JavaScriptCore/ChangeLog

   1 2009-07-03  Yong Li  <yong.li@torchmobile.com>
   2
   3         Reviewed by Maciej Stachowiak (and revised slightly)
   4
   5         RegExp::match to be optimized
   6         https://bugs.webkit.org/show_bug.cgi?id=26957
   7
   8         Allow regexp matching to use Vectors with inline capacity instead of
   9         allocating a new ovector buffer every time.
  10
  11         ~5% speedup on SunSpider string-unpack-code test, 0.3% on SunSpider overall.
  12
  13         * runtime/RegExp.cpp:
  14         (JSC::RegExp::match):
  15         * runtime/RegExp.h:
  16         * runtime/RegExpConstructor.cpp:
  17         (JSC::RegExpConstructorPrivate::RegExpConstructorPrivate):
  18         (JSC::RegExpConstructorPrivate::lastOvector):
  19         (JSC::RegExpConstructorPrivate::tempOvector):
  20         (JSC::RegExpConstructorPrivate::changeLastOvector):
  21         (JSC::RegExpConstructor::performMatch):
  22         (JSC::RegExpMatchesArray::RegExpMatchesArray):
  23         (JSC::RegExpMatchesArray::fillArrayInstance):
  24         (JSC::RegExpConstructor::getBackref):
  25         (JSC::RegExpConstructor::getLastParen):
  26         (JSC::RegExpConstructor::getLeftContext):
  27         (JSC::RegExpConstructor::getRightContext):
  28         * runtime/StringPrototype.cpp:
  29         (JSC::stringProtoFuncSplit):
  30
  31 2009-06-30  Kwang Yul Seo  <skyul@company100.net>
  32
  33         Reviewed by Eric Seidel.
  34
  35         Override operator new/delete with const std::nothrow_t& as the second
  36         argument.
  37         https://bugs.webkit.org/show_bug.cgi?id=26792
  38
  39         On Windows CE, operator new/delete, new[]/delete[] with const
  40         std::nothrow_t& must be overrided because some standard template
  41         libraries use these operators.
  42
  43         The problem occurs when memory allocated by new(size_t s, const
  44         std::nothrow_t&) is freed by delete(void* p). This causes the umatched
  45         malloc/free problem.
  46
  47         The patch overrides all new, delete, new[] and delete[] to use
  48         fastMaloc and fastFree consistently.
  49
  50         * wtf/FastMalloc.h:
  51         (throw):
  52
  53 2009-06-30  Gabor Loki  <loki@inf.u-szeged.hu>
  54
  55         Reviewed by Sam Weinig.
  56
  57         <https://bugs.webkit.org/show_bug.cgi?id=24986>
  58
  59         Remove unnecessary references to AssemblerBuffer.
  60
  61         * interpreter/Interpreter.cpp:
  62         * interpreter/Interpreter.h:
  63
  64 2009-06-29  David Levin  <levin@chromium.org>
  65
  66         Reviewed by Oliver Hunt.
  67
  68         Still seeing occasional leaks from UString::sharedBuffer code
  69         https://bugs.webkit.org/show_bug.cgi?id=26420
  70
  71         The problem is that the pointer to the memory allocation isn't visible
  72         by "leaks" due to the lower bits being used as flags.  The fix is to
  73         make the pointer visible in memory (in debug only). The downside of
  74         this fix that the memory allocated by sharedBuffer will still look like
  75         a leak in non-debug builds when any flags are set.
  76
  77         * wtf/PtrAndFlags.h:
  78         (WTF::PtrAndFlags::set):
  79
  80 2009-06-29  Sam Weinig  <sam@webkit.org>
  81
  82         Reviewed by Mark Rowe.
  83
  84         Remove more unused scons support.
  85
  86         * SConstruct: Removed.
  87
  88 2009-06-29  Oliver Hunt  <oliver@apple.com>
  89
  90         Reviewed by Gavin Barraclough.
  91
  92         <rdar://problem/7016214> JSON.parse fails to parse valid JSON with most Unicode characters
  93         <https://bugs.webkit.org/show_bug.cgi?id=26802>
  94
  95         In the original JSON.parse patch unicode was handled correctly, however in some last
  96         minute "clean up" I oversimplified isSafeStringCharacter.  This patch corrects this bug.
  97
  98         * runtime/LiteralParser.cpp:
  99         (JSC::isSafeStringCharacter):
 100         (JSC::LiteralParser::Lexer::lexString):
 101
 102 2009-06-26  Oliver Hunt  <oliver@apple.com>
 103
 104         Reviewed by Dan Bernstein.
 105
 106         <rdar://problem/7009684> REGRESSION(r45039): Crashes inside JSEvent::put on PowerPC (26746)
 107         <https://bugs.webkit.org/show_bug.cgi?id=26746>
 108
 109         Fix for r45039 incorrectly uncached a get_by_id by converting it to put_by_id.  Clearly this
 110         is less than correct.  This patch corrects that error.
 111
 112         * interpreter/Interpreter.cpp:
 113         (JSC::Interpreter::tryCacheGetByID):
 114
 115 2009-06-26  Eric Seidel  <eric@webkit.org>
 116
 117         No review, only rolling out r45259.
 118
 119         Roll out r45259 after crash appeared on the bots:
 120         plugins/undefined-property-crash.html
 121         ASSERTION FAILED: s <= HeapConstants<heapType>::cellSize
 122         (leopard-intel-debug-tests/build/JavaScriptCore/runtime/Collector.cpp:278
 123         void* JSC::Heap::heapAllocate(size_t) [with JSC::HeapType heapType = PrimaryHeap])
 124
 125         * runtime/DateInstance.cpp:
 126         * runtime/Identifier.cpp:
 127         * runtime/Lookup.h:
 128         * runtime/RegExpConstructor.cpp:
 129         * runtime/RegExpObject.h:
 130         * runtime/ScopeChain.h:
 131         * runtime/UString.h:
 132
 133 2009-06-26  Jedrzej Nowacki  <jedrzej.nowacki@nokia.com>
 134
 135         Reviewed by Simon Hausmann.
 136
 137         Add support for QDataStream operators to Vector.
 138
 139         * wtf/Vector.h:
 140         (WTF::operator<<):
 141         (WTF::operator>>):
 142
 143 2009-06-24  Sam Weinig  <sam@webkit.org>
 144
 145         Reviewed by Gavin Barraclough.
 146
 147         Make the opcode sampler work once again.
 148
 149         * jit/JIT.h:
 150         (JSC::JIT::compileGetByIdProto):
 151         (JSC::JIT::compileGetByIdSelfList):
 152         (JSC::JIT::compileGetByIdProtoList):
 153         (JSC::JIT::compileGetByIdChainList):
 154         (JSC::JIT::compileGetByIdChain):
 155         (JSC::JIT::compilePutByIdTransition):
 156         (JSC::JIT::compileCTIMachineTrampolines):
 157         (JSC::JIT::compilePatchGetArrayLength):
 158         * jit/JITStubCall.h:
 159         (JSC::JITStubCall::call):
 160
 161 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 162
 163         Reviewed by Maciej Stachowiak.
 164
 165         Extend FastAllocBase.h with 'using WTF::FastAllocBase' to avoid
 166         unnecessary WTF:: usings.
 167         Remove existing unnecessary WTF:: usings.
 168
 169         * interpreter/Interpreter.h:
 170         * profiler/CallIdentifier.h:
 171         * runtime/ScopeChain.h:
 172         * wtf/FastAllocBase.h:
 173
 174 2009-06-24  David Levin  <levin@chromium.org>
 175
 176         Fix all builds.
 177
 178         * bytecode/CodeBlock.h:
 179         * bytecompiler/BytecodeGenerator.h:
 180         * interpreter/Register.h:
 181
 182 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 183
 184         Reviewed by Maciej Stachowiak.
 185
 186         https://bugs.webkit.org/show_bug.cgi?id=26677
 187
 188         Inherits CodeBlock class from FastAllocBase because it
 189         has been instantiated by 'new' in JavaScriptCore/bytecode/CodeBlock.h:217.
 190
 191         * bytecode/CodeBlock.h:
 192
 193 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 194
 195         Reviewed by Maciej Stachowiak.
 196
 197         https://bugs.webkit.org/show_bug.cgi?id=26676
 198
 199         Inherits BytecodeGenerator class from FastAllocBase because it has been
 200         instantiated by 'new' in JavaScriptCore/parser/Nodes.cpp:1892.
 201
 202         * bytecompiler/BytecodeGenerator.h:
 203
 204 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 205
 206         Reviewed by Maciej Stachowiak.
 207
 208         https://bugs.webkit.org/show_bug.cgi?id=26675
 209
 210         Inherits Register class from FastAllocBase because it has been
 211         instantiated by 'new' in JavaScriptCore/runtime/JSVariableObject.h:149.
 212
 213         * interpreter/Register.h:
 214
 215 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 216
 217         Reviewed by Darin Adler.
 218
 219         https://bugs.webkit.org/show_bug.cgi?id=26674
 220
 221         Inherits HashMap class from FastAllocBase because it has been
 222         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:458.
 223
 224         * wtf/HashMap.h:
 225
 226 2009-06-24  Oliver Hunt  <oliver@apple.com>
 227
 228         Reviewed by Darin Adler.
 229
 230         <rdar://problem/6940519> REGRESSION (Safari 4 Public Beta - TOT): google.com/adplanner shows blank page instead of site details in "basic research'
 231
 232         The problem was caused by the page returned with a function using a
 233         var declaration list containing around ~3000 variables.  The solution
 234         to this is to flatten the comma expression representation and make
 235         codegen comma expressions and initializer lists iterative rather than
 236         recursive.
 237
 238         * parser/Grammar.y:
 239         * parser/NodeConstructors.h:
 240         (JSC::CommaNode::CommaNode):
 241         * parser/Nodes.cpp:
 242         (JSC::CommaNode::emitBytecode):
 243         * parser/Nodes.h:
 244         (JSC::ExpressionNode::isCommaNode):
 245         (JSC::CommaNode::isCommaNode):
 246         (JSC::CommaNode::append):
 247
 248 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 249
 250         Reviewed by Maciej Stachowiak.
 251
 252         https://bugs.webkit.org/show_bug.cgi?id=26645
 253
 254         Inherits ScopeChainNode class from FastAllocBase because it has been
 255         instantiated by 'new' in JavaScriptCore/runtime/ScopeChain.h:95.
 256
 257         * wtf/RefPtr.h:
 258
 259 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 260
 261         Reviewed by Darin Adler.
 262
 263         https://bugs.webkit.org/show_bug.cgi?id=26648
 264
 265         Inherits Deque class from FastAllocBase because it has been
 266         instantiated by 'new' with DEFINE_STATIC_LOCAL macro in
 267         JavaScriptCore/wtf/MainThread.cpp:62.
 268
 269         * wtf/Deque.h:
 270
 271 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 272
 273         Reviewed by Darin Adler.
 274
 275         https://bugs.webkit.org/show_bug.cgi?id=26644
 276
 277         Inherits RefPtr class from FastAllocBase because it has been
 278         instantiated by 'new' in JavaScriptCore/runtime/StructureChain.cpp:41.
 279
 280         * wtf/RefPtr.h:
 281
 282 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 283
 284         Reviewed by Darin Adler.
 285
 286         Inherits HashSet class from FastAllocBase, because it has been
 287         instantiated by 'new' in JavaScriptCore/runtime/Collector.h:116.
 288
 289         * wtf/HashSet.h:
 290
 291 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 292
 293         Reviewed by Darin Adler.
 294
 295         Inherits Vector class from FastAllocBase because it has been
 296         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:633.
 297
 298         * wtf/Vector.h:
 299
 300 2009-06-24  Norbert Leser  <norbert.leser@nokia.com>
 301
 302         Reviewed by Maciej Stachoviak.
 303
 304         The BytecodeGenerator objects were instantiated on stack, which takes up ~38kB per instance
 305         (each instance includes copy of JSC::CodeBlock with large SymbolTable, etc.).
 306         Specifically, since there is nested invocation (e.g., GlobalCode --> FunctionCode),
 307         the stack overflows immediately on Symbian hardware (max. 80 kB).
 308         Proposed change allocates generator objects on heap.
 309         Performance impact (if any) should be negligible and change is proposed as general fix,
 310         rather than ifdef'd for SYMBIAN.
 311
 312         * parser/Nodes.cpp:
 313         (JSC::ProgramNode::generateBytecode):
 314         (JSC::EvalNode::generateBytecode):
 315         (JSC::EvalNode::bytecodeForExceptionInfoReparse):
 316         (JSC::FunctionBodyNode::generateBytecode):
 317         (JSC::FunctionBodyNode::bytecodeForExceptionInfoReparse):
 318
 319 2009-06-23  Oliver Hunt  <oliver@apple.com>
 320
 321         Reviewed by Gavin Barraclough.
 322
 323         <rdar://problem/6992806> REGRESSION: Enumeration can skip new properties in cases of prototypes that have more than 64 (26593)
 324         <https://bugs.webkit.org/show_bug.cgi?id=26593>
 325
 326         Do not attempt to cache structure chains if they contain a dictionary at any level.
 327
 328         * interpreter/Interpreter.cpp:
 329         (JSC::Interpreter::tryCachePutByID):
 330         (JSC::Interpreter::tryCacheGetByID):
 331         * jit/JITStubs.cpp:
 332         (JSC::JITThunks::tryCachePutByID):
 333         * runtime/Structure.cpp:
 334         (JSC::Structure::getEnumerablePropertyNames):
 335         (JSC::Structure::addPropertyTransition):
 336         * runtime/StructureChain.cpp:
 337         (JSC::StructureChain::isCacheable):
 338         * runtime/StructureChain.h:
 339
 340 2009-06-23  Yong Li  <yong.li@torchmobile.com>
 341
 342         Reviewed by George Staikos.
 343
 344         https://bugs.webkit.org/show_bug.cgi?id=26654
 345         Add the proper export define for the JavaScriptCore API when building for WINCE.
 346
 347         * API/JSBase.h:
 348
 349 2009-06-23  Joe Mason  <joe.mason@torchmobile.com>
 350
 351         Reviewed by Adam Treat.
 352
 353         Authors: Yong Li <yong.li@torchmobile.com>, Joe Mason <joe.mason@torchmobile.com>
 354
 355         https://bugs.webkit.org/show_bug.cgi?id=26611
 356         Implement currentThreadStackBase on WINCE by adding a global,
 357         g_stackBase, which must be set to the address of a local variable
 358         by the caller before calling any WebKit function that invokes JSC.
 359
 360         * runtime/Collector.cpp:
 361         (JSC::isPageWritable):
 362         (JSC::getStackBase):
 363           Starts at the top of the stack and returns the entire range of
 364           consecutive writable pages as an estimate of the actual stack.
 365           This will be much bigger than the actual stack range, so some
 366           dead objects can't be collected, but it guarantees live objects
 367           aren't collected prematurely.
 368
 369         (JSC::currentThreadStackBase):
 370           On WinCE, returns g_stackBase if set or call getStackBase as a
 371           fallback if not.
 372
 373 2009-06-23  Oliver Hunt  <oliver@apple.com>
 374
 375         Reviewed by Alexey Proskuryakov.
 376
 377         Fix stupid performance problem in the LiteralParser
 378
 379         The LiteralParser was making a new UString in order to use
 380         toDouble, however UString's toDouble allows a much wider range
 381         of numberic strings than the LiteralParser accepts, and requires
 382         an additional heap allocation or two for the construciton of the
 383         UString.  To rectify this we just call WTF::dtoa directly using
 384         a stack allocated buffer to hold the validated numeric literal.
 385
 386         * runtime/LiteralParser.cpp:
 387         (JSC::LiteralParser::Lexer::lexNumber):
 388         (JSC::LiteralParser::parse):
 389         * runtime/LiteralParser.h:
 390
 391 2009-06-22  Oliver Hunt  <oliver@apple.com>
 392
 393         Reviewed by Alexey Proskuryakov.
 394
 395         Bug 26640: JSON.stringify needs to special case Boolean objects
 396         <https://bugs.webkit.org/show_bug.cgi?id=26640>
 397
 398         Add special case handling of the Boolean object so we match current
 399         ES5 errata.
 400
 401         * runtime/JSONObject.cpp:
 402         (JSC::unwrapBoxedPrimitive): renamed from unwrapNumberOrString
 403         (JSC::gap):
 404         (JSC::Stringifier::appendStringifiedValue):
 405
 406 2009-06-22  Oliver Hunt  <oliver@apple.com>
 407
 408         Reviewed by Darin Adler.
 409
 410         Bug 26591: Support revivers in JSON.parse
 411         <https://bugs.webkit.org/show_bug.cgi?id=26591>
 412
 413         Add reviver support to JSON.parse.  This completes the JSON object.
 414
 415         * runtime/JSONObject.cpp:
 416         (JSC::Walker::Walker):
 417         (JSC::Walker::callReviver):
 418         (JSC::Walker::walk):
 419         (JSC::JSONProtoFuncParse):
 420
 421 2009-06-21  Oliver Hunt  <oliver@apple.com>
 422
 423         Reviewed by Darin Adler.
 424
 425         Bug 26592: Support standard toJSON functions
 426         <https://bugs.webkit.org/show_bug.cgi?id=26592>
 427
 428         Add support for the standard Date.toJSON function.
 429
 430         * runtime/DatePrototype.cpp:
 431         (JSC::dateProtoFuncToJSON):
 432
 433 2009-06-21  Oliver Hunt  <oliver@apple.com>
 434
 435         Reviewed by Sam Weinig.
 436
 437         Bug 26594: JSC needs to support Date.toISOString
 438         <https://bugs.webkit.org/show_bug.cgi?id=26594>
 439
 440         Add support for Date.toISOString.
 441
 442         * runtime/DatePrototype.cpp:
 443         (JSC::dateProtoFuncToISOString):
 444
 445 2009-06-21  Oliver Hunt  <oliver@apple.com>
 446
 447         Reviewed by NOBODY (Build fix).
 448
 449         Remove dead code.
 450
 451         * runtime/LiteralParser.cpp:
 452         (JSC::LiteralParser::parse):
 453
 454 2009-06-21  Oliver Hunt  <oliver@apple.com>
 455
 456         Reviewed by Darin Adler and Cameron Zwarich.
 457
 458         Bug 26587: Support JSON.parse
 459         <https://bugs.webkit.org/show_bug.cgi?id=26587>
 460
 461         Extend the LiteralParser to support the full strict JSON
 462         grammar, fix a few places where the grammar was incorrectly
 463         lenient.   Doesn't yet support the JSON.parse reviver function
 464         but that does not block the JSON.parse functionality itself.
 465
 466         * interpreter/Interpreter.cpp:
 467         (JSC::Interpreter::callEval):
 468         * runtime/JSGlobalObjectFunctions.cpp:
 469         (JSC::globalFuncEval):
 470         * runtime/JSONObject.cpp:
 471         (JSC::JSONProtoFuncParse):
 472         * runtime/LiteralParser.cpp:
 473         (JSC::LiteralParser::Lexer::lex):
 474         (JSC::isSafeStringCharacter):
 475         (JSC::LiteralParser::Lexer::lexString):
 476         (JSC::LiteralParser::parse):
 477         * runtime/LiteralParser.h:
 478         (JSC::LiteralParser::LiteralParser):
 479         (JSC::LiteralParser::tryJSONParse):
 480         (JSC::LiteralParser::):
 481         (JSC::LiteralParser::Lexer::Lexer):
 482
 483 2009-06-21  David Levin  <levin@chromium.org>
 484
 485         Reviewed by NOBODY (speculative build fix for windows).
 486
 487         Simply removed some whitespace form this file to make windows build wtf and
 488         hopefully copy the new MessageQueque.h so that WebCore picks it up.
 489
 490         * wtf/Assertions.cpp:
 491
 492 2009-06-21  Drew Wilson  <atwilson@google.com>
 493
 494         Reviewed by David Levin.
 495
 496         <https://bugs.webkit.org/show_bug.cgi?id=25043>
 497         Added support for multi-threaded MessagePorts.
 498
 499         * wtf/MessageQueue.h:
 500         (WTF::::appendAndCheckEmpty):
 501             Added API to test whether the queue was empty before adding an element.
 502
 503 2009-06-20  David D. Kilzer  <ddkilzer@webkit.org>
 504
 505         Fix namespace comment in SegmentedVector.h
 506
 507         * wtf/SegmentedVector.h: Updated namespace comment to reflect
 508         new namespace after r44897.
 509
 510 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
 511
 512         Bug 24986: ARM JIT port
 513         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 514
 515         Reviewed by Oliver Hunt.
 516
 517         An Iterator added for SegmentedVector. Currently
 518         only the pre ++ operator is supported.
 519
 520         * wtf/SegmentedVector.h:
 521         (WTF::SegmentedVectorIterator::~SegmentedVectorIterator):
 522         (WTF::SegmentedVectorIterator::operator*):
 523         (WTF::SegmentedVectorIterator::operator->):
 524         (WTF::SegmentedVectorIterator::operator++):
 525         (WTF::SegmentedVectorIterator::operator==):
 526         (WTF::SegmentedVectorIterator::operator!=):
 527         (WTF::SegmentedVectorIterator::operator=):
 528         (WTF::SegmentedVectorIterator::SegmentedVectorIterator):
 529         (WTF::SegmentedVector::alloc):
 530         (WTF::SegmentedVector::begin):
 531         (WTF::SegmentedVector::end):
 532
 533 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
 534
 535         Bug 24986: ARM JIT port
 536         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 537
 538         Reviewed by Oliver Hunt.
 539
 540         Move SegmentedVector to /wtf subdirectory
 541         and change "namespace JSC" to "namespace WTF"
 542
 543         Additional build file updates by David Kilzer.
 544
 545         * GNUmakefile.am: Updated path to SegmentedVector.h.
 546         * JavaScriptCore.order: Updated SegmentedVector namespace from
 547         JSC to WTF in mangled C++ method name.
 548         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
 549         Removed reference to bytecompiler\SegmentedVector.h.
 550         * JavaScriptCore.vcproj/WTF/WTF.vcproj: Added reference to
 551         wtf\SegmentedVector.h.
 552         * JavaScriptCore.xcodeproj/project.pbxproj: Moved
 553         SegmentedVector.h definition from bytecompiler subdirectory to
 554         wtf subdirectory.
 555         * bytecompiler/BytecodeGenerator.h: Updated #include path to
 556         SegmentedVector.h and prepended WTF:: namespace to its use.
 557         * parser/Lexer.h: Ditto.
 558         * wtf/SegmentedVector.h: Renamed from JavaScriptCore/bytecompiler/SegmentedVector.h.
 559         (WTF::SegmentedVector::SegmentedVector):
 560         (WTF::SegmentedVector::~SegmentedVector):
 561         (WTF::SegmentedVector::size):
 562         (WTF::SegmentedVector::at):
 563         (WTF::SegmentedVector::operator[]):
 564         (WTF::SegmentedVector::last):
 565         (WTF::SegmentedVector::append):
 566         (WTF::SegmentedVector::removeLast):
 567         (WTF::SegmentedVector::grow):
 568         (WTF::SegmentedVector::clear):
 569         (WTF::SegmentedVector::deleteAllSegments):
 570         (WTF::SegmentedVector::segmentExistsFor):
 571         (WTF::SegmentedVector::segmentFor):
 572         (WTF::SegmentedVector::subscriptFor):
 573         (WTF::SegmentedVector::ensureSegmentsFor):
 574         (WTF::SegmentedVector::ensureSegment):
 575
 576 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 577
 578         Reviewed by NOBODY (build fix take 2 - rename FIELD_OFFSET to something that doesn't conflict with winnt.h).
 579
 580         * jit/JIT.cpp:
 581         (JSC::JIT::privateCompile):
 582         (JSC::JIT::privateCompileCTIMachineTrampolines):
 583         (JSC::JIT::emitGetVariableObjectRegister):
 584         (JSC::JIT::emitPutVariableObjectRegister):
 585         * jit/JIT.h:
 586         * jit/JITArithmetic.cpp:
 587         (JSC::JIT::emit_op_rshift):
 588         (JSC::JIT::emitSlow_op_jnless):
 589         (JSC::JIT::emitSlow_op_jnlesseq):
 590         (JSC::JIT::compileBinaryArithOp):
 591         * jit/JITCall.cpp:
 592         (JSC::JIT::compileOpCallInitializeCallFrame):
 593         (JSC::JIT::compileOpCall):
 594         * jit/JITInlineMethods.h:
 595         (JSC::JIT::restoreArgumentReference):
 596         (JSC::JIT::checkStructure):
 597         * jit/JITOpcodes.cpp:
 598         (JSC::JIT::emit_op_instanceof):
 599         (JSC::JIT::emit_op_get_scoped_var):
 600         (JSC::JIT::emit_op_put_scoped_var):
 601         (JSC::JIT::emit_op_construct_verify):
 602         (JSC::JIT::emit_op_resolve_global):
 603         (JSC::JIT::emit_op_jeq_null):
 604         (JSC::JIT::emit_op_jneq_null):
 605         (JSC::JIT::emit_op_to_jsnumber):
 606         (JSC::JIT::emit_op_catch):
 607         (JSC::JIT::emit_op_eq_null):
 608         (JSC::JIT::emit_op_neq_null):
 609         (JSC::JIT::emit_op_convert_this):
 610         (JSC::JIT::emit_op_profile_will_call):
 611         (JSC::JIT::emit_op_profile_did_call):
 612         (JSC::JIT::emitSlow_op_get_by_val):
 613         * jit/JITPropertyAccess.cpp:
 614         (JSC::JIT::emit_op_get_by_val):
 615         (JSC::JIT::emit_op_put_by_val):
 616         (JSC::JIT::emit_op_method_check):
 617         (JSC::JIT::compileGetByIdHotPath):
 618         (JSC::JIT::emit_op_put_by_id):
 619         (JSC::JIT::compilePutDirectOffset):
 620         (JSC::JIT::compileGetDirectOffset):
 621         (JSC::JIT::privateCompilePutByIdTransition):
 622         (JSC::JIT::privateCompilePatchGetArrayLength):
 623         * jit/JITStubs.cpp:
 624         (JSC::JITThunks::JITThunks):
 625
 626 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 627
 628         Reviewed by NOBODY (Windows build fix).
 629
 630         * jit/JIT.h:
 631         * jit/JITInlineMethods.h:
 632
 633 2009-06-19  Gabor Loki  <loki@inf.u-szeged.hu>
 634
 635         Reviewed by Gavin Barraclough.
 636
 637         Reorganize ARM architecture specific macros.
 638         Use PLATFORM_ARM_ARCH(7) instead of PLATFORM(ARM_V7).
 639
 640         Bug 24986: ARM JIT port
 641         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 642
 643         * assembler/ARMv7Assembler.h:
 644         * assembler/AbstractMacroAssembler.h:
 645         (JSC::AbstractMacroAssembler::Imm32::Imm32):
 646         * assembler/MacroAssembler.h:
 647         * assembler/MacroAssemblerCodeRef.h:
 648         (JSC::MacroAssemblerCodePtr::MacroAssemblerCodePtr):
 649         * jit/ExecutableAllocator.h:
 650         (JSC::ExecutableAllocator::cacheFlush):
 651         * jit/JIT.h:
 652         * jit/JITInlineMethods.h:
 653         (JSC::JIT::restoreArgumentReferenceForTrampoline):
 654         * jit/JITStubs.cpp:
 655         * jit/JITStubs.h:
 656         * wtf/Platform.h:
 657         * yarr/RegexJIT.cpp:
 658         (JSC::Yarr::RegexGenerator::generateEnter):
 659         (JSC::Yarr::RegexGenerator::generateReturn):
 660
 661 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
 662
 663         Reviewed by Oliver Hunt.
 664
 665         Fix armv7 JIT build issues.
 666
 667         Unfortunate the arm compiler does not like the use of offsetof on JITStackFrame (since it now contains non POD types),
 668         and the FIELD_OFFSET macro does not appear constantish enough for it to be happy with its use in COMPILE_ASSERT macros.
 669
 670         * Replace offsetofs with FIELD_OFFSETs (safe on C++ objects).
 671         * Move COMPILE_ASSERTs defending layout of JITStackFrame structure on armv7 into JITThunks constructor.
 672
 673         * jit/JIT.cpp:
 674         * jit/JIT.h:
 675         * jit/JITInlineMethods.h:
 676         (JSC::JIT::restoreArgumentReference):
 677         * jit/JITOpcodes.cpp:
 678         (JSC::JIT::emit_op_catch):
 679         * jit/JITStubs.cpp:
 680         (JSC::JITThunks::JITThunks):
 681
 682 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
 683
 684         Blind attempt at build fix.
 685
 686         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 687         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 688
 689 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 690
 691         Reviewed by Oliver Hunt.
 692
 693         Inherits CallIdentifier struct from FastAllocBase because it has been
 694         instantiated by 'new' in JavaScriptCore/profiler/CallIdentifier.h:86.
 695
 696         * wtf/HashCountedSet.h:
 697
 698 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
 699
 700         Reviewed by Oliver Hunt.
 701
 702         https://bugs.webkit.org/show_bug.cgi?id=26540
 703         Modify the test shell to add a new function 'checkSyntax' that will
 704         only parse the source instead of executing it. In this way we can test
 705         pure parsing performance against some of the larger scripts in the wild.
 706
 707         * jsc.cpp:
 708         (GlobalObject::GlobalObject):
 709         (functionCheckSyntax):
 710
 711 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 712
 713         Reviewed by Darin Adler.
 714
 715         Inherits HashCountedSet class from FastAllocBase because it has been
 716         instantiated by 'new' in JavaScriptCore/runtime/Collector.cpp:1095.
 717
 718         * wtf/HashCountedSet.h:
 719
 720 2009-06-19  Yong Li  <yong.li@torchmobile.com>
 721
 722         Reviewed by George Staikos.
 723
 724         https://bugs.webkit.org/show_bug.cgi?id=26558
 725         Declare these symbols extern for WINCE as they are provided by libce.
 726
 727         * runtime/DateConstructor.cpp:
 728         * runtime/DatePrototype.cpp:
 729         (JSC::formatLocaleDate):
 730
 731 2009-06-19  Oliver Hunt  <oliver@apple.com>
 732
 733         Reviewed by Maciej Stachowiak.
 734
 735         <rdar://problem/6988973> ScopeChain leak in interpreter builds
 736
 737         Move the Scopechain destruction code in JSFunction outside of the ENABLE(JIT)
 738         path.
 739
 740         * runtime/JSFunction.cpp:
 741         (JSC::JSFunction::~JSFunction):
 742         * wtf/Platform.h:
 743
 744 2009-06-19  Yong Li  <yong.li@torchmobile.com>
 745
 746         Reviewed by George Staikos.
 747
 748         https://bugs.webkit.org/show_bug.cgi?id=26543
 749         Windows CE uses 'GetLastError' instead of 'errno.'
 750
 751         * interpreter/RegisterFile.h:
 752         (JSC::RegisterFile::RegisterFile):
 753         (JSC::RegisterFile::grow):
 754
 755 2009-06-19  David Levin  <levin@chromium.org>
 756
 757         Reviewed by NOBODY (Windows build fix).
 758
 759         Add export for Windows corresponding to OSX export done in r44844.
 760         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 761         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 762
 763 2009-06-18  Oliver Hunt  <oliver@apple.com>
 764
 765         Reviewed by Gavin "Viceroy of Venezuela" Barraclough.
 766
 767         Bug 26532: Native functions do not correctly unlink from optimised callsites when they're collected
 768         <https://bugs.webkit.org/show_bug.cgi?id=26532> <rdar://problem/6625385>
 769
 770         We need to make sure that each native function instance correctly unlinks any references to it
 771         when it is collected.  Allowing this to happen required a few changes:
 772             * Every native function needs a codeblock to track the link information
 773             * To have this codeblock, every function now also needs its own functionbodynode
 774               so we no longer get to have a single shared instance.
 775             * Identifying a host function is now done by looking for CodeBlock::codeType() == NativeCode
 776
 777         * JavaScriptCore.exp:
 778         * bytecode/CodeBlock.cpp:
 779         (JSC::CodeBlock::CodeBlock):
 780            Constructor for NativeCode CodeBlock
 781         (JSC::CodeBlock::derefStructures):
 782         (JSC::CodeBlock::refStructures):
 783         (JSC::CodeBlock::reparseForExceptionInfoIfNecessary):
 784         (JSC::CodeBlock::handlerForBytecodeOffset):
 785         (JSC::CodeBlock::lineNumberForBytecodeOffset):
 786         (JSC::CodeBlock::expressionRangeForBytecodeOffset):
 787         (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset):
 788         (JSC::CodeBlock::functionRegisterForBytecodeOffset):
 789         (JSC::CodeBlock::hasGlobalResolveInstructionAtBytecodeOffset):
 790         (JSC::CodeBlock::hasGlobalResolveInfoAtBytecodeOffset):
 791         (JSC::CodeBlock::setJITCode):
 792            Add assertions to ensure we don't try and use NativeCode CodeBlocks as
 793            a normal codeblock.
 794
 795         * bytecode/CodeBlock.h:
 796         (JSC::):
 797         (JSC::CodeBlock::source):
 798         (JSC::CodeBlock::sourceOffset):
 799         (JSC::CodeBlock::evalCodeCache):
 800         (JSC::CodeBlock::createRareDataIfNecessary):
 801           More assertions.
 802
 803         * jit/JIT.cpp:
 804         (JSC::JIT::privateCompileCTIMachineTrampolines):
 805         (JSC::JIT::linkCall):
 806           Update logic to allow native function caching
 807
 808         * jit/JITStubs.cpp:
 809         * parser/Nodes.cpp:
 810         (JSC::FunctionBodyNode::createNativeThunk):
 811         (JSC::FunctionBodyNode::isHostFunction):
 812         * parser/Nodes.h:
 813         * runtime/JSFunction.cpp:
 814         (JSC::JSFunction::JSFunction):
 815         (JSC::JSFunction::~JSFunction):
 816         (JSC::JSFunction::mark):
 817         * runtime/JSGlobalData.cpp:
 818         (JSC::JSGlobalData::~JSGlobalData):
 819         * runtime/JSGlobalData.h:
 820
 821 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 822
 823         Reviewed by NOBODY (Windows build fix).
 824
 825         * wtf/DateMath.cpp:
 826         (WTF::calculateUTCOffset):
 827
 828 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 829
 830         Reviewed by Geoff Garen.
 831
 832         Timezone calculation incorrect in Venezuela.
 833
 834         https://bugs.webkit.org/show_bug.cgi?id=26531
 835         <rdar://problem/6646169> Time is incorrectly reported to JavaScript in both Safari 3 and Firefox 3
 836
 837         The problem is that we're calculating the timezone relative to 01/01/2000,
 838         but the VET timezone changed from -4 hours to -4:30 hours on 12/09/2007.
 839         According to the spec, section 15.9.1.9 states "the time since the beginning
 840         of the year", presumably meaning the *current* year.  Change the calculation
 841         to be based on whatever the current year is, rather than a canned date.
 842
 843         No performance impact.
 844
 845         * wtf/DateMath.cpp:
 846         (WTF::calculateUTCOffset):
 847
 848 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
 849
 850         Rubber Stamped by Mark Rowe (originally reviewed by Sam Weinig).
 851
 852         (Reintroducing patch added in r44492, and reverted in r44796.)
 853
 854         Change the implementation of op_throw so the stub function always modifies its
 855         return address - if it doesn't find a 'catch' it will switch to a trampoline
 856         to force a return from JIT execution.  This saves memory, by avoiding the need
 857         for a unique return for every op_throw.
 858
 859         * jit/JITOpcodes.cpp:
 860         (JSC::JIT::emit_op_throw):
 861             JITStubs::cti_op_throw now always changes its return address,
 862             remove return code generated after the stub call (this is now
 863             handled by ctiOpThrowNotCaught).
 864         * jit/JITStubs.cpp:
 865         (JSC::):
 866             Add ctiOpThrowNotCaught definitions.
 867         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
 868             Change cti_op_throw to always change its return address.
 869         * jit/JITStubs.h:
 870             Add ctiOpThrowNotCaught declaration.
 871
 872 2009-06-18  Kevin McCullough  <kmccullough@apple.com>
 873
 874         Reviewed by Oliver Hunt.
 875
 876         <rdar://problem/6940880> REGRESSION: Breakpoints don't break in 64-bit
 877
 878         - Exposed functions now needed by WebCore.
 879
 880         * JavaScriptCore.exp:
 881
 882 2009-06-17  Darin Adler  <darin@apple.com>
 883
 884         Reviewed by Oliver Hunt.
 885
 886         Bug 26429: Make JSON.stringify non-recursive so it can handle objects
 887         of arbitrary complexity
 888         https://bugs.webkit.org/show_bug.cgi?id=26429
 889
 890         For marking I decided not to use gcProtect, because this is inside the engine
 891         so it's easy enough to just do marking. And that darned gcProtect does locking!
 892         Oliver tried to convince me to used MarkedArgumentBuffer, but the constructor
 893         for that class says "FIXME: Remove all clients of this API, then remove this API."
 894
 895         * runtime/Collector.cpp:
 896         (JSC::Heap::collect): Add a call to JSONObject::markStringifiers.
 897
 898         * runtime/CommonIdentifiers.cpp:
 899         (JSC::CommonIdentifiers::CommonIdentifiers): Added emptyIdentifier.
 900         * runtime/CommonIdentifiers.h: Ditto.
 901
 902         * runtime/JSGlobalData.cpp:
 903         (JSC::JSGlobalData::JSGlobalData): Initialize firstStringifierToMark to 0.
 904         * runtime/JSGlobalData.h: Added firstStringifierToMark.
 905
 906         * runtime/JSONObject.cpp: Cut down the includes to the needed ones only.
 907         (JSC::unwrapNumberOrString): Added. Helper for unwrapping number and string
 908         objects to get their number and string values.
 909         (JSC::ReplacerPropertyName::ReplacerPropertyName): Added. The class is used
 910         to wrap an identifier or integer so we don't have to do any work unless we
 911         actually call a replacer.
 912         (JSC::ReplacerPropertyName::value): Added.
 913         (JSC::gap): Added. Helper function for the Stringifier constructor.
 914         (JSC::PropertyNameForFunctionCall::PropertyNameForFunctionCall): Added.
 915         The class is used to wrap an identifier or integer so we don't have to
 916         allocate a number or string until we actually call toJSON or a replacer.
 917         (JSC::PropertyNameForFunctionCall::asJSValue): Added.
 918         (JSC::Stringifier::Stringifier): Updated and moved out of the class
 919         definition. Added code to hook this into a singly linked list for marking.
 920         (JSC::Stringifier::~Stringifier): Remove from the singly linked list.
 921         (JSC::Stringifier::mark): Mark all the objects in the holder stacks.
 922         (JSC::Stringifier::stringify): Updated.
 923         (JSC::Stringifier::appendQuotedString): Tweaked and streamlined a bit.
 924         (JSC::Stringifier::toJSON): Renamed from toJSONValue.
 925         (JSC::Stringifier::appendStringifiedValue): Renamed from stringify.
 926         Added code to use the m_holderStack to do non-recursive stringify of
 927         objects and arrays. This code also uses the timeout checker since in
 928         pathological cases it could be slow even without calling into the
 929         JavaScript virtual machine.
 930         (JSC::Stringifier::willIndent): Added.
 931         (JSC::Stringifier::indent): Added.
 932         (JSC::Stringifier::unindent): Added.
 933         (JSC::Stringifier::startNewLine): Added.
 934         (JSC::Stringifier::Holder::Holder): Added.
 935         (JSC::Stringifier::Holder::appendNextProperty): Added. This is the
 936         function that handles the format of arrays and objects.
 937         (JSC::JSONObject::getOwnPropertySlot): Moved this down to the bottom
 938         of the file so the JSONObject class is not interleaved with the
 939         Stringifier class.
 940         (JSC::JSONObject::markStringifiers): Added. Calls mark.
 941         (JSC::JSONProtoFuncStringify): Streamlined the code here. The code
 942         to compute the gap string is now a separate function.
 943
 944         * runtime/JSONObject.h: Made everything private. Added markStringifiers.
 945
 946 2009-06-17  Oliver Hunt  <oliver@apple.com>
 947
 948         Reviewed by Gavin Barraclough.
 949
 950         <rdar://problem/6974140> REGRESSION(r43849): Crash in cti_op_call_NotJSFunction when getting directions on maps.google.com
 951
 952         Roll out r43849 as it appears that we cannot rely on the address of
 953         an objects property storage being constant even if the structure is
 954         unchanged.
 955
 956         * jit/JIT.h:
 957         * jit/JITPropertyAccess.cpp:
 958         (JSC::JIT::compileGetDirectOffset):
 959         (JSC::JIT::privateCompileGetByIdProto):
 960         (JSC::JIT::privateCompileGetByIdProtoList):
 961         (JSC::JIT::privateCompileGetByIdChainList):
 962         (JSC::JIT::privateCompileGetByIdChain):
 963
 964 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
 965
 966         Rubber Stamped by Mark Rowe.
 967
 968         Fully revert r44492 & r44748 while we fix a bug they cause on internal builds <rdar://problem/6955963>.
 969
 970         * jit/JITOpcodes.cpp:
 971         (JSC::JIT::emit_op_throw):
 972         * jit/JITStubs.cpp:
 973         (JSC::):
 974         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
 975         * jit/JITStubs.h:
 976
 977 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
 978
 979         Reviewed by Mark Rowe.
 980
 981         <rdar://problem/6947426> sunspider math-cordic.js exhibits different intermediate results running 32-bit vs. 64-bit
 982
 983         On 64-bit, NaN-encoded values must be detagged before they can be used in rshift.
 984
 985         No performance impact.
 986
 987         * jit/JITArithmetic.cpp:
 988         (JSC::JIT::emit_op_rshift):
 989
 990 2009-06-17  Adam Treat  <adam.treat@torchmobile.com>
 991
 992         Reviewed by George Staikos.
 993
 994         https://bugs.webkit.org/show_bug.cgi?id=23155
 995         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
 996
 997         * jsc.cpp:
 998         (main):
 999
1000 2009-06-17  George Staikos  <george.staikos@torchmobile.com>
1001
1002         Reviewed by Adam Treat.
1003
1004         https://bugs.webkit.org/show_bug.cgi?id=23155
1005         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
1006
1007         * config.h:
1008         * jsc.cpp:
1009         * wtf/Assertions.cpp:
1010         * wtf/Assertions.h:
1011         * wtf/CurrentTime.cpp:
1012         (WTF::lowResUTCTime):
1013         * wtf/DateMath.cpp:
1014         (WTF::getLocalTime):
1015         * wtf/MathExtras.h:
1016         * wtf/Platform.h:
1017         * wtf/StringExtras.h:
1018         * wtf/Threading.h:
1019         * wtf/win/MainThreadWin.cpp:
1020
1021 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
1022
1023         Reviewed by Oliver Hunt.
1024
1025         <rdar://problem/6974175> ASSERT in JITStubs.cpp at appsaccess.apple.com
1026
1027         Remove PropertySlot::putValue - PropertySlots should only be used for getting,
1028         not putting.  Rename JSGlobalObject::getOwnPropertySlot to hasOwnPropertyForWrite,
1029         which is what it really was being used to ask, and remove some other getOwnPropertySlot
1030         & getOwnPropertySlotForWrite methods, which were unused and likely to lead to confusion.
1031
1032         * runtime/JSGlobalObject.h:
1033         (JSC::JSGlobalObject::hasOwnPropertyForWrite):
1034         * runtime/JSObject.h:
1035         * runtime/JSStaticScopeObject.cpp:
1036         * runtime/JSStaticScopeObject.h:
1037         * runtime/PropertySlot.h:
1038
1039 2009-06-16  Gavin Barraclough  <barraclough@apple.com>
1040
1041         Reviewed by Oliver hunt.
1042
1043         Temporarily partially disable r44492, since this is causing some problems on internal builds.
1044
1045         * jit/JITOpcodes.cpp:
1046         (JSC::JIT::emit_op_throw):
1047         * jit/JITStubs.cpp:
1048         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
1049
1050 2009-06-16  Sam Weinig  <sam@webkit.org>
1051
1052         Fix windows build.
1053
1054         * jit/JIT.cpp:
1055         (JSC::JIT::JIT):
1056
1057 2009-06-16  Sam Weinig  <sam@webkit.org>
1058
1059         Reviewed by Oliver Hunt.
1060
1061         Initialize m_bytecodeIndex to -1 in JIT, and correctly initialize
1062         it for each type of stub using the return address to find the correct
1063         offset.
1064
1065         * jit/JIT.cpp:
1066         (JSC::JIT::JIT):
1067         * jit/JIT.h:
1068         (JSC::JIT::compileGetByIdProto):
1069         (JSC::JIT::compileGetByIdSelfList):
1070         (JSC::JIT::compileGetByIdProtoList):
1071         (JSC::JIT::compileGetByIdChainList):
1072         (JSC::JIT::compileGetByIdChain):
1073         (JSC::JIT::compilePutByIdTransition):
1074         (JSC::JIT::compileCTIMachineTrampolines):
1075         (JSC::JIT::compilePatchGetArrayLength):
1076         * jit/JITStubCall.h:
1077         (JSC::JITStubCall::call):
1078
1079 == Rolled over to ChangeLog-2009-06-16 ==