JavaScriptCore/ChangeLog

   1 2008-05-21  Oliver Hunt  <oliver@apple.com>
   2
   3         Reviewed by NOBODY (Build fix).
   4
   5         Add DebuggerCallFrame.{h,cpp} to the project file
   6
   7         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
   8
   9 2008-05-21  Alp Toker  <alp@nuanti.com>
  10
  11         GTK+ port build fixes following squirrelfish merge r33979.
  12
  13         * GNUmakefile.am:
  14
  15 2008-05-21  Maciej Stachowiak  <mjs@apple.com>
  16
  17         Reviewed by Darin.
  18
  19         - save a hash lookup wne writing to global properties
  20         0.3% speedup on SunSpider, 7% on bitops-bitwise-and
  21
  22         * VM/Machine.cpp:
  23         (KJS::resolveBase): Check for being a the end of the scope chain
  24         before hash lookup.
  25
  26 2008-05-21  Alp Toker  <alp@nuanti.com>
  27
  28         Rubber-stamped by Maciej.
  29
  30         Replace non-standard #pragma marks with comments to avoid compiler
  31         warnings.
  32
  33         * profiler/ProfileNode.cpp:
  34
  35 2008-05-21  Geoffrey Garen  <ggaren@apple.com>
  36
  37         Reviewed by Mark Rowe.
  38
  39         Fix layout test failure in fast/dom/getter-on-window-object2 introduced in r33961.
  40
  41         * JavaScriptCore.exp:
  42         * kjs/JSGlobalObject.cpp:
  43         (KJS::JSGlobalObject::defineGetter):
  44         (KJS::JSGlobalObject::defineSetter):
  45         * kjs/JSGlobalObject.h:
  46
  47 === End merge of squirrelfish ===
  48
  49 2008-05-21  Geoffrey Garen  <ggaren@apple.com>
  50
  51         Reviewed by Tim Hatcher.
  52
  53         Merged with trunk WebCore's new debugger.
  54
  55         * kjs/DebuggerCallFrame.cpp:
  56         (KJS::DebuggerCallFrame::evaluate): Changed this function to separate
  57         the exception value from the return value. The WebKit debugger treats
  58         them as one, but the WebCore debugger doesn't.
  59
  60         * kjs/DebuggerCallFrame.h:
  61         (KJS::DebuggerCallFrame::dynamicGlobalObject): Added a new accessor for
  62         the dynamic global object, since the debugger doesn't want the lexical
  63         global object.
  64
  65 2008-05-21  Oliver Hunt  <oliver@apple.com>
  66
  67         Reviewed by Maciej.
  68
  69         Bug 19116: SquirrelFish shouldn't regress on variable lookups
  70         <https://bugs.webkit.org/show_bug.cgi?id=19116>
  71
  72         Optimise cross scope assignment, 0.4% progression in sunspider.
  73
  74         * VM/CodeBlock.cpp:
  75         (KJS::CodeBlock::dump):
  76         * VM/CodeGenerator.cpp:
  77         (KJS::CodeGenerator::emitPutScopedVar):
  78         * VM/CodeGenerator.h:
  79         * VM/Machine.cpp:
  80         (KJS::Machine::privateExecute):
  81         * VM/Opcode.h:
  82         * kjs/nodes.cpp:
  83         (KJS::AssignResolveNode::emitCode):
  84
  85 2008-05-21  Maciej Stachowiak  <mjs@apple.com>
  86
  87         Reviewed by Oliver.
  88
  89         - check property map before symbol table in JSGlobalObject::getOwnPropertySlot
  90         0.5% speedup on SunSpider
  91
  92         * kjs/JSGlobalObject.h:
  93         (KJS::JSGlobalObject::getOwnPropertySlot): Check property map before symbol table
  94         because symbol table access is likely to have been optimized.
  95
  96 2008-05-21  Oliver Hunt  <oliver@apple.com>
  97
  98         Reviewed by Maciej.
  99
 100         Bug 19116: SquirrelFish shouldn't regress on variable lookups
 101         <https://bugs.webkit.org/show_bug.cgi?id=19116>
 102
 103         Optimise multiscope lookup of statically resolvable function calls.
 104         SunSpider reports a 1.5% improvement, including 37% on
 105         controlflow-recursive for some reason :D
 106
 107         * VM/CodeBlock.cpp:
 108         (KJS::CodeBlock::dump):
 109         * VM/CodeGenerator.cpp:
 110         (KJS::CodeGenerator::emitResolve):
 111         * VM/CodeGenerator.h:
 112         * kjs/nodes.cpp:
 113         (KJS::FunctionCallResolveNode::emitCode):
 114
 115 2008-05-21  Maciej Stachowiak  <mjs@apple.com>
 116
 117         Reviewed by Oliver.
 118
 119         - give JSGlobalObject a special version of getOwnPropertySlot that tells you if the slot is directly writable
 120         (WebCore change using this is a 2.6% speedup on in-browser SunSpider).
 121
 122         * JavaScriptCore.exp:
 123         * kjs/JSGlobalObject.h:
 124         (KJS::JSGlobalObject::getOwnPropertySlot):
 125         * kjs/JSVariableObject.h:
 126         (KJS::JSVariableObject::symbolTableGet):
 127         * kjs/object.h:
 128         (KJS::JSObject::getDirectLocation):
 129         (KJS::JSObject::getOwnPropertySlotForWrite):
 130         * kjs/property_map.cpp:
 131         (KJS::PropertyMap::getLocation):
 132         * kjs/property_map.h:
 133         * kjs/property_slot.h:
 134         (KJS::PropertySlot::putValue):
 135
 136 2008-05-20  Oliver Hunt  <oliver@apple.com>
 137
 138         Reviewed by Maciej.
 139
 140         Bug 19116: SquirrelFish shouldn't regress on variable lookups
 141         <https://bugs.webkit.org/show_bug.cgi?id=19116>
 142
 143         This restores multiscope optimisation to simple resolve, producing
 144         a 2.6% progression in SunSpider.  Have verified that none of the
 145         sites broken by the multiscope optimisation in trunk were effected
 146         by this change.
 147
 148         * VM/CodeBlock.cpp:
 149         (KJS::CodeBlock::dump):
 150         * VM/CodeBlock.h:
 151         (KJS::CodeBlock::CodeBlock):
 152         * VM/CodeGenerator.cpp:
 153         (KJS::CodeGenerator::findScopedProperty):
 154         (KJS::CodeGenerator::emitResolve):
 155         * VM/CodeGenerator.h:
 156         * VM/Machine.cpp:
 157         (KJS::resolve_n):
 158         (KJS::Machine::privateExecute):
 159         * VM/Opcode.h:
 160         * kjs/JSVariableObject.h:
 161
 162 2008-05-20  Oliver Hunt  <oliver@apple.com>
 163
 164         Reviewed by NOBODY (Build fix).
 165
 166         Fixerate the windows build.
 167
 168         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
 169         * VM/CodeGenerator.cpp:
 170         * VM/RegisterFile.h:
 171         * kjs/JSGlobalObject.h:
 172         * kjs/Parser.cpp:
 173         * kjs/interpreter.h:
 174
 175 2008-05-20  Oliver Hunt  <oliver@apple.com>
 176
 177         Reviewed by Geoff.
 178
 179         Bug 19110: SquirrelFish: Google Maps - no maps
 180         <https://bugs.webkit.org/show_bug.cgi?id=19110>
 181
 182         Correct a comedy of errors present in my original patch to "fix"
 183         exceptions occurring midway through pre and post increment. This
 184         solution is cleaner than the original, doesn't need the additional
 185         opcodes, and as an added benefit does not break Google Maps.
 186
 187         Sunspider reports a 0.4% progression.
 188
 189         * VM/CodeBlock.cpp:
 190         (KJS::CodeBlock::dump):
 191         * VM/CodeGenerator.cpp:
 192         * VM/CodeGenerator.h:
 193         * VM/Machine.cpp:
 194         (KJS::Machine::privateExecute):
 195         * VM/Opcode.h:
 196         * kjs/nodes.cpp:
 197         (KJS::PreIncResolveNode::emitCode):
 198         (KJS::PreDecResolveNode::emitCode):
 199         (KJS::PreIncBracketNode::emitCode):
 200         (KJS::PreDecBracketNode::emitCode):
 201         (KJS::PreIncDotNode::emitCode):
 202         (KJS::PreDecDotNode::emitCode):
 203
 204 2008-05-20  Maciej Stachowiak  <mjs@apple.com>
 205
 206         Reviewed by Oliver.
 207
 208         - inline JSGlobalObject::getOwnPropertySlot
 209         1% improvement on in-browser SunSpider (a wash command-line)
 210
 211         * kjs/JSGlobalObject.cpp:
 212         * kjs/JSGlobalObject.h:
 213         (KJS::JSGlobalObject::getOwnPropertySlot):
 214
 215 2008-05-18  Oliver Hunt  <oliver@apple.com>
 216
 217         Reviewed by Maciej.
 218
 219         Bug 18752: SQUIRRELFISH: exceptions are not always handled by the vm
 220         <https://bugs.webkit.org/show_bug.cgi?id=18752>
 221
 222         Handle exceptions thrown by toString conversion in subscript operators,
 223         this should basically complete exception handling in SquirrelFish.
 224
 225         Sunspider reports no regression.
 226
 227         * VM/Machine.cpp:
 228         (KJS::Machine::privateExecute):
 229
 230 2008-05-17  Geoffrey Garen  <ggaren@apple.com>
 231
 232         Reviewed by Oliver Hunt.
 233
 234         [Reapplying patch with previously missing files from r33553 -- Oliver]
 235
 236         Behold: debugging.
 237
 238         SunSpider reports no change.
 239
 240         * JavaScriptCore.xcodeproj/project.pbxproj: Added DebuggerCallFrame.h/.cpp,
 241         and created a debugger folder.
 242
 243         * VM/CodeGenerator.cpp:
 244         (KJS::CodeGenerator::generate): If the debugger is attached, always
 245         generate full scope chains for its sake.
 246
 247         * VM/Machine.cpp:
 248         (KJS::Machine::unwindCallFrame): Notify the debugger when unwinding
 249         due to an exception, so it doesn't keep stale call frames around.
 250
 251         (KJS::Machine::execute): Set Callee to 0 in eval frames, so the
 252         debugger can distinguish them from function call frames.
 253
 254         (KJS::Machine::debug): Simplified this function, since the debugger
 255         doesn't actually need all the information we used to provide.
 256
 257         (KJS::Machine::privateExecute): Treat debugging hooks like other function
 258         calls, so the code we hook into (the debugger UI) can be optimized.
 259
 260         * kjs/debugger.cpp: Nixed these default callback implementations and
 261         made the callbacks pure virtual instead, so the compiler could tell me
 262         if I made a mistake in one of the subclasses.
 263
 264         * kjs/debugger.h: Removed a bunch of irrelevent data from the debugger
 265         callbacks. Changed from passing an ExecState* to passing a
 266         DebuggerCallFrame*, since an ExecState* doesn't contain sufficient
 267         information anymore.
 268
 269         * kjs/function.cpp:
 270         (KJS::globalFuncEval): Easiest bug fix evar!
 271
 272         [Previously missing files from r33553]
 273         * kjs/DebuggerCallFrame.cpp: Copied from JavaScriptCore/profiler/FunctionCallProfile.h.
 274         (KJS::DebuggerCallFrame::functionName):
 275         (KJS::DebuggerCallFrame::thisObject):
 276         (KJS::DebuggerCallFrame::evaluateScript):
 277         * kjs/DebuggerCallFrame.h: Copied from JavaScriptCore/VM/Register.h.
 278         (KJS::DebuggerCallFrame::DebuggerCallFrame):
 279         (KJS::DebuggerCallFrame::scopeChain):
 280         (KJS::DebuggerCallFrame::exception):
 281
 282 2008-05-17  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
 283
 284         Reviewed by Oliver.
 285
 286         Bug 18991: SquirrelFish: Major codegen issue in a.b=expr, a[b]=expr
 287         <https://bugs.webkit.org/show_bug.cgi?id=18991>
 288
 289         Fix the last remaining blocking cases of this bug.
 290
 291         * kjs/grammar.y:
 292         * kjs/nodes.cpp:
 293         (KJS::ReadModifyResolveNode::emitCode):
 294
 295 2008-05-17  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
 296
 297         Reviewed by Oliver.
 298
 299         Partial fix for:
 300
 301         Bug 18991: SquirrelFish: Major codegen issue in a.b=expr, a[b]=expr
 302         <https://bugs.webkit.org/show_bug.cgi?id=18991>
 303
 304         Ensure that the code generated for assignments uses temporaries whenever
 305         necessary. This patch covers the vast majority of situations, but there
 306         are still a few left.
 307
 308         This patch also adds some missing cases to CodeBlock::dump().
 309
 310         * VM/CodeBlock.cpp:
 311         (KJS::CodeBlock::dump):
 312         * VM/CodeGenerator.h:
 313         (KJS::CodeGenerator::destinationForAssignResult):
 314         (KJS::CodeGenerator::leftHandSideNeedsCopy):
 315         (KJS::CodeGenerator::emitNodeForLeftHandSide):
 316         * kjs/NodeInfo.h:
 317         * kjs/grammar.y:
 318         * kjs/nodes.cpp:
 319         (KJS::AssignDotNode::emitCode):
 320         (KJS::ReadModifyDotNode::emitCode):
 321         (KJS::AssignBracketNode::emitCode):
 322         (KJS::ReadModifyBracketNode::emitCode):
 323         (KJS::ForInNode::ForInNode):
 324         * kjs/nodes.h:
 325         (KJS::ReadModifyResolveNode::):
 326         (KJS::AssignResolveNode::):
 327         (KJS::ReadModifyBracketNode::):
 328         (KJS::AssignBracketNode::):
 329         (KJS::AssignDotNode::):
 330         (KJS::ReadModifyDotNode::):
 331
 332 2008-05-17  Oliver Hunt  <oliver@apple.com>
 333
 334         Reviewed by Maciej.
 335
 336         Bug 19106: SquirrelFish: Activation is not marked correctly
 337         <https://bugs.webkit.org/show_bug.cgi?id=19106>
 338
 339         We can't rely on the symbol table for a count of the number of globals
 340         we need to mark as that misses duplicate parameters and 'this'.  Now we
 341         use the actual local register count from the codeBlock.
 342
 343         * kjs/JSActivation.cpp:
 344         (KJS::JSActivation::mark):
 345
 346 2008-05-16  Oliver Hunt  <oliver@apple.com>
 347
 348         Reviewed by Geoff.
 349
 350         Bug 19076: SquirrelFish: RegisterFile can be corrupted if implictly reenter global scope with no declared vars
 351         <https://bugs.webkit.org/show_bug.cgi?id=19076>
 352
 353         Don't delay allocation of initial global RegisterFile, as we can't guarantee we will be able
 354         to allocate the global 'this' register safely at any point after initialisation of the Global
 355         Object.
 356
 357         Unfortunately this initial allocation caused a regression of 0.2-0.3%, however this patch adds
 358         support for the static slot optimisation for the global Math object which brings it to a 0.3%
 359         progression.
 360
 361         * VM/CodeGenerator.cpp:
 362         (KJS::CodeGenerator::programCodeThis):
 363         (KJS::CodeGenerator::CodeGenerator):
 364         (KJS::CodeGenerator::addParameter):
 365         * VM/CodeGenerator.h:
 366         * VM/Machine.cpp:
 367         (KJS::Machine::execute):
 368         * kjs/ExecState.h:
 369         * kjs/JSGlobalObject.cpp:
 370         (KJS::JSGlobalObject::reset):
 371         * kjs/JSGlobalObject.h:
 372         (KJS::JSGlobalObject::GlobalPropertyInfo::GlobalPropertyInfo):
 373         (KJS::JSGlobalObject::addStaticGlobals):
 374         * kjs/nodes.cpp:
 375
 376 2008-05-16  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
 377
 378         Reviewed by Oliver Hunt.
 379
 380         Bug 19098: SquirrelFish: Ref'd temporaries can be clobbered
 381         <https://bugs.webkit.org/show_bug.cgi?id=19098>
 382
 383         When doing code generation for a statement list, increase the reference
 384         count on a register that might eventually be returned, so that it doesn't
 385         get clobbered by a request for a new temporary.
 386
 387         * kjs/nodes.cpp:
 388         (KJS::statementListEmitCode):
 389
 390 2008-05-16  Maciej Stachowiak  <mjs@apple.com>
 391
 392         Reviewed by Oliver.
 393
 394         - fixed Bug 19044: SquirrelFish: Bogus values enter evaluation when closing over scope with parameter and var with same name
 395         https://bugs.webkit.org/show_bug.cgi?id=19044
 396
 397         * kjs/JSActivation.cpp:
 398         (KJS::JSActivation::copyRegisters): Use numLocals from the code
 399         block rather than the size of the symbol table for the number of
 400         registers to copy, to account for duplicate parameters and vars
 401         with the same name as parameters (we still have potentially
 402         suboptimal codegen in that we allocate a local register for the
 403         var in the latter case but it is never used).
 404
 405 2008-05-15  Geoffrey Garen  <ggaren@apple.com>
 406
 407         Not reviewed.
 408
 409         We regret to inform you that your program is crashing because you were
 410         stupid.
 411
 412         * VM/Machine.cpp:
 413         (KJS::Machine::privateExecute): Math is hard.
 414
 415 2008-05-14  Geoffrey Garen  <ggaren@apple.com>
 416
 417         Reviewed by Oliver Hunt.
 418
 419         A little more debugger action: filled in op_debug. All debugger control
 420         flow works now, but variable inspection and backtraces still don't.
 421
 422         SunSpider reports no change.
 423
 424         * VM/CodeGenerator.cpp: Changed op_debug to accept line number parameters.
 425
 426         * VM/Machine.cpp:
 427         (KJS::Machine::getFunctionAndArguments): Moved op_debug into a
 428         NEVER_INLINE function to avoid a stunning 10% performance regression.
 429         Also factored out a common function for retrieving the function and
 430         arguments from a  call frame.
 431
 432         * kjs/JSActivation.cpp:
 433         (KJS::JSActivation::createArgumentsObject): Use the new factored out
 434         function mentioned above.
 435
 436         * kjs/Parser.cpp:
 437         (KJS::Parser::parse): Increment m_sourceId before assigning it, so the
 438         sourceId we send to the debugger matches the sourceId recorded in the
 439         node.
 440
 441         * kjs/nodes.cpp: Emit debugging hooks.
 442
 443 2008-05-14  Oliver Hunt  <oliver@apple.com>
 444
 445         Reviewed by Maciej.
 446
 447         Bug 19024: SQUIRRELFISH: ASSERTION FAILED: activation->isActivationObject() in Machine::unwindCallFrame
 448         <https://bugs.webkit.org/show_bug.cgi?id=19024>
 449
 450         This fixes a number of issues.  The most important is that we now check every register
 451         file for tainting rather than just looking for function register files as that was
 452         insufficient. Additionally guarded against implicit re-entry into Eval code.
 453
 454         Also added a few additional assertions to reduce the amout of time between something
 455         going wrong and us seeing the error.
 456
 457         * VM/Machine.cpp:
 458         (KJS::Machine::execute):
 459         (KJS::Machine::privateExecute):
 460         * VM/RegisterFile.cpp:
 461         (KJS::RegisterFile::growBuffer):
 462         (KJS::RegisterFile::addGlobalSlots):
 463         * VM/RegisterFileStack.cpp:
 464         (KJS::RegisterFileStack::pushGlobalRegisterFile):
 465         (KJS::RegisterFileStack::pushFunctionRegisterFile):
 466         * VM/RegisterFileStack.h:
 467         (KJS::RegisterFileStack::inImplicitCall):
 468
 469 2008-05-14  Geoffrey Garen  <ggaren@apple.com>
 470
 471         Reviewed by Oliver Hunt.
 472
 473         A little more debugger action: emit opcodes for debugger hooks. Right
 474         now, the opcode implementation is just a stub.
 475
 476         SunSpider reports no change.
 477
 478         Some example codegen for "function f() { 1; }":
 479
 480             [   0] dbg         DidEnterCallFrame
 481             [   2] dbg         WillExecuteStatement
 482             [   4] load        tr0, 1(@k0)
 483             [   7] load        tr0, undefined(@k1)
 484             [  10] dbg         WillLeaveCallFrame
 485             [  12] ret         tr0
 486
 487 2008-05-14  Oliver Hunt  <oliver@apple.com>
 488
 489         Reviewed by Geoff.
 490
 491         Bug 19025: SQUIRRELFISH: malformed syntax in onload handler causes crash
 492         <https://bugs.webkit.org/show_bug.cgi?id=19025>
 493
 494         Simple fix -- move the use of functionBodyNode to after the null check.
 495
 496         * kjs/function_object.cpp:
 497         (KJS::FunctionObjectImp::construct):
 498
 499 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 500
 501         Reviewed by Oliver Hunt.
 502
 503         Fixed a codegen crash with run-time parse errors.
 504
 505         SunSpider reports no change.
 506
 507         emitThrowError needs to return the temporary holding the error, not dst,
 508         since dst may be NULL. In fact, emitThrowError shouldn't take a dst
 509         parameter at all, since exceptions should not modify the destination
 510         register.
 511
 512 2008-05-13  Oliver Hunt  <oliver@apple.com>
 513
 514         Reviewed by Geoff.
 515
 516         Bug 19027: SquirrelFish: Incorrect codegen for pre-increment
 517         <https://bugs.webkit.org/show_bug.cgi?id=19027>
 518
 519         This fixes the codegen issues for the pre-inc/decrement operators
 520         to prevent incorrectly clobbering the destination in the event of
 521         an exception.
 522
 523         * VM/CodeBlock.cpp:
 524         (KJS::CodeBlock::dump):
 525         * VM/CodeGenerator.cpp:
 526         (KJS::CodeGenerator::emitPreInc):
 527         (KJS::CodeGenerator::emitPreDec):
 528         * VM/CodeGenerator.h:
 529         * VM/Machine.cpp:
 530         (KJS::Machine::privateExecute):
 531         * VM/Opcode.h:
 532         * kjs/nodes.cpp:
 533         (KJS::PreIncResolveNode::emitCode):
 534         (KJS::PreDecResolveNode::emitCode):
 535         (KJS::PreIncBracketNode::emitCode):
 536         (KJS::PreDecBracketNode::emitCode):
 537         (KJS::PreIncDotNode::emitCode):
 538         (KJS::PreDecDotNode::emitCode):
 539
 540 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 541
 542         Reviewed by Oliver Hunt.
 543
 544         A little more debugger action: supply a real line number, sourceId,
 545         and sourceURL in op_new_error.
 546
 547         SunSpider reports a .2% speedup. Not sure what that's about.
 548
 549         * VM/Machine.cpp:
 550         (KJS::Machine::privateExecute): Use the new good stuff in op_new_error.
 551
 552         * kjs/nodes.cpp:
 553         (KJS::RegExpNode::emitCode): Use the shared emitThrowError instead of
 554         rolling our own.
 555
 556 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 557
 558         Reviewed by Oliver Hunt.
 559
 560         A little more debugger action: implemented the exception callback.
 561
 562         SunSpider reports a .2% speedup. Not sure what that's about.
 563
 564         * VM/CodeBlock.h: A little refactoring here. Store a pointer to our
 565         owner ScopeNode so we can retrieve data from it. This allows us to
 566         stop storing copies of the data ourselves. Also, store a "this" register
 567         instead of a code type, since we were only using the code type to
 568         calculate the "this" register.
 569
 570         * VM/CodeGenerator.cpp:
 571         (KJS::CodeGenerator::generate): Calculate the "this" register mentioned
 572         above. Also, take care of removing "this" from the symbol table after
 573         codegen is done, since relying on the timing of a destructor for correct
 574         behavior is not so good.
 575
 576         * VM/Machine.cpp:
 577         (KJS::Machine::throwException): Invoke the debugger's exception callback.
 578         (KJS::Machine::privateExecute): Use the "this" register mentioned above.
 579
 580 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 581
 582         Reviewed by Oliver Hunt.
 583
 584         Removed some unused exception machinery.
 585
 586         SunSpider reports a .3% speedup.
 587
 588         * API/JSCallbackObject.h:
 589         * API/JSCallbackObjectFunctions.h:
 590         * JavaScriptCore.exp:
 591         * VM/Machine.cpp:
 592         (KJS::Machine::privateExecute):
 593         * kjs/internal.cpp:
 594         * kjs/object.cpp:
 595         * kjs/object.h:
 596         * kjs/value.h:
 597
 598 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 599
 600         Reviewed by Oliver Hunt.
 601
 602         A little more debugger action.
 603
 604         * kjs/debugger.cpp:
 605         * kjs/debugger.h: Removed debuggersPresent because it was unused.
 606         Replaced AttachedGlobalObject linked list with a HashSet because HashSet
 607         is faster and simpler. Changed all functions to return void instead of
 608         bool, because no clients ever return false, and we don't want to support
 609         it.
 610
 611         * kjs/nodes.cpp: Did some up-keep to avoid build bustage.
 612         (KJS::Node::handleException):
 613         (KJS::BreakpointCheckStatement::execute):
 614         (KJS::FunctionBodyNodeWithDebuggerHooks::execute):
 615
 616 2008-05-13  Oliver Hunt  <oliver@apple.com>
 617
 618         Reviewed by Darin.
 619
 620         Bug 18752: SQUIRRELFISH: exceptions are not always handled by the vm
 621         <https://bugs.webkit.org/show_bug.cgi?id=18752>
 622
 623         Replace old attempt at "branchless" exceptions as the extra information
 624         being passed made gcc an unhappy compiler, replacing these custom toNumber
 625         calls with ordinary toNumber logic (by relying on toNumber now preventing
 626         side effects after an exception has been thrown) provided sufficient leeway
 627         to add the additional checks for the remaining unchecked cases.
 628
 629         This leaves only toString conversions in certain contexts as possibly
 630         misbehaving.
 631
 632         * VM/Machine.cpp:
 633         (KJS::jsAdd):
 634         (KJS::resolve):
 635         (KJS::resolveBaseAndProperty):
 636         (KJS::resolveBaseAndFunc):
 637         (KJS::Machine::privateExecute):
 638         * VM/Opcode.h:
 639         * kjs/value.h:
 640         (KJS::JSValue::safeGetNumber):
 641
 642 2008-05-13  Geoffrey Garen  <ggaren@apple.com>
 643
 644         Reviewed by Oliver Hunt.
 645
 646         First steps toward supporting the debugger API: support the sourceParsed
 647         callback; plus some minor fixups.
 648
 649         SunSpider reports no regression.
 650
 651         * VM/CodeGenerator.h: Removed a misleading comment.
 652
 653         * kjs/Parser.h: Changed the parser to take an ExecState*, so it can
 654         implement the sourceParsed callback -- that way, we only have to
 655         implement the callback in one place.
 656
 657         * kjs/debugger.cpp: Nixed DebuggerImp, because its sole purpose in life
 658         was to demonstrate the misapplication of design patterns.
 659
 660         * kjs/debugger.h: Changed sourceParsed to take a SourceProvider, to
 661         reduce copying, and not to return a value, because pausing execution
 662         after parsing is complicated, and no clients needed that ability, anyway.
 663
 664         * kjs/grammar.y: Make sure never to pass a NULL SourceElements* to
 665         didFinishParsing -- that simplifies some code down the road.
 666
 667         * kjs/nodes.cpp: Don't generate special AST nodes just because the
 668         debugger is attached -- that's a relic of the old AST execution model,
 669         and those nodes haven't been maintained.
 670
 671 2008-05-13  Oliver Hunt  <oliver@apple.com>
 672
 673         Reviewed by Geoff.
 674
 675         Bug 18752: SQUIRRELFISH: exceptions are not always handled by the vm
 676         <https://bugs.webkit.org/show_bug.cgi?id=18752>
 677
 678         First step: prevent incorrect evaluation of valueOf/toString conversion
 679         in right hand side of expression after earlier conversion throws.
 680
 681         * API/JSCallbackObjectFunctions.h:
 682         (KJS::::toNumber):
 683         * kjs/object.cpp:
 684         (KJS::JSObject::defaultValue):
 685
 686 2008-05-12  Oliver Hunt  <oliver@apple.com>
 687
 688         Reviewed by Geoff.
 689
 690         Bug 18934: SQUIRRELFISH: ASSERT @ nytimes.com due to RegisterFile being clobbered
 691         <https://bugs.webkit.org/show_bug.cgi?id=18934>
 692
 693         Unfortunately we cannot create new statically optimised globals if there are any
 694         tainted RegisterFiles on the RegisterFileStack.  To handle this we re-introduce
 695         (in a slightly cleaner form) the inImplicitCall concept to the RegisterFileStack.
 696
 697         * VM/Machine.cpp:
 698         (KJS::Machine::execute):
 699         * VM/RegisterFileStack.cpp:
 700         (KJS::RegisterFileStack::pushFunctionRegisterFile):
 701         * VM/RegisterFileStack.h:
 702
 703 2008-05-12  Geoffrey Garen  <ggaren@apple.com>
 704
 705         Reviewed by Maciej Stachowiak.
 706
 707         Introduced support for function.caller.
 708
 709         Improved support for walking interesting scopes for function introspection.
 710
 711         This fixes all remaining layout tests not blocked by rebasing to trunk.
 712
 713         SunSpider reports no change.
 714
 715         * VM/Machine.cpp:
 716         (KJS::Machine::dumpRegisters): Fixed a spacing issue.
 717
 718 2008-05-11  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
 719
 720         Reviewed by Oliver.
 721
 722         Bug 18961: SQUIRRELFISH: Gmail doesn't load
 723         <https://bugs.webkit.org/show_bug.cgi?id=18961>
 724
 725         Fix codegen for logical nodes so that they don't use their destination
 726         as a temporary.
 727
 728         * kjs/nodes.cpp:
 729         (KJS::LogicalAndNode::emitCode):
 730         (KJS::LogicalOrNode::emitCode):
 731
 732 2008-05-10  Maciej Stachowiak  <mjs@apple.com>
 733
 734         Reviewed by Oliver.
 735
 736         - JavaScriptCore part of fix for: "SQUIRRELFISH: function toString broken after calling"
 737         https://bugs.webkit.org/show_bug.cgi?id=18869
 738
 739         Three layout tests are fixed:
 740           fast/js/toString-elision-trailing-comma.html
 741           fast/js/toString-prefix-postfix-preserve-parens.html
 742           fast/js/kde/lval-exceptions.html
 743
 744         Functions now save a shared subrange of the original source used
 745         to make them (so in the common case this adds no storage above the
 746         memory cache).
 747
 748         * kjs/SourceProvider.h: Added.
 749         (KJS::SourceProvider): New abstract base class for classes that provide on-demand access
 750         to the source for a JavaScript program. This allows function objects to have access to their
 751         original source without copying.
 752         (KJS::UStringSourceProvider): SourceProvider subclass backed by a KJS::UString.
 753         (KJS::UStringSourceProvider::create):
 754         (KJS::UStringSourceProvider::getRange):
 755         (KJS::UStringSourceProvider::data):
 756         (KJS::UStringSourceProvider::length):
 757         (KJS::UStringSourceProvider::UStringSourceProvider):
 758         * kjs/SourceRange.h: Added.
 759         (KJS::SourceRange::SourceRange): Class that holds a SourceProvider and a character range into
 760         the source, to encapsulate on-demand access to the source of a function.
 761         (KJS::SourceRange::toString):
 762         * VM/Machine.cpp:
 763         (KJS::eval): Pass a UStringSourceProvider to the parser.
 764         * kjs/Parser.cpp:
 765         (KJS::Parser::parse): Take a SourceProvider and pass it on to the lexer.
 766         * kjs/Parser.h:
 767         (KJS::Parser::parse): Take a SourceProvider.
 768         * kjs/lexer.cpp:
 769         (KJS::Lexer::setCode): Take a SourceProvider; keep it around, and
 770         use it to get the raw buffer and length.
 771         * kjs/lexer.h:
 772         (KJS::Lexer::sourceRange): Convenience function to get a source
 773         range based on the lexer's source provieder, and char offsets
 774         right before and after the desired range.
 775         * kjs/function.cpp:
 776         (KJS::globalFuncEval): Pass a UStringSourceProvider to the parser.
 777         * kjs/function_object.cpp:
 778         (KJS::functionProtoFuncToString): Use toSourceString to get the source.
 779         (KJS::FunctionObjectImp::construct): Give the parser a UStringSourceProvider.
 780         * kjs/grammar.y: When parsing a function declaration, function
 781         expression, or getter or setter, tell the function body about its
 782         SourceRange.
 783         * kjs/interpreter.cpp:
 784         (KJS::Interpreter::checkSyntax): Pass a SourceProvider to the parser.
 785         (KJS::Interpreter::evaluate): Pass a SourceProvider to the parser.
 786         * kjs/interpreter.h:
 787         * kjs/nodes.h:
 788         (KJS::FunctionBodyNode::setSource): Establish a SourceRange for this function.
 789         (KJS::FunctionBodyNode::toSourceString): Get the source string out
 790         of the SourceRange.
 791         (KJS::FuncExprNode::): Take a SourceRange and set it on the body.
 792         (KJS::FuncDeclNode::): ditto
 793         * kjs/testkjs.cpp:
 794         (prettyPrintScript): Use a SourceProvider appropriately.
 795         * JavaScriptCore.exp: Export new symbols.
 796         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj: Add new files.
 797         * JavaScriptCore.xcodeproj/project.pbxproj: Add new files.
 798
 799 2008-05-09  Oliver Hunt  <oliver@apple.com>
 800
 801         Reviewed by Maciej.
 802
 803         Bring back RegisterFile tainting in order to correctly handle
 804         natively implemented getters and setters that re-enter JavaScript
 805
 806         * VM/Machine.cpp:
 807         (KJS::Machine::privateExecute):
 808         * VM/RegisterFile.h:
 809         * kjs/function.cpp:
 810         (KJS::FunctionImp::callAsFunction):
 811         * kjs/object.cpp:
 812         (KJS::JSObject::put):
 813         (KJS::tryGetAndCallProperty):
 814         * kjs/property_slot.cpp:
 815         (KJS::PropertySlot::functionGetter):
 816
 817 2008-05-09  Maciej Stachowiak  <mjs@apple.com>
 818
 819         Reviewed by Oliver.
 820
 821         - track character offsets of open and close braces, in preparation for saving function source
 822
 823         I verified that there is no performance regression from this change.
 824
 825         * kjs/grammar.y:
 826         * kjs/lexer.cpp:
 827         (KJS::Lexer::lex):
 828         (KJS::Lexer::matchPunctuator):
 829         * kjs/lexer.h:
 830
 831 2008-05-09  Oliver Hunt  <oliver@apple.com>
 832
 833         Reviewed by Nobody (build fix).
 834
 835         Debug build fix
 836
 837         * kjs/JSGlobalObject.cpp:
 838         (KJS::JSGlobalObject::restoreLocalStorage):
 839
 840 2008-05-09  Oliver Hunt  <oliver@apple.com>
 841
 842         Reviewed by Geoff.
 843
 844         Build fixes for SquirrelFish on windows.
 845
 846         * JavaScriptCore.vcproj/JavaScriptCore/JavaScriptCore.vcproj:
 847         * JavaScriptCore.vcproj/testkjs/testkjs.vcproj:
 848         * VM/Register.h:
 849         * kjs/JSGlobalObject.cpp:
 850         (KJS::JSGlobalObject::restoreLocalStorage):
 851         * kjs/collector.cpp:
 852         (KJS::Collector::allocate):
 853         (KJS::Collector::allocateNumber):
 854         * kjs/collector.h:
 855         (KJS::Collector::allocate):
 856         (KJS::Collector::allocateNumber):
 857         * kjs/property_slot.cpp:
 858
 859 2008-05-08  Maciej Stachowiak  <mjs@apple.com>
 860
 861         Reviewed by Geoff.
 862
 863         - fix activation tearoff in the case where functions are called with too many arguments
 864
 865         Fixes:
 866         fast/canvas/patternfill-repeat.html
 867         fast/dom/SelectorAPI/bug-17313.html
 868
 869         * VM/Machine.cpp:
 870         (KJS::slideRegisterWindowForCall):
 871         (KJS::scopeChainForCall):
 872         (KJS::Machine::execute):
 873         (KJS::Machine::privateExecute):
 874
 875 2008-05-08  Geoffrey Garen  <ggaren@apple.com>
 876
 877         Reviewed by Oliver Hunt.
 878
 879         Fixed failure in fast/canvas/canvas-pattern-behaviour.html.
 880
 881         SunSpider reports a small speedup. Not sure what that's about.
 882
 883         * VM/CodeBlock.cpp:
 884         (KJS::CodeBlock::dump): Fixed op_call_eval to dump as "op_call_eval".
 885         This helped me while debugging.
 886
 887         * VM/Machine.cpp:
 888         (KJS::Machine::unwindCallFrame): When looking for an activation to tear
 889         off, don't use the scope chain. Inside eval, the scope chain doesn't
 890         belong to us; it belongs to our calling function.
 891
 892         Also, don't use the needsFullScopeChain flag to decide whether to tear
 893         off the activation. "function.arguments" can create an activation
 894         for a function whose needsFullScopeChain flag is set to false.
 895
 896 2008-05-08  Maciej Stachowiak  <mjs@apple.com>
 897
 898         Reviewed by Oliver.
 899
 900         - fix function.call for calls of more than 8 arguments
 901
 902         Fixes svg/carto.net/button.svg
 903
 904         * kjs/list.cpp:
 905         (KJS::List::getSlice): properly set up the m_buffer of the target list.
 906
 907 2008-05-08  Maciej Stachowiak  <mjs@apple.com>
 908
 909         Reviewed by Oliver.
 910
 911         - don't return a null RegisterID from RegExpNode in the exception case, since the caller may need a real register
 912
 913         Fixes:
 914         - fast/regex/early-acid3-86.html
 915         - http/tests/misc/acid3.html
 916
 917         * kjs/nodes.cpp:
 918         (KJS::RegExpNode::emitCode):
 919
 920 2008-05-07  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
 921
 922         Reviewed by Oliver.
 923
 924         Fix a performance regression caused by the introduction of property
 925         attributes to SymbolTable in r32859 by encoding the attributes and the
 926         register index into a single field of SymbolTableEntry.
 927
 928         This leaves Node::optimizeVariableAccess() definitely broken, although
 929         it was probably not entirely correct in SquirrelFish before this change.
 930
 931         * VM/CodeBlock.h:
 932         (KJS::missingThisObjectMarker):
 933         * VM/CodeGenerator.cpp:
 934         (KJS::CodeGenerator::addVar):
 935         (KJS::CodeGenerator::CodeGenerator):
 936         (KJS::CodeGenerator::registerForLocal):
 937         (KJS::CodeGenerator::registerForLocalConstInit):
 938         (KJS::CodeGenerator::isLocalConstant):
 939         (KJS::CodeGenerator::addConstant):
 940         (KJS::CodeGenerator::emitCall):
 941         * VM/CodeGenerator.h:
 942         (KJS::CodeGenerator::IdentifierMapIndexHashTraits::emptyValue):
 943         * VM/Machine.cpp:
 944         (KJS::Machine::privateExecute):
 945         * kjs/JSGlobalObject.cpp:
 946         (KJS::JSGlobalObject::saveLocalStorage):
 947         * kjs/JSVariableObject.cpp:
 948         (KJS::JSVariableObject::getPropertyNames):
 949         (KJS::JSVariableObject::getPropertyAttributes):
 950         * kjs/JSVariableObject.h:
 951         (KJS::JSVariableObject::symbolTableGet):
 952         (KJS::JSVariableObject::symbolTablePut):
 953         (KJS::JSVariableObject::symbolTablePutWithAttributes):
 954         * kjs/SymbolTable.h:
 955         (KJS::SymbolTableEntry::SymbolTableEntry):
 956         (KJS::SymbolTableEntry::isEmpty):
 957         (KJS::SymbolTableEntry::getIndex):
 958         (KJS::SymbolTableEntry::getAttributes):
 959         (KJS::SymbolTableEntry::setAttributes):
 960         (KJS::SymbolTableEntry::isReadOnly):
 961         * kjs/nodes.cpp:
 962         (KJS::getSymbolTableEntry):
 963         (KJS::PostIncResolveNode::optimizeVariableAccess):
 964         (KJS::PostDecResolveNode::optimizeVariableAccess):
 965         (KJS::DeleteResolveNode::optimizeVariableAccess):
 966         (KJS::TypeOfResolveNode::optimizeVariableAccess):
 967         (KJS::PreIncResolveNode::optimizeVariableAccess):
 968         (KJS::PreDecResolveNode::optimizeVariableAccess):
 969         (KJS::ReadModifyResolveNode::optimizeVariableAccess):
 970         (KJS::AssignResolveNode::optimizeVariableAccess):
 971         (KJS::ProgramNode::initializeSymbolTable):
 972
 973 2008-05-06  Maciej Stachowiak  <mjs@apple.com>
 974
 975         Rubber stamped by Oliver.
 976
 977         - add missing ! in an assert that I failed to reverse
 978
 979         * VM/CodeGenerator.cpp:
 980         (KJS::CodeGenerator::CodeGenerator):
 981
 982 2008-05-06  Maciej Stachowiak  <mjs@apple.com>
 983
 984         Reviewed by Oliver.
 985
 986         - fixed "SQUIRRELFISH: window.this shows up as a property, but it shouldn't"
 987         https://bugs.webkit.org/show_bug.cgi?id=18868
 988
 989         The basic approach is to have "this" only be present in the symbol
 990         table at compile time, not runtime.
 991
 992         * VM/CodeGenerator.cpp:
 993         (KJS::CodeGenerator::~CodeGenerator): Remove "this" from symbol table.
 994         (KJS::CodeGenerator::CodeGenerator): Add "this" back when re-using
 995         a symbol table.
 996         * VM/CodeGenerator.h:
 997         * VM/Machine.cpp:
 998         (KJS::Machine::execute): Don't assert that "this" is in the symbol table.
 999
1000 2008-05-06  Geoffrey Garen  <ggaren@apple.com>
1001
1002         Reviewed by Oliver Hunt.
1003
1004         Trivial support for function.arguments: Currently, we only support
1005         function.arguments from within the scope of function.
1006
1007         This fixes the remaining Mozilla JS test failures.
1008
1009         SunSpider reports no change.
1010
1011         * JavaScriptCore.exp:
1012
1013         * VM/Machine.cpp:
1014         (KJS::Machine::privateExecute): Separated scope chain deref from
1015         activation register copying: since it is now possible for client code
1016         to create an activation on behalf of a function that otherwise wouldn't
1017         need one, having an activation no longer necessarily means that you need
1018         to deref the scope chain.
1019
1020         (KJS::Machine::getCallFrame): For now, this function only examines the
1021         current scope. Walking parent scopes requires some refactoring in the
1022         way we track execution stacks.
1023
1024         * kjs/ExecState.cpp:
1025         (KJS::ExecState::ExecState): We use a negative call frame offset to
1026         indicate that a given scope is not a function call scope.
1027
1028 2008-05-05  Oliver Hunt  <oliver@apple.com>
1029
1030         Reviewed by Geoff.
1031
1032         Fix call frame set up for native -> JS function calls.
1033
1034         * VM/Machine.cpp:
1035         (KJS::Machine::execute):
1036
1037 2008-05-05  Geoffrey Garen  <ggaren@apple.com>
1038
1039         Reviewed by Maciej Stachowiak.
1040
1041         Fixed ecma_3/Object/8.6.2.6-001.js, and similar bugs.
1042
1043         SunSpider reports a .4% speedup. Not sure what that's about.
1044
1045         * VM/Machine.cpp:
1046         (KJS::Machine::privateExecute): Check for exception return from equal,
1047         since toPrimitive can throw.
1048
1049         * kjs/operations.cpp:
1050         (KJS::strictEqual): In response to an error I made in an earlier version
1051         of this patch, I changed strictEqual to make clear the fact that it
1052         performs no conversions and can't throw, making it slightly more efficient
1053         in the process.
1054
1055 2008-05-05  Maciej Stachowiak  <mjs@apple.com>
1056
1057         Reviewed by Oliver.
1058
1059         - fix some dumb mistakes in my last patch
1060
1061         * VM/CodeGenerator.cpp:
1062         (KJS::CodeGenerator::emitPushScope):
1063         (KJS::CodeGenerator::emitGetPropertyNames):
1064         * VM/Machine.cpp:
1065         (KJS::Machine::privateExecute):
1066
1067 2008-05-05  Maciej Stachowiak  <mjs@apple.com>
1068
1069         Reviewed by Oliver.
1070
1071         - document opcodes relating to jumps, scopes, and property name iteration
1072
1073         Documented jmp, jtrue, false, push_scope, pop_scope, get_pnames,
1074         next_pname and jmp_scopes.
1075
1076         * VM/CodeGenerator.cpp:
1077         (KJS::CodeGenerator::emitJump):
1078         (KJS::CodeGenerator::emitJumpIfTrue):
1079         (KJS::CodeGenerator::emitJumpIfFalse):
1080         (KJS::CodeGenerator::emitPushScope):
1081         (KJS::CodeGenerator::emitNextPropertyName):
1082         (KJS::CodeGenerator::emitGetPropertyNames):
1083         * VM/CodeGenerator.h:
1084         * VM/Machine.cpp:
1085         (KJS::Machine::privateExecute):
1086         * kjs/nodes.cpp:
1087         (KJS::LogicalAndNode::emitCode):
1088         (KJS::LogicalOrNode::emitCode):
1089         (KJS::ConditionalNode::emitCode):
1090         (KJS::IfNode::emitCode):
1091         (KJS::IfElseNode::emitCode):
1092         (KJS::DoWhileNode::emitCode):
1093         (KJS::WhileNode::emitCode):
1094         (KJS::ForNode::emitCode):
1095         (KJS::ForInNode::emitCode):
1096         (KJS::WithNode::emitCode):
1097
1098 2008-05-05  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
1099
1100         Reviewed by Oliver.
1101
1102         Bug 18749: SQUIRRELFISH: const support is broken
1103         <https://bugs.webkit.org/show_bug.cgi?id=18749>
1104
1105         Adds support for const during code generation.
1106
1107         Fixes 2 layout tests.
1108
1109         * ChangeLog:
1110         * VM/CodeGenerator.cpp:
1111         (KJS::CodeGenerator::addVar):
1112         (KJS::CodeGenerator::CodeGenerator):
1113         (KJS::CodeGenerator::isLocalConstant):
1114         * VM/CodeGenerator.h:
1115         (KJS::CodeGenerator::addVar):
1116         * kjs/nodes.cpp:
1117         (KJS::PostIncResolveNode::emitCode):
1118         (KJS::PostDecResolveNode::emitCode):
1119         (KJS::PreIncResolveNode::emitCode):
1120         (KJS::PreDecResolveNode::emitCode):
1121         (KJS::ReadModifyResolveNode::emitCode):
1122         (KJS::AssignResolveNode::emitCode):
1123
1124 2008-05-04  Maciej Stachowiak  <mjs@apple.com>
1125
1126         Reviewed by Geoff.
1127
1128         - document some more opcodes (and fix argument names)
1129
1130         Added docs for eq, neq, stricteq, nstriceq, less and lesseq.
1131
1132         * VM/CodeGenerator.cpp:
1133         (KJS::CodeGenerator::emitEqual):
1134         (KJS::CodeGenerator::emitNotEqual):
1135         (KJS::CodeGenerator::emitStrictEqual):
1136         (KJS::CodeGenerator::emitNotStrictEqual):
1137         (KJS::CodeGenerator::emitLess):
1138         (KJS::CodeGenerator::emitLessEq):
1139         * VM/CodeGenerator.h:
1140         * VM/Machine.cpp:
1141         (KJS::Machine::privateExecute):
1142         * kjs/nodes.cpp:
1143         (KJS::LessNode::emitCode):
1144         (KJS::GreaterNode::emitCode):
1145         (KJS::LessEqNode::emitCode):
1146         (KJS::GreaterEqNode::emitCode):
1147         (KJS::EqualNode::emitCode):
1148         (KJS::NotEqualNode::emitCode):
1149         (KJS::StrictEqualNode::emitCode):
1150         (KJS::NotStrictEqualNode::emitCode):
1151         (KJS::CaseBlockNode::emitCodeForBlock):
1152
1153 2008-05-04  Geoffrey Garen  <ggaren@apple.com>
1154
1155         Reviewed by Maciej Stachowiak.
1156
1157         More scaffolding for f.arguments.
1158
1159         Track the offset of the last call frame in the ExecState, so we can
1160         produce a backtrace at any time.
1161
1162         Also, record numLocals, the sum of numVars + numParameters, in each code
1163         block, to make updates to the ExecState a little cheaper than they
1164         would be otherwise.
1165
1166         We now use numLocals in a bunch of places where we used to calculate
1167         numVars + numParameters or -numVars - numParameters.
1168
1169         Reports are mixed, but all in all, this seems to be a wash on SunSpider.
1170
1171 2008-05-04  Oliver Hunt  <oliver@apple.com>
1172
1173         Reviewed by Geoff.
1174
1175         Whoops, correctly handle properties that don't exist in the
1176         symbol table.
1177
1178         * kjs/JSVariableObject.h:
1179         (KJS::JSVariableObject::symbolTablePutWithAttributes):
1180
1181 2008-05-04  Oliver Hunt  <oliver@apple.com>
1182
1183         Reviewed by Geoff.
1184
1185         Add attribute information to SymbolTable as ground work for
1186         various DontEnum and ReadOnly issues.
1187
1188         * VM/CodeGenerator.cpp:
1189         (KJS::CodeGenerator::addVar):
1190         (KJS::CodeGenerator::CodeGenerator):
1191         (KJS::CodeGenerator::registerForLocal):
1192         (KJS::CodeGenerator::registerForLocalConstInit):
1193         (KJS::CodeGenerator::addConstant):
1194         * VM/Machine.cpp:
1195         (KJS::Machine::execute):
1196         * kjs/JSGlobalObject.cpp:
1197         (KJS::JSGlobalObject::saveLocalStorage):
1198         * kjs/JSVariableObject.cpp:
1199         (KJS::JSVariableObject::getPropertyNames):
1200         (KJS::JSVariableObject::getPropertyAttributes):
1201         * kjs/JSVariableObject.h:
1202         (KJS::JSVariableObject::symbolTablePut):
1203         (KJS::JSVariableObject::symbolTablePutWithAttributes):
1204         * kjs/SymbolTable.h:
1205         (KJS::SymbolTableEntry::SymbolTableEntry):
1206         (KJS::SymbolTableIndexHashTraits::emptyValue):
1207         * kjs/nodes.cpp:
1208         (KJS::getSymbolTableEntry):
1209         (KJS::ReadModifyResolveNode::optimizeVariableAccess):
1210         (KJS::AssignResolveNode::optimizeVariableAccess):
1211         (KJS::ProgramNode::initializeSymbolTable):
1212
1213 2008-05-04  Geoffrey Garen  <ggaren@apple.com>
1214
1215         Reviewed by Oliver Hunt.
1216
1217         More scaffolding for f.arguments.
1218
1219         Store the register file associated with an ExecState in the ExecState.
1220
1221         SunSpider reports no change.
1222
1223         * kjs/JSGlobalObject.h:
1224         (KJS::JSGlobalObject::JSGlobalObjectData::JSGlobalObjectData): Moved
1225         registerFileStack above globalExec, so it gets initialized first.
1226         Removed remnants of old activation scheme.
1227
1228 2008-05-04  Maciej Stachowiak  <mjs@apple.com>
1229
1230         Rubber stamped by Oliver.
1231
1232         - renamed a few opcodes and fixed assembly formatting to accomodate the longest opcode
1233
1234         equal --> eq
1235         nequal --> neq
1236         resolve_base_and_property --> resolve_with_base
1237         resolve_base_and_func --> resolve_func
1238         get_prop_id --> get_by_id
1239         put_prop_id --> put_by_id
1240         delete_prop_id --> del_by_id
1241         get_prop_val --> get_by_val
1242         put_prop_val --> put_by_val
1243         delete_prop_val --> del_by_val
1244         put_prop_index --> put_by_index
1245
1246         * VM/CodeBlock.cpp:
1247         (KJS::printUnaryOp):
1248         (KJS::printBinaryOp):
1249         (KJS::printConditionalJump):
1250         (KJS::CodeBlock::dump):
1251         * VM/CodeGenerator.cpp:
1252         (KJS::CodeGenerator::emitEqual):
1253         (KJS::CodeGenerator::emitNotEqual):
1254         (KJS::CodeGenerator::emitResolveWithBase):
1255         (KJS::CodeGenerator::emitResolveFunction):
1256         (KJS::CodeGenerator::emitGetById):
1257         (KJS::CodeGenerator::emitPutById):
1258         (KJS::CodeGenerator::emitDeleteById):
1259         (KJS::CodeGenerator::emitGetByVal):
1260         (KJS::CodeGenerator::emitPutByVal):
1261         (KJS::CodeGenerator::emitDeleteByVal):
1262         (KJS::CodeGenerator::emitPutByIndex):
1263         * VM/CodeGenerator.h:
1264         * VM/Machine.cpp:
1265         (KJS::Machine::privateExecute):
1266         * VM/Opcode.h:
1267         * kjs/nodes.cpp:
1268         (KJS::ArrayNode::emitCode):
1269         (KJS::PropertyListNode::emitCode):
1270         (KJS::BracketAccessorNode::emitCode):
1271         (KJS::DotAccessorNode::emitCode):
1272         (KJS::EvalFunctionCallNode::emitCode):
1273         (KJS::FunctionCallResolveNode::emitCode):
1274         (KJS::FunctionCallBracketNode::emitCode):
1275         (KJS::FunctionCallDotNode::emitCode):
1276         (KJS::PostIncResolveNode::emitCode):
1277         (KJS::PostDecResolveNode::emitCode):
1278         (KJS::PostIncBracketNode::emitCode):
1279         (KJS::PostDecBracketNode::emitCode):
1280         (KJS::PostIncDotNode::emitCode):
1281         (KJS::PostDecDotNode::emitCode):
1282         (KJS::DeleteResolveNode::emitCode):
1283         (KJS::DeleteBracketNode::emitCode):
1284         (KJS::DeleteDotNode::emitCode):
1285         (KJS::TypeOfResolveNode::emitCode):
1286         (KJS::PreIncResolveNode::emitCode):
1287         (KJS::PreDecResolveNode::emitCode):
1288         (KJS::PreIncBracketNode::emitCode):
1289         (KJS::PreDecBracketNode::emitCode):
1290         (KJS::PreIncDotNode::emitCode):
1291         (KJS::PreDecDotNode::emitCode):
1292         (KJS::ReadModifyResolveNode::emitCode):
1293         (KJS::AssignResolveNode::emitCode):
1294         (KJS::AssignDotNode::emitCode):
1295         (KJS::ReadModifyDotNode::emitCode):
1296         (KJS::AssignBracketNode::emitCode):
1297         (KJS::ReadModifyBracketNode::emitCode):
1298         (KJS::ConstDeclNode::emitCodeSingle):
1299         (KJS::ForInNode::emitCode):
1300         (KJS::TryNode::emitCode):
1301
1302 2008-05-04  Oliver Hunt  <oliver@apple.com>
1303
1304         Reviewed by Maciej.
1305
1306         Fix assertion when accessing arguments object with too many arguments provided
1307
1308         The arguments constructor was assuming that the register offset given for argv
1309         was an absolute offset into the registerfile, rather than the offset from the
1310         frame.  This patches corrects that issue.
1311
1312         * kjs/JSActivation.cpp:
1313         (KJS::JSActivation::createArgumentsObject):
1314
1315 2008-05-04  Geoffrey Garen  <ggaren@apple.com>
1316
1317         Rubber stamped by Sam Weinig.
1318
1319         Cleaned up Machine.cpp according to our style guidelines: moved static
1320         data to the top of the file; moved stand-alone functions below that;
1321         moved the Machine constructor above other Machine member functions.
1322
1323 2008-05-03  Maciej Stachowiak  <mjs@apple.com>
1324
1325         Reviewed by Sam.
1326
1327         - fix accidental breakage from last patch
1328
1329         * VM/Machine.cpp:
1330         (KJS::Machine::privateExecute):
1331
1332 2008-05-03  Maciej Stachowiak  <mjs@apple.com>
1333
1334         Reviewed by Geoff.
1335
1336         - a bunch more opcode documentation and corresponding parameter name fixes
1337
1338         I renamed a few opcodes:
1339
1340         type_of --> typeof (that's what the JS operator is named)
1341         instance_of --> instanceof (ditto)
1342         create_error --> new_error (for consistency with other new_* opcodes)
1343
1344         I documented the following opcodes:
1345
1346         - load
1347         - new_object
1348         - new_array
1349         - new_regexp
1350         - mov
1351         - pre_inc
1352         - pre_dec
1353         - post_inc
1354         - post_dec
1355         - to_jsnumber
1356         - negate
1357         - bitnot
1358         - not
1359         - instanceof
1360         - typeof
1361         - in
1362         - new_func
1363         - new_funcexp
1364         - new_error
1365
1366         I also fixed formatting on some existing opcode docs.
1367
1368         * VM/CodeBlock.cpp:
1369         (KJS::CodeBlock::dump):
1370         * VM/CodeGenerator.cpp:
1371         (KJS::CodeGenerator::emitMove):
1372         (KJS::CodeGenerator::emitNot):
1373         (KJS::CodeGenerator::emitPreInc):
1374         (KJS::CodeGenerator::emitPreDec):
1375         (KJS::CodeGenerator::emitPostInc):
1376         (KJS::CodeGenerator::emitPostDec):
1377         (KJS::CodeGenerator::emitToJSNumber):
1378         (KJS::CodeGenerator::emitNegate):
1379         (KJS::CodeGenerator::emitBitNot):
1380         (KJS::CodeGenerator::emitInstanceOf):
1381         (KJS::CodeGenerator::emitTypeOf):
1382         (KJS::CodeGenerator::emitIn):
1383         (KJS::CodeGenerator::emitLoad):
1384         (KJS::CodeGenerator::emitNewObject):
1385         (KJS::CodeGenerator::emitNewArray):
1386         (KJS::CodeGenerator::emitNewRegExp):
1387         (KJS::CodeGenerator::emitNewError):
1388         * VM/CodeGenerator.h:
1389         (KJS::CodeGenerator::scopeDepth):
1390         (KJS::CodeGenerator::addVar):
1391         * VM/Machine.cpp:
1392         (KJS::Machine::privateExecute):
1393         * VM/Opcode.h:
1394         * kjs/nodes.cpp:
1395         (KJS::Node::emitThrowError):
1396         (KJS::RegExpNode::emitCode):
1397         (KJS::TypeOfValueNode::emitCode):
1398         (KJS::UnaryPlusNode::emitCode):
1399         (KJS::NegateNode::emitCode):
1400         (KJS::BitwiseNotNode::emitCode):
1401         (KJS::LogicalNotNode::emitCode):
1402         (KJS::InstanceOfNode::emitCode):
1403         (KJS::InNode::emitCode):
1404
1405 2008-05-03  Maciej Stachowiak  <mjs@apple.com>
1406
1407         Reviewed by Geoff and Sam.
1408
1409         - generate HTML bytecode docs at build time
1410
1411         * DerivedSources.make:
1412         * docs: Added.
1413         * docs/make-bytecode-docs.pl: Added.
1414
1415 2008-05-03  Geoffrey Garen  <ggaren@apple.com>
1416
1417         Reviewed by Sam Weinig.
1418
1419         Update ExecState::m_scopeChain when switching scope chains inside the
1420         machine.
1421
1422         This fixes uses of lexicalGlobalObject, such as, in a subframe
1423
1424             alert(top.makeArray() instanceof Array ? "FAIL" : "PASS");
1425
1426         and a bunch of the security failures listed in
1427         https://bugs.webkit.org/show_bug.cgi?id=18870. (Those tests still fail,
1428         seemingly because of regressions in exception messages).
1429
1430         SunSpider reports no change.
1431
1432         * VM/Machine.cpp: Factored out scope chain updating into a common
1433         function that takes care to update ExecState::m_scopeChain, too.
1434
1435         * kjs/ExecState.h: I made Machine a friend of ExecState so that Machine
1436         could update ExecState::m_scopeChain, even though that value is
1437         read-only for everyone else.
1438
1439         * kjs/JSGlobalObject.h:
1440         (KJS::JSGlobalObject::JSGlobalObjectData::JSGlobalObjectData): Changed
1441         this client to be a little friendlier to ExecState's internal
1442         storage type for scope chain data.
1443
1444 2008-05-03  Geoffrey Garen  <ggaren@apple.com>
1445
1446         Reviewed by Sam Weinig.
1447
1448         Fixed https://bugs.webkit.org/show_bug.cgi?id=18876
1449         Squirrelfish: ScopeChainNode leak in op_jmp_scopes.
1450
1451         SunSpider reports no change.
1452
1453         * VM/Machine.cpp:
1454         (KJS::Machine::privateExecute): Don't construct a ScopeChain object,
1455         since the direct threaded interpreter will goto across its destructor.
1456
1457 2008-05-03  Geoffrey Garen  <ggaren@apple.com>
1458
1459         Reviewed by Oliver Hunt.
1460
1461         A bit more efficient fix than r32832: Don't copy globals into function
1462         register files; instead, have the RegisterFileStack track only the base
1463         of the last *global* register file, so the global object's register
1464         references stay good.
1465
1466         SunSpider reports a .3% speedup. Not sure what that's about.
1467
1468 2008-05-03  Oliver Hunt  <oliver@apple.com>
1469
1470         Reviewed by Maciej.
1471
1472         Bug 18864: SquirrelFish: Support getter and setter definition in object literals
1473         <https://bugs.webkit.org/show_bug.cgi?id=18864>
1474
1475         Add new opcodes to allow us to add getters and setters to an object.  These are
1476         only used by the codegen for object literals.
1477
1478         * VM/CodeGenerator.cpp:
1479         (KJS::CodeGenerator::emitPutGetter):
1480         (KJS::CodeGenerator::emitPutSetter):
1481         * VM/CodeGenerator.h:
1482         * VM/Machine.cpp:
1483         (KJS::Machine::privateExecute):
1484         * VM/Opcode.h:
1485         * kjs/nodes.cpp:
1486         (KJS::PropertyListNode::emitCode):
1487
1488 2008-05-02  Maciej Stachowiak  <mjs@apple.com>
1489
1490         Reviewed by Oliver.
1491
1492         - properly copy globals into and out of implicit call register
1493         files, otherwise they will fail at global lookup
1494
1495         Fixes fast/js/array-tostring-and-join.html layout test.
1496
1497         * VM/RegisterFileStack.cpp:
1498         (KJS::RegisterFileStack::pushGlobalRegisterFile):
1499         (KJS::RegisterFileStack::popGlobalRegisterFile):
1500         (KJS::RegisterFileStack::pushFunctionRegisterFile):
1501         (KJS::RegisterFileStack::popFunctionRegisterFile):
1502
1503 2008-05-02  Geoffrey Garen  <ggaren@apple.com>
1504
1505         Reviewed by Oliver Hunt.
1506
1507         Fixed https://bugs.webkit.org/show_bug.cgi?id=18822
1508         SQUIRRELFISH: incorrect eval used in some cases
1509
1510         Changed all code inside the machine to fetch the lexical global object
1511         directly from the scope chain, instead of from the ExecState.
1512
1513         Clients who fetch the lexical global object through the ExecState
1514         still don't work.
1515
1516         SunSpider reports no change.
1517
1518         * VM/Machine.cpp:
1519         (KJS::Machine::privateExecute): Fetch the lexical global object from
1520         the scope chain.
1521
1522         * kjs/ExecState.h:
1523         (KJS::ExecState::ExecState::lexicalGlobalObject): Moved the logic for
1524         this function into ScopeChainNode, but kept this function around to
1525         support existing clients.
1526
1527 2008-05-02  Geoffrey Garen  <ggaren@apple.com>
1528
1529         Rubber stamped by Oliver Hunt.
1530
1531         Removed ExecState.cpp from AllInOneFile.cpp, for a .2% speedup.
1532
1533         * JavaScriptCore.xcodeproj/project.pbxproj:
1534         * kjs/AllInOneFile.cpp:
1535
1536 2008-05-01  Oliver Hunt  <oliver@apple.com>
1537
1538         Reviewed by Geoff and Maciej.
1539
1540         Bug 18827: SquirrelFish: Prevent getters and setters from destroying the current RegisterFile
1541         <https://bugs.webkit.org/show_bug.cgi?id=18827>
1542
1543         Remove safe/unsafe RegisterFile concept, and instead just add additional
1544         logic to ensure we always push/pop RegisterFiles when executing getters
1545         and setters, similar to the logic for valueOf and toString.
1546
1547         * VM/Machine.cpp:
1548         (KJS::Machine::privateExecute):
1549         * VM/RegisterFile.h:
1550         * kjs/function.cpp:
1551         (KJS::FunctionImp::callAsFunction):
1552         * kjs/object.cpp:
1553         (KJS::JSObject::put):
1554         * kjs/property_slot.cpp:
1555         (KJS::PropertySlot::functionGetter):
1556
1557 2008-05-01  Oliver Hunt  <oliver@apple.com>
1558
1559         RS=Geoff
1560
1561         Rename unsafeForReentry to safeForReentry to avoid double negatives.
1562
1563         * VM/Machine.cpp:
1564         (KJS::Machine::privateExecute):
1565         * VM/RegisterFile.h:
1566         * kjs/function.cpp:
1567         (KJS::FunctionImp::callAsFunction):
1568
1569 2008-05-01  Oliver Hunt  <oliver@apple.com>
1570
1571         Reviewed by Maciej.
1572
1573         Bug 18827: SquirrelFish: Prevent getters and setters from destroying the current RegisterFile
1574         <https://bugs.webkit.org/show_bug.cgi?id=18827>
1575
1576         This patch makes getters and setters work.  It does this by
1577         tracking whether the RegisterFile is "safe", that is whether
1578         the interpreter is in a state that in which it can handle
1579         the RegisterFile being reallocated.
1580
1581         * VM/Machine.cpp:
1582         (KJS::resolve):
1583         (KJS::Machine::privateExecute):
1584         * VM/RegisterFile.h:
1585         * kjs/function.cpp:
1586         (KJS::FunctionImp::callAsFunction):
1587
1588 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1589
1590         Release build fix: Always compile in "isGlobalObject", since it's
1591         listed in our .exp file.
1592
1593         * kjs/ExecState.cpp:
1594         (KJS::ExecState::isGlobalObject):
1595         * kjs/ExecState.h:
1596
1597 2008-04-30  Oliver Hunt  <oliver@apple.com>
1598
1599         Reviewed by Maciej.
1600
1601         Minor code restructuring to prepare for getters and setters,
1602         also helps exception semantics a bit.
1603
1604         * VM/Machine.cpp:
1605         (KJS::Machine::privateExecute):
1606
1607 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1608
1609         Fixed tyop.
1610
1611         * kjs/ExecState.h:
1612
1613 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1614
1615         Debug build fix: export a missing symbol.
1616
1617         * JavaScriptCore.exp:
1618
1619 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1620
1621         Reviewed by Oliver Hunt.
1622
1623         A little more ExecState refactoring: Now, only the global object creates
1624         an ExecState.
1625
1626         Also inlined ExecState::lexicalGlobalObject().
1627
1628         SunSpider reports no change.
1629
1630 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1631
1632         WebCore build fix: forward-declare ScopeChain.
1633
1634         * kjs/interpreter.h:
1635
1636 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1637
1638         Build fix for JavaScriptGlue: export a missing symbol.
1639
1640         * JavaScriptCore.exp:
1641
1642 2008-04-30  Geoffrey Garen  <ggaren@apple.com>
1643
1644         Reviewed by Oliver Hunt.
1645
1646         Removed a lot of unused bits from ExecState, moving them into
1647         OldInterpreterExecState, the fake scaffolding class.
1648
1649         The clutter was making it hard to see the forest from the trees.
1650
1651         .4% SunSpider speedup, probably because ExecState::lexicalGlobalObject()
1652         is faster now.
1653
1654 2008-04-29  Oliver Hunt  <oliver@apple.com>
1655
1656         Reviewed by Maciej.
1657
1658         Bug 18643: SQUIRRELFISH: need to support implicit function calls (valueOf, toString, getters/setters)
1659         <https://bugs.webkit.org/show_bug.cgi?id=18643>
1660
1661         Prevent static slot optimisation for new variables and functions in
1662         globally re-entrant code called from an an implicit function call.
1663
1664         This is necessary to prevent us from needing to resize the global
1665         slot portion of the root RegisterFile during an implicit (and hence
1666         unguarded) function call.
1667
1668         * VM/CodeGenerator.cpp:
1669         (KJS::CodeGenerator::CodeGenerator):
1670         * VM/CodeGenerator.h:
1671         * VM/Machine.cpp:
1672         (KJS::Machine::execute):
1673         * VM/RegisterFile.h:
1674         * VM/RegisterFileStack.cpp:
1675         (KJS::RegisterFileStack::pushGlobalRegisterFile):
1676         (KJS::RegisterFileStack::popGlobalRegisterFile):
1677         (KJS::RegisterFileStack::pushFunctionRegisterFile):
1678         (KJS::RegisterFileStack::popFunctionRegisterFile):
1679         * VM/RegisterFileStack.h:
1680         (KJS::RegisterFileStack::inImplicitFunctionCall):
1681         (KJS::RegisterFileStack::lastGlobal):
1682         * kjs/nodes.cpp:
1683         (KJS::ProgramNode::generateCode):
1684         * kjs/nodes.h:
1685         (KJS::ProgramNode::):
1686
1687 2008-04-29  Geoffrey Garen  <ggaren@apple.com>
1688
1689         Reviewed by Oliver Hunt.
1690
1691         In nested program code, don't propogate "this" back to the parent
1692         register file. ("this" should remain constant in the parent register
1693         file, regardless of the scripts it invokes.)
1694
1695         * VM/RegisterFile.cpp:
1696         (KJS::RegisterFile::copyGlobals):
1697
1698 2008-04-28  Oliver Hunt  <oliver@apple.com>
1699
1700         Reviewed by Geoff.
1701
1702         Restore base pointer when popping a global RegisterFile
1703
1704         * VM/RegisterFileStack.cpp:
1705         (KJS::RegisterFileStack::popGlobalRegisterFile):
1706
1707 2008-04-28  Oliver Hunt  <oliver@apple.com>
1708
1709         Reviewed by Geoff.
1710
1711         Bug 18643: SQUIRRELFISH: need to support implicit function calls (valueOf, toString, getters/setters)
1712         <https://bugs.webkit.org/show_bug.cgi?id=18643>
1713
1714         Partial fix.  This results in all implicit calls to toString or valueOf
1715         executing in a separate RegisterFile, so ensuring that the the pointers
1716         in the triggering interpreter don't get trashed.  This still leaves the
1717         task of preventing new global re-entry from toString and valueOf from
1718         clobbering the RegisterFile.
1719
1720         * VM/Machine.cpp:
1721         (KJS::Machine::execute):
1722         * VM/RegisterFileStack.cpp:
1723         (KJS::RegisterFileStack::pushFunctionRegisterFile):
1724         (KJS::RegisterFileStack::popFunctionRegisterFile):
1725         * VM/RegisterFileStack.h:
1726         * kjs/object.cpp:
1727         (KJS::tryGetAndCallProperty):
1728
1729 2008-04-28  Geoffrey Garen  <ggaren@apple.com>
1730
1731         Reviewed by Maciej Stachowiak.
1732
1733         Simplified activation object a bit: No need to store the callee
1734         in the activation object -- we can pull it out of the call frame
1735         when needed, instead.
1736
1737         SunSpider reports no change.
1738
1739 2008-04-28  Geoffrey Garen  <ggaren@apple.com>
1740
1741         Reviewed by Maciej Stachowiak.
1742
1743         RS by Oliver Hunt on moving JSArguments.cpp out of AllInOneFile.cpp.
1744
1745         Substantially more handling of "arguments": "arguments" works fully
1746         now, but "f.arguments" still doesn't work.
1747
1748         Fixes 10 regression tests.
1749
1750         SunSpider reports no regression.
1751
1752         * kjs/JSActivation.cpp:
1753         (KJS::JSActivation::createArgumentsObject): Reconstruct an arguments
1754         List to pass to the arguments object constructor.
1755
1756         * JavaScriptCore.xcodeproj/project.pbxproj:
1757         * kjs/AllInOneFile.cpp: Removed JSActivation.cpp from AllInOneFile.cpp
1758         because that seems to make GCC happy. (Previously, I had added
1759         JSActivation.cpp to AllInOneFile.cpp because *that* seemed to make GCC
1760         happy. So it goes.)
1761
1762 2008-04-28  Geoffrey Garen  <ggaren@apple.com>
1763
1764         Reviewed by Maciej Stachowiak.
1765
1766         Groundwork for more handling of "arguments". I'm not checking in the
1767         actual handling of "arguments" yet, because it still needs a little
1768         fiddling to avoid a performance regression.
1769
1770         SunSpider reports no change.
1771
1772         * VM/Machine.cpp:
1773         (KJS::initializeCallFrame): Put argc in the register file, so the
1774         arguments object can find it later, to determine arguments.length.
1775
1776         * kjs/nodes.h:
1777         (KJS::FunctionBodyNode::): Added a special code accessor for when you
1778         know the code has already been generated, and you don't have a scopeChain
1779         to supply for potential code generation. (This is the case when the
1780         activation object creates the arguments object.)
1781
1782 2008-04-28  Oliver Hunt  <oliver@apple.com>
1783
1784         Reviewed by Geoff.
1785
1786         Replace unsafe use of auto_ptr in Vector with manual memory
1787         management.
1788
1789         * VM/RegisterFileStack.cpp:
1790         (KJS::RegisterFileStack::~RegisterFileStack):
1791         (KJS::RegisterFileStack::popRegisterFile):
1792         * VM/RegisterFileStack.h:
1793
1794 2008-04-27  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
1795
1796         Reviewed by Maciej.
1797
1798         Bug 18746: SQUIRRELFISH: indirect eval used when direct eval should be used
1799         <https://bugs.webkit.org/show_bug.cgi?id=18746>
1800
1801         Change the base to the correct value of the 'this' object after the direct
1802         eval test instead of before.
1803
1804         Fixes 5 layout tests.
1805
1806         * VM/Machine.cpp:
1807         (KJS::Machine::privateExecute):
1808         * kjs/nodes.cpp:
1809         (KJS::EvalFunctionCallNode::emitCode):
1810
1811 2008-04-26  Maciej Stachowiak  <mjs@apple.com>
1812
1813         Reviewed by Oliver.
1814
1815         - document all property getting, setting and deleting opcodes
1816
1817         (And fix function parameter names to match corresponding opcode parameter names.)
1818
1819         * VM/CodeGenerator.cpp:
1820         (KJS::CodeGenerator::emitResolve):
1821         (KJS::CodeGenerator::emitResolveBase):
1822         (KJS::CodeGenerator::emitResolveBaseAndProperty):
1823         (KJS::CodeGenerator::emitResolveBaseAndFunc):
1824         (KJS::CodeGenerator::emitGetPropId):
1825         (KJS::CodeGenerator::emitPutPropId):
1826         (KJS::CodeGenerator::emitDeletePropId):
1827         (KJS::CodeGenerator::emitPutPropVal):
1828         * VM/CodeGenerator.h:
1829         * VM/Machine.cpp:
1830         (KJS::resolve):
1831         (KJS::resolveBase):
1832         (KJS::resolveBaseAndProperty):
1833         (KJS::resolveBaseAndFunc):
1834         (KJS::Machine::privateExecute):
1835         * kjs/nodes.cpp:
1836         (KJS::ResolveNode::emitCode):
1837         (KJS::ArrayNode::emitCode):
1838         (KJS::PropertyListNode::emitCode):
1839         (KJS::BracketAccessorNode::emitCode):
1840         (KJS::EvalFunctionCallNode::emitCode):
1841         (KJS::FunctionCallResolveNode::emitCode):
1842         (KJS::FunctionCallBracketNode::emitCode):
1843         (KJS::PostIncResolveNode::emitCode):
1844         (KJS::PostDecResolveNode::emitCode):
1845         (KJS::PostIncBracketNode::emitCode):
1846         (KJS::PostDecBracketNode::emitCode):
1847         (KJS::PostIncDotNode::emitCode):
1848         (KJS::PostDecDotNode::emitCode):
1849         (KJS::DeleteResolveNode::emitCode):
1850         (KJS::TypeOfResolveNode::emitCode):
1851         (KJS::PreIncResolveNode::emitCode):
1852         (KJS::PreDecResolveNode::emitCode):
1853         (KJS::PreIncBracketNode::emitCode):
1854         (KJS::PreDecBracketNode::emitCode):
1855         (KJS::AssignResolveNode::emitCode):
1856         (KJS::AssignDotNode::emitCode):
1857         (KJS::ReadModifyDotNode::emitCode):
1858         (KJS::AssignBracketNode::emitCode):
1859         (KJS::ReadModifyBracketNode::emitCode):
1860         (KJS::ConstDeclNode::emitCodeSingle):
1861
1862 2008-04-26  Oliver Hunt  <oliver@apple.com>
1863
1864         Reviewed by Maciej.
1865
1866         Bug 18628: SQUIRRELFISH: need to support recursion limit
1867         <https://bugs.webkit.org/show_bug.cgi?id=18628>
1868
1869         Basically completes recursion limiting.  There is still some
1870         tuning we may want to do to make things better in the face of
1871         very bad code, but certainly nothing worse than anything already
1872         possible in trunk.
1873
1874         Also fixes a WebKit test by fixing the exception text :D
1875
1876         * JavaScriptCore.exp:
1877         * VM/ExceptionHelpers.cpp:
1878         * VM/Machine.cpp:
1879         (KJS::Machine::execute):
1880         * VM/RegisterFile.cpp:
1881         (KJS::RegisterFile::growBuffer):
1882         (KJS::RegisterFile::addGlobalSlots):
1883         * VM/RegisterFile.h:
1884         (KJS::RegisterFile::grow):
1885         (KJS::RegisterFile::uncheckedGrow):
1886         * VM/RegisterFileStack.cpp:
1887         (KJS::RegisterFileStack::pushRegisterFile):
1888         * VM/RegisterFileStack.h:
1889
1890 2008-04-25  Oliver Hunt  <oliver@apple.com>
1891
1892         Reviewed by Geoff.
1893
1894         Bug 18628: SQUIRRELFISH: need to support recursion limit
1895         <https://bugs.webkit.org/show_bug.cgi?id=18628>
1896
1897         Put a limit on the level of reentry recursion.  128 levels of re-entrant recursion
1898         seems reasonable as it is greater than the old eval limit, and a long way short of
1899         the reentry depth needed to overflow the stack.
1900
1901         * VM/Machine.cpp:
1902         (KJS::Machine::execute):
1903         * VM/Machine.h:
1904
1905 2008-04-25  Geoffrey Garen  <ggaren@apple.com>
1906
1907         Reviewed by Sam Weinig.
1908
1909         A tiny bit of cleanup to the regexp code.
1910
1911         Removed some static_cast.
1912
1913         Removed createRegExpImp because it's no longer used.
1914
1915 2008-04-25  Oliver Hunt  <oliver@apple.com>
1916
1917         Reviewed by Maciej.
1918
1919         Bug 18736: SQUIRRELFISH: switch statements with no default have incorrect codegen
1920         <https://bugs.webkit.org/show_bug.cgi?id=18736>
1921
1922         Ensure the "default" target is correct in the absence of an explicit default handler.
1923
1924         * kjs/nodes.cpp:
1925         (KJS::CaseBlockNode::emitCodeForBlock):
1926
1927 2008-04-25  Oliver Hunt  <oliver@apple.com>
1928
1929         Reviewed by Maciej.
1930
1931         Bug 18628: SQUIRRELFISH: need to support recursion limit
1932         <https://bugs.webkit.org/show_bug.cgi?id=18628>
1933
1934         More bounds checking.
1935
1936         * VM/Machine.cpp:
1937         (KJS::Machine::execute):
1938         * VM/RegisterFile.cpp:
1939         (KJS::RegisterFile::growBuffer):
1940         * VM/RegisterFile.h:
1941
1942 2008-04-25  Maciej Stachowiak  <mjs@apple.com>
1943
1944         Reviewed by Oliver.
1945
1946         - fix signal catching magic
1947
1948         The signal handlers are restored to _exit but are only set when
1949         running under run-javascriptcore-tests. fprintf from a signal
1950         handler is not safe.
1951
1952         * kjs/testkjs.cpp:
1953         (main):
1954         (parseArguments):
1955         * tests/mozilla/jsDriver.pl:
1956
1957 2008-04-25  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
1958
1959         Reviewed by Maciej.
1960
1961         Bug 18732: SQUIRRELFISH: exceptions thrown by native constructors are ignored
1962         <https://bugs.webkit.org/show_bug.cgi?id=18732>
1963
1964         Fixes another regression test.
1965
1966         * VM/Machine.cpp:
1967         (KJS::Machine::privateExecute):
1968
1969 2008-04-25  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
1970
1971         Reviewed by Maciej.
1972
1973         Bug 18728: SQUIRRELFISH: invalid regular expression constants should throw exceptions
1974         <https://bugs.webkit.org/show_bug.cgi?id=18728>
1975
1976         Fixes another regression test.
1977
1978         * kjs/nodes.cpp:
1979         (KJS::RegExpNode::emitCode):
1980
1981 2008-04-24  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
1982
1983         Reviewed by Geoffrey Garen.
1984
1985         Bug 18735: SQUIRRELFISH: closures are sometimes given an incorrect 'this' value when called
1986         <https://bugs.webkit.org/show_bug.cgi?id=18735>
1987
1988         The overloaded toThisObject method was not copied over to JSActivation.
1989
1990         Fixes two regression tests.
1991
1992         * kjs/JSActivation.cpp:
1993         (KJS::JSActivation::toThisObject):
1994         * kjs/JSActivation.h:
1995
1996 2008-04-24  Geoffrey Garen  <ggaren@apple.com>
1997
1998         Reviewed by Oliver Hunt.
1999
2000         Added support for arguments.callee.
2001
2002 2008-04-24  Oliver Hunt  <oliver@apple.com>
2003
2004         Reviewed by Maciej.
2005
2006         Bug 18628: SQUIRRELFISH: need to support recursion limit
2007         <https://bugs.webkit.org/show_bug.cgi?id=18628>
2008
2009         Partial fix -- this gets us some of the required bounds checking, but not
2010         complete coverage.  But it does manage to do them without regressing :D
2011
2012         * VM/ExceptionHelpers.cpp:
2013         (KJS::createError):
2014         (KJS::createStackOverflowError):
2015         * VM/ExceptionHelpers.h:
2016         * VM/Machine.cpp:
2017         (KJS::slideRegisterWindowForCall):
2018         (KJS::Machine::execute):
2019         (KJS::Machine::privateExecute):
2020         * VM/RegisterFile.cpp:
2021         * VM/RegisterFile.h:
2022         (KJS::RegisterFile::):
2023         (KJS::RegisterFile::RegisterFile):
2024         (KJS::RegisterFile::grow):
2025
2026 2008-04-24  Geoffrey Garen  <ggaren@apple.com>
2027
2028         Reviewed by Oliver Hunt.
2029
2030         A tiny bit more handling of "arguments": create a real, but mostly
2031         hollow, arguments object.
2032
2033         Fixes 2 regression tests.
2034
2035 2008-04-24  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2036
2037         Reviewed by Oliver.
2038
2039         Bug 18717: SQUIRRELFISH: eval returns the wrong value for a variable declaration statement
2040         <https://bugs.webkit.org/show_bug.cgi?id=18717>
2041
2042         Fixes a regression test, but exposes the failure of another due to the
2043         lack of getters and setters.
2044
2045         * kjs/nodes.cpp:
2046         (KJS::ConstDeclNode::emitCodeSingle):
2047         (KJS::ConstDeclNode::emitCode):
2048         (KJS::ConstStatementNode::emitCode):
2049         (KJS::VarStatementNode::emitCode):
2050         * kjs/nodes.h:
2051
2052 2008-04-24  Geoffrey Garen  <ggaren@apple.com>
2053
2054         Reviewed by Sam Weinig.
2055
2056         Print a CRASH statement when crashing, so test failures are not a
2057         mystery.
2058
2059         * kjs/testkjs.cpp:
2060         (handleCrash):
2061         (main):
2062
2063 2008-04-24  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2064
2065         Reviewed by Geoffrey Garen.
2066
2067         Bug 18716: SQUIRRELFISH: typeof should return undefined for an undefined variable reference
2068         <https://bugs.webkit.org/show_bug.cgi?id=18716>
2069
2070         This fixes 2 more regression tests.
2071
2072         * kjs/nodes.cpp:
2073         (KJS::TypeOfResolveNode::emitCode):
2074
2075 2008-04-24  Geoffrey Garen  <ggaren@apple.com>
2076
2077         Reviewed by Sam Weinig.
2078
2079         Put the callee in the call frame.
2080
2081         Necessary in order to support "arguments" and "arguments.callee".
2082
2083         Also fixes a latent GC bug, where an executing function could be
2084         subject to GC if the register holding it were overwritten. Here's
2085         an example that would have caused problems:
2086
2087         function f()
2088         {
2089             // Flood the machine stack to eliminate any old pointers to f.
2090             g.call({});
2091
2092             // Overwrite f in the register file.
2093             f = 1;
2094
2095             // Force a GC.
2096             for (var i = 0; i < 5000; ++i) {
2097                 ({});
2098             }
2099
2100             // Welcome to crash-ville.
2101         }
2102
2103         function g()
2104         {
2105         }
2106
2107         f();
2108
2109         * VM/Machine.h: Changed the order of arguments to
2110         execute(FunctionBodyNode*...) to match the other execute functions.
2111         * kjs/function.cpp: Updated to match new argument requirements from
2112         execute(FunctionBodyNode*...). Renamed newObj to thisObj to match the
2113         rest of JavaScriptCore.
2114
2115         SunSpider reports no change.
2116
2117 2008-04-23  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2118
2119         Reviewed by Maciej.
2120
2121         Bug 18707: SQUIRRELFISH: eval always performs toString() on its argument
2122         <https://bugs.webkit.org/show_bug.cgi?id=18707>
2123
2124         This fixes 4 more regression tests.
2125
2126         * VM/Machine.cpp:
2127         (KJS::eval):
2128
2129 2008-04-23  Maciej Stachowiak  <mjs@apple.com>
2130
2131         Reviewed by Oliver.
2132
2133         - fix logic bug in SegmentedVector::grow which would sometimes fail to resize a segment when needed
2134
2135         Fixes 3 JSC tests.
2136
2137         * VM/SegmentedVector.h:
2138         (KJS::SegmentedVector::grow):
2139
2140 2008-04-23  Geoffrey Garen  <ggaren@apple.com>
2141
2142         Reviewed by Maciej Stachowiak.
2143
2144         Degenerate handling of "arguments" as a property of the activation
2145         object. Currently, we just return a vanilla object.
2146
2147         SunSpider reports no change.
2148
2149         Fixes:
2150
2151         ecma_3/Function/regress-94506.js.
2152
2153         Reveals to have been secretly broken:
2154
2155         ecma_3/Function/15.3.4.3-1.js
2156         ecma_3/Function/15.3.4.4-1.js
2157
2158         These tests were passing incorrectly. testkjs creates a global array
2159         named "arguments" to hold command-line arguments. That array was
2160         tricking these tests into thinking that an arguments object with length
2161         0 had been created. Since our new vanilla object shadows the global
2162         property named arguments, that object no longer fools these tests into
2163         passing.
2164
2165         Net change: +1 failing test.
2166
2167         * kjs/AllInOneFile.cpp: Had to put JSActivation.cpp into AllInOneFile.cpp
2168         to solve a surprising 8.6% regression in bitops-3bit-bits-in-byte.
2169
2170 2008-04-23  Maciej Stachowiak  <mjs@apple.com>
2171
2172         Reviewed by Oliver.
2173
2174         - save and restore callFrame
2175
2176         * VM/Machine.cpp:
2177         (KJS::slideRegisterWindowForCall):
2178         (KJS::Machine::execute):
2179         (KJS::Machine::privateExecute):
2180         * kjs/testkjs.cpp:
2181         (main):
2182
2183 2008-04-23  Geoffrey Garen  <ggaren@apple.com>
2184
2185         Reviewed by Maciej Stachowiak.
2186
2187         Fixed scopes for named function expressions.
2188
2189         Fixes one regression test.
2190
2191         Two changes here:
2192
2193         (1) The function's name is supposed to have attributes DontDelete,
2194         ReadOnly, regardless of the type of code executing.
2195
2196         (2) Push the name object on the function's scope chain, rather than
2197         the ExecState's scope chain because, well, that's where it belongs.
2198
2199 2008-04-23  Geoffrey Garen  <ggaren@apple.com>
2200
2201         Reviewed by Oliver Hunt.
2202
2203         Inlined JSObject::putDirect, for a .4% SunSpider speedup.
2204
2205         I did this as a first step toward removing nodes.cpp from
2206         AllInOneFile.cpp, but I'm putting that larger project aside for now.
2207
2208 2008-04-23  Maciej Stachowiak  <mjs@apple.com>
2209
2210         Rubber stamped by Geoff.
2211
2212         - add OldInterpreterExecState class and use it in dead code
2213
2214         This will allow removing things from the real ExecState class
2215         without having to figure out how to remove all this code without
2216         getting a perf regression.
2217
2218         * kjs/nodes.cpp:
2219         (KJS::ExpressionNode::evaluateToNumber):
2220         (KJS::ExpressionNode::evaluateToBoolean):
2221         (KJS::ExpressionNode::evaluateToInt32):
2222         (KJS::ExpressionNode::evaluateToUInt32):
2223         (KJS::Node::setErrorCompletion):
2224         (KJS::Node::throwError):
2225         (KJS::Node::throwUndefinedVariableError):
2226         (KJS::Node::handleException):
2227         (KJS::Node::rethrowException):
2228         (KJS::BreakpointCheckStatement::execute):
2229         (KJS::BreakpointCheckStatement::optimizeVariableAccess):
2230         (KJS::NullNode::evaluate):
2231         (KJS::FalseNode::evaluate):
2232         (KJS::TrueNode::evaluate):
2233         (KJS::NumberNode::evaluate):
2234         (KJS::NumberNode::evaluateToNumber):
2235         (KJS::NumberNode::evaluateToBoolean):
2236         (KJS::NumberNode::evaluateToInt32):
2237         (KJS::NumberNode::evaluateToUInt32):
2238         (KJS::ImmediateNumberNode::evaluate):
2239         (KJS::ImmediateNumberNode::evaluateToInt32):
2240         (KJS::ImmediateNumberNode::evaluateToUInt32):
2241         (KJS::StringNode::evaluate):
2242         (KJS::StringNode::evaluateToNumber):
2243         (KJS::StringNode::evaluateToBoolean):
2244         (KJS::RegExpNode::evaluate):
2245         (KJS::ThisNode::evaluate):
2246         (KJS::ResolveNode::inlineEvaluate):
2247         (KJS::ResolveNode::evaluate):
2248         (KJS::ResolveNode::evaluateToNumber):
2249         (KJS::ResolveNode::evaluateToBoolean):
2250         (KJS::ResolveNode::evaluateToInt32):
2251         (KJS::ResolveNode::evaluateToUInt32):
2252         (KJS::getSymbolTableEntry):
2253         (KJS::ResolveNode::optimizeVariableAccess):
2254         (KJS::LocalVarAccessNode::inlineEvaluate):
2255         (KJS::LocalVarAccessNode::evaluate):
2256         (KJS::LocalVarAccessNode::evaluateToNumber):
2257         (KJS::LocalVarAccessNode::evaluateToBoolean):
2258         (KJS::LocalVarAccessNode::evaluateToInt32):
2259         (KJS::LocalVarAccessNode::evaluateToUInt32):
2260         (KJS::getNonLocalSymbol):
2261         (KJS::ScopedVarAccessNode::inlineEvaluate):
2262         (KJS::ScopedVarAccessNode::evaluate):
2263         (KJS::ScopedVarAccessNode::evaluateToNumber):
2264         (KJS::ScopedVarAccessNode::evaluateToBoolean):
2265         (KJS::ScopedVarAccessNode::evaluateToInt32):
2266         (KJS::ScopedVarAccessNode::evaluateToUInt32):
2267         (KJS::NonLocalVarAccessNode::inlineEvaluate):
2268         (KJS::NonLocalVarAccessNode::evaluate):
2269         (KJS::NonLocalVarAccessNode::evaluateToNumber):
2270         (KJS::NonLocalVarAccessNode::evaluateToBoolean):
2271         (KJS::NonLocalVarAccessNode::evaluateToInt32):
2272         (KJS::NonLocalVarAccessNode::evaluateToUInt32):
2273         (KJS::ElementNode::optimizeVariableAccess):
2274         (KJS::ElementNode::evaluate):
2275         (KJS::ArrayNode::optimizeVariableAccess):
2276         (KJS::ArrayNode::evaluate):
2277         (KJS::ObjectLiteralNode::optimizeVariableAccess):
2278         (KJS::ObjectLiteralNode::evaluate):
2279         (KJS::PropertyListNode::optimizeVariableAccess):
2280         (KJS::PropertyListNode::evaluate):
2281         (KJS::PropertyNode::optimizeVariableAccess):
2282         (KJS::PropertyNode::evaluate):
2283         (KJS::BracketAccessorNode::optimizeVariableAccess):
2284         (KJS::BracketAccessorNode::inlineEvaluate):
2285         (KJS::BracketAccessorNode::evaluate):
2286         (KJS::BracketAccessorNode::evaluateToNumber):
2287         (KJS::BracketAccessorNode::evaluateToBoolean):
2288         (KJS::BracketAccessorNode::evaluateToInt32):
2289         (KJS::BracketAccessorNode::evaluateToUInt32):
2290         (KJS::DotAccessorNode::optimizeVariableAccess):
2291         (KJS::DotAccessorNode::inlineEvaluate):
2292         (KJS::DotAccessorNode::evaluate):
2293         (KJS::DotAccessorNode::evaluateToNumber):
2294         (KJS::DotAccessorNode::evaluateToBoolean):
2295         (KJS::DotAccessorNode::evaluateToInt32):
2296         (KJS::DotAccessorNode::evaluateToUInt32):
2297         (KJS::ArgumentListNode::optimizeVariableAccess):
2298         (KJS::ArgumentListNode::evaluateList):
2299         (KJS::ArgumentsNode::optimizeVariableAccess):
2300         (KJS::NewExprNode::optimizeVariableAccess):
2301         (KJS::NewExprNode::inlineEvaluate):
2302         (KJS::NewExprNode::evaluate):
2303         (KJS::NewExprNode::evaluateToNumber):
2304         (KJS::NewExprNode::evaluateToBoolean):
2305         (KJS::NewExprNode::evaluateToInt32):
2306         (KJS::NewExprNode::evaluateToUInt32):
2307         (KJS::ExpressionNode::resolveAndCall):
2308         (KJS::EvalFunctionCallNode::optimizeVariableAccess):
2309         (KJS::EvalFunctionCallNode::evaluate):
2310         (KJS::FunctionCallValueNode::optimizeVariableAccess):
2311         (KJS::FunctionCallValueNode::evaluate):
2312         (KJS::FunctionCallResolveNode::optimizeVariableAccess):
2313         (KJS::FunctionCallResolveNode::inlineEvaluate):
2314         (KJS::FunctionCallResolveNode::evaluate):
2315         (KJS::FunctionCallResolveNode::evaluateToNumber):
2316         (KJS::FunctionCallResolveNode::evaluateToBoolean):
2317         (KJS::FunctionCallResolveNode::evaluateToInt32):
2318         (KJS::FunctionCallResolveNode::evaluateToUInt32):
2319         (KJS::LocalVarFunctionCallNode::inlineEvaluate):
2320         (KJS::LocalVarFunctionCallNode::evaluate):
2321         (KJS::LocalVarFunctionCallNode::evaluateToNumber):
2322         (KJS::LocalVarFunctionCallNode::evaluateToBoolean):
2323         (KJS::LocalVarFunctionCallNode::evaluateToInt32):
2324         (KJS::LocalVarFunctionCallNode::evaluateToUInt32):
2325         (KJS::ScopedVarFunctionCallNode::inlineEvaluate):
2326         (KJS::ScopedVarFunctionCallNode::evaluate):
2327         (KJS::ScopedVarFunctionCallNode::evaluateToNumber):
2328         (KJS::ScopedVarFunctionCallNode::evaluateToBoolean):
2329         (KJS::ScopedVarFunctionCallNode::evaluateToInt32):
2330         (KJS::ScopedVarFunctionCallNode::evaluateToUInt32):
2331         (KJS::NonLocalVarFunctionCallNode::inlineEvaluate):
2332         (KJS::NonLocalVarFunctionCallNode::evaluate):
2333         (KJS::NonLocalVarFunctionCallNode::evaluateToNumber):
2334         (KJS::NonLocalVarFunctionCallNode::evaluateToBoolean):
2335         (KJS::NonLocalVarFunctionCallNode::evaluateToInt32):
2336         (KJS::NonLocalVarFunctionCallNode::evaluateToUInt32):
2337         (KJS::FunctionCallBracketNode::optimizeVariableAccess):
2338         (KJS::FunctionCallBracketNode::evaluate):
2339         (KJS::FunctionCallDotNode::optimizeVariableAccess):
2340         (KJS::FunctionCallDotNode::inlineEvaluate):
2341         (KJS::FunctionCallDotNode::evaluate):
2342         (KJS::FunctionCallDotNode::evaluateToNumber):
2343         (KJS::FunctionCallDotNode::evaluateToBoolean):
2344         (KJS::FunctionCallDotNode::evaluateToInt32):
2345         (KJS::FunctionCallDotNode::evaluateToUInt32):
2346         (KJS::PostIncResolveNode::optimizeVariableAccess):
2347         (KJS::PostIncResolveNode::evaluate):
2348         (KJS::PostIncLocalVarNode::evaluate):
2349         (KJS::PostDecResolveNode::optimizeVariableAccess):
2350         (KJS::PostDecResolveNode::evaluate):
2351         (KJS::PostDecLocalVarNode::evaluate):
2352         (KJS::PostDecLocalVarNode::inlineEvaluateToNumber):
2353         (KJS::PostDecLocalVarNode::evaluateToNumber):
2354         (KJS::PostDecLocalVarNode::evaluateToBoolean):
2355         (KJS::PostDecLocalVarNode::evaluateToInt32):
2356         (KJS::PostDecLocalVarNode::evaluateToUInt32):
2357         (KJS::PostfixBracketNode::optimizeVariableAccess):
2358         (KJS::PostIncBracketNode::evaluate):
2359         (KJS::PostDecBracketNode::evaluate):
2360         (KJS::PostfixDotNode::optimizeVariableAccess):
2361         (KJS::PostIncDotNode::evaluate):
2362         (KJS::PostDecDotNode::evaluate):
2363         (KJS::PostfixErrorNode::evaluate):
2364         (KJS::DeleteResolveNode::optimizeVariableAccess):
2365         (KJS::DeleteResolveNode::evaluate):
2366         (KJS::LocalVarDeleteNode::evaluate):
2367         (KJS::DeleteBracketNode::optimizeVariableAccess):
2368         (KJS::DeleteBracketNode::evaluate):
2369         (KJS::DeleteDotNode::optimizeVariableAccess):
2370         (KJS::DeleteDotNode::evaluate):
2371         (KJS::DeleteValueNode::optimizeVariableAccess):
2372         (KJS::DeleteValueNode::evaluate):
2373         (KJS::VoidNode::optimizeVariableAccess):
2374         (KJS::VoidNode::evaluate):
2375         (KJS::TypeOfValueNode::optimizeVariableAccess):
2376         (KJS::TypeOfResolveNode::optimizeVariableAccess):
2377         (KJS::LocalVarTypeOfNode::evaluate):
2378         (KJS::TypeOfResolveNode::evaluate):
2379         (KJS::TypeOfValueNode::evaluate):
2380         (KJS::PreIncResolveNode::optimizeVariableAccess):
2381         (KJS::PreIncLocalVarNode::evaluate):
2382         (KJS::PreIncResolveNode::evaluate):
2383         (KJS::PreDecResolveNode::optimizeVariableAccess):
2384         (KJS::PreDecLocalVarNode::evaluate):
2385         (KJS::PreDecResolveNode::evaluate):
2386         (KJS::PreIncConstNode::evaluate):
2387         (KJS::PreDecConstNode::evaluate):
2388         (KJS::PostIncConstNode::evaluate):
2389         (KJS::PostDecConstNode::evaluate):
2390         (KJS::PrefixBracketNode::optimizeVariableAccess):
2391         (KJS::PreIncBracketNode::evaluate):
2392         (KJS::PreDecBracketNode::evaluate):
2393         (KJS::PrefixDotNode::optimizeVariableAccess):
2394         (KJS::PreIncDotNode::evaluate):
2395         (KJS::PreDecDotNode::evaluate):
2396         (KJS::PrefixErrorNode::evaluate):
2397         (KJS::UnaryPlusNode::optimizeVariableAccess):
2398         (KJS::UnaryPlusNode::evaluate):
2399         (KJS::UnaryPlusNode::evaluateToBoolean):
2400         (KJS::UnaryPlusNode::evaluateToNumber):
2401         (KJS::UnaryPlusNode::evaluateToInt32):
2402         (KJS::UnaryPlusNode::evaluateToUInt32):
2403         (KJS::NegateNode::optimizeVariableAccess):
2404         (KJS::NegateNode::evaluate):
2405         (KJS::NegateNode::evaluateToNumber):
2406         (KJS::BitwiseNotNode::optimizeVariableAccess):
2407         (KJS::BitwiseNotNode::inlineEvaluateToInt32):
2408         (KJS::BitwiseNotNode::evaluate):
2409         (KJS::BitwiseNotNode::evaluateToNumber):
2410         (KJS::BitwiseNotNode::evaluateToBoolean):
2411         (KJS::BitwiseNotNode::evaluateToInt32):
2412         (KJS::BitwiseNotNode::evaluateToUInt32):
2413         (KJS::LogicalNotNode::optimizeVariableAccess):
2414         (KJS::LogicalNotNode::evaluate):
2415         (KJS::LogicalNotNode::evaluateToBoolean):
2416         (KJS::MultNode::optimizeVariableAccess):
2417         (KJS::MultNode::inlineEvaluateToNumber):
2418         (KJS::MultNode::evaluate):
2419         (KJS::MultNode::evaluateToNumber):
2420         (KJS::MultNode::evaluateToBoolean):
2421         (KJS::MultNode::evaluateToInt32):
2422         (KJS::MultNode::evaluateToUInt32):
2423         (KJS::DivNode::optimizeVariableAccess):
2424         (KJS::DivNode::inlineEvaluateToNumber):
2425         (KJS::DivNode::evaluate):
2426         (KJS::DivNode::evaluateToNumber):
2427         (KJS::DivNode::evaluateToInt32):
2428         (KJS::DivNode::evaluateToUInt32):
2429         (KJS::ModNode::optimizeVariableAccess):
2430         (KJS::ModNode::inlineEvaluateToNumber):
2431         (KJS::ModNode::evaluate):
2432         (KJS::ModNode::evaluateToNumber):
2433         (KJS::ModNode::evaluateToBoolean):
2434         (KJS::ModNode::evaluateToInt32):
2435         (KJS::ModNode::evaluateToUInt32):
2436         (KJS::throwOutOfMemoryErrorToNumber):
2437         (KJS::addSlowCase):
2438         (KJS::addSlowCaseToNumber):
2439         (KJS::add):
2440         (KJS::addToNumber):
2441         (KJS::AddNode::optimizeVariableAccess):
2442         (KJS::AddNode::evaluate):
2443         (KJS::AddNode::inlineEvaluateToNumber):
2444         (KJS::AddNode::evaluateToNumber):
2445         (KJS::AddNode::evaluateToInt32):
2446         (KJS::AddNode::evaluateToUInt32):
2447         (KJS::AddNumbersNode::inlineEvaluateToNumber):
2448         (KJS::AddNumbersNode::evaluate):
2449         (KJS::AddNumbersNode::evaluateToNumber):
2450         (KJS::AddNumbersNode::evaluateToInt32):
2451         (KJS::AddNumbersNode::evaluateToUInt32):
2452         (KJS::AddStringsNode::evaluate):
2453         (KJS::AddStringLeftNode::evaluate):
2454         (KJS::AddStringRightNode::evaluate):
2455         (KJS::SubNode::optimizeVariableAccess):
2456         (KJS::SubNode::inlineEvaluateToNumber):
2457         (KJS::SubNode::evaluate):
2458         (KJS::SubNode::evaluateToNumber):
2459         (KJS::SubNode::evaluateToInt32):
2460         (KJS::SubNode::evaluateToUInt32):
2461         (KJS::LeftShiftNode::optimizeVariableAccess):
2462         (KJS::LeftShiftNode::inlineEvaluateToInt32):
2463         (KJS::LeftShiftNode::evaluate):
2464         (KJS::LeftShiftNode::evaluateToNumber):
2465         (KJS::LeftShiftNode::evaluateToInt32):
2466         (KJS::LeftShiftNode::evaluateToUInt32):
2467         (KJS::RightShiftNode::optimizeVariableAccess):
2468         (KJS::RightShiftNode::inlineEvaluateToInt32):
2469         (KJS::RightShiftNode::evaluate):
2470         (KJS::RightShiftNode::evaluateToNumber):
2471         (KJS::RightShiftNode::evaluateToInt32):
2472         (KJS::RightShiftNode::evaluateToUInt32):
2473         (KJS::UnsignedRightShiftNode::optimizeVariableAccess):
2474         (KJS::UnsignedRightShiftNode::inlineEvaluateToUInt32):
2475         (KJS::UnsignedRightShiftNode::evaluate):
2476         (KJS::UnsignedRightShiftNode::evaluateToNumber):
2477         (KJS::UnsignedRightShiftNode::evaluateToInt32):
2478         (KJS::UnsignedRightShiftNode::evaluateToUInt32):
2479         (KJS::lessThan):
2480         (KJS::lessThanEq):
2481         (KJS::LessNode::optimizeVariableAccess):
2482         (KJS::LessNode::inlineEvaluateToBoolean):
2483         (KJS::LessNode::evaluate):
2484         (KJS::LessNode::evaluateToBoolean):
2485         (KJS::LessNumbersNode::inlineEvaluateToBoolean):
2486         (KJS::LessNumbersNode::evaluate):
2487         (KJS::LessNumbersNode::evaluateToBoolean):
2488         (KJS::LessStringsNode::inlineEvaluateToBoolean):
2489         (KJS::LessStringsNode::evaluate):
2490         (KJS::LessStringsNode::evaluateToBoolean):
2491         (KJS::GreaterNode::optimizeVariableAccess):
2492         (KJS::GreaterNode::inlineEvaluateToBoolean):
2493         (KJS::GreaterNode::evaluate):
2494         (KJS::GreaterNode::evaluateToBoolean):
2495         (KJS::LessEqNode::optimizeVariableAccess):
2496         (KJS::LessEqNode::inlineEvaluateToBoolean):
2497         (KJS::LessEqNode::evaluate):
2498         (KJS::LessEqNode::evaluateToBoolean):
2499         (KJS::GreaterEqNode::optimizeVariableAccess):
2500         (KJS::GreaterEqNode::inlineEvaluateToBoolean):
2501         (KJS::GreaterEqNode::evaluate):
2502         (KJS::GreaterEqNode::evaluateToBoolean):
2503         (KJS::InstanceOfNode::optimizeVariableAccess):
2504         (KJS::InstanceOfNode::evaluate):
2505         (KJS::InstanceOfNode::evaluateToBoolean):
2506         (KJS::InNode::optimizeVariableAccess):
2507         (KJS::InNode::evaluate):
2508         (KJS::InNode::evaluateToBoolean):
2509         (KJS::EqualNode::optimizeVariableAccess):
2510         (KJS::EqualNode::inlineEvaluateToBoolean):
2511         (KJS::EqualNode::evaluate):
2512         (KJS::EqualNode::evaluateToBoolean):
2513         (KJS::NotEqualNode::optimizeVariableAccess):
2514         (KJS::NotEqualNode::inlineEvaluateToBoolean):
2515         (KJS::NotEqualNode::evaluate):
2516         (KJS::NotEqualNode::evaluateToBoolean):
2517         (KJS::StrictEqualNode::optimizeVariableAccess):
2518         (KJS::StrictEqualNode::inlineEvaluateToBoolean):
2519         (KJS::StrictEqualNode::evaluate):
2520         (KJS::StrictEqualNode::evaluateToBoolean):
2521         (KJS::NotStrictEqualNode::optimizeVariableAccess):
2522         (KJS::NotStrictEqualNode::inlineEvaluateToBoolean):
2523         (KJS::NotStrictEqualNode::evaluate):
2524         (KJS::NotStrictEqualNode::evaluateToBoolean):
2525         (KJS::BitAndNode::optimizeVariableAccess):
2526         (KJS::BitAndNode::evaluate):
2527         (KJS::BitAndNode::inlineEvaluateToInt32):
2528         (KJS::BitAndNode::evaluateToNumber):
2529         (KJS::BitAndNode::evaluateToBoolean):
2530         (KJS::BitAndNode::evaluateToInt32):
2531         (KJS::BitAndNode::evaluateToUInt32):
2532         (KJS::BitXOrNode::optimizeVariableAccess):
2533         (KJS::BitXOrNode::inlineEvaluateToInt32):
2534         (KJS::BitXOrNode::evaluate):
2535         (KJS::BitXOrNode::evaluateToNumber):
2536         (KJS::BitXOrNode::evaluateToBoolean):
2537         (KJS::BitXOrNode::evaluateToInt32):
2538         (KJS::BitXOrNode::evaluateToUInt32):
2539         (KJS::BitOrNode::optimizeVariableAccess):
2540         (KJS::BitOrNode::inlineEvaluateToInt32):
2541         (KJS::BitOrNode::evaluate):
2542         (KJS::BitOrNode::evaluateToNumber):
2543         (KJS::BitOrNode::evaluateToBoolean):
2544         (KJS::BitOrNode::evaluateToInt32):
2545         (KJS::BitOrNode::evaluateToUInt32):
2546         (KJS::LogicalAndNode::optimizeVariableAccess):
2547         (KJS::LogicalAndNode::evaluate):
2548         (KJS::LogicalAndNode::evaluateToBoolean):
2549         (KJS::LogicalOrNode::optimizeVariableAccess):
2550         (KJS::LogicalOrNode::evaluate):
2551         (KJS::LogicalOrNode::evaluateToBoolean):
2552         (KJS::ConditionalNode::optimizeVariableAccess):
2553         (KJS::ConditionalNode::evaluate):
2554         (KJS::ConditionalNode::evaluateToBoolean):
2555         (KJS::ConditionalNode::evaluateToNumber):
2556         (KJS::ConditionalNode::evaluateToInt32):
2557         (KJS::ConditionalNode::evaluateToUInt32):
2558         (KJS::valueForReadModifyAssignment):
2559         (KJS::ReadModifyResolveNode::optimizeVariableAccess):
2560         (KJS::AssignResolveNode::optimizeVariableAccess):
2561         (KJS::ReadModifyLocalVarNode::evaluate):
2562         (KJS::AssignLocalVarNode::evaluate):
2563         (KJS::ReadModifyConstNode::evaluate):
2564         (KJS::AssignConstNode::evaluate):
2565         (KJS::ReadModifyResolveNode::evaluate):
2566         (KJS::AssignResolveNode::evaluate):
2567         (KJS::AssignDotNode::optimizeVariableAccess):
2568         (KJS::AssignDotNode::evaluate):
2569         (KJS::ReadModifyDotNode::optimizeVariableAccess):
2570         (KJS::ReadModifyDotNode::evaluate):
2571         (KJS::AssignErrorNode::evaluate):
2572         (KJS::AssignBracketNode::optimizeVariableAccess):
2573         (KJS::AssignBracketNode::evaluate):
2574         (KJS::ReadModifyBracketNode::optimizeVariableAccess):
2575         (KJS::ReadModifyBracketNode::evaluate):
2576         (KJS::CommaNode::optimizeVariableAccess):
2577         (KJS::CommaNode::evaluate):
2578         (KJS::ConstDeclNode::optimizeVariableAccess):
2579         (KJS::ConstDeclNode::handleSlowCase):
2580         (KJS::ConstDeclNode::evaluateSingle):
2581         (KJS::ConstDeclNode::evaluate):
2582         (KJS::ConstStatementNode::optimizeVariableAccess):
2583         (KJS::ConstStatementNode::execute):
2584         (KJS::statementListExecute):
2585         (KJS::BlockNode::optimizeVariableAccess):
2586         (KJS::BlockNode::execute):
2587         (KJS::EmptyStatementNode::execute):
2588         (KJS::ExprStatementNode::optimizeVariableAccess):
2589         (KJS::ExprStatementNode::execute):
2590         (KJS::VarStatementNode::optimizeVariableAccess):
2591         (KJS::VarStatementNode::execute):
2592         (KJS::IfNode::optimizeVariableAccess):
2593         (KJS::IfNode::execute):
2594         (KJS::IfElseNode::optimizeVariableAccess):
2595         (KJS::IfElseNode::execute):
2596         (KJS::DoWhileNode::optimizeVariableAccess):
2597         (KJS::DoWhileNode::execute):
2598         (KJS::WhileNode::optimizeVariableAccess):
2599         (KJS::WhileNode::execute):
2600         (KJS::ForNode::optimizeVariableAccess):
2601         (KJS::ForNode::execute):
2602         (KJS::ForInNode::optimizeVariableAccess):
2603         (KJS::ForInNode::execute):
2604         (KJS::ContinueNode::execute):
2605         (KJS::BreakNode::execute):
2606         (KJS::ReturnNode::optimizeVariableAccess):
2607         (KJS::ReturnNode::execute):
2608         (KJS::WithNode::optimizeVariableAccess):
2609         (KJS::WithNode::execute):
2610         (KJS::CaseClauseNode::optimizeVariableAccess):
2611         (KJS::CaseClauseNode::evaluate):
2612         (KJS::CaseClauseNode::executeStatements):
2613         (KJS::ClauseListNode::optimizeVariableAccess):
2614         (KJS::CaseBlockNode::optimizeVariableAccess):
2615         (KJS::CaseBlockNode::executeBlock):
2616         (KJS::SwitchNode::optimizeVariableAccess):
2617         (KJS::SwitchNode::execute):
2618         (KJS::LabelNode::optimizeVariableAccess):
2619         (KJS::LabelNode::execute):
2620         (KJS::ThrowNode::optimizeVariableAccess):
2621         (KJS::ThrowNode::execute):
2622         (KJS::TryNode::optimizeVariableAccess):
2623         (KJS::TryNode::execute):
2624         (KJS::ProgramNode::initializeSymbolTable):
2625         (KJS::ScopeNode::optimizeVariableAccess):
2626         (KJS::ProgramNode::processDeclarations):
2627         (KJS::EvalNode::processDeclarations):
2628         (KJS::ProgramNode::execute):
2629         (KJS::EvalNode::execute):
2630         (KJS::FunctionBodyNodeWithDebuggerHooks::execute):
2631         (KJS::FuncDeclNode::execute):
2632         (KJS::FuncExprNode::evaluate):
2633         * kjs/nodes.h:
2634         (KJS::Node::):
2635         (KJS::FalseNode::):
2636         (KJS::TrueNode::):
2637         (KJS::ArgumentsNode::):
2638
2639 2008-04-23  Oliver Hunt  <oliver@apple.com>
2640
2641         Reviewed by Geoff.
2642
2643         Bug 18672: SQUIRRELFISH: codegen fails with a large number of temporaries
2644         <https://bugs.webkit.org/show_bug.cgi?id=18672>
2645
2646         Add a SegmentedVector type, which provides a Vector<T> which maintains
2647         existing memory locations during resize.  This allows dynamically sizing
2648         local, temporary and label "vectors" in CodeGenerator.
2649
2650         * JavaScriptCore.xcodeproj/project.pbxproj:
2651         * VM/CodeGenerator.cpp:
2652         (KJS::CodeGenerator::addVar):
2653         (KJS::CodeGenerator::CodeGenerator):
2654         (KJS::CodeGenerator::newTemporary):
2655         (KJS::CodeGenerator::newLabel):
2656         * VM/CodeGenerator.h:
2657         * VM/SegmentedVector.h: Added.
2658         (KJS::SegmentedVector::SegmentedVector):
2659         (KJS::SegmentedVector::~SegmentedVector):
2660         (KJS::SegmentedVector::last):
2661         (KJS::SegmentedVector::append):
2662         (KJS::SegmentedVector::removeLast):
2663         (KJS::SegmentedVector::size):
2664         (KJS::SegmentedVector::operator[]):
2665         (KJS::SegmentedVector::resize):
2666         (KJS::SegmentedVector::shrink):
2667         (KJS::SegmentedVector::grow):
2668
2669 2008-04-23  Geoffrey Garen  <ggaren@apple.com>
2670
2671         Reviewed by Maciej Stachowiak.
2672
2673         A little refactoring in preparation for supporting 'arguments'.
2674
2675         Fixes 2 regression tests.
2676
2677         SunSpider reports no change.
2678
2679         We now check the activation register, instead of the codeBlock, to
2680         determine whether we need to tear off the activation. This is to support
2681         "f.arguments", which will create an activation/arguments pair for f,
2682         even though the needsFullScopeChain flag is false for f's codeBlock.
2683
2684         The test fixes resulted from calling initializeCallFrame for re-entrant
2685         function code, instead of initializing (not enough) parts of the call
2686         frame by hand.
2687
2688 2008-04-22  Maciej Stachowiak  <mjs@apple.com>
2689
2690         Reviewed by Sam.
2691
2692         - propagate the "this" value properly to local eval
2693
2694         (fixes a measly one regression test)
2695
2696         * VM/CodeBlock.h:
2697         (KJS::CodeBlock::CodeBlock):
2698         (KJS::ProgramCodeBlock::ProgramCodeBlock):
2699         (KJS::EvalCodeBlock::EvalCodeBlock):
2700         * VM/Machine.cpp:
2701         (KJS::Machine::privateExecute):
2702
2703 2008-04-22  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2704
2705         Reviewed by Maciej.
2706
2707         Add support for function declarations in eval code.
2708
2709         (this fixes 12 more regression tests)
2710
2711         * VM/CodeBlock.h:
2712         * VM/CodeGenerator.cpp:
2713         (KJS::CodeGenerator::CodeGenerator):
2714         * VM/CodeGenerator.h:
2715         * VM/Machine.cpp:
2716         (KJS::Machine::execute):
2717         * kjs/nodes.cpp:
2718         (KJS::EvalNode::generateCode):
2719
2720 2008-04-22  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2721
2722         Reviewed by Oliver.
2723
2724         Implement LabelNode.
2725
2726         * VM/CodeGenerator.cpp:
2727         (KJS::CodeGenerator::pushJumpContext):
2728         (KJS::CodeGenerator::jumpContextForContinue):
2729         (KJS::CodeGenerator::jumpContextForBreak):
2730         * VM/CodeGenerator.h:
2731         * kjs/nodes.cpp:
2732         (KJS::DoWhileNode::emitCode):
2733         (KJS::WhileNode::emitCode):
2734         (KJS::ForNode::emitCode):
2735         (KJS::ForInNode::emitCode):
2736         (KJS::ContinueNode::emitCode):
2737         (KJS::BreakNode::emitCode):
2738         (KJS::SwitchNode::emitCode):
2739         (KJS::LabelNode::emitCode):
2740
2741 2008-04-22  Geoffrey Garen  <ggaren@apple.com>
2742
2743         Reviewed by Oliver Hunt.
2744
2745         Fixed crash when unwinding from exceptions inside eval.
2746
2747         * VM/Machine.cpp:
2748         (KJS::Machine::unwindCallFrame): Don't assume that the top of the
2749         current call frame's scope chain is an activation: it can be the global
2750         object, instead.
2751
2752 2008-04-22  Maciej Stachowiak  <mjs@apple.com>
2753
2754         Reviewed by Geoff.
2755
2756         * kjs/testkjs.cpp:
2757         (main): Convert signals to exit codes, so that crashing tests are
2758         detected as regression test failures.
2759
2760 2008-04-22  Geoffrey Garen  <ggaren@apple.com>
2761
2762         Reviewed by Oliver Hunt and Maciej Stachowiak.
2763
2764         Renamed "needsActivation" to "needsFullScopeChain" because lying will
2765         make hair grow on the backs of your hands.
2766
2767 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2768
2769         Reviewed by Maciej Stachowiak.
2770
2771         Fixed ScopeChainNode lifetime problems:
2772
2773         (1) In "with" and "catch" scopes, we would construct a ScopeChain
2774         object and then jump across its destructor, leaking the ScopeChainNode
2775         we had pushed.
2776
2777         (2) In global and eval scopes, we would fail to initially ref
2778         "scopeChain", causing us to overrelease it later. Now that we ref
2779         "scopeChain" properly, we also need to deref it when the script
2780         terminates.
2781
2782         SunSpider reports a .2% regression, but an earlier round of ScopeChain
2783         refactoring was a .4% speedup, so there.
2784
2785 2008-04-22  Maciej Stachowiak  <mjs@apple.com>
2786
2787         Reviewed by Alexey.
2788
2789         - use global object instead of null for "this" on unqualified calls
2790
2791         This fixes 10 more JSC test regressions.
2792
2793         * VM/Machine.cpp:
2794         (KJS::Machine::privateExecute):
2795
2796 2008-04-22  Maciej Stachowiak  <mjs@apple.com>
2797
2798         Reviewed by Oliver.
2799
2800         - throw proper exceptions for objects that don't implement call or construct
2801
2802         This fixes 21 more JSC test regressions. It is also seemingly an
2803         0.5% progression.
2804
2805         * VM/ExceptionHelpers.cpp:
2806         (KJS::createNotAnObjectError):
2807         (KJS::createNotAConstructorError):
2808         (KJS::createNotAFunctionError):
2809         * VM/ExceptionHelpers.h:
2810         * VM/Machine.cpp:
2811         (KJS::Machine::privateExecute):
2812
2813 2008-04-21  Oliver Hunt  <oliver@apple.com>
2814
2815         Reviewed by Geoff.
2816
2817         Implement emitCode for ConstDeclNode.
2818
2819         This fixes the crash (assertion) in js1_5/Scope/scope-001.js
2820
2821         * VM/CodeGenerator.cpp:
2822         (KJS::CodeGenerator::registerForLocalConstInit):
2823         * VM/CodeGenerator.h:
2824         * kjs/nodes.cpp:
2825         (KJS::AssignResolveNode::emitCode):
2826         (KJS::ConstDeclNode::emitCodeSingle):
2827         (KJS::ConstDeclNode::emitCode):
2828         (KJS::ConstStatementNode::emitCode):
2829         * kjs/nodes.h:
2830
2831 2008-04-21  Maciej Stachowiak  <mjs@apple.com>
2832
2833         Reviewed by Sam.
2834
2835         - add some support for the split window object
2836
2837         This fixes many layout tests.
2838
2839         * VM/Machine.cpp:
2840         (KJS::resolveBaseAndFunc): Use toThisObject() to ensure we get the
2841         wrapper global, if one exists, as the "this" object.
2842         * kjs/function.cpp:
2843         (KJS::globalFuncEval): Use toGlobalObject() to handle the wrapper
2844         case properly.
2845
2846 2008-04-21  Maciej Stachowiak  <mjs@apple.com>
2847
2848         Reviewed by Oliver.
2849
2850         - restore ScopeChain::operator= to avoid crash on many layout tests
2851
2852         Otherwise, FunctionImp::setScope would cause a reference
2853         underflow. I implemented using the copy construct and swap idiom.
2854
2855         * kjs/scope_chain.h:
2856         (KJS::ScopeChain::swap):
2857         (KJS::ScopeChain::operator=):
2858
2859 2008-04-21  Oliver Hunt  <oliver@apple.com>
2860
2861         Reviewed by Geoff.
2862
2863         Bug 18649: SQUIRRELFISH: correctly handle exceptions in eval code
2864         <https://bugs.webkit.org/show_bug.cgi?id=18649>
2865
2866         Allocate a callframe for eval() and initialise with a null codeBlock to
2867         indicate native code.  This prevents the unwinder from clobbering the
2868         register stack.
2869
2870         * VM/Machine.cpp:
2871         (KJS::Machine::execute):
2872
2873 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2874
2875         Reviewed by Sam Weinig.
2876
2877         Removed ScopeChain::push(ScopeChain&) because it was unused. Moved
2878         ScopeChain::print to ScopeChainNode.
2879
2880         ScopeChain is now nothing more than a resource-handling wrapper around
2881         ScopeChainNode.
2882
2883 2008-04-21  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2884
2885         Reviewed by Maciej.
2886
2887         Bug 18671: SquirrelFish: continue inside switch fails
2888         <https://bugs.webkit.org/show_bug.cgi?id=18671>
2889
2890         * VM/CodeGenerator.cpp:
2891         (KJS::CodeGenerator::jumpContextForLabel):
2892         * VM/CodeGenerator.h:
2893         * kjs/nodes.cpp:
2894         (KJS::ContinueNode::emitCode):
2895
2896 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2897
2898         Reviewed by Sam Weinig.
2899
2900         Moved push(JSObject*) and pop() from ScopeChain to ScopeChainNode,
2901         rearranging scope_chain.h a bit.
2902
2903         SunSpider reports no change.
2904
2905 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2906
2907         Reviewed by Sam Weinig.
2908
2909         Moved bottom() from ScopeChain to ScopeChainNode, simplifying it based
2910         on the knowledge that the ScopeChain is never empty.
2911
2912         SunSpider reports no change.
2913
2914 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2915
2916         Reviewed by Oliver Hunt.
2917
2918         Moved begin() and end() from ScopeChain to ScopeChainNode.
2919
2920         Also marked a few methods "const".
2921
2922         SunSpider reports no change.
2923
2924 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2925
2926         Reviewed by Maciej Stachowiak.
2927
2928         Turned ScopeChain::depth into a stand-alone function, and simplified it
2929         a bit.
2930
2931         I also moved ScopeChain::depth to Machine.cpp because it doesn't report
2932         the true depth of the ScopeChain -- just the Machine's perspective of
2933         its depth within a given call frame.
2934
2935         SunSpider reports no change.
2936
2937 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2938
2939         Reviewed by Maciej Stachowiak.
2940
2941         Removed indirection in ScopeChain::ref / ScopeChain::deref.
2942
2943         SunSpider reports no change.
2944
2945         * kjs/scope_chain.h:
2946         (KJS::ScopeChain::ScopeChain):
2947         (KJS::ScopeChain::~ScopeChain):
2948         (KJS::ScopeChain::clear):
2949
2950 2008-04-21  Oliver Hunt  <oliver@apple.com>
2951
2952         Reviewed by NOBODY(Build fix)
2953
2954         Fix debug build
2955
2956         * kjs/nodes.cpp:
2957         (KJS::ConstDeclNode::evaluateSingle):
2958
2959 2008-04-21  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
2960
2961         Reviewed by Oliver.
2962
2963         Bug 18664: SQUIRRELFISH: correctly throw a SyntaxError when parsing of eval code fails
2964         <https://bugs.webkit.org/show_bug.cgi?id=18664>
2965
2966         Correctly throw a SyntaxError when parsing of eval code fails.
2967
2968         * VM/Machine.cpp:
2969         (KJS::eval):
2970
2971 2008-04-21  Oliver Hunt  <oliver@apple.com>
2972
2973         Reviewed by Geoff.
2974
2975         Partial fix for Bug 18649: SQUIRRELFISH: correctly handle exceptions in eval code
2976
2977         Make sure we correct the register state before jumping to vm_throw.
2978
2979         * VM/Machine.cpp:
2980         (KJS::Machine::privateExecute):
2981
2982 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2983
2984         Reviewed by Maciej Stachowiak.
2985
2986         Simplified ScopeChain ref/deref.
2987
2988         SunSpider reports a .4% speedup.
2989
2990         * kjs/scope_chain.h:
2991         (KJS::ScopeChainNode::ref): Removed this function because it was nonsense.
2992         ScopeChainNodes are initialized with a refCount of 1, so the loop was
2993         guaranteed to iterate exactly once.
2994
2995 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
2996
2997         Reviewed by Maciej Stachowiak.
2998
2999         Removed support for empty ScopeChains.
3000
3001         SunSpider reports no change.
3002
3003 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
3004
3005         Reviewed by Maciej Stachowiak.
3006
3007         Removed some completely unused ScopeChain member functions.
3008
3009         SunSpider reports no change.
3010
3011 2008-04-21  Geoffrey Garen  <ggaren@apple.com>
3012
3013         Reviewed by Maciej Stachowiak.
3014
3015         Avoid creating unnecessary ScopeChain objects, to reduce refcount churn.
3016
3017         SunSpider reports no change.
3018
3019 2008-04-21  Maciej Stachowiak  <mjs@apple.com>
3020
3021         Rubber stamped by Alexey.
3022
3023         Add some braces.x
3024
3025         * kjs/testkjs.cpp:
3026         (runWithScripts):
3027
3028 2008-04-21  Maciej Stachowiak  <mjs@apple.com>
3029
3030         Reviewed by Oliver.
3031
3032         - only print "End:" output when -d flag is passed.
3033
3034         This fixes half of our failing JSC regression tests.
3035
3036         * kjs/testkjs.cpp:
3037         (runWithScripts):
3038
3039 2008-04-21  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
3040
3041         Reviewed by Maciej.
3042
3043         Add support for variable declarations in eval code.
3044
3045         * VM/CodeBlock.h:
3046         (KJS::EvalCodeBlock::EvalCodeBlock):
3047         * VM/CodeGenerator.cpp:
3048         (KJS::CodeGenerator::CodeGenerator):
3049         * VM/CodeGenerator.h:
3050         * VM/Machine.cpp:
3051         (KJS::Machine::execute):
3052         * VM/Machine.h:
3053         * kjs/function.cpp:
3054         (KJS::globalFuncEval):
3055         * kjs/nodes.cpp:
3056         (KJS::EvalNode::generateCode):
3057         * kjs/nodes.h:
3058         (KJS::EvalNode::):
3059
3060 2008-04-20  Oliver Hunt  <oliver@apple.com>
3061
3062         Reviewed by Maciej.
3063
3064         Throw exceptions for invalid continue, break, and return statements.
3065
3066         Simple refactoring and extension of Cameron's AssignErrorNode, etc patch
3067
3068         * VM/CodeGenerator.cpp:
3069         (KJS::CodeGenerator::CodeGenerator):
3070         (KJS::CodeGenerator::pushJumpContext):
3071         (KJS::CodeGenerator::popJumpContext):
3072         (KJS::CodeGenerator::jumpContextForLabel):
3073         * VM/CodeGenerator.h:
3074         * kjs/nodes.cpp:
3075         (KJS::Node::emitThrowError):
3076         (KJS::ContinueNode::emitCode):
3077         (KJS::BreakNode::emitCode):
3078         (KJS::ReturnNode::emitCode):
3079         * kjs/nodes.h:
3080
3081 2008-04-20  Geoffrey Garen  <ggaren@apple.com>
3082
3083         Reviewed by Oliver Hunt.
3084
3085         Removed Machine.cpp from AllInOneFile.cpp, and manually inlined a few
3086         things that used to be inlined automatically.
3087
3088         1.9% speedup on SunSpider.
3089
3090         My hope is that we'll face fewer surprises in Machine.cpp codegen, now
3091         that GCC is making fewer decisions. The speedup seems to confirm that.
3092
3093 2008-04-20  Oliver Hunt  <oliver@apple.com>
3094
3095         Reviewed by Maciej.
3096
3097         Bug 18642: Iterator context may get placed into the return register, leading to much badness
3098         <https://bugs.webkit.org/show_bug.cgi?id=18642>
3099
3100         To prevent incorrectly reusing what will become the result register for
3101         eval and global code execution, we need to request and ref the destination
3102         in advance of codegen.  Unfortunately this may lead to unnecessary copying,
3103         although in future we can probably limit this.  Curiously SunSpider shows
3104         a progression in a number of tests, although it comes out as a wash overall.
3105
3106         * kjs/nodes.cpp:
3107         (KJS::EvalNode::emitCode):
3108         (KJS::ProgramNode::emitCode):
3109
3110 2008-04-20  Cameron Zwarich  <cwzwarich@uwaterloo.ca>
3111
3112         Reviewed by Maciej.
3113
3114         Add support for AssignErrorNode, PrefixErrorNode, and PostfixErrorNode.
3115
3116         * VM/CodeBlock.cpp:
3117         (KJS::CodeBlock::dump):
3118         * VM/CodeGenerator.cpp:
3119         (KJS::CodeGenerator::emitCreateError):
3120         * VM/CodeGenerator.h:
3121         * VM/Machine.cpp:
3122         (KJS::Machine::privateExecute):
3123         * VM/Opcode.h:
3124         * kjs/nodes.cpp:
3125         (KJS::PostfixErrorNode::emitCode):
3126         (KJS::PrefixErrorNode::emitCode):
3127         (KJS::AssignErrorNode::emitCode):
3128         * kjs/nodes.h:
3129
3130 2008-04-20  Oliver Hunt  <oliver@apple.com>
3131
3132         Reviewed by Geoff and Mark.
3133
3134         Provide line number information in exceptions
3135
3136         Simple patch, adds line number information metadata to CodeBlock
3137         and a simple method to get the line number responsible for a given
3138         Instruction*.
3139
3140         * VM/CodeBlock.cpp:
3141         (KJS::CodeBlock::lineNumberForVPC):
3142         * VM/CodeBlock.h:
3143         * VM/CodeGenerator.h:
3144         (KJS::CodeGenerator::emitNode):
3145         * VM/Machine.cpp:
3146         (KJS::Machine::throwException):
3147
3148 2008-04-20  Oliver Hunt  <oliver@apple.com>
3149
3150         Reviewed by Maciej.
3151
3152         Provide "sourceURL" in exceptions
3153
3154         * VM/CodeBlock.h:
3155         * VM/Machine.cpp:
3156         (KJS::Machine::throwException):
3157         * kjs/nodes.cpp:
3158         (KJS::EvalNode::generateCode):
3159         (KJS::ProgramNode::generateCode):
3160
3161 2008-04-19  Oliver Hunt  <oliver@apple.com>
3162
3163         Reviewed by Maciej.
3164
3165         Don't call emitCode directly on subnodes, instead use CodeGenerator::emitNode
3166
3167         This patch just a preparation for tracking line numbers.
3168
3169         * kjs/nodes.cpp:
3170         (KJS::ObjectLiteralNode::emitCode):
3171         (KJS::PropertyListNode::emitCode):
3172         (KJS::ArgumentListNode::emitCode):
3173         (KJS::TryNode::emitCode):
3174
3175 2008-04-19  Oliver Hunt  <oliver@apple.com>
3176
3177         Reviewed by Maciej.
3178
3179         Bug 18619: Support continue, break, and return in try .. finally blocks
3180         <https://bugs.webkit.org/show_bug.cgi?id=18619>
3181
3182         This patch replaces the current partial finally support (which uses code
3183         duplication to achieve what it does) with a subroutine based approach.
3184         This has a number of advantages over code duplication:
3185           * Reduced code size
3186           * Simplified exception handling as the finaliser code only exists in
3187             one place, so no "magic" is needed to get the correct handler for a
3188             finaliser.
3189           * When we support instruction to line number mapping we won't need to
3190             worry about the dramatic code movement caused by duplication
3191
3192         On the downside it is necessary to add two new opcodes, op_jsr and op_sret
3193         to enter and exit the finaliser subroutines, happily SunSpider reports
3194         a performance progression (gcc amazes me) and ubench reports a wash.
3195
3196         While jsr and sret provide a mechanism that allows us to enter and exit
3197         any arbitrary finaliser we need to, it was still necessary to increase
3198         the amount of information tracked when entering and exiting both finaliser
3199         scopes and dynamic scopes ("with").  This means "scopeDepth" is now
3200         the combination of "finaliserDepth" and "dynamicScopeDepth".  We also
3201         now use a scopeContextStack to ensure that we pop scopes and execute
3202         finalisers in the correct order.  This increases the cost of "with" nodes
3203         during codegen, but it should not be significant enough to effect real
3204         world performance and greatly simplifies codegen for return, break and
3205         continue when interacting with finalisers.
3206
3207         * VM/CodeBlock.cpp:
3208         (KJS::CodeBlock::dump):
3209           Pretty printing of jsr/sret opcodes
3210
3211         * VM/CodeGenerator.cpp:
3212         (KJS::CodeGenerator::CodeGenerator):
3213         (KJS::CodeGenerator::emitPushScope):
3214         (KJS::CodeGenerator::emitPopScope):
3215           Dynamic scopes need to be tracked on the scopeContextStack now
3216
3217         (KJS::CodeGenerator::pushFinallyContext):
3218         (KJS::CodeGenerator::popFinallyContext):
3219           Handle entry and exit from code regions with finalisers.  This is
3220           needed solely to support return, continue and break inside finaliser
3221           regions.
3222
3223         (KJS::CodeGenerator::emitComplexJumpScopes):
3224           Helper function for emitJumpScopes to handle the complex codegen
3225           needed to handle return, continue and break inside a finaliser region
3226
3227         (KJS::CodeGenerator::emitJumpScopes):
3228           Updated to be aware of finalisers, if a cross-scope jump occurs inside
3229           a finaliser we hand off codegen to emitComplexJumpScopes, otherwise
3230           we can handle the normal (trivial) case with a single instruction.
3231
3232         (KJS::CodeGenerator::emitJumpSubroutine):
3233         (KJS::CodeGenerator::emitSubroutineReturn):
3234           Trivial opcode emitter functions.
3235
3236         * VM/CodeGenerator.h:
3237         (KJS::CodeGenerator::scopeDepth):
3238         * VM/Machine.cpp:
3239         (KJS::Machine::privateExecute):
3240           Implement op_jsr and op_sret.
3241
3242         * VM/Opcode.h:
3243           Ad op_jsr and op_sret
3244
3245         * kjs/nodes.cpp:
3246         (KJS::TryNode::emitCode):
3247           Fix codegen for new finaliser model.
3248
3249 2008-04-17  Mark Rowe  <mrowe@apple.com>
3250
3251         Rubber-stamped by Oliver Hunt.
3252
3253         Remove unnecessary files from testkjs, testapi and minidom targets.
3254
3255         * JavaScriptCore.xcodeproj/project.pbxproj:
3256
3257 2008-04-17  Geoffrey Garen  <ggaren@apple.com>
3258
3259         Reviewed by Oliver Hunt.
3260
3261         Fixed ASSERT seen during run-sunspider of a debug build.
3262
3263         * VM/CodeGenerator.h: Made the default codegen buffers bigger. SunSpider
3264         runs all tests in one global environment, so you end up with more than
3265         128 locals. This is just a stop-gap until we code up a real
3266         solution to arbitrary symbol and label limits.
3267
3268 2008-04-17  Geoffrey Garen  <ggaren@apple.com>
3269
3270         Reviewed by Oliver Hunt.
3271
3272         Fixed a bug in exception unwinding, where we wouldn't deref the scope
3273         chain in global scope, so we would leak ScopeChainNodes when exceptions
3274         were thrown inside "with" and "catch" scopes.
3275
3276         Also did some cleanup of the unwinding code along the way.
3277
3278         Scope chain reference counting is still wrong in a few ways. I thought
3279         I would fix this portion of it first.
3280
3281         run-sunspider shows no change.
3282
3283         * VM/Machine.cpp:
3284         (KJS::Machine::unwindCallFrame):
3285         (KJS::Machine::throwException):
3286         (KJS::Machine::privateExecute):
3287         * VM/Machine.h:
3288
3289 2008-04-17  Oliver Hunt  <oliver@apple.com>
3290
3291         Reviewed by Maciej.
3292
3293         Add more exception checking to toNumber conversions
3294
3295         This corrects op_pre_dec, op_negate, op_mod and op_sub.
3296
3297         * VM/Machine.cpp:
3298         (KJS::Machine::privateExecute):
3299
3300 2008-04-17  Geoffrey Garen  <ggaren@apple.com> and Cameron Zwarich  <cwzwarich@uwaterloo.ca>
3301
3302         Reviewed by Oliver Hunt.
3303
3304         Behold: eval.
3305
3306         Introduced a new opcode: op_call_eval. In the normal case, it performs
3307         an eval. In the case where eval has been overridden in some way, it
3308         performs a function call.
3309
3310         * VM/CodeGenerator.h: Added a feature so the code generator knows not
3311         to optimized locals in eval code.
3312
3313 2008-04-17  Geoffrey Garen  <ggaren@apple.com>
3314
3315         Reviewed by Sam Weinig.
3316
3317         Added some ASSERTs to document codegen failures in
3318         run-javascriptcore-tests.
3319
3320         For all tests, program-level codegen now either succeeds, or fails with
3321         an ASSERT.
3322
3323         * VM/CodeGenerator.cpp:
3324         (KJS::CodeGenerator::addVar):
3325         (KJS::CodeGenerator::CodeGenerator):
3326         (KJS::CodeGenerator::newTemporary):
3327         (KJS::CodeGenerator::newLabel):
3328
3329 2008-04-17  Geoffrey Garen  <ggaren@apple.com>
3330
3331         Reviewed by Maciej Stachowiak.
3332
3333         Fixed another case of a dst register being an unreferenced temporary
3334         (caused an ASSERT when running the full sunspider suite).
3335
3336         * kjs/nodes.cpp:
3337         (KJS::CaseBlockNode::emitCodeForBlock):
3338
3339 2008-04-16  Maciej Stachowiak  <mjs@apple.com>
3340
3341         Reviewed by Geoff.
3342
3343         - add documentation (and meaningful parameter names) for arithmetic and bitwise binary ops
3344
3345         * VM/CodeBlock.cpp:
3346         (KJS::CodeBlock::dump):
3347         * VM/CodeGenerator.cpp:
3348         (KJS::CodeGenerator::emitMul):
3349         (KJS::CodeGenerator::emitDiv):
3350         (KJS::CodeGenerator::emitMod):
3351         (KJS::CodeGenerator::emitSub):
3352         (KJS::CodeGenerator::emitLeftShift):
3353         (KJS::CodeGenerator::emitRightShift):
3354         (KJS::CodeGenerator::emitUnsignedRightShift):
3355         (KJS::CodeGenerator::emitBitAnd):
3356         (KJS::CodeGenerator::emitBitXOr):
3357         (KJS::CodeGenerator::emitBitOr):
3358         * VM/CodeGenerator.h:
3359         * VM/Machine.cpp:
3360         (KJS::Machine::privateExecute):
3361         * VM/Opcode.h:
3362         * kjs/nodes.cpp:
3363         (KJS::MultNode::emitCode):
3364         (KJS::DivNode::emitCode):
3365         (KJS::ModNode::emitCode):
3366         (KJS::SubNode::emitCode):
3367         (KJS::LeftShiftNode::emitCode):
3368         (KJS::RightShiftNode::emitCode):
3369         (KJS::UnsignedRightShiftNode::emitCode):
3370         (KJS::BitAndNode::emitCode):
3371         (KJS::BitXOrNode::emitCode):
3372         (KJS::BitOrNode::emitCode):
3373         (KJS::emitReadModifyAssignment):
3374         (KJS::ReadModifyResolveNode::emitCode):
3375
3376 2008-04-16  Oliver Hunt  <oliver@apple.com>
3377
3378         Reviewed by Geoff.
3379
3380         Exception checks for toNumber in op_pre_inc
3381
3382         This is somewhat more convoluted than the simple hadException checks
3383         we currently use.  Instead we use special toNumber conversions that
3384         select between the exception and ordinary vPC.  This allows us to
3385         remove any branches in the common case (incrementing a number).
3386
3387         * API/JSCallbackObject.h:
3388         * API/JSCallbackObjectFunctions.h:
3389         (KJS::::toNumber):
3390         * ChangeLog:
3391         * JavaScriptCore.exp:
3392         * JavaScriptCore.xcodeproj/project.pbxproj:
3393         * VM/JSPropertyNameIterator.cpp:
3394         (KJS::JSPropertyNameIterator::toNumber):
3395         * VM/JSPropertyNameIterator.h:
3396         * VM/Machine.cpp:
3397         (KJS::Machine::privateExecute):
3398         * VM/Opcode.h:
3399         * kjs/ExecState.cpp:
3400         (KJS::ExecState::ExecState):
3401         * kjs/ExecState.h:
3402         * kjs/JSNotAnObject.cpp:
3403         (KJS::JSNotAnObject::toNumber):
3404         * kjs/JSNotAnObject.h:
3405         * kjs/internal.cpp:
3406         (KJS::StringImp::toNumber):
3407         (KJS::NumberImp::toNumber):
3408         (KJS::GetterSetterImp::toNumber):
3409         * kjs/internal.h:
3410         * kjs/object.cpp:
3411         (KJS::JSObject::toNumber):
3412         * kjs/object.h:
3413         * kjs/value.h:
3414         (KJS::JSValue::toNumber):
3415
3416 2008-04-16  Maciej Stachowiak  <mjs@apple.com>
3417
3418         Reviewed by Geoff.
3419
3420         - ensure that activations are kept in a register to protect them from GC
3421
3422         Also renamed OptionalCalleeScopeChain constant to OptionalCalleeActivation, since
3423         that is what is now kept there, and there is no more need to keep the scope chain in
3424         the register file.
3425
3426         * VM/Machine.cpp:
3427         (KJS::initializeCallFrame):
3428         (KJS::scopeChainForCall):
3429         * VM/Machine.h:
3430         (KJS::Machine::):
3431
3432 2008-04-16  Geoffrey Garen  <ggaren@apple.com>
3433
3434         Reviewed by Oliver Hunt.
3435
3436         Made "this" work in program code / global scope.
3437
3438         The machine can initialize "this" prior to execution because it knows
3439         that, for program code, "this" is always stored in lr1.
3440
3441         * VM/Machine.cpp:
3442         (KJS::Machine::execute):
3443         * VM/Machine.h:
3444         (KJS::Machine::):
3445         * kjs/interpreter.cpp:
3446         (KJS::Interpreter::evaluate):
3447
3448 2008-04-16  Geoffrey Garen  <ggaren@apple.com>
3449
3450         Reviewed by Oliver Hunt.
3451
3452         Fixed a codegen bug when returning from inside a dynamic scope (a with
3453         or catch block): we need to pop any dynamic scope(s) that have been
3454         added so op_ret can find the activation object at the top of the scope
3455         chain.
3456
3457         * kjs/nodes.cpp:
3458         (KJS::ReturnNode::emitCode): If we're returning from inside a dynamic
3459         scope, emit a jmp_scopes to take care of popping any dynamic scope(s)
3460         and then branching to the return instruction.
3461
3462 2008-04-16  Maciej Stachowiak  <mjs@apple.com>
3463
3464         Reviewed by Geoff.
3465
3466         - document the add and get_prop_id opcodes
3467
3468         In addition to adding documentation in comments, I changed
3469         references to register IDs or indices relating to these opcodes to
3470         have meaningful names instead of r0 r1 r2.
3471
3472         * VM/CodeGenerator.cpp:
3473         (KJS::CodeGenerator::emitAdd):
3474         * VM/CodeGenerator.h:
3475         * VM/Machine.cpp:
3476         (KJS::Machine::privateExecute):
3477         * kjs/nodes.cpp:
3478         (KJS::DotAccessorNode::emitCode):
3479         (KJS::FunctionCallDotNode::emitCode):
3480         (KJS::PostIncDotNode::emitCode):
3481         (KJS::PostDecDotNode::emitCode):
3482         (KJS::PreIncDotNode::emitCode):
3483         (KJS::PreDecDotNode::emitCode):
3484         (KJS::AddNode::emitCode):
3485         (KJS::ReadModifyDotNode::emitCode):
3486
3487 2008-04-15  Geoffrey Garen  <ggaren@apple.com>
3488
3489         Reviewed by Oliver Hunt and Maciej Stachowiak.
3490
3491         Fixed a codegen bug in with and switch, and added an ASSERT to
3492         make sure it doesn't happen again.
3493
3494         emitCode() assumes that dst, if non-zero, is either referenced or
3495         non-temporary (i.e., it assumes that newTemporary() will return a
3496         register not equal to dst). Certain callers to emitCode() weren't
3497         guaranteeing that to be so, so temporary register values were being
3498         overwritten.
3499
3500         * VM/CodeGenerator.h:
3501         (KJS::CodeGenerator::emitNode): ASSERT that dst is referenced or non-temporary.
3502
3503         * kjs/nodes.cpp:
3504         (KJS::CommaNode::emitCode): Reference the dst we pass.
3505
3506         (KJS::WithNode::emitCode): No need to pass an explicit dst register.
3507
3508         (KJS::CaseBlockNode::emitCodeForBlock): No need to pass an explicit dst register.
3509         (KJS::SwitchNode::emitCode): No need to pass an explicit dst register.
3510
3511         * kjs/nodes.h: Made dst the last parameter to emitCodeForBlock, to match
3512         emitCode.
3513
3514 2008-04-15  Oliver Hunt  <oliver@apple.com>
3515
3516         Reviewed by Maciej.
3517
3518         Bug 18526: Throw exceptions when resolve fails for op_resolve_base_and_func.
3519         <https://bugs.webkit.org/show_bug.cgi?id=18526>
3520
3521         Very simple fix, sunspider shows a 0.7% progression, ubench shows a 0.4% regression.
3522
3523         * VM/Machine.cpp:
3524         (KJS::resolveBaseAndFunc):
3525         (KJS::Machine::privateExecute):
3526
3527 2008-04-15  Maciej Stachowiak  <mjs@apple.com>
3528
3529         Reviewed by Oliver.
3530
3531         - fix incorrect result on 3d-raytrace test
3532
3533         Oliver found and tracked down this bug, I just typed in the fix.
3534
3535         * VM/Machine.cpp:
3536         (KJS::slideRegisterWindowForCall): When setting omitted parameters to undefined,
3537         account for the space for local variables.
3538
3539 2008-04-15  Maciej Stachowiak  <mjs@apple.com>
3540
3541         Reviewed by Oliver.
3542
3543         - fix codegen handling of dst registers
3544
3545         1.006x speedup (not sure why).
3546
3547         Most emitCode functions take an optional "dst" parameter that says
3548         where the output of the instruction should be written. I made some
3549         functions for convenient handling of the dst register:
3550
3551         * VM/CodeGenerator.h:
3552         (KJS::CodeGenerator::tempDestination): Takes the dst register. Returns it if
3553         it is not null and is a temporary, otherwise allocates a new temporary. This is
3554         intended for cases where an intermediate value might be written into the dst
3555
3556         (KJS::CodeGenerator::finalDestination): Takes the dst register and an optional
3557         register that was used as a temp destination. Picks the right thing for the final
3558         output. Intended to be used as the output register for the instruction that generates
3559         the final value of a particular node.
3560
3561         (KJS::CodeGenerator::moveToDestinationIfNeeded): Takes dst and a
3562         RegisterID; moves from the register to dst if dst is defined and