summaryrefslogtreecommitdiffstats
path: root/js/src/vm/BytecodeUtil.cpp
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-28 14:29:10 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-28 14:29:10 +0000
commit2aa4a82499d4becd2284cdb482213d541b8804dd (patch)
treeb80bf8bf13c3766139fbacc530efd0dd9d54394c /js/src/vm/BytecodeUtil.cpp
parentInitial commit. (diff)
downloadfirefox-2aa4a82499d4becd2284cdb482213d541b8804dd.tar.xz
firefox-2aa4a82499d4becd2284cdb482213d541b8804dd.zip
Adding upstream version 86.0.1.upstream/86.0.1upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'js/src/vm/BytecodeUtil.cpp')
-rw-r--r--js/src/vm/BytecodeUtil.cpp3036
1 files changed, 3036 insertions, 0 deletions
diff --git a/js/src/vm/BytecodeUtil.cpp b/js/src/vm/BytecodeUtil.cpp
new file mode 100644
index 0000000000..cb753a00ec
--- /dev/null
+++ b/js/src/vm/BytecodeUtil.cpp
@@ -0,0 +1,3036 @@
+/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*-
+ * vim: set ts=8 sts=2 et sw=2 tw=80:
+ * This Source Code Form is subject to the terms of the Mozilla Public
+ * License, v. 2.0. If a copy of the MPL was not distributed with this
+ * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
+
+/*
+ * JS bytecode descriptors, disassemblers, and (expression) decompilers.
+ */
+
+#include "vm/BytecodeUtil-inl.h"
+
+#define __STDC_FORMAT_MACROS
+
+#include "mozilla/Attributes.h"
+#include "mozilla/Maybe.h"
+#include "mozilla/ReverseIterator.h"
+#include "mozilla/Sprintf.h"
+#include "mozilla/Vector.h"
+
+#include <algorithm>
+#include <inttypes.h>
+#include <stdio.h>
+#include <string.h>
+#include <type_traits>
+
+#include "jsapi.h"
+#include "jsnum.h"
+#include "jstypes.h"
+
+#include "frontend/BytecodeCompiler.h"
+#include "frontend/SourceNotes.h" // SrcNote, SrcNoteType, SrcNoteIterator
+#include "gc/PublicIterators.h"
+#include "jit/IonScript.h" // IonBlockCounts
+#include "js/CharacterEncoding.h"
+#include "js/experimental/CodeCoverage.h"
+#include "js/friend/DumpFunctions.h" // js::DumpPC, js::DumpScript
+#include "js/friend/ErrorMessages.h" // js::GetErrorMessage, JSMSG_*
+#include "js/Printf.h"
+#include "js/Symbol.h"
+#include "util/DifferentialTesting.h"
+#include "util/Memory.h"
+#include "util/StringBuffer.h"
+#include "util/Text.h"
+#include "vm/BuiltinObjectKind.h"
+#include "vm/BytecodeIterator.h" // for AllBytecodesIterable
+#include "vm/BytecodeLocation.h"
+#include "vm/CodeCoverage.h"
+#include "vm/EnvironmentObject.h"
+#include "vm/FrameIter.h" // js::{,Script}FrameIter
+#include "vm/JSAtom.h"
+#include "vm/JSContext.h"
+#include "vm/JSFunction.h"
+#include "vm/JSObject.h"
+#include "vm/JSScript.h"
+#include "vm/Opcodes.h"
+#include "vm/Printer.h"
+#include "vm/Realm.h"
+#include "vm/Shape.h"
+#include "vm/ToSource.h" // js::ValueToSource
+
+#include "gc/GC-inl.h"
+#include "vm/BytecodeIterator-inl.h"
+#include "vm/BytecodeLocation-inl.h"
+#include "vm/JSContext-inl.h"
+#include "vm/JSObject-inl.h"
+#include "vm/JSScript-inl.h"
+#include "vm/Realm-inl.h"
+
+using namespace js;
+
+using js::frontend::IsIdentifier;
+
+/*
+ * Index limit must stay within 32 bits.
+ */
+static_assert(sizeof(uint32_t) * CHAR_BIT >= INDEX_LIMIT_LOG2 + 1);
+
+const JSCodeSpec js::CodeSpecTable[] = {
+#define MAKE_CODESPEC(op, op_snake, token, length, nuses, ndefs, format) \
+ {length, nuses, ndefs, format},
+ FOR_EACH_OPCODE(MAKE_CODESPEC)
+#undef MAKE_CODESPEC
+};
+
+/*
+ * Each element of the array is either a source literal associated with JS
+ * bytecode or null.
+ */
+static const char* const CodeToken[] = {
+#define TOKEN(op, op_snake, token, ...) token,
+ FOR_EACH_OPCODE(TOKEN)
+#undef TOKEN
+};
+
+/*
+ * Array of JS bytecode names used by PC count JSON, DEBUG-only Disassemble
+ * and JIT debug spew.
+ */
+const char* const js::CodeNameTable[] = {
+#define OPNAME(op, ...) #op,
+ FOR_EACH_OPCODE(OPNAME)
+#undef OPNAME
+};
+
+/************************************************************************/
+
+static bool DecompileArgumentFromStack(JSContext* cx, int formalIndex,
+ UniqueChars* res);
+
+/* static */ const char PCCounts::numExecName[] = "interp";
+
+static MOZ_MUST_USE bool DumpIonScriptCounts(Sprinter* sp, HandleScript script,
+ jit::IonScriptCounts* ionCounts) {
+ if (!sp->jsprintf("IonScript [%zu blocks]:\n", ionCounts->numBlocks())) {
+ return false;
+ }
+
+ for (size_t i = 0; i < ionCounts->numBlocks(); i++) {
+ const jit::IonBlockCounts& block = ionCounts->block(i);
+ unsigned lineNumber = 0, columnNumber = 0;
+ lineNumber = PCToLineNumber(script, script->offsetToPC(block.offset()),
+ &columnNumber);
+ if (!sp->jsprintf("BB #%" PRIu32 " [%05u,%u,%u]", block.id(),
+ block.offset(), lineNumber, columnNumber)) {
+ return false;
+ }
+ if (block.description()) {
+ if (!sp->jsprintf(" [inlined %s]", block.description())) {
+ return false;
+ }
+ }
+ for (size_t j = 0; j < block.numSuccessors(); j++) {
+ if (!sp->jsprintf(" -> #%" PRIu32, block.successor(j))) {
+ return false;
+ }
+ }
+ if (!sp->jsprintf(" :: %" PRIu64 " hits\n", block.hitCount())) {
+ return false;
+ }
+ if (!sp->jsprintf("%s\n", block.code())) {
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static MOZ_MUST_USE bool DumpPCCounts(JSContext* cx, HandleScript script,
+ Sprinter* sp) {
+ MOZ_ASSERT(script->hasScriptCounts());
+
+ // Ensure the Disassemble1 call below does not discard the script counts.
+ gc::AutoSuppressGC suppress(cx);
+
+#ifdef DEBUG
+ jsbytecode* pc = script->code();
+ while (pc < script->codeEnd()) {
+ jsbytecode* next = GetNextPc(pc);
+
+ if (!Disassemble1(cx, script, pc, script->pcToOffset(pc), true, sp)) {
+ return false;
+ }
+
+ if (!sp->put(" {")) {
+ return false;
+ }
+
+ PCCounts* counts = script->maybeGetPCCounts(pc);
+ if (double val = counts ? counts->numExec() : 0.0) {
+ if (!sp->jsprintf("\"%s\": %.0f", PCCounts::numExecName, val)) {
+ return false;
+ }
+ }
+ if (!sp->put("}\n")) {
+ return false;
+ }
+
+ pc = next;
+ }
+#endif
+
+ jit::IonScriptCounts* ionCounts = script->getIonCounts();
+ while (ionCounts) {
+ if (!DumpIonScriptCounts(sp, script, ionCounts)) {
+ return false;
+ }
+
+ ionCounts = ionCounts->previous();
+ }
+
+ return true;
+}
+
+bool js::DumpRealmPCCounts(JSContext* cx) {
+ Rooted<GCVector<JSScript*>> scripts(cx, GCVector<JSScript*>(cx));
+ for (auto base = cx->zone()->cellIter<BaseScript>(); !base.done();
+ base.next()) {
+ if (base->realm() != cx->realm()) {
+ continue;
+ }
+ MOZ_ASSERT_IF(base->hasScriptCounts(), base->hasBytecode());
+ if (base->hasScriptCounts()) {
+ if (!scripts.append(base->asJSScript())) {
+ return false;
+ }
+ }
+ }
+
+ for (uint32_t i = 0; i < scripts.length(); i++) {
+ HandleScript script = scripts[i];
+ Sprinter sprinter(cx);
+ if (!sprinter.init()) {
+ return false;
+ }
+
+ fprintf(stdout, "--- SCRIPT %s:%u ---\n", script->filename(),
+ script->lineno());
+ if (!DumpPCCounts(cx, script, &sprinter)) {
+ return false;
+ }
+ fputs(sprinter.string(), stdout);
+ fprintf(stdout, "--- END SCRIPT %s:%u ---\n", script->filename(),
+ script->lineno());
+ }
+
+ return true;
+}
+
+/////////////////////////////////////////////////////////////////////
+// Bytecode Parser
+/////////////////////////////////////////////////////////////////////
+
+// Stores the information about the stack slot, where the value comes from.
+// Elements of BytecodeParser::Bytecode.{offsetStack,offsetStackAfter} arrays.
+class OffsetAndDefIndex {
+ // The offset of the PC that pushed the value for this slot.
+ uint32_t offset_;
+
+ // The index in `ndefs` for the PC (0-origin)
+ uint8_t defIndex_;
+
+ enum : uint8_t {
+ Normal = 0,
+
+ // Ignored this value in the expression decompilation.
+ // Used by JSOp::NopDestructuring. See BytecodeParser::simulateOp.
+ Ignored,
+
+ // The value in this slot comes from 2 or more paths.
+ // offset_ and defIndex_ holds the information for the path that
+ // reaches here first.
+ Merged,
+ } type_;
+
+ public:
+ uint32_t offset() const {
+ MOZ_ASSERT(!isSpecial());
+ return offset_;
+ };
+ uint32_t specialOffset() const {
+ MOZ_ASSERT(isSpecial());
+ return offset_;
+ };
+
+ uint8_t defIndex() const {
+ MOZ_ASSERT(!isSpecial());
+ return defIndex_;
+ }
+ uint8_t specialDefIndex() const {
+ MOZ_ASSERT(isSpecial());
+ return defIndex_;
+ }
+
+ bool isSpecial() const { return type_ != Normal; }
+ bool isMerged() const { return type_ == Merged; }
+ bool isIgnored() const { return type_ == Ignored; }
+
+ void set(uint32_t aOffset, uint8_t aDefIndex) {
+ offset_ = aOffset;
+ defIndex_ = aDefIndex;
+ type_ = Normal;
+ }
+
+ // Keep offset_ and defIndex_ values for stack dump.
+ void setMerged() { type_ = Merged; }
+ void setIgnored() { type_ = Ignored; }
+
+ bool operator==(const OffsetAndDefIndex& rhs) const {
+ return offset_ == rhs.offset_ && defIndex_ == rhs.defIndex_;
+ }
+
+ bool operator!=(const OffsetAndDefIndex& rhs) const {
+ return !(*this == rhs);
+ }
+};
+
+namespace {
+
+class BytecodeParser {
+ public:
+ enum class JumpKind {
+ Simple,
+ SwitchCase,
+ SwitchDefault,
+ TryCatch,
+ TryFinally
+ };
+
+ private:
+ class Bytecode {
+ public:
+ explicit Bytecode(const LifoAllocPolicy<Fallible>& alloc)
+ : parsed(false),
+ stackDepth(0),
+ offsetStack(nullptr)
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ ,
+ stackDepthAfter(0),
+ offsetStackAfter(nullptr),
+ jumpOrigins(alloc)
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+ {
+ }
+
+ // Whether this instruction has been analyzed to get its output defines
+ // and stack.
+ bool parsed;
+
+ // Stack depth before this opcode.
+ uint32_t stackDepth;
+
+ // Pointer to array of |stackDepth| offsets. An element at position N
+ // in the array is the offset of the opcode that defined the
+ // corresponding stack slot. The top of the stack is at position
+ // |stackDepth - 1|.
+ OffsetAndDefIndex* offsetStack;
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ // stack depth after this opcode.
+ uint32_t stackDepthAfter;
+
+ // Pointer to array of |stackDepthAfter| offsets.
+ OffsetAndDefIndex* offsetStackAfter;
+
+ struct JumpInfo {
+ uint32_t from;
+ JumpKind kind;
+
+ JumpInfo(uint32_t from_, JumpKind kind_) : from(from_), kind(kind_) {}
+ };
+
+ // A list of offsets of the bytecode that jumps to this bytecode,
+ // exclusing previous bytecode.
+ Vector<JumpInfo, 0, LifoAllocPolicy<Fallible>> jumpOrigins;
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+
+ bool captureOffsetStack(LifoAlloc& alloc, const OffsetAndDefIndex* stack,
+ uint32_t depth) {
+ stackDepth = depth;
+ if (stackDepth) {
+ offsetStack = alloc.newArray<OffsetAndDefIndex>(stackDepth);
+ if (!offsetStack) {
+ return false;
+ }
+ for (uint32_t n = 0; n < stackDepth; n++) {
+ offsetStack[n] = stack[n];
+ }
+ }
+ return true;
+ }
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ bool captureOffsetStackAfter(LifoAlloc& alloc,
+ const OffsetAndDefIndex* stack,
+ uint32_t depth) {
+ stackDepthAfter = depth;
+ if (stackDepthAfter) {
+ offsetStackAfter = alloc.newArray<OffsetAndDefIndex>(stackDepthAfter);
+ if (!offsetStackAfter) {
+ return false;
+ }
+ for (uint32_t n = 0; n < stackDepthAfter; n++) {
+ offsetStackAfter[n] = stack[n];
+ }
+ }
+ return true;
+ }
+
+ bool addJump(uint32_t from, JumpKind kind) {
+ return jumpOrigins.append(JumpInfo(from, kind));
+ }
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+
+ // When control-flow merges, intersect the stacks, marking slots that
+ // are defined by different offsets and/or defIndices merged.
+ // This is sufficient for forward control-flow. It doesn't grok loops
+ // -- for that you would have to iterate to a fixed point -- but there
+ // shouldn't be operands on the stack at a loop back-edge anyway.
+ void mergeOffsetStack(const OffsetAndDefIndex* stack, uint32_t depth) {
+ MOZ_ASSERT(depth == stackDepth);
+ for (uint32_t n = 0; n < stackDepth; n++) {
+ if (stack[n].isIgnored()) {
+ continue;
+ }
+ if (offsetStack[n].isIgnored()) {
+ offsetStack[n] = stack[n];
+ }
+ if (offsetStack[n] != stack[n]) {
+ offsetStack[n].setMerged();
+ }
+ }
+ }
+ };
+
+ JSContext* cx_;
+ LifoAlloc& alloc_;
+ RootedScript script_;
+
+ Bytecode** codeArray_;
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ // Dedicated mode for stack dump.
+ // Capture stack after each opcode, and also enable special handling for
+ // some opcodes to make stack transition clearer.
+ bool isStackDump;
+#endif
+
+ public:
+ BytecodeParser(JSContext* cx, LifoAlloc& alloc, JSScript* script)
+ : cx_(cx),
+ alloc_(alloc),
+ script_(cx, script),
+ codeArray_(nullptr)
+#ifdef DEBUG
+ ,
+ isStackDump(false)
+#endif
+ {
+ }
+
+ bool parse();
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ bool isReachable(const jsbytecode* pc) const { return maybeCode(pc); }
+#endif
+
+ uint32_t stackDepthAtPC(uint32_t offset) const {
+ // Sometimes the code generator in debug mode asks about the stack depth
+ // of unreachable code (bug 932180 comment 22). Assume that unreachable
+ // code has no operands on the stack.
+ return getCode(offset).stackDepth;
+ }
+ uint32_t stackDepthAtPC(const jsbytecode* pc) const {
+ return stackDepthAtPC(script_->pcToOffset(pc));
+ }
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ uint32_t stackDepthAfterPC(uint32_t offset) const {
+ return getCode(offset).stackDepthAfter;
+ }
+ uint32_t stackDepthAfterPC(const jsbytecode* pc) const {
+ return stackDepthAfterPC(script_->pcToOffset(pc));
+ }
+#endif
+
+ const OffsetAndDefIndex& offsetForStackOperand(uint32_t offset,
+ int operand) const {
+ Bytecode& code = getCode(offset);
+ if (operand < 0) {
+ operand += code.stackDepth;
+ MOZ_ASSERT(operand >= 0);
+ }
+ MOZ_ASSERT(uint32_t(operand) < code.stackDepth);
+ return code.offsetStack[operand];
+ }
+ jsbytecode* pcForStackOperand(jsbytecode* pc, int operand,
+ uint8_t* defIndex) const {
+ size_t offset = script_->pcToOffset(pc);
+ const OffsetAndDefIndex& offsetAndDefIndex =
+ offsetForStackOperand(offset, operand);
+ if (offsetAndDefIndex.isSpecial()) {
+ return nullptr;
+ }
+ *defIndex = offsetAndDefIndex.defIndex();
+ return script_->offsetToPC(offsetAndDefIndex.offset());
+ }
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ const OffsetAndDefIndex& offsetForStackOperandAfterPC(uint32_t offset,
+ int operand) const {
+ Bytecode& code = getCode(offset);
+ if (operand < 0) {
+ operand += code.stackDepthAfter;
+ MOZ_ASSERT(operand >= 0);
+ }
+ MOZ_ASSERT(uint32_t(operand) < code.stackDepthAfter);
+ return code.offsetStackAfter[operand];
+ }
+
+ template <typename Callback>
+ bool forEachJumpOrigins(jsbytecode* pc, Callback callback) const {
+ Bytecode& code = getCode(script_->pcToOffset(pc));
+
+ for (Bytecode::JumpInfo& info : code.jumpOrigins) {
+ if (!callback(script_->offsetToPC(info.from), info.kind)) {
+ return false;
+ }
+ }
+
+ return true;
+ }
+
+ void setStackDump() { isStackDump = true; }
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+
+ private:
+ LifoAlloc& alloc() { return alloc_; }
+
+ void reportOOM() { ReportOutOfMemory(cx_); }
+
+ uint32_t maximumStackDepth() const {
+ return script_->nslots() - script_->nfixed();
+ }
+
+ Bytecode& getCode(uint32_t offset) const {
+ MOZ_ASSERT(offset < script_->length());
+ MOZ_ASSERT(codeArray_[offset]);
+ return *codeArray_[offset];
+ }
+
+ Bytecode* maybeCode(uint32_t offset) const {
+ MOZ_ASSERT(offset < script_->length());
+ return codeArray_[offset];
+ }
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ Bytecode* maybeCode(const jsbytecode* pc) const {
+ return maybeCode(script_->pcToOffset(pc));
+ }
+#endif
+
+ uint32_t simulateOp(JSOp op, uint32_t offset, OffsetAndDefIndex* offsetStack,
+ uint32_t stackDepth);
+
+ inline bool recordBytecode(uint32_t offset,
+ const OffsetAndDefIndex* offsetStack,
+ uint32_t stackDepth);
+
+ inline bool addJump(uint32_t offset, uint32_t stackDepth,
+ const OffsetAndDefIndex* offsetStack, jsbytecode* pc,
+ JumpKind kind);
+};
+
+} // anonymous namespace
+
+uint32_t BytecodeParser::simulateOp(JSOp op, uint32_t offset,
+ OffsetAndDefIndex* offsetStack,
+ uint32_t stackDepth) {
+ jsbytecode* pc = script_->offsetToPC(offset);
+ uint32_t nuses = GetUseCount(pc);
+ uint32_t ndefs = GetDefCount(pc);
+
+ MOZ_ASSERT(stackDepth >= nuses);
+ stackDepth -= nuses;
+ MOZ_ASSERT(stackDepth + ndefs <= maximumStackDepth());
+
+#ifdef DEBUG
+ if (isStackDump) {
+ // Opcodes that modifies the object but keeps it on the stack while
+ // initialization should be listed here instead of switch below.
+ // For error message, they shouldn't be shown as the original object
+ // after adding properties.
+ // For stack dump, keeping the input is better.
+ switch (op) {
+ case JSOp::InitHiddenProp:
+ case JSOp::InitHiddenPropGetter:
+ case JSOp::InitHiddenPropSetter:
+ case JSOp::InitLockedProp:
+ case JSOp::InitProp:
+ case JSOp::InitPropGetter:
+ case JSOp::InitPropSetter:
+ case JSOp::SetFunName:
+ // Keep the second value.
+ MOZ_ASSERT(nuses == 2);
+ MOZ_ASSERT(ndefs == 1);
+ goto end;
+
+ case JSOp::InitElem:
+ case JSOp::InitElemGetter:
+ case JSOp::InitElemSetter:
+ case JSOp::InitHiddenElem:
+ case JSOp::InitHiddenElemGetter:
+ case JSOp::InitHiddenElemSetter:
+ case JSOp::InitLockedElem:
+ // Keep the third value.
+ MOZ_ASSERT(nuses == 3);
+ MOZ_ASSERT(ndefs == 1);
+ goto end;
+
+ default:
+ break;
+ }
+ }
+#endif /* DEBUG */
+
+ // Mark the current offset as defining its values on the offset stack,
+ // unless it just reshuffles the stack. In that case we want to preserve
+ // the opcode that generated the original value.
+ switch (op) {
+ default:
+ for (uint32_t n = 0; n != ndefs; ++n) {
+ offsetStack[stackDepth + n].set(offset, n);
+ }
+ break;
+
+ case JSOp::NopDestructuring:
+ // Poison the last offset to not obfuscate the error message.
+ offsetStack[stackDepth - 1].setIgnored();
+ break;
+
+ case JSOp::Case:
+ // Keep the switch value.
+ MOZ_ASSERT(ndefs == 1);
+ break;
+
+ case JSOp::Dup:
+ MOZ_ASSERT(ndefs == 2);
+ offsetStack[stackDepth + 1] = offsetStack[stackDepth];
+ break;
+
+ case JSOp::Dup2:
+ MOZ_ASSERT(ndefs == 4);
+ offsetStack[stackDepth + 2] = offsetStack[stackDepth];
+ offsetStack[stackDepth + 3] = offsetStack[stackDepth + 1];
+ break;
+
+ case JSOp::DupAt: {
+ MOZ_ASSERT(ndefs == 1);
+ unsigned n = GET_UINT24(pc);
+ MOZ_ASSERT(n < stackDepth);
+ offsetStack[stackDepth] = offsetStack[stackDepth - 1 - n];
+ break;
+ }
+
+ case JSOp::Swap: {
+ MOZ_ASSERT(ndefs == 2);
+ OffsetAndDefIndex tmp = offsetStack[stackDepth + 1];
+ offsetStack[stackDepth + 1] = offsetStack[stackDepth];
+ offsetStack[stackDepth] = tmp;
+ break;
+ }
+
+ case JSOp::Pick: {
+ unsigned n = GET_UINT8(pc);
+ MOZ_ASSERT(ndefs == n + 1);
+ uint32_t top = stackDepth + n;
+ OffsetAndDefIndex tmp = offsetStack[stackDepth];
+ for (uint32_t i = stackDepth; i < top; i++) {
+ offsetStack[i] = offsetStack[i + 1];
+ }
+ offsetStack[top] = tmp;
+ break;
+ }
+
+ case JSOp::Unpick: {
+ unsigned n = GET_UINT8(pc);
+ MOZ_ASSERT(ndefs == n + 1);
+ uint32_t top = stackDepth + n;
+ OffsetAndDefIndex tmp = offsetStack[top];
+ for (uint32_t i = top; i > stackDepth; i--) {
+ offsetStack[i] = offsetStack[i - 1];
+ }
+ offsetStack[stackDepth] = tmp;
+ break;
+ }
+
+ case JSOp::And:
+ case JSOp::CheckIsObj:
+ case JSOp::CheckObjCoercible:
+ case JSOp::CheckThis:
+ case JSOp::CheckThisReinit:
+ case JSOp::CheckClassHeritage:
+ case JSOp::DebugCheckSelfHosted:
+ case JSOp::InitGLexical:
+ case JSOp::InitLexical:
+ case JSOp::Or:
+ case JSOp::Coalesce:
+ case JSOp::SetAliasedVar:
+ case JSOp::SetArg:
+ case JSOp::SetIntrinsic:
+ case JSOp::SetLocal:
+ case JSOp::InitAliasedLexical:
+ case JSOp::CheckLexical:
+ case JSOp::CheckAliasedLexical:
+ // Keep the top value.
+ MOZ_ASSERT(nuses == 1);
+ MOZ_ASSERT(ndefs == 1);
+ break;
+
+ case JSOp::InitHomeObject:
+ // Pop the top value, keep the other value.
+ MOZ_ASSERT(nuses == 2);
+ MOZ_ASSERT(ndefs == 1);
+ break;
+
+ case JSOp::CheckResumeKind:
+ // Pop the top two values, keep the other value.
+ MOZ_ASSERT(nuses == 3);
+ MOZ_ASSERT(ndefs == 1);
+ break;
+
+ case JSOp::SetGName:
+ case JSOp::SetName:
+ case JSOp::SetProp:
+ case JSOp::StrictSetGName:
+ case JSOp::StrictSetName:
+ case JSOp::StrictSetProp:
+ // Keep the top value, removing other 1 value.
+ MOZ_ASSERT(nuses == 2);
+ MOZ_ASSERT(ndefs == 1);
+ offsetStack[stackDepth] = offsetStack[stackDepth + 1];
+ break;
+
+ case JSOp::SetPropSuper:
+ case JSOp::StrictSetPropSuper:
+ // Keep the top value, removing other 2 values.
+ MOZ_ASSERT(nuses == 3);
+ MOZ_ASSERT(ndefs == 1);
+ offsetStack[stackDepth] = offsetStack[stackDepth + 2];
+ break;
+
+ case JSOp::SetElemSuper:
+ case JSOp::StrictSetElemSuper:
+ // Keep the top value, removing other 3 values.
+ MOZ_ASSERT(nuses == 4);
+ MOZ_ASSERT(ndefs == 1);
+ offsetStack[stackDepth] = offsetStack[stackDepth + 3];
+ break;
+
+ case JSOp::IsGenClosing:
+ case JSOp::IsNoIter:
+ case JSOp::MoreIter:
+ case JSOp::OptimizeSpreadCall:
+ // Keep the top value and push one more value.
+ MOZ_ASSERT(nuses == 1);
+ MOZ_ASSERT(ndefs == 2);
+ offsetStack[stackDepth + 1].set(offset, 1);
+ break;
+
+ case JSOp::CheckPrivateField:
+ // Keep the top two values, and push one new value.
+ MOZ_ASSERT(nuses == 2);
+ MOZ_ASSERT(ndefs == 3);
+ offsetStack[stackDepth + 2].set(offset, 2);
+ break;
+ }
+
+#ifdef DEBUG
+end:
+#endif /* DEBUG */
+
+ stackDepth += ndefs;
+ return stackDepth;
+}
+
+bool BytecodeParser::recordBytecode(uint32_t offset,
+ const OffsetAndDefIndex* offsetStack,
+ uint32_t stackDepth) {
+ MOZ_ASSERT(offset < script_->length());
+
+ Bytecode*& code = codeArray_[offset];
+ if (!code) {
+ code = alloc().new_<Bytecode>(alloc());
+ if (!code || !code->captureOffsetStack(alloc(), offsetStack, stackDepth)) {
+ reportOOM();
+ return false;
+ }
+ } else {
+ code->mergeOffsetStack(offsetStack, stackDepth);
+ }
+
+ return true;
+}
+
+bool BytecodeParser::addJump(uint32_t offset, uint32_t stackDepth,
+ const OffsetAndDefIndex* offsetStack,
+ jsbytecode* pc, JumpKind kind) {
+ if (!recordBytecode(offset, offsetStack, stackDepth)) {
+ return false;
+ }
+
+#ifdef DEBUG
+ uint32_t currentOffset = script_->pcToOffset(pc);
+ if (isStackDump) {
+ if (!codeArray_[offset]->addJump(currentOffset, kind)) {
+ reportOOM();
+ return false;
+ }
+ }
+
+ // If this is a backedge, assert we parsed the target JSOp::LoopHead.
+ MOZ_ASSERT_IF(offset < currentOffset, codeArray_[offset]->parsed);
+#endif /* DEBUG */
+
+ return true;
+}
+
+bool BytecodeParser::parse() {
+ MOZ_ASSERT(!codeArray_);
+
+ uint32_t length = script_->length();
+ codeArray_ = alloc().newArray<Bytecode*>(length);
+
+ if (!codeArray_) {
+ reportOOM();
+ return false;
+ }
+
+ mozilla::PodZero(codeArray_, length);
+
+ // Fill in stack depth and definitions at initial bytecode.
+ Bytecode* startcode = alloc().new_<Bytecode>(alloc());
+ if (!startcode) {
+ reportOOM();
+ return false;
+ }
+
+ // Fill in stack depth and definitions at initial bytecode.
+ OffsetAndDefIndex* offsetStack =
+ alloc().newArray<OffsetAndDefIndex>(maximumStackDepth());
+ if (maximumStackDepth() && !offsetStack) {
+ reportOOM();
+ return false;
+ }
+
+ startcode->stackDepth = 0;
+ codeArray_[0] = startcode;
+
+ for (uint32_t offset = 0, nextOffset = 0; offset < length;
+ offset = nextOffset) {
+ Bytecode* code = maybeCode(offset);
+ jsbytecode* pc = script_->offsetToPC(offset);
+
+ // Next bytecode to analyze.
+ nextOffset = offset + GetBytecodeLength(pc);
+
+ MOZ_ASSERT(*pc < JSOP_LIMIT);
+ JSOp op = JSOp(*pc);
+
+ if (!code) {
+ // Haven't found a path by which this bytecode is reachable.
+ continue;
+ }
+
+ // On a jump target, we reload the offsetStack saved for the current
+ // bytecode, as it contains either the original offset stack, or the
+ // merged offset stack.
+ if (BytecodeIsJumpTarget(op)) {
+ for (uint32_t n = 0; n < code->stackDepth; ++n) {
+ offsetStack[n] = code->offsetStack[n];
+ }
+ }
+
+ if (code->parsed) {
+ // No need to reparse.
+ continue;
+ }
+
+ code->parsed = true;
+
+ uint32_t stackDepth = simulateOp(op, offset, offsetStack, code->stackDepth);
+
+#ifdef DEBUG
+ if (isStackDump) {
+ if (!code->captureOffsetStackAfter(alloc(), offsetStack, stackDepth)) {
+ reportOOM();
+ return false;
+ }
+ }
+#endif /* DEBUG */
+
+ switch (op) {
+ case JSOp::TableSwitch: {
+ uint32_t defaultOffset = offset + GET_JUMP_OFFSET(pc);
+ jsbytecode* pc2 = pc + JUMP_OFFSET_LEN;
+ int32_t low = GET_JUMP_OFFSET(pc2);
+ pc2 += JUMP_OFFSET_LEN;
+ int32_t high = GET_JUMP_OFFSET(pc2);
+ pc2 += JUMP_OFFSET_LEN;
+
+ if (!addJump(defaultOffset, stackDepth, offsetStack, pc,
+ JumpKind::SwitchDefault)) {
+ return false;
+ }
+
+ uint32_t ncases = high - low + 1;
+
+ for (uint32_t i = 0; i < ncases; i++) {
+ uint32_t targetOffset = script_->tableSwitchCaseOffset(pc, i);
+ if (targetOffset != defaultOffset) {
+ if (!addJump(targetOffset, stackDepth, offsetStack, pc,
+ JumpKind::SwitchCase)) {
+ return false;
+ }
+ }
+ }
+ break;
+ }
+
+ case JSOp::Try: {
+ // Everything between a try and corresponding catch or finally is
+ // conditional. Note that there is no problem with code which is skipped
+ // by a thrown exception but is not caught by a later handler in the
+ // same function: no more code will execute, and it does not matter what
+ // is defined.
+ for (const TryNote& tn : script_->trynotes()) {
+ if (tn.start == offset + JSOpLength_Try) {
+ uint32_t catchOffset = tn.start + tn.length;
+ if (tn.kind() == TryNoteKind::Catch) {
+ if (!addJump(catchOffset, stackDepth, offsetStack, pc,
+ JumpKind::TryCatch)) {
+ return false;
+ }
+ } else if (tn.kind() == TryNoteKind::Finally) {
+ if (!addJump(catchOffset, stackDepth, offsetStack, pc,
+ JumpKind::TryFinally)) {
+ return false;
+ }
+ }
+ }
+ }
+ break;
+ }
+
+ default:
+ break;
+ }
+
+ // Check basic jump opcodes, which may or may not have a fallthrough.
+ if (IsJumpOpcode(op)) {
+ // Case instructions do not push the lvalue back when branching.
+ uint32_t newStackDepth = stackDepth;
+ if (op == JSOp::Case) {
+ newStackDepth--;
+ }
+
+ uint32_t targetOffset = offset + GET_JUMP_OFFSET(pc);
+ if (!addJump(targetOffset, newStackDepth, offsetStack, pc,
+ JumpKind::Simple)) {
+ return false;
+ }
+ }
+
+ // Handle any fallthrough from this opcode.
+ if (BytecodeFallsThrough(op)) {
+ if (!recordBytecode(nextOffset, offsetStack, stackDepth)) {
+ return false;
+ }
+ }
+ }
+
+ return true;
+}
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+
+bool js::ReconstructStackDepth(JSContext* cx, JSScript* script, jsbytecode* pc,
+ uint32_t* depth, bool* reachablePC) {
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ BytecodeParser parser(cx, allocScope.alloc(), script);
+ if (!parser.parse()) {
+ return false;
+ }
+
+ *reachablePC = parser.isReachable(pc);
+
+ if (*reachablePC) {
+ *depth = parser.stackDepthAtPC(pc);
+ }
+
+ return true;
+}
+
+static unsigned Disassemble1(JSContext* cx, HandleScript script, jsbytecode* pc,
+ unsigned loc, bool lines,
+ const BytecodeParser* parser, Sprinter* sp);
+
+/*
+ * If pc != nullptr, include a prefix indicating whether the PC is at the
+ * current line. If showAll is true, include the source note type and the
+ * entry stack depth.
+ */
+static MOZ_MUST_USE bool DisassembleAtPC(
+ JSContext* cx, JSScript* scriptArg, bool lines, const jsbytecode* pc,
+ bool showAll, Sprinter* sp,
+ DisassembleSkeptically skeptically = DisassembleSkeptically::No) {
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ RootedScript script(cx, scriptArg);
+ mozilla::Maybe<BytecodeParser> parser;
+
+ if (skeptically == DisassembleSkeptically::No) {
+ parser.emplace(cx, allocScope.alloc(), script);
+ parser->setStackDump();
+ if (!parser->parse()) {
+ return false;
+ }
+ }
+
+ if (showAll) {
+ if (!sp->jsprintf("%s:%u\n", script->filename(),
+ unsigned(script->lineno()))) {
+ return false;
+ }
+ }
+
+ if (pc != nullptr) {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+ if (showAll) {
+ if (!sp->put("sn stack ")) {
+ return false;
+ }
+ }
+ if (!sp->put("loc ")) {
+ return false;
+ }
+ if (lines) {
+ if (!sp->put("line")) {
+ return false;
+ }
+ }
+ if (!sp->put(" op\n")) {
+ return false;
+ }
+
+ if (pc != nullptr) {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+ if (showAll) {
+ if (!sp->put("-- ----- ")) {
+ return false;
+ }
+ }
+ if (!sp->put("----- ")) {
+ return false;
+ }
+ if (lines) {
+ if (!sp->put("----")) {
+ return false;
+ }
+ }
+ if (!sp->put(" --\n")) {
+ return false;
+ }
+
+ jsbytecode* next = script->code();
+ jsbytecode* end = script->codeEnd();
+ while (next < end) {
+ if (next == script->main()) {
+ if (!sp->put("main:\n")) {
+ return false;
+ }
+ }
+ if (pc != nullptr) {
+ if (!sp->put(pc == next ? "--> " : " ")) {
+ return false;
+ }
+ }
+ if (showAll) {
+ const SrcNote* sn = GetSrcNote(cx, script, next);
+ if (sn) {
+ MOZ_ASSERT(!sn->isTerminator());
+ SrcNoteIterator iter(sn);
+ while (true) {
+ ++iter;
+ auto next = *iter;
+ if (!(!next->isTerminator() && next->delta() == 0)) {
+ break;
+ }
+ if (!sp->jsprintf("%s\n ", sn->name())) {
+ return false;
+ }
+ sn = *iter;
+ }
+ if (!sp->jsprintf("%s ", sn->name())) {
+ return false;
+ }
+ } else {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+ if (parser && parser->isReachable(next)) {
+ if (!sp->jsprintf("%05u ", parser->stackDepthAtPC(next))) {
+ return false;
+ }
+ } else {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+ }
+ unsigned len = Disassemble1(cx, script, next, script->pcToOffset(next),
+ lines, parser.ptrOr(nullptr), sp);
+ if (!len) {
+ return false;
+ }
+
+ next += len;
+ }
+
+ return true;
+}
+
+bool js::Disassemble(JSContext* cx, HandleScript script, bool lines,
+ Sprinter* sp, DisassembleSkeptically skeptically) {
+ return DisassembleAtPC(cx, script, lines, nullptr, false, sp, skeptically);
+}
+
+JS_FRIEND_API bool js::DumpPC(JSContext* cx, FILE* fp) {
+ gc::AutoSuppressGC suppressGC(cx);
+ Sprinter sprinter(cx);
+ if (!sprinter.init()) {
+ return false;
+ }
+ ScriptFrameIter iter(cx);
+ if (iter.done()) {
+ fprintf(fp, "Empty stack.\n");
+ return true;
+ }
+ RootedScript script(cx, iter.script());
+ bool ok = DisassembleAtPC(cx, script, true, iter.pc(), false, &sprinter);
+ fprintf(fp, "%s", sprinter.string());
+ return ok;
+}
+
+JS_FRIEND_API bool js::DumpScript(JSContext* cx, JSScript* scriptArg,
+ FILE* fp) {
+ gc::AutoSuppressGC suppressGC(cx);
+ Sprinter sprinter(cx);
+ if (!sprinter.init()) {
+ return false;
+ }
+ RootedScript script(cx, scriptArg);
+ bool ok = Disassemble(cx, script, true, &sprinter);
+ fprintf(fp, "%s", sprinter.string());
+ return ok;
+}
+
+static UniqueChars ToDisassemblySource(JSContext* cx, HandleValue v) {
+ if (v.isString()) {
+ return QuoteString(cx, v.toString(), '"');
+ }
+
+ if (JS::RuntimeHeapIsBusy()) {
+ return DuplicateString(cx, "<value>");
+ }
+
+ if (v.isObject()) {
+ JSObject& obj = v.toObject();
+
+ if (obj.is<JSFunction>()) {
+ RootedFunction fun(cx, &obj.as<JSFunction>());
+ JSString* str = JS_DecompileFunction(cx, fun);
+ if (!str) {
+ return nullptr;
+ }
+ return QuoteString(cx, str);
+ }
+
+ if (obj.is<RegExpObject>()) {
+ Rooted<RegExpObject*> reobj(cx, &obj.as<RegExpObject>());
+ JSString* source = RegExpObject::toString(cx, reobj);
+ if (!source) {
+ return nullptr;
+ }
+ return QuoteString(cx, source);
+ }
+ }
+
+ JSString* str = ValueToSource(cx, v);
+ if (!str) {
+ return nullptr;
+ }
+ return QuoteString(cx, str);
+}
+
+static bool ToDisassemblySource(JSContext* cx, HandleScope scope,
+ UniqueChars* bytes) {
+ UniqueChars source = JS_smprintf("%s {", ScopeKindString(scope->kind()));
+ if (!source) {
+ ReportOutOfMemory(cx);
+ return false;
+ }
+
+ for (Rooted<BindingIter> bi(cx, BindingIter(scope)); bi; bi++) {
+ UniqueChars nameBytes = AtomToPrintableString(cx, bi.name());
+ if (!nameBytes) {
+ return false;
+ }
+
+ source = JS_sprintf_append(std::move(source), "%s: ", nameBytes.get());
+ if (!source) {
+ ReportOutOfMemory(cx);
+ return false;
+ }
+
+ BindingLocation loc = bi.location();
+ switch (loc.kind()) {
+ case BindingLocation::Kind::Global:
+ source = JS_sprintf_append(std::move(source), "global");
+ break;
+
+ case BindingLocation::Kind::Frame:
+ source =
+ JS_sprintf_append(std::move(source), "frame slot %u", loc.slot());
+ break;
+
+ case BindingLocation::Kind::Environment:
+ source =
+ JS_sprintf_append(std::move(source), "env slot %u", loc.slot());
+ break;
+
+ case BindingLocation::Kind::Argument:
+ source =
+ JS_sprintf_append(std::move(source), "arg slot %u", loc.slot());
+ break;
+
+ case BindingLocation::Kind::NamedLambdaCallee:
+ source = JS_sprintf_append(std::move(source), "named lambda callee");
+ break;
+
+ case BindingLocation::Kind::Import:
+ source = JS_sprintf_append(std::move(source), "import");
+ break;
+ }
+
+ if (!source) {
+ ReportOutOfMemory(cx);
+ return false;
+ }
+
+ if (!bi.isLast()) {
+ source = JS_sprintf_append(std::move(source), ", ");
+ if (!source) {
+ ReportOutOfMemory(cx);
+ return false;
+ }
+ }
+ }
+
+ source = JS_sprintf_append(std::move(source), "}");
+ if (!source) {
+ ReportOutOfMemory(cx);
+ return false;
+ }
+
+ *bytes = std::move(source);
+ return true;
+}
+
+static bool DumpJumpOrigins(HandleScript script, jsbytecode* pc,
+ const BytecodeParser* parser, Sprinter* sp) {
+ bool called = false;
+ auto callback = [&script, &sp, &called](jsbytecode* pc,
+ BytecodeParser::JumpKind kind) {
+ if (!called) {
+ called = true;
+ if (!sp->put("\n# ")) {
+ return false;
+ }
+ } else {
+ if (!sp->put(", ")) {
+ return false;
+ }
+ }
+
+ switch (kind) {
+ case BytecodeParser::JumpKind::Simple:
+ break;
+
+ case BytecodeParser::JumpKind::SwitchCase:
+ if (!sp->put("switch-case ")) {
+ return false;
+ }
+ break;
+
+ case BytecodeParser::JumpKind::SwitchDefault:
+ if (!sp->put("switch-default ")) {
+ return false;
+ }
+ break;
+
+ case BytecodeParser::JumpKind::TryCatch:
+ if (!sp->put("try-catch ")) {
+ return false;
+ }
+ break;
+
+ case BytecodeParser::JumpKind::TryFinally:
+ if (!sp->put("try-finally ")) {
+ return false;
+ }
+ break;
+ }
+
+ if (!sp->jsprintf("from %s @ %05u", CodeName(JSOp(*pc)),
+ unsigned(script->pcToOffset(pc)))) {
+ return false;
+ }
+
+ return true;
+ };
+ if (!parser->forEachJumpOrigins(pc, callback)) {
+ return false;
+ }
+ if (called) {
+ if (!sp->put("\n")) {
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static bool DecompileAtPCForStackDump(
+ JSContext* cx, HandleScript script,
+ const OffsetAndDefIndex& offsetAndDefIndex, Sprinter* sp);
+
+static unsigned Disassemble1(JSContext* cx, HandleScript script, jsbytecode* pc,
+ unsigned loc, bool lines,
+ const BytecodeParser* parser, Sprinter* sp) {
+ if (parser && parser->isReachable(pc)) {
+ if (!DumpJumpOrigins(script, pc, parser, sp)) {
+ return 0;
+ }
+ }
+
+ size_t before = sp->stringEnd() - sp->string();
+ bool stackDumped = false;
+ auto dumpStack = [&cx, &script, &pc, &parser, &sp, &before, &stackDumped]() {
+ if (!parser) {
+ return true;
+ }
+ if (stackDumped) {
+ return true;
+ }
+ stackDumped = true;
+
+ size_t after = sp->stringEnd() - sp->string();
+ MOZ_ASSERT(after >= before);
+
+ static const size_t stack_column = 40;
+ for (size_t i = after - before; i < stack_column - 1; i++) {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+
+ if (!sp->put(" # ")) {
+ return false;
+ }
+
+ if (!parser->isReachable(pc)) {
+ if (!sp->put("!!! UNREACHABLE !!!")) {
+ return false;
+ }
+ } else {
+ uint32_t depth = parser->stackDepthAfterPC(pc);
+
+ for (uint32_t i = 0; i < depth; i++) {
+ if (i) {
+ if (!sp->put(" ")) {
+ return false;
+ }
+ }
+
+ const OffsetAndDefIndex& offsetAndDefIndex =
+ parser->offsetForStackOperandAfterPC(script->pcToOffset(pc), i);
+ // This will decompile the stack for the same PC many times.
+ // We'll avoid optimizing it since this is a testing function
+ // and it won't be worth managing cached expression here.
+ if (!DecompileAtPCForStackDump(cx, script, offsetAndDefIndex, sp)) {
+ return false;
+ }
+ }
+ }
+
+ return true;
+ };
+
+ if (*pc >= JSOP_LIMIT) {
+ char numBuf1[12], numBuf2[12];
+ SprintfLiteral(numBuf1, "%d", int(*pc));
+ SprintfLiteral(numBuf2, "%d", JSOP_LIMIT);
+ JS_ReportErrorNumberASCII(cx, GetErrorMessage, nullptr,
+ JSMSG_BYTECODE_TOO_BIG, numBuf1, numBuf2);
+ return 0;
+ }
+ JSOp op = JSOp(*pc);
+ const JSCodeSpec& cs = CodeSpec(op);
+ const unsigned len = cs.length;
+ if (!sp->jsprintf("%05u:", loc)) {
+ return 0;
+ }
+ if (lines) {
+ if (!sp->jsprintf("%4u", PCToLineNumber(script, pc))) {
+ return 0;
+ }
+ }
+ if (!sp->jsprintf(" %s", CodeName(op))) {
+ return 0;
+ }
+
+ int i;
+ switch (JOF_TYPE(cs.format)) {
+ case JOF_BYTE:
+ break;
+
+ case JOF_JUMP: {
+ ptrdiff_t off = GET_JUMP_OFFSET(pc);
+ if (!sp->jsprintf(" %u (%+d)", unsigned(loc + int(off)), int(off))) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_SCOPE: {
+ RootedScope scope(cx, script->getScope(pc));
+ UniqueChars bytes;
+ if (!ToDisassemblySource(cx, scope, &bytes)) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s", bytes.get())) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_ENVCOORD: {
+ RootedValue v(cx, StringValue(EnvironmentCoordinateNameSlow(script, pc)));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ EnvironmentCoordinate ec(pc);
+ if (!sp->jsprintf(" %s (hops = %u, slot = %u)", bytes.get(), ec.hops(),
+ ec.slot())) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_ATOM: {
+ RootedValue v(cx, StringValue(script->getAtom(pc)));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s", bytes.get())) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_DOUBLE: {
+ double d = GET_INLINE_VALUE(pc).toDouble();
+ if (!sp->jsprintf(" %lf", d)) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_BIGINT: {
+ RootedValue v(cx, BigIntValue(script->getBigInt(pc)));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s", bytes.get())) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_OBJECT: {
+ JSObject* obj = script->getObject(pc);
+ {
+ RootedValue v(cx, ObjectValue(*obj));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s", bytes.get())) {
+ return 0;
+ }
+ }
+ break;
+ }
+
+ case JOF_REGEXP: {
+ js::RegExpObject* obj = script->getRegExp(pc);
+ RootedValue v(cx, ObjectValue(*obj));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s", bytes.get())) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_TABLESWITCH: {
+ int32_t i, low, high;
+
+ ptrdiff_t off = GET_JUMP_OFFSET(pc);
+ jsbytecode* pc2 = pc + JUMP_OFFSET_LEN;
+ low = GET_JUMP_OFFSET(pc2);
+ pc2 += JUMP_OFFSET_LEN;
+ high = GET_JUMP_OFFSET(pc2);
+ pc2 += JUMP_OFFSET_LEN;
+ if (!sp->jsprintf(" defaultOffset %d low %d high %d", int(off), low,
+ high)) {
+ return 0;
+ }
+
+ // Display stack dump before diplaying the offsets for each case.
+ if (!dumpStack()) {
+ return 0;
+ }
+
+ for (i = low; i <= high; i++) {
+ off =
+ script->tableSwitchCaseOffset(pc, i - low) - script->pcToOffset(pc);
+ if (!sp->jsprintf("\n\t%d: %d", i, int(off))) {
+ return 0;
+ }
+ }
+ break;
+ }
+
+ case JOF_QARG:
+ if (!sp->jsprintf(" %u", GET_ARGNO(pc))) {
+ return 0;
+ }
+ break;
+
+ case JOF_LOCAL:
+ if (!sp->jsprintf(" %u", GET_LOCALNO(pc))) {
+ return 0;
+ }
+ break;
+
+ case JOF_GCTHING:
+ if (!sp->jsprintf(" %u", unsigned(GET_GCTHING_INDEX(pc)))) {
+ return 0;
+ }
+ break;
+
+ case JOF_UINT32:
+ if (!sp->jsprintf(" %u", GET_UINT32(pc))) {
+ return 0;
+ }
+ break;
+
+ case JOF_ICINDEX:
+ if (!sp->jsprintf(" (ic: %u)", GET_ICINDEX(pc))) {
+ return 0;
+ }
+ break;
+
+ case JOF_LOOPHEAD:
+ if (!sp->jsprintf(" (ic: %u, depthHint: %u)", GET_ICINDEX(pc),
+ LoopHeadDepthHint(pc))) {
+ return 0;
+ }
+ break;
+
+ case JOF_CLASS_CTOR: {
+ GCThingIndex atomIndex;
+ uint32_t classStartOffset = 0, classEndOffset = 0;
+ GetClassConstructorOperands(pc, &atomIndex, &classStartOffset,
+ &classEndOffset);
+ RootedValue v(cx, StringValue(script->getAtom(atomIndex)));
+ UniqueChars bytes = ToDisassemblySource(cx, v);
+ if (!bytes) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %s (off: %u-%u)", bytes.get(), classStartOffset,
+ classEndOffset)) {
+ return 0;
+ }
+ break;
+ }
+ case JOF_TWO_UINT8: {
+ int one = (int)GET_UINT8(pc);
+ int two = (int)GET_UINT8(pc + 1);
+
+ if (!sp->jsprintf(" %d", one)) {
+ return 0;
+ }
+ if (!sp->jsprintf(" %d", two)) {
+ return 0;
+ }
+ break;
+ }
+
+ case JOF_ARGC:
+ case JOF_UINT16:
+ i = (int)GET_UINT16(pc);
+ goto print_int;
+
+ case JOF_RESUMEINDEX:
+ case JOF_UINT24:
+ MOZ_ASSERT(len == 4);
+ i = (int)GET_UINT24(pc);
+ goto print_int;
+
+ case JOF_UINT8:
+ i = GET_UINT8(pc);
+ goto print_int;
+
+ case JOF_INT8:
+ i = GET_INT8(pc);
+ goto print_int;
+
+ case JOF_INT32:
+ MOZ_ASSERT(op == JSOp::Int32);
+ i = GET_INT32(pc);
+ print_int:
+ if (!sp->jsprintf(" %d", i)) {
+ return 0;
+ }
+ break;
+
+ default: {
+ char numBuf[12];
+ SprintfLiteral(numBuf, "%x", cs.format);
+ JS_ReportErrorNumberASCII(cx, GetErrorMessage, nullptr,
+ JSMSG_UNKNOWN_FORMAT, numBuf);
+ return 0;
+ }
+ }
+
+ if (!dumpStack()) {
+ return 0;
+ }
+
+ if (!sp->put("\n")) {
+ return 0;
+ }
+ return len;
+}
+
+unsigned js::Disassemble1(JSContext* cx, JS::Handle<JSScript*> script,
+ jsbytecode* pc, unsigned loc, bool lines,
+ Sprinter* sp) {
+ return Disassemble1(cx, script, pc, loc, lines, nullptr, sp);
+}
+
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+
+namespace {
+/*
+ * The expression decompiler is invoked by error handling code to produce a
+ * string representation of the erroring expression. As it's only a debugging
+ * tool, it only supports basic expressions. For anything complicated, it simply
+ * puts "(intermediate value)" into the error result.
+ *
+ * Here's the basic algorithm:
+ *
+ * 1. Find the stack location of the value whose expression we wish to
+ * decompile. The error handler can explicitly pass this as an
+ * argument. Otherwise, we search backwards down the stack for the offending
+ * value.
+ *
+ * 2. Instantiate and run a BytecodeParser for the current frame. This creates a
+ * stack of pcs parallel to the interpreter stack; given an interpreter stack
+ * location, the corresponding pc stack location contains the opcode that pushed
+ * the value in the interpreter. Now, with the result of step 1, we have the
+ * opcode responsible for pushing the value we want to decompile.
+ *
+ * 3. Pass the opcode to decompilePC. decompilePC is the main decompiler
+ * routine, responsible for a string representation of the expression that
+ * generated a certain stack location. decompilePC looks at one opcode and
+ * returns the JS source equivalent of that opcode.
+ *
+ * 4. Expressions can, of course, contain subexpressions. For example, the
+ * literals "4" and "5" are subexpressions of the addition operator in "4 +
+ * 5". If we need to decompile a subexpression, we call decompilePC (step 2)
+ * recursively on the operands' pcs. The result is a depth-first traversal of
+ * the expression tree.
+ *
+ */
+struct ExpressionDecompiler {
+ JSContext* cx;
+ RootedScript script;
+ const BytecodeParser& parser;
+ Sprinter sprinter;
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ // Dedicated mode for stack dump.
+ // Generates an expression for stack dump, including internal state,
+ // and also disables special handling for self-hosted code.
+ bool isStackDump;
+#endif
+
+ ExpressionDecompiler(JSContext* cx, JSScript* script,
+ const BytecodeParser& parser)
+ : cx(cx),
+ script(cx, script),
+ parser(parser),
+ sprinter(cx)
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ ,
+ isStackDump(false)
+#endif
+ {
+ }
+ bool init();
+ bool decompilePCForStackOperand(jsbytecode* pc, int i);
+ bool decompilePC(jsbytecode* pc, uint8_t defIndex);
+ bool decompilePC(const OffsetAndDefIndex& offsetAndDefIndex);
+ JSAtom* getArg(unsigned slot);
+ JSAtom* loadAtom(jsbytecode* pc);
+ bool quote(JSString* s, char quote);
+ bool write(const char* s);
+ bool write(JSString* str);
+ UniqueChars getOutput();
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ void setStackDump() { isStackDump = true; }
+#endif
+};
+
+bool ExpressionDecompiler::decompilePCForStackOperand(jsbytecode* pc, int i) {
+ return decompilePC(parser.offsetForStackOperand(script->pcToOffset(pc), i));
+}
+
+bool ExpressionDecompiler::decompilePC(jsbytecode* pc, uint8_t defIndex) {
+ MOZ_ASSERT(script->containsPC(pc));
+
+ JSOp op = (JSOp)*pc;
+
+ if (const char* token = CodeToken[uint8_t(op)]) {
+ MOZ_ASSERT(defIndex == 0);
+ MOZ_ASSERT(CodeSpec(op).ndefs == 1);
+
+ // Handle simple cases of binary and unary operators.
+ switch (CodeSpec(op).nuses) {
+ case 2: {
+ const SrcNote* sn = GetSrcNote(cx, script, pc);
+ const char* extra =
+ sn && sn->type() == SrcNoteType::AssignOp ? "=" : "";
+ return write("(") && decompilePCForStackOperand(pc, -2) && write(" ") &&
+ write(token) && write(extra) && write(" ") &&
+ decompilePCForStackOperand(pc, -1) && write(")");
+ break;
+ }
+ case 1:
+ return write("(") && write(token) &&
+ decompilePCForStackOperand(pc, -1) && write(")");
+ default:
+ break;
+ }
+ }
+
+ switch (op) {
+ case JSOp::DelName:
+ return write("(delete ") && write(loadAtom(pc)) && write(")");
+
+ case JSOp::GetGName:
+ case JSOp::GetName:
+ case JSOp::GetIntrinsic:
+ return write(loadAtom(pc));
+ case JSOp::GetArg: {
+ unsigned slot = GET_ARGNO(pc);
+
+ // For self-hosted scripts that are called from non-self-hosted code,
+ // decompiling the parameter name in the self-hosted script is
+ // unhelpful. Decompile the argument name instead.
+ if (script->selfHosted()
+#ifdef DEBUG
+ // For stack dump, argument name is not necessary.
+ && !isStackDump
+#endif /* DEBUG */
+ ) {
+ UniqueChars result;
+ if (!DecompileArgumentFromStack(cx, slot, &result)) {
+ return false;
+ }
+
+ // Note that decompiling the argument in the parent frame might
+ // not succeed.
+ if (result) {
+ return write(result.get());
+ }
+
+ // If it fails, do not return parameter name and let the caller
+ // fallback.
+ return write("(intermediate value)");
+ }
+
+ JSAtom* atom = getArg(slot);
+ if (!atom) {
+ return false;
+ }
+ return write(atom);
+ }
+ case JSOp::GetLocal: {
+ JSAtom* atom = FrameSlotName(script, pc);
+ MOZ_ASSERT(atom);
+ return write(atom);
+ }
+ case JSOp::GetAliasedVar: {
+ JSAtom* atom = EnvironmentCoordinateNameSlow(script, pc);
+ MOZ_ASSERT(atom);
+ return write(atom);
+ }
+
+ case JSOp::DelProp:
+ case JSOp::StrictDelProp:
+ case JSOp::GetProp:
+ case JSOp::GetBoundName: {
+ bool hasDelete = op == JSOp::DelProp || op == JSOp::StrictDelProp;
+ RootedAtom prop(cx, loadAtom(pc));
+ MOZ_ASSERT(prop);
+ return (hasDelete ? write("(delete ") : true) &&
+ decompilePCForStackOperand(pc, -1) &&
+ (IsIdentifier(prop)
+ ? write(".") && quote(prop, '\0')
+ : write("[") && quote(prop, '\'') && write("]")) &&
+ (hasDelete ? write(")") : true);
+ }
+ case JSOp::GetPropSuper: {
+ RootedAtom prop(cx, loadAtom(pc));
+ return write("super.") && quote(prop, '\0');
+ }
+ case JSOp::SetElem:
+ case JSOp::StrictSetElem:
+ // NOTE: We don't show the right hand side of the operation because
+ // it's used in error messages like: "a[0] is not readable".
+ //
+ // We could though.
+ return decompilePCForStackOperand(pc, -3) && write("[") &&
+ decompilePCForStackOperand(pc, -2) && write("]");
+
+ case JSOp::DelElem:
+ case JSOp::StrictDelElem:
+ case JSOp::GetElem: {
+ bool hasDelete = (op == JSOp::DelElem || op == JSOp::StrictDelElem);
+ return (hasDelete ? write("(delete ") : true) &&
+ decompilePCForStackOperand(pc, -2) && write("[") &&
+ decompilePCForStackOperand(pc, -1) && write("]") &&
+ (hasDelete ? write(")") : true);
+ }
+
+ case JSOp::GetElemSuper:
+ return write("super[") && decompilePCForStackOperand(pc, -2) &&
+ write("]");
+ case JSOp::Null:
+ return write(js_null_str);
+ case JSOp::True:
+ return write(js_true_str);
+ case JSOp::False:
+ return write(js_false_str);
+ case JSOp::Zero:
+ case JSOp::One:
+ case JSOp::Int8:
+ case JSOp::Uint16:
+ case JSOp::Uint24:
+ case JSOp::Int32:
+ return sprinter.printf("%d", GetBytecodeInteger(pc));
+ case JSOp::String:
+ return quote(loadAtom(pc), '"');
+ case JSOp::Symbol: {
+ unsigned i = uint8_t(pc[1]);
+ MOZ_ASSERT(i < JS::WellKnownSymbolLimit);
+ if (i < JS::WellKnownSymbolLimit) {
+ return write(cx->names().wellKnownSymbolDescriptions()[i]);
+ }
+ break;
+ }
+ case JSOp::Undefined:
+ return write(js_undefined_str);
+ case JSOp::GlobalThis:
+ // |this| could convert to a very long object initialiser, so cite it by
+ // its keyword name.
+ return write(js_this_str);
+ case JSOp::NewTarget:
+ return write("new.target");
+ case JSOp::Call:
+ case JSOp::CallIgnoresRv:
+ case JSOp::CallIter:
+ case JSOp::FunCall:
+ case JSOp::FunApply: {
+ uint16_t argc = GET_ARGC(pc);
+ return decompilePCForStackOperand(pc, -int32_t(argc + 2)) &&
+ write(argc ? "(...)" : "()");
+ }
+ case JSOp::SpreadCall:
+ return decompilePCForStackOperand(pc, -3) && write("(...)");
+ case JSOp::NewArray:
+ return write("[]");
+ case JSOp::RegExp: {
+ Rooted<RegExpObject*> obj(cx, &script->getObject(pc)->as<RegExpObject>());
+ JSString* str = RegExpObject::toString(cx, obj);
+ if (!str) {
+ return false;
+ }
+ return write(str);
+ }
+ case JSOp::Object: {
+ JSObject* obj = script->getObject(pc);
+ RootedValue objv(cx, ObjectValue(*obj));
+ JSString* str = ValueToSource(cx, objv);
+ if (!str) {
+ return false;
+ }
+ return write(str);
+ }
+ case JSOp::Void:
+ return write("(void ") && decompilePCForStackOperand(pc, -1) &&
+ write(")");
+
+ case JSOp::SuperCall:
+ if (GET_ARGC(pc) == 0) {
+ return write("super()");
+ }
+ [[fallthrough]];
+ case JSOp::SpreadSuperCall:
+ return write("super(...)");
+ case JSOp::SuperFun:
+ return write("super");
+
+ case JSOp::Eval:
+ case JSOp::SpreadEval:
+ case JSOp::StrictEval:
+ case JSOp::StrictSpreadEval:
+ return write("eval(...)");
+
+ case JSOp::New: {
+ uint16_t argc = GET_ARGC(pc);
+ return write("(new ") &&
+ decompilePCForStackOperand(pc, -int32_t(argc + 3)) &&
+ write(argc ? "(...))" : "())");
+ }
+
+ case JSOp::SpreadNew:
+ return write("(new ") && decompilePCForStackOperand(pc, -4) &&
+ write("(...))");
+
+ case JSOp::Typeof:
+ case JSOp::TypeofExpr:
+ return write("(typeof ") && decompilePCForStackOperand(pc, -1) &&
+ write(")");
+
+ case JSOp::InitElemArray:
+ return write("[...]");
+
+ case JSOp::InitElemInc:
+ if (defIndex == 0) {
+ return write("[...]");
+ }
+ MOZ_ASSERT(defIndex == 1);
+#ifdef DEBUG
+ // INDEX won't be be exposed to error message.
+ if (isStackDump) {
+ return write("INDEX");
+ }
+#endif
+ break;
+
+ case JSOp::ToNumeric:
+ return write("(tonumeric ") && decompilePCForStackOperand(pc, -1) &&
+ write(")");
+
+ case JSOp::Inc:
+ return write("(inc ") && decompilePCForStackOperand(pc, -1) && write(")");
+
+ case JSOp::Dec:
+ return write("(dec ") && decompilePCForStackOperand(pc, -1) && write(")");
+
+ case JSOp::BigInt:
+#if defined(DEBUG) || defined(JS_JITSPEW)
+ // BigInt::dump() only available in this configuration.
+ script->getBigInt(pc)->dump(sprinter);
+ return !sprinter.hadOutOfMemory();
+#else
+ return write("[bigint]");
+#endif
+
+ case JSOp::BuiltinObject: {
+ auto kind = BuiltinObjectKind(GET_UINT8(pc));
+ return write(BuiltinObjectName(kind));
+ }
+
+ default:
+ break;
+ }
+
+#ifdef DEBUG
+ if (isStackDump) {
+ // Special decompilation for stack dump.
+ switch (op) {
+ case JSOp::Arguments:
+ return write("arguments");
+
+ case JSOp::BindGName:
+ return write("GLOBAL");
+
+ case JSOp::BindName:
+ case JSOp::BindVar:
+ return write("ENV");
+
+ case JSOp::Callee:
+ return write("CALLEE");
+
+ case JSOp::EnvCallee:
+ return write("ENVCALLEE");
+
+ case JSOp::CallSiteObj:
+ return write("OBJ");
+
+ case JSOp::ClassConstructor:
+ case JSOp::DerivedConstructor:
+ return write("CONSTRUCTOR");
+
+ case JSOp::Double:
+ return sprinter.printf("%lf", GET_INLINE_VALUE(pc).toDouble());
+
+ case JSOp::Exception:
+ return write("EXCEPTION");
+
+ case JSOp::Finally:
+ if (defIndex == 0) {
+ return write("THROWING");
+ }
+ MOZ_ASSERT(defIndex == 1);
+ return write("PC");
+
+ case JSOp::GImplicitThis:
+ case JSOp::FunctionThis:
+ case JSOp::ImplicitThis:
+ return write("THIS");
+
+ case JSOp::FunWithProto:
+ return write("FUN");
+
+ case JSOp::Generator:
+ return write("GENERATOR");
+
+ case JSOp::GetImport:
+ return write("VAL");
+
+ case JSOp::GetRval:
+ return write("RVAL");
+
+ case JSOp::Hole:
+ return write("HOLE");
+
+ case JSOp::IsGenClosing:
+ // For stack dump, defIndex == 0 is not used.
+ MOZ_ASSERT(defIndex == 1);
+ return write("ISGENCLOSING");
+
+ case JSOp::IsNoIter:
+ // For stack dump, defIndex == 0 is not used.
+ MOZ_ASSERT(defIndex == 1);
+ return write("ISNOITER");
+
+ case JSOp::IsConstructing:
+ return write("JS_IS_CONSTRUCTING");
+
+ case JSOp::Iter:
+ return write("ITER");
+
+ case JSOp::Lambda:
+ case JSOp::LambdaArrow:
+ return write("FUN");
+
+ case JSOp::ToAsyncIter:
+ return write("ASYNCITER");
+
+ case JSOp::MoreIter:
+ // For stack dump, defIndex == 0 is not used.
+ MOZ_ASSERT(defIndex == 1);
+ return write("MOREITER");
+
+ case JSOp::MutateProto:
+ return write("SUCCEEDED");
+
+ case JSOp::NewInit:
+ case JSOp::NewObject:
+ case JSOp::ObjWithProto:
+ return write("OBJ");
+
+ case JSOp::OptimizeSpreadCall:
+ // For stack dump, defIndex == 0 is not used.
+ MOZ_ASSERT(defIndex == 1);
+ return write("OPTIMIZED");
+
+ case JSOp::Rest:
+ return write("REST");
+
+ case JSOp::Resume:
+ return write("RVAL");
+
+ case JSOp::SuperBase:
+ return write("HOMEOBJECTPROTO");
+
+ case JSOp::ToPropertyKey:
+ return write("TOPROPERTYKEY(") && decompilePCForStackOperand(pc, -1) &&
+ write(")");
+ case JSOp::ToString:
+ return write("TOSTRING(") && decompilePCForStackOperand(pc, -1) &&
+ write(")");
+
+ case JSOp::Uninitialized:
+ return write("UNINITIALIZED");
+
+ case JSOp::InitialYield:
+ case JSOp::Await:
+ case JSOp::Yield:
+ // Printing "yield SOMETHING" is confusing since the operand doesn't
+ // match to the syntax, since the stack operand for "yield 10" is
+ // the result object, not 10.
+ if (defIndex == 0) {
+ return write("RVAL");
+ }
+ if (defIndex == 1) {
+ return write("GENERATOR");
+ }
+ MOZ_ASSERT(defIndex == 2);
+ return write("RESUMEKIND");
+
+ case JSOp::ResumeKind:
+ return write("RESUMEKIND");
+
+ case JSOp::AsyncAwait:
+ case JSOp::AsyncResolve:
+ return write("PROMISE");
+
+ case JSOp::CheckPrivateField:
+ return write("HasPrivateField");
+
+ default:
+ break;
+ }
+ return write("<unknown>");
+ }
+#endif /* DEBUG */
+
+ return write("(intermediate value)");
+}
+
+bool ExpressionDecompiler::decompilePC(
+ const OffsetAndDefIndex& offsetAndDefIndex) {
+ if (offsetAndDefIndex.isSpecial()) {
+#ifdef DEBUG
+ if (isStackDump) {
+ if (offsetAndDefIndex.isMerged()) {
+ if (!write("merged<")) {
+ return false;
+ }
+ } else if (offsetAndDefIndex.isIgnored()) {
+ if (!write("ignored<")) {
+ return false;
+ }
+ }
+
+ if (!decompilePC(script->offsetToPC(offsetAndDefIndex.specialOffset()),
+ offsetAndDefIndex.specialDefIndex())) {
+ return false;
+ }
+
+ if (!write(">")) {
+ return false;
+ }
+
+ return true;
+ }
+#endif /* DEBUG */
+ return write("(intermediate value)");
+ }
+
+ return decompilePC(script->offsetToPC(offsetAndDefIndex.offset()),
+ offsetAndDefIndex.defIndex());
+}
+
+bool ExpressionDecompiler::init() {
+ cx->check(script);
+ return sprinter.init();
+}
+
+bool ExpressionDecompiler::write(const char* s) { return sprinter.put(s); }
+
+bool ExpressionDecompiler::write(JSString* str) {
+ if (str == cx->names().dotThis) {
+ return write("this");
+ }
+ return sprinter.putString(str);
+}
+
+bool ExpressionDecompiler::quote(JSString* s, char quote) {
+ return QuoteString(&sprinter, s, quote);
+}
+
+JSAtom* ExpressionDecompiler::loadAtom(jsbytecode* pc) {
+ return script->getAtom(pc);
+}
+
+JSAtom* ExpressionDecompiler::getArg(unsigned slot) {
+ MOZ_ASSERT(script->isFunction());
+ MOZ_ASSERT(slot < script->numArgs());
+
+ for (PositionalFormalParameterIter fi(script); fi; fi++) {
+ if (fi.argumentSlot() == slot) {
+ if (!fi.isDestructured()) {
+ return fi.name();
+ }
+
+ // Destructured arguments have no single binding name.
+ static const char destructuredParam[] = "(destructured parameter)";
+ return Atomize(cx, destructuredParam, strlen(destructuredParam));
+ }
+ }
+
+ MOZ_CRASH("No binding");
+}
+
+UniqueChars ExpressionDecompiler::getOutput() {
+ ptrdiff_t len = sprinter.stringEnd() - sprinter.stringAt(0);
+ auto res = cx->make_pod_array<char>(len + 1);
+ if (!res) {
+ return nullptr;
+ }
+ js_memcpy(res.get(), sprinter.stringAt(0), len);
+ res[len] = 0;
+ return res;
+}
+
+} // anonymous namespace
+
+#if defined(DEBUG) || defined(JS_JITSPEW)
+static bool DecompileAtPCForStackDump(
+ JSContext* cx, HandleScript script,
+ const OffsetAndDefIndex& offsetAndDefIndex, Sprinter* sp) {
+ // The expression decompiler asserts the script is in the current realm.
+ AutoRealm ar(cx, script);
+
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ BytecodeParser parser(cx, allocScope.alloc(), script);
+ parser.setStackDump();
+ if (!parser.parse()) {
+ return false;
+ }
+
+ ExpressionDecompiler ed(cx, script, parser);
+ ed.setStackDump();
+ if (!ed.init()) {
+ return false;
+ }
+
+ if (!ed.decompilePC(offsetAndDefIndex)) {
+ return false;
+ }
+
+ UniqueChars result = ed.getOutput();
+ if (!result) {
+ return false;
+ }
+
+ return sp->put(result.get());
+}
+#endif /* defined(DEBUG) || defined(JS_JITSPEW) */
+
+static bool FindStartPC(JSContext* cx, const FrameIter& iter,
+ const BytecodeParser& parser, int spindex,
+ int skipStackHits, const Value& v, jsbytecode** valuepc,
+ uint8_t* defIndex) {
+ jsbytecode* current = *valuepc;
+ *valuepc = nullptr;
+ *defIndex = 0;
+
+ if (spindex < 0 && spindex + int(parser.stackDepthAtPC(current)) < 0) {
+ spindex = JSDVG_SEARCH_STACK;
+ }
+
+ if (spindex == JSDVG_SEARCH_STACK) {
+ size_t index = iter.numFrameSlots();
+
+ // The decompiler may be called from inside functions that are not
+ // called from script, but via the C++ API directly, such as
+ // Invoke. In that case, the youngest script frame may have a
+ // completely unrelated pc and stack depth, so we give up.
+ if (index < size_t(parser.stackDepthAtPC(current))) {
+ return true;
+ }
+
+ // We search from fp->sp to base to find the most recently calculated
+ // value matching v under assumption that it is the value that caused
+ // the exception.
+ int stackHits = 0;
+ Value s;
+ do {
+ if (!index) {
+ return true;
+ }
+ s = iter.frameSlotValue(--index);
+ } while (s != v || stackHits++ != skipStackHits);
+
+ // If the current PC has fewer values on the stack than the index we are
+ // looking for, the blamed value must be one pushed by the current
+ // bytecode (e.g. JSOp::MoreIter), so restore *valuepc.
+ if (index < size_t(parser.stackDepthAtPC(current))) {
+ *valuepc = parser.pcForStackOperand(current, index, defIndex);
+ } else {
+ *valuepc = current;
+ *defIndex = index - size_t(parser.stackDepthAtPC(current));
+ }
+ } else {
+ *valuepc = parser.pcForStackOperand(current, spindex, defIndex);
+ }
+ return true;
+}
+
+static bool DecompileExpressionFromStack(JSContext* cx, int spindex,
+ int skipStackHits, HandleValue v,
+ UniqueChars* res) {
+ MOZ_ASSERT(spindex < 0 || spindex == JSDVG_IGNORE_STACK ||
+ spindex == JSDVG_SEARCH_STACK);
+
+ *res = nullptr;
+
+ /*
+ * Give up if we need deterministic behavior for differential testing.
+ * IonMonkey doesn't use InterpreterFrames and this ensures we get the same
+ * error messages.
+ */
+ if (js::SupportDifferentialTesting()) {
+ return true;
+ }
+
+ if (spindex == JSDVG_IGNORE_STACK) {
+ return true;
+ }
+
+ FrameIter frameIter(cx);
+
+ if (frameIter.done() || !frameIter.hasScript() ||
+ frameIter.realm() != cx->realm()) {
+ return true;
+ }
+
+ /*
+ * FIXME: Fall back if iter.isIon(), since the stack snapshot may be for the
+ * previous pc (see bug 831120).
+ */
+ if (frameIter.isIon()) {
+ return true;
+ }
+
+ RootedScript script(cx, frameIter.script());
+ jsbytecode* valuepc = frameIter.pc();
+
+ MOZ_ASSERT(script->containsPC(valuepc));
+
+ // Give up if in prologue.
+ if (valuepc < script->main()) {
+ return true;
+ }
+
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ BytecodeParser parser(cx, allocScope.alloc(), frameIter.script());
+ if (!parser.parse()) {
+ return false;
+ }
+
+ uint8_t defIndex;
+ if (!FindStartPC(cx, frameIter, parser, spindex, skipStackHits, v, &valuepc,
+ &defIndex)) {
+ return false;
+ }
+ if (!valuepc) {
+ return true;
+ }
+
+ ExpressionDecompiler ed(cx, script, parser);
+ if (!ed.init()) {
+ return false;
+ }
+ if (!ed.decompilePC(valuepc, defIndex)) {
+ return false;
+ }
+
+ *res = ed.getOutput();
+ return *res != nullptr;
+}
+
+UniqueChars js::DecompileValueGenerator(JSContext* cx, int spindex,
+ HandleValue v, HandleString fallbackArg,
+ int skipStackHits) {
+ RootedString fallback(cx, fallbackArg);
+ {
+ UniqueChars result;
+ if (!DecompileExpressionFromStack(cx, spindex, skipStackHits, v, &result)) {
+ return nullptr;
+ }
+ if (result && strcmp(result.get(), "(intermediate value)")) {
+ return result;
+ }
+ }
+ if (!fallback) {
+ if (v.isUndefined()) {
+ return DuplicateString(
+ cx, js_undefined_str); // Prevent users from seeing "(void 0)"
+ }
+ fallback = ValueToSource(cx, v);
+ if (!fallback) {
+ return nullptr;
+ }
+ }
+
+ return StringToNewUTF8CharsZ(cx, *fallback);
+}
+
+static bool DecompileArgumentFromStack(JSContext* cx, int formalIndex,
+ UniqueChars* res) {
+ MOZ_ASSERT(formalIndex >= 0);
+
+ *res = nullptr;
+
+ /* See note in DecompileExpressionFromStack. */
+ if (js::SupportDifferentialTesting()) {
+ return true;
+ }
+
+ /*
+ * Settle on the nearest script frame, which should be the builtin that
+ * called the intrinsic.
+ */
+ FrameIter frameIter(cx);
+ MOZ_ASSERT(!frameIter.done());
+ MOZ_ASSERT(frameIter.script()->selfHosted());
+
+ /*
+ * Get the second-to-top frame, the non-self-hosted caller of the builtin
+ * that called the intrinsic.
+ */
+ ++frameIter;
+ if (frameIter.done() || !frameIter.hasScript() ||
+ frameIter.script()->selfHosted() || frameIter.realm() != cx->realm()) {
+ return true;
+ }
+
+ RootedScript script(cx, frameIter.script());
+ jsbytecode* current = frameIter.pc();
+
+ MOZ_ASSERT(script->containsPC(current));
+
+ if (current < script->main()) {
+ return true;
+ }
+
+ /* Don't handle getters, setters or calls from fun.call/fun.apply. */
+ JSOp op = JSOp(*current);
+ if (op != JSOp::Call && op != JSOp::CallIgnoresRv && op != JSOp::New) {
+ return true;
+ }
+
+ if (static_cast<unsigned>(formalIndex) >= GET_ARGC(current)) {
+ return true;
+ }
+
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ BytecodeParser parser(cx, allocScope.alloc(), script);
+ if (!parser.parse()) {
+ return false;
+ }
+
+ bool pushedNewTarget = op == JSOp::New;
+ int formalStackIndex = parser.stackDepthAtPC(current) - GET_ARGC(current) -
+ pushedNewTarget + formalIndex;
+ MOZ_ASSERT(formalStackIndex >= 0);
+ if (uint32_t(formalStackIndex) >= parser.stackDepthAtPC(current)) {
+ return true;
+ }
+
+ ExpressionDecompiler ed(cx, script, parser);
+ if (!ed.init()) {
+ return false;
+ }
+ if (!ed.decompilePCForStackOperand(current, formalStackIndex)) {
+ return false;
+ }
+
+ *res = ed.getOutput();
+ return *res != nullptr;
+}
+
+JSString* js::DecompileArgument(JSContext* cx, int formalIndex, HandleValue v) {
+ {
+ UniqueChars result;
+ if (!DecompileArgumentFromStack(cx, formalIndex, &result)) {
+ return nullptr;
+ }
+ if (result && strcmp(result.get(), "(intermediate value)")) {
+ JS::ConstUTF8CharsZ utf8chars(result.get(), strlen(result.get()));
+ return NewStringCopyUTF8Z<CanGC>(cx, utf8chars);
+ }
+ }
+ if (v.isUndefined()) {
+ return cx->names().undefined; // Prevent users from seeing "(void 0)"
+ }
+
+ return ValueToSource(cx, v);
+}
+
+extern bool js::IsValidBytecodeOffset(JSContext* cx, JSScript* script,
+ size_t offset) {
+ // This could be faster (by following jump instructions if the target
+ // is <= offset).
+ for (BytecodeRange r(cx, script); !r.empty(); r.popFront()) {
+ size_t here = r.frontOffset();
+ if (here >= offset) {
+ return here == offset;
+ }
+ }
+ return false;
+}
+
+/*
+ * There are three possible PCCount profiling states:
+ *
+ * 1. None: Neither scripts nor the runtime have count information.
+ * 2. Profile: Active scripts have count information, the runtime does not.
+ * 3. Query: Scripts do not have count information, the runtime does.
+ *
+ * When starting to profile scripts, counting begins immediately, with all JIT
+ * code discarded and recompiled with counts as necessary. Active interpreter
+ * frames will not begin profiling until they begin executing another script
+ * (via a call or return).
+ *
+ * The below API functions manage transitions to new states, according
+ * to the table below.
+ *
+ * Old State
+ * -------------------------
+ * Function None Profile Query
+ * --------
+ * StartPCCountProfiling Profile Profile Profile
+ * StopPCCountProfiling None Query Query
+ * PurgePCCounts None None None
+ */
+
+static void ReleaseScriptCounts(JSRuntime* rt) {
+ MOZ_ASSERT(rt->scriptAndCountsVector);
+
+ js_delete(rt->scriptAndCountsVector.ref());
+ rt->scriptAndCountsVector = nullptr;
+}
+
+JS_FRIEND_API void js::StartPCCountProfiling(JSContext* cx) {
+ JSRuntime* rt = cx->runtime();
+
+ if (rt->profilingScripts) {
+ return;
+ }
+
+ if (rt->scriptAndCountsVector) {
+ ReleaseScriptCounts(rt);
+ }
+
+ ReleaseAllJITCode(rt->defaultFreeOp());
+
+ rt->profilingScripts = true;
+}
+
+JS_FRIEND_API void js::StopPCCountProfiling(JSContext* cx) {
+ JSRuntime* rt = cx->runtime();
+
+ if (!rt->profilingScripts) {
+ return;
+ }
+ MOZ_ASSERT(!rt->scriptAndCountsVector);
+
+ ReleaseAllJITCode(rt->defaultFreeOp());
+
+ auto* vec = cx->new_<PersistentRooted<ScriptAndCountsVector>>(
+ cx, ScriptAndCountsVector());
+ if (!vec) {
+ return;
+ }
+
+ for (ZonesIter zone(rt, SkipAtoms); !zone.done(); zone.next()) {
+ for (auto base = zone->cellIter<BaseScript>(); !base.done(); base.next()) {
+ if (base->hasScriptCounts() && base->hasJitScript()) {
+ if (!vec->append(base->asJSScript())) {
+ return;
+ }
+ }
+ }
+ }
+
+ rt->profilingScripts = false;
+ rt->scriptAndCountsVector = vec;
+}
+
+JS_FRIEND_API void js::PurgePCCounts(JSContext* cx) {
+ JSRuntime* rt = cx->runtime();
+
+ if (!rt->scriptAndCountsVector) {
+ return;
+ }
+ MOZ_ASSERT(!rt->profilingScripts);
+
+ ReleaseScriptCounts(rt);
+}
+
+JS_FRIEND_API size_t js::GetPCCountScriptCount(JSContext* cx) {
+ JSRuntime* rt = cx->runtime();
+
+ if (!rt->scriptAndCountsVector) {
+ return 0;
+ }
+
+ return rt->scriptAndCountsVector->length();
+}
+
+static MOZ_MUST_USE bool JSONStringProperty(Sprinter& sp, JSONPrinter& json,
+ const char* name, JSString* str) {
+ json.beginStringProperty(name);
+ if (!JSONQuoteString(&sp, str)) {
+ return false;
+ }
+ json.endStringProperty();
+ return true;
+}
+
+JS_FRIEND_API JSString* js::GetPCCountScriptSummary(JSContext* cx,
+ size_t index) {
+ JSRuntime* rt = cx->runtime();
+
+ if (!rt->scriptAndCountsVector ||
+ index >= rt->scriptAndCountsVector->length()) {
+ JS_ReportErrorNumberASCII(cx, GetErrorMessage, nullptr,
+ JSMSG_BUFFER_TOO_SMALL);
+ return nullptr;
+ }
+
+ const ScriptAndCounts& sac = (*rt->scriptAndCountsVector)[index];
+ RootedScript script(cx, sac.script);
+
+ Sprinter sp(cx);
+ if (!sp.init()) {
+ return nullptr;
+ }
+
+ JSONPrinter json(sp, false);
+
+ json.beginObject();
+
+ RootedString filename(cx, NewStringCopyZ<CanGC>(cx, script->filename()));
+ if (!filename) {
+ return nullptr;
+ }
+ if (!JSONStringProperty(sp, json, "file", filename)) {
+ return nullptr;
+ }
+ json.property("line", script->lineno());
+
+ if (JSFunction* fun = script->function()) {
+ if (JSAtom* atom = fun->displayAtom()) {
+ if (!JSONStringProperty(sp, json, "name", atom)) {
+ return nullptr;
+ }
+ }
+ }
+
+ uint64_t total = 0;
+
+ AllBytecodesIterable iter(script);
+ for (BytecodeLocation loc : iter) {
+ if (const PCCounts* counts = sac.maybeGetPCCounts(loc.toRawBytecode())) {
+ total += counts->numExec();
+ }
+ }
+
+ json.beginObjectProperty("totals");
+
+ json.property(PCCounts::numExecName, total);
+
+ uint64_t ionActivity = 0;
+ jit::IonScriptCounts* ionCounts = sac.getIonCounts();
+ while (ionCounts) {
+ for (size_t i = 0; i < ionCounts->numBlocks(); i++) {
+ ionActivity += ionCounts->block(i).hitCount();
+ }
+ ionCounts = ionCounts->previous();
+ }
+ if (ionActivity) {
+ json.property("ion", ionActivity);
+ }
+
+ json.endObject();
+
+ json.endObject();
+
+ if (sp.hadOutOfMemory()) {
+ return nullptr;
+ }
+
+ return NewStringCopyZ<CanGC>(cx, sp.string());
+}
+
+static bool GetPCCountJSON(JSContext* cx, const ScriptAndCounts& sac,
+ Sprinter& sp) {
+ JSONPrinter json(sp, false);
+
+ RootedScript script(cx, sac.script);
+
+ LifoAllocScope allocScope(&cx->tempLifoAlloc());
+ BytecodeParser parser(cx, allocScope.alloc(), script);
+ if (!parser.parse()) {
+ return false;
+ }
+
+ json.beginObject();
+
+ JSString* str = JS_DecompileScript(cx, script);
+ if (!str) {
+ return false;
+ }
+
+ if (!JSONStringProperty(sp, json, "text", str)) {
+ return false;
+ }
+
+ json.property("line", script->lineno());
+
+ json.beginListProperty("opcodes");
+
+ uint64_t hits = 0;
+ for (BytecodeRangeWithPosition range(cx, script); !range.empty();
+ range.popFront()) {
+ jsbytecode* pc = range.frontPC();
+ size_t offset = script->pcToOffset(pc);
+ JSOp op = JSOp(*pc);
+
+ // If the current instruction is a jump target,
+ // then update the number of hits.
+ if (const PCCounts* counts = sac.maybeGetPCCounts(pc)) {
+ hits = counts->numExec();
+ }
+
+ json.beginObject();
+
+ json.property("id", offset);
+ json.property("line", range.frontLineNumber());
+ json.property("name", CodeName(op));
+
+ {
+ ExpressionDecompiler ed(cx, script, parser);
+ if (!ed.init()) {
+ return false;
+ }
+ // defIndex passed here is not used.
+ if (!ed.decompilePC(pc, /* defIndex = */ 0)) {
+ return false;
+ }
+ UniqueChars text = ed.getOutput();
+ if (!text) {
+ return false;
+ }
+
+ JS::ConstUTF8CharsZ utf8chars(text.get(), strlen(text.get()));
+ JSString* str = NewStringCopyUTF8Z<CanGC>(cx, utf8chars);
+ if (!str) {
+ return false;
+ }
+
+ if (!JSONStringProperty(sp, json, "text", str)) {
+ return false;
+ }
+ }
+
+ json.beginObjectProperty("counts");
+ if (hits > 0) {
+ json.property(PCCounts::numExecName, hits);
+ }
+ json.endObject();
+
+ json.endObject();
+
+ // If the current instruction has thrown,
+ // then decrement the hit counts with the number of throws.
+ if (const PCCounts* counts = sac.maybeGetThrowCounts(pc)) {
+ hits -= counts->numExec();
+ }
+ }
+
+ json.endList();
+
+ if (jit::IonScriptCounts* ionCounts = sac.getIonCounts()) {
+ json.beginListProperty("ion");
+
+ while (ionCounts) {
+ json.beginList();
+ for (size_t i = 0; i < ionCounts->numBlocks(); i++) {
+ const jit::IonBlockCounts& block = ionCounts->block(i);
+
+ json.beginObject();
+ json.property("id", block.id());
+ json.property("offset", block.offset());
+
+ json.beginListProperty("successors");
+ for (size_t j = 0; j < block.numSuccessors(); j++) {
+ json.value(block.successor(j));
+ }
+ json.endList();
+
+ json.property("hits", block.hitCount());
+
+ JSString* str = NewStringCopyZ<CanGC>(cx, block.code());
+ if (!str) {
+ return false;
+ }
+
+ if (!JSONStringProperty(sp, json, "code", str)) {
+ return false;
+ }
+
+ json.endObject();
+ }
+ json.endList();
+
+ ionCounts = ionCounts->previous();
+ }
+
+ json.endList();
+ }
+
+ json.endObject();
+
+ return true;
+}
+
+JS_FRIEND_API JSString* js::GetPCCountScriptContents(JSContext* cx,
+ size_t index) {
+ JSRuntime* rt = cx->runtime();
+
+ if (!rt->scriptAndCountsVector ||
+ index >= rt->scriptAndCountsVector->length()) {
+ JS_ReportErrorNumberASCII(cx, GetErrorMessage, nullptr,
+ JSMSG_BUFFER_TOO_SMALL);
+ return nullptr;
+ }
+
+ const ScriptAndCounts& sac = (*rt->scriptAndCountsVector)[index];
+ JSScript* script = sac.script;
+
+ Sprinter sp(cx);
+ if (!sp.init()) {
+ return nullptr;
+ }
+
+ {
+ AutoRealm ar(cx, &script->global());
+ if (!GetPCCountJSON(cx, sac, sp)) {
+ return nullptr;
+ }
+ }
+
+ if (sp.hadOutOfMemory()) {
+ return nullptr;
+ }
+
+ return NewStringCopyZ<CanGC>(cx, sp.string());
+}
+
+struct CollectedScripts {
+ MutableHandle<ScriptVector> scripts;
+ bool ok = true;
+
+ explicit CollectedScripts(MutableHandle<ScriptVector> scripts)
+ : scripts(scripts) {}
+
+ static void consider(JSRuntime* rt, void* data, BaseScript* script,
+ const JS::AutoRequireNoGC& nogc) {
+ auto self = static_cast<CollectedScripts*>(data);
+ if (!script->filename()) {
+ return;
+ }
+ if (!self->scripts.append(script->asJSScript())) {
+ self->ok = false;
+ }
+ }
+};
+
+static bool GenerateLcovInfo(JSContext* cx, JS::Realm* realm,
+ GenericPrinter& out) {
+ AutoRealmUnchecked ar(cx, realm);
+
+ // Collect the list of scripts which are part of the current realm.
+
+ MOZ_RELEASE_ASSERT(
+ coverage::IsLCovEnabled(),
+ "Coverage must be enabled for process before generating LCov info");
+
+ // Hold the scripts that we have already flushed, to avoid flushing them
+ // twice.
+ using JSScriptSet = GCHashSet<JSScript*>;
+ Rooted<JSScriptSet> scriptsDone(cx, JSScriptSet(cx));
+
+ Rooted<ScriptVector> queue(cx, ScriptVector(cx));
+
+ {
+ CollectedScripts result(&queue);
+ IterateScripts(cx, realm, &result, &CollectedScripts::consider);
+ if (!result.ok) {
+ return false;
+ }
+ }
+
+ if (queue.length() == 0) {
+ return true;
+ }
+
+ // Ensure the LCovRealm exists to collect info into.
+ coverage::LCovRealm* lcovRealm = realm->lcovRealm();
+ if (!lcovRealm) {
+ return false;
+ }
+
+ // Collect code coverage info for one realm.
+ do {
+ RootedScript script(cx, queue.popCopy());
+ RootedFunction fun(cx);
+
+ JSScriptSet::AddPtr entry = scriptsDone.lookupForAdd(script);
+ if (entry) {
+ continue;
+ }
+
+ if (!coverage::CollectScriptCoverage(script, false)) {
+ return false;
+ }
+
+ script->resetScriptCounts();
+
+ if (!scriptsDone.add(entry, script)) {
+ return false;
+ }
+
+ if (!script->isTopLevel()) {
+ continue;
+ }
+
+ // Iterate from the last to the first object in order to have
+ // the functions them visited in the opposite order when popping
+ // elements from the stack of remaining scripts, such that the
+ // functions are more-less listed with increasing line numbers.
+ auto gcthings = script->gcthings();
+ for (JS::GCCellPtr gcThing : mozilla::Reversed(gcthings)) {
+ if (!gcThing.is<JSObject>()) {
+ continue;
+ }
+ JSObject* obj = &gcThing.as<JSObject>();
+
+ if (!obj->is<JSFunction>()) {
+ continue;
+ }
+ fun = &obj->as<JSFunction>();
+
+ // Ignore asm.js functions
+ if (!fun->isInterpreted()) {
+ continue;
+ }
+
+ // Queue the script in the list of script associated to the
+ // current source.
+ JSScript* childScript = JSFunction::getOrCreateScript(cx, fun);
+ if (!childScript || !queue.append(childScript)) {
+ return false;
+ }
+ }
+ } while (!queue.empty());
+
+ bool isEmpty = true;
+ lcovRealm->exportInto(out, &isEmpty);
+ if (out.hadOutOfMemory()) {
+ return false;
+ }
+
+ return true;
+}
+
+JS_FRIEND_API UniqueChars js::GetCodeCoverageSummaryAll(JSContext* cx,
+ size_t* length) {
+ Sprinter out(cx);
+ if (!out.init()) {
+ return nullptr;
+ }
+
+ for (RealmsIter realm(cx->runtime()); !realm.done(); realm.next()) {
+ if (!GenerateLcovInfo(cx, realm, out)) {
+ JS_ReportOutOfMemory(cx);
+ return nullptr;
+ }
+ }
+
+ *length = out.getOffset();
+ return js::DuplicateString(cx, out.string(), *length);
+}
+
+JS_FRIEND_API UniqueChars js::GetCodeCoverageSummary(JSContext* cx,
+ size_t* length) {
+ Sprinter out(cx);
+ if (!out.init()) {
+ return nullptr;
+ }
+
+ if (!GenerateLcovInfo(cx, cx->realm(), out)) {
+ JS_ReportOutOfMemory(cx);
+ return nullptr;
+ }
+
+ *length = out.getOffset();
+ return js::DuplicateString(cx, out.string(), *length);
+}