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