JavaScriptCore/ChangeLog

   1 2009-06-24  Sam Weinig  <sam@webkit.org>
   2
   3         Reviewed by Gavin Barraclough.
   4
   5         Make the opcode sampler work once again.
   6
   7         * jit/JIT.h:
   8         (JSC::JIT::compileGetByIdProto):
   9         (JSC::JIT::compileGetByIdSelfList):
  10         (JSC::JIT::compileGetByIdProtoList):
  11         (JSC::JIT::compileGetByIdChainList):
  12         (JSC::JIT::compileGetByIdChain):
  13         (JSC::JIT::compilePutByIdTransition):
  14         (JSC::JIT::compileCTIMachineTrampolines):
  15         (JSC::JIT::compilePatchGetArrayLength):
  16         * jit/JITStubCall.h:
  17         (JSC::JITStubCall::call):
  18
  19 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  20
  21         Reviewed by Maciej Stachowiak.
  22
  23         Extend FastAllocBase.h with 'using WTF::FastAllocBase' to avoid
  24         unnecessary WTF:: usings.
  25         Remove existing unnecessary WTF:: usings.
  26
  27         * interpreter/Interpreter.h:
  28         * profiler/CallIdentifier.h:
  29         * runtime/ScopeChain.h:
  30         * wtf/FastAllocBase.h:
  31
  32 2009-06-24  David Levin  <levin@chromium.org>
  33
  34         Fix all builds.
  35
  36         * bytecode/CodeBlock.h:
  37         * bytecompiler/BytecodeGenerator.h:
  38         * interpreter/Register.h:
  39
  40 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  41
  42         Reviewed by Maciej Stachowiak.
  43
  44         https://bugs.webkit.org/show_bug.cgi?id=26677
  45
  46         Inherits CodeBlock class from FastAllocBase because it
  47         has been instantiated by 'new' in JavaScriptCore/bytecode/CodeBlock.h:217.
  48
  49         * bytecode/CodeBlock.h:
  50
  51 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  52
  53         Reviewed by Maciej Stachowiak.
  54
  55         https://bugs.webkit.org/show_bug.cgi?id=26676
  56
  57         Inherits BytecodeGenerator class from FastAllocBase because it has been
  58         instantiated by 'new' in JavaScriptCore/parser/Nodes.cpp:1892.
  59
  60         * bytecompiler/BytecodeGenerator.h:
  61
  62 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  63
  64         Reviewed by Maciej Stachowiak.
  65
  66         https://bugs.webkit.org/show_bug.cgi?id=26675
  67
  68         Inherits Register class from FastAllocBase because it has been
  69         instantiated by 'new' in JavaScriptCore/runtime/JSVariableObject.h:149.
  70
  71         * interpreter/Register.h:
  72
  73 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  74
  75         Reviewed by Darin Adler.
  76
  77         https://bugs.webkit.org/show_bug.cgi?id=26674
  78
  79         Inherits HashMap class from FastAllocBase because it has been
  80         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:458.
  81
  82         * wtf/HashMap.h:
  83
  84 2009-06-24  Oliver Hunt  <oliver@apple.com>
  85
  86         Reviewed by Darin Adler.
  87
  88         <rdar://problem/6940519> REGRESSION (Safari 4 Public Beta - TOT): google.com/adplanner shows blank page instead of site details in "basic research'
  89
  90         The problem was caused by the page returned with a function using a
  91         var declaration list containing around ~3000 variables.  The solution
  92         to this is to flatten the comma expression representation and make
  93         codegen comma expressions and initializer lists iterative rather than
  94         recursive.
  95
  96         * parser/Grammar.y:
  97         * parser/NodeConstructors.h:
  98         (JSC::CommaNode::CommaNode):
  99         * parser/Nodes.cpp:
 100         (JSC::CommaNode::emitBytecode):
 101         * parser/Nodes.h:
 102         (JSC::ExpressionNode::isCommaNode):
 103         (JSC::CommaNode::isCommaNode):
 104         (JSC::CommaNode::append):
 105
 106 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 107
 108         Reviewed by Maciej Stachowiak.
 109
 110         https://bugs.webkit.org/show_bug.cgi?id=26645
 111
 112         Inherits ScopeChainNode class from FastAllocBase because it has been
 113         instantiated by 'new' in JavaScriptCore/runtime/ScopeChain.h:95.
 114
 115         * wtf/RefPtr.h:
 116
 117 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 118
 119         Reviewed by Darin Adler.
 120
 121         https://bugs.webkit.org/show_bug.cgi?id=26648
 122
 123         Inherits Deque class from FastAllocBase because it has been
 124         instantiated by 'new' with DEFINE_STATIC_LOCAL macro in
 125         JavaScriptCore/wtf/MainThread.cpp:62.
 126
 127         * wtf/Deque.h:
 128
 129 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 130
 131         Reviewed by Darin Adler.
 132
 133         https://bugs.webkit.org/show_bug.cgi?id=26644
 134
 135         Inherits RefPtr class from FastAllocBase because it has been
 136         instantiated by 'new' in JavaScriptCore/runtime/StructureChain.cpp:41.
 137
 138         * wtf/RefPtr.h:
 139
 140 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 141
 142         Reviewed by Darin Adler.
 143
 144         Inherits HashSet class from FastAllocBase, because it has been
 145         instantiated by 'new' in JavaScriptCore/runtime/Collector.h:116.
 146
 147         * wtf/HashSet.h:
 148
 149 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 150
 151         Reviewed by Darin Adler.
 152
 153         Inherits Vector class from FastAllocBase because it has been
 154         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:633.
 155
 156         * wtf/Vector.h:
 157
 158 2009-06-24  Norbert Leser  <norbert.leser@nokia.com>
 159
 160         Reviewed by Maciej Stachoviak.
 161
 162         The BytecodeGenerator objects were instantiated on stack, which takes up ~38kB per instance
 163         (each instance includes copy of JSC::CodeBlock with large SymbolTable, etc.).
 164         Specifically, since there is nested invocation (e.g., GlobalCode --> FunctionCode),
 165         the stack overflows immediately on Symbian hardware (max. 80 kB).
 166         Proposed change allocates generator objects on heap.
 167         Performance impact (if any) should be negligible and change is proposed as general fix,
 168         rather than ifdef'd for SYMBIAN.
 169
 170         * parser/Nodes.cpp:
 171         (JSC::ProgramNode::generateBytecode):
 172         (JSC::EvalNode::generateBytecode):
 173         (JSC::EvalNode::bytecodeForExceptionInfoReparse):
 174         (JSC::FunctionBodyNode::generateBytecode):
 175         (JSC::FunctionBodyNode::bytecodeForExceptionInfoReparse):
 176
 177 2009-06-23  Oliver Hunt  <oliver@apple.com>
 178
 179         Reviewed by Gavin Barraclough.
 180
 181         <rdar://problem/6992806> REGRESSION: Enumeration can skip new properties in cases of prototypes that have more than 64 (26593)
 182         <https://bugs.webkit.org/show_bug.cgi?id=26593>
 183
 184         Do not attempt to cache structure chains if they contain a dictionary at any level.
 185
 186         * interpreter/Interpreter.cpp:
 187         (JSC::Interpreter::tryCachePutByID):
 188         (JSC::Interpreter::tryCacheGetByID):
 189         * jit/JITStubs.cpp:
 190         (JSC::JITThunks::tryCachePutByID):
 191         * runtime/Structure.cpp:
 192         (JSC::Structure::getEnumerablePropertyNames):
 193         (JSC::Structure::addPropertyTransition):
 194         * runtime/StructureChain.cpp:
 195         (JSC::StructureChain::isCacheable):
 196         * runtime/StructureChain.h:
 197
 198 2009-06-23  Yong Li  <yong.li@torchmobile.com>
 199
 200         Reviewed by George Staikos.
 201
 202         https://bugs.webkit.org/show_bug.cgi?id=26654
 203         Add the proper export define for the JavaScriptCore API when building for WINCE.
 204
 205         * API/JSBase.h:
 206
 207 2009-06-23  Joe Mason  <joe.mason@torchmobile.com>
 208
 209         Reviewed by Adam Treat.
 210
 211         Authors: Yong Li <yong.li@torchmobile.com>, Joe Mason <joe.mason@torchmobile.com>
 212
 213         https://bugs.webkit.org/show_bug.cgi?id=26611
 214         Implement currentThreadStackBase on WINCE by adding a global,
 215         g_stackBase, which must be set to the address of a local variable
 216         by the caller before calling any WebKit function that invokes JSC.
 217
 218         * runtime/Collector.cpp:
 219         (JSC::isPageWritable):
 220         (JSC::getStackBase):
 221           Starts at the top of the stack and returns the entire range of
 222           consecutive writable pages as an estimate of the actual stack.
 223           This will be much bigger than the actual stack range, so some
 224           dead objects can't be collected, but it guarantees live objects
 225           aren't collected prematurely.
 226
 227         (JSC::currentThreadStackBase):
 228           On WinCE, returns g_stackBase if set or call getStackBase as a
 229           fallback if not.
 230
 231 2009-06-23  Oliver Hunt  <oliver@apple.com>
 232
 233         Reviewed by Alexey Proskuryakov.
 234
 235         Fix stupid performance problem in the LiteralParser
 236
 237         The LiteralParser was making a new UString in order to use
 238         toDouble, however UString's toDouble allows a much wider range
 239         of numberic strings than the LiteralParser accepts, and requires
 240         an additional heap allocation or two for the construciton of the
 241         UString.  To rectify this we just call WTF::dtoa directly using
 242         a stack allocated buffer to hold the validated numeric literal.
 243
 244         * runtime/LiteralParser.cpp:
 245         (JSC::LiteralParser::Lexer::lexNumber):
 246         (JSC::LiteralParser::parse):
 247         * runtime/LiteralParser.h:
 248
 249 2009-06-22  Oliver Hunt  <oliver@apple.com>
 250
 251         Reviewed by Alexey Proskuryakov.
 252
 253         Bug 26640: JSON.stringify needs to special case Boolean objects
 254         <https://bugs.webkit.org/show_bug.cgi?id=26640>
 255
 256         Add special case handling of the Boolean object so we match current
 257         ES5 errata.
 258
 259         * runtime/JSONObject.cpp:
 260         (JSC::unwrapBoxedPrimitive): renamed from unwrapNumberOrString
 261         (JSC::gap):
 262         (JSC::Stringifier::appendStringifiedValue):
 263
 264 2009-06-22  Oliver Hunt  <oliver@apple.com>
 265
 266         Reviewed by Darin Adler.
 267
 268         Bug 26591: Support revivers in JSON.parse
 269         <https://bugs.webkit.org/show_bug.cgi?id=26591>
 270
 271         Add reviver support to JSON.parse.  This completes the JSON object.
 272
 273         * runtime/JSONObject.cpp:
 274         (JSC::Walker::Walker):
 275         (JSC::Walker::callReviver):
 276         (JSC::Walker::walk):
 277         (JSC::JSONProtoFuncParse):
 278
 279 2009-06-21  Oliver Hunt  <oliver@apple.com>
 280
 281         Reviewed by Darin Adler.
 282
 283         Bug 26592: Support standard toJSON functions
 284         <https://bugs.webkit.org/show_bug.cgi?id=26592>
 285
 286         Add support for the standard Date.toJSON function.
 287
 288         * runtime/DatePrototype.cpp:
 289         (JSC::dateProtoFuncToJSON):
 290
 291 2009-06-21  Oliver Hunt  <oliver@apple.com>
 292
 293         Reviewed by Sam Weinig.
 294
 295         Bug 26594: JSC needs to support Date.toISOString
 296         <https://bugs.webkit.org/show_bug.cgi?id=26594>
 297
 298         Add support for Date.toISOString.
 299
 300         * runtime/DatePrototype.cpp:
 301         (JSC::dateProtoFuncToISOString):
 302
 303 2009-06-21  Oliver Hunt  <oliver@apple.com>
 304
 305         Reviewed by NOBODY (Build fix).
 306
 307         Remove dead code.
 308
 309         * runtime/LiteralParser.cpp:
 310         (JSC::LiteralParser::parse):
 311
 312 2009-06-21  Oliver Hunt  <oliver@apple.com>
 313
 314         Reviewed by Darin Adler and Cameron Zwarich.
 315
 316         Bug 26587: Support JSON.parse
 317         <https://bugs.webkit.org/show_bug.cgi?id=26587>
 318
 319         Extend the LiteralParser to support the full strict JSON
 320         grammar, fix a few places where the grammar was incorrectly
 321         lenient.   Doesn't yet support the JSON.parse reviver function
 322         but that does not block the JSON.parse functionality itself.
 323
 324         * interpreter/Interpreter.cpp:
 325         (JSC::Interpreter::callEval):
 326         * runtime/JSGlobalObjectFunctions.cpp:
 327         (JSC::globalFuncEval):
 328         * runtime/JSONObject.cpp:
 329         (JSC::JSONProtoFuncParse):
 330         * runtime/LiteralParser.cpp:
 331         (JSC::LiteralParser::Lexer::lex):
 332         (JSC::isSafeStringCharacter):
 333         (JSC::LiteralParser::Lexer::lexString):
 334         (JSC::LiteralParser::parse):
 335         * runtime/LiteralParser.h:
 336         (JSC::LiteralParser::LiteralParser):
 337         (JSC::LiteralParser::tryJSONParse):
 338         (JSC::LiteralParser::):
 339         (JSC::LiteralParser::Lexer::Lexer):
 340
 341 2009-06-21  David Levin  <levin@chromium.org>
 342
 343         Reviewed by NOBODY (speculative build fix for windows).
 344
 345         Simply removed some whitespace form this file to make windows build wtf and
 346         hopefully copy the new MessageQueque.h so that WebCore picks it up.
 347
 348         * wtf/Assertions.cpp:
 349
 350 2009-06-21  Drew Wilson  <atwilson@google.com>
 351
 352         Reviewed by David Levin.
 353
 354         <https://bugs.webkit.org/show_bug.cgi?id=25043>
 355         Added support for multi-threaded MessagePorts.
 356
 357         * wtf/MessageQueue.h:
 358         (WTF::::appendAndCheckEmpty):
 359             Added API to test whether the queue was empty before adding an element.
 360
 361 2009-06-20  David D. Kilzer  <ddkilzer@webkit.org>
 362
 363         Fix namespace comment in SegmentedVector.h
 364
 365         * wtf/SegmentedVector.h: Updated namespace comment to reflect
 366         new namespace after r44897.
 367
 368 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
 369
 370         Bug 24986: ARM JIT port
 371         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 372
 373         Reviewed by Oliver Hunt.
 374
 375         An Iterator added for SegmentedVector. Currently
 376         only the pre ++ operator is supported.
 377
 378         * wtf/SegmentedVector.h:
 379         (WTF::SegmentedVectorIterator::~SegmentedVectorIterator):
 380         (WTF::SegmentedVectorIterator::operator*):
 381         (WTF::SegmentedVectorIterator::operator->):
 382         (WTF::SegmentedVectorIterator::operator++):
 383         (WTF::SegmentedVectorIterator::operator==):
 384         (WTF::SegmentedVectorIterator::operator!=):
 385         (WTF::SegmentedVectorIterator::operator=):
 386         (WTF::SegmentedVectorIterator::SegmentedVectorIterator):
 387         (WTF::SegmentedVector::alloc):
 388         (WTF::SegmentedVector::begin):
 389         (WTF::SegmentedVector::end):
 390
 391 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
 392
 393         Bug 24986: ARM JIT port
 394         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 395
 396         Reviewed by Oliver Hunt.
 397
 398         Move SegmentedVector to /wtf subdirectory
 399         and change "namespace JSC" to "namespace WTF"
 400
 401         Additional build file updates by David Kilzer.
 402
 403         * GNUmakefile.am: Updated path to SegmentedVector.h.
 404         * JavaScriptCore.order: Updated SegmentedVector namespace from
 405         JSC to WTF in mangled C++ method name.
 406         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
 407         Removed reference to bytecompiler\SegmentedVector.h.
 408         * JavaScriptCore.vcproj/WTF/WTF.vcproj: Added reference to
 409         wtf\SegmentedVector.h.
 410         * JavaScriptCore.xcodeproj/project.pbxproj: Moved
 411         SegmentedVector.h definition from bytecompiler subdirectory to
 412         wtf subdirectory.
 413         * bytecompiler/BytecodeGenerator.h: Updated #include path to
 414         SegmentedVector.h and prepended WTF:: namespace to its use.
 415         * parser/Lexer.h: Ditto.
 416         * wtf/SegmentedVector.h: Renamed from JavaScriptCore/bytecompiler/SegmentedVector.h.
 417         (WTF::SegmentedVector::SegmentedVector):
 418         (WTF::SegmentedVector::~SegmentedVector):
 419         (WTF::SegmentedVector::size):
 420         (WTF::SegmentedVector::at):
 421         (WTF::SegmentedVector::operator[]):
 422         (WTF::SegmentedVector::last):
 423         (WTF::SegmentedVector::append):
 424         (WTF::SegmentedVector::removeLast):
 425         (WTF::SegmentedVector::grow):
 426         (WTF::SegmentedVector::clear):
 427         (WTF::SegmentedVector::deleteAllSegments):
 428         (WTF::SegmentedVector::segmentExistsFor):
 429         (WTF::SegmentedVector::segmentFor):
 430         (WTF::SegmentedVector::subscriptFor):
 431         (WTF::SegmentedVector::ensureSegmentsFor):
 432         (WTF::SegmentedVector::ensureSegment):
 433
 434 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 435
 436         Reviewed by NOBODY (build fix take 2 - rename FIELD_OFFSET to something that doesn't conflict with winnt.h).
 437
 438         * jit/JIT.cpp:
 439         (JSC::JIT::privateCompile):
 440         (JSC::JIT::privateCompileCTIMachineTrampolines):
 441         (JSC::JIT::emitGetVariableObjectRegister):
 442         (JSC::JIT::emitPutVariableObjectRegister):
 443         * jit/JIT.h:
 444         * jit/JITArithmetic.cpp:
 445         (JSC::JIT::emit_op_rshift):
 446         (JSC::JIT::emitSlow_op_jnless):
 447         (JSC::JIT::emitSlow_op_jnlesseq):
 448         (JSC::JIT::compileBinaryArithOp):
 449         * jit/JITCall.cpp:
 450         (JSC::JIT::compileOpCallInitializeCallFrame):
 451         (JSC::JIT::compileOpCall):
 452         * jit/JITInlineMethods.h:
 453         (JSC::JIT::restoreArgumentReference):
 454         (JSC::JIT::checkStructure):
 455         * jit/JITOpcodes.cpp:
 456         (JSC::JIT::emit_op_instanceof):
 457         (JSC::JIT::emit_op_get_scoped_var):
 458         (JSC::JIT::emit_op_put_scoped_var):
 459         (JSC::JIT::emit_op_construct_verify):
 460         (JSC::JIT::emit_op_resolve_global):
 461         (JSC::JIT::emit_op_jeq_null):
 462         (JSC::JIT::emit_op_jneq_null):
 463         (JSC::JIT::emit_op_to_jsnumber):
 464         (JSC::JIT::emit_op_catch):
 465         (JSC::JIT::emit_op_eq_null):
 466         (JSC::JIT::emit_op_neq_null):
 467         (JSC::JIT::emit_op_convert_this):
 468         (JSC::JIT::emit_op_profile_will_call):
 469         (JSC::JIT::emit_op_profile_did_call):
 470         (JSC::JIT::emitSlow_op_get_by_val):
 471         * jit/JITPropertyAccess.cpp:
 472         (JSC::JIT::emit_op_get_by_val):
 473         (JSC::JIT::emit_op_put_by_val):
 474         (JSC::JIT::emit_op_method_check):
 475         (JSC::JIT::compileGetByIdHotPath):
 476         (JSC::JIT::emit_op_put_by_id):
 477         (JSC::JIT::compilePutDirectOffset):
 478         (JSC::JIT::compileGetDirectOffset):
 479         (JSC::JIT::privateCompilePutByIdTransition):
 480         (JSC::JIT::privateCompilePatchGetArrayLength):
 481         * jit/JITStubs.cpp:
 482         (JSC::JITThunks::JITThunks):
 483
 484 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 485
 486         Reviewed by NOBODY (Windows build fix).
 487
 488         * jit/JIT.h:
 489         * jit/JITInlineMethods.h:
 490
 491 2009-06-19  Gabor Loki  <loki@inf.u-szeged.hu>
 492
 493         Reviewed by Gavin Barraclough.
 494
 495         Reorganize ARM architecture specific macros.
 496         Use PLATFORM_ARM_ARCH(7) instead of PLATFORM(ARM_V7).
 497
 498         Bug 24986: ARM JIT port
 499         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 500
 501         * assembler/ARMv7Assembler.h:
 502         * assembler/AbstractMacroAssembler.h:
 503         (JSC::AbstractMacroAssembler::Imm32::Imm32):
 504         * assembler/MacroAssembler.h:
 505         * assembler/MacroAssemblerCodeRef.h:
 506         (JSC::MacroAssemblerCodePtr::MacroAssemblerCodePtr):
 507         * jit/ExecutableAllocator.h:
 508         (JSC::ExecutableAllocator::cacheFlush):
 509         * jit/JIT.h:
 510         * jit/JITInlineMethods.h:
 511         (JSC::JIT::restoreArgumentReferenceForTrampoline):
 512         * jit/JITStubs.cpp:
 513         * jit/JITStubs.h:
 514         * wtf/Platform.h:
 515         * yarr/RegexJIT.cpp:
 516         (JSC::Yarr::RegexGenerator::generateEnter):
 517         (JSC::Yarr::RegexGenerator::generateReturn):
 518
 519 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 520
 521         Reviewed by Oliver Hunt.
 522
 523         Fix armv7 JIT build issues.
 524
 525         Unfortunate the arm compiler does not like the use of offsetof on JITStackFrame (since it now contains non POD types),
 526         and the FIELD_OFFSET macro does not appear constantish enough for it to be happy with its use in COMPILE_ASSERT macros.
 527
 528         * Replace offsetofs with FIELD_OFFSETs (safe on C++ objects).
 529         * Move COMPILE_ASSERTs defending layout of JITStackFrame structure on armv7 into JITThunks constructor.
 530
 531         * jit/JIT.cpp:
 532         * jit/JIT.h:
 533         * jit/JITInlineMethods.h:
 534         (JSC::JIT::restoreArgumentReference):
 535         * jit/JITOpcodes.cpp:
 536         (JSC::JIT::emit_op_catch):
 537         * jit/JITStubs.cpp:
 538         (JSC::JITThunks::JITThunks):
 539
 540 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
 541
 542         Blind attempt at build fix.
 543
 544         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 545         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 546
 547 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 548
 549         Reviewed by Oliver Hunt.
 550
 551         Inherits CallIdentifier struct from FastAllocBase because it has been
 552         instantiated by 'new' in JavaScriptCore/profiler/CallIdentifier.h:86.
 553
 554         * wtf/HashCountedSet.h:
 555
 556 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
 557
 558         Reviewed by Oliver Hunt.
 559
 560         https://bugs.webkit.org/show_bug.cgi?id=26540
 561         Modify the test shell to add a new function 'checkSyntax' that will
 562         only parse the source instead of executing it. In this way we can test
 563         pure parsing performance against some of the larger scripts in the wild.
 564
 565         * jsc.cpp:
 566         (GlobalObject::GlobalObject):
 567         (functionCheckSyntax):
 568
 569 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 570
 571         Reviewed by Darin Adler.
 572
 573         Inherits HashCountedSet class from FastAllocBase because it has been
 574         instantiated by 'new' in JavaScriptCore/runtime/Collector.cpp:1095.
 575
 576         * wtf/HashCountedSet.h:
 577
 578 2009-06-19  Yong Li  <yong.li@torchmobile.com>
 579
 580         Reviewed by George Staikos.
 581
 582         https://bugs.webkit.org/show_bug.cgi?id=26558
 583         Declare these symbols extern for WINCE as they are provided by libce.
 584
 585         * runtime/DateConstructor.cpp:
 586         * runtime/DatePrototype.cpp:
 587         (JSC::formatLocaleDate):
 588
 589 2009-06-19  Oliver Hunt  <oliver@apple.com>
 590
 591         Reviewed by Maciej Stachowiak.
 592
 593         <rdar://problem/6988973> ScopeChain leak in interpreter builds
 594
 595         Move the Scopechain destruction code in JSFunction outside of the ENABLE(JIT)
 596         path.
 597
 598         * runtime/JSFunction.cpp:
 599         (JSC::JSFunction::~JSFunction):
 600         * wtf/Platform.h:
 601
 602 2009-06-19  Yong Li  <yong.li@torchmobile.com>
 603
 604         Reviewed by George Staikos.
 605
 606         https://bugs.webkit.org/show_bug.cgi?id=26543
 607         Windows CE uses 'GetLastError' instead of 'errno.'
 608
 609         * interpreter/RegisterFile.h:
 610         (JSC::RegisterFile::RegisterFile):
 611         (JSC::RegisterFile::grow):
 612
 613 2009-06-19  David Levin  <levin@chromium.org>
 614
 615         Reviewed by NOBODY (Windows build fix).
 616
 617         Add export for Windows corresponding to OSX export done in r44844.
 618         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 619         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 620
 621 2009-06-18  Oliver Hunt  <oliver@apple.com>
 622
 623         Reviewed by Gavin "Viceroy of Venezuela" Barraclough.
 624
 625         Bug 26532: Native functions do not correctly unlink from optimised callsites when they're collected
 626         <https://bugs.webkit.org/show_bug.cgi?id=26532> <rdar://problem/6625385>
 627
 628         We need to make sure that each native function instance correctly unlinks any references to it
 629         when it is collected.  Allowing this to happen required a few changes:
 630             * Every native function needs a codeblock to track the link information
 631             * To have this codeblock, every function now also needs its own functionbodynode
 632               so we no longer get to have a single shared instance.
 633             * Identifying a host function is now done by looking for CodeBlock::codeType() == NativeCode
 634
 635         * JavaScriptCore.exp:
 636         * bytecode/CodeBlock.cpp:
 637         (JSC::CodeBlock::CodeBlock):
 638            Constructor for NativeCode CodeBlock
 639         (JSC::CodeBlock::derefStructures):
 640         (JSC::CodeBlock::refStructures):
 641         (JSC::CodeBlock::reparseForExceptionInfoIfNecessary):
 642         (JSC::CodeBlock::handlerForBytecodeOffset):
 643         (JSC::CodeBlock::lineNumberForBytecodeOffset):
 644         (JSC::CodeBlock::expressionRangeForBytecodeOffset):
 645         (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset):
 646         (JSC::CodeBlock::functionRegisterForBytecodeOffset):
 647         (JSC::CodeBlock::hasGlobalResolveInstructionAtBytecodeOffset):
 648         (JSC::CodeBlock::hasGlobalResolveInfoAtBytecodeOffset):
 649         (JSC::CodeBlock::setJITCode):
 650            Add assertions to ensure we don't try and use NativeCode CodeBlocks as
 651            a normal codeblock.
 652
 653         * bytecode/CodeBlock.h:
 654         (JSC::):
 655         (JSC::CodeBlock::source):
 656         (JSC::CodeBlock::sourceOffset):
 657         (JSC::CodeBlock::evalCodeCache):
 658         (JSC::CodeBlock::createRareDataIfNecessary):
 659           More assertions.
 660
 661         * jit/JIT.cpp:
 662         (JSC::JIT::privateCompileCTIMachineTrampolines):
 663         (JSC::JIT::linkCall):
 664           Update logic to allow native function caching
 665
 666         * jit/JITStubs.cpp:
 667         * parser/Nodes.cpp:
 668         (JSC::FunctionBodyNode::createNativeThunk):
 669         (JSC::FunctionBodyNode::isHostFunction):
 670         * parser/Nodes.h:
 671         * runtime/JSFunction.cpp:
 672         (JSC::JSFunction::JSFunction):
 673         (JSC::JSFunction::~JSFunction):
 674         (JSC::JSFunction::mark):
 675         * runtime/JSGlobalData.cpp:
 676         (JSC::JSGlobalData::~JSGlobalData):
 677         * runtime/JSGlobalData.h:
 678
 679 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 680
 681         Reviewed by NOBODY (Windows build fix).
 682
 683         * wtf/DateMath.cpp:
 684         (WTF::calculateUTCOffset):
 685
 686 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 687
 688         Reviewed by Geoff Garen.
 689
 690         Timezone calculation incorrect in Venezuela.
 691
 692         https://bugs.webkit.org/show_bug.cgi?id=26531
 693         <rdar://problem/6646169> Time is incorrectly reported to JavaScript in both Safari 3 and Firefox 3
 694
 695         The problem is that we're calculating the timezone relative to 01/01/2000,
 696         but the VET timezone changed from -4 hours to -4:30 hours on 12/09/2007.
 697         According to the spec, section 15.9.1.9 states "the time since the beginning
 698         of the year", presumably meaning the *current* year.  Change the calculation
 699         to be based on whatever the current year is, rather than a canned date.
 700
 701         No performance impact.
 702
 703         * wtf/DateMath.cpp:
 704         (WTF::calculateUTCOffset):
 705
 706 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 707
 708         Rubber Stamped by Mark Rowe (originally reviewed by Sam Weinig).
 709
 710         (Reintroducing patch added in r44492, and reverted in r44796.)
 711
 712         Change the implementation of op_throw so the stub function always modifies its
 713         return address - if it doesn't find a 'catch' it will switch to a trampoline
 714         to force a return from JIT execution.  This saves memory, by avoiding the need
 715         for a unique return for every op_throw.
 716
 717         * jit/JITOpcodes.cpp:
 718         (JSC::JIT::emit_op_throw):
 719             JITStubs::cti_op_throw now always changes its return address,
 720             remove return code generated after the stub call (this is now
 721             handled by ctiOpThrowNotCaught).
 722         * jit/JITStubs.cpp:
 723         (JSC::):
 724             Add ctiOpThrowNotCaught definitions.
 725         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
 726             Change cti_op_throw to always change its return address.
 727         * jit/JITStubs.h:
 728             Add ctiOpThrowNotCaught declaration.
 729
 730 2009-06-18  Kevin McCullough  <kmccullough@apple.com>
 731
 732         Reviewed by Oliver Hunt.
 733
 734         <rdar://problem/6940880> REGRESSION: Breakpoints don't break in 64-bit
 735
 736         - Exposed functions now needed by WebCore.
 737
 738         * JavaScriptCore.exp:
 739
 740 2009-06-17  Darin Adler  <darin@apple.com>
 741
 742         Reviewed by Oliver Hunt.
 743
 744         Bug 26429: Make JSON.stringify non-recursive so it can handle objects
 745         of arbitrary complexity
 746         https://bugs.webkit.org/show_bug.cgi?id=26429
 747
 748         For marking I decided not to use gcProtect, because this is inside the engine
 749         so it's easy enough to just do marking. And that darned gcProtect does locking!
 750         Oliver tried to convince me to used MarkedArgumentBuffer, but the constructor
 751         for that class says "FIXME: Remove all clients of this API, then remove this API."
 752
 753         * runtime/Collector.cpp:
 754         (JSC::Heap::collect): Add a call to JSONObject::markStringifiers.
 755
 756         * runtime/CommonIdentifiers.cpp:
 757         (JSC::CommonIdentifiers::CommonIdentifiers): Added emptyIdentifier.
 758         * runtime/CommonIdentifiers.h: Ditto.
 759
 760         * runtime/JSGlobalData.cpp:
 761         (JSC::JSGlobalData::JSGlobalData): Initialize firstStringifierToMark to 0.
 762         * runtime/JSGlobalData.h: Added firstStringifierToMark.
 763
 764         * runtime/JSONObject.cpp: Cut down the includes to the needed ones only.
 765         (JSC::unwrapNumberOrString): Added. Helper for unwrapping number and string
 766         objects to get their number and string values.
 767         (JSC::ReplacerPropertyName::ReplacerPropertyName): Added. The class is used
 768         to wrap an identifier or integer so we don't have to do any work unless we
 769         actually call a replacer.
 770         (JSC::ReplacerPropertyName::value): Added.
 771         (JSC::gap): Added. Helper function for the Stringifier constructor.
 772         (JSC::PropertyNameForFunctionCall::PropertyNameForFunctionCall): Added.
 773         The class is used to wrap an identifier or integer so we don't have to
 774         allocate a number or string until we actually call toJSON or a replacer.
 775         (JSC::PropertyNameForFunctionCall::asJSValue): Added.
 776         (JSC::Stringifier::Stringifier): Updated and moved out of the class
 777         definition. Added code to hook this into a singly linked list for marking.
 778         (JSC::Stringifier::~Stringifier): Remove from the singly linked list.
 779         (JSC::Stringifier::mark): Mark all the objects in the holder stacks.
 780         (JSC::Stringifier::stringify): Updated.
 781         (JSC::Stringifier::appendQuotedString): Tweaked and streamlined a bit.
 782         (JSC::Stringifier::toJSON): Renamed from toJSONValue.
 783         (JSC::Stringifier::appendStringifiedValue): Renamed from stringify.
 784         Added code to use the m_holderStack to do non-recursive stringify of
 785         objects and arrays. This code also uses the timeout checker since in
 786         pathological cases it could be slow even without calling into the
 787         JavaScript virtual machine.
 788         (JSC::Stringifier::willIndent): Added.
 789         (JSC::Stringifier::indent): Added.
 790         (JSC::Stringifier::unindent): Added.
 791         (JSC::Stringifier::startNewLine): Added.
 792         (JSC::Stringifier::Holder::Holder): Added.
 793         (JSC::Stringifier::Holder::appendNextProperty): Added. This is the
 794         function that handles the format of arrays and objects.
 795         (JSC::JSONObject::getOwnPropertySlot): Moved this down to the bottom
 796         of the file so the JSONObject class is not interleaved with the
 797         Stringifier class.
 798         (JSC::JSONObject::markStringifiers): Added. Calls mark.
 799         (JSC::JSONProtoFuncStringify): Streamlined the code here. The code
 800         to compute the gap string is now a separate function.
 801
 802         * runtime/JSONObject.h: Made everything private. Added markStringifiers.
 803
 804 2009-06-17  Oliver Hunt  <oliver@apple.com>
 805
 806         Reviewed by Gavin Barraclough.
 807
 808         <rdar://problem/6974140> REGRESSION(r43849): Crash in cti_op_call_NotJSFunction when getting directions on maps.google.com
 809
 810         Roll out r43849 as it appears that we cannot rely on the address of
 811         an objects property storage being constant even if the structure is
 812         unchanged.
 813
 814         * jit/JIT.h:
 815         * jit/JITPropertyAccess.cpp:
 816         (JSC::JIT::compileGetDirectOffset):
 817         (JSC::JIT::privateCompileGetByIdProto):
 818         (JSC::JIT::privateCompileGetByIdProtoList):
 819         (JSC::JIT::privateCompileGetByIdChainList):
 820         (JSC::JIT::privateCompileGetByIdChain):
 821
 822 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
 823
 824         Rubber Stamped by Mark Rowe.
 825
 826         Fully revert r44492 & r44748 while we fix a bug they cause on internal builds <rdar://problem/6955963>.
 827
 828         * jit/JITOpcodes.cpp:
 829         (JSC::JIT::emit_op_throw):
 830         * jit/JITStubs.cpp:
 831         (JSC::):
 832         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
 833         * jit/JITStubs.h:
 834
 835 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
 836
 837         Reviewed by Mark Rowe.
 838
 839         <rdar://problem/6947426> sunspider math-cordic.js exhibits different intermediate results running 32-bit vs. 64-bit
 840
 841         On 64-bit, NaN-encoded values must be detagged before they can be used in rshift.
 842
 843         No performance impact.
 844
 845         * jit/JITArithmetic.cpp:
 846         (JSC::JIT::emit_op_rshift):
 847
 848 2009-06-17  Adam Treat  <adam.treat@torchmobile.com>
 849
 850         Reviewed by George Staikos.
 851
 852         https://bugs.webkit.org/show_bug.cgi?id=23155
 853         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
 854
 855         * jsc.cpp:
 856         (main):
 857
 858 2009-06-17  George Staikos  <george.staikos@torchmobile.com>
 859
 860         Reviewed by Adam Treat.
 861
 862         https://bugs.webkit.org/show_bug.cgi?id=23155
 863         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
 864
 865         * config.h:
 866         * jsc.cpp:
 867         * wtf/Assertions.cpp:
 868         * wtf/Assertions.h:
 869         * wtf/CurrentTime.cpp:
 870         (WTF::lowResUTCTime):
 871         * wtf/DateMath.cpp:
 872         (WTF::getLocalTime):
 873         * wtf/MathExtras.h:
 874         * wtf/Platform.h:
 875         * wtf/StringExtras.h:
 876         * wtf/Threading.h:
 877         * wtf/win/MainThreadWin.cpp:
 878
 879 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
 880
 881         Reviewed by Oliver Hunt.
 882
 883         <rdar://problem/6974175> ASSERT in JITStubs.cpp at appsaccess.apple.com
 884
 885         Remove PropertySlot::putValue - PropertySlots should only be used for getting,
 886         not putting.  Rename JSGlobalObject::getOwnPropertySlot to hasOwnPropertyForWrite,
 887         which is what it really was being used to ask, and remove some other getOwnPropertySlot
 888         & getOwnPropertySlotForWrite methods, which were unused and likely to lead to confusion.
 889
 890         * runtime/JSGlobalObject.h:
 891         (JSC::JSGlobalObject::hasOwnPropertyForWrite):
 892         * runtime/JSObject.h:
 893         * runtime/JSStaticScopeObject.cpp:
 894         * runtime/JSStaticScopeObject.h:
 895         * runtime/PropertySlot.h:
 896
 897 2009-06-16  Gavin Barraclough  <barraclough@apple.com>
 898
 899         Reviewed by Oliver hunt.
 900
 901         Temporarily partially disable r44492, since this is causing some problems on internal builds.
 902
 903         * jit/JITOpcodes.cpp:
 904         (JSC::JIT::emit_op_throw):
 905         * jit/JITStubs.cpp:
 906         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
 907
 908 2009-06-16  Sam Weinig  <sam@webkit.org>
 909
 910         Fix windows build.
 911
 912         * jit/JIT.cpp:
 913         (JSC::JIT::JIT):
 914
 915 2009-06-16  Sam Weinig  <sam@webkit.org>
 916
 917         Reviewed by Oliver Hunt.
 918
 919         Initialize m_bytecodeIndex to -1 in JIT, and correctly initialize
 920         it for each type of stub using the return address to find the correct
 921         offset.
 922
 923         * jit/JIT.cpp:
 924         (JSC::JIT::JIT):
 925         * jit/JIT.h:
 926         (JSC::JIT::compileGetByIdProto):
 927         (JSC::JIT::compileGetByIdSelfList):
 928         (JSC::JIT::compileGetByIdProtoList):
 929         (JSC::JIT::compileGetByIdChainList):
 930         (JSC::JIT::compileGetByIdChain):
 931         (JSC::JIT::compilePutByIdTransition):
 932         (JSC::JIT::compileCTIMachineTrampolines):
 933         (JSC::JIT::compilePatchGetArrayLength):
 934         * jit/JITStubCall.h:
 935         (JSC::JITStubCall::call):
 936
 937 == Rolled over to ChangeLog-2009-06-16 ==