Google Git
Sign in
cobalt / cobalt / e77c070364e97b7a7deea396227c08805cb9e66d / . / src / third_party / WebKit / Source / JavaScriptCore / ChangeLog-2008-08-10
blob: 0912aecff05efaa43bdcc24d77a1a972c2ca323d [file] [log] [blame]
2008-08-10 Jan Michael Alonzo <jmalonzo@webkit.org>
Reviewed (and updated) by Alp Toker.
https://bugs.webkit.org/show_bug.cgi?id=16620
[GTK] Autotools make dist and make check support
Get make dist working.
Note that not all possible configurations have been tested yet.
* GNUmakefile.am:
2008-08-09 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Sam Weinig.
Added same heap debug checks to more code paths.
* kjs/JSActivation.cpp:
(KJS::JSActivation::put):
(KJS::JSActivation::putWithAttributes):
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::putWithAttributes):
* kjs/JSObject.h:
(KJS::JSObject::putDirect):
* kjs/JSVariableObject.h:
(KJS::JSVariableObject::symbolTablePut):
(KJS::JSVariableObject::symbolTablePutWithAttributes):
2008-08-09 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Maciej.
Fix some style issues in the sampling tool.
* VM/SamplingTool.cpp:
(KJS::sleepForMicroseconds):
(KJS::SamplingTool::dump):
2008-08-09 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Oliver.
Revision 35651, despite being a rather trivial change, introduced a
large regression on the regexp-dna SunSpider test. This regression
stemmed from an increase in the size of CodeBlock::dump(). There is
no reason for this method (and several related methods) to be compiled
in non-debug builds with the sampling tool disabled. This patch
conditionally compiles them, reversing the regression on SunSpider.
* JavaScriptCore.exp:
* VM/CodeBlock.cpp:
* VM/CodeBlock.h:
* VM/Machine.cpp:
2008-08-08 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Oliver.
Bug 20330: JSCore crash loading any filehurricane media page
<https://bugs.webkit.org/show_bug.cgi?id=20330>
Fix a typo in the constant loading patch. Also, add a case for
op_unexpected_load to CodeBlock::dump().
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::addUnexpectedConstant):
2008-08-08 Matt Lilek <webkit@mattlilek.com>
Not reviewed, build fix.
* JavaScriptCore.exp:
2008-08-08 Oliver Hunt <oliver@apple.com>
Reviewed by Cameron Zwarich.
Improve performance of arithmetic operators
Added a fast (non-virtual) mechanism to determine if a non-immediate JSValue*
is a JSNumberCell. We then use this to allow improved specialisation in many
arithmetic operators. SunSpider reports a 2.5% progression overall, with greater
than 10% progressions on a number of arithmetic heavy tests.
* VM/Machine.cpp:
(KJS::fastIsNumber):
(KJS::fastToInt32):
(KJS::fastToUInt32):
(KJS::jsLess):
(KJS::jsLessEq):
(KJS::jsAdd):
(KJS::Machine::privateExecute):
* kjs/JSNumberCell.h:
(KJS::JSNumberCell::fastToInt32):
(KJS::JSNumberCell::fastToUInt32):
* kjs/collector.cpp:
(KJS::allocateBlock):
(KJS::Heap::heapAllocate):
* kjs/collector.h:
(KJS::Heap::fastIsNumber):
2008-08-06 Adam Roben <aroben@apple.com>
Try to fix the Windows build bots
* API/JSBase.cpp: Touch this to force JSC to rebuild and re-copy the
WTF headers.
2008-08-06 Tor Arne Vestbø <tavestbo@trolltech.com>
Revert change 35595.
* wtf/RetainPtr.h:
2008-08-06 Ariya Hidayat <ariya.hidayat@trolltech.com>
Fix non-Mac build.
* wtf/RetainPtr.h: CoreFoundation only for PLATFORM(MAC)
2008-08-06 Ariya Hidayat <ariya.hidayat@trolltech.com>
Fix non-Mac build.
* wtf/RetainPtr.h: CoreFoundation only for PLATFORM(MAC)
2008-08-06 Csaba Osztrogonac <oszi@inf.u-szeged.hu>
Reviewed by Darin. Landed by Cameron.
Bug 20272: typo in JavaScriptCore
<https://bugs.webkit.org/show_bug.cgi?id=20272>
Correct the documentation for op_not. (typo)
Fix #undef. (typo)
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-08-06 Cameron Zwarich <cwzwarich@webkit.org>
Reviewed by Maciej.
Bug 20286: Load constants all at once instead of using op_load
<https://bugs.webkit.org/show_bug.cgi?id=20286>
Load constants all at once into temporary registers instead of using
individual instances of op_load.
This is a 2.6% speedup on SunSpider.
* JavaScriptCore.exp:
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
(KJS::CodeBlock::mark):
* VM/CodeBlock.h:
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::CodeGenerator):
(KJS::CodeGenerator::newTemporary):
(KJS::CodeGenerator::addConstant):
(KJS::CodeGenerator::addUnexpectedConstant):
(KJS::CodeGenerator::emitLoad):
(KJS::CodeGenerator::emitUnexpectedLoad):
(KJS::CodeGenerator::emitNewError):
* VM/CodeGenerator.h:
* VM/Machine.cpp:
(KJS::slideRegisterWindowForCall):
(KJS::Machine::unwindCallFrame):
(KJS::Machine::throwException):
(KJS::Machine::execute):
(KJS::Machine::privateExecute):
* VM/Machine.h:
* VM/Opcode.h:
* VM/RegisterID.h:
(KJS::RegisterID::RegisterID):
(KJS::RegisterID::makeConstant):
(KJS::RegisterID::isTemporary):
* kjs/NodeInfo.h:
* kjs/Parser.cpp:
(KJS::Parser::didFinishParsing):
* kjs/Parser.h:
(KJS::Parser::parse):
* kjs/grammar.y:
* kjs/nodes.cpp:
(KJS::NullNode::emitCode):
(KJS::BooleanNode::emitCode):
(KJS::NumberNode::emitCode):
(KJS::StringNode::emitCode):
(KJS::ArrayNode::emitCode):
(KJS::DeleteResolveNode::emitCode):
(KJS::DeleteValueNode::emitCode):
(KJS::VoidNode::emitCode):
(KJS::ConstDeclNode::emitCodeSingle):
(KJS::ReturnNode::emitCode):
(KJS::ScopeNode::ScopeNode):
(KJS::ProgramNode::ProgramNode):
(KJS::ProgramNode::create):
(KJS::EvalNode::EvalNode):
(KJS::EvalNode::create):
(KJS::FunctionBodyNode::FunctionBodyNode):
(KJS::FunctionBodyNode::create):
(KJS::FunctionBodyNode::emitCode):
* kjs/nodes.h:
(KJS::ScopeNode::neededConstants):
2008-08-05 Maciej Stachowiak <mjs@apple.com>
Reviewed by Cameron.
- add fast path for immediates to % operator, as we have for many other math ops
This fixes handling for a 0 divisor relative to the last patch. Only an 0.2% speedup on SunSpider but
still a 1.4x win on Oliver's prime test.
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-08-05 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Darin.
Bug 20293: Crash in JavaScript codegen for eval("const a;")
<https://bugs.webkit.org/show_bug.cgi?id=20293>
Correctly handle constant declarations in eval code with no initializer.
* kjs/nodes.cpp:
(KJS::ConstDeclNode::emitCodeSingle):
2008-08-05 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Oliver.
Roll out r35555 because of correctness issues.
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-08-05 Maciej Stachowiak <mjs@apple.com>
Reviewed by Geoff.
- add fast path for immediates to % operator, as we have for many other math ops
0.6% speedup on SunSpider. 1.4x speedup on a prime testing torture test that Oliver whipped up.
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-07-31 Oliver Hunt <oliver@apple.com>
Reviewed by Cameron Zwarich.
Bug 19359: JavaScriptCore behaves differently from FF2/3 and IE when handling context in catch statement
<https://bugs.webkit.org/show_bug.cgi?id=19359>
Make our catch behave like Firefox and IE, we do this by using a StaticScopeObject
instead of a generic JSObject for the scope node. We still don't make use of the
fact that we have a static scope inside the catch block, so the internal performance
of the catch block is not improved, even though technically it would be possible to
do so.
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitPushNewScope):
* VM/CodeGenerator.h:
* VM/Machine.cpp:
(KJS::createExceptionScope):
(KJS::Machine::privateExecute):
* VM/Machine.h:
* VM/Opcode.h:
* kjs/JSStaticScopeObject.cpp:
(KJS::JSStaticScopeObject::toThisObject):
(KJS::JSStaticScopeObject::put):
* kjs/JSStaticScopeObject.h:
* kjs/nodes.cpp:
(KJS::TryNode::emitCode):
2008-08-02 Rob Gowin <robg@gowin.net>
Reviewed by Eric Seidel.
Added JavaScriptCore/API/WebKitAvailability to list of files in
javascriptcore_h_api.
* GNUmakefile.am:
2008-08-01 Alexey Proskuryakov <ap@webkit.org>
Rubber-stamped by Maciej.
Remove JSGlobalData::DataInstance. It was only needed when we had per-thread JSGlobalData
instances.
* kjs/JSGlobalData.h:
2008-07-31 Kevin Ollivier <kevino@theolliviers.com>
Second attempt at Windows/wx build fix. Instead of avoiding inclusion of windows.h,
use defines, etc. to avoid conflicts in each affected file. Also, change PLATFORM(WIN)
to PLATFORM(WIN_OS) so that other ports using Windows headers get the right impls.
* VM/SamplingTool.cpp:
* wtf/Threading.h:
2008-07-31 Anders Carlsson <andersca@apple.com>
Reviewed by Adam.
Fix Windows build.
* kjs/collector.h:
* wtf/FastMalloc.cpp:
2008-07-31 Csaba Osztrogonac <oszi@inf.u-szeged.hu>
Reviewed by Simon.
Bug 20170: [Qt] missing namespace defines in JavaScriptCore.pro
<https://bugs.webkit.org/show_bug.cgi?id=20170>
* JavaScriptCore.pro: Added missing define.
2008-07-31 Alexey Proskuryakov <ap@webkit.org>
Rubber-stamped by Maciej.
Eliminate JSLock (it was already disabled, removing the stub implementaion and all
call sites now).
* API/JSBase.cpp:
(JSEvaluateScript):
(JSCheckScriptSyntax):
(JSGarbageCollect):
* API/JSCallbackConstructor.cpp:
(KJS::constructJSCallback):
* API/JSCallbackFunction.cpp:
(KJS::JSCallbackFunction::call):
* API/JSCallbackObjectFunctions.h:
(KJS::::init):
(KJS::::getOwnPropertySlot):
(KJS::::put):
(KJS::::deleteProperty):
(KJS::::construct):
(KJS::::hasInstance):
(KJS::::call):
(KJS::::getPropertyNames):
(KJS::::toNumber):
(KJS::::toString):
(KJS::::staticValueGetter):
(KJS::::callbackGetter):
* API/JSContextRef.cpp:
(JSGlobalContextCreateInGroup):
(JSGlobalContextRetain):
(JSGlobalContextRelease):
* API/JSObjectRef.cpp:
(JSObjectMake):
(JSObjectMakeFunctionWithCallback):
(JSObjectMakeConstructor):
(JSObjectMakeFunction):
(JSObjectHasProperty):
(JSObjectGetProperty):
(JSObjectSetProperty):
(JSObjectGetPropertyAtIndex):
(JSObjectSetPropertyAtIndex):
(JSObjectDeleteProperty):
(JSObjectCallAsFunction):
(JSObjectCallAsConstructor):
(JSObjectCopyPropertyNames):
(JSPropertyNameArrayRelease):
(JSPropertyNameAccumulatorAddName):
* API/JSStringRef.cpp:
(JSStringRelease):
* API/JSValueRef.cpp:
(JSValueIsEqual):
(JSValueIsInstanceOfConstructor):
(JSValueMakeNumber):
(JSValueMakeString):
(JSValueToNumber):
(JSValueToStringCopy):
(JSValueToObject):
(JSValueProtect):
(JSValueUnprotect):
* ForwardingHeaders/JavaScriptCore/JSLock.h: Removed.
* GNUmakefile.am:
* JavaScriptCore.exp:
* JavaScriptCore.order:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
* kjs/JSGlobalData.h:
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::~JSGlobalObject):
(KJS::JSGlobalObject::init):
* kjs/JSLock.cpp: Removed.
* kjs/JSLock.h: Removed.
* kjs/Shell.cpp:
(functionGC):
(jscmain):
* kjs/collector.cpp:
(KJS::Heap::~Heap):
(KJS::Heap::heapAllocate):
(KJS::Heap::setGCProtectNeedsLocking):
(KJS::Heap::protect):
(KJS::Heap::unprotect):
(KJS::Heap::collect):
* kjs/identifier.cpp:
* kjs/interpreter.cpp:
(KJS::Interpreter::checkSyntax):
(KJS::Interpreter::evaluate):
2008-07-31 Alexey Proskuryakov <ap@webkit.org>
Rubber-stamped by Oliver Hunt.
Fix the Mac project to not display "test/" as part of file name for tests.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-31 Eric Seidel <eric@webkit.org>
Reviewed by Alexey Proskuryakov.
Rename USE(MULTIPLE_THREADS) to ENABLE(JSC_MULTIPLE_THREADS)
to better match the use/enable pattern (and better describe
the usage of the feature in question.)
I also fixed a couple other ENABLE_ macros to be pre-processor
definition override-able to match the rest of the ENABLE_ macros
since it seems to be our convention that build systems can set
ENABLE_ macros in Makefiles.
* kjs/InitializeThreading.cpp:
(KJS::initializeThreadingOnce):
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
(KJS::JSGlobalData::~JSGlobalData):
* kjs/MathObject.cpp:
* kjs/collector.cpp:
(KJS::Heap::Heap):
(KJS::Heap::~Heap):
(KJS::allocateBlock):
(KJS::Heap::markStackObjectsConservatively):
* kjs/collector.h:
* kjs/dtoa.cpp:
(KJS::pow5mult):
(KJS::rv_alloc):
(KJS::freedtoa):
(KJS::dtoa):
* wtf/FastMalloc.cpp:
* wtf/Platform.h:
* wtf/RefCountedLeakCounter.cpp:
2008-07-30 Eric Seidel <eric@webkit.org>
Reviewed by Mark Rowe.
Try to clean up our usage of USE(MULTIPLE_THREADS) vs. USE(PTHREADS) a little.
It looks like JSC assumes that if MULTIPLE_THREADS is defined, then pthreads will always be available
I'm not sure that's always the case for gtk, certainly not for Windows. We should eventually go back
and fix wtf/Threading.h to cover all these cases some day.
* kjs/JSLock.cpp:
* kjs/collector.h:
* wtf/Platform.h:
2008-07-30 Eric Seidel <eric@webkit.org>
Reviewed by Oliver.
MSVC warns when structs are called classes or vice versa.
Make all the source refer to JSGlobalData as a class.
* kjs/CommonIdentifiers.h:
* kjs/JSGlobalData.h:
* kjs/Parser.h:
* kjs/lexer.h:
2008-07-30 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
Add consistency checks to UString to document and enforce its design.
* kjs/ustring.cpp:
(KJS::UString::Rep::create):
(KJS::UString::Rep::destroy):
(KJS::UString::Rep::checkConsistency):
(KJS::UString::expandCapacity):
(KJS::UString::expandPreCapacity):
(KJS::UString::UString):
(KJS::UString::spliceSubstringsWithSeparators):
(KJS::UString::append):
* kjs/ustring.h:
(KJS::UString::Rep::checkConsistency):
2008-07-30 Gavin Barraclough <barraclough@apple.com>
Reviewed by Geoff Garen.
Fixes for Windows and non-AllInOne file build with SamplingTool, plus review fixes.
* GNUmakefile.am: Adding SamplingTool.cpp to build.
* JavaScriptCore.exp: Export hooks to init & control SamplingTool.
* JavaScriptCore.pri: Adding SamplingTool.cpp to build.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: Adding SamplingTool.cpp to build.
* JavaScriptCore.xcodeproj/project.pbxproj: Adding SamplingTool.cpp to build.
* JavaScriptCoreSources.bkl: Adding SamplingTool.cpp to build.
* VM/Machine.cpp: MACHINE_SAMPLING_callingNativeFunction renamed MACHINE_SAMPLING_callingHostFunction
* VM/Machine.h:
* VM/Opcode.cpp: SamplingTool moved to SamplingTool.cpp/.h, opcodeNames generated from FOR_EACH_OPCODE_ID.
* VM/Opcode.h:
* VM/SamplingTool.cpp: Added .cpp/.h for SamplingTool.
* VM/SamplingTool.h:
* kjs/Shell.cpp: Switched SAMPLING_TOOL_ENABLED to ENABLE_SAMPLING_TOOL.
* wtf/Platform.h: Added ENABLE_SAMPLING_TOOL config option.
* kjs/nodes.cpp: Header include to fix non-AllInOne builds.
2008-07-30 Ariya Hidayat <ariya.hidayat@trolltech.com>
Reviewed by Alexey Proskuryakov.
Fix compilation without multi-threading support.
* kjs/collector.cpp:
(KJS::Heap::Heap):
2008-07-30 Anders Carlsson <andersca@apple.com>
Add WebKitAvailability.h forwarding header.
* ForwardingHeaders/JavaScriptCore/WebKitAvailability.h: Added.
2008-07-30 Anders Carlsson <andersca@apple.com>
Fix the else.
* API/WebKitAvailability.h:
2008-07-30 Anders Carlsson <andersca@apple.com>
* API/WebKitAvailability.h:
Fix Windows (and other non-Mac builds).
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
Add WebKitAvailability.h to the project.
2008-07-30 Anders Carlsson <andersca@apple.com>
One step closer towards fixing the Windows build.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCoreGenerated.make:
Make sure to copy WebKitAvailability.h
2008-07-29 Gavin Barraclough <barraclough@apple.com>
Reviewed by Geoff Garen.
Bug 20209: Atomize constant strings
<https://bugs.webkit.org/show_bug.cgi?id=20209>
Prevents significant performance degradation seen when a script contains multiple
identical strings that are used as keys to identify properties on objects.
No performance change on SunSpider.
* kjs/nodes.cpp: Atomize constant strings.
2008-07-30 Oliver Hunt <oliver@apple.com>
Reviewed by Alexey Proskuryakov.
<rdar://problem/6111648> JavaScript exceptions fail if the scope chain includes the global object
In an attempt to remove the branch I just added to KJS::depth I
used the existence of a Variable Object at a point in the scope
chain as an indicator of function or global scope activation.
However this assumption results in incorrect behaviour if the
global object is injected into the scope chain with 'with'.
* VM/Machine.cpp:
(KJS::depth):
2008-07-30 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
Don't call JSGarbageCollect() on a released context.
* API/testapi.c: (main):
2008-07-29 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
Implement JSContextGroup APIs to make concurrent execution possible for
JavaScriptCore clients.
This changes the behavior of JSGlobalContextCreate(), so that it now uses a private context
group for each context, making JSlock implicit locking unnecessary.
* API/JSContextRef.h:
* API/JSContextRef.cpp:
(JSContextGroupCreate):
(JSContextGroupRetain):
(JSContextGroupRelease):
(JSGlobalContextCreate):
(JSGlobalContextCreateInGroup):
(JSGlobalContextRelease):
(JSContextGetGroup):
Added new methods. JSGlobalContextCreate() calls JSGlobalContextCreateInGroup() now.
* API/APICast.h: (toJS): (toRef): Added converters for JSContextGroupRef.
* API/JSBase.cpp: (JSGarbageCollect): JSGarbageCollect(0) is now a no-op, and the passed in
context is actually used.
* API/JSBase.h: Aded a typedef for JSContextGroupRef. Updated documentation for
JSGarbageCollect().
* JavaScriptCore.exp: Removed JSGlobalData::sharedInstance().
* kjs/JSGlobalData.cpp:
* kjs/JSGlobalData.h:
Removed support for JSGlobalData shared instance. JSGlobalData::isSharedInstance member
variable still remains, to be deleted in a followup patch.
* kjs/JSLock.cpp: (KJS::JSLock::JSLock): Disabled JSLock, to be deleted in a follow-up patch.
* kjs/collector.cpp:
(KJS::Heap::markOtherThreadConservatively): Removed an assertion that referenced
JSGlobalData::sharedInstance.
* kjs/collector.h: Made Heap destructor public, so that JSContextRelease can use it.
2008-07-29 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
Fix a leak of ThreadRegistrar objects.
As the heap is usually deleted when registered threads still exist, ThreadSpecific doesn't
have a chance to clean up per-thread object. Switched to native pthread calls, storing a
plain pointer that doesn't require cleanup.
* kjs/collector.cpp:
(KJS::PlatformThread::PlatformThread):
(KJS::Heap::Thread::Thread):
(KJS::Heap::Heap):
(KJS::Heap::~Heap):
(KJS::Heap::registerThread):
(KJS::Heap::unregisterThread):
* kjs/collector.h:
2008-07-29 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Sam Weinig.
https://bugs.webkit.org/show_bug.cgi?id=20169
Memory allocated with fastMalloc is freed with delete
* VM/JSPropertyNameIterator.cpp:
(KJS::JSPropertyNameIterator::invalidate): Free the array properly.
(KJS::JSPropertyNameIterator::~JSPropertyNameIterator): Delete the array by calling
invalidate().
2008-07-29 Mark Rowe <mrowe@apple.com>
Attempt to fix the Qt build.
* wtf/ThreadingQt.cpp: Add the extra argument to createThread.
2008-07-29 Adam Roben <aroben@apple.com>
Change Vector::find to return an index instead of an iterator
Indices are more natural than iterators when working with Vector.
Reviewed by John Sullivan.
* wtf/Vector.h:
(WTF::Vector::find): Changed to iterate the Vector manually and return
the index of the found item, rather than an iterator. When the item
could not be found, we return WTF::notFound.
2008-07-29 Adam Roben <aroben@apple.com>
Windows build fix
* wtf/ThreadingWin.cpp:
(WTF::setThreadName): Move a misplaced assertion to here...
(WTF::createThread): ...from here.
2008-07-29 Adam Roben <aroben@apple.com>
Add support for setting thread names on Windows
These thread names make it much easier to identify particular threads
in Visual Studio's Threads panel.
WTF::createThread now takes a const char* representing the thread's
name. On Windows, we throw a special exception to set this string as
the thread's name. Other platforms do nothing with this name for now.
Reviewed by Anders Carlsson.
* JavaScriptCore.exp: Export the new version of createThread that
takes 3 arguments (the old one continues to be exported for backward
compatibility).
* wtf/Threading.h: Add a threadName argument to createThread.
* wtf/ThreadingGtk.cpp:
(WTF::createThread):
* wtf/ThreadingNone.cpp:
(WTF::createThread):
Updated for function signature change.
* wtf/ThreadingPthreads.cpp:
(WTF::createThread): Updated for function signature change. We keep
around the old 2-argument version of createThread for backward
compatibility.
* wtf/ThreadingWin.cpp:
(WTF::setThreadName): Added. This function's implementation came from
MSDN.
(WTF::initializeThreading): Set the name of the main thread.
(WTF::createThread): Call setThreadName. We keep around the old
2-argument version of createThread for backward compatibility.
2008-07-29 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Oliver Hunt.
Store UString::Rep::isStatic bit in identifierTable pointer instead of reportedCost for
slightly nicer code and a 0.5% SunSpider improvement.
* API/JSClassRef.cpp:
(OpaqueJSClass::~OpaqueJSClass):
(OpaqueJSClassContextData::OpaqueJSClassContextData):
* API/JSStringRef.cpp:
(JSStringRelease):
* kjs/PropertyNameArray.cpp:
(KJS::PropertyNameArray::add):
* kjs/identifier.cpp:
(KJS::IdentifierTable::~IdentifierTable):
(KJS::IdentifierTable::add):
(KJS::Identifier::addSlowCase):
(KJS::Identifier::remove):
* kjs/identifier.h:
(KJS::Identifier::add):
* kjs/ustring.cpp:
(KJS::):
(KJS::UString::Rep::create):
(KJS::UString::Rep::destroy):
* kjs/ustring.h:
(KJS::UString::Rep::identifierTable):
(KJS::UString::Rep::setIdentifierTable):
(KJS::UString::Rep::isStatic):
(KJS::UString::Rep::setStatic):
(KJS::UString::cost):
2008-07-28 Geoffrey Garen <ggaren@apple.com>
Reviewed by Sam Weinig.
Renamed "ConstructTypeNative" => "ConstructTypeHost".
2008-07-26 Mark Rowe <mrowe@apple.com>
Speculative fix for the wx build.
* JavaScriptCoreSources.bkl: Add JSStaticScopeObject.cpp to the list of source files.
2008-07-25 Oliver Hunt <oliver@apple.com>
RS=Cameron Zwarich.
Whoops, forgot to save style correction.
* kjs/JSStaticScopeObject.h:
2008-07-25 Oliver Hunt <oliver@apple.com>
Reviewed by Cameron Zwarich.
Bug 19718: Named anonymous functions are slow accessing global variables
<https://bugs.webkit.org/show_bug.cgi?id=19718>
To fix this we switch over to an activation-like scope object for
on which we attach the function name property, and add logic to
prevent cross scope assignment to read only properties.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::findScopedProperty):
(KJS::CodeGenerator::emitResolve):
* VM/CodeGenerator.h:
* kjs/AllInOneFile.cpp:
* kjs/JSStaticScopeObject.cpp: Added.
(KJS::JSStaticScopeObject::putWithAttributes):
(KJS::JSStaticScopeObject::isDynamicScope):
(KJS::JSStaticScopeObject::~JSStaticScopeObject):
(KJS::JSStaticScopeObject::getOwnPropertySlot):
* kjs/JSStaticScopeObject.h: Added.
(KJS::JSStaticScopeObject::JSStaticScopeObjectData::JSStaticScopeObjectData):
(KJS::JSStaticScopeObject::JSStaticScopeObject):
* kjs/nodes.cpp:
(KJS::FunctionCallResolveNode::emitCode):
(KJS::PostfixResolveNode::emitCode):
(KJS::PrefixResolveNode::emitCode):
(KJS::ReadModifyResolveNode::emitCode):
(KJS::AssignResolveNode::emitCode):
(KJS::FuncExprNode::makeFunction):
2008-07-25 kevino <kevino@theolliviers.com>
wx build fix for Win.
On wx/Win, including windows.h in Threading.h causes multiply-defined symbol errors
for libjpeg and wx, and also wx needs to include windows.h itself first for wx
includes to work right. So until we can find a better solution to this problem,
on wx, we work around the need to include windows.h here.
* wtf/Threading.h:
2008-07-25 Adam Roben <aroben@apple.com>
Windows build fix
* JavaScriptCore.vcproj/testapi/testapi.vcproj: Add API/ to the
include path.
2008-07-25 Simon Hausmann <hausmann@webkit.org>
Fix the build of jsc on Qt/Windows, make sure os-win32 is in the
include search path (added by WebKit.pri).
* kjs/jsc.pro:
2008-07-25 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Simon Hausmann.
Move JavaScriptCore API tests into a subdirectory of their own to avoid header name
conflicts and developer confusion.
* API/JSNode.c: Removed.
* API/JSNode.h: Removed.
* API/JSNodeList.c: Removed.
* API/JSNodeList.h: Removed.
* API/Node.c: Removed.
* API/Node.h: Removed.
* API/NodeList.c: Removed.
* API/NodeList.h: Removed.
* API/minidom.c: Removed.
* API/minidom.html: Removed.
* API/minidom.js: Removed.
* API/testapi.c: Removed.
* API/testapi.js: Removed.
* API/tests: Added.
* API/tests/JSNode.c: Copied from JavaScriptCore/API/JSNode.c.
* API/tests/JSNode.h: Copied from JavaScriptCore/API/JSNode.h.
* API/tests/JSNodeList.c: Copied from JavaScriptCore/API/JSNodeList.c.
* API/tests/JSNodeList.h: Copied from JavaScriptCore/API/JSNodeList.h.
* API/tests/Node.c: Copied from JavaScriptCore/API/Node.c.
* API/tests/Node.h: Copied from JavaScriptCore/API/Node.h.
* API/tests/NodeList.c: Copied from JavaScriptCore/API/NodeList.c.
* API/tests/NodeList.h: Copied from JavaScriptCore/API/NodeList.h.
* API/tests/minidom.c: Copied from JavaScriptCore/API/minidom.c.
* API/tests/minidom.html: Copied from JavaScriptCore/API/minidom.html.
* API/tests/minidom.js: Copied from JavaScriptCore/API/minidom.js.
* API/tests/testapi.c: Copied from JavaScriptCore/API/testapi.c.
* API/tests/testapi.js: Copied from JavaScriptCore/API/testapi.js.
* GNUmakefile.am:
* JavaScriptCore.vcproj/testapi/testapi.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-25 Simon Hausmann <hausmann@webkit.org>
Prospective WX build fix, add JavaScriptCore/API to the include search
path.
* jscore.bkl:
2008-07-25 Simon Hausmann <hausmann@webkit.org>
Rubber-stamped by Lars.
Fix the build on Windows. operator new for ArgList is implemented using fastMalloc()
but operator delete was not implemented. Unfortunately MSVC decides to call/reference
the function, so a simple implementation using fastFree() fixes the build.
* kjs/ArgList.h:
(KJS::ArgList::operator delete):
2008-07-25 Simon Hausmann <hausmann@webkit.org>
Discussed with and rubber-stamped by Lars.
Fix the build system for the Qt port.
Recent JavaScriptCore changes require the addition of JavaScriptCore/API to the
include search path. With a build process that combines JavaScriptCore and
WebCore in one build process/Makefile the existance of
JavaScriptCore/API/Node.h and WebCore/dom/Node.h causes include conflicts.
This commit solves this by introducing a separate build of JavaScriptCore into
a static library.
As a result of the split-up a race-condition due to broken dependencies of
regular source files to header files of generated sources showed up very
frequently when doing parallel builds (which the buildbot does). This commit at
the same time tries to address the dependency problem by making the
addExtraCompiler() function also generate a pseudo extra compiler that
represents the header file output, so that qmake is aware of the creation of
the header file for dependency calculation.
At the same time I removed a lot of cruft from the pro files to ease maintenance.
* JavaScriptCore.pri:
* JavaScriptCore.pro: Added.
* kjs/jsc.pro:
2008-07-24 Geoffrey Garen <ggaren@apple.com>
Reviewed by Maciej Stachowiak.
Fixed a strict aliasing violation, which caused hash tables with floating
point keys not to find items that were indeed in the tables
(intermittently, and only in release builds, of course).
SunSpider reports no change.
This bug doesn't seem to affect any existing code, but it causes obvious
crashes in some new code I'm working on.
* wtf/HashFunctions.h:
(WTF::FloatHash::hash): Use a union when punning between a float / double
and an unsigned (bucket of bits). With strict aliasing enabled, unions
are the only safe way to do this kind of type punning.
* wtf/HashTable.h: When rehashing, ASSERT that the item we just added to
the table is indeed in the table. In the buggy case described above, this
ASSERT fires.
2008-07-24 Oliver Hunt <oliver@apple.com>
Reviewed by Alexey Proskuryakov.
Bug 20142: REGRESSION(r35245): /=/ weirdness
<https://bugs.webkit.org/show_bug.cgi?id=20142>
When adding all the meta data needed for exception error messages
I accidentally clobbered the handling of regex beginning with /=.
* kjs/grammar.y:
2008-07-23 Alp Toker <alp@nuanti.com>
Build fix after r35293: Add API/ to the include path.
* GNUmakefile.am:
2008-07-23 Adam Roben <aroben@apple.com>
Windows build fixes
Build fix after r35293:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: Add API/
to the include path.
Build fix after r35305:
* VM/Machine.cpp:
* VM/Machine.h:
* VM/Opcode.cpp:
* VM/Opcode.h:
Completely compile out all sampler-related code when
SAMPLING_TOOL_ENABLED is 0. The sampler code can't be compiled 1) on
non-AllInOne configurations due to circular header dependencies, and
2) on platforms that don't have a usleep() function, such as Windows.
2008-07-23 Oliver Hunt <oliver@apple.com>
Reviewed by Geoff Garen and Sam Weinig.
Improve switch performance.
Improve switch performance by converting to a hashmap based jump
table to avoid the sequence of dispatches that would otherwise be
needed. This results in a 9-19x performance win for string switches
based on ad hoc testing, and a 6x improvement for integer switch
statements. SunSpider reports a 1.2% progression.
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
(KJS::SimpleJumpTable::offsetForValue):
* VM/CodeBlock.h:
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::beginSwitch):
(KJS::prepareJumpTableForImmediateSwitch):
(KJS::prepareJumpTableForCharacterSwitch):
(KJS::prepareJumpTableForStringSwitch):
(KJS::CodeGenerator::endSwitch):
* VM/CodeGenerator.h:
* VM/Machine.cpp:
(KJS::offsetForStringSwitch):
(KJS::Machine::privateExecute):
* VM/Opcode.cpp:
(KJS::):
* VM/Opcode.h:
* kjs/JSImmediate.h:
* kjs/nodes.cpp:
(KJS::):
(KJS::processClauseList):
(KJS::CaseBlockNode::tryOptimisedSwitch):
(KJS::CaseBlockNode::emitCodeForBlock):
* kjs/nodes.h:
(KJS::SwitchInfo::):
2008-07-23 Gavin Barraclough <barraclough@apple.com>
Reviewed by Geoff Garen.
Sampling tool to analyze cost of instruction execution and identify hot regions of JS code.
Enable Switches by setting SAMPLING_TOOL_ENABLED in Opcode.h.
* JavaScriptCore.exp: Export symbols for Shell.cpp.
* VM/Machine.cpp: Added sampling hooks.
* VM/Machine.h: Machine contains a pointer to a sampler, when sampling.
* VM/Opcode.cpp: Tool implementation.
* VM/Opcode.h: Tool declaration.
* kjs/Shell.cpp: Initialize the sampler, if enabled.
* kjs/nodes.cpp: Added sampling hooks.
2008-07-23 Gabor Loki <loki@inf.u-szeged.hu>
Bug 20097: [Qt] 20% Sunspider slow-down
<https://bugs.webkit.org/show_bug.cgi?id=20097>
Reviewed by Simon Hausmann.
* kjs/jsc.pro: Added missing NDEBUG define for release builds.
2008-07-23 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
JSClassRef is created context-free, but gets infatuated with the first context it sees.
The implicit API contract is that JSClassRef can be used with any context on any thread.
This no longer worked, because UStrings in the class were turned into per-context
identifiers, and the cached JSObject prototype was tied to JSGlobalData, too.
* API/JSClassRef.h: Made a separate struct for context-dependent parts of OpaqueJSClass.
* API/JSClassRef.cpp:
(OpaqueJSClass::OpaqueJSClass): Updated for renames and changed member variable order.
(OpaqueJSClass::~OpaqueJSClass): Assert that string members are not identifiers.
(clearReferenceToPrototype): Update for the new reference location.
(OpaqueJSClassContextData::OpaqueJSClassContextData): Make a deep copy of all strings.
(OpaqueJSClass::contextData): Added a function that finds the per-context part of
OpaqueJSClass in JSGlobalData, or creates it if not found.
(OpaqueJSClass::className): Always make a deep copy. Callers of this function do not have
a way to access JSGlobalData, so a per-context copy could not be made.
(OpaqueJSClass::staticValues): Updated for new data location.
(OpaqueJSClass::staticFunctions): Ditto.
(OpaqueJSClass::prototype): Changed to take an internal type for consistency.
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
(KJS::JSGlobalData::~JSGlobalData):
* kjs/JSGlobalData.h:
Keep a HashMap to access per-context JSClass data given a pointr to the shared part.
* API/JSCallbackObjectFunctions.h:
(KJS::::className):
(KJS::::getOwnPropertySlot):
(KJS::::put):
(KJS::::deleteProperty):
(KJS::::getPropertyNames):
(KJS::::staticValueGetter):
(KJS::::staticFunctionGetter):j
Use function accessors instead of accessing OpaqueJSClass members directly.
* API/JSContextRef.cpp: (JSGlobalContextCreate): Updated for the change in
OpaqueJSClass::prototype() argument type.
* API/JSObjectRef.cpp:
(JSObjectMake): Updated for the change in OpaqueJSClass::prototype() argument type.
(JSObjectMakeConstructor): Ditto.
2008-07-23 Alexey Proskuryakov <ap@webkit.org>
Build fix.
* kjs/ArgList.h: (KJS::ArgList::operator new): removed an extraneous "ArgList::" inside the
class definition.
2008-07-22 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt and Sam Weinig.
Next step toward putting doubles in registers: Prepare the Register class
and its clients for registers that don't contain JSValue*s.
This means a few things:
1. Register::jsValue() clients, including ArgList clients, must now supply
an ExecState* when accessing an entry in an ArgList, in case the entry
will need to create a JSValue* on the fly.
2. Register clients that definitely don't want to create a JSValue* on
the fly now use different APIs: getJSValue() for clients that know
the register contains a JSValue*, and v() for clients who just want a
void*.
3. I had to change some headers around in order to resolve dependency
problems created by using a Register in the ArgList header.
SunSpider reports no change.
2008-07-22 Gavin Barraclough <barraclough@apple.com>
Reviewed by Alexey Proskuryakov.
Prevent integer overflow when reallocating storage vector for arrays.
Sunspider reports 1.005x as fast (no change expected).
* kjs/JSArray.cpp:
2008-07-21 Mark Rowe <mrowe@apple.com>
Reviewed by Sam Weinig.
<rdar://problem/6091287> Revamp the handling of CFBundleShortVersionString to be fixed at the major component of the version number.
* Configurations/Version.xcconfig:
* Info.plist:
2008-07-21 Adam Roben <aroben@apple.com>
Add Vector::find
This is a convenience wrapper around std::find.
Reviewed by Anders Carlsson.
* wtf/Vector.h:
2008-07-19 Oliver Hunt <oliver@apple.com>
Reviewed by Cameron Zwarich.
Bug 20104: Exception in tables/mozilla_expected_failures/bugs/bug92868_1.html includes the equals operator in the quoted expression
<https://bugs.webkit.org/show_bug.cgi?id=20104>
To make this correct we make the dot and bracket assign nodes emit the information to indicate
the failure range is the dot/bracket accessor.
* kjs/grammar.y:
2008-07-18 Steve Falkenburg <sfalken@apple.com>
Windows build fix.
* kjs/JSGlobalObjectFunctions.cpp:
(KJS::isStrWhiteSpace):
2008-07-18 Steve Falkenburg <sfalken@apple.com>
Windows build fix.
* kjs/nodes.h:
(KJS::ThrowableExpressionData::ThrowableExpressionData):
2008-07-18 Oliver Hunt <oliver@apple.com>
Reviewed by Cameron Zwarich.
Bug 18774: SQUIRRELFISH: print meaningful error messages <https://bugs.webkit.org/show_bug.cgi?id=18774>
<rdar://problem/5769353> SQUIRRELFISH: JavaScript error messages are missing informative text
Add support for decent error messages in JavaScript. This patch achieves this by providing
ensuring the common errors and exceptions have messages that provide the text of expression
that trigger the exception. In addition it attaches a number of properties to the exception
object detailing where in the source the expression came from.
* JavaScriptCore.exp:
* VM/CodeBlock.cpp:
(KJS::CodeBlock::lineNumberForVPC):
(KJS::CodeBlock::expressionRangeForVPC):
Function to recover the expression range for an instruction
that triggered an exception.
* VM/CodeBlock.h:
(KJS::ExpressionRangeInfo::):
(KJS::CodeBlock::CodeBlock):
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitCall):
(KJS::CodeGenerator::emitCallEval):
Emit call needed to be modified so to place the expression range info internally,
as the CodeGenerator emits the arguments nodes itself, rather than the various call
nodes.
* VM/CodeGenerator.h:
(KJS::CodeGenerator::emitExpressionInfo):
Record the expression range info.
* VM/ExceptionHelpers.cpp:
(KJS::createErrorMessage):
(KJS::createInvalidParamError):
(KJS::createUndefinedVariableError):
(KJS::createNotAConstructorError):
(KJS::createNotAFunctionError):
(KJS::createNotAnObjectErrorStub):
(KJS::createNotAnObjectError):
Rewrite all the code for the error messages so that they make use of the newly available
information.
* VM/ExceptionHelpers.h:
* VM/Machine.cpp:
(KJS::isNotObject): Now needs vPC and codeBlock
(KJS::Machine::throwException):
New logic to handle the NotAnObjectErrorStub and to handle the absurd "no default value" edge case
(KJS::Machine::privateExecute):
* VM/Machine.h:
* kjs/DebuggerCallFrame.cpp:
(KJS::DebuggerCallFrame::evaluate):
* kjs/Error.cpp:
(KJS::Error::create):
* kjs/Error.h:
* kjs/JSGlobalObjectFunctions.cpp:
* kjs/JSImmediate.cpp:
(KJS::JSImmediate::toObject):
(KJS::JSImmediate::prototype):
My changes to the JSNotAnObject constructor needed to be handled here.
* kjs/JSNotAnObject.h:
(KJS::JSNotAnObjectErrorStub::JSNotAnObjectErrorStub):
(KJS::JSNotAnObjectErrorStub::isNull):
(KJS::JSNotAnObjectErrorStub::isNotAnObjectErrorStub):
Added a JSNotAnObjectErrorStub class to ease the handling of toObject failure exceptions,
and potentially allow even more detailed error messages in future.
* kjs/JSObject.h:
* kjs/Parser.h:
(KJS::Parser::parse):
* kjs/SourceRange.h:
* kjs/grammar.y:
Large amounts of position propagation.
* kjs/lexer.cpp:
(KJS::Lexer::Lexer):
(KJS::Lexer::shift):
(KJS::Lexer::lex):
The lexer needed a few changes to be able to correctly track token character positions.
* kjs/lexer.h:
* kjs/nodes.cpp:
(KJS::ThrowableExpressionData::emitThrowError):
(KJS::StatementNode::StatementNode):
(KJS::ResolveNode::emitCode):
(KJS::BracketAccessorNode::emitCode):
(KJS::DotAccessorNode::emitCode):
(KJS::NewExprNode::emitCode):
(KJS::EvalFunctionCallNode::emitCode):
(KJS::FunctionCallValueNode::emitCode):
(KJS::FunctionCallResolveNode::emitCode):
(KJS::FunctionCallBracketNode::emitCode):
(KJS::FunctionCallDotNode::emitCode):
(KJS::PostfixResolveNode::emitCode):
(KJS::PostfixBracketNode::emitCode):
(KJS::PostfixDotNode::emitCode):
(KJS::DeleteResolveNode::emitCode):
(KJS::DeleteBracketNode::emitCode):
(KJS::DeleteDotNode::emitCode):
(KJS::PrefixResolveNode::emitCode):
(KJS::PrefixBracketNode::emitCode):
(KJS::PrefixDotNode::emitCode):
(KJS::ThrowableBinaryOpNode::emitCode):
(KJS::ReadModifyResolveNode::emitCode):
(KJS::AssignResolveNode::emitCode):
(KJS::AssignDotNode::emitCode):
(KJS::ReadModifyDotNode::emitCode):
(KJS::AssignBracketNode::emitCode):
(KJS::ReadModifyBracketNode::emitCode):
(KJS::ForInNode::ForInNode):
(KJS::ForInNode::emitCode):
(KJS::WithNode::emitCode):
(KJS::LabelNode::emitCode):
(KJS::ThrowNode::emitCode):
(KJS::ProgramNode::ProgramNode):
(KJS::ProgramNode::create):
(KJS::EvalNode::generateCode):
(KJS::FunctionBodyNode::create):
(KJS::FunctionBodyNode::generateCode):
(KJS::ProgramNode::generateCode):
All of these methods were handling the position information.
Constructors and create methods were modified to store the information.
All the emitCall implementations listed needed to be updated to actually
record the position information we have so carefully collected.
* kjs/nodes.h:
(KJS::ThrowableExpressionData::ThrowableExpressionData):
(KJS::ThrowableExpressionData::setExceptionSourceRange):
(KJS::ThrowableExpressionData::divot):
(KJS::ThrowableExpressionData::startOffset):
(KJS::ThrowableExpressionData::endOffset):
(KJS::ThrowableSubExpressionData::ThrowableSubExpressionData):
(KJS::ThrowableSubExpressionData::setSubexpressionInfo):
(KJS::ThrowablePrefixedSubExpressionData::ThrowablePrefixedSubExpressionData):
(KJS::ThrowablePrefixedSubExpressionData::setSubexpressionInfo):
ThrowableExpressionData is just a uniform mechanism for storing the position
information.
(KJS::ResolveNode::):
(KJS::PrePostResolveNode::):
(KJS::ThrowableBinaryOpNode::):
(KJS::WithNode::):
2008-07-18 Geoffrey Garen <ggaren@apple.com>
Reviewed by Cameron Zwarich.
Three renames:
"CallTypeNative" => "CallTypeHost"
"code" => "byteCode"
"generatedCode" => "generatedByteCode"
2008-07-18 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Optimized <= for immediate number cases.
SunSpider reports no overall change, but a 10% speedup on access-nsieve.
2008-07-18 Mark Rowe <mrowe@apple.com>
Rubber-stamped by Sam Weinig.
Fix some casts added in a previous build fix to match the style used
throughout WebKit.
* VM/Machine.cpp:
(KJS::Machine::initializeCallFrame):
* VM/Register.h:
(KJS::Register::Register):
2008-07-18 Landry Breuil <landry@openbsd.org>
Bug 19975: [OpenBSD] Patches to enable build of WebKit
<https://bugs.webkit.org/show_bug.cgi?id=19975>
Reviewed by David Kilzer.
Support for OpenBSD, mostly threading and libm tweaks.
* kjs/collector.cpp: #include <pthread.h>
(KJS::currentThreadStackBase): use pthread_stackseg_np() to get stack base
* kjs/config.h: OpenBSD also provides <pthread_np.h>
* wtf/MathExtras.h: #include <sys/types.h> and <machine/ieee.h>
(isfinite), (signbit): as long as we don't have those functions provide fallback implementations
* wtf/Platform.h: Add support for PLATFORM(OPENBSD) and PLATFORM(SPARC64) macro
2008-07-17 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Next step toward putting doubles in registers: Store constant pool
entries as registers, not JSValue*s.
SunSpider reports no change.
2008-07-17 Geoffrey Garen <ggaren@apple.com>
Reviewed by John Sullivan and Oliver Hunt.
A tiny bit of tidying in function call register allocation.
This patch saves one register when invoking a function expression and/or
a new expression that is stored in a temporary.
Since it's just one register, I can't make a testcase for it.
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitCall): No need to ref the function we're calling
or its base. We'd like the call frame to overlap with them, if possible.
op_call will read the function and its base before writing the call frame,
so this is safe.
* kjs/nodes.cpp:
(KJS::NewExprNode::emitCode): No need to ref the function we're new-ing,
for the same reasons stated above.
(KJS::FunctionCallValueNode::emitCode): ditto
2008-07-17 Steve Falkenburg <sfalken@apple.com>
Build fix.
* kjs/InternalFunction.cpp:
2008-07-17 Sam Weinig <sam@webkit.org>
Roll out r35199 as it is causing failures on the PPC build.
2008-07-17 Geoffrey Garen <ggaren@apple.com>
Reviewed by David Kilzer.
Fixed https://bugs.webkit.org/show_bug.cgi?id=20067
Support function.name (Firefox extension)
Pretty straight-forward.
2008-07-17 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Fixed <rdar://problem/6081636> Functions calls use more temporary
registers than necessary
Holding a reference to the last statement result register caused each
successive statement to output its result to an even higher register.
Happily, statements don't actually need to return a result register
at all. I hope to make this clearer in a future cleanup patch,
but this change will fix the major bug for now.
* kjs/nodes.cpp:
(KJS::statementListEmitCode):
2008-07-17 Gavin Barraclough <barraclough@apple.com>
Reviewed by Sam Weinig.
Merge pre&post dot nodes to simplify the parse tree.
Sunspider results show 0.6% progression (no performance change expected).
* kjs/grammar.y:
* kjs/nodes.cpp:
* kjs/nodes.h:
* kjs/nodes2string.cpp:
2008-07-17 Gavin Barraclough <barraclough@apple.com>
Reviewed by Cameron Zwarich.
Merge pre&post resolve nodes to simplify the parse tree.
Sunspider results show no performance change.
* kjs/grammar.y:
* kjs/nodes.cpp:
* kjs/nodes.h:
* kjs/nodes2string.cpp:
2008-07-17 Gavin Barraclough <barraclough@apple.com>
Reviewed by Cameron Zwarich.
Merge logical nodes to simplify the parse tree.
Sunspider results show 0.6% progression (no performance change expected).
* kjs/grammar.y:
* kjs/nodes.cpp:
* kjs/nodes.h:
* kjs/nodes2string.cpp:
2008-07-17 Ariya Hidayat <ariya.hidayat@trolltech.com>
Reviewed by Simon.
Fix MinGW build (broken in r35198) and simplify getLocalTime().
* kjs/DateMath.cpp:
(KJS::getLocalTime):
2008-07-17 Gavin Barraclough <barraclough@apple.com>
Reviewed by Sam Weinig.
Merge pre&post bracket nodes to simplify the parse tree.
Sunspider results show no performance change.
* kjs/grammar.y:
* kjs/nodes.cpp:
* kjs/nodes.h:
* kjs/nodes2string.cpp:
2008-07-17 Ariya Hidayat <ariya.hidayat@trolltech.com>
Reviewed by Simon.
Fix the 32-bit gcc builds, conversion from "long int" to Register is
ambiguous. Explicitly choose the intptr_t constructor.
* VM/Machine.cpp:
(KJS::Machine::initializeCallFrame):
* VM/Register.h:
(KJS::Register::Register):
2008-07-16 Mark Rowe <mrowe@apple.com>
Rubber-stamped by Geoff Garen.
Fix JavaScript in 64-bit by using a pointer-sized integer
type in the Register union. Also includes a rename of
the intType constant to IntType.
* VM/Machine.cpp:
(KJS::Machine::initializeCallFrame):
* VM/Register.h:
(KJS::Register::):
(KJS::Register::Register):
2008-07-17 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
First step toward putting doubles in registers: Turned Register into a
proper abstraction layer. It is no longer possible to cast a Register
to a JSValue*, or a Register& to a JSValue*&, or to access the union
inside a Register directly.
SunSpider reports no change.
In support of this change, I had to make the following mechanical changes
in a lot of places:
1. Clients now use explicit accessors to read data out of Registers, and
implicit copy constructors to write data into registers.
So, assignment that used to look like
x.u.jsValue = y;
now looks like
x = y;
And access that used to look like
x = y.u.jsValue;
now looks like
x = y.jsValue();
2. I made generic flow control specific in opcodes that made their flow
control generic by treating a Register& as a JSValue*&. This had the
added benefit of removing some exception checking branches from immediate
number code.
3. I beefed up PropertySlot to support storing a Register* in a property
slot. For now, only JSVariableObject's symbolTableGet and symbolTablePut
use this functionality, but I expect more clients to use it in the future.
4. I changed ArgList to be a buffer of Registers, not JSValue*'s, and I
changed ArgList iterator clients to iterate Registers, not JSValue*'s.
2008-07-16 Ada Chan <adachan@apple.com>
Fixed build.
* kjs/JSGlobalObject.cpp:
2008-07-16 Kevin McCullough <kmccullough@apple.com>
Reviewed by Sam and Geoff.
<rdar://problem/5958840> Navigating to another page while profiler is
attached results in slow JavaScript for all time.
- The UNLIKELY keeps this from being a sunspider performance regression.
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::~JSGlobalObject): Stop the profiler associated
with this exec state.
2008-07-16 Sam Weinig <sam@webkit.org>
Reviewed by Steve Falkenburg.
Replace adopting UString constructor in favor of explicit
static adopt method.
* API/JSStringRefCF.cpp:
(JSStringCreateWithCFString):
* kjs/StringConstructor.cpp:
(KJS::stringFromCharCode):
* kjs/StringPrototype.cpp:
(KJS::stringProtoFuncToLowerCase):
(KJS::stringProtoFuncToUpperCase):
(KJS::stringProtoFuncToLocaleLowerCase):
(KJS::stringProtoFuncToLocaleUpperCase):
* kjs/ustring.cpp:
(KJS::UString::adopt):
* kjs/ustring.h:
(KJS::UString::UString):
(KJS::UString::~UString):
2008-07-16 Ariya Hidayat <ariya.hidayat@trolltech.com>
Reviewed by Simon.
http://trolltech.com/developer/task-tracker/index_html?method=entry&id=216179
Fix potential crash (on Qt for Windows port) when performing JavaScript date
conversion.
* kjs/DateMath.cpp:
(KJS::getLocalTime): For the Qt port, prefer to use Windows code, i.e.
localtime_s() instead of localtime() since the latter might crash (on Windows)
given a non-sensible, e.g. NaN, argument.
2008-07-16 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Anders and Geoff.
https://bugs.webkit.org/show_bug.cgi?id=20023
Failed assertion in PropertyNameArray.cpp
This is already tested by testapi.
* API/JSObjectRef.cpp: (JSPropertyNameAccumulatorAddName): Add the string to identifier
table to appease PropertyNameArray.
2008-07-16 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff.
Dereference identifiers when deleting a hash table (fixes leaks with private JSGlobalData
objects).
* kjs/JSGlobalData.cpp: (KJS::JSGlobalData::~JSGlobalData):
* kjs/lookup.cpp: (KJS::HashTable::deleteTable):
* kjs/lookup.h:
* kjs/lexer.cpp: (KJS::Lexer::~Lexer)
HashTable cannot have a destructor, because check-for-global-initializers complains about
having a global constructor then.
2008-07-16 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff.
Check pthread_key_create return value.
This check was helpful when debugging a crash in run-webkit-tests --threaded that happened
because JSGlobalData objects were not deleted, and we were running out of pthread keys soon.
It also looks useful for production builds.
* wtf/ThreadSpecific.h: (WTF::::ThreadSpecific):
2008-07-15 Kevin McCullough <kmccullough@apple.com>
Reviewed by Geoff.
Rename pageGroupIdentifier to profileGroup to keep mention of a
pageGroup out of JavaScriptCore.
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::init):
* kjs/JSGlobalObject.h:
(KJS::JSGlobalObject::setProfileGroup):
(KJS::JSGlobalObject::profileGroup):
* profiler/ProfileGenerator.cpp:
(KJS::ProfileGenerator::create):
(KJS::ProfileGenerator::ProfileGenerator):
* profiler/ProfileGenerator.h:
(KJS::ProfileGenerator::profileGroup):
* profiler/Profiler.cpp:
(KJS::Profiler::startProfiling):
(KJS::dispatchFunctionToProfiles):
(KJS::Profiler::willExecute):
(KJS::Profiler::didExecute):
2008-07-14 Mark Rowe <mrowe@apple.com>
Reviewed by Sam Weinig.
Fix https://bugs.webkit.org/show_bug.cgi?id=20037
Bug 20037: GCC 4.2 build broken due to strict aliasing violation.
* kjs/ustring.cpp:
(KJS::UString::Rep::computeHash): Add a version of computeHash that takes a char* and explicit length.
* kjs/ustring.h:
* profiler/CallIdentifier.h:
(WTF::): Use new version of computeHash that takes a char* and explicit length to avoid unsafe aliasing.
2008-07-14 David Hyatt <hyatt@apple.com>
Fix a crashing bug in ListHashSet's -- operator. Make sure that end() can be -- by special-casing the null
position.
Reviewed by Maciej
* wtf/ListHashSet.h:
(WTF::ListHashSetConstIterator::operator--):
2008-07-14 David Hyatt <hyatt@apple.com>
Buidl fix. Make sure the second insertBefore method returns a value.
* wtf/ListHashSet.h:
(WTF::::insertBefore):
2008-07-14 Adam Roben <aroben@apple.com>
Windows build fix
* JavaScriptCore.vcproj/jsc/jsc.vcproj: Added include/pthreads to the
include path.
2008-07-14 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Kevin McCullough.
Make JSGlobalData refcounted in preparation to adding a way to create contexts that share
global data.
* JavaScriptCore.exp:
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::create):
* kjs/JSGlobalData.h:
Made contructor private, and added a static create() method. Made the class inherit from
RefCounted.
* kjs/JSGlobalObject.h:
(KJS::JSGlobalObject::globalData):
JSGlobalData is now owned by JSGlobalObject (except for the shared one, and the common
WebCore one, which are never deleted).
* kjs/Shell.cpp: (main): Create JSGlobalData with create() method.
2008-07-14 Simon Hausmann <hausmann@webkit.org>
Fix the single-threaded build.
* kjs/JSLock.cpp: Removed undeclared registerThread() function.
* kjs/collector.cpp:
(KJS::Heap::registerThread): Added dummy implementation.
2008-07-14 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff Garen.
Eliminate per-thread JavaScript global data instance support and make arbitrary
global data/global object combinations possible.
* kjs/collector.cpp:
(KJS::Heap::Heap): Store a JSGlobalData pointer instead of multiple pointers to its members.
This allows for going from any JS object to its associated global data, currently used in
JSGlobalObject constructor to initialize its JSGlobalData pointer.
(KJS::Heap::registerThread): Changed thread registration data to be per-heap. Previously,
only the shared heap could be used from multiple threads, so it was the only one that needed
thread registration, but now this can happen to any heap.
(KJS::Heap::unregisterThread): Ditto.
(KJS::Heap::markStackObjectsConservatively): Adapt for the above changes.
(KJS::Heap::setGCProtectNeedsLocking): Ditto.
(KJS::Heap::protect): Ditto.
(KJS::Heap::unprotect): Ditto.
(KJS::Heap::collect): Ditto.
(KJS::Heap::globalObjectCount): Use global object list associated with the current heap,
not the late per-thread one.
(KJS::Heap::protectedGlobalObjectCount): Ditto.
* kjs/collector.h:
(KJS::Heap::ThreadRegistrar): Added a helper object that unregisters a thread when it is
destroyed.
* kjs/JSLock.cpp:
(KJS::JSLock::JSLock):
* kjs/JSLock.h:
(KJS::JSLock::JSLock):
Don't use JSLock to implicitly register threads. I've added registerThread() calls to most
places that use JSLock - we cannot guarantee absolute safety unless we always mark all
threads in the process, but these implicit registration calls should cover reasonable usage
scenarios, I hope.
* API/JSBase.cpp:
(JSEvaluateScript): Explicitly register the current thread.
(JSCheckScriptSyntax): Explicitly register the current thread.
(JSGarbageCollect): Changed to use the passed in context. Unfortunately, this creates a race
condition for clients that pass an already released context to JSGarbageCollect - but it is
unlikely to create real life problems.
To maintain compatibility, the shared heap is collected if NULL is passed.
* API/JSContextRef.cpp:
(JSGlobalContextCreate): Use a new syntax for JSGlobalObject allocation.
(JSGlobalContextRetain): Register the thread.
(JSContextGetGlobalObject): Register the thread.
* API/JSObjectRef.cpp:
(JSObjectMake):
(JSObjectMakeFunctionWithCallback):
(JSObjectMakeConstructor):
(JSObjectMakeFunction):
(JSObjectHasProperty):
(JSObjectGetProperty):
(JSObjectSetProperty):
(JSObjectGetPropertyAtIndex):
(JSObjectSetPropertyAtIndex):
(JSObjectDeleteProperty):
(JSObjectCallAsFunction):
(JSObjectCallAsConstructor):
(JSObjectCopyPropertyNames):
(JSPropertyNameAccumulatorAddName):
* API/JSValueRef.cpp:
(JSValueIsEqual):
(JSValueIsInstanceOfConstructor):
(JSValueMakeNumber):
(JSValueMakeString):
(JSValueToNumber):
(JSValueToStringCopy):
(JSValueToObject):
(JSValueProtect):
(JSValueUnprotect):
Register the thread.
* API/JSStringRef.cpp: (JSStringRelease): Changed a comment to not mention per-thread contexts.
* API/JSStringRefCF.cpp: Removed an unnecessary include of JSLock.h.
* JavaScriptCore.exp: Export JSGlobalData constructor/destructor, now that anyone can have
their own instances. Adapt to other changes, too.
* JavaScriptCore.xcodeproj/project.pbxproj: Made ThreadSpecific.h private, as it is now
included by collector.h and is thus needed in other projects.
* kjs/InitializeThreading.cpp: (KJS::initializeThreadingOnce): Don't initialize per-thread
global data, as it no longer exists.
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
(KJS::JSGlobalData::~JSGlobalData):
* kjs/JSGlobalData.h:
Removed support for per-thread instance. Made constructor and destructor public.
* kjs/JSGlobalObject.cpp: (KJS::JSGlobalObject::init): Get to now arbitrary JSGlobalData
via the heap.
(KJS::JSGlobalObject::operator new): Changed ot take JSGlobalDatra pointer.
* kjs/JSGlobalObject.h:
* kjs/Shell.cpp:
(main):
(jscmain):
Changed to maintain a custom JSGlobalData pointer instead of a per-thread one.
2008-07-13 Ada Chan <adachan@apple.com>
Windows build fix: Add wtf/RefCountedLeakCounter to the project.
* JavaScriptCore.vcproj/WTF/WTF.vcproj:
2008-07-12 Jan Michael Alonzo <jmalonzo@webkit.org>
Gtk, Qt and Wx build fix: Add wtf/RefCountedLeakCounter in the
build scripts
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCoreSources.bkl:
2008-07-11 Stephanie Lewis <slewis@apple.com>
Reviewed by Darin Adler and Oliver Hunt.
Refactor RefCounting Leak counting code into a common class.
In order to export the symbols I needed to put the debug defines inside the function names
Before we had a separate channel for each Logging each Leak type. Since the leak channels were only used in one location, and only at quit for simplicity I combined them all into one leak channel.
* JavaScriptCore.exp:
* JavaScriptCore.xcodeproj/project.pbxproj: add new class
* kjs/nodes.cpp: remove old leak counting code
* wtf/RefCountedLeakCounter.cpp: Added. create a common leak counting class
* wtf/RefCountedLeakCounter.h: Added.
2008-07-11 David Hyatt <hyatt@apple.com>
Add an insertBefore method to ListHashSet to allow for insertions in the middle of the list (rather than just
at the end).
Reviewed by Anders
* wtf/ListHashSet.h:
(WTF::::insertBefore):
(WTF::::insertNodeBefore):
2008-07-11 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Adler.
Move call function to CallData.cpp and construct to ConstructData.cpp.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/CallData.cpp: Copied from kjs/JSValue.cpp.
* kjs/ConstructData.cpp: Copied from kjs/JSValue.cpp.
* kjs/JSValue.cpp:
2008-07-10 Mark Rowe <mrowe@apple.com>
Reviewed by Sam Weinig.
Define WEBKIT_VERSION_MIN_REQUIRED=WEBKIT_VERSION_LATEST when building WebKit to ensure that no symbols end up with the weak_import attribute.
* Configurations/Base.xcconfig:
2008-07-10 Mark Rowe <mrowe@apple.com>
Reviewed by Sam Weinig.
Fix the Tiger build by omitting annotations from methods declared in categories when using old versions of GCC.
* API/WebKitAvailability.h:
2008-07-10 Kevin McCullough <kmccullough@apple.com>
Reviewed by Darin.
-Minor cleanup. Renamed callTree() to head() and no longer use m_head
directly but instead keep it private and access via a method().
* profiler/HeavyProfile.cpp:
(KJS::HeavyProfile::HeavyProfile):
(KJS::HeavyProfile::generateHeavyStructure):
(KJS::HeavyProfile::addNode):
* profiler/Profile.h:
(KJS::Profile::head):
* profiler/ProfileGenerator.cpp:
(KJS::ProfileGenerator::ProfileGenerator):
2008-07-10 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Mark Rowe.
Eliminate CollectorHeapIntrospector.
CollectorHeapIntrospector was added primarily in the hopes to improve leaks tool output,
a result that it didn't deliver. Also, it helped by labeling JSC heap regions as reported by
vmmap tool, but at the same time, it made them mislabeled as malloc'd ones - the correct
way to label mapped regions is to use a VM tag.
So, it makes more sense to remove it completely than to make it work with multiple heaps.
* JavaScriptCore.exp:
* JavaScriptCore.xcodeproj/project.pbxproj:
* kjs/AllInOneFile.cpp:
* kjs/InitializeThreading.cpp:
(KJS::initializeThreading):
* kjs/collector.cpp:
* kjs/collector.h:
* kjs/CollectorHeapIntrospector.cpp: Removed.
* kjs/CollectorHeapIntrospector.h: Removed.
2008-07-09 Kevin McCullough <kmccullough@apple.com>
Reviewed by Darin.
<rdar://problem/5951532> JSProfiler: Implement heavy (or bottom-up)
view (19228)
- Implemented the time and call count portionof heavy. Now all that we
need is some UI.
* profiler/CallIdentifier.h: Removed an unused constructor.
* profiler/HeavyProfile.cpp:
(KJS::HeavyProfile::HeavyProfile): Set the initial time of the head
node so that percentages work correctly.
(KJS::HeavyProfile::mergeProfiles): Sum the times and call count of
nodes being merged.
* profiler/ProfileNode.cpp: Set the intital values of time and call
count when copying ProfileNodes.
(KJS::ProfileNode::ProfileNode):
2008-07-10 Jan Michael Alonzo <jmalonzo@webkit.org>
Gtk build fix.
* GNUmakefile.am: Add HeavyProfile.cpp
2008-07-09 Mark Rowe <mrowe@apple.com>
Reviewed by Geoff Garen.
Don't warn about deprecated functions in production builds.
* Configurations/Base.xcconfig:
* Configurations/DebugRelease.xcconfig:
2008-07-09 Darin Adler <darin@apple.com>
* JavaScriptCore.pri: Fix Qt build by adding HeavyProfile.cpp.
2008-07-09 Kevin Ollivier <kevino@theolliviers.com>
wx biuld fix. Add HeavyProfile.cpp to build files.
* JavaScriptCoreSources.bkl:
2008-07-09 Kevin McCullough <kmccullough@apple.com>
- Windows build fix.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
2008-07-09 Kevin McCullough <kmccullough@apple.com>
- Build fix.
* profiler/HeavyProfile.cpp:
(KJS::HeavyProfile::mergeProfiles):
2008-07-09 Kevin McCullough <kmccullough@apple.com>
Reviewed by Geoff and Adam.
<rdar://problem/5951532> JSProfiler: Implement Bottom-Up view (19228)
- This is the plumbing for bottom-up, but does not include calculating
time, mostly because I'm still undclear about what the end result should
look like.
- This, obviously, does not include the UI to expose this in the
inspector yet.
* JavaScriptCore.xcodeproj/project.pbxproj:
* profiler/CallIdentifier.h:
(KJS::CallIdentifier::CallIdentifier):
(WTF::): Added HashTraits for CallIdentifiers to be used by a HashMap.
* profiler/HeavyProfile.cpp: Added.
(KJS::HeavyProfile::HeavyProfile):
(KJS::HeavyProfile::generateHeavyStructure):
(KJS::HeavyProfile::addNode):
(KJS::HeavyProfile::mergeProfiles):
(KJS::HeavyProfile::addAncestorsAsChildren):
* profiler/HeavyProfile.h: Added.
(KJS::HeavyProfile::create):
(KJS::HeavyProfile::heavyProfile):
(KJS::HeavyProfile::treeProfile):
* profiler/Profile.cpp: Removed old commented out includes.
* profiler/Profile.h: The m_head is needed by the HeavyProfile so it
is now protected as opposed to private.
* profiler/ProfileNode.cpp:
(KJS::ProfileNode::ProfileNode): Created a constructor to copy
ProfileNodes.
(KJS::ProfileNode::findChild): Added a null check to make HeavyProfile
children finding easier and avoid a potential crasher.
* profiler/ProfileNode.h: Mostly moved things around but also added some
functionality needed by HeavyProfile.
(KJS::ProfileNode::create):
(KJS::ProfileNode::functionName):
(KJS::ProfileNode::url):
(KJS::ProfileNode::lineNumber):
(KJS::ProfileNode::head):
(KJS::ProfileNode::setHead):
(KJS::ProfileNode::setNextSibling):
(KJS::ProfileNode::actualTotalTime):
(KJS::ProfileNode::actualSelfTime):
* profiler/TreeProfile.cpp: Implemented the ability to get a
HeavyProfile.
(KJS::TreeProfile::heavyProfile):
* profiler/TreeProfile.h:
2008-07-08 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Added support for checking if an object has custom properties in its
property map. WebCore uses this to optimize marking DOM wrappers.
2008-07-08 Simon Hausmann <hausmann@webkit.org>
Prospective Gtk/Wx build fixes, add ProfileGenerator.cpp to the build.
* GNUmakefile.am:
* JavaScriptCoreSources.bkl:
2008-07-08 Simon Hausmann <hausmann@webkit.org>
Fix the Qt build, add ProfileGenerator.cpp to the build.
* JavaScriptCore.pri:
2008-07-07 David Kilzer <ddkilzer@apple.com>
releaseFastMallocFreeMemory() should always be defined
Reviewed by Darin.
* JavaScriptCore.exp: Changed to export C++ binding for
WTF::releaseFastMallocFreeMemory() instead of C binding for
releaseFastMallocFreeMemory().
* wtf/FastMalloc.cpp: Moved definitions of
releaseFastMallocFreeMemory() to be in the WTF namespace
regardless whether FORCE_SYSTEM_MALLOC is defined.
* wtf/FastMalloc.h: Moved releaseFastMallocFreeMemory() from
extern "C" binding to WTF::releaseFastMallocFreeMemory().
2008-07-07 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Geoff.
Bug 19926: URL causes crash within a minute
<https://bugs.webkit.org/show_bug.cgi?id=19926>
Add a check that lastGlobalObject is non-null in Machine::execute()
before copying its globals to the current register file.
In theory, it is possible to make a test case for this, but it will
take a while to get it right.
* VM/Machine.cpp:
(KJS::Machine::execute):
2008-07-07 Darin Adler <darin@apple.com>
Rubber stamped by Adele.
* VM/Machine.cpp:
(KJS::Machine::privateExecute): Fix a typo in a comment.
2008-07-07 Steve Falkenburg <sfalken@apple.com>
Build fixes.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.vcproj/testapi/testapi.vcproj:
2008-07-07 Kevin McCullough <kmccullough@apple.com>
Reviewed by Darin.
When the profiler is running it gathers information and creates a
Profile. After it finishes the Profile can be sorted and have other
data refinements run over it. Both of these were done in the same class
before. Now I split the gathering operations into a new class called
ProfileGenerator.
* JavaScriptCore.xcodeproj/project.pbxproj:
* profiler/Profile.cpp: Removed code related to the gather stage of a
Profile's creation.
(KJS::Profile::create):
(KJS::Profile::Profile):
* profiler/Profile.h: Ditto.
(KJS::Profile::title):
(KJS::Profile::callTree):
(KJS::Profile::setHead):
* profiler/ProfileGenerator.cpp: Added. This is the class that will
handle the stage of creating a Profile. Once the Profile is finished
being created, this class goes away.
(KJS::ProfileGenerator::create):
(KJS::ProfileGenerator::ProfileGenerator):
(KJS::ProfileGenerator::title):
(KJS::ProfileGenerator::willExecute):
(KJS::ProfileGenerator::didExecute):
(KJS::ProfileGenerator::stopProfiling):
(KJS::ProfileGenerator::didFinishAllExecution):
(KJS::ProfileGenerator::removeProfileStart):
(KJS::ProfileGenerator::removeProfileEnd):
* profiler/ProfileGenerator.h: Added.
(KJS::ProfileGenerator::profile):
(KJS::ProfileGenerator::originatingGlobalExec):
(KJS::ProfileGenerator::pageGroupIdentifier):
(KJS::ProfileGenerator::client):
(KJS::ProfileGenerator::stoppedProfiling):
* profiler/Profiler.cpp: Now operates with the ProfileGenerator instead
of the Profile.
(KJS::Profiler::startProfiling):
(KJS::Profiler::stopProfiling):
(KJS::Profiler::didFinishAllExecution): It is here that the Profile is
handed off to its client and the Profile Generator is no longer needed.
(KJS::dispatchFunctionToProfiles):
(KJS::Profiler::willExecute):
(KJS::Profiler::didExecute):
* profiler/Profiler.h: Cleaned up the includes and subsequently the
forward declarations. Also use the new ProfileGenerator.
(KJS::ProfilerClient::~ProfilerClient):
(KJS::Profiler::currentProfiles):
* profiler/TreeProfile.cpp: Use Profile's new interface.
(KJS::TreeProfile::create):
(KJS::TreeProfile::TreeProfile):
* profiler/TreeProfile.h:
2008-07-07 Sam Weinig <sam@webkit.org>
Reviewed by Cameron Zwarich.
Third step in broad cleanup effort.
[ File list elided ]
2008-07-06 Sam Weinig <sam@webkit.org>
Reviewed by Cameron Zwarich.
Second step in broad cleanup effort.
[ File list elided ]
2008-07-05 Sam Weinig <sam@webkit.org>
Reviewed by Cameron Zwarich.
First step in broad cleanup effort.
[ File list elided ]
2008-07-05 Sam Weinig <sam@webkit.org>
Rubber-stamped by Cameron Zwarich.
Rename list.h/cpp to ArgList.h/cpp.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Machine.h:
* kjs/AllInOneFile.cpp:
* kjs/ArgList.cpp: Copied from JavaScriptCore/kjs/list.cpp.
* kjs/ArgList.h: Copied from JavaScriptCore/kjs/list.h.
* kjs/IndexToNameMap.cpp:
* kjs/JSGlobalData.cpp:
* kjs/JSGlobalData.h:
* kjs/JSObject.h:
* kjs/collector.cpp:
* kjs/list.cpp: Removed.
* kjs/list.h: Removed.
2008-07-05 Sam Weinig <sam@webkit.org>
Fix non-AllInOne builds again.
* kjs/BooleanPrototype.cpp:
* kjs/ErrorPrototype.cpp:
* kjs/FunctionPrototype.cpp:
* kjs/NumberPrototype.cpp:
* kjs/ObjectPrototype.cpp:
2008-07-05 Sam Weinig <sam@webkit.org>
Fix build on case-sensitive build systems.
* kjs/IndexToNameMap.cpp:
2008-07-05 Sam Weinig <sam@webkit.org>
Fix build.
* kjs/Arguments.cpp:
* kjs/BooleanPrototype.cpp:
* kjs/DateConstructor.cpp:
* kjs/ErrorPrototype.cpp:
* kjs/FunctionPrototype.cpp:
* kjs/NumberPrototype.cpp:
* kjs/ObjectPrototype.cpp:
* kjs/RegExpPrototype.cpp:
* kjs/StringConstructor.cpp:
* kjs/lookup.cpp:
2008-07-05 Sam Weinig <sam@webkit.org>
Fix non-AllInOne build.
* kjs/JSGlobalObject.cpp:
2008-07-05 Sam Weinig <sam@webkit.org>
Rubber-stamped by Cameron Zwarich.
Split Arguments, IndexToNameMap, PrototypeFunction, GlobalEvalFunction and
the functions on the global object out of JSFunction.h/cpp.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Machine.cpp:
* kjs/AllInOneFile.cpp:
* kjs/Arguments.cpp: Copied from JavaScriptCore/kjs/JSFunction.cpp.
* kjs/Arguments.h: Copied from JavaScriptCore/kjs/JSFunction.h.
* kjs/GlobalEvalFunction.cpp: Copied from JavaScriptCore/kjs/JSFunction.cpp.
* kjs/GlobalEvalFunction.h: Copied from JavaScriptCore/kjs/JSFunction.h.
* kjs/IndexToNameMap.cpp: Copied from JavaScriptCore/kjs/JSFunction.cpp.
* kjs/IndexToNameMap.h: Copied from JavaScriptCore/kjs/JSFunction.h.
* kjs/JSActivation.cpp:
* kjs/JSFunction.cpp:
* kjs/JSFunction.h:
* kjs/JSGlobalObject.cpp:
* kjs/JSGlobalObjectFunctions.cpp: Copied from JavaScriptCore/kjs/JSFunction.cpp.
* kjs/JSGlobalObjectFunctions.h: Copied from JavaScriptCore/kjs/JSFunction.h.
The functions on the global object should be in JSGlobalObject.cpp, but putting them there
was a 0.5% regression.
* kjs/PrototypeFunction.cpp: Copied from JavaScriptCore/kjs/JSFunction.cpp.
* kjs/PrototypeFunction.h: Copied from JavaScriptCore/kjs/JSFunction.h.
* kjs/Shell.cpp:
* kjs/lexer.cpp:
* kjs/ustring.cpp:
2008-07-04 Sam Weinig <sam@webkit.org>
Really fix the mac build.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-04 Sam Weinig <sam@webkit.org>
Fix mac build.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-04 Sam Weinig <sam@webkit.org>
Fix non-AllInOne builds.
* kjs/Error.cpp:
* kjs/GetterSetter.cpp:
* kjs/JSImmediate.cpp:
* kjs/operations.cpp:
2008-07-04 Sam Weinig <sam@webkit.org>
Rubber-stamped by Dan Bernstein.
Split Error and GetterSetter out of JSObject.h.
* API/JSCallbackObjectFunctions.h:
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/ClassInfo.h: Copied from JavaScriptCore/kjs/JSObject.h.
* kjs/Error.cpp: Copied from JavaScriptCore/kjs/JSObject.cpp.
* kjs/Error.h: Copied from JavaScriptCore/kjs/JSObject.h.
* kjs/GetterSetter.cpp:
* kjs/GetterSetter.h: Copied from JavaScriptCore/kjs/JSObject.h.
* kjs/JSObject.cpp:
* kjs/JSObject.h:
* kjs/nodes.h:
2008-07-04 Simon Hausmann <hausmann@webkit.org>
Fix the Wx build, added TreeProfile.cpp to the build.
* JavaScriptCoreSources.bkl:
2008-07-03 Mark Rowe <mrowe@apple.com>
Reviewed by Oliver Hunt.
Fix output path of recently-added script phase to reference the correct file.
This prevents Xcode from running the script phase unnecessarily, which caused
the generated header to be recreated and lead to AllInOneFile.cpp rebuilding.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-03 Mark Rowe <mrowe@apple.com>
Follow-up to the 64-bit build fix. Use intptr_t rather than ssize_t as
the latter is non-standard and does not exist on Windows.
* kjs/JSLock.cpp:
(KJS::JSLock::lockCount):
(KJS::JSLock::lock):
(KJS::JSLock::unlock):
(KJS::JSLock::DropAllLocks::DropAllLocks):
* kjs/JSLock.h:
2008-07-02 Mark Rowe <mrowe@apple.com>
Fix the 64-bit build. pthread_getspecific works with pointer-sized values,
so use ssize_t rather than int to track the lock count to avoid warnings about
truncating the result of pthread_getspecific.
* kjs/JSLock.cpp:
(KJS::JSLock::lockCount):
(KJS::JSLock::lock):
(KJS::JSLock::unlock):
(KJS::JSLock::DropAllLocks::DropAllLocks):
* kjs/JSLock.h:
2008-07-03 Geoffrey Garen <ggaren@apple.com>
Reviewed by Sam Weinig.
Removed checking for the array get/put fast case from the array code.
Callers who want the fast case should call getIndex and/or setIndex
instead. (get_by_val and put_by_val already do this.)
SunSpider reports no change overall, but a 1.4% speedup on fannkuch and
a 3.6% speedup on nsieve.
2008-07-03 Dan Bernstein <mitz@apple.com>
- Windows build fix
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: Added TreeProfile.{h,cpp}.
2008-07-03 Dan Bernstein <mitz@apple.com>
Reviewed by Anders Carlsson.
- Windows build fix
* VM/Machine.cpp:
(KJS::Machine::Machine):
2008-07-03 Simon Hausmann <hausmann@webkit.org>
Reviewed by Alexey Proskuryakov.
Fix the non-threaded build.
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::threadInstanceInternal):
2008-07-03 Simon Hausmann <hausmann@webkit.org>
Fix the Qt build, added TreeProfile to the build.
* JavaScriptCore.pri:
2008-07-02 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff.
Don't create unnecessary JSGlobalData instances.
* kjs/JSGlobalData.h:
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::threadInstanceExists):
(KJS::JSGlobalData::sharedInstanceExists):
(KJS::JSGlobalData::threadInstance):
(KJS::JSGlobalData::sharedInstance):
(KJS::JSGlobalData::threadInstanceInternal):
(KJS::JSGlobalData::sharedInstanceInternal):
Added methods to query instance existence.
* kjs/InitializeThreading.cpp:
(KJS::initializeThreadingOnce):
Initialize thread instance static in a new way.
* API/JSBase.cpp:
(JSGarbageCollect):
* kjs/collector.cpp:
(KJS::Heap::collect):
Check for instance existence before accessing it.
2008-07-02 Geoffrey Garen <ggaren@apple.com>
Reviewed by Cameron Zwarich.
Fixed https://bugs.webkit.org/show_bug.cgi?id=19862
REGRESSION (r34907): Gmail crashes in JavaScriptCore code while editing drafts
I was never able to reproduce this issue, but Cameron could, and he says
that this patch fixes it.
The crash seems tied to a timer or event handler callback. In such a case,
the sole reference to the global object may be in the current call frame,
so we can't depend on the global object to mark the call frame area in
the register file.
The new GC marking rule is: the global object is not responsible for
marking the whole register file -- it's just responsible for the globals
section it's tied to. The heap is responsible for marking the call frame area.
2008-07-02 Mark Rowe <mrowe@apple.com>
Reviewed by Sam Weinig.
Add the ability to trace JavaScriptCore garabge collections using dtrace.
* JavaScriptCore.xcodeproj/project.pbxproj: Generate the dtrace probe header
file when building on a new enough version of Mac OS X.
* JavaScriptCorePrefix.h: Add our standard Mac OS X version detection macros.
* kjs/Tracing.d: Declare three dtrace probes.
* kjs/Tracing.h: Include the generated dtrace macros if dtrace is available,
otherwise provide versions that do nothing.
* kjs/collector.cpp:
(KJS::Heap::collect): Fire dtrace probes when starting a collection, after the
mark phase has completed, and when the collection is complete.
* wtf/Platform.h: Define HAVE_DTRACE when building on a new enough version of Mac OS X.
2008-07-02 Geoffrey Garen <ggaren@apple.com>
Rubber stamped by Oliver Hunt.
Reduced the max register file size from 8MB to 2MB.
We still allow about 20,000 levels of recursion.
2008-07-02 Alp Toker <alp@nuanti.com>
Build fix for r34960. Add TreeProfile.cpp to build.
* GNUmakefile.am:
2008-07-02 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Optimized a[n] get for cases when a is an array or a string. When a is
an array, we optimize both get and put. When a is a string, we only
optimize get, since you can't put to a string.
SunSpider says 3.4% faster.
2008-07-02 Kevin McCullough <kmccullough@apple.com>
Reviewed by Darin.
-Small cleanup in preparation for implementing Bottom-up.
* profiler/CallIdentifier.h: Rename debug function to make it clear of
its output and intention to be debug only.
(KJS::CallIdentifier::operator const char* ): Implement in terms of
c_str.
(KJS::CallIdentifier::c_str):
* profiler/ProfileNode.cpp: Impelment findChild() which will be needed
by the bottom-up implementation.
(KJS::ProfileNode::findChild):
* profiler/ProfileNode.h: Added comments to make the collections of
functions more clear.
(KJS::ProfileNode::operator==):
(KJS::ProfileNode::c_str):
2008-07-02 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Darin.
Bug 19776: Number.toExponential() is incorrect for numbers between 0.1 and 1
<https://bugs.webkit.org/show_bug.cgi?id=19776>
Perform the sign check for the exponent on the actual exponent value,
which is 1 less than the value of decimalPoint, instead of on the value
of decimalPoint itself.
* kjs/NumberPrototype.cpp:
(KJS::exponentialPartToString):
2008-07-02 Kevin McCullough <kmccullough@apple.com>
Reviewed by Darin.
<rdar://problem/5951532> JSProfiler: Implement Bottom-Up view (19228)
- Subclass TreeProfile as I prepare for a HeavyProfile to be comming
later.
* JavaScriptCore.xcodeproj/project.pbxproj:
* profiler/Profile.cpp: By default we create a TreeProfile.
(KJS::Profile::create):
* profiler/Profile.h: Changes to the Profile class to make it amenable
to be inherited from.
(KJS::Profile::~Profile):
* profiler/TreeProfile.cpp: Added.
(KJS::TreeProfile::create):
(KJS::TreeProfile::TreeProfile):
(KJS::TreeProfile::heavyProfile):
* profiler/TreeProfile.h: Added.
(KJS::TreeProfile::treeProfile):
2008-07-02 Kevin McCullough <kmccullough@apple.com>
Reviewed by Dan.
Broke CallIdentifier out into its own file. I did this because it's
going to grow a lot soon and I wanted this to be a separate patch.
* JavaScriptCore.xcodeproj/project.pbxproj:
* profiler/CallIdentifier.h: Added.
(KJS::CallIdentifier::CallIdentifier):
(KJS::CallIdentifier::operator==):
(KJS::CallIdentifier::operator!=):
(KJS::CallIdentifier::operator const char* ):
(KJS::CallIdentifier::toString):
* profiler/ProfileNode.h:
2008-07-02 Simon Hausmann <hausmann@webkit.org>
Build fix. Implemented missing functions for single-threaded build.
* kjs/JSLock.cpp:
(KJS::JSLock::JSLock):
(KJS::JSLock::lock):
(KJS::JSLock::unlock):
(KJS::JSLock::DropAllLocks::DropAllLocks):
2008-07-02 Alexey Proskuryakov <ap@webkit.org>
Another non-AllInOne build fix.
* kjs/JSGlobalObject.cpp: Include JSLock.h here, too.
2008-07-02 Alexey Proskuryakov <ap@webkit.org>
Non-AllInOne build fix.
* kjs/interpreter.cpp: Include JSLock.h.
2008-06-30 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Darin.
Disable JSLock for per-thread contexts.
No change on SunSpider.
* kjs/JSGlobalData.h:
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
(KJS::JSGlobalData::sharedInstance):
Added isSharedInstance as a better way to tell whether the instance is shared (legacy).
* kjs/JSLock.cpp:
(KJS::createJSLockCount):
(KJS::JSLock::lockCount):
(KJS::setLockCount):
(KJS::JSLock::JSLock):
(KJS::JSLock::lock):
(KJS::JSLock::unlock):
(KJS::JSLock::currentThreadIsHoldingLock):
(KJS::JSLock::DropAllLocks::DropAllLocks):
(KJS::JSLock::DropAllLocks::~DropAllLocks):
* kjs/JSLock.h:
(KJS::JSLock::JSLock):
(KJS::JSLock::~JSLock):
Made JSLock and JSLock::DropAllLocks constructors take a parameter to decide whether to
actually lock a mutex, or only to increment recursion count. We cannot turn it into no-op
if we want to keep existing assertions working.
Made recursion count per-thread, now that locks may not lock.
* API/JSBase.cpp:
(JSEvaluateScript): Take JSLock after casting JSContextRef to ExecState* (which doesn't need
locking in any case), so that a decision whether to actually lock can be made.
(JSCheckScriptSyntax): Ditto.
(JSGarbageCollect): Only lock while collecting the shared heap, not the per-thread one.
* API/JSObjectRef.cpp:
(JSClassCreate): Don't lock, as there is no reason to.
(JSClassRetain): Ditto.
(JSClassRelease): Ditto.
(JSPropertyNameArrayRetain): Ditto.
(JSPropertyNameArrayRelease): Only lock while deleting the array, as that may touch
identifier table.
(JSPropertyNameAccumulatorAddName): Adding a string also involves an identifier table
lookup, and possibly modification.
* API/JSStringRef.cpp:
(JSStringCreateWithCharacters):
(JSStringCreateWithUTF8CString):
(JSStringRetain):
(JSStringRelease):
(JSStringGetUTF8CString):
(JSStringIsEqual):
* API/JSStringRefCF.cpp:
(JSStringCreateWithCFString):
JSStringRef operations other than releasing do not need locking.
* VM/Machine.cpp: Don't include unused JSLock.h.
* kjs/CollectorHeapIntrospector.cpp: (KJS::CollectorHeapIntrospector::statistics):
Don't take the lock for real, as heap introspection pauses the process anyway. It seems that
the existing code could cause deadlocks.
* kjs/Shell.cpp:
(functionGC):
(main):
(jscmain):
The test tool uses a per-thread context, so no real locking is required.
* kjs/collector.h:
(KJS::Heap::setGCProtectNeedsLocking): Optionally protect m_protectedValues access with a
per-heap mutex. This is only needed for WebCore Database code, which violates the "no data
migration between threads" by using ProtectedPtr on a background thread.
(KJS::Heap::isShared): Keep a shared flag here, as well.
* kjs/protect.h:
(KJS::::ProtectedPtr):
(KJS::::~ProtectedPtr):
(KJS::::operator):
(KJS::operator==):
(KJS::operator!=):
ProtectedPtr is ony used from WebCore, so it doesn't need to take JSLock. An assertion in
Heap::protect/unprotect guards agains possible future unlocked uses of ProtectedPtr in JSC.
* kjs/collector.cpp:
(KJS::Heap::Heap): Initialize m_isShared.
(KJS::Heap::~Heap): No need to lock for real during destruction, but must keep assertions
in sweep() working.
(KJS::destroyRegisteredThread): Registered thread list is only accessed for shared heap,
so locking is always needed here.
(KJS::Heap::registerThread): Ditto.
(KJS::Heap::markStackObjectsConservatively): Use m_isShared instead of comparing to a shared
instance for a small speedup.
(KJS::Heap::setGCProtectNeedsLocking): Create m_protectedValuesMutex. There is currently no
way to undo this - and ideally, Database code will be fixed to lo longer require this quirk.
(KJS::Heap::protect): Take m_protectedValuesMutex (if it exists) while accessing
m_protectedValues.
(KJS::Heap::unprotect): Ditto.
(KJS::Heap::markProtectedObjects): Ditto.
(KJS::Heap::protectedGlobalObjectCount): Ditto.
(KJS::Heap::protectedObjectCount): Ditto.
(KJS::Heap::protectedObjectTypeCounts): Ditto.
* kjs/ustring.cpp:
* kjs/ustring.h:
Don't include JSLock.h, which is no longer used here. As a result, an explicit include had
to be added to many files in JavaScriptGlue, WebCore and WebKit.
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::init):
* API/JSCallbackConstructor.cpp:
(KJS::constructJSCallback):
* API/JSCallbackFunction.cpp:
(KJS::JSCallbackFunction::call):
* API/JSCallbackObjectFunctions.h:
(KJS::::init):
(KJS::::getOwnPropertySlot):
(KJS::::put):
(KJS::::deleteProperty):
(KJS::::construct):
(KJS::::hasInstance):
(KJS::::call):
(KJS::::getPropertyNames):
(KJS::::toNumber):
(KJS::::toString):
(KJS::::staticValueGetter):
(KJS::::callbackGetter):
* API/JSContextRef.cpp:
(JSGlobalContextCreate):
(JSGlobalContextRetain):
(JSGlobalContextRelease):
* API/JSValueRef.cpp:
(JSValueIsEqual):
(JSValueIsStrictEqual):
(JSValueIsInstanceOfConstructor):
(JSValueMakeNumber):
(JSValueMakeString):
(JSValueToNumber):
(JSValueToStringCopy):
(JSValueToObject):
(JSValueProtect):
(JSValueUnprotect):
* JavaScriptCore.exp:
* kjs/PropertyNameArray.h:
(KJS::PropertyNameArray::globalData):
* kjs/interpreter.cpp:
(KJS::Interpreter::checkSyntax):
(KJS::Interpreter::evaluate):
Pass a parameter to JSLock/JSLock::DropAllLocks to decide whether the lock needs to be taken.
2008-07-01 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Darin.
https://bugs.webkit.org/show_bug.cgi?id=19834
Failed assertion in JavaScriptCore/VM/SegmentedVector.h:82
Creating a global object with a custom prototype resets it twice (wasteful!).
So, addStaticGlobals() was called twice, but JSGlobalObject::reset() didn't reset
the register array.
* kjs/JSGlobalObject.cpp: (KJS::JSGlobalObject::reset): Call setRegisterArray(0, 0).
* kjs/JSVariableObject.h: Changed registerArray to OwnArrayPtr. Also, added private copy
constructor and operator= to ensure that no one attempts to copy this object (for whatever
reason, I couldn't make Noncopyable work).
* kjs/JSGlobalObject.h: (KJS::JSGlobalObject::addStaticGlobals): Allocate registerArray
with new[].
* kjs/JSVariableObject.cpp:
(KJS::JSVariableObject::copyRegisterArray): Allocate registerArray with new[].
(KJS::JSVariableObject::setRegisterArray): Avoid hitting an assertion in OwnArrayPtr when
"changing" the value from 0 to 0.
2008-07-01 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Removed and/or reordered exception checks in array-style a[n] access.
SunSpider says 1.4% faster.
* VM/Machine.cpp:
(KJS::Machine::privateExecute): No need to check for exceptions before
calling toString, toNumber and/or get. If the call ends up being observable
through toString, valueOf, or a getter, we short-circuit it there, instead.
In the op_del_by_val case, I removed the incorrect comment without actually
removing the code, since I didn't want to tempt the GCC fates!
* kjs/JSObject.cpp:
(KJS::callDefaultValueFunction): Added exception check to prevent
toString and valueOf functions from observing execution after an exception
has been thrown. This removes some of the burden of exception checking
from the machine.
(KJS::JSObject::defaultValue): Removed redundant exception check here.
* kjs/PropertySlot.cpp:
(KJS::PropertySlot::functionGetter): Added exception check to prevent
getter functions from observing execution after an exception has been
thrown. This removes some of the burden of exception checking from the
machine.
2008-07-01 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Optimized a[n] get and put for cases where n is an immediate unsigned
value.
SunSpider says 3.5% faster.
2008-07-01 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Darin.
Bug 19844: JavaScript Switch statement modifies "this"
<https://bugs.webkit.org/show_bug.cgi?id=19844>
Use a temporary when generating code for switch clauses to avoid
overwriting 'this' or a local variable.
* kjs/nodes.cpp:
(KJS::CaseBlockNode::emitCodeForBlock):
2008-07-01 Christian Dywan <christian@twotoasts.de>
Gtk+ build fix.
* kjs/list.cpp: Include "JSCell.h"
2008-07-01 Kevin McCullough <kmccullough@apple.com>
Build fix.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-07-01 Dan Bernstein <mitz@apple.com>
Reviewed by Anders Carlsson.
- Mac release build fix
* JavaScriptCore.exp:
2008-07-01 Sam Weinig <sam@webkit.org>
Try and fix mac builds.
* JavaScriptCore.exp:
2008-07-01 Sam Weinig <sam@webkit.org>
Fix non-AllInOne builds.
* kjs/DateMath.cpp:
2008-07-01 Sam Weinig <sam@webkit.org>
Reviewed by Darin Adler.
Split JSCell and JSNumberCell class declarations out of JSValue.h
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/JSPropertyNameIterator.h:
* kjs/AllInOneFile.cpp:
* kjs/JSCell.cpp: Copied from JavaScriptCore/kjs/JSValue.cpp.
* kjs/JSCell.h: Copied from JavaScriptCore/kjs/JSValue.h.
(KJS::JSValue::getJSNumber):
* kjs/JSNumberCell.cpp:
* kjs/JSNumberCell.h: Copied from JavaScriptCore/kjs/JSValue.h.
* kjs/JSObject.h:
* kjs/JSString.cpp:
(KJS::jsString):
(KJS::jsOwnedString):
* kjs/JSString.h:
(KJS::JSValue::toThisJSString):
* kjs/JSValue.cpp:
* kjs/JSValue.h:
2008-07-01 Anders Carlsson <andersca@apple.com>
Build fixes.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* kjs/JSGlobalObject.h:
(KJS::JSGlobalObject::addStaticGlobals):
2008-07-01 Simon Hausmann <hausmann@webkit.org>
Build fix, include OwnPtr.h.
* kjs/RegExpConstructor.h:
2008-06-30 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Fixed a global object leak caused by the switch to one register file.
Don't unconditionally mark the register file, since that logically
makes all global variables GC roots, even when their global object is
no longer reachable.
Instead, make the global object associated with the register file
responsible for marking the register file.
2008-06-30 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
Removed the "registerBase" abstraction. Since the register file never
reallocates, we can keep direct pointers into it, instead of
<registerBase, offset> tuples.
SunSpider says 0.8% faster.
2008-06-30 Oliver Hunt <oliver@apple.com>
Reviewed by NOBODY (build fix).
Fix build by adding all (hopefully) the missing includes.
* kjs/BooleanPrototype.cpp:
* kjs/DateConstructor.cpp:
* kjs/ErrorPrototype.cpp:
* kjs/FunctionPrototype.cpp:
* kjs/NativeErrorConstructor.cpp:
* kjs/NumberPrototype.cpp:
* kjs/ObjectPrototype.cpp:
* kjs/RegExpConstructor.cpp:
* kjs/StringConstructor.cpp:
* kjs/StringPrototype.cpp:
2008-06-30 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Oliver.
Bug 19830: REGRESSION (r34883): Google Reader doesn't show up feed list on sidebar
<https://bugs.webkit.org/show_bug.cgi?id=19830>
Ensure that we do not eliminate a write to a local register when doing
peephole optimizations.
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitJumpIfTrue):
(KJS::CodeGenerator::emitJumpIfFalse):
2008-06-30 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Alder.
Split InternalFunction into its own header file.
* API/JSCallbackFunction.h:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* kjs/ArrayConstructor.h:
* kjs/BooleanConstructor.h:
* kjs/DateConstructor.h:
* kjs/ErrorConstructor.h:
* kjs/FunctionConstructor.h:
* kjs/FunctionPrototype.h:
* kjs/InternalFunction.h: Copied from kjs/JSFunction.h.
* kjs/JSFunction.h:
* kjs/NativeErrorConstructor.h:
* kjs/NumberConstructor.h:
* kjs/ObjectConstructor.h:
* kjs/RegExpConstructor.h:
* kjs/StringConstructor.h:
* profiler/Profiler.cpp:
2008-06-30 Sam Weinig <sam@webkit.org>
Reviewed by Kevin McCullough.
Remove empty files Instruction.cpp, LabelID.cpp, Register.cpp and RegisterID.cpp.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Instruction.cpp: Removed.
* VM/LabelID.cpp: Removed.
* VM/Register.cpp: Removed.
* VM/RegisterID.cpp: Removed.
2008-06-30 Sam Weinig <sam@webkit.org>
Rubber-stamped (reluctantly) by Kevin McCullough.
Rename date_object.h/cpp to DateInstance.h/cpp
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/DateConstructor.cpp:
* kjs/DateInstance.cpp: Copied from kjs/date_object.cpp.
* kjs/DateInstance.h: Copied from kjs/date_object.h.
* kjs/DatePrototype.cpp:
* kjs/DatePrototype.h:
* kjs/date_object.cpp: Removed.
* kjs/date_object.h: Removed.
2008-06-30 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Adler.
Remove internal.cpp and move its contents to there own .cpp files.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/GetterSetter.cpp: Copied from kjs/internal.cpp.
* kjs/InternalFunction.cpp: Copied from kjs/internal.cpp.
* kjs/JSNumberCell.cpp: Copied from kjs/internal.cpp.
* kjs/JSString.cpp: Copied from kjs/internal.cpp.
* kjs/JSString.h:
* kjs/LabelStack.cpp: Copied from kjs/internal.cpp.
* kjs/NumberConstructor.cpp:
* kjs/NumberObject.cpp:
(KJS::constructNumber):
(KJS::constructNumberFromImmediateNumber):
* kjs/internal.cpp: Removed.
2008-06-30 Adam Roben <aroben@apple.com>
Fix <rdar://5954749> Assertion failure due to HashTable's use of
operator&
HashTable was passing &value to constructDeletedValue, which in
classes like WebCore::COMPtr would cause an assertion. We now pass
value by reference instead of by address so that the HashTraits
implementations have more flexibility in constructing the deleted
value.
Reviewed by Ada Chan.
* VM/CodeGenerator.h: Updated for changes to HashTraits.
* wtf/HashTable.h:
(WTF::::deleteBucket): Changed to pass bucket by reference instead of
by address.
(WTF::::checkKey): Ditto.
* wtf/HashTraits.h:
(WTF::): Updated HashTraits for HashTable change.
2008-07-01 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Cameron Zwarich.
Make RegisterFile really unmap memory on destruction.
This fixes run-webkit-tests --threaded, which ran out of address space in a few seconds.
* VM/RegisterFile.cpp: (KJS::RegisterFile::~RegisterFile): Unmap all the memory, not just
1/4 of it.
* kjs/JSGlobalObject.h: Don't include RegisterFile.h, so that changes to it don't make
half of WebCore rebuild.
* VM/Machine.h: Don't forward declare RegisterFile, as RegisterFile.h is included already.
* VM/RegisterFile.h: (KJS::RegisterFile::RegisterFile): Assert that the allocation succeeded.
2008-06-30 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Rubber-stamped by Oliver.
Correct the documentation for op_put_by_index.
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-06-29 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Oliver.
Bug 19821: Merge the instruction pair (less, jfalse)
<https://bugs.webkit.org/show_bug.cgi?id=19821>
This is a 2.4% win on SunSpider. I needed to add an ALWAYS_INLINE
intrinisc to CodeGenerator::rewindBinaryOp() to avoid a massive
regression in regexp-dna.
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::rewindBinaryOp):
(KJS::CodeGenerator::emitJumpIfFalse):
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
* VM/Opcode.cpp:
(KJS::):
* VM/Opcode.h:
2008-06-29 Sam Weinig <sam@webkit.org>
Fix non-AllInOne builds.
* kjs/JSObject.cpp:
* kjs/JSValue.cpp:
2008-06-29 Sam Weinig <sam@webkit.org>
Build fix for Qt.
* kjs/DateMath.cpp:
* kjs/DatePrototype.cpp:
2008-06-29 Sam Weinig <sam@webkit.org>
Rubber-stamped by Cameron Zwarich.
Splits ErrorConstructor, ErrorPrototype, NativeErrorConstructor and
NativeErrorPrototype out of error_object.h/cpp and renames it ErrorInstance.
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/ArrayConstructor.cpp:
* kjs/ArrayPrototype.cpp:
* kjs/BooleanPrototype.cpp:
* kjs/DatePrototype.cpp:
* kjs/ErrorConstructor.cpp: Copied from kjs/error_object.cpp.
* kjs/ErrorConstructor.h: Copied from kjs/error_object.h.
* kjs/ErrorInstance.cpp: Copied from kjs/error_object.cpp.
* kjs/ErrorInstance.h: Copied from kjs/error_object.h.
* kjs/ErrorPrototype.cpp: Copied from kjs/error_object.cpp.
* kjs/ErrorPrototype.h: Copied from kjs/error_object.h.
* kjs/JSGlobalObject.cpp:
* kjs/JSObject.cpp:
* kjs/JSValue.cpp:
* kjs/NativeErrorConstructor.cpp: Copied from kjs/error_object.cpp.
* kjs/NativeErrorConstructor.h: Copied from kjs/error_object.h.
* kjs/NativeErrorPrototype.cpp: Copied from kjs/error_object.cpp.
* kjs/NativeErrorPrototype.h: Copied from kjs/error_object.h.
* kjs/NumberPrototype.cpp:
* kjs/RegExpConstructor.cpp:
* kjs/RegExpObject.cpp:
* kjs/RegExpPrototype.cpp:
* kjs/StringPrototype.cpp:
* kjs/error_object.cpp: Removed.
* kjs/error_object.h: Removed.
* kjs/internal.cpp:
2008-06-29 Sam Weinig <sam@webkit.org>
Fix non-AllInOne build.
* kjs/DateConstructor.cpp:
* kjs/DateMath.cpp:
* kjs/JSObject.cpp:
2008-06-29 Sam Weinig <sam@webkit.org>
Rubber-stamped by Oliver Hunt.
Splits DateConstructor and DatePrototype out of date_object.h/cpp
Moves shared Date code into DateMath.
* DerivedSources.make:
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/DateConstructor.cpp: Copied from kjs/date_object.cpp.
* kjs/DateConstructor.h: Copied from kjs/date_object.h.
* kjs/DateMath.cpp:
(KJS::ymdhmsToSeconds):
(KJS::):
(KJS::skipSpacesAndComments):
(KJS::findMonth):
(KJS::parseDate):
(KJS::timeClip):
(KJS::formatDate):
(KJS::formatDateUTCVariant):
(KJS::formatTime):
* kjs/DateMath.h:
(KJS::gmtoffset):
* kjs/DatePrototype.cpp: Copied from kjs/date_object.cpp.
* kjs/DatePrototype.h: Copied from kjs/date_object.h.
* kjs/JSGlobalObject.cpp:
* kjs/JSObject.cpp:
* kjs/date_object.cpp:
* kjs/date_object.h:
* kjs/internal.cpp:
2008-06-29 Jan Michael Alonzo <jmalonzo@webkit.org>
Rubber-stamped by Cameron Zwarich
Fix Gtk non-AllInOne build
* GNUmakefile.am: include JSVariableObject.cpp
* kjs/RegExpConstructor.cpp: include RegExpObject.h
* kjs/RegExpObject.h: forward declare RegExpPrototype
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Sam and Cameron.
- fix https://bugs.webkit.org/show_bug.cgi?id=19805
Array.concat turns missing array elements into "undefined"
Test: fast/js/array-holes.html
* JavaScriptCore.exp: No longer export JSArray::getItem.
* kjs/ArrayPrototype.cpp:
(KJS::arrayProtoFuncConcat): Changed to use getProperty instead of
JSArray::getItem -- need to handle properties from the prototype chain
instead of ignoring them.
* kjs/JSArray.cpp: Removed getItem.
* kjs/JSArray.h: Ditto.
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Cameron.
- https://bugs.webkit.org/show_bug.cgi?id=19804
optimize access to arrays without "holes"
SunSpider says 1.8% faster.
* kjs/JSArray.cpp:
(KJS::JSArray::JSArray): Initialize m_fastAccessCutoff when creating
arrays. Also updated for new location of m_vectorLength.
(KJS::JSArray::getItem): Updated for new location of m_vectorLength.
(KJS::JSArray::getSlowCase): Added. Broke out the non-hot parts of
getOwnPropertySlot to make the hot part faster.
(KJS::JSArray::getOwnPropertySlot): Added a new faster case for
indices lower than m_fastAccessCutoff. We can do theese with no
additional checks or branches.
(KJS::JSArray::put): Added a new faster case for indices lower than
m_fastAccessCutoff. We can do theese with no additional checks or
branches. Moved the maxArrayIndex handling out of this function.
Added code to set m_fastAccessCutoff when the very last hole in
an array is filled; this is how the cutoff gets set for most arrays.
(KJS::JSArray::putSlowCase): Moved the rest of the put function logic
in here, to make the hot part of the put function faster.
(KJS::JSArray::deleteProperty): Added code to lower m_fastAccessCutoff
when a delete makes a new hole in the array.
(KJS::JSArray::getPropertyNames): Updated for new location of
m_vectorLength.
(KJS::JSArray::increaseVectorLength): Ditto.
(KJS::JSArray::setLength): Added code to lower m_fastAccessCutoff
when setLength makes the array smaller.
(KJS::JSArray::mark): Updated for new location of m_vectorLength.
(KJS::JSArray::sort): Ditto. Set m_fastAccessCutoff after moving
all the holes to the end of the array.
(KJS::JSArray::compactForSorting): Ditto.
(KJS::JSArray::checkConsistency): Added consistency checks fro
m_fastAccessCutoff and updated for the new location of m_vectorLength.
* kjs/JSArray.h: Added declarations for slow case functions.
Replaced m_vectorLength with m_fastAccessCutoff.
2008-06-28 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Sam.
When executing a native call, check for an exception before writing the
return value.
* VM/Machine.cpp:
(KJS::Machine::privateExecute):
2008-06-28 Mark Rowe <mrowe@apple.com>
Build fix. Flag headers as private or public as is appropriate.
These settings were accidentally removed during some project file cleanup.
* JavaScriptCore.xcodeproj/project.pbxproj:
2008-06-28 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Adler.
Splits RegExpConstructor and RegExpPrototype out of RegExpObject.h/cpp
* DerivedSources.make:
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Machine.cpp:
* kjs/AllInOneFile.cpp:
* kjs/JSGlobalObject.cpp:
* kjs/RegExpConstructor.cpp: Copied from kjs/RegExpObject.cpp.
* kjs/RegExpConstructor.h: Copied from kjs/RegExpObject.h.
* kjs/RegExpObject.cpp:
* kjs/RegExpObject.h:
* kjs/RegExpPrototype.cpp: Copied from kjs/RegExpObject.cpp.
* kjs/RegExpPrototype.h: Copied from kjs/RegExpObject.h.
* kjs/StringPrototype.cpp:
* kjs/internal.cpp:
2008-06-28 Sam Weinig <sam@webkit.org>
Fix non-AllInOne builds.
* kjs/StringConstructor.cpp:
2008-06-28 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Adler.
Rename string_object.h/cpp to StringObject.h/cpp and split out StringObjectThatMasqueradesAsUndefined,
StringConstructor and StringPrototype.
* DerivedSources.make:
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* kjs/AllInOneFile.cpp:
* kjs/JSGlobalObject.cpp:
* kjs/StringConstructor.cpp: Copied from JavaScriptCore/kjs/string_object.cpp.
* kjs/StringConstructor.h: Copied from JavaScriptCore/kjs/string_object.h.
* kjs/StringObject.cpp: Copied from JavaScriptCore/kjs/string_object.cpp.
* kjs/StringObject.h: Copied from JavaScriptCore/kjs/string_object.h.
* kjs/StringObjectThatMasqueradesAsUndefined.h: Copied from JavaScriptCore/kjs/string_object.h.
* kjs/StringPrototype.cpp: Copied from JavaScriptCore/kjs/string_object.cpp.
* kjs/StringPrototype.h: Copied from JavaScriptCore/kjs/string_object.h.
* kjs/internal.cpp:
* kjs/string_object.cpp: Removed.
* kjs/string_object.h: Removed.
2008-06-28 Jan Michael Alonzo <jmalonzo@webkit.org>
Gtk build fix: JSVariableObject is now part of AllInOne
* GNUmakefile.am:
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Oliver.
- https://bugs.webkit.org/show_bug.cgi?id=19801
add a feature so we can tell what regular expressions are taking time
* pcre/pcre_compile.cpp:
(jsRegExpCompile): Compile in the string if REGEXP_HISTOGRAM is on.
* pcre/pcre_exec.cpp:
(jsRegExpExecute): Add hook to time execution.
(Histogram::~Histogram): Print a sorted list of what took time.
(Histogram::add): Accumulate records of what took time.
(HistogramTimeLogger::~HistogramTimeLogger): Hook that calls
Histogram::add at the right moment and creates the global histogram
object.
* pcre/pcre_internal.h: Define REGEXP_HISTOGRAM.
* pcre/pcre_tables.cpp: Added missing include of "config.h". Not needed
any more, but an omissions an earlier version of this patch detected.
* pcre/pcre_ucp_searchfuncs.cpp: Ditto.
* pcre/pcre_xclass.cpp: Ditto.
2008-06-28 Sam Weinig <sam@webkit.org>
Try and fix the Windows build again.
* kjs/RegExpObject.cpp:
* kjs/date_object.cpp:
* kjs/error_object.cpp:
2008-06-28 Sam Weinig <sam@webkit.org>
Rubber-stamped by Darin Adler.
Remove unused StringConstructorFunction class.
* kjs/string_object.h:
2008-06-28 Sam Weinig <sam@webkit.org>
Fix windows build.
* kjs/ArrayPrototype.cpp:
* kjs/BooleanPrototype.cpp:
* kjs/BooleanPrototype.h:
* kjs/FunctionPrototype.cpp:
* kjs/JSImmediate.cpp:
* kjs/JSObject.cpp:
* kjs/MathObject.cpp:
* kjs/NumberPrototype.cpp:
* kjs/NumberPrototype.h:
* kjs/ObjectConstructor.cpp:
* kjs/RegExpObject.h:
* kjs/error_object.h:
* kjs/string_object.cpp:
2008-06-28 Sam Weinig <sam@webkit.org>
Rubber-stamped by Oliver Hunt.
Splits FunctionConstructor out of FunctionPrototype.h/cpp
Splits NumberConstructor and NumberPrototype out of NumberObject.h/cpp
Rename object_object.h/cpp to ObjectPrototype.h/cpp and split out ObjectConstructor.
* API/JSCallbackConstructor.cpp:
* API/JSClassRef.cpp:
* API/JSObjectRef.cpp:
* DerivedSources.make:
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Machine.cpp:
* kjs/AllInOneFile.cpp:
* kjs/ArrayConstructor.cpp:
* kjs/ArrayConstructor.h:
* kjs/FunctionConstructor.cpp: Copied from JavaScriptCore/kjs/FunctionPrototype.cpp.
* kjs/FunctionConstructor.h: Copied from JavaScriptCore/kjs/FunctionPrototype.h.
* kjs/FunctionPrototype.cpp:
* kjs/FunctionPrototype.h:
* kjs/JSFunction.cpp:
* kjs/JSGlobalObject.cpp:
* kjs/JSImmediate.cpp:
* kjs/MathObject.h:
* kjs/NumberConstructor.cpp: Copied from JavaScriptCore/kjs/NumberObject.cpp.
* kjs/NumberConstructor.h: Copied from JavaScriptCore/kjs/NumberObject.h.
* kjs/NumberObject.cpp:
* kjs/NumberObject.h:
* kjs/NumberPrototype.cpp: Copied from JavaScriptCore/kjs/NumberObject.cpp.
* kjs/NumberPrototype.h: Copied from JavaScriptCore/kjs/NumberObject.h.
* kjs/ObjectConstructor.cpp: Copied from JavaScriptCore/kjs/object_object.cpp.
* kjs/ObjectConstructor.h: Copied from JavaScriptCore/kjs/object_object.h.
* kjs/ObjectPrototype.cpp: Copied from JavaScriptCore/kjs/object_object.cpp.
* kjs/ObjectPrototype.h: Copied from JavaScriptCore/kjs/object_object.h.
* kjs/RegExpObject.h:
* kjs/Shell.cpp:
* kjs/error_object.h:
* kjs/internal.cpp:
* kjs/nodes.cpp:
* kjs/object_object.cpp: Removed.
* kjs/object_object.h: Removed.
* kjs/string_object.h:
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Oliver.
- fix https://bugs.webkit.org/show_bug.cgi?id=19796
optimize expressions with ignored results (especially post-increment)
SunSpider says 0.9% faster.
* VM/CodeGenerator.h:
(KJS::CodeGenerator::tempDestination): Create a new temporary for
ignoredResult() too, just as we would for 0.
(KJS::CodeGenerator::finalDestination): Use the temporary if the
register passed in is ignoredResult() too, just as we would for 0.
(KJS::CodeGenerator::destinationForAssignResult): Return 0 if the
passed in register is ignoredResult(), just as we would for 0.
(KJS::CodeGenerator::moveToDestinationIfNeeded): Return 0 if the
register passed in is ignoredResult(). What matters is that we
don't want to emit a move. The return value won't be looked at.
(KJS::CodeGenerator::emitNode): Allow ignoredResult() and pass it
through to the node's emitCode function.
* VM/RegisterID.h:
(KJS::ignoredResult): Added. Special value to indicate the result of
a node will be ignored and need not be put in any register.
* kjs/nodes.cpp:
(KJS::NullNode::emitCode): Do nothing if dst == ignoredResult().
(KJS::BooleanNode::emitCode): Ditto.
(KJS::NumberNode::emitCode): Ditto.
(KJS::StringNode::emitCode): Ditto.
(KJS::RegExpNode::emitCode): Ditto.
(KJS::ThisNode::emitCode): Ditto.
(KJS::ResolveNode::emitCode): Do nothing if dst == ignoredResult() and
the identifier resolves to a local variable.
(KJS::ObjectLiteralNode::emitCode): Do nothing if dst == ignoredResult()
and the object is empty.
(KJS::PostIncResolveNode::emitCode): If dst == ignoredResult(), then do
nothing for the local constant case, and do a pre-increment in all the
other cases.
(KJS::PostDecResolveNode::emitCode): Ditto.
(KJS::PostIncBracketNode::emitCode): Ditto.
(KJS::PostDecBracketNode::emitCode): Ditto.
(KJS::PostIncDotNode::emitCode): Ditto.
(KJS::PostDecDotNode::emitCode): Ditto.
(KJS::DeleteValueNode::emitCode): Pass ignoredResult() when evaluating
the expression.
(KJS::VoidNode::emitCode): Ditto.
(KJS::TypeOfResolveNode::emitCode): If dst == ignoredResult(), do nothing
if the identifier resolves to a local variable, and don't bother generating
a typeof opcode in the other case.
(KJS::TypeOfValueNode::emitCode): Ditto.
(KJS::PreIncResolveNode::emitCode): Do nothing if dst == ignoredResult() and
the identifier resolves to a local constant.
(KJS::PreDecResolveNode::emitCode): Ditto.
(KJS::AssignResolveNode::emitCode): Turn ignoredResult() into 0 in a couple
places, because we need to put the result into a register so we can assign
it. At other sites this is taken care of by functions like finalDestination.
(KJS::CommaNode::emitCode): Pass ignoredResult() when evaluating the first
expression.
(KJS::ForNode::emitCode): Pass ignoredResult() when evaluating the first and
third expressions.
(KJS::ForInNode::emitCode): Pass ignoredResult() when evaluating the first
expression.
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Oliver.
- https://bugs.webkit.org/show_bug.cgi?id=19787
create most arrays from values in registers rather than with multiple put operations
SunSpider says 0.8% faster.
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump): Added argv and argc parameters to new_array.
* VM/Machine.cpp:
(KJS::Machine::privateExecute): Ditto.
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitNewArray): Added.
* VM/CodeGenerator.h: Added ElementNode* argument to emitNewArray.
* kjs/nodes.cpp:
(KJS::ArrayNode::emitCode): Pass the ElementNode to emitNewArray so it can be
initialized with as many elements as possible. If the array doesn't have any
holes in it, that's all that's needed. If there are holes, then emit some separate
put operations for the other values in the array and for the length as needed.
* kjs/nodes.h: Added some accessors to ElementNode so the code generator can
iterate through elements and generate code to evaluate them. Now ArrayNode does
not need to be a friend. Also took out some unused PlacementNewAdoptType
constructors.
2008-06-28 Darin Adler <darin@apple.com>
Reviewed by Oliver.
* kjs/nodes.h: Remove obsolete PlacementNewAdopt constructors.
We no longer mutate the AST in place.
2008-06-28 Jan Michael Alonzo <jmalonzo@webkit.org>
Reviewed by Oliver Hunt.
Build fix
* VM/Machine.cpp: include stdio.h for printf
2008-06-27 Sam Weinig <sam@webkit.org>
Reviewed by Oliver Hunt.
Fix platforms that don't use AllInOne.cpp
* kjs/BooleanConstructor.h:
* kjs/BooleanPrototype.h:
* kjs/FunctionPrototype.cpp:
2008-06-27 Sam Weinig <sam@webkit.org>
Rubber-stamped by Oliver Hunt.
Splits ArrayConstructor out of ArrayPrototype.h/cpp
Splits BooleanConstructor and BooleanPrototype out of BooleanObject.h/cpp
* GNUmakefile.am:
* JavaScriptCore.pri:
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* JavaScriptCore.xcodeproj/project.pbxproj:
* JavaScriptCoreSources.bkl:
* VM/Machine.cpp:
* kjs/AllInOneFile.cpp:
* kjs/ArrayConstructor.cpp: Copied from kjs/ArrayPrototype.cpp.
* kjs/ArrayConstructor.h: Copied from kjs/ArrayPrototype.h.
* kjs/ArrayPrototype.cpp:
* kjs/ArrayPrototype.h:
* kjs/BooleanConstructor.cpp: Copied from kjs/BooleanObject.cpp.
* kjs/BooleanConstructor.h: Copied from kjs/BooleanObject.h.
* kjs/BooleanObject.cpp:
* kjs/BooleanObject.h:
* kjs/BooleanPrototype.cpp: Copied from kjs/BooleanObject.cpp.
* kjs/BooleanPrototype.h: Copied from kjs/BooleanObject.h.
* kjs/CommonIdentifiers.h:
* kjs/FunctionPrototype.cpp:
* kjs/JSArray.cpp:
* kjs/JSGlobalObject.cpp:
* kjs/JSImmediate.cpp:
* kjs/Shell.cpp:
* kjs/internal.cpp:
* kjs/nodes.cpp:
* kjs/string_object.cpp:
2008-06-27 Oliver Hunt <oliver@apple.com>
Reviewed by Sam.
Bug 18626: SQUIRRELFISH: support the "slow script" dialog <https://bugs.webkit.org/show_bug.cgi?id=18626>
<rdar://problem/5973931> Slow script dialog needs to be reimplemented for squirrelfish
Adds support for the slow script dialog in squirrelfish. This requires the addition
of three new op codes, op_loop, op_loop_if_true, and op_loop_if_less which have the
same behaviour as their simple jump equivalents but have an additional time out check.
Additional assertions were added to other jump instructions to prevent accidentally
creating loops with jump types that do not support time out checks.
Sunspider does not report a regression, however this appears very sensitive to code
layout and hardware, so i would expect up to a 1% regression on other systems.
Part of this required moving the old timeout logic from JSGlobalObject and into Machine
which is the cause of a number of the larger diff blocks.
* JavaScriptCore.exp:
* VM/CodeBlock.cpp:
(KJS::CodeBlock::dump):
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitJumpIfTrue):
(KJS::CodeGenerator::emitJumpScopes):
* VM/ExceptionHelpers.cpp:
(KJS::InterruptedExecutionError::isWatchdogException):
(KJS::createInterruptedExecutionException):
* VM/ExceptionHelpers.h:
* VM/LabelID.h:
* VM/Machine.cpp:
(KJS::Machine::Machine):
(KJS::Machine::throwException):
(KJS::Machine::resetTimeoutCheck):
(KJS::getCurrentTime):
(KJS::Machine::checkTimeout):
(KJS::Machine::privateExecute):
* VM/Machine.h:
(KJS::Machine::setTimeoutTime):
(KJS::Machine::startTimeoutCheck):
(KJS::Machine::stopTimeoutCheck):
(KJS::Machine::initTimeout):
* VM/Opcode.cpp:
(KJS::):
* VM/Opcode.h:
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::init):
(KJS::JSGlobalObject::setTimeoutTime):
(KJS::JSGlobalObject::startTimeoutCheck):
* kjs/JSGlobalObject.h:
* kjs/JSObject.h:
* kjs/interpreter.cpp:
(KJS::Interpreter::evaluate):
2008-06-27 Jan Michael Alonzo <jmalonzo@webkit.org>
Gtk and Qt build fix: Remove RegisterFileStack from the build
scripts.
* GNUmakefile.am:
* JavaScriptCore.pri:
2008-06-27 Adele Peterson <adele@apple.com>
Reviewed by Geoff.
Build fixes.
* JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
* VM/RegisterFile.h:
(KJS::RegisterFile::RegisterFile):
* kjs/JSGlobalObject.cpp:
* kjs/collector.cpp:
2008-06-27 Geoffrey Garen <ggaren@apple.com>
Reviewed by Oliver Hunt.
One RegisterFile to rule them all!
SunSpider reports a 0.2% speedup.
This patch removes the RegisterFileStack abstraction and replaces it with
a single register file that
(a) allocates a fixed storage area, including a fixed area for global
vars, so that no operation may cause the register file to reallocate
and
(b) swaps between global storage areas when executing code in different
global objects.
This patch also changes the layout of the register file so that all call
frames, including call frames for global code, get a header. This is
required to support re-entrant global code. It also just makes things simpler.
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::addGlobalVar): New function. Differs from addVar in
that
(a) global vars don't contribute to a CodeBlock's numLocals count, since
global storage is fixed and allocated at startup
and
(b) references to global vars get shifted to elide intermediate stack
between "r" and the global storage area.
* VM/Machine.cpp:
(KJS::Machine::dumpRegisters): Updated this function to match the new
register file layout, and added the ability to dump exact identifiers
for the different parts of a call frame.
(KJS::Machine::unwindCallFrame): Updated this function to match the new
register file layout.
(KJS::Machine::execute): Updated this function to initialize a call frame
header for global code, and to swap global storage areas when switching
to execution in a new global object.
(KJS::Machine::privateExecute): Got rid of "safeForReentry" and re-reading
of registerBase because the register file is always safe for reentry now,
and registerBase never changes.
* VM/Machine.h: Moved the call frame header enum from Machine to RegisterFile,
to resolve a header dependency problem (a good sign that the enum belonged
in RegisterFile all along!)
* VM/RegisterFile.cpp:
* VM/RegisterFile.h: Changed RegisterFile to mmap a fixed size register
area. This allows us to avoid re-allocting the register file later on.
Instead, we rely on the OS to allocate physical pages to the register
file as necessary.
* VM/RegisterFileStack.cpp: Removed. Tada!
* VM/RegisterFileStack.h: Removed. Tada!
* kjs/DebuggerCallFrame.cpp: Updated this class to match the new
register file layout, greatly simplifying it in the process.
* kjs/JSActivation.h:
* kjs/JSActivation.cpp: Moved some of this logic up to JSVariableObject,
since the global object now needs to be able to tear off its registers
just like the activation object.
* kjs/JSFunction.cpp: No need to fiddle with the register file anymore.
* kjs/JSGlobalObject.h:
* kjs/JSGlobalObject.cpp: Updated JSGlobalObject to support moving its
global storage area into and out of the register file.
* kjs/PropertySlot.cpp: No need to fiddle with the register file anymore.
* kjs/collector.cpp: Renamed markStackObjectConservatively to
markConservatively, since we don't just mark stack objects this way.
Also, added code to mark the machine's register file.
* kjs/config.h: Moved some platforms #defines from here...
* wtf/Platform.h: ...to here, to support mmap/VirtualAlloc detection
in RegisterFile.h.
2008-06-26 Mark Rowe <mrowe@apple.com>
Speculative fix for the Windows build.
* kjs/JSImmediate.cpp:
2008-06-26 Mark Rowe <mrowe@apple.com>
Reviewed by Darin Adler and Geoff Garen.
Fix the malloc zone introspection functions so that malloc_zone_statistics does not give
bogus output in an application that uses JavaScriptCore.
* kjs/CollectorHeapIntrospector.cpp:
(KJS::CollectorHeapIntrospector::statistics): Return statistics about memory allocated by the collector.
* kjs/CollectorHeapIntrospector.h:
* wtf/FastMalloc.cpp: Zero out the statistics. FastMalloc doesn't track this information at present.
Returning zero for all values is preferable to returning bogus data.
2008-06-26 Darin Adler <darin@apple.com>
Reviewed by Geoff.
- https://bugs.webkit.org/show_bug.cgi?id=19721
speed up JavaScriptCore by not wrapping strings in objects just
to call functions on them
- optimize UString append and the replace function a bit
SunSpider says 1.8% faster.
* JavaScriptCore.exp: Updated.
* VM/JSPropertyNameIterator.cpp: Added include of JSString.h, now needed
because jsString returns a JSString*.
* VM/Machine.cpp:
(KJS::Machine::privateExecute): Removed the toObject call from native
function calls. Also removed code to put the this value into a register.
* kjs/BooleanObject.cpp:
(KJS::booleanProtoFuncToString): Rewrite to handle false and true
separately.
* kjs/FunctionPrototype.cpp:
(KJS::constructFunction): Use single-character append rather than building
a string for each character.
* kjs/JSFunction.cpp:
(KJS::globalFuncUnescape): Ditto.
* kjs/JSImmediate.cpp:
(KJS::JSImmediate::prototype): Added. Gets the appropriate prototype for
use with an immediate value. To be used instead of toObject when doing a
get on an immediate value.
* kjs/JSImmediate.h: Added prototype.
* kjs/JSObject.cpp:
(KJS::JSObject::toString): Tweaked formatting.
* kjs/JSObject.h:
(KJS::JSValue::get): Use prototype instead of toObject to avoid creating
an object wrapper just to search for properties. This also saves an
unnecessary hash table lookup since the object wrappers themselves don't
have any properties.
* kjs/JSString.h: Added toThisString and toThisJSString.
* kjs/JSValue.cpp:
(KJS::JSCell::toThisString): Added.
(KJS::JSCell::toThisJSString): Added.
(KJS::JSCell::getJSNumber): Added.
(KJS::jsString): Changed return type to JSString*.
(KJS::jsOwnedString): Ditto.
* kjs/JSValue.h:
(KJS::JSValue::toThisString): Added.
(KJS::JSValue::toThisJSString): Added.
(KJS::JSValue::getJSNumber): Added.
* kjs/NumberObject.cpp:
(KJS::NumberObject::getJSNumber): Added.
(KJS::integer_part_noexp): Append C string directly rather than first
turning it into a UString.
(KJS::numberProtoFuncToString): Use getJSNumber to check if the value
is a number rather than isObject(&NumberObject::info). This works for
immediate numbers, number cells, and NumberObject instances.
(KJS::numberProtoFuncToLocaleString): Ditto.
(KJS::numberProtoFuncValueOf): Ditto.
(KJS::numberProtoFuncToFixed): Ditto.
(KJS::numberProtoFuncToExponential): Ditto.
(KJS::numberProtoFuncToPrecision): Ditto.
* kjs/NumberObject.h: Added getJSNumber.
* kjs/PropertySlot.cpp: Tweaked comment.
* kjs/internal.cpp:
(KJS::JSString::toThisString): Added.
(KJS::JSString::toThisJSString): Added.
(KJS::JSString::getOwnPropertySlot): Changed code that searches the
prototype chain to start with the string prototype and not create a
string object.
(KJS::JSNumberCell::toThisString): Added.
(KJS::JSNumberCell::getJSNumber): Added.
* kjs/lookup.cpp:
(KJS::staticFunctionGetter): Moved here, because there's no point in
having a function that's only used for a function pointer be inline.
(KJS::setUpStaticFunctionSlot): New function for getStaticFunctionSlot.
* kjs/lookup.h:
(KJS::staticValueGetter): Don't mark this inline. It doesn't make sense
to have a function that's only used for a function pointer be inline.
(KJS::getStaticFunctionSlot): Changed to get properties from the parent
first before doing any handling of functions. This is the fastest way
to return the function once the initial setup is done.
* kjs/string_object.cpp:
(KJS::StringObject::getPropertyNames): Call value() instead of getString(),
avoiding an unnecessary virtual function call (the call to the type()
function in the implementation of the isString() function).
(KJS::StringObject::toString): Added.
(KJS::StringObject::toThisString): Added.
(KJS::StringObject::toThisJSString): Added.
(KJS::substituteBackreferences): Rewrote to use a appending algorithm
instead of a the old one that tried to replace in place.
(KJS::stringProtoFuncReplace): Merged this function and the replace function.
Replaced the hand-rolled dynamic arrays for source ranges and replacements
with Vector.
(KJS::stringProtoFuncToString): Handle JSString as well as StringObject.
Removed the separate valueOf implementation, since it can just share this.
(KJS::stringProtoFuncCharAt): Use toThisString, which handles JSString as
well as StringObject, and is slightly more efficient than the old code too.
(KJS::stringProtoFuncCharCodeAt): Ditto.
(KJS::stringProtoFuncConcat): Ditto.
(KJS::stringProtoFuncIndexOf): Ditto.
(KJS::stringProtoFuncLastIndexOf): Ditto.
(KJS::stringProtoFuncMatch): Ditto.
(KJS::stringProtoFuncSearch): Ditto.
(KJS::stringProtoFuncSlice): Ditto.
(KJS::stringProtoFuncSplit): Ditto.
(KJS::stringProtoFuncSubstr): Ditto.
(KJS::stringProtoFuncSubstring): Ditto.
(KJS::stringProtoFuncToLowerCase): Use toThisJSString.
(KJS::stringProtoFuncToUpperCase): Ditto.
(KJS::stringProtoFuncToLocaleLowerCase): Ditto.
(KJS::stringProtoFuncToLocaleUpperCase): Ditto.
(KJS::stringProtoFuncLocaleCompare): Ditto.
(KJS::stringProtoFuncBig): Use toThisString.
(KJS::stringProtoFuncSmall): Ditto.
(KJS::stringProtoFuncBlink): Ditto.
(KJS::stringProtoFuncBold): Ditto.
(KJS::stringProtoFuncFixed): Ditto.
(KJS::stringProtoFuncItalics): Ditto.
(KJS::stringProtoFuncStrike): Ditto.
(KJS::stringProtoFuncSub): Ditto.
(KJS::stringProtoFuncSup): Ditto.
(KJS::stringProtoFuncFontcolor): Ditto.
(KJS::stringProtoFuncFontsize): Ditto.
(KJS::stringProtoFuncAnchor): Ditto.
(KJS::stringProtoFuncLink): Ditto.
* kjs/string_object.h: Added toString, toThisString, and toThisJSString.
* kjs/ustring.cpp:
(KJS::UString::append): Added a version that takes a character pointer and
size, so we don't have to create a UString just to append to another UString.
* kjs/ustring.h:
2008-06-26 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Maciej.
Make JSGlobalData per-thread.
No change on SunSpider total.
* wtf/ThreadSpecific.h: Re-enabled the actual implementation.
* kjs/JSGlobalObject.cpp:
(KJS::JSGlobalObject::~JSGlobalObject): Re-added a JSLock-related assertion. We'll probably
want to preserve these somehow to keep legacy behavior in working condition.
(KJS::JSGlobalObject::init): Initialize globalData pointer earlier, so that it is ready
when updating JSGlobalObject linked list.
* kjs/JSGlobalObject.h: (KJS::JSGlobalObject::head): Changed head() to be non-static, and
to use JSGlobalData associated with the current object.
* kjs/InitializeThreading.cpp: (KJS::initializeThreadingOnce): Removed a no longer needed
Heap::registerAsMainThread() call.
* kjs/JSGlobalData.h: Removed a lying lie comment - parserObjectExtraRefCounts is not
transient, and while newParserObjects may conceptually be such, there is still some node
manipulation going on outside Parser::parse which touches it.
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::~JSGlobalData): Delete recently added members.
(KJS::JSGlobalData::sharedInstance): Actually use a separate instance.
* kjs/collector.cpp:
(KJS::Heap::Heap):
(KJS::Heap::~Heap): Added a destructor, which unconditionally deletes everything.
(KJS::Heap::sweep): Removed code related to "collect on main thread only" logic.
(KJS::Heap::collect): Ditto.
(KJS::Heap::globalObjectCount): Explicitly use per-thread instance of JSGlobalObject linked
list now that JSGlobalObject::head() is not static. Curently, WebCoreStatistics methods only
work with the main thread currently anyway.
(KJS::Heap::protectedGlobalObjectCount): Ditto.
* kjs/collector.h: Removed code related to "collect on main thread only" logic.
* JavaScriptCore.exp: Removed Heap::collectOnMainThreadOnly.
2008-06-26 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Darin.
https://bugs.webkit.org/show_bug.cgi?id=19767
REGRESSION: Crash in sort() when visiting http://www.onnyturf.com/subway/
* kjs/JSArray.cpp: (KJS::AVLTreeAbstractorForArrayCompare::set_balance_factor):
Made changing balance factor from -1 to +1 work correctly.
* wtf/AVLTree.h: (KJS::AVLTreeDefaultBSet::operator[]): Added an assertion that catches
this slightly earlier.
2008-06-25 Timothy Hatcher <timothy@apple.com>
Fixes an ASSERT in the profiler when starting multiple profiles
with the same name inside the same function/program.
Reviewed by Kevin McCullough.
* profiler/Profile.cpp:
(KJS::Profile::Profile): Initialize m_stoppedCallDepth to zero.
(KJS::Profile::stopProfiling): Set the current node to the parent,
because we are in a call that will not get a didExecute call.
(KJS::Profile::removeProfile): Increment m_stoppedCallDepth to
account for didExecute not being called for profile.
(KJS::Profile::willExecute): Increment m_stoppedCallDepth if stopped.
(KJS::Profile::didExecute): Decrement m_stoppedCallDepth if stopped and
greater than zero, and return early.
* profiler/Profile.h: Added stoppedProfiling().
* profiler/Profiler.cpp:
(KJS::Profiler::findProfile): Removed.
(KJS::Profiler::startProfiling): Don't return early for stopped profiles.
(KJS::Profiler::stopProfiling): Skipp stopped profiles.
(KJS::Profiler::didFinishAllExecution): Code clean-up.
* profiler/Profiler.h: Removed findProfile.
2008-06-25 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Alexey Proskuryakov.
Attempt to fix Windows debug build. The compiler gives a warning when
Structured Exception Handling and destructors are used in the same
function. Using manual locking and unlocking instead of constructors
and destructors should fix the warning.
* kjs/Shell.cpp:
(main):
2008-06-25 Alexey Proskuryakov <ap@webkit.org>
Forgot to address a review comment about better names for tracked objects, doing it now.
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
* kjs/JSGlobalData.h:
* kjs/nodes.cpp:
(KJS::ParserRefCounted::ParserRefCounted):
(KJS::ParserRefCounted::ref):
(KJS::ParserRefCounted::deref):
(KJS::ParserRefCounted::hasOneRef):
(KJS::ParserRefCounted::deleteNewObjects):
2008-06-25 Alexey Proskuryakov <ap@webkit.org>
Reviewed by Geoff.
Remove more threadInstance() calls.
* kjs/JSFunction.cpp:
(KJS::JSFunction::getParameterName):
(KJS::IndexToNameMap::unMap):
(KJS::Arguments::deleteProperty):
* kjs/JSFunction.h:
Access nullIdentifier without going to thread specific storage.
* JavaScriptCore.exp:
* kjs/JSGlobalData.cpp:
(KJS::JSGlobalData::JSGlobalData):
* kjs/JSGlobalData.h:
* kjs/Parser.cpp:
(KJS::Parser::parse):
* kjs/Parser.h:
(KJS::ParserRefCountedData::ParserRefCountedData):
(KJS::Parser::parse):
* kjs/grammar.y:
* kjs/nodes.cpp:
(KJS::ParserRefCounted::ParserRefCounted):
(KJS::ParserRefCounted::ref):
(KJS::ParserRefCounted::deref):
(KJS::ParserRefCounted::hasOneRef):
(KJS::ParserRefCounted::deleteNewObjects):
(KJS::Node::Node):
(KJS::StatementNode::StatementNode):
(KJS::BreakpointCheckStatement::BreakpointCheckStatement):
(KJS::ConstDeclNode::ConstDeclNode):
(KJS::BlockNode::BlockNode):
(KJS::ForInNode::ForInNode):
(KJS::ScopeNode::ScopeNode):
(KJS::ProgramNode::ProgramNode):
(KJS::ProgramNode::create):
(KJS::EvalNode::EvalNode):
(KJS::EvalNode::create):
(KJS::FunctionBodyNode::FunctionBodyNode):
(KJS::FunctionBodyNode::create):
* kjs/nodes.h:
(KJS::ExpressionNode::):
(KJS::NullNode::):
(KJS::BooleanNode::):
(KJS::NumberNode::):
(KJS::ImmediateNumberNode::):
(KJS::StringNode::):
(KJS::RegExpNode::):
(KJS::ThisNode::):
(KJS::ResolveNode::):
(KJS::ElementNode::):
(KJS::ArrayNode::):
(KJS::PropertyNode::):
(KJS::PropertyListNode::):
(KJS::ObjectLiteralNode::):
(KJS::BracketAccessorNode::):
(KJS::DotAccessorNode::):
(KJS::ArgumentListNode::):
(KJS::ArgumentsNode::):
(KJS::NewExprNode::):
(KJS::EvalFunctionCallNode::):
(KJS::FunctionCallValueNode::):
(KJS::FunctionCallResolveNode::):
(KJS::FunctionCallBracketNode::):
(KJS::FunctionCallDotNode::):
(KJS::PrePostResolveNode::):
(KJS::PostIncResolveNode::):
(KJS::PostDecResolveNode::):
(KJS::PostfixBracketNode::):
(KJS::PostIncBracketNode::):
(KJS::PostDecBracketNode::):
(KJS::PostfixDotNode::):
(KJS::PostIncDotNode::):
(KJS::PostDecDotNode::):
(KJS::PostfixErrorNode::):
(KJS::DeleteResolveNode::):
(KJS::DeleteBracketNode::):
(KJS::DeleteDotNode::):
(KJS::DeleteValueNode::):
(KJS::VoidNode::):
(KJS::TypeOfResolveNode::):
(KJS::TypeOfValueNode::):
(KJS::PreIncResolveNode::):
(KJS::PreDecResolveNode::):
(KJS::PrefixBracketNode::):
(KJS::PreIncBracketNode::):
(KJS::PreDecBracketNode::):
(KJS::PrefixDotNode::):
(KJS::PreIncDotNode::):
(KJS::PreDecDotNode::):
(KJS::PrefixErrorNode::):
(KJS::UnaryOpNode::UnaryOpNode):
(KJS::UnaryPlusNode::):
(KJS::NegateNode::):
(KJS::BitwiseNotNode::):
(KJS::LogicalNotNode::):
(KJS::BinaryOpNode::BinaryOpNode):
(KJS::ReverseBinaryOpNode::ReverseBinaryOpNode):
(KJS::MultNode::):
(KJS::DivNode::):
(KJS::ModNode::):
(KJS::AddNode::):
(KJS::SubNode::):
(KJS::LeftShiftNode::):
(KJS::RightShiftNode::):
(KJS::UnsignedRightShiftNode::):
(KJS::LessNode::):
(KJS::GreaterNode::):
(KJS::LessEqNode::):
(KJS::GreaterEqNode::):
(KJS::InstanceOfNode::):
(KJS::InNode::):
(KJS::EqualNode::):
(KJS::NotEqualNode::):
(KJS::StrictEqualNode::):
(KJS::NotStrictEqualNode::):
(KJS::BitAndNode::):
(KJS::BitOrNode::):
(KJS::BitXOrNode::):
(KJS::LogicalAndNode::):
(KJS::LogicalOrNode::):
(KJS::ConditionalNode::):
(KJS::ReadModifyResolveNode::):
(KJS::AssignResolveNode::):
(KJS::ReadModifyBracketNode::):
(KJS::AssignBracketNode::):
(KJS::AssignDotNode::):
(KJS::ReadModifyDotNode::):
(KJS::AssignErrorNode::):
(KJS::CommaNode::):
(KJS::VarDeclCommaNode::):
(KJS::ConstStatementNode::):
(KJS::SourceElements::SourceElements):
(KJS::EmptyStatementNode::):
(KJS::DebuggerStatementNode::):
(KJS::ExprStatementNode::):
(KJS::VarStatementNode::):
(KJS::IfNode::):
(KJS::IfElseNode::):
(KJS::DoWhileNode::):
(KJS::WhileNode::):
(KJS::ForNode::):
(KJS::ContinueNode::):
(KJS::BreakNode::):
(KJS::ReturnNode::):
(KJS::WithNode::):
(KJS::LabelNode::):
(KJS::ThrowNode::):
(KJS::TryNode::):
(KJS::ParameterNode::):
(KJS::FuncExprNode::):
(KJS::FuncDeclNode::):
(KJS::CaseClauseNode::):
(KJS::ClauseListNode::):
(KJS::CaseBlockNode::):
(KJS::SwitchNode::):
Changed ParserRefCounted to hold a JSGlobalData pointer, and used it to replace
threadInstance calls.
2008-06-24 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Alexey Proskuryakov.
Make the JavaScript shell collect the heap from main() instead of
jscmain() to suppress leak messages in debug builds.
* kjs/Shell.cpp:
(main):
(jscmain):
2008-06-24 Cameron Zwarich <cwzwarich@uwaterloo.ca>
Reviewed by Maciej.
Make the conversion of the pair (less, jtrue) to jless use register
reference counting information for safety instead of requiring callers
to decide whether it is safe.
No changes on SunSpider codegen.
* VM/CodeGenerator.cpp:
(KJS::CodeGenerator::emitJumpIfTrue):
* VM/CodeGenerator.h:
* kjs/nodes.cpp:
(KJS::DoWhileNode::emitCode):
(KJS::WhileNode::emitCode):
(KJS::ForNode::emitCode):
(KJS::CaseBlockNode::emitCodeForBlock):
2008-06-24 Kevin McCullough <kmccullough@apple.com>
Reviewed by Tim.
<rdar://problem/6031594> JSProfiler: Profiler goes into an infinite
loop sometimes.
<rdar://problem/6031603> JSProfiler: Profiler asserts in debug and
give the wrong times in release
Fixed two issues found by Tim in the same test.
* profiler/Profile.cpp:
(KJS::Profile::removeProfileStart): No longer take profile's time from
all ancestors, but instead attribute it to its parent. Also add an
Assert to ensure we only delete the child we mean to.
(KJS::Profile::removeProfileEnd): Ditto for profileEnd.
(KJS::Profile::didExecute): Cleaned up the execution order and correctly
attribute all of the parent's time to the new node.
* profiler/ProfileNode.cpp: If this node does not have a startTime it
should not get a giant total time, but instead be 0.
(KJS::ProfileNode::endAndRecordCall):
* profiler/ProfileNode.h: