Google Git
Sign in
webkit / WebKit / 7794ba38fda6da562b59317fcbfc69439e38b8d9 / . / Source / JavaScriptCore / ChangeLog
blob: 59199c11651dd7f200f0e74c11654cc6b0674c23 [file] [log] [blame]
2020-07-01 Saam Barati <sbarati@apple.com>
Script to copy over testapi.js is redundant in xcodebuild
https://bugs.webkit.org/show_bug.cgi?id=213824
Reviewed by Keith Miller.
We're already copying over the entire testapiScripts directory.
No need to manually copy just one file from it, too.
* JavaScriptCore.xcodeproj/project.pbxproj:
2020-07-01 Tetsuharu Ohzeki <tetsuharu.ohzeki@gmail.com>
Use properly flag names for tests of Tools/Scripts/run-builtins-generator-tests
https://bugs.webkit.org/show_bug.cgi?id=213733
Reviewed by Youenn Fablet.
Test cases under Source/JavaScriptCore/Scripts/tests/builtins/ uses exist compilation flags.
But they can take an arbitary flag name and don't have to use an exist flag.
I think it's better to rename them to more proper ones.
* Scripts/tests/builtins/WebCore-ArbitraryConditionalGuard-Separate.js:
* Scripts/tests/builtins/WebCore-DuplicateKeyValueAnnotation-Separate.js:
* Scripts/tests/builtins/WebCore-GuardedBuiltin-Separate.js:
* Scripts/tests/builtins/WebCore-GuardedInternalBuiltin-Separate.js:
* Scripts/tests/builtins/WebCore-xmlCasingTest-Separate.js:
* Scripts/tests/builtins/expected/WebCore-ArbitraryConditionalGuard-Separate.js-result:
* Scripts/tests/builtins/expected/WebCore-GuardedBuiltin-Separate.js-result:
* Scripts/tests/builtins/expected/WebCore-GuardedInternalBuiltin-Separate.js-result:
* Scripts/tests/builtins/expected/WebCore-xmlCasingTest-Separate.js-result:
* Scripts/tests/builtins/expected/WebCoreJSBuiltins.h-result:
2020-06-30 Peng Liu <peng.liu6@apple.com>
Enable the support of FULLSCREEN_API in WebKitTestRunner
https://bugs.webkit.org/show_bug.cgi?id=213774
Reviewed by Youenn Fablet.
Replace the definition of ENABLE_FULLSCREEN_API in FeatureDefines.xcconfig with
the one in PlatformEnableCocoa.h. We have to do that because WebKitTestRunner
does not have a FeatureDefines.xcconfig but it uses "ENABLE(FULLSCREEN_API)"
to conditionally compile code to test the element fullscreen API.
WebKitTestRunner can use the macro defined in PlatformEnableCocoa.h.
* Configurations/FeatureDefines.xcconfig:
2020-06-30 Andy Estes <aestes@apple.com>
[Xcode] Enable the "My Mac (Mac Catalyst)" destination in WebKit Xcode projects
https://bugs.webkit.org/show_bug.cgi?id=213740
Reviewed by Darin Adler.
* Configurations/Base.xcconfig: Set SUPPORTS_MACCATALYST to YES to tell Xcode that this
project supports building for Mac Catalyst.
2020-06-29 Tetsuharu Ohzeki <tetsuharu.ohzeki@gmail.com>
Remove ENABLE_STREAMS_API compilation flag
https://bugs.webkit.org/show_bug.cgi?id=213728
Reviewed by Sam Weinig.
test cases under Scripts/tests/builtins/ does not uses
this removed compilation flag. So we don't have to touch them in this change.
But they are confusable so I plan to fix them in bug 213733.
* Configurations/FeatureDefines.xcconfig:
2020-06-29 Keith Miller <keith_miller@apple.com>
ConservativeRoots should mark any cell it finds an interior pointer to
https://bugs.webkit.org/show_bug.cgi?id=213686
Reviewed by Yusuke Suzuki.
Currently, if ConserativeRoots finds an interior pointer to a cell
it will only mark that cell if it's a butterfly of some
kind. However, this can cause problems if the C++ or B3 compilers
pre-compute the offset of some cell member they want to load from
after a call. If this happens and that interior pointer is the
only reference to the cell it can get collected while it is still
"alive".
A naive patch that doesn't return from
findGCObjectPointersForMarking after finding a live non-interior,
non-butterfly cell was a 2% regression on Speedometer2 and
JetStream2. So, this patch immediately returns after
marking some non-butterfly cell, which appears to have fixed the
regression locally. Given this was such a big regression (likely
from running MarkedBlock::isLive) more than once there's possibly
an optimization opportunity here. I filed
https://bugs.webkit.org/show_bug.cgi?id=213687 to investigate
further.
* heap/HeapCell.cpp:
(WTF::printInternal):
* heap/HeapCell.h:
(JSC::isJSCellKind):
(JSC::mayHaveIndexingHeader):
(JSC::hasInteriorPointers): Deleted.
* heap/HeapUtil.h:
(JSC::HeapUtil::findGCObjectPointersForMarking):
* heap/SlotVisitor.cpp:
(JSC::SlotVisitor::appendJSCellOrAuxiliary):
* runtime/VM.cpp:
(JSC::VM::VM):
2020-06-28 Geoffrey Garen <ggaren@apple.com>
Rename initializeThreading to initialize
https://bugs.webkit.org/show_bug.cgi?id=213674
Reviewed by Mark Lam.
* API/JSClassRef.cpp:
* API/JSContextRef.cpp:
(JSContextGroupCreate):
(JSGlobalContextCreate):
(JSGlobalContextCreateInGroup):
* API/JSObjectRef.cpp:
(JSClassCreate):
* API/JSStringRef.cpp:
(JSStringCreateWithCharacters):
(JSStringCreateWithUTF8CString):
(JSStringCreateWithCharactersNoCopy):
* API/JSStringRefCF.cpp:
(JSStringCreateWithCFString):
* API/tests/CompareAndSwapTest.cpp:
(testCompareAndSwap):
* API/tests/ExecutionTimeLimitTest.cpp:
(testExecutionTimeLimit):
* API/tests/FunctionOverridesTest.cpp:
(testFunctionOverrides):
* API/tests/MultithreadedMultiVMExecutionTest.cpp:
(startMultithreadedMultiVMExecutionTest):
* API/tests/PingPongStackOverflowTest.cpp:
(testPingPongStackOverflow):
* JavaScriptCore.order:
* assembler/testmasm.cpp:
(JSC::run):
* b3/air/testair.cpp:
(main):
* b3/testb3_1.cpp:
(main):
* dfg/testdfg.cpp:
(main):
* dynbench.cpp:
(main):
* inspector/remote/cocoa/RemoteInspectorCocoa.mm:
(Inspector::RemoteInspector::singleton):
* jsc.cpp:
(main):
(jscmain):
* runtime/InitializeThreading.cpp:
(JSC::initialize):
(JSC::initializeThreading): Deleted.
* runtime/InitializeThreading.h:
* runtime/JSCConfig.h:
* shell/playstation/TestShell.cpp:
(setupTestRun):
* testRegExp.cpp:
(main):
2020-06-27 Saam Barati <sbarati@apple.com>
BytecodeBasicBlock::addSuccessor should check if the successor already exists
https://bugs.webkit.org/show_bug.cgi?id=213670
Reviewed by Yusuke Suzuki.
It makes it nicer for algorithms using BytecodeGraph to not have to consider
whether or not there are duplicates in the successor list. Also, because of
this, bytecode liveness was doing extra work since it walked the successor list,
including any duplicates in it.
* bytecode/BytecodeBasicBlock.h:
(JSC::BytecodeBasicBlock::addSuccessor):
2020-06-27 Saam Barati <sbarati@apple.com>
Change bytecode dumping to dump the bytecode control flow graph
https://bugs.webkit.org/show_bug.cgi?id=213669
Reviewed by Yusuke Suzuki.
This makes the bytecode control flow graphs much easier to understand, and
puts bytecode dumping in more in line with how we dump other IRs.
The new dumps look like this:
```
foo#Ahf63N:[0x1035bc120->0x1035e5100, NoneFunctionCall, 36]: 13 instructions (0 16-bit instructions, 0 32-bit instructions, 1 instructions with metadata); 156 bytes (120 metadata bytes); 2 parameter(s); 8 callee register(s); 6 variable(s); scope at loc4
bb#1
[ 0] enter
[ 1] get_scope loc4
[ 3] mov loc5, loc4
[ 6] check_traps
[ 7] mov loc6, <JSValue()>(const0)
[ 10] mov loc6, Undefined(const1)
[ 13] mod loc7, arg1, Int32: 2(const2)
[ 17] jfalse loc7, 8(->25)
Successors: [ #3 #2 ]
bb#2
[ 20] mov loc6, Int32: 42(const3)
[ 23] jmp 5(->28)
Successors: [ #4 ]
bb#3
[ 25] mov loc6, Int32: 77(const4)
Successors: [ #4 ]
bb#4
[ 28] add loc7, arg1, loc6, OperandTypes(126, 126)
[ 34] ret loc7
Successors: [ ]
```
* bytecode/BytecodeDumper.cpp:
(JSC::dumpHeader):
(JSC::dumpFooter):
(JSC::CodeBlockBytecodeDumper<Block>::dumpBlock):
(JSC::CodeBlockBytecodeDumper<Block>::dumpGraph):
* bytecode/BytecodeDumper.h:
* bytecode/BytecodeGraph.h:
(JSC::BytecodeGraph::dump):
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::dumpBytecode):
2020-06-27 Stephan Szabo <stephan.szabo@sony.com>
[PlayStation] Update test runner for changes to Options and signing
https://bugs.webkit.org/show_bug.cgi?id=213650
Reviewed by Don Olmstead.
* shell/playstation/Initializer.cpp: Load ICU library
* shell/playstation/TestShell.cpp: Update between test options reset
2020-06-26 Geoffrey Garen <ggaren@apple.com>
Initializing the main thread should initialize the main run loop
https://bugs.webkit.org/show_bug.cgi?id=213637
Reviewed by Anders Carlsson.
* JavaScriptCore.order: Removed some defunct stuff.
* shell/playstation/TestShell.cpp:
(setupTestRun): Merged initializeThreading call with
initializeMainThread call because initializeMainThread is a superset.
2020-06-25 Yusuke Suzuki <ysuzuki@apple.com>
REGRESSION(r263035): stress/get-prototype-of.js broken on s390x
https://bugs.webkit.org/show_bug.cgi?id=213307
Reviewed by Ross Kirsling.
Structure::m_outOfLineTypeFlags is uint16_t. If we access this field as 32bit field, we have different value in big endian architectures.
Since we do not have half-size-load branch instructions, we should load this uint16_t value via `loadh` (which zero-extends the loaded value)
and perform branch onto that value.
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitLoadPrototype):
* llint/LowLevelInterpreter64.asm:
2020-06-25 Mark Lam <mark.lam@apple.com>
JSCell constructor needs to ensure that the passed in structure is still alive.
https://bugs.webkit.org/show_bug.cgi?id=213593
<rdar://problem/64597573>
Reviewed by Yusuke Suzuki.
Note that in the initializer list of the `JSCell(VM&, Structure*)` constructor,
we are only using values inside the passed in structure but not necessarily the
structure pointer itself. All these values are contained inside Structure::m_blob.
Note also that this constructor is an inline function. Hence, the compiler may
choose to pre-compute the address of structure->m_blob and discard the structure
pointer itself.
Here's an example:
0x10317a21e <+1054>: movq 0x18(%rsp), %rdx // rdx:vm = &vm
0x10317a223 <+1059>: addq $0x8, %r13 // r13 = &structure.m_blob <== pre-compute address of m_blob !!!
// r13 previously contained the structure pointer.
// Now, there's no more references to the structure base address.
0x10317a227 <+1063>: leaq 0x48(%rdx), %rdi // arg0:heap = &vm.heap
0x10317a22b <+1067>: movl $0x10, %edx // arg2:size = 16.
0x10317a230 <+1072>: movq %rdi, 0x28(%rsp)
0x10317a235 <+1077>: xorl %esi, %esi // arg1:deferralContext = 0
0x10317a237 <+1079>: callq 0x10317ae60 // call JSC::allocateCell<JSC::JSArray> <== Can GC here !!!
0x10317a23c <+1084>: movq %rax, %rbx // rbx:cell = rax:allocation result.
...
0x10317a253 <+1107>: movl (%r13), %eax // eax = m_blob.structureID <== Use pre-computed m_blob address here.
There's a chance that the GC may run while allocating this cell. In the event
that the structure is newly instantiated just before calling this constructor,
there may not be any other references to it. As a result, the structure may get
collected before the cell is even constructed. To avoid this possibility, we need
to ensure that the structure pointer is still alive by the time this constructor
is called.
I am not committing any tests for this issue because the test cases relies on:
1. Manually forcing an O3 ASan Release build.
2. Not running jsc with --useDollarVM=1. Note: the JSC test harness automatically
adds --useDollarVM=1 for all test runs.
3. Memory being allocated in a specific order. The recent Bitmap FreeList change
enabled this issue to manifest. The old linked list FreeList implementation
would have hidden the issue.
4. Adding some logging code can also make the issue stop manifesting.
In short, the test cases will not detect any regression even if we commit them
because the existing automatic regression test runs will not have the necessary
conditions for reproducing the issue. The tests are also somewhat fragile where
any changes in memory layout may stop the issue from manifesting in an observable
way.
* runtime/JSCellInlines.h:
(JSC::JSCell::JSCell):
2020-06-24 Ross Kirsling <ross.kirsling@sony.com>
[Intl] Disprefer using ICU enums directly as instance variables
https://bugs.webkit.org/show_bug.cgi?id=213587
Reviewed by Yusuke Suzuki.
* runtime/IntlPluralRules.cpp:
(JSC::IntlPluralRules::initializePluralRules):
(JSC::IntlPluralRules::resolvedOptions const):
* runtime/IntlPluralRules.h:
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::initializeRelativeTimeFormat):
(JSC::IntlRelativeTimeFormat::styleString): Renamed from JSC::styleString.
(JSC::IntlRelativeTimeFormat::resolvedOptions const):
(JSC::numericString): Deleted.
* runtime/IntlRelativeTimeFormat.h:
2020-06-24 Caitlin Potter <caitp@igalia.com>
[JSC] handle Put/DefinePrivateField in resetPutByID
https://bugs.webkit.org/show_bug.cgi?id=213583
Reviewed by Yusuke Suzuki.
r262613 extends and uses PutByValDirect to support updating and defining private fields, in order to reuse
the IC machinery. The necessary resetPutByID change was erroneously omitted, and is presented here.
* jit/Repatch.cpp:
(JSC::resetPutByID):
2020-06-24 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] llintTrue / jitTrue can encounter native functions
https://bugs.webkit.org/show_bug.cgi?id=213442
<rdar://problem/64257914>
Reviewed by Mark Lam.
If the CallFrame is for native function, associated CodeBlock is nullptr.
This patch fixes this case to handle it gracefully.
* tools/JSDollarVM.cpp:
(JSC::CallerFrameJITTypeFunctor::CallerFrameJITTypeFunctor):
(JSC::CallerFrameJITTypeFunctor::operator() const):
(JSC::functionBaselineJITTrue):
(JSC::JSDollarVM::finishCreation):
(JSC::functionJITTrue): Deleted.
2020-06-24 Umar Iqbal <uiqbal@apple.com>
We should resurrect the older patch that collects some statistics of web API calls
https://bugs.webkit.org/show_bug.cgi?id=213319
Reviewed by Brent Fulgham.
+ Enabled ENABLE_WEB_API_STATISTICS flag
* Configurations/FeatureDefines.xcconfig:
2020-06-24 Alexey Shvayka <shvaikalesh@gmail.com>
Add DFG/FTL fast path for GetPrototypeOf based on OverridesGetPrototype flag
https://bugs.webkit.org/show_bug.cgi?id=213191
Reviewed by Yusuke Suzuki.
This patch:
1. Introduces `loadInlineOffset` LLInt macro (64-bit only) and utilizes it in
`get_prototype_of` since we assert that `knownPolyProtoOffset` is an inline offset.
2. Brings baseline JIT fast path to 32-bit builds, progressing `super-getter.js`
microbenchmark by a factor of 10 (w/o DFG).
3. Adds GetPrototypeOf DFG/FTL fast paths that leverage OverridesGetPrototype type
info flag, advancing provided rare objects microbenchmark by ~46% (~37% w/o FTL).
Also, cleans up existing DFG fast path by using AssemblyHelpers::loadValue().
4. Extracts AssemblyHelpers::emitLoadPrototype() and uses it in baseline JIT, DFG, and
InstanceOfGeneric access case. With this change, `instanceof` access case handles all
[[GetPrototypeOf]] overrides (before: only Proxy objects), which is more correct, yet
not observable enough to provide a test case. `instanceof` microbenchmarks are neutral.
* bytecode/AccessCase.cpp:
(JSC::AccessCase::generateWithGuard):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileGetPrototypeOf):
* ftl/FTLAbstractHeapRepository.h:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileGetPrototypeOf):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitLoadPrototype):
* jit/AssemblyHelpers.h:
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_get_prototype_of):
* llint/LowLevelInterpreter64.asm:
2020-06-24 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Clobberize misses `write(Heap)` report in some nodes
https://bugs.webkit.org/show_bug.cgi?id=213525
<rdar://problem/64642067>
Reviewed by Mark Lam.
In some DFG nodes, clobberize phase misses `clobberTopFunctor` call while it is `write(Heap)`,
which confuses clobberizing validation.
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
2020-06-23 Mark Lam <mark.lam@apple.com>
Handle string overflow in DFG graph dump while validating AI.
https://bugs.webkit.org/show_bug.cgi?id=213524
<rdar://problem/64635620>
Not reviewed.
Applying refinement suggested by Darin in https://bugs.webkit.org/show_bug.cgi?id=213524#c3.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::validateAIState):
2020-06-23 Mark Lam <mark.lam@apple.com>
Handle string overflow in DFG graph dump while validating AI.
https://bugs.webkit.org/show_bug.cgi?id=213524
<rdar://problem/64635620>
Reviewed by Saam Barati.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::validateAIState):
2020-06-23 Devin Rousso <drousso@apple.com>
Keyframe animation doesn't 't show up in the Animations timeline
https://bugs.webkit.org/show_bug.cgi?id=213441
Reviewed by Brian Burg.
* inspector/protocol/Animation.json:
An `interationCount` of `Infinity` is not JSON serializable, so represent it as `-1` instead.
2020-06-22 Saam Barati <sbarati@apple.com>
Attempt to fix watchOS simulator build.
* assembler/FastJITPermissions.h:
(threadSelfRestrictRWXToRW):
(threadSelfRestrictRWXToRX):
2020-06-22 Saam Barati <sbarati@apple.com>
Allow building JavaScriptCore Mac+arm64 in public SDK build
https://bugs.webkit.org/show_bug.cgi?id=213472
Reviewed by Sam Weinig.
We used to only builld code for fast permission switching when using the
Apple internal SDK. However, with arm64 on macOS, this is no longer a viable
implementation strategy.
This patch makes it so we can build JSC on macOS+arm64 using the public Xcode
SDK.
- ENABLE_FAST_JIT_PERMISSIONS is removed. We now use runtime checks instead.
- In the new suite of OS betas, pthreads has added API for fast permissions
switching. We now use this API instead of using the non-public SDK found in
the kernel headers.
- We fall back to the separated W/X heaps when fast permissions checking is
not supported. This all happens at runtime.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* assembler/ARM64Assembler.h:
(JSC::ARM64Assembler::fillNops):
* assembler/ARMv7Assembler.h:
(JSC::ARMv7Assembler::fillNops):
* assembler/FastJITPermissions.h: Added.
(useFastJITPermissions):
(threadSelfRestrictRWXToRW):
(threadSelfRestrictRWXToRX):
(fastJITPermissionsIsSupported):
* assembler/LinkBuffer.cpp:
(JSC::memcpyWrapper):
(JSC::LinkBuffer::copyCompactAndLinkCode):
* assembler/MIPSAssembler.h:
(JSC::MIPSAssembler::fillNops):
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::link):
* assembler/MacroAssemblerARMv7.h:
(JSC::MacroAssemblerARMv7::link):
* jit/ExecutableAllocator.cpp:
(JSC::initializeJITPageReservation):
* jit/ExecutableAllocator.h:
(JSC::performJITMemcpy):
(JSC::useFastJITPermissions): Deleted.
* runtime/JSCConfig.h:
* runtime/Options.cpp:
(JSC::Options::recomputeDependentOptions):
* runtime/OptionsList.h:
2020-06-22 Tim Horton <timothy_horton@apple.com>
Disable the JS JIT when running in a translated process
https://bugs.webkit.org/show_bug.cgi?id=213478
Reviewed by Saam Barati.
* runtime/Options.cpp:
(JSC::Options::recomputeDependentOptions):
Based on our performance experiements, disable the JavaScript JIT
(but not the regular expression, DOM, or Wasm JIT) when running
in a translated process.
2020-06-22 Tim Horton <timothy_horton@apple.com>
Update macOS version macros
https://bugs.webkit.org/show_bug.cgi?id=213484
Reviewed by Alexey Proskuryakov.
* Configurations/Base.xcconfig:
* Configurations/DebugRelease.xcconfig:
* Configurations/Version.xcconfig:
* Configurations/WebKitTargetConditionals.xcconfig:
2020-06-19 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Check Gigacage usage before launching VM
https://bugs.webkit.org/show_bug.cgi?id=213410
Reviewed by Mark Lam.
Since VM allocates JSBigInt from Gigacage, it is possible that VM creation fails when Gigacage is exhausted.
As a work-around for internal testing, we insert ad-hoc Gigacage usage check before launching a new agent.
If 80% of Gigacage is used, we fail to launch a new VM gracefully.
* assembler/testmasm.cpp:
(JSC::testCagePreservesPACFailureBit):
* jsc.cpp:
(functionDollarAgentStart):
2020-06-19 James Darpinian <jdarpinian@chromium.org>
Typed array constructor behaves differently when length is not passed or when undefined is passed
https://bugs.webkit.org/show_bug.cgi?id=184232
Reviewed by Yusuke Suzuki.
Passing undefined for length should have the same effect as omitting the argument. It was being
treated as 0 instead.
* runtime/JSGenericTypedArrayViewConstructorInlines.h:
(JSC::constructGenericTypedArrayView):
2020-06-19 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Attempt to reduce timeout failures on Apple Watch Series 3
https://bugs.webkit.org/show_bug.cgi?id=213419
Reviewed by Mark Lam.
* tools/JSDollarVM.cpp:
(JSC::functionUseJIT):
(JSC::JSDollarVM::finishCreation):
2020-06-19 Mark Lam <mark.lam@apple.com>
toString of String doesn't check integrity of structureID in one path.
https://bugs.webkit.org/show_bug.cgi?id=213338
Reviewed by Saam Barati.
* runtime/StringPrototype.cpp:
(JSC::stringProtoFuncToString):
2020-06-19 Saam Barati <sbarati@apple.com>
Have a memory monitor thread in jsc shell when running tests using --memory-limited
https://bugs.webkit.org/show_bug.cgi?id=213389
Reviewed by Mark Lam.
When testing on iOS, there are times high memory usage from a JSC test
will jetsam our entire test runner. This makes it so we don't get any test
results from that test run, which can make it difficult to track testing
results.
This patch introduces an optional memory monitoring thread to the JSC
shell. It's a best effort approach. If memory usage exceeds the passed
in threshold, we crash the process. Similar to how the timeout mechanism
works. On Cocoa platforms, we also perform this check in the low memory
warning handler.
Currently, we use this feature when running JSC stress tests in
"--memory-limited" mode.
* jsc.cpp:
(crashIfExceedingMemoryLimit):
(startMemoryMonitoringThreadIfNeeded):
(jscmain):
2020-06-19 Mark Lam <mark.lam@apple.com>
Make $vm properties non-configurable, non-enumerable, and non-writable.
https://bugs.webkit.org/show_bug.cgi?id=213395
Reviewed by Saam Barati and Yusuke Suzuki.
$vm provides functions for test development and VM debugging. There's no reason
for them to be configurable, enumerable, and writable.
We particularly don't want them to be enumerable as this can trip up some fuzzers.
Fuzzers should not be fuzzing the $vm object which doesn't exist in real world
uses of JavaScriptCore.
* tools/JSDollarVM.cpp:
(JSC::JSDollarVM::finishCreation):
(JSC::JSDollarVM::addFunction):
(JSC::JSDollarVM::addConstructibleFunction):
2020-06-19 Tuomas Karkkainen <tuomas.webkit@apple.com>
functionCpuClflush checks that the second argument is Int32 but it actually expects it to be UInt32
https://bugs.webkit.org/show_bug.cgi?id=213388
Reviewed by Saam Barati.
This changes the check from isInt32() to isUInt32() so that the logic is consistent.
* tools/JSDollarVM.cpp:
2020-06-18 Mark Lam <mark.lam@apple.com>
Unify Bitmap math loops in MarkedBlock::Handle::specializedSweep().
https://bugs.webkit.org/show_bug.cgi?id=213345
Reviewed by Robin Morisset and Saam Barati.
This change appears to be performance neutral. However, we'll take the change
because we know that it does less work, and the new way of expressing the Bitmap
math in MarkedBlock::Handle::specializedSweep() does appear to be easier to
understand than the old code.
Also addressed feedback from Robin and Saam in https://bugs.webkit.org/show_bug.cgi?id=213071.
Changes made:
1. Use the new Bitmap::words() API to get direct access to the underlying bits
storage. With this, we can do the merging of the marked and newlyAllocated
bits with a single pass looping thru the bitmap words.
2. In MarkedBlock::Handle::specializedSweep()'s Bitmap free list code, moved the
implementation of handleDeadCells lambda down to the call to freeAtoms.forEachSetBit()
because this is the only place it is used.
3. Fixed MarkedBlock::Handle::specializedSweep()'s Bitmap free list code to
handle the dead cells unconditionally. This condition check was wrongly
adapted from the linked list implementation where handleDeadCell() was called
in 2 places depending on the destruction mode. With the Bitmap free list,
there is only once place to handle the dead cells, and it should be executed
unconditionally.
This fixes a bug where the FreeList::originalSize() never gets computed if the
cells in the block does not need destruction.
4. Renamed FreeList::bitmapRows() to FreeList::bitmapRowsMinusOne().
Renamed FreeList::offsetOfBitmapRows() to FreeList::offsetOfBitmapRowsMinusOne().
5. Also fixed some typos in comments.
* heap/FreeList.h:
(JSC::FreeList::bitmapIsEmpty const):
(JSC::FreeList::offsetOfBitmapRowsMinusOne):
(JSC::FreeList::bitmapRowsMinusOne const):
(JSC::FreeList::offsetOfBitmapRows): Deleted.
(JSC::FreeList::bitmapRows const): Deleted.
* heap/FreeListInlines.h:
(JSC::FreeList::allocate):
(JSC::FreeList::forEach const):
* heap/MarkedBlockInlines.h:
(JSC::MarkedBlock::Handle::specializedSweep):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitAllocateWithNonNullAllocator):
2020-06-18 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Remove dead non-ICU locale Date code since we are always using ICU version
https://bugs.webkit.org/show_bug.cgi?id=213362
Reviewed by Ross Kirsling.
There are old non-ICU version of Date locale code. But this is now dead code since we are always using ICU version,
which is invoked from builtin JS DatePrototype.js. We should remove these dead code.
* runtime/DatePrototype.cpp:
(JSC::DatePrototype::finishCreation):
(): Deleted.
(JSC::styleFromArgString): Deleted.
(JSC::formatLocaleDate): Deleted.
(JSC::dateProtoFuncToLocaleString): Deleted.
(JSC::dateProtoFuncToLocaleDateString): Deleted.
(JSC::dateProtoFuncToLocaleTimeString): Deleted.
2020-06-18 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed, address Darin's feedback on r263227.
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::UNumberFormatDeleter::operator() const):
(JSC::IntlRelativeTimeFormat::initializeRelativeTimeFormat):
(JSC::IntlRelativeTimeFormat::formatToParts const):
* runtime/IntlRelativeTimeFormat.h:
Keep ownership over our UNumberFormat instance after all,
to avoid relying on behavior ICU isn't explicitly guaranteeing.
2020-06-18 Ross Kirsling <ross.kirsling@sony.com>
[Intl] Enable RelativeTimeFormat and Locale by default
https://bugs.webkit.org/show_bug.cgi?id=213324
Reviewed by Yusuke Suzuki.
* runtime/IntlObject.cpp:
(JSC::createDateTimeFormatConstructor):
(JSC::createLocaleConstructor):
(JSC::createNumberFormatConstructor):
(JSC::createRelativeTimeFormatConstructor):
(JSC::IntlObject::finishCreation):
Unconditionalize creation of RelativeTimeFormat and Locale constructors.
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::initializeRelativeTimeFormat):
(JSC::IntlRelativeTimeFormat::formatToParts const):
(JSC::IntlRelativeTimeFormat::UNumberFormatDeleter::operator() const): Deleted.
* runtime/IntlRelativeTimeFormat.h:
Fix an actual bug -- URelativeDateTimeFormatter *adopts* the UNumberFormat it's instantiated with,
so we can't keep a unique_ptr to it.
* runtime/OptionsList.h:
Remove feature flags.
2020-06-18 Alexey Shvayka <shvaikalesh@gmail.com>
Promise built-in functions should be anonymous non-constructors
https://bugs.webkit.org/show_bug.cgi?id=213317
Reviewed by Darin Adler.
This patch makes userland-exposed Promise built-in functions
non-constructors and sets their "name" properties to empty strings
as per spec [1], aligning JSC with V8 and SpiderMonkey.
@createResolvingFunctionsWithoutPromise change is covered by test262's
async-generator/yield-thenable-create-resolving-functions-*.js cases.
Promise microbenchmarks are neutral. Promise constructors bytecode is
unchanged, while @createResolvingFunctions* bytecode is reduced by 2
instructions.
[1]: https://tc39.es/ecma262/#sec-ecmascript-standard-built-in-objects
* builtins/PromiseConstructor.js:
(nakedConstructor.Promise):
(nakedConstructor.InternalPromise):
* builtins/PromiseOperations.js:
(globalPrivate.newPromiseCapabilitySlow):
(globalPrivate.createResolvingFunctions):
(globalPrivate.createResolvingFunctionsWithoutPromise):
(globalPrivate.createResolvingFunctions.resolve): Deleted.
(globalPrivate.createResolvingFunctions.reject): Deleted.
(resolve): Deleted.
(reject): Deleted.
* builtins/PromisePrototype.js:
(globalPrivate.getThenFinally):
(globalPrivate.getCatchFinally):
(valueThunk): Deleted.
(thrower): Deleted.
2020-06-18 Alexey Shvayka <shvaikalesh@gmail.com>
TypedArray.prototype.set is incorrect with primitives
https://bugs.webkit.org/show_bug.cgi?id=212730
Reviewed by Yusuke Suzuki.
This change implements step 14 of %TypedArray%.prototype.set [1],
which coerces primitives to objects instead of throwing an error,
aligning JSC with V8 and SpiderMonkey.
[1]: https://tc39.es/ecma262/#sec-%typedarray%.prototype.set-array-offset
* runtime/JSGenericTypedArrayViewPrototypeFunctions.h:
(JSC::genericTypedArrayViewProtoFuncSet):
2020-06-17 Mark Lam <mark.lam@apple.com>
Replace JSC::FreeList linked list with a Bitmap.
https://bugs.webkit.org/show_bug.cgi?id=213071
Reviewed by Filip Pizlo.
Implement an alternative to the linked list FreeList. This alternative uses
a Bitmap to record which atom in the block is available for allocation.
The intuition here is that allocation using the Bitmap implementation will do:
2 loads - m_currentRowBitmap, m_currentMarkedBlockRowAddress
1 store - m_currentRowBitmap
whereas the linked list implementation will do:
3 loads - m_scrambledHead, m_secret, result->scrambledNext
1 store - m_scrambledHead
and result->scrambledNext is from a different region of code and therefore not
in the same cache line.
The downside of the Bitmap implementation is that it uses more instructions.
This change is currently only enabled for x86_64, which shows about a 0.8%
progression on Speedometer 2.
It appears to be about a 1% regression on ARM64E. Hence, for now, we keep the
linked list implementation for ARM64 builds.
This is how the Bitmap FreeList works:
1. The Bitmap implementation only replaces the linked list implementation. It
does not replace the bump allocator.
2. The Bitmap allocator keeps a m_bitmap that is initialized in
MarkedBlock::Handle::specializedSweep() to have a bit set for each atom
location that is available for allocation (i.e. is free). Note that a cell
is usually allocated using more than 1 atom. Only the bit corresponding to
the first atom (in that cell length range of free atoms) will be set.
This is consistent with how bits in MarkedBlock::Footer::m_marks and
MarkedBlock::Footer::m_newlyAllocated are set i.e. only the bit for the first
atom in the cell can be set.
3. The allocation algorithm thinks of the MarkedBlock as consisting of rows
of atoms, where the number of atoms in a row equals the number of bits in
a AtomsBitmap::Word. On 64-bit CPUs, this would be 64.
We will start allocating from the last (highest numbered) row down to the
first (row 0). As we allocate, we will only update m_currentRowIndex and
m_currentRowBitmap. m_bitmap will not be updated. This is so in order to
reduce the number of instructions executed during an allocation.
When m_currentRowIndex points to N, the AtomsBitmap::Word for row N in
m_bitmap will have been copied into m_currentRowBitmap. This is the row
that we will be allocating from until the row is exhausted.
This is how we know whether an atom is available for allocation or not:
i. Atoms in any rows above m_currentRowIndex are guaranteed to be
allocated already (because we allocate downwards), and hence, are not
available.
ii. For row m_currentRowIndex, m_currentRowBitmap is the source of truth
on which atoms in the row are available for allocation.
iii. For rows below m_currentRowIndex, m_bitmap is the source of truth on
which atoms are available for allocation.
When m_currentRowIndex reaches 0, the info in m_bitmap is completely
obsoleted, and m_currentRowBitmap holds the availability info for row 0.
When both m_currentRowIndex and m_currentRowBitmap are 0, then we have
completely exhausted the block and no more atoms are available for
allocation.
4. Allocation happens in 3 paths: fast, middle, slow.
The fast path checks m_currentRowBitmap. If it's not 0, then we compute the
bit number of the lowest set bit in it. That bit number will be used together
with m_currentMarkedBlockRowAddress to compute the address of the atom
location available for allocation. m_currentRowBitmap will be updated to clear
the bit for the atom that has just ben allocated.
If m_currentRowBitmap is 0, then we'll go to the middle path.
The middle path checks m_currentRowIndex to see if we have more rows to allocate
from. For each m_currentRowIndex, we check its corresponding AtomsBitmap::Word
in m_bitmap. If the word is non-zero, we copy it to m_currentRowBitmap and
jump to the fast path to do the allocation. The middle path will update
m_currentRowIndex to point to the current row we're allocating from.
If we have decremented m_currentRowIndex down to 0 but still can't find a
non-zero AtomsBitmap::Word in m_bitmap, then the block has been exhausted, and
we'll go to the slow path.
The slow path is analogous to the old slow path i.e. we try to refill the
LocalAllocator with a new MarkedBlock.
5. On the layout of fields in FreeList (see changes in FreeList.h), we try to
preserve the positions of the bump allocator fields. The only change we made
there is in the location of m_cellSize. It is now moved up next to m_remaining,
and m_originalSize is moved down. This is because m_originalSize is only
accessed in the slow path, and m_cellSize is accessed in the bump allocation
path.
Next, we try to put Bitmap allocation fields where the linked list fields
would have been. The one bit of trickiness is that we'll put
m_currentMarkedBlockRowAddress in a union with m_payloadEnd. This is because
m_payloadEnd is only used in the bump allocation path. If m_remaining is 0,
then we can reuse this location for m_currentMarkedBlockRowAddress.
With this, we would have 4 bytes of padding after m_currentRowIndex. For
compactness, we put m_originalSize there in that space. For builds that use
the linked list implementation, m_originalSize will be located below after
m_cellSize.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::allocateHeapCell):
* heap/FreeList.cpp:
(JSC::FreeList::clear):
(JSC::FreeList::initializeAtomsBitmap):
(JSC::FreeList::initializeBump):
(JSC::FreeList::contains const):
(JSC::FreeList::dump const):
* heap/FreeList.h:
(JSC::FreeList::bitmapIsEmpty const):
(JSC::FreeList::allocationWillFail const):
(JSC::FreeList::offsetOfCurrentRowBitmap):
(JSC::FreeList::offsetOfBitmapRows):
(JSC::FreeList::offsetOfCurrentRowIndex):
(JSC::FreeList::offsetOfCurrentMarkedBlockRowAddress):
(JSC::FreeList::offsetOfRemaining):
(JSC::FreeList::atomsBitmap):
(JSC::FreeList::bitmapRows const):
(JSC::FreeList::offsetOfOriginalSize): Deleted.
* heap/FreeListInlines.h:
(JSC::FreeList::allocate):
(JSC::FreeList::forEach const):
* heap/LocalAllocator.cpp:
(JSC::LocalAllocator::isFreeListedCell const):
* heap/MarkedBlock.h:
(JSC::MarkedBlock::Handle::atomAt const):
* heap/MarkedBlockInlines.h:
(JSC::MarkedBlock::Handle::specializedSweep):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitAllocateWithNonNullAllocator):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::emitAllocateWithNonNullAllocator):
2020-06-17 Mark Lam <mark.lam@apple.com>
StructureIDTable::validate() doesn't work when compiled with GCC.
https://bugs.webkit.org/show_bug.cgi?id=213302
<rdar://problem/64452172>
Reviewed by Yusuke Suzuki.
I was previously using ensureStillAliveHere() to force the validation load to
not be elided. However, this is not how ensureStillAliveHere() works. The proper
way to force the load is to use a volatile pointer instead, which is applied in
this patch.
With Clang, the ensureStillAliveHere() happened to do what I expected, but with
GCC it did not. The compiler is at liberty to elide the load because there is
no memory clobbering operation between the load and the call to
ensureStillAliveHere(). Switching to using the volatile pointer solution.
* runtime/StructureIDTable.h:
(JSC::StructureIDTable::validate):
2020-06-17 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Freeze JSBigInt when setting it as a constant in AI
https://bugs.webkit.org/show_bug.cgi?id=213310
<rdar://problem/64450410>
Reviewed by Mark Lam.
JSCells should be explicitly frozen via DFG::Graph::freeze or DFG::Graph::freezeStrong. And heap JSBigInt is JSCell.
We should freeze it before setting it as a parameter of setConstant in AI. We use DFG::Graph::freeze since we know
that this is coming from somewhere in DFG graph: this ToNumeric node itself is not newly producing this JSBigInt.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
2020-06-17 Keith Miller <keith_miller@apple.com>
$vm.haveABadTime/isHavingABadTime should work with non-globalObject parameters
https://bugs.webkit.org/show_bug.cgi?id=213304
Reviewed by Mark Lam.
Previously, $vm.haveABadTime would crash if passed a
non-globalObject object as the first parameter because it was
missing a `return` in front the error handling case. This patch
resolves that issue but also extends the semantics of
haveABadTime/isHavingABadTime to either use the global object of
the first parameter even if it's not a JSGlobalObject. If no
argument is passed, haveABadTime/isHavingABadTime instead use the
global object of the callee.
* tools/JSDollarVM.cpp:
(JSC::functionHaveABadTime):
(JSC::functionIsHavingABadTime):
2020-06-17 Mark Lam <mark.lam@apple.com>
Gardening: move some unused data inside ENABLE(JIT) to unbreak the CLoop build.
https://bugs.webkit.org/show_bug.cgi?id=213255
Not reviewed.
* assembler/testmasm.cpp:
2020-06-17 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, avoid node access in link-task
https://bugs.webkit.org/show_bug.cgi?id=213266
<rdar://problem/64453001>
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCheckJSCast):
2020-06-17 Mark Lam <mark.lam@apple.com>
Add a shiftAndAdd() emitter in AssemblyHelpers.
https://bugs.webkit.org/show_bug.cgi?id=213255
Reviewed by Michael Saboff.
void shiftAndAdd(RegisterID base, RegisterID index, uint8_t shift, RegisterID dest, Optional<RegisterID> = { });
Emits code to compute: dest = base + index << shift.
* assembler/testmasm.cpp:
(doubleOperands):
(floatOperands):
(int32Operands):
(int64Operands):
(JSC::testShiftAndAdd):
(JSC::run):
(JSC::doubleOperands): Deleted.
(JSC::floatOperands): Deleted.
(JSC::int32Operands): Deleted.
(JSC::int64Operands): Deleted.
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::shiftAndAdd):
2020-06-17 Michael Saboff <msaboff@apple.com>
[Wasm] Reduce the amount of memory used by the Air register coloring allocator
https://bugs.webkit.org/show_bug.cgi?id=212106
Reviewed by Yusuke Suzuki.
Changed InterferenceEdge to be a templated class so we can instantiate an unsigned
short version to cut memory in half for code that has less than 2^16 temps.
Through instrumentation, my testing showed that almost all compilations use the
16bit implementation. Although this change is for all B3/Air compilations at O2,
Wasm compilations are usally larger and therefore get the greatest benefit.
This allowed increasing the default value for the option webAssemblyBBQFallbackSize,
with a small increase in memory usage.
* b3/air/AirAllocateRegistersByGraphColoring.cpp:
* runtime/OptionsList.h:
2020-06-16 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Check NullSetterFunction under strict-mode context since structure / PropertyCondition are unaware of this
https://bugs.webkit.org/show_bug.cgi?id=213266
Reviewed by Mark Lam.
Our PropertyCondition is tracking the shape of Structure. This is enough for IC except for one case: throwing an error when invoking null setters in strict code.
"use strict";
var object = { get value() { return 42; } }
object.value = 42;
In the above case, we need to throw an error. Let's consider the following scenario.
1. Object has valid setter.
2. IC is buffering OPC which includes (1)'s object in [[Prototype]] hit.
3. IC commits buffered AccessCase with OPC. And PropertyCondition says Object + setter-offset => Presence.
4. Object deletes its setter.
5. Just after (4), DFG concurrently reads buffered committed OPCs.
6. DFG see that PropertyCondition is valid even after (4) since accessor property does exist.
7. Set up DFG sequence `GetSetter, Call`.
8. DFG calls null-setter under strict code, which is not assumed to be called.
In this patch, we insert NullSetterFunction check before setter invocation under strict mode. In IC, if we see NullSetterFunction,
we replace the calling target with special function which throws an error. In DFG / FTL, we emit `CheckNotJSCast` DFG node which
ensures that this setter is not null setter.
In IC code, we already have null-setter checking code before. So this change does not have any impact in terms of performance.
In DFG / FTL code, we only insert this check when we do not inline this setter. This is because inlining emits `CheckCell` anyway so
we can know that this is not NullSetterFunction. And this means that DFG Call opcode exists after CheckNotJSCast. Since Call opcode
loads the fields of call target anyway, this also does not affect on performance.
* bytecode/AccessCase.cpp:
(JSC::AccessCase::generateImpl):
* bytecode/PolymorphicAccess.cpp:
(JSC::PolymorphicAccess::regenerate):
* bytecode/PolymorphicAccess.h:
(JSC::AccessGenerationState::AccessGenerationState):
* bytecode/StructureStubInfo.cpp:
(JSC::StructureStubInfo::addAccessCase):
* bytecode/StructureStubInfo.h:
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleCall):
(JSC::DFG::ByteCodeParser::handleInlining):
(JSC::DFG::ByteCodeParser::handlePutById):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGNode.h:
(JSC::DFG::Node::hasClassInfo const):
* dfg/DFGNodeType.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileCheckJSCast):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGStructureAbstractValue.cpp:
(JSC::DFG::StructureAbstractValue::isNotSubClassOf const):
* dfg/DFGStructureAbstractValue.h:
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckJSCast):
* jit/Repatch.cpp:
(JSC::tryCacheGetBy):
(JSC::tryCacheArrayGetByVal):
(JSC::tryCachePutByID):
(JSC::tryCacheDeleteBy):
(JSC::tryCacheInByID):
(JSC::tryCacheInstanceOf):
* jit/ThunkGenerators.cpp:
(JSC::virtualThunkFor):
* runtime/InternalFunction.cpp:
(JSC::InternalFunction::finishCreation):
* runtime/JSCast.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
(JSC::JSGlobalObject::visitChildren):
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::nullSetterStrictFunction const):
* runtime/JSType.cpp:
(WTF::printInternal):
* runtime/JSType.h:
* runtime/NullSetterFunction.cpp:
(JSC::NullSetterFunctionInternal::callThrowError):
(JSC::NullSetterFunction::NullSetterFunction):
* runtime/NullSetterFunction.h:
2020-06-16 Mark Lam <mark.lam@apple.com>
Make Options::useJIT() be the canonical source of truth on whether we should use the JIT.
https://bugs.webkit.org/show_bug.cgi?id=212556
<rdar://problem/63780436>
Reviewed by Saam Barati.
After r263055, Options::useJIT() always equals VM::canUseJIT() after canUseJIT()
has been computed. This patch removes VM::canUseJIT(), and replaces all calls to
it with calls to Options::useJIT().
In the old code, VM::canUseJIT() would assert s_canUseJITIsSet to ensure that
its clients will not access s_canUseJIT before it is initialized. We not have an
equivalent mechanism with Options. This is how it works:
1. There are 2 new Options flags in the g_jscConfig:
g_jscConfig.options.isFinalized
g_jscConfig.options.allowUnfinalizedAccess
g_jscConfig.options.isFinalized means that all Options values are finalized
i.e. initialization is complete and ready to be frozen in the Config.
g_jscConfig.options.isFinalized is set by initializeThreading() by calling
Options::finalize() once options initialization is complete.
g_jscConfig.options.allowUnfinalizedAccess is an allowance for clients to
access Options values before they are finalized. This is only needed in
options initialization code where Options values are read and written to.
g_jscConfig.options.allowUnfinalizedAccess is set and cleared using the
Options::AllowUnfinalizedAccessScope RAII object. The few pieces of code that
do options initialization will instantiate this scope object.
2. All Options accessors (e.g. Option::useJIT()) will now assert that either
g_jscConfig.options.allowUnfinalizedAccess or g_jscConfig.options.isFinalized
is set.
3. Since r263055, Options::recomputeDependentOptions() ensures that if useJIT() is
false, all other JIT options (e.g. useBaselineJIT(), useDFTJIT(), useFTLJIT(),
etc.) are also false. This patch also adds useBBQJIT() and useOMGJIT() to that
list.
With this, checks for useJIT() are now redundant if there's also another JIT
option check, e.g. useRegExpJIT() or useDFGJIT(). When redundant, this patch
elides the useJIT() check (which used to be a VM::canUseJIT() check).
Ideally, we should also introduce a separate abstraction for requested option
values before finalization than the finalized option values that will be adopted
by the system. We'll do this as a separate exercise in a later patch.
* API/tests/ExecutionTimeLimitTest.cpp:
(testExecutionTimeLimit):
* API/tests/FunctionOverridesTest.cpp:
(testFunctionOverrides):
* API/tests/PingPongStackOverflowTest.cpp:
(testPingPongStackOverflow):
- Removed redundant calls to Options::initialize().
* API/tests/testapi.c:
(main):
- move the call to testExecutionTimeLimit() to after finalizeMultithreadedMultiVMExecutionTest()
returns. This is because testExecutionTimeLimit() modifies JIT options at runtime
as part of its testing. This can wreak havoc on the rest of the system that expects
the options to be frozen. Ideally, we'll find a way for testExecutionTimeLimit() to
do its work without changing JIT options, but that is not easy to do. For now,
we'll just run it at the end as a workaround.
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::setNumParameters):
* bytecode/CodeBlock.h:
(JSC::CodeBlock::numberOfArgumentValueProfiles):
(JSC::CodeBlock::valueProfileForArgument):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::isSupported):
* heap/Heap.cpp:
(JSC::Heap::completeAllJITPlans):
(JSC::Heap::iterateExecutingAndCompilingCodeBlocks):
(JSC::Heap::gatherScratchBufferRoots):
(JSC::Heap::removeDeadCompilerWorklistEntries):
(JSC::Heap::stopThePeriphery):
(JSC::Heap::suspendCompilerThreads):
(JSC::Heap::resumeCompilerThreads):
(JSC::Heap::addCoreConstraints):
* interpreter/AbstractPC.cpp:
(JSC::AbstractPC::AbstractPC):
* jit/JITThunks.cpp:
(JSC::JITThunks::ctiNativeCall):
(JSC::JITThunks::ctiNativeConstruct):
(JSC::JITThunks::ctiNativeTailCall):
(JSC::JITThunks::ctiNativeTailCallWithoutSavedTags):
(JSC::JITThunks::ctiInternalFunctionCall):
(JSC::JITThunks::ctiInternalFunctionConstruct):
(JSC::JITThunks::hostFunctionStub):
* jsc.cpp:
(CommandLine::parseArguments):
(jscmain):
* llint/LLIntEntrypoint.cpp:
(JSC::LLInt::setFunctionEntrypoint):
(JSC::LLInt::setEvalEntrypoint):
(JSC::LLInt::setProgramEntrypoint):
(JSC::LLInt::setModuleProgramEntrypoint):
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::shouldJIT):
(JSC::LLInt::jitCompileAndSetHeuristics):
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/JSCConfig.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::numberToStringWatchpointSet):
* runtime/Options.cpp:
(JSC::jitEnabledByDefault):
(JSC::disableAllJITOptions):
(JSC::Options::initialize):
- Move the calls to dumpOptionsIfNeeded() and ensureOptionsAreCoherent() to the
end after all the options have been initialized because this where they belong.
(JSC::Options::finalize):
(JSC::Options::setOptions):
(JSC::Options::setOption):
(JSC::Options::dumpAllOptions):
(JSC::Options::ensureOptionsAreCoherent):
* runtime/Options.h:
(JSC::Options::AllowUnfinalizedAccessScope::AllowUnfinalizedAccessScope):
(JSC::Options::AllowUnfinalizedAccessScope::~AllowUnfinalizedAccessScope):
* runtime/OptionsList.h:
* runtime/RegExp.cpp:
(JSC::RegExp::compile):
(JSC::RegExp::compileMatchOnly):
* runtime/SymbolTable.h:
(JSC::SymbolTableEntry::isWatchable const):
* runtime/VM.cpp:
(JSC::VM::computeCanUseJIT):
(JSC::VM::VM):
(JSC::VM::getHostFunction):
(JSC::VM::getCTIInternalFunctionTrampolineFor):
* runtime/VM.h:
(JSC::VM::isInMiniMode):
(JSC::VM::canUseJIT): Deleted.
* wasm/WasmCapabilities.h:
(JSC::Wasm::isSupported):
* wasm/WasmOperations.cpp:
(JSC::Wasm::shouldJIT):
* wasm/WasmSlowPaths.cpp:
(JSC::LLInt::shouldJIT):
2020-06-16 Robin Morisset <rmorisset@apple.com>
Optimize Air::TmpWidth analysis in IRC
https://bugs.webkit.org/show_bug.cgi?id=152478
Reviewed by Filip Pizlo.
AirTmpWidth currently uses a HashMap to map tmps to their width.
Since tmps have consecutive indices, we can instead use vectors (one for GP and one for FP tmps).
As a bonus, we can just compute the width of the tmps of the bank the register allocator is currently looking at.
This cuts the time spent in the register allocator in JetStream2 by about 100ms out of 3.4s
(or sometimes 80ms out of 2.4, the bimodality of the time spent is due to a huge function in tagcloud-SP which usually but not always reach the FTL, I'll check later if it can be fixed by tweaking the inliner).
* b3/air/AirAllocateRegistersByGraphColoring.cpp:
(JSC::B3::Air::allocateRegistersByGraphColoring):
* b3/air/AirTmpWidth.cpp:
(JSC::B3::Air::TmpWidth::TmpWidth):
(JSC::B3::Air::TmpWidth::recompute):
* b3/air/AirTmpWidth.h:
(JSC::B3::Air::TmpWidth::width const):
(JSC::B3::Air::TmpWidth::requiredWidth):
(JSC::B3::Air::TmpWidth::defWidth const):
(JSC::B3::Air::TmpWidth::useWidth const):
(JSC::B3::Air::TmpWidth::Widths::Widths):
(JSC::B3::Air::TmpWidth::widths):
(JSC::B3::Air::TmpWidth::widths const):
(JSC::B3::Air::TmpWidth::addWidths):
(JSC::B3::Air::TmpWidth::widthsVector):
2020-06-16 Fujii Hironori <Hironori.Fujii@sony.com>
[CMake][Visual Studio] CombinedDomains.json is generated twice in JavaScriptCore.vcxproj and InspectorBackendCommands.vcxproj
https://bugs.webkit.org/show_bug.cgi?id=213225
Reviewed by Don Olmstead.
Since r262203 (Bug 210014) added a new target
InspectorBackendCommands, CombinedDomains.json is generated twice
in JavaScriptCore.vcxproj and InspectorBackendCommands.vcxproj.
This caused unnecessary incremental builds.
The fundamental issue of this issue was fixed in CMake side.
<https://gitlab.kitware.com/cmake/cmake/issues/16767>
However, JavaScriptCore target needs to have a direct or indirect
dependency of InspectorBackendCommands target for CMake Visual
Studio generator to eliminate duplicated custom commands.
* CMakeLists.txt: Added add_dependencies(JavaScriptCore InspectorBackendCommands).
2020-06-16 Mark Lam <mark.lam@apple.com>
Add SIGABRT handler for non OS(DARWIN) builds to the jsc shell with the -s option.
https://bugs.webkit.org/show_bug.cgi?id=213200
Reviewed by Michael Catanzaro.
This is needed because non OS(DARWIN) builds uses abort as their "CRASH"ing
mechanism.
* jsc.cpp:
(CommandLine::parseArguments):
2020-06-15 Michael Catanzaro <mcatanzaro@gnome.org>
WTF signal machinery is guarded by #if USE(PTHREADS) && HAVE(MACHINE_CONTEXT) but does not use pthreads or machine context
https://bugs.webkit.org/show_bug.cgi?id=213223
Reviewed by Mark Lam.
Use #if OS(UNIX) here too. This should fix stress/ensure-crash.js when
HAVE(MACHINE_CONTEXT) is false.
* jsc.cpp:
(printUsageStatement):
(CommandLine::parseArguments):
2020-06-15 Pavel Feldman <pavel.feldman@gmail.com>
Web Inspector: introduce request interception
https://bugs.webkit.org/show_bug.cgi?id=207446
Reviewed by Devin Rousso.
This change introduces network request interception to the Network
protocol domain. It adds Network.interceptWithRequest notification that
can be continued, modified or fulfilled. NetworkStage enum can now have
'request' and 'response' values.
* inspector/protocol/Network.json:
2020-06-15 Tadeu Zagallo <tzagallo@apple.com>
op_iterator_open getNext checkpoint needs to declare it uses m_iterator
https://bugs.webkit.org/show_bug.cgi?id=213106
<rdar://problem/63416838>
Reviewed by Keith Miller.
Currently, we have no way of specifying that a checkpoint uses an operand defined at an earlier
point in the same bytecode, which is the case for op_iterator_open: we assume that it will have
already allocated the iterator and stored it in m_iterator by the time we get to the getNext
checkpoint. In order to support that, we change tmpLivenessForCheckpoint to livenessForCheckpoint
and allow it to also declare the use of the operands defined within the bytecode.
* bytecode/BytecodeLivenessAnalysis.cpp:
(JSC::livenessForCheckpoint):
(JSC::tmpLivenessForCheckpoint): Deleted.
* bytecode/BytecodeLivenessAnalysis.h:
* bytecode/FullBytecodeLiveness.h:
* dfg/DFGForAllKills.h:
(JSC::DFG::forAllKilledOperands):
* dfg/DFGGraph.cpp:
(JSC::DFG::Graph::isLiveInBytecode):
* dfg/DFGGraph.h:
2020-06-15 Alexey Shvayka <shvaikalesh@gmail.com>
Expand JSObject::defineOwnIndexedProperty() fast path for existing properties
https://bugs.webkit.org/show_bug.cgi?id=213133
Reviewed by Yusuke Suzuki.
This patch expands fast path of JSObject::defineOwnIndexedProperty() to cover existing properties
if given data descriptor has no falsy attributes, preventing the object from entering SparseMode.
The optimization is possible due to this invariant: indexed properties of non-SparseMode objects
have attributes of PropertyAttribute::None (except for typed arrays; added assert covers it).
PropertyDescriptor::attributesOverridingCurrent() with PropertyAttribute::None descriptor
is used to support partial descriptors like {value: 1, writable: true}.
This change advances Object.defineProperty microbenchmark by 35%; array read/write benchmark
following property redefinition is progressed by a factor of 16 due to avoiding SparseMode.
* runtime/JSObject.cpp:
(JSC::JSObject::defineOwnIndexedProperty):
2020-06-15 Robin Morisset <rmorisset@apple.com>
testB3::testReportUsedRegistersLateUseFollowedByEarlyDefDoesNotMarkUseAsDead() has a validation failure in debug mode
https://bugs.webkit.org/show_bug.cgi?id=196103
<rdar://problem/57808549>
Reviewed by Keith Miller.
The problem was trivial: patchpoints were referring to constants that were defined after them.
Just exchanging the order of the definition was enough to make this test pass.
* b3/testb3_1.cpp:
(shouldRun):
* b3/testb3_7.cpp:
(testReportUsedRegistersLateUseFollowedByEarlyDefDoesNotMarkUseAsDead):
2020-06-15 Mark Lam <mark.lam@apple.com>
Do not install the VMTraps signal handler if Options::useJIT=false.
https://bugs.webkit.org/show_bug.cgi?id=212543
<rdar://problem/63772519>
Reviewed by Keith Miller.
VMTraps is only needed for JITted code. Hence, if the JIT is disabled, we should
set Options::usePollingTraps() to true to indicate that we won't be using VMTraps.
With this change, we no longer install any signal handling machinery if
Options::useJIT() is false.
Because we may still disable the JIT even if useJIT() is true (due to failure to
allocate JIT memory or a number of other factors), we will also add a check of
VM::canUseJIT() in initializeThreading(), and disable useJIT() if needed. Of
course, this also means we need to call Options::recomputeDependentOptions() to
make other options consistent with useJIT() being false.
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/Options.cpp:
(JSC::disableAllJITOptions):
(JSC::Options::recomputeDependentOptions):
(JSC::recomputeDependentOptions): Deleted.
* runtime/Options.h:
* runtime/VMTraps.cpp:
(JSC::VMTraps::initializeSignals):
* tools/SigillCrashAnalyzer.cpp:
(JSC::SigillCrashAnalyzer::instance):
2020-06-15 Keith Miller <keith_miller@apple.com>
CheckIsConstant should not use BadCache exit kind
https://bugs.webkit.org/show_bug.cgi?id=213141
Reviewed by Yusuke Suzuki.
The BadCache exit kind causes the OSR exit compilers to try to
update ArrayProfiles. This is just incorrect for CheckIsConstant
since the node's origin may not even have an
ArrayProfile... BadCache also strongly assumes the value it's
profiling is a cell, which is clearly not always the case for
CheckIsConstant.
CheckIsConstant now uses the BadConstantValue (BadValue conflicts
with macros exported by X11 on GTK) exit kind for all use kinds,
which is just a rename of BadCell. All existing places where we
can emit a CheckIsConstant already have a story for BadConstantValue.
* bytecode/CallLinkStatus.cpp:
(JSC::CallLinkStatus::computeFromLLInt):
(JSC::CallLinkStatus::computeExitSiteData):
* bytecode/ExitKind.cpp:
(JSC::exitKindToString):
* bytecode/ExitKind.h:
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleInlining):
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
(JSC::DFG::ByteCodeParser::handleModuleNamespaceLoad):
(JSC::DFG::ByteCodeParser::parseBlock):
(JSC::DFG::ByteCodeParser::handlePutByVal):
(JSC::DFG::ByteCodeParser::handleCreateInternalFieldObject):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGNode.h:
(JSC::DFG::Node::isPseudoTerminal):
* dfg/DFGNodeType.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileCheckIsConstant):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckIsConstant):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckBadValue):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckBadCell): Deleted.
2020-06-15 Yusuke Suzuki <ysuzuki@apple.com>
Webkit Feature BigInt on webkit.org
https://bugs.webkit.org/show_bug.cgi?id=197546
Reviewed by Sam Weinig.
Add BigInt entry to JSC features.json.
* features.json:
2020-06-15 Keith Miller <keith_miller@apple.com>
JIT thunks should work on arm64_32
https://bugs.webkit.org/show_bug.cgi?id=213103
Reviewed by Saam Barati.
This patch fixes various issues when running JSC on arm64_32 with
useJIT=1 and useBaselineJIT=0. In particular this patch makes the
following changes:
1) ScalePtr is now just part of the Scale enum and is set based on
the size of the address space.
2) MacroAssembler::*Ptr functions call 32/64 bit variants based on
Address space size rather than cpu architecture. Vetting of callsites
using Ptr as 64 will happen in future patches since it's hard to
comprehensively vet.
3) Add some missing variants of functions for when pointers are 32-bit.
4) Add a load/storeReg function that stores a full register regardless
of pointer size for storing/loading callee saves.
5) numberOfDFGCompiles should report a big number for
useBaselineJIT=0 as some tests fail by default if useBaselineJIT=0
but useJIT=1.
6) Assert BaseIndex has a scale of PtrSize or TimesOne (for pre-scaled
values) when passed to a load/storePtr function.
* assembler/AbstractMacroAssembler.h:
(JSC::AbstractMacroAssembler::timesPtr): Deleted.
* assembler/MacroAssembler.h:
(JSC::MacroAssembler::rotateRightPtr):
(JSC::MacroAssembler::loadPtr):
(JSC::MacroAssembler::loadPtrWithCompactAddressOffsetPatch):
(JSC::MacroAssembler::branchPtr):
(JSC::MacroAssembler::storePtr):
(JSC::MacroAssembler::shouldBlindDouble):
(JSC::MacroAssembler::moveDouble):
(JSC::MacroAssembler::store64):
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::add32):
(JSC::MacroAssemblerARM64::signExtend32ToPtr):
(JSC::MacroAssemblerARM64::loadPtr):
(JSC::MacroAssemblerARM64::call):
(JSC::MacroAssemblerARM64::farJump):
* assembler/MacroAssemblerARMv7.h:
(JSC::MacroAssemblerARMv7::rotateRight32):
* assembler/MacroAssemblerMIPS.h:
(JSC::MacroAssemblerMIPS::rotateRight32):
* assembler/MacroAssemblerX86.h:
* assembler/MacroAssemblerX86_64.h:
* b3/B3LowerMacros.cpp:
* b3/testb3_6.cpp:
(testInterpreter):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::emitSwitchIntJump):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitLoadStructure):
(JSC::AssemblyHelpers::emitAllocateVariableSized):
(JSC::AssemblyHelpers::restoreCalleeSavesFromEntryFrameCalleeSavesBuffer):
(JSC::AssemblyHelpers::copyCalleeSavesToEntryFrameCalleeSavesBufferImpl):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::storeReg):
(JSC::AssemblyHelpers::loadReg):
(JSC::AssemblyHelpers::emitMaterializeTagCheckRegisters):
(JSC::AssemblyHelpers::emitGetFromCallFrameHeaderPtr):
(JSC::AssemblyHelpers::emitGetFromCallFrameHeader32):
(JSC::AssemblyHelpers::emitGetFromCallFrameHeader64):
(JSC::AssemblyHelpers::emitPutToCallFrameHeader):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::emit_op_enumerator_structure_pname):
(JSC::JIT::emit_op_enumerator_generic_pname):
* jit/ThunkGenerators.cpp:
(JSC::nativeForGenerator):
* runtime/TestRunnerUtils.cpp:
(JSC::numberOfDFGCompiles):
2020-06-15 Caitlin Potter <caitp@igalia.com>
[JSC] add machinery to disable JIT tiers when experimental features are enabled
https://bugs.webkit.org/show_bug.cgi?id=213193
Reviewed by Mark Lam.
A new macro FOR_EACH_JSC_EXPERIMENTAL_OPTION() supplies flags indicating the supported
JIT tiers (or, in the future, other options) of a particular feature,
in an easy to understand format. These flags are then used to
recompute dependent feature flags.
This should simplify the incremental development of language features.
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* runtime/Options.cpp:
(JSC::recomputeDependentOptions):
* runtime/OptionsList.h:
2020-06-15 Keith Miller <keith_miller@apple.com>
Signal handlers should have a two phase installation.
https://bugs.webkit.org/show_bug.cgi?id=213160
Reviewed by Mark Lam.
* jsc.cpp:
(CommandLine::parseArguments):
(jscmain):
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/VMTraps.cpp:
* tools/SigillCrashAnalyzer.cpp:
(JSC::installCrashHandler):
* wasm/WasmFaultSignalHandler.cpp:
(JSC::Wasm::enableFastMemory):
(JSC::Wasm::prepareFastMemory):
* wasm/WasmFaultSignalHandler.h:
2020-06-15 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, fix LLInt
https://bugs.webkit.org/show_bug.cgi?id=157972
loadi only takes address.
* llint/LowLevelInterpreter64.asm:
2020-06-15 Alexey Shvayka <shvaikalesh@gmail.com>
super should not depend on __proto__
https://bugs.webkit.org/show_bug.cgi?id=157972
Reviewed by Saam Barati.
Before this change, both super() call [1] and super.property [2] relied on
Object.prototype.__proto__ to acquire super base, which was observable and
incorrect if __proto__ gets removed.
This patch introduces get_prototype_of bytecode, ensuring returned values
are profiled so the op can be wired to existing DFG and FTL implementations.
In order to avoid performance regression w/o DFG (__proto__ is optimized via
IntrinsicGetterAccessCase), fast paths for LLInt and baseline JIT are added
(64-bit only), utilizing OverridesGetPrototypeOutOfLine type info flag.
This change aligns JSC with V8 and SpiderMonkey, progressing microbenchmarks/
super-get-by-{id,val}-with-this-monomorphic.js by 7-10%. SixSpeed is neutral.
Also, extracts JSValue::getPrototype() method to avoid code duplication and
utilizes it in objectConstructorGetPrototypeOf(), advancing provided
microbenchmark by 40%.
[1]: https://tc39.es/ecma262/#sec-getsuperconstructor (step 5)
[2]: https://tc39.es/ecma262/#sec-getsuperbase (step 5)
* builtins/BuiltinNames.h:
* bytecode/BytecodeIntrinsicRegistry.h:
* bytecode/BytecodeList.rb:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::finishCreation):
* bytecode/Opcode.h:
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitGetPrototypeOf):
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::emitSuperBaseForCallee):
(JSC::emitGetSuperFunctionForConstruct):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_getPrototypeOf):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicGetter):
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGOperations.cpp:
* jit/IntrinsicEmitter.cpp:
(JSC::IntrinsicGetterAccessCase::canEmitIntrinsicGetter):
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):
* jit/JIT.h:
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_get_prototype_of):
* llint/LowLevelInterpreter.asm:
* llint/LowLevelInterpreter64.asm:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/CommonSlowPaths.h:
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::getPrototype const):
* runtime/JSGlobalObjectFunctions.cpp:
(JSC::globalFuncProtoGetter):
* runtime/JSObject.cpp:
(JSC::JSObject::calculatedClassName):
* runtime/JSObject.h:
(JSC::JSObject::getPrototype):
* runtime/JSObjectInlines.h:
(JSC::JSObject::canPerformFastPutInlineExcludingProto):
(JSC::JSObject::getPropertySlot):
(JSC::JSObject::getNonIndexPropertySlot):
* runtime/JSProxy.h:
* runtime/JSTypeInfo.h:
(JSC::TypeInfo::overridesGetPrototype const):
* runtime/ObjectConstructor.cpp:
(JSC::objectConstructorGetPrototypeOf):
* runtime/ProxyObject.h:
* runtime/Structure.h:
* runtime/Structure.cpp:
(JSC::Structure::validateFlags):
2020-06-13 Devin Rousso <drousso@apple.com>
Make `errors` an own property of `AggregateError` instead of a prototype accessor
https://bugs.webkit.org/show_bug.cgi?id=212677
Reviewed by Yusuke Suzuki.
* runtime/AggregateError.h:
(JSC::AggregateError::destroy): Deleted.
(JSC::AggregateError::subspaceFor): Deleted.
(JSC::AggregateError::errors): Deleted.
* runtime/AggregateError.cpp:
(JSC::AggregateError::AggregateError):
(JSC::AggregateError::finishCreation): Added.
(JSC::AggregateError::visitChildren): Deleted.
* runtime/AggregateErrorPrototype.h:
* runtime/AggregateErrorPrototype.cpp:
(JSC::AggregateErrorPrototype::finishCreation):
(JSC::aggregateErrorPrototypeAccessorErrors): Deleted.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::initializeAggregateErrorConstructor):
* runtime/VM.h:
* runtime/VM.cpp:
(JSC::VM::VM):
* heap/Heap.cpp:
(JSC::Heap::finalizeUnconditionalFinalizers):
Remove `aggregateErrorSpace` since `AggregateError` doesn't add any new member variables.
Ensure that it can share an `IsoSubspace` with `ErrorInstance`.
* runtime/CommonIdentifiers.h:
Add `errors`.
2020-06-12 Robin Morisset <rmorisset@apple.com>
The ||= operator (and similar ones) should produce valid bytecode even if the right side is a static error
https://bugs.webkit.org/show_bug.cgi?id=213154
Reviewed by Devin Rousso.
There were two minor issues here that interacted:
- emitThrowReferenceError did not take an optional `dst` argument like everything else, and instead always returned a new temporary.
As a result, the various functions that sometimes did "return emitThrowReferenceError(..);" could return a different RegisterID than the one
provided to them through `dst`, breaking the invariant stated at the top of the file.
- ShortCircuitReadModifyResolveNode::emitBytecode used the result of such a function, unnecessarily, and (correctly) relied on the invariant being upheld.
The combination of these led to the bytecode trying to do a move of a temporary that was only defined in one of the predecessors of the basic block it was on,
which was caught by validateBytecode.
I fixed both issues, and verified that either fix is enough to stop the bug.
I fixed the first because other code may depend on that invariant in more subtle ways.
I fixed the second because it was just unnecessary complexity and made the code misleading.
I also reworded the comment at the top of NodesCodegen.cpp based on Keith's explanation and Mark's advice to make it less cryptic.
* bytecompiler/NodesCodegen.cpp:
(JSC::ThrowableExpressionData::emitThrowReferenceError):
(JSC::PostfixNode::emitBytecode):
(JSC::DeleteBracketNode::emitBytecode):
(JSC::DeleteDotNode::emitBytecode):
(JSC::PrefixNode::emitBytecode):
(JSC::ShortCircuitReadModifyResolveNode::emitBytecode):
(JSC::AssignErrorNode::emitBytecode):
* parser/Nodes.h:
2020-06-12 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] el(Greek) characters' upper-case conversion is locale-sensitive
https://bugs.webkit.org/show_bug.cgi?id=213155
<rdar://problem/55018467>
Reviewed by Darin Adler.
CLDR defines 4 locales which has language-sensitive case conversions. "az", "el", "lt", and "tr", where,
az = Azerbaijani
el = Greek
lt = Lithuanian
tr = Turkish
We can ensure it easily like this.
1. Download CLDR data
2. `ls common/transforms/*Upper.xml`
common/transforms/az-Upper.xml
common/transforms/el-Upper.xml
common/transforms/lt-Upper.xml
common/transforms/tr-Upper.xml
And ECMA-402 String.prototype.{toLocaleLowerCase,toLocaleUpperCase} requires these locales are listed as `availableLocales`.
> 7. Let availableLocales be a List with language tags that includes the languages for which the Unicode Character
> Database contains language sensitive case mappings. Implementations may add additional language tags if they
> support case mapping for additional locales.
https://tc39.es/ecma402/#sup-string.prototype.tolocalelowercase
This patch adds "el" to our maintained availableLocales list. Previously we only had "az", "lt", and "tr".
* runtime/StringPrototype.cpp:
(JSC::toLocaleCase):
(JSC::stringProtoFuncToLocaleUpperCase):
2020-06-12 Keith Miller <keith_miller@apple.com>
Tests expecting a crash should use a signal handler in the JSC CLI process
https://bugs.webkit.org/show_bug.cgi?id=212479
Reviewed by Yusuke Suzuki.
Have the -s option use WTF::Signals and make sure it adds breakpoint catching
as well.
* jsc.cpp:
(printUsageStatement):
(CommandLine::parseArguments):
* tools/SigillCrashAnalyzer.cpp:
(JSC::installCrashHandler):
2020-06-12 Alexey Shvayka <shvaikalesh@gmail.com>
AsyncGenerator should await "return" completions
https://bugs.webkit.org/show_bug.cgi?id=212774
Reviewed by Ross Kirsling.
This patch fixes 2 spec discrepancies, observable with async generators if the
value of "return" completion is a Promise, aligning JSC with V8 and SpiderMonkey.
* builtins/AsyncGeneratorPrototype.js:
(onFulfilled):
This change implements step 8 of AsyncGeneratorYield [1], that is executed after
step 15 of AsyncGeneratorResumeNext [2] (implemented as @doAsyncGeneratorBodyCall).
We are safe to rely on [[AsyncGeneratorState]] being "suspendedYield" (set in
step 6 of AsyncGeneratorYield [1]) instead of adding extra field to AsyncGenerator:
AsyncGeneratorResumeNext [2] does not overwrite "suspendedYield" state.
This change fixes most of test262 cases.
[1]: https://tc39.es/ecma262/#sec-asyncgeneratoryield
[2]: https://tc39.es/ecma262/#sec-asyncgeneratorresumenext
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitDelegateYield):
This change implements step 7.c.iii.1 of yield* runtime semantics [3], that is
observable only if [[Value]] has userland "then" getter. Awaited result is discarded.
This change fixes async-generator/yield-star-return-then-getter-ticks.js test262 case.
[3]: https://tc39.es/ecma262/#sec-generator-function-definitions-runtime-semantics-evaluation
2020-06-12 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed, address Darin's feedback on r262890.
* runtime/IntlObject.cpp:
(JSC::addScriptlessLocaleIfNeeded):
Use != instead of < for clarity.
2020-06-12 Adrian Perez de Castro <aperez@igalia.com>
Build is broken with EVENT_LOOP_TYPE=GLib
https://bugs.webkit.org/show_bug.cgi?id=212987
Reviewed by Konstantin Tokarev.
* PlatformJSCOnly.cmake: Add sources needed to support the remote inspector to
JavaScriptCore_SOURCES.
2020-06-11 Saam Barati <sbarati@apple.com>
Linear Scan uses the wrong Interval for spills for tmps with roles of early def or late use
https://bugs.webkit.org/show_bug.cgi?id=213055
<rdar://problem/59874018>
Reviewed by Yusuke Suzuki.
There was a bug in linear scan when computing the live range interval for
spill tmps that had early defs or late uses. When linear scan spills a
tmp, it creates a new tmp that it loads to and stores from, and replaces the old tmp
with the new tmp, and emits stores/loads around pertinent instructions. The live
interval for such tmps is small by nature, it's contained in the interval for the
instruction itself. However, we'd build this interval purely based off the
original tmp's arg timing. So, for example, let's consider a program like this:
RandoInsn: LateUse:Tmp1, Use:Tmp2, [early = N, late = N+1]
Let's say that Tmp1's last use is RandoInsn, and it had a def before
RandoInsn, therefore, its live range will be something like:
[J where J < N, N+1]
and now imagine we spilled Tmp1 for some reason, and rewrote the
program to be:
Move Addr(spill for Tmp1), TmpSpill
RandoInsn: LateUse:TmpSpill, Use:Tmp2, [early = N, late = N+1]
We used to incorrectly mark the live range for TmpSpill to just be [N+1, N+2).
However, the bug here is that we neglected that TmpSpill actually had an earlier
def at [N, N+1). So, the live range for TmpSpill was wrong. This could incorrectly
lead us to allocate Tmp2 and TmpSpill to the same register, since their live
ranges may not intersect if Tmp2 dies at RandoInsn.
We also had the symmetric bug for EarlyDefs: we wouldn't account for the
store-spill that'd happen after something like RandoInsn.
The fix is to account for the loads/stores of spill tmps when assigning
them a live range.
This patch contains a standalone test in testair. It also fixes crashes we had when
running B3O1 tests using typed arrays on arm64e since we had patchpoints that utilized
LateUse for signing and auth.
* b3/B3Procedure.h:
* b3/air/AirAllocateRegistersAndStackByLinearScan.cpp:
* b3/air/testair.cpp:
2020-06-11 Saam Barati <sbarati@apple.com>
Replace uses of black/white list with block/allow list
https://bugs.webkit.org/show_bug.cgi?id=213084
Reviewed by Keith Miller.
We should be using racially neutral names in our code. From Chromium style guide:
"Terms such as 'blacklist' and 'whitelist' reinforce the notion that
black==bad and white==good."
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* b3/air/AirLowerAfterRegAlloc.cpp:
(JSC::B3::Air::lowerAfterRegAlloc):
* dfg/DFGDriver.cpp:
(JSC::DFG::ensureGlobalDFGAllowlist):
(JSC::DFG::compileImpl):
(JSC::DFG::ensureGlobalDFGWhitelist): Deleted.
* dfg/DFGTierUpCheckInjectionPhase.cpp:
(JSC::DFG::ensureGlobalFTLAllowlist):
(JSC::DFG::TierUpCheckInjectionPhase::run):
(JSC::DFG::ensureGlobalFTLWhitelist): Deleted.
* heap/MachineStackMarker.cpp:
* inspector/scripts/codegen/objc_generator.py:
(ObjCGenerator.should_generate_types_for_domain):
(ObjCGenerator.should_generate_commands_for_domain):
(ObjCGenerator.should_generate_events_for_domain):
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::ensureGlobalJITAllowlist):
(JSC::LLInt::shouldJIT):
(JSC::LLInt::ensureGlobalJITWhitelist): Deleted.
* runtime/OptionsList.h:
* tools/FunctionAllowlist.cpp: Copied from Source/JavaScriptCore/tools/FunctionWhitelist.cpp.
(JSC::FunctionAllowlist::FunctionAllowlist):
(JSC::FunctionAllowlist::contains const):
(JSC::FunctionWhitelist::FunctionWhitelist): Deleted.
(JSC::FunctionWhitelist::contains const): Deleted.
* tools/FunctionAllowlist.h: Copied from Source/JavaScriptCore/tools/FunctionWhitelist.h.
* tools/FunctionWhitelist.cpp: Removed.
* tools/FunctionWhitelist.h: Removed.
2020-06-11 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Return DisposableCallSiteIndex when destroying GCAwareJITStubRoutineWithExceptionHandler
https://bugs.webkit.org/show_bug.cgi?id=213069
<rdar://problem/64205186>
Reviewed by Saam Barati.
Inside GCAwareJITStubRoutineWithExceptionHandler::observeZeroRefCount, we are returning DisposableCallSiteIndex to freelist.
However, GCAwareJITStubRoutineWithExceptionHandler::observeZeroRefCount can be called even if the code of GCAwareJITStubRoutineWithExceptionHandler is
on the stack. Let's consider the following scenario.
1. Execute GCAwareJITStubRoutineWithExceptionHandler's code. Set CallSiteIndex to the stack.
2. Execute more code. (1)'s GCAwareJITStubRoutineWithExceptionHandler's code is on the stack.
3. (1)'s GCAwareJITStubRoutineWithExceptionHandler's refcount becomes zero.
4. CallSiteIndex of GCAwareJITStubRoutineWithExceptionHandler is returned.
5. Execute StackVisitor to construct frames. But we cannot find CodeOrigin corresponding to CallSiteIndex stored in (1) since it is already returned.
DisposableCallSiteIndex should be returned after ensuring that GCAwareJITStubRoutineWithExceptionHandler's code is not on the stack. Detecting this is the functionality
what GCAwareJITStubRoutineWithExceptionHandler can offer. It is destroyed after ensuring that GCAwareJITStubRoutineWithExceptionHandler's code is not on the stack.
This patch delays DisposableCallSiteIndex returning until we destroy owner GCAwareJITStubRoutineWithExceptionHandler. But it is possible that CodeBlock* corresponding to
GCAwareJITStubRoutineWithExceptionHandler is already destroyed. To avoid this condition, we extract CodeOrigins vector as Ref<DFG::CodeOriginPool> and keep it alive from
GCAwareJITStubRoutineWithExceptionHandler too. And since CodeOrigin addition / removal happens only from the main thread after finishing the compilation, and
GCAwareJITStubRoutineWithExceptionHandler's destructor is called from the Heap's finalizer, which must be executed from the main thread, we can just modify it without a lock.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::newExceptionHandlingCallSiteIndex):
(JSC::CodeBlock::codeOrigins):
* bytecode/CodeBlock.h:
(JSC::CodeBlock::codeOrigin):
* dfg/DFGCodeOriginPool.cpp: Added.
(JSC::DFG::CodeOriginPool::addCodeOrigin):
(JSC::DFG::CodeOriginPool::addUniqueCallSiteIndex):
(JSC::DFG::CodeOriginPool::lastCallSite const):
(JSC::DFG::CodeOriginPool::addDisposableCallSiteIndex):
(JSC::DFG::CodeOriginPool::removeDisposableCallSiteIndex):
(JSC::DFG::CodeOriginPool::shrinkToFit):
* dfg/DFGCodeOriginPool.h: Added.
(JSC::DFG::CodeOriginPool::create):
(JSC::DFG::CodeOriginPool::get):
(JSC::DFG::CodeOriginPool::size const):
* dfg/DFGCommonData.cpp:
(JSC::DFG::CommonData::shrinkToFit):
(JSC::DFG::CommonData::addCodeOrigin): Deleted.
(JSC::DFG::CommonData::addUniqueCallSiteIndex): Deleted.
(JSC::DFG::CommonData::lastCallSite const): Deleted.
(JSC::DFG::CommonData::addDisposableCallSiteIndex): Deleted.
(JSC::DFG::CommonData::removeDisposableCallSiteIndex): Deleted.
* dfg/DFGCommonData.h:
(JSC::DFG::CommonData::CommonData):
* dfg/DFGJITCompiler.cpp:
(JSC::DFG::JITCompiler::exceptionCheck):
* dfg/DFGJITCompiler.h:
(JSC::DFG::JITCompiler::addCallSite):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compilePutById):
(JSC::FTL::DFG::LowerDFGToB3::compileGetByVal):
(JSC::FTL::DFG::LowerDFGToB3::compileDelBy):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstruct):
(JSC::FTL::DFG::LowerDFGToB3::compileDirectCallOrConstruct):
(JSC::FTL::DFG::LowerDFGToB3::compileTailCall):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargsSpread):
(JSC::FTL::DFG::LowerDFGToB3::compileCallOrConstructVarargs):
(JSC::FTL::DFG::LowerDFGToB3::compileCallEval):
(JSC::FTL::DFG::LowerDFGToB3::compileInById):
(JSC::FTL::DFG::LowerDFGToB3::compileInstanceOf):
(JSC::FTL::DFG::LowerDFGToB3::compileLogShadowChickenTail):
(JSC::FTL::DFG::LowerDFGToB3::getById):
(JSC::FTL::DFG::LowerDFGToB3::getByIdWithThis):
(JSC::FTL::DFG::LowerDFGToB3::lazySlowPath):
(JSC::FTL::DFG::LowerDFGToB3::callPreflight):
* ftl/FTLSlowPathCall.cpp:
(JSC::FTL::callSiteIndexForCodeOrigin):
* jit/GCAwareJITStubRoutine.cpp:
(JSC::GCAwareJITStubRoutineWithExceptionHandler::GCAwareJITStubRoutineWithExceptionHandler):
(JSC::GCAwareJITStubRoutineWithExceptionHandler::~GCAwareJITStubRoutineWithExceptionHandler):
(JSC::GCAwareJITStubRoutineWithExceptionHandler::aboutToDie):
(JSC::GCAwareJITStubRoutineWithExceptionHandler::observeZeroRefCount):
* jit/GCAwareJITStubRoutine.h:
2020-06-11 Alexey Shvayka <shvaikalesh@gmail.com>
RegExp.prototype getters should throw on cross-realm access
https://bugs.webkit.org/show_bug.cgi?id=213075
Reviewed by Saam Barati.
This patch makes RegExp.prototype getters throw TypeError when called on
RegExp.prototype object from another realm, aligning JSC with V8 and SpiderMonkey.
The spec [1] allows same-realm access to avoid breaking the web, while makes
RegExp.prototype an ordinary object (rather than RegExp instance) where possible.
[1]: https://tc39.es/ecma262/#sec-get-regexp.prototype.global (step 3.a)
* runtime/RegExpPrototype.cpp:
(JSC::regExpProtoGetterGlobal):
(JSC::regExpProtoGetterIgnoreCase):
(JSC::regExpProtoGetterMultiline):
(JSC::regExpProtoGetterDotAll):
(JSC::regExpProtoGetterSticky):
(JSC::regExpProtoGetterUnicode):
(JSC::regExpProtoGetterSource):
2020-06-11 Paulo Matos <pmatos@igalia.com>
Add missing include to JSONObject.cpp - non-unified build
https://bugs.webkit.org/show_bug.cgi?id=213073
Reviewed by Adrian Perez de Castro.
* runtime/JSONObject.cpp:
2020-06-10 Ross Kirsling <ross.kirsling@sony.com>
REGRESSION(r260697): [Intl] "missing script" locales like zh-TW are no longer mapped
https://bugs.webkit.org/show_bug.cgi?id=213007
Reviewed by Darin Adler.
addMissingScriptLocales was removed from IntlObject when changing our locale resolution to depend more directly
on ICU, but apparently even latest ICU won't perform this legacy "region implies script" mapping for us.
ICU 65+ does have uloc_openAvailableByType which will do the trick, so perhaps we should use this in the future,
but it still doesn't seem to help us with Collator, which has its own separate set of "available locales".
The exact set of locales which should be mapped is currently under discussion here:
https://github.com/tc39/ecma402/issues/159
But the crux seems to be that we should ensure we have an xx-ZZ alias for all available xx-Yyyy-ZZ locales.
* runtime/IntlObject.cpp:
(JSC::addScriptlessLocaleIfNeeded):
(JSC::intlAvailableLocales):
(JSC::intlCollatorAvailableLocales):
2020-06-10 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSCallbackObject::deleteProperty should redirect to Parent::deletePropertyByIndex if propertyName is index
https://bugs.webkit.org/show_bug.cgi?id=213041
<rdar://problem/64204300>
Reviewed by Darin Adler.
We have an infinite recursion here.
-> JSCallbackObject::deletePropertyByIndex
-> JSCell::deleteProperty
-> JSCallbackObject::deleteProperty
-> JSObject::deleteProperty
-> JSCallbackObject::deletePropertyByIndex
When propertyName in JSCallbackObject::deleteProperty is an index, we should go to JSObject::deletePropertyByIndex instead of JSObject::deleteProperty.
* API/JSCallbackObjectFunctions.h:
(JSC::JSCallbackObject<Parent>::deleteProperty):
2020-06-09 Mark Lam <mark.lam@apple.com>
Stringifier::appendStringifiedValue() should not assume it is always safe to recurse.
https://bugs.webkit.org/show_bug.cgi?id=213006
<rdar://problem/64154840>
Reviewed by Keith Miller.
In r262727, I suggested that Alexey Shvayka add an assertion in
Stringifier::appendStringifiedValue() to assert that it is safe to recurse because
we don't expect it to recurse into itself. Turns out this is a bad idea because
a client may be doing the recursing before calling Stringifier::appendStringifiedValue().
As a result, Stringifier::appendStringifiedValue() ends up being executed with
the stack pointer already in the reserved zone. This is legal, and is what the
reserved zone is intended for as long as we don't recurse from here. However,
this also means that asserting vm.isSafeToRecurseSoft() here will surely fail
because we are already in the reserved zone area. The fix is simply to remove
this faulty assertion.
* runtime/JSONObject.cpp:
(JSC::Stringifier::appendStringifiedValue):
2020-06-09 Mark Lam <mark.lam@apple.com>
Disambiguate the OverridesGetPropertyNames structure flag
https://bugs.webkit.org/show_bug.cgi?id=212909
<rdar://problem/63823557>
Reviewed by Saam Barati.
Previously, the OverridesGetPropertyNames structure flag could mean 2 different
things:
1. the getPropertyNames() method is overridden, or
2. any of the forms of getPropertyName() is overridden:
getPropertyName, getOwnPropertyNames, getOwnNonIndexPropertyNames
Some parts of the code expects one definition while other parts expect the other.
This patch disambiguates between the 2 by introducing OverridesAnyFormOfGetPropertyNames
for definition (2). OverridesGetPropertyNames now only means definition (1).
Note: we could have implemented overridesGetPropertyNames() by doing a comparison
of the getPropertyNames pointer in the MethodTable. This is a little slower than
checking a TypeInfo flag, but probably doesn't matter a lot in the code paths
where overridesGetPropertyNames() is called. However, we have bits in TypeInfo
left. So, we'll might as well use it.
This ambiguity resulted in JSObject::getPropertyNames() recursing infinitely
when it didn't think it could recurse. This is demonstrated in
JSTests/stress/unexpected-stack-overflow-below-JSObject-getPropertyNames.js as
follows:
1. The test case invokes JSObject::getPropertyNames on a JSArray.
2. In the while loop at the bottom of JSObject::getPropertynames(), we check
`if (prototype->structure(vm)->typeInfo().overridesGetPropertyNames()) {`.
3. The test overrides proto as follows:
`arg0.__proto__ = arr1` where both arg0 and arr1 are JArrays.
4. In the old code, JSArray sets OverridesGetPropertyNames but does not override
getPropertyNames(). It actually meant to set OverridesAnyFormOfGetPropertyNames
(after we disambiguated it) because JSArray overrides getOwnNonIndexPropertyNames().
5. When we get to the check at (2), we ask if the prototype overridesGetPropertyNames().
Since JSArray sets OverridesGetPropertyNames, the answer is yes / true.
JSObject::getPropertynames() then proceeds to invoke
`prototype->methodTable(vm)->getPropertyNames(prototype, globalObject, propertyNames, mode);`
But because JSArray does not actually overrides getPropertyNames(), we're
actually invoking JSObject::getPropertyNames() here. Viola! Infinite loop.
With this patch, JSArray is disambiguated to set OverridesAnyFormOfGetPropertyNames
instead of OverridesGetPropertyNames, and this infinite loop no longer exists.
This patch also made the following changes:
1. Templatized TypeInfo::isSetOnFlags1() and TypeInfo::isSetOnFlags2() so that
we can used static_asserts instead of a debug ASSERT to verify the integrity of
the flag we're checking against.
2. Added a Structure::validateFlags() called from the Structure constructor.
validateFlags() will verify the following:
a. OverridesGetOwnPropertySlot must be set in the flags if getOwnPropertySlot
is overridden in the MethodTable.
b. InterceptsGetOwnPropertySlotByIndexEvenWhenLengthIsNotZero must be set in
the flags if getOwnPropertySlotByIndex is overridden in the MethodTable.
c. HasPutPropertySecurityCheck must be set in the flags if doPutPropertySecurityCheck
is overridden in the MethodTable.
d. OverridesGetPropertyNames must be set in the flags if getPropertyNames
is overridden in the MethodTable.
e. OverridesAnyFormOfGetPropertyNames must be set in the flags if any of
getPropertyNames, getOwnPropertyNames, or getOwnNonIndexPropertyNames are
overridden in the MethodTable.
An alternate solution would be to automatically set these flags if we detect
their corresponding methods are overridden. However, this alternate solution
requires this laundry list to be checked every time a structure is constructed.
The current implementation of having the required flags already pre-determined
as a constant is more efficient in terms of performance and code space.
Also, it only takes one instantiation of the structure to verify that the flags
are valid. Since we only write JSCell / JSObject classes when we need them
and we always write tests to exercise new code (especially such classes), we're
guaranteed the flags validation will be exercised.
3. Made JSObject::getOwnPropertySlot() and JSObject::doPutPropertySecurityCheck()
not inlined when ASSERT_ENABLED. This is needed in order for Structure::validateFlags()
to do its checks using function pointer comparisons. Otherwise, the inline
functions can result in multiple instantiations of these functions. For
example, WebCore can get its own copy of JSObject::getOwnPropertySlot() and
the comparisons will think the function is overridden even when it's not.
4. Structure::validateFlags() found the following problems which are now fixed:
GetterSetter was not using its StructureFlags. As a result, it was missing the
OverridesGetOwnPropertySlot flag.
JSDataView did not define its StructureFlags. It was missing the
OverridesGetOwnPropertySlot and OverridesAnyFormOfGetPropertyNames flags.
5. Changed a TypeInfo constructor to not have a default argument for the flags value.
Also grepped for all uses of this constructor to make sure that it is passed
the StructureFlags field. This exercise found the following issue:
JSAPIValueWrapper was not using its StructureFlags when creating its structure.
Previously, it was just ignoring the StructureIsImmortal flag in StructureFlags.
6. Hardened the assertions for hasReadOnlyOrGetterSetterPropertiesExcludingProto()
and hasGetterSetterProperties() in the Structure constructor.
Previously, if the flag is set, it verifies that the ClassInfo has the
appropriate data expected by the flag. However, it does not assert the reverse
i.e. that if the ClassInfo data exists, then the flag must also be set.
The new assertions now checks both.
Moved the overridesGetCallData() assertion into Structure::validateFlags()
because it concerns the OverridesGetCallData flag. This assertion has also
ben hardened.
* API/JSAPIValueWrapper.h:
* API/JSCallbackObject.h:
* debugger/DebuggerScope.h:
* inspector/JSInjectedScriptHostPrototype.h:
* inspector/JSJavaScriptCallFramePrototype.h:
* runtime/ClonedArguments.h:
* runtime/ErrorInstance.h:
* runtime/GenericArguments.h:
* runtime/GetterSetter.h:
* runtime/JSArray.h:
* runtime/JSDataView.h:
* runtime/JSFunction.h:
* runtime/JSGenericTypedArrayView.h:
* runtime/JSGlobalObject.h:
* runtime/JSLexicalEnvironment.h:
* runtime/JSModuleEnvironment.h:
* runtime/JSModuleNamespaceObject.h:
* runtime/JSObject.cpp:
(JSC::JSObject::doPutPropertySecurityCheck):
(JSC::JSObject::getOwnPropertySlot):
* runtime/JSObject.h:
(JSC::JSObject::getOwnPropertySlotImpl):
(JSC::JSObject::getOwnPropertySlot):
* runtime/JSProxy.h:
* runtime/JSString.h:
* runtime/JSSymbolTableObject.h:
* runtime/JSTypeInfo.h:
(JSC::TypeInfo::TypeInfo):
(JSC::TypeInfo::masqueradesAsUndefined const):
(JSC::TypeInfo::implementsHasInstance const):
(JSC::TypeInfo::implementsDefaultHasInstance const):
(JSC::TypeInfo::overridesGetCallData const):
(JSC::TypeInfo::overridesToThis const):
(JSC::TypeInfo::structureIsImmortal const):
(JSC::TypeInfo::overridesGetPropertyNames const):
(JSC::TypeInfo::overridesAnyFormOfGetPropertyNames const):
(JSC::TypeInfo::prohibitsPropertyCaching const):
(JSC::TypeInfo::getOwnPropertySlotIsImpure const):
(JSC::TypeInfo::getOwnPropertySlotIsImpureForPropertyAbsence const):
(JSC::TypeInfo::hasPutPropertySecurityCheck const):
(JSC::TypeInfo::newImpurePropertyFiresWatchpoints const):
(JSC::TypeInfo::isImmutablePrototypeExoticObject const):
(JSC::TypeInfo::interceptsGetOwnPropertySlotByIndexEvenWhenLengthIsNotZero const):
(JSC::TypeInfo::isSetOnFlags1 const):
(JSC::TypeInfo::isSetOnFlags2 const):
* runtime/ObjectConstructor.cpp:
(JSC::objectConstructorAssign):
* runtime/ProxyObject.h:
* runtime/RegExpObject.h:
* runtime/StringObject.h:
* runtime/Structure.cpp:
(JSC::Structure::validateFlags):
(JSC::Structure::Structure):
* runtime/Structure.h:
* runtime/StructureInlines.h:
(JSC::Structure::canCacheOwnKeys const):
* tools/JSDollarVM.cpp:
2020-06-09 Jonathan Bedard <jbedard@apple.com>
JavaScriptCore: Support tvOS and watchOS builds with the public SDK
https://bugs.webkit.org/show_bug.cgi?id=212788
<rdar://problem/64000087>
Reviewed by Tim Horton.
* Configurations/Base.xcconfig: Link to tvOS and watchOS framework stubs.
* Configurations/JavaScriptCore.xcconfig: Use iOS flags for all embedded platforms.
2020-06-09 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Shrink __DATA,(__data,__bss,__common) more
https://bugs.webkit.org/show_bug.cgi?id=212863
Reviewed by Sam Weinig.
1. Use `unsigned` instead of `size_t` in GC size-class array. We know that this number never exceeds largeCutoff,
which must be much maller than UINT32_MAX.
2. Add missing const to various variables to put them DATA,__const instead of DATA,__data etc.
* heap/MarkedSpace.cpp:
(JSC::MarkedSpace::initializeSizeClassForStepSize):
* heap/MarkedSpace.h:
* heap/VisitRaceKey.cpp:
* heap/VisitRaceKey.h:
* inspector/agents/InspectorDebuggerAgent.cpp:
* inspector/agents/InspectorDebuggerAgent.h:
* runtime/PropertyDescriptor.cpp:
* runtime/PropertyDescriptor.h:
2020-06-08 Keith Miller <keith_miller@apple.com>
Removed unneeded POINTER_WIDTH macro from b3
https://bugs.webkit.org/show_bug.cgi?id=212927
Reviewed by Yusuke Suzuki.
C++20 has real constexpr functions so we don't need the
POINTER_WIDTH macro anymore.
* b3/B3Width.h:
(JSC::B3::pointerWidth):
* b3/air/opcode_generator.rb:
2020-06-08 Alexey Shvayka <shvaikalesh@gmail.com>
JSON.stringify should throw stack overflow error
https://bugs.webkit.org/show_bug.cgi?id=143511
Reviewed by Ross Kirsling and Mark Lam.
This change adds m_holderStack.size() check, reusing the limit of JSON.parse,
and throws StackOverflowError if exceeded, aligning JSC with V8 and SpiderMonkey.
Even with all the cyclic structure checks in place, excess is possible due to
very deeply nested object, user-provided "toJSON" method or functional replacer.
While Stringifier::appendStringifiedValue() and Holder::appendNextProperty()
mutually call each other, recursion is avoided by !holderStackWasEmpty check and
do/while loop at the end of appendStringifiedValue(), as well as cyclic structure
check as per spec [1].
[1]: https://tc39.es/ecma262/#sec-serializejsonobject (step 1)
* runtime/JSONObject.cpp:
(JSC::Stringifier::appendStringifiedValue):
(JSC::Walker::walk):
2020-06-08 Jonathan Bedard <jbedard@apple.com>
JavaScriptCore: Fix PLATFORM(TVOS) macro
https://bugs.webkit.org/show_bug.cgi?id=212900
<rdar://problem/64118879>
Unreviewed build fix.
* tools/JSDollarVM.cpp:
(JSC::functionIsMemoryLimited): PLATFORM(TVOS) should be PLATFORM(APPLETV).
2020-06-07 Philippe Normand <pnormand@igalia.com>
Remove ENABLE_VIDEO_TRACK ifdef guards
https://bugs.webkit.org/show_bug.cgi?id=212568
Reviewed by Youenn Fablet.
* Configurations/FeatureDefines.xcconfig: Remove ENABLE_VIDEO_TRACK, which is now enabled by
default under the ENABLE_VIDEO guard.
2020-06-07 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Checksum for generated files should be emitted at the end of the files
https://bugs.webkit.org/show_bug.cgi?id=212875
Reviewed by Mark Lam.
If the offlineasm file generation is interrupted in the middle of the generation, it already emitted checksum.
So next file generation can accept this broken file as a result of offlineasm and skip file generation.
We should emit checksum at the end of files. For now, this patch takes a quick way: just iterating lines, getting
a last line and use it for checksum comparison.
* generator/GeneratedFile.rb:
* offlineasm/asm.rb:
2020-06-06 Mark Lam <mark.lam@apple.com>
Make CodeBlockHash robust against unreasonably long source code.
https://bugs.webkit.org/show_bug.cgi?id=212847
<rdar://problem/64024279>
Reviewed by Saam Barati.
This patch adds a heuristic to avoid trying to compute the CodeBlockHash on
unreasonably long source code strings. This is done by first applying a length
check and, if needed, computing the hash with an alternate method.
This is OK to do because:
1. CodeBlockHash is not a critical hash.
2. In practice, reasonable source code are not that long.
3. And if they are that long, then we are still diversifying the hash on their
length. But if they do collide, it's OK.
The only invariant here is that we should always produce the same hash for the
same source string. Since the algorithm is deterministic, this invariant is not
violated.
* bytecode/CodeBlockHash.cpp:
(JSC::CodeBlockHash::CodeBlockHash):
2020-06-06 Devin Rousso <drousso@apple.com>
Web Inspector: unify the naming scheme for agents used by instrumentation
https://bugs.webkit.org/show_bug.cgi?id=212859
Reviewed by Timothy Hatcher.
Inspector agents fall into one of three categories:
- "persistent" when Web Inspector is connected
- "enabled" when that agent is `enable`d, such as if the corresponding tab is visible
- "tracking" when that agent is part of a timeline recording.
The only exception to this is the Console agent, as that exists regardless of whether Web
Inspector is connected as it needs to preserve messages logged before Web Inspector connects.
Also remove the "Inspector" prefix from getter/setter methods as it adds confusion if that
agent also has subclasses (e.g. `InspectorRuntimeAgent` and `PageRuntimeAgent`).
* inspector/JSGlobalObjectConsoleClient.h:
* inspector/JSGlobalObjectInspectorController.cpp:
* inspector/agents/InspectorConsoleAgent.h:
2020-06-05 Michael Saboff <msaboff@apple.com>
Make FAST_JIT_PERMISSIONS check in LinkBuffer::copyCompactAndLinkCode a runtime check
https://bugs.webkit.org/show_bug.cgi?id=212825
Reviewed by Saam Barati.
Added useFastJITPermissions() for runtime checks of FAST_JIT_PERMISSIONS
including the cases where it is conditional on OS settings. This is now
used in a few places to declutter the code.
When using the fast JIT permissions path, the JIT memory is the direct output
of the linking. Modified BranchCompactionLinkBuffer to hold a pointer to that
output "buffer" or a temporarily allocated buffer depending on if fast JIT
permissions are enabled.
Broke out the "verify hash" conditionally compiled code with a file local
ENABLE_VERIFY_JIT_HASH macro for readability.
* assembler/LinkBuffer.cpp:
(JSC::BranchCompactionLinkBuffer::BranchCompactionLinkBuffer):
(JSC::BranchCompactionLinkBuffer::~BranchCompactionLinkBuffer):
Changed this to use a provided buffer or a malloc'ed buffer. When using
a malloc'ed buffer, we put it in a thread local cache.
(JSC::LinkBuffer::copyCompactAndLinkCode):
* jit/ExecutableAllocator.h:
(JSC::useFastJITPermissions):
(JSC::performJITMemcpy):
2020-06-05 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Put dfgOpNames in __DATA,__const section instead of __DATA,__data
https://bugs.webkit.org/show_bug.cgi?id=212840
Reviewed by Saam Barati.
dfgOpNames array itself is not const annotated, and the compiler makes it __DATA,__data instead of __DATA,__const.
We should annotate it with const to ensure that this is compiled into __DATA,__const. We also remove unused CallFrame::describeFrame
since it allocates some bss memory, while we have more sophisticated mechanism (VMInspector) for this functionality and this function
is no longer used.
* dfg/DFGDoesGCCheck.cpp:
(JSC::DFG::DoesGCCheck::verifyCanGC):
* dfg/DFGGraph.cpp:
* dfg/DFGGraph.h:
* interpreter/CallFrame.cpp:
(JSC::CallFrame::describeFrame): Deleted.
* interpreter/CallFrame.h:
2020-06-05 Tadeu Zagallo <tzagallo@apple.com>
REGRESSION(r262523): Fix testb3
https://bugs.webkit.org/show_bug.cgi?id=212791
Reviewed by Mark Lam.
* b3/testb3_1.cpp:
(run):
(main):
2020-06-05 Paulo Matos <pmatos@igalia.com>
Add missing ECMAMode header to fix NonUnified Build
https://bugs.webkit.org/show_bug.cgi?id=212838
Reviewed by Darin Adler.
* bytecode/PutByValFlags.h:
2020-06-05 Saam Barati <sbarati@apple.com>
Audit safe to execute
https://bugs.webkit.org/show_bug.cgi?id=207075
<rdar://problem/59085094>
Reviewed by Yusuke Suzuki.
This audit found one interesting case for DOMJIT nodes. We emit safety checks
for CallDOM/CallDOMGetter inside fixup phase and the bytecode parser. When
determining if these nodes are safe to execute, we need to also ensure that
these checks hold.
I've also added a helper to JSDollarVM to ensure that this patch doesn't break
LICM of DOMJIT.
This patch also moves some nodes we will never hoist to return false.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleDOMJITGetter):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::attemptToMakeCallDOM):
* dfg/DFGNode.h:
(JSC::DFG::Node::classInfo):
(JSC::DFG::Node::requiredDOMJITClassInfo):
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* tools/JSDollarVM.cpp:
(JSC::functionCreateDOMJITGetterNoEffectsObject):
(JSC::JSDollarVM::finishCreation):
2020-06-05 Devin Rousso <drousso@apple.com>
Logical Assignment: perform NamedEvaluation of anonymous functions
https://bugs.webkit.org/show_bug.cgi?id=212679
Reviewed by Ross Kirsling.
* parser/ASTBuilder.h:
(JSC::ASTBuilder::makeAssignNode):
2020-06-05 Yusuke Suzuki <ysuzuki@apple.com>
DOM constructor should only accept Ref<> / ExceptionOr<Ref<>> for creation to ensure toJSNewlyCreated is always returning object
https://bugs.webkit.org/show_bug.cgi?id=212767
Reviewed by Darin Adler.
* runtime/JSObject.h:
(JSC::asObject):
2020-06-05 Andy Estes <aestes@apple.com>
[Apple Pay] Remove conditionals for ENABLE_APPLE_PAY_SESSION_V(3|4)
https://bugs.webkit.org/show_bug.cgi?id=212541
<rdar://problem/63781452>
Reviewed by Darin Adler.
APPLE_PAY_SESSION_V(3|4) is now enabled whenever APPLE_PAY itself is enabled.
* Configurations/FeatureDefines.xcconfig:
2020-06-05 Caitlin Potter <caitp@igalia.com>
[JSC] Add support for private class fields
https://bugs.webkit.org/show_bug.cgi?id=206431
Reviewed by Saam Barati.
Expanding upon the earlier public class fields patch, we implement the remaining (and
significant parts) of the instance fields (https://tc39.es/proposal-class-fields/).
There are a variety of key changes here:
- Parser now understands the concept of private names (Token PRIVATENAME).
- 1 new opcode (op_get_private_name), one changed opcode (op_put_by_val_direct).
- A method for creating Symbol objects with a null PrivateSymbolImpl is exposed as a
LinkTimeConstant (@createPrivateSymbol).
- Null Private Symbols are stored by name (not a valid identifier) in a JSScope, and
are loaded from the outer scope whenever they are used by the modified opcodes.
The changes to op_put_by_val_direct include a new bytecode operand (PutByValFlags) which are
used to distinguish between overwriting or defining a new private field. Specifically, when it
comes to private field accesses, it's necessary to throw an exception when accessing a field
which does not exist, or when attempting to define a private field which has already been
defined.
During the evaluation of a class expression, before the class element list is evaluated (in case
any computed property names expressions refer to a new private field), a new PrivateSymbol is
created for each individual private field name, and stored in the class lexical scope.
Private field names are loaded from scope before their use. This prevents multiple evaluations
of the same class source from accessing each other's private fields, because the values of the
symbols loaded from the class scope would be distinct. This is required by the proposal text,
and is the key reason why we use ByVal lookups rather than ById lookups.
To illustrate, typical private field access will look like:
<Field Reads>
resolve_scope <scope=>, <currentScope>, "#x", GlobalProperty, 0
get_from_scope <symbol=>, <scope>, "#x", 1050624<DoNotThrowIfNotFound|GlobalProperty|NotInitialization>, 0, 0
get_private_name <value=>, <receiver --- probably 'this'>, <symbol>
<Field Writes>
resolve_scope <scope=>, <currentScope>, "#x", GlobalProperty, 0
get_from_scope <symbol=>, <scope>, "#x", 1050624<DoNotThrowIfNotFound|GlobalProperty|NotInitialization>, 0, 0
put_by_val_direct <receiver, probably 'this'>, <symbol>, <value>, <PutByValPrivateName>
<Field Definition>
resolve_scope <scope=>, <currentScope>, "#x", GlobalProperty, 0
get_from_scope <symbol=>, <scope>, "#x", 1050624<DoNotThrowIfNotFound|GlobalProperty|NotInitialization>, 0, 0
put_by_val_direct <receiver, probably 'this'>, <symbol>, <value>, <PutByValPrivateName|PutByValThrowIfExists>
The feature is currently hidden behind the feature flag JSC::Options::usePrivateClassFields.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* builtins/BuiltinNames.h:
* bytecode/BytecodeList.rb:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::finalizeLLIntInlineCaches):
* bytecode/Fits.h:
* bytecode/LinkTimeConstant.h:
* bytecode/PutByValFlags.cpp: Copied from Source/JavaScriptCore/bytecode/PutKind.h.
(WTF::printInternal):
* bytecode/PutByValFlags.h: Added.
(JSC::PutByValFlags::create):
(JSC::PutByValFlags::createDirect):
(JSC::PutByValFlags::createDefinePrivateField):
(JSC::PutByValFlags::createPutPrivateField):
(JSC::PutByValFlags::isDirect const):
(JSC::PutByValFlags::ecmaMode const):
(JSC::PutByValFlags::privateFieldAccessKind const):
(JSC::PutByValFlags::isPrivateFieldAccess const):
(JSC::PutByValFlags::isPrivateFieldPut const):
(JSC::PutByValFlags::isPrivateFieldAdd const):
(JSC::PutByValFlags::PutByValFlags):
* bytecode/PutKind.h:
* bytecode/UnlinkedFunctionExecutable.cpp:
(JSC::generateUnlinkedFunctionCodeBlock):
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::instantiateLexicalVariables):
(JSC::BytecodeGenerator::emitDirectGetByVal):
(JSC::BytecodeGenerator::emitDirectPutByVal):
(JSC::BytecodeGenerator::emitDefinePrivateField):
(JSC::BytecodeGenerator::emitPrivateFieldPut):
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::PropertyListNode::emitDeclarePrivateFieldNames):
(JSC::PropertyListNode::emitBytecode):
(JSC::PropertyListNode::emitPutConstantProperty):
(JSC::DotAccessorNode::emitBytecode):
(JSC::BaseDotNode::emitGetPropertyValue):
(JSC::BaseDotNode::emitPutProperty):
(JSC::FunctionCallDotNode::emitBytecode):
(JSC::PostfixNode::emitDot):
(JSC::PrefixNode::emitDot):
(JSC::AssignDotNode::emitBytecode):
(JSC::ReadModifyDotNode::emitBytecode):
(JSC::DefineFieldNode::emitBytecode):
(JSC::ClassExprNode::emitBytecode):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ecmaMode):
(JSC::DFG::ecmaMode<OpPutByValDirect>):
(JSC::DFG::ByteCodeParser::handlePutByVal):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::cachedPutById):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compilePutById):
* generator/DSL.rb:
* jit/ICStats.h:
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
* jit/JIT.h:
* jit/JITInlineCacheGenerator.cpp:
(JSC::JITPutByIdGenerator::JITPutByIdGenerator):
(JSC::JITPutByIdGenerator::slowPathFunction):
* jit/JITInlineCacheGenerator.h:
(JSC::JITPutByIdGenerator::JITPutByIdGenerator):
* jit/JITInlines.h:
(JSC::JIT::ecmaMode):
(JSC::JIT::ecmaMode<OpPutById>):
(JSC::JIT::ecmaMode<OpPutByValDirect>):
(JSC::JIT::privateFieldAccessKind):
(JSC::JIT::privateFieldAccessKind<OpPutByValDirect>):
* jit/JITOperations.cpp:
(JSC::putPrivateField):
(JSC::definePrivateField):
* jit/JITOperations.h:
* jit/JITPropertyAccess.cpp:
(JSC::JIT::emitPutByValWithCachedId):
(JSC::JIT::emitSlow_op_put_by_val):
(JSC::JIT::emit_op_put_by_id):
* jit/JITPropertyAccess32_64.cpp:
(JSC::JIT::emitSlow_op_put_by_val):
(JSC::JIT::emit_op_put_by_id):
* jit/Repatch.cpp:
(JSC::appropriateGenericPutByIdFunction):
(JSC::appropriateOptimizingPutByIdFunction):
(JSC::tryCachePutByID):
* llint/LLIntOffsetsExtractor.cpp:
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::LLINT_SLOW_PATH_DECL):
* llint/LLIntSlowPaths.h:
* llint/LowLevelInterpreter32_64.asm:
* llint/LowLevelInterpreter64.asm:
* parser/ASTBuilder.h:
(JSC::ASTBuilder::createDotAccess):
(JSC::ASTBuilder::isPrivateLocation):
(JSC::ASTBuilder::makeFunctionCallNode):
(JSC::ASTBuilder::makeAssignNode):
* parser/Lexer.cpp:
(JSC::Lexer<LChar>::parseIdentifier):
(JSC::Lexer<UChar>::parseIdentifier):
(JSC::Lexer<CharacterType>::parseIdentifierSlowCase):
(JSC::Lexer<T>::lexWithoutClearingLineTerminator):
* parser/NodeConstructors.h:
(JSC::BaseDotNode::BaseDotNode):
(JSC::DotAccessorNode::DotAccessorNode):
(JSC::FunctionCallDotNode::FunctionCallDotNode):
(JSC::CallFunctionCallDotNode::CallFunctionCallDotNode):
(JSC::ApplyFunctionCallDotNode::ApplyFunctionCallDotNode):
(JSC::HasOwnPropertyFunctionCallDotNode::HasOwnPropertyFunctionCallDotNode):
(JSC::AssignDotNode::AssignDotNode):
(JSC::ReadModifyDotNode::ReadModifyDotNode):
* parser/Nodes.cpp:
(JSC::PropertyListNode::shouldCreateLexicalScopeForClass):
* parser/Nodes.h:
(JSC::ExpressionNode::isPrivateLocation const):
(JSC::BaseDotNode::base const):
(JSC::BaseDotNode::identifier const):
(JSC::BaseDotNode::type const):
(JSC::BaseDotNode::isPrivateField const):
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseVariableDeclarationList):
(JSC::Parser<LexerType>::parseDestructuringPattern):
(JSC::Parser<LexerType>::parseClass):
(JSC::Parser<LexerType>::parseInstanceFieldInitializerSourceElements):
(JSC::Parser<LexerType>::usePrivateName):
(JSC::Parser<LexerType>::parseMemberExpression):
(JSC::Parser<LexerType>::parseUnaryExpression):
(JSC::Parser<LexerType>::printUnexpectedTokenText):
* parser/Parser.h:
(JSC::Scope::isPrivateNameScope const):
(JSC::Scope::setIsPrivateNameScope):
(JSC::Scope::hasPrivateName):
(JSC::Scope::copyUndeclaredPrivateNamesTo):
(JSC::Scope::hasUsedButUndeclaredPrivateNames const):
(JSC::Scope::usePrivateName):
(JSC::Scope::declarePrivateName):
(JSC::Parser::findPrivateNameScope):
(JSC::Parser::privateNameScope):
(JSC::Parser::copyUndeclaredPrivateNamesToOuterScope):
(JSC::Parser::matchAndUpdate):
(JSC::Parser<LexerType>::parse):
(JSC::parse):
* parser/ParserTokens.h:
* parser/SyntaxChecker.h:
(JSC::SyntaxChecker::createDotAccess):
(JSC::SyntaxChecker::operatorStackPop):
* parser/VariableEnvironment.cpp:
(JSC::VariableEnvironment::operator=):
(JSC::VariableEnvironment::swap):
(JSC::CompactVariableEnvironment::CompactVariableEnvironment):
* parser/VariableEnvironment.h:
(JSC::VariableEnvironmentEntry::isPrivateName const):
(JSC::VariableEnvironmentEntry::setIsPrivateName):
(JSC::PrivateNameEntry::PrivateNameEntry):
(JSC::PrivateNameEntry::isUsed const):
(JSC::PrivateNameEntry::isDeclared const):
(JSC::PrivateNameEntry::setIsUsed):
(JSC::PrivateNameEntry::setIsDeclared):
(JSC::PrivateNameEntry::bits const):
(JSC::PrivateNameEntry::operator== const):
(JSC::VariableEnvironment::VariableEnvironment):
(JSC::VariableEnvironment::size const):
(JSC::VariableEnvironment::mapSize const):
(JSC::VariableEnvironment::declarePrivateName):
(JSC::VariableEnvironment::usePrivateName):
(JSC::VariableEnvironment::privateNames const):
(JSC::VariableEnvironment::privateNamesSize const):
(JSC::VariableEnvironment::hasPrivateName):
(JSC::VariableEnvironment::copyPrivateNamesTo const):
(JSC::VariableEnvironment::copyUndeclaredPrivateNamesTo const):
(JSC::VariableEnvironment::RareData::RareData):
(JSC::VariableEnvironment::getOrAddPrivateName):
* runtime/CachedTypes.cpp:
(JSC::CachedOptional::decodeAsPtr const):
(JSC::CachedVariableEnvironmentRareData::encode):
(JSC::CachedVariableEnvironmentRareData::decode const):
(JSC::CachedVariableEnvironment::encode):
(JSC::CachedVariableEnvironment::decode const):
(JSC::CachedSymbolTableRareData::encode):
(JSC::CachedSymbolTableRareData::decode const):
(JSC::CachedSymbolTable::encode):
(JSC::CachedSymbolTable::decode const):
* runtime/CodeCache.cpp:
(JSC::generateUnlinkedCodeBlockImpl):
* runtime/CommonIdentifiers.cpp:
(JSC::CommonIdentifiers::CommonIdentifiers):
* runtime/CommonIdentifiers.h:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/CommonSlowPaths.h:
* runtime/ExceptionHelpers.cpp:
(JSC::createInvalidPrivateNameError):
(JSC::createRedefinedPrivateNameError):
* runtime/ExceptionHelpers.h:
* runtime/JSGlobalObject.cpp:
(JSC::createPrivateSymbol):
(JSC::JSGlobalObject::init):
* runtime/JSObject.h:
* runtime/JSObjectInlines.h:
(JSC::JSObject::getPrivateFieldSlot):
(JSC::JSObject::getPrivateField):
(JSC::JSObject::putPrivateField):
(JSC::JSObject::definePrivateField):
* runtime/JSScope.cpp:
(JSC::JSScope::collectClosureVariablesUnderTDZ):
* runtime/OptionsList.h:
* runtime/SymbolTable.cpp:
(JSC::SymbolTable::cloneScopePart):
* runtime/SymbolTable.h:
2020-06-05 Paulo Matos <pmatos@igalia.com>
Fix includes to fix latest non-unified builds breakages
https://bugs.webkit.org/show_bug.cgi?id=212802
Reviewed by Adrian Perez de Castro.
* dfg/DFGDoesGCCheck.cpp:
* runtime/JSDateMath.h:
2020-06-04 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Report extra memory allocation from PropertyTable
https://bugs.webkit.org/show_bug.cgi?id=212793
Reviewed by Saam Barati.
This patch adds extra memory reporting from PropertyTable to make GC
responsive to the increase of memory in PropertyTable.
* runtime/PropertyMapHashTable.h:
(JSC::PropertyTable::add):
(JSC::PropertyTable::remove):
(JSC::PropertyTable::rehash):
(JSC::PropertyTable::dataSize):
* runtime/PropertyTable.cpp:
(JSC::PropertyTable::finishCreation):
(JSC::PropertyTable::visitChildren):
* runtime/Structure.cpp:
(JSC::Structure::materializePropertyTable):
* runtime/StructureInlines.h:
(JSC::Structure::add):
(JSC::Structure::remove):
2020-06-04 Commit Queue <commit-queue@webkit.org>
Unreviewed, reverting r262583.
https://bugs.webkit.org/show_bug.cgi?id=212799
Internal source code has the same bug, needs to be landed
after fixing internal source
Reverted changeset:
"DOM constructor should only accept Ref<> / ExceptionOr<Ref<>>
for creation to ensure toJSNewlyCreated is always returning
object"
https://bugs.webkit.org/show_bug.cgi?id=212767
https://trac.webkit.org/changeset/262583
2020-06-04 Michael Saboff <msaboff@apple.com>
Add a Thread Specific Cache for LinkBuffer::CompactAndLinkCode()
https://bugs.webkit.org/show_bug.cgi?id=212765
Reviewed by Saam Barati.
Added a thread local buffer for CPU types that use a second buffer when compacting.
This is very similary to the work done in https://bugs.webkit.org/show_bug.cgi?id=212562.
* assembler/LinkBuffer.cpp:
(JSC::threadSpecificBranchCompactionLinkBuffer):
(JSC::BranchCompactionLinkBuffer::BranchCompactionLinkBuffer):
(JSC::BranchCompactionLinkBuffer::~BranchCompactionLinkBuffer):
(JSC::BranchCompactionLinkBuffer::data):
(JSC::BranchCompactionLinkBuffer::takeBufferIfLarger):
(JSC::BranchCompactionLinkBuffer::size):
(JSC::LinkBuffer::copyCompactAndLinkCode):
2020-06-04 Mark Lam <mark.lam@apple.com>
Add Options::validateDoesGC() for turning DoesGC validation on/off.
https://bugs.webkit.org/show_bug.cgi?id=212773
Reviewed by Saam Barati.
It will default to on if ASSERT_ENABLED because we want testing to be done with
the validation on. When needed, we can turn it off if we need to e.g. to
de-clutter disassembly dumps while debugging.
If Options::validateDoesGC() is false, we turn off JIT code emission for this
check, as well as skip the validation checks. There are still places in C++
code that store to DoesGC::m_value without checking Options::validateDoesGC().
It doesn't hurt to just let these stores proceed, and performance-wise, it's
probably cheaper to just do the store unconditionally than to gate it on a load of
Options::validateDoesGC() first.
Also made it explicit that the check on validateDFGDoesGC is a constexpr check.
* dfg/DFGDoesGCCheck.cpp:
(JSC::DFG::DoesGCCheck::verifyCanGC):
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* runtime/OptionsList.h:
2020-06-04 Ross Kirsling <ross.kirsling@sony.com>
Intl classes should have meaningful @@toStringTag values
https://bugs.webkit.org/show_bug.cgi?id=212769
Reviewed by Yusuke Suzuki.
Implementation of https://github.com/tc39/ecma402/pull/430, which achieved consensus this week.
This ensures we get "[object Intl.Collator]" (etc.) instead "[object Object]" for older Intl classes.
* runtime/IntlCollatorPrototype.cpp:
* runtime/IntlDateTimeFormatPrototype.cpp:
* runtime/IntlNumberFormatPrototype.cpp:
* runtime/IntlPluralRulesPrototype.cpp:
2020-06-04 Alexey Shvayka <shvaikalesh@gmail.com>
GetMethod isn't performed properly on iterators
https://bugs.webkit.org/show_bug.cgi?id=212771
Reviewed by Saam Barati.
Before this change, iterator's "return" and "throw" methods with value of `null` were
considered incorrect rather than missing, causing TypeError to be thrown.
This patch aligns method lookup of iterators with the spec [1], V8, and SpiderMonkey
by utilizing isUndefinedOrNull(), which doesn't special-case [[IsHTMLDDA]] objects [2],
fixing a few Annex B tests.
for/of microbenchmarks are neutral.
[1]: https://tc39.es/ecma262/#sec-getmethod (step 3)
[2]: https://tc39.es/ecma262/#sec-IsHTMLDDA-internal-slot
* builtins/AsyncFromSyncIteratorPrototype.js:
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitIteratorGenericClose):
(JSC::BytecodeGenerator::emitGetAsyncIterator):
(JSC::BytecodeGenerator::emitDelegateYield):
* runtime/IteratorOperations.cpp:
(JSC::iteratorClose):
* runtime/JSGenericTypedArrayViewConstructorInlines.h:
(JSC::constructGenericTypedArrayViewWithArguments):
2020-06-04 Mark Lam <mark.lam@apple.com>
Reduce DFGDoesGCCheck to only storing a uint32_t.
https://bugs.webkit.org/show_bug.cgi?id=212734
Reviewed by Saam Barati and Caio Lima.
This patch changes the encoding of DoesGCCheck so that it will fit better in a
uint32_t. This has the following benefits:
1. speed improvement for debug builds because it now takes less instructions
(especially in JITted code) to store to DoesGCCheck::m_value.
2. enables this check for 32-bit platforms as well.
Fun fact: we currently have 373 DFG::NodeTypes. Hence, 9 bits for nodeOp.
The new encoding provides 21 bis for the nodeIndex. This gives us up to 2097152
node indexes. In my experience, I've never seen more than 3 decimal digits for
the nodeIndex so far. If we ever find that we need more than 21 bits of nodeIndex,
we have 2 options to deal with it:
1. We can just ignore the high bits. After all, it is the nodeOp that is the
most interesting piece of data we need to debug doesGC issues.
2. We can make DoesGCCheck use uint64_t for storage. This encoding automatically
scales to 64-bit, while still allowing the more efficient form of storing a
32-bit immediate to be used for the common cases.
This patch also makes ENABLE_DFG_DOES_GC_VALIDATION dependent on ENABLE(DFG_JIT).
DoesGC is only relevant for the DFG and FTL JITs.
* dfg/DFGDoesGCCheck.cpp:
(JSC::DFG::DoesGCCheck::verifyCanGC):
* dfg/DFGDoesGCCheck.h:
(JSC::DFG::DoesGCCheck::encode):
(JSC::DFG::DoesGCCheck::expectDoesGC const):
(JSC::DFG::DoesGCCheck::isSpecial const):
(JSC::DFG::DoesGCCheck::special):
(JSC::DFG::DoesGCCheck::nodeOp):
(JSC::DFG::DoesGCCheck::nodeIndex):
(JSC::DFG::DoesGCCheck::expectDoesGC): Deleted.
(JSC::DFG::DoesGCCheck::isSpecial): Deleted.
(JSC::DFG::DoesGCCheck::specialIndex): Deleted.
(JSC::DFG::DoesGCCheck::bits): Deleted.
* dfg/DFGNodeType.h:
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* heap/Heap.h:
2020-06-04 Tim Horton <timothy_horton@apple.com>
Work around broken system version macro
https://bugs.webkit.org/show_bug.cgi?id=212726
Reviewed by Dan Bernstein.
* Configurations/DebugRelease.xcconfig:
2020-06-04 Andy Estes <aestes@apple.com>
[watchOS] Re-enable content filtering in the simulator build
https://bugs.webkit.org/show_bug.cgi?id=212711
<rdar://problem/63938350>
Reviewed by Wenson Hsieh.
* Configurations/FeatureDefines.xcconfig:
2020-06-04 Mark Lam <mark.lam@apple.com>
SpeculativeJIT::compileDateGet()'s slow path does not need an exception check.
https://bugs.webkit.org/show_bug.cgi?id=212645
Reviewed by Yusuke Suzuki.
SpeculativeJIT::compileDateGet() implements a bunch of Date intrinsics which call
into a C++ operation function do their work. However, the call to these operation
functions were done using a slow path generator configured to automatically
emit exception checks after the call. These exception checks are unneeded because
those functions will not throw any exceptions.
This issue was found with JSC stress test runs on a debug build. The doesGC
verifier was failing on the exceptionFuzz/date-format-xparb.js test. The reason
is because doesGC does not expect any these Date intrinsics to throw any exceptions,
but SpeculativeJIT was emitting the unneeded exception checks there. These
exception check sites get turned into throw sites by the exceptionFuzzer, and
they allocate an Error object there. This allocation made the doesGC verifier
not happy.
This patch fixes this issue by changing SpeculativeJIT::compileDateGet() to
pass ExceptionCheckRequirement::CheckNotNeeded to the slow path generator.
The patch also proves that all the operation functions cannot throw any exceptions.
Previously, the operations passes a VM& to the Date functions. The purpose for
doing this is so that the Date functions can work with a few date cache data
structures stored as VM fields.
This patch refactors those VM fields into a VM::DateCache struct, and changed all
those Date functions to take a VM::DateCache& instead of a VM&. Since the Date
functions no longer take a VM&, this proves that they cannot throw because they
would need a VM& to make a ThrowScope in order to throw.
Update: Yusuke pointed out that the lack of a JSGlobalObject* argument is sufficient
to guarantee that the Date functions cannot throw. However, we'll keep this
DateCache refactoring since it provides additional info that the Date functions
only operate on the DateCache fields and nothing else in VM.
Also removed DFG::JITCompile's fastExceptionCheck() which is unused.
* dfg/DFGJITCompiler.h:
(JSC::DFG::JITCompiler::fastExceptionCheck): Deleted.
* dfg/DFGOperations.cpp:
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compileDateGet):
* runtime/DateConstructor.cpp:
(JSC::millisecondsFromComponents):
(JSC::callDate):
* runtime/DateInstance.cpp:
(JSC::DateInstance::calculateGregorianDateTime const):
(JSC::DateInstance::calculateGregorianDateTimeUTC const):
* runtime/DateInstance.h:
* runtime/DatePrototype.cpp:
(JSC::formatLocaleDate):
(JSC::formateDateInstance):
(JSC::dateProtoFuncToISOString):
(JSC::dateProtoFuncGetFullYear):
(JSC::dateProtoFuncGetUTCFullYear):
(JSC::dateProtoFuncGetMonth):
(JSC::dateProtoFuncGetUTCMonth):
(JSC::dateProtoFuncGetDate):
(JSC::dateProtoFuncGetUTCDate):
(JSC::dateProtoFuncGetDay):
(JSC::dateProtoFuncGetUTCDay):
(JSC::dateProtoFuncGetHours):
(JSC::dateProtoFuncGetUTCHours):
(JSC::dateProtoFuncGetMinutes):
(JSC::dateProtoFuncGetUTCMinutes):
(JSC::dateProtoFuncGetSeconds):
(JSC::dateProtoFuncGetUTCSeconds):
(JSC::dateProtoFuncGetTimezoneOffset):
(JSC::setNewValueFromTimeArgs):
(JSC::setNewValueFromDateArgs):
(JSC::dateProtoFuncSetYear):
(JSC::dateProtoFuncGetYear):
* runtime/JSDateMath.cpp:
(JSC::localTimeOffset):
(JSC::gregorianDateTimeToMS):
(JSC::msToGregorianDateTime):
(JSC::parseDate):
* runtime/JSDateMath.h:
* runtime/VM.cpp:
(JSC::VM::resetDateCache):
* runtime/VM.h:
2020-06-04 Paulo Matos <pmatos@igalia.com>
Fix 32bit build broken at r262513
https://bugs.webkit.org/show_bug.cgi?id=212735
Unreviewed Gardening.
Proper fix is being worked out under https://bugs.webkit.org/show_bug.cgi?id=212734
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::compileExit):
2020-06-03 Tadeu Zagallo <tzagallo@apple.com>
Disable B3 hoistLoopInvariantValues by default
https://bugs.webkit.org/show_bug.cgi?id=212511
<rdar://problem/63813245>
Reviewed by Mark Lam.
The hoistLoopInvariantValues optimization in B3 does not calculate the cost of hoisting the candidates.
For example, in the test case provided with the bug, a switch inside a loop can lead to hoisting the body
of several switch cases which would never be executed. Other than leading to worse runtime, this also
increases the pressure in the register allocate, leading to worse compile times (~10x worse in this case).
I have added a FIXME to consider adding cost calculation and re-enabling this pass, but given that we
already have LICM in DFG, it should be ok to disable it for now.
* b3/B3Generate.cpp:
(JSC::B3::generateToAir):
* runtime/OptionsList.h:
2020-06-03 Mark Lam <mark.lam@apple.com>
Gardening: fix broken Windows debug build.
https://bugs.webkit.org/show_bug.cgi?id=212680
Not reviewed.
* dfg/DFGDoesGCCheck.cpp:
(JSC::DFG::DoesGCCheck::verifyCanGC):
* dfg/DFGDoesGCCheck.h:
2020-06-03 Mark Lam <mark.lam@apple.com>
[Re-landing] Enhance DoesGC verification to print more useful info when verification fails.
https://bugs.webkit.org/show_bug.cgi?id=212680
Reviewed by Yusuke Susuki.
When DoesGC verification fails, the first step of debugging it would be to find
out what and which DFG node resulted in the failed verification. In pre-existing
code, all we get is an assertion failure.
This patch makes it so that the verifier will dump useful info. Here's an example:
Error: DoesGC failed @ D@34 DateGetInt32OrNaN in #DtCHMz:[0x1135bd1d0->0x1135bcab0->0x1135e5c80, DFGFunctionCall, 150 (DidTryToEnterInLoop)]
[0] frame 0x7ffee8285660 {
name:
sourceURL:
isInlinedFrame: false
callee: 0x1135f6820
returnPC: 0x50ce61248ae6
callerFrame: 0x7ffee82856f0
rawLocationBits: 5 0x5
codeBlock: 0x1135bd1d0 #DtCHMz:[0x1135bd1d0->0x1135bcab0->0x1135e5c80, DFGFunctionCall, 150 (DidTryToEnterInLoop)]
hasCodeOrigins: true
callSiteIndex: 5 of 13
jitCode: 0x113020200 start 0x50ce61214c60 end 0x50ce61219b00
line: 1
column: 60
EntryFrame: 0x7ffee8285860
}
[1] frame 0x7ffee82856f0 {
name:
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1135f65a0
returnPC: 0x50ce61227e99
callerFrame: 0x7ffee8285770
rawLocationBits: 4 0x4
codeBlock: 0x1135bd0a0 #BU6Zcd:[0x1135bd0a0->0x1135bc260->0x1135e5180, DFGFunctionCall, 112 (DidTryToEnterInLoop)]
hasCodeOrigins: true
callSiteIndex: 4 of 12
jitCode: 0x113004000 start 0x50ce61212c60 end 0x50ce61214960
line: 26
column: 22
EntryFrame: 0x7ffee8285860
}
[2] frame 0x7ffee8285770 {
name:
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1135f64e0
returnPC: 0x108058eb1
callerFrame: 0x7ffee82857e0
rawLocationBits: 1001 0x3e9
codeBlock: 0x1135bc130 #DAS9xe:[0x1135bc130->0x1135e5100, BaselineFunctionCall, 1149]
bc#1001 of 1149
line: 417
column: 38
EntryFrame: 0x7ffee8285860
}
[3] frame 0x7ffee82857e0 {
name: global code
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1130f97b8
returnPC: 0x108039043
callerFrame: 0x0
rawLocationBits: 23 0x17
codeBlock: 0x1135bc000 <global>#CukXvt:[0x1135bc000->0x1130cd768, LLIntGlobal, 81]
bc#23 of 81
line: 425
column: 3
EntryFrame: 0x7ffee8285860
}
ASSERTION FAILED: expectDoesGC()
The error message now comes with the node index, NodeType, codeBlock which this
failure was found in, and the JS call stack that led to the failure.
Changes made:
1. Introduced a DoesGCCheck value that is used to encode some of the above data.
Previously, we only recorded whether doesGC() returns true or false for the
Node. Now, we record the nodeIndex and nodeOp as well.
Note that we also set DoesGC expectations for OSR exits. So, DoesGCCheck
includes Special cases for those.
2. Added store64(TrustedImm64 imm, const void* address) emitters for X86_64 and ARM64.
Also added a test for this new emitter in testmasm.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::store64):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::store64):
* assembler/testmasm.cpp:
(JSC::testStore64Imm64AddressPointer):
(JSC::run):
* dfg/DFGDoesGCCheck.cpp: Copied from Source/JavaScriptCore/dfg/DFGDoesGCCheck.cpp.
* dfg/DFGDoesGCCheck.h: Copied from Source/JavaScriptCore/dfg/DFGDoesGCCheck.h.
* dfg/DFGGraph.cpp:
* dfg/DFGOSRExit.cpp:
(JSC::DFG::operationCompileOSRExit):
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
(JSC::FTL::operationCompileFTLOSRExit):
* heap/CompleteSubspace.cpp:
(JSC::CompleteSubspace::tryAllocateSlow):
(JSC::CompleteSubspace::reallocatePreciseAllocationNonVirtual):
* heap/CompleteSubspaceInlines.h:
(JSC::CompleteSubspace::allocateNonVirtual):
* heap/DeferGC.h:
(JSC::DeferGC::~DeferGC):
* heap/GCDeferralContextInlines.h:
(JSC::GCDeferralContext::~GCDeferralContext):
* heap/Heap.cpp:
(JSC::Heap::collectNow):
(JSC::Heap::collectAsync):
(JSC::Heap::collectSync):
(JSC::Heap::stopIfNecessarySlow):
(JSC::Heap::collectIfNecessaryOrDefer):
* heap/Heap.h:
(JSC::Heap::addressOfDoesGC):
(JSC::Heap::setDoesGCExpectation):
(JSC::Heap::verifyCanGC):
(JSC::Heap::expectDoesGC const): Deleted.
(JSC::Heap::setExpectDoesGC): Deleted.
(JSC::Heap::addressOfExpectDoesGC): Deleted.
* heap/HeapInlines.h:
(JSC::Heap::acquireAccess):
(JSC::Heap::stopIfNecessary):
* heap/LocalAllocatorInlines.h:
(JSC::LocalAllocator::allocate):
* heap/PreciseAllocation.cpp:
(JSC::PreciseAllocation::tryCreate):
(JSC::PreciseAllocation::createForLowerTier):
* runtime/JSString.h:
(JSC::jsSingleCharacterString):
(JSC::JSString::toAtomString const):
(JSC::JSString::toExistingAtomString const):
(JSC::JSString::value const):
(JSC::JSString::tryGetValue const):
(JSC::JSRopeString::unsafeView const):
(JSC::JSRopeString::viewWithUnderlyingString const):
(JSC::JSString::unsafeView const):
* runtime/RegExpMatchesArray.h:
(JSC::createRegExpMatchesArray):
2020-06-03 Mark Lam <mark.lam@apple.com>
DFGSSAConversionPhase.cpp needs to #include OperandsInlines.h.
https://bugs.webkit.org/show_bug.cgi?id=212687
Reviewed by Keith Miller.
Without this, strange build failures can happen with unified builds.
For example, the Windows build started failing due a linkage error in this file
when the patch from https://bugs.webkit.org/show_bug.cgi?id=212680 landed.
212680 introduced a new .cpp file, and that probably bumped DFGSSAConversionPhase.cpp
into another unified unit, thereby depriving it from seeing the OperandsInlines.h
#include'd by another .cpp.
* dfg/DFGSSAConversionPhase.cpp:
2020-06-03 Mark Lam <mark.lam@apple.com>
Fix non-unified --jsc-only build.
https://bugs.webkit.org/show_bug.cgi?id=212707
Reviewed by Yusuke Suzuki.
These files need JSGlobalObjectInlines.h. But rather than adding yet another
#include, we'll just remove many individual ones and just #include JSCInlines.h
instead.
* wasm/js/JSToWasmICCallee.cpp:
* wasm/js/WebAssemblyCompileErrorConstructor.cpp:
* wasm/js/WebAssemblyCompileErrorPrototype.cpp:
* wasm/js/WebAssemblyGlobalPrototype.cpp:
* wasm/js/WebAssemblyInstanceConstructor.cpp:
* wasm/js/WebAssemblyInstancePrototype.cpp:
* wasm/js/WebAssemblyLinkErrorConstructor.cpp:
* wasm/js/WebAssemblyLinkErrorPrototype.cpp:
* wasm/js/WebAssemblyModulePrototype.cpp:
* wasm/js/WebAssemblyRuntimeErrorConstructor.cpp:
* wasm/js/WebAssemblyRuntimeErrorPrototype.cpp:
2020-06-03 Rob Buis <rbuis@igalia.com>
Make generated C++ code use modern C++
https://bugs.webkit.org/show_bug.cgi?id=190714
Reviewed by Jonathan Bedard.
Update inspector protocol generator and rebaseline the tests.
* inspector/scripts/codegen/cpp_generator_templates.py:
* inspector/scripts/codegen/generate_cpp_protocol_types_header.py:
* inspector/scripts/tests/expected/commands-with-async-attribute.json-result:
* inspector/scripts/tests/expected/commands-with-optional-call-return-parameters.json-result:
* inspector/scripts/tests/expected/domains-with-varying-command-sizes.json-result:
* inspector/scripts/tests/expected/enum-values.json-result:
* inspector/scripts/tests/expected/events-with-optional-parameters.json-result:
* inspector/scripts/tests/expected/same-type-id-different-domain.json-result:
* inspector/scripts/tests/expected/type-declaration-aliased-primitive-type.json-result:
* inspector/scripts/tests/expected/type-declaration-array-type.json-result:
* inspector/scripts/tests/expected/type-declaration-enum-type.json-result:
* inspector/scripts/tests/expected/type-declaration-object-type.json-result:
* inspector/scripts/tests/expected/type-requiring-runtime-casts.json-result:
* yarr/generateYarrUnicodePropertyTables.py:
2020-06-02 Mark Lam <mark.lam@apple.com>
Rolling out r262475 to unbreak Windows bot.
https://bugs.webkit.org/show_bug.cgi?id=212680
Not reviewed.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* assembler/MacroAssemblerARM64.h:
* assembler/MacroAssemblerX86_64.h:
* assembler/testmasm.cpp:
(JSC::testCountTrailingZeros64WithoutNullCheck):
(JSC::run):
(JSC::testStore64Imm64AddressPointer): Deleted.
* dfg/DFGDoesGCCheck.cpp: Removed.
* dfg/DFGDoesGCCheck.h: Removed.
* dfg/DFGGraph.cpp:
* dfg/DFGOSRExit.cpp:
(JSC::DFG::operationCompileOSRExit):
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
(JSC::FTL::operationCompileFTLOSRExit):
* heap/CompleteSubspace.cpp:
(JSC::CompleteSubspace::tryAllocateSlow):
(JSC::CompleteSubspace::reallocatePreciseAllocationNonVirtual):
* heap/CompleteSubspaceInlines.h:
(JSC::CompleteSubspace::allocateNonVirtual):
* heap/DeferGC.h:
(JSC::DeferGC::~DeferGC):
* heap/GCDeferralContextInlines.h:
(JSC::GCDeferralContext::~GCDeferralContext):
* heap/Heap.cpp:
(JSC::Heap::collectNow):
(JSC::Heap::collectAsync):
(JSC::Heap::collectSync):
(JSC::Heap::stopIfNecessarySlow):
(JSC::Heap::collectIfNecessaryOrDefer):
* heap/Heap.h:
(JSC::Heap::expectDoesGC const):
(JSC::Heap::setExpectDoesGC):
(JSC::Heap::addressOfExpectDoesGC):
(JSC::Heap::addressOfDoesGC): Deleted.
(JSC::Heap::setDoesGCExpectation): Deleted.
(JSC::Heap::verifyCanGC): Deleted.
* heap/HeapInlines.h:
(JSC::Heap::acquireAccess):
(JSC::Heap::stopIfNecessary):
* heap/LocalAllocatorInlines.h:
(JSC::LocalAllocator::allocate):
* heap/PreciseAllocation.cpp:
(JSC::PreciseAllocation::tryCreate):
(JSC::PreciseAllocation::createForLowerTier):
* runtime/JSString.h:
(JSC::jsSingleCharacterString):
(JSC::JSString::toAtomString const):
(JSC::JSString::toExistingAtomString const):
(JSC::JSString::value const):
(JSC::JSString::tryGetValue const):
(JSC::JSRopeString::unsafeView const):
(JSC::JSRopeString::viewWithUnderlyingString const):
(JSC::JSString::unsafeView const):
* runtime/RegExpMatchesArray.h:
(JSC::createRegExpMatchesArray):
2020-06-02 Mark Lam <mark.lam@apple.com>
Enhance DoesGC verification to print more useful info when verification fails.
https://bugs.webkit.org/show_bug.cgi?id=212680
Reviewed by Yusuke Suzuki.
When DoesGC verification fails, the first step of debugging it would be to find
out what and which DFG node resulted in the failed verification. In pre-existing
code, all we get is an assertion failure.
This patch makes it so that the verifier will dump useful info. Here's an example:
Error: DoesGC failed @ D@34 DateGetInt32OrNaN in #DtCHMz:[0x1135bd1d0->0x1135bcab0->0x1135e5c80, DFGFunctionCall, 150 (DidTryToEnterInLoop)]
[0] frame 0x7ffee8285660 {
name:
sourceURL:
isInlinedFrame: false
callee: 0x1135f6820
returnPC: 0x50ce61248ae6
callerFrame: 0x7ffee82856f0
rawLocationBits: 5 0x5
codeBlock: 0x1135bd1d0 #DtCHMz:[0x1135bd1d0->0x1135bcab0->0x1135e5c80, DFGFunctionCall, 150 (DidTryToEnterInLoop)]
hasCodeOrigins: true
callSiteIndex: 5 of 13
jitCode: 0x113020200 start 0x50ce61214c60 end 0x50ce61219b00
line: 1
column: 60
EntryFrame: 0x7ffee8285860
}
[1] frame 0x7ffee82856f0 {
name:
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1135f65a0
returnPC: 0x50ce61227e99
callerFrame: 0x7ffee8285770
rawLocationBits: 4 0x4
codeBlock: 0x1135bd0a0 #BU6Zcd:[0x1135bd0a0->0x1135bc260->0x1135e5180, DFGFunctionCall, 112 (DidTryToEnterInLoop)]
hasCodeOrigins: true
callSiteIndex: 4 of 12
jitCode: 0x113004000 start 0x50ce61212c60 end 0x50ce61214960
line: 26
column: 22
EntryFrame: 0x7ffee8285860
}
[2] frame 0x7ffee8285770 {
name:
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1135f64e0
returnPC: 0x108058eb1
callerFrame: 0x7ffee82857e0
rawLocationBits: 1001 0x3e9
codeBlock: 0x1135bc130 #DAS9xe:[0x1135bc130->0x1135e5100, BaselineFunctionCall, 1149]
bc#1001 of 1149
line: 417
column: 38
EntryFrame: 0x7ffee8285860
}
[3] frame 0x7ffee82857e0 {
name: global code
sourceURL: date-format-xparb.js
isInlinedFrame: false
callee: 0x1130f97b8
returnPC: 0x108039043
callerFrame: 0x0
rawLocationBits: 23 0x17
codeBlock: 0x1135bc000 <global>#CukXvt:[0x1135bc000->0x1130cd768, LLIntGlobal, 81]
bc#23 of 81
line: 425
column: 3
EntryFrame: 0x7ffee8285860
}
ASSERTION FAILED: expectDoesGC()
The error message now comes with the node index, NodeType, codeBlock which this
failure was found in, and the JS call stack that led to the failure.
Changes made:
1. Introduced a DoesGCCheck value that is used to encode some of the above data.
Previously, we only recorded whether doesGC() returns true or false for the
Node. Now, we record the nodeIndex and nodeOp as well.
Note that we also set DoesGC expectations for OSR exits. So, DoesGCCheck
includes Special cases for those.
2. Added store64(TrustedImm64 imm, const void* address) emitters for X86_64 and ARM64.
Also added a test for this new emitter in testmasm.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::store64):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::store64):
* assembler/testmasm.cpp:
(JSC::testStore64Imm64AddressPointer):
(JSC::run):
* dfg/DFGDoesGCCheck.cpp: Added.
(JSC::DFG::DoesGCCheck::verifyCanGC):
* dfg/DFGDoesGCCheck.h: Added.
(JSC::DFG::DoesGCCheck::DoesGCCheck):
(JSC::DFG::DoesGCCheck::encode):
(JSC::DFG::DoesGCCheck::set):
(JSC::DFG::DoesGCCheck::expectDoesGC):
(JSC::DFG::DoesGCCheck::special):
(JSC::DFG::DoesGCCheck::nodeIndex):
(JSC::DFG::DoesGCCheck::nodeOp):
(JSC::DFG::DoesGCCheck::isSpecial):
(JSC::DFG::DoesGCCheck::specialIndex):
(JSC::DFG::DoesGCCheck::bits):
* dfg/DFGGraph.cpp:
* dfg/DFGOSRExit.cpp:
(JSC::DFG::operationCompileOSRExit):
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
(JSC::FTL::operationCompileFTLOSRExit):
* heap/CompleteSubspace.cpp:
(JSC::CompleteSubspace::tryAllocateSlow):
(JSC::CompleteSubspace::reallocatePreciseAllocationNonVirtual):
* heap/CompleteSubspaceInlines.h:
(JSC::CompleteSubspace::allocateNonVirtual):
* heap/DeferGC.h:
(JSC::DeferGC::~DeferGC):
* heap/GCDeferralContextInlines.h:
(JSC::GCDeferralContext::~GCDeferralContext):
* heap/Heap.cpp:
(JSC::Heap::collectNow):
(JSC::Heap::collectAsync):
(JSC::Heap::collectSync):
(JSC::Heap::stopIfNecessarySlow):
(JSC::Heap::collectIfNecessaryOrDefer):
* heap/Heap.h:
(JSC::Heap::addressOfDoesGC):
(JSC::Heap::setDoesGCExpectation):
(JSC::Heap::verifyCanGC):
(JSC::Heap::expectDoesGC const): Deleted.
(JSC::Heap::setExpectDoesGC): Deleted.
(JSC::Heap::addressOfExpectDoesGC): Deleted.
* heap/HeapInlines.h:
(JSC::Heap::acquireAccess):
(JSC::Heap::stopIfNecessary):
* heap/LocalAllocatorInlines.h:
(JSC::LocalAllocator::allocate):
* heap/PreciseAllocation.cpp:
(JSC::PreciseAllocation::tryCreate):
(JSC::PreciseAllocation::createForLowerTier):
* runtime/JSString.h:
(JSC::jsSingleCharacterString):
(JSC::JSString::toAtomString const):
(JSC::JSString::toExistingAtomString const):
(JSC::JSString::value const):
(JSC::JSString::tryGetValue const):
(JSC::JSRopeString::unsafeView const):
(JSC::JSRopeString::viewWithUnderlyingString const):
(JSC::JSString::unsafeView const):
* runtime/RegExpMatchesArray.h:
(JSC::createRegExpMatchesArray):
2020-06-02 Mark Lam <mark.lam@apple.com>
VMInspector APIs should be taking a VM* instead of a JSGlobalObject*.
https://bugs.webkit.org/show_bug.cgi?id=212676
Reviewed by Saam Barati and Robin Morisset.
This because:
1. None of the functions currently taking a JSGlobalObject* actually need the
globalObject. All of them need the VM.
2. The role of the VMInspector is to enable inspection of the VM. By requiring
that it be passed a JSGlobalObject*, we were actually preventing the VMInspector
from being used in code that have a VM to inspect but don't have a JSGlobalObject
to use.
The reason I'm choosing to pass VM* instead of VM& is because it makes these
functions trivial to call using lldb interactively. The VMInspector functions
are also intentionally designed so that they can be used for this purpose.
On occasion, I may have to cast literal numbers (addresses) to VM*. Technically,
I could cast a number to VM* and dereference it to get a VM& too. However, at
present, lldb is often buggy and not always reliable with casts. I would like to
lessen the chance that lldb fails on me when I'm deep in the middle of a debugging
session, and have a need to call one of these functions.
* tools/JSDollarVM.cpp:
(JSC::functionGC):
(JSC::functionEdenGC):
(JSC::functionCodeBlockForFrame):
(JSC::codeBlockFromArg):
(JSC::functionDumpCallFrame):
(JSC::functionDumpStack):
* tools/VMInspector.cpp:
(JSC::VMInspector::currentThreadOwnsJSLock):
(JSC::ensureCurrentThreadOwnsJSLock):
(JSC::VMInspector::gc):
(JSC::VMInspector::edenGC):
(JSC::VMInspector::isValidCodeBlock):
(JSC::VMInspector::codeBlockForFrame):
(JSC::VMInspector::dumpCallFrame):
(JSC::VMInspector::dumpStack):
* tools/VMInspector.h:
2020-06-02 Keith Rollin <krollin@apple.com>
Revert FEATURES_DEFINES related changes
https://bugs.webkit.org/show_bug.cgi?id=212664
<rdar://problem/63893033>
Reviewed by Andy Estes.
Bug 262310, Bug 262311, Bug 262318, and Bug 262331 involve changes to
FEATURE_DEFINES and how the values there relate to those found in the
Platform*.h files. Those changes break XCBuild (by removing the
.xcfilelist related to UnifiedSources and the process for generating
them), and so are being reverted.
* Configurations/FeatureDefines.xcconfig:
2020-06-02 Ryan Haddad <ryanhaddad@apple.com>
Unreviewed, reverting r262424.
Caused webkitpy test failure
Reverted changeset:
"Make generated C++ code use modern C++"
https://bugs.webkit.org/show_bug.cgi?id=190714
https://trac.webkit.org/changeset/262424
2020-06-02 Mark Lam <mark.lam@apple.com>
Change Gigacage::Config to use storage in WebConfig::g_config instead of its own.
https://bugs.webkit.org/show_bug.cgi?id=212585
<rdar://problem/63812487>
Reviewed by Yusuke Suzuki.
* assembler/testmasm.cpp:
(JSC::testCagePreservesPACFailureBit):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::cageTypedArrayStorage):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::caged):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::cageConditionally):
* llint/LowLevelInterpreter64.asm:
* runtime/JSCConfig.h:
(JSC::Config::isPermanentlyFrozen):
2020-06-02 Saam Barati <sbarati@apple.com>
MultiDeleteByOffset should not always def
https://bugs.webkit.org/show_bug.cgi?id=212621
<rdar://problem/63824182>
Reviewed by Yusuke Suzuki.
Clobberize used to claim that MultiDeleteByOffset always defd a value.
That's an incorrect modeling of MultiDeleteByOffset though, since it might
have delete misses in its variant list. This would lead us to incorrectly
CSE when we shouldn't. This patch fixes this by saying MultiDeleteByOffset
only defs when all its cases write out a value (are hits).
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGNode.cpp:
(JSC::DFG::MultiDeleteByOffsetData::allVariantsStoreEmpty const):
* dfg/DFGNode.h:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileMultiDeleteByOffset):
2020-06-02 Rob Buis <rbuis@igalia.com>
Make generated C++ code use modern C++
https://bugs.webkit.org/show_bug.cgi?id=190714
Reviewed by Sam Weinig.
Update inspector protocol generator and rebaseline the tests.
* inspector/scripts/codegen/cpp_generator_templates.py:
* inspector/scripts/codegen/generate_cpp_protocol_types_header.py:
* inspector/scripts/tests/expected/commands-with-async-attribute.json-result:
* inspector/scripts/tests/expected/commands-with-optional-call-return-parameters.json-result:
* inspector/scripts/tests/expected/domains-with-varying-command-sizes.json-result:
* inspector/scripts/tests/expected/enum-values.json-result:
* inspector/scripts/tests/expected/events-with-optional-parameters.json-result:
* inspector/scripts/tests/expected/same-type-id-different-domain.json-result:
* inspector/scripts/tests/expected/type-declaration-aliased-primitive-type.json-result:
* inspector/scripts/tests/expected/type-declaration-array-type.json-result:
* inspector/scripts/tests/expected/type-declaration-enum-type.json-result:
* inspector/scripts/tests/expected/type-declaration-object-type.json-result:
* inspector/scripts/tests/expected/type-requiring-runtime-casts.json-result:
* yarr/generateYarrUnicodePropertyTables.py:
2020-06-02 Paulo Matos <pmatos@igalia.com>
Fix assert message formatting
https://bugs.webkit.org/show_bug.cgi?id=212591
Reviewed by Adrian Perez de Castro.
Fixes warning by gcc - lineParts.size() is size_t, %zu should be used.
* runtime/FuzzerPredictions.cpp:
(JSC::FuzzerPredictions::FuzzerPredictions):
2020-06-01 Devin Rousso <drousso@apple.com>
Web Inspector: Graphics: should use the `id` (name) of the animation if it exists
https://bugs.webkit.org/show_bug.cgi?id=212618
Reviewed by Timothy Hatcher.
* inspector/protocol/Animation.json:
- added an optional `name` property to the `Animation.Animation` type
- created a new `Animation.nameChanged` event
2020-06-01 Saam Barati <sbarati@apple.com>
Correct misunderstandings on how ThreadSpecific work
https://bugs.webkit.org/show_bug.cgi?id=212616
Reviewed by Michael Saboff.
There were two misunderstandings I had when writing code using ThreadSpecific
when doing LLInt bytecode buffer caching in Wasm.
1. For ThreadSpecific<Vector>, I was calling Vector's constructor twice
unnecessarily, and incorrectly, since we ended up constructing over an
already constructed Vector for the second call. When doing operator* or
operator-> on a ThreadSpecific<T>, T() is called if it has not been
initialized yet. So there is no need to do manually call the constructor
the second time.
2. There is no need to try to destroy entries for ThreadSpecific manually
since we already run destructors when the thread goes away.
This patch removes code for (1) and (2) both from the Wasm bytecode
buffer and from AssemblerData.
* assembler/AssemblerBuffer.cpp:
(JSC::clearAssembleDataThreadSpecificCache): Deleted.
* assembler/AssemblerBuffer.h:
(JSC::AssemblerBuffer::AssemblerBuffer):
(JSC::AssemblerBuffer::~AssemblerBuffer):
(JSC::AssemblerBuffer::getThreadSpecificAssemblerData): Deleted.
* dfg/DFGWorklist.cpp:
* jit/JITWorklist.cpp:
* wasm/WasmLLIntGenerator.cpp:
(JSC::Wasm::LLIntGenerator::LLIntGenerator):
(JSC::Wasm::clearLLIntThreadSpecificCache): Deleted.
* wasm/WasmLLIntGenerator.h:
* wasm/WasmWorklist.cpp:
2020-06-01 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, fix build failure in ARMv7k
https://bugs.webkit.org/show_bug.cgi?id=212595
* runtime/JSCJSValue.cpp:
(JSC::JSValue::toThisSlowCase const):
2020-06-01 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSBigInt::rightTrim can miss |this| pointer and leads to incorrect GC collection
https://bugs.webkit.org/show_bug.cgi?id=212601
Reviewed by Saam Barati.
This is pretty rare case. But in some optimization level, JSBigInt::rightTrim could store |this| + offset pointer into the stack instead of |this|
and make conservative GC think that |this| JSBigInt is unreachable. We put ensureStillAliveHere(this) to ensure that this is alive.
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::rightTrim):
2020-06-01 Mark Lam <mark.lam@apple.com>
x86.rb's LabelReference.x86LoadOperand()'s address operand should be a pointer type.
https://bugs.webkit.org/show_bug.cgi?id=212603
Reviewed by Saam Barati.
The current implementation mistakenly sets the address type to that of the value
being loaded. I encountered this issue when I was trying to do a loadb from a
global address. Because of this bug, the emitted code was trying do a load using
%al (8 byte register) as the pointer to load from. With this fix, it now loads
from %rax.
* offlineasm/x86.rb:
2020-06-01 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSValue::toThis should not throw exception
https://bugs.webkit.org/show_bug.cgi?id=212595
Reviewed by Mark Lam.
Including WebCore code, there are a lot of code which assume JSValue::toThis should not throw an exception.
This assumption was now broken after making JSBigInt allocation graceful for OOM. But for this particular JSValue::toThis case,
we can make it non-throwing code.
This patch makes JSValue::toThis non-throwing code to fix exception-missing debug assertions.
We ensure that BigIntObject can hold BigInt32 (actually, it can already if toObjectSlowCase path is taken).
* runtime/BigIntObject.cpp:
(JSC::BigIntObject::create):
* runtime/JSCJSValue.cpp:
(JSC::JSValue::toThisSlowCase const):
2020-06-01 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] BigInt operations should handle exception correctly
https://bugs.webkit.org/show_bug.cgi?id=212596
Reviewed by Mark Lam.
Some places miss exception check / explicit scope-release while BigInt operations can now throw an exception.
This patch adds them. They are covered by existing stress tests with Debug build.
* runtime/Operations.h:
(JSC::compareBigIntToOtherPrimitive):
(JSC::compareBigInt32ToOtherPrimitive):
(JSC::jsInc):
(JSC::jsDec):
(JSC::jsBitwiseNot):
2020-05-31 Michael Saboff <msaboff@apple.com>
Consider a Thread Specific Cache for AssemblerBuffers
https://bugs.webkit.org/show_bug.cgi?id=212562
Reviewed by Filip Pizlo.
This patch creates a thread local cache of AssemblerData in the hopes that it will reduce
memory allocation churn. The cache is cleared when a thread is destroyed.
If an AssemblerData is destroyed in another thread, its storage is cached by the
destroying thread.
Made a few changes described below to facilite the swap as well as returning a
clear()'ed AssemblerData back to its original state.
Reviewed by Filip Pizlo.
* assembler/AssemblerBuffer.cpp:
(JSC::threadSpecificAssemblerData):
(JSC::clearAssembleDataThreadSpecificCache):
* assembler/AssemblerBuffer.h:
(JSC::AssemblerData::AssemblerData):
(JSC::AssemblerData::operator=):
The copy constructor and assignment operator now perform complete AssemblerBuffer swaps.
(JSC::AssemblerData::takeBufferIfLarger):
A new method that will conditionally copy the enclosed buffer of the argument to "this"
if the argument's buffer is larger than the current buffer of "this".
(JSC::AssemblerData::~AssemblerData):
(JSC::AssemblerData::clear):
The destructor now calls clear which has been changed to reset the buffer to one with
inline capacity.
(JSC::AssemblerBuffer::AssemblerBuffer):
Take the cached out of line buffer if there is one.
(JSC::AssemblerBuffer::~AssemblerBuffer):
Cache the enclosed out of line buffer if it is larger than the currently cached one.
(JSC::AssemblerBuffer::getThreadSpecificAssemblerData):
* dfg/DFGWorklist.cpp:
* jit/JITWorklist.cpp:
* wasm/WasmWorklist.cpp:
2020-05-31 Mark Lam <mark.lam@apple.com>
Change JSC::Config to use storage in WTF::Config instead of its own.
https://bugs.webkit.org/show_bug.cgi?id=212575
<rdar://problem/63796584>
Reviewed by Yusuke Suzuki.
Since Configs must be rounded up to CeilingOnPageSize, this will save us some
memory since the contents of both Configs do not add up to CeilingOnPageSize.
g_jscConfig is now located at g_wtfConfig.spaceForExtensions.
* runtime/JSCConfig.cpp:
(JSC::Config::disableFreezingForTesting):
(JSC::Config::enableRestrictedOptions):
(JSC::Config::permanentlyFreeze): Deleted.
* runtime/JSCConfig.h:
(JSC::Config::permanentlyFreeze):
(JSC::Config::isPermanentlyFrozen):
(): Deleted.
* runtime/Options.cpp:
(JSC::Options::setOptions):
* tools/JSDollarVM.cpp:
(JSC::functionCallWithStackSize):
2020-05-30 Mark Lam <mark.lam@apple.com>
Rename Signal::BadAccess to Signal::AccessFault.
https://bugs.webkit.org/show_bug.cgi?id=212577
Reviewed by Yusuke Suzuki.
* runtime/VMTraps.cpp:
* wasm/WasmFaultSignalHandler.cpp:
(JSC::Wasm::enableFastMemory):
2020-05-30 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] for-in should allocate new temporary register for base
https://bugs.webkit.org/show_bug.cgi?id=212519
<rdar://problem/63722044>
Reviewed by Saam Barati.
While r262233 keeps for-in's enumerated object in variable register if possible to use this register for heuristics driving an optimization,
for-in body can replace the content of this register during enumeration and confuse enumerator.
Instead, we record Variable information in StructureForInContext. This allows us to detect patterns using heap-variables too.
Further, this patch extends pattern-matching code to support ThisNode too.
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::pushStructureForInScope):
* bytecompiler/BytecodeGenerator.h:
(JSC::Variable::Variable):
(JSC::Variable::isResolved const):
(JSC::Variable::symbolTableConstantIndex const):
(JSC::Variable::ident const):
(JSC::Variable::offset const):
(JSC::Variable::isLocal const):
(JSC::Variable::local const):
(JSC::Variable::isReadOnly const):
(JSC::Variable::isSpecial const):
(JSC::Variable::isConst const):
(JSC::Variable::setIsReadOnly):
(JSC::Variable::operator== const):
(JSC::StructureForInContext::StructureForInContext):
(JSC::StructureForInContext::baseVariable const):
(JSC::StructureForInContext::base const): Deleted.
* bytecompiler/NodesCodegen.cpp:
(JSC::HasOwnPropertyFunctionCallDotNode::emitBytecode):
(JSC::ForInNode::emitBytecode):
* parser/ASTBuilder.h:
(JSC::ASTBuilder::makeFunctionCallNode):
* parser/Nodes.h:
(JSC::ExpressionNode::isThisNode const):
2020-05-30 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, fix JSC debug tests' exception checking
https://bugs.webkit.org/show_bug.cgi?id=212512
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createWithLength):
(JSC::JSBigInt::allocateFor):
2020-05-30 Mark Lam <mark.lam@apple.com>
AssemblyHelpers::callExceptionFuzz() is passing a wrong argument to operationExceptionFuzz().
https://bugs.webkit.org/show_bug.cgi?id=212561
Reviewed by Yusuke Suzuki.
There's 2 possible solution to this issue:
1. Thread the globalObject from all the way up the clients calling into
callExceptionFuzz(), or
2. Introduce a operationExceptionFuzzWithCallFrame() wrapper take receives a VM*
and CallFrame*, and use these to get the lexicalGlobalObject.
This patch applies solution 2.
Solution 1 is too unwieldy because it will cause the threading of the globalObject
argument to fan out to many clients, and almost all of those clients currently
do not need the globalObject. Hence, implementing this solution may incur some
performance penalty in normal code, for the sole benefit of this one fuzzing tool.
Secondly, the exception fuzzer doesn't really care which globalObject is used.
It only cares that an exception is thrown, and we need a globalObject in order to
throw that exception. Hence, there is no benefit to threading the globalObject
down from all the clients.
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::callExceptionFuzz):
* jit/JITOperations.cpp:
* jit/JITOperations.h:
2020-05-29 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSBigInt allocation should be graceful for OOM
https://bugs.webkit.org/show_bug.cgi?id=212512
Reviewed by Mark Lam.
This patch allows JSBigInt's storage allocation to fail gracefully if OOM condition happens.
We thread JSGlobalObject* instead of VM& and throw OOM error if storage allocation failed.
We also rename `JSGlobalObject* globalObject` parameter to `JSGlobalObject* nullOrGlobalObjectForOOM`
if it can be nullptr.
* jit/JITOperations.cpp:
* jsc.cpp:
(functionCreateHeapBigInt):
* parser/ParserArena.cpp:
(JSC::IdentifierArena::makeBigIntDecimalIdentifier):
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::callBigIntConstructor):
* runtime/BigIntPrototype.cpp:
(JSC::toThisBigIntValue):
(JSC::bigIntProtoFuncToString):
(JSC::bigIntProtoFuncToLocaleString):
(JSC::bigIntProtoFuncValueOf):
* runtime/CachedTypes.cpp:
(JSC::CachedBigInt::decode const):
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/IntlNumberFormatPrototype.cpp:
(JSC::IntlNumberFormatFuncFormat):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createZero):
(JSC::JSBigInt::tryCreateZero):
(JSC::JSBigInt::createWithLength):
(JSC::JSBigInt::tryCreateWithLength):
(JSC::JSBigInt::createFrom):
(JSC::JSBigInt::tryCreateFrom):
(JSC::JSBigInt::createFromImpl):
(JSC::JSBigInt::parseInt):
(JSC::HeapBigIntImpl::toHeapBigInt):
(JSC::Int32BigIntImpl::toHeapBigInt):
(JSC::zeroImpl):
(JSC::JSBigInt::exponentiateImpl):
(JSC::JSBigInt::multiplyImpl):
(JSC::JSBigInt::divideImpl):
(JSC::JSBigInt::copy):
(JSC::JSBigInt::unaryMinusImpl):
(JSC::JSBigInt::unaryMinus):
(JSC::JSBigInt::remainderImpl):
(JSC::JSBigInt::incImpl):
(JSC::JSBigInt::decImpl):
(JSC::JSBigInt::addImpl):
(JSC::JSBigInt::subImpl):
(JSC::JSBigInt::bitwiseAndImpl):
(JSC::JSBigInt::bitwiseOrImpl):
(JSC::JSBigInt::bitwiseXorImpl):
(JSC::JSBigInt::absoluteAdd):
(JSC::JSBigInt::absoluteSub):
(JSC::JSBigInt::absoluteDivWithDigitDivisor):
(JSC::JSBigInt::absoluteDivWithBigIntDivisor):
(JSC::JSBigInt::absoluteLeftShiftAlwaysCopy):
(JSC::JSBigInt::absoluteBitwiseOp):
(JSC::JSBigInt::absoluteAnd):
(JSC::JSBigInt::absoluteOr):
(JSC::JSBigInt::absoluteAndNot):
(JSC::JSBigInt::absoluteXor):
(JSC::JSBigInt::absoluteAddOne):
(JSC::JSBigInt::absoluteSubOne):
(JSC::JSBigInt::leftShiftByAbsolute):
(JSC::JSBigInt::rightShiftByAbsolute):
(JSC::JSBigInt::rightShiftByMaximum):
(JSC::JSBigInt::toStringBasePowerOfTwo):
(JSC::JSBigInt::toStringGeneric):
(JSC::JSBigInt::rightTrim):
(JSC::JSBigInt::tryRightTrim):
(JSC::JSBigInt::allocateFor):
(JSC::JSBigInt::asIntNImpl):
(JSC::JSBigInt::asUintNImpl):
(JSC::JSBigInt::truncateToNBits):
(JSC::JSBigInt::truncateAndSubFromPowerOfTwo):
(JSC::JSBigInt::createWithLengthUnchecked): Deleted.
* runtime/JSBigInt.h:
* runtime/JSCJSValue.cpp:
(JSC::JSValue::toThisSlowCase const):
* runtime/VM.cpp:
(JSC::VM::VM):
2020-05-29 Saam Barati <sbarati@apple.com>
We need to properly model heap ranges of Delete in DFG/B3
https://bugs.webkit.org/show_bug.cgi?id=212538
<rdar://problem/63670964>
Reviewed by Filip Pizlo.
We need to properly model the aliasing dependencies of an inlined delete
operation.
We had a bug in the B3 IR we generated from code like this for a delete
followed by a property addition:
```
const o = { y: 0 };
delete o.y;
o.z = 0;
```
generated:
```
note: bb#5 dominates bb#10, bb#10 dominates bb#15
bb#5
Void b@125 = Store($-562949953421312(b@282), b@112, offset = 16, ControlDependent|Writes:129, D@30)
bb#10
Void b@171 = Store($0(b@2), b@112, offset = 16, ControlDependent|Writes:129, D@37)
bb#15
Void b@217 = Store($-562949953421312(b@282), b@112, offset = 16, ControlDependent|Writes:130, D@44)
```
Notice that "y" and "z" ended up at the same property offset.
In the above program, B3 proves the pointer we're storing to is the same value
in all three stores (b@112). However, because of how it does store forwarding,
it determined it could eliminate b@217 because b@125 already stored the same
value to the same pointer. It didn't know that b@171 was a write because its
heap range is different than @217. Generally, when using two heap ranges, it's
telling B3 that two pointers don't alias.
```
@A, Heap_H
@B, Heap_H
```
In the above program,
- If @B reads H and @A writes H, then @B is dependent on @A.
- If @B writes H, then @B is dependent on @A if @A reads or writes H.
So for delete, we need to model the deletion of a property as actually
writing to all named properties that may exist at that slot given a
series of structure transitions. We model this by saying the PutStructure
for an inlined delete, or MultiDeleteByOffset, writes to all named properties
(which is a superset of all named properties that may exist at that slot
through a series of transitions).
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* ftl/FTLAbstractHeap.cpp:
(JSC::FTL::IndexedAbstractHeap::dump):
(JSC::FTL::NumberedAbstractHeap::dump):
(JSC::FTL::AbsoluteAbstractHeap::dump):
(JSC::FTL::IndexedAbstractHeap::dump const): Deleted.
(JSC::FTL::NumberedAbstractHeap::dump const): Deleted.
(JSC::FTL::AbsoluteAbstractHeap::dump const): Deleted.
* ftl/FTLAbstractHeap.h:
(JSC::FTL::IndexedAbstractHeap::atAnyIndex):
(JSC::FTL::NumberedAbstractHeap::atAnyNumber):
(JSC::FTL::AbsoluteAbstractHeap::atAnyAddress):
(JSC::FTL::IndexedAbstractHeap::atAnyIndex const): Deleted.
(JSC::FTL::NumberedAbstractHeap::atAnyNumber const): Deleted.
(JSC::FTL::AbsoluteAbstractHeap::atAnyAddress const): Deleted.
* ftl/FTLAbstractHeapRepository.cpp:
(JSC::FTL::AbstractHeapRepository::AbstractHeapRepository):
* ftl/FTLAbstractHeapRepository.h:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compilePutStructure):
(JSC::FTL::DFG::LowerDFGToB3::compileMultiDeleteByOffset):
2020-05-29 Andy Estes <aestes@apple.com>
[Apple Pay] Remove conditionals for ENABLE_APPLE_PAY_SESSION_V(3|4)
https://bugs.webkit.org/show_bug.cgi?id=212541
Reviewed by Darin Adler.
APPLE_PAY_SESSION_V(3|4) is now enabled whenever APPLE_PAY itself is enabled.
* Configurations/FeatureDefines.xcconfig:
2020-05-29 Darin Adler <darin@apple.com>
Remove things from FeatureDefines.xcconfig that are covered by PlatformEnableCocoa.h
https://bugs.webkit.org/show_bug.cgi?id=212418
Rubber-stamped by Simon Fraser.
* Configurations/FeatureDefines.xcconfig: Add back ENABLE_CSS_CONIC_GRADIENTS, removed
by accident.
2020-05-27 Darin Adler <darin@apple.com>
Remove things from FeatureDefines.xcconfig that are covered by PlatformEnableCocoa.h
https://bugs.webkit.org/show_bug.cgi?id=212418
Reviewed by Andy Estes.
* Configurations/FeatureDefines.xcconfig: Removed 83 of the 119 things defined in
this file. There are 36 more that are slightly more complex that we can remove
carefully later.
2020-05-29 Darin Adler <darin@apple.com>
[Cocoa] Pass all defines from Platform.h to various scripts, not just the ones from .xcconfig
https://bugs.webkit.org/show_bug.cgi?id=212451
Reviewed by Sam Weinig.
* DerivedSources.make: Run the preprocessor on Platform.h and parse the output into
FEATURE_AND_PLATFORM_DEFINES. Use that and FEATURE_AND_PLATFORM_DEFINE_DEPENDENCIES
whenever we need a list of defines. Also took out some Windows-specific stuff since
this is now only used on Mac platforms. Use ":=" when calling $(shell) to make sure
the same shell command is not invoked over and over again.
2020-05-29 Keith Rollin <krollin@apple.com>
Revert switch to XCBuild
https://bugs.webkit.org/show_bug.cgi?id=212530
<rdar://problem/63764632>
Unreviewed build fix.
Bug 209890 enabled the use of XCBuild by default. Since then, some
build issues have shown up. While addressing them, temporarily turn
off the use of XCBuild by default.
* Configurations/JavaScriptCore.xcconfig:
* JavaScriptCore.xcodeproj/project.pbxproj:
2020-05-29 Commit Queue <commit-queue@webkit.org>
Unreviewed, reverting r262245.
https://bugs.webkit.org/show_bug.cgi?id=212531
"Caused WebCore's 'Check .xcfilelists' build phase to be ~100x
slower"
Reverted changeset:
"[Cocoa] Pass all defines from Platform.h to various scripts,
not just the ones from .xcconfig"
https://bugs.webkit.org/show_bug.cgi?id=212451
https://trac.webkit.org/changeset/262245
2020-05-29 Devin Rousso <drousso@apple.com>
Web Inspector: add ITML debuggable/target type
https://bugs.webkit.org/show_bug.cgi?id=203300
<rdar://problem/56545896>
Reviewed by Joseph Pecoraro and Brian Burg.
* API/JSContextPrivate.h:
* API/JSContext.mm:
(-[JSContext _setITMLDebuggableType]): Added.
* runtime/JSGlobalObject.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::setIsITML): Added.
Create an SPI for marking a `JSContext` as an ITML context for Web Inspector.
* runtime/JSGlobalObjectDebuggable.h:
(isType):
* inspector/remote/RemoteControllableTarget.h:
* inspector/remote/RemoteInspectionTarget.h:
* inspector/remote/RemoteInspectorConstants.h:
* inspector/JSGlobalObjectInspectorController.cpp:
(Inspector::JSGlobalObjectInspectorController::connectFrontend):
Don't dispatch `Inspector.activateExtraDomains` unless we're a basic `JavaScript` debuggable.
* inspector/remote/cocoa/RemoteInspectorCocoa.mm:
(Inspector::RemoteInspector::listingForInspectionTarget const):
* inspector/scripts/codegen/models.py:
(validate_target_types):
* inspector/scripts/codegen/objc_generator.py:
(ObjCGenerator):
* inspector/scripts/tests/expected/fail-on-command-targetTypes-value.json-error:
* inspector/scripts/tests/expected/fail-on-domain-debuggableTypes-value.json-error:
* inspector/scripts/tests/expected/fail-on-domain-targetTypes-value.json-error:
* inspector/scripts/tests/expected/fail-on-event-targetTypes-value.json-error:
* inspector/protocol/Audit.json:
* inspector/protocol/CSS.json:
* inspector/protocol/Console.json:
* inspector/protocol/DOM.json:
* inspector/protocol/DOMStorage.json:
* inspector/protocol/Database.json:
* inspector/protocol/Debugger.json:
* inspector/protocol/Heap.json:
* inspector/protocol/Inspector.json:
* inspector/protocol/Network.json:
* inspector/protocol/Page.json:
* inspector/protocol/Runtime.json:
* inspector/protocol/Security.json:
Add support for `itml` debuggables and targets, marking non-ITML commands/events with `page`.
2020-05-29 Mark Lam <mark.lam@apple.com>
Add a check for errors when computing a utf string in jsc shell's runInteractive().
https://bugs.webkit.org/show_bug.cgi?id=212526
<rdar://problem/63757892>
Reviewed by Michael Saboff.
* jsc.cpp:
(runInteractive):
2020-05-28 Devin Rousso <drousso@apple.com>
Web Inspector: add missing condition guards when generating objc protocol files
https://bugs.webkit.org/show_bug.cgi?id=212494
Reviewed by Timothy Hatcher.
* inspector/scripts/codegen/generate_objc_backend_dispatcher_header.py:
(ObjCBackendDispatcherHeaderGenerator._generate_objc_forward_declarations):
* inspector/scripts/codegen/generate_objc_configuration_header.py:
(ObjCConfigurationHeaderGenerator._generate_properties_for_domain):
* inspector/scripts/codegen/generate_objc_configuration_implementation.py:
(ObjCConfigurationImplementationGenerator._generate_configuration_implementation_for_domains):
(ObjCConfigurationImplementationGenerator._generate_ivars):
(ObjCConfigurationImplementationGenerator._generate_dealloc):
(ObjCConfigurationImplementationGenerator._generate_event_dispatcher_getter_for_domain):
(ObjCConfigurationImplementationGenerator._variable_name_prefix_for_domain): Deleted.
* inspector/scripts/codegen/generate_objc_frontend_dispatcher_implementation.py:
(ObjCFrontendDispatcherImplementationGenerator._generate_event_dispatcher_implementations):
(ObjCFrontendDispatcherImplementationGenerator._generate_event):
* inspector/scripts/codegen/generate_objc_header.py:
(ObjCHeaderGenerator._generate_forward_declarations):
(ObjCHeaderGenerator._generate_enums):
(ObjCHeaderGenerator._generate_types):
(ObjCHeaderGenerator._generate_command_protocols):
(ObjCHeaderGenerator._generate_event_interfaces):
(ObjCHeaderGenerator._generate_single_event_interface):
* inspector/scripts/codegen/generate_objc_internal_header.py:
(ObjCInternalHeaderGenerator._generate_event_dispatcher_private_interfaces):
* inspector/scripts/codegen/generate_objc_protocol_type_conversions_header.py:
(ObjCProtocolTypeConversionsHeaderGenerator._generate_enum_conversion_functions):
* inspector/scripts/tests/definitions-with-mac-platform.json:
* inspector/scripts/tests/expected/definitions-with-mac-platform.json-result:
2020-05-27 Keith Miller <keith_miller@apple.com>
for-of should check the iterable is a JSArray for FastArray in DFG iterator_open
https://bugs.webkit.org/show_bug.cgi?id=212383
Reviewed by Saam Barati.
This patch fixes an issue where we didn't check that the iterable operand to
iterator_open was a JSArray when lowering the FastArray only variant of the bytecode to the DFG.
This meant we would OSR exit at the iterator_next's lowering then assertion failure in the
checkpoint OSR exit helper. To make this work, this patch extends (and renames from CheckSubClass)
CheckJSCast to use the same JSType information that we use for the jsCast function. In order to
get the JSType range from a ClassInfo* the macro that autogenerates MethodTable now also fills
a Optional<JSTypeRange> into the ClassInfo as well.
Lastly, speculationFromClassInfo was misused by AI. This patch
renames it to speculationFromClassInfoInheritance to better
reflect how AI was using it. The only other user of
speculationFromClassInfo was speculationFromStructure. Any case
where speculationFromClassInfoInteritance would differ from what a
Structure would tell you has been hoisted to
speculationFromStructure.
* assembler/testmasm.cpp:
(JSC::testBranchIfType):
(JSC::testBranchIfNotType):
* bytecode/SpeculatedType.cpp:
(JSC::speculationFromClassInfoInheritance):
(JSC::speculationFromStructure):
(JSC::speculationFromClassInfo): Deleted.
* bytecode/SpeculatedType.h:
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGAbstractValue.cpp:
(JSC::DFG::AbstractValue::filterClassInfo):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleDOMJITGetter):
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::attemptToMakeCallDOM):
(JSC::DFG::FixupPhase::fixupCheckJSCast):
(JSC::DFG::FixupPhase::fixupCheckSubClass): Deleted.
* dfg/DFGNode.h:
(JSC::DFG::Node::hasClassInfo const):
* dfg/DFGNodeType.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::checkArray):
(JSC::DFG::SpeculativeJIT::compileCheckJSCast):
(JSC::DFG::SpeculativeJIT::compileCheckSubClass): Deleted.
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckJSCast):
(JSC::FTL::DFG::LowerDFGToB3::isCellWithType):
(JSC::FTL::DFG::LowerDFGToB3::isTypedArrayView):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckSubClass): Deleted.
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::branchIfType):
(JSC::AssemblyHelpers::branchIfNotType):
* runtime/ClassInfo.h:
* runtime/JSCast.h:
(JSC::JSTypeRange::contains const):
(JSC::JSCastingHelpers::inheritsJSTypeImpl):
* runtime/JSFunction.cpp:
(JSC::JSFunction::assertTypeInfoFlagInvariants):
* tools/JSDollarVM.cpp:
(JSC::functionCreateDOMJITCheckJSCastObject):
(JSC::JSDollarVM::finishCreation):
(JSC::functionCreateDOMJITCheckSubClassObject): Deleted.
2020-05-27 Darin Adler <darin@apple.com>
[Cocoa] Pass all defines from Platform.h to various scripts, not just the ones from .xcconfig
https://bugs.webkit.org/show_bug.cgi?id=212451
Reviewed by Sam Weinig.
* DerivedSources.make: Run the preprocessor on Platform.h and parse the output into
FEATURE_AND_PLATFORM_DEFINES. Use that and FEATURE_AND_PLATFORM_DEFINE_DEPENDENCIES
whenever we need a list of defines. Also took out some Windows-specific stuff since
this is now only used on Mac platforms.
2020-05-28 Michael Catanzaro <mcatanzaro@gnome.org>
Web Inspector: generate_cpp_protocol_types_header.py:294: SyntaxWarning: "is" with a literal. Did you mean "=="?
https://bugs.webkit.org/show_bug.cgi?id=212468
Reviewed by Timothy Hatcher.
Use "==" instead of "is" to compare against 0.
* inspector/scripts/codegen/generate_cpp_protocol_types_header.py:
2020-05-28 Mark Lam <mark.lam@apple.com>
Gardening: Add an assertNoException() to placate the exception checker and green the bots.
https://bugs.webkit.org/show_bug.cgi?id=212248
Not reviewed.
This solution was pointed out by Caio Lima in https://bugs.webkit.org/show_bug.cgi?id=212248#c10.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
2020-05-27 Saam Barati <sbarati@apple.com>
hasOwnProperty inside structure property for-in loop should use an opcode like has_structure_property but for hasOwnProperty
https://bugs.webkit.org/show_bug.cgi?id=212248
Reviewed by Keith Miller.
This patch applies the same principles from r262083 but to hasOwnProperty.
In this patch, we have a fast path for this syntactic pattern when
iterating structure properties:
for (let <p> in <o>)
if (<o>.hasOwnProperty(<p>))
We look for both <p> and <o> as resolve nodes, and we look for them being the
same values both in the header and inside the body.
Using a simple static analysis, when we detect this pattern, we compare the
result of `<o>.hasOwnProperty` to the original hasOwnProperty function. If
it's the same, we execute the fast path new bytecode has_own_structure_property,
which on the fast path is two loads, a compare and branch, and a materialization of
the boolean constant true.
On the slow path, has_own_structure_property just executes the runtime code
for hasOwnProperty.
In my testing, this seems like it might be 3-5% faster on Speedometer 2's
react subtests. I was getting some noise when running the tests locally,
so I can't say for certain it's a definite speedup. But the data implies
it has a good chance at being a speedup.
* builtins/BuiltinNames.h:
* bytecode/BytecodeList.rb:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecode/LinkTimeConstant.h:
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitWideJumpIfNotFunctionHasOwnProperty):
(JSC::BytecodeGenerator::recordHasOwnStructurePropertyInForInLoop):
(JSC::BytecodeGenerator::emitHasOwnStructureProperty):
(JSC::BytecodeGenerator::pushStructureForInScope):
(JSC::StructureForInContext::finalize):
(JSC::BytecodeGenerator::findStructureForInContext):
* bytecompiler/BytecodeGenerator.h:
(JSC::StructureForInContext::StructureForInContext):
(JSC::StructureForInContext::base const):
(JSC::StructureForInContext::addHasOwnPropertyJump):
* bytecompiler/Label.h:
(JSC::GenericBoundLabel::GenericBoundLabel):
* bytecompiler/NodesCodegen.cpp:
(JSC::HasOwnPropertyFunctionCallDotNode::emitBytecode):
(JSC::ForInNode::emitBytecode):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGNodeType.h:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileHasStructureProperty):
(JSC::DFG::SpeculativeJIT::compileHasOwnStructurePropertyImpl):
(JSC::DFG::SpeculativeJIT::compileHasOwnStructureProperty):
(JSC::DFG::SpeculativeJIT::compileInStructureProperty):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileHasOwnStructureProperty):
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):
* jit/JIT.h:
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_has_structure_propertyImpl):
(JSC::JIT::emit_op_has_structure_property):
(JSC::JIT::emit_op_has_own_structure_property):
(JSC::JIT::emit_op_in_structure_property):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::emit_op_has_structure_propertyImpl):
(JSC::JIT::emit_op_has_structure_property):
(JSC::JIT::emit_op_has_own_structure_property):
(JSC::JIT::emit_op_in_structure_property):
* llint/LowLevelInterpreter.asm:
* llint/LowLevelInterpreter64.asm:
* parser/ASTBuilder.h:
(JSC::ASTBuilder::makeFunctionCallNode):
* parser/NodeConstructors.h:
(JSC::HasOwnPropertyFunctionCallDotNode::HasOwnPropertyFunctionCallDotNode):
* parser/Nodes.h:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/CommonSlowPaths.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
* runtime/ObjectPrototype.cpp:
(JSC::objectPrototypeHasOwnProperty):
(JSC::objectProtoFuncHasOwnProperty):
* runtime/ObjectPrototype.h:
2020-05-27 Yusuke Suzuki <ysuzuki@apple.com>
[ macOS iOS ] REGRESSION(r261600?): imported/w3c/web-platform-tests/html/dom/reflection-embedded.html & imported/w3c/web-platform-tests/html/dom/reflection-forms.html are flaky failures
https://bugs.webkit.org/show_bug.cgi?id=212430
Reviewed by Saam Barati.
r261600 added IsConstructor rule to DFG AI. That rule is saying that if the object is not JSFunction and not ProxyObject,
then it must not be a constructor. But this is wrong since any objects can implement getConstructData and DOM constructors
are actually implementing it while it is not JSFunction and it is not ProxyObject. This patch removes that rule.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* tools/JSDollarVM.cpp:
2020-05-27 Saam Barati <sbarati@apple.com>
Limit memory used by wasm/references/multitable.js on memory limited devices
https://bugs.webkit.org/show_bug.cgi?id=212437
Reviewed by Keith Miller.
* tools/JSDollarVM.cpp:
(JSC::functionIsMemoryLimited):
(JSC::JSDollarVM::finishCreation):
2020-05-27 David Kilzer <ddkilzer@apple.com>
REGRESSION: Build errors during JavaScriptCore "Generate Derived Sources" about missing Availability.h header and invalid SDKROOT [-Wmissing-sysroot]
<https://webkit.org/b/212436>
Unreviewed build fix.
* Scripts/generate-derived-sources.sh: Pass SDKROOT to `make`
command from the current environment.
2020-05-27 Mark Lam <mark.lam@apple.com>
Add missing #include for when LLINT_TRACING is enabled.
https://bugs.webkit.org/show_bug.cgi?id=212433
Reviewed by Tadeu Zagallo.
* llint/LLIntExceptions.cpp:
2020-05-27 Devin Rousso <drousso@apple.com>
Web Inspector: replace `featureGuard` and `availability` with a combined `condition` that accepts any macro
https://bugs.webkit.org/show_bug.cgi?id=210014
Reviewed by Brian Burg.
Previously, the generated InspectorBackendCommands.js would include code for things that the
backend doesn't actually support. By using actual macros and preprocessing that file, we can
ensure that the frontend doesn't incorrectly think that something is supported by the page
being inspected:
- the `Canvas` commands and events related to shader programs/pipelines should only exist
when the corresponding context type exists, namely `ENABLE(WEBGL)` and `ENABLE(WEBGPU)`.
- iOS doesn't support showing rulers, so create a variant of `DOM.setInspectModeEnabled`
that only exists for `PLATFORM(IOS_FAMILY)` that doesn't have the `showRulers` optional
parameter, as well as removing `Page.setShowRulers` entirely.
- setting the forced appearance should only be possible if dark mode is supported.
- web archives only exist if CF is used.
* inspector/protocol/CPUProfiler.json:
* inspector/protocol/Canvas.json:
* inspector/protocol/DOM.json:
* inspector/protocol/IndexedDB.json:
* inspector/protocol/Inspector.json:
* inspector/protocol/Memory.json:
* inspector/protocol/Page.json:
* inspector/protocol/ServiceWorker.json:
* Scripts/generate-derived-sources.sh:
Set `CC` if it hasn't already been set.
* DerivedSources.make:
* DerivedSources-input.xcfilelist:
Preprocess `InspectorBackendCommands.js.in` to get an accurate `InspectorBackendCommands.js`
that follows the logic/description above.
* CMakeLists.txt:
Create a new `InspectorBackendCommands` target now that `InspectorBackendCommands.js` is
generated seprately from the rest of the protocol files.
* Configurations/FeatureDefines.xcconfig:
Add `ENABLE_WEB_ARCHIVE` since it's always enabled in wtf/PlatformEnableCocoa.h.
* inspector/scripts/generate-inspector-protocol-bindings.py:
(generate_from_specification):
(generate_from_specification.load_specification):
* inspector/scripts/codegen/generator.py:
(Generator.__init__):
(Generator.model):
(Generator.set_generator_setting):
(Generator.type_declarations_for_domain):
(Generator.commands_for_domain):
(Generator.events_for_domain):
(Generator.wrap_with_guard_for_condition): Added.
(Generator.platform): Deleted.
(Generator.can_generate_platform): Deleted.
(Generator.wrap_with_guard_for_domain): Deleted.
(Generator.wrap_with_guard): Deleted.
* inspector/scripts/codegen/models.py:
(Frameworks):
(Protocol.parse_domain):
(Protocol.parse_type_declaration):
(Protocol.parse_command):
(Protocol.parse_event):
(Domain.__init__):
(TypeDeclaration.__init__):
(Command.__init__):
(Event.__init__):
(Platform): Deleted.
(Platform.__init__): Deleted.
(Platform.fromString): Deleted.
(Platforms): Deleted.
(Platforms.__metaclass__): Deleted.
(Platforms.__metaclass__.__iter__): Deleted.
* inspector/scripts/codegen/generator_templates.py:
Remove `platform` as it is handled by `condition`.
* inspector/scripts/codegen/preprocess.pl: Copied from Source/WebCore/bindings/scripts/preprocessor.pm.
* inspector/scripts/codegen/generate_js_backend_commands.py:
(JSBackendCommandsGenerator.output_filename):
(JSBackendCommandsGenerator.generate_domain):
Output to `InspectorBackendCommands.js.in` that includes `#if` for preprocessing.
* inspector/scripts/codegen/cpp_generator_templates.py:
* inspector/scripts/codegen/generate_cpp_alternate_backend_dispatcher_header.py:
(CppAlternateBackendDispatcherHeaderGenerator.generate_output):
(CppAlternateBackendDispatcherHeaderGenerator._generate_handler_declarations_for_domain):
(CppAlternateBackendDispatcherHeaderGenerator._generate_handler_declaration_for_command):
* inspector/scripts/codegen/generate_cpp_backend_dispatcher_header.py:
(CppBackendDispatcherHeaderGenerator._generate_alternate_handler_forward_declarations_for_domains.Alternate):
(CppBackendDispatcherHeaderGenerator._generate_handler_declarations_for_domain):
(CppBackendDispatcherHeaderGenerator._generate_handler_declaration_for_command):
(CppBackendDispatcherHeaderGenerator._generate_async_handler_declaration_for_command):
(CppBackendDispatcherHeaderGenerator._generate_dispatcher_declarations_for_domain):
(CppBackendDispatcherHeaderGenerator._generate_dispatcher_declaration_for_command):
* inspector/scripts/codegen/generate_cpp_backend_dispatcher_implementation.py:
(CppBackendDispatcherImplementationGenerator.generate_output):
(CppBackendDispatcherImplementationGenerator._generate_handler_class_destructor_for_domain):
(CppBackendDispatcherImplementationGenerator._generate_dispatcher_implementations_for_domain):
(CppBackendDispatcherImplementationGenerator._generate_small_dispatcher_switch_implementation_for_domain):
(CppBackendDispatcherImplementationGenerator._generate_large_dispatcher_switch_implementation_for_domain):
(CppBackendDispatcherImplementationGenerator._generate_async_dispatcher_class_for_domain):
(CppBackendDispatcherImplementationGenerator._generate_dispatcher_implementation_for_command):
* inspector/scripts/codegen/generate_cpp_frontend_dispatcher_header.py:
(CppFrontendDispatcherHeaderGenerator._generate_dispatcher_declarations_for_domain):
(CppFrontendDispatcherHeaderGenerator._generate_dispatcher_declaration_for_event):
* inspector/scripts/codegen/generate_cpp_frontend_dispatcher_implementation.py:
(CppFrontendDispatcherImplementationGenerator._generate_dispatcher_implementations_for_domain):
(CppFrontendDispatcherImplementationGenerator._generate_dispatcher_implementation_for_event):
* inspector/scripts/codegen/generate_cpp_protocol_types_header.py:
(CppProtocolTypesHeaderGenerator._generate_versions):
* inspector/scripts/codegen/generate_cpp_protocol_types_implementation.py:
(CppProtocolTypesImplementationGenerator._generate_enum_conversion_methods_for_domain.generate_conversion_method_body):
(CppProtocolTypesImplementationGenerator._generate_enum_conversion_methods_for_domain):
(CppProtocolTypesImplementationGenerator._generate_open_field_names):
(CppProtocolTypesImplementationGenerator._generate_builders_for_domain):
* inspector/scripts/codegen/objc_generator_templates.py:
* inspector/scripts/codegen/generate_objc_backend_dispatcher_header.py:
(ObjCBackendDispatcherHeaderGenerator._generate_objc_handler_declarations_for_domain):
(ObjCBackendDispatcherHeaderGenerator._generate_objc_handler_declaration_for_command):
* inspector/scripts/codegen/generate_objc_backend_dispatcher_implementation.py:
(ObjCBackendDispatcherImplementationGenerator._generate_handler_implementation_for_domain):
(ObjCBackendDispatcherImplementationGenerator._generate_handler_implementation_for_command):
* inspector/scripts/codegen/generate_objc_header.py:
(add_newline):
(ObjCHeaderGenerator.generate_output):
(ObjCHeaderGenerator._generate_forward_declarations):
(ObjCHeaderGenerator._generate_enums):
(ObjCHeaderGenerator._generate_types):
(ObjCHeaderGenerator._generate_type_interface):
(ObjCHeaderGenerator._generate_command_protocols):
(ObjCHeaderGenerator._generate_single_command_protocol):
(ObjCHeaderGenerator._generate_event_interfaces):
(ObjCHeaderGenerator._generate_single_event_interface):
(ObjCHeaderGenerator._generate_enum_for_platforms): Deleted.
* inspector/scripts/codegen/generate_objc_protocol_type_conversions_header.py:
(add_newline):
(ObjCProtocolTypeConversionsHeaderGenerator.generate_output):
(ObjCProtocolTypeConversionsHeaderGenerator._generate_enum_conversion_functions):
(ObjCProtocolTypeConversionsHeaderGenerator._generate_enum_conversion_for_platforms): Deleted.
* inspector/scripts/codegen/generate_objc_protocol_type_conversions_implementation.py:
(add_newline):
(ObjCProtocolTypeConversionsImplementationGenerator._generate_type_factory_category_interface):
(ObjCProtocolTypeConversionsImplementationGenerator._generate_type_factory_method_declaration):
(ObjCProtocolTypeConversionsImplementationGenerator._generate_type_factory_category_implementation):
(ObjCProtocolTypeConversionsImplementationGenerator._generate_type_factory_method_implementation):
* inspector/scripts/codegen/generate_objc_protocol_types_implementation.py:
(add_newline):
(ObjCProtocolTypesImplementationGenerator.generate_type_implementations):
(ObjCProtocolTypesImplementationGenerator.generate_type_implementation):
Wrap each domain, type, command, and event with the associated `condition` (if it exists).
* inspector/scripts/tests/command-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/command-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/commands-with-async-attribute.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/commands-with-async-attribute.json.
* inspector/scripts/tests/commands-with-optional-call-return-parameters.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/commands-with-optional-call-return-parameters.json.
* inspector/scripts/tests/definitions-with-mac-platform.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/mac/definitions-with-mac-platform.json.
* inspector/scripts/tests/domain-debuggableTypes.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/domain-debuggableTypes.json.
* inspector/scripts/tests/domain-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/domain-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/domain-targetTypes.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/domain-targetTypes.json.
* inspector/scripts/tests/domains-with-varying-command-sizes.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/domains-with-varying-command-sizes.json.
* inspector/scripts/tests/enum-values.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/enum-values.json.
* inspector/scripts/tests/event-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/event-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/events-with-optional-parameters.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/events-with-optional-parameters.json.
* inspector/scripts/tests/expected/command-targetType-matching-domain-debuggableType.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/command-targetType-matching-domain-debuggableType.json-result.
* inspector/scripts/tests/expected/commands-with-async-attribute.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/commands-with-async-attribute.json-result.
* inspector/scripts/tests/expected/commands-with-optional-call-return-parameters.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/commands-with-optional-call-return-parameters.json-result.
* inspector/scripts/tests/expected/definitions-with-mac-platform.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/mac/expected/definitions-with-mac-platform.json-result.
* inspector/scripts/tests/expected/domain-debuggableTypes.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/domain-debuggableTypes.json-result.
* inspector/scripts/tests/expected/domain-targetType-matching-domain-debuggableType.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/domain-targetType-matching-domain-debuggableType.json-result.
* inspector/scripts/tests/expected/domain-targetTypes.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/domain-targetTypes.json-result.
* inspector/scripts/tests/expected/domains-with-varying-command-sizes.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/domains-with-varying-command-sizes.json-result.
* inspector/scripts/tests/expected/enum-values.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/enum-values.json-result.
* inspector/scripts/tests/expected/event-targetType-matching-domain-debuggableType.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/event-targetType-matching-domain-debuggableType.json-result.
* inspector/scripts/tests/expected/events-with-optional-parameters.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/events-with-optional-parameters.json-result.
* inspector/scripts/tests/expected/fail-on-command-targetType-matching-domain-debuggableType.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-command-targetType-matching-domain-debuggableType.json-error.
* inspector/scripts/tests/expected/fail-on-command-targetTypes-type.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-command-targetTypes-type.json-error.
* inspector/scripts/tests/expected/fail-on-command-targetTypes-value.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-command-targetTypes-value.json-error.
* inspector/scripts/tests/expected/fail-on-domain-debuggableTypes-type.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-domain-debuggableTypes-type.json-error.
* inspector/scripts/tests/expected/fail-on-domain-debuggableTypes-value.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-domain-debuggableTypes-value.json-error.
* inspector/scripts/tests/expected/fail-on-domain-targetType-matching-domain-debuggableType.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-domain-targetType-matching-domain-debuggableType.json-error.
* inspector/scripts/tests/expected/fail-on-domain-targetTypes-type.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-domain-targetTypes-type.json-error.
* inspector/scripts/tests/expected/fail-on-domain-targetTypes-value.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-domain-targetTypes-value.json-error.
* inspector/scripts/tests/expected/fail-on-duplicate-command-call-parameter-names.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-duplicate-command-call-parameter-names.json-error.
* inspector/scripts/tests/expected/fail-on-duplicate-command-return-parameter-names.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-duplicate-command-return-parameter-names.json-error.
* inspector/scripts/tests/expected/fail-on-duplicate-event-parameter-names.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-duplicate-event-parameter-names.json-error.
* inspector/scripts/tests/expected/fail-on-duplicate-type-declarations.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-duplicate-type-declarations.json-error.
* inspector/scripts/tests/expected/fail-on-duplicate-type-member-names.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-duplicate-type-member-names.json-error.
* inspector/scripts/tests/expected/fail-on-enum-with-no-values.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-enum-with-no-values.json-error.
* inspector/scripts/tests/expected/fail-on-event-targetType-matching-domain-debuggableType.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-event-targetType-matching-domain-debuggableType.json-error.
* inspector/scripts/tests/expected/fail-on-event-targetTypes-type.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-event-targetTypes-type.json-error.
* inspector/scripts/tests/expected/fail-on-event-targetTypes-value.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-event-targetTypes-value.json-error.
* inspector/scripts/tests/expected/fail-on-number-typed-optional-parameter-flag.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-number-typed-optional-parameter-flag.json-error.
* inspector/scripts/tests/expected/fail-on-number-typed-optional-type-member.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-number-typed-optional-type-member.json-error.
* inspector/scripts/tests/expected/fail-on-string-typed-optional-parameter-flag.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-string-typed-optional-parameter-flag.json-error.
* inspector/scripts/tests/expected/fail-on-string-typed-optional-type-member.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-string-typed-optional-type-member.json-error.
* inspector/scripts/tests/expected/fail-on-type-declaration-using-type-reference.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-type-declaration-using-type-reference.json-error.
* inspector/scripts/tests/expected/fail-on-type-reference-as-primitive-type.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-type-reference-as-primitive-type.json-error.
* inspector/scripts/tests/expected/fail-on-type-with-lowercase-name.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-type-with-lowercase-name.json-error.
* inspector/scripts/tests/expected/fail-on-unknown-type-reference-in-type-declaration.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-unknown-type-reference-in-type-declaration.json-error.
* inspector/scripts/tests/expected/fail-on-unknown-type-reference-in-type-member.json-error: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/fail-on-unknown-type-reference-in-type-member.json-error.
* inspector/scripts/tests/expected/generate-domains-with-feature-guards.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/generate-domains-with-feature-guards.json-result.
* inspector/scripts/tests/expected/same-type-id-different-domain.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/same-type-id-different-domain.json-result.
* inspector/scripts/tests/expected/shadowed-optional-type-setters.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/shadowed-optional-type-setters.json-result.
* inspector/scripts/tests/expected/should-strip-comments.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/should-strip-comments.json-result.
* inspector/scripts/tests/expected/type-declaration-aliased-primitive-type.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-declaration-aliased-primitive-type.json-result.
* inspector/scripts/tests/expected/type-declaration-array-type.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-declaration-array-type.json-result.
* inspector/scripts/tests/expected/type-declaration-enum-type.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-declaration-enum-type.json-result.
* inspector/scripts/tests/expected/type-declaration-object-type.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-declaration-object-type.json-result.
* inspector/scripts/tests/expected/type-requiring-runtime-casts.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-requiring-runtime-casts.json-result.
* inspector/scripts/tests/expected/type-with-open-parameters.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/type-with-open-parameters.json-result.
* inspector/scripts/tests/expected/version.json-result: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/expected/version.json-result.
* inspector/scripts/tests/fail-on-command-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-command-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/fail-on-command-targetTypes-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-command-targetTypes-type.json.
* inspector/scripts/tests/fail-on-command-targetTypes-value.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-command-targetTypes-value.json.
* inspector/scripts/tests/fail-on-domain-debuggableTypes-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-domain-debuggableTypes-type.json.
* inspector/scripts/tests/fail-on-domain-debuggableTypes-value.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-domain-debuggableTypes-value.json.
* inspector/scripts/tests/fail-on-domain-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-domain-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/fail-on-domain-targetTypes-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-domain-targetTypes-type.json.
* inspector/scripts/tests/fail-on-domain-targetTypes-value.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-domain-targetTypes-value.json.
* inspector/scripts/tests/fail-on-duplicate-command-call-parameter-names.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-duplicate-command-call-parameter-names.json.
* inspector/scripts/tests/fail-on-duplicate-command-return-parameter-names.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-duplicate-command-return-parameter-names.json.
* inspector/scripts/tests/fail-on-duplicate-event-parameter-names.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-duplicate-event-parameter-names.json.
* inspector/scripts/tests/fail-on-duplicate-type-declarations.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-duplicate-type-declarations.json.
* inspector/scripts/tests/fail-on-duplicate-type-member-names.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-duplicate-type-member-names.json.
* inspector/scripts/tests/fail-on-enum-with-no-values.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-enum-with-no-values.json.
* inspector/scripts/tests/fail-on-event-targetType-matching-domain-debuggableType.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-event-targetType-matching-domain-debuggableType.json.
* inspector/scripts/tests/fail-on-event-targetTypes-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-event-targetTypes-type.json.
* inspector/scripts/tests/fail-on-event-targetTypes-value.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-event-targetTypes-value.json.
* inspector/scripts/tests/fail-on-number-typed-optional-parameter-flag.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-number-typed-optional-parameter-flag.json.
* inspector/scripts/tests/fail-on-number-typed-optional-type-member.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-number-typed-optional-type-member.json.
* inspector/scripts/tests/fail-on-string-typed-optional-parameter-flag.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-string-typed-optional-parameter-flag.json.
* inspector/scripts/tests/fail-on-string-typed-optional-type-member.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-string-typed-optional-type-member.json.
* inspector/scripts/tests/fail-on-type-declaration-using-type-reference.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-type-declaration-using-type-reference.json.
* inspector/scripts/tests/fail-on-type-reference-as-primitive-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-type-reference-as-primitive-type.json.
* inspector/scripts/tests/fail-on-type-with-lowercase-name.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-type-with-lowercase-name.json.
* inspector/scripts/tests/fail-on-unknown-type-reference-in-type-declaration.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-unknown-type-reference-in-type-declaration.json.
* inspector/scripts/tests/fail-on-unknown-type-reference-in-type-member.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/fail-on-unknown-type-reference-in-type-member.json.
* inspector/scripts/tests/generate-domains-with-feature-guards.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/generate-domains-with-feature-guards.json.
* inspector/scripts/tests/same-type-id-different-domain.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/same-type-id-different-domain.json.
* inspector/scripts/tests/shadowed-optional-type-setters.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/shadowed-optional-type-setters.json.
* inspector/scripts/tests/should-strip-comments.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/should-strip-comments.json.
* inspector/scripts/tests/type-declaration-aliased-primitive-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-declaration-aliased-primitive-type.json.
* inspector/scripts/tests/type-declaration-array-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-declaration-array-type.json.
* inspector/scripts/tests/type-declaration-enum-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-declaration-enum-type.json.
* inspector/scripts/tests/type-declaration-object-type.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-declaration-object-type.json.
* inspector/scripts/tests/type-requiring-runtime-casts.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-requiring-runtime-casts.json.
* inspector/scripts/tests/type-with-open-parameters.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/type-with-open-parameters.json.
* inspector/scripts/tests/version.json: Renamed from Source/JavaScriptCore/inspector/scripts/tests/generic/version.json.
* inspector/scripts/tests/generic/definitions-with-mac-platform.json: Removed.
* inspector/scripts/tests/generic/expected/definitions-with-mac-platform.json-result: Removed.
* inspector/scripts/tests/generic/fail-on-command-with-invalid-platform.json: Removed.
* inspector/scripts/tests/generic/expected/fail-on-command-with-invalid-platform.json-error: Removed.
* inspector/scripts/tests/generic/fail-on-type-with-invalid-platform.json: Removed.
* inspector/scripts/tests/generic/expected/fail-on-type-with-invalid-platform.json-error: Removed.
* inspector/scripts/tests/ios/definitions-with-mac-platform.json: Removed.
* inspector/scripts/tests/ios/expected/definitions-with-mac-platform.json-result: Removed.
* inspector/scripts/tests/all/definitions-with-mac-platform.json: Removed.
* inspector/scripts/tests/all/expected/definitions-with-mac-platform.json-result: Removed.
Don't separate the inspector generator tests by platform.
2020-05-27 Keith Miller <keith_miller@apple.com>
in_structure_property needs to handle constants on the RHS of the "in"
https://bugs.webkit.org/show_bug.cgi?id=212399
Reviewed by Saam Barati.
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
2020-05-27 Keith Rollin <krollin@apple.com>
Enable the use of XCBuild by default in Apple builds
https://bugs.webkit.org/show_bug.cgi?id=209890
<rdar://problem/44182078>
Unreviewed build fix. Check the value of XCODE_VERSION_ACTUAL rather
than XCODE_VERSION_MAJOR when determining whether to use the XCBuild
or non-XCBuild method of running header post-processing scripts.
* JavaScriptCore.xcodeproj/project.pbxproj:
2020-05-26 Mark Lam <mark.lam@apple.com>
Add some new emitters to the X86_64 and ARM64 MacroAssemblers.
https://bugs.webkit.org/show_bug.cgi?id=212385
Reviewed by Robin Morisset.
This patch adds these MacroAssembler emitters:
clearBit64
clearBits64WithMask
countTrailingZeros64WithoutNullCheck
clearBit64 clears a bit.
clearBits64WithMask does the equivalent of and64 with the 1's complement of the
provided mask.
countTrailingZeros64WithoutNullCheck does the same thing as countTrailingZeros64,
except that it assumes that the word in the register it is processing will never
be null, and therefore skips the null check. This is useful in code generation
that already does a null check ahead of time. So, there's no need to do a
redundant null check.
Also added testmasm tests for these emitters.
* assembler/AbortReason.h:
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::clearBit64):
(JSC::MacroAssemblerARM64::clearBits64WithMask):
(JSC::MacroAssemblerARM64::countTrailingZeros64WithoutNullCheck):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::countTrailingZeros64WithoutNullCheck):
(JSC::MacroAssemblerX86_64::clearBit64):
(JSC::MacroAssemblerX86_64::clearBits64WithMask):
* assembler/X86Assembler.h:
(JSC::X86Assembler::btrq_rr):
* assembler/testmasm.cpp:
(JSC::testClearBit64):
(JSC::testClearBits64WithMask):
(JSC::testClearBits64WithMaskTernary):
(JSC::testCountTrailingZeros64Impl):
(JSC::testCountTrailingZeros64):
(JSC::testCountTrailingZeros64WithoutNullCheck):
(JSC::run):
2020-05-26 Yoshiaki JITSUKAWA <yoshiaki.jitsukawa@sony.com>
[PlayStation] Enable RemoteWebInspector
https://bugs.webkit.org/show_bug.cgi?id=212312
Reviewed by Don Olmstead.
* API/JSRemoteInspectorServer.cpp:
Fix compile error.
* PlatformPlayStation.cmake:
Add JSRemoteInspectorServer.h to the public header list.
* inspector/remote/socket/posix/RemoteInspectorSocketPOSIX.cpp:
Set PlayStation specific socket option.
2020-05-26 Alexey Shvayka <shvaikalesh@gmail.com>
IteratorClose should suppress GetMethod errors
https://bugs.webkit.org/show_bug.cgi?id=212378
Reviewed by Keith Miller.
This patch implements recent spec change [1] that prevents "return" method lookup error
from overriding outer exception, aligning JSC with V8 and SpiderMonkey.
It is accomplished by moving pushTry() before emitGetById() in BytecodeGenerator.cpp
(covered by test262 suite) and removal of RETURN_IF_EXCEPTION in IteratorOperations.cpp
(added a stress test).
Before this patch, JSC partly implemented the spec change [1] by suppressing TypeError
if "return" method of iterator was not callable.
BytecodeGenerator::emitDelegateYield() is intentionally left unchanged.
Also, this patch utilizes emitIteratorGenericClose() to avoid code duplication.
for/of microbenchmarks are neutral.
[1]: https://github.com/tc39/ecma262/pull/1408
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitGenericEnumeration):
(JSC::BytecodeGenerator::emitEnumeration):
* runtime/IteratorOperations.cpp:
(JSC::iteratorClose):
2020-05-26 Mark Lam <mark.lam@apple.com>
SamplingProfiler::takeSample() should not assume that ENABLE(WEBASSEMBLY) means Wasm is enabled.
https://bugs.webkit.org/show_bug.cgi?id=212382
Reviewed by Saam Barati.
Wasm can still be disabled at runtime with JSC options. Fixing this will allow
sampling profiler tests to run with JSC_useJIT=0 without crashing.
* runtime/SamplingProfiler.cpp:
(JSC::FrameWalker::FrameWalker):
(JSC::FrameWalker::recordJITFrame):
(JSC::CFrameWalker::CFrameWalker):
(JSC::SamplingProfiler::takeSample):
2020-05-26 Keith Rollin <krollin@apple.com>
Enable the use of XCBuild by default in Apple builds
https://bugs.webkit.org/show_bug.cgi?id=209890
<rdar://problem/44182078>
Reviewed by Darin Adler.
Switch from the "legacy" Xcode build system to the "new" build system
(also known as "XCBuild"). Switching to the new system speeds up
builds by a small percentage, better validates projects for
build-related issues (such as dependency cycles), lets WebKit benefit
from future improvements in XCBuild such as those coming from the
underlying llbuild open source project, and prepares us for any other
tools built for this new ecosystem.
Specific changes:
- Remove Xcode project and workspace settings that selected the Build
system, allowing the default to take hold (which is currently the
New build system).
- Updated webkitdirs.pm with a terser check for Xcode version.
- Update build-webkit and Makefile.shared to be explicit when using
the old build system (no longer treat it as a default or fall-back
configuration).
- Update various xcconfig files similarly to treat the default as
using the new build system.
- Update various post-processing build steps to check for Xcode 11.4
and to no longer treat the default as using the old build system.
* Configurations/JavaScriptCore.xcconfig:
* JavaScriptCore.xcodeproj/project.pbxproj:
2020-05-23 Paulo Matos <pmatos@igalia.com>
Fix non-unified builds for x86_64
https://bugs.webkit.org/show_bug.cgi?id=212297
Reviewed by Adrian Perez de Castro.
* b3/B3BasicBlock.cpp:
* b3/B3CaseCollection.cpp:
* b3/B3DuplicateTails.cpp:
* b3/B3EnsureLoopPreHeaders.cpp:
* b3/B3FenceValue.cpp:
* b3/B3HoistLoopInvariantValues.cpp:
* b3/B3LegalizeMemoryOffsets.cpp:
* b3/B3LowerMacrosAfterOptimizations.cpp:
* b3/B3LowerToAir.cpp:
* b3/B3MathExtras.cpp:
* b3/B3MemoryValue.cpp:
* b3/B3Procedure.cpp:
* b3/B3StackmapValue.cpp:
* b3/B3SwitchValue.cpp:
* b3/B3UseCounts.cpp:
* b3/B3Validate.cpp:
* b3/B3VariableValue.cpp:
* b3/B3WasmAddressValue.cpp:
* b3/B3WasmBoundsCheckValue.cpp:
* ftl/FTLCommonValues.cpp:
* ftl/FTLCompile.cpp:
* ftl/FTLOSREntry.cpp:
* ftl/FTLOSRExitCompiler.cpp:
* wasm/WasmInstance.cpp:
* wasm/WasmStreamingParser.cpp:
* wasm/js/JSToWasm.cpp:
* wasm/js/JSToWasmICCallee.cpp:
* wasm/js/JSWebAssembly.cpp:
* wasm/js/JSWebAssemblyCodeBlock.cpp:
* wasm/js/WebAssemblyCompileErrorConstructor.cpp:
* wasm/js/WebAssemblyCompileErrorPrototype.cpp:
* wasm/js/WebAssemblyFunction.cpp:
* wasm/js/WebAssemblyFunctionBase.cpp:
* wasm/js/WebAssemblyGlobalPrototype.cpp:
* wasm/js/WebAssemblyInstanceConstructor.cpp:
* wasm/js/WebAssemblyInstancePrototype.cpp:
* wasm/js/WebAssemblyLinkErrorConstructor.cpp:
* wasm/js/WebAssemblyLinkErrorPrototype.cpp:
* wasm/js/WebAssemblyMemoryConstructor.cpp:
* wasm/js/WebAssemblyMemoryPrototype.cpp:
* wasm/js/WebAssemblyModuleConstructor.cpp:
* wasm/js/WebAssemblyModulePrototype.cpp:
* wasm/js/WebAssemblyModuleRecord.cpp:
* wasm/js/WebAssemblyRuntimeErrorConstructor.cpp:
* wasm/js/WebAssemblyRuntimeErrorPrototype.cpp:
* wasm/js/WebAssemblyTableConstructor.cpp:
* wasm/js/WebAssemblyTablePrototype.cpp:
2020-05-22 Yoshiaki JITSUKAWA <yoshiaki.jitsukawa@sony.com>
[PlayStation] Enable JSC shell to run
https://bugs.webkit.org/show_bug.cgi?id=212294
Reviewed by Ross Kirsling.
* shell/PlatformPlayStation.cmake:
Set working directory for Visual Studio
* shell/playstation/Initializer.cpp:
Load libJavaScriptCore as we now build it as SHARED.
2020-05-22 Alexey Shvayka <shvaikalesh@gmail.com>
Array.prototype.splice doesn't set "length" of returned object
https://bugs.webkit.org/show_bug.cgi?id=212285
Reviewed by Darin Adler.
This change implements step 12 of Array.prototype.splice [1], which is observable
if result object is not JSArray, aligning JSC with V8 and SpiderMonkey.
Only slow path of splice() is affected by this patch; zero-argument case already
performs setLength(). Microbenchmarks are neutral.
[1]: https://tc39.es/ecma262/#sec-array.prototype.splice
* runtime/ArrayPrototype.cpp:
(JSC::arrayProtoFuncSplice):
2020-05-22 Saam Barati <sbarati@apple.com>
in_by_val inside structure property for-in loop should use an opcode like has_structure_property but for "in"
https://bugs.webkit.org/show_bug.cgi?id=212239
Reviewed by Tadeu Zagallo.
There is code inside Speedometer 2 that is like:
```
for (let p in o) {
if (p in o2)
...
}
```
Where o and o2 frequently share the same structure. Elm does this when it's
diffing two objects. We already optimize o2[p] (get_by_val) in the above loop
for structure properties. This patch adds that same optimization for in_by_val.
Because we already emit a "structure" loop for for-in, where we iterate structure
properties, the fast path for the above, where o and o2 have the same
structure is simply a structure check followed by return true.
This patch introduces the new opcode: op_in_structure_property. Its fast path is identical
to op_has_structure_property. Its slow path, however, behaves like "in", which
uses the HasProperty internal method, unlike op_has_structure_property,
which uses the GetOwnProperty internal method. This behavior difference is
observable using Proxy.
This a 5% perf improvement in the Elm subtest in Speedometer 2.
* bytecode/BytecodeList.rb:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitInByVal):
(JSC::rewriteOp):
(JSC::StructureForInContext::finalize):
* bytecompiler/BytecodeGenerator.h:
(JSC::StructureForInContext::addInInst):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGNodeType.h:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileInStructureProperty):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileHasStructurePropertyImpl):
(JSC::FTL::DFG::LowerDFGToB3::compileHasStructureProperty):
(JSC::FTL::DFG::LowerDFGToB3::compileInStructureProperty):
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):
* jit/JIT.h:
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_in_structure_property):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::emit_op_in_structure_property):
* llint/LLIntOffsetsExtractor.cpp:
* llint/LowLevelInterpreter.asm:
* llint/LowLevelInterpreter64.asm:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/CommonSlowPaths.h:
* runtime/JSObject.cpp:
(JSC::JSObject::hasPropertyGeneric const):
* runtime/JSPropertyNameEnumerator.h:
2020-05-22 Keith Miller <keith_miller@apple.com>
Checkpoint inlined call return handler needs an exception check when dispatching
https://bugs.webkit.org/show_bug.cgi?id=212104
Reviewed by Yusuke Suzuki.
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::dispatchToNextInstruction):
(JSC::LLInt::slow_path_checkpoint_osr_exit_from_inlined_call):
(JSC::LLInt::slow_path_checkpoint_osr_exit):
2020-05-22 Paulo Matos <pmatos@igalia.com>
Fix non-unified builds for i386 build
https://bugs.webkit.org/show_bug.cgi?id=212258
Reviewed by Adrian Perez de Castro.
* API/JSContextRef.cpp:
* bytecode/IntrinsicGetterAccessCase.cpp:
* inspector/InjectedScriptHost.cpp:
* llint/LLIntData.cpp:
* llint/LLIntThunks.cpp:
* runtime/Exception.cpp:
* runtime/ExecutableBase.cpp:
* runtime/JSBigInt.cpp:
* runtime/JSInternalPromiseConstructor.cpp:
* runtime/JSString.cpp:
* runtime/ScopedArgumentsTable.cpp:
* runtime/ScriptExecutable.cpp:
* runtime/SparseArrayValueMap.cpp:
* runtime/StructureRareData.cpp:
2020-05-22 Paulo Matos <pmatos@igalia.com>
Fix typo in JSCVirtualMachine documentation
Unreviewed Typo Fix.
* API/glib/JSCVirtualMachine.cpp:
2020-05-21 Robin Morisset <rmorisset@apple.com>
Various compile-time boolean flags could/should be marked constexpr
https://bugs.webkit.org/show_bug.cgi?id=212244
Reviewed by Mark Lam.
This trivial patch saves roughly 16kB from the JavaScriptCore binary in release mode.
* b3/B3OptimizeAssociativeExpressionTrees.cpp:
* b3/air/AirAllocateRegistersByGraphColoring.cpp:
* b3/air/AirSimplifyCFG.cpp:
(JSC::B3::Air::simplifyCFG):
* b3/air/AirTmpWidth.cpp:
(JSC::B3::Air::TmpWidth::recompute):
* dfg/DFGPredictionPropagationPhase.cpp:
* heap/GCIncomingRefCountedInlines.h:
(JSC::GCIncomingRefCounted<T>::filterIncomingReferences):
* heap/Heap.cpp:
(JSC::Heap::updateAllocationLimits):
* wasm/WasmEntryPlan.cpp:
2020-05-21 Robin Morisset <rmorisset@apple.com>
Remove AssemblerBufferWithConstantPool.h (as it has been dead for years)
https://bugs.webkit.org/show_bug.cgi?id=212241
Reviewed by Yusuke Suzuki.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* assembler/AssemblerBufferWithConstantPool.h: Removed.
2020-05-21 Alexey Shvayka <shvaikalesh@gmail.com>
Use @isUndefinedOrNull instead of abstract equality with null
https://bugs.webkit.org/show_bug.cgi?id=210954
Reviewed by Yusuke Suzuki.
This patch:
a) Replaces 2 `!== @undefined` comparisons in String.prototype.{replace,replaceAll}
with @isUndefinedOrNull() as per spec [1], aligning JSC with V8 and SpiderMonkey.
b) Replaces 3 `!= @undefined` and 7 `!= null` comparisons with @isUndefinedOrNull()
as only the latter is correct with [[IsHTMLDDA]] aka MasqueradesAsUndefined objects [2].
c) Removes @isDictionary() since it is unused, easy to inline and its name is quite
misleading: one might expect it to perform Structure::isDictionary().
[1]: https://tc39.es/ecma262/#sec-getmethod (step 3)
[2]: https://tc39.es/ecma262/#sec-IsHTMLDDA-internal-slot
* builtins/ArrayConstructor.js:
* builtins/GlobalOperations.js:
(globalPrivate.isDictionary): Deleted.
* builtins/RegExpPrototype.js:
(Symbol.split): Unobservable as there is `=== null` check in @regExpExec.
* builtins/StringPrototype.js:
(match):
(replace):
(replaceAll):
(search):
(split):
* builtins/TypedArrayConstructor.js:
2020-05-21 Saam Barati <sbarati@apple.com>
Add an option that exposes functions on the global object to turn on and off the sampling profiler and the super sampler
https://bugs.webkit.org/show_bug.cgi?id=212178
Reviewed by Yusuke Suzuki.
When profiling things like Speedometer inside the browser, it's important to
to only enable the super sampler and the sampling profiler around the code
that you want profiled. Otherwise, you will be profiling things that aren't
relevant to the benchmark score. This patch adds a new option, exposeProfilersOnGlobalObject,
which when true, will expose JS functions on the global object that allow
enabling/disabling the super sampler and the sampling profiler. This way,
we can change the Speedometer source code locally such that these profilers
are only sampling code accounted for in the benchmark score.
* bytecode/SuperSampler.cpp:
(JSC::initializeSuperSampler):
(JSC::enableSuperSampler):
(JSC::disableSuperSampler):
* bytecode/SuperSampler.h:
* jsc.cpp:
(jscmain):
* runtime/JSGlobalObject.cpp:
(JSC::enableSamplingProfiler):
(JSC::disableSamplingProfiler):
(JSC::enableSuperSampler):
(JSC::disableSuperSampler):
(JSC::JSGlobalObject::init):
* runtime/OptionsList.h:
2020-05-21 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Fix 32bit JSBigInt with INT32_MAX < x <= UINT32_MAX
https://bugs.webkit.org/show_bug.cgi?id=212193
Reviewed by Mark Lam.
In 32bit architecture, we are creating one-length JSBigInt for INT32_MIN <= x <= INT32_MAX, and two-length JSBigInt otherwise.
This is wrong since one-length JSBigInt should cover from -UINT32_MAX <= x <= UINT32_MAX.
This patch fixes the bug and cleans up createFrom(VM&, int64_t). And it also adds JSBigInt::createFrom(VM&, uint64_t) in preparation for [1]
Currently, this path is not used while it was used previously because BigIntConstructor starts using JSBigInt::createFrom(VM&, double). But this
will be used in [1], and simply the existing implementation is wrong.
[1]: https://bugs.webkit.org/show_bug.cgi?id=190800
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createFromImpl):
(JSC::JSBigInt::createFrom):
* runtime/JSBigInt.h:
2020-05-21 Paulo Matos <pmatos@igalia.com>
Further non-unified build fixes
https://bugs.webkit.org/show_bug.cgi?id=212195
Reviewed by Adrian Perez de Castro.
* bytecode/InstanceOfStatus.cpp:
* heap/MarkedSpace.cpp:
* runtime/ObjectInitializationScope.cpp:
* runtime/ThrowScope.cpp:
2020-05-21 Alexey Shvayka <shvaikalesh@gmail.com>
Array.prototype.concat is incorrect with objects whose "length" exceeds 2 ** 32 - 1
https://bugs.webkit.org/show_bug.cgi?id=212167
Reviewed by Saam Barati.
This patch increases "length" limit of Array.prototype.concat result to @MAX_SAFE_INTEGER
and changes thrown error to TypeError, aligning JSC with the spec [1], V8, and SpiderMonkey.
Also, adds missing @MAX_SAFE_INTEGER overflow check in Array.from [2] (we implement similar
checks in other methods). SunSpider and microbenchmarks/concat-append-one.js are both neutral.
[1]: https://tc39.es/ecma262/#sec-array.prototype.concat (steps 5.c.iii, 5.d.ii)
[2]: https://tc39.es/ecma262/#sec-array.from (step 5.e.i)
* builtins/ArrayConstructor.js:
(from):
* builtins/ArrayPrototype.js:
(globalPrivate.concatSlowPath):
2020-05-20 Michael Saboff <msaboff@apple.com>
[Wasm] Limit the size of Wasm function we optimize in OMG mode
https://bugs.webkit.org/show_bug.cgi?id=212105
Reviewed by Filip Pizlo.
Given that memory grows O(N^2) compiling Wasm code through the OMG path,
we can run out of memory when compiling large Wasm functions. This change adds
a limit option, webAssemblyBBQFallbackSize, When the Wasm function size is
equal to or greater than this limit we always compile using BBQ optimization
parameters.
As part of this change, we still go through the OMG loop entry OSR code
generation path for functions that are at or above the threshold, but we
compile such functions with BBQ compilation optimization levels.
Also for Wasm functions at or above the threashold, we don't tier up to an
OMG compiled normal entry function. Instead we stay with the BBQ compiled version.
* runtime/OptionsList.h:
* wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::AirIRGenerator):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::B3IRGenerator):
(JSC::Wasm::parseAndCompile):
* wasm/WasmCompilationMode.cpp:
(JSC::Wasm::wasmFunctionSizeCanBeOMGCompiled):
* wasm/WasmCompilationMode.h:
* wasm/WasmOperations.cpp:
(JSC::Wasm::operationWasmTriggerOSREntryNow):
2020-05-19 Ross Kirsling <ross.kirsling@sony.com>
REGRESSION(r261755): Win/Linux non-unified builds have hundreds of link failures
https://bugs.webkit.org/show_bug.cgi?id=212111
Unreviewed build fix.
* API/:
* bindings/:
* bytecode/:
* bytecompiler/NodesCodegen.cpp:
* debugger/:
* dfg/:
* heap/:
* inspector/:
* interpreter/:
* jit/:
* llint/LLIntEntrypoint.cpp:
* parser/:
* profiler/:
* runtime/:
Restore *Inlines.h includes for >300 files,
but try to preserve the spirit of the original patch by pruning redundancies along the way.
2020-05-19 Mark Lam <mark.lam@apple.com>
Put PtrTagLookup data structures in Configs for freezing.
https://bugs.webkit.org/show_bug.cgi?id=212089
<rdar://problem/63401487>
Reviewed by Robin Morisset.
PtrTagLookup data structures were always meant to only be initialized once at
initialization time and never modified thereafter. This patch puts them in the
Configs for freezing to document and enforce this invariant.
* runtime/JSCConfig.h:
* runtime/JSCPtrTag.cpp:
(JSC::initializePtrTagLookup):
2020-05-19 Youenn Fablet <youenn@apple.com>
[ Mac wk1 Debug ] imported/w3c/web-platform-tests/fetch/api/basic/stream-safe-creation.any.html is flaky crashing with alerts - WTFCrashWithInfo - SC::JSObject::get(JSC::JSGlobalObject*, JSC::PropertyName)
https://bugs.webkit.org/show_bug.cgi?id=211923
<rdar://problem/63244249>
Reviewed by Mark Lam.
* runtime/JSObject.h:
(JSC::JSObject::get const):
When calling get, a terminate exception might happen if running in workers.
Return early in that case. Add an ASSERT that only terminated exceptions can actually happen.
2020-05-18 Andy Estes <aestes@apple.com>
http/tests/ssl/applepay/ApplePayInstallmentConfiguration.https.html fails in public SDK builds
https://bugs.webkit.org/show_bug.cgi?id=212000
<rdar://problem/63323082>
Reviewed by Youenn Fablet.
* Configurations/FeatureDefines.xcconfig:
2020-05-18 Saam Barati <sbarati@apple.com>
Do more speculation that a GetByVal/PutByVal will have an int32 index based on data from ArrayProfile
https://bugs.webkit.org/show_bug.cgi?id=211877
Reviewed by Yusuke Suzuki.
Before this patch, when a GetByVal or PutByVal had a non int32 prediction for
their incoming index, they'd fall completely off the fast path. However, there
are programs where an int32 is boxed inside a double, but our notion of
predicted types don't fully capture this fact. For example, if we have a double Add
to produce an array index, that double Add will predict a full double result,
not a SpecAnyIntAsDouble. However, for GetByVal and PutByVal, there is information
from ArrayProfile we can use to determine if the incoming value is expected to
be in int32 range. The heuristic this patch introduces is:
isFullNumberSpeculation(indexSpeculation)
&& node->arrayMode().isSpecific()
&& node->arrayMode().isInBounds()
&& !m_graph.hasExitSite(node->origin.semantic, Overflow) // DoubleAsInt32 will exit with Overflow on failure
If these conditions are met, we'll now emit a DoubleAsInt32 conversion node
for the index. This puts along the fast path for GetByVal and PutByVal on
array accesses where the incoming index is an int32 boxed in a double.
To make the above isFullNumberSpeculation check more robust, this patch also
makes it so non index double accesses result in marking the array profile as
out of bounds. So this means indices greater than max safe index, and also,
fractional doubles.
This is a 3.75x speedup on microbenchmarks/get-and-put-by-val-double-index-dont-fall-off-a-cliff.js
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* jit/JITOperations.cpp:
(JSC::getByVal):
2020-05-18 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] BigInt peephole compare should speculate appropriately
https://bugs.webkit.org/show_bug.cgi?id=212037
<rdar://problem/63346966>
Reviewed by Saam Barati.
SpeculativeJIT::nonSpeculativePeepholeBranch missed BigInt speculation. This patch renames it
to SpeculativeJIT::genericJSValuePeepholeBranch and adds speculation checks appropriately.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compilePeepHoleBranch):
(JSC::DFG::SpeculativeJIT::genericJSValuePeepholeBranch):
(JSC::DFG::SpeculativeJIT::nonSpeculativePeepholeBranch): Deleted.
* dfg/DFGSpeculativeJIT.h:
2020-05-18 Keith Miller <keith_miller@apple.com>
OSR loop entry to iterator_next generic needs to CheckNotEmpty on m_next
https://bugs.webkit.org/show_bug.cgi?id=212001
Reviewed by Saam Barati.
If we happen to OSR enter into iterator_next during a for-of loop
that has only profiled a generic iterator but is actually running
a fast iterator we will incorrectly perform the Call node This
could happen if we loop_hint OSR enter the first time have seen a
fast iterator. If this happens right now, we generate the following
code:
D@113:<!2:loc15> GetLocal(Check:Untyped:D@198, JS|MustGen|UseAsOther, Function|Empty, loc13(W~/FlushedJSValue), machine:loc10, R:Stack(loc13),Stack(loc5), bc#46, ExitValid) predicting Function|Empty
0x4913f1806151: mov -0x58(%rbp), %rsi
D@114:<!0:-> FilterCallLinkStatus(Check:Untyped:D@113, MustGen, (Function: Object: 0x1053f47e0 with butterfly 0x0 (Structure 0x1053f9260:[0x6dad, Function, {}, NonArray, Proto:0x1050fc248]), StructureID: 28077; Executable: next#Ddkruz:[0x1053c0480->0x1053e4a80, BaselineFunctionCall, 54 (StrictMode)]), R:Stack(loc5), W:SideState, bc#46, ExitValid)
D@115:<!6:loc15> Call(Check:Untyped:D@113, Check:Untyped:D@110, JS|MustGen|VarArgs|UseAsOther, Final, R:World,Stack(loc5), W:Heap, ExitsForExceptions, ClobbersExit, bc#46, ExitValid) predicting Final
0x4913f1806155: mov $0x1, 0x10(%rsp)
0x4913f180615d: mov %rax, 0x18(%rsp)
0x4913f1806162: mov %rsi, 0x8(%rsp)
0x4913f1806167: mov %rax, -0xa0(%rbp)
0x4913f180616e: mov $0x0, 0x24(%rbp)
0x4913f1806175: mov $0x0, %r11
0x4913f180617f: cmp %r11, %rsi
0x4913f1806182: jnz 0x4913f1806192
0x4913f1806188: call 0x4913f180618d
0x4913f180618d: jmp 0x4913f18061ae
0x4913f1806192: mov %rsi, %rax
0x4913f1806195: mov $0x1050cfcb0, %rdx
0x4913f180619f: mov $0x1052fab68, %rcx
0x4913f18061a9: call 0x4913f1801680
0x4913f18061ae: lea -0xd0(%rbp), %rsp
D@116:<!0:-> MovHint(Check:Untyped:D@115, MustGen, tmp0, R:Stack(loc5), W:SideState, ClobbersExit, bc#46, ExitInvalid)
D@332:<!0:-> InvalidationPoint(MustGen, R:Stack(loc5), W:SideState, Exits, bc#46, exit: bc#46cp#1, ExitValid)
D@335:<!0:-> CheckStructure(Check:Cell:D@115, MustGen, [%B2:Object], R:Stack(loc5),JSCell_structureID, Exits, bc#46, exit: bc#46cp#1, ExitValid)
0x4913f18061b5: test %rax, %r15
0x4913f18061b8: jnz 0x4913f18068db
0x4913f18061be: cmp $0xcaae, (%rax)
0x4913f18061c4: jnz 0x4913f18068f1
Loc13 in this IR is the location of the next function. Since it's
nullptr, we will pass the initial fast-path value of 0 and make a
garbage call. This is because Call does not know how to handle
empty values. Subsequently, we will fail a structure check for the
Call's result and OSR exit to the getDone checkpoint. The fix for
this is to simply put a CheckNotEmpty at the top of the generic
case. 99.9% of the time this check will be eliminated so it
doesn't really cost anything.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
2020-05-17 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, link fix for our internal Debug build
* heap/AlignedMemoryAllocator.cpp:
2020-05-17 Lauro Moura <lmoura@igalia.com>
[JSC] Silence unused-but-set-parameter warnings for older compilers
https://bugs.webkit.org/show_bug.cgi?id=212006
Reviewed by Mark Lam.
GCC up to 9.x will emit unused-but-set-parameter for the sources
parameter when NumberOfRegisters is zero (the if block is eliminated)
and for destinations when also ASSERT_ENABLED is false.
* jit/CCallHelpers.h:
(JSC::CCallHelpers::setupStubArgs):
2020-05-16 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Make OutOfMemory error as instance of RangeError
https://bugs.webkit.org/show_bug.cgi?id=211952
Reviewed by Mark Lam.
The spec sometimes requires "check parameters and throw RangeError" before allocating an object.
But we are just allocating an object and throwing an out-of-memory error since wrong parameter will
cause out-of-memory. If out-of-memory error is RangeError, then we can keep our current behavior while
we can make us spec compliant. And note that out-of-memory error is RangeError in SpiderMonkey and V8.
This patch makes out-of-memory error as RangeError instead of Error. We also fix @throwOutOfMemoryError
in builtin code: the previous thrown errors are not marked as out-of-memory error.
* bytecode/BytecodeList.rb:
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitThrowStaticError):
(JSC::BytecodeGenerator::emitThrowReferenceError):
(JSC::BytecodeGenerator::emitThrowTypeError):
(JSC::BytecodeGenerator::emitThrowRangeError):
(JSC::BytecodeGenerator::emitThrowOutOfMemoryError):
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::RegExpNode::emitBytecode):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_throwTypeError):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_throwRangeError):
* dfg/DFGOperations.cpp:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/Error.cpp:
(JSC::createError):
(JSC::createOutOfMemoryError):
* runtime/Error.h:
* runtime/ErrorType.cpp:
(JSC::errorTypeName):
(WTF::printInternal):
* runtime/ErrorType.h: We introduced ErrorTypeWithExtension separately from ErrorType to keep ErrorType one-on-one to spec-specified error types.
2020-05-15 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] getFunctionRealm should not use recursion
https://bugs.webkit.org/show_bug.cgi?id=211965
<rdar://problem/63268287>
Reviewed by Saam Barati.
This patch avoids using recursion in getFunctionRealm to avoid stack-overflow.
* runtime/InternalFunction.cpp:
(JSC::getFunctionRealm):
2020-05-15 Keith Miller <keith_miller@apple.com>
Unreviewed, fix internal arm64e build.
* dfg/DFGSpeculativeJIT.cpp:
2020-05-15 Keith Miller <keith_miller@apple.com>
Unreviewed, fix internal fast tls build.
* jit/AssemblyHelpers.cpp:
2020-05-15 Ross Kirsling <ross.kirsling@sony.com>
[IWYU] Remove unnecessary includes from JSC implementation files
https://bugs.webkit.org/show_bug.cgi?id=211867
Reviewed by Keith Miller.
* API/:
* assembler/:
* b3/:
* bindings/:
* builtins/BuiltinExecutables.cpp:
* bytecode/:
* bytecompiler/:
* debugger/:
* dfg/:
* disassembler/:
* ftl/:
* heap/:
* inspector/:
* interpreter/:
* jit/:
* jsc.cpp:
* llint/:
* parser/:
* profiler/:
* runtime/:
* testRegExp.cpp:
* tools/:
* wasm/:
* yarr/:
2020-05-15 Michael Catanzaro <mcatanzaro@gnome.org>
-Wtype-limits warning spam from CCallHelpers.h
https://bugs.webkit.org/show_bug.cgi?id=211701
Reviewed by Darin Adler.
Skip the problematic loops when TargetSize or NumberOfRegisters is 0 using constexpr if.
Solution suggested by Mark Lam.
* jit/CCallHelpers.h:
(JSC::CCallHelpers::setupStubArgs):
(JSC::CCallHelpers::clampArrayToSize):
2020-05-15 Mark Lam <mark.lam@apple.com>
Remove debugging dataLogs in LinkBuffer::copyCompactAndLinkCode() for release builds.
https://bugs.webkit.org/show_bug.cgi?id=211961
<rdar://problem/63264848>
Reviewed by Keith Miller.
* assembler/LinkBuffer.cpp:
(JSC::LinkBuffer::copyCompactAndLinkCode):
2020-05-15 Paulo Matos <pmatos@igalia.com>
Fix ARM NEON only assert
https://bugs.webkit.org/show_bug.cgi?id=211889
Reviewed by Mark Lam.
Fix assert that breaks if ARM does not contain NEON extensions -
the register d16 is only defined if NEON exists.
* assembler/ARMv7Assembler.h:
(JSC::RegisterNames::asSingle):
(JSC::RegisterNames::asSingleUpper):
2020-05-14 Saam Barati <sbarati@apple.com>
GetByVal and PutByVal runtime operations shouldn't fall off a performance cliff when the property is an integer boxed as a double
https://bugs.webkit.org/show_bug.cgi?id=211935
Reviewed by Yusuke Suzuki and Mark Lam.
There were parts in the runtime for get_by_val that weren't properly handling
ints boxed as doubles along the fast path. This could lead to terrible
performance as we could go from double -> string -> int while converting the
subscript into a property to access.
This patch fixes that, and removes the duplicate code we had throughout the
codebase that does this conversion. I'm adding a new functions tryGetAsUint32Index
and tryGetAsInt32 which will handle the double to int conversion.
This is a 10x speedup on the microbenchmark get-and-put-by-val-double-index-dont-fall-off-a-cliff.js
* dfg/DFGOperations.cpp:
(JSC::DFG::putByValInternal):
* jit/JITOperations.cpp:
(JSC::getByVal):
* jsc.cpp:
(functionAsDoubleNumber):
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::getByVal):
(JSC::LLInt::LLINT_SLOW_PATH_DECL):
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::tryGetAsUint32Index):
(JSC::JSValue::tryGetAsInt32):
2020-05-14 Devin Rousso <drousso@apple.com>
[ESNext] enable logical assignment operators by default
https://bugs.webkit.org/show_bug.cgi?id=211921
Reviewed by Yusuke Suzuki.
* runtime/OptionsList.h:
* parser/Lexer.cpp:
(JSC::Lexer<T>::lexWithoutClearingLineTerminator):
Remove `useLogicalAssignmentOperators` option.
2020-05-14 Keith Miller <keith_miller@apple.com>
Undecided Arrays shouldn't need to be OriginalArray to covert to GetArrayLength
https://bugs.webkit.org/show_bug.cgi?id=211914
Reviewed by Saam Barati.
Also, fix a bug that arrayModesThatPassFiltering() can't handle
Undecided arrays. Because we can now emit a CheckArray on
Undecided AI will try to figure out what types flow out of the
check. Since Undecided was unhandled by filtering, AI will assume
bottom is the only possible value and the DFG/FTL will insert a
breakpoint, causing a crash.
* dfg/DFGArrayMode.cpp:
(JSC::DFG::ArrayMode::refine const):
* dfg/DFGArrayMode.h:
(JSC::DFG::ArrayMode::arrayModesThatPassFiltering const):
2020-05-14 Keith Miller <keith_miller@apple.com>
GetArrayLength should be "blessed" during Fixup phase in the DFG
https://bugs.webkit.org/show_bug.cgi?id=211540
Reviewed by Saam Barati.
If we got an ArrayMode during bytecode parsing for-of that expects
to be configured during Fixup, then right now we will crash on
GetArrayLength. This fixes GetArrayLength to properly call
blessArrayOperation and fixes clobberize to know that
GetArrayLength could have a ForceExit ArrayMode briefly before
being cleaned up.
When blessing GetArrayLength we can now produce CheckArrays that
have an AnyTypedArray ArrayMode::Type. So this patch expands
CheckArray to properly handle that. To help with this we expand
branchIfType to have a starting JSType and an optional ending
JSType. Additionally, to prevent extra checks AI has been taught
to fold more ArrayModes so we should almost always avoid new
runtime checks.
Lastly, make sure that Undecided Arrays don't fall back to generic
because GetArrayLength can't be converted to...
GetArrayLenth. Also, GetArrayLength would previously pass it's own
speculation for the speculation of the index, which logically
doesn't make sense. So this patch adds a new constant, which is
SpecInt32Only, that can be passed if a DFG node doesn't have an
index.
* assembler/testmasm.cpp:
(JSC::testBranchIfType):
(JSC::testBranchIfNotType):
(JSC::run):
* dfg/DFGArrayMode.cpp:
(JSC::DFG::canBecomeGetArrayLength):
* dfg/DFGArrayMode.h:
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::blessArrayOperation):
(JSC::DFG::FixupPhase::attemptToMakeGetArrayLength):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::checkArray):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::isArrayTypeForCheckArray):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::branchIfType):
(JSC::AssemblyHelpers::branchIfNotType):
* runtime/JSType.h:
2020-05-13 Keith Miller <keith_miller@apple.com>
iteration bytecodes need to handle osr exiting from inlined getter frames
https://bugs.webkit.org/show_bug.cgi?id=211873
Reviewed by Saam Barati.
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::slow_path_checkpoint_osr_exit_from_inlined_call):
2020-05-13 Devin Rousso <drousso@apple.com>
Web Inspector: rename CSS.StyleSheetOrigin.Regular to CSS.StyleSheetOrigin.Author to match the spec
https://bugs.webkit.org/show_bug.cgi?id=211827
Reviewed by Timothy Hatcher.
* inspector/protocol/CSS.json:
2020-05-13 Yusuke Suzuki <ysuzuki@apple.com>
JSDOMWindowBase m_windowCloseWatchpoints must be Ref<>
https://bugs.webkit.org/show_bug.cgi?id=211844
Reviewed by Mark Lam.
* bytecode/Watchpoint.cpp:
(JSC::InlineWatchpointSet::inflateSlow):
* bytecode/Watchpoint.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::JSGlobalObject):
* runtime/Structure.cpp:
(JSC::Structure::ensurePropertyReplacementWatchpointSet):
* runtime/SymbolTable.cpp:
(JSC::SymbolTableEntry::prepareToWatch):
* runtime/VM.cpp:
(JSC::VM::ensureWatchpointSetForImpureProperty):
2020-05-13 Caio Lima <ticaiolima@gmail.com>
Making 32-bits JIT build without Unified Build system
https://bugs.webkit.org/show_bug.cgi?id=211853
Reviewed by Adrian Perez de Castro.
This patch is moving some templates to allow non-unified builds on
32-bits JIT configurations.
Those templates were from JITArithmetic32_64 and JITPropertyAccess32_64.
* jit/JITArithmetic.cpp:
(JSC::JIT::emit_compareAndJump):
(JSC::JIT::emit_compareUnsignedAndJump):
(JSC::JIT::emit_compareUnsigned):
(JSC::JIT::emit_compareAndJumpSlow):
(JSC::JIT::emitBinaryDoubleOp):
* jit/JITArithmetic32_64.cpp:
(JSC::JIT::emit_compareAndJump): Deleted.
(JSC::JIT::emit_compareUnsignedAndJump): Deleted.
(JSC::JIT::emit_compareUnsigned): Deleted.
(JSC::JIT::emit_compareAndJumpSlow): Deleted.
(JSC::JIT::emitBinaryDoubleOp): Deleted.
* jit/JITOpcodes32_64.cpp:
* jit/JITPropertyAccess.cpp:
(JSC::JIT::emitPutByValWithCachedId):
* jit/JITPropertyAccess32_64.cpp:
(JSC::JIT::emitPutByValWithCachedId): Deleted.
2020-05-13 Caio Lima <ticaiolima@gmail.com>
[JSC] Support delete by val/id IC on 32-bits
https://bugs.webkit.org/show_bug.cgi?id=208207
Reviewed by Saam Barati.
This patch implements DeleteById and DeleteByVal IC on 32-bits JIT. It
includes both Baseline and DFG changes.
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileDeleteById):
(JSC::DFG::SpeculativeJIT::compileDeleteByVal):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compileDeleteById): Deleted.
(JSC::DFG::SpeculativeJIT::compileDeleteByVal): Deleted.
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compileDeleteById): Deleted.
(JSC::DFG::SpeculativeJIT::compileDeleteByVal): Deleted.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileDelBy):
* jit/JITInlineCacheGenerator.cpp:
(JSC::JITDelByValGenerator::JITDelByValGenerator):
(JSC::JITDelByIdGenerator::JITDelByIdGenerator):
* jit/JITInlineCacheGenerator.h:
* jit/JITPropertyAccess.cpp:
(JSC::JIT::emit_op_del_by_id):
(JSC::JIT::emit_op_del_by_val):
* jit/JITPropertyAccess32_64.cpp:
(JSC::JIT::emit_op_del_by_id):
(JSC::JIT::emit_op_del_by_val):
(JSC::JIT::emitSlow_op_del_by_val):
(JSC::JIT::emitSlow_op_del_by_id):
2020-05-13 Saam Barati <sbarati@apple.com>
MovHint can see an arguments object be MovHinted to a Tmp
https://bugs.webkit.org/show_bug.cgi?id=211820
<rdar://problem/62882158>
Reviewed by Keith Miller.
We had an assert that it wasn't possible to have a MovHint from an arguments
object to a Tmp. However, this is possible with for-of. There is nothing
about the current algorithm that is specific to only VirtualRegisters. The
algorithm also works over Tmps. So I've generalized the algorithm to just work
over Operand.
* dfg/DFGVarargsForwardingPhase.cpp:
2020-05-13 Alexey Shvayka <shvaikalesh@gmail.com>
Move @isConstructor checks from fast paths of Array.from and Array.of
https://bugs.webkit.org/show_bug.cgi?id=211805
Reviewed by Keith Miller.
This semantically equivalent change advances provided Array.{from,of} microbenchmarks by ~60%.
Also, this patch removes @isConstructor check from @newPromiseCapabilitySlow (that is heavily
used by Promise subclasses) since it comes right before [[Construct]], its message doesn't add
more clarity, and constructability of its argument was likely checked by @speciesConstructor.
* builtins/ArrayConstructor.js:
(of):
(from):
* builtins/PromiseOperations.js:
(globalPrivate.newPromiseCapabilitySlow):
2020-05-12 Alexey Shvayka <shvaikalesh@gmail.com>
Implement @isConstructor bytecode intrinsic and bytecode for that
https://bugs.webkit.org/show_bug.cgi?id=144093
Reviewed by Keith Miller.
This change replaces @isConstructor link-time-constant with bytecode intrinsic and utilizes it
in ClassExprNode::emitBytecode() according to the spec [1], aligning JSC with V8 and SpiderMonkey.
Before this patch, we checked if "prototype" of superclass is an object, which is incorrect for
generators and bound non-constructor functions with own "prototype".
OpIsConstructor's fast path can't be easily compiled, and it's not a hot code anyway, so instead
we reduce code bloat by just calling slow ops from DFG and FTL (if we bail out, we slow down all
@isConstructor call sites). This advances microbenchmarks/is-constructor.js by ~35%.
[1]: https://tc39.es/ecma262/#sec-runtime-semantics-classdefinitionevaluation (step 5.f)
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* builtins/BuiltinNames.h:
* bytecode/BytecodeIntrinsicRegistry.h:
* bytecode/BytecodeList.rb:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitIsConstructor):
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::ClassExprNode::emitBytecode):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGHeapLocation.cpp:
(WTF::printInternal):
* dfg/DFGHeapLocation.h:
* dfg/DFGNodeType.h:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileIsConstructor):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileIsConstructor):
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
* llint/LowLevelInterpreter.asm:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/CommonSlowPaths.h:
* runtime/ECMAScriptSpecInternalFunctions.cpp: Removed.
* runtime/ECMAScriptSpecInternalFunctions.h: Removed.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
2020-05-12 Robin Morisset <rmorisset@apple.com>
Exception check for OOM is a bit too late in JSBigInt::exponentiate.
https://bugs.webkit.org/show_bug.cgi?id=211823
Reviewed by Mark Lam.
We were doing multiplyImpl(...).payload.asHeapBigInt(), but multiplyImpl can return a null payload if it causes an exception.
So we must first check whether an exception was raised, and only if not can we do asHeapBigInt.
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::exponentiateImpl):
2020-05-12 Saam Barati <sbarati@apple.com>
handling of Check in VarargsForwardingPhase is too pessimistic
https://bugs.webkit.org/show_bug.cgi?id=211810
Reviewed by Keith Miller and Filip Pizlo.
We were treating a check, even if it wasn't on the sink candidate,
as if it could escape the candidate. That's wrong. Only checks on the
candidate have the escaping ability.
* dfg/DFGVarargsForwardingPhase.cpp:
2020-05-12 Keith Miller <keith_miller@apple.com>
The bottom value set for m_value in iterator_next should be materialized after a done getter
https://bugs.webkit.org/show_bug.cgi?id=211811
Reviewed by Saam Barati.
Right now, if the done getter contains control flow, then we will
have the bottom value in a different block from the
MovHint/SetLocal and we will fail to validate.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
2020-05-12 Ross Kirsling <ross.kirsling@sony.com>
Fix existing usage of final/override/virtual in JSC and WTF
https://bugs.webkit.org/show_bug.cgi?id=211772
Reviewed by Darin Adler.
* API/JSAPIWrapperObject.mm:
* API/JSManagedValue.mm:
* API/JSScriptSourceProvider.h:
* API/ObjCCallbackFunction.mm:
* API/glib/JSAPIWrapperGlobalObject.cpp:
* API/glib/JSAPIWrapperObjectGLib.cpp:
* API/glib/JSCWeakValue.cpp:
* bytecode/AccessCaseSnippetParams.cpp:
* bytecode/AccessCaseSnippetParams.h:
* bytecode/CodeBlock.cpp:
* bytecode/StructureStubClearingWatchpoint.h:
* bytecode/VariableWriteFireDetail.h:
* bytecode/Watchpoint.h:
* dfg/DFGAdaptiveInferredPropertyValueWatchpoint.h:
* dfg/DFGArrayifySlowPathGenerator.h:
* dfg/DFGCallArrayAllocatorSlowPathGenerator.h:
* dfg/DFGCallCreateDirectArgumentsSlowPathGenerator.h:
* dfg/DFGSaneStringGetByValSlowPathGenerator.h:
* dfg/DFGSlowPathGenerator.h:
* dfg/DFGSnippetParams.h:
* dfg/DFGWorklist.cpp:
* ftl/FTLSnippetParams.h:
* heap/BlockDirectory.cpp:
* heap/EdenGCActivityCallback.h:
* heap/FullGCActivityCallback.h:
* heap/Heap.cpp:
* heap/Heap.h:
* heap/IncrementalSweeper.h:
* heap/IsoCellSet.cpp:
* heap/IsoCellSetInlines.h:
* heap/IsoHeapCellType.h:
* heap/IsoInlinedHeapCellType.h:
* heap/ParallelSourceAdapter.h:
* heap/StopIfNecessaryTimer.h:
* heap/Subspace.cpp:
* heap/SubspaceInlines.h:
* inspector/InjectedScript.h:
* inspector/JSGlobalObjectConsoleClient.h:
* inspector/JSGlobalObjectInspectorController.h:
* inspector/JSGlobalObjectScriptDebugServer.h:
* inspector/JSInjectedScriptHost.cpp:
* inspector/agents/InspectorAgent.h:
* inspector/agents/InspectorScriptProfilerAgent.h:
* inspector/agents/InspectorTargetAgent.h:
* inspector/agents/JSGlobalObjectAuditAgent.h:
* inspector/agents/JSGlobalObjectDebuggerAgent.h:
* inspector/agents/JSGlobalObjectRuntimeAgent.h:
* inspector/augmentable/AlternateDispatchableAgent.h:
* inspector/remote/RemoteConnectionToTarget.h:
* inspector/remote/RemoteInspector.h:
* inspector/remote/socket/RemoteInspectorServer.h:
* inspector/scripts/codegen/cpp_generator_templates.py:
* inspector/scripts/codegen/generate_objc_backend_dispatcher_header.py:
* inspector/scripts/tests/all/expected/definitions-with-mac-platform.json-result:
* inspector/scripts/tests/generic/expected/command-targetType-matching-domain-debuggableType.json-result:
* inspector/scripts/tests/generic/expected/commands-with-async-attribute.json-result:
* inspector/scripts/tests/generic/expected/commands-with-optional-call-return-parameters.json-result:
* inspector/scripts/tests/generic/expected/domain-debuggableTypes.json-result:
* inspector/scripts/tests/generic/expected/domain-targetType-matching-domain-debuggableType.json-result:
* inspector/scripts/tests/generic/expected/domain-targetTypes.json-result:
* inspector/scripts/tests/generic/expected/domains-with-varying-command-sizes.json-result:
* inspector/scripts/tests/generic/expected/enum-values.json-result:
* inspector/scripts/tests/generic/expected/event-targetType-matching-domain-debuggableType.json-result:
* inspector/scripts/tests/generic/expected/generate-domains-with-feature-guards.json-result:
* inspector/scripts/tests/mac/expected/definitions-with-mac-platform.json-result:
* jit/JITWorklist.cpp:
* parser/Nodes.h:
* parser/SourceProvider.h:
* runtime/DataView.h:
* runtime/DoublePredictionFuzzerAgent.h:
* runtime/FileBasedFuzzerAgent.h:
* runtime/GenericTypedArrayView.h:
* runtime/JSMicrotask.cpp:
* runtime/NarrowingNumberPredictionFuzzerAgent.h:
* runtime/ObjectPropertyChangeAdaptiveWatchpoint.h:
* runtime/PredictionFileCreatingFuzzerAgent.h:
* runtime/PromiseTimer.h:
* runtime/RandomizingFuzzerAgent.h:
* runtime/RegExpCache.h:
* runtime/Structure.cpp:
* runtime/StructureRareData.cpp:
* runtime/VMTraps.cpp:
* runtime/WideningNumberPredictionFuzzerAgent.h:
* tools/JSDollarVM.cpp:
* wasm/WasmBBQPlan.h:
* wasm/WasmCallee.h:
* wasm/WasmLLIntPlan.h:
* wasm/WasmOMGForOSREntryPlan.h:
* wasm/WasmOMGPlan.h:
* wasm/WasmWorklist.cpp:
* yarr/YarrJIT.cpp:
2020-05-12 Ross Kirsling <ross.kirsling@sony.com>
[clang-tidy] Run modernize-use-override over JSC, then ensure as much as possible is final
https://bugs.webkit.org/show_bug.cgi?id=211743
Reviewed by Saam Barati.
* API/JSScriptRef.cpp:
* b3/B3ArgumentRegValue.h:
* b3/B3AtomicValue.h:
* b3/B3CCallValue.h:
* b3/B3CheckSpecial.h:
* b3/B3CheckValue.h:
* b3/B3Const32Value.h:
* b3/B3Const64Value.h:
* b3/B3ConstDoubleValue.h:
* b3/B3ConstFloatValue.h:
* b3/B3DataSection.h:
* b3/B3ExtractValue.h:
* b3/B3FenceValue.h:
* b3/B3MemoryValue.h:
* b3/B3PatchpointSpecial.h:
* b3/B3PatchpointValue.h:
* b3/B3SlotBaseValue.h:
* b3/B3StackmapSpecial.h:
* b3/B3StackmapValue.h:
* b3/B3SwitchValue.h:
* b3/B3UpsilonValue.h:
* b3/B3VariableValue.h:
* b3/B3WasmAddressValue.h:
* b3/B3WasmBoundsCheckValue.h:
* b3/air/AirCCallSpecial.h:
* b3/air/AirPrintSpecial.h:
* bytecode/BytecodeDumper.h:
* bytecode/GetterSetterAccessCase.h:
* bytecode/InstanceOfAccessCase.h:
* bytecode/IntrinsicGetterAccessCase.h:
* bytecode/ModuleNamespaceAccessCase.h:
* bytecode/ProxyableAccessCase.h:
* bytecode/Watchpoint.h:
* dfg/DFGFailedFinalizer.h:
* dfg/DFGGraph.h:
* dfg/DFGJITCode.h:
* dfg/DFGJITFinalizer.h:
* dfg/DFGToFTLDeferredCompilationCallback.h:
* dfg/DFGToFTLForOSREntryDeferredCompilationCallback.h:
* ftl/FTLForOSREntryJITCode.h:
* ftl/FTLJITCode.h:
* ftl/FTLJITFinalizer.h:
* heap/CompleteSubspace.h:
* heap/FastMallocAlignedMemoryAllocator.h:
* heap/GigacageAlignedMemoryAllocator.h:
* heap/HeapSnapshotBuilder.h:
* heap/IsoAlignedMemoryAllocator.h:
* heap/IsoSubspace.h:
* heap/IsoSubspacePerVM.cpp:
* heap/IsoSubspacePerVM.h:
* heap/MarkStackMergingConstraint.h:
* heap/SimpleMarkingConstraint.h:
* heap/SpaceTimeMutatorScheduler.h:
* heap/StochasticSpaceTimeMutatorScheduler.h:
* heap/SynchronousStopTheWorldMutatorScheduler.h:
* jit/GCAwareJITStubRoutine.h:
* jit/JITCode.h:
* jit/JITThunks.h:
* jit/JITToDFGDeferredCompilationCallback.h:
* jit/PolymorphicCallStubRoutine.h:
* jsc.cpp:
* parser/Lexer.cpp: Address warning.
* runtime/JSDestructibleObjectHeapCellType.h:
* runtime/SimpleTypedArrayController.h:
* runtime/Structure.h:
* runtime/WeakGCMap.h:
* wasm/WasmEntryPlan.h:
2020-05-12 Michael Catanzaro <mcatanzaro@gnome.org>
-Wsign-compare warnings in FTLLowerDFGToB3.cpp and DFGSpeculativeJIT.cpp
https://bugs.webkit.org/show_bug.cgi?id=211783
Reviewed by Darin Adler.
This fixes -Wsign-compare warnings introduced in r260331.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileValueBitNot):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitNot):
2020-05-12 Truitt Savell <tsavell@apple.com>
Unreviewed, reverting r261542.
Broke internal builds
Reverted changeset:
"[clang-tidy] Run modernize-use-override over JSC, then ensure
as much as possible is final"
https://bugs.webkit.org/show_bug.cgi?id=211743
https://trac.webkit.org/changeset/261542
2020-05-12 Mark Lam <mark.lam@apple.com>
Wasm::enableFastMemory() was called too late.
https://bugs.webkit.org/show_bug.cgi?id=211773
Reviewed by Yusuke Suzuki.
If Wasm fast memory is to be enabled, we should just do it in initializeThreading()
just like for all the other signal handlers that need to be initialized for JSC.
This simplifies its initialization and ensures that it is done in a timely manner
before Configs are frozen.
* jsc.cpp:
(jscmain):
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
2020-05-11 Darin Adler <darin@apple.com>
Fix problems caught by replacing WTF::Optional with std::optional
https://bugs.webkit.org/show_bug.cgi?id=211703
Reviewed by Chris Dumez.
* runtime/MachineContext.h:
(JSC::MachineContext::instructionPointer): Use explcit makeOptional here,
to work around the fact that MacroAssemblerCodePtr uses an unusual technique
to disable conversions to everything except bool.
2020-05-11 Yoshiaki JITSUKAWA <yoshiaki.jitsukawa@sony.com>
Fix build errors after r260992
https://bugs.webkit.org/show_bug.cgi?id=211756
Reviewed by Darin Adler.
Add JSC namespace specifier to NonIntrinsic and PropertyAttribute
in the macros in JSObject.h since those can be used outside of
or without introducing JSC namespace.
* runtime/JSObject.h:
2020-05-11 Ross Kirsling <ross.kirsling@sony.com>
[clang-tidy] Run modernize-use-override over JSC, then ensure as much as possible is final
https://bugs.webkit.org/show_bug.cgi?id=211743
Reviewed by Saam Barati.
* API/JSScriptRef.cpp:
* b3/B3ArgumentRegValue.h:
* b3/B3AtomicValue.h:
* b3/B3CCallValue.h:
* b3/B3CheckSpecial.h:
* b3/B3CheckValue.h:
* b3/B3Const32Value.h:
* b3/B3Const64Value.h:
* b3/B3ConstDoubleValue.h:
* b3/B3ConstFloatValue.h:
* b3/B3DataSection.h:
* b3/B3ExtractValue.h:
* b3/B3FenceValue.h:
* b3/B3MemoryValue.h:
* b3/B3PatchpointSpecial.h:
* b3/B3PatchpointValue.h:
* b3/B3SlotBaseValue.h:
* b3/B3StackmapSpecial.h:
* b3/B3StackmapValue.h:
* b3/B3SwitchValue.h:
* b3/B3UpsilonValue.h:
* b3/B3VariableValue.h:
* b3/B3WasmAddressValue.h:
* b3/B3WasmBoundsCheckValue.h:
* b3/air/AirCCallSpecial.h:
* b3/air/AirPrintSpecial.h:
* bytecode/BytecodeDumper.h:
* bytecode/GetterSetterAccessCase.h:
* bytecode/InstanceOfAccessCase.h:
* bytecode/IntrinsicGetterAccessCase.h:
* bytecode/ModuleNamespaceAccessCase.h:
* bytecode/ProxyableAccessCase.h:
* bytecode/Watchpoint.h:
* dfg/DFGFailedFinalizer.h:
* dfg/DFGGraph.h:
* dfg/DFGJITCode.h:
* dfg/DFGJITFinalizer.h:
* dfg/DFGToFTLDeferredCompilationCallback.h:
* dfg/DFGToFTLForOSREntryDeferredCompilationCallback.h:
* ftl/FTLForOSREntryJITCode.h:
* ftl/FTLJITCode.h:
* ftl/FTLJITFinalizer.h:
* heap/CompleteSubspace.h:
* heap/FastMallocAlignedMemoryAllocator.h:
* heap/GigacageAlignedMemoryAllocator.h:
* heap/HeapSnapshotBuilder.h:
* heap/IsoAlignedMemoryAllocator.h:
* heap/IsoSubspace.h:
* heap/IsoSubspacePerVM.cpp:
* heap/IsoSubspacePerVM.h:
* heap/MarkStackMergingConstraint.h:
* heap/SimpleMarkingConstraint.h:
* heap/SpaceTimeMutatorScheduler.h:
* heap/StochasticSpaceTimeMutatorScheduler.h:
* heap/SynchronousStopTheWorldMutatorScheduler.h:
* jit/GCAwareJITStubRoutine.h:
* jit/JITCode.h:
* jit/JITThunks.h:
* jit/JITToDFGDeferredCompilationCallback.h:
* jit/PolymorphicCallStubRoutine.h:
* jsc.cpp:
* parser/Lexer.cpp: Address warning.
* runtime/JSDestructibleObjectHeapCellType.h:
* runtime/SimpleTypedArrayController.h:
* runtime/Structure.h:
* runtime/WeakGCMap.h:
* wasm/WasmEntryPlan.h:
2020-05-11 Mark Lam <mark.lam@apple.com>
Introduce WTF::Config and put Signal.cpp's init-once globals in it.
https://bugs.webkit.org/show_bug.cgi?id=211729
<rdar://problem/62938878>
Reviewed by Keith Miller and Saam Barati.
1. Initialize VMTraps' signals early now that we'll be freezing signals at the end
of the first VM initialization.
2. Move the !initializeThreadingHasBeenCalled RELEASE_ASSERT in initializeThreading()
to the bottom of the function. This way, we'll also catch bugs which may cause
us to jump into the middle of the function.
Added a compilerFence there to ensure that the RELEASE_ASSERT is only executed
after all initialization is done. This guarantees that it will only be executed
at the end.
3. Call WTF::Config::permanentlyFreeze() from JSC::Config::permanentlyFreeze()
for obvious reasons: freezing one should freeze the other.
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
* runtime/JSCConfig.cpp:
(JSC::Config::permanentlyFreeze):
* runtime/VMTraps.cpp:
(JSC::VMTraps::initializeSignals):
* runtime/VMTraps.h:
2020-05-11 Keith Miller <keith_miller@apple.com>
Remove unused BytecodeKills.h
https://bugs.webkit.org/show_bug.cgi?id=211753
Reviewed by Yusuke Suzuki.
No one uses this class anymore, we should get rid of it.
* JavaScriptCore.xcodeproj/project.pbxproj:
* bytecode/BytecodeKills.h: Removed.
* bytecode/BytecodeLivenessAnalysis.cpp:
(JSC::BytecodeLivenessAnalysis::computeKills): Deleted.
* bytecode/BytecodeLivenessAnalysis.h:
* dfg/DFGGraph.cpp:
(JSC::DFG::Graph::killsFor): Deleted.
* dfg/DFGGraph.h:
2020-05-10 Ross Kirsling <ross.kirsling@sony.com>
[clang-tidy] Run modernize-use-nullptr over JSC
https://bugs.webkit.org/show_bug.cgi?id=211706
Reviewed by Darin Adler.
* API/APICallbackFunction.h:
* API/JSAPIGlobalObject.h:
* API/JSBase.cpp:
* API/JSCallbackObjectFunctions.h:
* API/JSClassRef.cpp:
* API/JSContextRef.cpp:
* API/JSObjectRef.cpp:
* API/JSScriptRef.cpp:
* API/JSValueRef.cpp:
* API/JSWeakObjectMapRefPrivate.cpp:
* API/tests/ExecutionTimeLimitTest.cpp:
* API/tests/PingPongStackOverflowTest.cpp:
* assembler/AbstractMacroAssembler.h:
* assembler/CPU.cpp:
* bytecode/CodeBlock.cpp:
* bytecode/DeleteByIdVariant.cpp:
* bytecode/GetByIdVariant.cpp:
* bytecode/InByIdVariant.cpp:
* bytecode/InlineCallFrame.cpp:
* bytecode/LazyOperandValueProfile.cpp:
* bytecode/PutByIdVariant.cpp:
* bytecode/ValueProfile.h:
* bytecode/ValueRecovery.cpp:
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
* debugger/DebuggerScope.h:
* dfg/DFGAbstractValue.cpp:
* dfg/DFGAdjacencyList.h:
* dfg/DFGArgumentPosition.h:
* dfg/DFGArrayifySlowPathGenerator.h:
* dfg/DFGAvailability.h:
* dfg/DFGByteCodeParser.cpp:
* dfg/DFGCFGSimplificationPhase.cpp:
* dfg/DFGCPSRethreadingPhase.cpp:
* dfg/DFGCompilationKey.h:
* dfg/DFGConstantFoldingPhase.cpp:
* dfg/DFGDisassembler.cpp:
* dfg/DFGDoubleFormatState.h:
* dfg/DFGEdge.h:
* dfg/DFGFixupPhase.cpp:
* dfg/DFGFrozenValue.cpp:
* dfg/DFGGenerationInfo.h:
* dfg/DFGGraph.h:
* dfg/DFGInPlaceAbstractState.cpp:
* dfg/DFGIntegerCheckCombiningPhase.cpp:
* dfg/DFGLazyJSValue.cpp:
* dfg/DFGNode.h:
* dfg/DFGOSREntrypointCreationPhase.cpp:
* dfg/DFGOSRExit.cpp:
* dfg/DFGOperations.cpp:
* dfg/DFGSilentRegisterSavePlan.h:
* dfg/DFGSpeculativeJIT.cpp:
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT64.cpp:
* dfg/DFGStructureAbstractValue.cpp:
* dfg/DFGTransition.cpp:
* dfg/DFGTypeCheckHoistingPhase.cpp:
* dfg/DFGWorklist.cpp:
* ftl/FTLAbstractHeapRepository.h:
* ftl/FTLAvailableRecovery.h:
* ftl/FTLExitValue.cpp:
* ftl/FTLFormattedValue.h:
* ftl/FTLJITCode.cpp:
* ftl/FTLLink.cpp:
* ftl/FTLLowerDFGToB3.cpp:
* ftl/FTLLoweredNodeValue.h:
* ftl/FTLOSREntry.cpp:
* ftl/FTLOSRExitCompiler.cpp:
* ftl/FTLTypedPointer.h:
* ftl/FTLValueFromBlock.h:
* ftl/FTLValueRange.h:
* heap/GCSegmentedArray.h:
* heap/Handle.h:
* heap/HandleSet.h:
* heap/HandleTypes.h:
* heap/HeapSnapshotBuilder.cpp:
* heap/MarkedBlockInlines.h:
* heap/Strong.h:
* heap/WeakImpl.h:
* heap/WeakInlines.h:
* heap/WeakSet.cpp:
* heap/WeakSet.h:
* interpreter/CallFrame.cpp:
* interpreter/CallFrame.h:
* interpreter/Interpreter.cpp:
* interpreter/ProtoCallFrame.h:
* interpreter/StackVisitor.cpp:
* interpreter/StackVisitor.h:
* jit/AssemblyHelpers.h:
* jit/CCallHelpers.h:
* jit/JITCode.cpp:
* jit/JITOperations.cpp:
* jit/Repatch.cpp:
* jit/ThunkGenerators.cpp:
* jsc.cpp:
* llint/LLIntSlowPaths.cpp:
* parser/ASTBuilder.h:
* parser/Lexer.cpp:
* parser/Lexer.h:
* parser/Nodes.cpp:
* parser/Nodes.h:
* parser/Parser.cpp:
* parser/Parser.h:
* parser/ParserArena.cpp:
* parser/ParserArena.h:
* parser/ParserFunctionInfo.h:
* parser/SyntaxChecker.h:
* parser/UnlinkedSourceCode.h:
* profiler/ProfilerBytecodeSequence.cpp:
* profiler/ProfilerCompilation.cpp:
* profiler/ProfilerDatabase.cpp:
* profiler/ProfilerOSRExitSite.cpp:
* profiler/ProfilerOriginStack.cpp:
* runtime/ArgList.h:
* runtime/ArrayPrototype.cpp:
* runtime/ClonedArguments.cpp:
* runtime/CommonSlowPaths.cpp:
* runtime/Completion.h:
* runtime/DataView.h:
* runtime/DatePrototype.cpp:
* runtime/DirectEvalExecutable.cpp:
* runtime/DumpContext.cpp:
* runtime/FunctionExecutable.cpp:
* runtime/IndirectEvalExecutable.cpp:
* runtime/JSArray.cpp:
* runtime/JSArrayBufferView.cpp:
* runtime/JSCJSValue.cpp:
* runtime/JSCJSValueInlines.h:
* runtime/JSCell.cpp:
* runtime/JSDataView.cpp:
* runtime/JSDestructibleObject.h:
* runtime/JSFunction.cpp:
* runtime/JSGlobalObject.cpp:
* runtime/JSGlobalObject.h:
* runtime/JSONObject.cpp:
* runtime/JSObject.cpp:
* runtime/JSObject.h:
* runtime/JSScope.cpp:
* runtime/JSScope.h:
* runtime/LiteralParser.cpp:
* runtime/OptionsList.h:
* runtime/PropertyDescriptor.cpp:
* runtime/PropertyMapHashTable.h:
* runtime/PropertySlot.h:
* runtime/PutPropertySlot.h:
* runtime/RegExpMatchesArray.h:
* runtime/RegExpPrototype.cpp:
* runtime/StringPrototype.cpp:
* runtime/Structure.cpp:
* runtime/Structure.h:
* runtime/TestRunnerUtils.cpp:
* runtime/TypedArrayType.cpp:
* runtime/VM.cpp:
* runtime/Watchdog.cpp:
* runtime/Watchdog.h:
* runtime/WriteBarrier.h:
* testRegExp.cpp:
* tools/JSDollarVM.cpp:
* wasm/WasmSlowPaths.cpp:
* yarr/RegularExpression.h:
* yarr/YarrInterpreter.cpp:
* yarr/YarrJIT.cpp:
* yarr/YarrJIT.h:
* yarr/YarrPattern.cpp:
* yarr/YarrPattern.h:
2020-05-09 Ross Kirsling <ross.kirsling@sony.com>
Fix build errors and warnings for non-unified JSCOnly
https://bugs.webkit.org/show_bug.cgi?id=211655
Reviewed by Darin Adler and Yusuke Suzuki.
* bytecode/BytecodeDumper.cpp:
(JSC::isConstantRegisterIndex): Deleted.
Remove unused function.
* llint/LLIntEntrypoint.cpp:
* llint/LLIntThunks.cpp:
* llint/LLIntThunks.h:
* runtime/AggregateErrorConstructor.cpp:
* runtime/AggregateErrorPrototype.cpp:
* wasm/js/WebAssemblyFunction.cpp:
Fix includes.
* tools/JSDollarVM.cpp:
Deal with "unused constant" warnings for needsDestruction.
* wasm/WasmLLIntPlan.cpp:
* wasm/WasmSignature.cpp:
Remove unused constants.
2020-05-08 Darin Adler <darin@apple.com>
Streamline MarkupAccumulator to improve efficiency a bit
https://bugs.webkit.org/show_bug.cgi?id=211656
Reviewed by Anders Carlsson.
* b3/air/AirFixPartialRegisterStalls.h: Fix spelling of "explicitly".
2020-05-08 Alexey Shvayka <shvaikalesh@gmail.com>
Array.prototype.concat fast path checks should not be observable
https://bugs.webkit.org/show_bug.cgi?id=211643
Reviewed by Ross Kirsling.
This change utilizes @tryGetByIdWithWellKnownSymbol intrinsic to make
off the spec Symbol.isConcatSpreadable lookups unobservable to userland code,
aligning JSC with V8 and SpiderMonkey.
Since @tryGetById uses PropertySlot::getPureResult(), which returns `null`
for Proxy [[Get]] traps and JS getters (covered by stress/try-get-by-id.js),
we can safely compare its result `undefined`. Also, this allows us to remove
@isProxyObject check as Proxy argument is never a fast path anyway.
This patch is neutral on microbenchmarks/concat-append-one.js.
* builtins/ArrayPrototype.js:
(concat):
2020-05-07 Michael Catanzaro <mcatanzaro@gnome.org>
Simplify preprocessor guards in GCMemoryOperations.h
https://bugs.webkit.org/show_bug.cgi?id=211588
Reviewed by Darin Adler.
If we adjust the guards a bit, then we don't need to repeat the fallback path.
* heap/GCMemoryOperations.h:
(JSC::gcSafeMemmove):
(JSC::gcSafeZeroMemory):
2020-05-07 Mark Lam <mark.lam@apple.com>
Give the DFG and FTL WorkList threads more stack space on ASAN builds.
https://bugs.webkit.org/show_bug.cgi?id=211535
<rdar://problem/62947884>
Reviewed by Geoffrey Garen.
* dfg/DFGWorklist.cpp:
(JSC::DFG::Worklist::ThreadBody::ThreadBody):
- Mark the AutomaticThread as ThreadType::Compiler.
2020-05-07 Daniel Kolesa <daniel@octaforge.org>
REGRESSION(r251875): Crash in JSC::StructureIDTable::get on ppc64le: gcSafeMemcpy broken on JSVALUE64 platforms other than x86_64 and aarch64
https://bugs.webkit.org/show_bug.cgi?id=210685
Reviewed by Michael Catanzaro.
Fix gcSafeMemcpy on non-x86_64/aarch64 64-bit architectures.
We were hitting an incorrect x86_64 assertion on values larger than
mediumCutoff on JSVALUE64 architectures other than x86_64 and aarch64,
as the control flow is wrong.
* heap/GCMemoryOperations.h:
(JSC::gcSafeMemcpy):
2020-05-07 Mark Lam <mark.lam@apple.com>
Add stack checks to the DFG and FTL bytecode parser.
https://bugs.webkit.org/show_bug.cgi?id=211547
<rdar://problem/62958880>
Reviewed by Yusuke Suzuki.
Inlining can cause some level of recursion of the DFG bytecode parser. We should
do a stack check at each inlining check before recursing. If a stack overflow
appears to be imminent, then just refuse to inline, and therefore, don't recurse
deeper into the parser.
This issue is more noticeable on ASan debug builds where stack frames can be
humongous.
Removed the SUPPRESS_ASAN on cloberrize() and the associated comment from r260692.
It was a mis-diagnosis. The stack checks are what we need.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleVarargsInlining):
(JSC::DFG::ByteCodeParser::handleInlining):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGGraph.h:
2020-05-07 Darin Adler <darin@apple.com>
REGRESSION (r261257): Lifetime problem with upconverted characters in toLocaleCase
https://bugs.webkit.org/show_bug.cgi?id=211580
rdar://62980449
Reviewed by Yusuke Suzuki.
The problem comes from the fact that callBufferProducingFunction is moving the same
arguments multiple times. At the moment, this works around the only practical
problem with that, but later it should be fixed in callBufferProducingFunction.
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::initializeDateTimeFormat): Work around mistakes in how
callBufferProducingFunction works with arguments by calling get() explicitly on the
result of upconvertedCharacters. Later we could fix callBufferProducingFunction to
be safer, but for now this solves the problem.
* runtime/StringPrototype.cpp:
(JSC::toLocaleCase): Ditto.
2020-05-07 Keith Miller <keith_miller@apple.com>
Fix ArrayMode nodes after r261260
https://bugs.webkit.org/show_bug.cgi?id=211543
Reviewed by Yusuke Suzuki.
I accidentally ran tests with a release build rather than
release+assert when uploading r261260. This patch skips the
CheckArray node in the ArrayMode clobbersTop() logic before
Fixup. And also marks a GetArrayLength in the TypedArray
intrsinics as ExitOK.
This patch also relands r261260.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicGetter):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
2020-05-07 Ryan Haddad <ryanhaddad@apple.com>
Unreviewed, reverting r261260.
Caused 26 JSC test failures
Reverted changeset:
"DFG ByVal nodes with ArrayModes should clobberTop until Fixup
phase runs."
https://bugs.webkit.org/show_bug.cgi?id=211531
https://trac.webkit.org/changeset/261260
2020-05-07 Mark Lam <mark.lam@apple.com>
Fix broken exceptionFuzz tests.
https://bugs.webkit.org/show_bug.cgi?id=211550
Reviewed by Yusuke Suzuki.
Remove the bad and now unused utility function to set Options::useExceptionFuzz().
* tools/JSDollarVM.cpp:
(JSC::JSDollarVM::finishCreation):
(JSC::functionEnableExceptionFuzz): Deleted.
2020-05-06 Keith Miller <keith_miller@apple.com>
DFG ByVal nodes with ArrayModes should clobberTop until Fixup phase runs.
https://bugs.webkit.org/show_bug.cgi?id=211531
Reviewed by Yusuke Suzuki.
When parsing bytecode we may pick a relatively constrained
ArrayMode based on our profiling. Some of these modes may not
clobber exit state. However, Fixup sometimes wants to widen this
to a more generic mode based on other data. This causes us to
think it was valid to exit immediately after the
GetByVal/HasIndexedProperty, which would be wrong with the wider
ArrayMode. We may also incorrectly insert invalidition points
if clobberize gives us the wrong data.
To fix this clobberize should say All ByVal nodes clobberTop()
until after fixup. Additionally, this patch adds an assertion that
nodes don't go from not clobbering exit state to clobbering exit
state during fixup.
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::performFixup):
* dfg/DFGGraph.h:
2020-05-06 Darin Adler <darin@apple.com>
Make a helper for the pattern of ICU functions that may need to be called twice to populate a buffer
https://bugs.webkit.org/show_bug.cgi?id=211499
Reviewed by Ross Kirsling.
* runtime/IntlDateTimeFormat.cpp:
(JSC::defaultTimeZone): Use callBufferProducingFunction.
(JSC::canonicalizeTimeZoneName): Ditto.
(JSC::IntlDateTimeFormat::initializeDateTimeFormat): Ditto.
(JSC::IntlDateTimeFormat::format const): Ditto.
(JSC::IntlDateTimeFormat::formatToParts const): Ditto.
* runtime/IntlLocale.cpp:
(JSC::LocaleIDBuilder::toCanonical): Ditto.
(JSC::IntlLocale::language): Ditto.
(JSC::IntlLocale::script): Ditto.
(JSC::IntlLocale::region): Ditto.
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::format const): Ditto.
(JSC::IntlNumberFormat::formatToParts const): Ditto.
* runtime/IntlObject.cpp:
(JSC::languageTagForLocaleID): Ditto.
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::formatInternal const): Ditto.
(JSC::IntlRelativeTimeFormat::formatToParts const): Ditto.
* runtime/StringPrototype.cpp:
(JSC::toLocaleCase): Ditto.
2020-05-06 Devin Rousso <drousso@apple.com>
ASSERT_WITH_MESSAGE(m_isOwnedByMainThread == isMainThread()) when web inspecting
https://bugs.webkit.org/show_bug.cgi?id=203638
<rdar://problem/56761893>
Reviewed by Brian Burg.
Mark the `InspectorEnvironment::executionStopwatch` abstract function as `const` and have it
return a `Stopwatch&` instead of a `RefPtr<Stopwatch>&` as callers assume that it exists.
By not using a `RefPtr`, an additional `copyRef` can be avoided.
* inspector/InspectorEnvironment.h:
* inspector/JSGlobalObjectInspectorController.h:
* inspector/JSGlobalObjectInspectorController.cpp:
(Inspector::JSGlobalObjectInspectorController::executionStopwatch const): Added.
(Inspector::JSGlobalObjectInspectorController::executionStopwatch): Deleted.
* inspector/agents/InspectorDebuggerAgent.cpp:
(Inspector::InspectorDebuggerAgent::didPause):
(Inspector::InspectorDebuggerAgent::breakpointActionProbe):
(Inspector::InspectorDebuggerAgent::didContinue):
* inspector/agents/InspectorHeapAgent.cpp:
(Inspector::InspectorHeapAgent::snapshot):
(Inspector::InspectorHeapAgent::willGarbageCollect):
(Inspector::InspectorHeapAgent::didGarbageCollect):
* inspector/agents/InspectorScriptProfilerAgent.cpp:
(Inspector::InspectorScriptProfilerAgent::startTracking):
(Inspector::InspectorScriptProfilerAgent::willEvaluateScript):
(Inspector::InspectorScriptProfilerAgent::didEvaluateScript):
(Inspector::InspectorScriptProfilerAgent::trackingComplete):
* runtime/SamplingProfiler.h:
* runtime/SamplingProfiler.cpp:
(JSC::SamplingProfiler::SamplingProfiler):
* runtime/VM.h:
* runtime/VM.cpp:
(JSC::VM::ensureSamplingProfiler):
2020-05-05 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] Implement Intl.Locale
https://bugs.webkit.org/show_bug.cgi?id=209772
Reviewed by Darin Adler and Saam Barati.
This patch implements the recent ECMA-402 feature Intl.Locale.
This is effectively a wrapper class for all the pieces of uloc.h that ECMA-402 cares about.
(If we used the C++ API, there's a LocaleBuilder that would make this much easier, but in sticking to the C API,
it's basically an object that has an ICU localeID as data and uloc_* functions as methods / getters.
Furthermore, there's no way to modify said data, so every method / getter can be lazy and cache its result.)
Usage example:
>>> locale = new Intl.Locale('ja', { region: 'JP', calendar: 'japanese', numeric: false })
"ja-JP-u-ca-japanese-kn-false"
>>> locale.baseName
"ja-JP"
Intl.Locale can be used anywhere that Intl APIs accept locale strings as input parameters,
and is moreover hoped to be the class by which future Web APIs will handle the current locale.
This feature is runtime-guarded by the `useIntlLocale` option.
* CMakeLists.txt:
* DerivedSources-input.xcfilelist:
* DerivedSources-output.xcfilelist:
* DerivedSources.make:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* runtime/CommonIdentifiers.h:
* runtime/IntlLocale.cpp: Added.
* runtime/IntlLocale.h: Added.
* runtime/IntlLocaleConstructor.cpp: Added.
* runtime/IntlLocaleConstructor.h: Added.
* runtime/IntlLocalePrototype.cpp: Added.
* runtime/IntlLocalePrototype.h: Added.
* runtime/IntlObject.cpp:
(JSC::IntlObject::finishCreation):
(JSC::localeIDBufferForLanguageTag): Added.
(JSC::languageTagForLocaleID): Renamed from JSC::convertICULocaleToBCP47LanguageTag.
(JSC::intlAvailableLocales):
(JSC::intlCollatorAvailableLocales):
(JSC::canonicalizeLanguageTag):
(JSC::canonicalizeLocaleList):
(JSC::defaultLocale):
* runtime/IntlObject.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
(JSC::JSGlobalObject::visitChildren):
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::collatorStructure):
(JSC::JSGlobalObject::numberFormatStructure):
(JSC::JSGlobalObject::localeStructure):
* runtime/OptionsList.h:
* runtime/VM.cpp:
(JSC::VM::VM):
* runtime/VM.h:
2020-05-05 Keith Miller <keith_miller@apple.com>
clobberize validator should use branchTest8 directly.
https://bugs.webkit.org/show_bug.cgi?id=211469
Reviewed by Yusuke Suzuki.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileCurrentBlock):
2020-05-05 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Implement BigInt.asIntN and BigInt.asUintN
https://bugs.webkit.org/show_bug.cgi?id=181144
Reviewed by Darin Adler.
This patch implements BigInt.asIntN[1] and BigInt.asUintN[2] features.
As the same to the other BigInt runtime C++ code, we port V8 code to JSC to implement both.
BigInt.asIntN is `static_cast<intN_t>(BigInt value)` and BigInt.asUintN is `static_cast<uintN_t>(BigInt value)`.
They are getting slice of N bits from two's complement representation of the given BigInt. The difference between
asIntN and asUintN is asIntN renders MSB as a sign.
This patch is once rolled out due to ARM64_32 build failure, which is caused by the existing bug[3]. Relanding it
since it is now fixed.
[1]: https://tc39.es/ecma262/#sec-bigint.asintn
[2]: https://tc39.es/ecma262/#sec-bigint.asuintn
[3]: https://trac.webkit.org/changeset/261174/webkit
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::bigIntConstructorFuncAsUintN):
(JSC::bigIntConstructorFuncAsIntN):
* runtime/JSBigInt.cpp:
(JSC::zeroImpl):
(JSC::JSBigInt::divideImpl):
(JSC::JSBigInt::unaryMinusImpl):
(JSC::JSBigInt::remainderImpl):
(JSC::JSBigInt::digitDiv):
(JSC::JSBigInt::absoluteSub):
(JSC::JSBigInt::asIntNImpl):
(JSC::JSBigInt::asUintNImpl):
(JSC::JSBigInt::truncateToNBits):
(JSC::JSBigInt::truncateAndSubFromPowerOfTwo):
(JSC::JSBigInt::asIntN):
(JSC::JSBigInt::asUintN):
* runtime/JSBigInt.h:
2020-05-05 Ross Kirsling <ross.kirsling@sony.com>
[Intl] Alphabetize extension keys and correctly mark const methods
https://bugs.webkit.org/show_bug.cgi?id=211359
Reviewed by Darin Adler.
Two cleanup items for Intl classes:
1. Ensure `resolvedOptions().locale` returns relevant extension keys in alphabetical order.
ICU does this for us via Intl.getCanonicalLocales / Intl.*.supportedLocalesOf but not via ResolveLocale.
However, we don't need to do any sorting in ResolveLocale; we can just pre-alphabetize relevantExtensionKeys.
(See also https://github.com/tc39/ecma402/pull/433.)
2. Ensure Intl classes are marking const methods correctly.
* runtime/IntlCollator.cpp:
(JSC::IntlCollator::sortLocaleData):
(JSC::IntlCollator::searchLocaleData):
(JSC::IntlCollator::compareStrings const): Add const specifier.
(JSC::IntlCollator::resolvedOptions const): Add const specifier.
* runtime/IntlCollator.h:
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::localeData):
(JSC::IntlDateTimeFormat::resolvedOptions const): Add const specifier.
(JSC::IntlDateTimeFormat::format const): Add const specifier.
(JSC::IntlDateTimeFormat::formatToParts const): Add const specifier.
* runtime/IntlDateTimeFormat.h:
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::format const): Add const specifier.
(JSC::IntlNumberFormat::resolvedOptions const): Add const specifier.
(JSC::IntlNumberFormat::formatToParts const): Add const specifier.
* runtime/IntlNumberFormat.h:
* runtime/IntlPluralRules.cpp:
(JSC::IntlPluralRules::resolvedOptions const): Add const specifier.
(JSC::IntlPluralRules::select const): Add const specifier.
* runtime/IntlPluralRules.h:
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::resolvedOptions const): Add const specifier.
(JSC::IntlRelativeTimeFormat::formatInternal const): Add const specifier.
(JSC::IntlRelativeTimeFormat::format const): Add const specifier.
(JSC::IntlRelativeTimeFormat::formatToParts const): Add const specifier.
* runtime/IntlRelativeTimeFormat.h:
2020-05-05 Keith Miller <keith_miller@apple.com>
Add Clobberize validator for clobber top.
https://bugs.webkit.org/show_bug.cgi?id=209432
Reviewed by Yusuke Suzuki.
* assembler/MacroAssemblerARMv7.h:
(JSC::MacroAssemblerARMv7::scratchRegister):
* assembler/MacroAssemblerMIPS.h:
(JSC::MacroAssemblerMIPS::scratchRegister):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileCurrentBlock):
* dfg/DFGSpeculativeJIT64.cpp:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::lower):
(JSC::FTL::DFG::LowerDFGToB3::compileBlock):
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
* interpreter/Interpreter.cpp:
(JSC::eval):
(JSC::Interpreter::executeProgram):
(JSC::Interpreter::executeCall):
(JSC::Interpreter::executeConstruct):
(JSC::Interpreter::execute):
(JSC::Interpreter::executeModuleProgram):
* jit/JITCodeInlines.h:
(JSC::JITCode::execute):
* llint/LLIntThunks.h:
(JSC::vmEntryToWasm):
* runtime/OptionsList.h:
* runtime/VM.h:
2020-05-05 Mark Lam <mark.lam@apple.com>
Allow Bitmap to use up to a UCPURegister word size for internal bit storage.
https://bugs.webkit.org/show_bug.cgi?id=211328
<rdar://problem/62755865>
Reviewed by Yusuke Suzuki.
* assembler/CPU.h:
2020-05-05 Keith Miller <keith_miller@apple.com>
iterator_open should remap the symbolIterator argument correctly when inlined.
https://bugs.webkit.org/show_bug.cgi?id=211308
<rdar://problem/62287877>
Reviewed by Mark Lam.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
2020-05-05 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSBigInt::maxLengthBits and JSBigInt::maxLength are wrong
https://bugs.webkit.org/show_bug.cgi?id=211445
Reviewed by Mark Lam.
JSBigInt::maxLengthBits and JSBigInt::maxLength definitions are wrong.
1. We are defining maxLength and maxLengthBits as an unrelated value to each other. This is wrong.
maxLength should be defined as maxLengthBits / (sizeof(Digit) * bitsPerByte).
2. We use `sizeof(void*)` and assume that `sizeof(Digit) == sizeof(void*)`. This is wrong in ARM64_32 environment
where Digit size is sizeof(uint64_t) while the pointer size is sizeof(uint32_t). This causes compile errors in ARM64_32
when the code is using these values with static_assert.
* runtime/JSBigInt.h:
2020-05-05 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, reverting r261156.
Break ARM64_32 build due to existing bug
Reverted changeset:
"[JSC] Implement BigInt.asIntN and BigInt.asUintN"
https://bugs.webkit.org/show_bug.cgi?id=181144
https://trac.webkit.org/changeset/261156
2020-05-05 Alexey Shvayka <shvaikalesh@gmail.com>
Object.prototype.toString is not spec-perfect
https://bugs.webkit.org/show_bug.cgi?id=199138
Reviewed by Darin Adler and Keith Miller.
Before ES6, Object.prototype.toString relied only on internal [[Class]] slot. Starting with ES6,
Object.prototype.toString checks for a handful of internal slots, mimicing [[Class]], to ensure
backwards compatibility for pre-ES6 instances. Newly-added built-ins provide @@toStringTag for
the method to use.
Before this change, Object.prototype.toString in JSC relied on className() a.k.a [[Class]] for
all instances. For (almost all) new built-ins, it was overriden by toStringName() returning
"Object", while @@toStringTag was set to correct value. This is quite an error-prone approach
and observable spec discrepancy if @@toStringTag is deleted or set to a non-string.
This change eliminates the above-mentioned discrepancy and fixes Object.prototype.toString
to return "[object Function]" for callable Proxy objects, aligning JSC with the spec [1], V8,
and SpiderMonkey.
For Object.prototype.toString to work through DebuggerScope and JSProxy, we perform all checks
in JSObject::toStringName(). Given that isArray() may throw a TypeError [2], we invoke
toStringName() before @@toStringTag lookup to accomodate revoked Proxy case.
Also, this patch defines @@toStringTag for WebAssembly namespace object (to match Chrome),
JSC shell, and ConsoleObject.
[1]: https://tc39.es/ecma262/#sec-object.prototype.tostring
[2]: https://tc39.es/ecma262/#sec-isarray (step 3.a)
* jsc.cpp:
* runtime/BigIntObject.cpp:
(JSC::BigIntObject::toStringName): Deleted.
* runtime/BigIntObject.h:
* runtime/BooleanObject.cpp:
(JSC::BooleanObject::toStringName):
* runtime/BooleanObject.h:
* runtime/ConsoleObject.cpp:
(JSC::ConsoleObject::finishCreation):
* runtime/DateInstance.cpp:
(JSC::DateInstance::toStringName):
* runtime/DateInstance.h:
* runtime/ErrorInstance.cpp:
(JSC::ErrorInstance::toStringName):
* runtime/ErrorInstance.h:
* runtime/JSArrayBufferView.cpp:
(JSC::JSArrayBufferView::toStringName): Deleted.
* runtime/JSArrayBufferView.h:
* runtime/JSMap.cpp:
(JSC::JSMap::toStringName): Deleted.
* runtime/JSMap.h:
* runtime/JSObject.cpp:
(JSC::JSObject::toStringName):
* runtime/JSSet.cpp:
(JSC::JSSet::toStringName): Deleted.
* runtime/JSSet.h:
* runtime/JSWeakMap.cpp:
(JSC::JSWeakMap::toStringName): Deleted.
* runtime/JSWeakMap.h:
* runtime/JSWeakObjectRef.cpp:
(JSC::JSWeakObjectRef::toStringName): Deleted.
* runtime/JSWeakObjectRef.h:
* runtime/JSWeakSet.cpp:
(JSC::JSWeakSet::toStringName): Deleted.
* runtime/JSWeakSet.h:
* runtime/NumberObject.cpp:
(JSC::NumberObject::toStringName):
* runtime/NumberObject.h:
* runtime/ObjectPrototype.cpp:
(JSC::objectProtoFuncToString):
* runtime/ProxyObject.cpp:
(JSC::ProxyObject::toStringName): Deleted.
* runtime/ProxyObject.h:
* runtime/RegExpObject.cpp:
(JSC::RegExpObject::toStringName):
* runtime/RegExpObject.h:
* runtime/StringObject.cpp:
(JSC::StringObject::toStringName):
* runtime/StringObject.h:
* runtime/SymbolObject.cpp:
(JSC::SymbolObject::toStringName): Deleted.
* runtime/SymbolObject.h:
* wasm/js/JSWebAssembly.cpp:
(JSC::JSWebAssembly::finishCreation):
2020-05-04 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Implement BigInt.asIntN and BigInt.asUintN
https://bugs.webkit.org/show_bug.cgi?id=181144
Reviewed by Darin Adler.
This patch implements BigInt.asIntN[1] and BigInt.asUintN[2] features.
As the same to the other BigInt runtime C++ code, we port V8 code to JSC to implement both.
BigInt.asIntN is `static_cast<intN_t>(BigInt value)` and BigInt.asUintN is `static_cast<uintN_t>(BigInt value)`.
They are getting slice of N bits from two's complement representation of the given BigInt. The difference between
asIntN and asUintN is asIntN renders MSB as a sign.
[1]: https://tc39.es/ecma262/#sec-bigint.asintn
[2]: https://tc39.es/ecma262/#sec-bigint.asuintn
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::bigIntConstructorFuncAsUintN):
(JSC::bigIntConstructorFuncAsIntN):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::zeroImpl):
(JSC::JSBigInt::divideImpl):
(JSC::JSBigInt::unaryMinusImpl):
(JSC::JSBigInt::remainderImpl):
(JSC::JSBigInt::digitDiv):
(JSC::JSBigInt::asIntNImpl):
(JSC::JSBigInt::asUintNImpl):
(JSC::JSBigInt::truncateToNBits):
(JSC::JSBigInt::truncateAndSubFromPowerOfTwo):
(JSC::JSBigInt::asIntN):
(JSC::JSBigInt::asUintN):
* runtime/JSBigInt.h:
2020-05-04 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] DFG NotCellUse is used without considering about BigInt32
https://bugs.webkit.org/show_bug.cgi?id=211395
Reviewed by Saam Barati.
When we see CompareXXX(BigInt32, Double), we are emitting CompareXXX(DoubleRep(BigInt:NotCellUse), Double). But this has two problems.
1. We should emit CompareXXX(UntypedUse, UntypedUse) in this case.
2. DoubleRep(NotCellUse) does not support converting BigInt32 to double. Since DoubleRep's semantics is for ToNumber, it should not
accept BigInt32 since it should throw an error. However, DoubleRep currently assumes that NotCellUse value can be converted to double
without any errors.
To keep DoubleRep's semantics ToNumber, we replace NotCellUse with NotCellNorBigIntUse, which rejects BigInt32. This patch also uses NotCellNorBigIntUse
for ValueToInt32 because of the same reason.
For CompareXXX and CompareEq nodes, we can optimize it if we introduce new DoubleRepAcceptingBigInt32 DFG node which can convert BigInt32 to Double, since
CompareXXX and CompareEq are not requiring toNumber semantics. This should be done in a separate bug https://bugs.webkit.org/show_bug.cgi?id=211407.
* bytecode/SpeculatedType.h:
(JSC::isNotCellNorBigIntSpeculation):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::fixIntConvertingEdge):
(JSC::DFG::FixupPhase::fixupChecksInBlock):
* dfg/DFGNode.h:
(JSC::DFG::Node::shouldSpeculateNotCellNorBigInt):
* dfg/DFGSafeToExecute.h:
(JSC::DFG::SafeToExecuteEdge::operator()):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileValueToInt32):
(JSC::DFG::SpeculativeJIT::compileDoubleRep):
(JSC::DFG::SpeculativeJIT::speculateNotCellNorBigInt):
(JSC::DFG::SpeculativeJIT::speculate):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGUseKind.cpp:
(WTF::printInternal):
* dfg/DFGUseKind.h:
(JSC::DFG::typeFilterFor):
(JSC::DFG::checkMayCrashIfInputIsEmpty):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileDoubleRep):
(JSC::FTL::DFG::LowerDFGToB3::compileValueToInt32):
(JSC::FTL::DFG::LowerDFGToB3::numberOrNotCellNorBigIntToInt32):
(JSC::FTL::DFG::LowerDFGToB3::speculate):
(JSC::FTL::DFG::LowerDFGToB3::speculateNotCellNorBigInt):
(JSC::FTL::DFG::LowerDFGToB3::numberOrNotCellToInt32): Deleted.
2020-05-04 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Add @@toStringTag to WebAssembly.Global
https://bugs.webkit.org/show_bug.cgi?id=211372
Reviewed by Sam Weinig.
As r260992 did for the other wasm prototypes, we should put @@toStringTag to WebAssembly.Global's prototype too.
* wasm/js/WebAssemblyGlobalPrototype.cpp:
(JSC::WebAssemblyGlobalPrototype::finishCreation):
2020-05-04 Devin Rousso <drousso@apple.com>
Web Inspector: Worker: should use the name of the worker if it exists
https://bugs.webkit.org/show_bug.cgi?id=211244
Reviewed by Brian Burg.
* inspector/protocol/Worker.json:
Include the `name` in `Worker.workerCreated`.
2020-05-04 Devin Rousso <drousso@apple.com>
Web Inspector: provide a way for inspector to turn on/off ITP debug mode and AdClickAttribution debug mode
https://bugs.webkit.org/show_bug.cgi?id=209763
Reviewed by Brian Burg.
* inspector/protocol/Page.json:
Add new enum values to `Page.Setting`:
- `AdClickAttributionDebugModeEnabled`
- `ITPDebugModeEnabled`
2020-05-03 Maciej Stachowiak <mjs@apple.com>
Remove no longer needed WebKitAdditions include for JavaScriptCorePrefix.h
https://bugs.webkit.org/show_bug.cgi?id=211357
Reviewed by Mark Lam.
* JavaScriptCorePrefix.h:
2020-05-02 Mark Lam <mark.lam@apple.com>
Gardening: rolling out r261050 and r261051.
https://bugs.webkit.org/show_bug.cgi?id=211328
<rdar://problem/62755865>
Not reviewed.
* assembler/CPU.h:
2020-05-01 Mark Lam <mark.lam@apple.com>
Allow Bitmap to use up to a UCPURegister word size for internal bit storage.
https://bugs.webkit.org/show_bug.cgi?id=211328
<rdar://problem/62755865>
Reviewed by Yusuke Suzuki.
* assembler/CPU.h:
2020-05-01 Saam Barati <sbarati@apple.com>
Have a thread local cache for the Wasm LLInt bytecode buffer
https://bugs.webkit.org/show_bug.cgi?id=211317
Reviewed by Filip Pizlo and Mark Lam.
One of the main things slowing down Wasm compile times is the banging
on bmalloc's global heap lock. This patch makes it so for the bytecode
instruction buffer, we keep a thread local cache with latest capacity
the thread needed to compile. This makes it so that in the average case,
we only do one malloc at the end of a compile to memcpy the final result.
We clear these thread local caches when the WasmWorklist's automatic threads
underlying machine thread is destroyed.
This is a 15% speedup in zen garden compile times on a 16-core Mac Pro.
This is a 4-5% speedup in zen garden compile times on a 6-core MBP.
* bytecode/InstructionStream.h:
(JSC::InstructionStreamWriter::setInstructionBuffer):
(JSC::InstructionStreamWriter::finalize):
* wasm/WasmLLIntGenerator.cpp:
(JSC::Wasm::threadSpecificBuffer):
(JSC::Wasm::clearLLIntThreadSpecificCache):
(JSC::Wasm::LLIntGenerator::LLIntGenerator):
(JSC::Wasm::LLIntGenerator::finalize):
* wasm/WasmLLIntGenerator.h:
* wasm/WasmWorklist.cpp:
2020-05-01 Per Arne Vollan <pvollan@apple.com>
[Win] Fix AppleWin build
https://bugs.webkit.org/show_bug.cgi?id=211324
Reviewed by Don Olmstead.
Check if target WTF_CopyHeaders exists before using it.
* CMakeLists.txt:
2020-05-01 Don Olmstead <don.olmstead@sony.com>
[GTK] Add additional exports to support hidden visibility
https://bugs.webkit.org/show_bug.cgi?id=211246
Reviewed by Michael Catanzaro.
* API/glib/JSCContextPrivate.h:
* API/glib/JSCValuePrivate.h:
* inspector/remote/glib/RemoteInspectorServer.h:
* inspector/remote/glib/RemoteInspectorUtils.h:
2020-05-01 Don Olmstead <don.olmstead@sony.com>
Use export macros on all platforms
https://bugs.webkit.org/show_bug.cgi?id=211293
Reviewed by Michael Catanzaro.
Allow overriding of JS_EXPORT_PRIVATE if desired otherwise use the defaults.
* runtime/JSExportMacros.h:
2020-05-01 Saam Barati <sbarati@apple.com>
Unreviewed. Non-speculative build fix for watchOS build.
* runtime/ArrayPrototype.cpp:
(JSC::shift):
(JSC::unshift):
(JSC::arrayProtoFuncToLocaleString):
(JSC::arrayProtoFuncReverse):
(JSC::arrayProtoFuncSlice):
(JSC::arrayProtoFuncSplice):
* runtime/JSONObject.cpp:
(JSC::Stringifier::Stringifier):
2020-05-01 Saam Barati <sbarati@apple.com>
Unreviewed. Speculative build fix for watchOS build.
* runtime/ArrayPrototype.cpp:
(JSC::shift):
2020-05-01 Alexey Shvayka <shvaikalesh@gmail.com>
[WebIDL] Interface prototype objects should define @@toStringTag
https://bugs.webkit.org/show_bug.cgi?id=211020
Unreviewed follow-up to r260992.
* runtime/JSArrayBufferPrototype.cpp:
(JSC::JSArrayBufferPrototype::finishCreation): Revert change in attempt to fix ARMv7 test.
2020-05-01 David Kilzer <ddkilzer@apple.com>
JSC::PropertySlot::m_attributes is uninitialized in constructor
<https://webkit.org/b/211267>
Reviewed by Mark Lam.
* runtime/PropertySlot.h:
(JSC::PropertySlot::PropertySlot):
- Initialize m_attributes and m_additionalData, and make use of
default initializers.
2020-05-01 Alexey Shvayka <shvaikalesh@gmail.com>
[WebIDL] Interface prototype objects should define @@toStringTag
https://bugs.webkit.org/show_bug.cgi?id=211020
Reviewed by Darin Adler.
WebIDL spec was recently updated [1] to define @@toStringTag on interface prototype objects.
This change aligns WebIDL with ECMA-262 built-ins and Blink's behavior. Gecko have also
expressed implementation commitment.
This patch implements the spec change, making `X.prototype.toString()` return "[object X]"
instead of "[object XPrototype]", where X is WebIDL interface. This behavior is proven to
be web compatible (shipping in Chrome since Q2 2016) and matches class strings of iterator
prototype objects [2] introduced in r253855.
We define @@toStringTag for all WebAssembly interfaces but Error subclasses since they
are not defined using WebIDL [3].
This change also introduces JSC_TO_STRING_TAG_WITHOUT_TRANSITION() macro that sets up
@@toStringTag using ClassInfo to avoid extra strings creation, ensuring `className` equality
between prototype and instance classes (fixing a few discrepancies), as well as correct
descriptors. It also ensures using faster jsNontrivialString() and relieves from putting
more code into CodeGeneratorJS.pm.
[1]: https://github.com/heycam/webidl/pull/357
[2]: https://heycam.github.io/webidl/#es-iterator-prototype-object
[3]: https://webassembly.github.io/spec/js-api/#error-objects
Tests: imported/w3c/web-platform-tests/wasm/jsapi/instance/toString.any.js
imported/w3c/web-platform-tests/wasm/jsapi/memory/toString.any.js
imported/w3c/web-platform-tests/wasm/jsapi/module/toString.any.js
imported/w3c/web-platform-tests/wasm/jsapi/table/toString.any.js
* runtime/ArrayIteratorPrototype.cpp:
(JSC::ArrayIteratorPrototype::finishCreation):
* runtime/AsyncFunctionPrototype.cpp:
(JSC::AsyncFunctionPrototype::finishCreation):
* runtime/AsyncGeneratorFunctionPrototype.cpp:
(JSC::AsyncGeneratorFunctionPrototype::finishCreation):
* runtime/AsyncGeneratorPrototype.cpp:
(JSC::AsyncGeneratorPrototype::finishCreation):
* runtime/BigIntPrototype.cpp:
(JSC::BigIntPrototype::finishCreation):
* runtime/GeneratorFunctionPrototype.cpp:
(JSC::GeneratorFunctionPrototype::finishCreation):
* runtime/GeneratorPrototype.cpp:
(JSC::GeneratorPrototype::finishCreation):
* runtime/IntlCollatorPrototype.cpp:
(JSC::IntlCollatorPrototype::finishCreation):
* runtime/IntlDateTimeFormatPrototype.cpp:
(JSC::IntlDateTimeFormatPrototype::finishCreation):
* runtime/IntlNumberFormatPrototype.cpp:
(JSC::IntlNumberFormatPrototype::finishCreation):
* runtime/IntlPluralRulesPrototype.cpp:
(JSC::IntlPluralRulesPrototype::finishCreation):
* runtime/IntlRelativeTimeFormatPrototype.cpp:
(JSC::IntlRelativeTimeFormatPrototype::finishCreation):
* runtime/JSArrayBufferPrototype.cpp:
(JSC::JSArrayBufferPrototype::finishCreation):
* runtime/JSDataViewPrototype.cpp:
(JSC::JSDataViewPrototype::finishCreation):
* runtime/JSONObject.cpp:
(JSC::JSONObject::finishCreation):
* runtime/JSObject.h:
* runtime/JSPromisePrototype.cpp:
(JSC::JSPromisePrototype::finishCreation):
* runtime/MapIteratorPrototype.cpp:
(JSC::MapIteratorPrototype::finishCreation):
* runtime/MapPrototype.cpp:
(JSC::MapPrototype::finishCreation):
* runtime/MathObject.cpp:
(JSC::MathObject::finishCreation):
* runtime/RegExpStringIteratorPrototype.cpp:
(JSC::RegExpStringIteratorPrototype::finishCreation):
* runtime/SetIteratorPrototype.cpp:
(JSC::SetIteratorPrototype::finishCreation):
* runtime/SetPrototype.cpp:
(JSC::SetPrototype::finishCreation):
* runtime/StringIteratorPrototype.cpp:
(JSC::StringIteratorPrototype::finishCreation):
* runtime/SymbolPrototype.cpp:
(JSC::SymbolPrototype::finishCreation):
* runtime/WeakMapPrototype.cpp:
(JSC::WeakMapPrototype::finishCreation):
* runtime/WeakObjectRefPrototype.cpp:
(JSC::WeakObjectRefPrototype::finishCreation):
* runtime/WeakSetPrototype.cpp:
(JSC::WeakSetPrototype::finishCreation):
* wasm/js/WebAssemblyInstancePrototype.cpp:
(JSC::WebAssemblyInstancePrototype::finishCreation):
* wasm/js/WebAssemblyMemoryPrototype.cpp:
(JSC::WebAssemblyMemoryPrototype::finishCreation):
* wasm/js/WebAssemblyModulePrototype.cpp:
(JSC::WebAssemblyModulePrototype::finishCreation):
* wasm/js/WebAssemblyTablePrototype.cpp:
(JSC::WebAssemblyTablePrototype::finishCreation):
2020-05-01 Saam Barati <sbarati@apple.com>
We can't cast toLength result to unsigned
https://bugs.webkit.org/show_bug.cgi?id=211205
<rdar://problem/62625562>
Reviewed by Yusuke Suzuki.
toLength, according to the spec, returns a 53-bit integer. In our
implementation, we return a double. However, there were many callsites
that did something like:
```
unsigned length = toLength(obj);
```
This is bad for a few reasons:
- Casting to unsigned from double is undefined behavior when the integer
is greater than UINT_MAX. In practice, this means that we'd have different
engine behavior depending on what architecture we'd be running on. For
example, if the length were UINT_MAX + 1, on x86, we'd treat the
length as zero. On arm64, we'd treat it as UINT_MAX. Both are wrong.
- We weren't spec compliant. We were just ignoring that these numbers could
be 53-bit integers.
This patch addresses each bad use of the undefined behavior, and by doing so,
makes us more spec compliant.
* dfg/DFGOperations.cpp:
* jit/JITOperations.cpp:
(JSC::getByVal):
* runtime/ArrayPrototype.cpp:
(JSC::getProperty):
(JSC::setLength):
(JSC::argumentClampedIndexFromStartOrEnd):
(JSC::shift):
(JSC::unshift):
(JSC::arrayProtoFuncToLocaleString):
(JSC::arrayProtoFuncPop):
(JSC::arrayProtoFuncPush):
(JSC::arrayProtoFuncReverse):
(JSC::arrayProtoFuncShift):
(JSC::arrayProtoFuncSlice):
(JSC::arrayProtoFuncSplice):
(JSC::arrayProtoFuncUnShift):
(JSC::fastIndexOf):
(JSC::arrayProtoFuncIndexOf):
(JSC::arrayProtoFuncLastIndexOf):
* runtime/Identifier.h:
(JSC::Identifier::from):
* runtime/IntlObject.cpp:
(JSC::canonicalizeLocaleList):
* runtime/JSONObject.cpp:
(JSC::Stringifier::Stringifier):
(JSC::Stringifier::Holder::appendNextProperty):
(JSC::Walker::walk):
* runtime/JSObject.cpp:
(JSC::JSObject::hasProperty const):
* runtime/JSObject.h:
(JSC::JSObject::putByIndexInline):
(JSC::JSObject::putDirectIndex):
(JSC::JSObject::canGetIndexQuickly const):
(JSC::JSObject::tryGetIndexQuickly const):
* runtime/JSObjectInlines.h:
(JSC::JSObject::getPropertySlot):
(JSC::JSObject::deleteProperty):
(JSC::JSObject::get const):
* runtime/PropertySlot.h:
(JSC::PropertySlot::getValue const):
* tools/JSDollarVM.cpp:
(JSC::functionSetUserPreferredLanguages):
2020-04-30 Ross Kirsling <ross.kirsling@sony.com>
TriState should be an enum class and use "Indeterminate" instead of "Mixed"
https://bugs.webkit.org/show_bug.cgi?id=211268
Reviewed by Mark Lam.
* b3/B3Const32Value.cpp:
(JSC::B3::Const32Value::equalConstant const):
(JSC::B3::Const32Value::notEqualConstant const):
(JSC::B3::Const32Value::lessThanConstant const):
(JSC::B3::Const32Value::greaterThanConstant const):
(JSC::B3::Const32Value::lessEqualConstant const):
(JSC::B3::Const32Value::greaterEqualConstant const):
(JSC::B3::Const32Value::aboveConstant const):
(JSC::B3::Const32Value::belowConstant const):
(JSC::B3::Const32Value::aboveEqualConstant const):
(JSC::B3::Const32Value::belowEqualConstant const):
* b3/B3Const64Value.cpp:
(JSC::B3::Const64Value::equalConstant const):
(JSC::B3::Const64Value::notEqualConstant const):
(JSC::B3::Const64Value::lessThanConstant const):
(JSC::B3::Const64Value::greaterThanConstant const):
(JSC::B3::Const64Value::lessEqualConstant const):
(JSC::B3::Const64Value::greaterEqualConstant const):
(JSC::B3::Const64Value::aboveConstant const):
(JSC::B3::Const64Value::belowConstant const):
(JSC::B3::Const64Value::aboveEqualConstant const):
(JSC::B3::Const64Value::belowEqualConstant const):
* b3/B3ConstDoubleValue.cpp:
(JSC::B3::ConstDoubleValue::equalConstant const):
(JSC::B3::ConstDoubleValue::notEqualConstant const):
(JSC::B3::ConstDoubleValue::lessThanConstant const):
(JSC::B3::ConstDoubleValue::greaterThanConstant const):
(JSC::B3::ConstDoubleValue::lessEqualConstant const):
(JSC::B3::ConstDoubleValue::greaterEqualConstant const):
(JSC::B3::ConstDoubleValue::equalOrUnorderedConstant const):
* b3/B3ConstFloatValue.cpp:
(JSC::B3::ConstFloatValue::equalConstant const):
(JSC::B3::ConstFloatValue::notEqualConstant const):
(JSC::B3::ConstFloatValue::lessThanConstant const):
(JSC::B3::ConstFloatValue::greaterThanConstant const):
(JSC::B3::ConstFloatValue::lessEqualConstant const):
(JSC::B3::ConstFloatValue::greaterEqualConstant const):
(JSC::B3::ConstFloatValue::equalOrUnorderedConstant const):
* b3/B3Procedure.cpp:
(JSC::B3::Procedure::addBoolConstant):
* b3/B3Procedure.h:
* b3/B3ReduceStrength.cpp:
* b3/B3Value.cpp:
(JSC::B3::Value::equalConstant const):
(JSC::B3::Value::notEqualConstant const):
(JSC::B3::Value::lessThanConstant const):
(JSC::B3::Value::greaterThanConstant const):
(JSC::B3::Value::lessEqualConstant const):
(JSC::B3::Value::greaterEqualConstant const):
(JSC::B3::Value::aboveConstant const):
(JSC::B3::Value::belowConstant const):
(JSC::B3::Value::aboveEqualConstant const):
(JSC::B3::Value::belowEqualConstant const):
(JSC::B3::Value::equalOrUnorderedConstant const):
(JSC::B3::Value::asTriState const):
* b3/B3Value.h:
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::~CodeBlock):
(JSC::CodeBlock::thresholdForJIT):
* bytecode/UnlinkedCodeBlock.cpp:
(JSC::UnlinkedCodeBlock::UnlinkedCodeBlock):
* bytecode/UnlinkedFunctionExecutable.cpp:
(JSC::UnlinkedFunctionExecutable::visitChildren):
* bytecompiler/NodesCodegen.cpp:
(JSC::ConstantNode::emitBytecodeInConditionContext):
(JSC::BinaryOpNode::emitBytecodeInConditionContext):
(JSC::BinaryOpNode::tryFoldToBranch):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
* dfg/DFGCFGSimplificationPhase.cpp:
(JSC::DFG::CFGSimplificationPhase::run):
* dfg/DFGLazyJSValue.cpp:
(JSC::DFG::equalToSingleCharacter):
(JSC::DFG::equalToStringImpl):
(JSC::DFG::LazyJSValue::strictEqual const):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileDataViewGet):
(JSC::FTL::DFG::LowerDFGToB3::compileDataViewSet):
* ftl/FTLOutput.cpp:
(JSC::FTL::Output::equal):
(JSC::FTL::Output::notEqual):
(JSC::FTL::Output::above):
(JSC::FTL::Output::aboveOrEqual):
(JSC::FTL::Output::below):
(JSC::FTL::Output::belowOrEqual):
(JSC::FTL::Output::greaterThan):
(JSC::FTL::Output::greaterThanOrEqual):
(JSC::FTL::Output::lessThan):
(JSC::FTL::Output::lessThanOrEqual):
* jit/JITOperations.cpp:
* runtime/CachedTypes.cpp:
(JSC::UnlinkedCodeBlock::UnlinkedCodeBlock):
* runtime/DefinePropertyAttributes.h:
(JSC::DefinePropertyAttributes::DefinePropertyAttributes):
(JSC::DefinePropertyAttributes::hasWritable const):
(JSC::DefinePropertyAttributes::writable const):
(JSC::DefinePropertyAttributes::hasConfigurable const):
(JSC::DefinePropertyAttributes::configurable const):
(JSC::DefinePropertyAttributes::hasEnumerable const):
(JSC::DefinePropertyAttributes::enumerable const):
(JSC::DefinePropertyAttributes::setWritable):
(JSC::DefinePropertyAttributes::setConfigurable):
(JSC::DefinePropertyAttributes::setEnumerable):
* runtime/IntlCollator.cpp:
(JSC::IntlCollator::initializeCollator):
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::initializeNumberFormat):
* runtime/IntlObject.cpp:
(JSC::intlBooleanOption):
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::pureStrictEqual):
(JSC::JSValue::pureToBoolean const):
* runtime/JSCellInlines.h:
(JSC::JSCell::pureToBoolean const):
2020-04-30 Ross Kirsling <ross.kirsling@sony.com>
[JSC] intlBooleanOption should return TriState instead of taking an out param
https://bugs.webkit.org/show_bug.cgi?id=211256
Reviewed by Darin Adler and Mark Lam.
Boolean options for Intl constructors can have default values of true, false, or undefined.
To handle the undefined case, intlBooleanOption currently has a `bool& usesFallback` param;
we should have the return type simply be a TriState instead.
* runtime/IntlCollator.cpp:
(JSC::IntlCollator::initializeCollator):
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::initializeNumberFormat):
* runtime/IntlObject.cpp:
(JSC::intlBooleanOption):
* runtime/IntlObject.h:
2020-04-30 Devin Rousso <drousso@apple.com>
WebKit.WebContent process crashes when web developer tools are opened in Safari
https://bugs.webkit.org/show_bug.cgi?id=210794
<rdar://problem/62214651>
Reviewed by Brian Burg.
* inspector/InjectedScriptManager.cpp:
(Inspector::InjectedScriptManager::injectedScriptFor):
Don't crash if a `TerminatedExecutionError` is thrown.
* inspector/InjectedScriptBase.cpp:
(Inspector::InjectedScriptBase::makeCall):
Report the actual error message. Check that the result has a value before attempting to make
a `JSON::Value` out of it.
2020-04-29 Ross Kirsling <ross.kirsling@sony.com>
Ensure Intl classes don't have naming conflicts with unified builds
https://bugs.webkit.org/show_bug.cgi?id=211213
Reviewed by Yusuke Suzuki.
Each Intl class usually has an array named relevantExtensionsKeys and a function named localeData.
This can result in redefinition errors when unified builds put two of them into the same translation unit.
Some are already guarding against this with an internal namespace while others are not.
As a uniform approach, this patch makes each localeData function a static method and
puts each relevantExtensionsKeys array (as well as any constants for its indices) into an internal namespace.
Furthermore, since three different classes are defining an identical UFieldPositionIteratorDeleter,
this patch consolidates them into one definition in IntlObject.
* runtime/IntlCollator.cpp:
(JSC::IntlCollator::sortLocaleData): Renamed from JSC::sortLocaleData.
(JSC::IntlCollator::searchLocaleData): Renamed from JSC::searchLocaleData.
(JSC::IntlCollator::initializeCollator):
* runtime/IntlCollator.h:
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::localeData): Renamed from JSC::IntlDTFInternal::localeData.
(JSC::toDateTimeOptionsAnyDate): Renamed from JSC::IntlDTFInternal::toDateTimeOptionsAnyDate.
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
(JSC::UFieldPositionIteratorDeleter::operator() const): Deleted.
* runtime/IntlDateTimeFormat.h:
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::localeData): Renamed from JSC::IntlNFInternal::localeData.
(JSC::IntlNumberFormat::initializeNumberFormat):
(JSC::UFieldPositionIteratorDeleter::operator() const): Deleted.
* runtime/IntlNumberFormat.h:
* runtime/IntlObject.cpp:
(JSC::UFieldPositionIteratorDeleter::operator() const): Added.
* runtime/IntlObject.h:
* runtime/IntlPluralRules.cpp:
(JSC::IntlPluralRules::localeData): Renamed from JSC::localeData.
* runtime/IntlPluralRules.h:
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::localeData): Renamed from JSC::localeData.
(JSC::IntlRelativeTimeFormat::initializeRelativeTimeFormat):
(JSC::UFieldPositionIteratorDeleter::operator() const): Deleted.
* runtime/IntlRelativeTimeFormat.h:
2020-04-29 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed follow-up to r260848.
LowerDFGToB3 has its own isFunction which should NOT have been renamed.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCreateThis):
(JSC::FTL::DFG::LowerDFGToB3::compileCreatePromise):
(JSC::FTL::DFG::LowerDFGToB3::compileCreateInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileIsObjectOrNull):
(JSC::FTL::DFG::LowerDFGToB3::compileIsFunction):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
(JSC::FTL::DFG::LowerDFGToB3::isFunction): Renamed from isCallable.
2020-04-29 Alexey Shvayka <shvaikalesh@gmail.com>
AsyncFromSyncIterator methods should not pass absent values
https://bugs.webkit.org/show_bug.cgi?id=211147
Reviewed by Ross Kirsling.
This patch implements minor spec change [1] to match async and sync iteration
from the perspective of userland `next` and `return` iterator methods.
`throw` method always receives an argument, yet we align with others to be
consistent and future-proof.
This change is already implemented in SpiderMonkey.
[1]: https://github.com/tc39/ecma262/pull/1776
* builtins/AsyncFromSyncIteratorPrototype.js:
2020-04-29 Mark Lam <mark.lam@apple.com>
Freezing of Gigacage and JSC Configs should be thread safe.
https://bugs.webkit.org/show_bug.cgi?id=211201
<rdar://problem/62597619>
Reviewed by Yusuke Suzuki.
If a client creates multiple VM instances in different threads concurrently, the
following race can occur:
Config::permanentlyFreeze() contains the following code:
if (!g_jscConfig.isPermanentlyFrozen) // Point P1
g_jscConfig.isPermanentlyFrozen = true; // Point P2
Let's say there are 2 threads T1 and T2.
1. T1 creates a VM and gets to point P1, and sees that g_jscConfig.isPermanentlyFrozen is not set.
T1 is about to execute P2 when it gets pre-empted.
2. T2 creates a VM and gets to point P1, and sees that g_jscConfig.isPermanentlyFrozen is not set.
T2 proceeds to point P2 and sets g_jscConfig.isPermanentlyFrozen to true.
T2 goes on to freeze the Config and makes it not writable.
3. T1 gets to run again, and proceeds to point P2.
T1 tries to set g_jscConfig.isPermanentlyFrozen to true.
But because the Config has been frozen against writes, the write to
g_jscConfig.isPermanentlyFrozen results in a crash.
This is a classic TOCTOU bug. The fix is simply to ensure that only one thread
can enter Config::permanentlyFreeze() at a time.
Ditto for Gigacage::permanentlyFreezeGigacageConfig().
* runtime/JSCConfig.cpp:
(JSC::Config::permanentlyFreeze):
2020-04-29 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSStringJoiner is missing BigInt handling
https://bugs.webkit.org/show_bug.cgi?id=211174
Reviewed by Mark Lam.
JSStringJoiner missed handling of BigInt (specifically BigInt32) and appending empty string incorrectly.
In debug build, assertion hits. We should support BigInt in JSStringJoiner.
* runtime/JSStringJoiner.h:
(JSC::JSStringJoiner::appendWithoutSideEffects):
2020-04-29 Saam Barati <sbarati@apple.com>
U_STRING_NOT_TERMINATED_WARNING ICU must be handled when using the output buffer as a C string
https://bugs.webkit.org/show_bug.cgi?id=211142
<rdar://problem/62530860>
Reviewed by Darin Adler.
* runtime/IntlDateTimeFormat.cpp:
(JSC::defaultTimeZone):
(JSC::canonicalizeTimeZoneName):
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
(JSC::IntlDateTimeFormat::format):
(JSC::IntlDateTimeFormat::formatToParts):
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::format):
(JSC::IntlNumberFormat::formatToParts):
* runtime/IntlObject.cpp:
(JSC::convertICULocaleToBCP47LanguageTag):
(JSC::canonicalizeLanguageTag):
* runtime/IntlRelativeTimeFormat.cpp:
(JSC::IntlRelativeTimeFormat::formatInternal):
(JSC::IntlRelativeTimeFormat::formatToParts):
* runtime/StringPrototype.cpp:
(JSC::toLocaleCase):
(JSC::normalize):
2020-04-28 Saam Barati <sbarati@apple.com>
Unreviewed. Fix 32-bit build.
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createFrom):
(JSC::Int32BigIntImpl::digit):
2020-04-28 Commit Queue <commit-queue@webkit.org>
Unreviewed, reverting r260876 and r260877.
https://bugs.webkit.org/show_bug.cgi?id=211165
Broke build (Requested by yusukesuzuki on #webkit).
Reverted changesets:
"Unreviewed, build fix on watchOS"
https://bugs.webkit.org/show_bug.cgi?id=210978
https://trac.webkit.org/changeset/260876
"Unreviewed, speculative build fix on watchOS part 2"
https://bugs.webkit.org/show_bug.cgi?id=210978
https://trac.webkit.org/changeset/260877
2020-04-28 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, speculative build fix on watchOS part 2
https://bugs.webkit.org/show_bug.cgi?id=210978
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createFrom):
(JSC::Int32BigIntImpl::digit):
* runtime/JSBigInt.h:
2020-04-28 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, build fix on watchOS
https://bugs.webkit.org/show_bug.cgi?id=210978
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createFrom):
(JSC::Int32BigIntImpl::digit):
* runtime/JSBigInt.h:
2020-04-28 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] BigInt constructor should accept larger integers than safe-integers
https://bugs.webkit.org/show_bug.cgi?id=210755
Reviewed by Darin Adler.
While our implementation of BigInt constructor only accepts safe integers, it should accept all integers.
This patch implements it by creating JSBigInt::createFrom(double). We port double bit processing part from
V8 as the same to the other part of JSBigInt.
* runtime/BigIntConstructor.cpp:
(JSC::callBigIntConstructor):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createFrom):
* runtime/JSBigInt.h:
* runtime/MathCommon.h:
(JSC::isInteger):
(JSC::isSafeInteger):
* runtime/NumberConstructor.cpp:
(JSC::numberConstructorFuncIsSafeInteger):
* runtime/NumberConstructor.h:
2020-04-28 Ross Kirsling <ross.kirsling@sony.com>
[JSC] Align upon the name isCallable instead of isFunction
https://bugs.webkit.org/show_bug.cgi?id=211140
Reviewed by Darin Adler.
Follow-up to r260722. Usage is now cleanly separated between isFunction / getCallData,
but the name isCallable is still clearer than isFunction so let's flip that after all.
* API/JSContextRef.cpp:
(JSGlobalContextSetUnhandledRejectionCallback):
* API/JSObjectRef.cpp:
(JSObjectIsFunction):
* dfg/DFGOperations.cpp:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCreateThis):
(JSC::FTL::DFG::LowerDFGToB3::compileCreatePromise):
(JSC::FTL::DFG::LowerDFGToB3::compileCreateInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileIsObjectOrNull):
(JSC::FTL::DFG::LowerDFGToB3::compileIsFunction):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
(JSC::FTL::DFG::LowerDFGToB3::isCallable):
(JSC::FTL::DFG::LowerDFGToB3::isFunction): Deleted.
* ftl/FTLOperations.cpp:
(JSC::FTL::operationTypeOfObjectAsTypeofType):
* jsc.cpp:
(functionSetUnhandledRejectionCallback):
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/ExceptionHelpers.cpp:
(JSC::errorDescriptionForValue):
* runtime/FunctionPrototype.cpp:
(JSC::functionProtoFuncToString):
* runtime/InternalFunction.cpp:
(JSC::getFunctionRealm):
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::isCallable const):
(JSC::JSValue::isFunction const): Deleted.
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::JSCell::isCallable):
(JSC::JSCell::isFunction): Deleted.
* runtime/JSONObject.cpp:
(JSC::Stringifier::appendStringifiedValue):
* runtime/ObjectConstructor.cpp:
(JSC::toPropertyDescriptor):
* runtime/ObjectPrototype.cpp:
(JSC::objectProtoFuncDefineGetter):
(JSC::objectProtoFuncDefineSetter):
* runtime/Operations.cpp:
(JSC::jsTypeStringForValue):
(JSC::jsIsObjectTypeOrNull):
* runtime/ProxyObject.cpp:
(JSC::ProxyObject::structureForTarget):
(JSC::ProxyObject::finishCreation):
* runtime/RuntimeType.cpp:
(JSC::runtimeTypeForValue):
* tools/JSDollarVM.cpp:
(JSC::functionCallWithStackSize):
(JSC::functionFindTypeForExpression):
(JSC::functionReturnTypeFor):
(JSC::functionHasBasicBlockExecuted):
(JSC::functionBasicBlockExecutionCount):
* wasm/WasmInstance.cpp:
(JSC::Wasm::Instance::setFunctionWrapper):
* wasm/WasmOperations.cpp:
(JSC::Wasm::operationIterateResults):
(JSC::Wasm::operationWasmRefFunc):
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::link):
* wasm/js/WebAssemblyWrapperFunction.cpp:
(JSC::WebAssemblyWrapperFunction::finishCreation):
2020-04-28 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] NumberConstructor should accept BigInt
https://bugs.webkit.org/show_bug.cgi?id=210835
Reviewed by Mark Lam.
This patch fixes our Number constructor behavior to accept BigInt. According to the spec[1],
Number constructor should accept BigInt and should generate numbers from that.
We port V8's BigInt to double conversion code as we did for the other HeapBigInt runtime functions.
And we introduce CallNumberConstructor DFG node and handle Number constructor call with BigInt correctly
in DFG and FTL. Previously we were emitting ToNumber DFG node for Number constructor. But this is wrong
now since ToNumber does not accept BigInt and throws an error, and Number constructor should not use
ToNumber to implement its implementation. So we should introduce slightly different semantics: CallNumberConstructor
as we introduced CallStringConstructor in addition to ToString DFG node. And we add appropriate BigInt32 path
to emit efficient CallNumberConstructor machine code.
[1]: https://tc39.es/ecma262/#sec-number-constructor-number-value
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGBackwardsPropagationPhase.cpp:
(JSC::DFG::BackwardsPropagationPhase::propagate):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleConstantInternalFunction):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::fixupToNumberOrToNumericOrCallNumberConstructor):
(JSC::DFG::FixupPhase::fixupToNumeric): Deleted.
(JSC::DFG::FixupPhase::fixupToNumber): Deleted.
* dfg/DFGNode.h:
(JSC::DFG::Node::hasHeapPrediction):
* dfg/DFGNodeType.h:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileToNumeric):
(JSC::DFG::SpeculativeJIT::compileCallNumberConstructor):
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCallNumberConstructor):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::decideRounding):
(JSC::JSBigInt::toNumberHeap):
* runtime/JSBigInt.h:
* runtime/NumberConstructor.cpp:
(JSC::constructNumberConstructor):
(JSC::callNumberConstructor):
2020-04-27 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Throw OutOfMemoryError instead of RangeError if BigInt is too big
https://bugs.webkit.org/show_bug.cgi?id=211111
Reviewed by Saam Barati.
Currently, we are throwing a RangeError if we detect that JSBigInt becomes too large. But this is not consistent with our JSString's policy.
We should throw OutOfMemoryError in this case. This also makes DFG simple since DFG allows throwing OutOfMemoryError in any places which node
is even removed.
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* runtime/ExceptionHelpers.cpp:
(JSC::throwOutOfMemoryError):
* runtime/ExceptionHelpers.h:
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::tryCreateWithLength):
(JSC::JSBigInt::exponentiateHeap):
(JSC::JSBigInt::leftShiftByAbsolute):
(JSC::JSBigInt::allocateFor):
2020-04-27 Saam Barati <sbarati@apple.com>
BigInt math runtime shouldn't convert BigInt32 input operands to a heap cell when doing math
https://bugs.webkit.org/show_bug.cgi?id=210978
Reviewed by Yusuke Suzuki.
This patch adds support in the runtime for doing alomst all BigInt math
operations on the inputs either being Int32, HeapBigInt, or a mixing
of both. Before, if we detected a binary operation on an Int32 and a
HeapBigInt, this would lead us to convert the Int32 operand into a HeapBigInt.
This is especially bad because we'd repeat this for all math ops. For example,
if x is a BigInt32, and all rhs are a HeapBigInt, we'd repeatedly convert x
to a HeapBigInt for each operation:
```
x + y
x * y
x - y
x >> y
x << y
etc
```
To teach the runtime how to operate both over a BigInt32 and a HeapBigInt, I
templatized the runtime math operations to work both over BigInt32 and
HeapBigInt wrapper classes that expose the same interface.
This is a ~28% speedup on microbenchmarks/sunspider-sha1-big-int.js
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compare):
* jit/JITOperations.cpp:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/JSBigInt.cpp:
(JSC::HeapBigIntImpl::HeapBigIntImpl):
(JSC::HeapBigIntImpl::isZero):
(JSC::HeapBigIntImpl::sign):
(JSC::HeapBigIntImpl::length):
(JSC::HeapBigIntImpl::digit):
(JSC::HeapBigIntImpl::toHeapBigInt):
(JSC::Int32BigIntImpl::Int32BigIntImpl):
(JSC::Int32BigIntImpl::isZero):
(JSC::Int32BigIntImpl::sign):
(JSC::Int32BigIntImpl::length):
(JSC::Int32BigIntImpl::digit):
(JSC::Int32BigIntImpl::toHeapBigInt):
(JSC::JSBigInt::ImplResult::ImplResult):
(JSC::tryConvertToBigInt32):
(JSC::JSBigInt::inplaceMultiplyAdd):
(JSC::JSBigInt::exponentiateImpl):
(JSC::JSBigInt::exponentiate):
(JSC::JSBigInt::multiplyImpl):
(JSC::JSBigInt::multiply):
(JSC::JSBigInt::divideImpl):
(JSC::JSBigInt::divide):
(JSC::JSBigInt::copy):
(JSC::JSBigInt::unaryMinusImpl):
(JSC::JSBigInt::unaryMinus):
(JSC::JSBigInt::remainderImpl):
(JSC::JSBigInt::remainder):
(JSC::JSBigInt::incImpl):
(JSC::JSBigInt::inc):
(JSC::JSBigInt::decImpl):
(JSC::JSBigInt::dec):
(JSC::JSBigInt::addImpl):
(JSC::JSBigInt::add):
(JSC::JSBigInt::subImpl):
(JSC::JSBigInt::sub):
(JSC::JSBigInt::bitwiseAndImpl):
(JSC::JSBigInt::bitwiseAnd):
(JSC::JSBigInt::bitwiseOrImpl):
(JSC::JSBigInt::bitwiseOr):
(JSC::JSBigInt::bitwiseXorImpl):
(JSC::JSBigInt::bitwiseXor):
(JSC::JSBigInt::leftShiftImpl):
(JSC::JSBigInt::leftShift):
(JSC::JSBigInt::leftShiftSlow):
(JSC::JSBigInt::signedRightShiftImpl):
(JSC::JSBigInt::signedRightShift):
(JSC::JSBigInt::bitwiseNotImpl):
(JSC::JSBigInt::bitwiseNot):
(JSC::JSBigInt::internalMultiplyAdd):
(JSC::JSBigInt::multiplyAccumulate):
(JSC::JSBigInt::absoluteCompare):
(JSC::JSBigInt::compareImpl):
(JSC::JSBigInt::compare):
(JSC::JSBigInt::absoluteAdd):
(JSC::JSBigInt::absoluteSub):
(JSC::JSBigInt::absoluteDivWithDigitDivisor):
(JSC::JSBigInt::absoluteDivWithBigIntDivisor):
(JSC::JSBigInt::absoluteLeftShiftAlwaysCopy):
(JSC::JSBigInt::absoluteBitwiseOp):
(JSC::JSBigInt::absoluteAnd):
(JSC::JSBigInt::absoluteOr):
(JSC::JSBigInt::absoluteAndNot):
(JSC::JSBigInt::absoluteXor):
(JSC::JSBigInt::absoluteAddOne):
(JSC::JSBigInt::absoluteSubOne):
(JSC::JSBigInt::leftShiftByAbsolute):
(JSC::JSBigInt::rightShiftByAbsolute):
(JSC::JSBigInt::rightShiftByMaximum):
(JSC::JSBigInt::toStringGeneric):
(JSC::JSBigInt::toShiftAmount):
(JSC::JSBigInt::exponentiateHeap): Deleted.
(JSC::JSBigInt::multiplyHeap): Deleted.
(JSC::JSBigInt::divideHeap): Deleted.
(JSC::JSBigInt::unaryMinusHeap): Deleted.
(JSC::JSBigInt::remainderHeap): Deleted.
(JSC::JSBigInt::incHeap): Deleted.
(JSC::JSBigInt::decHeap): Deleted.
(JSC::JSBigInt::addHeap): Deleted.
(JSC::JSBigInt::subHeap): Deleted.
(JSC::JSBigInt::bitwiseAndHeap): Deleted.
(JSC::JSBigInt::bitwiseOrHeap): Deleted.
(JSC::JSBigInt::bitwiseXorHeap): Deleted.
(JSC::JSBigInt::leftShiftHeap): Deleted.
(JSC::JSBigInt::signedRightShiftHeap): Deleted.
(JSC::JSBigInt::bitwiseNotHeap): Deleted.
(JSC::JSBigInt::compareToInt32): Deleted.
* runtime/JSBigInt.h:
* runtime/Operations.cpp:
(JSC::jsAddSlowCase):
* runtime/Operations.h:
(JSC::compareBigInt):
(JSC::compareBigInt32ToOtherPrimitive):
(JSC::arithmeticBinaryOp):
(JSC::jsSub):
(JSC::jsMul):
(JSC::jsDiv):
(JSC::jsRemainder):
(JSC::jsPow):
(JSC::jsInc):
(JSC::jsDec):
(JSC::jsBitwiseNot):
(JSC::shift):
(JSC::jsLShift):
(JSC::jsRShift):
(JSC::bitwiseBinaryOp):
(JSC::jsBitwiseAnd):
(JSC::jsBitwiseOr):
(JSC::jsBitwiseXor):
2020-04-27 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] >>> should call ToNumeric
https://bugs.webkit.org/show_bug.cgi?id=211065
Reviewed by Ross Kirsling.
While BigInt does not support >>> operator, >>> operator should call ToNumeric (in this case, toBigIntOrInt32) for both before throwing an error.
We call toBigIntOrInt32 for both operands, and throw an error. And after that, casting int32_t to uint32_t to perform >>> operator. This is correct
since the only difference between toUint32 and toInt32 is casting int32_t result to uint32_t.
* dfg/DFGOperations.cpp:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/Operations.h:
(JSC::shift):
(JSC::jsURShift):
2020-04-27 Keith Miller <keith_miller@apple.com>
OSR Exit compiler should know and print the exiting DFG node's index
https://bugs.webkit.org/show_bug.cgi?id=210998
Reviewed by Mark Lam.
The only interesting thing here is that we set the node to index 0 if there is no node.
AFAICT, we only don't have a node when we are checking arguments.
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::OSRExit):
(JSC::DFG::operationCompileOSRExit):
* dfg/DFGOSRExitBase.h:
(JSC::DFG::OSRExitBase::OSRExitBase):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileInvalidationPoint):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckSubClass):
(JSC::FTL::DFG::LowerDFGToB3::blessSpeculation):
* ftl/FTLOSRExit.cpp:
(JSC::FTL::OSRExitDescriptor::emitOSRExit):
(JSC::FTL::OSRExitDescriptor::emitOSRExitLater):
(JSC::FTL::OSRExitDescriptor::prepareOSRExitHandle):
(JSC::FTL::OSRExit::OSRExit):
* ftl/FTLOSRExit.h:
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
2020-04-27 Saam Barati <sbarati@apple.com>
compilePeepHoleBigInt32Branch needs to handle all conditions
https://bugs.webkit.org/show_bug.cgi?id=211096
<rdar://problem/62469971>
Reviewed by Yusuke Suzuki.
We were falling through to the generic path for all conditions which
weren't Equal/NotEqual. The generic path does not do speculation, so
it was leading to potential miscompiles because we omitted a type check.
Defining compilePeepHoleBigInt32Branch for other conditions is trivial,
so this patch just implements that.
This failure is caught by microbenchmarks/sunspider-sha1-big-int.js
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compilePeepHoleBranch):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compilePeepHoleBigInt32Branch):
2020-04-27 Jason Lawrence <lawrence.j@apple.com>
Unreviewed, reverting r260772.
This commit caused tests to start failing internally.
Reverted changeset:
"OSR Exit compiler should know and print the exiting DFG
node's index"
https://bugs.webkit.org/show_bug.cgi?id=210998
https://trac.webkit.org/changeset/260772
2020-04-27 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Add $vm.assertEnabled() to suppress Debug crash expected tests in release+assert build
https://bugs.webkit.org/show_bug.cgi?id=211089
Reviewed by Keith Miller.
Expose ASSERT_ENABLED condition to the shell to control crash expected tests.
* tools/JSDollarVM.cpp:
(JSC::functionAssertEnabled):
(JSC::JSDollarVM::finishCreation):
2020-04-27 Keith Miller <keith_miller@apple.com>
OSR Exit compiler should know and print the exiting DFG node's index
https://bugs.webkit.org/show_bug.cgi?id=210998
Reviewed by Mark Lam.
The only interesting thing here is that we set the node to index 0 if there is no node.
AFAICT, we only don't have a node when we are checking arguments.
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::OSRExit):
(JSC::DFG::operationCompileOSRExit):
* dfg/DFGOSRExitBase.h:
(JSC::DFG::OSRExitBase::OSRExitBase):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileInvalidationPoint):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckSubClass):
(JSC::FTL::DFG::LowerDFGToB3::blessSpeculation):
* ftl/FTLOSRExit.cpp:
(JSC::FTL::OSRExitDescriptor::emitOSRExit):
(JSC::FTL::OSRExitDescriptor::emitOSRExitLater):
(JSC::FTL::OSRExitDescriptor::prepareOSRExitHandle):
(JSC::FTL::OSRExit::OSRExit):
* ftl/FTLOSRExit.h:
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
2020-04-27 Ross Kirsling <ross.kirsling@sony.com>
[JSC] CallData/ConstructData should include CallType/ConstructType
https://bugs.webkit.org/show_bug.cgi?id=211059
Reviewed by Darin Adler.
getCallData/getConstructData return a CallType/ConstructType and have a CallData/ConstructData out param,
and then *both* of these are passed side-by-side to `call`/`construct`, which all seems a bit silly.
This patch merges CallType/ConstructType into CallData/ConstructData such that getCallData/getConstructData
no longer need an out param and `call`/`construct` require one less overt parameter.
In so doing, it also:
- removes ConstructData entirely as it's an exact duplicate of CallData
- renames enum value Host to Native in alignment with CallData's union
* API/JSCallbackConstructor.cpp:
(JSC::JSCallbackConstructor::getConstructData):
* API/JSCallbackConstructor.h:
* API/JSCallbackObject.h:
* API/JSCallbackObjectFunctions.h:
(JSC::JSCallbackObject<Parent>::getConstructData):
(JSC::JSCallbackObject<Parent>::getCallData):
* API/JSObjectRef.cpp:
(JSObjectCallAsFunction):
(JSObjectCallAsConstructor):
* bindings/ScriptFunctionCall.cpp:
(Deprecated::ScriptFunctionCall::call):
* bindings/ScriptFunctionCall.h:
* dfg/DFGOperations.cpp:
* inspector/InjectedScriptManager.cpp:
(Inspector::InjectedScriptManager::createInjectedScript):
* inspector/InspectorEnvironment.h:
* interpreter/Interpreter.cpp:
(JSC::Interpreter::executeProgram):
(JSC::Interpreter::executeCall):
(JSC::Interpreter::executeConstruct):
* interpreter/Interpreter.h:
* jit/JITOperations.cpp:
* jsc.cpp:
(functionDollarAgentReceiveBroadcast):
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::handleHostCall):
* runtime/ArrayPrototype.cpp:
(JSC::arrayProtoFuncToString):
(JSC::arrayProtoFuncToLocaleString):
* runtime/CallData.cpp:
(JSC::call):
(JSC::profiledCall):
* runtime/CallData.h:
* runtime/ClassInfo.h:
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/ConstructData.cpp:
(JSC::construct):
(JSC::profiledConstruct):
* runtime/ConstructData.h:
(JSC::construct):
(JSC::profiledConstruct):
(): Deleted.
* runtime/DatePrototype.cpp:
(JSC::dateProtoFuncToJSON):
* runtime/GetterSetter.cpp:
(JSC::callGetter):
(JSC::callSetter):
* runtime/InternalFunction.cpp:
(JSC::InternalFunction::getCallData):
(JSC::InternalFunction::getConstructData):
* runtime/InternalFunction.h:
* runtime/IteratorOperations.cpp:
(JSC::iteratorNext):
(JSC::iteratorClose):
(JSC::hasIteratorMethod):
(JSC::iteratorMethod):
(JSC::iteratorForIterable):
* runtime/JSBoundFunction.cpp:
(JSC::boundThisNoArgsFunctionCall):
(JSC::boundFunctionCall):
(JSC::boundThisNoArgsFunctionConstruct):
(JSC::boundFunctionConstruct):
* runtime/JSCJSValue.h:
* runtime/JSCell.cpp:
(JSC::JSCell::getCallData):
(JSC::JSCell::getConstructData):
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::JSCell::isFunction):
(JSC::JSCell::isConstructor):
* runtime/JSFunction.cpp:
(JSC::JSFunction::getCallData):
(JSC::JSFunction::getConstructData):
* runtime/JSFunction.h:
* runtime/JSInternalPromise.cpp:
(JSC::JSInternalPromise::then):
* runtime/JSMicrotask.cpp:
(JSC::JSMicrotask::run):
* runtime/JSModuleLoader.cpp:
(JSC::JSModuleLoader::dependencyKeysIfEvaluated):
(JSC::JSModuleLoader::provideFetch):
(JSC::JSModuleLoader::loadAndEvaluateModule):
(JSC::JSModuleLoader::loadModule):
(JSC::JSModuleLoader::linkAndEvaluateModule):
(JSC::JSModuleLoader::requestImportModule):
* runtime/JSONObject.cpp:
(JSC::Stringifier::isCallableReplacer const):
(JSC::Stringifier::Stringifier):
(JSC::Stringifier::toJSON):
(JSC::Stringifier::appendStringifiedValue):
(JSC::Walker::Walker):
(JSC::Walker::callReviver):
(JSC::JSONProtoFuncParse):
* runtime/JSObject.cpp:
(JSC::ordinarySetSlow):
(JSC::callToPrimitiveFunction):
(JSC::JSObject::hasInstance):
(JSC::JSObject::getMethod):
* runtime/JSObject.h:
* runtime/JSObjectInlines.h:
(JSC::getCallData):
(JSC::getConstructData):
* runtime/JSPromise.cpp:
(JSC::JSPromise::createDeferredData):
(JSC::JSPromise::resolvedPromise):
(JSC::callFunction):
* runtime/MapConstructor.cpp:
(JSC::constructMap):
* runtime/ObjectPrototype.cpp:
(JSC::objectProtoFuncToLocaleString):
* runtime/ProxyObject.cpp:
(JSC::performProxyGet):
(JSC::ProxyObject::performInternalMethodGetOwnProperty):
(JSC::ProxyObject::performHasProperty):
(JSC::ProxyObject::performPut):
(JSC::performProxyCall):
(JSC::ProxyObject::getCallData):
(JSC::performProxyConstruct):
(JSC::ProxyObject::getConstructData):
(JSC::ProxyObject::performDelete):
(JSC::ProxyObject::performPreventExtensions):
(JSC::ProxyObject::performIsExtensible):
(JSC::ProxyObject::performDefineOwnProperty):
(JSC::ProxyObject::performGetOwnPropertyNames):
(JSC::ProxyObject::performSetPrototype):
(JSC::ProxyObject::performGetPrototype):
* runtime/ProxyObject.h:
* runtime/ReflectObject.cpp:
(JSC::reflectObjectConstruct):
* runtime/SamplingProfiler.cpp:
(JSC::SamplingProfiler::processUnverifiedStackTraces):
* runtime/SetConstructor.cpp:
(JSC::constructSet):
* runtime/StringPrototype.cpp:
(JSC::replaceUsingRegExpSearch):
(JSC::operationStringProtoFuncReplaceRegExpEmptyStr):
(JSC::operationStringProtoFuncReplaceRegExpString):
(JSC::replaceUsingStringSearch):
* runtime/VM.cpp:
(JSC::VM::callPromiseRejectionCallback):
* runtime/WeakMapConstructor.cpp:
(JSC::constructWeakMap):
* runtime/WeakSetConstructor.cpp:
(JSC::constructWeakSet):
* tools/JSDollarVM.cpp:
(JSC::callWithStackSizeProbeFunction):
* wasm/js/WebAssemblyModuleRecord.cpp:
(JSC::WebAssemblyModuleRecord::evaluate):
* wasm/js/WebAssemblyWrapperFunction.cpp:
(JSC::callWebAssemblyWrapperFunction):
2020-04-26 Ross Kirsling <ross.kirsling@sony.com>
[JSC] Clearly distinguish isConstructor from getConstructData
https://bugs.webkit.org/show_bug.cgi?id=211053
Reviewed by Sam Weinig.
Follow-up to r260722. Remove the isConstructor overload that duplicates getConstructData
and clearly distinguish the usage of these two functions.
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::JSCell::isConstructor):
Remove isConstructor overload.
* runtime/JSBoundFunction.cpp:
(JSC::JSBoundFunction::create):
Don't use getConstructData if you don't need ConstructData.
* runtime/ReflectObject.cpp:
(JSC::reflectObjectConstruct):
Use getConstructData if you need ConstructData.
* API/JSObjectRef.cpp:
(JSObjectIsFunction):
Use isFunction (leftover spot from last patch).
2020-04-26 Alexey Shvayka <shvaikalesh@gmail.com>
Symbol should have [[Construct]] internal method
https://bugs.webkit.org/show_bug.cgi?id=211050
Reviewed by Yusuke Suzuki.
This change introduces constructSymbol() method, which unconditionally throws
a TypeError, since its presence is observable when, for example, Symbol is a
[[ProxyTarget]] itself [1]. Aligns JSC with the spec [2], V8, and SpiderMonkey.
[1]: https://tc39.es/ecma262/#sec-proxycreate (step 7.b)
[2]: https://tc39.es/ecma262/#constructor
* runtime/SymbolConstructor.cpp:
(JSC::SymbolConstructor::SymbolConstructor):
(JSC::constructSymbol):
2020-04-26 Alexey Shvayka <shvaikalesh@gmail.com>
InternalFunction::createSubclassStructure should use newTarget's globalObject
https://bugs.webkit.org/show_bug.cgi?id=202599
Reviewed by Yusuke Suzuki.
If "prototype" of NewTarget is not an object, built-in constructors [1] should acquire
default [[Prototype]] from realm of NewTarget, utilizing GetFunctionRealm helper [2].
Before this change, realm of active constructor was used instead. This patch introduces
GetFunctionRealm and aligns all subclassable constructors with the spec, V8, and SpiderMonkey.
This change inlines fast paths checks of InternalFunction::createSubclassStructure() and
simplifies its signature; getFunctionRealm() is invoked in slow paths only.
While a dynamically created function uses NewTarget's realm for its default [[Prototype]]
similar to other built-ins, its "prototype" object inherit from ObjectPrototype
of active constructor's realm [3] (just like their scope), making it retain references
to 2 different global objects. To accomodate this behavior, this change introduces
`scopeGlobalObject` in JSFunction.cpp methods.
Above-mentioned behavior also simplifies creation of JSGenerator and JSAsyncGenerator
instances since NewTarget's realm is irrelevant to them.
IntlCollatorConstructor::collatorStructure() and 6 similar methods are removed:
a) to impose good practice of using newTarget's globalObject;
b) with this change, each of them have 1 call site max;
c) other JSC constructors have no methods alike.
[1]: https://tc39.es/ecma262/#sec-map-constructor (step 2)
[2]: https://tc39.es/ecma262/#sec-getfunctionrealm
[3]: https://tc39.es/ecma262/#sec-createdynamicfunction (steps 23-25)
* dfg/DFGOperations.cpp:
* runtime/AggregateErrorConstructor.cpp:
(JSC::callAggregateErrorConstructor):
(JSC::constructAggregateErrorConstructor):
* runtime/AggregateErrorConstructor.h:
* runtime/AsyncFunctionConstructor.cpp:
(JSC::constructAsyncFunctionConstructor):
* runtime/AsyncGeneratorFunctionConstructor.cpp:
(JSC::constructAsyncGeneratorFunctionConstructor):
* runtime/BooleanConstructor.cpp:
(JSC::constructWithBooleanConstructor):
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
(JSC::createInternalFieldObject):
* runtime/DateConstructor.cpp:
(JSC::constructDate):
* runtime/ErrorConstructor.cpp:
(JSC::constructErrorConstructor):
* runtime/FunctionConstructor.cpp:
(JSC::constructFunctionSkippingEvalEnabledCheck):
* runtime/InternalFunction.cpp:
(JSC::InternalFunction::createSubclassStructure):
(JSC::getFunctionRealm):
(JSC::InternalFunction::createSubclassStructureSlow): Deleted.
* runtime/InternalFunction.h:
(JSC::InternalFunction::createSubclassStructure): Deleted.
* runtime/IntlCollatorConstructor.cpp:
(JSC::constructIntlCollator):
(JSC::callIntlCollator):
* runtime/IntlCollatorConstructor.h:
* runtime/IntlDateTimeFormatConstructor.cpp:
(JSC::constructIntlDateTimeFormat):
(JSC::callIntlDateTimeFormat):
* runtime/IntlDateTimeFormatConstructor.h:
* runtime/IntlNumberFormatConstructor.cpp:
(JSC::constructIntlNumberFormat):
(JSC::callIntlNumberFormat):
* runtime/IntlNumberFormatConstructor.h:
* runtime/IntlPluralRulesConstructor.cpp:
(JSC::constructIntlPluralRules):
* runtime/IntlPluralRulesConstructor.h:
* runtime/IntlRelativeTimeFormatConstructor.cpp:
(JSC::constructIntlRelativeTimeFormat):
* runtime/IntlRelativeTimeFormatConstructor.h:
* runtime/JSArrayBufferConstructor.cpp:
(JSC::JSGenericArrayBufferConstructor<sharingMode>::constructArrayBuffer):
* runtime/JSFunction.cpp:
(JSC::JSFunction::prototypeForConstruction):
(JSC::JSFunction::getOwnPropertySlot):
* runtime/JSGenericTypedArrayViewConstructorInlines.h:
(JSC::constructGenericTypedArrayView):
* runtime/JSGlobalObjectInlines.h:
(JSC::JSGlobalObject::arrayStructureForIndexingTypeDuringAllocation const):
* runtime/MapConstructor.cpp:
(JSC::constructMap):
* runtime/NativeErrorConstructor.cpp:
(JSC::NativeErrorConstructor<errorType>::constructNativeErrorConstructor):
(JSC::NativeErrorConstructor<errorType>::callNativeErrorConstructor):
* runtime/NativeErrorConstructor.h:
* runtime/NumberConstructor.cpp:
(JSC::constructNumberConstructor):
* runtime/ObjectConstructor.cpp:
(JSC::constructObjectWithNewTarget):
* runtime/RegExpConstructor.cpp:
(JSC::getRegExpStructure):
(JSC::constructRegExp):
(JSC::esSpecRegExpCreate):
* runtime/RegExpConstructor.h:
* runtime/SetConstructor.cpp:
(JSC::constructSet):
* runtime/StringConstructor.cpp:
(JSC::constructWithStringConstructor):
* runtime/WeakMapConstructor.cpp:
(JSC::constructWeakMap):
* runtime/WeakObjectRefConstructor.cpp:
(JSC::constructWeakRef):
* runtime/WeakSetConstructor.cpp:
(JSC::constructWeakSet):
* wasm/js/WebAssemblyCompileErrorConstructor.cpp:
(JSC::constructJSWebAssemblyCompileError):
* wasm/js/WebAssemblyInstanceConstructor.cpp:
(JSC::constructJSWebAssemblyInstance):
* wasm/js/WebAssemblyLinkErrorConstructor.cpp:
(JSC::constructJSWebAssemblyLinkError):
* wasm/js/WebAssemblyModuleConstructor.cpp:
(JSC::WebAssemblyModuleConstructor::createModule):
* wasm/js/WebAssemblyRuntimeErrorConstructor.cpp:
(JSC::constructJSWebAssemblyRuntimeError):
2020-04-26 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] ValueAdd, VaueSub, ValueMul, Inc, Dec should say SpecBigInt32 prediction based on ArithProfile
https://bugs.webkit.org/show_bug.cgi?id=211038
Reviewed by Filip Pizlo.
This patch adds profile feedback to ValueAdd, ValueSub, ValueMul, Inc, Dec to say SpecBigInt32 prediction.
Our HeapBigInt v.s. BigInt32 strategy is simpler than Double v.s. Int32 strategy: we always
prefer BigInt32 over HeapBigInt. This is because HeapBigInt calculation and conversion require
much higher cost than BigInt32. This tradeoff is largely different from Double v.s. Int32.
So keeping HeapBigInt is simply inefficient when we can use BigInt32.
This means that ArithProfile's feedback is also very simple. If we see HeapBigInt, this means
overflow happens. In DFG, we propagate this information to ValueAdd, ValueSub, and ValueMul nodes
and record it in DFGNodeFlags. And based on this information, we change the prediction and
speculation in prediction propagation and fixup phase.
We change exit reason from Overflow to BigInt32Overflow since Overflow is solely used for Int32 case,
and we have Int52Overflow for Int52 case. We should have BigInt32Overflow for BigInt32 to precisely
record and tell about what happens in DFG as a feedback for the next compilation.
We add BigInt32 speculation for ValueSub. Previously, we missed that in fixup phase and we always
speculate ValueSub with AnyBigIntUse or HeapBigIntUse. Now it can use BigInt32Use.
We also fix Inc / Dec's fixup phase to use BigInt path. Previously, it was always using UntypedUse since
`node->child1()->shouldSpeculateUntypedForArithmetic()` returns true for BigInt. We fix the ordering of
speculation attempts as it is done in the other places in fixup phase.
This patch offers 7.9% performance improvement in sunspider-sha1-big-int.
ToT Patched
sunspider-sha1-big-int 134.5668+-2.8695 ^ 124.6743+-0.7541 ^ definitely 1.0793x faster
* bytecode/ExitKind.cpp:
(JSC::exitKindToString):
* bytecode/ExitKind.h:
* bytecode/SpeculatedType.h:
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::makeSafe):
(JSC::DFG::ByteCodeParser::makeDivSafe):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGGraph.h:
(JSC::DFG::Graph::binaryArithShouldSpeculateBigInt32):
(JSC::DFG::Graph::unaryArithShouldSpeculateBigInt32):
* dfg/DFGNode.h:
(JSC::DFG::Node::mayHaveBigInt32Result):
(JSC::DFG::Node::mayHaveHeapBigIntResult):
(JSC::DFG::Node::mayHaveBigIntResult):
(JSC::DFG::Node::canSpeculateBigInt32):
(JSC::DFG::Node::canSpeculateInt52):
* dfg/DFGNodeFlags.cpp:
(JSC::DFG::dumpNodeFlags):
* dfg/DFGNodeFlags.h:
(JSC::DFG::nodeMayHaveHeapBigInt):
(JSC::DFG::nodeCanSpeculateBigInt32):
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileValueSub):
(JSC::DFG::SpeculativeJIT::compileValueMul):
(JSC::DFG::SpeculativeJIT::compileValueDiv):
(JSC::DFG::SpeculativeJIT::speculateHeapBigInt):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileValueAdd):
(JSC::FTL::DFG::LowerDFGToB3::compileValueSub):
(JSC::FTL::DFG::LowerDFGToB3::compileValueMul):
(JSC::FTL::DFG::LowerDFGToB3::compileValueDiv):
2020-04-25 Ross Kirsling <ross.kirsling@sony.com>
[JSC] isCallable is redundant with isFunction
https://bugs.webkit.org/show_bug.cgi?id=211037
Reviewed by Yusuke Suzuki.
isCallable is only being used in two places and has the same definition as isFunction (aside from out params).
Where CallData is needed, getCallData should be used; where CallData is not needed, isFunction should be used.
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::isCallable const): Deleted.
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::JSCell::isCallable): Deleted.
Remove isCallable.
* runtime/JSONObject.cpp:
(JSC::Stringifier::Stringifier):
(JSC::Stringifier::toJSON):
Use getCallData if you need CallData.
* runtime/ExceptionHelpers.cpp:
(JSC::errorDescriptionForValue):
* runtime/ObjectConstructor.cpp:
(JSC::toPropertyDescriptor):
* runtime/ObjectPrototype.cpp:
(JSC::objectProtoFuncDefineGetter):
(JSC::objectProtoFuncDefineSetter):
Don't use getCallData if you don't need CallData.
2020-04-25 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Handle BigInt32 INT32_MIN shift amount
https://bugs.webkit.org/show_bug.cgi?id=211030
Reviewed by Darin Adler.
Our BigInt shift-operation does not correctly handle INT32_MIN shift amount, and producing a wrong result.
This patch fixes it.
* runtime/Operations.h:
(JSC::shift):
2020-04-25 Darin Adler <darin@apple.com>
[Cocoa] Deal with another round of Xcode upgrade checks
https://bugs.webkit.org/show_bug.cgi?id=211027
Reviewed by Alexey Proskuryakov.
* JavaScriptCore.xcodeproj/project.pbxproj: Bump the upgrade check version.
Add a harmless base localization; this project contains nothing localized.
2020-04-25 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Add fast path for BigInt32 left-shift
https://bugs.webkit.org/show_bug.cgi?id=211029
Reviewed by Saam Barati.
Currently, the left-shift operation misses the fast path for BigInt32 <> BigInt32 case. This patch adds it. We also fixes
prediction-propagation for left/right shift to use existing heap prediction instead of polluting the result with SpecBigInt.
This offer 4.5% improvement in microbenchmarks/sunspider-sha1-big-int.js.
* dfg/DFGPredictionPropagationPhase.cpp:
* runtime/Operations.h:
(JSC::shift):
2020-04-25 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed fix for JSC Debug tests following r210853.
* runtime/IntlObject.cpp:
(JSC::canonicalizeLanguageTag):
(JSC::canonicalizeLocaleList):
(JSC::defaultLocale):
Deal with unchecked exception by moving tryGetUtf8 call out of canonicalizeLanguageTag; it's meant to
verify the user input from canonicalizeLocaleList and needn't change the noexcept-ness of defaultLocale.
2020-04-25 Alex Christensen <achristensen@webkit.org>
Prepare to remove automatic URL->String conversion operators
https://bugs.webkit.org/show_bug.cgi?id=211007
Reviewed by Darin Adler.
* API/JSAPIGlobalObject.mm:
(JSC::JSAPIGlobalObject::moduleLoaderResolve):
(JSC::JSAPIGlobalObject::moduleLoaderImportModule):
* API/JSScript.mm:
(validateBytecodeCachePath):
(+[JSScript scriptOfType:memoryMappedFromASCIIFile:withSourceURL:andBytecodeCache:inVirtualMachine:error:]):
* inspector/ScriptDebugServer.cpp:
(Inspector::ScriptDebugServer::sourceParsed):
* parser/Nodes.h:
(JSC::ScopeNode::sourceURL const):
* runtime/CachedTypes.cpp:
(JSC::CachedSourceProviderShape::encode):
* runtime/Error.cpp:
(JSC::addErrorInfo):
* runtime/ScriptExecutable.h:
(JSC::ScriptExecutable::sourceURL const):
2020-04-25 Ross Kirsling <ross.kirsling@sony.com>
[Intl] Locale validation/canonicalization should defer to ICU
https://bugs.webkit.org/show_bug.cgi?id=210853
Reviewed by Darin Adler.
The mappings for locale canonicalization in latest CLDR are sufficiently complex
that it really no longer makes sense not to have ICU do this work for us.
This means the UTS 35 canonicalization desired by ECMA-402 will not be fully achievable until ICU ~67,
but it's better than reaching right into CLDR and pretending that we *are* ICU.
(On this point, we thus align with V8 and diverge from SM.)
Of course, we can still add our own pre-validations / post-canonicalizations if desired.
* CMakeLists.txt:
* DerivedSources-input.xcfilelist:
* DerivedSources-output.xcfilelist:
* DerivedSources.make:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Scripts/generateIntlCanonicalizeLanguage.py: Removed.
* runtime/IntlObject.cpp:
(JSC::intlAvailableLocales):
(JSC::intlCollatorAvailableLocales):
(JSC::canonicalizeLanguageTag):
(JSC::canonicalizeLocaleList):
(JSC::defaultLocale):
(JSC::removeUnicodeLocaleExtension):
(JSC::addMissingScriptLocales): Deleted. This one was ostensibly a fix for an old ICU bug.
(JSC::privateUseLangTag): Deleted.
(JSC::preferredLanguage): Deleted.
(JSC::preferredRegion): Deleted.
(JSC::canonicalLangTag): Deleted.
* ucd/language-subtag-registry.txt: Removed.
2020-04-24 Yusuke Suzuki <ysuzuki@apple.com>
Fix internal build by using strcmp instead of using string literal comparison
https://bugs.webkit.org/show_bug.cgi?id=211011
Reviewed by Keith Miller.
Use strcmp for string literal comparison to expect that this is fully handled by compiler and converted into constant at compile time.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
2020-04-24 Mark Lam <mark.lam@apple.com>
Suppress ASan on DFG::clobberize() to work around an ASan bug.
https://bugs.webkit.org/show_bug.cgi?id=211012
<rdar://problem/62275430>
Reviewed by Yusuke Suzuki.
ASan was incorrectly thinking that we're accessing invalid stack memory when we're not.
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
2020-04-24 Alexey Shvayka <shvaikalesh@gmail.com>
Fix WASM Error classes and re-sync wpt/wasm/jsapi from upstream
https://bugs.webkit.org/show_bug.cgi?id=210980
Reviewed by Keith Miller.
assert_throws_js() harness, which is extensively used by wpt/wasm/jsapi tests,
was recently updated to assert that passed constructors subclass Error in
spec-perfect way.
With this patch, WebAssembly errors have Error as [[Prototype]] of their constructors
and define correct "name" and "message" properties on their prototypes, aligning JSC
with the spec [1], V8 and SpiderMonkey.
[1]: https://webassembly.github.io/spec/js-api/#error-objects
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
* wasm/js/WebAssemblyCompileErrorPrototype.cpp:
(JSC::WebAssemblyCompileErrorPrototype::finishCreation):
* wasm/js/WebAssemblyLinkErrorPrototype.cpp:
(JSC::WebAssemblyLinkErrorPrototype::finishCreation):
* wasm/js/WebAssemblyRuntimeErrorPrototype.cpp:
(JSC::WebAssemblyRuntimeErrorPrototype::finishCreation):
2020-04-24 Saam Barati <sbarati@apple.com>
Return BigInt32 whenever we can
https://bugs.webkit.org/show_bug.cgi?id=210922
Reviewed by Yusuke Suzuki.
This patch makes it so our runtime functions for big int math on heap
big ints converts the result to a big int 32 when possible.
The inspiration for this patch came from converting SunSpider's sha1 benchmark to
using big ints. I found that that original implementation of big int 32
was a ~35% slowdown here. This patch speeds it up by 86% from ToT, and
36% faster than before big int 32 was introduced.
To make this sound in the DFG/FTL, we are currently reporting that all
HeapBigInt math ops return SpecBigInt, instead of SpecHeapBigInt.
However, we want to do better in a follow up. We need some kind of profiling
system where we determine if we should speculate if the result is big int
32, a heap big int, or both:
https://bugs.webkit.org/show_bug.cgi?id=210982
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileValueBitNot):
(JSC::DFG::SpeculativeJIT::compileValueBitwiseOp):
(JSC::DFG::SpeculativeJIT::compileValueLShiftOp):
(JSC::DFG::SpeculativeJIT::compileValueBitRShift):
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileValueSub):
(JSC::DFG::SpeculativeJIT::compileValueMul):
(JSC::DFG::SpeculativeJIT::compileValueDiv):
(JSC::DFG::SpeculativeJIT::compileValueMod):
(JSC::DFG::SpeculativeJIT::compileValuePow):
* jit/JITOperations.cpp:
* jsc.cpp:
(functionCreateBigInt32):
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::callBigIntConstructor):
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::exponentiateHeap):
(JSC::JSBigInt::multiplyHeap):
(JSC::JSBigInt::divideHeap):
(JSC::JSBigInt::unaryMinusHeap):
(JSC::JSBigInt::remainderHeap):
(JSC::JSBigInt::incHeap):
(JSC::JSBigInt::decHeap):
(JSC::JSBigInt::addHeap):
(JSC::JSBigInt::subHeap):
(JSC::JSBigInt::bitwiseAndHeap):
(JSC::JSBigInt::bitwiseOrHeap):
(JSC::JSBigInt::bitwiseXorHeap):
(JSC::JSBigInt::leftShiftHeap):
(JSC::JSBigInt::signedRightShiftHeap):
(JSC::JSBigInt::bitwiseNotHeap):
(JSC::JSBigInt::absoluteAdd):
(JSC::JSBigInt::absoluteSub):
(JSC::JSBigInt::parseInt):
(JSC::JSBigInt::exponentiate): Deleted.
(JSC::JSBigInt::multiply): Deleted.
(JSC::JSBigInt::divide): Deleted.
(JSC::JSBigInt::unaryMinus): Deleted.
(JSC::JSBigInt::remainder): Deleted.
(JSC::JSBigInt::inc): Deleted.
(JSC::JSBigInt::dec): Deleted.
(JSC::JSBigInt::add): Deleted.
(JSC::JSBigInt::sub): Deleted.
(JSC::JSBigInt::bitwiseAnd): Deleted.
(JSC::JSBigInt::bitwiseOr): Deleted.
(JSC::JSBigInt::bitwiseXor): Deleted.
(JSC::JSBigInt::leftShift): Deleted.
(JSC::JSBigInt::signedRightShift): Deleted.
(JSC::JSBigInt::bitwiseNot): Deleted.
* runtime/JSBigInt.h:
* runtime/JSCJSValue.h:
(JSC::jsBigInt32):
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::JSValue):
* runtime/Operations.cpp:
(JSC::jsAddSlowCase):
* runtime/Operations.h:
(JSC::jsSub):
(JSC::jsMul):
(JSC::jsDiv):
(JSC::jsInc):
(JSC::jsDec):
(JSC::jsBitwiseNot):
(JSC::shift):
(JSC::bitwiseBinaryOp):
2020-04-24 Michael Catanzaro <mcatanzaro@gnome.org>
[GTK][WPE][JSCOnly] compile error when -DWTF_CPU_ARM64_CORTEXA53=ON set for arm64
https://bugs.webkit.org/show_bug.cgi?id=197192
Reviewed by Yusuke Suzuki.
This workaround is supposed to fix WebKit on old Cortex A53 CPUs, but it has been broken
since 2018, and people would like to use WebKit on modern Cortex A53. If anyone using WebKit
on the original hardware wants to fix and reimplement the workaround, feel free.
* assembler/ARM64Assembler.h:
(JSC::ARM64Assembler::adrp):
(JSC::ARM64Assembler::madd):
(JSC::ARM64Assembler::msub):
(JSC::ARM64Assembler::smaddl):
(JSC::ARM64Assembler::smsubl):
(JSC::ARM64Assembler::umaddl):
(JSC::ARM64Assembler::umsubl):
(JSC::ARM64Assembler::nopCortexA53Fix835769): Deleted.
(JSC::ARM64Assembler::nopCortexA53Fix843419): Deleted.
* offlineasm/arm64.rb:
* offlineasm/instructions.rb:
2020-04-24 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Fix DataFormatJSBigInt32 missing part
https://bugs.webkit.org/show_bug.cgi?id=210986
Reviewed by Mark Lam.
Add missing part of DataFormatJSBigInt32 implementation.
* bytecode/DataFormat.h:
(JSC::dataFormatToString):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::checkGeneratedTypeForToInt32):
2020-04-24 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, build fix in Windows
https://bugs.webkit.org/show_bug.cgi?id=210892
Windows MSVC does not have proper understanding of IGNORE_RETURN_TYPE_WARNINGS_BEGIN.
* runtime/JSBigInt.h:
(JSC::invertBigIntCompareResult):
2020-04-24 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] DFG compare should speculate BigInt well
https://bugs.webkit.org/show_bug.cgi?id=210892
Reviewed by Saam Barati.
Compare operations in DFG does not support BigInt related speculations. As a result, DFG fixup phase emits DoubleRep for operands, and
causes OSR exit. This patch adds BigInt32, HeapBigInt, and AnyBigIntUse support to DFG compare operations to avoid OSR exits.
We also introduce JSBigInt::compareToInt32 to avoid allocating JSBigInt only for comparison, and optimize C++ runtime for JSBigInt comparison.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileValueSub):
(JSC::DFG::SpeculativeJIT::compileValueMul):
(JSC::DFG::SpeculativeJIT::compare):
(JSC::DFG::SpeculativeJIT::genericJSValueNonPeepholeCompare):
(JSC::DFG::SpeculativeJIT::nonSpeculativeNonPeepholeCompare): Deleted.
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compileBigInt32Compare):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCompareEq):
(JSC::FTL::DFG::LowerDFGToB3::compileCompareStrictEq):
(JSC::FTL::DFG::LowerDFGToB3::compare):
(JSC::FTL::DFG::LowerDFGToB3::genericJSValueCompare):
(JSC::FTL::DFG::LowerDFGToB3::nonSpeculativeCompare): Deleted.
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::unboxBigInt32):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::compareToInt32):
* runtime/JSBigInt.h:
(JSC::swapBigIntCompareResult):
* runtime/Operations.h:
(JSC::compareBigInt):
(JSC::compareBigInt32ToOtherPrimitive):
(JSC::bigIntCompare):
2020-04-24 Alexey Shvayka <shvaikalesh@gmail.com>
Proxy.revocable should not have [[Construct]] slot
https://bugs.webkit.org/show_bug.cgi?id=210959
Reviewed by Darin Adler.
This change removes proxyRevocableConstructorThrowError() since its presence is
observable when, for example, Proxy.revocable is a [[ProxyTarget]] itself [1].
Also removes unnecessary newTarget() check in constructProxyObject() and
2 extra ArgList instances.
This patch aligns JSC with the spec [2], V8 and SpiderMonkey.
[1]: https://tc39.es/ecma262/#sec-proxycreate (step 7.b)
[2]: https://tc39.es/ecma262/#sec-ecmascript-standard-built-in-objects
* runtime/ProxyConstructor.cpp:
(JSC::makeRevocableProxy):
(JSC::ProxyConstructor::finishCreation):
(JSC::constructProxyObject):
(JSC::proxyRevocableConstructorThrowError): Deleted.
2020-04-24 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] DFG AI for some bitops + BigInt32 should be precise
https://bugs.webkit.org/show_bug.cgi?id=210956
Reviewed by Keith Miller.
Use SpecBigInt32 for ValueBitXor, ValueBitAnd, and ValueBitOr since they are always producing BigInt32 and they have inlined implementations in DFG / FTL.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
2020-04-23 Alexey Shvayka <shvaikalesh@gmail.com>
Remove revoked Proxy checks from ProxyCreate
https://bugs.webkit.org/show_bug.cgi?id=210862
Reviewed by Ross Kirsling.
This change removes revoked Proxy checks from ProxyCreate [1], implementing
https://github.com/tc39/ecma262/pull/1814 and aligning JSC with SpiderMonkey.
Also cleans up ProxyObject creation by using isFunction() instead of
isCallable(), which are identical.
[1]: https://tc39.es/ecma262/#sec-proxycreate (steps 2, 4)
* runtime/ProxyObject.cpp:
(JSC::ProxyObject::structureForTarget):
(JSC::ProxyObject::finishCreation):
2020-04-22 Keith Miller <keith_miller@apple.com>
Fix OSR exiting/iterator object checks in for-of bytecodes
https://bugs.webkit.org/show_bug.cgi?id=210882
Reviewed by Saam Barati.
This patch fixes some bugs in the DFGBytecodeParser where we would
set the exit origin for the SetLocal following the iterator_open/next
first call to the next bytecode. This meant that if out-of-line
Symbol.iterator or next functions returned an unexpected non-cell
we would OSR past the rest of the next bytecode rather than to the
first checkpoint.
This patch also makes sure we properly throw for non-objects returned
from either of the above functions in all tiers (and adds tests).
Finally, this patch makes a small optimization where we just ArithBitOr the
iterator's closed state (index == -1) and index is out of bounds. We can't
do a CompareBelow check because the index is effectively an int33_t.
* bytecode/BytecodeIndex.h:
(JSC::BytecodeIndex::withCheckpoint const):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::nextOpcodeIndex const):
(JSC::DFG::ByteCodeParser::nextCheckpoint const):
(JSC::DFG::ByteCodeParser::progressToNextCheckpoint):
(JSC::DFG::ByteCodeParser::handleCall):
(JSC::DFG::ByteCodeParser::handleCallVariant):
(JSC::DFG::ByteCodeParser::handleInlining):
(JSC::DFG::ByteCodeParser::handleGetById):
(JSC::DFG::ByteCodeParser::handlePutById):
(JSC::DFG::ByteCodeParser::parseGetById):
(JSC::DFG::ByteCodeParser::parseBlock):
(JSC::DFG::ByteCodeParser::handlePutByVal):
* jit/JITCall.cpp:
(JSC::JIT::emitSlow_op_iterator_open):
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::LLINT_SLOW_PATH_DECL):
(JSC::LLInt::handleIteratorNextCheckpoint):
2020-04-22 Darin Adler <darin@apple.com>
[Cocoa] Build with UChar as char16_t even in builds that use Apple's internal SDK
https://bugs.webkit.org/show_bug.cgi?id=210845
Reviewed by Anders Carlsson.
* Configurations/Base.xcconfig: Move ICU-configuring macros to Platform.h.
2020-04-22 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] branchIfBigInt32 can use BigInt32Mask and remove branchIfNumber filter
https://bugs.webkit.org/show_bug.cgi?id=210870
Reviewed by Saam Barati.
By using BigInt32Mask, we can detect BigInt32 without filtering Numbers. In this patch,
1. Remove branchIfBigInt32KnownNotNumber and branchIfNotBigInt32KnownNotNumber. And always use branchBigInt32 and branchNotBigInt32 instead.
2. Remove branchIfNumber type filtering in DFG.
3. Use BigInt32Mask based scheme in FTL.
4. Add and64(TrustedImm64, RegisterID) implementations in MacroAssembler.
5. Add TagRegistersMode version in branchIfBigInt. We use numberTagRegister to produce really efficient code[1] by avoiding large constant materialization.
[1]: From
mov %rax, %rdx
mov $0xfffe000000000012, %r11
and %r11, %rdx
cmp $0x12, %rdx
To
lea 0x12(%r14), %rdx
and %rax, %rdx
cmp $0x12, %rdx
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::and64):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::and64):
* bytecode/ArithProfile.cpp:
(JSC::ArithProfile<BitfieldType>::emitObserveResult):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::fillSpeculateBigInt32):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileToNumeric):
(JSC::FTL::DFG::LowerDFGToB3::compileCompareStrictEq):
(JSC::FTL::DFG::LowerDFGToB3::compileIsBigInt):
(JSC::FTL::DFG::LowerDFGToB3::boolify):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
(JSC::FTL::DFG::LowerDFGToB3::lowBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isNotBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isNotAnyBigInt):
(JSC::FTL::DFG::LowerDFGToB3::speculateBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::speculateAnyBigInt):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32KnownNotCell): Deleted.
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32KnownNotNumber): Deleted.
(JSC::FTL::DFG::LowerDFGToB3::isNotBigInt32KnownNotNumber): Deleted.
(JSC::FTL::DFG::LowerDFGToB3::isNotAnyBigIntKnownNotNumber): Deleted.
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitConvertValueToBoolean):
(JSC::AssemblyHelpers::branchIfValue):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::branchIfBigInt32):
(JSC::AssemblyHelpers::branchIfNotBigInt32):
(JSC::AssemblyHelpers::emitTypeOf):
(JSC::AssemblyHelpers::branchIfBigInt32KnownNotNumber): Deleted.
(JSC::AssemblyHelpers::branchIfNotBigInt32KnownNotNumber): Deleted.
2020-04-22 Saam Barati <sbarati@apple.com>
BigInt32 parsing should be precise
https://bugs.webkit.org/show_bug.cgi?id=210869
Reviewed by Robin Morisset.
Our algorithm before was conservative, and might produce a heap big int even
if the value could be an int32. This patch makes the algorithm precise on
64-bit, always producing a bigint32 if the number is indeed an int32.
* jsc.cpp:
(functionUseBigInt32):
(functionIsBigInt32):
(functionIsHeapBigInt):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::parseInt):
2020-04-22 Saam Barati <sbarati@apple.com>
Edge use kind asserts are wrong for BigInt32 on ValueBitLShift
https://bugs.webkit.org/show_bug.cgi?id=210872
Reviewed by Yusuke Suzuki, Mark Lam, and Robin Morisset.
This is already covered by the v8 tests Yusuke checked in.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::emitUntypedOrAnyBigIntBitOp):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitLShift):
(JSC::FTL::DFG::LowerDFGToB3::emitBinaryBitOpSnippet):
2020-04-22 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] JSBigInt inc operation does not produce right HeapBigInt zero
https://bugs.webkit.org/show_bug.cgi?id=210860
Reviewed by Mark Lam.
JSBigInt::inc can produce signed HeapBigInt zero, which is not meeting the invariant of JSBigInt.
This patch fixes it by checking zero status before setting `setSign(true)`.
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::inc):
* runtime/JSCJSValue.cpp:
(JSC::JSValue::dumpInContextAssumingStructure const):
2020-04-22 Devin Rousso <drousso@apple.com>
Web Inspector: Debugger: Step Over should only step through comma expressions if they are comma statements
https://bugs.webkit.org/show_bug.cgi?id=210588
Reviewed by Brian Burg.
* parser/Nodes.h:
(JSC::ExpressionNode::isStatement const): Added.
(JSC::ExpressionNode::setIsStatement): Added.
* parser/NodeConstructors.h:
(JSC::ExprStatementNode::ExprStatementNode):
(JSC::DeclarationStatement::DeclarationStatement):
(JSC::ReturnNode::ReturnNode):
(JSC::ThrowNode::ThrowNode):
* bytecompiler/NodesCodegen.cpp:
(JSC::CommaNode::emitBytecode):
Only emit `WillExecuteStatement` debug hooks inside `CommaNode` if it's the only child of a
statement parent node (e.g. `a(), b(), c()` vs `true && (a(), b(), c()) && true`).
* parser/Parser.h:
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseReturnStatement):
(JSC::Parser<LexerType>::parseThrowStatement):
(JSC::Parser<LexerType>::parseExpressionOrLabelStatement):
(JSC::Parser<LexerType>::parseExpressionStatement):
(JSC::Parser<LexerType>::parseExpression):
Only record a pause location for each sub-expression in a comma separated expression if it's
the only child of a statement (e.g. `a(), b(), c()` vs `true && (a(), b(), c()) && true`).
2020-04-22 Saam Barati <sbarati@apple.com>
ValueBitNot is wrong in FTL with AnyBigIntUse
https://bugs.webkit.org/show_bug.cgi?id=210846
Reviewed by Yusuke Suzuki.
We forgot to speculate.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitNot):
2020-04-22 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] AI results of BigInt32 Bitwise shift operation does not match to runtime results
https://bugs.webkit.org/show_bug.cgi?id=210839
Reviewed by Saam Barati.
While runtime function of bitwise ops with BigInt32 sometimes returns HeapBigInt, DFG AI is setting SpecBigInt32
as a result value. This leads to miscompilation particularly in FTL since FTL uses this information to remove
a lot of branches.
And we found that FTL BigInt32 predicate is not correctly checking state. This patch fixes it too.
Added test case found this (v8-bigint32-sar.js).
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitRShift):
(JSC::FTL::DFG::LowerDFGToB3::compileToNumeric):
(JSC::FTL::DFG::LowerDFGToB3::compileCompareStrictEq):
(JSC::FTL::DFG::LowerDFGToB3::compileIsBigInt):
(JSC::FTL::DFG::LowerDFGToB3::boolify):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
(JSC::FTL::DFG::LowerDFGToB3::lowBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32KnownNotCell):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32KnownNotNumber):
(JSC::FTL::DFG::LowerDFGToB3::isNotBigInt32KnownNotNumber):
(JSC::FTL::DFG::LowerDFGToB3::isNotAnyBigIntKnownNotNumber):
(JSC::FTL::DFG::LowerDFGToB3::isNotHeapBigIntUnknownWhetherCell):
(JSC::FTL::DFG::LowerDFGToB3::speculateBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::speculateAnyBigInt):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32): Deleted.
(JSC::FTL::DFG::LowerDFGToB3::isNotBigInt32): Deleted.
(JSC::FTL::DFG::LowerDFGToB3::isNotAnyBigInt): Deleted.
2020-04-21 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, build fix for watchOS
https://bugs.webkit.org/show_bug.cgi?id=210832
If function is not defined, static declaration should not be declared, otherwise, unused-function-error happens.
* jsc.cpp:
2020-04-21 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewd, speculative Windows build fix part 2
https://bugs.webkit.org/show_bug.cgi?id=210834
* runtime/Options.cpp:
(JSC::strncasecmp):
2020-04-21 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, fix windows build failure
https://bugs.webkit.org/show_bug.cgi?id=210834
* runtime/Options.cpp:
(JSC::strncasecmp):
2020-04-21 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] SpeculativeJIT::nonSpeculativeNonPeepholeStrictEq should expect AnyBigIntUse
https://bugs.webkit.org/show_bug.cgi?id=210832
Reviewed by Mark Lam.
SpeculativeJIT::nonSpeculativeNonPeepholeStrictEq will get AnyBigIntUse now. We should use ManualOperandSpeculation
and speculate function to perform speculation check.
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::nonSpeculativePeepholeStrictEq):
(JSC::DFG::SpeculativeJIT::nonSpeculativeNonPeepholeStrictEq):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::nonSpeculativeNonPeepholeStrictEq):
* jsc.cpp:
(functionCreateHeapBigInt):
(functionCreateBigInt32):
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::callBigIntConstructor):
* runtime/BigIntConstructor.h:
* runtime/JSBigInt.h:
2020-04-21 Yusuke Suzuki <ysuzuki@apple.com>
Canonicalize JSBigInt generated by structured-cloning by calling rightTrim
https://bugs.webkit.org/show_bug.cgi?id=210816
Reviewed by Keith Miller and Darin Adler.
* runtime/JSBigInt.h:
2020-04-21 Peng Liu <peng.liu6@apple.com>
Fix MACCATALYST build failures
https://bugs.webkit.org/show_bug.cgi?id=210815
Reviewed by Tim Horton.
* Configurations/FeatureDefines.xcconfig:
2020-04-21 Keith Miller <keith_miller@apple.com>
JSC's options should be case insensitive
https://bugs.webkit.org/show_bug.cgi?id=210834
Reviewed by Yusuke Suzuki.
* runtime/Options.cpp:
(JSC::Options::setOptionWithoutAlias):
(JSC::Options::setAliasedOption):
* runtime/OptionsList.h:
2020-04-21 Alexey Shvayka <shvaikalesh@gmail.com>
constructObjectFromPropertyDescriptor() is incorrect with partial descriptors
https://bugs.webkit.org/show_bug.cgi?id=184629
Reviewed by Ross Kirsling.
Before this change, constructObjectFromPropertyDescriptor() serialized a value-only descriptor
with nullish m_seenAttributes to {value, writable: false, enumerable: false, configurable: false}
instead of just {value}. This was observable when ordinarySetSlow() was called on a Proxy
`receiver` with "defineProperty" trap.
This patch makes constructObjectFromPropertyDescriptor() 1:1 with the spec [2], aligning JSC
with V8 and SpiderMonkey, and also cleans up its call sites from handling exceptions and
`undefined` value returns.
[1]: https://tc39.es/ecma262/#sec-ordinarysetwithowndescriptor (step 3.d.iv)
[2]: https://tc39.es/ecma262/#sec-frompropertydescriptor
* runtime/ObjectConstructor.cpp:
(JSC::objectConstructorGetOwnPropertyDescriptor):
(JSC::objectConstructorGetOwnPropertyDescriptors):
* runtime/ObjectConstructor.h:
(JSC::constructObjectFromPropertyDescriptor):
* runtime/ProxyObject.cpp:
(JSC::ProxyObject::performDefineOwnProperty):
2020-04-20 Yusuke Suzuki <ysuzuki@apple.com>
Check Structure attributes in Object.assign exhaustively
https://bugs.webkit.org/show_bug.cgi?id=210782
<rdar://problem/62065853>
Reviewed by Mark Lam.
* runtime/ObjectConstructor.cpp:
(JSC::objectConstructorAssign):
2020-04-21 Adrian Perez de Castro <aperez@igalia.com>
Non-unified build fixes late February 2020 edition
https://bugs.webkit.org/show_bug.cgi?id=210767
Unreviewed build fix.
* dfg/DFGValueRepReductionPhase.cpp: Add missing JSCJSValueInlines.h header.
* jit/JITCall.cpp: Add missing SlowPathCall.h header.
* runtime/AggregateError.cpp: Add missing JSCJSValueInlines.h, JSCellInlines.h, and
JSGlobalObjectInlines.h headers.
* runtime/AggregateErrorConstructor.cpp: Added missing JSCJSValueInlines.h, JSCellInlines.h,
and VMInlines.h headers.
* runtime/AggregateErrorPrototype.cpp: Added missing AggregateError.h, IdentifierInlines.h,
JSCJSValueInlines.h, JSCellInlines.h, JSGlobalObjectInlines.h, and VMInlines.h headers.
* runtime/Intrinsic.h: Added missing wtf/Optional.h header.
2020-04-20 Ross Kirsling <ross.kirsling@sony.com>
Classes marked final should not use protected access specifier
https://bugs.webkit.org/show_bug.cgi?id=210775
Reviewed by Daniel Bates.
* API/JSAPIValueWrapper.h:
* API/JSCallbackConstructor.h:
* API/JSCallbackObject.h:
* b3/B3ExtractValue.h:
* bytecode/UnlinkedFunctionExecutable.h:
* inspector/JSGlobalObjectConsoleClient.h:
* inspector/JSInjectedScriptHost.h:
* inspector/JSJavaScriptCallFrame.h:
* jsc.cpp:
* runtime/AggregateError.h:
* runtime/AggregateErrorPrototype.h:
* runtime/ArrayConstructor.h:
* runtime/ArrayPrototype.h:
* runtime/AsyncFunctionPrototype.h:
* runtime/AsyncGeneratorFunctionPrototype.h:
* runtime/AtomicsObject.h:
* runtime/BigIntConstructor.h:
* runtime/BigIntObject.h:
* runtime/BigIntPrototype.h:
* runtime/BooleanConstructor.h:
* runtime/BooleanPrototype.h:
* runtime/ConsoleObject.h:
* runtime/DateConstructor.h:
* runtime/DatePrototype.h:
* runtime/ErrorConstructor.h:
* runtime/ErrorPrototype.h:
* runtime/FileBasedFuzzerAgent.h:
* runtime/FunctionPrototype.h:
* runtime/FunctionRareData.h:
* runtime/GeneratorFunctionPrototype.h:
* runtime/GenericTypedArrayView.h:
* runtime/InspectorInstrumentationObject.h:
* runtime/IntlCollator.h:
* runtime/IntlCollatorConstructor.h:
* runtime/IntlCollatorPrototype.h:
* runtime/IntlDateTimeFormat.h:
* runtime/IntlDateTimeFormatConstructor.h:
* runtime/IntlDateTimeFormatPrototype.h:
* runtime/IntlNumberFormat.h:
* runtime/IntlNumberFormatConstructor.h:
* runtime/IntlNumberFormatPrototype.h:
* runtime/IntlPluralRules.h:
* runtime/IntlPluralRulesConstructor.h:
* runtime/IntlPluralRulesPrototype.h:
* runtime/IntlRelativeTimeFormatConstructor.h:
* runtime/IntlRelativeTimeFormatPrototype.h:
* runtime/JSArrayBuffer.h:
* runtime/JSArrayBufferConstructor.h:
* runtime/JSArrayBufferPrototype.h:
* runtime/JSAsyncGenerator.h:
* runtime/JSBoundFunction.h:
* runtime/JSCustomGetterSetterFunction.h:
* runtime/JSDataView.h:
* runtime/JSDataViewPrototype.h:
* runtime/JSGenerator.h:
* runtime/JSGenericTypedArrayView.h:
* runtime/JSGenericTypedArrayViewConstructor.h:
* runtime/JSGenericTypedArrayViewPrototype.h:
* runtime/JSGlobalLexicalEnvironment.h:
* runtime/JSModuleLoader.h:
* runtime/JSModuleNamespaceObject.h:
* runtime/JSNativeStdFunction.h:
* runtime/JSONObject.h:
* runtime/JSObject.h:
* runtime/JSTemplateObjectDescriptor.h:
* runtime/JSTypedArrayViewConstructor.h:
* runtime/JSTypedArrayViewPrototype.h:
* runtime/MathObject.h:
* runtime/NativeExecutable.h:
* runtime/NumberConstructor.h:
* runtime/NumberPrototype.h:
* runtime/ObjectConstructor.h:
* runtime/ObjectPrototype.h:
* runtime/PredictionFileCreatingFuzzerAgent.h:
* runtime/ReflectObject.h:
* runtime/RegExp.h:
* runtime/RegExpConstructor.h:
* runtime/RegExpObject.h:
* runtime/RegExpPrototype.h:
* runtime/StringPrototype.h:
* runtime/Structure.h:
* runtime/Symbol.h:
* runtime/SymbolConstructor.h:
* runtime/SymbolObject.h:
* runtime/SymbolPrototype.h:
* runtime/VMTraps.cpp:
* testRegExp.cpp:
* wasm/WasmBBQPlan.h:
* wasm/WasmLLIntPlan.h:
* wasm/WasmWorklist.cpp:
* wasm/js/JSWebAssembly.h:
* wasm/js/JSWebAssemblyCompileError.h:
* wasm/js/JSWebAssemblyInstance.h:
* wasm/js/JSWebAssemblyLinkError.h:
* wasm/js/JSWebAssemblyRuntimeError.h:
* wasm/js/WebAssemblyCompileErrorConstructor.h:
* wasm/js/WebAssemblyCompileErrorPrototype.h:
* wasm/js/WebAssemblyGlobalConstructor.h:
* wasm/js/WebAssemblyGlobalPrototype.h:
* wasm/js/WebAssemblyInstanceConstructor.h:
* wasm/js/WebAssemblyInstancePrototype.h:
* wasm/js/WebAssemblyLinkErrorConstructor.h:
* wasm/js/WebAssemblyLinkErrorPrototype.h:
* wasm/js/WebAssemblyMemoryConstructor.h:
* wasm/js/WebAssemblyMemoryPrototype.h:
* wasm/js/WebAssemblyModuleConstructor.h:
* wasm/js/WebAssemblyModulePrototype.h:
* wasm/js/WebAssemblyRuntimeErrorConstructor.h:
* wasm/js/WebAssemblyRuntimeErrorPrototype.h:
* wasm/js/WebAssemblyTableConstructor.h:
* wasm/js/WebAssemblyTablePrototype.h:
* wasm/js/WebAssemblyWrapperFunction.h:
2020-04-20 Peng Liu <peng.liu6@apple.com>
Fix build failures when video fullscreen and picture-in-picture is disabled
https://bugs.webkit.org/show_bug.cgi?id=210777
Reviewed by Eric Carlson.
* Configurations/FeatureDefines.xcconfig:
2020-04-20 Ross Kirsling <ross.kirsling@sony.com>
Intl classes shouldn't need an m_initialized* field
https://bugs.webkit.org/show_bug.cgi?id=210764
Reviewed by Darin Adler.
Existing Intl classes each have a field like m_initializedNumberFormat, but this is unnecessary on two levels:
1. The thing that gets initialized is a unique pointer to an ICU struct, so we can check it directly.
2. Everywhere we're checking this is redundant since we've already done the same check on the prototype side,
therefore we can just ASSERT before using said ICU struct.
While we're at it, clean up other stuff like:
- Move stuff that doesn't need to be part of the class to the CPP file (e.g. UFieldPositionIteratorDeleter).
- Merge createCollator into initializeCollator (seems like this is probably the oldest code in this space).
* runtime/IntlCollator.cpp:
(JSC::IntlCollator::initializeCollator):
(JSC::IntlCollator::compareStrings):
(JSC::IntlCollator::resolvedOptions):
(JSC::IntlCollator::createCollator): Deleted.
* runtime/IntlCollator.h:
* runtime/IntlDateTimeFormat.cpp:
(JSC::UFieldPositionIteratorDeleter::operator() const):
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
(JSC::IntlDateTimeFormat::resolvedOptions):
(JSC::IntlDateTimeFormat::format):
(JSC::partTypeString):
(JSC::IntlDateTimeFormat::formatToParts):
(JSC::IntlDateTimeFormat::UFieldPositionIteratorDeleter::operator() const): Deleted.
(JSC::IntlDateTimeFormat::partTypeString): Deleted.
* runtime/IntlDateTimeFormat.h:
* runtime/IntlNumberFormat.cpp:
(JSC::UFieldPositionIteratorDeleter::operator() const):
(JSC::IntlNumberFormatField::IntlNumberFormatField):
(JSC::IntlNumberFormat::initializeNumberFormat):
(JSC::IntlNumberFormat::format):
(JSC::IntlNumberFormat::resolvedOptions):
(JSC::partTypeString):
(JSC::IntlNumberFormat::formatToParts):
(JSC::IntlNumberFormat::UFieldPositionIteratorDeleter::operator() const): Deleted.
(JSC::IntlNumberFormat::partTypeString): Deleted.
* runtime/IntlNumberFormat.h:
* runtime/IntlPluralRules.cpp:
(JSC::localeData):
(JSC::IntlPluralRules::initializePluralRules):
(JSC::IntlPluralRules::resolvedOptions):
(JSC::IntlPluralRules::select):
(JSC::IntlPRInternal::localeData): Deleted.
* runtime/IntlPluralRules.h:
2020-04-20 Keith Miller <keith_miller@apple.com>
FTL doesn't observe the use kind of CheckIsConstant's child1
https://bugs.webkit.org/show_bug.cgi?id=210763
Reviewed by Yusuke Suzuki.
Somehow, this didn't get added when I changed CheckIsConstant and didn't show up
when I tested r260377 because I tested in release. Fortunately, the produced
DFG IR will be the same.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileCheckIsConstant):
2020-04-20 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Skip test262 for non-safe-integer range BigIntConstructor
https://bugs.webkit.org/show_bug.cgi?id=210749
Reviewed by Keith Miller.
* runtime/BigIntConstructor.cpp:
(JSC::callBigIntConstructor):
2020-04-20 Keith Miller <keith_miller@apple.com>
Fix CheckIsConstant for non-constant values and checking for empty
https://bugs.webkit.org/show_bug.cgi?id=210752
Reviewed by Saam Barati.
We need to make sure that we only have one speculated type if our value
is empty.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
2020-04-20 Darin Adler <darin@apple.com>
Use #import instead of #include in Objective-C and don't use #pragma once
https://bugs.webkit.org/show_bug.cgi?id=210724
Reviewed by David Kilzer.
* API/JSAPIWrapperObject.mm:
* API/JSContext.h:
* API/JSContext.mm:
* API/JSScriptInternal.h:
* API/JSValue.mm:
* API/JSVirtualMachine.mm:
* API/JSVirtualMachinePrivate.h:
* API/JSWrapperMap.mm:
* API/ObjCCallbackFunction.mm:
* API/tests/CurrentThisInsideBlockGetterTest.mm:
More #import, less #pragma once.
2020-04-20 Yusuke Suzuki <ysuzuki@apple.com>
StructuredClone algorithm should be aware of BigInt
https://bugs.webkit.org/show_bug.cgi?id=210728
Reviewed by Mark Lam.
* CMakeLists.txt:
* runtime/BigIntObject.h:
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::digit): Deleted.
(JSC::JSBigInt::setDigit): Deleted.
* runtime/JSBigInt.h:
(JSC::JSBigInt::digit):
(JSC::JSBigInt::setDigit):
2020-04-19 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] Intl.RelativeTimeFormat missing in WebKit
https://bugs.webkit.org/show_bug.cgi?id=209770
Reviewed by Darin Adler.
This patch implements the recent ECMA-402 feature Intl.RelativeTimeFormat.
RelativeTimeFormat has format / formatToParts functions like NumberFormat / DateTimeFormat
and is used to turn a number and unit into a formatted relative time string, e.g.:
new Intl.RelativeTimeFormat('en').format(10, 'day')
> 'in 10 days'
new Intl.RelativeTimeFormat('en', { numeric: 'auto' }).format(0, 'day')
> 'today'
Implementation of RelativeTimeFormat#formatToParts makes direct use of NumberFormat#formatToParts,
as the relative time string consists of at most one formatted number with optional literal text on either side.
This feature is runtime-guarded by the `useIntlRelativeTimeFormat` option.
* CMakeLists.txt:
* DerivedSources-input.xcfilelist:
* DerivedSources-output.xcfilelist:
* DerivedSources.make:
* JavaScriptCore.xcodeproj/project.pbxproj:
* Sources.txt:
* runtime/CommonIdentifiers.h:
* runtime/IntlRelativeTimeFormat.cpp: Added.
* runtime/IntlRelativeTimeFormat.h: Added.
* runtime/IntlRelativeTimeFormatConstructor.cpp: Added.
* runtime/IntlRelativeTimeFormatConstructor.h: Added.
* runtime/IntlRelativeTimeFormatPrototype.cpp: Added.
* runtime/IntlRelativeTimeFormatPrototype.h: Added.
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
(JSC::JSGlobalObject::visitChildren):
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::relativeTimeFormatStructure):
* runtime/OptionsList.h:
* runtime/VM.cpp:
(JSC::VM::VM):
* runtime/VM.h:
Add feature and runtime option.
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::formatToParts):
* runtime/IntlPluralRules.cpp:
(JSC::IntlPluralRules::initializePluralRules):
(JSC::IntlPluralRules::resolvedOptions):
Make "type" a property name.
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::initializeNumberFormat):
(JSC::IntlNumberFormat::resolvedOptions):
(JSC::IntlNumberFormat::formatToPartsInternal):
(JSC::IntlNumberFormat::formatToParts):
* runtime/IntlNumberFormat.h:
Factor out formatToPartsInternal so that RelativeTimeFormat can use it with its own UNumberFormat.
(This logic is too complicated to duplicate; it's because ICU won't split, e.g., "10,000" into parts for us.)
* runtime/IntlObject.cpp:
(JSC::IntlObject::IntlObject):
(JSC::IntlObject::create):
(JSC::IntlObject::finishCreation):
(JSC::intlAvailableLocales):
(JSC::intlCollatorAvailableLocales):
(JSC::isUnicodeLocaleIdentifierType):
(JSC::supportedLocales):
(JSC::intlDateTimeFormatAvailableLocales): Deleted.
(JSC::intlNumberFormatAvailableLocales): Deleted.
* runtime/IntlObject.h:
(JSC::intlDateTimeFormatAvailableLocales):
(JSC::intlNumberFormatAvailableLocales):
(JSC::intlPluralRulesAvailableLocales):
(JSC::intlRelativeTimeFormatAvailableLocales):
Perform three corrections for Intl classes:
1. Collator should be the only class with unique "available locales".
[unum|udat]_getAvailable exist but they've deferred to uloc_getAvailable for 20 years.
2. isUnicodeLocaleIdentifierType isn't just `alphanum{3,8}` but rather `alphanum{3,8} (sep alphanum{3,8})*`.
This is my own mistake from r239941.
3. supportedLocalesOf entries should not be frozen.
Changed in https://github.com/tc39/ecma402/pull/278.
* tools/JSDollarVM.cpp:
(JSC::functionICUVersion):
(JSC::JSDollarVM::finishCreation):
Add $vm.icuVersion so that we can add per-line skips to stress tests.
2020-04-19 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] SlowPathCall is not supported by callOperation in Windows
https://bugs.webkit.org/show_bug.cgi?id=210727
Reviewed by Ross Kirsling.
In Windows, SlowPathCall should be handled by JITSlowPathCall, otherwise, stack is not correctly allocated.
* jit/JITCall.cpp:
(JSC::JIT::emit_op_iterator_open):
(JSC::JIT::emit_op_iterator_next):
* jit/SlowPathCall.h:
(JSC::JITSlowPathCall::call):
2020-04-19 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Enable BigInt
https://bugs.webkit.org/show_bug.cgi?id=210726
Reviewed by Mark Lam.
* runtime/OptionsList.h:
2020-04-19 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] LLInt slow path call should not have third argument
https://bugs.webkit.org/show_bug.cgi?id=210721
Reviewed by Mark Lam.
LLInt callSlowPath does not work with third argument in Windows, CLoop etc. LLInt slow-path should not take third argument,
instead, use `bytecode.metadata(...)` to get metadata.
* jit/JITCall.cpp:
(JSC::JIT::emit_op_iterator_open):
(JSC::JIT::emit_op_iterator_next):
* llint/LowLevelInterpreter64.asm:
* runtime/CommonSlowPaths.cpp:
(JSC::iterator_open_try_fast):
(JSC::SLOW_PATH_DECL):
(JSC::iterator_next_try_fast):
(JSC::iterator_open_try_fast_narrow): Deleted.
(JSC::iterator_open_try_fast_wide16): Deleted.
(JSC::iterator_open_try_fast_wide32): Deleted.
(JSC::iterator_next_try_fast_narrow): Deleted.
(JSC::iterator_next_try_fast_wide16): Deleted.
(JSC::iterator_next_try_fast_wide32): Deleted.
* runtime/CommonSlowPaths.h:
2020-04-19 Mark Lam <mark.lam@apple.com>
Fix missing exception checks and handling in JSC APIs.
https://bugs.webkit.org/show_bug.cgi?id=210715
<rdar://problem/61599658>
Reviewed by Saam Barati.
* API/APICallbackFunction.h:
(JSC::APICallbackFunction::call):
- We should return early if an exception was thrown. We should not be using the
result in any way since we cannot rely on it having any sane value.
(JSC::APICallbackFunction::construct):
- For consistency, also return an undefined here when an exception was thrown.
* API/JSCallbackObjectFunctions.h:
(JSC::JSCallbackObject<Parent>::construct):
(JSC::JSCallbackObject<Parent>::call):
- Return an undefined if an exception was thrown. Don't return the potentially
garbage result value. Who knows what the client code will do with it. Returning
an undefined here makes the code more robust.
* API/JSObjectRef.cpp:
(JSObjectGetProperty):
(JSObjectHasPropertyForKey):
(JSObjectGetPropertyForKey):
(JSObjectDeletePropertyForKey):
(JSObjectGetPropertyAtIndex):
(JSObjectDeleteProperty):
- Explicitly return a nullptr if an exception was thrown. The toRef() on the
result that follows the exception check may or may not return a nullptr
(also see toRef(JSC::VM& vm, JSC::JSValue v) for !CPU(ADDRESS64)).
* API/JSValueRef.cpp:
(JSValueIsEqual):
(JSValueIsInstanceOfConstructor):
- For consistency, make these return false if an exception is thrown.
* API/ObjCCallbackFunction.mm:
(JSC::objCCallbackFunctionCallAsFunction):
(JSC::objCCallbackFunctionCallAsConstructor):
(JSC::ObjCCallbackFunctionImpl::call):
- Add some assertions and return early if an exception was thrown.
2020-04-18 Keith Miller <keith_miller@apple.com>
Fix CLoop build for iterator opcodes
https://bugs.webkit.org/show_bug.cgi?id=210709
Reviewed by Robin Morisset.
We need to add a default paramater for the metadata pointer
in the CLoop build. Additionally, the helper declarations need
to be in the various slow path header files. Lastly we need
opcode labels for our new JS call return points.
* bytecode/BytecodeList.rb:
* llint/LLIntSlowPaths.cpp:
* llint/LLIntSlowPaths.h:
* runtime/CommonSlowPaths.h:
2020-04-18 Robin Morisset <rmorisset@apple.com>
Support an inlined representation in JSValue of small BigInts ("BigInt32")
https://bugs.webkit.org/show_bug.cgi?id=206182
Reviewed by Yusuke Suzuki.
This patch attempts to optimize the performance of BigInts, when they are small (32 bit or less).
It works by inlining them into JSValue on 64-bit platforms, avoiding the allocation of a JSBigInt.
The bit pattern we use is 0000:XXXX:XXXX:0012
This representation works because of the following things:
- It cannot be confused with a Double or Integer thanks to the top bits
- It cannot be confused with a pointer to a Cell, thanks to bit 1 which is set to true
- It cannot be confused with a pointer to wasm thanks to bit 0 which is set to false
- It cannot be confused with true/false because bit 2 is set to false
- It cannot be confused for null/undefined because bit 4 is set to true
This entire change is gated by USE(BIGINT32), to make it easier to disable if it turns out to have bugs.
It should also make it much easier to verify if a given bug comes from it or from something else.
Note that in this patch we create BigInt32s when parsing small BigInt constants, and most operations (e.g. Add or BitOr) produce a BigInt32 if both of their operands are BigInt32,
but we don't produce a BigInt32 from for example the substraction/division of two large heap-allocated JSBigInts, even if the result fits in 32-bits.
As a result, small BigInts can now either be heap-allocated or inlined in the JSValue.
This patch includes a significant refactor of various slow paths, which are now grouped together in Operations.h
Because this increased the size of Operations.h significantly, I split the parts of Operations.h which are only used by the GC into Scribble.h, to avoid bloating compile times.
In the DFG and FTL we now have 3 UseKinds for BigInts: HeapBigIntUse, BigInt32Use and AnyBigIntUse.
The latter is useful when we know that we are receiving BigInts, but speculation indicates a mix of heap-allocated and small (inlined) big-ints.
Unfortunately, a naive implementation of this patch significantly regresses the performance of StrictEq (and its variants), as it is no longer true that a cell and a non-cell cannot be equal.
Before this patch, the code was jumping to a slow path if either:
- at least one operand is a double
- or both operands are cells
Now, it also needs to jump to the slow path if at least one is a cell.
To recover this performance cost, I significantly rewrote this code, from
if (left is Cell && right is Cell) {
if (left == right)
return true;
goto slowPath;
}
if (! left is Int32) {
if (left is Number)
goto slowPath
}
if (! right is Int32) {
if (right is Number)
goto slowPath
}
return left == right
To the following:
if (left is Double || right is Double)
goto slowPath
if (left == right)
return true;
if (left is Cell || right is Cell)
goto slowPath
return false;
I believe this to be faster than just replacing (left is Cell && right is Cell) by an ||, because I found a bit-trick to check (left is Double || right is Double) which should help reduce the pressure on the branch predictor.
Early JetStream2 tests appear to confirm that this patch is roughly neutral while it was a 0.5% regression before I used this trick, but the numbers are still too noisy, I plan to do more measurements before landing this patch.
I don't yet have performance numbers for this patch on a BigInt benchmark, I will get such numbers before trying to land it, but I'd like some review in the meantime.
* JavaScriptCore.xcodeproj/project.pbxproj:
* assembler/X86Assembler.h:
(JSC::X86Assembler::X86InstructionFormatter::SingleInstructionBufferWriter::memoryModRM):
* bytecode/ArithProfile.cpp:
(JSC::ArithProfile<BitfieldType>::emitObserveResult):
(JSC::ArithProfile<BitfieldType>::shouldEmitSetBigInt32 const):
(JSC::ArithProfile<BitfieldType>::shouldEmitSetHeapBigInt const):
(JSC::ArithProfile<BitfieldType>::emitSetHeapBigInt const):
(JSC::ArithProfile<BitfieldType>::emitSetBigInt32 const):
(WTF::printInternal):
* bytecode/ArithProfile.h:
(JSC::ObservedResults::didObserveNonInt32):
(JSC::ObservedResults::didObserveBigInt):
(JSC::ObservedResults::didObserveHeapBigInt):
(JSC::ObservedResults::didObserveBigInt32):
(JSC::ArithProfile::didObserveHeapBigInt const):
(JSC::ArithProfile::didObserveBigInt32 const):
(JSC::ArithProfile::setObservedHeapBigInt):
(JSC::ArithProfile::setObservedBigInt32):
(JSC::ArithProfile::observeResult):
* bytecode/BytecodeList.rb:
* bytecode/BytecodeLivenessAnalysisInlines.h:
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecode/CodeBlock.cpp:
* bytecode/DataFormat.h:
* bytecode/MethodOfGettingAValueProfile.cpp:
(JSC::MethodOfGettingAValueProfile::emitReportValue const):
* bytecode/MethodOfGettingAValueProfile.h:
* bytecode/SpeculatedType.cpp:
(JSC::dumpSpeculation):
(JSC::speculationFromClassInfo):
(JSC::speculationFromStructure):
(JSC::speculationFromValue):
(JSC::speculationFromJSType):
(JSC::leastUpperBoundOfStrictlyEquivalentSpeculations):
* bytecode/SpeculatedType.h:
(JSC::isBigInt32Speculation):
(JSC::isHeapBigIntSpeculation):
(JSC::isBigIntSpeculation):
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitEqualityOpImpl):
(JSC::BytecodeGenerator::addBigIntConstant):
* bytecompiler/BytecodeGenerator.h:
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::isToThisAnIdentity):
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::fixupToThis):
(JSC::DFG::FixupPhase::fixupToNumeric):
(JSC::DFG::FixupPhase::observeUseKindOnNode):
(JSC::DFG::FixupPhase::fixupCompareStrictEqAndSameValue):
* dfg/DFGMayExit.cpp:
* dfg/DFGNode.h:
(JSC::DFG::Node::shouldSpeculateBigInt32):
(JSC::DFG::Node::shouldSpeculateHeapBigInt):
* dfg/DFGNodeType.h:
* dfg/DFGOSRExit.cpp:
(JSC::DFG::OSRExit::compileExit):
* dfg/DFGOSRExit.h:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::SafeToExecuteEdge::operator()):
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compilePeepHoleBranch):
(JSC::DFG::SpeculativeJIT::compileValueBitNot):
(JSC::DFG::SpeculativeJIT::emitUntypedOrAnyBigIntBitOp):
(JSC::DFG::SpeculativeJIT::compileValueBitwiseOp):
(JSC::DFG::SpeculativeJIT::emitUntypedOrBigIntRightShiftBitOp):
(JSC::DFG::SpeculativeJIT::compileValueLShiftOp):
(JSC::DFG::SpeculativeJIT::compileValueBitRShift):
(JSC::DFG::SpeculativeJIT::compileShiftOp):
(JSC::DFG::SpeculativeJIT::compileValueAdd):
(JSC::DFG::SpeculativeJIT::compileValueSub):
(JSC::DFG::SpeculativeJIT::compileIncOrDec):
(JSC::DFG::SpeculativeJIT::compileValueNegate):
(JSC::DFG::SpeculativeJIT::compileValueMul):
(JSC::DFG::SpeculativeJIT::compileValueDiv):
(JSC::DFG::SpeculativeJIT::compileValueMod):
(JSC::DFG::SpeculativeJIT::compileValuePow):
(JSC::DFG::SpeculativeJIT::compare):
(JSC::DFG::SpeculativeJIT::compileStrictEq):
(JSC::DFG::SpeculativeJIT::speculateHeapBigInt):
(JSC::DFG::SpeculativeJIT::speculate):
(JSC::DFG::SpeculativeJIT::compileToNumeric):
(JSC::DFG::SpeculativeJIT::compileHeapBigIntEquality):
* dfg/DFGSpeculativeJIT.h:
(JSC::DFG::SpeculateBigInt32Operand::SpeculateBigInt32Operand):
(JSC::DFG::SpeculateBigInt32Operand::~SpeculateBigInt32Operand):
(JSC::DFG::SpeculateBigInt32Operand::edge const):
(JSC::DFG::SpeculateBigInt32Operand::node const):
(JSC::DFG::SpeculateBigInt32Operand::gpr):
(JSC::DFG::SpeculateBigInt32Operand::use):
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::fillJSValue):
(JSC::DFG::SpeculativeJIT::nonSpeculativePeepholeStrictEq):
(JSC::DFG::SpeculativeJIT::nonSpeculativeNonPeepholeStrictEq):
(JSC::DFG::SpeculativeJIT::fillSpeculateInt32Internal):
(JSC::DFG::SpeculativeJIT::fillSpeculateCell):
(JSC::DFG::SpeculativeJIT::fillSpeculateBoolean):
(JSC::DFG::SpeculativeJIT::speculateBigInt32):
(JSC::DFG::SpeculativeJIT::speculateAnyBigInt):
(JSC::DFG::SpeculativeJIT::fillSpeculateBigInt32):
(JSC::DFG::SpeculativeJIT::compileBigInt32Compare):
(JSC::DFG::SpeculativeJIT::compilePeepHoleBigInt32Branch):
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGStrengthReductionPhase.cpp:
(JSC::DFG::StrengthReductionPhase::handleNode):
* dfg/DFGUseKind.cpp:
(WTF::printInternal):
* dfg/DFGUseKind.h:
(JSC::DFG::typeFilterFor):
(JSC::DFG::isCell):
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLCommonValues.cpp:
(JSC::FTL::CommonValues::initializeConstants):
* ftl/FTLCommonValues.h:
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileValueAdd):
(JSC::FTL::DFG::LowerDFGToB3::compileValueSub):
(JSC::FTL::DFG::LowerDFGToB3::compileValueMul):
(JSC::FTL::DFG::LowerDFGToB3::compileBinaryMathIC):
(JSC::FTL::DFG::LowerDFGToB3::compileValueDiv):
(JSC::FTL::DFG::LowerDFGToB3::compileValueMod):
(JSC::FTL::DFG::LowerDFGToB3::compileValuePow):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitNot):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitAnd):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitOr):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitXor):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitRShift):
(JSC::FTL::DFG::LowerDFGToB3::compileArithBitRShift):
(JSC::FTL::DFG::LowerDFGToB3::compileArithBitLShift):
(JSC::FTL::DFG::LowerDFGToB3::compileValueBitLShift):
(JSC::FTL::DFG::LowerDFGToB3::compileBitURShift):
(JSC::FTL::DFG::LowerDFGToB3::compileToNumeric):
(JSC::FTL::DFG::LowerDFGToB3::compileCompareEq):
(JSC::FTL::DFG::LowerDFGToB3::compileCompareStrictEq):
(JSC::FTL::DFG::LowerDFGToB3::compileIsBigInt):
(JSC::FTL::DFG::LowerDFGToB3::emitBinarySnippet):
(JSC::FTL::DFG::LowerDFGToB3::emitBinaryBitOpSnippet):
(JSC::FTL::DFG::LowerDFGToB3::boolify):
(JSC::FTL::DFG::LowerDFGToB3::buildTypeOf):
(JSC::FTL::DFG::LowerDFGToB3::lowHeapBigInt):
(JSC::FTL::DFG::LowerDFGToB3::lowBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isNotBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::unboxBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::boxBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::isNotAnyBigInt):
(JSC::FTL::DFG::LowerDFGToB3::speculate):
(JSC::FTL::DFG::LowerDFGToB3::isNotHeapBigIntUnknownWhetherCell):
(JSC::FTL::DFG::LowerDFGToB3::isNotHeapBigInt):
(JSC::FTL::DFG::LowerDFGToB3::isHeapBigInt):
(JSC::FTL::DFG::LowerDFGToB3::speculateHeapBigInt):
(JSC::FTL::DFG::LowerDFGToB3::speculateHeapBigIntUnknownWhetherCell):
(JSC::FTL::DFG::LowerDFGToB3::speculateBigInt32):
(JSC::FTL::DFG::LowerDFGToB3::speculateAnyBigInt):
* ftl/FTLOSRExitCompiler.cpp:
(JSC::FTL::compileStub):
* heap/HeapSnapshotBuilder.cpp:
(JSC::HeapSnapshotBuilder::json):
* heap/MarkedBlockInlines.h:
* heap/PreciseAllocation.cpp:
* inspector/agents/InspectorHeapAgent.cpp:
(Inspector::InspectorHeapAgent::getPreview):
* interpreter/Interpreter.cpp:
(JSC::sizeOfVarargs):
* jit/AssemblyHelpers.cpp:
(JSC::AssemblyHelpers::emitConvertValueToBoolean):
(JSC::AssemblyHelpers::branchIfValue):
* jit/AssemblyHelpers.h:
(JSC::AssemblyHelpers::branchIfBigInt32):
(JSC::AssemblyHelpers::branchIfBigInt32KnownNotNumber):
(JSC::AssemblyHelpers::branchIfNotBigInt32KnownNotNumber):
(JSC::AssemblyHelpers::branchIfHeapBigInt):
(JSC::AssemblyHelpers::branchIfNotHeapBigInt):
(JSC::AssemblyHelpers::unboxBigInt32):
(JSC::AssemblyHelpers::boxBigInt32):
(JSC::AssemblyHelpers::emitTypeOf):
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
* jit/JIT.h:
* jit/JITArithmetic.cpp:
(JSC::JIT::emit_op_negate):
(JSC::JIT::emitSlow_op_negate):
* jit/JITOpcodes.cpp:
(JSC::JIT::emit_op_is_big_int):
(JSC::JIT::compileOpStrictEq):
(JSC::JIT::compileOpStrictEqJump):
(JSC::JIT::emit_op_to_numeric):
* jit/JITOpcodes32_64.cpp:
(JSC::JIT::emit_op_is_big_int):
(JSC::JIT::emit_op_to_numeric):
* jit/JITOperations.cpp:
* jit/JITOperations.h:
* llint/LLIntOfflineAsmConfig.h:
* llint/LowLevelInterpreter.asm:
* llint/LowLevelInterpreter64.asm:
* parser/ParserArena.cpp:
(JSC::IdentifierArena::makeBigIntDecimalIdentifier):
* runtime/ArrayPrototype.cpp:
* runtime/BigIntConstructor.cpp:
(JSC::toBigInt):
(JSC::callBigIntConstructor):
* runtime/BigIntObject.cpp:
(JSC::BigIntObject::create):
(JSC::BigIntObject::finishCreation):
* runtime/BigIntObject.h:
* runtime/BigIntPrototype.cpp:
(JSC::toThisBigIntValue):
(JSC::bigIntProtoFuncToStringImpl):
* runtime/CommonSlowPaths.cpp:
(JSC::SLOW_PATH_DECL):
(JSC::updateArithProfileForUnaryArithOp):
(JSC::updateArithProfileForBinaryArithOp):
* runtime/JSBigInt.cpp:
(JSC::JSBigInt::createStructure):
(JSC::JSBigInt::parseInt):
(JSC::JSBigInt::stringToBigInt):
(JSC::JSBigInt::inc):
(JSC::JSBigInt::dec):
(JSC::JSBigInt::bitwiseAnd):
(JSC::JSBigInt::toStringGeneric):
(JSC::JSBigInt::equalsToNumber):
(JSC::JSBigInt::equalsToInt32):
* runtime/JSBigInt.h:
(JSC::asHeapBigInt):
* runtime/JSCJSValue.cpp:
(JSC::JSValue::toNumberSlowCase const):
(JSC::JSValue::toObjectSlowCase const):
(JSC::JSValue::toThisSlowCase const):
(JSC::JSValue::synthesizePrototype const):
(JSC::JSValue::dumpInContextAssumingStructure const):
(JSC::JSValue::dumpForBacktrace const):
(JSC::JSValue::toStringSlowCase const):
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::JSValue):
(JSC::JSValue::asHeapBigInt const):
(JSC::JSValue::isBigInt const):
(JSC::JSValue::isHeapBigInt const):
(JSC::JSValue::isBigInt32 const):
(JSC::JSValue::bigInt32AsInt32 const):
(JSC::JSValue::isPrimitive const):
(JSC::JSValue::getPrimitiveNumber):
(JSC::JSValue::toNumeric const):
(JSC::JSValue::toBigIntOrInt32 const):
(JSC::JSValue::equalSlowCaseInline):
(JSC::JSValue::strictEqualForCells):
(JSC::JSValue::strictEqual):
(JSC::JSValue::pureStrictEqual):
(JSC::JSValue::pureToBoolean const):
* runtime/JSCell.cpp:
(JSC::JSCell::put):
(JSC::JSCell::putByIndex):
(JSC::JSCell::toPrimitive const):
(JSC::JSCell::getPrimitiveNumber const):
(JSC::JSCell::toNumber const):
(JSC::JSCell::toObjectSlow const):
* runtime/JSCell.h:
* runtime/JSCellInlines.h:
(JSC::JSCell::isHeapBigInt const):
(JSC::JSCell::toBoolean const):
(JSC::JSCell::pureToBoolean const):
* runtime/JSString.h:
(JSC::JSValue::toBoolean const):
* runtime/JSType.cpp:
(WTF::printInternal):
* runtime/JSType.h:
* runtime/JSTypeInfo.h:
* runtime/ObjectInitializationScope.cpp:
* runtime/Operations.cpp:
(JSC::jsAddSlowCase):
(JSC::jsIsObjectTypeOrNull):
* runtime/Operations.h:
(JSC::compareBigIntToOtherPrimitive):
(JSC::bigIntCompare):
(JSC::jsLess):
(JSC::jsLessEq):
(JSC::arithmeticBinaryOp):
(JSC::jsSub):
(JSC::jsMul):
(JSC::jsDiv):
(JSC::jsRemainder):
(JSC::jsPow):
(JSC::jsInc):
(JSC::jsDec):
(JSC::jsBitwiseNot):
(JSC::shift):
(JSC::jsLShift):
(JSC::jsRShift):
(JSC::bitwiseBinaryOp):
(JSC::jsBitwiseAnd):
(JSC::jsBitwiseOr):
(JSC::jsBitwiseXor):
* runtime/Scribble.h: Copied from Source/JavaScriptCore/runtime/BigIntObject.h.
(JSC::scribbleFreeCells):
(JSC::isScribbledValue):
(JSC::scribble):
* runtime/StructureInlines.h:
(JSC::prototypeForLookupPrimitiveImpl):
2020-04-18 Keith Miller <keith_miller@apple.com>
Unreviewed, remove commented out/dead code that didn't failed to
get removed when landing r260323.
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::LLINT_SLOW_PATH_DECL):
* runtime/CommonSlowPaths.cpp:
(JSC::iterator_next_try_fast):
2020-04-18 Keith Miller <keith_miller@apple.com>
Redesign how we do for-of iteration for JSArrays
https://bugs.webkit.org/show_bug.cgi?id=175454
Reviewed by Filip Pizlo and Saam Barati.
This patch intrinsics for-of iteration for JSArrays when they are
being iterated with the built-in Symbol.iterator. We do this by
adding two new bytecodes op_iterator_open and
op_iterator_next. These bytecodes are essentially a fused set of
existing bytecodes with a special case for our intrinsiced JSArray
case. This patch only adds support for these instructions on
64-bit.
The op_iterator_open bytecode is semantically the same as:
iterator = symbolIterator.@call(iterable);
next = iterator.next;
where iterable is the rhs of the for-of and symbolIterator is the
result of running iterable.symbolIterator;
The op_iterator_next bytecode is semantically the same as:
nextResult = next.@call(iterator);
done = nextResult.done;
value = done ? (undefined / bottom) : nextResult.value;
where nextResult is a temporary (the value VirtualRegister in the
LLInt/Baseline and a tmp in the DFG).
In order to make sure these bytecodes have the same perfomance as
the existing bytecode sequence, we need to make sure we have the
same profiling data and inline caching. Most of the existing
get_by_id code assumed a particular bytecode member name was the
same in each flavor get_by_id access. This patch adds template
specialized functions that vend the correct
Profile/VirtualRegister for the current bytecode/checkpoint. This
means we can have meaningful names for our Bytecode structs and
still use the generic functions.
In the LLInt most of the logic for calls/get_by_id had to be
factored into helper macros, so we could have bytecodes that are
some combination of those.
The trickiest part of this patch was getting the hand rolled DFG
IR to work correctly. This is because we don't have a great way to
express large chucks of DFG graph that doesn't involve manually
tracking all the DFG's invariants. Such as:
1) Flushing/Phantoming values at the end of each block.
2) Rolling forwards and backwards the BytecodeIndex when switching
blocks.
3) Remembering to GetLocal each variable at the top of every block.
4) Ensuring that the JSValue stored to the op_iterator_next.m_value
local does not cause us to OSR exit at the set local.
(4) is handled by a new function, bottomValueMatchingSpeculation,
on DFGGraph that produces a FrozenValue that is roughly the bottom
for a given speculated type. In a future patch we should make this
more complete, probably by adding a VM::bottomCellForSetLocal that
prediction propagation and AI know how treat as a true bottom
value. See: https://bugs.webkit.org/show_bug.cgi?id=210694
Lastly, this patch changes the DFG NodeType, CheckCell to be
CheckIsConstant. CheckIsConstant is equivalent to the == operator
on JSValue where it just checks the register values are the
same. In order to keep the same perf that we had for CheckCell,
CheckIsConstant supports CellUse.
* CMakeLists.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
* assembler/MacroAssemblerARM64.h:
(JSC::MacroAssemblerARM64::or8):
(JSC::MacroAssemblerARM64::store8):
* assembler/MacroAssemblerX86_64.h:
(JSC::MacroAssemblerX86_64::or8):
* bytecode/ArrayProfile.h:
(JSC::ArrayProfile::observeStructureID):
(JSC::ArrayProfile::observeStructure):
* bytecode/BytecodeList.rb:
* bytecode/BytecodeLivenessAnalysis.cpp:
(JSC::tmpLivenessForCheckpoint):
* bytecode/BytecodeOperandsForCheckpoint.h: Added.
(JSC::arrayProfileForImpl):
(JSC::hasArrayProfileFor):
(JSC::arrayProfileFor):
(JSC::valueProfileForImpl):
(JSC::hasValueProfileFor):
(JSC::valueProfileFor):
(JSC::destinationFor):
(JSC::calleeFor):
(JSC::argumentCountIncludingThisFor):
(JSC::stackOffsetInRegistersForCall):
(JSC::callLinkInfoFor):
* bytecode/BytecodeUseDef.cpp:
(JSC::computeUsesForBytecodeIndexImpl):
(JSC::computeDefsForBytecodeIndexImpl):
* bytecode/CallLinkInfo.cpp:
(JSC::CallLinkInfo::callTypeFor):
* bytecode/CallLinkStatus.cpp:
(JSC::CallLinkStatus::computeFromLLInt):
* bytecode/CodeBlock.cpp:
(JSC::CodeBlock::finishCreation):
(JSC::CodeBlock::finalizeLLIntInlineCaches):
(JSC::CodeBlock::tryGetValueProfileForBytecodeIndex):
* bytecode/CodeBlock.h:
(JSC::CodeBlock::instructionAt const):
* bytecode/CodeBlockInlines.h:
(JSC::CodeBlock::forEachValueProfile):
(JSC::CodeBlock::forEachArrayProfile):
* bytecode/GetByStatus.cpp:
(JSC::GetByStatus::computeFromLLInt):
* bytecode/Instruction.h:
(JSC::BaseInstruction::width const):
(JSC::BaseInstruction::hasCheckpoints const):
(JSC::BaseInstruction::asKnownWidth const):
(JSC::BaseInstruction::wide16 const):
(JSC::BaseInstruction::wide32 const):
* bytecode/InstructionStream.h:
* bytecode/IterationModeMetadata.h: Copied from Source/JavaScriptCore/bytecode/SuperSampler.h.
* bytecode/LLIntPrototypeLoadAdaptiveStructureWatchpoint.cpp:
(JSC::LLIntPrototypeLoadAdaptiveStructureWatchpoint::fireInternal):
(JSC::LLIntPrototypeLoadAdaptiveStructureWatchpoint::clearLLIntGetByIdCache):
* bytecode/LLIntPrototypeLoadAdaptiveStructureWatchpoint.h:
* bytecode/Opcode.h:
* bytecode/SpeculatedType.h:
(JSC::isSubtypeSpeculation):
(JSC::speculationContains):
* bytecode/SuperSampler.h:
(JSC::SuperSamplerScope::release):
* bytecompiler/BytecodeGenerator.cpp:
(JSC::BytecodeGenerator::emitGenericEnumeration):
(JSC::BytecodeGenerator::emitEnumeration):
(JSC::BytecodeGenerator::emitIsEmpty):
(JSC::BytecodeGenerator::emitIteratorOpen):
(JSC::BytecodeGenerator::emitIteratorNext):
(JSC::BytecodeGenerator::emitGetGenericIterator):
(JSC::BytecodeGenerator::emitIteratorGenericNext):
(JSC::BytecodeGenerator::emitIteratorGenericNextWithValue):
(JSC::BytecodeGenerator::emitIteratorGenericClose):
(JSC::BytecodeGenerator::emitGetAsyncIterator):
(JSC::BytecodeGenerator::emitDelegateYield):
(JSC::BytecodeGenerator::emitIteratorNextWithValue): Deleted.
(JSC::BytecodeGenerator::emitIteratorClose): Deleted.
(JSC::BytecodeGenerator::emitGetIterator): Deleted.
* bytecompiler/BytecodeGenerator.h:
* bytecompiler/NodesCodegen.cpp:
(JSC::ArrayPatternNode::bindValue const):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
(JSC::DFG::AbstractInterpreter<AbstractStateType>::forAllValues):
* dfg/DFGAtTailAbstractState.h:
(JSC::DFG::AtTailAbstractState::size const):
(JSC::DFG::AtTailAbstractState::numberOfTmps const):
(JSC::DFG::AtTailAbstractState::atIndex):
(JSC::DFG::AtTailAbstractState::tmp):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::progressToNextCheckpoint):
(JSC::DFG::ByteCodeParser::get):
(JSC::DFG::ByteCodeParser::set):
(JSC::DFG::ByteCodeParser::jsConstant):
(JSC::DFG::ByteCodeParser::weakJSConstant):
(JSC::DFG::ByteCodeParser::addCall):
(JSC::DFG::ByteCodeParser::allocateUntargetableBlock):
(JSC::DFG::ByteCodeParser::handleCall):
(JSC::DFG::ByteCodeParser::emitFunctionChecks):
(JSC::DFG::ByteCodeParser::inlineCall):
(JSC::DFG::ByteCodeParser::handleCallVariant):
(JSC::DFG::ByteCodeParser::handleVarargsInlining):
(JSC::DFG::ByteCodeParser::handleInlining):
(JSC::DFG::ByteCodeParser::handleMinMax):
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
(JSC::DFG::ByteCodeParser::handleDOMJITCall):
(JSC::DFG::ByteCodeParser::handleIntrinsicGetter):
(JSC::DFG::ByteCodeParser::handleDOMJITGetter):
(JSC::DFG::ByteCodeParser::handleModuleNamespaceLoad):
(JSC::DFG::ByteCodeParser::handleTypedArrayConstructor):
(JSC::DFG::ByteCodeParser::handleConstantInternalFunction):
(JSC::DFG::ByteCodeParser::handleGetById):
(JSC::DFG::ByteCodeParser::parseBlock):
(JSC::DFG::ByteCodeParser::InlineStackEntry::InlineStackEntry):
(JSC::DFG::ByteCodeParser::handlePutByVal):
(JSC::DFG::ByteCodeParser::handleCreateInternalFieldObject):
(JSC::DFG::ByteCodeParser::parse):
* dfg/DFGCFGSimplificationPhase.cpp:
(JSC::DFG::CFGSimplificationPhase::keepOperandAlive):
(JSC::DFG::CFGSimplificationPhase::jettisonBlock):
(JSC::DFG::CFGSimplificationPhase::mergeBlocks):
* dfg/DFGCapabilities.cpp:
(JSC::DFG::capabilityLevel):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
(JSC::DFG::FixupPhase::addStringReplacePrimordialChecks):
* dfg/DFGForAllKills.h:
(JSC::DFG::forAllKilledOperands):
* dfg/DFGGraph.cpp:
(JSC::DFG::Graph::bottomValueMatchingSpeculation):
* dfg/DFGGraph.h:
* dfg/DFGInPlaceAbstractState.cpp:
(JSC::DFG::InPlaceAbstractState::beginBasicBlock):
(JSC::DFG::InPlaceAbstractState::initialize):
(JSC::DFG::InPlaceAbstractState::endBasicBlock):
(JSC::DFG::InPlaceAbstractState::merge):
* dfg/DFGInPlaceAbstractState.h:
(JSC::DFG::InPlaceAbstractState::size const):
(JSC::DFG::InPlaceAbstractState::numberOfTmps const):
(JSC::DFG::InPlaceAbstractState::atIndex):
(JSC::DFG::InPlaceAbstractState::operand):
(JSC::DFG::InPlaceAbstractState::local):
(JSC::DFG::InPlaceAbstractState::argument):
(JSC::DFG::InPlaceAbstractState::variableAt): Deleted.
* dfg/DFGLazyJSValue.h:
(JSC::DFG::LazyJSValue::speculatedType const):
* dfg/DFGNode.h:
(JSC::DFG::Node::hasConstant):
(JSC::DFG::Node::hasCellOperand):
* dfg/DFGNodeType.h:
* dfg/DFGOSRExitCompilerCommon.cpp:
(JSC::DFG::callerReturnPC):
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileCheckIsConstant):
(JSC::DFG::SpeculativeJIT::compileCheckCell): Deleted.
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGValidate.cpp:
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckIsConstant):
(JSC::FTL::DFG::LowerDFGToB3::compileCheckCell): Deleted.
* generator/DSL.rb:
* generator/Metadata.rb:
* generator/Section.rb:
* jit/JIT.cpp:
(JSC::JIT::privateCompileMainPass):
(JSC::JIT::privateCompileSlowCases):
* jit/JIT.h:
* jit/JITCall.cpp:
(JSC::JIT::emitPutCallResult):
(JSC::JIT::compileSetupFrame):
(JSC::JIT::compileOpCall):
(JSC::JIT::emit_op_iterator_open):
(JSC::JIT::emitSlow_op_iterator_open):
(JSC::JIT::emit_op_iterator_next):
(JSC::JIT::emitSlow_op_iterator_next):
* jit/JITCall32_64.cpp:
(JSC::JIT::emit_op_iterator_open):
(JSC::JIT::emitSlow_op_iterator_open):
(JSC::JIT::emit_op_iterator_next):
(JSC::JIT::emitSlow_op_iterator_next):
* jit/JITInlines.h:
(JSC::JIT::updateTopCallFrame):
(JSC::JIT::advanceToNextCheckpoint):
(JSC::JIT::emitJumpSlowToHotForCheckpoint):
(JSC::JIT::emitValueProfilingSite):
* jit/JITOperations.cpp:
* jit/JITOperations.h:
* llint/LLIntSlowPaths.cpp:
(JSC::LLInt::setupGetByIdPrototypeCache):
(JSC::LLInt::performLLIntGetByID):
(JSC::LLInt::LLINT_SLOW_PATH_DECL):
(JSC::LLInt::genericCall):
(JSC::LLInt::handleIteratorOpenCheckpoint):
(JSC::LLInt::handleIteratorNextCheckpoint):
(JSC::LLInt::slow_path_checkpoint_osr_exit):
(JSC::LLInt::llint_dump_value):
* llint/LowLevelInterpreter.asm:
* llint/LowLevelInterpreter32_64.asm:
* llint/LowLevelInterpreter64.asm:
* offlineasm/transform.rb:
* runtime/CommonSlowPaths.cpp:
(JSC::iterator_open_try_fast):
(JSC::iterator_open_try_fast_narrow):
(JSC::iterator_open_try_fast_wide16):
(JSC::iterator_open_try_fast_wide32):
(JSC::iterator_next_try_fast):
(JSC::iterator_next_try_fast_narrow):
(JSC::iterator_next_try_fast_wide16):
(JSC::iterator_next_try_fast_wide32):
* runtime/CommonSlowPaths.h:
* runtime/Intrinsic.cpp:
(JSC::interationKindForIntrinsic):
* runtime/Intrinsic.h:
* runtime/JSArrayIterator.h:
* runtime/JSCJSValue.h:
* runtime/JSCJSValueInlines.h:
(JSC::JSValue::isCallable const):
* runtime/JSCast.h:
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::arrayProtoValuesFunctionConcurrently const):
* runtime/OptionsList.h:
* runtime/Structure.cpp:
(JSC::Structure::dumpBrief const):
2020-04-18 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Replace DFG NewPromise with NewInternalFieldObject
https://bugs.webkit.org/show_bug.cgi?id=210687
Reviewed by Saam Barati.
The feature of DFG::NewPromise can be implemented completely with DFG::NewInternalFieldObject. This reduces code duplication, and furthermore,
this offers Object Allocation Sinking support for free. This patch replaces DFG::NewPromise with DFG::NewInternalFieldObject and remove DFG::NewPromise
completely.
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::parseBlock):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGClobbersExitState.cpp:
(JSC::DFG::clobbersExitState):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGNode.h:
(JSC::DFG::Node::convertToNewInternalFieldObject):
(JSC::DFG::Node::convertToNewInternalFieldObjectWithInlineFields):
(JSC::DFG::Node::hasIsInternalPromise):
(JSC::DFG::Node::hasStructure):
(JSC::DFG::Node::convertToNewPromise): Deleted.
* dfg/DFGNodeType.h:
* dfg/DFGObjectAllocationSinkingPhase.cpp:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileNewInternalFieldObject):
(JSC::DFG::SpeculativeJIT::compileNewPromise): Deleted.
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGStoreBarrierInsertionPhase.cpp:
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileNewInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileMaterializeNewInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileNewPromise): Deleted.
* ftl/FTLOperations.cpp:
(JSC::FTL::operationPopulateObjectInOSR):
(JSC::FTL::operationMaterializeObjectInOSR):
* runtime/JSInternalPromise.cpp:
(JSC::JSInternalPromise::createWithInitialValues):
* runtime/JSInternalPromise.h:
* runtime/JSPromise.cpp:
(JSC::JSPromise::createWithInitialValues):
(JSC::JSPromise::finishCreation):
(JSC::JSPromise::status const):
(JSC::JSPromise::result const):
(JSC::JSPromise::flags const):
(JSC::JSPromise::resolve):
(JSC::JSPromise::reject):
(JSC::JSPromise::rejectAsHandled):
* runtime/JSPromise.h:
(JSC::JSPromise::initialValues):
(JSC::JSPromise::internalField const):
(JSC::JSPromise::internalField):
2020-04-18 Yusuke Suzuki <ysuzuki@apple.com>
Unreviewed, build fix for ARM64E after r260310
https://bugs.webkit.org/show_bug.cgi?id=207330
r260310 uses undefined function Instruction.cloneWithNewOperands in arm64e.rb and throws an error.
This patch calls `node.cloneWithNewOperands`.
* offlineasm/arm64e.rb:
2020-04-18 Alexey Shvayka <shvaikalesh@gmail.com>
RegExp.prototype[@@search] should use SameValue
https://bugs.webkit.org/show_bug.cgi?id=173226
Reviewed by Yusuke Suzuki.
This change exposes Object.is implementation as link-time-constant @sameValue and utilizes
it in RegExp.prototype[@@search] per spec [1], aligning JSC with V8 and SpiderMonkey.
[1]: https://tc39.es/ecma262/#sec-regexp.prototype-@@search (steps 5, 8)
* builtins/BuiltinNames.h:
* builtins/RegExpPrototype.js:
(Symbol.search):
* bytecode/LinkTimeConstant.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
* runtime/ObjectConstructor.cpp:
* runtime/ObjectConstructor.h:
2020-04-18 Angelos Oikonomopoulos <angelos@igalia.com>
Fix code origin when lowering offlineasm instructions on MIPS/ARM64E
https://bugs.webkit.org/show_bug.cgi?id=207330
Reviewed by Mark Lam.
Instruction operands are mapped to RegisterID in RegisterID.forName
and the operation is memoized. Therefore, we can't use the codeOrigin
of the operand at that point. Use the codeOrigin of the original
instruction instead.
* offlineasm/arm64e.rb:
* offlineasm/ast.rb:
* offlineasm/mips.rb:
* offlineasm/risc.rb:
2020-04-18 Angelos Oikonomopoulos <angelos@igalia.com>
REGRESSION(r260246): It broke build on MIPS32
https://bugs.webkit.org/show_bug.cgi?id=210665
Reviewed by Aakash Jain.
The mnemonic for 'store halfword' is 'sh', not 'shv'. This appears to
be a typo in a path that was never exercised.
Exposed by r260246; riscLowerMisplacedAddresses now calls into
riscAsRegisters with an 'h' suffix, which results in a 'storeh'
instruction. The storeh is then lowered to the non-existent 'shv'.
* offlineasm/mips.rb:
2020-04-17 Commit Queue <commit-queue@webkit.org>
Unreviewed, reverting r260279.
https://bugs.webkit.org/show_bug.cgi?id=210678
Throwing error would be more efficient, having a generic code
is still worth doing (Requested by yusukesuzuki on #webkit).
Reverted changeset:
"[JSC] We do not need to have exit-check for Map/Set iterator
functions"
https://bugs.webkit.org/show_bug.cgi?id=210667
https://trac.webkit.org/changeset/260279
2020-04-17 Saam Barati <sbarati@apple.com>
GetTypedArrayByteOffset is broken on arm64e
https://bugs.webkit.org/show_bug.cgi?id=210631
Reviewed by Mark Lam.
The vector of JSArrayBufferView is signed even when null on arm64e. However, we were
comparing against zero, which is wrong. This patch changes it so we do the right thing
and instead compare against whatever constant (ptr=nullptr,size=0) signs as.
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileGetTypedArrayByteOffset):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileGetTypedArrayByteOffset):
* runtime/CagedBarrierPtr.h:
(JSC::CagedBarrierPtr::rawBits const):
* runtime/JSArrayBufferView.h:
(JSC::JSArrayBufferView::nullVectorPtr):
2020-04-17 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] We do not need to have exit-check for Map/Set iterator functions
https://bugs.webkit.org/show_bug.cgi?id=210667
Reviewed by Michael Saboff.
If the intrinsic's DFG node does not support general cases, we should check exit-frequency to avoid exit-recompile loop.
However, Map/Set iterator creation functions (values, keys, entries) always require Map / Set types. And throwing an error
when this is not met. So, the current DFG nodes for these intrinsic supports all the cases except for the case throwing an
error, and error will exit anyway. So we do not need to have this exit-frequency guard here.
This path is already tested by map-iterator-check-before-fail.js / set-iterator-check-before-fail.js.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
2020-04-17 Devin Rousso <drousso@apple.com>
Rename NullishEq / NULLISHEQUAL to CoalesceEq / COALESCEEQUAL to match the spec
https://bugs.webkit.org/show_bug.cgi?id=210663
Reviewed by Ross Kirsling.
* bytecompiler/NodesCodegen.cpp:
(JSC::emitShortCircuitAssignment):
* parser/ASTBuilder.h:
(JSC::ASTBuilder::makeAssignNode):
* parser/Lexer.cpp:
(JSC::Lexer<T>::lexWithoutClearingLineTerminator):
* parser/Nodes.h:
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseAssignmentExpression):
* parser/ParserTokens.h:
2020-04-17 Devin Rousso <drousso@apple.com>
Implement Promise.any and AggregateError
https://bugs.webkit.org/show_bug.cgi?id=202566
Reviewed by Yusuke Suzuki.
`Promise.any` resolves when any of the given `promises` resolve, but only rejects if _all_
of the given `promises` reject. In order to support aggregating all of the `reason` values
for all of the rejections, a new error type `AggregateError` is introduced which has an
`get errors` that returns an aggregated array of the `reason` values.
* builtins/PromiseConstructor.js:
(all.newResolveElement):
(allSettled.newResolveRejectElements):
(any): Added.
(any.newRejectElement): Added.
* runtime/JSPromiseConstructor.cpp:
* builtins/BuiltinNames.h:
* bytecode/LinkTimeConstant.h:
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::errorStructure const):
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::initializeAggregateErrorConstructor): Added.
(JSC::JSGlobalObject::init):
(JSC::JSGlobalObject::visitChildren):
Expose `@AggregateError` for builtins.
* runtime/AggregateError.h: Added.
(JSC::AggregateError::destroy):
(JSC::AggregateError::subspaceFor):
(JSC::AggregateError::createStructure):
(JSC::AggregateError::create):
(JSC::AggregateError::errors const):
* runtime/AggregateError.cpp: Added.
(JSC::AggregateError::AggregateError):
(JSC::AggregateError::visitChildren):
(JSC::AggregateError::create):
(JSC::AggregateError::finishCreation):
* runtime/AggregateErrorPrototype.h: Added.
* runtime/AggregateErrorPrototype.cpp: Added.
(JSC::AggregateErrorPrototype::AggregateErrorPrototype):
(JSC::AggregateErrorPrototype::finishCreation):
(JSC::aggregateErrorPrototypeAccessorErrors):
* runtime/AggregateErrorConstructor.h: Added.
* runtime/AggregateErrorConstructor.cpp: Added.
(JSC::callAggregateErrorConstructor):
(JSC::constructAggregateErrorConstructor):
(JSC::AggregateErrorConstructor::AggregateErrorConstructor):
(JSC::AggregateErrorConstructor::finishCreation):
* runtime/ErrorType.h:
* runtime/ErrorType.cpp:
(JSC::errorTypeName):
* runtime/VM.h:
* runtime/VM.cpp:
(JSC::VM::VM):
Make an `IsoSubspace` for `AggregateError` as it has a different size than `ErrorInstance`.
* runtime/ErrorInstance.h:
(JSC::ErrorInstance::create):
* runtime/ErrorInstance.cpp:
(JSC::ErrorInstance::finishCreation):
* wasm/js/JSWebAssemblyCompileError.cpp:
(JSC::JSWebAssemblyCompileError::create):
* wasm/js/JSWebAssemblyLinkError.cpp:
(JSC::JSWebAssemblyLinkError::create):
* wasm/js/JSWebAssemblyRuntimeError.cpp:
(JSC::JSWebAssemblyRuntimeError::create):
Assign to `ErrorInstance` member variables inside `ErrorInstance::finishCreation` instead of
inside `ErrorInstance::create` so that subclasses don't have to do the work as well.
* runtime/Error.cpp:
(JSC::createError):
* runtime/ErrorPrototype.h:
(JSC::ErrorPrototype::createStructure):
* runtime/NativeErrorPrototype.h:
(JSC::NativeErrorPrototype::createStructure):
Drive-by: fix incorrect usage of `ErrorInstanceType` since `ErrorPrototype` does not inherit
from `ErrorInstance` (and therefore neither does `NativeErrorPrototype`).
* runtime/ArgList.h:
Add `WTF_MAKE_NONMOVABLE` to `MarkedArgumentBuffer`.
* Sources.txt:
* JavaScriptCore.xcodeproj/project.pbxproj:
2020-04-17 Ross Kirsling <ross.kirsling@sony.com>
Clean up some Intl classes following the ICU upgrade
https://bugs.webkit.org/show_bug.cgi?id=210637
Reviewed by Yusuke Suzuki.
In r259606, I removed the compile-time guards for {DateTimeFormat, NumberFormat}.prototype.formatToParts,
but I forgot to move the method setup back to the lookup table.
This patch addresses that and prunes various other unnecessary includes and forward declarations.
* runtime/IntlCollator.h:
* runtime/IntlCollatorConstructor.h:
* runtime/IntlDateTimeFormat.h:
* runtime/IntlDateTimeFormatConstructor.h:
* runtime/IntlDateTimeFormatPrototype.cpp:
(JSC::IntlDateTimeFormatPrototype::create):
(JSC::IntlDateTimeFormatPrototype::finishCreation):
* runtime/IntlDateTimeFormatPrototype.h:
* runtime/IntlNumberFormat.h:
* runtime/IntlNumberFormatConstructor.h:
* runtime/IntlNumberFormatPrototype.cpp:
(JSC::IntlNumberFormatPrototype::create):
(JSC::IntlNumberFormatPrototype::finishCreation):
* runtime/IntlNumberFormatPrototype.h:
* runtime/IntlObject.h:
* runtime/IntlPluralRules.h:
* runtime/IntlPluralRulesConstructor.h:
* runtime/IntlPluralRulesPrototype.cpp:
(JSC::IntlPluralRulesPrototype::create):
(JSC::IntlPluralRulesPrototype::finishCreation):
* runtime/IntlPluralRulesPrototype.h:
2020-04-17 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Map/Set iterator creation functions should fail with BadType etc. before executing insertChecks
https://bugs.webkit.org/show_bug.cgi?id=210649
<rdar://problem/61925452>
Reviewed by Mark Lam.
Since insertChecks adds some DFG nodes, we should determine whether this intrinsic handling is OK or not before executing insertChecks.
Otherwise, we will hit an assertion with `!didInsertChecks`.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
2020-04-17 Mark Lam <mark.lam@apple.com>
offlineasm is generating the wrong load/store for the "orh" instruction.
https://bugs.webkit.org/show_bug.cgi?id=210639
<rdar://problem/21501876>
Reviewed by Robin Morisset.
For example, on ARM64E, the "orh" instruction was generating the following:
"\tldr w17, [x1, #0]\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
"\torr w17, w17, #64\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
"\tstr w17, [x1, #0]\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
i.e. a 32-bit load, followed by a 32-bit OR, followed by a 32-bit store.
Instead, it should be generating the following:
"\tldrh w17, [x1, #0]\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
"\torr w17, w17, #64\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
"\tstrh w17, [x1, #0]\n" // JavaScriptCore/llint/LowLevelInterpreter64.asm:919
i.e. a 16-bit load, followed by a 32-bit OR, followed by a 16-bit store.
This bug also affects ARM64, ARMv7, and MIPS (basically any backend that uses
riscLowerMisplacedAddresses() from rise.rb). It does not affect x86, x86_64, and
C_LOOP (which was written based on x86).
* offlineasm/risc.rb:
2020-04-16 Ross Kirsling <ross.kirsling@sony.com>
REGRESSION(r259480): Two new failing i18n tests
https://bugs.webkit.org/show_bug.cgi?id=210605
Reviewed by Darin Adler.
* runtime/IntlDateTimeFormat.cpp:
(JSC::isUTCEquivalent):
(JSC::defaultTimeZone):
(JSC::canonicalizeTimeZoneName):
The default time zone needs to be canonicalized too.
* runtime/IntlObject.cpp:
(JSC::canonicalLangTag):
(JSC::resolveLocale):
Deal with some odd ""_s cases from my previous patch.
(Drive-by fix inspired by Darin's comments on this one.)
2020-04-16 Sergio Villar Senin <svillar@igalia.com>
Unreviewed build fix for non unified builds.
* dfg/DFGOperations.cpp: Added missing includes.
2020-04-16 Mark Lam <mark.lam@apple.com>
[Re-landing] Use more PAC diversity for JIT probe code.
https://bugs.webkit.org/show_bug.cgi?id=210252
<rdar://problem/54490367>
Reviewed by Keith Miller.
Introducing new PtrTags:
JITProbePtrTag - for the client probe function.
JITProbeTrampolinePtrTag - for calling the ctiMasmProbeTrampoline.
JITProbeExecutorPtrTag - for calling the probe executor.
Currently, this is only the Probe::executeProbe().
JITProbeStackInitializationFunctionPtrTag - for calling the optional stack
initialization function that the client probe function may set.
We'll now use these in the JIT probe mechanism instead of adopting the default
CFunctionPtrTag.
Fixed an assert in MacroAssemblerARM64.cpp which does not apply to non ARM64E
builds.
* assembler/MacroAssembler.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerARM64.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerPrinter.h:
(JSC::MacroAssembler::print):
* assembler/ProbeContext.h:
* runtime/JSCPtrTag.h:
* tools/JSDollarVM.cpp:
(JSC::callWithStackSizeProbeFunction):
* wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::emitLoopTierUpCheck):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::emitLoopTierUpCheck):
2020-04-16 Mark Lam <mark.lam@apple.com>
Rolling out r259897: Causing crashes on iOS.
https://bugs.webkit.org/show_bug.cgi?id=210252
Not reviewed.
* assembler/MacroAssembler.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerARM64.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerPrinter.h:
(JSC::MacroAssembler::print):
* assembler/ProbeContext.h:
* runtime/JSCPtrTag.h:
* tools/JSDollarVM.cpp:
(JSC::callWithStackSizeProbeFunction):
* wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::emitLoopTierUpCheck):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::emitLoopTierUpCheck):
2020-04-16 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Implement JSMapIterator/JSSetIterator with JSInternalFieldObjectImpl
https://bugs.webkit.org/show_bug.cgi?id=210023
Reviewed by Keith Miller.
This patch reimplement JSMapIterator/JSSetIterator with JSInternalFieldObjectImpl.
This makes current JSFinalObject-based Map/SetIterator simple and small.
We generalize NewArrayIterator/PhantomNewArrayIterator to convert them to NewInternalFieldObject/PhantomNewInternalFieldObject
to support JSMapIterator/JSSetIterator too in DFG / FTL. This makes allocation efficient and object-allocation-sinking aware.
* builtins/BuiltinNames.h:
* builtins/MapIteratorPrototype.js:
(globalPrivate.mapIteratorNext):
(next):
* builtins/MapPrototype.js:
(globalPrivate.MapIterator): Deleted.
(values): Deleted.
(keys): Deleted.
(entries): Deleted.
* builtins/SetIteratorPrototype.js:
(globalPrivate.setIteratorNext):
(next):
* builtins/SetPrototype.js:
(globalPrivate.SetIterator): Deleted.
(values): Deleted.
(entries): Deleted.
* bytecode/BytecodeIntrinsicRegistry.cpp:
(JSC::BytecodeIntrinsicRegistry::BytecodeIntrinsicRegistry):
* bytecode/BytecodeIntrinsicRegistry.h:
* bytecompiler/BytecodeGenerator.h:
(JSC::BytecodeGenerator::emitIsMapIterator):
(JSC::BytecodeGenerator::emitIsSetIterator):
* bytecompiler/NodesCodegen.cpp:
(JSC::mapIteratorInternalFieldIndex):
(JSC::setIteratorInternalFieldIndex):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_getMapIteratorInternalField):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_getSetIteratorInternalField):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_putMapIteratorInternalField):
(JSC::BytecodeIntrinsicNode::emit_intrinsic_putSetIteratorInternalField):
* dfg/DFGAbstractInterpreterInlines.h:
(JSC::DFG::AbstractInterpreter<AbstractStateType>::executeEffects):
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handleIntrinsicCall):
* dfg/DFGClobberize.h:
(JSC::DFG::clobberize):
* dfg/DFGClobbersExitState.cpp:
(JSC::DFG::clobbersExitState):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::foldConstants):
* dfg/DFGDoesGC.cpp:
(JSC::DFG::doesGC):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::fixupNode):
* dfg/DFGMayExit.cpp:
* dfg/DFGNode.h:
(JSC::DFG::Node::convertToPhantomNewInternalFieldObject):
(JSC::DFG::Node::hasStructure):
(JSC::DFG::Node::isPhantomAllocation):
(JSC::DFG::Node::convertToPhantomNewArrayIterator): Deleted.
* dfg/DFGNodeType.h:
* dfg/DFGObjectAllocationSinkingPhase.cpp:
* dfg/DFGOperations.cpp:
* dfg/DFGOperations.h:
* dfg/DFGPredictionPropagationPhase.cpp:
* dfg/DFGSafeToExecute.h:
(JSC::DFG::safeToExecute):
* dfg/DFGSpeculativeJIT.cpp:
(JSC::DFG::SpeculativeJIT::compileNewInternalFieldObjectImpl):
(JSC::DFG::SpeculativeJIT::compileNewGenerator):
(JSC::DFG::SpeculativeJIT::compileNewAsyncGenerator):
(JSC::DFG::SpeculativeJIT::compileNewInternalFieldObject):
(JSC::DFG::SpeculativeJIT::compileNewArrayIterator): Deleted.
* dfg/DFGSpeculativeJIT.h:
* dfg/DFGSpeculativeJIT32_64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGSpeculativeJIT64.cpp:
(JSC::DFG::SpeculativeJIT::compile):
* dfg/DFGStoreBarrierInsertionPhase.cpp:
* ftl/FTLCapabilities.cpp:
(JSC::FTL::canCompile):
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileNode):
(JSC::FTL::DFG::LowerDFGToB3::compileNewInternalFieldObjectImpl):
(JSC::FTL::DFG::LowerDFGToB3::compileNewGenerator):
(JSC::FTL::DFG::LowerDFGToB3::compileNewAsyncGenerator):
(JSC::FTL::DFG::LowerDFGToB3::compileNewInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileMaterializeNewInternalFieldObject):
(JSC::FTL::DFG::LowerDFGToB3::compileNewArrayIterator): Deleted.
* ftl/FTLOperations.cpp:
(JSC::FTL::operationPopulateObjectInOSR):
(JSC::FTL::operationMaterializeObjectInOSR):
* inspector/JSInjectedScriptHost.cpp:
(Inspector::JSInjectedScriptHost::subtype):
(Inspector::JSInjectedScriptHost::getInternalProperties):
(Inspector::cloneMapIteratorObject):
(Inspector::cloneSetIteratorObject):
(Inspector::JSInjectedScriptHost::iteratorEntries):
* runtime/CommonIdentifiers.h:
* runtime/Intrinsic.cpp:
(JSC::intrinsicName):
* runtime/Intrinsic.h:
* runtime/JSArrayIterator.h:
* runtime/JSGlobalObject.cpp:
(JSC::JSGlobalObject::init):
(JSC::JSGlobalObject::visitChildren):
* runtime/JSGlobalObject.h:
(JSC::JSGlobalObject::mapIteratorPrototype const):
(JSC::JSGlobalObject::setIteratorPrototype const):
(JSC::JSGlobalObject::mapIteratorStructure const):
(JSC::JSGlobalObject::setIteratorStructure const):
* runtime/JSMapIterator.cpp:
(JSC::JSMapIterator::createWithInitialValues):
(JSC::JSMapIterator::finishCreation):
(JSC::JSMapIterator::visitChildren):
* runtime/JSMapIterator.h:
* runtime/JSSetIterator.cpp:
(JSC::JSSetIterator::createWithInitialValues):
(JSC::JSSetIterator::finishCreation):
(JSC::JSSetIterator::visitChildren):
* runtime/JSSetIterator.h:
* runtime/JSType.cpp:
(WTF::printInternal):
* runtime/JSType.h:
* runtime/JSTypedArrayViewPrototype.cpp:
(JSC::JSTypedArrayViewPrototype::finishCreation):
* runtime/MapPrototype.cpp:
(JSC::MapPrototype::finishCreation):
(JSC::createMapIteratorObject):
(JSC::mapProtoFuncValues):
(JSC::mapProtoFuncKeys):
(JSC::mapProtoFuncEntries):
* runtime/SetPrototype.cpp:
(JSC::SetPrototype::finishCreation):
(JSC::createSetIteratorObject):
(JSC::setProtoFuncValues):
(JSC::setProtoFuncEntries):
* runtime/VM.cpp:
(JSC::VM::setIteratorStructureSlow): Deleted.
(JSC::VM::mapIteratorStructureSlow): Deleted.
* runtime/VM.h:
(JSC::VM::setIteratorStructure): Deleted.
(JSC::VM::mapIteratorStructure): Deleted.
2020-04-15 Yusuke Suzuki <ysuzuki@apple.com>
[JSC] Use ensureStillAliveHere in FTL when content of storage should be kept alive
https://bugs.webkit.org/show_bug.cgi?id=210583
<rdar://problem/61831515>
Reviewed by Mark Lam.
The content of Butterfly / ArrayStorage is kept alive only when the owner JSCell is alive.
This means that we should keep the owner JSCell alive if we are loading content of storage
which includes JSCells. This patch inserts ensureStillAliveHere in FTL to ensure this invariant.
* ftl/FTLJITCode.cpp:
(JSC::FTL::JITCode::~JITCode): Found that we get crash with `dumpDisassembly` if FTL::JITCode is destroyed while it fails to generate code while testing this.
* ftl/FTLLowerDFGToB3.cpp:
(JSC::FTL::DFG::LowerDFGToB3::compileGetByVal):
(JSC::FTL::DFG::LowerDFGToB3::compileArrayIndexOf):
(JSC::FTL::DFG::LowerDFGToB3::compileArrayPop):
(JSC::FTL::DFG::LowerDFGToB3::compileStringCharAt):
(JSC::FTL::DFG::LowerDFGToB3::compileStringCharCodeAt):
(JSC::FTL::DFG::LowerDFGToB3::compileStringCodePointAt):
(JSC::FTL::DFG::LowerDFGToB3::compileGetByOffset):
(JSC::FTL::DFG::LowerDFGToB3::compileMultiGetByOffset):
2020-04-15 Keith Miller <keith_miller@apple.com>
Disable Store-load pair auto-vectorization for JSC
https://bugs.webkit.org/show_bug.cgi?id=210574
Reviewed by Geoffrey Garen.
slp-vectorization appears to make our slow path code significantly
slower. That's because when we materialize our constant bytecode
structs into C++ we load all the fields at the same time then
widen them to the struct's member C++ size. Since we have 3
different possible sizes Clang generates a total mess of
code. Disabling this does not appear to be a regression on any
platform I tested and improves the performance of slow path code
significantly in micro benchmarks.
* CMakeLists.txt:
* Configurations/JavaScriptCore.xcconfig:
2020-04-15 Robin Morisset <rmorisset@apple.com>
Flaky Test: fetch/fetch-worker-crash.html
https://bugs.webkit.org/show_bug.cgi?id=187257
<rdar://problem/48527526>
Reviewed by Yusuke Suzuki.
The crash is coming from setExceptionPorts which is inlined in WTF::registerThreadForMachExceptionHandling.
From the error message we know that the problem is an "invalid port right".
http://web.mit.edu/darwin/src/modules/xnu/osfmk/man/thread_set_exception_ports.html tells us that the "port right" is the third parameter to thread_set_exception_ports, which is exceptionPort in our case.
exceptionPort is a global variable defined at the top of Signals.cpp:
static mach_port_t exceptionPort;
It is set in exactly one place:
kern_return_t kr = mach_port_allocate(mach_task_self(), MACH_PORT_RIGHT_RECEIVE, &exceptionPort);
in a std::call_once, in startMachExceptionHandlerThread().
Note that startMachExceptionHandlerThread() is called from the main thread just before the point where we are stuck.. and there is no synchronization to make sure it completed and its effect is visible to the worker thread before it uses exceptionPort.
So I think the crash is due to this race between allocating exceptionPort and using it, resulting in an invalid exceptionPort being sometimes passed to the kernel.
So this patch is a simple speculative fix, by running startMachExceptionHandlerThread() in initializeThreading(), before JSLock()::lock() can be run.
* runtime/InitializeThreading.cpp:
(JSC::initializeThreading):
2020-04-15 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed build fix for r260161.
* runtime/IntlObject.cpp:
(JSC::canonicalLangTag):
2020-04-15 Ross Kirsling <ross.kirsling@sony.com>
Unreviewed, address Darin's feedback on r260151.
* runtime/IntlObject.cpp:
(JSC::canonicalLangTag):
2020-04-15 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] Extension values should default to true, canonicalize without "-true"
https://bugs.webkit.org/show_bug.cgi?id=210457
Reviewed by Yusuke Suzuki.
This patch implements two simple intertwining updates to ECMA-402:
- Valueless extension keys should not be dropped when resolving locale
https://tc39.es/ecma402/#sec-resolvelocale (9.h.4.b)
- Following UTS 35, "-true" should not appear in canonicalized locale ids
https://tc39.es/ecma402/#sec-canonicalizeunicodelocaleid
https://unicode.org/reports/tr35/#Canonical_Unicode_Locale_Identifiers
('Any type or tfield value "true" is removed.')
* runtime/IntlObject.cpp:
(JSC::canonicalLangTag):
(JSC::resolveLocale):
2020-04-15 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] Fix Intl.DateTimeFormat patterns and fields in WebKit
https://bugs.webkit.org/show_bug.cgi?id=209783
Reviewed by Keith Miller.
This patch implements two intertwining normative changes to Intl.DateTimeFormat:
- Calendar setting must be taken into account when choosing a date-time pattern
https://github.com/tc39/ecma402/pull/349
- formatToParts must recognize relatedYear and yearName parts
https://github.com/tc39/ecma402/pull/349
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
(JSC::IntlDateTimeFormat::partTypeString):
2020-04-15 Devin Rousso <drousso@apple.com>
[ESNext] Implement logical assignment operators
https://bugs.webkit.org/show_bug.cgi?id=209716
Reviewed by Ross Kirsling.
Implement the logical assignment operators proposal, which is now Stage 3. It introduces
three new assignment operators which will only store the result of the rhs in the lhs if the
lhs meets the given condition:
- `??=`, for if the lhs is nullish (`null` or `undefined`)
- `||=`, for if the lhs is falsy
- `&&=`, for if the lhs is truthy
This short circuiting can be beneficial as it can avoid a redundant store when used in the
common JavaScript programming pattern of "defaulting" a parameter.
```js
function foo(x) {
x = x || 42;
}
```
If `x` is a truthy value, it would result in the rhs `x` being stored back into the lhs `x`.
In some situations, this can have negative unintended side-effects, such as for `innerHTML`.
Logical assignment operators, however, are defined such that they only store if the rhs is
to actually be needed/used, skipping the redundant store and simply returning lhs otherwise.
In the case of readonly references, this means that an error is only thrown when the
assignment occurs, meaning that if the lhs already satisfies the condition it will be used
and returned with no error.
* parser/ParserTokens.h:
* parser/Lexer.cpp:
(JSC::Lexer<T>::lexWithoutClearingLineTerminator):
* parser/Parser.cpp:
(JSC::Parser<LexerType>::parseAssignmentExpression):
* parser/ASTBuilder.h:
(JSC::ASTBuilder::makeAssignNode):
* parser/Nodes.h:
* parser/NodeConstructors.h:
(JSC::ShortCircuitReadModifyResolveNode::ShortCircuitReadModifyResolveNode): Added.
(JSC::ShortCircuitReadModifyBracketNode::ShortCircuitReadModifyBracketNode): Added.
(JSC::ShortCircuitReadModifyDotNode::ShortCircuitReadModifyDotNode): Added.
* bytecompiler/NodesCodegen.cpp:
(JSC::emitShortCircuitAssignment): Added.
(JSC::ShortCircuitReadModifyResolveNode::emitBytecode): Added.
(JSC::ShortCircuitReadModifyDotNode::emitBytecode): Added.
(JSC::ShortCircuitReadModifyBracketNode::emitBytecode): Added.
* runtime/OptionsList.h:
Add a `useLogicalAssignmentOperators` setting for controlling this feature.
2020-04-14 Devin Rousso <drousso@apple.com>
Web Inspector: Debugger: add a Step next that steps by expression
https://bugs.webkit.org/show_bug.cgi?id=210324
Reviewed by Timothy Hatcher.
Step next is a hybrid of Step over and Step into which continues execution to the next pause
opportunity within the current (or ancestor) call frame. It is especially useful when trying
to debug minified code, such as trying to continue to `c()` in `a() && b() && c();`, where
Step over would continue to the next statement (i.e. after the `;`) and Step in would
continue to the first line inside `a()` (and would require a Step out to get back).
* inspector/protocol/Debugger.json:
* inspector/agents/InspectorDebuggerAgent.h:
* inspector/agents/InspectorDebuggerAgent.cpp:
(Inspector::InspectorDebuggerAgent::stepNext): Added.
* debugger/Debugger.h:
* debugger/Debugger.cpp:
(JSC::Debugger::stepNextExpression): Added.
(JSC::Debugger::atExpression):
(JSC::Debugger::clearNextPauseState):
2020-04-13 Alexey Shvayka <shvaikalesh@gmail.com>
REGRESSION (r259587): bterlson/eshost throws during init in strict mode
https://bugs.webkit.org/show_bug.cgi?id=210470
Reviewed by Ross Kirsling.
This change makes $262.IsHTMLDDA of JSC shell a CustomValue, allowing it to be reassigned
and restoring compatibility with any version of https://github.com/bterlson/eshost.
Since putDirectCustomAccessor() is now used instead of putGetter(), scope exception assert
is no longer needed and can be safely removed, as well as JSObject::putGetter() export.
* jsc.cpp:
* runtime/JSObject.h:
2020-04-13 David Kilzer <ddkilzer@apple.com>
Replace use of Checked<size_t, RecordOverflow> with CheckedSize
<https://webkit.org/b/210461>
Reviewed by Mark Lam.
* heap/Heap.cpp:
(JSC::Heap::deprecatedReportExtraMemorySlowCase):
(JSC::Heap::extraMemorySize):
(JSC::Heap::updateAllocationLimits):
(JSC::Heap::reportExtraMemoryVisited):
* heap/SlotVisitor.h:
* runtime/ArgList.cpp:
(JSC::MarkedArgumentBuffer::expandCapacity):
2020-04-10 Michael Saboff <msaboff@apple.com>
[YARR] Allow for Unicode named capture group identifiers in non-Unicode regular expressions
https://bugs.webkit.org/show_bug.cgi?id=210309
Reviewed by Ross Kirsling.
Update YARR pattern processing to allow for non-BMP unicode identifier characters in named capture groups.
This change was discussed and approved at the March/April 2020 TC-39 meeting.
See https://github.com/tc39/ecma262/pull/1869 for the discussion and change.
Updated tryConsumeUnicodeEscape() to allow for unicode escapes in non-unicode flagged regex's.
Added the same support to consumePossibleSurrogatePair().
* yarr/YarrParser.h:
(JSC::Yarr::Parser::consumePossibleSurrogatePair):
(JSC::Yarr::Parser::parseCharacterClass):
(JSC::Yarr::Parser::parseTokens):
(JSC::Yarr::Parser::tryConsumeUnicodeEscape):
(JSC::Yarr::Parser::tryConsumeIdentifierCharacter):
2020-04-13 Michael Catanzaro <mcatanzaro@gnome.org>
Fix various build warnings
https://bugs.webkit.org/show_bug.cgi?id=210429
Reviewed by Mark Lam.
Fix -Wimplicit-fallthrough warning by adding a default case CRASH() to prevent the inner
switch from falling through to the outer switch.
* dfg/DFGArrayMode.cpp:
(JSC::DFG::ArrayMode::alreadyChecked const):
2020-04-12 Mark Lam <mark.lam@apple.com>
Enable the ability to build the ASM LLInt for ARMv7k.
https://bugs.webkit.org/show_bug.cgi?id=210412
Reviewed by Sam Weinig.
Fix the offlineasm so that it can build the ASM LLInt for ARMv7k. This patch does
not actually enable the ASM LLInt. The ARMv7k port still build the C Loop LLInt.
Also, the ARMv7k ASM LLInt is still broken and needs additional work before it
can run. This patch only fixes things so that it will build.
* JavaScriptCore.xcodeproj/project.pbxproj:
- Added generate_settings_extractor.rb to the project so that we can view it from
inside Xcode.
* offlineasm/arm.rb:
- Added support for the globaladdr LLInt instruction for ARMv7k.
* offlineasm/backends.rb:
- Fix the backend to enable ARMV7 also when building for ARMv7k.
2020-04-12 Darin Adler <darin@apple.com>
Fix a few mispellings of descendant and propagation
https://bugs.webkit.org/show_bug.cgi?id=210409
Reviewed by Mark Lam.
* ftl/FTLAbstractHeap.h: "descendants"
* offlineasm/ast.rb: "descendants"
2020-04-12 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] WebKit Intl does not allow calendar and numberingSystem options
https://bugs.webkit.org/show_bug.cgi?id=209784
Reviewed by Myles C. Maxfield.
As an alternative to using `ca` and `nu` extensions in the locale string:
- the Intl.DateTimeFormat constructor needs to be able to take `calendar` and `numberingSystem` options
https://tc39.es/ecma402/#sec-initializedatetimeformat
- the Intl.NumberFormat needs to be able to take a `numberingSystem` option
https://tc39.es/ecma402/#sec-initializenumberformat
Since we already support `ca` and `nu`, this is a very simple addition.
The only interesting part is that we must verify that values for these options are 3-8 alphanumeric characters.
* runtime/IntlDateTimeFormat.cpp:
(JSC::IntlDateTimeFormat::initializeDateTimeFormat):
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::initializeNumberFormat):
(JSC::IntlNumberFormat::resolvedOptions):
* runtime/IntlObject.cpp:
(JSC::isUnicodeLocaleIdentifierType):
* runtime/IntlObject.h:
2020-04-10 Ross Kirsling <ross.kirsling@sony.com>
[ECMA-402] Properly implement BigInt.prototype.toLocaleString
https://bugs.webkit.org/show_bug.cgi?id=209782
Reviewed by Darin Adler.
Our BigInt's toLocaleString has been simply falling back to toString instead of following ECMA-402.
(https://tc39.es/ecma402/#sup-bigint.prototype.tolocalestring)
Since {Number, BigInt}.prototype.toLocaleString are internally the same as Intl.NumberFormat.prototype.format,
this patch simultaneously lets the latter method take a BigInt argument.
(https://tc39.es/ecma402/#sec-number-format-functions)
This patch continues to use the old unum_* API instead of ICU 62's new unumf_* API,
as the latter would require a large refactor as well as fallback paths.
(This will, however, be a prerequisite for https://bugs.webkit.org/show_bug.cgi?id=209774.)
* runtime/BigIntPrototype.cpp:
(JSC::bigIntProtoFuncToString):
(JSC::bigIntProtoFuncToLocaleString):
(JSC::bigIntProtoFuncToStringImpl): Deleted.
* runtime/IntlNumberFormat.cpp:
(JSC::IntlNumberFormat::format):
(JSC::IntlNumberFormat::formatNumber): Deleted.
* runtime/IntlNumberFormat.h:
* runtime/IntlNumberFormatPrototype.cpp:
(JSC::IntlNumberFormatFuncFormat):
(JSC::IntlNumberFormatPrototypeGetterFormat):
(JSC::IntlNumberFormatFuncFormatNumber): Deleted.
* runtime/NumberPrototype.cpp:
(JSC::numberProtoFuncToLocaleString):
2020-04-10 Devin Rousso <drousso@apple.com>
The rhs in `ReadModifyResolveNode` should be evaluated before throwing an exception if the lhs is read-only
https://bugs.webkit.org/show_bug.cgi?id=210317
Reviewed by Ross Kirsling.
* bytecompiler/NodesCodegen.cpp:
(JSC::emitReadModifyAssignment):
(JSC::ReadModifyResolveNode::emitBytecode):
If the corresponding `Variable` is read-only, pass it to `emitReadModifyAssignment` as an
additional optionl argument, where it will be used to `emitReadOnlyExceptionIfNeeded` after
the rhs is emitted.
2020-04-10 Mark Lam <mark.lam@apple.com>
Use more PAC diversity for JIT probe code.
https://bugs.webkit.org/show_bug.cgi?id=210252
<rdar://problem/54490367>
Reviewed by Keith Miller.
Introducing new PtrTags:
JITProbePtrTag - for the client probe function.
JITProbeTrampolinePtrTag - for calling the ctiMasmProbeTrampoline.
JITProbeExecutorPtrTag - for calling the probe executor.
Currently, this is only the Probe::executeProbe().
JITProbeStackInitializationFunctionPtrTag - for calling the optional stack
initialization function that the client probe function may set.
We'll now use these in the JIT probe mechanism instead of adopting the default
CFunctionPtrTag.
* assembler/MacroAssembler.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerARM64.cpp:
(JSC::MacroAssembler::probe):
* assembler/MacroAssemblerPrinter.h:
(JSC::MacroAssembler::print):
* assembler/ProbeContext.h:
* runtime/JSCPtrTag.h:
* tools/JSDollarVM.cpp:
(JSC::callWithStackSizeProbeFunction):
* wasm/WasmAirIRGenerator.cpp:
(JSC::Wasm::AirIRGenerator::emitLoopTierUpCheck):
* wasm/WasmB3IRGenerator.cpp:
(JSC::Wasm::B3IRGenerator::emitLoopTierUpCheck):
2020-04-10 Mark Lam <mark.lam@apple.com>
[Follow up] Fix bad tests in testmasm's testCagePreservesPACFailureBit().
https://bugs.webkit.org/show_bug.cgi?id=210314
<rdar://problem/61556785>
Not reviewed.
Applying Keith's feedback in https://bugs.webkit.org/show_bug.cgi?id=210314#c5:
added the stronger test but kept the weaker one as well.
* assembler/testmasm.cpp:
(JSC::testCagePreservesPACFailureBit):
== Rolled over to ChangeLog-2020-04-10 ==