AsmParser.cpp revision 3f2d5f60b31fd057c10f77b2e607b23a8c94f6d3
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> 39using namespace llvm; 40 41static cl::opt<bool> 42FatalAssemblerWarnings("fatal-assembler-warnings", 43 cl::desc("Consider warnings as error")); 44 45namespace { 46 47/// \brief Helper class for tracking macro definitions. 48struct Macro { 49 StringRef Name; 50 StringRef Body; 51 std::vector<StringRef> Parameters; 52 53public: 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. 60struct 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 73public: 74 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 75 MemoryBuffer *I); 76}; 77 78/// \brief The concrete assembly parser instance. 79class AsmParser : public MCAsmParser { 80 friend class GenericAsmParser; 81 82 AsmParser(const AsmParser &); // DO NOT IMPLEMENT 83 void operator=(const AsmParser &); // DO NOT IMPLEMENT 84private: 85 AsmLexer Lexer; 86 MCContext &Ctx; 87 MCStreamer &Out; 88 const MCAsmInfo &MAI; 89 SourceMgr &SrcMgr; 90 MCAsmParserExtension *GenericParser; 91 MCAsmParserExtension *PlatformParser; 92 93 /// This is the current buffer index we're lexing from as managed by the 94 /// SourceMgr object. 95 int CurBuffer; 96 97 AsmCond TheCondState; 98 std::vector<AsmCond> TheCondStack; 99 100 /// DirectiveMap - This is a table handlers for directives. Each handler is 101 /// invoked after the directive identifier is read and is responsible for 102 /// parsing and validating the rest of the directive. The handler is passed 103 /// in the directive name and the location of the directive keyword. 104 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap; 105 106 /// MacroMap - Map of currently defined macros. 107 StringMap<Macro*> MacroMap; 108 109 /// ActiveMacros - Stack of active macro instantiations. 110 std::vector<MacroInstantiation*> ActiveMacros; 111 112 /// Boolean tracking whether macro substitution is enabled. 113 unsigned MacrosEnabled : 1; 114 115 /// Flag tracking whether any errors have been encountered. 116 unsigned HadError : 1; 117 118 /// The values from the last parsed cpp hash file line comment if any. 119 StringRef CppHashFilename; 120 int64_t CppHashLineNumber; 121 SMLoc CppHashLoc; 122 123public: 124 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out, 125 const MCAsmInfo &MAI); 126 ~AsmParser(); 127 128 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false); 129 130 void AddDirectiveHandler(MCAsmParserExtension *Object, 131 StringRef Directive, 132 DirectiveHandler Handler) { 133 DirectiveMap[Directive] = std::make_pair(Object, Handler); 134 } 135 136public: 137 /// @name MCAsmParser Interface 138 /// { 139 140 virtual SourceMgr &getSourceManager() { return SrcMgr; } 141 virtual MCAsmLexer &getLexer() { return Lexer; } 142 virtual MCContext &getContext() { return Ctx; } 143 virtual MCStreamer &getStreamer() { return Out; } 144 145 virtual bool Warning(SMLoc L, const Twine &Msg, 146 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 147 virtual bool Error(SMLoc L, const Twine &Msg, 148 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()); 149 150 const AsmToken &Lex(); 151 152 bool ParseExpression(const MCExpr *&Res); 153 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc); 154 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc); 155 virtual bool ParseAbsoluteExpression(int64_t &Res); 156 157 /// } 158 159private: 160 void CheckForValidSection(); 161 162 bool ParseStatement(); 163 void EatToEndOfLine(); 164 bool ParseCppHashLineFilenameComment(const SMLoc &L); 165 166 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M); 167 bool expandMacro(SmallString<256> &Buf, StringRef Body, 168 const std::vector<StringRef> &Parameters, 169 const std::vector<std::vector<AsmToken> > &A, 170 const SMLoc &L); 171 void HandleMacroExit(); 172 173 void PrintMacroInstantiations(); 174 void PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg, 175 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>(), 176 bool ShowLine = true) const { 177 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges, ShowLine); 178 } 179 static void DiagHandler(const SMDiagnostic &Diag, void *Context); 180 181 /// EnterIncludeFile - Enter the specified file. This returns true on failure. 182 bool EnterIncludeFile(const std::string &Filename); 183 184 /// \brief Reset the current lexer position to that given by \arg Loc. The 185 /// current token is not set; clients should ensure Lex() is called 186 /// subsequently. 187 void JumpToLoc(SMLoc Loc); 188 189 void EatToEndOfStatement(); 190 191 /// \brief Parse up to the end of statement and a return the contents from the 192 /// current token until the end of the statement; the current token on exit 193 /// will be either the EndOfStatement or EOF. 194 StringRef ParseStringToEndOfStatement(); 195 196 bool ParseAssignment(StringRef Name, bool allow_redef); 197 198 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc); 199 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc); 200 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc); 201 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc); 202 203 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier) 204 /// and set \arg Res to the identifier contents. 205 bool ParseIdentifier(StringRef &Res); 206 207 // Directive Parsing. 208 209 // ".ascii", ".asciiz", ".string" 210 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated); 211 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ... 212 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ... 213 bool ParseDirectiveFill(); // ".fill" 214 bool ParseDirectiveSpace(); // ".space" 215 bool ParseDirectiveZero(); // ".zero" 216 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv" 217 bool ParseDirectiveOrg(); // ".org" 218 // ".align{,32}", ".p2align{,w,l}" 219 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize); 220 221 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which 222 /// accepts a single symbol (which should be a label or an external). 223 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr); 224 225 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm" 226 227 bool ParseDirectiveAbort(); // ".abort" 228 bool ParseDirectiveInclude(); // ".include" 229 230 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if" 231 // ".ifdef" or ".ifndef", depending on expect_defined 232 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined); 233 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif" 234 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else" 235 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif 236 237 /// ParseEscapedString - Parse the current token as a string which may include 238 /// escaped characters and return the string contents. 239 bool ParseEscapedString(std::string &Data); 240 241 const MCExpr *ApplyModifierToExpr(const MCExpr *E, 242 MCSymbolRefExpr::VariantKind Variant); 243}; 244 245/// \brief Generic implementations of directive handling, etc. which is shared 246/// (or the default, at least) for all assembler parser. 247class GenericAsmParser : public MCAsmParserExtension { 248 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)> 249 void AddDirectiveHandler(StringRef Directive) { 250 getParser().AddDirectiveHandler(this, Directive, 251 HandleDirective<GenericAsmParser, Handler>); 252 } 253public: 254 GenericAsmParser() {} 255 256 AsmParser &getParser() { 257 return (AsmParser&) this->MCAsmParserExtension::getParser(); 258 } 259 260 virtual void Initialize(MCAsmParser &Parser) { 261 // Call the base implementation. 262 this->MCAsmParserExtension::Initialize(Parser); 263 264 // Debugging directives. 265 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file"); 266 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line"); 267 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc"); 268 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs"); 269 270 // CFI directives. 271 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>( 272 ".cfi_sections"); 273 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>( 274 ".cfi_startproc"); 275 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>( 276 ".cfi_endproc"); 277 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>( 278 ".cfi_def_cfa"); 279 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>( 280 ".cfi_def_cfa_offset"); 281 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>( 282 ".cfi_adjust_cfa_offset"); 283 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>( 284 ".cfi_def_cfa_register"); 285 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>( 286 ".cfi_offset"); 287 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>( 288 ".cfi_rel_offset"); 289 AddDirectiveHandler< 290 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality"); 291 AddDirectiveHandler< 292 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda"); 293 AddDirectiveHandler< 294 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state"); 295 AddDirectiveHandler< 296 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state"); 297 AddDirectiveHandler< 298 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value"); 299 300 // Macro directives. 301 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 302 ".macros_on"); 303 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>( 304 ".macros_off"); 305 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro"); 306 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm"); 307 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro"); 308 309 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128"); 310 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128"); 311 } 312 313 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc); 314 315 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc); 316 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc); 317 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc); 318 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc); 319 bool ParseDirectiveCFISections(StringRef, SMLoc DirectiveLoc); 320 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc); 321 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc); 322 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc); 323 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc); 324 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc); 325 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc); 326 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc); 327 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc); 328 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc); 329 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc); 330 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc); 331 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc); 332 333 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc); 334 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc); 335 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc); 336 337 bool ParseDirectiveLEB128(StringRef, SMLoc); 338}; 339 340} 341 342namespace llvm { 343 344extern MCAsmParserExtension *createDarwinAsmParser(); 345extern MCAsmParserExtension *createELFAsmParser(); 346extern MCAsmParserExtension *createCOFFAsmParser(); 347 348} 349 350enum { DEFAULT_ADDRSPACE = 0 }; 351 352AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx, 353 MCStreamer &_Out, const MCAsmInfo &_MAI) 354 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM), 355 GenericParser(new GenericAsmParser), PlatformParser(0), 356 CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0) { 357 SrcMgr.setDiagHandler(DiagHandler, this); 358 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 359 360 // Initialize the generic parser. 361 GenericParser->Initialize(*this); 362 363 // Initialize the platform / file format parser. 364 // 365 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are 366 // created. 367 if (_MAI.hasMicrosoftFastStdCallMangling()) { 368 PlatformParser = createCOFFAsmParser(); 369 PlatformParser->Initialize(*this); 370 } else if (_MAI.hasSubsectionsViaSymbols()) { 371 PlatformParser = createDarwinAsmParser(); 372 PlatformParser->Initialize(*this); 373 } else { 374 PlatformParser = createELFAsmParser(); 375 PlatformParser->Initialize(*this); 376 } 377} 378 379AsmParser::~AsmParser() { 380 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!"); 381 382 // Destroy any macros. 383 for (StringMap<Macro*>::iterator it = MacroMap.begin(), 384 ie = MacroMap.end(); it != ie; ++it) 385 delete it->getValue(); 386 387 delete PlatformParser; 388 delete GenericParser; 389} 390 391void AsmParser::PrintMacroInstantiations() { 392 // Print the active macro instantiation stack. 393 for (std::vector<MacroInstantiation*>::const_reverse_iterator 394 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it) 395 PrintMessage((*it)->InstantiationLoc, SourceMgr::DK_Note, 396 "while in macro instantiation"); 397} 398 399bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 400 if (FatalAssemblerWarnings) 401 return Error(L, Msg, Ranges); 402 PrintMessage(L, SourceMgr::DK_Warning, Msg, Ranges); 403 PrintMacroInstantiations(); 404 return false; 405} 406 407bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) { 408 HadError = true; 409 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges); 410 PrintMacroInstantiations(); 411 return true; 412} 413 414bool AsmParser::EnterIncludeFile(const std::string &Filename) { 415 std::string IncludedFile; 416 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile); 417 if (NewBuf == -1) 418 return true; 419 420 CurBuffer = NewBuf; 421 422 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 423 424 return false; 425} 426 427void AsmParser::JumpToLoc(SMLoc Loc) { 428 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc); 429 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer()); 430} 431 432const AsmToken &AsmParser::Lex() { 433 const AsmToken *tok = &Lexer.Lex(); 434 435 if (tok->is(AsmToken::Eof)) { 436 // If this is the end of an included file, pop the parent file off the 437 // include stack. 438 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer); 439 if (ParentIncludeLoc != SMLoc()) { 440 JumpToLoc(ParentIncludeLoc); 441 tok = &Lexer.Lex(); 442 } 443 } 444 445 if (tok->is(AsmToken::Error)) 446 Error(Lexer.getErrLoc(), Lexer.getErr()); 447 448 return *tok; 449} 450 451bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) { 452 // Create the initial section, if requested. 453 if (!NoInitialTextSection) 454 Out.InitSections(); 455 456 // Prime the lexer. 457 Lex(); 458 459 HadError = false; 460 AsmCond StartingCondState = TheCondState; 461 462 // While we have input, parse each statement. 463 while (Lexer.isNot(AsmToken::Eof)) { 464 if (!ParseStatement()) continue; 465 466 // We had an error, validate that one was emitted and recover by skipping to 467 // the next line. 468 assert(HadError && "Parse statement returned an error, but none emitted!"); 469 EatToEndOfStatement(); 470 } 471 472 if (TheCondState.TheCond != StartingCondState.TheCond || 473 TheCondState.Ignore != StartingCondState.Ignore) 474 return TokError("unmatched .ifs or .elses"); 475 476 // Check to see there are no empty DwarfFile slots. 477 const std::vector<MCDwarfFile *> &MCDwarfFiles = 478 getContext().getMCDwarfFiles(); 479 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) { 480 if (!MCDwarfFiles[i]) 481 TokError("unassigned file number: " + Twine(i) + " for .file directives"); 482 } 483 484 // Check to see that all assembler local symbols were actually defined. 485 // Targets that don't do subsections via symbols may not want this, though, 486 // so conservatively exclude them. Only do this if we're finalizing, though, 487 // as otherwise we won't necessarilly have seen everything yet. 488 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) { 489 const MCContext::SymbolTable &Symbols = getContext().getSymbols(); 490 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(), 491 e = Symbols.end(); 492 i != e; ++i) { 493 MCSymbol *Sym = i->getValue(); 494 // Variable symbols may not be marked as defined, so check those 495 // explicitly. If we know it's a variable, we have a definition for 496 // the purposes of this check. 497 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined()) 498 // FIXME: We would really like to refer back to where the symbol was 499 // first referenced for a source location. We need to add something 500 // to track that. Currently, we just point to the end of the file. 501 PrintMessage(getLexer().getLoc(), SourceMgr::DK_Error, 502 "assembler local symbol '" + Sym->getName() + 503 "' not defined"); 504 } 505 } 506 507 508 // Finalize the output stream if there are no errors and if the client wants 509 // us to. 510 if (!HadError && !NoFinalize) 511 Out.Finish(); 512 513 return HadError; 514} 515 516void AsmParser::CheckForValidSection() { 517 if (!getStreamer().getCurrentSection()) { 518 TokError("expected section directive before assembly directive"); 519 Out.SwitchSection(Ctx.getMachOSection( 520 "__TEXT", "__text", 521 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS, 522 0, SectionKind::getText())); 523 } 524} 525 526/// EatToEndOfStatement - Throw away the rest of the line for testing purposes. 527void AsmParser::EatToEndOfStatement() { 528 while (Lexer.isNot(AsmToken::EndOfStatement) && 529 Lexer.isNot(AsmToken::Eof)) 530 Lex(); 531 532 // Eat EOL. 533 if (Lexer.is(AsmToken::EndOfStatement)) 534 Lex(); 535} 536 537StringRef AsmParser::ParseStringToEndOfStatement() { 538 const char *Start = getTok().getLoc().getPointer(); 539 540 while (Lexer.isNot(AsmToken::EndOfStatement) && 541 Lexer.isNot(AsmToken::Eof)) 542 Lex(); 543 544 const char *End = getTok().getLoc().getPointer(); 545 return StringRef(Start, End - Start); 546} 547 548/// ParseParenExpr - Parse a paren expression and return it. 549/// NOTE: This assumes the leading '(' has already been consumed. 550/// 551/// parenexpr ::= expr) 552/// 553bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { 554 if (ParseExpression(Res)) return true; 555 if (Lexer.isNot(AsmToken::RParen)) 556 return TokError("expected ')' in parentheses expression"); 557 EndLoc = Lexer.getLoc(); 558 Lex(); 559 return false; 560} 561 562/// ParseBracketExpr - Parse a bracket expression and return it. 563/// NOTE: This assumes the leading '[' has already been consumed. 564/// 565/// bracketexpr ::= expr] 566/// 567bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { 568 if (ParseExpression(Res)) return true; 569 if (Lexer.isNot(AsmToken::RBrac)) 570 return TokError("expected ']' in brackets expression"); 571 EndLoc = Lexer.getLoc(); 572 Lex(); 573 return false; 574} 575 576/// ParsePrimaryExpr - Parse a primary expression and return it. 577/// primaryexpr ::= (parenexpr 578/// primaryexpr ::= symbol 579/// primaryexpr ::= number 580/// primaryexpr ::= '.' 581/// primaryexpr ::= ~,+,- primaryexpr 582bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) { 583 switch (Lexer.getKind()) { 584 default: 585 return TokError("unknown token in expression"); 586 // If we have an error assume that we've already handled it. 587 case AsmToken::Error: 588 return true; 589 case AsmToken::Exclaim: 590 Lex(); // Eat the operator. 591 if (ParsePrimaryExpr(Res, EndLoc)) 592 return true; 593 Res = MCUnaryExpr::CreateLNot(Res, getContext()); 594 return false; 595 case AsmToken::Dollar: 596 case AsmToken::String: 597 case AsmToken::Identifier: { 598 EndLoc = Lexer.getLoc(); 599 600 StringRef Identifier; 601 if (ParseIdentifier(Identifier)) 602 return true; 603 604 // This is a symbol reference. 605 std::pair<StringRef, StringRef> Split = Identifier.split('@'); 606 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first); 607 608 // Lookup the symbol variant if used. 609 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; 610 if (Split.first.size() != Identifier.size()) { 611 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second); 612 if (Variant == MCSymbolRefExpr::VK_Invalid) { 613 Variant = MCSymbolRefExpr::VK_None; 614 return TokError("invalid variant '" + Split.second + "'"); 615 } 616 } 617 618 // If this is an absolute variable reference, substitute it now to preserve 619 // semantics in the face of reassignment. 620 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) { 621 if (Variant) 622 return Error(EndLoc, "unexpected modifier on variable reference"); 623 624 Res = Sym->getVariableValue(); 625 return false; 626 } 627 628 // Otherwise create a symbol ref. 629 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext()); 630 return false; 631 } 632 case AsmToken::Integer: { 633 SMLoc Loc = getTok().getLoc(); 634 int64_t IntVal = getTok().getIntVal(); 635 Res = MCConstantExpr::Create(IntVal, getContext()); 636 EndLoc = Lexer.getLoc(); 637 Lex(); // Eat token. 638 // Look for 'b' or 'f' following an Integer as a directional label 639 if (Lexer.getKind() == AsmToken::Identifier) { 640 StringRef IDVal = getTok().getString(); 641 if (IDVal == "f" || IDVal == "b"){ 642 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal, 643 IDVal == "f" ? 1 : 0); 644 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, 645 getContext()); 646 if(IDVal == "b" && Sym->isUndefined()) 647 return Error(Loc, "invalid reference to undefined symbol"); 648 EndLoc = Lexer.getLoc(); 649 Lex(); // Eat identifier. 650 } 651 } 652 return false; 653 } 654 case AsmToken::Real: { 655 APFloat RealVal(APFloat::IEEEdouble, getTok().getString()); 656 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); 657 Res = MCConstantExpr::Create(IntVal, getContext()); 658 Lex(); // Eat token. 659 return false; 660 } 661 case AsmToken::Dot: { 662 // This is a '.' reference, which references the current PC. Emit a 663 // temporary label to the streamer and refer to it. 664 MCSymbol *Sym = Ctx.CreateTempSymbol(); 665 Out.EmitLabel(Sym); 666 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext()); 667 EndLoc = Lexer.getLoc(); 668 Lex(); // Eat identifier. 669 return false; 670 } 671 case AsmToken::LParen: 672 Lex(); // Eat the '('. 673 return ParseParenExpr(Res, EndLoc); 674 case AsmToken::LBrac: 675 if (!PlatformParser->HasBracketExpressions()) 676 return TokError("brackets expression not supported on this target"); 677 Lex(); // Eat the '['. 678 return ParseBracketExpr(Res, EndLoc); 679 case AsmToken::Minus: 680 Lex(); // Eat the operator. 681 if (ParsePrimaryExpr(Res, EndLoc)) 682 return true; 683 Res = MCUnaryExpr::CreateMinus(Res, getContext()); 684 return false; 685 case AsmToken::Plus: 686 Lex(); // Eat the operator. 687 if (ParsePrimaryExpr(Res, EndLoc)) 688 return true; 689 Res = MCUnaryExpr::CreatePlus(Res, getContext()); 690 return false; 691 case AsmToken::Tilde: 692 Lex(); // Eat the operator. 693 if (ParsePrimaryExpr(Res, EndLoc)) 694 return true; 695 Res = MCUnaryExpr::CreateNot(Res, getContext()); 696 return false; 697 } 698} 699 700bool AsmParser::ParseExpression(const MCExpr *&Res) { 701 SMLoc EndLoc; 702 return ParseExpression(Res, EndLoc); 703} 704 705const MCExpr * 706AsmParser::ApplyModifierToExpr(const MCExpr *E, 707 MCSymbolRefExpr::VariantKind Variant) { 708 // Recurse over the given expression, rebuilding it to apply the given variant 709 // if there is exactly one symbol. 710 switch (E->getKind()) { 711 case MCExpr::Target: 712 case MCExpr::Constant: 713 return 0; 714 715 case MCExpr::SymbolRef: { 716 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E); 717 718 if (SRE->getKind() != MCSymbolRefExpr::VK_None) { 719 TokError("invalid variant on expression '" + 720 getTok().getIdentifier() + "' (already modified)"); 721 return E; 722 } 723 724 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext()); 725 } 726 727 case MCExpr::Unary: { 728 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E); 729 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant); 730 if (!Sub) 731 return 0; 732 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext()); 733 } 734 735 case MCExpr::Binary: { 736 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E); 737 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant); 738 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant); 739 740 if (!LHS && !RHS) 741 return 0; 742 743 if (!LHS) LHS = BE->getLHS(); 744 if (!RHS) RHS = BE->getRHS(); 745 746 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext()); 747 } 748 } 749 750 assert(0 && "Invalid expression kind!"); 751 return 0; 752} 753 754/// ParseExpression - Parse an expression and return it. 755/// 756/// expr ::= expr &&,|| expr -> lowest. 757/// expr ::= expr |,^,&,! expr 758/// expr ::= expr ==,!=,<>,<,<=,>,>= expr 759/// expr ::= expr <<,>> expr 760/// expr ::= expr +,- expr 761/// expr ::= expr *,/,% expr -> highest. 762/// expr ::= primaryexpr 763/// 764bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) { 765 // Parse the expression. 766 Res = 0; 767 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc)) 768 return true; 769 770 // As a special case, we support 'a op b @ modifier' by rewriting the 771 // expression to include the modifier. This is inefficient, but in general we 772 // expect users to use 'a@modifier op b'. 773 if (Lexer.getKind() == AsmToken::At) { 774 Lex(); 775 776 if (Lexer.isNot(AsmToken::Identifier)) 777 return TokError("unexpected symbol modifier following '@'"); 778 779 MCSymbolRefExpr::VariantKind Variant = 780 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier()); 781 if (Variant == MCSymbolRefExpr::VK_Invalid) 782 return TokError("invalid variant '" + getTok().getIdentifier() + "'"); 783 784 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant); 785 if (!ModifiedRes) { 786 return TokError("invalid modifier '" + getTok().getIdentifier() + 787 "' (no symbols present)"); 788 return true; 789 } 790 791 Res = ModifiedRes; 792 Lex(); 793 } 794 795 // Try to constant fold it up front, if possible. 796 int64_t Value; 797 if (Res->EvaluateAsAbsolute(Value)) 798 Res = MCConstantExpr::Create(Value, getContext()); 799 800 return false; 801} 802 803bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { 804 Res = 0; 805 return ParseParenExpr(Res, EndLoc) || 806 ParseBinOpRHS(1, Res, EndLoc); 807} 808 809bool AsmParser::ParseAbsoluteExpression(int64_t &Res) { 810 const MCExpr *Expr; 811 812 SMLoc StartLoc = Lexer.getLoc(); 813 if (ParseExpression(Expr)) 814 return true; 815 816 if (!Expr->EvaluateAsAbsolute(Res)) 817 return Error(StartLoc, "expected absolute expression"); 818 819 return false; 820} 821 822static unsigned getBinOpPrecedence(AsmToken::TokenKind K, 823 MCBinaryExpr::Opcode &Kind) { 824 switch (K) { 825 default: 826 return 0; // not a binop. 827 828 // Lowest Precedence: &&, || 829 case AsmToken::AmpAmp: 830 Kind = MCBinaryExpr::LAnd; 831 return 1; 832 case AsmToken::PipePipe: 833 Kind = MCBinaryExpr::LOr; 834 return 1; 835 836 837 // Low Precedence: |, &, ^ 838 // 839 // FIXME: gas seems to support '!' as an infix operator? 840 case AsmToken::Pipe: 841 Kind = MCBinaryExpr::Or; 842 return 2; 843 case AsmToken::Caret: 844 Kind = MCBinaryExpr::Xor; 845 return 2; 846 case AsmToken::Amp: 847 Kind = MCBinaryExpr::And; 848 return 2; 849 850 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >= 851 case AsmToken::EqualEqual: 852 Kind = MCBinaryExpr::EQ; 853 return 3; 854 case AsmToken::ExclaimEqual: 855 case AsmToken::LessGreater: 856 Kind = MCBinaryExpr::NE; 857 return 3; 858 case AsmToken::Less: 859 Kind = MCBinaryExpr::LT; 860 return 3; 861 case AsmToken::LessEqual: 862 Kind = MCBinaryExpr::LTE; 863 return 3; 864 case AsmToken::Greater: 865 Kind = MCBinaryExpr::GT; 866 return 3; 867 case AsmToken::GreaterEqual: 868 Kind = MCBinaryExpr::GTE; 869 return 3; 870 871 // Intermediate Precedence: <<, >> 872 case AsmToken::LessLess: 873 Kind = MCBinaryExpr::Shl; 874 return 4; 875 case AsmToken::GreaterGreater: 876 Kind = MCBinaryExpr::Shr; 877 return 4; 878 879 // High Intermediate Precedence: +, - 880 case AsmToken::Plus: 881 Kind = MCBinaryExpr::Add; 882 return 5; 883 case AsmToken::Minus: 884 Kind = MCBinaryExpr::Sub; 885 return 5; 886 887 // Highest Precedence: *, /, % 888 case AsmToken::Star: 889 Kind = MCBinaryExpr::Mul; 890 return 6; 891 case AsmToken::Slash: 892 Kind = MCBinaryExpr::Div; 893 return 6; 894 case AsmToken::Percent: 895 Kind = MCBinaryExpr::Mod; 896 return 6; 897 } 898} 899 900 901/// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'. 902/// Res contains the LHS of the expression on input. 903bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, 904 SMLoc &EndLoc) { 905 while (1) { 906 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; 907 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind); 908 909 // If the next token is lower precedence than we are allowed to eat, return 910 // successfully with what we ate already. 911 if (TokPrec < Precedence) 912 return false; 913 914 Lex(); 915 916 // Eat the next primary expression. 917 const MCExpr *RHS; 918 if (ParsePrimaryExpr(RHS, EndLoc)) return true; 919 920 // If BinOp binds less tightly with RHS than the operator after RHS, let 921 // the pending operator take RHS as its LHS. 922 MCBinaryExpr::Opcode Dummy; 923 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy); 924 if (TokPrec < NextTokPrec) { 925 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true; 926 } 927 928 // Merge LHS and RHS according to operator. 929 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext()); 930 } 931} 932 933 934 935 936/// ParseStatement: 937/// ::= EndOfStatement 938/// ::= Label* Directive ...Operands... EndOfStatement 939/// ::= Label* Identifier OperandList* EndOfStatement 940bool AsmParser::ParseStatement() { 941 if (Lexer.is(AsmToken::EndOfStatement)) { 942 Out.AddBlankLine(); 943 Lex(); 944 return false; 945 } 946 947 // Statements always start with an identifier or are a full line comment. 948 AsmToken ID = getTok(); 949 SMLoc IDLoc = ID.getLoc(); 950 StringRef IDVal; 951 int64_t LocalLabelVal = -1; 952 // A full line comment is a '#' as the first token. 953 if (Lexer.is(AsmToken::Hash)) 954 return ParseCppHashLineFilenameComment(IDLoc); 955 956 // Allow an integer followed by a ':' as a directional local label. 957 if (Lexer.is(AsmToken::Integer)) { 958 LocalLabelVal = getTok().getIntVal(); 959 if (LocalLabelVal < 0) { 960 if (!TheCondState.Ignore) 961 return TokError("unexpected token at start of statement"); 962 IDVal = ""; 963 } 964 else { 965 IDVal = getTok().getString(); 966 Lex(); // Consume the integer token to be used as an identifier token. 967 if (Lexer.getKind() != AsmToken::Colon) { 968 if (!TheCondState.Ignore) 969 return TokError("unexpected token at start of statement"); 970 } 971 } 972 973 } else if (Lexer.is(AsmToken::Dot)) { 974 // Treat '.' as a valid identifier in this context. 975 Lex(); 976 IDVal = "."; 977 978 } else if (ParseIdentifier(IDVal)) { 979 if (!TheCondState.Ignore) 980 return TokError("unexpected token at start of statement"); 981 IDVal = ""; 982 } 983 984 985 // Handle conditional assembly here before checking for skipping. We 986 // have to do this so that .endif isn't skipped in a ".if 0" block for 987 // example. 988 if (IDVal == ".if") 989 return ParseDirectiveIf(IDLoc); 990 if (IDVal == ".ifdef") 991 return ParseDirectiveIfdef(IDLoc, true); 992 if (IDVal == ".ifndef" || IDVal == ".ifnotdef") 993 return ParseDirectiveIfdef(IDLoc, false); 994 if (IDVal == ".elseif") 995 return ParseDirectiveElseIf(IDLoc); 996 if (IDVal == ".else") 997 return ParseDirectiveElse(IDLoc); 998 if (IDVal == ".endif") 999 return ParseDirectiveEndIf(IDLoc); 1000 1001 // If we are in a ".if 0" block, ignore this statement. 1002 if (TheCondState.Ignore) { 1003 EatToEndOfStatement(); 1004 return false; 1005 } 1006 1007 // FIXME: Recurse on local labels? 1008 1009 // See what kind of statement we have. 1010 switch (Lexer.getKind()) { 1011 case AsmToken::Colon: { 1012 CheckForValidSection(); 1013 1014 // identifier ':' -> Label. 1015 Lex(); 1016 1017 // Diagnose attempt to use '.' as a label. 1018 if (IDVal == ".") 1019 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label"); 1020 1021 // Diagnose attempt to use a variable as a label. 1022 // 1023 // FIXME: Diagnostics. Note the location of the definition as a label. 1024 // FIXME: This doesn't diagnose assignment to a symbol which has been 1025 // implicitly marked as external. 1026 MCSymbol *Sym; 1027 if (LocalLabelVal == -1) 1028 Sym = getContext().GetOrCreateSymbol(IDVal); 1029 else 1030 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal); 1031 if (!Sym->isUndefined() || Sym->isVariable()) 1032 return Error(IDLoc, "invalid symbol redefinition"); 1033 1034 // Emit the label. 1035 Out.EmitLabel(Sym); 1036 1037 // Consume any end of statement token, if present, to avoid spurious 1038 // AddBlankLine calls(). 1039 if (Lexer.is(AsmToken::EndOfStatement)) { 1040 Lex(); 1041 if (Lexer.is(AsmToken::Eof)) 1042 return false; 1043 } 1044 1045 return ParseStatement(); 1046 } 1047 1048 case AsmToken::Equal: 1049 // identifier '=' ... -> assignment statement 1050 Lex(); 1051 1052 return ParseAssignment(IDVal, true); 1053 1054 default: // Normal instruction or directive. 1055 break; 1056 } 1057 1058 // If macros are enabled, check to see if this is a macro instantiation. 1059 if (MacrosEnabled) 1060 if (const Macro *M = MacroMap.lookup(IDVal)) 1061 return HandleMacroEntry(IDVal, IDLoc, M); 1062 1063 // Otherwise, we have a normal instruction or directive. 1064 if (IDVal[0] == '.' && IDVal != ".") { 1065 // Assembler features 1066 if (IDVal == ".set" || IDVal == ".equ") 1067 return ParseDirectiveSet(IDVal, true); 1068 if (IDVal == ".equiv") 1069 return ParseDirectiveSet(IDVal, false); 1070 1071 // Data directives 1072 1073 if (IDVal == ".ascii") 1074 return ParseDirectiveAscii(IDVal, false); 1075 if (IDVal == ".asciz" || IDVal == ".string") 1076 return ParseDirectiveAscii(IDVal, true); 1077 1078 if (IDVal == ".byte") 1079 return ParseDirectiveValue(1); 1080 if (IDVal == ".short") 1081 return ParseDirectiveValue(2); 1082 if (IDVal == ".value") 1083 return ParseDirectiveValue(2); 1084 if (IDVal == ".2byte") 1085 return ParseDirectiveValue(2); 1086 if (IDVal == ".long") 1087 return ParseDirectiveValue(4); 1088 if (IDVal == ".int") 1089 return ParseDirectiveValue(4); 1090 if (IDVal == ".4byte") 1091 return ParseDirectiveValue(4); 1092 if (IDVal == ".quad") 1093 return ParseDirectiveValue(8); 1094 if (IDVal == ".8byte") 1095 return ParseDirectiveValue(8); 1096 if (IDVal == ".single" || IDVal == ".float") 1097 return ParseDirectiveRealValue(APFloat::IEEEsingle); 1098 if (IDVal == ".double") 1099 return ParseDirectiveRealValue(APFloat::IEEEdouble); 1100 1101 if (IDVal == ".align") { 1102 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1103 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1); 1104 } 1105 if (IDVal == ".align32") { 1106 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes(); 1107 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4); 1108 } 1109 if (IDVal == ".balign") 1110 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1); 1111 if (IDVal == ".balignw") 1112 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2); 1113 if (IDVal == ".balignl") 1114 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4); 1115 if (IDVal == ".p2align") 1116 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1); 1117 if (IDVal == ".p2alignw") 1118 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2); 1119 if (IDVal == ".p2alignl") 1120 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4); 1121 1122 if (IDVal == ".org") 1123 return ParseDirectiveOrg(); 1124 1125 if (IDVal == ".fill") 1126 return ParseDirectiveFill(); 1127 if (IDVal == ".space" || IDVal == ".skip") 1128 return ParseDirectiveSpace(); 1129 if (IDVal == ".zero") 1130 return ParseDirectiveZero(); 1131 1132 // Symbol attribute directives 1133 1134 if (IDVal == ".globl" || IDVal == ".global") 1135 return ParseDirectiveSymbolAttribute(MCSA_Global); 1136 if (IDVal == ".indirect_symbol") 1137 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol); 1138 if (IDVal == ".lazy_reference") 1139 return ParseDirectiveSymbolAttribute(MCSA_LazyReference); 1140 if (IDVal == ".no_dead_strip") 1141 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip); 1142 if (IDVal == ".symbol_resolver") 1143 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver); 1144 if (IDVal == ".private_extern") 1145 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern); 1146 if (IDVal == ".reference") 1147 return ParseDirectiveSymbolAttribute(MCSA_Reference); 1148 if (IDVal == ".weak_definition") 1149 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition); 1150 if (IDVal == ".weak_reference") 1151 return ParseDirectiveSymbolAttribute(MCSA_WeakReference); 1152 if (IDVal == ".weak_def_can_be_hidden") 1153 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate); 1154 1155 if (IDVal == ".comm" || IDVal == ".common") 1156 return ParseDirectiveComm(/*IsLocal=*/false); 1157 if (IDVal == ".lcomm") 1158 return ParseDirectiveComm(/*IsLocal=*/true); 1159 1160 if (IDVal == ".abort") 1161 return ParseDirectiveAbort(); 1162 if (IDVal == ".include") 1163 return ParseDirectiveInclude(); 1164 1165 if (IDVal == ".code16") 1166 return TokError(Twine(IDVal) + " not supported yet"); 1167 1168 // Look up the handler in the handler table. 1169 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler = 1170 DirectiveMap.lookup(IDVal); 1171 if (Handler.first) 1172 return (*Handler.second)(Handler.first, IDVal, IDLoc); 1173 1174 // Target hook for parsing target specific directives. 1175 if (!getTargetParser().ParseDirective(ID)) 1176 return false; 1177 1178 bool retval = Warning(IDLoc, "ignoring directive for now"); 1179 EatToEndOfStatement(); 1180 return retval; 1181 } 1182 1183 CheckForValidSection(); 1184 1185 // Canonicalize the opcode to lower case. 1186 SmallString<128> Opcode; 1187 for (unsigned i = 0, e = IDVal.size(); i != e; ++i) 1188 Opcode.push_back(tolower(IDVal[i])); 1189 1190 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands; 1191 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc, 1192 ParsedOperands); 1193 1194 // Dump the parsed representation, if requested. 1195 if (getShowParsedOperands()) { 1196 SmallString<256> Str; 1197 raw_svector_ostream OS(Str); 1198 OS << "parsed instruction: ["; 1199 for (unsigned i = 0; i != ParsedOperands.size(); ++i) { 1200 if (i != 0) 1201 OS << ", "; 1202 ParsedOperands[i]->print(OS); 1203 } 1204 OS << "]"; 1205 1206 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str()); 1207 } 1208 1209 // If parsing succeeded, match the instruction. 1210 if (!HadError) 1211 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands, 1212 Out); 1213 1214 // Free any parsed operands. 1215 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i) 1216 delete ParsedOperands[i]; 1217 1218 // Don't skip the rest of the line, the instruction parser is responsible for 1219 // that. 1220 return false; 1221} 1222 1223/// EatToEndOfLine uses the Lexer to eat the characters to the end of the line 1224/// since they may not be able to be tokenized to get to the end of line token. 1225void AsmParser::EatToEndOfLine() { 1226 Lexer.LexUntilEndOfLine(); 1227 // Eat EOL. 1228 Lex(); 1229} 1230 1231/// ParseCppHashLineFilenameComment as this: 1232/// ::= # number "filename" 1233/// or just as a full line comment if it doesn't have a number and a string. 1234bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) { 1235 Lex(); // Eat the hash token. 1236 1237 if (getLexer().isNot(AsmToken::Integer)) { 1238 // Consume the line since in cases it is not a well-formed line directive, 1239 // as if were simply a full line comment. 1240 EatToEndOfLine(); 1241 return false; 1242 } 1243 1244 int64_t LineNumber = getTok().getIntVal(); 1245 Lex(); 1246 1247 if (getLexer().isNot(AsmToken::String)) { 1248 EatToEndOfLine(); 1249 return false; 1250 } 1251 1252 StringRef Filename = getTok().getString(); 1253 // Get rid of the enclosing quotes. 1254 Filename = Filename.substr(1, Filename.size()-2); 1255 1256 // Save the SMLoc, Filename and LineNumber for later use by diagnostics. 1257 CppHashLoc = L; 1258 CppHashFilename = Filename; 1259 CppHashLineNumber = LineNumber; 1260 1261 // Ignore any trailing characters, they're just comment. 1262 EatToEndOfLine(); 1263 return false; 1264} 1265 1266/// DiagHandler - will use the the last parsed cpp hash line filename comment 1267/// for the Filename and LineNo if any in the diagnostic. 1268void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { 1269 const AsmParser *Parser = static_cast<const AsmParser*>(Context); 1270 raw_ostream &OS = errs(); 1271 1272 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); 1273 const SMLoc &DiagLoc = Diag.getLoc(); 1274 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1275 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc); 1276 1277 // Like SourceMgr::PrintMessage() we need to print the include stack if any 1278 // before printing the message. 1279 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc); 1280 if (DiagCurBuffer > 0) { 1281 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer); 1282 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS); 1283 } 1284 1285 // If we have not parsed a cpp hash line filename comment or the source 1286 // manager changed or buffer changed (like in a nested include) then just 1287 // print the normal diagnostic using its Filename and LineNo. 1288 if (!Parser->CppHashLineNumber || 1289 &DiagSrcMgr != &Parser->SrcMgr || 1290 DiagBuf != CppHashBuf) { 1291 Diag.print(0, OS); 1292 return; 1293 } 1294 1295 // Use the CppHashFilename and calculate a line number based on the 1296 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for 1297 // the diagnostic. 1298 const std::string Filename = Parser->CppHashFilename; 1299 1300 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf); 1301 int CppHashLocLineNo = 1302 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf); 1303 int LineNo = Parser->CppHashLineNumber - 1 + 1304 (DiagLocLineNo - CppHashLocLineNo); 1305 1306 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), 1307 Filename, LineNo, Diag.getColumnNo(), 1308 Diag.getKind(), Diag.getMessage(), 1309 Diag.getLineContents(), 1310 Diag.getRanges(), Diag.getShowLine()); 1311 1312 NewDiag.print(0, OS); 1313} 1314 1315bool AsmParser::expandMacro(SmallString<256> &Buf, StringRef Body, 1316 const std::vector<StringRef> &Parameters, 1317 const std::vector<std::vector<AsmToken> > &A, 1318 const SMLoc &L) { 1319 raw_svector_ostream OS(Buf); 1320 unsigned NParameters = Parameters.size(); 1321 if (NParameters != 0 && NParameters != A.size()) 1322 return Error(L, "Wrong number of arguments"); 1323 1324 while (!Body.empty()) { 1325 // Scan for the next substitution. 1326 std::size_t End = Body.size(), Pos = 0; 1327 for (; Pos != End; ++Pos) { 1328 // Check for a substitution or escape. 1329 if (!NParameters) { 1330 // This macro has no parameters, look for $0, $1, etc. 1331 if (Body[Pos] != '$' || Pos + 1 == End) 1332 continue; 1333 1334 char Next = Body[Pos + 1]; 1335 if (Next == '$' || Next == 'n' || isdigit(Next)) 1336 break; 1337 } else { 1338 // This macro has parameters, look for \foo, \bar, etc. 1339 if (Body[Pos] == '\\' && Pos + 1 != End) 1340 break; 1341 } 1342 } 1343 1344 // Add the prefix. 1345 OS << Body.slice(0, Pos); 1346 1347 // Check if we reached the end. 1348 if (Pos == End) 1349 break; 1350 1351 if (!NParameters) { 1352 switch (Body[Pos+1]) { 1353 // $$ => $ 1354 case '$': 1355 OS << '$'; 1356 break; 1357 1358 // $n => number of arguments 1359 case 'n': 1360 OS << A.size(); 1361 break; 1362 1363 // $[0-9] => argument 1364 default: { 1365 // Missing arguments are ignored. 1366 unsigned Index = Body[Pos+1] - '0'; 1367 if (Index >= A.size()) 1368 break; 1369 1370 // Otherwise substitute with the token values, with spaces eliminated. 1371 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(), 1372 ie = A[Index].end(); it != ie; ++it) 1373 OS << it->getString(); 1374 break; 1375 } 1376 } 1377 Pos += 2; 1378 } else { 1379 unsigned I = Pos + 1; 1380 while (isalnum(Body[I]) && I + 1 != End) 1381 ++I; 1382 1383 const char *Begin = Body.data() + Pos +1; 1384 StringRef Argument(Begin, I - (Pos +1)); 1385 unsigned Index = 0; 1386 for (; Index < NParameters; ++Index) 1387 if (Parameters[Index] == Argument) 1388 break; 1389 1390 // FIXME: We should error at the macro definition. 1391 if (Index == NParameters) 1392 return Error(L, "Parameter not found"); 1393 1394 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(), 1395 ie = A[Index].end(); it != ie; ++it) 1396 OS << it->getString(); 1397 1398 Pos += 1 + Argument.size(); 1399 } 1400 // Update the scan point. 1401 Body = Body.substr(Pos); 1402 } 1403 1404 // We include the .endmacro in the buffer as our queue to exit the macro 1405 // instantiation. 1406 OS << ".endmacro\n"; 1407 return false; 1408} 1409 1410MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL, 1411 MemoryBuffer *I) 1412 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL) 1413{ 1414} 1415 1416bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc, 1417 const Macro *M) { 1418 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate 1419 // this, although we should protect against infinite loops. 1420 if (ActiveMacros.size() == 20) 1421 return TokError("macros cannot be nested more than 20 levels deep"); 1422 1423 // Parse the macro instantiation arguments. 1424 std::vector<std::vector<AsmToken> > MacroArguments; 1425 MacroArguments.push_back(std::vector<AsmToken>()); 1426 unsigned ParenLevel = 0; 1427 for (;;) { 1428 if (Lexer.is(AsmToken::Eof)) 1429 return TokError("unexpected token in macro instantiation"); 1430 if (Lexer.is(AsmToken::EndOfStatement)) 1431 break; 1432 1433 // If we aren't inside parentheses and this is a comma, start a new token 1434 // list. 1435 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) { 1436 MacroArguments.push_back(std::vector<AsmToken>()); 1437 } else { 1438 // Adjust the current parentheses level. 1439 if (Lexer.is(AsmToken::LParen)) 1440 ++ParenLevel; 1441 else if (Lexer.is(AsmToken::RParen) && ParenLevel) 1442 --ParenLevel; 1443 1444 // Append the token to the current argument list. 1445 MacroArguments.back().push_back(getTok()); 1446 } 1447 Lex(); 1448 } 1449 1450 // Macro instantiation is lexical, unfortunately. We construct a new buffer 1451 // to hold the macro body with substitutions. 1452 SmallString<256> Buf; 1453 StringRef Body = M->Body; 1454 1455 if (expandMacro(Buf, Body, M->Parameters, MacroArguments, getTok().getLoc())) 1456 return true; 1457 1458 MemoryBuffer *Instantiation = 1459 MemoryBuffer::getMemBufferCopy(Buf.str(), "<instantiation>"); 1460 1461 // Create the macro instantiation object and add to the current macro 1462 // instantiation stack. 1463 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc, 1464 getTok().getLoc(), 1465 Instantiation); 1466 ActiveMacros.push_back(MI); 1467 1468 // Jump to the macro instantiation and prime the lexer. 1469 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc()); 1470 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)); 1471 Lex(); 1472 1473 return false; 1474} 1475 1476void AsmParser::HandleMacroExit() { 1477 // Jump to the EndOfStatement we should return to, and consume it. 1478 JumpToLoc(ActiveMacros.back()->ExitLoc); 1479 Lex(); 1480 1481 // Pop the instantiation entry. 1482 delete ActiveMacros.back(); 1483 ActiveMacros.pop_back(); 1484} 1485 1486static void MarkUsed(const MCExpr *Value) { 1487 switch (Value->getKind()) { 1488 case MCExpr::Binary: 1489 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getLHS()); 1490 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getRHS()); 1491 break; 1492 case MCExpr::Target: 1493 case MCExpr::Constant: 1494 break; 1495 case MCExpr::SymbolRef: { 1496 static_cast<const MCSymbolRefExpr*>(Value)->getSymbol().setUsed(true); 1497 break; 1498 } 1499 case MCExpr::Unary: 1500 MarkUsed(static_cast<const MCUnaryExpr*>(Value)->getSubExpr()); 1501 break; 1502 } 1503} 1504 1505bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) { 1506 // FIXME: Use better location, we should use proper tokens. 1507 SMLoc EqualLoc = Lexer.getLoc(); 1508 1509 const MCExpr *Value; 1510 if (ParseExpression(Value)) 1511 return true; 1512 1513 MarkUsed(Value); 1514 1515 if (Lexer.isNot(AsmToken::EndOfStatement)) 1516 return TokError("unexpected token in assignment"); 1517 1518 // Error on assignment to '.'. 1519 if (Name == ".") { 1520 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported " 1521 "(use '.space' or '.org').)")); 1522 } 1523 1524 // Eat the end of statement marker. 1525 Lex(); 1526 1527 // Validate that the LHS is allowed to be a variable (either it has not been 1528 // used as a symbol, or it is an absolute symbol). 1529 MCSymbol *Sym = getContext().LookupSymbol(Name); 1530 if (Sym) { 1531 // Diagnose assignment to a label. 1532 // 1533 // FIXME: Diagnostics. Note the location of the definition as a label. 1534 // FIXME: Diagnose assignment to protected identifier (e.g., register name). 1535 if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable()) 1536 ; // Allow redefinitions of undefined symbols only used in directives. 1537 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef)) 1538 return Error(EqualLoc, "redefinition of '" + Name + "'"); 1539 else if (!Sym->isVariable()) 1540 return Error(EqualLoc, "invalid assignment to '" + Name + "'"); 1541 else if (!isa<MCConstantExpr>(Sym->getVariableValue())) 1542 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" + 1543 Name + "'"); 1544 1545 // Don't count these checks as uses. 1546 Sym->setUsed(false); 1547 } else 1548 Sym = getContext().GetOrCreateSymbol(Name); 1549 1550 // FIXME: Handle '.'. 1551 1552 // Do the assignment. 1553 Out.EmitAssignment(Sym, Value); 1554 1555 return false; 1556} 1557 1558/// ParseIdentifier: 1559/// ::= identifier 1560/// ::= string 1561bool AsmParser::ParseIdentifier(StringRef &Res) { 1562 // The assembler has relaxed rules for accepting identifiers, in particular we 1563 // allow things like '.globl $foo', which would normally be separate 1564 // tokens. At this level, we have already lexed so we cannot (currently) 1565 // handle this as a context dependent token, instead we detect adjacent tokens 1566 // and return the combined identifier. 1567 if (Lexer.is(AsmToken::Dollar)) { 1568 SMLoc DollarLoc = getLexer().getLoc(); 1569 1570 // Consume the dollar sign, and check for a following identifier. 1571 Lex(); 1572 if (Lexer.isNot(AsmToken::Identifier)) 1573 return true; 1574 1575 // We have a '$' followed by an identifier, make sure they are adjacent. 1576 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer()) 1577 return true; 1578 1579 // Construct the joined identifier and consume the token. 1580 Res = StringRef(DollarLoc.getPointer(), 1581 getTok().getIdentifier().size() + 1); 1582 Lex(); 1583 return false; 1584 } 1585 1586 if (Lexer.isNot(AsmToken::Identifier) && 1587 Lexer.isNot(AsmToken::String)) 1588 return true; 1589 1590 Res = getTok().getIdentifier(); 1591 1592 Lex(); // Consume the identifier token. 1593 1594 return false; 1595} 1596 1597/// ParseDirectiveSet: 1598/// ::= .equ identifier ',' expression 1599/// ::= .equiv identifier ',' expression 1600/// ::= .set identifier ',' expression 1601bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) { 1602 StringRef Name; 1603 1604 if (ParseIdentifier(Name)) 1605 return TokError("expected identifier after '" + Twine(IDVal) + "'"); 1606 1607 if (getLexer().isNot(AsmToken::Comma)) 1608 return TokError("unexpected token in '" + Twine(IDVal) + "'"); 1609 Lex(); 1610 1611 return ParseAssignment(Name, allow_redef); 1612} 1613 1614bool AsmParser::ParseEscapedString(std::string &Data) { 1615 assert(getLexer().is(AsmToken::String) && "Unexpected current token!"); 1616 1617 Data = ""; 1618 StringRef Str = getTok().getStringContents(); 1619 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 1620 if (Str[i] != '\\') { 1621 Data += Str[i]; 1622 continue; 1623 } 1624 1625 // Recognize escaped characters. Note that this escape semantics currently 1626 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes. 1627 ++i; 1628 if (i == e) 1629 return TokError("unexpected backslash at end of string"); 1630 1631 // Recognize octal sequences. 1632 if ((unsigned) (Str[i] - '0') <= 7) { 1633 // Consume up to three octal characters. 1634 unsigned Value = Str[i] - '0'; 1635 1636 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1637 ++i; 1638 Value = Value * 8 + (Str[i] - '0'); 1639 1640 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) { 1641 ++i; 1642 Value = Value * 8 + (Str[i] - '0'); 1643 } 1644 } 1645 1646 if (Value > 255) 1647 return TokError("invalid octal escape sequence (out of range)"); 1648 1649 Data += (unsigned char) Value; 1650 continue; 1651 } 1652 1653 // Otherwise recognize individual escapes. 1654 switch (Str[i]) { 1655 default: 1656 // Just reject invalid escape sequences for now. 1657 return TokError("invalid escape sequence (unrecognized character)"); 1658 1659 case 'b': Data += '\b'; break; 1660 case 'f': Data += '\f'; break; 1661 case 'n': Data += '\n'; break; 1662 case 'r': Data += '\r'; break; 1663 case 't': Data += '\t'; break; 1664 case '"': Data += '"'; break; 1665 case '\\': Data += '\\'; break; 1666 } 1667 } 1668 1669 return false; 1670} 1671 1672/// ParseDirectiveAscii: 1673/// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] 1674bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { 1675 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1676 CheckForValidSection(); 1677 1678 for (;;) { 1679 if (getLexer().isNot(AsmToken::String)) 1680 return TokError("expected string in '" + Twine(IDVal) + "' directive"); 1681 1682 std::string Data; 1683 if (ParseEscapedString(Data)) 1684 return true; 1685 1686 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE); 1687 if (ZeroTerminated) 1688 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE); 1689 1690 Lex(); 1691 1692 if (getLexer().is(AsmToken::EndOfStatement)) 1693 break; 1694 1695 if (getLexer().isNot(AsmToken::Comma)) 1696 return TokError("unexpected token in '" + Twine(IDVal) + "' directive"); 1697 Lex(); 1698 } 1699 } 1700 1701 Lex(); 1702 return false; 1703} 1704 1705/// ParseDirectiveValue 1706/// ::= (.byte | .short | ... ) [ expression (, expression)* ] 1707bool AsmParser::ParseDirectiveValue(unsigned Size) { 1708 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1709 CheckForValidSection(); 1710 1711 for (;;) { 1712 const MCExpr *Value; 1713 SMLoc ExprLoc = getLexer().getLoc(); 1714 if (ParseExpression(Value)) 1715 return true; 1716 1717 // Special case constant expressions to match code generator. 1718 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 1719 assert(Size <= 8 && "Invalid size"); 1720 uint64_t IntValue = MCE->getValue(); 1721 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue)) 1722 return Error(ExprLoc, "literal value out of range for directive"); 1723 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE); 1724 } else 1725 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE); 1726 1727 if (getLexer().is(AsmToken::EndOfStatement)) 1728 break; 1729 1730 // FIXME: Improve diagnostic. 1731 if (getLexer().isNot(AsmToken::Comma)) 1732 return TokError("unexpected token in directive"); 1733 Lex(); 1734 } 1735 } 1736 1737 Lex(); 1738 return false; 1739} 1740 1741/// ParseDirectiveRealValue 1742/// ::= (.single | .double) [ expression (, expression)* ] 1743bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) { 1744 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1745 CheckForValidSection(); 1746 1747 for (;;) { 1748 // We don't truly support arithmetic on floating point expressions, so we 1749 // have to manually parse unary prefixes. 1750 bool IsNeg = false; 1751 if (getLexer().is(AsmToken::Minus)) { 1752 Lex(); 1753 IsNeg = true; 1754 } else if (getLexer().is(AsmToken::Plus)) 1755 Lex(); 1756 1757 if (getLexer().isNot(AsmToken::Integer) && 1758 getLexer().isNot(AsmToken::Real) && 1759 getLexer().isNot(AsmToken::Identifier)) 1760 return TokError("unexpected token in directive"); 1761 1762 // Convert to an APFloat. 1763 APFloat Value(Semantics); 1764 StringRef IDVal = getTok().getString(); 1765 if (getLexer().is(AsmToken::Identifier)) { 1766 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf")) 1767 Value = APFloat::getInf(Semantics); 1768 else if (!IDVal.compare_lower("nan")) 1769 Value = APFloat::getNaN(Semantics, false, ~0); 1770 else 1771 return TokError("invalid floating point literal"); 1772 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) == 1773 APFloat::opInvalidOp) 1774 return TokError("invalid floating point literal"); 1775 if (IsNeg) 1776 Value.changeSign(); 1777 1778 // Consume the numeric token. 1779 Lex(); 1780 1781 // Emit the value as an integer. 1782 APInt AsInt = Value.bitcastToAPInt(); 1783 getStreamer().EmitIntValue(AsInt.getLimitedValue(), 1784 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE); 1785 1786 if (getLexer().is(AsmToken::EndOfStatement)) 1787 break; 1788 1789 if (getLexer().isNot(AsmToken::Comma)) 1790 return TokError("unexpected token in directive"); 1791 Lex(); 1792 } 1793 } 1794 1795 Lex(); 1796 return false; 1797} 1798 1799/// ParseDirectiveSpace 1800/// ::= .space expression [ , expression ] 1801bool AsmParser::ParseDirectiveSpace() { 1802 CheckForValidSection(); 1803 1804 int64_t NumBytes; 1805 if (ParseAbsoluteExpression(NumBytes)) 1806 return true; 1807 1808 int64_t FillExpr = 0; 1809 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1810 if (getLexer().isNot(AsmToken::Comma)) 1811 return TokError("unexpected token in '.space' directive"); 1812 Lex(); 1813 1814 if (ParseAbsoluteExpression(FillExpr)) 1815 return true; 1816 1817 if (getLexer().isNot(AsmToken::EndOfStatement)) 1818 return TokError("unexpected token in '.space' directive"); 1819 } 1820 1821 Lex(); 1822 1823 if (NumBytes <= 0) 1824 return TokError("invalid number of bytes in '.space' directive"); 1825 1826 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0. 1827 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE); 1828 1829 return false; 1830} 1831 1832/// ParseDirectiveZero 1833/// ::= .zero expression 1834bool AsmParser::ParseDirectiveZero() { 1835 CheckForValidSection(); 1836 1837 int64_t NumBytes; 1838 if (ParseAbsoluteExpression(NumBytes)) 1839 return true; 1840 1841 int64_t Val = 0; 1842 if (getLexer().is(AsmToken::Comma)) { 1843 Lex(); 1844 if (ParseAbsoluteExpression(Val)) 1845 return true; 1846 } 1847 1848 if (getLexer().isNot(AsmToken::EndOfStatement)) 1849 return TokError("unexpected token in '.zero' directive"); 1850 1851 Lex(); 1852 1853 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE); 1854 1855 return false; 1856} 1857 1858/// ParseDirectiveFill 1859/// ::= .fill expression , expression , expression 1860bool AsmParser::ParseDirectiveFill() { 1861 CheckForValidSection(); 1862 1863 int64_t NumValues; 1864 if (ParseAbsoluteExpression(NumValues)) 1865 return true; 1866 1867 if (getLexer().isNot(AsmToken::Comma)) 1868 return TokError("unexpected token in '.fill' directive"); 1869 Lex(); 1870 1871 int64_t FillSize; 1872 if (ParseAbsoluteExpression(FillSize)) 1873 return true; 1874 1875 if (getLexer().isNot(AsmToken::Comma)) 1876 return TokError("unexpected token in '.fill' directive"); 1877 Lex(); 1878 1879 int64_t FillExpr; 1880 if (ParseAbsoluteExpression(FillExpr)) 1881 return true; 1882 1883 if (getLexer().isNot(AsmToken::EndOfStatement)) 1884 return TokError("unexpected token in '.fill' directive"); 1885 1886 Lex(); 1887 1888 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8) 1889 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8"); 1890 1891 for (uint64_t i = 0, e = NumValues; i != e; ++i) 1892 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE); 1893 1894 return false; 1895} 1896 1897/// ParseDirectiveOrg 1898/// ::= .org expression [ , expression ] 1899bool AsmParser::ParseDirectiveOrg() { 1900 CheckForValidSection(); 1901 1902 const MCExpr *Offset; 1903 if (ParseExpression(Offset)) 1904 return true; 1905 1906 // Parse optional fill expression. 1907 int64_t FillExpr = 0; 1908 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1909 if (getLexer().isNot(AsmToken::Comma)) 1910 return TokError("unexpected token in '.org' directive"); 1911 Lex(); 1912 1913 if (ParseAbsoluteExpression(FillExpr)) 1914 return true; 1915 1916 if (getLexer().isNot(AsmToken::EndOfStatement)) 1917 return TokError("unexpected token in '.org' directive"); 1918 } 1919 1920 Lex(); 1921 1922 // FIXME: Only limited forms of relocatable expressions are accepted here, it 1923 // has to be relative to the current section. 1924 getStreamer().EmitValueToOffset(Offset, FillExpr); 1925 1926 return false; 1927} 1928 1929/// ParseDirectiveAlign 1930/// ::= {.align, ...} expression [ , expression [ , expression ]] 1931bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) { 1932 CheckForValidSection(); 1933 1934 SMLoc AlignmentLoc = getLexer().getLoc(); 1935 int64_t Alignment; 1936 if (ParseAbsoluteExpression(Alignment)) 1937 return true; 1938 1939 SMLoc MaxBytesLoc; 1940 bool HasFillExpr = false; 1941 int64_t FillExpr = 0; 1942 int64_t MaxBytesToFill = 0; 1943 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1944 if (getLexer().isNot(AsmToken::Comma)) 1945 return TokError("unexpected token in directive"); 1946 Lex(); 1947 1948 // The fill expression can be omitted while specifying a maximum number of 1949 // alignment bytes, e.g: 1950 // .align 3,,4 1951 if (getLexer().isNot(AsmToken::Comma)) { 1952 HasFillExpr = true; 1953 if (ParseAbsoluteExpression(FillExpr)) 1954 return true; 1955 } 1956 1957 if (getLexer().isNot(AsmToken::EndOfStatement)) { 1958 if (getLexer().isNot(AsmToken::Comma)) 1959 return TokError("unexpected token in directive"); 1960 Lex(); 1961 1962 MaxBytesLoc = getLexer().getLoc(); 1963 if (ParseAbsoluteExpression(MaxBytesToFill)) 1964 return true; 1965 1966 if (getLexer().isNot(AsmToken::EndOfStatement)) 1967 return TokError("unexpected token in directive"); 1968 } 1969 } 1970 1971 Lex(); 1972 1973 if (!HasFillExpr) 1974 FillExpr = 0; 1975 1976 // Compute alignment in bytes. 1977 if (IsPow2) { 1978 // FIXME: Diagnose overflow. 1979 if (Alignment >= 32) { 1980 Error(AlignmentLoc, "invalid alignment value"); 1981 Alignment = 31; 1982 } 1983 1984 Alignment = 1ULL << Alignment; 1985 } 1986 1987 // Diagnose non-sensical max bytes to align. 1988 if (MaxBytesLoc.isValid()) { 1989 if (MaxBytesToFill < 1) { 1990 Error(MaxBytesLoc, "alignment directive can never be satisfied in this " 1991 "many bytes, ignoring maximum bytes expression"); 1992 MaxBytesToFill = 0; 1993 } 1994 1995 if (MaxBytesToFill >= Alignment) { 1996 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and " 1997 "has no effect"); 1998 MaxBytesToFill = 0; 1999 } 2000 } 2001 2002 // Check whether we should use optimal code alignment for this .align 2003 // directive. 2004 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign(); 2005 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) && 2006 ValueSize == 1 && UseCodeAlign) { 2007 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill); 2008 } else { 2009 // FIXME: Target specific behavior about how the "extra" bytes are filled. 2010 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize, 2011 MaxBytesToFill); 2012 } 2013 2014 return false; 2015} 2016 2017/// ParseDirectiveSymbolAttribute 2018/// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] 2019bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) { 2020 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2021 for (;;) { 2022 StringRef Name; 2023 SMLoc Loc = getTok().getLoc(); 2024 2025 if (ParseIdentifier(Name)) 2026 return Error(Loc, "expected identifier in directive"); 2027 2028 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2029 2030 // Assembler local symbols don't make any sense here. Complain loudly. 2031 if (Sym->isTemporary()) 2032 return Error(Loc, "non-local symbol required in directive"); 2033 2034 getStreamer().EmitSymbolAttribute(Sym, Attr); 2035 2036 if (getLexer().is(AsmToken::EndOfStatement)) 2037 break; 2038 2039 if (getLexer().isNot(AsmToken::Comma)) 2040 return TokError("unexpected token in directive"); 2041 Lex(); 2042 } 2043 } 2044 2045 Lex(); 2046 return false; 2047} 2048 2049/// ParseDirectiveComm 2050/// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] 2051bool AsmParser::ParseDirectiveComm(bool IsLocal) { 2052 CheckForValidSection(); 2053 2054 SMLoc IDLoc = getLexer().getLoc(); 2055 StringRef Name; 2056 if (ParseIdentifier(Name)) 2057 return TokError("expected identifier in directive"); 2058 2059 // Handle the identifier as the key symbol. 2060 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2061 2062 if (getLexer().isNot(AsmToken::Comma)) 2063 return TokError("unexpected token in directive"); 2064 Lex(); 2065 2066 int64_t Size; 2067 SMLoc SizeLoc = getLexer().getLoc(); 2068 if (ParseAbsoluteExpression(Size)) 2069 return true; 2070 2071 int64_t Pow2Alignment = 0; 2072 SMLoc Pow2AlignmentLoc; 2073 if (getLexer().is(AsmToken::Comma)) { 2074 Lex(); 2075 Pow2AlignmentLoc = getLexer().getLoc(); 2076 if (ParseAbsoluteExpression(Pow2Alignment)) 2077 return true; 2078 2079 // If this target takes alignments in bytes (not log) validate and convert. 2080 if (Lexer.getMAI().getAlignmentIsInBytes()) { 2081 if (!isPowerOf2_64(Pow2Alignment)) 2082 return Error(Pow2AlignmentLoc, "alignment must be a power of 2"); 2083 Pow2Alignment = Log2_64(Pow2Alignment); 2084 } 2085 } 2086 2087 if (getLexer().isNot(AsmToken::EndOfStatement)) 2088 return TokError("unexpected token in '.comm' or '.lcomm' directive"); 2089 2090 Lex(); 2091 2092 // NOTE: a size of zero for a .comm should create a undefined symbol 2093 // but a size of .lcomm creates a bss symbol of size zero. 2094 if (Size < 0) 2095 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't " 2096 "be less than zero"); 2097 2098 // NOTE: The alignment in the directive is a power of 2 value, the assembler 2099 // may internally end up wanting an alignment in bytes. 2100 // FIXME: Diagnose overflow. 2101 if (Pow2Alignment < 0) 2102 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive " 2103 "alignment, can't be less than zero"); 2104 2105 if (!Sym->isUndefined()) 2106 return Error(IDLoc, "invalid symbol redefinition"); 2107 2108 // '.lcomm' is equivalent to '.zerofill'. 2109 // Create the Symbol as a common or local common with Size and Pow2Alignment 2110 if (IsLocal) { 2111 getStreamer().EmitZerofill(Ctx.getMachOSection( 2112 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL, 2113 0, SectionKind::getBSS()), 2114 Sym, Size, 1 << Pow2Alignment); 2115 return false; 2116 } 2117 2118 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment); 2119 return false; 2120} 2121 2122/// ParseDirectiveAbort 2123/// ::= .abort [... message ...] 2124bool AsmParser::ParseDirectiveAbort() { 2125 // FIXME: Use loc from directive. 2126 SMLoc Loc = getLexer().getLoc(); 2127 2128 StringRef Str = ParseStringToEndOfStatement(); 2129 if (getLexer().isNot(AsmToken::EndOfStatement)) 2130 return TokError("unexpected token in '.abort' directive"); 2131 2132 Lex(); 2133 2134 if (Str.empty()) 2135 Error(Loc, ".abort detected. Assembly stopping."); 2136 else 2137 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping."); 2138 // FIXME: Actually abort assembly here. 2139 2140 return false; 2141} 2142 2143/// ParseDirectiveInclude 2144/// ::= .include "filename" 2145bool AsmParser::ParseDirectiveInclude() { 2146 if (getLexer().isNot(AsmToken::String)) 2147 return TokError("expected string in '.include' directive"); 2148 2149 std::string Filename = getTok().getString(); 2150 SMLoc IncludeLoc = getLexer().getLoc(); 2151 Lex(); 2152 2153 if (getLexer().isNot(AsmToken::EndOfStatement)) 2154 return TokError("unexpected token in '.include' directive"); 2155 2156 // Strip the quotes. 2157 Filename = Filename.substr(1, Filename.size()-2); 2158 2159 // Attempt to switch the lexer to the included file before consuming the end 2160 // of statement to avoid losing it when we switch. 2161 if (EnterIncludeFile(Filename)) { 2162 Error(IncludeLoc, "Could not find include file '" + Filename + "'"); 2163 return true; 2164 } 2165 2166 return false; 2167} 2168 2169/// ParseDirectiveIf 2170/// ::= .if expression 2171bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) { 2172 TheCondStack.push_back(TheCondState); 2173 TheCondState.TheCond = AsmCond::IfCond; 2174 if(TheCondState.Ignore) { 2175 EatToEndOfStatement(); 2176 } 2177 else { 2178 int64_t ExprValue; 2179 if (ParseAbsoluteExpression(ExprValue)) 2180 return true; 2181 2182 if (getLexer().isNot(AsmToken::EndOfStatement)) 2183 return TokError("unexpected token in '.if' directive"); 2184 2185 Lex(); 2186 2187 TheCondState.CondMet = ExprValue; 2188 TheCondState.Ignore = !TheCondState.CondMet; 2189 } 2190 2191 return false; 2192} 2193 2194bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { 2195 StringRef Name; 2196 TheCondStack.push_back(TheCondState); 2197 TheCondState.TheCond = AsmCond::IfCond; 2198 2199 if (TheCondState.Ignore) { 2200 EatToEndOfStatement(); 2201 } else { 2202 if (ParseIdentifier(Name)) 2203 return TokError("expected identifier after '.ifdef'"); 2204 2205 Lex(); 2206 2207 MCSymbol *Sym = getContext().LookupSymbol(Name); 2208 2209 if (expect_defined) 2210 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined()); 2211 else 2212 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined()); 2213 TheCondState.Ignore = !TheCondState.CondMet; 2214 } 2215 2216 return false; 2217} 2218 2219/// ParseDirectiveElseIf 2220/// ::= .elseif expression 2221bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) { 2222 if (TheCondState.TheCond != AsmCond::IfCond && 2223 TheCondState.TheCond != AsmCond::ElseIfCond) 2224 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or " 2225 " an .elseif"); 2226 TheCondState.TheCond = AsmCond::ElseIfCond; 2227 2228 bool LastIgnoreState = false; 2229 if (!TheCondStack.empty()) 2230 LastIgnoreState = TheCondStack.back().Ignore; 2231 if (LastIgnoreState || TheCondState.CondMet) { 2232 TheCondState.Ignore = true; 2233 EatToEndOfStatement(); 2234 } 2235 else { 2236 int64_t ExprValue; 2237 if (ParseAbsoluteExpression(ExprValue)) 2238 return true; 2239 2240 if (getLexer().isNot(AsmToken::EndOfStatement)) 2241 return TokError("unexpected token in '.elseif' directive"); 2242 2243 Lex(); 2244 TheCondState.CondMet = ExprValue; 2245 TheCondState.Ignore = !TheCondState.CondMet; 2246 } 2247 2248 return false; 2249} 2250 2251/// ParseDirectiveElse 2252/// ::= .else 2253bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) { 2254 if (getLexer().isNot(AsmToken::EndOfStatement)) 2255 return TokError("unexpected token in '.else' directive"); 2256 2257 Lex(); 2258 2259 if (TheCondState.TheCond != AsmCond::IfCond && 2260 TheCondState.TheCond != AsmCond::ElseIfCond) 2261 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an " 2262 ".elseif"); 2263 TheCondState.TheCond = AsmCond::ElseCond; 2264 bool LastIgnoreState = false; 2265 if (!TheCondStack.empty()) 2266 LastIgnoreState = TheCondStack.back().Ignore; 2267 if (LastIgnoreState || TheCondState.CondMet) 2268 TheCondState.Ignore = true; 2269 else 2270 TheCondState.Ignore = false; 2271 2272 return false; 2273} 2274 2275/// ParseDirectiveEndIf 2276/// ::= .endif 2277bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) { 2278 if (getLexer().isNot(AsmToken::EndOfStatement)) 2279 return TokError("unexpected token in '.endif' directive"); 2280 2281 Lex(); 2282 2283 if ((TheCondState.TheCond == AsmCond::NoCond) || 2284 TheCondStack.empty()) 2285 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or " 2286 ".else"); 2287 if (!TheCondStack.empty()) { 2288 TheCondState = TheCondStack.back(); 2289 TheCondStack.pop_back(); 2290 } 2291 2292 return false; 2293} 2294 2295/// ParseDirectiveFile 2296/// ::= .file [number] string 2297bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) { 2298 // FIXME: I'm not sure what this is. 2299 int64_t FileNumber = -1; 2300 SMLoc FileNumberLoc = getLexer().getLoc(); 2301 if (getLexer().is(AsmToken::Integer)) { 2302 FileNumber = getTok().getIntVal(); 2303 Lex(); 2304 2305 if (FileNumber < 1) 2306 return TokError("file number less than one"); 2307 } 2308 2309 if (getLexer().isNot(AsmToken::String)) 2310 return TokError("unexpected token in '.file' directive"); 2311 2312 StringRef Filename = getTok().getString(); 2313 Filename = Filename.substr(1, Filename.size()-2); 2314 Lex(); 2315 2316 if (getLexer().isNot(AsmToken::EndOfStatement)) 2317 return TokError("unexpected token in '.file' directive"); 2318 2319 if (FileNumber == -1) 2320 getStreamer().EmitFileDirective(Filename); 2321 else { 2322 if (getStreamer().EmitDwarfFileDirective(FileNumber, Filename)) 2323 Error(FileNumberLoc, "file number already allocated"); 2324 } 2325 2326 return false; 2327} 2328 2329/// ParseDirectiveLine 2330/// ::= .line [number] 2331bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) { 2332 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2333 if (getLexer().isNot(AsmToken::Integer)) 2334 return TokError("unexpected token in '.line' directive"); 2335 2336 int64_t LineNumber = getTok().getIntVal(); 2337 (void) LineNumber; 2338 Lex(); 2339 2340 // FIXME: Do something with the .line. 2341 } 2342 2343 if (getLexer().isNot(AsmToken::EndOfStatement)) 2344 return TokError("unexpected token in '.line' directive"); 2345 2346 return false; 2347} 2348 2349 2350/// ParseDirectiveLoc 2351/// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] 2352/// [epilogue_begin] [is_stmt VALUE] [isa VALUE] 2353/// The first number is a file number, must have been previously assigned with 2354/// a .file directive, the second number is the line number and optionally the 2355/// third number is a column position (zero if not specified). The remaining 2356/// optional items are .loc sub-directives. 2357bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) { 2358 2359 if (getLexer().isNot(AsmToken::Integer)) 2360 return TokError("unexpected token in '.loc' directive"); 2361 int64_t FileNumber = getTok().getIntVal(); 2362 if (FileNumber < 1) 2363 return TokError("file number less than one in '.loc' directive"); 2364 if (!getContext().isValidDwarfFileNumber(FileNumber)) 2365 return TokError("unassigned file number in '.loc' directive"); 2366 Lex(); 2367 2368 int64_t LineNumber = 0; 2369 if (getLexer().is(AsmToken::Integer)) { 2370 LineNumber = getTok().getIntVal(); 2371 if (LineNumber < 1) 2372 return TokError("line number less than one in '.loc' directive"); 2373 Lex(); 2374 } 2375 2376 int64_t ColumnPos = 0; 2377 if (getLexer().is(AsmToken::Integer)) { 2378 ColumnPos = getTok().getIntVal(); 2379 if (ColumnPos < 0) 2380 return TokError("column position less than zero in '.loc' directive"); 2381 Lex(); 2382 } 2383 2384 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0; 2385 unsigned Isa = 0; 2386 int64_t Discriminator = 0; 2387 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2388 for (;;) { 2389 if (getLexer().is(AsmToken::EndOfStatement)) 2390 break; 2391 2392 StringRef Name; 2393 SMLoc Loc = getTok().getLoc(); 2394 if (getParser().ParseIdentifier(Name)) 2395 return TokError("unexpected token in '.loc' directive"); 2396 2397 if (Name == "basic_block") 2398 Flags |= DWARF2_FLAG_BASIC_BLOCK; 2399 else if (Name == "prologue_end") 2400 Flags |= DWARF2_FLAG_PROLOGUE_END; 2401 else if (Name == "epilogue_begin") 2402 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; 2403 else if (Name == "is_stmt") { 2404 SMLoc Loc = getTok().getLoc(); 2405 const MCExpr *Value; 2406 if (getParser().ParseExpression(Value)) 2407 return true; 2408 // The expression must be the constant 0 or 1. 2409 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2410 int Value = MCE->getValue(); 2411 if (Value == 0) 2412 Flags &= ~DWARF2_FLAG_IS_STMT; 2413 else if (Value == 1) 2414 Flags |= DWARF2_FLAG_IS_STMT; 2415 else 2416 return Error(Loc, "is_stmt value not 0 or 1"); 2417 } 2418 else { 2419 return Error(Loc, "is_stmt value not the constant value of 0 or 1"); 2420 } 2421 } 2422 else if (Name == "isa") { 2423 SMLoc Loc = getTok().getLoc(); 2424 const MCExpr *Value; 2425 if (getParser().ParseExpression(Value)) 2426 return true; 2427 // The expression must be a constant greater or equal to 0. 2428 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) { 2429 int Value = MCE->getValue(); 2430 if (Value < 0) 2431 return Error(Loc, "isa number less than zero"); 2432 Isa = Value; 2433 } 2434 else { 2435 return Error(Loc, "isa number not a constant value"); 2436 } 2437 } 2438 else if (Name == "discriminator") { 2439 if (getParser().ParseAbsoluteExpression(Discriminator)) 2440 return true; 2441 } 2442 else { 2443 return Error(Loc, "unknown sub-directive in '.loc' directive"); 2444 } 2445 2446 if (getLexer().is(AsmToken::EndOfStatement)) 2447 break; 2448 } 2449 } 2450 2451 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags, 2452 Isa, Discriminator, StringRef()); 2453 2454 return false; 2455} 2456 2457/// ParseDirectiveStabs 2458/// ::= .stabs string, number, number, number 2459bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive, 2460 SMLoc DirectiveLoc) { 2461 return TokError("unsupported directive '" + Directive + "'"); 2462} 2463 2464/// ParseDirectiveCFISections 2465/// ::= .cfi_sections section [, section] 2466bool GenericAsmParser::ParseDirectiveCFISections(StringRef, 2467 SMLoc DirectiveLoc) { 2468 StringRef Name; 2469 bool EH = false; 2470 bool Debug = false; 2471 2472 if (getParser().ParseIdentifier(Name)) 2473 return TokError("Expected an identifier"); 2474 2475 if (Name == ".eh_frame") 2476 EH = true; 2477 else if (Name == ".debug_frame") 2478 Debug = true; 2479 2480 if (getLexer().is(AsmToken::Comma)) { 2481 Lex(); 2482 2483 if (getParser().ParseIdentifier(Name)) 2484 return TokError("Expected an identifier"); 2485 2486 if (Name == ".eh_frame") 2487 EH = true; 2488 else if (Name == ".debug_frame") 2489 Debug = true; 2490 } 2491 2492 getStreamer().EmitCFISections(EH, Debug); 2493 2494 return false; 2495} 2496 2497/// ParseDirectiveCFIStartProc 2498/// ::= .cfi_startproc 2499bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef, 2500 SMLoc DirectiveLoc) { 2501 getStreamer().EmitCFIStartProc(); 2502 return false; 2503} 2504 2505/// ParseDirectiveCFIEndProc 2506/// ::= .cfi_endproc 2507bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) { 2508 getStreamer().EmitCFIEndProc(); 2509 return false; 2510} 2511 2512/// ParseRegisterOrRegisterNumber - parse register name or number. 2513bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register, 2514 SMLoc DirectiveLoc) { 2515 unsigned RegNo; 2516 2517 if (getLexer().isNot(AsmToken::Integer)) { 2518 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc, 2519 DirectiveLoc)) 2520 return true; 2521 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true); 2522 } else 2523 return getParser().ParseAbsoluteExpression(Register); 2524 2525 return false; 2526} 2527 2528/// ParseDirectiveCFIDefCfa 2529/// ::= .cfi_def_cfa register, offset 2530bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef, 2531 SMLoc DirectiveLoc) { 2532 int64_t Register = 0; 2533 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2534 return true; 2535 2536 if (getLexer().isNot(AsmToken::Comma)) 2537 return TokError("unexpected token in directive"); 2538 Lex(); 2539 2540 int64_t Offset = 0; 2541 if (getParser().ParseAbsoluteExpression(Offset)) 2542 return true; 2543 2544 getStreamer().EmitCFIDefCfa(Register, Offset); 2545 return false; 2546} 2547 2548/// ParseDirectiveCFIDefCfaOffset 2549/// ::= .cfi_def_cfa_offset offset 2550bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef, 2551 SMLoc DirectiveLoc) { 2552 int64_t Offset = 0; 2553 if (getParser().ParseAbsoluteExpression(Offset)) 2554 return true; 2555 2556 getStreamer().EmitCFIDefCfaOffset(Offset); 2557 return false; 2558} 2559 2560/// ParseDirectiveCFIAdjustCfaOffset 2561/// ::= .cfi_adjust_cfa_offset adjustment 2562bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef, 2563 SMLoc DirectiveLoc) { 2564 int64_t Adjustment = 0; 2565 if (getParser().ParseAbsoluteExpression(Adjustment)) 2566 return true; 2567 2568 getStreamer().EmitCFIAdjustCfaOffset(Adjustment); 2569 return false; 2570} 2571 2572/// ParseDirectiveCFIDefCfaRegister 2573/// ::= .cfi_def_cfa_register register 2574bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef, 2575 SMLoc DirectiveLoc) { 2576 int64_t Register = 0; 2577 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2578 return true; 2579 2580 getStreamer().EmitCFIDefCfaRegister(Register); 2581 return false; 2582} 2583 2584/// ParseDirectiveCFIOffset 2585/// ::= .cfi_offset register, offset 2586bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) { 2587 int64_t Register = 0; 2588 int64_t Offset = 0; 2589 2590 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2591 return true; 2592 2593 if (getLexer().isNot(AsmToken::Comma)) 2594 return TokError("unexpected token in directive"); 2595 Lex(); 2596 2597 if (getParser().ParseAbsoluteExpression(Offset)) 2598 return true; 2599 2600 getStreamer().EmitCFIOffset(Register, Offset); 2601 return false; 2602} 2603 2604/// ParseDirectiveCFIRelOffset 2605/// ::= .cfi_rel_offset register, offset 2606bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef, 2607 SMLoc DirectiveLoc) { 2608 int64_t Register = 0; 2609 2610 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2611 return true; 2612 2613 if (getLexer().isNot(AsmToken::Comma)) 2614 return TokError("unexpected token in directive"); 2615 Lex(); 2616 2617 int64_t Offset = 0; 2618 if (getParser().ParseAbsoluteExpression(Offset)) 2619 return true; 2620 2621 getStreamer().EmitCFIRelOffset(Register, Offset); 2622 return false; 2623} 2624 2625static bool isValidEncoding(int64_t Encoding) { 2626 if (Encoding & ~0xff) 2627 return false; 2628 2629 if (Encoding == dwarf::DW_EH_PE_omit) 2630 return true; 2631 2632 const unsigned Format = Encoding & 0xf; 2633 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && 2634 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && 2635 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && 2636 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) 2637 return false; 2638 2639 const unsigned Application = Encoding & 0x70; 2640 if (Application != dwarf::DW_EH_PE_absptr && 2641 Application != dwarf::DW_EH_PE_pcrel) 2642 return false; 2643 2644 return true; 2645} 2646 2647/// ParseDirectiveCFIPersonalityOrLsda 2648/// ::= .cfi_personality encoding, [symbol_name] 2649/// ::= .cfi_lsda encoding, [symbol_name] 2650bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal, 2651 SMLoc DirectiveLoc) { 2652 int64_t Encoding = 0; 2653 if (getParser().ParseAbsoluteExpression(Encoding)) 2654 return true; 2655 if (Encoding == dwarf::DW_EH_PE_omit) 2656 return false; 2657 2658 if (!isValidEncoding(Encoding)) 2659 return TokError("unsupported encoding."); 2660 2661 if (getLexer().isNot(AsmToken::Comma)) 2662 return TokError("unexpected token in directive"); 2663 Lex(); 2664 2665 StringRef Name; 2666 if (getParser().ParseIdentifier(Name)) 2667 return TokError("expected identifier in directive"); 2668 2669 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name); 2670 2671 if (IDVal == ".cfi_personality") 2672 getStreamer().EmitCFIPersonality(Sym, Encoding); 2673 else { 2674 assert(IDVal == ".cfi_lsda"); 2675 getStreamer().EmitCFILsda(Sym, Encoding); 2676 } 2677 return false; 2678} 2679 2680/// ParseDirectiveCFIRememberState 2681/// ::= .cfi_remember_state 2682bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal, 2683 SMLoc DirectiveLoc) { 2684 getStreamer().EmitCFIRememberState(); 2685 return false; 2686} 2687 2688/// ParseDirectiveCFIRestoreState 2689/// ::= .cfi_remember_state 2690bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal, 2691 SMLoc DirectiveLoc) { 2692 getStreamer().EmitCFIRestoreState(); 2693 return false; 2694} 2695 2696/// ParseDirectiveCFISameValue 2697/// ::= .cfi_same_value register 2698bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal, 2699 SMLoc DirectiveLoc) { 2700 int64_t Register = 0; 2701 2702 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc)) 2703 return true; 2704 2705 getStreamer().EmitCFISameValue(Register); 2706 2707 return false; 2708} 2709 2710/// ParseDirectiveMacrosOnOff 2711/// ::= .macros_on 2712/// ::= .macros_off 2713bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive, 2714 SMLoc DirectiveLoc) { 2715 if (getLexer().isNot(AsmToken::EndOfStatement)) 2716 return Error(getLexer().getLoc(), 2717 "unexpected token in '" + Directive + "' directive"); 2718 2719 getParser().MacrosEnabled = Directive == ".macros_on"; 2720 2721 return false; 2722} 2723 2724/// ParseDirectiveMacro 2725/// ::= .macro name [parameters] 2726bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive, 2727 SMLoc DirectiveLoc) { 2728 StringRef Name; 2729 if (getParser().ParseIdentifier(Name)) 2730 return TokError("expected identifier in directive"); 2731 2732 std::vector<StringRef> Parameters; 2733 if (getLexer().isNot(AsmToken::EndOfStatement)) { 2734 for(;;) { 2735 StringRef Parameter; 2736 if (getParser().ParseIdentifier(Parameter)) 2737 return TokError("expected identifier in directive"); 2738 Parameters.push_back(Parameter); 2739 2740 if (getLexer().isNot(AsmToken::Comma)) 2741 break; 2742 Lex(); 2743 } 2744 } 2745 2746 if (getLexer().isNot(AsmToken::EndOfStatement)) 2747 return TokError("unexpected token in '.macro' directive"); 2748 2749 // Eat the end of statement. 2750 Lex(); 2751 2752 AsmToken EndToken, StartToken = getTok(); 2753 2754 // Lex the macro definition. 2755 for (;;) { 2756 // Check whether we have reached the end of the file. 2757 if (getLexer().is(AsmToken::Eof)) 2758 return Error(DirectiveLoc, "no matching '.endmacro' in definition"); 2759 2760 // Otherwise, check whether we have reach the .endmacro. 2761 if (getLexer().is(AsmToken::Identifier) && 2762 (getTok().getIdentifier() == ".endm" || 2763 getTok().getIdentifier() == ".endmacro")) { 2764 EndToken = getTok(); 2765 Lex(); 2766 if (getLexer().isNot(AsmToken::EndOfStatement)) 2767 return TokError("unexpected token in '" + EndToken.getIdentifier() + 2768 "' directive"); 2769 break; 2770 } 2771 2772 // Otherwise, scan til the end of the statement. 2773 getParser().EatToEndOfStatement(); 2774 } 2775 2776 if (getParser().MacroMap.lookup(Name)) { 2777 return Error(DirectiveLoc, "macro '" + Name + "' is already defined"); 2778 } 2779 2780 const char *BodyStart = StartToken.getLoc().getPointer(); 2781 const char *BodyEnd = EndToken.getLoc().getPointer(); 2782 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); 2783 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters); 2784 return false; 2785} 2786 2787/// ParseDirectiveEndMacro 2788/// ::= .endm 2789/// ::= .endmacro 2790bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive, 2791 SMLoc DirectiveLoc) { 2792 if (getLexer().isNot(AsmToken::EndOfStatement)) 2793 return TokError("unexpected token in '" + Directive + "' directive"); 2794 2795 // If we are inside a macro instantiation, terminate the current 2796 // instantiation. 2797 if (!getParser().ActiveMacros.empty()) { 2798 getParser().HandleMacroExit(); 2799 return false; 2800 } 2801 2802 // Otherwise, this .endmacro is a stray entry in the file; well formed 2803 // .endmacro directives are handled during the macro definition parsing. 2804 return TokError("unexpected '" + Directive + "' in file, " 2805 "no current macro definition"); 2806} 2807 2808bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) { 2809 getParser().CheckForValidSection(); 2810 2811 const MCExpr *Value; 2812 2813 if (getParser().ParseExpression(Value)) 2814 return true; 2815 2816 if (getLexer().isNot(AsmToken::EndOfStatement)) 2817 return TokError("unexpected token in directive"); 2818 2819 if (DirName[1] == 's') 2820 getStreamer().EmitSLEB128Value(Value); 2821 else 2822 getStreamer().EmitULEB128Value(Value); 2823 2824 return false; 2825} 2826 2827 2828/// \brief Create an MCAsmParser instance. 2829MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM, 2830 MCContext &C, MCStreamer &Out, 2831 const MCAsmInfo &MAI) { 2832 return new AsmParser(SM, C, Out, MAI); 2833} 2834