AsmParser.cpp revision aa7a2f2ba308656e206338fe65c422e0b6781c64
1//===- AsmParser.cpp - Parser for Assembly Files --------------------------===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9// 10// This class implements the parser for assembly files. 11// 12//===----------------------------------------------------------------------===// 13 14#include "llvm/ADT/APFloat.h" 15#include "llvm/ADT/SmallString.h" 16#include "llvm/ADT/StringMap.h" 17#include "llvm/ADT/Twine.h" 18#include "llvm/MC/MCAsmInfo.h" 19#include "llvm/MC/MCContext.h" 20#include "llvm/MC/MCDwarf.h" 21#include "llvm/MC/MCExpr.h" 22#include "llvm/MC/MCParser/AsmCond.h" 23#include "llvm/MC/MCParser/AsmLexer.h" 24#include "llvm/MC/MCParser/MCAsmParser.h" 25#include "llvm/MC/MCParser/MCParsedAsmOperand.h" 26#include "llvm/MC/MCRegisterInfo.h" 27#include "llvm/MC/MCSectionMachO.h" 28#include "llvm/MC/MCStreamer.h" 29#include "llvm/MC/MCSymbol.h" 30#include "llvm/MC/MCTargetAsmParser.h" 31#include "llvm/Support/CommandLine.h" 32#include "llvm/Support/ErrorHandling.h" 33#include "llvm/Support/MathExtras.h" 34#include "llvm/Support/MemoryBuffer.h" 35#include "llvm/Support/SourceMgr.h" 36#include "llvm/Support/raw_ostream.h" 37#include <cctype> 38#include <vector> 39using namespace llvm; 40 41static cl::opt<bool> 42FatalAssemblerWarnings("fatal-assembler-warnings", 43 cl::desc("Consider warnings as error")); 44 45namespace { 46 47/// \brief Helper class for tracking macro definitions. 48typedef std::vector<AsmToken> MacroArgument; 49 50struct Macro { 51 StringRef Name; 52 StringRef Body; 53 std::vector<StringRef> Parameters; 54 55public: 56 Macro(StringRef N, StringRef B, const std::vector<StringRef> &P) : 57 Name(N), Body(B), Parameters(P) {} 58}; 59 60/// \brief Helper class for storing information about an active macro 61/// instantiation. 62struct MacroInstantiation { 63 /// The macro being instantiated. 64 const Macro *TheMacro; 65 66 /// The macro instantiation with substitutions. 67 MemoryBuffer *Instantiation; 68 69 /// The location of the instantiation. 70 SMLoc InstantiationLoc; 71 72 /// The location where parsing should resume upon instantiation completion. 73 SMLoc ExitLoc; 74 75public: 76 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 77 MemoryBuffer *I); 78}; 79 80/// \brief The concrete assembly parser instance. 81class AsmParser : public MCAsmParser { 82 friend class GenericAsmParser; 83 84 AsmParser(const AsmParser &); // DO NOT IMPLEMENT 85 void operator=(const AsmParser &); // DO NOT IMPLEMENT 86private: 87 AsmLexer Lexer; 88 MCContext &Ctx; 89 MCStreamer &Out; 90 const MCAsmInfo &MAI; 91 SourceMgr &SrcMgr; 92 SourceMgr::DiagHandlerTy SavedDiagHandler; 93 void *SavedDiagContext; 94 MCAsmParserExtension *GenericParser; 95 MCAsmParserExtension *PlatformParser; 96 97 /// This is the current buffer index we're lexing from as managed by the 98 /// SourceMgr object. 99 int CurBuffer; 100 101 AsmCond TheCondState; 102 std::vector<AsmCond> TheCondStack; 103 104 /// DirectiveMap - This is a table handlers for directives. Each handler is 105 /// invoked after the directive identifier is read and is responsible for 106 /// parsing and validating the rest of the directive. The handler is passed 107 /// in the directive name and the location of the directive keyword. 108 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap; 109 110 /// MacroMap - Map of currently defined macros. 111 StringMap<Macro*> MacroMap; 112 113 /// ActiveMacros - Stack of active macro instantiations. 114 std::vector<MacroInstantiation*> ActiveMacros; 115 116 /// Boolean tracking whether macro substitution is enabled. 117 unsigned MacrosEnabled : 1; 118 119 /// Flag tracking whether any errors have been encountered. 120 unsigned HadError : 1; 121 122 /// The values from the last parsed cpp hash file line comment if any. 123 StringRef CppHashFilename; 124 int64_t CppHashLineNumber; 125 SMLoc CppHashLoc; 126 127 /// AssemblerDialect. ~OU means unset value and use value provided by MAI. 128 unsigned AssemblerDialect; 129 130public: 131 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 132 const MCAsmInfo &MAI); 133 ~AsmParser(); 134 135 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false); 136 137 void AddDirectiveHandler(MCAsmParserExtension *Object, 138 StringRef Directive, 139 DirectiveHandler Handler) { 140 DirectiveMap[Directive] = std::make_pair(Object, Handler); 141 } 142 143public: 144 /// @name MCAsmParser Interface 145 /// { 146 147 virtual SourceMgr &getSourceManager() { return SrcMgr; } 148 virtual MCAsmLexer &getLexer() { return Lexer; } 149 virtual MCContext &getContext() { return Ctx; } 150 virtual MCStreamer &getStreamer() { return Out; } 151 virtual unsigned getAssemblerDialect() { 152 if (AssemblerDialect == ~0U) 153 return MAI.getAssemblerDialect(); 154 else 155 return AssemblerDialect; 156 } 157 virtual void setAssemblerDialect(unsigned i) { 158 AssemblerDialect = i; 159 } 160 161 virtual bool Warning(SMLoc L, const Twine &Msg, 162 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 163 virtual bool Error(SMLoc L, const Twine &Msg, 164 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 165 166 const AsmToken &Lex(); 167 168 bool ParseExpression(const MCExpr *&Res); 169 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc); 170 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc); 171 virtual bool ParseAbsoluteExpression(int64_t &Res); 172 173 /// } 174 175private: 176 void CheckForValidSection(); 177 178 bool ParseStatement(); 179 void EatToEndOfLine(); 180 bool ParseCppHashLineFilenameComment(const SMLoc &L); 181 182 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M); 183 bool expandMacro(raw_svector_ostream &OS, StringRef Body, 184 const std::vector<StringRef> &Parameters, 185 const std::vector<MacroArgument> &A, 186 const SMLoc &L); 187 void HandleMacroExit(); 188 189 void PrintMacroInstantiations(); 190 void PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 191 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) const { 192 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges); 193 } 194 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 195 196 /// EnterIncludeFile - Enter the specified file. This returns true on failure. 197 bool EnterIncludeFile(const std::string &Filename); 198 /// ProcessIncbinFile - Process the specified file for the .incbin directive. 199 /// This returns true on failure. 200 bool ProcessIncbinFile(const std::string &Filename); 201 202 /// \brief Reset the current lexer position to that given by \arg Loc. The 203 /// current token is not set; clients should ensure Lex() is called 204 /// subsequently. 205 void JumpToLoc(SMLoc Loc); 206 207 void EatToEndOfStatement(); 208 209 bool ParseMacroArgument(MacroArgument &MA); 210 bool ParseMacroArguments(const Macro *M, std::vector<MacroArgument> &A); 211 212 /// \brief Parse up to the end of statement and a return the contents from the 213 /// current token until the end of the statement; the current token on exit 214 /// will be either the EndOfStatement or EOF. 215 StringRef ParseStringToEndOfStatement(); 216 217 /// \brief Parse until the end of a statement or a comma is encountered, 218 /// return the contents from the current token up to the end or comma. 219 StringRef ParseStringToComma(); 220 221 bool ParseAssignment(StringRef Name, bool allow_redef); 222 223 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc); 224 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 225 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 226 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 227 228 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier) 229 /// and set \arg Res to the identifier contents. 230 bool ParseIdentifier(StringRef &Res); 231 232 // Directive Parsing. 233 234 // ".ascii", ".asciiz", ".string" 235 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 236 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ... 237 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ... 238 bool ParseDirectiveFill(); // ".fill" 239 bool ParseDirectiveSpace(); // ".space" 240 bool ParseDirectiveZero(); // ".zero" 241 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv" 242 bool ParseDirectiveOrg(); // ".org" 243 // ".align{,32}", ".p2align{,w,l}" 244 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize); 245 246 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which 247 /// accepts a single symbol (which should be a label or an external). 248 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr); 249 250 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 251 252 bool ParseDirectiveAbort(); // ".abort" 253 bool ParseDirectiveInclude(); // ".include" 254 bool ParseDirectiveIncbin(); // ".incbin" 255 256 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if" 257 // ".ifb" or ".ifnb", depending on ExpectBlank. 258 bool ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank); 259 // ".ifc" or ".ifnc", depending on ExpectEqual. 260 bool ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual); 261 // ".ifdef" or ".ifndef", depending on expect_defined 262 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 263 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 264 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else" 265 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 266 267 /// ParseEscapedString - Parse the current token as a string which may include 268 /// escaped characters and return the string contents. 269 bool ParseEscapedString(std::string &Data); 270 271 const MCExpr *ApplyModifierToExpr(const MCExpr *E, 272 MCSymbolRefExpr::VariantKind Variant); 273 274 // Macro-like directives 275 Macro *ParseMacroLikeBody(SMLoc DirectiveLoc); 276 void InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc, 277 raw_svector_ostream &OS); 278 bool ParseDirectiveRept(SMLoc DirectiveLoc); // ".rept" 279 bool ParseDirectiveIrp(SMLoc DirectiveLoc); // ".irp" 280 bool ParseDirectiveEndr(SMLoc DirectiveLoc); // ".endr" 281}; 282 283/// \brief Generic implementations of directive handling, etc. which is shared 284/// (or the default, at least) for all assembler parser. 285class GenericAsmParser : public MCAsmParserExtension { 286 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)> 287 void AddDirectiveHandler(StringRef Directive) { 288 getParser().AddDirectiveHandler(this, Directive, 289 HandleDirective<GenericAsmParser, Handler>); 290 } 291public: 292 GenericAsmParser() {} 293 294 AsmParser &getParser() { 295 return (AsmParser&) this->MCAsmParserExtension::getParser(); 296 } 297 298 virtual void Initialize(MCAsmParser &Parser) { 299 // Call the base implementation. 300 this->MCAsmParserExtension::Initialize(Parser); 301 302 // Debugging directives. 303 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file"); 304 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line"); 305 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc"); 306 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs"); 307 308 // CFI directives. 309 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>( 310 ".cfi_sections"); 311 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>( 312 ".cfi_startproc"); 313 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>( 314 ".cfi_endproc"); 315 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>( 316 ".cfi_def_cfa"); 317 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>( 318 ".cfi_def_cfa_offset"); 319 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>( 320 ".cfi_adjust_cfa_offset"); 321 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>( 322 ".cfi_def_cfa_register"); 323 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>( 324 ".cfi_offset"); 325 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>( 326 ".cfi_rel_offset"); 327 AddDirectiveHandler< 328 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality"); 329 AddDirectiveHandler< 330 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda"); 331 AddDirectiveHandler< 332 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state"); 333 AddDirectiveHandler< 334 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state"); 335 AddDirectiveHandler< 336 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value"); 337 AddDirectiveHandler< 338 &GenericAsmParser::ParseDirectiveCFIRestore>(".cfi_restore"); 339 AddDirectiveHandler< 340 &GenericAsmParser::ParseDirectiveCFIEscape>(".cfi_escape"); 341 AddDirectiveHandler< 342 &GenericAsmParser::ParseDirectiveCFISignalFrame>(".cfi_signal_frame"); 343 344 // Macro directives. 345 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 346 ".macros_on"); 347 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 348 ".macros_off"); 349 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro"); 350 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm"); 351 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro"); 352 AddDirectiveHandler<&GenericAsmParser::ParseDirectivePurgeMacro>(".purgem"); 353 354 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128"); 355 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128"); 356 } 357 358 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 359 360 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc); 361 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc); 362 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc); 363 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc); 364 bool ParseDirectiveCFISections(StringRef, SMLoc DirectiveLoc); 365 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc); 366 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc); 367 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc); 368 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc); 369 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc); 370 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc); 371 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc); 372 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc); 373 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc); 374 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc); 375 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc); 376 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc); 377 bool ParseDirectiveCFIRestore(StringRef, SMLoc DirectiveLoc); 378 bool ParseDirectiveCFIEscape(StringRef, SMLoc DirectiveLoc); 379 bool ParseDirectiveCFISignalFrame(StringRef, SMLoc DirectiveLoc); 380 381 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc); 382 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc); 383 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc); 384 bool ParseDirectivePurgeMacro(StringRef, SMLoc DirectiveLoc); 385 386 bool ParseDirectiveLEB128(StringRef, SMLoc); 387}; 388 389} 390 391namespace llvm { 392 393extern MCAsmParserExtension *createDarwinAsmParser(); 394extern MCAsmParserExtension *createELFAsmParser(); 395extern MCAsmParserExtension *createCOFFAsmParser(); 396 397} 398 399enum { DEFAULT_ADDRSPACE = 0 }; 400 401AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx, 402 MCStreamer &_Out, const MCAsmInfo &_MAI) 403 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM), 404 GenericParser(new GenericAsmParser), PlatformParser(0), 405 CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0), 406 AssemblerDialect(~0U) { 407 // Save the old handler. 408 SavedDiagHandler = SrcMgr.getDiagHandler(); 409 SavedDiagContext = SrcMgr.getDiagContext(); 410 // Set our own handler which calls the saved handler. 411 SrcMgr.setDiagHandler(DiagHandler, this); 412 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 413 414 // Initialize the generic parser. 415 GenericParser->Initialize(*this); 416 417 // Initialize the platform / file format parser. 418 // 419 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are 420 // created. 421 if (_MAI.hasMicrosoftFastStdCallMangling()) { 422 PlatformParser = createCOFFAsmParser(); 423 PlatformParser->Initialize(*this); 424 } else if (_MAI.hasSubsectionsViaSymbols()) { 425 PlatformParser = createDarwinAsmParser(); 426 PlatformParser->Initialize(*this); 427 } else { 428 PlatformParser = createELFAsmParser(); 429 PlatformParser->Initialize(*this); 430 } 431} 432 433AsmParser::~AsmParser() { 434 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!"); 435 436 // Destroy any macros. 437 for (StringMap<Macro*>::iterator it = MacroMap.begin(), 438 ie = MacroMap.end(); it != ie; ++it) 439 delete it->getValue(); 440 441 delete PlatformParser; 442 delete GenericParser; 443} 444 445void AsmParser::PrintMacroInstantiations() { 446 // Print the active macro instantiation stack. 447 for (std::vector<MacroInstantiation*>::const_reverse_iterator 448 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it) 449 PrintMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 450 "while in macro instantiation"); 451} 452 453bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 454 if (FatalAssemblerWarnings) 455 return Error(L, Msg, Ranges); 456 PrintMessage(L, SourceMgr::DK_Warning, Msg, Ranges); 457 PrintMacroInstantiations(); 458 return false; 459} 460 461bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 462 HadError = true; 463 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges); 464 PrintMacroInstantiations(); 465 return true; 466} 467 468bool AsmParser::EnterIncludeFile(const std::string &Filename) { 469 std::string IncludedFile; 470 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 471 if (NewBuf == -1) 472 return true; 473 474 CurBuffer = NewBuf; 475 476 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 477 478 return false; 479} 480 481/// Process the specified .incbin file by seaching for it in the include paths 482/// then just emitting the byte contents of the file to the streamer. This 483/// returns true on failure. 484bool AsmParser::ProcessIncbinFile(const std::string &Filename) { 485 std::string IncludedFile; 486 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 487 if (NewBuf == -1) 488 return true; 489 490 // Pick up the bytes from the file and emit them. 491 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(), 492 DEFAULT_ADDRSPACE); 493 return false; 494} 495 496void AsmParser::JumpToLoc(SMLoc Loc) { 497 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc); 498 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer()); 499} 500 501const AsmToken &AsmParser::Lex() { 502 const AsmToken *tok = &Lexer.Lex(); 503 504 if (tok->is(AsmToken::Eof)) { 505 // If this is the end of an included file, pop the parent file off the 506 // include stack. 507 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 508 if (ParentIncludeLoc != SMLoc()) { 509 JumpToLoc(ParentIncludeLoc); 510 tok = &Lexer.Lex(); 511 } 512 } 513 514 if (tok->is(AsmToken::Error)) 515 Error(Lexer.getErrLoc(), Lexer.getErr()); 516 517 return *tok; 518} 519 520bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 521 // Create the initial section, if requested. 522 if (!NoInitialTextSection) 523 Out.InitSections(); 524 525 // Prime the lexer. 526 Lex(); 527 528 HadError = false; 529 AsmCond StartingCondState = TheCondState; 530 531 // If we are generating dwarf for assembly source files save the initial text 532 // section and generate a .file directive. 533 if (getContext().getGenDwarfForAssembly()) { 534 getContext().setGenDwarfSection(getStreamer().getCurrentSection()); 535 MCSymbol *SectionStartSym = getContext().CreateTempSymbol(); 536 getStreamer().EmitLabel(SectionStartSym); 537 getContext().setGenDwarfSectionStartSym(SectionStartSym); 538 getStreamer().EmitDwarfFileDirective(getContext().nextGenDwarfFileNumber(), 539 StringRef(), SrcMgr.getMemoryBuffer(CurBuffer)->getBufferIdentifier()); 540 } 541 542 // While we have input, parse each statement. 543 while (Lexer.isNot(AsmToken::Eof)) { 544 if (!ParseStatement()) continue; 545 546 // We had an error, validate that one was emitted and recover by skipping to 547 // the next line. 548 assert(HadError && "Parse statement returned an error, but none emitted!"); 549 EatToEndOfStatement(); 550 } 551 552 if (TheCondState.TheCond != StartingCondState.TheCond || 553 TheCondState.Ignore != StartingCondState.Ignore) 554 return TokError("unmatched .ifs or .elses"); 555 556 // Check to see there are no empty DwarfFile slots. 557 const std::vector<MCDwarfFile *> &MCDwarfFiles = 558 getContext().getMCDwarfFiles(); 559 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) { 560 if (!MCDwarfFiles[i]) 561 TokError("unassigned file number: " + Twine(i) + " for .file directives"); 562 } 563 564 // Check to see that all assembler local symbols were actually defined. 565 // Targets that don't do subsections via symbols may not want this, though, 566 // so conservatively exclude them. Only do this if we're finalizing, though, 567 // as otherwise we won't necessarilly have seen everything yet. 568 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) { 569 const MCContext::SymbolTable &Symbols = getContext().getSymbols(); 570 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(), 571 e = Symbols.end(); 572 i != e; ++i) { 573 MCSymbol *Sym = i->getValue(); 574 // Variable symbols may not be marked as defined, so check those 575 // explicitly. If we know it's a variable, we have a definition for 576 // the purposes of this check. 577 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 578 // FIXME: We would really like to refer back to where the symbol was 579 // first referenced for a source location. We need to add something 580 // to track that. Currently, we just point to the end of the file. 581 PrintMessage(getLexer().getLoc(), SourceMgr::DK_Error, 582 "assembler local symbol '" + Sym->getName() + 583 "' not defined"); 584 } 585 } 586 587 588 // Finalize the output stream if there are no errors and if the client wants 589 // us to. 590 if (!HadError && !NoFinalize) 591 Out.Finish(); 592 593 return HadError; 594} 595 596void AsmParser::CheckForValidSection() { 597 if (!getStreamer().getCurrentSection()) { 598 TokError("expected section directive before assembly directive"); 599 Out.SwitchSection(Ctx.getMachOSection( 600 "__TEXT", "__text", 601 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS, 602 0, SectionKind::getText())); 603 } 604} 605 606/// EatToEndOfStatement - Throw away the rest of the line for testing purposes. 607void AsmParser::EatToEndOfStatement() { 608 while (Lexer.isNot(AsmToken::EndOfStatement) && 609 Lexer.isNot(AsmToken::Eof)) 610 Lex(); 611 612 // Eat EOL. 613 if (Lexer.is(AsmToken::EndOfStatement)) 614 Lex(); 615} 616 617StringRef AsmParser::ParseStringToEndOfStatement() { 618 const char *Start = getTok().getLoc().getPointer(); 619 620 while (Lexer.isNot(AsmToken::EndOfStatement) && 621 Lexer.isNot(AsmToken::Eof)) 622 Lex(); 623 624 const char *End = getTok().getLoc().getPointer(); 625 return StringRef(Start, End - Start); 626} 627 628StringRef AsmParser::ParseStringToComma() { 629 const char *Start = getTok().getLoc().getPointer(); 630 631 while (Lexer.isNot(AsmToken::EndOfStatement) && 632 Lexer.isNot(AsmToken::Comma) && 633 Lexer.isNot(AsmToken::Eof)) 634 Lex(); 635 636 const char *End = getTok().getLoc().getPointer(); 637 return StringRef(Start, End - Start); 638} 639 640/// ParseParenExpr - Parse a paren expression and return it. 641/// NOTE: This assumes the leading '(' has already been consumed. 642/// 643/// parenexpr ::= expr) 644/// 645bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 646 if (ParseExpression(Res)) return true; 647 if (Lexer.isNot(AsmToken::RParen)) 648 return TokError("expected ')' in parentheses expression"); 649 EndLoc = Lexer.getLoc(); 650 Lex(); 651 return false; 652} 653 654/// ParseBracketExpr - Parse a bracket expression and return it. 655/// NOTE: This assumes the leading '[' has already been consumed. 656/// 657/// bracketexpr ::= expr] 658/// 659bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 660 if (ParseExpression(Res)) return true; 661 if (Lexer.isNot(AsmToken::RBrac)) 662 return TokError("expected ']' in brackets expression"); 663 EndLoc = Lexer.getLoc(); 664 Lex(); 665 return false; 666} 667 668/// ParsePrimaryExpr - Parse a primary expression and return it. 669/// primaryexpr ::= (parenexpr 670/// primaryexpr ::= symbol 671/// primaryexpr ::= number 672/// primaryexpr ::= '.' 673/// primaryexpr ::= ~,+,- primaryexpr 674bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 675 switch (Lexer.getKind()) { 676 default: 677 return TokError("unknown token in expression"); 678 // If we have an error assume that we've already handled it. 679 case AsmToken::Error: 680 return true; 681 case AsmToken::Exclaim: 682 Lex(); // Eat the operator. 683 if (ParsePrimaryExpr(Res, EndLoc)) 684 return true; 685 Res = MCUnaryExpr::CreateLNot(Res, getContext()); 686 return false; 687 case AsmToken::Dollar: 688 case AsmToken::String: 689 case AsmToken::Identifier: { 690 EndLoc = Lexer.getLoc(); 691 692 StringRef Identifier; 693 if (ParseIdentifier(Identifier)) 694 return true; 695 696 // This is a symbol reference. 697 std::pair<StringRef, StringRef> Split = Identifier.split('@'); 698 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first); 699 700 // Lookup the symbol variant if used. 701 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 702 if (Split.first.size() != Identifier.size()) { 703 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 704 if (Variant == MCSymbolRefExpr::VK_Invalid) { 705 Variant = MCSymbolRefExpr::VK_None; 706 return TokError("invalid variant '" + Split.second + "'"); 707 } 708 } 709 710 // If this is an absolute variable reference, substitute it now to preserve 711 // semantics in the face of reassignment. 712 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) { 713 if (Variant) 714 return Error(EndLoc, "unexpected modifier on variable reference"); 715 716 Res = Sym->getVariableValue(); 717 return false; 718 } 719 720 // Otherwise create a symbol ref. 721 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext()); 722 return false; 723 } 724 case AsmToken::Integer: { 725 SMLoc Loc = getTok().getLoc(); 726 int64_t IntVal = getTok().getIntVal(); 727 Res = MCConstantExpr::Create(IntVal, getContext()); 728 EndLoc = Lexer.getLoc(); 729 Lex(); // Eat token. 730 // Look for 'b' or 'f' following an Integer as a directional label 731 if (Lexer.getKind() == AsmToken::Identifier) { 732 StringRef IDVal = getTok().getString(); 733 if (IDVal == "f" || IDVal == "b"){ 734 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal, 735 IDVal == "f" ? 1 : 0); 736 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, 737 getContext()); 738 if (IDVal == "b" && Sym->isUndefined()) 739 return Error(Loc, "invalid reference to undefined symbol"); 740 EndLoc = Lexer.getLoc(); 741 Lex(); // Eat identifier. 742 } 743 } 744 return false; 745 } 746 case AsmToken::Real: { 747 APFloat RealVal(APFloat::IEEEdouble, getTok().getString()); 748 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 749 Res = MCConstantExpr::Create(IntVal, getContext()); 750 Lex(); // Eat token. 751 return false; 752 } 753 case AsmToken::Dot: { 754 // This is a '.' reference, which references the current PC. Emit a 755 // temporary label to the streamer and refer to it. 756 MCSymbol *Sym = Ctx.CreateTempSymbol(); 757 Out.EmitLabel(Sym); 758 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext()); 759 EndLoc = Lexer.getLoc(); 760 Lex(); // Eat identifier. 761 return false; 762 } 763 case AsmToken::LParen: 764 Lex(); // Eat the '('. 765 return ParseParenExpr(Res, EndLoc); 766 case AsmToken::LBrac: 767 if (!PlatformParser->HasBracketExpressions()) 768 return TokError("brackets expression not supported on this target"); 769 Lex(); // Eat the '['. 770 return ParseBracketExpr(Res, EndLoc); 771 case AsmToken::Minus: 772 Lex(); // Eat the operator. 773 if (ParsePrimaryExpr(Res, EndLoc)) 774 return true; 775 Res = MCUnaryExpr::CreateMinus(Res, getContext()); 776 return false; 777 case AsmToken::Plus: 778 Lex(); // Eat the operator. 779 if (ParsePrimaryExpr(Res, EndLoc)) 780 return true; 781 Res = MCUnaryExpr::CreatePlus(Res, getContext()); 782 return false; 783 case AsmToken::Tilde: 784 Lex(); // Eat the operator. 785 if (ParsePrimaryExpr(Res, EndLoc)) 786 return true; 787 Res = MCUnaryExpr::CreateNot(Res, getContext()); 788 return false; 789 } 790} 791 792bool AsmParser::ParseExpression(const MCExpr *&Res) { 793 SMLoc EndLoc; 794 return ParseExpression(Res, EndLoc); 795} 796 797const MCExpr * 798AsmParser::ApplyModifierToExpr(const MCExpr *E, 799 MCSymbolRefExpr::VariantKind Variant) { 800 // Recurse over the given expression, rebuilding it to apply the given variant 801 // if there is exactly one symbol. 802 switch (E->getKind()) { 803 case MCExpr::Target: 804 case MCExpr::Constant: 805 return 0; 806 807 case MCExpr::SymbolRef: { 808 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 809 810 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 811 TokError("invalid variant on expression '" + 812 getTok().getIdentifier() + "' (already modified)"); 813 return E; 814 } 815 816 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext()); 817 } 818 819 case MCExpr::Unary: { 820 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 821 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant); 822 if (!Sub) 823 return 0; 824 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext()); 825 } 826 827 case MCExpr::Binary: { 828 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 829 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant); 830 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant); 831 832 if (!LHS && !RHS) 833 return 0; 834 835 if (!LHS) LHS = BE->getLHS(); 836 if (!RHS) RHS = BE->getRHS(); 837 838 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext()); 839 } 840 } 841 842 llvm_unreachable("Invalid expression kind!"); 843} 844 845/// ParseExpression - Parse an expression and return it. 846/// 847/// expr ::= expr &&,|| expr -> lowest. 848/// expr ::= expr |,^,&,! expr 849/// expr ::= expr ==,!=,<>,<,<=,>,>= expr 850/// expr ::= expr <<,>> expr 851/// expr ::= expr +,- expr 852/// expr ::= expr *,/,% expr -> highest. 853/// expr ::= primaryexpr 854/// 855bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 856 // Parse the expression. 857 Res = 0; 858 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc)) 859 return true; 860 861 // As a special case, we support 'a op b @ modifier' by rewriting the 862 // expression to include the modifier. This is inefficient, but in general we 863 // expect users to use 'a@modifier op b'. 864 if (Lexer.getKind() == AsmToken::At) { 865 Lex(); 866 867 if (Lexer.isNot(AsmToken::Identifier)) 868 return TokError("unexpected symbol modifier following '@'"); 869 870 MCSymbolRefExpr::VariantKind Variant = 871 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 872 if (Variant == MCSymbolRefExpr::VK_Invalid) 873 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 874 875 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant); 876 if (!ModifiedRes) { 877 return TokError("invalid modifier '" + getTok().getIdentifier() + 878 "' (no symbols present)"); 879 } 880 881 Res = ModifiedRes; 882 Lex(); 883 } 884 885 // Try to constant fold it up front, if possible. 886 int64_t Value; 887 if (Res->EvaluateAsAbsolute(Value)) 888 Res = MCConstantExpr::Create(Value, getContext()); 889 890 return false; 891} 892 893bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 894 Res = 0; 895 return ParseParenExpr(Res, EndLoc) || 896 ParseBinOpRHS(1, Res, EndLoc); 897} 898 899bool AsmParser::ParseAbsoluteExpression(int64_t &Res) { 900 const MCExpr *Expr; 901 902 SMLoc StartLoc = Lexer.getLoc(); 903 if (ParseExpression(Expr)) 904 return true; 905 906 if (!Expr->EvaluateAsAbsolute(Res)) 907 return Error(StartLoc, "expected absolute expression"); 908 909 return false; 910} 911 912static unsigned getBinOpPrecedence(AsmToken::TokenKind K, 913 MCBinaryExpr::Opcode &Kind) { 914 switch (K) { 915 default: 916 return 0; // not a binop. 917 918 // Lowest Precedence: &&, || 919 case AsmToken::AmpAmp: 920 Kind = MCBinaryExpr::LAnd; 921 return 1; 922 case AsmToken::PipePipe: 923 Kind = MCBinaryExpr::LOr; 924 return 1; 925 926 927 // Low Precedence: |, &, ^ 928 // 929 // FIXME: gas seems to support '!' as an infix operator? 930 case AsmToken::Pipe: 931 Kind = MCBinaryExpr::Or; 932 return 2; 933 case AsmToken::Caret: 934 Kind = MCBinaryExpr::Xor; 935 return 2; 936 case AsmToken::Amp: 937 Kind = MCBinaryExpr::And; 938 return 2; 939 940 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 941 case AsmToken::EqualEqual: 942 Kind = MCBinaryExpr::EQ; 943 return 3; 944 case AsmToken::ExclaimEqual: 945 case AsmToken::LessGreater: 946 Kind = MCBinaryExpr::NE; 947 return 3; 948 case AsmToken::Less: 949 Kind = MCBinaryExpr::LT; 950 return 3; 951 case AsmToken::LessEqual: 952 Kind = MCBinaryExpr::LTE; 953 return 3; 954 case AsmToken::Greater: 955 Kind = MCBinaryExpr::GT; 956 return 3; 957 case AsmToken::GreaterEqual: 958 Kind = MCBinaryExpr::GTE; 959 return 3; 960 961 // Intermediate Precedence: <<, >> 962 case AsmToken::LessLess: 963 Kind = MCBinaryExpr::Shl; 964 return 4; 965 case AsmToken::GreaterGreater: 966 Kind = MCBinaryExpr::Shr; 967 return 4; 968 969 // High Intermediate Precedence: +, - 970 case AsmToken::Plus: 971 Kind = MCBinaryExpr::Add; 972 return 5; 973 case AsmToken::Minus: 974 Kind = MCBinaryExpr::Sub; 975 return 5; 976 977 // Highest Precedence: *, /, % 978 case AsmToken::Star: 979 Kind = MCBinaryExpr::Mul; 980 return 6; 981 case AsmToken::Slash: 982 Kind = MCBinaryExpr::Div; 983 return 6; 984 case AsmToken::Percent: 985 Kind = MCBinaryExpr::Mod; 986 return 6; 987 } 988} 989 990 991/// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'. 992/// Res contains the LHS of the expression on input. 993bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 994 SMLoc &EndLoc) { 995 while (1) { 996 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 997 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 998 999 // If the next token is lower precedence than we are allowed to eat, return 1000 // successfully with what we ate already. 1001 if (TokPrec < Precedence) 1002 return false; 1003 1004 Lex(); 1005 1006 // Eat the next primary expression. 1007 const MCExpr *RHS; 1008 if (ParsePrimaryExpr(RHS, EndLoc)) return true; 1009 1010 // If BinOp binds less tightly with RHS than the operator after RHS, let 1011 // the pending operator take RHS as its LHS. 1012 MCBinaryExpr::Opcode Dummy; 1013 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 1014 if (TokPrec < NextTokPrec) { 1015 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true; 1016 } 1017 1018 // Merge LHS and RHS according to operator. 1019 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext()); 1020 } 1021} 1022 1023 1024 1025 1026/// ParseStatement: 1027/// ::= EndOfStatement 1028/// ::= Label* Directive ...Operands... EndOfStatement 1029/// ::= Label* Identifier OperandList* EndOfStatement 1030bool AsmParser::ParseStatement() { 1031 if (Lexer.is(AsmToken::EndOfStatement)) { 1032 Out.AddBlankLine(); 1033 Lex(); 1034 return false; 1035 } 1036 1037 // Statements always start with an identifier or are a full line comment. 1038 AsmToken ID = getTok(); 1039 SMLoc IDLoc = ID.getLoc(); 1040 StringRef IDVal; 1041 int64_t LocalLabelVal = -1; 1042 // A full line comment is a '#' as the first token. 1043 if (Lexer.is(AsmToken::Hash)) 1044 return ParseCppHashLineFilenameComment(IDLoc); 1045 1046 // Allow an integer followed by a ':' as a directional local label. 1047 if (Lexer.is(AsmToken::Integer)) { 1048 LocalLabelVal = getTok().getIntVal(); 1049 if (LocalLabelVal < 0) { 1050 if (!TheCondState.Ignore) 1051 return TokError("unexpected token at start of statement"); 1052 IDVal = ""; 1053 } 1054 else { 1055 IDVal = getTok().getString(); 1056 Lex(); // Consume the integer token to be used as an identifier token. 1057 if (Lexer.getKind() != AsmToken::Colon) { 1058 if (!TheCondState.Ignore) 1059 return TokError("unexpected token at start of statement"); 1060 } 1061 } 1062 1063 } else if (Lexer.is(AsmToken::Dot)) { 1064 // Treat '.' as a valid identifier in this context. 1065 Lex(); 1066 IDVal = "."; 1067 1068 } else if (ParseIdentifier(IDVal)) { 1069 if (!TheCondState.Ignore) 1070 return TokError("unexpected token at start of statement"); 1071 IDVal = ""; 1072 } 1073 1074 1075 // Handle conditional assembly here before checking for skipping. We 1076 // have to do this so that .endif isn't skipped in a ".if 0" block for 1077 // example. 1078 if (IDVal == ".if") 1079 return ParseDirectiveIf(IDLoc); 1080 if (IDVal == ".ifb") 1081 return ParseDirectiveIfb(IDLoc, true); 1082 if (IDVal == ".ifnb") 1083 return ParseDirectiveIfb(IDLoc, false); 1084 if (IDVal == ".ifc") 1085 return ParseDirectiveIfc(IDLoc, true); 1086 if (IDVal == ".ifnc") 1087 return ParseDirectiveIfc(IDLoc, false); 1088 if (IDVal == ".ifdef") 1089 return ParseDirectiveIfdef(IDLoc, true); 1090 if (IDVal == ".ifndef" || IDVal == ".ifnotdef") 1091 return ParseDirectiveIfdef(IDLoc, false); 1092 if (IDVal == ".elseif") 1093 return ParseDirectiveElseIf(IDLoc); 1094 if (IDVal == ".else") 1095 return ParseDirectiveElse(IDLoc); 1096 if (IDVal == ".endif") 1097 return ParseDirectiveEndIf(IDLoc); 1098 1099 // If we are in a ".if 0" block, ignore this statement. 1100 if (TheCondState.Ignore) { 1101 EatToEndOfStatement(); 1102 return false; 1103 } 1104 1105 // FIXME: Recurse on local labels? 1106 1107 // See what kind of statement we have. 1108 switch (Lexer.getKind()) { 1109 case AsmToken::Colon: { 1110 CheckForValidSection(); 1111 1112 // identifier ':' -> Label. 1113 Lex(); 1114 1115 // Diagnose attempt to use '.' as a label. 1116 if (IDVal == ".") 1117 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1118 1119 // Diagnose attempt to use a variable as a label. 1120 // 1121 // FIXME: Diagnostics. Note the location of the definition as a label. 1122 // FIXME: This doesn't diagnose assignment to a symbol which has been 1123 // implicitly marked as external. 1124 MCSymbol *Sym; 1125 if (LocalLabelVal == -1) 1126 Sym = getContext().GetOrCreateSymbol(IDVal); 1127 else 1128 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal); 1129 if (!Sym->isUndefined() || Sym->isVariable()) 1130 return Error(IDLoc, "invalid symbol redefinition"); 1131 1132 // Emit the label. 1133 Out.EmitLabel(Sym); 1134 1135 // If we are generating dwarf for assembly source files then gather the 1136 // info to make a dwarf label entry for this label if needed. 1137 if (getContext().getGenDwarfForAssembly()) 1138 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(), 1139 IDLoc); 1140 1141 // Consume any end of statement token, if present, to avoid spurious 1142 // AddBlankLine calls(). 1143 if (Lexer.is(AsmToken::EndOfStatement)) { 1144 Lex(); 1145 if (Lexer.is(AsmToken::Eof)) 1146 return false; 1147 } 1148 1149 return ParseStatement(); 1150 } 1151 1152 case AsmToken::Equal: 1153 // identifier '=' ... -> assignment statement 1154 Lex(); 1155 1156 return ParseAssignment(IDVal, true); 1157 1158 default: // Normal instruction or directive. 1159 break; 1160 } 1161 1162 // If macros are enabled, check to see if this is a macro instantiation. 1163 if (MacrosEnabled) 1164 if (const Macro *M = MacroMap.lookup(IDVal)) 1165 return HandleMacroEntry(IDVal, IDLoc, M); 1166 1167 // Otherwise, we have a normal instruction or directive. 1168 if (IDVal[0] == '.' && IDVal != ".") { 1169 // Assembler features 1170 if (IDVal == ".set" || IDVal == ".equ") 1171 return ParseDirectiveSet(IDVal, true); 1172 if (IDVal == ".equiv") 1173 return ParseDirectiveSet(IDVal, false); 1174 1175 // Data directives 1176 1177 if (IDVal == ".ascii") 1178 return ParseDirectiveAscii(IDVal, false); 1179 if (IDVal == ".asciz" || IDVal == ".string") 1180 return ParseDirectiveAscii(IDVal, true); 1181 1182 if (IDVal == ".byte") 1183 return ParseDirectiveValue(1); 1184 if (IDVal == ".short") 1185 return ParseDirectiveValue(2); 1186 if (IDVal == ".value") 1187 return ParseDirectiveValue(2); 1188 if (IDVal == ".2byte") 1189 return ParseDirectiveValue(2); 1190 if (IDVal == ".long") 1191 return ParseDirectiveValue(4); 1192 if (IDVal == ".int") 1193 return ParseDirectiveValue(4); 1194 if (IDVal == ".4byte") 1195 return ParseDirectiveValue(4); 1196 if (IDVal == ".quad") 1197 return ParseDirectiveValue(8); 1198 if (IDVal == ".8byte") 1199 return ParseDirectiveValue(8); 1200 if (IDVal == ".single" || IDVal == ".float") 1201 return ParseDirectiveRealValue(APFloat::IEEEsingle); 1202 if (IDVal == ".double") 1203 return ParseDirectiveRealValue(APFloat::IEEEdouble); 1204 1205 if (IDVal == ".align") { 1206 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1207 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1208 } 1209 if (IDVal == ".align32") { 1210 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1211 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1212 } 1213 if (IDVal == ".balign") 1214 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1215 if (IDVal == ".balignw") 1216 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1217 if (IDVal == ".balignl") 1218 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1219 if (IDVal == ".p2align") 1220 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1221 if (IDVal == ".p2alignw") 1222 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1223 if (IDVal == ".p2alignl") 1224 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1225 1226 if (IDVal == ".org") 1227 return ParseDirectiveOrg(); 1228 1229 if (IDVal == ".fill") 1230 return ParseDirectiveFill(); 1231 if (IDVal == ".space" || IDVal == ".skip") 1232 return ParseDirectiveSpace(); 1233 if (IDVal == ".zero") 1234 return ParseDirectiveZero(); 1235 1236 // Symbol attribute directives 1237 1238 if (IDVal == ".extern") { 1239 EatToEndOfStatement(); // .extern is the default, ignore it. 1240 return false; 1241 } 1242 if (IDVal == ".globl" || IDVal == ".global") 1243 return ParseDirectiveSymbolAttribute(MCSA_Global); 1244 if (IDVal == ".indirect_symbol") 1245 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol); 1246 if (IDVal == ".lazy_reference") 1247 return ParseDirectiveSymbolAttribute(MCSA_LazyReference); 1248 if (IDVal == ".no_dead_strip") 1249 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 1250 if (IDVal == ".symbol_resolver") 1251 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver); 1252 if (IDVal == ".private_extern") 1253 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern); 1254 if (IDVal == ".reference") 1255 return ParseDirectiveSymbolAttribute(MCSA_Reference); 1256 if (IDVal == ".weak_definition") 1257 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition); 1258 if (IDVal == ".weak_reference") 1259 return ParseDirectiveSymbolAttribute(MCSA_WeakReference); 1260 if (IDVal == ".weak_def_can_be_hidden") 1261 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 1262 1263 if (IDVal == ".comm" || IDVal == ".common") 1264 return ParseDirectiveComm(/*IsLocal=*/false); 1265 if (IDVal == ".lcomm") 1266 return ParseDirectiveComm(/*IsLocal=*/true); 1267 1268 if (IDVal == ".abort") 1269 return ParseDirectiveAbort(); 1270 if (IDVal == ".include") 1271 return ParseDirectiveInclude(); 1272 if (IDVal == ".incbin") 1273 return ParseDirectiveIncbin(); 1274 1275 if (IDVal == ".code16" || IDVal == ".code16gcc") 1276 return TokError(Twine(IDVal) + " not supported yet"); 1277 1278 // Macro-like directives 1279 if (IDVal == ".rept") 1280 return ParseDirectiveRept(IDLoc); 1281 if (IDVal == ".irp") 1282 return ParseDirectiveIrp(IDLoc); 1283 if (IDVal == ".endr") 1284 return ParseDirectiveEndr(IDLoc); 1285 1286 // Look up the handler in the handler table. 1287 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler = 1288 DirectiveMap.lookup(IDVal); 1289 if (Handler.first) 1290 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1291 1292 // Target hook for parsing target specific directives. 1293 if (!getTargetParser().ParseDirective(ID)) 1294 return false; 1295 1296 return Error(IDLoc, "unknown directive"); 1297 } 1298 1299 CheckForValidSection(); 1300 1301 // Canonicalize the opcode to lower case. 1302 SmallString<128> Opcode; 1303 for (unsigned i = 0, e = IDVal.size(); i != e; ++i) 1304 Opcode.push_back(tolower(IDVal[i])); 1305 1306 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands; 1307 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc, 1308 ParsedOperands); 1309 1310 // Dump the parsed representation, if requested. 1311 if (getShowParsedOperands()) { 1312 SmallString<256> Str; 1313 raw_svector_ostream OS(Str); 1314 OS << "parsed instruction: ["; 1315 for (unsigned i = 0; i != ParsedOperands.size(); ++i) { 1316 if (i != 0) 1317 OS << ", "; 1318 ParsedOperands[i]->print(OS); 1319 } 1320 OS << "]"; 1321 1322 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 1323 } 1324 1325 // If we are generating dwarf for assembly source files and the current 1326 // section is the initial text section then generate a .loc directive for 1327 // the instruction. 1328 if (!HadError && getContext().getGenDwarfForAssembly() && 1329 getContext().getGenDwarfSection() == getStreamer().getCurrentSection() ) { 1330 getStreamer().EmitDwarfLocDirective(getContext().getGenDwarfFileNumber(), 1331 SrcMgr.FindLineNumber(IDLoc, CurBuffer), 1332 0, DWARF2_LINE_DEFAULT_IS_STMT ? 1333 DWARF2_FLAG_IS_STMT : 0, 0, 0, 1334 StringRef()); 1335 } 1336 1337 // If parsing succeeded, match the instruction. 1338 if (!HadError) 1339 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands, 1340 Out); 1341 1342 // Free any parsed operands. 1343 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i) 1344 delete ParsedOperands[i]; 1345 1346 // Don't skip the rest of the line, the instruction parser is responsible for 1347 // that. 1348 return false; 1349} 1350 1351/// EatToEndOfLine uses the Lexer to eat the characters to the end of the line 1352/// since they may not be able to be tokenized to get to the end of line token. 1353void AsmParser::EatToEndOfLine() { 1354 if (!Lexer.is(AsmToken::EndOfStatement)) 1355 Lexer.LexUntilEndOfLine(); 1356 // Eat EOL. 1357 Lex(); 1358} 1359 1360/// ParseCppHashLineFilenameComment as this: 1361/// ::= # number "filename" 1362/// or just as a full line comment if it doesn't have a number and a string. 1363bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) { 1364 Lex(); // Eat the hash token. 1365 1366 if (getLexer().isNot(AsmToken::Integer)) { 1367 // Consume the line since in cases it is not a well-formed line directive, 1368 // as if were simply a full line comment. 1369 EatToEndOfLine(); 1370 return false; 1371 } 1372 1373 int64_t LineNumber = getTok().getIntVal(); 1374 Lex(); 1375 1376 if (getLexer().isNot(AsmToken::String)) { 1377 EatToEndOfLine(); 1378 return false; 1379 } 1380 1381 StringRef Filename = getTok().getString(); 1382 // Get rid of the enclosing quotes. 1383 Filename = Filename.substr(1, Filename.size()-2); 1384 1385 // Save the SMLoc, Filename and LineNumber for later use by diagnostics. 1386 CppHashLoc = L; 1387 CppHashFilename = Filename; 1388 CppHashLineNumber = LineNumber; 1389 1390 // Ignore any trailing characters, they're just comment. 1391 EatToEndOfLine(); 1392 return false; 1393} 1394 1395/// DiagHandler - will use the the last parsed cpp hash line filename comment 1396/// for the Filename and LineNo if any in the diagnostic. 1397void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 1398 const AsmParser *Parser = static_cast<const AsmParser*>(Context); 1399 raw_ostream &OS = errs(); 1400 1401 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 1402 const SMLoc &DiagLoc = Diag.getLoc(); 1403 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1404 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc); 1405 1406 // Like SourceMgr::PrintMessage() we need to print the include stack if any 1407 // before printing the message. 1408 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1409 if (!Parser->SavedDiagHandler && DiagCurBuffer > 0) { 1410 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 1411 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 1412 } 1413 1414 // If we have not parsed a cpp hash line filename comment or the source 1415 // manager changed or buffer changed (like in a nested include) then just 1416 // print the normal diagnostic using its Filename and LineNo. 1417 if (!Parser->CppHashLineNumber || 1418 &DiagSrcMgr != &Parser->SrcMgr || 1419 DiagBuf != CppHashBuf) { 1420 if (Parser->SavedDiagHandler) 1421 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); 1422 else 1423 Diag.print(0, OS); 1424 return; 1425 } 1426 1427 // Use the CppHashFilename and calculate a line number based on the 1428 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for 1429 // the diagnostic. 1430 const std::string Filename = Parser->CppHashFilename; 1431 1432 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 1433 int CppHashLocLineNo = 1434 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf); 1435 int LineNo = Parser->CppHashLineNumber - 1 + 1436 (DiagLocLineNo - CppHashLocLineNo); 1437 1438 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), 1439 Filename, LineNo, Diag.getColumnNo(), 1440 Diag.getKind(), Diag.getMessage(), 1441 Diag.getLineContents(), Diag.getRanges()); 1442 1443 if (Parser->SavedDiagHandler) 1444 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); 1445 else 1446 NewDiag.print(0, OS); 1447} 1448 1449bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, 1450 const std::vector<StringRef> &Parameters, 1451 const std::vector<MacroArgument> &A, 1452 const SMLoc &L) { 1453 unsigned NParameters = Parameters.size(); 1454 if (NParameters != 0 && NParameters != A.size()) 1455 return Error(L, "Wrong number of arguments"); 1456 1457 while (!Body.empty()) { 1458 // Scan for the next substitution. 1459 std::size_t End = Body.size(), Pos = 0; 1460 for (; Pos != End; ++Pos) { 1461 // Check for a substitution or escape. 1462 if (!NParameters) { 1463 // This macro has no parameters, look for $0, $1, etc. 1464 if (Body[Pos] != '$' || Pos + 1 == End) 1465 continue; 1466 1467 char Next = Body[Pos + 1]; 1468 if (Next == '$' || Next == 'n' || isdigit(Next)) 1469 break; 1470 } else { 1471 // This macro has parameters, look for \foo, \bar, etc. 1472 if (Body[Pos] == '\\' && Pos + 1 != End) 1473 break; 1474 } 1475 } 1476 1477 // Add the prefix. 1478 OS << Body.slice(0, Pos); 1479 1480 // Check if we reached the end. 1481 if (Pos == End) 1482 break; 1483 1484 if (!NParameters) { 1485 switch (Body[Pos+1]) { 1486 // $$ => $ 1487 case '$': 1488 OS << '$'; 1489 break; 1490 1491 // $n => number of arguments 1492 case 'n': 1493 OS << A.size(); 1494 break; 1495 1496 // $[0-9] => argument 1497 default: { 1498 // Missing arguments are ignored. 1499 unsigned Index = Body[Pos+1] - '0'; 1500 if (Index >= A.size()) 1501 break; 1502 1503 // Otherwise substitute with the token values, with spaces eliminated. 1504 for (MacroArgument::const_iterator it = A[Index].begin(), 1505 ie = A[Index].end(); it != ie; ++it) 1506 OS << it->getString(); 1507 break; 1508 } 1509 } 1510 Pos += 2; 1511 } else { 1512 unsigned I = Pos + 1; 1513 while (isalnum(Body[I]) && I + 1 != End) 1514 ++I; 1515 1516 const char *Begin = Body.data() + Pos +1; 1517 StringRef Argument(Begin, I - (Pos +1)); 1518 unsigned Index = 0; 1519 for (; Index < NParameters; ++Index) 1520 if (Parameters[Index] == Argument) 1521 break; 1522 1523 // FIXME: We should error at the macro definition. 1524 if (Index == NParameters) 1525 return Error(L, "Parameter not found"); 1526 1527 for (MacroArgument::const_iterator it = A[Index].begin(), 1528 ie = A[Index].end(); it != ie; ++it) 1529 OS << it->getString(); 1530 1531 Pos += 1 + Argument.size(); 1532 } 1533 // Update the scan point. 1534 Body = Body.substr(Pos); 1535 } 1536 1537 return false; 1538} 1539 1540MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 1541 MemoryBuffer *I) 1542 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL) 1543{ 1544} 1545 1546/// ParseMacroArgument - Extract AsmTokens for a macro argument. 1547/// This is used for both default macro parameter values and the 1548/// arguments in macro invocations 1549bool AsmParser::ParseMacroArgument(MacroArgument &MA) { 1550 unsigned ParenLevel = 0; 1551 1552 for (;;) { 1553 SMLoc LastTokenLoc; 1554 1555 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal)) 1556 return TokError("unexpected token in macro instantiation"); 1557 1558 // HandleMacroEntry relies on not advancing the lexer here 1559 // to be able to fill in the remaining default parameter values 1560 if (Lexer.is(AsmToken::EndOfStatement)) 1561 break; 1562 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) 1563 break; 1564 1565 // Adjust the current parentheses level. 1566 if (Lexer.is(AsmToken::LParen)) 1567 ++ParenLevel; 1568 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 1569 --ParenLevel; 1570 1571 // Append the token to the current argument list. 1572 MA.push_back(getTok()); 1573 Lex(); 1574 } 1575 if (ParenLevel != 0) 1576 return TokError("unbalanced parenthesises in macro argument"); 1577 return false; 1578} 1579 1580// Parse the macro instantiation arguments. 1581bool AsmParser::ParseMacroArguments(const Macro *M, 1582 std::vector<MacroArgument> &A) { 1583 const unsigned NParameters = M ? M->Parameters.size() : 0; 1584 1585 // Parse two kinds of macro invocations: 1586 // - macros defined without any parameters accept an arbitrary number of them 1587 // - macros defined with parameters accept at most that many of them 1588 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; 1589 ++Parameter) { 1590 MacroArgument MA; 1591 1592 if (ParseMacroArgument(MA)) 1593 return true; 1594 1595 if (!MA.empty()) 1596 A.push_back(MA); 1597 if (Lexer.is(AsmToken::EndOfStatement)) 1598 return false; 1599 1600 if (Lexer.is(AsmToken::Comma)) 1601 Lex(); 1602 } 1603 return TokError("Too many arguments"); 1604} 1605 1606bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc, 1607 const Macro *M) { 1608 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate 1609 // this, although we should protect against infinite loops. 1610 if (ActiveMacros.size() == 20) 1611 return TokError("macros cannot be nested more than 20 levels deep"); 1612 1613 std::vector<MacroArgument> MacroArguments; 1614 if (ParseMacroArguments(M, MacroArguments)) 1615 return true; 1616 1617 // Macro instantiation is lexical, unfortunately. We construct a new buffer 1618 // to hold the macro body with substitutions. 1619 SmallString<256> Buf; 1620 StringRef Body = M->Body; 1621 raw_svector_ostream OS(Buf); 1622 1623 if (expandMacro(OS, Body, M->Parameters, MacroArguments, getTok().getLoc())) 1624 return true; 1625 1626 // We include the .endmacro in the buffer as our queue to exit the macro 1627 // instantiation. 1628 OS << ".endmacro\n"; 1629 1630 MemoryBuffer *Instantiation = 1631 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 1632 1633 // Create the macro instantiation object and add to the current macro 1634 // instantiation stack. 1635 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc, 1636 getTok().getLoc(), 1637 Instantiation); 1638 ActiveMacros.push_back(MI); 1639 1640 // Jump to the macro instantiation and prime the lexer. 1641 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc()); 1642 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 1643 Lex(); 1644 1645 return false; 1646} 1647 1648void AsmParser::HandleMacroExit() { 1649 // Jump to the EndOfStatement we should return to, and consume it. 1650 JumpToLoc(ActiveMacros.back()->ExitLoc); 1651 Lex(); 1652 1653 // Pop the instantiation entry. 1654 delete ActiveMacros.back(); 1655 ActiveMacros.pop_back(); 1656} 1657 1658static bool IsUsedIn(const MCSymbol *Sym, const MCExpr *Value) { 1659 switch (Value->getKind()) { 1660 case MCExpr::Binary: { 1661 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr*>(Value); 1662 return IsUsedIn(Sym, BE->getLHS()) || IsUsedIn(Sym, BE->getRHS()); 1663 break; 1664 } 1665 case MCExpr::Target: 1666 case MCExpr::Constant: 1667 return false; 1668 case MCExpr::SymbolRef: { 1669 const MCSymbol &S = static_cast<const MCSymbolRefExpr*>(Value)->getSymbol(); 1670 if (S.isVariable()) 1671 return IsUsedIn(Sym, S.getVariableValue()); 1672 return &S == Sym; 1673 } 1674 case MCExpr::Unary: 1675 return IsUsedIn(Sym, static_cast<const MCUnaryExpr*>(Value)->getSubExpr()); 1676 } 1677 1678 llvm_unreachable("Unknown expr kind!"); 1679} 1680 1681bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) { 1682 // FIXME: Use better location, we should use proper tokens. 1683 SMLoc EqualLoc = Lexer.getLoc(); 1684 1685 const MCExpr *Value; 1686 if (ParseExpression(Value)) 1687 return true; 1688 1689 // Note: we don't count b as used in "a = b". This is to allow 1690 // a = b 1691 // b = c 1692 1693 if (Lexer.isNot(AsmToken::EndOfStatement)) 1694 return TokError("unexpected token in assignment"); 1695 1696 // Error on assignment to '.'. 1697 if (Name == ".") { 1698 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported " 1699 "(use '.space' or '.org').)")); 1700 } 1701 1702 // Eat the end of statement marker. 1703 Lex(); 1704 1705 // Validate that the LHS is allowed to be a variable (either it has not been 1706 // used as a symbol, or it is an absolute symbol). 1707 MCSymbol *Sym = getContext().LookupSymbol(Name); 1708 if (Sym) { 1709 // Diagnose assignment to a label. 1710 // 1711 // FIXME: Diagnostics. Note the location of the definition as a label. 1712 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 1713 if (IsUsedIn(Sym, Value)) 1714 return Error(EqualLoc, "Recursive use of '" + Name + "'"); 1715 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable()) 1716 ; // Allow redefinitions of undefined symbols only used in directives. 1717 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef) 1718 ; // Allow redefinitions of variables that haven't yet been used. 1719 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 1720 return Error(EqualLoc, "redefinition of '" + Name + "'"); 1721 else if (!Sym->isVariable()) 1722 return Error(EqualLoc, "invalid assignment to '" + Name + "'"); 1723 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 1724 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" + 1725 Name + "'"); 1726 1727 // Don't count these checks as uses. 1728 Sym->setUsed(false); 1729 } else 1730 Sym = getContext().GetOrCreateSymbol(Name); 1731 1732 // FIXME: Handle '.'. 1733 1734 // Do the assignment. 1735 Out.EmitAssignment(Sym, Value); 1736 1737 return false; 1738} 1739 1740/// ParseIdentifier: 1741/// ::= identifier 1742/// ::= string 1743bool AsmParser::ParseIdentifier(StringRef &Res) { 1744 // The assembler has relaxed rules for accepting identifiers, in particular we 1745 // allow things like '.globl $foo', which would normally be separate 1746 // tokens. At this level, we have already lexed so we cannot (currently) 1747 // handle this as a context dependent token, instead we detect adjacent tokens 1748 // and return the combined identifier. 1749 if (Lexer.is(AsmToken::Dollar)) { 1750 SMLoc DollarLoc = getLexer().getLoc(); 1751 1752 // Consume the dollar sign, and check for a following identifier. 1753 Lex(); 1754 if (Lexer.isNot(AsmToken::Identifier)) 1755 return true; 1756 1757 // We have a '$' followed by an identifier, make sure they are adjacent. 1758 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer()) 1759 return true; 1760 1761 // Construct the joined identifier and consume the token. 1762 Res = StringRef(DollarLoc.getPointer(), 1763 getTok().getIdentifier().size() + 1); 1764 Lex(); 1765 return false; 1766 } 1767 1768 if (Lexer.isNot(AsmToken::Identifier) && 1769 Lexer.isNot(AsmToken::String)) 1770 return true; 1771 1772 Res = getTok().getIdentifier(); 1773 1774 Lex(); // Consume the identifier token. 1775 1776 return false; 1777} 1778 1779/// ParseDirectiveSet: 1780/// ::= .equ identifier ',' expression 1781/// ::= .equiv identifier ',' expression 1782/// ::= .set identifier ',' expression 1783bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) { 1784 StringRef Name; 1785 1786 if (ParseIdentifier(Name)) 1787 return TokError("expected identifier after '" + Twine(IDVal) + "'"); 1788 1789 if (getLexer().isNot(AsmToken::Comma)) 1790 return TokError("unexpected token in '" + Twine(IDVal) + "'"); 1791 Lex(); 1792 1793 return ParseAssignment(Name, allow_redef); 1794} 1795 1796bool AsmParser::ParseEscapedString(std::string &Data) { 1797 assert(getLexer().is(AsmToken::String) && "Unexpected current token!"); 1798 1799 Data = ""; 1800 StringRef Str = getTok().getStringContents(); 1801 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 1802 if (Str[i] != '\\') { 1803 Data += Str[i]; 1804 continue; 1805 } 1806 1807 // Recognize escaped characters. Note that this escape semantics currently 1808 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes. 1809 ++i; 1810 if (i == e) 1811 return TokError("unexpected backslash at end of string"); 1812 1813 // Recognize octal sequences. 1814 if ((unsigned) (Str[i] - '0') <= 7) { 1815 // Consume up to three octal characters. 1816 unsigned Value = Str[i] - '0'; 1817 1818 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1819 ++i; 1820 Value = Value * 8 + (Str[i] - '0'); 1821 1822 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1823 ++i; 1824 Value = Value * 8 + (Str[i] - '0'); 1825 } 1826 } 1827 1828 if (Value > 255) 1829 return TokError("invalid octal escape sequence (out of range)"); 1830 1831 Data += (unsigned char) Value; 1832 continue; 1833 } 1834 1835 // Otherwise recognize individual escapes. 1836 switch (Str[i]) { 1837 default: 1838 // Just reject invalid escape sequences for now. 1839 return TokError("invalid escape sequence (unrecognized character)"); 1840 1841 case 'b': Data += '\b'; break; 1842 case 'f': Data += '\f'; break; 1843 case 'n': Data += '\n'; break; 1844 case 'r': Data += '\r'; break; 1845 case 't': Data += '\t'; break; 1846 case '"': Data += '"'; break; 1847 case '\\': Data += '\\'; break; 1848 } 1849 } 1850 1851 return false; 1852} 1853 1854/// ParseDirectiveAscii: 1855/// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 1856bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 1857 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1858 CheckForValidSection(); 1859 1860 for (;;) { 1861 if (getLexer().isNot(AsmToken::String)) 1862 return TokError("expected string in '" + Twine(IDVal) + "' directive"); 1863 1864 std::string Data; 1865 if (ParseEscapedString(Data)) 1866 return true; 1867 1868 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE); 1869 if (ZeroTerminated) 1870 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE); 1871 1872 Lex(); 1873 1874 if (getLexer().is(AsmToken::EndOfStatement)) 1875 break; 1876 1877 if (getLexer().isNot(AsmToken::Comma)) 1878 return TokError("unexpected token in '" + Twine(IDVal) + "' directive"); 1879 Lex(); 1880 } 1881 } 1882 1883 Lex(); 1884 return false; 1885} 1886 1887/// ParseDirectiveValue 1888/// ::= (.byte | .short | ... ) [ expression (, expression)* ] 1889bool AsmParser::ParseDirectiveValue(unsigned Size) { 1890 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1891 CheckForValidSection(); 1892 1893 for (;;) { 1894 const MCExpr *Value; 1895 SMLoc ExprLoc = getLexer().getLoc(); 1896 if (ParseExpression(Value)) 1897 return true; 1898 1899 // Special case constant expressions to match code generator. 1900 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 1901 assert(Size <= 8 && "Invalid size"); 1902 uint64_t IntValue = MCE->getValue(); 1903 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 1904 return Error(ExprLoc, "literal value out of range for directive"); 1905 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE); 1906 } else 1907 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE); 1908 1909 if (getLexer().is(AsmToken::EndOfStatement)) 1910 break; 1911 1912 // FIXME: Improve diagnostic. 1913 if (getLexer().isNot(AsmToken::Comma)) 1914 return TokError("unexpected token in directive"); 1915 Lex(); 1916 } 1917 } 1918 1919 Lex(); 1920 return false; 1921} 1922 1923/// ParseDirectiveRealValue 1924/// ::= (.single | .double) [ expression (, expression)* ] 1925bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) { 1926 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1927 CheckForValidSection(); 1928 1929 for (;;) { 1930 // We don't truly support arithmetic on floating point expressions, so we 1931 // have to manually parse unary prefixes. 1932 bool IsNeg = false; 1933 if (getLexer().is(AsmToken::Minus)) { 1934 Lex(); 1935 IsNeg = true; 1936 } else if (getLexer().is(AsmToken::Plus)) 1937 Lex(); 1938 1939 if (getLexer().isNot(AsmToken::Integer) && 1940 getLexer().isNot(AsmToken::Real) && 1941 getLexer().isNot(AsmToken::Identifier)) 1942 return TokError("unexpected token in directive"); 1943 1944 // Convert to an APFloat. 1945 APFloat Value(Semantics); 1946 StringRef IDVal = getTok().getString(); 1947 if (getLexer().is(AsmToken::Identifier)) { 1948 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 1949 Value = APFloat::getInf(Semantics); 1950 else if (!IDVal.compare_lower("nan")) 1951 Value = APFloat::getNaN(Semantics, false, ~0); 1952 else 1953 return TokError("invalid floating point literal"); 1954 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) == 1955 APFloat::opInvalidOp) 1956 return TokError("invalid floating point literal"); 1957 if (IsNeg) 1958 Value.changeSign(); 1959 1960 // Consume the numeric token. 1961 Lex(); 1962 1963 // Emit the value as an integer. 1964 APInt AsInt = Value.bitcastToAPInt(); 1965 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 1966 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE); 1967 1968 if (getLexer().is(AsmToken::EndOfStatement)) 1969 break; 1970 1971 if (getLexer().isNot(AsmToken::Comma)) 1972 return TokError("unexpected token in directive"); 1973 Lex(); 1974 } 1975 } 1976 1977 Lex(); 1978 return false; 1979} 1980 1981/// ParseDirectiveSpace 1982/// ::= .space expression [ , expression ] 1983bool AsmParser::ParseDirectiveSpace() { 1984 CheckForValidSection(); 1985 1986 int64_t NumBytes; 1987 if (ParseAbsoluteExpression(NumBytes)) 1988 return true; 1989 1990 int64_t FillExpr = 0; 1991 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1992 if (getLexer().isNot(AsmToken::Comma)) 1993 return TokError("unexpected token in '.space' directive"); 1994 Lex(); 1995 1996 if (ParseAbsoluteExpression(FillExpr)) 1997 return true; 1998 1999 if (getLexer().isNot(AsmToken::EndOfStatement)) 2000 return TokError("unexpected token in '.space' directive"); 2001 } 2002 2003 Lex(); 2004 2005 if (NumBytes <= 0) 2006 return TokError("invalid number of bytes in '.space' directive"); 2007 2008 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 2009 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE); 2010 2011 return false; 2012} 2013 2014/// ParseDirectiveZero 2015/// ::= .zero expression 2016bool AsmParser::ParseDirectiveZero() { 2017 CheckForValidSection(); 2018 2019 int64_t NumBytes; 2020 if (ParseAbsoluteExpression(NumBytes)) 2021 return true; 2022 2023 int64_t Val = 0; 2024 if (getLexer().is(AsmToken::Comma)) { 2025 Lex(); 2026 if (ParseAbsoluteExpression(Val)) 2027 return true; 2028 } 2029 2030 if (getLexer().isNot(AsmToken::EndOfStatement)) 2031 return TokError("unexpected token in '.zero' directive"); 2032 2033 Lex(); 2034 2035 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE); 2036 2037 return false; 2038} 2039 2040/// ParseDirectiveFill 2041/// ::= .fill expression , expression , expression 2042bool AsmParser::ParseDirectiveFill() { 2043 CheckForValidSection(); 2044 2045 int64_t NumValues; 2046 if (ParseAbsoluteExpression(NumValues)) 2047 return true; 2048 2049 if (getLexer().isNot(AsmToken::Comma)) 2050 return TokError("unexpected token in '.fill' directive"); 2051 Lex(); 2052 2053 int64_t FillSize; 2054 if (ParseAbsoluteExpression(FillSize)) 2055 return true; 2056 2057 if (getLexer().isNot(AsmToken::Comma)) 2058 return TokError("unexpected token in '.fill' directive"); 2059 Lex(); 2060 2061 int64_t FillExpr; 2062 if (ParseAbsoluteExpression(FillExpr)) 2063 return true; 2064 2065 if (getLexer().isNot(AsmToken::EndOfStatement)) 2066 return TokError("unexpected token in '.fill' directive"); 2067 2068 Lex(); 2069 2070 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8) 2071 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8"); 2072 2073 for (uint64_t i = 0, e = NumValues; i != e; ++i) 2074 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE); 2075 2076 return false; 2077} 2078 2079/// ParseDirectiveOrg 2080/// ::= .org expression [ , expression ] 2081bool AsmParser::ParseDirectiveOrg() { 2082 CheckForValidSection(); 2083 2084 const MCExpr *Offset; 2085 SMLoc Loc = getTok().getLoc(); 2086 if (ParseExpression(Offset)) 2087 return true; 2088 2089 // Parse optional fill expression. 2090 int64_t FillExpr = 0; 2091 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2092 if (getLexer().isNot(AsmToken::Comma)) 2093 return TokError("unexpected token in '.org' directive"); 2094 Lex(); 2095 2096 if (ParseAbsoluteExpression(FillExpr)) 2097 return true; 2098 2099 if (getLexer().isNot(AsmToken::EndOfStatement)) 2100 return TokError("unexpected token in '.org' directive"); 2101 } 2102 2103 Lex(); 2104 2105 // Only limited forms of relocatable expressions are accepted here, it 2106 // has to be relative to the current section. The streamer will return 2107 // 'true' if the expression wasn't evaluatable. 2108 if (getStreamer().EmitValueToOffset(Offset, FillExpr)) 2109 return Error(Loc, "expected assembly-time absolute expression"); 2110 2111 return false; 2112} 2113 2114/// ParseDirectiveAlign 2115/// ::= {.align, ...} expression [ , expression [ , expression ]] 2116bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 2117 CheckForValidSection(); 2118 2119 SMLoc AlignmentLoc = getLexer().getLoc(); 2120 int64_t Alignment; 2121 if (ParseAbsoluteExpression(Alignment)) 2122 return true; 2123 2124 SMLoc MaxBytesLoc; 2125 bool HasFillExpr = false; 2126 int64_t FillExpr = 0; 2127 int64_t MaxBytesToFill = 0; 2128 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2129 if (getLexer().isNot(AsmToken::Comma)) 2130 return TokError("unexpected token in directive"); 2131 Lex(); 2132 2133 // The fill expression can be omitted while specifying a maximum number of 2134 // alignment bytes, e.g: 2135 // .align 3,,4 2136 if (getLexer().isNot(AsmToken::Comma)) { 2137 HasFillExpr = true; 2138 if (ParseAbsoluteExpression(FillExpr)) 2139 return true; 2140 } 2141 2142 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2143 if (getLexer().isNot(AsmToken::Comma)) 2144 return TokError("unexpected token in directive"); 2145 Lex(); 2146 2147 MaxBytesLoc = getLexer().getLoc(); 2148 if (ParseAbsoluteExpression(MaxBytesToFill)) 2149 return true; 2150 2151 if (getLexer().isNot(AsmToken::EndOfStatement)) 2152 return TokError("unexpected token in directive"); 2153 } 2154 } 2155 2156 Lex(); 2157 2158 if (!HasFillExpr) 2159 FillExpr = 0; 2160 2161 // Compute alignment in bytes. 2162 if (IsPow2) { 2163 // FIXME: Diagnose overflow. 2164 if (Alignment >= 32) { 2165 Error(AlignmentLoc, "invalid alignment value"); 2166 Alignment = 31; 2167 } 2168 2169 Alignment = 1ULL << Alignment; 2170 } 2171 2172 // Diagnose non-sensical max bytes to align. 2173 if (MaxBytesLoc.isValid()) { 2174 if (MaxBytesToFill < 1) { 2175 Error(MaxBytesLoc, "alignment directive can never be satisfied in this " 2176 "many bytes, ignoring maximum bytes expression"); 2177 MaxBytesToFill = 0; 2178 } 2179 2180 if (MaxBytesToFill >= Alignment) { 2181 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 2182 "has no effect"); 2183 MaxBytesToFill = 0; 2184 } 2185 } 2186 2187 // Check whether we should use optimal code alignment for this .align 2188 // directive. 2189 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign(); 2190 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 2191 ValueSize == 1 && UseCodeAlign) { 2192 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 2193 } else { 2194 // FIXME: Target specific behavior about how the "extra" bytes are filled. 2195 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 2196 MaxBytesToFill); 2197 } 2198 2199 return false; 2200} 2201 2202/// ParseDirectiveSymbolAttribute 2203/// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 2204bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 2205 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2206 for (;;) { 2207 StringRef Name; 2208 SMLoc Loc = getTok().getLoc(); 2209 2210 if (ParseIdentifier(Name)) 2211 return Error(Loc, "expected identifier in directive"); 2212 2213 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2214 2215 // Assembler local symbols don't make any sense here. Complain loudly. 2216 if (Sym->isTemporary()) 2217 return Error(Loc, "non-local symbol required in directive"); 2218 2219 getStreamer().EmitSymbolAttribute(Sym, Attr); 2220 2221 if (getLexer().is(AsmToken::EndOfStatement)) 2222 break; 2223 2224 if (getLexer().isNot(AsmToken::Comma)) 2225 return TokError("unexpected token in directive"); 2226 Lex(); 2227 } 2228 } 2229 2230 Lex(); 2231 return false; 2232} 2233 2234/// ParseDirectiveComm 2235/// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 2236bool AsmParser::ParseDirectiveComm(bool IsLocal) { 2237 CheckForValidSection(); 2238 2239 SMLoc IDLoc = getLexer().getLoc(); 2240 StringRef Name; 2241 if (ParseIdentifier(Name)) 2242 return TokError("expected identifier in directive"); 2243 2244 // Handle the identifier as the key symbol. 2245 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2246 2247 if (getLexer().isNot(AsmToken::Comma)) 2248 return TokError("unexpected token in directive"); 2249 Lex(); 2250 2251 int64_t Size; 2252 SMLoc SizeLoc = getLexer().getLoc(); 2253 if (ParseAbsoluteExpression(Size)) 2254 return true; 2255 2256 int64_t Pow2Alignment = 0; 2257 SMLoc Pow2AlignmentLoc; 2258 if (getLexer().is(AsmToken::Comma)) { 2259 Lex(); 2260 Pow2AlignmentLoc = getLexer().getLoc(); 2261 if (ParseAbsoluteExpression(Pow2Alignment)) 2262 return true; 2263 2264 // If this target takes alignments in bytes (not log) validate and convert. 2265 if (Lexer.getMAI().getAlignmentIsInBytes()) { 2266 if (!isPowerOf2_64(Pow2Alignment)) 2267 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 2268 Pow2Alignment = Log2_64(Pow2Alignment); 2269 } 2270 } 2271 2272 if (getLexer().isNot(AsmToken::EndOfStatement)) 2273 return TokError("unexpected token in '.comm' or '.lcomm' directive"); 2274 2275 Lex(); 2276 2277 // NOTE: a size of zero for a .comm should create a undefined symbol 2278 // but a size of .lcomm creates a bss symbol of size zero. 2279 if (Size < 0) 2280 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 2281 "be less than zero"); 2282 2283 // NOTE: The alignment in the directive is a power of 2 value, the assembler 2284 // may internally end up wanting an alignment in bytes. 2285 // FIXME: Diagnose overflow. 2286 if (Pow2Alignment < 0) 2287 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 2288 "alignment, can't be less than zero"); 2289 2290 if (!Sym->isUndefined()) 2291 return Error(IDLoc, "invalid symbol redefinition"); 2292 2293 // '.lcomm' is equivalent to '.zerofill'. 2294 // Create the Symbol as a common or local common with Size and Pow2Alignment 2295 if (IsLocal) { 2296 getStreamer().EmitZerofill(Ctx.getMachOSection( 2297 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL, 2298 0, SectionKind::getBSS()), 2299 Sym, Size, 1 << Pow2Alignment); 2300 return false; 2301 } 2302 2303 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 2304 return false; 2305} 2306 2307/// ParseDirectiveAbort 2308/// ::= .abort [... message ...] 2309bool AsmParser::ParseDirectiveAbort() { 2310 // FIXME: Use loc from directive. 2311 SMLoc Loc = getLexer().getLoc(); 2312 2313 StringRef Str = ParseStringToEndOfStatement(); 2314 if (getLexer().isNot(AsmToken::EndOfStatement)) 2315 return TokError("unexpected token in '.abort' directive"); 2316 2317 Lex(); 2318 2319 if (Str.empty()) 2320 Error(Loc, ".abort detected. Assembly stopping."); 2321 else 2322 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 2323 // FIXME: Actually abort assembly here. 2324 2325 return false; 2326} 2327 2328/// ParseDirectiveInclude 2329/// ::= .include "filename" 2330bool AsmParser::ParseDirectiveInclude() { 2331 if (getLexer().isNot(AsmToken::String)) 2332 return TokError("expected string in '.include' directive"); 2333 2334 std::string Filename = getTok().getString(); 2335 SMLoc IncludeLoc = getLexer().getLoc(); 2336 Lex(); 2337 2338 if (getLexer().isNot(AsmToken::EndOfStatement)) 2339 return TokError("unexpected token in '.include' directive"); 2340 2341 // Strip the quotes. 2342 Filename = Filename.substr(1, Filename.size()-2); 2343 2344 // Attempt to switch the lexer to the included file before consuming the end 2345 // of statement to avoid losing it when we switch. 2346 if (EnterIncludeFile(Filename)) { 2347 Error(IncludeLoc, "Could not find include file '" + Filename + "'"); 2348 return true; 2349 } 2350 2351 return false; 2352} 2353 2354/// ParseDirectiveIncbin 2355/// ::= .incbin "filename" 2356bool AsmParser::ParseDirectiveIncbin() { 2357 if (getLexer().isNot(AsmToken::String)) 2358 return TokError("expected string in '.incbin' directive"); 2359 2360 std::string Filename = getTok().getString(); 2361 SMLoc IncbinLoc = getLexer().getLoc(); 2362 Lex(); 2363 2364 if (getLexer().isNot(AsmToken::EndOfStatement)) 2365 return TokError("unexpected token in '.incbin' directive"); 2366 2367 // Strip the quotes. 2368 Filename = Filename.substr(1, Filename.size()-2); 2369 2370 // Attempt to process the included file. 2371 if (ProcessIncbinFile(Filename)) { 2372 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'"); 2373 return true; 2374 } 2375 2376 return false; 2377} 2378 2379/// ParseDirectiveIf 2380/// ::= .if expression 2381bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) { 2382 TheCondStack.push_back(TheCondState); 2383 TheCondState.TheCond = AsmCond::IfCond; 2384 if (TheCondState.Ignore) { 2385 EatToEndOfStatement(); 2386 } else { 2387 int64_t ExprValue; 2388 if (ParseAbsoluteExpression(ExprValue)) 2389 return true; 2390 2391 if (getLexer().isNot(AsmToken::EndOfStatement)) 2392 return TokError("unexpected token in '.if' directive"); 2393 2394 Lex(); 2395 2396 TheCondState.CondMet = ExprValue; 2397 TheCondState.Ignore = !TheCondState.CondMet; 2398 } 2399 2400 return false; 2401} 2402 2403/// ParseDirectiveIfb 2404/// ::= .ifb string 2405bool AsmParser::ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { 2406 TheCondStack.push_back(TheCondState); 2407 TheCondState.TheCond = AsmCond::IfCond; 2408 2409 if (TheCondState.Ignore) { 2410 EatToEndOfStatement(); 2411 } else { 2412 StringRef Str = ParseStringToEndOfStatement(); 2413 2414 if (getLexer().isNot(AsmToken::EndOfStatement)) 2415 return TokError("unexpected token in '.ifb' directive"); 2416 2417 Lex(); 2418 2419 TheCondState.CondMet = ExpectBlank == Str.empty(); 2420 TheCondState.Ignore = !TheCondState.CondMet; 2421 } 2422 2423 return false; 2424} 2425 2426/// ParseDirectiveIfc 2427/// ::= .ifc string1, string2 2428bool AsmParser::ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) { 2429 TheCondStack.push_back(TheCondState); 2430 TheCondState.TheCond = AsmCond::IfCond; 2431 2432 if (TheCondState.Ignore) { 2433 EatToEndOfStatement(); 2434 } else { 2435 StringRef Str1 = ParseStringToComma(); 2436 2437 if (getLexer().isNot(AsmToken::Comma)) 2438 return TokError("unexpected token in '.ifc' directive"); 2439 2440 Lex(); 2441 2442 StringRef Str2 = ParseStringToEndOfStatement(); 2443 2444 if (getLexer().isNot(AsmToken::EndOfStatement)) 2445 return TokError("unexpected token in '.ifc' directive"); 2446 2447 Lex(); 2448 2449 TheCondState.CondMet = ExpectEqual == (Str1 == Str2); 2450 TheCondState.Ignore = !TheCondState.CondMet; 2451 } 2452 2453 return false; 2454} 2455 2456/// ParseDirectiveIfdef 2457/// ::= .ifdef symbol 2458bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 2459 StringRef Name; 2460 TheCondStack.push_back(TheCondState); 2461 TheCondState.TheCond = AsmCond::IfCond; 2462 2463 if (TheCondState.Ignore) { 2464 EatToEndOfStatement(); 2465 } else { 2466 if (ParseIdentifier(Name)) 2467 return TokError("expected identifier after '.ifdef'"); 2468 2469 Lex(); 2470 2471 MCSymbol *Sym = getContext().LookupSymbol(Name); 2472 2473 if (expect_defined) 2474 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined()); 2475 else 2476 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined()); 2477 TheCondState.Ignore = !TheCondState.CondMet; 2478 } 2479 2480 return false; 2481} 2482 2483/// ParseDirectiveElseIf 2484/// ::= .elseif expression 2485bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) { 2486 if (TheCondState.TheCond != AsmCond::IfCond && 2487 TheCondState.TheCond != AsmCond::ElseIfCond) 2488 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or " 2489 " an .elseif"); 2490 TheCondState.TheCond = AsmCond::ElseIfCond; 2491 2492 bool LastIgnoreState = false; 2493 if (!TheCondStack.empty()) 2494 LastIgnoreState = TheCondStack.back().Ignore; 2495 if (LastIgnoreState || TheCondState.CondMet) { 2496 TheCondState.Ignore = true; 2497 EatToEndOfStatement(); 2498 } 2499 else { 2500 int64_t ExprValue; 2501 if (ParseAbsoluteExpression(ExprValue)) 2502 return true; 2503 2504 if (getLexer().isNot(AsmToken::EndOfStatement)) 2505 return TokError("unexpected token in '.elseif' directive"); 2506 2507 Lex(); 2508 TheCondState.CondMet = ExprValue; 2509 TheCondState.Ignore = !TheCondState.CondMet; 2510 } 2511 2512 return false; 2513} 2514 2515/// ParseDirectiveElse 2516/// ::= .else 2517bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) { 2518 if (getLexer().isNot(AsmToken::EndOfStatement)) 2519 return TokError("unexpected token in '.else' directive"); 2520 2521 Lex(); 2522 2523 if (TheCondState.TheCond != AsmCond::IfCond && 2524 TheCondState.TheCond != AsmCond::ElseIfCond) 2525 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an " 2526 ".elseif"); 2527 TheCondState.TheCond = AsmCond::ElseCond; 2528 bool LastIgnoreState = false; 2529 if (!TheCondStack.empty()) 2530 LastIgnoreState = TheCondStack.back().Ignore; 2531 if (LastIgnoreState || TheCondState.CondMet) 2532 TheCondState.Ignore = true; 2533 else 2534 TheCondState.Ignore = false; 2535 2536 return false; 2537} 2538 2539/// ParseDirectiveEndIf 2540/// ::= .endif 2541bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) { 2542 if (getLexer().isNot(AsmToken::EndOfStatement)) 2543 return TokError("unexpected token in '.endif' directive"); 2544 2545 Lex(); 2546 2547 if ((TheCondState.TheCond == AsmCond::NoCond) || 2548 TheCondStack.empty()) 2549 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or " 2550 ".else"); 2551 if (!TheCondStack.empty()) { 2552 TheCondState = TheCondStack.back(); 2553 TheCondStack.pop_back(); 2554 } 2555 2556 return false; 2557} 2558 2559/// ParseDirectiveFile 2560/// ::= .file [number] filename 2561/// ::= .file number directory filename 2562bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) { 2563 // FIXME: I'm not sure what this is. 2564 int64_t FileNumber = -1; 2565 SMLoc FileNumberLoc = getLexer().getLoc(); 2566 if (getLexer().is(AsmToken::Integer)) { 2567 FileNumber = getTok().getIntVal(); 2568 Lex(); 2569 2570 if (FileNumber < 1) 2571 return TokError("file number less than one"); 2572 } 2573 2574 if (getLexer().isNot(AsmToken::String)) 2575 return TokError("unexpected token in '.file' directive"); 2576 2577 // Usually the directory and filename together, otherwise just the directory. 2578 StringRef Path = getTok().getString(); 2579 Path = Path.substr(1, Path.size()-2); 2580 Lex(); 2581 2582 StringRef Directory; 2583 StringRef Filename; 2584 if (getLexer().is(AsmToken::String)) { 2585 if (FileNumber == -1) 2586 return TokError("explicit path specified, but no file number"); 2587 Filename = getTok().getString(); 2588 Filename = Filename.substr(1, Filename.size()-2); 2589 Directory = Path; 2590 Lex(); 2591 } else { 2592 Filename = Path; 2593 } 2594 2595 if (getLexer().isNot(AsmToken::EndOfStatement)) 2596 return TokError("unexpected token in '.file' directive"); 2597 2598 if (FileNumber == -1) 2599 getStreamer().EmitFileDirective(Filename); 2600 else { 2601 if (getContext().getGenDwarfForAssembly() == true) 2602 Error(DirectiveLoc, "input can't have .file dwarf directives when -g is " 2603 "used to generate dwarf debug info for assembly code"); 2604 2605 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename)) 2606 Error(FileNumberLoc, "file number already allocated"); 2607 } 2608 2609 return false; 2610} 2611 2612/// ParseDirectiveLine 2613/// ::= .line [number] 2614bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) { 2615 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2616 if (getLexer().isNot(AsmToken::Integer)) 2617 return TokError("unexpected token in '.line' directive"); 2618 2619 int64_t LineNumber = getTok().getIntVal(); 2620 (void) LineNumber; 2621 Lex(); 2622 2623 // FIXME: Do something with the .line. 2624 } 2625 2626 if (getLexer().isNot(AsmToken::EndOfStatement)) 2627 return TokError("unexpected token in '.line' directive"); 2628 2629 return false; 2630} 2631 2632 2633/// ParseDirectiveLoc 2634/// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 2635/// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 2636/// The first number is a file number, must have been previously assigned with 2637/// a .file directive, the second number is the line number and optionally the 2638/// third number is a column position (zero if not specified). The remaining 2639/// optional items are .loc sub-directives. 2640bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) { 2641 2642 if (getLexer().isNot(AsmToken::Integer)) 2643 return TokError("unexpected token in '.loc' directive"); 2644 int64_t FileNumber = getTok().getIntVal(); 2645 if (FileNumber < 1) 2646 return TokError("file number less than one in '.loc' directive"); 2647 if (!getContext().isValidDwarfFileNumber(FileNumber)) 2648 return TokError("unassigned file number in '.loc' directive"); 2649 Lex(); 2650 2651 int64_t LineNumber = 0; 2652 if (getLexer().is(AsmToken::Integer)) { 2653 LineNumber = getTok().getIntVal(); 2654 if (LineNumber < 1) 2655 return TokError("line number less than one in '.loc' directive"); 2656 Lex(); 2657 } 2658 2659 int64_t ColumnPos = 0; 2660 if (getLexer().is(AsmToken::Integer)) { 2661 ColumnPos = getTok().getIntVal(); 2662 if (ColumnPos < 0) 2663 return TokError("column position less than zero in '.loc' directive"); 2664 Lex(); 2665 } 2666 2667 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 2668 unsigned Isa = 0; 2669 int64_t Discriminator = 0; 2670 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2671 for (;;) { 2672 if (getLexer().is(AsmToken::EndOfStatement)) 2673 break; 2674 2675 StringRef Name; 2676 SMLoc Loc = getTok().getLoc(); 2677 if (getParser().ParseIdentifier(Name)) 2678 return TokError("unexpected token in '.loc' directive"); 2679 2680 if (Name == "basic_block") 2681 Flags |= DWARF2_FLAG_BASIC_BLOCK; 2682 else if (Name == "prologue_end") 2683 Flags |= DWARF2_FLAG_PROLOGUE_END; 2684 else if (Name == "epilogue_begin") 2685 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 2686 else if (Name == "is_stmt") { 2687 SMLoc Loc = getTok().getLoc(); 2688 const MCExpr *Value; 2689 if (getParser().ParseExpression(Value)) 2690 return true; 2691 // The expression must be the constant 0 or 1. 2692 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2693 int Value = MCE->getValue(); 2694 if (Value == 0) 2695 Flags &= ~DWARF2_FLAG_IS_STMT; 2696 else if (Value == 1) 2697 Flags |= DWARF2_FLAG_IS_STMT; 2698 else 2699 return Error(Loc, "is_stmt value not 0 or 1"); 2700 } 2701 else { 2702 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 2703 } 2704 } 2705 else if (Name == "isa") { 2706 SMLoc Loc = getTok().getLoc(); 2707 const MCExpr *Value; 2708 if (getParser().ParseExpression(Value)) 2709 return true; 2710 // The expression must be a constant greater or equal to 0. 2711 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2712 int Value = MCE->getValue(); 2713 if (Value < 0) 2714 return Error(Loc, "isa number less than zero"); 2715 Isa = Value; 2716 } 2717 else { 2718 return Error(Loc, "isa number not a constant value"); 2719 } 2720 } 2721 else if (Name == "discriminator") { 2722 if (getParser().ParseAbsoluteExpression(Discriminator)) 2723 return true; 2724 } 2725 else { 2726 return Error(Loc, "unknown sub-directive in '.loc' directive"); 2727 } 2728 2729 if (getLexer().is(AsmToken::EndOfStatement)) 2730 break; 2731 } 2732 } 2733 2734 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 2735 Isa, Discriminator, StringRef()); 2736 2737 return false; 2738} 2739 2740/// ParseDirectiveStabs 2741/// ::= .stabs string, number, number, number 2742bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive, 2743 SMLoc DirectiveLoc) { 2744 return TokError("unsupported directive '" + Directive + "'"); 2745} 2746 2747/// ParseDirectiveCFISections 2748/// ::= .cfi_sections section [, section] 2749bool GenericAsmParser::ParseDirectiveCFISections(StringRef, 2750 SMLoc DirectiveLoc) { 2751 StringRef Name; 2752 bool EH = false; 2753 bool Debug = false; 2754 2755 if (getParser().ParseIdentifier(Name)) 2756 return TokError("Expected an identifier"); 2757 2758 if (Name == ".eh_frame") 2759 EH = true; 2760 else if (Name == ".debug_frame") 2761 Debug = true; 2762 2763 if (getLexer().is(AsmToken::Comma)) { 2764 Lex(); 2765 2766 if (getParser().ParseIdentifier(Name)) 2767 return TokError("Expected an identifier"); 2768 2769 if (Name == ".eh_frame") 2770 EH = true; 2771 else if (Name == ".debug_frame") 2772 Debug = true; 2773 } 2774 2775 getStreamer().EmitCFISections(EH, Debug); 2776 2777 return false; 2778} 2779 2780/// ParseDirectiveCFIStartProc 2781/// ::= .cfi_startproc 2782bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef, 2783 SMLoc DirectiveLoc) { 2784 getStreamer().EmitCFIStartProc(); 2785 return false; 2786} 2787 2788/// ParseDirectiveCFIEndProc 2789/// ::= .cfi_endproc 2790bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) { 2791 getStreamer().EmitCFIEndProc(); 2792 return false; 2793} 2794 2795/// ParseRegisterOrRegisterNumber - parse register name or number. 2796bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register, 2797 SMLoc DirectiveLoc) { 2798 unsigned RegNo; 2799 2800 if (getLexer().isNot(AsmToken::Integer)) { 2801 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc, 2802 DirectiveLoc)) 2803 return true; 2804 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true); 2805 } else 2806 return getParser().ParseAbsoluteExpression(Register); 2807 2808 return false; 2809} 2810 2811/// ParseDirectiveCFIDefCfa 2812/// ::= .cfi_def_cfa register, offset 2813bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef, 2814 SMLoc DirectiveLoc) { 2815 int64_t Register = 0; 2816 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2817 return true; 2818 2819 if (getLexer().isNot(AsmToken::Comma)) 2820 return TokError("unexpected token in directive"); 2821 Lex(); 2822 2823 int64_t Offset = 0; 2824 if (getParser().ParseAbsoluteExpression(Offset)) 2825 return true; 2826 2827 getStreamer().EmitCFIDefCfa(Register, Offset); 2828 return false; 2829} 2830 2831/// ParseDirectiveCFIDefCfaOffset 2832/// ::= .cfi_def_cfa_offset offset 2833bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef, 2834 SMLoc DirectiveLoc) { 2835 int64_t Offset = 0; 2836 if (getParser().ParseAbsoluteExpression(Offset)) 2837 return true; 2838 2839 getStreamer().EmitCFIDefCfaOffset(Offset); 2840 return false; 2841} 2842 2843/// ParseDirectiveCFIAdjustCfaOffset 2844/// ::= .cfi_adjust_cfa_offset adjustment 2845bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef, 2846 SMLoc DirectiveLoc) { 2847 int64_t Adjustment = 0; 2848 if (getParser().ParseAbsoluteExpression(Adjustment)) 2849 return true; 2850 2851 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 2852 return false; 2853} 2854 2855/// ParseDirectiveCFIDefCfaRegister 2856/// ::= .cfi_def_cfa_register register 2857bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef, 2858 SMLoc DirectiveLoc) { 2859 int64_t Register = 0; 2860 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2861 return true; 2862 2863 getStreamer().EmitCFIDefCfaRegister(Register); 2864 return false; 2865} 2866 2867/// ParseDirectiveCFIOffset 2868/// ::= .cfi_offset register, offset 2869bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) { 2870 int64_t Register = 0; 2871 int64_t Offset = 0; 2872 2873 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2874 return true; 2875 2876 if (getLexer().isNot(AsmToken::Comma)) 2877 return TokError("unexpected token in directive"); 2878 Lex(); 2879 2880 if (getParser().ParseAbsoluteExpression(Offset)) 2881 return true; 2882 2883 getStreamer().EmitCFIOffset(Register, Offset); 2884 return false; 2885} 2886 2887/// ParseDirectiveCFIRelOffset 2888/// ::= .cfi_rel_offset register, offset 2889bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef, 2890 SMLoc DirectiveLoc) { 2891 int64_t Register = 0; 2892 2893 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2894 return true; 2895 2896 if (getLexer().isNot(AsmToken::Comma)) 2897 return TokError("unexpected token in directive"); 2898 Lex(); 2899 2900 int64_t Offset = 0; 2901 if (getParser().ParseAbsoluteExpression(Offset)) 2902 return true; 2903 2904 getStreamer().EmitCFIRelOffset(Register, Offset); 2905 return false; 2906} 2907 2908static bool isValidEncoding(int64_t Encoding) { 2909 if (Encoding & ~0xff) 2910 return false; 2911 2912 if (Encoding == dwarf::DW_EH_PE_omit) 2913 return true; 2914 2915 const unsigned Format = Encoding & 0xf; 2916 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 2917 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 2918 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 2919 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 2920 return false; 2921 2922 const unsigned Application = Encoding & 0x70; 2923 if (Application != dwarf::DW_EH_PE_absptr && 2924 Application != dwarf::DW_EH_PE_pcrel) 2925 return false; 2926 2927 return true; 2928} 2929 2930/// ParseDirectiveCFIPersonalityOrLsda 2931/// ::= .cfi_personality encoding, [symbol_name] 2932/// ::= .cfi_lsda encoding, [symbol_name] 2933bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal, 2934 SMLoc DirectiveLoc) { 2935 int64_t Encoding = 0; 2936 if (getParser().ParseAbsoluteExpression(Encoding)) 2937 return true; 2938 if (Encoding == dwarf::DW_EH_PE_omit) 2939 return false; 2940 2941 if (!isValidEncoding(Encoding)) 2942 return TokError("unsupported encoding."); 2943 2944 if (getLexer().isNot(AsmToken::Comma)) 2945 return TokError("unexpected token in directive"); 2946 Lex(); 2947 2948 StringRef Name; 2949 if (getParser().ParseIdentifier(Name)) 2950 return TokError("expected identifier in directive"); 2951 2952 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2953 2954 if (IDVal == ".cfi_personality") 2955 getStreamer().EmitCFIPersonality(Sym, Encoding); 2956 else { 2957 assert(IDVal == ".cfi_lsda"); 2958 getStreamer().EmitCFILsda(Sym, Encoding); 2959 } 2960 return false; 2961} 2962 2963/// ParseDirectiveCFIRememberState 2964/// ::= .cfi_remember_state 2965bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal, 2966 SMLoc DirectiveLoc) { 2967 getStreamer().EmitCFIRememberState(); 2968 return false; 2969} 2970 2971/// ParseDirectiveCFIRestoreState 2972/// ::= .cfi_remember_state 2973bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal, 2974 SMLoc DirectiveLoc) { 2975 getStreamer().EmitCFIRestoreState(); 2976 return false; 2977} 2978 2979/// ParseDirectiveCFISameValue 2980/// ::= .cfi_same_value register 2981bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal, 2982 SMLoc DirectiveLoc) { 2983 int64_t Register = 0; 2984 2985 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2986 return true; 2987 2988 getStreamer().EmitCFISameValue(Register); 2989 2990 return false; 2991} 2992 2993/// ParseDirectiveCFIRestore 2994/// ::= .cfi_restore register 2995bool GenericAsmParser::ParseDirectiveCFIRestore(StringRef IDVal, 2996 SMLoc DirectiveLoc) { 2997 int64_t Register = 0; 2998 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2999 return true; 3000 3001 getStreamer().EmitCFIRestore(Register); 3002 3003 return false; 3004} 3005 3006/// ParseDirectiveCFIEscape 3007/// ::= .cfi_escape expression[,...] 3008bool GenericAsmParser::ParseDirectiveCFIEscape(StringRef IDVal, 3009 SMLoc DirectiveLoc) { 3010 std::string Values; 3011 int64_t CurrValue; 3012 if (getParser().ParseAbsoluteExpression(CurrValue)) 3013 return true; 3014 3015 Values.push_back((uint8_t)CurrValue); 3016 3017 while (getLexer().is(AsmToken::Comma)) { 3018 Lex(); 3019 3020 if (getParser().ParseAbsoluteExpression(CurrValue)) 3021 return true; 3022 3023 Values.push_back((uint8_t)CurrValue); 3024 } 3025 3026 getStreamer().EmitCFIEscape(Values); 3027 return false; 3028} 3029 3030/// ParseDirectiveCFISignalFrame 3031/// ::= .cfi_signal_frame 3032bool GenericAsmParser::ParseDirectiveCFISignalFrame(StringRef Directive, 3033 SMLoc DirectiveLoc) { 3034 if (getLexer().isNot(AsmToken::EndOfStatement)) 3035 return Error(getLexer().getLoc(), 3036 "unexpected token in '" + Directive + "' directive"); 3037 3038 getStreamer().EmitCFISignalFrame(); 3039 3040 return false; 3041} 3042 3043/// ParseDirectiveMacrosOnOff 3044/// ::= .macros_on 3045/// ::= .macros_off 3046bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive, 3047 SMLoc DirectiveLoc) { 3048 if (getLexer().isNot(AsmToken::EndOfStatement)) 3049 return Error(getLexer().getLoc(), 3050 "unexpected token in '" + Directive + "' directive"); 3051 3052 getParser().MacrosEnabled = Directive == ".macros_on"; 3053 3054 return false; 3055} 3056 3057/// ParseDirectiveMacro 3058/// ::= .macro name [parameters] 3059bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive, 3060 SMLoc DirectiveLoc) { 3061 StringRef Name; 3062 if (getParser().ParseIdentifier(Name)) 3063 return TokError("expected identifier in directive"); 3064 3065 std::vector<StringRef> Parameters; 3066 if (getLexer().isNot(AsmToken::EndOfStatement)) { 3067 for(;;) { 3068 StringRef Parameter; 3069 if (getParser().ParseIdentifier(Parameter)) 3070 return TokError("expected identifier in directive"); 3071 Parameters.push_back(Parameter); 3072 3073 if (getLexer().isNot(AsmToken::Comma)) 3074 break; 3075 Lex(); 3076 } 3077 } 3078 3079 if (getLexer().isNot(AsmToken::EndOfStatement)) 3080 return TokError("unexpected token in '.macro' directive"); 3081 3082 // Eat the end of statement. 3083 Lex(); 3084 3085 AsmToken EndToken, StartToken = getTok(); 3086 3087 // Lex the macro definition. 3088 for (;;) { 3089 // Check whether we have reached the end of the file. 3090 if (getLexer().is(AsmToken::Eof)) 3091 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 3092 3093 // Otherwise, check whether we have reach the .endmacro. 3094 if (getLexer().is(AsmToken::Identifier) && 3095 (getTok().getIdentifier() == ".endm" || 3096 getTok().getIdentifier() == ".endmacro")) { 3097 EndToken = getTok(); 3098 Lex(); 3099 if (getLexer().isNot(AsmToken::EndOfStatement)) 3100 return TokError("unexpected token in '" + EndToken.getIdentifier() + 3101 "' directive"); 3102 break; 3103 } 3104 3105 // Otherwise, scan til the end of the statement. 3106 getParser().EatToEndOfStatement(); 3107 } 3108 3109 if (getParser().MacroMap.lookup(Name)) { 3110 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 3111 } 3112 3113 const char *BodyStart = StartToken.getLoc().getPointer(); 3114 const char *BodyEnd = EndToken.getLoc().getPointer(); 3115 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 3116 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters); 3117 return false; 3118} 3119 3120/// ParseDirectiveEndMacro 3121/// ::= .endm 3122/// ::= .endmacro 3123bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive, 3124 SMLoc DirectiveLoc) { 3125 if (getLexer().isNot(AsmToken::EndOfStatement)) 3126 return TokError("unexpected token in '" + Directive + "' directive"); 3127 3128 // If we are inside a macro instantiation, terminate the current 3129 // instantiation. 3130 if (!getParser().ActiveMacros.empty()) { 3131 getParser().HandleMacroExit(); 3132 return false; 3133 } 3134 3135 // Otherwise, this .endmacro is a stray entry in the file; well formed 3136 // .endmacro directives are handled during the macro definition parsing. 3137 return TokError("unexpected '" + Directive + "' in file, " 3138 "no current macro definition"); 3139} 3140 3141/// ParseDirectivePurgeMacro 3142/// ::= .purgem 3143bool GenericAsmParser::ParseDirectivePurgeMacro(StringRef Directive, 3144 SMLoc DirectiveLoc) { 3145 StringRef Name; 3146 if (getParser().ParseIdentifier(Name)) 3147 return TokError("expected identifier in '.purgem' directive"); 3148 3149 if (getLexer().isNot(AsmToken::EndOfStatement)) 3150 return TokError("unexpected token in '.purgem' directive"); 3151 3152 StringMap<Macro*>::iterator I = getParser().MacroMap.find(Name); 3153 if (I == getParser().MacroMap.end()) 3154 return Error(DirectiveLoc, "macro '" + Name + "' is not defined"); 3155 3156 // Undefine the macro. 3157 delete I->getValue(); 3158 getParser().MacroMap.erase(I); 3159 return false; 3160} 3161 3162bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) { 3163 getParser().CheckForValidSection(); 3164 3165 const MCExpr *Value; 3166 3167 if (getParser().ParseExpression(Value)) 3168 return true; 3169 3170 if (getLexer().isNot(AsmToken::EndOfStatement)) 3171 return TokError("unexpected token in directive"); 3172 3173 if (DirName[1] == 's') 3174 getStreamer().EmitSLEB128Value(Value); 3175 else 3176 getStreamer().EmitULEB128Value(Value); 3177 3178 return false; 3179} 3180 3181Macro *AsmParser::ParseMacroLikeBody(SMLoc DirectiveLoc) { 3182 AsmToken EndToken, StartToken = getTok(); 3183 3184 unsigned NestLevel = 0; 3185 for (;;) { 3186 // Check whether we have reached the end of the file. 3187 if (getLexer().is(AsmToken::Eof)) { 3188 Error(DirectiveLoc, "no matching '.endr' in definition"); 3189 return 0; 3190 } 3191 3192 if (Lexer.is(AsmToken::Identifier) && 3193 (getTok().getIdentifier() == ".rept")) { 3194 ++NestLevel; 3195 } 3196 3197 // Otherwise, check whether we have reached the .endr. 3198 if (Lexer.is(AsmToken::Identifier) && 3199 getTok().getIdentifier() == ".endr") { 3200 if (NestLevel == 0) { 3201 EndToken = getTok(); 3202 Lex(); 3203 if (Lexer.isNot(AsmToken::EndOfStatement)) { 3204 TokError("unexpected token in '.endr' directive"); 3205 return 0; 3206 } 3207 break; 3208 } 3209 --NestLevel; 3210 } 3211 3212 // Otherwise, scan till the end of the statement. 3213 EatToEndOfStatement(); 3214 } 3215 3216 const char *BodyStart = StartToken.getLoc().getPointer(); 3217 const char *BodyEnd = EndToken.getLoc().getPointer(); 3218 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 3219 3220 // We Are Anonymous. 3221 StringRef Name; 3222 std::vector<StringRef> Parameters; 3223 return new Macro(Name, Body, Parameters); 3224} 3225 3226void AsmParser::InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc, 3227 raw_svector_ostream &OS) { 3228 OS << ".endr\n"; 3229 3230 MemoryBuffer *Instantiation = 3231 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>"); 3232 3233 // Create the macro instantiation object and add to the current macro 3234 // instantiation stack. 3235 MacroInstantiation *MI = new MacroInstantiation(M, DirectiveLoc, 3236 getTok().getLoc(), 3237 Instantiation); 3238 ActiveMacros.push_back(MI); 3239 3240 // Jump to the macro instantiation and prime the lexer. 3241 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc()); 3242 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 3243 Lex(); 3244} 3245 3246bool AsmParser::ParseDirectiveRept(SMLoc DirectiveLoc) { 3247 int64_t Count; 3248 if (ParseAbsoluteExpression(Count)) 3249 return TokError("unexpected token in '.rept' directive"); 3250 3251 if (Count < 0) 3252 return TokError("Count is negative"); 3253 3254 if (Lexer.isNot(AsmToken::EndOfStatement)) 3255 return TokError("unexpected token in '.rept' directive"); 3256 3257 // Eat the end of statement. 3258 Lex(); 3259 3260 // Lex the rept definition. 3261 Macro *M = ParseMacroLikeBody(DirectiveLoc); 3262 if (!M) 3263 return true; 3264 3265 // Macro instantiation is lexical, unfortunately. We construct a new buffer 3266 // to hold the macro body with substitutions. 3267 SmallString<256> Buf; 3268 std::vector<StringRef> Parameters; 3269 const std::vector<MacroArgument> A; 3270 raw_svector_ostream OS(Buf); 3271 while (Count--) { 3272 if (expandMacro(OS, M->Body, Parameters, A, getTok().getLoc())) 3273 return true; 3274 } 3275 InstantiateMacroLikeBody(M, DirectiveLoc, OS); 3276 3277 return false; 3278} 3279 3280/// ParseDirectiveIrp 3281/// ::= .irp symbol,values 3282bool AsmParser::ParseDirectiveIrp(SMLoc DirectiveLoc) { 3283 std::vector<StringRef> Parameters; 3284 StringRef Parameter; 3285 3286 if (ParseIdentifier(Parameter)) 3287 return TokError("expected identifier in '.irp' directive"); 3288 3289 Parameters.push_back(Parameter); 3290 3291 if (Lexer.isNot(AsmToken::Comma)) 3292 return TokError("expected comma in '.irp' directive"); 3293 3294 Lex(); 3295 3296 std::vector<MacroArgument> A; 3297 if (ParseMacroArguments(0, A)) 3298 return true; 3299 3300 // Eat the end of statement. 3301 Lex(); 3302 3303 // Lex the irp definition. 3304 Macro *M = ParseMacroLikeBody(DirectiveLoc); 3305 if (!M) 3306 return true; 3307 3308 // Macro instantiation is lexical, unfortunately. We construct a new buffer 3309 // to hold the macro body with substitutions. 3310 SmallString<256> Buf; 3311 raw_svector_ostream OS(Buf); 3312 3313 for (std::vector<MacroArgument>::iterator i = A.begin(), e = A.end(); i != e; 3314 ++i) { 3315 std::vector<MacroArgument> Args; 3316 Args.push_back(*i); 3317 3318 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc())) 3319 return true; 3320 } 3321 3322 InstantiateMacroLikeBody(M, DirectiveLoc, OS); 3323 3324 return false; 3325} 3326 3327bool AsmParser::ParseDirectiveEndr(SMLoc DirectiveLoc) { 3328 if (ActiveMacros.empty()) 3329 return TokError("unexpected '.endr' directive, no current .rept"); 3330 3331 // The only .repl that should get here are the ones created by 3332 // InstantiateMacroLikeBody. 3333 assert(getLexer().is(AsmToken::EndOfStatement)); 3334 3335 HandleMacroExit(); 3336 return false; 3337} 3338 3339/// \brief Create an MCAsmParser instance. 3340MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, 3341 MCContext &C, MCStreamer &Out, 3342 const MCAsmInfo &MAI) { 3343 return new AsmParser(SM, C, Out, MAI); 3344} 3345