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 "AsmParser.h"
17 #include "llvm/ADT/Twine.h"
18 #include "llvm/MC/MCContext.h"
19 #include "llvm/MC/MCInst.h"
20 #include "llvm/MC/MCSectionMachO.h"
21 #include "llvm/MC/MCStreamer.h"
22 #include "llvm/MC/MCSymbol.h"
23 #include "llvm/Support/SourceMgr.h"
24 #include "llvm/Support/raw_ostream.h"
25 #include "llvm/Target/TargetAsmParser.h"
28 void AsmParser::Warning(SMLoc L, const Twine &Msg) {
29 Lexer.PrintMessage(L, Msg.str(), "warning");
32 bool AsmParser::Error(SMLoc L, const Twine &Msg) {
33 Lexer.PrintMessage(L, Msg.str(), "error");
37 bool AsmParser::TokError(const char *Msg) {
38 Lexer.PrintMessage(Lexer.getLoc(), Msg, "error");
42 bool AsmParser::Run() {
46 bool HadError = false;
48 AsmCond StartingCondState = TheCondState;
50 // While we have input, parse each statement.
51 while (Lexer.isNot(AsmToken::Eof)) {
52 // Handle conditional assembly here before calling ParseStatement()
53 if (Lexer.getKind() == AsmToken::Identifier) {
54 // If we have an identifier, handle it as the key symbol.
55 AsmToken ID = Lexer.getTok();
56 SMLoc IDLoc = ID.getLoc();
57 StringRef IDVal = ID.getString();
63 if (!ParseConditionalAssemblyDirectives(IDVal, IDLoc))
66 EatToEndOfStatement();
70 if (TheCondState.Ignore) {
71 EatToEndOfStatement();
75 if (!ParseStatement()) continue;
77 // We had an error, remember it and recover by skipping to the next line.
79 EatToEndOfStatement();
82 if (TheCondState.TheCond != StartingCondState.TheCond ||
83 TheCondState.Ignore != StartingCondState.Ignore)
84 return TokError("unmatched .ifs or .elses");
89 /// ParseConditionalAssemblyDirectives - parse the conditional assembly
91 bool AsmParser::ParseConditionalAssemblyDirectives(StringRef Directive,
93 if (Directive == ".if")
94 return ParseDirectiveIf(DirectiveLoc);
95 if (Directive == ".elseif")
96 return ParseDirectiveElseIf(DirectiveLoc);
97 if (Directive == ".else")
98 return ParseDirectiveElse(DirectiveLoc);
99 if (Directive == ".endif")
100 return ParseDirectiveEndIf(DirectiveLoc);
104 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
105 void AsmParser::EatToEndOfStatement() {
106 while (Lexer.isNot(AsmToken::EndOfStatement) &&
107 Lexer.isNot(AsmToken::Eof))
111 if (Lexer.is(AsmToken::EndOfStatement))
116 /// ParseParenExpr - Parse a paren expression and return it.
117 /// NOTE: This assumes the leading '(' has already been consumed.
119 /// parenexpr ::= expr)
121 bool AsmParser::ParseParenExpr(AsmExpr *&Res) {
122 if (ParseExpression(Res)) return true;
123 if (Lexer.isNot(AsmToken::RParen))
124 return TokError("expected ')' in parentheses expression");
129 /// ParsePrimaryExpr - Parse a primary expression and return it.
130 /// primaryexpr ::= (parenexpr
131 /// primaryexpr ::= symbol
132 /// primaryexpr ::= number
133 /// primaryexpr ::= ~,+,- primaryexpr
134 bool AsmParser::ParsePrimaryExpr(AsmExpr *&Res) {
135 switch (Lexer.getKind()) {
137 return TokError("unknown token in expression");
138 case AsmToken::Exclaim:
139 Lexer.Lex(); // Eat the operator.
140 if (ParsePrimaryExpr(Res))
142 Res = new AsmUnaryExpr(AsmUnaryExpr::LNot, Res);
144 case AsmToken::String:
145 case AsmToken::Identifier: {
146 // This is a label, this should be parsed as part of an expression, to
147 // handle things like LFOO+4.
148 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getTok().getIdentifier());
150 // If this is use of an undefined symbol then mark it external.
151 if (!Sym->getSection() && !Ctx.GetSymbolValue(Sym))
152 Sym->setExternal(true);
154 Res = new AsmSymbolRefExpr(Sym);
155 Lexer.Lex(); // Eat identifier.
158 case AsmToken::Integer:
159 Res = new AsmConstantExpr(Lexer.getTok().getIntVal());
160 Lexer.Lex(); // Eat token.
162 case AsmToken::LParen:
163 Lexer.Lex(); // Eat the '('.
164 return ParseParenExpr(Res);
165 case AsmToken::Minus:
166 Lexer.Lex(); // Eat the operator.
167 if (ParsePrimaryExpr(Res))
169 Res = new AsmUnaryExpr(AsmUnaryExpr::Minus, Res);
172 Lexer.Lex(); // Eat the operator.
173 if (ParsePrimaryExpr(Res))
175 Res = new AsmUnaryExpr(AsmUnaryExpr::Plus, Res);
177 case AsmToken::Tilde:
178 Lexer.Lex(); // Eat the operator.
179 if (ParsePrimaryExpr(Res))
181 Res = new AsmUnaryExpr(AsmUnaryExpr::Not, Res);
186 /// ParseExpression - Parse an expression and return it.
188 /// expr ::= expr +,- expr -> lowest.
189 /// expr ::= expr |,^,&,! expr -> middle.
190 /// expr ::= expr *,/,%,<<,>> expr -> highest.
191 /// expr ::= primaryexpr
193 bool AsmParser::ParseExpression(AsmExpr *&Res) {
195 return ParsePrimaryExpr(Res) ||
196 ParseBinOpRHS(1, Res);
199 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
202 SMLoc StartLoc = Lexer.getLoc();
203 if (ParseExpression(Expr))
206 if (!Expr->EvaluateAsAbsolute(Ctx, Res))
207 return Error(StartLoc, "expected absolute expression");
212 bool AsmParser::ParseRelocatableExpression(MCValue &Res) {
215 SMLoc StartLoc = Lexer.getLoc();
216 if (ParseExpression(Expr))
219 if (!Expr->EvaluateAsRelocatable(Ctx, Res))
220 return Error(StartLoc, "expected relocatable expression");
225 bool AsmParser::ParseParenRelocatableExpression(MCValue &Res) {
228 SMLoc StartLoc = Lexer.getLoc();
229 if (ParseParenExpr(Expr))
232 if (!Expr->EvaluateAsRelocatable(Ctx, Res))
233 return Error(StartLoc, "expected relocatable expression");
238 static unsigned getBinOpPrecedence(AsmToken::TokenKind K,
239 AsmBinaryExpr::Opcode &Kind) {
241 default: return 0; // not a binop.
243 // Lowest Precedence: &&, ||
244 case AsmToken::AmpAmp:
245 Kind = AsmBinaryExpr::LAnd;
247 case AsmToken::PipePipe:
248 Kind = AsmBinaryExpr::LOr;
251 // Low Precedence: +, -, ==, !=, <>, <, <=, >, >=
253 Kind = AsmBinaryExpr::Add;
255 case AsmToken::Minus:
256 Kind = AsmBinaryExpr::Sub;
258 case AsmToken::EqualEqual:
259 Kind = AsmBinaryExpr::EQ;
261 case AsmToken::ExclaimEqual:
262 case AsmToken::LessGreater:
263 Kind = AsmBinaryExpr::NE;
266 Kind = AsmBinaryExpr::LT;
268 case AsmToken::LessEqual:
269 Kind = AsmBinaryExpr::LTE;
271 case AsmToken::Greater:
272 Kind = AsmBinaryExpr::GT;
274 case AsmToken::GreaterEqual:
275 Kind = AsmBinaryExpr::GTE;
278 // Intermediate Precedence: |, &, ^
280 // FIXME: gas seems to support '!' as an infix operator?
282 Kind = AsmBinaryExpr::Or;
284 case AsmToken::Caret:
285 Kind = AsmBinaryExpr::Xor;
288 Kind = AsmBinaryExpr::And;
291 // Highest Precedence: *, /, %, <<, >>
293 Kind = AsmBinaryExpr::Mul;
295 case AsmToken::Slash:
296 Kind = AsmBinaryExpr::Div;
298 case AsmToken::Percent:
299 Kind = AsmBinaryExpr::Mod;
301 case AsmToken::LessLess:
302 Kind = AsmBinaryExpr::Shl;
304 case AsmToken::GreaterGreater:
305 Kind = AsmBinaryExpr::Shr;
311 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
312 /// Res contains the LHS of the expression on input.
313 bool AsmParser::ParseBinOpRHS(unsigned Precedence, AsmExpr *&Res) {
315 AsmBinaryExpr::Opcode Kind = AsmBinaryExpr::Add;
316 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
318 // If the next token is lower precedence than we are allowed to eat, return
319 // successfully with what we ate already.
320 if (TokPrec < Precedence)
325 // Eat the next primary expression.
327 if (ParsePrimaryExpr(RHS)) return true;
329 // If BinOp binds less tightly with RHS than the operator after RHS, let
330 // the pending operator take RHS as its LHS.
331 AsmBinaryExpr::Opcode Dummy;
332 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
333 if (TokPrec < NextTokPrec) {
334 if (ParseBinOpRHS(Precedence+1, RHS)) return true;
337 // Merge LHS and RHS according to operator.
338 Res = new AsmBinaryExpr(Kind, Res, RHS);
346 /// ::= EndOfStatement
347 /// ::= Label* Directive ...Operands... EndOfStatement
348 /// ::= Label* Identifier OperandList* EndOfStatement
349 bool AsmParser::ParseStatement() {
350 if (Lexer.is(AsmToken::EndOfStatement)) {
355 // Statements always start with an identifier.
356 AsmToken ID = Lexer.getTok();
357 SMLoc IDLoc = ID.getLoc();
359 if (ParseIdentifier(IDVal))
360 return TokError("unexpected token at start of statement");
362 // FIXME: Recurse on local labels?
364 // See what kind of statement we have.
365 switch (Lexer.getKind()) {
366 case AsmToken::Colon: {
367 // identifier ':' -> Label.
370 // Diagnose attempt to use a variable as a label.
372 // FIXME: Diagnostics. Note the location of the definition as a label.
373 // FIXME: This doesn't diagnose assignment to a symbol which has been
374 // implicitly marked as external.
375 MCSymbol *Sym = Ctx.GetOrCreateSymbol(IDVal);
376 if (Sym->getSection())
377 return Error(IDLoc, "invalid symbol redefinition");
378 if (Ctx.GetSymbolValue(Sym))
379 return Error(IDLoc, "symbol already used as assembler variable");
381 // Since we saw a label, create a symbol and emit it.
382 // FIXME: If the label starts with L it is an assembler temporary label.
383 // Why does the client of this api need to know this?
386 return ParseStatement();
389 case AsmToken::Equal:
390 // identifier '=' ... -> assignment statement
393 return ParseAssignment(IDVal, false);
395 default: // Normal instruction or directive.
399 // Otherwise, we have a normal instruction or directive.
400 if (IDVal[0] == '.') {
401 // FIXME: This should be driven based on a hash lookup and callback.
402 if (IDVal == ".section")
403 return ParseDirectiveDarwinSection();
404 if (IDVal == ".text")
405 // FIXME: This changes behavior based on the -static flag to the
407 return ParseDirectiveSectionSwitch("__TEXT", "__text",
408 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS);
409 if (IDVal == ".const")
410 return ParseDirectiveSectionSwitch("__TEXT", "__const", 0);
411 if (IDVal == ".static_const")
412 return ParseDirectiveSectionSwitch("__TEXT", "__static_const", 0);
413 if (IDVal == ".cstring")
414 return ParseDirectiveSectionSwitch("__TEXT","__cstring",
415 MCSectionMachO::S_CSTRING_LITERALS);
416 if (IDVal == ".literal4")
417 return ParseDirectiveSectionSwitch("__TEXT", "__literal4",
418 MCSectionMachO::S_4BYTE_LITERALS);
419 if (IDVal == ".literal8")
420 return ParseDirectiveSectionSwitch("__TEXT", "__literal8",
421 MCSectionMachO::S_8BYTE_LITERALS);
422 if (IDVal == ".literal16")
423 return ParseDirectiveSectionSwitch("__TEXT","__literal16",
424 MCSectionMachO::S_16BYTE_LITERALS);
425 if (IDVal == ".constructor")
426 return ParseDirectiveSectionSwitch("__TEXT","__constructor", 0);
427 if (IDVal == ".destructor")
428 return ParseDirectiveSectionSwitch("__TEXT","__destructor", 0);
429 if (IDVal == ".fvmlib_init0")
430 return ParseDirectiveSectionSwitch("__TEXT","__fvmlib_init0", 0);
431 if (IDVal == ".fvmlib_init1")
432 return ParseDirectiveSectionSwitch("__TEXT","__fvmlib_init1", 0);
433 if (IDVal == ".symbol_stub")
434 return ParseDirectiveSectionSwitch("__TEXT","__symbol_stub",
435 MCSectionMachO::S_SYMBOL_STUBS |
436 MCSectionMachO::S_ATTR_SELF_MODIFYING_CODE |
437 MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
438 // FIXME: Different on PPC and ARM.
440 // FIXME: .picsymbol_stub on PPC.
441 if (IDVal == ".data")
442 return ParseDirectiveSectionSwitch("__DATA", "__data", 0);
443 if (IDVal == ".static_data")
444 return ParseDirectiveSectionSwitch("__DATA", "__static_data", 0);
445 if (IDVal == ".non_lazy_symbol_pointer")
446 return ParseDirectiveSectionSwitch("__DATA", "__nl_symbol_pointer",
447 MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS);
448 if (IDVal == ".lazy_symbol_pointer")
449 return ParseDirectiveSectionSwitch("__DATA", "__la_symbol_pointer",
450 MCSectionMachO::S_LAZY_SYMBOL_POINTERS);
451 if (IDVal == ".dyld")
452 return ParseDirectiveSectionSwitch("__DATA", "__dyld", 0);
453 if (IDVal == ".mod_init_func")
454 return ParseDirectiveSectionSwitch("__DATA", "__mod_init_func",
455 MCSectionMachO::S_MOD_INIT_FUNC_POINTERS);
456 if (IDVal == ".mod_term_func")
457 return ParseDirectiveSectionSwitch("__DATA", "__mod_term_func",
458 MCSectionMachO::S_MOD_TERM_FUNC_POINTERS);
459 if (IDVal == ".const_data")
460 return ParseDirectiveSectionSwitch("__DATA", "__const", 0);
463 // FIXME: Verify attributes on sections.
464 if (IDVal == ".objc_class")
465 return ParseDirectiveSectionSwitch("__OBJC", "__class", 0);
466 if (IDVal == ".objc_meta_class")
467 return ParseDirectiveSectionSwitch("__OBJC", "__meta_class", 0);
468 if (IDVal == ".objc_cat_cls_meth")
469 return ParseDirectiveSectionSwitch("__OBJC", "__cat_cls_meth", 0);
470 if (IDVal == ".objc_cat_inst_meth")
471 return ParseDirectiveSectionSwitch("__OBJC", "__cat_inst_meth", 0);
472 if (IDVal == ".objc_protocol")
473 return ParseDirectiveSectionSwitch("__OBJC", "__protocol", 0);
474 if (IDVal == ".objc_string_object")
475 return ParseDirectiveSectionSwitch("__OBJC", "__string_object", 0);
476 if (IDVal == ".objc_cls_meth")
477 return ParseDirectiveSectionSwitch("__OBJC", "__cls_meth", 0);
478 if (IDVal == ".objc_inst_meth")
479 return ParseDirectiveSectionSwitch("__OBJC", "__inst_meth", 0);
480 if (IDVal == ".objc_cls_refs")
481 return ParseDirectiveSectionSwitch("__OBJC", "__cls_refs", 0);
482 if (IDVal == ".objc_message_refs")
483 return ParseDirectiveSectionSwitch("__OBJC", "__message_refs", 0);
484 if (IDVal == ".objc_symbols")
485 return ParseDirectiveSectionSwitch("__OBJC", "__symbols", 0);
486 if (IDVal == ".objc_category")
487 return ParseDirectiveSectionSwitch("__OBJC", "__category", 0);
488 if (IDVal == ".objc_class_vars")
489 return ParseDirectiveSectionSwitch("__OBJC", "__class_vars", 0);
490 if (IDVal == ".objc_instance_vars")
491 return ParseDirectiveSectionSwitch("__OBJC", "__instance_vars", 0);
492 if (IDVal == ".objc_module_info")
493 return ParseDirectiveSectionSwitch("__OBJC", "__module_info", 0);
494 if (IDVal == ".objc_class_names")
495 return ParseDirectiveSectionSwitch("__TEXT", "__cstring",
496 MCSectionMachO::S_CSTRING_LITERALS);
497 if (IDVal == ".objc_meth_var_types")
498 return ParseDirectiveSectionSwitch("__TEXT", "__cstring",
499 MCSectionMachO::S_CSTRING_LITERALS);
500 if (IDVal == ".objc_meth_var_names")
501 return ParseDirectiveSectionSwitch("__TEXT", "__cstring",
502 MCSectionMachO::S_CSTRING_LITERALS);
503 if (IDVal == ".objc_selector_strs")
504 return ParseDirectiveSectionSwitch("__OBJC", "__selector_strs", 0);
506 // Assembler features
508 return ParseDirectiveSet();
512 if (IDVal == ".ascii")
513 return ParseDirectiveAscii(false);
514 if (IDVal == ".asciz")
515 return ParseDirectiveAscii(true);
517 // FIXME: Target hooks for size? Also for "word", "hword".
518 if (IDVal == ".byte")
519 return ParseDirectiveValue(1);
520 if (IDVal == ".short")
521 return ParseDirectiveValue(2);
522 if (IDVal == ".long")
523 return ParseDirectiveValue(4);
524 if (IDVal == ".quad")
525 return ParseDirectiveValue(8);
527 // FIXME: Target hooks for IsPow2.
528 if (IDVal == ".align")
529 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
530 if (IDVal == ".align32")
531 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
532 if (IDVal == ".balign")
533 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
534 if (IDVal == ".balignw")
535 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
536 if (IDVal == ".balignl")
537 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
538 if (IDVal == ".p2align")
539 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
540 if (IDVal == ".p2alignw")
541 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
542 if (IDVal == ".p2alignl")
543 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
546 return ParseDirectiveOrg();
548 if (IDVal == ".fill")
549 return ParseDirectiveFill();
550 if (IDVal == ".space")
551 return ParseDirectiveSpace();
553 // Symbol attribute directives
554 if (IDVal == ".globl" || IDVal == ".global")
555 return ParseDirectiveSymbolAttribute(MCStreamer::Global);
556 if (IDVal == ".hidden")
557 return ParseDirectiveSymbolAttribute(MCStreamer::Hidden);
558 if (IDVal == ".indirect_symbol")
559 return ParseDirectiveSymbolAttribute(MCStreamer::IndirectSymbol);
560 if (IDVal == ".internal")
561 return ParseDirectiveSymbolAttribute(MCStreamer::Internal);
562 if (IDVal == ".lazy_reference")
563 return ParseDirectiveSymbolAttribute(MCStreamer::LazyReference);
564 if (IDVal == ".no_dead_strip")
565 return ParseDirectiveSymbolAttribute(MCStreamer::NoDeadStrip);
566 if (IDVal == ".private_extern")
567 return ParseDirectiveSymbolAttribute(MCStreamer::PrivateExtern);
568 if (IDVal == ".protected")
569 return ParseDirectiveSymbolAttribute(MCStreamer::Protected);
570 if (IDVal == ".reference")
571 return ParseDirectiveSymbolAttribute(MCStreamer::Reference);
572 if (IDVal == ".weak")
573 return ParseDirectiveSymbolAttribute(MCStreamer::Weak);
574 if (IDVal == ".weak_definition")
575 return ParseDirectiveSymbolAttribute(MCStreamer::WeakDefinition);
576 if (IDVal == ".weak_reference")
577 return ParseDirectiveSymbolAttribute(MCStreamer::WeakReference);
579 if (IDVal == ".comm")
580 return ParseDirectiveComm(/*IsLocal=*/false);
581 if (IDVal == ".lcomm")
582 return ParseDirectiveComm(/*IsLocal=*/true);
583 if (IDVal == ".zerofill")
584 return ParseDirectiveDarwinZerofill();
585 if (IDVal == ".desc")
586 return ParseDirectiveDarwinSymbolDesc();
587 if (IDVal == ".lsym")
588 return ParseDirectiveDarwinLsym();
590 if (IDVal == ".subsections_via_symbols")
591 return ParseDirectiveDarwinSubsectionsViaSymbols();
592 if (IDVal == ".abort")
593 return ParseDirectiveAbort();
594 if (IDVal == ".include")
595 return ParseDirectiveInclude();
596 if (IDVal == ".dump")
597 return ParseDirectiveDarwinDumpOrLoad(IDLoc, /*IsDump=*/true);
598 if (IDVal == ".load")
599 return ParseDirectiveDarwinDumpOrLoad(IDLoc, /*IsLoad=*/false);
601 Warning(IDLoc, "ignoring directive for now");
602 EatToEndOfStatement();
607 if (getTargetParser().ParseInstruction(IDVal, Inst))
610 if (Lexer.isNot(AsmToken::EndOfStatement))
611 return TokError("unexpected token in argument list");
613 // Eat the end of statement marker.
616 // Instruction is good, process it.
617 Out.EmitInstruction(Inst);
619 // Skip to end of line for now.
623 bool AsmParser::ParseAssignment(const StringRef &Name, bool IsDotSet) {
624 // FIXME: Use better location, we should use proper tokens.
625 SMLoc EqualLoc = Lexer.getLoc();
628 if (ParseRelocatableExpression(Value))
631 if (Lexer.isNot(AsmToken::EndOfStatement))
632 return TokError("unexpected token in assignment");
634 // Eat the end of statement marker.
637 // Diagnose assignment to a label.
639 // FIXME: Diagnostics. Note the location of the definition as a label.
640 // FIXME: This doesn't diagnose assignment to a symbol which has been
641 // implicitly marked as external.
642 // FIXME: Handle '.'.
643 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
644 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
645 if (Sym->getSection())
646 return Error(EqualLoc, "invalid assignment to symbol emitted as a label");
647 if (Sym->isExternal())
648 return Error(EqualLoc, "invalid assignment to external symbol");
650 // Do the assignment.
651 Out.EmitAssignment(Sym, Value, IsDotSet);
659 bool AsmParser::ParseIdentifier(StringRef &Res) {
660 if (Lexer.isNot(AsmToken::Identifier) &&
661 Lexer.isNot(AsmToken::String))
664 Res = Lexer.getTok().getIdentifier();
666 Lexer.Lex(); // Consume the identifier token.
671 /// ParseDirectiveSet:
672 /// ::= .set identifier ',' expression
673 bool AsmParser::ParseDirectiveSet() {
676 if (ParseIdentifier(Name))
677 return TokError("expected identifier after '.set' directive");
679 if (Lexer.isNot(AsmToken::Comma))
680 return TokError("unexpected token in '.set'");
683 return ParseAssignment(Name, true);
686 /// ParseDirectiveSection:
687 /// ::= .section identifier (',' identifier)*
688 /// FIXME: This should actually parse out the segment, section, attributes and
689 /// sizeof_stub fields.
690 bool AsmParser::ParseDirectiveDarwinSection() {
691 StringRef SectionName;
693 if (Lexer.isNot(AsmToken::Identifier))
694 return TokError("expected identifier after '.section' directive");
696 std::string SectionSpec = SectionName;
697 StringRef EOL = Lexer.LexUntilEndOfStatement();
698 SectionSpec.append(EOL.begin(), EOL.end());
700 if (Lexer.isNot(AsmToken::EndOfStatement))
701 return TokError("unexpected token in '.section' directive");
705 StringRef Segment, Section;
706 unsigned TAA, StubSize;
707 std::string ErrorStr =
708 MCSectionMachO::ParseSectionSpecifier(SectionSpec, Segment, Section,
711 if (!ErrorStr.empty())
712 return TokError(ErrorStr.c_str());
714 // FIXME: CACHE THESE.
716 // FIXME: Arch specific.
717 MCSection *S = 0; //Ctx.GetSection(Section);
719 S = MCSectionMachO::Create(Segment, Section, TAA, StubSize,
722 Out.SwitchSection(S);
726 /// ParseDirectiveSectionSwitch -
728 /// FIXME! Many of these directives implicitly cause a ".align" directive to get
729 /// emitted, we don't do this yet which can lead to subtle miscompiles.
730 bool AsmParser::ParseDirectiveSectionSwitch(const char *Segment,
732 unsigned TAA, unsigned StubSize) {
733 if (Lexer.isNot(AsmToken::EndOfStatement))
734 return TokError("unexpected token in section switching directive");
737 // FIXME: Arch specific.
738 // FIXME: Cache this!
739 MCSection *S = 0; // Ctx.GetSection(Section);
741 S = MCSectionMachO::Create(Segment, Section, TAA, StubSize,
744 Out.SwitchSection(S);
748 /// ParseDirectiveAscii:
749 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
750 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
751 if (Lexer.isNot(AsmToken::EndOfStatement)) {
753 if (Lexer.isNot(AsmToken::String))
754 return TokError("expected string in '.ascii' or '.asciz' directive");
756 // FIXME: This shouldn't use a const char* + strlen, the string could have
758 // FIXME: Should have accessor for getting string contents.
759 StringRef Str = Lexer.getTok().getString();
760 Out.EmitBytes(Str.substr(1, Str.size() - 2));
762 Out.EmitBytes(StringRef("\0", 1));
766 if (Lexer.is(AsmToken::EndOfStatement))
769 if (Lexer.isNot(AsmToken::Comma))
770 return TokError("unexpected token in '.ascii' or '.asciz' directive");
779 /// ParseDirectiveValue
780 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
781 bool AsmParser::ParseDirectiveValue(unsigned Size) {
782 if (Lexer.isNot(AsmToken::EndOfStatement)) {
785 if (ParseRelocatableExpression(Expr))
788 Out.EmitValue(Expr, Size);
790 if (Lexer.is(AsmToken::EndOfStatement))
793 // FIXME: Improve diagnostic.
794 if (Lexer.isNot(AsmToken::Comma))
795 return TokError("unexpected token in directive");
804 /// ParseDirectiveSpace
805 /// ::= .space expression [ , expression ]
806 bool AsmParser::ParseDirectiveSpace() {
808 if (ParseAbsoluteExpression(NumBytes))
811 int64_t FillExpr = 0;
812 bool HasFillExpr = false;
813 if (Lexer.isNot(AsmToken::EndOfStatement)) {
814 if (Lexer.isNot(AsmToken::Comma))
815 return TokError("unexpected token in '.space' directive");
818 if (ParseAbsoluteExpression(FillExpr))
823 if (Lexer.isNot(AsmToken::EndOfStatement))
824 return TokError("unexpected token in '.space' directive");
830 return TokError("invalid number of bytes in '.space' directive");
832 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
833 for (uint64_t i = 0, e = NumBytes; i != e; ++i)
834 Out.EmitValue(MCValue::get(FillExpr), 1);
839 /// ParseDirectiveFill
840 /// ::= .fill expression , expression , expression
841 bool AsmParser::ParseDirectiveFill() {
843 if (ParseAbsoluteExpression(NumValues))
846 if (Lexer.isNot(AsmToken::Comma))
847 return TokError("unexpected token in '.fill' directive");
851 if (ParseAbsoluteExpression(FillSize))
854 if (Lexer.isNot(AsmToken::Comma))
855 return TokError("unexpected token in '.fill' directive");
859 if (ParseAbsoluteExpression(FillExpr))
862 if (Lexer.isNot(AsmToken::EndOfStatement))
863 return TokError("unexpected token in '.fill' directive");
867 if (FillSize != 1 && FillSize != 2 && FillSize != 4)
868 return TokError("invalid '.fill' size, expected 1, 2, or 4");
870 for (uint64_t i = 0, e = NumValues; i != e; ++i)
871 Out.EmitValue(MCValue::get(FillExpr), FillSize);
876 /// ParseDirectiveOrg
877 /// ::= .org expression [ , expression ]
878 bool AsmParser::ParseDirectiveOrg() {
880 if (ParseRelocatableExpression(Offset))
883 // Parse optional fill expression.
884 int64_t FillExpr = 0;
885 if (Lexer.isNot(AsmToken::EndOfStatement)) {
886 if (Lexer.isNot(AsmToken::Comma))
887 return TokError("unexpected token in '.org' directive");
890 if (ParseAbsoluteExpression(FillExpr))
893 if (Lexer.isNot(AsmToken::EndOfStatement))
894 return TokError("unexpected token in '.org' directive");
899 // FIXME: Only limited forms of relocatable expressions are accepted here, it
900 // has to be relative to the current section.
901 Out.EmitValueToOffset(Offset, FillExpr);
906 /// ParseDirectiveAlign
907 /// ::= {.align, ...} expression [ , expression [ , expression ]]
908 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
910 if (ParseAbsoluteExpression(Alignment))
914 bool HasFillExpr = false;
915 int64_t FillExpr = 0;
916 int64_t MaxBytesToFill = 0;
917 if (Lexer.isNot(AsmToken::EndOfStatement)) {
918 if (Lexer.isNot(AsmToken::Comma))
919 return TokError("unexpected token in directive");
922 // The fill expression can be omitted while specifying a maximum number of
923 // alignment bytes, e.g:
925 if (Lexer.isNot(AsmToken::Comma)) {
927 if (ParseAbsoluteExpression(FillExpr))
931 if (Lexer.isNot(AsmToken::EndOfStatement)) {
932 if (Lexer.isNot(AsmToken::Comma))
933 return TokError("unexpected token in directive");
936 MaxBytesLoc = Lexer.getLoc();
937 if (ParseAbsoluteExpression(MaxBytesToFill))
940 if (Lexer.isNot(AsmToken::EndOfStatement))
941 return TokError("unexpected token in directive");
948 // FIXME: Sometimes fill with nop.
952 // Compute alignment in bytes.
954 // FIXME: Diagnose overflow.
955 Alignment = 1LL << Alignment;
958 // Diagnose non-sensical max bytes to fill.
959 if (MaxBytesLoc.isValid()) {
960 if (MaxBytesToFill < 1) {
961 Warning(MaxBytesLoc, "alignment directive can never be satisfied in this "
962 "many bytes, ignoring");
966 if (MaxBytesToFill >= Alignment) {
967 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
973 // FIXME: Target specific behavior about how the "extra" bytes are filled.
974 Out.EmitValueToAlignment(Alignment, FillExpr, ValueSize, MaxBytesToFill);
979 /// ParseDirectiveSymbolAttribute
980 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
981 bool AsmParser::ParseDirectiveSymbolAttribute(MCStreamer::SymbolAttr Attr) {
982 if (Lexer.isNot(AsmToken::EndOfStatement)) {
986 if (ParseIdentifier(Name))
987 return TokError("expected identifier in directive");
989 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
991 // If this is use of an undefined symbol then mark it external.
992 if (!Sym->getSection() && !Ctx.GetSymbolValue(Sym))
993 Sym->setExternal(true);
995 Out.EmitSymbolAttribute(Sym, Attr);
997 if (Lexer.is(AsmToken::EndOfStatement))
1000 if (Lexer.isNot(AsmToken::Comma))
1001 return TokError("unexpected token in directive");
1010 /// ParseDirectiveDarwinSymbolDesc
1011 /// ::= .desc identifier , expression
1012 bool AsmParser::ParseDirectiveDarwinSymbolDesc() {
1014 if (ParseIdentifier(Name))
1015 return TokError("expected identifier in directive");
1017 // Handle the identifier as the key symbol.
1018 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
1020 if (Lexer.isNot(AsmToken::Comma))
1021 return TokError("unexpected token in '.desc' directive");
1024 SMLoc DescLoc = Lexer.getLoc();
1026 if (ParseAbsoluteExpression(DescValue))
1029 if (Lexer.isNot(AsmToken::EndOfStatement))
1030 return TokError("unexpected token in '.desc' directive");
1034 // Set the n_desc field of this Symbol to this DescValue
1035 Out.EmitSymbolDesc(Sym, DescValue);
1040 /// ParseDirectiveComm
1041 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
1042 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
1043 SMLoc IDLoc = Lexer.getLoc();
1045 if (ParseIdentifier(Name))
1046 return TokError("expected identifier in directive");
1048 // Handle the identifier as the key symbol.
1049 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
1051 if (Lexer.isNot(AsmToken::Comma))
1052 return TokError("unexpected token in directive");
1056 SMLoc SizeLoc = Lexer.getLoc();
1057 if (ParseAbsoluteExpression(Size))
1060 int64_t Pow2Alignment = 0;
1061 SMLoc Pow2AlignmentLoc;
1062 if (Lexer.is(AsmToken::Comma)) {
1064 Pow2AlignmentLoc = Lexer.getLoc();
1065 if (ParseAbsoluteExpression(Pow2Alignment))
1069 if (Lexer.isNot(AsmToken::EndOfStatement))
1070 return TokError("unexpected token in '.comm' or '.lcomm' directive");
1074 // NOTE: a size of zero for a .comm should create a undefined symbol
1075 // but a size of .lcomm creates a bss symbol of size zero.
1077 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
1078 "be less than zero");
1080 // NOTE: The alignment in the directive is a power of 2 value, the assember
1081 // may internally end up wanting an alignment in bytes.
1082 // FIXME: Diagnose overflow.
1083 if (Pow2Alignment < 0)
1084 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
1085 "alignment, can't be less than zero");
1087 // TODO: Symbol must be undefined or it is a error to re-defined the symbol
1088 if (Sym->getSection() || Ctx.GetSymbolValue(Sym))
1089 return Error(IDLoc, "invalid symbol redefinition");
1091 // Create the Symbol as a common or local common with Size and Pow2Alignment
1092 Out.EmitCommonSymbol(Sym, Size, Pow2Alignment, IsLocal);
1097 /// ParseDirectiveDarwinZerofill
1098 /// ::= .zerofill segname , sectname [, identifier , size_expression [
1099 /// , align_expression ]]
1100 bool AsmParser::ParseDirectiveDarwinZerofill() {
1101 // FIXME: Handle quoted names here.
1103 if (Lexer.isNot(AsmToken::Identifier))
1104 return TokError("expected segment name after '.zerofill' directive");
1105 StringRef Segment = Lexer.getTok().getString();
1108 if (Lexer.isNot(AsmToken::Comma))
1109 return TokError("unexpected token in directive");
1112 if (Lexer.isNot(AsmToken::Identifier))
1113 return TokError("expected section name after comma in '.zerofill' "
1115 StringRef Section = Lexer.getTok().getString();
1118 // If this is the end of the line all that was wanted was to create the
1119 // the section but with no symbol.
1120 if (Lexer.is(AsmToken::EndOfStatement)) {
1121 // FIXME: CACHE THIS.
1122 MCSection *S = 0; //Ctx.GetSection(Section);
1124 S = MCSectionMachO::Create(Segment, Section,
1125 MCSectionMachO::S_ZEROFILL, 0,
1126 SectionKind(), Ctx);
1128 // Create the zerofill section but no symbol
1129 Out.EmitZerofill(S);
1133 if (Lexer.isNot(AsmToken::Comma))
1134 return TokError("unexpected token in directive");
1137 if (Lexer.isNot(AsmToken::Identifier))
1138 return TokError("expected identifier in directive");
1140 // handle the identifier as the key symbol.
1141 SMLoc IDLoc = Lexer.getLoc();
1142 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getTok().getString());
1145 if (Lexer.isNot(AsmToken::Comma))
1146 return TokError("unexpected token in directive");
1150 SMLoc SizeLoc = Lexer.getLoc();
1151 if (ParseAbsoluteExpression(Size))
1154 int64_t Pow2Alignment = 0;
1155 SMLoc Pow2AlignmentLoc;
1156 if (Lexer.is(AsmToken::Comma)) {
1158 Pow2AlignmentLoc = Lexer.getLoc();
1159 if (ParseAbsoluteExpression(Pow2Alignment))
1163 if (Lexer.isNot(AsmToken::EndOfStatement))
1164 return TokError("unexpected token in '.zerofill' directive");
1169 return Error(SizeLoc, "invalid '.zerofill' directive size, can't be less "
1172 // NOTE: The alignment in the directive is a power of 2 value, the assember
1173 // may internally end up wanting an alignment in bytes.
1174 // FIXME: Diagnose overflow.
1175 if (Pow2Alignment < 0)
1176 return Error(Pow2AlignmentLoc, "invalid '.zerofill' directive alignment, "
1177 "can't be less than zero");
1179 // TODO: Symbol must be undefined or it is a error to re-defined the symbol
1180 if (Sym->getSection() || Ctx.GetSymbolValue(Sym))
1181 return Error(IDLoc, "invalid symbol redefinition");
1183 // FIXME: Arch specific.
1185 MCSection *S = 0; //Ctx.GetSection(Section);
1187 S = MCSectionMachO::Create(Segment, Section,
1188 MCSectionMachO::S_ZEROFILL, 0,
1189 SectionKind(), Ctx);
1191 // Create the zerofill Symbol with Size and Pow2Alignment
1192 Out.EmitZerofill(S, Sym, Size, Pow2Alignment);
1197 /// ParseDirectiveDarwinSubsectionsViaSymbols
1198 /// ::= .subsections_via_symbols
1199 bool AsmParser::ParseDirectiveDarwinSubsectionsViaSymbols() {
1200 if (Lexer.isNot(AsmToken::EndOfStatement))
1201 return TokError("unexpected token in '.subsections_via_symbols' directive");
1205 Out.EmitAssemblerFlag(MCStreamer::SubsectionsViaSymbols);
1210 /// ParseDirectiveAbort
1211 /// ::= .abort [ "abort_string" ]
1212 bool AsmParser::ParseDirectiveAbort() {
1213 // FIXME: Use loc from directive.
1214 SMLoc Loc = Lexer.getLoc();
1217 if (Lexer.isNot(AsmToken::EndOfStatement)) {
1218 if (Lexer.isNot(AsmToken::String))
1219 return TokError("expected string in '.abort' directive");
1221 Str = Lexer.getTok().getString();
1226 if (Lexer.isNot(AsmToken::EndOfStatement))
1227 return TokError("unexpected token in '.abort' directive");
1231 // FIXME: Handle here.
1233 Error(Loc, ".abort detected. Assembly stopping.");
1235 Error(Loc, ".abort '" + Str + "' detected. Assembly stopping.");
1240 /// ParseDirectiveLsym
1241 /// ::= .lsym identifier , expression
1242 bool AsmParser::ParseDirectiveDarwinLsym() {
1244 if (ParseIdentifier(Name))
1245 return TokError("expected identifier in directive");
1247 // Handle the identifier as the key symbol.
1248 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
1250 if (Lexer.isNot(AsmToken::Comma))
1251 return TokError("unexpected token in '.lsym' directive");
1255 if (ParseRelocatableExpression(Expr))
1258 if (Lexer.isNot(AsmToken::EndOfStatement))
1259 return TokError("unexpected token in '.lsym' directive");
1263 // Create the Sym with the value of the Expr
1264 Out.EmitLocalSymbol(Sym, Expr);
1269 /// ParseDirectiveInclude
1270 /// ::= .include "filename"
1271 bool AsmParser::ParseDirectiveInclude() {
1272 if (Lexer.isNot(AsmToken::String))
1273 return TokError("expected string in '.include' directive");
1275 std::string Filename = Lexer.getTok().getString();
1276 SMLoc IncludeLoc = Lexer.getLoc();
1279 if (Lexer.isNot(AsmToken::EndOfStatement))
1280 return TokError("unexpected token in '.include' directive");
1282 // Strip the quotes.
1283 Filename = Filename.substr(1, Filename.size()-2);
1285 // Attempt to switch the lexer to the included file before consuming the end
1286 // of statement to avoid losing it when we switch.
1287 if (Lexer.EnterIncludeFile(Filename)) {
1288 Lexer.PrintMessage(IncludeLoc,
1289 "Could not find include file '" + Filename + "'",
1297 /// ParseDirectiveDarwinDumpOrLoad
1298 /// ::= ( .dump | .load ) "filename"
1299 bool AsmParser::ParseDirectiveDarwinDumpOrLoad(SMLoc IDLoc, bool IsDump) {
1300 if (Lexer.isNot(AsmToken::String))
1301 return TokError("expected string in '.dump' or '.load' directive");
1305 if (Lexer.isNot(AsmToken::EndOfStatement))
1306 return TokError("unexpected token in '.dump' or '.load' directive");
1310 // FIXME: If/when .dump and .load are implemented they will be done in the
1311 // the assembly parser and not have any need for an MCStreamer API.
1313 Warning(IDLoc, "ignoring directive .dump for now");
1315 Warning(IDLoc, "ignoring directive .load for now");
1320 /// ParseDirectiveIf
1321 /// ::= .if expression
1322 bool AsmParser::ParseDirectiveIf(SMLoc DirectiveLoc) {
1323 // Consume the identifier that was the .if directive
1326 TheCondStack.push_back(TheCondState);
1327 TheCondState.TheCond = AsmCond::IfCond;
1328 if(TheCondState.Ignore) {
1329 EatToEndOfStatement();
1333 if (ParseAbsoluteExpression(ExprValue))
1336 if (Lexer.isNot(AsmToken::EndOfStatement))
1337 return TokError("unexpected token in '.if' directive");
1341 TheCondState.CondMet = ExprValue;
1342 TheCondState.Ignore = !TheCondState.CondMet;
1348 /// ParseDirectiveElseIf
1349 /// ::= .elseif expression
1350 bool AsmParser::ParseDirectiveElseIf(SMLoc DirectiveLoc) {
1351 if (TheCondState.TheCond != AsmCond::IfCond &&
1352 TheCondState.TheCond != AsmCond::ElseIfCond)
1353 Error(DirectiveLoc, "Encountered a .elseif that doesn't follow a .if or "
1355 TheCondState.TheCond = AsmCond::ElseIfCond;
1357 // Consume the identifier that was the .elseif directive
1360 bool LastIgnoreState = false;
1361 if (!TheCondStack.empty())
1362 LastIgnoreState = TheCondStack.back().Ignore;
1363 if (LastIgnoreState || TheCondState.CondMet) {
1364 TheCondState.Ignore = true;
1365 EatToEndOfStatement();
1369 if (ParseAbsoluteExpression(ExprValue))
1372 if (Lexer.isNot(AsmToken::EndOfStatement))
1373 return TokError("unexpected token in '.elseif' directive");
1376 TheCondState.CondMet = ExprValue;
1377 TheCondState.Ignore = !TheCondState.CondMet;
1383 /// ParseDirectiveElse
1385 bool AsmParser::ParseDirectiveElse(SMLoc DirectiveLoc) {
1386 // Consume the identifier that was the .else directive
1389 if (Lexer.isNot(AsmToken::EndOfStatement))
1390 return TokError("unexpected token in '.else' directive");
1394 if (TheCondState.TheCond != AsmCond::IfCond &&
1395 TheCondState.TheCond != AsmCond::ElseIfCond)
1396 Error(DirectiveLoc, "Encountered a .else that doesn't follow a .if or an "
1398 TheCondState.TheCond = AsmCond::ElseCond;
1399 bool LastIgnoreState = false;
1400 if (!TheCondStack.empty())
1401 LastIgnoreState = TheCondStack.back().Ignore;
1402 if (LastIgnoreState || TheCondState.CondMet)
1403 TheCondState.Ignore = true;
1405 TheCondState.Ignore = false;
1410 /// ParseDirectiveEndIf
1412 bool AsmParser::ParseDirectiveEndIf(SMLoc DirectiveLoc) {
1413 // Consume the identifier that was the .endif directive
1416 if (Lexer.isNot(AsmToken::EndOfStatement))
1417 return TokError("unexpected token in '.endif' directive");
1421 if ((TheCondState.TheCond == AsmCond::NoCond) ||
1422 TheCondStack.empty())
1423 Error(DirectiveLoc, "Encountered a .endif that doesn't follow a .if or "
1425 if (!TheCondStack.empty()) {
1426 TheCondState = TheCondStack.back();
1427 TheCondStack.pop_back();