1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This class implements the parser for assembly files. 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "llvm/ADT/APFloat.h" 14 #include "llvm/ADT/APInt.h" 15 #include "llvm/ADT/ArrayRef.h" 16 #include "llvm/ADT/None.h" 17 #include "llvm/ADT/STLExtras.h" 18 #include "llvm/ADT/SmallString.h" 19 #include "llvm/ADT/SmallVector.h" 20 #include "llvm/ADT/StringExtras.h" 21 #include "llvm/ADT/StringMap.h" 22 #include "llvm/ADT/StringRef.h" 23 #include "llvm/ADT/Twine.h" 24 #include "llvm/BinaryFormat/Dwarf.h" 25 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 26 #include "llvm/MC/MCAsmInfo.h" 27 #include "llvm/MC/MCCodeView.h" 28 #include "llvm/MC/MCContext.h" 29 #include "llvm/MC/MCDirectives.h" 30 #include "llvm/MC/MCDwarf.h" 31 #include "llvm/MC/MCExpr.h" 32 #include "llvm/MC/MCInstPrinter.h" 33 #include "llvm/MC/MCInstrDesc.h" 34 #include "llvm/MC/MCInstrInfo.h" 35 #include "llvm/MC/MCObjectFileInfo.h" 36 #include "llvm/MC/MCParser/AsmCond.h" 37 #include "llvm/MC/MCParser/AsmLexer.h" 38 #include "llvm/MC/MCParser/MCAsmLexer.h" 39 #include "llvm/MC/MCParser/MCAsmParser.h" 40 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 41 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 42 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 43 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 44 #include "llvm/MC/MCRegisterInfo.h" 45 #include "llvm/MC/MCSection.h" 46 #include "llvm/MC/MCStreamer.h" 47 #include "llvm/MC/MCSymbol.h" 48 #include "llvm/MC/MCTargetOptions.h" 49 #include "llvm/MC/MCValue.h" 50 #include "llvm/Support/Casting.h" 51 #include "llvm/Support/CommandLine.h" 52 #include "llvm/Support/ErrorHandling.h" 53 #include "llvm/Support/MD5.h" 54 #include "llvm/Support/MathExtras.h" 55 #include "llvm/Support/MemoryBuffer.h" 56 #include "llvm/Support/SMLoc.h" 57 #include "llvm/Support/SourceMgr.h" 58 #include "llvm/Support/raw_ostream.h" 59 #include <algorithm> 60 #include <cassert> 61 #include <cctype> 62 #include <climits> 63 #include <cstddef> 64 #include <cstdint> 65 #include <deque> 66 #include <memory> 67 #include <sstream> 68 #include <string> 69 #include <tuple> 70 #include <utility> 71 #include <vector> 72 73 using namespace llvm; 74 75 extern cl::opt<unsigned> AsmMacroMaxNestingDepth; 76 77 namespace { 78 79 /// Helper types for tracking macro definitions. 80 typedef std::vector<AsmToken> MCAsmMacroArgument; 81 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 82 83 /// Helper class for storing information about an active macro instantiation. 84 struct MacroInstantiation { 85 /// The location of the instantiation. 86 SMLoc InstantiationLoc; 87 88 /// The buffer where parsing should resume upon instantiation completion. 89 unsigned ExitBuffer; 90 91 /// The location where parsing should resume upon instantiation completion. 92 SMLoc ExitLoc; 93 94 /// The depth of TheCondStack at the start of the instantiation. 95 size_t CondStackDepth; 96 }; 97 98 struct ParseStatementInfo { 99 /// The parsed operands from the last parsed statement. 100 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 101 102 /// The opcode from the last parsed instruction. 103 unsigned Opcode = ~0U; 104 105 /// Was there an error parsing the inline assembly? 106 bool ParseError = false; 107 108 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 109 110 ParseStatementInfo() = delete; 111 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 112 : AsmRewrites(rewrites) {} 113 }; 114 115 enum FieldType { 116 FT_INTEGRAL, // Initializer: integer expression, stored as an MCExpr. 117 FT_REAL, // Initializer: real number, stored as an APInt. 118 FT_STRUCT // Initializer: struct initializer, stored recursively. 119 }; 120 121 struct FieldInfo; 122 struct StructInfo { 123 StringRef Name; 124 bool IsUnion = false; 125 size_t Alignment = 0; 126 size_t Size = 0; 127 std::vector<FieldInfo> Fields; 128 StringMap<size_t> FieldsByName; 129 130 FieldInfo &addField(StringRef FieldName, FieldType FT); 131 132 StructInfo() = default; 133 134 StructInfo(StringRef StructName, bool Union, unsigned AlignmentValue) 135 : Name(StructName), IsUnion(Union), Alignment(AlignmentValue) {} 136 }; 137 138 // FIXME: This should probably use a class hierarchy, raw pointers between the 139 // objects, and dynamic type resolution instead of a union. On the other hand, 140 // ownership then becomes much more complicated; the obvious thing would be to 141 // use BumpPtrAllocator, but the lack of a destructor makes that messy. 142 143 struct StructInitializer; 144 struct IntFieldInfo { 145 SmallVector<const MCExpr *, 1> Values; 146 147 IntFieldInfo() = default; 148 IntFieldInfo(const SmallVector<const MCExpr *, 1> &V) { Values = V; } 149 IntFieldInfo(SmallVector<const MCExpr *, 1> &&V) { Values = V; } 150 }; 151 struct RealFieldInfo { 152 SmallVector<APInt, 1> AsIntValues; 153 154 RealFieldInfo() = default; 155 RealFieldInfo(const SmallVector<APInt, 1> &V) { AsIntValues = V; } 156 RealFieldInfo(SmallVector<APInt, 1> &&V) { AsIntValues = V; } 157 }; 158 struct StructFieldInfo { 159 std::vector<StructInitializer> Initializers; 160 StructInfo Structure; 161 162 StructFieldInfo() = default; 163 StructFieldInfo(const std::vector<StructInitializer> &V, StructInfo S) { 164 Initializers = V; 165 Structure = S; 166 } 167 StructFieldInfo(std::vector<StructInitializer> &&V, StructInfo S) { 168 Initializers = V; 169 Structure = S; 170 } 171 }; 172 173 class FieldInitializer { 174 public: 175 FieldType FT; 176 union { 177 IntFieldInfo IntInfo; 178 RealFieldInfo RealInfo; 179 StructFieldInfo StructInfo; 180 }; 181 182 ~FieldInitializer() { 183 switch (FT) { 184 case FT_INTEGRAL: 185 IntInfo.~IntFieldInfo(); 186 break; 187 case FT_REAL: 188 RealInfo.~RealFieldInfo(); 189 break; 190 case FT_STRUCT: 191 StructInfo.~StructFieldInfo(); 192 break; 193 } 194 } 195 196 FieldInitializer(FieldType FT) : FT(FT) { 197 switch (FT) { 198 case FT_INTEGRAL: 199 new (&IntInfo) IntFieldInfo(); 200 break; 201 case FT_REAL: 202 new (&RealInfo) RealFieldInfo(); 203 break; 204 case FT_STRUCT: 205 new (&StructInfo) StructFieldInfo(); 206 break; 207 } 208 } 209 210 FieldInitializer(SmallVector<const MCExpr *, 1> &&Values) : FT(FT_INTEGRAL) { 211 new (&IntInfo) IntFieldInfo(Values); 212 } 213 214 FieldInitializer(SmallVector<APInt, 1> &&AsIntValues) : FT(FT_REAL) { 215 new (&RealInfo) RealFieldInfo(AsIntValues); 216 } 217 218 FieldInitializer(std::vector<StructInitializer> &&Initializers, 219 struct StructInfo Structure) 220 : FT(FT_STRUCT) { 221 new (&StructInfo) StructFieldInfo(Initializers, Structure); 222 } 223 224 FieldInitializer(const FieldInitializer &Initializer) : FT(Initializer.FT) { 225 switch (FT) { 226 case FT_INTEGRAL: 227 new (&IntInfo) IntFieldInfo(Initializer.IntInfo); 228 break; 229 case FT_REAL: 230 new (&RealInfo) RealFieldInfo(Initializer.RealInfo); 231 break; 232 case FT_STRUCT: 233 new (&StructInfo) StructFieldInfo(Initializer.StructInfo); 234 break; 235 } 236 } 237 238 FieldInitializer(FieldInitializer &&Initializer) : FT(Initializer.FT) { 239 switch (FT) { 240 case FT_INTEGRAL: 241 new (&IntInfo) IntFieldInfo(Initializer.IntInfo); 242 break; 243 case FT_REAL: 244 new (&RealInfo) RealFieldInfo(Initializer.RealInfo); 245 break; 246 case FT_STRUCT: 247 new (&StructInfo) StructFieldInfo(Initializer.StructInfo); 248 break; 249 } 250 } 251 252 FieldInitializer &operator=(const FieldInitializer &Initializer) { 253 if (FT != Initializer.FT) { 254 switch (FT) { 255 case FT_INTEGRAL: 256 IntInfo.~IntFieldInfo(); 257 break; 258 case FT_REAL: 259 RealInfo.~RealFieldInfo(); 260 break; 261 case FT_STRUCT: 262 StructInfo.~StructFieldInfo(); 263 break; 264 } 265 } 266 FT = Initializer.FT; 267 switch (FT) { 268 case FT_INTEGRAL: 269 IntInfo = Initializer.IntInfo; 270 break; 271 case FT_REAL: 272 RealInfo = Initializer.RealInfo; 273 break; 274 case FT_STRUCT: 275 StructInfo = Initializer.StructInfo; 276 break; 277 } 278 return *this; 279 } 280 281 FieldInitializer &operator=(FieldInitializer &&Initializer) { 282 if (FT != Initializer.FT) { 283 switch (FT) { 284 case FT_INTEGRAL: 285 IntInfo.~IntFieldInfo(); 286 break; 287 case FT_REAL: 288 RealInfo.~RealFieldInfo(); 289 break; 290 case FT_STRUCT: 291 StructInfo.~StructFieldInfo(); 292 break; 293 } 294 } 295 FT = Initializer.FT; 296 switch (FT) { 297 case FT_INTEGRAL: 298 IntInfo = Initializer.IntInfo; 299 break; 300 case FT_REAL: 301 RealInfo = Initializer.RealInfo; 302 break; 303 case FT_STRUCT: 304 StructInfo = Initializer.StructInfo; 305 break; 306 } 307 return *this; 308 } 309 }; 310 311 struct StructInitializer { 312 std::vector<FieldInitializer> FieldInitializers; 313 }; 314 315 struct FieldInfo { 316 // Offset of the field within the containing STRUCT. 317 size_t Offset = 0; 318 319 // Total size of the field (= LengthOf * Type). 320 size_t SizeOf = 0; 321 322 // Number of elements in the field (1 if scalar, >1 if an array). 323 size_t LengthOf = 0; 324 325 // Size of a single entry in this field, in bytes ("type" in MASM standards). 326 size_t Type = 0; 327 328 FieldInitializer Contents; 329 330 FieldInfo(FieldType FT) : Contents(FT) {} 331 }; 332 333 FieldInfo &StructInfo::addField(StringRef FieldName, FieldType FT) { 334 if (!FieldName.empty()) 335 FieldsByName[FieldName] = Fields.size(); 336 Fields.emplace_back(FT); 337 FieldInfo &Field = Fields.back(); 338 if (IsUnion) { 339 Field.Offset = 0; 340 } else { 341 Size = llvm::alignTo(Size, Alignment); 342 Field.Offset = Size; 343 } 344 return Field; 345 } 346 347 /// The concrete assembly parser instance. 348 // Note that this is a full MCAsmParser, not an MCAsmParserExtension! 349 // It's a peer of AsmParser, not of COFFAsmParser, WasmAsmParser, etc. 350 class MasmParser : public MCAsmParser { 351 private: 352 AsmLexer Lexer; 353 MCContext &Ctx; 354 MCStreamer &Out; 355 const MCAsmInfo &MAI; 356 SourceMgr &SrcMgr; 357 SourceMgr::DiagHandlerTy SavedDiagHandler; 358 void *SavedDiagContext; 359 std::unique_ptr<MCAsmParserExtension> PlatformParser; 360 361 /// This is the current buffer index we're lexing from as managed by the 362 /// SourceMgr object. 363 unsigned CurBuffer; 364 365 AsmCond TheCondState; 366 std::vector<AsmCond> TheCondStack; 367 368 /// maps directive names to handler methods in parser 369 /// extensions. Extensions register themselves in this map by calling 370 /// addDirectiveHandler. 371 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 372 373 /// maps assembly-time variable names to variables. 374 struct Variable { 375 StringRef Name; 376 bool Redefinable = true; 377 bool IsText = false; 378 int64_t NumericValue = 0; 379 std::string TextValue; 380 }; 381 StringMap<Variable> Variables; 382 383 /// Stack of active struct definitions. 384 SmallVector<StructInfo, 1> StructInProgress; 385 386 /// Maps struct tags to struct definitions. 387 StringMap<StructInfo> Structs; 388 389 /// Maps data location names to user-defined types. 390 StringMap<const StructInfo *> KnownType; 391 392 /// Stack of active macro instantiations. 393 std::vector<MacroInstantiation*> ActiveMacros; 394 395 /// List of bodies of anonymous macros. 396 std::deque<MCAsmMacro> MacroLikeBodies; 397 398 /// Keeps track of how many .macro's have been instantiated. 399 unsigned NumOfMacroInstantiations; 400 401 /// The values from the last parsed cpp hash file line comment if any. 402 struct CppHashInfoTy { 403 StringRef Filename; 404 int64_t LineNumber; 405 SMLoc Loc; 406 unsigned Buf; 407 CppHashInfoTy() : Filename(), LineNumber(0), Loc(), Buf(0) {} 408 }; 409 CppHashInfoTy CppHashInfo; 410 411 /// The filename from the first cpp hash file line comment, if any. 412 StringRef FirstCppHashFilename; 413 414 /// List of forward directional labels for diagnosis at the end. 415 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 416 417 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 418 /// Defaults to 1U, meaning Intel. 419 unsigned AssemblerDialect = 1U; 420 421 /// is Darwin compatibility enabled? 422 bool IsDarwin = false; 423 424 /// Are we parsing ms-style inline assembly? 425 bool ParsingMSInlineAsm = false; 426 427 /// Did we already inform the user about inconsistent MD5 usage? 428 bool ReportedInconsistentMD5 = false; 429 430 // Is alt macro mode enabled. 431 bool AltMacroMode = false; 432 433 // Current <...> expression depth. 434 unsigned AngleBracketDepth = 0U; 435 436 public: 437 MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 438 const MCAsmInfo &MAI, unsigned CB); 439 MasmParser(const MasmParser &) = delete; 440 MasmParser &operator=(const MasmParser &) = delete; 441 ~MasmParser() override; 442 443 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 444 445 void addDirectiveHandler(StringRef Directive, 446 ExtensionDirectiveHandler Handler) override { 447 ExtensionDirectiveMap[Directive] = Handler; 448 if (DirectiveKindMap.find(Directive) == DirectiveKindMap.end()) { 449 DirectiveKindMap[Directive] = DK_HANDLER_DIRECTIVE; 450 } 451 } 452 453 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 454 DirectiveKindMap[Directive] = DirectiveKindMap[Alias]; 455 } 456 457 /// @name MCAsmParser Interface 458 /// { 459 460 SourceMgr &getSourceManager() override { return SrcMgr; } 461 MCAsmLexer &getLexer() override { return Lexer; } 462 MCContext &getContext() override { return Ctx; } 463 MCStreamer &getStreamer() override { return Out; } 464 465 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 466 467 unsigned getAssemblerDialect() override { 468 if (AssemblerDialect == ~0U) 469 return MAI.getAssemblerDialect(); 470 else 471 return AssemblerDialect; 472 } 473 void setAssemblerDialect(unsigned i) override { 474 AssemblerDialect = i; 475 } 476 477 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 478 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 479 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 480 481 const AsmToken &Lex() override; 482 483 void setParsingMSInlineAsm(bool V) override { 484 ParsingMSInlineAsm = V; 485 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 486 // hex integer literals. 487 Lexer.setLexMasmIntegers(V); 488 } 489 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; } 490 491 bool isParsingMasm() const override { return true; } 492 493 bool lookUpField(StringRef Name, StringRef &Type, 494 unsigned &Offset) const override; 495 bool lookUpField(StringRef Base, StringRef Member, StringRef &Type, 496 unsigned &Offset) const override; 497 498 bool parseMSInlineAsm(void *AsmLoc, std::string &AsmString, 499 unsigned &NumOutputs, unsigned &NumInputs, 500 SmallVectorImpl<std::pair<void *,bool>> &OpDecls, 501 SmallVectorImpl<std::string> &Constraints, 502 SmallVectorImpl<std::string> &Clobbers, 503 const MCInstrInfo *MII, const MCInstPrinter *IP, 504 MCAsmParserSemaCallback &SI) override; 505 506 bool parseExpression(const MCExpr *&Res); 507 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 508 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) override; 509 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 510 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 511 SMLoc &EndLoc) override; 512 bool parseAbsoluteExpression(int64_t &Res) override; 513 514 /// Parse a floating point expression using the float \p Semantics 515 /// and set \p Res to the value. 516 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 517 518 /// Parse an identifier or string (as a quoted identifier) 519 /// and set \p Res to the identifier contents. 520 bool parseIdentifier(StringRef &Res) override; 521 void eatToEndOfStatement() override; 522 523 bool checkForValidSection() override; 524 525 /// } 526 527 private: 528 bool parseStatement(ParseStatementInfo &Info, 529 MCAsmParserSemaCallback *SI); 530 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 531 bool parseCppHashLineFilenameComment(SMLoc L); 532 533 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 534 ArrayRef<MCAsmMacroParameter> Parameters); 535 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 536 ArrayRef<MCAsmMacroParameter> Parameters, 537 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 538 SMLoc L); 539 540 /// Are we inside a macro instantiation? 541 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 542 543 /// Handle entry to macro instantiation. 544 /// 545 /// \param M The macro. 546 /// \param NameLoc Instantiation location. 547 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 548 549 /// Handle exit from macro instantiation. 550 void handleMacroExit(); 551 552 /// Extract AsmTokens for a macro argument. 553 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 554 555 /// Parse all macro arguments for a given macro. 556 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 557 558 void printMacroInstantiations(); 559 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 560 SMRange Range = None) const { 561 ArrayRef<SMRange> Ranges(Range); 562 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 563 } 564 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 565 566 bool lookUpField(const StructInfo &Structure, StringRef Member, 567 StringRef &Type, unsigned &Offset) const; 568 569 /// Should we emit DWARF describing this assembler source? (Returns false if 570 /// the source has .file directives, which means we don't want to generate 571 /// info describing the assembler source itself.) 572 bool enabledGenDwarfForAssembly(); 573 574 /// Enter the specified file. This returns true on failure. 575 bool enterIncludeFile(const std::string &Filename); 576 577 /// Reset the current lexer position to that given by \p Loc. The 578 /// current token is not set; clients should ensure Lex() is called 579 /// subsequently. 580 /// 581 /// \param InBuffer If not 0, should be the known buffer id that contains the 582 /// location. 583 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 584 585 /// Parse up to the end of statement and a return the contents from the 586 /// current token until the end of the statement; the current token on exit 587 /// will be either the EndOfStatement or EOF. 588 StringRef parseStringToEndOfStatement() override; 589 590 bool parseTextItem(std::string &Data); 591 592 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 593 MCBinaryExpr::Opcode &Kind); 594 595 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 596 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 597 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 598 599 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 600 601 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 602 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 603 604 // Generic (target and platform independent) directive parsing. 605 enum DirectiveKind { 606 DK_NO_DIRECTIVE, // Placeholder 607 DK_HANDLER_DIRECTIVE, 608 DK_ASSIGN, 609 DK_EQU, 610 DK_TEXTEQU, 611 DK_ASCII, 612 DK_ASCIZ, 613 DK_STRING, 614 DK_BYTE, 615 DK_SBYTE, 616 DK_WORD, 617 DK_SWORD, 618 DK_DWORD, 619 DK_SDWORD, 620 DK_FWORD, 621 DK_QWORD, 622 DK_SQWORD, 623 DK_DB, 624 DK_DD, 625 DK_DQ, 626 DK_DW, 627 DK_REAL4, 628 DK_REAL8, 629 DK_ALIGN, 630 DK_ORG, 631 DK_ENDR, 632 DK_EXTERN, 633 DK_PUBLIC, 634 DK_COMM, 635 DK_COMMENT, 636 DK_INCLUDE, 637 DK_REPT, 638 DK_IRP, 639 DK_IRPC, 640 DK_IF, 641 DK_IFE, 642 DK_IFB, 643 DK_IFNB, 644 DK_IFDEF, 645 DK_IFNDEF, 646 DK_IFDIF, 647 DK_IFDIFI, 648 DK_IFIDN, 649 DK_IFIDNI, 650 DK_ELSEIF, 651 DK_ELSEIFE, 652 DK_ELSEIFB, 653 DK_ELSEIFNB, 654 DK_ELSEIFDEF, 655 DK_ELSEIFNDEF, 656 DK_ELSEIFDIF, 657 DK_ELSEIFDIFI, 658 DK_ELSEIFIDN, 659 DK_ELSEIFIDNI, 660 DK_ELSE, 661 DK_ENDIF, 662 DK_FILE, 663 DK_LINE, 664 DK_LOC, 665 DK_STABS, 666 DK_CV_FILE, 667 DK_CV_FUNC_ID, 668 DK_CV_INLINE_SITE_ID, 669 DK_CV_LOC, 670 DK_CV_LINETABLE, 671 DK_CV_INLINE_LINETABLE, 672 DK_CV_DEF_RANGE, 673 DK_CV_STRINGTABLE, 674 DK_CV_STRING, 675 DK_CV_FILECHECKSUMS, 676 DK_CV_FILECHECKSUM_OFFSET, 677 DK_CV_FPO_DATA, 678 DK_CFI_SECTIONS, 679 DK_CFI_STARTPROC, 680 DK_CFI_ENDPROC, 681 DK_CFI_DEF_CFA, 682 DK_CFI_DEF_CFA_OFFSET, 683 DK_CFI_ADJUST_CFA_OFFSET, 684 DK_CFI_DEF_CFA_REGISTER, 685 DK_CFI_OFFSET, 686 DK_CFI_REL_OFFSET, 687 DK_CFI_PERSONALITY, 688 DK_CFI_LSDA, 689 DK_CFI_REMEMBER_STATE, 690 DK_CFI_RESTORE_STATE, 691 DK_CFI_SAME_VALUE, 692 DK_CFI_RESTORE, 693 DK_CFI_ESCAPE, 694 DK_CFI_RETURN_COLUMN, 695 DK_CFI_SIGNAL_FRAME, 696 DK_CFI_UNDEFINED, 697 DK_CFI_REGISTER, 698 DK_CFI_WINDOW_SAVE, 699 DK_CFI_B_KEY_FRAME, 700 DK_ALTMACRO, 701 DK_NOALTMACRO, 702 DK_MACRO, 703 DK_EXITM, 704 DK_ENDM, 705 DK_PURGEM, 706 DK_ERR, 707 DK_ERRB, 708 DK_ERRNB, 709 DK_ERRDEF, 710 DK_ERRNDEF, 711 DK_ERRDIF, 712 DK_ERRDIFI, 713 DK_ERRIDN, 714 DK_ERRIDNI, 715 DK_ERRE, 716 DK_ERRNZ, 717 DK_ECHO, 718 DK_STRUCT, 719 DK_UNION, 720 DK_ENDS, 721 DK_END 722 }; 723 724 /// Maps directive name --> DirectiveKind enum, for directives parsed by this 725 /// class. 726 StringMap<DirectiveKind> DirectiveKindMap; 727 728 // Codeview def_range type parsing. 729 enum CVDefRangeType { 730 CVDR_DEFRANGE = 0, // Placeholder 731 CVDR_DEFRANGE_REGISTER, 732 CVDR_DEFRANGE_FRAMEPOINTER_REL, 733 CVDR_DEFRANGE_SUBFIELD_REGISTER, 734 CVDR_DEFRANGE_REGISTER_REL 735 }; 736 737 /// Maps Codeview def_range types --> CVDefRangeType enum, for Codeview 738 /// def_range types parsed by this class. 739 StringMap<CVDefRangeType> CVDefRangeTypeMap; 740 741 bool parseInitValue(unsigned Size); 742 743 // ".ascii", ".asciz", ".string" 744 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 745 746 // "byte", "word", ... 747 bool emitIntValue(const MCExpr *Value, unsigned Size); 748 bool parseScalarInitializer(unsigned Size, 749 SmallVectorImpl<const MCExpr *> &Values, 750 unsigned StringPadLength = 0); 751 bool parseScalarInstList( 752 unsigned Size, SmallVectorImpl<const MCExpr *> &Values, 753 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 754 bool emitIntegralValues(unsigned Size); 755 bool addIntegralField(StringRef Name, unsigned Size); 756 bool parseDirectiveValue(StringRef IDVal, unsigned Size); 757 bool parseDirectiveNamedValue(StringRef IDVal, unsigned Size, StringRef Name, 758 SMLoc NameLoc); 759 760 // "real4", "real8" 761 bool emitRealValues(const fltSemantics &Semantics); 762 bool addRealField(StringRef Name, const fltSemantics &Semantics); 763 bool parseDirectiveRealValue(StringRef IDVal, const fltSemantics &Semantics); 764 bool parseRealInstList( 765 const fltSemantics &Semantics, SmallVectorImpl<APInt> &Values, 766 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 767 bool parseDirectiveNamedRealValue(StringRef IDVal, 768 const fltSemantics &Semantics, 769 StringRef Name, SMLoc NameLoc); 770 771 bool parseOptionalAngleBracketOpen(); 772 bool parseAngleBracketClose(const Twine &Msg = "expected '>'"); 773 774 bool parseFieldInitializer(const FieldInfo &Field, 775 FieldInitializer &Initializer); 776 bool parseFieldInitializer(const FieldInfo &Field, 777 const IntFieldInfo &Contents, 778 FieldInitializer &Initializer); 779 bool parseFieldInitializer(const FieldInfo &Field, 780 const RealFieldInfo &Contents, 781 FieldInitializer &Initializer); 782 bool parseFieldInitializer(const FieldInfo &Field, 783 const StructFieldInfo &Contents, 784 FieldInitializer &Initializer); 785 786 bool parseStructInitializer(const StructInfo &Structure, 787 StructInitializer &Initializer); 788 bool parseStructInstList( 789 const StructInfo &Structure, std::vector<StructInitializer> &Initializers, 790 const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); 791 792 bool emitFieldValue(const FieldInfo &Field); 793 bool emitFieldValue(const FieldInfo &Field, const IntFieldInfo &Contents); 794 bool emitFieldValue(const FieldInfo &Field, const RealFieldInfo &Contents); 795 bool emitFieldValue(const FieldInfo &Field, const StructFieldInfo &Contents); 796 797 bool emitStructValue(const StructInfo &Structure); 798 799 bool emitFieldInitializer(const FieldInfo &Field, 800 const FieldInitializer &Initializer); 801 bool emitFieldInitializer(const FieldInfo &Field, 802 const IntFieldInfo &Contents, 803 const IntFieldInfo &Initializer); 804 bool emitFieldInitializer(const FieldInfo &Field, 805 const RealFieldInfo &Contents, 806 const RealFieldInfo &Initializer); 807 bool emitFieldInitializer(const FieldInfo &Field, 808 const StructFieldInfo &Contents, 809 const StructFieldInfo &Initializer); 810 811 bool emitStructInitializer(const StructInfo &Structure, 812 const StructInitializer &Initializer); 813 814 // User-defined types (structs, unions): 815 bool emitStructValues(const StructInfo &Structure); 816 bool addStructField(StringRef Name, const StructInfo &Structure); 817 bool parseDirectiveStructValue(const StructInfo &Structure, 818 StringRef Directive, SMLoc DirLoc); 819 bool parseDirectiveNamedStructValue(const StructInfo &Structure, 820 StringRef Directive, SMLoc DirLoc, 821 StringRef Name); 822 823 // "=", "equ", "textequ" 824 bool parseDirectiveEquate(StringRef IDVal, StringRef Name, 825 DirectiveKind DirKind); 826 827 bool parseDirectiveOrg(); // ".org" 828 bool parseDirectiveAlign(); // "align" 829 830 // ".file", ".line", ".loc", ".stabs" 831 bool parseDirectiveFile(SMLoc DirectiveLoc); 832 bool parseDirectiveLine(); 833 bool parseDirectiveLoc(); 834 bool parseDirectiveStabs(); 835 836 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 837 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 838 bool parseDirectiveCVFile(); 839 bool parseDirectiveCVFuncId(); 840 bool parseDirectiveCVInlineSiteId(); 841 bool parseDirectiveCVLoc(); 842 bool parseDirectiveCVLinetable(); 843 bool parseDirectiveCVInlineLinetable(); 844 bool parseDirectiveCVDefRange(); 845 bool parseDirectiveCVString(); 846 bool parseDirectiveCVStringTable(); 847 bool parseDirectiveCVFileChecksums(); 848 bool parseDirectiveCVFileChecksumOffset(); 849 bool parseDirectiveCVFPOData(); 850 851 // .cfi directives 852 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 853 bool parseDirectiveCFIWindowSave(); 854 bool parseDirectiveCFISections(); 855 bool parseDirectiveCFIStartProc(); 856 bool parseDirectiveCFIEndProc(); 857 bool parseDirectiveCFIDefCfaOffset(); 858 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 859 bool parseDirectiveCFIAdjustCfaOffset(); 860 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 861 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 862 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 863 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 864 bool parseDirectiveCFIRememberState(); 865 bool parseDirectiveCFIRestoreState(); 866 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 867 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 868 bool parseDirectiveCFIEscape(); 869 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 870 bool parseDirectiveCFISignalFrame(); 871 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 872 873 // macro directives 874 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 875 bool parseDirectiveExitMacro(StringRef Directive); 876 bool parseDirectiveEndMacro(StringRef Directive); 877 bool parseDirectiveMacro(SMLoc DirectiveLoc); 878 // alternate macro mode directives 879 bool parseDirectiveAltmacro(StringRef Directive); 880 881 bool parseDirectiveStruct(StringRef Directive, DirectiveKind DirKind, 882 StringRef Name, SMLoc NameLoc); 883 bool parseDirectiveNestedStruct(StringRef Directive, DirectiveKind DirKind); 884 bool parseDirectiveEnds(StringRef Name, SMLoc NameLoc); 885 bool parseDirectiveNestedEnds(); 886 887 /// Parse a directive like ".globl" which accepts a single symbol (which 888 /// should be a label or an external). 889 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 890 891 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 892 893 bool parseDirectiveComment(SMLoc DirectiveLoc); // "comment" 894 895 bool parseDirectiveInclude(); // "include" 896 897 // "if" or "ife" 898 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 899 // "ifb" or "ifnb", depending on ExpectBlank. 900 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 901 // "ifidn", "ifdif", "ifidni", or "ifdifi", depending on ExpectEqual and 902 // CaseInsensitive. 903 bool parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 904 bool CaseInsensitive); 905 // "ifdef" or "ifndef", depending on expect_defined 906 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 907 // "elseif" or "elseife" 908 bool parseDirectiveElseIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 909 // "elseifb" or "elseifnb", depending on ExpectBlank. 910 bool parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank); 911 // ".elseifdef" or ".elseifndef", depending on expect_defined 912 bool parseDirectiveElseIfdef(SMLoc DirectiveLoc, bool expect_defined); 913 // "elseifidn", "elseifdif", "elseifidni", or "elseifdifi", depending on 914 // ExpectEqual and CaseInsensitive. 915 bool parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 916 bool CaseInsensitive); 917 bool parseDirectiveElse(SMLoc DirectiveLoc); // "else" 918 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // "endif" 919 bool parseEscapedString(std::string &Data) override; 920 bool parseAngleBracketString(std::string &Data) override; 921 922 // Macro-like directives 923 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 924 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 925 raw_svector_ostream &OS); 926 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 927 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 928 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 929 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 930 931 // "_emit" or "__emit" 932 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 933 size_t Len); 934 935 // "align" 936 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 937 938 // "end" 939 bool parseDirectiveEnd(SMLoc DirectiveLoc); 940 941 // ".err" 942 bool parseDirectiveError(SMLoc DirectiveLoc); 943 // ".errb" or ".errnb", depending on ExpectBlank. 944 bool parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank); 945 // ".errdef" or ".errndef", depending on ExpectBlank. 946 bool parseDirectiveErrorIfdef(SMLoc DirectiveLoc, bool ExpectDefined); 947 // ".erridn", ".errdif", ".erridni", or ".errdifi", depending on ExpectEqual 948 // and CaseInsensitive. 949 bool parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 950 bool CaseInsensitive); 951 // ".erre" or ".errnz", depending on ExpectZero. 952 bool parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero); 953 954 // "echo" 955 bool parseDirectiveEcho(); 956 957 void initializeDirectiveKindMap(); 958 void initializeCVDefRangeTypeMap(); 959 }; 960 961 } // end anonymous namespace 962 963 namespace llvm { 964 965 extern MCAsmParserExtension *createCOFFMasmParser(); 966 967 } // end namespace llvm 968 969 enum { DEFAULT_ADDRSPACE = 0 }; 970 971 MasmParser::MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 972 const MCAsmInfo &MAI, unsigned CB = 0) 973 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 974 CurBuffer(CB ? CB : SM.getMainFileID()) { 975 HadError = false; 976 // Save the old handler. 977 SavedDiagHandler = SrcMgr.getDiagHandler(); 978 SavedDiagContext = SrcMgr.getDiagContext(); 979 // Set our own handler which calls the saved handler. 980 SrcMgr.setDiagHandler(DiagHandler, this); 981 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 982 983 // Initialize the platform / file format parser. 984 switch (Ctx.getObjectFileInfo()->getObjectFileType()) { 985 case MCObjectFileInfo::IsCOFF: 986 PlatformParser.reset(createCOFFMasmParser()); 987 break; 988 default: 989 report_fatal_error("llvm-ml currently supports only COFF output."); 990 break; 991 } 992 993 initializeDirectiveKindMap(); 994 PlatformParser->Initialize(*this); 995 initializeCVDefRangeTypeMap(); 996 997 NumOfMacroInstantiations = 0; 998 } 999 1000 MasmParser::~MasmParser() { 1001 assert((HadError || ActiveMacros.empty()) && 1002 "Unexpected active macro instantiation!"); 1003 1004 // Restore the saved diagnostics handler and context for use during 1005 // finalization. 1006 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 1007 } 1008 1009 void MasmParser::printMacroInstantiations() { 1010 // Print the active macro instantiation stack. 1011 for (std::vector<MacroInstantiation *>::const_reverse_iterator 1012 it = ActiveMacros.rbegin(), 1013 ie = ActiveMacros.rend(); 1014 it != ie; ++it) 1015 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 1016 "while in macro instantiation"); 1017 } 1018 1019 void MasmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 1020 printPendingErrors(); 1021 printMessage(L, SourceMgr::DK_Note, Msg, Range); 1022 printMacroInstantiations(); 1023 } 1024 1025 bool MasmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 1026 if (getTargetParser().getTargetOptions().MCNoWarn) 1027 return false; 1028 if (getTargetParser().getTargetOptions().MCFatalWarnings) 1029 return Error(L, Msg, Range); 1030 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 1031 printMacroInstantiations(); 1032 return false; 1033 } 1034 1035 bool MasmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 1036 HadError = true; 1037 printMessage(L, SourceMgr::DK_Error, Msg, Range); 1038 printMacroInstantiations(); 1039 return true; 1040 } 1041 1042 bool MasmParser::enterIncludeFile(const std::string &Filename) { 1043 std::string IncludedFile; 1044 unsigned NewBuf = 1045 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 1046 if (!NewBuf) 1047 return true; 1048 1049 CurBuffer = NewBuf; 1050 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 1051 return false; 1052 } 1053 1054 void MasmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 1055 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 1056 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 1057 Loc.getPointer()); 1058 } 1059 1060 const AsmToken &MasmParser::Lex() { 1061 if (Lexer.getTok().is(AsmToken::Error)) 1062 Error(Lexer.getErrLoc(), Lexer.getErr()); 1063 1064 // if it's a end of statement with a comment in it 1065 if (getTok().is(AsmToken::EndOfStatement)) { 1066 // if this is a line comment output it. 1067 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 1068 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 1069 Out.addExplicitComment(Twine(getTok().getString())); 1070 } 1071 1072 const AsmToken *tok = &Lexer.Lex(); 1073 1074 while (tok->is(AsmToken::Identifier)) { 1075 auto it = Variables.find(tok->getIdentifier()); 1076 if (it != Variables.end() && it->second.IsText) { 1077 std::unique_ptr<MemoryBuffer> Instantiation = 1078 MemoryBuffer::getMemBufferCopy(it->second.TextValue, 1079 "<instantiation>"); 1080 1081 // Jump to the macro instantiation and prime the lexer. 1082 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), 1083 getTok().getEndLoc()); 1084 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), nullptr, 1085 /*EndStatementAtEOF=*/false); 1086 tok = &Lexer.Lex(); 1087 } else { 1088 break; 1089 } 1090 } 1091 1092 // Parse comments here to be deferred until end of next statement. 1093 while (tok->is(AsmToken::Comment)) { 1094 if (MAI.preserveAsmComments()) 1095 Out.addExplicitComment(Twine(tok->getString())); 1096 tok = &Lexer.Lex(); 1097 } 1098 1099 if (tok->is(AsmToken::Eof)) { 1100 // If this is the end of an included file, pop the parent file off the 1101 // include stack. 1102 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 1103 if (ParentIncludeLoc != SMLoc()) { 1104 jumpToLoc(ParentIncludeLoc); 1105 return Lex(); 1106 } 1107 } 1108 1109 return *tok; 1110 } 1111 1112 bool MasmParser::enabledGenDwarfForAssembly() { 1113 // Check whether the user specified -g. 1114 if (!getContext().getGenDwarfForAssembly()) 1115 return false; 1116 // If we haven't encountered any .file directives (which would imply that 1117 // the assembler source was produced with debug info already) then emit one 1118 // describing the assembler source file itself. 1119 if (getContext().getGenDwarfFileNumber() == 0) { 1120 // Use the first #line directive for this, if any. It's preprocessed, so 1121 // there is no checksum, and of course no source directive. 1122 if (!FirstCppHashFilename.empty()) 1123 getContext().setMCLineTableRootFile(/*CUID=*/0, 1124 getContext().getCompilationDir(), 1125 FirstCppHashFilename, 1126 /*Cksum=*/None, /*Source=*/None); 1127 const MCDwarfFile &RootFile = 1128 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 1129 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective( 1130 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 1131 RootFile.Checksum, RootFile.Source)); 1132 } 1133 return true; 1134 } 1135 1136 bool MasmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 1137 // Create the initial section, if requested. 1138 if (!NoInitialTextSection) 1139 Out.InitSections(false); 1140 1141 // Prime the lexer. 1142 Lex(); 1143 1144 HadError = false; 1145 AsmCond StartingCondState = TheCondState; 1146 SmallVector<AsmRewrite, 4> AsmStrRewrites; 1147 1148 // If we are generating dwarf for assembly source files save the initial text 1149 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 1150 // emitting any actual debug info yet and haven't had a chance to parse any 1151 // embedded .file directives.) 1152 if (getContext().getGenDwarfForAssembly()) { 1153 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 1154 if (!Sec->getBeginSymbol()) { 1155 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 1156 getStreamer().emitLabel(SectionStartSym); 1157 Sec->setBeginSymbol(SectionStartSym); 1158 } 1159 bool InsertResult = getContext().addGenDwarfSection(Sec); 1160 assert(InsertResult && ".text section should not have debug info yet"); 1161 (void)InsertResult; 1162 } 1163 1164 // While we have input, parse each statement. 1165 while (Lexer.isNot(AsmToken::Eof)) { 1166 ParseStatementInfo Info(&AsmStrRewrites); 1167 bool Parsed = parseStatement(Info, nullptr); 1168 1169 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 1170 // for printing ErrMsg via Lex() only if no (presumably better) parser error 1171 // exists. 1172 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 1173 Lex(); 1174 } 1175 1176 // parseStatement returned true so may need to emit an error. 1177 printPendingErrors(); 1178 1179 // Skipping to the next line if needed. 1180 if (Parsed && !getLexer().isAtStartOfStatement()) 1181 eatToEndOfStatement(); 1182 } 1183 1184 getTargetParser().onEndOfFile(); 1185 printPendingErrors(); 1186 1187 // All errors should have been emitted. 1188 assert(!hasPendingError() && "unexpected error from parseStatement"); 1189 1190 getTargetParser().flushPendingInstructions(getStreamer()); 1191 1192 if (TheCondState.TheCond != StartingCondState.TheCond || 1193 TheCondState.Ignore != StartingCondState.Ignore) 1194 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 1195 // Check to see there are no empty DwarfFile slots. 1196 const auto &LineTables = getContext().getMCDwarfLineTables(); 1197 if (!LineTables.empty()) { 1198 unsigned Index = 0; 1199 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 1200 if (File.Name.empty() && Index != 0) 1201 printError(getTok().getLoc(), "unassigned file number: " + 1202 Twine(Index) + 1203 " for .file directives"); 1204 ++Index; 1205 } 1206 } 1207 1208 // Check to see that all assembler local symbols were actually defined. 1209 // Targets that don't do subsections via symbols may not want this, though, 1210 // so conservatively exclude them. Only do this if we're finalizing, though, 1211 // as otherwise we won't necessarilly have seen everything yet. 1212 if (!NoFinalize) { 1213 if (MAI.hasSubsectionsViaSymbols()) { 1214 for (const auto &TableEntry : getContext().getSymbols()) { 1215 MCSymbol *Sym = TableEntry.getValue(); 1216 // Variable symbols may not be marked as defined, so check those 1217 // explicitly. If we know it's a variable, we have a definition for 1218 // the purposes of this check. 1219 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 1220 // FIXME: We would really like to refer back to where the symbol was 1221 // first referenced for a source location. We need to add something 1222 // to track that. Currently, we just point to the end of the file. 1223 printError(getTok().getLoc(), "assembler local symbol '" + 1224 Sym->getName() + "' not defined"); 1225 } 1226 } 1227 1228 // Temporary symbols like the ones for directional jumps don't go in the 1229 // symbol table. They also need to be diagnosed in all (final) cases. 1230 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 1231 if (std::get<2>(LocSym)->isUndefined()) { 1232 // Reset the state of any "# line file" directives we've seen to the 1233 // context as it was at the diagnostic site. 1234 CppHashInfo = std::get<1>(LocSym); 1235 printError(std::get<0>(LocSym), "directional label undefined"); 1236 } 1237 } 1238 } 1239 1240 // Finalize the output stream if there are no errors and if the client wants 1241 // us to. 1242 if (!HadError && !NoFinalize) 1243 Out.Finish(); 1244 1245 return HadError || getContext().hadError(); 1246 } 1247 1248 bool MasmParser::checkForValidSection() { 1249 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1250 Out.InitSections(false); 1251 return Error(getTok().getLoc(), 1252 "expected section directive before assembly directive"); 1253 } 1254 return false; 1255 } 1256 1257 /// Throw away the rest of the line for testing purposes. 1258 void MasmParser::eatToEndOfStatement() { 1259 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1260 Lexer.Lex(); 1261 1262 // Eat EOL. 1263 if (Lexer.is(AsmToken::EndOfStatement)) 1264 Lexer.Lex(); 1265 } 1266 1267 StringRef MasmParser::parseStringToEndOfStatement() { 1268 const char *Start = getTok().getLoc().getPointer(); 1269 1270 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1271 Lexer.Lex(); 1272 1273 const char *End = getTok().getLoc().getPointer(); 1274 return StringRef(Start, End - Start); 1275 } 1276 1277 /// Parse a paren expression and return it. 1278 /// NOTE: This assumes the leading '(' has already been consumed. 1279 /// 1280 /// parenexpr ::= expr) 1281 /// 1282 bool MasmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1283 if (parseExpression(Res)) 1284 return true; 1285 if (Lexer.isNot(AsmToken::RParen)) 1286 return TokError("expected ')' in parentheses expression"); 1287 EndLoc = Lexer.getTok().getEndLoc(); 1288 Lex(); 1289 return false; 1290 } 1291 1292 /// Parse a bracket expression and return it. 1293 /// NOTE: This assumes the leading '[' has already been consumed. 1294 /// 1295 /// bracketexpr ::= expr] 1296 /// 1297 bool MasmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1298 if (parseExpression(Res)) 1299 return true; 1300 EndLoc = getTok().getEndLoc(); 1301 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1302 return true; 1303 return false; 1304 } 1305 1306 /// Parse a primary expression and return it. 1307 /// primaryexpr ::= (parenexpr 1308 /// primaryexpr ::= symbol 1309 /// primaryexpr ::= number 1310 /// primaryexpr ::= '.' 1311 /// primaryexpr ::= ~,+,- primaryexpr 1312 bool MasmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1313 SMLoc FirstTokenLoc = getLexer().getLoc(); 1314 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1315 switch (FirstTokenKind) { 1316 default: 1317 return TokError("unknown token in expression"); 1318 // If we have an error assume that we've already handled it. 1319 case AsmToken::Error: 1320 return true; 1321 case AsmToken::Exclaim: 1322 Lex(); // Eat the operator. 1323 if (parsePrimaryExpr(Res, EndLoc)) 1324 return true; 1325 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1326 return false; 1327 case AsmToken::Dollar: 1328 case AsmToken::At: 1329 case AsmToken::String: 1330 case AsmToken::Identifier: { 1331 StringRef Identifier; 1332 if (parseIdentifier(Identifier)) { 1333 // We may have failed but $ may be a valid token. 1334 if (getTok().is(AsmToken::Dollar)) { 1335 if (Lexer.getMAI().getDollarIsPC()) { 1336 Lex(); 1337 // This is a '$' reference, which references the current PC. Emit a 1338 // temporary label to the streamer and refer to it. 1339 MCSymbol *Sym = Ctx.createTempSymbol(); 1340 Out.emitLabel(Sym); 1341 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1342 getContext()); 1343 EndLoc = FirstTokenLoc; 1344 return false; 1345 } 1346 return Error(FirstTokenLoc, "invalid token in expression"); 1347 } 1348 } 1349 // Parse symbol variant. 1350 std::pair<StringRef, StringRef> Split; 1351 if (!MAI.useParensForSymbolVariant()) { 1352 if (FirstTokenKind == AsmToken::String) { 1353 if (Lexer.is(AsmToken::At)) { 1354 Lex(); // eat @ 1355 SMLoc AtLoc = getLexer().getLoc(); 1356 StringRef VName; 1357 if (parseIdentifier(VName)) 1358 return Error(AtLoc, "expected symbol variant after '@'"); 1359 1360 Split = std::make_pair(Identifier, VName); 1361 } 1362 } else { 1363 Split = Identifier.split('@'); 1364 } 1365 } else if (Lexer.is(AsmToken::LParen)) { 1366 Lex(); // eat '('. 1367 StringRef VName; 1368 parseIdentifier(VName); 1369 // eat ')'. 1370 if (parseToken(AsmToken::RParen, 1371 "unexpected token in variant, expected ')'")) 1372 return true; 1373 Split = std::make_pair(Identifier, VName); 1374 } 1375 1376 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1377 1378 // This is a symbol reference. 1379 StringRef SymbolName = Identifier; 1380 if (SymbolName.empty()) 1381 return Error(getLexer().getLoc(), "expected a symbol reference"); 1382 1383 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1384 1385 // Look up the symbol variant if used. 1386 if (!Split.second.empty()) { 1387 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1388 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1389 SymbolName = Split.first; 1390 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1391 Variant = MCSymbolRefExpr::VK_None; 1392 } else { 1393 return Error(SMLoc::getFromPointer(Split.second.begin()), 1394 "invalid variant '" + Split.second + "'"); 1395 } 1396 } 1397 1398 // Find the field offset if used. 1399 StringRef Type; 1400 unsigned Offset = 0; 1401 Split = SymbolName.split('.'); 1402 if (!Split.second.empty()) { 1403 SymbolName = Split.first; 1404 if (Structs.count(SymbolName.lower()) && 1405 !lookUpField(SymbolName, Split.second, Type, Offset)) { 1406 // This is actually a reference to a field offset. 1407 Res = MCConstantExpr::create(Offset, getContext()); 1408 return false; 1409 } 1410 1411 auto TypeIt = KnownType.find(SymbolName); 1412 if (TypeIt == KnownType.end() || 1413 lookUpField(*TypeIt->second, Split.second, Type, Offset)) { 1414 std::pair<StringRef, StringRef> BaseMember = Split.second.split('.'); 1415 StringRef Base = BaseMember.first, Member = BaseMember.second; 1416 lookUpField(Base, Member, Type, Offset); 1417 } 1418 } 1419 1420 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1421 if (!Sym) 1422 Sym = getContext().getOrCreateSymbol(SymbolName); 1423 1424 // If this is an absolute variable reference, substitute it now to preserve 1425 // semantics in the face of reassignment. 1426 if (Sym->isVariable()) { 1427 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1428 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1429 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1430 DoInline = TV->inlineAssignedExpr(); 1431 if (DoInline) { 1432 if (Variant) 1433 return Error(EndLoc, "unexpected modifier on variable reference"); 1434 Res = Sym->getVariableValue(/*SetUsed*/ false); 1435 return false; 1436 } 1437 } 1438 1439 // Otherwise create a symbol ref. 1440 const MCExpr *SymRef = 1441 MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1442 if (Offset) { 1443 Res = MCBinaryExpr::create(MCBinaryExpr::Add, SymRef, 1444 MCConstantExpr::create(Offset, getContext()), 1445 getContext()); 1446 } else { 1447 Res = SymRef; 1448 } 1449 return false; 1450 } 1451 case AsmToken::BigNum: 1452 return TokError("literal value out of range for directive"); 1453 case AsmToken::Integer: { 1454 SMLoc Loc = getTok().getLoc(); 1455 int64_t IntVal = getTok().getIntVal(); 1456 Res = MCConstantExpr::create(IntVal, getContext()); 1457 EndLoc = Lexer.getTok().getEndLoc(); 1458 Lex(); // Eat token. 1459 // Look for 'b' or 'f' following an Integer as a directional label. 1460 if (Lexer.getKind() == AsmToken::Identifier) { 1461 StringRef IDVal = getTok().getString(); 1462 // Look up the symbol variant if used. 1463 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1464 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1465 if (Split.first.size() != IDVal.size()) { 1466 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1467 if (Variant == MCSymbolRefExpr::VK_Invalid) 1468 return TokError("invalid variant '" + Split.second + "'"); 1469 IDVal = Split.first; 1470 } 1471 if (IDVal == "f" || IDVal == "b") { 1472 MCSymbol *Sym = 1473 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1474 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1475 if (IDVal == "b" && Sym->isUndefined()) 1476 return Error(Loc, "directional label undefined"); 1477 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1478 EndLoc = Lexer.getTok().getEndLoc(); 1479 Lex(); // Eat identifier. 1480 } 1481 } 1482 return false; 1483 } 1484 case AsmToken::Real: { 1485 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1486 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1487 Res = MCConstantExpr::create(IntVal, getContext()); 1488 EndLoc = Lexer.getTok().getEndLoc(); 1489 Lex(); // Eat token. 1490 return false; 1491 } 1492 case AsmToken::Dot: { 1493 // This is a '.' reference, which references the current PC. Emit a 1494 // temporary label to the streamer and refer to it. 1495 MCSymbol *Sym = Ctx.createTempSymbol(); 1496 Out.emitLabel(Sym); 1497 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1498 EndLoc = Lexer.getTok().getEndLoc(); 1499 Lex(); // Eat identifier. 1500 return false; 1501 } 1502 case AsmToken::LParen: 1503 Lex(); // Eat the '('. 1504 return parseParenExpr(Res, EndLoc); 1505 case AsmToken::LBrac: 1506 if (!PlatformParser->HasBracketExpressions()) 1507 return TokError("brackets expression not supported on this target"); 1508 Lex(); // Eat the '['. 1509 return parseBracketExpr(Res, EndLoc); 1510 case AsmToken::Minus: 1511 Lex(); // Eat the operator. 1512 if (parsePrimaryExpr(Res, EndLoc)) 1513 return true; 1514 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1515 return false; 1516 case AsmToken::Plus: 1517 Lex(); // Eat the operator. 1518 if (parsePrimaryExpr(Res, EndLoc)) 1519 return true; 1520 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1521 return false; 1522 case AsmToken::Tilde: 1523 Lex(); // Eat the operator. 1524 if (parsePrimaryExpr(Res, EndLoc)) 1525 return true; 1526 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1527 return false; 1528 // MIPS unary expression operators. The lexer won't generate these tokens if 1529 // MCAsmInfo::HasMipsExpressions is false for the target. 1530 case AsmToken::PercentCall16: 1531 case AsmToken::PercentCall_Hi: 1532 case AsmToken::PercentCall_Lo: 1533 case AsmToken::PercentDtprel_Hi: 1534 case AsmToken::PercentDtprel_Lo: 1535 case AsmToken::PercentGot: 1536 case AsmToken::PercentGot_Disp: 1537 case AsmToken::PercentGot_Hi: 1538 case AsmToken::PercentGot_Lo: 1539 case AsmToken::PercentGot_Ofst: 1540 case AsmToken::PercentGot_Page: 1541 case AsmToken::PercentGottprel: 1542 case AsmToken::PercentGp_Rel: 1543 case AsmToken::PercentHi: 1544 case AsmToken::PercentHigher: 1545 case AsmToken::PercentHighest: 1546 case AsmToken::PercentLo: 1547 case AsmToken::PercentNeg: 1548 case AsmToken::PercentPcrel_Hi: 1549 case AsmToken::PercentPcrel_Lo: 1550 case AsmToken::PercentTlsgd: 1551 case AsmToken::PercentTlsldm: 1552 case AsmToken::PercentTprel_Hi: 1553 case AsmToken::PercentTprel_Lo: 1554 Lex(); // Eat the operator. 1555 if (Lexer.isNot(AsmToken::LParen)) 1556 return TokError("expected '(' after operator"); 1557 Lex(); // Eat the operator. 1558 if (parseExpression(Res, EndLoc)) 1559 return true; 1560 if (Lexer.isNot(AsmToken::RParen)) 1561 return TokError("expected ')'"); 1562 Lex(); // Eat the operator. 1563 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1564 return !Res; 1565 } 1566 } 1567 1568 bool MasmParser::parseExpression(const MCExpr *&Res) { 1569 SMLoc EndLoc; 1570 return parseExpression(Res, EndLoc); 1571 } 1572 1573 /// This function checks if the next token is <string> type or arithmetic. 1574 /// string that begin with character '<' must end with character '>'. 1575 /// otherwise it is arithmetics. 1576 /// If the function returns a 'true' value, 1577 /// the End argument will be filled with the last location pointed to the '>' 1578 /// character. 1579 1580 /// There is a gap between the AltMacro's documentation and the single quote 1581 /// implementation. GCC does not fully support this feature and so we will not 1582 /// support it. 1583 /// TODO: Adding single quote as a string. 1584 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) { 1585 assert((StrLoc.getPointer() != nullptr) && 1586 "Argument to the function cannot be a NULL value"); 1587 const char *CharPtr = StrLoc.getPointer(); 1588 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1589 (*CharPtr != '\0')) { 1590 if (*CharPtr == '!') 1591 CharPtr++; 1592 CharPtr++; 1593 } 1594 if (*CharPtr == '>') { 1595 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1596 return true; 1597 } 1598 return false; 1599 } 1600 1601 /// creating a string without the escape characters '!'. 1602 static std::string angleBracketString(StringRef AltMacroStr) { 1603 std::string Res; 1604 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1605 if (AltMacroStr[Pos] == '!') 1606 Pos++; 1607 Res += AltMacroStr[Pos]; 1608 } 1609 return Res; 1610 } 1611 1612 /// Parse an expression and return it. 1613 /// 1614 /// expr ::= expr &&,|| expr -> lowest. 1615 /// expr ::= expr |,^,&,! expr 1616 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1617 /// expr ::= expr <<,>> expr 1618 /// expr ::= expr +,- expr 1619 /// expr ::= expr *,/,% expr -> highest. 1620 /// expr ::= primaryexpr 1621 /// 1622 bool MasmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1623 // Parse the expression. 1624 Res = nullptr; 1625 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1626 parseBinOpRHS(1, Res, EndLoc)) 1627 return true; 1628 1629 // Try to constant fold it up front, if possible. Do not exploit 1630 // assembler here. 1631 int64_t Value; 1632 if (Res->evaluateAsAbsolute(Value)) 1633 Res = MCConstantExpr::create(Value, getContext()); 1634 1635 return false; 1636 } 1637 1638 bool MasmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1639 Res = nullptr; 1640 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1641 } 1642 1643 bool MasmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1644 SMLoc &EndLoc) { 1645 if (parseParenExpr(Res, EndLoc)) 1646 return true; 1647 1648 for (; ParenDepth > 0; --ParenDepth) { 1649 if (parseBinOpRHS(1, Res, EndLoc)) 1650 return true; 1651 1652 // We don't Lex() the last RParen. 1653 // This is the same behavior as parseParenExpression(). 1654 if (ParenDepth - 1 > 0) { 1655 EndLoc = getTok().getEndLoc(); 1656 if (parseToken(AsmToken::RParen, 1657 "expected ')' in parentheses expression")) 1658 return true; 1659 } 1660 } 1661 return false; 1662 } 1663 1664 bool MasmParser::parseAbsoluteExpression(int64_t &Res) { 1665 const MCExpr *Expr; 1666 1667 SMLoc StartLoc = Lexer.getLoc(); 1668 if (parseExpression(Expr)) 1669 return true; 1670 1671 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1672 return Error(StartLoc, "expected absolute expression"); 1673 1674 return false; 1675 } 1676 1677 static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, 1678 MCBinaryExpr::Opcode &Kind, 1679 bool ShouldUseLogicalShr, 1680 bool EndExpressionAtGreater) { 1681 switch (K) { 1682 default: 1683 return 0; // not a binop. 1684 1685 // Lowest Precedence: &&, || 1686 case AsmToken::AmpAmp: 1687 Kind = MCBinaryExpr::LAnd; 1688 return 2; 1689 case AsmToken::PipePipe: 1690 Kind = MCBinaryExpr::LOr; 1691 return 1; 1692 1693 // Low Precedence: ==, !=, <>, <, <=, >, >= 1694 case AsmToken::EqualEqual: 1695 Kind = MCBinaryExpr::EQ; 1696 return 3; 1697 case AsmToken::ExclaimEqual: 1698 case AsmToken::LessGreater: 1699 Kind = MCBinaryExpr::NE; 1700 return 3; 1701 case AsmToken::Less: 1702 Kind = MCBinaryExpr::LT; 1703 return 3; 1704 case AsmToken::LessEqual: 1705 Kind = MCBinaryExpr::LTE; 1706 return 3; 1707 case AsmToken::Greater: 1708 if (EndExpressionAtGreater) 1709 return 0; 1710 Kind = MCBinaryExpr::GT; 1711 return 3; 1712 case AsmToken::GreaterEqual: 1713 Kind = MCBinaryExpr::GTE; 1714 return 3; 1715 1716 // Low Intermediate Precedence: +, - 1717 case AsmToken::Plus: 1718 Kind = MCBinaryExpr::Add; 1719 return 4; 1720 case AsmToken::Minus: 1721 Kind = MCBinaryExpr::Sub; 1722 return 4; 1723 1724 // High Intermediate Precedence: |, &, ^ 1725 // 1726 // FIXME: gas seems to support '!' as an infix operator? 1727 case AsmToken::Pipe: 1728 Kind = MCBinaryExpr::Or; 1729 return 5; 1730 case AsmToken::Caret: 1731 Kind = MCBinaryExpr::Xor; 1732 return 5; 1733 case AsmToken::Amp: 1734 Kind = MCBinaryExpr::And; 1735 return 5; 1736 1737 // Highest Precedence: *, /, %, <<, >> 1738 case AsmToken::Star: 1739 Kind = MCBinaryExpr::Mul; 1740 return 6; 1741 case AsmToken::Slash: 1742 Kind = MCBinaryExpr::Div; 1743 return 6; 1744 case AsmToken::Percent: 1745 Kind = MCBinaryExpr::Mod; 1746 return 6; 1747 case AsmToken::LessLess: 1748 Kind = MCBinaryExpr::Shl; 1749 return 6; 1750 case AsmToken::GreaterGreater: 1751 if (EndExpressionAtGreater) 1752 return 0; 1753 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1754 return 6; 1755 } 1756 } 1757 1758 unsigned MasmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1759 MCBinaryExpr::Opcode &Kind) { 1760 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1761 return getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr, 1762 AngleBracketDepth > 0); 1763 } 1764 1765 /// Parse all binary operators with precedence >= 'Precedence'. 1766 /// Res contains the LHS of the expression on input. 1767 bool MasmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1768 SMLoc &EndLoc) { 1769 SMLoc StartLoc = Lexer.getLoc(); 1770 while (true) { 1771 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1772 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1773 1774 // If the next token is lower precedence than we are allowed to eat, return 1775 // successfully with what we ate already. 1776 if (TokPrec < Precedence) 1777 return false; 1778 1779 Lex(); 1780 1781 // Eat the next primary expression. 1782 const MCExpr *RHS; 1783 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1784 return true; 1785 1786 // If BinOp binds less tightly with RHS than the operator after RHS, let 1787 // the pending operator take RHS as its LHS. 1788 MCBinaryExpr::Opcode Dummy; 1789 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1790 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1791 return true; 1792 1793 // Merge LHS and RHS according to operator. 1794 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1795 } 1796 } 1797 1798 /// ParseStatement: 1799 /// ::= EndOfStatement 1800 /// ::= Label* Directive ...Operands... EndOfStatement 1801 /// ::= Label* Identifier OperandList* EndOfStatement 1802 bool MasmParser::parseStatement(ParseStatementInfo &Info, 1803 MCAsmParserSemaCallback *SI) { 1804 assert(!hasPendingError() && "parseStatement started with pending error"); 1805 // Eat initial spaces and comments. 1806 while (Lexer.is(AsmToken::Space)) 1807 Lex(); 1808 if (Lexer.is(AsmToken::EndOfStatement)) { 1809 // If this is a line comment we can drop it safely. 1810 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1811 getTok().getString().front() == '\n') 1812 Out.AddBlankLine(); 1813 Lex(); 1814 return false; 1815 } 1816 // Statements always start with an identifier, unless we're dealing with a 1817 // processor directive (.386, .686, etc.) that lexes as a real. 1818 AsmToken ID = getTok(); 1819 SMLoc IDLoc = ID.getLoc(); 1820 StringRef IDVal; 1821 int64_t LocalLabelVal = -1; 1822 if (Lexer.is(AsmToken::HashDirective)) 1823 return parseCppHashLineFilenameComment(IDLoc); 1824 // Allow an integer followed by a ':' as a directional local label. 1825 if (Lexer.is(AsmToken::Integer)) { 1826 LocalLabelVal = getTok().getIntVal(); 1827 if (LocalLabelVal < 0) { 1828 if (!TheCondState.Ignore) { 1829 Lex(); // always eat a token 1830 return Error(IDLoc, "unexpected token at start of statement"); 1831 } 1832 IDVal = ""; 1833 } else { 1834 IDVal = getTok().getString(); 1835 Lex(); // Consume the integer token to be used as an identifier token. 1836 if (Lexer.getKind() != AsmToken::Colon) { 1837 if (!TheCondState.Ignore) { 1838 Lex(); // always eat a token 1839 return Error(IDLoc, "unexpected token at start of statement"); 1840 } 1841 } 1842 } 1843 } else if (Lexer.is(AsmToken::Dot)) { 1844 // Treat '.' as a valid identifier in this context. 1845 Lex(); 1846 IDVal = "."; 1847 } else if (Lexer.is(AsmToken::LCurly)) { 1848 // Treat '{' as a valid identifier in this context. 1849 Lex(); 1850 IDVal = "{"; 1851 1852 } else if (Lexer.is(AsmToken::RCurly)) { 1853 // Treat '}' as a valid identifier in this context. 1854 Lex(); 1855 IDVal = "}"; 1856 } else if (Lexer.is(AsmToken::Star) && 1857 getTargetParser().starIsStartOfStatement()) { 1858 // Accept '*' as a valid start of statement. 1859 Lex(); 1860 IDVal = "*"; 1861 } else if (Lexer.is(AsmToken::Real)) { 1862 // Treat ".<number>" as a valid identifier in this context. 1863 IDVal = getTok().getString(); 1864 Lex(); // always eat a token 1865 if (!IDVal.startswith(".")) 1866 return Error(IDLoc, "unexpected token at start of statement"); 1867 } else if (parseIdentifier(IDVal)) { 1868 if (!TheCondState.Ignore) { 1869 Lex(); // always eat a token 1870 return Error(IDLoc, "unexpected token at start of statement"); 1871 } 1872 IDVal = ""; 1873 } 1874 1875 // Handle conditional assembly here before checking for skipping. We 1876 // have to do this so that .endif isn't skipped in a ".if 0" block for 1877 // example. 1878 StringMap<DirectiveKind>::const_iterator DirKindIt = 1879 DirectiveKindMap.find(IDVal.lower()); 1880 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1881 ? DK_NO_DIRECTIVE 1882 : DirKindIt->getValue(); 1883 switch (DirKind) { 1884 default: 1885 break; 1886 case DK_IF: 1887 case DK_IFE: 1888 return parseDirectiveIf(IDLoc, DirKind); 1889 case DK_IFB: 1890 return parseDirectiveIfb(IDLoc, true); 1891 case DK_IFNB: 1892 return parseDirectiveIfb(IDLoc, false); 1893 case DK_IFDEF: 1894 return parseDirectiveIfdef(IDLoc, true); 1895 case DK_IFNDEF: 1896 return parseDirectiveIfdef(IDLoc, false); 1897 case DK_IFDIF: 1898 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false, 1899 /*CaseInsensitive=*/false); 1900 case DK_IFDIFI: 1901 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false, 1902 /*CaseInsensitive=*/true); 1903 case DK_IFIDN: 1904 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true, 1905 /*CaseInsensitive=*/false); 1906 case DK_IFIDNI: 1907 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true, 1908 /*CaseInsensitive=*/true); 1909 case DK_ELSEIF: 1910 case DK_ELSEIFE: 1911 return parseDirectiveElseIf(IDLoc, DirKind); 1912 case DK_ELSEIFB: 1913 return parseDirectiveElseIfb(IDLoc, true); 1914 case DK_ELSEIFNB: 1915 return parseDirectiveElseIfb(IDLoc, false); 1916 case DK_ELSEIFDEF: 1917 return parseDirectiveElseIfdef(IDLoc, true); 1918 case DK_ELSEIFNDEF: 1919 return parseDirectiveElseIfdef(IDLoc, false); 1920 case DK_ELSEIFDIF: 1921 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false, 1922 /*CaseInsensitive=*/false); 1923 case DK_ELSEIFDIFI: 1924 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false, 1925 /*CaseInsensitive=*/true); 1926 case DK_ELSEIFIDN: 1927 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true, 1928 /*CaseInsensitive=*/false); 1929 case DK_ELSEIFIDNI: 1930 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true, 1931 /*CaseInsensitive=*/true); 1932 case DK_ELSE: 1933 return parseDirectiveElse(IDLoc); 1934 case DK_ENDIF: 1935 return parseDirectiveEndIf(IDLoc); 1936 } 1937 1938 // Ignore the statement if in the middle of inactive conditional 1939 // (e.g. ".if 0"). 1940 if (TheCondState.Ignore) { 1941 eatToEndOfStatement(); 1942 return false; 1943 } 1944 1945 // FIXME: Recurse on local labels? 1946 1947 // See what kind of statement we have. 1948 switch (Lexer.getKind()) { 1949 case AsmToken::Colon: { 1950 if (!getTargetParser().isLabel(ID)) 1951 break; 1952 if (checkForValidSection()) 1953 return true; 1954 1955 // identifier ':' -> Label. 1956 Lex(); 1957 1958 // Diagnose attempt to use '.' as a label. 1959 if (IDVal == ".") 1960 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1961 1962 // Diagnose attempt to use a variable as a label. 1963 // 1964 // FIXME: Diagnostics. Note the location of the definition as a label. 1965 // FIXME: This doesn't diagnose assignment to a symbol which has been 1966 // implicitly marked as external. 1967 MCSymbol *Sym; 1968 if (LocalLabelVal == -1) { 1969 if (ParsingMSInlineAsm && SI) { 1970 StringRef RewrittenLabel = 1971 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1972 assert(!RewrittenLabel.empty() && 1973 "We should have an internal name here."); 1974 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1975 RewrittenLabel); 1976 IDVal = RewrittenLabel; 1977 } 1978 Sym = getContext().getOrCreateSymbol(IDVal); 1979 } else 1980 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1981 // End of Labels should be treated as end of line for lexing 1982 // purposes but that information is not available to the Lexer who 1983 // does not understand Labels. This may cause us to see a Hash 1984 // here instead of a preprocessor line comment. 1985 if (getTok().is(AsmToken::Hash)) { 1986 StringRef CommentStr = parseStringToEndOfStatement(); 1987 Lexer.Lex(); 1988 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1989 } 1990 1991 // Consume any end of statement token, if present, to avoid spurious 1992 // AddBlankLine calls(). 1993 if (getTok().is(AsmToken::EndOfStatement)) { 1994 Lex(); 1995 } 1996 1997 getTargetParser().doBeforeLabelEmit(Sym); 1998 1999 // Emit the label. 2000 if (!getTargetParser().isParsingMSInlineAsm()) 2001 Out.emitLabel(Sym, IDLoc); 2002 2003 // If we are generating dwarf for assembly source files then gather the 2004 // info to make a dwarf label entry for this label if needed. 2005 if (enabledGenDwarfForAssembly()) 2006 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 2007 IDLoc); 2008 2009 getTargetParser().onLabelParsed(Sym); 2010 2011 return false; 2012 } 2013 2014 default: // Normal instruction or directive. 2015 break; 2016 } 2017 2018 // If macros are enabled, check to see if this is a macro instantiation. 2019 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 2020 return handleMacroEntry(M, IDLoc); 2021 } 2022 2023 // Otherwise, we have a normal instruction or directive. 2024 2025 if (DirKind != DK_NO_DIRECTIVE) { 2026 // There are several entities interested in parsing directives: 2027 // 2028 // 1. Asm parser extensions. For example, platform-specific parsers 2029 // (like the ELF parser) register themselves as extensions. 2030 // 2. The target-specific assembly parser. Some directives are target 2031 // specific or may potentially behave differently on certain targets. 2032 // 3. The generic directive parser implemented by this class. These are 2033 // all the directives that behave in a target and platform independent 2034 // manner, or at least have a default behavior that's shared between 2035 // all targets and platforms. 2036 2037 getTargetParser().flushPendingInstructions(getStreamer()); 2038 2039 // Special-case handling of structure-end directives at higher priority, 2040 // since ENDS is overloaded as a segment-end directive. 2041 if (IDVal.equals_lower("ends") && StructInProgress.size() > 1 && 2042 getTok().is(AsmToken::EndOfStatement)) { 2043 return parseDirectiveNestedEnds(); 2044 } 2045 2046 // First, check the extension directive map to see if any extension has 2047 // registered itself to parse this directive. 2048 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2049 ExtensionDirectiveMap.lookup(IDVal.lower()); 2050 if (Handler.first) 2051 return (*Handler.second)(Handler.first, IDVal, IDLoc); 2052 2053 // Next, let the target-specific assembly parser try. 2054 SMLoc StartTokLoc = getTok().getLoc(); 2055 bool TPDirectiveReturn = 2056 ID.is(AsmToken::Identifier) && getTargetParser().ParseDirective(ID); 2057 2058 if (hasPendingError()) 2059 return true; 2060 // Currently the return value should be true if we are 2061 // uninterested but as this is at odds with the standard parsing 2062 // convention (return true = error) we have instances of a parsed 2063 // directive that fails returning true as an error. Catch these 2064 // cases as best as possible errors here. 2065 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 2066 return true; 2067 // Return if we did some parsing or believe we succeeded. 2068 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 2069 return false; 2070 2071 // Finally, if no one else is interested in this directive, it must be 2072 // generic and familiar to this class. 2073 switch (DirKind) { 2074 default: 2075 break; 2076 case DK_ASCII: 2077 return parseDirectiveAscii(IDVal, false); 2078 case DK_ASCIZ: 2079 case DK_STRING: 2080 return parseDirectiveAscii(IDVal, true); 2081 case DK_BYTE: 2082 case DK_SBYTE: 2083 case DK_DB: 2084 return parseDirectiveValue(IDVal, 1); 2085 case DK_WORD: 2086 case DK_SWORD: 2087 case DK_DW: 2088 return parseDirectiveValue(IDVal, 2); 2089 case DK_DWORD: 2090 case DK_SDWORD: 2091 case DK_DD: 2092 return parseDirectiveValue(IDVal, 4); 2093 case DK_FWORD: 2094 return parseDirectiveValue(IDVal, 6); 2095 case DK_QWORD: 2096 case DK_SQWORD: 2097 case DK_DQ: 2098 return parseDirectiveValue(IDVal, 8); 2099 case DK_REAL4: 2100 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 2101 case DK_REAL8: 2102 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 2103 case DK_STRUCT: 2104 case DK_UNION: 2105 return parseDirectiveNestedStruct(IDVal, DirKind); 2106 case DK_ENDS: 2107 return parseDirectiveNestedEnds(); 2108 case DK_ALIGN: 2109 return parseDirectiveAlign(); 2110 case DK_ORG: 2111 return parseDirectiveOrg(); 2112 case DK_EXTERN: 2113 eatToEndOfStatement(); // .extern is the default, ignore it. 2114 return false; 2115 case DK_PUBLIC: 2116 return parseDirectiveSymbolAttribute(MCSA_Global); 2117 case DK_COMM: 2118 return parseDirectiveComm(/*IsLocal=*/false); 2119 case DK_COMMENT: 2120 return parseDirectiveComment(IDLoc); 2121 case DK_INCLUDE: 2122 return parseDirectiveInclude(); 2123 case DK_REPT: 2124 return parseDirectiveRept(IDLoc, IDVal); 2125 case DK_IRP: 2126 return parseDirectiveIrp(IDLoc); 2127 case DK_IRPC: 2128 return parseDirectiveIrpc(IDLoc); 2129 case DK_ENDR: 2130 return parseDirectiveEndr(IDLoc); 2131 case DK_FILE: 2132 return parseDirectiveFile(IDLoc); 2133 case DK_LINE: 2134 return parseDirectiveLine(); 2135 case DK_LOC: 2136 return parseDirectiveLoc(); 2137 case DK_STABS: 2138 return parseDirectiveStabs(); 2139 case DK_CV_FILE: 2140 return parseDirectiveCVFile(); 2141 case DK_CV_FUNC_ID: 2142 return parseDirectiveCVFuncId(); 2143 case DK_CV_INLINE_SITE_ID: 2144 return parseDirectiveCVInlineSiteId(); 2145 case DK_CV_LOC: 2146 return parseDirectiveCVLoc(); 2147 case DK_CV_LINETABLE: 2148 return parseDirectiveCVLinetable(); 2149 case DK_CV_INLINE_LINETABLE: 2150 return parseDirectiveCVInlineLinetable(); 2151 case DK_CV_DEF_RANGE: 2152 return parseDirectiveCVDefRange(); 2153 case DK_CV_STRING: 2154 return parseDirectiveCVString(); 2155 case DK_CV_STRINGTABLE: 2156 return parseDirectiveCVStringTable(); 2157 case DK_CV_FILECHECKSUMS: 2158 return parseDirectiveCVFileChecksums(); 2159 case DK_CV_FILECHECKSUM_OFFSET: 2160 return parseDirectiveCVFileChecksumOffset(); 2161 case DK_CV_FPO_DATA: 2162 return parseDirectiveCVFPOData(); 2163 case DK_CFI_SECTIONS: 2164 return parseDirectiveCFISections(); 2165 case DK_CFI_STARTPROC: 2166 return parseDirectiveCFIStartProc(); 2167 case DK_CFI_ENDPROC: 2168 return parseDirectiveCFIEndProc(); 2169 case DK_CFI_DEF_CFA: 2170 return parseDirectiveCFIDefCfa(IDLoc); 2171 case DK_CFI_DEF_CFA_OFFSET: 2172 return parseDirectiveCFIDefCfaOffset(); 2173 case DK_CFI_ADJUST_CFA_OFFSET: 2174 return parseDirectiveCFIAdjustCfaOffset(); 2175 case DK_CFI_DEF_CFA_REGISTER: 2176 return parseDirectiveCFIDefCfaRegister(IDLoc); 2177 case DK_CFI_OFFSET: 2178 return parseDirectiveCFIOffset(IDLoc); 2179 case DK_CFI_REL_OFFSET: 2180 return parseDirectiveCFIRelOffset(IDLoc); 2181 case DK_CFI_PERSONALITY: 2182 return parseDirectiveCFIPersonalityOrLsda(true); 2183 case DK_CFI_LSDA: 2184 return parseDirectiveCFIPersonalityOrLsda(false); 2185 case DK_CFI_REMEMBER_STATE: 2186 return parseDirectiveCFIRememberState(); 2187 case DK_CFI_RESTORE_STATE: 2188 return parseDirectiveCFIRestoreState(); 2189 case DK_CFI_SAME_VALUE: 2190 return parseDirectiveCFISameValue(IDLoc); 2191 case DK_CFI_RESTORE: 2192 return parseDirectiveCFIRestore(IDLoc); 2193 case DK_CFI_ESCAPE: 2194 return parseDirectiveCFIEscape(); 2195 case DK_CFI_RETURN_COLUMN: 2196 return parseDirectiveCFIReturnColumn(IDLoc); 2197 case DK_CFI_SIGNAL_FRAME: 2198 return parseDirectiveCFISignalFrame(); 2199 case DK_CFI_UNDEFINED: 2200 return parseDirectiveCFIUndefined(IDLoc); 2201 case DK_CFI_REGISTER: 2202 return parseDirectiveCFIRegister(IDLoc); 2203 case DK_CFI_WINDOW_SAVE: 2204 return parseDirectiveCFIWindowSave(); 2205 case DK_MACRO: 2206 return parseDirectiveMacro(IDLoc); 2207 case DK_ALTMACRO: 2208 case DK_NOALTMACRO: 2209 return parseDirectiveAltmacro(IDVal); 2210 case DK_EXITM: 2211 return parseDirectiveExitMacro(IDVal); 2212 case DK_ENDM: 2213 return parseDirectiveEndMacro(IDVal); 2214 case DK_PURGEM: 2215 return parseDirectivePurgeMacro(IDLoc); 2216 case DK_END: 2217 return parseDirectiveEnd(IDLoc); 2218 case DK_ERR: 2219 return parseDirectiveError(IDLoc); 2220 case DK_ERRB: 2221 return parseDirectiveErrorIfb(IDLoc, true); 2222 case DK_ERRNB: 2223 return parseDirectiveErrorIfb(IDLoc, false); 2224 case DK_ERRDEF: 2225 return parseDirectiveErrorIfdef(IDLoc, true); 2226 case DK_ERRNDEF: 2227 return parseDirectiveErrorIfdef(IDLoc, false); 2228 case DK_ERRDIF: 2229 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false, 2230 /*CaseInsensitive=*/false); 2231 case DK_ERRDIFI: 2232 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false, 2233 /*CaseInsensitive=*/true); 2234 case DK_ERRIDN: 2235 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true, 2236 /*CaseInsensitive=*/false); 2237 case DK_ERRIDNI: 2238 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true, 2239 /*CaseInsensitive=*/true); 2240 case DK_ERRE: 2241 return parseDirectiveErrorIfe(IDLoc, true); 2242 case DK_ERRNZ: 2243 return parseDirectiveErrorIfe(IDLoc, false); 2244 case DK_ECHO: 2245 return parseDirectiveEcho(); 2246 } 2247 2248 return Error(IDLoc, "unknown directive"); 2249 } 2250 2251 // We also check if this is allocating memory with user-defined type. 2252 auto IDIt = Structs.find(IDVal.lower()); 2253 if (IDIt != Structs.end()) 2254 return parseDirectiveStructValue(/*Structure=*/IDIt->getValue(), IDVal, 2255 IDLoc); 2256 2257 // Non-conditional Microsoft directives sometimes follow their first argument. 2258 const AsmToken nextTok = getTok(); 2259 const StringRef nextVal = nextTok.getString(); 2260 const SMLoc nextLoc = nextTok.getLoc(); 2261 2262 // There are several entities interested in parsing infix directives: 2263 // 2264 // 1. Asm parser extensions. For example, platform-specific parsers 2265 // (like the ELF parser) register themselves as extensions. 2266 // 2. The generic directive parser implemented by this class. These are 2267 // all the directives that behave in a target and platform independent 2268 // manner, or at least have a default behavior that's shared between 2269 // all targets and platforms. 2270 2271 getTargetParser().flushPendingInstructions(getStreamer()); 2272 2273 // Special-case handling of structure-end directives at higher priority, since 2274 // ENDS is overloaded as a segment-end directive. 2275 if (nextVal.equals_lower("ends") && StructInProgress.size() == 1) { 2276 Lex(); 2277 return parseDirectiveEnds(IDVal, IDLoc); 2278 } 2279 2280 // First, check the extension directive map to see if any extension has 2281 // registered itself to parse this directive. 2282 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2283 ExtensionDirectiveMap.lookup(nextVal.lower()); 2284 if (Handler.first) { 2285 Lex(); 2286 Lexer.UnLex(ID); 2287 return (*Handler.second)(Handler.first, nextVal, nextLoc); 2288 } 2289 2290 // If no one else is interested in this directive, it must be 2291 // generic and familiar to this class. 2292 DirKindIt = DirectiveKindMap.find(nextVal.lower()); 2293 DirKind = (DirKindIt == DirectiveKindMap.end()) 2294 ? DK_NO_DIRECTIVE 2295 : DirKindIt->getValue(); 2296 switch (DirKind) { 2297 default: 2298 break; 2299 case DK_ASSIGN: 2300 case DK_EQU: 2301 case DK_TEXTEQU: 2302 Lex(); 2303 return parseDirectiveEquate(nextVal, IDVal, DirKind); 2304 case DK_BYTE: 2305 case DK_DB: 2306 Lex(); 2307 return parseDirectiveNamedValue(nextVal, 1, IDVal, IDLoc); 2308 case DK_WORD: 2309 case DK_DW: 2310 Lex(); 2311 return parseDirectiveNamedValue(nextVal, 2, IDVal, IDLoc); 2312 case DK_DWORD: 2313 case DK_DD: 2314 Lex(); 2315 return parseDirectiveNamedValue(nextVal, 4, IDVal, IDLoc); 2316 case DK_FWORD: 2317 Lex(); 2318 return parseDirectiveNamedValue(nextVal, 6, IDVal, IDLoc); 2319 case DK_QWORD: 2320 case DK_DQ: 2321 Lex(); 2322 return parseDirectiveNamedValue(nextVal, 8, IDVal, IDLoc); 2323 case DK_REAL4: 2324 Lex(); 2325 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEsingle(), IDVal, 2326 IDLoc); 2327 case DK_REAL8: 2328 Lex(); 2329 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEdouble(), IDVal, 2330 IDLoc); 2331 case DK_STRUCT: 2332 case DK_UNION: 2333 Lex(); 2334 return parseDirectiveStruct(nextVal, DirKind, IDVal, IDLoc); 2335 case DK_ENDS: 2336 Lex(); 2337 return parseDirectiveEnds(IDVal, IDLoc); 2338 } 2339 2340 // Finally, we check if this is allocating a variable with user-defined type. 2341 auto NextIt = Structs.find(nextVal.lower()); 2342 if (NextIt != Structs.end()) { 2343 Lex(); 2344 return parseDirectiveNamedStructValue(/*Structure=*/NextIt->getValue(), 2345 nextVal, nextLoc, IDVal); 2346 } 2347 2348 // __asm _emit or __asm __emit 2349 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2350 IDVal == "_EMIT" || IDVal == "__EMIT")) 2351 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2352 2353 // __asm align 2354 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2355 return parseDirectiveMSAlign(IDLoc, Info); 2356 2357 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2358 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2359 if (checkForValidSection()) 2360 return true; 2361 2362 // Canonicalize the opcode to lower case. 2363 std::string OpcodeStr = IDVal.lower(); 2364 ParseInstructionInfo IInfo(Info.AsmRewrites); 2365 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2366 Info.ParsedOperands); 2367 Info.ParseError = ParseHadError; 2368 2369 // Dump the parsed representation, if requested. 2370 if (getShowParsedOperands()) { 2371 SmallString<256> Str; 2372 raw_svector_ostream OS(Str); 2373 OS << "parsed instruction: ["; 2374 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2375 if (i != 0) 2376 OS << ", "; 2377 Info.ParsedOperands[i]->print(OS); 2378 } 2379 OS << "]"; 2380 2381 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2382 } 2383 2384 // Fail even if ParseInstruction erroneously returns false. 2385 if (hasPendingError() || ParseHadError) 2386 return true; 2387 2388 // If we are generating dwarf for the current section then generate a .loc 2389 // directive for the instruction. 2390 if (!ParseHadError && enabledGenDwarfForAssembly() && 2391 getContext().getGenDwarfSectionSyms().count( 2392 getStreamer().getCurrentSectionOnly())) { 2393 unsigned Line; 2394 if (ActiveMacros.empty()) 2395 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2396 else 2397 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2398 ActiveMacros.front()->ExitBuffer); 2399 2400 // If we previously parsed a cpp hash file line comment then make sure the 2401 // current Dwarf File is for the CppHashFilename if not then emit the 2402 // Dwarf File table for it and adjust the line number for the .loc. 2403 if (!CppHashInfo.Filename.empty()) { 2404 unsigned FileNumber = getStreamer().emitDwarfFileDirective( 2405 0, StringRef(), CppHashInfo.Filename); 2406 getContext().setGenDwarfFileNumber(FileNumber); 2407 2408 unsigned CppHashLocLineNo = 2409 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2410 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2411 } 2412 2413 getStreamer().emitDwarfLocDirective( 2414 getContext().getGenDwarfFileNumber(), Line, 0, 2415 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2416 StringRef()); 2417 } 2418 2419 // If parsing succeeded, match the instruction. 2420 if (!ParseHadError) { 2421 uint64_t ErrorInfo; 2422 if (getTargetParser().MatchAndEmitInstruction( 2423 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2424 getTargetParser().isParsingMSInlineAsm())) 2425 return true; 2426 } 2427 return false; 2428 } 2429 2430 // Parse and erase curly braces marking block start/end. 2431 bool MasmParser::parseCurlyBlockScope( 2432 SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2433 // Identify curly brace marking block start/end. 2434 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2435 return false; 2436 2437 SMLoc StartLoc = Lexer.getLoc(); 2438 Lex(); // Eat the brace. 2439 if (Lexer.is(AsmToken::EndOfStatement)) 2440 Lex(); // Eat EndOfStatement following the brace. 2441 2442 // Erase the block start/end brace from the output asm string. 2443 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2444 StartLoc.getPointer()); 2445 return true; 2446 } 2447 2448 /// parseCppHashLineFilenameComment as this: 2449 /// ::= # number "filename" 2450 bool MasmParser::parseCppHashLineFilenameComment(SMLoc L) { 2451 Lex(); // Eat the hash token. 2452 // Lexer only ever emits HashDirective if it fully formed if it's 2453 // done the checking already so this is an internal error. 2454 assert(getTok().is(AsmToken::Integer) && 2455 "Lexing Cpp line comment: Expected Integer"); 2456 int64_t LineNumber = getTok().getIntVal(); 2457 Lex(); 2458 assert(getTok().is(AsmToken::String) && 2459 "Lexing Cpp line comment: Expected String"); 2460 StringRef Filename = getTok().getString(); 2461 Lex(); 2462 2463 // Get rid of the enclosing quotes. 2464 Filename = Filename.substr(1, Filename.size() - 2); 2465 2466 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2467 // and possibly DWARF file info. 2468 CppHashInfo.Loc = L; 2469 CppHashInfo.Filename = Filename; 2470 CppHashInfo.LineNumber = LineNumber; 2471 CppHashInfo.Buf = CurBuffer; 2472 if (FirstCppHashFilename.empty()) 2473 FirstCppHashFilename = Filename; 2474 return false; 2475 } 2476 2477 /// will use the last parsed cpp hash line filename comment 2478 /// for the Filename and LineNo if any in the diagnostic. 2479 void MasmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2480 const MasmParser *Parser = static_cast<const MasmParser *>(Context); 2481 raw_ostream &OS = errs(); 2482 2483 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2484 SMLoc DiagLoc = Diag.getLoc(); 2485 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2486 unsigned CppHashBuf = 2487 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2488 2489 // Like SourceMgr::printMessage() we need to print the include stack if any 2490 // before printing the message. 2491 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2492 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2493 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2494 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2495 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2496 } 2497 2498 // If we have not parsed a cpp hash line filename comment or the source 2499 // manager changed or buffer changed (like in a nested include) then just 2500 // print the normal diagnostic using its Filename and LineNo. 2501 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || 2502 DiagBuf != CppHashBuf) { 2503 if (Parser->SavedDiagHandler) 2504 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2505 else 2506 Diag.print(nullptr, OS); 2507 return; 2508 } 2509 2510 // Use the CppHashFilename and calculate a line number based on the 2511 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2512 // for the diagnostic. 2513 const std::string &Filename = std::string(Parser->CppHashInfo.Filename); 2514 2515 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2516 int CppHashLocLineNo = 2517 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2518 int LineNo = 2519 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2520 2521 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2522 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2523 Diag.getLineContents(), Diag.getRanges()); 2524 2525 if (Parser->SavedDiagHandler) 2526 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 2527 else 2528 NewDiag.print(nullptr, OS); 2529 } 2530 2531 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2532 // difference being that that function accepts '@' as part of identifiers and 2533 // we can't do that. AsmLexer.cpp should probably be changed to handle 2534 // '@' as a special case when needed. 2535 static bool isIdentifierChar(char c) { 2536 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2537 c == '.'; 2538 } 2539 2540 bool MasmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2541 ArrayRef<MCAsmMacroParameter> Parameters, 2542 ArrayRef<MCAsmMacroArgument> A, 2543 bool EnableAtPseudoVariable, SMLoc L) { 2544 unsigned NParameters = Parameters.size(); 2545 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2546 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2547 return Error(L, "Wrong number of arguments"); 2548 2549 // A macro without parameters is handled differently on Darwin: 2550 // gas accepts no arguments and does no substitutions 2551 while (!Body.empty()) { 2552 // Scan for the next substitution. 2553 std::size_t End = Body.size(), Pos = 0; 2554 for (; Pos != End; ++Pos) { 2555 // Check for a substitution or escape. 2556 if (IsDarwin && !NParameters) { 2557 // This macro has no parameters, look for $0, $1, etc. 2558 if (Body[Pos] != '$' || Pos + 1 == End) 2559 continue; 2560 2561 char Next = Body[Pos + 1]; 2562 if (Next == '$' || Next == 'n' || 2563 isdigit(static_cast<unsigned char>(Next))) 2564 break; 2565 } else { 2566 // This macro has parameters, look for \foo, \bar, etc. 2567 if (Body[Pos] == '\\' && Pos + 1 != End) 2568 break; 2569 } 2570 } 2571 2572 // Add the prefix. 2573 OS << Body.slice(0, Pos); 2574 2575 // Check if we reached the end. 2576 if (Pos == End) 2577 break; 2578 2579 if (IsDarwin && !NParameters) { 2580 switch (Body[Pos + 1]) { 2581 // $$ => $ 2582 case '$': 2583 OS << '$'; 2584 break; 2585 2586 // $n => number of arguments 2587 case 'n': 2588 OS << A.size(); 2589 break; 2590 2591 // $[0-9] => argument 2592 default: { 2593 // Missing arguments are ignored. 2594 unsigned Index = Body[Pos + 1] - '0'; 2595 if (Index >= A.size()) 2596 break; 2597 2598 // Otherwise substitute with the token values, with spaces eliminated. 2599 for (const AsmToken &Token : A[Index]) 2600 OS << Token.getString(); 2601 break; 2602 } 2603 } 2604 Pos += 2; 2605 } else { 2606 unsigned I = Pos + 1; 2607 2608 // Check for the \@ pseudo-variable. 2609 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2610 ++I; 2611 else 2612 while (isIdentifierChar(Body[I]) && I + 1 != End) 2613 ++I; 2614 2615 const char *Begin = Body.data() + Pos + 1; 2616 StringRef Argument(Begin, I - (Pos + 1)); 2617 unsigned Index = 0; 2618 2619 if (Argument == "@") { 2620 OS << NumOfMacroInstantiations; 2621 Pos += 2; 2622 } else { 2623 for (; Index < NParameters; ++Index) 2624 if (Parameters[Index].Name == Argument) 2625 break; 2626 2627 if (Index == NParameters) { 2628 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2629 Pos += 3; 2630 else { 2631 OS << '\\' << Argument; 2632 Pos = I; 2633 } 2634 } else { 2635 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2636 for (const AsmToken &Token : A[Index]) 2637 // For altmacro mode, you can write '%expr'. 2638 // The prefix '%' evaluates the expression 'expr' 2639 // and uses the result as a string (e.g. replace %(1+2) with the 2640 // string "3"). 2641 // Here, we identify the integer token which is the result of the 2642 // absolute expression evaluation and replace it with its string 2643 // representation. 2644 if (AltMacroMode && Token.getString().front() == '%' && 2645 Token.is(AsmToken::Integer)) 2646 // Emit an integer value to the buffer. 2647 OS << Token.getIntVal(); 2648 // Only Token that was validated as a string and begins with '<' 2649 // is considered altMacroString!!! 2650 else if (AltMacroMode && Token.getString().front() == '<' && 2651 Token.is(AsmToken::String)) { 2652 OS << angleBracketString(Token.getStringContents()); 2653 } 2654 // We expect no quotes around the string's contents when 2655 // parsing for varargs. 2656 else if (Token.isNot(AsmToken::String) || VarargParameter) 2657 OS << Token.getString(); 2658 else 2659 OS << Token.getStringContents(); 2660 2661 Pos += 1 + Argument.size(); 2662 } 2663 } 2664 } 2665 // Update the scan point. 2666 Body = Body.substr(Pos); 2667 } 2668 2669 return false; 2670 } 2671 2672 static bool isOperator(AsmToken::TokenKind kind) { 2673 switch (kind) { 2674 default: 2675 return false; 2676 case AsmToken::Plus: 2677 case AsmToken::Minus: 2678 case AsmToken::Tilde: 2679 case AsmToken::Slash: 2680 case AsmToken::Star: 2681 case AsmToken::Dot: 2682 case AsmToken::Equal: 2683 case AsmToken::EqualEqual: 2684 case AsmToken::Pipe: 2685 case AsmToken::PipePipe: 2686 case AsmToken::Caret: 2687 case AsmToken::Amp: 2688 case AsmToken::AmpAmp: 2689 case AsmToken::Exclaim: 2690 case AsmToken::ExclaimEqual: 2691 case AsmToken::Less: 2692 case AsmToken::LessEqual: 2693 case AsmToken::LessLess: 2694 case AsmToken::LessGreater: 2695 case AsmToken::Greater: 2696 case AsmToken::GreaterEqual: 2697 case AsmToken::GreaterGreater: 2698 return true; 2699 } 2700 } 2701 2702 namespace { 2703 2704 class AsmLexerSkipSpaceRAII { 2705 public: 2706 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2707 Lexer.setSkipSpace(SkipSpace); 2708 } 2709 2710 ~AsmLexerSkipSpaceRAII() { 2711 Lexer.setSkipSpace(true); 2712 } 2713 2714 private: 2715 AsmLexer &Lexer; 2716 }; 2717 2718 } // end anonymous namespace 2719 2720 bool MasmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2721 2722 if (Vararg) { 2723 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2724 StringRef Str = parseStringToEndOfStatement(); 2725 MA.emplace_back(AsmToken::String, Str); 2726 } 2727 return false; 2728 } 2729 2730 unsigned ParenLevel = 0; 2731 2732 // Darwin doesn't use spaces to delmit arguments. 2733 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2734 2735 bool SpaceEaten; 2736 2737 while (true) { 2738 SpaceEaten = false; 2739 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2740 return TokError("unexpected token in macro instantiation"); 2741 2742 if (ParenLevel == 0) { 2743 2744 if (Lexer.is(AsmToken::Comma)) 2745 break; 2746 2747 if (Lexer.is(AsmToken::Space)) { 2748 SpaceEaten = true; 2749 Lexer.Lex(); // Eat spaces. 2750 } 2751 2752 // Spaces can delimit parameters, but could also be part an expression. 2753 // If the token after a space is an operator, add the token and the next 2754 // one into this argument 2755 if (!IsDarwin) { 2756 if (isOperator(Lexer.getKind())) { 2757 MA.push_back(getTok()); 2758 Lexer.Lex(); 2759 2760 // Whitespace after an operator can be ignored. 2761 if (Lexer.is(AsmToken::Space)) 2762 Lexer.Lex(); 2763 2764 continue; 2765 } 2766 } 2767 if (SpaceEaten) 2768 break; 2769 } 2770 2771 // handleMacroEntry relies on not advancing the lexer here 2772 // to be able to fill in the remaining default parameter values 2773 if (Lexer.is(AsmToken::EndOfStatement)) 2774 break; 2775 2776 // Adjust the current parentheses level. 2777 if (Lexer.is(AsmToken::LParen)) 2778 ++ParenLevel; 2779 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2780 --ParenLevel; 2781 2782 // Append the token to the current argument list. 2783 MA.push_back(getTok()); 2784 Lexer.Lex(); 2785 } 2786 2787 if (ParenLevel != 0) 2788 return TokError("unbalanced parentheses in macro argument"); 2789 return false; 2790 } 2791 2792 // Parse the macro instantiation arguments. 2793 bool MasmParser::parseMacroArguments(const MCAsmMacro *M, 2794 MCAsmMacroArguments &A) { 2795 const unsigned NParameters = M ? M->Parameters.size() : 0; 2796 bool NamedParametersFound = false; 2797 SmallVector<SMLoc, 4> FALocs; 2798 2799 A.resize(NParameters); 2800 FALocs.resize(NParameters); 2801 2802 // Parse two kinds of macro invocations: 2803 // - macros defined without any parameters accept an arbitrary number of them 2804 // - macros defined with parameters accept at most that many of them 2805 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2806 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2807 ++Parameter) { 2808 SMLoc IDLoc = Lexer.getLoc(); 2809 MCAsmMacroParameter FA; 2810 2811 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2812 if (parseIdentifier(FA.Name)) 2813 return Error(IDLoc, "invalid argument identifier for formal argument"); 2814 2815 if (Lexer.isNot(AsmToken::Equal)) 2816 return TokError("expected '=' after formal parameter identifier"); 2817 2818 Lex(); 2819 2820 NamedParametersFound = true; 2821 } 2822 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2823 2824 if (NamedParametersFound && FA.Name.empty()) 2825 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2826 2827 SMLoc StrLoc = Lexer.getLoc(); 2828 SMLoc EndLoc; 2829 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2830 const MCExpr *AbsoluteExp; 2831 int64_t Value; 2832 /// Eat '%'. 2833 Lex(); 2834 if (parseExpression(AbsoluteExp, EndLoc)) 2835 return false; 2836 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2837 getStreamer().getAssemblerPtr())) 2838 return Error(StrLoc, "expected absolute expression"); 2839 const char *StrChar = StrLoc.getPointer(); 2840 const char *EndChar = EndLoc.getPointer(); 2841 AsmToken newToken(AsmToken::Integer, 2842 StringRef(StrChar, EndChar - StrChar), Value); 2843 FA.Value.push_back(newToken); 2844 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2845 isAngleBracketString(StrLoc, EndLoc)) { 2846 const char *StrChar = StrLoc.getPointer(); 2847 const char *EndChar = EndLoc.getPointer(); 2848 jumpToLoc(EndLoc, CurBuffer); 2849 /// Eat from '<' to '>'. 2850 Lex(); 2851 AsmToken newToken(AsmToken::String, 2852 StringRef(StrChar, EndChar - StrChar)); 2853 FA.Value.push_back(newToken); 2854 } else if(parseMacroArgument(FA.Value, Vararg)) 2855 return true; 2856 2857 unsigned PI = Parameter; 2858 if (!FA.Name.empty()) { 2859 unsigned FAI = 0; 2860 for (FAI = 0; FAI < NParameters; ++FAI) 2861 if (M->Parameters[FAI].Name == FA.Name) 2862 break; 2863 2864 if (FAI >= NParameters) { 2865 assert(M && "expected macro to be defined"); 2866 return Error(IDLoc, "parameter named '" + FA.Name + 2867 "' does not exist for macro '" + M->Name + "'"); 2868 } 2869 PI = FAI; 2870 } 2871 2872 if (!FA.Value.empty()) { 2873 if (A.size() <= PI) 2874 A.resize(PI + 1); 2875 A[PI] = FA.Value; 2876 2877 if (FALocs.size() <= PI) 2878 FALocs.resize(PI + 1); 2879 2880 FALocs[PI] = Lexer.getLoc(); 2881 } 2882 2883 // At the end of the statement, fill in remaining arguments that have 2884 // default values. If there aren't any, then the next argument is 2885 // required but missing 2886 if (Lexer.is(AsmToken::EndOfStatement)) { 2887 bool Failure = false; 2888 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2889 if (A[FAI].empty()) { 2890 if (M->Parameters[FAI].Required) { 2891 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2892 "missing value for required parameter " 2893 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2894 Failure = true; 2895 } 2896 2897 if (!M->Parameters[FAI].Value.empty()) 2898 A[FAI] = M->Parameters[FAI].Value; 2899 } 2900 } 2901 return Failure; 2902 } 2903 2904 if (Lexer.is(AsmToken::Comma)) 2905 Lex(); 2906 } 2907 2908 return TokError("too many positional arguments"); 2909 } 2910 2911 bool MasmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2912 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2913 // eliminate this, although we should protect against infinite loops. 2914 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2915 if (ActiveMacros.size() == MaxNestingDepth) { 2916 std::ostringstream MaxNestingDepthError; 2917 MaxNestingDepthError << "macros cannot be nested more than " 2918 << MaxNestingDepth << " levels deep." 2919 << " Use -asm-macro-max-nesting-depth to increase " 2920 "this limit."; 2921 return TokError(MaxNestingDepthError.str()); 2922 } 2923 2924 MCAsmMacroArguments A; 2925 if (parseMacroArguments(M, A)) 2926 return true; 2927 2928 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2929 // to hold the macro body with substitutions. 2930 SmallString<256> Buf; 2931 StringRef Body = M->Body; 2932 raw_svector_ostream OS(Buf); 2933 2934 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2935 return true; 2936 2937 // We include the .endmacro in the buffer as our cue to exit the macro 2938 // instantiation. 2939 OS << ".endmacro\n"; 2940 2941 std::unique_ptr<MemoryBuffer> Instantiation = 2942 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2943 2944 // Create the macro instantiation object and add to the current macro 2945 // instantiation stack. 2946 MacroInstantiation *MI = new MacroInstantiation{ 2947 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 2948 ActiveMacros.push_back(MI); 2949 2950 ++NumOfMacroInstantiations; 2951 2952 // Jump to the macro instantiation and prime the lexer. 2953 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2954 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2955 Lex(); 2956 2957 return false; 2958 } 2959 2960 void MasmParser::handleMacroExit() { 2961 // Jump to the EndOfStatement we should return to, and consume it. 2962 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2963 Lex(); 2964 2965 // Pop the instantiation entry. 2966 delete ActiveMacros.back(); 2967 ActiveMacros.pop_back(); 2968 } 2969 2970 /// parseIdentifier: 2971 /// ::= identifier 2972 /// ::= string 2973 bool MasmParser::parseIdentifier(StringRef &Res) { 2974 // The assembler has relaxed rules for accepting identifiers, in particular we 2975 // allow things like '.globl $foo' and '.def @feat.00', which would normally 2976 // be separate tokens. At this level, we have already lexed so we cannot 2977 // (currently) handle this as a context dependent token, instead we detect 2978 // adjacent tokens and return the combined identifier. 2979 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2980 SMLoc PrefixLoc = getLexer().getLoc(); 2981 2982 // Consume the prefix character, and check for a following identifier. 2983 2984 AsmToken Buf[1]; 2985 Lexer.peekTokens(Buf, false); 2986 2987 if (Buf[0].isNot(AsmToken::Identifier)) 2988 return true; 2989 2990 // We have a '$' or '@' followed by an identifier, make sure they are adjacent. 2991 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2992 return true; 2993 2994 // eat $ or @ 2995 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 2996 // Construct the joined identifier and consume the token. 2997 Res = 2998 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); 2999 Lex(); // Parser Lex to maintain invariants. 3000 return false; 3001 } 3002 3003 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 3004 return true; 3005 3006 Res = getTok().getIdentifier(); 3007 3008 Lex(); // Consume the identifier token. 3009 3010 return false; 3011 } 3012 3013 /// parseDirectiveEquate: 3014 /// ::= name "=" expression 3015 /// | name "equ" expression (not redefinable) 3016 /// | name "equ" text-list 3017 /// | name "textequ" text-list 3018 bool MasmParser::parseDirectiveEquate(StringRef IDVal, StringRef Name, 3019 DirectiveKind DirKind) { 3020 Variable &Var = Variables[Name]; 3021 if (Var.Name.empty()) { 3022 Var.Name = Name; 3023 } else if (!Var.Redefinable) { 3024 return TokError("invalid variable redefinition"); 3025 } 3026 Var.Redefinable = (DirKind != DK_EQU); 3027 3028 if (DirKind == DK_EQU || DirKind == DK_TEXTEQU) { 3029 // "equ" and "textequ" both allow text expressions. 3030 std::string Value; 3031 if (!parseTextItem(Value)) { 3032 Var.IsText = true; 3033 Var.TextValue = Value; 3034 3035 // Accept a text-list, not just one text-item. 3036 auto parseItem = [&]() -> bool { 3037 if (parseTextItem(Value)) 3038 return true; 3039 Var.TextValue += Value; 3040 return false; 3041 }; 3042 if (parseOptionalToken(AsmToken::Comma) && parseMany(parseItem)) 3043 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3044 3045 return false; 3046 } 3047 } 3048 if (DirKind == DK_TEXTEQU) 3049 return TokError("expected <text> in '" + Twine(IDVal) + "' directive"); 3050 3051 // Parse as expression assignment. 3052 const MCExpr *Expr; 3053 SMLoc EndLoc, StartLoc = Lexer.getLoc(); 3054 if (parseExpression(Expr, EndLoc)) 3055 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3056 if (Expr->evaluateAsAbsolute(Var.NumericValue, 3057 getStreamer().getAssemblerPtr())) 3058 return false; 3059 3060 // Not an absolute expression; define as a text replacement. 3061 Var.IsText = true; 3062 Var.TextValue = StringRef(StartLoc.getPointer(), 3063 EndLoc.getPointer() - StartLoc.getPointer()).str(); 3064 return false; 3065 } 3066 3067 bool MasmParser::parseEscapedString(std::string &Data) { 3068 if (check(getTok().isNot(AsmToken::String), "expected string")) 3069 return true; 3070 3071 Data = ""; 3072 StringRef Str = getTok().getStringContents(); 3073 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 3074 if (Str[i] != '\\') { 3075 Data += Str[i]; 3076 continue; 3077 } 3078 3079 // Recognize escaped characters. Note that this escape semantics currently 3080 // loosely follows Darwin 'as'. 3081 ++i; 3082 if (i == e) 3083 return TokError("unexpected backslash at end of string"); 3084 3085 // Recognize hex sequences similarly to GNU 'as'. 3086 if (Str[i] == 'x' || Str[i] == 'X') { 3087 size_t length = Str.size(); 3088 if (i + 1 >= length || !isHexDigit(Str[i + 1])) 3089 return TokError("invalid hexadecimal escape sequence"); 3090 3091 // Consume hex characters. GNU 'as' reads all hexadecimal characters and 3092 // then truncates to the lower 16 bits. Seems reasonable. 3093 unsigned Value = 0; 3094 while (i + 1 < length && isHexDigit(Str[i + 1])) 3095 Value = Value * 16 + hexDigitValue(Str[++i]); 3096 3097 Data += (unsigned char)(Value & 0xFF); 3098 continue; 3099 } 3100 3101 // Recognize octal sequences. 3102 if ((unsigned)(Str[i] - '0') <= 7) { 3103 // Consume up to three octal characters. 3104 unsigned Value = Str[i] - '0'; 3105 3106 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3107 ++i; 3108 Value = Value * 8 + (Str[i] - '0'); 3109 3110 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3111 ++i; 3112 Value = Value * 8 + (Str[i] - '0'); 3113 } 3114 } 3115 3116 if (Value > 255) 3117 return TokError("invalid octal escape sequence (out of range)"); 3118 3119 Data += (unsigned char)Value; 3120 continue; 3121 } 3122 3123 // Otherwise recognize individual escapes. 3124 switch (Str[i]) { 3125 default: 3126 // Just reject invalid escape sequences for now. 3127 return TokError("invalid escape sequence (unrecognized character)"); 3128 3129 case 'b': Data += '\b'; break; 3130 case 'f': Data += '\f'; break; 3131 case 'n': Data += '\n'; break; 3132 case 'r': Data += '\r'; break; 3133 case 't': Data += '\t'; break; 3134 case '"': Data += '"'; break; 3135 case '\\': Data += '\\'; break; 3136 } 3137 } 3138 3139 Lex(); 3140 return false; 3141 } 3142 3143 bool MasmParser::parseAngleBracketString(std::string &Data) { 3144 SMLoc EndLoc, StartLoc = getTok().getLoc(); 3145 if (isAngleBracketString(StartLoc, EndLoc)) { 3146 const char *StartChar = StartLoc.getPointer() + 1; 3147 const char *EndChar = EndLoc.getPointer() - 1; 3148 jumpToLoc(EndLoc, CurBuffer); 3149 // Eat from '<' to '>'. 3150 Lex(); 3151 3152 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar)); 3153 return false; 3154 } 3155 return true; 3156 } 3157 3158 /// textItem ::= textLiteral | textMacroID | % constExpr 3159 bool MasmParser::parseTextItem(std::string &Data) { 3160 // TODO(epastor): Support textMacroID and % expansion of expressions. 3161 return parseAngleBracketString(Data); 3162 } 3163 3164 /// parseDirectiveAscii: 3165 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 3166 bool MasmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 3167 auto parseOp = [&]() -> bool { 3168 std::string Data; 3169 if (checkForValidSection() || parseEscapedString(Data)) 3170 return true; 3171 getStreamer().emitBytes(Data); 3172 if (ZeroTerminated) 3173 getStreamer().emitBytes(StringRef("\0", 1)); 3174 return false; 3175 }; 3176 3177 if (parseMany(parseOp)) 3178 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3179 return false; 3180 } 3181 3182 bool MasmParser::emitIntValue(const MCExpr *Value, unsigned Size) { 3183 // Special case constant expressions to match code generator. 3184 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3185 assert(Size <= 8 && "Invalid size"); 3186 int64_t IntValue = MCE->getValue(); 3187 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3188 return Error(MCE->getLoc(), "out of range literal value"); 3189 getStreamer().emitIntValue(IntValue, Size); 3190 } else { 3191 const MCSymbolRefExpr *MSE = dyn_cast<MCSymbolRefExpr>(Value); 3192 if (MSE && MSE->getSymbol().getName() == "?") { 3193 // ? initializer; treat as 0. 3194 getStreamer().emitIntValue(0, Size); 3195 } else { 3196 getStreamer().emitValue(Value, Size, Value->getLoc()); 3197 } 3198 } 3199 return false; 3200 } 3201 3202 bool MasmParser::parseScalarInitializer(unsigned Size, 3203 SmallVectorImpl<const MCExpr *> &Values, 3204 unsigned StringPadLength) { 3205 if (getTok().is(AsmToken::String)) { 3206 StringRef Value = getTok().getStringContents(); 3207 if (Size == 1) { 3208 // Treat each character as an initializer. 3209 for (const char CharVal : Value) 3210 Values.push_back(MCConstantExpr::create(CharVal, getContext())); 3211 3212 // Pad the string with spaces to the specified length. 3213 for (size_t i = Value.size(); i < StringPadLength; ++i) 3214 Values.push_back(MCConstantExpr::create(' ', getContext())); 3215 } else { 3216 // Treat the string as an initial value in big-endian representation. 3217 if (Value.size() > Size) 3218 return Error(getTok().getLoc(), "out of range literal value"); 3219 3220 uint64_t IntValue = 0; 3221 for (const unsigned char CharVal : Value.bytes()) 3222 IntValue = (IntValue << 8) | CharVal; 3223 Values.push_back(MCConstantExpr::create(IntValue, getContext())); 3224 } 3225 Lex(); 3226 } else { 3227 const MCExpr *Value; 3228 if (checkForValidSection() || parseExpression(Value)) 3229 return true; 3230 if (getTok().is(AsmToken::Identifier) && 3231 getTok().getString().equals_lower("dup")) { 3232 Lex(); // Eat 'dup'. 3233 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3234 if (!MCE) 3235 return Error(Value->getLoc(), 3236 "cannot repeat value a non-constant number of times"); 3237 const int64_t Repetitions = MCE->getValue(); 3238 if (Repetitions < 0) 3239 return Error(Value->getLoc(), 3240 "cannot repeat value a negative number of times"); 3241 3242 SmallVector<const MCExpr *, 1> DuplicatedValues; 3243 if (parseToken(AsmToken::LParen, 3244 "parentheses required for 'dup' contents") || 3245 parseScalarInstList(Size, DuplicatedValues) || 3246 parseToken(AsmToken::RParen, "unmatched parentheses")) 3247 return true; 3248 3249 for (int i = 0; i < Repetitions; ++i) 3250 Values.append(DuplicatedValues.begin(), DuplicatedValues.end()); 3251 } else { 3252 Values.push_back(Value); 3253 } 3254 } 3255 return false; 3256 } 3257 3258 bool MasmParser::parseScalarInstList(unsigned Size, 3259 SmallVectorImpl<const MCExpr *> &Values, 3260 const AsmToken::TokenKind EndToken) { 3261 while (getTok().isNot(EndToken) && 3262 (EndToken != AsmToken::Greater || 3263 getTok().isNot(AsmToken::GreaterGreater))) { 3264 parseScalarInitializer(Size, Values); 3265 3266 // If we see a comma, continue, and allow line continuation. 3267 if (!parseOptionalToken(AsmToken::Comma)) 3268 break; 3269 parseOptionalToken(AsmToken::EndOfStatement); 3270 } 3271 return false; 3272 } 3273 3274 bool MasmParser::emitIntegralValues(unsigned Size) { 3275 SmallVector<const MCExpr *, 1> Values; 3276 if (checkForValidSection() || parseScalarInstList(Size, Values)) 3277 return true; 3278 3279 for (auto Value : Values) { 3280 emitIntValue(Value, Size); 3281 } 3282 return false; 3283 } 3284 3285 // Add a field to the current structure. 3286 bool MasmParser::addIntegralField(StringRef Name, unsigned Size) { 3287 StructInfo &Struct = StructInProgress.back(); 3288 FieldInfo &Field = Struct.addField(Name, FT_INTEGRAL); 3289 IntFieldInfo &IntInfo = Field.Contents.IntInfo; 3290 3291 Field.Type = Size; 3292 3293 if (parseScalarInstList(Size, IntInfo.Values)) 3294 return true; 3295 3296 Field.SizeOf = Field.Type * IntInfo.Values.size(); 3297 Field.LengthOf = IntInfo.Values.size(); 3298 if (Struct.IsUnion) 3299 Struct.Size = std::max(Struct.Size, Field.SizeOf); 3300 else 3301 Struct.Size += Field.SizeOf; 3302 return false; 3303 } 3304 3305 /// parseDirectiveValue 3306 /// ::= (byte | word | ... ) [ expression (, expression)* ] 3307 bool MasmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3308 if (StructInProgress.empty()) { 3309 // Initialize data value. 3310 if (emitIntegralValues(Size)) 3311 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3312 } else if (addIntegralField("", Size)) { 3313 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3314 } 3315 3316 return false; 3317 } 3318 3319 /// parseDirectiveNamedValue 3320 /// ::= name (byte | word | ... ) [ expression (, expression)* ] 3321 bool MasmParser::parseDirectiveNamedValue(StringRef IDVal, unsigned Size, 3322 StringRef Name, SMLoc NameLoc) { 3323 if (StructInProgress.empty()) { 3324 // Initialize named data value. 3325 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 3326 getStreamer().emitLabel(Sym); 3327 if (emitIntegralValues(Size)) 3328 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3329 } else if (addIntegralField(Name, Size)) { 3330 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3331 } 3332 3333 return false; 3334 } 3335 3336 static bool parseHexOcta(MasmParser &Asm, uint64_t &hi, uint64_t &lo) { 3337 if (Asm.getTok().isNot(AsmToken::Integer) && 3338 Asm.getTok().isNot(AsmToken::BigNum)) 3339 return Asm.TokError("unknown token in expression"); 3340 SMLoc ExprLoc = Asm.getTok().getLoc(); 3341 APInt IntValue = Asm.getTok().getAPIntVal(); 3342 Asm.Lex(); 3343 if (!IntValue.isIntN(128)) 3344 return Asm.Error(ExprLoc, "out of range literal value"); 3345 if (!IntValue.isIntN(64)) { 3346 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3347 lo = IntValue.getLoBits(64).getZExtValue(); 3348 } else { 3349 hi = 0; 3350 lo = IntValue.getZExtValue(); 3351 } 3352 return false; 3353 } 3354 3355 bool MasmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3356 // We don't truly support arithmetic on floating point expressions, so we 3357 // have to manually parse unary prefixes. 3358 bool IsNeg = false; 3359 if (getLexer().is(AsmToken::Minus)) { 3360 Lexer.Lex(); 3361 IsNeg = true; 3362 } else if (getLexer().is(AsmToken::Plus)) { 3363 Lexer.Lex(); 3364 } 3365 3366 if (Lexer.is(AsmToken::Error)) 3367 return TokError(Lexer.getErr()); 3368 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3369 Lexer.isNot(AsmToken::Identifier)) 3370 return TokError("unexpected token in directive"); 3371 3372 // Convert to an APFloat. 3373 APFloat Value(Semantics); 3374 StringRef IDVal = getTok().getString(); 3375 if (getLexer().is(AsmToken::Identifier)) { 3376 if (IDVal.equals_lower("infinity") || IDVal.equals_lower("inf")) 3377 Value = APFloat::getInf(Semantics); 3378 else if (IDVal.equals_lower("nan")) 3379 Value = APFloat::getNaN(Semantics, false, ~0); 3380 else if (IDVal.equals_lower("?")) 3381 Value = APFloat::getZero(Semantics); 3382 else 3383 return TokError("invalid floating point literal"); 3384 } else if (errorToBool( 3385 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) 3386 .takeError())) { 3387 return TokError("invalid floating point literal"); 3388 } 3389 if (IsNeg) 3390 Value.changeSign(); 3391 3392 // Consume the numeric token. 3393 Lex(); 3394 3395 Res = Value.bitcastToAPInt(); 3396 3397 return false; 3398 } 3399 3400 bool MasmParser::parseRealInstList(const fltSemantics &Semantics, 3401 SmallVectorImpl<APInt> &ValuesAsInt, 3402 const AsmToken::TokenKind EndToken) { 3403 while (getTok().isNot(EndToken) || 3404 (EndToken == AsmToken::Greater && 3405 getTok().isNot(AsmToken::GreaterGreater))) { 3406 const AsmToken NextTok = Lexer.peekTok(); 3407 if (NextTok.is(AsmToken::Identifier) && 3408 NextTok.getString().equals_lower("dup")) { 3409 const MCExpr *Value; 3410 if (parseExpression(Value) || parseToken(AsmToken::Identifier)) 3411 return true; 3412 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3413 if (!MCE) 3414 return Error(Value->getLoc(), 3415 "cannot repeat value a non-constant number of times"); 3416 const int64_t Repetitions = MCE->getValue(); 3417 if (Repetitions < 0) 3418 return Error(Value->getLoc(), 3419 "cannot repeat value a negative number of times"); 3420 3421 SmallVector<APInt, 1> DuplicatedValues; 3422 if (parseToken(AsmToken::LParen, 3423 "parentheses required for 'dup' contents") || 3424 parseRealInstList(Semantics, DuplicatedValues) || 3425 parseToken(AsmToken::RParen, "unmatched parentheses")) 3426 return true; 3427 3428 for (int i = 0; i < Repetitions; ++i) 3429 ValuesAsInt.append(DuplicatedValues.begin(), DuplicatedValues.end()); 3430 } else { 3431 APInt AsInt; 3432 if (parseRealValue(Semantics, AsInt)) 3433 return true; 3434 ValuesAsInt.push_back(AsInt); 3435 } 3436 3437 // Continue if we see a comma. (Also, allow line continuation.) 3438 if (!parseOptionalToken(AsmToken::Comma)) 3439 break; 3440 parseOptionalToken(AsmToken::EndOfStatement); 3441 } 3442 3443 return false; 3444 } 3445 3446 // Initialize real data values. 3447 bool MasmParser::emitRealValues(const fltSemantics &Semantics) { 3448 SmallVector<APInt, 1> ValuesAsInt; 3449 if (parseRealInstList(Semantics, ValuesAsInt)) 3450 return true; 3451 3452 for (const APInt &AsInt : ValuesAsInt) { 3453 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3454 AsInt.getBitWidth() / 8); 3455 } 3456 return false; 3457 } 3458 3459 // Add a real field to the current struct. 3460 bool MasmParser::addRealField(StringRef Name, const fltSemantics &Semantics) { 3461 StructInfo &Struct = StructInProgress.back(); 3462 FieldInfo &Field = Struct.addField(Name, FT_REAL); 3463 RealFieldInfo &RealInfo = Field.Contents.RealInfo; 3464 3465 Field.SizeOf = 0; 3466 3467 if (checkForValidSection() || 3468 parseRealInstList(Semantics, RealInfo.AsIntValues)) 3469 return true; 3470 3471 Field.Type = RealInfo.AsIntValues.back().getBitWidth() / 8; 3472 Field.LengthOf = RealInfo.AsIntValues.size(); 3473 Field.SizeOf = Field.Type * Field.LengthOf; 3474 if (Struct.IsUnion) 3475 Struct.Size = std::max(Struct.Size, Field.SizeOf); 3476 else 3477 Struct.Size += Field.SizeOf; 3478 return false; 3479 } 3480 3481 /// parseDirectiveRealValue 3482 /// ::= (real4 | real8) [ expression (, expression)* ] 3483 bool MasmParser::parseDirectiveRealValue(StringRef IDVal, 3484 const fltSemantics &Semantics) { 3485 if (checkForValidSection()) 3486 return true; 3487 3488 if (StructInProgress.empty()) { 3489 // Initialize data value. 3490 if (emitRealValues(Semantics)) 3491 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3492 } else if (addRealField("", Semantics)) { 3493 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3494 } 3495 return false; 3496 } 3497 3498 /// parseDirectiveNamedRealValue 3499 /// ::= name (real4 | real8) [ expression (, expression)* ] 3500 bool MasmParser::parseDirectiveNamedRealValue(StringRef IDVal, 3501 const fltSemantics &Semantics, 3502 StringRef Name, SMLoc NameLoc) { 3503 if (checkForValidSection()) 3504 return true; 3505 3506 if (StructInProgress.empty()) { 3507 // Initialize named data value. 3508 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 3509 getStreamer().emitLabel(Sym); 3510 if (emitRealValues(Semantics)) 3511 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3512 } else if (addRealField(Name, Semantics)) { 3513 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive"); 3514 } 3515 return false; 3516 } 3517 3518 bool MasmParser::parseOptionalAngleBracketOpen() { 3519 const AsmToken Tok = getTok(); 3520 if (parseOptionalToken(AsmToken::LessLess)) { 3521 AngleBracketDepth++; 3522 Lexer.UnLex(AsmToken(AsmToken::Less, Tok.getString().substr(1))); 3523 return true; 3524 } else if (parseOptionalToken(AsmToken::LessGreater)) { 3525 AngleBracketDepth++; 3526 Lexer.UnLex(AsmToken(AsmToken::Greater, Tok.getString().substr(1))); 3527 return true; 3528 } else if (parseOptionalToken(AsmToken::Less)) { 3529 AngleBracketDepth++; 3530 return true; 3531 } 3532 3533 return false; 3534 } 3535 3536 bool MasmParser::parseAngleBracketClose(const Twine &Msg) { 3537 const AsmToken Tok = getTok(); 3538 if (parseOptionalToken(AsmToken::GreaterGreater)) { 3539 Lexer.UnLex(AsmToken(AsmToken::Greater, Tok.getString().substr(1))); 3540 } else if (parseToken(AsmToken::Greater, Msg)) { 3541 return true; 3542 } 3543 AngleBracketDepth--; 3544 return false; 3545 } 3546 3547 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3548 const IntFieldInfo &Contents, 3549 FieldInitializer &Initializer) { 3550 SMLoc Loc = getTok().getLoc(); 3551 3552 SmallVector<const MCExpr *, 1> Values; 3553 if (parseOptionalToken(AsmToken::LCurly)) { 3554 if (Field.LengthOf == 1 && Field.Type > 1) 3555 return Error(Loc, "Cannot initialize scalar field with array value"); 3556 if (parseScalarInstList(Field.Type, Values, AsmToken::RCurly) || 3557 parseToken(AsmToken::RCurly)) 3558 return true; 3559 } else if (parseOptionalAngleBracketOpen()) { 3560 if (Field.LengthOf == 1 && Field.Type > 1) 3561 return Error(Loc, "Cannot initialize scalar field with array value"); 3562 if (parseScalarInstList(Field.Type, Values, AsmToken::Greater) || 3563 parseAngleBracketClose()) 3564 return true; 3565 } else if (Field.LengthOf > 1 && Field.Type > 1) { 3566 return Error(Loc, "Cannot initialize array field with scalar value"); 3567 } else if (parseScalarInitializer(Field.Type, Values, 3568 /*StringPadLength=*/Field.LengthOf)) { 3569 return true; 3570 } 3571 3572 if (Values.size() > Field.LengthOf) { 3573 return Error(Loc, "Initializer too long for field; expected at most " + 3574 std::to_string(Field.LengthOf) + " elements, got " + 3575 std::to_string(Values.size())); 3576 } 3577 // Default-initialize all remaining values. 3578 Values.append(Contents.Values.begin() + Values.size(), Contents.Values.end()); 3579 3580 Initializer = FieldInitializer(std::move(Values)); 3581 return false; 3582 } 3583 3584 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3585 const RealFieldInfo &Contents, 3586 FieldInitializer &Initializer) { 3587 const fltSemantics &Semantics = 3588 (Field.Type == 4) ? APFloat::IEEEsingle() : APFloat::IEEEdouble(); 3589 3590 SMLoc Loc = getTok().getLoc(); 3591 3592 SmallVector<APInt, 1> AsIntValues; 3593 if (parseOptionalToken(AsmToken::LCurly)) { 3594 if (Field.LengthOf == 1) 3595 return Error(Loc, "Cannot initialize scalar field with array value"); 3596 if (parseRealInstList(Semantics, AsIntValues, AsmToken::RCurly) || 3597 parseToken(AsmToken::RCurly)) 3598 return true; 3599 } else if (parseOptionalAngleBracketOpen()) { 3600 if (Field.LengthOf == 1) 3601 return Error(Loc, "Cannot initialize scalar field with array value"); 3602 if (parseRealInstList(Semantics, AsIntValues, AsmToken::Greater) || 3603 parseAngleBracketClose()) 3604 return true; 3605 } else if (Field.LengthOf > 1) { 3606 return Error(Loc, "Cannot initialize array field with scalar value"); 3607 } else { 3608 AsIntValues.emplace_back(); 3609 if (parseRealValue(Semantics, AsIntValues.back())) 3610 return true; 3611 } 3612 3613 if (AsIntValues.size() > Field.LengthOf) { 3614 return Error(Loc, "Initializer too long for field; expected at most " + 3615 std::to_string(Field.LengthOf) + " elements, got " + 3616 std::to_string(AsIntValues.size())); 3617 } 3618 // Default-initialize all remaining values. 3619 AsIntValues.append(Contents.AsIntValues.begin() + AsIntValues.size(), 3620 Contents.AsIntValues.end()); 3621 3622 Initializer = FieldInitializer(std::move(AsIntValues)); 3623 return false; 3624 } 3625 3626 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3627 const StructFieldInfo &Contents, 3628 FieldInitializer &Initializer) { 3629 SMLoc Loc = getTok().getLoc(); 3630 3631 std::vector<StructInitializer> Initializers; 3632 if (Field.LengthOf > 1) { 3633 if (parseOptionalToken(AsmToken::LCurly)) { 3634 if (parseStructInstList(Contents.Structure, Initializers, 3635 AsmToken::RCurly) || 3636 parseToken(AsmToken::RCurly)) 3637 return true; 3638 } else if (parseOptionalAngleBracketOpen()) { 3639 if (parseStructInstList(Contents.Structure, Initializers, 3640 AsmToken::Greater) || 3641 parseAngleBracketClose()) 3642 return true; 3643 } else { 3644 return Error(Loc, "Cannot initialize array field with scalar value"); 3645 } 3646 } else { 3647 Initializers.emplace_back(); 3648 if (parseStructInitializer(Contents.Structure, Initializers.back())) 3649 return true; 3650 } 3651 3652 if (Initializers.size() > Field.LengthOf) { 3653 return Error(Loc, "Initializer too long for field; expected at most " + 3654 std::to_string(Field.LengthOf) + " elements, got " + 3655 std::to_string(Initializers.size())); 3656 } 3657 // Default-initialize all remaining values. 3658 Initializers.insert(Initializers.end(), 3659 Contents.Initializers.begin() + Initializers.size(), 3660 Contents.Initializers.end()); 3661 3662 Initializer = FieldInitializer(std::move(Initializers), Contents.Structure); 3663 return false; 3664 } 3665 3666 bool MasmParser::parseFieldInitializer(const FieldInfo &Field, 3667 FieldInitializer &Initializer) { 3668 switch (Field.Contents.FT) { 3669 case FT_INTEGRAL: 3670 return parseFieldInitializer(Field, Field.Contents.IntInfo, Initializer); 3671 case FT_REAL: 3672 return parseFieldInitializer(Field, Field.Contents.RealInfo, Initializer); 3673 case FT_STRUCT: 3674 return parseFieldInitializer(Field, Field.Contents.StructInfo, Initializer); 3675 } 3676 llvm_unreachable("Unhandled FieldType enum"); 3677 } 3678 3679 bool MasmParser::parseStructInitializer(const StructInfo &Structure, 3680 StructInitializer &Initializer) { 3681 const AsmToken FirstToken = getTok(); 3682 3683 Optional<AsmToken::TokenKind> EndToken; 3684 if (parseOptionalToken(AsmToken::LCurly)) { 3685 EndToken = AsmToken::RCurly; 3686 } else if (parseOptionalAngleBracketOpen()) { 3687 EndToken = AsmToken::Greater; 3688 AngleBracketDepth++; 3689 } else if (FirstToken.is(AsmToken::Identifier) && 3690 FirstToken.getString() == "?") { 3691 // ? initializer; leave EndToken uninitialized to treat as empty. 3692 if (parseToken(AsmToken::Identifier)) 3693 return true; 3694 } else { 3695 return Error(FirstToken.getLoc(), "Expected struct initializer"); 3696 } 3697 3698 auto &FieldInitializers = Initializer.FieldInitializers; 3699 size_t FieldIndex = 0; 3700 if (EndToken.hasValue()) { 3701 // Initialize all fields with given initializers. 3702 while (getTok().isNot(EndToken.getValue()) && 3703 FieldIndex < Structure.Fields.size()) { 3704 const FieldInfo &Field = Structure.Fields[FieldIndex++]; 3705 if (parseOptionalToken(AsmToken::Comma)) { 3706 // Empty initializer; use the default and continue. (Also, allow line 3707 // continuation.) 3708 FieldInitializers.push_back(Field.Contents); 3709 parseOptionalToken(AsmToken::EndOfStatement); 3710 continue; 3711 } 3712 FieldInitializers.emplace_back(Field.Contents.FT); 3713 if (parseFieldInitializer(Field, FieldInitializers.back())) 3714 return true; 3715 3716 // Continue if we see a comma. (Also, allow line continuation.) 3717 SMLoc CommaLoc = getTok().getLoc(); 3718 if (!parseOptionalToken(AsmToken::Comma)) 3719 break; 3720 if (FieldIndex == Structure.Fields.size()) 3721 return Error(CommaLoc, "'" + Structure.Name + 3722 "' initializer initializes too many fields"); 3723 parseOptionalToken(AsmToken::EndOfStatement); 3724 } 3725 } 3726 // Default-initialize all remaining fields. 3727 for (auto It = Structure.Fields.begin() + FieldIndex; 3728 It != Structure.Fields.end(); ++It) { 3729 const FieldInfo &Field = *It; 3730 FieldInitializers.push_back(Field.Contents); 3731 } 3732 3733 if (EndToken.hasValue()) { 3734 if (EndToken.getValue() == AsmToken::Greater) 3735 return parseAngleBracketClose(); 3736 3737 return parseToken(EndToken.getValue()); 3738 } 3739 3740 return false; 3741 } 3742 3743 bool MasmParser::parseStructInstList( 3744 const StructInfo &Structure, std::vector<StructInitializer> &Initializers, 3745 const AsmToken::TokenKind EndToken) { 3746 while (getTok().isNot(EndToken) || 3747 (EndToken == AsmToken::Greater && 3748 getTok().isNot(AsmToken::GreaterGreater))) { 3749 const AsmToken NextTok = Lexer.peekTok(); 3750 if (NextTok.is(AsmToken::Identifier) && 3751 NextTok.getString().equals_lower("dup")) { 3752 const MCExpr *Value; 3753 if (parseExpression(Value) || parseToken(AsmToken::Identifier)) 3754 return true; 3755 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 3756 if (!MCE) 3757 return Error(Value->getLoc(), 3758 "cannot repeat value a non-constant number of times"); 3759 const int64_t Repetitions = MCE->getValue(); 3760 if (Repetitions < 0) 3761 return Error(Value->getLoc(), 3762 "cannot repeat value a negative number of times"); 3763 3764 std::vector<StructInitializer> DuplicatedValues; 3765 if (parseToken(AsmToken::LParen, 3766 "parentheses required for 'dup' contents") || 3767 parseStructInstList(Structure, DuplicatedValues) || 3768 parseToken(AsmToken::RParen, "unmatched parentheses")) 3769 return true; 3770 3771 for (int i = 0; i < Repetitions; ++i) 3772 Initializers.insert(Initializers.end(), DuplicatedValues.begin(), 3773 DuplicatedValues.end()); 3774 } else { 3775 Initializers.emplace_back(); 3776 if (parseStructInitializer(Structure, Initializers.back())) 3777 return true; 3778 } 3779 3780 // Continue if we see a comma. (Also, allow line continuation.) 3781 if (!parseOptionalToken(AsmToken::Comma)) 3782 break; 3783 parseOptionalToken(AsmToken::EndOfStatement); 3784 } 3785 3786 return false; 3787 } 3788 3789 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3790 const IntFieldInfo &Contents) { 3791 // Default-initialize all values. 3792 for (const MCExpr *Value : Contents.Values) { 3793 if (emitIntValue(Value, Field.Type)) 3794 return true; 3795 } 3796 return false; 3797 } 3798 3799 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3800 const RealFieldInfo &Contents) { 3801 for (const APInt &AsInt : Contents.AsIntValues) { 3802 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3803 AsInt.getBitWidth() / 8); 3804 } 3805 return false; 3806 } 3807 3808 bool MasmParser::emitFieldValue(const FieldInfo &Field, 3809 const StructFieldInfo &Contents) { 3810 for (const auto &Initializer : Contents.Initializers) { 3811 size_t Index = 0, Offset = 0; 3812 for (const auto &SubField : Contents.Structure.Fields) { 3813 getStreamer().emitZeros(SubField.Offset - Offset); 3814 Offset = SubField.Offset + SubField.SizeOf; 3815 emitFieldInitializer(SubField, Initializer.FieldInitializers[Index++]); 3816 } 3817 } 3818 return false; 3819 } 3820 3821 bool MasmParser::emitFieldValue(const FieldInfo &Field) { 3822 switch (Field.Contents.FT) { 3823 case FT_INTEGRAL: 3824 return emitFieldValue(Field, Field.Contents.IntInfo); 3825 case FT_REAL: 3826 return emitFieldValue(Field, Field.Contents.RealInfo); 3827 case FT_STRUCT: 3828 return emitFieldValue(Field, Field.Contents.StructInfo); 3829 } 3830 llvm_unreachable("Unhandled FieldType enum"); 3831 } 3832 3833 bool MasmParser::emitStructValue(const StructInfo &Structure) { 3834 size_t Offset = 0; 3835 for (const auto &Field : Structure.Fields) { 3836 getStreamer().emitZeros(Field.Offset - Offset); 3837 if (emitFieldValue(Field)) 3838 return true; 3839 Offset = Field.Offset + Field.SizeOf; 3840 } 3841 // Add final padding. 3842 if (Offset != Structure.Size) 3843 getStreamer().emitZeros(Structure.Size - Offset); 3844 return false; 3845 } 3846 3847 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3848 const IntFieldInfo &Contents, 3849 const IntFieldInfo &Initializer) { 3850 for (const auto &Value : Initializer.Values) { 3851 if (emitIntValue(Value, Field.Type)) 3852 return true; 3853 } 3854 // Default-initialize all remaining values. 3855 for (auto it = Contents.Values.begin() + Initializer.Values.size(); 3856 it != Contents.Values.end(); ++it) { 3857 const auto &Value = *it; 3858 if (emitIntValue(Value, Field.Type)) 3859 return true; 3860 } 3861 return false; 3862 } 3863 3864 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3865 const RealFieldInfo &Contents, 3866 const RealFieldInfo &Initializer) { 3867 for (const auto &AsInt : Initializer.AsIntValues) { 3868 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3869 AsInt.getBitWidth() / 8); 3870 } 3871 // Default-initialize all remaining values. 3872 for (auto It = Contents.AsIntValues.begin() + Initializer.AsIntValues.size(); 3873 It != Contents.AsIntValues.end(); ++It) { 3874 const auto &AsInt = *It; 3875 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3876 AsInt.getBitWidth() / 8); 3877 } 3878 return false; 3879 } 3880 3881 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3882 const StructFieldInfo &Contents, 3883 const StructFieldInfo &Initializer) { 3884 for (const auto &Init : Initializer.Initializers) { 3885 emitStructInitializer(Contents.Structure, Init); 3886 } 3887 // Default-initialize all remaining values. 3888 for (auto It = 3889 Contents.Initializers.begin() + Initializer.Initializers.size(); 3890 It != Contents.Initializers.end(); ++It) { 3891 const auto &Init = *It; 3892 emitStructInitializer(Contents.Structure, Init); 3893 } 3894 return false; 3895 } 3896 3897 bool MasmParser::emitFieldInitializer(const FieldInfo &Field, 3898 const FieldInitializer &Initializer) { 3899 switch (Field.Contents.FT) { 3900 case FT_INTEGRAL: 3901 return emitFieldInitializer(Field, Field.Contents.IntInfo, 3902 Initializer.IntInfo); 3903 case FT_REAL: 3904 return emitFieldInitializer(Field, Field.Contents.RealInfo, 3905 Initializer.RealInfo); 3906 case FT_STRUCT: 3907 return emitFieldInitializer(Field, Field.Contents.StructInfo, 3908 Initializer.StructInfo); 3909 } 3910 llvm_unreachable("Unhandled FieldType enum"); 3911 } 3912 3913 bool MasmParser::emitStructInitializer(const StructInfo &Structure, 3914 const StructInitializer &Initializer) { 3915 size_t Index = 0, Offset = 0; 3916 for (const auto &Init : Initializer.FieldInitializers) { 3917 const auto &Field = Structure.Fields[Index++]; 3918 getStreamer().emitZeros(Field.Offset - Offset); 3919 Offset = Field.Offset + Field.SizeOf; 3920 if (emitFieldInitializer(Field, Init)) 3921 return true; 3922 } 3923 // Default-initialize all remaining fields. 3924 for (auto It = 3925 Structure.Fields.begin() + Initializer.FieldInitializers.size(); 3926 It != Structure.Fields.end(); ++It) { 3927 const auto &Field = *It; 3928 getStreamer().emitZeros(Field.Offset - Offset); 3929 Offset = Field.Offset + Field.SizeOf; 3930 if (emitFieldValue(Field)) 3931 return true; 3932 } 3933 // Add final padding. 3934 if (Offset != Structure.Size) 3935 getStreamer().emitZeros(Structure.Size - Offset); 3936 return false; 3937 } 3938 3939 // Set data values from initializers. 3940 bool MasmParser::emitStructValues(const StructInfo &Structure) { 3941 std::vector<StructInitializer> Initializers; 3942 if (parseStructInstList(Structure, Initializers)) 3943 return true; 3944 3945 for (const auto &Initializer : Initializers) { 3946 if (emitStructInitializer(Structure, Initializer)) 3947 return true; 3948 } 3949 3950 return false; 3951 } 3952 3953 // Declare a field in the current struct. 3954 bool MasmParser::addStructField(StringRef Name, const StructInfo &Structure) { 3955 StructInfo &OwningStruct = StructInProgress.back(); 3956 FieldInfo &Field = OwningStruct.addField(Name, FT_STRUCT); 3957 StructFieldInfo &StructInfo = Field.Contents.StructInfo; 3958 3959 StructInfo.Structure = Structure; 3960 Field.Type = Structure.Size; 3961 3962 if (parseStructInstList(Structure, StructInfo.Initializers)) 3963 return true; 3964 3965 Field.LengthOf = StructInfo.Initializers.size(); 3966 Field.SizeOf = Field.Type * Field.LengthOf; 3967 if (OwningStruct.IsUnion) 3968 OwningStruct.Size = std::max(OwningStruct.Size, Field.SizeOf); 3969 else 3970 OwningStruct.Size += Field.SizeOf; 3971 3972 return false; 3973 } 3974 3975 /// parseDirectiveStructValue 3976 /// ::= struct-id (<struct-initializer> | {struct-initializer}) 3977 /// [, (<struct-initializer> | {struct-initializer})]* 3978 bool MasmParser::parseDirectiveStructValue(const StructInfo &Structure, 3979 StringRef Directive, SMLoc DirLoc) { 3980 if (StructInProgress.empty()) { 3981 if (emitStructValues(Structure)) 3982 return true; 3983 } else if (addStructField("", Structure)) { 3984 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 3985 } 3986 3987 return false; 3988 } 3989 3990 /// parseDirectiveNamedValue 3991 /// ::= name (byte | word | ... ) [ expression (, expression)* ] 3992 bool MasmParser::parseDirectiveNamedStructValue(const StructInfo &Structure, 3993 StringRef Directive, 3994 SMLoc DirLoc, StringRef Name) { 3995 if (StructInProgress.empty()) { 3996 // Initialize named data value. 3997 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 3998 getStreamer().emitLabel(Sym); 3999 KnownType[Name] = &Structure; 4000 if (emitStructValues(Structure)) 4001 return true; 4002 } else if (addStructField(Name, Structure)) { 4003 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4004 } 4005 4006 return false; 4007 } 4008 4009 /// parseDirectiveStruct 4010 /// ::= <name> (STRUC | STRUCT | UNION) [fieldAlign] [, NONUNIQUE] 4011 /// (dataDir | generalDir | offsetDir | nestedStruct)+ 4012 /// <name> ENDS 4013 ////// dataDir = data declaration 4014 ////// offsetDir = EVEN, ORG, ALIGN 4015 bool MasmParser::parseDirectiveStruct(StringRef Directive, 4016 DirectiveKind DirKind, StringRef Name, 4017 SMLoc NameLoc) { 4018 // We ignore NONUNIQUE; we do not support OPTION M510 or OPTION OLDSTRUCTS 4019 // anyway, so all field accesses must be qualified. 4020 AsmToken NextTok = getTok(); 4021 int64_t AlignmentValue = 1; 4022 if (NextTok.isNot(AsmToken::Comma) && 4023 NextTok.isNot(AsmToken::EndOfStatement) && 4024 parseAbsoluteExpression(AlignmentValue)) { 4025 return addErrorSuffix(" in alignment value for '" + Twine(Directive) + 4026 "' directive"); 4027 } 4028 if (!isPowerOf2_64(AlignmentValue)) { 4029 return Error(NextTok.getLoc(), "alignment must be a power of two; was " + 4030 std::to_string(AlignmentValue)); 4031 } 4032 4033 StringRef Qualifier; 4034 SMLoc QualifierLoc; 4035 if (parseOptionalToken(AsmToken::Comma)) { 4036 QualifierLoc = getTok().getLoc(); 4037 if (parseIdentifier(Qualifier)) 4038 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4039 if (!Qualifier.equals_lower("nonunique")) 4040 return Error(QualifierLoc, "Unrecognized qualifier for '" + 4041 Twine(Directive) + 4042 "' directive; expected none or NONUNIQUE"); 4043 } 4044 4045 if (parseToken(AsmToken::EndOfStatement)) 4046 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4047 4048 StructInProgress.emplace_back(Name, DirKind == DK_UNION, AlignmentValue); 4049 return false; 4050 } 4051 4052 /// parseDirectiveNestedStruct 4053 /// ::= (STRUC | STRUCT | UNION) [name] 4054 /// (dataDir | generalDir | offsetDir | nestedStruct)+ 4055 /// ENDS 4056 bool MasmParser::parseDirectiveNestedStruct(StringRef Directive, 4057 DirectiveKind DirKind) { 4058 if (StructInProgress.empty()) 4059 return TokError("missing name in top-level '" + Twine(Directive) + 4060 "' directive"); 4061 4062 StringRef Name; 4063 if (getTok().is(AsmToken::Identifier)) { 4064 Name = getTok().getIdentifier(); 4065 parseToken(AsmToken::Identifier); 4066 } 4067 if (parseToken(AsmToken::EndOfStatement)) 4068 return addErrorSuffix(" in '" + Twine(Directive) + "' directive"); 4069 4070 StructInProgress.emplace_back(Name, DirKind == DK_UNION, 4071 StructInProgress.back().Alignment); 4072 return false; 4073 } 4074 4075 bool MasmParser::parseDirectiveEnds(StringRef Name, SMLoc NameLoc) { 4076 if (StructInProgress.empty()) 4077 return Error(NameLoc, "ENDS directive without matching STRUC/STRUCT/UNION"); 4078 if (StructInProgress.size() > 1) 4079 return Error(NameLoc, "unexpected name in nested ENDS directive"); 4080 if (StructInProgress.back().Name.compare_lower(Name)) 4081 return Error(NameLoc, "mismatched name in ENDS directive; expected '" + 4082 StructInProgress.back().Name + "'"); 4083 StructInfo Structure = StructInProgress.pop_back_val(); 4084 // Pad to make the structure's size divisible by its alignment. 4085 Structure.Size = llvm::alignTo(Structure.Size, Structure.Alignment); 4086 Structs[Name.lower()] = Structure; 4087 4088 if (parseToken(AsmToken::EndOfStatement)) 4089 return addErrorSuffix(" in ENDS directive"); 4090 4091 return false; 4092 } 4093 4094 bool MasmParser::parseDirectiveNestedEnds() { 4095 if (StructInProgress.empty()) 4096 return TokError("ENDS directive without matching STRUC/STRUCT/UNION"); 4097 if (StructInProgress.size() == 1) 4098 return TokError("missing name in top-level ENDS directive"); 4099 4100 if (parseToken(AsmToken::EndOfStatement)) 4101 return addErrorSuffix(" in nested ENDS directive"); 4102 4103 StructInfo Structure = StructInProgress.pop_back_val(); 4104 // Pad to make the structure's size divisible by its alignment. 4105 Structure.Size = llvm::alignTo(Structure.Size, Structure.Alignment); 4106 4107 StructInfo &ParentStruct = StructInProgress.back(); 4108 if (Structure.Name.empty()) { 4109 const size_t OldFields = ParentStruct.Fields.size(); 4110 ParentStruct.Fields.insert( 4111 ParentStruct.Fields.end(), 4112 std::make_move_iterator(Structure.Fields.begin()), 4113 std::make_move_iterator(Structure.Fields.end())); 4114 for (const auto &FieldByName : Structure.FieldsByName) { 4115 ParentStruct.FieldsByName[FieldByName.getKey()] = 4116 FieldByName.getValue() + OldFields; 4117 } 4118 if (!ParentStruct.IsUnion) { 4119 for (auto FieldIter = ParentStruct.Fields.begin() + OldFields; 4120 FieldIter != ParentStruct.Fields.end(); ++FieldIter) { 4121 FieldIter->Offset += ParentStruct.Size; 4122 } 4123 } 4124 4125 if (ParentStruct.IsUnion) 4126 ParentStruct.Size = std::max(ParentStruct.Size, Structure.Size); 4127 else 4128 ParentStruct.Size += Structure.Size; 4129 } else { 4130 FieldInfo &Field = ParentStruct.addField(Structure.Name, FT_STRUCT); 4131 StructFieldInfo &StructInfo = Field.Contents.StructInfo; 4132 Field.Type = Structure.Size; 4133 Field.LengthOf = 1; 4134 Field.SizeOf = Structure.Size; 4135 4136 if (ParentStruct.IsUnion) 4137 ParentStruct.Size = std::max(ParentStruct.Size, Field.SizeOf); 4138 else 4139 ParentStruct.Size += Field.SizeOf; 4140 4141 StructInfo.Structure = Structure; 4142 StructInfo.Initializers.emplace_back(); 4143 auto &FieldInitializers = StructInfo.Initializers.back().FieldInitializers; 4144 for (const auto &SubField : Structure.Fields) { 4145 FieldInitializers.push_back(SubField.Contents); 4146 } 4147 } 4148 4149 return false; 4150 } 4151 4152 /// parseDirectiveOrg 4153 /// ::= .org expression [ , expression ] 4154 bool MasmParser::parseDirectiveOrg() { 4155 const MCExpr *Offset; 4156 SMLoc OffsetLoc = Lexer.getLoc(); 4157 if (checkForValidSection() || parseExpression(Offset)) 4158 return true; 4159 4160 // Parse optional fill expression. 4161 int64_t FillExpr = 0; 4162 if (parseOptionalToken(AsmToken::Comma)) 4163 if (parseAbsoluteExpression(FillExpr)) 4164 return addErrorSuffix(" in '.org' directive"); 4165 if (parseToken(AsmToken::EndOfStatement)) 4166 return addErrorSuffix(" in '.org' directive"); 4167 4168 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 4169 return false; 4170 } 4171 4172 /// parseDirectiveAlign 4173 /// ::= align expression 4174 bool MasmParser::parseDirectiveAlign() { 4175 SMLoc AlignmentLoc = getLexer().getLoc(); 4176 int64_t Alignment; 4177 4178 if (checkForValidSection()) 4179 return addErrorSuffix(" in align directive"); 4180 // Ignore empty 'align' directives. 4181 if (getTok().is(AsmToken::EndOfStatement)) { 4182 Warning(AlignmentLoc, "align directive with no operand is ignored"); 4183 return parseToken(AsmToken::EndOfStatement); 4184 } 4185 if (parseAbsoluteExpression(Alignment) || 4186 parseToken(AsmToken::EndOfStatement)) 4187 return addErrorSuffix(" in align directive"); 4188 4189 // Always emit an alignment here even if we thrown an error. 4190 bool ReturnVal = false; 4191 4192 // Reject alignments that aren't either a power of two or zero, for gas 4193 // compatibility. Alignment of zero is silently rounded up to one. 4194 if (Alignment == 0) 4195 Alignment = 1; 4196 if (!isPowerOf2_64(Alignment)) 4197 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 4198 4199 // Check whether we should use optimal code alignment for this align 4200 // directive. 4201 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 4202 assert(Section && "must have section to emit alignment"); 4203 if (Section->UseCodeAlign()) { 4204 getStreamer().emitCodeAlignment(Alignment, /*MaxBytesToEmit=*/0); 4205 } else { 4206 // FIXME: Target specific behavior about how the "extra" bytes are filled. 4207 getStreamer().emitValueToAlignment(Alignment, /*Value=*/0, /*ValueSize=*/1, 4208 /*MaxBytesToEmit=*/0); 4209 } 4210 4211 return ReturnVal; 4212 } 4213 4214 /// parseDirectiveFile 4215 /// ::= .file filename 4216 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 4217 bool MasmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 4218 // FIXME: I'm not sure what this is. 4219 int64_t FileNumber = -1; 4220 if (getLexer().is(AsmToken::Integer)) { 4221 FileNumber = getTok().getIntVal(); 4222 Lex(); 4223 4224 if (FileNumber < 0) 4225 return TokError("negative file number"); 4226 } 4227 4228 std::string Path; 4229 4230 // Usually the directory and filename together, otherwise just the directory. 4231 // Allow the strings to have escaped octal character sequence. 4232 if (check(getTok().isNot(AsmToken::String), 4233 "unexpected token in '.file' directive") || 4234 parseEscapedString(Path)) 4235 return true; 4236 4237 StringRef Directory; 4238 StringRef Filename; 4239 std::string FilenameData; 4240 if (getLexer().is(AsmToken::String)) { 4241 if (check(FileNumber == -1, 4242 "explicit path specified, but no file number") || 4243 parseEscapedString(FilenameData)) 4244 return true; 4245 Filename = FilenameData; 4246 Directory = Path; 4247 } else { 4248 Filename = Path; 4249 } 4250 4251 uint64_t MD5Hi, MD5Lo; 4252 bool HasMD5 = false; 4253 4254 Optional<StringRef> Source; 4255 bool HasSource = false; 4256 std::string SourceString; 4257 4258 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 4259 StringRef Keyword; 4260 if (check(getTok().isNot(AsmToken::Identifier), 4261 "unexpected token in '.file' directive") || 4262 parseIdentifier(Keyword)) 4263 return true; 4264 if (Keyword == "md5") { 4265 HasMD5 = true; 4266 if (check(FileNumber == -1, 4267 "MD5 checksum specified, but no file number") || 4268 parseHexOcta(*this, MD5Hi, MD5Lo)) 4269 return true; 4270 } else if (Keyword == "source") { 4271 HasSource = true; 4272 if (check(FileNumber == -1, 4273 "source specified, but no file number") || 4274 check(getTok().isNot(AsmToken::String), 4275 "unexpected token in '.file' directive") || 4276 parseEscapedString(SourceString)) 4277 return true; 4278 } else { 4279 return TokError("unexpected token in '.file' directive"); 4280 } 4281 } 4282 4283 if (FileNumber == -1) { 4284 // Ignore the directive if there is no number and the target doesn't support 4285 // numberless .file directives. This allows some portability of assembler 4286 // between different object file formats. 4287 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 4288 getStreamer().emitFileDirective(Filename); 4289 } else { 4290 // In case there is a -g option as well as debug info from directive .file, 4291 // we turn off the -g option, directly use the existing debug info instead. 4292 // Throw away any implicit file table for the assembler source. 4293 if (Ctx.getGenDwarfForAssembly()) { 4294 Ctx.getMCDwarfLineTable(0).resetFileTable(); 4295 Ctx.setGenDwarfForAssembly(false); 4296 } 4297 4298 Optional<MD5::MD5Result> CKMem; 4299 if (HasMD5) { 4300 MD5::MD5Result Sum; 4301 for (unsigned i = 0; i != 8; ++i) { 4302 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 4303 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 4304 } 4305 CKMem = Sum; 4306 } 4307 if (HasSource) { 4308 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 4309 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 4310 Source = StringRef(SourceBuf, SourceString.size()); 4311 } 4312 if (FileNumber == 0) { 4313 if (Ctx.getDwarfVersion() < 5) 4314 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5"); 4315 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 4316 } else { 4317 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 4318 FileNumber, Directory, Filename, CKMem, Source); 4319 if (!FileNumOrErr) 4320 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 4321 } 4322 // Alert the user if there are some .file directives with MD5 and some not. 4323 // But only do that once. 4324 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 4325 ReportedInconsistentMD5 = true; 4326 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 4327 } 4328 } 4329 4330 return false; 4331 } 4332 4333 /// parseDirectiveLine 4334 /// ::= .line [number] 4335 bool MasmParser::parseDirectiveLine() { 4336 int64_t LineNumber; 4337 if (getLexer().is(AsmToken::Integer)) { 4338 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 4339 return true; 4340 (void)LineNumber; 4341 // FIXME: Do something with the .line. 4342 } 4343 if (parseToken(AsmToken::EndOfStatement, 4344 "unexpected token in '.line' directive")) 4345 return true; 4346 4347 return false; 4348 } 4349 4350 /// parseDirectiveLoc 4351 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 4352 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 4353 /// The first number is a file number, must have been previously assigned with 4354 /// a .file directive, the second number is the line number and optionally the 4355 /// third number is a column position (zero if not specified). The remaining 4356 /// optional items are .loc sub-directives. 4357 bool MasmParser::parseDirectiveLoc() { 4358 int64_t FileNumber = 0, LineNumber = 0; 4359 SMLoc Loc = getTok().getLoc(); 4360 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 4361 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 4362 "file number less than one in '.loc' directive") || 4363 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 4364 "unassigned file number in '.loc' directive")) 4365 return true; 4366 4367 // optional 4368 if (getLexer().is(AsmToken::Integer)) { 4369 LineNumber = getTok().getIntVal(); 4370 if (LineNumber < 0) 4371 return TokError("line number less than zero in '.loc' directive"); 4372 Lex(); 4373 } 4374 4375 int64_t ColumnPos = 0; 4376 if (getLexer().is(AsmToken::Integer)) { 4377 ColumnPos = getTok().getIntVal(); 4378 if (ColumnPos < 0) 4379 return TokError("column position less than zero in '.loc' directive"); 4380 Lex(); 4381 } 4382 4383 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags(); 4384 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT; 4385 unsigned Isa = 0; 4386 int64_t Discriminator = 0; 4387 4388 auto parseLocOp = [&]() -> bool { 4389 StringRef Name; 4390 SMLoc Loc = getTok().getLoc(); 4391 if (parseIdentifier(Name)) 4392 return TokError("unexpected token in '.loc' directive"); 4393 4394 if (Name == "basic_block") 4395 Flags |= DWARF2_FLAG_BASIC_BLOCK; 4396 else if (Name == "prologue_end") 4397 Flags |= DWARF2_FLAG_PROLOGUE_END; 4398 else if (Name == "epilogue_begin") 4399 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 4400 else if (Name == "is_stmt") { 4401 Loc = getTok().getLoc(); 4402 const MCExpr *Value; 4403 if (parseExpression(Value)) 4404 return true; 4405 // The expression must be the constant 0 or 1. 4406 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4407 int Value = MCE->getValue(); 4408 if (Value == 0) 4409 Flags &= ~DWARF2_FLAG_IS_STMT; 4410 else if (Value == 1) 4411 Flags |= DWARF2_FLAG_IS_STMT; 4412 else 4413 return Error(Loc, "is_stmt value not 0 or 1"); 4414 } else { 4415 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 4416 } 4417 } else if (Name == "isa") { 4418 Loc = getTok().getLoc(); 4419 const MCExpr *Value; 4420 if (parseExpression(Value)) 4421 return true; 4422 // The expression must be a constant greater or equal to 0. 4423 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4424 int Value = MCE->getValue(); 4425 if (Value < 0) 4426 return Error(Loc, "isa number less than zero"); 4427 Isa = Value; 4428 } else { 4429 return Error(Loc, "isa number not a constant value"); 4430 } 4431 } else if (Name == "discriminator") { 4432 if (parseAbsoluteExpression(Discriminator)) 4433 return true; 4434 } else { 4435 return Error(Loc, "unknown sub-directive in '.loc' directive"); 4436 } 4437 return false; 4438 }; 4439 4440 if (parseMany(parseLocOp, false /*hasComma*/)) 4441 return true; 4442 4443 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 4444 Isa, Discriminator, StringRef()); 4445 4446 return false; 4447 } 4448 4449 /// parseDirectiveStabs 4450 /// ::= .stabs string, number, number, number 4451 bool MasmParser::parseDirectiveStabs() { 4452 return TokError("unsupported directive '.stabs'"); 4453 } 4454 4455 /// parseDirectiveCVFile 4456 /// ::= .cv_file number filename [checksum] [checksumkind] 4457 bool MasmParser::parseDirectiveCVFile() { 4458 SMLoc FileNumberLoc = getTok().getLoc(); 4459 int64_t FileNumber; 4460 std::string Filename; 4461 std::string Checksum; 4462 int64_t ChecksumKind = 0; 4463 4464 if (parseIntToken(FileNumber, 4465 "expected file number in '.cv_file' directive") || 4466 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 4467 check(getTok().isNot(AsmToken::String), 4468 "unexpected token in '.cv_file' directive") || 4469 parseEscapedString(Filename)) 4470 return true; 4471 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4472 if (check(getTok().isNot(AsmToken::String), 4473 "unexpected token in '.cv_file' directive") || 4474 parseEscapedString(Checksum) || 4475 parseIntToken(ChecksumKind, 4476 "expected checksum kind in '.cv_file' directive") || 4477 parseToken(AsmToken::EndOfStatement, 4478 "unexpected token in '.cv_file' directive")) 4479 return true; 4480 } 4481 4482 Checksum = fromHex(Checksum); 4483 void *CKMem = Ctx.allocate(Checksum.size(), 1); 4484 memcpy(CKMem, Checksum.data(), Checksum.size()); 4485 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 4486 Checksum.size()); 4487 4488 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 4489 static_cast<uint8_t>(ChecksumKind))) 4490 return Error(FileNumberLoc, "file number already allocated"); 4491 4492 return false; 4493 } 4494 4495 bool MasmParser::parseCVFunctionId(int64_t &FunctionId, 4496 StringRef DirectiveName) { 4497 SMLoc Loc; 4498 return parseTokenLoc(Loc) || 4499 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 4500 "' directive") || 4501 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 4502 "expected function id within range [0, UINT_MAX)"); 4503 } 4504 4505 bool MasmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 4506 SMLoc Loc; 4507 return parseTokenLoc(Loc) || 4508 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 4509 "' directive") || 4510 check(FileNumber < 1, Loc, "file number less than one in '" + 4511 DirectiveName + "' directive") || 4512 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 4513 "unassigned file number in '" + DirectiveName + "' directive"); 4514 } 4515 4516 /// parseDirectiveCVFuncId 4517 /// ::= .cv_func_id FunctionId 4518 /// 4519 /// Introduces a function ID that can be used with .cv_loc. 4520 bool MasmParser::parseDirectiveCVFuncId() { 4521 SMLoc FunctionIdLoc = getTok().getLoc(); 4522 int64_t FunctionId; 4523 4524 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 4525 parseToken(AsmToken::EndOfStatement, 4526 "unexpected token in '.cv_func_id' directive")) 4527 return true; 4528 4529 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 4530 return Error(FunctionIdLoc, "function id already allocated"); 4531 4532 return false; 4533 } 4534 4535 /// parseDirectiveCVInlineSiteId 4536 /// ::= .cv_inline_site_id FunctionId 4537 /// "within" IAFunc 4538 /// "inlined_at" IAFile IALine [IACol] 4539 /// 4540 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 4541 /// at" source location information for use in the line table of the caller, 4542 /// whether the caller is a real function or another inlined call site. 4543 bool MasmParser::parseDirectiveCVInlineSiteId() { 4544 SMLoc FunctionIdLoc = getTok().getLoc(); 4545 int64_t FunctionId; 4546 int64_t IAFunc; 4547 int64_t IAFile; 4548 int64_t IALine; 4549 int64_t IACol = 0; 4550 4551 // FunctionId 4552 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 4553 return true; 4554 4555 // "within" 4556 if (check((getLexer().isNot(AsmToken::Identifier) || 4557 getTok().getIdentifier() != "within"), 4558 "expected 'within' identifier in '.cv_inline_site_id' directive")) 4559 return true; 4560 Lex(); 4561 4562 // IAFunc 4563 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 4564 return true; 4565 4566 // "inlined_at" 4567 if (check((getLexer().isNot(AsmToken::Identifier) || 4568 getTok().getIdentifier() != "inlined_at"), 4569 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 4570 "directive") ) 4571 return true; 4572 Lex(); 4573 4574 // IAFile IALine 4575 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 4576 parseIntToken(IALine, "expected line number after 'inlined_at'")) 4577 return true; 4578 4579 // [IACol] 4580 if (getLexer().is(AsmToken::Integer)) { 4581 IACol = getTok().getIntVal(); 4582 Lex(); 4583 } 4584 4585 if (parseToken(AsmToken::EndOfStatement, 4586 "unexpected token in '.cv_inline_site_id' directive")) 4587 return true; 4588 4589 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 4590 IALine, IACol, FunctionIdLoc)) 4591 return Error(FunctionIdLoc, "function id already allocated"); 4592 4593 return false; 4594 } 4595 4596 /// parseDirectiveCVLoc 4597 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 4598 /// [is_stmt VALUE] 4599 /// The first number is a file number, must have been previously assigned with 4600 /// a .file directive, the second number is the line number and optionally the 4601 /// third number is a column position (zero if not specified). The remaining 4602 /// optional items are .loc sub-directives. 4603 bool MasmParser::parseDirectiveCVLoc() { 4604 SMLoc DirectiveLoc = getTok().getLoc(); 4605 int64_t FunctionId, FileNumber; 4606 if (parseCVFunctionId(FunctionId, ".cv_loc") || 4607 parseCVFileId(FileNumber, ".cv_loc")) 4608 return true; 4609 4610 int64_t LineNumber = 0; 4611 if (getLexer().is(AsmToken::Integer)) { 4612 LineNumber = getTok().getIntVal(); 4613 if (LineNumber < 0) 4614 return TokError("line number less than zero in '.cv_loc' directive"); 4615 Lex(); 4616 } 4617 4618 int64_t ColumnPos = 0; 4619 if (getLexer().is(AsmToken::Integer)) { 4620 ColumnPos = getTok().getIntVal(); 4621 if (ColumnPos < 0) 4622 return TokError("column position less than zero in '.cv_loc' directive"); 4623 Lex(); 4624 } 4625 4626 bool PrologueEnd = false; 4627 uint64_t IsStmt = 0; 4628 4629 auto parseOp = [&]() -> bool { 4630 StringRef Name; 4631 SMLoc Loc = getTok().getLoc(); 4632 if (parseIdentifier(Name)) 4633 return TokError("unexpected token in '.cv_loc' directive"); 4634 if (Name == "prologue_end") 4635 PrologueEnd = true; 4636 else if (Name == "is_stmt") { 4637 Loc = getTok().getLoc(); 4638 const MCExpr *Value; 4639 if (parseExpression(Value)) 4640 return true; 4641 // The expression must be the constant 0 or 1. 4642 IsStmt = ~0ULL; 4643 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 4644 IsStmt = MCE->getValue(); 4645 4646 if (IsStmt > 1) 4647 return Error(Loc, "is_stmt value not 0 or 1"); 4648 } else { 4649 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 4650 } 4651 return false; 4652 }; 4653 4654 if (parseMany(parseOp, false /*hasComma*/)) 4655 return true; 4656 4657 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber, 4658 ColumnPos, PrologueEnd, IsStmt, StringRef(), 4659 DirectiveLoc); 4660 return false; 4661 } 4662 4663 /// parseDirectiveCVLinetable 4664 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 4665 bool MasmParser::parseDirectiveCVLinetable() { 4666 int64_t FunctionId; 4667 StringRef FnStartName, FnEndName; 4668 SMLoc Loc = getTok().getLoc(); 4669 if (parseCVFunctionId(FunctionId, ".cv_linetable") || 4670 parseToken(AsmToken::Comma, 4671 "unexpected token in '.cv_linetable' directive") || 4672 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 4673 "expected identifier in directive") || 4674 parseToken(AsmToken::Comma, 4675 "unexpected token in '.cv_linetable' directive") || 4676 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 4677 "expected identifier in directive")) 4678 return true; 4679 4680 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 4681 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 4682 4683 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 4684 return false; 4685 } 4686 4687 /// parseDirectiveCVInlineLinetable 4688 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 4689 bool MasmParser::parseDirectiveCVInlineLinetable() { 4690 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 4691 StringRef FnStartName, FnEndName; 4692 SMLoc Loc = getTok().getLoc(); 4693 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 4694 parseTokenLoc(Loc) || 4695 parseIntToken( 4696 SourceFileId, 4697 "expected SourceField in '.cv_inline_linetable' directive") || 4698 check(SourceFileId <= 0, Loc, 4699 "File id less than zero in '.cv_inline_linetable' directive") || 4700 parseTokenLoc(Loc) || 4701 parseIntToken( 4702 SourceLineNum, 4703 "expected SourceLineNum in '.cv_inline_linetable' directive") || 4704 check(SourceLineNum < 0, Loc, 4705 "Line number less than zero in '.cv_inline_linetable' directive") || 4706 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 4707 "expected identifier in directive") || 4708 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 4709 "expected identifier in directive")) 4710 return true; 4711 4712 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4713 return true; 4714 4715 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 4716 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 4717 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 4718 SourceLineNum, FnStartSym, 4719 FnEndSym); 4720 return false; 4721 } 4722 4723 void MasmParser::initializeCVDefRangeTypeMap() { 4724 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 4725 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 4726 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 4727 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 4728 } 4729 4730 /// parseDirectiveCVDefRange 4731 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 4732 bool MasmParser::parseDirectiveCVDefRange() { 4733 SMLoc Loc; 4734 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 4735 while (getLexer().is(AsmToken::Identifier)) { 4736 Loc = getLexer().getLoc(); 4737 StringRef GapStartName; 4738 if (parseIdentifier(GapStartName)) 4739 return Error(Loc, "expected identifier in directive"); 4740 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 4741 4742 Loc = getLexer().getLoc(); 4743 StringRef GapEndName; 4744 if (parseIdentifier(GapEndName)) 4745 return Error(Loc, "expected identifier in directive"); 4746 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 4747 4748 Ranges.push_back({GapStartSym, GapEndSym}); 4749 } 4750 4751 StringRef CVDefRangeTypeStr; 4752 if (parseToken( 4753 AsmToken::Comma, 4754 "expected comma before def_range type in .cv_def_range directive") || 4755 parseIdentifier(CVDefRangeTypeStr)) 4756 return Error(Loc, "expected def_range type in directive"); 4757 4758 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 4759 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 4760 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 4761 ? CVDR_DEFRANGE 4762 : CVTypeIt->getValue(); 4763 switch (CVDRType) { 4764 case CVDR_DEFRANGE_REGISTER: { 4765 int64_t DRRegister; 4766 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4767 ".cv_def_range directive") || 4768 parseAbsoluteExpression(DRRegister)) 4769 return Error(Loc, "expected register number"); 4770 4771 codeview::DefRangeRegisterHeader DRHdr; 4772 DRHdr.Register = DRRegister; 4773 DRHdr.MayHaveNoName = 0; 4774 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4775 break; 4776 } 4777 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 4778 int64_t DROffset; 4779 if (parseToken(AsmToken::Comma, 4780 "expected comma before offset in .cv_def_range directive") || 4781 parseAbsoluteExpression(DROffset)) 4782 return Error(Loc, "expected offset value"); 4783 4784 codeview::DefRangeFramePointerRelHeader DRHdr; 4785 DRHdr.Offset = DROffset; 4786 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4787 break; 4788 } 4789 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 4790 int64_t DRRegister; 4791 int64_t DROffsetInParent; 4792 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4793 ".cv_def_range directive") || 4794 parseAbsoluteExpression(DRRegister)) 4795 return Error(Loc, "expected register number"); 4796 if (parseToken(AsmToken::Comma, 4797 "expected comma before offset in .cv_def_range directive") || 4798 parseAbsoluteExpression(DROffsetInParent)) 4799 return Error(Loc, "expected offset value"); 4800 4801 codeview::DefRangeSubfieldRegisterHeader DRHdr; 4802 DRHdr.Register = DRRegister; 4803 DRHdr.MayHaveNoName = 0; 4804 DRHdr.OffsetInParent = DROffsetInParent; 4805 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4806 break; 4807 } 4808 case CVDR_DEFRANGE_REGISTER_REL: { 4809 int64_t DRRegister; 4810 int64_t DRFlags; 4811 int64_t DRBasePointerOffset; 4812 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4813 ".cv_def_range directive") || 4814 parseAbsoluteExpression(DRRegister)) 4815 return Error(Loc, "expected register value"); 4816 if (parseToken( 4817 AsmToken::Comma, 4818 "expected comma before flag value in .cv_def_range directive") || 4819 parseAbsoluteExpression(DRFlags)) 4820 return Error(Loc, "expected flag value"); 4821 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 4822 "in .cv_def_range directive") || 4823 parseAbsoluteExpression(DRBasePointerOffset)) 4824 return Error(Loc, "expected base pointer offset value"); 4825 4826 codeview::DefRangeRegisterRelHeader DRHdr; 4827 DRHdr.Register = DRRegister; 4828 DRHdr.Flags = DRFlags; 4829 DRHdr.BasePointerOffset = DRBasePointerOffset; 4830 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4831 break; 4832 } 4833 default: 4834 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 4835 } 4836 return true; 4837 } 4838 4839 /// parseDirectiveCVString 4840 /// ::= .cv_stringtable "string" 4841 bool MasmParser::parseDirectiveCVString() { 4842 std::string Data; 4843 if (checkForValidSection() || parseEscapedString(Data)) 4844 return addErrorSuffix(" in '.cv_string' directive"); 4845 4846 // Put the string in the table and emit the offset. 4847 std::pair<StringRef, unsigned> Insertion = 4848 getCVContext().addToStringTable(Data); 4849 getStreamer().emitIntValue(Insertion.second, 4); 4850 return false; 4851 } 4852 4853 /// parseDirectiveCVStringTable 4854 /// ::= .cv_stringtable 4855 bool MasmParser::parseDirectiveCVStringTable() { 4856 getStreamer().emitCVStringTableDirective(); 4857 return false; 4858 } 4859 4860 /// parseDirectiveCVFileChecksums 4861 /// ::= .cv_filechecksums 4862 bool MasmParser::parseDirectiveCVFileChecksums() { 4863 getStreamer().emitCVFileChecksumsDirective(); 4864 return false; 4865 } 4866 4867 /// parseDirectiveCVFileChecksumOffset 4868 /// ::= .cv_filechecksumoffset fileno 4869 bool MasmParser::parseDirectiveCVFileChecksumOffset() { 4870 int64_t FileNo; 4871 if (parseIntToken(FileNo, "expected identifier in directive")) 4872 return true; 4873 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4874 return true; 4875 getStreamer().emitCVFileChecksumOffsetDirective(FileNo); 4876 return false; 4877 } 4878 4879 /// parseDirectiveCVFPOData 4880 /// ::= .cv_fpo_data procsym 4881 bool MasmParser::parseDirectiveCVFPOData() { 4882 SMLoc DirLoc = getLexer().getLoc(); 4883 StringRef ProcName; 4884 if (parseIdentifier(ProcName)) 4885 return TokError("expected symbol name"); 4886 if (parseEOL("unexpected tokens")) 4887 return addErrorSuffix(" in '.cv_fpo_data' directive"); 4888 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4889 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 4890 return false; 4891 } 4892 4893 /// parseDirectiveCFISections 4894 /// ::= .cfi_sections section [, section] 4895 bool MasmParser::parseDirectiveCFISections() { 4896 StringRef Name; 4897 bool EH = false; 4898 bool Debug = false; 4899 4900 if (parseIdentifier(Name)) 4901 return TokError("Expected an identifier"); 4902 4903 if (Name == ".eh_frame") 4904 EH = true; 4905 else if (Name == ".debug_frame") 4906 Debug = true; 4907 4908 if (getLexer().is(AsmToken::Comma)) { 4909 Lex(); 4910 4911 if (parseIdentifier(Name)) 4912 return TokError("Expected an identifier"); 4913 4914 if (Name == ".eh_frame") 4915 EH = true; 4916 else if (Name == ".debug_frame") 4917 Debug = true; 4918 } 4919 4920 getStreamer().emitCFISections(EH, Debug); 4921 return false; 4922 } 4923 4924 /// parseDirectiveCFIStartProc 4925 /// ::= .cfi_startproc [simple] 4926 bool MasmParser::parseDirectiveCFIStartProc() { 4927 StringRef Simple; 4928 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4929 if (check(parseIdentifier(Simple) || Simple != "simple", 4930 "unexpected token") || 4931 parseToken(AsmToken::EndOfStatement)) 4932 return addErrorSuffix(" in '.cfi_startproc' directive"); 4933 } 4934 4935 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4936 // being produced if this directive is emitted as part of preprocessor macro 4937 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4938 // Tools like llvm-mc on the other hand are not affected by it, and report 4939 // correct context information. 4940 getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4941 return false; 4942 } 4943 4944 /// parseDirectiveCFIEndProc 4945 /// ::= .cfi_endproc 4946 bool MasmParser::parseDirectiveCFIEndProc() { 4947 getStreamer().emitCFIEndProc(); 4948 return false; 4949 } 4950 4951 /// parse register name or number. 4952 bool MasmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4953 SMLoc DirectiveLoc) { 4954 unsigned RegNo; 4955 4956 if (getLexer().isNot(AsmToken::Integer)) { 4957 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4958 return true; 4959 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4960 } else 4961 return parseAbsoluteExpression(Register); 4962 4963 return false; 4964 } 4965 4966 /// parseDirectiveCFIDefCfa 4967 /// ::= .cfi_def_cfa register, offset 4968 bool MasmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4969 int64_t Register = 0, Offset = 0; 4970 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 4971 parseToken(AsmToken::Comma, "unexpected token in directive") || 4972 parseAbsoluteExpression(Offset)) 4973 return true; 4974 4975 getStreamer().emitCFIDefCfa(Register, Offset); 4976 return false; 4977 } 4978 4979 /// parseDirectiveCFIDefCfaOffset 4980 /// ::= .cfi_def_cfa_offset offset 4981 bool MasmParser::parseDirectiveCFIDefCfaOffset() { 4982 int64_t Offset = 0; 4983 if (parseAbsoluteExpression(Offset)) 4984 return true; 4985 4986 getStreamer().emitCFIDefCfaOffset(Offset); 4987 return false; 4988 } 4989 4990 /// parseDirectiveCFIRegister 4991 /// ::= .cfi_register register, register 4992 bool MasmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4993 int64_t Register1 = 0, Register2 = 0; 4994 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || 4995 parseToken(AsmToken::Comma, "unexpected token in directive") || 4996 parseRegisterOrRegisterNumber(Register2, DirectiveLoc)) 4997 return true; 4998 4999 getStreamer().emitCFIRegister(Register1, Register2); 5000 return false; 5001 } 5002 5003 /// parseDirectiveCFIWindowSave 5004 /// ::= .cfi_window_save 5005 bool MasmParser::parseDirectiveCFIWindowSave() { 5006 getStreamer().emitCFIWindowSave(); 5007 return false; 5008 } 5009 5010 /// parseDirectiveCFIAdjustCfaOffset 5011 /// ::= .cfi_adjust_cfa_offset adjustment 5012 bool MasmParser::parseDirectiveCFIAdjustCfaOffset() { 5013 int64_t Adjustment = 0; 5014 if (parseAbsoluteExpression(Adjustment)) 5015 return true; 5016 5017 getStreamer().emitCFIAdjustCfaOffset(Adjustment); 5018 return false; 5019 } 5020 5021 /// parseDirectiveCFIDefCfaRegister 5022 /// ::= .cfi_def_cfa_register register 5023 bool MasmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 5024 int64_t Register = 0; 5025 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5026 return true; 5027 5028 getStreamer().emitCFIDefCfaRegister(Register); 5029 return false; 5030 } 5031 5032 /// parseDirectiveCFIOffset 5033 /// ::= .cfi_offset register, offset 5034 bool MasmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 5035 int64_t Register = 0; 5036 int64_t Offset = 0; 5037 5038 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 5039 parseToken(AsmToken::Comma, "unexpected token in directive") || 5040 parseAbsoluteExpression(Offset)) 5041 return true; 5042 5043 getStreamer().emitCFIOffset(Register, Offset); 5044 return false; 5045 } 5046 5047 /// parseDirectiveCFIRelOffset 5048 /// ::= .cfi_rel_offset register, offset 5049 bool MasmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 5050 int64_t Register = 0, Offset = 0; 5051 5052 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || 5053 parseToken(AsmToken::Comma, "unexpected token in directive") || 5054 parseAbsoluteExpression(Offset)) 5055 return true; 5056 5057 getStreamer().emitCFIRelOffset(Register, Offset); 5058 return false; 5059 } 5060 5061 static bool isValidEncoding(int64_t Encoding) { 5062 if (Encoding & ~0xff) 5063 return false; 5064 5065 if (Encoding == dwarf::DW_EH_PE_omit) 5066 return true; 5067 5068 const unsigned Format = Encoding & 0xf; 5069 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 5070 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 5071 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 5072 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 5073 return false; 5074 5075 const unsigned Application = Encoding & 0x70; 5076 if (Application != dwarf::DW_EH_PE_absptr && 5077 Application != dwarf::DW_EH_PE_pcrel) 5078 return false; 5079 5080 return true; 5081 } 5082 5083 /// parseDirectiveCFIPersonalityOrLsda 5084 /// IsPersonality true for cfi_personality, false for cfi_lsda 5085 /// ::= .cfi_personality encoding, [symbol_name] 5086 /// ::= .cfi_lsda encoding, [symbol_name] 5087 bool MasmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 5088 int64_t Encoding = 0; 5089 if (parseAbsoluteExpression(Encoding)) 5090 return true; 5091 if (Encoding == dwarf::DW_EH_PE_omit) 5092 return false; 5093 5094 StringRef Name; 5095 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 5096 parseToken(AsmToken::Comma, "unexpected token in directive") || 5097 check(parseIdentifier(Name), "expected identifier in directive")) 5098 return true; 5099 5100 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5101 5102 if (IsPersonality) 5103 getStreamer().emitCFIPersonality(Sym, Encoding); 5104 else 5105 getStreamer().emitCFILsda(Sym, Encoding); 5106 return false; 5107 } 5108 5109 /// parseDirectiveCFIRememberState 5110 /// ::= .cfi_remember_state 5111 bool MasmParser::parseDirectiveCFIRememberState() { 5112 getStreamer().emitCFIRememberState(); 5113 return false; 5114 } 5115 5116 /// parseDirectiveCFIRestoreState 5117 /// ::= .cfi_remember_state 5118 bool MasmParser::parseDirectiveCFIRestoreState() { 5119 getStreamer().emitCFIRestoreState(); 5120 return false; 5121 } 5122 5123 /// parseDirectiveCFISameValue 5124 /// ::= .cfi_same_value register 5125 bool MasmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 5126 int64_t Register = 0; 5127 5128 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5129 return true; 5130 5131 getStreamer().emitCFISameValue(Register); 5132 return false; 5133 } 5134 5135 /// parseDirectiveCFIRestore 5136 /// ::= .cfi_restore register 5137 bool MasmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 5138 int64_t Register = 0; 5139 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5140 return true; 5141 5142 getStreamer().emitCFIRestore(Register); 5143 return false; 5144 } 5145 5146 /// parseDirectiveCFIEscape 5147 /// ::= .cfi_escape expression[,...] 5148 bool MasmParser::parseDirectiveCFIEscape() { 5149 std::string Values; 5150 int64_t CurrValue; 5151 if (parseAbsoluteExpression(CurrValue)) 5152 return true; 5153 5154 Values.push_back((uint8_t)CurrValue); 5155 5156 while (getLexer().is(AsmToken::Comma)) { 5157 Lex(); 5158 5159 if (parseAbsoluteExpression(CurrValue)) 5160 return true; 5161 5162 Values.push_back((uint8_t)CurrValue); 5163 } 5164 5165 getStreamer().emitCFIEscape(Values); 5166 return false; 5167 } 5168 5169 /// parseDirectiveCFIReturnColumn 5170 /// ::= .cfi_return_column register 5171 bool MasmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 5172 int64_t Register = 0; 5173 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5174 return true; 5175 getStreamer().emitCFIReturnColumn(Register); 5176 return false; 5177 } 5178 5179 /// parseDirectiveCFISignalFrame 5180 /// ::= .cfi_signal_frame 5181 bool MasmParser::parseDirectiveCFISignalFrame() { 5182 if (parseToken(AsmToken::EndOfStatement, 5183 "unexpected token in '.cfi_signal_frame'")) 5184 return true; 5185 5186 getStreamer().emitCFISignalFrame(); 5187 return false; 5188 } 5189 5190 /// parseDirectiveCFIUndefined 5191 /// ::= .cfi_undefined register 5192 bool MasmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 5193 int64_t Register = 0; 5194 5195 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) 5196 return true; 5197 5198 getStreamer().emitCFIUndefined(Register); 5199 return false; 5200 } 5201 5202 /// parseDirectiveAltmacro 5203 /// ::= .altmacro 5204 /// ::= .noaltmacro 5205 bool MasmParser::parseDirectiveAltmacro(StringRef Directive) { 5206 if (getLexer().isNot(AsmToken::EndOfStatement)) 5207 return TokError("unexpected token in '" + Directive + "' directive"); 5208 AltMacroMode = (Directive == ".altmacro"); 5209 return false; 5210 } 5211 5212 /// parseDirectiveMacro 5213 /// ::= .macro name[,] [parameters] 5214 bool MasmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 5215 StringRef Name; 5216 if (parseIdentifier(Name)) 5217 return TokError("expected identifier in '.macro' directive"); 5218 5219 if (getLexer().is(AsmToken::Comma)) 5220 Lex(); 5221 5222 MCAsmMacroParameters Parameters; 5223 while (getLexer().isNot(AsmToken::EndOfStatement)) { 5224 5225 if (!Parameters.empty() && Parameters.back().Vararg) 5226 return Error(Lexer.getLoc(), 5227 "Vararg parameter '" + Parameters.back().Name + 5228 "' should be last one in the list of parameters."); 5229 5230 MCAsmMacroParameter Parameter; 5231 if (parseIdentifier(Parameter.Name)) 5232 return TokError("expected identifier in '.macro' directive"); 5233 5234 // Emit an error if two (or more) named parameters share the same name. 5235 for (const MCAsmMacroParameter& CurrParam : Parameters) 5236 if (CurrParam.Name.equals(Parameter.Name)) 5237 return TokError("macro '" + Name + "' has multiple parameters" 5238 " named '" + Parameter.Name + "'"); 5239 5240 if (Lexer.is(AsmToken::Colon)) { 5241 Lex(); // consume ':' 5242 5243 SMLoc QualLoc; 5244 StringRef Qualifier; 5245 5246 QualLoc = Lexer.getLoc(); 5247 if (parseIdentifier(Qualifier)) 5248 return Error(QualLoc, "missing parameter qualifier for " 5249 "'" + Parameter.Name + "' in macro '" + Name + "'"); 5250 5251 if (Qualifier == "req") 5252 Parameter.Required = true; 5253 else if (Qualifier == "vararg") 5254 Parameter.Vararg = true; 5255 else 5256 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 5257 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 5258 } 5259 5260 if (getLexer().is(AsmToken::Equal)) { 5261 Lex(); 5262 5263 SMLoc ParamLoc; 5264 5265 ParamLoc = Lexer.getLoc(); 5266 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 5267 return true; 5268 5269 if (Parameter.Required) 5270 Warning(ParamLoc, "pointless default value for required parameter " 5271 "'" + Parameter.Name + "' in macro '" + Name + "'"); 5272 } 5273 5274 Parameters.push_back(std::move(Parameter)); 5275 5276 if (getLexer().is(AsmToken::Comma)) 5277 Lex(); 5278 } 5279 5280 // Eat just the end of statement. 5281 Lexer.Lex(); 5282 5283 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors. 5284 AsmToken EndToken, StartToken = getTok(); 5285 unsigned MacroDepth = 0; 5286 // Lex the macro definition. 5287 while (true) { 5288 // Ignore Lexing errors in macros. 5289 while (Lexer.is(AsmToken::Error)) { 5290 Lexer.Lex(); 5291 } 5292 5293 // Check whether we have reached the end of the file. 5294 if (getLexer().is(AsmToken::Eof)) 5295 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 5296 5297 // Otherwise, check whether we have reach the .endmacro. 5298 if (getLexer().is(AsmToken::Identifier)) { 5299 if (getTok().getIdentifier() == ".endm" || 5300 getTok().getIdentifier() == ".endmacro") { 5301 if (MacroDepth == 0) { // Outermost macro. 5302 EndToken = getTok(); 5303 Lexer.Lex(); 5304 if (getLexer().isNot(AsmToken::EndOfStatement)) 5305 return TokError("unexpected token in '" + EndToken.getIdentifier() + 5306 "' directive"); 5307 break; 5308 } else { 5309 // Otherwise we just found the end of an inner macro. 5310 --MacroDepth; 5311 } 5312 } else if (getTok().getIdentifier() == ".macro") { 5313 // We allow nested macros. Those aren't instantiated until the outermost 5314 // macro is expanded so just ignore them for now. 5315 ++MacroDepth; 5316 } 5317 } 5318 5319 // Otherwise, scan til the end of the statement. 5320 eatToEndOfStatement(); 5321 } 5322 5323 if (getContext().lookupMacro(Name)) { 5324 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 5325 } 5326 5327 const char *BodyStart = StartToken.getLoc().getPointer(); 5328 const char *BodyEnd = EndToken.getLoc().getPointer(); 5329 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5330 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 5331 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 5332 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 5333 Macro.dump()); 5334 getContext().defineMacro(Name, std::move(Macro)); 5335 return false; 5336 } 5337 5338 /// checkForBadMacro 5339 /// 5340 /// With the support added for named parameters there may be code out there that 5341 /// is transitioning from positional parameters. In versions of gas that did 5342 /// not support named parameters they would be ignored on the macro definition. 5343 /// But to support both styles of parameters this is not possible so if a macro 5344 /// definition has named parameters but does not use them and has what appears 5345 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 5346 /// warning that the positional parameter found in body which have no effect. 5347 /// Hoping the developer will either remove the named parameters from the macro 5348 /// definition so the positional parameters get used if that was what was 5349 /// intended or change the macro to use the named parameters. It is possible 5350 /// this warning will trigger when the none of the named parameters are used 5351 /// and the strings like $1 are infact to simply to be passed trough unchanged. 5352 void MasmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 5353 StringRef Body, 5354 ArrayRef<MCAsmMacroParameter> Parameters) { 5355 // If this macro is not defined with named parameters the warning we are 5356 // checking for here doesn't apply. 5357 unsigned NParameters = Parameters.size(); 5358 if (NParameters == 0) 5359 return; 5360 5361 bool NamedParametersFound = false; 5362 bool PositionalParametersFound = false; 5363 5364 // Look at the body of the macro for use of both the named parameters and what 5365 // are likely to be positional parameters. This is what expandMacro() is 5366 // doing when it finds the parameters in the body. 5367 while (!Body.empty()) { 5368 // Scan for the next possible parameter. 5369 std::size_t End = Body.size(), Pos = 0; 5370 for (; Pos != End; ++Pos) { 5371 // Check for a substitution or escape. 5372 // This macro is defined with parameters, look for \foo, \bar, etc. 5373 if (Body[Pos] == '\\' && Pos + 1 != End) 5374 break; 5375 5376 // This macro should have parameters, but look for $0, $1, ..., $n too. 5377 if (Body[Pos] != '$' || Pos + 1 == End) 5378 continue; 5379 char Next = Body[Pos + 1]; 5380 if (Next == '$' || Next == 'n' || 5381 isdigit(static_cast<unsigned char>(Next))) 5382 break; 5383 } 5384 5385 // Check if we reached the end. 5386 if (Pos == End) 5387 break; 5388 5389 if (Body[Pos] == '$') { 5390 switch (Body[Pos + 1]) { 5391 // $$ => $ 5392 case '$': 5393 break; 5394 5395 // $n => number of arguments 5396 case 'n': 5397 PositionalParametersFound = true; 5398 break; 5399 5400 // $[0-9] => argument 5401 default: { 5402 PositionalParametersFound = true; 5403 break; 5404 } 5405 } 5406 Pos += 2; 5407 } else { 5408 unsigned I = Pos + 1; 5409 while (isIdentifierChar(Body[I]) && I + 1 != End) 5410 ++I; 5411 5412 const char *Begin = Body.data() + Pos + 1; 5413 StringRef Argument(Begin, I - (Pos + 1)); 5414 unsigned Index = 0; 5415 for (; Index < NParameters; ++Index) 5416 if (Parameters[Index].Name == Argument) 5417 break; 5418 5419 if (Index == NParameters) { 5420 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 5421 Pos += 3; 5422 else { 5423 Pos = I; 5424 } 5425 } else { 5426 NamedParametersFound = true; 5427 Pos += 1 + Argument.size(); 5428 } 5429 } 5430 // Update the scan point. 5431 Body = Body.substr(Pos); 5432 } 5433 5434 if (!NamedParametersFound && PositionalParametersFound) 5435 Warning(DirectiveLoc, "macro defined with named parameters which are not " 5436 "used in macro body, possible positional parameter " 5437 "found in body which will have no effect"); 5438 } 5439 5440 /// parseDirectiveExitMacro 5441 /// ::= .exitm 5442 bool MasmParser::parseDirectiveExitMacro(StringRef Directive) { 5443 if (parseToken(AsmToken::EndOfStatement, 5444 "unexpected token in '" + Directive + "' directive")) 5445 return true; 5446 5447 if (!isInsideMacroInstantiation()) 5448 return TokError("unexpected '" + Directive + "' in file, " 5449 "no current macro definition"); 5450 5451 // Exit all conditionals that are active in the current macro. 5452 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 5453 TheCondState = TheCondStack.back(); 5454 TheCondStack.pop_back(); 5455 } 5456 5457 handleMacroExit(); 5458 return false; 5459 } 5460 5461 /// parseDirectiveEndMacro 5462 /// ::= .endm 5463 /// ::= .endmacro 5464 bool MasmParser::parseDirectiveEndMacro(StringRef Directive) { 5465 if (getLexer().isNot(AsmToken::EndOfStatement)) 5466 return TokError("unexpected token in '" + Directive + "' directive"); 5467 5468 // If we are inside a macro instantiation, terminate the current 5469 // instantiation. 5470 if (isInsideMacroInstantiation()) { 5471 handleMacroExit(); 5472 return false; 5473 } 5474 5475 // Otherwise, this .endmacro is a stray entry in the file; well formed 5476 // .endmacro directives are handled during the macro definition parsing. 5477 return TokError("unexpected '" + Directive + "' in file, " 5478 "no current macro definition"); 5479 } 5480 5481 /// parseDirectivePurgeMacro 5482 /// ::= .purgem 5483 bool MasmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 5484 StringRef Name; 5485 SMLoc Loc; 5486 if (parseTokenLoc(Loc) || 5487 check(parseIdentifier(Name), Loc, 5488 "expected identifier in '.purgem' directive") || 5489 parseToken(AsmToken::EndOfStatement, 5490 "unexpected token in '.purgem' directive")) 5491 return true; 5492 5493 if (!getContext().lookupMacro(Name)) 5494 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 5495 5496 getContext().undefineMacro(Name); 5497 DEBUG_WITH_TYPE("asm-macros", dbgs() 5498 << "Un-defining macro: " << Name << "\n"); 5499 return false; 5500 } 5501 5502 /// parseDirectiveSymbolAttribute 5503 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 5504 bool MasmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 5505 auto parseOp = [&]() -> bool { 5506 StringRef Name; 5507 SMLoc Loc = getTok().getLoc(); 5508 if (parseIdentifier(Name)) 5509 return Error(Loc, "expected identifier"); 5510 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5511 5512 // Assembler local symbols don't make any sense here. Complain loudly. 5513 if (Sym->isTemporary()) 5514 return Error(Loc, "non-local symbol required"); 5515 5516 if (!getStreamer().emitSymbolAttribute(Sym, Attr)) 5517 return Error(Loc, "unable to emit symbol attribute"); 5518 return false; 5519 }; 5520 5521 if (parseMany(parseOp)) 5522 return addErrorSuffix(" in directive"); 5523 return false; 5524 } 5525 5526 /// parseDirectiveComm 5527 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 5528 bool MasmParser::parseDirectiveComm(bool IsLocal) { 5529 if (checkForValidSection()) 5530 return true; 5531 5532 SMLoc IDLoc = getLexer().getLoc(); 5533 StringRef Name; 5534 if (parseIdentifier(Name)) 5535 return TokError("expected identifier in directive"); 5536 5537 // Handle the identifier as the key symbol. 5538 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5539 5540 if (getLexer().isNot(AsmToken::Comma)) 5541 return TokError("unexpected token in directive"); 5542 Lex(); 5543 5544 int64_t Size; 5545 SMLoc SizeLoc = getLexer().getLoc(); 5546 if (parseAbsoluteExpression(Size)) 5547 return true; 5548 5549 int64_t Pow2Alignment = 0; 5550 SMLoc Pow2AlignmentLoc; 5551 if (getLexer().is(AsmToken::Comma)) { 5552 Lex(); 5553 Pow2AlignmentLoc = getLexer().getLoc(); 5554 if (parseAbsoluteExpression(Pow2Alignment)) 5555 return true; 5556 5557 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 5558 if (IsLocal && LCOMM == LCOMM::NoAlignment) 5559 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 5560 5561 // If this target takes alignments in bytes (not log) validate and convert. 5562 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 5563 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 5564 if (!isPowerOf2_64(Pow2Alignment)) 5565 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 5566 Pow2Alignment = Log2_64(Pow2Alignment); 5567 } 5568 } 5569 5570 if (parseToken(AsmToken::EndOfStatement, 5571 "unexpected token in '.comm' or '.lcomm' directive")) 5572 return true; 5573 5574 // NOTE: a size of zero for a .comm should create a undefined symbol 5575 // but a size of .lcomm creates a bss symbol of size zero. 5576 if (Size < 0) 5577 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 5578 "be less than zero"); 5579 5580 // NOTE: The alignment in the directive is a power of 2 value, the assembler 5581 // may internally end up wanting an alignment in bytes. 5582 // FIXME: Diagnose overflow. 5583 if (Pow2Alignment < 0) 5584 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 5585 "alignment, can't be less than zero"); 5586 5587 Sym->redefineIfPossible(); 5588 if (!Sym->isUndefined()) 5589 return Error(IDLoc, "invalid symbol redefinition"); 5590 5591 // Create the Symbol as a common or local common with Size and Pow2Alignment. 5592 if (IsLocal) { 5593 getStreamer().emitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5594 return false; 5595 } 5596 5597 getStreamer().emitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5598 return false; 5599 } 5600 5601 /// parseDirectiveComment 5602 /// ::= comment delimiter [[text]] 5603 /// [[text]] 5604 /// [[text]] delimiter [[text]] 5605 bool MasmParser::parseDirectiveComment(SMLoc DirectiveLoc) { 5606 StringRef FirstLine = parseStringToEndOfStatement(); 5607 size_t DelimiterEnd = FirstLine.find_first_of("\b\t\v\f\r\x1A "); 5608 StringRef Delimiter = FirstLine.take_front(DelimiterEnd); 5609 if (Delimiter.empty()) 5610 return Error(DirectiveLoc, "no delimiter in 'comment' directive"); 5611 do { 5612 if (getTok().is(AsmToken::Eof)) 5613 return Error(DirectiveLoc, "unmatched delimiter in 'comment' directive"); 5614 Lex(); // eat end of statement 5615 } while (!parseStringToEndOfStatement().contains(Delimiter)); 5616 return parseToken(AsmToken::EndOfStatement, 5617 "unexpected token in 'comment' directive"); 5618 } 5619 5620 /// parseDirectiveInclude 5621 /// ::= include <filename> 5622 /// | include filename 5623 bool MasmParser::parseDirectiveInclude() { 5624 // Allow the strings to have escaped octal character sequence. 5625 std::string Filename; 5626 SMLoc IncludeLoc = getTok().getLoc(); 5627 5628 if (!parseAngleBracketString(Filename)) 5629 Filename = parseStringToEndOfStatement().str(); 5630 if (check(!Filename.empty(), "missing filename in 'include' directive") || 5631 check(getTok().isNot(AsmToken::EndOfStatement), 5632 "unexpected token in 'include' directive") || 5633 // Attempt to switch the lexer to the included file before consuming the 5634 // end of statement to avoid losing it when we switch. 5635 check(enterIncludeFile(Filename), IncludeLoc, 5636 "Could not find include file '" + Filename + "'")) 5637 return true; 5638 5639 return false; 5640 } 5641 5642 /// parseDirectiveIf 5643 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5644 bool MasmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5645 TheCondStack.push_back(TheCondState); 5646 TheCondState.TheCond = AsmCond::IfCond; 5647 if (TheCondState.Ignore) { 5648 eatToEndOfStatement(); 5649 } else { 5650 int64_t ExprValue; 5651 if (parseAbsoluteExpression(ExprValue) || 5652 parseToken(AsmToken::EndOfStatement, 5653 "unexpected token in '.if' directive")) 5654 return true; 5655 5656 switch (DirKind) { 5657 default: 5658 llvm_unreachable("unsupported directive"); 5659 case DK_IF: 5660 break; 5661 case DK_IFE: 5662 ExprValue = ExprValue == 0; 5663 break; 5664 } 5665 5666 TheCondState.CondMet = ExprValue; 5667 TheCondState.Ignore = !TheCondState.CondMet; 5668 } 5669 5670 return false; 5671 } 5672 5673 /// parseDirectiveIfb 5674 /// ::= .ifb string 5675 bool MasmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5676 TheCondStack.push_back(TheCondState); 5677 TheCondState.TheCond = AsmCond::IfCond; 5678 5679 if (TheCondState.Ignore) { 5680 eatToEndOfStatement(); 5681 } else { 5682 std::string Str; 5683 if (parseTextItem(Str)) 5684 return TokError("expected string parameter for 'ifb' directive"); 5685 5686 if (parseToken(AsmToken::EndOfStatement, 5687 "unexpected token in 'ifb' directive")) 5688 return true; 5689 5690 TheCondState.CondMet = ExpectBlank == Str.empty(); 5691 TheCondState.Ignore = !TheCondState.CondMet; 5692 } 5693 5694 return false; 5695 } 5696 5697 /// parseDirectiveIfidn 5698 /// ::= ifidn string1, string2 5699 bool MasmParser::parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual, bool CaseInsensitive) { 5700 std::string String1, String2; 5701 5702 if (parseTextItem(String1)) { 5703 if (ExpectEqual) 5704 return TokError("expected string parameter for 'ifidn' directive"); 5705 return TokError("expected string parameter for 'ifdif' directive"); 5706 } 5707 5708 if (Lexer.isNot(AsmToken::Comma)) { 5709 if (ExpectEqual) 5710 return TokError( 5711 "expected comma after first string for 'ifidn' directive"); 5712 return TokError("expected comma after first string for 'ifdif' directive"); 5713 } 5714 Lex(); 5715 5716 if (parseTextItem(String2)) { 5717 if (ExpectEqual) 5718 return TokError("expected string parameter for 'ifidn' directive"); 5719 return TokError("expected string parameter for 'ifdif' directive"); 5720 } 5721 5722 TheCondStack.push_back(TheCondState); 5723 TheCondState.TheCond = AsmCond::IfCond; 5724 if (CaseInsensitive) 5725 TheCondState.CondMet = 5726 ExpectEqual == (StringRef(String1).equals_lower(String2)); 5727 else 5728 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5729 TheCondState.Ignore = !TheCondState.CondMet; 5730 5731 return false; 5732 } 5733 5734 /// parseDirectiveIfdef 5735 /// ::= ifdef symbol 5736 /// | ifdef variable 5737 bool MasmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5738 TheCondStack.push_back(TheCondState); 5739 TheCondState.TheCond = AsmCond::IfCond; 5740 5741 if (TheCondState.Ignore) { 5742 eatToEndOfStatement(); 5743 } else { 5744 bool is_defined = false; 5745 unsigned RegNo; 5746 SMLoc StartLoc, EndLoc; 5747 is_defined = (getTargetParser().tryParseRegister( 5748 RegNo, StartLoc, EndLoc) == MatchOperand_Success); 5749 if (!is_defined) { 5750 StringRef Name; 5751 if (check(parseIdentifier(Name), "expected identifier after 'ifdef'") || 5752 parseToken(AsmToken::EndOfStatement, "unexpected token in 'ifdef'")) 5753 return true; 5754 5755 if (Variables.find(Name) != Variables.end()) { 5756 is_defined = true; 5757 } else { 5758 MCSymbol *Sym = getContext().lookupSymbol(Name); 5759 is_defined = (Sym && !Sym->isUndefined(false)); 5760 } 5761 } 5762 5763 TheCondState.CondMet = (is_defined == expect_defined); 5764 TheCondState.Ignore = !TheCondState.CondMet; 5765 } 5766 5767 return false; 5768 } 5769 5770 /// parseDirectiveElseIf 5771 /// ::= elseif expression 5772 bool MasmParser::parseDirectiveElseIf(SMLoc DirectiveLoc, 5773 DirectiveKind DirKind) { 5774 if (TheCondState.TheCond != AsmCond::IfCond && 5775 TheCondState.TheCond != AsmCond::ElseIfCond) 5776 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5777 " .if or an .elseif"); 5778 TheCondState.TheCond = AsmCond::ElseIfCond; 5779 5780 bool LastIgnoreState = false; 5781 if (!TheCondStack.empty()) 5782 LastIgnoreState = TheCondStack.back().Ignore; 5783 if (LastIgnoreState || TheCondState.CondMet) { 5784 TheCondState.Ignore = true; 5785 eatToEndOfStatement(); 5786 } else { 5787 int64_t ExprValue; 5788 if (parseAbsoluteExpression(ExprValue)) 5789 return true; 5790 5791 if (parseToken(AsmToken::EndOfStatement, 5792 "unexpected token in '.elseif' directive")) 5793 return true; 5794 5795 switch (DirKind) { 5796 default: 5797 llvm_unreachable("unsupported directive"); 5798 case DK_ELSEIF: 5799 break; 5800 case DK_ELSEIFE: 5801 ExprValue = ExprValue == 0; 5802 break; 5803 } 5804 5805 TheCondState.CondMet = ExprValue; 5806 TheCondState.Ignore = !TheCondState.CondMet; 5807 } 5808 5809 return false; 5810 } 5811 5812 /// parseDirectiveElseIfb 5813 /// ::= elseifb expression 5814 bool MasmParser::parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5815 if (TheCondState.TheCond != AsmCond::IfCond && 5816 TheCondState.TheCond != AsmCond::ElseIfCond) 5817 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5818 " if or an elseif"); 5819 TheCondState.TheCond = AsmCond::ElseIfCond; 5820 5821 bool LastIgnoreState = false; 5822 if (!TheCondStack.empty()) 5823 LastIgnoreState = TheCondStack.back().Ignore; 5824 if (LastIgnoreState || TheCondState.CondMet) { 5825 TheCondState.Ignore = true; 5826 eatToEndOfStatement(); 5827 } else { 5828 std::string Str; 5829 if (parseTextItem(Str)) 5830 return TokError("expected string parameter for 'elseifb' directive"); 5831 5832 if (parseToken(AsmToken::EndOfStatement, 5833 "unexpected token in 'elseifb' directive")) 5834 return true; 5835 5836 TheCondState.CondMet = ExpectBlank == Str.empty(); 5837 TheCondState.Ignore = !TheCondState.CondMet; 5838 } 5839 5840 return false; 5841 } 5842 5843 /// parseDirectiveElseIfdef 5844 /// ::= elseifdef symbol 5845 /// | elseifdef variable 5846 bool MasmParser::parseDirectiveElseIfdef(SMLoc DirectiveLoc, 5847 bool expect_defined) { 5848 if (TheCondState.TheCond != AsmCond::IfCond && 5849 TheCondState.TheCond != AsmCond::ElseIfCond) 5850 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5851 " if or an elseif"); 5852 TheCondState.TheCond = AsmCond::ElseIfCond; 5853 5854 bool LastIgnoreState = false; 5855 if (!TheCondStack.empty()) 5856 LastIgnoreState = TheCondStack.back().Ignore; 5857 if (LastIgnoreState || TheCondState.CondMet) { 5858 TheCondState.Ignore = true; 5859 eatToEndOfStatement(); 5860 } else { 5861 bool is_defined = false; 5862 unsigned RegNo; 5863 SMLoc StartLoc, EndLoc; 5864 is_defined = (getTargetParser().tryParseRegister(RegNo, StartLoc, EndLoc) == 5865 MatchOperand_Success); 5866 if (!is_defined) { 5867 StringRef Name; 5868 if (check(parseIdentifier(Name), 5869 "expected identifier after 'elseifdef'") || 5870 parseToken(AsmToken::EndOfStatement, 5871 "unexpected token in 'elseifdef'")) 5872 return true; 5873 5874 if (Variables.find(Name) != Variables.end()) { 5875 is_defined = true; 5876 } else { 5877 MCSymbol *Sym = getContext().lookupSymbol(Name); 5878 is_defined = (Sym && !Sym->isUndefined(false)); 5879 } 5880 } 5881 5882 TheCondState.CondMet = (is_defined == expect_defined); 5883 TheCondState.Ignore = !TheCondState.CondMet; 5884 } 5885 5886 return false; 5887 } 5888 5889 /// parseDirectiveElseIfidn 5890 /// ::= elseifidn string1, string2 5891 bool MasmParser::parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 5892 bool CaseInsensitive) { 5893 if (TheCondState.TheCond != AsmCond::IfCond && 5894 TheCondState.TheCond != AsmCond::ElseIfCond) 5895 return Error(DirectiveLoc, "Encountered an elseif that doesn't follow an" 5896 " if or an elseif"); 5897 TheCondState.TheCond = AsmCond::ElseIfCond; 5898 5899 bool LastIgnoreState = false; 5900 if (!TheCondStack.empty()) 5901 LastIgnoreState = TheCondStack.back().Ignore; 5902 if (LastIgnoreState || TheCondState.CondMet) { 5903 TheCondState.Ignore = true; 5904 eatToEndOfStatement(); 5905 } else { 5906 std::string String1, String2; 5907 5908 if (parseTextItem(String1)) { 5909 if (ExpectEqual) 5910 return TokError("expected string parameter for 'elseifidn' directive"); 5911 return TokError("expected string parameter for 'elseifdif' directive"); 5912 } 5913 5914 if (Lexer.isNot(AsmToken::Comma)) { 5915 if (ExpectEqual) 5916 return TokError( 5917 "expected comma after first string for 'elseifidn' directive"); 5918 return TokError( 5919 "expected comma after first string for 'elseifdif' directive"); 5920 } 5921 Lex(); 5922 5923 if (parseTextItem(String2)) { 5924 if (ExpectEqual) 5925 return TokError("expected string parameter for 'elseifidn' directive"); 5926 return TokError("expected string parameter for 'elseifdif' directive"); 5927 } 5928 5929 if (CaseInsensitive) 5930 TheCondState.CondMet = 5931 ExpectEqual == (StringRef(String1).equals_lower(String2)); 5932 else 5933 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5934 TheCondState.Ignore = !TheCondState.CondMet; 5935 } 5936 5937 return false; 5938 } 5939 5940 /// parseDirectiveElse 5941 /// ::= else 5942 bool MasmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5943 if (parseToken(AsmToken::EndOfStatement, 5944 "unexpected token in 'else' directive")) 5945 return true; 5946 5947 if (TheCondState.TheCond != AsmCond::IfCond && 5948 TheCondState.TheCond != AsmCond::ElseIfCond) 5949 return Error(DirectiveLoc, "Encountered an else that doesn't follow an if" 5950 " or an elseif"); 5951 TheCondState.TheCond = AsmCond::ElseCond; 5952 bool LastIgnoreState = false; 5953 if (!TheCondStack.empty()) 5954 LastIgnoreState = TheCondStack.back().Ignore; 5955 if (LastIgnoreState || TheCondState.CondMet) 5956 TheCondState.Ignore = true; 5957 else 5958 TheCondState.Ignore = false; 5959 5960 return false; 5961 } 5962 5963 /// parseDirectiveEnd 5964 /// ::= end 5965 bool MasmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5966 if (parseToken(AsmToken::EndOfStatement, 5967 "unexpected token in 'end' directive")) 5968 return true; 5969 5970 while (Lexer.isNot(AsmToken::Eof)) 5971 Lexer.Lex(); 5972 5973 return false; 5974 } 5975 5976 /// parseDirectiveError 5977 /// ::= .err [message] 5978 bool MasmParser::parseDirectiveError(SMLoc DirectiveLoc) { 5979 if (!TheCondStack.empty()) { 5980 if (TheCondStack.back().Ignore) { 5981 eatToEndOfStatement(); 5982 return false; 5983 } 5984 } 5985 5986 StringRef Message = ".err directive invoked in source file"; 5987 if (Lexer.isNot(AsmToken::EndOfStatement)) 5988 Message = parseStringToEndOfStatement(); 5989 Lex(); 5990 5991 return Error(DirectiveLoc, Message); 5992 } 5993 5994 /// parseDirectiveErrorIfb 5995 /// ::= .errb textitem[, message] 5996 bool MasmParser::parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5997 if (!TheCondStack.empty()) { 5998 if (TheCondStack.back().Ignore) { 5999 eatToEndOfStatement(); 6000 return false; 6001 } 6002 } 6003 6004 std::string Text; 6005 if (parseTextItem(Text)) 6006 return Error(getTok().getLoc(), "missing text item in '.errb' directive"); 6007 6008 StringRef Message = ".errb directive invoked in source file"; 6009 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6010 if (parseToken(AsmToken::Comma)) 6011 return addErrorSuffix(" in '.errb' directive"); 6012 Message = parseStringToEndOfStatement(); 6013 } 6014 Lex(); 6015 6016 if (Text.empty() == ExpectBlank) 6017 return Error(DirectiveLoc, Message); 6018 return false; 6019 } 6020 6021 /// parseDirectiveErrorIfdef 6022 /// ::= .errdef name[, message] 6023 bool MasmParser::parseDirectiveErrorIfdef(SMLoc DirectiveLoc, 6024 bool ExpectDefined) { 6025 if (!TheCondStack.empty()) { 6026 if (TheCondStack.back().Ignore) { 6027 eatToEndOfStatement(); 6028 return false; 6029 } 6030 } 6031 6032 bool IsDefined = false; 6033 unsigned RegNo; 6034 SMLoc StartLoc, EndLoc; 6035 IsDefined = (getTargetParser().tryParseRegister(RegNo, StartLoc, EndLoc) == 6036 MatchOperand_Success); 6037 if (!IsDefined) { 6038 StringRef Name; 6039 if (check(parseIdentifier(Name), "expected identifier after '.errdef'")) 6040 return true; 6041 6042 if (Variables.find(Name) != Variables.end()) { 6043 IsDefined = true; 6044 } else { 6045 MCSymbol *Sym = getContext().lookupSymbol(Name); 6046 IsDefined = (Sym && !Sym->isUndefined(false)); 6047 } 6048 } 6049 6050 StringRef Message = ".errdef directive invoked in source file"; 6051 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6052 if (parseToken(AsmToken::Comma)) 6053 return addErrorSuffix(" in '.errdef' directive"); 6054 Message = parseStringToEndOfStatement(); 6055 } 6056 Lex(); 6057 6058 if (IsDefined == ExpectDefined) 6059 return Error(DirectiveLoc, Message); 6060 return false; 6061 } 6062 6063 /// parseDirectiveErrorIfidn 6064 /// ::= .erridn textitem1, textitem2[, message] 6065 bool MasmParser::parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual, 6066 bool CaseInsensitive) { 6067 if (!TheCondStack.empty()) { 6068 if (TheCondStack.back().Ignore) { 6069 eatToEndOfStatement(); 6070 return false; 6071 } 6072 } 6073 6074 std::string String1, String2; 6075 6076 if (parseTextItem(String1)) { 6077 if (ExpectEqual) 6078 return TokError("expected string parameter for '.erridn' directive"); 6079 return TokError("expected string parameter for '.errdif' directive"); 6080 } 6081 6082 if (Lexer.isNot(AsmToken::Comma)) { 6083 if (ExpectEqual) 6084 return TokError( 6085 "expected comma after first string for '.erridn' directive"); 6086 return TokError( 6087 "expected comma after first string for '.errdif' directive"); 6088 } 6089 Lex(); 6090 6091 if (parseTextItem(String2)) { 6092 if (ExpectEqual) 6093 return TokError("expected string parameter for '.erridn' directive"); 6094 return TokError("expected string parameter for '.errdif' directive"); 6095 } 6096 6097 StringRef Message; 6098 if (ExpectEqual) 6099 Message = ".erridn directive invoked in source file"; 6100 else 6101 Message = ".errdif directive invoked in source file"; 6102 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6103 if (parseToken(AsmToken::Comma)) 6104 return addErrorSuffix(" in '.erridn' directive"); 6105 Message = parseStringToEndOfStatement(); 6106 } 6107 Lex(); 6108 6109 if (CaseInsensitive) 6110 TheCondState.CondMet = 6111 ExpectEqual == (StringRef(String1).equals_lower(String2)); 6112 else 6113 TheCondState.CondMet = ExpectEqual == (String1 == String2); 6114 TheCondState.Ignore = !TheCondState.CondMet; 6115 6116 if ((CaseInsensitive && 6117 ExpectEqual == StringRef(String1).equals_lower(String2)) || 6118 (ExpectEqual == (String1 == String2))) 6119 return Error(DirectiveLoc, Message); 6120 return false; 6121 } 6122 6123 /// parseDirectiveErrorIfe 6124 /// ::= .erre expression[, message] 6125 bool MasmParser::parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero) { 6126 if (!TheCondStack.empty()) { 6127 if (TheCondStack.back().Ignore) { 6128 eatToEndOfStatement(); 6129 return false; 6130 } 6131 } 6132 6133 int64_t ExprValue; 6134 if (parseAbsoluteExpression(ExprValue)) 6135 return addErrorSuffix(" in '.erre' directive"); 6136 6137 StringRef Message = ".erre directive invoked in source file"; 6138 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6139 if (parseToken(AsmToken::Comma)) 6140 return addErrorSuffix(" in '.erre' directive"); 6141 Message = parseStringToEndOfStatement(); 6142 } 6143 Lex(); 6144 6145 if ((ExprValue == 0) == ExpectZero) 6146 return Error(DirectiveLoc, Message); 6147 return false; 6148 } 6149 6150 /// parseDirectiveEndIf 6151 /// ::= .endif 6152 bool MasmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 6153 if (parseToken(AsmToken::EndOfStatement, 6154 "unexpected token in '.endif' directive")) 6155 return true; 6156 6157 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 6158 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 6159 "an .if or .else"); 6160 if (!TheCondStack.empty()) { 6161 TheCondState = TheCondStack.back(); 6162 TheCondStack.pop_back(); 6163 } 6164 6165 return false; 6166 } 6167 6168 void MasmParser::initializeDirectiveKindMap() { 6169 DirectiveKindMap["="] = DK_ASSIGN; 6170 DirectiveKindMap["equ"] = DK_EQU; 6171 DirectiveKindMap["textequ"] = DK_TEXTEQU; 6172 // DirectiveKindMap[".ascii"] = DK_ASCII; 6173 // DirectiveKindMap[".asciz"] = DK_ASCIZ; 6174 // DirectiveKindMap[".string"] = DK_STRING; 6175 DirectiveKindMap["byte"] = DK_BYTE; 6176 DirectiveKindMap["sbyte"] = DK_SBYTE; 6177 DirectiveKindMap["word"] = DK_WORD; 6178 DirectiveKindMap["sword"] = DK_SWORD; 6179 DirectiveKindMap["dword"] = DK_DWORD; 6180 DirectiveKindMap["sdword"] = DK_SDWORD; 6181 DirectiveKindMap["fword"] = DK_FWORD; 6182 DirectiveKindMap["qword"] = DK_QWORD; 6183 DirectiveKindMap["sqword"] = DK_SQWORD; 6184 DirectiveKindMap["real4"] = DK_REAL4; 6185 DirectiveKindMap["real8"] = DK_REAL8; 6186 DirectiveKindMap["align"] = DK_ALIGN; 6187 // DirectiveKindMap[".org"] = DK_ORG; 6188 DirectiveKindMap["extern"] = DK_EXTERN; 6189 DirectiveKindMap["public"] = DK_PUBLIC; 6190 // DirectiveKindMap[".comm"] = DK_COMM; 6191 DirectiveKindMap["comment"] = DK_COMMENT; 6192 DirectiveKindMap["include"] = DK_INCLUDE; 6193 // DirectiveKindMap[".rept"] = DK_REPT; 6194 // DirectiveKindMap[".rep"] = DK_REPT; 6195 // DirectiveKindMap[".irp"] = DK_IRP; 6196 // DirectiveKindMap[".irpc"] = DK_IRPC; 6197 // DirectiveKindMap[".endr"] = DK_ENDR; 6198 DirectiveKindMap["if"] = DK_IF; 6199 DirectiveKindMap["ife"] = DK_IFE; 6200 DirectiveKindMap["ifb"] = DK_IFB; 6201 DirectiveKindMap["ifnb"] = DK_IFNB; 6202 DirectiveKindMap["ifdef"] = DK_IFDEF; 6203 DirectiveKindMap["ifndef"] = DK_IFNDEF; 6204 DirectiveKindMap["ifdif"] = DK_IFDIF; 6205 DirectiveKindMap["ifdifi"] = DK_IFDIFI; 6206 DirectiveKindMap["ifidn"] = DK_IFIDN; 6207 DirectiveKindMap["ifidni"] = DK_IFIDNI; 6208 DirectiveKindMap["elseif"] = DK_ELSEIF; 6209 DirectiveKindMap["elseifdef"] = DK_ELSEIFDEF; 6210 DirectiveKindMap["elseifndef"] = DK_ELSEIFNDEF; 6211 DirectiveKindMap["elseifdif"] = DK_ELSEIFDIF; 6212 DirectiveKindMap["elseifidn"] = DK_ELSEIFIDN; 6213 DirectiveKindMap["else"] = DK_ELSE; 6214 DirectiveKindMap["end"] = DK_END; 6215 DirectiveKindMap["endif"] = DK_ENDIF; 6216 // DirectiveKindMap[".file"] = DK_FILE; 6217 // DirectiveKindMap[".line"] = DK_LINE; 6218 // DirectiveKindMap[".loc"] = DK_LOC; 6219 // DirectiveKindMap[".stabs"] = DK_STABS; 6220 // DirectiveKindMap[".cv_file"] = DK_CV_FILE; 6221 // DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 6222 // DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 6223 // DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 6224 // DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 6225 // DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 6226 // DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 6227 // DirectiveKindMap[".cv_string"] = DK_CV_STRING; 6228 // DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 6229 // DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 6230 // DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 6231 // DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 6232 // DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 6233 // DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 6234 // DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 6235 // DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 6236 // DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 6237 // DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 6238 // DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 6239 // DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 6240 // DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 6241 // DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 6242 // DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 6243 // DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 6244 // DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 6245 // DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 6246 // DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 6247 // DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 6248 // DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 6249 // DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 6250 // DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 6251 // DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 6252 // DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 6253 // DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 6254 // DirectiveKindMap[".macro"] = DK_MACRO; 6255 // DirectiveKindMap[".exitm"] = DK_EXITM; 6256 // DirectiveKindMap[".endm"] = DK_ENDM; 6257 // DirectiveKindMap[".purgem"] = DK_PURGEM; 6258 DirectiveKindMap[".err"] = DK_ERR; 6259 DirectiveKindMap[".errb"] = DK_ERRB; 6260 DirectiveKindMap[".errnb"] = DK_ERRNB; 6261 DirectiveKindMap[".errdef"] = DK_ERRDEF; 6262 DirectiveKindMap[".errndef"] = DK_ERRNDEF; 6263 DirectiveKindMap[".errdif"] = DK_ERRDIF; 6264 DirectiveKindMap[".errdifi"] = DK_ERRDIFI; 6265 DirectiveKindMap[".erridn"] = DK_ERRIDN; 6266 DirectiveKindMap[".erridni"] = DK_ERRIDNI; 6267 DirectiveKindMap[".erre"] = DK_ERRE; 6268 DirectiveKindMap[".errnz"] = DK_ERRNZ; 6269 // DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 6270 // DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 6271 DirectiveKindMap["db"] = DK_DB; 6272 DirectiveKindMap["dd"] = DK_DD; 6273 DirectiveKindMap["dq"] = DK_DQ; 6274 DirectiveKindMap["dw"] = DK_DW; 6275 DirectiveKindMap["echo"] = DK_ECHO; 6276 DirectiveKindMap["struc"] = DK_STRUCT; 6277 DirectiveKindMap["struct"] = DK_STRUCT; 6278 DirectiveKindMap["union"] = DK_UNION; 6279 DirectiveKindMap["ends"] = DK_ENDS; 6280 } 6281 6282 MCAsmMacro *MasmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 6283 AsmToken EndToken, StartToken = getTok(); 6284 6285 unsigned NestLevel = 0; 6286 while (true) { 6287 // Check whether we have reached the end of the file. 6288 if (getLexer().is(AsmToken::Eof)) { 6289 printError(DirectiveLoc, "no matching '.endr' in definition"); 6290 return nullptr; 6291 } 6292 6293 if (Lexer.is(AsmToken::Identifier) && 6294 (getTok().getIdentifier() == ".rep" || 6295 getTok().getIdentifier() == ".rept" || 6296 getTok().getIdentifier() == ".irp" || 6297 getTok().getIdentifier() == ".irpc")) { 6298 ++NestLevel; 6299 } 6300 6301 // Otherwise, check whether we have reached the .endr. 6302 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 6303 if (NestLevel == 0) { 6304 EndToken = getTok(); 6305 Lex(); 6306 if (Lexer.isNot(AsmToken::EndOfStatement)) { 6307 printError(getTok().getLoc(), 6308 "unexpected token in '.endr' directive"); 6309 return nullptr; 6310 } 6311 break; 6312 } 6313 --NestLevel; 6314 } 6315 6316 // Otherwise, scan till the end of the statement. 6317 eatToEndOfStatement(); 6318 } 6319 6320 const char *BodyStart = StartToken.getLoc().getPointer(); 6321 const char *BodyEnd = EndToken.getLoc().getPointer(); 6322 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 6323 6324 // We Are Anonymous. 6325 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 6326 return &MacroLikeBodies.back(); 6327 } 6328 6329 void MasmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 6330 raw_svector_ostream &OS) { 6331 OS << ".endr\n"; 6332 6333 std::unique_ptr<MemoryBuffer> Instantiation = 6334 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 6335 6336 // Create the macro instantiation object and add to the current macro 6337 // instantiation stack. 6338 MacroInstantiation *MI = new MacroInstantiation{ 6339 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 6340 ActiveMacros.push_back(MI); 6341 6342 // Jump to the macro instantiation and prime the lexer. 6343 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 6344 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 6345 Lex(); 6346 } 6347 6348 /// parseDirectiveRept 6349 /// ::= .rep | .rept count 6350 bool MasmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 6351 const MCExpr *CountExpr; 6352 SMLoc CountLoc = getTok().getLoc(); 6353 if (parseExpression(CountExpr)) 6354 return true; 6355 6356 int64_t Count; 6357 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 6358 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 6359 } 6360 6361 if (check(Count < 0, CountLoc, "Count is negative") || 6362 parseToken(AsmToken::EndOfStatement, 6363 "unexpected token in '" + Dir + "' directive")) 6364 return true; 6365 6366 // Lex the rept definition. 6367 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6368 if (!M) 6369 return true; 6370 6371 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6372 // to hold the macro body with substitutions. 6373 SmallString<256> Buf; 6374 raw_svector_ostream OS(Buf); 6375 while (Count--) { 6376 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 6377 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 6378 return true; 6379 } 6380 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6381 6382 return false; 6383 } 6384 6385 /// parseDirectiveIrp 6386 /// ::= .irp symbol,values 6387 bool MasmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 6388 MCAsmMacroParameter Parameter; 6389 MCAsmMacroArguments A; 6390 if (check(parseIdentifier(Parameter.Name), 6391 "expected identifier in '.irp' directive") || 6392 parseToken(AsmToken::Comma, "expected comma in '.irp' directive") || 6393 parseMacroArguments(nullptr, A) || 6394 parseToken(AsmToken::EndOfStatement, "expected End of Statement")) 6395 return true; 6396 6397 // Lex the irp definition. 6398 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6399 if (!M) 6400 return true; 6401 6402 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6403 // to hold the macro body with substitutions. 6404 SmallString<256> Buf; 6405 raw_svector_ostream OS(Buf); 6406 6407 for (const MCAsmMacroArgument &Arg : A) { 6408 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 6409 // This is undocumented, but GAS seems to support it. 6410 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 6411 return true; 6412 } 6413 6414 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6415 6416 return false; 6417 } 6418 6419 /// parseDirectiveIrpc 6420 /// ::= .irpc symbol,values 6421 bool MasmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 6422 MCAsmMacroParameter Parameter; 6423 MCAsmMacroArguments A; 6424 6425 if (check(parseIdentifier(Parameter.Name), 6426 "expected identifier in '.irpc' directive") || 6427 parseToken(AsmToken::Comma, "expected comma in '.irpc' directive") || 6428 parseMacroArguments(nullptr, A)) 6429 return true; 6430 6431 if (A.size() != 1 || A.front().size() != 1) 6432 return TokError("unexpected token in '.irpc' directive"); 6433 6434 // Eat the end of statement. 6435 if (parseToken(AsmToken::EndOfStatement, "expected end of statement")) 6436 return true; 6437 6438 // Lex the irpc definition. 6439 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 6440 if (!M) 6441 return true; 6442 6443 // Macro instantiation is lexical, unfortunately. We construct a new buffer 6444 // to hold the macro body with substitutions. 6445 SmallString<256> Buf; 6446 raw_svector_ostream OS(Buf); 6447 6448 StringRef Values = A.front().front().getString(); 6449 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 6450 MCAsmMacroArgument Arg; 6451 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 6452 6453 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 6454 // This is undocumented, but GAS seems to support it. 6455 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 6456 return true; 6457 } 6458 6459 instantiateMacroLikeBody(M, DirectiveLoc, OS); 6460 6461 return false; 6462 } 6463 6464 bool MasmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 6465 if (ActiveMacros.empty()) 6466 return TokError("unmatched '.endr' directive"); 6467 6468 // The only .repl that should get here are the ones created by 6469 // instantiateMacroLikeBody. 6470 assert(getLexer().is(AsmToken::EndOfStatement)); 6471 6472 handleMacroExit(); 6473 return false; 6474 } 6475 6476 bool MasmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 6477 size_t Len) { 6478 const MCExpr *Value; 6479 SMLoc ExprLoc = getLexer().getLoc(); 6480 if (parseExpression(Value)) 6481 return true; 6482 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 6483 if (!MCE) 6484 return Error(ExprLoc, "unexpected expression in _emit"); 6485 uint64_t IntValue = MCE->getValue(); 6486 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 6487 return Error(ExprLoc, "literal value out of range for directive"); 6488 6489 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 6490 return false; 6491 } 6492 6493 bool MasmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 6494 const MCExpr *Value; 6495 SMLoc ExprLoc = getLexer().getLoc(); 6496 if (parseExpression(Value)) 6497 return true; 6498 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 6499 if (!MCE) 6500 return Error(ExprLoc, "unexpected expression in align"); 6501 uint64_t IntValue = MCE->getValue(); 6502 if (!isPowerOf2_64(IntValue)) 6503 return Error(ExprLoc, "literal value not a power of two greater then zero"); 6504 6505 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 6506 return false; 6507 } 6508 6509 bool MasmParser::parseDirectiveEcho() { 6510 StringRef Message = parseStringToEndOfStatement(); 6511 Lex(); // eat end of statement 6512 llvm::outs() << Message << '\n'; 6513 return false; 6514 } 6515 6516 // We are comparing pointers, but the pointers are relative to a single string. 6517 // Thus, this should always be deterministic. 6518 static int rewritesSort(const AsmRewrite *AsmRewriteA, 6519 const AsmRewrite *AsmRewriteB) { 6520 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 6521 return -1; 6522 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 6523 return 1; 6524 6525 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 6526 // rewrite to the same location. Make sure the SizeDirective rewrite is 6527 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 6528 // ensures the sort algorithm is stable. 6529 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 6530 AsmRewritePrecedence[AsmRewriteB->Kind]) 6531 return -1; 6532 6533 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 6534 AsmRewritePrecedence[AsmRewriteB->Kind]) 6535 return 1; 6536 llvm_unreachable("Unstable rewrite sort."); 6537 } 6538 6539 bool MasmParser::lookUpField(StringRef Name, StringRef &Type, 6540 unsigned &Offset) const { 6541 const std::pair<StringRef, StringRef> BaseMember = Name.split('.'); 6542 const StringRef Base = BaseMember.first, Member = BaseMember.second; 6543 return lookUpField(Base, Member, Type, Offset); 6544 } 6545 6546 bool MasmParser::lookUpField(StringRef Base, StringRef Member, StringRef &Type, 6547 unsigned &Offset) const { 6548 if (Base.empty()) 6549 return true; 6550 6551 unsigned BaseOffset = 0; 6552 if (Base.contains('.') && !lookUpField(Base, Type, BaseOffset)) 6553 Base = Type; 6554 6555 auto TypeIt = KnownType.find(Base); 6556 if (TypeIt != KnownType.end()) 6557 return lookUpField(*TypeIt->second, Member, Type, Offset); 6558 6559 auto StructIt = Structs.find(Base.lower()); 6560 if (StructIt != Structs.end()) 6561 return lookUpField(StructIt->second, Member, Type, Offset); 6562 6563 return true; 6564 } 6565 6566 bool MasmParser::lookUpField(const StructInfo &Structure, StringRef Member, 6567 StringRef &Type, unsigned &Offset) const { 6568 if (Member.empty()) { 6569 Type = Structure.Name; 6570 return false; 6571 } 6572 6573 std::pair<StringRef, StringRef> Split = Member.split('.'); 6574 const StringRef FieldName = Split.first, FieldMember = Split.second; 6575 6576 auto StructIt = Structs.find(FieldName.lower()); 6577 if (StructIt != Structs.end()) 6578 return lookUpField(StructIt->second, FieldMember, Type, Offset); 6579 6580 auto FieldIt = Structure.FieldsByName.find(FieldName.lower()); 6581 if (FieldIt == Structure.FieldsByName.end()) 6582 return true; 6583 6584 const FieldInfo &Field = Structure.Fields[FieldIt->second]; 6585 if (FieldMember.empty()) { 6586 Offset += Field.Offset; 6587 if (Field.Contents.FT == FT_STRUCT) 6588 Type = Field.Contents.StructInfo.Structure.Name; 6589 return false; 6590 } 6591 6592 if (Field.Contents.FT != FT_STRUCT) 6593 return true; 6594 const StructFieldInfo &StructInfo = Field.Contents.StructInfo; 6595 6596 bool Result = lookUpField(StructInfo.Structure, FieldMember, Type, Offset); 6597 if (Result) 6598 return true; 6599 6600 Offset += Field.Offset; 6601 return false; 6602 } 6603 6604 bool MasmParser::parseMSInlineAsm( 6605 void *AsmLoc, std::string &AsmString, unsigned &NumOutputs, 6606 unsigned &NumInputs, SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 6607 SmallVectorImpl<std::string> &Constraints, 6608 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 6609 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 6610 SmallVector<void *, 4> InputDecls; 6611 SmallVector<void *, 4> OutputDecls; 6612 SmallVector<bool, 4> InputDeclsAddressOf; 6613 SmallVector<bool, 4> OutputDeclsAddressOf; 6614 SmallVector<std::string, 4> InputConstraints; 6615 SmallVector<std::string, 4> OutputConstraints; 6616 SmallVector<unsigned, 4> ClobberRegs; 6617 6618 SmallVector<AsmRewrite, 4> AsmStrRewrites; 6619 6620 // Prime the lexer. 6621 Lex(); 6622 6623 // While we have input, parse each statement. 6624 unsigned InputIdx = 0; 6625 unsigned OutputIdx = 0; 6626 while (getLexer().isNot(AsmToken::Eof)) { 6627 // Parse curly braces marking block start/end. 6628 if (parseCurlyBlockScope(AsmStrRewrites)) 6629 continue; 6630 6631 ParseStatementInfo Info(&AsmStrRewrites); 6632 bool StatementErr = parseStatement(Info, &SI); 6633 6634 if (StatementErr || Info.ParseError) { 6635 // Emit pending errors if any exist. 6636 printPendingErrors(); 6637 return true; 6638 } 6639 6640 // No pending error should exist here. 6641 assert(!hasPendingError() && "unexpected error from parseStatement"); 6642 6643 if (Info.Opcode == ~0U) 6644 continue; 6645 6646 const MCInstrDesc &Desc = MII->get(Info.Opcode); 6647 6648 // Build the list of clobbers, outputs and inputs. 6649 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 6650 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 6651 6652 // Register operand. 6653 if (Operand.isReg() && !Operand.needAddressOf() && 6654 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 6655 unsigned NumDefs = Desc.getNumDefs(); 6656 // Clobber. 6657 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 6658 ClobberRegs.push_back(Operand.getReg()); 6659 continue; 6660 } 6661 6662 // Expr/Input or Output. 6663 StringRef SymName = Operand.getSymName(); 6664 if (SymName.empty()) 6665 continue; 6666 6667 void *OpDecl = Operand.getOpDecl(); 6668 if (!OpDecl) 6669 continue; 6670 6671 StringRef Constraint = Operand.getConstraint(); 6672 if (Operand.isImm()) { 6673 // Offset as immediate. 6674 if (Operand.isOffsetOfLocal()) 6675 Constraint = "r"; 6676 else 6677 Constraint = "i"; 6678 } 6679 6680 bool isOutput = (i == 1) && Desc.mayStore(); 6681 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 6682 if (isOutput) { 6683 ++InputIdx; 6684 OutputDecls.push_back(OpDecl); 6685 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 6686 OutputConstraints.push_back(("=" + Constraint).str()); 6687 AsmStrRewrites.emplace_back(AOK_Output, Start, SymName.size()); 6688 } else { 6689 InputDecls.push_back(OpDecl); 6690 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 6691 InputConstraints.push_back(Constraint.str()); 6692 if (Desc.OpInfo[i - 1].isBranchTarget()) 6693 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size()); 6694 else 6695 AsmStrRewrites.emplace_back(AOK_Input, Start, SymName.size()); 6696 } 6697 } 6698 6699 // Consider implicit defs to be clobbers. Think of cpuid and push. 6700 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 6701 Desc.getNumImplicitDefs()); 6702 ClobberRegs.insert(ClobberRegs.end(), ImpDefs.begin(), ImpDefs.end()); 6703 } 6704 6705 // Set the number of Outputs and Inputs. 6706 NumOutputs = OutputDecls.size(); 6707 NumInputs = InputDecls.size(); 6708 6709 // Set the unique clobbers. 6710 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 6711 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 6712 ClobberRegs.end()); 6713 Clobbers.assign(ClobberRegs.size(), std::string()); 6714 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 6715 raw_string_ostream OS(Clobbers[I]); 6716 IP->printRegName(OS, ClobberRegs[I]); 6717 } 6718 6719 // Merge the various outputs and inputs. Output are expected first. 6720 if (NumOutputs || NumInputs) { 6721 unsigned NumExprs = NumOutputs + NumInputs; 6722 OpDecls.resize(NumExprs); 6723 Constraints.resize(NumExprs); 6724 for (unsigned i = 0; i < NumOutputs; ++i) { 6725 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 6726 Constraints[i] = OutputConstraints[i]; 6727 } 6728 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 6729 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 6730 Constraints[j] = InputConstraints[i]; 6731 } 6732 } 6733 6734 // Build the IR assembly string. 6735 std::string AsmStringIR; 6736 raw_string_ostream OS(AsmStringIR); 6737 StringRef ASMString = 6738 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 6739 const char *AsmStart = ASMString.begin(); 6740 const char *AsmEnd = ASMString.end(); 6741 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 6742 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) { 6743 const AsmRewrite &AR = *it; 6744 // Check if this has already been covered by another rewrite... 6745 if (AR.Done) 6746 continue; 6747 AsmRewriteKind Kind = AR.Kind; 6748 6749 const char *Loc = AR.Loc.getPointer(); 6750 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 6751 6752 // Emit everything up to the immediate/expression. 6753 if (unsigned Len = Loc - AsmStart) 6754 OS << StringRef(AsmStart, Len); 6755 6756 // Skip the original expression. 6757 if (Kind == AOK_Skip) { 6758 AsmStart = Loc + AR.Len; 6759 continue; 6760 } 6761 6762 unsigned AdditionalSkip = 0; 6763 // Rewrite expressions in $N notation. 6764 switch (Kind) { 6765 default: 6766 break; 6767 case AOK_IntelExpr: 6768 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 6769 if (AR.IntelExp.NeedBracs) 6770 OS << "["; 6771 if (AR.IntelExp.hasBaseReg()) 6772 OS << AR.IntelExp.BaseReg; 6773 if (AR.IntelExp.hasIndexReg()) 6774 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 6775 << AR.IntelExp.IndexReg; 6776 if (AR.IntelExp.Scale > 1) 6777 OS << " * $$" << AR.IntelExp.Scale; 6778 if (AR.IntelExp.hasOffset()) { 6779 if (AR.IntelExp.hasRegs()) 6780 OS << " + "; 6781 // Fuse this rewrite with a rewrite of the offset name, if present. 6782 StringRef OffsetName = AR.IntelExp.OffsetName; 6783 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data()); 6784 size_t OffsetLen = OffsetName.size(); 6785 auto rewrite_it = std::find_if( 6786 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) { 6787 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && 6788 (FusingAR.Kind == AOK_Input || 6789 FusingAR.Kind == AOK_CallInput); 6790 }); 6791 if (rewrite_it == AsmStrRewrites.end()) { 6792 OS << "offset " << OffsetName; 6793 } else if (rewrite_it->Kind == AOK_CallInput) { 6794 OS << "${" << InputIdx++ << ":P}"; 6795 rewrite_it->Done = true; 6796 } else { 6797 OS << '$' << InputIdx++; 6798 rewrite_it->Done = true; 6799 } 6800 } 6801 if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) 6802 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm; 6803 if (AR.IntelExp.NeedBracs) 6804 OS << "]"; 6805 break; 6806 case AOK_Label: 6807 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 6808 break; 6809 case AOK_Input: 6810 OS << '$' << InputIdx++; 6811 break; 6812 case AOK_CallInput: 6813 OS << "${" << InputIdx++ << ":P}"; 6814 break; 6815 case AOK_Output: 6816 OS << '$' << OutputIdx++; 6817 break; 6818 case AOK_SizeDirective: 6819 switch (AR.Val) { 6820 default: break; 6821 case 8: OS << "byte ptr "; break; 6822 case 16: OS << "word ptr "; break; 6823 case 32: OS << "dword ptr "; break; 6824 case 64: OS << "qword ptr "; break; 6825 case 80: OS << "xword ptr "; break; 6826 case 128: OS << "xmmword ptr "; break; 6827 case 256: OS << "ymmword ptr "; break; 6828 } 6829 break; 6830 case AOK_Emit: 6831 OS << ".byte"; 6832 break; 6833 case AOK_Align: { 6834 // MS alignment directives are measured in bytes. If the native assembler 6835 // measures alignment in bytes, we can pass it straight through. 6836 OS << ".align"; 6837 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 6838 break; 6839 6840 // Alignment is in log2 form, so print that instead and skip the original 6841 // immediate. 6842 unsigned Val = AR.Val; 6843 OS << ' ' << Val; 6844 assert(Val < 10 && "Expected alignment less then 2^10."); 6845 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 6846 break; 6847 } 6848 case AOK_EVEN: 6849 OS << ".even"; 6850 break; 6851 case AOK_EndOfStatement: 6852 OS << "\n\t"; 6853 break; 6854 } 6855 6856 // Skip the original expression. 6857 AsmStart = Loc + AR.Len + AdditionalSkip; 6858 } 6859 6860 // Emit the remainder of the asm string. 6861 if (AsmStart != AsmEnd) 6862 OS << StringRef(AsmStart, AsmEnd - AsmStart); 6863 6864 AsmString = OS.str(); 6865 return false; 6866 } 6867 6868 /// Create an MCAsmParser instance. 6869 MCAsmParser *llvm::createMCMasmParser(SourceMgr &SM, MCContext &C, 6870 MCStreamer &Out, const MCAsmInfo &MAI, 6871 unsigned CB) { 6872 return new MasmParser(SM, C, Out, MAI, CB); 6873 } 6874