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/Twine.h"
18 #include "llvm/MC/MCAsmInfo.h"
19 #include "llvm/MC/MCContext.h"
20 #include "llvm/MC/MCDwarf.h"
21 #include "llvm/MC/MCExpr.h"
22 #include "llvm/MC/MCParser/AsmCond.h"
23 #include "llvm/MC/MCParser/AsmLexer.h"
24 #include "llvm/MC/MCParser/MCAsmParser.h"
25 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
26 #include "llvm/MC/MCRegisterInfo.h"
27 #include "llvm/MC/MCSectionMachO.h"
28 #include "llvm/MC/MCStreamer.h"
29 #include "llvm/MC/MCSymbol.h"
30 #include "llvm/MC/MCTargetAsmParser.h"
31 #include "llvm/Support/CommandLine.h"
32 #include "llvm/Support/ErrorHandling.h"
33 #include "llvm/Support/MathExtras.h"
34 #include "llvm/Support/MemoryBuffer.h"
35 #include "llvm/Support/SourceMgr.h"
36 #include "llvm/Support/raw_ostream.h"
42 FatalAssemblerWarnings("fatal-assembler-warnings",
43 cl::desc("Consider warnings as error"));
47 /// \brief Helper class for tracking macro definitions.
48 typedef std::vector<AsmToken> MacroArgument;
49 typedef std::vector<MacroArgument> MacroArguments;
50 typedef StringRef MacroParameter;
51 typedef std::vector<MacroParameter> MacroParameters;
56 MacroParameters Parameters;
59 Macro(StringRef N, StringRef B, const MacroParameters &P) :
60 Name(N), Body(B), Parameters(P) {}
63 /// \brief Helper class for storing information about an active macro
65 struct MacroInstantiation {
66 /// The macro being instantiated.
67 const Macro *TheMacro;
69 /// The macro instantiation with substitutions.
70 MemoryBuffer *Instantiation;
72 /// The location of the instantiation.
73 SMLoc InstantiationLoc;
75 /// The location where parsing should resume upon instantiation completion.
79 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
83 /// \brief The concrete assembly parser instance.
84 class AsmParser : public MCAsmParser {
85 friend class GenericAsmParser;
87 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
88 void operator=(const AsmParser &); // DO NOT IMPLEMENT
95 SourceMgr::DiagHandlerTy SavedDiagHandler;
96 void *SavedDiagContext;
97 MCAsmParserExtension *GenericParser;
98 MCAsmParserExtension *PlatformParser;
100 /// This is the current buffer index we're lexing from as managed by the
101 /// SourceMgr object.
104 AsmCond TheCondState;
105 std::vector<AsmCond> TheCondStack;
107 /// DirectiveMap - This is a table handlers for directives. Each handler is
108 /// invoked after the directive identifier is read and is responsible for
109 /// parsing and validating the rest of the directive. The handler is passed
110 /// in the directive name and the location of the directive keyword.
111 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
113 /// MacroMap - Map of currently defined macros.
114 StringMap<Macro*> MacroMap;
116 /// ActiveMacros - Stack of active macro instantiations.
117 std::vector<MacroInstantiation*> ActiveMacros;
119 /// Boolean tracking whether macro substitution is enabled.
120 unsigned MacrosEnabled : 1;
122 /// Flag tracking whether any errors have been encountered.
123 unsigned HadError : 1;
125 /// The values from the last parsed cpp hash file line comment if any.
126 StringRef CppHashFilename;
127 int64_t CppHashLineNumber;
130 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
131 unsigned AssemblerDialect;
134 AsmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
135 const MCAsmInfo &MAI);
138 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
140 void AddDirectiveHandler(MCAsmParserExtension *Object,
142 DirectiveHandler Handler) {
143 DirectiveMap[Directive] = std::make_pair(Object, Handler);
147 /// @name MCAsmParser Interface
150 virtual SourceMgr &getSourceManager() { return SrcMgr; }
151 virtual MCAsmLexer &getLexer() { return Lexer; }
152 virtual MCContext &getContext() { return Ctx; }
153 virtual MCStreamer &getStreamer() { return Out; }
154 virtual unsigned getAssemblerDialect() {
155 if (AssemblerDialect == ~0U)
156 return MAI.getAssemblerDialect();
158 return AssemblerDialect;
160 virtual void setAssemblerDialect(unsigned i) {
161 AssemblerDialect = i;
164 virtual bool Warning(SMLoc L, const Twine &Msg,
165 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>());
166 virtual bool Error(SMLoc L, const Twine &Msg,
167 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>());
169 const AsmToken &Lex();
171 bool ParseExpression(const MCExpr *&Res);
172 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
173 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
174 virtual bool ParseAbsoluteExpression(int64_t &Res);
179 void CheckForValidSection();
181 bool ParseStatement();
182 void EatToEndOfLine();
183 bool ParseCppHashLineFilenameComment(const SMLoc &L);
185 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
186 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
187 const MacroParameters &Parameters,
188 const MacroArguments &A,
190 void HandleMacroExit();
192 void PrintMacroInstantiations();
193 void PrintMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
194 ArrayRef<SMRange> Ranges = ArrayRef<SMRange>()) const {
195 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
197 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
199 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
200 bool EnterIncludeFile(const std::string &Filename);
201 /// ProcessIncbinFile - Process the specified file for the .incbin directive.
202 /// This returns true on failure.
203 bool ProcessIncbinFile(const std::string &Filename);
205 /// \brief Reset the current lexer position to that given by \arg Loc. The
206 /// current token is not set; clients should ensure Lex() is called
208 void JumpToLoc(SMLoc Loc);
210 void EatToEndOfStatement();
212 bool ParseMacroArgument(MacroArgument &MA);
213 bool ParseMacroArguments(const Macro *M, MacroArguments &A);
215 /// \brief Parse up to the end of statement and a return the contents from the
216 /// current token until the end of the statement; the current token on exit
217 /// will be either the EndOfStatement or EOF.
218 StringRef ParseStringToEndOfStatement();
220 /// \brief Parse until the end of a statement or a comma is encountered,
221 /// return the contents from the current token up to the end or comma.
222 StringRef ParseStringToComma();
224 bool ParseAssignment(StringRef Name, bool allow_redef);
226 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
227 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
228 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
229 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
231 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
232 /// and set \arg Res to the identifier contents.
233 bool ParseIdentifier(StringRef &Res);
235 // Directive Parsing.
237 // ".ascii", ".asciiz", ".string"
238 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
239 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
240 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
241 bool ParseDirectiveFill(); // ".fill"
242 bool ParseDirectiveSpace(); // ".space"
243 bool ParseDirectiveZero(); // ".zero"
244 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv"
245 bool ParseDirectiveOrg(); // ".org"
246 // ".align{,32}", ".p2align{,w,l}"
247 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
249 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
250 /// accepts a single symbol (which should be a label or an external).
251 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
253 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
255 bool ParseDirectiveAbort(); // ".abort"
256 bool ParseDirectiveInclude(); // ".include"
257 bool ParseDirectiveIncbin(); // ".incbin"
259 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
260 // ".ifb" or ".ifnb", depending on ExpectBlank.
261 bool ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
262 // ".ifc" or ".ifnc", depending on ExpectEqual.
263 bool ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual);
264 // ".ifdef" or ".ifndef", depending on expect_defined
265 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
266 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
267 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
268 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
270 /// ParseEscapedString - Parse the current token as a string which may include
271 /// escaped characters and return the string contents.
272 bool ParseEscapedString(std::string &Data);
274 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
275 MCSymbolRefExpr::VariantKind Variant);
277 // Macro-like directives
278 Macro *ParseMacroLikeBody(SMLoc DirectiveLoc);
279 void InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc,
280 raw_svector_ostream &OS);
281 bool ParseDirectiveRept(SMLoc DirectiveLoc); // ".rept"
282 bool ParseDirectiveIrp(SMLoc DirectiveLoc); // ".irp"
283 bool ParseDirectiveIrpc(SMLoc DirectiveLoc); // ".irpc"
284 bool ParseDirectiveEndr(SMLoc DirectiveLoc); // ".endr"
287 /// \brief Generic implementations of directive handling, etc. which is shared
288 /// (or the default, at least) for all assembler parser.
289 class GenericAsmParser : public MCAsmParserExtension {
290 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
291 void AddDirectiveHandler(StringRef Directive) {
292 getParser().AddDirectiveHandler(this, Directive,
293 HandleDirective<GenericAsmParser, Handler>);
296 GenericAsmParser() {}
298 AsmParser &getParser() {
299 return (AsmParser&) this->MCAsmParserExtension::getParser();
302 virtual void Initialize(MCAsmParser &Parser) {
303 // Call the base implementation.
304 this->MCAsmParserExtension::Initialize(Parser);
306 // Debugging directives.
307 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
308 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
309 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
310 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
313 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFISections>(
315 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>(
317 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>(
319 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>(
321 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>(
322 ".cfi_def_cfa_offset");
323 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>(
324 ".cfi_adjust_cfa_offset");
325 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>(
326 ".cfi_def_cfa_register");
327 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>(
329 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>(
332 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality");
334 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda");
336 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state");
338 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state");
340 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value");
342 &GenericAsmParser::ParseDirectiveCFIRestore>(".cfi_restore");
344 &GenericAsmParser::ParseDirectiveCFIEscape>(".cfi_escape");
346 &GenericAsmParser::ParseDirectiveCFISignalFrame>(".cfi_signal_frame");
349 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
351 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
353 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
354 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
355 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
356 AddDirectiveHandler<&GenericAsmParser::ParseDirectivePurgeMacro>(".purgem");
358 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
359 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
362 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
364 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
365 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
366 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
367 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
368 bool ParseDirectiveCFISections(StringRef, SMLoc DirectiveLoc);
369 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc);
370 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc);
371 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc);
372 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc);
373 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc);
374 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc);
375 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc);
376 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc);
377 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc);
378 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc);
379 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc);
380 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc);
381 bool ParseDirectiveCFIRestore(StringRef, SMLoc DirectiveLoc);
382 bool ParseDirectiveCFIEscape(StringRef, SMLoc DirectiveLoc);
383 bool ParseDirectiveCFISignalFrame(StringRef, SMLoc DirectiveLoc);
385 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
386 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
387 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
388 bool ParseDirectivePurgeMacro(StringRef, SMLoc DirectiveLoc);
390 bool ParseDirectiveLEB128(StringRef, SMLoc);
397 extern MCAsmParserExtension *createDarwinAsmParser();
398 extern MCAsmParserExtension *createELFAsmParser();
399 extern MCAsmParserExtension *createCOFFAsmParser();
403 enum { DEFAULT_ADDRSPACE = 0 };
405 AsmParser::AsmParser(SourceMgr &_SM, MCContext &_Ctx,
406 MCStreamer &_Out, const MCAsmInfo &_MAI)
407 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), MAI(_MAI), SrcMgr(_SM),
408 GenericParser(new GenericAsmParser), PlatformParser(0),
409 CurBuffer(0), MacrosEnabled(true), CppHashLineNumber(0),
410 AssemblerDialect(~0U) {
411 // Save the old handler.
412 SavedDiagHandler = SrcMgr.getDiagHandler();
413 SavedDiagContext = SrcMgr.getDiagContext();
414 // Set our own handler which calls the saved handler.
415 SrcMgr.setDiagHandler(DiagHandler, this);
416 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
418 // Initialize the generic parser.
419 GenericParser->Initialize(*this);
421 // Initialize the platform / file format parser.
423 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
425 if (_MAI.hasMicrosoftFastStdCallMangling()) {
426 PlatformParser = createCOFFAsmParser();
427 PlatformParser->Initialize(*this);
428 } else if (_MAI.hasSubsectionsViaSymbols()) {
429 PlatformParser = createDarwinAsmParser();
430 PlatformParser->Initialize(*this);
432 PlatformParser = createELFAsmParser();
433 PlatformParser->Initialize(*this);
437 AsmParser::~AsmParser() {
438 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
440 // Destroy any macros.
441 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
442 ie = MacroMap.end(); it != ie; ++it)
443 delete it->getValue();
445 delete PlatformParser;
446 delete GenericParser;
449 void AsmParser::PrintMacroInstantiations() {
450 // Print the active macro instantiation stack.
451 for (std::vector<MacroInstantiation*>::const_reverse_iterator
452 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
453 PrintMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
454 "while in macro instantiation");
457 bool AsmParser::Warning(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
458 if (FatalAssemblerWarnings)
459 return Error(L, Msg, Ranges);
460 PrintMessage(L, SourceMgr::DK_Warning, Msg, Ranges);
461 PrintMacroInstantiations();
465 bool AsmParser::Error(SMLoc L, const Twine &Msg, ArrayRef<SMRange> Ranges) {
467 PrintMessage(L, SourceMgr::DK_Error, Msg, Ranges);
468 PrintMacroInstantiations();
472 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
473 std::string IncludedFile;
474 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
480 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
485 /// Process the specified .incbin file by seaching for it in the include paths
486 /// then just emitting the byte contents of the file to the streamer. This
487 /// returns true on failure.
488 bool AsmParser::ProcessIncbinFile(const std::string &Filename) {
489 std::string IncludedFile;
490 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
494 // Pick up the bytes from the file and emit them.
495 getStreamer().EmitBytes(SrcMgr.getMemoryBuffer(NewBuf)->getBuffer(),
500 void AsmParser::JumpToLoc(SMLoc Loc) {
501 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
502 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
505 const AsmToken &AsmParser::Lex() {
506 const AsmToken *tok = &Lexer.Lex();
508 if (tok->is(AsmToken::Eof)) {
509 // If this is the end of an included file, pop the parent file off the
511 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
512 if (ParentIncludeLoc != SMLoc()) {
513 JumpToLoc(ParentIncludeLoc);
518 if (tok->is(AsmToken::Error))
519 Error(Lexer.getErrLoc(), Lexer.getErr());
524 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
525 // Create the initial section, if requested.
526 if (!NoInitialTextSection)
533 AsmCond StartingCondState = TheCondState;
535 // If we are generating dwarf for assembly source files save the initial text
536 // section and generate a .file directive.
537 if (getContext().getGenDwarfForAssembly()) {
538 getContext().setGenDwarfSection(getStreamer().getCurrentSection());
539 MCSymbol *SectionStartSym = getContext().CreateTempSymbol();
540 getStreamer().EmitLabel(SectionStartSym);
541 getContext().setGenDwarfSectionStartSym(SectionStartSym);
542 getStreamer().EmitDwarfFileDirective(getContext().nextGenDwarfFileNumber(),
543 StringRef(), SrcMgr.getMemoryBuffer(CurBuffer)->getBufferIdentifier());
546 // While we have input, parse each statement.
547 while (Lexer.isNot(AsmToken::Eof)) {
548 if (!ParseStatement()) continue;
550 // We had an error, validate that one was emitted and recover by skipping to
552 assert(HadError && "Parse statement returned an error, but none emitted!");
553 EatToEndOfStatement();
556 if (TheCondState.TheCond != StartingCondState.TheCond ||
557 TheCondState.Ignore != StartingCondState.Ignore)
558 return TokError("unmatched .ifs or .elses");
560 // Check to see there are no empty DwarfFile slots.
561 const std::vector<MCDwarfFile *> &MCDwarfFiles =
562 getContext().getMCDwarfFiles();
563 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
564 if (!MCDwarfFiles[i])
565 TokError("unassigned file number: " + Twine(i) + " for .file directives");
568 // Check to see that all assembler local symbols were actually defined.
569 // Targets that don't do subsections via symbols may not want this, though,
570 // so conservatively exclude them. Only do this if we're finalizing, though,
571 // as otherwise we won't necessarilly have seen everything yet.
572 if (!NoFinalize && MAI.hasSubsectionsViaSymbols()) {
573 const MCContext::SymbolTable &Symbols = getContext().getSymbols();
574 for (MCContext::SymbolTable::const_iterator i = Symbols.begin(),
577 MCSymbol *Sym = i->getValue();
578 // Variable symbols may not be marked as defined, so check those
579 // explicitly. If we know it's a variable, we have a definition for
580 // the purposes of this check.
581 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
582 // FIXME: We would really like to refer back to where the symbol was
583 // first referenced for a source location. We need to add something
584 // to track that. Currently, we just point to the end of the file.
585 PrintMessage(getLexer().getLoc(), SourceMgr::DK_Error,
586 "assembler local symbol '" + Sym->getName() +
592 // Finalize the output stream if there are no errors and if the client wants
594 if (!HadError && !NoFinalize)
600 void AsmParser::CheckForValidSection() {
601 if (!getStreamer().getCurrentSection()) {
602 TokError("expected section directive before assembly directive");
603 Out.SwitchSection(Ctx.getMachOSection(
605 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
606 0, SectionKind::getText()));
610 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
611 void AsmParser::EatToEndOfStatement() {
612 while (Lexer.isNot(AsmToken::EndOfStatement) &&
613 Lexer.isNot(AsmToken::Eof))
617 if (Lexer.is(AsmToken::EndOfStatement))
621 StringRef AsmParser::ParseStringToEndOfStatement() {
622 const char *Start = getTok().getLoc().getPointer();
624 while (Lexer.isNot(AsmToken::EndOfStatement) &&
625 Lexer.isNot(AsmToken::Eof))
628 const char *End = getTok().getLoc().getPointer();
629 return StringRef(Start, End - Start);
632 StringRef AsmParser::ParseStringToComma() {
633 const char *Start = getTok().getLoc().getPointer();
635 while (Lexer.isNot(AsmToken::EndOfStatement) &&
636 Lexer.isNot(AsmToken::Comma) &&
637 Lexer.isNot(AsmToken::Eof))
640 const char *End = getTok().getLoc().getPointer();
641 return StringRef(Start, End - Start);
644 /// ParseParenExpr - Parse a paren expression and return it.
645 /// NOTE: This assumes the leading '(' has already been consumed.
647 /// parenexpr ::= expr)
649 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
650 if (ParseExpression(Res)) return true;
651 if (Lexer.isNot(AsmToken::RParen))
652 return TokError("expected ')' in parentheses expression");
653 EndLoc = Lexer.getLoc();
658 /// ParseBracketExpr - Parse a bracket expression and return it.
659 /// NOTE: This assumes the leading '[' has already been consumed.
661 /// bracketexpr ::= expr]
663 bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
664 if (ParseExpression(Res)) return true;
665 if (Lexer.isNot(AsmToken::RBrac))
666 return TokError("expected ']' in brackets expression");
667 EndLoc = Lexer.getLoc();
672 /// ParsePrimaryExpr - Parse a primary expression and return it.
673 /// primaryexpr ::= (parenexpr
674 /// primaryexpr ::= symbol
675 /// primaryexpr ::= number
676 /// primaryexpr ::= '.'
677 /// primaryexpr ::= ~,+,- primaryexpr
678 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
679 switch (Lexer.getKind()) {
681 return TokError("unknown token in expression");
682 // If we have an error assume that we've already handled it.
683 case AsmToken::Error:
685 case AsmToken::Exclaim:
686 Lex(); // Eat the operator.
687 if (ParsePrimaryExpr(Res, EndLoc))
689 Res = MCUnaryExpr::CreateLNot(Res, getContext());
691 case AsmToken::Dollar:
692 case AsmToken::String:
693 case AsmToken::Identifier: {
694 EndLoc = Lexer.getLoc();
696 StringRef Identifier;
697 if (ParseIdentifier(Identifier))
700 // This is a symbol reference.
701 std::pair<StringRef, StringRef> Split = Identifier.split('@');
702 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
704 // Lookup the symbol variant if used.
705 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
706 if (Split.first.size() != Identifier.size()) {
707 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
708 if (Variant == MCSymbolRefExpr::VK_Invalid) {
709 Variant = MCSymbolRefExpr::VK_None;
710 return TokError("invalid variant '" + Split.second + "'");
714 // If this is an absolute variable reference, substitute it now to preserve
715 // semantics in the face of reassignment.
716 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
718 return Error(EndLoc, "unexpected modifier on variable reference");
720 Res = Sym->getVariableValue();
724 // Otherwise create a symbol ref.
725 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
728 case AsmToken::Integer: {
729 SMLoc Loc = getTok().getLoc();
730 int64_t IntVal = getTok().getIntVal();
731 Res = MCConstantExpr::Create(IntVal, getContext());
732 EndLoc = Lexer.getLoc();
734 // Look for 'b' or 'f' following an Integer as a directional label
735 if (Lexer.getKind() == AsmToken::Identifier) {
736 StringRef IDVal = getTok().getString();
737 if (IDVal == "f" || IDVal == "b"){
738 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
739 IDVal == "f" ? 1 : 0);
740 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
742 if (IDVal == "b" && Sym->isUndefined())
743 return Error(Loc, "invalid reference to undefined symbol");
744 EndLoc = Lexer.getLoc();
745 Lex(); // Eat identifier.
750 case AsmToken::Real: {
751 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
752 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
753 Res = MCConstantExpr::Create(IntVal, getContext());
757 case AsmToken::Dot: {
758 // This is a '.' reference, which references the current PC. Emit a
759 // temporary label to the streamer and refer to it.
760 MCSymbol *Sym = Ctx.CreateTempSymbol();
762 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
763 EndLoc = Lexer.getLoc();
764 Lex(); // Eat identifier.
767 case AsmToken::LParen:
768 Lex(); // Eat the '('.
769 return ParseParenExpr(Res, EndLoc);
770 case AsmToken::LBrac:
771 if (!PlatformParser->HasBracketExpressions())
772 return TokError("brackets expression not supported on this target");
773 Lex(); // Eat the '['.
774 return ParseBracketExpr(Res, EndLoc);
775 case AsmToken::Minus:
776 Lex(); // Eat the operator.
777 if (ParsePrimaryExpr(Res, EndLoc))
779 Res = MCUnaryExpr::CreateMinus(Res, getContext());
782 Lex(); // Eat the operator.
783 if (ParsePrimaryExpr(Res, EndLoc))
785 Res = MCUnaryExpr::CreatePlus(Res, getContext());
787 case AsmToken::Tilde:
788 Lex(); // Eat the operator.
789 if (ParsePrimaryExpr(Res, EndLoc))
791 Res = MCUnaryExpr::CreateNot(Res, getContext());
796 bool AsmParser::ParseExpression(const MCExpr *&Res) {
798 return ParseExpression(Res, EndLoc);
802 AsmParser::ApplyModifierToExpr(const MCExpr *E,
803 MCSymbolRefExpr::VariantKind Variant) {
804 // Recurse over the given expression, rebuilding it to apply the given variant
805 // if there is exactly one symbol.
806 switch (E->getKind()) {
808 case MCExpr::Constant:
811 case MCExpr::SymbolRef: {
812 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
814 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
815 TokError("invalid variant on expression '" +
816 getTok().getIdentifier() + "' (already modified)");
820 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
823 case MCExpr::Unary: {
824 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
825 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
828 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
831 case MCExpr::Binary: {
832 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
833 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
834 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
839 if (!LHS) LHS = BE->getLHS();
840 if (!RHS) RHS = BE->getRHS();
842 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
846 llvm_unreachable("Invalid expression kind!");
849 /// ParseExpression - Parse an expression and return it.
851 /// expr ::= expr &&,|| expr -> lowest.
852 /// expr ::= expr |,^,&,! expr
853 /// expr ::= expr ==,!=,<>,<,<=,>,>= expr
854 /// expr ::= expr <<,>> expr
855 /// expr ::= expr +,- expr
856 /// expr ::= expr *,/,% expr -> highest.
857 /// expr ::= primaryexpr
859 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
860 // Parse the expression.
862 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
865 // As a special case, we support 'a op b @ modifier' by rewriting the
866 // expression to include the modifier. This is inefficient, but in general we
867 // expect users to use 'a@modifier op b'.
868 if (Lexer.getKind() == AsmToken::At) {
871 if (Lexer.isNot(AsmToken::Identifier))
872 return TokError("unexpected symbol modifier following '@'");
874 MCSymbolRefExpr::VariantKind Variant =
875 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
876 if (Variant == MCSymbolRefExpr::VK_Invalid)
877 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
879 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
881 return TokError("invalid modifier '" + getTok().getIdentifier() +
882 "' (no symbols present)");
889 // Try to constant fold it up front, if possible.
891 if (Res->EvaluateAsAbsolute(Value))
892 Res = MCConstantExpr::Create(Value, getContext());
897 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
899 return ParseParenExpr(Res, EndLoc) ||
900 ParseBinOpRHS(1, Res, EndLoc);
903 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
906 SMLoc StartLoc = Lexer.getLoc();
907 if (ParseExpression(Expr))
910 if (!Expr->EvaluateAsAbsolute(Res))
911 return Error(StartLoc, "expected absolute expression");
916 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
917 MCBinaryExpr::Opcode &Kind) {
920 return 0; // not a binop.
922 // Lowest Precedence: &&, ||
923 case AsmToken::AmpAmp:
924 Kind = MCBinaryExpr::LAnd;
926 case AsmToken::PipePipe:
927 Kind = MCBinaryExpr::LOr;
931 // Low Precedence: |, &, ^
933 // FIXME: gas seems to support '!' as an infix operator?
935 Kind = MCBinaryExpr::Or;
937 case AsmToken::Caret:
938 Kind = MCBinaryExpr::Xor;
941 Kind = MCBinaryExpr::And;
944 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
945 case AsmToken::EqualEqual:
946 Kind = MCBinaryExpr::EQ;
948 case AsmToken::ExclaimEqual:
949 case AsmToken::LessGreater:
950 Kind = MCBinaryExpr::NE;
953 Kind = MCBinaryExpr::LT;
955 case AsmToken::LessEqual:
956 Kind = MCBinaryExpr::LTE;
958 case AsmToken::Greater:
959 Kind = MCBinaryExpr::GT;
961 case AsmToken::GreaterEqual:
962 Kind = MCBinaryExpr::GTE;
965 // Intermediate Precedence: <<, >>
966 case AsmToken::LessLess:
967 Kind = MCBinaryExpr::Shl;
969 case AsmToken::GreaterGreater:
970 Kind = MCBinaryExpr::Shr;
973 // High Intermediate Precedence: +, -
975 Kind = MCBinaryExpr::Add;
977 case AsmToken::Minus:
978 Kind = MCBinaryExpr::Sub;
981 // Highest Precedence: *, /, %
983 Kind = MCBinaryExpr::Mul;
985 case AsmToken::Slash:
986 Kind = MCBinaryExpr::Div;
988 case AsmToken::Percent:
989 Kind = MCBinaryExpr::Mod;
995 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
996 /// Res contains the LHS of the expression on input.
997 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
1000 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
1001 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
1003 // If the next token is lower precedence than we are allowed to eat, return
1004 // successfully with what we ate already.
1005 if (TokPrec < Precedence)
1010 // Eat the next primary expression.
1012 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
1014 // If BinOp binds less tightly with RHS than the operator after RHS, let
1015 // the pending operator take RHS as its LHS.
1016 MCBinaryExpr::Opcode Dummy;
1017 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
1018 if (TokPrec < NextTokPrec) {
1019 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
1022 // Merge LHS and RHS according to operator.
1023 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
1031 /// ::= EndOfStatement
1032 /// ::= Label* Directive ...Operands... EndOfStatement
1033 /// ::= Label* Identifier OperandList* EndOfStatement
1034 bool AsmParser::ParseStatement() {
1035 if (Lexer.is(AsmToken::EndOfStatement)) {
1041 // Statements always start with an identifier or are a full line comment.
1042 AsmToken ID = getTok();
1043 SMLoc IDLoc = ID.getLoc();
1045 int64_t LocalLabelVal = -1;
1046 // A full line comment is a '#' as the first token.
1047 if (Lexer.is(AsmToken::Hash))
1048 return ParseCppHashLineFilenameComment(IDLoc);
1050 // Allow an integer followed by a ':' as a directional local label.
1051 if (Lexer.is(AsmToken::Integer)) {
1052 LocalLabelVal = getTok().getIntVal();
1053 if (LocalLabelVal < 0) {
1054 if (!TheCondState.Ignore)
1055 return TokError("unexpected token at start of statement");
1059 IDVal = getTok().getString();
1060 Lex(); // Consume the integer token to be used as an identifier token.
1061 if (Lexer.getKind() != AsmToken::Colon) {
1062 if (!TheCondState.Ignore)
1063 return TokError("unexpected token at start of statement");
1067 } else if (Lexer.is(AsmToken::Dot)) {
1068 // Treat '.' as a valid identifier in this context.
1072 } else if (ParseIdentifier(IDVal)) {
1073 if (!TheCondState.Ignore)
1074 return TokError("unexpected token at start of statement");
1079 // Handle conditional assembly here before checking for skipping. We
1080 // have to do this so that .endif isn't skipped in a ".if 0" block for
1083 return ParseDirectiveIf(IDLoc);
1084 if (IDVal == ".ifb")
1085 return ParseDirectiveIfb(IDLoc, true);
1086 if (IDVal == ".ifnb")
1087 return ParseDirectiveIfb(IDLoc, false);
1088 if (IDVal == ".ifc")
1089 return ParseDirectiveIfc(IDLoc, true);
1090 if (IDVal == ".ifnc")
1091 return ParseDirectiveIfc(IDLoc, false);
1092 if (IDVal == ".ifdef")
1093 return ParseDirectiveIfdef(IDLoc, true);
1094 if (IDVal == ".ifndef" || IDVal == ".ifnotdef")
1095 return ParseDirectiveIfdef(IDLoc, false);
1096 if (IDVal == ".elseif")
1097 return ParseDirectiveElseIf(IDLoc);
1098 if (IDVal == ".else")
1099 return ParseDirectiveElse(IDLoc);
1100 if (IDVal == ".endif")
1101 return ParseDirectiveEndIf(IDLoc);
1103 // If we are in a ".if 0" block, ignore this statement.
1104 if (TheCondState.Ignore) {
1105 EatToEndOfStatement();
1109 // FIXME: Recurse on local labels?
1111 // See what kind of statement we have.
1112 switch (Lexer.getKind()) {
1113 case AsmToken::Colon: {
1114 CheckForValidSection();
1116 // identifier ':' -> Label.
1119 // Diagnose attempt to use '.' as a label.
1121 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
1123 // Diagnose attempt to use a variable as a label.
1125 // FIXME: Diagnostics. Note the location of the definition as a label.
1126 // FIXME: This doesn't diagnose assignment to a symbol which has been
1127 // implicitly marked as external.
1129 if (LocalLabelVal == -1)
1130 Sym = getContext().GetOrCreateSymbol(IDVal);
1132 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
1133 if (!Sym->isUndefined() || Sym->isVariable())
1134 return Error(IDLoc, "invalid symbol redefinition");
1139 // If we are generating dwarf for assembly source files then gather the
1140 // info to make a dwarf label entry for this label if needed.
1141 if (getContext().getGenDwarfForAssembly())
1142 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
1145 // Consume any end of statement token, if present, to avoid spurious
1146 // AddBlankLine calls().
1147 if (Lexer.is(AsmToken::EndOfStatement)) {
1149 if (Lexer.is(AsmToken::Eof))
1153 return ParseStatement();
1156 case AsmToken::Equal:
1157 // identifier '=' ... -> assignment statement
1160 return ParseAssignment(IDVal, true);
1162 default: // Normal instruction or directive.
1166 // If macros are enabled, check to see if this is a macro instantiation.
1168 if (const Macro *M = MacroMap.lookup(IDVal))
1169 return HandleMacroEntry(IDVal, IDLoc, M);
1171 // Otherwise, we have a normal instruction or directive.
1172 if (IDVal[0] == '.' && IDVal != ".") {
1174 // Target hook for parsing target specific directives.
1175 if (!getTargetParser().ParseDirective(ID))
1178 // Assembler features
1179 if (IDVal == ".set" || IDVal == ".equ")
1180 return ParseDirectiveSet(IDVal, true);
1181 if (IDVal == ".equiv")
1182 return ParseDirectiveSet(IDVal, false);
1186 if (IDVal == ".ascii")
1187 return ParseDirectiveAscii(IDVal, false);
1188 if (IDVal == ".asciz" || IDVal == ".string")
1189 return ParseDirectiveAscii(IDVal, true);
1191 if (IDVal == ".byte")
1192 return ParseDirectiveValue(1);
1193 if (IDVal == ".short")
1194 return ParseDirectiveValue(2);
1195 if (IDVal == ".value")
1196 return ParseDirectiveValue(2);
1197 if (IDVal == ".2byte")
1198 return ParseDirectiveValue(2);
1199 if (IDVal == ".long")
1200 return ParseDirectiveValue(4);
1201 if (IDVal == ".int")
1202 return ParseDirectiveValue(4);
1203 if (IDVal == ".4byte")
1204 return ParseDirectiveValue(4);
1205 if (IDVal == ".quad")
1206 return ParseDirectiveValue(8);
1207 if (IDVal == ".8byte")
1208 return ParseDirectiveValue(8);
1209 if (IDVal == ".single" || IDVal == ".float")
1210 return ParseDirectiveRealValue(APFloat::IEEEsingle);
1211 if (IDVal == ".double")
1212 return ParseDirectiveRealValue(APFloat::IEEEdouble);
1214 if (IDVal == ".align") {
1215 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1216 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1218 if (IDVal == ".align32") {
1219 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1220 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1222 if (IDVal == ".balign")
1223 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1224 if (IDVal == ".balignw")
1225 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1226 if (IDVal == ".balignl")
1227 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1228 if (IDVal == ".p2align")
1229 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1230 if (IDVal == ".p2alignw")
1231 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1232 if (IDVal == ".p2alignl")
1233 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1235 if (IDVal == ".org")
1236 return ParseDirectiveOrg();
1238 if (IDVal == ".fill")
1239 return ParseDirectiveFill();
1240 if (IDVal == ".space" || IDVal == ".skip")
1241 return ParseDirectiveSpace();
1242 if (IDVal == ".zero")
1243 return ParseDirectiveZero();
1245 // Symbol attribute directives
1247 if (IDVal == ".extern") {
1248 EatToEndOfStatement(); // .extern is the default, ignore it.
1251 if (IDVal == ".globl" || IDVal == ".global")
1252 return ParseDirectiveSymbolAttribute(MCSA_Global);
1253 if (IDVal == ".indirect_symbol")
1254 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
1255 if (IDVal == ".lazy_reference")
1256 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
1257 if (IDVal == ".no_dead_strip")
1258 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1259 if (IDVal == ".symbol_resolver")
1260 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1261 if (IDVal == ".private_extern")
1262 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1263 if (IDVal == ".reference")
1264 return ParseDirectiveSymbolAttribute(MCSA_Reference);
1265 if (IDVal == ".weak_definition")
1266 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1267 if (IDVal == ".weak_reference")
1268 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
1269 if (IDVal == ".weak_def_can_be_hidden")
1270 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1272 if (IDVal == ".comm" || IDVal == ".common")
1273 return ParseDirectiveComm(/*IsLocal=*/false);
1274 if (IDVal == ".lcomm")
1275 return ParseDirectiveComm(/*IsLocal=*/true);
1277 if (IDVal == ".abort")
1278 return ParseDirectiveAbort();
1279 if (IDVal == ".include")
1280 return ParseDirectiveInclude();
1281 if (IDVal == ".incbin")
1282 return ParseDirectiveIncbin();
1284 if (IDVal == ".code16" || IDVal == ".code16gcc")
1285 return TokError(Twine(IDVal) + " not supported yet");
1287 // Macro-like directives
1288 if (IDVal == ".rept")
1289 return ParseDirectiveRept(IDLoc);
1290 if (IDVal == ".irp")
1291 return ParseDirectiveIrp(IDLoc);
1292 if (IDVal == ".irpc")
1293 return ParseDirectiveIrpc(IDLoc);
1294 if (IDVal == ".endr")
1295 return ParseDirectiveEndr(IDLoc);
1297 // Look up the handler in the handler table.
1298 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1299 DirectiveMap.lookup(IDVal);
1301 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1304 return Error(IDLoc, "unknown directive");
1307 CheckForValidSection();
1309 // Canonicalize the opcode to lower case.
1310 SmallString<128> Opcode;
1311 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1312 Opcode.push_back(tolower(IDVal[i]));
1314 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1315 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1318 // Dump the parsed representation, if requested.
1319 if (getShowParsedOperands()) {
1320 SmallString<256> Str;
1321 raw_svector_ostream OS(Str);
1322 OS << "parsed instruction: [";
1323 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1326 ParsedOperands[i]->print(OS);
1330 PrintMessage(IDLoc, SourceMgr::DK_Note, OS.str());
1333 // If we are generating dwarf for assembly source files and the current
1334 // section is the initial text section then generate a .loc directive for
1336 if (!HadError && getContext().getGenDwarfForAssembly() &&
1337 getContext().getGenDwarfSection() == getStreamer().getCurrentSection() ) {
1338 getStreamer().EmitDwarfLocDirective(getContext().getGenDwarfFileNumber(),
1339 SrcMgr.FindLineNumber(IDLoc, CurBuffer),
1340 0, DWARF2_LINE_DEFAULT_IS_STMT ?
1341 DWARF2_FLAG_IS_STMT : 0, 0, 0,
1345 // If parsing succeeded, match the instruction.
1347 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1350 // Free any parsed operands.
1351 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1352 delete ParsedOperands[i];
1354 // Don't skip the rest of the line, the instruction parser is responsible for
1359 /// EatToEndOfLine uses the Lexer to eat the characters to the end of the line
1360 /// since they may not be able to be tokenized to get to the end of line token.
1361 void AsmParser::EatToEndOfLine() {
1362 if (!Lexer.is(AsmToken::EndOfStatement))
1363 Lexer.LexUntilEndOfLine();
1368 /// ParseCppHashLineFilenameComment as this:
1369 /// ::= # number "filename"
1370 /// or just as a full line comment if it doesn't have a number and a string.
1371 bool AsmParser::ParseCppHashLineFilenameComment(const SMLoc &L) {
1372 Lex(); // Eat the hash token.
1374 if (getLexer().isNot(AsmToken::Integer)) {
1375 // Consume the line since in cases it is not a well-formed line directive,
1376 // as if were simply a full line comment.
1381 int64_t LineNumber = getTok().getIntVal();
1384 if (getLexer().isNot(AsmToken::String)) {
1389 StringRef Filename = getTok().getString();
1390 // Get rid of the enclosing quotes.
1391 Filename = Filename.substr(1, Filename.size()-2);
1393 // Save the SMLoc, Filename and LineNumber for later use by diagnostics.
1395 CppHashFilename = Filename;
1396 CppHashLineNumber = LineNumber;
1398 // Ignore any trailing characters, they're just comment.
1403 /// DiagHandler - will use the last parsed cpp hash line filename comment
1404 /// for the Filename and LineNo if any in the diagnostic.
1405 void AsmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
1406 const AsmParser *Parser = static_cast<const AsmParser*>(Context);
1407 raw_ostream &OS = errs();
1409 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
1410 const SMLoc &DiagLoc = Diag.getLoc();
1411 int DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1412 int CppHashBuf = Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashLoc);
1414 // Like SourceMgr::PrintMessage() we need to print the include stack if any
1415 // before printing the message.
1416 int DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
1417 if (!Parser->SavedDiagHandler && DiagCurBuffer > 0) {
1418 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
1419 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
1422 // If we have not parsed a cpp hash line filename comment or the source
1423 // manager changed or buffer changed (like in a nested include) then just
1424 // print the normal diagnostic using its Filename and LineNo.
1425 if (!Parser->CppHashLineNumber ||
1426 &DiagSrcMgr != &Parser->SrcMgr ||
1427 DiagBuf != CppHashBuf) {
1428 if (Parser->SavedDiagHandler)
1429 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
1435 // Use the CppHashFilename and calculate a line number based on the
1436 // CppHashLoc and CppHashLineNumber relative to this Diag's SMLoc for
1438 const std::string Filename = Parser->CppHashFilename;
1440 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
1441 int CppHashLocLineNo =
1442 Parser->SrcMgr.FindLineNumber(Parser->CppHashLoc, CppHashBuf);
1443 int LineNo = Parser->CppHashLineNumber - 1 +
1444 (DiagLocLineNo - CppHashLocLineNo);
1446 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(),
1447 Filename, LineNo, Diag.getColumnNo(),
1448 Diag.getKind(), Diag.getMessage(),
1449 Diag.getLineContents(), Diag.getRanges());
1451 if (Parser->SavedDiagHandler)
1452 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
1454 NewDiag.print(0, OS);
1457 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1458 const MacroParameters &Parameters,
1459 const MacroArguments &A,
1461 unsigned NParameters = Parameters.size();
1462 if (NParameters != 0 && NParameters != A.size())
1463 return Error(L, "Wrong number of arguments");
1465 while (!Body.empty()) {
1466 // Scan for the next substitution.
1467 std::size_t End = Body.size(), Pos = 0;
1468 for (; Pos != End; ++Pos) {
1469 // Check for a substitution or escape.
1471 // This macro has no parameters, look for $0, $1, etc.
1472 if (Body[Pos] != '$' || Pos + 1 == End)
1475 char Next = Body[Pos + 1];
1476 if (Next == '$' || Next == 'n' || isdigit(Next))
1479 // This macro has parameters, look for \foo, \bar, etc.
1480 if (Body[Pos] == '\\' && Pos + 1 != End)
1486 OS << Body.slice(0, Pos);
1488 // Check if we reached the end.
1493 switch (Body[Pos+1]) {
1499 // $n => number of arguments
1504 // $[0-9] => argument
1506 // Missing arguments are ignored.
1507 unsigned Index = Body[Pos+1] - '0';
1508 if (Index >= A.size())
1511 // Otherwise substitute with the token values, with spaces eliminated.
1512 for (MacroArgument::const_iterator it = A[Index].begin(),
1513 ie = A[Index].end(); it != ie; ++it)
1514 OS << it->getString();
1520 unsigned I = Pos + 1;
1521 while (isalnum(Body[I]) && I + 1 != End)
1524 const char *Begin = Body.data() + Pos +1;
1525 StringRef Argument(Begin, I - (Pos +1));
1527 for (; Index < NParameters; ++Index)
1528 if (Parameters[Index] == Argument)
1531 // FIXME: We should error at the macro definition.
1532 if (Index == NParameters)
1533 return Error(L, "Parameter not found");
1535 for (MacroArgument::const_iterator it = A[Index].begin(),
1536 ie = A[Index].end(); it != ie; ++it)
1537 OS << it->getString();
1539 Pos += 1 + Argument.size();
1541 // Update the scan point.
1542 Body = Body.substr(Pos);
1548 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1550 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL)
1554 /// ParseMacroArgument - Extract AsmTokens for a macro argument.
1555 /// This is used for both default macro parameter values and the
1556 /// arguments in macro invocations
1557 bool AsmParser::ParseMacroArgument(MacroArgument &MA) {
1558 unsigned ParenLevel = 0;
1563 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
1564 return TokError("unexpected token in macro instantiation");
1566 // HandleMacroEntry relies on not advancing the lexer here
1567 // to be able to fill in the remaining default parameter values
1568 if (Lexer.is(AsmToken::EndOfStatement))
1570 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma))
1573 // Adjust the current parentheses level.
1574 if (Lexer.is(AsmToken::LParen))
1576 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1579 // Append the token to the current argument list.
1580 MA.push_back(getTok());
1583 if (ParenLevel != 0)
1584 return TokError("unbalanced parenthesises in macro argument");
1588 // Parse the macro instantiation arguments.
1589 bool AsmParser::ParseMacroArguments(const Macro *M, MacroArguments &A) {
1590 const unsigned NParameters = M ? M->Parameters.size() : 0;
1592 // Parse two kinds of macro invocations:
1593 // - macros defined without any parameters accept an arbitrary number of them
1594 // - macros defined with parameters accept at most that many of them
1595 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
1599 if (ParseMacroArgument(MA))
1604 if (Lexer.is(AsmToken::EndOfStatement))
1607 if (Lexer.is(AsmToken::Comma))
1610 return TokError("Too many arguments");
1613 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1615 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1616 // this, although we should protect against infinite loops.
1617 if (ActiveMacros.size() == 20)
1618 return TokError("macros cannot be nested more than 20 levels deep");
1621 if (ParseMacroArguments(M, A))
1624 // Remove any trailing empty arguments. Do this after-the-fact as we have
1625 // to keep empty arguments in the middle of the list or positionality
1626 // gets off. e.g., "foo 1, , 2" vs. "foo 1, 2,"
1627 while (!A.empty() && A.back().empty())
1630 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1631 // to hold the macro body with substitutions.
1632 SmallString<256> Buf;
1633 StringRef Body = M->Body;
1634 raw_svector_ostream OS(Buf);
1636 if (expandMacro(OS, Body, M->Parameters, A, getTok().getLoc()))
1639 // We include the .endmacro in the buffer as our queue to exit the macro
1641 OS << ".endmacro\n";
1643 MemoryBuffer *Instantiation =
1644 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1646 // Create the macro instantiation object and add to the current macro
1647 // instantiation stack.
1648 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1651 ActiveMacros.push_back(MI);
1653 // Jump to the macro instantiation and prime the lexer.
1654 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1655 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1661 void AsmParser::HandleMacroExit() {
1662 // Jump to the EndOfStatement we should return to, and consume it.
1663 JumpToLoc(ActiveMacros.back()->ExitLoc);
1666 // Pop the instantiation entry.
1667 delete ActiveMacros.back();
1668 ActiveMacros.pop_back();
1671 static bool IsUsedIn(const MCSymbol *Sym, const MCExpr *Value) {
1672 switch (Value->getKind()) {
1673 case MCExpr::Binary: {
1674 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr*>(Value);
1675 return IsUsedIn(Sym, BE->getLHS()) || IsUsedIn(Sym, BE->getRHS());
1678 case MCExpr::Target:
1679 case MCExpr::Constant:
1681 case MCExpr::SymbolRef: {
1682 const MCSymbol &S = static_cast<const MCSymbolRefExpr*>(Value)->getSymbol();
1684 return IsUsedIn(Sym, S.getVariableValue());
1688 return IsUsedIn(Sym, static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1691 llvm_unreachable("Unknown expr kind!");
1694 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) {
1695 // FIXME: Use better location, we should use proper tokens.
1696 SMLoc EqualLoc = Lexer.getLoc();
1698 const MCExpr *Value;
1699 if (ParseExpression(Value))
1702 // Note: we don't count b as used in "a = b". This is to allow
1706 if (Lexer.isNot(AsmToken::EndOfStatement))
1707 return TokError("unexpected token in assignment");
1709 // Error on assignment to '.'.
1711 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported "
1712 "(use '.space' or '.org').)"));
1715 // Eat the end of statement marker.
1718 // Validate that the LHS is allowed to be a variable (either it has not been
1719 // used as a symbol, or it is an absolute symbol).
1720 MCSymbol *Sym = getContext().LookupSymbol(Name);
1722 // Diagnose assignment to a label.
1724 // FIXME: Diagnostics. Note the location of the definition as a label.
1725 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1726 if (IsUsedIn(Sym, Value))
1727 return Error(EqualLoc, "Recursive use of '" + Name + "'");
1728 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
1729 ; // Allow redefinitions of undefined symbols only used in directives.
1730 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
1731 ; // Allow redefinitions of variables that haven't yet been used.
1732 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
1733 return Error(EqualLoc, "redefinition of '" + Name + "'");
1734 else if (!Sym->isVariable())
1735 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1736 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1737 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1740 // Don't count these checks as uses.
1741 Sym->setUsed(false);
1743 Sym = getContext().GetOrCreateSymbol(Name);
1745 // FIXME: Handle '.'.
1747 // Do the assignment.
1748 Out.EmitAssignment(Sym, Value);
1753 /// ParseIdentifier:
1756 bool AsmParser::ParseIdentifier(StringRef &Res) {
1757 // The assembler has relaxed rules for accepting identifiers, in particular we
1758 // allow things like '.globl $foo', which would normally be separate
1759 // tokens. At this level, we have already lexed so we cannot (currently)
1760 // handle this as a context dependent token, instead we detect adjacent tokens
1761 // and return the combined identifier.
1762 if (Lexer.is(AsmToken::Dollar)) {
1763 SMLoc DollarLoc = getLexer().getLoc();
1765 // Consume the dollar sign, and check for a following identifier.
1767 if (Lexer.isNot(AsmToken::Identifier))
1770 // We have a '$' followed by an identifier, make sure they are adjacent.
1771 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1774 // Construct the joined identifier and consume the token.
1775 Res = StringRef(DollarLoc.getPointer(),
1776 getTok().getIdentifier().size() + 1);
1781 if (Lexer.isNot(AsmToken::Identifier) &&
1782 Lexer.isNot(AsmToken::String))
1785 Res = getTok().getIdentifier();
1787 Lex(); // Consume the identifier token.
1792 /// ParseDirectiveSet:
1793 /// ::= .equ identifier ',' expression
1794 /// ::= .equiv identifier ',' expression
1795 /// ::= .set identifier ',' expression
1796 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1799 if (ParseIdentifier(Name))
1800 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1802 if (getLexer().isNot(AsmToken::Comma))
1803 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1806 return ParseAssignment(Name, allow_redef);
1809 bool AsmParser::ParseEscapedString(std::string &Data) {
1810 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1813 StringRef Str = getTok().getStringContents();
1814 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1815 if (Str[i] != '\\') {
1820 // Recognize escaped characters. Note that this escape semantics currently
1821 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1824 return TokError("unexpected backslash at end of string");
1826 // Recognize octal sequences.
1827 if ((unsigned) (Str[i] - '0') <= 7) {
1828 // Consume up to three octal characters.
1829 unsigned Value = Str[i] - '0';
1831 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1833 Value = Value * 8 + (Str[i] - '0');
1835 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1837 Value = Value * 8 + (Str[i] - '0');
1842 return TokError("invalid octal escape sequence (out of range)");
1844 Data += (unsigned char) Value;
1848 // Otherwise recognize individual escapes.
1851 // Just reject invalid escape sequences for now.
1852 return TokError("invalid escape sequence (unrecognized character)");
1854 case 'b': Data += '\b'; break;
1855 case 'f': Data += '\f'; break;
1856 case 'n': Data += '\n'; break;
1857 case 'r': Data += '\r'; break;
1858 case 't': Data += '\t'; break;
1859 case '"': Data += '"'; break;
1860 case '\\': Data += '\\'; break;
1867 /// ParseDirectiveAscii:
1868 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1869 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1870 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1871 CheckForValidSection();
1874 if (getLexer().isNot(AsmToken::String))
1875 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1878 if (ParseEscapedString(Data))
1881 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1883 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1887 if (getLexer().is(AsmToken::EndOfStatement))
1890 if (getLexer().isNot(AsmToken::Comma))
1891 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1900 /// ParseDirectiveValue
1901 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1902 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1903 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1904 CheckForValidSection();
1907 const MCExpr *Value;
1908 SMLoc ExprLoc = getLexer().getLoc();
1909 if (ParseExpression(Value))
1912 // Special case constant expressions to match code generator.
1913 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
1914 assert(Size <= 8 && "Invalid size");
1915 uint64_t IntValue = MCE->getValue();
1916 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
1917 return Error(ExprLoc, "literal value out of range for directive");
1918 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE);
1920 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1922 if (getLexer().is(AsmToken::EndOfStatement))
1925 // FIXME: Improve diagnostic.
1926 if (getLexer().isNot(AsmToken::Comma))
1927 return TokError("unexpected token in directive");
1936 /// ParseDirectiveRealValue
1937 /// ::= (.single | .double) [ expression (, expression)* ]
1938 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1939 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1940 CheckForValidSection();
1943 // We don't truly support arithmetic on floating point expressions, so we
1944 // have to manually parse unary prefixes.
1946 if (getLexer().is(AsmToken::Minus)) {
1949 } else if (getLexer().is(AsmToken::Plus))
1952 if (getLexer().isNot(AsmToken::Integer) &&
1953 getLexer().isNot(AsmToken::Real) &&
1954 getLexer().isNot(AsmToken::Identifier))
1955 return TokError("unexpected token in directive");
1957 // Convert to an APFloat.
1958 APFloat Value(Semantics);
1959 StringRef IDVal = getTok().getString();
1960 if (getLexer().is(AsmToken::Identifier)) {
1961 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
1962 Value = APFloat::getInf(Semantics);
1963 else if (!IDVal.compare_lower("nan"))
1964 Value = APFloat::getNaN(Semantics, false, ~0);
1966 return TokError("invalid floating point literal");
1967 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
1968 APFloat::opInvalidOp)
1969 return TokError("invalid floating point literal");
1973 // Consume the numeric token.
1976 // Emit the value as an integer.
1977 APInt AsInt = Value.bitcastToAPInt();
1978 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1979 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1981 if (getLexer().is(AsmToken::EndOfStatement))
1984 if (getLexer().isNot(AsmToken::Comma))
1985 return TokError("unexpected token in directive");
1994 /// ParseDirectiveSpace
1995 /// ::= .space expression [ , expression ]
1996 bool AsmParser::ParseDirectiveSpace() {
1997 CheckForValidSection();
2000 if (ParseAbsoluteExpression(NumBytes))
2003 int64_t FillExpr = 0;
2004 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2005 if (getLexer().isNot(AsmToken::Comma))
2006 return TokError("unexpected token in '.space' directive");
2009 if (ParseAbsoluteExpression(FillExpr))
2012 if (getLexer().isNot(AsmToken::EndOfStatement))
2013 return TokError("unexpected token in '.space' directive");
2019 return TokError("invalid number of bytes in '.space' directive");
2021 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
2022 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
2027 /// ParseDirectiveZero
2028 /// ::= .zero expression
2029 bool AsmParser::ParseDirectiveZero() {
2030 CheckForValidSection();
2033 if (ParseAbsoluteExpression(NumBytes))
2037 if (getLexer().is(AsmToken::Comma)) {
2039 if (ParseAbsoluteExpression(Val))
2043 if (getLexer().isNot(AsmToken::EndOfStatement))
2044 return TokError("unexpected token in '.zero' directive");
2048 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
2053 /// ParseDirectiveFill
2054 /// ::= .fill expression , expression , expression
2055 bool AsmParser::ParseDirectiveFill() {
2056 CheckForValidSection();
2059 if (ParseAbsoluteExpression(NumValues))
2062 if (getLexer().isNot(AsmToken::Comma))
2063 return TokError("unexpected token in '.fill' directive");
2067 if (ParseAbsoluteExpression(FillSize))
2070 if (getLexer().isNot(AsmToken::Comma))
2071 return TokError("unexpected token in '.fill' directive");
2075 if (ParseAbsoluteExpression(FillExpr))
2078 if (getLexer().isNot(AsmToken::EndOfStatement))
2079 return TokError("unexpected token in '.fill' directive");
2083 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
2084 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
2086 for (uint64_t i = 0, e = NumValues; i != e; ++i)
2087 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
2092 /// ParseDirectiveOrg
2093 /// ::= .org expression [ , expression ]
2094 bool AsmParser::ParseDirectiveOrg() {
2095 CheckForValidSection();
2097 const MCExpr *Offset;
2098 SMLoc Loc = getTok().getLoc();
2099 if (ParseExpression(Offset))
2102 // Parse optional fill expression.
2103 int64_t FillExpr = 0;
2104 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2105 if (getLexer().isNot(AsmToken::Comma))
2106 return TokError("unexpected token in '.org' directive");
2109 if (ParseAbsoluteExpression(FillExpr))
2112 if (getLexer().isNot(AsmToken::EndOfStatement))
2113 return TokError("unexpected token in '.org' directive");
2118 // Only limited forms of relocatable expressions are accepted here, it
2119 // has to be relative to the current section. The streamer will return
2120 // 'true' if the expression wasn't evaluatable.
2121 if (getStreamer().EmitValueToOffset(Offset, FillExpr))
2122 return Error(Loc, "expected assembly-time absolute expression");
2127 /// ParseDirectiveAlign
2128 /// ::= {.align, ...} expression [ , expression [ , expression ]]
2129 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2130 CheckForValidSection();
2132 SMLoc AlignmentLoc = getLexer().getLoc();
2134 if (ParseAbsoluteExpression(Alignment))
2138 bool HasFillExpr = false;
2139 int64_t FillExpr = 0;
2140 int64_t MaxBytesToFill = 0;
2141 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2142 if (getLexer().isNot(AsmToken::Comma))
2143 return TokError("unexpected token in directive");
2146 // The fill expression can be omitted while specifying a maximum number of
2147 // alignment bytes, e.g:
2149 if (getLexer().isNot(AsmToken::Comma)) {
2151 if (ParseAbsoluteExpression(FillExpr))
2155 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2156 if (getLexer().isNot(AsmToken::Comma))
2157 return TokError("unexpected token in directive");
2160 MaxBytesLoc = getLexer().getLoc();
2161 if (ParseAbsoluteExpression(MaxBytesToFill))
2164 if (getLexer().isNot(AsmToken::EndOfStatement))
2165 return TokError("unexpected token in directive");
2174 // Compute alignment in bytes.
2176 // FIXME: Diagnose overflow.
2177 if (Alignment >= 32) {
2178 Error(AlignmentLoc, "invalid alignment value");
2182 Alignment = 1ULL << Alignment;
2185 // Diagnose non-sensical max bytes to align.
2186 if (MaxBytesLoc.isValid()) {
2187 if (MaxBytesToFill < 1) {
2188 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2189 "many bytes, ignoring maximum bytes expression");
2193 if (MaxBytesToFill >= Alignment) {
2194 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2200 // Check whether we should use optimal code alignment for this .align
2202 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
2203 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2204 ValueSize == 1 && UseCodeAlign) {
2205 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2207 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2208 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2215 /// ParseDirectiveSymbolAttribute
2216 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
2217 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
2218 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2221 SMLoc Loc = getTok().getLoc();
2223 if (ParseIdentifier(Name))
2224 return Error(Loc, "expected identifier in directive");
2226 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2228 // Assembler local symbols don't make any sense here. Complain loudly.
2229 if (Sym->isTemporary())
2230 return Error(Loc, "non-local symbol required in directive");
2232 getStreamer().EmitSymbolAttribute(Sym, Attr);
2234 if (getLexer().is(AsmToken::EndOfStatement))
2237 if (getLexer().isNot(AsmToken::Comma))
2238 return TokError("unexpected token in directive");
2247 /// ParseDirectiveComm
2248 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
2249 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
2250 CheckForValidSection();
2252 SMLoc IDLoc = getLexer().getLoc();
2254 if (ParseIdentifier(Name))
2255 return TokError("expected identifier in directive");
2257 // Handle the identifier as the key symbol.
2258 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2260 if (getLexer().isNot(AsmToken::Comma))
2261 return TokError("unexpected token in directive");
2265 SMLoc SizeLoc = getLexer().getLoc();
2266 if (ParseAbsoluteExpression(Size))
2269 int64_t Pow2Alignment = 0;
2270 SMLoc Pow2AlignmentLoc;
2271 if (getLexer().is(AsmToken::Comma)) {
2273 Pow2AlignmentLoc = getLexer().getLoc();
2274 if (ParseAbsoluteExpression(Pow2Alignment))
2277 // If this target takes alignments in bytes (not log) validate and convert.
2278 if (Lexer.getMAI().getAlignmentIsInBytes()) {
2279 if (!isPowerOf2_64(Pow2Alignment))
2280 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
2281 Pow2Alignment = Log2_64(Pow2Alignment);
2285 if (getLexer().isNot(AsmToken::EndOfStatement))
2286 return TokError("unexpected token in '.comm' or '.lcomm' directive");
2290 // NOTE: a size of zero for a .comm should create a undefined symbol
2291 // but a size of .lcomm creates a bss symbol of size zero.
2293 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
2294 "be less than zero");
2296 // NOTE: The alignment in the directive is a power of 2 value, the assembler
2297 // may internally end up wanting an alignment in bytes.
2298 // FIXME: Diagnose overflow.
2299 if (Pow2Alignment < 0)
2300 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
2301 "alignment, can't be less than zero");
2303 if (!Sym->isUndefined())
2304 return Error(IDLoc, "invalid symbol redefinition");
2306 // '.lcomm' is equivalent to '.zerofill'.
2307 // Create the Symbol as a common or local common with Size and Pow2Alignment
2309 getStreamer().EmitZerofill(Ctx.getMachOSection(
2310 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
2311 0, SectionKind::getBSS()),
2312 Sym, Size, 1 << Pow2Alignment);
2316 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
2320 /// ParseDirectiveAbort
2321 /// ::= .abort [... message ...]
2322 bool AsmParser::ParseDirectiveAbort() {
2323 // FIXME: Use loc from directive.
2324 SMLoc Loc = getLexer().getLoc();
2326 StringRef Str = ParseStringToEndOfStatement();
2327 if (getLexer().isNot(AsmToken::EndOfStatement))
2328 return TokError("unexpected token in '.abort' directive");
2333 Error(Loc, ".abort detected. Assembly stopping.");
2335 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
2336 // FIXME: Actually abort assembly here.
2341 /// ParseDirectiveInclude
2342 /// ::= .include "filename"
2343 bool AsmParser::ParseDirectiveInclude() {
2344 if (getLexer().isNot(AsmToken::String))
2345 return TokError("expected string in '.include' directive");
2347 std::string Filename = getTok().getString();
2348 SMLoc IncludeLoc = getLexer().getLoc();
2351 if (getLexer().isNot(AsmToken::EndOfStatement))
2352 return TokError("unexpected token in '.include' directive");
2354 // Strip the quotes.
2355 Filename = Filename.substr(1, Filename.size()-2);
2357 // Attempt to switch the lexer to the included file before consuming the end
2358 // of statement to avoid losing it when we switch.
2359 if (EnterIncludeFile(Filename)) {
2360 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
2367 /// ParseDirectiveIncbin
2368 /// ::= .incbin "filename"
2369 bool AsmParser::ParseDirectiveIncbin() {
2370 if (getLexer().isNot(AsmToken::String))
2371 return TokError("expected string in '.incbin' directive");
2373 std::string Filename = getTok().getString();
2374 SMLoc IncbinLoc = getLexer().getLoc();
2377 if (getLexer().isNot(AsmToken::EndOfStatement))
2378 return TokError("unexpected token in '.incbin' directive");
2380 // Strip the quotes.
2381 Filename = Filename.substr(1, Filename.size()-2);
2383 // Attempt to process the included file.
2384 if (ProcessIncbinFile(Filename)) {
2385 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
2392 /// ParseDirectiveIf
2393 /// ::= .if expression
2394 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
2395 TheCondStack.push_back(TheCondState);
2396 TheCondState.TheCond = AsmCond::IfCond;
2397 if (TheCondState.Ignore) {
2398 EatToEndOfStatement();
2401 if (ParseAbsoluteExpression(ExprValue))
2404 if (getLexer().isNot(AsmToken::EndOfStatement))
2405 return TokError("unexpected token in '.if' directive");
2409 TheCondState.CondMet = ExprValue;
2410 TheCondState.Ignore = !TheCondState.CondMet;
2416 /// ParseDirectiveIfb
2418 bool AsmParser::ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
2419 TheCondStack.push_back(TheCondState);
2420 TheCondState.TheCond = AsmCond::IfCond;
2422 if (TheCondState.Ignore) {
2423 EatToEndOfStatement();
2425 StringRef Str = ParseStringToEndOfStatement();
2427 if (getLexer().isNot(AsmToken::EndOfStatement))
2428 return TokError("unexpected token in '.ifb' directive");
2432 TheCondState.CondMet = ExpectBlank == Str.empty();
2433 TheCondState.Ignore = !TheCondState.CondMet;
2439 /// ParseDirectiveIfc
2440 /// ::= .ifc string1, string2
2441 bool AsmParser::ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
2442 TheCondStack.push_back(TheCondState);
2443 TheCondState.TheCond = AsmCond::IfCond;
2445 if (TheCondState.Ignore) {
2446 EatToEndOfStatement();
2448 StringRef Str1 = ParseStringToComma();
2450 if (getLexer().isNot(AsmToken::Comma))
2451 return TokError("unexpected token in '.ifc' directive");
2455 StringRef Str2 = ParseStringToEndOfStatement();
2457 if (getLexer().isNot(AsmToken::EndOfStatement))
2458 return TokError("unexpected token in '.ifc' directive");
2462 TheCondState.CondMet = ExpectEqual == (Str1 == Str2);
2463 TheCondState.Ignore = !TheCondState.CondMet;
2469 /// ParseDirectiveIfdef
2470 /// ::= .ifdef symbol
2471 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
2473 TheCondStack.push_back(TheCondState);
2474 TheCondState.TheCond = AsmCond::IfCond;
2476 if (TheCondState.Ignore) {
2477 EatToEndOfStatement();
2479 if (ParseIdentifier(Name))
2480 return TokError("expected identifier after '.ifdef'");
2484 MCSymbol *Sym = getContext().LookupSymbol(Name);
2487 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
2489 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
2490 TheCondState.Ignore = !TheCondState.CondMet;
2496 /// ParseDirectiveElseIf
2497 /// ::= .elseif expression
2498 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
2499 if (TheCondState.TheCond != AsmCond::IfCond &&
2500 TheCondState.TheCond != AsmCond::ElseIfCond)
2501 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
2503 TheCondState.TheCond = AsmCond::ElseIfCond;
2505 bool LastIgnoreState = false;
2506 if (!TheCondStack.empty())
2507 LastIgnoreState = TheCondStack.back().Ignore;
2508 if (LastIgnoreState || TheCondState.CondMet) {
2509 TheCondState.Ignore = true;
2510 EatToEndOfStatement();
2514 if (ParseAbsoluteExpression(ExprValue))
2517 if (getLexer().isNot(AsmToken::EndOfStatement))
2518 return TokError("unexpected token in '.elseif' directive");
2521 TheCondState.CondMet = ExprValue;
2522 TheCondState.Ignore = !TheCondState.CondMet;
2528 /// ParseDirectiveElse
2530 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2531 if (getLexer().isNot(AsmToken::EndOfStatement))
2532 return TokError("unexpected token in '.else' directive");
2536 if (TheCondState.TheCond != AsmCond::IfCond &&
2537 TheCondState.TheCond != AsmCond::ElseIfCond)
2538 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
2540 TheCondState.TheCond = AsmCond::ElseCond;
2541 bool LastIgnoreState = false;
2542 if (!TheCondStack.empty())
2543 LastIgnoreState = TheCondStack.back().Ignore;
2544 if (LastIgnoreState || TheCondState.CondMet)
2545 TheCondState.Ignore = true;
2547 TheCondState.Ignore = false;
2552 /// ParseDirectiveEndIf
2554 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2555 if (getLexer().isNot(AsmToken::EndOfStatement))
2556 return TokError("unexpected token in '.endif' directive");
2560 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2561 TheCondStack.empty())
2562 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2564 if (!TheCondStack.empty()) {
2565 TheCondState = TheCondStack.back();
2566 TheCondStack.pop_back();
2572 /// ParseDirectiveFile
2573 /// ::= .file [number] filename
2574 /// ::= .file number directory filename
2575 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
2576 // FIXME: I'm not sure what this is.
2577 int64_t FileNumber = -1;
2578 SMLoc FileNumberLoc = getLexer().getLoc();
2579 if (getLexer().is(AsmToken::Integer)) {
2580 FileNumber = getTok().getIntVal();
2584 return TokError("file number less than one");
2587 if (getLexer().isNot(AsmToken::String))
2588 return TokError("unexpected token in '.file' directive");
2590 // Usually the directory and filename together, otherwise just the directory.
2591 StringRef Path = getTok().getString();
2592 Path = Path.substr(1, Path.size()-2);
2595 StringRef Directory;
2597 if (getLexer().is(AsmToken::String)) {
2598 if (FileNumber == -1)
2599 return TokError("explicit path specified, but no file number");
2600 Filename = getTok().getString();
2601 Filename = Filename.substr(1, Filename.size()-2);
2608 if (getLexer().isNot(AsmToken::EndOfStatement))
2609 return TokError("unexpected token in '.file' directive");
2611 if (FileNumber == -1)
2612 getStreamer().EmitFileDirective(Filename);
2614 if (getContext().getGenDwarfForAssembly() == true)
2615 Error(DirectiveLoc, "input can't have .file dwarf directives when -g is "
2616 "used to generate dwarf debug info for assembly code");
2618 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename))
2619 Error(FileNumberLoc, "file number already allocated");
2625 /// ParseDirectiveLine
2626 /// ::= .line [number]
2627 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2628 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2629 if (getLexer().isNot(AsmToken::Integer))
2630 return TokError("unexpected token in '.line' directive");
2632 int64_t LineNumber = getTok().getIntVal();
2636 // FIXME: Do something with the .line.
2639 if (getLexer().isNot(AsmToken::EndOfStatement))
2640 return TokError("unexpected token in '.line' directive");
2646 /// ParseDirectiveLoc
2647 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2648 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2649 /// The first number is a file number, must have been previously assigned with
2650 /// a .file directive, the second number is the line number and optionally the
2651 /// third number is a column position (zero if not specified). The remaining
2652 /// optional items are .loc sub-directives.
2653 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2655 if (getLexer().isNot(AsmToken::Integer))
2656 return TokError("unexpected token in '.loc' directive");
2657 int64_t FileNumber = getTok().getIntVal();
2659 return TokError("file number less than one in '.loc' directive");
2660 if (!getContext().isValidDwarfFileNumber(FileNumber))
2661 return TokError("unassigned file number in '.loc' directive");
2664 int64_t LineNumber = 0;
2665 if (getLexer().is(AsmToken::Integer)) {
2666 LineNumber = getTok().getIntVal();
2668 return TokError("line number less than one in '.loc' directive");
2672 int64_t ColumnPos = 0;
2673 if (getLexer().is(AsmToken::Integer)) {
2674 ColumnPos = getTok().getIntVal();
2676 return TokError("column position less than zero in '.loc' directive");
2680 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2682 int64_t Discriminator = 0;
2683 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2685 if (getLexer().is(AsmToken::EndOfStatement))
2689 SMLoc Loc = getTok().getLoc();
2690 if (getParser().ParseIdentifier(Name))
2691 return TokError("unexpected token in '.loc' directive");
2693 if (Name == "basic_block")
2694 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2695 else if (Name == "prologue_end")
2696 Flags |= DWARF2_FLAG_PROLOGUE_END;
2697 else if (Name == "epilogue_begin")
2698 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2699 else if (Name == "is_stmt") {
2700 SMLoc Loc = getTok().getLoc();
2701 const MCExpr *Value;
2702 if (getParser().ParseExpression(Value))
2704 // The expression must be the constant 0 or 1.
2705 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2706 int Value = MCE->getValue();
2708 Flags &= ~DWARF2_FLAG_IS_STMT;
2709 else if (Value == 1)
2710 Flags |= DWARF2_FLAG_IS_STMT;
2712 return Error(Loc, "is_stmt value not 0 or 1");
2715 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2718 else if (Name == "isa") {
2719 SMLoc Loc = getTok().getLoc();
2720 const MCExpr *Value;
2721 if (getParser().ParseExpression(Value))
2723 // The expression must be a constant greater or equal to 0.
2724 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2725 int Value = MCE->getValue();
2727 return Error(Loc, "isa number less than zero");
2731 return Error(Loc, "isa number not a constant value");
2734 else if (Name == "discriminator") {
2735 if (getParser().ParseAbsoluteExpression(Discriminator))
2739 return Error(Loc, "unknown sub-directive in '.loc' directive");
2742 if (getLexer().is(AsmToken::EndOfStatement))
2747 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2748 Isa, Discriminator, StringRef());
2753 /// ParseDirectiveStabs
2754 /// ::= .stabs string, number, number, number
2755 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2756 SMLoc DirectiveLoc) {
2757 return TokError("unsupported directive '" + Directive + "'");
2760 /// ParseDirectiveCFISections
2761 /// ::= .cfi_sections section [, section]
2762 bool GenericAsmParser::ParseDirectiveCFISections(StringRef,
2763 SMLoc DirectiveLoc) {
2768 if (getParser().ParseIdentifier(Name))
2769 return TokError("Expected an identifier");
2771 if (Name == ".eh_frame")
2773 else if (Name == ".debug_frame")
2776 if (getLexer().is(AsmToken::Comma)) {
2779 if (getParser().ParseIdentifier(Name))
2780 return TokError("Expected an identifier");
2782 if (Name == ".eh_frame")
2784 else if (Name == ".debug_frame")
2788 getStreamer().EmitCFISections(EH, Debug);
2793 /// ParseDirectiveCFIStartProc
2794 /// ::= .cfi_startproc
2795 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2796 SMLoc DirectiveLoc) {
2797 getStreamer().EmitCFIStartProc();
2801 /// ParseDirectiveCFIEndProc
2802 /// ::= .cfi_endproc
2803 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2804 getStreamer().EmitCFIEndProc();
2808 /// ParseRegisterOrRegisterNumber - parse register name or number.
2809 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2810 SMLoc DirectiveLoc) {
2813 if (getLexer().isNot(AsmToken::Integer)) {
2814 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2817 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true);
2819 return getParser().ParseAbsoluteExpression(Register);
2824 /// ParseDirectiveCFIDefCfa
2825 /// ::= .cfi_def_cfa register, offset
2826 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
2827 SMLoc DirectiveLoc) {
2828 int64_t Register = 0;
2829 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2832 if (getLexer().isNot(AsmToken::Comma))
2833 return TokError("unexpected token in directive");
2837 if (getParser().ParseAbsoluteExpression(Offset))
2840 getStreamer().EmitCFIDefCfa(Register, Offset);
2844 /// ParseDirectiveCFIDefCfaOffset
2845 /// ::= .cfi_def_cfa_offset offset
2846 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
2847 SMLoc DirectiveLoc) {
2849 if (getParser().ParseAbsoluteExpression(Offset))
2852 getStreamer().EmitCFIDefCfaOffset(Offset);
2856 /// ParseDirectiveCFIAdjustCfaOffset
2857 /// ::= .cfi_adjust_cfa_offset adjustment
2858 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef,
2859 SMLoc DirectiveLoc) {
2860 int64_t Adjustment = 0;
2861 if (getParser().ParseAbsoluteExpression(Adjustment))
2864 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
2868 /// ParseDirectiveCFIDefCfaRegister
2869 /// ::= .cfi_def_cfa_register register
2870 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
2871 SMLoc DirectiveLoc) {
2872 int64_t Register = 0;
2873 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2876 getStreamer().EmitCFIDefCfaRegister(Register);
2880 /// ParseDirectiveCFIOffset
2881 /// ::= .cfi_offset register, offset
2882 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
2883 int64_t Register = 0;
2886 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2889 if (getLexer().isNot(AsmToken::Comma))
2890 return TokError("unexpected token in directive");
2893 if (getParser().ParseAbsoluteExpression(Offset))
2896 getStreamer().EmitCFIOffset(Register, Offset);
2900 /// ParseDirectiveCFIRelOffset
2901 /// ::= .cfi_rel_offset register, offset
2902 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef,
2903 SMLoc DirectiveLoc) {
2904 int64_t Register = 0;
2906 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2909 if (getLexer().isNot(AsmToken::Comma))
2910 return TokError("unexpected token in directive");
2914 if (getParser().ParseAbsoluteExpression(Offset))
2917 getStreamer().EmitCFIRelOffset(Register, Offset);
2921 static bool isValidEncoding(int64_t Encoding) {
2922 if (Encoding & ~0xff)
2925 if (Encoding == dwarf::DW_EH_PE_omit)
2928 const unsigned Format = Encoding & 0xf;
2929 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
2930 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
2931 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
2932 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
2935 const unsigned Application = Encoding & 0x70;
2936 if (Application != dwarf::DW_EH_PE_absptr &&
2937 Application != dwarf::DW_EH_PE_pcrel)
2943 /// ParseDirectiveCFIPersonalityOrLsda
2944 /// ::= .cfi_personality encoding, [symbol_name]
2945 /// ::= .cfi_lsda encoding, [symbol_name]
2946 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
2947 SMLoc DirectiveLoc) {
2948 int64_t Encoding = 0;
2949 if (getParser().ParseAbsoluteExpression(Encoding))
2951 if (Encoding == dwarf::DW_EH_PE_omit)
2954 if (!isValidEncoding(Encoding))
2955 return TokError("unsupported encoding.");
2957 if (getLexer().isNot(AsmToken::Comma))
2958 return TokError("unexpected token in directive");
2962 if (getParser().ParseIdentifier(Name))
2963 return TokError("expected identifier in directive");
2965 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2967 if (IDVal == ".cfi_personality")
2968 getStreamer().EmitCFIPersonality(Sym, Encoding);
2970 assert(IDVal == ".cfi_lsda");
2971 getStreamer().EmitCFILsda(Sym, Encoding);
2976 /// ParseDirectiveCFIRememberState
2977 /// ::= .cfi_remember_state
2978 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
2979 SMLoc DirectiveLoc) {
2980 getStreamer().EmitCFIRememberState();
2984 /// ParseDirectiveCFIRestoreState
2985 /// ::= .cfi_remember_state
2986 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
2987 SMLoc DirectiveLoc) {
2988 getStreamer().EmitCFIRestoreState();
2992 /// ParseDirectiveCFISameValue
2993 /// ::= .cfi_same_value register
2994 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal,
2995 SMLoc DirectiveLoc) {
2996 int64_t Register = 0;
2998 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3001 getStreamer().EmitCFISameValue(Register);
3006 /// ParseDirectiveCFIRestore
3007 /// ::= .cfi_restore register
3008 bool GenericAsmParser::ParseDirectiveCFIRestore(StringRef IDVal,
3009 SMLoc DirectiveLoc) {
3010 int64_t Register = 0;
3011 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3014 getStreamer().EmitCFIRestore(Register);
3019 /// ParseDirectiveCFIEscape
3020 /// ::= .cfi_escape expression[,...]
3021 bool GenericAsmParser::ParseDirectiveCFIEscape(StringRef IDVal,
3022 SMLoc DirectiveLoc) {
3025 if (getParser().ParseAbsoluteExpression(CurrValue))
3028 Values.push_back((uint8_t)CurrValue);
3030 while (getLexer().is(AsmToken::Comma)) {
3033 if (getParser().ParseAbsoluteExpression(CurrValue))
3036 Values.push_back((uint8_t)CurrValue);
3039 getStreamer().EmitCFIEscape(Values);
3043 /// ParseDirectiveCFISignalFrame
3044 /// ::= .cfi_signal_frame
3045 bool GenericAsmParser::ParseDirectiveCFISignalFrame(StringRef Directive,
3046 SMLoc DirectiveLoc) {
3047 if (getLexer().isNot(AsmToken::EndOfStatement))
3048 return Error(getLexer().getLoc(),
3049 "unexpected token in '" + Directive + "' directive");
3051 getStreamer().EmitCFISignalFrame();
3056 /// ParseDirectiveMacrosOnOff
3059 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
3060 SMLoc DirectiveLoc) {
3061 if (getLexer().isNot(AsmToken::EndOfStatement))
3062 return Error(getLexer().getLoc(),
3063 "unexpected token in '" + Directive + "' directive");
3065 getParser().MacrosEnabled = Directive == ".macros_on";
3070 /// ParseDirectiveMacro
3071 /// ::= .macro name [parameters]
3072 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
3073 SMLoc DirectiveLoc) {
3075 if (getParser().ParseIdentifier(Name))
3076 return TokError("expected identifier in directive");
3078 MacroParameters Parameters;
3079 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3081 StringRef Parameter;
3082 if (getParser().ParseIdentifier(Parameter))
3083 return TokError("expected identifier in directive");
3084 Parameters.push_back(Parameter);
3086 if (getLexer().isNot(AsmToken::Comma))
3092 if (getLexer().isNot(AsmToken::EndOfStatement))
3093 return TokError("unexpected token in '.macro' directive");
3095 // Eat the end of statement.
3098 AsmToken EndToken, StartToken = getTok();
3100 // Lex the macro definition.
3102 // Check whether we have reached the end of the file.
3103 if (getLexer().is(AsmToken::Eof))
3104 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3106 // Otherwise, check whether we have reach the .endmacro.
3107 if (getLexer().is(AsmToken::Identifier) &&
3108 (getTok().getIdentifier() == ".endm" ||
3109 getTok().getIdentifier() == ".endmacro")) {
3110 EndToken = getTok();
3112 if (getLexer().isNot(AsmToken::EndOfStatement))
3113 return TokError("unexpected token in '" + EndToken.getIdentifier() +
3118 // Otherwise, scan til the end of the statement.
3119 getParser().EatToEndOfStatement();
3122 if (getParser().MacroMap.lookup(Name)) {
3123 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3126 const char *BodyStart = StartToken.getLoc().getPointer();
3127 const char *BodyEnd = EndToken.getLoc().getPointer();
3128 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3129 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters);
3133 /// ParseDirectiveEndMacro
3136 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
3137 SMLoc DirectiveLoc) {
3138 if (getLexer().isNot(AsmToken::EndOfStatement))
3139 return TokError("unexpected token in '" + Directive + "' directive");
3141 // If we are inside a macro instantiation, terminate the current
3143 if (!getParser().ActiveMacros.empty()) {
3144 getParser().HandleMacroExit();
3148 // Otherwise, this .endmacro is a stray entry in the file; well formed
3149 // .endmacro directives are handled during the macro definition parsing.
3150 return TokError("unexpected '" + Directive + "' in file, "
3151 "no current macro definition");
3154 /// ParseDirectivePurgeMacro
3156 bool GenericAsmParser::ParseDirectivePurgeMacro(StringRef Directive,
3157 SMLoc DirectiveLoc) {
3159 if (getParser().ParseIdentifier(Name))
3160 return TokError("expected identifier in '.purgem' directive");
3162 if (getLexer().isNot(AsmToken::EndOfStatement))
3163 return TokError("unexpected token in '.purgem' directive");
3165 StringMap<Macro*>::iterator I = getParser().MacroMap.find(Name);
3166 if (I == getParser().MacroMap.end())
3167 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3169 // Undefine the macro.
3170 delete I->getValue();
3171 getParser().MacroMap.erase(I);
3175 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
3176 getParser().CheckForValidSection();
3178 const MCExpr *Value;
3180 if (getParser().ParseExpression(Value))
3183 if (getLexer().isNot(AsmToken::EndOfStatement))
3184 return TokError("unexpected token in directive");
3186 if (DirName[1] == 's')
3187 getStreamer().EmitSLEB128Value(Value);
3189 getStreamer().EmitULEB128Value(Value);
3194 Macro *AsmParser::ParseMacroLikeBody(SMLoc DirectiveLoc) {
3195 AsmToken EndToken, StartToken = getTok();
3197 unsigned NestLevel = 0;
3199 // Check whether we have reached the end of the file.
3200 if (getLexer().is(AsmToken::Eof)) {
3201 Error(DirectiveLoc, "no matching '.endr' in definition");
3205 if (Lexer.is(AsmToken::Identifier) &&
3206 (getTok().getIdentifier() == ".rept")) {
3210 // Otherwise, check whether we have reached the .endr.
3211 if (Lexer.is(AsmToken::Identifier) &&
3212 getTok().getIdentifier() == ".endr") {
3213 if (NestLevel == 0) {
3214 EndToken = getTok();
3216 if (Lexer.isNot(AsmToken::EndOfStatement)) {
3217 TokError("unexpected token in '.endr' directive");
3225 // Otherwise, scan till the end of the statement.
3226 EatToEndOfStatement();
3229 const char *BodyStart = StartToken.getLoc().getPointer();
3230 const char *BodyEnd = EndToken.getLoc().getPointer();
3231 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3233 // We Are Anonymous.
3235 MacroParameters Parameters;
3236 return new Macro(Name, Body, Parameters);
3239 void AsmParser::InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc,
3240 raw_svector_ostream &OS) {
3243 MemoryBuffer *Instantiation =
3244 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
3246 // Create the macro instantiation object and add to the current macro
3247 // instantiation stack.
3248 MacroInstantiation *MI = new MacroInstantiation(M, DirectiveLoc,
3251 ActiveMacros.push_back(MI);
3253 // Jump to the macro instantiation and prime the lexer.
3254 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
3255 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
3259 bool AsmParser::ParseDirectiveRept(SMLoc DirectiveLoc) {
3261 if (ParseAbsoluteExpression(Count))
3262 return TokError("unexpected token in '.rept' directive");
3265 return TokError("Count is negative");
3267 if (Lexer.isNot(AsmToken::EndOfStatement))
3268 return TokError("unexpected token in '.rept' directive");
3270 // Eat the end of statement.
3273 // Lex the rept definition.
3274 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3278 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3279 // to hold the macro body with substitutions.
3280 SmallString<256> Buf;
3281 MacroParameters Parameters;
3283 raw_svector_ostream OS(Buf);
3285 if (expandMacro(OS, M->Body, Parameters, A, getTok().getLoc()))
3288 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3293 /// ParseDirectiveIrp
3294 /// ::= .irp symbol,values
3295 bool AsmParser::ParseDirectiveIrp(SMLoc DirectiveLoc) {
3296 MacroParameters Parameters;
3297 MacroParameter Parameter;
3299 if (ParseIdentifier(Parameter))
3300 return TokError("expected identifier in '.irp' directive");
3302 Parameters.push_back(Parameter);
3304 if (Lexer.isNot(AsmToken::Comma))
3305 return TokError("expected comma in '.irp' directive");
3310 if (ParseMacroArguments(0, A))
3313 // Eat the end of statement.
3316 // Lex the irp definition.
3317 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3321 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3322 // to hold the macro body with substitutions.
3323 SmallString<256> Buf;
3324 raw_svector_ostream OS(Buf);
3326 for (std::vector<MacroArgument>::iterator i = A.begin(), e = A.end(); i != e;
3328 std::vector<MacroArgument> Args;
3331 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3335 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3340 /// ParseDirectiveIrpc
3341 /// ::= .irpc symbol,values
3342 bool AsmParser::ParseDirectiveIrpc(SMLoc DirectiveLoc) {
3343 MacroParameters Parameters;
3344 MacroParameter Parameter;
3346 if (ParseIdentifier(Parameter))
3347 return TokError("expected identifier in '.irpc' directive");
3349 Parameters.push_back(Parameter);
3351 if (Lexer.isNot(AsmToken::Comma))
3352 return TokError("expected comma in '.irpc' directive");
3357 if (ParseMacroArguments(0, A))
3360 if (A.size() != 1 || A.front().size() != 1)
3361 return TokError("unexpected token in '.irpc' directive");
3363 // Eat the end of statement.
3366 // Lex the irpc definition.
3367 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3371 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3372 // to hold the macro body with substitutions.
3373 SmallString<256> Buf;
3374 raw_svector_ostream OS(Buf);
3376 StringRef Values = A.front().front().getString();
3377 std::size_t I, End = Values.size();
3378 for (I = 0; I < End; ++I) {
3380 Arg.push_back(AsmToken(AsmToken::Identifier, Values.slice(I, I+1)));
3382 MacroArguments Args;
3383 Args.push_back(Arg);
3385 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3389 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3394 bool AsmParser::ParseDirectiveEndr(SMLoc DirectiveLoc) {
3395 if (ActiveMacros.empty())
3396 return TokError("unexpected '.endr' directive, no current .rept");
3398 // The only .repl that should get here are the ones created by
3399 // InstantiateMacroLikeBody.
3400 assert(getLexer().is(AsmToken::EndOfStatement));
3406 /// \brief Create an MCAsmParser instance.
3407 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM,
3408 MCContext &C, MCStreamer &Out,
3409 const MCAsmInfo &MAI) {
3410 return new AsmParser(SM, C, Out, MAI);