Source/JavaScriptCore/ChangeLog

   1 2011-06-04  Kevin Ollivier  <kevino@theolliviers.com>
   2
   3         [wx] Unreviewed build fix. Restore the PPC build and allow users to specify architectures
   4         to build on Mac.
   5
   6         * wtf/Platform.h:
   7
   8 2011-06-04  Gustavo Noronha Silva  <gns@gnome.org>
   9
  10         Unreviewed, MIPS build fix.
  11
  12         WebKitGTK+ tarball fails to build on MIPS.
  13         https://buildd.debian.org/status/fetch.php?pkg=webkitgtk%2B&arch=mips&ver=1.4.0-1&stamp=1304786691
  14
  15         * GNUmakefile.list.am: Add missing MIPS-related file to the list
  16         of files that are added to the tarball on make dist, and fix
  17         sorting.
  18
  19 2011-06-04  Sam Weinig  <sam@webkit.org>
  20
  21         Reviewed by Darin Adler.
  22
  23         Fix formatting of the output generated by KeywordLookupGenerator.py
  24         https://bugs.webkit.org/show_bug.cgi?id=62083
  25
  26         - Uses correct year for copyright.
  27         - Puts ending brace on same line as "else if"
  28         - Puts starting brace of function on its own line.
  29         - Adds some tasteful whitespace.
  30         - Adds comments to make clear that scopes are ending
  31         - Make macros actually split on two lines.
  32
  33         * KeywordLookupGenerator.py:
  34
  35 2011-06-04  Adam Barth  <abarth@webkit.org>
  36
  37         Reviewed by Eric Seidel.
  38
  39         KeywordLookupGenerator.py spams stdout in Chromium Linux build
  40         https://bugs.webkit.org/show_bug.cgi?id=62087
  41
  42         This action does not appear to be needed.
  43
  44         * JavaScriptCore.gyp/JavaScriptCore.gyp:
  45
  46 2011-06-03  Oliver Hunt  <oliver@apple.com>
  47
  48         Reviewed by Maciej Stachowiak.
  49
  50         Lexer needs to provide Identifier for reserved words
  51         https://bugs.webkit.org/show_bug.cgi?id=62086
  52
  53         Alas it is necessary to provide an Identifier reference for keywords
  54         so that we can do the right thing when they're used in object literals.
  55         We now keep Identifiers for all reserved words in the CommonIdentifiers
  56         structure so that we can access them without a hash lookup.
  57
  58         * KeywordLookupGenerator.py:
  59         * parser/Lexer.cpp:
  60         (JSC::Lexer::parseIdentifier):
  61         * parser/Lexer.h:
  62         * runtime/CommonIdentifiers.cpp:
  63         (JSC::CommonIdentifiers::CommonIdentifiers):
  64         * runtime/CommonIdentifiers.h:
  65
  66 2011-06-03  Gavin Barraclough  <barraclough@apple.com>
  67
  68         Reviewed by Sam Weinig.
  69
  70         Add debug code to break on speculation failures.
  71
  72         * dfg/DFGJITCompiler.cpp:
  73         (JSC::DFG::JITCompiler::jumpFromSpeculativeToNonSpeculative):
  74         (JSC::DFG::JITCompiler::compileFunction):
  75         * dfg/DFGNode.h:
  76
  77 2011-06-03  Gavin Barraclough  <barraclough@apple.com>
  78
  79         Reviewed by Sam Weinig.
  80
  81         https://bugs.webkit.org/show_bug.cgi?id=62082
  82         DFG JIT - bug passing arguments that need swap
  83
  84         This is really just a typo.
  85         When setting up the arguments for a call out to a C operation, we'll
  86         fail to swap arguments where this is necessary. For example, in the
  87         case of 2 arg calls, where the first argument is in %rdx & the second
  88         is in %rsi we should swap (exec will be passed in %rdi), but we don't.
  89
  90         This can also affect function calls passing three arguments.
  91
  92         * dfg/DFGJITCodeGenerator.h:
  93         (JSC::DFG::JITCodeGenerator::setupTwoStubArgs):
  94             - Call swap with the correct arguments.
  95
  96 2011-06-03  Oliver Hunt  <oliver@apple.com>
  97
  98         Reviewed by Gavin Barraclough.
  99
 100         Force inlining of some hot lexer functions
 101         https://bugs.webkit.org/show_bug.cgi?id=62079
 102
 103         Fix more GCC stupidity
 104
 105         * parser/Lexer.h:
 106         (JSC::Lexer::isWhiteSpace):
 107         (JSC::Lexer::isLineTerminator):
 108
 109 2011-06-03  Oliver Hunt  <oliver@apple.com>
 110
 111         Reviewed by Gavin Barraclough.
 112
 113         GCC not inlining some functions that it really should be
 114         https://bugs.webkit.org/show_bug.cgi?id=62075
 115
 116         Add ALWAYS_INLINE to a number of parsing and lexing functions
 117         that should always be inlined.  This gets us ~1.4% on my ad hoc
 118         parser test.
 119
 120         * KeywordLookupGenerator.py:
 121         * parser/JSParser.cpp:
 122         (JSC::JSParser::next):
 123         (JSC::JSParser::nextTokenIsColon):
 124         (JSC::JSParser::consume):
 125         (JSC::JSParser::match):
 126         (JSC::JSParser::tokenStart):
 127         (JSC::JSParser::tokenLine):
 128         (JSC::JSParser::tokenEnd):
 129         * parser/Lexer.cpp:
 130         (JSC::isIdentPart):
 131
 132 2011-06-03  Oliver Hunt  <oliver@apple.com>
 133
 134         Whoops, fix last minute bug.
 135
 136         * parser/Lexer.cpp:
 137         (JSC::Lexer::parseIdentifier):
 138
 139 2011-06-03  Martin Robinson  <mrobinson@igalia.com>
 140
 141         Try to fix the GTK+ build.
 142
 143         * GNUmakefile.am: Clean up some spaces that should be tabs.
 144         * GNUmakefile.list.am: Add KeywordLookup.h to the source list
 145         and clean up some spaces that should be tabs.
 146
 147 2011-06-03  Oliver Hunt  <oliver@apple.com>
 148
 149         Reviewed by Geoffrey Garen.
 150
 151         Improve keyword lookup
 152         https://bugs.webkit.org/show_bug.cgi?id=61913
 153
 154         Rather than doing multiple hash lookups as we currently
 155         do when trying to identify keywords we now use an
 156         automatically generated decision tree (essentially it's
 157         a hard coded patricia trie).  We still use the regular
 158         lookup table for the last few characters of an input as
 159         this allows us to completely skip all bounds checks.
 160
 161         * CMakeLists.txt:
 162         * DerivedSources.make:
 163         * DerivedSources.pro:
 164         * GNUmakefile.am:
 165         * JavaScriptCore.gyp/JavaScriptCore.gyp:
 166         * JavaScriptCore.xcodeproj/project.pbxproj:
 167         * KeywordLookupGenerator.py: Added.
 168         * make-generated-sources.sh:
 169         * parser/Lexer.cpp:
 170         (JSC::Lexer::internalShift):
 171         (JSC::Lexer::shift):
 172         (JSC::Lexer::parseIdentifier):
 173         * parser/Lexer.h:
 174
 175 2011-06-03  Siddharth Mathur  <siddharth.mathur@nokia.com>
 176
 177         Reviewed by Benjamin Poulain.
 178
 179         [Qt] Build flag for experimental ICU library support
 180         https://bugs.webkit.org/show_bug.cgi?id=60786
 181
 182         Adds a build-time flag (CONFIG+=use_system_icu) that enables experimental
 183         ICU powered Unicode support.
 184
 185         * JavaScriptCore.pri: Support for use_system_icu CONFIG flag.
 186         * wtf/unicode/qt4/UnicodeQt4.h: Guard an include file with USE(ICU_UNICODE).
 187
 188 2011-06-03  Alexis Menard  <alexis.menard@openbossa.org>
 189
 190         Reviewed by Benjamin Poulain.
 191
 192         [Qt] Build fix of QtWebKit 2.2 when inside Qt tree with GCC 4.6.
 193         https://bugs.webkit.org/show_bug.cgi?id=61957
 194
 195         When building inside the Qt source tree, qmake always append the mkspecs
 196         defines after ours. We have to workaround and make sure that we append
 197         our flags after the qmake variable used inside Qt. This workaround was provided
 198         by our qmake folks. We need to append in both case because qmake behave differently
 199         when called with -spec or via SUBDIR+=. This patch unbreak r87950 on Mac for Qt port.
 200
 201         * JavaScriptCore.pro:
 202
 203 2011-06-02  Jay Civelli  <jcivelli@chromium.org>
 204
 205         Reviewed by Adam Barth.
 206
 207         Added a method to generate RFC 2822 compliant date strings.
 208         https://bugs.webkit.org/show_bug.cgi?id=7169
 209
 210         * wtf/DateMath.cpp:
 211         (WTF::twoDigitStringFromNumber):
 212         (WTF::makeRFC2822DateString):
 213         * wtf/DateMath.h:
 214
 215 2011-06-02  Alexis Menard  <alexis.menard@openbossa.org>
 216
 217         Reviewed by Andreas Kling.
 218
 219         [Qt] Build fix of QtWebKit 2.2 when inside Qt tree with GCC 4.6.
 220         https://bugs.webkit.org/show_bug.cgi?id=61957
 221
 222         When building inside the Qt source tree, qmake always append the mkspecs
 223         defines after ours. We have to workaround and make sure that we append
 224         our flags after the qmake variable used inside Qt. This workaround was provided
 225         by our qmake folks.
 226
 227         * JavaScriptCore.pro:
 228
 229 2011-06-01  Oliver Hunt  <oliver@apple.com>
 230
 231         Reviewed by Geoffrey Garen.
 232
 233         Add single character lookup cache to IdentifierArena
 234         https://bugs.webkit.org/show_bug.cgi?id=61879
 235
 236         Add a simple lookup cache for single ascii character
 237         identifiers.  Produces around a 2% improvement in parse
 238         time for my adhoc parser test.
 239
 240         * parser/ParserArena.h:
 241         (JSC::IdentifierArena::IdentifierArena):
 242         (JSC::IdentifierArena::clear):
 243         (JSC::IdentifierArena::makeIdentifier):
 244
 245 2011-05-31  Oliver Hunt  <oliver@apple.com>
 246
 247         Reviewed by Geoffrey Garen.
 248
 249         Freezing a function and its prototype causes browser to crash.
 250         https://bugs.webkit.org/show_bug.cgi?id=61758
 251
 252         Make JSObject::preventExtensions virtual so that we can override it
 253         and instantiate all lazy
 254
 255         * JavaScriptCore.exp:
 256         * runtime/JSFunction.cpp:
 257         (JSC::createPrototypeProperty):
 258         (JSC::JSFunction::preventExtensions):
 259         (JSC::JSFunction::getOwnPropertySlot):
 260         * runtime/JSFunction.h:
 261         * runtime/JSObject.h:
 262         * runtime/JSObject.cpp:
 263         (JSC::JSObject::seal):
 264         (JSC::JSObject::seal):
 265
 266 2011-06-01  Sheriff Bot  <webkit.review.bot@gmail.com>
 267
 268         Unreviewed, rolling out r87788.
 269         http://trac.webkit.org/changeset/87788
 270         https://bugs.webkit.org/show_bug.cgi?id=61856
 271
 272         breaks windows chromium canary (Requested by jknotten on
 273         #webkit).
 274
 275         * wtf/DateMath.cpp:
 276         (WTF::timeClip):
 277         * wtf/DateMath.h:
 278
 279 2011-06-01  Jay Civelli  <jcivelli@chromium.org>
 280
 281         Reviewed by Adam Barth.
 282
 283         Added a method to generate RFC 2822 compliant date strings.
 284         https://bugs.webkit.org/show_bug.cgi?id=7169
 285
 286         * wtf/DateMath.cpp:
 287         (WTF::twoDigitStringFromNumber):
 288         (WTF::makeRFC2822DateString):
 289         * wtf/DateMath.h:
 290
 291 2011-05-31  Yong Li  <yoli@rim.com>
 292
 293         Reviewed by Eric Seidel.
 294
 295         https://bugs.webkit.org/show_bug.cgi?id=54807
 296         We have been assuming plain bitfields (like "int a : 31") are always signed integers.
 297         However some compilers can treat them as unsigned. For example, RVCT 4.0 states plain
 298         bitfields (declared without either signed or unsigned qualifiers) are treats as unsigned.
 299         http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0348c/Babjddhe.html
 300         Although we can use "--signed-bitfields" flag to make RVCT 4.0 behave as most other compilers,
 301         always using "signed"/"unsigned" qualifier to declare integral type bitfields is still a good
 302         rule we should have in order to make our code independent from compilers and compiler flags.
 303
 304         No new test added because this change is not known to fix any issue.
 305
 306         * bytecode/StructureStubInfo.h:
 307
 308 2011-05-30  Hojong Han  <hojong.han@samsung.com>
 309
 310         Reviewed by Geoffrey Garen.
 311
 312         [JSC] malfunction during arithmetic condition check with negative number (-2147483648)
 313         https://bugs.webkit.org/show_bug.cgi?id=61416
 314
 315         * assembler/MacroAssemblerARM.h:
 316         (JSC::MacroAssemblerARM::branch32):
 317         * tests/mozilla/ecma/Expressions/11.12-1.js:
 318         (getTestCases):
 319
 320 2011-05-29  Geoffrey Garen  <ggaren@apple.com>
 321
 322         Reviewed by Sam Weinig.
 323
 324         Some heap refactoring
 325         https://bugs.webkit.org/show_bug.cgi?id=61704
 326
 327         SunSpider says no change.
 328
 329         * JavaScriptCore.exp: Export!
 330
 331         * heap/Heap.cpp: COLLECT_ON_EVERY_ALLOCATION can actually do so now.
 332
 333         (JSC::Heap::Heap): Changed Heap sub-objects to point to the heap.
 334
 335         (JSC::Heap::allocate): Changed inline allocation code to only select the
 336         size class, since this can be optimized out at compile time -- everything
 337         else is now inlined into this out-of-line function.
 338
 339         No need to duplicate ASSERTs made in our caller.
 340
 341         * heap/Heap.h:
 342         (JSC::Heap::heap):
 343         (JSC::Heap::isMarked):
 344         (JSC::Heap::testAndSetMarked):
 345         (JSC::Heap::testAndClearMarked):
 346         (JSC::Heap::setMarked): Call directly into MarkedBlock instead of adding
 347         a layer of indirection through MarkedSpace.
 348
 349         (JSC::Heap::allocate): See above.
 350
 351         * heap/MarkedBlock.cpp:
 352         (JSC::MarkedBlock::create):
 353         (JSC::MarkedBlock::MarkedBlock):
 354         * heap/MarkedBlock.h: Changed Heap sub-objects to point to the heap.
 355
 356         * heap/MarkedSpace.cpp:
 357         (JSC::MarkedSpace::MarkedSpace):
 358         (JSC::MarkedSpace::allocateBlock):
 359         * heap/MarkedSpace.h:
 360         (JSC::MarkedSpace::allocate): Updated to match changes above.
 361
 362 2011-05-28  David Kilzer  <ddkilzer@apple.com>
 363
 364         BUILD FIX when building only the interpreter
 365
 366         Fixes the following compiler warning:
 367
 368             JavaScriptCore/runtime/JSGlobalData.cpp:462:6: error: no previous prototype for function 'releaseExecutableMemory' [-Werror,-Wmissing-prototypes,3]
 369              void releaseExecutableMemory(JSGlobalData& globalData)
 370                   ^
 371
 372         * jit/ExecutableAllocator.h: Moved declaration of
 373         JSC::releaseExecutableMemory().
 374
 375 2011-05-28  David Kilzer  <ddkilzer@apple.com>
 376
 377         BUILD FIX after r87527 with ENABLE(BRANCH_COMPACTION)
 378
 379         * assembler/LinkBuffer.h:
 380         (JSC::LinkBuffer::linkCode): Added missing argument.
 381
 382 2011-05-27  Geoffrey Garen  <ggaren@apple.com>
 383
 384         Reviewed by Oliver Hunt.
 385
 386         JS API is too aggressive about throwing exceptions for NULL get or set operations
 387         https://bugs.webkit.org/show_bug.cgi?id=61678
 388
 389         * API/JSCallbackObject.h: Changed our staticValueGetter to a regular
 390         function that returns a JSValue, so it can fail and still forward to
 391         normal property lookup.
 392
 393         * API/JSCallbackObjectFunctions.h:
 394         (JSC::::getOwnPropertySlot): Don't throw an exception when failing to
 395         access a static property -- just forward the access. This allows objects
 396         to observe get/set operations but still let the JS object manage lifetime.
 397
 398         (JSC::::put): Ditto.
 399
 400         (JSC::::getStaticValue): Same as JSCallbackObject.h.
 401
 402         * API/tests/testapi.c:
 403         (MyObject_set_nullGetForwardSet):
 404         * API/tests/testapi.js: Updated tests to reflect slightly less strict
 405         behavior, which matches headerdoc claims.
 406
 407 2011-05-27  Geoffrey Garen  <ggaren@apple.com>
 408
 409         Reviewed by Oliver Hunt.
 410
 411         Property caching is too aggressive for API objects
 412         https://bugs.webkit.org/show_bug.cgi?id=61677
 413
 414         * API/JSCallbackObject.h: Opt in to ProhibitsPropertyCaching, since our
 415         callback APIs allow the client to change its mind about our propertis at
 416         any time.
 417
 418         * API/tests/testapi.c:
 419         (PropertyCatchalls_getProperty):
 420         (PropertyCatchalls_setProperty):
 421         (PropertyCatchalls_getPropertyNames):
 422         (PropertyCatchalls_class):
 423         (main):
 424         * API/tests/testapi.js: Some tests for dynamic API objects.
 425
 426         * interpreter/Interpreter.cpp:
 427         (JSC::Interpreter::tryCachePutByID):
 428         (JSC::Interpreter::tryCacheGetByID):
 429         * jit/JITStubs.cpp:
 430         (JSC::JITThunks::tryCachePutByID):
 431         (JSC::JITThunks::tryCacheGetByID):
 432         (JSC::DEFINE_STUB_FUNCTION): Opt out of property caching if the client
 433         requires it.
 434
 435         * runtime/JSTypeInfo.h:
 436         (JSC::TypeInfo::TypeInfo):
 437         (JSC::TypeInfo::isFinal):
 438         (JSC::TypeInfo::prohibitsPropertyCaching):
 439         (JSC::TypeInfo::flags): Added a flag to track opting out of property
 440         caching. Fixed an "&&" vs "&" typo that was previously harmless, but
 441         is now harmful since m_flags2 can have more than one bit set.
 442
 443 2011-05-27  Stephanie Lewis  <slewis@apple.com>
 444
 445         Unreviewed.
 446
 447         Fix a typo in the order_file flag.
 448
 449         * Configurations/Base.xcconfig:
 450
 451 2011-05-27  Patrick Gansterer  <paroga@webkit.org>
 452
 453         Unreviewed. Build fix for !ENABLE(ASSEMBLER) after r87527.
 454
 455         * runtime/JSGlobalData.cpp:
 456         (JSGlobalData::JSGlobalData):
 457
 458 2011-05-27  Oliver Hunt  <oliver@apple.com>
 459
 460         Reviewed by Geoffrey Garen.
 461
 462         Add a few validity assertions to JSCallbackObject
 463         https://bugs.webkit.org/show_bug.cgi?id=61659
 464
 465         * API/JSCallbackObject.h:
 466         (JSC::JSCallbackObject::visitChildren):
 467
 468 2011-05-27  Oliver Hunt  <oliver@apple.com>
 469
 470         Build fix
 471
 472         * runtime/RegExpCache.cpp:
 473         (JSC::RegExpCache::invalidateCode):
 474
 475 2011-05-27  Oliver Hunt  <oliver@apple.com>
 476
 477         Reviewed by Geoffrey Garen.
 478
 479         Try to release unused executable memory when the FixedVMPool allocator is under pressure
 480         https://bugs.webkit.org/show_bug.cgi?id=61651
 481
 482         Rather than crashing when full the FixedVMPool allocator now returns a null
 483         allocation.  We replace the code that used to CRASH() on null allocations
 484         with logic that asks the provided globalData to release any executable memory
 485         that it can.  Currently this just means throwing away all regexp code, but
 486         in future we'll try to be more aggressive.
 487
 488         * assembler/ARMAssembler.cpp:
 489         (JSC::ARMAssembler::executableCopy):
 490         * assembler/ARMAssembler.h:
 491         * assembler/AssemblerBuffer.h:
 492         (JSC::AssemblerBuffer::executableCopy):
 493         * assembler/AssemblerBufferWithConstantPool.h:
 494         * assembler/LinkBuffer.h:
 495         (JSC::LinkBuffer::LinkBuffer):
 496         (JSC::LinkBuffer::linkCode):
 497         * assembler/MIPSAssembler.h:
 498         (JSC::MIPSAssembler::executableCopy):
 499         * assembler/SH4Assembler.h:
 500         (JSC::SH4Assembler::executableCopy):
 501         * assembler/X86Assembler.h:
 502         (JSC::X86Assembler::executableCopy):
 503         (JSC::X86Assembler::X86InstructionFormatter::executableCopy):
 504         * dfg/DFGJITCompiler.cpp:
 505         (JSC::DFG::JITCompiler::compileFunction):
 506         * jit/ExecutableAllocator.h:
 507         (JSC::ExecutablePool::create):
 508         (JSC::ExecutablePool::alloc):
 509         (JSC::ExecutableAllocator::ExecutableAllocator):
 510         (JSC::ExecutableAllocator::poolForSize):
 511         (JSC::ExecutablePool::ExecutablePool):
 512         (JSC::ExecutablePool::poolAllocate):
 513         * jit/ExecutableAllocatorFixedVMPool.cpp:
 514         (JSC::FixedVMPoolAllocator::alloc):
 515         * jit/JIT.cpp:
 516         (JSC::JIT::privateCompile):
 517         * jit/JITOpcodes.cpp:
 518         (JSC::JIT::privateCompileCTIMachineTrampolines):
 519         * jit/JITOpcodes32_64.cpp:
 520         (JSC::JIT::privateCompileCTIMachineTrampolines):
 521         (JSC::JIT::privateCompileCTINativeCall):
 522         * jit/JITPropertyAccess.cpp:
 523         (JSC::JIT::stringGetByValStubGenerator):
 524         (JSC::JIT::privateCompilePutByIdTransition):
 525         (JSC::JIT::privateCompilePatchGetArrayLength):
 526         (JSC::JIT::privateCompileGetByIdProto):
 527         (JSC::JIT::privateCompileGetByIdSelfList):
 528         (JSC::JIT::privateCompileGetByIdProtoList):
 529         (JSC::JIT::privateCompileGetByIdChainList):
 530         (JSC::JIT::privateCompileGetByIdChain):
 531         * jit/JITPropertyAccess32_64.cpp:
 532         (JSC::JIT::stringGetByValStubGenerator):
 533         (JSC::JIT::privateCompilePutByIdTransition):
 534         (JSC::JIT::privateCompilePatchGetArrayLength):
 535         (JSC::JIT::privateCompileGetByIdProto):
 536         (JSC::JIT::privateCompileGetByIdSelfList):
 537         (JSC::JIT::privateCompileGetByIdProtoList):
 538         (JSC::JIT::privateCompileGetByIdChainList):
 539         (JSC::JIT::privateCompileGetByIdChain):
 540         * jit/SpecializedThunkJIT.h:
 541         (JSC::SpecializedThunkJIT::finalize):
 542         * jit/ThunkGenerators.cpp:
 543         (JSC::charCodeAtThunkGenerator):
 544         (JSC::charAtThunkGenerator):
 545         (JSC::fromCharCodeThunkGenerator):
 546         (JSC::sqrtThunkGenerator):
 547         (JSC::powThunkGenerator):
 548         * runtime/JSGlobalData.cpp:
 549         (JSC::JSGlobalData::JSGlobalData):
 550         (JSC::JSGlobalData::releaseExecutableMemory):
 551         (JSC::releaseExecutableMemory):
 552         * runtime/JSGlobalData.h:
 553         * runtime/RegExpCache.cpp:
 554         (JSC::RegExpCache::invalidateCode):
 555         * runtime/RegExpCache.h:
 556         * yarr/YarrJIT.cpp:
 557         (JSC::Yarr::YarrGenerator::compile):
 558
 559 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 560
 561         Reviewed by Oliver Hunt.
 562
 563         Optimized ConservativeSet to avoid double-visiting objects
 564         https://bugs.webkit.org/show_bug.cgi?id=61592
 565
 566         SunSpider thinks this might be a 1% speedup
 567
 568         * heap/ConservativeRoots.h:
 569         (JSC::ConservativeRoots::add): Use testAndClearMarked to avoid double-visiting
 570         an object.
 571
 572         * heap/Heap.h:
 573         (JSC::Heap::isMarked):
 574         (JSC::Heap::testAndSetMarked):
 575         (JSC::Heap::testAndClearMarked):
 576         (JSC::Heap::setMarked): Added testAndClearMarked. Changed argument type
 577         to void*, since clients want to ask questions about arbitrary pointers
 578         into the heap, even when they aren't known to be JSCells.
 579
 580         * heap/MarkedBlock.h:
 581         (JSC::MarkedBlock::testAndClearMarked):
 582         * heap/MarkedSpace.h:
 583         (JSC::MarkedSpace::isMarked):
 584         (JSC::MarkedSpace::testAndSetMarked):
 585         (JSC::MarkedSpace::testAndClearMarked):
 586         (JSC::MarkedSpace::setMarked):
 587         (JSC::MarkedSpace::contains): Ditto.
 588
 589         * wtf/Bitmap.h:
 590         (WTF::::testAndClear): New function for ConservativeRoots's inverted
 591         marking pass.
 592
 593 2011-05-27  Stephanie Lewis  <slewis@apple.com>
 594
 595         Rubber Stamped by Adam Roben.
 596
 597         Update Order Files.  Use -order_file flag since it can order more of the binary.
 598
 599         * Configurations/Base.xcconfig:
 600         * JavaScriptCore.order:
 601
 602 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 603
 604         Reviewed by Oliver Hunt.
 605
 606         Renamed heapRootMarker to heapRootVisitor to match its class name
 607         https://bugs.webkit.org/show_bug.cgi?id=61584
 608
 609         * heap/Heap.cpp:
 610         (JSC::Heap::markProtectedObjects):
 611         (JSC::Heap::markTempSortVectors):
 612         (JSC::Heap::markRoots):
 613
 614 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 615
 616         Reviewed by Oliver Hunt.
 617
 618         Removed some interdependency between Heap and SmallStrings by simplifying
 619         the SmallStrings lifetime model
 620         https://bugs.webkit.org/show_bug.cgi?id=61579
 621
 622         SunSpider reports no change.
 623
 624         Using Weak<T> could accomplish this too, but we're not sure it will give
 625         us the performance we need. This is a first step, and it accomplishes
 626         most of the value of using Weak<T>.
 627
 628         * heap/Heap.cpp:
 629         (JSC::Heap::destroy):
 630         (JSC::Heap::markRoots):
 631         (JSC::Heap::reset): Finalize small strings just like other weak handles.
 632
 633         * runtime/SmallStrings.cpp:
 634         (JSC::finalize):
 635         (JSC::SmallStrings::finalizeSmallStrings):
 636         * runtime/SmallStrings.h: Make all small strings trivially weak, instead
 637         of having an "all for one, one for all" memory model.
 638
 639 2011-05-26  Oliver Hunt  <oliver@apple.com>
 640
 641         Reviewed by Geoffrey Garen.
 642
 643         Make RegExpCache a weak map
 644         https://bugs.webkit.org/show_bug.cgi?id=61554
 645
 646         Switch to a weak map for the regexp cache, and hide that
 647         behaviour behind RegExp::create.
 648
 649         When a RegExp is compiled it attempts to add itself to
 650         the "strong" cache.  This cache is a simple round-robin
 651         buffer as was the old strong cache.  Happily this can
 652         be smaller than the old strong cache as RegExps are only
 653         added when they're compiled so it is under less pressure
 654         to evict.
 655
 656         * bytecompiler/NodesCodegen.cpp:
 657         (JSC::RegExpNode::emitBytecode):
 658         * runtime/RegExp.cpp:
 659         (JSC::RegExp::RegExp):
 660         (JSC::RegExp::create):
 661         (JSC::RegExp::match):
 662         * runtime/RegExp.h:
 663         (JSC::RegExp::gcShouldInvalidateCode):
 664         (JSC::RegExp::hasCode):
 665         (JSC::RegExp::key):
 666         * runtime/RegExpCache.cpp:
 667         (JSC::RegExpCache::lookupOrCreate):
 668         (JSC::RegExpCache::RegExpCache):
 669         (JSC::RegExpCache::isReachableFromOpaqueRoots):
 670         (JSC::RegExpCache::finalize):
 671         * runtime/RegExpCache.h:
 672         * runtime/RegExpConstructor.cpp:
 673         (JSC::constructRegExp):
 674         * runtime/RegExpPrototype.cpp:
 675         (JSC::regExpProtoFuncCompile):
 676         * runtime/StringPrototype.cpp:
 677         (JSC::stringProtoFuncMatch):
 678         (JSC::stringProtoFuncSearch):
 679
 680 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 681
 682         Reviewed by Oliver Hunt.
 683
 684         Moved Heap-related functions out of JSCell.h and into respective header files
 685         https://bugs.webkit.org/show_bug.cgi?id=61567
 686
 687         * heap/Heap.h:
 688         (JSC::Heap::allocate):
 689         (JSC::Heap::heap):
 690         * heap/MarkedBlock.h:
 691         (JSC::MarkedBlock::allocate):
 692         * heap/MarkedSpace.h:
 693         (JSC::MarkedSpace::sizeClassFor):
 694         (JSC::MarkedSpace::allocate):
 695         * runtime/JSCell.h:
 696         (JSC::JSCell::destructor):
 697
 698 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 699
 700         Try to fix Windows build.
 701
 702         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
 703
 704 2011-05-26  Ryosuke Niwa  <rniwa@webkit.org>
 705
 706         Reviewed by Eric Seidel.
 707
 708         [debug feature] WTFString should have show() method
 709         https://bugs.webkit.org/show_bug.cgi?id=61149
 710
 711         Added String::show and AtomicString::show in NDEBUG.
 712
 713         * wtf/text/AtomicString.cpp:
 714         (WTF::AtomicString::show):
 715         * wtf/text/AtomicString.h:
 716         * wtf/text/WTFString.cpp:
 717         (String::show):
 718         * wtf/text/WTFString.h:
 719
 720 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 721
 722         Reviewed by Geoffrey Garen.
 723
 724         Factored out some Heap ASSERTs
 725         https://bugs.webkit.org/show_bug.cgi?id=61565
 726
 727         * JavaScriptCore.exp:
 728         * heap/Heap.cpp:
 729         (JSC::isValidSharedInstanceThreadState):
 730         (JSC::isValidThreadState):
 731         (JSC::Heap::markRoots):
 732         (JSC::Heap::isValidAllocation):
 733         * heap/Heap.h:
 734         * runtime/JSCell.h:
 735         (JSC::JSCell::Heap::allocate):
 736
 737 2011-05-26  Gavin Barraclough  <barraclough@apple.com>
 738
 739         Reviewed by Geoff Garen.
 740
 741         https://bugs.webkit.org/show_bug.cgi?id=61508
 742         DFG JIT - Add support for get by id self caching.
 743
 744         Change the call out to be an unexpected call (using silent spill/fill functions),
 745         add a structure check & compact load to the JIT code, and add repatching mechanisms.
 746         Since DFGOperations may want to be be implemented in asm, make these symbols be extern
 747         "C". Add an asm wrapper to pass the return address to the optimizing get-by-id operation,
 748         so that it can look up its StructureStubInfo.
 749
 750         * JavaScriptCore.xcodeproj/project.pbxproj:
 751             - Added new files.
 752         * bytecode/StructureStubInfo.h:
 753             - Added 'unset' entries to union.
 754         * dfg/DFGJITCodeGenerator.h:
 755         (JSC::DFG::JITCodeGenerator::appendCallWithExceptionCheck):
 756             - Return the call, we need this to populate the StructureStubInfo.
 757         * dfg/DFGJITCompiler.cpp:
 758         (JSC::DFG::JITCompiler::compileFunction):
 759             - Populate the CodebBlock's StructureStubInfo Vector.
 760         * dfg/DFGJITCompiler.h:
 761         (JSC::DFG::JITCompiler::appendCallWithExceptionCheck):
 762             - Return the call, we need this to populate the StructureStubInfo.
 763         (JSC::DFG::JITCompiler::addPropertyAccess):
 764         (JSC::DFG::JITCompiler::PropertyAccessRecord::PropertyAccessRecord):
 765             - Add structures to record property access info during compilation.
 766         * dfg/DFGOperations.cpp:
 767             - Made all external methods extern "C".
 768         (JSC::DFG::operationPutByValInternal):
 769             - Moved outside of the extern "C" block.
 770         * dfg/DFGOperations.h:
 771             - Made all external methods extern "C".
 772         * dfg/DFGRepatch.cpp: Added.
 773         (JSC::DFG::dfgRepatchCall):
 774             - repatch a call to link to a new callee function.
 775         (JSC::DFG::dfgRepatchGetByIdSelf):
 776             - Modify the JIT code to optimize self accesses.
 777         (JSC::DFG::tryCacheGetByID):
 778             - Internal implementation of dfgRepatchGetByID (factor out failing cases).
 779         (JSC::DFG::dfgRepatchGetByID):
 780             - Used to optimize 'operationGetByIdOptimize' - repatches to 'operationGetById', and tries to optimize self accesses!
 781         * dfg/DFGRepatch.h: Added.
 782             - Expose dfgRepatchGetByID.
 783         * dfg/DFGSpeculativeJIT.cpp:
 784         (JSC::DFG::SpeculativeJIT::compile):
 785             - Changed implementation of GetById ops.
 786
 787 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 788
 789         Rolled back in http://trac.webkit.org/changeset/87408 with Windows build fixed.
 790
 791         * heap/MarkedBlock.cpp:
 792         (JSC::MarkedBlock::MarkedBlock):
 793         * heap/MarkedBlock.h:
 794         * wtf/DoublyLinkedList.h:
 795         (WTF::::DoublyLinkedListNode):
 796         (WTF::::setPrev):
 797         (WTF::::setNext):
 798         (WTF::::prev):
 799         (WTF::::next):
 800         (WTF::::DoublyLinkedList):
 801         (WTF::::isEmpty):
 802         (WTF::::size):
 803         (WTF::::clear):
 804         (WTF::::head):
 805         (WTF::::append):
 806         (WTF::::remove):
 807         (WTF::::removeHead):
 808
 809 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 810
 811         Rolled out http://trac.webkit.org/changeset/87408 because it broke the
 812         Windows build.
 813
 814         * heap/MarkedBlock.cpp:
 815         (JSC::MarkedBlock::MarkedBlock):
 816         * heap/MarkedBlock.h:
 817         (JSC::MarkedBlock::setPrev):
 818         (JSC::MarkedBlock::setNext):
 819         (JSC::MarkedBlock::prev):
 820         (JSC::MarkedBlock::next):
 821         * wtf/DoublyLinkedList.h:
 822         (WTF::::DoublyLinkedList):
 823         (WTF::::isEmpty):
 824         (WTF::::head):
 825         (WTF::::append):
 826         (WTF::::remove):
 827
 828 2011-05-26  Geoffrey Garen  <ggaren@apple.com>
 829
 830         Reviewed by Oliver Hunt.
 831
 832         Provide a real owner when copying a property table, for the sake of
 833         write barriers.
 834         https://bugs.webkit.org/show_bug.cgi?id=61547
 835
 836         No test because we can't enable the writeBarrier() ASSERT just yet.
 837
 838         * runtime/Structure.cpp:
 839         (JSC::Structure::addPropertyTransition):
 840
 841 2011-05-26  Adam Roben  <aroben@apple.com>
 842
 843         Windows build fix after r87346
 844
 845         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def: Fixed up exports to match
 846         reality.
 847
 848 2011-05-26  Patrick Gansterer  <paroga@webkit.org>
 849
 850         Reviewed by Adam Barth.
 851
 852         ASSERT(isMainThread()) when using single threaded jsc executable
 853         https://bugs.webkit.org/show_bug.cgi?id=60846
 854
 855         Remove the ASSERT since we do not have the concept of MainThread in JSC.
 856
 857         * wtf/CryptographicallyRandomNumber.cpp:
 858         (WTF::ARC4Stream::ARC4RandomNumberGenerator::randomNumber):
 859         (WTF::ARC4Stream::ARC4RandomNumberGenerator::randomValues):
 860
 861 2011-05-25  Gavin Barraclough  <barraclough@apple.com>
 862
 863         Reviewed by Sam Weinig.
 864
 865         https://bugs.webkit.org/show_bug.cgi?id=61506
 866
 867         Move the silent spill/fill methods in the DFG JIT to the JITCodeGenerator
 868         so that they are available to the SpeculativeJIT.
 869
 870         * dfg/DFGJITCodeGenerator.h:
 871         (JSC::DFG::JITCodeGenerator::silentSpillGPR):
 872         (JSC::DFG::JITCodeGenerator::silentSpillFPR):
 873         (JSC::DFG::JITCodeGenerator::silentFillGPR):
 874         (JSC::DFG::JITCodeGenerator::silentFillFPR):
 875         (JSC::DFG::JITCodeGenerator::silentSpillAllRegisters):
 876         (JSC::DFG::JITCodeGenerator::silentFillAllRegisters):
 877         * dfg/DFGNonSpeculativeJIT.h:
 878
 879 2011-05-25  Ryosuke Niwa  <rniwa@webkit.org>
 880
 881         An attempt to revive Windows bots.
 882
 883         * runtime/RegExp.cpp:
 884         * runtime/RegExp.h:
 885
 886 2011-05-25  Gavin Barraclough  <barraclough@apple.com>
 887
 888         Reviewed by Sam Weinig.
 889
 890         Bug 61503 - Move population of CodeBlock::m_structureStubInfos into JIT
 891
 892         This data structure, used at runtime by the JIT, is currently unnecessarily populated
 893         with default entries during byte compilation.
 894
 895         Aside from meaning that there is JIT specific code in the bytecompiler, this also ties
 896         us to one entry per corresponding bytecode op, which may be undesirable. Instead,
 897         populate this array from the JIT.
 898
 899         The type StructureStubInfo has two unused states, one for gets & one for puts. Unify
 900         these, so that the class can have a default constructor (and to simply switch statements
 901         in code walking over the table).
 902
 903         This change has ramification for the DFG JIT, in that the DFG JIT used this datastructure
 904         to check for functions containing property access. Instead do so in the DFGByteCodeParser.
 905
 906         * bytecode/CodeBlock.cpp:
 907         (JSC::printStructureStubInfo):
 908         * bytecode/CodeBlock.h:
 909         (JSC::CodeBlock::setNumberOfStructureStubInfos):
 910         (JSC::CodeBlock::numberOfStructureStubInfos):
 911         * bytecode/StructureStubInfo.cpp:
 912         (JSC::StructureStubInfo::deref):
 913         (JSC::StructureStubInfo::visitAggregate):
 914         * bytecode/StructureStubInfo.h:
 915         (JSC::StructureStubInfo::StructureStubInfo):
 916         * bytecompiler/BytecodeGenerator.cpp:
 917         (JSC::BytecodeGenerator::emitGetById):
 918         (JSC::BytecodeGenerator::emitPutById):
 919         (JSC::BytecodeGenerator::emitDirectPutById):
 920         * dfg/DFGByteCodeParser.cpp:
 921         (JSC::DFG::ByteCodeParser::parseBlock):
 922         * jit/JIT.cpp:
 923         (JSC::JIT::JIT):
 924         (JSC::JIT::privateCompileMainPass):
 925         (JSC::JIT::privateCompileSlowCases):
 926         (JSC::JIT::privateCompile):
 927         * jit/JIT.h:
 928         * jit/JITPropertyAccess.cpp:
 929         (JSC::JIT::emit_op_get_by_id):
 930         (JSC::JIT::emit_op_put_by_id):
 931         (JSC::JIT::emit_op_method_check):
 932         (JSC::JIT::compileGetByIdHotPath):
 933         (JSC::JIT::compileGetByIdSlowCase):
 934         (JSC::JIT::emitSlow_op_put_by_id):
 935         * jit/JITPropertyAccess32_64.cpp:
 936         (JSC::JIT::emit_op_get_by_id):
 937         (JSC::JIT::emitSlow_op_get_by_id):
 938         (JSC::JIT::emit_op_put_by_id):
 939         (JSC::JIT::emitSlow_op_put_by_id):
 940         (JSC::JIT::emit_op_method_check):
 941         (JSC::JIT::compileGetByIdHotPath):
 942         (JSC::JIT::compileGetByIdSlowCase):
 943         * runtime/Executable.cpp:
 944         (JSC::tryDFGCompile):
 945
 946 2011-05-25  Gavin Barraclough  <barraclough@apple.com>
 947
 948         Reviewed by Sam Weinig.
 949
 950         Bug 61501 - Unify AbstractMacroAssembler::differenceBetween methods.
 951
 952         * assembler/AbstractMacroAssembler.h:
 953         (JSC::AbstractMacroAssembler::Call::Call):
 954         (JSC::AbstractMacroAssembler::Call::fromTailJump):
 955         (JSC::AbstractMacroAssembler::Jump::Jump):
 956         (JSC::AbstractMacroAssembler::Jump::link):
 957         (JSC::AbstractMacroAssembler::Jump::linkTo):
 958         (JSC::AbstractMacroAssembler::Jump::isSet):
 959         (JSC::AbstractMacroAssembler::differenceBetween):
 960         (JSC::AbstractMacroAssembler::linkJump):
 961         (JSC::AbstractMacroAssembler::getLinkerCallReturnOffset):
 962         * assembler/LinkBuffer.h:
 963         (JSC::LinkBuffer::link):
 964         (JSC::LinkBuffer::locationOf):
 965         (JSC::LinkBuffer::locationOfNearCall):
 966         (JSC::LinkBuffer::returnAddressOffset):
 967         * assembler/MacroAssemblerARM.h:
 968         (JSC::MacroAssemblerARM::linkCall):
 969         * assembler/MacroAssemblerARMv7.h:
 970         (JSC::MacroAssemblerARMv7::linkCall):
 971         * assembler/MacroAssemblerMIPS.h:
 972         (JSC::MacroAssemblerMIPS::linkCall):
 973         * assembler/MacroAssemblerSH4.cpp:
 974         (JSC::MacroAssemblerSH4::linkCall):
 975         * assembler/MacroAssemblerX86.h:
 976         (JSC::MacroAssemblerX86::linkCall):
 977         * assembler/MacroAssemblerX86_64.h:
 978         (JSC::MacroAssemblerX86_64::linkCall):
 979
 980 2011-05-25  Gavin Barraclough  <barraclough@apple.com>
 981
 982         Reviewed by Sam Weinig.
 983
 984         https://bugs.webkit.org/show_bug.cgi?id=61500
 985         Add JSObject::offsetOfPropertyStorage
 986
 987         * jit/JITPropertyAccess.cpp:
 988         (JSC::JIT::compileGetDirectOffset):
 989         (JSC::JIT::compileGetByIdHotPath):
 990         (JSC::JIT::emit_op_put_by_id):
 991         (JSC::JIT::compilePutDirectOffset):
 992         * jit/JITPropertyAccess32_64.cpp:
 993         (JSC::JIT::compileGetByIdHotPath):
 994         (JSC::JIT::emit_op_put_by_id):
 995         (JSC::JIT::compilePutDirectOffset):
 996         (JSC::JIT::compileGetDirectOffset):
 997         * runtime/JSObject.h:
 998         (JSC::JSObject::offsetOfPropertyStorage):
 999
1000 2011-05-25  Oliver Hunt  <oliver@apple.com>
1001
1002         Reviewed by Geoffrey Garen.
1003
1004         Make RegExp GC allocated
1005         https://bugs.webkit.org/show_bug.cgi?id=61490
1006
1007         Make RegExp GC allocated.  Basically mechanical change to replace
1008         most use of [Pass]RefPtr<RegExp> with RegExp* or WriteBarrier<RegExp>
1009         where actual ownership happens.
1010
1011         Made the RegExpCache use Strong<> references currently to avoid any
1012         changes in behaviour.
1013
1014         * JavaScriptCore.exp:
1015         * bytecode/CodeBlock.cpp:
1016         (JSC::CodeBlock::visitAggregate):
1017         * bytecode/CodeBlock.h:
1018         (JSC::CodeBlock::addRegExp):
1019         * bytecompiler/BytecodeGenerator.cpp:
1020         (JSC::BytecodeGenerator::addRegExp):
1021         (JSC::BytecodeGenerator::emitNewRegExp):
1022         * bytecompiler/BytecodeGenerator.h:
1023         * runtime/JSCell.h:
1024         * runtime/JSGlobalData.cpp:
1025         (JSC::JSGlobalData::JSGlobalData):
1026         (JSC::JSGlobalData::clearBuiltinStructures):
1027         (JSC::JSGlobalData::addRegExpToTrace):
1028         * runtime/JSGlobalData.h:
1029         * runtime/JSGlobalObject.cpp:
1030         (JSC::JSGlobalObject::reset):
1031         * runtime/RegExp.cpp:
1032         (JSC::RegExp::RegExp):
1033         (JSC::RegExp::create):
1034         (JSC::RegExp::invalidateCode):
1035         * runtime/RegExp.h:
1036         (JSC::RegExp::createStructure):
1037         * runtime/RegExpCache.cpp:
1038         (JSC::RegExpCache::lookupOrCreate):
1039         (JSC::RegExpCache::create):
1040         * runtime/RegExpCache.h:
1041         * runtime/RegExpConstructor.cpp:
1042         (JSC::constructRegExp):
1043         * runtime/RegExpObject.cpp:
1044         (JSC::RegExpObject::RegExpObject):
1045         (JSC::RegExpObject::visitChildren):
1046         * runtime/RegExpObject.h:
1047         (JSC::RegExpObject::setRegExp):
1048         (JSC::RegExpObject::RegExpObjectData::RegExpObjectData):
1049         * runtime/RegExpPrototype.cpp:
1050         (JSC::RegExpPrototype::RegExpPrototype):
1051         (JSC::regExpProtoFuncCompile):
1052         * runtime/RegExpPrototype.h:
1053         * runtime/StringPrototype.cpp:
1054         (JSC::stringProtoFuncMatch):
1055         (JSC::stringProtoFuncSearch):
1056
1057 2011-05-25  Oliver Hunt  <oliver@apple.com>
1058
1059         Reviewed by Geoffrey Garen.
1060
1061         Generate regexp code lazily
1062         https://bugs.webkit.org/show_bug.cgi?id=61476
1063
1064         RegExp construction now simply validates the RegExp, it does
1065         not perform actual codegen.
1066
1067         * runtime/RegExp.cpp:
1068         (JSC::RegExp::RegExp):
1069         (JSC::RegExp::recompile):
1070         (JSC::RegExp::compile):
1071         (JSC::RegExp::match):
1072         * runtime/RegExp.h:
1073         (JSC::RegExp::recompileIfNecessary):
1074         * runtime/RegExpConstructor.h:
1075         (JSC::RegExpConstructor::performMatch):
1076         * runtime/RegExpObject.cpp:
1077         (JSC::RegExpObject::match):
1078         * runtime/StringPrototype.cpp:
1079         (JSC::stringProtoFuncReplace):
1080         (JSC::stringProtoFuncMatch):
1081         (JSC::stringProtoFuncSearch):
1082         (JSC::stringProtoFuncSplit):
1083
1084 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1085
1086         Reviewed by Geoffrey Garen.
1087
1088         Removed MarkSetProperties because it was unused
1089         https://bugs.webkit.org/show_bug.cgi?id=61418
1090
1091         * heap/MarkStack.h:
1092         (JSC::MarkSet::MarkSet):
1093         (JSC::MarkStack::append):
1094         * runtime/JSActivation.cpp:
1095         (JSC::JSActivation::visitChildren):
1096         * runtime/JSArray.h:
1097         (JSC::JSArray::visitChildrenDirect):
1098         * runtime/JSPropertyNameIterator.cpp:
1099         (JSC::JSPropertyNameIterator::visitChildren):
1100         * runtime/WriteBarrier.h:
1101         (JSC::MarkStack::appendValues):
1102
1103 2011-05-25  Oliver Hunt  <oliver@apple.com>
1104
1105         Reviewed by Geoffrey Garen.
1106
1107         Make allocations with guard pages ensure that the allocation succeeded
1108         https://bugs.webkit.org/show_bug.cgi?id=61453
1109
1110         Add null checks, and make PageBlock's operator bool() use
1111         the realbase, rather than the start of usable memory.
1112
1113         * wtf/OSAllocatorPosix.cpp:
1114         (WTF::OSAllocator::reserveAndCommit):
1115         * wtf/PageBlock.h:
1116         (WTF::PageBlock::operator bool):
1117         (WTF::PageBlock::PageBlock):
1118
1119 2011-04-10  Kevin Ollivier  <kevino@theolliviers.com>
1120
1121         Reviewed by Eric Seidel.
1122
1123         Add JS_EXPORT_PRIVATE macro for exported methods in bytecompiler headers.
1124
1125         https://bugs.webkit.org/show_bug.cgi?id=27551
1126
1127         * bytecompiler/BytecodeGenerator.h:
1128
1129 2011-05-24  Keishi Hattori  <keishi@webkit.org>
1130
1131         Reviewed by Kent Tamura.
1132
1133         Disable textfield implementation of <input type=color>. Add INPUT_COLOR feature flag. Add input color sanitizer.
1134         https://bugs.webkit.org/show_bug.cgi?id=61273
1135
1136         * Configurations/FeatureDefines.xcconfig: Added COLOR_INPUT feature flag.
1137
1138 2011-05-24  Kevin Ollivier  <kevino@theolliviers.com>
1139
1140         Reviewed by Eric Seidel.
1141
1142         Add export macros to WTFString.h.
1143
1144         https://bugs.webkit.org/show_bug.cgi?id=27551
1145
1146         * wtf/text/WTFString.h:
1147         (WTF::String::String):
1148         (WTF::String::findIgnoringCase):
1149         (WTF::String::isHashTableDeletedValue):
1150
1151 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1152
1153         Maybe fix the Mac build now?
1154
1155         * JavaScriptCore.xcodeproj/project.pbxproj:
1156
1157 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1158
1159         Maybe fix the Mac build?
1160
1161         * JavaScriptCore.xcodeproj/project.pbxproj:
1162
1163 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1164
1165         Reviewed by Oliver Hunt.
1166
1167         Split HeapRootVisitor into its own class
1168         https://bugs.webkit.org/show_bug.cgi?id=61399
1169
1170         * GNUmakefile.list.am:
1171         * JavaScriptCore.gypi:
1172         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
1173         * JavaScriptCore.xcodeproj/project.pbxproj:
1174         * heap/HandleHeap.cpp:
1175         * heap/HandleStack.cpp:
1176         * heap/Heap.cpp:
1177         * heap/HeapRootVisitor.h: Copied from Source/JavaScriptCore/heap/MarkStack.h.
1178         * heap/MarkStack.h:
1179         * runtime/ArgList.cpp:
1180         * runtime/SmallStrings.cpp:
1181
1182 2011-05-24  Jay Civelli  <jcivelli@chromium.org>
1183
1184         Rubberstamped by David Kilzer.
1185
1186         Updated some files that I forgot in my previous MHTML CL.
1187
1188         * Configurations/FeatureDefines.xcconfig:
1189
1190 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1191
1192         Fix the Mac build: Yes, please do remove these files, svn.
1193
1194         * JavaScriptCore.xcodeproj/project.pbxproj:
1195
1196 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1197
1198         Reviewed by Oliver Hunt.
1199
1200         Let's just have one way to get the system page size, bokay?
1201         https://bugs.webkit.org/show_bug.cgi?id=61384
1202
1203         * CMakeListsEfl.txt:
1204         * CMakeListsWinCE.txt:
1205         * GNUmakefile.list.am:
1206         * JavaScriptCore.exp:
1207         * JavaScriptCore.gypi:
1208         * JavaScriptCore.pro:
1209         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: MarkStack[Platform].cpp
1210         is gone completely now, since it only existed to provide a duplicate way
1211         to access the system page size.
1212
1213         * heap/MarkStack.cpp:
1214         (JSC::MarkStack::reset):
1215         * heap/MarkStack.h:
1216         (JSC::::MarkStackArray):
1217         (JSC::::shrinkAllocation): Use WTF::pageSize.
1218
1219         * heap/MarkStackPosix.cpp:
1220         * heap/MarkStackSymbian.cpp:
1221         * heap/MarkStackWin.cpp: Removed now-empty files.
1222
1223         * jit/ExecutableAllocator.cpp:
1224         (JSC::ExecutableAllocator::reprotectRegion):
1225         * jit/ExecutableAllocator.h:
1226         (JSC::ExecutableAllocator::ExecutableAllocator):
1227         (JSC::ExecutablePool::ExecutablePool):
1228         (JSC::ExecutablePool::poolAllocate):
1229         * jit/ExecutableAllocatorFixedVMPool.cpp: Use WTF::pageSize.
1230
1231         * wscript: Removed now-empty files.
1232
1233         * wtf/PageBlock.cpp:
1234         (WTF::systemPageSize): Integrated questionable Symbian page size rule
1235         from ExecutableAllocator, because that seems like what the original
1236         author should have done.
1237
1238 2011-05-24  Oliver Hunt  <oliver@apple.com>
1239
1240         Reviewed by Gavin Barraclough.
1241
1242         Interpreter crashes with gc validation enabled due to failure to mark initial cache structure
1243         https://bugs.webkit.org/show_bug.cgi?id=61385
1244
1245         The interpreter uses the structure slot of get_by_id and put_by_id to hold
1246         the initial structure it encountered so that it can identify whether a
1247         given access is stable.
1248
1249         When marking though we only visit the slot when we've decided to cache, and
1250         so this value could die.  This was "safe" as the value was only used for a
1251         pointer compare, but it was incorrect.  We now just mark the slot like we
1252         should have been doing already.
1253
1254         * bytecode/CodeBlock.cpp:
1255         (JSC::CodeBlock::visitStructures):
1256
1257 2011-05-24  Adam Roben  <aroben@apple.com>
1258
1259         Windows build fix
1260
1261         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def: Removed now-inline functions.
1262
1263 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1264
1265         Windows build fix: update the #if OS(WINDOWS) section to match my last patch.
1266
1267         * heap/MarkStack.h:
1268         (JSC::::shrinkAllocation):
1269
1270 2011-05-24  Geoffrey Garen  <ggaren@apple.com>
1271
1272         Rubber-stamped by Oliver Hunt.
1273
1274         Split out function definitions and class definitions from class
1275         declarations in MarkStack.h, for readability.
1276
1277         * heap/MarkStack.h:
1278         (JSC::MarkStack::MarkStack):
1279         (JSC::MarkStack::~MarkStack):
1280         (JSC::MarkStack::addOpaqueRoot):
1281         (JSC::MarkStack::containsOpaqueRoot):
1282         (JSC::MarkStack::opaqueRootCount):
1283         (JSC::MarkSet::MarkSet):
1284         (JSC::MarkStack::allocateStack):
1285         (JSC::MarkStack::releaseStack):
1286         (JSC::MarkStack::pageSize):
1287         (JSC::::MarkStackArray):
1288         (JSC::::~MarkStackArray):
1289         (JSC::::expand):
1290         (JSC::::append):
1291         (JSC::::removeLast):
1292         (JSC::::last):
1293         (JSC::::isEmpty):
1294         (JSC::::size):
1295         (JSC::::shrinkAllocation):
1296
1297 2011-05-24  Oliver Hunt  <oliver@apple.com>
1298
1299         Reviewed by Geoffrey Garen.
1300
1301         Avoid creating unnecessary identifiers and strings in the syntax checker
1302         https://bugs.webkit.org/show_bug.cgi?id=61378
1303
1304         Selectively tell the lexer that there are some places it does not need to
1305         do the real work of creating Identifiers for IDENT and STRING tokens.
1306
1307         Make parseString and parseIdentifier templatized on whether they should
1308         do real work, or merely validate the tokens.
1309
1310         SunSpider --parse-only reports ~5-8% win depending on hardware.
1311
1312         * parser/ASTBuilder.h:
1313         (JSC::ASTBuilder::createDotAccess):
1314         * parser/JSParser.cpp:
1315         (JSC::JSParser::next):
1316         (JSC::JSParser::consume):
1317         (JSC::JSParser::parseVarDeclarationList):
1318         (JSC::JSParser::parseConstDeclarationList):
1319         (JSC::JSParser::parseExpression):
1320         (JSC::JSParser::parseAssignmentExpression):
1321         (JSC::JSParser::parseConditionalExpression):
1322         (JSC::JSParser::parseBinaryExpression):
1323         (JSC::JSParser::parseProperty):
1324         (JSC::JSParser::parseObjectLiteral):
1325         (JSC::JSParser::parseArrayLiteral):
1326         (JSC::JSParser::parseArguments):
1327         (JSC::JSParser::parseMemberExpression):
1328         * parser/Lexer.cpp:
1329         (JSC::Lexer::parseIdentifier):
1330         (JSC::Lexer::parseString):
1331         (JSC::Lexer::lex):
1332         * parser/Lexer.h:
1333         * parser/SyntaxChecker.h:
1334         (JSC::SyntaxChecker::createDotAccess):
1335         (JSC::SyntaxChecker::createProperty):
1336
1337 2011-05-23  Michael Saboff  <msaboff@apple.com>
1338
1339         Reviewed by Mark Rowe.
1340
1341         Safari often freezes when clicking "Return free memory" in Caches dialog
1342         https://bugs.webkit.org/show_bug.cgi?id=61325
1343
1344         There are two fixes and improvement in instrumentation code used to find
1345         one of the problems.
1346         Changed ReleaseFreeList() to set the "decommitted" bit when releasing
1347         pages to the system and moving Spans from the normal list to the returned
1348         list.
1349         Added a "not making forward progress" check to TCMalloc_PageHeap::scavenge
1350         to eliminate an infinite loop if we can't meet the pagesToRelease target.
1351         Added a check for the decommitted bit being set properly in
1352         TCMalloc_PageHeap::CheckList.
1353
1354         * wtf/FastMalloc.cpp:
1355         (WTF::TCMalloc_PageHeap::scavenge):
1356         (WTF::TCMalloc_PageHeap::Check):
1357         (WTF::TCMalloc_PageHeap::CheckList):
1358         (WTF::ReleaseFreeList):
1359
1360 2011-05-23  Gavin Barraclough  <barraclough@apple.com>
1361
1362         Reviewed by Geoff Garen.
1363
1364         https://bugs.webkit.org/show_bug.cgi?id=61306
1365
1366         The begin characters optimization currently has issues (#61129),
1367         and does not appear to still be a performance win. The prudent
1368         next step seems to be to disable while we ascertain whether this
1369         is still a useful performance optimization.
1370
1371         * yarr/YarrInterpreter.cpp:
1372         (JSC::Yarr::Interpreter::matchDisjunction):
1373         (JSC::Yarr::Interpreter::interpret):
1374         * yarr/YarrInterpreter.h:
1375         (JSC::Yarr::BytecodePattern::BytecodePattern):
1376         * yarr/YarrPattern.cpp:
1377         (JSC::Yarr::YarrPatternConstructor::YarrPatternConstructor):
1378         (JSC::Yarr::YarrPattern::compile):
1379         (JSC::Yarr::YarrPattern::YarrPattern):
1380         * yarr/YarrPattern.h:
1381         (JSC::Yarr::YarrPattern::reset):
1382
1383 2011-05-23  Matthew Delaney  <mdelaney@apple.com>
1384
1385         Reviewed by Simon Fraser.
1386
1387         Remove safeFloatToInt() in FloatRect.cpp and replace with working version of clampToInteger()
1388         https://bugs.webkit.org/show_bug.cgi?id=58216
1389
1390         * wtf/MathExtras.h:
1391         (clampToInteger):
1392         (clampToPositiveInteger):
1393
1394 2011-05-23  Ruben  <chromium@hybridsource.org>
1395
1396         Reviewed by Tony Chang.
1397
1398         Chromium gyp patch to use new POSIX defines toolkit_uses_gtk and os_posix
1399         https://bugs.webkit.org/show_bug.cgi?id=61219
1400
1401         * JavaScriptCore.gyp/JavaScriptCore.gyp:
1402
1403 2011-05-23  Thouraya ANDOLSI  <thouraya.andolsi@st.com>
1404
1405         Reviewed by Gavin Barraclough.
1406
1407         [SH4] AssemblerLabel does not name a type
1408         https://bugs.webkit.org/show_bug.cgi?id=59927
1409
1410         SH4Assembler.h file shoold be included before AbstractMacroAssembler.h.
1411
1412         * assembler/MacroAssemblerSH4.h:
1413
1414 2011-05-23  Ryuan Choi  <ryuan.choi@samsung.com>
1415
1416         Rubber stamped by Eric Seidel.
1417
1418         [CMAKE] Refactoring wtf related code.
1419         https://bugs.webkit.org/show_bug.cgi?id=60146
1420
1421         Move wtf-files to Source/JavaScriptCore/wtf/CMakeLists.txt.
1422
1423         * CMakeLists.txt:
1424         * CMakeListsEfl.txt:
1425         * wtf/CMakeLists.txt:
1426         * wtf/CMakeListsEfl.txt:
1427
1428 2011-05-22  Adam Barth  <abarth@webkit.org>
1429
1430         Enable strict PassOwnPtr for everyone.  I expect this patch will need
1431         some followups to make the GTK and EFL bots green again.
1432
1433         * wtf/PassOwnPtr.h:
1434
1435 2011-05-20  Oliver Hunt  <oliver@apple.com>
1436
1437         Reviewed by Gavin Barraclough.
1438
1439         Reduce size of inline cache path of get_by_id on ARMv7
1440         https://bugs.webkit.org/show_bug.cgi?id=61221
1441
1442         This reduces the code size of get_by_id by 20 bytes
1443
1444         * assembler/ARMv7Assembler.h:
1445         (JSC::ARMv7Assembler::ldrCompact):
1446         (JSC::ARMv7Assembler::repatchCompact):
1447         (JSC::ARMv7Assembler::setUInt7ForLoad):
1448         * assembler/MacroAssemblerARMv7.h:
1449         (JSC::MacroAssemblerARMv7::load32WithCompactAddressOffsetPatch):
1450         * jit/JIT.h:
1451
1452 2011-05-20  Zoltan Herczeg  <zherczeg@inf.u-szeged.hu>
1453
1454         Reviewed by Oliver Hunt.
1455
1456         Zombies should "live" forever
1457         https://bugs.webkit.org/show_bug.cgi?id=61170
1458
1459         Reusing zombie cells could still hide garbage
1460         collected cell related bugs.
1461
1462         * JavaScriptCore.pro:
1463         * heap/MarkedBlock.cpp:
1464         (JSC::MarkedBlock::clearMarks):
1465         * heap/MarkedBlock.h:
1466         * heap/MarkedSpace.cpp:
1467         (JSC::MarkedSpace::destroy):
1468         * runtime/JSCell.h:
1469         (JSC::JSCell::JSValue::isZombie):
1470         * runtime/JSZombie.h:
1471         (JSC::JSZombie::~JSZombie):
1472         * runtime/WriteBarrier.h:
1473         (JSC::WriteBarrierBase::setWithoutWriteBarrier):
1474
1475 2011-05-20  Brady Eidson  <beidson@apple.com>
1476
1477         Reviewed by Sam Weinig.
1478
1479         <rdar://problem/9472883> and https://bugs.webkit.org/show_bug.cgi?id=61203
1480         Horrendous bug in callOnMainThreadAndWait
1481
1482         * wtf/MainThread.cpp:
1483         (WTF::dispatchFunctionsFromMainThread): Before signaling the background thread with the
1484           syncFlag condition, reacquire the mutex first.
1485
1486 2011-05-20  Oliver Hunt  <oliver@apple.com>
1487
1488         Reviewed by Sam Weinig.
1489
1490         Remove unnecessary double->int conversion at the end of op_div
1491         https://bugs.webkit.org/show_bug.cgi?id=61198
1492
1493         We don't attempt this conversion on 64bit, removing it actually speeds
1494         up sunspider and v8 slightly, and it reduces code size.
1495
1496         * jit/JITArithmetic32_64.cpp:
1497         (JSC::JIT::emit_op_div):
1498
1499 2011-05-19  Evan Martin  <evan@chromium.org>
1500
1501         Reviewed by Tony Chang.
1502
1503         [chromium] remove <(library) variable
1504         https://bugs.webkit.org/show_bug.cgi?id=61158
1505
1506         This was for a build experiment; we can just use the correct value now.
1507
1508         * JavaScriptCore.gyp/JavaScriptCore.gyp:
1509
1510 2011-05-20  Oliver Hunt  <oliver@apple.com>
1511
1512         Reviewed by Sam Weinig.
1513
1514         Interpreter uses wrong bytecode offset for determining exception handler
1515         https://bugs.webkit.org/show_bug.cgi?id=61191
1516
1517         The bytecode offset given for the returnPC from the JIT is
1518         actually the offset for the start of the instruction triggering
1519         the call, whereas in the interpreter it is the actual return
1520         VPC.  This means if the next instruction following a call was
1521         in an exception region we would incorrectly redirect to its
1522         handler.  Long term we want to completely redo how exceptions
1523         are handled anyway so the simplest and lowest risk fix here is
1524         to simply subtract one from the return vPC so that we have an
1525         offset in the triggering instruction.
1526
1527         It turns out this is caught by a couple of tests already.
1528
1529         * interpreter/Interpreter.cpp:
1530         (JSC::Interpreter::unwindCallFrame):
1531
1532 2011-05-20  Xan Lopez  <xlopez@igalia.com>
1533
1534         Reviewed by Oliver Hunt.
1535
1536         JIT requires VM overcommit (particularly on x86-64), Linux does not by default support this without swap?
1537         https://bugs.webkit.org/show_bug.cgi?id=42756
1538
1539         Use the MAP_NORESERVE flag for mmap on Linux to skip the kernel
1540         check of the available memory. This should give us an
1541         overcommit-like behavior in most systems, which is what we want.
1542
1543         * wtf/OSAllocatorPosix.cpp:
1544         (WTF::OSAllocator::reserveAndCommit): pass MAP_NORSERVE to mmap.
1545
1546 2011-05-19  Gabor Loki  <loki@webkit.org>
1547
1548         Fix ARM build after r86919
1549
1550         * assembler/ARMAssembler.h:
1551         (JSC::ARMAssembler::nop):
1552
1553 2011-05-19  Oliver Hunt  <oliver@apple.com>
1554
1555         Reviewed by Gavin Barraclough.
1556
1557         Randomise code starting location a little
1558         https://bugs.webkit.org/show_bug.cgi?id=61161
1559
1560         Add a nop() function to the Assemblers so that we
1561         can randomise code offsets slightly at no real cost.
1562
1563         * assembler/ARMAssembler.h:
1564         (JSC::ARMAssembler::nop):
1565         * assembler/ARMv7Assembler.h:
1566         (JSC::ARMv7Assembler::nop):
1567         * assembler/MacroAssemblerARM.h:
1568         (JSC::MacroAssemblerARM::nop):
1569         * assembler/MacroAssemblerARMv7.h:
1570         (JSC::MacroAssemblerARMv7::nop):
1571         * assembler/MacroAssemblerMIPS.h:
1572         (JSC::MacroAssemblerMIPS::nop):
1573         * assembler/MacroAssemblerSH4.h:
1574         (JSC::MacroAssemblerSH4::nop):
1575         * assembler/MacroAssemblerX86Common.h:
1576         (JSC::MacroAssemblerX86Common::nop):
1577         * assembler/X86Assembler.h:
1578         (JSC::X86Assembler::nop):
1579         * jit/JIT.cpp:
1580         (JSC::JIT::JIT):
1581         (JSC::JIT::privateCompile):
1582         * jit/JIT.h:
1583         * runtime/WeakRandom.h:
1584         (JSC::WeakRandom::getUint32):
1585
1586 2011-05-19  Oliver Hunt  <oliver@apple.com>
1587
1588         Fix windows build.
1589
1590         * wtf/OSAllocatorWin.cpp:
1591         (WTF::OSAllocator::reserveUncommitted):
1592         (WTF::OSAllocator::reserveAndCommit):
1593
1594 2011-05-19  Oliver Hunt  <oliver@apple.com>
1595
1596         Reviewed by Gavin Barraclough.
1597
1598         Add guard pages to each end of the memory region used by the fixedvm allocator
1599         https://bugs.webkit.org/show_bug.cgi?id=61150
1600
1601         Add mechanism to notify the OSAllocator that pages at either end of an
1602         allocation should be considered guard pages.  Update PageReservation,
1603         PageAllocation, etc to handle this.
1604
1605         * JavaScriptCore.exp:
1606         * jit/ExecutableAllocatorFixedVMPool.cpp:
1607         (JSC::FixedVMPoolAllocator::FixedVMPoolAllocator):
1608         * wtf/OSAllocator.h:
1609         * wtf/OSAllocatorPosix.cpp:
1610         (WTF::OSAllocator::reserveUncommitted):
1611         (WTF::OSAllocator::reserveAndCommit):
1612         * wtf/PageAllocation.h:
1613         (WTF::PageAllocation::PageAllocation):
1614         * wtf/PageAllocationAligned.h:
1615         (WTF::PageAllocationAligned::PageAllocationAligned):
1616         * wtf/PageBlock.h:
1617         (WTF::PageBlock::PageBlock):
1618         * wtf/PageReservation.h:
1619         (WTF::PageReservation::reserve):
1620         (WTF::PageReservation::reserveWithGuardPages):
1621             Add a new function to make a reservation that will add guard
1622             pages to the ends of an allocation.
1623         (WTF::PageReservation::PageReservation):
1624
1625 2011-05-19  Oliver Hunt  <oliver@apple.com>
1626
1627         Reviewed by Geoffrey Garen.
1628
1629         Make Executables release their JIT code as soon as they become dead
1630         https://bugs.webkit.org/show_bug.cgi?id=61134
1631
1632         Add an ability to clear an Executable's jit code without requiring
1633         it to be destroyed, and then call that from a finalizer.
1634
1635         * heap/Weak.h:
1636         (JSC::Weak::Weak):
1637         (JSC::Weak::leak):
1638         * jit/JITCode.h:
1639         (JSC::JITCode::clear):
1640         * runtime/Executable.cpp:
1641         (JSC::ExecutableFinalizer::finalize):
1642         (JSC::ExecutableBase::executableFinalizer):
1643         * runtime/Executable.h:
1644         (JSC::ExecutableBase::ExecutableBase):
1645         (JSC::ExecutableBase::clearExecutableCode):
1646
1647 2011-05-19  Adam Roben  <aroben@apple.com>
1648
1649         Remove a redundant and broken data export
1650
1651         Data can't be exported from JavaScriptCore.dll by listing it in the .def file. The
1652         JS_EXPORTDATA macro must be used instead. (In this case it was already being used, leading
1653         to a linker warning about multiple definitions.)
1654
1655         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def: Removed JSGlobalData::s_info.
1656
1657 2011-05-18  Oliver Hunt  <oliver@apple.com>
1658
1659         Reviewed by Gavin Barraclough.
1660
1661         Some tests crashing in JSC::MarkStack::validateValue beneath ScriptController::clearWindowShell on SnowLeopard Intel Release (WebKit2 Tests)
1662         https://bugs.webkit.org/show_bug.cgi?id=61064
1663
1664         Switch NonFinalObject to using WriteBarrier<> rather than WriteBarrierBase<>
1665         for its inline storage.  This resolves the problem of GC occurring before
1666         a subclass has initialised its anonymous storage.
1667
1668         * runtime/JSObject.h:
1669
1670 2011-05-18  Adam Barth  <abarth@webkit.org>
1671
1672         Reviewed by Sam Weinig.
1673
1674         Delete WTFURL
1675         https://bugs.webkit.org/show_bug.cgi?id=61084
1676
1677         It's been a year and we've failed to complete this project.  It's time
1678         to throw in the towel.
1679
1680         * JavaScriptCore.xcodeproj/project.pbxproj:
1681         * wtf/url: Removed.
1682         * wtf/url/api: Removed.
1683         * wtf/url/api/ParsedURL.cpp: Removed.
1684         * wtf/url/api/ParsedURL.h: Removed.
1685         * wtf/url/api/URLString.h: Removed.
1686         * wtf/url/src: Removed.
1687         * wtf/url/src/RawURLBuffer.h: Removed.
1688         * wtf/url/src/URLBuffer.h: Removed.
1689         * wtf/url/src/URLCharacterTypes.cpp: Removed.
1690         * wtf/url/src/URLCharacterTypes.h: Removed.
1691         * wtf/url/src/URLComponent.h: Removed.
1692         * wtf/url/src/URLEscape.cpp: Removed.
1693         * wtf/url/src/URLEscape.h: Removed.
1694         * wtf/url/src/URLParser.h: Removed.
1695         * wtf/url/src/URLQueryCanonicalizer.h: Removed.
1696         * wtf/url/src/URLSegments.cpp: Removed.
1697         * wtf/url/src/URLSegments.h: Removed.
1698         * wtf/url/wtfurl.gyp: Removed.
1699
1700 2011-05-18  Oliver Hunt  <oliver@apple.com>
1701
1702         Reviewed by Sam Weinig.
1703
1704         JSGlobalObject and some others do GC allocation during initialization, which can cause heap corruption
1705         https://bugs.webkit.org/show_bug.cgi?id=61090
1706
1707         Remove the Structure-free JSGlobalObject constructor and instead always
1708         pass the structure into the JSGlobalObject constructor.
1709         Stop DebuggerActivation creating a new structure every time, and simply
1710         use a single shared structure held by the GlobalData.
1711
1712         * API/JSContextRef.cpp:
1713         * debugger/DebuggerActivation.cpp:
1714         (JSC::DebuggerActivation::DebuggerActivation):
1715         * jsc.cpp:
1716         (GlobalObject::GlobalObject):
1717         (functionRun):
1718         (jscmain):
1719         * runtime/JSGlobalData.cpp:
1720         (JSC::JSGlobalData::JSGlobalData):
1721         (JSC::JSGlobalData::clearBuiltinStructures):
1722         * runtime/JSGlobalData.h:
1723         * runtime/JSGlobalObject.h:
1724
1725 2011-05-18  Oliver Hunt  <oliver@apple.com>
1726
1727         Reviewed by Adam Roben.
1728
1729         Disable gc validation in release builds
1730         https://bugs.webkit.org/show_bug.cgi?id=60680
1731
1732         Add back the NDEBUG check
1733
1734         * wtf/Platform.h:
1735
1736 2011-05-17  Geoffrey Garen  <ggaren@apple.com>
1737
1738         Rolled out attempts to fix EFL build because they're not enough -- the
1739         build script needs to be fixed.
1740
1741         * runtime/BooleanPrototype.cpp:
1742         * runtime/DateConstructor.cpp:
1743         * runtime/ErrorPrototype.cpp:
1744
1745 2011-05-17  Geoffrey Garen  <ggaren@apple.com>
1746
1747         More attempts to work around the EFL build system being borken.
1748
1749         * runtime/DateConstructor.cpp:
1750         * runtime/ErrorPrototype.cpp:
1751
1752 2011-05-17  Geoffrey Garen  <ggaren@apple.com>
1753
1754         Try to fix the EFL build.
1755
1756         * runtime/BooleanPrototype.cpp:
1757
1758 2011-05-16  Geoffrey Garen  <ggaren@apple.com>
1759
1760         Rolling back in r86653 with build fixed.
1761
1762         Reviewed by Gavin Barraclough and Oliver Hunt.
1763
1764         Global object initialization is expensive
1765         https://bugs.webkit.org/show_bug.cgi?id=60933
1766
1767         Changed a bunch of globals to allocate their properties lazily, and changed
1768         the global object to allocate a bunch of its globals lazily.
1769
1770         This reduces the footprint of a global object from 287 objects with 58
1771         functions for 24K to 173 objects with 20 functions for 15K.
1772
1773         Large patch, but it's all mechanical.
1774
1775         * DerivedSources.make:
1776         * JavaScriptCore.exp: Build!
1777
1778         * create_hash_table: Added a special case for fromCharCode, since it uses
1779         a custom "thunk generator".
1780
1781         * heap/Heap.cpp:
1782         (JSC::TypeCounter::operator()): Fixed a bug where the type counter would
1783         overcount objects that were owned through more than one mechanism because
1784         it was getting in the way of counting the results for this patch.
1785
1786         * interpreter/CallFrame.h:
1787         (JSC::ExecState::arrayConstructorTable):
1788         (JSC::ExecState::arrayPrototypeTable):
1789         (JSC::ExecState::booleanPrototypeTable):
1790         (JSC::ExecState::dateConstructorTable):
1791         (JSC::ExecState::errorPrototypeTable):
1792         (JSC::ExecState::globalObjectTable):
1793         (JSC::ExecState::numberConstructorTable):
1794         (JSC::ExecState::numberPrototypeTable):
1795         (JSC::ExecState::objectPrototypeTable):
1796         (JSC::ExecState::regExpPrototypeTable):
1797         (JSC::ExecState::stringConstructorTable): Added new tables.
1798
1799         * runtime/ArrayConstructor.cpp:
1800         (JSC::ArrayConstructor::ArrayConstructor):
1801         (JSC::ArrayConstructor::getOwnPropertySlot):
1802         (JSC::ArrayConstructor::getOwnPropertyDescriptor):
1803         * runtime/ArrayConstructor.h:
1804         (JSC::ArrayConstructor::createStructure):
1805         * runtime/ArrayPrototype.cpp:
1806         (JSC::ArrayPrototype::getOwnPropertySlot):
1807         (JSC::ArrayPrototype::getOwnPropertyDescriptor):
1808         * runtime/ArrayPrototype.h:
1809         * runtime/BooleanPrototype.cpp:
1810         (JSC::BooleanPrototype::BooleanPrototype):
1811         (JSC::BooleanPrototype::getOwnPropertySlot):
1812         (JSC::BooleanPrototype::getOwnPropertyDescriptor):
1813         * runtime/BooleanPrototype.h:
1814         (JSC::BooleanPrototype::createStructure):
1815         * runtime/DateConstructor.cpp:
1816         (JSC::DateConstructor::DateConstructor):
1817         (JSC::DateConstructor::getOwnPropertySlot):
1818         (JSC::DateConstructor::getOwnPropertyDescriptor):
1819         * runtime/DateConstructor.h:
1820         (JSC::DateConstructor::createStructure):
1821         * runtime/ErrorPrototype.cpp:
1822         (JSC::ErrorPrototype::ErrorPrototype):
1823         (JSC::ErrorPrototype::getOwnPropertySlot):
1824         (JSC::ErrorPrototype::getOwnPropertyDescriptor):
1825         * runtime/ErrorPrototype.h:
1826         (JSC::ErrorPrototype::createStructure): Standardized these objects
1827         to use static tables for function properties.
1828
1829         * runtime/JSGlobalData.cpp:
1830         (JSC::JSGlobalData::JSGlobalData):
1831         (JSC::JSGlobalData::~JSGlobalData):
1832         * runtime/JSGlobalData.h: Added new tables.
1833
1834         * runtime/JSGlobalObject.cpp:
1835         (JSC::JSGlobalObject::reset):
1836         (JSC::JSGlobalObject::addStaticGlobals):
1837         (JSC::JSGlobalObject::getOwnPropertySlot):
1838         (JSC::JSGlobalObject::getOwnPropertyDescriptor):
1839         * runtime/JSGlobalObject.h:
1840         * runtime/JSGlobalObjectFunctions.cpp:
1841         * runtime/JSGlobalObjectFunctions.h: Changed JSGlobalObject to use a
1842         static table for its global functions. This required uninlining some
1843         things to avoid a circular header dependency. However, those things
1844         probably shouldn't have been inlined in the first place.
1845
1846         Even more global object properties can be made lazy, but that requires
1847         more in-depth changes.
1848
1849         * runtime/MathObject.cpp:
1850         * runtime/NumberConstructor.cpp:
1851         (JSC::NumberConstructor::getOwnPropertySlot):
1852         (JSC::NumberConstructor::getOwnPropertyDescriptor):
1853         * runtime/NumberPrototype.cpp:
1854         (JSC::NumberPrototype::NumberPrototype):
1855         (JSC::NumberPrototype::getOwnPropertySlot):
1856         (JSC::NumberPrototype::getOwnPropertyDescriptor):
1857         * runtime/NumberPrototype.h:
1858         (JSC::NumberPrototype::createStructure):
1859         * runtime/ObjectPrototype.cpp:
1860         (JSC::ObjectPrototype::ObjectPrototype):
1861         (JSC::ObjectPrototype::put):
1862         (JSC::ObjectPrototype::getOwnPropertySlot):
1863         (JSC::ObjectPrototype::getOwnPropertyDescriptor):
1864         * runtime/ObjectPrototype.h:
1865         (JSC::ObjectPrototype::createStructure):
1866         * runtime/RegExpPrototype.cpp:
1867         (JSC::RegExpPrototype::RegExpPrototype):
1868         (JSC::RegExpPrototype::getOwnPropertySlot):
1869         (JSC::RegExpPrototype::getOwnPropertyDescriptor):
1870         * runtime/RegExpPrototype.h:
1871         (JSC::RegExpPrototype::createStructure):
1872         * runtime/StringConstructor.cpp:
1873         (JSC::StringConstructor::StringConstructor):
1874         (JSC::StringConstructor::getOwnPropertySlot):
1875         (JSC::StringConstructor::getOwnPropertyDescriptor):
1876         * runtime/StringConstructor.h:
1877         (JSC::StringConstructor::createStructure): Standardized these objects
1878         to use static tables for function properties.
1879
1880 2011-05-17  Sam Weinig  <sam@webkit.org>
1881
1882         Reviewed by Oliver Hunt.
1883
1884         JSGlobalContextRelease should not trigger a synchronous garbage collection
1885         https://bugs.webkit.org/show_bug.cgi?id=60990
1886
1887         * API/JSContextRef.cpp:
1888         Change synchronous call to collectAllGarbage to a call to trigger the
1889         activityCallback.
1890
1891 2011-05-16  Oliver Hunt  <oliver@apple.com>
1892
1893         Reviewed by Gavin Barraclough.
1894
1895         Reduce code size for inline cache
1896         https://bugs.webkit.org/show_bug.cgi?id=60942
1897
1898         This patch introduces the concept of a "compact" address that
1899         allows individual architectures to control the maximum offset
1900         used for the inline path of get_by_id.  This reduces the code
1901         size of get_by_id by 3 bytes on x86 and x86_64 and slightly
1902         improves performance on v8 tests.
1903
1904         * assembler/ARMAssembler.h:
1905         (JSC::ARMAssembler::repatchCompact):
1906         * assembler/ARMv7Assembler.h:
1907         (JSC::ARMv7Assembler::repatchCompact):
1908         * assembler/AbstractMacroAssembler.h:
1909         (JSC::AbstractMacroAssembler::DataLabelCompact::DataLabelCompact):
1910         (JSC::AbstractMacroAssembler::differenceBetween):
1911         (JSC::AbstractMacroAssembler::repatchCompact):
1912         * assembler/CodeLocation.h:
1913         (JSC::CodeLocationDataLabelCompact::CodeLocationDataLabelCompact):
1914         (JSC::CodeLocationCommon::dataLabelCompactAtOffset):
1915         * assembler/LinkBuffer.h:
1916         (JSC::LinkBuffer::locationOf):
1917         * assembler/MIPSAssembler.h:
1918         (JSC::MIPSAssembler::repatchCompact):
1919         * assembler/MacroAssembler.h:
1920         (JSC::MacroAssembler::loadPtrWithCompactAddressOffsetPatch):
1921         * assembler/MacroAssemblerARM.h:
1922         (JSC::MacroAssemblerARM::load32WithCompactAddressOffsetPatch):
1923         * assembler/MacroAssemblerARMv7.h:
1924         (JSC::MacroAssemblerARMv7::load32WithCompactAddressOffsetPatch):
1925         * assembler/MacroAssemblerMIPS.h:
1926         (JSC::MacroAssemblerMIPS::load32WithCompactAddressOffsetPatch):
1927         * assembler/MacroAssemblerSH4.h:
1928         (JSC::MacroAssemblerSH4::load32WithAddressOffsetPatch):
1929         * assembler/MacroAssemblerX86.h:
1930         (JSC::MacroAssemblerX86::repatchCompact):
1931         * assembler/MacroAssemblerX86Common.h:
1932         (JSC::MacroAssemblerX86Common::loadCompactWithAddressOffsetPatch):
1933         * assembler/MacroAssemblerX86_64.h:
1934         (JSC::MacroAssemblerX86_64::loadPtrWithCompactAddressOffsetPatch):
1935         * assembler/RepatchBuffer.h:
1936         (JSC::RepatchBuffer::repatch):
1937         * assembler/SH4Assembler.h:
1938         (JSC::SH4Assembler::repatchCompact):
1939         * assembler/X86Assembler.h:
1940         (JSC::X86Assembler::movl_mr_disp8):
1941         (JSC::X86Assembler::movq_mr_disp8):
1942         (JSC::X86Assembler::repatchCompact):
1943         (JSC::X86Assembler::setInt8):
1944         (JSC::X86Assembler::X86InstructionFormatter::oneByteOp_disp8):
1945         (JSC::X86Assembler::X86InstructionFormatter::oneByteOp64_disp8):
1946         (JSC::X86Assembler::X86InstructionFormatter::memoryModRM):
1947         * jit/JIT.h:
1948         * jit/JITPropertyAccess.cpp:
1949         (JSC::JIT::compileGetByIdHotPath):
1950         (JSC::JIT::emit_op_put_by_id):
1951         (JSC::JIT::patchGetByIdSelf):
1952         * jit/JITPropertyAccess32_64.cpp:
1953         (JSC::JIT::compileGetByIdHotPath):
1954         (JSC::JIT::emit_op_put_by_id):
1955         (JSC::JIT::patchGetByIdSelf):
1956         * jit/JITStubs.cpp:
1957         (JSC::JITThunks::tryCacheGetByID):
1958
1959 2011-05-16  Sheriff Bot  <webkit.review.bot@gmail.com>
1960
1961         Unreviewed, rolling out r86653.
1962         http://trac.webkit.org/changeset/86653
1963         https://bugs.webkit.org/show_bug.cgi?id=60944
1964
1965         "Caused regressions on Windows, OSX and EFL" (Requested by
1966         yutak on #webkit).
1967
1968         * DerivedSources.make:
1969         * DerivedSources.pro:
1970         * GNUmakefile.am:
1971         * GNUmakefile.list.am:
1972         * JavaScriptCore.exp:
1973         * JavaScriptCore.gypi:
1974         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
1975         * create_hash_table:
1976         * heap/Heap.cpp:
1977         (JSC::TypeCounter::operator()):
1978         * interpreter/CallFrame.h:
1979         (JSC::ExecState::arrayTable):
1980         (JSC::ExecState::numberTable):
1981         * runtime/ArrayConstructor.cpp:
1982         (JSC::ArrayConstructor::ArrayConstructor):
1983         * runtime/ArrayConstructor.h:
1984         * runtime/ArrayPrototype.cpp:
1985         (JSC::ArrayPrototype::getOwnPropertySlot):
1986         (JSC::ArrayPrototype::getOwnPropertyDescriptor):
1987         * runtime/ArrayPrototype.h:
1988         * runtime/BooleanPrototype.cpp:
1989         (JSC::BooleanPrototype::BooleanPrototype):
1990         * runtime/BooleanPrototype.h:
1991         * runtime/DateConstructor.cpp:
1992         (JSC::DateConstructor::DateConstructor):
1993         * runtime/DateConstructor.h:
1994         * runtime/ErrorPrototype.cpp:
1995         (JSC::ErrorPrototype::ErrorPrototype):
1996         * runtime/ErrorPrototype.h:
1997         * runtime/JSGlobalData.cpp:
1998         (JSC::JSGlobalData::JSGlobalData):
1999         (JSC::JSGlobalData::~JSGlobalData):
2000         * runtime/JSGlobalData.h:
2001         * runtime/JSGlobalObject.cpp:
2002         (JSC::JSGlobalObject::reset):
2003         * runtime/JSGlobalObject.h:
2004         (JSC::JSGlobalObject::addStaticGlobals):
2005         (JSC::JSGlobalObject::getOwnPropertySlot):
2006         (JSC::JSGlobalObject::getOwnPropertyDescriptor):
2007         * runtime/JSGlobalObjectFunctions.cpp:
2008         (JSC::globalFuncJSCPrint):
2009         * runtime/JSGlobalObjectFunctions.h:
2010         * runtime/MathObject.cpp:
2011         * runtime/NumberConstructor.cpp:
2012         (JSC::NumberConstructor::getOwnPropertySlot):
2013         (JSC::NumberConstructor::getOwnPropertyDescriptor):
2014         * runtime/NumberPrototype.cpp:
2015         (JSC::NumberPrototype::NumberPrototype):
2016         * runtime/NumberPrototype.h:
2017         * runtime/ObjectPrototype.cpp:
2018         (JSC::ObjectPrototype::ObjectPrototype):
2019         (JSC::ObjectPrototype::put):
2020         (JSC::ObjectPrototype::getOwnPropertySlot):
2021         * runtime/ObjectPrototype.h:
2022         * runtime/RegExpPrototype.cpp:
2023         (JSC::RegExpPrototype::RegExpPrototype):
2024         * runtime/RegExpPrototype.h:
2025         * runtime/StringConstructor.cpp:
2026         (JSC::StringConstructor::StringConstructor):
2027         * runtime/StringConstructor.h:
2028
2029 2011-05-16  Geoffrey Garen  <ggaren@apple.com>
2030
2031         Reviewed by Geoffrey Garen.
2032
2033         Global object initialization is expensive
2034         https://bugs.webkit.org/show_bug.cgi?id=60933
2035
2036         Changed a bunch of globals to allocate their properties lazily, and changed
2037         the global object to allocate a bunch of its globals lazily.
2038
2039         This reduces the footprint of a global object from 287 objects with 58
2040         functions for 24K to 173 objects with 20 functions for 15K.
2041
2042         Large patch, but it's all mechanical.
2043
2044         * DerivedSources.make:
2045         * JavaScriptCore.exp: Build!
2046
2047         * create_hash_table: Added a special case for fromCharCode, since it uses
2048         a custom "thunk generator".
2049
2050         * heap/Heap.cpp:
2051         (JSC::TypeCounter::operator()): Fixed a bug where the type counter would
2052         overcount objects that were owned through more than one mechanism because
2053         it was getting in the way of counting the results for this patch.
2054
2055         * interpreter/CallFrame.h:
2056         (JSC::ExecState::arrayConstructorTable):
2057         (JSC::ExecState::arrayPrototypeTable):
2058         (JSC::ExecState::booleanPrototypeTable):
2059         (JSC::ExecState::dateConstructorTable):
2060         (JSC::ExecState::errorPrototypeTable):
2061         (JSC::ExecState::globalObjectTable):
2062         (JSC::ExecState::numberConstructorTable):
2063         (JSC::ExecState::numberPrototypeTable):
2064         (JSC::ExecState::objectPrototypeTable):
2065         (JSC::ExecState::regExpPrototypeTable):
2066         (JSC::ExecState::stringConstructorTable): Added new tables.
2067
2068         * runtime/ArrayConstructor.cpp:
2069         (JSC::ArrayConstructor::ArrayConstructor):
2070         (JSC::ArrayConstructor::getOwnPropertySlot):
2071         (JSC::ArrayConstructor::getOwnPropertyDescriptor):
2072         * runtime/ArrayConstructor.h:
2073         (JSC::ArrayConstructor::createStructure):
2074         * runtime/ArrayPrototype.cpp:
2075         (JSC::ArrayPrototype::getOwnPropertySlot):
2076         (JSC::ArrayPrototype::getOwnPropertyDescriptor):
2077         * runtime/ArrayPrototype.h:
2078         * runtime/BooleanPrototype.cpp:
2079         (JSC::BooleanPrototype::BooleanPrototype):
2080         (JSC::BooleanPrototype::getOwnPropertySlot):
2081         (JSC::BooleanPrototype::getOwnPropertyDescriptor):
2082         * runtime/BooleanPrototype.h:
2083         (JSC::BooleanPrototype::createStructure):
2084         * runtime/DateConstructor.cpp:
2085         (JSC::DateConstructor::DateConstructor):
2086         (JSC::DateConstructor::getOwnPropertySlot):
2087         (JSC::DateConstructor::getOwnPropertyDescriptor):
2088         * runtime/DateConstructor.h:
2089         (JSC::DateConstructor::createStructure):
2090         * runtime/ErrorPrototype.cpp:
2091         (JSC::ErrorPrototype::ErrorPrototype):
2092         (JSC::ErrorPrototype::getOwnPropertySlot):
2093         (JSC::ErrorPrototype::getOwnPropertyDescriptor):
2094         * runtime/ErrorPrototype.h:
2095         (JSC::ErrorPrototype::createStructure): Standardized these objects
2096         to use static tables for function properties.
2097
2098         * runtime/JSGlobalData.cpp:
2099         (JSC::JSGlobalData::JSGlobalData):
2100         (JSC::JSGlobalData::~JSGlobalData):
2101         * runtime/JSGlobalData.h: Added new tables.
2102
2103         * runtime/JSGlobalObject.cpp:
2104         (JSC::JSGlobalObject::reset):
2105         (JSC::JSGlobalObject::addStaticGlobals):
2106         (JSC::JSGlobalObject::getOwnPropertySlot):
2107         (JSC::JSGlobalObject::getOwnPropertyDescriptor):
2108         * runtime/JSGlobalObject.h:
2109         * runtime/JSGlobalObjectFunctions.cpp:
2110         * runtime/JSGlobalObjectFunctions.h: Changed JSGlobalObject to use a
2111         static table for its global functions. This required uninlining some
2112         things to avoid a circular header dependency. However, those things
2113         probably shouldn't have been inlined in the first place.
2114
2115         Even more global object properties can be made lazy, but that requires
2116         more in-depth changes.
2117
2118         * runtime/MathObject.cpp:
2119         * runtime/NumberConstructor.cpp:
2120         (JSC::NumberConstructor::getOwnPropertySlot):
2121         (JSC::NumberConstructor::getOwnPropertyDescriptor):
2122         * runtime/NumberPrototype.cpp:
2123         (JSC::NumberPrototype::NumberPrototype):
2124         (JSC::NumberPrototype::getOwnPropertySlot):
2125         (JSC::NumberPrototype::getOwnPropertyDescriptor):
2126         * runtime/NumberPrototype.h:
2127         (JSC::NumberPrototype::createStructure):
2128         * runtime/ObjectPrototype.cpp:
2129         (JSC::ObjectPrototype::ObjectPrototype):
2130         (JSC::ObjectPrototype::put):
2131         (JSC::ObjectPrototype::getOwnPropertySlot):
2132         (JSC::ObjectPrototype::getOwnPropertyDescriptor):
2133         * runtime/ObjectPrototype.h:
2134         (JSC::ObjectPrototype::createStructure):
2135         * runtime/RegExpPrototype.cpp:
2136         (JSC::RegExpPrototype::RegExpPrototype):
2137         (JSC::RegExpPrototype::getOwnPropertySlot):
2138         (JSC::RegExpPrototype::getOwnPropertyDescriptor):
2139         * runtime/RegExpPrototype.h:
2140         (JSC::RegExpPrototype::createStructure):
2141         * runtime/StringConstructor.cpp:
2142         (JSC::StringConstructor::StringConstructor):
2143         (JSC::StringConstructor::getOwnPropertySlot):
2144         (JSC::StringConstructor::getOwnPropertyDescriptor):
2145         * runtime/StringConstructor.h:
2146         (JSC::StringConstructor::createStructure): Standardized these objects
2147         to use static tables for function properties.
2148
2149 2011-05-16  David Kilzer  <ddkilzer@apple.com>
2150
2151         <http://webkit.org/b/60913> C++ exceptions should not be enabled when building with llvm-gcc-4.2
2152         <rdar://problem/9446430>
2153
2154         Reviewed by Mark Rowe.
2155
2156         * Configurations/Base.xcconfig: Fixed typo.
2157
2158 2011-05-16  Oliver Hunt  <oliver@apple.com>
2159
2160         Reviewed by Geoffrey Garen.
2161
2162         JSWeakObjectMap finalisation may occur while gc is in inconsistent state
2163         https://bugs.webkit.org/show_bug.cgi?id=60908
2164         <rdar://problem/9409491>
2165
2166         We need to ensure that we have called all the weak map finalizers while
2167         the global object (and hence global context) is still in a consistent
2168         state.  The best way to achieve this is to simply use a weak handle and
2169         finalizer on the global object.
2170
2171         * JavaScriptCore.exp:
2172         * runtime/JSGlobalObject.cpp:
2173         (JSC::JSGlobalObject::WeakMapFinalizer::finalize):
2174         * runtime/JSGlobalObject.h:
2175         (JSC::JSGlobalObject::registerWeakMap):
2176
2177 2011-05-16  Siddharth Mathur  <siddharth.mathur@nokia.com>
2178
2179         Reviewed by Laszlo Gombos.
2180
2181         [Qt][WK2][Symbian] Shared memory implementation for Symbian
2182         https://bugs.webkit.org/show_bug.cgi?id=55875
2183
2184         * wtf/Platform.h: Exclude Symbian OS from USE(UNIX_DOMAIN_SOCKETS) users
2185
2186 2011-05-16  Gavin Barraclough  <barraclough@apple.com>
2187
2188         Rubber stamped by Geoff Garen.
2189
2190         https://bugs.webkit.org/show_bug.cgi?id=60866
2191         Evaluation order broken for empty alternatives in subpatterns
2192
2193         Reverting https://bugs.webkit.org/show_bug.cgi?id=51395
2194
2195         * yarr/YarrPattern.cpp:
2196         (JSC::Yarr::YarrPatternConstructor::atomParenthesesEnd):
2197
2198 2011-05-15  Gavin Barraclough  <barraclough@apple.com>
2199
2200         Reviewed by Geoff Garen & Michael Saboff.
2201
2202         https://bugs.webkit.org/show_bug.cgi?id=60860
2203         Simplify backtracking in YARR JIT
2204
2205         YARR JIT currently performs a single pass of code generation over the pattern,
2206         with special handling to allow the code generation for some backtracking code
2207         out of line. We can simplify things by moving to a common mechanism whereby all
2208         forwards matching code is generated in one pass, and all backtracking code is
2209         generated in another. Backtracking code can be generated in reverse order, to
2210         optimized the common fall-through case.
2211
2212         To make it easier to walk over the pattern, we can first convert to a more
2213         byte-code like format before JIT generating. In time we should unify this with
2214         the YARR interpreter to more closely unify the two.
2215
2216         * yarr/YarrJIT.cpp:
2217         (JSC::Yarr::YarrGenerator::jumpIfNoAvailableInput):
2218         (JSC::Yarr::YarrGenerator::YarrOp::YarrOp):
2219         (JSC::Yarr::YarrGenerator::BacktrackingState::BacktrackingState):
2220         (JSC::Yarr::YarrGenerator::BacktrackingState::append):
2221         (JSC::Yarr::YarrGenerator::BacktrackingState::fallthrough):
2222         (JSC::Yarr::YarrGenerator::BacktrackingState::link):
2223         (JSC::Yarr::YarrGenerator::BacktrackingState::linkTo):
2224         (JSC::Yarr::YarrGenerator::BacktrackingState::takeBacktracksToJumpList):
2225         (JSC::Yarr::YarrGenerator::BacktrackingState::isEmpty):
2226         (JSC::Yarr::YarrGenerator::BacktrackingState::linkDataLabels):
2227         (JSC::Yarr::YarrGenerator::BacktrackingState::ReturnAddressRecord::ReturnAddressRecord):
2228         (JSC::Yarr::YarrGenerator::generateAssertionBOL):
2229         (JSC::Yarr::YarrGenerator::backtrackAssertionBOL):
2230         (JSC::Yarr::YarrGenerator::generateAssertionEOL):
2231         (JSC::Yarr::YarrGenerator::backtrackAssertionEOL):
2232         (JSC::Yarr::YarrGenerator::matchAssertionWordchar):
2233         (JSC::Yarr::YarrGenerator::generateAssertionWordBoundary):
2234         (JSC::Yarr::YarrGenerator::backtrackAssertionWordBoundary):
2235         (JSC::Yarr::YarrGenerator::generatePatternCharacterOnce):
2236         (JSC::Yarr::YarrGenerator::backtrackPatternCharacterOnce):
2237         (JSC::Yarr::YarrGenerator::generatePatternCharacterFixed):
2238         (JSC::Yarr::YarrGenerator::backtrackPatternCharacterFixed):
2239         (JSC::Yarr::YarrGenerator::generatePatternCharacterGreedy):
2240         (JSC::Yarr::YarrGenerator::backtrackPatternCharacterGreedy):
2241         (JSC::Yarr::YarrGenerator::generatePatternCharacterNonGreedy):
2242         (JSC::Yarr::YarrGenerator::backtrackPatternCharacterNonGreedy):
2243         (JSC::Yarr::YarrGenerator::generateCharacterClassOnce):
2244         (JSC::Yarr::YarrGenerator::backtrackCharacterClassOnce):
2245         (JSC::Yarr::YarrGenerator::generateCharacterClassFixed):
2246         (JSC::Yarr::YarrGenerator::backtrackCharacterClassFixed):
2247         (JSC::Yarr::YarrGenerator::generateCharacterClassGreedy):
2248         (JSC::Yarr::YarrGenerator::backtrackCharacterClassGreedy):
2249         (JSC::Yarr::YarrGenerator::generateCharacterClassNonGreedy):
2250         (JSC::Yarr::YarrGenerator::backtrackCharacterClassNonGreedy):
2251         (JSC::Yarr::YarrGenerator::generateTerm):
2252         (JSC::Yarr::YarrGenerator::backtrackTerm):
2253         (JSC::Yarr::YarrGenerator::generate):
2254         (JSC::Yarr::YarrGenerator::backtrack):
2255         (JSC::Yarr::YarrGenerator::opCompileParenthesesSubpattern):
2256         (JSC::Yarr::YarrGenerator::opCompileParentheticalAssertion):
2257         (JSC::Yarr::YarrGenerator::opCompileAlternative):
2258         (JSC::Yarr::YarrGenerator::opCompileBody):
2259         (JSC::Yarr::YarrGenerator::YarrGenerator):
2260         (JSC::Yarr::YarrGenerator::compile):
2261
2262 2011-05-15  Adam Barth  <abarth@webkit.org>
2263
2264         Enable strict PassOwnPtr on Qt.  (Build fixes to follow.)
2265
2266         * wtf/PassOwnPtr.h:
2267
2268 2011-05-15  Geoffrey Garen  <ggaren@apple.com>
2269
2270         Reviewed by Maciej Stachowiak.
2271
2272         Partial fix for <rdar://problem/9417875> REGRESSION: SunSpider ~17% slower
2273         in browser than on command line
2274
2275         This patch fixes a few issues in generated code that could unreasonably
2276         prolong object lifetimes.
2277
2278         * heap/Heap.cpp:
2279         (JSC::Heap::collectAllGarbage): Throw away all function code before doing
2280         a major collection. We want to clear polymorphic caches, since they can
2281         keep alive large object graphs that have gone "stale". For the same reason,
2282         but to a lesser extent, we also want to clear linked functions and other
2283         one-off caches.
2284
2285         This has the side-benefit of reducing memory footprint from run-once
2286         functions, and of allowing predictions and caches that have failed to
2287         re-specialize.
2288
2289         Eventually, if compilation costs rise far enough, we may want a more
2290         limited strategy for de-specializing code without throwing it away
2291         completely, but this works for now, and it's the simplest solution.
2292
2293         * jit/JITStubs.cpp:
2294         (JSC::JITThunks::hostFunctionStub):
2295         * jit/JITStubs.h:
2296         * runtime/JSFunction.cpp: Made the host function stub cache weak --
2297         otherwise it's effectively a memory leak that can seriously fragment the
2298         GC and JIT heaps.
2299
2300         (JSC::JSFunction::JSFunction):
2301         (JSC::JSFunction::visitChildren): Cleared up some comments that confused
2302         me when working with this code.
2303
2304 2011-05-13  Oliver Hunt  <oliver@apple.com>
2305
2306         Reviewed by Geoffrey Garen.
2307
2308         Make GC validation more aggressive
2309         https://bugs.webkit.org/show_bug.cgi?id=60802
2310
2311         This patch makes the checks performed under GC_VALIDATION
2312         much more aggressive, and adds the checks to more places
2313         in order to allow us to catch GC bugs much closer to the
2314         point of failure.
2315
2316         * JavaScriptCore.exp:
2317         * JavaScriptCore.xcodeproj/project.pbxproj:
2318         * debugger/DebuggerActivation.cpp:
2319         (JSC::DebuggerActivation::visitChildren):
2320         * heap/MarkedBlock.cpp:
2321         (JSC::MarkedBlock::MarkedBlock):
2322         * heap/MarkedSpace.cpp:
2323         * runtime/Arguments.cpp:
2324         (JSC::Arguments::visitChildren):
2325         * runtime/Executable.cpp:
2326         (JSC::EvalExecutable::visitChildren):
2327         (JSC::ProgramExecutable::visitChildren):
2328         (JSC::FunctionExecutable::visitChildren):
2329         * runtime/Executable.h:
2330         * runtime/GetterSetter.cpp:
2331         (JSC::GetterSetter::visitChildren):
2332         * runtime/GetterSetter.h:
2333         * runtime/JSAPIValueWrapper.h:
2334         (JSC::JSAPIValueWrapper::createStructure):
2335         (JSC::JSAPIValueWrapper::JSAPIValueWrapper):
2336         * runtime/JSActivation.cpp:
2337         (JSC::JSActivation::visitChildren):
2338         * runtime/JSArray.cpp:
2339         (JSC::JSArray::visitChildren):
2340         * runtime/JSCell.cpp:
2341         (JSC::slowValidateCell):
2342         * runtime/JSCell.h:
2343         (JSC::JSCell::JSCell::unvalidatedStructure):
2344         (JSC::JSCell::JSCell::JSCell):
2345         * runtime/JSFunction.cpp:
2346         (JSC::JSFunction::visitChildren):
2347         * runtime/JSGlobalObject.cpp:
2348         (JSC::JSGlobalObject::visitChildren):
2349         (JSC::slowValidateCell):
2350         * runtime/JSONObject.h:
2351         * runtime/JSObject.cpp:
2352         (JSC::JSObject::visitChildren):
2353         * runtime/JSPropertyNameIterator.cpp:
2354         (JSC::JSPropertyNameIterator::visitChildren):
2355         * runtime/JSPropertyNameIterator.h:
2356         * runtime/JSStaticScopeObject.cpp:
2357         (JSC::JSStaticScopeObject::visitChildren):
2358         * runtime/JSString.h:
2359         (JSC::RopeBuilder::JSString):
2360         * runtime/JSWrapperObject.cpp:
2361         (JSC::JSWrapperObject::visitChildren):
2362         * runtime/NativeErrorConstructor.cpp:
2363         (JSC::NativeErrorConstructor::visitChildren):
2364         * runtime/PropertyMapHashTable.h:
2365         (JSC::PropertyMapEntry::PropertyMapEntry):
2366         * runtime/RegExpObject.cpp:
2367         (JSC::RegExpObject::visitChildren):
2368         * runtime/ScopeChain.cpp:
2369         (JSC::ScopeChainNode::visitChildren):
2370         * runtime/ScopeChain.h:
2371         (JSC::ScopeChainNode::ScopeChainNode):
2372         * runtime/Structure.cpp:
2373         (JSC::Structure::Structure):
2374         (JSC::Structure::addPropertyTransition):
2375         (JSC::Structure::visitChildren):
2376         * runtime/Structure.h:
2377         (JSC::JSCell::classInfo):
2378         * runtime/StructureChain.cpp:
2379         (JSC::StructureChain::visitChildren):
2380         * runtime/StructureChain.h:
2381         * runtime/WriteBarrier.h:
2382         (JSC::validateCell):
2383         (JSC::JSCell):
2384         (JSC::JSGlobalObject):
2385         (JSC::WriteBarrierBase::set):
2386         (JSC::WriteBarrierBase::setMayBeNull):
2387         (JSC::WriteBarrierBase::setEarlyValue):
2388         (JSC::WriteBarrierBase::get):
2389         (JSC::WriteBarrierBase::operator*):
2390         (JSC::WriteBarrierBase::operator->):
2391         (JSC::WriteBarrierBase::unvalidatedGet):
2392         (JSC::WriteBarrier::WriteBarrier):
2393         * wtf/Assertions.h:
2394
2395 2011-05-13  Oliver Hunt  <oliver@apple.com>
2396
2397         Reviewed by Geoffrey Garen.
2398
2399         Make GC validation more aggressive
2400         https://bugs.webkit.org/show_bug.cgi?id=60802
2401
2402         This patch makes the checks performed under GC_VALIDATION
2403         much more aggressive, and adds the checks to more places
2404         in order to allow us to catch GC bugs much closer to the
2405         point of failure.
2406
2407         * JavaScriptCore.exp:
2408         * JavaScriptCore.xcodeproj/project.pbxproj:
2409         * debugger/DebuggerActivation.cpp:
2410         (JSC::DebuggerActivation::visitChildren):
2411         * heap/MarkedBlock.cpp:
2412         (JSC::MarkedBlock::MarkedBlock):
2413         * heap/MarkedSpace.cpp:
2414         * runtime/Arguments.cpp:
2415         (JSC::Arguments::visitChildren):
2416         * runtime/Executable.cpp:
2417         (JSC::EvalExecutable::visitChildren):
2418         (JSC::ProgramExecutable::visitChildren):
2419         (JSC::FunctionExecutable::visitChildren):
2420         * runtime/Executable.h:
2421         * runtime/GetterSetter.cpp:
2422         (JSC::GetterSetter::visitChildren):
2423         * runtime/GetterSetter.h:
2424         * runtime/JSAPIValueWrapper.h:
2425         (JSC::JSAPIValueWrapper::createStructure):
2426         (JSC::JSAPIValueWrapper::JSAPIValueWrapper):
2427         * runtime/JSActivation.cpp:
2428         (JSC::JSActivation::visitChildren):
2429         * runtime/JSArray.cpp:
2430         (JSC::JSArray::visitChildren):
2431         * runtime/JSCell.cpp:
2432         (JSC::slowValidateCell):
2433         * runtime/JSCell.h:
2434         (JSC::JSCell::JSCell::unvalidatedStructure):
2435         (JSC::JSCell::JSCell::JSCell):
2436         * runtime/JSFunction.cpp:
2437         (JSC::JSFunction::visitChildren):
2438         * runtime/JSGlobalObject.cpp:
2439         (JSC::JSGlobalObject::visitChildren):
2440         (JSC::slowValidateCell):
2441         * runtime/JSONObject.h:
2442         * runtime/JSObject.cpp:
2443         (JSC::JSObject::visitChildren):
2444         * runtime/JSPropertyNameIterator.cpp:
2445         (JSC::JSPropertyNameIterator::visitChildren):
2446         * runtime/JSPropertyNameIterator.h:
2447         * runtime/JSStaticScopeObject.cpp:
2448         (JSC::JSStaticScopeObject::visitChildren):
2449         * runtime/JSString.h:
2450         (JSC::RopeBuilder::JSString):
2451         * runtime/JSWrapperObject.cpp:
2452         (JSC::JSWrapperObject::visitChildren):
2453         * runtime/NativeErrorConstructor.cpp:
2454         (JSC::NativeErrorConstructor::visitChildren):
2455         * runtime/PropertyMapHashTable.h:
2456         (JSC::PropertyMapEntry::PropertyMapEntry):
2457         * runtime/RegExpObject.cpp:
2458         (JSC::RegExpObject::visitChildren):
2459         * runtime/ScopeChain.cpp:
2460         (JSC::ScopeChainNode::visitChildren):
2461         * runtime/ScopeChain.h:
2462         (JSC::ScopeChainNode::ScopeChainNode):
2463         * runtime/Structure.cpp:
2464         (JSC::Structure::Structure):
2465         (JSC::Structure::addPropertyTransition):
2466         (JSC::Structure::visitChildren):
2467         * runtime/Structure.h:
2468         (JSC::JSCell::classInfo):
2469         * runtime/StructureChain.cpp:
2470         (JSC::StructureChain::visitChildren):
2471         * runtime/StructureChain.h:
2472         * runtime/WriteBarrier.h:
2473         (JSC::validateCell):
2474         (JSC::JSCell):
2475         (JSC::JSGlobalObject):
2476         (JSC::WriteBarrierBase::set):
2477         (JSC::WriteBarrierBase::setMayBeNull):
2478         (JSC::WriteBarrierBase::setEarlyValue):
2479         (JSC::WriteBarrierBase::get):
2480         (JSC::WriteBarrierBase::operator*):
2481         (JSC::WriteBarrierBase::operator->):
2482         (JSC::WriteBarrierBase::unvalidatedGet):
2483         (JSC::WriteBarrier::WriteBarrier):
2484         * wtf/Assertions.h:
2485
2486 2011-05-14  Csaba Osztrogonác  <ossy@webkit.org>
2487
2488         Unreviewed, rolling out r86469 and r86471, because they made hundreds tests crash on Qt.
2489
2490         Make GC validation more aggressive
2491         https://bugs.webkit.org/show_bug.cgi?id=60802
2492
2493         * JavaScriptCore.exp:
2494         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
2495         * JavaScriptCore.xcodeproj/project.pbxproj:
2496         * debugger/DebuggerActivation.cpp:
2497         (JSC::DebuggerActivation::visitChildren):
2498         * heap/MarkedBlock.cpp:
2499         (JSC::MarkedBlock::MarkedBlock):
2500         * heap/MarkedSpace.cpp:
2501         * runtime/Arguments.cpp:
2502         (JSC::Arguments::visitChildren):
2503         * runtime/Executable.cpp:
2504         (JSC::EvalExecutable::visitChildren):
2505         (JSC::ProgramExecutable::visitChildren):
2506         (JSC::FunctionExecutable::visitChildren):
2507         * runtime/Executable.h:
2508         (JSC::ProgramExecutable::createStructure):
2509         (JSC::FunctionExecutable::createStructure):
2510         * runtime/GetterSetter.cpp:
2511         (JSC::GetterSetter::visitChildren):
2512         * runtime/GetterSetter.h:
2513         (JSC::GetterSetter::createStructure):
2514         * runtime/JSAPIValueWrapper.h:
2515         (JSC::JSAPIValueWrapper::createStructure):
2516         * runtime/JSActivation.cpp:
2517         (JSC::JSActivation::visitChildren):
2518         * runtime/JSArray.cpp:
2519         (JSC::JSArray::visitChildren):
2520         * runtime/JSCell.cpp:
2521         * runtime/JSCell.h:
2522         (JSC::JSCell::JSCell::JSCell):
2523         * runtime/JSFunction.cpp:
2524         (JSC::JSFunction::visitChildren):
2525         * runtime/JSGlobalObject.cpp:
2526         (JSC::JSGlobalObject::visitChildren):
2527         * runtime/JSONObject.h:
2528         (JSC::JSONObject::createStructure):
2529         * runtime/JSObject.cpp:
2530         (JSC::JSObject::visitChildren):
2531         * runtime/JSPropertyNameIterator.cpp:
2532         (JSC::JSPropertyNameIterator::visitChildren):
2533         * runtime/JSPropertyNameIterator.h:
2534         * runtime/JSStaticScopeObject.cpp:
2535         (JSC::JSStaticScopeObject::visitChildren):
2536         * runtime/JSString.h:
2537         (JSC::RopeBuilder::createStructure):
2538         * runtime/JSWrapperObject.cpp:
2539         (JSC::JSWrapperObject::visitChildren):
2540         * runtime/NativeErrorConstructor.cpp:
2541         (JSC::NativeErrorConstructor::visitChildren):
2542         * runtime/PropertyMapHashTable.h:
2543         (JSC::PropertyMapEntry::PropertyMapEntry):
2544         * runtime/RegExpObject.cpp:
2545         (JSC::RegExpObject::visitChildren):
2546         * runtime/ScopeChain.cpp:
2547         (JSC::ScopeChainNode::visitChildren):
2548         * runtime/ScopeChain.h:
2549         (JSC::ScopeChainNode::ScopeChainNode):
2550         * runtime/Structure.cpp:
2551         (JSC::Structure::Structure):
2552         (JSC::Structure::addPropertyTransition):
2553         (JSC::Structure::visitChildren):
2554         * runtime/Structure.h:
2555         (JSC::Structure::createStructure):
2556         (JSC::JSCell::classInfo):
2557         * runtime/StructureChain.cpp:
2558         (JSC::StructureChain::visitChildren):
2559         * runtime/StructureChain.h:
2560         * runtime/WriteBarrier.h:
2561         (JSC::WriteBarrierBase::set):
2562         (JSC::WriteBarrierBase::get):
2563         (JSC::WriteBarrierBase::operator*):
2564         (JSC::WriteBarrierBase::operator->):
2565         (JSC::WriteBarrier::WriteBarrier):
2566         * wtf/Assertions.h:
2567
2568 2011-05-13  Oliver Hunt  <oliver@apple.com>
2569
2570         Reviewed by Geoffrey Garen.
2571
2572         Make GC validation more aggressive
2573         https://bugs.webkit.org/show_bug.cgi?id=60802
2574
2575         This patch makes the checks performed under GC_VALIDATION
2576         much more aggressive, and adds the checks to more places
2577         in order to allow us to catch GC bugs much closer to the
2578         point of failure.
2579
2580         * JavaScriptCore.exp:
2581         * JavaScriptCore.xcodeproj/project.pbxproj:
2582         * debugger/DebuggerActivation.cpp:
2583         (JSC::DebuggerActivation::visitChildren):
2584         * heap/MarkedBlock.cpp:
2585         (JSC::MarkedBlock::MarkedBlock):
2586         * heap/MarkedSpace.cpp:
2587         * runtime/Arguments.cpp:
2588         (JSC::Arguments::visitChildren):
2589         * runtime/Executable.cpp:
2590         (JSC::EvalExecutable::visitChildren):
2591         (JSC::ProgramExecutable::visitChildren):
2592         (JSC::FunctionExecutable::visitChildren):
2593         * runtime/Executable.h:
2594         * runtime/GetterSetter.cpp:
2595         (JSC::GetterSetter::visitChildren):
2596         * runtime/GetterSetter.h:
2597         * runtime/JSAPIValueWrapper.h:
2598         (JSC::JSAPIValueWrapper::createStructure):
2599         (JSC::JSAPIValueWrapper::JSAPIValueWrapper):
2600         * runtime/JSActivation.cpp:
2601         (JSC::JSActivation::visitChildren):
2602         * runtime/JSArray.cpp:
2603         (JSC::JSArray::visitChildren):
2604         * runtime/JSCell.cpp:
2605         (JSC::slowValidateCell):
2606         * runtime/JSCell.h:
2607         (JSC::JSCell::JSCell::unvalidatedStructure):
2608         (JSC::JSCell::JSCell::JSCell):
2609         * runtime/JSFunction.cpp:
2610         (JSC::JSFunction::visitChildren):
2611         * runtime/JSGlobalObject.cpp:
2612         (JSC::JSGlobalObject::visitChildren):
2613         (JSC::slowValidateCell):
2614         * runtime/JSONObject.h:
2615         * runtime/JSObject.cpp:
2616         (JSC::JSObject::visitChildren):
2617         * runtime/JSPropertyNameIterator.cpp:
2618         (JSC::JSPropertyNameIterator::visitChildren):
2619         * runtime/JSPropertyNameIterator.h:
2620         * runtime/JSStaticScopeObject.cpp:
2621         (JSC::JSStaticScopeObject::visitChildren):
2622         * runtime/JSString.h:
2623         (JSC::RopeBuilder::JSString):
2624         * runtime/JSWrapperObject.cpp:
2625         (JSC::JSWrapperObject::visitChildren):
2626         * runtime/NativeErrorConstructor.cpp:
2627         (JSC::NativeErrorConstructor::visitChildren):
2628         * runtime/PropertyMapHashTable.h:
2629         (JSC::PropertyMapEntry::PropertyMapEntry):
2630         * runtime/RegExpObject.cpp:
2631         (JSC::RegExpObject::visitChildren):
2632         * runtime/ScopeChain.cpp:
2633         (JSC::ScopeChainNode::visitChildren):
2634         * runtime/ScopeChain.h:
2635         (JSC::ScopeChainNode::ScopeChainNode):
2636         * runtime/Structure.cpp:
2637         (JSC::Structure::Structure):
2638         (JSC::Structure::addPropertyTransition):
2639         (JSC::Structure::visitChildren):
2640         * runtime/Structure.h:
2641         (JSC::JSCell::classInfo):
2642         * runtime/StructureChain.cpp:
2643         (JSC::StructureChain::visitChildren):
2644         * runtime/StructureChain.h:
2645         * runtime/WriteBarrier.h:
2646         (JSC::validateCell):
2647         (JSC::JSCell):
2648         (JSC::JSGlobalObject):
2649         (JSC::WriteBarrierBase::set):
2650         (JSC::WriteBarrierBase::setMayBeNull):
2651         (JSC::WriteBarrierBase::setEarlyValue):
2652         (JSC::WriteBarrierBase::get):
2653         (JSC::WriteBarrierBase::operator*):
2654         (JSC::WriteBarrierBase::operator->):
2655         (JSC::WriteBarrierBase::unvalidatedGet):
2656         (JSC::WriteBarrier::WriteBarrier):
2657         * wtf/Assertions.h:
2658
2659 2011-05-01  Holger Hans Peter Freyther  <holger@moiji-mobile.com>
2660
2661         Reviewed by Steve Block.
2662
2663         [android] OS(ANDROID) does not imply PLATFORM(ANDROID)
2664         https://bugs.webkit.org/show_bug.cgi?id=59888
2665
2666         It is possible to build QtWebKit and others for OS(ANDROID). Let
2667         the buildsystem decide which platform is to be build.
2668
2669         * wtf/Platform.h:
2670
2671 2011-05-12  Maciej Stachowiak  <mjs@apple.com>
2672
2673         Reviewed by Darin Adler.
2674
2675         XMLDocumentParserLibxml2 should play nice with strict OwnPtrs
2676         https://bugs.webkit.org/show_bug.cgi?id=59394
2677
2678         This portion of the change introduces a PassTraits template, which
2679         is used to enable takeFirst() to work for a Deque holding OwnPtrs,
2680         and optimize it for a Deque holding RefPtrs. In the future it can
2681         be deployed elsewhere to make our data structures work better with
2682         our smart pointers.
2683
2684         * GNUmakefile.list.am:
2685         * JavaScriptCore.gypi:
2686         * JavaScriptCore.vcproj/WTF/WTF.vcproj:
2687         * JavaScriptCore.xcodeproj/project.pbxproj:
2688         * wtf/CMakeLists.txt:
2689         * wtf/Deque.h:
2690         (WTF::::takeFirst):
2691         * wtf/PassTraits.h: Added.
2692         (WTF::PassTraits::transfer):
2693
2694 2011-05-12  Nikolas Zimmermann  <nzimmermann@rim.com>
2695
2696         Not reviewed.
2697
2698         Revert r86334, it broke the win build. WinCE build is fixed even without this patch. WinCairo remains broken atm, everything else works.
2699
2700         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
2701
2702 2011-05-12  Nikolas Zimmermann  <nzimmermann@rim.com>
2703
2704         Not reviewed.
2705
2706         String operator+ reallocates unnecessarily when concatting > 2 strings
2707         https://bugs.webkit.org/show_bug.cgi?id=58420
2708
2709         Try to fix WinCE/WinCairo linking by exporting three symbols, not sure whether it's correct though. Win worked just fine before.
2710
2711         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
2712
2713 2011-05-12  Nikolas Zimmermann  <nzimmermann@rim.com>
2714
2715         Reviewed by Darin Adler.
2716
2717         String operator+ reallocates unnecessarily when concatting > 2 strings
2718         https://bugs.webkit.org/show_bug.cgi?id=58420
2719
2720         Provide a faster String append operator.
2721         Up until now, "String operator+(const String& a, const String& b)" copied String a into a temporary
2722         object, and used a.append(b), which reallocates a new buffer of aLength+bLength. When concatting
2723         N strings using operator+, this leads to N-1 reallocations.
2724
2725         Replace this with a flexible operator+ implementation, that avoids these reallocations.
2726         When concatting a 'String' with any string type (char*, UChar, Vector<char>, String, AtomicString, etc..)
2727         a StringAppend<String, T> object is created, which holds the intermediate string objects, and delays
2728         creation of the final string, until operator String() is invoked.
2729
2730         template<typename T>
2731         StringAppend<String, T> operator+(const String& string1, T string2)
2732         {
2733             return StringAppend<String, T>(string1, string2);
2734         }
2735
2736         template<typename U, typename V, typename W>
2737         StringAppend<U, StringAppend<V, W> > operator+(U string1, const StringAppend<V, W>& string2)
2738         {
2739             return StringAppend<U, StringAppend<V, W> >(string1, string2);
2740         }
2741
2742         When concatting three strings - "String a, b, c; String result = a + b + c;" following happens:
2743         first a StringAppend<String, String> object is created by operator+(const String& string1, String string2).
2744         Then operator+(String string1, const StringAppend<String, String>& string2) is invoked, which returns
2745         a StringAppend<String, StringAppend<String, String> > object.
2746         Then operator String() is invoked, which allocates a StringImpl object, once, large enough to hold the
2747         final string - it uses tryMakeString provided by StringConcatenate.h under the hoods, which guards us
2748         against too big string allocations, etc.
2749
2750         Note that the second template, defines a recursive way to concat an arbitary number of strings
2751         into a single String with just one allocation.
2752
2753         * GNUmakefile.list.am: Add StringOperators.h to build.
2754         * JavaScriptCore.exp: Export WTF::emptyString(). Remove no longer needed symbols.
2755         * JavaScriptCore.gypi: Add StringOperators.h to build.
2756         * JavaScriptCore.vcproj/WTF/WTF.vcproj: Ditto.
2757         * JavaScriptCore.xcodeproj/project.pbxproj: Ditto.
2758         * wtf/text/AtomicString.h: Pull in StringConcatenate.h at the end of the file.
2759         * wtf/text/StringConcatenate.h: Conditionally include AtomicString.h to avoid a cyclic dependency. Pull in StringOperators.h at the end of the file.
2760         * wtf/text/StringOperators.h: Added. This is never meant to be included directly, including either WTFString.h or AtomicString.h automatically pulls in this file.
2761         (WTF::StringAppend::StringAppend):
2762         (WTF::StringAppend::operator String):
2763         (WTF::StringAppend::operator AtomicString):
2764         (WTF::StringAppend::writeTo):
2765         (WTF::StringAppend::length):
2766         (WTF::operator+):
2767         * wtf/text/WTFString.cpp: Remove operator+ implementations that use String::append().
2768         (WTF::emptyString): Add new shared empty string free function.
2769         * wtf/text/WTFString.h: Replace operator+ implementations by StringAppend template solution. Pull in AtomicString.h at the end of the file.
2770
2771 2011-05-12  Philippe Normand  <pnormand@igalia.com>
2772
2773         Unreviewed, GTK build fix.
2774
2775         * wtf/Platform.h:
2776
2777 2011-05-12  Keith Kyzivat  <keith.kyzivat@nokia.com>
2778
2779         Reviewed by Csaba Osztrogonác.
2780
2781         [Qt] Arm debug build failing on ARMAssembler::debugOffset()
2782         https://bugs.webkit.org/show_bug.cgi?id=60688
2783
2784         Related to svn rev 85523
2785
2786         * assembler/ARMAssembler.h:
2787         (JSC::ARMAssembler::debugOffset):
2788
2789 2011-05-11  Igor Oliveira  <igor.oliveira@openbossa.org>
2790
2791         Reviewed by Eric Seidel.
2792
2793         WebKit does not build with GCCE
2794         https://bugs.webkit.org/show_bug.cgi?id=60667
2795
2796         Allow compile WebKit with GCCE
2797
2798         * wtf/Alignment.h:
2799         * wtf/Platform.h:
2800
2801 2011-05-11  Adam Barth  <abarth@webkit.org>
2802
2803         Reviewed by Eric Seidel.
2804
2805         Enable strict PassOwnPtr on Mac
2806         https://bugs.webkit.org/show_bug.cgi?id=60684
2807
2808         This should build cleanly now.
2809
2810         * wtf/PassOwnPtr.h:
2811
2812 2011-05-11  Oliver Hunt  <oliver@apple.com>
2813
2814         Reviewed by Darin Adler.
2815
2816         Protect JSC from WebCore executing JS during JS wrapper finalization
2817         https://bugs.webkit.org/show_bug.cgi?id=60672
2818         <rdar://problem/9350997>
2819
2820         Detect when we're trying to execute JS during GC and prevent the
2821         execution from happening.  We also assert that this isn't happening
2822         as it implies incorrect behaviour of an object's destructor.
2823
2824         * JavaScriptCore.exp:
2825         * heap/Heap.cpp:
2826         * heap/Heap.h:
2827         (JSC::Heap::isBusy):
2828         * interpreter/Interpreter.cpp:
2829         (JSC::Interpreter::execute):
2830         (JSC::Interpreter::executeCall):
2831         (JSC::Interpreter::executeConstruct):
2832         * runtime/JSGlobalData.h:
2833         (JSC::JSGlobalData::isCollectorBusy):
2834
2835 2011-05-11  Oliver Hunt  <oliver@apple.com>
2836
2837         Reviewed by Gavin Barraclough.
2838
2839         Enable gc mark validation in temporarily in release builds
2840         https://bugs.webkit.org/show_bug.cgi?id=60678
2841
2842         Make it easier to turn the gc mark validation on and off, and
2843         temporarily turn it on for all builds.
2844
2845         * heap/MarkStack.cpp:
2846         * heap/MarkStack.h:
2847         (JSC::MarkStack::append):
2848         (JSC::MarkStack::internalAppend):
2849         * runtime/WriteBarrier.h:
2850         (JSC::MarkStack::appendValues):
2851         * wtf/Platform.h:
2852
2853 2011-05-11  Geoffrey Garen  <ggaren@apple.com>
2854
2855         Reviewed by Oliver Hunt.
2856
2857         <rdar://problem/9331651> REGRESSION: RPRVT grows by 1MB / sec @ dvd2blu.com
2858
2859         SunSpider reports no change.
2860
2861         This bug was caused by changing Structure and Executable to being GC
2862         objects, and by a long-standing bug that would thrash the global object
2863         between dictionary and non-dictionary states.
2864
2865         * runtime/BatchedTransitionOptimizer.h:
2866         (JSC::BatchedTransitionOptimizer::BatchedTransitionOptimizer): Don't
2867         eagerly transition to dictionary -- this can cause pathological dictionary
2868         churn, and it's not necessary, since objects know how to automatically
2869         transition to dictionary when necessary.
2870
2871         * runtime/Executable.cpp:
2872         (JSC::EvalExecutable::compileInternal):
2873         (JSC::ProgramExecutable::compileInternal):
2874         (JSC::FunctionExecutable::compileForCallInternal):
2875         (JSC::FunctionExecutable::compileForConstructInternal): Be sure to report
2876         extra cost from compilation, because it can be quite high. This is especially
2877         important for program code, since DOM timers can repeatedly allocate
2878         program code without allocating any other objects.
2879
2880         * runtime/JSObject.cpp:
2881         (JSC::JSObject::removeDirect): Don't transition to the uncacheable state
2882         if the thing we're trying to remove doesn't exist. This can happen during
2883         compilation, since the compiler needs to ensure that no pre-existing
2884         conflicting definitions exist for certain declarations.
2885
2886 2011-05-11  Oliver Hunt  <oliver@apple.com>
2887
2888         Reviewed by Gavin Barraclough.
2889
2890         Make mark stack validation functions do something useful in a release build
2891         https://bugs.webkit.org/show_bug.cgi?id=60645
2892
2893         Turn ASSERTs into actual if(...) CRASH(); statements.
2894
2895         * heap/MarkStack.cpp:
2896         (JSC::MarkStack::validateValue):
2897
2898 2011-05-11  Xan Lopez  <xlopez@igalia.com>
2899
2900         Reviewed by Martin Robinson.
2901
2902         Fix copy&paste error in comment.
2903
2904         * jit/JITPropertyAccess.cpp:
2905         (JSC::JIT::stringGetByValStubGenerator): the value is stored in
2906         regT2, not regT1.
2907
2908 2011-05-11  Adam Roben  <aroben@apple.com>
2909
2910         WinCE build fixes for strict PassOwnPtr
2911
2912         * wtf/unicode/CollatorDefault.cpp:
2913         (WTF::Collator::userDefault): Use adoptPtr.
2914
2915 2011-05-11  Holger Hans Peter Freyther  <holger@moiji-mobile.com>
2916
2917         Unreviewed build fix.
2918
2919         [MIPS] Fix compilation of the MIPS JIT
2920
2921         Include the MIPSAssembler.h first to indirectly include
2922         AssemblerBuffer.h before the AbstractMacroAssembler.h. This
2923         order is used for the ARM and X86 MacroAssembler*.h
2924
2925         * assembler/MacroAssemblerMIPS.h:
2926
2927 2011-05-11  Adam Roben  <aroben@apple.com>
2928
2929         Turn on strict PassOwnPtr on Windows
2930
2931         Fixes <http://webkit.org/b/60632> Windows should build with strict PassOwnPtr enabled
2932
2933         Reviewed by Adam Barth.
2934
2935         * wtf/PassOwnPtr.h:
2936
2937 2011-05-10  Stephanie Lewis  <slewis@apple.com>
2938
2939         Unreviewed.
2940
2941         Revert accidental JavaScriptCore change in http://trac.webkit.org/changeset/86130
2942
2943         * Configurations/JavaScriptCore.xcconfig:
2944
2945 2011-05-10  Adam Barth  <abarth@webkit.org>
2946
2947         Reviewed by David Levin.
2948
2949         Enable strict PassOwnPtr on Chromium
2950         https://bugs.webkit.org/show_bug.cgi?id=60502
2951
2952         Other platforms to follow.
2953
2954         * wtf/PassOwnPtr.h:
2955
2956 2011-05-10  Geoffrey Garen  <ggaren@apple.com>
2957
2958         Reviewed by Darin Adler.
2959
2960         Fixed up some #include dependencies so the WriteBarrier class can actually call Heap::writeBarrier
2961         https://bugs.webkit.org/show_bug.cgi?id=60532
2962
2963         * GNUmakefile.list.am:
2964         * JavaScriptCore.gypi:
2965         * JavaScriptCore.xcodeproj/project.pbxproj: Build!
2966
2967         * heap/Handle.h: Moved HandleTypes to its own header because that's the
2968         WebKit style, and it was necessary to resolve a circular dependency
2969         between Handle.h and WriteBarrier.h.
2970
2971         * heap/Heap.h:
2972         (JSC::Heap::writeBarrier): Added an inline no-op writeBarrier(), to
2973         verify that all the code is in the right place.
2974
2975         * heap/MarkStack.h: Moved WriteBarrier operations to WriteBarrier.h to
2976         resolve a circular dependency.
2977
2978         * runtime/ArgList.h:
2979         * runtime/JSCell.h: #include WriteBarrier.h since we don't get it for
2980         free anymore.
2981
2982         * runtime/PropertyMapHashTable.h:
2983         (JSC::PropertyTable::PropertyTable): Call the real writeBarrier()
2984         function, now that it exists.
2985
2986         * runtime/SmallStrings.h: Removed a stray #include to resolve a circular
2987         dependency.
2988
2989         * runtime/WriteBarrier.h:
2990         (JSC::WriteBarrierBase::set):
2991         (JSC::MarkStack::append):
2992         (JSC::MarkStack::appendValues): Updated to match the changes above.
2993
2994 2011-05-10  Oliver Hunt  <oliver@apple.com>
2995
2996         Build fix.
2997
2998         * heap/MarkStack.cpp:
2999         (JSC::MarkStack::validateValue):
3000
3001 2011-05-10  Oliver Hunt  <oliver@apple.com>
3002
3003         Reviewed by Gavin Barraclough.
3004
3005         Add some aggressive GC validation to debug builds.
3006         https://bugs.webkit.org/show_bug.cgi?id=60601
3007
3008         When assertions are enabled we now do some validity checking
3009         of objects being added to the mark stack.
3010
3011         * bytecode/Instruction.h:
3012         (JSC::PolymorphicAccessStructureList::PolymorphicStubInfo::PolymorphicStubInfo):
3013         (JSC::PolymorphicAccessStructureList::visitAggregate):
3014         * heap/MarkStack.cpp:
3015         (JSC::MarkStack::validateSet):
3016         (JSC::MarkStack::validateValue):
3017         * heap/MarkStack.h:
3018         (JSC::MarkStack::appendValues):
3019         (JSC::MarkStack::append):
3020         (JSC::MarkStack::internalAppend):
3021
3022 2011-05-09  Darin Adler  <darin@apple.com>
3023
3024         Reviewed by Oliver Hunt.
3025
3026         http://bugs.webkit.org/show_bug.cgi?id=60509
3027         Wrong type used for return value from strlen
3028
3029         * wtf/FastMalloc.cpp:
3030         (WTF::fastStrDup): Use size_t. Also don't bother checking for failure since
3031         fastMalloc won't return if it fails.
3032
3033 2011-05-09  Adam Barth  <abarth@webkit.org>
3034
3035         Reviewed by Eric Seidel.
3036
3037         CSP should block Function constructor
3038         https://bugs.webkit.org/show_bug.cgi?id=60240
3039
3040         When eval is disabled, we need to block the use of the function
3041         constructor.  However, the WebCore JSC bindings call the function
3042         constructor directly to create inline event listeners.  To support that
3043         use, this patch adds an entrypoint that bypasses the check for whether
3044         eval is enabled.
3045
3046         * JavaScriptCore.exp:
3047         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
3048         * runtime/FunctionConstructor.cpp:
3049         (JSC::constructFunction):
3050         (JSC::constructFunctionSkippingEvalEnabledCheck):
3051         * runtime/FunctionConstructor.h:
3052
3053 2011-05-09  Adam Roben  <aroben@apple.com>
3054
3055         Automatically touch WebKit.idl whenever any other WebKit1 IDL file changes
3056
3057         Fixes <http://webkit.org/b/60468> WebKit.idl needs to be manually touched whenever any other
3058         WebKit1 IDL file changes to avoid build errors
3059
3060         Reviewed by Tim Hatcher.
3061
3062         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCoreGenerated.make:
3063         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCoreGenerated.vcproj:
3064         Updated for script rename.
3065
3066         * JavaScriptCore.vcproj/JavaScriptCore/react-to-vsprops-changes.py: Removed.
3067         * JavaScriptCore.vcproj/JavaScriptCore/work-around-vs-dependency-tracking-bugs.py: Renamed
3068         from react-to-vsprops-changes.py.
3069         (top level): Moved a constant here from main.
3070         (main): Moved most code from here to react_to_vsprops_changes. Added a call to the new
3071         react_to_webkit1_interface_changes function.
3072         (react_to_vsprops_changes): Moved code here from main. Updated to use the
3073         TOP_LEVEL_DIRECTORY global. Moved some code from here to mtime_of_newest_file_matching_globa
3074         and touch_if_older_than.
3075         (react_to_webkit1_interface_changes): Added. Touches WebKit.idl if any other WebKit1 IDL
3076         file has changed.
3077         (mtime_of_newest_file_matching_glob): Added. Code came from main.
3078         (touch_if_older_than): Added. Code came from main.
3079
3080 2011-05-08  Jessie Berlin  <jberlin@apple.com>
3081
3082         Reviewed by Dan Bernstein.
3083
3084         Make JSRetainPtr work with JSGlobalContextRefs.
3085         https://bugs.webkit.org/show_bug.cgi?id=60452
3086
3087         Add specialized functions for JSRetain and JSRelease when dealing with JSGlobalContextRefs.
3088
3089         * API/JSRetainPtr.h:
3090         (JSRetain):
3091         (JSRelease):
3092
3093 2011-05-07  Dawit Alemayehu  <adawit@kde.org>
3094
3095         Reviewed by Daniel Bates.
3096
3097         Fix compile with GCC 4.6.0
3098         https://bugs.webkit.org/show_bug.cgi?id=60380
3099
3100         Remove unused local variable from code.
3101
3102         * runtime/StringPrototype.cpp:
3103         (JSC::stringProtoFuncMatch):
3104
3105 2011-05-06  Alexis Menard  <alexis.menard@openbossa.org>
3106
3107         Unreviewed build fix with gcc 4.6.0 on linux and c++0x support.
3108
3109         std::tr1::has_trivial_constructor is in <tr1/memory>.
3110
3111         * wtf/TypeTraits.h:
3112
3113 2011-05-05  Jay Civelli  <jcivelli@chromium.org>
3114
3115         Reviewed by Adam Barth.
3116
3117         Added convenience methods to convert from a byte to hex ASCII digit
3118         characters and vice-versa.
3119         https://bugs.webkit.org/show_bug.cgi?id=59834
3120
3121         * wtf/ASCIICType.h:
3122         (WTF::toASCIIHexValue):
3123         (WTF::lowerNibbleToASCIIHexDigit):
3124         (WTF::upperNibbleToASCIIHexDigit):
3125
3126 2011-05-05  Alexis Menard  <alexis.menard@openbossa.org>
3127
3128         Reviewed by Benjamin Poulain.
3129
3130         [Qt] Make QtWebKit build when using gcc 4.6.0
3131         https://bugs.webkit.org/show_bug.cgi?id=60265
3132
3133         If QtWebKit is compiled with gcc 4.6.0 or later we don't want to deactivate
3134         the c++0x support because it works.
3135
3136         * JavaScriptCore.pro:
3137
3138 2011-05-04  Fridrich Strba  <fridrich.strba@bluewin.ch>
3139
3140         Reviewed by Geoffrey Garen.
3141
3142         Port MachineStackMarker.cpp to Windows x64
3143         https://bugs.webkit.org/show_bug.cgi?id=60216
3144
3145         * heap/MachineStackMarker.cpp:
3146         (JSC::getPlatformThreadRegisters): the CONTEXT struct is usable also
3147         on 64-bit Windows.
3148         (JSC::otherThreadStackPointer): return the Rsp register on Windows x64.
3149
3150 2011-05-04  Fridrich Strba  <fridrich.strba@bluewin.ch>
3151
3152         Reviewed by Martin Robinson.
3153
3154         Link libjavascriptcoregtk on Windows with winmm.dll
3155         https://bugs.webkit.org/show_bug.cgi?id=60215
3156
3157         * GNUmakefile.am:
3158
3159 2011-05-04  Tao Bai  <michaelbai@chromium.org>
3160
3161         Reviewed by David Kilzer.
3162
3163         Populate touch-icon url to FrameLoaderClient
3164         https://bugs.webkit.org/show_bug.cgi?id=59143
3165
3166         * Configurations/FeatureDefines.xcconfig:
3167
3168 2011-05-03  Geoffrey Garen  <ggaren@apple.com>
3169
3170         Reviewed by Darin Adler.
3171
3172         <rdar://problem/9366557> Various crashes due to bad DFG codegen at canalplus.fr
3173
3174         * dfg/DFGSpeculativeJIT.cpp:
3175         (JSC::DFG::SpeculativeJIT::checkArgumentTypes): Removed a stray line of
3176         code that accidentally survived the conversion to a switch statement,
3177         causing a lot of important code not to run most of the time.
3178
3179         Since this is not a trivial finger-picking mistake, I will not call it a
3180         typo.
3181
3182 2011-05-04  Adam Roben  <aroben@apple.com>
3183
3184         Another attempted build fix
3185
3186         * wtf/OwnPtr.h:
3187         (WTF::OwnPtr::operator==):
3188         (WTF::OwnPtr::operator!=):
3189         * wtf/PassOwnPtr.h:
3190         (WTF::PassOwnPtr::operator==):
3191         (WTF::PassOwnPtr::operator!=):
3192         Added a return statement. And made a tweak based on a suggestion from Anders Carlsson.
3193
3194 2011-05-04  Adam Roben  <aroben@apple.com>
3195
3196         Try to fix Leopard, Qt, and probably others
3197
3198         * wtf/OwnPtr.h:
3199         (WTF::OwnPtr::operator==):
3200         (WTF::OwnPtr::operator!=):
3201         * wtf/PassOwnPtr.h:
3202         (WTF::PassOwnPtr::operator==):
3203         (WTF::PassOwnPtr::operator!=):
3204         Try to get the compiler not to instantiate these function templates unnecessarily.
3205
3206 2011-05-03  Adam Roben  <aroben@apple.com>
3207
3208         Disallow equality comparisons between [Pass]OwnPtrs
3209
3210         If you have two OwnPtrs that are equal, you've already lost. (Unless you're doing something
3211         really sneaky, in which case you should stop!)
3212
3213         Fixes <http://webkit.org/b/60053> Testing OwnPtrs for equality should cause a compiler error
3214
3215         Reviewed by Anders Carlsson and Antti Koivisto.
3216
3217         * wtf/OwnPtr.h:
3218         (WTF::OwnPtr::operator==):
3219         (WTF::OwnPtr::operator!=):
3220         * wtf/PassOwnPtr.h:
3221         (WTF::PassOwnPtr::operator==):
3222         (WTF::PassOwnPtr::operator!=):
3223         Added private equality operators that fail to compile when used. (When not used, the
3224         compiler will skip over them because they are function templates.)
3225
3226 2011-05-04  Alexis Menard  <alexis.menard@openbossa.org>
3227
3228         Reviewed by Gavin Barraclough.
3229
3230         JITArithmetic.cpp produces a warning on a unused variable.
3231         https://bugs.webkit.org/show_bug.cgi?id=60060
3232
3233         Just properly use what we already have converted.
3234
3235         * jit/JITArithmetic.cpp:
3236         (JSC::JIT::emitSlow_op_add):
3237         (JSC::JIT::emitSlow_op_mul):
3238
3239 2011-05-04  Alexis Menard  <alexis.menard@openbossa.org>
3240
3241         Reviewed by Geoffrey Garen.
3242
3243         JITPropertyAccess produces a unused but set variable warning in gcc 4.6.0.
3244         https://bugs.webkit.org/show_bug.cgi?id=60050
3245
3246         This patch fix a compilation warning. The new warning scenario -Wunused-but-set-variable
3247         in gcc 4.6.0 is included in -Wall and therefore stops the compilation when warnings are treated
3248         as errors. The patch introduces a new macro ASSERT_JIT_OFFSET_UNUSED and ASSERT_WITH_MESSAGE_UNUSED
3249         which copy the idea of ASSERT_UNUSED.
3250
3251         * jit/JIT.h:
3252         * jit/JITPropertyAccess.cpp:
3253         (JSC::JIT::emit_op_method_check):
3254         (JSC::JIT::compileGetByIdHotPath):
3255         (JSC::JIT::emit_op_put_by_id):
3256         * wtf/Assertions.h:
3257         (assertWithMessageUnused):
3258
3259 2011-04-29  Jer Noble  <jer.noble@apple.com>
3260
3261         Reviewed by Eric Seidel.
3262
3263         Implement FULLSCREEN_API on Windows, Part 4: Enable it
3264         https://bugs.webkit.org/show_bug.cgi?id=59798
3265
3266         * wtf/Platform.h: Set ENABLE_FULLSCREEN_API on win.
3267
3268 2011-05-03  Alexis Menard  <alexis.menard@openbossa.org>
3269
3270         Reviewed by Eric Seidel.
3271
3272         Unused but set variable warning in MacroAssemberX86_64
3273         https://bugs.webkit.org/show_bug.cgi?id=59482
3274
3275         * assembler/MacroAssemblerX86_64.h:
3276         (JSC::MacroAssemblerX86_64::call):
3277         (JSC::MacroAssemblerX86_64::tailRecursiveCall):
3278         (JSC::MacroAssemblerX86_64::makeTailRecursiveCall):
3279
3280 2011-05-03  Oliver Hunt  <oliver@apple.com>
3281
3282         Reviewed by Geoffrey Garen.
3283
3284         Make malloc validation useful
3285         https://bugs.webkit.org/show_bug.cgi?id=57502
3286
3287         Reland this patch (rolled out in 82905) without
3288         turning it on by default.
3289
3290         * JavaScriptCore.exp:
3291         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
3292         * wtf/FastMalloc.cpp:
3293         (WTF::tryFastMalloc):
3294         (WTF::fastMalloc):
3295         (WTF::tryFastCalloc):
3296         (WTF::fastCalloc):
3297         (WTF::fastFree):
3298         (WTF::tryFastRealloc):
3299         (WTF::fastRealloc):
3300         (WTF::fastMallocSize):
3301         (WTF::TCMalloc_PageHeap::isScavengerSuspended):
3302         (WTF::TCMalloc_PageHeap::scheduleScavenger):
3303         (WTF::TCMalloc_PageHeap::suspendScavenger):
3304         (WTF::TCMalloc_PageHeap::signalScavenger):
3305         (WTF::TCMallocStats::malloc):
3306         (WTF::TCMallocStats::free):
3307         (WTF::TCMallocStats::fastCalloc):
3308         (WTF::TCMallocStats::tryFastCalloc):
3309         (WTF::TCMallocStats::calloc):
3310         (WTF::TCMallocStats::fastRealloc):
3311         (WTF::TCMallocStats::tryFastRealloc):
3312         (WTF::TCMallocStats::realloc):
3313         (WTF::TCMallocStats::fastMallocSize):
3314         * wtf/FastMalloc.h:
3315         (WTF::Internal::fastMallocValidationHeader):
3316         (WTF::Internal::fastMallocValidationSuffix):
3317         (WTF::Internal::fastMallocMatchValidationType):
3318         (WTF::Internal::setFastMallocMatchValidationType):
3319         (WTF::fastMallocMatchValidateFree):
3320         (WTF::fastMallocValidate):
3321
3322 2011-05-03  Xan Lopez  <xlopez@igalia.com>
3323
3324         Reviewed by Anders Carlsson.
3325
3326         Compile error with GCC 4.6.0, tries to assign unsigned& to bitfield
3327         https://bugs.webkit.org/show_bug.cgi?id=59261
3328
3329         Use unary '+' to force proper type detection in template arguments
3330         with GCC 4.6.0. See bug report for more details.
3331
3332         * runtime/Structure.cpp:
3333         (JSC::StructureTransitionTable::remove): Use '+' to force precise type detection.
3334         (JSC::StructureTransitionTable::add): ditto.
3335         * runtime/Structure.h:
3336         (JSC::StructureTransitionTable::keyForWeakGCMapFinalizer): ditto.
3337
3338 2011-05-03  Jessie Berlin  <jberlin@apple.com>
3339
3340         Rubber-stamped by Adam Roben.
3341
3342         Revert r85550 and r85575.
3343
3344         Variables cannot be exported via the .def file. Instead, they should be annotated with
3345         JS_EXPORTDATA.
3346
3347         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
3348         * runtime/Structure.cpp:
3349         (JSC::Structure::materializePropertyMap):
3350         * runtime/Structure.h:
3351         (JSC::Structure::typeInfo):
3352         (JSC::Structure::previousID):
3353         (JSC::Structure::propertyStorageCapacity):
3354         (JSC::Structure::propertyStorageSize):
3355         (JSC::Structure::get):
3356         (JSC::Structure::materializePropertyMapIfNecessary):
3357
3358 2011-05-02  Adam Roben  <aroben@apple.com>
3359
3360         Allow implicit conversion from nullptr_t to PassOwnPtr
3361
3362         This makes it a lot easier to write code that just wants a null PassOwnPtr, especially in
3363         strict PassOwnPtr mode.
3364
3365         Fixes <http://webkit.org/b/59964> Implicit conversion from std::nullptr_t to PassOwnPtr
3366         doesn't work, but should
3367
3368         Reviewed by Adam Barth.
3369
3370         * wtf/PassOwnPtr.h:
3371         (WTF::PassOwnPtr::PassOwnPtr): Added a non-explicit constructor that takes a nullptr_t.
3372
3373         * wtf/MessageQueue.h:
3374         (WTF::::waitForMessageFilteredWithTimeout):
3375         (WTF::::tryGetMessage):
3376         Use the new implicit conversion.
3377
3378 2011-05-02  Jessie Berlin  <jberlin@apple.com>
3379
3380         Rubber-stamped by Oliver Hunt.
3381
3382         Remove an assertion that Windows was hitting on launch.
3383
3384         * runtime/Structure.cpp:
3385         (JSC::Structure::materializePropertyMap):
3386         * runtime/Structure.h:
3387         (JSC::Structure::typeInfo):
3388         (JSC::Structure::previousID):
3389         (JSC::Structure::propertyStorageCapacity):
3390         (JSC::Structure::propertyStorageSize):
3391         (JSC::Structure::get):
3392         (JSC::Structure::materializePropertyMapIfNecessary):
3393
3394 2011-05-02  Mark Rowe  <mrowe@apple.com>
3395
3396         Reviewed by Geoff Garen.
3397
3398         <rdar://problem/9371948> JavaScriptCore should build with GCC 4.2
3399
3400         * Configurations/CompilerVersion.xcconfig:
3401
3402 2011-05-02  Gavin Barraclough  <barraclough@apple.com>
3403
3404         ARMv7 build fix.
3405
3406         * assembler/AbstractMacroAssembler.h:
3407         (JSC::AbstractMacroAssembler::Jump::link):
3408         (JSC::AbstractMacroAssembler::Jump::linkTo):
3409
3410 2011-05-02  Oliver Hunt  <oliver@apple.com>
3411
3412         Windows build fix.
3413
3414         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.def:
3415
3416 2011-05-02  Michael Saboff  <msaboff@apple.com>
3417
3418         Reviewed by Geoffrey Garen.
3419
3420         crash in JSC::RegExp::match
3421         https://bugs.webkit.org/show_bug.cgi?id=58922
3422
3423         Cleared chained backtrack data label when linking label even if that
3424         label doesn't chain itself.  This is needed so that subsequent
3425         backtrack data labels point to the next outer paren and not within
3426         the current paren.
3427
3428         * yarr/YarrJIT.cpp:
3429         (JSC::Yarr::YarrGenerator::TermGenerationState::linkDataLabelToBacktrackIfExists):
3430
3431 2011-05-02  Geoffrey Garen  <ggaren@apple.com>
3432
3433         Reviewed by Oliver Hunt.
3434
3435         Tiny bit of heap cleanup.
3436
3437         * heap/MarkedBlock.h:
3438         (JSC::MarkedBlock::contains): Tightened up an assertion and a comment.
3439
3440         * heap/MarkedSpace.h:
3441         (JSC::MarkedSpace::globalData):
3442         (JSC::MarkedSpace::highWaterMark):
3443         (JSC::MarkedSpace::setHighWaterMark): Moved inlines out of the class
3444         definition, for better clarity.
3445
3446 2011-05-02  Oliver Hunt  <oliver@apple.com>
3447
3448         Reviewed by Gavin Barraclough.
3449
3450         Correct marking of interpreter data in mixed mode builds
3451         https://bugs.webkit.org/show_bug.cgi?id=59962
3452
3453         We had a few places in mixed mode builds where we would not
3454         track data used by the interpreter for marking.  This patch
3455         corrects the problem and adds a number of assertions to catch
3456         live Structures being collected.
3457
3458         * JavaScriptCore.exp:
3459         * assembler/ARMv7Assembler.h:
3460         (JSC::ARMv7Assembler::ARMInstructionFormatter::debugOffset):
3461         * bytecode/CodeBlock.cpp:
3462         (JSC::CodeBlock::dump):
3463         * bytecode/CodeBlock.h:
3464         (JSC::CodeBlock::addPropertyAccessInstruction):
3465         (JSC::CodeBlock::addGlobalResolveInstruction):
3466         (JSC::CodeBlock::addStructureStubInfo):
3467         (JSC::CodeBlock::addGlobalResolveInfo):
3468         * bytecompiler/BytecodeGenerator.cpp:
3469         (JSC::BytecodeGenerator::emitResolve):
3470         (JSC::BytecodeGenerator::emitResolveWithBase):
3471         (JSC::BytecodeGenerator::emitGetById):
3472         (JSC::BytecodeGenerator::emitPutById):
3473         (JSC::BytecodeGenerator::emitDirectPutById):
3474         * runtime/Structure.cpp:
3475         (JSC::Structure::materializePropertyMap):
3476         * runtime/Structure.h:
3477         (JSC::Structure::typeInfo):
3478         (JSC::Structure::previousID):
3479         (JSC::Structure::propertyStorageCapacity):
3480         (JSC::Structure::propertyStorageSize):
3481         (JSC::Structure::get):
3482         (JSC::Structure::materializePropertyMapIfNecessary):
3483
3484 2011-05-02  Xan Lopez  <xlopez@igalia.com>
3485
3486         Reviewed by Alexey Proskuryakov.