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 // FIXME: This is mostly duplicated from the function in AsmLexer.cpp. The
1458 // difference being that that function accepts '@' as part of identifiers and
1459 // we can't do that. AsmLexer.cpp should probably be changed to handle
1460 // '@' as a special case when needed.
1461 static bool isIdentifierChar(char c) {
1462 return isalnum(c) || c == '_' || c == '$' || c == '.';
1465 bool AsmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
1466 const MacroParameters &Parameters,
1467 const MacroArguments &A,
1469 unsigned NParameters = Parameters.size();
1470 if (NParameters != 0 && NParameters != A.size())
1471 return Error(L, "Wrong number of arguments");
1473 while (!Body.empty()) {
1474 // Scan for the next substitution.
1475 std::size_t End = Body.size(), Pos = 0;
1476 for (; Pos != End; ++Pos) {
1477 // Check for a substitution or escape.
1479 // This macro has no parameters, look for $0, $1, etc.
1480 if (Body[Pos] != '$' || Pos + 1 == End)
1483 char Next = Body[Pos + 1];
1484 if (Next == '$' || Next == 'n' || isdigit(Next))
1487 // This macro has parameters, look for \foo, \bar, etc.
1488 if (Body[Pos] == '\\' && Pos + 1 != End)
1494 OS << Body.slice(0, Pos);
1496 // Check if we reached the end.
1501 switch (Body[Pos+1]) {
1507 // $n => number of arguments
1512 // $[0-9] => argument
1514 // Missing arguments are ignored.
1515 unsigned Index = Body[Pos+1] - '0';
1516 if (Index >= A.size())
1519 // Otherwise substitute with the token values, with spaces eliminated.
1520 for (MacroArgument::const_iterator it = A[Index].begin(),
1521 ie = A[Index].end(); it != ie; ++it)
1522 OS << it->getString();
1528 unsigned I = Pos + 1;
1529 while (isIdentifierChar(Body[I]) && I + 1 != End)
1532 const char *Begin = Body.data() + Pos +1;
1533 StringRef Argument(Begin, I - (Pos +1));
1535 for (; Index < NParameters; ++Index)
1536 if (Parameters[Index] == Argument)
1539 // FIXME: We should error at the macro definition.
1540 if (Index == NParameters)
1541 return Error(L, "Parameter not found");
1543 for (MacroArgument::const_iterator it = A[Index].begin(),
1544 ie = A[Index].end(); it != ie; ++it)
1545 OS << it->getString();
1547 Pos += 1 + Argument.size();
1549 // Update the scan point.
1550 Body = Body.substr(Pos);
1556 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1558 : TheMacro(M), Instantiation(I), InstantiationLoc(IL), ExitLoc(EL)
1562 /// ParseMacroArgument - Extract AsmTokens for a macro argument.
1563 /// This is used for both default macro parameter values and the
1564 /// arguments in macro invocations
1565 bool AsmParser::ParseMacroArgument(MacroArgument &MA) {
1566 unsigned ParenLevel = 0;
1569 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
1570 return TokError("unexpected token in macro instantiation");
1572 // HandleMacroEntry relies on not advancing the lexer here
1573 // to be able to fill in the remaining default parameter values
1574 if (Lexer.is(AsmToken::EndOfStatement))
1576 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma))
1579 // Adjust the current parentheses level.
1580 if (Lexer.is(AsmToken::LParen))
1582 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1585 // Append the token to the current argument list.
1586 MA.push_back(getTok());
1589 if (ParenLevel != 0)
1590 return TokError("unbalanced parentheses in macro argument");
1594 // Parse the macro instantiation arguments.
1595 bool AsmParser::ParseMacroArguments(const Macro *M, MacroArguments &A) {
1596 const unsigned NParameters = M ? M->Parameters.size() : 0;
1598 // Parse two kinds of macro invocations:
1599 // - macros defined without any parameters accept an arbitrary number of them
1600 // - macros defined with parameters accept at most that many of them
1601 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
1605 if (ParseMacroArgument(MA))
1610 if (Lexer.is(AsmToken::EndOfStatement))
1613 if (Lexer.is(AsmToken::Comma))
1616 return TokError("Too many arguments");
1619 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1621 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1622 // this, although we should protect against infinite loops.
1623 if (ActiveMacros.size() == 20)
1624 return TokError("macros cannot be nested more than 20 levels deep");
1627 if (ParseMacroArguments(M, A))
1630 // Remove any trailing empty arguments. Do this after-the-fact as we have
1631 // to keep empty arguments in the middle of the list or positionality
1632 // gets off. e.g., "foo 1, , 2" vs. "foo 1, 2,"
1633 while (!A.empty() && A.back().empty())
1636 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1637 // to hold the macro body with substitutions.
1638 SmallString<256> Buf;
1639 StringRef Body = M->Body;
1640 raw_svector_ostream OS(Buf);
1642 if (expandMacro(OS, Body, M->Parameters, A, getTok().getLoc()))
1645 // We include the .endmacro in the buffer as our queue to exit the macro
1647 OS << ".endmacro\n";
1649 MemoryBuffer *Instantiation =
1650 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1652 // Create the macro instantiation object and add to the current macro
1653 // instantiation stack.
1654 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1657 ActiveMacros.push_back(MI);
1659 // Jump to the macro instantiation and prime the lexer.
1660 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1661 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1667 void AsmParser::HandleMacroExit() {
1668 // Jump to the EndOfStatement we should return to, and consume it.
1669 JumpToLoc(ActiveMacros.back()->ExitLoc);
1672 // Pop the instantiation entry.
1673 delete ActiveMacros.back();
1674 ActiveMacros.pop_back();
1677 static bool IsUsedIn(const MCSymbol *Sym, const MCExpr *Value) {
1678 switch (Value->getKind()) {
1679 case MCExpr::Binary: {
1680 const MCBinaryExpr *BE = static_cast<const MCBinaryExpr*>(Value);
1681 return IsUsedIn(Sym, BE->getLHS()) || IsUsedIn(Sym, BE->getRHS());
1684 case MCExpr::Target:
1685 case MCExpr::Constant:
1687 case MCExpr::SymbolRef: {
1688 const MCSymbol &S = static_cast<const MCSymbolRefExpr*>(Value)->getSymbol();
1690 return IsUsedIn(Sym, S.getVariableValue());
1694 return IsUsedIn(Sym, static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1697 llvm_unreachable("Unknown expr kind!");
1700 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) {
1701 // FIXME: Use better location, we should use proper tokens.
1702 SMLoc EqualLoc = Lexer.getLoc();
1704 const MCExpr *Value;
1705 if (ParseExpression(Value))
1708 // Note: we don't count b as used in "a = b". This is to allow
1712 if (Lexer.isNot(AsmToken::EndOfStatement))
1713 return TokError("unexpected token in assignment");
1715 // Error on assignment to '.'.
1717 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported "
1718 "(use '.space' or '.org').)"));
1721 // Eat the end of statement marker.
1724 // Validate that the LHS is allowed to be a variable (either it has not been
1725 // used as a symbol, or it is an absolute symbol).
1726 MCSymbol *Sym = getContext().LookupSymbol(Name);
1728 // Diagnose assignment to a label.
1730 // FIXME: Diagnostics. Note the location of the definition as a label.
1731 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1732 if (IsUsedIn(Sym, Value))
1733 return Error(EqualLoc, "Recursive use of '" + Name + "'");
1734 else if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
1735 ; // Allow redefinitions of undefined symbols only used in directives.
1736 else if (Sym->isVariable() && !Sym->isUsed() && allow_redef)
1737 ; // Allow redefinitions of variables that haven't yet been used.
1738 else if (!Sym->isUndefined() && (!Sym->isVariable() || !allow_redef))
1739 return Error(EqualLoc, "redefinition of '" + Name + "'");
1740 else if (!Sym->isVariable())
1741 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1742 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1743 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1746 // Don't count these checks as uses.
1747 Sym->setUsed(false);
1749 Sym = getContext().GetOrCreateSymbol(Name);
1751 // FIXME: Handle '.'.
1753 // Do the assignment.
1754 Out.EmitAssignment(Sym, Value);
1759 /// ParseIdentifier:
1762 bool AsmParser::ParseIdentifier(StringRef &Res) {
1763 // The assembler has relaxed rules for accepting identifiers, in particular we
1764 // allow things like '.globl $foo', which would normally be separate
1765 // tokens. At this level, we have already lexed so we cannot (currently)
1766 // handle this as a context dependent token, instead we detect adjacent tokens
1767 // and return the combined identifier.
1768 if (Lexer.is(AsmToken::Dollar)) {
1769 SMLoc DollarLoc = getLexer().getLoc();
1771 // Consume the dollar sign, and check for a following identifier.
1773 if (Lexer.isNot(AsmToken::Identifier))
1776 // We have a '$' followed by an identifier, make sure they are adjacent.
1777 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1780 // Construct the joined identifier and consume the token.
1781 Res = StringRef(DollarLoc.getPointer(),
1782 getTok().getIdentifier().size() + 1);
1787 if (Lexer.isNot(AsmToken::Identifier) &&
1788 Lexer.isNot(AsmToken::String))
1791 Res = getTok().getIdentifier();
1793 Lex(); // Consume the identifier token.
1798 /// ParseDirectiveSet:
1799 /// ::= .equ identifier ',' expression
1800 /// ::= .equiv identifier ',' expression
1801 /// ::= .set identifier ',' expression
1802 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1805 if (ParseIdentifier(Name))
1806 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1808 if (getLexer().isNot(AsmToken::Comma))
1809 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1812 return ParseAssignment(Name, allow_redef);
1815 bool AsmParser::ParseEscapedString(std::string &Data) {
1816 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1819 StringRef Str = getTok().getStringContents();
1820 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1821 if (Str[i] != '\\') {
1826 // Recognize escaped characters. Note that this escape semantics currently
1827 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1830 return TokError("unexpected backslash at end of string");
1832 // Recognize octal sequences.
1833 if ((unsigned) (Str[i] - '0') <= 7) {
1834 // Consume up to three octal characters.
1835 unsigned Value = Str[i] - '0';
1837 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1839 Value = Value * 8 + (Str[i] - '0');
1841 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1843 Value = Value * 8 + (Str[i] - '0');
1848 return TokError("invalid octal escape sequence (out of range)");
1850 Data += (unsigned char) Value;
1854 // Otherwise recognize individual escapes.
1857 // Just reject invalid escape sequences for now.
1858 return TokError("invalid escape sequence (unrecognized character)");
1860 case 'b': Data += '\b'; break;
1861 case 'f': Data += '\f'; break;
1862 case 'n': Data += '\n'; break;
1863 case 'r': Data += '\r'; break;
1864 case 't': Data += '\t'; break;
1865 case '"': Data += '"'; break;
1866 case '\\': Data += '\\'; break;
1873 /// ParseDirectiveAscii:
1874 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1875 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1876 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1877 CheckForValidSection();
1880 if (getLexer().isNot(AsmToken::String))
1881 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1884 if (ParseEscapedString(Data))
1887 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1889 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1893 if (getLexer().is(AsmToken::EndOfStatement))
1896 if (getLexer().isNot(AsmToken::Comma))
1897 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1906 /// ParseDirectiveValue
1907 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1908 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1909 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1910 CheckForValidSection();
1913 const MCExpr *Value;
1914 SMLoc ExprLoc = getLexer().getLoc();
1915 if (ParseExpression(Value))
1918 // Special case constant expressions to match code generator.
1919 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
1920 assert(Size <= 8 && "Invalid size");
1921 uint64_t IntValue = MCE->getValue();
1922 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
1923 return Error(ExprLoc, "literal value out of range for directive");
1924 getStreamer().EmitIntValue(IntValue, Size, DEFAULT_ADDRSPACE);
1926 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1928 if (getLexer().is(AsmToken::EndOfStatement))
1931 // FIXME: Improve diagnostic.
1932 if (getLexer().isNot(AsmToken::Comma))
1933 return TokError("unexpected token in directive");
1942 /// ParseDirectiveRealValue
1943 /// ::= (.single | .double) [ expression (, expression)* ]
1944 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1945 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1946 CheckForValidSection();
1949 // We don't truly support arithmetic on floating point expressions, so we
1950 // have to manually parse unary prefixes.
1952 if (getLexer().is(AsmToken::Minus)) {
1955 } else if (getLexer().is(AsmToken::Plus))
1958 if (getLexer().isNot(AsmToken::Integer) &&
1959 getLexer().isNot(AsmToken::Real) &&
1960 getLexer().isNot(AsmToken::Identifier))
1961 return TokError("unexpected token in directive");
1963 // Convert to an APFloat.
1964 APFloat Value(Semantics);
1965 StringRef IDVal = getTok().getString();
1966 if (getLexer().is(AsmToken::Identifier)) {
1967 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
1968 Value = APFloat::getInf(Semantics);
1969 else if (!IDVal.compare_lower("nan"))
1970 Value = APFloat::getNaN(Semantics, false, ~0);
1972 return TokError("invalid floating point literal");
1973 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
1974 APFloat::opInvalidOp)
1975 return TokError("invalid floating point literal");
1979 // Consume the numeric token.
1982 // Emit the value as an integer.
1983 APInt AsInt = Value.bitcastToAPInt();
1984 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1985 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1987 if (getLexer().is(AsmToken::EndOfStatement))
1990 if (getLexer().isNot(AsmToken::Comma))
1991 return TokError("unexpected token in directive");
2000 /// ParseDirectiveSpace
2001 /// ::= .space expression [ , expression ]
2002 bool AsmParser::ParseDirectiveSpace() {
2003 CheckForValidSection();
2006 if (ParseAbsoluteExpression(NumBytes))
2009 int64_t FillExpr = 0;
2010 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2011 if (getLexer().isNot(AsmToken::Comma))
2012 return TokError("unexpected token in '.space' directive");
2015 if (ParseAbsoluteExpression(FillExpr))
2018 if (getLexer().isNot(AsmToken::EndOfStatement))
2019 return TokError("unexpected token in '.space' directive");
2025 return TokError("invalid number of bytes in '.space' directive");
2027 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
2028 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
2033 /// ParseDirectiveZero
2034 /// ::= .zero expression
2035 bool AsmParser::ParseDirectiveZero() {
2036 CheckForValidSection();
2039 if (ParseAbsoluteExpression(NumBytes))
2043 if (getLexer().is(AsmToken::Comma)) {
2045 if (ParseAbsoluteExpression(Val))
2049 if (getLexer().isNot(AsmToken::EndOfStatement))
2050 return TokError("unexpected token in '.zero' directive");
2054 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
2059 /// ParseDirectiveFill
2060 /// ::= .fill expression , expression , expression
2061 bool AsmParser::ParseDirectiveFill() {
2062 CheckForValidSection();
2065 if (ParseAbsoluteExpression(NumValues))
2068 if (getLexer().isNot(AsmToken::Comma))
2069 return TokError("unexpected token in '.fill' directive");
2073 if (ParseAbsoluteExpression(FillSize))
2076 if (getLexer().isNot(AsmToken::Comma))
2077 return TokError("unexpected token in '.fill' directive");
2081 if (ParseAbsoluteExpression(FillExpr))
2084 if (getLexer().isNot(AsmToken::EndOfStatement))
2085 return TokError("unexpected token in '.fill' directive");
2089 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
2090 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
2092 for (uint64_t i = 0, e = NumValues; i != e; ++i)
2093 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
2098 /// ParseDirectiveOrg
2099 /// ::= .org expression [ , expression ]
2100 bool AsmParser::ParseDirectiveOrg() {
2101 CheckForValidSection();
2103 const MCExpr *Offset;
2104 SMLoc Loc = getTok().getLoc();
2105 if (ParseExpression(Offset))
2108 // Parse optional fill expression.
2109 int64_t FillExpr = 0;
2110 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2111 if (getLexer().isNot(AsmToken::Comma))
2112 return TokError("unexpected token in '.org' directive");
2115 if (ParseAbsoluteExpression(FillExpr))
2118 if (getLexer().isNot(AsmToken::EndOfStatement))
2119 return TokError("unexpected token in '.org' directive");
2124 // Only limited forms of relocatable expressions are accepted here, it
2125 // has to be relative to the current section. The streamer will return
2126 // 'true' if the expression wasn't evaluatable.
2127 if (getStreamer().EmitValueToOffset(Offset, FillExpr))
2128 return Error(Loc, "expected assembly-time absolute expression");
2133 /// ParseDirectiveAlign
2134 /// ::= {.align, ...} expression [ , expression [ , expression ]]
2135 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
2136 CheckForValidSection();
2138 SMLoc AlignmentLoc = getLexer().getLoc();
2140 if (ParseAbsoluteExpression(Alignment))
2144 bool HasFillExpr = false;
2145 int64_t FillExpr = 0;
2146 int64_t MaxBytesToFill = 0;
2147 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2148 if (getLexer().isNot(AsmToken::Comma))
2149 return TokError("unexpected token in directive");
2152 // The fill expression can be omitted while specifying a maximum number of
2153 // alignment bytes, e.g:
2155 if (getLexer().isNot(AsmToken::Comma)) {
2157 if (ParseAbsoluteExpression(FillExpr))
2161 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2162 if (getLexer().isNot(AsmToken::Comma))
2163 return TokError("unexpected token in directive");
2166 MaxBytesLoc = getLexer().getLoc();
2167 if (ParseAbsoluteExpression(MaxBytesToFill))
2170 if (getLexer().isNot(AsmToken::EndOfStatement))
2171 return TokError("unexpected token in directive");
2180 // Compute alignment in bytes.
2182 // FIXME: Diagnose overflow.
2183 if (Alignment >= 32) {
2184 Error(AlignmentLoc, "invalid alignment value");
2188 Alignment = 1ULL << Alignment;
2191 // Diagnose non-sensical max bytes to align.
2192 if (MaxBytesLoc.isValid()) {
2193 if (MaxBytesToFill < 1) {
2194 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
2195 "many bytes, ignoring maximum bytes expression");
2199 if (MaxBytesToFill >= Alignment) {
2200 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
2206 // Check whether we should use optimal code alignment for this .align
2208 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
2209 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
2210 ValueSize == 1 && UseCodeAlign) {
2211 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
2213 // FIXME: Target specific behavior about how the "extra" bytes are filled.
2214 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
2221 /// ParseDirectiveSymbolAttribute
2222 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
2223 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
2224 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2227 SMLoc Loc = getTok().getLoc();
2229 if (ParseIdentifier(Name))
2230 return Error(Loc, "expected identifier in directive");
2232 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2234 // Assembler local symbols don't make any sense here. Complain loudly.
2235 if (Sym->isTemporary())
2236 return Error(Loc, "non-local symbol required in directive");
2238 getStreamer().EmitSymbolAttribute(Sym, Attr);
2240 if (getLexer().is(AsmToken::EndOfStatement))
2243 if (getLexer().isNot(AsmToken::Comma))
2244 return TokError("unexpected token in directive");
2253 /// ParseDirectiveComm
2254 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
2255 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
2256 CheckForValidSection();
2258 SMLoc IDLoc = getLexer().getLoc();
2260 if (ParseIdentifier(Name))
2261 return TokError("expected identifier in directive");
2263 // Handle the identifier as the key symbol.
2264 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2266 if (getLexer().isNot(AsmToken::Comma))
2267 return TokError("unexpected token in directive");
2271 SMLoc SizeLoc = getLexer().getLoc();
2272 if (ParseAbsoluteExpression(Size))
2275 int64_t Pow2Alignment = 0;
2276 SMLoc Pow2AlignmentLoc;
2277 if (getLexer().is(AsmToken::Comma)) {
2279 Pow2AlignmentLoc = getLexer().getLoc();
2280 if (ParseAbsoluteExpression(Pow2Alignment))
2283 // If this target takes alignments in bytes (not log) validate and convert.
2284 if (Lexer.getMAI().getAlignmentIsInBytes()) {
2285 if (!isPowerOf2_64(Pow2Alignment))
2286 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
2287 Pow2Alignment = Log2_64(Pow2Alignment);
2291 if (getLexer().isNot(AsmToken::EndOfStatement))
2292 return TokError("unexpected token in '.comm' or '.lcomm' directive");
2296 // NOTE: a size of zero for a .comm should create a undefined symbol
2297 // but a size of .lcomm creates a bss symbol of size zero.
2299 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
2300 "be less than zero");
2302 // NOTE: The alignment in the directive is a power of 2 value, the assembler
2303 // may internally end up wanting an alignment in bytes.
2304 // FIXME: Diagnose overflow.
2305 if (Pow2Alignment < 0)
2306 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
2307 "alignment, can't be less than zero");
2309 if (!Sym->isUndefined())
2310 return Error(IDLoc, "invalid symbol redefinition");
2312 // '.lcomm' is equivalent to '.zerofill'.
2313 // Create the Symbol as a common or local common with Size and Pow2Alignment
2315 getStreamer().EmitZerofill(Ctx.getMachOSection(
2316 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
2317 0, SectionKind::getBSS()),
2318 Sym, Size, 1 << Pow2Alignment);
2322 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
2326 /// ParseDirectiveAbort
2327 /// ::= .abort [... message ...]
2328 bool AsmParser::ParseDirectiveAbort() {
2329 // FIXME: Use loc from directive.
2330 SMLoc Loc = getLexer().getLoc();
2332 StringRef Str = ParseStringToEndOfStatement();
2333 if (getLexer().isNot(AsmToken::EndOfStatement))
2334 return TokError("unexpected token in '.abort' directive");
2339 Error(Loc, ".abort detected. Assembly stopping.");
2341 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
2342 // FIXME: Actually abort assembly here.
2347 /// ParseDirectiveInclude
2348 /// ::= .include "filename"
2349 bool AsmParser::ParseDirectiveInclude() {
2350 if (getLexer().isNot(AsmToken::String))
2351 return TokError("expected string in '.include' directive");
2353 std::string Filename = getTok().getString();
2354 SMLoc IncludeLoc = getLexer().getLoc();
2357 if (getLexer().isNot(AsmToken::EndOfStatement))
2358 return TokError("unexpected token in '.include' directive");
2360 // Strip the quotes.
2361 Filename = Filename.substr(1, Filename.size()-2);
2363 // Attempt to switch the lexer to the included file before consuming the end
2364 // of statement to avoid losing it when we switch.
2365 if (EnterIncludeFile(Filename)) {
2366 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
2373 /// ParseDirectiveIncbin
2374 /// ::= .incbin "filename"
2375 bool AsmParser::ParseDirectiveIncbin() {
2376 if (getLexer().isNot(AsmToken::String))
2377 return TokError("expected string in '.incbin' directive");
2379 std::string Filename = getTok().getString();
2380 SMLoc IncbinLoc = getLexer().getLoc();
2383 if (getLexer().isNot(AsmToken::EndOfStatement))
2384 return TokError("unexpected token in '.incbin' directive");
2386 // Strip the quotes.
2387 Filename = Filename.substr(1, Filename.size()-2);
2389 // Attempt to process the included file.
2390 if (ProcessIncbinFile(Filename)) {
2391 Error(IncbinLoc, "Could not find incbin file '" + Filename + "'");
2398 /// ParseDirectiveIf
2399 /// ::= .if expression
2400 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
2401 TheCondStack.push_back(TheCondState);
2402 TheCondState.TheCond = AsmCond::IfCond;
2403 if (TheCondState.Ignore) {
2404 EatToEndOfStatement();
2407 if (ParseAbsoluteExpression(ExprValue))
2410 if (getLexer().isNot(AsmToken::EndOfStatement))
2411 return TokError("unexpected token in '.if' directive");
2415 TheCondState.CondMet = ExprValue;
2416 TheCondState.Ignore = !TheCondState.CondMet;
2422 /// ParseDirectiveIfb
2424 bool AsmParser::ParseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) {
2425 TheCondStack.push_back(TheCondState);
2426 TheCondState.TheCond = AsmCond::IfCond;
2428 if (TheCondState.Ignore) {
2429 EatToEndOfStatement();
2431 StringRef Str = ParseStringToEndOfStatement();
2433 if (getLexer().isNot(AsmToken::EndOfStatement))
2434 return TokError("unexpected token in '.ifb' directive");
2438 TheCondState.CondMet = ExpectBlank == Str.empty();
2439 TheCondState.Ignore = !TheCondState.CondMet;
2445 /// ParseDirectiveIfc
2446 /// ::= .ifc string1, string2
2447 bool AsmParser::ParseDirectiveIfc(SMLoc DirectiveLoc, bool ExpectEqual) {
2448 TheCondStack.push_back(TheCondState);
2449 TheCondState.TheCond = AsmCond::IfCond;
2451 if (TheCondState.Ignore) {
2452 EatToEndOfStatement();
2454 StringRef Str1 = ParseStringToComma();
2456 if (getLexer().isNot(AsmToken::Comma))
2457 return TokError("unexpected token in '.ifc' directive");
2461 StringRef Str2 = ParseStringToEndOfStatement();
2463 if (getLexer().isNot(AsmToken::EndOfStatement))
2464 return TokError("unexpected token in '.ifc' directive");
2468 TheCondState.CondMet = ExpectEqual == (Str1 == Str2);
2469 TheCondState.Ignore = !TheCondState.CondMet;
2475 /// ParseDirectiveIfdef
2476 /// ::= .ifdef symbol
2477 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
2479 TheCondStack.push_back(TheCondState);
2480 TheCondState.TheCond = AsmCond::IfCond;
2482 if (TheCondState.Ignore) {
2483 EatToEndOfStatement();
2485 if (ParseIdentifier(Name))
2486 return TokError("expected identifier after '.ifdef'");
2490 MCSymbol *Sym = getContext().LookupSymbol(Name);
2493 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
2495 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
2496 TheCondState.Ignore = !TheCondState.CondMet;
2502 /// ParseDirectiveElseIf
2503 /// ::= .elseif expression
2504 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
2505 if (TheCondState.TheCond != AsmCond::IfCond &&
2506 TheCondState.TheCond != AsmCond::ElseIfCond)
2507 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
2509 TheCondState.TheCond = AsmCond::ElseIfCond;
2511 bool LastIgnoreState = false;
2512 if (!TheCondStack.empty())
2513 LastIgnoreState = TheCondStack.back().Ignore;
2514 if (LastIgnoreState || TheCondState.CondMet) {
2515 TheCondState.Ignore = true;
2516 EatToEndOfStatement();
2520 if (ParseAbsoluteExpression(ExprValue))
2523 if (getLexer().isNot(AsmToken::EndOfStatement))
2524 return TokError("unexpected token in '.elseif' directive");
2527 TheCondState.CondMet = ExprValue;
2528 TheCondState.Ignore = !TheCondState.CondMet;
2534 /// ParseDirectiveElse
2536 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2537 if (getLexer().isNot(AsmToken::EndOfStatement))
2538 return TokError("unexpected token in '.else' directive");
2542 if (TheCondState.TheCond != AsmCond::IfCond &&
2543 TheCondState.TheCond != AsmCond::ElseIfCond)
2544 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
2546 TheCondState.TheCond = AsmCond::ElseCond;
2547 bool LastIgnoreState = false;
2548 if (!TheCondStack.empty())
2549 LastIgnoreState = TheCondStack.back().Ignore;
2550 if (LastIgnoreState || TheCondState.CondMet)
2551 TheCondState.Ignore = true;
2553 TheCondState.Ignore = false;
2558 /// ParseDirectiveEndIf
2560 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2561 if (getLexer().isNot(AsmToken::EndOfStatement))
2562 return TokError("unexpected token in '.endif' directive");
2566 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2567 TheCondStack.empty())
2568 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2570 if (!TheCondStack.empty()) {
2571 TheCondState = TheCondStack.back();
2572 TheCondStack.pop_back();
2578 /// ParseDirectiveFile
2579 /// ::= .file [number] filename
2580 /// ::= .file number directory filename
2581 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
2582 // FIXME: I'm not sure what this is.
2583 int64_t FileNumber = -1;
2584 SMLoc FileNumberLoc = getLexer().getLoc();
2585 if (getLexer().is(AsmToken::Integer)) {
2586 FileNumber = getTok().getIntVal();
2590 return TokError("file number less than one");
2593 if (getLexer().isNot(AsmToken::String))
2594 return TokError("unexpected token in '.file' directive");
2596 // Usually the directory and filename together, otherwise just the directory.
2597 StringRef Path = getTok().getString();
2598 Path = Path.substr(1, Path.size()-2);
2601 StringRef Directory;
2603 if (getLexer().is(AsmToken::String)) {
2604 if (FileNumber == -1)
2605 return TokError("explicit path specified, but no file number");
2606 Filename = getTok().getString();
2607 Filename = Filename.substr(1, Filename.size()-2);
2614 if (getLexer().isNot(AsmToken::EndOfStatement))
2615 return TokError("unexpected token in '.file' directive");
2617 if (FileNumber == -1)
2618 getStreamer().EmitFileDirective(Filename);
2620 if (getContext().getGenDwarfForAssembly() == true)
2621 Error(DirectiveLoc, "input can't have .file dwarf directives when -g is "
2622 "used to generate dwarf debug info for assembly code");
2624 if (getStreamer().EmitDwarfFileDirective(FileNumber, Directory, Filename))
2625 Error(FileNumberLoc, "file number already allocated");
2631 /// ParseDirectiveLine
2632 /// ::= .line [number]
2633 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2634 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2635 if (getLexer().isNot(AsmToken::Integer))
2636 return TokError("unexpected token in '.line' directive");
2638 int64_t LineNumber = getTok().getIntVal();
2642 // FIXME: Do something with the .line.
2645 if (getLexer().isNot(AsmToken::EndOfStatement))
2646 return TokError("unexpected token in '.line' directive");
2652 /// ParseDirectiveLoc
2653 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2654 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2655 /// The first number is a file number, must have been previously assigned with
2656 /// a .file directive, the second number is the line number and optionally the
2657 /// third number is a column position (zero if not specified). The remaining
2658 /// optional items are .loc sub-directives.
2659 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2661 if (getLexer().isNot(AsmToken::Integer))
2662 return TokError("unexpected token in '.loc' directive");
2663 int64_t FileNumber = getTok().getIntVal();
2665 return TokError("file number less than one in '.loc' directive");
2666 if (!getContext().isValidDwarfFileNumber(FileNumber))
2667 return TokError("unassigned file number in '.loc' directive");
2670 int64_t LineNumber = 0;
2671 if (getLexer().is(AsmToken::Integer)) {
2672 LineNumber = getTok().getIntVal();
2674 return TokError("line number less than one in '.loc' directive");
2678 int64_t ColumnPos = 0;
2679 if (getLexer().is(AsmToken::Integer)) {
2680 ColumnPos = getTok().getIntVal();
2682 return TokError("column position less than zero in '.loc' directive");
2686 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2688 int64_t Discriminator = 0;
2689 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2691 if (getLexer().is(AsmToken::EndOfStatement))
2695 SMLoc Loc = getTok().getLoc();
2696 if (getParser().ParseIdentifier(Name))
2697 return TokError("unexpected token in '.loc' directive");
2699 if (Name == "basic_block")
2700 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2701 else if (Name == "prologue_end")
2702 Flags |= DWARF2_FLAG_PROLOGUE_END;
2703 else if (Name == "epilogue_begin")
2704 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2705 else if (Name == "is_stmt") {
2706 SMLoc Loc = getTok().getLoc();
2707 const MCExpr *Value;
2708 if (getParser().ParseExpression(Value))
2710 // The expression must be the constant 0 or 1.
2711 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2712 int Value = MCE->getValue();
2714 Flags &= ~DWARF2_FLAG_IS_STMT;
2715 else if (Value == 1)
2716 Flags |= DWARF2_FLAG_IS_STMT;
2718 return Error(Loc, "is_stmt value not 0 or 1");
2721 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2724 else if (Name == "isa") {
2725 SMLoc Loc = getTok().getLoc();
2726 const MCExpr *Value;
2727 if (getParser().ParseExpression(Value))
2729 // The expression must be a constant greater or equal to 0.
2730 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2731 int Value = MCE->getValue();
2733 return Error(Loc, "isa number less than zero");
2737 return Error(Loc, "isa number not a constant value");
2740 else if (Name == "discriminator") {
2741 if (getParser().ParseAbsoluteExpression(Discriminator))
2745 return Error(Loc, "unknown sub-directive in '.loc' directive");
2748 if (getLexer().is(AsmToken::EndOfStatement))
2753 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2754 Isa, Discriminator, StringRef());
2759 /// ParseDirectiveStabs
2760 /// ::= .stabs string, number, number, number
2761 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2762 SMLoc DirectiveLoc) {
2763 return TokError("unsupported directive '" + Directive + "'");
2766 /// ParseDirectiveCFISections
2767 /// ::= .cfi_sections section [, section]
2768 bool GenericAsmParser::ParseDirectiveCFISections(StringRef,
2769 SMLoc DirectiveLoc) {
2774 if (getParser().ParseIdentifier(Name))
2775 return TokError("Expected an identifier");
2777 if (Name == ".eh_frame")
2779 else if (Name == ".debug_frame")
2782 if (getLexer().is(AsmToken::Comma)) {
2785 if (getParser().ParseIdentifier(Name))
2786 return TokError("Expected an identifier");
2788 if (Name == ".eh_frame")
2790 else if (Name == ".debug_frame")
2794 getStreamer().EmitCFISections(EH, Debug);
2799 /// ParseDirectiveCFIStartProc
2800 /// ::= .cfi_startproc
2801 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2802 SMLoc DirectiveLoc) {
2803 getStreamer().EmitCFIStartProc();
2807 /// ParseDirectiveCFIEndProc
2808 /// ::= .cfi_endproc
2809 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2810 getStreamer().EmitCFIEndProc();
2814 /// ParseRegisterOrRegisterNumber - parse register name or number.
2815 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2816 SMLoc DirectiveLoc) {
2819 if (getLexer().isNot(AsmToken::Integer)) {
2820 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2823 Register = getContext().getRegisterInfo().getDwarfRegNum(RegNo, true);
2825 return getParser().ParseAbsoluteExpression(Register);
2830 /// ParseDirectiveCFIDefCfa
2831 /// ::= .cfi_def_cfa register, offset
2832 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
2833 SMLoc DirectiveLoc) {
2834 int64_t Register = 0;
2835 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2838 if (getLexer().isNot(AsmToken::Comma))
2839 return TokError("unexpected token in directive");
2843 if (getParser().ParseAbsoluteExpression(Offset))
2846 getStreamer().EmitCFIDefCfa(Register, Offset);
2850 /// ParseDirectiveCFIDefCfaOffset
2851 /// ::= .cfi_def_cfa_offset offset
2852 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
2853 SMLoc DirectiveLoc) {
2855 if (getParser().ParseAbsoluteExpression(Offset))
2858 getStreamer().EmitCFIDefCfaOffset(Offset);
2862 /// ParseDirectiveCFIAdjustCfaOffset
2863 /// ::= .cfi_adjust_cfa_offset adjustment
2864 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef,
2865 SMLoc DirectiveLoc) {
2866 int64_t Adjustment = 0;
2867 if (getParser().ParseAbsoluteExpression(Adjustment))
2870 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
2874 /// ParseDirectiveCFIDefCfaRegister
2875 /// ::= .cfi_def_cfa_register register
2876 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
2877 SMLoc DirectiveLoc) {
2878 int64_t Register = 0;
2879 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2882 getStreamer().EmitCFIDefCfaRegister(Register);
2886 /// ParseDirectiveCFIOffset
2887 /// ::= .cfi_offset register, offset
2888 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
2889 int64_t Register = 0;
2892 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2895 if (getLexer().isNot(AsmToken::Comma))
2896 return TokError("unexpected token in directive");
2899 if (getParser().ParseAbsoluteExpression(Offset))
2902 getStreamer().EmitCFIOffset(Register, Offset);
2906 /// ParseDirectiveCFIRelOffset
2907 /// ::= .cfi_rel_offset register, offset
2908 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef,
2909 SMLoc DirectiveLoc) {
2910 int64_t Register = 0;
2912 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2915 if (getLexer().isNot(AsmToken::Comma))
2916 return TokError("unexpected token in directive");
2920 if (getParser().ParseAbsoluteExpression(Offset))
2923 getStreamer().EmitCFIRelOffset(Register, Offset);
2927 static bool isValidEncoding(int64_t Encoding) {
2928 if (Encoding & ~0xff)
2931 if (Encoding == dwarf::DW_EH_PE_omit)
2934 const unsigned Format = Encoding & 0xf;
2935 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
2936 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
2937 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
2938 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
2941 const unsigned Application = Encoding & 0x70;
2942 if (Application != dwarf::DW_EH_PE_absptr &&
2943 Application != dwarf::DW_EH_PE_pcrel)
2949 /// ParseDirectiveCFIPersonalityOrLsda
2950 /// ::= .cfi_personality encoding, [symbol_name]
2951 /// ::= .cfi_lsda encoding, [symbol_name]
2952 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
2953 SMLoc DirectiveLoc) {
2954 int64_t Encoding = 0;
2955 if (getParser().ParseAbsoluteExpression(Encoding))
2957 if (Encoding == dwarf::DW_EH_PE_omit)
2960 if (!isValidEncoding(Encoding))
2961 return TokError("unsupported encoding.");
2963 if (getLexer().isNot(AsmToken::Comma))
2964 return TokError("unexpected token in directive");
2968 if (getParser().ParseIdentifier(Name))
2969 return TokError("expected identifier in directive");
2971 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2973 if (IDVal == ".cfi_personality")
2974 getStreamer().EmitCFIPersonality(Sym, Encoding);
2976 assert(IDVal == ".cfi_lsda");
2977 getStreamer().EmitCFILsda(Sym, Encoding);
2982 /// ParseDirectiveCFIRememberState
2983 /// ::= .cfi_remember_state
2984 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
2985 SMLoc DirectiveLoc) {
2986 getStreamer().EmitCFIRememberState();
2990 /// ParseDirectiveCFIRestoreState
2991 /// ::= .cfi_remember_state
2992 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
2993 SMLoc DirectiveLoc) {
2994 getStreamer().EmitCFIRestoreState();
2998 /// ParseDirectiveCFISameValue
2999 /// ::= .cfi_same_value register
3000 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal,
3001 SMLoc DirectiveLoc) {
3002 int64_t Register = 0;
3004 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3007 getStreamer().EmitCFISameValue(Register);
3012 /// ParseDirectiveCFIRestore
3013 /// ::= .cfi_restore register
3014 bool GenericAsmParser::ParseDirectiveCFIRestore(StringRef IDVal,
3015 SMLoc DirectiveLoc) {
3016 int64_t Register = 0;
3017 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
3020 getStreamer().EmitCFIRestore(Register);
3025 /// ParseDirectiveCFIEscape
3026 /// ::= .cfi_escape expression[,...]
3027 bool GenericAsmParser::ParseDirectiveCFIEscape(StringRef IDVal,
3028 SMLoc DirectiveLoc) {
3031 if (getParser().ParseAbsoluteExpression(CurrValue))
3034 Values.push_back((uint8_t)CurrValue);
3036 while (getLexer().is(AsmToken::Comma)) {
3039 if (getParser().ParseAbsoluteExpression(CurrValue))
3042 Values.push_back((uint8_t)CurrValue);
3045 getStreamer().EmitCFIEscape(Values);
3049 /// ParseDirectiveCFISignalFrame
3050 /// ::= .cfi_signal_frame
3051 bool GenericAsmParser::ParseDirectiveCFISignalFrame(StringRef Directive,
3052 SMLoc DirectiveLoc) {
3053 if (getLexer().isNot(AsmToken::EndOfStatement))
3054 return Error(getLexer().getLoc(),
3055 "unexpected token in '" + Directive + "' directive");
3057 getStreamer().EmitCFISignalFrame();
3062 /// ParseDirectiveMacrosOnOff
3065 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
3066 SMLoc DirectiveLoc) {
3067 if (getLexer().isNot(AsmToken::EndOfStatement))
3068 return Error(getLexer().getLoc(),
3069 "unexpected token in '" + Directive + "' directive");
3071 getParser().MacrosEnabled = Directive == ".macros_on";
3076 /// ParseDirectiveMacro
3077 /// ::= .macro name [parameters]
3078 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
3079 SMLoc DirectiveLoc) {
3081 if (getParser().ParseIdentifier(Name))
3082 return TokError("expected identifier in '.macro' directive");
3084 MacroParameters Parameters;
3085 if (getLexer().isNot(AsmToken::EndOfStatement)) {
3087 MacroParameter Parameter;
3088 if (getParser().ParseIdentifier(Parameter))
3089 return TokError("expected identifier in '.macro' directive");
3090 Parameters.push_back(Parameter);
3092 if (getLexer().isNot(AsmToken::Comma))
3098 if (getLexer().isNot(AsmToken::EndOfStatement))
3099 return TokError("unexpected token in '.macro' directive");
3101 // Eat the end of statement.
3104 AsmToken EndToken, StartToken = getTok();
3106 // Lex the macro definition.
3108 // Check whether we have reached the end of the file.
3109 if (getLexer().is(AsmToken::Eof))
3110 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
3112 // Otherwise, check whether we have reach the .endmacro.
3113 if (getLexer().is(AsmToken::Identifier) &&
3114 (getTok().getIdentifier() == ".endm" ||
3115 getTok().getIdentifier() == ".endmacro")) {
3116 EndToken = getTok();
3118 if (getLexer().isNot(AsmToken::EndOfStatement))
3119 return TokError("unexpected token in '" + EndToken.getIdentifier() +
3124 // Otherwise, scan til the end of the statement.
3125 getParser().EatToEndOfStatement();
3128 if (getParser().MacroMap.lookup(Name)) {
3129 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
3132 const char *BodyStart = StartToken.getLoc().getPointer();
3133 const char *BodyEnd = EndToken.getLoc().getPointer();
3134 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3135 getParser().MacroMap[Name] = new Macro(Name, Body, Parameters);
3139 /// ParseDirectiveEndMacro
3142 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
3143 SMLoc DirectiveLoc) {
3144 if (getLexer().isNot(AsmToken::EndOfStatement))
3145 return TokError("unexpected token in '" + Directive + "' directive");
3147 // If we are inside a macro instantiation, terminate the current
3149 if (!getParser().ActiveMacros.empty()) {
3150 getParser().HandleMacroExit();
3154 // Otherwise, this .endmacro is a stray entry in the file; well formed
3155 // .endmacro directives are handled during the macro definition parsing.
3156 return TokError("unexpected '" + Directive + "' in file, "
3157 "no current macro definition");
3160 /// ParseDirectivePurgeMacro
3162 bool GenericAsmParser::ParseDirectivePurgeMacro(StringRef Directive,
3163 SMLoc DirectiveLoc) {
3165 if (getParser().ParseIdentifier(Name))
3166 return TokError("expected identifier in '.purgem' directive");
3168 if (getLexer().isNot(AsmToken::EndOfStatement))
3169 return TokError("unexpected token in '.purgem' directive");
3171 StringMap<Macro*>::iterator I = getParser().MacroMap.find(Name);
3172 if (I == getParser().MacroMap.end())
3173 return Error(DirectiveLoc, "macro '" + Name + "' is not defined");
3175 // Undefine the macro.
3176 delete I->getValue();
3177 getParser().MacroMap.erase(I);
3181 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
3182 getParser().CheckForValidSection();
3184 const MCExpr *Value;
3186 if (getParser().ParseExpression(Value))
3189 if (getLexer().isNot(AsmToken::EndOfStatement))
3190 return TokError("unexpected token in directive");
3192 if (DirName[1] == 's')
3193 getStreamer().EmitSLEB128Value(Value);
3195 getStreamer().EmitULEB128Value(Value);
3200 Macro *AsmParser::ParseMacroLikeBody(SMLoc DirectiveLoc) {
3201 AsmToken EndToken, StartToken = getTok();
3203 unsigned NestLevel = 0;
3205 // Check whether we have reached the end of the file.
3206 if (getLexer().is(AsmToken::Eof)) {
3207 Error(DirectiveLoc, "no matching '.endr' in definition");
3211 if (Lexer.is(AsmToken::Identifier) &&
3212 (getTok().getIdentifier() == ".rept")) {
3216 // Otherwise, check whether we have reached the .endr.
3217 if (Lexer.is(AsmToken::Identifier) &&
3218 getTok().getIdentifier() == ".endr") {
3219 if (NestLevel == 0) {
3220 EndToken = getTok();
3222 if (Lexer.isNot(AsmToken::EndOfStatement)) {
3223 TokError("unexpected token in '.endr' directive");
3231 // Otherwise, scan till the end of the statement.
3232 EatToEndOfStatement();
3235 const char *BodyStart = StartToken.getLoc().getPointer();
3236 const char *BodyEnd = EndToken.getLoc().getPointer();
3237 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
3239 // We Are Anonymous.
3241 MacroParameters Parameters;
3242 return new Macro(Name, Body, Parameters);
3245 void AsmParser::InstantiateMacroLikeBody(Macro *M, SMLoc DirectiveLoc,
3246 raw_svector_ostream &OS) {
3249 MemoryBuffer *Instantiation =
3250 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
3252 // Create the macro instantiation object and add to the current macro
3253 // instantiation stack.
3254 MacroInstantiation *MI = new MacroInstantiation(M, DirectiveLoc,
3257 ActiveMacros.push_back(MI);
3259 // Jump to the macro instantiation and prime the lexer.
3260 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
3261 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
3265 bool AsmParser::ParseDirectiveRept(SMLoc DirectiveLoc) {
3267 if (ParseAbsoluteExpression(Count))
3268 return TokError("unexpected token in '.rept' directive");
3271 return TokError("Count is negative");
3273 if (Lexer.isNot(AsmToken::EndOfStatement))
3274 return TokError("unexpected token in '.rept' directive");
3276 // Eat the end of statement.
3279 // Lex the rept definition.
3280 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3284 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3285 // to hold the macro body with substitutions.
3286 SmallString<256> Buf;
3287 MacroParameters Parameters;
3289 raw_svector_ostream OS(Buf);
3291 if (expandMacro(OS, M->Body, Parameters, A, getTok().getLoc()))
3294 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3299 /// ParseDirectiveIrp
3300 /// ::= .irp symbol,values
3301 bool AsmParser::ParseDirectiveIrp(SMLoc DirectiveLoc) {
3302 MacroParameters Parameters;
3303 MacroParameter Parameter;
3305 if (ParseIdentifier(Parameter))
3306 return TokError("expected identifier in '.irp' directive");
3308 Parameters.push_back(Parameter);
3310 if (Lexer.isNot(AsmToken::Comma))
3311 return TokError("expected comma in '.irp' directive");
3316 if (ParseMacroArguments(0, A))
3319 // Eat the end of statement.
3322 // Lex the irp definition.
3323 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3327 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3328 // to hold the macro body with substitutions.
3329 SmallString<256> Buf;
3330 raw_svector_ostream OS(Buf);
3332 for (MacroArguments::iterator i = A.begin(), e = A.end(); i != e; ++i) {
3333 MacroArguments Args;
3336 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3340 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3345 /// ParseDirectiveIrpc
3346 /// ::= .irpc symbol,values
3347 bool AsmParser::ParseDirectiveIrpc(SMLoc DirectiveLoc) {
3348 MacroParameters Parameters;
3349 MacroParameter Parameter;
3351 if (ParseIdentifier(Parameter))
3352 return TokError("expected identifier in '.irpc' directive");
3354 Parameters.push_back(Parameter);
3356 if (Lexer.isNot(AsmToken::Comma))
3357 return TokError("expected comma in '.irpc' directive");
3362 if (ParseMacroArguments(0, A))
3365 if (A.size() != 1 || A.front().size() != 1)
3366 return TokError("unexpected token in '.irpc' directive");
3368 // Eat the end of statement.
3371 // Lex the irpc definition.
3372 Macro *M = ParseMacroLikeBody(DirectiveLoc);
3376 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3377 // to hold the macro body with substitutions.
3378 SmallString<256> Buf;
3379 raw_svector_ostream OS(Buf);
3381 StringRef Values = A.front().front().getString();
3382 std::size_t I, End = Values.size();
3383 for (I = 0; I < End; ++I) {
3385 Arg.push_back(AsmToken(AsmToken::Identifier, Values.slice(I, I+1)));
3387 MacroArguments Args;
3388 Args.push_back(Arg);
3390 if (expandMacro(OS, M->Body, Parameters, Args, getTok().getLoc()))
3394 InstantiateMacroLikeBody(M, DirectiveLoc, OS);
3399 bool AsmParser::ParseDirectiveEndr(SMLoc DirectiveLoc) {
3400 if (ActiveMacros.empty())
3401 return TokError("unexpected '.endr' directive, no current .rept");
3403 // The only .repl that should get here are the ones created by
3404 // InstantiateMacroLikeBody.
3405 assert(getLexer().is(AsmToken::EndOfStatement));
3411 /// \brief Create an MCAsmParser instance.
3412 MCAsmParser *llvm::createMCAsmParser(SourceMgr &SM,
3413 MCContext &C, MCStreamer &Out,
3414 const MCAsmInfo &MAI) {
3415 return new AsmParser(SM, C, Out, MAI);