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 a parser for assembly files similar to gas syntax. 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/SmallSet.h" 19 #include "llvm/ADT/SmallString.h" 20 #include "llvm/ADT/SmallVector.h" 21 #include "llvm/ADT/StringExtras.h" 22 #include "llvm/ADT/StringMap.h" 23 #include "llvm/ADT/StringRef.h" 24 #include "llvm/ADT/Twine.h" 25 #include "llvm/BinaryFormat/Dwarf.h" 26 #include "llvm/DebugInfo/CodeView/SymbolRecord.h" 27 #include "llvm/MC/MCAsmInfo.h" 28 #include "llvm/MC/MCCodeView.h" 29 #include "llvm/MC/MCContext.h" 30 #include "llvm/MC/MCDirectives.h" 31 #include "llvm/MC/MCDwarf.h" 32 #include "llvm/MC/MCExpr.h" 33 #include "llvm/MC/MCInstPrinter.h" 34 #include "llvm/MC/MCInstrDesc.h" 35 #include "llvm/MC/MCInstrInfo.h" 36 #include "llvm/MC/MCObjectFileInfo.h" 37 #include "llvm/MC/MCParser/AsmCond.h" 38 #include "llvm/MC/MCParser/AsmLexer.h" 39 #include "llvm/MC/MCParser/MCAsmLexer.h" 40 #include "llvm/MC/MCParser/MCAsmParser.h" 41 #include "llvm/MC/MCParser/MCAsmParserExtension.h" 42 #include "llvm/MC/MCParser/MCAsmParserUtils.h" 43 #include "llvm/MC/MCParser/MCParsedAsmOperand.h" 44 #include "llvm/MC/MCParser/MCTargetAsmParser.h" 45 #include "llvm/MC/MCRegisterInfo.h" 46 #include "llvm/MC/MCSection.h" 47 #include "llvm/MC/MCStreamer.h" 48 #include "llvm/MC/MCSymbol.h" 49 #include "llvm/MC/MCTargetOptions.h" 50 #include "llvm/MC/MCValue.h" 51 #include "llvm/Support/Casting.h" 52 #include "llvm/Support/CommandLine.h" 53 #include "llvm/Support/ErrorHandling.h" 54 #include "llvm/Support/MD5.h" 55 #include "llvm/Support/MathExtras.h" 56 #include "llvm/Support/MemoryBuffer.h" 57 #include "llvm/Support/SMLoc.h" 58 #include "llvm/Support/SourceMgr.h" 59 #include "llvm/Support/raw_ostream.h" 60 #include <algorithm> 61 #include <cassert> 62 #include <cctype> 63 #include <climits> 64 #include <cstddef> 65 #include <cstdint> 66 #include <deque> 67 #include <memory> 68 #include <sstream> 69 #include <string> 70 #include <tuple> 71 #include <utility> 72 #include <vector> 73 74 using namespace llvm; 75 76 MCAsmParserSemaCallback::~MCAsmParserSemaCallback() = default; 77 78 extern cl::opt<unsigned> AsmMacroMaxNestingDepth; 79 80 namespace { 81 82 /// Helper types for tracking macro definitions. 83 typedef std::vector<AsmToken> MCAsmMacroArgument; 84 typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; 85 86 /// Helper class for storing information about an active macro 87 /// instantiation. 88 struct MacroInstantiation { 89 /// The location of the instantiation. 90 SMLoc InstantiationLoc; 91 92 /// The buffer where parsing should resume upon instantiation completion. 93 unsigned ExitBuffer; 94 95 /// The location where parsing should resume upon instantiation completion. 96 SMLoc ExitLoc; 97 98 /// The depth of TheCondStack at the start of the instantiation. 99 size_t CondStackDepth; 100 }; 101 102 struct ParseStatementInfo { 103 /// The parsed operands from the last parsed statement. 104 SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; 105 106 /// The opcode from the last parsed instruction. 107 unsigned Opcode = ~0U; 108 109 /// Was there an error parsing the inline assembly? 110 bool ParseError = false; 111 112 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr; 113 114 ParseStatementInfo() = delete; 115 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites) 116 : AsmRewrites(rewrites) {} 117 }; 118 119 /// The concrete assembly parser instance. 120 class AsmParser : public MCAsmParser { 121 private: 122 AsmLexer Lexer; 123 MCContext &Ctx; 124 MCStreamer &Out; 125 const MCAsmInfo &MAI; 126 SourceMgr &SrcMgr; 127 SourceMgr::DiagHandlerTy SavedDiagHandler; 128 void *SavedDiagContext; 129 std::unique_ptr<MCAsmParserExtension> PlatformParser; 130 SMLoc StartTokLoc; 131 132 /// This is the current buffer index we're lexing from as managed by the 133 /// SourceMgr object. 134 unsigned CurBuffer; 135 136 AsmCond TheCondState; 137 std::vector<AsmCond> TheCondStack; 138 139 /// maps directive names to handler methods in parser 140 /// extensions. Extensions register themselves in this map by calling 141 /// addDirectiveHandler. 142 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap; 143 144 /// Stack of active macro instantiations. 145 std::vector<MacroInstantiation*> ActiveMacros; 146 147 /// List of bodies of anonymous macros. 148 std::deque<MCAsmMacro> MacroLikeBodies; 149 150 /// Boolean tracking whether macro substitution is enabled. 151 unsigned MacrosEnabledFlag : 1; 152 153 /// Keeps track of how many .macro's have been instantiated. 154 unsigned NumOfMacroInstantiations; 155 156 /// The values from the last parsed cpp hash file line comment if any. 157 struct CppHashInfoTy { 158 StringRef Filename; 159 int64_t LineNumber; 160 SMLoc Loc; 161 unsigned Buf; 162 CppHashInfoTy() : LineNumber(0), Buf(0) {} 163 }; 164 CppHashInfoTy CppHashInfo; 165 166 /// The filename from the first cpp hash file line comment, if any. 167 StringRef FirstCppHashFilename; 168 169 /// List of forward directional labels for diagnosis at the end. 170 SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; 171 172 SmallSet<StringRef, 2> LTODiscardSymbols; 173 174 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 175 unsigned AssemblerDialect = ~0U; 176 177 /// is Darwin compatibility enabled? 178 bool IsDarwin = false; 179 180 /// Are we parsing ms-style inline assembly? 181 bool ParsingMSInlineAsm = false; 182 183 /// Did we already inform the user about inconsistent MD5 usage? 184 bool ReportedInconsistentMD5 = false; 185 186 // Is alt macro mode enabled. 187 bool AltMacroMode = false; 188 189 protected: 190 virtual bool parseStatement(ParseStatementInfo &Info, 191 MCAsmParserSemaCallback *SI); 192 193 /// This routine uses the target specific ParseInstruction function to 194 /// parse an instruction into Operands, and then call the target specific 195 /// MatchAndEmit function to match and emit the instruction. 196 bool parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info, 197 StringRef IDVal, AsmToken ID, 198 SMLoc IDLoc); 199 200 /// Should we emit DWARF describing this assembler source? (Returns false if 201 /// the source has .file directives, which means we don't want to generate 202 /// info describing the assembler source itself.) 203 bool enabledGenDwarfForAssembly(); 204 205 public: 206 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 207 const MCAsmInfo &MAI, unsigned CB); 208 AsmParser(const AsmParser &) = delete; 209 AsmParser &operator=(const AsmParser &) = delete; 210 ~AsmParser() override; 211 212 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override; 213 214 void addDirectiveHandler(StringRef Directive, 215 ExtensionDirectiveHandler Handler) override { 216 ExtensionDirectiveMap[Directive] = Handler; 217 } 218 219 void addAliasForDirective(StringRef Directive, StringRef Alias) override { 220 DirectiveKindMap[Directive.lower()] = DirectiveKindMap[Alias.lower()]; 221 } 222 223 /// @name MCAsmParser Interface 224 /// { 225 226 SourceMgr &getSourceManager() override { return SrcMgr; } 227 MCAsmLexer &getLexer() override { return Lexer; } 228 MCContext &getContext() override { return Ctx; } 229 MCStreamer &getStreamer() override { return Out; } 230 231 CodeViewContext &getCVContext() { return Ctx.getCVContext(); } 232 233 unsigned getAssemblerDialect() override { 234 if (AssemblerDialect == ~0U) 235 return MAI.getAssemblerDialect(); 236 else 237 return AssemblerDialect; 238 } 239 void setAssemblerDialect(unsigned i) override { 240 AssemblerDialect = i; 241 } 242 243 void Note(SMLoc L, const Twine &Msg, SMRange Range = None) override; 244 bool Warning(SMLoc L, const Twine &Msg, SMRange Range = None) override; 245 bool printError(SMLoc L, const Twine &Msg, SMRange Range = None) override; 246 247 const AsmToken &Lex() override; 248 249 void setParsingMSInlineAsm(bool V) override { 250 ParsingMSInlineAsm = V; 251 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and 252 // hex integer literals. 253 Lexer.setLexMasmIntegers(V); 254 } 255 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; } 256 257 bool discardLTOSymbol(StringRef Name) const override { 258 return LTODiscardSymbols.contains(Name); 259 } 260 261 bool parseMSInlineAsm(std::string &AsmString, unsigned &NumOutputs, 262 unsigned &NumInputs, 263 SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 264 SmallVectorImpl<std::string> &Constraints, 265 SmallVectorImpl<std::string> &Clobbers, 266 const MCInstrInfo *MII, const MCInstPrinter *IP, 267 MCAsmParserSemaCallback &SI) override; 268 269 bool parseExpression(const MCExpr *&Res); 270 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 271 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc, 272 AsmTypeInfo *TypeInfo) override; 273 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override; 274 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 275 SMLoc &EndLoc) override; 276 bool parseAbsoluteExpression(int64_t &Res) override; 277 278 /// Parse a floating point expression using the float \p Semantics 279 /// and set \p Res to the value. 280 bool parseRealValue(const fltSemantics &Semantics, APInt &Res); 281 282 /// Parse an identifier or string (as a quoted identifier) 283 /// and set \p Res to the identifier contents. 284 bool parseIdentifier(StringRef &Res) override; 285 void eatToEndOfStatement() override; 286 287 bool checkForValidSection() override; 288 289 /// } 290 291 private: 292 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); 293 bool parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo = true); 294 295 void checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, StringRef Body, 296 ArrayRef<MCAsmMacroParameter> Parameters); 297 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 298 ArrayRef<MCAsmMacroParameter> Parameters, 299 ArrayRef<MCAsmMacroArgument> A, bool EnableAtPseudoVariable, 300 SMLoc L); 301 302 /// Are macros enabled in the parser? 303 bool areMacrosEnabled() {return MacrosEnabledFlag;} 304 305 /// Control a flag in the parser that enables or disables macros. 306 void setMacrosEnabled(bool Flag) {MacrosEnabledFlag = Flag;} 307 308 /// Are we inside a macro instantiation? 309 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();} 310 311 /// Handle entry to macro instantiation. 312 /// 313 /// \param M The macro. 314 /// \param NameLoc Instantiation location. 315 bool handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc); 316 317 /// Handle exit from macro instantiation. 318 void handleMacroExit(); 319 320 /// Extract AsmTokens for a macro argument. 321 bool parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg); 322 323 /// Parse all macro arguments for a given macro. 324 bool parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A); 325 326 void printMacroInstantiations(); 327 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 328 SMRange Range = None) const { 329 ArrayRef<SMRange> Ranges(Range); 330 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 331 } 332 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 333 334 /// Enter the specified file. This returns true on failure. 335 bool enterIncludeFile(const std::string &Filename); 336 337 /// Process the specified file for the .incbin directive. 338 /// This returns true on failure. 339 bool processIncbinFile(const std::string &Filename, int64_t Skip = 0, 340 const MCExpr *Count = nullptr, SMLoc Loc = SMLoc()); 341 342 /// Reset the current lexer position to that given by \p Loc. The 343 /// current token is not set; clients should ensure Lex() is called 344 /// subsequently. 345 /// 346 /// \param InBuffer If not 0, should be the known buffer id that contains the 347 /// location. 348 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0); 349 350 /// Parse up to the end of statement and a return the contents from the 351 /// current token until the end of the statement; the current token on exit 352 /// will be either the EndOfStatement or EOF. 353 StringRef parseStringToEndOfStatement() override; 354 355 /// Parse until the end of a statement or a comma is encountered, 356 /// return the contents from the current token up to the end or comma. 357 StringRef parseStringToComma(); 358 359 enum class AssignmentKind { 360 Set, 361 Equiv, 362 Equal, 363 LTOSetConditional, 364 }; 365 366 bool parseAssignment(StringRef Name, AssignmentKind Kind); 367 368 unsigned getBinOpPrecedence(AsmToken::TokenKind K, 369 MCBinaryExpr::Opcode &Kind); 370 371 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 372 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 373 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 374 375 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 376 377 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName); 378 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName); 379 380 // Generic (target and platform independent) directive parsing. 381 enum DirectiveKind { 382 DK_NO_DIRECTIVE, // Placeholder 383 DK_SET, 384 DK_EQU, 385 DK_EQUIV, 386 DK_ASCII, 387 DK_ASCIZ, 388 DK_STRING, 389 DK_BYTE, 390 DK_SHORT, 391 DK_RELOC, 392 DK_VALUE, 393 DK_2BYTE, 394 DK_LONG, 395 DK_INT, 396 DK_4BYTE, 397 DK_QUAD, 398 DK_8BYTE, 399 DK_OCTA, 400 DK_DC, 401 DK_DC_A, 402 DK_DC_B, 403 DK_DC_D, 404 DK_DC_L, 405 DK_DC_S, 406 DK_DC_W, 407 DK_DC_X, 408 DK_DCB, 409 DK_DCB_B, 410 DK_DCB_D, 411 DK_DCB_L, 412 DK_DCB_S, 413 DK_DCB_W, 414 DK_DCB_X, 415 DK_DS, 416 DK_DS_B, 417 DK_DS_D, 418 DK_DS_L, 419 DK_DS_P, 420 DK_DS_S, 421 DK_DS_W, 422 DK_DS_X, 423 DK_SINGLE, 424 DK_FLOAT, 425 DK_DOUBLE, 426 DK_ALIGN, 427 DK_ALIGN32, 428 DK_BALIGN, 429 DK_BALIGNW, 430 DK_BALIGNL, 431 DK_P2ALIGN, 432 DK_P2ALIGNW, 433 DK_P2ALIGNL, 434 DK_ORG, 435 DK_FILL, 436 DK_ENDR, 437 DK_BUNDLE_ALIGN_MODE, 438 DK_BUNDLE_LOCK, 439 DK_BUNDLE_UNLOCK, 440 DK_ZERO, 441 DK_EXTERN, 442 DK_GLOBL, 443 DK_GLOBAL, 444 DK_LAZY_REFERENCE, 445 DK_NO_DEAD_STRIP, 446 DK_SYMBOL_RESOLVER, 447 DK_PRIVATE_EXTERN, 448 DK_REFERENCE, 449 DK_WEAK_DEFINITION, 450 DK_WEAK_REFERENCE, 451 DK_WEAK_DEF_CAN_BE_HIDDEN, 452 DK_COLD, 453 DK_COMM, 454 DK_COMMON, 455 DK_LCOMM, 456 DK_ABORT, 457 DK_INCLUDE, 458 DK_INCBIN, 459 DK_CODE16, 460 DK_CODE16GCC, 461 DK_REPT, 462 DK_IRP, 463 DK_IRPC, 464 DK_IF, 465 DK_IFEQ, 466 DK_IFGE, 467 DK_IFGT, 468 DK_IFLE, 469 DK_IFLT, 470 DK_IFNE, 471 DK_IFB, 472 DK_IFNB, 473 DK_IFC, 474 DK_IFEQS, 475 DK_IFNC, 476 DK_IFNES, 477 DK_IFDEF, 478 DK_IFNDEF, 479 DK_IFNOTDEF, 480 DK_ELSEIF, 481 DK_ELSE, 482 DK_ENDIF, 483 DK_SPACE, 484 DK_SKIP, 485 DK_FILE, 486 DK_LINE, 487 DK_LOC, 488 DK_STABS, 489 DK_CV_FILE, 490 DK_CV_FUNC_ID, 491 DK_CV_INLINE_SITE_ID, 492 DK_CV_LOC, 493 DK_CV_LINETABLE, 494 DK_CV_INLINE_LINETABLE, 495 DK_CV_DEF_RANGE, 496 DK_CV_STRINGTABLE, 497 DK_CV_STRING, 498 DK_CV_FILECHECKSUMS, 499 DK_CV_FILECHECKSUM_OFFSET, 500 DK_CV_FPO_DATA, 501 DK_CFI_SECTIONS, 502 DK_CFI_STARTPROC, 503 DK_CFI_ENDPROC, 504 DK_CFI_DEF_CFA, 505 DK_CFI_DEF_CFA_OFFSET, 506 DK_CFI_ADJUST_CFA_OFFSET, 507 DK_CFI_DEF_CFA_REGISTER, 508 DK_CFI_LLVM_DEF_ASPACE_CFA, 509 DK_CFI_OFFSET, 510 DK_CFI_REL_OFFSET, 511 DK_CFI_PERSONALITY, 512 DK_CFI_LSDA, 513 DK_CFI_REMEMBER_STATE, 514 DK_CFI_RESTORE_STATE, 515 DK_CFI_SAME_VALUE, 516 DK_CFI_RESTORE, 517 DK_CFI_ESCAPE, 518 DK_CFI_RETURN_COLUMN, 519 DK_CFI_SIGNAL_FRAME, 520 DK_CFI_UNDEFINED, 521 DK_CFI_REGISTER, 522 DK_CFI_WINDOW_SAVE, 523 DK_CFI_B_KEY_FRAME, 524 DK_MACROS_ON, 525 DK_MACROS_OFF, 526 DK_ALTMACRO, 527 DK_NOALTMACRO, 528 DK_MACRO, 529 DK_EXITM, 530 DK_ENDM, 531 DK_ENDMACRO, 532 DK_PURGEM, 533 DK_SLEB128, 534 DK_ULEB128, 535 DK_ERR, 536 DK_ERROR, 537 DK_WARNING, 538 DK_PRINT, 539 DK_ADDRSIG, 540 DK_ADDRSIG_SYM, 541 DK_PSEUDO_PROBE, 542 DK_LTO_DISCARD, 543 DK_LTO_SET_CONDITIONAL, 544 DK_END 545 }; 546 547 /// Maps directive name --> DirectiveKind enum, for 548 /// directives parsed by this class. 549 StringMap<DirectiveKind> DirectiveKindMap; 550 551 // Codeview def_range type parsing. 552 enum CVDefRangeType { 553 CVDR_DEFRANGE = 0, // Placeholder 554 CVDR_DEFRANGE_REGISTER, 555 CVDR_DEFRANGE_FRAMEPOINTER_REL, 556 CVDR_DEFRANGE_SUBFIELD_REGISTER, 557 CVDR_DEFRANGE_REGISTER_REL 558 }; 559 560 /// Maps Codeview def_range types --> CVDefRangeType enum, for 561 /// Codeview def_range types parsed by this class. 562 StringMap<CVDefRangeType> CVDefRangeTypeMap; 563 564 // ".ascii", ".asciz", ".string" 565 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 566 bool parseDirectiveReloc(SMLoc DirectiveLoc); // ".reloc" 567 bool parseDirectiveValue(StringRef IDVal, 568 unsigned Size); // ".byte", ".long", ... 569 bool parseDirectiveOctaValue(StringRef IDVal); // ".octa", ... 570 bool parseDirectiveRealValue(StringRef IDVal, 571 const fltSemantics &); // ".single", ... 572 bool parseDirectiveFill(); // ".fill" 573 bool parseDirectiveZero(); // ".zero" 574 // ".set", ".equ", ".equiv", ".lto_set_conditional" 575 bool parseDirectiveSet(StringRef IDVal, AssignmentKind Kind); 576 bool parseDirectiveOrg(); // ".org" 577 // ".align{,32}", ".p2align{,w,l}" 578 bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize); 579 580 // ".file", ".line", ".loc", ".stabs" 581 bool parseDirectiveFile(SMLoc DirectiveLoc); 582 bool parseDirectiveLine(); 583 bool parseDirectiveLoc(); 584 bool parseDirectiveStabs(); 585 586 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable", 587 // ".cv_inline_linetable", ".cv_def_range", ".cv_string" 588 bool parseDirectiveCVFile(); 589 bool parseDirectiveCVFuncId(); 590 bool parseDirectiveCVInlineSiteId(); 591 bool parseDirectiveCVLoc(); 592 bool parseDirectiveCVLinetable(); 593 bool parseDirectiveCVInlineLinetable(); 594 bool parseDirectiveCVDefRange(); 595 bool parseDirectiveCVString(); 596 bool parseDirectiveCVStringTable(); 597 bool parseDirectiveCVFileChecksums(); 598 bool parseDirectiveCVFileChecksumOffset(); 599 bool parseDirectiveCVFPOData(); 600 601 // .cfi directives 602 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc); 603 bool parseDirectiveCFIWindowSave(); 604 bool parseDirectiveCFISections(); 605 bool parseDirectiveCFIStartProc(); 606 bool parseDirectiveCFIEndProc(); 607 bool parseDirectiveCFIDefCfaOffset(); 608 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc); 609 bool parseDirectiveCFIAdjustCfaOffset(); 610 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc); 611 bool parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc); 612 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc); 613 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc); 614 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality); 615 bool parseDirectiveCFIRememberState(); 616 bool parseDirectiveCFIRestoreState(); 617 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc); 618 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc); 619 bool parseDirectiveCFIEscape(); 620 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc); 621 bool parseDirectiveCFISignalFrame(); 622 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc); 623 624 // macro directives 625 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc); 626 bool parseDirectiveExitMacro(StringRef Directive); 627 bool parseDirectiveEndMacro(StringRef Directive); 628 bool parseDirectiveMacro(SMLoc DirectiveLoc); 629 bool parseDirectiveMacrosOnOff(StringRef Directive); 630 // alternate macro mode directives 631 bool parseDirectiveAltmacro(StringRef Directive); 632 // ".bundle_align_mode" 633 bool parseDirectiveBundleAlignMode(); 634 // ".bundle_lock" 635 bool parseDirectiveBundleLock(); 636 // ".bundle_unlock" 637 bool parseDirectiveBundleUnlock(); 638 639 // ".space", ".skip" 640 bool parseDirectiveSpace(StringRef IDVal); 641 642 // ".dcb" 643 bool parseDirectiveDCB(StringRef IDVal, unsigned Size); 644 bool parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &); 645 // ".ds" 646 bool parseDirectiveDS(StringRef IDVal, unsigned Size); 647 648 // .sleb128 (Signed=true) and .uleb128 (Signed=false) 649 bool parseDirectiveLEB128(bool Signed); 650 651 /// Parse a directive like ".globl" which 652 /// accepts a single symbol (which should be a label or an external). 653 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr); 654 655 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 656 657 bool parseDirectiveAbort(); // ".abort" 658 bool parseDirectiveInclude(); // ".include" 659 bool parseDirectiveIncbin(); // ".incbin" 660 661 // ".if", ".ifeq", ".ifge", ".ifgt" , ".ifle", ".iflt" or ".ifne" 662 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind); 663 // ".ifb" or ".ifnb", depending on ExpectBlank. 664 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 665 // ".ifc" or ".ifnc", depending on ExpectEqual. 666 bool parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 667 // ".ifeqs" or ".ifnes", depending on ExpectEqual. 668 bool parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual); 669 // ".ifdef" or ".ifndef", depending on expect_defined 670 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 671 bool parseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 672 bool parseDirectiveElse(SMLoc DirectiveLoc); // ".else" 673 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 674 bool parseEscapedString(std::string &Data) override; 675 bool parseAngleBracketString(std::string &Data) override; 676 677 const MCExpr *applyModifierToExpr(const MCExpr *E, 678 MCSymbolRefExpr::VariantKind Variant); 679 680 // Macro-like directives 681 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc); 682 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 683 raw_svector_ostream &OS); 684 bool parseDirectiveRept(SMLoc DirectiveLoc, StringRef Directive); 685 bool parseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 686 bool parseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc" 687 bool parseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 688 689 // "_emit" or "__emit" 690 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info, 691 size_t Len); 692 693 // "align" 694 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info); 695 696 // "end" 697 bool parseDirectiveEnd(SMLoc DirectiveLoc); 698 699 // ".err" or ".error" 700 bool parseDirectiveError(SMLoc DirectiveLoc, bool WithMessage); 701 702 // ".warning" 703 bool parseDirectiveWarning(SMLoc DirectiveLoc); 704 705 // .print <double-quotes-string> 706 bool parseDirectivePrint(SMLoc DirectiveLoc); 707 708 // .pseudoprobe 709 bool parseDirectivePseudoProbe(); 710 711 // ".lto_discard" 712 bool parseDirectiveLTODiscard(); 713 714 // Directives to support address-significance tables. 715 bool parseDirectiveAddrsig(); 716 bool parseDirectiveAddrsigSym(); 717 718 void initializeDirectiveKindMap(); 719 void initializeCVDefRangeTypeMap(); 720 }; 721 722 class HLASMAsmParser final : public AsmParser { 723 private: 724 MCAsmLexer &Lexer; 725 MCStreamer &Out; 726 727 void lexLeadingSpaces() { 728 while (Lexer.is(AsmToken::Space)) 729 Lexer.Lex(); 730 } 731 732 bool parseAsHLASMLabel(ParseStatementInfo &Info, MCAsmParserSemaCallback *SI); 733 bool parseAsMachineInstruction(ParseStatementInfo &Info, 734 MCAsmParserSemaCallback *SI); 735 736 public: 737 HLASMAsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 738 const MCAsmInfo &MAI, unsigned CB = 0) 739 : AsmParser(SM, Ctx, Out, MAI, CB), Lexer(getLexer()), Out(Out) { 740 Lexer.setSkipSpace(false); 741 Lexer.setAllowHashInIdentifier(true); 742 Lexer.setLexHLASMIntegers(true); 743 Lexer.setLexHLASMStrings(true); 744 } 745 746 ~HLASMAsmParser() { Lexer.setSkipSpace(true); } 747 748 bool parseStatement(ParseStatementInfo &Info, 749 MCAsmParserSemaCallback *SI) override; 750 }; 751 752 } // end anonymous namespace 753 754 namespace llvm { 755 756 extern MCAsmParserExtension *createDarwinAsmParser(); 757 extern MCAsmParserExtension *createELFAsmParser(); 758 extern MCAsmParserExtension *createCOFFAsmParser(); 759 extern MCAsmParserExtension *createGOFFAsmParser(); 760 extern MCAsmParserExtension *createXCOFFAsmParser(); 761 extern MCAsmParserExtension *createWasmAsmParser(); 762 763 } // end namespace llvm 764 765 enum { DEFAULT_ADDRSPACE = 0 }; 766 767 AsmParser::AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 768 const MCAsmInfo &MAI, unsigned CB = 0) 769 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM), 770 CurBuffer(CB ? CB : SM.getMainFileID()), MacrosEnabledFlag(true) { 771 HadError = false; 772 // Save the old handler. 773 SavedDiagHandler = SrcMgr.getDiagHandler(); 774 SavedDiagContext = SrcMgr.getDiagContext(); 775 // Set our own handler which calls the saved handler. 776 SrcMgr.setDiagHandler(DiagHandler, this); 777 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 778 // Make MCStreamer aware of the StartTokLoc for locations in diagnostics. 779 Out.setStartTokLocPtr(&StartTokLoc); 780 781 // Initialize the platform / file format parser. 782 switch (Ctx.getObjectFileType()) { 783 case MCContext::IsCOFF: 784 PlatformParser.reset(createCOFFAsmParser()); 785 break; 786 case MCContext::IsMachO: 787 PlatformParser.reset(createDarwinAsmParser()); 788 IsDarwin = true; 789 break; 790 case MCContext::IsELF: 791 PlatformParser.reset(createELFAsmParser()); 792 break; 793 case MCContext::IsGOFF: 794 PlatformParser.reset(createGOFFAsmParser()); 795 break; 796 case MCContext::IsWasm: 797 PlatformParser.reset(createWasmAsmParser()); 798 break; 799 case MCContext::IsXCOFF: 800 PlatformParser.reset(createXCOFFAsmParser()); 801 break; 802 } 803 804 PlatformParser->Initialize(*this); 805 initializeDirectiveKindMap(); 806 initializeCVDefRangeTypeMap(); 807 808 NumOfMacroInstantiations = 0; 809 } 810 811 AsmParser::~AsmParser() { 812 assert((HadError || ActiveMacros.empty()) && 813 "Unexpected active macro instantiation!"); 814 815 // Remove MCStreamer's reference to the parser SMLoc. 816 Out.setStartTokLocPtr(nullptr); 817 // Restore the saved diagnostics handler and context for use during 818 // finalization. 819 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext); 820 } 821 822 void AsmParser::printMacroInstantiations() { 823 // Print the active macro instantiation stack. 824 for (std::vector<MacroInstantiation *>::const_reverse_iterator 825 it = ActiveMacros.rbegin(), 826 ie = ActiveMacros.rend(); 827 it != ie; ++it) 828 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 829 "while in macro instantiation"); 830 } 831 832 void AsmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { 833 printPendingErrors(); 834 printMessage(L, SourceMgr::DK_Note, Msg, Range); 835 printMacroInstantiations(); 836 } 837 838 bool AsmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) { 839 if(getTargetParser().getTargetOptions().MCNoWarn) 840 return false; 841 if (getTargetParser().getTargetOptions().MCFatalWarnings) 842 return Error(L, Msg, Range); 843 printMessage(L, SourceMgr::DK_Warning, Msg, Range); 844 printMacroInstantiations(); 845 return false; 846 } 847 848 bool AsmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { 849 HadError = true; 850 printMessage(L, SourceMgr::DK_Error, Msg, Range); 851 printMacroInstantiations(); 852 return true; 853 } 854 855 bool AsmParser::enterIncludeFile(const std::string &Filename) { 856 std::string IncludedFile; 857 unsigned NewBuf = 858 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 859 if (!NewBuf) 860 return true; 861 862 CurBuffer = NewBuf; 863 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 864 return false; 865 } 866 867 /// Process the specified .incbin file by searching for it in the include paths 868 /// then just emitting the byte contents of the file to the streamer. This 869 /// returns true on failure. 870 bool AsmParser::processIncbinFile(const std::string &Filename, int64_t Skip, 871 const MCExpr *Count, SMLoc Loc) { 872 std::string IncludedFile; 873 unsigned NewBuf = 874 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 875 if (!NewBuf) 876 return true; 877 878 // Pick up the bytes from the file and emit them. 879 StringRef Bytes = SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(); 880 Bytes = Bytes.drop_front(Skip); 881 if (Count) { 882 int64_t Res; 883 if (!Count->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 884 return Error(Loc, "expected absolute expression"); 885 if (Res < 0) 886 return Warning(Loc, "negative count has no effect"); 887 Bytes = Bytes.take_front(Res); 888 } 889 getStreamer().emitBytes(Bytes); 890 return false; 891 } 892 893 void AsmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer) { 894 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); 895 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), 896 Loc.getPointer()); 897 } 898 899 const AsmToken &AsmParser::Lex() { 900 if (Lexer.getTok().is(AsmToken::Error)) 901 Error(Lexer.getErrLoc(), Lexer.getErr()); 902 903 // if it's a end of statement with a comment in it 904 if (getTok().is(AsmToken::EndOfStatement)) { 905 // if this is a line comment output it. 906 if (!getTok().getString().empty() && getTok().getString().front() != '\n' && 907 getTok().getString().front() != '\r' && MAI.preserveAsmComments()) 908 Out.addExplicitComment(Twine(getTok().getString())); 909 } 910 911 const AsmToken *tok = &Lexer.Lex(); 912 913 // Parse comments here to be deferred until end of next statement. 914 while (tok->is(AsmToken::Comment)) { 915 if (MAI.preserveAsmComments()) 916 Out.addExplicitComment(Twine(tok->getString())); 917 tok = &Lexer.Lex(); 918 } 919 920 if (tok->is(AsmToken::Eof)) { 921 // If this is the end of an included file, pop the parent file off the 922 // include stack. 923 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 924 if (ParentIncludeLoc != SMLoc()) { 925 jumpToLoc(ParentIncludeLoc); 926 return Lex(); 927 } 928 } 929 930 return *tok; 931 } 932 933 bool AsmParser::enabledGenDwarfForAssembly() { 934 // Check whether the user specified -g. 935 if (!getContext().getGenDwarfForAssembly()) 936 return false; 937 // If we haven't encountered any .file directives (which would imply that 938 // the assembler source was produced with debug info already) then emit one 939 // describing the assembler source file itself. 940 if (getContext().getGenDwarfFileNumber() == 0) { 941 // Use the first #line directive for this, if any. It's preprocessed, so 942 // there is no checksum, and of course no source directive. 943 if (!FirstCppHashFilename.empty()) 944 getContext().setMCLineTableRootFile(/*CUID=*/0, 945 getContext().getCompilationDir(), 946 FirstCppHashFilename, 947 /*Cksum=*/None, /*Source=*/None); 948 const MCDwarfFile &RootFile = 949 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); 950 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective( 951 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name, 952 RootFile.Checksum, RootFile.Source)); 953 } 954 return true; 955 } 956 957 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 958 LTODiscardSymbols.clear(); 959 960 // Create the initial section, if requested. 961 if (!NoInitialTextSection) 962 Out.initSections(false, getTargetParser().getSTI()); 963 964 // Prime the lexer. 965 Lex(); 966 967 HadError = false; 968 AsmCond StartingCondState = TheCondState; 969 SmallVector<AsmRewrite, 4> AsmStrRewrites; 970 971 // If we are generating dwarf for assembly source files save the initial text 972 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't 973 // emitting any actual debug info yet and haven't had a chance to parse any 974 // embedded .file directives.) 975 if (getContext().getGenDwarfForAssembly()) { 976 MCSection *Sec = getStreamer().getCurrentSectionOnly(); 977 if (!Sec->getBeginSymbol()) { 978 MCSymbol *SectionStartSym = getContext().createTempSymbol(); 979 getStreamer().emitLabel(SectionStartSym); 980 Sec->setBeginSymbol(SectionStartSym); 981 } 982 bool InsertResult = getContext().addGenDwarfSection(Sec); 983 assert(InsertResult && ".text section should not have debug info yet"); 984 (void)InsertResult; 985 } 986 987 getTargetParser().onBeginOfFile(); 988 989 // While we have input, parse each statement. 990 while (Lexer.isNot(AsmToken::Eof)) { 991 ParseStatementInfo Info(&AsmStrRewrites); 992 bool Parsed = parseStatement(Info, nullptr); 993 994 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error 995 // for printing ErrMsg via Lex() only if no (presumably better) parser error 996 // exists. 997 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) { 998 Lex(); 999 } 1000 1001 // parseStatement returned true so may need to emit an error. 1002 printPendingErrors(); 1003 1004 // Skipping to the next line if needed. 1005 if (Parsed && !getLexer().isAtStartOfStatement()) 1006 eatToEndOfStatement(); 1007 } 1008 1009 getTargetParser().onEndOfFile(); 1010 printPendingErrors(); 1011 1012 // All errors should have been emitted. 1013 assert(!hasPendingError() && "unexpected error from parseStatement"); 1014 1015 getTargetParser().flushPendingInstructions(getStreamer()); 1016 1017 if (TheCondState.TheCond != StartingCondState.TheCond || 1018 TheCondState.Ignore != StartingCondState.Ignore) 1019 printError(getTok().getLoc(), "unmatched .ifs or .elses"); 1020 // Check to see there are no empty DwarfFile slots. 1021 const auto &LineTables = getContext().getMCDwarfLineTables(); 1022 if (!LineTables.empty()) { 1023 unsigned Index = 0; 1024 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) { 1025 if (File.Name.empty() && Index != 0) 1026 printError(getTok().getLoc(), "unassigned file number: " + 1027 Twine(Index) + 1028 " for .file directives"); 1029 ++Index; 1030 } 1031 } 1032 1033 // Check to see that all assembler local symbols were actually defined. 1034 // Targets that don't do subsections via symbols may not want this, though, 1035 // so conservatively exclude them. Only do this if we're finalizing, though, 1036 // as otherwise we won't necessarilly have seen everything yet. 1037 if (!NoFinalize) { 1038 if (MAI.hasSubsectionsViaSymbols()) { 1039 for (const auto &TableEntry : getContext().getSymbols()) { 1040 MCSymbol *Sym = TableEntry.getValue(); 1041 // Variable symbols may not be marked as defined, so check those 1042 // explicitly. If we know it's a variable, we have a definition for 1043 // the purposes of this check. 1044 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 1045 // FIXME: We would really like to refer back to where the symbol was 1046 // first referenced for a source location. We need to add something 1047 // to track that. Currently, we just point to the end of the file. 1048 printError(getTok().getLoc(), "assembler local symbol '" + 1049 Sym->getName() + "' not defined"); 1050 } 1051 } 1052 1053 // Temporary symbols like the ones for directional jumps don't go in the 1054 // symbol table. They also need to be diagnosed in all (final) cases. 1055 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { 1056 if (std::get<2>(LocSym)->isUndefined()) { 1057 // Reset the state of any "# line file" directives we've seen to the 1058 // context as it was at the diagnostic site. 1059 CppHashInfo = std::get<1>(LocSym); 1060 printError(std::get<0>(LocSym), "directional label undefined"); 1061 } 1062 } 1063 } 1064 // Finalize the output stream if there are no errors and if the client wants 1065 // us to. 1066 if (!HadError && !NoFinalize) { 1067 if (auto *TS = Out.getTargetStreamer()) 1068 TS->emitConstantPools(); 1069 1070 Out.Finish(Lexer.getLoc()); 1071 } 1072 1073 return HadError || getContext().hadError(); 1074 } 1075 1076 bool AsmParser::checkForValidSection() { 1077 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) { 1078 Out.initSections(false, getTargetParser().getSTI()); 1079 return Error(getTok().getLoc(), 1080 "expected section directive before assembly directive"); 1081 } 1082 return false; 1083 } 1084 1085 /// Throw away the rest of the line for testing purposes. 1086 void AsmParser::eatToEndOfStatement() { 1087 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1088 Lexer.Lex(); 1089 1090 // Eat EOL. 1091 if (Lexer.is(AsmToken::EndOfStatement)) 1092 Lexer.Lex(); 1093 } 1094 1095 StringRef AsmParser::parseStringToEndOfStatement() { 1096 const char *Start = getTok().getLoc().getPointer(); 1097 1098 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof)) 1099 Lexer.Lex(); 1100 1101 const char *End = getTok().getLoc().getPointer(); 1102 return StringRef(Start, End - Start); 1103 } 1104 1105 StringRef AsmParser::parseStringToComma() { 1106 const char *Start = getTok().getLoc().getPointer(); 1107 1108 while (Lexer.isNot(AsmToken::EndOfStatement) && 1109 Lexer.isNot(AsmToken::Comma) && Lexer.isNot(AsmToken::Eof)) 1110 Lexer.Lex(); 1111 1112 const char *End = getTok().getLoc().getPointer(); 1113 return StringRef(Start, End - Start); 1114 } 1115 1116 /// Parse a paren expression and return it. 1117 /// NOTE: This assumes the leading '(' has already been consumed. 1118 /// 1119 /// parenexpr ::= expr) 1120 /// 1121 bool AsmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1122 if (parseExpression(Res)) 1123 return true; 1124 EndLoc = Lexer.getTok().getEndLoc(); 1125 return parseRParen(); 1126 } 1127 1128 /// Parse a bracket expression and return it. 1129 /// NOTE: This assumes the leading '[' has already been consumed. 1130 /// 1131 /// bracketexpr ::= expr] 1132 /// 1133 bool AsmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 1134 if (parseExpression(Res)) 1135 return true; 1136 EndLoc = getTok().getEndLoc(); 1137 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression")) 1138 return true; 1139 return false; 1140 } 1141 1142 /// Parse a primary expression and return it. 1143 /// primaryexpr ::= (parenexpr 1144 /// primaryexpr ::= symbol 1145 /// primaryexpr ::= number 1146 /// primaryexpr ::= '.' 1147 /// primaryexpr ::= ~,+,- primaryexpr 1148 bool AsmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc, 1149 AsmTypeInfo *TypeInfo) { 1150 SMLoc FirstTokenLoc = getLexer().getLoc(); 1151 AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); 1152 switch (FirstTokenKind) { 1153 default: 1154 return TokError("unknown token in expression"); 1155 // If we have an error assume that we've already handled it. 1156 case AsmToken::Error: 1157 return true; 1158 case AsmToken::Exclaim: 1159 Lex(); // Eat the operator. 1160 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1161 return true; 1162 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc); 1163 return false; 1164 case AsmToken::Dollar: 1165 case AsmToken::Star: 1166 case AsmToken::At: 1167 case AsmToken::String: 1168 case AsmToken::Identifier: { 1169 StringRef Identifier; 1170 if (parseIdentifier(Identifier)) { 1171 // We may have failed but '$'|'*' may be a valid token in context of 1172 // the current PC. 1173 if (getTok().is(AsmToken::Dollar) || getTok().is(AsmToken::Star)) { 1174 bool ShouldGenerateTempSymbol = false; 1175 if ((getTok().is(AsmToken::Dollar) && MAI.getDollarIsPC()) || 1176 (getTok().is(AsmToken::Star) && MAI.getStarIsPC())) 1177 ShouldGenerateTempSymbol = true; 1178 1179 if (!ShouldGenerateTempSymbol) 1180 return Error(FirstTokenLoc, "invalid token in expression"); 1181 1182 // Eat the '$'|'*' token. 1183 Lex(); 1184 // This is either a '$'|'*' reference, which references the current PC. 1185 // Emit a temporary label to the streamer and refer to it. 1186 MCSymbol *Sym = Ctx.createTempSymbol(); 1187 Out.emitLabel(Sym); 1188 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, 1189 getContext()); 1190 EndLoc = FirstTokenLoc; 1191 return false; 1192 } 1193 } 1194 // Parse symbol variant 1195 std::pair<StringRef, StringRef> Split; 1196 if (!MAI.useParensForSymbolVariant()) { 1197 if (FirstTokenKind == AsmToken::String) { 1198 if (Lexer.is(AsmToken::At)) { 1199 Lex(); // eat @ 1200 SMLoc AtLoc = getLexer().getLoc(); 1201 StringRef VName; 1202 if (parseIdentifier(VName)) 1203 return Error(AtLoc, "expected symbol variant after '@'"); 1204 1205 Split = std::make_pair(Identifier, VName); 1206 } 1207 } else { 1208 Split = Identifier.split('@'); 1209 } 1210 } else if (Lexer.is(AsmToken::LParen)) { 1211 Lex(); // eat '('. 1212 StringRef VName; 1213 parseIdentifier(VName); 1214 if (parseRParen()) 1215 return true; 1216 Split = std::make_pair(Identifier, VName); 1217 } 1218 1219 EndLoc = SMLoc::getFromPointer(Identifier.end()); 1220 1221 // This is a symbol reference. 1222 StringRef SymbolName = Identifier; 1223 if (SymbolName.empty()) 1224 return Error(getLexer().getLoc(), "expected a symbol reference"); 1225 1226 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1227 1228 // Lookup the symbol variant if used. 1229 if (!Split.second.empty()) { 1230 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1231 if (Variant != MCSymbolRefExpr::VK_Invalid) { 1232 SymbolName = Split.first; 1233 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { 1234 Variant = MCSymbolRefExpr::VK_None; 1235 } else { 1236 return Error(SMLoc::getFromPointer(Split.second.begin()), 1237 "invalid variant '" + Split.second + "'"); 1238 } 1239 } 1240 1241 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName); 1242 if (!Sym) 1243 Sym = getContext().getOrCreateSymbol( 1244 MAI.shouldEmitLabelsInUpperCase() ? SymbolName.upper() : SymbolName); 1245 1246 // If this is an absolute variable reference, substitute it now to preserve 1247 // semantics in the face of reassignment. 1248 if (Sym->isVariable()) { 1249 auto V = Sym->getVariableValue(/*SetUsed*/ false); 1250 bool DoInline = isa<MCConstantExpr>(V) && !Variant; 1251 if (auto TV = dyn_cast<MCTargetExpr>(V)) 1252 DoInline = TV->inlineAssignedExpr(); 1253 if (DoInline) { 1254 if (Variant) 1255 return Error(EndLoc, "unexpected modifier on variable reference"); 1256 Res = Sym->getVariableValue(/*SetUsed*/ false); 1257 return false; 1258 } 1259 } 1260 1261 // Otherwise create a symbol ref. 1262 Res = MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc); 1263 return false; 1264 } 1265 case AsmToken::BigNum: 1266 return TokError("literal value out of range for directive"); 1267 case AsmToken::Integer: { 1268 SMLoc Loc = getTok().getLoc(); 1269 int64_t IntVal = getTok().getIntVal(); 1270 Res = MCConstantExpr::create(IntVal, getContext()); 1271 EndLoc = Lexer.getTok().getEndLoc(); 1272 Lex(); // Eat token. 1273 // Look for 'b' or 'f' following an Integer as a directional label 1274 if (Lexer.getKind() == AsmToken::Identifier) { 1275 StringRef IDVal = getTok().getString(); 1276 // Lookup the symbol variant if used. 1277 std::pair<StringRef, StringRef> Split = IDVal.split('@'); 1278 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 1279 if (Split.first.size() != IDVal.size()) { 1280 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 1281 if (Variant == MCSymbolRefExpr::VK_Invalid) 1282 return TokError("invalid variant '" + Split.second + "'"); 1283 IDVal = Split.first; 1284 } 1285 if (IDVal == "f" || IDVal == "b") { 1286 MCSymbol *Sym = 1287 Ctx.getDirectionalLocalSymbol(IntVal, IDVal == "b"); 1288 Res = MCSymbolRefExpr::create(Sym, Variant, getContext()); 1289 if (IDVal == "b" && Sym->isUndefined()) 1290 return Error(Loc, "directional label undefined"); 1291 DirLabels.push_back(std::make_tuple(Loc, CppHashInfo, Sym)); 1292 EndLoc = Lexer.getTok().getEndLoc(); 1293 Lex(); // Eat identifier. 1294 } 1295 } 1296 return false; 1297 } 1298 case AsmToken::Real: { 1299 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); 1300 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 1301 Res = MCConstantExpr::create(IntVal, getContext()); 1302 EndLoc = Lexer.getTok().getEndLoc(); 1303 Lex(); // Eat token. 1304 return false; 1305 } 1306 case AsmToken::Dot: { 1307 if (!MAI.getDotIsPC()) 1308 return TokError("cannot use . as current PC"); 1309 1310 // This is a '.' reference, which references the current PC. Emit a 1311 // temporary label to the streamer and refer to it. 1312 MCSymbol *Sym = Ctx.createTempSymbol(); 1313 Out.emitLabel(Sym); 1314 Res = MCSymbolRefExpr::create(Sym, MCSymbolRefExpr::VK_None, getContext()); 1315 EndLoc = Lexer.getTok().getEndLoc(); 1316 Lex(); // Eat identifier. 1317 return false; 1318 } 1319 case AsmToken::LParen: 1320 Lex(); // Eat the '('. 1321 return parseParenExpr(Res, EndLoc); 1322 case AsmToken::LBrac: 1323 if (!PlatformParser->HasBracketExpressions()) 1324 return TokError("brackets expression not supported on this target"); 1325 Lex(); // Eat the '['. 1326 return parseBracketExpr(Res, EndLoc); 1327 case AsmToken::Minus: 1328 Lex(); // Eat the operator. 1329 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1330 return true; 1331 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc); 1332 return false; 1333 case AsmToken::Plus: 1334 Lex(); // Eat the operator. 1335 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1336 return true; 1337 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc); 1338 return false; 1339 case AsmToken::Tilde: 1340 Lex(); // Eat the operator. 1341 if (parsePrimaryExpr(Res, EndLoc, TypeInfo)) 1342 return true; 1343 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc); 1344 return false; 1345 // MIPS unary expression operators. The lexer won't generate these tokens if 1346 // MCAsmInfo::HasMipsExpressions is false for the target. 1347 case AsmToken::PercentCall16: 1348 case AsmToken::PercentCall_Hi: 1349 case AsmToken::PercentCall_Lo: 1350 case AsmToken::PercentDtprel_Hi: 1351 case AsmToken::PercentDtprel_Lo: 1352 case AsmToken::PercentGot: 1353 case AsmToken::PercentGot_Disp: 1354 case AsmToken::PercentGot_Hi: 1355 case AsmToken::PercentGot_Lo: 1356 case AsmToken::PercentGot_Ofst: 1357 case AsmToken::PercentGot_Page: 1358 case AsmToken::PercentGottprel: 1359 case AsmToken::PercentGp_Rel: 1360 case AsmToken::PercentHi: 1361 case AsmToken::PercentHigher: 1362 case AsmToken::PercentHighest: 1363 case AsmToken::PercentLo: 1364 case AsmToken::PercentNeg: 1365 case AsmToken::PercentPcrel_Hi: 1366 case AsmToken::PercentPcrel_Lo: 1367 case AsmToken::PercentTlsgd: 1368 case AsmToken::PercentTlsldm: 1369 case AsmToken::PercentTprel_Hi: 1370 case AsmToken::PercentTprel_Lo: 1371 Lex(); // Eat the operator. 1372 if (Lexer.isNot(AsmToken::LParen)) 1373 return TokError("expected '(' after operator"); 1374 Lex(); // Eat the operator. 1375 if (parseExpression(Res, EndLoc)) 1376 return true; 1377 if (parseRParen()) 1378 return true; 1379 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx); 1380 return !Res; 1381 } 1382 } 1383 1384 bool AsmParser::parseExpression(const MCExpr *&Res) { 1385 SMLoc EndLoc; 1386 return parseExpression(Res, EndLoc); 1387 } 1388 1389 const MCExpr * 1390 AsmParser::applyModifierToExpr(const MCExpr *E, 1391 MCSymbolRefExpr::VariantKind Variant) { 1392 // Ask the target implementation about this expression first. 1393 const MCExpr *NewE = getTargetParser().applyModifierToExpr(E, Variant, Ctx); 1394 if (NewE) 1395 return NewE; 1396 // Recurse over the given expression, rebuilding it to apply the given variant 1397 // if there is exactly one symbol. 1398 switch (E->getKind()) { 1399 case MCExpr::Target: 1400 case MCExpr::Constant: 1401 return nullptr; 1402 1403 case MCExpr::SymbolRef: { 1404 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 1405 1406 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 1407 TokError("invalid variant on expression '" + getTok().getIdentifier() + 1408 "' (already modified)"); 1409 return E; 1410 } 1411 1412 return MCSymbolRefExpr::create(&SRE->getSymbol(), Variant, getContext()); 1413 } 1414 1415 case MCExpr::Unary: { 1416 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 1417 const MCExpr *Sub = applyModifierToExpr(UE->getSubExpr(), Variant); 1418 if (!Sub) 1419 return nullptr; 1420 return MCUnaryExpr::create(UE->getOpcode(), Sub, getContext()); 1421 } 1422 1423 case MCExpr::Binary: { 1424 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 1425 const MCExpr *LHS = applyModifierToExpr(BE->getLHS(), Variant); 1426 const MCExpr *RHS = applyModifierToExpr(BE->getRHS(), Variant); 1427 1428 if (!LHS && !RHS) 1429 return nullptr; 1430 1431 if (!LHS) 1432 LHS = BE->getLHS(); 1433 if (!RHS) 1434 RHS = BE->getRHS(); 1435 1436 return MCBinaryExpr::create(BE->getOpcode(), LHS, RHS, getContext()); 1437 } 1438 } 1439 1440 llvm_unreachable("Invalid expression kind!"); 1441 } 1442 1443 /// This function checks if the next token is <string> type or arithmetic. 1444 /// string that begin with character '<' must end with character '>'. 1445 /// otherwise it is arithmetics. 1446 /// If the function returns a 'true' value, 1447 /// the End argument will be filled with the last location pointed to the '>' 1448 /// character. 1449 1450 /// There is a gap between the AltMacro's documentation and the single quote 1451 /// implementation. GCC does not fully support this feature and so we will not 1452 /// support it. 1453 /// TODO: Adding single quote as a string. 1454 static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) { 1455 assert((StrLoc.getPointer() != nullptr) && 1456 "Argument to the function cannot be a NULL value"); 1457 const char *CharPtr = StrLoc.getPointer(); 1458 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && 1459 (*CharPtr != '\0')) { 1460 if (*CharPtr == '!') 1461 CharPtr++; 1462 CharPtr++; 1463 } 1464 if (*CharPtr == '>') { 1465 EndLoc = StrLoc.getFromPointer(CharPtr + 1); 1466 return true; 1467 } 1468 return false; 1469 } 1470 1471 /// creating a string without the escape characters '!'. 1472 static std::string angleBracketString(StringRef AltMacroStr) { 1473 std::string Res; 1474 for (size_t Pos = 0; Pos < AltMacroStr.size(); Pos++) { 1475 if (AltMacroStr[Pos] == '!') 1476 Pos++; 1477 Res += AltMacroStr[Pos]; 1478 } 1479 return Res; 1480 } 1481 1482 /// Parse an expression and return it. 1483 /// 1484 /// expr ::= expr &&,|| expr -> lowest. 1485 /// expr ::= expr |,^,&,! expr 1486 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr 1487 /// expr ::= expr <<,>> expr 1488 /// expr ::= expr +,- expr 1489 /// expr ::= expr *,/,% expr -> highest. 1490 /// expr ::= primaryexpr 1491 /// 1492 bool AsmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1493 // Parse the expression. 1494 Res = nullptr; 1495 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || 1496 parseBinOpRHS(1, Res, EndLoc)) 1497 return true; 1498 1499 // As a special case, we support 'a op b @ modifier' by rewriting the 1500 // expression to include the modifier. This is inefficient, but in general we 1501 // expect users to use 'a@modifier op b'. 1502 if (Lexer.getKind() == AsmToken::At) { 1503 Lex(); 1504 1505 if (Lexer.isNot(AsmToken::Identifier)) 1506 return TokError("unexpected symbol modifier following '@'"); 1507 1508 MCSymbolRefExpr::VariantKind Variant = 1509 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 1510 if (Variant == MCSymbolRefExpr::VK_Invalid) 1511 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 1512 1513 const MCExpr *ModifiedRes = applyModifierToExpr(Res, Variant); 1514 if (!ModifiedRes) { 1515 return TokError("invalid modifier '" + getTok().getIdentifier() + 1516 "' (no symbols present)"); 1517 } 1518 1519 Res = ModifiedRes; 1520 Lex(); 1521 } 1522 1523 // Try to constant fold it up front, if possible. Do not exploit 1524 // assembler here. 1525 int64_t Value; 1526 if (Res->evaluateAsAbsolute(Value)) 1527 Res = MCConstantExpr::create(Value, getContext()); 1528 1529 return false; 1530 } 1531 1532 bool AsmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 1533 Res = nullptr; 1534 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc); 1535 } 1536 1537 bool AsmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, 1538 SMLoc &EndLoc) { 1539 if (parseParenExpr(Res, EndLoc)) 1540 return true; 1541 1542 for (; ParenDepth > 0; --ParenDepth) { 1543 if (parseBinOpRHS(1, Res, EndLoc)) 1544 return true; 1545 1546 // We don't Lex() the last RParen. 1547 // This is the same behavior as parseParenExpression(). 1548 if (ParenDepth - 1 > 0) { 1549 EndLoc = getTok().getEndLoc(); 1550 if (parseRParen()) 1551 return true; 1552 } 1553 } 1554 return false; 1555 } 1556 1557 bool AsmParser::parseAbsoluteExpression(int64_t &Res) { 1558 const MCExpr *Expr; 1559 1560 SMLoc StartLoc = Lexer.getLoc(); 1561 if (parseExpression(Expr)) 1562 return true; 1563 1564 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr())) 1565 return Error(StartLoc, "expected absolute expression"); 1566 1567 return false; 1568 } 1569 1570 static unsigned getDarwinBinOpPrecedence(AsmToken::TokenKind K, 1571 MCBinaryExpr::Opcode &Kind, 1572 bool ShouldUseLogicalShr) { 1573 switch (K) { 1574 default: 1575 return 0; // not a binop. 1576 1577 // Lowest Precedence: &&, || 1578 case AsmToken::AmpAmp: 1579 Kind = MCBinaryExpr::LAnd; 1580 return 1; 1581 case AsmToken::PipePipe: 1582 Kind = MCBinaryExpr::LOr; 1583 return 1; 1584 1585 // Low Precedence: |, &, ^ 1586 case AsmToken::Pipe: 1587 Kind = MCBinaryExpr::Or; 1588 return 2; 1589 case AsmToken::Caret: 1590 Kind = MCBinaryExpr::Xor; 1591 return 2; 1592 case AsmToken::Amp: 1593 Kind = MCBinaryExpr::And; 1594 return 2; 1595 1596 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 1597 case AsmToken::EqualEqual: 1598 Kind = MCBinaryExpr::EQ; 1599 return 3; 1600 case AsmToken::ExclaimEqual: 1601 case AsmToken::LessGreater: 1602 Kind = MCBinaryExpr::NE; 1603 return 3; 1604 case AsmToken::Less: 1605 Kind = MCBinaryExpr::LT; 1606 return 3; 1607 case AsmToken::LessEqual: 1608 Kind = MCBinaryExpr::LTE; 1609 return 3; 1610 case AsmToken::Greater: 1611 Kind = MCBinaryExpr::GT; 1612 return 3; 1613 case AsmToken::GreaterEqual: 1614 Kind = MCBinaryExpr::GTE; 1615 return 3; 1616 1617 // Intermediate Precedence: <<, >> 1618 case AsmToken::LessLess: 1619 Kind = MCBinaryExpr::Shl; 1620 return 4; 1621 case AsmToken::GreaterGreater: 1622 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1623 return 4; 1624 1625 // High Intermediate Precedence: +, - 1626 case AsmToken::Plus: 1627 Kind = MCBinaryExpr::Add; 1628 return 5; 1629 case AsmToken::Minus: 1630 Kind = MCBinaryExpr::Sub; 1631 return 5; 1632 1633 // Highest Precedence: *, /, % 1634 case AsmToken::Star: 1635 Kind = MCBinaryExpr::Mul; 1636 return 6; 1637 case AsmToken::Slash: 1638 Kind = MCBinaryExpr::Div; 1639 return 6; 1640 case AsmToken::Percent: 1641 Kind = MCBinaryExpr::Mod; 1642 return 6; 1643 } 1644 } 1645 1646 static unsigned getGNUBinOpPrecedence(const MCAsmInfo &MAI, 1647 AsmToken::TokenKind K, 1648 MCBinaryExpr::Opcode &Kind, 1649 bool ShouldUseLogicalShr) { 1650 switch (K) { 1651 default: 1652 return 0; // not a binop. 1653 1654 // Lowest Precedence: &&, || 1655 case AsmToken::AmpAmp: 1656 Kind = MCBinaryExpr::LAnd; 1657 return 2; 1658 case AsmToken::PipePipe: 1659 Kind = MCBinaryExpr::LOr; 1660 return 1; 1661 1662 // Low Precedence: ==, !=, <>, <, <=, >, >= 1663 case AsmToken::EqualEqual: 1664 Kind = MCBinaryExpr::EQ; 1665 return 3; 1666 case AsmToken::ExclaimEqual: 1667 case AsmToken::LessGreater: 1668 Kind = MCBinaryExpr::NE; 1669 return 3; 1670 case AsmToken::Less: 1671 Kind = MCBinaryExpr::LT; 1672 return 3; 1673 case AsmToken::LessEqual: 1674 Kind = MCBinaryExpr::LTE; 1675 return 3; 1676 case AsmToken::Greater: 1677 Kind = MCBinaryExpr::GT; 1678 return 3; 1679 case AsmToken::GreaterEqual: 1680 Kind = MCBinaryExpr::GTE; 1681 return 3; 1682 1683 // Low Intermediate Precedence: +, - 1684 case AsmToken::Plus: 1685 Kind = MCBinaryExpr::Add; 1686 return 4; 1687 case AsmToken::Minus: 1688 Kind = MCBinaryExpr::Sub; 1689 return 4; 1690 1691 // High Intermediate Precedence: |, !, &, ^ 1692 // 1693 case AsmToken::Pipe: 1694 Kind = MCBinaryExpr::Or; 1695 return 5; 1696 case AsmToken::Exclaim: 1697 // Hack to support ARM compatible aliases (implied 'sp' operand in 'srs*' 1698 // instructions like 'srsda #31!') and not parse ! as an infix operator. 1699 if (MAI.getCommentString() == "@") 1700 return 0; 1701 Kind = MCBinaryExpr::OrNot; 1702 return 5; 1703 case AsmToken::Caret: 1704 Kind = MCBinaryExpr::Xor; 1705 return 5; 1706 case AsmToken::Amp: 1707 Kind = MCBinaryExpr::And; 1708 return 5; 1709 1710 // Highest Precedence: *, /, %, <<, >> 1711 case AsmToken::Star: 1712 Kind = MCBinaryExpr::Mul; 1713 return 6; 1714 case AsmToken::Slash: 1715 Kind = MCBinaryExpr::Div; 1716 return 6; 1717 case AsmToken::Percent: 1718 Kind = MCBinaryExpr::Mod; 1719 return 6; 1720 case AsmToken::LessLess: 1721 Kind = MCBinaryExpr::Shl; 1722 return 6; 1723 case AsmToken::GreaterGreater: 1724 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; 1725 return 6; 1726 } 1727 } 1728 1729 unsigned AsmParser::getBinOpPrecedence(AsmToken::TokenKind K, 1730 MCBinaryExpr::Opcode &Kind) { 1731 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); 1732 return IsDarwin ? getDarwinBinOpPrecedence(K, Kind, ShouldUseLogicalShr) 1733 : getGNUBinOpPrecedence(MAI, K, Kind, ShouldUseLogicalShr); 1734 } 1735 1736 /// Parse all binary operators with precedence >= 'Precedence'. 1737 /// Res contains the LHS of the expression on input. 1738 bool AsmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 1739 SMLoc &EndLoc) { 1740 SMLoc StartLoc = Lexer.getLoc(); 1741 while (true) { 1742 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 1743 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 1744 1745 // If the next token is lower precedence than we are allowed to eat, return 1746 // successfully with what we ate already. 1747 if (TokPrec < Precedence) 1748 return false; 1749 1750 Lex(); 1751 1752 // Eat the next primary expression. 1753 const MCExpr *RHS; 1754 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc)) 1755 return true; 1756 1757 // If BinOp binds less tightly with RHS than the operator after RHS, let 1758 // the pending operator take RHS as its LHS. 1759 MCBinaryExpr::Opcode Dummy; 1760 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1761 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc)) 1762 return true; 1763 1764 // Merge LHS and RHS according to operator. 1765 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc); 1766 } 1767 } 1768 1769 /// ParseStatement: 1770 /// ::= EndOfStatement 1771 /// ::= Label* Directive ...Operands... EndOfStatement 1772 /// ::= Label* Identifier OperandList* EndOfStatement 1773 bool AsmParser::parseStatement(ParseStatementInfo &Info, 1774 MCAsmParserSemaCallback *SI) { 1775 assert(!hasPendingError() && "parseStatement started with pending error"); 1776 // Eat initial spaces and comments 1777 while (Lexer.is(AsmToken::Space)) 1778 Lex(); 1779 if (Lexer.is(AsmToken::EndOfStatement)) { 1780 // if this is a line comment we can drop it safely 1781 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 1782 getTok().getString().front() == '\n') 1783 Out.AddBlankLine(); 1784 Lex(); 1785 return false; 1786 } 1787 // Statements always start with an identifier. 1788 AsmToken ID = getTok(); 1789 SMLoc IDLoc = ID.getLoc(); 1790 StringRef IDVal; 1791 int64_t LocalLabelVal = -1; 1792 StartTokLoc = ID.getLoc(); 1793 if (Lexer.is(AsmToken::HashDirective)) 1794 return parseCppHashLineFilenameComment(IDLoc, 1795 !isInsideMacroInstantiation()); 1796 1797 // Allow an integer followed by a ':' as a directional local label. 1798 if (Lexer.is(AsmToken::Integer)) { 1799 LocalLabelVal = getTok().getIntVal(); 1800 if (LocalLabelVal < 0) { 1801 if (!TheCondState.Ignore) { 1802 Lex(); // always eat a token 1803 return Error(IDLoc, "unexpected token at start of statement"); 1804 } 1805 IDVal = ""; 1806 } else { 1807 IDVal = getTok().getString(); 1808 Lex(); // Consume the integer token to be used as an identifier token. 1809 if (Lexer.getKind() != AsmToken::Colon) { 1810 if (!TheCondState.Ignore) { 1811 Lex(); // always eat a token 1812 return Error(IDLoc, "unexpected token at start of statement"); 1813 } 1814 } 1815 } 1816 } else if (Lexer.is(AsmToken::Dot)) { 1817 // Treat '.' as a valid identifier in this context. 1818 Lex(); 1819 IDVal = "."; 1820 } else if (Lexer.is(AsmToken::LCurly)) { 1821 // Treat '{' as a valid identifier in this context. 1822 Lex(); 1823 IDVal = "{"; 1824 1825 } else if (Lexer.is(AsmToken::RCurly)) { 1826 // Treat '}' as a valid identifier in this context. 1827 Lex(); 1828 IDVal = "}"; 1829 } else if (Lexer.is(AsmToken::Star) && 1830 getTargetParser().starIsStartOfStatement()) { 1831 // Accept '*' as a valid start of statement. 1832 Lex(); 1833 IDVal = "*"; 1834 } else if (parseIdentifier(IDVal)) { 1835 if (!TheCondState.Ignore) { 1836 Lex(); // always eat a token 1837 return Error(IDLoc, "unexpected token at start of statement"); 1838 } 1839 IDVal = ""; 1840 } 1841 1842 // Handle conditional assembly here before checking for skipping. We 1843 // have to do this so that .endif isn't skipped in a ".if 0" block for 1844 // example. 1845 StringMap<DirectiveKind>::const_iterator DirKindIt = 1846 DirectiveKindMap.find(IDVal.lower()); 1847 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) 1848 ? DK_NO_DIRECTIVE 1849 : DirKindIt->getValue(); 1850 switch (DirKind) { 1851 default: 1852 break; 1853 case DK_IF: 1854 case DK_IFEQ: 1855 case DK_IFGE: 1856 case DK_IFGT: 1857 case DK_IFLE: 1858 case DK_IFLT: 1859 case DK_IFNE: 1860 return parseDirectiveIf(IDLoc, DirKind); 1861 case DK_IFB: 1862 return parseDirectiveIfb(IDLoc, true); 1863 case DK_IFNB: 1864 return parseDirectiveIfb(IDLoc, false); 1865 case DK_IFC: 1866 return parseDirectiveIfc(IDLoc, true); 1867 case DK_IFEQS: 1868 return parseDirectiveIfeqs(IDLoc, true); 1869 case DK_IFNC: 1870 return parseDirectiveIfc(IDLoc, false); 1871 case DK_IFNES: 1872 return parseDirectiveIfeqs(IDLoc, false); 1873 case DK_IFDEF: 1874 return parseDirectiveIfdef(IDLoc, true); 1875 case DK_IFNDEF: 1876 case DK_IFNOTDEF: 1877 return parseDirectiveIfdef(IDLoc, false); 1878 case DK_ELSEIF: 1879 return parseDirectiveElseIf(IDLoc); 1880 case DK_ELSE: 1881 return parseDirectiveElse(IDLoc); 1882 case DK_ENDIF: 1883 return parseDirectiveEndIf(IDLoc); 1884 } 1885 1886 // Ignore the statement if in the middle of inactive conditional 1887 // (e.g. ".if 0"). 1888 if (TheCondState.Ignore) { 1889 eatToEndOfStatement(); 1890 return false; 1891 } 1892 1893 // FIXME: Recurse on local labels? 1894 1895 // See what kind of statement we have. 1896 switch (Lexer.getKind()) { 1897 case AsmToken::Colon: { 1898 if (!getTargetParser().isLabel(ID)) 1899 break; 1900 if (checkForValidSection()) 1901 return true; 1902 1903 // identifier ':' -> Label. 1904 Lex(); 1905 1906 // Diagnose attempt to use '.' as a label. 1907 if (IDVal == ".") 1908 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1909 1910 // Diagnose attempt to use a variable as a label. 1911 // 1912 // FIXME: Diagnostics. Note the location of the definition as a label. 1913 // FIXME: This doesn't diagnose assignment to a symbol which has been 1914 // implicitly marked as external. 1915 MCSymbol *Sym; 1916 if (LocalLabelVal == -1) { 1917 if (ParsingMSInlineAsm && SI) { 1918 StringRef RewrittenLabel = 1919 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true); 1920 assert(!RewrittenLabel.empty() && 1921 "We should have an internal name here."); 1922 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(), 1923 RewrittenLabel); 1924 IDVal = RewrittenLabel; 1925 } 1926 Sym = getContext().getOrCreateSymbol(IDVal); 1927 } else 1928 Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal); 1929 // End of Labels should be treated as end of line for lexing 1930 // purposes but that information is not available to the Lexer who 1931 // does not understand Labels. This may cause us to see a Hash 1932 // here instead of a preprocessor line comment. 1933 if (getTok().is(AsmToken::Hash)) { 1934 StringRef CommentStr = parseStringToEndOfStatement(); 1935 Lexer.Lex(); 1936 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr)); 1937 } 1938 1939 // Consume any end of statement token, if present, to avoid spurious 1940 // AddBlankLine calls(). 1941 if (getTok().is(AsmToken::EndOfStatement)) { 1942 Lex(); 1943 } 1944 1945 if (discardLTOSymbol(IDVal)) 1946 return false; 1947 1948 getTargetParser().doBeforeLabelEmit(Sym); 1949 1950 // Emit the label. 1951 if (!getTargetParser().isParsingMSInlineAsm()) 1952 Out.emitLabel(Sym, IDLoc); 1953 1954 // If we are generating dwarf for assembly source files then gather the 1955 // info to make a dwarf label entry for this label if needed. 1956 if (enabledGenDwarfForAssembly()) 1957 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1958 IDLoc); 1959 1960 getTargetParser().onLabelParsed(Sym); 1961 1962 return false; 1963 } 1964 1965 case AsmToken::Equal: 1966 if (!getTargetParser().equalIsAsmAssignment()) 1967 break; 1968 // identifier '=' ... -> assignment statement 1969 Lex(); 1970 1971 return parseAssignment(IDVal, AssignmentKind::Equal); 1972 1973 default: // Normal instruction or directive. 1974 break; 1975 } 1976 1977 // If macros are enabled, check to see if this is a macro instantiation. 1978 if (areMacrosEnabled()) 1979 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal)) { 1980 return handleMacroEntry(M, IDLoc); 1981 } 1982 1983 // Otherwise, we have a normal instruction or directive. 1984 1985 // Directives start with "." 1986 if (IDVal.startswith(".") && IDVal != ".") { 1987 // There are several entities interested in parsing directives: 1988 // 1989 // 1. The target-specific assembly parser. Some directives are target 1990 // specific or may potentially behave differently on certain targets. 1991 // 2. Asm parser extensions. For example, platform-specific parsers 1992 // (like the ELF parser) register themselves as extensions. 1993 // 3. The generic directive parser implemented by this class. These are 1994 // all the directives that behave in a target and platform independent 1995 // manner, or at least have a default behavior that's shared between 1996 // all targets and platforms. 1997 1998 getTargetParser().flushPendingInstructions(getStreamer()); 1999 2000 SMLoc StartTokLoc = getTok().getLoc(); 2001 bool TPDirectiveReturn = getTargetParser().ParseDirective(ID); 2002 2003 if (hasPendingError()) 2004 return true; 2005 // Currently the return value should be true if we are 2006 // uninterested but as this is at odds with the standard parsing 2007 // convention (return true = error) we have instances of a parsed 2008 // directive that fails returning true as an error. Catch these 2009 // cases as best as possible errors here. 2010 if (TPDirectiveReturn && StartTokLoc != getTok().getLoc()) 2011 return true; 2012 // Return if we did some parsing or believe we succeeded. 2013 if (!TPDirectiveReturn || StartTokLoc != getTok().getLoc()) 2014 return false; 2015 2016 // Next, check the extension directive map to see if any extension has 2017 // registered itself to parse this directive. 2018 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = 2019 ExtensionDirectiveMap.lookup(IDVal); 2020 if (Handler.first) 2021 return (*Handler.second)(Handler.first, IDVal, IDLoc); 2022 2023 // Finally, if no one else is interested in this directive, it must be 2024 // generic and familiar to this class. 2025 switch (DirKind) { 2026 default: 2027 break; 2028 case DK_SET: 2029 case DK_EQU: 2030 return parseDirectiveSet(IDVal, AssignmentKind::Set); 2031 case DK_EQUIV: 2032 return parseDirectiveSet(IDVal, AssignmentKind::Equiv); 2033 case DK_LTO_SET_CONDITIONAL: 2034 return parseDirectiveSet(IDVal, AssignmentKind::LTOSetConditional); 2035 case DK_ASCII: 2036 return parseDirectiveAscii(IDVal, false); 2037 case DK_ASCIZ: 2038 case DK_STRING: 2039 return parseDirectiveAscii(IDVal, true); 2040 case DK_BYTE: 2041 case DK_DC_B: 2042 return parseDirectiveValue(IDVal, 1); 2043 case DK_DC: 2044 case DK_DC_W: 2045 case DK_SHORT: 2046 case DK_VALUE: 2047 case DK_2BYTE: 2048 return parseDirectiveValue(IDVal, 2); 2049 case DK_LONG: 2050 case DK_INT: 2051 case DK_4BYTE: 2052 case DK_DC_L: 2053 return parseDirectiveValue(IDVal, 4); 2054 case DK_QUAD: 2055 case DK_8BYTE: 2056 return parseDirectiveValue(IDVal, 8); 2057 case DK_DC_A: 2058 return parseDirectiveValue( 2059 IDVal, getContext().getAsmInfo()->getCodePointerSize()); 2060 case DK_OCTA: 2061 return parseDirectiveOctaValue(IDVal); 2062 case DK_SINGLE: 2063 case DK_FLOAT: 2064 case DK_DC_S: 2065 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle()); 2066 case DK_DOUBLE: 2067 case DK_DC_D: 2068 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble()); 2069 case DK_ALIGN: { 2070 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 2071 return parseDirectiveAlign(IsPow2, /*ExprSize=*/1); 2072 } 2073 case DK_ALIGN32: { 2074 bool IsPow2 = !getContext().getAsmInfo()->getAlignmentIsInBytes(); 2075 return parseDirectiveAlign(IsPow2, /*ExprSize=*/4); 2076 } 2077 case DK_BALIGN: 2078 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 2079 case DK_BALIGNW: 2080 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 2081 case DK_BALIGNL: 2082 return parseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 2083 case DK_P2ALIGN: 2084 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 2085 case DK_P2ALIGNW: 2086 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 2087 case DK_P2ALIGNL: 2088 return parseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 2089 case DK_ORG: 2090 return parseDirectiveOrg(); 2091 case DK_FILL: 2092 return parseDirectiveFill(); 2093 case DK_ZERO: 2094 return parseDirectiveZero(); 2095 case DK_EXTERN: 2096 eatToEndOfStatement(); // .extern is the default, ignore it. 2097 return false; 2098 case DK_GLOBL: 2099 case DK_GLOBAL: 2100 return parseDirectiveSymbolAttribute(MCSA_Global); 2101 case DK_LAZY_REFERENCE: 2102 return parseDirectiveSymbolAttribute(MCSA_LazyReference); 2103 case DK_NO_DEAD_STRIP: 2104 return parseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 2105 case DK_SYMBOL_RESOLVER: 2106 return parseDirectiveSymbolAttribute(MCSA_SymbolResolver); 2107 case DK_PRIVATE_EXTERN: 2108 return parseDirectiveSymbolAttribute(MCSA_PrivateExtern); 2109 case DK_REFERENCE: 2110 return parseDirectiveSymbolAttribute(MCSA_Reference); 2111 case DK_WEAK_DEFINITION: 2112 return parseDirectiveSymbolAttribute(MCSA_WeakDefinition); 2113 case DK_WEAK_REFERENCE: 2114 return parseDirectiveSymbolAttribute(MCSA_WeakReference); 2115 case DK_WEAK_DEF_CAN_BE_HIDDEN: 2116 return parseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 2117 case DK_COLD: 2118 return parseDirectiveSymbolAttribute(MCSA_Cold); 2119 case DK_COMM: 2120 case DK_COMMON: 2121 return parseDirectiveComm(/*IsLocal=*/false); 2122 case DK_LCOMM: 2123 return parseDirectiveComm(/*IsLocal=*/true); 2124 case DK_ABORT: 2125 return parseDirectiveAbort(); 2126 case DK_INCLUDE: 2127 return parseDirectiveInclude(); 2128 case DK_INCBIN: 2129 return parseDirectiveIncbin(); 2130 case DK_CODE16: 2131 case DK_CODE16GCC: 2132 return TokError(Twine(IDVal) + 2133 " not currently supported for this target"); 2134 case DK_REPT: 2135 return parseDirectiveRept(IDLoc, IDVal); 2136 case DK_IRP: 2137 return parseDirectiveIrp(IDLoc); 2138 case DK_IRPC: 2139 return parseDirectiveIrpc(IDLoc); 2140 case DK_ENDR: 2141 return parseDirectiveEndr(IDLoc); 2142 case DK_BUNDLE_ALIGN_MODE: 2143 return parseDirectiveBundleAlignMode(); 2144 case DK_BUNDLE_LOCK: 2145 return parseDirectiveBundleLock(); 2146 case DK_BUNDLE_UNLOCK: 2147 return parseDirectiveBundleUnlock(); 2148 case DK_SLEB128: 2149 return parseDirectiveLEB128(true); 2150 case DK_ULEB128: 2151 return parseDirectiveLEB128(false); 2152 case DK_SPACE: 2153 case DK_SKIP: 2154 return parseDirectiveSpace(IDVal); 2155 case DK_FILE: 2156 return parseDirectiveFile(IDLoc); 2157 case DK_LINE: 2158 return parseDirectiveLine(); 2159 case DK_LOC: 2160 return parseDirectiveLoc(); 2161 case DK_STABS: 2162 return parseDirectiveStabs(); 2163 case DK_CV_FILE: 2164 return parseDirectiveCVFile(); 2165 case DK_CV_FUNC_ID: 2166 return parseDirectiveCVFuncId(); 2167 case DK_CV_INLINE_SITE_ID: 2168 return parseDirectiveCVInlineSiteId(); 2169 case DK_CV_LOC: 2170 return parseDirectiveCVLoc(); 2171 case DK_CV_LINETABLE: 2172 return parseDirectiveCVLinetable(); 2173 case DK_CV_INLINE_LINETABLE: 2174 return parseDirectiveCVInlineLinetable(); 2175 case DK_CV_DEF_RANGE: 2176 return parseDirectiveCVDefRange(); 2177 case DK_CV_STRING: 2178 return parseDirectiveCVString(); 2179 case DK_CV_STRINGTABLE: 2180 return parseDirectiveCVStringTable(); 2181 case DK_CV_FILECHECKSUMS: 2182 return parseDirectiveCVFileChecksums(); 2183 case DK_CV_FILECHECKSUM_OFFSET: 2184 return parseDirectiveCVFileChecksumOffset(); 2185 case DK_CV_FPO_DATA: 2186 return parseDirectiveCVFPOData(); 2187 case DK_CFI_SECTIONS: 2188 return parseDirectiveCFISections(); 2189 case DK_CFI_STARTPROC: 2190 return parseDirectiveCFIStartProc(); 2191 case DK_CFI_ENDPROC: 2192 return parseDirectiveCFIEndProc(); 2193 case DK_CFI_DEF_CFA: 2194 return parseDirectiveCFIDefCfa(IDLoc); 2195 case DK_CFI_DEF_CFA_OFFSET: 2196 return parseDirectiveCFIDefCfaOffset(); 2197 case DK_CFI_ADJUST_CFA_OFFSET: 2198 return parseDirectiveCFIAdjustCfaOffset(); 2199 case DK_CFI_DEF_CFA_REGISTER: 2200 return parseDirectiveCFIDefCfaRegister(IDLoc); 2201 case DK_CFI_LLVM_DEF_ASPACE_CFA: 2202 return parseDirectiveCFILLVMDefAspaceCfa(IDLoc); 2203 case DK_CFI_OFFSET: 2204 return parseDirectiveCFIOffset(IDLoc); 2205 case DK_CFI_REL_OFFSET: 2206 return parseDirectiveCFIRelOffset(IDLoc); 2207 case DK_CFI_PERSONALITY: 2208 return parseDirectiveCFIPersonalityOrLsda(true); 2209 case DK_CFI_LSDA: 2210 return parseDirectiveCFIPersonalityOrLsda(false); 2211 case DK_CFI_REMEMBER_STATE: 2212 return parseDirectiveCFIRememberState(); 2213 case DK_CFI_RESTORE_STATE: 2214 return parseDirectiveCFIRestoreState(); 2215 case DK_CFI_SAME_VALUE: 2216 return parseDirectiveCFISameValue(IDLoc); 2217 case DK_CFI_RESTORE: 2218 return parseDirectiveCFIRestore(IDLoc); 2219 case DK_CFI_ESCAPE: 2220 return parseDirectiveCFIEscape(); 2221 case DK_CFI_RETURN_COLUMN: 2222 return parseDirectiveCFIReturnColumn(IDLoc); 2223 case DK_CFI_SIGNAL_FRAME: 2224 return parseDirectiveCFISignalFrame(); 2225 case DK_CFI_UNDEFINED: 2226 return parseDirectiveCFIUndefined(IDLoc); 2227 case DK_CFI_REGISTER: 2228 return parseDirectiveCFIRegister(IDLoc); 2229 case DK_CFI_WINDOW_SAVE: 2230 return parseDirectiveCFIWindowSave(); 2231 case DK_MACROS_ON: 2232 case DK_MACROS_OFF: 2233 return parseDirectiveMacrosOnOff(IDVal); 2234 case DK_MACRO: 2235 return parseDirectiveMacro(IDLoc); 2236 case DK_ALTMACRO: 2237 case DK_NOALTMACRO: 2238 return parseDirectiveAltmacro(IDVal); 2239 case DK_EXITM: 2240 return parseDirectiveExitMacro(IDVal); 2241 case DK_ENDM: 2242 case DK_ENDMACRO: 2243 return parseDirectiveEndMacro(IDVal); 2244 case DK_PURGEM: 2245 return parseDirectivePurgeMacro(IDLoc); 2246 case DK_END: 2247 return parseDirectiveEnd(IDLoc); 2248 case DK_ERR: 2249 return parseDirectiveError(IDLoc, false); 2250 case DK_ERROR: 2251 return parseDirectiveError(IDLoc, true); 2252 case DK_WARNING: 2253 return parseDirectiveWarning(IDLoc); 2254 case DK_RELOC: 2255 return parseDirectiveReloc(IDLoc); 2256 case DK_DCB: 2257 case DK_DCB_W: 2258 return parseDirectiveDCB(IDVal, 2); 2259 case DK_DCB_B: 2260 return parseDirectiveDCB(IDVal, 1); 2261 case DK_DCB_D: 2262 return parseDirectiveRealDCB(IDVal, APFloat::IEEEdouble()); 2263 case DK_DCB_L: 2264 return parseDirectiveDCB(IDVal, 4); 2265 case DK_DCB_S: 2266 return parseDirectiveRealDCB(IDVal, APFloat::IEEEsingle()); 2267 case DK_DC_X: 2268 case DK_DCB_X: 2269 return TokError(Twine(IDVal) + 2270 " not currently supported for this target"); 2271 case DK_DS: 2272 case DK_DS_W: 2273 return parseDirectiveDS(IDVal, 2); 2274 case DK_DS_B: 2275 return parseDirectiveDS(IDVal, 1); 2276 case DK_DS_D: 2277 return parseDirectiveDS(IDVal, 8); 2278 case DK_DS_L: 2279 case DK_DS_S: 2280 return parseDirectiveDS(IDVal, 4); 2281 case DK_DS_P: 2282 case DK_DS_X: 2283 return parseDirectiveDS(IDVal, 12); 2284 case DK_PRINT: 2285 return parseDirectivePrint(IDLoc); 2286 case DK_ADDRSIG: 2287 return parseDirectiveAddrsig(); 2288 case DK_ADDRSIG_SYM: 2289 return parseDirectiveAddrsigSym(); 2290 case DK_PSEUDO_PROBE: 2291 return parseDirectivePseudoProbe(); 2292 case DK_LTO_DISCARD: 2293 return parseDirectiveLTODiscard(); 2294 } 2295 2296 return Error(IDLoc, "unknown directive"); 2297 } 2298 2299 // __asm _emit or __asm __emit 2300 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || 2301 IDVal == "_EMIT" || IDVal == "__EMIT")) 2302 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size()); 2303 2304 // __asm align 2305 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN")) 2306 return parseDirectiveMSAlign(IDLoc, Info); 2307 2308 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN")) 2309 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4); 2310 if (checkForValidSection()) 2311 return true; 2312 2313 return parseAndMatchAndEmitTargetInstruction(Info, IDVal, ID, IDLoc); 2314 } 2315 2316 bool AsmParser::parseAndMatchAndEmitTargetInstruction(ParseStatementInfo &Info, 2317 StringRef IDVal, 2318 AsmToken ID, 2319 SMLoc IDLoc) { 2320 // Canonicalize the opcode to lower case. 2321 std::string OpcodeStr = IDVal.lower(); 2322 ParseInstructionInfo IInfo(Info.AsmRewrites); 2323 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID, 2324 Info.ParsedOperands); 2325 Info.ParseError = ParseHadError; 2326 2327 // Dump the parsed representation, if requested. 2328 if (getShowParsedOperands()) { 2329 SmallString<256> Str; 2330 raw_svector_ostream OS(Str); 2331 OS << "parsed instruction: ["; 2332 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { 2333 if (i != 0) 2334 OS << ", "; 2335 Info.ParsedOperands[i]->print(OS); 2336 } 2337 OS << "]"; 2338 2339 printMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 2340 } 2341 2342 // Fail even if ParseInstruction erroneously returns false. 2343 if (hasPendingError() || ParseHadError) 2344 return true; 2345 2346 // If we are generating dwarf for the current section then generate a .loc 2347 // directive for the instruction. 2348 if (!ParseHadError && enabledGenDwarfForAssembly() && 2349 getContext().getGenDwarfSectionSyms().count( 2350 getStreamer().getCurrentSectionOnly())) { 2351 unsigned Line; 2352 if (ActiveMacros.empty()) 2353 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer); 2354 else 2355 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc, 2356 ActiveMacros.front()->ExitBuffer); 2357 2358 // If we previously parsed a cpp hash file line comment then make sure the 2359 // current Dwarf File is for the CppHashFilename if not then emit the 2360 // Dwarf File table for it and adjust the line number for the .loc. 2361 if (!CppHashInfo.Filename.empty()) { 2362 unsigned FileNumber = getStreamer().emitDwarfFileDirective( 2363 0, StringRef(), CppHashInfo.Filename); 2364 getContext().setGenDwarfFileNumber(FileNumber); 2365 2366 unsigned CppHashLocLineNo = 2367 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf); 2368 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); 2369 } 2370 2371 getStreamer().emitDwarfLocDirective( 2372 getContext().getGenDwarfFileNumber(), Line, 0, 2373 DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, 0, 0, 2374 StringRef()); 2375 } 2376 2377 // If parsing succeeded, match the instruction. 2378 if (!ParseHadError) { 2379 uint64_t ErrorInfo; 2380 if (getTargetParser().MatchAndEmitInstruction( 2381 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo, 2382 getTargetParser().isParsingMSInlineAsm())) 2383 return true; 2384 } 2385 return false; 2386 } 2387 2388 // Parse and erase curly braces marking block start/end 2389 bool 2390 AsmParser::parseCurlyBlockScope(SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { 2391 // Identify curly brace marking block start/end 2392 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly)) 2393 return false; 2394 2395 SMLoc StartLoc = Lexer.getLoc(); 2396 Lex(); // Eat the brace 2397 if (Lexer.is(AsmToken::EndOfStatement)) 2398 Lex(); // Eat EndOfStatement following the brace 2399 2400 // Erase the block start/end brace from the output asm string 2401 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() - 2402 StartLoc.getPointer()); 2403 return true; 2404 } 2405 2406 /// parseCppHashLineFilenameComment as this: 2407 /// ::= # number "filename" 2408 bool AsmParser::parseCppHashLineFilenameComment(SMLoc L, bool SaveLocInfo) { 2409 Lex(); // Eat the hash token. 2410 // Lexer only ever emits HashDirective if it fully formed if it's 2411 // done the checking already so this is an internal error. 2412 assert(getTok().is(AsmToken::Integer) && 2413 "Lexing Cpp line comment: Expected Integer"); 2414 int64_t LineNumber = getTok().getIntVal(); 2415 Lex(); 2416 assert(getTok().is(AsmToken::String) && 2417 "Lexing Cpp line comment: Expected String"); 2418 StringRef Filename = getTok().getString(); 2419 Lex(); 2420 2421 if (!SaveLocInfo) 2422 return false; 2423 2424 // Get rid of the enclosing quotes. 2425 Filename = Filename.substr(1, Filename.size() - 2); 2426 2427 // Save the SMLoc, Filename and LineNumber for later use by diagnostics 2428 // and possibly DWARF file info. 2429 CppHashInfo.Loc = L; 2430 CppHashInfo.Filename = Filename; 2431 CppHashInfo.LineNumber = LineNumber; 2432 CppHashInfo.Buf = CurBuffer; 2433 if (FirstCppHashFilename.empty()) 2434 FirstCppHashFilename = Filename; 2435 return false; 2436 } 2437 2438 /// will use the last parsed cpp hash line filename comment 2439 /// for the Filename and LineNo if any in the diagnostic. 2440 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 2441 auto *Parser = static_cast<AsmParser *>(Context); 2442 raw_ostream &OS = errs(); 2443 2444 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 2445 SMLoc DiagLoc = Diag.getLoc(); 2446 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2447 unsigned CppHashBuf = 2448 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc); 2449 2450 // Like SourceMgr::printMessage() we need to print the include stack if any 2451 // before printing the message. 2452 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 2453 if (!Parser->SavedDiagHandler && DiagCurBuffer && 2454 DiagCurBuffer != DiagSrcMgr.getMainFileID()) { 2455 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 2456 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 2457 } 2458 2459 // If we have not parsed a cpp hash line filename comment or the source 2460 // manager changed or buffer changed (like in a nested include) then just 2461 // print the normal diagnostic using its Filename and LineNo. 2462 if (!Parser->CppHashInfo.LineNumber || DiagBuf != CppHashBuf) { 2463 if (Parser->SavedDiagHandler) 2464 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2465 else 2466 Parser->getContext().diagnose(Diag); 2467 return; 2468 } 2469 2470 // Use the CppHashFilename and calculate a line number based on the 2471 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc 2472 // for the diagnostic. 2473 const std::string &Filename = std::string(Parser->CppHashInfo.Filename); 2474 2475 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 2476 int CppHashLocLineNo = 2477 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf); 2478 int LineNo = 2479 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); 2480 2481 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, 2482 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), 2483 Diag.getLineContents(), Diag.getRanges()); 2484 2485 if (Parser->SavedDiagHandler) 2486 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 2487 else 2488 Parser->getContext().diagnose(NewDiag); 2489 } 2490 2491 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The 2492 // difference being that that function accepts '@' as part of identifiers and 2493 // we can't do that. AsmLexer.cpp should probably be changed to handle 2494 // '@' as a special case when needed. 2495 static bool isIdentifierChar(char c) { 2496 return isalnum(static_cast<unsigned char>(c)) || c == '_' || c == '$' || 2497 c == '.'; 2498 } 2499 2500 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 2501 ArrayRef<MCAsmMacroParameter> Parameters, 2502 ArrayRef<MCAsmMacroArgument> A, 2503 bool EnableAtPseudoVariable, SMLoc L) { 2504 unsigned NParameters = Parameters.size(); 2505 bool HasVararg = NParameters ? Parameters.back().Vararg : false; 2506 if ((!IsDarwin || NParameters != 0) && NParameters != A.size()) 2507 return Error(L, "Wrong number of arguments"); 2508 2509 // A macro without parameters is handled differently on Darwin: 2510 // gas accepts no arguments and does no substitutions 2511 while (!Body.empty()) { 2512 // Scan for the next substitution. 2513 std::size_t End = Body.size(), Pos = 0; 2514 for (; Pos != End; ++Pos) { 2515 // Check for a substitution or escape. 2516 if (IsDarwin && !NParameters) { 2517 // This macro has no parameters, look for $0, $1, etc. 2518 if (Body[Pos] != '$' || Pos + 1 == End) 2519 continue; 2520 2521 char Next = Body[Pos + 1]; 2522 if (Next == '$' || Next == 'n' || 2523 isdigit(static_cast<unsigned char>(Next))) 2524 break; 2525 } else { 2526 // This macro has parameters, look for \foo, \bar, etc. 2527 if (Body[Pos] == '\\' && Pos + 1 != End) 2528 break; 2529 } 2530 } 2531 2532 // Add the prefix. 2533 OS << Body.slice(0, Pos); 2534 2535 // Check if we reached the end. 2536 if (Pos == End) 2537 break; 2538 2539 if (IsDarwin && !NParameters) { 2540 switch (Body[Pos + 1]) { 2541 // $$ => $ 2542 case '$': 2543 OS << '$'; 2544 break; 2545 2546 // $n => number of arguments 2547 case 'n': 2548 OS << A.size(); 2549 break; 2550 2551 // $[0-9] => argument 2552 default: { 2553 // Missing arguments are ignored. 2554 unsigned Index = Body[Pos + 1] - '0'; 2555 if (Index >= A.size()) 2556 break; 2557 2558 // Otherwise substitute with the token values, with spaces eliminated. 2559 for (const AsmToken &Token : A[Index]) 2560 OS << Token.getString(); 2561 break; 2562 } 2563 } 2564 Pos += 2; 2565 } else { 2566 unsigned I = Pos + 1; 2567 2568 // Check for the \@ pseudo-variable. 2569 if (EnableAtPseudoVariable && Body[I] == '@' && I + 1 != End) 2570 ++I; 2571 else 2572 while (isIdentifierChar(Body[I]) && I + 1 != End) 2573 ++I; 2574 2575 const char *Begin = Body.data() + Pos + 1; 2576 StringRef Argument(Begin, I - (Pos + 1)); 2577 unsigned Index = 0; 2578 2579 if (Argument == "@") { 2580 OS << NumOfMacroInstantiations; 2581 Pos += 2; 2582 } else { 2583 for (; Index < NParameters; ++Index) 2584 if (Parameters[Index].Name == Argument) 2585 break; 2586 2587 if (Index == NParameters) { 2588 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 2589 Pos += 3; 2590 else { 2591 OS << '\\' << Argument; 2592 Pos = I; 2593 } 2594 } else { 2595 bool VarargParameter = HasVararg && Index == (NParameters - 1); 2596 for (const AsmToken &Token : A[Index]) 2597 // For altmacro mode, you can write '%expr'. 2598 // The prefix '%' evaluates the expression 'expr' 2599 // and uses the result as a string (e.g. replace %(1+2) with the 2600 // string "3"). 2601 // Here, we identify the integer token which is the result of the 2602 // absolute expression evaluation and replace it with its string 2603 // representation. 2604 if (AltMacroMode && Token.getString().front() == '%' && 2605 Token.is(AsmToken::Integer)) 2606 // Emit an integer value to the buffer. 2607 OS << Token.getIntVal(); 2608 // Only Token that was validated as a string and begins with '<' 2609 // is considered altMacroString!!! 2610 else if (AltMacroMode && Token.getString().front() == '<' && 2611 Token.is(AsmToken::String)) { 2612 OS << angleBracketString(Token.getStringContents()); 2613 } 2614 // We expect no quotes around the string's contents when 2615 // parsing for varargs. 2616 else if (Token.isNot(AsmToken::String) || VarargParameter) 2617 OS << Token.getString(); 2618 else 2619 OS << Token.getStringContents(); 2620 2621 Pos += 1 + Argument.size(); 2622 } 2623 } 2624 } 2625 // Update the scan point. 2626 Body = Body.substr(Pos); 2627 } 2628 2629 return false; 2630 } 2631 2632 static bool isOperator(AsmToken::TokenKind kind) { 2633 switch (kind) { 2634 default: 2635 return false; 2636 case AsmToken::Plus: 2637 case AsmToken::Minus: 2638 case AsmToken::Tilde: 2639 case AsmToken::Slash: 2640 case AsmToken::Star: 2641 case AsmToken::Dot: 2642 case AsmToken::Equal: 2643 case AsmToken::EqualEqual: 2644 case AsmToken::Pipe: 2645 case AsmToken::PipePipe: 2646 case AsmToken::Caret: 2647 case AsmToken::Amp: 2648 case AsmToken::AmpAmp: 2649 case AsmToken::Exclaim: 2650 case AsmToken::ExclaimEqual: 2651 case AsmToken::Less: 2652 case AsmToken::LessEqual: 2653 case AsmToken::LessLess: 2654 case AsmToken::LessGreater: 2655 case AsmToken::Greater: 2656 case AsmToken::GreaterEqual: 2657 case AsmToken::GreaterGreater: 2658 return true; 2659 } 2660 } 2661 2662 namespace { 2663 2664 class AsmLexerSkipSpaceRAII { 2665 public: 2666 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { 2667 Lexer.setSkipSpace(SkipSpace); 2668 } 2669 2670 ~AsmLexerSkipSpaceRAII() { 2671 Lexer.setSkipSpace(true); 2672 } 2673 2674 private: 2675 AsmLexer &Lexer; 2676 }; 2677 2678 } // end anonymous namespace 2679 2680 bool AsmParser::parseMacroArgument(MCAsmMacroArgument &MA, bool Vararg) { 2681 2682 if (Vararg) { 2683 if (Lexer.isNot(AsmToken::EndOfStatement)) { 2684 StringRef Str = parseStringToEndOfStatement(); 2685 MA.emplace_back(AsmToken::String, Str); 2686 } 2687 return false; 2688 } 2689 2690 unsigned ParenLevel = 0; 2691 2692 // Darwin doesn't use spaces to delmit arguments. 2693 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); 2694 2695 bool SpaceEaten; 2696 2697 while (true) { 2698 SpaceEaten = false; 2699 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 2700 return TokError("unexpected token in macro instantiation"); 2701 2702 if (ParenLevel == 0) { 2703 2704 if (Lexer.is(AsmToken::Comma)) 2705 break; 2706 2707 if (Lexer.is(AsmToken::Space)) { 2708 SpaceEaten = true; 2709 Lexer.Lex(); // Eat spaces 2710 } 2711 2712 // Spaces can delimit parameters, but could also be part an expression. 2713 // If the token after a space is an operator, add the token and the next 2714 // one into this argument 2715 if (!IsDarwin) { 2716 if (isOperator(Lexer.getKind())) { 2717 MA.push_back(getTok()); 2718 Lexer.Lex(); 2719 2720 // Whitespace after an operator can be ignored. 2721 if (Lexer.is(AsmToken::Space)) 2722 Lexer.Lex(); 2723 2724 continue; 2725 } 2726 } 2727 if (SpaceEaten) 2728 break; 2729 } 2730 2731 // handleMacroEntry relies on not advancing the lexer here 2732 // to be able to fill in the remaining default parameter values 2733 if (Lexer.is(AsmToken::EndOfStatement)) 2734 break; 2735 2736 // Adjust the current parentheses level. 2737 if (Lexer.is(AsmToken::LParen)) 2738 ++ParenLevel; 2739 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 2740 --ParenLevel; 2741 2742 // Append the token to the current argument list. 2743 MA.push_back(getTok()); 2744 Lexer.Lex(); 2745 } 2746 2747 if (ParenLevel != 0) 2748 return TokError("unbalanced parentheses in macro argument"); 2749 return false; 2750 } 2751 2752 // Parse the macro instantiation arguments. 2753 bool AsmParser::parseMacroArguments(const MCAsmMacro *M, 2754 MCAsmMacroArguments &A) { 2755 const unsigned NParameters = M ? M->Parameters.size() : 0; 2756 bool NamedParametersFound = false; 2757 SmallVector<SMLoc, 4> FALocs; 2758 2759 A.resize(NParameters); 2760 FALocs.resize(NParameters); 2761 2762 // Parse two kinds of macro invocations: 2763 // - macros defined without any parameters accept an arbitrary number of them 2764 // - macros defined with parameters accept at most that many of them 2765 bool HasVararg = NParameters ? M->Parameters.back().Vararg : false; 2766 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 2767 ++Parameter) { 2768 SMLoc IDLoc = Lexer.getLoc(); 2769 MCAsmMacroParameter FA; 2770 2771 if (Lexer.is(AsmToken::Identifier) && Lexer.peekTok().is(AsmToken::Equal)) { 2772 if (parseIdentifier(FA.Name)) 2773 return Error(IDLoc, "invalid argument identifier for formal argument"); 2774 2775 if (Lexer.isNot(AsmToken::Equal)) 2776 return TokError("expected '=' after formal parameter identifier"); 2777 2778 Lex(); 2779 2780 NamedParametersFound = true; 2781 } 2782 bool Vararg = HasVararg && Parameter == (NParameters - 1); 2783 2784 if (NamedParametersFound && FA.Name.empty()) 2785 return Error(IDLoc, "cannot mix positional and keyword arguments"); 2786 2787 SMLoc StrLoc = Lexer.getLoc(); 2788 SMLoc EndLoc; 2789 if (AltMacroMode && Lexer.is(AsmToken::Percent)) { 2790 const MCExpr *AbsoluteExp; 2791 int64_t Value; 2792 /// Eat '%' 2793 Lex(); 2794 if (parseExpression(AbsoluteExp, EndLoc)) 2795 return false; 2796 if (!AbsoluteExp->evaluateAsAbsolute(Value, 2797 getStreamer().getAssemblerPtr())) 2798 return Error(StrLoc, "expected absolute expression"); 2799 const char *StrChar = StrLoc.getPointer(); 2800 const char *EndChar = EndLoc.getPointer(); 2801 AsmToken newToken(AsmToken::Integer, 2802 StringRef(StrChar, EndChar - StrChar), Value); 2803 FA.Value.push_back(newToken); 2804 } else if (AltMacroMode && Lexer.is(AsmToken::Less) && 2805 isAngleBracketString(StrLoc, EndLoc)) { 2806 const char *StrChar = StrLoc.getPointer(); 2807 const char *EndChar = EndLoc.getPointer(); 2808 jumpToLoc(EndLoc, CurBuffer); 2809 /// Eat from '<' to '>' 2810 Lex(); 2811 AsmToken newToken(AsmToken::String, 2812 StringRef(StrChar, EndChar - StrChar)); 2813 FA.Value.push_back(newToken); 2814 } else if(parseMacroArgument(FA.Value, Vararg)) 2815 return true; 2816 2817 unsigned PI = Parameter; 2818 if (!FA.Name.empty()) { 2819 unsigned FAI = 0; 2820 for (FAI = 0; FAI < NParameters; ++FAI) 2821 if (M->Parameters[FAI].Name == FA.Name) 2822 break; 2823 2824 if (FAI >= NParameters) { 2825 assert(M && "expected macro to be defined"); 2826 return Error(IDLoc, "parameter named '" + FA.Name + 2827 "' does not exist for macro '" + M->Name + "'"); 2828 } 2829 PI = FAI; 2830 } 2831 2832 if (!FA.Value.empty()) { 2833 if (A.size() <= PI) 2834 A.resize(PI + 1); 2835 A[PI] = FA.Value; 2836 2837 if (FALocs.size() <= PI) 2838 FALocs.resize(PI + 1); 2839 2840 FALocs[PI] = Lexer.getLoc(); 2841 } 2842 2843 // At the end of the statement, fill in remaining arguments that have 2844 // default values. If there aren't any, then the next argument is 2845 // required but missing 2846 if (Lexer.is(AsmToken::EndOfStatement)) { 2847 bool Failure = false; 2848 for (unsigned FAI = 0; FAI < NParameters; ++FAI) { 2849 if (A[FAI].empty()) { 2850 if (M->Parameters[FAI].Required) { 2851 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), 2852 "missing value for required parameter " 2853 "'" + M->Parameters[FAI].Name + "' in macro '" + M->Name + "'"); 2854 Failure = true; 2855 } 2856 2857 if (!M->Parameters[FAI].Value.empty()) 2858 A[FAI] = M->Parameters[FAI].Value; 2859 } 2860 } 2861 return Failure; 2862 } 2863 2864 if (Lexer.is(AsmToken::Comma)) 2865 Lex(); 2866 } 2867 2868 return TokError("too many positional arguments"); 2869 } 2870 2871 bool AsmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc) { 2872 // Arbitrarily limit macro nesting depth (default matches 'as'). We can 2873 // eliminate this, although we should protect against infinite loops. 2874 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; 2875 if (ActiveMacros.size() == MaxNestingDepth) { 2876 std::ostringstream MaxNestingDepthError; 2877 MaxNestingDepthError << "macros cannot be nested more than " 2878 << MaxNestingDepth << " levels deep." 2879 << " Use -asm-macro-max-nesting-depth to increase " 2880 "this limit."; 2881 return TokError(MaxNestingDepthError.str()); 2882 } 2883 2884 MCAsmMacroArguments A; 2885 if (parseMacroArguments(M, A)) 2886 return true; 2887 2888 // Macro instantiation is lexical, unfortunately. We construct a new buffer 2889 // to hold the macro body with substitutions. 2890 SmallString<256> Buf; 2891 StringRef Body = M->Body; 2892 raw_svector_ostream OS(Buf); 2893 2894 if (expandMacro(OS, Body, M->Parameters, A, true, getTok().getLoc())) 2895 return true; 2896 2897 // We include the .endmacro in the buffer as our cue to exit the macro 2898 // instantiation. 2899 OS << ".endmacro\n"; 2900 2901 std::unique_ptr<MemoryBuffer> Instantiation = 2902 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 2903 2904 // Create the macro instantiation object and add to the current macro 2905 // instantiation stack. 2906 MacroInstantiation *MI = new MacroInstantiation{ 2907 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 2908 ActiveMacros.push_back(MI); 2909 2910 ++NumOfMacroInstantiations; 2911 2912 // Jump to the macro instantiation and prime the lexer. 2913 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 2914 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 2915 Lex(); 2916 2917 return false; 2918 } 2919 2920 void AsmParser::handleMacroExit() { 2921 // Jump to the EndOfStatement we should return to, and consume it. 2922 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer); 2923 Lex(); 2924 2925 // Pop the instantiation entry. 2926 delete ActiveMacros.back(); 2927 ActiveMacros.pop_back(); 2928 } 2929 2930 bool AsmParser::parseAssignment(StringRef Name, AssignmentKind Kind) { 2931 MCSymbol *Sym; 2932 const MCExpr *Value; 2933 SMLoc ExprLoc = getTok().getLoc(); 2934 bool AllowRedef = 2935 Kind == AssignmentKind::Set || Kind == AssignmentKind::Equal; 2936 if (MCParserUtils::parseAssignmentExpression(Name, AllowRedef, *this, Sym, 2937 Value)) 2938 return true; 2939 2940 if (!Sym) { 2941 // In the case where we parse an expression starting with a '.', we will 2942 // not generate an error, nor will we create a symbol. In this case we 2943 // should just return out. 2944 return false; 2945 } 2946 2947 if (discardLTOSymbol(Name)) 2948 return false; 2949 2950 // Do the assignment. 2951 switch (Kind) { 2952 case AssignmentKind::Equal: 2953 Out.emitAssignment(Sym, Value); 2954 break; 2955 case AssignmentKind::Set: 2956 case AssignmentKind::Equiv: 2957 Out.emitAssignment(Sym, Value); 2958 Out.emitSymbolAttribute(Sym, MCSA_NoDeadStrip); 2959 break; 2960 case AssignmentKind::LTOSetConditional: 2961 if (Value->getKind() != MCExpr::SymbolRef) 2962 return Error(ExprLoc, "expected identifier"); 2963 2964 Out.emitConditionalAssignment(Sym, Value); 2965 break; 2966 } 2967 2968 return false; 2969 } 2970 2971 /// parseIdentifier: 2972 /// ::= identifier 2973 /// ::= string 2974 bool AsmParser::parseIdentifier(StringRef &Res) { 2975 // The assembler has relaxed rules for accepting identifiers, in particular we 2976 // allow things like '.globl $foo' and '.def @feat.00', which would normally be 2977 // separate tokens. At this level, we have already lexed so we cannot (currently) 2978 // handle this as a context dependent token, instead we detect adjacent tokens 2979 // and return the combined identifier. 2980 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) { 2981 SMLoc PrefixLoc = getLexer().getLoc(); 2982 2983 // Consume the prefix character, and check for a following identifier. 2984 2985 AsmToken Buf[1]; 2986 Lexer.peekTokens(Buf, false); 2987 2988 if (Buf[0].isNot(AsmToken::Identifier) && Buf[0].isNot(AsmToken::Integer)) 2989 return true; 2990 2991 // We have a '$' or '@' followed by an identifier or integer token, make 2992 // sure they are adjacent. 2993 if (PrefixLoc.getPointer() + 1 != Buf[0].getLoc().getPointer()) 2994 return true; 2995 2996 // eat $ or @ 2997 Lexer.Lex(); // Lexer's Lex guarantees consecutive token. 2998 // Construct the joined identifier and consume the token. 2999 Res = StringRef(PrefixLoc.getPointer(), getTok().getString().size() + 1); 3000 Lex(); // Parser Lex to maintain invariants. 3001 return false; 3002 } 3003 3004 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String)) 3005 return true; 3006 3007 Res = getTok().getIdentifier(); 3008 3009 Lex(); // Consume the identifier token. 3010 3011 return false; 3012 } 3013 3014 /// parseDirectiveSet: 3015 /// ::= .equ identifier ',' expression 3016 /// ::= .equiv identifier ',' expression 3017 /// ::= .set identifier ',' expression 3018 /// ::= .lto_set_conditional identifier ',' expression 3019 bool AsmParser::parseDirectiveSet(StringRef IDVal, AssignmentKind Kind) { 3020 StringRef Name; 3021 if (check(parseIdentifier(Name), "expected identifier") || parseComma() || 3022 parseAssignment(Name, Kind)) 3023 return true; 3024 return false; 3025 } 3026 3027 bool AsmParser::parseEscapedString(std::string &Data) { 3028 if (check(getTok().isNot(AsmToken::String), "expected string")) 3029 return true; 3030 3031 Data = ""; 3032 StringRef Str = getTok().getStringContents(); 3033 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 3034 if (Str[i] != '\\') { 3035 Data += Str[i]; 3036 continue; 3037 } 3038 3039 // Recognize escaped characters. Note that this escape semantics currently 3040 // loosely follows Darwin 'as'. 3041 ++i; 3042 if (i == e) 3043 return TokError("unexpected backslash at end of string"); 3044 3045 // Recognize hex sequences similarly to GNU 'as'. 3046 if (Str[i] == 'x' || Str[i] == 'X') { 3047 size_t length = Str.size(); 3048 if (i + 1 >= length || !isHexDigit(Str[i + 1])) 3049 return TokError("invalid hexadecimal escape sequence"); 3050 3051 // Consume hex characters. GNU 'as' reads all hexadecimal characters and 3052 // then truncates to the lower 16 bits. Seems reasonable. 3053 unsigned Value = 0; 3054 while (i + 1 < length && isHexDigit(Str[i + 1])) 3055 Value = Value * 16 + hexDigitValue(Str[++i]); 3056 3057 Data += (unsigned char)(Value & 0xFF); 3058 continue; 3059 } 3060 3061 // Recognize octal sequences. 3062 if ((unsigned)(Str[i] - '0') <= 7) { 3063 // Consume up to three octal characters. 3064 unsigned Value = Str[i] - '0'; 3065 3066 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3067 ++i; 3068 Value = Value * 8 + (Str[i] - '0'); 3069 3070 if (i + 1 != e && ((unsigned)(Str[i + 1] - '0')) <= 7) { 3071 ++i; 3072 Value = Value * 8 + (Str[i] - '0'); 3073 } 3074 } 3075 3076 if (Value > 255) 3077 return TokError("invalid octal escape sequence (out of range)"); 3078 3079 Data += (unsigned char)Value; 3080 continue; 3081 } 3082 3083 // Otherwise recognize individual escapes. 3084 switch (Str[i]) { 3085 default: 3086 // Just reject invalid escape sequences for now. 3087 return TokError("invalid escape sequence (unrecognized character)"); 3088 3089 case 'b': Data += '\b'; break; 3090 case 'f': Data += '\f'; break; 3091 case 'n': Data += '\n'; break; 3092 case 'r': Data += '\r'; break; 3093 case 't': Data += '\t'; break; 3094 case '"': Data += '"'; break; 3095 case '\\': Data += '\\'; break; 3096 } 3097 } 3098 3099 Lex(); 3100 return false; 3101 } 3102 3103 bool AsmParser::parseAngleBracketString(std::string &Data) { 3104 SMLoc EndLoc, StartLoc = getTok().getLoc(); 3105 if (isAngleBracketString(StartLoc, EndLoc)) { 3106 const char *StartChar = StartLoc.getPointer() + 1; 3107 const char *EndChar = EndLoc.getPointer() - 1; 3108 jumpToLoc(EndLoc, CurBuffer); 3109 /// Eat from '<' to '>' 3110 Lex(); 3111 3112 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar)); 3113 return false; 3114 } 3115 return true; 3116 } 3117 3118 /// parseDirectiveAscii: 3119 // ::= .ascii [ "string"+ ( , "string"+ )* ] 3120 /// ::= ( .asciz | .string ) [ "string" ( , "string" )* ] 3121 bool AsmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 3122 auto parseOp = [&]() -> bool { 3123 std::string Data; 3124 if (checkForValidSection()) 3125 return true; 3126 // Only support spaces as separators for .ascii directive for now. See the 3127 // discusssion at https://reviews.llvm.org/D91460 for more details. 3128 do { 3129 if (parseEscapedString(Data)) 3130 return true; 3131 getStreamer().emitBytes(Data); 3132 } while (!ZeroTerminated && getTok().is(AsmToken::String)); 3133 if (ZeroTerminated) 3134 getStreamer().emitBytes(StringRef("\0", 1)); 3135 return false; 3136 }; 3137 3138 return parseMany(parseOp); 3139 } 3140 3141 /// parseDirectiveReloc 3142 /// ::= .reloc expression , identifier [ , expression ] 3143 bool AsmParser::parseDirectiveReloc(SMLoc DirectiveLoc) { 3144 const MCExpr *Offset; 3145 const MCExpr *Expr = nullptr; 3146 SMLoc OffsetLoc = Lexer.getTok().getLoc(); 3147 3148 if (parseExpression(Offset)) 3149 return true; 3150 if (parseComma() || 3151 check(getTok().isNot(AsmToken::Identifier), "expected relocation name")) 3152 return true; 3153 3154 SMLoc NameLoc = Lexer.getTok().getLoc(); 3155 StringRef Name = Lexer.getTok().getIdentifier(); 3156 Lex(); 3157 3158 if (Lexer.is(AsmToken::Comma)) { 3159 Lex(); 3160 SMLoc ExprLoc = Lexer.getLoc(); 3161 if (parseExpression(Expr)) 3162 return true; 3163 3164 MCValue Value; 3165 if (!Expr->evaluateAsRelocatable(Value, nullptr, nullptr)) 3166 return Error(ExprLoc, "expression must be relocatable"); 3167 } 3168 3169 if (parseEOL()) 3170 return true; 3171 3172 const MCTargetAsmParser &MCT = getTargetParser(); 3173 const MCSubtargetInfo &STI = MCT.getSTI(); 3174 if (Optional<std::pair<bool, std::string>> Err = 3175 getStreamer().emitRelocDirective(*Offset, Name, Expr, DirectiveLoc, 3176 STI)) 3177 return Error(Err->first ? NameLoc : OffsetLoc, Err->second); 3178 3179 return false; 3180 } 3181 3182 /// parseDirectiveValue 3183 /// ::= (.byte | .short | ... ) [ expression (, expression)* ] 3184 bool AsmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { 3185 auto parseOp = [&]() -> bool { 3186 const MCExpr *Value; 3187 SMLoc ExprLoc = getLexer().getLoc(); 3188 if (checkForValidSection() || parseExpression(Value)) 3189 return true; 3190 // Special case constant expressions to match code generator. 3191 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3192 assert(Size <= 8 && "Invalid size"); 3193 uint64_t IntValue = MCE->getValue(); 3194 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 3195 return Error(ExprLoc, "out of range literal value"); 3196 getStreamer().emitIntValue(IntValue, Size); 3197 } else 3198 getStreamer().emitValue(Value, Size, ExprLoc); 3199 return false; 3200 }; 3201 3202 return parseMany(parseOp); 3203 } 3204 3205 static bool parseHexOcta(AsmParser &Asm, uint64_t &hi, uint64_t &lo) { 3206 if (Asm.getTok().isNot(AsmToken::Integer) && 3207 Asm.getTok().isNot(AsmToken::BigNum)) 3208 return Asm.TokError("unknown token in expression"); 3209 SMLoc ExprLoc = Asm.getTok().getLoc(); 3210 APInt IntValue = Asm.getTok().getAPIntVal(); 3211 Asm.Lex(); 3212 if (!IntValue.isIntN(128)) 3213 return Asm.Error(ExprLoc, "out of range literal value"); 3214 if (!IntValue.isIntN(64)) { 3215 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue(); 3216 lo = IntValue.getLoBits(64).getZExtValue(); 3217 } else { 3218 hi = 0; 3219 lo = IntValue.getZExtValue(); 3220 } 3221 return false; 3222 } 3223 3224 /// ParseDirectiveOctaValue 3225 /// ::= .octa [ hexconstant (, hexconstant)* ] 3226 3227 bool AsmParser::parseDirectiveOctaValue(StringRef IDVal) { 3228 auto parseOp = [&]() -> bool { 3229 if (checkForValidSection()) 3230 return true; 3231 uint64_t hi, lo; 3232 if (parseHexOcta(*this, hi, lo)) 3233 return true; 3234 if (MAI.isLittleEndian()) { 3235 getStreamer().emitInt64(lo); 3236 getStreamer().emitInt64(hi); 3237 } else { 3238 getStreamer().emitInt64(hi); 3239 getStreamer().emitInt64(lo); 3240 } 3241 return false; 3242 }; 3243 3244 return parseMany(parseOp); 3245 } 3246 3247 bool AsmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { 3248 // We don't truly support arithmetic on floating point expressions, so we 3249 // have to manually parse unary prefixes. 3250 bool IsNeg = false; 3251 if (getLexer().is(AsmToken::Minus)) { 3252 Lexer.Lex(); 3253 IsNeg = true; 3254 } else if (getLexer().is(AsmToken::Plus)) 3255 Lexer.Lex(); 3256 3257 if (Lexer.is(AsmToken::Error)) 3258 return TokError(Lexer.getErr()); 3259 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) && 3260 Lexer.isNot(AsmToken::Identifier)) 3261 return TokError("unexpected token in directive"); 3262 3263 // Convert to an APFloat. 3264 APFloat Value(Semantics); 3265 StringRef IDVal = getTok().getString(); 3266 if (getLexer().is(AsmToken::Identifier)) { 3267 if (!IDVal.compare_insensitive("infinity") || 3268 !IDVal.compare_insensitive("inf")) 3269 Value = APFloat::getInf(Semantics); 3270 else if (!IDVal.compare_insensitive("nan")) 3271 Value = APFloat::getNaN(Semantics, false, ~0); 3272 else 3273 return TokError("invalid floating point literal"); 3274 } else if (errorToBool( 3275 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) 3276 .takeError())) 3277 return TokError("invalid floating point literal"); 3278 if (IsNeg) 3279 Value.changeSign(); 3280 3281 // Consume the numeric token. 3282 Lex(); 3283 3284 Res = Value.bitcastToAPInt(); 3285 3286 return false; 3287 } 3288 3289 /// parseDirectiveRealValue 3290 /// ::= (.single | .double) [ expression (, expression)* ] 3291 bool AsmParser::parseDirectiveRealValue(StringRef IDVal, 3292 const fltSemantics &Semantics) { 3293 auto parseOp = [&]() -> bool { 3294 APInt AsInt; 3295 if (checkForValidSection() || parseRealValue(Semantics, AsInt)) 3296 return true; 3297 getStreamer().emitIntValue(AsInt.getLimitedValue(), 3298 AsInt.getBitWidth() / 8); 3299 return false; 3300 }; 3301 3302 return parseMany(parseOp); 3303 } 3304 3305 /// parseDirectiveZero 3306 /// ::= .zero expression 3307 bool AsmParser::parseDirectiveZero() { 3308 SMLoc NumBytesLoc = Lexer.getLoc(); 3309 const MCExpr *NumBytes; 3310 if (checkForValidSection() || parseExpression(NumBytes)) 3311 return true; 3312 3313 int64_t Val = 0; 3314 if (getLexer().is(AsmToken::Comma)) { 3315 Lex(); 3316 if (parseAbsoluteExpression(Val)) 3317 return true; 3318 } 3319 3320 if (parseEOL()) 3321 return true; 3322 getStreamer().emitFill(*NumBytes, Val, NumBytesLoc); 3323 3324 return false; 3325 } 3326 3327 /// parseDirectiveFill 3328 /// ::= .fill expression [ , expression [ , expression ] ] 3329 bool AsmParser::parseDirectiveFill() { 3330 SMLoc NumValuesLoc = Lexer.getLoc(); 3331 const MCExpr *NumValues; 3332 if (checkForValidSection() || parseExpression(NumValues)) 3333 return true; 3334 3335 int64_t FillSize = 1; 3336 int64_t FillExpr = 0; 3337 3338 SMLoc SizeLoc, ExprLoc; 3339 3340 if (parseOptionalToken(AsmToken::Comma)) { 3341 SizeLoc = getTok().getLoc(); 3342 if (parseAbsoluteExpression(FillSize)) 3343 return true; 3344 if (parseOptionalToken(AsmToken::Comma)) { 3345 ExprLoc = getTok().getLoc(); 3346 if (parseAbsoluteExpression(FillExpr)) 3347 return true; 3348 } 3349 } 3350 if (parseEOL()) 3351 return true; 3352 3353 if (FillSize < 0) { 3354 Warning(SizeLoc, "'.fill' directive with negative size has no effect"); 3355 return false; 3356 } 3357 if (FillSize > 8) { 3358 Warning(SizeLoc, "'.fill' directive with size greater than 8 has been truncated to 8"); 3359 FillSize = 8; 3360 } 3361 3362 if (!isUInt<32>(FillExpr) && FillSize > 4) 3363 Warning(ExprLoc, "'.fill' directive pattern has been truncated to 32-bits"); 3364 3365 getStreamer().emitFill(*NumValues, FillSize, FillExpr, NumValuesLoc); 3366 3367 return false; 3368 } 3369 3370 /// parseDirectiveOrg 3371 /// ::= .org expression [ , expression ] 3372 bool AsmParser::parseDirectiveOrg() { 3373 const MCExpr *Offset; 3374 SMLoc OffsetLoc = Lexer.getLoc(); 3375 if (checkForValidSection() || parseExpression(Offset)) 3376 return true; 3377 3378 // Parse optional fill expression. 3379 int64_t FillExpr = 0; 3380 if (parseOptionalToken(AsmToken::Comma)) 3381 if (parseAbsoluteExpression(FillExpr)) 3382 return true; 3383 if (parseEOL()) 3384 return true; 3385 3386 getStreamer().emitValueToOffset(Offset, FillExpr, OffsetLoc); 3387 return false; 3388 } 3389 3390 /// parseDirectiveAlign 3391 /// ::= {.align, ...} expression [ , expression [ , expression ]] 3392 bool AsmParser::parseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 3393 SMLoc AlignmentLoc = getLexer().getLoc(); 3394 int64_t Alignment; 3395 SMLoc MaxBytesLoc; 3396 bool HasFillExpr = false; 3397 int64_t FillExpr = 0; 3398 int64_t MaxBytesToFill = 0; 3399 3400 auto parseAlign = [&]() -> bool { 3401 if (parseAbsoluteExpression(Alignment)) 3402 return true; 3403 if (parseOptionalToken(AsmToken::Comma)) { 3404 // The fill expression can be omitted while specifying a maximum number of 3405 // alignment bytes, e.g: 3406 // .align 3,,4 3407 if (getTok().isNot(AsmToken::Comma)) { 3408 HasFillExpr = true; 3409 if (parseAbsoluteExpression(FillExpr)) 3410 return true; 3411 } 3412 if (parseOptionalToken(AsmToken::Comma)) 3413 if (parseTokenLoc(MaxBytesLoc) || 3414 parseAbsoluteExpression(MaxBytesToFill)) 3415 return true; 3416 } 3417 return parseEOL(); 3418 }; 3419 3420 if (checkForValidSection()) 3421 return true; 3422 // Ignore empty '.p2align' directives for GNU-as compatibility 3423 if (IsPow2 && (ValueSize == 1) && getTok().is(AsmToken::EndOfStatement)) { 3424 Warning(AlignmentLoc, "p2align directive with no operand(s) is ignored"); 3425 return parseEOL(); 3426 } 3427 if (parseAlign()) 3428 return true; 3429 3430 // Always emit an alignment here even if we thrown an error. 3431 bool ReturnVal = false; 3432 3433 // Compute alignment in bytes. 3434 if (IsPow2) { 3435 // FIXME: Diagnose overflow. 3436 if (Alignment >= 32) { 3437 ReturnVal |= Error(AlignmentLoc, "invalid alignment value"); 3438 Alignment = 31; 3439 } 3440 3441 Alignment = 1ULL << Alignment; 3442 } else { 3443 // Reject alignments that aren't either a power of two or zero, 3444 // for gas compatibility. Alignment of zero is silently rounded 3445 // up to one. 3446 if (Alignment == 0) 3447 Alignment = 1; 3448 if (!isPowerOf2_64(Alignment)) 3449 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2"); 3450 if (!isUInt<32>(Alignment)) 3451 ReturnVal |= Error(AlignmentLoc, "alignment must be smaller than 2**32"); 3452 } 3453 3454 // Diagnose non-sensical max bytes to align. 3455 if (MaxBytesLoc.isValid()) { 3456 if (MaxBytesToFill < 1) { 3457 ReturnVal |= Error(MaxBytesLoc, 3458 "alignment directive can never be satisfied in this " 3459 "many bytes, ignoring maximum bytes expression"); 3460 MaxBytesToFill = 0; 3461 } 3462 3463 if (MaxBytesToFill >= Alignment) { 3464 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 3465 "has no effect"); 3466 MaxBytesToFill = 0; 3467 } 3468 } 3469 3470 // Check whether we should use optimal code alignment for this .align 3471 // directive. 3472 const MCSection *Section = getStreamer().getCurrentSectionOnly(); 3473 assert(Section && "must have section to emit alignment"); 3474 bool UseCodeAlign = Section->UseCodeAlign(); 3475 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 3476 ValueSize == 1 && UseCodeAlign) { 3477 getStreamer().emitCodeAlignment(Alignment, &getTargetParser().getSTI(), 3478 MaxBytesToFill); 3479 } else { 3480 // FIXME: Target specific behavior about how the "extra" bytes are filled. 3481 getStreamer().emitValueToAlignment(Alignment, FillExpr, ValueSize, 3482 MaxBytesToFill); 3483 } 3484 3485 return ReturnVal; 3486 } 3487 3488 /// parseDirectiveFile 3489 /// ::= .file filename 3490 /// ::= .file number [directory] filename [md5 checksum] [source source-text] 3491 bool AsmParser::parseDirectiveFile(SMLoc DirectiveLoc) { 3492 // FIXME: I'm not sure what this is. 3493 int64_t FileNumber = -1; 3494 if (getLexer().is(AsmToken::Integer)) { 3495 FileNumber = getTok().getIntVal(); 3496 Lex(); 3497 3498 if (FileNumber < 0) 3499 return TokError("negative file number"); 3500 } 3501 3502 std::string Path; 3503 3504 // Usually the directory and filename together, otherwise just the directory. 3505 // Allow the strings to have escaped octal character sequence. 3506 if (parseEscapedString(Path)) 3507 return true; 3508 3509 StringRef Directory; 3510 StringRef Filename; 3511 std::string FilenameData; 3512 if (getLexer().is(AsmToken::String)) { 3513 if (check(FileNumber == -1, 3514 "explicit path specified, but no file number") || 3515 parseEscapedString(FilenameData)) 3516 return true; 3517 Filename = FilenameData; 3518 Directory = Path; 3519 } else { 3520 Filename = Path; 3521 } 3522 3523 uint64_t MD5Hi, MD5Lo; 3524 bool HasMD5 = false; 3525 3526 Optional<StringRef> Source; 3527 bool HasSource = false; 3528 std::string SourceString; 3529 3530 while (!parseOptionalToken(AsmToken::EndOfStatement)) { 3531 StringRef Keyword; 3532 if (check(getTok().isNot(AsmToken::Identifier), 3533 "unexpected token in '.file' directive") || 3534 parseIdentifier(Keyword)) 3535 return true; 3536 if (Keyword == "md5") { 3537 HasMD5 = true; 3538 if (check(FileNumber == -1, 3539 "MD5 checksum specified, but no file number") || 3540 parseHexOcta(*this, MD5Hi, MD5Lo)) 3541 return true; 3542 } else if (Keyword == "source") { 3543 HasSource = true; 3544 if (check(FileNumber == -1, 3545 "source specified, but no file number") || 3546 check(getTok().isNot(AsmToken::String), 3547 "unexpected token in '.file' directive") || 3548 parseEscapedString(SourceString)) 3549 return true; 3550 } else { 3551 return TokError("unexpected token in '.file' directive"); 3552 } 3553 } 3554 3555 if (FileNumber == -1) { 3556 // Ignore the directive if there is no number and the target doesn't support 3557 // numberless .file directives. This allows some portability of assembler 3558 // between different object file formats. 3559 if (getContext().getAsmInfo()->hasSingleParameterDotFile()) 3560 getStreamer().emitFileDirective(Filename); 3561 } else { 3562 // In case there is a -g option as well as debug info from directive .file, 3563 // we turn off the -g option, directly use the existing debug info instead. 3564 // Throw away any implicit file table for the assembler source. 3565 if (Ctx.getGenDwarfForAssembly()) { 3566 Ctx.getMCDwarfLineTable(0).resetFileTable(); 3567 Ctx.setGenDwarfForAssembly(false); 3568 } 3569 3570 Optional<MD5::MD5Result> CKMem; 3571 if (HasMD5) { 3572 MD5::MD5Result Sum; 3573 for (unsigned i = 0; i != 8; ++i) { 3574 Sum.Bytes[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); 3575 Sum.Bytes[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); 3576 } 3577 CKMem = Sum; 3578 } 3579 if (HasSource) { 3580 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size())); 3581 memcpy(SourceBuf, SourceString.data(), SourceString.size()); 3582 Source = StringRef(SourceBuf, SourceString.size()); 3583 } 3584 if (FileNumber == 0) { 3585 // Upgrade to Version 5 for assembly actions like clang -c a.s. 3586 if (Ctx.getDwarfVersion() < 5) 3587 Ctx.setDwarfVersion(5); 3588 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source); 3589 } else { 3590 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( 3591 FileNumber, Directory, Filename, CKMem, Source); 3592 if (!FileNumOrErr) 3593 return Error(DirectiveLoc, toString(FileNumOrErr.takeError())); 3594 } 3595 // Alert the user if there are some .file directives with MD5 and some not. 3596 // But only do that once. 3597 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) { 3598 ReportedInconsistentMD5 = true; 3599 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums"); 3600 } 3601 } 3602 3603 return false; 3604 } 3605 3606 /// parseDirectiveLine 3607 /// ::= .line [number] 3608 bool AsmParser::parseDirectiveLine() { 3609 int64_t LineNumber; 3610 if (getLexer().is(AsmToken::Integer)) { 3611 if (parseIntToken(LineNumber, "unexpected token in '.line' directive")) 3612 return true; 3613 (void)LineNumber; 3614 // FIXME: Do something with the .line. 3615 } 3616 return parseEOL(); 3617 } 3618 3619 /// parseDirectiveLoc 3620 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 3621 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 3622 /// The first number is a file number, must have been previously assigned with 3623 /// a .file directive, the second number is the line number and optionally the 3624 /// third number is a column position (zero if not specified). The remaining 3625 /// optional items are .loc sub-directives. 3626 bool AsmParser::parseDirectiveLoc() { 3627 int64_t FileNumber = 0, LineNumber = 0; 3628 SMLoc Loc = getTok().getLoc(); 3629 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") || 3630 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, 3631 "file number less than one in '.loc' directive") || 3632 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc, 3633 "unassigned file number in '.loc' directive")) 3634 return true; 3635 3636 // optional 3637 if (getLexer().is(AsmToken::Integer)) { 3638 LineNumber = getTok().getIntVal(); 3639 if (LineNumber < 0) 3640 return TokError("line number less than zero in '.loc' directive"); 3641 Lex(); 3642 } 3643 3644 int64_t ColumnPos = 0; 3645 if (getLexer().is(AsmToken::Integer)) { 3646 ColumnPos = getTok().getIntVal(); 3647 if (ColumnPos < 0) 3648 return TokError("column position less than zero in '.loc' directive"); 3649 Lex(); 3650 } 3651 3652 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags(); 3653 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT; 3654 unsigned Isa = 0; 3655 int64_t Discriminator = 0; 3656 3657 auto parseLocOp = [&]() -> bool { 3658 StringRef Name; 3659 SMLoc Loc = getTok().getLoc(); 3660 if (parseIdentifier(Name)) 3661 return TokError("unexpected token in '.loc' directive"); 3662 3663 if (Name == "basic_block") 3664 Flags |= DWARF2_FLAG_BASIC_BLOCK; 3665 else if (Name == "prologue_end") 3666 Flags |= DWARF2_FLAG_PROLOGUE_END; 3667 else if (Name == "epilogue_begin") 3668 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 3669 else if (Name == "is_stmt") { 3670 Loc = getTok().getLoc(); 3671 const MCExpr *Value; 3672 if (parseExpression(Value)) 3673 return true; 3674 // The expression must be the constant 0 or 1. 3675 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3676 int Value = MCE->getValue(); 3677 if (Value == 0) 3678 Flags &= ~DWARF2_FLAG_IS_STMT; 3679 else if (Value == 1) 3680 Flags |= DWARF2_FLAG_IS_STMT; 3681 else 3682 return Error(Loc, "is_stmt value not 0 or 1"); 3683 } else { 3684 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 3685 } 3686 } else if (Name == "isa") { 3687 Loc = getTok().getLoc(); 3688 const MCExpr *Value; 3689 if (parseExpression(Value)) 3690 return true; 3691 // The expression must be a constant greater or equal to 0. 3692 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 3693 int Value = MCE->getValue(); 3694 if (Value < 0) 3695 return Error(Loc, "isa number less than zero"); 3696 Isa = Value; 3697 } else { 3698 return Error(Loc, "isa number not a constant value"); 3699 } 3700 } else if (Name == "discriminator") { 3701 if (parseAbsoluteExpression(Discriminator)) 3702 return true; 3703 } else { 3704 return Error(Loc, "unknown sub-directive in '.loc' directive"); 3705 } 3706 return false; 3707 }; 3708 3709 if (parseMany(parseLocOp, false /*hasComma*/)) 3710 return true; 3711 3712 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 3713 Isa, Discriminator, StringRef()); 3714 3715 return false; 3716 } 3717 3718 /// parseDirectiveStabs 3719 /// ::= .stabs string, number, number, number 3720 bool AsmParser::parseDirectiveStabs() { 3721 return TokError("unsupported directive '.stabs'"); 3722 } 3723 3724 /// parseDirectiveCVFile 3725 /// ::= .cv_file number filename [checksum] [checksumkind] 3726 bool AsmParser::parseDirectiveCVFile() { 3727 SMLoc FileNumberLoc = getTok().getLoc(); 3728 int64_t FileNumber; 3729 std::string Filename; 3730 std::string Checksum; 3731 int64_t ChecksumKind = 0; 3732 3733 if (parseIntToken(FileNumber, 3734 "expected file number in '.cv_file' directive") || 3735 check(FileNumber < 1, FileNumberLoc, "file number less than one") || 3736 check(getTok().isNot(AsmToken::String), 3737 "unexpected token in '.cv_file' directive") || 3738 parseEscapedString(Filename)) 3739 return true; 3740 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 3741 if (check(getTok().isNot(AsmToken::String), 3742 "unexpected token in '.cv_file' directive") || 3743 parseEscapedString(Checksum) || 3744 parseIntToken(ChecksumKind, 3745 "expected checksum kind in '.cv_file' directive") || 3746 parseToken(AsmToken::EndOfStatement, 3747 "unexpected token in '.cv_file' directive")) 3748 return true; 3749 } 3750 3751 Checksum = fromHex(Checksum); 3752 void *CKMem = Ctx.allocate(Checksum.size(), 1); 3753 memcpy(CKMem, Checksum.data(), Checksum.size()); 3754 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), 3755 Checksum.size()); 3756 3757 if (!getStreamer().EmitCVFileDirective(FileNumber, Filename, ChecksumAsBytes, 3758 static_cast<uint8_t>(ChecksumKind))) 3759 return Error(FileNumberLoc, "file number already allocated"); 3760 3761 return false; 3762 } 3763 3764 bool AsmParser::parseCVFunctionId(int64_t &FunctionId, 3765 StringRef DirectiveName) { 3766 SMLoc Loc; 3767 return parseTokenLoc(Loc) || 3768 parseIntToken(FunctionId, "expected function id in '" + DirectiveName + 3769 "' directive") || 3770 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc, 3771 "expected function id within range [0, UINT_MAX)"); 3772 } 3773 3774 bool AsmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { 3775 SMLoc Loc; 3776 return parseTokenLoc(Loc) || 3777 parseIntToken(FileNumber, "expected integer in '" + DirectiveName + 3778 "' directive") || 3779 check(FileNumber < 1, Loc, "file number less than one in '" + 3780 DirectiveName + "' directive") || 3781 check(!getCVContext().isValidFileNumber(FileNumber), Loc, 3782 "unassigned file number in '" + DirectiveName + "' directive"); 3783 } 3784 3785 /// parseDirectiveCVFuncId 3786 /// ::= .cv_func_id FunctionId 3787 /// 3788 /// Introduces a function ID that can be used with .cv_loc. 3789 bool AsmParser::parseDirectiveCVFuncId() { 3790 SMLoc FunctionIdLoc = getTok().getLoc(); 3791 int64_t FunctionId; 3792 3793 if (parseCVFunctionId(FunctionId, ".cv_func_id") || 3794 parseToken(AsmToken::EndOfStatement, 3795 "unexpected token in '.cv_func_id' directive")) 3796 return true; 3797 3798 if (!getStreamer().EmitCVFuncIdDirective(FunctionId)) 3799 return Error(FunctionIdLoc, "function id already allocated"); 3800 3801 return false; 3802 } 3803 3804 /// parseDirectiveCVInlineSiteId 3805 /// ::= .cv_inline_site_id FunctionId 3806 /// "within" IAFunc 3807 /// "inlined_at" IAFile IALine [IACol] 3808 /// 3809 /// Introduces a function ID that can be used with .cv_loc. Includes "inlined 3810 /// at" source location information for use in the line table of the caller, 3811 /// whether the caller is a real function or another inlined call site. 3812 bool AsmParser::parseDirectiveCVInlineSiteId() { 3813 SMLoc FunctionIdLoc = getTok().getLoc(); 3814 int64_t FunctionId; 3815 int64_t IAFunc; 3816 int64_t IAFile; 3817 int64_t IALine; 3818 int64_t IACol = 0; 3819 3820 // FunctionId 3821 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id")) 3822 return true; 3823 3824 // "within" 3825 if (check((getLexer().isNot(AsmToken::Identifier) || 3826 getTok().getIdentifier() != "within"), 3827 "expected 'within' identifier in '.cv_inline_site_id' directive")) 3828 return true; 3829 Lex(); 3830 3831 // IAFunc 3832 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id")) 3833 return true; 3834 3835 // "inlined_at" 3836 if (check((getLexer().isNot(AsmToken::Identifier) || 3837 getTok().getIdentifier() != "inlined_at"), 3838 "expected 'inlined_at' identifier in '.cv_inline_site_id' " 3839 "directive") ) 3840 return true; 3841 Lex(); 3842 3843 // IAFile IALine 3844 if (parseCVFileId(IAFile, ".cv_inline_site_id") || 3845 parseIntToken(IALine, "expected line number after 'inlined_at'")) 3846 return true; 3847 3848 // [IACol] 3849 if (getLexer().is(AsmToken::Integer)) { 3850 IACol = getTok().getIntVal(); 3851 Lex(); 3852 } 3853 3854 if (parseToken(AsmToken::EndOfStatement, 3855 "unexpected token in '.cv_inline_site_id' directive")) 3856 return true; 3857 3858 if (!getStreamer().EmitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, 3859 IALine, IACol, FunctionIdLoc)) 3860 return Error(FunctionIdLoc, "function id already allocated"); 3861 3862 return false; 3863 } 3864 3865 /// parseDirectiveCVLoc 3866 /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] 3867 /// [is_stmt VALUE] 3868 /// The first number is a file number, must have been previously assigned with 3869 /// a .file directive, the second number is the line number and optionally the 3870 /// third number is a column position (zero if not specified). The remaining 3871 /// optional items are .loc sub-directives. 3872 bool AsmParser::parseDirectiveCVLoc() { 3873 SMLoc DirectiveLoc = getTok().getLoc(); 3874 int64_t FunctionId, FileNumber; 3875 if (parseCVFunctionId(FunctionId, ".cv_loc") || 3876 parseCVFileId(FileNumber, ".cv_loc")) 3877 return true; 3878 3879 int64_t LineNumber = 0; 3880 if (getLexer().is(AsmToken::Integer)) { 3881 LineNumber = getTok().getIntVal(); 3882 if (LineNumber < 0) 3883 return TokError("line number less than zero in '.cv_loc' directive"); 3884 Lex(); 3885 } 3886 3887 int64_t ColumnPos = 0; 3888 if (getLexer().is(AsmToken::Integer)) { 3889 ColumnPos = getTok().getIntVal(); 3890 if (ColumnPos < 0) 3891 return TokError("column position less than zero in '.cv_loc' directive"); 3892 Lex(); 3893 } 3894 3895 bool PrologueEnd = false; 3896 uint64_t IsStmt = 0; 3897 3898 auto parseOp = [&]() -> bool { 3899 StringRef Name; 3900 SMLoc Loc = getTok().getLoc(); 3901 if (parseIdentifier(Name)) 3902 return TokError("unexpected token in '.cv_loc' directive"); 3903 if (Name == "prologue_end") 3904 PrologueEnd = true; 3905 else if (Name == "is_stmt") { 3906 Loc = getTok().getLoc(); 3907 const MCExpr *Value; 3908 if (parseExpression(Value)) 3909 return true; 3910 // The expression must be the constant 0 or 1. 3911 IsStmt = ~0ULL; 3912 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value)) 3913 IsStmt = MCE->getValue(); 3914 3915 if (IsStmt > 1) 3916 return Error(Loc, "is_stmt value not 0 or 1"); 3917 } else { 3918 return Error(Loc, "unknown sub-directive in '.cv_loc' directive"); 3919 } 3920 return false; 3921 }; 3922 3923 if (parseMany(parseOp, false /*hasComma*/)) 3924 return true; 3925 3926 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber, 3927 ColumnPos, PrologueEnd, IsStmt, StringRef(), 3928 DirectiveLoc); 3929 return false; 3930 } 3931 3932 /// parseDirectiveCVLinetable 3933 /// ::= .cv_linetable FunctionId, FnStart, FnEnd 3934 bool AsmParser::parseDirectiveCVLinetable() { 3935 int64_t FunctionId; 3936 StringRef FnStartName, FnEndName; 3937 SMLoc Loc = getTok().getLoc(); 3938 if (parseCVFunctionId(FunctionId, ".cv_linetable") || parseComma() || 3939 parseTokenLoc(Loc) || 3940 check(parseIdentifier(FnStartName), Loc, 3941 "expected identifier in directive") || 3942 parseComma() || parseTokenLoc(Loc) || 3943 check(parseIdentifier(FnEndName), Loc, 3944 "expected identifier in directive")) 3945 return true; 3946 3947 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3948 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3949 3950 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym); 3951 return false; 3952 } 3953 3954 /// parseDirectiveCVInlineLinetable 3955 /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd 3956 bool AsmParser::parseDirectiveCVInlineLinetable() { 3957 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; 3958 StringRef FnStartName, FnEndName; 3959 SMLoc Loc = getTok().getLoc(); 3960 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") || 3961 parseTokenLoc(Loc) || 3962 parseIntToken( 3963 SourceFileId, 3964 "expected SourceField in '.cv_inline_linetable' directive") || 3965 check(SourceFileId <= 0, Loc, 3966 "File id less than zero in '.cv_inline_linetable' directive") || 3967 parseTokenLoc(Loc) || 3968 parseIntToken( 3969 SourceLineNum, 3970 "expected SourceLineNum in '.cv_inline_linetable' directive") || 3971 check(SourceLineNum < 0, Loc, 3972 "Line number less than zero in '.cv_inline_linetable' directive") || 3973 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc, 3974 "expected identifier in directive") || 3975 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc, 3976 "expected identifier in directive")) 3977 return true; 3978 3979 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 3980 return true; 3981 3982 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName); 3983 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName); 3984 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, 3985 SourceLineNum, FnStartSym, 3986 FnEndSym); 3987 return false; 3988 } 3989 3990 void AsmParser::initializeCVDefRangeTypeMap() { 3991 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER; 3992 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL; 3993 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER; 3994 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL; 3995 } 3996 3997 /// parseDirectiveCVDefRange 3998 /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* 3999 bool AsmParser::parseDirectiveCVDefRange() { 4000 SMLoc Loc; 4001 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; 4002 while (getLexer().is(AsmToken::Identifier)) { 4003 Loc = getLexer().getLoc(); 4004 StringRef GapStartName; 4005 if (parseIdentifier(GapStartName)) 4006 return Error(Loc, "expected identifier in directive"); 4007 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName); 4008 4009 Loc = getLexer().getLoc(); 4010 StringRef GapEndName; 4011 if (parseIdentifier(GapEndName)) 4012 return Error(Loc, "expected identifier in directive"); 4013 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName); 4014 4015 Ranges.push_back({GapStartSym, GapEndSym}); 4016 } 4017 4018 StringRef CVDefRangeTypeStr; 4019 if (parseToken( 4020 AsmToken::Comma, 4021 "expected comma before def_range type in .cv_def_range directive") || 4022 parseIdentifier(CVDefRangeTypeStr)) 4023 return Error(Loc, "expected def_range type in directive"); 4024 4025 StringMap<CVDefRangeType>::const_iterator CVTypeIt = 4026 CVDefRangeTypeMap.find(CVDefRangeTypeStr); 4027 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) 4028 ? CVDR_DEFRANGE 4029 : CVTypeIt->getValue(); 4030 switch (CVDRType) { 4031 case CVDR_DEFRANGE_REGISTER: { 4032 int64_t DRRegister; 4033 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4034 ".cv_def_range directive") || 4035 parseAbsoluteExpression(DRRegister)) 4036 return Error(Loc, "expected register number"); 4037 4038 codeview::DefRangeRegisterHeader DRHdr; 4039 DRHdr.Register = DRRegister; 4040 DRHdr.MayHaveNoName = 0; 4041 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4042 break; 4043 } 4044 case CVDR_DEFRANGE_FRAMEPOINTER_REL: { 4045 int64_t DROffset; 4046 if (parseToken(AsmToken::Comma, 4047 "expected comma before offset in .cv_def_range directive") || 4048 parseAbsoluteExpression(DROffset)) 4049 return Error(Loc, "expected offset value"); 4050 4051 codeview::DefRangeFramePointerRelHeader DRHdr; 4052 DRHdr.Offset = DROffset; 4053 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4054 break; 4055 } 4056 case CVDR_DEFRANGE_SUBFIELD_REGISTER: { 4057 int64_t DRRegister; 4058 int64_t DROffsetInParent; 4059 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4060 ".cv_def_range directive") || 4061 parseAbsoluteExpression(DRRegister)) 4062 return Error(Loc, "expected register number"); 4063 if (parseToken(AsmToken::Comma, 4064 "expected comma before offset in .cv_def_range directive") || 4065 parseAbsoluteExpression(DROffsetInParent)) 4066 return Error(Loc, "expected offset value"); 4067 4068 codeview::DefRangeSubfieldRegisterHeader DRHdr; 4069 DRHdr.Register = DRRegister; 4070 DRHdr.MayHaveNoName = 0; 4071 DRHdr.OffsetInParent = DROffsetInParent; 4072 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4073 break; 4074 } 4075 case CVDR_DEFRANGE_REGISTER_REL: { 4076 int64_t DRRegister; 4077 int64_t DRFlags; 4078 int64_t DRBasePointerOffset; 4079 if (parseToken(AsmToken::Comma, "expected comma before register number in " 4080 ".cv_def_range directive") || 4081 parseAbsoluteExpression(DRRegister)) 4082 return Error(Loc, "expected register value"); 4083 if (parseToken( 4084 AsmToken::Comma, 4085 "expected comma before flag value in .cv_def_range directive") || 4086 parseAbsoluteExpression(DRFlags)) 4087 return Error(Loc, "expected flag value"); 4088 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset " 4089 "in .cv_def_range directive") || 4090 parseAbsoluteExpression(DRBasePointerOffset)) 4091 return Error(Loc, "expected base pointer offset value"); 4092 4093 codeview::DefRangeRegisterRelHeader DRHdr; 4094 DRHdr.Register = DRRegister; 4095 DRHdr.Flags = DRFlags; 4096 DRHdr.BasePointerOffset = DRBasePointerOffset; 4097 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); 4098 break; 4099 } 4100 default: 4101 return Error(Loc, "unexpected def_range type in .cv_def_range directive"); 4102 } 4103 return true; 4104 } 4105 4106 /// parseDirectiveCVString 4107 /// ::= .cv_stringtable "string" 4108 bool AsmParser::parseDirectiveCVString() { 4109 std::string Data; 4110 if (checkForValidSection() || parseEscapedString(Data)) 4111 return true; 4112 4113 // Put the string in the table and emit the offset. 4114 std::pair<StringRef, unsigned> Insertion = 4115 getCVContext().addToStringTable(Data); 4116 getStreamer().emitInt32(Insertion.second); 4117 return false; 4118 } 4119 4120 /// parseDirectiveCVStringTable 4121 /// ::= .cv_stringtable 4122 bool AsmParser::parseDirectiveCVStringTable() { 4123 getStreamer().emitCVStringTableDirective(); 4124 return false; 4125 } 4126 4127 /// parseDirectiveCVFileChecksums 4128 /// ::= .cv_filechecksums 4129 bool AsmParser::parseDirectiveCVFileChecksums() { 4130 getStreamer().emitCVFileChecksumsDirective(); 4131 return false; 4132 } 4133 4134 /// parseDirectiveCVFileChecksumOffset 4135 /// ::= .cv_filechecksumoffset fileno 4136 bool AsmParser::parseDirectiveCVFileChecksumOffset() { 4137 int64_t FileNo; 4138 if (parseIntToken(FileNo, "expected identifier in directive")) 4139 return true; 4140 if (parseToken(AsmToken::EndOfStatement, "Expected End of Statement")) 4141 return true; 4142 getStreamer().emitCVFileChecksumOffsetDirective(FileNo); 4143 return false; 4144 } 4145 4146 /// parseDirectiveCVFPOData 4147 /// ::= .cv_fpo_data procsym 4148 bool AsmParser::parseDirectiveCVFPOData() { 4149 SMLoc DirLoc = getLexer().getLoc(); 4150 StringRef ProcName; 4151 if (parseIdentifier(ProcName)) 4152 return TokError("expected symbol name"); 4153 if (parseEOL()) 4154 return true; 4155 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName); 4156 getStreamer().EmitCVFPOData(ProcSym, DirLoc); 4157 return false; 4158 } 4159 4160 /// parseDirectiveCFISections 4161 /// ::= .cfi_sections section [, section] 4162 bool AsmParser::parseDirectiveCFISections() { 4163 StringRef Name; 4164 bool EH = false; 4165 bool Debug = false; 4166 4167 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4168 for (;;) { 4169 if (parseIdentifier(Name)) 4170 return TokError("expected .eh_frame or .debug_frame"); 4171 if (Name == ".eh_frame") 4172 EH = true; 4173 else if (Name == ".debug_frame") 4174 Debug = true; 4175 if (parseOptionalToken(AsmToken::EndOfStatement)) 4176 break; 4177 if (parseComma()) 4178 return true; 4179 } 4180 } 4181 getStreamer().emitCFISections(EH, Debug); 4182 return false; 4183 } 4184 4185 /// parseDirectiveCFIStartProc 4186 /// ::= .cfi_startproc [simple] 4187 bool AsmParser::parseDirectiveCFIStartProc() { 4188 StringRef Simple; 4189 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4190 if (check(parseIdentifier(Simple) || Simple != "simple", 4191 "unexpected token") || 4192 parseEOL()) 4193 return true; 4194 } 4195 4196 // TODO(kristina): Deal with a corner case of incorrect diagnostic context 4197 // being produced if this directive is emitted as part of preprocessor macro 4198 // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. 4199 // Tools like llvm-mc on the other hand are not affected by it, and report 4200 // correct context information. 4201 getStreamer().emitCFIStartProc(!Simple.empty(), Lexer.getLoc()); 4202 return false; 4203 } 4204 4205 /// parseDirectiveCFIEndProc 4206 /// ::= .cfi_endproc 4207 bool AsmParser::parseDirectiveCFIEndProc() { 4208 if (parseEOL()) 4209 return true; 4210 getStreamer().emitCFIEndProc(); 4211 return false; 4212 } 4213 4214 /// parse register name or number. 4215 bool AsmParser::parseRegisterOrRegisterNumber(int64_t &Register, 4216 SMLoc DirectiveLoc) { 4217 unsigned RegNo; 4218 4219 if (getLexer().isNot(AsmToken::Integer)) { 4220 if (getTargetParser().ParseRegister(RegNo, DirectiveLoc, DirectiveLoc)) 4221 return true; 4222 Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNo, true); 4223 } else 4224 return parseAbsoluteExpression(Register); 4225 4226 return false; 4227 } 4228 4229 /// parseDirectiveCFIDefCfa 4230 /// ::= .cfi_def_cfa register, offset 4231 bool AsmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { 4232 int64_t Register = 0, Offset = 0; 4233 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4234 parseAbsoluteExpression(Offset) || parseEOL()) 4235 return true; 4236 4237 getStreamer().emitCFIDefCfa(Register, Offset); 4238 return false; 4239 } 4240 4241 /// parseDirectiveCFIDefCfaOffset 4242 /// ::= .cfi_def_cfa_offset offset 4243 bool AsmParser::parseDirectiveCFIDefCfaOffset() { 4244 int64_t Offset = 0; 4245 if (parseAbsoluteExpression(Offset) || parseEOL()) 4246 return true; 4247 4248 getStreamer().emitCFIDefCfaOffset(Offset); 4249 return false; 4250 } 4251 4252 /// parseDirectiveCFIRegister 4253 /// ::= .cfi_register register, register 4254 bool AsmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { 4255 int64_t Register1 = 0, Register2 = 0; 4256 if (parseRegisterOrRegisterNumber(Register1, DirectiveLoc) || parseComma() || 4257 parseRegisterOrRegisterNumber(Register2, DirectiveLoc) || parseEOL()) 4258 return true; 4259 4260 getStreamer().emitCFIRegister(Register1, Register2); 4261 return false; 4262 } 4263 4264 /// parseDirectiveCFIWindowSave 4265 /// ::= .cfi_window_save 4266 bool AsmParser::parseDirectiveCFIWindowSave() { 4267 if (parseEOL()) 4268 return true; 4269 getStreamer().emitCFIWindowSave(); 4270 return false; 4271 } 4272 4273 /// parseDirectiveCFIAdjustCfaOffset 4274 /// ::= .cfi_adjust_cfa_offset adjustment 4275 bool AsmParser::parseDirectiveCFIAdjustCfaOffset() { 4276 int64_t Adjustment = 0; 4277 if (parseAbsoluteExpression(Adjustment) || parseEOL()) 4278 return true; 4279 4280 getStreamer().emitCFIAdjustCfaOffset(Adjustment); 4281 return false; 4282 } 4283 4284 /// parseDirectiveCFIDefCfaRegister 4285 /// ::= .cfi_def_cfa_register register 4286 bool AsmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { 4287 int64_t Register = 0; 4288 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4289 return true; 4290 4291 getStreamer().emitCFIDefCfaRegister(Register); 4292 return false; 4293 } 4294 4295 /// parseDirectiveCFILLVMDefAspaceCfa 4296 /// ::= .cfi_llvm_def_aspace_cfa register, offset, address_space 4297 bool AsmParser::parseDirectiveCFILLVMDefAspaceCfa(SMLoc DirectiveLoc) { 4298 int64_t Register = 0, Offset = 0, AddressSpace = 0; 4299 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4300 parseAbsoluteExpression(Offset) || parseComma() || 4301 parseAbsoluteExpression(AddressSpace) || parseEOL()) 4302 return true; 4303 4304 getStreamer().emitCFILLVMDefAspaceCfa(Register, Offset, AddressSpace); 4305 return false; 4306 } 4307 4308 /// parseDirectiveCFIOffset 4309 /// ::= .cfi_offset register, offset 4310 bool AsmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { 4311 int64_t Register = 0; 4312 int64_t Offset = 0; 4313 4314 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4315 parseAbsoluteExpression(Offset) || parseEOL()) 4316 return true; 4317 4318 getStreamer().emitCFIOffset(Register, Offset); 4319 return false; 4320 } 4321 4322 /// parseDirectiveCFIRelOffset 4323 /// ::= .cfi_rel_offset register, offset 4324 bool AsmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { 4325 int64_t Register = 0, Offset = 0; 4326 4327 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseComma() || 4328 parseAbsoluteExpression(Offset) || parseEOL()) 4329 return true; 4330 4331 getStreamer().emitCFIRelOffset(Register, Offset); 4332 return false; 4333 } 4334 4335 static bool isValidEncoding(int64_t Encoding) { 4336 if (Encoding & ~0xff) 4337 return false; 4338 4339 if (Encoding == dwarf::DW_EH_PE_omit) 4340 return true; 4341 4342 const unsigned Format = Encoding & 0xf; 4343 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 4344 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 4345 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 4346 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 4347 return false; 4348 4349 const unsigned Application = Encoding & 0x70; 4350 if (Application != dwarf::DW_EH_PE_absptr && 4351 Application != dwarf::DW_EH_PE_pcrel) 4352 return false; 4353 4354 return true; 4355 } 4356 4357 /// parseDirectiveCFIPersonalityOrLsda 4358 /// IsPersonality true for cfi_personality, false for cfi_lsda 4359 /// ::= .cfi_personality encoding, [symbol_name] 4360 /// ::= .cfi_lsda encoding, [symbol_name] 4361 bool AsmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { 4362 int64_t Encoding = 0; 4363 if (parseAbsoluteExpression(Encoding)) 4364 return true; 4365 if (Encoding == dwarf::DW_EH_PE_omit) 4366 return false; 4367 4368 StringRef Name; 4369 if (check(!isValidEncoding(Encoding), "unsupported encoding.") || 4370 parseComma() || 4371 check(parseIdentifier(Name), "expected identifier in directive") || 4372 parseEOL()) 4373 return true; 4374 4375 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4376 4377 if (IsPersonality) 4378 getStreamer().emitCFIPersonality(Sym, Encoding); 4379 else 4380 getStreamer().emitCFILsda(Sym, Encoding); 4381 return false; 4382 } 4383 4384 /// parseDirectiveCFIRememberState 4385 /// ::= .cfi_remember_state 4386 bool AsmParser::parseDirectiveCFIRememberState() { 4387 if (parseEOL()) 4388 return true; 4389 getStreamer().emitCFIRememberState(); 4390 return false; 4391 } 4392 4393 /// parseDirectiveCFIRestoreState 4394 /// ::= .cfi_remember_state 4395 bool AsmParser::parseDirectiveCFIRestoreState() { 4396 if (parseEOL()) 4397 return true; 4398 getStreamer().emitCFIRestoreState(); 4399 return false; 4400 } 4401 4402 /// parseDirectiveCFISameValue 4403 /// ::= .cfi_same_value register 4404 bool AsmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { 4405 int64_t Register = 0; 4406 4407 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4408 return true; 4409 4410 getStreamer().emitCFISameValue(Register); 4411 return false; 4412 } 4413 4414 /// parseDirectiveCFIRestore 4415 /// ::= .cfi_restore register 4416 bool AsmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { 4417 int64_t Register = 0; 4418 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4419 return true; 4420 4421 getStreamer().emitCFIRestore(Register); 4422 return false; 4423 } 4424 4425 /// parseDirectiveCFIEscape 4426 /// ::= .cfi_escape expression[,...] 4427 bool AsmParser::parseDirectiveCFIEscape() { 4428 std::string Values; 4429 int64_t CurrValue; 4430 if (parseAbsoluteExpression(CurrValue)) 4431 return true; 4432 4433 Values.push_back((uint8_t)CurrValue); 4434 4435 while (getLexer().is(AsmToken::Comma)) { 4436 Lex(); 4437 4438 if (parseAbsoluteExpression(CurrValue)) 4439 return true; 4440 4441 Values.push_back((uint8_t)CurrValue); 4442 } 4443 4444 getStreamer().emitCFIEscape(Values); 4445 return false; 4446 } 4447 4448 /// parseDirectiveCFIReturnColumn 4449 /// ::= .cfi_return_column register 4450 bool AsmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { 4451 int64_t Register = 0; 4452 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4453 return true; 4454 getStreamer().emitCFIReturnColumn(Register); 4455 return false; 4456 } 4457 4458 /// parseDirectiveCFISignalFrame 4459 /// ::= .cfi_signal_frame 4460 bool AsmParser::parseDirectiveCFISignalFrame() { 4461 if (parseEOL()) 4462 return true; 4463 4464 getStreamer().emitCFISignalFrame(); 4465 return false; 4466 } 4467 4468 /// parseDirectiveCFIUndefined 4469 /// ::= .cfi_undefined register 4470 bool AsmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { 4471 int64_t Register = 0; 4472 4473 if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || parseEOL()) 4474 return true; 4475 4476 getStreamer().emitCFIUndefined(Register); 4477 return false; 4478 } 4479 4480 /// parseDirectiveAltmacro 4481 /// ::= .altmacro 4482 /// ::= .noaltmacro 4483 bool AsmParser::parseDirectiveAltmacro(StringRef Directive) { 4484 if (parseEOL()) 4485 return true; 4486 AltMacroMode = (Directive == ".altmacro"); 4487 return false; 4488 } 4489 4490 /// parseDirectiveMacrosOnOff 4491 /// ::= .macros_on 4492 /// ::= .macros_off 4493 bool AsmParser::parseDirectiveMacrosOnOff(StringRef Directive) { 4494 if (parseEOL()) 4495 return true; 4496 setMacrosEnabled(Directive == ".macros_on"); 4497 return false; 4498 } 4499 4500 /// parseDirectiveMacro 4501 /// ::= .macro name[,] [parameters] 4502 bool AsmParser::parseDirectiveMacro(SMLoc DirectiveLoc) { 4503 StringRef Name; 4504 if (parseIdentifier(Name)) 4505 return TokError("expected identifier in '.macro' directive"); 4506 4507 if (getLexer().is(AsmToken::Comma)) 4508 Lex(); 4509 4510 MCAsmMacroParameters Parameters; 4511 while (getLexer().isNot(AsmToken::EndOfStatement)) { 4512 4513 if (!Parameters.empty() && Parameters.back().Vararg) 4514 return Error(Lexer.getLoc(), "vararg parameter '" + 4515 Parameters.back().Name + 4516 "' should be the last parameter"); 4517 4518 MCAsmMacroParameter Parameter; 4519 if (parseIdentifier(Parameter.Name)) 4520 return TokError("expected identifier in '.macro' directive"); 4521 4522 // Emit an error if two (or more) named parameters share the same name 4523 for (const MCAsmMacroParameter& CurrParam : Parameters) 4524 if (CurrParam.Name.equals(Parameter.Name)) 4525 return TokError("macro '" + Name + "' has multiple parameters" 4526 " named '" + Parameter.Name + "'"); 4527 4528 if (Lexer.is(AsmToken::Colon)) { 4529 Lex(); // consume ':' 4530 4531 SMLoc QualLoc; 4532 StringRef Qualifier; 4533 4534 QualLoc = Lexer.getLoc(); 4535 if (parseIdentifier(Qualifier)) 4536 return Error(QualLoc, "missing parameter qualifier for " 4537 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4538 4539 if (Qualifier == "req") 4540 Parameter.Required = true; 4541 else if (Qualifier == "vararg") 4542 Parameter.Vararg = true; 4543 else 4544 return Error(QualLoc, Qualifier + " is not a valid parameter qualifier " 4545 "for '" + Parameter.Name + "' in macro '" + Name + "'"); 4546 } 4547 4548 if (getLexer().is(AsmToken::Equal)) { 4549 Lex(); 4550 4551 SMLoc ParamLoc; 4552 4553 ParamLoc = Lexer.getLoc(); 4554 if (parseMacroArgument(Parameter.Value, /*Vararg=*/false )) 4555 return true; 4556 4557 if (Parameter.Required) 4558 Warning(ParamLoc, "pointless default value for required parameter " 4559 "'" + Parameter.Name + "' in macro '" + Name + "'"); 4560 } 4561 4562 Parameters.push_back(std::move(Parameter)); 4563 4564 if (getLexer().is(AsmToken::Comma)) 4565 Lex(); 4566 } 4567 4568 // Eat just the end of statement. 4569 Lexer.Lex(); 4570 4571 // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors 4572 AsmToken EndToken, StartToken = getTok(); 4573 unsigned MacroDepth = 0; 4574 // Lex the macro definition. 4575 while (true) { 4576 // Ignore Lexing errors in macros. 4577 while (Lexer.is(AsmToken::Error)) { 4578 Lexer.Lex(); 4579 } 4580 4581 // Check whether we have reached the end of the file. 4582 if (getLexer().is(AsmToken::Eof)) 4583 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 4584 4585 // Otherwise, check whether we have reach the .endmacro or the start of a 4586 // preprocessor line marker. 4587 if (getLexer().is(AsmToken::Identifier)) { 4588 if (getTok().getIdentifier() == ".endm" || 4589 getTok().getIdentifier() == ".endmacro") { 4590 if (MacroDepth == 0) { // Outermost macro. 4591 EndToken = getTok(); 4592 Lexer.Lex(); 4593 if (getLexer().isNot(AsmToken::EndOfStatement)) 4594 return TokError("unexpected token in '" + EndToken.getIdentifier() + 4595 "' directive"); 4596 break; 4597 } else { 4598 // Otherwise we just found the end of an inner macro. 4599 --MacroDepth; 4600 } 4601 } else if (getTok().getIdentifier() == ".macro") { 4602 // We allow nested macros. Those aren't instantiated until the outermost 4603 // macro is expanded so just ignore them for now. 4604 ++MacroDepth; 4605 } 4606 } else if (Lexer.is(AsmToken::HashDirective)) { 4607 (void)parseCppHashLineFilenameComment(getLexer().getLoc()); 4608 } 4609 4610 // Otherwise, scan til the end of the statement. 4611 eatToEndOfStatement(); 4612 } 4613 4614 if (getContext().lookupMacro(Name)) { 4615 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 4616 } 4617 4618 const char *BodyStart = StartToken.getLoc().getPointer(); 4619 const char *BodyEnd = EndToken.getLoc().getPointer(); 4620 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 4621 checkForBadMacro(DirectiveLoc, Name, Body, Parameters); 4622 MCAsmMacro Macro(Name, Body, std::move(Parameters)); 4623 DEBUG_WITH_TYPE("asm-macros", dbgs() << "Defining new macro:\n"; 4624 Macro.dump()); 4625 getContext().defineMacro(Name, std::move(Macro)); 4626 return false; 4627 } 4628 4629 /// checkForBadMacro 4630 /// 4631 /// With the support added for named parameters there may be code out there that 4632 /// is transitioning from positional parameters. In versions of gas that did 4633 /// not support named parameters they would be ignored on the macro definition. 4634 /// But to support both styles of parameters this is not possible so if a macro 4635 /// definition has named parameters but does not use them and has what appears 4636 /// to be positional parameters, strings like $1, $2, ... and $n, then issue a 4637 /// warning that the positional parameter found in body which have no effect. 4638 /// Hoping the developer will either remove the named parameters from the macro 4639 /// definition so the positional parameters get used if that was what was 4640 /// intended or change the macro to use the named parameters. It is possible 4641 /// this warning will trigger when the none of the named parameters are used 4642 /// and the strings like $1 are infact to simply to be passed trough unchanged. 4643 void AsmParser::checkForBadMacro(SMLoc DirectiveLoc, StringRef Name, 4644 StringRef Body, 4645 ArrayRef<MCAsmMacroParameter> Parameters) { 4646 // If this macro is not defined with named parameters the warning we are 4647 // checking for here doesn't apply. 4648 unsigned NParameters = Parameters.size(); 4649 if (NParameters == 0) 4650 return; 4651 4652 bool NamedParametersFound = false; 4653 bool PositionalParametersFound = false; 4654 4655 // Look at the body of the macro for use of both the named parameters and what 4656 // are likely to be positional parameters. This is what expandMacro() is 4657 // doing when it finds the parameters in the body. 4658 while (!Body.empty()) { 4659 // Scan for the next possible parameter. 4660 std::size_t End = Body.size(), Pos = 0; 4661 for (; Pos != End; ++Pos) { 4662 // Check for a substitution or escape. 4663 // This macro is defined with parameters, look for \foo, \bar, etc. 4664 if (Body[Pos] == '\\' && Pos + 1 != End) 4665 break; 4666 4667 // This macro should have parameters, but look for $0, $1, ..., $n too. 4668 if (Body[Pos] != '$' || Pos + 1 == End) 4669 continue; 4670 char Next = Body[Pos + 1]; 4671 if (Next == '$' || Next == 'n' || 4672 isdigit(static_cast<unsigned char>(Next))) 4673 break; 4674 } 4675 4676 // Check if we reached the end. 4677 if (Pos == End) 4678 break; 4679 4680 if (Body[Pos] == '$') { 4681 switch (Body[Pos + 1]) { 4682 // $$ => $ 4683 case '$': 4684 break; 4685 4686 // $n => number of arguments 4687 case 'n': 4688 PositionalParametersFound = true; 4689 break; 4690 4691 // $[0-9] => argument 4692 default: { 4693 PositionalParametersFound = true; 4694 break; 4695 } 4696 } 4697 Pos += 2; 4698 } else { 4699 unsigned I = Pos + 1; 4700 while (isIdentifierChar(Body[I]) && I + 1 != End) 4701 ++I; 4702 4703 const char *Begin = Body.data() + Pos + 1; 4704 StringRef Argument(Begin, I - (Pos + 1)); 4705 unsigned Index = 0; 4706 for (; Index < NParameters; ++Index) 4707 if (Parameters[Index].Name == Argument) 4708 break; 4709 4710 if (Index == NParameters) { 4711 if (Body[Pos + 1] == '(' && Body[Pos + 2] == ')') 4712 Pos += 3; 4713 else { 4714 Pos = I; 4715 } 4716 } else { 4717 NamedParametersFound = true; 4718 Pos += 1 + Argument.size(); 4719 } 4720 } 4721 // Update the scan point. 4722 Body = Body.substr(Pos); 4723 } 4724 4725 if (!NamedParametersFound && PositionalParametersFound) 4726 Warning(DirectiveLoc, "macro defined with named parameters which are not " 4727 "used in macro body, possible positional parameter " 4728 "found in body which will have no effect"); 4729 } 4730 4731 /// parseDirectiveExitMacro 4732 /// ::= .exitm 4733 bool AsmParser::parseDirectiveExitMacro(StringRef Directive) { 4734 if (parseEOL()) 4735 return true; 4736 4737 if (!isInsideMacroInstantiation()) 4738 return TokError("unexpected '" + Directive + "' in file, " 4739 "no current macro definition"); 4740 4741 // Exit all conditionals that are active in the current macro. 4742 while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { 4743 TheCondState = TheCondStack.back(); 4744 TheCondStack.pop_back(); 4745 } 4746 4747 handleMacroExit(); 4748 return false; 4749 } 4750 4751 /// parseDirectiveEndMacro 4752 /// ::= .endm 4753 /// ::= .endmacro 4754 bool AsmParser::parseDirectiveEndMacro(StringRef Directive) { 4755 if (getLexer().isNot(AsmToken::EndOfStatement)) 4756 return TokError("unexpected token in '" + Directive + "' directive"); 4757 4758 // If we are inside a macro instantiation, terminate the current 4759 // instantiation. 4760 if (isInsideMacroInstantiation()) { 4761 handleMacroExit(); 4762 return false; 4763 } 4764 4765 // Otherwise, this .endmacro is a stray entry in the file; well formed 4766 // .endmacro directives are handled during the macro definition parsing. 4767 return TokError("unexpected '" + Directive + "' in file, " 4768 "no current macro definition"); 4769 } 4770 4771 /// parseDirectivePurgeMacro 4772 /// ::= .purgem name 4773 bool AsmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { 4774 StringRef Name; 4775 SMLoc Loc; 4776 if (parseTokenLoc(Loc) || 4777 check(parseIdentifier(Name), Loc, 4778 "expected identifier in '.purgem' directive") || 4779 parseEOL()) 4780 return true; 4781 4782 if (!getContext().lookupMacro(Name)) 4783 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 4784 4785 getContext().undefineMacro(Name); 4786 DEBUG_WITH_TYPE("asm-macros", dbgs() 4787 << "Un-defining macro: " << Name << "\n"); 4788 return false; 4789 } 4790 4791 /// parseDirectiveBundleAlignMode 4792 /// ::= {.bundle_align_mode} expression 4793 bool AsmParser::parseDirectiveBundleAlignMode() { 4794 // Expect a single argument: an expression that evaluates to a constant 4795 // in the inclusive range 0-30. 4796 SMLoc ExprLoc = getLexer().getLoc(); 4797 int64_t AlignSizePow2; 4798 if (checkForValidSection() || parseAbsoluteExpression(AlignSizePow2) || 4799 parseEOL() || 4800 check(AlignSizePow2 < 0 || AlignSizePow2 > 30, ExprLoc, 4801 "invalid bundle alignment size (expected between 0 and 30)")) 4802 return true; 4803 4804 // Because of AlignSizePow2's verified range we can safely truncate it to 4805 // unsigned. 4806 getStreamer().emitBundleAlignMode(static_cast<unsigned>(AlignSizePow2)); 4807 return false; 4808 } 4809 4810 /// parseDirectiveBundleLock 4811 /// ::= {.bundle_lock} [align_to_end] 4812 bool AsmParser::parseDirectiveBundleLock() { 4813 if (checkForValidSection()) 4814 return true; 4815 bool AlignToEnd = false; 4816 4817 StringRef Option; 4818 SMLoc Loc = getTok().getLoc(); 4819 const char *kInvalidOptionError = 4820 "invalid option for '.bundle_lock' directive"; 4821 4822 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 4823 if (check(parseIdentifier(Option), Loc, kInvalidOptionError) || 4824 check(Option != "align_to_end", Loc, kInvalidOptionError) || parseEOL()) 4825 return true; 4826 AlignToEnd = true; 4827 } 4828 4829 getStreamer().emitBundleLock(AlignToEnd); 4830 return false; 4831 } 4832 4833 /// parseDirectiveBundleLock 4834 /// ::= {.bundle_lock} 4835 bool AsmParser::parseDirectiveBundleUnlock() { 4836 if (checkForValidSection() || parseEOL()) 4837 return true; 4838 4839 getStreamer().emitBundleUnlock(); 4840 return false; 4841 } 4842 4843 /// parseDirectiveSpace 4844 /// ::= (.skip | .space) expression [ , expression ] 4845 bool AsmParser::parseDirectiveSpace(StringRef IDVal) { 4846 SMLoc NumBytesLoc = Lexer.getLoc(); 4847 const MCExpr *NumBytes; 4848 if (checkForValidSection() || parseExpression(NumBytes)) 4849 return true; 4850 4851 int64_t FillExpr = 0; 4852 if (parseOptionalToken(AsmToken::Comma)) 4853 if (parseAbsoluteExpression(FillExpr)) 4854 return true; 4855 if (parseEOL()) 4856 return true; 4857 4858 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 4859 getStreamer().emitFill(*NumBytes, FillExpr, NumBytesLoc); 4860 4861 return false; 4862 } 4863 4864 /// parseDirectiveDCB 4865 /// ::= .dcb.{b, l, w} expression, expression 4866 bool AsmParser::parseDirectiveDCB(StringRef IDVal, unsigned Size) { 4867 SMLoc NumValuesLoc = Lexer.getLoc(); 4868 int64_t NumValues; 4869 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4870 return true; 4871 4872 if (NumValues < 0) { 4873 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4874 return false; 4875 } 4876 4877 if (parseComma()) 4878 return true; 4879 4880 const MCExpr *Value; 4881 SMLoc ExprLoc = getLexer().getLoc(); 4882 if (parseExpression(Value)) 4883 return true; 4884 4885 // Special case constant expressions to match code generator. 4886 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 4887 assert(Size <= 8 && "Invalid size"); 4888 uint64_t IntValue = MCE->getValue(); 4889 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 4890 return Error(ExprLoc, "literal value out of range for directive"); 4891 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4892 getStreamer().emitIntValue(IntValue, Size); 4893 } else { 4894 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4895 getStreamer().emitValue(Value, Size, ExprLoc); 4896 } 4897 4898 return parseEOL(); 4899 } 4900 4901 /// parseDirectiveRealDCB 4902 /// ::= .dcb.{d, s} expression, expression 4903 bool AsmParser::parseDirectiveRealDCB(StringRef IDVal, const fltSemantics &Semantics) { 4904 SMLoc NumValuesLoc = Lexer.getLoc(); 4905 int64_t NumValues; 4906 if (checkForValidSection() || parseAbsoluteExpression(NumValues)) 4907 return true; 4908 4909 if (NumValues < 0) { 4910 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4911 return false; 4912 } 4913 4914 if (parseComma()) 4915 return true; 4916 4917 APInt AsInt; 4918 if (parseRealValue(Semantics, AsInt) || parseEOL()) 4919 return true; 4920 4921 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4922 getStreamer().emitIntValue(AsInt.getLimitedValue(), 4923 AsInt.getBitWidth() / 8); 4924 4925 return false; 4926 } 4927 4928 /// parseDirectiveDS 4929 /// ::= .ds.{b, d, l, p, s, w, x} expression 4930 bool AsmParser::parseDirectiveDS(StringRef IDVal, unsigned Size) { 4931 SMLoc NumValuesLoc = Lexer.getLoc(); 4932 int64_t NumValues; 4933 if (checkForValidSection() || parseAbsoluteExpression(NumValues) || 4934 parseEOL()) 4935 return true; 4936 4937 if (NumValues < 0) { 4938 Warning(NumValuesLoc, "'" + Twine(IDVal) + "' directive with negative repeat count has no effect"); 4939 return false; 4940 } 4941 4942 for (uint64_t i = 0, e = NumValues; i != e; ++i) 4943 getStreamer().emitFill(Size, 0); 4944 4945 return false; 4946 } 4947 4948 /// parseDirectiveLEB128 4949 /// ::= (.sleb128 | .uleb128) [ expression (, expression)* ] 4950 bool AsmParser::parseDirectiveLEB128(bool Signed) { 4951 if (checkForValidSection()) 4952 return true; 4953 4954 auto parseOp = [&]() -> bool { 4955 const MCExpr *Value; 4956 if (parseExpression(Value)) 4957 return true; 4958 if (Signed) 4959 getStreamer().emitSLEB128Value(Value); 4960 else 4961 getStreamer().emitULEB128Value(Value); 4962 return false; 4963 }; 4964 4965 return parseMany(parseOp); 4966 } 4967 4968 /// parseDirectiveSymbolAttribute 4969 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 4970 bool AsmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 4971 auto parseOp = [&]() -> bool { 4972 StringRef Name; 4973 SMLoc Loc = getTok().getLoc(); 4974 if (parseIdentifier(Name)) 4975 return Error(Loc, "expected identifier"); 4976 4977 if (discardLTOSymbol(Name)) 4978 return false; 4979 4980 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 4981 4982 // Assembler local symbols don't make any sense here. Complain loudly. 4983 if (Sym->isTemporary()) 4984 return Error(Loc, "non-local symbol required"); 4985 4986 if (!getStreamer().emitSymbolAttribute(Sym, Attr)) 4987 return Error(Loc, "unable to emit symbol attribute"); 4988 return false; 4989 }; 4990 4991 return parseMany(parseOp); 4992 } 4993 4994 /// parseDirectiveComm 4995 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 4996 bool AsmParser::parseDirectiveComm(bool IsLocal) { 4997 if (checkForValidSection()) 4998 return true; 4999 5000 SMLoc IDLoc = getLexer().getLoc(); 5001 StringRef Name; 5002 if (parseIdentifier(Name)) 5003 return TokError("expected identifier in directive"); 5004 5005 // Handle the identifier as the key symbol. 5006 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5007 5008 if (parseComma()) 5009 return true; 5010 5011 int64_t Size; 5012 SMLoc SizeLoc = getLexer().getLoc(); 5013 if (parseAbsoluteExpression(Size)) 5014 return true; 5015 5016 int64_t Pow2Alignment = 0; 5017 SMLoc Pow2AlignmentLoc; 5018 if (getLexer().is(AsmToken::Comma)) { 5019 Lex(); 5020 Pow2AlignmentLoc = getLexer().getLoc(); 5021 if (parseAbsoluteExpression(Pow2Alignment)) 5022 return true; 5023 5024 LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); 5025 if (IsLocal && LCOMM == LCOMM::NoAlignment) 5026 return Error(Pow2AlignmentLoc, "alignment not supported on this target"); 5027 5028 // If this target takes alignments in bytes (not log) validate and convert. 5029 if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || 5030 (IsLocal && LCOMM == LCOMM::ByteAlignment)) { 5031 if (!isPowerOf2_64(Pow2Alignment)) 5032 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 5033 Pow2Alignment = Log2_64(Pow2Alignment); 5034 } 5035 } 5036 5037 if (parseEOL()) 5038 return true; 5039 5040 // NOTE: a size of zero for a .comm should create a undefined symbol 5041 // but a size of .lcomm creates a bss symbol of size zero. 5042 if (Size < 0) 5043 return Error(SizeLoc, "size must be non-negative"); 5044 5045 Sym->redefineIfPossible(); 5046 if (!Sym->isUndefined()) 5047 return Error(IDLoc, "invalid symbol redefinition"); 5048 5049 // Create the Symbol as a common or local common with Size and Pow2Alignment 5050 if (IsLocal) { 5051 getStreamer().emitLocalCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5052 return false; 5053 } 5054 5055 getStreamer().emitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 5056 return false; 5057 } 5058 5059 /// parseDirectiveAbort 5060 /// ::= .abort [... message ...] 5061 bool AsmParser::parseDirectiveAbort() { 5062 // FIXME: Use loc from directive. 5063 SMLoc Loc = getLexer().getLoc(); 5064 5065 StringRef Str = parseStringToEndOfStatement(); 5066 if (parseEOL()) 5067 return true; 5068 5069 if (Str.empty()) 5070 return Error(Loc, ".abort detected. Assembly stopping."); 5071 else 5072 return Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 5073 // FIXME: Actually abort assembly here. 5074 5075 return false; 5076 } 5077 5078 /// parseDirectiveInclude 5079 /// ::= .include "filename" 5080 bool AsmParser::parseDirectiveInclude() { 5081 // Allow the strings to have escaped octal character sequence. 5082 std::string Filename; 5083 SMLoc IncludeLoc = getTok().getLoc(); 5084 5085 if (check(getTok().isNot(AsmToken::String), 5086 "expected string in '.include' directive") || 5087 parseEscapedString(Filename) || 5088 check(getTok().isNot(AsmToken::EndOfStatement), 5089 "unexpected token in '.include' directive") || 5090 // Attempt to switch the lexer to the included file before consuming the 5091 // end of statement to avoid losing it when we switch. 5092 check(enterIncludeFile(Filename), IncludeLoc, 5093 "Could not find include file '" + Filename + "'")) 5094 return true; 5095 5096 return false; 5097 } 5098 5099 /// parseDirectiveIncbin 5100 /// ::= .incbin "filename" [ , skip [ , count ] ] 5101 bool AsmParser::parseDirectiveIncbin() { 5102 // Allow the strings to have escaped octal character sequence. 5103 std::string Filename; 5104 SMLoc IncbinLoc = getTok().getLoc(); 5105 if (check(getTok().isNot(AsmToken::String), 5106 "expected string in '.incbin' directive") || 5107 parseEscapedString(Filename)) 5108 return true; 5109 5110 int64_t Skip = 0; 5111 const MCExpr *Count = nullptr; 5112 SMLoc SkipLoc, CountLoc; 5113 if (parseOptionalToken(AsmToken::Comma)) { 5114 // The skip expression can be omitted while specifying the count, e.g: 5115 // .incbin "filename",,4 5116 if (getTok().isNot(AsmToken::Comma)) { 5117 if (parseTokenLoc(SkipLoc) || parseAbsoluteExpression(Skip)) 5118 return true; 5119 } 5120 if (parseOptionalToken(AsmToken::Comma)) { 5121 CountLoc = getTok().getLoc(); 5122 if (parseExpression(Count)) 5123 return true; 5124 } 5125 } 5126 5127 if (parseEOL()) 5128 return true; 5129 5130 if (check(Skip < 0, SkipLoc, "skip is negative")) 5131 return true; 5132 5133 // Attempt to process the included file. 5134 if (processIncbinFile(Filename, Skip, Count, CountLoc)) 5135 return Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 5136 return false; 5137 } 5138 5139 /// parseDirectiveIf 5140 /// ::= .if{,eq,ge,gt,le,lt,ne} expression 5141 bool AsmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { 5142 TheCondStack.push_back(TheCondState); 5143 TheCondState.TheCond = AsmCond::IfCond; 5144 if (TheCondState.Ignore) { 5145 eatToEndOfStatement(); 5146 } else { 5147 int64_t ExprValue; 5148 if (parseAbsoluteExpression(ExprValue) || parseEOL()) 5149 return true; 5150 5151 switch (DirKind) { 5152 default: 5153 llvm_unreachable("unsupported directive"); 5154 case DK_IF: 5155 case DK_IFNE: 5156 break; 5157 case DK_IFEQ: 5158 ExprValue = ExprValue == 0; 5159 break; 5160 case DK_IFGE: 5161 ExprValue = ExprValue >= 0; 5162 break; 5163 case DK_IFGT: 5164 ExprValue = ExprValue > 0; 5165 break; 5166 case DK_IFLE: 5167 ExprValue = ExprValue <= 0; 5168 break; 5169 case DK_IFLT: 5170 ExprValue = ExprValue < 0; 5171 break; 5172 } 5173 5174 TheCondState.CondMet = ExprValue; 5175 TheCondState.Ignore = !TheCondState.CondMet; 5176 } 5177 5178 return false; 5179 } 5180 5181 /// parseDirectiveIfb 5182 /// ::= .ifb string 5183 bool AsmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 5184 TheCondStack.push_back(TheCondState); 5185 TheCondState.TheCond = AsmCond::IfCond; 5186 5187 if (TheCondState.Ignore) { 5188 eatToEndOfStatement(); 5189 } else { 5190 StringRef Str = parseStringToEndOfStatement(); 5191 5192 if (parseEOL()) 5193 return true; 5194 5195 TheCondState.CondMet = ExpectBlank == Str.empty(); 5196 TheCondState.Ignore = !TheCondState.CondMet; 5197 } 5198 5199 return false; 5200 } 5201 5202 /// parseDirectiveIfc 5203 /// ::= .ifc string1, string2 5204 /// ::= .ifnc string1, string2 5205 bool AsmParser::parseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 5206 TheCondStack.push_back(TheCondState); 5207 TheCondState.TheCond = AsmCond::IfCond; 5208 5209 if (TheCondState.Ignore) { 5210 eatToEndOfStatement(); 5211 } else { 5212 StringRef Str1 = parseStringToComma(); 5213 5214 if (parseComma()) 5215 return true; 5216 5217 StringRef Str2 = parseStringToEndOfStatement(); 5218 5219 if (parseEOL()) 5220 return true; 5221 5222 TheCondState.CondMet = ExpectEqual == (Str1.trim() == Str2.trim()); 5223 TheCondState.Ignore = !TheCondState.CondMet; 5224 } 5225 5226 return false; 5227 } 5228 5229 /// parseDirectiveIfeqs 5230 /// ::= .ifeqs string1, string2 5231 bool AsmParser::parseDirectiveIfeqs(SMLoc DirectiveLoc, bool ExpectEqual) { 5232 if (Lexer.isNot(AsmToken::String)) { 5233 if (ExpectEqual) 5234 return TokError("expected string parameter for '.ifeqs' directive"); 5235 return TokError("expected string parameter for '.ifnes' directive"); 5236 } 5237 5238 StringRef String1 = getTok().getStringContents(); 5239 Lex(); 5240 5241 if (Lexer.isNot(AsmToken::Comma)) { 5242 if (ExpectEqual) 5243 return TokError( 5244 "expected comma after first string for '.ifeqs' directive"); 5245 return TokError("expected comma after first string for '.ifnes' directive"); 5246 } 5247 5248 Lex(); 5249 5250 if (Lexer.isNot(AsmToken::String)) { 5251 if (ExpectEqual) 5252 return TokError("expected string parameter for '.ifeqs' directive"); 5253 return TokError("expected string parameter for '.ifnes' directive"); 5254 } 5255 5256 StringRef String2 = getTok().getStringContents(); 5257 Lex(); 5258 5259 TheCondStack.push_back(TheCondState); 5260 TheCondState.TheCond = AsmCond::IfCond; 5261 TheCondState.CondMet = ExpectEqual == (String1 == String2); 5262 TheCondState.Ignore = !TheCondState.CondMet; 5263 5264 return false; 5265 } 5266 5267 /// parseDirectiveIfdef 5268 /// ::= .ifdef symbol 5269 bool AsmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 5270 StringRef Name; 5271 TheCondStack.push_back(TheCondState); 5272 TheCondState.TheCond = AsmCond::IfCond; 5273 5274 if (TheCondState.Ignore) { 5275 eatToEndOfStatement(); 5276 } else { 5277 if (check(parseIdentifier(Name), "expected identifier after '.ifdef'") || 5278 parseEOL()) 5279 return true; 5280 5281 MCSymbol *Sym = getContext().lookupSymbol(Name); 5282 5283 if (expect_defined) 5284 TheCondState.CondMet = (Sym && !Sym->isUndefined(false)); 5285 else 5286 TheCondState.CondMet = (!Sym || Sym->isUndefined(false)); 5287 TheCondState.Ignore = !TheCondState.CondMet; 5288 } 5289 5290 return false; 5291 } 5292 5293 /// parseDirectiveElseIf 5294 /// ::= .elseif expression 5295 bool AsmParser::parseDirectiveElseIf(SMLoc DirectiveLoc) { 5296 if (TheCondState.TheCond != AsmCond::IfCond && 5297 TheCondState.TheCond != AsmCond::ElseIfCond) 5298 return Error(DirectiveLoc, "Encountered a .elseif that doesn't follow an" 5299 " .if or an .elseif"); 5300 TheCondState.TheCond = AsmCond::ElseIfCond; 5301 5302 bool LastIgnoreState = false; 5303 if (!TheCondStack.empty()) 5304 LastIgnoreState = TheCondStack.back().Ignore; 5305 if (LastIgnoreState || TheCondState.CondMet) { 5306 TheCondState.Ignore = true; 5307 eatToEndOfStatement(); 5308 } else { 5309 int64_t ExprValue; 5310 if (parseAbsoluteExpression(ExprValue)) 5311 return true; 5312 5313 if (parseEOL()) 5314 return true; 5315 5316 TheCondState.CondMet = ExprValue; 5317 TheCondState.Ignore = !TheCondState.CondMet; 5318 } 5319 5320 return false; 5321 } 5322 5323 /// parseDirectiveElse 5324 /// ::= .else 5325 bool AsmParser::parseDirectiveElse(SMLoc DirectiveLoc) { 5326 if (parseEOL()) 5327 return true; 5328 5329 if (TheCondState.TheCond != AsmCond::IfCond && 5330 TheCondState.TheCond != AsmCond::ElseIfCond) 5331 return Error(DirectiveLoc, "Encountered a .else that doesn't follow " 5332 " an .if or an .elseif"); 5333 TheCondState.TheCond = AsmCond::ElseCond; 5334 bool LastIgnoreState = false; 5335 if (!TheCondStack.empty()) 5336 LastIgnoreState = TheCondStack.back().Ignore; 5337 if (LastIgnoreState || TheCondState.CondMet) 5338 TheCondState.Ignore = true; 5339 else 5340 TheCondState.Ignore = false; 5341 5342 return false; 5343 } 5344 5345 /// parseDirectiveEnd 5346 /// ::= .end 5347 bool AsmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { 5348 if (parseEOL()) 5349 return true; 5350 5351 while (Lexer.isNot(AsmToken::Eof)) 5352 Lexer.Lex(); 5353 5354 return false; 5355 } 5356 5357 /// parseDirectiveError 5358 /// ::= .err 5359 /// ::= .error [string] 5360 bool AsmParser::parseDirectiveError(SMLoc L, bool WithMessage) { 5361 if (!TheCondStack.empty()) { 5362 if (TheCondStack.back().Ignore) { 5363 eatToEndOfStatement(); 5364 return false; 5365 } 5366 } 5367 5368 if (!WithMessage) 5369 return Error(L, ".err encountered"); 5370 5371 StringRef Message = ".error directive invoked in source file"; 5372 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5373 if (Lexer.isNot(AsmToken::String)) 5374 return TokError(".error argument must be a string"); 5375 5376 Message = getTok().getStringContents(); 5377 Lex(); 5378 } 5379 5380 return Error(L, Message); 5381 } 5382 5383 /// parseDirectiveWarning 5384 /// ::= .warning [string] 5385 bool AsmParser::parseDirectiveWarning(SMLoc L) { 5386 if (!TheCondStack.empty()) { 5387 if (TheCondStack.back().Ignore) { 5388 eatToEndOfStatement(); 5389 return false; 5390 } 5391 } 5392 5393 StringRef Message = ".warning directive invoked in source file"; 5394 5395 if (!parseOptionalToken(AsmToken::EndOfStatement)) { 5396 if (Lexer.isNot(AsmToken::String)) 5397 return TokError(".warning argument must be a string"); 5398 5399 Message = getTok().getStringContents(); 5400 Lex(); 5401 if (parseEOL()) 5402 return true; 5403 } 5404 5405 return Warning(L, Message); 5406 } 5407 5408 /// parseDirectiveEndIf 5409 /// ::= .endif 5410 bool AsmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { 5411 if (parseEOL()) 5412 return true; 5413 5414 if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) 5415 return Error(DirectiveLoc, "Encountered a .endif that doesn't follow " 5416 "an .if or .else"); 5417 if (!TheCondStack.empty()) { 5418 TheCondState = TheCondStack.back(); 5419 TheCondStack.pop_back(); 5420 } 5421 5422 return false; 5423 } 5424 5425 void AsmParser::initializeDirectiveKindMap() { 5426 /* Lookup will be done with the directive 5427 * converted to lower case, so all these 5428 * keys should be lower case. 5429 * (target specific directives are handled 5430 * elsewhere) 5431 */ 5432 DirectiveKindMap[".set"] = DK_SET; 5433 DirectiveKindMap[".equ"] = DK_EQU; 5434 DirectiveKindMap[".equiv"] = DK_EQUIV; 5435 DirectiveKindMap[".ascii"] = DK_ASCII; 5436 DirectiveKindMap[".asciz"] = DK_ASCIZ; 5437 DirectiveKindMap[".string"] = DK_STRING; 5438 DirectiveKindMap[".byte"] = DK_BYTE; 5439 DirectiveKindMap[".short"] = DK_SHORT; 5440 DirectiveKindMap[".value"] = DK_VALUE; 5441 DirectiveKindMap[".2byte"] = DK_2BYTE; 5442 DirectiveKindMap[".long"] = DK_LONG; 5443 DirectiveKindMap[".int"] = DK_INT; 5444 DirectiveKindMap[".4byte"] = DK_4BYTE; 5445 DirectiveKindMap[".quad"] = DK_QUAD; 5446 DirectiveKindMap[".8byte"] = DK_8BYTE; 5447 DirectiveKindMap[".octa"] = DK_OCTA; 5448 DirectiveKindMap[".single"] = DK_SINGLE; 5449 DirectiveKindMap[".float"] = DK_FLOAT; 5450 DirectiveKindMap[".double"] = DK_DOUBLE; 5451 DirectiveKindMap[".align"] = DK_ALIGN; 5452 DirectiveKindMap[".align32"] = DK_ALIGN32; 5453 DirectiveKindMap[".balign"] = DK_BALIGN; 5454 DirectiveKindMap[".balignw"] = DK_BALIGNW; 5455 DirectiveKindMap[".balignl"] = DK_BALIGNL; 5456 DirectiveKindMap[".p2align"] = DK_P2ALIGN; 5457 DirectiveKindMap[".p2alignw"] = DK_P2ALIGNW; 5458 DirectiveKindMap[".p2alignl"] = DK_P2ALIGNL; 5459 DirectiveKindMap[".org"] = DK_ORG; 5460 DirectiveKindMap[".fill"] = DK_FILL; 5461 DirectiveKindMap[".zero"] = DK_ZERO; 5462 DirectiveKindMap[".extern"] = DK_EXTERN; 5463 DirectiveKindMap[".globl"] = DK_GLOBL; 5464 DirectiveKindMap[".global"] = DK_GLOBAL; 5465 DirectiveKindMap[".lazy_reference"] = DK_LAZY_REFERENCE; 5466 DirectiveKindMap[".no_dead_strip"] = DK_NO_DEAD_STRIP; 5467 DirectiveKindMap[".symbol_resolver"] = DK_SYMBOL_RESOLVER; 5468 DirectiveKindMap[".private_extern"] = DK_PRIVATE_EXTERN; 5469 DirectiveKindMap[".reference"] = DK_REFERENCE; 5470 DirectiveKindMap[".weak_definition"] = DK_WEAK_DEFINITION; 5471 DirectiveKindMap[".weak_reference"] = DK_WEAK_REFERENCE; 5472 DirectiveKindMap[".weak_def_can_be_hidden"] = DK_WEAK_DEF_CAN_BE_HIDDEN; 5473 DirectiveKindMap[".cold"] = DK_COLD; 5474 DirectiveKindMap[".comm"] = DK_COMM; 5475 DirectiveKindMap[".common"] = DK_COMMON; 5476 DirectiveKindMap[".lcomm"] = DK_LCOMM; 5477 DirectiveKindMap[".abort"] = DK_ABORT; 5478 DirectiveKindMap[".include"] = DK_INCLUDE; 5479 DirectiveKindMap[".incbin"] = DK_INCBIN; 5480 DirectiveKindMap[".code16"] = DK_CODE16; 5481 DirectiveKindMap[".code16gcc"] = DK_CODE16GCC; 5482 DirectiveKindMap[".rept"] = DK_REPT; 5483 DirectiveKindMap[".rep"] = DK_REPT; 5484 DirectiveKindMap[".irp"] = DK_IRP; 5485 DirectiveKindMap[".irpc"] = DK_IRPC; 5486 DirectiveKindMap[".endr"] = DK_ENDR; 5487 DirectiveKindMap[".bundle_align_mode"] = DK_BUNDLE_ALIGN_MODE; 5488 DirectiveKindMap[".bundle_lock"] = DK_BUNDLE_LOCK; 5489 DirectiveKindMap[".bundle_unlock"] = DK_BUNDLE_UNLOCK; 5490 DirectiveKindMap[".if"] = DK_IF; 5491 DirectiveKindMap[".ifeq"] = DK_IFEQ; 5492 DirectiveKindMap[".ifge"] = DK_IFGE; 5493 DirectiveKindMap[".ifgt"] = DK_IFGT; 5494 DirectiveKindMap[".ifle"] = DK_IFLE; 5495 DirectiveKindMap[".iflt"] = DK_IFLT; 5496 DirectiveKindMap[".ifne"] = DK_IFNE; 5497 DirectiveKindMap[".ifb"] = DK_IFB; 5498 DirectiveKindMap[".ifnb"] = DK_IFNB; 5499 DirectiveKindMap[".ifc"] = DK_IFC; 5500 DirectiveKindMap[".ifeqs"] = DK_IFEQS; 5501 DirectiveKindMap[".ifnc"] = DK_IFNC; 5502 DirectiveKindMap[".ifnes"] = DK_IFNES; 5503 DirectiveKindMap[".ifdef"] = DK_IFDEF; 5504 DirectiveKindMap[".ifndef"] = DK_IFNDEF; 5505 DirectiveKindMap[".ifnotdef"] = DK_IFNOTDEF; 5506 DirectiveKindMap[".elseif"] = DK_ELSEIF; 5507 DirectiveKindMap[".else"] = DK_ELSE; 5508 DirectiveKindMap[".end"] = DK_END; 5509 DirectiveKindMap[".endif"] = DK_ENDIF; 5510 DirectiveKindMap[".skip"] = DK_SKIP; 5511 DirectiveKindMap[".space"] = DK_SPACE; 5512 DirectiveKindMap[".file"] = DK_FILE; 5513 DirectiveKindMap[".line"] = DK_LINE; 5514 DirectiveKindMap[".loc"] = DK_LOC; 5515 DirectiveKindMap[".stabs"] = DK_STABS; 5516 DirectiveKindMap[".cv_file"] = DK_CV_FILE; 5517 DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; 5518 DirectiveKindMap[".cv_loc"] = DK_CV_LOC; 5519 DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; 5520 DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; 5521 DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; 5522 DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; 5523 DirectiveKindMap[".cv_string"] = DK_CV_STRING; 5524 DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; 5525 DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; 5526 DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; 5527 DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; 5528 DirectiveKindMap[".sleb128"] = DK_SLEB128; 5529 DirectiveKindMap[".uleb128"] = DK_ULEB128; 5530 DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; 5531 DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; 5532 DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; 5533 DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; 5534 DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; 5535 DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; 5536 DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; 5537 DirectiveKindMap[".cfi_llvm_def_aspace_cfa"] = DK_CFI_LLVM_DEF_ASPACE_CFA; 5538 DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; 5539 DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; 5540 DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; 5541 DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; 5542 DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; 5543 DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; 5544 DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; 5545 DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; 5546 DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; 5547 DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; 5548 DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; 5549 DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; 5550 DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; 5551 DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; 5552 DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; 5553 DirectiveKindMap[".macros_on"] = DK_MACROS_ON; 5554 DirectiveKindMap[".macros_off"] = DK_MACROS_OFF; 5555 DirectiveKindMap[".macro"] = DK_MACRO; 5556 DirectiveKindMap[".exitm"] = DK_EXITM; 5557 DirectiveKindMap[".endm"] = DK_ENDM; 5558 DirectiveKindMap[".endmacro"] = DK_ENDMACRO; 5559 DirectiveKindMap[".purgem"] = DK_PURGEM; 5560 DirectiveKindMap[".err"] = DK_ERR; 5561 DirectiveKindMap[".error"] = DK_ERROR; 5562 DirectiveKindMap[".warning"] = DK_WARNING; 5563 DirectiveKindMap[".altmacro"] = DK_ALTMACRO; 5564 DirectiveKindMap[".noaltmacro"] = DK_NOALTMACRO; 5565 DirectiveKindMap[".reloc"] = DK_RELOC; 5566 DirectiveKindMap[".dc"] = DK_DC; 5567 DirectiveKindMap[".dc.a"] = DK_DC_A; 5568 DirectiveKindMap[".dc.b"] = DK_DC_B; 5569 DirectiveKindMap[".dc.d"] = DK_DC_D; 5570 DirectiveKindMap[".dc.l"] = DK_DC_L; 5571 DirectiveKindMap[".dc.s"] = DK_DC_S; 5572 DirectiveKindMap[".dc.w"] = DK_DC_W; 5573 DirectiveKindMap[".dc.x"] = DK_DC_X; 5574 DirectiveKindMap[".dcb"] = DK_DCB; 5575 DirectiveKindMap[".dcb.b"] = DK_DCB_B; 5576 DirectiveKindMap[".dcb.d"] = DK_DCB_D; 5577 DirectiveKindMap[".dcb.l"] = DK_DCB_L; 5578 DirectiveKindMap[".dcb.s"] = DK_DCB_S; 5579 DirectiveKindMap[".dcb.w"] = DK_DCB_W; 5580 DirectiveKindMap[".dcb.x"] = DK_DCB_X; 5581 DirectiveKindMap[".ds"] = DK_DS; 5582 DirectiveKindMap[".ds.b"] = DK_DS_B; 5583 DirectiveKindMap[".ds.d"] = DK_DS_D; 5584 DirectiveKindMap[".ds.l"] = DK_DS_L; 5585 DirectiveKindMap[".ds.p"] = DK_DS_P; 5586 DirectiveKindMap[".ds.s"] = DK_DS_S; 5587 DirectiveKindMap[".ds.w"] = DK_DS_W; 5588 DirectiveKindMap[".ds.x"] = DK_DS_X; 5589 DirectiveKindMap[".print"] = DK_PRINT; 5590 DirectiveKindMap[".addrsig"] = DK_ADDRSIG; 5591 DirectiveKindMap[".addrsig_sym"] = DK_ADDRSIG_SYM; 5592 DirectiveKindMap[".pseudoprobe"] = DK_PSEUDO_PROBE; 5593 DirectiveKindMap[".lto_discard"] = DK_LTO_DISCARD; 5594 DirectiveKindMap[".lto_set_conditional"] = DK_LTO_SET_CONDITIONAL; 5595 } 5596 5597 MCAsmMacro *AsmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { 5598 AsmToken EndToken, StartToken = getTok(); 5599 5600 unsigned NestLevel = 0; 5601 while (true) { 5602 // Check whether we have reached the end of the file. 5603 if (getLexer().is(AsmToken::Eof)) { 5604 printError(DirectiveLoc, "no matching '.endr' in definition"); 5605 return nullptr; 5606 } 5607 5608 if (Lexer.is(AsmToken::Identifier) && 5609 (getTok().getIdentifier() == ".rep" || 5610 getTok().getIdentifier() == ".rept" || 5611 getTok().getIdentifier() == ".irp" || 5612 getTok().getIdentifier() == ".irpc")) { 5613 ++NestLevel; 5614 } 5615 5616 // Otherwise, check whether we have reached the .endr. 5617 if (Lexer.is(AsmToken::Identifier) && getTok().getIdentifier() == ".endr") { 5618 if (NestLevel == 0) { 5619 EndToken = getTok(); 5620 Lex(); 5621 if (Lexer.isNot(AsmToken::EndOfStatement)) { 5622 printError(getTok().getLoc(), 5623 "unexpected token in '.endr' directive"); 5624 return nullptr; 5625 } 5626 break; 5627 } 5628 --NestLevel; 5629 } 5630 5631 // Otherwise, scan till the end of the statement. 5632 eatToEndOfStatement(); 5633 } 5634 5635 const char *BodyStart = StartToken.getLoc().getPointer(); 5636 const char *BodyEnd = EndToken.getLoc().getPointer(); 5637 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 5638 5639 // We Are Anonymous. 5640 MacroLikeBodies.emplace_back(StringRef(), Body, MCAsmMacroParameters()); 5641 return &MacroLikeBodies.back(); 5642 } 5643 5644 void AsmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, 5645 raw_svector_ostream &OS) { 5646 OS << ".endr\n"; 5647 5648 std::unique_ptr<MemoryBuffer> Instantiation = 5649 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 5650 5651 // Create the macro instantiation object and add to the current macro 5652 // instantiation stack. 5653 MacroInstantiation *MI = new MacroInstantiation{ 5654 DirectiveLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()}; 5655 ActiveMacros.push_back(MI); 5656 5657 // Jump to the macro instantiation and prime the lexer. 5658 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc()); 5659 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer()); 5660 Lex(); 5661 } 5662 5663 /// parseDirectiveRept 5664 /// ::= .rep | .rept count 5665 bool AsmParser::parseDirectiveRept(SMLoc DirectiveLoc, StringRef Dir) { 5666 const MCExpr *CountExpr; 5667 SMLoc CountLoc = getTok().getLoc(); 5668 if (parseExpression(CountExpr)) 5669 return true; 5670 5671 int64_t Count; 5672 if (!CountExpr->evaluateAsAbsolute(Count, getStreamer().getAssemblerPtr())) { 5673 return Error(CountLoc, "unexpected token in '" + Dir + "' directive"); 5674 } 5675 5676 if (check(Count < 0, CountLoc, "Count is negative") || parseEOL()) 5677 return true; 5678 5679 // Lex the rept definition. 5680 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5681 if (!M) 5682 return true; 5683 5684 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5685 // to hold the macro body with substitutions. 5686 SmallString<256> Buf; 5687 raw_svector_ostream OS(Buf); 5688 while (Count--) { 5689 // Note that the AtPseudoVariable is disabled for instantiations of .rep(t). 5690 if (expandMacro(OS, M->Body, None, None, false, getTok().getLoc())) 5691 return true; 5692 } 5693 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5694 5695 return false; 5696 } 5697 5698 /// parseDirectiveIrp 5699 /// ::= .irp symbol,values 5700 bool AsmParser::parseDirectiveIrp(SMLoc DirectiveLoc) { 5701 MCAsmMacroParameter Parameter; 5702 MCAsmMacroArguments A; 5703 if (check(parseIdentifier(Parameter.Name), 5704 "expected identifier in '.irp' directive") || 5705 parseComma() || parseMacroArguments(nullptr, A) || parseEOL()) 5706 return true; 5707 5708 // Lex the irp definition. 5709 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5710 if (!M) 5711 return true; 5712 5713 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5714 // to hold the macro body with substitutions. 5715 SmallString<256> Buf; 5716 raw_svector_ostream OS(Buf); 5717 5718 for (const MCAsmMacroArgument &Arg : A) { 5719 // Note that the AtPseudoVariable is enabled for instantiations of .irp. 5720 // This is undocumented, but GAS seems to support it. 5721 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5722 return true; 5723 } 5724 5725 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5726 5727 return false; 5728 } 5729 5730 /// parseDirectiveIrpc 5731 /// ::= .irpc symbol,values 5732 bool AsmParser::parseDirectiveIrpc(SMLoc DirectiveLoc) { 5733 MCAsmMacroParameter Parameter; 5734 MCAsmMacroArguments A; 5735 5736 if (check(parseIdentifier(Parameter.Name), 5737 "expected identifier in '.irpc' directive") || 5738 parseComma() || parseMacroArguments(nullptr, A)) 5739 return true; 5740 5741 if (A.size() != 1 || A.front().size() != 1) 5742 return TokError("unexpected token in '.irpc' directive"); 5743 if (parseEOL()) 5744 return true; 5745 5746 // Lex the irpc definition. 5747 MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); 5748 if (!M) 5749 return true; 5750 5751 // Macro instantiation is lexical, unfortunately. We construct a new buffer 5752 // to hold the macro body with substitutions. 5753 SmallString<256> Buf; 5754 raw_svector_ostream OS(Buf); 5755 5756 StringRef Values = A.front().front().getString(); 5757 for (std::size_t I = 0, End = Values.size(); I != End; ++I) { 5758 MCAsmMacroArgument Arg; 5759 Arg.emplace_back(AsmToken::Identifier, Values.slice(I, I + 1)); 5760 5761 // Note that the AtPseudoVariable is enabled for instantiations of .irpc. 5762 // This is undocumented, but GAS seems to support it. 5763 if (expandMacro(OS, M->Body, Parameter, Arg, true, getTok().getLoc())) 5764 return true; 5765 } 5766 5767 instantiateMacroLikeBody(M, DirectiveLoc, OS); 5768 5769 return false; 5770 } 5771 5772 bool AsmParser::parseDirectiveEndr(SMLoc DirectiveLoc) { 5773 if (ActiveMacros.empty()) 5774 return TokError("unmatched '.endr' directive"); 5775 5776 // The only .repl that should get here are the ones created by 5777 // instantiateMacroLikeBody. 5778 assert(getLexer().is(AsmToken::EndOfStatement)); 5779 5780 handleMacroExit(); 5781 return false; 5782 } 5783 5784 bool AsmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, 5785 size_t Len) { 5786 const MCExpr *Value; 5787 SMLoc ExprLoc = getLexer().getLoc(); 5788 if (parseExpression(Value)) 5789 return true; 5790 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5791 if (!MCE) 5792 return Error(ExprLoc, "unexpected expression in _emit"); 5793 uint64_t IntValue = MCE->getValue(); 5794 if (!isUInt<8>(IntValue) && !isInt<8>(IntValue)) 5795 return Error(ExprLoc, "literal value out of range for directive"); 5796 5797 Info.AsmRewrites->emplace_back(AOK_Emit, IDLoc, Len); 5798 return false; 5799 } 5800 5801 bool AsmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { 5802 const MCExpr *Value; 5803 SMLoc ExprLoc = getLexer().getLoc(); 5804 if (parseExpression(Value)) 5805 return true; 5806 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value); 5807 if (!MCE) 5808 return Error(ExprLoc, "unexpected expression in align"); 5809 uint64_t IntValue = MCE->getValue(); 5810 if (!isPowerOf2_64(IntValue)) 5811 return Error(ExprLoc, "literal value not a power of two greater then zero"); 5812 5813 Info.AsmRewrites->emplace_back(AOK_Align, IDLoc, 5, Log2_64(IntValue)); 5814 return false; 5815 } 5816 5817 bool AsmParser::parseDirectivePrint(SMLoc DirectiveLoc) { 5818 const AsmToken StrTok = getTok(); 5819 Lex(); 5820 if (StrTok.isNot(AsmToken::String) || StrTok.getString().front() != '"') 5821 return Error(DirectiveLoc, "expected double quoted string after .print"); 5822 if (parseEOL()) 5823 return true; 5824 llvm::outs() << StrTok.getStringContents() << '\n'; 5825 return false; 5826 } 5827 5828 bool AsmParser::parseDirectiveAddrsig() { 5829 if (parseEOL()) 5830 return true; 5831 getStreamer().emitAddrsig(); 5832 return false; 5833 } 5834 5835 bool AsmParser::parseDirectiveAddrsigSym() { 5836 StringRef Name; 5837 if (check(parseIdentifier(Name), "expected identifier") || parseEOL()) 5838 return true; 5839 MCSymbol *Sym = getContext().getOrCreateSymbol(Name); 5840 getStreamer().emitAddrsigSym(Sym); 5841 return false; 5842 } 5843 5844 bool AsmParser::parseDirectivePseudoProbe() { 5845 int64_t Guid; 5846 int64_t Index; 5847 int64_t Type; 5848 int64_t Attr; 5849 5850 if (getLexer().is(AsmToken::Integer)) { 5851 if (parseIntToken(Guid, "unexpected token in '.pseudoprobe' directive")) 5852 return true; 5853 } 5854 5855 if (getLexer().is(AsmToken::Integer)) { 5856 if (parseIntToken(Index, "unexpected token in '.pseudoprobe' directive")) 5857 return true; 5858 } 5859 5860 if (getLexer().is(AsmToken::Integer)) { 5861 if (parseIntToken(Type, "unexpected token in '.pseudoprobe' directive")) 5862 return true; 5863 } 5864 5865 if (getLexer().is(AsmToken::Integer)) { 5866 if (parseIntToken(Attr, "unexpected token in '.pseudoprobe' directive")) 5867 return true; 5868 } 5869 5870 // Parse inline stack like @ GUID:11:12 @ GUID:1:11 @ GUID:3:21 5871 MCPseudoProbeInlineStack InlineStack; 5872 5873 while (getLexer().is(AsmToken::At)) { 5874 // eat @ 5875 Lex(); 5876 5877 int64_t CallerGuid = 0; 5878 if (getLexer().is(AsmToken::Integer)) { 5879 if (parseIntToken(CallerGuid, 5880 "unexpected token in '.pseudoprobe' directive")) 5881 return true; 5882 } 5883 5884 // eat colon 5885 if (getLexer().is(AsmToken::Colon)) 5886 Lex(); 5887 5888 int64_t CallerProbeId = 0; 5889 if (getLexer().is(AsmToken::Integer)) { 5890 if (parseIntToken(CallerProbeId, 5891 "unexpected token in '.pseudoprobe' directive")) 5892 return true; 5893 } 5894 5895 InlineSite Site(CallerGuid, CallerProbeId); 5896 InlineStack.push_back(Site); 5897 } 5898 5899 if (parseEOL()) 5900 return true; 5901 5902 getStreamer().emitPseudoProbe(Guid, Index, Type, Attr, InlineStack); 5903 return false; 5904 } 5905 5906 /// parseDirectiveLTODiscard 5907 /// ::= ".lto_discard" [ identifier ( , identifier )* ] 5908 /// The LTO library emits this directive to discard non-prevailing symbols. 5909 /// We ignore symbol assignments and attribute changes for the specified 5910 /// symbols. 5911 bool AsmParser::parseDirectiveLTODiscard() { 5912 auto ParseOp = [&]() -> bool { 5913 StringRef Name; 5914 SMLoc Loc = getTok().getLoc(); 5915 if (parseIdentifier(Name)) 5916 return Error(Loc, "expected identifier"); 5917 LTODiscardSymbols.insert(Name); 5918 return false; 5919 }; 5920 5921 LTODiscardSymbols.clear(); 5922 return parseMany(ParseOp); 5923 } 5924 5925 // We are comparing pointers, but the pointers are relative to a single string. 5926 // Thus, this should always be deterministic. 5927 static int rewritesSort(const AsmRewrite *AsmRewriteA, 5928 const AsmRewrite *AsmRewriteB) { 5929 if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) 5930 return -1; 5931 if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) 5932 return 1; 5933 5934 // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output 5935 // rewrite to the same location. Make sure the SizeDirective rewrite is 5936 // performed first, then the Imm/ImmPrefix and finally the Input/Output. This 5937 // ensures the sort algorithm is stable. 5938 if (AsmRewritePrecedence[AsmRewriteA->Kind] > 5939 AsmRewritePrecedence[AsmRewriteB->Kind]) 5940 return -1; 5941 5942 if (AsmRewritePrecedence[AsmRewriteA->Kind] < 5943 AsmRewritePrecedence[AsmRewriteB->Kind]) 5944 return 1; 5945 llvm_unreachable("Unstable rewrite sort."); 5946 } 5947 5948 bool AsmParser::parseMSInlineAsm( 5949 std::string &AsmString, unsigned &NumOutputs, unsigned &NumInputs, 5950 SmallVectorImpl<std::pair<void *, bool>> &OpDecls, 5951 SmallVectorImpl<std::string> &Constraints, 5952 SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, 5953 const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { 5954 SmallVector<void *, 4> InputDecls; 5955 SmallVector<void *, 4> OutputDecls; 5956 SmallVector<bool, 4> InputDeclsAddressOf; 5957 SmallVector<bool, 4> OutputDeclsAddressOf; 5958 SmallVector<std::string, 4> InputConstraints; 5959 SmallVector<std::string, 4> OutputConstraints; 5960 SmallVector<unsigned, 4> ClobberRegs; 5961 5962 SmallVector<AsmRewrite, 4> AsmStrRewrites; 5963 5964 // Prime the lexer. 5965 Lex(); 5966 5967 // While we have input, parse each statement. 5968 unsigned InputIdx = 0; 5969 unsigned OutputIdx = 0; 5970 while (getLexer().isNot(AsmToken::Eof)) { 5971 // Parse curly braces marking block start/end 5972 if (parseCurlyBlockScope(AsmStrRewrites)) 5973 continue; 5974 5975 ParseStatementInfo Info(&AsmStrRewrites); 5976 bool StatementErr = parseStatement(Info, &SI); 5977 5978 if (StatementErr || Info.ParseError) { 5979 // Emit pending errors if any exist. 5980 printPendingErrors(); 5981 return true; 5982 } 5983 5984 // No pending error should exist here. 5985 assert(!hasPendingError() && "unexpected error from parseStatement"); 5986 5987 if (Info.Opcode == ~0U) 5988 continue; 5989 5990 const MCInstrDesc &Desc = MII->get(Info.Opcode); 5991 5992 // Build the list of clobbers, outputs and inputs. 5993 for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { 5994 MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; 5995 5996 // Register operand. 5997 if (Operand.isReg() && !Operand.needAddressOf() && 5998 !getTargetParser().OmitRegisterFromClobberLists(Operand.getReg())) { 5999 unsigned NumDefs = Desc.getNumDefs(); 6000 // Clobber. 6001 if (NumDefs && Operand.getMCOperandNum() < NumDefs) 6002 ClobberRegs.push_back(Operand.getReg()); 6003 continue; 6004 } 6005 6006 // Expr/Input or Output. 6007 StringRef SymName = Operand.getSymName(); 6008 if (SymName.empty()) 6009 continue; 6010 6011 void *OpDecl = Operand.getOpDecl(); 6012 if (!OpDecl) 6013 continue; 6014 6015 StringRef Constraint = Operand.getConstraint(); 6016 if (Operand.isImm()) { 6017 // Offset as immediate 6018 if (Operand.isOffsetOfLocal()) 6019 Constraint = "r"; 6020 else 6021 Constraint = "i"; 6022 } 6023 6024 bool isOutput = (i == 1) && Desc.mayStore(); 6025 SMLoc Start = SMLoc::getFromPointer(SymName.data()); 6026 int64_t Size = Operand.isMemPlaceholder(Desc) ? 0 : SymName.size(); 6027 if (isOutput) { 6028 ++InputIdx; 6029 OutputDecls.push_back(OpDecl); 6030 OutputDeclsAddressOf.push_back(Operand.needAddressOf()); 6031 OutputConstraints.push_back(("=" + Constraint).str()); 6032 AsmStrRewrites.emplace_back(AOK_Output, Start, Size); 6033 } else { 6034 InputDecls.push_back(OpDecl); 6035 InputDeclsAddressOf.push_back(Operand.needAddressOf()); 6036 InputConstraints.push_back(Constraint.str()); 6037 if (Desc.OpInfo[i - 1].isBranchTarget()) 6038 AsmStrRewrites.emplace_back(AOK_CallInput, Start, SymName.size()); 6039 else 6040 AsmStrRewrites.emplace_back(AOK_Input, Start, Size); 6041 } 6042 } 6043 6044 // Consider implicit defs to be clobbers. Think of cpuid and push. 6045 ArrayRef<MCPhysReg> ImpDefs(Desc.getImplicitDefs(), 6046 Desc.getNumImplicitDefs()); 6047 llvm::append_range(ClobberRegs, ImpDefs); 6048 } 6049 6050 // Set the number of Outputs and Inputs. 6051 NumOutputs = OutputDecls.size(); 6052 NumInputs = InputDecls.size(); 6053 6054 // Set the unique clobbers. 6055 array_pod_sort(ClobberRegs.begin(), ClobberRegs.end()); 6056 ClobberRegs.erase(std::unique(ClobberRegs.begin(), ClobberRegs.end()), 6057 ClobberRegs.end()); 6058 Clobbers.assign(ClobberRegs.size(), std::string()); 6059 for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { 6060 raw_string_ostream OS(Clobbers[I]); 6061 IP->printRegName(OS, ClobberRegs[I]); 6062 } 6063 6064 // Merge the various outputs and inputs. Output are expected first. 6065 if (NumOutputs || NumInputs) { 6066 unsigned NumExprs = NumOutputs + NumInputs; 6067 OpDecls.resize(NumExprs); 6068 Constraints.resize(NumExprs); 6069 for (unsigned i = 0; i < NumOutputs; ++i) { 6070 OpDecls[i] = std::make_pair(OutputDecls[i], OutputDeclsAddressOf[i]); 6071 Constraints[i] = OutputConstraints[i]; 6072 } 6073 for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { 6074 OpDecls[j] = std::make_pair(InputDecls[i], InputDeclsAddressOf[i]); 6075 Constraints[j] = InputConstraints[i]; 6076 } 6077 } 6078 6079 // Build the IR assembly string. 6080 std::string AsmStringIR; 6081 raw_string_ostream OS(AsmStringIR); 6082 StringRef ASMString = 6083 SrcMgr.getMemoryBuffer(SrcMgr.getMainFileID())->getBuffer(); 6084 const char *AsmStart = ASMString.begin(); 6085 const char *AsmEnd = ASMString.end(); 6086 array_pod_sort(AsmStrRewrites.begin(), AsmStrRewrites.end(), rewritesSort); 6087 for (auto it = AsmStrRewrites.begin(); it != AsmStrRewrites.end(); ++it) { 6088 const AsmRewrite &AR = *it; 6089 // Check if this has already been covered by another rewrite... 6090 if (AR.Done) 6091 continue; 6092 AsmRewriteKind Kind = AR.Kind; 6093 6094 const char *Loc = AR.Loc.getPointer(); 6095 assert(Loc >= AsmStart && "Expected Loc to be at or after Start!"); 6096 6097 // Emit everything up to the immediate/expression. 6098 if (unsigned Len = Loc - AsmStart) 6099 OS << StringRef(AsmStart, Len); 6100 6101 // Skip the original expression. 6102 if (Kind == AOK_Skip) { 6103 AsmStart = Loc + AR.Len; 6104 continue; 6105 } 6106 6107 unsigned AdditionalSkip = 0; 6108 // Rewrite expressions in $N notation. 6109 switch (Kind) { 6110 default: 6111 break; 6112 case AOK_IntelExpr: 6113 assert(AR.IntelExp.isValid() && "cannot write invalid intel expression"); 6114 if (AR.IntelExp.NeedBracs) 6115 OS << "["; 6116 if (AR.IntelExp.hasBaseReg()) 6117 OS << AR.IntelExp.BaseReg; 6118 if (AR.IntelExp.hasIndexReg()) 6119 OS << (AR.IntelExp.hasBaseReg() ? " + " : "") 6120 << AR.IntelExp.IndexReg; 6121 if (AR.IntelExp.Scale > 1) 6122 OS << " * $$" << AR.IntelExp.Scale; 6123 if (AR.IntelExp.hasOffset()) { 6124 if (AR.IntelExp.hasRegs()) 6125 OS << " + "; 6126 // Fuse this rewrite with a rewrite of the offset name, if present. 6127 StringRef OffsetName = AR.IntelExp.OffsetName; 6128 SMLoc OffsetLoc = SMLoc::getFromPointer(AR.IntelExp.OffsetName.data()); 6129 size_t OffsetLen = OffsetName.size(); 6130 auto rewrite_it = std::find_if( 6131 it, AsmStrRewrites.end(), [&](const AsmRewrite &FusingAR) { 6132 return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && 6133 (FusingAR.Kind == AOK_Input || 6134 FusingAR.Kind == AOK_CallInput); 6135 }); 6136 if (rewrite_it == AsmStrRewrites.end()) { 6137 OS << "offset " << OffsetName; 6138 } else if (rewrite_it->Kind == AOK_CallInput) { 6139 OS << "${" << InputIdx++ << ":P}"; 6140 rewrite_it->Done = true; 6141 } else { 6142 OS << '$' << InputIdx++; 6143 rewrite_it->Done = true; 6144 } 6145 } 6146 if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) 6147 OS << (AR.IntelExp.emitImm() ? "$$" : " + $$") << AR.IntelExp.Imm; 6148 if (AR.IntelExp.NeedBracs) 6149 OS << "]"; 6150 break; 6151 case AOK_Label: 6152 OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; 6153 break; 6154 case AOK_Input: 6155 if (AR.Len) 6156 OS << '$' << InputIdx; 6157 ++InputIdx; 6158 break; 6159 case AOK_CallInput: 6160 OS << "${" << InputIdx++ << ":P}"; 6161 break; 6162 case AOK_Output: 6163 if (AR.Len) 6164 OS << '$' << OutputIdx; 6165 ++OutputIdx; 6166 break; 6167 case AOK_SizeDirective: 6168 switch (AR.Val) { 6169 default: break; 6170 case 8: OS << "byte ptr "; break; 6171 case 16: OS << "word ptr "; break; 6172 case 32: OS << "dword ptr "; break; 6173 case 64: OS << "qword ptr "; break; 6174 case 80: OS << "xword ptr "; break; 6175 case 128: OS << "xmmword ptr "; break; 6176 case 256: OS << "ymmword ptr "; break; 6177 } 6178 break; 6179 case AOK_Emit: 6180 OS << ".byte"; 6181 break; 6182 case AOK_Align: { 6183 // MS alignment directives are measured in bytes. If the native assembler 6184 // measures alignment in bytes, we can pass it straight through. 6185 OS << ".align"; 6186 if (getContext().getAsmInfo()->getAlignmentIsInBytes()) 6187 break; 6188 6189 // Alignment is in log2 form, so print that instead and skip the original 6190 // immediate. 6191 unsigned Val = AR.Val; 6192 OS << ' ' << Val; 6193 assert(Val < 10 && "Expected alignment less then 2^10."); 6194 AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; 6195 break; 6196 } 6197 case AOK_EVEN: 6198 OS << ".even"; 6199 break; 6200 case AOK_EndOfStatement: 6201 OS << "\n\t"; 6202 break; 6203 } 6204 6205 // Skip the original expression. 6206 AsmStart = Loc + AR.Len + AdditionalSkip; 6207 } 6208 6209 // Emit the remainder of the asm string. 6210 if (AsmStart != AsmEnd) 6211 OS << StringRef(AsmStart, AsmEnd - AsmStart); 6212 6213 AsmString = OS.str(); 6214 return false; 6215 } 6216 6217 bool HLASMAsmParser::parseAsHLASMLabel(ParseStatementInfo &Info, 6218 MCAsmParserSemaCallback *SI) { 6219 AsmToken LabelTok = getTok(); 6220 SMLoc LabelLoc = LabelTok.getLoc(); 6221 StringRef LabelVal; 6222 6223 if (parseIdentifier(LabelVal)) 6224 return Error(LabelLoc, "The HLASM Label has to be an Identifier"); 6225 6226 // We have validated whether the token is an Identifier. 6227 // Now we have to validate whether the token is a 6228 // valid HLASM Label. 6229 if (!getTargetParser().isLabel(LabelTok) || checkForValidSection()) 6230 return true; 6231 6232 // Lex leading spaces to get to the next operand. 6233 lexLeadingSpaces(); 6234 6235 // We shouldn't emit the label if there is nothing else after the label. 6236 // i.e asm("<token>\n") 6237 if (getTok().is(AsmToken::EndOfStatement)) 6238 return Error(LabelLoc, 6239 "Cannot have just a label for an HLASM inline asm statement"); 6240 6241 MCSymbol *Sym = getContext().getOrCreateSymbol( 6242 getContext().getAsmInfo()->shouldEmitLabelsInUpperCase() 6243 ? LabelVal.upper() 6244 : LabelVal); 6245 6246 getTargetParser().doBeforeLabelEmit(Sym); 6247 6248 // Emit the label. 6249 Out.emitLabel(Sym, LabelLoc); 6250 6251 // If we are generating dwarf for assembly source files then gather the 6252 // info to make a dwarf label entry for this label if needed. 6253 if (enabledGenDwarfForAssembly()) 6254 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 6255 LabelLoc); 6256 6257 getTargetParser().onLabelParsed(Sym); 6258 6259 return false; 6260 } 6261 6262 bool HLASMAsmParser::parseAsMachineInstruction(ParseStatementInfo &Info, 6263 MCAsmParserSemaCallback *SI) { 6264 AsmToken OperationEntryTok = Lexer.getTok(); 6265 SMLoc OperationEntryLoc = OperationEntryTok.getLoc(); 6266 StringRef OperationEntryVal; 6267 6268 // Attempt to parse the first token as an Identifier 6269 if (parseIdentifier(OperationEntryVal)) 6270 return Error(OperationEntryLoc, "unexpected token at start of statement"); 6271 6272 // Once we've parsed the operation entry successfully, lex 6273 // any spaces to get to the OperandEntries. 6274 lexLeadingSpaces(); 6275 6276 return parseAndMatchAndEmitTargetInstruction( 6277 Info, OperationEntryVal, OperationEntryTok, OperationEntryLoc); 6278 } 6279 6280 bool HLASMAsmParser::parseStatement(ParseStatementInfo &Info, 6281 MCAsmParserSemaCallback *SI) { 6282 assert(!hasPendingError() && "parseStatement started with pending error"); 6283 6284 // Should the first token be interpreted as a HLASM Label. 6285 bool ShouldParseAsHLASMLabel = false; 6286 6287 // If a Name Entry exists, it should occur at the very 6288 // start of the string. In this case, we should parse the 6289 // first non-space token as a Label. 6290 // If the Name entry is missing (i.e. there's some other 6291 // token), then we attempt to parse the first non-space 6292 // token as a Machine Instruction. 6293 if (getTok().isNot(AsmToken::Space)) 6294 ShouldParseAsHLASMLabel = true; 6295 6296 // If we have an EndOfStatement (which includes the target's comment 6297 // string) we can appropriately lex it early on) 6298 if (Lexer.is(AsmToken::EndOfStatement)) { 6299 // if this is a line comment we can drop it safely 6300 if (getTok().getString().empty() || getTok().getString().front() == '\r' || 6301 getTok().getString().front() == '\n') 6302 Out.AddBlankLine(); 6303 Lex(); 6304 return false; 6305 } 6306 6307 // We have established how to parse the inline asm statement. 6308 // Now we can safely lex any leading spaces to get to the 6309 // first token. 6310 lexLeadingSpaces(); 6311 6312 // If we see a new line or carriage return as the first operand, 6313 // after lexing leading spaces, emit the new line and lex the 6314 // EndOfStatement token. 6315 if (Lexer.is(AsmToken::EndOfStatement)) { 6316 if (getTok().getString().front() == '\n' || 6317 getTok().getString().front() == '\r') { 6318 Out.AddBlankLine(); 6319 Lex(); 6320 return false; 6321 } 6322 } 6323 6324 // Handle the label first if we have to before processing the rest 6325 // of the tokens as a machine instruction. 6326 if (ShouldParseAsHLASMLabel) { 6327 // If there were any errors while handling and emitting the label, 6328 // early return. 6329 if (parseAsHLASMLabel(Info, SI)) { 6330 // If we know we've failed in parsing, simply eat until end of the 6331 // statement. This ensures that we don't process any other statements. 6332 eatToEndOfStatement(); 6333 return true; 6334 } 6335 } 6336 6337 return parseAsMachineInstruction(Info, SI); 6338 } 6339 6340 namespace llvm { 6341 namespace MCParserUtils { 6342 6343 /// Returns whether the given symbol is used anywhere in the given expression, 6344 /// or subexpressions. 6345 static bool isSymbolUsedInExpression(const MCSymbol *Sym, const MCExpr *Value) { 6346 switch (Value->getKind()) { 6347 case MCExpr::Binary: { 6348 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr *>(Value); 6349 return isSymbolUsedInExpression(Sym, BE->getLHS()) || 6350 isSymbolUsedInExpression(Sym, BE->getRHS()); 6351 } 6352 case MCExpr::Target: 6353 case MCExpr::Constant: 6354 return false; 6355 case MCExpr::SymbolRef: { 6356 const MCSymbol &S = 6357 static_cast<const MCSymbolRefExpr *>(Value)->getSymbol(); 6358 if (S.isVariable()) 6359 return isSymbolUsedInExpression(Sym, S.getVariableValue()); 6360 return &S == Sym; 6361 } 6362 case MCExpr::Unary: 6363 return isSymbolUsedInExpression( 6364 Sym, static_cast<const MCUnaryExpr *>(Value)->getSubExpr()); 6365 } 6366 6367 llvm_unreachable("Unknown expr kind!"); 6368 } 6369 6370 bool parseAssignmentExpression(StringRef Name, bool allow_redef, 6371 MCAsmParser &Parser, MCSymbol *&Sym, 6372 const MCExpr *&Value) { 6373 6374 // FIXME: Use better location, we should use proper tokens. 6375 SMLoc EqualLoc = Parser.getTok().getLoc(); 6376 if (Parser.parseExpression(Value)) 6377 return Parser.TokError("missing expression"); 6378 6379 // Note: we don't count b as used in "a = b". This is to allow 6380 // a = b 6381 // b = c 6382 6383 if (Parser.parseEOL()) 6384 return true; 6385 6386 // Validate that the LHS is allowed to be a variable (either it has not been 6387 // used as a symbol, or it is an absolute symbol). 6388 Sym = Parser.getContext().lookupSymbol(Name); 6389 if (Sym) { 6390 // Diagnose assignment to a label. 6391 // 6392 // FIXME: Diagnostics. Note the location of the definition as a label. 6393 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 6394 if (isSymbolUsedInExpression(Sym, Value)) 6395 return Parser.Error(EqualLoc, "Recursive use of '" + Name + "'"); 6396 else if (Sym->isUndefined(/*SetUsed*/ false) && !Sym->isUsed() && 6397 !Sym->isVariable()) 6398 ; // Allow redefinitions of undefined symbols only used in directives. 6399 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 6400 ; // Allow redefinitions of variables that haven't yet been used. 6401 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 6402 return Parser.Error(EqualLoc, "redefinition of '" + Name + "'"); 6403 else if (!Sym->isVariable()) 6404 return Parser.Error(EqualLoc, "invalid assignment to '" + Name + "'"); 6405 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 6406 return Parser.Error(EqualLoc, 6407 "invalid reassignment of non-absolute variable '" + 6408 Name + "'"); 6409 } else if (Name == ".") { 6410 Parser.getStreamer().emitValueToOffset(Value, 0, EqualLoc); 6411 return false; 6412 } else 6413 Sym = Parser.getContext().getOrCreateSymbol(Name); 6414 6415 Sym->setRedefinable(allow_redef); 6416 6417 return false; 6418 } 6419 6420 } // end namespace MCParserUtils 6421 } // end namespace llvm 6422 6423 /// Create an MCAsmParser instance. 6424 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, MCContext &C, 6425 MCStreamer &Out, const MCAsmInfo &MAI, 6426 unsigned CB) { 6427 if (C.getTargetTriple().isSystemZ() && C.getTargetTriple().isOSzOS()) 6428 return new HLASMAsmParser(SM, C, Out, MAI, CB); 6429 6430 return new AsmParser(SM, C, Out, MAI, CB); 6431 } 6432