzz_test(189); + fuzz_test(189, &mut rng); } } @@ -984,35 +986,34 @@ fn drain_filter_complex() { #[test] fn drain_filter_drop_panic_leak() { - static mut DROPS: i32 = 0; - - struct D(bool); - - impl Drop for D { - fn drop(&mut self) { - unsafe { - DROPS += 1; - } - - if self.0 { - panic!("panic in `drop`"); - } - } - } - + let d0 = CrashTestDummy::new(0); + let d1 = CrashTestDummy::new(1); + let d2 = CrashTestDummy::new(2); + let d3 = CrashTestDummy::new(3); + let d4 = CrashTestDummy::new(4); + let d5 = CrashTestDummy::new(5); + let d6 = CrashTestDummy::new(6); + let d7 = CrashTestDummy::new(7); let mut q = LinkedList::new(); - q.push_back(D(false)); - q.push_back(D(false)); - q.push_back(D(false)); - q.push_back(D(false)); - q.push_back(D(false)); - q.push_front(D(false)); - q.push_front(D(true)); - q.push_front(D(false)); - - catch_unwind(AssertUnwindSafe(|| drop(q.drain_filter(|_| true)))).ok(); - - assert_eq!(unsafe { DROPS }, 8); + q.push_back(d3.spawn(Panic::Never)); + q.push_back(d4.spawn(Panic::Never)); + q.push_back(d5.spawn(Panic::Never)); + q.push_back(d6.spawn(Panic::Never)); + q.push_back(d7.spawn(Panic::Never)); + q.push_front(d2.spawn(Panic::Never)); + q.push_front(d1.spawn(Panic::InDrop)); + q.push_front(d0.spawn(Panic::Never)); + + catch_unwind(AssertUnwindSafe(|| drop(q.drain_filter(|_| true)))).unwrap_err(); + + assert_eq!(d0.dropped(), 1); + assert_eq!(d1.dropped(), 1); + assert_eq!(d2.dropped(), 1); + assert_eq!(d3.dropped(), 1); + assert_eq!(d4.dropped(), 1); + assert_eq!(d5.dropped(), 1); + assert_eq!(d6.dropped(), 1); + assert_eq!(d7.dropped(), 1); assert!(q.is_empty()); } diff --git a/library/alloc/src/collections/vec_deque/into_iter.rs b/library/alloc/src/collections/vec_deque/into_iter.rs index 55f6138cd..e54880e86 100644 --- a/library/alloc/src/collections/vec_deque/into_iter.rs +++ b/library/alloc/src/collections/vec_deque/into_iter.rs @@ -25,6 +25,10 @@ impl IntoIter { pub(super) fn new(inner: VecDeque) -> Self { IntoIter { inner } } + + pub(super) fn into_vecdeque(self) -> VecDeque { + self.inner + } } #[stable(feature = "collection_debug", since = "1.17.0")] diff --git a/library/alloc/src/collections/vec_deque/mod.rs b/library/alloc/src/collections/vec_deque/mod.rs index 4866c53e7..8317ac431 100644 --- a/library/alloc/src/collections/vec_deque/mod.rs +++ b/library/alloc/src/collections/vec_deque/mod.rs @@ -55,6 +55,10 @@ use self::spec_extend::SpecExtend; mod spec_extend; +use self::spec_from_iter::SpecFromIter; + +mod spec_from_iter; + #[cfg(test)] mod tests; @@ -531,12 +535,13 @@ impl VecDeque { /// /// let deque: VecDeque = VecDeque::new(); /// ``` - // FIXME: This should probably be const #[inline] #[stable(feature = "rust1", since = "1.0.0")] + #[rustc_const_stable(feature = "const_vec_deque_new", since = "1.68.0")] #[must_use] - pub fn new() -> VecDeque { - VecDeque::new_in(Global) + pub const fn new() -> VecDeque { + // FIXME: This should just be `VecDeque::new_in(Global)` once that hits stable. + VecDeque { head: 0, len: 0, buf: RawVec::NEW } } /// Creates an empty deque with space for at least `capacity` elements. @@ -586,6 +591,38 @@ impl VecDeque { VecDeque { head: 0, len: 0, buf: RawVec::with_capacity_in(capacity, alloc) } } + /// Creates a `VecDeque` from a raw allocation, when the initialized + /// part of that allocation forms a *contiguous* subslice thereof. + /// + /// For use by `vec::IntoIter::into_vecdeque` + /// + /// # Safety + /// + /// All the usual requirements on the allocated memory like in + /// `Vec::from_raw_parts_in`, but takes a *range* of elements that are + /// initialized rather than only supporting `0..len`. Requires that + /// `initialized.start` ≤ `initialized.end` ≤ `capacity`. + #[inline] + pub(crate) unsafe fn from_contiguous_raw_parts_in( + ptr: *mut T, + initialized: Range, + capacity: usize, + alloc: A, + ) -> Self { + debug_assert!(initialized.start <= initialized.end); + debug_assert!(initialized.end <= capacity); + + // SAFETY: Our safety precondition guarantees the range length won't wrap, + // and that the allocation is valid for use in `RawVec`. + unsafe { + VecDeque { + head: initialized.start, + len: initialized.end.unchecked_sub(initialized.start), + buf: RawVec::from_raw_parts_in(ptr, capacity, alloc), + } + } + } + /// Provides a reference to the element at the given index. /// /// Element at index 0 is the front of the queue. @@ -599,6 +636,7 @@ impl VecDeque { /// buf.push_back(3); /// buf.push_back(4); /// buf.push_back(5); + /// buf.push_back(6); /// assert_eq!(buf.get(1), Some(&4)); /// ``` #[stable(feature = "rust1", since = "1.0.0")] @@ -624,10 +662,11 @@ impl VecDeque { /// buf.push_back(3); /// buf.push_back(4); /// buf.push_back(5); + /// buf.push_back(6); + /// assert_eq!(buf[1], 4); /// if let Some(elem) = buf.get_mut(1) { /// *elem = 7; /// } - /// /// assert_eq!(buf[1], 7); /// ``` #[stable(feature = "rust1", since = "1.0.0")] @@ -905,65 +944,72 @@ impl VecDeque { return; } - if target_cap < self.capacity() { - // There are three cases of interest: - // All elements are out of desired bounds - // Elements are contiguous, and head is out of desired bounds - // Elements are discontiguous, and tail is out of desired bounds + // There are three cases of interest: + // All elements are out of desired bounds + // Elements are contiguous, and tail is out of desired bounds + // Elements are discontiguous + // + // At all other times, element positions are unaffected. + + // `head` and `len` are at most `isize::MAX` and `target_cap < self.capacity()`, so nothing can + // overflow. + let tail_outside = (target_cap + 1..=self.capacity()).contains(&(self.head + self.len)); + + if self.len == 0 { + self.head = 0; + } else if self.head >= target_cap && tail_outside { + // Head and tail are both out of bounds, so copy all of them to the front. // - // At all other times, element positions are unaffected. + // H := head + // L := last element + // H L + // [. . . . . . . . o o o o o o o . ] + // H L + // [o o o o o o o . ] + unsafe { + // nonoverlapping because `self.head >= target_cap >= self.len`. + self.copy_nonoverlapping(self.head, 0, self.len); + } + self.head = 0; + } else if self.head < target_cap && tail_outside { + // Head is in bounds, tail is out of bounds. + // Copy the overflowing part to the beginning of the + // buffer. This won't overlap because `target_cap >= self.len`. // - // Indicates that elements at the head should be moved. - - let tail_outside = (target_cap + 1..=self.capacity()).contains(&(self.head + self.len)); - // Move elements from out of desired bounds (positions after target_cap) - if self.len == 0 { - self.head = 0; - } else if self.head >= target_cap && tail_outside { - // H := head - // L := last element - // H L - // [. . . . . . . . o o o o o o o . ] - // H L - // [o o o o o o o . ] - unsafe { - // nonoverlapping because self.head >= target_cap >= self.len - self.copy_nonoverlapping(self.head, 0, self.len); - } - self.head = 0; - } else if self.head < target_cap && tail_outside { - // H := head - // L := last element - // H L - // [. . . o o o o o o o . . . . . . ] - // L H - // [o o . o o o o o ] - let len = self.head + self.len - target_cap; - unsafe { - self.copy_nonoverlapping(target_cap, 0, len); - } - } else if self.head >= target_cap { - // H := head - // L := last element - // L H - // [o o o o o . . . . . . . . . o o ] - // L H - // [o o o o o . o o ] - let len = self.capacity() - self.head; - let new_head = target_cap - len; - unsafe { - // can't use copy_nonoverlapping here for the same reason - // as in `handle_capacity_increase()` - self.copy(self.head, new_head, len); - } - self.head = new_head; + // H := head + // L := last element + // H L + // [. . . o o o o o o o . . . . . . ] + // L H + // [o o . o o o o o ] + let len = self.head + self.len - target_cap; + unsafe { + self.copy_nonoverlapping(target_cap, 0, len); } - - self.buf.shrink_to_fit(target_cap); - - debug_assert!(self.head < self.capacity() || self.capacity() == 0); - debug_assert!(self.len <= self.capacity()); + } else if !self.is_contiguous() { + // The head slice is at least partially out of bounds, tail is in bounds. + // Copy the head backwards so it lines up with the target capacity. + // This won't overlap because `target_cap >= self.len`. + // + // H := head + // L := last element + // L H + // [o o o o o . . . . . . . . . o o ] + // L H + // [o o o o o . o o ] + let head_len = self.capacity() - self.head; + let new_head = target_cap - head_len; + unsafe { + // can't use `copy_nonoverlapping()` here because the new and old + // regions for the head might overlap. + self.copy(self.head, new_head, head_len); + } + self.head = new_head; } + self.buf.shrink_to_fit(target_cap); + + debug_assert!(self.head < self.capacity() || self.capacity() == 0); + debug_assert!(self.len <= self.capacity()); } /// Shortens the deque, keeping the first `len` elements and dropping @@ -1878,7 +1924,7 @@ impl VecDeque { #[stable(feature = "append", since = "1.4.0")] pub fn append(&mut self, other: &mut Self) { if T::IS_ZST { - self.len += other.len; + self.len = self.len.checked_add(other.len).expect("capacity overflow"); other.len = 0; other.head = 0; return; @@ -2505,7 +2551,7 @@ impl VecDeque { /// The deque is assumed to be partitioned according to the given predicate. /// This means that all elements for which the predicate returns true are at the start of the deque /// and all elements for which the predicate returns false are at the end. - /// For example, [7, 15, 3, 5, 4, 12, 6] is a partitioned under the predicate x % 2 != 0 + /// For example, `[7, 15, 3, 5, 4, 12, 6]` is partitioned under the predicate `x % 2 != 0` /// (all odd numbers are at the start, all even at the end). /// /// If the deque is not partitioned, the returned result is unspecified and meaningless, @@ -2699,18 +2745,8 @@ impl IndexMut for VecDeque { #[stable(feature = "rust1", since = "1.0.0")] impl FromIterator for VecDeque { - #[inline] fn from_iter>(iter: I) -> VecDeque { - // Since converting is O(1) now, might as well re-use that logic - // (including things like the `vec::IntoIter`→`Vec` specialization) - // especially as that could save us some monomorphiziation work - // if one uses the same iterators (like slice ones) with both. - return from_iter_via_vec(iter.into_iter()); - - #[inline] - fn from_iter_via_vec(iter: impl Iterator) -> VecDeque { - Vec::from_iter(iter).into() - } + SpecFromIter::spec_from_iter(iter.into_iter()) } } @@ -2794,9 +2830,9 @@ impl From> for VecDeque { /// [`Vec`]: crate::vec::Vec /// [`VecDeque`]: crate::collections::VecDeque /// - /// In its current implementation, this is a very cheap - /// conversion. This isn't yet a guarantee though, and - /// shouldn't be relied on. + /// This conversion is guaranteed to run in *O*(1) time + /// and to not re-allocate the `Vec`'s buffer or allocate + /// any additional memory. #[inline] fn from(other: Vec) -> Self { let (ptr, len, cap, alloc) = other.into_raw_parts_with_alloc(); diff --git a/library/alloc/src/collections/vec_deque/spec_from_iter.rs b/library/alloc/src/collections/vec_deque/spec_from_iter.rs new file mode 100644 index 000000000..7650492eb --- /dev/null +++ b/library/alloc/src/collections/vec_deque/spec_from_iter.rs @@ -0,0 +1,33 @@ +use super::{IntoIter, VecDeque}; + +/// Specialization trait used for `VecDeque::from_iter` +pub(super) trait SpecFromIter { + fn spec_from_iter(iter: I) -> Self; +} + +impl SpecFromIter for VecDeque +where + I: Iterator, +{ + default fn spec_from_iter(iterator: I) -> Self { + // Since converting is O(1) now, just re-use the `Vec` logic for + // anything where we can't do something extra-special for `VecDeque`, + // especially as that could save us some monomorphiziation work + // if one uses the same iterators (like slice ones) with both. + crate::vec::Vec::from_iter(iterator).into() + } +} + +impl SpecFromIter> for VecDeque { + #[inline] + fn spec_from_iter(iterator: crate::vec::IntoIter) -> Self { + iterator.into_vecdeque() + } +} + +impl SpecFromIter> for VecDeque { + #[inline] + fn spec_from_iter(iterator: IntoIter) -> Self { + iterator.into_vecdeque() + } +} diff --git a/library/alloc/src/collections/vec_deque/tests.rs b/library/alloc/src/collections/vec_deque/tests.rs index 220ad71be..205a8ff3c 100644 --- a/library/alloc/src/collections/vec_deque/tests.rs +++ b/library/alloc/src/collections/vec_deque/tests.rs @@ -748,6 +748,48 @@ fn test_drain() { } } +#[test] +fn issue_108453() { + let mut deque = VecDeque::with_capacity(10); + + deque.push_back(1u8); + deque.push_back(2); + deque.push_back(3); + + deque.push_front(10); + deque.push_front(9); + + deque.shrink_to(9); + + assert_eq!(deque.into_iter().collect::>(), vec![9, 10, 1, 2, 3]); +} + +#[test] +fn test_shrink_to() { + // test deques with capacity 16 with all possible head positions, lengths and target capacities. + let cap = 16; + + for len in 0..cap { + for head in 0..cap { + let expected = (1..=len).collect::>(); + + for target_cap in len..cap { + let mut deque = VecDeque::with_capacity(cap); + // currently, `with_capacity` always allocates the exact capacity if it's greater than 8. + assert_eq!(deque.capacity(), cap); + + // we can let the head point anywhere in the buffer since the deque is empty. + deque.head = head; + deque.extend(1..=len); + + deque.shrink_to(target_cap); + + assert_eq!(deque, expected); + } + } + } +} + #[test] fn test_shrink_to_fit() { // This test checks that every single combination of head and tail position, diff --git a/library/alloc/src/fmt.rs b/library/alloc/src/fmt.rs index 799ce9d5d..eadb35cb9 100644 --- a/library/alloc/src/fmt.rs +++ b/library/alloc/src/fmt.rs @@ -419,7 +419,7 @@ //! // documentation for details, and the function `pad` can be used //! // to pad strings. //! let decimals = f.precision().unwrap_or(3); -//! let string = format!("{:.*}", decimals, magnitude); +//! let string = format!("{magnitude:.decimals$}"); //! f.pad_integral(true, "", &string) //! } //! } @@ -518,7 +518,7 @@ //! write!(&mut some_writer, "{}", format_args!("print with a {}", "macro")); //! //! fn my_fmt_fn(args: fmt::Arguments) { -//! write!(&mut io::stdout(), "{}", args); +//! write!(&mut io::stdout(), "{args}"); //! } //! my_fmt_fn(format_args!(", or a {} too", "function")); //! ``` diff --git a/library/alloc/src/lib.rs b/library/alloc/src/lib.rs index 96960d43f..ca75c3895 100644 --- a/library/alloc/src/lib.rs +++ b/library/alloc/src/lib.rs @@ -3,7 +3,7 @@ //! This library provides smart pointers and collections for managing //! heap-allocated values. //! -//! This library, like libcore, normally doesn’t need to be used directly +//! This library, like core, normally doesn’t need to be used directly //! since its contents are re-exported in the [`std` crate](../std/index.html). //! Crates that use the `#![no_std]` attribute however will typically //! not depend on `std`, so they’d use this crate instead. @@ -75,7 +75,7 @@ ))] #![no_std] #![needs_allocator] -// To run liballoc tests without x.py without ending up with two copies of liballoc, Miri needs to be +// To run alloc tests without x.py without ending up with two copies of alloc, Miri needs to be // able to "empty" this crate. See . // rustc itself never sets the feature, so this line has no affect there. #![cfg(any(not(feature = "miri-test-libstd"), test, doctest))] @@ -106,10 +106,12 @@ #![feature(const_size_of_val)] #![feature(const_align_of_val)] #![feature(const_ptr_read)] +#![feature(const_maybe_uninit_zeroed)] #![feature(const_maybe_uninit_write)] #![feature(const_maybe_uninit_as_mut_ptr)] #![feature(const_refs_to_cell)] #![feature(core_intrinsics)] +#![feature(core_panic)] #![feature(const_eval_select)] #![feature(const_pin)] #![feature(const_waker)] @@ -122,7 +124,9 @@ #![feature(fmt_internals)] #![feature(fn_traits)] #![feature(hasher_prefixfree_extras)] +#![feature(inline_const)] #![feature(inplace_iteration)] +#![cfg_attr(test, feature(is_sorted))] #![feature(iter_advance_by)] #![feature(iter_next_chunk)] #![feature(iter_repeat_n)] @@ -152,7 +156,7 @@ #![feature(trusted_len)] #![feature(trusted_random_access)] #![feature(try_trait_v2)] -#![cfg_attr(not(bootstrap), feature(tuple_trait))] +#![feature(tuple_trait)] #![feature(unchecked_math)] #![feature(unicode_internals)] #![feature(unsize)] @@ -190,6 +194,7 @@ #![feature(unsized_fn_params)] #![feature(c_unwind)] #![feature(with_negative_coherence)] +#![cfg_attr(test, feature(panic_update_hook))] // // Rustdoc features: #![feature(doc_cfg)] @@ -206,6 +211,8 @@ extern crate std; #[cfg(test)] extern crate test; +#[cfg(test)] +mod testing; // Module with internal macros used by other modules (needs to be included before other modules). #[macro_use] @@ -251,3 +258,20 @@ pub mod vec; pub mod __export { pub use core::format_args; } + +#[cfg(test)] +#[allow(dead_code)] // Not used in all configurations +pub(crate) mod test_helpers { + /// Copied from `std::test_helpers::test_rng`, since these tests rely on the + /// seed not being the same for every RNG invocation too. + pub(crate) fn test_rng() -> rand_xorshift::XorShiftRng { + use std::hash::{BuildHasher, Hash, Hasher}; + let mut hasher = std::collections::hash_map::RandomState::new().build_hasher(); + std::panic::Location::caller().hash(&mut hasher); + let hc64 = hasher.finish(); + let seed_vec = + hc64.to_le_bytes().into_iter().chain(0u8..8).collect::>(); + let seed: [u8; 16] = seed_vec.as_slice().try_into().unwrap(); + rand::SeedableRng::from_seed(seed) + } +} diff --git a/library/alloc/src/rc.rs b/library/alloc/src/rc.rs index 38e31b180..c9aa23fc4 100644 --- a/library/alloc/src/rc.rs +++ b/library/alloc/src/rc.rs @@ -336,7 +336,7 @@ impl UnwindSafe for Rc {} #[stable(feature = "rc_ref_unwind_safe", since = "1.58.0")] impl RefUnwindSafe for Rc {} -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized> for Rc {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] @@ -2179,7 +2179,7 @@ pub struct Weak { // This is a `NonNull` to allow optimizing the size of this type in enums, // but it is not necessarily a valid pointer. // `Weak::new` sets this to `usize::MAX` so that it doesn’t need - // to allocate space on the heap. That's not a value a real pointer + // to allocate space on the heap. That's not a value a real pointer // will ever have because RcBox has alignment at least 2. // This is only possible when `T: Sized`; unsized `T` never dangle. ptr: NonNull>, @@ -2190,7 +2190,7 @@ impl !marker::Send for Weak {} #[stable(feature = "rc_weak", since = "1.4.0")] impl !marker::Sync for Weak {} -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized> for Weak {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] @@ -2561,7 +2561,7 @@ impl Clone for Weak { } #[stable(feature = "rc_weak", since = "1.4.0")] -impl fmt::Debug for Weak { +impl fmt::Debug for Weak { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "(Weak)") } diff --git a/library/alloc/src/slice.rs b/library/alloc/src/slice.rs index 1b61ede34..fecacc2bb 100644 --- a/library/alloc/src/slice.rs +++ b/library/alloc/src/slice.rs @@ -19,15 +19,20 @@ use core::cmp::Ordering::{self, Less}; use core::mem::{self, SizedTypeProperties}; #[cfg(not(no_global_oom_handling))] use core::ptr; +#[cfg(not(no_global_oom_handling))] +use core::slice::sort; use crate::alloc::Allocator; #[cfg(not(no_global_oom_handling))] -use crate::alloc::Global; +use crate::alloc::{self, Global}; #[cfg(not(no_global_oom_handling))] use crate::borrow::ToOwned; use crate::boxed::Box; use crate::vec::Vec; +#[cfg(test)] +mod tests; + #[unstable(feature = "slice_range", issue = "76393")] pub use core::slice::range; #[unstable(feature = "array_chunks", issue = "74985")] @@ -203,7 +208,7 @@ impl [T] { where T: Ord, { - merge_sort(self, T::lt); + stable_sort(self, T::lt); } /// Sorts the slice with a comparator function. @@ -259,7 +264,7 @@ impl [T] { where F: FnMut(&T, &T) -> Ordering, { - merge_sort(self, |a, b| compare(a, b) == Less); + stable_sort(self, |a, b| compare(a, b) == Less); } /// Sorts the slice with a key extraction function. @@ -302,7 +307,7 @@ impl [T] { F: FnMut(&T) -> K, K: Ord, { - merge_sort(self, |a, b| f(a).lt(&f(b))); + stable_sort(self, |a, b| f(a).lt(&f(b))); } /// Sorts the slice with a key extraction function. @@ -653,7 +658,7 @@ impl [u8] { /// /// ```error /// error[E0207]: the type parameter `T` is not constrained by the impl trait, self type, or predica -/// --> src/liballoc/slice.rs:608:6 +/// --> library/alloc/src/slice.rs:608:6 /// | /// 608 | impl> Concat for [V] { /// | ^ unconstrained type parameter @@ -809,324 +814,52 @@ impl ToOwned for [T] { // Sorting //////////////////////////////////////////////////////////////////////////////// -/// Inserts `v[0]` into pre-sorted sequence `v[1..]` so that whole `v[..]` becomes sorted. -/// -/// This is the integral subroutine of insertion sort. -#[cfg(not(no_global_oom_handling))] -fn insert_head(v: &mut [T], is_less: &mut F) -where - F: FnMut(&T, &T) -> bool, -{ - if v.len() >= 2 && is_less(&v[1], &v[0]) { - unsafe { - // There are three ways to implement insertion here: - // - // 1. Swap adjacent elements until the first one gets to its final destination. - // However, this way we copy data around more than is necessary. If elements are big - // structures (costly to copy), this method will be slow. - // - // 2. Iterate until the right place for the first element is found. Then shift the - // elements succeeding it to make room for it and finally place it into the - // remaining hole. This is a good method. - // - // 3. Copy the first element into a temporary variable. Iterate until the right place - // for it is found. As we go along, copy every traversed element into the slot - // preceding it. Finally, copy data from the temporary variable into the remaining - // hole. This method is very good. Benchmarks demonstrated slightly better - // performance than with the 2nd method. - // - // All methods were benchmarked, and the 3rd showed best results. So we chose that one. - let tmp = mem::ManuallyDrop::new(ptr::read(&v[0])); - - // Intermediate state of the insertion process is always tracked by `hole`, which - // serves two purposes: - // 1. Protects integrity of `v` from panics in `is_less`. - // 2. Fills the remaining hole in `v` in the end. - // - // Panic safety: - // - // If `is_less` panics at any point during the process, `hole` will get dropped and - // fill the hole in `v` with `tmp`, thus ensuring that `v` still holds every object it - // initially held exactly once. - let mut hole = InsertionHole { src: &*tmp, dest: &mut v[1] }; - ptr::copy_nonoverlapping(&v[1], &mut v[0], 1); - - for i in 2..v.len() { - if !is_less(&v[i], &*tmp) { - break; - } - ptr::copy_nonoverlapping(&v[i], &mut v[i - 1], 1); - hole.dest = &mut v[i]; - } - // `hole` gets dropped and thus copies `tmp` into the remaining hole in `v`. - } - } - - // When dropped, copies from `src` into `dest`. - struct InsertionHole { - src: *const T, - dest: *mut T, - } - - impl Drop for InsertionHole { - fn drop(&mut self) { - unsafe { - ptr::copy_nonoverlapping(self.src, self.dest, 1); - } - } - } -} - -/// Merges non-decreasing runs `v[..mid]` and `v[mid..]` using `buf` as temporary storage, and -/// stores the result into `v[..]`. -/// -/// # Safety -/// -/// The two slices must be non-empty and `mid` must be in bounds. Buffer `buf` must be long enough -/// to hold a copy of the shorter slice. Also, `T` must not be a zero-sized type. -#[cfg(not(no_global_oom_handling))] -unsafe fn merge(v: &mut [T], mid: usize, buf: *mut T, is_less: &mut F) -where - F: FnMut(&T, &T) -> bool, -{ - let len = v.len(); - let v = v.as_mut_ptr(); - let (v_mid, v_end) = unsafe { (v.add(mid), v.add(len)) }; - - // The merge process first copies the shorter run into `buf`. Then it traces the newly copied - // run and the longer run forwards (or backwards), comparing their next unconsumed elements and - // copying the lesser (or greater) one into `v`. - // - // As soon as the shorter run is fully consumed, the process is done. If the longer run gets - // consumed first, then we must copy whatever is left of the shorter run into the remaining - // hole in `v`. - // - // Intermediate state of the process is always tracked by `hole`, which serves two purposes: - // 1. Protects integrity of `v` from panics in `is_less`. - // 2. Fills the remaining hole in `v` if the longer run gets consumed first. - // - // Panic safety: - // - // If `is_less` panics at any point during the process, `hole` will get dropped and fill the - // hole in `v` with the unconsumed range in `buf`, thus ensuring that `v` still holds every - // object it initially held exactly once. - let mut hole; - - if mid <= len - mid { - // The left run is shorter. - unsafe { - ptr::copy_nonoverlapping(v, buf, mid); - hole = MergeHole { start: buf, end: buf.add(mid), dest: v }; - } - - // Initially, these pointers point to the beginnings of their arrays. - let left = &mut hole.start; - let mut right = v_mid; - let out = &mut hole.dest; - - while *left < hole.end && right < v_end { - // Consume the lesser side. - // If equal, prefer the left run to maintain stability. - unsafe { - let to_copy = if is_less(&*right, &**left) { - get_and_increment(&mut right) - } else { - get_and_increment(left) - }; - ptr::copy_nonoverlapping(to_copy, get_and_increment(out), 1); - } - } - } else { - // The right run is shorter. - unsafe { - ptr::copy_nonoverlapping(v_mid, buf, len - mid); - hole = MergeHole { start: buf, end: buf.add(len - mid), dest: v_mid }; - } - - // Initially, these pointers point past the ends of their arrays. - let left = &mut hole.dest; - let right = &mut hole.end; - let mut out = v_end; - - while v < *left && buf < *right { - // Consume the greater side. - // If equal, prefer the right run to maintain stability. - unsafe { - let to_copy = if is_less(&*right.sub(1), &*left.sub(1)) { - decrement_and_get(left) - } else { - decrement_and_get(right) - }; - ptr::copy_nonoverlapping(to_copy, decrement_and_get(&mut out), 1); - } - } - } - // Finally, `hole` gets dropped. If the shorter run was not fully consumed, whatever remains of - // it will now be copied into the hole in `v`. - - unsafe fn get_and_increment(ptr: &mut *mut T) -> *mut T { - let old = *ptr; - *ptr = unsafe { ptr.add(1) }; - old - } - - unsafe fn decrement_and_get(ptr: &mut *mut T) -> *mut T { - *ptr = unsafe { ptr.sub(1) }; - *ptr - } - - // When dropped, copies the range `start..end` into `dest..`. - struct MergeHole { - start: *mut T, - end: *mut T, - dest: *mut T, - } - - impl Drop for MergeHole { - fn drop(&mut self) { - // `T` is not a zero-sized type, and these are pointers into a slice's elements. - unsafe { - let len = self.end.sub_ptr(self.start); - ptr::copy_nonoverlapping(self.start, self.dest, len); - } - } - } -} - -/// This merge sort borrows some (but not all) ideas from TimSort, which is described in detail -/// [here](https://github.com/python/cpython/blob/main/Objects/listsort.txt). -/// -/// The algorithm identifies strictly descending and non-descending subsequences, which are called -/// natural runs. There is a stack of pending runs yet to be merged. Each newly found run is pushed -/// onto the stack, and then some pairs of adjacent runs are merged until these two invariants are -/// satisfied: -/// -/// 1. for every `i` in `1..runs.len()`: `runs[i - 1].len > runs[i].len` -/// 2. for every `i` in `2..runs.len()`: `runs[i - 2].len > runs[i - 1].len + runs[i].len` -/// -/// The invariants ensure that the total running time is *O*(*n* \* log(*n*)) worst-case. +#[inline] #[cfg(not(no_global_oom_handling))] -fn merge_sort(v: &mut [T], mut is_less: F) +fn stable_sort(v: &mut [T], mut is_less: F) where F: FnMut(&T, &T) -> bool, { - // Slices of up to this length get sorted using insertion sort. - const MAX_INSERTION: usize = 20; - // Very short runs are extended using insertion sort to span at least this many elements. - const MIN_RUN: usize = 10; - - // Sorting has no meaningful behavior on zero-sized types. if T::IS_ZST { + // Sorting has no meaningful behavior on zero-sized types. Do nothing. return; } - let len = v.len(); + let elem_alloc_fn = |len: usize| -> *mut T { + // SAFETY: Creating the layout is safe as long as merge_sort never calls this with len > + // v.len(). Alloc in general will only be used as 'shadow-region' to store temporary swap + // elements. + unsafe { alloc::alloc(alloc::Layout::array::(len).unwrap_unchecked()) as *mut T } + }; - // Short arrays get sorted in-place via insertion sort to avoid allocations. - if len <= MAX_INSERTION { - if len >= 2 { - for i in (0..len - 1).rev() { - insert_head(&mut v[i..], &mut is_less); - } - } - return; - } - - // Allocate a buffer to use as scratch memory. We keep the length 0 so we can keep in it - // shallow copies of the contents of `v` without risking the dtors running on copies if - // `is_less` panics. When merging two sorted runs, this buffer holds a copy of the shorter run, - // which will always have length at most `len / 2`. - let mut buf = Vec::with_capacity(len / 2); - - // In order to identify natural runs in `v`, we traverse it backwards. That might seem like a - // strange decision, but consider the fact that merges more often go in the opposite direction - // (forwards). According to benchmarks, merging forwards is slightly faster than merging - // backwards. To conclude, identifying runs by traversing backwards improves performance. - let mut runs = vec![]; - let mut end = len; - while end > 0 { - // Find the next natural run, and reverse it if it's strictly descending. - let mut start = end - 1; - if start > 0 { - start -= 1; - unsafe { - if is_less(v.get_unchecked(start + 1), v.get_unchecked(start)) { - while start > 0 && is_less(v.get_unchecked(start), v.get_unchecked(start - 1)) { - start -= 1; - } - v[start..end].reverse(); - } else { - while start > 0 && !is_less(v.get_unchecked(start), v.get_unchecked(start - 1)) - { - start -= 1; - } - } - } - } - - // Insert some more elements into the run if it's too short. Insertion sort is faster than - // merge sort on short sequences, so this significantly improves performance. - while start > 0 && end - start < MIN_RUN { - start -= 1; - insert_head(&mut v[start..end], &mut is_less); + let elem_dealloc_fn = |buf_ptr: *mut T, len: usize| { + // SAFETY: Creating the layout is safe as long as merge_sort never calls this with len > + // v.len(). The caller must ensure that buf_ptr was created by elem_alloc_fn with the same + // len. + unsafe { + alloc::dealloc(buf_ptr as *mut u8, alloc::Layout::array::(len).unwrap_unchecked()); } + }; - // Push this run onto the stack. - runs.push(Run { start, len: end - start }); - end = start; - - // Merge some pairs of adjacent runs to satisfy the invariants. - while let Some(r) = collapse(&runs) { - let left = runs[r + 1]; - let right = runs[r]; - unsafe { - merge( - &mut v[left.start..right.start + right.len], - left.len, - buf.as_mut_ptr(), - &mut is_less, - ); - } - runs[r] = Run { start: left.start, len: left.len + right.len }; - runs.remove(r + 1); + let run_alloc_fn = |len: usize| -> *mut sort::TimSortRun { + // SAFETY: Creating the layout is safe as long as merge_sort never calls this with an + // obscene length or 0. + unsafe { + alloc::alloc(alloc::Layout::array::(len).unwrap_unchecked()) + as *mut sort::TimSortRun } - } - - // Finally, exactly one run must remain in the stack. - debug_assert!(runs.len() == 1 && runs[0].start == 0 && runs[0].len == len); + }; - // Examines the stack of runs and identifies the next pair of runs to merge. More specifically, - // if `Some(r)` is returned, that means `runs[r]` and `runs[r + 1]` must be merged next. If the - // algorithm should continue building a new run instead, `None` is returned. - // - // TimSort is infamous for its buggy implementations, as described here: - // http://envisage-project.eu/timsort-specification-and-verification/ - // - // The gist of the story is: we must enforce the invariants on the top four runs on the stack. - // Enforcing them on just top three is not sufficient to ensure that the invariants will still - // hold for *all* runs in the stack. - // - // This function correctly checks invariants for the top four runs. Additionally, if the top - // run starts at index 0, it will always demand a merge operation until the stack is fully - // collapsed, in order to complete the sort. - #[inline] - fn collapse(runs: &[Run]) -> Option { - let n = runs.len(); - if n >= 2 - && (runs[n - 1].start == 0 - || runs[n - 2].len <= runs[n - 1].len - || (n >= 3 && runs[n - 3].len <= runs[n - 2].len + runs[n - 1].len) - || (n >= 4 && runs[n - 4].len <= runs[n - 3].len + runs[n - 2].len)) - { - if n >= 3 && runs[n - 3].len < runs[n - 1].len { Some(n - 3) } else { Some(n - 2) } - } else { - None + let run_dealloc_fn = |buf_ptr: *mut sort::TimSortRun, len: usize| { + // SAFETY: The caller must ensure that buf_ptr was created by elem_alloc_fn with the same + // len. + unsafe { + alloc::dealloc( + buf_ptr as *mut u8, + alloc::Layout::array::(len).unwrap_unchecked(), + ); } - } + }; - #[derive(Clone, Copy)] - struct Run { - start: usize, - len: usize, - } + sort::merge_sort(v, &mut is_less, elem_alloc_fn, elem_dealloc_fn, run_alloc_fn, run_dealloc_fn); } diff --git a/library/alloc/src/slice/tests.rs b/library/alloc/src/slice/tests.rs new file mode 100644 index 000000000..f674530aa --- /dev/null +++ b/library/alloc/src/slice/tests.rs @@ -0,0 +1,359 @@ +use crate::borrow::ToOwned; +use crate::rc::Rc; +use crate::string::ToString; +use crate::test_helpers::test_rng; +use crate::vec::Vec; + +use core::cell::Cell; +use core::cmp::Ordering::{self, Equal, Greater, Less}; +use core::convert::identity; +use core::fmt; +use core::mem; +use core::sync::atomic::{AtomicUsize, Ordering::Relaxed}; +use rand::{distributions::Standard, prelude::*, Rng, RngCore}; +use std::panic; + +macro_rules! do_test { + ($input:ident, $func:ident) => { + let len = $input.len(); + + // Work out the total number of comparisons required to sort + // this array... + let mut count = 0usize; + $input.to_owned().$func(|a, b| { + count += 1; + a.cmp(b) + }); + + // ... and then panic on each and every single one. + for panic_countdown in 0..count { + // Refresh the counters. + VERSIONS.store(0, Relaxed); + for i in 0..len { + DROP_COUNTS[i].store(0, Relaxed); + } + + let v = $input.to_owned(); + let _ = std::panic::catch_unwind(move || { + let mut v = v; + let mut panic_countdown = panic_countdown; + v.$func(|a, b| { + if panic_countdown == 0 { + SILENCE_PANIC.with(|s| s.set(true)); + panic!(); + } + panic_countdown -= 1; + a.cmp(b) + }) + }); + + // Check that the number of things dropped is exactly + // what we expect (i.e., the contents of `v`). + for (i, c) in DROP_COUNTS.iter().enumerate().take(len) { + let count = c.load(Relaxed); + assert!(count == 1, "found drop count == {} for i == {}, len == {}", count, i, len); + } + + // Check that the most recent versions of values were dropped. + assert_eq!(VERSIONS.load(Relaxed), 0); + } + }; +} + +const MAX_LEN: usize = 80; + +static DROP_COUNTS: [AtomicUsize; MAX_LEN] = [ + // FIXME(RFC 1109): AtomicUsize is not Copy. + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), + AtomicUsize::new(0), +]; + +static VERSIONS: AtomicUsize = AtomicUsize::new(0); + +#[derive(Clone, Eq)] +struct DropCounter { + x: u32, + id: usize, + version: Cell, +} + +impl PartialEq for DropCounter { + fn eq(&self, other: &Self) -> bool { + self.partial_cmp(other) == Some(Ordering::Equal) + } +} + +impl PartialOrd for DropCounter { + fn partial_cmp(&self, other: &Self) -> Option { + self.version.set(self.version.get() + 1); + other.version.set(other.version.get() + 1); + VERSIONS.fetch_add(2, Relaxed); + self.x.partial_cmp(&other.x) + } +} + +impl Ord for DropCounter { + fn cmp(&self, other: &Self) -> Ordering { + self.partial_cmp(other).unwrap() + } +} + +impl Drop for DropCounter { + fn drop(&mut self) { + DROP_COUNTS[self.id].fetch_add(1, Relaxed); + VERSIONS.fetch_sub(self.version.get(), Relaxed); + } +} + +std::thread_local!(static SILENCE_PANIC: Cell = Cell::new(false)); + +#[test] +#[cfg_attr(target_os = "emscripten", ignore)] // no threads +fn panic_safe() { + panic::update_hook(move |prev, info| { + if !SILENCE_PANIC.with(|s| s.get()) { + prev(info); + } + }); + + let mut rng = test_rng(); + + // Miri is too slow (but still need to `chain` to make the types match) + let lens = if cfg!(miri) { (1..10).chain(0..0) } else { (1..20).chain(70..MAX_LEN) }; + let moduli: &[u32] = if cfg!(miri) { &[5] } else { &[5, 20, 50] }; + + for len in lens { + for &modulus in moduli { + for &has_runs in &[false, true] { + let mut input = (0..len) + .map(|id| DropCounter { + x: rng.next_u32() % modulus, + id: id, + version: Cell::new(0), + }) + .collect::>(); + + if has_runs { + for c in &mut input { + c.x = c.id as u32; + } + + for _ in 0..5 { + let a = rng.gen::() % len; + let b = rng.gen::() % len; + if a < b { + input[a..b].reverse(); + } else { + input.swap(a, b); + } + } + } + + do_test!(input, sort_by); + do_test!(input, sort_unstable_by); + } + } + } + + // Set default panic hook again. + drop(panic::take_hook()); +} + +#[test] +#[cfg_attr(miri, ignore)] // Miri is too slow +fn test_sort() { + let mut rng = test_rng(); + + for len in (2..25).chain(500..510) { + for &modulus in &[5, 10, 100, 1000] { + for _ in 0..10 { + let orig: Vec<_> = (&mut rng) + .sample_iter::(&Standard) + .map(|x| x % modulus) + .take(len) + .collect(); + + // Sort in default order. + let mut v = orig.clone(); + v.sort(); + assert!(v.windows(2).all(|w| w[0] <= w[1])); + + // Sort in ascending order. + let mut v = orig.clone(); + v.sort_by(|a, b| a.cmp(b)); + assert!(v.windows(2).all(|w| w[0] <= w[1])); + + // Sort in descending order. + let mut v = orig.clone(); + v.sort_by(|a, b| b.cmp(a)); + assert!(v.windows(2).all(|w| w[0] >= w[1])); + + // Sort in lexicographic order. + let mut v1 = orig.clone(); + let mut v2 = orig.clone(); + v1.sort_by_key(|x| x.to_string()); + v2.sort_by_cached_key(|x| x.to_string()); + assert!(v1.windows(2).all(|w| w[0].to_string() <= w[1].to_string())); + assert!(v1 == v2); + + // Sort with many pre-sorted runs. + let mut v = orig.clone(); + v.sort(); + v.reverse(); + for _ in 0..5 { + let a = rng.gen::() % len; + let b = rng.gen::() % len; + if a < b { + v[a..b].reverse(); + } else { + v.swap(a, b); + } + } + v.sort(); + assert!(v.windows(2).all(|w| w[0] <= w[1])); + } + } + } + + // Sort using a completely random comparison function. + // This will reorder the elements *somehow*, but won't panic. + let mut v = [0; 500]; + for i in 0..v.len() { + v[i] = i as i32; + } + v.sort_by(|_, _| *[Less, Equal, Greater].choose(&mut rng).unwrap()); + v.sort(); + for i in 0..v.len() { + assert_eq!(v[i], i as i32); + } + + // Should not panic. + [0i32; 0].sort(); + [(); 10].sort(); + [(); 100].sort(); + + let mut v = [0xDEADBEEFu64]; + v.sort(); + assert!(v == [0xDEADBEEF]); +} + +#[test] +fn test_sort_stability() { + // Miri is too slow + let large_range = if cfg!(miri) { 0..0 } else { 500..510 }; + let rounds = if cfg!(miri) { 1 } else { 10 }; + + let mut rng = test_rng(); + for len in (2..25).chain(large_range) { + for _ in 0..rounds { + let mut counts = [0; 10]; + + // create a vector like [(6, 1), (5, 1), (6, 2), ...], + // where the first item of each tuple is random, but + // the second item represents which occurrence of that + // number this element is, i.e., the second elements + // will occur in sorted order. + let orig: Vec<_> = (0..len) + .map(|_| { + let n = rng.gen::() % 10; + counts[n] += 1; + (n, counts[n]) + }) + .collect(); + + let mut v = orig.clone(); + // Only sort on the first element, so an unstable sort + // may mix up the counts. + v.sort_by(|&(a, _), &(b, _)| a.cmp(&b)); + + // This comparison includes the count (the second item + // of the tuple), so elements with equal first items + // will need to be ordered with increasing + // counts... i.e., exactly asserting that this sort is + // stable. + assert!(v.windows(2).all(|w| w[0] <= w[1])); + + let mut v = orig.clone(); + v.sort_by_cached_key(|&(x, _)| x); + assert!(v.windows(2).all(|w| w[0] <= w[1])); + } + } +} diff --git a/library/alloc/src/str.rs b/library/alloc/src/str.rs index b28d20cda..afbe5cfaf 100644 --- a/library/alloc/src/str.rs +++ b/library/alloc/src/str.rs @@ -559,10 +559,9 @@ impl str { #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")] #[inline] pub fn to_ascii_uppercase(&self) -> String { - let mut bytes = self.as_bytes().to_vec(); - bytes.make_ascii_uppercase(); - // make_ascii_uppercase() preserves the UTF-8 invariant. - unsafe { String::from_utf8_unchecked(bytes) } + let mut s = self.to_owned(); + s.make_ascii_uppercase(); + s } /// Returns a copy of this string where each character is mapped to its @@ -592,10 +591,9 @@ impl str { #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")] #[inline] pub fn to_ascii_lowercase(&self) -> String { - let mut bytes = self.as_bytes().to_vec(); - bytes.make_ascii_lowercase(); - // make_ascii_lowercase() preserves the UTF-8 invariant. - unsafe { String::from_utf8_unchecked(bytes) } + let mut s = self.to_owned(); + s.make_ascii_lowercase(); + s } } diff --git a/library/alloc/src/string.rs b/library/alloc/src/string.rs index 7a8e6f088..ca182c810 100644 --- a/library/alloc/src/string.rs +++ b/library/alloc/src/string.rs @@ -363,7 +363,7 @@ use crate::vec::Vec; /// [`as_str()`]: String::as_str #[derive(PartialOrd, Eq, Ord)] #[stable(feature = "rust1", since = "1.0.0")] -#[cfg_attr(all(not(bootstrap), not(test)), lang = "String")] +#[cfg_attr(not(test), lang = "String")] pub struct String { vec: Vec, } @@ -2548,6 +2548,15 @@ impl ToString for char { } } +#[cfg(not(no_global_oom_handling))] +#[stable(feature = "bool_to_string_specialization", since = "1.68.0")] +impl ToString for bool { + #[inline] + fn to_string(&self) -> String { + String::from(if *self { "true" } else { "false" }) + } +} + #[cfg(not(no_global_oom_handling))] #[stable(feature = "u8_to_string_specialization", since = "1.54.0")] impl ToString for u8 { @@ -2678,7 +2687,7 @@ impl From<&String> for String { } } -// note: test pulls in libstd, which causes errors here +// note: test pulls in std, which causes errors here #[cfg(not(test))] #[stable(feature = "string_from_box", since = "1.18.0")] impl From> for String { diff --git a/library/alloc/src/sync.rs b/library/alloc/src/sync.rs index f7dc4d109..bab7f5f53 100644 --- a/library/alloc/src/sync.rs +++ b/library/alloc/src/sync.rs @@ -254,7 +254,7 @@ unsafe impl Sync for Arc {} #[stable(feature = "catch_unwind", since = "1.9.0")] impl UnwindSafe for Arc {} -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized> for Arc {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] @@ -295,7 +295,7 @@ pub struct Weak { // This is a `NonNull` to allow optimizing the size of this type in enums, // but it is not necessarily a valid pointer. // `Weak::new` sets this to `usize::MAX` so that it doesn’t need - // to allocate space on the heap. That's not a value a real pointer + // to allocate space on the heap. That's not a value a real pointer // will ever have because RcBox has alignment at least 2. // This is only possible when `T: Sized`; unsized `T` never dangle. ptr: NonNull>, @@ -306,13 +306,13 @@ unsafe impl Send for Weak {} #[stable(feature = "arc_weak", since = "1.4.0")] unsafe impl Sync for Weak {} -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized> for Weak {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] impl, U: ?Sized> DispatchFromDyn> for Weak {} #[stable(feature = "arc_weak", since = "1.4.0")] -impl fmt::Debug for Weak { +impl fmt::Debug for Weak { fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { write!(f, "(Weak)") } @@ -1656,7 +1656,7 @@ impl Arc { // // The acquire label here ensures a happens-before relationship with any // writes to `strong` (in particular in `Weak::upgrade`) prior to decrements - // of the `weak` count (via `Weak::drop`, which uses release). If the upgraded + // of the `weak` count (via `Weak::drop`, which uses release). If the upgraded // weak ref was never dropped, the CAS here will fail so we do not care to synchronize. if self.inner().weak.compare_exchange(1, usize::MAX, Acquire, Relaxed).is_ok() { // This needs to be an `Acquire` to synchronize with the decrement of the `strong` @@ -1712,7 +1712,7 @@ unsafe impl<#[may_dangle] T: ?Sized> Drop for Arc { } // This fence is needed to prevent reordering of use of the data and - // deletion of the data. Because it is marked `Release`, the decreasing + // deletion of the data. Because it is marked `Release`, the decreasing // of the reference count synchronizes with this `Acquire` fence. This // means that use of the data happens before decreasing the reference // count, which happens before this fence, which happens before the @@ -2172,7 +2172,7 @@ impl Clone for Weak { } else { return Weak { ptr: self.ptr }; }; - // See comments in Arc::clone() for why this is relaxed. This can use a + // See comments in Arc::clone() for why this is relaxed. This can use a // fetch_add (ignoring the lock) because the weak count is only locked // where are *no other* weak pointers in existence. (So we can't be // running this code in that case). diff --git a/library/alloc/src/testing/crash_test.rs b/library/alloc/src/testing/crash_test.rs new file mode 100644 index 000000000..bcf5f5f72 --- /dev/null +++ b/library/alloc/src/testing/crash_test.rs @@ -0,0 +1,119 @@ +// We avoid relying on anything else in the crate, apart from the `Debug` trait. +use crate::fmt::Debug; +use std::cmp::Ordering; +use std::sync::atomic::{AtomicUsize, Ordering::SeqCst}; + +/// A blueprint for crash test dummy instances that monitor particular events. +/// Some instances may be configured to panic at some point. +/// Events are `clone`, `drop` or some anonymous `query`. +/// +/// Crash test dummies are identified and ordered by an id, so they can be used +/// as keys in a BTreeMap. +#[derive(Debug)] +pub struct CrashTestDummy { + pub id: usize, + cloned: AtomicUsize, + dropped: AtomicUsize, + queried: AtomicUsize, +} + +impl CrashTestDummy { + /// Creates a crash test dummy design. The `id` determines order and equality of instances. + pub fn new(id: usize) -> CrashTestDummy { + CrashTestDummy { + id, + cloned: AtomicUsize::new(0), + dropped: AtomicUsize::new(0), + queried: AtomicUsize::new(0), + } + } + + /// Creates an instance of a crash test dummy that records what events it experiences + /// and optionally panics. + pub fn spawn(&self, panic: Panic) -> Instance<'_> { + Instance { origin: self, panic } + } + + /// Returns how many times instances of the dummy have been cloned. + pub fn cloned(&self) -> usize { + self.cloned.load(SeqCst) + } + + /// Returns how many times instances of the dummy have been dropped. + pub fn dropped(&self) -> usize { + self.dropped.load(SeqCst) + } + + /// Returns how many times instances of the dummy have had their `query` member invoked. + pub fn queried(&self) -> usize { + self.queried.load(SeqCst) + } +} + +#[derive(Debug)] +pub struct Instance<'a> { + origin: &'a CrashTestDummy, + panic: Panic, +} + +#[derive(Copy, Clone, Debug, PartialEq, Eq)] +pub enum Panic { + Never, + InClone, + InDrop, + InQuery, +} + +impl Instance<'_> { + pub fn id(&self) -> usize { + self.origin.id + } + + /// Some anonymous query, the result of which is already given. + pub fn query(&self, result: R) -> R { + self.origin.queried.fetch_add(1, SeqCst); + if self.panic == Panic::InQuery { + panic!("panic in `query`"); + } + result + } +} + +impl Clone for Instance<'_> { + fn clone(&self) -> Self { + self.origin.cloned.fetch_add(1, SeqCst); + if self.panic == Panic::InClone { + panic!("panic in `clone`"); + } + Self { origin: self.origin, panic: Panic::Never } + } +} + +impl Drop for Instance<'_> { + fn drop(&mut self) { + self.origin.dropped.fetch_add(1, SeqCst); + if self.panic == Panic::InDrop { + panic!("panic in `drop`"); + } + } +} + +impl PartialOrd for Instance<'_> { + fn partial_cmp(&self, other: &Self) -> Option { + self.id().partial_cmp(&other.id()) + } +} + +impl Ord for Instance<'_> { + fn cmp(&self, other: &Self) -> Ordering { + self.id().cmp(&other.id()) + } +} + +impl PartialEq for Instance<'_> { + fn eq(&self, other: &Self) -> bool { + self.id().eq(&other.id()) + } +} + +impl Eq for Instance<'_> {} diff --git a/library/alloc/src/testing/mod.rs b/library/alloc/src/testing/mod.rs new file mode 100644 index 000000000..7a094f8a5 --- /dev/null +++ b/library/alloc/src/testing/mod.rs @@ -0,0 +1,3 @@ +pub mod crash_test; +pub mod ord_chaos; +pub mod rng; diff --git a/library/alloc/src/testing/ord_chaos.rs b/library/alloc/src/testing/ord_chaos.rs new file mode 100644 index 000000000..96ce7c157 --- /dev/null +++ b/library/alloc/src/testing/ord_chaos.rs @@ -0,0 +1,81 @@ +use std::cell::Cell; +use std::cmp::Ordering::{self, *}; +use std::ptr; + +// Minimal type with an `Ord` implementation violating transitivity. +#[derive(Debug)] +pub enum Cyclic3 { + A, + B, + C, +} +use Cyclic3::*; + +impl PartialOrd for Cyclic3 { + fn partial_cmp(&self, other: &Self) -> Option { + Some(self.cmp(other)) + } +} + +impl Ord for Cyclic3 { + fn cmp(&self, other: &Self) -> Ordering { + match (self, other) { + (A, A) | (B, B) | (C, C) => Equal, + (A, B) | (B, C) | (C, A) => Less, + (A, C) | (B, A) | (C, B) => Greater, + } + } +} + +impl PartialEq for Cyclic3 { + fn eq(&self, other: &Self) -> bool { + self.cmp(&other) == Equal + } +} + +impl Eq for Cyclic3 {} + +// Controls the ordering of values wrapped by `Governed`. +#[derive(Debug)] +pub struct Governor { + flipped: Cell, +} + +impl Governor { + pub fn new() -> Self { + Governor { flipped: Cell::new(false) } + } + + pub fn flip(&self) { + self.flipped.set(!self.flipped.get()); + } +} + +// Type with an `Ord` implementation that forms a total order at any moment +// (assuming that `T` respects total order), but can suddenly be made to invert +// that total order. +#[derive(Debug)] +pub struct Governed<'a, T>(pub T, pub &'a Governor); + +impl PartialOrd for Governed<'_, T> { + fn partial_cmp(&self, other: &Self) -> Option { + Some(self.cmp(other)) + } +} + +impl Ord for Governed<'_, T> { + fn cmp(&self, other: &Self) -> Ordering { + assert!(ptr::eq(self.1, other.1)); + let ord = self.0.cmp(&other.0); + if self.1.flipped.get() { ord.reverse() } else { ord } + } +} + +impl PartialEq for Governed<'_, T> { + fn eq(&self, other: &Self) -> bool { + assert!(ptr::eq(self.1, other.1)); + self.0.eq(&other.0) + } +} + +impl Eq for Governed<'_, T> {} diff --git a/library/alloc/src/testing/rng.rs b/library/alloc/src/testing/rng.rs new file mode 100644 index 000000000..ecf543bee --- /dev/null +++ b/library/alloc/src/testing/rng.rs @@ -0,0 +1,28 @@ +/// XorShiftRng +pub struct DeterministicRng { + count: usize, + x: u32, + y: u32, + z: u32, + w: u32, +} + +impl DeterministicRng { + pub fn new() -> Self { + DeterministicRng { count: 0, x: 0x193a6754, y: 0xa8a7d469, z: 0x97830e05, w: 0x113ba7bb } + } + + /// Guarantees that each returned number is unique. + pub fn next(&mut self) -> u32 { + self.count += 1; + assert!(self.count <= 70029); + let x = self.x; + let t = x ^ (x << 11); + self.x = self.y; + self.y = self.z; + self.z = self.w; + let w_ = self.w; + self.w = w_ ^ (w_ >> 19) ^ (t ^ (t >> 8)); + self.w + } +} diff --git a/library/alloc/src/vec/drain.rs b/library/alloc/src/vec/drain.rs index 541f99bcf..2b1a787cc 100644 --- a/library/alloc/src/vec/drain.rs +++ b/library/alloc/src/vec/drain.rs @@ -223,9 +223,9 @@ impl Drop for Drain<'_, T, A> { } // as_slice() must only be called when iter.len() is > 0 because - // vec::Splice modifies vec::Drain fields and may grow the vec which would invalidate - // the iterator's internal pointers. Creating a reference to deallocated memory - // is invalid even when it is zero-length + // it also gets touched by vec::Splice which may turn it into a dangling pointer + // which would make it and the vec pointer point to different allocations which would + // lead to invalid pointer arithmetic below. let drop_ptr = iter.as_slice().as_ptr(); unsafe { diff --git a/library/alloc/src/vec/into_iter.rs b/library/alloc/src/vec/into_iter.rs index 02cc7691a..37966007e 100644 --- a/library/alloc/src/vec/into_iter.rs +++ b/library/alloc/src/vec/into_iter.rs @@ -1,6 +1,8 @@ #[cfg(not(no_global_oom_handling))] use super::AsVecIntoIter; use crate::alloc::{Allocator, Global}; +#[cfg(not(no_global_oom_handling))] +use crate::collections::VecDeque; use crate::raw_vec::RawVec; use core::array; use core::fmt; @@ -38,7 +40,9 @@ pub struct IntoIter< // to avoid dropping the allocator twice we need to wrap it into ManuallyDrop pub(super) alloc: ManuallyDrop, pub(super) ptr: *const T, - pub(super) end: *const T, + pub(super) end: *const T, // If T is a ZST, this is actually ptr+len. This encoding is picked so that + // ptr == end is a quick test for the Iterator being empty, that works + // for both ZST and non-ZST. } #[stable(feature = "vec_intoiter_debug", since = "1.13.0")] @@ -130,7 +134,36 @@ impl IntoIter { /// Forgets to Drop the remaining elements while still allowing the backing allocation to be freed. pub(crate) fn forget_remaining_elements(&mut self) { - self.ptr = self.end; + // For th ZST case, it is crucial that we mutate `end` here, not `ptr`. + // `ptr` must stay aligned, while `end` may be unaligned. + self.end = self.ptr; + } + + #[cfg(not(no_global_oom_handling))] + #[inline] + pub(crate) fn into_vecdeque(self) -> VecDeque { + // Keep our `Drop` impl from dropping the elements and the allocator + let mut this = ManuallyDrop::new(self); + + // SAFETY: This allocation originally came from a `Vec`, so it passes + // all those checks. We have `this.buf` ≤ `this.ptr` ≤ `this.end`, + // so the `sub_ptr`s below cannot wrap, and will produce a well-formed + // range. `end` ≤ `buf + cap`, so the range will be in-bounds. + // Taking `alloc` is ok because nothing else is going to look at it, + // since our `Drop` impl isn't going to run so there's no more code. + unsafe { + let buf = this.buf.as_ptr(); + let initialized = if T::IS_ZST { + // All the pointers are the same for ZSTs, so it's fine to + // say that they're all at the beginning of the "allocation". + 0..this.len() + } else { + this.ptr.sub_ptr(buf)..this.end.sub_ptr(buf) + }; + let cap = this.cap; + let alloc = ManuallyDrop::take(&mut this.alloc); + VecDeque::from_contiguous_raw_parts_in(buf, initialized, cap, alloc) + } } } @@ -155,10 +188,9 @@ impl Iterator for IntoIter { if self.ptr == self.end { None } else if T::IS_ZST { - // purposefully don't use 'ptr.offset' because for - // vectors with 0-size elements this would return the - // same pointer. - self.ptr = self.ptr.wrapping_byte_add(1); + // `ptr` has to stay where it is to remain aligned, so we reduce the length by 1 by + // reducing the `end`. + self.end = self.end.wrapping_byte_sub(1); // Make up a value of this ZST. Some(unsafe { mem::zeroed() }) @@ -185,10 +217,8 @@ impl Iterator for IntoIter { let step_size = self.len().min(n); let to_drop = ptr::slice_from_raw_parts_mut(self.ptr as *mut T, step_size); if T::IS_ZST { - // SAFETY: due to unchecked casts of unsigned amounts to signed offsets the wraparound - // effectively results in unsigned pointers representing positions 0..usize::MAX, - // which is valid for ZSTs. - self.ptr = self.ptr.wrapping_byte_add(step_size); + // See `next` for why we sub `end` here. + self.end = self.end.wrapping_byte_sub(step_size); } else { // SAFETY: the min() above ensures that step_size is in bounds self.ptr = unsafe { self.ptr.add(step_size) }; @@ -221,7 +251,7 @@ impl Iterator for IntoIter { return Err(unsafe { array::IntoIter::new_unchecked(raw_ary, 0..len) }); } - self.ptr = self.ptr.wrapping_byte_add(N); + self.end = self.end.wrapping_byte_sub(N); // Safety: ditto return Ok(unsafe { raw_ary.transpose().assume_init() }); } diff --git a/library/alloc/src/vec/is_zero.rs b/library/alloc/src/vec/is_zero.rs index 8e652d676..cb9adf05c 100644 --- a/library/alloc/src/vec/is_zero.rs +++ b/library/alloc/src/vec/is_zero.rs @@ -4,7 +4,8 @@ use crate::boxed::Box; #[rustc_specialization_trait] pub(super) unsafe trait IsZero { - /// Whether this value's representation is all zeros + /// Whether this value's representation is all zeros, + /// or can be represented with all zeroes. fn is_zero(&self) -> bool; } @@ -57,7 +58,7 @@ unsafe impl IsZero for [T; N] { #[inline] fn is_zero(&self) -> bool { // Because this is generated as a runtime check, it's not obvious that - // it's worth doing if the array is really long. The threshold here + // it's worth doing if the array is really long. The threshold here // is largely arbitrary, but was picked because as of 2022-07-01 LLVM // fails to const-fold the check in `vec![[1; 32]; n]` // See https://github.com/rust-lang/rust/pull/97581#issuecomment-1166628022 @@ -147,6 +148,23 @@ impl_is_zero_option_of_nonzero!( NonZeroIsize, ); +macro_rules! impl_is_zero_option_of_num { + ($($t:ty,)+) => {$( + unsafe impl IsZero for Option<$t> { + #[inline] + fn is_zero(&self) -> bool { + const { + let none: Self = unsafe { core::mem::MaybeUninit::zeroed().assume_init() }; + assert!(none.is_none()); + } + self.is_none() + } + } + )+}; +} + +impl_is_zero_option_of_num!(u8, u16, u32, u64, u128, i8, i16, i32, i64, i128, usize, isize,); + unsafe impl IsZero for Wrapping { #[inline] fn is_zero(&self) -> bool { diff --git a/library/alloc/src/vec/mod.rs b/library/alloc/src/vec/mod.rs index ba34ab680..36b0b3c9e 100644 --- a/library/alloc/src/vec/mod.rs +++ b/library/alloc/src/vec/mod.rs @@ -166,7 +166,7 @@ mod spec_extend; /// vec[0] = 7; /// assert_eq!(vec[0], 7); /// -/// vec.extend([1, 2, 3].iter().copied()); +/// vec.extend([1, 2, 3]); /// /// for x in &vec { /// println!("{x}"); @@ -490,6 +490,8 @@ impl Vec { /// This is highly unsafe, due to the number of invariants that aren't /// checked: /// + /// * `ptr` must have been allocated using the global allocator, such as via + /// the [`alloc::alloc`] function. /// * `T` needs to have the same alignment as what `ptr` was allocated with. /// (`T` having a less strict alignment is not sufficient, the alignment really /// needs to be equal to satisfy the [`dealloc`] requirement that memory must be @@ -526,6 +528,7 @@ impl Vec { /// function. /// /// [`String`]: crate::string::String + /// [`alloc::alloc`]: crate::alloc::alloc /// [`dealloc`]: crate::alloc::GlobalAlloc::dealloc /// /// # Examples @@ -681,6 +684,7 @@ impl Vec { /// This is highly unsafe, due to the number of invariants that aren't /// checked: /// + /// * `ptr` must be [*currently allocated*] via the given allocator `alloc`. /// * `T` needs to have the same alignment as what `ptr` was allocated with. /// (`T` having a less strict alignment is not sufficient, the alignment really /// needs to be equal to satisfy the [`dealloc`] requirement that memory must be @@ -714,6 +718,7 @@ impl Vec { /// /// [`String`]: crate::string::String /// [`dealloc`]: crate::alloc::GlobalAlloc::dealloc + /// [*currently allocated*]: crate::alloc::Allocator#currently-allocated-memory /// [*fit*]: crate::alloc::Allocator#memory-fitting /// /// # Examples @@ -2424,7 +2429,7 @@ impl Vec { self.reserve(range.len()); // SAFETY: - // - `slice::range` guarantees that the given range is valid for indexing self + // - `slice::range` guarantees that the given range is valid for indexing self unsafe { self.spec_extend_from_within(range); } @@ -2681,7 +2686,7 @@ impl Clone for Vec { // HACK(japaric): with cfg(test) the inherent `[T]::to_vec` method, which is // required for this method definition, is not available. Instead use the - // `slice::to_vec` function which is only available with cfg(test) + // `slice::to_vec` function which is only available with cfg(test) // NB see the slice::hack module in slice.rs for more information #[cfg(test)] fn clone(&self) -> Self { @@ -3191,7 +3196,7 @@ where } } -// note: test pulls in libstd, which causes errors here +// note: test pulls in std, which causes errors here #[cfg(not(test))] #[stable(feature = "vec_from_box", since = "1.18.0")] impl From> for Vec { @@ -3209,7 +3214,7 @@ impl From> for Vec { } } -// note: test pulls in libstd, which causes errors here +// note: test pulls in std, which causes errors here #[cfg(not(no_global_oom_handling))] #[cfg(not(test))] #[stable(feature = "box_from_vec", since = "1.20.0")] diff --git a/library/alloc/src/vec/splice.rs b/library/alloc/src/vec/splice.rs index bad765c7f..1861147fe 100644 --- a/library/alloc/src/vec/splice.rs +++ b/library/alloc/src/vec/splice.rs @@ -54,6 +54,12 @@ impl ExactSizeIterator for Splice<'_, I, A> {} impl Drop for Splice<'_, I, A> { fn drop(&mut self) { self.drain.by_ref().for_each(drop); + // At this point draining is done and the only remaining tasks are splicing + // and moving things into the final place. + // Which means we can replace the slice::Iter with pointers that won't point to deallocated + // memory, so that Drain::drop is still allowed to call iter.len(), otherwise it would break + // the ptr.sub_ptr contract. + self.drain.iter = (&[]).iter(); unsafe { if self.drain.tail_len == 0 { diff --git a/library/alloc/tests/autotraits.rs b/library/alloc/tests/autotraits.rs index 8ff5f0abe..879e32b3f 100644 --- a/library/alloc/tests/autotraits.rs +++ b/library/alloc/tests/autotraits.rs @@ -32,7 +32,7 @@ fn test_btree_map() { // spawn(f()); // } // - // where with some unintentionally overconstrained Send impls in liballoc's + // where with some unintentionally overconstrained Send impls in alloc's // internals, the future might incorrectly not be Send even though every // single type involved in the program is Send and Sync. require_send_sync(async { diff --git a/library/alloc/tests/lib.rs b/library/alloc/tests/lib.rs index d6d2b055b..2a93a242d 100644 --- a/library/alloc/tests/lib.rs +++ b/library/alloc/tests/lib.rs @@ -1,7 +1,6 @@ #![feature(allocator_api)] #![feature(alloc_layout_extra)] #![feature(assert_matches)] -#![feature(box_syntax)] #![feature(btree_drain_filter)] #![feature(cow_is_borrowed)] #![feature(const_box)] diff --git a/library/alloc/tests/slice.rs b/library/alloc/tests/slice.rs index 21f894343..0693beb48 100644 --- a/library/alloc/tests/slice.rs +++ b/library/alloc/tests/slice.rs @@ -1,15 +1,9 @@ -use std::cell::Cell; -use std::cmp::Ordering::{self, Equal, Greater, Less}; +use std::cmp::Ordering::{Equal, Greater, Less}; use std::convert::identity; use std::fmt; use std::mem; use std::panic; use std::rc::Rc; -use std::sync::atomic::{AtomicUsize, Ordering::Relaxed}; - -use rand::distributions::Standard; -use rand::seq::SliceRandom; -use rand::{thread_rng, Rng, RngCore}; fn square(n: usize) -> usize { n * n @@ -388,123 +382,6 @@ fn test_reverse() { assert_eq!(v, (-50..51i16).rev().collect::>()); } -#[test] -#[cfg_attr(miri, ignore)] // Miri is too slow -fn test_sort() { - let mut rng = thread_rng(); - - for len in (2..25).chain(500..510) { - for &modulus in &[5, 10, 100, 1000] { - for _ in 0..10 { - let orig: Vec<_> = - rng.sample_iter::(&Standard).map(|x| x % modulus).take(len).collect(); - - // Sort in default order. - let mut v = orig.clone(); - v.sort(); - assert!(v.windows(2).all(|w| w[0] <= w[1])); - - // Sort in ascending order. - let mut v = orig.clone(); - v.sort_by(|a, b| a.cmp(b)); - assert!(v.windows(2).all(|w| w[0] <= w[1])); - - // Sort in descending order. - let mut v = orig.clone(); - v.sort_by(|a, b| b.cmp(a)); - assert!(v.windows(2).all(|w| w[0] >= w[1])); - - // Sort in lexicographic order. - let mut v1 = orig.clone(); - let mut v2 = orig.clone(); - v1.sort_by_key(|x| x.to_string()); - v2.sort_by_cached_key(|x| x.to_string()); - assert!(v1.windows(2).all(|w| w[0].to_string() <= w[1].to_string())); - assert!(v1 == v2); - - // Sort with many pre-sorted runs. - let mut v = orig.clone(); - v.sort(); - v.reverse(); - for _ in 0..5 { - let a = rng.gen::() % len; - let b = rng.gen::() % len; - if a < b { - v[a..b].reverse(); - } else { - v.swap(a, b); - } - } - v.sort(); - assert!(v.windows(2).all(|w| w[0] <= w[1])); - } - } - } - - // Sort using a completely random comparison function. - // This will reorder the elements *somehow*, but won't panic. - let mut v = [0; 500]; - for i in 0..v.len() { - v[i] = i as i32; - } - v.sort_by(|_, _| *[Less, Equal, Greater].choose(&mut rng).unwrap()); - v.sort(); - for i in 0..v.len() { - assert_eq!(v[i], i as i32); - } - - // Should not panic. - [0i32; 0].sort(); - [(); 10].sort(); - [(); 100].sort(); - - let mut v = [0xDEADBEEFu64]; - v.sort(); - assert!(v == [0xDEADBEEF]); -} - -#[test] -fn test_sort_stability() { - // Miri is too slow - let large_range = if cfg!(miri) { 0..0 } else { 500..510 }; - let rounds = if cfg!(miri) { 1 } else { 10 }; - - for len in (2..25).chain(large_range) { - for _ in 0..rounds { - let mut counts = [0; 10]; - - // create a vector like [(6, 1), (5, 1), (6, 2), ...], - // where the first item of each tuple is random, but - // the second item represents which occurrence of that - // number this element is, i.e., the second elements - // will occur in sorted order. - let orig: Vec<_> = (0..len) - .map(|_| { - let n = thread_rng().gen::() % 10; - counts[n] += 1; - (n, counts[n]) - }) - .collect(); - - let mut v = orig.clone(); - // Only sort on the first element, so an unstable sort - // may mix up the counts. - v.sort_by(|&(a, _), &(b, _)| a.cmp(&b)); - - // This comparison includes the count (the second item - // of the tuple), so elements with equal first items - // will need to be ordered with increasing - // counts... i.e., exactly asserting that this sort is - // stable. - assert!(v.windows(2).all(|w| w[0] <= w[1])); - - let mut v = orig.clone(); - v.sort_by_cached_key(|&(x, _)| x); - assert!(v.windows(2).all(|w| w[0] <= w[1])); - } - } -} - #[test] fn test_rotate_left() { let expected: Vec<_> = (0..13).collect(); @@ -1608,230 +1485,6 @@ fn test_copy_from_slice_dst_shorter() { dst.copy_from_slice(&src); } -const MAX_LEN: usize = 80; - -static DROP_COUNTS: [AtomicUsize; MAX_LEN] = [ - // FIXME(RFC 1109): AtomicUsize is not Copy. - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), - AtomicUsize::new(0), -]; - -static VERSIONS: AtomicUsize = AtomicUsize::new(0); - -#[derive(Clone, Eq)] -struct DropCounter { - x: u32, - id: usize, - version: Cell, -} - -impl PartialEq for DropCounter { - fn eq(&self, other: &Self) -> bool { - self.partial_cmp(other) == Some(Ordering::Equal) - } -} - -impl PartialOrd for DropCounter { - fn partial_cmp(&self, other: &Self) -> Option { - self.version.set(self.version.get() + 1); - other.version.set(other.version.get() + 1); - VERSIONS.fetch_add(2, Relaxed); - self.x.partial_cmp(&other.x) - } -} - -impl Ord for DropCounter { - fn cmp(&self, other: &Self) -> Ordering { - self.partial_cmp(other).unwrap() - } -} - -impl Drop for DropCounter { - fn drop(&mut self) { - DROP_COUNTS[self.id].fetch_add(1, Relaxed); - VERSIONS.fetch_sub(self.version.get(), Relaxed); - } -} - -macro_rules! test { - ($input:ident, $func:ident) => { - let len = $input.len(); - - // Work out the total number of comparisons required to sort - // this array... - let mut count = 0usize; - $input.to_owned().$func(|a, b| { - count += 1; - a.cmp(b) - }); - - // ... and then panic on each and every single one. - for panic_countdown in 0..count { - // Refresh the counters. - VERSIONS.store(0, Relaxed); - for i in 0..len { - DROP_COUNTS[i].store(0, Relaxed); - } - - let v = $input.to_owned(); - let _ = std::panic::catch_unwind(move || { - let mut v = v; - let mut panic_countdown = panic_countdown; - v.$func(|a, b| { - if panic_countdown == 0 { - SILENCE_PANIC.with(|s| s.set(true)); - panic!(); - } - panic_countdown -= 1; - a.cmp(b) - }) - }); - - // Check that the number of things dropped is exactly - // what we expect (i.e., the contents of `v`). - for (i, c) in DROP_COUNTS.iter().enumerate().take(len) { - let count = c.load(Relaxed); - assert!(count == 1, "found drop count == {} for i == {}, len == {}", count, i, len); - } - - // Check that the most recent versions of values were dropped. - assert_eq!(VERSIONS.load(Relaxed), 0); - } - }; -} - -thread_local!(static SILENCE_PANIC: Cell = Cell::new(false)); - -#[test] -#[cfg_attr(target_os = "emscripten", ignore)] // no threads -fn panic_safe() { - panic::update_hook(move |prev, info| { - if !SILENCE_PANIC.with(|s| s.get()) { - prev(info); - } - }); - - let mut rng = thread_rng(); - - // Miri is too slow (but still need to `chain` to make the types match) - let lens = if cfg!(miri) { (1..10).chain(0..0) } else { (1..20).chain(70..MAX_LEN) }; - let moduli: &[u32] = if cfg!(miri) { &[5] } else { &[5, 20, 50] }; - - for len in lens { - for &modulus in moduli { - for &has_runs in &[false, true] { - let mut input = (0..len) - .map(|id| DropCounter { - x: rng.next_u32() % modulus, - id: id, - version: Cell::new(0), - }) - .collect::>(); - - if has_runs { - for c in &mut input { - c.x = c.id as u32; - } - - for _ in 0..5 { - let a = rng.gen::() % len; - let b = rng.gen::() % len; - if a < b { - input[a..b].reverse(); - } else { - input.swap(a, b); - } - } - } - - test!(input, sort_by); - test!(input, sort_unstable_by); - } - } - } - - // Set default panic hook again. - drop(panic::take_hook()); -} - #[test] fn repeat_generic_slice() { assert_eq!([1, 2].repeat(2), vec![1, 2, 1, 2]); diff --git a/library/alloc/tests/vec.rs b/library/alloc/tests/vec.rs index 7ebed0d5c..2f07c2911 100644 --- a/library/alloc/tests/vec.rs +++ b/library/alloc/tests/vec.rs @@ -7,7 +7,9 @@ use std::borrow::Cow; use std::cell::Cell; use std::collections::TryReserveErrorKind::*; use std::fmt::Debug; +use std::hint; use std::iter::InPlaceIterable; +use std::mem; use std::mem::{size_of, swap}; use std::ops::Bound::*; use std::panic::{catch_unwind, AssertUnwindSafe}; @@ -1107,8 +1109,31 @@ fn test_into_iter_drop_allocator() { #[test] fn test_into_iter_zst() { - for _ in vec![[0u64; 0]].into_iter() {} - for _ in vec![[0u64; 0]; 5].into_iter().rev() {} + #[derive(Debug, Clone)] + struct AlignedZstWithDrop([u64; 0]); + impl Drop for AlignedZstWithDrop { + fn drop(&mut self) { + let addr = self as *mut _ as usize; + assert!(hint::black_box(addr) % mem::align_of::() == 0); + } + } + + const C: AlignedZstWithDrop = AlignedZstWithDrop([0u64; 0]); + + for _ in vec![C].into_iter() {} + for _ in vec![C; 5].into_iter().rev() {} + + let mut it = vec![C, C].into_iter(); + it.advance_by(1).unwrap(); + drop(it); + + let mut it = vec![C, C].into_iter(); + it.next_chunk::<1>().unwrap(); + drop(it); + + let mut it = vec![C, C].into_iter(); + it.next_chunk::<4>().unwrap_err(); + drop(it); } #[test] @@ -1824,7 +1849,7 @@ fn test_stable_pointers() { } // Test that, if we reserved enough space, adding and removing elements does not - // invalidate references into the vector (such as `v0`). This test also + // invalidate references into the vector (such as `v0`). This test also // runs in Miri, which would detect such problems. // Note that this test does *not* constitute a stable guarantee that all these functions do not // reallocate! Only what is explicitly documented at diff --git a/library/alloc/tests/vec_deque.rs b/library/alloc/tests/vec_deque.rs index d04de5a07..5a0b852e8 100644 --- a/library/alloc/tests/vec_deque.rs +++ b/library/alloc/tests/vec_deque.rs @@ -1045,6 +1045,20 @@ fn test_append_double_drop() { assert_eq!(count_b, 1); } +#[test] +#[should_panic] +fn test_append_zst_capacity_overflow() { + let mut v = Vec::with_capacity(usize::MAX); + // note: using resize instead of set_len here would + // be *extremely* slow in unoptimized builds. + // SAFETY: `v` has capacity `usize::MAX`, and no initialization + // is needed for empty tuples. + unsafe { v.set_len(usize::MAX) }; + let mut v = VecDeque::from(v); + let mut w = vec![()].into(); + v.append(&mut w); +} + #[test] fn test_retain() { let mut buf = VecDeque::new(); @@ -1736,3 +1750,39 @@ fn test_resize_keeps_reserved_space_from_item() { d.resize(1, v); assert_eq!(d[0].capacity(), 1234); } + +#[test] +fn test_collect_from_into_iter_keeps_allocation() { + let mut v = Vec::with_capacity(13); + v.extend(0..7); + check(v.as_ptr(), v.last().unwrap(), v.into_iter()); + + let mut v = VecDeque::with_capacity(13); + v.extend(0..7); + check(&v[0], &v[v.len() - 1], v.into_iter()); + + fn check(buf: *const i32, last: *const i32, mut it: impl Iterator) { + assert_eq!(it.next(), Some(0)); + assert_eq!(it.next(), Some(1)); + + let mut v: VecDeque = it.collect(); + assert_eq!(v.capacity(), 13); + assert_eq!(v.as_slices().0.as_ptr(), buf.wrapping_add(2)); + assert_eq!(&v[v.len() - 1] as *const _, last); + + assert_eq!(v.as_slices(), ([2, 3, 4, 5, 6].as_slice(), [].as_slice())); + v.push_front(7); + assert_eq!(v.as_slices(), ([7, 2, 3, 4, 5, 6].as_slice(), [].as_slice())); + v.push_front(8); + assert_eq!(v.as_slices(), ([8, 7, 2, 3, 4, 5, 6].as_slice(), [].as_slice())); + + // Now that we've adding thing in place of the two that we removed from + // the front of the iterator, we're back to matching the buffer pointer. + assert_eq!(v.as_slices().0.as_ptr(), buf); + assert_eq!(&v[v.len() - 1] as *const _, last); + + v.push_front(9); + assert_eq!(v.as_slices(), ([9].as_slice(), [8, 7, 2, 3, 4, 5, 6].as_slice())); + assert_eq!(v.capacity(), 13); + } +} diff --git a/library/core/Cargo.toml b/library/core/Cargo.toml index 2a7df9556..3dc8c84e0 100644 --- a/library/core/Cargo.toml +++ b/library/core/Cargo.toml @@ -24,8 +24,8 @@ path = "benches/lib.rs" test = true [dev-dependencies] -rand = "0.7" -rand_xorshift = "0.2" +rand = { version = "0.8.5", default-features = false } +rand_xorshift = { version = "0.3.0", default-features = false } [features] # Make panics and failed asserts immediately abort without formatting any message diff --git a/library/core/benches/num/int_log/mod.rs b/library/core/benches/num/int_log/mod.rs index 3c01e2998..bb61224b5 100644 --- a/library/core/benches/num/int_log/mod.rs +++ b/library/core/benches/num/int_log/mod.rs @@ -21,7 +21,7 @@ macro_rules! int_log_bench { /* Exponentially distributed random numbers from the whole range of the type. */ let numbers: Vec<$t> = (0..256) .map(|_| { - let x = rng.gen::<$t>() >> rng.gen_range(0, <$t>::BITS); + let x = rng.gen::<$t>() >> rng.gen_range(0..<$t>::BITS); if x != 0 { x } else { 1 } }) .collect(); @@ -38,7 +38,7 @@ macro_rules! int_log_bench { /* Exponentially distributed random numbers from the range 0..256. */ let numbers: Vec<$t> = (0..256) .map(|_| { - let x = (rng.gen::() >> rng.gen_range(0, u8::BITS)) as $t; + let x = (rng.gen::() >> rng.gen_range(0..u8::BITS)) as $t; if x != 0 { x } else { 1 } }) .collect(); diff --git a/library/core/src/any.rs b/library/core/src/any.rs index 1a379ecc1..c0fb0d993 100644 --- a/library/core/src/any.rs +++ b/library/core/src/any.rs @@ -148,7 +148,7 @@ //! ``` //! //! In this example, if the concrete type of `obj` in `use_my_trait` is `SomeConcreteType`, then -//! the `get_context_ref` call will return a reference to `obj.some_string` with type `&String`. +//! the `get_context_by_ref` call will return a reference to `obj.some_string` with type `&String`. #![stable(feature = "rust1", since = "1.0.0")] @@ -662,7 +662,8 @@ impl dyn Any + Send + Sync { /// While `TypeId` implements `Hash`, `PartialOrd`, and `Ord`, it is worth /// noting that the hashes and ordering will vary between Rust releases. Beware /// of relying on them inside of your code! -#[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Debug, Hash)] +#[derive(Clone, Copy, Debug, Hash, Eq)] +#[derive_const(PartialEq, PartialOrd, Ord)] #[stable(feature = "rust1", since = "1.0.0")] pub struct TypeId { t: u64, diff --git a/library/core/src/array/iter.rs b/library/core/src/array/iter.rs index b91c63018..8259c087d 100644 --- a/library/core/src/array/iter.rs +++ b/library/core/src/array/iter.rs @@ -109,8 +109,8 @@ impl IntoIter { /// use std::array::IntoIter; /// use std::mem::MaybeUninit; /// - /// # // Hi! Thanks for reading the code. This is restricted to `Copy` because - /// # // otherwise it could leak. A fully-general version this would need a drop + /// # // Hi! Thanks for reading the code. This is restricted to `Copy` because + /// # // otherwise it could leak. A fully-general version this would need a drop /// # // guard to handle panics from the iterator, but this works for an example. /// fn next_chunk( /// it: &mut impl Iterator, @@ -211,7 +211,7 @@ impl IntoIter { let initialized = 0..0; // SAFETY: We're telling it that none of the elements are initialized, - // which is trivially true. And ∀N: usize, 0 <= N. + // which is trivially true. And ∀N: usize, 0 <= N. unsafe { Self::new_unchecked(buffer, initialized) } } diff --git a/library/core/src/array/mod.rs b/library/core/src/array/mod.rs index 94a1a1d32..2825e0bbb 100644 --- a/library/core/src/array/mod.rs +++ b/library/core/src/array/mod.rs @@ -69,7 +69,7 @@ where /// if any element creation was unsuccessful. /// /// The return type of this function depends on the return type of the closure. -/// If you return `Result` from the closure, you'll get a `Result<[T; N]; E>`. +/// If you return `Result` from the closure, you'll get a `Result<[T; N], E>`. /// If you return `Option` from the closure, you'll get an `Option<[T; N]>`. /// /// # Arguments @@ -522,7 +522,7 @@ impl [T; N] { /// return an array the same size as `self` or the first error encountered. /// /// The return type of this function depends on the return type of the closure. - /// If you return `Result` from the closure, you'll get a `Result<[T; N]; E>`. + /// If you return `Result` from the closure, you'll get a `Result<[T; N], E>`. /// If you return `Option` from the closure, you'll get an `Option<[T; N]>`. /// /// # Examples diff --git a/library/core/src/borrow.rs b/library/core/src/borrow.rs index fdd56cb4e..4a8302ee4 100644 --- a/library/core/src/borrow.rs +++ b/library/core/src/borrow.rs @@ -26,7 +26,7 @@ /// to be modified, it can additionally implement [`BorrowMut`]. /// /// Further, when providing implementations for additional traits, it needs -/// to be considered whether they should behave identical to those of the +/// to be considered whether they should behave identically to those of the /// underlying type as a consequence of acting as a representation of that /// underlying type. Generic code typically uses `Borrow` when it relies /// on the identical behavior of these additional trait implementations. diff --git a/library/core/src/cell.rs b/library/core/src/cell.rs index 47cce2aa3..129213fde 100644 --- a/library/core/src/cell.rs +++ b/library/core/src/cell.rs @@ -568,7 +568,7 @@ impl Cell { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U> CoerceUnsized> for Cell {} impl Cell<[T]> { @@ -807,7 +807,8 @@ impl RefCell { /// /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently borrowed, or + /// if `self` and `other` point to the same `RefCell`. /// /// # Examples /// @@ -1193,7 +1194,7 @@ impl Default for RefCell { impl PartialEq for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn eq(&self, other: &RefCell) -> bool { *self.borrow() == *other.borrow() @@ -1207,7 +1208,7 @@ impl Eq for RefCell {} impl PartialOrd for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn partial_cmp(&self, other: &RefCell) -> Option { self.borrow().partial_cmp(&*other.borrow()) @@ -1215,7 +1216,7 @@ impl PartialOrd for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn lt(&self, other: &RefCell) -> bool { *self.borrow() < *other.borrow() @@ -1223,7 +1224,7 @@ impl PartialOrd for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn le(&self, other: &RefCell) -> bool { *self.borrow() <= *other.borrow() @@ -1231,7 +1232,7 @@ impl PartialOrd for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn gt(&self, other: &RefCell) -> bool { *self.borrow() > *other.borrow() @@ -1239,7 +1240,7 @@ impl PartialOrd for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn ge(&self, other: &RefCell) -> bool { *self.borrow() >= *other.borrow() @@ -1250,7 +1251,7 @@ impl PartialOrd for RefCell { impl Ord for RefCell { /// # Panics /// - /// Panics if the value in either `RefCell` is currently borrowed. + /// Panics if the value in either `RefCell` is currently mutably borrowed. #[inline] fn cmp(&self, other: &RefCell) -> Ordering { self.borrow().cmp(&*other.borrow()) @@ -1266,7 +1267,7 @@ impl const From for RefCell { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U> CoerceUnsized> for RefCell {} struct BorrowRef<'b> { @@ -1492,7 +1493,7 @@ impl<'b, T: ?Sized> Ref<'b, T> { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'b, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized> for Ref<'b, T> {} #[stable(feature = "std_guard_impls", since = "1.20.0")] @@ -1738,7 +1739,7 @@ impl DerefMut for RefMut<'_, T> { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'b, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized> for RefMut<'b, T> {} #[stable(feature = "std_guard_impls", since = "1.20.0")] @@ -1783,7 +1784,7 @@ impl fmt::Display for RefMut<'_, T> { /// until the reference expires. As a special exception, given an `&T`, any part of it that is /// inside an `UnsafeCell<_>` may be deallocated during the lifetime of the reference, after the /// last time the reference is used (dereferenced or reborrowed). Since you cannot deallocate a part -/// of what a reference points to, this means the memory an `&T` points to can be deallocted only if +/// of what a reference points to, this means the memory an `&T` points to can be deallocated only if /// *every part of it* (including padding) is inside an `UnsafeCell`. /// /// However, whenever a `&UnsafeCell` is constructed or dereferenced, it must still point to @@ -1993,7 +1994,7 @@ impl UnsafeCell { #[rustc_const_stable(feature = "const_unsafecell_get", since = "1.32.0")] pub const fn get(&self) -> *mut T { // We can just cast the pointer from `UnsafeCell` to `T` because of - // #[repr(transparent)]. This exploits libstd's special status, there is + // #[repr(transparent)]. This exploits std's special status, there is // no guarantee for user code that this will work in future versions of the compiler! self as *const UnsafeCell as *const T as *mut T } @@ -2051,7 +2052,7 @@ impl UnsafeCell { #[rustc_const_stable(feature = "unsafe_cell_raw_get", since = "1.56.0")] pub const fn raw_get(this: *const Self) -> *mut T { // We can just cast the pointer from `UnsafeCell` to `T` because of - // #[repr(transparent)]. This exploits libstd's special status, there is + // #[repr(transparent)]. This exploits std's special status, there is // no guarantee for user code that this will work in future versions of the compiler! this as *const T as *mut T } @@ -2074,7 +2075,7 @@ impl const From for UnsafeCell { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U> CoerceUnsized> for UnsafeCell {} /// [`UnsafeCell`], but [`Sync`]. @@ -2164,7 +2165,7 @@ impl const From for SyncUnsafeCell { } } -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] //#[unstable(feature = "sync_unsafe_cell", issue = "95439")] impl, U> CoerceUnsized> for SyncUnsafeCell {} diff --git a/library/core/src/cell/lazy.rs b/library/core/src/cell/lazy.rs index b355d94ce..65d12c25c 100644 --- a/library/core/src/cell/lazy.rs +++ b/library/core/src/cell/lazy.rs @@ -35,7 +35,7 @@ pub struct LazyCell T> { init: Cell>, } -impl LazyCell { +impl T> LazyCell { /// Creates a new lazy value with the given initializing function. /// /// # Examples @@ -51,13 +51,12 @@ impl LazyCell { /// /// assert_eq!(&*lazy, "HELLO, WORLD!"); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub const fn new(init: F) -> LazyCell { LazyCell { cell: OnceCell::new(), init: Cell::new(Some(init)) } } -} -impl T> LazyCell { /// Forces the evaluation of this lazy value and returns a reference to /// the result. /// @@ -75,6 +74,7 @@ impl T> LazyCell { /// assert_eq!(LazyCell::force(&lazy), &92); /// assert_eq!(&*lazy, &92); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn force(this: &LazyCell) -> &T { this.cell.get_or_init(|| match this.init.take() { @@ -87,6 +87,7 @@ impl T> LazyCell { #[unstable(feature = "once_cell", issue = "74465")] impl T> Deref for LazyCell { type Target = T; + #[inline] fn deref(&self) -> &T { LazyCell::force(self) } @@ -95,6 +96,7 @@ impl T> Deref for LazyCell { #[unstable(feature = "once_cell", issue = "74465")] impl Default for LazyCell { /// Creates a new lazy value using `Default` as the initializing function. + #[inline] fn default() -> LazyCell { LazyCell::new(T::default) } diff --git a/library/core/src/cell/once.rs b/library/core/src/cell/once.rs index 8c01643c7..7757068a4 100644 --- a/library/core/src/cell/once.rs +++ b/library/core/src/cell/once.rs @@ -37,8 +37,9 @@ pub struct OnceCell { impl OnceCell { /// Creates a new empty cell. - #[unstable(feature = "once_cell", issue = "74465")] + #[inline] #[must_use] + #[unstable(feature = "once_cell", issue = "74465")] pub const fn new() -> OnceCell { OnceCell { inner: UnsafeCell::new(None) } } @@ -46,6 +47,7 @@ impl OnceCell { /// Gets the reference to the underlying value. /// /// Returns `None` if the cell is empty. + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get(&self) -> Option<&T> { // SAFETY: Safe due to `inner`'s invariant @@ -55,6 +57,7 @@ impl OnceCell { /// Gets the mutable reference to the underlying value. /// /// Returns `None` if the cell is empty. + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get_mut(&mut self) -> Option<&mut T> { self.inner.get_mut().as_mut() @@ -82,6 +85,7 @@ impl OnceCell { /// /// assert!(cell.get().is_some()); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn set(&self, value: T) -> Result<(), T> { // SAFETY: Safe because we cannot have overlapping mutable borrows @@ -123,6 +127,7 @@ impl OnceCell { /// let value = cell.get_or_init(|| unreachable!()); /// assert_eq!(value, &92); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get_or_init(&self, f: F) -> &T where @@ -205,6 +210,7 @@ impl OnceCell { /// cell.set("hello".to_string()).unwrap(); /// assert_eq!(cell.into_inner(), Some("hello".to_string())); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn into_inner(self) -> Option { // Because `into_inner` takes `self` by value, the compiler statically verifies @@ -233,6 +239,7 @@ impl OnceCell { /// assert_eq!(cell.take(), Some("hello".to_string())); /// assert_eq!(cell.get(), None); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn take(&mut self) -> Option { mem::take(self).into_inner() @@ -241,6 +248,7 @@ impl OnceCell { #[unstable(feature = "once_cell", issue = "74465")] impl Default for OnceCell { + #[inline] fn default() -> Self { Self::new() } @@ -258,6 +266,7 @@ impl fmt::Debug for OnceCell { #[unstable(feature = "once_cell", issue = "74465")] impl Clone for OnceCell { + #[inline] fn clone(&self) -> OnceCell { let res = OnceCell::new(); if let Some(value) = self.get() { @@ -272,6 +281,7 @@ impl Clone for OnceCell { #[unstable(feature = "once_cell", issue = "74465")] impl PartialEq for OnceCell { + #[inline] fn eq(&self, other: &Self) -> bool { self.get() == other.get() } @@ -283,6 +293,7 @@ impl Eq for OnceCell {} #[unstable(feature = "once_cell", issue = "74465")] impl const From for OnceCell { /// Creates a new `OnceCell` which already contains the given `value`. + #[inline] fn from(value: T) -> Self { OnceCell { inner: UnsafeCell::new(Some(value)) } } diff --git a/library/core/src/char/decode.rs b/library/core/src/char/decode.rs index 11f1c30f6..eeb088030 100644 --- a/library/core/src/char/decode.rs +++ b/library/core/src/char/decode.rs @@ -67,7 +67,7 @@ impl> Iterator for DecodeUtf16 { } // all ok, so lets decode it. - let c = (((u - 0xD800) as u32) << 10 | (u2 - 0xDC00) as u32) + 0x1_0000; + let c = (((u & 0x3ff) as u32) << 10 | (u2 & 0x3ff) as u32) + 0x1_0000; // SAFETY: we checked that it's a legal unicode value Some(Ok(unsafe { from_u32_unchecked(c) })) } diff --git a/library/core/src/cmp.rs b/library/core/src/cmp.rs index 949896e57..a7d6fec7d 100644 --- a/library/core/src/cmp.rs +++ b/library/core/src/cmp.rs @@ -24,8 +24,6 @@ use crate::const_closure::ConstFnMutClosure; use crate::marker::Destruct; -#[cfg(bootstrap)] -use crate::marker::StructuralPartialEq; use self::Ordering::*; @@ -333,7 +331,7 @@ pub struct AssertParamIsEq { /// assert_eq!(Ordering::Greater, result); /// ``` #[derive(Clone, Copy, Eq, Debug, Hash)] -#[cfg_attr(not(bootstrap), derive_const(PartialOrd, Ord, PartialEq))] +#[derive_const(PartialOrd, Ord, PartialEq)] #[stable(feature = "rust1", since = "1.0.0")] #[repr(i8)] pub enum Ordering { @@ -800,9 +798,12 @@ pub trait Ord: Eq + PartialOrd { Self: Sized, Self: ~const Destruct, { - // HACK(fee1-dead): go back to using `self.max_by(other, Ord::cmp)` - // when trait methods are allowed to be used when a const closure is - // expected. + #[cfg(not(bootstrap))] + { + max_by(self, other, Ord::cmp) + } + + #[cfg(bootstrap)] match self.cmp(&other) { Ordering::Less | Ordering::Equal => other, Ordering::Greater => self, @@ -827,9 +828,12 @@ pub trait Ord: Eq + PartialOrd { Self: Sized, Self: ~const Destruct, { - // HACK(fee1-dead): go back to using `self.min_by(other, Ord::cmp)` - // when trait methods are allowed to be used when a const closure is - // expected. + #[cfg(not(bootstrap))] + { + min_by(self, other, Ord::cmp) + } + + #[cfg(bootstrap)] match self.cmp(&other) { Ordering::Less | Ordering::Equal => self, Ordering::Greater => other, @@ -879,40 +883,6 @@ pub macro Ord($item:item) { /* compiler built-in */ } -#[stable(feature = "rust1", since = "1.0.0")] -#[cfg(bootstrap)] -impl StructuralPartialEq for Ordering {} - -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_const_unstable(feature = "const_cmp", issue = "92391")] -#[cfg(bootstrap)] -impl const PartialEq for Ordering { - #[inline] - fn eq(&self, other: &Self) -> bool { - (*self as i32).eq(&(*other as i32)) - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_const_unstable(feature = "const_cmp", issue = "92391")] -#[cfg(bootstrap)] -impl const Ord for Ordering { - #[inline] - fn cmp(&self, other: &Ordering) -> Ordering { - (*self as i32).cmp(&(*other as i32)) - } -} - -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_const_unstable(feature = "const_cmp", issue = "92391")] -#[cfg(bootstrap)] -impl const PartialOrd for Ordering { - #[inline] - fn partial_cmp(&self, other: &Ordering) -> Option { - (*self as i32).partial_cmp(&(*other as i32)) - } -} - /// Trait for types that form a [partial order](https://en.wikipedia.org/wiki/Partial_order). /// /// The `lt`, `le`, `gt`, and `ge` methods of this trait can be called using @@ -1264,17 +1234,23 @@ where F: ~const Destruct, K: ~const Destruct, { - const fn imp K, K: ~const Ord>( - f: &mut F, - (v1, v2): (&T, &T), - ) -> Ordering - where - T: ~const Destruct, - K: ~const Destruct, - { - f(v1).cmp(&f(v2)) + cfg_if! { + if #[cfg(bootstrap)] { + const fn imp K, K: ~const Ord>( + f: &mut F, + (v1, v2): (&T, &T), + ) -> Ordering + where + T: ~const Destruct, + K: ~const Destruct, + { + f(v1).cmp(&f(v2)) + } + min_by(v1, v2, ConstFnMutClosure::new(&mut f, imp)) + } else { + min_by(v1, v2, const |v1, v2| f(v1).cmp(&f(v2))) + } } - min_by(v1, v2, ConstFnMutClosure::new(&mut f, imp)) } /// Compares and returns the maximum of two values. diff --git a/library/core/src/const_closure.rs b/library/core/src/const_closure.rs index 151c8e6d8..97900a486 100644 --- a/library/core/src/const_closure.rs +++ b/library/core/src/const_closure.rs @@ -1,5 +1,4 @@ use crate::marker::Destruct; -#[cfg(not(bootstrap))] use crate::marker::Tuple; /// Struct representing a closure with mutably borrowed data. @@ -46,33 +45,6 @@ impl<'a, CapturedData: ?Sized, Function> ConstFnMutClosure<&'a mut CapturedData, macro_rules! impl_fn_mut_tuple { ($($var:ident)*) => { - #[cfg(bootstrap)] - #[allow(unused_parens)] - impl<'a, $($var,)* ClosureArguments, Function, ClosureReturnValue> const - FnOnce for ConstFnMutClosure<($(&'a mut $var),*), Function> - where - Function: ~const Fn(($(&mut $var),*), ClosureArguments) -> ClosureReturnValue+ ~const Destruct, - { - type Output = ClosureReturnValue; - - extern "rust-call" fn call_once(mut self, args: ClosureArguments) -> Self::Output { - self.call_mut(args) - } - } - #[cfg(bootstrap)] - #[allow(unused_parens)] - impl<'a, $($var,)* ClosureArguments, Function, ClosureReturnValue> const - FnMut for ConstFnMutClosure<($(&'a mut $var),*), Function> - where - Function: ~const Fn(($(&mut $var),*), ClosureArguments)-> ClosureReturnValue, - { - extern "rust-call" fn call_mut(&mut self, args: ClosureArguments) -> Self::Output { - #[allow(non_snake_case)] - let ($($var),*) = &mut self.data; - (self.func)(($($var),*), args) - } - } - #[cfg(not(bootstrap))] #[allow(unused_parens)] impl<'a, $($var,)* ClosureArguments: Tuple, Function, ClosureReturnValue> const FnOnce for ConstFnMutClosure<($(&'a mut $var),*), Function> @@ -85,7 +57,6 @@ macro_rules! impl_fn_mut_tuple { self.call_mut(args) } } - #[cfg(not(bootstrap))] #[allow(unused_parens)] impl<'a, $($var,)* ClosureArguments: Tuple, Function, ClosureReturnValue> const FnMut for ConstFnMutClosure<($(&'a mut $var),*), Function> diff --git a/library/core/src/convert/num.rs b/library/core/src/convert/num.rs index 9c0d7e9a1..4da7c3234 100644 --- a/library/core/src/convert/num.rs +++ b/library/core/src/convert/num.rs @@ -168,6 +168,26 @@ impl_from! { u32, f64, #[stable(feature = "lossless_float_conv", since = "1.6.0" // Float -> Float impl_from! { f32, f64, #[stable(feature = "lossless_float_conv", since = "1.6.0")] } +// bool -> Float +#[stable(feature = "float_from_bool", since = "1.68.0")] +#[rustc_const_unstable(feature = "const_num_from_num", issue = "87852")] +impl const From for f32 { + /// Converts `bool` to `f32` losslessly. + #[inline] + fn from(small: bool) -> Self { + small as u8 as Self + } +} +#[stable(feature = "float_from_bool", since = "1.68.0")] +#[rustc_const_unstable(feature = "const_num_from_num", issue = "87852")] +impl const From for f64 { + /// Converts `bool` to `f64` losslessly. + #[inline] + fn from(small: bool) -> Self { + small as u8 as Self + } +} + // no possible bounds violation macro_rules! try_from_unbounded { ($source:ty, $($target:ty),*) => {$( diff --git a/library/core/src/ffi/mod.rs b/library/core/src/ffi/mod.rs index ec1eaa99f..76daceecd 100644 --- a/library/core/src/ffi/mod.rs +++ b/library/core/src/ffi/mod.rs @@ -227,7 +227,12 @@ impl fmt::Debug for c_void { /// Basic implementation of a `va_list`. // The name is WIP, using `VaListImpl` for now. #[cfg(any( - all(not(target_arch = "aarch64"), not(target_arch = "powerpc"), not(target_arch = "x86_64")), + all( + not(target_arch = "aarch64"), + not(target_arch = "powerpc"), + not(target_arch = "s390x"), + not(target_arch = "x86_64") + ), all(target_arch = "aarch64", any(target_os = "macos", target_os = "ios")), target_family = "wasm", target_arch = "asmjs", @@ -251,7 +256,12 @@ pub struct VaListImpl<'f> { } #[cfg(any( - all(not(target_arch = "aarch64"), not(target_arch = "powerpc"), not(target_arch = "x86_64")), + all( + not(target_arch = "aarch64"), + not(target_arch = "powerpc"), + not(target_arch = "s390x"), + not(target_arch = "x86_64") + ), all(target_arch = "aarch64", any(target_os = "macos", target_os = "ios")), target_family = "wasm", target_arch = "asmjs", @@ -319,6 +329,25 @@ pub struct VaListImpl<'f> { _marker: PhantomData<&'f mut &'f c_void>, } +/// s390x ABI implementation of a `va_list`. +#[cfg(target_arch = "s390x")] +#[repr(C)] +#[derive(Debug)] +#[unstable( + feature = "c_variadic", + reason = "the `c_variadic` feature has not been properly tested on \ + all supported platforms", + issue = "44930" +)] +#[lang = "va_list"] +pub struct VaListImpl<'f> { + gpr: i64, + fpr: i64, + overflow_arg_area: *mut c_void, + reg_save_area: *mut c_void, + _marker: PhantomData<&'f mut &'f c_void>, +} + /// x86_64 ABI implementation of a `va_list`. #[cfg(all(target_arch = "x86_64", not(target_os = "uefi"), not(windows)))] #[repr(C)] @@ -352,6 +381,7 @@ pub struct VaList<'a, 'f: 'a> { all( not(target_arch = "aarch64"), not(target_arch = "powerpc"), + not(target_arch = "s390x"), not(target_arch = "x86_64") ), all(target_arch = "aarch64", any(target_os = "macos", target_os = "ios")), @@ -363,7 +393,12 @@ pub struct VaList<'a, 'f: 'a> { inner: VaListImpl<'f>, #[cfg(all( - any(target_arch = "aarch64", target_arch = "powerpc", target_arch = "x86_64"), + any( + target_arch = "aarch64", + target_arch = "powerpc", + target_arch = "s390x", + target_arch = "x86_64" + ), any(not(target_arch = "aarch64"), not(any(target_os = "macos", target_os = "ios"))), not(target_family = "wasm"), not(target_arch = "asmjs"), @@ -376,7 +411,12 @@ pub struct VaList<'a, 'f: 'a> { } #[cfg(any( - all(not(target_arch = "aarch64"), not(target_arch = "powerpc"), not(target_arch = "x86_64")), + all( + not(target_arch = "aarch64"), + not(target_arch = "powerpc"), + not(target_arch = "s390x"), + not(target_arch = "x86_64") + ), all(target_arch = "aarch64", any(target_os = "macos", target_os = "ios")), target_family = "wasm", target_arch = "asmjs", @@ -398,7 +438,12 @@ impl<'f> VaListImpl<'f> { } #[cfg(all( - any(target_arch = "aarch64", target_arch = "powerpc", target_arch = "x86_64"), + any( + target_arch = "aarch64", + target_arch = "powerpc", + target_arch = "s390x", + target_arch = "x86_64" + ), any(not(target_arch = "aarch64"), not(any(target_os = "macos", target_os = "ios"))), not(target_family = "wasm"), not(target_arch = "asmjs"), diff --git a/library/core/src/fmt/mod.rs b/library/core/src/fmt/mod.rs index 48b617743..fa5073e33 100644 --- a/library/core/src/fmt/mod.rs +++ b/library/core/src/fmt/mod.rs @@ -174,6 +174,11 @@ pub trait Write { /// This method should generally not be invoked manually, but rather through /// the [`write!`] macro itself. /// + /// # Errors + /// + /// This function will return an instance of [`Error`] on error. Please see + /// [write_str](Write::write_str) for details. + /// /// # Examples /// /// ``` @@ -405,7 +410,7 @@ impl<'a> Arguments<'a> { /// 1. The `pieces` slice must be at least as long as `fmt`. /// 2. Every [`rt::v1::Argument::position`] value within `fmt` must be a /// valid index of `args`. - /// 3. Every [`Count::Param`] within `fmt` must contain a valid index of + /// 3. Every [`rt::v1::Count::Param`] within `fmt` must contain a valid index of /// `args`. #[doc(hidden)] #[inline] @@ -558,7 +563,7 @@ impl Display for Arguments<'_> { /// /// Derived `Debug` formats are not stable, and so may change with future Rust /// versions. Additionally, `Debug` implementations of types provided by the -/// standard library (`libstd`, `libcore`, `liballoc`, etc.) are not stable, and +/// standard library (`std`, `core`, `alloc`, etc.) are not stable, and /// may also change with future Rust versions. /// /// # Examples @@ -2471,8 +2476,8 @@ impl Display for char { #[stable(feature = "rust1", since = "1.0.0")] impl Pointer for *const T { fn fmt(&self, f: &mut Formatter<'_>) -> Result { - // Cast is needed here because `.addr()` requires `T: Sized`. - pointer_fmt_inner((*self as *const ()).addr(), f) + // Cast is needed here because `.expose_addr()` requires `T: Sized`. + pointer_fmt_inner((*self as *const ()).expose_addr(), f) } } diff --git a/library/core/src/future/future.rs b/library/core/src/future/future.rs index f29d3e1e9..8c7111cb3 100644 --- a/library/core/src/future/future.rs +++ b/library/core/src/future/future.rs @@ -37,6 +37,7 @@ use crate::task::{Context, Poll}; pub trait Future { /// The type of value produced on completion. #[stable(feature = "futures_api", since = "1.36.0")] + #[rustc_diagnostic_item = "FutureOutput"] type Output; /// Attempt to resolve the future to a final value, registering diff --git a/library/core/src/future/mod.rs b/library/core/src/future/mod.rs index f2b961d62..c4fb36209 100644 --- a/library/core/src/future/mod.rs +++ b/library/core/src/future/mod.rs @@ -44,7 +44,7 @@ pub use poll_fn::{poll_fn, PollFn}; /// non-Send/Sync as well, and we don't want that. /// /// It also simplifies the HIR lowering of `.await`. -#[cfg_attr(not(bootstrap), lang = "ResumeTy")] +#[lang = "ResumeTy"] #[doc(hidden)] #[unstable(feature = "gen_future", issue = "50547")] #[derive(Debug, Copy, Clone)] @@ -61,7 +61,6 @@ unsafe impl Sync for ResumeTy {} /// This function returns a `GenFuture` underneath, but hides it in `impl Trait` to give /// better error messages (`impl Future` rather than `GenFuture<[closure.....]>`). // This is `const` to avoid extra errors after we recover from `const async fn` -#[cfg_attr(bootstrap, lang = "from_generator")] #[doc(hidden)] #[unstable(feature = "gen_future", issue = "50547")] #[rustc_const_unstable(feature = "gen_future", issue = "50547")] @@ -113,10 +112,14 @@ pub unsafe fn get_context<'a, 'b>(cx: ResumeTy) -> &'a mut Context<'b> { unsafe { &mut *cx.0.as_ptr().cast() } } -#[cfg_attr(not(bootstrap), lang = "identity_future")] +// FIXME(swatinem): This fn is currently needed to work around shortcomings +// in type and lifetime inference. +// See the comment at the bottom of `LoweringContext::make_async_expr` and +// . #[doc(hidden)] #[unstable(feature = "gen_future", issue = "50547")] #[inline] +#[lang = "identity_future"] pub const fn identity_future>(f: Fut) -> Fut { f } diff --git a/library/core/src/hash/mod.rs b/library/core/src/hash/mod.rs index c755afa39..71a0d1825 100644 --- a/library/core/src/hash/mod.rs +++ b/library/core/src/hash/mod.rs @@ -199,7 +199,7 @@ pub trait Hash { /// println!("Hash is {:x}!", hasher.finish()); /// ``` #[stable(feature = "rust1", since = "1.0.0")] - fn hash(&self, state: &mut H); + fn hash(&self, state: &mut H); /// Feeds a slice of this type into the given [`Hasher`]. /// @@ -980,7 +980,7 @@ mod impls { #[rustc_const_unstable(feature = "const_hash", issue = "104061")] impl const Hash for &mut T { #[inline] - fn hash(&self, state: &mut H) { + fn hash(&self, state: &mut H) { (**self).hash(state); } } diff --git a/library/core/src/hint.rs b/library/core/src/hint.rs index e8d724ab1..5a76e8669 100644 --- a/library/core/src/hint.rs +++ b/library/core/src/hint.rs @@ -219,6 +219,75 @@ pub fn spin_loop() { /// backend used. Programs cannot rely on `black_box` for *correctness* in any way. /// /// [`std::convert::identity`]: crate::convert::identity +/// +/// # When is this useful? +/// +/// First and foremost: `black_box` does _not_ guarantee any exact behavior and, in some cases, may +/// do nothing at all. As such, it **must not be relied upon to control critical program behavior.** +/// This _immediately_ precludes any direct use of this function for cryptographic or security +/// purposes. +/// +/// While not suitable in those mission-critical cases, `back_box`'s functionality can generally be +/// relied upon for benchmarking, and should be used there. It will try to ensure that the +/// compiler doesn't optimize away part of the intended test code based on context. For +/// example: +/// +/// ``` +/// fn contains(haystack: &[&str], needle: &str) -> bool { +/// haystack.iter().any(|x| x == &needle) +/// } +/// +/// pub fn benchmark() { +/// let haystack = vec!["abc", "def", "ghi", "jkl", "mno"]; +/// let needle = "ghi"; +/// for _ in 0..10 { +/// contains(&haystack, needle); +/// } +/// } +/// ``` +/// +/// The compiler could theoretically make optimizations like the following: +/// +/// - `needle` and `haystack` are always the same, move the call to `contains` outside the loop and +/// delete the loop +/// - Inline `contains` +/// - `needle` and `haystack` have values known at compile time, `contains` is always true. Remove +/// the call and replace with `true` +/// - Nothing is done with the result of `contains`: delete this function call entirely +/// - `benchmark` now has no purpose: delete this function +/// +/// It is not likely that all of the above happens, but the compiler is definitely able to make some +/// optimizations that could result in a very inaccurate benchmark. This is where `black_box` comes +/// in: +/// +/// ``` +/// use std::hint::black_box; +/// +/// // Same `contains` function +/// fn contains(haystack: &[&str], needle: &str) -> bool { +/// haystack.iter().any(|x| x == &needle) +/// } +/// +/// pub fn benchmark() { +/// let haystack = vec!["abc", "def", "ghi", "jkl", "mno"]; +/// let needle = "ghi"; +/// for _ in 0..10 { +/// // Adjust our benchmark loop contents +/// black_box(contains(black_box(&haystack), black_box(needle))); +/// } +/// } +/// ``` +/// +/// This essentially tells the compiler to block optimizations across any calls to `black_box`. So, +/// it now: +/// +/// - Treats both arguments to `contains` as unpredictable: the body of `contains` can no longer be +/// optimized based on argument values +/// - Treats the call to `contains` and its result as volatile: the body of `benchmark` cannot +/// optimize this away +/// +/// This makes our benchmark much more realistic to how the function would be used in situ, where +/// arguments are usually not known at compile time and the result is used in some way. #[inline] #[stable(feature = "bench_black_box", since = "1.66.0")] #[rustc_const_unstable(feature = "const_black_box", issue = "none")] diff --git a/library/core/src/intrinsics.rs b/library/core/src/intrinsics.rs index 7ed7d767f..a315a28fb 100644 --- a/library/core/src/intrinsics.rs +++ b/library/core/src/intrinsics.rs @@ -55,7 +55,6 @@ #![allow(missing_docs)] use crate::marker::DiscriminantKind; -#[cfg(not(bootstrap))] use crate::marker::Tuple; use crate::mem; @@ -959,13 +958,13 @@ extern "rust-intrinsic" { #[rustc_safe_intrinsic] pub fn assert_zero_valid(); - /// A guard for unsafe functions that cannot ever be executed if `T` has invalid - /// bit patterns: This will statically either panic, or do nothing. + /// A guard for `std::mem::uninitialized`. This will statically either panic, or do nothing. /// /// This intrinsic does not have a stable counterpart. #[rustc_const_unstable(feature = "const_assert_type2", issue = "none")] #[rustc_safe_intrinsic] - pub fn assert_uninit_valid(); + #[cfg(not(bootstrap))] + pub fn assert_mem_uninitialized_valid(); /// Gets a reference to a static `Location` indicating where it was called. /// @@ -2175,66 +2174,6 @@ extern "rust-intrinsic" { /// `unreachable_unchecked` is actually being reached. The bug is in *crate A*, /// which violates the principle that a `const fn` must behave the same at /// compile-time and at run-time. The unsafe code in crate B is fine. - #[cfg(bootstrap)] - #[rustc_const_unstable(feature = "const_eval_select", issue = "none")] - pub fn const_eval_select(arg: ARG, called_in_const: F, called_at_rt: G) -> RET - where - G: FnOnce, - F: FnOnce; - - /// Selects which function to call depending on the context. - /// - /// If this function is evaluated at compile-time, then a call to this - /// intrinsic will be replaced with a call to `called_in_const`. It gets - /// replaced with a call to `called_at_rt` otherwise. - /// - /// # Type Requirements - /// - /// The two functions must be both function items. They cannot be function - /// pointers or closures. The first function must be a `const fn`. - /// - /// `arg` will be the tupled arguments that will be passed to either one of - /// the two functions, therefore, both functions must accept the same type of - /// arguments. Both functions must return RET. - /// - /// # Safety - /// - /// The two functions must behave observably equivalent. Safe code in other - /// crates may assume that calling a `const fn` at compile-time and at run-time - /// produces the same result. A function that produces a different result when - /// evaluated at run-time, or has any other observable side-effects, is - /// *unsound*. - /// - /// Here is an example of how this could cause a problem: - /// ```no_run - /// #![feature(const_eval_select)] - /// #![feature(core_intrinsics)] - /// use std::hint::unreachable_unchecked; - /// use std::intrinsics::const_eval_select; - /// - /// // Crate A - /// pub const fn inconsistent() -> i32 { - /// fn runtime() -> i32 { 1 } - /// const fn compiletime() -> i32 { 2 } - /// - /// unsafe { - // // âš  This code violates the required equivalence of `compiletime` - /// // and `runtime`. - /// const_eval_select((), compiletime, runtime) - /// } - /// } - /// - /// // Crate B - /// const X: i32 = inconsistent(); - /// let x = inconsistent(); - /// if x != X { unsafe { unreachable_unchecked(); }} - /// ``` - /// - /// This code causes Undefined Behavior when being run, since the - /// `unreachable_unchecked` is actually being reached. The bug is in *crate A*, - /// which violates the principle that a `const fn` must behave the same at - /// compile-time and at run-time. The unsafe code in crate B is fine. - #[cfg(not(bootstrap))] #[rustc_const_unstable(feature = "const_eval_select", issue = "none")] pub fn const_eval_select( arg: ARG, @@ -2281,7 +2220,7 @@ macro_rules! assert_unsafe_precondition { fn runtime$(<$($tt)*>)?($($i:$ty),*) { if !$e { // don't unwind to reduce impact on code size - ::core::panicking::panic_str_nounwind( + ::core::panicking::panic_nounwind( concat!("unsafe precondition(s) violated: ", $name) ); } diff --git a/library/core/src/intrinsics/mir.rs b/library/core/src/intrinsics/mir.rs index 8ba1c1228..e3157b669 100644 --- a/library/core/src/intrinsics/mir.rs +++ b/library/core/src/intrinsics/mir.rs @@ -21,15 +21,14 @@ //! #[custom_mir(dialect = "built")] //! pub fn simple(x: i32) -> i32 { //! mir!( -//! let temp1: i32; -//! let temp2: _; +//! let temp2: i32; //! //! { -//! temp1 = x; -//! Goto(exit) +//! let temp1 = x; +//! Goto(my_second_block) //! } //! -//! exit = { +//! my_second_block = { //! temp2 = Move(temp1); //! RET = temp2; //! Return() @@ -38,22 +37,200 @@ //! } //! ``` //! -//! Hopefully most of this is fairly self-explanatory. Expanding on some notable details: +//! The `custom_mir` attribute tells the compiler to treat the function as being custom MIR. This +//! attribute only works on functions - there is no way to insert custom MIR into the middle of +//! another function. The `dialect` and `phase` parameters indicate which [version of MIR][dialect +//! docs] you are inserting here. Generally you'll want to use `#![custom_mir(dialect = "built")]` +//! if you want your MIR to be modified by the full MIR pipeline, or `#![custom_mir(dialect = +//! "runtime", phase = "optimized")] if you don't. //! -//! - The `custom_mir` attribute tells the compiler to treat the function as being custom MIR. This -//! attribute only works on functions - there is no way to insert custom MIR into the middle of -//! another function. -//! - The `dialect` and `phase` parameters indicate which version of MIR you are inserting here. -//! This will normally be the phase that corresponds to the thing you are trying to test. The -//! phase can be omitted for dialects that have just one. -//! - You should define your function signature like you normally would. Externally, this function -//! can be called like any other function. -//! - Type inference works - you don't have to spell out the type of all of your locals. +//! [dialect docs]: +//! https://doc.rust-lang.org/nightly/nightly-rustc/rustc_middle/mir/enum.MirPhase.html //! -//! For now, all statements and terminators are parsed from nested invocations of the special -//! functions provided in this module. We additionally want to (but do not yet) support more -//! "normal" Rust syntax in places where it makes sense. Also, most kinds of instructions are not -//! supported yet. +//! The input to the [`mir!`] macro is: +//! +//! - A possibly empty list of local declarations. Locals can also be declared inline on +//! assignments via `let`. Type inference generally works. Shadowing does not. +//! - A list of basic blocks. The first of these is the start block and is where execution begins. +//! All blocks other than the start block need to be given a name, so that they can be referred +//! to later. +//! - Each block is a list of semicolon terminated statements, followed by a terminator. The +//! syntax for the various statements and terminators is designed to be as similar as possible +//! to the syntax for analogous concepts in native Rust. See below for a list. +//! +//! # Examples +//! +#![cfg_attr(bootstrap, doc = "```rust,compile_fail")] +#![cfg_attr(not(bootstrap), doc = "```rust")] +//! #![feature(core_intrinsics, custom_mir)] +//! +//! extern crate core; +//! use core::intrinsics::mir::*; +//! +//! #[custom_mir(dialect = "built")] +//! pub fn choose_load(a: &i32, b: &i32, c: bool) -> i32 { +//! mir!( +//! { +//! match c { +//! true => t, +//! _ => f, +//! } +//! } +//! +//! t = { +//! let temp = a; +//! Goto(load_and_exit) +//! } +//! +//! f = { +//! temp = b; +//! Goto(load_and_exit) +//! } +//! +//! load_and_exit = { +//! RET = *temp; +//! Return() +//! } +//! ) +//! } +//! +//! #[custom_mir(dialect = "built")] +//! fn unwrap_unchecked(opt: Option) -> T { +//! mir!({ +//! RET = Move(Field(Variant(opt, 1), 0)); +//! Return() +//! }) +//! } +//! +//! #[custom_mir(dialect = "runtime", phase = "optimized")] +//! fn push_and_pop(v: &mut Vec, value: T) { +//! mir!( +//! let unused; +//! let popped; +//! +//! { +//! Call(unused, pop, Vec::push(v, value)) +//! } +//! +//! pop = { +//! Call(popped, drop, Vec::pop(v)) +//! } +//! +//! drop = { +//! Drop(popped, ret) +//! } +//! +//! ret = { +//! Return() +//! } +//! ) +//! } +//! ``` +//! +//! We can also set off compilation failures that happen in sufficiently late stages of the +//! compiler: +//! +//! ```rust,compile_fail +//! #![feature(core_intrinsics, custom_mir)] +//! +//! extern crate core; +//! use core::intrinsics::mir::*; +//! +//! #[custom_mir(dialect = "built")] +//! fn borrow_error(should_init: bool) -> i32 { +//! mir!( +//! let temp: i32; +//! +//! { +//! match should_init { +//! true => init, +//! _ => use_temp, +//! } +//! } +//! +//! init = { +//! temp = 0; +//! Goto(use_temp) +//! } +//! +//! use_temp = { +//! RET = temp; +//! Return() +//! } +//! ) +//! } +//! ``` +//! +//! ```text +//! error[E0381]: used binding is possibly-uninitialized +//! --> test.rs:24:13 +//! | +//! 8 | / mir!( +//! 9 | | let temp: i32; +//! 10 | | +//! 11 | | { +//! ... | +//! 19 | | temp = 0; +//! | | -------- binding initialized here in some conditions +//! ... | +//! 24 | | RET = temp; +//! | | ^^^^^^^^^^ value used here but it is possibly-uninitialized +//! 25 | | Return() +//! 26 | | } +//! 27 | | ) +//! | |_____- binding declared here but left uninitialized +//! +//! error: aborting due to previous error +//! +//! For more information about this error, try `rustc --explain E0381`. +//! ``` +//! +//! # Syntax +//! +//! The lists below are an exhaustive description of how various MIR constructs can be created. +//! Anything missing from the list should be assumed to not be supported, PRs welcome. +//! +//! #### Locals +//! +//! - The `_0` return local can always be accessed via `RET`. +//! - Arguments can be accessed via their regular name. +//! - All other locals need to be declared with `let` somewhere and then can be accessed by name. +//! +//! #### Places +//! - Locals implicit convert to places. +//! - Field accesses, derefs, and indexing work normally. +//! - Fields in variants can be accessed via the [`Variant`] and [`Field`] associated functions, +//! see their documentation for details. +//! +//! #### Operands +//! - Places implicitly convert to `Copy` operands. +//! - `Move` operands can be created via [`Move`]. +//! - Const blocks, literals, named constants, and const params all just work. +//! - [`Static`] and [`StaticMut`] can be used to create `&T` and `*mut T`s to statics. These are +//! constants in MIR and the only way to access statics. +//! +//! #### Statements +//! - Assign statements work via normal Rust assignment. +//! - [`Retag`] statements have an associated function. +//! +//! #### Rvalues +//! +//! - Operands implicitly convert to `Use` rvalues. +//! - `&`, `&mut`, `addr_of!`, and `addr_of_mut!` all work to create their associated rvalue. +//! - [`Discriminant`] has an associated function. +//! +//! #### Terminators +//! +//! Custom MIR does not currently support cleanup blocks or non-trivial unwind paths. As such, there +//! are no resume and abort terminators, and terminators that might unwind do not have any way to +//! indicate the unwind block. +//! +//! - [`Goto`], [`Return`], [`Unreachable`], [`Drop`](Drop()), and [`DropAndReplace`] have associated functions. +//! - `match some_int_operand` becomes a `SwitchInt`. Each arm should be `literal => basic_block` +//! - The exception is the last arm, which must be `_ => basic_block` and corresponds to the +//! otherwise branch. +//! - [`Call`] has an associated function as well. The third argument of this function is a normal +//! function call expresion, for example `my_other_function(a, 5)`. //! #![unstable( @@ -69,21 +246,93 @@ pub struct BasicBlock; macro_rules! define { - ($name:literal, $($sig:tt)*) => { + ($name:literal, $( #[ $meta:meta ] )* fn $($sig:tt)*) => { #[rustc_diagnostic_item = $name] - pub $($sig)* { panic!() } + $( #[ $meta ] )* + pub fn $($sig)* { panic!() } } } define!("mir_return", fn Return() -> BasicBlock); define!("mir_goto", fn Goto(destination: BasicBlock) -> BasicBlock); +define!("mir_unreachable", fn Unreachable() -> BasicBlock); +define!("mir_drop", fn Drop(place: T, goto: BasicBlock)); +define!("mir_drop_and_replace", fn DropAndReplace(place: T, value: T, goto: BasicBlock)); +define!("mir_call", fn Call(place: T, goto: BasicBlock, call: T)); +define!("mir_storage_live", fn StorageLive(local: T)); +define!("mir_storage_dead", fn StorageDead(local: T)); define!("mir_retag", fn Retag(place: T)); -define!("mir_retag_raw", fn RetagRaw(place: T)); define!("mir_move", fn Move(place: T) -> T); define!("mir_static", fn Static(s: T) -> &'static T); define!("mir_static_mut", fn StaticMut(s: T) -> *mut T); +define!( + "mir_discriminant", + /// Gets the discriminant of a place. + fn Discriminant(place: T) -> ::Discriminant +); +define!("mir_set_discriminant", fn SetDiscriminant(place: T, index: u32)); +define!( + "mir_field", + /// Access the field with the given index of some place. + /// + /// This only makes sense to use in conjunction with [`Variant`]. If the type you are looking to + /// access the field of does not have variants, you can use normal field projection syntax. + /// + /// There is no proper way to do a place projection to a variant in Rust, and so these two + /// functions are a workaround. You can access a field of a variant via `Field(Variant(place, + /// var_idx), field_idx)`, where `var_idx` and `field_idx` are appropriate literals. Some + /// caveats: + /// + /// - The return type of `Variant` is always `()`. Don't worry about that, the correct MIR will + /// still be generated. + /// - In some situations, the return type of `Field` cannot be inferred. You may need to + /// annotate it on the function in these cases. + /// - Since `Field` is a function call which is not a place expression, using this on the left + /// hand side of an expression is rejected by the compiler. [`place!`] is a macro provided to + /// work around that issue. Wrap the left hand side of an assignment in the macro to convince + /// the compiler that it's ok. + /// + /// # Examples + /// + #[cfg_attr(bootstrap, doc = "```rust,compile_fail")] + #[cfg_attr(not(bootstrap), doc = "```rust")] + /// #![feature(custom_mir, core_intrinsics)] + /// + /// extern crate core; + /// use core::intrinsics::mir::*; + /// + /// #[custom_mir(dialect = "built")] + /// fn unwrap_deref(opt: Option<&i32>) -> i32 { + /// mir!({ + /// RET = *Field::<&i32>(Variant(opt, 1), 0); + /// Return() + /// }) + /// } + /// + /// #[custom_mir(dialect = "built")] + /// fn set(opt: &mut Option) { + /// mir!({ + /// place!(Field(Variant(*opt, 1), 0)) = 5; + /// Return() + /// }) + /// } + /// ``` + fn Field(place: (), field: u32) -> F +); +define!( + "mir_variant", + /// Adds a variant projection with the given index to the place. + /// + /// See [`Field`] for documentation. + fn Variant(place: T, index: u32) -> () +); +define!( + "mir_make_place", + #[doc(hidden)] + fn __internal_make_place(place: T) -> *mut T +); -/// Convenience macro for generating custom MIR. +/// Macro for generating custom MIR. /// /// See the module documentation for syntax details. This macro is not magic - it only transforms /// your MIR into something that is easier to parse in the compiler. @@ -139,6 +388,13 @@ pub macro mir { }} } +/// Helper macro that allows you to treat a value expression like a place expression. +/// +/// See the documentation on [`Variant`] for why this is necessary and how to use it. +pub macro place($e:expr) { + (*::core::intrinsics::mir::__internal_make_place($e)) +} + /// Helper macro that extracts the `let` declarations out of a bunch of statements. /// /// This macro is written using the "statement muncher" strategy. Each invocation parses the first diff --git a/library/core/src/iter/range.rs b/library/core/src/iter/range.rs index ac7b389b1..b5739f2f3 100644 --- a/library/core/src/iter/range.rs +++ b/library/core/src/iter/range.rs @@ -756,7 +756,7 @@ impl Iterator for ops::Range { where Self: TrustedRandomAccessNoCoerce, { - // SAFETY: The TrustedRandomAccess contract requires that callers only pass an index + // SAFETY: The TrustedRandomAccess contract requires that callers only pass an index // that is in bounds. // Additionally Self: TrustedRandomAccess is only implemented for Copy types // which means even repeated reads of the same index would be safe. diff --git a/library/core/src/iter/sources/empty.rs b/library/core/src/iter/sources/empty.rs index 98734c527..617dfd123 100644 --- a/library/core/src/iter/sources/empty.rs +++ b/library/core/src/iter/sources/empty.rs @@ -22,17 +22,12 @@ pub const fn empty() -> Empty { Empty(marker::PhantomData) } -// Newtype for use in `PhantomData` to avoid -// > error: const-stable function cannot use `#[feature(const_fn_fn_ptr_basics)]` -// in `const fn empty()` above. -struct FnReturning(fn() -> T); - /// An iterator that yields nothing. /// /// This `struct` is created by the [`empty()`] function. See its documentation for more. #[must_use = "iterators are lazy and do nothing unless consumed"] #[stable(feature = "iter_empty", since = "1.2.0")] -pub struct Empty(marker::PhantomData>); +pub struct Empty(marker::PhantomData T>); #[stable(feature = "core_impl_debug", since = "1.9.0")] impl fmt::Debug for Empty { diff --git a/library/core/src/iter/sources/from_generator.rs b/library/core/src/iter/sources/from_generator.rs index 8e7cbd34a..4cbe731b2 100644 --- a/library/core/src/iter/sources/from_generator.rs +++ b/library/core/src/iter/sources/from_generator.rs @@ -1,3 +1,4 @@ +use crate::fmt; use crate::ops::{Generator, GeneratorState}; use crate::pin::Pin; @@ -23,14 +24,21 @@ use crate::pin::Pin; /// ``` #[inline] #[unstable(feature = "iter_from_generator", issue = "43122", reason = "generators are unstable")] -pub fn from_generator + Unpin>( - generator: G, -) -> impl Iterator { +pub fn from_generator + Unpin>(generator: G) -> FromGenerator { FromGenerator(generator) } -struct FromGenerator(G); +/// An iterator over the values yielded by an underlying generator. +/// +/// This `struct` is created by the [`iter::from_generator()`] function. See its documentation for +/// more. +/// +/// [`iter::from_generator()`]: from_generator +#[unstable(feature = "iter_from_generator", issue = "43122", reason = "generators are unstable")] +#[derive(Clone)] +pub struct FromGenerator(G); +#[unstable(feature = "iter_from_generator", issue = "43122", reason = "generators are unstable")] impl + Unpin> Iterator for FromGenerator { type Item = G::Yield; @@ -41,3 +49,10 @@ impl + Unpin> Iterator for FromGenerator { } } } + +#[unstable(feature = "iter_from_generator", issue = "43122", reason = "generators are unstable")] +impl fmt::Debug for FromGenerator { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("FromGenerator").finish() + } +} diff --git a/library/core/src/iter/sources/once_with.rs b/library/core/src/iter/sources/once_with.rs index d79f85c25..9309a06c8 100644 --- a/library/core/src/iter/sources/once_with.rs +++ b/library/core/src/iter/sources/once_with.rs @@ -1,3 +1,4 @@ +use crate::fmt; use crate::iter::{FusedIterator, TrustedLen}; /// Creates an iterator that lazily generates a value exactly once by invoking @@ -66,12 +67,23 @@ pub fn once_with A>(gen: F) -> OnceWith { /// /// This `struct` is created by the [`once_with()`] function. /// See its documentation for more. -#[derive(Clone, Debug)] +#[derive(Clone)] #[stable(feature = "iter_once_with", since = "1.43.0")] pub struct OnceWith { gen: Option, } +#[stable(feature = "iter_once_with_debug", since = "1.68.0")] +impl fmt::Debug for OnceWith { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + if self.gen.is_some() { + f.write_str("OnceWith(Some(_))") + } else { + f.write_str("OnceWith(None)") + } + } +} + #[stable(feature = "iter_once_with", since = "1.43.0")] impl A> Iterator for OnceWith { type Item = A; diff --git a/library/core/src/iter/sources/repeat_n.rs b/library/core/src/iter/sources/repeat_n.rs index fd8d25ce1..dc61d6065 100644 --- a/library/core/src/iter/sources/repeat_n.rs +++ b/library/core/src/iter/sources/repeat_n.rs @@ -126,7 +126,7 @@ impl Iterator for RepeatN { // zero so it won't be dropped later, and thus it's okay to take it here. unsafe { ManuallyDrop::take(&mut self.element) } } else { - A::clone(&mut self.element) + A::clone(&self.element) }) } diff --git a/library/core/src/iter/sources/repeat_with.rs b/library/core/src/iter/sources/repeat_with.rs index ab2d0472b..3f34105a3 100644 --- a/library/core/src/iter/sources/repeat_with.rs +++ b/library/core/src/iter/sources/repeat_with.rs @@ -1,3 +1,4 @@ +use crate::fmt; use crate::iter::{FusedIterator, TrustedLen}; use crate::ops::Try; @@ -71,12 +72,19 @@ pub fn repeat_with A>(repeater: F) -> RepeatWith { /// /// This `struct` is created by the [`repeat_with()`] function. /// See its documentation for more. -#[derive(Copy, Clone, Debug)] +#[derive(Copy, Clone)] #[stable(feature = "iterator_repeat_with", since = "1.28.0")] pub struct RepeatWith { repeater: F, } +#[stable(feature = "iterator_repeat_with_debug", since = "1.68.0")] +impl fmt::Debug for RepeatWith { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("RepeatWith").finish_non_exhaustive() + } +} + #[stable(feature = "iterator_repeat_with", since = "1.28.0")] impl A> Iterator for RepeatWith { type Item = A; diff --git a/library/core/src/iter/traits/accum.rs b/library/core/src/iter/traits/accum.rs index 84d83ee39..e31669b39 100644 --- a/library/core/src/iter/traits/accum.rs +++ b/library/core/src/iter/traits/accum.rs @@ -10,6 +10,10 @@ use crate::num::Wrapping; /// [`sum()`]: Iterator::sum /// [`FromIterator`]: iter::FromIterator #[stable(feature = "iter_arith_traits", since = "1.12.0")] +#[rustc_on_unimplemented( + message = "a value of type `{Self}` cannot be made by summing an iterator over elements of type `{A}`", + label = "value of type `{Self}` cannot be made by summing a `std::iter::Iterator`" +)] pub trait Sum: Sized { /// Method which takes an iterator and generates `Self` from the elements by /// "summing up" the items. @@ -27,6 +31,10 @@ pub trait Sum: Sized { /// [`product()`]: Iterator::product /// [`FromIterator`]: iter::FromIterator #[stable(feature = "iter_arith_traits", since = "1.12.0")] +#[rustc_on_unimplemented( + message = "a value of type `{Self}` cannot be made by multiplying all elements of type `{A}` from an iterator", + label = "value of type `{Self}` cannot be made by multiplying all elements from a `std::iter::Iterator`" +)] pub trait Product: Sized { /// Method which takes an iterator and generates `Self` from the elements by /// multiplying the items. diff --git a/library/core/src/iter/traits/iterator.rs b/library/core/src/iter/traits/iterator.rs index 83c7e8977..a4a665d48 100644 --- a/library/core/src/iter/traits/iterator.rs +++ b/library/core/src/iter/traits/iterator.rs @@ -58,6 +58,11 @@ fn _assert_is_object_safe(_: &dyn Iterator) {} note = "if you want to iterate between `start` until a value `end`, use the exclusive range \ syntax `start..end` or the inclusive range syntax `start..=end`" ), + on( + _Self = "{float}", + note = "if you want to iterate between `start` until a value `end`, use the exclusive range \ + syntax `start..end` or the inclusive range syntax `start..=end`" + ), label = "`{Self}` is not an iterator", message = "`{Self}` is not an iterator" )] @@ -66,6 +71,7 @@ fn _assert_is_object_safe(_: &dyn Iterator) {} #[must_use = "iterators are lazy and do nothing unless consumed"] pub trait Iterator { /// The type of the elements being iterated over. + #[rustc_diagnostic_item = "IteratorItem"] #[stable(feature = "rust1", since = "1.0.0")] type Item; @@ -803,7 +809,7 @@ pub trait Iterator { /// (0..5).map(|x| x * 2 + 1) /// .for_each(move |x| tx.send(x).unwrap()); /// - /// let v: Vec<_> = rx.iter().collect(); + /// let v: Vec<_> = rx.iter().collect(); /// assert_eq!(v, vec![1, 3, 5, 7, 9]); /// ``` /// @@ -1380,8 +1386,8 @@ pub trait Iterator { Take::new(self, n) } - /// An iterator adapter similar to [`fold`] that holds internal state and - /// produces a new iterator. + /// An iterator adapter which, like [`fold`], holds internal state, but + /// unlike [`fold`], produces a new iterator. /// /// [`fold`]: Iterator::fold /// @@ -1393,20 +1399,25 @@ pub trait Iterator { /// /// On iteration, the closure will be applied to each element of the /// iterator and the return value from the closure, an [`Option`], is - /// yielded by the iterator. + /// returned by the `next` method. Thus the closure can return + /// `Some(value)` to yield `value`, or `None` to end the iteration. /// /// # Examples /// /// Basic usage: /// /// ``` - /// let a = [1, 2, 3]; + /// let a = [1, 2, 3, 4]; /// /// let mut iter = a.iter().scan(1, |state, &x| { - /// // each iteration, we'll multiply the state by the element + /// // each iteration, we'll multiply the state by the element ... /// *state = *state * x; /// - /// // then, we'll yield the negation of the state + /// // ... and terminate if the state exceeds 6 + /// if *state > 6 { + /// return None; + /// } + /// // ... else yield the negation of the state /// Some(-*state) /// }); /// @@ -1508,6 +1519,18 @@ pub trait Iterator { /// assert_eq!(merged, "alphabetagamma"); /// ``` /// + /// Flattening works on any `IntoIterator` type, including `Option` and `Result`: + /// + /// ``` + /// let options = vec![Some(123), Some(321), None, Some(231)]; + /// let flattened_options: Vec<_> = options.into_iter().flatten().collect(); + /// assert_eq!(flattened_options, vec![123, 321, 231]); + /// + /// let results = vec![Ok(123), Ok(321), Err(456), Ok(231)]; + /// let flattened_results: Vec<_> = results.into_iter().flatten().collect(); + /// assert_eq!(flattened_results, vec![123, 321, 231]); + /// ``` + /// /// Flattening only removes one level of nesting at a time: /// /// ``` @@ -1829,6 +1852,7 @@ pub trait Iterator { #[inline] #[stable(feature = "rust1", since = "1.0.0")] #[must_use = "if you really need to exhaust the iterator, consider `.for_each(drop)` instead"] + #[cfg_attr(not(test), rustc_diagnostic_item = "iterator_collect_fn")] fn collect>(self) -> B where Self: Sized, @@ -2652,7 +2676,10 @@ pub trait Iterator { /// argument is a double reference. You can see this effect in the /// examples below, with `&&x`. /// + /// If you need the index of the element, see [`position()`]. + /// /// [`Some(element)`]: Some + /// [`position()`]: Iterator::position /// /// # Examples /// @@ -2733,7 +2760,7 @@ pub trait Iterator { /// the first true result or the first error. /// /// The return type of this method depends on the return type of the closure. - /// If you return `Result` from the closure, you'll get a `Result; E>`. + /// If you return `Result` from the closure, you'll get a `Result, E>`. /// If you return `Option` from the closure, you'll get an `Option>`. /// /// # Examples diff --git a/library/core/src/lib.rs b/library/core/src/lib.rs index 1823fd300..8790649ab 100644 --- a/library/core/src/lib.rs +++ b/library/core/src/lib.rs @@ -38,18 +38,18 @@ //! which do not trigger a panic can be assured that this function is never //! called. The `lang` attribute is called `eh_personality`. -// Since libcore defines many fundamental lang items, all tests live in a -// separate crate, libcoretest, to avoid bizarre issues. +// Since core defines many fundamental lang items, all tests live in a +// separate crate, libcoretest (library/core/tests), to avoid bizarre issues. // // Here we explicitly #[cfg]-out this whole crate when testing. If we don't do // this, both the generated test artifact and the linked libtest (which -// transitively includes libcore) will both define the same set of lang items, +// transitively includes core) will both define the same set of lang items, // and this will cause the E0152 "found duplicate lang item" error. See // discussion in #50466 for details. // // This cfg won't affect doc tests. #![cfg(not(test))] -// To run libcore tests without x.py without ending up with two copies of libcore, Miri needs to be +// To run core tests without x.py without ending up with two copies of core, Miri needs to be // able to "empty" this crate. See . // rustc itself never sets the feature, so this line has no affect there. #![cfg(any(not(feature = "miri-test-libstd"), test, doctest))] @@ -158,6 +158,7 @@ #![feature(const_unsafecell_get_mut)] #![feature(const_waker)] #![feature(core_panic)] +#![feature(char_indices_offset)] #![feature(duration_consts_float)] #![feature(maybe_uninit_uninit_array)] #![feature(ptr_alignment_type)] @@ -166,6 +167,8 @@ #![feature(slice_ptr_get)] #![feature(slice_split_at_unchecked)] #![feature(str_internals)] +#![feature(str_split_remainder)] +#![feature(str_split_inclusive_remainder)] #![feature(strict_provenance)] #![feature(utf16_extra)] #![feature(utf16_extra_const)] @@ -188,13 +191,14 @@ #![feature(cfg_sanitize)] #![feature(cfg_target_has_atomic)] #![feature(cfg_target_has_atomic_equal_alignment)] +#![cfg_attr(not(bootstrap), feature(const_closures))] #![feature(const_fn_floating_point_arithmetic)] #![feature(const_mut_refs)] #![feature(const_precise_live_drops)] #![feature(const_refs_to_cell)] #![feature(decl_macro)] #![feature(deprecated_suggestion)] -#![cfg_attr(not(bootstrap), feature(derive_const))] +#![feature(derive_const)] #![feature(doc_cfg)] #![feature(doc_notable_trait)] #![feature(rustdoc_internals)] @@ -236,7 +240,6 @@ #![feature(arm_target_feature)] #![feature(avx512_target_feature)] #![feature(cmpxchg16b_target_feature)] -#![feature(f16c_target_feature)] #![feature(hexagon_target_feature)] #![feature(mips_target_feature)] #![feature(powerpc_target_feature)] @@ -245,6 +248,7 @@ #![feature(sse4a_target_feature)] #![feature(tbm_target_feature)] #![feature(wasm_target_feature)] +#![cfg_attr(bootstrap, feature(f16c_target_feature))] // allow using `core::` in intra-doc links #[allow(unused_extern_crates)] @@ -309,7 +313,7 @@ pub mod f64; #[macro_use] pub mod num; -/* The libcore prelude, not as all-encompassing as the libstd prelude */ +/* The core prelude, not as all-encompassing as the std prelude */ pub mod prelude; @@ -376,12 +380,12 @@ mod const_closure; #[stable(feature = "core_primitive", since = "1.43.0")] pub mod primitive; -// Pull in the `core_arch` crate directly into libcore. The contents of +// Pull in the `core_arch` crate directly into core. The contents of // `core_arch` are in a different repository: rust-lang/stdarch. // -// `core_arch` depends on libcore, but the contents of this module are +// `core_arch` depends on core, but the contents of this module are // set up in such a way that directly pulling it here works such that the -// crate uses the this crate as its libcore. +// crate uses the this crate as its core. #[path = "../../stdarch/crates/core_arch/src/mod.rs"] #[allow( missing_docs, @@ -400,12 +404,12 @@ mod core_arch; #[stable(feature = "simd_arch", since = "1.27.0")] pub mod arch; -// Pull in the `core_simd` crate directly into libcore. The contents of +// Pull in the `core_simd` crate directly into core. The contents of // `core_simd` are in a different repository: rust-lang/portable-simd. // -// `core_simd` depends on libcore, but the contents of this module are +// `core_simd` depends on core, but the contents of this module are // set up in such a way that directly pulling it here works such that the -// crate uses this crate as its libcore. +// crate uses this crate as its core. #[path = "../../portable-simd/crates/core_simd/src/mod.rs"] #[allow(missing_debug_implementations, dead_code, unsafe_op_in_unsafe_fn, unused_unsafe)] #[allow(rustdoc::bare_urls)] diff --git a/library/core/src/macros/mod.rs b/library/core/src/macros/mod.rs index f29cd357d..3b026bc0e 100644 --- a/library/core/src/macros/mod.rs +++ b/library/core/src/macros/mod.rs @@ -1315,22 +1315,41 @@ pub(crate) mod builtin { /// Parses a file as an expression or an item according to the context. /// - /// The file is located relative to the current file (similarly to how - /// modules are found). The provided path is interpreted in a platform-specific - /// way at compile time. So, for instance, an invocation with a Windows path - /// containing backslashes `\` would not compile correctly on Unix. + /// **Warning**: For multi-file Rust projects, the `include!` macro is probably not what you + /// are looking for. Usually, multi-file Rust projects use + /// [modules](https://doc.rust-lang.org/reference/items/modules.html). Multi-file projects and + /// modules are explained in the Rust-by-Example book + /// [here](https://doc.rust-lang.org/rust-by-example/mod/split.html) and the module system is + /// explained in the Rust Book + /// [here](https://doc.rust-lang.org/book/ch07-02-defining-modules-to-control-scope-and-privacy.html). + /// + /// The included file is placed in the surrounding code + /// [unhygienically](https://doc.rust-lang.org/reference/macros-by-example.html#hygiene). If + /// the included file is parsed as an expression and variables or functions share names across + /// both files, it could result in variables or functions being different from what the + /// included file expected. + /// + /// The included file is located relative to the current file (similarly to how modules are + /// found). The provided path is interpreted in a platform-specific way at compile time. So, + /// for instance, an invocation with a Windows path containing backslashes `\` would not + /// compile correctly on Unix. /// - /// Using this macro is often a bad idea, because if the file is - /// parsed as an expression, it is going to be placed in the - /// surrounding code unhygienically. This could result in variables - /// or functions being different from what the file expected if - /// there are variables or functions that have the same name in - /// the current file. + /// # Uses + /// + /// The `include!` macro is primarily used for two purposes. It is used to include + /// documentation that is written in a separate file and it is used to include [build artifacts + /// usually as a result from the `build.rs` + /// script](https://doc.rust-lang.org/cargo/reference/build-scripts.html#outputs-of-the-build-script). + /// + /// When using the `include` macro to include stretches of documentation, remember that the + /// included file still needs to be a valid rust syntax. It is also possible to + /// use the [`include_str`] macro as `#![doc = include_str!("...")]` (at the module level) or + /// `#[doc = include_str!("...")]` (at the item level) to include documentation from a plain + /// text or markdown file. /// /// # Examples /// - /// Assume there are two files in the same directory with the following - /// contents: + /// Assume there are two files in the same directory with the following contents: /// /// File 'monkeys.in': /// @@ -1461,7 +1480,6 @@ pub(crate) mod builtin { /// [the reference]: ../../../reference/attributes/derive.html #[unstable(feature = "derive_const", issue = "none")] #[rustc_builtin_macro] - #[cfg(not(bootstrap))] pub macro derive_const($item:item) { /* compiler built-in */ } @@ -1516,7 +1534,6 @@ pub(crate) mod builtin { /// Attribute macro applied to a function to register it as a handler for allocation failure. /// /// See also [`std::alloc::handle_alloc_error`](../../../std/alloc/fn.handle_alloc_error.html). - #[cfg(not(bootstrap))] #[unstable(feature = "alloc_error_handler", issue = "51540")] #[allow_internal_unstable(rustc_attrs)] #[rustc_builtin_macro] @@ -1553,7 +1570,6 @@ pub(crate) mod builtin { issue = "23416", reason = "placeholder syntax for type ascription" )] - #[cfg(not(bootstrap))] pub macro type_ascribe($expr:expr, $ty:ty) { /* compiler built-in */ } diff --git a/library/core/src/marker.rs b/library/core/src/marker.rs index 42c342801..1326fc9ab 100644 --- a/library/core/src/marker.rs +++ b/library/core/src/marker.rs @@ -96,7 +96,7 @@ unsafe impl Send for &T {} )] #[fundamental] // for Default, for example, which requires that `[T]: !Default` be evaluatable #[rustc_specialization_trait] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait Sized { // Empty. } @@ -126,9 +126,9 @@ pub trait Sized { /// [`Rc`]: ../../std/rc/struct.Rc.html /// [RFC982]: https://github.com/rust-lang/rfcs/blob/master/text/0982-dst-coercion.md /// [nomicon-coerce]: ../../nomicon/coercions.html -#[unstable(feature = "unsize", issue = "27732")] +#[unstable(feature = "unsize", issue = "18598")] #[lang = "unsize"] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait Unsize { // Empty. } @@ -623,6 +623,12 @@ impl !Sync for *mut T {} /// (ideally) or `PhantomData<*const T>` (if no lifetime applies), so /// as not to indicate ownership. /// +/// ## Layout +/// +/// For all `T`, the following are guaranteed: +/// * `size_of::>() == 0` +/// * `align_of::>() == 1` +/// /// [drop check]: ../../nomicon/dropck.html #[lang = "phantom_data"] #[stable(feature = "rust1", since = "1.0.0")] @@ -695,7 +701,7 @@ impl StructuralEq for PhantomData {} reason = "this trait is unlikely to ever be stabilized, use `mem::discriminant` instead" )] #[lang = "discriminant_kind"] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait DiscriminantKind { /// The type of the discriminant, which must satisfy the trait /// bounds required by `mem::Discriminant`. @@ -796,7 +802,7 @@ impl Unpin for *mut T {} #[lang = "destruct"] #[rustc_on_unimplemented(message = "can't drop `{Self}`", append_const_msg)] #[const_trait] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait Destruct {} /// A marker for tuple types. @@ -806,12 +812,12 @@ pub trait Destruct {} #[unstable(feature = "tuple_trait", issue = "none")] #[lang = "tuple_trait"] #[rustc_on_unimplemented(message = "`{Self}` is not a tuple")] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait Tuple {} /// A marker for things #[unstable(feature = "pointer_sized_trait", issue = "none")] -#[cfg_attr(not(bootstrap), lang = "pointer_sized")] +#[lang = "pointer_sized"] #[rustc_on_unimplemented( message = "`{Self}` needs to be a pointer-sized type", label = "`{Self}` needs to be a pointer-sized type" diff --git a/library/core/src/mem/mod.rs b/library/core/src/mem/mod.rs index 383bdc7b6..5e01ccc07 100644 --- a/library/core/src/mem/mod.rs +++ b/library/core/src/mem/mod.rs @@ -682,7 +682,8 @@ pub unsafe fn zeroed() -> T { pub unsafe fn uninitialized() -> T { // SAFETY: the caller must guarantee that an uninitialized value is valid for `T`. unsafe { - intrinsics::assert_uninit_valid::(); + #[cfg(not(bootstrap))] // If the compiler hits this itself then it deserves the UB. + intrinsics::assert_mem_uninitialized_valid::(); let mut val = MaybeUninit::::uninit(); // Fill memory with 0x01, as an imperfect mitigation for old code that uses this function on diff --git a/library/core/src/num/dec2flt/fpu.rs b/library/core/src/num/dec2flt/fpu.rs index ec5fa45fd..3806977f7 100644 --- a/library/core/src/num/dec2flt/fpu.rs +++ b/library/core/src/num/dec2flt/fpu.rs @@ -26,7 +26,7 @@ mod fpu_precision { /// Developer's Manual (Volume 1). /// /// The only field which is relevant for the following code is PC, Precision Control. This - /// field determines the precision of the operations performed by the FPU. It can be set to: + /// field determines the precision of the operations performed by the FPU. It can be set to: /// - 0b00, single precision i.e., 32-bits /// - 0b10, double precision i.e., 64-bits /// - 0b11, double extended precision i.e., 80-bits (default state) diff --git a/library/core/src/num/f32.rs b/library/core/src/num/f32.rs index 2c6a0ba64..1308b0770 100644 --- a/library/core/src/num/f32.rs +++ b/library/core/src/num/f32.rs @@ -428,7 +428,7 @@ impl f32 { self != self } - // FIXME(#50145): `abs` is publicly unavailable in libcore due to + // FIXME(#50145): `abs` is publicly unavailable in core due to // concerns about portability, so this implementation is for // private use internally. #[inline] diff --git a/library/core/src/num/f64.rs b/library/core/src/num/f64.rs index fd3c18ce2..2a22c4302 100644 --- a/library/core/src/num/f64.rs +++ b/library/core/src/num/f64.rs @@ -427,7 +427,7 @@ impl f64 { self != self } - // FIXME(#50145): `abs` is publicly unavailable in libcore due to + // FIXME(#50145): `abs` is publicly unavailable in core due to // concerns about portability, so this implementation is for // private use internally. #[inline] diff --git a/library/core/src/num/int_macros.rs b/library/core/src/num/int_macros.rs index 57096f439..2cae98b8e 100644 --- a/library/core/src/num/int_macros.rs +++ b/library/core/src/num/int_macros.rs @@ -1514,37 +1514,50 @@ macro_rules! int_impl { (a as Self, b) } - /// Calculates `self + rhs + carry` without the ability to overflow. + /// Calculates `self` + `rhs` + `carry` and checks for overflow. /// - /// Performs "signed ternary addition" which takes in an extra bit to add, and may return an - /// additional bit of overflow. This signed function is used only on the highest-ordered data, - /// for which the signed overflow result indicates whether the big integer overflowed or not. + /// Performs "ternary addition" of two integer operands and a carry-in + /// bit, and returns a tuple of the sum along with a boolean indicating + /// whether an arithmetic overflow would occur. On overflow, the wrapped + /// value is returned. /// - /// # Examples + /// This allows chaining together multiple additions to create a wider + /// addition, and can be useful for bignum addition. This method should + /// only be used for the most significant word; for the less significant + /// words the unsigned method + #[doc = concat!("[`", stringify!($UnsignedT), "::carrying_add`]")] + /// should be used. /// - /// Basic usage: + /// The output boolean returned by this method is *not* a carry flag, + /// and should *not* be added to a more significant word. /// - /// ``` - /// #![feature(bigint_helper_methods)] - #[doc = concat!("assert_eq!(5", stringify!($SelfT), ".carrying_add(2, false), (7, false));")] - #[doc = concat!("assert_eq!(5", stringify!($SelfT), ".carrying_add(2, true), (8, false));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.carrying_add(1, false), (", stringify!($SelfT), "::MIN, true));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.carrying_add(0, true), (", stringify!($SelfT), "::MIN, true));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.carrying_add(1, true), (", stringify!($SelfT), "::MIN + 1, true));")] - #[doc = concat!("assert_eq!(", - stringify!($SelfT), "::MAX.carrying_add(", stringify!($SelfT), "::MAX, true), ", - "(-1, true));" - )] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MIN.carrying_add(-1, true), (", stringify!($SelfT), "::MIN, false));")] - #[doc = concat!("assert_eq!(0", stringify!($SelfT), ".carrying_add(", stringify!($SelfT), "::MAX, true), (", stringify!($SelfT), "::MIN, true));")] - /// ``` + /// If the input carry is false, this method is equivalent to + /// [`overflowing_add`](Self::overflowing_add). /// - /// If `carry` is false, this method is equivalent to [`overflowing_add`](Self::overflowing_add): + /// # Examples /// /// ``` /// #![feature(bigint_helper_methods)] - #[doc = concat!("assert_eq!(5_", stringify!($SelfT), ".carrying_add(2, false), 5_", stringify!($SelfT), ".overflowing_add(2));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.carrying_add(1, false), ", stringify!($SelfT), "::MAX.overflowing_add(1));")] + /// // Only the most significant word is signed. + /// // + #[doc = concat!("// 10 MAX (a = 10 × 2^", stringify!($BITS), " + 2^", stringify!($BITS), " - 1)")] + #[doc = concat!("// + -5 9 (b = -5 × 2^", stringify!($BITS), " + 9)")] + /// // --------- + #[doc = concat!("// 6 8 (sum = 6 × 2^", stringify!($BITS), " + 8)")] + /// + #[doc = concat!("let (a1, a0): (", stringify!($SelfT), ", ", stringify!($UnsignedT), ") = (10, ", stringify!($UnsignedT), "::MAX);")] + #[doc = concat!("let (b1, b0): (", stringify!($SelfT), ", ", stringify!($UnsignedT), ") = (-5, 9);")] + /// let carry0 = false; + /// + #[doc = concat!("// ", stringify!($UnsignedT), "::carrying_add for the less significant words")] + /// let (sum0, carry1) = a0.carrying_add(b0, carry0); + /// assert_eq!(carry1, true); + /// + #[doc = concat!("// ", stringify!($SelfT), "::carrying_add for the most significant word")] + /// let (sum1, overflow) = a1.carrying_add(b1, carry1); + /// assert_eq!(overflow, false); + /// + /// assert_eq!((sum1, sum0), (6, 8)); /// ``` #[unstable(feature = "bigint_helper_methods", issue = "85532")] #[rustc_const_unstable(feature = "const_bigint_helper_methods", issue = "85532")] @@ -1608,25 +1621,51 @@ macro_rules! int_impl { (a as Self, b) } - /// Calculates `self - rhs - borrow` without the ability to overflow. + /// Calculates `self` − `rhs` − `borrow` and checks for + /// overflow. /// - /// Performs "signed ternary subtraction" which takes in an extra bit to subtract, and may return an - /// additional bit of overflow. This signed function is used only on the highest-ordered data, - /// for which the signed overflow result indicates whether the big integer overflowed or not. + /// Performs "ternary subtraction" by subtracting both an integer + /// operand and a borrow-in bit from `self`, and returns a tuple of the + /// difference along with a boolean indicating whether an arithmetic + /// overflow would occur. On overflow, the wrapped value is returned. /// - /// # Examples + /// This allows chaining together multiple subtractions to create a + /// wider subtraction, and can be useful for bignum subtraction. This + /// method should only be used for the most significant word; for the + /// less significant words the unsigned method + #[doc = concat!("[`", stringify!($UnsignedT), "::borrowing_sub`]")] + /// should be used. /// - /// Basic usage: + /// The output boolean returned by this method is *not* a borrow flag, + /// and should *not* be subtracted from a more significant word. + /// + /// If the input borrow is false, this method is equivalent to + /// [`overflowing_sub`](Self::overflowing_sub). + /// + /// # Examples /// /// ``` /// #![feature(bigint_helper_methods)] - #[doc = concat!("assert_eq!(5", stringify!($SelfT), ".borrowing_sub(2, false), (3, false));")] - #[doc = concat!("assert_eq!(5", stringify!($SelfT), ".borrowing_sub(2, true), (2, false));")] - #[doc = concat!("assert_eq!(0", stringify!($SelfT), ".borrowing_sub(1, false), (-1, false));")] - #[doc = concat!("assert_eq!(0", stringify!($SelfT), ".borrowing_sub(1, true), (-2, false));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MIN.borrowing_sub(1, true), (", stringify!($SelfT), "::MAX - 1, true));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.borrowing_sub(-1, false), (", stringify!($SelfT), "::MIN, true));")] - #[doc = concat!("assert_eq!(", stringify!($SelfT), "::MAX.borrowing_sub(-1, true), (", stringify!($SelfT), "::MAX, false));")] + /// // Only the most significant word is signed. + /// // + #[doc = concat!("// 6 8 (a = 6 × 2^", stringify!($BITS), " + 8)")] + #[doc = concat!("// - -5 9 (b = -5 × 2^", stringify!($BITS), " + 9)")] + /// // --------- + #[doc = concat!("// 10 MAX (diff = 10 × 2^", stringify!($BITS), " + 2^", stringify!($BITS), " - 1)")] + /// + #[doc = concat!("let (a1, a0): (", stringify!($SelfT), ", ", stringify!($UnsignedT), ") = (6, 8);")] + #[doc = concat!("let (b1, b0): (", stringify!($SelfT), ", ", stringify!($UnsignedT), ") = (-5, 9);")] + /// let borrow0 = false; + /// + #[doc = concat!("// ", stringify!($UnsignedT), "::borrowing_sub for the less significant words")] + /// let (diff0, borrow1) = a0.borrowing_sub(b0, borrow0); + /// assert_eq!(borrow1, true); + /// + #[doc = concat!("// ", stringify!($SelfT), "::borrowing_sub for the most significant word")] + /// let (diff1, overflow) = a1.borrowing_sub(b1, borrow1); + /// assert_eq!(overflow, false); + /// + #[doc = concat!("assert_eq!((diff1, diff0), (10, ", stringify!($UnsignedT), "::MAX));")] /// ``` #[unstable(feature = "bigint_helper_methods", issue = "85532")] #[rustc_const_unstable(feature = "const_bigint_helper_methods", issue = "85532")] diff --git a/library/core/src/ops/function.rs b/library/core/src/ops/function.rs index 127b047db..b7e1aee9d 100644 --- a/library/core/src/ops/function.rs +++ b/library/core/src/ops/function.rs @@ -1,4 +1,3 @@ -#[cfg(not(bootstrap))] use crate::marker::Tuple; /// The version of the call operator that takes an immutable receiver. @@ -54,87 +53,6 @@ use crate::marker::Tuple; /// let double = |x| x * 2; /// assert_eq!(call_with_one(double), 2); /// ``` -#[cfg(bootstrap)] -#[lang = "fn"] -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_paren_sugar] -#[rustc_on_unimplemented( - on( - Args = "()", - note = "wrap the `{Self}` in a closure with no arguments: `|| {{ /* code */ }}`" - ), - on( - _Self = "unsafe fn", - note = "unsafe function cannot be called generically without an unsafe block", - // SAFETY: tidy is not smart enough to tell that the below unsafe block is a string - label = "call the function in a closure: `|| unsafe {{ /* code */ }}`" - ), - message = "expected a `{Fn}<{Args}>` closure, found `{Self}`", - label = "expected an `Fn<{Args}>` closure, found `{Self}`" -)] -#[fundamental] // so that regex can rely that `&str: !FnMut` -#[must_use = "closures are lazy and do nothing unless called"] -#[const_trait] -pub trait Fn: FnMut { - /// Performs the call operation. - #[unstable(feature = "fn_traits", issue = "29625")] - extern "rust-call" fn call(&self, args: Args) -> Self::Output; -} - -/// The version of the call operator that takes an immutable receiver. -/// -/// Instances of `Fn` can be called repeatedly without mutating state. -/// -/// *This trait (`Fn`) is not to be confused with [function pointers] -/// (`fn`).* -/// -/// `Fn` is implemented automatically by closures which only take immutable -/// references to captured variables or don't capture anything at all, as well -/// as (safe) [function pointers] (with some caveats, see their documentation -/// for more details). Additionally, for any type `F` that implements `Fn`, `&F` -/// implements `Fn`, too. -/// -/// Since both [`FnMut`] and [`FnOnce`] are supertraits of `Fn`, any -/// instance of `Fn` can be used as a parameter where a [`FnMut`] or [`FnOnce`] -/// is expected. -/// -/// Use `Fn` as a bound when you want to accept a parameter of function-like -/// type and need to call it repeatedly and without mutating state (e.g., when -/// calling it concurrently). If you do not need such strict requirements, use -/// [`FnMut`] or [`FnOnce`] as bounds. -/// -/// See the [chapter on closures in *The Rust Programming Language*][book] for -/// some more information on this topic. -/// -/// Also of note is the special syntax for `Fn` traits (e.g. -/// `Fn(usize, bool) -> usize`). Those interested in the technical details of -/// this can refer to [the relevant section in the *Rustonomicon*][nomicon]. -/// -/// [book]: ../../book/ch13-01-closures.html -/// [function pointers]: fn -/// [nomicon]: ../../nomicon/hrtb.html -/// -/// # Examples -/// -/// ## Calling a closure -/// -/// ``` -/// let square = |x| x * x; -/// assert_eq!(square(5), 25); -/// ``` -/// -/// ## Using a `Fn` parameter -/// -/// ``` -/// fn call_with_one(func: F) -> usize -/// where F: Fn(usize) -> usize { -/// func(1) -/// } -/// -/// let double = |x| x * 2; -/// assert_eq!(call_with_one(double), 2); -/// ``` -#[cfg(not(bootstrap))] #[lang = "fn"] #[stable(feature = "rust1", since = "1.0.0")] #[rustc_paren_sugar] @@ -222,95 +140,6 @@ pub trait Fn: FnMut { /// /// assert_eq!(x, 5); /// ``` -#[cfg(bootstrap)] -#[lang = "fn_mut"] -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_paren_sugar] -#[rustc_on_unimplemented( - on( - Args = "()", - note = "wrap the `{Self}` in a closure with no arguments: `|| {{ /* code */ }}`" - ), - on( - _Self = "unsafe fn", - note = "unsafe function cannot be called generically without an unsafe block", - // SAFETY: tidy is not smart enough to tell that the below unsafe block is a string - label = "call the function in a closure: `|| unsafe {{ /* code */ }}`" - ), - message = "expected a `{FnMut}<{Args}>` closure, found `{Self}`", - label = "expected an `FnMut<{Args}>` closure, found `{Self}`" -)] -#[fundamental] // so that regex can rely that `&str: !FnMut` -#[must_use = "closures are lazy and do nothing unless called"] -#[const_trait] -pub trait FnMut: FnOnce { - /// Performs the call operation. - #[unstable(feature = "fn_traits", issue = "29625")] - extern "rust-call" fn call_mut(&mut self, args: Args) -> Self::Output; -} - -/// The version of the call operator that takes a mutable receiver. -/// -/// Instances of `FnMut` can be called repeatedly and may mutate state. -/// -/// `FnMut` is implemented automatically by closures which take mutable -/// references to captured variables, as well as all types that implement -/// [`Fn`], e.g., (safe) [function pointers] (since `FnMut` is a supertrait of -/// [`Fn`]). Additionally, for any type `F` that implements `FnMut`, `&mut F` -/// implements `FnMut`, too. -/// -/// Since [`FnOnce`] is a supertrait of `FnMut`, any instance of `FnMut` can be -/// used where a [`FnOnce`] is expected, and since [`Fn`] is a subtrait of -/// `FnMut`, any instance of [`Fn`] can be used where `FnMut` is expected. -/// -/// Use `FnMut` as a bound when you want to accept a parameter of function-like -/// type and need to call it repeatedly, while allowing it to mutate state. -/// If you don't want the parameter to mutate state, use [`Fn`] as a -/// bound; if you don't need to call it repeatedly, use [`FnOnce`]. -/// -/// See the [chapter on closures in *The Rust Programming Language*][book] for -/// some more information on this topic. -/// -/// Also of note is the special syntax for `Fn` traits (e.g. -/// `Fn(usize, bool) -> usize`). Those interested in the technical details of -/// this can refer to [the relevant section in the *Rustonomicon*][nomicon]. -/// -/// [book]: ../../book/ch13-01-closures.html -/// [function pointers]: fn -/// [nomicon]: ../../nomicon/hrtb.html -/// -/// # Examples -/// -/// ## Calling a mutably capturing closure -/// -/// ``` -/// let mut x = 5; -/// { -/// let mut square_x = || x *= x; -/// square_x(); -/// } -/// assert_eq!(x, 25); -/// ``` -/// -/// ## Using a `FnMut` parameter -/// -/// ``` -/// fn do_twice(mut func: F) -/// where F: FnMut() -/// { -/// func(); -/// func(); -/// } -/// -/// let mut x: usize = 1; -/// { -/// let add_two_to_x = || x += 2; -/// do_twice(add_two_to_x); -/// } -/// -/// assert_eq!(x, 5); -/// ``` -#[cfg(not(bootstrap))] #[lang = "fn_mut"] #[stable(feature = "rust1", since = "1.0.0")] #[rustc_paren_sugar] @@ -390,92 +219,6 @@ pub trait FnMut: FnOnce { /// /// // `consume_and_return_x` can no longer be invoked at this point /// ``` -#[cfg(bootstrap)] -#[lang = "fn_once"] -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_paren_sugar] -#[rustc_on_unimplemented( - on( - Args = "()", - note = "wrap the `{Self}` in a closure with no arguments: `|| {{ /* code */ }}`" - ), - on( - _Self = "unsafe fn", - note = "unsafe function cannot be called generically without an unsafe block", - // SAFETY: tidy is not smart enough to tell that the below unsafe block is a string - label = "call the function in a closure: `|| unsafe {{ /* code */ }}`" - ), - message = "expected a `{FnOnce}<{Args}>` closure, found `{Self}`", - label = "expected an `FnOnce<{Args}>` closure, found `{Self}`" -)] -#[fundamental] // so that regex can rely that `&str: !FnMut` -#[must_use = "closures are lazy and do nothing unless called"] -#[const_trait] -pub trait FnOnce { - /// The returned type after the call operator is used. - #[lang = "fn_once_output"] - #[stable(feature = "fn_once_output", since = "1.12.0")] - type Output; - - /// Performs the call operation. - #[unstable(feature = "fn_traits", issue = "29625")] - extern "rust-call" fn call_once(self, args: Args) -> Self::Output; -} - -/// The version of the call operator that takes a by-value receiver. -/// -/// Instances of `FnOnce` can be called, but might not be callable multiple -/// times. Because of this, if the only thing known about a type is that it -/// implements `FnOnce`, it can only be called once. -/// -/// `FnOnce` is implemented automatically by closures that might consume captured -/// variables, as well as all types that implement [`FnMut`], e.g., (safe) -/// [function pointers] (since `FnOnce` is a supertrait of [`FnMut`]). -/// -/// Since both [`Fn`] and [`FnMut`] are subtraits of `FnOnce`, any instance of -/// [`Fn`] or [`FnMut`] can be used where a `FnOnce` is expected. -/// -/// Use `FnOnce` as a bound when you want to accept a parameter of function-like -/// type and only need to call it once. If you need to call the parameter -/// repeatedly, use [`FnMut`] as a bound; if you also need it to not mutate -/// state, use [`Fn`]. -/// -/// See the [chapter on closures in *The Rust Programming Language*][book] for -/// some more information on this topic. -/// -/// Also of note is the special syntax for `Fn` traits (e.g. -/// `Fn(usize, bool) -> usize`). Those interested in the technical details of -/// this can refer to [the relevant section in the *Rustonomicon*][nomicon]. -/// -/// [book]: ../../book/ch13-01-closures.html -/// [function pointers]: fn -/// [nomicon]: ../../nomicon/hrtb.html -/// -/// # Examples -/// -/// ## Using a `FnOnce` parameter -/// -/// ``` -/// fn consume_with_relish(func: F) -/// where F: FnOnce() -> String -/// { -/// // `func` consumes its captured variables, so it cannot be run more -/// // than once. -/// println!("Consumed: {}", func()); -/// -/// println!("Delicious!"); -/// -/// // Attempting to invoke `func()` again will throw a `use of moved -/// // value` error for `func`. -/// } -/// -/// let x = String::from("x"); -/// let consume_and_return_x = move || x; -/// consume_with_relish(consume_and_return_x); -/// -/// // `consume_and_return_x` can no longer be invoked at this point -/// ``` -#[cfg(not(bootstrap))] #[lang = "fn_once"] #[stable(feature = "rust1", since = "1.0.0")] #[rustc_paren_sugar] @@ -507,68 +250,6 @@ pub trait FnOnce { extern "rust-call" fn call_once(self, args: Args) -> Self::Output; } -#[cfg(bootstrap)] -mod impls { - #[stable(feature = "rust1", since = "1.0.0")] - #[rustc_const_unstable(feature = "const_fn_trait_ref_impls", issue = "101803")] - impl const Fn for &F - where - F: ~const Fn, - { - extern "rust-call" fn call(&self, args: A) -> F::Output { - (**self).call(args) - } - } - - #[stable(feature = "rust1", since = "1.0.0")] - #[rustc_const_unstable(feature = "const_fn_trait_ref_impls", issue = "101803")] - impl const FnMut for &F - where - F: ~const Fn, - { - extern "rust-call" fn call_mut(&mut self, args: A) -> F::Output { - (**self).call(args) - } - } - - #[stable(feature = "rust1", since = "1.0.0")] - #[rustc_const_unstable(feature = "const_fn_trait_ref_impls", issue = "101803")] - impl const FnOnce for &F - where - F: ~const Fn, - { - type Output = F::Output; - - extern "rust-call" fn call_once(self, args: A) -> F::Output { - (*self).call(args) - } - } - - #[stable(feature = "rust1", since = "1.0.0")] - #[rustc_const_unstable(feature = "const_fn_trait_ref_impls", issue = "101803")] - impl const FnMut for &mut F - where - F: ~const FnMut, - { - extern "rust-call" fn call_mut(&mut self, args: A) -> F::Output { - (*self).call_mut(args) - } - } - - #[stable(feature = "rust1", since = "1.0.0")] - #[rustc_const_unstable(feature = "const_fn_trait_ref_impls", issue = "101803")] - impl const FnOnce for &mut F - where - F: ~const FnMut, - { - type Output = F::Output; - extern "rust-call" fn call_once(self, args: A) -> F::Output { - (*self).call_mut(args) - } - } -} - -#[cfg(not(bootstrap))] mod impls { use crate::marker::Tuple; diff --git a/library/core/src/ops/index.rs b/library/core/src/ops/index.rs index 5e3dc48b6..228efb0bc 100644 --- a/library/core/src/ops/index.rs +++ b/library/core/src/ops/index.rs @@ -165,7 +165,7 @@ see chapter in The Book : Index { +pub trait IndexMut: ~const Index { /// Performs the mutable indexing (`container[index]`) operation. /// /// # Panics diff --git a/library/core/src/ops/mod.rs b/library/core/src/ops/mod.rs index a5e5b13b3..97d9b750d 100644 --- a/library/core/src/ops/mod.rs +++ b/library/core/src/ops/mod.rs @@ -17,10 +17,10 @@ //! should have some resemblance to multiplication (and share expected //! properties like associativity). //! -//! Note that the `&&` and `||` operators short-circuit, i.e., they only -//! evaluate their second operand if it contributes to the result. Since this -//! behavior is not enforceable by traits, `&&` and `||` are not supported as -//! overloadable operators. +//! Note that the `&&` and `||` operators are currently not supported for +//! overloading. Due to their short circuiting nature, they require a different +//! design from traits for other operators like [`BitAnd`]. Designs for them are +//! under discussion. //! //! Many of the operators take their operands by value. In non-generic //! contexts involving built-in types, this is usually not a problem. @@ -201,7 +201,7 @@ pub(crate) use self::try_trait::{ChangeOutputType, NeverShortCircuit}; #[unstable(feature = "generator_trait", issue = "43122")] pub use self::generator::{Generator, GeneratorState}; -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] pub use self::unsize::CoerceUnsized; #[unstable(feature = "dispatch_from_dyn", issue = "none")] diff --git a/library/core/src/ops/unsize.rs b/library/core/src/ops/unsize.rs index a920b9165..b51f12580 100644 --- a/library/core/src/ops/unsize.rs +++ b/library/core/src/ops/unsize.rs @@ -31,41 +31,41 @@ use crate::marker::Unsize; /// [dst-coerce]: https://github.com/rust-lang/rfcs/blob/master/text/0982-dst-coercion.md /// [unsize]: crate::marker::Unsize /// [nomicon-coerce]: ../../nomicon/coercions.html -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] #[lang = "coerce_unsized"] pub trait CoerceUnsized { // Empty. } // &mut T -> &mut U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<&'a mut U> for &'a mut T {} // &mut T -> &U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, 'b: 'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<&'a U> for &'b mut T {} // &mut T -> *mut U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<*mut U> for &'a mut T {} // &mut T -> *const U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<*const U> for &'a mut T {} // &T -> &U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, 'b: 'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<&'a U> for &'b T {} // &T -> *const U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl<'a, T: ?Sized + Unsize, U: ?Sized> CoerceUnsized<*const U> for &'a T {} // *mut T -> *mut U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized<*mut U> for *mut T {} // *mut T -> *const U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized<*const U> for *mut T {} // *const T -> *const U -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl, U: ?Sized> CoerceUnsized<*const U> for *const T {} /// `DispatchFromDyn` is used in the implementation of object safety checks (specifically allowing diff --git a/library/core/src/option.rs b/library/core/src/option.rs index 505d964e5..7cc00e3f8 100644 --- a/library/core/src/option.rs +++ b/library/core/src/option.rs @@ -72,6 +72,50 @@ //! } //! ``` //! +//! # The question mark operator, `?` +//! +//! Similar to the [`Result`] type, when writing code that calls many functions that return the +//! [`Option`] type, handling `Some`/`None` can be tedious. The question mark +//! operator, [`?`], hides some of the boilerplate of propagating values +//! up the call stack. +//! +//! It replaces this: +//! +//! ``` +//! # #![allow(dead_code)] +//! fn add_last_numbers(stack: &mut Vec) -> Option { +//! let a = stack.pop(); +//! let b = stack.pop(); +//! +//! match (a, b) { +//! (Some(x), Some(y)) => Some(x + y), +//! _ => None, +//! } +//! } +//! +//! ``` +//! +//! With this: +//! +//! ``` +//! # #![allow(dead_code)] +//! fn add_last_numbers(stack: &mut Vec) -> Option { +//! Some(stack.pop()? + stack.pop()?) +//! } +//! ``` +//! +//! *It's much nicer!* +//! +//! Ending the expression with [`?`] will result in the [`Some`]'s unwrapped value, unless the +//! result is [`None`], in which case [`None`] is returned early from the enclosing function. +//! +//! [`?`] can be used in functions that return [`Option`] because of the +//! early return of [`None`] that it provides. +//! +//! [`?`]: crate::ops::Try +//! [`Some`]: Some +//! [`None`]: None +//! //! # Representation //! //! Rust guarantees to optimize the following types `T` such that @@ -608,13 +652,14 @@ impl Option { /// /// # Examples /// - /// Converts an Option<[String]> into an Option<[usize]>, preserving - /// the original. The [`map`] method takes the `self` argument by value, consuming the original, - /// so this technique uses `as_ref` to first take an `Option` to a reference - /// to the value inside the original. + /// Calculates the length of an Option<[String]> as an Option<[usize]> + /// without moving the [`String`]. The [`map`] method takes the `self` argument by value, + /// consuming the original, so this technique uses `as_ref` to first take an `Option` to a + /// reference to the value inside the original. /// /// [`map`]: Option::map /// [String]: ../../std/string/struct.String.html "String" + /// [`String`]: ../../std/string/struct.String.html "String" /// /// ``` /// let text: Option = Some("Hello, world!".to_string()); @@ -902,8 +947,8 @@ impl Option { /// /// # Examples /// - /// Converts an Option<[String]> into an Option<[usize]>, consuming - /// the original: + /// Calculates the length of an Option<[String]> as an + /// Option<[usize]>, consuming the original: /// /// [String]: ../../std/string/struct.String.html "String" /// ``` diff --git a/library/core/src/panic.rs b/library/core/src/panic.rs index 461b70c32..8338a5d7e 100644 --- a/library/core/src/panic.rs +++ b/library/core/src/panic.rs @@ -90,14 +90,14 @@ pub macro unreachable_2021 { ), } -/// An internal trait used by libstd to pass data from libstd to `panic_unwind` -/// and other panic runtimes. Not intended to be stabilized any time soon, do -/// not use. +/// An internal trait used by std to pass data from std to `panic_unwind` and +/// other panic runtimes. Not intended to be stabilized any time soon, do not +/// use. #[unstable(feature = "std_internals", issue = "none")] #[doc(hidden)] pub unsafe trait BoxMeUp { /// Take full ownership of the contents. - /// The return type is actually `Box`, but we cannot use `Box` in libcore. + /// The return type is actually `Box`, but we cannot use `Box` in core. /// /// After this method got called, only some dummy default value is left in `self`. /// Calling this method twice, or calling `get` after calling this method, is an error. diff --git a/library/core/src/panic/panic_info.rs b/library/core/src/panic/panic_info.rs index 1923155eb..0d385c9d1 100644 --- a/library/core/src/panic/panic_info.rs +++ b/library/core/src/panic/panic_info.rs @@ -157,7 +157,7 @@ impl fmt::Display for PanicInfo<'_> { write!(formatter, "'{}', ", payload)? } // NOTE: we cannot use downcast_ref::() here - // since String is not available in libcore! + // since String is not available in core! // The payload is a String when `std::panic!` is called with multiple arguments, // but in that case the message is also available. diff --git a/library/core/src/panicking.rs b/library/core/src/panicking.rs index a704a00fa..48e90e6d7 100644 --- a/library/core/src/panicking.rs +++ b/library/core/src/panicking.rs @@ -1,8 +1,8 @@ -//! Panic support for libcore +//! Panic support for core //! //! The core library cannot define panicking, but it does *declare* panicking. This -//! means that the functions inside of libcore are allowed to panic, but to be -//! useful an upstream crate must define panicking for libcore to use. The current +//! means that the functions inside of core are allowed to panic, but to be +//! useful an upstream crate must define panicking for core to use. The current //! interface for panicking is: //! //! ``` @@ -13,7 +13,7 @@ //! This definition allows for panicking with any general message, but it does not //! allow for failing with a `Box` value. (`PanicInfo` just contains a `&(dyn Any + Send)`, //! for which we fill in a dummy value in `PanicInfo::internal_constructor`.) -//! The reason for this is that libcore is not allowed to allocate. +//! The reason for this is that core is not allowed to allocate. //! //! This module contains a few other panicking functions, but these are just the //! necessary lang items for the compiler. All panics are funneled through this @@ -64,12 +64,17 @@ pub const fn panic_fmt(fmt: fmt::Arguments<'_>) -> ! { unsafe { panic_impl(&pi) } } -/// Like panic_fmt, but without unwinding and track_caller to reduce the impact on codesize. -/// Also just works on `str`, as a `fmt::Arguments` needs more space to be passed. +/// Like `panic_fmt`, but for non-unwinding panics. +/// +/// Has to be a separate function so that it can carry the `rustc_nounwind` attribute. #[cfg_attr(not(feature = "panic_immediate_abort"), inline(never), cold)] #[cfg_attr(feature = "panic_immediate_abort", inline)] +#[track_caller] +// This attribute has the key side-effect that if the panic handler ignores `can_unwind` +// and unwinds anyway, we will hit the "unwinding out of nounwind function" guard, +// which causes a "panic in a function that cannot unwind". #[rustc_nounwind] -pub fn panic_str_nounwind(msg: &'static str) -> ! { +pub fn panic_nounwind_fmt(fmt: fmt::Arguments<'_>) -> ! { if cfg!(feature = "panic_immediate_abort") { super::intrinsics::abort() } @@ -82,8 +87,6 @@ pub fn panic_str_nounwind(msg: &'static str) -> ! { } // PanicInfo with the `can_unwind` flag set to false forces an abort. - let pieces = [msg]; - let fmt = fmt::Arguments::new_v1(&pieces, &[]); let pi = PanicInfo::internal_constructor(Some(&fmt), Location::caller(), false); // SAFETY: `panic_impl` is defined in safe Rust code and thus is safe to call. @@ -93,7 +96,7 @@ pub fn panic_str_nounwind(msg: &'static str) -> ! { // Next we define a bunch of higher-level wrappers that all bottom out in the two core functions // above. -/// The underlying implementation of libcore's `panic!` macro when no formatting is used. +/// The underlying implementation of core's `panic!` macro when no formatting is used. // never inline unless panic_immediate_abort to avoid code // bloat at the call sites as much as possible #[cfg_attr(not(feature = "panic_immediate_abort"), inline(never), cold)] @@ -111,6 +114,15 @@ pub const fn panic(expr: &'static str) -> ! { panic_fmt(fmt::Arguments::new_v1(&[expr], &[])); } +/// Like `panic`, but without unwinding and track_caller to reduce the impact on codesize. +#[cfg_attr(not(feature = "panic_immediate_abort"), inline(never), cold)] +#[cfg_attr(feature = "panic_immediate_abort", inline)] +#[cfg_attr(not(bootstrap), lang = "panic_nounwind")] // needed by codegen for non-unwinding panics +#[rustc_nounwind] +pub fn panic_nounwind(expr: &'static str) -> ! { + panic_nounwind_fmt(fmt::Arguments::new_v1(&[expr], &[])); +} + #[inline] #[track_caller] #[rustc_diagnostic_item = "panic_str"] @@ -153,10 +165,11 @@ fn panic_bounds_check(index: usize, len: usize) -> ! { /// any extra arguments (including those synthesized by track_caller). #[cfg_attr(not(feature = "panic_immediate_abort"), inline(never), cold)] #[cfg_attr(feature = "panic_immediate_abort", inline)] -#[lang = "panic_no_unwind"] // needed by codegen for panic in nounwind function +#[cfg_attr(bootstrap, lang = "panic_no_unwind")] // needed by codegen for panic in nounwind function +#[cfg_attr(not(bootstrap), lang = "panic_cannot_unwind")] // needed by codegen for panic in nounwind function #[rustc_nounwind] -fn panic_no_unwind() -> ! { - panic_str_nounwind("panic in a function that cannot unwind") +fn panic_cannot_unwind() -> ! { + panic_nounwind("panic in a function that cannot unwind") } /// This function is used instead of panic_fmt in const eval. diff --git a/library/core/src/pin.rs b/library/core/src/pin.rs index 4524fa4c4..febe57dc9 100644 --- a/library/core/src/pin.rs +++ b/library/core/src/pin.rs @@ -485,6 +485,16 @@ impl> Pin

{ /// /// Unlike `Pin::new_unchecked`, this method is safe because the pointer /// `P` dereferences to an [`Unpin`] type, which cancels the pinning guarantees. + /// + /// # Examples + /// + /// ``` + /// use std::pin::Pin; + /// + /// let mut val: u8 = 5; + /// // We can pin the value, since it doesn't care about being moved + /// let mut pinned: Pin<&mut u8> = Pin::new(&mut val); + /// ``` #[inline(always)] #[rustc_const_unstable(feature = "const_pin", issue = "76654")] #[stable(feature = "pin", since = "1.33.0")] @@ -496,8 +506,20 @@ impl> Pin

{ /// Unwraps this `Pin

` returning the underlying pointer. /// - /// This requires that the data inside this `Pin` is [`Unpin`] so that we + /// This requires that the data inside this `Pin` implements [`Unpin`] so that we /// can ignore the pinning invariants when unwrapping it. + /// + /// # Examples + /// + /// ``` + /// use std::pin::Pin; + /// + /// let mut val: u8 = 5; + /// let pinned: Pin<&mut u8> = Pin::new(&mut val); + /// // Unwrap the pin to get a reference to the value + /// let r = Pin::into_inner(pinned); + /// assert_eq!(*r, 5); + /// ``` #[inline(always)] #[rustc_const_unstable(feature = "const_pin", issue = "76654")] #[stable(feature = "pin_into_inner", since = "1.39.0")] @@ -600,9 +622,8 @@ impl Pin

{ /// that the closure is pinned. /// /// The better alternative is to avoid all that trouble and do the pinning in the outer function - /// instead (here using the unstable `pin` macro): + /// instead (here using the [`pin!`][crate::pin::pin] macro): /// ``` - /// #![feature(pin_macro)] /// use std::pin::pin; /// use std::task::Context; /// use std::future::Future; @@ -707,6 +728,18 @@ impl Pin

{ /// /// This overwrites pinned data, but that is okay: its destructor gets /// run before being overwritten, so no pinning guarantee is violated. + /// + /// # Example + /// + /// ``` + /// use std::pin::Pin; + /// + /// let mut val: u8 = 5; + /// let mut pinned: Pin<&mut u8> = Pin::new(&mut val); + /// println!("{}", pinned); // 5 + /// pinned.as_mut().set(10); + /// println!("{}", pinned); // 10 + /// ``` #[stable(feature = "pin", since = "1.33.0")] #[inline(always)] pub fn set(&mut self, value: P::Target) @@ -720,7 +753,7 @@ impl Pin

{ impl<'a, T: ?Sized> Pin<&'a T> { /// Constructs a new pin by mapping the interior value. /// - /// For example, if you wanted to get a `Pin` of a field of something, + /// For example, if you wanted to get a `Pin` of a field of something, /// you could use this to get access to that field in one line of code. /// However, there are several gotchas with these "pinning projections"; /// see the [`pin` module] documentation for further details on that topic. @@ -823,7 +856,7 @@ impl<'a, T: ?Sized> Pin<&'a mut T> { /// Construct a new pin by mapping the interior value. /// - /// For example, if you wanted to get a `Pin` of a field of something, + /// For example, if you wanted to get a `Pin` of a field of something, /// you could use this to get access to that field in one line of code. /// However, there are several gotchas with these "pinning projections"; /// see the [`pin` module] documentation for further details on that topic. @@ -992,7 +1025,6 @@ impl DispatchFromDyn> for Pin

where P: DispatchFromDyn {} /// ### Basic usage /// /// ```rust -/// #![feature(pin_macro)] /// # use core::marker::PhantomPinned as Foo; /// use core::pin::{pin, Pin}; /// @@ -1010,7 +1042,6 @@ impl DispatchFromDyn> for Pin

where P: DispatchFromDyn {} /// ### Manually polling a `Future` (without `Unpin` bounds) /// /// ```rust -/// #![feature(pin_macro)] /// use std::{ /// future::Future, /// pin::pin, @@ -1049,7 +1080,7 @@ impl DispatchFromDyn> for Pin

where P: DispatchFromDyn {} /// ### With `Generator`s /// /// ```rust -/// #![feature(generators, generator_trait, pin_macro)] +/// #![feature(generators, generator_trait)] /// use core::{ /// ops::{Generator, GeneratorState}, /// pin::pin, @@ -1092,7 +1123,6 @@ impl DispatchFromDyn> for Pin

where P: DispatchFromDyn {} /// The following, for instance, fails to compile: /// /// ```rust,compile_fail -/// #![feature(pin_macro)] /// use core::pin::{pin, Pin}; /// # use core::{marker::PhantomPinned as Foo, mem::drop as stuff}; /// @@ -1134,7 +1164,7 @@ impl DispatchFromDyn> for Pin

where P: DispatchFromDyn {} /// constructor. /// /// [`Box::pin`]: ../../std/boxed/struct.Box.html#method.pin -#[unstable(feature = "pin_macro", issue = "93178")] +#[stable(feature = "pin_macro", since = "1.68.0")] #[rustc_macro_transparency = "semitransparent"] #[allow_internal_unstable(unsafe_pin_internals)] pub macro pin($value:expr $(,)?) { diff --git a/library/core/src/prelude/mod.rs b/library/core/src/prelude/mod.rs index 3cd3a3b78..12f762ef1 100644 --- a/library/core/src/prelude/mod.rs +++ b/library/core/src/prelude/mod.rs @@ -1,8 +1,8 @@ -//! The libcore prelude +//! The core prelude //! -//! This module is intended for users of libcore which do not link to libstd as -//! well. This module is imported by default when `#![no_std]` is used in the -//! same manner as the standard library's prelude. +//! This module is intended for users of core which do not link to std as well. +//! This module is imported by default when `#![no_std]` is used in the same +//! manner as the standard library's prelude. #![stable(feature = "core_prelude", since = "1.4.0")] diff --git a/library/core/src/prelude/v1.rs b/library/core/src/prelude/v1.rs index 2d67d742c..10525a16f 100644 --- a/library/core/src/prelude/v1.rs +++ b/library/core/src/prelude/v1.rs @@ -75,14 +75,12 @@ pub use crate::macros::builtin::{RustcDecodable, RustcEncodable}; // Do not `doc(no_inline)` so that they become doc items on their own // (no public module for them to be re-exported from). -#[cfg(not(bootstrap))] #[stable(feature = "builtin_macro_prelude", since = "1.38.0")] -pub use crate::macros::builtin::alloc_error_handler; -#[stable(feature = "builtin_macro_prelude", since = "1.38.0")] -pub use crate::macros::builtin::{bench, derive, global_allocator, test, test_case}; +pub use crate::macros::builtin::{ + alloc_error_handler, bench, derive, global_allocator, test, test_case, +}; #[unstable(feature = "derive_const", issue = "none")] -#[cfg(not(bootstrap))] pub use crate::macros::builtin::derive_const; #[unstable( @@ -104,5 +102,4 @@ pub use crate::macros::builtin::cfg_eval; issue = "23416", reason = "placeholder syntax for type ascription" )] -#[cfg(not(bootstrap))] pub use crate::macros::builtin::type_ascribe; diff --git a/library/core/src/ptr/alignment.rs b/library/core/src/ptr/alignment.rs index 64a5290c3..2123147c7 100644 --- a/library/core/src/ptr/alignment.rs +++ b/library/core/src/ptr/alignment.rs @@ -10,8 +10,7 @@ use crate::{cmp, fmt, hash, mem, num}; /// are likely not to be supported by actual allocators and linkers. #[unstable(feature = "ptr_alignment_type", issue = "102070")] #[derive(Copy, Clone, Eq)] -#[cfg_attr(bootstrap, derive(PartialEq))] -#[cfg_attr(not(bootstrap), derive_const(PartialEq))] +#[derive_const(PartialEq)] #[repr(transparent)] pub struct Alignment(AlignmentEnum); @@ -203,8 +202,7 @@ type AlignmentEnum = AlignmentEnum32; type AlignmentEnum = AlignmentEnum64; #[derive(Copy, Clone, Eq)] -#[cfg_attr(bootstrap, derive(PartialEq))] -#[cfg_attr(not(bootstrap), derive_const(PartialEq))] +#[derive_const(PartialEq)] #[repr(u16)] enum AlignmentEnum16 { _Align1Shl0 = 1 << 0, @@ -226,8 +224,7 @@ enum AlignmentEnum16 { } #[derive(Copy, Clone, Eq)] -#[cfg_attr(bootstrap, derive(PartialEq))] -#[cfg_attr(not(bootstrap), derive_const(PartialEq))] +#[derive_const(PartialEq)] #[repr(u32)] enum AlignmentEnum32 { _Align1Shl0 = 1 << 0, @@ -265,8 +262,7 @@ enum AlignmentEnum32 { } #[derive(Copy, Clone, Eq)] -#[cfg_attr(bootstrap, derive(PartialEq))] -#[cfg_attr(not(bootstrap), derive_const(PartialEq))] +#[derive_const(PartialEq)] #[repr(u64)] enum AlignmentEnum64 { _Align1Shl0 = 1 << 0, diff --git a/library/core/src/ptr/const_ptr.rs b/library/core/src/ptr/const_ptr.rs index d34813599..7b1cb5488 100644 --- a/library/core/src/ptr/const_ptr.rs +++ b/library/core/src/ptr/const_ptr.rs @@ -1,6 +1,6 @@ use super::*; use crate::cmp::Ordering::{self, Equal, Greater, Less}; -use crate::intrinsics; +use crate::intrinsics::{self, const_eval_select}; use crate::mem; use crate::slice::{self, SliceIndex}; @@ -34,12 +34,23 @@ impl *const T { #[rustc_const_unstable(feature = "const_ptr_is_null", issue = "74939")] #[inline] pub const fn is_null(self) -> bool { - // Compare via a cast to a thin pointer, so fat pointers are only - // considering their "data" part for null-ness. - match (self as *const u8).guaranteed_eq(null()) { - None => false, - Some(res) => res, + #[inline] + fn runtime_impl(ptr: *const u8) -> bool { + ptr.addr() == 0 } + + #[inline] + const fn const_impl(ptr: *const u8) -> bool { + // Compare via a cast to a thin pointer, so fat pointers are only + // considering their "data" part for null-ness. + match (ptr).guaranteed_eq(null_mut()) { + None => false, + Some(res) => res, + } + } + + // SAFETY: The two versions are equivalent at runtime. + unsafe { const_eval_select((self as *const u8,), const_impl, runtime_impl) } } /// Casts to a pointer of another type. @@ -191,14 +202,11 @@ impl *const T { #[must_use] #[inline(always)] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn addr(self) -> usize - where - T: Sized, - { + pub fn addr(self) -> usize { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. // SAFETY: Pointer-to-integer transmutes are valid (if you are okay with losing the // provenance). - unsafe { mem::transmute(self) } + unsafe { mem::transmute(self.cast::<()>()) } } /// Gets the "address" portion of the pointer, and 'exposes' the "provenance" part for future @@ -228,12 +236,9 @@ impl *const T { #[must_use] #[inline(always)] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn expose_addr(self) -> usize - where - T: Sized, - { + pub fn expose_addr(self) -> usize { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. - self as usize + self.cast::<()>() as usize } /// Creates a new pointer with the given address. @@ -251,10 +256,7 @@ impl *const T { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn with_addr(self, addr: usize) -> Self - where - T: Sized, - { + pub fn with_addr(self, addr: usize) -> Self { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. // // In the mean-time, this operation is defined to be "as if" it was @@ -277,10 +279,7 @@ impl *const T { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn map_addr(self, f: impl FnOnce(usize) -> usize) -> Self - where - T: Sized, - { + pub fn map_addr(self, f: impl FnOnce(usize) -> usize) -> Self { self.with_addr(f(self.addr())) } @@ -1008,7 +1007,7 @@ impl *const T { #[stable(feature = "pointer_methods", since = "1.26.0")] #[must_use = "returns a new pointer rather than modifying its argument"] #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")] - #[inline] + #[inline(always)] #[cfg_attr(miri, track_caller)] // even without panics, this helps for Miri backtraces pub const unsafe fn sub(self, count: usize) -> Self where @@ -1173,7 +1172,7 @@ impl *const T { #[stable(feature = "pointer_methods", since = "1.26.0")] #[must_use = "returns a new pointer rather than modifying its argument"] #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")] - #[inline] + #[inline(always)] pub const fn wrapping_sub(self, count: usize) -> Self where T: Sized, @@ -1350,26 +1349,6 @@ impl *const T { panic!("align_offset: align is not a power-of-two"); } - #[cfg(bootstrap)] - { - fn rt_impl(p: *const T, align: usize) -> usize { - // SAFETY: `align` has been checked to be a power of 2 above - unsafe { align_offset(p, align) } - } - - const fn ctfe_impl(_: *const T, _: usize) -> usize { - usize::MAX - } - - // SAFETY: - // It is permissible for `align_offset` to always return `usize::MAX`, - // algorithm correctness can not depend on `align_offset` returning non-max values. - // - // As such the behaviour can't change after replacing `align_offset` with `usize::MAX`, only performance can. - unsafe { intrinsics::const_eval_select((self, align), ctfe_impl, rt_impl) } - } - - #[cfg(not(bootstrap))] { // SAFETY: `align` has been checked to be a power of 2 above unsafe { align_offset(self, align) } @@ -1406,8 +1385,7 @@ impl *const T { /// is never aligned if cast to a type with a stricter alignment than the reference's /// underlying allocation. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1433,8 +1411,7 @@ impl *const T { /// Due to this behavior, it is possible that a runtime pointer derived from a compiletime /// pointer is aligned, even if the compiletime pointer wasn't aligned. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1460,8 +1437,7 @@ impl *const T { /// If a pointer is created from a fixed address, this function behaves the same during /// runtime and compiletime. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1537,8 +1513,7 @@ impl *const T { /// return `true` if the pointer is guaranteed to be aligned. This means that the pointer /// cannot be stricter aligned than the reference's underlying allocation. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1563,8 +1538,7 @@ impl *const T { /// Due to this behavior, it is possible that a runtime pointer derived from a compiletime /// pointer is aligned, even if the compiletime pointer wasn't aligned. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1588,8 +1562,7 @@ impl *const T { /// If a pointer is created from a fixed address, this function behaves the same during /// runtime and compiletime. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1613,11 +1586,22 @@ impl *const T { panic!("is_aligned_to: align is not a power-of-two"); } - // We can't use the address of `self` in a `const fn`, so we use `align_offset` instead. - // The cast to `()` is used to - // 1. deal with fat pointers; and - // 2. ensure that `align_offset` doesn't actually try to compute an offset. - self.cast::<()>().align_offset(align) == 0 + #[inline] + fn runtime_impl(ptr: *const (), align: usize) -> bool { + ptr.addr() & (align - 1) == 0 + } + + #[inline] + const fn const_impl(ptr: *const (), align: usize) -> bool { + // We can't use the address of `self` in a `const fn`, so we use `align_offset` instead. + // The cast to `()` is used to + // 1. deal with fat pointers; and + // 2. ensure that `align_offset` doesn't actually try to compute an offset. + ptr.align_offset(align) == 0 + } + + // SAFETY: The two versions are equivalent at runtime. + unsafe { const_eval_select((self.cast::<()>(), align), const_impl, runtime_impl) } } } diff --git a/library/core/src/ptr/metadata.rs b/library/core/src/ptr/metadata.rs index a8604843e..2ea032d4a 100644 --- a/library/core/src/ptr/metadata.rs +++ b/library/core/src/ptr/metadata.rs @@ -50,7 +50,7 @@ use crate::hash::{Hash, Hasher}; /// /// [`to_raw_parts`]: *const::to_raw_parts #[lang = "pointee_trait"] -#[cfg_attr(not(bootstrap), rustc_deny_explicit_impl)] +#[rustc_deny_explicit_impl] pub trait Pointee { /// The type for metadata in pointers and references to `Self`. #[lang = "metadata_type"] diff --git a/library/core/src/ptr/mod.rs b/library/core/src/ptr/mod.rs index 48b2e88da..1ad9af154 100644 --- a/library/core/src/ptr/mod.rs +++ b/library/core/src/ptr/mod.rs @@ -516,6 +516,27 @@ pub const fn null() -> *const T { from_raw_parts(invalid(0), ()) } +/// Creates a null mutable raw pointer. +/// +/// # Examples +/// +/// ``` +/// use std::ptr; +/// +/// let p: *mut i32 = ptr::null_mut(); +/// assert!(p.is_null()); +/// ``` +#[inline(always)] +#[must_use] +#[stable(feature = "rust1", since = "1.0.0")] +#[rustc_promotable] +#[rustc_const_stable(feature = "const_ptr_null", since = "1.24.0")] +#[rustc_allow_const_fn_unstable(ptr_metadata)] +#[rustc_diagnostic_item = "ptr_null_mut"] +pub const fn null_mut() -> *mut T { + from_raw_parts_mut(invalid_mut(0), ()) +} + /// Creates an invalid pointer with the given address. /// /// This is different from `addr as *const T`, which creates a pointer that picks up a previously @@ -663,25 +684,26 @@ where addr as *mut T } -/// Creates a null mutable raw pointer. +/// Convert a reference to a raw pointer. /// -/// # Examples -/// -/// ``` -/// use std::ptr; +/// This is equivalent to `r as *const T`, but is a bit safer since it will never silently change +/// type or mutability, in particular if the code is refactored. +#[inline(always)] +#[must_use] +#[unstable(feature = "ptr_from_ref", issue = "106116")] +pub fn from_ref(r: &T) -> *const T { + r +} + +/// Convert a mutable reference to a raw pointer. /// -/// let p: *mut i32 = ptr::null_mut(); -/// assert!(p.is_null()); -/// ``` +/// This is equivalent to `r as *mut T`, but is a bit safer since it will never silently change +/// type or mutability, in particular if the code is refactored. #[inline(always)] #[must_use] -#[stable(feature = "rust1", since = "1.0.0")] -#[rustc_promotable] -#[rustc_const_stable(feature = "const_ptr_null", since = "1.24.0")] -#[rustc_allow_const_fn_unstable(ptr_metadata)] -#[rustc_diagnostic_item = "ptr_null_mut"] -pub const fn null_mut() -> *mut T { - from_raw_parts_mut(invalid_mut(0), ()) +#[unstable(feature = "ptr_from_ref", issue = "106116")] +pub fn from_mut(r: &mut T) -> *mut T { + r } /// Forms a raw slice from a pointer and a length. @@ -1679,7 +1701,7 @@ pub(crate) const unsafe fn align_offset(p: *const T, a: usize) -> usiz // offset is not a multiple of `stride`, the input pointer was misaligned and no pointer // offset will be able to produce a `p` aligned to the specified `a`. // - // The naive `-p (mod a)` equation inhibits LLVM's ability to select instructions + // The naive `-p (mod a)` equation inhibits LLVM's ability to select instructions // like `lea`. We compute `(round_up_to_next_alignment(p, a) - p)` instead. This // redistributes operations around the load-bearing, but pessimizing `and` instruction // sufficiently for LLVM to be able to utilize the various optimizations it knows about. diff --git a/library/core/src/ptr/mut_ptr.rs b/library/core/src/ptr/mut_ptr.rs index c924a90b1..ed1e3bd48 100644 --- a/library/core/src/ptr/mut_ptr.rs +++ b/library/core/src/ptr/mut_ptr.rs @@ -1,6 +1,6 @@ use super::*; use crate::cmp::Ordering::{self, Equal, Greater, Less}; -use crate::intrinsics; +use crate::intrinsics::{self, const_eval_select}; use crate::slice::{self, SliceIndex}; impl *mut T { @@ -33,12 +33,23 @@ impl *mut T { #[rustc_const_unstable(feature = "const_ptr_is_null", issue = "74939")] #[inline] pub const fn is_null(self) -> bool { - // Compare via a cast to a thin pointer, so fat pointers are only - // considering their "data" part for null-ness. - match (self as *mut u8).guaranteed_eq(null_mut()) { - None => false, - Some(res) => res, + #[inline] + fn runtime_impl(ptr: *mut u8) -> bool { + ptr.addr() == 0 } + + #[inline] + const fn const_impl(ptr: *mut u8) -> bool { + // Compare via a cast to a thin pointer, so fat pointers are only + // considering their "data" part for null-ness. + match (ptr).guaranteed_eq(null_mut()) { + None => false, + Some(res) => res, + } + } + + // SAFETY: The two versions are equivalent at runtime. + unsafe { const_eval_select((self as *mut u8,), const_impl, runtime_impl) } } /// Casts to a pointer of another type. @@ -197,14 +208,11 @@ impl *mut T { #[must_use] #[inline(always)] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn addr(self) -> usize - where - T: Sized, - { + pub fn addr(self) -> usize { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. // SAFETY: Pointer-to-integer transmutes are valid (if you are okay with losing the // provenance). - unsafe { mem::transmute(self) } + unsafe { mem::transmute(self.cast::<()>()) } } /// Gets the "address" portion of the pointer, and 'exposes' the "provenance" part for future @@ -234,12 +242,9 @@ impl *mut T { #[must_use] #[inline(always)] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn expose_addr(self) -> usize - where - T: Sized, - { + pub fn expose_addr(self) -> usize { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. - self as usize + self.cast::<()>() as usize } /// Creates a new pointer with the given address. @@ -257,10 +262,7 @@ impl *mut T { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn with_addr(self, addr: usize) -> Self - where - T: Sized, - { + pub fn with_addr(self, addr: usize) -> Self { // FIXME(strict_provenance_magic): I am magic and should be a compiler intrinsic. // // In the mean-time, this operation is defined to be "as if" it was @@ -283,10 +285,7 @@ impl *mut T { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn map_addr(self, f: impl FnOnce(usize) -> usize) -> Self - where - T: Sized, - { + pub fn map_addr(self, f: impl FnOnce(usize) -> usize) -> Self { self.with_addr(f(self.addr())) } @@ -1110,7 +1109,7 @@ impl *mut T { #[stable(feature = "pointer_methods", since = "1.26.0")] #[must_use = "returns a new pointer rather than modifying its argument"] #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")] - #[inline] + #[inline(always)] #[cfg_attr(miri, track_caller)] // even without panics, this helps for Miri backtraces pub const unsafe fn sub(self, count: usize) -> Self where @@ -1275,7 +1274,7 @@ impl *mut T { #[stable(feature = "pointer_methods", since = "1.26.0")] #[must_use = "returns a new pointer rather than modifying its argument"] #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")] - #[inline] + #[inline(always)] pub const fn wrapping_sub(self, count: usize) -> Self where T: Sized, @@ -1618,26 +1617,6 @@ impl *mut T { panic!("align_offset: align is not a power-of-two"); } - #[cfg(bootstrap)] - { - fn rt_impl(p: *mut T, align: usize) -> usize { - // SAFETY: `align` has been checked to be a power of 2 above - unsafe { align_offset(p, align) } - } - - const fn ctfe_impl(_: *mut T, _: usize) -> usize { - usize::MAX - } - - // SAFETY: - // It is permissible for `align_offset` to always return `usize::MAX`, - // algorithm correctness can not depend on `align_offset` returning non-max values. - // - // As such the behaviour can't change after replacing `align_offset` with `usize::MAX`, only performance can. - unsafe { intrinsics::const_eval_select((self, align), ctfe_impl, rt_impl) } - } - - #[cfg(not(bootstrap))] { // SAFETY: `align` has been checked to be a power of 2 above unsafe { align_offset(self, align) } @@ -1674,8 +1653,7 @@ impl *mut T { /// is never aligned if cast to a type with a stricter alignment than the reference's /// underlying allocation. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// #![feature(const_mut_refs)] @@ -1702,8 +1680,7 @@ impl *mut T { /// Due to this behavior, it is possible that a runtime pointer derived from a compiletime /// pointer is aligned, even if the compiletime pointer wasn't aligned. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1730,8 +1707,7 @@ impl *mut T { /// If a pointer is created from a fixed address, this function behaves the same during /// runtime and compiletime. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1807,8 +1783,7 @@ impl *mut T { /// return `true` if the pointer is guaranteed to be aligned. This means that the pointer /// cannot be stricter aligned than the reference's underlying allocation. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// #![feature(const_mut_refs)] @@ -1834,8 +1809,7 @@ impl *mut T { /// Due to this behavior, it is possible that a runtime pointer derived from a compiletime /// pointer is aligned, even if the compiletime pointer wasn't aligned. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1860,8 +1834,7 @@ impl *mut T { /// If a pointer is created from a fixed address, this function behaves the same during /// runtime and compiletime. /// - #[cfg_attr(bootstrap, doc = "```ignore")] - #[cfg_attr(not(bootstrap), doc = "```")] + /// ``` /// #![feature(pointer_is_aligned)] /// #![feature(const_pointer_is_aligned)] /// @@ -1885,11 +1858,22 @@ impl *mut T { panic!("is_aligned_to: align is not a power-of-two"); } - // We can't use the address of `self` in a `const fn`, so we use `align_offset` instead. - // The cast to `()` is used to - // 1. deal with fat pointers; and - // 2. ensure that `align_offset` doesn't actually try to compute an offset. - self.cast::<()>().align_offset(align) == 0 + #[inline] + fn runtime_impl(ptr: *mut (), align: usize) -> bool { + ptr.addr() & (align - 1) == 0 + } + + #[inline] + const fn const_impl(ptr: *mut (), align: usize) -> bool { + // We can't use the address of `self` in a `const fn`, so we use `align_offset` instead. + // The cast to `()` is used to + // 1. deal with fat pointers; and + // 2. ensure that `align_offset` doesn't actually try to compute an offset. + ptr.align_offset(align) == 0 + } + + // SAFETY: The two versions are equivalent at runtime. + unsafe { const_eval_select((self.cast::<()>(), align), const_impl, runtime_impl) } } } diff --git a/library/core/src/ptr/non_null.rs b/library/core/src/ptr/non_null.rs index c4348169c..8c1a64886 100644 --- a/library/core/src/ptr/non_null.rs +++ b/library/core/src/ptr/non_null.rs @@ -268,10 +268,7 @@ impl NonNull { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn addr(self) -> NonZeroUsize - where - T: Sized, - { + pub fn addr(self) -> NonZeroUsize { // SAFETY: The pointer is guaranteed by the type to be non-null, // meaning that the address will be non-zero. unsafe { NonZeroUsize::new_unchecked(self.pointer.addr()) } @@ -286,10 +283,7 @@ impl NonNull { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn with_addr(self, addr: NonZeroUsize) -> Self - where - T: Sized, - { + pub fn with_addr(self, addr: NonZeroUsize) -> Self { // SAFETY: The result of `ptr::from::with_addr` is non-null because `addr` is guaranteed to be non-zero. unsafe { NonNull::new_unchecked(self.pointer.with_addr(addr.get()) as *mut _) } } @@ -303,10 +297,7 @@ impl NonNull { #[must_use] #[inline] #[unstable(feature = "strict_provenance", issue = "95228")] - pub fn map_addr(self, f: impl FnOnce(NonZeroUsize) -> NonZeroUsize) -> Self - where - T: Sized, - { + pub fn map_addr(self, f: impl FnOnce(NonZeroUsize) -> NonZeroUsize) -> Self { self.with_addr(f(self.addr())) } @@ -712,7 +703,7 @@ impl const Clone for NonNull { #[stable(feature = "nonnull", since = "1.25.0")] impl Copy for NonNull {} -#[unstable(feature = "coerce_unsized", issue = "27732")] +#[unstable(feature = "coerce_unsized", issue = "18598")] impl CoerceUnsized> for NonNull where T: Unsize {} #[unstable(feature = "dispatch_from_dyn", issue = "none")] diff --git a/library/core/src/result.rs b/library/core/src/result.rs index 3f33c5fd6..f00c40f35 100644 --- a/library/core/src/result.rs +++ b/library/core/src/result.rs @@ -209,11 +209,10 @@ //! //! *It's much nicer!* //! -//! Ending the expression with [`?`] will result in the unwrapped -//! success ([`Ok`]) value, unless the result is [`Err`], in which case -//! [`Err`] is returned early from the enclosing function. +//! Ending the expression with [`?`] will result in the [`Ok`]'s unwrapped value, unless the result +//! is [`Err`], in which case [`Err`] is returned early from the enclosing function. //! -//! [`?`] can only be used in functions that return [`Result`] because of the +//! [`?`] can be used in functions that return [`Result`] because of the //! early return of [`Err`] that it provides. //! //! [`expect`]: Result::expect diff --git a/library/core/src/slice/iter.rs b/library/core/src/slice/iter.rs index 062289767..90ab43d12 100644 --- a/library/core/src/slice/iter.rs +++ b/library/core/src/slice/iter.rs @@ -6,7 +6,7 @@ mod macros; use crate::cmp; use crate::cmp::Ordering; use crate::fmt; -use crate::intrinsics::{assume, exact_div, unchecked_sub}; +use crate::intrinsics::assume; use crate::iter::{FusedIterator, TrustedLen, TrustedRandomAccess, TrustedRandomAccessNoCoerce}; use crate::marker::{PhantomData, Send, Sized, Sync}; use crate::mem::{self, SizedTypeProperties}; @@ -35,12 +35,6 @@ impl<'a, T> IntoIterator for &'a mut [T] { } } -// Macro helper functions -#[inline(always)] -fn size_from_ptr(_: *const T) -> usize { - mem::size_of::() -} - /// Immutable slice iterator /// /// This struct is created by the [`iter`] method on [slices]. @@ -65,7 +59,7 @@ fn size_from_ptr(_: *const T) -> usize { #[must_use = "iterators are lazy and do nothing unless consumed"] pub struct Iter<'a, T: 'a> { ptr: NonNull, - end: *const T, // If T is a ZST, this is actually ptr+len. This encoding is picked so that + end: *const T, // If T is a ZST, this is actually ptr+len. This encoding is picked so that // ptr == end is a quick test for the Iterator being empty, that works // for both ZST and non-ZST. _marker: PhantomData<&'a T>, @@ -186,7 +180,7 @@ impl AsRef<[T]> for Iter<'_, T> { #[must_use = "iterators are lazy and do nothing unless consumed"] pub struct IterMut<'a, T: 'a> { ptr: NonNull, - end: *mut T, // If T is a ZST, this is actually ptr+len. This encoding is picked so that + end: *mut T, // If T is a ZST, this is actually ptr+len. This encoding is picked so that // ptr == end is a quick test for the Iterator being empty, that works // for both ZST and non-ZST. _marker: PhantomData<&'a mut T>, diff --git a/library/core/src/slice/iter/macros.rs b/library/core/src/slice/iter/macros.rs index ce51d48e3..0fd57b197 100644 --- a/library/core/src/slice/iter/macros.rs +++ b/library/core/src/slice/iter/macros.rs @@ -9,30 +9,20 @@ macro_rules! is_empty { }; } -// To get rid of some bounds checks (see `position`), we compute the length in a somewhat -// unexpected way. (Tested by `codegen/slice-position-bounds-check`.) macro_rules! len { ($self: ident) => {{ #![allow(unused_unsafe)] // we're sometimes used within an unsafe block let start = $self.ptr; - let size = size_from_ptr(start.as_ptr()); - if size == 0 { - // This _cannot_ use `unchecked_sub` because we depend on wrapping + if T::IS_ZST { + // This _cannot_ use `ptr_sub` because we depend on wrapping // to represent the length of long ZST slice iterators. $self.end.addr().wrapping_sub(start.as_ptr().addr()) } else { - // We know that `start <= end`, so can do better than `offset_from`, - // which needs to deal in signed. By setting appropriate flags here - // we can tell LLVM this, which helps it remove bounds checks. - // SAFETY: By the type invariant, `start <= end` - let diff = unsafe { unchecked_sub($self.end.addr(), start.as_ptr().addr()) }; - // By also telling LLVM that the pointers are apart by an exact - // multiple of the type size, it can optimize `len() == 0` down to - // `start == end` instead of `(end - start) < size`. - // SAFETY: By the type invariant, the pointers are aligned so the - // distance between them must be a multiple of pointee size - unsafe { exact_div(diff, size) } + // To get rid of some bounds checks (see `position`), we use ptr_sub instead of + // offset_from (Tested by `codegen/slice-position-bounds-check`.) + // SAFETY: by the type invariant pointers are aligned and `start <= end` + unsafe { $self.end.sub_ptr(start.as_ptr()) } } }}; } diff --git a/library/core/src/slice/mod.rs b/library/core/src/slice/mod.rs index d9281a925..d93a3a57e 100644 --- a/library/core/src/slice/mod.rs +++ b/library/core/src/slice/mod.rs @@ -29,13 +29,19 @@ use crate::slice; /// Pure rust memchr implementation, taken from rust-memchr pub mod memchr; +#[unstable( + feature = "slice_internals", + issue = "none", + reason = "exposed from core to be reused in std;" +)] +pub mod sort; + mod ascii; mod cmp; mod index; mod iter; mod raw; mod rotate; -mod sort; mod specialize; #[stable(feature = "rust1", since = "1.0.0")] @@ -703,7 +709,7 @@ impl [T] { // Because this function is first compiled in isolation, // this check tells LLVM that the indexing below is - // in-bounds. Then after inlining -- once the actual + // in-bounds. Then after inlining -- once the actual // lengths of the slices are known -- it's removed. let (a, b) = (&mut a[..n], &mut b[..n]); @@ -781,6 +787,22 @@ impl [T] { /// let mut iter = slice.windows(4); /// assert!(iter.next().is_none()); /// ``` + /// + /// There's no `windows_mut`, as that existing would let safe code violate the + /// "only one `&mut` at a time to the same thing" rule. However, you can sometimes + /// use [`Cell::as_slice_of_cells`](crate::cell::Cell::as_slice_of_cells) in + /// conjunction with `windows` to accomplish something similar: + /// ``` + /// use std::cell::Cell; + /// + /// let mut array = ['R', 'u', 's', 't', ' ', '2', '0', '1', '5']; + /// let slice = &mut array[..]; + /// let slice_of_cells: &[Cell] = Cell::from_mut(slice).as_slice_of_cells(); + /// for w in slice_of_cells.windows(3) { + /// Cell::swap(&w[0], &w[2]); + /// } + /// assert_eq!(array, ['s', 't', ' ', '2', '0', '1', '5', 'u', 'R']); + /// ``` #[stable(feature = "rust1", since = "1.0.0")] #[inline] pub fn windows(&self, size: usize) -> Windows<'_, T> { @@ -893,7 +915,7 @@ impl [T] { #[stable(feature = "chunks_exact", since = "1.31.0")] #[inline] pub fn chunks_exact(&self, chunk_size: usize) -> ChunksExact<'_, T> { - assert_ne!(chunk_size, 0); + assert_ne!(chunk_size, 0, "chunks cannot have a size of zero"); ChunksExact::new(self, chunk_size) } @@ -935,7 +957,7 @@ impl [T] { #[stable(feature = "chunks_exact", since = "1.31.0")] #[inline] pub fn chunks_exact_mut(&mut self, chunk_size: usize) -> ChunksExactMut<'_, T> { - assert_ne!(chunk_size, 0); + assert_ne!(chunk_size, 0, "chunks cannot have a size of zero"); ChunksExactMut::new(self, chunk_size) } @@ -1002,11 +1024,22 @@ impl [T] { /// assert_eq!(chunks, &[['l', 'o'], ['r', 'e']]); /// assert_eq!(remainder, &['m']); /// ``` + /// + /// If you expect the slice to be an exact multiple, you can combine + /// `let`-`else` with an empty slice pattern: + /// ``` + /// #![feature(slice_as_chunks)] + /// let slice = ['R', 'u', 's', 't']; + /// let (chunks, []) = slice.as_chunks::<2>() else { + /// panic!("slice didn't have even length") + /// }; + /// assert_eq!(chunks, &[['R', 'u'], ['s', 't']]); + /// ``` #[unstable(feature = "slice_as_chunks", issue = "74985")] #[inline] #[must_use] pub fn as_chunks(&self) -> (&[[T; N]], &[T]) { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); let len = self.len() / N; let (multiple_of_n, remainder) = self.split_at(len * N); // SAFETY: We already panicked for zero, and ensured by construction @@ -1037,7 +1070,7 @@ impl [T] { #[inline] #[must_use] pub fn as_rchunks(&self) -> (&[T], &[[T; N]]) { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); let len = self.len() / N; let (remainder, multiple_of_n) = self.split_at(self.len() - len * N); // SAFETY: We already panicked for zero, and ensured by construction @@ -1076,7 +1109,7 @@ impl [T] { #[unstable(feature = "array_chunks", issue = "74985")] #[inline] pub fn array_chunks(&self) -> ArrayChunks<'_, T, N> { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); ArrayChunks::new(self) } @@ -1155,7 +1188,7 @@ impl [T] { #[inline] #[must_use] pub fn as_chunks_mut(&mut self) -> (&mut [[T; N]], &mut [T]) { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); let len = self.len() / N; let (multiple_of_n, remainder) = self.split_at_mut(len * N); // SAFETY: We already panicked for zero, and ensured by construction @@ -1192,7 +1225,7 @@ impl [T] { #[inline] #[must_use] pub fn as_rchunks_mut(&mut self) -> (&mut [T], &mut [[T; N]]) { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); let len = self.len() / N; let (remainder, multiple_of_n) = self.split_at_mut(self.len() - len * N); // SAFETY: We already panicked for zero, and ensured by construction @@ -1233,11 +1266,11 @@ impl [T] { #[unstable(feature = "array_chunks", issue = "74985")] #[inline] pub fn array_chunks_mut(&mut self) -> ArrayChunksMut<'_, T, N> { - assert_ne!(N, 0); + assert_ne!(N, 0, "chunks cannot have a size of zero"); ArrayChunksMut::new(self) } - /// Returns an iterator over overlapping windows of `N` elements of a slice, + /// Returns an iterator over overlapping windows of `N` elements of a slice, /// starting at the beginning of the slice. /// /// This is the const generic equivalent of [`windows`]. @@ -1265,7 +1298,7 @@ impl [T] { #[unstable(feature = "array_windows", issue = "75027")] #[inline] pub fn array_windows(&self) -> ArrayWindows<'_, T, N> { - assert_ne!(N, 0); + assert_ne!(N, 0, "windows cannot have a size of zero"); ArrayWindows::new(self) } @@ -2465,7 +2498,7 @@ impl [T] { let mid = left + size / 2; // SAFETY: the while condition means `size` is strictly positive, so - // `size/2 < size`. Thus `left + size/2 < left + size`, which + // `size/2 < size`. Thus `left + size/2 < left + size`, which // coupled with the `left + size <= self.len()` invariant means // we have `left + size/2 < self.len()`, and this is in-bounds. let cmp = f(unsafe { self.get_unchecked(mid) }); @@ -3795,7 +3828,7 @@ impl [T] { /// The slice is assumed to be partitioned according to the given predicate. /// This means that all elements for which the predicate returns true are at the start of the slice /// and all elements for which the predicate returns false are at the end. - /// For example, [7, 15, 3, 5, 4, 12, 6] is a partitioned under the predicate x % 2 != 0 + /// For example, `[7, 15, 3, 5, 4, 12, 6]` is partitioned under the predicate `x % 2 != 0` /// (all odd numbers are at the start, all even at the end). /// /// If this slice is not partitioned, the returned result is unspecified and meaningless, diff --git a/library/core/src/slice/sort.rs b/library/core/src/slice/sort.rs index 87f77b7f2..2181f9a81 100644 --- a/library/core/src/slice/sort.rs +++ b/library/core/src/slice/sort.rs @@ -3,8 +3,11 @@ //! This module contains a sorting algorithm based on Orson Peters' pattern-defeating quicksort, //! published at: //! -//! Unstable sorting is compatible with libcore because it doesn't allocate memory, unlike our +//! Unstable sorting is compatible with core because it doesn't allocate memory, unlike our //! stable sorting implementation. +//! +//! In addition it also contains the core logic of the stable sort used by `slice::sort` based on +//! TimSort. use crate::cmp; use crate::mem::{self, MaybeUninit, SizedTypeProperties}; @@ -18,9 +21,9 @@ struct CopyOnDrop { impl Drop for CopyOnDrop { fn drop(&mut self) { - // SAFETY: This is a helper class. - // Please refer to its usage for correctness. - // Namely, one must be sure that `src` and `dst` does not overlap as required by `ptr::copy_nonoverlapping`. + // SAFETY: This is a helper class. + // Please refer to its usage for correctness. + // Namely, one must be sure that `src` and `dst` does not overlap as required by `ptr::copy_nonoverlapping`. unsafe { ptr::copy_nonoverlapping(self.src, self.dest, 1); } @@ -831,6 +834,15 @@ fn partition_at_index_loop<'a, T, F>( ) where F: FnMut(&T, &T) -> bool, { + // Limit the amount of iterations and fall back to heapsort, similarly to `slice::sort_unstable`. + // This lowers the worst case running time from O(n^2) to O(n log n). + // FIXME: Investigate whether it would be better to use something like Median of Medians + // or Fast Deterministic Selection to guarantee O(n) worst case. + let mut limit = usize::BITS - v.len().leading_zeros(); + + // True if the last partitioning was reasonably balanced. + let mut was_balanced = true; + loop { // For slices of up to this length it's probably faster to simply sort them. const MAX_INSERTION: usize = 10; @@ -839,6 +851,18 @@ fn partition_at_index_loop<'a, T, F>( return; } + if limit == 0 { + heapsort(v, is_less); + return; + } + + // If the last partitioning was imbalanced, try breaking patterns in the slice by shuffling + // some elements around. Hopefully we'll choose a better pivot this time. + if !was_balanced { + break_patterns(v); + limit -= 1; + } + // Choose a pivot let (pivot, _) = choose_pivot(v, is_less); @@ -863,6 +887,7 @@ fn partition_at_index_loop<'a, T, F>( } let (mid, _) = partition(v, pivot, is_less); + was_balanced = cmp::min(mid, v.len() - mid) >= v.len() / 8; // Split the slice into `left`, `pivot`, and `right`. let (left, right) = v.split_at_mut(mid); @@ -883,6 +908,7 @@ fn partition_at_index_loop<'a, T, F>( } } +/// Reorder the slice such that the element at `index` is at its final sorted position. pub fn partition_at_index( v: &mut [T], index: usize, @@ -927,3 +953,513 @@ where let pivot = &mut pivot[0]; (left, pivot, right) } + +/// Inserts `v[0]` into pre-sorted sequence `v[1..]` so that whole `v[..]` becomes sorted. +/// +/// This is the integral subroutine of insertion sort. +fn insert_head(v: &mut [T], is_less: &mut F) +where + F: FnMut(&T, &T) -> bool, +{ + if v.len() >= 2 && is_less(&v[1], &v[0]) { + // SAFETY: Copy tmp back even if panic, and ensure unique observation. + unsafe { + // There are three ways to implement insertion here: + // + // 1. Swap adjacent elements until the first one gets to its final destination. + // However, this way we copy data around more than is necessary. If elements are big + // structures (costly to copy), this method will be slow. + // + // 2. Iterate until the right place for the first element is found. Then shift the + // elements succeeding it to make room for it and finally place it into the + // remaining hole. This is a good method. + // + // 3. Copy the first element into a temporary variable. Iterate until the right place + // for it is found. As we go along, copy every traversed element into the slot + // preceding it. Finally, copy data from the temporary variable into the remaining + // hole. This method is very good. Benchmarks demonstrated slightly better + // performance than with the 2nd method. + // + // All methods were benchmarked, and the 3rd showed best results. So we chose that one. + let tmp = mem::ManuallyDrop::new(ptr::read(&v[0])); + + // Intermediate state of the insertion process is always tracked by `hole`, which + // serves two purposes: + // 1. Protects integrity of `v` from panics in `is_less`. + // 2. Fills the remaining hole in `v` in the end. + // + // Panic safety: + // + // If `is_less` panics at any point during the process, `hole` will get dropped and + // fill the hole in `v` with `tmp`, thus ensuring that `v` still holds every object it + // initially held exactly once. + let mut hole = InsertionHole { src: &*tmp, dest: &mut v[1] }; + ptr::copy_nonoverlapping(&v[1], &mut v[0], 1); + + for i in 2..v.len() { + if !is_less(&v[i], &*tmp) { + break; + } + ptr::copy_nonoverlapping(&v[i], &mut v[i - 1], 1); + hole.dest = &mut v[i]; + } + // `hole` gets dropped and thus copies `tmp` into the remaining hole in `v`. + } + } + + // When dropped, copies from `src` into `dest`. + struct InsertionHole { + src: *const T, + dest: *mut T, + } + + impl Drop for InsertionHole { + fn drop(&mut self) { + // SAFETY: The caller must ensure that src and dest are correctly set. + unsafe { + ptr::copy_nonoverlapping(self.src, self.dest, 1); + } + } + } +} + +/// Merges non-decreasing runs `v[..mid]` and `v[mid..]` using `buf` as temporary storage, and +/// stores the result into `v[..]`. +/// +/// # Safety +/// +/// The two slices must be non-empty and `mid` must be in bounds. Buffer `buf` must be long enough +/// to hold a copy of the shorter slice. Also, `T` must not be a zero-sized type. +unsafe fn merge(v: &mut [T], mid: usize, buf: *mut T, is_less: &mut F) +where + F: FnMut(&T, &T) -> bool, +{ + let len = v.len(); + let v = v.as_mut_ptr(); + + // SAFETY: mid and len must be in-bounds of v. + let (v_mid, v_end) = unsafe { (v.add(mid), v.add(len)) }; + + // The merge process first copies the shorter run into `buf`. Then it traces the newly copied + // run and the longer run forwards (or backwards), comparing their next unconsumed elements and + // copying the lesser (or greater) one into `v`. + // + // As soon as the shorter run is fully consumed, the process is done. If the longer run gets + // consumed first, then we must copy whatever is left of the shorter run into the remaining + // hole in `v`. + // + // Intermediate state of the process is always tracked by `hole`, which serves two purposes: + // 1. Protects integrity of `v` from panics in `is_less`. + // 2. Fills the remaining hole in `v` if the longer run gets consumed first. + // + // Panic safety: + // + // If `is_less` panics at any point during the process, `hole` will get dropped and fill the + // hole in `v` with the unconsumed range in `buf`, thus ensuring that `v` still holds every + // object it initially held exactly once. + let mut hole; + + if mid <= len - mid { + // The left run is shorter. + + // SAFETY: buf must have enough capacity for `v[..mid]`. + unsafe { + ptr::copy_nonoverlapping(v, buf, mid); + hole = MergeHole { start: buf, end: buf.add(mid), dest: v }; + } + + // Initially, these pointers point to the beginnings of their arrays. + let left = &mut hole.start; + let mut right = v_mid; + let out = &mut hole.dest; + + while *left < hole.end && right < v_end { + // Consume the lesser side. + // If equal, prefer the left run to maintain stability. + + // SAFETY: left and right must be valid and part of v same for out. + unsafe { + let to_copy = if is_less(&*right, &**left) { + get_and_increment(&mut right) + } else { + get_and_increment(left) + }; + ptr::copy_nonoverlapping(to_copy, get_and_increment(out), 1); + } + } + } else { + // The right run is shorter. + + // SAFETY: buf must have enough capacity for `v[mid..]`. + unsafe { + ptr::copy_nonoverlapping(v_mid, buf, len - mid); + hole = MergeHole { start: buf, end: buf.add(len - mid), dest: v_mid }; + } + + // Initially, these pointers point past the ends of their arrays. + let left = &mut hole.dest; + let right = &mut hole.end; + let mut out = v_end; + + while v < *left && buf < *right { + // Consume the greater side. + // If equal, prefer the right run to maintain stability. + + // SAFETY: left and right must be valid and part of v same for out. + unsafe { + let to_copy = if is_less(&*right.sub(1), &*left.sub(1)) { + decrement_and_get(left) + } else { + decrement_and_get(right) + }; + ptr::copy_nonoverlapping(to_copy, decrement_and_get(&mut out), 1); + } + } + } + // Finally, `hole` gets dropped. If the shorter run was not fully consumed, whatever remains of + // it will now be copied into the hole in `v`. + + unsafe fn get_and_increment(ptr: &mut *mut T) -> *mut T { + let old = *ptr; + + // SAFETY: ptr.add(1) must still be a valid pointer and part of `v`. + *ptr = unsafe { ptr.add(1) }; + old + } + + unsafe fn decrement_and_get(ptr: &mut *mut T) -> *mut T { + // SAFETY: ptr.sub(1) must still be a valid pointer and part of `v`. + *ptr = unsafe { ptr.sub(1) }; + *ptr + } + + // When dropped, copies the range `start..end` into `dest..`. + struct MergeHole { + start: *mut T, + end: *mut T, + dest: *mut T, + } + + impl Drop for MergeHole { + fn drop(&mut self) { + // SAFETY: `T` is not a zero-sized type, and these are pointers into a slice's elements. + unsafe { + let len = self.end.sub_ptr(self.start); + ptr::copy_nonoverlapping(self.start, self.dest, len); + } + } + } +} + +/// This merge sort borrows some (but not all) ideas from TimSort, which used to be described in +/// detail [here](https://github.com/python/cpython/blob/main/Objects/listsort.txt). However Python +/// has switched to a Powersort based implementation. +/// +/// The algorithm identifies strictly descending and non-descending subsequences, which are called +/// natural runs. There is a stack of pending runs yet to be merged. Each newly found run is pushed +/// onto the stack, and then some pairs of adjacent runs are merged until these two invariants are +/// satisfied: +/// +/// 1. for every `i` in `1..runs.len()`: `runs[i - 1].len > runs[i].len` +/// 2. for every `i` in `2..runs.len()`: `runs[i - 2].len > runs[i - 1].len + runs[i].len` +/// +/// The invariants ensure that the total running time is *O*(*n* \* log(*n*)) worst-case. +pub fn merge_sort( + v: &mut [T], + is_less: &mut CmpF, + elem_alloc_fn: ElemAllocF, + elem_dealloc_fn: ElemDeallocF, + run_alloc_fn: RunAllocF, + run_dealloc_fn: RunDeallocF, +) where + CmpF: FnMut(&T, &T) -> bool, + ElemAllocF: Fn(usize) -> *mut T, + ElemDeallocF: Fn(*mut T, usize), + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), +{ + // Slices of up to this length get sorted using insertion sort. + const MAX_INSERTION: usize = 20; + // Very short runs are extended using insertion sort to span at least this many elements. + const MIN_RUN: usize = 10; + + // The caller should have already checked that. + debug_assert!(!T::IS_ZST); + + let len = v.len(); + + // Short arrays get sorted in-place via insertion sort to avoid allocations. + if len <= MAX_INSERTION { + if len >= 2 { + for i in (0..len - 1).rev() { + insert_head(&mut v[i..], is_less); + } + } + return; + } + + // Allocate a buffer to use as scratch memory. We keep the length 0 so we can keep in it + // shallow copies of the contents of `v` without risking the dtors running on copies if + // `is_less` panics. When merging two sorted runs, this buffer holds a copy of the shorter run, + // which will always have length at most `len / 2`. + let buf = BufGuard::new(len / 2, elem_alloc_fn, elem_dealloc_fn); + let buf_ptr = buf.buf_ptr; + + let mut runs = RunVec::new(run_alloc_fn, run_dealloc_fn); + + // In order to identify natural runs in `v`, we traverse it backwards. That might seem like a + // strange decision, but consider the fact that merges more often go in the opposite direction + // (forwards). According to benchmarks, merging forwards is slightly faster than merging + // backwards. To conclude, identifying runs by traversing backwards improves performance. + let mut end = len; + while end > 0 { + // Find the next natural run, and reverse it if it's strictly descending. + let mut start = end - 1; + if start > 0 { + start -= 1; + + // SAFETY: The v.get_unchecked must be fed with correct inbound indicies. + unsafe { + if is_less(v.get_unchecked(start + 1), v.get_unchecked(start)) { + while start > 0 && is_less(v.get_unchecked(start), v.get_unchecked(start - 1)) { + start -= 1; + } + v[start..end].reverse(); + } else { + while start > 0 && !is_less(v.get_unchecked(start), v.get_unchecked(start - 1)) + { + start -= 1; + } + } + } + } + + // Insert some more elements into the run if it's too short. Insertion sort is faster than + // merge sort on short sequences, so this significantly improves performance. + while start > 0 && end - start < MIN_RUN { + start -= 1; + insert_head(&mut v[start..end], is_less); + } + + // Push this run onto the stack. + runs.push(TimSortRun { start, len: end - start }); + end = start; + + // Merge some pairs of adjacent runs to satisfy the invariants. + while let Some(r) = collapse(runs.as_slice()) { + let left = runs[r + 1]; + let right = runs[r]; + // SAFETY: `buf_ptr` must hold enough capacity for the shorter of the two sides, and + // neither side may be on length 0. + unsafe { + merge(&mut v[left.start..right.start + right.len], left.len, buf_ptr, is_less); + } + runs[r] = TimSortRun { start: left.start, len: left.len + right.len }; + runs.remove(r + 1); + } + } + + // Finally, exactly one run must remain in the stack. + debug_assert!(runs.len() == 1 && runs[0].start == 0 && runs[0].len == len); + + // Examines the stack of runs and identifies the next pair of runs to merge. More specifically, + // if `Some(r)` is returned, that means `runs[r]` and `runs[r + 1]` must be merged next. If the + // algorithm should continue building a new run instead, `None` is returned. + // + // TimSort is infamous for its buggy implementations, as described here: + // http://envisage-project.eu/timsort-specification-and-verification/ + // + // The gist of the story is: we must enforce the invariants on the top four runs on the stack. + // Enforcing them on just top three is not sufficient to ensure that the invariants will still + // hold for *all* runs in the stack. + // + // This function correctly checks invariants for the top four runs. Additionally, if the top + // run starts at index 0, it will always demand a merge operation until the stack is fully + // collapsed, in order to complete the sort. + #[inline] + fn collapse(runs: &[TimSortRun]) -> Option { + let n = runs.len(); + if n >= 2 + && (runs[n - 1].start == 0 + || runs[n - 2].len <= runs[n - 1].len + || (n >= 3 && runs[n - 3].len <= runs[n - 2].len + runs[n - 1].len) + || (n >= 4 && runs[n - 4].len <= runs[n - 3].len + runs[n - 2].len)) + { + if n >= 3 && runs[n - 3].len < runs[n - 1].len { Some(n - 3) } else { Some(n - 2) } + } else { + None + } + } + + // Extremely basic versions of Vec. + // Their use is super limited and by having the code here, it allows reuse between the sort + // implementations. + struct BufGuard + where + ElemDeallocF: Fn(*mut T, usize), + { + buf_ptr: *mut T, + capacity: usize, + elem_dealloc_fn: ElemDeallocF, + } + + impl BufGuard + where + ElemDeallocF: Fn(*mut T, usize), + { + fn new( + len: usize, + elem_alloc_fn: ElemAllocF, + elem_dealloc_fn: ElemDeallocF, + ) -> Self + where + ElemAllocF: Fn(usize) -> *mut T, + { + Self { buf_ptr: elem_alloc_fn(len), capacity: len, elem_dealloc_fn } + } + } + + impl Drop for BufGuard + where + ElemDeallocF: Fn(*mut T, usize), + { + fn drop(&mut self) { + (self.elem_dealloc_fn)(self.buf_ptr, self.capacity); + } + } + + struct RunVec + where + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), + { + buf_ptr: *mut TimSortRun, + capacity: usize, + len: usize, + run_alloc_fn: RunAllocF, + run_dealloc_fn: RunDeallocF, + } + + impl RunVec + where + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), + { + fn new(run_alloc_fn: RunAllocF, run_dealloc_fn: RunDeallocF) -> Self { + // Most slices can be sorted with at most 16 runs in-flight. + const START_RUN_CAPACITY: usize = 16; + + Self { + buf_ptr: run_alloc_fn(START_RUN_CAPACITY), + capacity: START_RUN_CAPACITY, + len: 0, + run_alloc_fn, + run_dealloc_fn, + } + } + + fn push(&mut self, val: TimSortRun) { + if self.len == self.capacity { + let old_capacity = self.capacity; + let old_buf_ptr = self.buf_ptr; + + self.capacity = self.capacity * 2; + self.buf_ptr = (self.run_alloc_fn)(self.capacity); + + // SAFETY: buf_ptr new and old were correctly allocated and old_buf_ptr has + // old_capacity valid elements. + unsafe { + ptr::copy_nonoverlapping(old_buf_ptr, self.buf_ptr, old_capacity); + } + + (self.run_dealloc_fn)(old_buf_ptr, old_capacity); + } + + // SAFETY: The invariant was just checked. + unsafe { + self.buf_ptr.add(self.len).write(val); + } + self.len += 1; + } + + fn remove(&mut self, index: usize) { + if index >= self.len { + panic!("Index out of bounds"); + } + + // SAFETY: buf_ptr needs to be valid and len invariant upheld. + unsafe { + // the place we are taking from. + let ptr = self.buf_ptr.add(index); + + // Shift everything down to fill in that spot. + ptr::copy(ptr.add(1), ptr, self.len - index - 1); + } + self.len -= 1; + } + + fn as_slice(&self) -> &[TimSortRun] { + // SAFETY: Safe as long as buf_ptr is valid and len invariant was upheld. + unsafe { &*ptr::slice_from_raw_parts(self.buf_ptr, self.len) } + } + + fn len(&self) -> usize { + self.len + } + } + + impl core::ops::Index for RunVec + where + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), + { + type Output = TimSortRun; + + fn index(&self, index: usize) -> &Self::Output { + if index < self.len { + // SAFETY: buf_ptr and len invariant must be upheld. + unsafe { + return &*(self.buf_ptr.add(index)); + } + } + + panic!("Index out of bounds"); + } + } + + impl core::ops::IndexMut for RunVec + where + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), + { + fn index_mut(&mut self, index: usize) -> &mut Self::Output { + if index < self.len { + // SAFETY: buf_ptr and len invariant must be upheld. + unsafe { + return &mut *(self.buf_ptr.add(index)); + } + } + + panic!("Index out of bounds"); + } + } + + impl Drop for RunVec + where + RunAllocF: Fn(usize) -> *mut TimSortRun, + RunDeallocF: Fn(*mut TimSortRun, usize), + { + fn drop(&mut self) { + // As long as TimSortRun is Copy we don't need to drop them individually but just the + // whole allocation. + (self.run_dealloc_fn)(self.buf_ptr, self.capacity); + } + } +} + +/// Internal type used by merge_sort. +#[derive(Clone, Copy, Debug)] +pub struct TimSortRun { + len: usize, + start: usize, +} diff --git a/library/core/src/str/iter.rs b/library/core/src/str/iter.rs index 24083ee6a..d969475aa 100644 --- a/library/core/src/str/iter.rs +++ b/library/core/src/str/iter.rs @@ -585,16 +585,17 @@ where impl<'a, P: Pattern<'a>> SplitInternal<'a, P> { #[inline] fn get_end(&mut self) -> Option<&'a str> { - if !self.finished && (self.allow_trailing_empty || self.end - self.start > 0) { + if !self.finished { self.finished = true; - // SAFETY: `self.start` and `self.end` always lie on unicode boundaries. - unsafe { - let string = self.matcher.haystack().get_unchecked(self.start..self.end); - Some(string) + + if self.allow_trailing_empty || self.end - self.start > 0 { + // SAFETY: `self.start` and `self.end` always lie on unicode boundaries. + let string = unsafe { self.matcher.haystack().get_unchecked(self.start..self.end) }; + return Some(string); } - } else { - None } + + None } #[inline] @@ -716,14 +717,14 @@ impl<'a, P: Pattern<'a>> SplitInternal<'a, P> { } #[inline] - fn as_str(&self) -> &'a str { + fn remainder(&self) -> Option<&'a str> { // `Self::get_end` doesn't change `self.start` if self.finished { - return ""; + return None; } // SAFETY: `self.start` and `self.end` always lie on unicode boundaries. - unsafe { self.matcher.haystack().get_unchecked(self.start..self.end) } + Some(unsafe { self.matcher.haystack().get_unchecked(self.start..self.end) }) } } @@ -746,44 +747,48 @@ generate_pattern_iterators! { } impl<'a, P: Pattern<'a>> Split<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "Mary had a little lamb".split(' '); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// split.next(); - /// assert_eq!(split.as_str(), "had a little lamb"); + /// assert_eq!(split.remainder(), Some("had a little lamb")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } impl<'a, P: Pattern<'a>> RSplit<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "Mary had a little lamb".rsplit(' '); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// split.next(); - /// assert_eq!(split.as_str(), "Mary had a little"); + /// assert_eq!(split.remainder(), Some("Mary had a little")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } @@ -806,44 +811,48 @@ generate_pattern_iterators! { } impl<'a, P: Pattern<'a>> SplitTerminator<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "A..B..".split_terminator('.'); - /// assert_eq!(split.as_str(), "A..B.."); + /// assert_eq!(split.remainder(), Some("A..B..")); /// split.next(); - /// assert_eq!(split.as_str(), ".B.."); + /// assert_eq!(split.remainder(), Some(".B..")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } impl<'a, P: Pattern<'a>> RSplitTerminator<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "A..B..".rsplit_terminator('.'); - /// assert_eq!(split.as_str(), "A..B.."); + /// assert_eq!(split.remainder(), Some("A..B..")); /// split.next(); - /// assert_eq!(split.as_str(), "A..B"); + /// assert_eq!(split.remainder(), Some("A..B")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } @@ -905,8 +914,8 @@ impl<'a, P: Pattern<'a>> SplitNInternal<'a, P> { } #[inline] - fn as_str(&self) -> &'a str { - self.iter.as_str() + fn remainder(&self) -> Option<&'a str> { + self.iter.remainder() } } @@ -929,44 +938,48 @@ generate_pattern_iterators! { } impl<'a, P: Pattern<'a>> SplitN<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "Mary had a little lamb".splitn(3, ' '); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// split.next(); - /// assert_eq!(split.as_str(), "had a little lamb"); + /// assert_eq!(split.remainder(), Some("had a little lamb")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } impl<'a, P: Pattern<'a>> RSplitN<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_as_str)] + /// #![feature(str_split_remainder)] /// let mut split = "Mary had a little lamb".rsplitn(3, ' '); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// split.next(); - /// assert_eq!(split.as_str(), "Mary had a little"); + /// assert_eq!(split.remainder(), Some("Mary had a little")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } @@ -1239,22 +1252,22 @@ impl<'a> SplitWhitespace<'a> { /// # Examples /// /// ``` - /// #![feature(str_split_whitespace_as_str)] + /// #![feature(str_split_whitespace_remainder)] /// /// let mut split = "Mary had a little lamb".split_whitespace(); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// /// split.next(); - /// assert_eq!(split.as_str(), "had a little lamb"); + /// assert_eq!(split.remainder(), Some("had a little lamb")); /// /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] #[must_use] - #[unstable(feature = "str_split_whitespace_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.inner.iter.as_str() + #[unstable(feature = "str_split_whitespace_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.inner.iter.remainder() } } @@ -1290,32 +1303,34 @@ impl<'a> DoubleEndedIterator for SplitAsciiWhitespace<'a> { impl FusedIterator for SplitAsciiWhitespace<'_> {} impl<'a> SplitAsciiWhitespace<'a> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_whitespace_as_str)] + /// #![feature(str_split_whitespace_remainder)] /// /// let mut split = "Mary had a little lamb".split_ascii_whitespace(); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// /// split.next(); - /// assert_eq!(split.as_str(), "had a little lamb"); + /// assert_eq!(split.remainder(), Some("had a little lamb")); /// /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] #[must_use] - #[unstable(feature = "str_split_whitespace_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { + #[unstable(feature = "str_split_whitespace_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { if self.inner.iter.iter.finished { - return ""; + return None; } // SAFETY: Slice is created from str. - unsafe { crate::str::from_utf8_unchecked(&self.inner.iter.iter.v) } + Some(unsafe { crate::str::from_utf8_unchecked(&self.inner.iter.iter.v) }) } } @@ -1358,23 +1373,25 @@ impl<'a, P: Pattern<'a, Searcher: ReverseSearcher<'a>>> DoubleEndedIterator impl<'a, P: Pattern<'a>> FusedIterator for SplitInclusive<'a, P> {} impl<'a, P: Pattern<'a>> SplitInclusive<'a, P> { - /// Returns remainder of the split string + /// Returns remainder of the split string. + /// + /// If the iterator is empty, returns `None`. /// /// # Examples /// /// ``` - /// #![feature(str_split_inclusive_as_str)] + /// #![feature(str_split_inclusive_remainder)] /// let mut split = "Mary had a little lamb".split_inclusive(' '); - /// assert_eq!(split.as_str(), "Mary had a little lamb"); + /// assert_eq!(split.remainder(), Some("Mary had a little lamb")); /// split.next(); - /// assert_eq!(split.as_str(), "had a little lamb"); + /// assert_eq!(split.remainder(), Some("had a little lamb")); /// split.by_ref().for_each(drop); - /// assert_eq!(split.as_str(), ""); + /// assert_eq!(split.remainder(), None); /// ``` #[inline] - #[unstable(feature = "str_split_inclusive_as_str", issue = "77998")] - pub fn as_str(&self) -> &'a str { - self.0.as_str() + #[unstable(feature = "str_split_inclusive_remainder", issue = "77998")] + pub fn remainder(&self) -> Option<&'a str> { + self.0.remainder() } } diff --git a/library/core/src/str/mod.rs b/library/core/src/str/mod.rs index 45fd2caae..ab2f8520e 100644 --- a/library/core/src/str/mod.rs +++ b/library/core/src/str/mod.rs @@ -368,7 +368,7 @@ impl str { #[inline(always)] pub unsafe fn as_bytes_mut(&mut self) -> &mut [u8] { // SAFETY: the cast from `&str` to `&[u8]` is safe since `str` - // has the same layout as `&[u8]` (only libstd can make this guarantee). + // has the same layout as `&[u8]` (only std can make this guarantee). // The pointer dereference is safe since it comes from a mutable reference which // is guaranteed to be valid for writes. unsafe { &mut *(self as *mut str as *mut [u8]) } @@ -970,8 +970,10 @@ impl str { /// An iterator over the lines of a string, as string slices. /// - /// Lines are ended with either a newline (`\n`) or a carriage return with - /// a line feed (`\r\n`). + /// Lines are split at line endings that are either newlines (`\n`) or + /// sequences of a carriage return followed by a line feed (`\r\n`). + /// + /// Line terminators are not included in the lines returned by the iterator. /// /// The final line ending is optional. A string that ends with a final line /// ending will return the same lines as an otherwise identical string diff --git a/library/core/src/sync/atomic.rs b/library/core/src/sync/atomic.rs index edc68d6fa..14367eb09 100644 --- a/library/core/src/sync/atomic.rs +++ b/library/core/src/sync/atomic.rs @@ -1786,6 +1786,42 @@ impl AtomicPtr { // SAFETY: data races are prevented by atomic intrinsics. unsafe { atomic_xor(self.p.get(), core::ptr::invalid_mut(val), order).cast() } } + + /// Returns a mutable pointer to the underlying pointer. + /// + /// Doing non-atomic reads and writes on the resulting integer can be a data race. + /// This method is mostly useful for FFI, where the function signature may use + /// `*mut *mut T` instead of `&AtomicPtr`. + /// + /// Returning an `*mut` pointer from a shared reference to this atomic is safe because the + /// atomic types work with interior mutability. All modifications of an atomic change the value + /// through a shared reference, and can do so safely as long as they use atomic operations. Any + /// use of the returned raw pointer requires an `unsafe` block and still has to uphold the same + /// restriction: operations on it must be atomic. + /// + /// # Examples + /// + /// ```ignore (extern-declaration) + /// #![feature(atomic_mut_ptr)] + //// use std::sync::atomic::AtomicPtr; + /// + /// extern "C" { + /// fn my_atomic_op(arg: *mut *mut u32); + /// } + /// + /// let mut value = 17; + /// let atomic = AtomicPtr::new(&mut value); + /// + /// // SAFETY: Safe as long as `my_atomic_op` is atomic. + /// unsafe { + /// my_atomic_op(atomic.as_mut_ptr()); + /// } + /// ``` + #[inline] + #[unstable(feature = "atomic_mut_ptr", reason = "recently added", issue = "66893")] + pub fn as_mut_ptr(&self) -> *mut *mut T { + self.p.get() + } } #[cfg(target_has_atomic_load_store = "8")] @@ -2678,9 +2714,9 @@ macro_rules! atomic_int { #[doc = concat!(" fn my_atomic_op(arg: *mut ", stringify!($int_type), ");")] /// } /// - #[doc = concat!("let mut atomic = ", stringify!($atomic_type), "::new(1);")] + #[doc = concat!("let atomic = ", stringify!($atomic_type), "::new(1);")] /// - // SAFETY: Safe as long as `my_atomic_op` is atomic. + /// // SAFETY: Safe as long as `my_atomic_op` is atomic. /// unsafe { /// my_atomic_op(atomic.as_mut_ptr()); /// } diff --git a/library/core/src/sync/exclusive.rs b/library/core/src/sync/exclusive.rs index c65c27500..301ad41c9 100644 --- a/library/core/src/sync/exclusive.rs +++ b/library/core/src/sync/exclusive.rs @@ -138,7 +138,7 @@ impl Exclusive { unsafe { Pin::new_unchecked(&mut self.get_unchecked_mut().inner) } } - /// Build a _mutable_ references to an `Exclusive` from + /// Build a _mutable_ reference to an `Exclusive` from /// a _mutable_ reference to a `T`. This allows you to skip /// building an `Exclusive` with [`Exclusive::new`]. #[unstable(feature = "exclusive_wrapper", issue = "98407")] @@ -149,7 +149,7 @@ impl Exclusive { unsafe { &mut *(r as *mut T as *mut Exclusive) } } - /// Build a _pinned mutable_ references to an `Exclusive` from + /// Build a _pinned mutable_ reference to an `Exclusive` from /// a _pinned mutable_ reference to a `T`. This allows you to skip /// building an `Exclusive` with [`Exclusive::new`]. #[unstable(feature = "exclusive_wrapper", issue = "98407")] diff --git a/library/core/src/task/poll.rs b/library/core/src/task/poll.rs index f1dc4f7b5..25b61c0e6 100644 --- a/library/core/src/task/poll.rs +++ b/library/core/src/task/poll.rs @@ -9,7 +9,7 @@ use crate::task::Ready; /// scheduled to receive a wakeup instead. #[must_use = "this `Poll` may be a `Pending` variant, which should be handled"] #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)] -#[cfg_attr(not(bootstrap), lang = "Poll")] +#[lang = "Poll"] #[stable(feature = "futures_api", since = "1.36.0")] pub enum Poll { /// Represents that a value is immediately ready. diff --git a/library/core/src/task/wake.rs b/library/core/src/task/wake.rs index 0cff972df..89adfccd9 100644 --- a/library/core/src/task/wake.rs +++ b/library/core/src/task/wake.rs @@ -104,7 +104,7 @@ pub struct RawWakerVTable { /// pointer. wake_by_ref: unsafe fn(*const ()), - /// This function gets called when a [`RawWaker`] gets dropped. + /// This function gets called when a [`Waker`] gets dropped. /// /// The implementation of this function must make sure to release any /// resources that are associated with this instance of a [`RawWaker`] and @@ -151,7 +151,7 @@ impl RawWakerVTable { /// /// # `drop` /// - /// This function gets called when a [`RawWaker`] gets dropped. + /// This function gets called when a [`Waker`] gets dropped. /// /// The implementation of this function must make sure to release any /// resources that are associated with this instance of a [`RawWaker`] and @@ -174,6 +174,7 @@ impl RawWakerVTable { /// Currently, `Context` only serves to provide access to a [`&Waker`](Waker) /// which can be used to wake the current task. #[stable(feature = "futures_api", since = "1.36.0")] +#[cfg_attr(not(bootstrap), lang = "Context")] pub struct Context<'a> { waker: &'a Waker, // Ensure we future-proof against variance changes by forcing @@ -181,6 +182,9 @@ pub struct Context<'a> { // are contravariant while return-position lifetimes are // covariant). _marker: PhantomData &'a ()>, + // Ensure `Context` is `!Send` and `!Sync` in order to allow + // for future `!Send` and / or `!Sync` fields. + _marker2: PhantomData<*mut ()>, } impl<'a> Context<'a> { @@ -190,7 +194,7 @@ impl<'a> Context<'a> { #[must_use] #[inline] pub const fn from_waker(waker: &'a Waker) -> Self { - Context { waker, _marker: PhantomData } + Context { waker, _marker: PhantomData, _marker2: PhantomData } } /// Returns a reference to the [`Waker`] for the current task. diff --git a/library/core/src/unicode/mod.rs b/library/core/src/unicode/mod.rs index 72fa059b7..e1faa407d 100644 --- a/library/core/src/unicode/mod.rs +++ b/library/core/src/unicode/mod.rs @@ -17,7 +17,7 @@ mod unicode_data; #[stable(feature = "unicode_version", since = "1.45.0")] pub const UNICODE_VERSION: (u8, u8, u8) = unicode_data::UNICODE_VERSION; -// For use in liballoc, not re-exported in libstd. +// For use in alloc, not re-exported in std. pub use unicode_data::{ case_ignorable::lookup as Case_Ignorable, cased::lookup as Cased, conversions, }; diff --git a/library/core/tests/any.rs b/library/core/tests/any.rs index e98dac8d1..a8f6b7ebb 100644 --- a/library/core/tests/any.rs +++ b/library/core/tests/any.rs @@ -131,7 +131,6 @@ fn distinct_type_names() { assert_ne!(type_name_of_val(Velocity), type_name_of_val(Velocity(0.0, -9.8)),); } -#[cfg(not(bootstrap))] #[test] fn dyn_type_name() { trait Foo { diff --git a/library/core/tests/char.rs b/library/core/tests/char.rs index 8542e5c70..ac0b2ca16 100644 --- a/library/core/tests/char.rs +++ b/library/core/tests/char.rs @@ -306,6 +306,10 @@ fn test_decode_utf16() { } check(&[0xD800, 0x41, 0x42], &[Err(0xD800), Ok('A'), Ok('B')]); check(&[0xD800, 0], &[Err(0xD800), Ok('\0')]); + check(&[0xD800], &[Err(0xD800)]); + check(&[0xD840, 0xDC00], &[Ok('\u{20000}')]); + check(&[0xD840, 0xD840, 0xDC00], &[Err(0xD840), Ok('\u{20000}')]); + check(&[0xDC00, 0xD840], &[Err(0xDC00), Err(0xD840)]); } #[test] diff --git a/library/core/tests/lazy.rs b/library/core/tests/lazy.rs index 70fcc6d2d..c7c3c479b 100644 --- a/library/core/tests/lazy.rs +++ b/library/core/tests/lazy.rs @@ -106,6 +106,12 @@ fn lazy_new() { assert_eq!(called.get(), 1); } +// Check that we can infer `T` from closure's type. +#[test] +fn lazy_type_inference() { + let _ = LazyCell::new(|| ()); +} + #[test] fn aliasing_in_get() { let x = OnceCell::new(); diff --git a/library/core/tests/lib.rs b/library/core/tests/lib.rs index 99d4a40c4..42a26ae16 100644 --- a/library/core/tests/lib.rs +++ b/library/core/tests/lib.rs @@ -48,7 +48,6 @@ #![feature(is_sorted)] #![feature(layout_for_ptr)] #![feature(pattern)] -#![feature(pin_macro)] #![feature(sort_internals)] #![feature(slice_take)] #![feature(slice_from_ptr_range)] @@ -155,3 +154,16 @@ mod time; mod tuple; mod unicode; mod waker; + +/// Copied from `std::test_helpers::test_rng`, see that function for rationale. +#[track_caller] +#[allow(dead_code)] // Not used in all configurations. +pub(crate) fn test_rng() -> rand_xorshift::XorShiftRng { + use core::hash::{BuildHasher, Hash, Hasher}; + let mut hasher = std::collections::hash_map::RandomState::new().build_hasher(); + core::panic::Location::caller().hash(&mut hasher); + let hc64 = hasher.finish(); + let seed_vec = hc64.to_le_bytes().into_iter().chain(0u8..8).collect::>(); + let seed: [u8; 16] = seed_vec.as_slice().try_into().unwrap(); + rand::SeedableRng::from_seed(seed) +} diff --git a/library/core/tests/mem.rs b/library/core/tests/mem.rs index 1cfb4fd9f..f7740a114 100644 --- a/library/core/tests/mem.rs +++ b/library/core/tests/mem.rs @@ -77,7 +77,6 @@ fn align_of_val_basic() { } #[test] -#[cfg(not(bootstrap))] // stage 0 doesn't have the fix yet, so the test fails fn align_of_val_raw_packed() { #[repr(C, packed)] struct B { diff --git a/library/core/tests/num/flt2dec/random.rs b/library/core/tests/num/flt2dec/random.rs index d09500393..0084c1c81 100644 --- a/library/core/tests/num/flt2dec/random.rs +++ b/library/core/tests/num/flt2dec/random.rs @@ -9,8 +9,6 @@ use core::num::flt2dec::MAX_SIG_DIGITS; use core::num::flt2dec::{decode, DecodableFloat, Decoded, FullDecoded}; use rand::distributions::{Distribution, Uniform}; -use rand::rngs::StdRng; -use rand::SeedableRng; pub fn decode_finite(v: T) -> Decoded { match decode(v).1 { @@ -92,7 +90,7 @@ where if cfg!(target_os = "emscripten") { return; // using rng pulls in i128 support, which doesn't work } - let mut rng = StdRng::from_entropy(); + let mut rng = crate::test_rng(); let f32_range = Uniform::new(0x0000_0001u32, 0x7f80_0000); iterate("f32_random_equivalence_test", k, n, f, g, |_| { let x = f32::from_bits(f32_range.sample(&mut rng)); @@ -108,7 +106,7 @@ where if cfg!(target_os = "emscripten") { return; // using rng pulls in i128 support, which doesn't work } - let mut rng = StdRng::from_entropy(); + let mut rng = crate::test_rng(); let f64_range = Uniform::new(0x0000_0000_0000_0001u64, 0x7ff0_0000_0000_0000); iterate("f64_random_equivalence_test", k, n, f, g, |_| { let x = f64::from_bits(f64_range.sample(&mut rng)); diff --git a/library/core/tests/ptr.rs b/library/core/tests/ptr.rs index 90bc83510..80d30f14c 100644 --- a/library/core/tests/ptr.rs +++ b/library/core/tests/ptr.rs @@ -359,7 +359,6 @@ fn align_offset_zst() { } #[test] -#[cfg(not(bootstrap))] fn align_offset_zst_const() { const { // For pointers of stride = 0, the pointer is already aligned or it cannot be aligned at @@ -397,7 +396,6 @@ fn align_offset_stride_one() { } #[test] -#[cfg(not(bootstrap))] fn align_offset_stride_one_const() { const { // For pointers of stride = 1, the pointer can always be aligned. The offset is equal to @@ -493,7 +491,6 @@ fn align_offset_various_strides() { } #[test] -#[cfg(not(bootstrap))] fn align_offset_various_strides_const() { const unsafe fn test_stride(ptr: *const T, numptr: usize, align: usize) { let mut expected = usize::MAX; @@ -561,7 +558,6 @@ fn align_offset_various_strides_const() { } #[test] -#[cfg(not(bootstrap))] fn align_offset_with_provenance_const() { const { // On some platforms (e.g. msp430-none-elf), the alignment of `i32` is less than 4. @@ -681,7 +677,6 @@ fn align_offset_issue_103361() { } #[test] -#[cfg(not(bootstrap))] fn align_offset_issue_103361_const() { #[cfg(target_pointer_width = "64")] const SIZE: usize = 1 << 47; @@ -715,7 +710,6 @@ fn is_aligned() { } #[test] -#[cfg(not(bootstrap))] fn is_aligned_const() { const { let data = 42; @@ -734,18 +728,6 @@ fn is_aligned_const() { } } -#[test] -#[cfg(bootstrap)] -fn is_aligned_const() { - const { - let data = 42; - let ptr: *const i32 = &data; - // The bootstrap compiler always returns false for is_aligned. - assert!(!ptr.is_aligned()); - assert!(!ptr.is_aligned_to(1)); - } -} - #[test] fn offset_from() { let mut a = [0; 5]; @@ -825,7 +807,7 @@ fn ptr_metadata_bounds() { } // "Synthetic" trait impls generated by the compiler like those of `Pointee` // are not checked for bounds of associated type. - // So with a buggy libcore we could have both: + // So with a buggy core we could have both: // * `::Metadata == DynMetadata` // * `DynMetadata: !PartialEq` // … and cause an ICE here: diff --git a/library/core/tests/slice.rs b/library/core/tests/slice.rs index 4e06e0f43..39559cdbb 100644 --- a/library/core/tests/slice.rs +++ b/library/core/tests/slice.rs @@ -1488,7 +1488,7 @@ mod slice_index { // optional: // // one or more similar inputs for which data[input] succeeds, - // and the corresponding output as an array. This helps validate + // and the corresponding output as an array. This helps validate // "critical points" where an input range straddles the boundary // between valid and invalid. // (such as the input `len..len`, which is just barely valid) @@ -1805,7 +1805,7 @@ fn brute_force_rotate_test_1() { fn sort_unstable() { use core::cmp::Ordering::{Equal, Greater, Less}; use core::slice::heapsort; - use rand::{rngs::StdRng, seq::SliceRandom, Rng, SeedableRng}; + use rand::{seq::SliceRandom, Rng}; // Miri is too slow (but still need to `chain` to make the types match) let lens = if cfg!(miri) { (2..20).chain(0..0) } else { (2..25).chain(500..510) }; @@ -1813,7 +1813,7 @@ fn sort_unstable() { let mut v = [0; 600]; let mut tmp = [0; 600]; - let mut rng = StdRng::from_entropy(); + let mut rng = crate::test_rng(); for len in lens { let v = &mut v[0..len]; @@ -1879,11 +1879,10 @@ fn sort_unstable() { #[cfg_attr(miri, ignore)] // Miri is too slow fn select_nth_unstable() { use core::cmp::Ordering::{Equal, Greater, Less}; - use rand::rngs::StdRng; use rand::seq::SliceRandom; - use rand::{Rng, SeedableRng}; + use rand::Rng; - let mut rng = StdRng::from_entropy(); + let mut rng = crate::test_rng(); for len in (2..21).chain(500..501) { let mut orig = vec![0; len]; diff --git a/library/core/tests/str.rs b/library/core/tests/str.rs index ed939ca71..f5066343a 100644 --- a/library/core/tests/str.rs +++ b/library/core/tests/str.rs @@ -1 +1 @@ -// All `str` tests live in liballoc/tests +// All `str` tests live in library/alloc/tests/str.rs diff --git a/library/core/tests/task.rs b/library/core/tests/task.rs index 56be30e92..163b34c96 100644 --- a/library/core/tests/task.rs +++ b/library/core/tests/task.rs @@ -1,4 +1,4 @@ -use core::task::{Context, Poll, RawWaker, RawWakerVTable, Waker}; +use core::task::{Poll, RawWaker, RawWakerVTable, Waker}; #[test] fn poll_const() { @@ -21,9 +21,5 @@ fn waker_const() { static WAKER: Waker = unsafe { Waker::from_raw(VOID_WAKER) }; - static CONTEXT: Context<'static> = Context::from_waker(&WAKER); - - static WAKER_REF: &'static Waker = CONTEXT.waker(); - - WAKER_REF.wake_by_ref(); + WAKER.wake_by_ref(); } diff --git a/library/core/tests/time.rs b/library/core/tests/time.rs index a05128de4..2975c81f8 100644 --- a/library/core/tests/time.rs +++ b/library/core/tests/time.rs @@ -173,6 +173,32 @@ fn div() { assert_eq!(Duration::new(99, 999_999_000) / 100, Duration::new(0, 999_999_990)); } +#[test] +fn div_duration_f32() { + assert_eq!(Duration::ZERO.div_duration_f32(Duration::MAX), 0.0); + assert_eq!(Duration::MAX.div_duration_f32(Duration::ZERO), f32::INFINITY); + assert_eq!((Duration::SECOND * 2).div_duration_f32(Duration::SECOND), 2.0); + assert!(Duration::ZERO.div_duration_f32(Duration::ZERO).is_nan()); + // These tests demonstrate it doesn't panic with extreme values. + // Accuracy of the computed value is not a huge concern, we know floats don't work well + // at these extremes. + assert!((Duration::MAX).div_duration_f32(Duration::NANOSECOND) > 10.0f32.powf(28.0)); + assert!((Duration::NANOSECOND).div_duration_f32(Duration::MAX) < 0.1); +} + +#[test] +fn div_duration_f64() { + assert_eq!(Duration::ZERO.div_duration_f64(Duration::MAX), 0.0); + assert_eq!(Duration::MAX.div_duration_f64(Duration::ZERO), f64::INFINITY); + assert_eq!((Duration::SECOND * 2).div_duration_f64(Duration::SECOND), 2.0); + assert!(Duration::ZERO.div_duration_f64(Duration::ZERO).is_nan()); + // These tests demonstrate it doesn't panic with extreme values. + // Accuracy of the computed value is not a huge concern, we know floats don't work well + // at these extremes. + assert!((Duration::MAX).div_duration_f64(Duration::NANOSECOND) > 10.0f64.powf(28.0)); + assert!((Duration::NANOSECOND).div_duration_f64(Duration::MAX) < 0.1); +} + #[test] fn checked_div() { assert_eq!(Duration::new(2, 0).checked_div(2), Some(Duration::new(1, 0))); diff --git a/library/panic_abort/src/lib.rs b/library/panic_abort/src/lib.rs index cba8ef25d..a3cebf99c 100644 --- a/library/panic_abort/src/lib.rs +++ b/library/panic_abort/src/lib.rs @@ -61,7 +61,7 @@ pub unsafe fn __rust_start_panic(_payload: *mut &mut dyn BoxMeUp) -> u32 { // // https://docs.microsoft.com/en-us/cpp/intrinsics/fastfail // - // Note: this is the same implementation as in libstd's `abort_internal` + // Note: this is the same implementation as in std's `abort_internal` unsafe fn abort() -> ! { #[allow(unused)] const FAST_FAIL_FATAL_APP_EXIT: usize = 7; @@ -89,7 +89,7 @@ pub unsafe fn __rust_start_panic(_payload: *mut &mut dyn BoxMeUp) -> u32 { // This... is a bit of an oddity. The tl;dr; is that this is required to link // correctly, the longer explanation is below. // -// Right now the binaries of libcore/libstd that we ship are all compiled with +// Right now the binaries of core/std that we ship are all compiled with // `-C panic=unwind`. This is done to ensure that the binaries are maximally // compatible with as many situations as possible. The compiler, however, // requires a "personality function" for all functions compiled with `-C @@ -109,7 +109,7 @@ pub unsafe fn __rust_start_panic(_payload: *mut &mut dyn BoxMeUp) -> u32 { // library just defines this symbol so there's at least some personality // somewhere. // -// Essentially this symbol is just defined to get wired up to libcore/libstd +// Essentially this symbol is just defined to get wired up to core/std // binaries, but it should never be called as we don't link in an unwinding // runtime at all. pub mod personalities { diff --git a/library/panic_unwind/src/lib.rs b/library/panic_unwind/src/lib.rs index 7e7180a38..ea3c9a7a6 100644 --- a/library/panic_unwind/src/lib.rs +++ b/library/panic_unwind/src/lib.rs @@ -82,11 +82,11 @@ cfg_if::cfg_if! { } extern "C" { - /// Handler in libstd called when a panic object is dropped outside of + /// Handler in std called when a panic object is dropped outside of /// `catch_unwind`. fn __rust_drop_panic() -> !; - /// Handler in libstd called when a foreign exception is caught. + /// Handler in std called when a foreign exception is caught. fn __rust_foreign_exception() -> !; } diff --git a/library/proc_macro/src/bridge/client.rs b/library/proc_macro/src/bridge/client.rs index 506b2a773..52a08cad9 100644 --- a/library/proc_macro/src/bridge/client.rs +++ b/library/proc_macro/src/bridge/client.rs @@ -356,7 +356,7 @@ impl Clone for Client { fn maybe_install_panic_hook(force_show_panics: bool) { // Hide the default panic output within `proc_macro` expansions. - // NB. the server can't do this because it may use a different libstd. + // NB. the server can't do this because it may use a different std. static HIDE_PANICS_DURING_EXPANSION: Once = Once::new(); HIDE_PANICS_DURING_EXPANSION.call_once(|| { let prev = panic::take_hook(); diff --git a/library/proc_macro/src/bridge/fxhash.rs b/library/proc_macro/src/bridge/fxhash.rs index 4b1e412e2..17bd0a1b3 100644 --- a/library/proc_macro/src/bridge/fxhash.rs +++ b/library/proc_macro/src/bridge/fxhash.rs @@ -15,9 +15,9 @@ use std::ops::BitXor; /// Type alias for a hashmap using the `fx` hash algorithm. pub type FxHashMap = HashMap>; -/// A speedy hash algorithm for use within rustc. The hashmap in liballoc -/// by default uses SipHash which isn't quite as speedy as we want. In the -/// compiler we're not really worried about DOS attempts, so we use a fast +/// A speedy hash algorithm for use within rustc. The hashmap in alloc by +/// default uses SipHash which isn't quite as speedy as we want. In the compiler +/// we're not really worried about DOS attempts, so we use a fast /// non-cryptographic hash. /// /// This is the same as the algorithm used by Firefox -- which is a homespun diff --git a/library/proc_macro/src/bridge/server.rs b/library/proc_macro/src/bridge/server.rs index 8202c40d6..2ea87d866 100644 --- a/library/proc_macro/src/bridge/server.rs +++ b/library/proc_macro/src/bridge/server.rs @@ -112,7 +112,7 @@ macro_rules! define_dispatcher_impl { $name::$method(server, $($arg),*) }; // HACK(eddyb) don't use `panic::catch_unwind` in a panic. - // If client and server happen to use the same `libstd`, + // If client and server happen to use the same `std`, // `catch_unwind` asserts that the panic counter was 0, // even when the closure passed to it didn't panic. let r = if thread::panicking() { diff --git a/library/proc_macro/src/lib.rs b/library/proc_macro/src/lib.rs index 0d3fc2c52..f0e4f5d8a 100644 --- a/library/proc_macro/src/lib.rs +++ b/library/proc_macro/src/lib.rs @@ -1493,7 +1493,7 @@ pub mod tracked_env { use std::ffi::OsStr; /// Retrieve an environment variable and add it to build dependency info. - /// Build system executing the compiler will know that the variable was accessed during + /// The build system executing the compiler will know that the variable was accessed during /// compilation, and will be able to rerun the build when the value of that variable changes. /// Besides the dependency tracking this function should be equivalent to `env::var` from the /// standard library, except that the argument must be UTF-8. diff --git a/library/rustc-std-workspace-alloc/lib.rs b/library/rustc-std-workspace-alloc/lib.rs index c38a8d2f2..87db7af44 100644 --- a/library/rustc-std-workspace-alloc/lib.rs +++ b/library/rustc-std-workspace-alloc/lib.rs @@ -3,7 +3,7 @@ // See rustc-std-workspace-core for why this crate is needed. -// Rename the crate to avoid conflicting with the alloc module in liballoc. +// Rename the crate to avoid conflicting with the alloc module in alloc. extern crate alloc as foo; pub use foo::*; diff --git a/library/std/Cargo.toml b/library/std/Cargo.toml index a7aefc26b..adf521d9b 100644 --- a/library/std/Cargo.toml +++ b/library/std/Cargo.toml @@ -16,7 +16,7 @@ panic_unwind = { path = "../panic_unwind", optional = true } panic_abort = { path = "../panic_abort" } core = { path = "../core" } libc = { version = "0.2.138", default-features = false, features = ['rustc-dep-of-std'] } -compiler_builtins = { version = "0.1.82" } +compiler_builtins = { version = "0.1.85" } profiler_builtins = { path = "../profiler_builtins", optional = true } unwind = { path = "../unwind" } hashbrown = { version = "0.12", default-features = false, features = ['rustc-dep-of-std'] } @@ -33,7 +33,8 @@ default-features = false features = ['read_core', 'elf', 'macho', 'pe', 'unaligned', 'archive'] [dev-dependencies] -rand = "0.7" +rand = { version = "0.8.5", default-features = false, features = ["alloc"] } +rand_xorshift = "0.3.0" [target.'cfg(any(all(target_family = "wasm", not(target_os = "emscripten")), all(target_vendor = "fortanix", target_env = "sgx")))'.dependencies] dlmalloc = { version = "0.2.3", features = ['rustc-dep-of-std'] } diff --git a/library/std/src/alloc.rs b/library/std/src/alloc.rs index 61c1ff578..c5a5991cc 100644 --- a/library/std/src/alloc.rs +++ b/library/std/src/alloc.rs @@ -338,7 +338,7 @@ fn default_alloc_error_hook(layout: Layout) { #[allow(unused_unsafe)] if unsafe { __rust_alloc_error_handler_should_panic != 0 } { - panic!("memory allocation of {} bytes failed\n", layout.size()); + panic!("memory allocation of {} bytes failed", layout.size()); } else { rtprintpanic!("memory allocation of {} bytes failed\n", layout.size()); } diff --git a/library/std/src/backtrace.rs b/library/std/src/backtrace.rs index 9cb74f951..7543ffadd 100644 --- a/library/std/src/backtrace.rs +++ b/library/std/src/backtrace.rs @@ -23,10 +23,10 @@ //! //! ## Platform support //! -//! Not all platforms that libstd compiles for support capturing backtraces. -//! Some platforms simply do nothing when capturing a backtrace. To check -//! whether the platform supports capturing backtraces you can consult the -//! `BacktraceStatus` enum as a result of `Backtrace::status`. +//! Not all platforms that std compiles for support capturing backtraces. Some +//! platforms simply do nothing when capturing a backtrace. To check whether the +//! platform supports capturing backtraces you can consult the `BacktraceStatus` +//! enum as a result of `Backtrace::status`. //! //! Like above with accuracy platform support is done on a best effort basis. //! Sometimes libraries might not be available at runtime or something may go diff --git a/library/std/src/collections/hash/map/tests.rs b/library/std/src/collections/hash/map/tests.rs index 65634f206..6b89518e2 100644 --- a/library/std/src/collections/hash/map/tests.rs +++ b/library/std/src/collections/hash/map/tests.rs @@ -3,7 +3,8 @@ use super::HashMap; use super::RandomState; use crate::assert_matches::assert_matches; use crate::cell::RefCell; -use rand::{thread_rng, Rng}; +use crate::test_helpers::test_rng; +use rand::Rng; use realstd::collections::TryReserveErrorKind::*; // https://github.com/rust-lang/rust/issues/62301 @@ -710,16 +711,16 @@ fn test_entry_take_doesnt_corrupt() { } let mut m = HashMap::new(); - let mut rng = thread_rng(); + let mut rng = test_rng(); // Populate the map with some items. for _ in 0..50 { - let x = rng.gen_range(-10, 10); + let x = rng.gen_range(-10..10); m.insert(x, ()); } for _ in 0..1000 { - let x = rng.gen_range(-10, 10); + let x = rng.gen_range(-10..10); match m.entry(x) { Vacant(_) => {} Occupied(e) => { diff --git a/library/std/src/collections/hash/set.rs b/library/std/src/collections/hash/set.rs index cee884145..b59f89d32 100644 --- a/library/std/src/collections/hash/set.rs +++ b/library/std/src/collections/hash/set.rs @@ -317,7 +317,7 @@ impl HashSet { /// /// let mut set = HashSet::from([1, 2, 3, 4, 5, 6]); /// set.retain(|&k| k % 2 == 0); - /// assert_eq!(set.len(), 3); + /// assert_eq!(set, HashSet::from([2, 4, 6])); /// ``` /// /// # Performance diff --git a/library/std/src/env.rs b/library/std/src/env.rs index 6eb7cbea6..183f9ab3b 100644 --- a/library/std/src/env.rs +++ b/library/std/src/env.rs @@ -570,6 +570,13 @@ impl Error for JoinPathsError { /// /// [msdn]: https://docs.microsoft.com/en-us/windows/win32/api/userenv/nf-userenv-getuserprofiledirectorya /// +/// # Deprecation +/// +/// This function is deprecated because the behaviour on Windows is not correct. +/// The 'HOME' environment variable is not standard on Windows, and may not produce +/// desired results; for instance, under Cygwin or Mingw it will return `/home/you` +/// when it should return `C:\Users\you`. +/// /// # Examples /// /// ``` @@ -582,7 +589,7 @@ impl Error for JoinPathsError { /// ``` #[deprecated( since = "1.29.0", - note = "This function's behavior is unexpected and probably not what you want. \ + note = "This function's behavior may be unexpected on Windows. \ Consider using a crate from crates.io instead." )] #[must_use] diff --git a/library/std/src/fs.rs b/library/std/src/fs.rs index f357d505f..286ad68fd 100644 --- a/library/std/src/fs.rs +++ b/library/std/src/fs.rs @@ -249,8 +249,9 @@ pub struct DirBuilder { pub fn read>(path: P) -> io::Result> { fn inner(path: &Path) -> io::Result> { let mut file = File::open(path)?; - let mut bytes = Vec::new(); - file.read_to_end(&mut bytes)?; + let size = file.metadata().map(|m| m.len()).unwrap_or(0); + let mut bytes = Vec::with_capacity(size as usize); + io::default_read_to_end(&mut file, &mut bytes)?; Ok(bytes) } inner(path.as_ref()) @@ -288,8 +289,9 @@ pub fn read>(path: P) -> io::Result> { pub fn read_to_string>(path: P) -> io::Result { fn inner(path: &Path) -> io::Result { let mut file = File::open(path)?; - let mut string = String::new(); - file.read_to_string(&mut string)?; + let size = file.metadata().map(|m| m.len()).unwrap_or(0); + let mut string = String::with_capacity(size as usize); + io::default_read_to_string(&mut file, &mut string)?; Ok(string) } inner(path.as_ref()) @@ -1510,7 +1512,7 @@ impl FileType { } /// Tests whether this file type represents a regular file. - /// The result is mutually exclusive to the results of + /// The result is mutually exclusive to the results of /// [`is_dir`] and [`is_symlink`]; only zero or one of these /// tests may pass. /// diff --git a/library/std/src/fs/tests.rs b/library/std/src/fs/tests.rs index b8959316d..eb582be01 100644 --- a/library/std/src/fs/tests.rs +++ b/library/std/src/fs/tests.rs @@ -10,7 +10,7 @@ use crate::sys_common::io::test::{tmpdir, TempDir}; use crate::thread; use crate::time::{Duration, Instant}; -use rand::{rngs::StdRng, RngCore, SeedableRng}; +use rand::RngCore; #[cfg(unix)] use crate::os::unix::fs::symlink as symlink_dir; @@ -1181,7 +1181,7 @@ fn _assert_send_sync() { #[test] fn binary_file() { let mut bytes = [0; 1024]; - StdRng::from_entropy().fill_bytes(&mut bytes); + crate::test_helpers::test_rng().fill_bytes(&mut bytes); let tmpdir = tmpdir(); @@ -1194,7 +1194,7 @@ fn binary_file() { #[test] fn write_then_read() { let mut bytes = [0; 1024]; - StdRng::from_entropy().fill_bytes(&mut bytes); + crate::test_helpers::test_rng().fill_bytes(&mut bytes); let tmpdir = tmpdir(); @@ -1551,3 +1551,47 @@ fn hiberfil_sys() { fs::metadata(hiberfil).unwrap(); assert_eq!(true, hiberfil.exists()); } + +/// Test that two different ways of obtaining the FileType give the same result. +/// Cf. https://github.com/rust-lang/rust/issues/104900 +#[test] +fn test_eq_direntry_metadata() { + let tmpdir = tmpdir(); + let file_path = tmpdir.join("file"); + File::create(file_path).unwrap(); + for e in fs::read_dir(tmpdir.path()).unwrap() { + let e = e.unwrap(); + let p = e.path(); + let ft1 = e.file_type().unwrap(); + let ft2 = p.metadata().unwrap().file_type(); + assert_eq!(ft1, ft2); + } +} + +/// Regression test for https://github.com/rust-lang/rust/issues/50619. +#[test] +#[cfg(target_os = "linux")] +fn test_read_dir_infinite_loop() { + use crate::io::ErrorKind; + use crate::process::Command; + + // Create a zombie child process + let Ok(mut child) = Command::new("echo").spawn() else { return }; + + // Make sure the process is (un)dead + match child.kill() { + // InvalidInput means the child already exited + Err(e) if e.kind() != ErrorKind::InvalidInput => return, + _ => {} + } + + // open() on this path will succeed, but readdir() will fail + let id = child.id(); + let path = format!("/proc/{id}/net"); + + // Skip the test if we can't open the directory in the first place + let Ok(dir) = fs::read_dir(path) else { return }; + + // Check for duplicate errors + assert!(dir.filter(|e| e.is_err()).take(2).count() < 2); +} diff --git a/library/std/src/io/buffered/tests.rs b/library/std/src/io/buffered/tests.rs index f4e688eb9..4c1b7d576 100644 --- a/library/std/src/io/buffered/tests.rs +++ b/library/std/src/io/buffered/tests.rs @@ -288,8 +288,8 @@ fn test_buffered_reader_seek_underflow_discard_buffer_between_seeks() { let mut reader = BufReader::with_capacity(5, ErrAfterFirstSeekReader { first_seek: true }); assert_eq!(reader.fill_buf().ok(), Some(&[0, 0, 0, 0, 0][..])); - // The following seek will require two underlying seeks. The first will - // succeed but the second will fail. This should still invalidate the + // The following seek will require two underlying seeks. The first will + // succeed but the second will fail. This should still invalidate the // buffer. assert!(reader.seek(SeekFrom::Current(i64::MIN)).is_err()); assert_eq!(reader.buffer().len(), 0); diff --git a/library/std/src/io/error/repr_bitpacked.rs b/library/std/src/io/error/repr_bitpacked.rs index 781ae03ad..358148405 100644 --- a/library/std/src/io/error/repr_bitpacked.rs +++ b/library/std/src/io/error/repr_bitpacked.rs @@ -166,7 +166,7 @@ impl Repr { // `new_unchecked` is safe. let res = Self(unsafe { NonNull::new_unchecked(tagged) }, PhantomData); // quickly smoke-check we encoded the right thing (This generally will - // only run in libstd's tests, unless the user uses -Zbuild-std) + // only run in std's tests, unless the user uses -Zbuild-std) debug_assert!(matches!(res.data(), ErrorData::Custom(_)), "repr(custom) encoding failed"); res } @@ -177,7 +177,7 @@ impl Repr { // Safety: `TAG_OS` is not zero, so the result of the `|` is not 0. let res = Self(unsafe { NonNull::new_unchecked(ptr::invalid_mut(utagged)) }, PhantomData); // quickly smoke-check we encoded the right thing (This generally will - // only run in libstd's tests, unless the user uses -Zbuild-std) + // only run in std's tests, unless the user uses -Zbuild-std) debug_assert!( matches!(res.data(), ErrorData::Os(c) if c == code), "repr(os) encoding failed for {code}" @@ -191,7 +191,7 @@ impl Repr { // Safety: `TAG_SIMPLE` is not zero, so the result of the `|` is not 0. let res = Self(unsafe { NonNull::new_unchecked(ptr::invalid_mut(utagged)) }, PhantomData); // quickly smoke-check we encoded the right thing (This generally will - // only run in libstd's tests, unless the user uses -Zbuild-std) + // only run in std's tests, unless the user uses -Zbuild-std) debug_assert!( matches!(res.data(), ErrorData::Simple(k) if k == kind), "repr(simple) encoding failed {:?}", @@ -348,7 +348,7 @@ fn kind_from_prim(ek: u32) -> Option { // that our encoding relies on for correctness and soundness. (Some of these are // a bit overly thorough/cautious, admittedly) // -// If any of these are hit on a platform that libstd supports, we should likely +// If any of these are hit on a platform that std supports, we should likely // just use `repr_unpacked.rs` there instead (unless the fix is easy). macro_rules! static_assert { ($condition:expr) => { @@ -374,10 +374,10 @@ static_assert!((TAG_MASK + 1).is_power_of_two()); static_assert!(align_of::() >= TAG_MASK + 1); static_assert!(align_of::() >= TAG_MASK + 1); -static_assert!(@usize_eq: (TAG_MASK & TAG_SIMPLE_MESSAGE), TAG_SIMPLE_MESSAGE); -static_assert!(@usize_eq: (TAG_MASK & TAG_CUSTOM), TAG_CUSTOM); -static_assert!(@usize_eq: (TAG_MASK & TAG_OS), TAG_OS); -static_assert!(@usize_eq: (TAG_MASK & TAG_SIMPLE), TAG_SIMPLE); +static_assert!(@usize_eq: TAG_MASK & TAG_SIMPLE_MESSAGE, TAG_SIMPLE_MESSAGE); +static_assert!(@usize_eq: TAG_MASK & TAG_CUSTOM, TAG_CUSTOM); +static_assert!(@usize_eq: TAG_MASK & TAG_OS, TAG_OS); +static_assert!(@usize_eq: TAG_MASK & TAG_SIMPLE, TAG_SIMPLE); // This is obviously true (`TAG_CUSTOM` is `0b01`), but in `Repr::new_custom` we // offset a pointer by this value, and expect it to both be within the same diff --git a/library/std/src/io/mod.rs b/library/std/src/io/mod.rs index 23a13523f..de528e853 100644 --- a/library/std/src/io/mod.rs +++ b/library/std/src/io/mod.rs @@ -2137,8 +2137,10 @@ pub trait BufRead: Read { } /// Read all bytes until a newline (the `0xA` byte) is reached, and append - /// them to the provided buffer. You do not need to clear the buffer before - /// appending. + /// them to the provided `String` buffer. + /// + /// Previous content of the buffer will be preserved. To avoid appending to + /// the buffer, you need to [`clear`] it first. /// /// This function will read bytes from the underlying stream until the /// newline delimiter (the `0xA` byte) or EOF is found. Once found, all bytes @@ -2151,9 +2153,11 @@ pub trait BufRead: Read { /// /// This function is blocking and should be used carefully: it is possible for /// an attacker to continuously send bytes without ever sending a newline - /// or EOF. + /// or EOF. You can use [`take`] to limit the maximum number of bytes read. /// /// [`Ok(0)`]: Ok + /// [`clear`]: String::clear + /// [`take`]: crate::io::Read::take /// /// # Errors /// diff --git a/library/std/src/io/stdio.rs b/library/std/src/io/stdio.rs index 1141a957d..14bfef4c7 100644 --- a/library/std/src/io/stdio.rs +++ b/library/std/src/io/stdio.rs @@ -10,9 +10,8 @@ use crate::fmt; use crate::fs::File; use crate::io::{self, BufReader, IoSlice, IoSliceMut, LineWriter, Lines}; use crate::sync::atomic::{AtomicBool, Ordering}; -use crate::sync::{Arc, Mutex, MutexGuard, OnceLock}; +use crate::sync::{Arc, Mutex, MutexGuard, OnceLock, ReentrantMutex, ReentrantMutexGuard}; use crate::sys::stdio; -use crate::sys_common::remutex::{ReentrantMutex, ReentrantMutexGuard}; type LocalStream = Arc>>; diff --git a/library/std/src/lib.rs b/library/std/src/lib.rs index 65d4c3c89..a7e13f5b8 100644 --- a/library/std/src/lib.rs +++ b/library/std/src/lib.rs @@ -14,7 +14,7 @@ //! # How to read this documentation //! //! If you already know the name of what you are looking for, the fastest way to -//! find it is to use the search +//! find it is to use the search //! bar at the top of the page. //! //! Otherwise, you may want to jump to one of these useful sections: @@ -202,7 +202,7 @@ no_global_oom_handling, not(no_global_oom_handling) ))] -// To run libstd tests without x.py without ending up with two copies of libstd, Miri needs to be +// To run std tests without x.py without ending up with two copies of std, Miri needs to be // able to "empty" this crate. See . // rustc itself never sets the feature, so this line has no affect there. #![cfg(any(not(feature = "miri-test-libstd"), test, doctest))] @@ -532,7 +532,7 @@ pub mod process; pub mod sync; pub mod time; -// Pull in `std_float` crate into libstd. The contents of +// Pull in `std_float` crate into std. The contents of // `std_float` are in a different repository: rust-lang/portable-simd. #[path = "../../portable-simd/crates/std_float/src/lib.rs"] #[allow(missing_debug_implementations, dead_code, unsafe_op_in_unsafe_fn, unused_unsafe)] @@ -602,7 +602,7 @@ mod personality; #[allow(dead_code, unused_attributes, fuzzy_provenance_casts)] mod backtrace_rs; -// Re-export macros defined in libcore. +// Re-export macros defined in core. #[stable(feature = "rust1", since = "1.0.0")] #[allow(deprecated, deprecated_in_future)] pub use core::{ @@ -610,7 +610,7 @@ pub use core::{ unimplemented, unreachable, write, writeln, }; -// Re-export built-in macros defined through libcore. +// Re-export built-in macros defined through core. #[stable(feature = "builtin_macro_prelude", since = "1.38.0")] #[allow(deprecated)] pub use core::{ @@ -652,3 +652,30 @@ mod sealed { #[unstable(feature = "sealed", issue = "none")] pub trait Sealed {} } + +#[cfg(test)] +#[allow(dead_code)] // Not used in all configurations. +pub(crate) mod test_helpers { + /// Test-only replacement for `rand::thread_rng()`, which is unusable for + /// us, as we want to allow running stdlib tests on tier-3 targets which may + /// not have `getrandom` support. + /// + /// Does a bit of a song and dance to ensure that the seed is different on + /// each call (as some tests sadly rely on this), but doesn't try that hard. + /// + /// This is duplicated in the `core`, `alloc` test suites (as well as + /// `std`'s integration tests), but figuring out a mechanism to share these + /// seems far more painful than copy-pasting a 7 line function a couple + /// times, given that even under a perma-unstable feature, I don't think we + /// want to expose types from `rand` from `std`. + #[track_caller] + pub(crate) fn test_rng() -> rand_xorshift::XorShiftRng { + use core::hash::{BuildHasher, Hash, Hasher}; + let mut hasher = crate::collections::hash_map::RandomState::new().build_hasher(); + core::panic::Location::caller().hash(&mut hasher); + let hc64 = hasher.finish(); + let seed_vec = hc64.to_le_bytes().into_iter().chain(0u8..8).collect::>(); + let seed: [u8; 16] = seed_vec.as_slice().try_into().unwrap(); + rand::SeedableRng::from_seed(seed) + } +} diff --git a/library/std/src/macros.rs b/library/std/src/macros.rs index 6e4ba1404..fcc5cfafd 100644 --- a/library/std/src/macros.rs +++ b/library/std/src/macros.rs @@ -3,6 +3,7 @@ //! This module contains a set of macros which are exported from the standard //! library. Each macro is available for use when linking against the standard //! library. +// ignore-tidy-dbg #[doc = include_str!("../../core/src/macros/panic.md")] #[macro_export] diff --git a/library/std/src/net/ip_addr.rs b/library/std/src/net/ip_addr.rs index 5453853e1..07f08c1b5 100644 --- a/library/std/src/net/ip_addr.rs +++ b/library/std/src/net/ip_addr.rs @@ -1195,6 +1195,9 @@ impl Ipv6Addr { /// An IPv6 address representing localhost: `::1`. /// + /// This corresponds to constant `IN6ADDR_LOOPBACK_INIT` or `in6addr_loopback` in other + /// languages. + /// /// # Examples /// /// ``` @@ -1203,11 +1206,15 @@ impl Ipv6Addr { /// let addr = Ipv6Addr::LOCALHOST; /// assert_eq!(addr, Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 1)); /// ``` + #[doc(alias = "IN6ADDR_LOOPBACK_INIT")] + #[doc(alias = "in6addr_loopback")] #[stable(feature = "ip_constructors", since = "1.30.0")] pub const LOCALHOST: Self = Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 1); /// An IPv6 address representing the unspecified address: `::` /// + /// This corresponds to constant `IN6ADDR_ANY_INIT` or `in6addr_any` in other languages. + /// /// # Examples /// /// ``` @@ -1216,6 +1223,8 @@ impl Ipv6Addr { /// let addr = Ipv6Addr::UNSPECIFIED; /// assert_eq!(addr, Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 0)); /// ``` + #[doc(alias = "IN6ADDR_ANY_INIT")] + #[doc(alias = "in6addr_any")] #[stable(feature = "ip_constructors", since = "1.30.0")] pub const UNSPECIFIED: Self = Ipv6Addr::new(0, 0, 0, 0, 0, 0, 0, 0); diff --git a/library/std/src/os/fd/mod.rs b/library/std/src/os/fd/mod.rs index c6aa7c77d..35de4860f 100644 --- a/library/std/src/os/fd/mod.rs +++ b/library/std/src/os/fd/mod.rs @@ -3,7 +3,7 @@ //! This module is supported on Unix platforms and WASI, which both use a //! similar file descriptor system for referencing OS resources. -#![stable(feature = "io_safety", since = "1.63.0")] +#![stable(feature = "os_fd", since = "1.66.0")] #![deny(unsafe_op_in_unsafe_fn)] // `RawFd`, `AsRawFd`, etc. @@ -19,7 +19,7 @@ mod net; mod tests; // Export the types and traits for the public API. -#[unstable(feature = "os_fd", issue = "98699")] +#[stable(feature = "os_fd", since = "1.66.0")] pub use owned::*; -#[unstable(feature = "os_fd", issue = "98699")] +#[stable(feature = "os_fd", since = "1.66.0")] pub use raw::*; diff --git a/library/std/src/os/fd/owned.rs b/library/std/src/os/fd/owned.rs index c16518577..c41e093a7 100644 --- a/library/std/src/os/fd/owned.rs +++ b/library/std/src/os/fd/owned.rs @@ -100,7 +100,7 @@ impl BorrowedFd<'_> { // For ESP-IDF, F_DUPFD is used instead, because the CLOEXEC semantics // will never be supported, as this is a bare metal framework with - // no capabilities for multi-process execution. While F_DUPFD is also + // no capabilities for multi-process execution. While F_DUPFD is also // not supported yet, it might be (currently it returns ENOSYS). #[cfg(target_os = "espidf")] let cmd = libc::F_DUPFD; diff --git a/library/std/src/os/net/linux_ext/addr.rs b/library/std/src/os/net/linux_ext/addr.rs index df3fc8e6a..85065984f 100644 --- a/library/std/src/os/net/linux_ext/addr.rs +++ b/library/std/src/os/net/linux_ext/addr.rs @@ -38,7 +38,7 @@ pub trait SocketAddrExt: Sealed { /// Ok(()) /// } /// ``` - fn from_abstract_name(name: &N) -> crate::io::Result + fn from_abstract_name(name: N) -> crate::io::Result where N: AsRef<[u8]>; diff --git a/library/std/src/os/unix/net/addr.rs b/library/std/src/os/unix/net/addr.rs index 81ac829d2..ece2b33bd 100644 --- a/library/std/src/os/unix/net/addr.rs +++ b/library/std/src/os/unix/net/addr.rs @@ -256,7 +256,7 @@ impl linux_ext::addr::SocketAddrExt for SocketAddr { if let AddressKind::Abstract(name) = self.address() { Some(name) } else { None } } - fn from_abstract_name(name: &N) -> crate::io::Result + fn from_abstract_name(name: N) -> crate::io::Result where N: AsRef<[u8]>, { diff --git a/library/std/src/panic.rs b/library/std/src/panic.rs index c4f022de0..9fa8f5702 100644 --- a/library/std/src/panic.rs +++ b/library/std/src/panic.rs @@ -114,6 +114,9 @@ where /// aborting the process as well. This function *only* catches unwinding panics, /// not those that abort the process. /// +/// Note that if a custom panic hook has been set, it will be invoked before +/// the panic is caught, before unwinding. +/// /// Also note that unwinding into Rust code with a foreign exception (e.g. /// an exception thrown from C++ code) is undefined behavior. /// diff --git a/library/std/src/panicking.rs b/library/std/src/panicking.rs index 1039835bb..b0db3112e 100644 --- a/library/std/src/panicking.rs +++ b/library/std/src/panicking.rs @@ -306,11 +306,11 @@ pub mod panic_count { // and after increase and decrease, but not necessarily during their execution. // // Additionally, the top bit of GLOBAL_PANIC_COUNT (GLOBAL_ALWAYS_ABORT_FLAG) - // records whether panic::always_abort() has been called. This can only be + // records whether panic::always_abort() has been called. This can only be // set, never cleared. // panic::always_abort() is usually called to prevent memory allocations done by // the panic handling in the child created by `libc::fork`. - // Memory allocations performed in a child created with `libc::fork` are undefined + // Memory allocations performed in a child created with `libc::fork` are undefined // behavior in most operating systems. // Accessing LOCAL_PANIC_COUNT in a child created by `libc::fork` would lead to a memory // allocation. Only GLOBAL_PANIC_COUNT can be accessed in this situation. This is @@ -517,7 +517,7 @@ pub fn panicking() -> bool { !panic_count::count_is_zero() } -/// Entry point of panics from the libcore crate (`panic_impl` lang item). +/// Entry point of panics from the core crate (`panic_impl` lang item). #[cfg(not(test))] #[panic_handler] pub fn begin_panic_handler(info: &PanicInfo<'_>) -> ! { @@ -699,7 +699,11 @@ fn rust_panic_with_hook( // have limited options. Currently our preference is to // just abort. In the future we may consider resuming // unwinding or otherwise exiting the thread cleanly. - rtprintpanic!("thread panicked while panicking. aborting.\n"); + if !can_unwind { + rtprintpanic!("thread caused non-unwinding panic. aborting.\n"); + } else { + rtprintpanic!("thread panicked while panicking. aborting.\n"); + } crate::sys::abort_internal(); } diff --git a/library/std/src/path.rs b/library/std/src/path.rs index 6c957c2fa..4778114b4 100644 --- a/library/std/src/path.rs +++ b/library/std/src/path.rs @@ -271,7 +271,7 @@ pub const MAIN_SEPARATOR: char = crate::sys::path::MAIN_SEP; /// The primary separator of path components for the current platform. /// /// For example, `/` on Unix and `\` on Windows. -#[unstable(feature = "main_separator_str", issue = "94071")] +#[stable(feature = "main_separator_str", since = "1.68.0")] pub const MAIN_SEPARATOR_STR: &str = crate::sys::path::MAIN_SEP_STR; //////////////////////////////////////////////////////////////////////////////// @@ -306,7 +306,7 @@ unsafe fn u8_slice_as_os_str(s: &[u8]) -> &OsStr { // This casts are safe as OsStr is internally a wrapper around [u8] on all // platforms. // - // Note that currently this relies on the special knowledge that libstd has; + // Note that currently this relies on the special knowledge that std has; // these types are single-element structs but are not marked // repr(transparent) or repr(C) which would make these casts not allowable // outside std. @@ -607,7 +607,7 @@ pub struct Components<'a> { // true if path *physically* has a root separator; for most Windows // prefixes, it may have a "logical" root separator for the purposes of - // normalization, e.g., \\server\share == \\server\share\. + // normalization, e.g., \\server\share == \\server\share\. has_physical_root: bool, // The iterator is double-ended, and these two states keep track of what has @@ -1246,6 +1246,9 @@ impl PathBuf { /// and `path` is not empty, the new path is normalized: all references /// to `.` and `..` are removed. /// + /// Consider using [`Path::join`] if you need a new `PathBuf` instead of + /// using this function on a cloned `PathBuf`. + /// /// # Examples /// /// Pushing a relative path extends the existing path: @@ -1411,7 +1414,8 @@ impl PathBuf { self.push(file_name); } - /// Updates [`self.extension`] to `extension`. + /// Updates [`self.extension`] to `Some(extension)` or to `None` if + /// `extension` is empty. /// /// Returns `false` and does nothing if [`self.file_name`] is [`None`], /// returns `true` and updates the extension otherwise. @@ -1419,6 +1423,20 @@ impl PathBuf { /// If [`self.extension`] is [`None`], the extension is added; otherwise /// it is replaced. /// + /// If `extension` is the empty string, [`self.extension`] will be [`None`] + /// afterwards, not `Some("")`. + /// + /// # Caveats + /// + /// The new `extension` may contain dots and will be used in its entirety, + /// but only the part after the final dot will be reflected in + /// [`self.extension`]. + /// + /// If the file stem contains internal dots and `extension` is empty, part + /// of the old file stem will be considered the new [`self.extension`]. + /// + /// See the examples below. + /// /// [`self.file_name`]: Path::file_name /// [`self.extension`]: Path::extension /// @@ -1432,8 +1450,20 @@ impl PathBuf { /// p.set_extension("force"); /// assert_eq!(Path::new("/feel/the.force"), p.as_path()); /// - /// p.set_extension("dark_side"); - /// assert_eq!(Path::new("/feel/the.dark_side"), p.as_path()); + /// p.set_extension("dark.side"); + /// assert_eq!(Path::new("/feel/the.dark.side"), p.as_path()); + /// + /// p.set_extension("cookie"); + /// assert_eq!(Path::new("/feel/the.dark.cookie"), p.as_path()); + /// + /// p.set_extension(""); + /// assert_eq!(Path::new("/feel/the.dark"), p.as_path()); + /// + /// p.set_extension(""); + /// assert_eq!(Path::new("/feel/the"), p.as_path()); + /// + /// p.set_extension(""); + /// assert_eq!(Path::new("/feel/the"), p.as_path()); /// ``` #[stable(feature = "rust1", since = "1.0.0")] pub fn set_extension>(&mut self, extension: S) -> bool { @@ -1748,6 +1778,14 @@ impl ops::Deref for PathBuf { } } +#[stable(feature = "path_buf_deref_mut", since = "1.68.0")] +impl ops::DerefMut for PathBuf { + #[inline] + fn deref_mut(&mut self) -> &mut Path { + Path::from_inner_mut(&mut self.inner) + } +} + #[stable(feature = "rust1", since = "1.0.0")] impl Borrow for PathBuf { #[inline] @@ -2000,6 +2038,12 @@ impl Path { unsafe { &*(s.as_ref() as *const OsStr as *const Path) } } + fn from_inner_mut(inner: &mut OsStr) -> &mut Path { + // SAFETY: Path is just a wrapper around OsStr, + // therefore converting &mut OsStr to &mut Path is safe. + unsafe { &mut *(inner as *mut OsStr as *mut Path) } + } + /// Yields the underlying [`OsStr`] slice. /// /// # Examples @@ -2025,12 +2069,12 @@ impl Path { /// #![feature(path_as_mut_os_str)] /// use std::path::{Path, PathBuf}; /// - /// let mut path = PathBuf::from("/Foo.TXT").into_boxed_path(); + /// let mut path = PathBuf::from("Foo.TXT"); /// - /// assert_ne!(&*path, Path::new("/foo.txt")); + /// assert_ne!(path, Path::new("foo.txt")); /// /// path.as_mut_os_str().make_ascii_lowercase(); - /// assert_eq!(&*path, Path::new("/foo.txt")); + /// assert_eq!(path, Path::new("foo.txt")); /// ``` #[unstable(feature = "path_as_mut_os_str", issue = "105021")] #[must_use] @@ -3133,9 +3177,9 @@ impl<'a> IntoIterator for &'a Path { } macro_rules! impl_cmp { - ($lhs:ty, $rhs: ty) => { + (<$($life:lifetime),*> $lhs:ty, $rhs: ty) => { #[stable(feature = "partialeq_path", since = "1.6.0")] - impl<'a, 'b> PartialEq<$rhs> for $lhs { + impl<$($life),*> PartialEq<$rhs> for $lhs { #[inline] fn eq(&self, other: &$rhs) -> bool { ::eq(self, other) @@ -3143,7 +3187,7 @@ macro_rules! impl_cmp { } #[stable(feature = "partialeq_path", since = "1.6.0")] - impl<'a, 'b> PartialEq<$lhs> for $rhs { + impl<$($life),*> PartialEq<$lhs> for $rhs { #[inline] fn eq(&self, other: &$lhs) -> bool { ::eq(self, other) @@ -3151,7 +3195,7 @@ macro_rules! impl_cmp { } #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialOrd<$rhs> for $lhs { + impl<$($life),*> PartialOrd<$rhs> for $lhs { #[inline] fn partial_cmp(&self, other: &$rhs) -> Option { ::partial_cmp(self, other) @@ -3159,7 +3203,7 @@ macro_rules! impl_cmp { } #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialOrd<$lhs> for $rhs { + impl<$($life),*> PartialOrd<$lhs> for $rhs { #[inline] fn partial_cmp(&self, other: &$lhs) -> Option { ::partial_cmp(self, other) @@ -3168,16 +3212,16 @@ macro_rules! impl_cmp { }; } -impl_cmp!(PathBuf, Path); -impl_cmp!(PathBuf, &'a Path); -impl_cmp!(Cow<'a, Path>, Path); -impl_cmp!(Cow<'a, Path>, &'b Path); -impl_cmp!(Cow<'a, Path>, PathBuf); +impl_cmp!(<> PathBuf, Path); +impl_cmp!(<'a> PathBuf, &'a Path); +impl_cmp!(<'a> Cow<'a, Path>, Path); +impl_cmp!(<'a, 'b> Cow<'a, Path>, &'b Path); +impl_cmp!(<'a> Cow<'a, Path>, PathBuf); macro_rules! impl_cmp_os_str { - ($lhs:ty, $rhs: ty) => { + (<$($life:lifetime),*> $lhs:ty, $rhs: ty) => { #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialEq<$rhs> for $lhs { + impl<$($life),*> PartialEq<$rhs> for $lhs { #[inline] fn eq(&self, other: &$rhs) -> bool { ::eq(self, other.as_ref()) @@ -3185,7 +3229,7 @@ macro_rules! impl_cmp_os_str { } #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialEq<$lhs> for $rhs { + impl<$($life),*> PartialEq<$lhs> for $rhs { #[inline] fn eq(&self, other: &$lhs) -> bool { ::eq(self.as_ref(), other) @@ -3193,7 +3237,7 @@ macro_rules! impl_cmp_os_str { } #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialOrd<$rhs> for $lhs { + impl<$($life),*> PartialOrd<$rhs> for $lhs { #[inline] fn partial_cmp(&self, other: &$rhs) -> Option { ::partial_cmp(self, other.as_ref()) @@ -3201,7 +3245,7 @@ macro_rules! impl_cmp_os_str { } #[stable(feature = "cmp_path", since = "1.8.0")] - impl<'a, 'b> PartialOrd<$lhs> for $rhs { + impl<$($life),*> PartialOrd<$lhs> for $rhs { #[inline] fn partial_cmp(&self, other: &$lhs) -> Option { ::partial_cmp(self.as_ref(), other) @@ -3210,20 +3254,20 @@ macro_rules! impl_cmp_os_str { }; } -impl_cmp_os_str!(PathBuf, OsStr); -impl_cmp_os_str!(PathBuf, &'a OsStr); -impl_cmp_os_str!(PathBuf, Cow<'a, OsStr>); -impl_cmp_os_str!(PathBuf, OsString); -impl_cmp_os_str!(Path, OsStr); -impl_cmp_os_str!(Path, &'a OsStr); -impl_cmp_os_str!(Path, Cow<'a, OsStr>); -impl_cmp_os_str!(Path, OsString); -impl_cmp_os_str!(&'a Path, OsStr); -impl_cmp_os_str!(&'a Path, Cow<'b, OsStr>); -impl_cmp_os_str!(&'a Path, OsString); -impl_cmp_os_str!(Cow<'a, Path>, OsStr); -impl_cmp_os_str!(Cow<'a, Path>, &'b OsStr); -impl_cmp_os_str!(Cow<'a, Path>, OsString); +impl_cmp_os_str!(<> PathBuf, OsStr); +impl_cmp_os_str!(<'a> PathBuf, &'a OsStr); +impl_cmp_os_str!(<'a> PathBuf, Cow<'a, OsStr>); +impl_cmp_os_str!(<> PathBuf, OsString); +impl_cmp_os_str!(<> Path, OsStr); +impl_cmp_os_str!(<'a> Path, &'a OsStr); +impl_cmp_os_str!(<'a> Path, Cow<'a, OsStr>); +impl_cmp_os_str!(<> Path, OsString); +impl_cmp_os_str!(<'a> &'a Path, OsStr); +impl_cmp_os_str!(<'a, 'b> &'a Path, Cow<'b, OsStr>); +impl_cmp_os_str!(<'a> &'a Path, OsString); +impl_cmp_os_str!(<'a> Cow<'a, Path>, OsStr); +impl_cmp_os_str!(<'a, 'b> Cow<'a, Path>, &'b OsStr); +impl_cmp_os_str!(<'a> Cow<'a, Path>, OsString); #[stable(since = "1.7.0", feature = "strip_prefix")] impl fmt::Display for StripPrefixError { diff --git a/library/std/src/personality/dwarf/eh.rs b/library/std/src/personality/dwarf/eh.rs index a783e1870..87585a8fc 100644 --- a/library/std/src/personality/dwarf/eh.rs +++ b/library/std/src/personality/dwarf/eh.rs @@ -84,7 +84,7 @@ pub unsafe fn find_eh_action(lsda: *const u8, context: &EHContext<'_>) -> Result let cs_start = read_encoded_pointer(&mut reader, context, call_site_encoding)?; let cs_len = read_encoded_pointer(&mut reader, context, call_site_encoding)?; let cs_lpad = read_encoded_pointer(&mut reader, context, call_site_encoding)?; - let cs_action = reader.read_uleb128(); + let cs_action_entry = reader.read_uleb128(); // Callsite table is sorted by cs_start, so if we've passed the ip, we // may stop searching. if ip < func_start + cs_start { @@ -95,7 +95,7 @@ pub unsafe fn find_eh_action(lsda: *const u8, context: &EHContext<'_>) -> Result return Ok(EHAction::None); } else { let lpad = lpad_base + cs_lpad; - return Ok(interpret_cs_action(cs_action, lpad)); + return Ok(interpret_cs_action(action_table as *mut u8, cs_action_entry, lpad)); } } } @@ -113,26 +113,39 @@ pub unsafe fn find_eh_action(lsda: *const u8, context: &EHContext<'_>) -> Result let mut idx = ip; loop { let cs_lpad = reader.read_uleb128(); - let cs_action = reader.read_uleb128(); + let cs_action_entry = reader.read_uleb128(); idx -= 1; if idx == 0 { // Can never have null landing pad for sjlj -- that would have // been indicated by a -1 call site index. let lpad = (cs_lpad + 1) as usize; - return Ok(interpret_cs_action(cs_action, lpad)); + return Ok(interpret_cs_action(action_table as *mut u8, cs_action_entry, lpad)); } } } } -fn interpret_cs_action(cs_action: u64, lpad: usize) -> EHAction { - if cs_action == 0 { - // If cs_action is 0 then this is a cleanup (Drop::drop). We run these +unsafe fn interpret_cs_action( + action_table: *mut u8, + cs_action_entry: u64, + lpad: usize, +) -> EHAction { + if cs_action_entry == 0 { + // If cs_action_entry is 0 then this is a cleanup (Drop::drop). We run these // for both Rust panics and foreign exceptions. EHAction::Cleanup(lpad) } else { - // Stop unwinding Rust panics at catch_unwind. - EHAction::Catch(lpad) + // If lpad != 0 and cs_action_entry != 0, we have to check ttype_index. + // If ttype_index == 0 under the condition, we take cleanup action. + let action_record = (action_table as *mut u8).offset(cs_action_entry as isize - 1); + let mut action_reader = DwarfReader::new(action_record); + let ttype_index = action_reader.read_sleb128(); + if ttype_index == 0 { + EHAction::Cleanup(lpad) + } else { + // Stop unwinding Rust panics at catch_unwind. + EHAction::Catch(lpad) + } } } diff --git a/library/std/src/prelude/v1.rs b/library/std/src/prelude/v1.rs index a5a798078..2aefd7c51 100644 --- a/library/std/src/prelude/v1.rs +++ b/library/std/src/prelude/v1.rs @@ -59,14 +59,12 @@ pub use core::prelude::v1::{RustcDecodable, RustcEncodable}; // Do not `doc(no_inline)` so that they become doc items on their own // (no public module for them to be re-exported from). -#[cfg(not(bootstrap))] #[stable(feature = "builtin_macro_prelude", since = "1.38.0")] -pub use core::prelude::v1::alloc_error_handler; -#[stable(feature = "builtin_macro_prelude", since = "1.38.0")] -pub use core::prelude::v1::{bench, derive, global_allocator, test, test_case}; +pub use core::prelude::v1::{ + alloc_error_handler, bench, derive, global_allocator, test, test_case, +}; #[unstable(feature = "derive_const", issue = "none")] -#[cfg(not(bootstrap))] pub use core::prelude::v1::derive_const; // Do not `doc(no_inline)` either. @@ -91,7 +89,6 @@ pub use core::prelude::v1::cfg_eval; issue = "23416", reason = "placeholder syntax for type ascription" )] -#[cfg(not(bootstrap))] pub use core::prelude::v1::type_ascribe; // The file so far is equivalent to src/libcore/prelude/v1.rs, diff --git a/library/std/src/process.rs b/library/std/src/process.rs index 400d25beb..62ce2cb33 100644 --- a/library/std/src/process.rs +++ b/library/std/src/process.rs @@ -362,6 +362,10 @@ impl Read for ChildStdout { fn is_read_vectored(&self) -> bool { self.inner.is_read_vectored() } + + fn read_to_end(&mut self, buf: &mut Vec) -> io::Result { + self.inner.read_to_end(buf) + } } impl AsInner for ChildStdout { @@ -907,10 +911,8 @@ impl Command { /// ``` #[stable(feature = "process", since = "1.0.0")] pub fn output(&mut self) -> io::Result { - self.inner - .spawn(imp::Stdio::MakePipe, false) - .map(Child::from_inner) - .and_then(|p| p.wait_with_output()) + let (status, stdout, stderr) = self.inner.output()?; + Ok(Output { status: ExitStatus(status), stdout, stderr }) } /// Executes a command as a child process, waiting for it to finish and @@ -1036,6 +1038,15 @@ impl fmt::Debug for Command { /// Format the program and arguments of a Command for display. Any /// non-utf8 data is lossily converted using the utf8 replacement /// character. + /// + /// The default format approximates a shell invocation of the program along with its + /// arguments. It does not include most of the other command properties. The output is not guaranteed to work + /// (e.g. due to lack of shell-escaping or differences in path resolution) + /// On some platforms you can use [the alternate syntax] to show more fields. + /// + /// Note that the debug implementation is platform-specific. + /// + /// [the alternate syntax]: fmt#sign0 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { self.inner.fmt(f) } @@ -2153,18 +2164,11 @@ pub fn id() -> u32 { /// to provide similar functionality. #[cfg_attr(not(test), lang = "termination")] #[stable(feature = "termination_trait_lib", since = "1.61.0")] -#[rustc_on_unimplemented( - on( - all(not(bootstrap), cause = "MainFunctionType"), - message = "`main` has invalid return type `{Self}`", - label = "`main` can only return types that implement `{Termination}`" - ), - on( - bootstrap, - message = "`main` has invalid return type `{Self}`", - label = "`main` can only return types that implement `{Termination}`" - ) -)] +#[rustc_on_unimplemented(on( + cause = "MainFunctionType", + message = "`main` has invalid return type `{Self}`", + label = "`main` can only return types that implement `{Termination}`" +))] pub trait Termination { /// Is called to get the representation of the value as status code. /// This status code is returned to the operating system. diff --git a/library/std/src/process/tests.rs b/library/std/src/process/tests.rs index 955ad6891..b4f6cc2da 100644 --- a/library/std/src/process/tests.rs +++ b/library/std/src/process/tests.rs @@ -417,6 +417,100 @@ fn env_empty() { assert!(p.is_ok()); } +#[test] +#[cfg(not(windows))] +#[cfg_attr(any(target_os = "emscripten", target_env = "sgx"), ignore)] +fn main() { + const PIDFD: &'static str = + if cfg!(target_os = "linux") { " create_pidfd: false,\n" } else { "" }; + + let mut command = Command::new("some-boring-name"); + + assert_eq!(format!("{command:?}"), format!(r#""some-boring-name""#)); + + assert_eq!( + format!("{command:#?}"), + format!( + r#"Command {{ + program: "some-boring-name", + args: [ + "some-boring-name", + ], +{PIDFD}}}"# + ) + ); + + command.args(&["1", "2", "3"]); + + assert_eq!(format!("{command:?}"), format!(r#""some-boring-name" "1" "2" "3""#)); + + assert_eq!( + format!("{command:#?}"), + format!( + r#"Command {{ + program: "some-boring-name", + args: [ + "some-boring-name", + "1", + "2", + "3", + ], +{PIDFD}}}"# + ) + ); + + crate::os::unix::process::CommandExt::arg0(&mut command, "exciting-name"); + + assert_eq!( + format!("{command:?}"), + format!(r#"["some-boring-name"] "exciting-name" "1" "2" "3""#) + ); + + assert_eq!( + format!("{command:#?}"), + format!( + r#"Command {{ + program: "some-boring-name", + args: [ + "exciting-name", + "1", + "2", + "3", + ], +{PIDFD}}}"# + ) + ); + + let mut command_with_env_and_cwd = Command::new("boring-name"); + command_with_env_and_cwd.current_dir("/some/path").env("FOO", "bar"); + assert_eq!( + format!("{command_with_env_and_cwd:?}"), + r#"cd "/some/path" && FOO="bar" "boring-name""# + ); + assert_eq!( + format!("{command_with_env_and_cwd:#?}"), + format!( + r#"Command {{ + program: "boring-name", + args: [ + "boring-name", + ], + env: CommandEnv {{ + clear: false, + vars: {{ + "FOO": Some( + "bar", + ), + }}, + }}, + cwd: Some( + "/some/path", + ), +{PIDFD}}}"# + ) + ); +} + // See issue #91991 #[test] #[cfg(windows)] diff --git a/library/std/src/rt.rs b/library/std/src/rt.rs index 9c2f0c1dd..f1eeb75be 100644 --- a/library/std/src/rt.rs +++ b/library/std/src/rt.rs @@ -139,9 +139,9 @@ fn lang_start_internal( // mechanism itself. // // There are a couple of instances where unwinding can begin. First is inside of the - // `rt::init`, `rt::cleanup` and similar functions controlled by libstd. In those instances a - // panic is a libstd implementation bug. A quite likely one too, as there isn't any way to - // prevent libstd from accidentally introducing a panic to these functions. Another is from + // `rt::init`, `rt::cleanup` and similar functions controlled by bstd. In those instances a + // panic is a std implementation bug. A quite likely one too, as there isn't any way to + // prevent std from accidentally introducing a panic to these functions. Another is from // user code from `main` or, more nefariously, as described in e.g. issue #86030. // SAFETY: Only called once during runtime initialization. panic::catch_unwind(move || unsafe { init(argc, argv, sigpipe) }).map_err(rt_abort)?; diff --git a/library/std/src/sync/lazy_lock.rs b/library/std/src/sync/lazy_lock.rs index c8d3289ca..4a1530530 100644 --- a/library/std/src/sync/lazy_lock.rs +++ b/library/std/src/sync/lazy_lock.rs @@ -46,17 +46,15 @@ pub struct LazyLock T> { cell: OnceLock, init: Cell>, } - -impl LazyLock { +impl T> LazyLock { /// Creates a new lazy value with the given initializing /// function. + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub const fn new(f: F) -> LazyLock { LazyLock { cell: OnceLock::new(), init: Cell::new(Some(f)) } } -} -impl T> LazyLock { /// Forces the evaluation of this lazy value and /// returns a reference to result. This is equivalent /// to the `Deref` impl, but is explicit. @@ -73,6 +71,7 @@ impl T> LazyLock { /// assert_eq!(LazyLock::force(&lazy), &92); /// assert_eq!(&*lazy, &92); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn force(this: &LazyLock) -> &T { this.cell.get_or_init(|| match this.init.take() { @@ -85,6 +84,8 @@ impl T> LazyLock { #[unstable(feature = "once_cell", issue = "74465")] impl T> Deref for LazyLock { type Target = T; + + #[inline] fn deref(&self) -> &T { LazyLock::force(self) } @@ -93,6 +94,7 @@ impl T> Deref for LazyLock { #[unstable(feature = "once_cell", issue = "74465")] impl Default for LazyLock { /// Creates a new lazy value using `Default` as the initializing function. + #[inline] fn default() -> LazyLock { LazyLock::new(T::default) } diff --git a/library/std/src/sync/lazy_lock/tests.rs b/library/std/src/sync/lazy_lock/tests.rs index f11b66bfc..a5d4e25c5 100644 --- a/library/std/src/sync/lazy_lock/tests.rs +++ b/library/std/src/sync/lazy_lock/tests.rs @@ -136,6 +136,12 @@ fn sync_lazy_poisoning() { } } +// Check that we can infer `T` from closure's type. +#[test] +fn lazy_type_inference() { + let _ = LazyCell::new(|| ()); +} + #[test] fn is_sync_send() { fn assert_traits() {} diff --git a/library/std/src/sync/mod.rs b/library/std/src/sync/mod.rs index 4fee8d3e9..ba20bab87 100644 --- a/library/std/src/sync/mod.rs +++ b/library/std/src/sync/mod.rs @@ -177,6 +177,8 @@ pub use self::lazy_lock::LazyLock; #[unstable(feature = "once_cell", issue = "74465")] pub use self::once_lock::OnceLock; +pub(crate) use self::remutex::{ReentrantMutex, ReentrantMutexGuard}; + pub mod mpsc; mod barrier; @@ -187,4 +189,5 @@ mod mutex; mod once; mod once_lock; mod poison; +mod remutex; mod rwlock; diff --git a/library/std/src/sync/mutex/tests.rs b/library/std/src/sync/mutex/tests.rs index 93900566f..1786a3c09 100644 --- a/library/std/src/sync/mutex/tests.rs +++ b/library/std/src/sync/mutex/tests.rs @@ -181,7 +181,7 @@ fn test_mutex_arc_poison() { let arc2 = arc.clone(); let _ = thread::spawn(move || { let lock = arc2.lock().unwrap(); - assert_eq!(*lock, 2); + assert_eq!(*lock, 2); // deliberate assertion failure to poison the mutex }) .join(); assert!(arc.lock().is_err()); diff --git a/library/std/src/sync/once_lock.rs b/library/std/src/sync/once_lock.rs index 16d1fd2a5..ed339ca5d 100644 --- a/library/std/src/sync/once_lock.rs +++ b/library/std/src/sync/once_lock.rs @@ -61,8 +61,9 @@ pub struct OnceLock { impl OnceLock { /// Creates a new empty cell. - #[unstable(feature = "once_cell", issue = "74465")] + #[inline] #[must_use] + #[unstable(feature = "once_cell", issue = "74465")] pub const fn new() -> OnceLock { OnceLock { once: Once::new(), @@ -75,6 +76,7 @@ impl OnceLock { /// /// Returns `None` if the cell is empty, or being initialized. This /// method never blocks. + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get(&self) -> Option<&T> { if self.is_initialized() { @@ -88,6 +90,7 @@ impl OnceLock { /// Gets the mutable reference to the underlying value. /// /// Returns `None` if the cell is empty. This method never blocks. + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get_mut(&mut self) -> Option<&mut T> { if self.is_initialized() { @@ -125,6 +128,7 @@ impl OnceLock { /// assert_eq!(CELL.get(), Some(&92)); /// } /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn set(&self, value: T) -> Result<(), T> { let mut value = Some(value); @@ -164,6 +168,7 @@ impl OnceLock { /// let value = cell.get_or_init(|| unreachable!()); /// assert_eq!(value, &92); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get_or_init(&self, f: F) -> &T where @@ -203,6 +208,7 @@ impl OnceLock { /// assert_eq!(value, Ok(&92)); /// assert_eq!(cell.get(), Some(&92)) /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn get_or_try_init(&self, f: F) -> Result<&T, E> where @@ -241,6 +247,7 @@ impl OnceLock { /// cell.set("hello".to_string()).unwrap(); /// assert_eq!(cell.into_inner(), Some("hello".to_string())); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn into_inner(mut self) -> Option { self.take() @@ -267,6 +274,7 @@ impl OnceLock { /// assert_eq!(cell.take(), Some("hello".to_string())); /// assert_eq!(cell.get(), None); /// ``` + #[inline] #[unstable(feature = "once_cell", issue = "74465")] pub fn take(&mut self) -> Option { if self.is_initialized() { @@ -315,6 +323,7 @@ impl OnceLock { /// # Safety /// /// The value must be initialized + #[inline] unsafe fn get_unchecked(&self) -> &T { debug_assert!(self.is_initialized()); (&*self.value.get()).assume_init_ref() @@ -323,6 +332,7 @@ impl OnceLock { /// # Safety /// /// The value must be initialized + #[inline] unsafe fn get_unchecked_mut(&mut self) -> &mut T { debug_assert!(self.is_initialized()); (&mut *self.value.get()).assume_init_mut() @@ -360,6 +370,7 @@ impl const Default for OnceLock { /// assert_eq!(OnceLock::<()>::new(), OnceLock::default()); /// } /// ``` + #[inline] fn default() -> OnceLock { OnceLock::new() } @@ -377,6 +388,7 @@ impl fmt::Debug for OnceLock { #[unstable(feature = "once_cell", issue = "74465")] impl Clone for OnceLock { + #[inline] fn clone(&self) -> OnceLock { let cell = Self::new(); if let Some(value) = self.get() { @@ -408,6 +420,7 @@ impl From for OnceLock { /// Ok(()) /// # } /// ``` + #[inline] fn from(value: T) -> Self { let cell = Self::new(); match cell.set(value) { @@ -419,6 +432,7 @@ impl From for OnceLock { #[unstable(feature = "once_cell", issue = "74465")] impl PartialEq for OnceLock { + #[inline] fn eq(&self, other: &OnceLock) -> bool { self.get() == other.get() } @@ -429,6 +443,7 @@ impl Eq for OnceLock {} #[unstable(feature = "once_cell", issue = "74465")] unsafe impl<#[may_dangle] T> Drop for OnceLock { + #[inline] fn drop(&mut self) { if self.is_initialized() { // SAFETY: The cell is initialized and being dropped, so it can't diff --git a/library/std/src/sync/remutex.rs b/library/std/src/sync/remutex.rs new file mode 100644 index 000000000..4c054da64 --- /dev/null +++ b/library/std/src/sync/remutex.rs @@ -0,0 +1,178 @@ +#[cfg(all(test, not(target_os = "emscripten")))] +mod tests; + +use crate::cell::UnsafeCell; +use crate::ops::Deref; +use crate::panic::{RefUnwindSafe, UnwindSafe}; +use crate::sync::atomic::{AtomicUsize, Ordering::Relaxed}; +use crate::sys::locks as sys; + +/// A re-entrant mutual exclusion +/// +/// This mutex will block *other* threads waiting for the lock to become +/// available. The thread which has already locked the mutex can lock it +/// multiple times without blocking, preventing a common source of deadlocks. +/// +/// This is used by stdout().lock() and friends. +/// +/// ## Implementation details +/// +/// The 'owner' field tracks which thread has locked the mutex. +/// +/// We use current_thread_unique_ptr() as the thread identifier, +/// which is just the address of a thread local variable. +/// +/// If `owner` is set to the identifier of the current thread, +/// we assume the mutex is already locked and instead of locking it again, +/// we increment `lock_count`. +/// +/// When unlocking, we decrement `lock_count`, and only unlock the mutex when +/// it reaches zero. +/// +/// `lock_count` is protected by the mutex and only accessed by the thread that has +/// locked the mutex, so needs no synchronization. +/// +/// `owner` can be checked by other threads that want to see if they already +/// hold the lock, so needs to be atomic. If it compares equal, we're on the +/// same thread that holds the mutex and memory access can use relaxed ordering +/// since we're not dealing with multiple threads. If it compares unequal, +/// synchronization is left to the mutex, making relaxed memory ordering for +/// the `owner` field fine in all cases. +pub struct ReentrantMutex { + mutex: sys::Mutex, + owner: AtomicUsize, + lock_count: UnsafeCell, + data: T, +} + +unsafe impl Send for ReentrantMutex {} +unsafe impl Sync for ReentrantMutex {} + +impl UnwindSafe for ReentrantMutex {} +impl RefUnwindSafe for ReentrantMutex {} + +/// An RAII implementation of a "scoped lock" of a mutex. When this structure is +/// dropped (falls out of scope), the lock will be unlocked. +/// +/// The data protected by the mutex can be accessed through this guard via its +/// Deref implementation. +/// +/// # Mutability +/// +/// Unlike `MutexGuard`, `ReentrantMutexGuard` does not implement `DerefMut`, +/// because implementation of the trait would violate Rust’s reference aliasing +/// rules. Use interior mutability (usually `RefCell`) in order to mutate the +/// guarded data. +#[must_use = "if unused the ReentrantMutex will immediately unlock"] +pub struct ReentrantMutexGuard<'a, T: 'a> { + lock: &'a ReentrantMutex, +} + +impl !Send for ReentrantMutexGuard<'_, T> {} + +impl ReentrantMutex { + /// Creates a new reentrant mutex in an unlocked state. + pub const fn new(t: T) -> ReentrantMutex { + ReentrantMutex { + mutex: sys::Mutex::new(), + owner: AtomicUsize::new(0), + lock_count: UnsafeCell::new(0), + data: t, + } + } + + /// Acquires a mutex, blocking the current thread until it is able to do so. + /// + /// This function will block the caller until it is available to acquire the mutex. + /// Upon returning, the thread is the only thread with the mutex held. When the thread + /// calling this method already holds the lock, the call shall succeed without + /// blocking. + /// + /// # Errors + /// + /// If another user of this mutex panicked while holding the mutex, then + /// this call will return failure if the mutex would otherwise be + /// acquired. + pub fn lock(&self) -> ReentrantMutexGuard<'_, T> { + let this_thread = current_thread_unique_ptr(); + // Safety: We only touch lock_count when we own the lock. + unsafe { + if self.owner.load(Relaxed) == this_thread { + self.increment_lock_count(); + } else { + self.mutex.lock(); + self.owner.store(this_thread, Relaxed); + debug_assert_eq!(*self.lock_count.get(), 0); + *self.lock_count.get() = 1; + } + } + ReentrantMutexGuard { lock: self } + } + + /// Attempts to acquire this lock. + /// + /// If the lock could not be acquired at this time, then `Err` is returned. + /// Otherwise, an RAII guard is returned. + /// + /// This function does not block. + /// + /// # Errors + /// + /// If another user of this mutex panicked while holding the mutex, then + /// this call will return failure if the mutex would otherwise be + /// acquired. + pub fn try_lock(&self) -> Option> { + let this_thread = current_thread_unique_ptr(); + // Safety: We only touch lock_count when we own the lock. + unsafe { + if self.owner.load(Relaxed) == this_thread { + self.increment_lock_count(); + Some(ReentrantMutexGuard { lock: self }) + } else if self.mutex.try_lock() { + self.owner.store(this_thread, Relaxed); + debug_assert_eq!(*self.lock_count.get(), 0); + *self.lock_count.get() = 1; + Some(ReentrantMutexGuard { lock: self }) + } else { + None + } + } + } + + unsafe fn increment_lock_count(&self) { + *self.lock_count.get() = (*self.lock_count.get()) + .checked_add(1) + .expect("lock count overflow in reentrant mutex"); + } +} + +impl Deref for ReentrantMutexGuard<'_, T> { + type Target = T; + + fn deref(&self) -> &T { + &self.lock.data + } +} + +impl Drop for ReentrantMutexGuard<'_, T> { + #[inline] + fn drop(&mut self) { + // Safety: We own the lock. + unsafe { + *self.lock.lock_count.get() -= 1; + if *self.lock.lock_count.get() == 0 { + self.lock.owner.store(0, Relaxed); + self.lock.mutex.unlock(); + } + } + } +} + +/// Get an address that is unique per running thread. +/// +/// This can be used as a non-null usize-sized ID. +pub fn current_thread_unique_ptr() -> usize { + // Use a non-drop type to make sure it's still available during thread destruction. + thread_local! { static X: u8 = const { 0 } } + X.with(|x| <*const _>::addr(x)) +} diff --git a/library/std/src/sync/remutex/tests.rs b/library/std/src/sync/remutex/tests.rs new file mode 100644 index 000000000..fc553081d --- /dev/null +++ b/library/std/src/sync/remutex/tests.rs @@ -0,0 +1,60 @@ +use super::{ReentrantMutex, ReentrantMutexGuard}; +use crate::cell::RefCell; +use crate::sync::Arc; +use crate::thread; + +#[test] +fn smoke() { + let m = ReentrantMutex::new(()); + { + let a = m.lock(); + { + let b = m.lock(); + { + let c = m.lock(); + assert_eq!(*c, ()); + } + assert_eq!(*b, ()); + } + assert_eq!(*a, ()); + } +} + +#[test] +fn is_mutex() { + let m = Arc::new(ReentrantMutex::new(RefCell::new(0))); + let m2 = m.clone(); + let lock = m.lock(); + let child = thread::spawn(move || { + let lock = m2.lock(); + assert_eq!(*lock.borrow(), 4950); + }); + for i in 0..100 { + let lock = m.lock(); + *lock.borrow_mut() += i; + } + drop(lock); + child.join().unwrap(); +} + +#[test] +fn trylock_works() { + let m = Arc::new(ReentrantMutex::new(())); + let m2 = m.clone(); + let _lock = m.try_lock(); + let _lock2 = m.try_lock(); + thread::spawn(move || { + let lock = m2.try_lock(); + assert!(lock.is_none()); + }) + .join() + .unwrap(); + let _lock3 = m.try_lock(); +} + +pub struct Answer<'a>(pub ReentrantMutexGuard<'a, RefCell>); +impl Drop for Answer<'_> { + fn drop(&mut self) { + *self.0.borrow_mut() = 42; + } +} diff --git a/library/std/src/sync/rwlock/tests.rs b/library/std/src/sync/rwlock/tests.rs index b5b3ad989..1a9d3d3f1 100644 --- a/library/std/src/sync/rwlock/tests.rs +++ b/library/std/src/sync/rwlock/tests.rs @@ -2,7 +2,7 @@ use crate::sync::atomic::{AtomicUsize, Ordering}; use crate::sync::mpsc::channel; use crate::sync::{Arc, RwLock, RwLockReadGuard, TryLockError}; use crate::thread; -use rand::{self, Rng}; +use rand::Rng; #[derive(Eq, PartialEq, Debug)] struct NonCopy(i32); @@ -28,7 +28,7 @@ fn frob() { let tx = tx.clone(); let r = r.clone(); thread::spawn(move || { - let mut rng = rand::thread_rng(); + let mut rng = crate::test_helpers::test_rng(); for _ in 0..M { if rng.gen_bool(1.0 / (N as f64)) { drop(r.write().unwrap()); diff --git a/library/std/src/sys/itron/thread.rs b/library/std/src/sys/itron/thread.rs index c2b366808..19350b83f 100644 --- a/library/std/src/sys/itron/thread.rs +++ b/library/std/src/sys/itron/thread.rs @@ -119,7 +119,7 @@ impl Thread { let old_lifecycle = inner .lifecycle - .swap(LIFECYCLE_EXITED_OR_FINISHED_OR_JOIN_FINALIZE, Ordering::Release); + .swap(LIFECYCLE_EXITED_OR_FINISHED_OR_JOIN_FINALIZE, Ordering::AcqRel); match old_lifecycle { LIFECYCLE_DETACHED => { @@ -129,9 +129,9 @@ impl Thread { // In this case, `*p_inner`'s ownership has been moved to // us, and we are responsible for dropping it. The acquire - // ordering is not necessary because the parent thread made - // no memory access needing synchronization since the call - // to `acre_tsk`. + // ordering ensures that the swap operation that wrote + // `LIFECYCLE_DETACHED` happens-before `Box::from_raw( + // p_inner)`. // Safety: See above. let _ = unsafe { Box::from_raw(p_inner) }; @@ -151,6 +151,9 @@ impl Thread { // Since the parent might drop `*inner` and terminate us as // soon as it sees `JOIN_FINALIZE`, the release ordering // must be used in the above `swap` call. + // + // To make the task referred to by `parent_tid` visible, we + // must use the acquire ordering in the above `swap` call. // [JOINING → JOIN_FINALIZE] // Wake up the parent task. @@ -218,11 +221,15 @@ impl Thread { let current_task = current_task as usize; - match inner.lifecycle.swap(current_task, Ordering::Acquire) { + match inner.lifecycle.swap(current_task, Ordering::AcqRel) { LIFECYCLE_INIT => { // [INIT → JOINING] // The child task will transition the state to `JOIN_FINALIZE` // and wake us up. + // + // To make the task referred to by `current_task` visible from + // the child task's point of view, we must use the release + // ordering in the above `swap` call. loop { expect_success_aborting(unsafe { abi::slp_tsk() }, &"slp_tsk"); // To synchronize with the child task's memory accesses to @@ -267,15 +274,15 @@ impl Drop for Thread { let inner = unsafe { self.p_inner.as_ref() }; // Detach the thread. - match inner.lifecycle.swap(LIFECYCLE_DETACHED_OR_JOINED, Ordering::Acquire) { + match inner.lifecycle.swap(LIFECYCLE_DETACHED_OR_JOINED, Ordering::AcqRel) { LIFECYCLE_INIT => { // [INIT → DETACHED] // When the time comes, the child will figure out that no // one will ever join it. // The ownership of `*p_inner` is moved to the child thread. - // However, the release ordering is not necessary because we - // made no memory access needing synchronization since the call - // to `acre_tsk`. + // The release ordering ensures that the above swap operation on + // `lifecycle` happens-before the child thread's + // `Box::from_raw(p_inner)`. } LIFECYCLE_FINISHED => { // [FINISHED → JOINED] @@ -287,7 +294,7 @@ impl Drop for Thread { // Terminate and delete the task // Safety: `self.task` still represents a task we own (because // this method or `join_inner` is called only once for - // each `Thread`). The task indicated that it's safe to + // each `Thread`). The task indicated that it's safe to // delete by entering the `FINISHED` state. unsafe { terminate_and_delete_task(self.task) }; diff --git a/library/std/src/sys/sgx/mod.rs b/library/std/src/sys/sgx/mod.rs index 01e4ffe3d..9865a945b 100644 --- a/library/std/src/sys/sgx/mod.rs +++ b/library/std/src/sys/sgx/mod.rs @@ -34,6 +34,7 @@ pub mod process; pub mod stdio; pub mod thread; pub mod thread_local_key; +pub mod thread_parking; pub mod time; mod condvar; diff --git a/library/std/src/sys/sgx/thread.rs b/library/std/src/sys/sgx/thread.rs index d745a6196..1608b8cb6 100644 --- a/library/std/src/sys/sgx/thread.rs +++ b/library/std/src/sys/sgx/thread.rs @@ -65,39 +65,36 @@ mod task_queue { /// execution. The signal is sent once all TLS destructors have finished at /// which point no new thread locals should be created. pub mod wait_notify { - use super::super::waitqueue::{SpinMutex, WaitQueue, WaitVariable}; + use crate::pin::Pin; use crate::sync::Arc; + use crate::sys_common::thread_parking::Parker; - pub struct Notifier(Arc>>); + pub struct Notifier(Arc); impl Notifier { /// Notify the waiter. The waiter is either notified right away (if /// currently blocked in `Waiter::wait()`) or later when it calls the /// `Waiter::wait()` method. pub fn notify(self) { - let mut guard = self.0.lock(); - *guard.lock_var_mut() = true; - let _ = WaitQueue::notify_one(guard); + Pin::new(&*self.0).unpark() } } - pub struct Waiter(Arc>>); + pub struct Waiter(Arc); impl Waiter { /// Wait for a notification. If `Notifier::notify()` has already been /// called, this will return immediately, otherwise the current thread /// is blocked until notified. pub fn wait(self) { - let guard = self.0.lock(); - if *guard.lock_var() { - return; - } - WaitQueue::wait(guard, || {}); + // SAFETY: + // This is only ever called on one thread. + unsafe { Pin::new(&*self.0).park() } } } pub fn new() -> (Notifier, Waiter) { - let inner = Arc::new(SpinMutex::new(WaitVariable::new(false))); + let inner = Arc::new(Parker::new()); (Notifier(inner.clone()), Waiter(inner)) } } diff --git a/library/std/src/sys/sgx/thread_parking.rs b/library/std/src/sys/sgx/thread_parking.rs new file mode 100644 index 000000000..0006cd4f1 --- /dev/null +++ b/library/std/src/sys/sgx/thread_parking.rs @@ -0,0 +1,23 @@ +use super::abi::usercalls; +use crate::io::ErrorKind; +use crate::time::Duration; +use fortanix_sgx_abi::{EV_UNPARK, WAIT_INDEFINITE}; + +pub type ThreadId = fortanix_sgx_abi::Tcs; + +pub use super::abi::thread::current; + +pub fn park(_hint: usize) { + usercalls::wait(EV_UNPARK, WAIT_INDEFINITE).unwrap(); +} + +pub fn park_timeout(dur: Duration, _hint: usize) { + let timeout = u128::min(dur.as_nanos(), WAIT_INDEFINITE as u128 - 1) as u64; + if let Err(e) = usercalls::wait(EV_UNPARK, timeout) { + assert!(matches!(e.kind(), ErrorKind::TimedOut | ErrorKind::WouldBlock)) + } +} + +pub fn unpark(tid: ThreadId, _hint: usize) { + let _ = usercalls::send(EV_UNPARK, Some(tid)); +} diff --git a/library/std/src/sys/unix/android.rs b/library/std/src/sys/unix/android.rs index 73ff10ab8..0f704994f 100644 --- a/library/std/src/sys/unix/android.rs +++ b/library/std/src/sys/unix/android.rs @@ -1,7 +1,7 @@ //! Android ABI-compatibility module //! -//! The ABI of Android has changed quite a bit over time, and libstd attempts to -//! be both forwards and backwards compatible as much as possible. We want to +//! The ABI of Android has changed quite a bit over time, and std attempts to be +//! both forwards and backwards compatible as much as possible. We want to //! always work with the most recent version of Android, but we also want to //! work with older versions of Android for whenever projects need to. //! diff --git a/library/std/src/sys/unix/fs.rs b/library/std/src/sys/unix/fs.rs index 37a49f2d7..8e1f35d6c 100644 --- a/library/std/src/sys/unix/fs.rs +++ b/library/std/src/sys/unix/fs.rs @@ -149,12 +149,13 @@ cfg_has_statx! {{ ) -> Option> { use crate::sync::atomic::{AtomicU8, Ordering}; - // Linux kernel prior to 4.11 or glibc prior to glibc 2.28 don't support `statx` - // We store the availability in global to avoid unnecessary syscalls. - // 0: Unknown - // 1: Not available - // 2: Available - static STATX_STATE: AtomicU8 = AtomicU8::new(0); + // Linux kernel prior to 4.11 or glibc prior to glibc 2.28 don't support `statx`. + // We check for it on first failure and remember availability to avoid having to + // do it again. + #[repr(u8)] + enum STATX_STATE{ Unknown = 0, Present, Unavailable } + static STATX_SAVED_STATE: AtomicU8 = AtomicU8::new(STATX_STATE::Unknown as u8); + syscall! { fn statx( fd: c_int, @@ -165,31 +166,44 @@ cfg_has_statx! {{ ) -> c_int } - match STATX_STATE.load(Ordering::Relaxed) { - 0 => { - // It is a trick to call `statx` with null pointers to check if the syscall - // is available. According to the manual, it is expected to fail with EFAULT. - // We do this mainly for performance, since it is nearly hundreds times - // faster than a normal successful call. - let err = cvt(statx(0, ptr::null(), 0, libc::STATX_ALL, ptr::null_mut())) - .err() - .and_then(|e| e.raw_os_error()); - // We don't check `err == Some(libc::ENOSYS)` because the syscall may be limited - // and returns `EPERM`. Listing all possible errors seems not a good idea. - // See: https://github.com/rust-lang/rust/issues/65662 - if err != Some(libc::EFAULT) { - STATX_STATE.store(1, Ordering::Relaxed); - return None; - } - STATX_STATE.store(2, Ordering::Relaxed); - } - 1 => return None, - _ => {} + if STATX_SAVED_STATE.load(Ordering::Relaxed) == STATX_STATE::Unavailable as u8 { + return None; } let mut buf: libc::statx = mem::zeroed(); if let Err(err) = cvt(statx(fd, path, flags, mask, &mut buf)) { - return Some(Err(err)); + if STATX_SAVED_STATE.load(Ordering::Relaxed) == STATX_STATE::Present as u8 { + return Some(Err(err)); + } + + // Availability not checked yet. + // + // First try the cheap way. + if err.raw_os_error() == Some(libc::ENOSYS) { + STATX_SAVED_STATE.store(STATX_STATE::Unavailable as u8, Ordering::Relaxed); + return None; + } + + // Error other than `ENOSYS` is not a good enough indicator -- it is + // known that `EPERM` can be returned as a result of using seccomp to + // block the syscall. + // Availability is checked by performing a call which expects `EFAULT` + // if the syscall is usable. + // See: https://github.com/rust-lang/rust/issues/65662 + // FIXME this can probably just do the call if `EPERM` was received, but + // previous iteration of the code checked it for all errors and for now + // this is retained. + // FIXME what about transient conditions like `ENOMEM`? + let err2 = cvt(statx(0, ptr::null(), 0, libc::STATX_ALL, ptr::null_mut())) + .err() + .and_then(|e| e.raw_os_error()); + if err2 == Some(libc::EFAULT) { + STATX_SAVED_STATE.store(STATX_STATE::Present as u8, Ordering::Relaxed); + return Some(Err(err)); + } else { + STATX_SAVED_STATE.store(STATX_STATE::Unavailable as u8, Ordering::Relaxed); + return None; + } } // We cannot fill `stat64` exhaustively because of private padding fields. @@ -243,17 +257,15 @@ struct InnerReadDir { pub struct ReadDir { inner: Arc, - #[cfg(not(any( - target_os = "android", - target_os = "linux", - target_os = "solaris", - target_os = "illumos", - target_os = "fuchsia", - target_os = "redox", - )))] end_of_stream: bool, } +impl ReadDir { + fn new(inner: InnerReadDir) -> Self { + Self { inner: Arc::new(inner), end_of_stream: false } + } +} + struct Dir(*mut libc::DIR); unsafe impl Send for Dir {} @@ -332,11 +344,23 @@ pub struct FileTimes { modified: Option, } -#[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)] +#[derive(Copy, Clone, Eq, Debug)] pub struct FileType { mode: mode_t, } +impl PartialEq for FileType { + fn eq(&self, other: &Self) -> bool { + self.masked() == other.masked() + } +} + +impl core::hash::Hash for FileType { + fn hash(&self, state: &mut H) { + self.masked().hash(state); + } +} + #[derive(Debug)] pub struct DirBuilder { mode: mode_t, @@ -548,7 +572,11 @@ impl FileType { } pub fn is(&self, mode: mode_t) -> bool { - self.mode & libc::S_IFMT == mode + self.masked() == mode + } + + fn masked(&self) -> mode_t { + self.mode & libc::S_IFMT } } @@ -578,18 +606,26 @@ impl Iterator for ReadDir { target_os = "illumos" ))] fn next(&mut self) -> Option> { + if self.end_of_stream { + return None; + } + unsafe { loop { // As of POSIX.1-2017, readdir() is not required to be thread safe; only // readdir_r() is. However, readdir_r() cannot correctly handle platforms - // with unlimited or variable NAME_MAX. Many modern platforms guarantee + // with unlimited or variable NAME_MAX. Many modern platforms guarantee // thread safety for readdir() as long an individual DIR* is not accessed // concurrently, which is sufficient for Rust. super::os::set_errno(0); let entry_ptr = readdir64(self.inner.dirp.0); if entry_ptr.is_null() { - // null can mean either the end is reached or an error occurred. - // So we had to clear errno beforehand to check for an error now. + // We either encountered an error, or reached the end. Either way, + // the next call to next() should return None. + self.end_of_stream = true; + + // To distinguish between errors and end-of-directory, we had to clear + // errno beforehand to check for an error now. return match super::os::errno() { 0 => None, e => Some(Err(Error::from_raw_os_error(e))), @@ -1347,18 +1383,7 @@ pub fn readdir(path: &Path) -> io::Result { } else { let root = path.to_path_buf(); let inner = InnerReadDir { dirp: Dir(ptr), root }; - Ok(ReadDir { - inner: Arc::new(inner), - #[cfg(not(any( - target_os = "android", - target_os = "linux", - target_os = "solaris", - target_os = "illumos", - target_os = "fuchsia", - target_os = "redox", - )))] - end_of_stream: false, - }) + Ok(ReadDir::new(inner)) } } @@ -1739,12 +1764,16 @@ mod remove_dir_impl { use crate::os::unix::io::{AsRawFd, FromRawFd, IntoRawFd}; use crate::os::unix::prelude::{OwnedFd, RawFd}; use crate::path::{Path, PathBuf}; - use crate::sync::Arc; use crate::sys::common::small_c_string::run_path_with_cstr; use crate::sys::{cvt, cvt_r}; - #[cfg(not(all(target_os = "macos", not(target_arch = "aarch64")),))] + #[cfg(not(any( + all(target_os = "linux", target_env = "gnu"), + all(target_os = "macos", not(target_arch = "aarch64")) + )))] use libc::{fdopendir, openat, unlinkat}; + #[cfg(all(target_os = "linux", target_env = "gnu"))] + use libc::{fdopendir, openat64 as openat, unlinkat}; #[cfg(all(target_os = "macos", not(target_arch = "aarch64")))] use macos_weak::{fdopendir, openat, unlinkat}; @@ -1811,21 +1840,8 @@ mod remove_dir_impl { // a valid root is not needed because we do not call any functions involving the full path // of the DirEntrys. let dummy_root = PathBuf::new(); - Ok(( - ReadDir { - inner: Arc::new(InnerReadDir { dirp, root: dummy_root }), - #[cfg(not(any( - target_os = "android", - target_os = "linux", - target_os = "solaris", - target_os = "illumos", - target_os = "fuchsia", - target_os = "redox", - )))] - end_of_stream: false, - }, - new_parent_fd, - )) + let inner = InnerReadDir { dirp, root: dummy_root }; + Ok((ReadDir::new(inner), new_parent_fd)) } #[cfg(any( diff --git a/library/std/src/sys/unix/kernel_copy.rs b/library/std/src/sys/unix/kernel_copy.rs index 94546ca09..73b9bef7e 100644 --- a/library/std/src/sys/unix/kernel_copy.rs +++ b/library/std/src/sys/unix/kernel_copy.rs @@ -61,6 +61,10 @@ use crate::ptr; use crate::sync::atomic::{AtomicBool, AtomicU8, Ordering}; use crate::sys::cvt; use crate::sys::weak::syscall; +#[cfg(not(all(target_os = "linux", target_env = "gnu")))] +use libc::sendfile as sendfile64; +#[cfg(all(target_os = "linux", target_env = "gnu"))] +use libc::sendfile64; use libc::{EBADF, EINVAL, ENOSYS, EOPNOTSUPP, EOVERFLOW, EPERM, EXDEV}; #[cfg(test)] @@ -583,7 +587,7 @@ pub(super) fn copy_regular_files(reader: RawFd, writer: RawFd, max_len: u64) -> // - copy_file_range file is immutable or syscall is blocked by seccomp¹ (EPERM) // - copy_file_range cannot be used with pipes or device nodes (EINVAL) // - the writer fd was opened with O_APPEND (EBADF²) - // and no bytes were written successfully yet. (All these errnos should + // and no bytes were written successfully yet. (All these errnos should // not be returned if something was already written, but they happen in // the wild, see #91152.) // @@ -647,7 +651,7 @@ fn sendfile_splice(mode: SpliceMode, reader: RawFd, writer: RawFd, len: u64) -> let result = match mode { SpliceMode::Sendfile => { - cvt(unsafe { libc::sendfile(writer, reader, ptr::null_mut(), chunk_size) }) + cvt(unsafe { sendfile64(writer, reader, ptr::null_mut(), chunk_size) }) } SpliceMode::Splice => cvt(unsafe { splice(reader, ptr::null_mut(), writer, ptr::null_mut(), chunk_size, 0) diff --git a/library/std/src/sys/unix/locks/pthread_condvar.rs b/library/std/src/sys/unix/locks/pthread_condvar.rs index 1ddb09905..6be1abc2b 100644 --- a/library/std/src/sys/unix/locks/pthread_condvar.rs +++ b/library/std/src/sys/unix/locks/pthread_condvar.rs @@ -2,6 +2,7 @@ use crate::cell::UnsafeCell; use crate::ptr; use crate::sync::atomic::{AtomicPtr, Ordering::Relaxed}; use crate::sys::locks::{pthread_mutex, Mutex}; +use crate::sys::time::TIMESPEC_MAX; use crate::sys_common::lazy_box::{LazyBox, LazyInit}; use crate::time::Duration; @@ -12,13 +13,6 @@ pub struct Condvar { mutex: AtomicPtr, } -const TIMESPEC_MAX: libc::timespec = - libc::timespec { tv_sec: ::MAX, tv_nsec: 1_000_000_000 - 1 }; - -fn saturating_cast_to_time_t(value: u64) -> libc::time_t { - if value > ::MAX as u64 { ::MAX } else { value as libc::time_t } -} - #[inline] fn raw(c: &Condvar) -> *mut libc::pthread_cond_t { c.inner.0.get() @@ -133,26 +127,15 @@ impl Condvar { target_os = "horizon" )))] pub unsafe fn wait_timeout(&self, mutex: &Mutex, dur: Duration) -> bool { - use crate::mem; + use crate::sys::time::Timespec; let mutex = pthread_mutex::raw(mutex); self.verify(mutex); - let mut now: libc::timespec = mem::zeroed(); - let r = libc::clock_gettime(libc::CLOCK_MONOTONIC, &mut now); - assert_eq!(r, 0); - - // Nanosecond calculations can't overflow because both values are below 1e9. - let nsec = dur.subsec_nanos() + now.tv_nsec as u32; - - let sec = saturating_cast_to_time_t(dur.as_secs()) - .checked_add((nsec / 1_000_000_000) as libc::time_t) - .and_then(|s| s.checked_add(now.tv_sec)); - let nsec = nsec % 1_000_000_000; - - let timeout = - sec.map(|s| libc::timespec { tv_sec: s, tv_nsec: nsec as _ }).unwrap_or(TIMESPEC_MAX); - + let timeout = Timespec::now(libc::CLOCK_MONOTONIC) + .checked_add_duration(&dur) + .and_then(|t| t.to_timespec()) + .unwrap_or(TIMESPEC_MAX); let r = libc::pthread_cond_timedwait(raw(self), mutex, &timeout); assert!(r == libc::ETIMEDOUT || r == 0); r == 0 @@ -169,57 +152,41 @@ impl Condvar { target_os = "espidf", target_os = "horizon" ))] - pub unsafe fn wait_timeout(&self, mutex: &Mutex, mut dur: Duration) -> bool { + pub unsafe fn wait_timeout(&self, mutex: &Mutex, dur: Duration) -> bool { + use crate::sys::time::SystemTime; use crate::time::Instant; let mutex = pthread_mutex::raw(mutex); self.verify(mutex); - // 1000 years - let max_dur = Duration::from_secs(1000 * 365 * 86400); - - if dur > max_dur { - // OSX implementation of `pthread_cond_timedwait` is buggy - // with super long durations. When duration is greater than - // 0x100_0000_0000_0000 seconds, `pthread_cond_timedwait` - // in macOS Sierra return error 316. - // - // This program demonstrates the issue: - // https://gist.github.com/stepancheg/198db4623a20aad2ad7cddb8fda4a63c - // - // To work around this issue, and possible bugs of other OSes, timeout - // is clamped to 1000 years, which is allowable per the API of `wait_timeout` - // because of spurious wakeups. - - dur = max_dur; - } - - // First, figure out what time it currently is, in both system and - // stable time. pthread_cond_timedwait uses system time, but we want to - // report timeout based on stable time. - let mut sys_now = libc::timeval { tv_sec: 0, tv_usec: 0 }; - let stable_now = Instant::now(); - let r = libc::gettimeofday(&mut sys_now, ptr::null_mut()); - assert_eq!(r, 0, "unexpected error: {:?}", crate::io::Error::last_os_error()); - - let nsec = dur.subsec_nanos() as libc::c_long + (sys_now.tv_usec * 1000) as libc::c_long; - let extra = (nsec / 1_000_000_000) as libc::time_t; - let nsec = nsec % 1_000_000_000; - let seconds = saturating_cast_to_time_t(dur.as_secs()); - - let timeout = sys_now - .tv_sec - .checked_add(extra) - .and_then(|s| s.checked_add(seconds)) - .map(|s| libc::timespec { tv_sec: s, tv_nsec: nsec }) + // OSX implementation of `pthread_cond_timedwait` is buggy + // with super long durations. When duration is greater than + // 0x100_0000_0000_0000 seconds, `pthread_cond_timedwait` + // in macOS Sierra returns error 316. + // + // This program demonstrates the issue: + // https://gist.github.com/stepancheg/198db4623a20aad2ad7cddb8fda4a63c + // + // To work around this issue, and possible bugs of other OSes, timeout + // is clamped to 1000 years, which is allowable per the API of `wait_timeout` + // because of spurious wakeups. + let dur = Duration::min(dur, Duration::from_secs(1000 * 365 * 86400)); + + // pthread_cond_timedwait uses system time, but we want to report timeout + // based on stable time. + let now = Instant::now(); + + let timeout = SystemTime::now() + .t + .checked_add_duration(&dur) + .and_then(|t| t.to_timespec()) .unwrap_or(TIMESPEC_MAX); - // And wait! let r = libc::pthread_cond_timedwait(raw(self), mutex, &timeout); debug_assert!(r == libc::ETIMEDOUT || r == 0); // ETIMEDOUT is not a totally reliable method of determining timeout due // to clock shifts, so do the check ourselves - stable_now.elapsed() < dur + now.elapsed() < dur } } diff --git a/library/std/src/sys/unix/mod.rs b/library/std/src/sys/unix/mod.rs index 9055a011c..30a96be14 100644 --- a/library/std/src/sys/unix/mod.rs +++ b/library/std/src/sys/unix/mod.rs @@ -40,7 +40,7 @@ pub mod stdio; pub mod thread; pub mod thread_local_dtor; pub mod thread_local_key; -pub mod thread_parker; +pub mod thread_parking; pub mod time; #[cfg(target_os = "espidf")] @@ -95,6 +95,10 @@ pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { )))] 'poll: { use crate::sys::os::errno; + #[cfg(not(all(target_os = "linux", target_env = "gnu")))] + use libc::open as open64; + #[cfg(all(target_os = "linux", target_env = "gnu"))] + use libc::open64; let pfds: &mut [_] = &mut [ libc::pollfd { fd: 0, events: 0, revents: 0 }, libc::pollfd { fd: 1, events: 0, revents: 0 }, @@ -116,7 +120,7 @@ pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { if pfd.revents & libc::POLLNVAL == 0 { continue; } - if libc::open("/dev/null\0".as_ptr().cast(), libc::O_RDWR, 0) == -1 { + if open64("/dev/null\0".as_ptr().cast(), libc::O_RDWR, 0) == -1 { // If the stream is closed but we failed to reopen it, abort the // process. Otherwise we wouldn't preserve the safety of // operations on the corresponding Rust object Stdin, Stdout, or @@ -139,9 +143,13 @@ pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { )))] { use crate::sys::os::errno; + #[cfg(not(all(target_os = "linux", target_env = "gnu")))] + use libc::open as open64; + #[cfg(all(target_os = "linux", target_env = "gnu"))] + use libc::open64; for fd in 0..3 { if libc::fcntl(fd, libc::F_GETFD) == -1 && errno() == libc::EBADF { - if libc::open("/dev/null\0".as_ptr().cast(), libc::O_RDWR, 0) == -1 { + if open64("/dev/null\0".as_ptr().cast(), libc::O_RDWR, 0) == -1 { // If the stream is closed but we failed to reopen it, abort the // process. Otherwise we wouldn't preserve the safety of // operations on the corresponding Rust object Stdin, Stdout, or @@ -156,7 +164,7 @@ pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { unsafe fn reset_sigpipe(#[allow(unused_variables)] sigpipe: u8) { #[cfg(not(any(target_os = "emscripten", target_os = "fuchsia", target_os = "horizon")))] { - // We don't want to add this as a public type to libstd, nor do we + // We don't want to add this as a public type to std, nor do we // want to `include!` a file from the compiler (which would break // Miri and xargo for example), so we choose to duplicate these // constants from `compiler/rustc_session/src/config/sigpipe.rs`. @@ -176,12 +184,7 @@ pub unsafe fn init(argc: isize, argv: *const *const u8, sigpipe: u8) { sigpipe::SIG_DFL => (true, Some(libc::SIG_DFL)), _ => unreachable!(), }; - // The bootstrap compiler doesn't know about sigpipe::DEFAULT, and always passes in - // SIG_IGN. This causes some tests to fail because they expect SIGPIPE to be reset to - // default on process spawning (which doesn't happen if #[unix_sigpipe] is specified). - // Since we can't differentiate between the cases here, treat SIG_IGN as DEFAULT - // unconditionally. - if sigpipe_attr_specified && !(cfg!(bootstrap) && sigpipe == sigpipe::SIG_IGN) { + if sigpipe_attr_specified { UNIX_SIGPIPE_ATTR_SPECIFIED.store(true, crate::sync::atomic::Ordering::Relaxed); } if let Some(handler) = handler { diff --git a/library/std/src/sys/unix/net.rs b/library/std/src/sys/unix/net.rs index b84bf8f92..c86f80972 100644 --- a/library/std/src/sys/unix/net.rs +++ b/library/std/src/sys/unix/net.rs @@ -512,7 +512,7 @@ impl FromRawFd for Socket { // A workaround for this bug is to call the res_init libc function, to clear // the cached configs. Unfortunately, while we believe glibc's implementation // of res_init is thread-safe, we know that other implementations are not -// (https://github.com/rust-lang/rust/issues/43592). Code here in libstd could +// (https://github.com/rust-lang/rust/issues/43592). Code here in std could // try to synchronize its res_init calls with a Mutex, but that wouldn't // protect programs that call into libc in other ways. So instead of calling // res_init unconditionally, we call it only when we detect we're linking diff --git a/library/std/src/sys/unix/pipe.rs b/library/std/src/sys/unix/pipe.rs index a56c275c9..a744d0ab6 100644 --- a/library/std/src/sys/unix/pipe.rs +++ b/library/std/src/sys/unix/pipe.rs @@ -58,6 +58,10 @@ impl AnonPipe { self.0.is_read_vectored() } + pub fn read_to_end(&self, buf: &mut Vec) -> io::Result { + self.0.read_to_end(buf) + } + pub fn write(&self, buf: &[u8]) -> io::Result { self.0.write(buf) } diff --git a/library/std/src/sys/unix/process/process_common.rs b/library/std/src/sys/unix/process/process_common.rs index 848adca78..afd03d79c 100644 --- a/library/std/src/sys/unix/process/process_common.rs +++ b/library/std/src/sys/unix/process/process_common.rs @@ -144,6 +144,7 @@ pub enum ChildStdio { Null, } +#[derive(Debug)] pub enum Stdio { Inherit, Null, @@ -510,16 +511,68 @@ impl ChildStdio { } impl fmt::Debug for Command { + // show all attributes but `self.closures` which does not implement `Debug` + // and `self.argv` which is not useful for debugging fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - if self.program != self.args[0] { - write!(f, "[{:?}] ", self.program)?; - } - write!(f, "{:?}", self.args[0])?; + if f.alternate() { + let mut debug_command = f.debug_struct("Command"); + debug_command.field("program", &self.program).field("args", &self.args); + if !self.env.is_unchanged() { + debug_command.field("env", &self.env); + } + + if self.cwd.is_some() { + debug_command.field("cwd", &self.cwd); + } + if self.uid.is_some() { + debug_command.field("uid", &self.uid); + } + if self.gid.is_some() { + debug_command.field("gid", &self.gid); + } + + if self.groups.is_some() { + debug_command.field("groups", &self.groups); + } + + if self.stdin.is_some() { + debug_command.field("stdin", &self.stdin); + } + if self.stdout.is_some() { + debug_command.field("stdout", &self.stdout); + } + if self.stderr.is_some() { + debug_command.field("stderr", &self.stderr); + } + if self.pgroup.is_some() { + debug_command.field("pgroup", &self.pgroup); + } + + #[cfg(target_os = "linux")] + { + debug_command.field("create_pidfd", &self.create_pidfd); + } - for arg in &self.args[1..] { - write!(f, " {:?}", arg)?; + debug_command.finish() + } else { + if let Some(ref cwd) = self.cwd { + write!(f, "cd {cwd:?} && ")?; + } + for (key, value_opt) in self.get_envs() { + if let Some(value) = value_opt { + write!(f, "{}={value:?} ", key.to_string_lossy())?; + } + } + if self.program != self.args[0] { + write!(f, "[{:?}] ", self.program)?; + } + write!(f, "{:?}", self.args[0])?; + + for arg in &self.args[1..] { + write!(f, " {:?}", arg)?; + } + Ok(()) } - Ok(()) } } diff --git a/library/std/src/sys/unix/process/process_fuchsia.rs b/library/std/src/sys/unix/process/process_fuchsia.rs index 66ea3db20..d4c7e58b3 100644 --- a/library/std/src/sys/unix/process/process_fuchsia.rs +++ b/library/std/src/sys/unix/process/process_fuchsia.rs @@ -35,6 +35,11 @@ impl Command { Ok((Process { handle: Handle::new(process_handle) }, ours)) } + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + let (proc, pipes) = self.spawn(Stdio::MakePipe, false)?; + crate::sys_common::process::wait_with_output(proc, pipes) + } + pub fn exec(&mut self, default: Stdio) -> io::Error { if self.saw_nul() { return io::const_io_error!( @@ -257,7 +262,7 @@ impl ExitStatus { // available on Fuchsia. // // It does not appear that Fuchsia is Unix-like enough to implement ExitStatus (or indeed many - // other things from std::os::unix) properly. This veneer is always going to be a bodge. So + // other things from std::os::unix) properly. This veneer is always going to be a bodge. So // while I don't know if these implementations are actually correct, I think they will do for // now at least. pub fn core_dumped(&self) -> bool { @@ -272,9 +277,9 @@ impl ExitStatus { pub fn into_raw(&self) -> c_int { // We don't know what someone who calls into_raw() will do with this value, but it should - // have the conventional Unix representation. Despite the fact that this is not + // have the conventional Unix representation. Despite the fact that this is not // standardised in SuS or POSIX, all Unix systems encode the signal and exit status the - // same way. (Ie the WIFEXITED, WEXITSTATUS etc. macros have identical behaviour on every + // same way. (Ie the WIFEXITED, WEXITSTATUS etc. macros have identical behaviour on every // Unix.) // // The caller of `std::os::unix::into_raw` is probably wanting a Unix exit status, and may @@ -282,14 +287,14 @@ impl ExitStatus { // different Unix variant. // // The other view would be to say that the caller on Fuchsia ought to know that `into_raw` - // will give a raw Fuchsia status (whatever that is - I don't know, personally). That is + // will give a raw Fuchsia status (whatever that is - I don't know, personally). That is // not possible here because we must return a c_int because that's what Unix (including // SuS and POSIX) say a wait status is, but Fuchsia apparently uses a u64, so it won't // necessarily fit. // // It seems to me that the right answer would be to provide std::os::fuchsia with its // own ExitStatusExt, rather that trying to provide a not very convincing imitation of - // Unix. Ie, std::os::unix::process:ExitStatusExt ought not to exist on Fuchsia. But + // Unix. Ie, std::os::unix::process:ExitStatusExt ought not to exist on Fuchsia. But // fixing this up that is beyond the scope of my efforts now. let exit_status_as_if_unix: u8 = self.0.try_into().expect("Fuchsia process return code bigger than 8 bits, but std::os::unix::ExitStatusExt::into_raw() was called to try to convert the value into a traditional Unix-style wait status, which cannot represent values greater than 255."); let wait_status_as_if_unix = (exit_status_as_if_unix as c_int) << 8; diff --git a/library/std/src/sys/unix/process/process_unix.rs b/library/std/src/sys/unix/process/process_unix.rs index 56a805cef..3bc17b775 100644 --- a/library/std/src/sys/unix/process/process_unix.rs +++ b/library/std/src/sys/unix/process/process_unix.rs @@ -66,14 +66,15 @@ impl Command { // // Note that as soon as we're done with the fork there's no need to hold // a lock any more because the parent won't do anything and the child is - // in its own process. Thus the parent drops the lock guard while the child - // forgets it to avoid unlocking it on a new thread, which would be invalid. + // in its own process. Thus the parent drops the lock guard immediately. + // The child calls `mem::forget` to leak the lock, which is crucial because + // releasing a lock is not async-signal-safe. let env_lock = sys::os::env_read_lock(); let (pid, pidfd) = unsafe { self.do_fork()? }; if pid == 0 { crate::panic::always_abort(); - mem::forget(env_lock); + mem::forget(env_lock); // avoid non-async-signal-safe unlocking drop(input); let Err(err) = unsafe { self.do_exec(theirs, envp.as_ref()) }; let errno = err.raw_os_error().unwrap_or(libc::EINVAL) as u32; @@ -132,6 +133,11 @@ impl Command { } } + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + let (proc, pipes) = self.spawn(Stdio::MakePipe, false)?; + crate::sys_common::process::wait_with_output(proc, pipes) + } + // Attempts to fork the process. If successful, returns Ok((0, -1)) // in the child, and Ok((child_pid, -1)) in the parent. #[cfg(not(target_os = "linux"))] @@ -660,11 +666,11 @@ impl ExitStatus { } pub fn exit_ok(&self) -> Result<(), ExitStatusError> { - // This assumes that WIFEXITED(status) && WEXITSTATUS==0 corresponds to status==0. This is + // This assumes that WIFEXITED(status) && WEXITSTATUS==0 corresponds to status==0. This is // true on all actual versions of Unix, is widely assumed, and is specified in SuS - // https://pubs.opengroup.org/onlinepubs/9699919799/functions/wait.html . If it is not + // https://pubs.opengroup.org/onlinepubs/9699919799/functions/wait.html. If it is not // true for a platform pretending to be Unix, the tests (our doctests, and also - // procsss_unix/tests.rs) will spot it. `ExitStatusError::code` assumes this too. + // procsss_unix/tests.rs) will spot it. `ExitStatusError::code` assumes this too. match NonZero_c_int::try_from(self.0) { /* was nonzero */ Ok(failure) => Err(ExitStatusError(failure)), /* was zero, couldn't convert */ Err(_) => Ok(()), @@ -740,6 +746,8 @@ fn signal_string(signal: i32) -> &'static str { libc::SIGWINCH => " (SIGWINCH)", #[cfg(not(target_os = "haiku"))] libc::SIGIO => " (SIGIO)", + #[cfg(target_os = "haiku")] + libc::SIGPOLL => " (SIGPOLL)", libc::SIGSYS => " (SIGSYS)", // For information on Linux signals, run `man 7 signal` #[cfg(all( diff --git a/library/std/src/sys/unix/process/process_unix/tests.rs b/library/std/src/sys/unix/process/process_unix/tests.rs index e0e2d478f..e5e1f956b 100644 --- a/library/std/src/sys/unix/process/process_unix/tests.rs +++ b/library/std/src/sys/unix/process/process_unix/tests.rs @@ -3,7 +3,7 @@ use crate::panic::catch_unwind; use crate::process::Command; // Many of the other aspects of this situation, including heap alloc concurrency -// safety etc., are tested in src/test/ui/process/process-panic-after-fork.rs +// safety etc., are tested in tests/ui/process/process-panic-after-fork.rs #[test] fn exitstatus_display_tests() { @@ -19,17 +19,17 @@ fn exitstatus_display_tests() { t(0x00000, "exit status: 0"); t(0x0ff00, "exit status: 255"); - // On MacOS, 0x0137f is WIFCONTINUED, not WIFSTOPPED. Probably *BSD is similar. + // On MacOS, 0x0137f is WIFCONTINUED, not WIFSTOPPED. Probably *BSD is similar. // https://github.com/rust-lang/rust/pull/82749#issuecomment-790525956 // The purpose of this test is to test our string formatting, not our understanding of the wait - // status magic numbers. So restrict these to Linux. + // status magic numbers. So restrict these to Linux. if cfg!(target_os = "linux") { t(0x0137f, "stopped (not terminated) by signal: 19 (SIGSTOP)"); t(0x0ffff, "continued (WIFCONTINUED)"); } // Testing "unrecognised wait status" is hard because the wait.h macros typically - // assume that the value came from wait and isn't mad. With the glibc I have here + // assume that the value came from wait and isn't mad. With the glibc I have here // this works: if cfg!(all(target_os = "linux", target_env = "gnu")) { t(0x000ff, "unrecognised wait status: 255 0xff"); diff --git a/library/std/src/sys/unix/process/process_unsupported.rs b/library/std/src/sys/unix/process/process_unsupported.rs index 72f9f3f9c..f28ca58d0 100644 --- a/library/std/src/sys/unix/process/process_unsupported.rs +++ b/library/std/src/sys/unix/process/process_unsupported.rs @@ -20,6 +20,10 @@ impl Command { unsupported() } + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + unsupported() + } + pub fn exec(&mut self, _default: Stdio) -> io::Error { unsupported_err() } diff --git a/library/std/src/sys/unix/process/process_vxworks.rs b/library/std/src/sys/unix/process/process_vxworks.rs index 200ef6719..569a4b149 100644 --- a/library/std/src/sys/unix/process/process_vxworks.rs +++ b/library/std/src/sys/unix/process/process_vxworks.rs @@ -108,6 +108,11 @@ impl Command { } } + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + let (proc, pipes) = self.spawn(Stdio::MakePipe, false)?; + crate::sys_common::process::wait_with_output(proc, pipes) + } + pub fn exec(&mut self, default: Stdio) -> io::Error { let ret = Command::spawn(self, default, false); match ret { @@ -190,11 +195,11 @@ impl ExitStatus { } pub fn exit_ok(&self) -> Result<(), ExitStatusError> { - // This assumes that WIFEXITED(status) && WEXITSTATUS==0 corresponds to status==0. This is + // This assumes that WIFEXITED(status) && WEXITSTATUS==0 corresponds to status==0. This is // true on all actual versions of Unix, is widely assumed, and is specified in SuS - // https://pubs.opengroup.org/onlinepubs/9699919799/functions/wait.html . If it is not + // https://pubs.opengroup.org/onlinepubs/9699919799/functions/wait.html. If it is not // true for a platform pretending to be Unix, the tests (our doctests, and also - // procsss_unix/tests.rs) will spot it. `ExitStatusError::code` assumes this too. + // procsss_unix/tests.rs) will spot it. `ExitStatusError::code` assumes this too. match NonZero_c_int::try_from(self.0) { Ok(failure) => Err(ExitStatusError(failure)), Err(_) => Ok(()), diff --git a/library/std/src/sys/unix/stack_overflow.rs b/library/std/src/sys/unix/stack_overflow.rs index 75a5c0f92..b59d4ba26 100644 --- a/library/std/src/sys/unix/stack_overflow.rs +++ b/library/std/src/sys/unix/stack_overflow.rs @@ -45,7 +45,10 @@ mod imp { use crate::thread; use libc::MAP_FAILED; - use libc::{mmap, munmap}; + #[cfg(not(all(target_os = "linux", target_env = "gnu")))] + use libc::{mmap as mmap64, munmap}; + #[cfg(all(target_os = "linux", target_env = "gnu"))] + use libc::{mmap64, munmap}; use libc::{sigaction, sighandler_t, SA_ONSTACK, SA_SIGINFO, SIGBUS, SIG_DFL}; use libc::{sigaltstack, SIGSTKSZ, SS_DISABLE}; use libc::{MAP_ANON, MAP_PRIVATE, PROT_NONE, PROT_READ, PROT_WRITE, SIGSEGV}; @@ -135,7 +138,7 @@ mod imp { #[cfg(not(any(target_os = "openbsd", target_os = "netbsd", target_os = "linux",)))] let flags = MAP_PRIVATE | MAP_ANON; let stackp = - mmap(ptr::null_mut(), SIGSTKSZ + page_size(), PROT_READ | PROT_WRITE, flags, -1, 0); + mmap64(ptr::null_mut(), SIGSTKSZ + page_size(), PROT_READ | PROT_WRITE, flags, -1, 0); if stackp == MAP_FAILED { panic!("failed to allocate an alternative stack: {}", io::Error::last_os_error()); } diff --git a/library/std/src/sys/unix/thread.rs b/library/std/src/sys/unix/thread.rs index c1d30dd9d..2a1830d06 100644 --- a/library/std/src/sys/unix/thread.rs +++ b/library/std/src/sys/unix/thread.rs @@ -73,7 +73,7 @@ impl Thread { n => { assert_eq!(n, libc::EINVAL); // EINVAL means |stack_size| is either too small or not a - // multiple of the system page size. Because it's definitely + // multiple of the system page size. Because it's definitely // >= PTHREAD_STACK_MIN, it must be an alignment issue. // Round up to the nearest page and try again. let page_size = os::page_size(); @@ -136,7 +136,7 @@ impl Thread { unsafe { // Available since glibc 2.12, musl 1.1.16, and uClibc 1.0.20. - let name = truncate_cstr(name, TASK_COMM_LEN); + let name = truncate_cstr::<{ TASK_COMM_LEN }>(name); let res = libc::pthread_setname_np(libc::pthread_self(), name.as_ptr()); // We have no good way of propagating errors here, but in debug-builds let's check that this actually worked. debug_assert_eq!(res, 0); @@ -153,7 +153,7 @@ impl Thread { #[cfg(any(target_os = "macos", target_os = "ios", target_os = "watchos"))] pub fn set_name(name: &CStr) { unsafe { - let name = truncate_cstr(name, libc::MAXTHREADNAMESIZE); + let name = truncate_cstr::<{ libc::MAXTHREADNAMESIZE }>(name); let res = libc::pthread_setname_np(name.as_ptr()); // We have no good way of propagating errors here, but in debug-builds let's check that this actually worked. debug_assert_eq!(res, 0); @@ -285,17 +285,12 @@ impl Drop for Thread { } #[cfg(any(target_os = "linux", target_os = "macos", target_os = "ios", target_os = "watchos"))] -fn truncate_cstr(cstr: &CStr, max_with_nul: usize) -> crate::borrow::Cow<'_, CStr> { - use crate::{borrow::Cow, ffi::CString}; - - if cstr.to_bytes_with_nul().len() > max_with_nul { - let bytes = cstr.to_bytes()[..max_with_nul - 1].to_vec(); - // SAFETY: the non-nul bytes came straight from a CStr. - // (CString will add the terminating nul.) - Cow::Owned(unsafe { CString::from_vec_unchecked(bytes) }) - } else { - Cow::Borrowed(cstr) +fn truncate_cstr(cstr: &CStr) -> [libc::c_char; MAX_WITH_NUL] { + let mut result = [0; MAX_WITH_NUL]; + for (src, dst) in cstr.to_bytes().iter().zip(&mut result[..MAX_WITH_NUL - 1]) { + *dst = *src as libc::c_char; } + result } pub fn available_parallelism() -> io::Result { @@ -510,7 +505,7 @@ mod cgroups { let limit = raw_quota.next()?; let period = raw_quota.next()?; match (limit.parse::(), period.parse::()) { - (Ok(limit), Ok(period)) => { + (Ok(limit), Ok(period)) if period > 0 => { quota = quota.min(limit / period); } _ => {} @@ -570,7 +565,7 @@ mod cgroups { let period = parse_file("cpu.cfs_period_us"); match (limit, period) { - (Some(limit), Some(period)) => quota = quota.min(limit / period), + (Some(limit), Some(period)) if period > 0 => quota = quota.min(limit / period), _ => {} } @@ -658,7 +653,10 @@ pub mod guard { ))] #[cfg_attr(test, allow(dead_code))] pub mod guard { - use libc::{mmap, mprotect}; + #[cfg(not(all(target_os = "linux", target_env = "gnu")))] + use libc::{mmap as mmap64, mprotect}; + #[cfg(all(target_os = "linux", target_env = "gnu"))] + use libc::{mmap64, mprotect}; use libc::{MAP_ANON, MAP_FAILED, MAP_FIXED, MAP_PRIVATE, PROT_NONE, PROT_READ, PROT_WRITE}; use crate::io; @@ -757,10 +755,10 @@ pub mod guard { if cfg!(all(target_os = "linux", not(target_env = "musl"))) { // Linux doesn't allocate the whole stack right away, and // the kernel has its own stack-guard mechanism to fault - // when growing too close to an existing mapping. If we map + // when growing too close to an existing mapping. If we map // our own guard, then the kernel starts enforcing a rather // large gap above that, rendering much of the possible - // stack space useless. See #43052. + // stack space useless. See #43052. // // Instead, we'll just note where we expect rlimit to start // faulting, so our handler can report "stack overflow", and @@ -776,14 +774,14 @@ pub mod guard { None } else if cfg!(target_os = "freebsd") { // FreeBSD's stack autogrows, and optionally includes a guard page - // at the bottom. If we try to remap the bottom of the stack - // ourselves, FreeBSD's guard page moves upwards. So we'll just use + // at the bottom. If we try to remap the bottom of the stack + // ourselves, FreeBSD's guard page moves upwards. So we'll just use // the builtin guard page. let stackptr = get_stack_start_aligned()?; let guardaddr = stackptr.addr(); // Technically the number of guard pages is tunable and controlled // by the security.bsd.stack_guard_page sysctl, but there are - // few reasons to change it from the default. The default value has + // few reasons to change it from the default. The default value has // been 1 ever since FreeBSD 11.1 and 10.4. const GUARD_PAGES: usize = 1; let guard = guardaddr..guardaddr + GUARD_PAGES * page_size; @@ -808,7 +806,7 @@ pub mod guard { // read/write permissions and only then mprotect() it to // no permissions at all. See issue #50313. let stackptr = get_stack_start_aligned()?; - let result = mmap( + let result = mmap64( stackptr, page_size, PROT_READ | PROT_WRITE, @@ -879,9 +877,9 @@ pub mod guard { } else if cfg!(all(target_os = "linux", any(target_env = "gnu", target_env = "uclibc"))) { // glibc used to include the guard area within the stack, as noted in the BUGS - // section of `man pthread_attr_getguardsize`. This has been corrected starting + // section of `man pthread_attr_getguardsize`. This has been corrected starting // with glibc 2.27, and in some distro backports, so the guard is now placed at the - // end (below) the stack. There's no easy way for us to know which we have at + // end (below) the stack. There's no easy way for us to know which we have at // runtime, so we'll just match any fault in the range right above or below the // stack base to call that fault a stack overflow. Some(stackaddr - guardsize..stackaddr + guardsize) diff --git a/library/std/src/sys/unix/thread_parker/darwin.rs b/library/std/src/sys/unix/thread_parker/darwin.rs deleted file mode 100644 index 2f5356fe2..000000000 --- a/library/std/src/sys/unix/thread_parker/darwin.rs +++ /dev/null @@ -1,131 +0,0 @@ -//! Thread parking for Darwin-based systems. -//! -//! Darwin actually has futex syscalls (`__ulock_wait`/`__ulock_wake`), but they -//! cannot be used in `std` because they are non-public (their use will lead to -//! rejection from the App Store) and because they are only available starting -//! with macOS version 10.12, even though the minimum target version is 10.7. -//! -//! Therefore, we need to look for other synchronization primitives. Luckily, Darwin -//! supports semaphores, which allow us to implement the behaviour we need with -//! only one primitive (as opposed to a mutex-condvar pair). We use the semaphore -//! provided by libdispatch, as the underlying Mach semaphore is only dubiously -//! public. - -use crate::pin::Pin; -use crate::sync::atomic::{ - AtomicI8, - Ordering::{Acquire, Release}, -}; -use crate::time::Duration; - -type dispatch_semaphore_t = *mut crate::ffi::c_void; -type dispatch_time_t = u64; - -const DISPATCH_TIME_NOW: dispatch_time_t = 0; -const DISPATCH_TIME_FOREVER: dispatch_time_t = !0; - -// Contained in libSystem.dylib, which is linked by default. -extern "C" { - fn dispatch_time(when: dispatch_time_t, delta: i64) -> dispatch_time_t; - fn dispatch_semaphore_create(val: isize) -> dispatch_semaphore_t; - fn dispatch_semaphore_wait(dsema: dispatch_semaphore_t, timeout: dispatch_time_t) -> isize; - fn dispatch_semaphore_signal(dsema: dispatch_semaphore_t) -> isize; - fn dispatch_release(object: *mut crate::ffi::c_void); -} - -const EMPTY: i8 = 0; -const NOTIFIED: i8 = 1; -const PARKED: i8 = -1; - -pub struct Parker { - semaphore: dispatch_semaphore_t, - state: AtomicI8, -} - -unsafe impl Sync for Parker {} -unsafe impl Send for Parker {} - -impl Parker { - pub unsafe fn new(parker: *mut Parker) { - let semaphore = dispatch_semaphore_create(0); - assert!( - !semaphore.is_null(), - "failed to create dispatch semaphore for thread synchronization" - ); - parker.write(Parker { semaphore, state: AtomicI8::new(EMPTY) }) - } - - // Does not need `Pin`, but other implementation do. - pub unsafe fn park(self: Pin<&Self>) { - // The semaphore counter must be zero at this point, because unparking - // threads will not actually increase it until we signalled that we - // are waiting. - - // Change NOTIFIED to EMPTY and EMPTY to PARKED. - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - - // Another thread may increase the semaphore counter from this point on. - // If it is faster than us, we will decrement it again immediately below. - // If we are faster, we wait. - - // Ensure that the semaphore counter has actually been decremented, even - // if the call timed out for some reason. - while dispatch_semaphore_wait(self.semaphore, DISPATCH_TIME_FOREVER) != 0 {} - - // At this point, the semaphore counter is zero again. - - // We were definitely woken up, so we don't need to check the state. - // Still, we need to reset the state using a swap to observe the state - // change with acquire ordering. - self.state.swap(EMPTY, Acquire); - } - - // Does not need `Pin`, but other implementation do. - pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - - let nanos = dur.as_nanos().try_into().unwrap_or(i64::MAX); - let timeout = dispatch_time(DISPATCH_TIME_NOW, nanos); - - let timeout = dispatch_semaphore_wait(self.semaphore, timeout) != 0; - - let state = self.state.swap(EMPTY, Acquire); - if state == NOTIFIED && timeout { - // If the state was NOTIFIED but semaphore_wait returned without - // decrementing the count because of a timeout, it means another - // thread is about to call semaphore_signal. We must wait for that - // to happen to ensure the semaphore count is reset. - while dispatch_semaphore_wait(self.semaphore, DISPATCH_TIME_FOREVER) != 0 {} - } else { - // Either a timeout occurred and we reset the state before any thread - // tried to wake us up, or we were woken up and reset the state, - // making sure to observe the state change with acquire ordering. - // Either way, the semaphore counter is now zero again. - } - } - - // Does not need `Pin`, but other implementation do. - pub fn unpark(self: Pin<&Self>) { - let state = self.state.swap(NOTIFIED, Release); - if state == PARKED { - unsafe { - dispatch_semaphore_signal(self.semaphore); - } - } - } -} - -impl Drop for Parker { - fn drop(&mut self) { - // SAFETY: - // We always ensure that the semaphore count is reset, so this will - // never cause an exception. - unsafe { - dispatch_release(self.semaphore); - } - } -} diff --git a/library/std/src/sys/unix/thread_parker/mod.rs b/library/std/src/sys/unix/thread_parker/mod.rs deleted file mode 100644 index 35f1e68a8..000000000 --- a/library/std/src/sys/unix/thread_parker/mod.rs +++ /dev/null @@ -1,32 +0,0 @@ -//! Thread parking on systems without futex support. - -#![cfg(not(any( - target_os = "linux", - target_os = "android", - all(target_os = "emscripten", target_feature = "atomics"), - target_os = "freebsd", - target_os = "openbsd", - target_os = "dragonfly", - target_os = "fuchsia", -)))] - -cfg_if::cfg_if! { - if #[cfg(all( - any( - target_os = "macos", - target_os = "ios", - target_os = "watchos", - target_os = "tvos", - ), - not(miri), - ))] { - mod darwin; - pub use darwin::Parker; - } else if #[cfg(target_os = "netbsd")] { - mod netbsd; - pub use netbsd::Parker; - } else { - mod pthread; - pub use pthread::Parker; - } -} diff --git a/library/std/src/sys/unix/thread_parker/netbsd.rs b/library/std/src/sys/unix/thread_parker/netbsd.rs deleted file mode 100644 index 7657605b5..000000000 --- a/library/std/src/sys/unix/thread_parker/netbsd.rs +++ /dev/null @@ -1,113 +0,0 @@ -use crate::ffi::{c_int, c_void}; -use crate::pin::Pin; -use crate::ptr::{null, null_mut}; -use crate::sync::atomic::{ - AtomicU64, - Ordering::{Acquire, Relaxed, Release}, -}; -use crate::time::Duration; -use libc::{_lwp_self, clockid_t, lwpid_t, time_t, timespec, CLOCK_MONOTONIC}; - -extern "C" { - fn ___lwp_park60( - clock_id: clockid_t, - flags: c_int, - ts: *mut timespec, - unpark: lwpid_t, - hint: *const c_void, - unparkhint: *const c_void, - ) -> c_int; - fn _lwp_unpark(lwp: lwpid_t, hint: *const c_void) -> c_int; -} - -/// The thread is not parked and the token is not available. -/// -/// Zero cannot be a valid LWP id, since it is used as empty value for the unpark -/// argument in _lwp_park. -const EMPTY: u64 = 0; -/// The token is available. Do not park anymore. -const NOTIFIED: u64 = u64::MAX; - -pub struct Parker { - /// The parker state. Contains either one of the two state values above or the LWP - /// id of the parked thread. - state: AtomicU64, -} - -impl Parker { - pub unsafe fn new(parker: *mut Parker) { - parker.write(Parker { state: AtomicU64::new(EMPTY) }) - } - - // Does not actually need `unsafe` or `Pin`, but the pthread implementation does. - pub unsafe fn park(self: Pin<&Self>) { - // If the token has already been made available, we can skip - // a bit of work, so check for it here. - if self.state.load(Acquire) != NOTIFIED { - let parked = _lwp_self() as u64; - let hint = self.state.as_mut_ptr().cast(); - if self.state.compare_exchange(EMPTY, parked, Relaxed, Acquire).is_ok() { - // Loop to guard against spurious wakeups. - loop { - ___lwp_park60(0, 0, null_mut(), 0, hint, null()); - if self.state.load(Acquire) == NOTIFIED { - break; - } - } - } - } - - // At this point, the change to NOTIFIED has always been observed with acquire - // ordering, so we can just use a relaxed store here (instead of a swap). - self.state.store(EMPTY, Relaxed); - } - - // Does not actually need `unsafe` or `Pin`, but the pthread implementation does. - pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { - if self.state.load(Acquire) != NOTIFIED { - let parked = _lwp_self() as u64; - let hint = self.state.as_mut_ptr().cast(); - let mut timeout = timespec { - // Saturate so that the operation will definitely time out - // (even if it is after the heat death of the universe). - tv_sec: dur.as_secs().try_into().ok().unwrap_or(time_t::MAX), - tv_nsec: dur.subsec_nanos().into(), - }; - - if self.state.compare_exchange(EMPTY, parked, Relaxed, Acquire).is_ok() { - // Timeout needs to be mutable since it is modified on NetBSD 9.0 and - // above. - ___lwp_park60(CLOCK_MONOTONIC, 0, &mut timeout, 0, hint, null()); - // Use a swap to get acquire ordering even if the token was set after - // the timeout occurred. - self.state.swap(EMPTY, Acquire); - return; - } - } - - self.state.store(EMPTY, Relaxed); - } - - // Does not actually need `Pin`, but the pthread implementation does. - pub fn unpark(self: Pin<&Self>) { - let state = self.state.swap(NOTIFIED, Release); - if !matches!(state, EMPTY | NOTIFIED) { - let lwp = state as lwpid_t; - let hint = self.state.as_mut_ptr().cast(); - - // If the parking thread terminated and did not actually park, this will - // probably return an error, which is OK. In the worst case, another - // thread has received the same LWP id. It will then receive a spurious - // wakeup, but those are allowable per the API contract. The same reasoning - // applies if a timeout occurred before this call, but the state was not - // yet reset. - - // SAFETY: - // The syscall has no invariants to hold. Only unsafe because it is an - // extern function. - unsafe { - _lwp_unpark(lwp, hint); - } - } - } -} diff --git a/library/std/src/sys/unix/thread_parker/pthread.rs b/library/std/src/sys/unix/thread_parker/pthread.rs deleted file mode 100644 index 3dfc0026e..000000000 --- a/library/std/src/sys/unix/thread_parker/pthread.rs +++ /dev/null @@ -1,271 +0,0 @@ -//! Thread parking without `futex` using the `pthread` synchronization primitives. - -use crate::cell::UnsafeCell; -use crate::marker::PhantomPinned; -use crate::pin::Pin; -use crate::ptr::addr_of_mut; -use crate::sync::atomic::AtomicUsize; -use crate::sync::atomic::Ordering::SeqCst; -use crate::time::Duration; - -const EMPTY: usize = 0; -const PARKED: usize = 1; -const NOTIFIED: usize = 2; - -unsafe fn lock(lock: *mut libc::pthread_mutex_t) { - let r = libc::pthread_mutex_lock(lock); - debug_assert_eq!(r, 0); -} - -unsafe fn unlock(lock: *mut libc::pthread_mutex_t) { - let r = libc::pthread_mutex_unlock(lock); - debug_assert_eq!(r, 0); -} - -unsafe fn notify_one(cond: *mut libc::pthread_cond_t) { - let r = libc::pthread_cond_signal(cond); - debug_assert_eq!(r, 0); -} - -unsafe fn wait(cond: *mut libc::pthread_cond_t, lock: *mut libc::pthread_mutex_t) { - let r = libc::pthread_cond_wait(cond, lock); - debug_assert_eq!(r, 0); -} - -const TIMESPEC_MAX: libc::timespec = - libc::timespec { tv_sec: ::MAX, tv_nsec: 1_000_000_000 - 1 }; - -unsafe fn wait_timeout( - cond: *mut libc::pthread_cond_t, - lock: *mut libc::pthread_mutex_t, - dur: Duration, -) { - // Use the system clock on systems that do not support pthread_condattr_setclock. - // This unfortunately results in problems when the system time changes. - #[cfg(any( - target_os = "macos", - target_os = "ios", - target_os = "watchos", - target_os = "espidf" - ))] - let (now, dur) = { - use crate::cmp::min; - use crate::sys::time::SystemTime; - - // OSX implementation of `pthread_cond_timedwait` is buggy - // with super long durations. When duration is greater than - // 0x100_0000_0000_0000 seconds, `pthread_cond_timedwait` - // in macOS Sierra return error 316. - // - // This program demonstrates the issue: - // https://gist.github.com/stepancheg/198db4623a20aad2ad7cddb8fda4a63c - // - // To work around this issue, and possible bugs of other OSes, timeout - // is clamped to 1000 years, which is allowable per the API of `park_timeout` - // because of spurious wakeups. - let dur = min(dur, Duration::from_secs(1000 * 365 * 86400)); - let now = SystemTime::now().t; - (now, dur) - }; - // Use the monotonic clock on other systems. - #[cfg(not(any( - target_os = "macos", - target_os = "ios", - target_os = "watchos", - target_os = "espidf" - )))] - let (now, dur) = { - use crate::sys::time::Timespec; - - (Timespec::now(libc::CLOCK_MONOTONIC), dur) - }; - - let timeout = - now.checked_add_duration(&dur).and_then(|t| t.to_timespec()).unwrap_or(TIMESPEC_MAX); - let r = libc::pthread_cond_timedwait(cond, lock, &timeout); - debug_assert!(r == libc::ETIMEDOUT || r == 0); -} - -pub struct Parker { - state: AtomicUsize, - lock: UnsafeCell, - cvar: UnsafeCell, - // The `pthread` primitives require a stable address, so make this struct `!Unpin`. - _pinned: PhantomPinned, -} - -impl Parker { - /// Construct the UNIX parker in-place. - /// - /// # Safety - /// The constructed parker must never be moved. - pub unsafe fn new(parker: *mut Parker) { - // Use the default mutex implementation to allow for simpler initialization. - // This could lead to undefined behaviour when deadlocking. This is avoided - // by not deadlocking. Note in particular the unlocking operation before any - // panic, as code after the panic could try to park again. - addr_of_mut!((*parker).state).write(AtomicUsize::new(EMPTY)); - addr_of_mut!((*parker).lock).write(UnsafeCell::new(libc::PTHREAD_MUTEX_INITIALIZER)); - - cfg_if::cfg_if! { - if #[cfg(any( - target_os = "macos", - target_os = "ios", - target_os = "watchos", - target_os = "l4re", - target_os = "android", - target_os = "redox" - ))] { - addr_of_mut!((*parker).cvar).write(UnsafeCell::new(libc::PTHREAD_COND_INITIALIZER)); - } else if #[cfg(any(target_os = "espidf", target_os = "horizon"))] { - let r = libc::pthread_cond_init(addr_of_mut!((*parker).cvar).cast(), crate::ptr::null()); - assert_eq!(r, 0); - } else { - use crate::mem::MaybeUninit; - let mut attr = MaybeUninit::::uninit(); - let r = libc::pthread_condattr_init(attr.as_mut_ptr()); - assert_eq!(r, 0); - let r = libc::pthread_condattr_setclock(attr.as_mut_ptr(), libc::CLOCK_MONOTONIC); - assert_eq!(r, 0); - let r = libc::pthread_cond_init(addr_of_mut!((*parker).cvar).cast(), attr.as_ptr()); - assert_eq!(r, 0); - let r = libc::pthread_condattr_destroy(attr.as_mut_ptr()); - assert_eq!(r, 0); - } - } - } - - // This implementation doesn't require `unsafe`, but other implementations - // may assume this is only called by the thread that owns the Parker. - pub unsafe fn park(self: Pin<&Self>) { - // If we were previously notified then we consume this notification and - // return quickly. - if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { - return; - } - - // Otherwise we need to coordinate going to sleep - lock(self.lock.get()); - match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { - Ok(_) => {} - Err(NOTIFIED) => { - // We must read here, even though we know it will be `NOTIFIED`. - // This is because `unpark` may have been called again since we read - // `NOTIFIED` in the `compare_exchange` above. We must perform an - // acquire operation that synchronizes with that `unpark` to observe - // any writes it made before the call to unpark. To do that we must - // read from the write it made to `state`. - let old = self.state.swap(EMPTY, SeqCst); - - unlock(self.lock.get()); - - assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); - return; - } // should consume this notification, so prohibit spurious wakeups in next park. - Err(_) => { - unlock(self.lock.get()); - - panic!("inconsistent park state") - } - } - - loop { - wait(self.cvar.get(), self.lock.get()); - - match self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst) { - Ok(_) => break, // got a notification - Err(_) => {} // spurious wakeup, go back to sleep - } - } - - unlock(self.lock.get()); - } - - // This implementation doesn't require `unsafe`, but other implementations - // may assume this is only called by the thread that owns the Parker. Use - // `Pin` to guarantee a stable address for the mutex and condition variable. - pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { - // Like `park` above we have a fast path for an already-notified thread, and - // afterwards we start coordinating for a sleep. - // return quickly. - if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { - return; - } - - lock(self.lock.get()); - match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { - Ok(_) => {} - Err(NOTIFIED) => { - // We must read again here, see `park`. - let old = self.state.swap(EMPTY, SeqCst); - unlock(self.lock.get()); - - assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); - return; - } // should consume this notification, so prohibit spurious wakeups in next park. - Err(_) => { - unlock(self.lock.get()); - panic!("inconsistent park_timeout state") - } - } - - // Wait with a timeout, and if we spuriously wake up or otherwise wake up - // from a notification we just want to unconditionally set the state back to - // empty, either consuming a notification or un-flagging ourselves as - // parked. - wait_timeout(self.cvar.get(), self.lock.get(), dur); - - match self.state.swap(EMPTY, SeqCst) { - NOTIFIED => unlock(self.lock.get()), // got a notification, hurray! - PARKED => unlock(self.lock.get()), // no notification, alas - n => { - unlock(self.lock.get()); - panic!("inconsistent park_timeout state: {n}") - } - } - } - - pub fn unpark(self: Pin<&Self>) { - // To ensure the unparked thread will observe any writes we made - // before this call, we must perform a release operation that `park` - // can synchronize with. To do that we must write `NOTIFIED` even if - // `state` is already `NOTIFIED`. That is why this must be a swap - // rather than a compare-and-swap that returns if it reads `NOTIFIED` - // on failure. - match self.state.swap(NOTIFIED, SeqCst) { - EMPTY => return, // no one was waiting - NOTIFIED => return, // already unparked - PARKED => {} // gotta go wake someone up - _ => panic!("inconsistent state in unpark"), - } - - // There is a period between when the parked thread sets `state` to - // `PARKED` (or last checked `state` in the case of a spurious wake - // up) and when it actually waits on `cvar`. If we were to notify - // during this period it would be ignored and then when the parked - // thread went to sleep it would never wake up. Fortunately, it has - // `lock` locked at this stage so we can acquire `lock` to wait until - // it is ready to receive the notification. - // - // Releasing `lock` before the call to `notify_one` means that when the - // parked thread wakes it doesn't get woken only to have to wait for us - // to release `lock`. - unsafe { - lock(self.lock.get()); - unlock(self.lock.get()); - notify_one(self.cvar.get()); - } - } -} - -impl Drop for Parker { - fn drop(&mut self) { - unsafe { - libc::pthread_cond_destroy(self.cvar.get_mut()); - libc::pthread_mutex_destroy(self.lock.get_mut()); - } - } -} - -unsafe impl Sync for Parker {} -unsafe impl Send for Parker {} diff --git a/library/std/src/sys/unix/thread_parking/darwin.rs b/library/std/src/sys/unix/thread_parking/darwin.rs new file mode 100644 index 000000000..b709fada3 --- /dev/null +++ b/library/std/src/sys/unix/thread_parking/darwin.rs @@ -0,0 +1,131 @@ +//! Thread parking for Darwin-based systems. +//! +//! Darwin actually has futex syscalls (`__ulock_wait`/`__ulock_wake`), but they +//! cannot be used in `std` because they are non-public (their use will lead to +//! rejection from the App Store) and because they are only available starting +//! with macOS version 10.12, even though the minimum target version is 10.7. +//! +//! Therefore, we need to look for other synchronization primitives. Luckily, Darwin +//! supports semaphores, which allow us to implement the behaviour we need with +//! only one primitive (as opposed to a mutex-condvar pair). We use the semaphore +//! provided by libdispatch, as the underlying Mach semaphore is only dubiously +//! public. + +use crate::pin::Pin; +use crate::sync::atomic::{ + AtomicI8, + Ordering::{Acquire, Release}, +}; +use crate::time::Duration; + +type dispatch_semaphore_t = *mut crate::ffi::c_void; +type dispatch_time_t = u64; + +const DISPATCH_TIME_NOW: dispatch_time_t = 0; +const DISPATCH_TIME_FOREVER: dispatch_time_t = !0; + +// Contained in libSystem.dylib, which is linked by default. +extern "C" { + fn dispatch_time(when: dispatch_time_t, delta: i64) -> dispatch_time_t; + fn dispatch_semaphore_create(val: isize) -> dispatch_semaphore_t; + fn dispatch_semaphore_wait(dsema: dispatch_semaphore_t, timeout: dispatch_time_t) -> isize; + fn dispatch_semaphore_signal(dsema: dispatch_semaphore_t) -> isize; + fn dispatch_release(object: *mut crate::ffi::c_void); +} + +const EMPTY: i8 = 0; +const NOTIFIED: i8 = 1; +const PARKED: i8 = -1; + +pub struct Parker { + semaphore: dispatch_semaphore_t, + state: AtomicI8, +} + +unsafe impl Sync for Parker {} +unsafe impl Send for Parker {} + +impl Parker { + pub unsafe fn new_in_place(parker: *mut Parker) { + let semaphore = dispatch_semaphore_create(0); + assert!( + !semaphore.is_null(), + "failed to create dispatch semaphore for thread synchronization" + ); + parker.write(Parker { semaphore, state: AtomicI8::new(EMPTY) }) + } + + // Does not need `Pin`, but other implementation do. + pub unsafe fn park(self: Pin<&Self>) { + // The semaphore counter must be zero at this point, because unparking + // threads will not actually increase it until we signalled that we + // are waiting. + + // Change NOTIFIED to EMPTY and EMPTY to PARKED. + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + + // Another thread may increase the semaphore counter from this point on. + // If it is faster than us, we will decrement it again immediately below. + // If we are faster, we wait. + + // Ensure that the semaphore counter has actually been decremented, even + // if the call timed out for some reason. + while dispatch_semaphore_wait(self.semaphore, DISPATCH_TIME_FOREVER) != 0 {} + + // At this point, the semaphore counter is zero again. + + // We were definitely woken up, so we don't need to check the state. + // Still, we need to reset the state using a swap to observe the state + // change with acquire ordering. + self.state.swap(EMPTY, Acquire); + } + + // Does not need `Pin`, but other implementation do. + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + + let nanos = dur.as_nanos().try_into().unwrap_or(i64::MAX); + let timeout = dispatch_time(DISPATCH_TIME_NOW, nanos); + + let timeout = dispatch_semaphore_wait(self.semaphore, timeout) != 0; + + let state = self.state.swap(EMPTY, Acquire); + if state == NOTIFIED && timeout { + // If the state was NOTIFIED but semaphore_wait returned without + // decrementing the count because of a timeout, it means another + // thread is about to call semaphore_signal. We must wait for that + // to happen to ensure the semaphore count is reset. + while dispatch_semaphore_wait(self.semaphore, DISPATCH_TIME_FOREVER) != 0 {} + } else { + // Either a timeout occurred and we reset the state before any thread + // tried to wake us up, or we were woken up and reset the state, + // making sure to observe the state change with acquire ordering. + // Either way, the semaphore counter is now zero again. + } + } + + // Does not need `Pin`, but other implementation do. + pub fn unpark(self: Pin<&Self>) { + let state = self.state.swap(NOTIFIED, Release); + if state == PARKED { + unsafe { + dispatch_semaphore_signal(self.semaphore); + } + } + } +} + +impl Drop for Parker { + fn drop(&mut self) { + // SAFETY: + // We always ensure that the semaphore count is reset, so this will + // never cause an exception. + unsafe { + dispatch_release(self.semaphore); + } + } +} diff --git a/library/std/src/sys/unix/thread_parking/mod.rs b/library/std/src/sys/unix/thread_parking/mod.rs new file mode 100644 index 000000000..185333c07 --- /dev/null +++ b/library/std/src/sys/unix/thread_parking/mod.rs @@ -0,0 +1,32 @@ +//! Thread parking on systems without futex support. + +#![cfg(not(any( + target_os = "linux", + target_os = "android", + all(target_os = "emscripten", target_feature = "atomics"), + target_os = "freebsd", + target_os = "openbsd", + target_os = "dragonfly", + target_os = "fuchsia", +)))] + +cfg_if::cfg_if! { + if #[cfg(all( + any( + target_os = "macos", + target_os = "ios", + target_os = "watchos", + target_os = "tvos", + ), + not(miri), + ))] { + mod darwin; + pub use darwin::Parker; + } else if #[cfg(target_os = "netbsd")] { + mod netbsd; + pub use netbsd::{current, park, park_timeout, unpark, ThreadId}; + } else { + mod pthread; + pub use pthread::Parker; + } +} diff --git a/library/std/src/sys/unix/thread_parking/netbsd.rs b/library/std/src/sys/unix/thread_parking/netbsd.rs new file mode 100644 index 000000000..3be081221 --- /dev/null +++ b/library/std/src/sys/unix/thread_parking/netbsd.rs @@ -0,0 +1,52 @@ +use crate::ffi::{c_int, c_void}; +use crate::ptr; +use crate::time::Duration; +use libc::{_lwp_self, clockid_t, lwpid_t, time_t, timespec, CLOCK_MONOTONIC}; + +extern "C" { + fn ___lwp_park60( + clock_id: clockid_t, + flags: c_int, + ts: *mut timespec, + unpark: lwpid_t, + hint: *const c_void, + unparkhint: *const c_void, + ) -> c_int; + fn _lwp_unpark(lwp: lwpid_t, hint: *const c_void) -> c_int; +} + +pub type ThreadId = lwpid_t; + +#[inline] +pub fn current() -> ThreadId { + unsafe { _lwp_self() } +} + +#[inline] +pub fn park(hint: usize) { + unsafe { + ___lwp_park60(0, 0, ptr::null_mut(), 0, ptr::invalid(hint), ptr::null()); + } +} + +pub fn park_timeout(dur: Duration, hint: usize) { + let mut timeout = timespec { + // Saturate so that the operation will definitely time out + // (even if it is after the heat death of the universe). + tv_sec: dur.as_secs().try_into().ok().unwrap_or(time_t::MAX), + tv_nsec: dur.subsec_nanos().into(), + }; + + // Timeout needs to be mutable since it is modified on NetBSD 9.0 and + // above. + unsafe { + ___lwp_park60(CLOCK_MONOTONIC, 0, &mut timeout, 0, ptr::invalid(hint), ptr::null()); + } +} + +#[inline] +pub fn unpark(tid: ThreadId, hint: usize) { + unsafe { + _lwp_unpark(tid, ptr::invalid(hint)); + } +} diff --git a/library/std/src/sys/unix/thread_parking/pthread.rs b/library/std/src/sys/unix/thread_parking/pthread.rs new file mode 100644 index 000000000..082d25e68 --- /dev/null +++ b/library/std/src/sys/unix/thread_parking/pthread.rs @@ -0,0 +1,271 @@ +//! Thread parking without `futex` using the `pthread` synchronization primitives. + +use crate::cell::UnsafeCell; +use crate::marker::PhantomPinned; +use crate::pin::Pin; +use crate::ptr::addr_of_mut; +use crate::sync::atomic::AtomicUsize; +use crate::sync::atomic::Ordering::SeqCst; +use crate::sys::time::TIMESPEC_MAX; +use crate::time::Duration; + +const EMPTY: usize = 0; +const PARKED: usize = 1; +const NOTIFIED: usize = 2; + +unsafe fn lock(lock: *mut libc::pthread_mutex_t) { + let r = libc::pthread_mutex_lock(lock); + debug_assert_eq!(r, 0); +} + +unsafe fn unlock(lock: *mut libc::pthread_mutex_t) { + let r = libc::pthread_mutex_unlock(lock); + debug_assert_eq!(r, 0); +} + +unsafe fn notify_one(cond: *mut libc::pthread_cond_t) { + let r = libc::pthread_cond_signal(cond); + debug_assert_eq!(r, 0); +} + +unsafe fn wait(cond: *mut libc::pthread_cond_t, lock: *mut libc::pthread_mutex_t) { + let r = libc::pthread_cond_wait(cond, lock); + debug_assert_eq!(r, 0); +} + +unsafe fn wait_timeout( + cond: *mut libc::pthread_cond_t, + lock: *mut libc::pthread_mutex_t, + dur: Duration, +) { + // Use the system clock on systems that do not support pthread_condattr_setclock. + // This unfortunately results in problems when the system time changes. + #[cfg(any( + target_os = "macos", + target_os = "ios", + target_os = "watchos", + target_os = "espidf", + target_os = "horizon", + ))] + let (now, dur) = { + use crate::cmp::min; + use crate::sys::time::SystemTime; + + // OSX implementation of `pthread_cond_timedwait` is buggy + // with super long durations. When duration is greater than + // 0x100_0000_0000_0000 seconds, `pthread_cond_timedwait` + // in macOS Sierra return error 316. + // + // This program demonstrates the issue: + // https://gist.github.com/stepancheg/198db4623a20aad2ad7cddb8fda4a63c + // + // To work around this issue, and possible bugs of other OSes, timeout + // is clamped to 1000 years, which is allowable per the API of `park_timeout` + // because of spurious wakeups. + let dur = min(dur, Duration::from_secs(1000 * 365 * 86400)); + let now = SystemTime::now().t; + (now, dur) + }; + // Use the monotonic clock on other systems. + #[cfg(not(any( + target_os = "macos", + target_os = "ios", + target_os = "watchos", + target_os = "espidf", + target_os = "horizon", + )))] + let (now, dur) = { + use crate::sys::time::Timespec; + + (Timespec::now(libc::CLOCK_MONOTONIC), dur) + }; + + let timeout = + now.checked_add_duration(&dur).and_then(|t| t.to_timespec()).unwrap_or(TIMESPEC_MAX); + let r = libc::pthread_cond_timedwait(cond, lock, &timeout); + debug_assert!(r == libc::ETIMEDOUT || r == 0); +} + +pub struct Parker { + state: AtomicUsize, + lock: UnsafeCell, + cvar: UnsafeCell, + // The `pthread` primitives require a stable address, so make this struct `!Unpin`. + _pinned: PhantomPinned, +} + +impl Parker { + /// Construct the UNIX parker in-place. + /// + /// # Safety + /// The constructed parker must never be moved. + pub unsafe fn new_in_place(parker: *mut Parker) { + // Use the default mutex implementation to allow for simpler initialization. + // This could lead to undefined behaviour when deadlocking. This is avoided + // by not deadlocking. Note in particular the unlocking operation before any + // panic, as code after the panic could try to park again. + addr_of_mut!((*parker).state).write(AtomicUsize::new(EMPTY)); + addr_of_mut!((*parker).lock).write(UnsafeCell::new(libc::PTHREAD_MUTEX_INITIALIZER)); + + cfg_if::cfg_if! { + if #[cfg(any( + target_os = "macos", + target_os = "ios", + target_os = "watchos", + target_os = "l4re", + target_os = "android", + target_os = "redox" + ))] { + addr_of_mut!((*parker).cvar).write(UnsafeCell::new(libc::PTHREAD_COND_INITIALIZER)); + } else if #[cfg(any(target_os = "espidf", target_os = "horizon"))] { + let r = libc::pthread_cond_init(addr_of_mut!((*parker).cvar).cast(), crate::ptr::null()); + assert_eq!(r, 0); + } else { + use crate::mem::MaybeUninit; + let mut attr = MaybeUninit::::uninit(); + let r = libc::pthread_condattr_init(attr.as_mut_ptr()); + assert_eq!(r, 0); + let r = libc::pthread_condattr_setclock(attr.as_mut_ptr(), libc::CLOCK_MONOTONIC); + assert_eq!(r, 0); + let r = libc::pthread_cond_init(addr_of_mut!((*parker).cvar).cast(), attr.as_ptr()); + assert_eq!(r, 0); + let r = libc::pthread_condattr_destroy(attr.as_mut_ptr()); + assert_eq!(r, 0); + } + } + } + + // This implementation doesn't require `unsafe`, but other implementations + // may assume this is only called by the thread that owns the Parker. + pub unsafe fn park(self: Pin<&Self>) { + // If we were previously notified then we consume this notification and + // return quickly. + if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { + return; + } + + // Otherwise we need to coordinate going to sleep + lock(self.lock.get()); + match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { + Ok(_) => {} + Err(NOTIFIED) => { + // We must read here, even though we know it will be `NOTIFIED`. + // This is because `unpark` may have been called again since we read + // `NOTIFIED` in the `compare_exchange` above. We must perform an + // acquire operation that synchronizes with that `unpark` to observe + // any writes it made before the call to unpark. To do that we must + // read from the write it made to `state`. + let old = self.state.swap(EMPTY, SeqCst); + + unlock(self.lock.get()); + + assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); + return; + } // should consume this notification, so prohibit spurious wakeups in next park. + Err(_) => { + unlock(self.lock.get()); + + panic!("inconsistent park state") + } + } + + loop { + wait(self.cvar.get(), self.lock.get()); + + match self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst) { + Ok(_) => break, // got a notification + Err(_) => {} // spurious wakeup, go back to sleep + } + } + + unlock(self.lock.get()); + } + + // This implementation doesn't require `unsafe`, but other implementations + // may assume this is only called by the thread that owns the Parker. Use + // `Pin` to guarantee a stable address for the mutex and condition variable. + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + // Like `park` above we have a fast path for an already-notified thread, and + // afterwards we start coordinating for a sleep. + // return quickly. + if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { + return; + } + + lock(self.lock.get()); + match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { + Ok(_) => {} + Err(NOTIFIED) => { + // We must read again here, see `park`. + let old = self.state.swap(EMPTY, SeqCst); + unlock(self.lock.get()); + + assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); + return; + } // should consume this notification, so prohibit spurious wakeups in next park. + Err(_) => { + unlock(self.lock.get()); + panic!("inconsistent park_timeout state") + } + } + + // Wait with a timeout, and if we spuriously wake up or otherwise wake up + // from a notification we just want to unconditionally set the state back to + // empty, either consuming a notification or un-flagging ourselves as + // parked. + wait_timeout(self.cvar.get(), self.lock.get(), dur); + + match self.state.swap(EMPTY, SeqCst) { + NOTIFIED => unlock(self.lock.get()), // got a notification, hurray! + PARKED => unlock(self.lock.get()), // no notification, alas + n => { + unlock(self.lock.get()); + panic!("inconsistent park_timeout state: {n}") + } + } + } + + pub fn unpark(self: Pin<&Self>) { + // To ensure the unparked thread will observe any writes we made + // before this call, we must perform a release operation that `park` + // can synchronize with. To do that we must write `NOTIFIED` even if + // `state` is already `NOTIFIED`. That is why this must be a swap + // rather than a compare-and-swap that returns if it reads `NOTIFIED` + // on failure. + match self.state.swap(NOTIFIED, SeqCst) { + EMPTY => return, // no one was waiting + NOTIFIED => return, // already unparked + PARKED => {} // gotta go wake someone up + _ => panic!("inconsistent state in unpark"), + } + + // There is a period between when the parked thread sets `state` to + // `PARKED` (or last checked `state` in the case of a spurious wake + // up) and when it actually waits on `cvar`. If we were to notify + // during this period it would be ignored and then when the parked + // thread went to sleep it would never wake up. Fortunately, it has + // `lock` locked at this stage so we can acquire `lock` to wait until + // it is ready to receive the notification. + // + // Releasing `lock` before the call to `notify_one` means that when the + // parked thread wakes it doesn't get woken only to have to wait for us + // to release `lock`. + unsafe { + lock(self.lock.get()); + unlock(self.lock.get()); + notify_one(self.cvar.get()); + } + } +} + +impl Drop for Parker { + fn drop(&mut self) { + unsafe { + libc::pthread_cond_destroy(self.cvar.get_mut()); + libc::pthread_mutex_destroy(self.lock.get_mut()); + } + } +} + +unsafe impl Sync for Parker {} +unsafe impl Send for Parker {} diff --git a/library/std/src/sys/unix/time.rs b/library/std/src/sys/unix/time.rs index d5abd9b58..2daad981b 100644 --- a/library/std/src/sys/unix/time.rs +++ b/library/std/src/sys/unix/time.rs @@ -5,6 +5,9 @@ pub use self::inner::Instant; const NSEC_PER_SEC: u64 = 1_000_000_000; pub const UNIX_EPOCH: SystemTime = SystemTime { t: Timespec::zero() }; +#[allow(dead_code)] // Used for pthread condvar timeouts +pub const TIMESPEC_MAX: libc::timespec = + libc::timespec { tv_sec: ::MAX, tv_nsec: 1_000_000_000 - 1 }; #[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)] #[repr(transparent)] diff --git a/library/std/src/sys/unix/weak.rs b/library/std/src/sys/unix/weak.rs index f5a4ce929..62ffee70b 100644 --- a/library/std/src/sys/unix/weak.rs +++ b/library/std/src/sys/unix/weak.rs @@ -1,9 +1,8 @@ //! Support for "weak linkage" to symbols on Unix //! -//! Some I/O operations we do in libstd require newer versions of OSes but we -//! need to maintain binary compatibility with older releases for now. In order -//! to use the new functionality when available we use this module for -//! detection. +//! Some I/O operations we do in std require newer versions of OSes but we need +//! to maintain binary compatibility with older releases for now. In order to +//! use the new functionality when available we use this module for detection. //! //! One option to use here is weak linkage, but that is unfortunately only //! really workable with ELF. Otherwise, use dlsym to get the symbol value at @@ -29,7 +28,7 @@ use crate::ptr; use crate::sync::atomic::{self, AtomicPtr, Ordering}; // We can use true weak linkage on ELF targets. -#[cfg(all(not(any(target_os = "macos", target_os = "ios")), not(bootstrap)))] +#[cfg(not(any(target_os = "macos", target_os = "ios")))] pub(crate) macro weak { (fn $name:ident($($t:ty),*) -> $ret:ty) => ( let ref $name: ExternWeak $ret> = { @@ -43,30 +42,14 @@ pub(crate) macro weak { ) } -#[cfg(all(not(any(target_os = "macos", target_os = "ios")), bootstrap))] -pub(crate) macro weak { - (fn $name:ident($($t:ty),*) -> $ret:ty) => ( - let ref $name: ExternWeak $ret> = { - extern "C" { - #[linkage = "extern_weak"] - static $name: *const libc::c_void; - } - #[allow(unused_unsafe)] - ExternWeak::new(unsafe { $name }) - }; - ) -} - // On non-ELF targets, use the dlsym approximation of weak linkage. #[cfg(any(target_os = "macos", target_os = "ios"))] pub(crate) use self::dlsym as weak; -#[cfg(not(bootstrap))] pub(crate) struct ExternWeak { weak_ptr: Option, } -#[cfg(not(bootstrap))] impl ExternWeak { #[inline] pub(crate) fn new(weak_ptr: Option) -> Self { @@ -79,34 +62,6 @@ impl ExternWeak { } } -#[cfg(bootstrap)] -pub(crate) struct ExternWeak { - weak_ptr: *const libc::c_void, - _marker: PhantomData, -} - -#[cfg(bootstrap)] -impl ExternWeak { - #[inline] - pub(crate) fn new(weak_ptr: *const libc::c_void) -> Self { - ExternWeak { weak_ptr, _marker: PhantomData } - } -} - -#[cfg(bootstrap)] -impl ExternWeak { - #[inline] - pub(crate) fn get(&self) -> Option { - unsafe { - if self.weak_ptr.is_null() { - None - } else { - Some(mem::transmute_copy::<*const libc::c_void, F>(&self.weak_ptr)) - } - } - } -} - pub(crate) macro dlsym { (fn $name:ident($($t:ty),*) -> $ret:ty) => ( dlsym!(fn $name($($t),*) -> $ret, stringify!($name)); diff --git a/library/std/src/sys/unsupported/mod.rs b/library/std/src/sys/unsupported/mod.rs index 7bf6d40b7..15b22c620 100644 --- a/library/std/src/sys/unsupported/mod.rs +++ b/library/std/src/sys/unsupported/mod.rs @@ -9,6 +9,7 @@ pub mod fs; pub mod io; pub mod locks; pub mod net; +pub mod once; pub mod os; #[path = "../unix/os_str.rs"] pub mod os_str; diff --git a/library/std/src/sys/unsupported/once.rs b/library/std/src/sys/unsupported/once.rs new file mode 100644 index 000000000..b4bb4975f --- /dev/null +++ b/library/std/src/sys/unsupported/once.rs @@ -0,0 +1,89 @@ +use crate::cell::Cell; +use crate::sync as public; + +pub struct Once { + state: Cell, +} + +pub struct OnceState { + poisoned: bool, + set_state_to: Cell, +} + +#[derive(Clone, Copy, PartialEq, Eq)] +enum State { + Incomplete, + Poisoned, + Running, + Complete, +} + +struct CompletionGuard<'a> { + state: &'a Cell, + set_state_on_drop_to: State, +} + +impl<'a> Drop for CompletionGuard<'a> { + fn drop(&mut self) { + self.state.set(self.set_state_on_drop_to); + } +} + +// Safety: threads are not supported on this platform. +unsafe impl Sync for Once {} + +impl Once { + #[inline] + #[rustc_const_stable(feature = "const_once_new", since = "1.32.0")] + pub const fn new() -> Once { + Once { state: Cell::new(State::Incomplete) } + } + + #[inline] + pub fn is_completed(&self) -> bool { + self.state.get() == State::Complete + } + + #[cold] + #[track_caller] + pub fn call(&self, ignore_poisoning: bool, f: &mut impl FnMut(&public::OnceState)) { + let state = self.state.get(); + match state { + State::Poisoned if !ignore_poisoning => { + // Panic to propagate the poison. + panic!("Once instance has previously been poisoned"); + } + State::Incomplete | State::Poisoned => { + self.state.set(State::Running); + // `guard` will set the new state on drop. + let mut guard = + CompletionGuard { state: &self.state, set_state_on_drop_to: State::Poisoned }; + // Run the function, letting it know if we're poisoned or not. + let f_state = public::OnceState { + inner: OnceState { + poisoned: state == State::Poisoned, + set_state_to: Cell::new(State::Complete), + }, + }; + f(&f_state); + guard.set_state_on_drop_to = f_state.inner.set_state_to.get(); + } + State::Running => { + panic!("one-time initialization may not be performed recursively"); + } + State::Complete => {} + } + } +} + +impl OnceState { + #[inline] + pub fn is_poisoned(&self) -> bool { + self.poisoned + } + + #[inline] + pub fn poison(&self) { + self.set_state_to.set(State::Poisoned) + } +} diff --git a/library/std/src/sys/unsupported/pipe.rs b/library/std/src/sys/unsupported/pipe.rs index 25514c232..0bba673b4 100644 --- a/library/std/src/sys/unsupported/pipe.rs +++ b/library/std/src/sys/unsupported/pipe.rs @@ -15,6 +15,10 @@ impl AnonPipe { self.0 } + pub fn read_to_end(&self, _buf: &mut Vec) -> io::Result { + self.0 + } + pub fn write(&self, _buf: &[u8]) -> io::Result { self.0 } diff --git a/library/std/src/sys/unsupported/process.rs b/library/std/src/sys/unsupported/process.rs index 633f17c05..a494f2d6b 100644 --- a/library/std/src/sys/unsupported/process.rs +++ b/library/std/src/sys/unsupported/process.rs @@ -75,6 +75,10 @@ impl Command { ) -> io::Result<(Process, StdioPipes)> { unsupported() } + + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + unsupported() + } } impl From for Stdio { diff --git a/library/std/src/sys/wasi/mod.rs b/library/std/src/sys/wasi/mod.rs index c8c47763a..1dc3f2b20 100644 --- a/library/std/src/sys/wasi/mod.rs +++ b/library/std/src/sys/wasi/mod.rs @@ -32,6 +32,8 @@ pub mod io; #[path = "../unsupported/locks/mod.rs"] pub mod locks; pub mod net; +#[path = "../unsupported/once.rs"] +pub mod once; pub mod os; #[path = "../unix/os_str.rs"] pub mod os_str; diff --git a/library/std/src/sys/wasm/mod.rs b/library/std/src/sys/wasm/mod.rs index d68c3e5f1..77ebe3c4a 100644 --- a/library/std/src/sys/wasm/mod.rs +++ b/library/std/src/sys/wasm/mod.rs @@ -66,6 +66,8 @@ cfg_if::cfg_if! { } else { #[path = "../unsupported/locks/mod.rs"] pub mod locks; + #[path = "../unsupported/once.rs"] + pub mod once; #[path = "../unsupported/thread.rs"] pub mod thread; } diff --git a/library/std/src/sys/windows/c.rs b/library/std/src/sys/windows/c.rs index 81461de4f..f58dcf128 100644 --- a/library/std/src/sys/windows/c.rs +++ b/library/std/src/sys/windows/c.rs @@ -295,8 +295,6 @@ pub fn nt_success(status: NTSTATUS) -> bool { status >= 0 } -// "RNG\0" -pub const BCRYPT_RNG_ALGORITHM: &[u16] = &[b'R' as u16, b'N' as u16, b'G' as u16, 0]; pub const BCRYPT_USE_SYSTEM_PREFERRED_RNG: DWORD = 0x00000002; #[repr(C)] @@ -834,6 +832,10 @@ if #[cfg(not(target_vendor = "uwp"))] { #[link(name = "advapi32")] extern "system" { + // Forbidden when targeting UWP + #[link_name = "SystemFunction036"] + pub fn RtlGenRandom(RandomBuffer: *mut u8, RandomBufferLength: ULONG) -> BOOLEAN; + // Allowed but unused by UWP pub fn OpenProcessToken( ProcessHandle: HANDLE, @@ -1258,13 +1260,6 @@ extern "system" { cbBuffer: ULONG, dwFlags: ULONG, ) -> NTSTATUS; - pub fn BCryptOpenAlgorithmProvider( - phalgorithm: *mut BCRYPT_ALG_HANDLE, - pszAlgId: LPCWSTR, - pszimplementation: LPCWSTR, - dwflags: ULONG, - ) -> NTSTATUS; - pub fn BCryptCloseAlgorithmProvider(hAlgorithm: BCRYPT_ALG_HANDLE, dwFlags: ULONG) -> NTSTATUS; } // Functions that aren't available on every version of Windows that we support, diff --git a/library/std/src/sys/windows/mod.rs b/library/std/src/sys/windows/mod.rs index e67411e16..77359abe4 100644 --- a/library/std/src/sys/windows/mod.rs +++ b/library/std/src/sys/windows/mod.rs @@ -33,7 +33,7 @@ pub mod stdio; pub mod thread; pub mod thread_local_dtor; pub mod thread_local_key; -pub mod thread_parker; +pub mod thread_parking; pub mod time; cfg_if::cfg_if! { if #[cfg(not(target_vendor = "uwp"))] { diff --git a/library/std/src/sys/windows/os.rs b/library/std/src/sys/windows/os.rs index 352337ba3..d7adeb266 100644 --- a/library/std/src/sys/windows/os.rs +++ b/library/std/src/sys/windows/os.rs @@ -157,7 +157,7 @@ impl<'a> Iterator for SplitPaths<'a> { // Double quotes are used as a way of introducing literal semicolons // (since c:\some;dir is a valid Windows path). Double quotes are not // themselves permitted in path names, so there is no way to escape a - // double quote. Quoted regions can appear in arbitrary locations, so + // double quote. Quoted regions can appear in arbitrary locations, so // // c:\foo;c:\som"e;di"r;c:\bar // diff --git a/library/std/src/sys/windows/pipe.rs b/library/std/src/sys/windows/pipe.rs index 9f26acc45..7b25edaa5 100644 --- a/library/std/src/sys/windows/pipe.rs +++ b/library/std/src/sys/windows/pipe.rs @@ -1,7 +1,7 @@ use crate::os::windows::prelude::*; use crate::ffi::OsStr; -use crate::io::{self, IoSlice, IoSliceMut}; +use crate::io::{self, IoSlice, IoSliceMut, Read}; use crate::mem; use crate::path::Path; use crate::ptr; @@ -261,6 +261,10 @@ impl AnonPipe { self.inner.is_read_vectored() } + pub fn read_to_end(&self, buf: &mut Vec) -> io::Result { + self.handle().read_to_end(buf) + } + pub fn write(&self, buf: &[u8]) -> io::Result { unsafe { let len = crate::cmp::min(buf.len(), c::DWORD::MAX as usize) as c::DWORD; diff --git a/library/std/src/sys/windows/process.rs b/library/std/src/sys/windows/process.rs index 31e9b34fb..10bc949e1 100644 --- a/library/std/src/sys/windows/process.rs +++ b/library/std/src/sys/windows/process.rs @@ -351,6 +351,11 @@ impl Command { )) } } + + pub fn output(&mut self) -> io::Result<(ExitStatus, Vec, Vec)> { + let (proc, pipes) = self.spawn(Stdio::MakePipe, false)?; + crate::sys_common::process::wait_with_output(proc, pipes) + } } impl fmt::Debug for Command { diff --git a/library/std/src/sys/windows/rand.rs b/library/std/src/sys/windows/rand.rs index b5a49489d..cdf37cfe9 100644 --- a/library/std/src/sys/windows/rand.rs +++ b/library/std/src/sys/windows/rand.rs @@ -1,106 +1,39 @@ -//! # Random key generation -//! -//! This module wraps the RNG provided by the OS. There are a few different -//! ways to interface with the OS RNG so it's worth exploring each of the options. -//! Note that at the time of writing these all go through the (undocumented) -//! `bcryptPrimitives.dll` but they use different route to get there. -//! -//! Originally we were using [`RtlGenRandom`], however that function is -//! deprecated and warns it "may be altered or unavailable in subsequent versions". -//! -//! So we switched to [`BCryptGenRandom`] with the `BCRYPT_USE_SYSTEM_PREFERRED_RNG` -//! flag to query and find the system configured RNG. However, this change caused a small -//! but significant number of users to experience panics caused by a failure of -//! this function. See [#94098]. -//! -//! The current version falls back to using `BCryptOpenAlgorithmProvider` if -//! `BCRYPT_USE_SYSTEM_PREFERRED_RNG` fails for any reason. -//! -//! [#94098]: https://github.com/rust-lang/rust/issues/94098 -//! [`RtlGenRandom`]: https://docs.microsoft.com/en-us/windows/win32/api/ntsecapi/nf-ntsecapi-rtlgenrandom -//! [`BCryptGenRandom`]: https://docs.microsoft.com/en-us/windows/win32/api/bcrypt/nf-bcrypt-bcryptgenrandom +use crate::io; use crate::mem; use crate::ptr; use crate::sys::c; -/// Generates high quality secure random keys for use by [`HashMap`]. -/// -/// This is used to seed the default [`RandomState`]. -/// -/// [`HashMap`]: crate::collections::HashMap -/// [`RandomState`]: crate::collections::hash_map::RandomState pub fn hashmap_random_keys() -> (u64, u64) { - Rng::SYSTEM.gen_random_keys().unwrap_or_else(fallback_rng) + let mut v = (0, 0); + let ret = unsafe { + c::BCryptGenRandom( + ptr::null_mut(), + &mut v as *mut _ as *mut u8, + mem::size_of_val(&v) as c::ULONG, + c::BCRYPT_USE_SYSTEM_PREFERRED_RNG, + ) + }; + if c::nt_success(ret) { v } else { fallback_rng() } } -struct Rng { - algorithm: c::BCRYPT_ALG_HANDLE, - flags: u32, -} -impl Rng { - const SYSTEM: Self = unsafe { Self::new(ptr::null_mut(), c::BCRYPT_USE_SYSTEM_PREFERRED_RNG) }; - - /// Create the RNG from an existing algorithm handle. - /// - /// # Safety - /// - /// The handle must either be null or a valid algorithm handle. - const unsafe fn new(algorithm: c::BCRYPT_ALG_HANDLE, flags: u32) -> Self { - Self { algorithm, flags } - } - - /// Open a handle to the RNG algorithm. - fn open() -> Result { - use crate::sync::atomic::AtomicPtr; - use crate::sync::atomic::Ordering::{Acquire, Release}; - - // An atomic is used so we don't need to reopen the handle every time. - static HANDLE: AtomicPtr = AtomicPtr::new(ptr::null_mut()); - - let mut handle = HANDLE.load(Acquire); - if handle.is_null() { - let status = unsafe { - c::BCryptOpenAlgorithmProvider( - &mut handle, - c::BCRYPT_RNG_ALGORITHM.as_ptr(), - ptr::null(), - 0, - ) - }; - if c::nt_success(status) { - // If another thread opens a handle first then use that handle instead. - let result = HANDLE.compare_exchange(ptr::null_mut(), handle, Release, Acquire); - if let Err(previous_handle) = result { - // Close our handle and return the previous one. - unsafe { c::BCryptCloseAlgorithmProvider(handle, 0) }; - handle = previous_handle; - } - Ok(unsafe { Self::new(handle, 0) }) - } else { - Err(status) - } - } else { - Ok(unsafe { Self::new(handle, 0) }) - } - } +/// Generate random numbers using the fallback RNG function (RtlGenRandom) +/// +/// This is necessary because of a failure to load the SysWOW64 variant of the +/// bcryptprimitives.dll library from code that lives in bcrypt.dll +/// See +#[cfg(not(target_vendor = "uwp"))] +#[inline(never)] +fn fallback_rng() -> (u64, u64) { + let mut v = (0, 0); + let ret = + unsafe { c::RtlGenRandom(&mut v as *mut _ as *mut u8, mem::size_of_val(&v) as c::ULONG) }; - fn gen_random_keys(self) -> Result<(u64, u64), c::NTSTATUS> { - let mut v = (0, 0); - let status = unsafe { - let size = mem::size_of_val(&v).try_into().unwrap(); - c::BCryptGenRandom(self.algorithm, ptr::addr_of_mut!(v).cast(), size, self.flags) - }; - if c::nt_success(status) { Ok(v) } else { Err(status) } - } + if ret != 0 { v } else { panic!("fallback RNG broken: {}", io::Error::last_os_error()) } } -/// Generate random numbers using the fallback RNG function +/// We can't use RtlGenRandom with UWP, so there is no fallback +#[cfg(target_vendor = "uwp")] #[inline(never)] -fn fallback_rng(rng_status: c::NTSTATUS) -> (u64, u64) { - match Rng::open().and_then(|rng| rng.gen_random_keys()) { - Ok(keys) => keys, - Err(status) => { - panic!("RNG broken: {rng_status:#x}, fallback RNG broken: {status:#x}") - } - } +fn fallback_rng() -> (u64, u64) { + panic!("fallback RNG broken: RtlGenRandom() not supported on UWP"); } diff --git a/library/std/src/sys/windows/thread.rs b/library/std/src/sys/windows/thread.rs index c5c9e97e6..1cb576c95 100644 --- a/library/std/src/sys/windows/thread.rs +++ b/library/std/src/sys/windows/thread.rs @@ -26,7 +26,7 @@ impl Thread { // FIXME On UNIX, we guard against stack sizes that are too small but // that's because pthreads enforces that stacks are at least - // PTHREAD_STACK_MIN bytes big. Windows has no such lower limit, it's + // PTHREAD_STACK_MIN bytes big. Windows has no such lower limit, it's // just that below a certain threshold you can't do anything useful. // That threshold is application and architecture-specific, however. let ret = c::CreateThread( diff --git a/library/std/src/sys/windows/thread_parker.rs b/library/std/src/sys/windows/thread_parker.rs deleted file mode 100644 index 2f7ae863b..000000000 --- a/library/std/src/sys/windows/thread_parker.rs +++ /dev/null @@ -1,253 +0,0 @@ -// Thread parker implementation for Windows. -// -// This uses WaitOnAddress and WakeByAddressSingle if available (Windows 8+). -// This modern API is exactly the same as the futex syscalls the Linux thread -// parker uses. When These APIs are available, the implementation of this -// thread parker matches the Linux thread parker exactly. -// -// However, when the modern API is not available, this implementation falls -// back to NT Keyed Events, which are similar, but have some important -// differences. These are available since Windows XP. -// -// WaitOnAddress first checks the state of the thread parker to make sure it no -// WakeByAddressSingle calls can be missed between updating the parker state -// and calling the function. -// -// NtWaitForKeyedEvent does not have this option, and unconditionally blocks -// without checking the parker state first. Instead, NtReleaseKeyedEvent -// (unlike WakeByAddressSingle) *blocks* until it woke up a thread waiting for -// it by NtWaitForKeyedEvent. This way, we can be sure no events are missed, -// but we need to be careful not to block unpark() if park_timeout() was woken -// up by a timeout instead of unpark(). -// -// Unlike WaitOnAddress, NtWaitForKeyedEvent/NtReleaseKeyedEvent operate on a -// HANDLE (created with NtCreateKeyedEvent). This means that we can be sure -// a successfully awoken park() was awoken by unpark() and not a -// NtReleaseKeyedEvent call from some other code, as these events are not only -// matched by the key (address of the parker (state)), but also by this HANDLE. -// We lazily allocate this handle the first time it is needed. -// -// The fast path (calling park() after unpark() was already called) and the -// possible states are the same for both implementations. This is used here to -// make sure the fast path does not even check which API to use, but can return -// right away, independent of the used API. Only the slow paths (which will -// actually block/wake a thread) check which API is available and have -// different implementations. -// -// Unfortunately, NT Keyed Events are an undocumented Windows API. However: -// - This API is relatively simple with obvious behaviour, and there are -// several (unofficial) articles documenting the details. [1] -// - `parking_lot` has been using this API for years (on Windows versions -// before Windows 8). [2] Many big projects extensively use parking_lot, -// such as servo and the Rust compiler itself. -// - It is the underlying API used by Windows SRW locks and Windows critical -// sections. [3] [4] -// - The source code of the implementations of Wine, ReactOs, and Windows XP -// are available and match the expected behaviour. -// - The main risk with an undocumented API is that it might change in the -// future. But since we only use it for older versions of Windows, that's not -// a problem. -// - Even if these functions do not block or wake as we expect (which is -// unlikely, see all previous points), this implementation would still be -// memory safe. The NT Keyed Events API is only used to sleep/block in the -// right place. -// -// [1]: http://www.locklessinc.com/articles/keyed_events/ -// [2]: https://github.com/Amanieu/parking_lot/commit/43abbc964e -// [3]: https://docs.microsoft.com/en-us/archive/msdn-magazine/2012/november/windows-with-c-the-evolution-of-synchronization-in-windows-and-c -// [4]: Windows Internals, Part 1, ISBN 9780735671300 - -use crate::pin::Pin; -use crate::ptr; -use crate::sync::atomic::{ - AtomicI8, AtomicPtr, - Ordering::{Acquire, Relaxed, Release}, -}; -use crate::sys::{c, dur2timeout}; -use crate::time::Duration; - -pub struct Parker { - state: AtomicI8, -} - -const PARKED: i8 = -1; -const EMPTY: i8 = 0; -const NOTIFIED: i8 = 1; - -// Notes about memory ordering: -// -// Memory ordering is only relevant for the relative ordering of operations -// between different variables. Even Ordering::Relaxed guarantees a -// monotonic/consistent order when looking at just a single atomic variable. -// -// So, since this parker is just a single atomic variable, we only need to look -// at the ordering guarantees we need to provide to the 'outside world'. -// -// The only memory ordering guarantee that parking and unparking provide, is -// that things which happened before unpark() are visible on the thread -// returning from park() afterwards. Otherwise, it was effectively unparked -// before unpark() was called while still consuming the 'token'. -// -// In other words, unpark() needs to synchronize with the part of park() that -// consumes the token and returns. -// -// This is done with a release-acquire synchronization, by using -// Ordering::Release when writing NOTIFIED (the 'token') in unpark(), and using -// Ordering::Acquire when reading this state in park() after waking up. -impl Parker { - /// Construct the Windows parker. The UNIX parker implementation - /// requires this to happen in-place. - pub unsafe fn new(parker: *mut Parker) { - parker.write(Self { state: AtomicI8::new(EMPTY) }); - } - - // Assumes this is only called by the thread that owns the Parker, - // which means that `self.state != PARKED`. This implementation doesn't require `Pin`, - // but other implementations do. - pub unsafe fn park(self: Pin<&Self>) { - // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the - // first case. - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - - if let Some(wait_on_address) = c::WaitOnAddress::option() { - loop { - // Wait for something to happen, assuming it's still set to PARKED. - wait_on_address(self.ptr(), &PARKED as *const _ as c::LPVOID, 1, c::INFINITE); - // Change NOTIFIED=>EMPTY but leave PARKED alone. - if self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Acquire).is_ok() { - // Actually woken up by unpark(). - return; - } else { - // Spurious wake up. We loop to try again. - } - } - } else { - // Wait for unpark() to produce this event. - c::NtWaitForKeyedEvent(keyed_event_handle(), self.ptr(), 0, ptr::null_mut()); - // Set the state back to EMPTY (from either PARKED or NOTIFIED). - // Note that we don't just write EMPTY, but use swap() to also - // include an acquire-ordered read to synchronize with unpark()'s - // release-ordered write. - self.state.swap(EMPTY, Acquire); - } - } - - // Assumes this is only called by the thread that owns the Parker, - // which means that `self.state != PARKED`. This implementation doesn't require `Pin`, - // but other implementations do. - pub unsafe fn park_timeout(self: Pin<&Self>, timeout: Duration) { - // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the - // first case. - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - - if let Some(wait_on_address) = c::WaitOnAddress::option() { - // Wait for something to happen, assuming it's still set to PARKED. - wait_on_address(self.ptr(), &PARKED as *const _ as c::LPVOID, 1, dur2timeout(timeout)); - // Set the state back to EMPTY (from either PARKED or NOTIFIED). - // Note that we don't just write EMPTY, but use swap() to also - // include an acquire-ordered read to synchronize with unpark()'s - // release-ordered write. - if self.state.swap(EMPTY, Acquire) == NOTIFIED { - // Actually woken up by unpark(). - } else { - // Timeout or spurious wake up. - // We return either way, because we can't easily tell if it was the - // timeout or not. - } - } else { - // Need to wait for unpark() using NtWaitForKeyedEvent. - let handle = keyed_event_handle(); - - // NtWaitForKeyedEvent uses a unit of 100ns, and uses negative - // values to indicate a relative time on the monotonic clock. - // This is documented here for the underlying KeWaitForSingleObject function: - // https://docs.microsoft.com/en-us/windows-hardware/drivers/ddi/wdm/nf-wdm-kewaitforsingleobject - let mut timeout = match i64::try_from((timeout.as_nanos() + 99) / 100) { - Ok(t) => -t, - Err(_) => i64::MIN, - }; - - // Wait for unpark() to produce this event. - let unparked = - c::NtWaitForKeyedEvent(handle, self.ptr(), 0, &mut timeout) == c::STATUS_SUCCESS; - - // Set the state back to EMPTY (from either PARKED or NOTIFIED). - let prev_state = self.state.swap(EMPTY, Acquire); - - if !unparked && prev_state == NOTIFIED { - // We were awoken by a timeout, not by unpark(), but the state - // was set to NOTIFIED, which means we *just* missed an - // unpark(), which is now blocked on us to wait for it. - // Wait for it to consume the event and unblock that thread. - c::NtWaitForKeyedEvent(handle, self.ptr(), 0, ptr::null_mut()); - } - } - } - - // This implementation doesn't require `Pin`, but other implementations do. - pub fn unpark(self: Pin<&Self>) { - // Change PARKED=>NOTIFIED, EMPTY=>NOTIFIED, or NOTIFIED=>NOTIFIED, and - // wake the thread in the first case. - // - // Note that even NOTIFIED=>NOTIFIED results in a write. This is on - // purpose, to make sure every unpark() has a release-acquire ordering - // with park(). - if self.state.swap(NOTIFIED, Release) == PARKED { - unsafe { - if let Some(wake_by_address_single) = c::WakeByAddressSingle::option() { - wake_by_address_single(self.ptr()); - } else { - // If we run NtReleaseKeyedEvent before the waiting thread runs - // NtWaitForKeyedEvent, this (shortly) blocks until we can wake it up. - // If the waiting thread wakes up before we run NtReleaseKeyedEvent - // (e.g. due to a timeout), this blocks until we do wake up a thread. - // To prevent this thread from blocking indefinitely in that case, - // park_impl() will, after seeing the state set to NOTIFIED after - // waking up, call NtWaitForKeyedEvent again to unblock us. - c::NtReleaseKeyedEvent(keyed_event_handle(), self.ptr(), 0, ptr::null_mut()); - } - } - } - } - - fn ptr(&self) -> c::LPVOID { - &self.state as *const _ as c::LPVOID - } -} - -fn keyed_event_handle() -> c::HANDLE { - const INVALID: c::HANDLE = ptr::invalid_mut(!0); - static HANDLE: AtomicPtr = AtomicPtr::new(INVALID); - match HANDLE.load(Relaxed) { - INVALID => { - let mut handle = c::INVALID_HANDLE_VALUE; - unsafe { - match c::NtCreateKeyedEvent( - &mut handle, - c::GENERIC_READ | c::GENERIC_WRITE, - ptr::null_mut(), - 0, - ) { - c::STATUS_SUCCESS => {} - r => panic!("Unable to create keyed event handle: error {r}"), - } - } - match HANDLE.compare_exchange(INVALID, handle, Relaxed, Relaxed) { - Ok(_) => handle, - Err(h) => { - // Lost the race to another thread initializing HANDLE before we did. - // Closing our handle and using theirs instead. - unsafe { - c::CloseHandle(handle); - } - h - } - } - } - handle => handle, - } -} diff --git a/library/std/src/sys/windows/thread_parking.rs b/library/std/src/sys/windows/thread_parking.rs new file mode 100644 index 000000000..5d43676ad --- /dev/null +++ b/library/std/src/sys/windows/thread_parking.rs @@ -0,0 +1,253 @@ +// Thread parker implementation for Windows. +// +// This uses WaitOnAddress and WakeByAddressSingle if available (Windows 8+). +// This modern API is exactly the same as the futex syscalls the Linux thread +// parker uses. When These APIs are available, the implementation of this +// thread parker matches the Linux thread parker exactly. +// +// However, when the modern API is not available, this implementation falls +// back to NT Keyed Events, which are similar, but have some important +// differences. These are available since Windows XP. +// +// WaitOnAddress first checks the state of the thread parker to make sure it no +// WakeByAddressSingle calls can be missed between updating the parker state +// and calling the function. +// +// NtWaitForKeyedEvent does not have this option, and unconditionally blocks +// without checking the parker state first. Instead, NtReleaseKeyedEvent +// (unlike WakeByAddressSingle) *blocks* until it woke up a thread waiting for +// it by NtWaitForKeyedEvent. This way, we can be sure no events are missed, +// but we need to be careful not to block unpark() if park_timeout() was woken +// up by a timeout instead of unpark(). +// +// Unlike WaitOnAddress, NtWaitForKeyedEvent/NtReleaseKeyedEvent operate on a +// HANDLE (created with NtCreateKeyedEvent). This means that we can be sure +// a successfully awoken park() was awoken by unpark() and not a +// NtReleaseKeyedEvent call from some other code, as these events are not only +// matched by the key (address of the parker (state)), but also by this HANDLE. +// We lazily allocate this handle the first time it is needed. +// +// The fast path (calling park() after unpark() was already called) and the +// possible states are the same for both implementations. This is used here to +// make sure the fast path does not even check which API to use, but can return +// right away, independent of the used API. Only the slow paths (which will +// actually block/wake a thread) check which API is available and have +// different implementations. +// +// Unfortunately, NT Keyed Events are an undocumented Windows API. However: +// - This API is relatively simple with obvious behaviour, and there are +// several (unofficial) articles documenting the details. [1] +// - `parking_lot` has been using this API for years (on Windows versions +// before Windows 8). [2] Many big projects extensively use parking_lot, +// such as servo and the Rust compiler itself. +// - It is the underlying API used by Windows SRW locks and Windows critical +// sections. [3] [4] +// - The source code of the implementations of Wine, ReactOs, and Windows XP +// are available and match the expected behaviour. +// - The main risk with an undocumented API is that it might change in the +// future. But since we only use it for older versions of Windows, that's not +// a problem. +// - Even if these functions do not block or wake as we expect (which is +// unlikely, see all previous points), this implementation would still be +// memory safe. The NT Keyed Events API is only used to sleep/block in the +// right place. +// +// [1]: http://www.locklessinc.com/articles/keyed_events/ +// [2]: https://github.com/Amanieu/parking_lot/commit/43abbc964e +// [3]: https://docs.microsoft.com/en-us/archive/msdn-magazine/2012/november/windows-with-c-the-evolution-of-synchronization-in-windows-and-c +// [4]: Windows Internals, Part 1, ISBN 9780735671300 + +use crate::pin::Pin; +use crate::ptr; +use crate::sync::atomic::{ + AtomicI8, AtomicPtr, + Ordering::{Acquire, Relaxed, Release}, +}; +use crate::sys::{c, dur2timeout}; +use crate::time::Duration; + +pub struct Parker { + state: AtomicI8, +} + +const PARKED: i8 = -1; +const EMPTY: i8 = 0; +const NOTIFIED: i8 = 1; + +// Notes about memory ordering: +// +// Memory ordering is only relevant for the relative ordering of operations +// between different variables. Even Ordering::Relaxed guarantees a +// monotonic/consistent order when looking at just a single atomic variable. +// +// So, since this parker is just a single atomic variable, we only need to look +// at the ordering guarantees we need to provide to the 'outside world'. +// +// The only memory ordering guarantee that parking and unparking provide, is +// that things which happened before unpark() are visible on the thread +// returning from park() afterwards. Otherwise, it was effectively unparked +// before unpark() was called while still consuming the 'token'. +// +// In other words, unpark() needs to synchronize with the part of park() that +// consumes the token and returns. +// +// This is done with a release-acquire synchronization, by using +// Ordering::Release when writing NOTIFIED (the 'token') in unpark(), and using +// Ordering::Acquire when reading this state in park() after waking up. +impl Parker { + /// Construct the Windows parker. The UNIX parker implementation + /// requires this to happen in-place. + pub unsafe fn new_in_place(parker: *mut Parker) { + parker.write(Self { state: AtomicI8::new(EMPTY) }); + } + + // Assumes this is only called by the thread that owns the Parker, + // which means that `self.state != PARKED`. This implementation doesn't require `Pin`, + // but other implementations do. + pub unsafe fn park(self: Pin<&Self>) { + // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the + // first case. + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + + if let Some(wait_on_address) = c::WaitOnAddress::option() { + loop { + // Wait for something to happen, assuming it's still set to PARKED. + wait_on_address(self.ptr(), &PARKED as *const _ as c::LPVOID, 1, c::INFINITE); + // Change NOTIFIED=>EMPTY but leave PARKED alone. + if self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Acquire).is_ok() { + // Actually woken up by unpark(). + return; + } else { + // Spurious wake up. We loop to try again. + } + } + } else { + // Wait for unpark() to produce this event. + c::NtWaitForKeyedEvent(keyed_event_handle(), self.ptr(), 0, ptr::null_mut()); + // Set the state back to EMPTY (from either PARKED or NOTIFIED). + // Note that we don't just write EMPTY, but use swap() to also + // include an acquire-ordered read to synchronize with unpark()'s + // release-ordered write. + self.state.swap(EMPTY, Acquire); + } + } + + // Assumes this is only called by the thread that owns the Parker, + // which means that `self.state != PARKED`. This implementation doesn't require `Pin`, + // but other implementations do. + pub unsafe fn park_timeout(self: Pin<&Self>, timeout: Duration) { + // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the + // first case. + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + + if let Some(wait_on_address) = c::WaitOnAddress::option() { + // Wait for something to happen, assuming it's still set to PARKED. + wait_on_address(self.ptr(), &PARKED as *const _ as c::LPVOID, 1, dur2timeout(timeout)); + // Set the state back to EMPTY (from either PARKED or NOTIFIED). + // Note that we don't just write EMPTY, but use swap() to also + // include an acquire-ordered read to synchronize with unpark()'s + // release-ordered write. + if self.state.swap(EMPTY, Acquire) == NOTIFIED { + // Actually woken up by unpark(). + } else { + // Timeout or spurious wake up. + // We return either way, because we can't easily tell if it was the + // timeout or not. + } + } else { + // Need to wait for unpark() using NtWaitForKeyedEvent. + let handle = keyed_event_handle(); + + // NtWaitForKeyedEvent uses a unit of 100ns, and uses negative + // values to indicate a relative time on the monotonic clock. + // This is documented here for the underlying KeWaitForSingleObject function: + // https://docs.microsoft.com/en-us/windows-hardware/drivers/ddi/wdm/nf-wdm-kewaitforsingleobject + let mut timeout = match i64::try_from((timeout.as_nanos() + 99) / 100) { + Ok(t) => -t, + Err(_) => i64::MIN, + }; + + // Wait for unpark() to produce this event. + let unparked = + c::NtWaitForKeyedEvent(handle, self.ptr(), 0, &mut timeout) == c::STATUS_SUCCESS; + + // Set the state back to EMPTY (from either PARKED or NOTIFIED). + let prev_state = self.state.swap(EMPTY, Acquire); + + if !unparked && prev_state == NOTIFIED { + // We were awoken by a timeout, not by unpark(), but the state + // was set to NOTIFIED, which means we *just* missed an + // unpark(), which is now blocked on us to wait for it. + // Wait for it to consume the event and unblock that thread. + c::NtWaitForKeyedEvent(handle, self.ptr(), 0, ptr::null_mut()); + } + } + } + + // This implementation doesn't require `Pin`, but other implementations do. + pub fn unpark(self: Pin<&Self>) { + // Change PARKED=>NOTIFIED, EMPTY=>NOTIFIED, or NOTIFIED=>NOTIFIED, and + // wake the thread in the first case. + // + // Note that even NOTIFIED=>NOTIFIED results in a write. This is on + // purpose, to make sure every unpark() has a release-acquire ordering + // with park(). + if self.state.swap(NOTIFIED, Release) == PARKED { + unsafe { + if let Some(wake_by_address_single) = c::WakeByAddressSingle::option() { + wake_by_address_single(self.ptr()); + } else { + // If we run NtReleaseKeyedEvent before the waiting thread runs + // NtWaitForKeyedEvent, this (shortly) blocks until we can wake it up. + // If the waiting thread wakes up before we run NtReleaseKeyedEvent + // (e.g. due to a timeout), this blocks until we do wake up a thread. + // To prevent this thread from blocking indefinitely in that case, + // park_impl() will, after seeing the state set to NOTIFIED after + // waking up, call NtWaitForKeyedEvent again to unblock us. + c::NtReleaseKeyedEvent(keyed_event_handle(), self.ptr(), 0, ptr::null_mut()); + } + } + } + } + + fn ptr(&self) -> c::LPVOID { + &self.state as *const _ as c::LPVOID + } +} + +fn keyed_event_handle() -> c::HANDLE { + const INVALID: c::HANDLE = ptr::invalid_mut(!0); + static HANDLE: AtomicPtr = AtomicPtr::new(INVALID); + match HANDLE.load(Relaxed) { + INVALID => { + let mut handle = c::INVALID_HANDLE_VALUE; + unsafe { + match c::NtCreateKeyedEvent( + &mut handle, + c::GENERIC_READ | c::GENERIC_WRITE, + ptr::null_mut(), + 0, + ) { + c::STATUS_SUCCESS => {} + r => panic!("Unable to create keyed event handle: error {r}"), + } + } + match HANDLE.compare_exchange(INVALID, handle, Relaxed, Relaxed) { + Ok(_) => handle, + Err(h) => { + // Lost the race to another thread initializing HANDLE before we did. + // Closing our handle and using theirs instead. + unsafe { + c::CloseHandle(handle); + } + h + } + } + } + handle => handle, + } +} diff --git a/library/std/src/sys_common/backtrace.rs b/library/std/src/sys_common/backtrace.rs index 8807077cb..f1d804ef4 100644 --- a/library/std/src/sys_common/backtrace.rs +++ b/library/std/src/sys_common/backtrace.rs @@ -20,7 +20,7 @@ pub fn lock() -> impl Drop { /// Prints the current backtrace. pub fn print(w: &mut dyn Write, format: PrintFmt) -> io::Result<()> { // There are issues currently linking libbacktrace into tests, and in - // general during libstd's own unit tests we're not testing this path. In + // general during std's own unit tests we're not testing this path. In // test mode immediately return here to optimize away any references to the // libbacktrace symbols if cfg!(test) { @@ -111,7 +111,7 @@ unsafe fn _print_fmt(fmt: &mut fmt::Formatter<'_>, print_fmt: PrintFmt) -> fmt:: } /// Fixed frame used to clean the backtrace with `RUST_BACKTRACE=1`. Note that -/// this is only inline(never) when backtraces in libstd are enabled, otherwise +/// this is only inline(never) when backtraces in std are enabled, otherwise /// it's fine to optimize away. #[cfg_attr(feature = "backtrace", inline(never))] pub fn __rust_begin_short_backtrace(f: F) -> T @@ -127,7 +127,7 @@ where } /// Fixed frame used to clean the backtrace with `RUST_BACKTRACE=1`. Note that -/// this is only inline(never) when backtraces in libstd are enabled, otherwise +/// this is only inline(never) when backtraces in std are enabled, otherwise /// it's fine to optimize away. #[cfg_attr(feature = "backtrace", inline(never))] pub fn __rust_end_short_backtrace(f: F) -> T diff --git a/library/std/src/sys_common/io.rs b/library/std/src/sys_common/io.rs index d1e9fed41..4a42ff3c6 100644 --- a/library/std/src/sys_common/io.rs +++ b/library/std/src/sys_common/io.rs @@ -39,9 +39,10 @@ pub mod test { } } + #[track_caller] // for `test_rng` pub fn tmpdir() -> TempDir { let p = env::temp_dir(); - let mut r = rand::thread_rng(); + let mut r = crate::test_helpers::test_rng(); let ret = p.join(&format!("rust-{}", r.next_u32())); fs::create_dir(&ret).unwrap(); TempDir(ret) diff --git a/library/std/src/sys_common/mod.rs b/library/std/src/sys_common/mod.rs index b1987aa0f..6b24b0e9a 100644 --- a/library/std/src/sys_common/mod.rs +++ b/library/std/src/sys_common/mod.rs @@ -27,11 +27,10 @@ pub mod lazy_box; pub mod memchr; pub mod once; pub mod process; -pub mod remutex; pub mod thread; pub mod thread_info; pub mod thread_local_dtor; -pub mod thread_parker; +pub mod thread_parking; pub mod wstr; pub mod wtf8; diff --git a/library/std/src/sys_common/once/generic.rs b/library/std/src/sys_common/once/generic.rs deleted file mode 100644 index d953a6745..000000000 --- a/library/std/src/sys_common/once/generic.rs +++ /dev/null @@ -1,283 +0,0 @@ -// Each `Once` has one word of atomic state, and this state is CAS'd on to -// determine what to do. There are four possible state of a `Once`: -// -// * Incomplete - no initialization has run yet, and no thread is currently -// using the Once. -// * Poisoned - some thread has previously attempted to initialize the Once, but -// it panicked, so the Once is now poisoned. There are no other -// threads currently accessing this Once. -// * Running - some thread is currently attempting to run initialization. It may -// succeed, so all future threads need to wait for it to finish. -// Note that this state is accompanied with a payload, described -// below. -// * Complete - initialization has completed and all future calls should finish -// immediately. -// -// With 4 states we need 2 bits to encode this, and we use the remaining bits -// in the word we have allocated as a queue of threads waiting for the thread -// responsible for entering the RUNNING state. This queue is just a linked list -// of Waiter nodes which is monotonically increasing in size. Each node is -// allocated on the stack, and whenever the running closure finishes it will -// consume the entire queue and notify all waiters they should try again. -// -// You'll find a few more details in the implementation, but that's the gist of -// it! -// -// Atomic orderings: -// When running `Once` we deal with multiple atomics: -// `Once.state_and_queue` and an unknown number of `Waiter.signaled`. -// * `state_and_queue` is used (1) as a state flag, (2) for synchronizing the -// result of the `Once`, and (3) for synchronizing `Waiter` nodes. -// - At the end of the `call` function we have to make sure the result -// of the `Once` is acquired. So every load which can be the only one to -// load COMPLETED must have at least acquire ordering, which means all -// three of them. -// - `WaiterQueue::drop` is the only place that may store COMPLETED, and -// must do so with release ordering to make the result available. -// - `wait` inserts `Waiter` nodes as a pointer in `state_and_queue`, and -// needs to make the nodes available with release ordering. The load in -// its `compare_exchange` can be relaxed because it only has to compare -// the atomic, not to read other data. -// - `WaiterQueue::drop` must see the `Waiter` nodes, so it must load -// `state_and_queue` with acquire ordering. -// - There is just one store where `state_and_queue` is used only as a -// state flag, without having to synchronize data: switching the state -// from INCOMPLETE to RUNNING in `call`. This store can be Relaxed, -// but the read has to be Acquire because of the requirements mentioned -// above. -// * `Waiter.signaled` is both used as a flag, and to protect a field with -// interior mutability in `Waiter`. `Waiter.thread` is changed in -// `WaiterQueue::drop` which then sets `signaled` with release ordering. -// After `wait` loads `signaled` with acquire ordering and sees it is true, -// it needs to see the changes to drop the `Waiter` struct correctly. -// * There is one place where the two atomics `Once.state_and_queue` and -// `Waiter.signaled` come together, and might be reordered by the compiler or -// processor. Because both use acquire ordering such a reordering is not -// allowed, so no need for `SeqCst`. - -use crate::cell::Cell; -use crate::fmt; -use crate::ptr; -use crate::sync as public; -use crate::sync::atomic::{AtomicBool, AtomicPtr, Ordering}; -use crate::thread::{self, Thread}; - -type Masked = (); - -pub struct Once { - state_and_queue: AtomicPtr, -} - -pub struct OnceState { - poisoned: bool, - set_state_on_drop_to: Cell<*mut Masked>, -} - -// Four states that a Once can be in, encoded into the lower bits of -// `state_and_queue` in the Once structure. -const INCOMPLETE: usize = 0x0; -const POISONED: usize = 0x1; -const RUNNING: usize = 0x2; -const COMPLETE: usize = 0x3; - -// Mask to learn about the state. All other bits are the queue of waiters if -// this is in the RUNNING state. -const STATE_MASK: usize = 0x3; - -// Representation of a node in the linked list of waiters, used while in the -// RUNNING state. -// Note: `Waiter` can't hold a mutable pointer to the next thread, because then -// `wait` would both hand out a mutable reference to its `Waiter` node, and keep -// a shared reference to check `signaled`. Instead we hold shared references and -// use interior mutability. -#[repr(align(4))] // Ensure the two lower bits are free to use as state bits. -struct Waiter { - thread: Cell>, - signaled: AtomicBool, - next: *const Waiter, -} - -// Head of a linked list of waiters. -// Every node is a struct on the stack of a waiting thread. -// Will wake up the waiters when it gets dropped, i.e. also on panic. -struct WaiterQueue<'a> { - state_and_queue: &'a AtomicPtr, - set_state_on_drop_to: *mut Masked, -} - -impl Once { - #[inline] - #[rustc_const_stable(feature = "const_once_new", since = "1.32.0")] - pub const fn new() -> Once { - Once { state_and_queue: AtomicPtr::new(ptr::invalid_mut(INCOMPLETE)) } - } - - #[inline] - pub fn is_completed(&self) -> bool { - // An `Acquire` load is enough because that makes all the initialization - // operations visible to us, and, this being a fast path, weaker - // ordering helps with performance. This `Acquire` synchronizes with - // `Release` operations on the slow path. - self.state_and_queue.load(Ordering::Acquire).addr() == COMPLETE - } - - // This is a non-generic function to reduce the monomorphization cost of - // using `call_once` (this isn't exactly a trivial or small implementation). - // - // Additionally, this is tagged with `#[cold]` as it should indeed be cold - // and it helps let LLVM know that calls to this function should be off the - // fast path. Essentially, this should help generate more straight line code - // in LLVM. - // - // Finally, this takes an `FnMut` instead of a `FnOnce` because there's - // currently no way to take an `FnOnce` and call it via virtual dispatch - // without some allocation overhead. - #[cold] - #[track_caller] - pub fn call(&self, ignore_poisoning: bool, init: &mut dyn FnMut(&public::OnceState)) { - let mut state_and_queue = self.state_and_queue.load(Ordering::Acquire); - loop { - match state_and_queue.addr() { - COMPLETE => break, - POISONED if !ignore_poisoning => { - // Panic to propagate the poison. - panic!("Once instance has previously been poisoned"); - } - POISONED | INCOMPLETE => { - // Try to register this thread as the one RUNNING. - let exchange_result = self.state_and_queue.compare_exchange( - state_and_queue, - ptr::invalid_mut(RUNNING), - Ordering::Acquire, - Ordering::Acquire, - ); - if let Err(old) = exchange_result { - state_and_queue = old; - continue; - } - // `waiter_queue` will manage other waiting threads, and - // wake them up on drop. - let mut waiter_queue = WaiterQueue { - state_and_queue: &self.state_and_queue, - set_state_on_drop_to: ptr::invalid_mut(POISONED), - }; - // Run the initialization function, letting it know if we're - // poisoned or not. - let init_state = public::OnceState { - inner: OnceState { - poisoned: state_and_queue.addr() == POISONED, - set_state_on_drop_to: Cell::new(ptr::invalid_mut(COMPLETE)), - }, - }; - init(&init_state); - waiter_queue.set_state_on_drop_to = init_state.inner.set_state_on_drop_to.get(); - break; - } - _ => { - // All other values must be RUNNING with possibly a - // pointer to the waiter queue in the more significant bits. - assert!(state_and_queue.addr() & STATE_MASK == RUNNING); - wait(&self.state_and_queue, state_and_queue); - state_and_queue = self.state_and_queue.load(Ordering::Acquire); - } - } - } - } -} - -fn wait(state_and_queue: &AtomicPtr, mut current_state: *mut Masked) { - // Note: the following code was carefully written to avoid creating a - // mutable reference to `node` that gets aliased. - loop { - // Don't queue this thread if the status is no longer running, - // otherwise we will not be woken up. - if current_state.addr() & STATE_MASK != RUNNING { - return; - } - - // Create the node for our current thread. - let node = Waiter { - thread: Cell::new(Some(thread::current())), - signaled: AtomicBool::new(false), - next: current_state.with_addr(current_state.addr() & !STATE_MASK) as *const Waiter, - }; - let me = &node as *const Waiter as *const Masked as *mut Masked; - - // Try to slide in the node at the head of the linked list, making sure - // that another thread didn't just replace the head of the linked list. - let exchange_result = state_and_queue.compare_exchange( - current_state, - me.with_addr(me.addr() | RUNNING), - Ordering::Release, - Ordering::Relaxed, - ); - if let Err(old) = exchange_result { - current_state = old; - continue; - } - - // We have enqueued ourselves, now lets wait. - // It is important not to return before being signaled, otherwise we - // would drop our `Waiter` node and leave a hole in the linked list - // (and a dangling reference). Guard against spurious wakeups by - // reparking ourselves until we are signaled. - while !node.signaled.load(Ordering::Acquire) { - // If the managing thread happens to signal and unpark us before we - // can park ourselves, the result could be this thread never gets - // unparked. Luckily `park` comes with the guarantee that if it got - // an `unpark` just before on an unparked thread it does not park. - thread::park(); - } - break; - } -} - -#[stable(feature = "std_debug", since = "1.16.0")] -impl fmt::Debug for Once { - fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { - f.debug_struct("Once").finish_non_exhaustive() - } -} - -impl Drop for WaiterQueue<'_> { - fn drop(&mut self) { - // Swap out our state with however we finished. - let state_and_queue = - self.state_and_queue.swap(self.set_state_on_drop_to, Ordering::AcqRel); - - // We should only ever see an old state which was RUNNING. - assert_eq!(state_and_queue.addr() & STATE_MASK, RUNNING); - - // Walk the entire linked list of waiters and wake them up (in lifo - // order, last to register is first to wake up). - unsafe { - // Right after setting `node.signaled = true` the other thread may - // free `node` if there happens to be has a spurious wakeup. - // So we have to take out the `thread` field and copy the pointer to - // `next` first. - let mut queue = - state_and_queue.with_addr(state_and_queue.addr() & !STATE_MASK) as *const Waiter; - while !queue.is_null() { - let next = (*queue).next; - let thread = (*queue).thread.take().unwrap(); - (*queue).signaled.store(true, Ordering::Release); - // ^- FIXME (maybe): This is another case of issue #55005 - // `store()` has a potentially dangling ref to `signaled`. - queue = next; - thread.unpark(); - } - } - } -} - -impl OnceState { - #[inline] - pub fn is_poisoned(&self) -> bool { - self.poisoned - } - - #[inline] - pub fn poison(&self) { - self.set_state_on_drop_to.set(ptr::invalid_mut(POISONED)); - } -} diff --git a/library/std/src/sys_common/once/mod.rs b/library/std/src/sys_common/once/mod.rs index 8742e68cc..359697d83 100644 --- a/library/std/src/sys_common/once/mod.rs +++ b/library/std/src/sys_common/once/mod.rs @@ -6,22 +6,6 @@ // As a result, we end up implementing it ourselves in the standard library. // This also gives us the opportunity to optimize the implementation a bit which // should help the fast path on call sites. -// -// So to recap, the guarantees of a Once are that it will call the -// initialization closure at most once, and it will never return until the one -// that's running has finished running. This means that we need some form of -// blocking here while the custom callback is running at the very least. -// Additionally, we add on the restriction of **poisoning**. Whenever an -// initialization closure panics, the Once enters a "poisoned" state which means -// that all future calls will immediately panic as well. -// -// So to implement this, one might first reach for a `Mutex`, but those cannot -// be put into a `static`. It also gets a lot harder with poisoning to figure -// out when the mutex needs to be deallocated because it's not after the closure -// finishes, but after the first successful closure finishes. -// -// All in all, this is instead implemented with atomics and lock-free -// operations! Whee! cfg_if::cfg_if! { if #[cfg(any( @@ -36,8 +20,15 @@ cfg_if::cfg_if! { ))] { mod futex; pub use futex::{Once, OnceState}; + } else if #[cfg(any( + windows, + target_family = "unix", + all(target_vendor = "fortanix", target_env = "sgx"), + target_os = "solid_asp3", + ))] { + mod queue; + pub use queue::{Once, OnceState}; } else { - mod generic; - pub use generic::{Once, OnceState}; + pub use crate::sys::once::{Once, OnceState}; } } diff --git a/library/std/src/sys_common/once/queue.rs b/library/std/src/sys_common/once/queue.rs new file mode 100644 index 000000000..d953a6745 --- /dev/null +++ b/library/std/src/sys_common/once/queue.rs @@ -0,0 +1,283 @@ +// Each `Once` has one word of atomic state, and this state is CAS'd on to +// determine what to do. There are four possible state of a `Once`: +// +// * Incomplete - no initialization has run yet, and no thread is currently +// using the Once. +// * Poisoned - some thread has previously attempted to initialize the Once, but +// it panicked, so the Once is now poisoned. There are no other +// threads currently accessing this Once. +// * Running - some thread is currently attempting to run initialization. It may +// succeed, so all future threads need to wait for it to finish. +// Note that this state is accompanied with a payload, described +// below. +// * Complete - initialization has completed and all future calls should finish +// immediately. +// +// With 4 states we need 2 bits to encode this, and we use the remaining bits +// in the word we have allocated as a queue of threads waiting for the thread +// responsible for entering the RUNNING state. This queue is just a linked list +// of Waiter nodes which is monotonically increasing in size. Each node is +// allocated on the stack, and whenever the running closure finishes it will +// consume the entire queue and notify all waiters they should try again. +// +// You'll find a few more details in the implementation, but that's the gist of +// it! +// +// Atomic orderings: +// When running `Once` we deal with multiple atomics: +// `Once.state_and_queue` and an unknown number of `Waiter.signaled`. +// * `state_and_queue` is used (1) as a state flag, (2) for synchronizing the +// result of the `Once`, and (3) for synchronizing `Waiter` nodes. +// - At the end of the `call` function we have to make sure the result +// of the `Once` is acquired. So every load which can be the only one to +// load COMPLETED must have at least acquire ordering, which means all +// three of them. +// - `WaiterQueue::drop` is the only place that may store COMPLETED, and +// must do so with release ordering to make the result available. +// - `wait` inserts `Waiter` nodes as a pointer in `state_and_queue`, and +// needs to make the nodes available with release ordering. The load in +// its `compare_exchange` can be relaxed because it only has to compare +// the atomic, not to read other data. +// - `WaiterQueue::drop` must see the `Waiter` nodes, so it must load +// `state_and_queue` with acquire ordering. +// - There is just one store where `state_and_queue` is used only as a +// state flag, without having to synchronize data: switching the state +// from INCOMPLETE to RUNNING in `call`. This store can be Relaxed, +// but the read has to be Acquire because of the requirements mentioned +// above. +// * `Waiter.signaled` is both used as a flag, and to protect a field with +// interior mutability in `Waiter`. `Waiter.thread` is changed in +// `WaiterQueue::drop` which then sets `signaled` with release ordering. +// After `wait` loads `signaled` with acquire ordering and sees it is true, +// it needs to see the changes to drop the `Waiter` struct correctly. +// * There is one place where the two atomics `Once.state_and_queue` and +// `Waiter.signaled` come together, and might be reordered by the compiler or +// processor. Because both use acquire ordering such a reordering is not +// allowed, so no need for `SeqCst`. + +use crate::cell::Cell; +use crate::fmt; +use crate::ptr; +use crate::sync as public; +use crate::sync::atomic::{AtomicBool, AtomicPtr, Ordering}; +use crate::thread::{self, Thread}; + +type Masked = (); + +pub struct Once { + state_and_queue: AtomicPtr, +} + +pub struct OnceState { + poisoned: bool, + set_state_on_drop_to: Cell<*mut Masked>, +} + +// Four states that a Once can be in, encoded into the lower bits of +// `state_and_queue` in the Once structure. +const INCOMPLETE: usize = 0x0; +const POISONED: usize = 0x1; +const RUNNING: usize = 0x2; +const COMPLETE: usize = 0x3; + +// Mask to learn about the state. All other bits are the queue of waiters if +// this is in the RUNNING state. +const STATE_MASK: usize = 0x3; + +// Representation of a node in the linked list of waiters, used while in the +// RUNNING state. +// Note: `Waiter` can't hold a mutable pointer to the next thread, because then +// `wait` would both hand out a mutable reference to its `Waiter` node, and keep +// a shared reference to check `signaled`. Instead we hold shared references and +// use interior mutability. +#[repr(align(4))] // Ensure the two lower bits are free to use as state bits. +struct Waiter { + thread: Cell>, + signaled: AtomicBool, + next: *const Waiter, +} + +// Head of a linked list of waiters. +// Every node is a struct on the stack of a waiting thread. +// Will wake up the waiters when it gets dropped, i.e. also on panic. +struct WaiterQueue<'a> { + state_and_queue: &'a AtomicPtr, + set_state_on_drop_to: *mut Masked, +} + +impl Once { + #[inline] + #[rustc_const_stable(feature = "const_once_new", since = "1.32.0")] + pub const fn new() -> Once { + Once { state_and_queue: AtomicPtr::new(ptr::invalid_mut(INCOMPLETE)) } + } + + #[inline] + pub fn is_completed(&self) -> bool { + // An `Acquire` load is enough because that makes all the initialization + // operations visible to us, and, this being a fast path, weaker + // ordering helps with performance. This `Acquire` synchronizes with + // `Release` operations on the slow path. + self.state_and_queue.load(Ordering::Acquire).addr() == COMPLETE + } + + // This is a non-generic function to reduce the monomorphization cost of + // using `call_once` (this isn't exactly a trivial or small implementation). + // + // Additionally, this is tagged with `#[cold]` as it should indeed be cold + // and it helps let LLVM know that calls to this function should be off the + // fast path. Essentially, this should help generate more straight line code + // in LLVM. + // + // Finally, this takes an `FnMut` instead of a `FnOnce` because there's + // currently no way to take an `FnOnce` and call it via virtual dispatch + // without some allocation overhead. + #[cold] + #[track_caller] + pub fn call(&self, ignore_poisoning: bool, init: &mut dyn FnMut(&public::OnceState)) { + let mut state_and_queue = self.state_and_queue.load(Ordering::Acquire); + loop { + match state_and_queue.addr() { + COMPLETE => break, + POISONED if !ignore_poisoning => { + // Panic to propagate the poison. + panic!("Once instance has previously been poisoned"); + } + POISONED | INCOMPLETE => { + // Try to register this thread as the one RUNNING. + let exchange_result = self.state_and_queue.compare_exchange( + state_and_queue, + ptr::invalid_mut(RUNNING), + Ordering::Acquire, + Ordering::Acquire, + ); + if let Err(old) = exchange_result { + state_and_queue = old; + continue; + } + // `waiter_queue` will manage other waiting threads, and + // wake them up on drop. + let mut waiter_queue = WaiterQueue { + state_and_queue: &self.state_and_queue, + set_state_on_drop_to: ptr::invalid_mut(POISONED), + }; + // Run the initialization function, letting it know if we're + // poisoned or not. + let init_state = public::OnceState { + inner: OnceState { + poisoned: state_and_queue.addr() == POISONED, + set_state_on_drop_to: Cell::new(ptr::invalid_mut(COMPLETE)), + }, + }; + init(&init_state); + waiter_queue.set_state_on_drop_to = init_state.inner.set_state_on_drop_to.get(); + break; + } + _ => { + // All other values must be RUNNING with possibly a + // pointer to the waiter queue in the more significant bits. + assert!(state_and_queue.addr() & STATE_MASK == RUNNING); + wait(&self.state_and_queue, state_and_queue); + state_and_queue = self.state_and_queue.load(Ordering::Acquire); + } + } + } + } +} + +fn wait(state_and_queue: &AtomicPtr, mut current_state: *mut Masked) { + // Note: the following code was carefully written to avoid creating a + // mutable reference to `node` that gets aliased. + loop { + // Don't queue this thread if the status is no longer running, + // otherwise we will not be woken up. + if current_state.addr() & STATE_MASK != RUNNING { + return; + } + + // Create the node for our current thread. + let node = Waiter { + thread: Cell::new(Some(thread::current())), + signaled: AtomicBool::new(false), + next: current_state.with_addr(current_state.addr() & !STATE_MASK) as *const Waiter, + }; + let me = &node as *const Waiter as *const Masked as *mut Masked; + + // Try to slide in the node at the head of the linked list, making sure + // that another thread didn't just replace the head of the linked list. + let exchange_result = state_and_queue.compare_exchange( + current_state, + me.with_addr(me.addr() | RUNNING), + Ordering::Release, + Ordering::Relaxed, + ); + if let Err(old) = exchange_result { + current_state = old; + continue; + } + + // We have enqueued ourselves, now lets wait. + // It is important not to return before being signaled, otherwise we + // would drop our `Waiter` node and leave a hole in the linked list + // (and a dangling reference). Guard against spurious wakeups by + // reparking ourselves until we are signaled. + while !node.signaled.load(Ordering::Acquire) { + // If the managing thread happens to signal and unpark us before we + // can park ourselves, the result could be this thread never gets + // unparked. Luckily `park` comes with the guarantee that if it got + // an `unpark` just before on an unparked thread it does not park. + thread::park(); + } + break; + } +} + +#[stable(feature = "std_debug", since = "1.16.0")] +impl fmt::Debug for Once { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + f.debug_struct("Once").finish_non_exhaustive() + } +} + +impl Drop for WaiterQueue<'_> { + fn drop(&mut self) { + // Swap out our state with however we finished. + let state_and_queue = + self.state_and_queue.swap(self.set_state_on_drop_to, Ordering::AcqRel); + + // We should only ever see an old state which was RUNNING. + assert_eq!(state_and_queue.addr() & STATE_MASK, RUNNING); + + // Walk the entire linked list of waiters and wake them up (in lifo + // order, last to register is first to wake up). + unsafe { + // Right after setting `node.signaled = true` the other thread may + // free `node` if there happens to be has a spurious wakeup. + // So we have to take out the `thread` field and copy the pointer to + // `next` first. + let mut queue = + state_and_queue.with_addr(state_and_queue.addr() & !STATE_MASK) as *const Waiter; + while !queue.is_null() { + let next = (*queue).next; + let thread = (*queue).thread.take().unwrap(); + (*queue).signaled.store(true, Ordering::Release); + // ^- FIXME (maybe): This is another case of issue #55005 + // `store()` has a potentially dangling ref to `signaled`. + queue = next; + thread.unpark(); + } + } + } +} + +impl OnceState { + #[inline] + pub fn is_poisoned(&self) -> bool { + self.poisoned + } + + #[inline] + pub fn poison(&self) { + self.set_state_on_drop_to.set(ptr::invalid_mut(POISONED)); + } +} diff --git a/library/std/src/sys_common/process.rs b/library/std/src/sys_common/process.rs index 9f978789a..18883048d 100644 --- a/library/std/src/sys_common/process.rs +++ b/library/std/src/sys_common/process.rs @@ -4,10 +4,13 @@ use crate::collections::BTreeMap; use crate::env; use crate::ffi::{OsStr, OsString}; -use crate::sys::process::EnvKey; +use crate::fmt; +use crate::io; +use crate::sys::pipe::read2; +use crate::sys::process::{EnvKey, ExitStatus, Process, StdioPipes}; // Stores a set of changes to an environment -#[derive(Clone, Debug)] +#[derive(Clone)] pub struct CommandEnv { clear: bool, saw_path: bool, @@ -20,6 +23,14 @@ impl Default for CommandEnv { } } +impl fmt::Debug for CommandEnv { + fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { + let mut debug_command_env = f.debug_struct("CommandEnv"); + debug_command_env.field("clear", &self.clear).field("vars", &self.vars); + debug_command_env.finish() + } +} + impl CommandEnv { // Capture the current environment with these changes applied pub fn capture(&self) -> BTreeMap { @@ -117,3 +128,30 @@ impl<'a> ExactSizeIterator for CommandEnvs<'a> { self.iter.is_empty() } } + +pub fn wait_with_output( + mut process: Process, + mut pipes: StdioPipes, +) -> io::Result<(ExitStatus, Vec, Vec)> { + drop(pipes.stdin.take()); + + let (mut stdout, mut stderr) = (Vec::new(), Vec::new()); + match (pipes.stdout.take(), pipes.stderr.take()) { + (None, None) => {} + (Some(out), None) => { + let res = out.read_to_end(&mut stdout); + res.unwrap(); + } + (None, Some(err)) => { + let res = err.read_to_end(&mut stderr); + res.unwrap(); + } + (Some(out), Some(err)) => { + let res = read2(out, &mut stdout, err, &mut stderr); + res.unwrap(); + } + } + + let status = process.wait()?; + Ok((status, stdout, stderr)) +} diff --git a/library/std/src/sys_common/remutex.rs b/library/std/src/sys_common/remutex.rs deleted file mode 100644 index 4c054da64..000000000 --- a/library/std/src/sys_common/remutex.rs +++ /dev/null @@ -1,178 +0,0 @@ -#[cfg(all(test, not(target_os = "emscripten")))] -mod tests; - -use crate::cell::UnsafeCell; -use crate::ops::Deref; -use crate::panic::{RefUnwindSafe, UnwindSafe}; -use crate::sync::atomic::{AtomicUsize, Ordering::Relaxed}; -use crate::sys::locks as sys; - -/// A re-entrant mutual exclusion -/// -/// This mutex will block *other* threads waiting for the lock to become -/// available. The thread which has already locked the mutex can lock it -/// multiple times without blocking, preventing a common source of deadlocks. -/// -/// This is used by stdout().lock() and friends. -/// -/// ## Implementation details -/// -/// The 'owner' field tracks which thread has locked the mutex. -/// -/// We use current_thread_unique_ptr() as the thread identifier, -/// which is just the address of a thread local variable. -/// -/// If `owner` is set to the identifier of the current thread, -/// we assume the mutex is already locked and instead of locking it again, -/// we increment `lock_count`. -/// -/// When unlocking, we decrement `lock_count`, and only unlock the mutex when -/// it reaches zero. -/// -/// `lock_count` is protected by the mutex and only accessed by the thread that has -/// locked the mutex, so needs no synchronization. -/// -/// `owner` can be checked by other threads that want to see if they already -/// hold the lock, so needs to be atomic. If it compares equal, we're on the -/// same thread that holds the mutex and memory access can use relaxed ordering -/// since we're not dealing with multiple threads. If it compares unequal, -/// synchronization is left to the mutex, making relaxed memory ordering for -/// the `owner` field fine in all cases. -pub struct ReentrantMutex { - mutex: sys::Mutex, - owner: AtomicUsize, - lock_count: UnsafeCell, - data: T, -} - -unsafe impl Send for ReentrantMutex {} -unsafe impl Sync for ReentrantMutex {} - -impl UnwindSafe for ReentrantMutex {} -impl RefUnwindSafe for ReentrantMutex {} - -/// An RAII implementation of a "scoped lock" of a mutex. When this structure is -/// dropped (falls out of scope), the lock will be unlocked. -/// -/// The data protected by the mutex can be accessed through this guard via its -/// Deref implementation. -/// -/// # Mutability -/// -/// Unlike `MutexGuard`, `ReentrantMutexGuard` does not implement `DerefMut`, -/// because implementation of the trait would violate Rust’s reference aliasing -/// rules. Use interior mutability (usually `RefCell`) in order to mutate the -/// guarded data. -#[must_use = "if unused the ReentrantMutex will immediately unlock"] -pub struct ReentrantMutexGuard<'a, T: 'a> { - lock: &'a ReentrantMutex, -} - -impl !Send for ReentrantMutexGuard<'_, T> {} - -impl ReentrantMutex { - /// Creates a new reentrant mutex in an unlocked state. - pub const fn new(t: T) -> ReentrantMutex { - ReentrantMutex { - mutex: sys::Mutex::new(), - owner: AtomicUsize::new(0), - lock_count: UnsafeCell::new(0), - data: t, - } - } - - /// Acquires a mutex, blocking the current thread until it is able to do so. - /// - /// This function will block the caller until it is available to acquire the mutex. - /// Upon returning, the thread is the only thread with the mutex held. When the thread - /// calling this method already holds the lock, the call shall succeed without - /// blocking. - /// - /// # Errors - /// - /// If another user of this mutex panicked while holding the mutex, then - /// this call will return failure if the mutex would otherwise be - /// acquired. - pub fn lock(&self) -> ReentrantMutexGuard<'_, T> { - let this_thread = current_thread_unique_ptr(); - // Safety: We only touch lock_count when we own the lock. - unsafe { - if self.owner.load(Relaxed) == this_thread { - self.increment_lock_count(); - } else { - self.mutex.lock(); - self.owner.store(this_thread, Relaxed); - debug_assert_eq!(*self.lock_count.get(), 0); - *self.lock_count.get() = 1; - } - } - ReentrantMutexGuard { lock: self } - } - - /// Attempts to acquire this lock. - /// - /// If the lock could not be acquired at this time, then `Err` is returned. - /// Otherwise, an RAII guard is returned. - /// - /// This function does not block. - /// - /// # Errors - /// - /// If another user of this mutex panicked while holding the mutex, then - /// this call will return failure if the mutex would otherwise be - /// acquired. - pub fn try_lock(&self) -> Option> { - let this_thread = current_thread_unique_ptr(); - // Safety: We only touch lock_count when we own the lock. - unsafe { - if self.owner.load(Relaxed) == this_thread { - self.increment_lock_count(); - Some(ReentrantMutexGuard { lock: self }) - } else if self.mutex.try_lock() { - self.owner.store(this_thread, Relaxed); - debug_assert_eq!(*self.lock_count.get(), 0); - *self.lock_count.get() = 1; - Some(ReentrantMutexGuard { lock: self }) - } else { - None - } - } - } - - unsafe fn increment_lock_count(&self) { - *self.lock_count.get() = (*self.lock_count.get()) - .checked_add(1) - .expect("lock count overflow in reentrant mutex"); - } -} - -impl Deref for ReentrantMutexGuard<'_, T> { - type Target = T; - - fn deref(&self) -> &T { - &self.lock.data - } -} - -impl Drop for ReentrantMutexGuard<'_, T> { - #[inline] - fn drop(&mut self) { - // Safety: We own the lock. - unsafe { - *self.lock.lock_count.get() -= 1; - if *self.lock.lock_count.get() == 0 { - self.lock.owner.store(0, Relaxed); - self.lock.mutex.unlock(); - } - } - } -} - -/// Get an address that is unique per running thread. -/// -/// This can be used as a non-null usize-sized ID. -pub fn current_thread_unique_ptr() -> usize { - // Use a non-drop type to make sure it's still available during thread destruction. - thread_local! { static X: u8 = const { 0 } } - X.with(|x| <*const _>::addr(x)) -} diff --git a/library/std/src/sys_common/remutex/tests.rs b/library/std/src/sys_common/remutex/tests.rs deleted file mode 100644 index 8e97ce11c..000000000 --- a/library/std/src/sys_common/remutex/tests.rs +++ /dev/null @@ -1,60 +0,0 @@ -use crate::cell::RefCell; -use crate::sync::Arc; -use crate::sys_common::remutex::{ReentrantMutex, ReentrantMutexGuard}; -use crate::thread; - -#[test] -fn smoke() { - let m = ReentrantMutex::new(()); - { - let a = m.lock(); - { - let b = m.lock(); - { - let c = m.lock(); - assert_eq!(*c, ()); - } - assert_eq!(*b, ()); - } - assert_eq!(*a, ()); - } -} - -#[test] -fn is_mutex() { - let m = Arc::new(ReentrantMutex::new(RefCell::new(0))); - let m2 = m.clone(); - let lock = m.lock(); - let child = thread::spawn(move || { - let lock = m2.lock(); - assert_eq!(*lock.borrow(), 4950); - }); - for i in 0..100 { - let lock = m.lock(); - *lock.borrow_mut() += i; - } - drop(lock); - child.join().unwrap(); -} - -#[test] -fn trylock_works() { - let m = Arc::new(ReentrantMutex::new(())); - let m2 = m.clone(); - let _lock = m.try_lock(); - let _lock2 = m.try_lock(); - thread::spawn(move || { - let lock = m2.try_lock(); - assert!(lock.is_none()); - }) - .join() - .unwrap(); - let _lock3 = m.try_lock(); -} - -pub struct Answer<'a>(pub ReentrantMutexGuard<'a, RefCell>); -impl Drop for Answer<'_> { - fn drop(&mut self) { - *self.0.borrow_mut() = 42; - } -} diff --git a/library/std/src/sys_common/thread_local_key.rs b/library/std/src/sys_common/thread_local_key.rs index 747579f17..2672a2a75 100644 --- a/library/std/src/sys_common/thread_local_key.rs +++ b/library/std/src/sys_common/thread_local_key.rs @@ -117,10 +117,14 @@ pub struct Key { /// This value specifies no destructor by default. pub const INIT: StaticKey = StaticKey::new(None); +// Define a sentinel value that is unlikely to be returned +// as a TLS key (but it may be returned). +const KEY_SENTVAL: usize = 0; + impl StaticKey { #[rustc_const_unstable(feature = "thread_local_internals", issue = "none")] pub const fn new(dtor: Option) -> StaticKey { - StaticKey { key: atomic::AtomicUsize::new(0), dtor } + StaticKey { key: atomic::AtomicUsize::new(KEY_SENTVAL), dtor } } /// Gets the value associated with this TLS key @@ -144,31 +148,36 @@ impl StaticKey { #[inline] unsafe fn key(&self) -> imp::Key { match self.key.load(Ordering::Relaxed) { - 0 => self.lazy_init() as imp::Key, + KEY_SENTVAL => self.lazy_init() as imp::Key, n => n as imp::Key, } } unsafe fn lazy_init(&self) -> usize { - // POSIX allows the key created here to be 0, but the compare_exchange - // below relies on using 0 as a sentinel value to check who won the + // POSIX allows the key created here to be KEY_SENTVAL, but the compare_exchange + // below relies on using KEY_SENTVAL as a sentinel value to check who won the // race to set the shared TLS key. As far as I know, there is no // guaranteed value that cannot be returned as a posix_key_create key, // so there is no value we can initialize the inner key with to // prove that it has not yet been set. As such, we'll continue using a - // value of 0, but with some gyrations to make sure we have a non-0 + // value of KEY_SENTVAL, but with some gyrations to make sure we have a non-KEY_SENTVAL // value returned from the creation routine. // FIXME: this is clearly a hack, and should be cleaned up. let key1 = imp::create(self.dtor); - let key = if key1 != 0 { + let key = if key1 as usize != KEY_SENTVAL { key1 } else { let key2 = imp::create(self.dtor); imp::destroy(key1); key2 }; - rtassert!(key != 0); - match self.key.compare_exchange(0, key as usize, Ordering::SeqCst, Ordering::SeqCst) { + rtassert!(key as usize != KEY_SENTVAL); + match self.key.compare_exchange( + KEY_SENTVAL, + key as usize, + Ordering::SeqCst, + Ordering::SeqCst, + ) { // The CAS succeeded, so we've created the actual key Ok(_) => key as usize, // If someone beat us to the punch, use their key instead diff --git a/library/std/src/sys_common/thread_parker/futex.rs b/library/std/src/sys_common/thread_parker/futex.rs deleted file mode 100644 index d9e2f39e3..000000000 --- a/library/std/src/sys_common/thread_parker/futex.rs +++ /dev/null @@ -1,97 +0,0 @@ -use crate::pin::Pin; -use crate::sync::atomic::AtomicU32; -use crate::sync::atomic::Ordering::{Acquire, Release}; -use crate::sys::futex::{futex_wait, futex_wake}; -use crate::time::Duration; - -const PARKED: u32 = u32::MAX; -const EMPTY: u32 = 0; -const NOTIFIED: u32 = 1; - -pub struct Parker { - state: AtomicU32, -} - -// Notes about memory ordering: -// -// Memory ordering is only relevant for the relative ordering of operations -// between different variables. Even Ordering::Relaxed guarantees a -// monotonic/consistent order when looking at just a single atomic variable. -// -// So, since this parker is just a single atomic variable, we only need to look -// at the ordering guarantees we need to provide to the 'outside world'. -// -// The only memory ordering guarantee that parking and unparking provide, is -// that things which happened before unpark() are visible on the thread -// returning from park() afterwards. Otherwise, it was effectively unparked -// before unpark() was called while still consuming the 'token'. -// -// In other words, unpark() needs to synchronize with the part of park() that -// consumes the token and returns. -// -// This is done with a release-acquire synchronization, by using -// Ordering::Release when writing NOTIFIED (the 'token') in unpark(), and using -// Ordering::Acquire when checking for this state in park(). -impl Parker { - /// Construct the futex parker. The UNIX parker implementation - /// requires this to happen in-place. - pub unsafe fn new(parker: *mut Parker) { - parker.write(Self { state: AtomicU32::new(EMPTY) }); - } - - // Assumes this is only called by the thread that owns the Parker, - // which means that `self.state != PARKED`. - pub unsafe fn park(self: Pin<&Self>) { - // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the - // first case. - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - loop { - // Wait for something to happen, assuming it's still set to PARKED. - futex_wait(&self.state, PARKED, None); - // Change NOTIFIED=>EMPTY and return in that case. - if self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Acquire).is_ok() { - return; - } else { - // Spurious wake up. We loop to try again. - } - } - } - - // Assumes this is only called by the thread that owns the Parker, - // which means that `self.state != PARKED`. This implementation doesn't - // require `Pin`, but other implementations do. - pub unsafe fn park_timeout(self: Pin<&Self>, timeout: Duration) { - // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the - // first case. - if self.state.fetch_sub(1, Acquire) == NOTIFIED { - return; - } - // Wait for something to happen, assuming it's still set to PARKED. - futex_wait(&self.state, PARKED, Some(timeout)); - // This is not just a store, because we need to establish a - // release-acquire ordering with unpark(). - if self.state.swap(EMPTY, Acquire) == NOTIFIED { - // Woke up because of unpark(). - } else { - // Timeout or spurious wake up. - // We return either way, because we can't easily tell if it was the - // timeout or not. - } - } - - // This implementation doesn't require `Pin`, but other implementations do. - #[inline] - pub fn unpark(self: Pin<&Self>) { - // Change PARKED=>NOTIFIED, EMPTY=>NOTIFIED, or NOTIFIED=>NOTIFIED, and - // wake the thread in the first case. - // - // Note that even NOTIFIED=>NOTIFIED results in a write. This is on - // purpose, to make sure every unpark() has a release-acquire ordering - // with park(). - if self.state.swap(NOTIFIED, Release) == PARKED { - futex_wake(&self.state); - } - } -} diff --git a/library/std/src/sys_common/thread_parker/generic.rs b/library/std/src/sys_common/thread_parker/generic.rs deleted file mode 100644 index f3d8b34d3..000000000 --- a/library/std/src/sys_common/thread_parker/generic.rs +++ /dev/null @@ -1,125 +0,0 @@ -//! Parker implementation based on a Mutex and Condvar. - -use crate::pin::Pin; -use crate::sync::atomic::AtomicUsize; -use crate::sync::atomic::Ordering::SeqCst; -use crate::sync::{Condvar, Mutex}; -use crate::time::Duration; - -const EMPTY: usize = 0; -const PARKED: usize = 1; -const NOTIFIED: usize = 2; - -pub struct Parker { - state: AtomicUsize, - lock: Mutex<()>, - cvar: Condvar, -} - -impl Parker { - /// Construct the generic parker. The UNIX parker implementation - /// requires this to happen in-place. - pub unsafe fn new(parker: *mut Parker) { - parker.write(Parker { - state: AtomicUsize::new(EMPTY), - lock: Mutex::new(()), - cvar: Condvar::new(), - }); - } - - // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. - pub unsafe fn park(self: Pin<&Self>) { - // If we were previously notified then we consume this notification and - // return quickly. - if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { - return; - } - - // Otherwise we need to coordinate going to sleep - let mut m = self.lock.lock().unwrap(); - match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { - Ok(_) => {} - Err(NOTIFIED) => { - // We must read here, even though we know it will be `NOTIFIED`. - // This is because `unpark` may have been called again since we read - // `NOTIFIED` in the `compare_exchange` above. We must perform an - // acquire operation that synchronizes with that `unpark` to observe - // any writes it made before the call to unpark. To do that we must - // read from the write it made to `state`. - let old = self.state.swap(EMPTY, SeqCst); - assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); - return; - } // should consume this notification, so prohibit spurious wakeups in next park. - Err(_) => panic!("inconsistent park state"), - } - loop { - m = self.cvar.wait(m).unwrap(); - match self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst) { - Ok(_) => return, // got a notification - Err(_) => {} // spurious wakeup, go back to sleep - } - } - } - - // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. - pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { - // Like `park` above we have a fast path for an already-notified thread, and - // afterwards we start coordinating for a sleep. - // return quickly. - if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { - return; - } - let m = self.lock.lock().unwrap(); - match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { - Ok(_) => {} - Err(NOTIFIED) => { - // We must read again here, see `park`. - let old = self.state.swap(EMPTY, SeqCst); - assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); - return; - } // should consume this notification, so prohibit spurious wakeups in next park. - Err(_) => panic!("inconsistent park_timeout state"), - } - - // Wait with a timeout, and if we spuriously wake up or otherwise wake up - // from a notification we just want to unconditionally set the state back to - // empty, either consuming a notification or un-flagging ourselves as - // parked. - let (_m, _result) = self.cvar.wait_timeout(m, dur).unwrap(); - match self.state.swap(EMPTY, SeqCst) { - NOTIFIED => {} // got a notification, hurray! - PARKED => {} // no notification, alas - n => panic!("inconsistent park_timeout state: {n}"), - } - } - - // This implementation doesn't require `Pin`, but other implementations do. - pub fn unpark(self: Pin<&Self>) { - // To ensure the unparked thread will observe any writes we made - // before this call, we must perform a release operation that `park` - // can synchronize with. To do that we must write `NOTIFIED` even if - // `state` is already `NOTIFIED`. That is why this must be a swap - // rather than a compare-and-swap that returns if it reads `NOTIFIED` - // on failure. - match self.state.swap(NOTIFIED, SeqCst) { - EMPTY => return, // no one was waiting - NOTIFIED => return, // already unparked - PARKED => {} // gotta go wake someone up - _ => panic!("inconsistent state in unpark"), - } - - // There is a period between when the parked thread sets `state` to - // `PARKED` (or last checked `state` in the case of a spurious wake - // up) and when it actually waits on `cvar`. If we were to notify - // during this period it would be ignored and then when the parked - // thread went to sleep it would never wake up. Fortunately, it has - // `lock` locked at this stage so we can acquire `lock` to wait until - // it is ready to receive the notification. - // - // Releasing `lock` before the call to `notify_one` means that when the - // parked thread wakes it doesn't get woken only to have to wait for us - // to release `lock`. - drop(self.lock.lock().unwrap()); - self.cvar.notify_one() - } -} diff --git a/library/std/src/sys_common/thread_parker/mod.rs b/library/std/src/sys_common/thread_parker/mod.rs deleted file mode 100644 index f86a9a555..000000000 --- a/library/std/src/sys_common/thread_parker/mod.rs +++ /dev/null @@ -1,23 +0,0 @@ -cfg_if::cfg_if! { - if #[cfg(any( - target_os = "linux", - target_os = "android", - all(target_arch = "wasm32", target_feature = "atomics"), - target_os = "freebsd", - target_os = "openbsd", - target_os = "dragonfly", - target_os = "fuchsia", - target_os = "hermit", - ))] { - mod futex; - pub use futex::Parker; - } else if #[cfg(target_os = "solid_asp3")] { - mod wait_flag; - pub use wait_flag::Parker; - } else if #[cfg(any(windows, target_family = "unix"))] { - pub use crate::sys::thread_parker::Parker; - } else { - mod generic; - pub use generic::Parker; - } -} diff --git a/library/std/src/sys_common/thread_parker/wait_flag.rs b/library/std/src/sys_common/thread_parker/wait_flag.rs deleted file mode 100644 index 6561c1866..000000000 --- a/library/std/src/sys_common/thread_parker/wait_flag.rs +++ /dev/null @@ -1,102 +0,0 @@ -//! A wait-flag-based thread parker. -//! -//! Some operating systems provide low-level parking primitives like wait counts, -//! event flags or semaphores which are not susceptible to race conditions (meaning -//! the wakeup can occur before the wait operation). To implement the `std` thread -//! parker on top of these primitives, we only have to ensure that parking is fast -//! when the thread token is available, the atomic ordering guarantees are maintained -//! and spurious wakeups are minimized. -//! -//! To achieve this, this parker uses an atomic variable with three states: `EMPTY`, -//! `PARKED` and `NOTIFIED`: -//! * `EMPTY` means the token has not been made available, but the thread is not -//! currently waiting on it. -//! * `PARKED` means the token is not available and the thread is parked. -//! * `NOTIFIED` means the token is available. -//! -//! `park` and `park_timeout` change the state from `EMPTY` to `PARKED` and from -//! `NOTIFIED` to `EMPTY`. If the state was `NOTIFIED`, the thread was unparked and -//! execution can continue without calling into the OS. If the state was `EMPTY`, -//! the token is not available and the thread waits on the primitive (here called -//! "wait flag"). -//! -//! `unpark` changes the state to `NOTIFIED`. If the state was `PARKED`, the thread -//! is or will be sleeping on the wait flag, so we raise it. - -use crate::pin::Pin; -use crate::sync::atomic::AtomicI8; -use crate::sync::atomic::Ordering::{Acquire, Relaxed, Release}; -use crate::sys::wait_flag::WaitFlag; -use crate::time::Duration; - -const EMPTY: i8 = 0; -const PARKED: i8 = -1; -const NOTIFIED: i8 = 1; - -pub struct Parker { - state: AtomicI8, - wait_flag: WaitFlag, -} - -impl Parker { - /// Construct a parker for the current thread. The UNIX parker - /// implementation requires this to happen in-place. - pub unsafe fn new(parker: *mut Parker) { - parker.write(Parker { state: AtomicI8::new(EMPTY), wait_flag: WaitFlag::new() }) - } - - // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. - pub unsafe fn park(self: Pin<&Self>) { - match self.state.fetch_sub(1, Acquire) { - // NOTIFIED => EMPTY - NOTIFIED => return, - // EMPTY => PARKED - EMPTY => (), - _ => panic!("inconsistent park state"), - } - - // Avoid waking up from spurious wakeups (these are quite likely, see below). - loop { - self.wait_flag.wait(); - - match self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Relaxed) { - Ok(_) => return, - Err(PARKED) => (), - Err(_) => panic!("inconsistent park state"), - } - } - } - - // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. - pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { - match self.state.fetch_sub(1, Acquire) { - NOTIFIED => return, - EMPTY => (), - _ => panic!("inconsistent park state"), - } - - self.wait_flag.wait_timeout(dur); - - // Either a wakeup or a timeout occurred. Wakeups may be spurious, as there can be - // a race condition when `unpark` is performed between receiving the timeout and - // resetting the state, resulting in the eventflag being set unnecessarily. `park` - // is protected against this by looping until the token is actually given, but - // here we cannot easily tell. - - // Use `swap` to provide acquire ordering. - match self.state.swap(EMPTY, Acquire) { - NOTIFIED => (), - PARKED => (), - _ => panic!("inconsistent park state"), - } - } - - // This implementation doesn't require `Pin`, but other implementations do. - pub fn unpark(self: Pin<&Self>) { - let state = self.state.swap(NOTIFIED, Release); - - if state == PARKED { - self.wait_flag.raise(); - } - } -} diff --git a/library/std/src/sys_common/thread_parking/futex.rs b/library/std/src/sys_common/thread_parking/futex.rs new file mode 100644 index 000000000..588e7b278 --- /dev/null +++ b/library/std/src/sys_common/thread_parking/futex.rs @@ -0,0 +1,97 @@ +use crate::pin::Pin; +use crate::sync::atomic::AtomicU32; +use crate::sync::atomic::Ordering::{Acquire, Release}; +use crate::sys::futex::{futex_wait, futex_wake}; +use crate::time::Duration; + +const PARKED: u32 = u32::MAX; +const EMPTY: u32 = 0; +const NOTIFIED: u32 = 1; + +pub struct Parker { + state: AtomicU32, +} + +// Notes about memory ordering: +// +// Memory ordering is only relevant for the relative ordering of operations +// between different variables. Even Ordering::Relaxed guarantees a +// monotonic/consistent order when looking at just a single atomic variable. +// +// So, since this parker is just a single atomic variable, we only need to look +// at the ordering guarantees we need to provide to the 'outside world'. +// +// The only memory ordering guarantee that parking and unparking provide, is +// that things which happened before unpark() are visible on the thread +// returning from park() afterwards. Otherwise, it was effectively unparked +// before unpark() was called while still consuming the 'token'. +// +// In other words, unpark() needs to synchronize with the part of park() that +// consumes the token and returns. +// +// This is done with a release-acquire synchronization, by using +// Ordering::Release when writing NOTIFIED (the 'token') in unpark(), and using +// Ordering::Acquire when checking for this state in park(). +impl Parker { + /// Construct the futex parker. The UNIX parker implementation + /// requires this to happen in-place. + pub unsafe fn new_in_place(parker: *mut Parker) { + parker.write(Self { state: AtomicU32::new(EMPTY) }); + } + + // Assumes this is only called by the thread that owns the Parker, + // which means that `self.state != PARKED`. + pub unsafe fn park(self: Pin<&Self>) { + // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the + // first case. + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + loop { + // Wait for something to happen, assuming it's still set to PARKED. + futex_wait(&self.state, PARKED, None); + // Change NOTIFIED=>EMPTY and return in that case. + if self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Acquire).is_ok() { + return; + } else { + // Spurious wake up. We loop to try again. + } + } + } + + // Assumes this is only called by the thread that owns the Parker, + // which means that `self.state != PARKED`. This implementation doesn't + // require `Pin`, but other implementations do. + pub unsafe fn park_timeout(self: Pin<&Self>, timeout: Duration) { + // Change NOTIFIED=>EMPTY or EMPTY=>PARKED, and directly return in the + // first case. + if self.state.fetch_sub(1, Acquire) == NOTIFIED { + return; + } + // Wait for something to happen, assuming it's still set to PARKED. + futex_wait(&self.state, PARKED, Some(timeout)); + // This is not just a store, because we need to establish a + // release-acquire ordering with unpark(). + if self.state.swap(EMPTY, Acquire) == NOTIFIED { + // Woke up because of unpark(). + } else { + // Timeout or spurious wake up. + // We return either way, because we can't easily tell if it was the + // timeout or not. + } + } + + // This implementation doesn't require `Pin`, but other implementations do. + #[inline] + pub fn unpark(self: Pin<&Self>) { + // Change PARKED=>NOTIFIED, EMPTY=>NOTIFIED, or NOTIFIED=>NOTIFIED, and + // wake the thread in the first case. + // + // Note that even NOTIFIED=>NOTIFIED results in a write. This is on + // purpose, to make sure every unpark() has a release-acquire ordering + // with park(). + if self.state.swap(NOTIFIED, Release) == PARKED { + futex_wake(&self.state); + } + } +} diff --git a/library/std/src/sys_common/thread_parking/generic.rs b/library/std/src/sys_common/thread_parking/generic.rs new file mode 100644 index 000000000..3209bffe3 --- /dev/null +++ b/library/std/src/sys_common/thread_parking/generic.rs @@ -0,0 +1,125 @@ +//! Parker implementation based on a Mutex and Condvar. + +use crate::pin::Pin; +use crate::sync::atomic::AtomicUsize; +use crate::sync::atomic::Ordering::SeqCst; +use crate::sync::{Condvar, Mutex}; +use crate::time::Duration; + +const EMPTY: usize = 0; +const PARKED: usize = 1; +const NOTIFIED: usize = 2; + +pub struct Parker { + state: AtomicUsize, + lock: Mutex<()>, + cvar: Condvar, +} + +impl Parker { + /// Construct the generic parker. The UNIX parker implementation + /// requires this to happen in-place. + pub unsafe fn new_in_place(parker: *mut Parker) { + parker.write(Parker { + state: AtomicUsize::new(EMPTY), + lock: Mutex::new(()), + cvar: Condvar::new(), + }); + } + + // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. + pub unsafe fn park(self: Pin<&Self>) { + // If we were previously notified then we consume this notification and + // return quickly. + if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { + return; + } + + // Otherwise we need to coordinate going to sleep + let mut m = self.lock.lock().unwrap(); + match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { + Ok(_) => {} + Err(NOTIFIED) => { + // We must read here, even though we know it will be `NOTIFIED`. + // This is because `unpark` may have been called again since we read + // `NOTIFIED` in the `compare_exchange` above. We must perform an + // acquire operation that synchronizes with that `unpark` to observe + // any writes it made before the call to unpark. To do that we must + // read from the write it made to `state`. + let old = self.state.swap(EMPTY, SeqCst); + assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); + return; + } // should consume this notification, so prohibit spurious wakeups in next park. + Err(_) => panic!("inconsistent park state"), + } + loop { + m = self.cvar.wait(m).unwrap(); + match self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst) { + Ok(_) => return, // got a notification + Err(_) => {} // spurious wakeup, go back to sleep + } + } + } + + // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + // Like `park` above we have a fast path for an already-notified thread, and + // afterwards we start coordinating for a sleep. + // return quickly. + if self.state.compare_exchange(NOTIFIED, EMPTY, SeqCst, SeqCst).is_ok() { + return; + } + let m = self.lock.lock().unwrap(); + match self.state.compare_exchange(EMPTY, PARKED, SeqCst, SeqCst) { + Ok(_) => {} + Err(NOTIFIED) => { + // We must read again here, see `park`. + let old = self.state.swap(EMPTY, SeqCst); + assert_eq!(old, NOTIFIED, "park state changed unexpectedly"); + return; + } // should consume this notification, so prohibit spurious wakeups in next park. + Err(_) => panic!("inconsistent park_timeout state"), + } + + // Wait with a timeout, and if we spuriously wake up or otherwise wake up + // from a notification we just want to unconditionally set the state back to + // empty, either consuming a notification or un-flagging ourselves as + // parked. + let (_m, _result) = self.cvar.wait_timeout(m, dur).unwrap(); + match self.state.swap(EMPTY, SeqCst) { + NOTIFIED => {} // got a notification, hurray! + PARKED => {} // no notification, alas + n => panic!("inconsistent park_timeout state: {n}"), + } + } + + // This implementation doesn't require `Pin`, but other implementations do. + pub fn unpark(self: Pin<&Self>) { + // To ensure the unparked thread will observe any writes we made + // before this call, we must perform a release operation that `park` + // can synchronize with. To do that we must write `NOTIFIED` even if + // `state` is already `NOTIFIED`. That is why this must be a swap + // rather than a compare-and-swap that returns if it reads `NOTIFIED` + // on failure. + match self.state.swap(NOTIFIED, SeqCst) { + EMPTY => return, // no one was waiting + NOTIFIED => return, // already unparked + PARKED => {} // gotta go wake someone up + _ => panic!("inconsistent state in unpark"), + } + + // There is a period between when the parked thread sets `state` to + // `PARKED` (or last checked `state` in the case of a spurious wake + // up) and when it actually waits on `cvar`. If we were to notify + // during this period it would be ignored and then when the parked + // thread went to sleep it would never wake up. Fortunately, it has + // `lock` locked at this stage so we can acquire `lock` to wait until + // it is ready to receive the notification. + // + // Releasing `lock` before the call to `notify_one` means that when the + // parked thread wakes it doesn't get woken only to have to wait for us + // to release `lock`. + drop(self.lock.lock().unwrap()); + self.cvar.notify_one() + } +} diff --git a/library/std/src/sys_common/thread_parking/id.rs b/library/std/src/sys_common/thread_parking/id.rs new file mode 100644 index 000000000..e98169597 --- /dev/null +++ b/library/std/src/sys_common/thread_parking/id.rs @@ -0,0 +1,108 @@ +//! Thread parking using thread ids. +//! +//! Some platforms (notably NetBSD) have thread parking primitives whose semantics +//! match those offered by `thread::park`, with the difference that the thread to +//! be unparked is referenced by a platform-specific thread id. Since the thread +//! parker is constructed before that id is known, an atomic state variable is used +//! to manage the park state and propagate the thread id. This also avoids platform +//! calls in the case where `unpark` is called before `park`. + +use crate::cell::UnsafeCell; +use crate::pin::Pin; +use crate::sync::atomic::{ + fence, AtomicI8, + Ordering::{Acquire, Relaxed, Release}, +}; +use crate::sys::thread_parking::{current, park, park_timeout, unpark, ThreadId}; +use crate::time::Duration; + +pub struct Parker { + state: AtomicI8, + tid: UnsafeCell>, +} + +const PARKED: i8 = -1; +const EMPTY: i8 = 0; +const NOTIFIED: i8 = 1; + +impl Parker { + pub fn new() -> Parker { + Parker { state: AtomicI8::new(EMPTY), tid: UnsafeCell::new(None) } + } + + /// Create a new thread parker. UNIX requires this to happen in-place. + pub unsafe fn new_in_place(parker: *mut Parker) { + parker.write(Parker::new()) + } + + /// # Safety + /// * must always be called from the same thread + /// * must be called before the state is set to PARKED + unsafe fn init_tid(&self) { + // The field is only ever written to from this thread, so we don't need + // synchronization to read it here. + if self.tid.get().read().is_none() { + // Because this point is only reached once, before the state is set + // to PARKED for the first time, the non-atomic write here can not + // conflict with reads by other threads. + self.tid.get().write(Some(current())); + // Ensure that the write can be observed by all threads reading the + // state. Synchronizes with the acquire barrier in `unpark`. + fence(Release); + } + } + + pub unsafe fn park(self: Pin<&Self>) { + self.init_tid(); + + // Changes NOTIFIED to EMPTY and EMPTY to PARKED. + let mut state = self.state.fetch_sub(1, Acquire).wrapping_sub(1); + if state == PARKED { + // Loop to guard against spurious wakeups. + while state == PARKED { + park(self.state.as_mut_ptr().addr()); + state = self.state.load(Acquire); + } + + // Since the state change has already been observed with acquire + // ordering, the state can be reset with a relaxed store instead + // of a swap. + self.state.store(EMPTY, Relaxed); + } + } + + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + self.init_tid(); + + let state = self.state.fetch_sub(1, Acquire).wrapping_sub(1); + if state == PARKED { + park_timeout(dur, self.state.as_mut_ptr().addr()); + // Swap to ensure that we observe all state changes with acquire + // ordering, even if the state has been changed after the timeout + // occured. + self.state.swap(EMPTY, Acquire); + } + } + + pub fn unpark(self: Pin<&Self>) { + let state = self.state.swap(NOTIFIED, Release); + if state == PARKED { + // Synchronize with the release fence in `init_tid` to observe the + // write to `tid`. + fence(Acquire); + // # Safety + // The thread id is initialized before the state is set to `PARKED` + // for the first time and is not written to from that point on + // (negating the need for an atomic read). + let tid = unsafe { self.tid.get().read().unwrap_unchecked() }; + // It is possible that the waiting thread woke up because of a timeout + // and terminated before this call is made. This call then returns an + // error or wakes up an unrelated thread. The platform API and + // environment does allow this, however. + unpark(tid, self.state.as_mut_ptr().addr()); + } + } +} + +unsafe impl Send for Parker {} +unsafe impl Sync for Parker {} diff --git a/library/std/src/sys_common/thread_parking/mod.rs b/library/std/src/sys_common/thread_parking/mod.rs new file mode 100644 index 000000000..0ead6633c --- /dev/null +++ b/library/std/src/sys_common/thread_parking/mod.rs @@ -0,0 +1,29 @@ +cfg_if::cfg_if! { + if #[cfg(any( + target_os = "linux", + target_os = "android", + all(target_arch = "wasm32", target_feature = "atomics"), + target_os = "freebsd", + target_os = "openbsd", + target_os = "dragonfly", + target_os = "fuchsia", + target_os = "hermit", + ))] { + mod futex; + pub use futex::Parker; + } else if #[cfg(any( + target_os = "netbsd", + all(target_vendor = "fortanix", target_env = "sgx"), + ))] { + mod id; + pub use id::Parker; + } else if #[cfg(target_os = "solid_asp3")] { + mod wait_flag; + pub use wait_flag::Parker; + } else if #[cfg(any(windows, target_family = "unix"))] { + pub use crate::sys::thread_parking::Parker; + } else { + mod generic; + pub use generic::Parker; + } +} diff --git a/library/std/src/sys_common/thread_parking/wait_flag.rs b/library/std/src/sys_common/thread_parking/wait_flag.rs new file mode 100644 index 000000000..d0f8899a9 --- /dev/null +++ b/library/std/src/sys_common/thread_parking/wait_flag.rs @@ -0,0 +1,102 @@ +//! A wait-flag-based thread parker. +//! +//! Some operating systems provide low-level parking primitives like wait counts, +//! event flags or semaphores which are not susceptible to race conditions (meaning +//! the wakeup can occur before the wait operation). To implement the `std` thread +//! parker on top of these primitives, we only have to ensure that parking is fast +//! when the thread token is available, the atomic ordering guarantees are maintained +//! and spurious wakeups are minimized. +//! +//! To achieve this, this parker uses an atomic variable with three states: `EMPTY`, +//! `PARKED` and `NOTIFIED`: +//! * `EMPTY` means the token has not been made available, but the thread is not +//! currently waiting on it. +//! * `PARKED` means the token is not available and the thread is parked. +//! * `NOTIFIED` means the token is available. +//! +//! `park` and `park_timeout` change the state from `EMPTY` to `PARKED` and from +//! `NOTIFIED` to `EMPTY`. If the state was `NOTIFIED`, the thread was unparked and +//! execution can continue without calling into the OS. If the state was `EMPTY`, +//! the token is not available and the thread waits on the primitive (here called +//! "wait flag"). +//! +//! `unpark` changes the state to `NOTIFIED`. If the state was `PARKED`, the thread +//! is or will be sleeping on the wait flag, so we raise it. + +use crate::pin::Pin; +use crate::sync::atomic::AtomicI8; +use crate::sync::atomic::Ordering::{Acquire, Relaxed, Release}; +use crate::sys::wait_flag::WaitFlag; +use crate::time::Duration; + +const EMPTY: i8 = 0; +const PARKED: i8 = -1; +const NOTIFIED: i8 = 1; + +pub struct Parker { + state: AtomicI8, + wait_flag: WaitFlag, +} + +impl Parker { + /// Construct a parker for the current thread. The UNIX parker + /// implementation requires this to happen in-place. + pub unsafe fn new_in_place(parker: *mut Parker) { + parker.write(Parker { state: AtomicI8::new(EMPTY), wait_flag: WaitFlag::new() }) + } + + // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. + pub unsafe fn park(self: Pin<&Self>) { + match self.state.fetch_sub(1, Acquire) { + // NOTIFIED => EMPTY + NOTIFIED => return, + // EMPTY => PARKED + EMPTY => (), + _ => panic!("inconsistent park state"), + } + + // Avoid waking up from spurious wakeups (these are quite likely, see below). + loop { + self.wait_flag.wait(); + + match self.state.compare_exchange(NOTIFIED, EMPTY, Acquire, Relaxed) { + Ok(_) => return, + Err(PARKED) => (), + Err(_) => panic!("inconsistent park state"), + } + } + } + + // This implementation doesn't require `unsafe` and `Pin`, but other implementations do. + pub unsafe fn park_timeout(self: Pin<&Self>, dur: Duration) { + match self.state.fetch_sub(1, Acquire) { + NOTIFIED => return, + EMPTY => (), + _ => panic!("inconsistent park state"), + } + + self.wait_flag.wait_timeout(dur); + + // Either a wakeup or a timeout occurred. Wakeups may be spurious, as there can be + // a race condition when `unpark` is performed between receiving the timeout and + // resetting the state, resulting in the eventflag being set unnecessarily. `park` + // is protected against this by looping until the token is actually given, but + // here we cannot easily tell. + + // Use `swap` to provide acquire ordering. + match self.state.swap(EMPTY, Acquire) { + NOTIFIED => (), + PARKED => (), + _ => panic!("inconsistent park state"), + } + } + + // This implementation doesn't require `Pin`, but other implementations do. + pub fn unpark(self: Pin<&Self>) { + let state = self.state.swap(NOTIFIED, Release); + + if state == PARKED { + self.wait_flag.raise(); + } + } +} diff --git a/library/std/src/thread/local.rs b/library/std/src/thread/local.rs index 5d267891b..b30bb7b77 100644 --- a/library/std/src/thread/local.rs +++ b/library/std/src/thread/local.rs @@ -905,9 +905,8 @@ pub mod statik { pub mod fast { use super::lazy::LazyKeyInner; use crate::cell::Cell; - use crate::fmt; - use crate::mem; use crate::sys::thread_local_dtor::register_dtor; + use crate::{fmt, mem, panic}; #[derive(Copy, Clone)] enum DtorState { @@ -950,7 +949,7 @@ pub mod fast { // note that this is just a publicly-callable function only for the // const-initialized form of thread locals, basically a way to call the - // free `register_dtor` function defined elsewhere in libstd. + // free `register_dtor` function defined elsewhere in std. pub unsafe fn register_dtor(a: *mut u8, dtor: unsafe extern "C" fn(*mut u8)) { unsafe { register_dtor(a, dtor); @@ -1028,10 +1027,15 @@ pub mod fast { // `Option` to `None`, and `dtor_state` to `RunningOrHasRun`. This // causes future calls to `get` to run `try_initialize_drop` again, // which will now fail, and return `None`. - unsafe { + // + // Wrap the call in a catch to ensure unwinding is caught in the event + // a panic takes place in a destructor. + if let Err(_) = panic::catch_unwind(panic::AssertUnwindSafe(|| unsafe { let value = (*ptr).inner.take(); (*ptr).dtor_state.set(DtorState::RunningOrHasRun); drop(value); + })) { + rtabort!("thread local panicked on drop"); } } } @@ -1044,10 +1048,8 @@ pub mod fast { pub mod os { use super::lazy::LazyKeyInner; use crate::cell::Cell; - use crate::fmt; - use crate::marker; - use crate::ptr; use crate::sys_common::thread_local_key::StaticKey as OsStaticKey; + use crate::{fmt, marker, panic, ptr}; /// Use a regular global static to store this key; the state provided will then be /// thread-local. @@ -1137,12 +1139,17 @@ pub mod os { // // Note that to prevent an infinite loop we reset it back to null right // before we return from the destructor ourselves. - unsafe { + // + // Wrap the call in a catch to ensure unwinding is caught in the event + // a panic takes place in a destructor. + if let Err(_) = panic::catch_unwind(|| unsafe { let ptr = Box::from_raw(ptr as *mut Value); let key = ptr.key; key.os.set(ptr::invalid_mut(1)); drop(ptr); key.os.set(ptr::null_mut()); + }) { + rtabort!("thread local panicked on drop"); } } } diff --git a/library/std/src/thread/local/tests.rs b/library/std/src/thread/local/tests.rs index 80dc4c038..964c7fc5b 100644 --- a/library/std/src/thread/local/tests.rs +++ b/library/std/src/thread/local/tests.rs @@ -23,11 +23,11 @@ impl Signal { } } -struct Foo(Signal); +struct NotifyOnDrop(Signal); -impl Drop for Foo { +impl Drop for NotifyOnDrop { fn drop(&mut self) { - let Foo(ref f) = *self; + let NotifyOnDrop(ref f) = *self; f.notify(); } } @@ -82,18 +82,18 @@ fn states() { #[test] fn smoke_dtor() { - thread_local!(static FOO: UnsafeCell> = UnsafeCell::new(None)); + thread_local!(static FOO: UnsafeCell> = UnsafeCell::new(None)); run(&FOO); - thread_local!(static FOO2: UnsafeCell> = const { UnsafeCell::new(None) }); + thread_local!(static FOO2: UnsafeCell> = const { UnsafeCell::new(None) }); run(&FOO2); - fn run(key: &'static LocalKey>>) { + fn run(key: &'static LocalKey>>) { let signal = Signal::default(); let signal2 = signal.clone(); let t = thread::spawn(move || unsafe { let mut signal = Some(signal2); key.with(|f| { - *f.get() = Some(Foo(signal.take().unwrap())); + *f.get() = Some(NotifyOnDrop(signal.take().unwrap())); }); }); signal.wait(); @@ -187,13 +187,13 @@ fn self_referential() { fn dtors_in_dtors_in_dtors() { struct S1(Signal); thread_local!(static K1: UnsafeCell> = UnsafeCell::new(None)); - thread_local!(static K2: UnsafeCell> = UnsafeCell::new(None)); + thread_local!(static K2: UnsafeCell> = UnsafeCell::new(None)); impl Drop for S1 { fn drop(&mut self) { let S1(ref signal) = *self; unsafe { - let _ = K2.try_with(|s| *s.get() = Some(Foo(signal.clone()))); + let _ = K2.try_with(|s| *s.get() = Some(NotifyOnDrop(signal.clone()))); } } } @@ -211,13 +211,13 @@ fn dtors_in_dtors_in_dtors() { fn dtors_in_dtors_in_dtors_const_init() { struct S1(Signal); thread_local!(static K1: UnsafeCell> = const { UnsafeCell::new(None) }); - thread_local!(static K2: UnsafeCell> = const { UnsafeCell::new(None) }); + thread_local!(static K2: UnsafeCell> = const { UnsafeCell::new(None) }); impl Drop for S1 { fn drop(&mut self) { let S1(ref signal) = *self; unsafe { - let _ = K2.try_with(|s| *s.get() = Some(Foo(signal.clone()))); + let _ = K2.try_with(|s| *s.get() = Some(NotifyOnDrop(signal.clone()))); } } } diff --git a/library/std/src/thread/mod.rs b/library/std/src/thread/mod.rs index 34bdb8bd4..692ff0cbc 100644 --- a/library/std/src/thread/mod.rs +++ b/library/std/src/thread/mod.rs @@ -173,10 +173,16 @@ use crate::sync::Arc; use crate::sys::thread as imp; use crate::sys_common::thread; use crate::sys_common::thread_info; -use crate::sys_common::thread_parker::Parker; +use crate::sys_common::thread_parking::Parker; use crate::sys_common::{AsInner, IntoInner}; use crate::time::Duration; +#[stable(feature = "scoped_threads", since = "1.63.0")] +mod scoped; + +#[stable(feature = "scoped_threads", since = "1.63.0")] +pub use scoped::{scope, Scope, ScopedJoinHandle}; + //////////////////////////////////////////////////////////////////////////////// // Thread-local storage //////////////////////////////////////////////////////////////////////////////// @@ -184,12 +190,6 @@ use crate::time::Duration; #[macro_use] mod local; -#[stable(feature = "scoped_threads", since = "1.63.0")] -mod scoped; - -#[stable(feature = "scoped_threads", since = "1.63.0")] -pub use scoped::{scope, Scope, ScopedJoinHandle}; - #[stable(feature = "rust1", since = "1.0.0")] pub use self::local::{AccessError, LocalKey}; @@ -209,7 +209,6 @@ pub use self::local::{AccessError, LocalKey}; ))] #[doc(hidden)] pub use self::local::fast::Key as __FastLocalKeyInner; - // when building for tests, use real std's type #[unstable(feature = "libstd_thread_internals", issue = "none")] #[cfg(test)] @@ -220,12 +219,21 @@ pub use self::local::fast::Key as __FastLocalKeyInner; pub use realstd::thread::__FastLocalKeyInner; #[unstable(feature = "libstd_thread_internals", issue = "none")] +#[cfg(not(test))] #[cfg(all( not(target_thread_local), not(all(target_family = "wasm", not(target_feature = "atomics"))), ))] #[doc(hidden)] pub use self::local::os::Key as __OsLocalKeyInner; +// when building for tests, use real std's type +#[unstable(feature = "libstd_thread_internals", issue = "none")] +#[cfg(test)] +#[cfg(all( + not(target_thread_local), + not(all(target_family = "wasm", not(target_feature = "atomics"))), +))] +pub use realstd::thread::__OsLocalKeyInner; #[unstable(feature = "libstd_thread_internals", issue = "none")] #[cfg(all(target_family = "wasm", not(target_feature = "atomics")))] @@ -1216,7 +1224,7 @@ impl Thread { let ptr = Arc::get_mut_unchecked(&mut arc).as_mut_ptr(); addr_of_mut!((*ptr).name).write(name); addr_of_mut!((*ptr).id).write(ThreadId::new()); - Parker::new(addr_of_mut!((*ptr).parker)); + Parker::new_in_place(addr_of_mut!((*ptr).parker)); Pin::new_unchecked(arc.assume_init()) }; diff --git a/library/std/tests/env.rs b/library/std/tests/env.rs index b095c2dde..aae2c49d8 100644 --- a/library/std/tests/env.rs +++ b/library/std/tests/env.rs @@ -1,12 +1,24 @@ use std::env::*; use std::ffi::{OsStr, OsString}; -use rand::distributions::Alphanumeric; -use rand::{thread_rng, Rng}; +use rand::distributions::{Alphanumeric, DistString}; + +/// Copied from `std::test_helpers::test_rng`, since these tests rely on the +/// seed not being the same for every RNG invocation too. +#[track_caller] +pub(crate) fn test_rng() -> rand_xorshift::XorShiftRng { + use core::hash::{BuildHasher, Hash, Hasher}; + let mut hasher = std::collections::hash_map::RandomState::new().build_hasher(); + core::panic::Location::caller().hash(&mut hasher); + let hc64 = hasher.finish(); + let seed_vec = hc64.to_le_bytes().into_iter().chain(0u8..8).collect::>(); + let seed: [u8; 16] = seed_vec.as_slice().try_into().unwrap(); + rand::SeedableRng::from_seed(seed) +} +#[track_caller] fn make_rand_name() -> OsString { - let rng = thread_rng(); - let n = format!("TEST{}", rng.sample_iter(&Alphanumeric).take(10).collect::()); + let n = format!("TEST{}", Alphanumeric.sample_string(&mut test_rng(), 10)); let n = OsString::from(n); assert!(var_os(&n).is_none()); n diff --git a/library/std/tests/run-time-detect.rs b/library/std/tests/run-time-detect.rs index 02c076f1b..1a2c12556 100644 --- a/library/std/tests/run-time-detect.rs +++ b/library/std/tests/run-time-detect.rs @@ -1,9 +1,8 @@ -//! These tests just check that the macros are available in libstd. +//! These tests just check that the macros are available in std. #![cfg_attr( any( all(target_arch = "arm", any(target_os = "linux", target_os = "android")), - all(bootstrap, target_arch = "aarch64", any(target_os = "linux", target_os = "android")), all(target_arch = "powerpc", target_os = "linux"), all(target_arch = "powerpc64", target_os = "linux"), ), diff --git a/library/test/src/cli.rs b/library/test/src/cli.rs index 524658bce..796796e07 100644 --- a/library/test/src/cli.rs +++ b/library/test/src/cli.rs @@ -354,8 +354,7 @@ fn get_shuffle_seed(matches: &getopts::Matches, allow_unstable: bool) -> OptPart Err(e) => { return Err(format!( "argument for --shuffle-seed must be a number \ - (error: {})", - e + (error: {e})" )); } }, @@ -383,8 +382,7 @@ fn get_test_threads(matches: &getopts::Matches) -> OptPartRes> { Err(e) => { return Err(format!( "argument for --test-threads must be a number > 0 \ - (error: {})", - e + (error: {e})" )); } }, @@ -418,8 +416,7 @@ fn get_format( Some(v) => { return Err(format!( "argument for --format must be pretty, terse, json or junit (was \ - {})", - v + {v})" )); } }; @@ -436,8 +433,7 @@ fn get_color_config(matches: &getopts::Matches) -> OptPartRes { Some(v) => { return Err(format!( "argument for --color must be auto, always, or never (was \ - {})", - v + {v})" )); } }; diff --git a/library/test/src/console.rs b/library/test/src/console.rs index a3c39f71f..24cbe035f 100644 --- a/library/test/src/console.rs +++ b/library/test/src/console.rs @@ -147,7 +147,7 @@ pub fn list_tests_console(opts: &TestOpts, tests: Vec) -> io::Res let mut ntest = 0; let mut nbench = 0; - for test in filter_tests(&opts, tests).into_iter() { + for test in filter_tests(opts, tests).into_iter() { use crate::TestFn::*; let TestDescAndFn { desc: TestDesc { name, .. }, testfn } = test; @@ -244,7 +244,7 @@ fn on_test_event( let stdout = &completed_test.stdout; st.write_log_result(test, result, exec_time.as_ref())?; - out.write_result(test, result, exec_time.as_ref(), &*stdout, st)?; + out.write_result(test, result, exec_time.as_ref(), stdout, st)?; handle_test_result(st, completed_test); } } @@ -262,7 +262,7 @@ pub fn run_tests_console(opts: &TestOpts, tests: Vec) -> io::Resu let max_name_len = tests .iter() - .max_by_key(|t| len_if_padded(*t)) + .max_by_key(|t| len_if_padded(t)) .map(|t| t.desc.name.as_slice().len()) .unwrap_or(0); diff --git a/library/test/src/formatters/json.rs b/library/test/src/formatters/json.rs index c07fdafb1..95d2faf25 100644 --- a/library/test/src/formatters/json.rs +++ b/library/test/src/formatters/json.rs @@ -40,20 +40,20 @@ impl JsonFormatter { extra: Option<&str>, ) -> io::Result<()> { // A doc test's name includes a filename which must be escaped for correct json. - self.write_message(&*format!( + self.write_message(&format!( r#"{{ "type": "{}", "name": "{}", "event": "{}""#, ty, EscapedString(name), evt ))?; if let Some(exec_time) = exec_time { - self.write_message(&*format!(r#", "exec_time": {}"#, exec_time.0.as_secs_f64()))?; + self.write_message(&format!(r#", "exec_time": {}"#, exec_time.0.as_secs_f64()))?; } if let Some(stdout) = stdout { - self.write_message(&*format!(r#", "stdout": "{}""#, EscapedString(stdout)))?; + self.write_message(&format!(r#", "stdout": "{}""#, EscapedString(stdout)))?; } if let Some(extra) = extra { - self.write_message(&*format!(r#", {}"#, extra))?; + self.write_message(&format!(r#", {extra}"#))?; } self.writeln_message(" }") } @@ -62,18 +62,17 @@ impl JsonFormatter { impl OutputFormatter for JsonFormatter { fn write_run_start(&mut self, test_count: usize, shuffle_seed: Option) -> io::Result<()> { let shuffle_seed_json = if let Some(shuffle_seed) = shuffle_seed { - format!(r#", "shuffle_seed": {}"#, shuffle_seed) + format!(r#", "shuffle_seed": {shuffle_seed}"#) } else { String::new() }; - self.writeln_message(&*format!( - r#"{{ "type": "suite", "event": "started", "test_count": {}{} }}"#, - test_count, shuffle_seed_json + self.writeln_message(&format!( + r#"{{ "type": "suite", "event": "started", "test_count": {test_count}{shuffle_seed_json} }}"# )) } fn write_test_start(&mut self, desc: &TestDesc) -> io::Result<()> { - self.writeln_message(&*format!( + self.writeln_message(&format!( r#"{{ "type": "test", "event": "started", "name": "{}" }}"#, EscapedString(desc.name.as_slice()) )) @@ -152,20 +151,20 @@ impl OutputFormatter for JsonFormatter { mbps ); - self.writeln_message(&*line) + self.writeln_message(&line) } } } fn write_timeout(&mut self, desc: &TestDesc) -> io::Result<()> { - self.writeln_message(&*format!( + self.writeln_message(&format!( r#"{{ "type": "test", "event": "timeout", "name": "{}" }}"#, EscapedString(desc.name.as_slice()) )) } fn write_run_finish(&mut self, state: &ConsoleTestState) -> io::Result { - self.write_message(&*format!( + self.write_message(&format!( "{{ \"type\": \"suite\", \ \"event\": \"{}\", \ \"passed\": {}, \ diff --git a/library/test/src/formatters/junit.rs b/library/test/src/formatters/junit.rs index e6fb4f570..7a40ce33c 100644 --- a/library/test/src/formatters/junit.rs +++ b/library/test/src/formatters/junit.rs @@ -64,7 +64,7 @@ impl OutputFormatter for JunitFormatter { fn write_run_finish(&mut self, state: &ConsoleTestState) -> io::Result { self.write_message("")?; - self.write_message(&*format!( + self.write_message(&format!( " OutputFormatter for JunitFormatter { >", state.failed, state.total, state.ignored ))?; - for (desc, result, duration) in std::mem::replace(&mut self.results, Vec::new()) { + for (desc, result, duration) in std::mem::take(&mut self.results) { let (class_name, test_name) = parse_class_name(&desc); match result { TestResult::TrIgnored => { /* no-op */ } TestResult::TrFailed => { - self.write_message(&*format!( + self.write_message(&format!( "", class_name, @@ -90,19 +90,19 @@ impl OutputFormatter for JunitFormatter { } TestResult::TrFailedMsg(ref m) => { - self.write_message(&*format!( + self.write_message(&format!( "", class_name, test_name, duration.as_secs_f64() ))?; - self.write_message(&*format!(""))?; + self.write_message(&format!(""))?; self.write_message("")?; } TestResult::TrTimedFail => { - self.write_message(&*format!( + self.write_message(&format!( "", class_name, @@ -114,7 +114,7 @@ impl OutputFormatter for JunitFormatter { } TestResult::TrBench(ref b) => { - self.write_message(&*format!( + self.write_message(&format!( "", class_name, test_name, b.ns_iter_summ.sum @@ -122,7 +122,7 @@ impl OutputFormatter for JunitFormatter { } TestResult::TrOk => { - self.write_message(&*format!( + self.write_message(&format!( "", class_name, diff --git a/library/test/src/formatters/mod.rs b/library/test/src/formatters/mod.rs index cb8085975..cb67b6491 100644 --- a/library/test/src/formatters/mod.rs +++ b/library/test/src/formatters/mod.rs @@ -38,5 +38,5 @@ pub(crate) fn write_stderr_delimiter(test_output: &mut Vec, test_name: &Test Some(_) => test_output.push(b'\n'), None => (), } - writeln!(test_output, "---- {} stderr ----", test_name).unwrap(); + writeln!(test_output, "---- {test_name} stderr ----").unwrap(); } diff --git a/library/test/src/formatters/pretty.rs b/library/test/src/formatters/pretty.rs index 694202229..247778e51 100644 --- a/library/test/src/formatters/pretty.rs +++ b/library/test/src/formatters/pretty.rs @@ -47,7 +47,7 @@ impl PrettyFormatter { pub fn write_ignored(&mut self, message: Option<&'static str>) -> io::Result<()> { if let Some(message) = message { - self.write_short_result(&format!("ignored, {}", message), term::color::YELLOW) + self.write_short_result(&format!("ignored, {message}"), term::color::YELLOW) } else { self.write_short_result("ignored", term::color::YELLOW) } @@ -134,7 +134,7 @@ impl PrettyFormatter { let mut results = Vec::new(); let mut stdouts = String::new(); - for &(ref f, ref stdout) in inputs { + for (f, stdout) in inputs { results.push(f.name.to_string()); if !stdout.is_empty() { stdouts.push_str(&format!("---- {} stdout ----\n", f.name)); @@ -171,9 +171,9 @@ impl PrettyFormatter { fn write_test_name(&mut self, desc: &TestDesc) -> io::Result<()> { let name = desc.padded_name(self.max_name_len, desc.name.padding()); if let Some(test_mode) = desc.test_mode() { - self.write_plain(&format!("test {name} - {test_mode} ... "))?; + self.write_plain(format!("test {name} - {test_mode} ... "))?; } else { - self.write_plain(&format!("test {name} ... "))?; + self.write_plain(format!("test {name} ... "))?; } Ok(()) @@ -188,7 +188,7 @@ impl OutputFormatter for PrettyFormatter { } else { String::new() }; - self.write_plain(&format!("\nrunning {test_count} {noun}{shuffle_seed_msg}\n")) + self.write_plain(format!("\nrunning {test_count} {noun}{shuffle_seed_msg}\n")) } fn write_test_start(&mut self, desc: &TestDesc) -> io::Result<()> { @@ -221,7 +221,7 @@ impl OutputFormatter for PrettyFormatter { TestResult::TrIgnored => self.write_ignored(desc.ignore_message)?, TestResult::TrBench(ref bs) => { self.write_bench()?; - self.write_plain(&format!(": {}", fmt_bench_samples(bs)))?; + self.write_plain(format!(": {}", fmt_bench_samples(bs)))?; } TestResult::TrTimedFail => self.write_time_failed()?, } @@ -231,7 +231,7 @@ impl OutputFormatter for PrettyFormatter { } fn write_timeout(&mut self, desc: &TestDesc) -> io::Result<()> { - self.write_plain(&format!( + self.write_plain(format!( "test {} has been running for over {} seconds\n", desc.name, time::TEST_WARN_TIMEOUT_S @@ -267,11 +267,11 @@ impl OutputFormatter for PrettyFormatter { state.passed, state.failed, state.ignored, state.measured, state.filtered_out ); - self.write_plain(&s)?; + self.write_plain(s)?; if let Some(ref exec_time) = state.exec_time { let time_str = format!("; finished in {exec_time}"); - self.write_plain(&time_str)?; + self.write_plain(time_str)?; } self.write_plain("\n\n")?; diff --git a/library/test/src/formatters/terse.rs b/library/test/src/formatters/terse.rs index 5dace8bae..0837ab169 100644 --- a/library/test/src/formatters/terse.rs +++ b/library/test/src/formatters/terse.rs @@ -70,7 +70,7 @@ impl TerseFormatter { // screen when dealing with line-buffered output (e.g., piping to // `stamp` in the rust CI). let out = format!(" {}/{}\n", self.test_count + 1, self.total_test_count); - self.write_plain(&out)?; + self.write_plain(out)?; } self.test_count += 1; @@ -106,7 +106,7 @@ impl TerseFormatter { self.write_plain("\nsuccesses:\n")?; let mut successes = Vec::new(); let mut stdouts = String::new(); - for &(ref f, ref stdout) in &state.not_failures { + for (f, stdout) in &state.not_failures { successes.push(f.name.to_string()); if !stdout.is_empty() { stdouts.push_str(&format!("---- {} stdout ----\n", f.name)); @@ -132,7 +132,7 @@ impl TerseFormatter { self.write_plain("\nfailures:\n")?; let mut failures = Vec::new(); let mut fail_out = String::new(); - for &(ref f, ref stdout) in &state.failures { + for (f, stdout) in &state.failures { failures.push(f.name.to_string()); if !stdout.is_empty() { fail_out.push_str(&format!("---- {} stdout ----\n", f.name)); @@ -157,9 +157,9 @@ impl TerseFormatter { fn write_test_name(&mut self, desc: &TestDesc) -> io::Result<()> { let name = desc.padded_name(self.max_name_len, desc.name.padding()); if let Some(test_mode) = desc.test_mode() { - self.write_plain(&format!("test {name} - {test_mode} ... "))?; + self.write_plain(format!("test {name} - {test_mode} ... "))?; } else { - self.write_plain(&format!("test {name} ... "))?; + self.write_plain(format!("test {name} ... "))?; } Ok(()) @@ -175,7 +175,7 @@ impl OutputFormatter for TerseFormatter { } else { String::new() }; - self.write_plain(&format!("\nrunning {test_count} {noun}{shuffle_seed_msg}\n")) + self.write_plain(format!("\nrunning {test_count} {noun}{shuffle_seed_msg}\n")) } fn write_test_start(&mut self, desc: &TestDesc) -> io::Result<()> { @@ -209,13 +209,13 @@ impl OutputFormatter for TerseFormatter { self.write_test_name(desc)?; } self.write_bench()?; - self.write_plain(&format!(": {}\n", fmt_bench_samples(bs))) + self.write_plain(format!(": {}\n", fmt_bench_samples(bs))) } } } fn write_timeout(&mut self, desc: &TestDesc) -> io::Result<()> { - self.write_plain(&format!( + self.write_plain(format!( "test {} has been running for over {} seconds\n", desc.name, time::TEST_WARN_TIMEOUT_S @@ -245,11 +245,11 @@ impl OutputFormatter for TerseFormatter { state.passed, state.failed, state.ignored, state.measured, state.filtered_out ); - self.write_plain(&s)?; + self.write_plain(s)?; if let Some(ref exec_time) = state.exec_time { let time_str = format!("; finished in {exec_time}"); - self.write_plain(&time_str)?; + self.write_plain(time_str)?; } self.write_plain("\n\n")?; diff --git a/library/test/src/lib.rs b/library/test/src/lib.rs index 256c9e8d1..69fb529d7 100644 --- a/library/test/src/lib.rs +++ b/library/test/src/lib.rs @@ -116,7 +116,7 @@ pub fn test_main(args: &[String], tests: Vec, options: Option(result: T) -> Result<(), String> { } else { Err(format!( "the test returned a termination value with a non-zero status code \ - ({}) which indicates a failure", - code + ({code}) which indicates a failure" )) } } @@ -750,7 +749,7 @@ fn spawn_test_subprocess( })() { Ok(r) => r, Err(e) => { - write!(&mut test_output, "Unexpected error: {}", e).unwrap(); + write!(&mut test_output, "Unexpected error: {e}").unwrap(); TrFailed } }; @@ -790,7 +789,7 @@ fn run_test_in_spawned_subprocess( } }); let record_result2 = record_result.clone(); - panic::set_hook(Box::new(move |info| record_result2(Some(&info)))); + panic::set_hook(Box::new(move |info| record_result2(Some(info)))); if let Err(message) = testfn() { panic!("{}", message); } diff --git a/library/test/src/term/terminfo/mod.rs b/library/test/src/term/terminfo/mod.rs index 355859019..67ba89410 100644 --- a/library/test/src/term/terminfo/mod.rs +++ b/library/test/src/term/terminfo/mod.rs @@ -149,7 +149,7 @@ impl Terminal for TerminfoTerminal { // are there any terminals that have color/attrs and not sgr0? // Try falling back to sgr, then op let cmd = match ["sgr0", "sgr", "op"].iter().find_map(|cap| self.ti.strings.get(*cap)) { - Some(op) => match expand(&op, &[], &mut Variables::new()) { + Some(op) => match expand(op, &[], &mut Variables::new()) { Ok(cmd) => cmd, Err(e) => return Err(io::Error::new(io::ErrorKind::InvalidData, e)), }, @@ -180,12 +180,12 @@ impl TerminfoTerminal { } fn dim_if_necessary(&self, color: color::Color) -> color::Color { - if color >= self.num_colors && color >= 8 && color < 16 { color - 8 } else { color } + if color >= self.num_colors && (8..16).contains(&color) { color - 8 } else { color } } fn apply_cap(&mut self, cmd: &str, params: &[Param]) -> io::Result { match self.ti.strings.get(cmd) { - Some(cmd) => match expand(&cmd, params, &mut Variables::new()) { + Some(cmd) => match expand(cmd, params, &mut Variables::new()) { Ok(s) => self.out.write_all(&s).and(Ok(true)), Err(e) => Err(io::Error::new(io::ErrorKind::InvalidData, e)), }, diff --git a/library/test/src/term/terminfo/parm.rs b/library/test/src/term/terminfo/parm.rs index 0756c8374..2815f6cfc 100644 --- a/library/test/src/term/terminfo/parm.rs +++ b/library/test/src/term/terminfo/parm.rs @@ -282,14 +282,14 @@ pub(crate) fn expand( ); } SetVar => { - if cur >= 'A' && cur <= 'Z' { + if cur.is_ascii_uppercase() { if let Some(arg) = stack.pop() { let idx = (cur as u8) - b'A'; vars.sta_va[idx as usize] = arg; } else { return Err("stack is empty".to_string()); } - } else if cur >= 'a' && cur <= 'z' { + } else if cur.is_ascii_lowercase() { if let Some(arg) = stack.pop() { let idx = (cur as u8) - b'a'; vars.dyn_va[idx as usize] = arg; @@ -301,10 +301,10 @@ pub(crate) fn expand( } } GetVar => { - if cur >= 'A' && cur <= 'Z' { + if cur.is_ascii_uppercase() { let idx = (cur as u8) - b'A'; stack.push(vars.sta_va[idx as usize].clone()); - } else if cur >= 'a' && cur <= 'z' { + } else if cur.is_ascii_lowercase() { let idx = (cur as u8) - b'a'; stack.push(vars.dyn_va[idx as usize].clone()); } else { diff --git a/library/test/src/term/terminfo/searcher.rs b/library/test/src/term/terminfo/searcher.rs index 68e181a68..3e8ccc91a 100644 --- a/library/test/src/term/terminfo/searcher.rs +++ b/library/test/src/term/terminfo/searcher.rs @@ -22,7 +22,7 @@ pub(crate) fn get_dbpath_for_term(term: &str) -> Option { if let Ok(dirs) = env::var("TERMINFO_DIRS") { for i in dirs.split(':') { - if i == "" { + if i.is_empty() { dirs_to_search.push(PathBuf::from("/usr/share/terminfo")); } else { dirs_to_search.push(PathBuf::from(i)); @@ -30,7 +30,7 @@ pub(crate) fn get_dbpath_for_term(term: &str) -> Option { } } else { // Found nothing in TERMINFO_DIRS, use the default paths: - // According to /etc/terminfo/README, after looking at + // According to /etc/terminfo/README, after looking at // ~/.terminfo, ncurses will search /etc/terminfo, then // /lib/terminfo, and eventually /usr/share/terminfo. // On Haiku the database can be found at /boot/system/data/terminfo @@ -49,7 +49,7 @@ pub(crate) fn get_dbpath_for_term(term: &str) -> Option { for mut p in dirs_to_search { if fs::metadata(&p).is_ok() { p.push(&first_char.to_string()); - p.push(&term); + p.push(term); if fs::metadata(&p).is_ok() { return Some(p); } diff --git a/library/test/src/test_result.rs b/library/test/src/test_result.rs index 7f44d6e3d..1da238e3e 100644 --- a/library/test/src/test_result.rs +++ b/library/test/src/test_result.rs @@ -33,7 +33,7 @@ pub fn calc_result<'a>( ) -> TestResult { let result = match (&desc.should_panic, task_result) { (&ShouldPanic::No, Ok(())) | (&ShouldPanic::Yes, Err(_)) => TestResult::TrOk, - (&ShouldPanic::YesWithMessage(msg), Err(ref err)) => { + (&ShouldPanic::YesWithMessage(msg), Err(err)) => { let maybe_panic_str = err .downcast_ref::() .map(|e| &**e) @@ -44,16 +44,15 @@ pub fn calc_result<'a>( } else if let Some(panic_str) = maybe_panic_str { TestResult::TrFailedMsg(format!( r#"panic did not contain expected string - panic message: `{:?}`, - expected substring: `{:?}`"#, - panic_str, msg + panic message: `{panic_str:?}`, + expected substring: `{msg:?}`"# )) } else { TestResult::TrFailedMsg(format!( r#"expected panic with string value, found non-string value: `{:?}` expected substring: `{:?}`"#, - (**err).type_id(), + (*err).type_id(), msg )) } diff --git a/library/test/src/time.rs b/library/test/src/time.rs index 8c64e5d1b..7fd69d7f7 100644 --- a/library/test/src/time.rs +++ b/library/test/src/time.rs @@ -107,16 +107,14 @@ impl TimeThreshold { let durations_str = env::var(env_var_name).ok()?; let (warn_str, critical_str) = durations_str.split_once(',').unwrap_or_else(|| { panic!( - "Duration variable {} expected to have 2 numbers separated by comma, but got {}", - env_var_name, durations_str + "Duration variable {env_var_name} expected to have 2 numbers separated by comma, but got {durations_str}" ) }); let parse_u64 = |v| { u64::from_str(v).unwrap_or_else(|_| { panic!( - "Duration value in variable {} is expected to be a number, but got {}", - env_var_name, v + "Duration value in variable {env_var_name} is expected to be a number, but got {v}" ) }) }; diff --git a/library/test/src/types.rs b/library/test/src/types.rs index 888afff79..6f2e03309 100644 --- a/library/test/src/types.rs +++ b/library/test/src/types.rs @@ -47,7 +47,7 @@ impl TestName { match *self { StaticTestName(s) => s, DynTestName(ref s) => s, - AlignedTestName(ref s, _) => &*s, + AlignedTestName(ref s, _) => s, } } diff --git a/library/unwind/src/libunwind.rs b/library/unwind/src/libunwind.rs index 0fa11f376..036a35869 100644 --- a/library/unwind/src/libunwind.rs +++ b/library/unwind/src/libunwind.rs @@ -88,7 +88,7 @@ pub type _Unwind_Exception_Cleanup_Fn = extern "C" fn(unwind_code: _Unwind_Reason_Code, exception: *mut _Unwind_Exception); // FIXME: The `#[link]` attributes on `extern "C"` block marks those symbols declared in -// the block are reexported in dylib build of libstd. This is needed when build rustc with +// the block are reexported in dylib build of std. This is needed when build rustc with // feature `llvm-libunwind', as no other cdylib will provided those _Unwind_* symbols. // However the `link` attribute is duplicated multiple times and does not just export symbol, // a better way to manually export symbol would be another attribute like `#[export]`. -- cgit v1.2.3