Merge branch 'master' into systemz
[llvm/systemz.git] / tools / llvm-mc / AsmParser.cpp
blob68eb9d56c07b8ab0a6f98d70e84a2373e402bd95
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 "AsmParser.h"
16 #include "AsmExpr.h"
17 #include "llvm/MC/MCContext.h"
18 #include "llvm/MC/MCInst.h"
19 #include "llvm/MC/MCStreamer.h"
20 #include "llvm/MC/MCSymbol.h"
21 #include "llvm/Support/SourceMgr.h"
22 #include "llvm/Support/raw_ostream.h"
23 using namespace llvm;
25 void AsmParser::Warning(SMLoc L, const char *Msg) {
26 Lexer.PrintMessage(L, Msg, "warning");
29 bool AsmParser::Error(SMLoc L, const char *Msg) {
30 Lexer.PrintMessage(L, Msg, "error");
31 return true;
34 bool AsmParser::TokError(const char *Msg) {
35 Lexer.PrintMessage(Lexer.getLoc(), Msg, "error");
36 return true;
39 bool AsmParser::Run() {
40 // Prime the lexer.
41 Lexer.Lex();
43 bool HadError = false;
45 // While we have input, parse each statement.
46 while (Lexer.isNot(asmtok::Eof)) {
47 if (!ParseStatement()) continue;
49 // If we had an error, remember it and recover by skipping to the next line.
50 HadError = true;
51 EatToEndOfStatement();
54 return HadError;
57 /// EatToEndOfStatement - Throw away the rest of the line for testing purposes.
58 void AsmParser::EatToEndOfStatement() {
59 while (Lexer.isNot(asmtok::EndOfStatement) &&
60 Lexer.isNot(asmtok::Eof))
61 Lexer.Lex();
63 // Eat EOL.
64 if (Lexer.is(asmtok::EndOfStatement))
65 Lexer.Lex();
69 /// ParseParenExpr - Parse a paren expression and return it.
70 /// NOTE: This assumes the leading '(' has already been consumed.
71 ///
72 /// parenexpr ::= expr)
73 ///
74 bool AsmParser::ParseParenExpr(AsmExpr *&Res) {
75 if (ParseExpression(Res)) return true;
76 if (Lexer.isNot(asmtok::RParen))
77 return TokError("expected ')' in parentheses expression");
78 Lexer.Lex();
79 return false;
82 /// ParsePrimaryExpr - Parse a primary expression and return it.
83 /// primaryexpr ::= (parenexpr
84 /// primaryexpr ::= symbol
85 /// primaryexpr ::= number
86 /// primaryexpr ::= ~,+,- primaryexpr
87 bool AsmParser::ParsePrimaryExpr(AsmExpr *&Res) {
88 switch (Lexer.getKind()) {
89 default:
90 return TokError("unknown token in expression");
91 case asmtok::Exclaim:
92 Lexer.Lex(); // Eat the operator.
93 if (ParsePrimaryExpr(Res))
94 return true;
95 Res = new AsmUnaryExpr(AsmUnaryExpr::LNot, Res);
96 return false;
97 case asmtok::Identifier: {
98 // This is a label, this should be parsed as part of an expression, to
99 // handle things like LFOO+4.
100 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
102 // If this is use of an undefined symbol then mark it external.
103 if (!Sym->getSection() && !Ctx.GetSymbolValue(Sym))
104 Sym->setExternal(true);
106 Res = new AsmSymbolRefExpr(Sym);
107 Lexer.Lex(); // Eat identifier.
108 return false;
110 case asmtok::IntVal:
111 Res = new AsmConstantExpr(Lexer.getCurIntVal());
112 Lexer.Lex(); // Eat identifier.
113 return false;
114 case asmtok::LParen:
115 Lexer.Lex(); // Eat the '('.
116 return ParseParenExpr(Res);
117 case asmtok::Minus:
118 Lexer.Lex(); // Eat the operator.
119 if (ParsePrimaryExpr(Res))
120 return true;
121 Res = new AsmUnaryExpr(AsmUnaryExpr::Minus, Res);
122 return false;
123 case asmtok::Plus:
124 Lexer.Lex(); // Eat the operator.
125 if (ParsePrimaryExpr(Res))
126 return true;
127 Res = new AsmUnaryExpr(AsmUnaryExpr::Plus, Res);
128 return false;
129 case asmtok::Tilde:
130 Lexer.Lex(); // Eat the operator.
131 if (ParsePrimaryExpr(Res))
132 return true;
133 Res = new AsmUnaryExpr(AsmUnaryExpr::Not, Res);
134 return false;
138 /// ParseExpression - Parse an expression and return it.
139 ///
140 /// expr ::= expr +,- expr -> lowest.
141 /// expr ::= expr |,^,&,! expr -> middle.
142 /// expr ::= expr *,/,%,<<,>> expr -> highest.
143 /// expr ::= primaryexpr
145 bool AsmParser::ParseExpression(AsmExpr *&Res) {
146 Res = 0;
147 return ParsePrimaryExpr(Res) ||
148 ParseBinOpRHS(1, Res);
151 bool AsmParser::ParseAbsoluteExpression(int64_t &Res) {
152 AsmExpr *Expr;
154 SMLoc StartLoc = Lexer.getLoc();
155 if (ParseExpression(Expr))
156 return true;
158 if (!Expr->EvaluateAsAbsolute(Ctx, Res))
159 return Error(StartLoc, "expected absolute expression");
161 return false;
164 bool AsmParser::ParseRelocatableExpression(MCValue &Res) {
165 AsmExpr *Expr;
167 SMLoc StartLoc = Lexer.getLoc();
168 if (ParseExpression(Expr))
169 return true;
171 if (!Expr->EvaluateAsRelocatable(Ctx, Res))
172 return Error(StartLoc, "expected relocatable expression");
174 return false;
177 bool AsmParser::ParseParenRelocatableExpression(MCValue &Res) {
178 AsmExpr *Expr;
180 SMLoc StartLoc = Lexer.getLoc();
181 if (ParseParenExpr(Expr))
182 return true;
184 if (!Expr->EvaluateAsRelocatable(Ctx, Res))
185 return Error(StartLoc, "expected relocatable expression");
187 return false;
190 static unsigned getBinOpPrecedence(asmtok::TokKind K,
191 AsmBinaryExpr::Opcode &Kind) {
192 switch (K) {
193 default: return 0; // not a binop.
195 // Lowest Precedence: &&, ||
196 case asmtok::AmpAmp:
197 Kind = AsmBinaryExpr::LAnd;
198 return 1;
199 case asmtok::PipePipe:
200 Kind = AsmBinaryExpr::LOr;
201 return 1;
203 // Low Precedence: +, -, ==, !=, <>, <, <=, >, >=
204 case asmtok::Plus:
205 Kind = AsmBinaryExpr::Add;
206 return 2;
207 case asmtok::Minus:
208 Kind = AsmBinaryExpr::Sub;
209 return 2;
210 case asmtok::EqualEqual:
211 Kind = AsmBinaryExpr::EQ;
212 return 2;
213 case asmtok::ExclaimEqual:
214 case asmtok::LessGreater:
215 Kind = AsmBinaryExpr::NE;
216 return 2;
217 case asmtok::Less:
218 Kind = AsmBinaryExpr::LT;
219 return 2;
220 case asmtok::LessEqual:
221 Kind = AsmBinaryExpr::LTE;
222 return 2;
223 case asmtok::Greater:
224 Kind = AsmBinaryExpr::GT;
225 return 2;
226 case asmtok::GreaterEqual:
227 Kind = AsmBinaryExpr::GTE;
228 return 2;
230 // Intermediate Precedence: |, &, ^
232 // FIXME: gas seems to support '!' as an infix operator?
233 case asmtok::Pipe:
234 Kind = AsmBinaryExpr::Or;
235 return 3;
236 case asmtok::Caret:
237 Kind = AsmBinaryExpr::Xor;
238 return 3;
239 case asmtok::Amp:
240 Kind = AsmBinaryExpr::And;
241 return 3;
243 // Highest Precedence: *, /, %, <<, >>
244 case asmtok::Star:
245 Kind = AsmBinaryExpr::Mul;
246 return 4;
247 case asmtok::Slash:
248 Kind = AsmBinaryExpr::Div;
249 return 4;
250 case asmtok::Percent:
251 Kind = AsmBinaryExpr::Mod;
252 return 4;
253 case asmtok::LessLess:
254 Kind = AsmBinaryExpr::Shl;
255 return 4;
256 case asmtok::GreaterGreater:
257 Kind = AsmBinaryExpr::Shr;
258 return 4;
263 /// ParseBinOpRHS - Parse all binary operators with precedence >= 'Precedence'.
264 /// Res contains the LHS of the expression on input.
265 bool AsmParser::ParseBinOpRHS(unsigned Precedence, AsmExpr *&Res) {
266 while (1) {
267 AsmBinaryExpr::Opcode Kind = AsmBinaryExpr::Add;
268 unsigned TokPrec = getBinOpPrecedence(Lexer.getKind(), Kind);
270 // If the next token is lower precedence than we are allowed to eat, return
271 // successfully with what we ate already.
272 if (TokPrec < Precedence)
273 return false;
275 Lexer.Lex();
277 // Eat the next primary expression.
278 AsmExpr *RHS;
279 if (ParsePrimaryExpr(RHS)) return true;
281 // If BinOp binds less tightly with RHS than the operator after RHS, let
282 // the pending operator take RHS as its LHS.
283 AsmBinaryExpr::Opcode Dummy;
284 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
285 if (TokPrec < NextTokPrec) {
286 if (ParseBinOpRHS(Precedence+1, RHS)) return true;
289 // Merge LHS and RHS according to operator.
290 Res = new AsmBinaryExpr(Kind, Res, RHS);
297 /// ParseStatement:
298 /// ::= EndOfStatement
299 /// ::= Label* Directive ...Operands... EndOfStatement
300 /// ::= Label* Identifier OperandList* EndOfStatement
301 bool AsmParser::ParseStatement() {
302 switch (Lexer.getKind()) {
303 default:
304 return TokError("unexpected token at start of statement");
305 case asmtok::EndOfStatement:
306 Lexer.Lex();
307 return false;
308 case asmtok::Identifier:
309 break;
310 // TODO: Recurse on local labels etc.
313 // If we have an identifier, handle it as the key symbol.
314 SMLoc IDLoc = Lexer.getLoc();
315 const char *IDVal = Lexer.getCurStrVal();
317 // Consume the identifier, see what is after it.
318 switch (Lexer.Lex()) {
319 case asmtok::Colon: {
320 // identifier ':' -> Label.
321 Lexer.Lex();
323 // Diagnose attempt to use a variable as a label.
325 // FIXME: Diagnostics. Note the location of the definition as a label.
326 // FIXME: This doesn't diagnose assignment to a symbol which has been
327 // implicitly marked as external.
328 MCSymbol *Sym = Ctx.GetOrCreateSymbol(IDVal);
329 if (Sym->getSection())
330 return Error(IDLoc, "invalid symbol redefinition");
331 if (Ctx.GetSymbolValue(Sym))
332 return Error(IDLoc, "symbol already used as assembler variable");
334 // Since we saw a label, create a symbol and emit it.
335 // FIXME: If the label starts with L it is an assembler temporary label.
336 // Why does the client of this api need to know this?
337 Out.EmitLabel(Sym);
339 return ParseStatement();
342 case asmtok::Equal:
343 // identifier '=' ... -> assignment statement
344 Lexer.Lex();
346 return ParseAssignment(IDVal, false);
348 default: // Normal instruction or directive.
349 break;
352 // Otherwise, we have a normal instruction or directive.
353 if (IDVal[0] == '.') {
354 // FIXME: This should be driven based on a hash lookup and callback.
355 if (!strcmp(IDVal, ".section"))
356 return ParseDirectiveDarwinSection();
357 if (!strcmp(IDVal, ".text"))
358 // FIXME: This changes behavior based on the -static flag to the
359 // assembler.
360 return ParseDirectiveSectionSwitch("__TEXT,__text",
361 "regular,pure_instructions");
362 if (!strcmp(IDVal, ".const"))
363 return ParseDirectiveSectionSwitch("__TEXT,__const");
364 if (!strcmp(IDVal, ".static_const"))
365 return ParseDirectiveSectionSwitch("__TEXT,__static_const");
366 if (!strcmp(IDVal, ".cstring"))
367 return ParseDirectiveSectionSwitch("__TEXT,__cstring",
368 "cstring_literals");
369 if (!strcmp(IDVal, ".literal4"))
370 return ParseDirectiveSectionSwitch("__TEXT,__literal4", "4byte_literals");
371 if (!strcmp(IDVal, ".literal8"))
372 return ParseDirectiveSectionSwitch("__TEXT,__literal8", "8byte_literals");
373 if (!strcmp(IDVal, ".literal16"))
374 return ParseDirectiveSectionSwitch("__TEXT,__literal16",
375 "16byte_literals");
376 if (!strcmp(IDVal, ".constructor"))
377 return ParseDirectiveSectionSwitch("__TEXT,__constructor");
378 if (!strcmp(IDVal, ".destructor"))
379 return ParseDirectiveSectionSwitch("__TEXT,__destructor");
380 if (!strcmp(IDVal, ".fvmlib_init0"))
381 return ParseDirectiveSectionSwitch("__TEXT,__fvmlib_init0");
382 if (!strcmp(IDVal, ".fvmlib_init1"))
383 return ParseDirectiveSectionSwitch("__TEXT,__fvmlib_init1");
384 if (!strcmp(IDVal, ".symbol_stub")) // FIXME: Different on PPC.
385 return ParseDirectiveSectionSwitch("__IMPORT,__jump_table,symbol_stubs",
386 "self_modifying_code+pure_instructions,5");
387 // FIXME: .picsymbol_stub on PPC.
388 if (!strcmp(IDVal, ".data"))
389 return ParseDirectiveSectionSwitch("__DATA,__data");
390 if (!strcmp(IDVal, ".static_data"))
391 return ParseDirectiveSectionSwitch("__DATA,__static_data");
392 if (!strcmp(IDVal, ".non_lazy_symbol_pointer"))
393 return ParseDirectiveSectionSwitch("__DATA,__nl_symbol_pointer",
394 "non_lazy_symbol_pointers");
395 if (!strcmp(IDVal, ".lazy_symbol_pointer"))
396 return ParseDirectiveSectionSwitch("__DATA,__la_symbol_pointer",
397 "lazy_symbol_pointers");
398 if (!strcmp(IDVal, ".dyld"))
399 return ParseDirectiveSectionSwitch("__DATA,__dyld");
400 if (!strcmp(IDVal, ".mod_init_func"))
401 return ParseDirectiveSectionSwitch("__DATA,__mod_init_func",
402 "mod_init_funcs");
403 if (!strcmp(IDVal, ".mod_term_func"))
404 return ParseDirectiveSectionSwitch("__DATA,__mod_term_func",
405 "mod_term_funcs");
406 if (!strcmp(IDVal, ".const_data"))
407 return ParseDirectiveSectionSwitch("__DATA,__const", "regular");
410 // FIXME: Verify attributes on sections.
411 if (!strcmp(IDVal, ".objc_class"))
412 return ParseDirectiveSectionSwitch("__OBJC,__class");
413 if (!strcmp(IDVal, ".objc_meta_class"))
414 return ParseDirectiveSectionSwitch("__OBJC,__meta_class");
415 if (!strcmp(IDVal, ".objc_cat_cls_meth"))
416 return ParseDirectiveSectionSwitch("__OBJC,__cat_cls_meth");
417 if (!strcmp(IDVal, ".objc_cat_inst_meth"))
418 return ParseDirectiveSectionSwitch("__OBJC,__cat_inst_meth");
419 if (!strcmp(IDVal, ".objc_protocol"))
420 return ParseDirectiveSectionSwitch("__OBJC,__protocol");
421 if (!strcmp(IDVal, ".objc_string_object"))
422 return ParseDirectiveSectionSwitch("__OBJC,__string_object");
423 if (!strcmp(IDVal, ".objc_cls_meth"))
424 return ParseDirectiveSectionSwitch("__OBJC,__cls_meth");
425 if (!strcmp(IDVal, ".objc_inst_meth"))
426 return ParseDirectiveSectionSwitch("__OBJC,__inst_meth");
427 if (!strcmp(IDVal, ".objc_cls_refs"))
428 return ParseDirectiveSectionSwitch("__OBJC,__cls_refs");
429 if (!strcmp(IDVal, ".objc_message_refs"))
430 return ParseDirectiveSectionSwitch("__OBJC,__message_refs");
431 if (!strcmp(IDVal, ".objc_symbols"))
432 return ParseDirectiveSectionSwitch("__OBJC,__symbols");
433 if (!strcmp(IDVal, ".objc_category"))
434 return ParseDirectiveSectionSwitch("__OBJC,__category");
435 if (!strcmp(IDVal, ".objc_class_vars"))
436 return ParseDirectiveSectionSwitch("__OBJC,__class_vars");
437 if (!strcmp(IDVal, ".objc_instance_vars"))
438 return ParseDirectiveSectionSwitch("__OBJC,__instance_vars");
439 if (!strcmp(IDVal, ".objc_module_info"))
440 return ParseDirectiveSectionSwitch("__OBJC,__module_info");
441 if (!strcmp(IDVal, ".objc_class_names"))
442 return ParseDirectiveSectionSwitch("__TEXT,__cstring","cstring_literals");
443 if (!strcmp(IDVal, ".objc_meth_var_types"))
444 return ParseDirectiveSectionSwitch("__TEXT,__cstring","cstring_literals");
445 if (!strcmp(IDVal, ".objc_meth_var_names"))
446 return ParseDirectiveSectionSwitch("__TEXT,__cstring","cstring_literals");
447 if (!strcmp(IDVal, ".objc_selector_strs"))
448 return ParseDirectiveSectionSwitch("__OBJC,__selector_strs");
450 // Assembler features
451 if (!strcmp(IDVal, ".set"))
452 return ParseDirectiveSet();
454 // Data directives
456 if (!strcmp(IDVal, ".ascii"))
457 return ParseDirectiveAscii(false);
458 if (!strcmp(IDVal, ".asciz"))
459 return ParseDirectiveAscii(true);
461 // FIXME: Target hooks for size? Also for "word", "hword".
462 if (!strcmp(IDVal, ".byte"))
463 return ParseDirectiveValue(1);
464 if (!strcmp(IDVal, ".short"))
465 return ParseDirectiveValue(2);
466 if (!strcmp(IDVal, ".long"))
467 return ParseDirectiveValue(4);
468 if (!strcmp(IDVal, ".quad"))
469 return ParseDirectiveValue(8);
471 // FIXME: Target hooks for IsPow2.
472 if (!strcmp(IDVal, ".align"))
473 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
474 if (!strcmp(IDVal, ".align32"))
475 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
476 if (!strcmp(IDVal, ".balign"))
477 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/1);
478 if (!strcmp(IDVal, ".balignw"))
479 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/2);
480 if (!strcmp(IDVal, ".balignl"))
481 return ParseDirectiveAlign(/*IsPow2=*/false, /*ExprSize=*/4);
482 if (!strcmp(IDVal, ".p2align"))
483 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/1);
484 if (!strcmp(IDVal, ".p2alignw"))
485 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/2);
486 if (!strcmp(IDVal, ".p2alignl"))
487 return ParseDirectiveAlign(/*IsPow2=*/true, /*ExprSize=*/4);
489 if (!strcmp(IDVal, ".org"))
490 return ParseDirectiveOrg();
492 if (!strcmp(IDVal, ".fill"))
493 return ParseDirectiveFill();
494 if (!strcmp(IDVal, ".space"))
495 return ParseDirectiveSpace();
497 // Symbol attribute directives
498 if (!strcmp(IDVal, ".globl") || !strcmp(IDVal, ".global"))
499 return ParseDirectiveSymbolAttribute(MCStreamer::Global);
500 if (!strcmp(IDVal, ".hidden"))
501 return ParseDirectiveSymbolAttribute(MCStreamer::Hidden);
502 if (!strcmp(IDVal, ".indirect_symbol"))
503 return ParseDirectiveSymbolAttribute(MCStreamer::IndirectSymbol);
504 if (!strcmp(IDVal, ".internal"))
505 return ParseDirectiveSymbolAttribute(MCStreamer::Internal);
506 if (!strcmp(IDVal, ".lazy_reference"))
507 return ParseDirectiveSymbolAttribute(MCStreamer::LazyReference);
508 if (!strcmp(IDVal, ".no_dead_strip"))
509 return ParseDirectiveSymbolAttribute(MCStreamer::NoDeadStrip);
510 if (!strcmp(IDVal, ".private_extern"))
511 return ParseDirectiveSymbolAttribute(MCStreamer::PrivateExtern);
512 if (!strcmp(IDVal, ".protected"))
513 return ParseDirectiveSymbolAttribute(MCStreamer::Protected);
514 if (!strcmp(IDVal, ".reference"))
515 return ParseDirectiveSymbolAttribute(MCStreamer::Reference);
516 if (!strcmp(IDVal, ".weak"))
517 return ParseDirectiveSymbolAttribute(MCStreamer::Weak);
518 if (!strcmp(IDVal, ".weak_definition"))
519 return ParseDirectiveSymbolAttribute(MCStreamer::WeakDefinition);
520 if (!strcmp(IDVal, ".weak_reference"))
521 return ParseDirectiveSymbolAttribute(MCStreamer::WeakReference);
523 if (!strcmp(IDVal, ".comm"))
524 return ParseDirectiveComm(/*IsLocal=*/false);
525 if (!strcmp(IDVal, ".lcomm"))
526 return ParseDirectiveComm(/*IsLocal=*/true);
527 if (!strcmp(IDVal, ".zerofill"))
528 return ParseDirectiveDarwinZerofill();
529 if (!strcmp(IDVal, ".desc"))
530 return ParseDirectiveDarwinSymbolDesc();
531 if (!strcmp(IDVal, ".lsym"))
532 return ParseDirectiveDarwinLsym();
534 if (!strcmp(IDVal, ".subsections_via_symbols"))
535 return ParseDirectiveDarwinSubsectionsViaSymbols();
536 if (!strcmp(IDVal, ".abort"))
537 return ParseDirectiveAbort();
538 if (!strcmp(IDVal, ".include"))
539 return ParseDirectiveInclude();
540 if (!strcmp(IDVal, ".dump"))
541 return ParseDirectiveDarwinDumpOrLoad(/*IsDump=*/true);
542 if (!strcmp(IDVal, ".load"))
543 return ParseDirectiveDarwinDumpOrLoad(/*IsLoad=*/false);
545 Warning(IDLoc, "ignoring directive for now");
546 EatToEndOfStatement();
547 return false;
550 MCInst Inst;
551 if (ParseX86InstOperands(IDVal, Inst))
552 return true;
554 if (Lexer.isNot(asmtok::EndOfStatement))
555 return TokError("unexpected token in argument list");
557 // Eat the end of statement marker.
558 Lexer.Lex();
560 // Instruction is good, process it.
561 Out.EmitInstruction(Inst);
563 // Skip to end of line for now.
564 return false;
567 bool AsmParser::ParseAssignment(const char *Name, bool IsDotSet) {
568 // FIXME: Use better location, we should use proper tokens.
569 SMLoc EqualLoc = Lexer.getLoc();
571 MCValue Value;
572 if (ParseRelocatableExpression(Value))
573 return true;
575 if (Lexer.isNot(asmtok::EndOfStatement))
576 return TokError("unexpected token in assignment");
578 // Eat the end of statement marker.
579 Lexer.Lex();
581 // Diagnose assignment to a label.
583 // FIXME: Diagnostics. Note the location of the definition as a label.
584 // FIXME: This doesn't diagnose assignment to a symbol which has been
585 // implicitly marked as external.
586 // FIXME: Handle '.'.
587 // FIXME: Diagnose assignment to protected identifier (e.g., register name).
588 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name);
589 if (Sym->getSection())
590 return Error(EqualLoc, "invalid assignment to symbol emitted as a label");
591 if (Sym->isExternal())
592 return Error(EqualLoc, "invalid assignment to external symbol");
594 // Do the assignment.
595 Out.EmitAssignment(Sym, Value, IsDotSet);
597 return false;
600 /// ParseDirectiveSet:
601 /// ::= .set identifier ',' expression
602 bool AsmParser::ParseDirectiveSet() {
603 if (Lexer.isNot(asmtok::Identifier))
604 return TokError("expected identifier after '.set' directive");
606 const char *Name = Lexer.getCurStrVal();
608 if (Lexer.Lex() != asmtok::Comma)
609 return TokError("unexpected token in '.set'");
610 Lexer.Lex();
612 return ParseAssignment(Name, true);
615 /// ParseDirectiveSection:
616 /// ::= .section identifier (',' identifier)*
617 /// FIXME: This should actually parse out the segment, section, attributes and
618 /// sizeof_stub fields.
619 bool AsmParser::ParseDirectiveDarwinSection() {
620 if (Lexer.isNot(asmtok::Identifier))
621 return TokError("expected identifier after '.section' directive");
623 std::string Section = Lexer.getCurStrVal();
624 Lexer.Lex();
626 // Accept a comma separated list of modifiers.
627 while (Lexer.is(asmtok::Comma)) {
628 Lexer.Lex();
630 if (Lexer.isNot(asmtok::Identifier))
631 return TokError("expected identifier in '.section' directive");
632 Section += ',';
633 Section += Lexer.getCurStrVal();
634 Lexer.Lex();
637 if (Lexer.isNot(asmtok::EndOfStatement))
638 return TokError("unexpected token in '.section' directive");
639 Lexer.Lex();
641 Out.SwitchSection(Ctx.GetSection(Section.c_str()));
642 return false;
645 bool AsmParser::ParseDirectiveSectionSwitch(const char *Section,
646 const char *Directives) {
647 if (Lexer.isNot(asmtok::EndOfStatement))
648 return TokError("unexpected token in section switching directive");
649 Lexer.Lex();
651 std::string SectionStr = Section;
652 if (Directives && Directives[0]) {
653 SectionStr += ",";
654 SectionStr += Directives;
657 Out.SwitchSection(Ctx.GetSection(Section));
658 return false;
661 /// ParseDirectiveAscii:
662 /// ::= ( .ascii | .asciz ) [ "string" ( , "string" )* ]
663 bool AsmParser::ParseDirectiveAscii(bool ZeroTerminated) {
664 if (Lexer.isNot(asmtok::EndOfStatement)) {
665 for (;;) {
666 if (Lexer.isNot(asmtok::String))
667 return TokError("expected string in '.ascii' or '.asciz' directive");
669 // FIXME: This shouldn't use a const char* + strlen, the string could have
670 // embedded nulls.
671 // FIXME: Should have accessor for getting string contents.
672 const char *Str = Lexer.getCurStrVal();
673 Out.EmitBytes(Str + 1, strlen(Str) - 2);
674 if (ZeroTerminated)
675 Out.EmitBytes("\0", 1);
677 Lexer.Lex();
679 if (Lexer.is(asmtok::EndOfStatement))
680 break;
682 if (Lexer.isNot(asmtok::Comma))
683 return TokError("unexpected token in '.ascii' or '.asciz' directive");
684 Lexer.Lex();
688 Lexer.Lex();
689 return false;
692 /// ParseDirectiveValue
693 /// ::= (.byte | .short | ... ) [ expression (, expression)* ]
694 bool AsmParser::ParseDirectiveValue(unsigned Size) {
695 if (Lexer.isNot(asmtok::EndOfStatement)) {
696 for (;;) {
697 MCValue Expr;
698 if (ParseRelocatableExpression(Expr))
699 return true;
701 Out.EmitValue(Expr, Size);
703 if (Lexer.is(asmtok::EndOfStatement))
704 break;
706 // FIXME: Improve diagnostic.
707 if (Lexer.isNot(asmtok::Comma))
708 return TokError("unexpected token in directive");
709 Lexer.Lex();
713 Lexer.Lex();
714 return false;
717 /// ParseDirectiveSpace
718 /// ::= .space expression [ , expression ]
719 bool AsmParser::ParseDirectiveSpace() {
720 int64_t NumBytes;
721 if (ParseAbsoluteExpression(NumBytes))
722 return true;
724 int64_t FillExpr = 0;
725 bool HasFillExpr = false;
726 if (Lexer.isNot(asmtok::EndOfStatement)) {
727 if (Lexer.isNot(asmtok::Comma))
728 return TokError("unexpected token in '.space' directive");
729 Lexer.Lex();
731 if (ParseAbsoluteExpression(FillExpr))
732 return true;
734 HasFillExpr = true;
736 if (Lexer.isNot(asmtok::EndOfStatement))
737 return TokError("unexpected token in '.space' directive");
740 Lexer.Lex();
742 if (NumBytes <= 0)
743 return TokError("invalid number of bytes in '.space' directive");
745 // FIXME: Sometimes the fill expr is 'nop' if it isn't supplied, instead of 0.
746 for (uint64_t i = 0, e = NumBytes; i != e; ++i)
747 Out.EmitValue(MCValue::get(FillExpr), 1);
749 return false;
752 /// ParseDirectiveFill
753 /// ::= .fill expression , expression , expression
754 bool AsmParser::ParseDirectiveFill() {
755 int64_t NumValues;
756 if (ParseAbsoluteExpression(NumValues))
757 return true;
759 if (Lexer.isNot(asmtok::Comma))
760 return TokError("unexpected token in '.fill' directive");
761 Lexer.Lex();
763 int64_t FillSize;
764 if (ParseAbsoluteExpression(FillSize))
765 return true;
767 if (Lexer.isNot(asmtok::Comma))
768 return TokError("unexpected token in '.fill' directive");
769 Lexer.Lex();
771 int64_t FillExpr;
772 if (ParseAbsoluteExpression(FillExpr))
773 return true;
775 if (Lexer.isNot(asmtok::EndOfStatement))
776 return TokError("unexpected token in '.fill' directive");
778 Lexer.Lex();
780 if (FillSize != 1 && FillSize != 2 && FillSize != 4)
781 return TokError("invalid '.fill' size, expected 1, 2, or 4");
783 for (uint64_t i = 0, e = NumValues; i != e; ++i)
784 Out.EmitValue(MCValue::get(FillExpr), FillSize);
786 return false;
789 /// ParseDirectiveOrg
790 /// ::= .org expression [ , expression ]
791 bool AsmParser::ParseDirectiveOrg() {
792 MCValue Offset;
793 if (ParseRelocatableExpression(Offset))
794 return true;
796 // Parse optional fill expression.
797 int64_t FillExpr = 0;
798 if (Lexer.isNot(asmtok::EndOfStatement)) {
799 if (Lexer.isNot(asmtok::Comma))
800 return TokError("unexpected token in '.org' directive");
801 Lexer.Lex();
803 if (ParseAbsoluteExpression(FillExpr))
804 return true;
806 if (Lexer.isNot(asmtok::EndOfStatement))
807 return TokError("unexpected token in '.org' directive");
810 Lexer.Lex();
812 // FIXME: Only limited forms of relocatable expressions are accepted here, it
813 // has to be relative to the current section.
814 Out.EmitValueToOffset(Offset, FillExpr);
816 return false;
819 /// ParseDirectiveAlign
820 /// ::= {.align, ...} expression [ , expression [ , expression ]]
821 bool AsmParser::ParseDirectiveAlign(bool IsPow2, unsigned ValueSize) {
822 int64_t Alignment;
823 if (ParseAbsoluteExpression(Alignment))
824 return true;
826 SMLoc MaxBytesLoc;
827 bool HasFillExpr = false;
828 int64_t FillExpr = 0;
829 int64_t MaxBytesToFill = 0;
830 if (Lexer.isNot(asmtok::EndOfStatement)) {
831 if (Lexer.isNot(asmtok::Comma))
832 return TokError("unexpected token in directive");
833 Lexer.Lex();
835 // The fill expression can be omitted while specifying a maximum number of
836 // alignment bytes, e.g:
837 // .align 3,,4
838 if (Lexer.isNot(asmtok::Comma)) {
839 HasFillExpr = true;
840 if (ParseAbsoluteExpression(FillExpr))
841 return true;
844 if (Lexer.isNot(asmtok::EndOfStatement)) {
845 if (Lexer.isNot(asmtok::Comma))
846 return TokError("unexpected token in directive");
847 Lexer.Lex();
849 MaxBytesLoc = Lexer.getLoc();
850 if (ParseAbsoluteExpression(MaxBytesToFill))
851 return true;
853 if (Lexer.isNot(asmtok::EndOfStatement))
854 return TokError("unexpected token in directive");
858 Lexer.Lex();
860 if (!HasFillExpr) {
861 // FIXME: Sometimes fill with nop.
862 FillExpr = 0;
865 // Compute alignment in bytes.
866 if (IsPow2) {
867 // FIXME: Diagnose overflow.
868 Alignment = 1LL << Alignment;
871 // Diagnose non-sensical max bytes to fill.
872 if (MaxBytesLoc.isValid()) {
873 if (MaxBytesToFill < 1) {
874 Warning(MaxBytesLoc, "alignment directive can never be satisfied in this "
875 "many bytes, ignoring");
876 return false;
879 if (MaxBytesToFill >= Alignment) {
880 Warning(MaxBytesLoc, "maximum bytes expression exceeds alignment and "
881 "has no effect");
882 MaxBytesToFill = 0;
886 // FIXME: Target specific behavior about how the "extra" bytes are filled.
887 Out.EmitValueToAlignment(Alignment, FillExpr, ValueSize, MaxBytesToFill);
889 return false;
892 /// ParseDirectiveSymbolAttribute
893 /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ]
894 bool AsmParser::ParseDirectiveSymbolAttribute(MCStreamer::SymbolAttr Attr) {
895 if (Lexer.isNot(asmtok::EndOfStatement)) {
896 for (;;) {
897 if (Lexer.isNot(asmtok::Identifier))
898 return TokError("expected identifier in directive");
900 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
901 Lexer.Lex();
903 // If this is use of an undefined symbol then mark it external.
904 if (!Sym->getSection() && !Ctx.GetSymbolValue(Sym))
905 Sym->setExternal(true);
907 Out.EmitSymbolAttribute(Sym, Attr);
909 if (Lexer.is(asmtok::EndOfStatement))
910 break;
912 if (Lexer.isNot(asmtok::Comma))
913 return TokError("unexpected token in directive");
914 Lexer.Lex();
918 Lexer.Lex();
919 return false;
922 /// ParseDirectiveDarwinSymbolDesc
923 /// ::= .desc identifier , expression
924 bool AsmParser::ParseDirectiveDarwinSymbolDesc() {
925 if (Lexer.isNot(asmtok::Identifier))
926 return TokError("expected identifier in directive");
928 // handle the identifier as the key symbol.
929 SMLoc IDLoc = Lexer.getLoc();
930 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
931 Lexer.Lex();
933 if (Lexer.isNot(asmtok::Comma))
934 return TokError("unexpected token in '.desc' directive");
935 Lexer.Lex();
937 SMLoc DescLoc = Lexer.getLoc();
938 int64_t DescValue;
939 if (ParseAbsoluteExpression(DescValue))
940 return true;
942 if (Lexer.isNot(asmtok::EndOfStatement))
943 return TokError("unexpected token in '.desc' directive");
945 Lexer.Lex();
947 // Set the n_desc field of this Symbol to this DescValue
948 Out.EmitSymbolDesc(Sym, DescValue);
950 return false;
953 /// ParseDirectiveComm
954 /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ]
955 bool AsmParser::ParseDirectiveComm(bool IsLocal) {
956 if (Lexer.isNot(asmtok::Identifier))
957 return TokError("expected identifier in directive");
959 // handle the identifier as the key symbol.
960 SMLoc IDLoc = Lexer.getLoc();
961 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
962 Lexer.Lex();
964 if (Lexer.isNot(asmtok::Comma))
965 return TokError("unexpected token in directive");
966 Lexer.Lex();
968 int64_t Size;
969 SMLoc SizeLoc = Lexer.getLoc();
970 if (ParseAbsoluteExpression(Size))
971 return true;
973 int64_t Pow2Alignment = 0;
974 SMLoc Pow2AlignmentLoc;
975 if (Lexer.is(asmtok::Comma)) {
976 Lexer.Lex();
977 Pow2AlignmentLoc = Lexer.getLoc();
978 if (ParseAbsoluteExpression(Pow2Alignment))
979 return true;
982 if (Lexer.isNot(asmtok::EndOfStatement))
983 return TokError("unexpected token in '.comm' or '.lcomm' directive");
985 Lexer.Lex();
987 // NOTE: a size of zero for a .comm should create a undefined symbol
988 // but a size of .lcomm creates a bss symbol of size zero.
989 if (Size < 0)
990 return Error(SizeLoc, "invalid '.comm' or '.lcomm' directive size, can't "
991 "be less than zero");
993 // NOTE: The alignment in the directive is a power of 2 value, the assember
994 // may internally end up wanting an alignment in bytes.
995 // FIXME: Diagnose overflow.
996 if (Pow2Alignment < 0)
997 return Error(Pow2AlignmentLoc, "invalid '.comm' or '.lcomm' directive "
998 "alignment, can't be less than zero");
1000 // TODO: Symbol must be undefined or it is a error to re-defined the symbol
1001 if (Sym->getSection() || Ctx.GetSymbolValue(Sym))
1002 return Error(IDLoc, "invalid symbol redefinition");
1004 // Create the Symbol as a common or local common with Size and Pow2Alignment
1005 Out.EmitCommonSymbol(Sym, Size, Pow2Alignment, IsLocal);
1007 return false;
1010 /// ParseDirectiveDarwinZerofill
1011 /// ::= .zerofill segname , sectname [, identifier , size_expression [
1012 /// , align_expression ]]
1013 bool AsmParser::ParseDirectiveDarwinZerofill() {
1014 if (Lexer.isNot(asmtok::Identifier))
1015 return TokError("expected segment name after '.zerofill' directive");
1016 std::string Section = Lexer.getCurStrVal();
1017 Lexer.Lex();
1019 if (Lexer.isNot(asmtok::Comma))
1020 return TokError("unexpected token in directive");
1021 Section += ',';
1022 Lexer.Lex();
1024 if (Lexer.isNot(asmtok::Identifier))
1025 return TokError("expected section name after comma in '.zerofill' "
1026 "directive");
1027 Section += Lexer.getCurStrVal();
1028 Lexer.Lex();
1030 // FIXME: we will need to tell GetSection() that this is to be created with or
1031 // must have the Mach-O section type of S_ZEROFILL. Something like the code
1032 // below could be done but for now it is not as EmitZerofill() does not know
1033 // how to deal with a section type in the section name like
1034 // ParseDirectiveDarwinSection() allows.
1035 // Section += ',';
1036 // Section += "zerofill";
1038 // If this is the end of the line all that was wanted was to create the
1039 // the section but with no symbol.
1040 if (Lexer.is(asmtok::EndOfStatement)) {
1041 // Create the zerofill section but no symbol
1042 Out.EmitZerofill(Ctx.GetSection(Section.c_str()));
1043 return false;
1046 if (Lexer.isNot(asmtok::Comma))
1047 return TokError("unexpected token in directive");
1048 Lexer.Lex();
1050 if (Lexer.isNot(asmtok::Identifier))
1051 return TokError("expected identifier in directive");
1053 // handle the identifier as the key symbol.
1054 SMLoc IDLoc = Lexer.getLoc();
1055 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
1056 Lexer.Lex();
1058 if (Lexer.isNot(asmtok::Comma))
1059 return TokError("unexpected token in directive");
1060 Lexer.Lex();
1062 int64_t Size;
1063 SMLoc SizeLoc = Lexer.getLoc();
1064 if (ParseAbsoluteExpression(Size))
1065 return true;
1067 int64_t Pow2Alignment = 0;
1068 SMLoc Pow2AlignmentLoc;
1069 if (Lexer.is(asmtok::Comma)) {
1070 Lexer.Lex();
1071 Pow2AlignmentLoc = Lexer.getLoc();
1072 if (ParseAbsoluteExpression(Pow2Alignment))
1073 return true;
1076 if (Lexer.isNot(asmtok::EndOfStatement))
1077 return TokError("unexpected token in '.zerofill' directive");
1079 Lexer.Lex();
1081 if (Size < 0)
1082 return Error(SizeLoc, "invalid '.zerofill' directive size, can't be less "
1083 "than zero");
1085 // NOTE: The alignment in the directive is a power of 2 value, the assember
1086 // may internally end up wanting an alignment in bytes.
1087 // FIXME: Diagnose overflow.
1088 if (Pow2Alignment < 0)
1089 return Error(Pow2AlignmentLoc, "invalid '.zerofill' directive alignment, "
1090 "can't be less than zero");
1092 // TODO: Symbol must be undefined or it is a error to re-defined the symbol
1093 if (Sym->getSection() || Ctx.GetSymbolValue(Sym))
1094 return Error(IDLoc, "invalid symbol redefinition");
1096 // Create the zerofill Symbol with Size and Pow2Alignment
1097 Out.EmitZerofill(Ctx.GetSection(Section.c_str()), Sym, Size, Pow2Alignment);
1099 return false;
1102 /// ParseDirectiveDarwinSubsectionsViaSymbols
1103 /// ::= .subsections_via_symbols
1104 bool AsmParser::ParseDirectiveDarwinSubsectionsViaSymbols() {
1105 if (Lexer.isNot(asmtok::EndOfStatement))
1106 return TokError("unexpected token in '.subsections_via_symbols' directive");
1108 Lexer.Lex();
1110 Out.SubsectionsViaSymbols();
1112 return false;
1115 /// ParseDirectiveAbort
1116 /// ::= .abort [ "abort_string" ]
1117 bool AsmParser::ParseDirectiveAbort() {
1118 const char *Str = NULL;
1119 if (Lexer.isNot(asmtok::EndOfStatement)) {
1120 if (Lexer.isNot(asmtok::String))
1121 return TokError("expected string in '.abort' directive");
1123 Str = Lexer.getCurStrVal();
1125 Lexer.Lex();
1128 if (Lexer.isNot(asmtok::EndOfStatement))
1129 return TokError("unexpected token in '.abort' directive");
1131 Lexer.Lex();
1133 Out.AbortAssembly(Str);
1135 return false;
1138 /// ParseDirectiveLsym
1139 /// ::= .lsym identifier , expression
1140 bool AsmParser::ParseDirectiveDarwinLsym() {
1141 if (Lexer.isNot(asmtok::Identifier))
1142 return TokError("expected identifier in directive");
1144 // handle the identifier as the key symbol.
1145 SMLoc IDLoc = Lexer.getLoc();
1146 MCSymbol *Sym = Ctx.GetOrCreateSymbol(Lexer.getCurStrVal());
1147 Lexer.Lex();
1149 if (Lexer.isNot(asmtok::Comma))
1150 return TokError("unexpected token in '.lsym' directive");
1151 Lexer.Lex();
1153 MCValue Expr;
1154 if (ParseRelocatableExpression(Expr))
1155 return true;
1157 if (Lexer.isNot(asmtok::EndOfStatement))
1158 return TokError("unexpected token in '.lsym' directive");
1160 Lexer.Lex();
1162 // Create the Sym with the value of the Expr
1163 Out.EmitLocalSymbol(Sym, Expr);
1165 return false;
1168 /// ParseDirectiveInclude
1169 /// ::= .include "filename"
1170 bool AsmParser::ParseDirectiveInclude() {
1171 if (Lexer.isNot(asmtok::String))
1172 return TokError("expected string in '.include' directive");
1174 std::string Filename = Lexer.getCurStrVal();
1175 SMLoc IncludeLoc = Lexer.getLoc();
1176 Lexer.Lex();
1178 if (Lexer.isNot(asmtok::EndOfStatement))
1179 return TokError("unexpected token in '.include' directive");
1181 // Strip the quotes.
1182 Filename = Filename.substr(1, Filename.size()-2);
1184 // Attempt to switch the lexer to the included file before consuming the end
1185 // of statement to avoid losing it when we switch.
1186 if (Lexer.EnterIncludeFile(Filename)) {
1187 Lexer.PrintMessage(IncludeLoc,
1188 "Could not find include file '" + Filename + "'",
1189 "error");
1190 return true;
1193 return false;
1196 /// ParseDirectiveDarwinDumpOrLoad
1197 /// ::= ( .dump | .load ) "filename"
1198 bool AsmParser::ParseDirectiveDarwinDumpOrLoad(bool IsDump) {
1199 const char *Str;
1201 if (Lexer.isNot(asmtok::String))
1202 return TokError("expected string in '.dump' or '.load' directive");
1204 Str = Lexer.getCurStrVal();
1206 Lexer.Lex();
1208 if (Lexer.isNot(asmtok::EndOfStatement))
1209 return TokError("unexpected token in '.dump' or '.load' directive");
1211 Lexer.Lex();
1213 if (IsDump)
1214 Out.DumpSymbolsandMacros(Str);
1215 else
1216 Out.LoadSymbolsandMacros(Str);
1218 return false;