Use %ull here.
[llvm/stm8.git] / lib / MC / MCParser / AsmParser.cpp
blob9d6c086c623f6fa021f37349234ab25eabcaa3a2
1 //===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This class implements the parser for assembly files.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/ADT/APFloat.h"
15 #include "llvm/ADT/SmallString.h"
16 #include "llvm/ADT/StringMap.h"
17 #include "llvm/ADT/StringSwitch.h"
18 #include "llvm/ADT/Twine.h"
19 #include "llvm/MC/MCAsmInfo.h"
20 #include "llvm/MC/MCContext.h"
21 #include "llvm/MC/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/MCSectionMachO.h"
27 #include "llvm/MC/MCStreamer.h"
28 #include "llvm/MC/MCSymbol.h"
29 #include "llvm/MC/MCDwarf.h"
30 #include "llvm/Support/MemoryBuffer.h"
31 #include "llvm/Support/SourceMgr.h"
32 #include "llvm/Support/raw_ostream.h"
33 #include "llvm/Target/TargetAsmInfo.h"
34 #include "llvm/Target/TargetAsmParser.h"
35 #include <cctype>
36 #include <vector>
37 using namespace llvm;
39 namespace {
41 /// \brief Helper class for tracking macro definitions.
42 struct Macro {
43 StringRef Name;
44 StringRef Body;
46 public:
47 Macro(StringRef N, StringRef B) : Name(N), Body(B) {}
50 /// \brief Helper class for storing information about an active macro
51 /// instantiation.
52 struct MacroInstantiation {
53 /// The macro being instantiated.
54 const Macro *TheMacro;
56 /// The macro instantiation with substitutions.
57 MemoryBuffer *Instantiation;
59 /// The location of the instantiation.
60 SMLoc InstantiationLoc;
62 /// The location where parsing should resume upon instantiation completion.
63 SMLoc ExitLoc;
65 public:
66 MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
67 const std::vector<std::vector<AsmToken> > &A);
70 /// \brief The concrete assembly parser instance.
71 class AsmParser : public MCAsmParser {
72 friend class GenericAsmParser;
74 AsmParser(const AsmParser &); // DO NOT IMPLEMENT
75 void operator=(const AsmParser &); // DO NOT IMPLEMENT
76 private:
77 AsmLexer Lexer;
78 MCContext &Ctx;
79 MCStreamer &Out;
80 SourceMgr &SrcMgr;
81 MCAsmParserExtension *GenericParser;
82 MCAsmParserExtension *PlatformParser;
84 /// This is the current buffer index we're lexing from as managed by the
85 /// SourceMgr object.
86 int CurBuffer;
88 AsmCond TheCondState;
89 std::vector<AsmCond> TheCondStack;
91 /// DirectiveMap - This is a table handlers for directives. Each handler is
92 /// invoked after the directive identifier is read and is responsible for
93 /// parsing and validating the rest of the directive. The handler is passed
94 /// in the directive name and the location of the directive keyword.
95 StringMap<std::pair<MCAsmParserExtension*, DirectiveHandler> > DirectiveMap;
97 /// MacroMap - Map of currently defined macros.
98 StringMap<Macro*> MacroMap;
100 /// ActiveMacros - Stack of active macro instantiations.
101 std::vector<MacroInstantiation*> ActiveMacros;
103 /// Boolean tracking whether macro substitution is enabled.
104 unsigned MacrosEnabled : 1;
106 /// Flag tracking whether any errors have been encountered.
107 unsigned HadError : 1;
109 public:
110 AsmParser(const Target &T, SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
111 const MCAsmInfo &MAI);
112 ~AsmParser();
114 virtual bool Run(bool NoInitialTextSection, bool NoFinalize = false);
116 void AddDirectiveHandler(MCAsmParserExtension *Object,
117 StringRef Directive,
118 DirectiveHandler Handler) {
119 DirectiveMap[Directive] = std::make_pair(Object, Handler);
122 public:
123 /// @name MCAsmParser Interface
124 /// {
126 virtual SourceMgr &getSourceManager() { return SrcMgr; }
127 virtual MCAsmLexer &getLexer() { return Lexer; }
128 virtual MCContext &getContext() { return Ctx; }
129 virtual MCStreamer &getStreamer() { return Out; }
131 virtual void Warning(SMLoc L, const Twine &Meg);
132 virtual bool Error(SMLoc L, const Twine &Msg);
134 const AsmToken &Lex();
136 bool ParseExpression(const MCExpr *&Res);
137 virtual bool ParseExpression(const MCExpr *&Res, SMLoc &EndLoc);
138 virtual bool ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc);
139 virtual bool ParseAbsoluteExpression(int64_t &Res);
141 /// }
143 private:
144 void CheckForValidSection();
146 bool ParseStatement();
148 bool HandleMacroEntry(StringRef Name, SMLoc NameLoc, const Macro *M);
149 void HandleMacroExit();
151 void PrintMacroInstantiations();
152 void PrintMessage(SMLoc Loc, const Twine &Msg, const char *Type) const {
153 SrcMgr.PrintMessage(Loc, Msg, Type);
156 /// EnterIncludeFile - Enter the specified file. This returns true on failure.
157 bool EnterIncludeFile(const std::string &Filename);
159 /// \brief Reset the current lexer position to that given by \arg Loc. The
160 /// current token is not set; clients should ensure Lex() is called
161 /// subsequently.
162 void JumpToLoc(SMLoc Loc);
164 void EatToEndOfStatement();
166 /// \brief Parse up to the end of statement and a return the contents from the
167 /// current token until the end of the statement; the current token on exit
168 /// will be either the EndOfStatement or EOF.
169 StringRef ParseStringToEndOfStatement();
171 bool ParseAssignment(StringRef Name, bool allow_redef);
173 bool ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc);
174 bool ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
175 bool ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
176 bool ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
178 /// ParseIdentifier - Parse an identifier or string (as a quoted identifier)
179 /// and set \arg Res to the identifier contents.
180 bool ParseIdentifier(StringRef &Res);
182 // Directive Parsing.
184 // ".ascii", ".asciiz", ".string"
185 bool ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
186 bool ParseDirectiveValue(unsigned Size); // ".byte", ".long", ...
187 bool ParseDirectiveRealValue(const fltSemantics &); // ".single", ...
188 bool ParseDirectiveFill(); // ".fill"
189 bool ParseDirectiveSpace(); // ".space"
190 bool ParseDirectiveZero(); // ".zero"
191 bool ParseDirectiveSet(StringRef IDVal, bool allow_redef); // ".set", ".equ", ".equiv"
192 bool ParseDirectiveOrg(); // ".org"
193 // ".align{,32}", ".p2align{,w,l}"
194 bool ParseDirectiveAlign(bool IsPow2, unsigned ValueSize);
196 /// ParseDirectiveSymbolAttribute - Parse a directive like ".globl" which
197 /// accepts a single symbol (which should be a label or an external).
198 bool ParseDirectiveSymbolAttribute(MCSymbolAttr Attr);
200 bool ParseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
202 bool ParseDirectiveAbort(); // ".abort"
203 bool ParseDirectiveInclude(); // ".include"
205 bool ParseDirectiveIf(SMLoc DirectiveLoc); // ".if"
206 // ".ifdef" or ".ifndef", depending on expect_defined
207 bool ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
208 bool ParseDirectiveElseIf(SMLoc DirectiveLoc); // ".elseif"
209 bool ParseDirectiveElse(SMLoc DirectiveLoc); // ".else"
210 bool ParseDirectiveEndIf(SMLoc DirectiveLoc); // .endif
212 /// ParseEscapedString - Parse the current token as a string which may include
213 /// escaped characters and return the string contents.
214 bool ParseEscapedString(std::string &Data);
216 const MCExpr *ApplyModifierToExpr(const MCExpr *E,
217 MCSymbolRefExpr::VariantKind Variant);
220 /// \brief Generic implementations of directive handling, etc. which is shared
221 /// (or the default, at least) for all assembler parser.
222 class GenericAsmParser : public MCAsmParserExtension {
223 template<bool (GenericAsmParser::*Handler)(StringRef, SMLoc)>
224 void AddDirectiveHandler(StringRef Directive) {
225 getParser().AddDirectiveHandler(this, Directive,
226 HandleDirective<GenericAsmParser, Handler>);
228 public:
229 GenericAsmParser() {}
231 AsmParser &getParser() {
232 return (AsmParser&) this->MCAsmParserExtension::getParser();
235 virtual void Initialize(MCAsmParser &Parser) {
236 // Call the base implementation.
237 this->MCAsmParserExtension::Initialize(Parser);
239 // Debugging directives.
240 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveFile>(".file");
241 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLine>(".line");
242 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLoc>(".loc");
243 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveStabs>(".stabs");
245 // CFI directives.
246 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIStartProc>(
247 ".cfi_startproc");
248 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIEndProc>(
249 ".cfi_endproc");
250 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfa>(
251 ".cfi_def_cfa");
252 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaOffset>(
253 ".cfi_def_cfa_offset");
254 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset>(
255 ".cfi_adjust_cfa_offset");
256 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIDefCfaRegister>(
257 ".cfi_def_cfa_register");
258 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIOffset>(
259 ".cfi_offset");
260 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveCFIRelOffset>(
261 ".cfi_rel_offset");
262 AddDirectiveHandler<
263 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_personality");
264 AddDirectiveHandler<
265 &GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda>(".cfi_lsda");
266 AddDirectiveHandler<
267 &GenericAsmParser::ParseDirectiveCFIRememberState>(".cfi_remember_state");
268 AddDirectiveHandler<
269 &GenericAsmParser::ParseDirectiveCFIRestoreState>(".cfi_restore_state");
270 AddDirectiveHandler<
271 &GenericAsmParser::ParseDirectiveCFISameValue>(".cfi_same_value");
273 // Macro directives.
274 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
275 ".macros_on");
276 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacrosOnOff>(
277 ".macros_off");
278 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveMacro>(".macro");
279 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endm");
280 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveEndMacro>(".endmacro");
282 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".sleb128");
283 AddDirectiveHandler<&GenericAsmParser::ParseDirectiveLEB128>(".uleb128");
286 bool ParseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
288 bool ParseDirectiveFile(StringRef, SMLoc DirectiveLoc);
289 bool ParseDirectiveLine(StringRef, SMLoc DirectiveLoc);
290 bool ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc);
291 bool ParseDirectiveStabs(StringRef, SMLoc DirectiveLoc);
292 bool ParseDirectiveCFIStartProc(StringRef, SMLoc DirectiveLoc);
293 bool ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc);
294 bool ParseDirectiveCFIDefCfa(StringRef, SMLoc DirectiveLoc);
295 bool ParseDirectiveCFIDefCfaOffset(StringRef, SMLoc DirectiveLoc);
296 bool ParseDirectiveCFIAdjustCfaOffset(StringRef, SMLoc DirectiveLoc);
297 bool ParseDirectiveCFIDefCfaRegister(StringRef, SMLoc DirectiveLoc);
298 bool ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc);
299 bool ParseDirectiveCFIRelOffset(StringRef, SMLoc DirectiveLoc);
300 bool ParseDirectiveCFIPersonalityOrLsda(StringRef, SMLoc DirectiveLoc);
301 bool ParseDirectiveCFIRememberState(StringRef, SMLoc DirectiveLoc);
302 bool ParseDirectiveCFIRestoreState(StringRef, SMLoc DirectiveLoc);
303 bool ParseDirectiveCFISameValue(StringRef, SMLoc DirectiveLoc);
305 bool ParseDirectiveMacrosOnOff(StringRef, SMLoc DirectiveLoc);
306 bool ParseDirectiveMacro(StringRef, SMLoc DirectiveLoc);
307 bool ParseDirectiveEndMacro(StringRef, SMLoc DirectiveLoc);
309 bool ParseDirectiveLEB128(StringRef, SMLoc);
314 namespace llvm {
316 extern MCAsmParserExtension *createDarwinAsmParser();
317 extern MCAsmParserExtension *createELFAsmParser();
318 extern MCAsmParserExtension *createCOFFAsmParser();
322 enum { DEFAULT_ADDRSPACE = 0 };
324 AsmParser::AsmParser(const Target &T, SourceMgr &_SM, MCContext &_Ctx,
325 MCStreamer &_Out, const MCAsmInfo &_MAI)
326 : Lexer(_MAI), Ctx(_Ctx), Out(_Out), SrcMgr(_SM),
327 GenericParser(new GenericAsmParser), PlatformParser(0),
328 CurBuffer(0), MacrosEnabled(true) {
329 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
331 // Initialize the generic parser.
332 GenericParser->Initialize(*this);
334 // Initialize the platform / file format parser.
336 // FIXME: This is a hack, we need to (majorly) cleanup how these objects are
337 // created.
338 if (_MAI.hasMicrosoftFastStdCallMangling()) {
339 PlatformParser = createCOFFAsmParser();
340 PlatformParser->Initialize(*this);
341 } else if (_MAI.hasSubsectionsViaSymbols()) {
342 PlatformParser = createDarwinAsmParser();
343 PlatformParser->Initialize(*this);
344 } else {
345 PlatformParser = createELFAsmParser();
346 PlatformParser->Initialize(*this);
350 AsmParser::~AsmParser() {
351 assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
353 // Destroy any macros.
354 for (StringMap<Macro*>::iterator it = MacroMap.begin(),
355 ie = MacroMap.end(); it != ie; ++it)
356 delete it->getValue();
358 delete PlatformParser;
359 delete GenericParser;
362 void AsmParser::PrintMacroInstantiations() {
363 // Print the active macro instantiation stack.
364 for (std::vector<MacroInstantiation*>::const_reverse_iterator
365 it = ActiveMacros.rbegin(), ie = ActiveMacros.rend(); it != ie; ++it)
366 PrintMessage((*it)->InstantiationLoc, "while in macro instantiation",
367 "note");
370 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
371 PrintMessage(L, Msg, "warning");
372 PrintMacroInstantiations();
375 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
376 HadError = true;
377 PrintMessage(L, Msg, "error");
378 PrintMacroInstantiations();
379 return true;
382 bool AsmParser::EnterIncludeFile(const std::string &Filename) {
383 int NewBuf = SrcMgr.AddIncludeFile(Filename, Lexer.getLoc());
384 if (NewBuf == -1)
385 return true;
387 CurBuffer = NewBuf;
389 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
391 return false;
394 void AsmParser::JumpToLoc(SMLoc Loc) {
395 CurBuffer = SrcMgr.FindBufferContainingLoc(Loc);
396 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer), Loc.getPointer());
399 const AsmToken &AsmParser::Lex() {
400 const AsmToken *tok = &Lexer.Lex();
402 if (tok->is(AsmToken::Eof)) {
403 // If this is the end of an included file, pop the parent file off the
404 // include stack.
405 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
406 if (ParentIncludeLoc != SMLoc()) {
407 JumpToLoc(ParentIncludeLoc);
408 tok = &Lexer.Lex();
412 if (tok->is(AsmToken::Error))
413 Error(Lexer.getErrLoc(), Lexer.getErr());
415 return *tok;
418 bool AsmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
419 // Create the initial section, if requested.
420 if (!NoInitialTextSection)
421 Out.InitSections();
423 // Prime the lexer.
424 Lex();
426 HadError = false;
427 AsmCond StartingCondState = TheCondState;
429 // While we have input, parse each statement.
430 while (Lexer.isNot(AsmToken::Eof)) {
431 if (!ParseStatement()) continue;
433 // We had an error, validate that one was emitted and recover by skipping to
434 // the next line.
435 assert(HadError && "Parse statement returned an error, but none emitted!");
436 EatToEndOfStatement();
439 if (TheCondState.TheCond != StartingCondState.TheCond ||
440 TheCondState.Ignore != StartingCondState.Ignore)
441 return TokError("unmatched .ifs or .elses");
443 // Check to see there are no empty DwarfFile slots.
444 const std::vector<MCDwarfFile *> &MCDwarfFiles =
445 getContext().getMCDwarfFiles();
446 for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
447 if (!MCDwarfFiles[i])
448 TokError("unassigned file number: " + Twine(i) + " for .file directives");
451 // Finalize the output stream if there are no errors and if the client wants
452 // us to.
453 if (!HadError && !NoFinalize)
454 Out.Finish();
456 return HadError;
459 void AsmParser::CheckForValidSection() {
460 if (!getStreamer().getCurrentSection()) {
461 TokError("expected section directive before assembly directive");
462 Out.SwitchSection(Ctx.getMachOSection(
463 "__TEXT", "__text",
464 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
465 0, SectionKind::getText()));
469 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
470 void AsmParser::EatToEndOfStatement() {
471 while (Lexer.isNot(AsmToken::EndOfStatement) &&
472 Lexer.isNot(AsmToken::Eof))
473 Lex();
475 // Eat EOL.
476 if (Lexer.is(AsmToken::EndOfStatement))
477 Lex();
480 StringRef AsmParser::ParseStringToEndOfStatement() {
481 const char *Start = getTok().getLoc().getPointer();
483 while (Lexer.isNot(AsmToken::EndOfStatement) &&
484 Lexer.isNot(AsmToken::Eof))
485 Lex();
487 const char *End = getTok().getLoc().getPointer();
488 return StringRef(Start, End - Start);
491 /// ParseParenExpr - Parse a paren expression and return it.
492 /// NOTE: This assumes the leading '(' has already been consumed.
494 /// parenexpr ::= expr)
496 bool AsmParser::ParseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
497 if (ParseExpression(Res)) return true;
498 if (Lexer.isNot(AsmToken::RParen))
499 return TokError("expected ')' in parentheses expression");
500 EndLoc = Lexer.getLoc();
501 Lex();
502 return false;
505 /// ParseBracketExpr - Parse a bracket expression and return it.
506 /// NOTE: This assumes the leading '[' has already been consumed.
508 /// bracketexpr ::= expr]
510 bool AsmParser::ParseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
511 if (ParseExpression(Res)) return true;
512 if (Lexer.isNot(AsmToken::RBrac))
513 return TokError("expected ']' in brackets expression");
514 EndLoc = Lexer.getLoc();
515 Lex();
516 return false;
519 /// ParsePrimaryExpr - Parse a primary expression and return it.
520 /// primaryexpr ::= (parenexpr
521 /// primaryexpr ::= symbol
522 /// primaryexpr ::= number
523 /// primaryexpr ::= '.'
524 /// primaryexpr ::= ~,+,- primaryexpr
525 bool AsmParser::ParsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc) {
526 switch (Lexer.getKind()) {
527 default:
528 return TokError("unknown token in expression");
529 // If we have an error assume that we've already handled it.
530 case AsmToken::Error:
531 return true;
532 case AsmToken::Exclaim:
533 Lex(); // Eat the operator.
534 if (ParsePrimaryExpr(Res, EndLoc))
535 return true;
536 Res = MCUnaryExpr::CreateLNot(Res, getContext());
537 return false;
538 case AsmToken::Dollar:
539 case AsmToken::String:
540 case AsmToken::Identifier: {
541 EndLoc = Lexer.getLoc();
543 StringRef Identifier;
544 if (ParseIdentifier(Identifier))
545 return false;
547 // This is a symbol reference.
548 std::pair<StringRef, StringRef> Split = Identifier.split('@');
549 MCSymbol *Sym = getContext().GetOrCreateSymbol(Split.first);
551 // Lookup the symbol variant if used.
552 MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None;
553 if (Split.first.size() != Identifier.size()) {
554 Variant = MCSymbolRefExpr::getVariantKindForName(Split.second);
555 if (Variant == MCSymbolRefExpr::VK_Invalid) {
556 Variant = MCSymbolRefExpr::VK_None;
557 return TokError("invalid variant '" + Split.second + "'");
561 // If this is an absolute variable reference, substitute it now to preserve
562 // semantics in the face of reassignment.
563 if (Sym->isVariable() && isa<MCConstantExpr>(Sym->getVariableValue())) {
564 if (Variant)
565 return Error(EndLoc, "unexpected modifier on variable reference");
567 Res = Sym->getVariableValue();
568 return false;
571 // Otherwise create a symbol ref.
572 Res = MCSymbolRefExpr::Create(Sym, Variant, getContext());
573 return false;
575 case AsmToken::Integer: {
576 SMLoc Loc = getTok().getLoc();
577 int64_t IntVal = getTok().getIntVal();
578 Res = MCConstantExpr::Create(IntVal, getContext());
579 EndLoc = Lexer.getLoc();
580 Lex(); // Eat token.
581 // Look for 'b' or 'f' following an Integer as a directional label
582 if (Lexer.getKind() == AsmToken::Identifier) {
583 StringRef IDVal = getTok().getString();
584 if (IDVal == "f" || IDVal == "b"){
585 MCSymbol *Sym = Ctx.GetDirectionalLocalSymbol(IntVal,
586 IDVal == "f" ? 1 : 0);
587 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None,
588 getContext());
589 if(IDVal == "b" && Sym->isUndefined())
590 return Error(Loc, "invalid reference to undefined symbol");
591 EndLoc = Lexer.getLoc();
592 Lex(); // Eat identifier.
595 return false;
597 case AsmToken::Real: {
598 APFloat RealVal(APFloat::IEEEdouble, getTok().getString());
599 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
600 Res = MCConstantExpr::Create(IntVal, getContext());
601 Lex(); // Eat token.
602 return false;
604 case AsmToken::Dot: {
605 // This is a '.' reference, which references the current PC. Emit a
606 // temporary label to the streamer and refer to it.
607 MCSymbol *Sym = Ctx.CreateTempSymbol();
608 Out.EmitLabel(Sym);
609 Res = MCSymbolRefExpr::Create(Sym, MCSymbolRefExpr::VK_None, getContext());
610 EndLoc = Lexer.getLoc();
611 Lex(); // Eat identifier.
612 return false;
614 case AsmToken::LParen:
615 Lex(); // Eat the '('.
616 return ParseParenExpr(Res, EndLoc);
617 case AsmToken::LBrac:
618 if (!PlatformParser->HasBracketExpressions())
619 return TokError("brackets expression not supported on this target");
620 Lex(); // Eat the '['.
621 return ParseBracketExpr(Res, EndLoc);
622 case AsmToken::Minus:
623 Lex(); // Eat the operator.
624 if (ParsePrimaryExpr(Res, EndLoc))
625 return true;
626 Res = MCUnaryExpr::CreateMinus(Res, getContext());
627 return false;
628 case AsmToken::Plus:
629 Lex(); // Eat the operator.
630 if (ParsePrimaryExpr(Res, EndLoc))
631 return true;
632 Res = MCUnaryExpr::CreatePlus(Res, getContext());
633 return false;
634 case AsmToken::Tilde:
635 Lex(); // Eat the operator.
636 if (ParsePrimaryExpr(Res, EndLoc))
637 return true;
638 Res = MCUnaryExpr::CreateNot(Res, getContext());
639 return false;
643 bool AsmParser::ParseExpression(const MCExpr *&Res) {
644 SMLoc EndLoc;
645 return ParseExpression(Res, EndLoc);
648 const MCExpr *
649 AsmParser::ApplyModifierToExpr(const MCExpr *E,
650 MCSymbolRefExpr::VariantKind Variant) {
651 // Recurse over the given expression, rebuilding it to apply the given variant
652 // if there is exactly one symbol.
653 switch (E->getKind()) {
654 case MCExpr::Target:
655 case MCExpr::Constant:
656 return 0;
658 case MCExpr::SymbolRef: {
659 const MCSymbolRefExpr *SRE = cast<MCSymbolRefExpr>(E);
661 if (SRE->getKind() != MCSymbolRefExpr::VK_None) {
662 TokError("invalid variant on expression '" +
663 getTok().getIdentifier() + "' (already modified)");
664 return E;
667 return MCSymbolRefExpr::Create(&SRE->getSymbol(), Variant, getContext());
670 case MCExpr::Unary: {
671 const MCUnaryExpr *UE = cast<MCUnaryExpr>(E);
672 const MCExpr *Sub = ApplyModifierToExpr(UE->getSubExpr(), Variant);
673 if (!Sub)
674 return 0;
675 return MCUnaryExpr::Create(UE->getOpcode(), Sub, getContext());
678 case MCExpr::Binary: {
679 const MCBinaryExpr *BE = cast<MCBinaryExpr>(E);
680 const MCExpr *LHS = ApplyModifierToExpr(BE->getLHS(), Variant);
681 const MCExpr *RHS = ApplyModifierToExpr(BE->getRHS(), Variant);
683 if (!LHS && !RHS)
684 return 0;
686 if (!LHS) LHS = BE->getLHS();
687 if (!RHS) RHS = BE->getRHS();
689 return MCBinaryExpr::Create(BE->getOpcode(), LHS, RHS, getContext());
693 assert(0 && "Invalid expression kind!");
694 return 0;
697 /// ParseExpression - Parse an expression and return it.
699 /// expr ::= expr +,- expr -> lowest.
700 /// expr ::= expr |,^,&,! expr -> middle.
701 /// expr ::= expr *,/,%,<<,>> expr -> highest.
702 /// expr ::= primaryexpr
704 bool AsmParser::ParseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
705 // Parse the expression.
706 Res = 0;
707 if (ParsePrimaryExpr(Res, EndLoc) || ParseBinOpRHS(1, Res, EndLoc))
708 return true;
710 // As a special case, we support 'a op b @ modifier' by rewriting the
711 // expression to include the modifier. This is inefficient, but in general we
712 // expect users to use 'a@modifier op b'.
713 if (Lexer.getKind() == AsmToken::At) {
714 Lex();
716 if (Lexer.isNot(AsmToken::Identifier))
717 return TokError("unexpected symbol modifier following '@'");
719 MCSymbolRefExpr::VariantKind Variant =
720 MCSymbolRefExpr::getVariantKindForName(getTok().getIdentifier());
721 if (Variant == MCSymbolRefExpr::VK_Invalid)
722 return TokError("invalid variant '" + getTok().getIdentifier() + "'");
724 const MCExpr *ModifiedRes = ApplyModifierToExpr(Res, Variant);
725 if (!ModifiedRes) {
726 return TokError("invalid modifier '" + getTok().getIdentifier() +
727 "' (no symbols present)");
728 return true;
731 Res = ModifiedRes;
732 Lex();
735 // Try to constant fold it up front, if possible.
736 int64_t Value;
737 if (Res->EvaluateAsAbsolute(Value))
738 Res = MCConstantExpr::Create(Value, getContext());
740 return false;
743 bool AsmParser::ParseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
744 Res = 0;
745 return ParseParenExpr(Res, EndLoc) ||
746 ParseBinOpRHS(1, Res, EndLoc);
749 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
750 const MCExpr *Expr;
752 SMLoc StartLoc = Lexer.getLoc();
753 if (ParseExpression(Expr))
754 return true;
756 if (!Expr->EvaluateAsAbsolute(Res))
757 return Error(StartLoc, "expected absolute expression");
759 return false;
762 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
763 MCBinaryExpr::Opcode &Kind) {
764 switch (K) {
765 default:
766 return 0; // not a binop.
768 // Lowest Precedence: &&, ||, @
769 case AsmToken::AmpAmp:
770 Kind = MCBinaryExpr::LAnd;
771 return 1;
772 case AsmToken::PipePipe:
773 Kind = MCBinaryExpr::LOr;
774 return 1;
777 // Low Precedence: |, &, ^
779 // FIXME: gas seems to support '!' as an infix operator?
780 case AsmToken::Pipe:
781 Kind = MCBinaryExpr::Or;
782 return 2;
783 case AsmToken::Caret:
784 Kind = MCBinaryExpr::Xor;
785 return 2;
786 case AsmToken::Amp:
787 Kind = MCBinaryExpr::And;
788 return 2;
790 // Low Intermediate Precedence: ==, !=, <>, <, <=, >, >=
791 case AsmToken::EqualEqual:
792 Kind = MCBinaryExpr::EQ;
793 return 3;
794 case AsmToken::ExclaimEqual:
795 case AsmToken::LessGreater:
796 Kind = MCBinaryExpr::NE;
797 return 3;
798 case AsmToken::Less:
799 Kind = MCBinaryExpr::LT;
800 return 3;
801 case AsmToken::LessEqual:
802 Kind = MCBinaryExpr::LTE;
803 return 3;
804 case AsmToken::Greater:
805 Kind = MCBinaryExpr::GT;
806 return 3;
807 case AsmToken::GreaterEqual:
808 Kind = MCBinaryExpr::GTE;
809 return 3;
811 // High Intermediate Precedence: +, -
812 case AsmToken::Plus:
813 Kind = MCBinaryExpr::Add;
814 return 4;
815 case AsmToken::Minus:
816 Kind = MCBinaryExpr::Sub;
817 return 4;
819 // Highest Precedence: *, /, %, <<, >>
820 case AsmToken::Star:
821 Kind = MCBinaryExpr::Mul;
822 return 5;
823 case AsmToken::Slash:
824 Kind = MCBinaryExpr::Div;
825 return 5;
826 case AsmToken::Percent:
827 Kind = MCBinaryExpr::Mod;
828 return 5;
829 case AsmToken::LessLess:
830 Kind = MCBinaryExpr::Shl;
831 return 5;
832 case AsmToken::GreaterGreater:
833 Kind = MCBinaryExpr::Shr;
834 return 5;
839 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
840 /// Res contains the LHS of the expression on input.
841 bool AsmParser::ParseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
842 SMLoc &EndLoc) {
843 while (1) {
844 MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add;
845 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
847 // If the next token is lower precedence than we are allowed to eat, return
848 // successfully with what we ate already.
849 if (TokPrec < Precedence)
850 return false;
852 Lex();
854 // Eat the next primary expression.
855 const MCExpr *RHS;
856 if (ParsePrimaryExpr(RHS, EndLoc)) return true;
858 // If BinOp binds less tightly with RHS than the operator after RHS, let
859 // the pending operator take RHS as its LHS.
860 MCBinaryExpr::Opcode Dummy;
861 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
862 if (TokPrec < NextTokPrec) {
863 if (ParseBinOpRHS(Precedence+1, RHS, EndLoc)) return true;
866 // Merge LHS and RHS according to operator.
867 Res = MCBinaryExpr::Create(Kind, Res, RHS, getContext());
874 /// ParseStatement:
875 /// ::= EndOfStatement
876 /// ::= Label* Directive ...Operands... EndOfStatement
877 /// ::= Label* Identifier OperandList* EndOfStatement
878 bool AsmParser::ParseStatement() {
879 if (Lexer.is(AsmToken::EndOfStatement)) {
880 Out.AddBlankLine();
881 Lex();
882 return false;
885 // Statements always start with an identifier or are a full line comment.
886 AsmToken ID = getTok();
887 SMLoc IDLoc = ID.getLoc();
888 StringRef IDVal;
889 int64_t LocalLabelVal = -1;
890 // A full line comment is a '#' as the first token.
891 if (Lexer.is(AsmToken::Hash)) {
892 EatToEndOfStatement();
893 return false;
896 // Allow an integer followed by a ':' as a directional local label.
897 if (Lexer.is(AsmToken::Integer)) {
898 LocalLabelVal = getTok().getIntVal();
899 if (LocalLabelVal < 0) {
900 if (!TheCondState.Ignore)
901 return TokError("unexpected token at start of statement");
902 IDVal = "";
904 else {
905 IDVal = getTok().getString();
906 Lex(); // Consume the integer token to be used as an identifier token.
907 if (Lexer.getKind() != AsmToken::Colon) {
908 if (!TheCondState.Ignore)
909 return TokError("unexpected token at start of statement");
913 } else if (Lexer.is(AsmToken::Dot)) {
914 // Treat '.' as a valid identifier in this context.
915 Lex();
916 IDVal = ".";
918 } else if (ParseIdentifier(IDVal)) {
919 if (!TheCondState.Ignore)
920 return TokError("unexpected token at start of statement");
921 IDVal = "";
925 // Handle conditional assembly here before checking for skipping. We
926 // have to do this so that .endif isn't skipped in a ".if 0" block for
927 // example.
928 if (IDVal == ".if")
929 return ParseDirectiveIf(IDLoc);
930 if (IDVal == ".ifdef")
931 return ParseDirectiveIfdef(IDLoc, true);
932 if (IDVal == ".ifndef" || IDVal == ".ifnotdef")
933 return ParseDirectiveIfdef(IDLoc, false);
934 if (IDVal == ".elseif")
935 return ParseDirectiveElseIf(IDLoc);
936 if (IDVal == ".else")
937 return ParseDirectiveElse(IDLoc);
938 if (IDVal == ".endif")
939 return ParseDirectiveEndIf(IDLoc);
941 // If we are in a ".if 0" block, ignore this statement.
942 if (TheCondState.Ignore) {
943 EatToEndOfStatement();
944 return false;
947 // FIXME: Recurse on local labels?
949 // See what kind of statement we have.
950 switch (Lexer.getKind()) {
951 case AsmToken::Colon: {
952 CheckForValidSection();
954 // identifier ':' -> Label.
955 Lex();
957 // Diagnose attempt to use '.' as a label.
958 if (IDVal == ".")
959 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
961 // Diagnose attempt to use a variable as a label.
963 // FIXME: Diagnostics. Note the location of the definition as a label.
964 // FIXME: This doesn't diagnose assignment to a symbol which has been
965 // implicitly marked as external.
966 MCSymbol *Sym;
967 if (LocalLabelVal == -1)
968 Sym = getContext().GetOrCreateSymbol(IDVal);
969 else
970 Sym = Ctx.CreateDirectionalLocalSymbol(LocalLabelVal);
971 if (!Sym->isUndefined() || Sym->isVariable())
972 return Error(IDLoc, "invalid symbol redefinition");
974 // Emit the label.
975 Out.EmitLabel(Sym);
977 // Consume any end of statement token, if present, to avoid spurious
978 // AddBlankLine calls().
979 if (Lexer.is(AsmToken::EndOfStatement)) {
980 Lex();
981 if (Lexer.is(AsmToken::Eof))
982 return false;
985 return ParseStatement();
988 case AsmToken::Equal:
989 // identifier '=' ... -> assignment statement
990 Lex();
992 return ParseAssignment(IDVal, true);
994 default: // Normal instruction or directive.
995 break;
998 // If macros are enabled, check to see if this is a macro instantiation.
999 if (MacrosEnabled)
1000 if (const Macro *M = MacroMap.lookup(IDVal))
1001 return HandleMacroEntry(IDVal, IDLoc, M);
1003 // Otherwise, we have a normal instruction or directive.
1004 if (IDVal[0] == '.' && IDVal != ".") {
1005 // Assembler features
1006 if (IDVal == ".set" || IDVal == ".equ")
1007 return ParseDirectiveSet(IDVal, true);
1008 if (IDVal == ".equiv")
1009 return ParseDirectiveSet(IDVal, false);
1011 // Data directives
1013 if (IDVal == ".ascii")
1014 return ParseDirectiveAscii(IDVal, false);
1015 if (IDVal == ".asciz" || IDVal == ".string")
1016 return ParseDirectiveAscii(IDVal, true);
1018 if (IDVal == ".byte")
1019 return ParseDirectiveValue(1);
1020 if (IDVal == ".short")
1021 return ParseDirectiveValue(2);
1022 if (IDVal == ".value")
1023 return ParseDirectiveValue(2);
1024 if (IDVal == ".2byte")
1025 return ParseDirectiveValue(2);
1026 if (IDVal == ".long")
1027 return ParseDirectiveValue(4);
1028 if (IDVal == ".int")
1029 return ParseDirectiveValue(4);
1030 if (IDVal == ".4byte")
1031 return ParseDirectiveValue(4);
1032 if (IDVal == ".quad")
1033 return ParseDirectiveValue(8);
1034 if (IDVal == ".8byte")
1035 return ParseDirectiveValue(8);
1036 if (IDVal == ".single" || IDVal == ".float")
1037 return ParseDirectiveRealValue(APFloat::IEEEsingle);
1038 if (IDVal == ".double")
1039 return ParseDirectiveRealValue(APFloat::IEEEdouble);
1041 if (IDVal == ".align") {
1042 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1043 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/1);
1045 if (IDVal == ".align32") {
1046 bool IsPow2 = !getContext().getAsmInfo().getAlignmentIsInBytes();
1047 return ParseDirectiveAlign(IsPow2, /*ExprSize=*/4);
1049 if (IDVal == ".balign")
1050 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
1051 if (IDVal == ".balignw")
1052 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
1053 if (IDVal == ".balignl")
1054 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
1055 if (IDVal == ".p2align")
1056 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
1057 if (IDVal == ".p2alignw")
1058 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
1059 if (IDVal == ".p2alignl")
1060 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
1062 if (IDVal == ".org")
1063 return ParseDirectiveOrg();
1065 if (IDVal == ".fill")
1066 return ParseDirectiveFill();
1067 if (IDVal == ".space" || IDVal == ".skip")
1068 return ParseDirectiveSpace();
1069 if (IDVal == ".zero")
1070 return ParseDirectiveZero();
1072 // Symbol attribute directives
1074 if (IDVal == ".globl" || IDVal == ".global")
1075 return ParseDirectiveSymbolAttribute(MCSA_Global);
1076 // ELF only? Should it be here?
1077 if (IDVal == ".local")
1078 return ParseDirectiveSymbolAttribute(MCSA_Local);
1079 if (IDVal == ".hidden")
1080 return ParseDirectiveSymbolAttribute(MCSA_Hidden);
1081 if (IDVal == ".indirect_symbol")
1082 return ParseDirectiveSymbolAttribute(MCSA_IndirectSymbol);
1083 if (IDVal == ".internal")
1084 return ParseDirectiveSymbolAttribute(MCSA_Internal);
1085 if (IDVal == ".lazy_reference")
1086 return ParseDirectiveSymbolAttribute(MCSA_LazyReference);
1087 if (IDVal == ".no_dead_strip")
1088 return ParseDirectiveSymbolAttribute(MCSA_NoDeadStrip);
1089 if (IDVal == ".symbol_resolver")
1090 return ParseDirectiveSymbolAttribute(MCSA_SymbolResolver);
1091 if (IDVal == ".private_extern")
1092 return ParseDirectiveSymbolAttribute(MCSA_PrivateExtern);
1093 if (IDVal == ".protected")
1094 return ParseDirectiveSymbolAttribute(MCSA_Protected);
1095 if (IDVal == ".reference")
1096 return ParseDirectiveSymbolAttribute(MCSA_Reference);
1097 if (IDVal == ".weak")
1098 return ParseDirectiveSymbolAttribute(MCSA_Weak);
1099 if (IDVal == ".weak_definition")
1100 return ParseDirectiveSymbolAttribute(MCSA_WeakDefinition);
1101 if (IDVal == ".weak_reference")
1102 return ParseDirectiveSymbolAttribute(MCSA_WeakReference);
1103 if (IDVal == ".weak_def_can_be_hidden")
1104 return ParseDirectiveSymbolAttribute(MCSA_WeakDefAutoPrivate);
1106 if (IDVal == ".comm")
1107 return ParseDirectiveComm(/*IsLocal=*/false);
1108 if (IDVal == ".lcomm")
1109 return ParseDirectiveComm(/*IsLocal=*/true);
1111 if (IDVal == ".abort")
1112 return ParseDirectiveAbort();
1113 if (IDVal == ".include")
1114 return ParseDirectiveInclude();
1116 if (IDVal == ".code16" || IDVal == ".code32" || IDVal == ".code64")
1117 return TokError(Twine(IDVal) + " not supported yet");
1119 // Look up the handler in the handler table.
1120 std::pair<MCAsmParserExtension*, DirectiveHandler> Handler =
1121 DirectiveMap.lookup(IDVal);
1122 if (Handler.first)
1123 return (*Handler.second)(Handler.first, IDVal, IDLoc);
1125 // Target hook for parsing target specific directives.
1126 if (!getTargetParser().ParseDirective(ID))
1127 return false;
1129 Warning(IDLoc, "ignoring directive for now");
1130 EatToEndOfStatement();
1131 return false;
1134 CheckForValidSection();
1136 // Canonicalize the opcode to lower case.
1137 SmallString<128> Opcode;
1138 for (unsigned i = 0, e = IDVal.size(); i != e; ++i)
1139 Opcode.push_back(tolower(IDVal[i]));
1141 SmallVector<MCParsedAsmOperand*, 8> ParsedOperands;
1142 bool HadError = getTargetParser().ParseInstruction(Opcode.str(), IDLoc,
1143 ParsedOperands);
1145 // Dump the parsed representation, if requested.
1146 if (getShowParsedOperands()) {
1147 SmallString<256> Str;
1148 raw_svector_ostream OS(Str);
1149 OS << "parsed instruction: [";
1150 for (unsigned i = 0; i != ParsedOperands.size(); ++i) {
1151 if (i != 0)
1152 OS << ", ";
1153 ParsedOperands[i]->dump(OS);
1155 OS << "]";
1157 PrintMessage(IDLoc, OS.str(), "note");
1160 // If parsing succeeded, match the instruction.
1161 if (!HadError)
1162 HadError = getTargetParser().MatchAndEmitInstruction(IDLoc, ParsedOperands,
1163 Out);
1165 // Free any parsed operands.
1166 for (unsigned i = 0, e = ParsedOperands.size(); i != e; ++i)
1167 delete ParsedOperands[i];
1169 // Don't skip the rest of the line, the instruction parser is responsible for
1170 // that.
1171 return false;
1174 MacroInstantiation::MacroInstantiation(const Macro *M, SMLoc IL, SMLoc EL,
1175 const std::vector<std::vector<AsmToken> > &A)
1176 : TheMacro(M), InstantiationLoc(IL), ExitLoc(EL)
1178 // Macro instantiation is lexical, unfortunately. We construct a new buffer
1179 // to hold the macro body with substitutions.
1180 SmallString<256> Buf;
1181 raw_svector_ostream OS(Buf);
1183 StringRef Body = M->Body;
1184 while (!Body.empty()) {
1185 // Scan for the next substitution.
1186 std::size_t End = Body.size(), Pos = 0;
1187 for (; Pos != End; ++Pos) {
1188 // Check for a substitution or escape.
1189 if (Body[Pos] != '$' || Pos + 1 == End)
1190 continue;
1192 char Next = Body[Pos + 1];
1193 if (Next == '$' || Next == 'n' || isdigit(Next))
1194 break;
1197 // Add the prefix.
1198 OS << Body.slice(0, Pos);
1200 // Check if we reached the end.
1201 if (Pos == End)
1202 break;
1204 switch (Body[Pos+1]) {
1205 // $$ => $
1206 case '$':
1207 OS << '$';
1208 break;
1210 // $n => number of arguments
1211 case 'n':
1212 OS << A.size();
1213 break;
1215 // $[0-9] => argument
1216 default: {
1217 // Missing arguments are ignored.
1218 unsigned Index = Body[Pos+1] - '0';
1219 if (Index >= A.size())
1220 break;
1222 // Otherwise substitute with the token values, with spaces eliminated.
1223 for (std::vector<AsmToken>::const_iterator it = A[Index].begin(),
1224 ie = A[Index].end(); it != ie; ++it)
1225 OS << it->getString();
1226 break;
1230 // Update the scan point.
1231 Body = Body.substr(Pos + 2);
1234 // We include the .endmacro in the buffer as our queue to exit the macro
1235 // instantiation.
1236 OS << ".endmacro\n";
1238 Instantiation = MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
1241 bool AsmParser::HandleMacroEntry(StringRef Name, SMLoc NameLoc,
1242 const Macro *M) {
1243 // Arbitrarily limit macro nesting depth, to match 'as'. We can eliminate
1244 // this, although we should protect against infinite loops.
1245 if (ActiveMacros.size() == 20)
1246 return TokError("macros cannot be nested more than 20 levels deep");
1248 // Parse the macro instantiation arguments.
1249 std::vector<std::vector<AsmToken> > MacroArguments;
1250 MacroArguments.push_back(std::vector<AsmToken>());
1251 unsigned ParenLevel = 0;
1252 for (;;) {
1253 if (Lexer.is(AsmToken::Eof))
1254 return TokError("unexpected token in macro instantiation");
1255 if (Lexer.is(AsmToken::EndOfStatement))
1256 break;
1258 // If we aren't inside parentheses and this is a comma, start a new token
1259 // list.
1260 if (ParenLevel == 0 && Lexer.is(AsmToken::Comma)) {
1261 MacroArguments.push_back(std::vector<AsmToken>());
1262 } else {
1263 // Adjust the current parentheses level.
1264 if (Lexer.is(AsmToken::LParen))
1265 ++ParenLevel;
1266 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
1267 --ParenLevel;
1269 // Append the token to the current argument list.
1270 MacroArguments.back().push_back(getTok());
1272 Lex();
1275 // Create the macro instantiation object and add to the current macro
1276 // instantiation stack.
1277 MacroInstantiation *MI = new MacroInstantiation(M, NameLoc,
1278 getTok().getLoc(),
1279 MacroArguments);
1280 ActiveMacros.push_back(MI);
1282 // Jump to the macro instantiation and prime the lexer.
1283 CurBuffer = SrcMgr.AddNewSourceBuffer(MI->Instantiation, SMLoc());
1284 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer));
1285 Lex();
1287 return false;
1290 void AsmParser::HandleMacroExit() {
1291 // Jump to the EndOfStatement we should return to, and consume it.
1292 JumpToLoc(ActiveMacros.back()->ExitLoc);
1293 Lex();
1295 // Pop the instantiation entry.
1296 delete ActiveMacros.back();
1297 ActiveMacros.pop_back();
1300 static void MarkUsed(const MCExpr *Value) {
1301 switch (Value->getKind()) {
1302 case MCExpr::Binary:
1303 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getLHS());
1304 MarkUsed(static_cast<const MCBinaryExpr*>(Value)->getRHS());
1305 break;
1306 case MCExpr::Target:
1307 case MCExpr::Constant:
1308 break;
1309 case MCExpr::SymbolRef: {
1310 static_cast<const MCSymbolRefExpr*>(Value)->getSymbol().setUsed(true);
1311 break;
1313 case MCExpr::Unary:
1314 MarkUsed(static_cast<const MCUnaryExpr*>(Value)->getSubExpr());
1315 break;
1319 bool AsmParser::ParseAssignment(StringRef Name, bool allow_redef) {
1320 // FIXME: Use better location, we should use proper tokens.
1321 SMLoc EqualLoc = Lexer.getLoc();
1323 const MCExpr *Value;
1324 if (ParseExpression(Value))
1325 return true;
1327 MarkUsed(Value);
1329 if (Lexer.isNot(AsmToken::EndOfStatement))
1330 return TokError("unexpected token in assignment");
1332 // Error on assignment to '.'.
1333 if (Name == ".") {
1334 return Error(EqualLoc, ("assignment to pseudo-symbol '.' is unsupported "
1335 "(use '.space' or '.org').)"));
1338 // Eat the end of statement marker.
1339 Lex();
1341 // Validate that the LHS is allowed to be a variable (either it has not been
1342 // used as a symbol, or it is an absolute symbol).
1343 MCSymbol *Sym = getContext().LookupSymbol(Name);
1344 if (Sym) {
1345 // Diagnose assignment to a label.
1347 // FIXME: Diagnostics. Note the location of the definition as a label.
1348 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
1349 if (Sym->isUndefined() && !Sym->isUsed() && !Sym->isVariable())
1350 ; // Allow redefinitions of undefined symbols only used in directives.
1351 else if (!Sym->isUndefined() && (!Sym->isAbsolute() || !allow_redef))
1352 return Error(EqualLoc, "redefinition of '" + Name + "'");
1353 else if (!Sym->isVariable())
1354 return Error(EqualLoc, "invalid assignment to '" + Name + "'");
1355 else if (!isa<MCConstantExpr>(Sym->getVariableValue()))
1356 return Error(EqualLoc, "invalid reassignment of non-absolute variable '" +
1357 Name + "'");
1359 // Don't count these checks as uses.
1360 Sym->setUsed(false);
1361 } else
1362 Sym = getContext().GetOrCreateSymbol(Name);
1364 // FIXME: Handle '.'.
1366 // Do the assignment.
1367 Out.EmitAssignment(Sym, Value);
1369 return false;
1372 /// ParseIdentifier:
1373 /// ::= identifier
1374 /// ::= string
1375 bool AsmParser::ParseIdentifier(StringRef &Res) {
1376 // The assembler has relaxed rules for accepting identifiers, in particular we
1377 // allow things like '.globl $foo', which would normally be separate
1378 // tokens. At this level, we have already lexed so we cannot (currently)
1379 // handle this as a context dependent token, instead we detect adjacent tokens
1380 // and return the combined identifier.
1381 if (Lexer.is(AsmToken::Dollar)) {
1382 SMLoc DollarLoc = getLexer().getLoc();
1384 // Consume the dollar sign, and check for a following identifier.
1385 Lex();
1386 if (Lexer.isNot(AsmToken::Identifier))
1387 return true;
1389 // We have a '$' followed by an identifier, make sure they are adjacent.
1390 if (DollarLoc.getPointer() + 1 != getTok().getLoc().getPointer())
1391 return true;
1393 // Construct the joined identifier and consume the token.
1394 Res = StringRef(DollarLoc.getPointer(),
1395 getTok().getIdentifier().size() + 1);
1396 Lex();
1397 return false;
1400 if (Lexer.isNot(AsmToken::Identifier) &&
1401 Lexer.isNot(AsmToken::String))
1402 return true;
1404 Res = getTok().getIdentifier();
1406 Lex(); // Consume the identifier token.
1408 return false;
1411 /// ParseDirectiveSet:
1412 /// ::= .equ identifier ',' expression
1413 /// ::= .equiv identifier ',' expression
1414 /// ::= .set identifier ',' expression
1415 bool AsmParser::ParseDirectiveSet(StringRef IDVal, bool allow_redef) {
1416 StringRef Name;
1418 if (ParseIdentifier(Name))
1419 return TokError("expected identifier after '" + Twine(IDVal) + "'");
1421 if (getLexer().isNot(AsmToken::Comma))
1422 return TokError("unexpected token in '" + Twine(IDVal) + "'");
1423 Lex();
1425 return ParseAssignment(Name, allow_redef);
1428 bool AsmParser::ParseEscapedString(std::string &Data) {
1429 assert(getLexer().is(AsmToken::String) && "Unexpected current token!");
1431 Data = "";
1432 StringRef Str = getTok().getStringContents();
1433 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
1434 if (Str[i] != '\\') {
1435 Data += Str[i];
1436 continue;
1439 // Recognize escaped characters. Note that this escape semantics currently
1440 // loosely follows Darwin 'as'. Notably, it doesn't support hex escapes.
1441 ++i;
1442 if (i == e)
1443 return TokError("unexpected backslash at end of string");
1445 // Recognize octal sequences.
1446 if ((unsigned) (Str[i] - '0') <= 7) {
1447 // Consume up to three octal characters.
1448 unsigned Value = Str[i] - '0';
1450 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1451 ++i;
1452 Value = Value * 8 + (Str[i] - '0');
1454 if (i + 1 != e && ((unsigned) (Str[i + 1] - '0')) <= 7) {
1455 ++i;
1456 Value = Value * 8 + (Str[i] - '0');
1460 if (Value > 255)
1461 return TokError("invalid octal escape sequence (out of range)");
1463 Data += (unsigned char) Value;
1464 continue;
1467 // Otherwise recognize individual escapes.
1468 switch (Str[i]) {
1469 default:
1470 // Just reject invalid escape sequences for now.
1471 return TokError("invalid escape sequence (unrecognized character)");
1473 case 'b': Data += '\b'; break;
1474 case 'f': Data += '\f'; break;
1475 case 'n': Data += '\n'; break;
1476 case 'r': Data += '\r'; break;
1477 case 't': Data += '\t'; break;
1478 case '"': Data += '"'; break;
1479 case '\\': Data += '\\'; break;
1483 return false;
1486 /// ParseDirectiveAscii:
1487 /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
1488 bool AsmParser::ParseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
1489 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1490 CheckForValidSection();
1492 for (;;) {
1493 if (getLexer().isNot(AsmToken::String))
1494 return TokError("expected string in '" + Twine(IDVal) + "' directive");
1496 std::string Data;
1497 if (ParseEscapedString(Data))
1498 return true;
1500 getStreamer().EmitBytes(Data, DEFAULT_ADDRSPACE);
1501 if (ZeroTerminated)
1502 getStreamer().EmitBytes(StringRef("\0", 1), DEFAULT_ADDRSPACE);
1504 Lex();
1506 if (getLexer().is(AsmToken::EndOfStatement))
1507 break;
1509 if (getLexer().isNot(AsmToken::Comma))
1510 return TokError("unexpected token in '" + Twine(IDVal) + "' directive");
1511 Lex();
1515 Lex();
1516 return false;
1519 /// ParseDirectiveValue
1520 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
1521 bool AsmParser::ParseDirectiveValue(unsigned Size) {
1522 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1523 CheckForValidSection();
1525 for (;;) {
1526 const MCExpr *Value;
1527 if (ParseExpression(Value))
1528 return true;
1530 // Special case constant expressions to match code generator.
1531 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value))
1532 getStreamer().EmitIntValue(MCE->getValue(), Size, DEFAULT_ADDRSPACE);
1533 else
1534 getStreamer().EmitValue(Value, Size, DEFAULT_ADDRSPACE);
1536 if (getLexer().is(AsmToken::EndOfStatement))
1537 break;
1539 // FIXME: Improve diagnostic.
1540 if (getLexer().isNot(AsmToken::Comma))
1541 return TokError("unexpected token in directive");
1542 Lex();
1546 Lex();
1547 return false;
1550 /// ParseDirectiveRealValue
1551 /// ::= (.single | .double) [ expression (, expression)* ]
1552 bool AsmParser::ParseDirectiveRealValue(const fltSemantics &Semantics) {
1553 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1554 CheckForValidSection();
1556 for (;;) {
1557 // We don't truly support arithmetic on floating point expressions, so we
1558 // have to manually parse unary prefixes.
1559 bool IsNeg = false;
1560 if (getLexer().is(AsmToken::Minus)) {
1561 Lex();
1562 IsNeg = true;
1563 } else if (getLexer().is(AsmToken::Plus))
1564 Lex();
1566 if (getLexer().isNot(AsmToken::Integer) &&
1567 getLexer().isNot(AsmToken::Real) &&
1568 getLexer().isNot(AsmToken::Identifier))
1569 return TokError("unexpected token in directive");
1571 // Convert to an APFloat.
1572 APFloat Value(Semantics);
1573 StringRef IDVal = getTok().getString();
1574 if (getLexer().is(AsmToken::Identifier)) {
1575 if (!IDVal.compare_lower("infinity") || !IDVal.compare_lower("inf"))
1576 Value = APFloat::getInf(Semantics);
1577 else if (!IDVal.compare_lower("nan"))
1578 Value = APFloat::getNaN(Semantics, false, ~0);
1579 else
1580 return TokError("invalid floating point literal");
1581 } else if (Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) ==
1582 APFloat::opInvalidOp)
1583 return TokError("invalid floating point literal");
1584 if (IsNeg)
1585 Value.changeSign();
1587 // Consume the numeric token.
1588 Lex();
1590 // Emit the value as an integer.
1591 APInt AsInt = Value.bitcastToAPInt();
1592 getStreamer().EmitIntValue(AsInt.getLimitedValue(),
1593 AsInt.getBitWidth() / 8, DEFAULT_ADDRSPACE);
1595 if (getLexer().is(AsmToken::EndOfStatement))
1596 break;
1598 if (getLexer().isNot(AsmToken::Comma))
1599 return TokError("unexpected token in directive");
1600 Lex();
1604 Lex();
1605 return false;
1608 /// ParseDirectiveSpace
1609 /// ::= .space expression [ , expression ]
1610 bool AsmParser::ParseDirectiveSpace() {
1611 CheckForValidSection();
1613 int64_t NumBytes;
1614 if (ParseAbsoluteExpression(NumBytes))
1615 return true;
1617 int64_t FillExpr = 0;
1618 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1619 if (getLexer().isNot(AsmToken::Comma))
1620 return TokError("unexpected token in '.space' directive");
1621 Lex();
1623 if (ParseAbsoluteExpression(FillExpr))
1624 return true;
1626 if (getLexer().isNot(AsmToken::EndOfStatement))
1627 return TokError("unexpected token in '.space' directive");
1630 Lex();
1632 if (NumBytes <= 0)
1633 return TokError("invalid number of bytes in '.space' directive");
1635 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
1636 getStreamer().EmitFill(NumBytes, FillExpr, DEFAULT_ADDRSPACE);
1638 return false;
1641 /// ParseDirectiveZero
1642 /// ::= .zero expression
1643 bool AsmParser::ParseDirectiveZero() {
1644 CheckForValidSection();
1646 int64_t NumBytes;
1647 if (ParseAbsoluteExpression(NumBytes))
1648 return true;
1650 int64_t Val = 0;
1651 if (getLexer().is(AsmToken::Comma)) {
1652 Lex();
1653 if (ParseAbsoluteExpression(Val))
1654 return true;
1657 if (getLexer().isNot(AsmToken::EndOfStatement))
1658 return TokError("unexpected token in '.zero' directive");
1660 Lex();
1662 getStreamer().EmitFill(NumBytes, Val, DEFAULT_ADDRSPACE);
1664 return false;
1667 /// ParseDirectiveFill
1668 /// ::= .fill expression , expression , expression
1669 bool AsmParser::ParseDirectiveFill() {
1670 CheckForValidSection();
1672 int64_t NumValues;
1673 if (ParseAbsoluteExpression(NumValues))
1674 return true;
1676 if (getLexer().isNot(AsmToken::Comma))
1677 return TokError("unexpected token in '.fill' directive");
1678 Lex();
1680 int64_t FillSize;
1681 if (ParseAbsoluteExpression(FillSize))
1682 return true;
1684 if (getLexer().isNot(AsmToken::Comma))
1685 return TokError("unexpected token in '.fill' directive");
1686 Lex();
1688 int64_t FillExpr;
1689 if (ParseAbsoluteExpression(FillExpr))
1690 return true;
1692 if (getLexer().isNot(AsmToken::EndOfStatement))
1693 return TokError("unexpected token in '.fill' directive");
1695 Lex();
1697 if (FillSize != 1 && FillSize != 2 && FillSize != 4 && FillSize != 8)
1698 return TokError("invalid '.fill' size, expected 1, 2, 4, or 8");
1700 for (uint64_t i = 0, e = NumValues; i != e; ++i)
1701 getStreamer().EmitIntValue(FillExpr, FillSize, DEFAULT_ADDRSPACE);
1703 return false;
1706 /// ParseDirectiveOrg
1707 /// ::= .org expression [ , expression ]
1708 bool AsmParser::ParseDirectiveOrg() {
1709 CheckForValidSection();
1711 const MCExpr *Offset;
1712 if (ParseExpression(Offset))
1713 return true;
1715 // Parse optional fill expression.
1716 int64_t FillExpr = 0;
1717 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1718 if (getLexer().isNot(AsmToken::Comma))
1719 return TokError("unexpected token in '.org' directive");
1720 Lex();
1722 if (ParseAbsoluteExpression(FillExpr))
1723 return true;
1725 if (getLexer().isNot(AsmToken::EndOfStatement))
1726 return TokError("unexpected token in '.org' directive");
1729 Lex();
1731 // FIXME: Only limited forms of relocatable expressions are accepted here, it
1732 // has to be relative to the current section.
1733 getStreamer().EmitValueToOffset(Offset, FillExpr);
1735 return false;
1738 /// ParseDirectiveAlign
1739 /// ::= {.align, ...} expression [ , expression [ , expression ]]
1740 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
1741 CheckForValidSection();
1743 SMLoc AlignmentLoc = getLexer().getLoc();
1744 int64_t Alignment;
1745 if (ParseAbsoluteExpression(Alignment))
1746 return true;
1748 SMLoc MaxBytesLoc;
1749 bool HasFillExpr = false;
1750 int64_t FillExpr = 0;
1751 int64_t MaxBytesToFill = 0;
1752 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1753 if (getLexer().isNot(AsmToken::Comma))
1754 return TokError("unexpected token in directive");
1755 Lex();
1757 // The fill expression can be omitted while specifying a maximum number of
1758 // alignment bytes, e.g:
1759 // .align 3,,4
1760 if (getLexer().isNot(AsmToken::Comma)) {
1761 HasFillExpr = true;
1762 if (ParseAbsoluteExpression(FillExpr))
1763 return true;
1766 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1767 if (getLexer().isNot(AsmToken::Comma))
1768 return TokError("unexpected token in directive");
1769 Lex();
1771 MaxBytesLoc = getLexer().getLoc();
1772 if (ParseAbsoluteExpression(MaxBytesToFill))
1773 return true;
1775 if (getLexer().isNot(AsmToken::EndOfStatement))
1776 return TokError("unexpected token in directive");
1780 Lex();
1782 if (!HasFillExpr)
1783 FillExpr = 0;
1785 // Compute alignment in bytes.
1786 if (IsPow2) {
1787 // FIXME: Diagnose overflow.
1788 if (Alignment >= 32) {
1789 Error(AlignmentLoc, "invalid alignment value");
1790 Alignment = 31;
1793 Alignment = 1ULL << Alignment;
1796 // Diagnose non-sensical max bytes to align.
1797 if (MaxBytesLoc.isValid()) {
1798 if (MaxBytesToFill < 1) {
1799 Error(MaxBytesLoc, "alignment directive can never be satisfied in this "
1800 "many bytes, ignoring maximum bytes expression");
1801 MaxBytesToFill = 0;
1804 if (MaxBytesToFill >= Alignment) {
1805 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
1806 "has no effect");
1807 MaxBytesToFill = 0;
1811 // Check whether we should use optimal code alignment for this .align
1812 // directive.
1813 bool UseCodeAlign = getStreamer().getCurrentSection()->UseCodeAlign();
1814 if ((!HasFillExpr || Lexer.getMAI().getTextAlignFillValue() == FillExpr) &&
1815 ValueSize == 1 && UseCodeAlign) {
1816 getStreamer().EmitCodeAlignment(Alignment, MaxBytesToFill);
1817 } else {
1818 // FIXME: Target specific behavior about how the "extra" bytes are filled.
1819 getStreamer().EmitValueToAlignment(Alignment, FillExpr, ValueSize,
1820 MaxBytesToFill);
1823 return false;
1826 /// ParseDirectiveSymbolAttribute
1827 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
1828 bool AsmParser::ParseDirectiveSymbolAttribute(MCSymbolAttr Attr) {
1829 if (getLexer().isNot(AsmToken::EndOfStatement)) {
1830 for (;;) {
1831 StringRef Name;
1833 if (ParseIdentifier(Name))
1834 return TokError("expected identifier in directive");
1836 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1838 getStreamer().EmitSymbolAttribute(Sym, Attr);
1840 if (getLexer().is(AsmToken::EndOfStatement))
1841 break;
1843 if (getLexer().isNot(AsmToken::Comma))
1844 return TokError("unexpected token in directive");
1845 Lex();
1849 Lex();
1850 return false;
1853 /// ParseDirectiveComm
1854 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1855 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1856 CheckForValidSection();
1858 SMLoc IDLoc = getLexer().getLoc();
1859 StringRef Name;
1860 if (ParseIdentifier(Name))
1861 return TokError("expected identifier in directive");
1863 // Handle the identifier as the key symbol.
1864 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
1866 if (getLexer().isNot(AsmToken::Comma))
1867 return TokError("unexpected token in directive");
1868 Lex();
1870 int64_t Size;
1871 SMLoc SizeLoc = getLexer().getLoc();
1872 if (ParseAbsoluteExpression(Size))
1873 return true;
1875 int64_t Pow2Alignment = 0;
1876 SMLoc Pow2AlignmentLoc;
1877 if (getLexer().is(AsmToken::Comma)) {
1878 Lex();
1879 Pow2AlignmentLoc = getLexer().getLoc();
1880 if (ParseAbsoluteExpression(Pow2Alignment))
1881 return true;
1883 // If this target takes alignments in bytes (not log) validate and convert.
1884 if (Lexer.getMAI().getAlignmentIsInBytes()) {
1885 if (!isPowerOf2_64(Pow2Alignment))
1886 return Error(Pow2AlignmentLoc, "alignment must be a power of 2");
1887 Pow2Alignment = Log2_64(Pow2Alignment);
1891 if (getLexer().isNot(AsmToken::EndOfStatement))
1892 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1894 Lex();
1896 // NOTE: a size of zero for a .comm should create a undefined symbol
1897 // but a size of .lcomm creates a bss symbol of size zero.
1898 if (Size < 0)
1899 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1900 "be less than zero");
1902 // NOTE: The alignment in the directive is a power of 2 value, the assembler
1903 // may internally end up wanting an alignment in bytes.
1904 // FIXME: Diagnose overflow.
1905 if (Pow2Alignment < 0)
1906 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1907 "alignment, can't be less than zero");
1909 if (!Sym->isUndefined())
1910 return Error(IDLoc, "invalid symbol redefinition");
1912 // '.lcomm' is equivalent to '.zerofill'.
1913 // Create the Symbol as a common or local common with Size and Pow2Alignment
1914 if (IsLocal) {
1915 getStreamer().EmitZerofill(Ctx.getMachOSection(
1916 "__DATA", "__bss", MCSectionMachO::S_ZEROFILL,
1917 0, SectionKind::getBSS()),
1918 Sym, Size, 1 << Pow2Alignment);
1919 return false;
1922 getStreamer().EmitCommonSymbol(Sym, Size, 1 << Pow2Alignment);
1923 return false;
1926 /// ParseDirectiveAbort
1927 /// ::= .abort [... message ...]
1928 bool AsmParser::ParseDirectiveAbort() {
1929 // FIXME: Use loc from directive.
1930 SMLoc Loc = getLexer().getLoc();
1932 StringRef Str = ParseStringToEndOfStatement();
1933 if (getLexer().isNot(AsmToken::EndOfStatement))
1934 return TokError("unexpected token in '.abort' directive");
1936 Lex();
1938 if (Str.empty())
1939 Error(Loc, ".abort detected. Assembly stopping.");
1940 else
1941 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1942 // FIXME: Actually abort assembly here.
1944 return false;
1947 /// ParseDirectiveInclude
1948 /// ::= .include "filename"
1949 bool AsmParser::ParseDirectiveInclude() {
1950 if (getLexer().isNot(AsmToken::String))
1951 return TokError("expected string in '.include' directive");
1953 std::string Filename = getTok().getString();
1954 SMLoc IncludeLoc = getLexer().getLoc();
1955 Lex();
1957 if (getLexer().isNot(AsmToken::EndOfStatement))
1958 return TokError("unexpected token in '.include' directive");
1960 // Strip the quotes.
1961 Filename = Filename.substr(1, Filename.size()-2);
1963 // Attempt to switch the lexer to the included file before consuming the end
1964 // of statement to avoid losing it when we switch.
1965 if (EnterIncludeFile(Filename)) {
1966 Error(IncludeLoc, "Could not find include file '" + Filename + "'");
1967 return true;
1970 return false;
1973 /// ParseDirectiveIf
1974 /// ::= .if expression
1975 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1976 TheCondStack.push_back(TheCondState);
1977 TheCondState.TheCond = AsmCond::IfCond;
1978 if(TheCondState.Ignore) {
1979 EatToEndOfStatement();
1981 else {
1982 int64_t ExprValue;
1983 if (ParseAbsoluteExpression(ExprValue))
1984 return true;
1986 if (getLexer().isNot(AsmToken::EndOfStatement))
1987 return TokError("unexpected token in '.if' directive");
1989 Lex();
1991 TheCondState.CondMet = ExprValue;
1992 TheCondState.Ignore = !TheCondState.CondMet;
1995 return false;
1998 bool AsmParser::ParseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) {
1999 StringRef Name;
2000 TheCondStack.push_back(TheCondState);
2001 TheCondState.TheCond = AsmCond::IfCond;
2003 if (TheCondState.Ignore) {
2004 EatToEndOfStatement();
2005 } else {
2006 if (ParseIdentifier(Name))
2007 return TokError("expected identifier after '.ifdef'");
2009 Lex();
2011 MCSymbol *Sym = getContext().LookupSymbol(Name);
2013 if (expect_defined)
2014 TheCondState.CondMet = (Sym != NULL && !Sym->isUndefined());
2015 else
2016 TheCondState.CondMet = (Sym == NULL || Sym->isUndefined());
2017 TheCondState.Ignore = !TheCondState.CondMet;
2020 return false;
2023 /// ParseDirectiveElseIf
2024 /// ::= .elseif expression
2025 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
2026 if (TheCondState.TheCond != AsmCond::IfCond &&
2027 TheCondState.TheCond != AsmCond::ElseIfCond)
2028 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
2029 " an .elseif");
2030 TheCondState.TheCond = AsmCond::ElseIfCond;
2032 bool LastIgnoreState = false;
2033 if (!TheCondStack.empty())
2034 LastIgnoreState = TheCondStack.back().Ignore;
2035 if (LastIgnoreState || TheCondState.CondMet) {
2036 TheCondState.Ignore = true;
2037 EatToEndOfStatement();
2039 else {
2040 int64_t ExprValue;
2041 if (ParseAbsoluteExpression(ExprValue))
2042 return true;
2044 if (getLexer().isNot(AsmToken::EndOfStatement))
2045 return TokError("unexpected token in '.elseif' directive");
2047 Lex();
2048 TheCondState.CondMet = ExprValue;
2049 TheCondState.Ignore = !TheCondState.CondMet;
2052 return false;
2055 /// ParseDirectiveElse
2056 /// ::= .else
2057 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
2058 if (getLexer().isNot(AsmToken::EndOfStatement))
2059 return TokError("unexpected token in '.else' directive");
2061 Lex();
2063 if (TheCondState.TheCond != AsmCond::IfCond &&
2064 TheCondState.TheCond != AsmCond::ElseIfCond)
2065 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
2066 ".elseif");
2067 TheCondState.TheCond = AsmCond::ElseCond;
2068 bool LastIgnoreState = false;
2069 if (!TheCondStack.empty())
2070 LastIgnoreState = TheCondStack.back().Ignore;
2071 if (LastIgnoreState || TheCondState.CondMet)
2072 TheCondState.Ignore = true;
2073 else
2074 TheCondState.Ignore = false;
2076 return false;
2079 /// ParseDirectiveEndIf
2080 /// ::= .endif
2081 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
2082 if (getLexer().isNot(AsmToken::EndOfStatement))
2083 return TokError("unexpected token in '.endif' directive");
2085 Lex();
2087 if ((TheCondState.TheCond == AsmCond::NoCond) ||
2088 TheCondStack.empty())
2089 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
2090 ".else");
2091 if (!TheCondStack.empty()) {
2092 TheCondState = TheCondStack.back();
2093 TheCondStack.pop_back();
2096 return false;
2099 /// ParseDirectiveFile
2100 /// ::= .file [number] string
2101 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
2102 // FIXME: I'm not sure what this is.
2103 int64_t FileNumber = -1;
2104 SMLoc FileNumberLoc = getLexer().getLoc();
2105 if (getLexer().is(AsmToken::Integer)) {
2106 FileNumber = getTok().getIntVal();
2107 Lex();
2109 if (FileNumber < 1)
2110 return TokError("file number less than one");
2113 if (getLexer().isNot(AsmToken::String))
2114 return TokError("unexpected token in '.file' directive");
2116 StringRef Filename = getTok().getString();
2117 Filename = Filename.substr(1, Filename.size()-2);
2118 Lex();
2120 if (getLexer().isNot(AsmToken::EndOfStatement))
2121 return TokError("unexpected token in '.file' directive");
2123 if (FileNumber == -1)
2124 getStreamer().EmitFileDirective(Filename);
2125 else {
2126 if (getStreamer().EmitDwarfFileDirective(FileNumber, Filename))
2127 Error(FileNumberLoc, "file number already allocated");
2130 return false;
2133 /// ParseDirectiveLine
2134 /// ::= .line [number]
2135 bool GenericAsmParser::ParseDirectiveLine(StringRef, SMLoc DirectiveLoc) {
2136 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2137 if (getLexer().isNot(AsmToken::Integer))
2138 return TokError("unexpected token in '.line' directive");
2140 int64_t LineNumber = getTok().getIntVal();
2141 (void) LineNumber;
2142 Lex();
2144 // FIXME: Do something with the .line.
2147 if (getLexer().isNot(AsmToken::EndOfStatement))
2148 return TokError("unexpected token in '.line' directive");
2150 return false;
2154 /// ParseDirectiveLoc
2155 /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
2156 /// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
2157 /// The first number is a file number, must have been previously assigned with
2158 /// a .file directive, the second number is the line number and optionally the
2159 /// third number is a column position (zero if not specified). The remaining
2160 /// optional items are .loc sub-directives.
2161 bool GenericAsmParser::ParseDirectiveLoc(StringRef, SMLoc DirectiveLoc) {
2163 if (getLexer().isNot(AsmToken::Integer))
2164 return TokError("unexpected token in '.loc' directive");
2165 int64_t FileNumber = getTok().getIntVal();
2166 if (FileNumber < 1)
2167 return TokError("file number less than one in '.loc' directive");
2168 if (!getContext().isValidDwarfFileNumber(FileNumber))
2169 return TokError("unassigned file number in '.loc' directive");
2170 Lex();
2172 int64_t LineNumber = 0;
2173 if (getLexer().is(AsmToken::Integer)) {
2174 LineNumber = getTok().getIntVal();
2175 if (LineNumber < 1)
2176 return TokError("line number less than one in '.loc' directive");
2177 Lex();
2180 int64_t ColumnPos = 0;
2181 if (getLexer().is(AsmToken::Integer)) {
2182 ColumnPos = getTok().getIntVal();
2183 if (ColumnPos < 0)
2184 return TokError("column position less than zero in '.loc' directive");
2185 Lex();
2188 unsigned Flags = DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0;
2189 unsigned Isa = 0;
2190 int64_t Discriminator = 0;
2191 if (getLexer().isNot(AsmToken::EndOfStatement)) {
2192 for (;;) {
2193 if (getLexer().is(AsmToken::EndOfStatement))
2194 break;
2196 StringRef Name;
2197 SMLoc Loc = getTok().getLoc();
2198 if (getParser().ParseIdentifier(Name))
2199 return TokError("unexpected token in '.loc' directive");
2201 if (Name == "basic_block")
2202 Flags |= DWARF2_FLAG_BASIC_BLOCK;
2203 else if (Name == "prologue_end")
2204 Flags |= DWARF2_FLAG_PROLOGUE_END;
2205 else if (Name == "epilogue_begin")
2206 Flags |= DWARF2_FLAG_EPILOGUE_BEGIN;
2207 else if (Name == "is_stmt") {
2208 SMLoc Loc = getTok().getLoc();
2209 const MCExpr *Value;
2210 if (getParser().ParseExpression(Value))
2211 return true;
2212 // The expression must be the constant 0 or 1.
2213 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2214 int Value = MCE->getValue();
2215 if (Value == 0)
2216 Flags &= ~DWARF2_FLAG_IS_STMT;
2217 else if (Value == 1)
2218 Flags |= DWARF2_FLAG_IS_STMT;
2219 else
2220 return Error(Loc, "is_stmt value not 0 or 1");
2222 else {
2223 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
2226 else if (Name == "isa") {
2227 SMLoc Loc = getTok().getLoc();
2228 const MCExpr *Value;
2229 if (getParser().ParseExpression(Value))
2230 return true;
2231 // The expression must be a constant greater or equal to 0.
2232 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
2233 int Value = MCE->getValue();
2234 if (Value < 0)
2235 return Error(Loc, "isa number less than zero");
2236 Isa = Value;
2238 else {
2239 return Error(Loc, "isa number not a constant value");
2242 else if (Name == "discriminator") {
2243 if (getParser().ParseAbsoluteExpression(Discriminator))
2244 return true;
2246 else {
2247 return Error(Loc, "unknown sub-directive in '.loc' directive");
2250 if (getLexer().is(AsmToken::EndOfStatement))
2251 break;
2255 getStreamer().EmitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
2256 Isa, Discriminator);
2258 return false;
2261 /// ParseDirectiveStabs
2262 /// ::= .stabs string, number, number, number
2263 bool GenericAsmParser::ParseDirectiveStabs(StringRef Directive,
2264 SMLoc DirectiveLoc) {
2265 return TokError("unsupported directive '" + Directive + "'");
2268 /// ParseDirectiveCFIStartProc
2269 /// ::= .cfi_startproc
2270 bool GenericAsmParser::ParseDirectiveCFIStartProc(StringRef,
2271 SMLoc DirectiveLoc) {
2272 getStreamer().EmitCFIStartProc();
2273 return false;
2276 /// ParseDirectiveCFIEndProc
2277 /// ::= .cfi_endproc
2278 bool GenericAsmParser::ParseDirectiveCFIEndProc(StringRef, SMLoc DirectiveLoc) {
2279 getStreamer().EmitCFIEndProc();
2280 return false;
2283 /// ParseRegisterOrRegisterNumber - parse register name or number.
2284 bool GenericAsmParser::ParseRegisterOrRegisterNumber(int64_t &Register,
2285 SMLoc DirectiveLoc) {
2286 unsigned RegNo;
2288 if (getLexer().is(AsmToken::Percent)) {
2289 if (getParser().getTargetParser().ParseRegister(RegNo, DirectiveLoc,
2290 DirectiveLoc))
2291 return true;
2292 Register = getContext().getTargetAsmInfo().getDwarfRegNum(RegNo, true);
2293 } else
2294 return getParser().ParseAbsoluteExpression(Register);
2296 return false;
2299 /// ParseDirectiveCFIDefCfa
2300 /// ::= .cfi_def_cfa register, offset
2301 bool GenericAsmParser::ParseDirectiveCFIDefCfa(StringRef,
2302 SMLoc DirectiveLoc) {
2303 int64_t Register = 0;
2304 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2305 return true;
2307 if (getLexer().isNot(AsmToken::Comma))
2308 return TokError("unexpected token in directive");
2309 Lex();
2311 int64_t Offset = 0;
2312 if (getParser().ParseAbsoluteExpression(Offset))
2313 return true;
2315 getStreamer().EmitCFIDefCfa(Register, Offset);
2316 return false;
2319 /// ParseDirectiveCFIDefCfaOffset
2320 /// ::= .cfi_def_cfa_offset offset
2321 bool GenericAsmParser::ParseDirectiveCFIDefCfaOffset(StringRef,
2322 SMLoc DirectiveLoc) {
2323 int64_t Offset = 0;
2324 if (getParser().ParseAbsoluteExpression(Offset))
2325 return true;
2327 getStreamer().EmitCFIDefCfaOffset(Offset);
2328 return false;
2331 /// ParseDirectiveCFIAdjustCfaOffset
2332 /// ::= .cfi_adjust_cfa_offset adjustment
2333 bool GenericAsmParser::ParseDirectiveCFIAdjustCfaOffset(StringRef,
2334 SMLoc DirectiveLoc) {
2335 int64_t Adjustment = 0;
2336 if (getParser().ParseAbsoluteExpression(Adjustment))
2337 return true;
2339 getStreamer().EmitCFIAdjustCfaOffset(Adjustment);
2340 return false;
2343 /// ParseDirectiveCFIDefCfaRegister
2344 /// ::= .cfi_def_cfa_register register
2345 bool GenericAsmParser::ParseDirectiveCFIDefCfaRegister(StringRef,
2346 SMLoc DirectiveLoc) {
2347 int64_t Register = 0;
2348 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2349 return true;
2351 getStreamer().EmitCFIDefCfaRegister(Register);
2352 return false;
2355 /// ParseDirectiveCFIOffset
2356 /// ::= .cfi_offset register, offset
2357 bool GenericAsmParser::ParseDirectiveCFIOffset(StringRef, SMLoc DirectiveLoc) {
2358 int64_t Register = 0;
2359 int64_t Offset = 0;
2361 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2362 return true;
2364 if (getLexer().isNot(AsmToken::Comma))
2365 return TokError("unexpected token in directive");
2366 Lex();
2368 if (getParser().ParseAbsoluteExpression(Offset))
2369 return true;
2371 getStreamer().EmitCFIOffset(Register, Offset);
2372 return false;
2375 /// ParseDirectiveCFIRelOffset
2376 /// ::= .cfi_rel_offset register, offset
2377 bool GenericAsmParser::ParseDirectiveCFIRelOffset(StringRef,
2378 SMLoc DirectiveLoc) {
2379 int64_t Register = 0;
2381 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2382 return true;
2384 if (getLexer().isNot(AsmToken::Comma))
2385 return TokError("unexpected token in directive");
2386 Lex();
2388 int64_t Offset = 0;
2389 if (getParser().ParseAbsoluteExpression(Offset))
2390 return true;
2392 getStreamer().EmitCFIRelOffset(Register, Offset);
2393 return false;
2396 static bool isValidEncoding(int64_t Encoding) {
2397 if (Encoding & ~0xff)
2398 return false;
2400 if (Encoding == dwarf::DW_EH_PE_omit)
2401 return true;
2403 const unsigned Format = Encoding & 0xf;
2404 if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 &&
2405 Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 &&
2406 Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 &&
2407 Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed)
2408 return false;
2410 const unsigned Application = Encoding & 0x70;
2411 if (Application != dwarf::DW_EH_PE_absptr &&
2412 Application != dwarf::DW_EH_PE_pcrel)
2413 return false;
2415 return true;
2418 /// ParseDirectiveCFIPersonalityOrLsda
2419 /// ::= .cfi_personality encoding, [symbol_name]
2420 /// ::= .cfi_lsda encoding, [symbol_name]
2421 bool GenericAsmParser::ParseDirectiveCFIPersonalityOrLsda(StringRef IDVal,
2422 SMLoc DirectiveLoc) {
2423 int64_t Encoding = 0;
2424 if (getParser().ParseAbsoluteExpression(Encoding))
2425 return true;
2426 if (Encoding == dwarf::DW_EH_PE_omit)
2427 return false;
2429 if (!isValidEncoding(Encoding))
2430 return TokError("unsupported encoding.");
2432 if (getLexer().isNot(AsmToken::Comma))
2433 return TokError("unexpected token in directive");
2434 Lex();
2436 StringRef Name;
2437 if (getParser().ParseIdentifier(Name))
2438 return TokError("expected identifier in directive");
2440 MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
2442 if (IDVal == ".cfi_personality")
2443 getStreamer().EmitCFIPersonality(Sym, Encoding);
2444 else {
2445 assert(IDVal == ".cfi_lsda");
2446 getStreamer().EmitCFILsda(Sym, Encoding);
2448 return false;
2451 /// ParseDirectiveCFIRememberState
2452 /// ::= .cfi_remember_state
2453 bool GenericAsmParser::ParseDirectiveCFIRememberState(StringRef IDVal,
2454 SMLoc DirectiveLoc) {
2455 getStreamer().EmitCFIRememberState();
2456 return false;
2459 /// ParseDirectiveCFIRestoreState
2460 /// ::= .cfi_remember_state
2461 bool GenericAsmParser::ParseDirectiveCFIRestoreState(StringRef IDVal,
2462 SMLoc DirectiveLoc) {
2463 getStreamer().EmitCFIRestoreState();
2464 return false;
2467 /// ParseDirectiveCFISameValue
2468 /// ::= .cfi_same_value register
2469 bool GenericAsmParser::ParseDirectiveCFISameValue(StringRef IDVal,
2470 SMLoc DirectiveLoc) {
2471 int64_t Register = 0;
2473 if (ParseRegisterOrRegisterNumber(Register, DirectiveLoc))
2474 return true;
2476 getStreamer().EmitCFISameValue(Register);
2478 return false;
2481 /// ParseDirectiveMacrosOnOff
2482 /// ::= .macros_on
2483 /// ::= .macros_off
2484 bool GenericAsmParser::ParseDirectiveMacrosOnOff(StringRef Directive,
2485 SMLoc DirectiveLoc) {
2486 if (getLexer().isNot(AsmToken::EndOfStatement))
2487 return Error(getLexer().getLoc(),
2488 "unexpected token in '" + Directive + "' directive");
2490 getParser().MacrosEnabled = Directive == ".macros_on";
2492 return false;
2495 /// ParseDirectiveMacro
2496 /// ::= .macro name
2497 bool GenericAsmParser::ParseDirectiveMacro(StringRef Directive,
2498 SMLoc DirectiveLoc) {
2499 StringRef Name;
2500 if (getParser().ParseIdentifier(Name))
2501 return TokError("expected identifier in directive");
2503 if (getLexer().isNot(AsmToken::EndOfStatement))
2504 return TokError("unexpected token in '.macro' directive");
2506 // Eat the end of statement.
2507 Lex();
2509 AsmToken EndToken, StartToken = getTok();
2511 // Lex the macro definition.
2512 for (;;) {
2513 // Check whether we have reached the end of the file.
2514 if (getLexer().is(AsmToken::Eof))
2515 return Error(DirectiveLoc, "no matching '.endmacro' in definition");
2517 // Otherwise, check whether we have reach the .endmacro.
2518 if (getLexer().is(AsmToken::Identifier) &&
2519 (getTok().getIdentifier() == ".endm" ||
2520 getTok().getIdentifier() == ".endmacro")) {
2521 EndToken = getTok();
2522 Lex();
2523 if (getLexer().isNot(AsmToken::EndOfStatement))
2524 return TokError("unexpected token in '" + EndToken.getIdentifier() +
2525 "' directive");
2526 break;
2529 // Otherwise, scan til the end of the statement.
2530 getParser().EatToEndOfStatement();
2533 if (getParser().MacroMap.lookup(Name)) {
2534 return Error(DirectiveLoc, "macro '" + Name + "' is already defined");
2537 const char *BodyStart = StartToken.getLoc().getPointer();
2538 const char *BodyEnd = EndToken.getLoc().getPointer();
2539 StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart);
2540 getParser().MacroMap[Name] = new Macro(Name, Body);
2541 return false;
2544 /// ParseDirectiveEndMacro
2545 /// ::= .endm
2546 /// ::= .endmacro
2547 bool GenericAsmParser::ParseDirectiveEndMacro(StringRef Directive,
2548 SMLoc DirectiveLoc) {
2549 if (getLexer().isNot(AsmToken::EndOfStatement))
2550 return TokError("unexpected token in '" + Directive + "' directive");
2552 // If we are inside a macro instantiation, terminate the current
2553 // instantiation.
2554 if (!getParser().ActiveMacros.empty()) {
2555 getParser().HandleMacroExit();
2556 return false;
2559 // Otherwise, this .endmacro is a stray entry in the file; well formed
2560 // .endmacro directives are handled during the macro definition parsing.
2561 return TokError("unexpected '" + Directive + "' in file, "
2562 "no current macro definition");
2565 bool GenericAsmParser::ParseDirectiveLEB128(StringRef DirName, SMLoc) {
2566 getParser().CheckForValidSection();
2568 const MCExpr *Value;
2570 if (getParser().ParseExpression(Value))
2571 return true;
2573 if (getLexer().isNot(AsmToken::EndOfStatement))
2574 return TokError("unexpected token in directive");
2576 if (DirName[1] == 's')
2577 getStreamer().EmitSLEB128Value(Value);
2578 else
2579 getStreamer().EmitULEB128Value(Value);
2581 return false;
2585 /// \brief Create an MCAsmParser instance.
2586 MCAsmParser *llvm::createMCAsmParser(const Target &T, SourceMgr &SM,
2587 MCContext &C, MCStreamer &Out,
2588 const MCAsmInfo &MAI) {
2589 return new AsmParser(T, SM, C, Out, MAI);