1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This class implements the parser for assembly files.
12 //===----------------------------------------------------------------------===//
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"
42 FatalAssemblerWarnings("fatal-assembler-warnings",
43 cl::desc("Consider warnings as error"));
47 /// \brief Helper class for tracking macro definitions.
51 std::vector<StringRef> Parameters;
54 Macro(StringRef N, StringRef B, const std::vector<StringRef> &P) :
55 Name(N), Body(B), Parameters(P) {}
58 /// \brief Helper class for storing information about an active macro
60 struct MacroInstantiation {
61 /// The macro being instantiated.
62 const Macro *TheMacro;
64 /// The macro instantiation with substitutions.
65 MemoryBuffer *Instantiation;
67 /// The location of the instantiation.
68 SMLoc InstantiationLoc;
70 /// The location where parsing should resume upon instantiation completion.
74 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
78 /// \brief The concrete assembly parser instance.
79 class AsmParser : public MCAsmParser {
80 friend class GenericAsmParser;
82 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
83 void operator=(const AsmParser &); // DO NOT IMPLEMENT
90 MCAsmParserExtension *GenericParser;
91 MCAsmParserExtension *PlatformParser;
93 /// This is the current buffer index we're lexing from as managed by the
98 std::vector<AsmCond> TheCondStack;
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;
106 /// MacroMap - Map of currently defined macros.
107 StringMap<Macro*> MacroMap;
109 /// ActiveMacros - Stack of active macro instantiations.
110 std::vector<MacroInstantiation*> ActiveMacros;
112 /// Boolean tracking whether macro substitution is enabled.
113 unsigned MacrosEnabled : 1;
115 /// Flag tracking whether any errors have been encountered.
116 unsigned HadError : 1;
118 /// The values from the last parsed cpp hash file line comment if any.
119 StringRef CppHashFilename;
120 int64_t CppHashLineNumber;
124 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
125 const MCAsmInfo &MAI);
128 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
130 void AddDirectiveHandler(MCAsmParserExtension *Object,
132 DirectiveHandler Handler) {
133 DirectiveMap[Directive] = std::make_pair(Object, Handler);
137 /// @name MCAsmParser Interface
140 virtual SourceMgr &getSourceManager() { return SrcMgr; }
141 virtual MCAsmLexer &getLexer() { return Lexer; }
142 virtual MCContext &getContext() { return Ctx; }
143 virtual MCStreamer &getStreamer() { return Out; }
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>());
150 const AsmToken &Lex();
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);
160 void CheckForValidSection();
162 bool ParseStatement();
163 void EatToEndOfLine();
164 bool ParseCppHashLineFilenameComment(const SMLoc &L);
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,
171 void HandleMacroExit();
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);
179 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
181 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
182 bool EnterIncludeFile(const std::string &Filename);
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
187 void JumpToLoc(SMLoc Loc);
189 void EatToEndOfStatement();
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();
196 bool ParseAssignment(StringRef Name, bool allow_redef);
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);
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);
207 // Directive Parsing.
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);
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);
225 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
227 bool ParseDirectiveAbort(); // ".abort"
228 bool ParseDirectiveInclude(); // ".include"
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
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);
241 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
242 MCSymbolRefExpr::VariantKind Variant);
245 /// \brief Generic implementations of directive handling, etc. which is shared
246 /// (or the default, at least) for all assembler parser.
247 class GenericAsmParser : public MCAsmParserExtension {
248 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
249 void AddDirectiveHandler(StringRef Directive) {
250 getParser().AddDirectiveHandler(this, Directive,
251 HandleDirective<GenericAsmParser, Handler>);
254 GenericAsmParser() {}
256 AsmParser &getParser() {
257 return (AsmParser&) this->MCAsmParserExtension::getParser();
260 virtual void Initialize(MCAsmParser &Parser) {
261 // Call the base implementation.
262 this->MCAsmParserExtension::Initialize(Parser);
264 // Debugging directives.
265 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
266 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
267 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
268 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
271 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>(
273 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>(
275 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>(
277 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>(
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>(
287 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>(
290 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality");
292 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda");
294 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state");
296 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state");
298 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value");
301 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
303 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
305 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
306 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
307 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
309 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
310 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
313 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
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);
333 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
334 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
335 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
337 bool ParseDirectiveLEB128(StringRef, SMLoc);
344 extern MCAsmParserExtension *createDarwinAsmParser();
345 extern MCAsmParserExtension *createELFAsmParser();
346 extern MCAsmParserExtension *createCOFFAsmParser();
350 enum { DEFAULT_ADDRSPACE = 0 };
352 AsmParser::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));
360 // Initialize the generic parser.
361 GenericParser->Initialize(*this);
363 // Initialize the platform / file format parser.
365 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
367 if (_MAI.hasMicrosoftFastStdCallMangling()) {
368 PlatformParser = createCOFFAsmParser();
369 PlatformParser->Initialize(*this);
370 } else if (_MAI.hasSubsectionsViaSymbols()) {
371 PlatformParser = createDarwinAsmParser();
372 PlatformParser->Initialize(*this);
374 PlatformParser = createELFAsmParser();
375 PlatformParser->Initialize(*this);
379 AsmParser::~AsmParser() {
380 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
382 // Destroy any macros.
383 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
384 ie = MacroMap.end(); it != ie; ++it)
385 delete it->getValue();
387 delete PlatformParser;
388 delete GenericParser;
391 void 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");
399 bool 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();
407 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
409 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges);
410 PrintMacroInstantiations();
414 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
415 std::string IncludedFile;
416 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
422 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
427 void AsmParser::JumpToLoc(SMLoc Loc) {
428 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
429 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
432 const AsmToken &AsmParser::Lex() {
433 const AsmToken *tok = &Lexer.Lex();
435 if (tok->is(AsmToken::Eof)) {
436 // If this is the end of an included file, pop the parent file off the
438 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
439 if (ParentIncludeLoc != SMLoc()) {
440 JumpToLoc(ParentIncludeLoc);
445 if (tok->is(AsmToken::Error))
446 Error(Lexer.getErrLoc(), Lexer.getErr());
451 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
452 // Create the initial section, if requested.
453 if (!NoInitialTextSection)
460 AsmCond StartingCondState = TheCondState;
462 // While we have input, parse each statement.
463 while (Lexer.isNot(AsmToken::Eof)) {
464 if (!ParseStatement()) continue;
466 // We had an error, validate that one was emitted and recover by skipping to
468 assert(HadError && "Parse statement returned an error, but none emitted!");
469 EatToEndOfStatement();
472 if (TheCondState.TheCond != StartingCondState.TheCond ||
473 TheCondState.Ignore != StartingCondState.Ignore)
474 return TokError("unmatched .ifs or .elses");
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");
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(),
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() +
508 // Finalize the output stream if there are no errors and if the client wants
510 if (!HadError && !NoFinalize)
516 void AsmParser::CheckForValidSection() {
517 if (!getStreamer().getCurrentSection()) {
518 TokError("expected section directive before assembly directive");
519 Out.SwitchSection(Ctx.getMachOSection(
521 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
522 0, SectionKind::getText()));
526 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
527 void AsmParser::EatToEndOfStatement() {
528 while (Lexer.isNot(AsmToken::EndOfStatement) &&
529 Lexer.isNot(AsmToken::Eof))
533 if (Lexer.is(AsmToken::EndOfStatement))
537 StringRef AsmParser::ParseStringToEndOfStatement() {
538 const char *Start = getTok().getLoc().getPointer();
540 while (Lexer.isNot(AsmToken::EndOfStatement) &&
541 Lexer.isNot(AsmToken::Eof))
544 const char *End = getTok().getLoc().getPointer();
545 return StringRef(Start, End - Start);
548 /// ParseParenExpr - Parse a paren expression and return it.
549 /// NOTE: This assumes the leading '(' has already been consumed.
551 /// parenexpr ::= expr)
553 bool 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();
562 /// ParseBracketExpr - Parse a bracket expression and return it.
563 /// NOTE: This assumes the leading '[' has already been consumed.
565 /// bracketexpr ::= expr]
567 bool 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();
576 /// ParsePrimaryExpr - Parse a primary expression and return it.
577 /// primaryexpr ::= (parenexpr
578 /// primaryexpr ::= symbol
579 /// primaryexpr ::= number
580 /// primaryexpr ::= '.'
581 /// primaryexpr ::= ~,+,- primaryexpr
582 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
583 switch (Lexer.getKind()) {
585 return TokError("unknown token in expression");
586 // If we have an error assume that we've already handled it.
587 case AsmToken::Error:
589 case AsmToken::Exclaim:
590 Lex(); // Eat the operator.
591 if (ParsePrimaryExpr(Res, EndLoc))
593 Res = MCUnaryExpr::CreateLNot(Res, getContext());
595 case AsmToken::Dollar:
596 case AsmToken::String:
597 case AsmToken::Identifier: {
598 EndLoc = Lexer.getLoc();
600 StringRef Identifier;
601 if (ParseIdentifier(Identifier))
604 // This is a symbol reference.
605 std::pair<StringRef, StringRef> Split = Identifier.split('@');
606 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
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 + "'");
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())) {
622 return Error(EndLoc, "unexpected modifier on variable reference");
624 Res = Sym->getVariableValue();
628 // Otherwise create a symbol ref.
629 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
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();
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,
646 if(IDVal == "b" && Sym->isUndefined())
647 return Error(Loc, "invalid reference to undefined symbol");
648 EndLoc = Lexer.getLoc();
649 Lex(); // Eat identifier.
654 case AsmToken::Real: {
655 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
656 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
657 Res = MCConstantExpr::Create(IntVal, getContext());
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();
666 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
667 EndLoc = Lexer.getLoc();
668 Lex(); // Eat identifier.
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))
683 Res = MCUnaryExpr::CreateMinus(Res, getContext());
686 Lex(); // Eat the operator.
687 if (ParsePrimaryExpr(Res, EndLoc))
689 Res = MCUnaryExpr::CreatePlus(Res, getContext());
691 case AsmToken::Tilde:
692 Lex(); // Eat the operator.
693 if (ParsePrimaryExpr(Res, EndLoc))
695 Res = MCUnaryExpr::CreateNot(Res, getContext());
700 bool AsmParser::ParseExpression(const MCExpr *&Res) {
702 return ParseExpression(Res, EndLoc);
706 AsmParser::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()) {
712 case MCExpr::Constant:
715 case MCExpr::SymbolRef: {
716 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
718 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
719 TokError("invalid variant on expression '" +
720 getTok().getIdentifier() + "' (already modified)");
724 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
727 case MCExpr::Unary: {
728 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
729 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
732 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
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);
743 if (!LHS) LHS = BE->getLHS();
744 if (!RHS) RHS = BE->getRHS();
746 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
750 assert(0 && "Invalid expression kind!");
754 /// ParseExpression - Parse an expression and return it.
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
764 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
765 // Parse the expression.
767 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
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) {
776 if (Lexer.isNot(AsmToken::Identifier))
777 return TokError("unexpected symbol modifier following '@'");
779 MCSymbolRefExpr::VariantKind Variant =
780 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
781 if (Variant == MCSymbolRefExpr::VK_Invalid)
782 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
784 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
786 return TokError("invalid modifier '" + getTok().getIdentifier() +
787 "' (no symbols present)");
795 // Try to constant fold it up front, if possible.
797 if (Res->EvaluateAsAbsolute(Value))
798 Res = MCConstantExpr::Create(Value, getContext());
803 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
805 return ParseParenExpr(Res, EndLoc) ||
806 ParseBinOpRHS(1, Res, EndLoc);
809 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
812 SMLoc StartLoc = Lexer.getLoc();
813 if (ParseExpression(Expr))
816 if (!Expr->EvaluateAsAbsolute(Res))
817 return Error(StartLoc, "expected absolute expression");
822 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
823 MCBinaryExpr::Opcode &Kind) {
826 return 0; // not a binop.
828 // Lowest Precedence: &&, ||
829 case AsmToken::AmpAmp:
830 Kind = MCBinaryExpr::LAnd;
832 case AsmToken::PipePipe:
833 Kind = MCBinaryExpr::LOr;
837 // Low Precedence: |, &, ^
839 // FIXME: gas seems to support '!' as an infix operator?
841 Kind = MCBinaryExpr::Or;
843 case AsmToken::Caret:
844 Kind = MCBinaryExpr::Xor;
847 Kind = MCBinaryExpr::And;
850 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
851 case AsmToken::EqualEqual:
852 Kind = MCBinaryExpr::EQ;
854 case AsmToken::ExclaimEqual:
855 case AsmToken::LessGreater:
856 Kind = MCBinaryExpr::NE;
859 Kind = MCBinaryExpr::LT;
861 case AsmToken::LessEqual:
862 Kind = MCBinaryExpr::LTE;
864 case AsmToken::Greater:
865 Kind = MCBinaryExpr::GT;
867 case AsmToken::GreaterEqual:
868 Kind = MCBinaryExpr::GTE;
871 // Intermediate Precedence: <<, >>
872 case AsmToken::LessLess:
873 Kind = MCBinaryExpr::Shl;
875 case AsmToken::GreaterGreater:
876 Kind = MCBinaryExpr::Shr;
879 // High Intermediate Precedence: +, -
881 Kind = MCBinaryExpr::Add;
883 case AsmToken::Minus:
884 Kind = MCBinaryExpr::Sub;
887 // Highest Precedence: *, /, %
889 Kind = MCBinaryExpr::Mul;
891 case AsmToken::Slash:
892 Kind = MCBinaryExpr::Div;
894 case AsmToken::Percent:
895 Kind = MCBinaryExpr::Mod;
901 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
902 /// Res contains the LHS of the expression on input.
903 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
906 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
907 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
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)
916 // Eat the next primary expression.
918 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
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;
928 // Merge LHS and RHS according to operator.
929 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
937 /// ::= EndOfStatement
938 /// ::= Label* Directive ...Operands... EndOfStatement
939 /// ::= Label* Identifier OperandList* EndOfStatement
940 bool AsmParser::ParseStatement() {
941 if (Lexer.is(AsmToken::EndOfStatement)) {
947 // Statements always start with an identifier or are a full line comment.
948 AsmToken ID = getTok();
949 SMLoc IDLoc = ID.getLoc();
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);
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");
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");
973 } else if (Lexer.is(AsmToken::Dot)) {
974 // Treat '.' as a valid identifier in this context.
978 } else if (ParseIdentifier(IDVal)) {
979 if (!TheCondState.Ignore)
980 return TokError("unexpected token at start of statement");
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
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);
1001 // If we are in a ".if 0" block, ignore this statement.
1002 if (TheCondState.Ignore) {
1003 EatToEndOfStatement();
1007 // FIXME: Recurse on local labels?
1009 // See what kind of statement we have.
1010 switch (Lexer.getKind()) {
1011 case AsmToken::Colon: {
1012 CheckForValidSection();
1014 // identifier ':' -> Label.
1017 // Diagnose attempt to use '.' as a label.
1019 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1021 // Diagnose attempt to use a variable as a label.
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.
1027 if (LocalLabelVal == -1)
1028 Sym = getContext().GetOrCreateSymbol(IDVal);
1030 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
1031 if (!Sym->isUndefined() || Sym->isVariable())
1032 return Error(IDLoc, "invalid symbol redefinition");
1037 // Consume any end of statement token, if present, to avoid spurious
1038 // AddBlankLine calls().
1039 if (Lexer.is(AsmToken::EndOfStatement)) {
1041 if (Lexer.is(AsmToken::Eof))
1045 return ParseStatement();
1048 case AsmToken::Equal:
1049 // identifier '=' ... -> assignment statement
1052 return ParseAssignment(IDVal, true);
1054 default: // Normal instruction or directive.
1058 // If macros are enabled, check to see if this is a macro instantiation.
1060 if (const Macro *M = MacroMap.lookup(IDVal))
1061 return HandleMacroEntry(IDVal, IDLoc, M);
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);
1073 if (IDVal == ".ascii")
1074 return ParseDirectiveAscii(IDVal, false);
1075 if (IDVal == ".asciz" || IDVal == ".string")
1076 return ParseDirectiveAscii(IDVal, true);
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);
1101 if (IDVal == ".align") {
1102 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1103 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1105 if (IDVal == ".align32") {
1106 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1107 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
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);
1122 if (IDVal == ".org")
1123 return ParseDirectiveOrg();
1125 if (IDVal == ".fill")
1126 return ParseDirectiveFill();
1127 if (IDVal == ".space" || IDVal == ".skip")
1128 return ParseDirectiveSpace();
1129 if (IDVal == ".zero")
1130 return ParseDirectiveZero();
1132 // Symbol attribute directives
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);
1155 if (IDVal == ".comm" || IDVal == ".common")
1156 return ParseDirectiveComm(/*IsLocal=*/false);
1157 if (IDVal == ".lcomm")
1158 return ParseDirectiveComm(/*IsLocal=*/true);
1160 if (IDVal == ".abort")
1161 return ParseDirectiveAbort();
1162 if (IDVal == ".include")
1163 return ParseDirectiveInclude();
1165 if (IDVal == ".code16")
1166 return TokError(Twine(IDVal) + " not supported yet");
1168 // Look up the handler in the handler table.
1169 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1170 DirectiveMap.lookup(IDVal);
1172 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1174 // Target hook for parsing target specific directives.
1175 if (!getTargetParser().ParseDirective(ID))
1178 bool retval = Warning(IDLoc, "ignoring directive for now");
1179 EatToEndOfStatement();
1183 CheckForValidSection();
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]));
1190 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1191 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
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) {
1202 ParsedOperands[i]->print(OS);
1206 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1209 // If parsing succeeded, match the instruction.
1211 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1214 // Free any parsed operands.
1215 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1216 delete ParsedOperands[i];
1218 // Don't skip the rest of the line, the instruction parser is responsible for
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.
1225 void AsmParser::EatToEndOfLine() {
1226 Lexer.LexUntilEndOfLine();
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.
1234 bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) {
1235 Lex(); // Eat the hash token.
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.
1244 int64_t LineNumber = getTok().getIntVal();
1247 if (getLexer().isNot(AsmToken::String)) {
1252 StringRef Filename = getTok().getString();
1253 // Get rid of the enclosing quotes.
1254 Filename = Filename.substr(1, Filename.size()-2);
1256 // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1258 CppHashFilename = Filename;
1259 CppHashLineNumber = LineNumber;
1261 // Ignore any trailing characters, they're just comment.
1266 /// DiagHandler - will use the the last parsed cpp hash line filename comment
1267 /// for the Filename and LineNo if any in the diagnostic.
1268 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1269 const AsmParser *Parser = static_cast<const AsmParser*>(Context);
1270 raw_ostream &OS = errs();
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);
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);
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) {
1295 // Use the CppHashFilename and calculate a line number based on the
1296 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1298 const std::string Filename = Parser->CppHashFilename;
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);
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());
1312 NewDiag.print(0, OS);
1315 bool AsmParser::expandMacro(SmallString<256> &Buf, StringRef Body,
1316 const std::vector<StringRef> &Parameters,
1317 const std::vector<std::vector<AsmToken> > &A,
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");
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.
1330 // This macro has no parameters, look for $0, $1, etc.
1331 if (Body[Pos] != '$' || Pos + 1 == End)
1334 char Next = Body[Pos + 1];
1335 if (Next == '$' || Next == 'n' || isdigit(Next))
1338 // This macro has parameters, look for \foo, \bar, etc.
1339 if (Body[Pos] == '\\' && Pos + 1 != End)
1345 OS << Body.slice(0, Pos);
1347 // Check if we reached the end.
1352 switch (Body[Pos+1]) {
1358 // $n => number of arguments
1363 // $[0-9] => argument
1365 // Missing arguments are ignored.
1366 unsigned Index = Body[Pos+1] - '0';
1367 if (Index >= A.size())
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();
1379 unsigned I = Pos + 1;
1380 while (isalnum(Body[I]) && I + 1 != End)
1383 const char *Begin = Body.data() + Pos +1;
1384 StringRef Argument(Begin, I - (Pos +1));
1386 for (; Index < NParameters; ++Index)
1387 if (Parameters[Index] == Argument)
1390 // FIXME: We should error at the macro definition.
1391 if (Index == NParameters)
1392 return Error(L, "Parameter not found");
1394 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1395 ie = A[Index].end(); it != ie; ++it)
1396 OS << it->getString();
1398 Pos += 1 + Argument.size();
1400 // Update the scan point.
1401 Body = Body.substr(Pos);
1404 // We include the .endmacro in the buffer as our queue to exit the macro
1406 OS << ".endmacro\n";
1410 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1412 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL)
1416 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
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");
1423 // Parse the macro instantiation arguments.
1424 std::vector<std::vector<AsmToken> > MacroArguments;
1425 MacroArguments.push_back(std::vector<AsmToken>());
1426 unsigned ParenLevel = 0;
1428 if (Lexer.is(AsmToken::Eof))
1429 return TokError("unexpected token in macro instantiation");
1430 if (Lexer.is(AsmToken::EndOfStatement))
1433 // If we aren't inside parentheses and this is a comma, start a new token
1435 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1436 MacroArguments.push_back(std::vector<AsmToken>());
1438 // Adjust the current parentheses level.
1439 if (Lexer.is(AsmToken::LParen))
1441 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1444 // Append the token to the current argument list.
1445 MacroArguments.back().push_back(getTok());
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;
1455 if (expandMacro(Buf, Body, M->Parameters, MacroArguments, getTok().getLoc()))
1458 MemoryBuffer *Instantiation =
1459 MemoryBuffer::getMemBufferCopy(Buf.str(), "<instantiation>");
1461 // Create the macro instantiation object and add to the current macro
1462 // instantiation stack.
1463 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1466 ActiveMacros.push_back(MI);
1468 // Jump to the macro instantiation and prime the lexer.
1469 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1470 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1476 void AsmParser::HandleMacroExit() {
1477 // Jump to the EndOfStatement we should return to, and consume it.
1478 JumpToLoc(ActiveMacros.back()->ExitLoc);
1481 // Pop the instantiation entry.
1482 delete ActiveMacros.back();
1483 ActiveMacros.pop_back();
1486 static 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());
1492 case MCExpr::Target:
1493 case MCExpr::Constant:
1495 case MCExpr::SymbolRef: {
1496 static_cast<const MCSymbolRefExpr*>(Value)->getSymbol().setUsed(true);
1500 MarkUsed(static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1505 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) {
1506 // FIXME: Use better location, we should use proper tokens.
1507 SMLoc EqualLoc = Lexer.getLoc();
1509 const MCExpr *Value;
1510 if (ParseExpression(Value))
1515 if (Lexer.isNot(AsmToken::EndOfStatement))
1516 return TokError("unexpected token in assignment");
1518 // Error on assignment to '.'.
1520 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported "
1521 "(use '.space' or '.org').)"));
1524 // Eat the end of statement marker.
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);
1531 // Diagnose assignment to a label.
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 '" +
1545 // Don't count these checks as uses.
1546 Sym->setUsed(false);
1548 Sym = getContext().GetOrCreateSymbol(Name);
1550 // FIXME: Handle '.'.
1552 // Do the assignment.
1553 Out.EmitAssignment(Sym, Value);
1558 /// ParseIdentifier:
1561 bool AsmParser::ParseIdentifier(StringRef &Res) {
1562 // The assembler has relaxed rules for accepting identifiers, in particular we
1563 // allow things like '.globl $foo', which would normally be separate
1564 // tokens. At this level, we have already lexed so we cannot (currently)
1565 // handle this as a context dependent token, instead we detect adjacent tokens
1566 // and return the combined identifier.
1567 if (Lexer.is(AsmToken::Dollar)) {
1568 SMLoc DollarLoc = getLexer().getLoc();
1570 // Consume the dollar sign, and check for a following identifier.
1572 if (Lexer.isNot(AsmToken::Identifier))
1575 // We have a '$' followed by an identifier, make sure they are adjacent.
1576 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1579 // Construct the joined identifier and consume the token.
1580 Res = StringRef(DollarLoc.getPointer(),
1581 getTok().getIdentifier().size() + 1);
1586 if (Lexer.isNot(AsmToken::Identifier) &&
1587 Lexer.isNot(AsmToken::String))
1590 Res = getTok().getIdentifier();
1592 Lex(); // Consume the identifier token.
1597 /// ParseDirectiveSet:
1598 /// ::= .equ identifier ',' expression
1599 /// ::= .equiv identifier ',' expression
1600 /// ::= .set identifier ',' expression
1601 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1604 if (ParseIdentifier(Name))
1605 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1607 if (getLexer().isNot(AsmToken::Comma))
1608 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1611 return ParseAssignment(Name, allow_redef);
1614 bool AsmParser::ParseEscapedString(std::string &Data) {
1615 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1618 StringRef Str = getTok().getStringContents();
1619 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1620 if (Str[i] != '\\') {
1625 // Recognize escaped characters. Note that this escape semantics currently
1626 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1629 return TokError("unexpected backslash at end of string");
1631 // Recognize octal sequences.
1632 if ((unsigned) (Str[i] - '0') <= 7) {
1633 // Consume up to three octal characters.
1634 unsigned Value = Str[i] - '0';
1636 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1638 Value = Value * 8 + (Str[i] - '0');
1640 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1642 Value = Value * 8 + (Str[i] - '0');
1647 return TokError("invalid octal escape sequence (out of range)");
1649 Data += (unsigned char) Value;
1653 // Otherwise recognize individual escapes.
1656 // Just reject invalid escape sequences for now.
1657 return TokError("invalid escape sequence (unrecognized character)");
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;
1672 /// ParseDirectiveAscii:
1673 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1674 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1675 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1676 CheckForValidSection();
1679 if (getLexer().isNot(AsmToken::String))
1680 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1683 if (ParseEscapedString(Data))
1686 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1688 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1692 if (getLexer().is(AsmToken::EndOfStatement))
1695 if (getLexer().isNot(AsmToken::Comma))
1696 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1705 /// ParseDirectiveValue
1706 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1707 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1708 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1709 CheckForValidSection();
1712 const MCExpr *Value;
1713 SMLoc ExprLoc = getLexer().getLoc();
1714 if (ParseExpression(Value))
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);
1725 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1727 if (getLexer().is(AsmToken::EndOfStatement))
1730 // FIXME: Improve diagnostic.
1731 if (getLexer().isNot(AsmToken::Comma))
1732 return TokError("unexpected token in directive");
1741 /// ParseDirectiveRealValue
1742 /// ::= (.single | .double) [ expression (, expression)* ]
1743 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1744 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1745 CheckForValidSection();
1748 // We don't truly support arithmetic on floating point expressions, so we
1749 // have to manually parse unary prefixes.
1751 if (getLexer().is(AsmToken::Minus)) {
1754 } else if (getLexer().is(AsmToken::Plus))
1757 if (getLexer().isNot(AsmToken::Integer) &&
1758 getLexer().isNot(AsmToken::Real) &&
1759 getLexer().isNot(AsmToken::Identifier))
1760 return TokError("unexpected token in directive");
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);
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");
1778 // Consume the numeric token.
1781 // Emit the value as an integer.
1782 APInt AsInt = Value.bitcastToAPInt();
1783 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1784 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1786 if (getLexer().is(AsmToken::EndOfStatement))
1789 if (getLexer().isNot(AsmToken::Comma))
1790 return TokError("unexpected token in directive");
1799 /// ParseDirectiveSpace
1800 /// ::= .space expression [ , expression ]
1801 bool AsmParser::ParseDirectiveSpace() {
1802 CheckForValidSection();
1805 if (ParseAbsoluteExpression(NumBytes))
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");
1814 if (ParseAbsoluteExpression(FillExpr))
1817 if (getLexer().isNot(AsmToken::EndOfStatement))
1818 return TokError("unexpected token in '.space' directive");
1824 return TokError("invalid number of bytes in '.space' directive");
1826 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1827 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1832 /// ParseDirectiveZero
1833 /// ::= .zero expression
1834 bool AsmParser::ParseDirectiveZero() {
1835 CheckForValidSection();
1838 if (ParseAbsoluteExpression(NumBytes))
1842 if (getLexer().is(AsmToken::Comma)) {
1844 if (ParseAbsoluteExpression(Val))
1848 if (getLexer().isNot(AsmToken::EndOfStatement))
1849 return TokError("unexpected token in '.zero' directive");
1853 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1858 /// ParseDirectiveFill
1859 /// ::= .fill expression , expression , expression
1860 bool AsmParser::ParseDirectiveFill() {
1861 CheckForValidSection();
1864 if (ParseAbsoluteExpression(NumValues))
1867 if (getLexer().isNot(AsmToken::Comma))
1868 return TokError("unexpected token in '.fill' directive");
1872 if (ParseAbsoluteExpression(FillSize))
1875 if (getLexer().isNot(AsmToken::Comma))
1876 return TokError("unexpected token in '.fill' directive");
1880 if (ParseAbsoluteExpression(FillExpr))
1883 if (getLexer().isNot(AsmToken::EndOfStatement))
1884 return TokError("unexpected token in '.fill' directive");
1888 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1889 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1891 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1892 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1897 /// ParseDirectiveOrg
1898 /// ::= .org expression [ , expression ]
1899 bool AsmParser::ParseDirectiveOrg() {
1900 CheckForValidSection();
1902 const MCExpr *Offset;
1903 if (ParseExpression(Offset))
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");
1913 if (ParseAbsoluteExpression(FillExpr))
1916 if (getLexer().isNot(AsmToken::EndOfStatement))
1917 return TokError("unexpected token in '.org' directive");
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);
1929 /// ParseDirectiveAlign
1930 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1931 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1932 CheckForValidSection();
1934 SMLoc AlignmentLoc = getLexer().getLoc();
1936 if (ParseAbsoluteExpression(Alignment))
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");
1948 // The fill expression can be omitted while specifying a maximum number of
1949 // alignment bytes, e.g:
1951 if (getLexer().isNot(AsmToken::Comma)) {
1953 if (ParseAbsoluteExpression(FillExpr))
1957 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1958 if (getLexer().isNot(AsmToken::Comma))
1959 return TokError("unexpected token in directive");
1962 MaxBytesLoc = getLexer().getLoc();
1963 if (ParseAbsoluteExpression(MaxBytesToFill))
1966 if (getLexer().isNot(AsmToken::EndOfStatement))
1967 return TokError("unexpected token in directive");
1976 // Compute alignment in bytes.
1978 // FIXME: Diagnose overflow.
1979 if (Alignment >= 32) {
1980 Error(AlignmentLoc, "invalid alignment value");
1984 Alignment = 1ULL << Alignment;
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");
1995 if (MaxBytesToFill >= Alignment) {
1996 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2002 // Check whether we should use optimal code alignment for this .align
2004 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
2005 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2006 ValueSize == 1 && UseCodeAlign) {
2007 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2009 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2010 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2017 /// ParseDirectiveSymbolAttribute
2018 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
2019 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
2020 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2023 SMLoc Loc = getTok().getLoc();
2025 if (ParseIdentifier(Name))
2026 return Error(Loc, "expected identifier in directive");
2028 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
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");
2034 getStreamer().EmitSymbolAttribute(Sym, Attr);
2036 if (getLexer().is(AsmToken::EndOfStatement))
2039 if (getLexer().isNot(AsmToken::Comma))
2040 return TokError("unexpected token in directive");
2049 /// ParseDirectiveComm
2050 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
2051 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
2052 CheckForValidSection();
2054 SMLoc IDLoc = getLexer().getLoc();
2056 if (ParseIdentifier(Name))
2057 return TokError("expected identifier in directive");
2059 // Handle the identifier as the key symbol.
2060 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2062 if (getLexer().isNot(AsmToken::Comma))
2063 return TokError("unexpected token in directive");
2067 SMLoc SizeLoc = getLexer().getLoc();
2068 if (ParseAbsoluteExpression(Size))
2071 int64_t Pow2Alignment = 0;
2072 SMLoc Pow2AlignmentLoc;
2073 if (getLexer().is(AsmToken::Comma)) {
2075 Pow2AlignmentLoc = getLexer().getLoc();
2076 if (ParseAbsoluteExpression(Pow2Alignment))
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);
2087 if (getLexer().isNot(AsmToken::EndOfStatement))
2088 return TokError("unexpected token in '.comm' or '.lcomm' directive");
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.
2095 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
2096 "be less than zero");
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");
2105 if (!Sym->isUndefined())
2106 return Error(IDLoc, "invalid symbol redefinition");
2108 // '.lcomm' is equivalent to '.zerofill'.
2109 // Create the Symbol as a common or local common with Size and Pow2Alignment
2111 getStreamer().EmitZerofill(Ctx.getMachOSection(
2112 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
2113 0, SectionKind::getBSS()),
2114 Sym, Size, 1 << Pow2Alignment);
2118 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
2122 /// ParseDirectiveAbort
2123 /// ::= .abort [... message ...]
2124 bool AsmParser::ParseDirectiveAbort() {
2125 // FIXME: Use loc from directive.
2126 SMLoc Loc = getLexer().getLoc();
2128 StringRef Str = ParseStringToEndOfStatement();
2129 if (getLexer().isNot(AsmToken::EndOfStatement))
2130 return TokError("unexpected token in '.abort' directive");
2135 Error(Loc, ".abort detected. Assembly stopping.");
2137 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
2138 // FIXME: Actually abort assembly here.
2143 /// ParseDirectiveInclude
2144 /// ::= .include "filename"
2145 bool AsmParser::ParseDirectiveInclude() {
2146 if (getLexer().isNot(AsmToken::String))
2147 return TokError("expected string in '.include' directive");
2149 std::string Filename = getTok().getString();
2150 SMLoc IncludeLoc = getLexer().getLoc();
2153 if (getLexer().isNot(AsmToken::EndOfStatement))
2154 return TokError("unexpected token in '.include' directive");
2156 // Strip the quotes.
2157 Filename = Filename.substr(1, Filename.size()-2);
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 + "'");
2169 /// ParseDirectiveIf
2170 /// ::= .if expression
2171 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
2172 TheCondStack.push_back(TheCondState);
2173 TheCondState.TheCond = AsmCond::IfCond;
2174 if(TheCondState.Ignore) {
2175 EatToEndOfStatement();
2179 if (ParseAbsoluteExpression(ExprValue))
2182 if (getLexer().isNot(AsmToken::EndOfStatement))
2183 return TokError("unexpected token in '.if' directive");
2187 TheCondState.CondMet = ExprValue;
2188 TheCondState.Ignore = !TheCondState.CondMet;
2194 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
2196 TheCondStack.push_back(TheCondState);
2197 TheCondState.TheCond = AsmCond::IfCond;
2199 if (TheCondState.Ignore) {
2200 EatToEndOfStatement();
2202 if (ParseIdentifier(Name))
2203 return TokError("expected identifier after '.ifdef'");
2207 MCSymbol *Sym = getContext().LookupSymbol(Name);
2210 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
2212 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
2213 TheCondState.Ignore = !TheCondState.CondMet;
2219 /// ParseDirectiveElseIf
2220 /// ::= .elseif expression
2221 bool 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 "
2226 TheCondState.TheCond = AsmCond::ElseIfCond;
2228 bool LastIgnoreState = false;
2229 if (!TheCondStack.empty())
2230 LastIgnoreState = TheCondStack.back().Ignore;
2231 if (LastIgnoreState || TheCondState.CondMet) {
2232 TheCondState.Ignore = true;
2233 EatToEndOfStatement();
2237 if (ParseAbsoluteExpression(ExprValue))
2240 if (getLexer().isNot(AsmToken::EndOfStatement))
2241 return TokError("unexpected token in '.elseif' directive");
2244 TheCondState.CondMet = ExprValue;
2245 TheCondState.Ignore = !TheCondState.CondMet;
2251 /// ParseDirectiveElse
2253 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2254 if (getLexer().isNot(AsmToken::EndOfStatement))
2255 return TokError("unexpected token in '.else' directive");
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 "
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;
2270 TheCondState.Ignore = false;
2275 /// ParseDirectiveEndIf
2277 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2278 if (getLexer().isNot(AsmToken::EndOfStatement))
2279 return TokError("unexpected token in '.endif' directive");
2283 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2284 TheCondStack.empty())
2285 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2287 if (!TheCondStack.empty()) {
2288 TheCondState = TheCondStack.back();
2289 TheCondStack.pop_back();
2295 /// ParseDirectiveFile
2296 /// ::= .file [number] string
2297 bool 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();
2306 return TokError("file number less than one");
2309 if (getLexer().isNot(AsmToken::String))
2310 return TokError("unexpected token in '.file' directive");
2312 StringRef Filename = getTok().getString();
2313 Filename = Filename.substr(1, Filename.size()-2);
2316 if (getLexer().isNot(AsmToken::EndOfStatement))
2317 return TokError("unexpected token in '.file' directive");
2319 if (FileNumber == -1)
2320 getStreamer().EmitFileDirective(Filename);
2322 if (getStreamer().EmitDwarfFileDirective(FileNumber, Filename))
2323 Error(FileNumberLoc, "file number already allocated");
2329 /// ParseDirectiveLine
2330 /// ::= .line [number]
2331 bool 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");
2336 int64_t LineNumber = getTok().getIntVal();
2340 // FIXME: Do something with the .line.
2343 if (getLexer().isNot(AsmToken::EndOfStatement))
2344 return TokError("unexpected token in '.line' directive");
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.
2357 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2359 if (getLexer().isNot(AsmToken::Integer))
2360 return TokError("unexpected token in '.loc' directive");
2361 int64_t FileNumber = getTok().getIntVal();
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");
2368 int64_t LineNumber = 0;
2369 if (getLexer().is(AsmToken::Integer)) {
2370 LineNumber = getTok().getIntVal();
2372 return TokError("line number less than one in '.loc' directive");
2376 int64_t ColumnPos = 0;
2377 if (getLexer().is(AsmToken::Integer)) {
2378 ColumnPos = getTok().getIntVal();
2380 return TokError("column position less than zero in '.loc' directive");
2384 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2386 int64_t Discriminator = 0;
2387 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2389 if (getLexer().is(AsmToken::EndOfStatement))
2393 SMLoc Loc = getTok().getLoc();
2394 if (getParser().ParseIdentifier(Name))
2395 return TokError("unexpected token in '.loc' directive");
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))
2408 // The expression must be the constant 0 or 1.
2409 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2410 int Value = MCE->getValue();
2412 Flags &= ~DWARF2_FLAG_IS_STMT;
2413 else if (Value == 1)
2414 Flags |= DWARF2_FLAG_IS_STMT;
2416 return Error(Loc, "is_stmt value not 0 or 1");
2419 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2422 else if (Name == "isa") {
2423 SMLoc Loc = getTok().getLoc();
2424 const MCExpr *Value;
2425 if (getParser().ParseExpression(Value))
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();
2431 return Error(Loc, "isa number less than zero");
2435 return Error(Loc, "isa number not a constant value");
2438 else if (Name == "discriminator") {
2439 if (getParser().ParseAbsoluteExpression(Discriminator))
2443 return Error(Loc, "unknown sub-directive in '.loc' directive");
2446 if (getLexer().is(AsmToken::EndOfStatement))
2451 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2452 Isa, Discriminator, StringRef());
2457 /// ParseDirectiveStabs
2458 /// ::= .stabs string, number, number, number
2459 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2460 SMLoc DirectiveLoc) {
2461 return TokError("unsupported directive '" + Directive + "'");
2464 /// ParseDirectiveCFISections
2465 /// ::= .cfi_sections section [, section]
2466 bool GenericAsmParser::ParseDirectiveCFISections(StringRef,
2467 SMLoc DirectiveLoc) {
2472 if (getParser().ParseIdentifier(Name))
2473 return TokError("Expected an identifier");
2475 if (Name == ".eh_frame")
2477 else if (Name == ".debug_frame")
2480 if (getLexer().is(AsmToken::Comma)) {
2483 if (getParser().ParseIdentifier(Name))
2484 return TokError("Expected an identifier");
2486 if (Name == ".eh_frame")
2488 else if (Name == ".debug_frame")
2492 getStreamer().EmitCFISections(EH, Debug);
2497 /// ParseDirectiveCFIStartProc
2498 /// ::= .cfi_startproc
2499 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2500 SMLoc DirectiveLoc) {
2501 getStreamer().EmitCFIStartProc();
2505 /// ParseDirectiveCFIEndProc
2506 /// ::= .cfi_endproc
2507 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2508 getStreamer().EmitCFIEndProc();
2512 /// ParseRegisterOrRegisterNumber - parse register name or number.
2513 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2514 SMLoc DirectiveLoc) {
2517 if (getLexer().isNot(AsmToken::Integer)) {
2518 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2521 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true);
2523 return getParser().ParseAbsoluteExpression(Register);
2528 /// ParseDirectiveCFIDefCfa
2529 /// ::= .cfi_def_cfa register, offset
2530 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
2531 SMLoc DirectiveLoc) {
2532 int64_t Register = 0;
2533 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2536 if (getLexer().isNot(AsmToken::Comma))
2537 return TokError("unexpected token in directive");
2541 if (getParser().ParseAbsoluteExpression(Offset))
2544 getStreamer().EmitCFIDefCfa(Register, Offset);
2548 /// ParseDirectiveCFIDefCfaOffset
2549 /// ::= .cfi_def_cfa_offset offset
2550 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
2551 SMLoc DirectiveLoc) {
2553 if (getParser().ParseAbsoluteExpression(Offset))
2556 getStreamer().EmitCFIDefCfaOffset(Offset);
2560 /// ParseDirectiveCFIAdjustCfaOffset
2561 /// ::= .cfi_adjust_cfa_offset adjustment
2562 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef,
2563 SMLoc DirectiveLoc) {
2564 int64_t Adjustment = 0;
2565 if (getParser().ParseAbsoluteExpression(Adjustment))
2568 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
2572 /// ParseDirectiveCFIDefCfaRegister
2573 /// ::= .cfi_def_cfa_register register
2574 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
2575 SMLoc DirectiveLoc) {
2576 int64_t Register = 0;
2577 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2580 getStreamer().EmitCFIDefCfaRegister(Register);
2584 /// ParseDirectiveCFIOffset
2585 /// ::= .cfi_offset register, offset
2586 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
2587 int64_t Register = 0;
2590 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2593 if (getLexer().isNot(AsmToken::Comma))
2594 return TokError("unexpected token in directive");
2597 if (getParser().ParseAbsoluteExpression(Offset))
2600 getStreamer().EmitCFIOffset(Register, Offset);
2604 /// ParseDirectiveCFIRelOffset
2605 /// ::= .cfi_rel_offset register, offset
2606 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef,
2607 SMLoc DirectiveLoc) {
2608 int64_t Register = 0;
2610 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2613 if (getLexer().isNot(AsmToken::Comma))
2614 return TokError("unexpected token in directive");
2618 if (getParser().ParseAbsoluteExpression(Offset))
2621 getStreamer().EmitCFIRelOffset(Register, Offset);
2625 static bool isValidEncoding(int64_t Encoding) {
2626 if (Encoding & ~0xff)
2629 if (Encoding == dwarf::DW_EH_PE_omit)
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)
2639 const unsigned Application = Encoding & 0x70;
2640 if (Application != dwarf::DW_EH_PE_absptr &&
2641 Application != dwarf::DW_EH_PE_pcrel)
2647 /// ParseDirectiveCFIPersonalityOrLsda
2648 /// ::= .cfi_personality encoding, [symbol_name]
2649 /// ::= .cfi_lsda encoding, [symbol_name]
2650 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
2651 SMLoc DirectiveLoc) {
2652 int64_t Encoding = 0;
2653 if (getParser().ParseAbsoluteExpression(Encoding))
2655 if (Encoding == dwarf::DW_EH_PE_omit)
2658 if (!isValidEncoding(Encoding))
2659 return TokError("unsupported encoding.");
2661 if (getLexer().isNot(AsmToken::Comma))
2662 return TokError("unexpected token in directive");
2666 if (getParser().ParseIdentifier(Name))
2667 return TokError("expected identifier in directive");
2669 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2671 if (IDVal == ".cfi_personality")
2672 getStreamer().EmitCFIPersonality(Sym, Encoding);
2674 assert(IDVal == ".cfi_lsda");
2675 getStreamer().EmitCFILsda(Sym, Encoding);
2680 /// ParseDirectiveCFIRememberState
2681 /// ::= .cfi_remember_state
2682 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
2683 SMLoc DirectiveLoc) {
2684 getStreamer().EmitCFIRememberState();
2688 /// ParseDirectiveCFIRestoreState
2689 /// ::= .cfi_remember_state
2690 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
2691 SMLoc DirectiveLoc) {
2692 getStreamer().EmitCFIRestoreState();
2696 /// ParseDirectiveCFISameValue
2697 /// ::= .cfi_same_value register
2698 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal,
2699 SMLoc DirectiveLoc) {
2700 int64_t Register = 0;
2702 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2705 getStreamer().EmitCFISameValue(Register);
2710 /// ParseDirectiveMacrosOnOff
2713 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2714 SMLoc DirectiveLoc) {
2715 if (getLexer().isNot(AsmToken::EndOfStatement))
2716 return Error(getLexer().getLoc(),
2717 "unexpected token in '" + Directive + "' directive");
2719 getParser().MacrosEnabled = Directive == ".macros_on";
2724 /// ParseDirectiveMacro
2725 /// ::= .macro name [parameters]
2726 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2727 SMLoc DirectiveLoc) {
2729 if (getParser().ParseIdentifier(Name))
2730 return TokError("expected identifier in directive");
2732 std::vector<StringRef> Parameters;
2733 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2735 StringRef Parameter;
2736 if (getParser().ParseIdentifier(Parameter))
2737 return TokError("expected identifier in directive");
2738 Parameters.push_back(Parameter);
2740 if (getLexer().isNot(AsmToken::Comma))
2746 if (getLexer().isNot(AsmToken::EndOfStatement))
2747 return TokError("unexpected token in '.macro' directive");
2749 // Eat the end of statement.
2752 AsmToken EndToken, StartToken = getTok();
2754 // Lex the macro definition.
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");
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();
2766 if (getLexer().isNot(AsmToken::EndOfStatement))
2767 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2772 // Otherwise, scan til the end of the statement.
2773 getParser().EatToEndOfStatement();
2776 if (getParser().MacroMap.lookup(Name)) {
2777 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
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);
2787 /// ParseDirectiveEndMacro
2790 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2791 SMLoc DirectiveLoc) {
2792 if (getLexer().isNot(AsmToken::EndOfStatement))
2793 return TokError("unexpected token in '" + Directive + "' directive");
2795 // If we are inside a macro instantiation, terminate the current
2797 if (!getParser().ActiveMacros.empty()) {
2798 getParser().HandleMacroExit();
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");
2808 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2809 getParser().CheckForValidSection();
2811 const MCExpr *Value;
2813 if (getParser().ParseExpression(Value))
2816 if (getLexer().isNot(AsmToken::EndOfStatement))
2817 return TokError("unexpected token in directive");
2819 if (DirName[1] == 's')
2820 getStreamer().EmitSLEB128Value(Value);
2822 getStreamer().EmitULEB128Value(Value);
2828 /// \brief Create an MCAsmParser instance.
2829 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM,
2830 MCContext &C, MCStreamer &Out,
2831 const MCAsmInfo &MAI) {
2832 return new AsmParser(SM, C, Out, MAI);