JavaScriptCore/ChangeLog

   1 2009-07-14  Steve Falkenburg  <sfalken@apple.com>
   2
   3         Reorganize JavaScriptCore headers into:
   4         API: include/JavaScriptCore/
   5         Private: include/private/JavaScriptCore/
   6
   7         Reviewed by Darin Adler.
   8
   9         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
  10         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCoreCommon.vsprops:
  11         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCoreGenerated.make:
  12         * JavaScriptCore.vcproj/testapi/testapi.vcproj:
  13         * JavaScriptCore.vcproj/testapi/testapiCommon.vsprops:
  14
  15 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  16
  17         Reviewed by Darin Adler.
  18
  19         Change JSCell's superclass to NoncopyableCustomAllocated
  20         https://bugs.webkit.org/show_bug.cgi?id=27248
  21
  22         JSCell class customizes operator new, since Noncopyable will be
  23         inherited from FastAllocBase, NoncopyableCustomAllocated has
  24         to be used.
  25
  26         * runtime/JSCell.h:
  27
  28 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  29
  30         Reviewed by Darin Adler.
  31
  32         Change all Noncopyable inheriting visibility to public.
  33         https://bugs.webkit.org/show_bug.cgi?id=27225
  34
  35         Change all Noncopyable inheriting visibility to public because
  36         it is needed to the custom allocation framework (bug #20422).
  37
  38         * bytecode/SamplingTool.h:
  39         * bytecompiler/RegisterID.h:
  40         * interpreter/CachedCall.h:
  41         * interpreter/RegisterFile.h:
  42         * parser/Lexer.h:
  43         * parser/Parser.h:
  44         * runtime/ArgList.h:
  45         * runtime/BatchedTransitionOptimizer.h:
  46         * runtime/Collector.h:
  47         * runtime/CommonIdentifiers.h:
  48         * runtime/JSCell.h:
  49         * runtime/JSGlobalObject.h:
  50         * runtime/JSLock.h:
  51         * runtime/JSONObject.cpp:
  52         * runtime/SmallStrings.cpp:
  53         * runtime/SmallStrings.h:
  54         * wtf/CrossThreadRefCounted.h:
  55         * wtf/GOwnPtr.h:
  56         * wtf/Locker.h:
  57         * wtf/MessageQueue.h:
  58         * wtf/OwnArrayPtr.h:
  59         * wtf/OwnFastMallocPtr.h:
  60         * wtf/OwnPtr.h:
  61         * wtf/RefCounted.h:
  62         * wtf/ThreadSpecific.h:
  63         * wtf/Threading.h:
  64         * wtf/Vector.h:
  65         * wtf/unicode/Collator.h:
  66
  67 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  68
  69         Reviewed by Darin Adler.
  70
  71         Change ParserArenaRefCounted's superclass to RefCountedCustomAllocated
  72         https://bugs.webkit.org/show_bug.cgi?id=27249
  73
  74         ParserArenaDeletable customizes operator new, to avoid double inheritance
  75         ParserArenaDeletable's superclass has been changed to RefCountedCustomAllocated.
  76
  77         * parser/Nodes.h:
  78
  79 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  80
  81         Reviewed by Darin Adler.
  82
  83         Add RefCountedCustomAllocated to RefCounted.h
  84         https://bugs.webkit.org/show_bug.cgi?id=27232
  85
  86         Some class which are inherited from RefCounted customize
  87         operator new, but RefCounted is inherited from Noncopyable
  88         which will be inherited from FastAllocBase. To avoid
  89         conflicts Noncopyable inheriting was moved down to RefCounted
  90         and to avoid double inheritance this class has been added.
  91
  92         * wtf/RefCounted.h:
  93         (WTF::RefCountedCustomAllocated::deref):
  94         (WTF::RefCountedCustomAllocated::~RefCountedCustomAllocated):
  95
  96 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
  97
  98         Reviewed by Darin Adler.
  99
 100         Add NoncopyableCustomAllocated to Noncopyable.h.
 101         https://bugs.webkit.org/show_bug.cgi?id=27228
 102
 103         Some classes which inherited from Noncopyable overrides operator new
 104         since Noncopyable'll be inherited from FastAllocBase, Noncopyable.h
 105         needs to be extended with this new class to support the overriding.
 106
 107         * wtf/Noncopyable.h:
 108         (WTFNoncopyable::NoncopyableCustomAllocated::NoncopyableCustomAllocated):
 109         (WTFNoncopyable::NoncopyableCustomAllocated::~NoncopyableCustomAllocated):
 110
 111 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 112
 113         Reviewed by Darin Adler.
 114
 115         Allow custom memory allocation control for JavaScriptCore's IdentifierTable class
 116         https://bugs.webkit.org/show_bug.cgi?id=27260
 117
 118         Inherits IdentifierTable class from FastAllocBase because it has been
 119         instantiated by 'new' in JavaScriptCore/runtime/Identifier.cpp:70.
 120
 121         * runtime/Identifier.cpp:
 122
 123 2009-07-14  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 124
 125         Reviewed by Darin Adler.
 126
 127         Allow custom memory allocation control for JavaScriptCore's Profiler class
 128         https://bugs.webkit.org/show_bug.cgi?id=27253
 129
 130         Inherits Profiler class from FastAllocBase because it has been instantiated by
 131         'new' in JavaScriptCore/profiler/Profiler.cpp:56.
 132
 133         * profiler/Profiler.h:
 134
 135 2009-07-06  George Staikos  <george.staikos@torchmobile.com>
 136
 137         Reviewed by Adam Treat.
 138
 139         Authors: George Staikos <george.staikos@torchmobile.com>, Joe Mason <joe.mason@torchmobile.com>, Makoto Matsumoto <matumoto@math.keio.ac.jp>, Takuji Nishimura
 140
 141         https://bugs.webkit.org/show_bug.cgi?id=27030
 142         Implement custom RNG for WinCE using Mersenne Twister
 143
 144         * wtf/RandomNumber.cpp:
 145         (WTF::randomNumber):
 146         * wtf/RandomNumberSeed.h:
 147         (WTF::initializeRandomNumberGenerator):
 148         * wtf/wince/mt19937ar.c: Added.
 149         (init_genrand):
 150         (init_by_array):
 151         (genrand_int32):
 152         (genrand_int31):
 153         (genrand_real1):
 154         (genrand_real2):
 155         (genrand_real3):
 156         (genrand_res53):
 157
 158 2009-07-13  Gustavo Noronha Silva  <gustavo.noronha@collabora.co.uk>
 159
 160         Unreviewed make dist build fix.
 161
 162         * GNUmakefile.am:
 163
 164 2009-07-13  Drew Wilson  <atwilson@google.com>
 165
 166         Reviewed by David Levin.
 167
 168         Add ENABLE(SHARED_WORKERS) flag and define SharedWorker APIs
 169         https://bugs.webkit.org/show_bug.cgi?id=26932
 170
 171         Added ENABLE(SHARED_WORKERS) flag (off by default).
 172
 173         * Configurations/FeatureDefines.xcconfig:
 174
 175 2009-07-07  Norbert Leser  <norbert.leser@nokia.com>
 176
 177         Reviewed by Maciej Stachoviak.
 178
 179         https://bugs.webkit.org/show_bug.cgi?id=27058
 180
 181         Removed superfluous parenthesis around single expression.
 182         Compilers on Symbian platform fail to properly parse and compile.
 183
 184         * JavaScriptCore/wtf/Platform.h:
 185
 186 2009-07-13  Norbert Leser  <norbert.leser@nokia.com>
 187
 188         Reviewed by Maciej Stachoviak.
 189
 190         https://bugs.webkit.org/show_bug.cgi?id=27054
 191
 192         Renamed Translator to HashTranslator
 193
 194         Codewarrior compiler (WINSCW) latest b482 cannot resolve typename
 195         mismatch between template declaration and definition
 196         (HashTranslator / Translator)
 197
 198         * wtf/HashSet.h:
 199
 200 2009-07-13  Norbert Leser  <norbert.leser@nokia.com>
 201
 202         Reviewed by Eric Seidel.
 203
 204         https://bugs.webkit.org/show_bug.cgi?id=27053
 205
 206         Ambiguity in LabelScope initialization
 207
 208         Codewarrior compiler (WINSCW) latest b482 on Symbian cannot resolve
 209         type of "0" unambiguously. Set expression explicitly to
 210         PassRefPtr<Label>::PassRefPtr()
 211
 212         * bytecompiler/BytecodeGenerator.cpp
 213
 214 2009-07-11  Simon Fraser  <simon.fraser@apple.com>
 215
 216         Enable support for accelerated compositing and 3d transforms on Leopard.
 217         <https://bugs.webkit.org/show_bug.cgi?id=20166>
 218         <rdar://problem/6120614>
 219
 220         Reviewed by Oliver Hunt.
 221
 222         * Configurations/FeatureDefines.xcconfig:
 223         * wtf/Platform.h:
 224
 225 2009-07-10  Mark Rowe  <mrowe@apple.com>
 226
 227         Second part of the "make Windows happier" dance.
 228
 229         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 230         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 231
 232 2009-07-10  Mark Rowe  <mrowe@apple.com>
 233
 234         Try and make the Windows build happy.
 235
 236         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 237         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
 238
 239 2009-07-10  Kevin McCullough  <kmccullough@apple.com>
 240
 241         Reviewed by Geoffrey Garen.
 242
 243         * debugger/Debugger.h: Made this function virtual for use in WebCore's
 244         WebInspector.
 245
 246 2009-07-10  Kwang Yul Seo  <skyul@company100.net>
 247
 248         Reviewed by Darin Adler.
 249
 250         ParserArenaDeletable should override delete
 251         https://bugs.webkit.org/show_bug.cgi?id=26790
 252
 253         ParserArenaDeletable overrides new, but it does not override delete.
 254         ParserArenaDeletable must be freed by fastFree
 255         because it is allocated by fastMalloc.
 256
 257         * parser/NodeConstructors.h:
 258         (JSC::ParserArenaDeletable::operator delete):
 259         * parser/Nodes.h:
 260
 261 2009-07-10  Adam Roben  <aroben@apple.com>
 262
 263         Sort all our Xcode projects
 264
 265         Accomplished using sort-Xcode-project-file.
 266
 267         Requested by Dave Kilzer.
 268
 269         * JavaScriptCore.xcodeproj/project.pbxproj:
 270
 271 2009-07-09  Maciej Stachowiak  <mjs@apple.com>
 272
 273         Not reviewed, build fix.
 274
 275         Windows build fix for the last change.
 276
 277         * wtf/dtoa.cpp: Forgot to include Vector.h
 278
 279 2009-07-09  Maciej Stachowiak  <mjs@apple.com>
 280
 281         Reviewed by Darin Adler.
 282
 283         REGRESSION: crash in edge cases of floating point parsing.
 284         https://bugs.webkit.org/show_bug.cgi?id=27110
 285         <rdar://problem/7044458>
 286
 287         Tests: fast/css/number-parsing-crash.html
 288                fast/css/number-parsing-crash.html
 289                fast/js/number-parsing-crash.html
 290
 291         * wtf/dtoa.cpp:
 292         (WTF::BigInt::BigInt): Converted this to more a proper class, using a Vector
 293         with inline capacity
 294
 295         (WTF::lshift): Rearranged logic somewhat nontrivially to deal with the new way of sizing BigInts.
 296         Added an assertion to verify that invariants are maintained.
 297
 298         All other functions are adapted fairly mechanically to the above changes.
 299         (WTF::BigInt::clear):
 300         (WTF::BigInt::size):
 301         (WTF::BigInt::resize):
 302         (WTF::BigInt::words):
 303         (WTF::BigInt::append):
 304         (WTF::multadd):
 305         (WTF::s2b):
 306         (WTF::i2b):
 307         (WTF::mult):
 308         (WTF::cmp):
 309         (WTF::diff):
 310         (WTF::b2d):
 311         (WTF::d2b):
 312         (WTF::ratio):
 313         (WTF::strtod):
 314         (WTF::quorem):
 315         (WTF::dtoa):
 316
 317 2009-07-09  Drew Wilson  <atwilson@google.com>
 318
 319         Reviewed by Alexey Proskuryakov.
 320
 321         Turned on CHANNEL_MESSAGING by default because the MessageChannel API
 322         can now be implemented for Web Workers and is reasonably stable.
 323
 324         * Configurations/FeatureDefines.xcconfig:
 325
 326 2009-07-09  Oliver Hunt  <oliver@apple.com>
 327
 328         Reviewed by NOBODY (Build fix).
 329
 330         * interpreter/Interpreter.cpp:
 331         (JSC::Interpreter::privateExecute):
 332
 333 2009-07-09  Oliver Hunt  <oliver@apple.com>
 334
 335         Reviewed by Darin Adler.
 336
 337         Bug 27016 - Interpreter crashes due to invalid array indexes
 338         <https://bugs.webkit.org/show_bug.cgi?id=27016>
 339
 340         Unsigned vs signed conversions results in incorrect behaviour in
 341         64bit interpreter builds.
 342
 343         * interpreter/Interpreter.cpp:
 344         (JSC::Interpreter::privateExecute):
 345
 346 2009-07-09  Dimitri Glazkov  <dglazkov@chromium.org>
 347
 348         Reviewed by Darin Fisher.
 349
 350         [Chromium] Upstream JavaScriptCore.gypi, the project file for Chromium build.
 351         https://bugs.webkit.org/show_bug.cgi?id=27135
 352
 353         * JavaScriptCore.gypi: Added.
 354
 355 2009-07-09  Joe Mason  <joe.mason@torchmobile.com>
 356
 357         Reviewed by George Staikos.
 358
 359         Authors: Yong Li <yong.li@torchmobile.com>, Joe Mason <joe.mason@torchmobile.com>
 360
 361         https://bugs.webkit.org/show_bug.cgi?id=27031
 362         Add an override for deleteOwnedPtr(HDC) on Windows
 363
 364         * wtf/OwnPtrCommon.h:
 365         * wtf/OwnPtrWin.cpp:
 366         (WTF::deleteOwnedPtr):
 367
 368 2009-07-09  Laszlo Gombos  <laszlo.1.gombos@nokia.com>
 369
 370         Reviewed by Darin Adler.
 371
 372         Guard singal.h dependency with HAVE(SIGNAL_H) to enable building jsc
 373         on SYMBIAN.
 374
 375         https://bugs.webkit.org/show_bug.cgi?id=27026
 376
 377         Based on Norbert Leser's work.
 378
 379         * jsc.cpp:
 380         (printUsageStatement):
 381         (parseArguments):
 382         * wtf/Platform.h:
 383
 384 2009-07-07  Gavin Barraclough  <barraclough@apple.com>
 385
 386         Reviewed by Sam Weinig.
 387
 388         Stop loading constants into the register file.
 389
 390         Instead, use high register values (highest bit bar the sign bit set) to indicate
 391         constants in the instruction stream, and when we encounter such a value load it
 392         directly from the CodeBlock.
 393
 394         Since constants are no longer copied into the register file, this patch renders
 395         the 'unexpected constant' mechanism redundant, and removes it.
 396
 397         2% improvement, thanks to Sam Weinig.
 398
 399         * bytecode/CodeBlock.cpp:
 400         (JSC::CodeBlock::dump):
 401         (JSC::CodeBlock::CodeBlock):
 402         (JSC::CodeBlock::mark):
 403         (JSC::CodeBlock::shrinkToFit):
 404         * bytecode/CodeBlock.h:
 405         (JSC::CodeBlock::isTemporaryRegisterIndex):
 406         (JSC::CodeBlock::constantRegister):
 407         (JSC::CodeBlock::isConstantRegisterIndex):
 408         (JSC::CodeBlock::getConstant):
 409         (JSC::ExecState::r):
 410         * bytecode/Opcode.h:
 411         * bytecompiler/BytecodeGenerator.cpp:
 412         (JSC::BytecodeGenerator::preserveLastVar):
 413         (JSC::BytecodeGenerator::BytecodeGenerator):
 414         (JSC::BytecodeGenerator::addConstantValue):
 415         (JSC::BytecodeGenerator::emitEqualityOp):
 416         (JSC::BytecodeGenerator::emitLoad):
 417         (JSC::BytecodeGenerator::emitResolveBase):
 418         (JSC::BytecodeGenerator::emitResolveWithBase):
 419         (JSC::BytecodeGenerator::emitNewError):
 420         * bytecompiler/BytecodeGenerator.h:
 421         (JSC::BytecodeGenerator::emitNode):
 422         * interpreter/CallFrame.h:
 423         (JSC::ExecState::noCaller):
 424         (JSC::ExecState::hasHostCallFrameFlag):
 425         (JSC::ExecState::addHostCallFrameFlag):
 426         (JSC::ExecState::removeHostCallFrameFlag):
 427         * interpreter/Interpreter.cpp:
 428         (JSC::Interpreter::resolve):
 429         (JSC::Interpreter::resolveSkip):
 430         (JSC::Interpreter::resolveGlobal):
 431         (JSC::Interpreter::resolveBase):
 432         (JSC::Interpreter::resolveBaseAndProperty):
 433         (JSC::Interpreter::resolveBaseAndFunc):
 434         (JSC::Interpreter::dumpRegisters):
 435         (JSC::Interpreter::throwException):
 436         (JSC::Interpreter::createExceptionScope):
 437         (JSC::Interpreter::privateExecute):
 438         (JSC::Interpreter::retrieveArguments):
 439         * jit/JIT.cpp:
 440         (JSC::JIT::privateCompileMainPass):
 441         * jit/JITInlineMethods.h:
 442         (JSC::JIT::emitLoadDouble):
 443         (JSC::JIT::emitLoadInt32ToDouble):
 444         * jit/JITOpcodes.cpp:
 445         (JSC::JIT::emit_op_new_error):
 446         (JSC::JIT::emit_op_enter):
 447         (JSC::JIT::emit_op_enter_with_activation):
 448         * parser/Nodes.cpp:
 449         (JSC::DeleteResolveNode::emitBytecode):
 450         (JSC::DeleteValueNode::emitBytecode):
 451         (JSC::PrefixResolveNode::emitBytecode):
 452         * runtime/JSActivation.cpp:
 453         (JSC::JSActivation::JSActivation):
 454         * wtf/Platform.h:
 455
 456 2009-07-07  Mark Rowe  <mrowe@apple.com>
 457
 458         Reviewed by Darin Adler.
 459
 460         Fix <https://bugs.webkit.org/show_bug.cgi?id=27025> / <rdar://problem/7033448>.
 461         Bug 27025: Crashes and regression test failures related to regexps in 64-bit
 462
 463         For x86_64 RegexGenerator uses rbx, a callee-save register, as a scratch register but
 464         neglects to save and restore it.  The change in handling of the output vector in r45545
 465         altered code generation so that the RegExp::match was now storing important data in rbx,
 466         which caused crashes and bogus results when it was clobbered.
 467
 468         * yarr/RegexJIT.cpp:
 469         (JSC::Yarr::RegexGenerator::generateEnter): Save rbx.
 470         (JSC::Yarr::RegexGenerator::generateReturn): Restore rbx.
 471
 472 2009-07-06  Ada Chan  <adachan@apple.com>
 473
 474         Reviewed by Darin Adler and Mark Rowe.
 475
 476         Decommitted spans are added to the list of normal spans rather than
 477         the returned spans in TCMalloc_PageHeap::Delete().
 478         https://bugs.webkit.org/show_bug.cgi?id=26998
 479
 480         In TCMalloc_PageHeap::Delete(), the deleted span can be decommitted in
 481         the process of merging with neighboring spans that are also decommitted.
 482         The merged span needs to be placed in the list of returned spans (spans
 483         whose memory has been returned to the system).  Right now it's always added
 484         to the list of the normal spans which can theoretically cause thrashing.
 485
 486         * wtf/FastMalloc.cpp:
 487         (WTF::TCMalloc_PageHeap::Delete):
 488
 489 2009-07-05  Lars Knoll  <lars.knoll@nokia.com>
 490
 491         Reviewed by Maciej Stachowiak.
 492
 493         https://bugs.webkit.org/show_bug.cgi?id=26843
 494
 495         Fix run-time crashes in JavaScriptCore with the Metrowerks compiler on Symbian.
 496
 497         The Metrowerks compiler on the Symbian platform moves the globally
 498         defined Hashtables into read-only memory, despite one of the members
 499         being mutable. This causes crashes at run-time due to write access to
 500         read-only memory.
 501
 502         Avoid the use of const with this compiler by introducing the
 503         JSC_CONST_HASHTABLE macro.
 504
 505         Based on idea by Norbert Leser.
 506
 507         * runtime/Lookup.h: Define JSC_CONST_HASHTABLE as const for !WINSCW.
 508         * create_hash_table: Use JSC_CONST_HASHTABLE for hashtables.
 509         * runtime/JSGlobalData.cpp: Import various global hashtables via the macro.
 510
 511 2009-07-04  Dan Bernstein  <mitz@apple.com>
 512
 513         - debug build fix
 514
 515         * runtime/RegExpConstructor.cpp:
 516         (JSC::RegExpConstructor::getLastParen):
 517
 518 2009-07-03  Yong Li  <yong.li@torchmobile.com>
 519
 520         Reviewed by Maciej Stachowiak (and revised slightly)
 521
 522         RegExp::match to be optimized
 523         https://bugs.webkit.org/show_bug.cgi?id=26957
 524
 525         Allow regexp matching to use Vectors with inline capacity instead of
 526         allocating a new ovector buffer every time.
 527
 528         ~5% speedup on SunSpider string-unpack-code test, 0.3% on SunSpider overall.
 529
 530         * runtime/RegExp.cpp:
 531         (JSC::RegExp::match):
 532         * runtime/RegExp.h:
 533         * runtime/RegExpConstructor.cpp:
 534         (JSC::RegExpConstructorPrivate::RegExpConstructorPrivate):
 535         (JSC::RegExpConstructorPrivate::lastOvector):
 536         (JSC::RegExpConstructorPrivate::tempOvector):
 537         (JSC::RegExpConstructorPrivate::changeLastOvector):
 538         (JSC::RegExpConstructor::performMatch):
 539         (JSC::RegExpMatchesArray::RegExpMatchesArray):
 540         (JSC::RegExpMatchesArray::fillArrayInstance):
 541         (JSC::RegExpConstructor::getBackref):
 542         (JSC::RegExpConstructor::getLastParen):
 543         (JSC::RegExpConstructor::getLeftContext):
 544         (JSC::RegExpConstructor::getRightContext):
 545         * runtime/StringPrototype.cpp:
 546         (JSC::stringProtoFuncSplit):
 547
 548 2009-06-30  Kwang Yul Seo  <skyul@company100.net>
 549
 550         Reviewed by Eric Seidel.
 551
 552         Override operator new/delete with const std::nothrow_t& as the second
 553         argument.
 554         https://bugs.webkit.org/show_bug.cgi?id=26792
 555
 556         On Windows CE, operator new/delete, new[]/delete[] with const
 557         std::nothrow_t& must be overrided because some standard template
 558         libraries use these operators.
 559
 560         The problem occurs when memory allocated by new(size_t s, const
 561         std::nothrow_t&) is freed by delete(void* p). This causes the umatched
 562         malloc/free problem.
 563
 564         The patch overrides all new, delete, new[] and delete[] to use
 565         fastMaloc and fastFree consistently.
 566
 567         * wtf/FastMalloc.h:
 568         (throw):
 569
 570 2009-06-30  Gabor Loki  <loki@inf.u-szeged.hu>
 571
 572         Reviewed by Sam Weinig.
 573
 574         <https://bugs.webkit.org/show_bug.cgi?id=24986>
 575
 576         Remove unnecessary references to AssemblerBuffer.
 577
 578         * interpreter/Interpreter.cpp:
 579         * interpreter/Interpreter.h:
 580
 581 2009-06-29  David Levin  <levin@chromium.org>
 582
 583         Reviewed by Oliver Hunt.
 584
 585         Still seeing occasional leaks from UString::sharedBuffer code
 586         https://bugs.webkit.org/show_bug.cgi?id=26420
 587
 588         The problem is that the pointer to the memory allocation isn't visible
 589         by "leaks" due to the lower bits being used as flags.  The fix is to
 590         make the pointer visible in memory (in debug only). The downside of
 591         this fix that the memory allocated by sharedBuffer will still look like
 592         a leak in non-debug builds when any flags are set.
 593
 594         * wtf/PtrAndFlags.h:
 595         (WTF::PtrAndFlags::set):
 596
 597 2009-06-29  Sam Weinig  <sam@webkit.org>
 598
 599         Reviewed by Mark Rowe.
 600
 601         Remove more unused scons support.
 602
 603         * SConstruct: Removed.
 604
 605 2009-06-29  Oliver Hunt  <oliver@apple.com>
 606
 607         Reviewed by Gavin Barraclough.
 608
 609         <rdar://problem/7016214> JSON.parse fails to parse valid JSON with most Unicode characters
 610         <https://bugs.webkit.org/show_bug.cgi?id=26802>
 611
 612         In the original JSON.parse patch unicode was handled correctly, however in some last
 613         minute "clean up" I oversimplified isSafeStringCharacter.  This patch corrects this bug.
 614
 615         * runtime/LiteralParser.cpp:
 616         (JSC::isSafeStringCharacter):
 617         (JSC::LiteralParser::Lexer::lexString):
 618
 619 2009-06-26  Oliver Hunt  <oliver@apple.com>
 620
 621         Reviewed by Dan Bernstein.
 622
 623         <rdar://problem/7009684> REGRESSION(r45039): Crashes inside JSEvent::put on PowerPC (26746)
 624         <https://bugs.webkit.org/show_bug.cgi?id=26746>
 625
 626         Fix for r45039 incorrectly uncached a get_by_id by converting it to put_by_id.  Clearly this
 627         is less than correct.  This patch corrects that error.
 628
 629         * interpreter/Interpreter.cpp:
 630         (JSC::Interpreter::tryCacheGetByID):
 631
 632 2009-06-26  Eric Seidel  <eric@webkit.org>
 633
 634         No review, only rolling out r45259.
 635
 636         Roll out r45259 after crash appeared on the bots:
 637         plugins/undefined-property-crash.html
 638         ASSERTION FAILED: s <= HeapConstants<heapType>::cellSize
 639         (leopard-intel-debug-tests/build/JavaScriptCore/runtime/Collector.cpp:278
 640         void* JSC::Heap::heapAllocate(size_t) [with JSC::HeapType heapType = PrimaryHeap])
 641
 642         * runtime/DateInstance.cpp:
 643         * runtime/Identifier.cpp:
 644         * runtime/Lookup.h:
 645         * runtime/RegExpConstructor.cpp:
 646         * runtime/RegExpObject.h:
 647         * runtime/ScopeChain.h:
 648         * runtime/UString.h:
 649
 650 2009-06-26  Jedrzej Nowacki  <jedrzej.nowacki@nokia.com>
 651
 652         Reviewed by Simon Hausmann.
 653
 654         Add support for QDataStream operators to Vector.
 655
 656         * wtf/Vector.h:
 657         (WTF::operator<<):
 658         (WTF::operator>>):
 659
 660 2009-06-24  Sam Weinig  <sam@webkit.org>
 661
 662         Reviewed by Gavin Barraclough.
 663
 664         Make the opcode sampler work once again.
 665
 666         * jit/JIT.h:
 667         (JSC::JIT::compileGetByIdProto):
 668         (JSC::JIT::compileGetByIdSelfList):
 669         (JSC::JIT::compileGetByIdProtoList):
 670         (JSC::JIT::compileGetByIdChainList):
 671         (JSC::JIT::compileGetByIdChain):
 672         (JSC::JIT::compilePutByIdTransition):
 673         (JSC::JIT::compileCTIMachineTrampolines):
 674         (JSC::JIT::compilePatchGetArrayLength):
 675         * jit/JITStubCall.h:
 676         (JSC::JITStubCall::call):
 677
 678 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 679
 680         Reviewed by Maciej Stachowiak.
 681
 682         Extend FastAllocBase.h with 'using WTF::FastAllocBase' to avoid
 683         unnecessary WTF:: usings.
 684         Remove existing unnecessary WTF:: usings.
 685
 686         * interpreter/Interpreter.h:
 687         * profiler/CallIdentifier.h:
 688         * runtime/ScopeChain.h:
 689         * wtf/FastAllocBase.h:
 690
 691 2009-06-24  David Levin  <levin@chromium.org>
 692
 693         Fix all builds.
 694
 695         * bytecode/CodeBlock.h:
 696         * bytecompiler/BytecodeGenerator.h:
 697         * interpreter/Register.h:
 698
 699 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 700
 701         Reviewed by Maciej Stachowiak.
 702
 703         https://bugs.webkit.org/show_bug.cgi?id=26677
 704
 705         Inherits CodeBlock class from FastAllocBase because it
 706         has been instantiated by 'new' in JavaScriptCore/bytecode/CodeBlock.h:217.
 707
 708         * bytecode/CodeBlock.h:
 709
 710 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 711
 712         Reviewed by Maciej Stachowiak.
 713
 714         https://bugs.webkit.org/show_bug.cgi?id=26676
 715
 716         Inherits BytecodeGenerator class from FastAllocBase because it has been
 717         instantiated by 'new' in JavaScriptCore/parser/Nodes.cpp:1892.
 718
 719         * bytecompiler/BytecodeGenerator.h:
 720
 721 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 722
 723         Reviewed by Maciej Stachowiak.
 724
 725         https://bugs.webkit.org/show_bug.cgi?id=26675
 726
 727         Inherits Register class from FastAllocBase because it has been
 728         instantiated by 'new' in JavaScriptCore/runtime/JSVariableObject.h:149.
 729
 730         * interpreter/Register.h:
 731
 732 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 733
 734         Reviewed by Darin Adler.
 735
 736         https://bugs.webkit.org/show_bug.cgi?id=26674
 737
 738         Inherits HashMap class from FastAllocBase because it has been
 739         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:458.
 740
 741         * wtf/HashMap.h:
 742
 743 2009-06-24  Oliver Hunt  <oliver@apple.com>
 744
 745         Reviewed by Darin Adler.
 746
 747         <rdar://problem/6940519> REGRESSION (Safari 4 Public Beta - TOT): google.com/adplanner shows blank page instead of site details in "basic research'
 748
 749         The problem was caused by the page returned with a function using a
 750         var declaration list containing around ~3000 variables.  The solution
 751         to this is to flatten the comma expression representation and make
 752         codegen comma expressions and initializer lists iterative rather than
 753         recursive.
 754
 755         * parser/Grammar.y:
 756         * parser/NodeConstructors.h:
 757         (JSC::CommaNode::CommaNode):
 758         * parser/Nodes.cpp:
 759         (JSC::CommaNode::emitBytecode):
 760         * parser/Nodes.h:
 761         (JSC::ExpressionNode::isCommaNode):
 762         (JSC::CommaNode::isCommaNode):
 763         (JSC::CommaNode::append):
 764
 765 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 766
 767         Reviewed by Maciej Stachowiak.
 768
 769         https://bugs.webkit.org/show_bug.cgi?id=26645
 770
 771         Inherits ScopeChainNode class from FastAllocBase because it has been
 772         instantiated by 'new' in JavaScriptCore/runtime/ScopeChain.h:95.
 773
 774         * wtf/RefPtr.h:
 775
 776 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 777
 778         Reviewed by Darin Adler.
 779
 780         https://bugs.webkit.org/show_bug.cgi?id=26648
 781
 782         Inherits Deque class from FastAllocBase because it has been
 783         instantiated by 'new' with DEFINE_STATIC_LOCAL macro in
 784         JavaScriptCore/wtf/MainThread.cpp:62.
 785
 786         * wtf/Deque.h:
 787
 788 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 789
 790         Reviewed by Darin Adler.
 791
 792         https://bugs.webkit.org/show_bug.cgi?id=26644
 793
 794         Inherits RefPtr class from FastAllocBase because it has been
 795         instantiated by 'new' in JavaScriptCore/runtime/StructureChain.cpp:41.
 796
 797         * wtf/RefPtr.h:
 798
 799 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 800
 801         Reviewed by Darin Adler.
 802
 803         Inherits HashSet class from FastAllocBase, because it has been
 804         instantiated by 'new' in JavaScriptCore/runtime/Collector.h:116.
 805
 806         * wtf/HashSet.h:
 807
 808 2009-06-24  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
 809
 810         Reviewed by Darin Adler.
 811
 812         Inherits Vector class from FastAllocBase because it has been
 813         instantiated by 'new' in JavaScriptCore/runtime/Structure.cpp:633.
 814
 815         * wtf/Vector.h:
 816
 817 2009-06-24  Norbert Leser  <norbert.leser@nokia.com>
 818
 819         Reviewed by Maciej Stachoviak.
 820
 821         The BytecodeGenerator objects were instantiated on stack, which takes up ~38kB per instance
 822         (each instance includes copy of JSC::CodeBlock with large SymbolTable, etc.).
 823         Specifically, since there is nested invocation (e.g., GlobalCode --> FunctionCode),
 824         the stack overflows immediately on Symbian hardware (max. 80 kB).
 825         Proposed change allocates generator objects on heap.
 826         Performance impact (if any) should be negligible and change is proposed as general fix,
 827         rather than ifdef'd for SYMBIAN.
 828
 829         * parser/Nodes.cpp:
 830         (JSC::ProgramNode::generateBytecode):
 831         (JSC::EvalNode::generateBytecode):
 832         (JSC::EvalNode::bytecodeForExceptionInfoReparse):
 833         (JSC::FunctionBodyNode::generateBytecode):
 834         (JSC::FunctionBodyNode::bytecodeForExceptionInfoReparse):
 835
 836 2009-06-23  Oliver Hunt  <oliver@apple.com>
 837
 838         Reviewed by Gavin Barraclough.
 839
 840         <rdar://problem/6992806> REGRESSION: Enumeration can skip new properties in cases of prototypes that have more than 64 (26593)
 841         <https://bugs.webkit.org/show_bug.cgi?id=26593>
 842
 843         Do not attempt to cache structure chains if they contain a dictionary at any level.
 844
 845         * interpreter/Interpreter.cpp:
 846         (JSC::Interpreter::tryCachePutByID):
 847         (JSC::Interpreter::tryCacheGetByID):
 848         * jit/JITStubs.cpp:
 849         (JSC::JITThunks::tryCachePutByID):
 850         * runtime/Structure.cpp:
 851         (JSC::Structure::getEnumerablePropertyNames):
 852         (JSC::Structure::addPropertyTransition):
 853         * runtime/StructureChain.cpp:
 854         (JSC::StructureChain::isCacheable):
 855         * runtime/StructureChain.h:
 856
 857 2009-06-23  Yong Li  <yong.li@torchmobile.com>
 858
 859         Reviewed by George Staikos.
 860
 861         https://bugs.webkit.org/show_bug.cgi?id=26654
 862         Add the proper export define for the JavaScriptCore API when building for WINCE.
 863
 864         * API/JSBase.h:
 865
 866 2009-06-23  Joe Mason  <joe.mason@torchmobile.com>
 867
 868         Reviewed by Adam Treat.
 869
 870         Authors: Yong Li <yong.li@torchmobile.com>, Joe Mason <joe.mason@torchmobile.com>
 871
 872         https://bugs.webkit.org/show_bug.cgi?id=26611
 873         Implement currentThreadStackBase on WINCE by adding a global,
 874         g_stackBase, which must be set to the address of a local variable
 875         by the caller before calling any WebKit function that invokes JSC.
 876
 877         * runtime/Collector.cpp:
 878         (JSC::isPageWritable):
 879         (JSC::getStackBase):
 880           Starts at the top of the stack and returns the entire range of
 881           consecutive writable pages as an estimate of the actual stack.
 882           This will be much bigger than the actual stack range, so some
 883           dead objects can't be collected, but it guarantees live objects
 884           aren't collected prematurely.
 885
 886         (JSC::currentThreadStackBase):
 887           On WinCE, returns g_stackBase if set or call getStackBase as a
 888           fallback if not.
 889
 890 2009-06-23  Oliver Hunt  <oliver@apple.com>
 891
 892         Reviewed by Alexey Proskuryakov.
 893
 894         Fix stupid performance problem in the LiteralParser
 895
 896         The LiteralParser was making a new UString in order to use
 897         toDouble, however UString's toDouble allows a much wider range
 898         of numberic strings than the LiteralParser accepts, and requires
 899         an additional heap allocation or two for the construciton of the
 900         UString.  To rectify this we just call WTF::dtoa directly using
 901         a stack allocated buffer to hold the validated numeric literal.
 902
 903         * runtime/LiteralParser.cpp:
 904         (JSC::LiteralParser::Lexer::lexNumber):
 905         (JSC::LiteralParser::parse):
 906         * runtime/LiteralParser.h:
 907
 908 2009-06-22  Oliver Hunt  <oliver@apple.com>
 909
 910         Reviewed by Alexey Proskuryakov.
 911
 912         Bug 26640: JSON.stringify needs to special case Boolean objects
 913         <https://bugs.webkit.org/show_bug.cgi?id=26640>
 914
 915         Add special case handling of the Boolean object so we match current
 916         ES5 errata.
 917
 918         * runtime/JSONObject.cpp:
 919         (JSC::unwrapBoxedPrimitive): renamed from unwrapNumberOrString
 920         (JSC::gap):
 921         (JSC::Stringifier::appendStringifiedValue):
 922
 923 2009-06-22  Oliver Hunt  <oliver@apple.com>
 924
 925         Reviewed by Darin Adler.
 926
 927         Bug 26591: Support revivers in JSON.parse
 928         <https://bugs.webkit.org/show_bug.cgi?id=26591>
 929
 930         Add reviver support to JSON.parse.  This completes the JSON object.
 931
 932         * runtime/JSONObject.cpp:
 933         (JSC::Walker::Walker):
 934         (JSC::Walker::callReviver):
 935         (JSC::Walker::walk):
 936         (JSC::JSONProtoFuncParse):
 937
 938 2009-06-21  Oliver Hunt  <oliver@apple.com>
 939
 940         Reviewed by Darin Adler.
 941
 942         Bug 26592: Support standard toJSON functions
 943         <https://bugs.webkit.org/show_bug.cgi?id=26592>
 944
 945         Add support for the standard Date.toJSON function.
 946
 947         * runtime/DatePrototype.cpp:
 948         (JSC::dateProtoFuncToJSON):
 949
 950 2009-06-21  Oliver Hunt  <oliver@apple.com>
 951
 952         Reviewed by Sam Weinig.
 953
 954         Bug 26594: JSC needs to support Date.toISOString
 955         <https://bugs.webkit.org/show_bug.cgi?id=26594>
 956
 957         Add support for Date.toISOString.
 958
 959         * runtime/DatePrototype.cpp:
 960         (JSC::dateProtoFuncToISOString):
 961
 962 2009-06-21  Oliver Hunt  <oliver@apple.com>
 963
 964         Reviewed by NOBODY (Build fix).
 965
 966         Remove dead code.
 967
 968         * runtime/LiteralParser.cpp:
 969         (JSC::LiteralParser::parse):
 970
 971 2009-06-21  Oliver Hunt  <oliver@apple.com>
 972
 973         Reviewed by Darin Adler and Cameron Zwarich.
 974
 975         Bug 26587: Support JSON.parse
 976         <https://bugs.webkit.org/show_bug.cgi?id=26587>
 977
 978         Extend the LiteralParser to support the full strict JSON
 979         grammar, fix a few places where the grammar was incorrectly
 980         lenient.   Doesn't yet support the JSON.parse reviver function
 981         but that does not block the JSON.parse functionality itself.
 982
 983         * interpreter/Interpreter.cpp:
 984         (JSC::Interpreter::callEval):
 985         * runtime/JSGlobalObjectFunctions.cpp:
 986         (JSC::globalFuncEval):
 987         * runtime/JSONObject.cpp:
 988         (JSC::JSONProtoFuncParse):
 989         * runtime/LiteralParser.cpp:
 990         (JSC::LiteralParser::Lexer::lex):
 991         (JSC::isSafeStringCharacter):
 992         (JSC::LiteralParser::Lexer::lexString):
 993         (JSC::LiteralParser::parse):
 994         * runtime/LiteralParser.h:
 995         (JSC::LiteralParser::LiteralParser):
 996         (JSC::LiteralParser::tryJSONParse):
 997         (JSC::LiteralParser::):
 998         (JSC::LiteralParser::Lexer::Lexer):
 999
1000 2009-06-21  David Levin  <levin@chromium.org>
1001
1002         Reviewed by NOBODY (speculative build fix for windows).
1003
1004         Simply removed some whitespace form this file to make windows build wtf and
1005         hopefully copy the new MessageQueque.h so that WebCore picks it up.
1006
1007         * wtf/Assertions.cpp:
1008
1009 2009-06-21  Drew Wilson  <atwilson@google.com>
1010
1011         Reviewed by David Levin.
1012
1013         <https://bugs.webkit.org/show_bug.cgi?id=25043>
1014         Added support for multi-threaded MessagePorts.
1015
1016         * wtf/MessageQueue.h:
1017         (WTF::::appendAndCheckEmpty):
1018             Added API to test whether the queue was empty before adding an element.
1019
1020 2009-06-20  David D. Kilzer  <ddkilzer@webkit.org>
1021
1022         Fix namespace comment in SegmentedVector.h
1023
1024         * wtf/SegmentedVector.h: Updated namespace comment to reflect
1025         new namespace after r44897.
1026
1027 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
1028
1029         Bug 24986: ARM JIT port
1030         <https://bugs.webkit.org/show_bug.cgi?id=24986>
1031
1032         Reviewed by Oliver Hunt.
1033
1034         An Iterator added for SegmentedVector. Currently
1035         only the pre ++ operator is supported.
1036
1037         * wtf/SegmentedVector.h:
1038         (WTF::SegmentedVectorIterator::~SegmentedVectorIterator):
1039         (WTF::SegmentedVectorIterator::operator*):
1040         (WTF::SegmentedVectorIterator::operator->):
1041         (WTF::SegmentedVectorIterator::operator++):
1042         (WTF::SegmentedVectorIterator::operator==):
1043         (WTF::SegmentedVectorIterator::operator!=):
1044         (WTF::SegmentedVectorIterator::operator=):
1045         (WTF::SegmentedVectorIterator::SegmentedVectorIterator):
1046         (WTF::SegmentedVector::alloc):
1047         (WTF::SegmentedVector::begin):
1048         (WTF::SegmentedVector::end):
1049
1050 2009-06-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
1051
1052         Bug 24986: ARM JIT port
1053         <https://bugs.webkit.org/show_bug.cgi?id=24986>
1054
1055         Reviewed by Oliver Hunt.
1056
1057         Move SegmentedVector to /wtf subdirectory
1058         and change "namespace JSC" to "namespace WTF"
1059
1060         Additional build file updates by David Kilzer.
1061
1062         * GNUmakefile.am: Updated path to SegmentedVector.h.
1063         * JavaScriptCore.order: Updated SegmentedVector namespace from
1064         JSC to WTF in mangled C++ method name.
1065         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
1066         Removed reference to bytecompiler\SegmentedVector.h.
1067         * JavaScriptCore.vcproj/WTF/WTF.vcproj: Added reference to
1068         wtf\SegmentedVector.h.
1069         * JavaScriptCore.xcodeproj/project.pbxproj: Moved
1070         SegmentedVector.h definition from bytecompiler subdirectory to
1071         wtf subdirectory.
1072         * bytecompiler/BytecodeGenerator.h: Updated #include path to
1073         SegmentedVector.h and prepended WTF:: namespace to its use.
1074         * parser/Lexer.h: Ditto.
1075         * wtf/SegmentedVector.h: Renamed from JavaScriptCore/bytecompiler/SegmentedVector.h.
1076         (WTF::SegmentedVector::SegmentedVector):
1077         (WTF::SegmentedVector::~SegmentedVector):
1078         (WTF::SegmentedVector::size):
1079         (WTF::SegmentedVector::at):
1080         (WTF::SegmentedVector::operator[]):
1081         (WTF::SegmentedVector::last):
1082         (WTF::SegmentedVector::append):
1083         (WTF::SegmentedVector::removeLast):
1084         (WTF::SegmentedVector::grow):
1085         (WTF::SegmentedVector::clear):
1086         (WTF::SegmentedVector::deleteAllSegments):
1087         (WTF::SegmentedVector::segmentExistsFor):
1088         (WTF::SegmentedVector::segmentFor):
1089         (WTF::SegmentedVector::subscriptFor):
1090         (WTF::SegmentedVector::ensureSegmentsFor):
1091         (WTF::SegmentedVector::ensureSegment):
1092
1093 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
1094
1095         Reviewed by NOBODY (build fix take 2 - rename FIELD_OFFSET to something that doesn't conflict with winnt.h).
1096
1097         * jit/JIT.cpp:
1098         (JSC::JIT::privateCompile):
1099         (JSC::JIT::privateCompileCTIMachineTrampolines):
1100         (JSC::JIT::emitGetVariableObjectRegister):
1101         (JSC::JIT::emitPutVariableObjectRegister):
1102         * jit/JIT.h:
1103         * jit/JITArithmetic.cpp:
1104         (JSC::JIT::emit_op_rshift):
1105         (JSC::JIT::emitSlow_op_jnless):
1106         (JSC::JIT::emitSlow_op_jnlesseq):
1107         (JSC::JIT::compileBinaryArithOp):
1108         * jit/JITCall.cpp:
1109         (JSC::JIT::compileOpCallInitializeCallFrame):
1110         (JSC::JIT::compileOpCall):
1111         * jit/JITInlineMethods.h:
1112         (JSC::JIT::restoreArgumentReference):
1113         (JSC::JIT::checkStructure):
1114         * jit/JITOpcodes.cpp:
1115         (JSC::JIT::emit_op_instanceof):
1116         (JSC::JIT::emit_op_get_scoped_var):
1117         (JSC::JIT::emit_op_put_scoped_var):
1118         (JSC::JIT::emit_op_construct_verify):
1119         (JSC::JIT::emit_op_resolve_global):
1120         (JSC::JIT::emit_op_jeq_null):
1121         (JSC::JIT::emit_op_jneq_null):
1122         (JSC::JIT::emit_op_to_jsnumber):
1123         (JSC::JIT::emit_op_catch):
1124         (JSC::JIT::emit_op_eq_null):
1125         (JSC::JIT::emit_op_neq_null):
1126         (JSC::JIT::emit_op_convert_this):
1127         (JSC::JIT::emit_op_profile_will_call):
1128         (JSC::JIT::emit_op_profile_did_call):
1129         (JSC::JIT::emitSlow_op_get_by_val):
1130         * jit/JITPropertyAccess.cpp:
1131         (JSC::JIT::emit_op_get_by_val):
1132         (JSC::JIT::emit_op_put_by_val):
1133         (JSC::JIT::emit_op_method_check):
1134         (JSC::JIT::compileGetByIdHotPath):
1135         (JSC::JIT::emit_op_put_by_id):
1136         (JSC::JIT::compilePutDirectOffset):
1137         (JSC::JIT::compileGetDirectOffset):
1138         (JSC::JIT::privateCompilePutByIdTransition):
1139         (JSC::JIT::privateCompilePatchGetArrayLength):
1140         * jit/JITStubs.cpp:
1141         (JSC::JITThunks::JITThunks):
1142
1143 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
1144
1145         Reviewed by NOBODY (Windows build fix).
1146
1147         * jit/JIT.h:
1148         * jit/JITInlineMethods.h:
1149
1150 2009-06-19  Gabor Loki  <loki@inf.u-szeged.hu>
1151
1152         Reviewed by Gavin Barraclough.
1153
1154         Reorganize ARM architecture specific macros.
1155         Use PLATFORM_ARM_ARCH(7) instead of PLATFORM(ARM_V7).
1156
1157         Bug 24986: ARM JIT port
1158         <https://bugs.webkit.org/show_bug.cgi?id=24986>
1159
1160         * assembler/ARMv7Assembler.h:
1161         * assembler/AbstractMacroAssembler.h:
1162         (JSC::AbstractMacroAssembler::Imm32::Imm32):
1163         * assembler/MacroAssembler.h:
1164         * assembler/MacroAssemblerCodeRef.h:
1165         (JSC::MacroAssemblerCodePtr::MacroAssemblerCodePtr):
1166         * jit/ExecutableAllocator.h:
1167         (JSC::ExecutableAllocator::cacheFlush):
1168         * jit/JIT.h:
1169         * jit/JITInlineMethods.h:
1170         (JSC::JIT::restoreArgumentReferenceForTrampoline):
1171         * jit/JITStubs.cpp:
1172         * jit/JITStubs.h:
1173         * wtf/Platform.h:
1174         * yarr/RegexJIT.cpp:
1175         (JSC::Yarr::RegexGenerator::generateEnter):
1176         (JSC::Yarr::RegexGenerator::generateReturn):
1177
1178 2009-06-19  Gavin Barraclough  <barraclough@apple.com>
1179
1180         Reviewed by Oliver Hunt.
1181
1182         Fix armv7 JIT build issues.
1183
1184         Unfortunate the arm compiler does not like the use of offsetof on JITStackFrame (since it now contains non POD types),
1185         and the FIELD_OFFSET macro does not appear constantish enough for it to be happy with its use in COMPILE_ASSERT macros.
1186
1187         * Replace offsetofs with FIELD_OFFSETs (safe on C++ objects).
1188         * Move COMPILE_ASSERTs defending layout of JITStackFrame structure on armv7 into JITThunks constructor.
1189
1190         * jit/JIT.cpp:
1191         * jit/JIT.h:
1192         * jit/JITInlineMethods.h:
1193         (JSC::JIT::restoreArgumentReference):
1194         * jit/JITOpcodes.cpp:
1195         (JSC::JIT::emit_op_catch):
1196         * jit/JITStubs.cpp:
1197         (JSC::JITThunks::JITThunks):
1198
1199 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
1200
1201         Blind attempt at build fix.
1202
1203         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
1204         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
1205
1206 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
1207
1208         Reviewed by Oliver Hunt.
1209
1210         Inherits CallIdentifier struct from FastAllocBase because it has been
1211         instantiated by 'new' in JavaScriptCore/profiler/CallIdentifier.h:86.
1212
1213         * wtf/HashCountedSet.h:
1214
1215 2009-06-19  Adam Treat  <adam.treat@torchmobile.com>
1216
1217         Reviewed by Oliver Hunt.
1218
1219         https://bugs.webkit.org/show_bug.cgi?id=26540
1220         Modify the test shell to add a new function 'checkSyntax' that will
1221         only parse the source instead of executing it. In this way we can test
1222         pure parsing performance against some of the larger scripts in the wild.
1223
1224         * jsc.cpp:
1225         (GlobalObject::GlobalObject):
1226         (functionCheckSyntax):
1227
1228 2009-06-19  Zoltan Horvath  <hzoltan@inf.u-szeged.hu>
1229
1230         Reviewed by Darin Adler.
1231
1232         Inherits HashCountedSet class from FastAllocBase because it has been
1233         instantiated by 'new' in JavaScriptCore/runtime/Collector.cpp:1095.
1234
1235         * wtf/HashCountedSet.h:
1236
1237 2009-06-19  Yong Li  <yong.li@torchmobile.com>
1238
1239         Reviewed by George Staikos.
1240
1241         https://bugs.webkit.org/show_bug.cgi?id=26558
1242         Declare these symbols extern for WINCE as they are provided by libce.
1243
1244         * runtime/DateConstructor.cpp:
1245         * runtime/DatePrototype.cpp:
1246         (JSC::formatLocaleDate):
1247
1248 2009-06-19  Oliver Hunt  <oliver@apple.com>
1249
1250         Reviewed by Maciej Stachowiak.
1251
1252         <rdar://problem/6988973> ScopeChain leak in interpreter builds
1253
1254         Move the Scopechain destruction code in JSFunction outside of the ENABLE(JIT)
1255         path.
1256
1257         * runtime/JSFunction.cpp:
1258         (JSC::JSFunction::~JSFunction):
1259         * wtf/Platform.h:
1260
1261 2009-06-19  Yong Li  <yong.li@torchmobile.com>
1262
1263         Reviewed by George Staikos.
1264
1265         https://bugs.webkit.org/show_bug.cgi?id=26543
1266         Windows CE uses 'GetLastError' instead of 'errno.'
1267
1268         * interpreter/RegisterFile.h:
1269         (JSC::RegisterFile::RegisterFile):
1270         (JSC::RegisterFile::grow):
1271
1272 2009-06-19  David Levin  <levin@chromium.org>
1273
1274         Reviewed by NOBODY (Windows build fix).
1275
1276         Add export for Windows corresponding to OSX export done in r44844.
1277         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
1278         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore_debug.def:
1279
1280 2009-06-18  Oliver Hunt  <oliver@apple.com>
1281
1282         Reviewed by Gavin "Viceroy of Venezuela" Barraclough.
1283
1284         Bug 26532: Native functions do not correctly unlink from optimised callsites when they're collected
1285         <https://bugs.webkit.org/show_bug.cgi?id=26532> <rdar://problem/6625385>
1286
1287         We need to make sure that each native function instance correctly unlinks any references to it
1288         when it is collected.  Allowing this to happen required a few changes:
1289             * Every native function needs a codeblock to track the link information
1290             * To have this codeblock, every function now also needs its own functionbodynode
1291               so we no longer get to have a single shared instance.
1292             * Identifying a host function is now done by looking for CodeBlock::codeType() == NativeCode
1293
1294         * JavaScriptCore.exp:
1295         * bytecode/CodeBlock.cpp:
1296         (JSC::CodeBlock::CodeBlock):
1297            Constructor for NativeCode CodeBlock
1298         (JSC::CodeBlock::derefStructures):
1299         (JSC::CodeBlock::refStructures):
1300         (JSC::CodeBlock::reparseForExceptionInfoIfNecessary):
1301         (JSC::CodeBlock::handlerForBytecodeOffset):
1302         (JSC::CodeBlock::lineNumberForBytecodeOffset):
1303         (JSC::CodeBlock::expressionRangeForBytecodeOffset):
1304         (JSC::CodeBlock::getByIdExceptionInfoForBytecodeOffset):
1305         (JSC::CodeBlock::functionRegisterForBytecodeOffset):
1306         (JSC::CodeBlock::hasGlobalResolveInstructionAtBytecodeOffset):
1307         (JSC::CodeBlock::hasGlobalResolveInfoAtBytecodeOffset):
1308         (JSC::CodeBlock::setJITCode):
1309            Add assertions to ensure we don't try and use NativeCode CodeBlocks as
1310            a normal codeblock.
1311
1312         * bytecode/CodeBlock.h:
1313         (JSC::):
1314         (JSC::CodeBlock::source):
1315         (JSC::CodeBlock::sourceOffset):
1316         (JSC::CodeBlock::evalCodeCache):
1317         (JSC::CodeBlock::createRareDataIfNecessary):
1318           More assertions.
1319
1320         * jit/JIT.cpp:
1321         (JSC::JIT::privateCompileCTIMachineTrampolines):
1322         (JSC::JIT::linkCall):
1323           Update logic to allow native function caching
1324
1325         * jit/JITStubs.cpp:
1326         * parser/Nodes.cpp:
1327         (JSC::FunctionBodyNode::createNativeThunk):
1328         (JSC::FunctionBodyNode::isHostFunction):
1329         * parser/Nodes.h:
1330         * runtime/JSFunction.cpp:
1331         (JSC::JSFunction::JSFunction):
1332         (JSC::JSFunction::~JSFunction):
1333         (JSC::JSFunction::mark):
1334         * runtime/JSGlobalData.cpp:
1335         (JSC::JSGlobalData::~JSGlobalData):
1336         * runtime/JSGlobalData.h:
1337
1338 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
1339
1340         Reviewed by NOBODY (Windows build fix).
1341
1342         * wtf/DateMath.cpp:
1343         (WTF::calculateUTCOffset):
1344
1345 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
1346
1347         Reviewed by Geoff Garen.
1348
1349         Timezone calculation incorrect in Venezuela.
1350
1351         https://bugs.webkit.org/show_bug.cgi?id=26531
1352         <rdar://problem/6646169> Time is incorrectly reported to JavaScript in both Safari 3 and Firefox 3
1353
1354         The problem is that we're calculating the timezone relative to 01/01/2000,
1355         but the VET timezone changed from -4 hours to -4:30 hours on 12/09/2007.
1356         According to the spec, section 15.9.1.9 states "the time since the beginning
1357         of the year", presumably meaning the *current* year.  Change the calculation
1358         to be based on whatever the current year is, rather than a canned date.
1359
1360         No performance impact.
1361
1362         * wtf/DateMath.cpp:
1363         (WTF::calculateUTCOffset):
1364
1365 2009-06-18  Gavin Barraclough  <barraclough@apple.com>
1366
1367         Rubber Stamped by Mark Rowe (originally reviewed by Sam Weinig).
1368
1369         (Reintroducing patch added in r44492, and reverted in r44796.)
1370
1371         Change the implementation of op_throw so the stub function always modifies its
1372         return address - if it doesn't find a 'catch' it will switch to a trampoline
1373         to force a return from JIT execution.  This saves memory, by avoiding the need
1374         for a unique return for every op_throw.
1375
1376         * jit/JITOpcodes.cpp:
1377         (JSC::JIT::emit_op_throw):
1378             JITStubs::cti_op_throw now always changes its return address,
1379             remove return code generated after the stub call (this is now
1380             handled by ctiOpThrowNotCaught).
1381         * jit/JITStubs.cpp:
1382         (JSC::):
1383             Add ctiOpThrowNotCaught definitions.
1384         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
1385             Change cti_op_throw to always change its return address.
1386         * jit/JITStubs.h:
1387             Add ctiOpThrowNotCaught declaration.
1388
1389 2009-06-18  Kevin McCullough  <kmccullough@apple.com>
1390
1391         Reviewed by Oliver Hunt.
1392
1393         <rdar://problem/6940880> REGRESSION: Breakpoints don't break in 64-bit
1394
1395         - Exposed functions now needed by WebCore.
1396
1397         * JavaScriptCore.exp:
1398
1399 2009-06-17  Darin Adler  <darin@apple.com>
1400
1401         Reviewed by Oliver Hunt.
1402
1403         Bug 26429: Make JSON.stringify non-recursive so it can handle objects
1404         of arbitrary complexity
1405         https://bugs.webkit.org/show_bug.cgi?id=26429
1406
1407         For marking I decided not to use gcProtect, because this is inside the engine
1408         so it's easy enough to just do marking. And that darned gcProtect does locking!
1409         Oliver tried to convince me to used MarkedArgumentBuffer, but the constructor
1410         for that class says "FIXME: Remove all clients of this API, then remove this API."
1411
1412         * runtime/Collector.cpp:
1413         (JSC::Heap::collect): Add a call to JSONObject::markStringifiers.
1414
1415         * runtime/CommonIdentifiers.cpp:
1416         (JSC::CommonIdentifiers::CommonIdentifiers): Added emptyIdentifier.
1417         * runtime/CommonIdentifiers.h: Ditto.
1418
1419         * runtime/JSGlobalData.cpp:
1420         (JSC::JSGlobalData::JSGlobalData): Initialize firstStringifierToMark to 0.
1421         * runtime/JSGlobalData.h: Added firstStringifierToMark.
1422
1423         * runtime/JSONObject.cpp: Cut down the includes to the needed ones only.
1424         (JSC::unwrapNumberOrString): Added. Helper for unwrapping number and string
1425         objects to get their number and string values.
1426         (JSC::ReplacerPropertyName::ReplacerPropertyName): Added. The class is used
1427         to wrap an identifier or integer so we don't have to do any work unless we
1428         actually call a replacer.
1429         (JSC::ReplacerPropertyName::value): Added.
1430         (JSC::gap): Added. Helper function for the Stringifier constructor.
1431         (JSC::PropertyNameForFunctionCall::PropertyNameForFunctionCall): Added.
1432         The class is used to wrap an identifier or integer so we don't have to
1433         allocate a number or string until we actually call toJSON or a replacer.
1434         (JSC::PropertyNameForFunctionCall::asJSValue): Added.
1435         (JSC::Stringifier::Stringifier): Updated and moved out of the class
1436         definition. Added code to hook this into a singly linked list for marking.
1437         (JSC::Stringifier::~Stringifier): Remove from the singly linked list.
1438         (JSC::Stringifier::mark): Mark all the objects in the holder stacks.
1439         (JSC::Stringifier::stringify): Updated.
1440         (JSC::Stringifier::appendQuotedString): Tweaked and streamlined a bit.
1441         (JSC::Stringifier::toJSON): Renamed from toJSONValue.
1442         (JSC::Stringifier::appendStringifiedValue): Renamed from stringify.
1443         Added code to use the m_holderStack to do non-recursive stringify of
1444         objects and arrays. This code also uses the timeout checker since in
1445         pathological cases it could be slow even without calling into the
1446         JavaScript virtual machine.
1447         (JSC::Stringifier::willIndent): Added.
1448         (JSC::Stringifier::indent): Added.
1449         (JSC::Stringifier::unindent): Added.
1450         (JSC::Stringifier::startNewLine): Added.
1451         (JSC::Stringifier::Holder::Holder): Added.
1452         (JSC::Stringifier::Holder::appendNextProperty): Added. This is the
1453         function that handles the format of arrays and objects.
1454         (JSC::JSONObject::getOwnPropertySlot): Moved this down to the bottom
1455         of the file so the JSONObject class is not interleaved with the
1456         Stringifier class.
1457         (JSC::JSONObject::markStringifiers): Added. Calls mark.
1458         (JSC::JSONProtoFuncStringify): Streamlined the code here. The code
1459         to compute the gap string is now a separate function.
1460
1461         * runtime/JSONObject.h: Made everything private. Added markStringifiers.
1462
1463 2009-06-17  Oliver Hunt  <oliver@apple.com>
1464
1465         Reviewed by Gavin Barraclough.
1466
1467         <rdar://problem/6974140> REGRESSION(r43849): Crash in cti_op_call_NotJSFunction when getting directions on maps.google.com
1468
1469         Roll out r43849 as it appears that we cannot rely on the address of
1470         an objects property storage being constant even if the structure is
1471         unchanged.
1472
1473         * jit/JIT.h:
1474         * jit/JITPropertyAccess.cpp:
1475         (JSC::JIT::compileGetDirectOffset):
1476         (JSC::JIT::privateCompileGetByIdProto):
1477         (JSC::JIT::privateCompileGetByIdProtoList):
1478         (JSC::JIT::privateCompileGetByIdChainList):
1479         (JSC::JIT::privateCompileGetByIdChain):
1480
1481 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
1482
1483         Rubber Stamped by Mark Rowe.
1484
1485         Fully revert r44492 & r44748 while we fix a bug they cause on internal builds <rdar://problem/6955963>.
1486
1487         * jit/JITOpcodes.cpp:
1488         (JSC::JIT::emit_op_throw):
1489         * jit/JITStubs.cpp:
1490         (JSC::):
1491         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
1492         * jit/JITStubs.h:
1493
1494 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
1495
1496         Reviewed by Mark Rowe.
1497
1498         <rdar://problem/6947426> sunspider math-cordic.js exhibits different intermediate results running 32-bit vs. 64-bit
1499
1500         On 64-bit, NaN-encoded values must be detagged before they can be used in rshift.
1501
1502         No performance impact.
1503
1504         * jit/JITArithmetic.cpp:
1505         (JSC::JIT::emit_op_rshift):
1506
1507 2009-06-17  Adam Treat  <adam.treat@torchmobile.com>
1508
1509         Reviewed by George Staikos.
1510
1511         https://bugs.webkit.org/show_bug.cgi?id=23155
1512         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
1513
1514         * jsc.cpp:
1515         (main):
1516
1517 2009-06-17  George Staikos  <george.staikos@torchmobile.com>
1518
1519         Reviewed by Adam Treat.
1520
1521         https://bugs.webkit.org/show_bug.cgi?id=23155
1522         Move WIN_CE -> WINCE as previously discussed with Qt WINCE folks.
1523
1524         * config.h:
1525         * jsc.cpp:
1526         * wtf/Assertions.cpp:
1527         * wtf/Assertions.h:
1528         * wtf/CurrentTime.cpp:
1529         (WTF::lowResUTCTime):
1530         * wtf/DateMath.cpp:
1531         (WTF::getLocalTime):
1532         * wtf/MathExtras.h:
1533         * wtf/Platform.h:
1534         * wtf/StringExtras.h:
1535         * wtf/Threading.h:
1536         * wtf/win/MainThreadWin.cpp:
1537
1538 2009-06-17  Gavin Barraclough  <barraclough@apple.com>
1539
1540         Reviewed by Oliver Hunt.
1541
1542         <rdar://problem/6974175> ASSERT in JITStubs.cpp at appsaccess.apple.com
1543
1544         Remove PropertySlot::putValue - PropertySlots should only be used for getting,
1545         not putting.  Rename JSGlobalObject::getOwnPropertySlot to hasOwnPropertyForWrite,
1546         which is what it really was being used to ask, and remove some other getOwnPropertySlot
1547         & getOwnPropertySlotForWrite methods, which were unused and likely to lead to confusion.
1548
1549         * runtime/JSGlobalObject.h:
1550         (JSC::JSGlobalObject::hasOwnPropertyForWrite):
1551         * runtime/JSObject.h:
1552         * runtime/JSStaticScopeObject.cpp:
1553         * runtime/JSStaticScopeObject.h:
1554         * runtime/PropertySlot.h:
1555
1556 2009-06-16  Gavin Barraclough  <barraclough@apple.com>
1557
1558         Reviewed by Oliver hunt.
1559
1560         Temporarily partially disable r44492, since this is causing some problems on internal builds.
1561
1562         * jit/JITOpcodes.cpp:
1563         (JSC::JIT::emit_op_throw):
1564         * jit/JITStubs.cpp:
1565         (JSC::JITStubs::DEFINE_STUB_FUNCTION):
1566
1567 2009-06-16  Sam Weinig  <sam@webkit.org>
1568
1569         Fix windows build.
1570
1571         * jit/JIT.cpp:
1572         (JSC::JIT::JIT):
1573
1574 2009-06-16  Sam Weinig  <sam@webkit.org>
1575
1576         Reviewed by Oliver Hunt.
1577
1578         Initialize m_bytecodeIndex to -1 in JIT, and correctly initialize
1579         it for each type of stub using the return address to find the correct
1580         offset.
1581
1582         * jit/JIT.cpp:
1583         (JSC::JIT::JIT):
1584         * jit/JIT.h:
1585         (JSC::JIT::compileGetByIdProto):
1586         (JSC::JIT::compileGetByIdSelfList):
1587         (JSC::JIT::compileGetByIdProtoList):
1588         (JSC::JIT::compileGetByIdChainList):
1589         (JSC::JIT::compileGetByIdChain):
1590         (JSC::JIT::compilePutByIdTransition):
1591         (JSC::JIT::compileCTIMachineTrampolines):
1592         (JSC::JIT::compilePatchGetArrayLength):
1593         * jit/JITStubCall.h:
1594         (JSC::JITStubCall::call):
1595
1596 == Rolled over to ChangeLog-2009-06-16 ==