LLVM 18.0.0git
MasmParser.cpp
Go to the documentation of this file.
1//===- AsmParser.cpp - Parser for Assembly Files --------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This class implements the parser for assembly files.
10//
11//===----------------------------------------------------------------------===//
12
13#include "llvm/ADT/APFloat.h"
14#include "llvm/ADT/APInt.h"
15#include "llvm/ADT/ArrayRef.h"
16#include "llvm/ADT/BitVector.h"
17#include "llvm/ADT/STLExtras.h"
21#include "llvm/ADT/StringMap.h"
22#include "llvm/ADT/StringRef.h"
24#include "llvm/ADT/Twine.h"
27#include "llvm/MC/MCAsmInfo.h"
28#include "llvm/MC/MCCodeView.h"
29#include "llvm/MC/MCContext.h"
31#include "llvm/MC/MCDwarf.h"
32#include "llvm/MC/MCExpr.h"
34#include "llvm/MC/MCInstrDesc.h"
35#include "llvm/MC/MCInstrInfo.h"
44#include "llvm/MC/MCSection.h"
45#include "llvm/MC/MCStreamer.h"
47#include "llvm/MC/MCSymbol.h"
52#include "llvm/Support/Format.h"
53#include "llvm/Support/MD5.h"
56#include "llvm/Support/Path.h"
57#include "llvm/Support/SMLoc.h"
60#include <algorithm>
61#include <cassert>
62#include <climits>
63#include <cstddef>
64#include <cstdint>
65#include <ctime>
66#include <deque>
67#include <memory>
68#include <optional>
69#include <sstream>
70#include <string>
71#include <tuple>
72#include <utility>
73#include <vector>
74
75using namespace llvm;
76
77namespace {
78
79/// Helper types for tracking macro definitions.
80typedef std::vector<AsmToken> MCAsmMacroArgument;
81typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments;
82
83/// Helper class for storing information about an active macro instantiation.
84struct MacroInstantiation {
85 /// The location of the instantiation.
86 SMLoc InstantiationLoc;
87
88 /// The buffer where parsing should resume upon instantiation completion.
89 unsigned ExitBuffer;
90
91 /// The location where parsing should resume upon instantiation completion.
92 SMLoc ExitLoc;
93
94 /// The depth of TheCondStack at the start of the instantiation.
95 size_t CondStackDepth;
96};
97
98struct ParseStatementInfo {
99 /// The parsed operands from the last parsed statement.
101
102 /// The opcode from the last parsed instruction.
103 unsigned Opcode = ~0U;
104
105 /// Was there an error parsing the inline assembly?
106 bool ParseError = false;
107
108 /// The value associated with a macro exit.
109 std::optional<std::string> ExitValue;
110
111 SmallVectorImpl<AsmRewrite> *AsmRewrites = nullptr;
112
113 ParseStatementInfo() = delete;
114 ParseStatementInfo(SmallVectorImpl<AsmRewrite> *rewrites)
115 : AsmRewrites(rewrites) {}
116};
117
118enum FieldType {
119 FT_INTEGRAL, // Initializer: integer expression, stored as an MCExpr.
120 FT_REAL, // Initializer: real number, stored as an APInt.
121 FT_STRUCT // Initializer: struct initializer, stored recursively.
122};
123
124struct FieldInfo;
125struct StructInfo {
127 bool IsUnion = false;
128 bool Initializable = true;
129 unsigned Alignment = 0;
130 unsigned AlignmentSize = 0;
131 unsigned NextOffset = 0;
132 unsigned Size = 0;
133 std::vector<FieldInfo> Fields;
134 StringMap<size_t> FieldsByName;
135
136 FieldInfo &addField(StringRef FieldName, FieldType FT,
137 unsigned FieldAlignmentSize);
138
139 StructInfo() = default;
140 StructInfo(StringRef StructName, bool Union, unsigned AlignmentValue);
141};
142
143// FIXME: This should probably use a class hierarchy, raw pointers between the
144// objects, and dynamic type resolution instead of a union. On the other hand,
145// ownership then becomes much more complicated; the obvious thing would be to
146// use BumpPtrAllocator, but the lack of a destructor makes that messy.
147
148struct StructInitializer;
149struct IntFieldInfo {
151
152 IntFieldInfo() = default;
153 IntFieldInfo(const SmallVector<const MCExpr *, 1> &V) { Values = V; }
154 IntFieldInfo(SmallVector<const MCExpr *, 1> &&V) { Values = std::move(V); }
155};
156struct RealFieldInfo {
157 SmallVector<APInt, 1> AsIntValues;
158
159 RealFieldInfo() = default;
160 RealFieldInfo(const SmallVector<APInt, 1> &V) { AsIntValues = V; }
161 RealFieldInfo(SmallVector<APInt, 1> &&V) { AsIntValues = std::move(V); }
162};
163struct StructFieldInfo {
164 std::vector<StructInitializer> Initializers;
165 StructInfo Structure;
166
167 StructFieldInfo() = default;
168 StructFieldInfo(std::vector<StructInitializer> V, StructInfo S);
169};
170
171class FieldInitializer {
172public:
173 FieldType FT;
174 union {
175 IntFieldInfo IntInfo;
176 RealFieldInfo RealInfo;
177 StructFieldInfo StructInfo;
178 };
179
180 ~FieldInitializer();
181 FieldInitializer(FieldType FT);
182
183 FieldInitializer(SmallVector<const MCExpr *, 1> &&Values);
184 FieldInitializer(SmallVector<APInt, 1> &&AsIntValues);
185 FieldInitializer(std::vector<StructInitializer> &&Initializers,
186 struct StructInfo Structure);
187
188 FieldInitializer(const FieldInitializer &Initializer);
189 FieldInitializer(FieldInitializer &&Initializer);
190
191 FieldInitializer &operator=(const FieldInitializer &Initializer);
192 FieldInitializer &operator=(FieldInitializer &&Initializer);
193};
194
195struct StructInitializer {
196 std::vector<FieldInitializer> FieldInitializers;
197};
198
199struct FieldInfo {
200 // Offset of the field within the containing STRUCT.
201 unsigned Offset = 0;
202
203 // Total size of the field (= LengthOf * Type).
204 unsigned SizeOf = 0;
205
206 // Number of elements in the field (1 if scalar, >1 if an array).
207 unsigned LengthOf = 0;
208
209 // Size of a single entry in this field, in bytes ("type" in MASM standards).
210 unsigned Type = 0;
211
212 FieldInitializer Contents;
213
214 FieldInfo(FieldType FT) : Contents(FT) {}
215};
216
217StructFieldInfo::StructFieldInfo(std::vector<StructInitializer> V,
218 StructInfo S) {
219 Initializers = std::move(V);
220 Structure = S;
221}
222
223StructInfo::StructInfo(StringRef StructName, bool Union,
224 unsigned AlignmentValue)
225 : Name(StructName), IsUnion(Union), Alignment(AlignmentValue) {}
226
227FieldInfo &StructInfo::addField(StringRef FieldName, FieldType FT,
228 unsigned FieldAlignmentSize) {
229 if (!FieldName.empty())
230 FieldsByName[FieldName.lower()] = Fields.size();
231 Fields.emplace_back(FT);
232 FieldInfo &Field = Fields.back();
233 Field.Offset =
234 llvm::alignTo(NextOffset, std::min(Alignment, FieldAlignmentSize));
235 if (!IsUnion) {
236 NextOffset = std::max(NextOffset, Field.Offset);
237 }
238 AlignmentSize = std::max(AlignmentSize, FieldAlignmentSize);
239 return Field;
240}
241
242FieldInitializer::~FieldInitializer() {
243 switch (FT) {
244 case FT_INTEGRAL:
245 IntInfo.~IntFieldInfo();
246 break;
247 case FT_REAL:
248 RealInfo.~RealFieldInfo();
249 break;
250 case FT_STRUCT:
251 StructInfo.~StructFieldInfo();
252 break;
253 }
254}
255
256FieldInitializer::FieldInitializer(FieldType FT) : FT(FT) {
257 switch (FT) {
258 case FT_INTEGRAL:
259 new (&IntInfo) IntFieldInfo();
260 break;
261 case FT_REAL:
262 new (&RealInfo) RealFieldInfo();
263 break;
264 case FT_STRUCT:
265 new (&StructInfo) StructFieldInfo();
266 break;
267 }
268}
269
270FieldInitializer::FieldInitializer(SmallVector<const MCExpr *, 1> &&Values)
271 : FT(FT_INTEGRAL) {
272 new (&IntInfo) IntFieldInfo(std::move(Values));
273}
274
275FieldInitializer::FieldInitializer(SmallVector<APInt, 1> &&AsIntValues)
276 : FT(FT_REAL) {
277 new (&RealInfo) RealFieldInfo(std::move(AsIntValues));
278}
279
280FieldInitializer::FieldInitializer(
281 std::vector<StructInitializer> &&Initializers, struct StructInfo Structure)
282 : FT(FT_STRUCT) {
283 new (&StructInfo) StructFieldInfo(std::move(Initializers), Structure);
284}
285
286FieldInitializer::FieldInitializer(const FieldInitializer &Initializer)
287 : FT(Initializer.FT) {
288 switch (FT) {
289 case FT_INTEGRAL:
290 new (&IntInfo) IntFieldInfo(Initializer.IntInfo);
291 break;
292 case FT_REAL:
293 new (&RealInfo) RealFieldInfo(Initializer.RealInfo);
294 break;
295 case FT_STRUCT:
296 new (&StructInfo) StructFieldInfo(Initializer.StructInfo);
297 break;
298 }
299}
300
301FieldInitializer::FieldInitializer(FieldInitializer &&Initializer)
302 : FT(Initializer.FT) {
303 switch (FT) {
304 case FT_INTEGRAL:
305 new (&IntInfo) IntFieldInfo(Initializer.IntInfo);
306 break;
307 case FT_REAL:
308 new (&RealInfo) RealFieldInfo(Initializer.RealInfo);
309 break;
310 case FT_STRUCT:
311 new (&StructInfo) StructFieldInfo(Initializer.StructInfo);
312 break;
313 }
314}
315
316FieldInitializer &
317FieldInitializer::operator=(const FieldInitializer &Initializer) {
318 if (FT != Initializer.FT) {
319 switch (FT) {
320 case FT_INTEGRAL:
321 IntInfo.~IntFieldInfo();
322 break;
323 case FT_REAL:
324 RealInfo.~RealFieldInfo();
325 break;
326 case FT_STRUCT:
327 StructInfo.~StructFieldInfo();
328 break;
329 }
330 }
331 FT = Initializer.FT;
332 switch (FT) {
333 case FT_INTEGRAL:
334 IntInfo = Initializer.IntInfo;
335 break;
336 case FT_REAL:
337 RealInfo = Initializer.RealInfo;
338 break;
339 case FT_STRUCT:
340 StructInfo = Initializer.StructInfo;
341 break;
342 }
343 return *this;
344}
345
346FieldInitializer &FieldInitializer::operator=(FieldInitializer &&Initializer) {
347 if (FT != Initializer.FT) {
348 switch (FT) {
349 case FT_INTEGRAL:
350 IntInfo.~IntFieldInfo();
351 break;
352 case FT_REAL:
353 RealInfo.~RealFieldInfo();
354 break;
355 case FT_STRUCT:
356 StructInfo.~StructFieldInfo();
357 break;
358 }
359 }
360 FT = Initializer.FT;
361 switch (FT) {
362 case FT_INTEGRAL:
363 IntInfo = Initializer.IntInfo;
364 break;
365 case FT_REAL:
366 RealInfo = Initializer.RealInfo;
367 break;
368 case FT_STRUCT:
369 StructInfo = Initializer.StructInfo;
370 break;
371 }
372 return *this;
373}
374
375/// The concrete assembly parser instance.
376// Note that this is a full MCAsmParser, not an MCAsmParserExtension!
377// It's a peer of AsmParser, not of COFFAsmParser, WasmAsmParser, etc.
378class MasmParser : public MCAsmParser {
379private:
380 AsmLexer Lexer;
381 MCContext &Ctx;
382 MCStreamer &Out;
383 const MCAsmInfo &MAI;
385 SourceMgr::DiagHandlerTy SavedDiagHandler;
386 void *SavedDiagContext;
387 std::unique_ptr<MCAsmParserExtension> PlatformParser;
388
389 /// This is the current buffer index we're lexing from as managed by the
390 /// SourceMgr object.
391 unsigned CurBuffer;
392
393 /// time of assembly
394 struct tm TM;
395
396 BitVector EndStatementAtEOFStack;
397
398 AsmCond TheCondState;
399 std::vector<AsmCond> TheCondStack;
400
401 /// maps directive names to handler methods in parser
402 /// extensions. Extensions register themselves in this map by calling
403 /// addDirectiveHandler.
404 StringMap<ExtensionDirectiveHandler> ExtensionDirectiveMap;
405
406 /// maps assembly-time variable names to variables.
407 struct Variable {
408 enum RedefinableKind { NOT_REDEFINABLE, WARN_ON_REDEFINITION, REDEFINABLE };
409
411 RedefinableKind Redefinable = REDEFINABLE;
412 bool IsText = false;
413 std::string TextValue;
414 };
415 StringMap<Variable> Variables;
416
417 /// Stack of active struct definitions.
418 SmallVector<StructInfo, 1> StructInProgress;
419
420 /// Maps struct tags to struct definitions.
421 StringMap<StructInfo> Structs;
422
423 /// Maps data location names to types.
424 StringMap<AsmTypeInfo> KnownType;
425
426 /// Stack of active macro instantiations.
427 std::vector<MacroInstantiation*> ActiveMacros;
428
429 /// List of bodies of anonymous macros.
430 std::deque<MCAsmMacro> MacroLikeBodies;
431
432 /// Keeps track of how many .macro's have been instantiated.
433 unsigned NumOfMacroInstantiations;
434
435 /// The values from the last parsed cpp hash file line comment if any.
436 struct CppHashInfoTy {
438 int64_t LineNumber;
439 SMLoc Loc;
440 unsigned Buf;
441 CppHashInfoTy() : LineNumber(0), Buf(0) {}
442 };
443 CppHashInfoTy CppHashInfo;
444
445 /// The filename from the first cpp hash file line comment, if any.
446 StringRef FirstCppHashFilename;
447
448 /// List of forward directional labels for diagnosis at the end.
450
451 /// AssemblerDialect. ~OU means unset value and use value provided by MAI.
452 /// Defaults to 1U, meaning Intel.
453 unsigned AssemblerDialect = 1U;
454
455 /// is Darwin compatibility enabled?
456 bool IsDarwin = false;
457
458 /// Are we parsing ms-style inline assembly?
459 bool ParsingMSInlineAsm = false;
460
461 /// Did we already inform the user about inconsistent MD5 usage?
462 bool ReportedInconsistentMD5 = false;
463
464 // Current <...> expression depth.
465 unsigned AngleBracketDepth = 0U;
466
467 // Number of locals defined.
468 uint16_t LocalCounter = 0;
469
470public:
471 MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
472 const MCAsmInfo &MAI, struct tm TM, unsigned CB = 0);
473 MasmParser(const MasmParser &) = delete;
474 MasmParser &operator=(const MasmParser &) = delete;
475 ~MasmParser() override;
476
477 bool Run(bool NoInitialTextSection, bool NoFinalize = false) override;
478
480 ExtensionDirectiveHandler Handler) override {
481 ExtensionDirectiveMap[Directive] = Handler;
482 if (!DirectiveKindMap.contains(Directive)) {
483 DirectiveKindMap[Directive] = DK_HANDLER_DIRECTIVE;
484 }
485 }
486
487 void addAliasForDirective(StringRef Directive, StringRef Alias) override {
488 DirectiveKindMap[Directive] = DirectiveKindMap[Alias];
489 }
490
491 /// @name MCAsmParser Interface
492 /// {
493
494 SourceMgr &getSourceManager() override { return SrcMgr; }
495 MCAsmLexer &getLexer() override { return Lexer; }
496 MCContext &getContext() override { return Ctx; }
497 MCStreamer &getStreamer() override { return Out; }
498
499 CodeViewContext &getCVContext() { return Ctx.getCVContext(); }
500
501 unsigned getAssemblerDialect() override {
502 if (AssemblerDialect == ~0U)
503 return MAI.getAssemblerDialect();
504 else
505 return AssemblerDialect;
506 }
507 void setAssemblerDialect(unsigned i) override {
508 AssemblerDialect = i;
509 }
510
511 void Note(SMLoc L, const Twine &Msg, SMRange Range = std::nullopt) override;
512 bool Warning(SMLoc L, const Twine &Msg,
513 SMRange Range = std::nullopt) override;
514 bool printError(SMLoc L, const Twine &Msg,
515 SMRange Range = std::nullopt) override;
516
517 enum ExpandKind { ExpandMacros, DoNotExpandMacros };
518 const AsmToken &Lex(ExpandKind ExpandNextToken);
519 const AsmToken &Lex() override { return Lex(ExpandMacros); }
520
521 void setParsingMSInlineAsm(bool V) override {
522 ParsingMSInlineAsm = V;
523 // When parsing MS inline asm, we must lex 0b1101 and 0ABCH as binary and
524 // hex integer literals.
525 Lexer.setLexMasmIntegers(V);
526 }
527 bool isParsingMSInlineAsm() override { return ParsingMSInlineAsm; }
528
529 bool isParsingMasm() const override { return true; }
530
531 bool defineMacro(StringRef Name, StringRef Value) override;
532
533 bool lookUpField(StringRef Name, AsmFieldInfo &Info) const override;
534 bool lookUpField(StringRef Base, StringRef Member,
535 AsmFieldInfo &Info) const override;
536
537 bool lookUpType(StringRef Name, AsmTypeInfo &Info) const override;
538
539 bool parseMSInlineAsm(std::string &AsmString, unsigned &NumOutputs,
540 unsigned &NumInputs,
541 SmallVectorImpl<std::pair<void *, bool>> &OpDecls,
542 SmallVectorImpl<std::string> &Constraints,
544 const MCInstrInfo *MII, const MCInstPrinter *IP,
545 MCAsmParserSemaCallback &SI) override;
546
547 bool parseExpression(const MCExpr *&Res);
548 bool parseExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
549 bool parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
550 AsmTypeInfo *TypeInfo) override;
551 bool parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) override;
552 bool parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
553 SMLoc &EndLoc) override;
554 bool parseAbsoluteExpression(int64_t &Res) override;
555
556 /// Parse a floating point expression using the float \p Semantics
557 /// and set \p Res to the value.
558 bool parseRealValue(const fltSemantics &Semantics, APInt &Res);
559
560 /// Parse an identifier or string (as a quoted identifier)
561 /// and set \p Res to the identifier contents.
562 enum IdentifierPositionKind { StandardPosition, StartOfStatement };
563 bool parseIdentifier(StringRef &Res, IdentifierPositionKind Position);
564 bool parseIdentifier(StringRef &Res) override {
565 return parseIdentifier(Res, StandardPosition);
566 }
567 void eatToEndOfStatement() override;
568
569 bool checkForValidSection() override;
570
571 /// }
572
573private:
574 bool expandMacros();
575 const AsmToken peekTok(bool ShouldSkipSpace = true);
576
577 bool parseStatement(ParseStatementInfo &Info,
579 bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites);
580 bool parseCppHashLineFilenameComment(SMLoc L);
581
582 bool expandMacro(raw_svector_ostream &OS, StringRef Body,
585 const std::vector<std::string> &Locals, SMLoc L);
586
587 /// Are we inside a macro instantiation?
588 bool isInsideMacroInstantiation() {return !ActiveMacros.empty();}
589
590 /// Handle entry to macro instantiation.
591 ///
592 /// \param M The macro.
593 /// \param NameLoc Instantiation location.
594 bool handleMacroEntry(
595 const MCAsmMacro *M, SMLoc NameLoc,
597
598 /// Handle invocation of macro function.
599 ///
600 /// \param M The macro.
601 /// \param NameLoc Invocation location.
602 bool handleMacroInvocation(const MCAsmMacro *M, SMLoc NameLoc);
603
604 /// Handle exit from macro instantiation.
605 void handleMacroExit();
606
607 /// Extract AsmTokens for a macro argument.
608 bool
609 parseMacroArgument(const MCAsmMacroParameter *MP, MCAsmMacroArgument &MA,
611
612 /// Parse all macro arguments for a given macro.
613 bool
614 parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A,
616
617 void printMacroInstantiations();
618
619 bool expandStatement(SMLoc Loc);
620
621 void printMessage(SMLoc Loc, SourceMgr::DiagKind Kind, const Twine &Msg,
622 SMRange Range = std::nullopt) const {
624 SrcMgr.PrintMessage(Loc, Kind, Msg, Ranges);
625 }
626 static void DiagHandler(const SMDiagnostic &Diag, void *Context);
627
628 bool lookUpField(const StructInfo &Structure, StringRef Member,
629 AsmFieldInfo &Info) const;
630
631 /// Should we emit DWARF describing this assembler source? (Returns false if
632 /// the source has .file directives, which means we don't want to generate
633 /// info describing the assembler source itself.)
634 bool enabledGenDwarfForAssembly();
635
636 /// Enter the specified file. This returns true on failure.
637 bool enterIncludeFile(const std::string &Filename);
638
639 /// Reset the current lexer position to that given by \p Loc. The
640 /// current token is not set; clients should ensure Lex() is called
641 /// subsequently.
642 ///
643 /// \param InBuffer If not 0, should be the known buffer id that contains the
644 /// location.
645 void jumpToLoc(SMLoc Loc, unsigned InBuffer = 0,
646 bool EndStatementAtEOF = true);
647
648 /// Parse up to a token of kind \p EndTok and return the contents from the
649 /// current token up to (but not including) this token; the current token on
650 /// exit will be either this kind or EOF. Reads through instantiated macro
651 /// functions and text macros.
652 SmallVector<StringRef, 1> parseStringRefsTo(AsmToken::TokenKind EndTok);
653 std::string parseStringTo(AsmToken::TokenKind EndTok);
654
655 /// Parse up to the end of statement and return the contents from the current
656 /// token until the end of the statement; the current token on exit will be
657 /// either the EndOfStatement or EOF.
659
660 bool parseTextItem(std::string &Data);
661
662 unsigned getBinOpPrecedence(AsmToken::TokenKind K,
664
665 bool parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, SMLoc &EndLoc);
666 bool parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc);
667 bool parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc);
668
669 bool parseRegisterOrRegisterNumber(int64_t &Register, SMLoc DirectiveLoc);
670
671 bool parseCVFunctionId(int64_t &FunctionId, StringRef DirectiveName);
672 bool parseCVFileId(int64_t &FileId, StringRef DirectiveName);
673
674 // Generic (target and platform independent) directive parsing.
675 enum DirectiveKind {
676 DK_NO_DIRECTIVE, // Placeholder
677 DK_HANDLER_DIRECTIVE,
678 DK_ASSIGN,
679 DK_EQU,
680 DK_TEXTEQU,
681 DK_ASCII,
682 DK_ASCIZ,
683 DK_STRING,
684 DK_BYTE,
685 DK_SBYTE,
686 DK_WORD,
687 DK_SWORD,
688 DK_DWORD,
689 DK_SDWORD,
690 DK_FWORD,
691 DK_QWORD,
692 DK_SQWORD,
693 DK_DB,
694 DK_DD,
695 DK_DF,
696 DK_DQ,
697 DK_DW,
698 DK_REAL4,
699 DK_REAL8,
700 DK_REAL10,
701 DK_ALIGN,
702 DK_EVEN,
703 DK_ORG,
704 DK_ENDR,
705 DK_EXTERN,
706 DK_PUBLIC,
707 DK_COMM,
708 DK_COMMENT,
709 DK_INCLUDE,
710 DK_REPEAT,
711 DK_WHILE,
712 DK_FOR,
713 DK_FORC,
714 DK_IF,
715 DK_IFE,
716 DK_IFB,
717 DK_IFNB,
718 DK_IFDEF,
719 DK_IFNDEF,
720 DK_IFDIF,
721 DK_IFDIFI,
722 DK_IFIDN,
723 DK_IFIDNI,
724 DK_ELSEIF,
725 DK_ELSEIFE,
726 DK_ELSEIFB,
727 DK_ELSEIFNB,
728 DK_ELSEIFDEF,
729 DK_ELSEIFNDEF,
730 DK_ELSEIFDIF,
731 DK_ELSEIFDIFI,
732 DK_ELSEIFIDN,
733 DK_ELSEIFIDNI,
734 DK_ELSE,
735 DK_ENDIF,
736 DK_FILE,
737 DK_LINE,
738 DK_LOC,
739 DK_STABS,
740 DK_CV_FILE,
741 DK_CV_FUNC_ID,
742 DK_CV_INLINE_SITE_ID,
743 DK_CV_LOC,
744 DK_CV_LINETABLE,
745 DK_CV_INLINE_LINETABLE,
746 DK_CV_DEF_RANGE,
747 DK_CV_STRINGTABLE,
748 DK_CV_STRING,
749 DK_CV_FILECHECKSUMS,
750 DK_CV_FILECHECKSUM_OFFSET,
751 DK_CV_FPO_DATA,
752 DK_CFI_SECTIONS,
753 DK_CFI_STARTPROC,
754 DK_CFI_ENDPROC,
755 DK_CFI_DEF_CFA,
756 DK_CFI_DEF_CFA_OFFSET,
757 DK_CFI_ADJUST_CFA_OFFSET,
758 DK_CFI_DEF_CFA_REGISTER,
759 DK_CFI_OFFSET,
760 DK_CFI_REL_OFFSET,
761 DK_CFI_PERSONALITY,
762 DK_CFI_LSDA,
763 DK_CFI_REMEMBER_STATE,
764 DK_CFI_RESTORE_STATE,
765 DK_CFI_SAME_VALUE,
766 DK_CFI_RESTORE,
767 DK_CFI_ESCAPE,
768 DK_CFI_RETURN_COLUMN,
769 DK_CFI_SIGNAL_FRAME,
770 DK_CFI_UNDEFINED,
771 DK_CFI_REGISTER,
772 DK_CFI_WINDOW_SAVE,
773 DK_CFI_B_KEY_FRAME,
774 DK_MACRO,
775 DK_EXITM,
776 DK_ENDM,
777 DK_PURGE,
778 DK_ERR,
779 DK_ERRB,
780 DK_ERRNB,
781 DK_ERRDEF,
782 DK_ERRNDEF,
783 DK_ERRDIF,
784 DK_ERRDIFI,
785 DK_ERRIDN,
786 DK_ERRIDNI,
787 DK_ERRE,
788 DK_ERRNZ,
789 DK_ECHO,
790 DK_STRUCT,
791 DK_UNION,
792 DK_ENDS,
793 DK_END,
794 DK_PUSHFRAME,
795 DK_PUSHREG,
796 DK_SAVEREG,
797 DK_SAVEXMM128,
798 DK_SETFRAME,
799 DK_RADIX,
800 };
801
802 /// Maps directive name --> DirectiveKind enum, for directives parsed by this
803 /// class.
804 StringMap<DirectiveKind> DirectiveKindMap;
805
806 bool isMacroLikeDirective();
807
808 // Codeview def_range type parsing.
809 enum CVDefRangeType {
810 CVDR_DEFRANGE = 0, // Placeholder
811 CVDR_DEFRANGE_REGISTER,
812 CVDR_DEFRANGE_FRAMEPOINTER_REL,
813 CVDR_DEFRANGE_SUBFIELD_REGISTER,
814 CVDR_DEFRANGE_REGISTER_REL
815 };
816
817 /// Maps Codeview def_range types --> CVDefRangeType enum, for Codeview
818 /// def_range types parsed by this class.
819 StringMap<CVDefRangeType> CVDefRangeTypeMap;
820
821 // Generic (target and platform independent) directive parsing.
822 enum BuiltinSymbol {
823 BI_NO_SYMBOL, // Placeholder
824 BI_DATE,
825 BI_TIME,
826 BI_VERSION,
827 BI_FILECUR,
828 BI_FILENAME,
829 BI_LINE,
830 BI_CURSEG,
831 BI_CPU,
832 BI_INTERFACE,
833 BI_CODE,
834 BI_DATA,
835 BI_FARDATA,
836 BI_WORDSIZE,
837 BI_CODESIZE,
838 BI_DATASIZE,
839 BI_MODEL,
840 BI_STACK,
841 };
842
843 /// Maps builtin name --> BuiltinSymbol enum, for builtins handled by this
844 /// class.
845 StringMap<BuiltinSymbol> BuiltinSymbolMap;
846
847 const MCExpr *evaluateBuiltinValue(BuiltinSymbol Symbol, SMLoc StartLoc);
848
849 std::optional<std::string> evaluateBuiltinTextMacro(BuiltinSymbol Symbol,
850 SMLoc StartLoc);
851
852 // ".ascii", ".asciz", ".string"
853 bool parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated);
854
855 // "byte", "word", ...
856 bool emitIntValue(const MCExpr *Value, unsigned Size);
857 bool parseScalarInitializer(unsigned Size,
859 unsigned StringPadLength = 0);
860 bool parseScalarInstList(
861 unsigned Size, SmallVectorImpl<const MCExpr *> &Values,
863 bool emitIntegralValues(unsigned Size, unsigned *Count = nullptr);
864 bool addIntegralField(StringRef Name, unsigned Size);
865 bool parseDirectiveValue(StringRef IDVal, unsigned Size);
866 bool parseDirectiveNamedValue(StringRef TypeName, unsigned Size,
867 StringRef Name, SMLoc NameLoc);
868
869 // "real4", "real8", "real10"
870 bool emitRealValues(const fltSemantics &Semantics, unsigned *Count = nullptr);
871 bool addRealField(StringRef Name, const fltSemantics &Semantics, size_t Size);
872 bool parseDirectiveRealValue(StringRef IDVal, const fltSemantics &Semantics,
873 size_t Size);
874 bool parseRealInstList(
875 const fltSemantics &Semantics, SmallVectorImpl<APInt> &Values,
877 bool parseDirectiveNamedRealValue(StringRef TypeName,
878 const fltSemantics &Semantics,
879 unsigned Size, StringRef Name,
880 SMLoc NameLoc);
881
882 bool parseOptionalAngleBracketOpen();
883 bool parseAngleBracketClose(const Twine &Msg = "expected '>'");
884
885 bool parseFieldInitializer(const FieldInfo &Field,
886 FieldInitializer &Initializer);
887 bool parseFieldInitializer(const FieldInfo &Field,
888 const IntFieldInfo &Contents,
889 FieldInitializer &Initializer);
890 bool parseFieldInitializer(const FieldInfo &Field,
891 const RealFieldInfo &Contents,
892 FieldInitializer &Initializer);
893 bool parseFieldInitializer(const FieldInfo &Field,
894 const StructFieldInfo &Contents,
895 FieldInitializer &Initializer);
896
897 bool parseStructInitializer(const StructInfo &Structure,
898 StructInitializer &Initializer);
899 bool parseStructInstList(
900 const StructInfo &Structure, std::vector<StructInitializer> &Initializers,
902
903 bool emitFieldValue(const FieldInfo &Field);
904 bool emitFieldValue(const FieldInfo &Field, const IntFieldInfo &Contents);
905 bool emitFieldValue(const FieldInfo &Field, const RealFieldInfo &Contents);
906 bool emitFieldValue(const FieldInfo &Field, const StructFieldInfo &Contents);
907
908 bool emitFieldInitializer(const FieldInfo &Field,
909 const FieldInitializer &Initializer);
910 bool emitFieldInitializer(const FieldInfo &Field,
911 const IntFieldInfo &Contents,
912 const IntFieldInfo &Initializer);
913 bool emitFieldInitializer(const FieldInfo &Field,
914 const RealFieldInfo &Contents,
915 const RealFieldInfo &Initializer);
916 bool emitFieldInitializer(const FieldInfo &Field,
917 const StructFieldInfo &Contents,
918 const StructFieldInfo &Initializer);
919
920 bool emitStructInitializer(const StructInfo &Structure,
921 const StructInitializer &Initializer);
922
923 // User-defined types (structs, unions):
924 bool emitStructValues(const StructInfo &Structure, unsigned *Count = nullptr);
925 bool addStructField(StringRef Name, const StructInfo &Structure);
926 bool parseDirectiveStructValue(const StructInfo &Structure,
927 StringRef Directive, SMLoc DirLoc);
928 bool parseDirectiveNamedStructValue(const StructInfo &Structure,
929 StringRef Directive, SMLoc DirLoc,
931
932 // "=", "equ", "textequ"
933 bool parseDirectiveEquate(StringRef IDVal, StringRef Name,
934 DirectiveKind DirKind, SMLoc NameLoc);
935
936 bool parseDirectiveOrg(); // "org"
937
938 bool emitAlignTo(int64_t Alignment);
939 bool parseDirectiveAlign(); // "align"
940 bool parseDirectiveEven(); // "even"
941
942 // ".file", ".line", ".loc", ".stabs"
943 bool parseDirectiveFile(SMLoc DirectiveLoc);
944 bool parseDirectiveLine();
945 bool parseDirectiveLoc();
946 bool parseDirectiveStabs();
947
948 // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable",
949 // ".cv_inline_linetable", ".cv_def_range", ".cv_string"
950 bool parseDirectiveCVFile();
951 bool parseDirectiveCVFuncId();
952 bool parseDirectiveCVInlineSiteId();
953 bool parseDirectiveCVLoc();
954 bool parseDirectiveCVLinetable();
955 bool parseDirectiveCVInlineLinetable();
956 bool parseDirectiveCVDefRange();
957 bool parseDirectiveCVString();
958 bool parseDirectiveCVStringTable();
959 bool parseDirectiveCVFileChecksums();
960 bool parseDirectiveCVFileChecksumOffset();
961 bool parseDirectiveCVFPOData();
962
963 // .cfi directives
964 bool parseDirectiveCFIRegister(SMLoc DirectiveLoc);
965 bool parseDirectiveCFIWindowSave(SMLoc DirectiveLoc);
966 bool parseDirectiveCFISections();
967 bool parseDirectiveCFIStartProc();
968 bool parseDirectiveCFIEndProc();
969 bool parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc);
970 bool parseDirectiveCFIDefCfa(SMLoc DirectiveLoc);
971 bool parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc);
972 bool parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc);
973 bool parseDirectiveCFIOffset(SMLoc DirectiveLoc);
974 bool parseDirectiveCFIRelOffset(SMLoc DirectiveLoc);
975 bool parseDirectiveCFIPersonalityOrLsda(bool IsPersonality);
976 bool parseDirectiveCFIRememberState(SMLoc DirectiveLoc);
977 bool parseDirectiveCFIRestoreState(SMLoc DirectiveLoc);
978 bool parseDirectiveCFISameValue(SMLoc DirectiveLoc);
979 bool parseDirectiveCFIRestore(SMLoc DirectiveLoc);
980 bool parseDirectiveCFIEscape(SMLoc DirectiveLoc);
981 bool parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc);
982 bool parseDirectiveCFISignalFrame();
983 bool parseDirectiveCFIUndefined(SMLoc DirectiveLoc);
984
985 // macro directives
986 bool parseDirectivePurgeMacro(SMLoc DirectiveLoc);
987 bool parseDirectiveExitMacro(SMLoc DirectiveLoc, StringRef Directive,
988 std::string &Value);
989 bool parseDirectiveEndMacro(StringRef Directive);
990 bool parseDirectiveMacro(StringRef Name, SMLoc NameLoc);
991
992 bool parseDirectiveStruct(StringRef Directive, DirectiveKind DirKind,
993 StringRef Name, SMLoc NameLoc);
994 bool parseDirectiveNestedStruct(StringRef Directive, DirectiveKind DirKind);
995 bool parseDirectiveEnds(StringRef Name, SMLoc NameLoc);
996 bool parseDirectiveNestedEnds();
997
998 bool parseDirectiveExtern();
999
1000 /// Parse a directive like ".globl" which accepts a single symbol (which
1001 /// should be a label or an external).
1002 bool parseDirectiveSymbolAttribute(MCSymbolAttr Attr);
1003
1004 bool parseDirectiveComm(bool IsLocal); // ".comm" and ".lcomm"
1005
1006 bool parseDirectiveComment(SMLoc DirectiveLoc); // "comment"
1007
1008 bool parseDirectiveInclude(); // "include"
1009
1010 // "if" or "ife"
1011 bool parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
1012 // "ifb" or "ifnb", depending on ExpectBlank.
1013 bool parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank);
1014 // "ifidn", "ifdif", "ifidni", or "ifdifi", depending on ExpectEqual and
1015 // CaseInsensitive.
1016 bool parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual,
1017 bool CaseInsensitive);
1018 // "ifdef" or "ifndef", depending on expect_defined
1019 bool parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined);
1020 // "elseif" or "elseife"
1021 bool parseDirectiveElseIf(SMLoc DirectiveLoc, DirectiveKind DirKind);
1022 // "elseifb" or "elseifnb", depending on ExpectBlank.
1023 bool parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank);
1024 // ".elseifdef" or ".elseifndef", depending on expect_defined
1025 bool parseDirectiveElseIfdef(SMLoc DirectiveLoc, bool expect_defined);
1026 // "elseifidn", "elseifdif", "elseifidni", or "elseifdifi", depending on
1027 // ExpectEqual and CaseInsensitive.
1028 bool parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual,
1029 bool CaseInsensitive);
1030 bool parseDirectiveElse(SMLoc DirectiveLoc); // "else"
1031 bool parseDirectiveEndIf(SMLoc DirectiveLoc); // "endif"
1032 bool parseEscapedString(std::string &Data) override;
1033 bool parseAngleBracketString(std::string &Data) override;
1034
1035 // Macro-like directives
1036 MCAsmMacro *parseMacroLikeBody(SMLoc DirectiveLoc);
1037 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
1039 void instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc,
1040 SMLoc ExitLoc, raw_svector_ostream &OS);
1041 bool parseDirectiveRepeat(SMLoc DirectiveLoc, StringRef Directive);
1042 bool parseDirectiveFor(SMLoc DirectiveLoc, StringRef Directive);
1043 bool parseDirectiveForc(SMLoc DirectiveLoc, StringRef Directive);
1044 bool parseDirectiveWhile(SMLoc DirectiveLoc);
1045
1046 // "_emit" or "__emit"
1047 bool parseDirectiveMSEmit(SMLoc DirectiveLoc, ParseStatementInfo &Info,
1048 size_t Len);
1049
1050 // "align"
1051 bool parseDirectiveMSAlign(SMLoc DirectiveLoc, ParseStatementInfo &Info);
1052
1053 // "end"
1054 bool parseDirectiveEnd(SMLoc DirectiveLoc);
1055
1056 // ".err"
1057 bool parseDirectiveError(SMLoc DirectiveLoc);
1058 // ".errb" or ".errnb", depending on ExpectBlank.
1059 bool parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank);
1060 // ".errdef" or ".errndef", depending on ExpectBlank.
1061 bool parseDirectiveErrorIfdef(SMLoc DirectiveLoc, bool ExpectDefined);
1062 // ".erridn", ".errdif", ".erridni", or ".errdifi", depending on ExpectEqual
1063 // and CaseInsensitive.
1064 bool parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual,
1065 bool CaseInsensitive);
1066 // ".erre" or ".errnz", depending on ExpectZero.
1067 bool parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero);
1068
1069 // ".radix"
1070 bool parseDirectiveRadix(SMLoc DirectiveLoc);
1071
1072 // "echo"
1073 bool parseDirectiveEcho(SMLoc DirectiveLoc);
1074
1075 void initializeDirectiveKindMap();
1076 void initializeCVDefRangeTypeMap();
1077 void initializeBuiltinSymbolMap();
1078};
1079
1080} // end anonymous namespace
1081
1082namespace llvm {
1083
1085
1087
1088} // end namespace llvm
1089
1091
1092MasmParser::MasmParser(SourceMgr &SM, MCContext &Ctx, MCStreamer &Out,
1093 const MCAsmInfo &MAI, struct tm TM, unsigned CB)
1094 : Lexer(MAI), Ctx(Ctx), Out(Out), MAI(MAI), SrcMgr(SM),
1095 CurBuffer(CB ? CB : SM.getMainFileID()), TM(TM) {
1096 HadError = false;
1097 // Save the old handler.
1098 SavedDiagHandler = SrcMgr.getDiagHandler();
1099 SavedDiagContext = SrcMgr.getDiagContext();
1100 // Set our own handler which calls the saved handler.
1102 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
1103 EndStatementAtEOFStack.push_back(true);
1104
1105 // Initialize the platform / file format parser.
1106 switch (Ctx.getObjectFileType()) {
1107 case MCContext::IsCOFF:
1108 PlatformParser.reset(createCOFFMasmParser());
1109 break;
1110 default:
1111 report_fatal_error("llvm-ml currently supports only COFF output.");
1112 break;
1113 }
1114
1115 initializeDirectiveKindMap();
1116 PlatformParser->Initialize(*this);
1117 initializeCVDefRangeTypeMap();
1118 initializeBuiltinSymbolMap();
1119
1120 NumOfMacroInstantiations = 0;
1121}
1122
1123MasmParser::~MasmParser() {
1124 assert((HadError || ActiveMacros.empty()) &&
1125 "Unexpected active macro instantiation!");
1126
1127 // Restore the saved diagnostics handler and context for use during
1128 // finalization.
1129 SrcMgr.setDiagHandler(SavedDiagHandler, SavedDiagContext);
1130}
1131
1132void MasmParser::printMacroInstantiations() {
1133 // Print the active macro instantiation stack.
1134 for (std::vector<MacroInstantiation *>::const_reverse_iterator
1135 it = ActiveMacros.rbegin(),
1136 ie = ActiveMacros.rend();
1137 it != ie; ++it)
1138 printMessage((*it)->InstantiationLoc, SourceMgr::DK_Note,
1139 "while in macro instantiation");
1140}
1141
1142void MasmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) {
1143 printPendingErrors();
1144 printMessage(L, SourceMgr::DK_Note, Msg, Range);
1145 printMacroInstantiations();
1146}
1147
1148bool MasmParser::Warning(SMLoc L, const Twine &Msg, SMRange Range) {
1149 if (getTargetParser().getTargetOptions().MCNoWarn)
1150 return false;
1151 if (getTargetParser().getTargetOptions().MCFatalWarnings)
1152 return Error(L, Msg, Range);
1153 printMessage(L, SourceMgr::DK_Warning, Msg, Range);
1154 printMacroInstantiations();
1155 return false;
1156}
1157
1158bool MasmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) {
1159 HadError = true;
1160 printMessage(L, SourceMgr::DK_Error, Msg, Range);
1161 printMacroInstantiations();
1162 return true;
1163}
1164
1165bool MasmParser::enterIncludeFile(const std::string &Filename) {
1166 std::string IncludedFile;
1167 unsigned NewBuf =
1168 SrcMgr.AddIncludeFile(Filename, Lexer.getLoc(), IncludedFile);
1169 if (!NewBuf)
1170 return true;
1171
1172 CurBuffer = NewBuf;
1173 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
1174 EndStatementAtEOFStack.push_back(true);
1175 return false;
1176}
1177
1178void MasmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer,
1179 bool EndStatementAtEOF) {
1180 CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc);
1181 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(),
1182 Loc.getPointer(), EndStatementAtEOF);
1183}
1184
1185bool MasmParser::expandMacros() {
1186 const AsmToken &Tok = getTok();
1187 const std::string IDLower = Tok.getIdentifier().lower();
1188
1189 const llvm::MCAsmMacro *M = getContext().lookupMacro(IDLower);
1190 if (M && M->IsFunction && peekTok().is(AsmToken::LParen)) {
1191 // This is a macro function invocation; expand it in place.
1192 const SMLoc MacroLoc = Tok.getLoc();
1193 const StringRef MacroId = Tok.getIdentifier();
1194 Lexer.Lex();
1195 if (handleMacroInvocation(M, MacroLoc)) {
1196 Lexer.UnLex(AsmToken(AsmToken::Error, MacroId));
1197 Lexer.Lex();
1198 }
1199 return false;
1200 }
1201
1202 std::optional<std::string> ExpandedValue;
1203 auto BuiltinIt = BuiltinSymbolMap.find(IDLower);
1204 if (BuiltinIt != BuiltinSymbolMap.end()) {
1205 ExpandedValue =
1206 evaluateBuiltinTextMacro(BuiltinIt->getValue(), Tok.getLoc());
1207 } else {
1208 auto VarIt = Variables.find(IDLower);
1209 if (VarIt != Variables.end() && VarIt->getValue().IsText) {
1210 ExpandedValue = VarIt->getValue().TextValue;
1211 }
1212 }
1213
1214 if (!ExpandedValue)
1215 return true;
1216 std::unique_ptr<MemoryBuffer> Instantiation =
1217 MemoryBuffer::getMemBufferCopy(*ExpandedValue, "<instantiation>");
1218
1219 // Jump to the macro instantiation and prime the lexer.
1220 CurBuffer =
1221 SrcMgr.AddNewSourceBuffer(std::move(Instantiation), Tok.getEndLoc());
1222 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), nullptr,
1223 /*EndStatementAtEOF=*/false);
1224 EndStatementAtEOFStack.push_back(false);
1225 Lexer.Lex();
1226 return false;
1227}
1228
1229const AsmToken &MasmParser::Lex(ExpandKind ExpandNextToken) {
1230 if (Lexer.getTok().is(AsmToken::Error))
1231 Error(Lexer.getErrLoc(), Lexer.getErr());
1232
1233 // if it's a end of statement with a comment in it
1234 if (getTok().is(AsmToken::EndOfStatement)) {
1235 // if this is a line comment output it.
1236 if (!getTok().getString().empty() && getTok().getString().front() != '\n' &&
1237 getTok().getString().front() != '\r' && MAI.preserveAsmComments())
1238 Out.addExplicitComment(Twine(getTok().getString()));
1239 }
1240
1241 const AsmToken *tok = &Lexer.Lex();
1242 bool StartOfStatement = Lexer.isAtStartOfStatement();
1243
1244 while (ExpandNextToken == ExpandMacros && tok->is(AsmToken::Identifier)) {
1245 if (StartOfStatement) {
1246 AsmToken NextTok;
1247 MutableArrayRef<AsmToken> Buf(NextTok);
1248 size_t ReadCount = Lexer.peekTokens(Buf);
1249 if (ReadCount && NextTok.is(AsmToken::Identifier) &&
1250 (NextTok.getString().equals_insensitive("equ") ||
1251 NextTok.getString().equals_insensitive("textequ"))) {
1252 // This looks like an EQU or TEXTEQU directive; don't expand the
1253 // identifier, allowing for redefinitions.
1254 break;
1255 }
1256 }
1257 if (expandMacros())
1258 break;
1259 }
1260
1261 // Parse comments here to be deferred until end of next statement.
1262 while (tok->is(AsmToken::Comment)) {
1263 if (MAI.preserveAsmComments())
1264 Out.addExplicitComment(Twine(tok->getString()));
1265 tok = &Lexer.Lex();
1266 }
1267
1268 // Recognize and bypass line continuations.
1269 while (tok->is(AsmToken::BackSlash) &&
1270 peekTok().is(AsmToken::EndOfStatement)) {
1271 // Eat both the backslash and the end of statement.
1272 Lexer.Lex();
1273 tok = &Lexer.Lex();
1274 }
1275
1276 if (tok->is(AsmToken::Eof)) {
1277 // If this is the end of an included file, pop the parent file off the
1278 // include stack.
1279 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
1280 if (ParentIncludeLoc != SMLoc()) {
1281 EndStatementAtEOFStack.pop_back();
1282 jumpToLoc(ParentIncludeLoc, 0, EndStatementAtEOFStack.back());
1283 return Lex();
1284 }
1285 EndStatementAtEOFStack.pop_back();
1286 assert(EndStatementAtEOFStack.empty());
1287 }
1288
1289 return *tok;
1290}
1291
1292const AsmToken MasmParser::peekTok(bool ShouldSkipSpace) {
1293 AsmToken Tok;
1294
1296 size_t ReadCount = Lexer.peekTokens(Buf, ShouldSkipSpace);
1297
1298 if (ReadCount == 0) {
1299 // If this is the end of an included file, pop the parent file off the
1300 // include stack.
1301 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
1302 if (ParentIncludeLoc != SMLoc()) {
1303 EndStatementAtEOFStack.pop_back();
1304 jumpToLoc(ParentIncludeLoc, 0, EndStatementAtEOFStack.back());
1305 return peekTok(ShouldSkipSpace);
1306 }
1307 EndStatementAtEOFStack.pop_back();
1308 assert(EndStatementAtEOFStack.empty());
1309 }
1310
1311 assert(ReadCount == 1);
1312 return Tok;
1313}
1314
1315bool MasmParser::enabledGenDwarfForAssembly() {
1316 // Check whether the user specified -g.
1317 if (!getContext().getGenDwarfForAssembly())
1318 return false;
1319 // If we haven't encountered any .file directives (which would imply that
1320 // the assembler source was produced with debug info already) then emit one
1321 // describing the assembler source file itself.
1322 if (getContext().getGenDwarfFileNumber() == 0) {
1323 // Use the first #line directive for this, if any. It's preprocessed, so
1324 // there is no checksum, and of course no source directive.
1325 if (!FirstCppHashFilename.empty())
1326 getContext().setMCLineTableRootFile(
1327 /*CUID=*/0, getContext().getCompilationDir(), FirstCppHashFilename,
1328 /*Cksum=*/std::nullopt, /*Source=*/std::nullopt);
1329 const MCDwarfFile &RootFile =
1330 getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile();
1331 getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective(
1332 /*CUID=*/0, getContext().getCompilationDir(), RootFile.Name,
1333 RootFile.Checksum, RootFile.Source));
1334 }
1335 return true;
1336}
1337
1338bool MasmParser::Run(bool NoInitialTextSection, bool NoFinalize) {
1339 // Create the initial section, if requested.
1340 if (!NoInitialTextSection)
1341 Out.initSections(false, getTargetParser().getSTI());
1342
1343 // Prime the lexer.
1344 Lex();
1345
1346 HadError = false;
1347 AsmCond StartingCondState = TheCondState;
1348 SmallVector<AsmRewrite, 4> AsmStrRewrites;
1349
1350 // If we are generating dwarf for assembly source files save the initial text
1351 // section. (Don't use enabledGenDwarfForAssembly() here, as we aren't
1352 // emitting any actual debug info yet and haven't had a chance to parse any
1353 // embedded .file directives.)
1354 if (getContext().getGenDwarfForAssembly()) {
1355 MCSection *Sec = getStreamer().getCurrentSectionOnly();
1356 if (!Sec->getBeginSymbol()) {
1357 MCSymbol *SectionStartSym = getContext().createTempSymbol();
1358 getStreamer().emitLabel(SectionStartSym);
1359 Sec->setBeginSymbol(SectionStartSym);
1360 }
1361 bool InsertResult = getContext().addGenDwarfSection(Sec);
1362 assert(InsertResult && ".text section should not have debug info yet");
1363 (void)InsertResult;
1364 }
1365
1366 getTargetParser().onBeginOfFile();
1367
1368 // While we have input, parse each statement.
1369 while (Lexer.isNot(AsmToken::Eof) ||
1370 SrcMgr.getParentIncludeLoc(CurBuffer) != SMLoc()) {
1371 // Skip through the EOF at the end of an inclusion.
1372 if (Lexer.is(AsmToken::Eof))
1373 Lex();
1374
1375 ParseStatementInfo Info(&AsmStrRewrites);
1376 bool Parsed = parseStatement(Info, nullptr);
1377
1378 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error
1379 // for printing ErrMsg via Lex() only if no (presumably better) parser error
1380 // exists.
1381 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) {
1382 Lex();
1383 }
1384
1385 // parseStatement returned true so may need to emit an error.
1386 printPendingErrors();
1387
1388 // Skipping to the next line if needed.
1389 if (Parsed && !getLexer().isAtStartOfStatement())
1390 eatToEndOfStatement();
1391 }
1392
1393 getTargetParser().onEndOfFile();
1394 printPendingErrors();
1395
1396 // All errors should have been emitted.
1397 assert(!hasPendingError() && "unexpected error from parseStatement");
1398
1399 getTargetParser().flushPendingInstructions(getStreamer());
1400
1401 if (TheCondState.TheCond != StartingCondState.TheCond ||
1402 TheCondState.Ignore != StartingCondState.Ignore)
1403 printError(getTok().getLoc(), "unmatched .ifs or .elses");
1404 // Check to see there are no empty DwarfFile slots.
1405 const auto &LineTables = getContext().getMCDwarfLineTables();
1406 if (!LineTables.empty()) {
1407 unsigned Index = 0;
1408 for (const auto &File : LineTables.begin()->second.getMCDwarfFiles()) {
1409 if (File.Name.empty() && Index != 0)
1410 printError(getTok().getLoc(), "unassigned file number: " +
1411 Twine(Index) +
1412 " for .file directives");
1413 ++Index;
1414 }
1415 }
1416
1417 // Check to see that all assembler local symbols were actually defined.
1418 // Targets that don't do subsections via symbols may not want this, though,
1419 // so conservatively exclude them. Only do this if we're finalizing, though,
1420 // as otherwise we won't necessarilly have seen everything yet.
1421 if (!NoFinalize) {
1422 if (MAI.hasSubsectionsViaSymbols()) {
1423 for (const auto &TableEntry : getContext().getSymbols()) {
1424 MCSymbol *Sym = TableEntry.getValue();
1425 // Variable symbols may not be marked as defined, so check those
1426 // explicitly. If we know it's a variable, we have a definition for
1427 // the purposes of this check.
1428 if (Sym->isTemporary() && !Sym->isVariable() && !Sym->isDefined())
1429 // FIXME: We would really like to refer back to where the symbol was
1430 // first referenced for a source location. We need to add something
1431 // to track that. Currently, we just point to the end of the file.
1432 printError(getTok().getLoc(), "assembler local symbol '" +
1433 Sym->getName() + "' not defined");
1434 }
1435 }
1436
1437 // Temporary symbols like the ones for directional jumps don't go in the
1438 // symbol table. They also need to be diagnosed in all (final) cases.
1439 for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) {
1440 if (std::get<2>(LocSym)->isUndefined()) {
1441 // Reset the state of any "# line file" directives we've seen to the
1442 // context as it was at the diagnostic site.
1443 CppHashInfo = std::get<1>(LocSym);
1444 printError(std::get<0>(LocSym), "directional label undefined");
1445 }
1446 }
1447 }
1448
1449 // Finalize the output stream if there are no errors and if the client wants
1450 // us to.
1451 if (!HadError && !NoFinalize)
1452 Out.finish(Lexer.getLoc());
1453
1454 return HadError || getContext().hadError();
1455}
1456
1457bool MasmParser::checkForValidSection() {
1458 if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) {
1459 Out.initSections(false, getTargetParser().getSTI());
1460 return Error(getTok().getLoc(),
1461 "expected section directive before assembly directive");
1462 }
1463 return false;
1464}
1465
1466/// Throw away the rest of the line for testing purposes.
1467void MasmParser::eatToEndOfStatement() {
1468 while (Lexer.isNot(AsmToken::EndOfStatement)) {
1469 if (Lexer.is(AsmToken::Eof)) {
1470 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
1471 if (ParentIncludeLoc == SMLoc()) {
1472 break;
1473 }
1474
1475 EndStatementAtEOFStack.pop_back();
1476 jumpToLoc(ParentIncludeLoc, 0, EndStatementAtEOFStack.back());
1477 }
1478
1479 Lexer.Lex();
1480 }
1481
1482 // Eat EOL.
1483 if (Lexer.is(AsmToken::EndOfStatement))
1484 Lexer.Lex();
1485}
1486
1488MasmParser::parseStringRefsTo(AsmToken::TokenKind EndTok) {
1490 const char *Start = getTok().getLoc().getPointer();
1491 while (Lexer.isNot(EndTok)) {
1492 if (Lexer.is(AsmToken::Eof)) {
1493 SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(CurBuffer);
1494 if (ParentIncludeLoc == SMLoc()) {
1495 break;
1496 }
1497 Refs.emplace_back(Start, getTok().getLoc().getPointer() - Start);
1498
1499 EndStatementAtEOFStack.pop_back();
1500 jumpToLoc(ParentIncludeLoc, 0, EndStatementAtEOFStack.back());
1501 Lexer.Lex();
1502 Start = getTok().getLoc().getPointer();
1503 } else {
1504 Lexer.Lex();
1505 }
1506 }
1507 Refs.emplace_back(Start, getTok().getLoc().getPointer() - Start);
1508 return Refs;
1509}
1510
1511std::string MasmParser::parseStringTo(AsmToken::TokenKind EndTok) {
1512 SmallVector<StringRef, 1> Refs = parseStringRefsTo(EndTok);
1513 std::string Str;
1514 for (StringRef S : Refs) {
1515 Str.append(S.str());
1516 }
1517 return Str;
1518}
1519
1520StringRef MasmParser::parseStringToEndOfStatement() {
1521 const char *Start = getTok().getLoc().getPointer();
1522
1523 while (Lexer.isNot(AsmToken::EndOfStatement) && Lexer.isNot(AsmToken::Eof))
1524 Lexer.Lex();
1525
1526 const char *End = getTok().getLoc().getPointer();
1527 return StringRef(Start, End - Start);
1528}
1529
1530/// Parse a paren expression and return it.
1531/// NOTE: This assumes the leading '(' has already been consumed.
1532///
1533/// parenexpr ::= expr)
1534///
1535bool MasmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1536 if (parseExpression(Res))
1537 return true;
1538 EndLoc = Lexer.getTok().getEndLoc();
1539 return parseRParen();
1540}
1541
1542/// Parse a bracket expression and return it.
1543/// NOTE: This assumes the leading '[' has already been consumed.
1544///
1545/// bracketexpr ::= expr]
1546///
1547bool MasmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) {
1548 if (parseExpression(Res))
1549 return true;
1550 EndLoc = getTok().getEndLoc();
1551 if (parseToken(AsmToken::RBrac, "expected ']' in brackets expression"))
1552 return true;
1553 return false;
1554}
1555
1556/// Parse a primary expression and return it.
1557/// primaryexpr ::= (parenexpr
1558/// primaryexpr ::= symbol
1559/// primaryexpr ::= number
1560/// primaryexpr ::= '.'
1561/// primaryexpr ::= ~,+,-,'not' primaryexpr
1562/// primaryexpr ::= string
1563/// (a string is interpreted as a 64-bit number in big-endian base-256)
1564bool MasmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc,
1565 AsmTypeInfo *TypeInfo) {
1566 SMLoc FirstTokenLoc = getLexer().getLoc();
1567 AsmToken::TokenKind FirstTokenKind = Lexer.getKind();
1568 switch (FirstTokenKind) {
1569 default:
1570 return TokError("unknown token in expression");
1571 // If we have an error assume that we've already handled it.
1572 case AsmToken::Error:
1573 return true;
1574 case AsmToken::Exclaim:
1575 Lex(); // Eat the operator.
1576 if (parsePrimaryExpr(Res, EndLoc, nullptr))
1577 return true;
1578 Res = MCUnaryExpr::createLNot(Res, getContext(), FirstTokenLoc);
1579 return false;
1580 case AsmToken::Dollar:
1581 case AsmToken::At:
1582 case AsmToken::Identifier: {
1584 if (parseIdentifier(Identifier)) {
1585 // We may have failed but $ may be a valid token.
1586 if (getTok().is(AsmToken::Dollar)) {
1587 if (Lexer.getMAI().getDollarIsPC()) {
1588 Lex();
1589 // This is a '$' reference, which references the current PC. Emit a
1590 // temporary label to the streamer and refer to it.
1591 MCSymbol *Sym = Ctx.createTempSymbol();
1592 Out.emitLabel(Sym);
1594 getContext());
1595 EndLoc = FirstTokenLoc;
1596 return false;
1597 }
1598 return Error(FirstTokenLoc, "invalid token in expression");
1599 }
1600 }
1601 // Parse named bitwise negation.
1602 if (Identifier.equals_insensitive("not")) {
1603 if (parsePrimaryExpr(Res, EndLoc, nullptr))
1604 return true;
1605 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc);
1606 return false;
1607 }
1608 // Parse directional local label references.
1609 if (Identifier.equals_insensitive("@b") ||
1610 Identifier.equals_insensitive("@f")) {
1611 bool Before = Identifier.equals_insensitive("@b");
1612 MCSymbol *Sym = getContext().getDirectionalLocalSymbol(0, Before);
1613 if (Before && Sym->isUndefined())
1614 return Error(FirstTokenLoc, "Expected @@ label before @B reference");
1615 Res = MCSymbolRefExpr::create(Sym, getContext());
1616 return false;
1617 }
1618 // Parse symbol variant.
1619 std::pair<StringRef, StringRef> Split;
1620 if (!MAI.useParensForSymbolVariant()) {
1621 Split = Identifier.split('@');
1622 } else if (Lexer.is(AsmToken::LParen)) {
1623 Lex(); // eat '('.
1624 StringRef VName;
1625 parseIdentifier(VName);
1626 // eat ')'.
1627 if (parseToken(AsmToken::RParen,
1628 "unexpected token in variant, expected ')'"))
1629 return true;
1630 Split = std::make_pair(Identifier, VName);
1631 }
1632
1633 EndLoc = SMLoc::getFromPointer(Identifier.end());
1634
1635 // This is a symbol reference.
1637 if (SymbolName.empty())
1638 return Error(getLexer().getLoc(), "expected a symbol reference");
1639
1641
1642 // Look up the symbol variant if used.
1643 if (!Split.second.empty()) {
1645 if (Variant != MCSymbolRefExpr::VK_Invalid) {
1646 SymbolName = Split.first;
1647 } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) {
1649 } else {
1650 return Error(SMLoc::getFromPointer(Split.second.begin()),
1651 "invalid variant '" + Split.second + "'");
1652 }
1653 }
1654
1655 // Find the field offset if used.
1657 Split = SymbolName.split('.');
1658 if (Split.second.empty()) {
1659 } else {
1660 SymbolName = Split.first;
1661 if (lookUpField(SymbolName, Split.second, Info)) {
1662 std::pair<StringRef, StringRef> BaseMember = Split.second.split('.');
1663 StringRef Base = BaseMember.first, Member = BaseMember.second;
1664 lookUpField(Base, Member, Info);
1665 } else if (Structs.count(SymbolName.lower())) {
1666 // This is actually a reference to a field offset.
1667 Res = MCConstantExpr::create(Info.Offset, getContext());
1668 return false;
1669 }
1670 }
1671
1672 MCSymbol *Sym = getContext().getInlineAsmLabel(SymbolName);
1673 if (!Sym) {
1674 // If this is a built-in numeric value, treat it as a constant.
1675 auto BuiltinIt = BuiltinSymbolMap.find(SymbolName.lower());
1676 const BuiltinSymbol Symbol = (BuiltinIt == BuiltinSymbolMap.end())
1677 ? BI_NO_SYMBOL
1678 : BuiltinIt->getValue();
1679 if (Symbol != BI_NO_SYMBOL) {
1680 const MCExpr *Value = evaluateBuiltinValue(Symbol, FirstTokenLoc);
1681 if (Value) {
1682 Res = Value;
1683 return false;
1684 }
1685 }
1686
1687 // Variables use case-insensitive symbol names; if this is a variable, we
1688 // find the symbol using its canonical name.
1689 auto VarIt = Variables.find(SymbolName.lower());
1690 if (VarIt != Variables.end())
1691 SymbolName = VarIt->second.Name;
1692 Sym = getContext().getOrCreateSymbol(SymbolName);
1693 }
1694
1695 // If this is an absolute variable reference, substitute it now to preserve
1696 // semantics in the face of reassignment.
1697 if (Sym->isVariable()) {
1698 auto V = Sym->getVariableValue(/*SetUsed=*/false);
1699 bool DoInline = isa<MCConstantExpr>(V) && !Variant;
1700 if (auto TV = dyn_cast<MCTargetExpr>(V))
1701 DoInline = TV->inlineAssignedExpr();
1702 if (DoInline) {
1703 if (Variant)
1704 return Error(EndLoc, "unexpected modifier on variable reference");
1705 Res = Sym->getVariableValue(/*SetUsed=*/false);
1706 return false;
1707 }
1708 }
1709
1710 // Otherwise create a symbol ref.
1711 const MCExpr *SymRef =
1712 MCSymbolRefExpr::create(Sym, Variant, getContext(), FirstTokenLoc);
1713 if (Info.Offset) {
1715 MCBinaryExpr::Add, SymRef,
1716 MCConstantExpr::create(Info.Offset, getContext()), getContext());
1717 } else {
1718 Res = SymRef;
1719 }
1720 if (TypeInfo) {
1721 if (Info.Type.Name.empty()) {
1722 auto TypeIt = KnownType.find(Identifier.lower());
1723 if (TypeIt != KnownType.end()) {
1724 Info.Type = TypeIt->second;
1725 }
1726 }
1727
1728 *TypeInfo = Info.Type;
1729 }
1730 return false;
1731 }
1732 case AsmToken::BigNum:
1733 return TokError("literal value out of range for directive");
1734 case AsmToken::Integer: {
1735 int64_t IntVal = getTok().getIntVal();
1736 Res = MCConstantExpr::create(IntVal, getContext());
1737 EndLoc = Lexer.getTok().getEndLoc();
1738 Lex(); // Eat token.
1739 return false;
1740 }
1741 case AsmToken::String: {
1742 // MASM strings (used as constants) are interpreted as big-endian base-256.
1743 SMLoc ValueLoc = getTok().getLoc();
1744 std::string Value;
1745 if (parseEscapedString(Value))
1746 return true;
1747 if (Value.size() > 8)
1748 return Error(ValueLoc, "literal value out of range");
1749 uint64_t IntValue = 0;
1750 for (const unsigned char CharVal : Value)
1751 IntValue = (IntValue << 8) | CharVal;
1752 Res = MCConstantExpr::create(IntValue, getContext());
1753 return false;
1754 }
1755 case AsmToken::Real: {
1756 APFloat RealVal(APFloat::IEEEdouble(), getTok().getString());
1757 uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue();
1758 Res = MCConstantExpr::create(IntVal, getContext());
1759 EndLoc = Lexer.getTok().getEndLoc();
1760 Lex(); // Eat token.
1761 return false;
1762 }
1763 case AsmToken::Dot: {
1764 // This is a '.' reference, which references the current PC. Emit a
1765 // temporary label to the streamer and refer to it.
1766 MCSymbol *Sym = Ctx.createTempSymbol();
1767 Out.emitLabel(Sym);
1769 EndLoc = Lexer.getTok().getEndLoc();
1770 Lex(); // Eat identifier.
1771 return false;
1772 }
1773 case AsmToken::LParen:
1774 Lex(); // Eat the '('.
1775 return parseParenExpr(Res, EndLoc);
1776 case AsmToken::LBrac:
1777 if (!PlatformParser->HasBracketExpressions())
1778 return TokError("brackets expression not supported on this target");
1779 Lex(); // Eat the '['.
1780 return parseBracketExpr(Res, EndLoc);
1781 case AsmToken::Minus:
1782 Lex(); // Eat the operator.
1783 if (parsePrimaryExpr(Res, EndLoc, nullptr))
1784 return true;
1785 Res = MCUnaryExpr::createMinus(Res, getContext(), FirstTokenLoc);
1786 return false;
1787 case AsmToken::Plus:
1788 Lex(); // Eat the operator.
1789 if (parsePrimaryExpr(Res, EndLoc, nullptr))
1790 return true;
1791 Res = MCUnaryExpr::createPlus(Res, getContext(), FirstTokenLoc);
1792 return false;
1793 case AsmToken::Tilde:
1794 Lex(); // Eat the operator.
1795 if (parsePrimaryExpr(Res, EndLoc, nullptr))
1796 return true;
1797 Res = MCUnaryExpr::createNot(Res, getContext(), FirstTokenLoc);
1798 return false;
1799 // MIPS unary expression operators. The lexer won't generate these tokens if
1800 // MCAsmInfo::HasMipsExpressions is false for the target.
1825 Lex(); // Eat the operator.
1826 if (Lexer.isNot(AsmToken::LParen))
1827 return TokError("expected '(' after operator");
1828 Lex(); // Eat the operator.
1829 if (parseExpression(Res, EndLoc))
1830 return true;
1831 if (parseRParen())
1832 return true;
1833 Res = getTargetParser().createTargetUnaryExpr(Res, FirstTokenKind, Ctx);
1834 return !Res;
1835 }
1836}
1837
1838bool MasmParser::parseExpression(const MCExpr *&Res) {
1839 SMLoc EndLoc;
1840 return parseExpression(Res, EndLoc);
1841}
1842
1843/// This function checks if the next token is <string> type or arithmetic.
1844/// string that begin with character '<' must end with character '>'.
1845/// otherwise it is arithmetics.
1846/// If the function returns a 'true' value,
1847/// the End argument will be filled with the last location pointed to the '>'
1848/// character.
1849static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) {
1850 assert((StrLoc.getPointer() != nullptr) &&
1851 "Argument to the function cannot be a NULL value");
1852 const char *CharPtr = StrLoc.getPointer();
1853 while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') &&
1854 (*CharPtr != '\0')) {
1855 if (*CharPtr == '!')
1856 CharPtr++;
1857 CharPtr++;
1858 }
1859 if (*CharPtr == '>') {
1860 EndLoc = StrLoc.getFromPointer(CharPtr + 1);
1861 return true;
1862 }
1863 return false;
1864}
1865
1866/// creating a string without the escape characters '!'.
1867static std::string angleBracketString(StringRef BracketContents) {
1868 std::string Res;
1869 for (size_t Pos = 0; Pos < BracketContents.size(); Pos++) {
1870 if (BracketContents[Pos] == '!')
1871 Pos++;
1872 Res += BracketContents[Pos];
1873 }
1874 return Res;
1875}
1876
1877/// Parse an expression and return it.
1878///
1879/// expr ::= expr &&,|| expr -> lowest.
1880/// expr ::= expr |,^,&,! expr
1881/// expr ::= expr ==,!=,<>,<,<=,>,>= expr
1882/// expr ::= expr <<,>> expr
1883/// expr ::= expr +,- expr
1884/// expr ::= expr *,/,% expr -> highest.
1885/// expr ::= primaryexpr
1886///
1887bool MasmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1888 // Parse the expression.
1889 Res = nullptr;
1890 if (getTargetParser().parsePrimaryExpr(Res, EndLoc) ||
1891 parseBinOpRHS(1, Res, EndLoc))
1892 return true;
1893
1894 // Try to constant fold it up front, if possible. Do not exploit
1895 // assembler here.
1896 int64_t Value;
1897 if (Res->evaluateAsAbsolute(Value))
1898 Res = MCConstantExpr::create(Value, getContext());
1899
1900 return false;
1901}
1902
1903bool MasmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) {
1904 Res = nullptr;
1905 return parseParenExpr(Res, EndLoc) || parseBinOpRHS(1, Res, EndLoc);
1906}
1907
1908bool MasmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res,
1909 SMLoc &EndLoc) {
1910 if (parseParenExpr(Res, EndLoc))
1911 return true;
1912
1913 for (; ParenDepth > 0; --ParenDepth) {
1914 if (parseBinOpRHS(1, Res, EndLoc))
1915 return true;
1916
1917 // We don't Lex() the last RParen.
1918 // This is the same behavior as parseParenExpression().
1919 if (ParenDepth - 1 > 0) {
1920 EndLoc = getTok().getEndLoc();
1921 if (parseRParen())
1922 return true;
1923 }
1924 }
1925 return false;
1926}
1927
1928bool MasmParser::parseAbsoluteExpression(int64_t &Res) {
1929 const MCExpr *Expr;
1930
1931 SMLoc StartLoc = Lexer.getLoc();
1932 if (parseExpression(Expr))
1933 return true;
1934
1935 if (!Expr->evaluateAsAbsolute(Res, getStreamer().getAssemblerPtr()))
1936 return Error(StartLoc, "expected absolute expression");
1937
1938 return false;
1939}
1940
1943 bool ShouldUseLogicalShr,
1944 bool EndExpressionAtGreater) {
1945 switch (K) {
1946 default:
1947 return 0; // not a binop.
1948
1949 // Lowest Precedence: &&, ||
1950 case AsmToken::AmpAmp:
1951 Kind = MCBinaryExpr::LAnd;
1952 return 2;
1953 case AsmToken::PipePipe:
1954 Kind = MCBinaryExpr::LOr;
1955 return 1;
1956
1957 // Low Precedence: ==, !=, <>, <, <=, >, >=
1959 Kind = MCBinaryExpr::EQ;
1960 return 3;
1963 Kind = MCBinaryExpr::NE;
1964 return 3;
1965 case AsmToken::Less:
1966 Kind = MCBinaryExpr::LT;
1967 return 3;
1969 Kind = MCBinaryExpr::LTE;
1970 return 3;
1971 case AsmToken::Greater:
1972 if (EndExpressionAtGreater)
1973 return 0;
1974 Kind = MCBinaryExpr::GT;
1975 return 3;
1977 Kind = MCBinaryExpr::GTE;
1978 return 3;
1979
1980 // Low Intermediate Precedence: +, -
1981 case AsmToken::Plus:
1982 Kind = MCBinaryExpr::Add;
1983 return 4;
1984 case AsmToken::Minus:
1985 Kind = MCBinaryExpr::Sub;
1986 return 4;
1987
1988 // High Intermediate Precedence: |, &, ^
1989 case AsmToken::Pipe:
1990 Kind = MCBinaryExpr::Or;
1991 return 5;
1992 case AsmToken::Caret:
1993 Kind = MCBinaryExpr::Xor;
1994 return 5;
1995 case AsmToken::Amp:
1996 Kind = MCBinaryExpr::And;
1997 return 5;
1998
1999 // Highest Precedence: *, /, %, <<, >>
2000 case AsmToken::Star:
2001 Kind = MCBinaryExpr::Mul;
2002 return 6;
2003 case AsmToken::Slash:
2004 Kind = MCBinaryExpr::Div;
2005 return 6;
2006 case AsmToken::Percent:
2007 Kind = MCBinaryExpr::Mod;
2008 return 6;
2009 case AsmToken::LessLess:
2010 Kind = MCBinaryExpr::Shl;
2011 return 6;
2013 if (EndExpressionAtGreater)
2014 return 0;
2015 Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr;
2016 return 6;
2017 }
2018}
2019
2020unsigned MasmParser::getBinOpPrecedence(AsmToken::TokenKind K,
2021 MCBinaryExpr::Opcode &Kind) {
2022 bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr();
2023 return getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr,
2024 AngleBracketDepth > 0);
2025}
2026
2027/// Parse all binary operators with precedence >= 'Precedence'.
2028/// Res contains the LHS of the expression on input.
2029bool MasmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res,
2030 SMLoc &EndLoc) {
2031 SMLoc StartLoc = Lexer.getLoc();
2032 while (true) {
2034 if (Lexer.getKind() == AsmToken::Identifier) {
2036 .CaseLower("and", AsmToken::Amp)
2039 .CaseLower("xor", AsmToken::Caret)
2048 .Default(TokKind);
2049 }
2051 unsigned TokPrec = getBinOpPrecedence(TokKind, Kind);
2052
2053 // If the next token is lower precedence than we are allowed to eat, return
2054 // successfully with what we ate already.
2055 if (TokPrec < Precedence)
2056 return false;
2057
2058 Lex();
2059
2060 // Eat the next primary expression.
2061 const MCExpr *RHS;
2062 if (getTargetParser().parsePrimaryExpr(RHS, EndLoc))
2063 return true;
2064
2065 // If BinOp binds less tightly with RHS than the operator after RHS, let
2066 // the pending operator take RHS as its LHS.
2068 unsigned NextTokPrec = getBinOpPrecedence(Lexer.getKind(), Dummy);
2069 if (TokPrec < NextTokPrec && parseBinOpRHS(TokPrec + 1, RHS, EndLoc))
2070 return true;
2071
2072 // Merge LHS and RHS according to operator.
2073 Res = MCBinaryExpr::create(Kind, Res, RHS, getContext(), StartLoc);
2074 }
2075}
2076
2077/// ParseStatement:
2078/// ::= % statement
2079/// ::= EndOfStatement
2080/// ::= Label* Directive ...Operands... EndOfStatement
2081/// ::= Label* Identifier OperandList* EndOfStatement
2082bool MasmParser::parseStatement(ParseStatementInfo &Info,
2084 assert(!hasPendingError() && "parseStatement started with pending error");
2085 // Eat initial spaces and comments.
2086 while (Lexer.is(AsmToken::Space))
2087 Lex();
2088 if (Lexer.is(AsmToken::EndOfStatement)) {
2089 // If this is a line comment we can drop it safely.
2090 if (getTok().getString().empty() || getTok().getString().front() == '\r' ||
2091 getTok().getString().front() == '\n')
2092 Out.addBlankLine();
2093 Lex();
2094 return false;
2095 }
2096
2097 // If preceded by an expansion operator, first expand all text macros and
2098 // macro functions.
2099 if (getTok().is(AsmToken::Percent)) {
2100 SMLoc ExpansionLoc = getTok().getLoc();
2101 if (parseToken(AsmToken::Percent) || expandStatement(ExpansionLoc))
2102 return true;
2103 }
2104
2105 // Statements always start with an identifier, unless we're dealing with a
2106 // processor directive (.386, .686, etc.) that lexes as a real.
2107 AsmToken ID = getTok();
2108 SMLoc IDLoc = ID.getLoc();
2109 StringRef IDVal;
2110 if (Lexer.is(AsmToken::HashDirective))
2111 return parseCppHashLineFilenameComment(IDLoc);
2112 if (Lexer.is(AsmToken::Dot)) {
2113 // Treat '.' as a valid identifier in this context.
2114 Lex();
2115 IDVal = ".";
2116 } else if (Lexer.is(AsmToken::Real)) {
2117 // Treat ".<number>" as a valid identifier in this context.
2118 IDVal = getTok().getString();
2119 Lex(); // always eat a token
2120 if (!IDVal.startswith("."))
2121 return Error(IDLoc, "unexpected token at start of statement");
2122 } else if (parseIdentifier(IDVal, StartOfStatement)) {
2123 if (!TheCondState.Ignore) {
2124 Lex(); // always eat a token
2125 return Error(IDLoc, "unexpected token at start of statement");
2126 }
2127 IDVal = "";
2128 }
2129
2130 // Handle conditional assembly here before checking for skipping. We
2131 // have to do this so that .endif isn't skipped in a ".if 0" block for
2132 // example.
2134 DirectiveKindMap.find(IDVal.lower());
2135 DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end())
2136 ? DK_NO_DIRECTIVE
2137 : DirKindIt->getValue();
2138 switch (DirKind) {
2139 default:
2140 break;
2141 case DK_IF:
2142 case DK_IFE:
2143 return parseDirectiveIf(IDLoc, DirKind);
2144 case DK_IFB:
2145 return parseDirectiveIfb(IDLoc, true);
2146 case DK_IFNB:
2147 return parseDirectiveIfb(IDLoc, false);
2148 case DK_IFDEF:
2149 return parseDirectiveIfdef(IDLoc, true);
2150 case DK_IFNDEF:
2151 return parseDirectiveIfdef(IDLoc, false);
2152 case DK_IFDIF:
2153 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false,
2154 /*CaseInsensitive=*/false);
2155 case DK_IFDIFI:
2156 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/false,
2157 /*CaseInsensitive=*/true);
2158 case DK_IFIDN:
2159 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true,
2160 /*CaseInsensitive=*/false);
2161 case DK_IFIDNI:
2162 return parseDirectiveIfidn(IDLoc, /*ExpectEqual=*/true,
2163 /*CaseInsensitive=*/true);
2164 case DK_ELSEIF:
2165 case DK_ELSEIFE:
2166 return parseDirectiveElseIf(IDLoc, DirKind);
2167 case DK_ELSEIFB:
2168 return parseDirectiveElseIfb(IDLoc, true);
2169 case DK_ELSEIFNB:
2170 return parseDirectiveElseIfb(IDLoc, false);
2171 case DK_ELSEIFDEF:
2172 return parseDirectiveElseIfdef(IDLoc, true);
2173 case DK_ELSEIFNDEF:
2174 return parseDirectiveElseIfdef(IDLoc, false);
2175 case DK_ELSEIFDIF:
2176 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false,
2177 /*CaseInsensitive=*/false);
2178 case DK_ELSEIFDIFI:
2179 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/false,
2180 /*CaseInsensitive=*/true);
2181 case DK_ELSEIFIDN:
2182 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true,
2183 /*CaseInsensitive=*/false);
2184 case DK_ELSEIFIDNI:
2185 return parseDirectiveElseIfidn(IDLoc, /*ExpectEqual=*/true,
2186 /*CaseInsensitive=*/true);
2187 case DK_ELSE:
2188 return parseDirectiveElse(IDLoc);
2189 case DK_ENDIF:
2190 return parseDirectiveEndIf(IDLoc);
2191 }
2192
2193 // Ignore the statement if in the middle of inactive conditional
2194 // (e.g. ".if 0").
2195 if (TheCondState.Ignore) {
2196 eatToEndOfStatement();
2197 return false;
2198 }
2199
2200 // FIXME: Recurse on local labels?
2201
2202 // Check for a label.
2203 // ::= identifier ':'
2204 // ::= number ':'
2205 if (Lexer.is(AsmToken::Colon) && getTargetParser().isLabel(ID)) {
2206 if (checkForValidSection())
2207 return true;
2208
2209 // identifier ':' -> Label.
2210 Lex();
2211
2212 // Diagnose attempt to use '.' as a label.
2213 if (IDVal == ".")
2214 return Error(IDLoc, "invalid use of pseudo-symbol '.' as a label");
2215
2216 // Diagnose attempt to use a variable as a label.
2217 //
2218 // FIXME: Diagnostics. Note the location of the definition as a label.
2219 // FIXME: This doesn't diagnose assignment to a symbol which has been
2220 // implicitly marked as external.
2221 MCSymbol *Sym;
2222 if (ParsingMSInlineAsm && SI) {
2223 StringRef RewrittenLabel =
2224 SI->LookupInlineAsmLabel(IDVal, getSourceManager(), IDLoc, true);
2225 assert(!RewrittenLabel.empty() &&
2226 "We should have an internal name here.");
2227 Info.AsmRewrites->emplace_back(AOK_Label, IDLoc, IDVal.size(),
2228 RewrittenLabel);
2229 IDVal = RewrittenLabel;
2230 }
2231 // Handle directional local labels
2232 if (IDVal == "@@") {
2234 } else {
2235 Sym = getContext().getOrCreateSymbol(IDVal);
2236 }
2237
2238 // End of Labels should be treated as end of line for lexing
2239 // purposes but that information is not available to the Lexer who
2240 // does not understand Labels. This may cause us to see a Hash
2241 // here instead of a preprocessor line comment.
2242 if (getTok().is(AsmToken::Hash)) {
2243 std::string CommentStr = parseStringTo(AsmToken::EndOfStatement);
2244 Lexer.Lex();
2245 Lexer.UnLex(AsmToken(AsmToken::EndOfStatement, CommentStr));
2246 }
2247
2248 // Consume any end of statement token, if present, to avoid spurious
2249 // addBlankLine calls().
2250 if (getTok().is(AsmToken::EndOfStatement)) {
2251 Lex();
2252 }
2253
2254 getTargetParser().doBeforeLabelEmit(Sym, IDLoc);
2255
2256 // Emit the label.
2257 if (!getTargetParser().isParsingMSInlineAsm())
2258 Out.emitLabel(Sym, IDLoc);
2259
2260 // If we are generating dwarf for assembly source files then gather the
2261 // info to make a dwarf label entry for this label if needed.
2262 if (enabledGenDwarfForAssembly())
2263 MCGenDwarfLabelEntry::Make(Sym, &getStreamer(), getSourceManager(),
2264 IDLoc);
2265
2266 getTargetParser().onLabelParsed(Sym);
2267
2268 return false;
2269 }
2270
2271 // If macros are enabled, check to see if this is a macro instantiation.
2272 if (const MCAsmMacro *M = getContext().lookupMacro(IDVal.lower())) {
2273 return handleMacroEntry(M, IDLoc);
2274 }
2275
2276 // Otherwise, we have a normal instruction or directive.
2277
2278 if (DirKind != DK_NO_DIRECTIVE) {
2279 // There are several entities interested in parsing directives:
2280 //
2281 // 1. Asm parser extensions. For example, platform-specific parsers
2282 // (like the ELF parser) register themselves as extensions.
2283 // 2. The target-specific assembly parser. Some directives are target
2284 // specific or may potentially behave differently on certain targets.
2285 // 3. The generic directive parser implemented by this class. These are
2286 // all the directives that behave in a target and platform independent
2287 // manner, or at least have a default behavior that's shared between
2288 // all targets and platforms.
2289
2290 getTargetParser().flushPendingInstructions(getStreamer());
2291
2292 // Special-case handling of structure-end directives at higher priority,
2293 // since ENDS is overloaded as a segment-end directive.
2294 if (IDVal.equals_insensitive("ends") && StructInProgress.size() > 1 &&
2295 getTok().is(AsmToken::EndOfStatement)) {
2296 return parseDirectiveNestedEnds();
2297 }
2298
2299 // First, check the extension directive map to see if any extension has
2300 // registered itself to parse this directive.
2301 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
2302 ExtensionDirectiveMap.lookup(IDVal.lower());
2303 if (Handler.first)
2304 return (*Handler.second)(Handler.first, IDVal, IDLoc);
2305
2306 // Next, let the target-specific assembly parser try.
2307 if (ID.isNot(AsmToken::Identifier))
2308 return false;
2309
2310 ParseStatus TPDirectiveReturn = getTargetParser().parseDirective(ID);
2311 assert(TPDirectiveReturn.isFailure() == hasPendingError() &&
2312 "Should only return Failure iff there was an error");
2313 if (TPDirectiveReturn.isFailure())
2314 return true;
2315 if (TPDirectiveReturn.isSuccess())
2316 return false;
2317
2318 // Finally, if no one else is interested in this directive, it must be
2319 // generic and familiar to this class.
2320 switch (DirKind) {
2321 default:
2322 break;
2323 case DK_ASCII:
2324 return parseDirectiveAscii(IDVal, false);
2325 case DK_ASCIZ:
2326 case DK_STRING:
2327 return parseDirectiveAscii(IDVal, true);
2328 case DK_BYTE:
2329 case DK_SBYTE:
2330 case DK_DB:
2331 return parseDirectiveValue(IDVal, 1);
2332 case DK_WORD:
2333 case DK_SWORD:
2334 case DK_DW:
2335 return parseDirectiveValue(IDVal, 2);
2336 case DK_DWORD:
2337 case DK_SDWORD:
2338 case DK_DD:
2339 return parseDirectiveValue(IDVal, 4);
2340 case DK_FWORD:
2341 case DK_DF:
2342 return parseDirectiveValue(IDVal, 6);
2343 case DK_QWORD:
2344 case DK_SQWORD:
2345 case DK_DQ:
2346 return parseDirectiveValue(IDVal, 8);
2347 case DK_REAL4:
2348 return parseDirectiveRealValue(IDVal, APFloat::IEEEsingle(), 4);
2349 case DK_REAL8:
2350 return parseDirectiveRealValue(IDVal, APFloat::IEEEdouble(), 8);
2351 case DK_REAL10:
2352 return parseDirectiveRealValue(IDVal, APFloat::x87DoubleExtended(), 10);
2353 case DK_STRUCT:
2354 case DK_UNION:
2355 return parseDirectiveNestedStruct(IDVal, DirKind);
2356 case DK_ENDS:
2357 return parseDirectiveNestedEnds();
2358 case DK_ALIGN:
2359 return parseDirectiveAlign();
2360 case DK_EVEN:
2361 return parseDirectiveEven();
2362 case DK_ORG:
2363 return parseDirectiveOrg();
2364 case DK_EXTERN:
2365 return parseDirectiveExtern();
2366 case DK_PUBLIC:
2367 return parseDirectiveSymbolAttribute(MCSA_Global);
2368 case DK_COMM:
2369 return parseDirectiveComm(/*IsLocal=*/false);
2370 case DK_COMMENT:
2371 return parseDirectiveComment(IDLoc);
2372 case DK_INCLUDE:
2373 return parseDirectiveInclude();
2374 case DK_REPEAT:
2375 return parseDirectiveRepeat(IDLoc, IDVal);
2376 case DK_WHILE:
2377 return parseDirectiveWhile(IDLoc);
2378 case DK_FOR:
2379 return parseDirectiveFor(IDLoc, IDVal);
2380 case DK_FORC:
2381 return parseDirectiveForc(IDLoc, IDVal);
2382 case DK_FILE:
2383 return parseDirectiveFile(IDLoc);
2384 case DK_LINE:
2385 return parseDirectiveLine();
2386 case DK_LOC:
2387 return parseDirectiveLoc();
2388 case DK_STABS:
2389 return parseDirectiveStabs();
2390 case DK_CV_FILE:
2391 return parseDirectiveCVFile();
2392 case DK_CV_FUNC_ID:
2393 return parseDirectiveCVFuncId();
2394 case DK_CV_INLINE_SITE_ID:
2395 return parseDirectiveCVInlineSiteId();
2396 case DK_CV_LOC:
2397 return parseDirectiveCVLoc();
2398 case DK_CV_LINETABLE:
2399 return parseDirectiveCVLinetable();
2400 case DK_CV_INLINE_LINETABLE:
2401 return parseDirectiveCVInlineLinetable();
2402 case DK_CV_DEF_RANGE:
2403 return parseDirectiveCVDefRange();
2404 case DK_CV_STRING:
2405 return parseDirectiveCVString();
2406 case DK_CV_STRINGTABLE:
2407 return parseDirectiveCVStringTable();
2408 case DK_CV_FILECHECKSUMS:
2409 return parseDirectiveCVFileChecksums();
2410 case DK_CV_FILECHECKSUM_OFFSET:
2411 return parseDirectiveCVFileChecksumOffset();
2412 case DK_CV_FPO_DATA:
2413 return parseDirectiveCVFPOData();
2414 case DK_CFI_SECTIONS:
2415 return parseDirectiveCFISections();
2416 case DK_CFI_STARTPROC:
2417 return parseDirectiveCFIStartProc();
2418 case DK_CFI_ENDPROC:
2419 return parseDirectiveCFIEndProc();
2420 case DK_CFI_DEF_CFA:
2421 return parseDirectiveCFIDefCfa(IDLoc);
2422 case DK_CFI_DEF_CFA_OFFSET:
2423 return parseDirectiveCFIDefCfaOffset(IDLoc);
2424 case DK_CFI_ADJUST_CFA_OFFSET:
2425 return parseDirectiveCFIAdjustCfaOffset(IDLoc);
2426 case DK_CFI_DEF_CFA_REGISTER:
2427 return parseDirectiveCFIDefCfaRegister(IDLoc);
2428 case DK_CFI_OFFSET:
2429 return parseDirectiveCFIOffset(IDLoc);
2430 case DK_CFI_REL_OFFSET:
2431 return parseDirectiveCFIRelOffset(IDLoc);
2432 case DK_CFI_PERSONALITY:
2433 return parseDirectiveCFIPersonalityOrLsda(true);
2434 case DK_CFI_LSDA:
2435 return parseDirectiveCFIPersonalityOrLsda(false);
2436 case DK_CFI_REMEMBER_STATE:
2437 return parseDirectiveCFIRememberState(IDLoc);
2438 case DK_CFI_RESTORE_STATE:
2439 return parseDirectiveCFIRestoreState(IDLoc);
2440 case DK_CFI_SAME_VALUE:
2441 return parseDirectiveCFISameValue(IDLoc);
2442 case DK_CFI_RESTORE:
2443 return parseDirectiveCFIRestore(IDLoc);
2444 case DK_CFI_ESCAPE:
2445 return parseDirectiveCFIEscape(IDLoc);
2446 case DK_CFI_RETURN_COLUMN:
2447 return parseDirectiveCFIReturnColumn(IDLoc);
2448 case DK_CFI_SIGNAL_FRAME:
2449 return parseDirectiveCFISignalFrame();
2450 case DK_CFI_UNDEFINED:
2451 return parseDirectiveCFIUndefined(IDLoc);
2452 case DK_CFI_REGISTER:
2453 return parseDirectiveCFIRegister(IDLoc);
2454 case DK_CFI_WINDOW_SAVE:
2455 return parseDirectiveCFIWindowSave(IDLoc);
2456 case DK_EXITM:
2457 Info.ExitValue = "";
2458 return parseDirectiveExitMacro(IDLoc, IDVal, *Info.ExitValue);
2459 case DK_ENDM:
2460 Info.ExitValue = "";
2461 return parseDirectiveEndMacro(IDVal);
2462 case DK_PURGE:
2463 return parseDirectivePurgeMacro(IDLoc);
2464 case DK_END:
2465 return parseDirectiveEnd(IDLoc);
2466 case DK_ERR:
2467 return parseDirectiveError(IDLoc);
2468 case DK_ERRB:
2469 return parseDirectiveErrorIfb(IDLoc, true);
2470 case DK_ERRNB:
2471 return parseDirectiveErrorIfb(IDLoc, false);
2472 case DK_ERRDEF:
2473 return parseDirectiveErrorIfdef(IDLoc, true);
2474 case DK_ERRNDEF:
2475 return parseDirectiveErrorIfdef(IDLoc, false);
2476 case DK_ERRDIF:
2477 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false,
2478 /*CaseInsensitive=*/false);
2479 case DK_ERRDIFI:
2480 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/false,
2481 /*CaseInsensitive=*/true);
2482 case DK_ERRIDN:
2483 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true,
2484 /*CaseInsensitive=*/false);
2485 case DK_ERRIDNI:
2486 return parseDirectiveErrorIfidn(IDLoc, /*ExpectEqual=*/true,
2487 /*CaseInsensitive=*/true);
2488 case DK_ERRE:
2489 return parseDirectiveErrorIfe(IDLoc, true);
2490 case DK_ERRNZ:
2491 return parseDirectiveErrorIfe(IDLoc, false);
2492 case DK_RADIX:
2493 return parseDirectiveRadix(IDLoc);
2494 case DK_ECHO:
2495 return parseDirectiveEcho(IDLoc);
2496 }
2497
2498 return Error(IDLoc, "unknown directive");
2499 }
2500
2501 // We also check if this is allocating memory with user-defined type.
2502 auto IDIt = Structs.find(IDVal.lower());
2503 if (IDIt != Structs.end())
2504 return parseDirectiveStructValue(/*Structure=*/IDIt->getValue(), IDVal,
2505 IDLoc);
2506
2507 // Non-conditional Microsoft directives sometimes follow their first argument.
2508 const AsmToken nextTok = getTok();
2509 const StringRef nextVal = nextTok.getString();
2510 const SMLoc nextLoc = nextTok.getLoc();
2511
2512 const AsmToken afterNextTok = peekTok();
2513
2514 // There are several entities interested in parsing infix directives:
2515 //
2516 // 1. Asm parser extensions. For example, platform-specific parsers
2517 // (like the ELF parser) register themselves as extensions.
2518 // 2. The generic directive parser implemented by this class. These are
2519 // all the directives that behave in a target and platform independent
2520 // manner, or at least have a default behavior that's shared between
2521 // all targets and platforms.
2522
2523 getTargetParser().flushPendingInstructions(getStreamer());
2524
2525 // Special-case handling of structure-end directives at higher priority, since
2526 // ENDS is overloaded as a segment-end directive.
2527 if (nextVal.equals_insensitive("ends") && StructInProgress.size() == 1) {
2528 Lex();
2529 return parseDirectiveEnds(IDVal, IDLoc);
2530 }
2531
2532 // First, check the extension directive map to see if any extension has
2533 // registered itself to parse this directive.
2534 std::pair<MCAsmParserExtension *, DirectiveHandler> Handler =
2535 ExtensionDirectiveMap.lookup(nextVal.lower());
2536 if (Handler.first) {
2537 Lex();
2538 Lexer.UnLex(ID);
2539 return (*Handler.second)(Handler.first, nextVal, nextLoc);
2540 }
2541
2542 // If no one else is interested in this directive, it must be
2543 // generic and familiar to this class.
2544 DirKindIt = DirectiveKindMap.find(nextVal.lower());
2545 DirKind = (DirKindIt == DirectiveKindMap.end())
2546 ? DK_NO_DIRECTIVE
2547 : DirKindIt->getValue();
2548 switch (DirKind) {
2549 default:
2550 break;
2551 case DK_ASSIGN:
2552 case DK_EQU:
2553 case DK_TEXTEQU:
2554 Lex();
2555 return parseDirectiveEquate(nextVal, IDVal, DirKind, IDLoc);
2556 case DK_BYTE:
2557 if (afterNextTok.is(AsmToken::Identifier) &&
2558 afterNextTok.getString().equals_insensitive("ptr")) {
2559 // Size directive; part of an instruction.
2560 break;
2561 }
2562 [[fallthrough]];
2563 case DK_SBYTE:
2564 case DK_DB:
2565 Lex();
2566 return parseDirectiveNamedValue(nextVal, 1, IDVal, IDLoc);
2567 case DK_WORD:
2568 if (afterNextTok.is(AsmToken::Identifier) &&
2569 afterNextTok.getString().equals_insensitive("ptr")) {
2570 // Size directive; part of an instruction.
2571 break;
2572 }
2573 [[fallthrough]];
2574 case DK_SWORD:
2575 case DK_DW:
2576 Lex();
2577 return parseDirectiveNamedValue(nextVal, 2, IDVal, IDLoc);
2578 case DK_DWORD:
2579 if (afterNextTok.is(AsmToken::Identifier) &&
2580 afterNextTok.getString().equals_insensitive("ptr")) {
2581 // Size directive; part of an instruction.
2582 break;
2583 }
2584 [[fallthrough]];
2585 case DK_SDWORD:
2586 case DK_DD:
2587 Lex();
2588 return parseDirectiveNamedValue(nextVal, 4, IDVal, IDLoc);
2589 case DK_FWORD:
2590 if (afterNextTok.is(AsmToken::Identifier) &&
2591 afterNextTok.getString().equals_insensitive("ptr")) {
2592 // Size directive; part of an instruction.
2593 break;
2594 }
2595 [[fallthrough]];
2596 case DK_DF:
2597 Lex();
2598 return parseDirectiveNamedValue(nextVal, 6, IDVal, IDLoc);
2599 case DK_QWORD:
2600 if (afterNextTok.is(AsmToken::Identifier) &&
2601 afterNextTok.getString().equals_insensitive("ptr")) {
2602 // Size directive; part of an instruction.
2603 break;
2604 }
2605 [[fallthrough]];
2606 case DK_SQWORD:
2607 case DK_DQ:
2608 Lex();
2609 return parseDirectiveNamedValue(nextVal, 8, IDVal, IDLoc);
2610 case DK_REAL4:
2611 Lex();
2612 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEsingle(), 4,
2613 IDVal, IDLoc);
2614 case DK_REAL8:
2615 Lex();
2616 return parseDirectiveNamedRealValue(nextVal, APFloat::IEEEdouble(), 8,
2617 IDVal, IDLoc);
2618 case DK_REAL10:
2619 Lex();
2620 return parseDirectiveNamedRealValue(nextVal, APFloat::x87DoubleExtended(),
2621 10, IDVal, IDLoc);
2622 case DK_STRUCT:
2623 case DK_UNION:
2624 Lex();
2625 return parseDirectiveStruct(nextVal, DirKind, IDVal, IDLoc);
2626 case DK_ENDS:
2627 Lex();
2628 return parseDirectiveEnds(IDVal, IDLoc);
2629 case DK_MACRO:
2630 Lex();
2631 return parseDirectiveMacro(IDVal, IDLoc);
2632 }
2633
2634 // Finally, we check if this is allocating a variable with user-defined type.
2635 auto NextIt = Structs.find(nextVal.lower());
2636 if (NextIt != Structs.end()) {
2637 Lex();
2638 return parseDirectiveNamedStructValue(/*Structure=*/NextIt->getValue(),
2639 nextVal, nextLoc, IDVal);
2640 }
2641
2642 // __asm _emit or __asm __emit
2643 if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" ||
2644 IDVal == "_EMIT" || IDVal == "__EMIT"))
2645 return parseDirectiveMSEmit(IDLoc, Info, IDVal.size());
2646
2647 // __asm align
2648 if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN"))
2649 return parseDirectiveMSAlign(IDLoc, Info);
2650
2651 if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN"))
2652 Info.AsmRewrites->emplace_back(AOK_EVEN, IDLoc, 4);
2653 if (checkForValidSection())
2654 return true;
2655
2656 // Canonicalize the opcode to lower case.
2657 std::string OpcodeStr = IDVal.lower();
2658 ParseInstructionInfo IInfo(Info.AsmRewrites);
2659 bool ParseHadError = getTargetParser().ParseInstruction(IInfo, OpcodeStr, ID,
2660 Info.ParsedOperands);
2661 Info.ParseError = ParseHadError;
2662
2663 // Dump the parsed representation, if requested.
2664 if (getShowParsedOperands()) {
2665 SmallString<256> Str;
2667 OS << "parsed instruction: [";
2668 for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) {
2669 if (i != 0)
2670 OS << ", ";
2671 Info.ParsedOperands[i]->print(OS);
2672 }
2673 OS << "]";
2674
2675 printMessage(IDLoc, SourceMgr::DK_Note, OS.str());
2676 }
2677
2678 // Fail even if ParseInstruction erroneously returns false.
2679 if (hasPendingError() || ParseHadError)
2680 return true;
2681
2682 // If we are generating dwarf for the current section then generate a .loc
2683 // directive for the instruction.
2684 if (!ParseHadError && enabledGenDwarfForAssembly() &&
2685 getContext().getGenDwarfSectionSyms().count(
2686 getStreamer().getCurrentSectionOnly())) {
2687 unsigned Line;
2688 if (ActiveMacros.empty())
2689 Line = SrcMgr.FindLineNumber(IDLoc, CurBuffer);
2690 else
2691 Line = SrcMgr.FindLineNumber(ActiveMacros.front()->InstantiationLoc,
2692 ActiveMacros.front()->ExitBuffer);
2693
2694 // If we previously parsed a cpp hash file line comment then make sure the
2695 // current Dwarf File is for the CppHashFilename if not then emit the
2696 // Dwarf File table for it and adjust the line number for the .loc.
2697 if (!CppHashInfo.Filename.empty()) {
2698 unsigned FileNumber = getStreamer().emitDwarfFileDirective(
2699 0, StringRef(), CppHashInfo.Filename);
2700 getContext().setGenDwarfFileNumber(FileNumber);
2701
2702 unsigned CppHashLocLineNo =
2703 SrcMgr.FindLineNumber(CppHashInfo.Loc, CppHashInfo.Buf);
2704 Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo);
2705 }
2706
2707 getStreamer().emitDwarfLocDirective(
2708 getContext().getGenDwarfFileNumber(), Line, 0,
2710 StringRef());
2711 }
2712
2713 // If parsing succeeded, match the instruction.
2714 if (!ParseHadError) {
2716 if (getTargetParser().MatchAndEmitInstruction(
2717 IDLoc, Info.Opcode, Info.ParsedOperands, Out, ErrorInfo,
2718 getTargetParser().isParsingMSInlineAsm()))
2719 return true;
2720 }
2721 return false;
2722}
2723
2724// Parse and erase curly braces marking block start/end.
2725bool MasmParser::parseCurlyBlockScope(
2726 SmallVectorImpl<AsmRewrite> &AsmStrRewrites) {
2727 // Identify curly brace marking block start/end.
2728 if (Lexer.isNot(AsmToken::LCurly) && Lexer.isNot(AsmToken::RCurly))
2729 return false;
2730
2731 SMLoc StartLoc = Lexer.getLoc();
2732 Lex(); // Eat the brace.
2733 if (Lexer.is(AsmToken::EndOfStatement))
2734 Lex(); // Eat EndOfStatement following the brace.
2735
2736 // Erase the block start/end brace from the output asm string.
2737 AsmStrRewrites.emplace_back(AOK_Skip, StartLoc, Lexer.getLoc().getPointer() -
2738 StartLoc.getPointer());
2739 return true;
2740}
2741
2742/// parseCppHashLineFilenameComment as this:
2743/// ::= # number "filename"
2744bool MasmParser::parseCppHashLineFilenameComment(SMLoc L) {
2745 Lex(); // Eat the hash token.
2746 // Lexer only ever emits HashDirective if it fully formed if it's
2747 // done the checking already so this is an internal error.
2748 assert(getTok().is(AsmToken::Integer) &&
2749 "Lexing Cpp line comment: Expected Integer");
2750 int64_t LineNumber = getTok().getIntVal();
2751 Lex();
2752 assert(getTok().is(AsmToken::String) &&
2753 "Lexing Cpp line comment: Expected String");
2754 StringRef Filename = getTok().getString();
2755 Lex();
2756
2757 // Get rid of the enclosing quotes.
2758 Filename = Filename.substr(1, Filename.size() - 2);
2759
2760 // Save the SMLoc, Filename and LineNumber for later use by diagnostics
2761 // and possibly DWARF file info.
2762 CppHashInfo.Loc = L;
2763 CppHashInfo.Filename = Filename;
2764 CppHashInfo.LineNumber = LineNumber;
2765 CppHashInfo.Buf = CurBuffer;
2766 if (FirstCppHashFilename.empty())
2767 FirstCppHashFilename = Filename;
2768 return false;
2769}
2770
2771/// will use the last parsed cpp hash line filename comment
2772/// for the Filename and LineNo if any in the diagnostic.
2773void MasmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) {
2774 const MasmParser *Parser = static_cast<const MasmParser *>(Context);
2775 raw_ostream &OS = errs();
2776
2777 const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr();
2778 SMLoc DiagLoc = Diag.getLoc();
2779 unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2780 unsigned CppHashBuf =
2781 Parser->SrcMgr.FindBufferContainingLoc(Parser->CppHashInfo.Loc);
2782
2783 // Like SourceMgr::printMessage() we need to print the include stack if any
2784 // before printing the message.
2785 unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(DiagLoc);
2786 if (!Parser->SavedDiagHandler && DiagCurBuffer &&
2787 DiagCurBuffer != DiagSrcMgr.getMainFileID()) {
2788 SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(DiagCurBuffer);
2789 DiagSrcMgr.PrintIncludeStack(ParentIncludeLoc, OS);
2790 }
2791
2792 // If we have not parsed a cpp hash line filename comment or the source
2793 // manager changed or buffer changed (like in a nested include) then just
2794 // print the normal diagnostic using its Filename and LineNo.
2795 if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr ||
2796 DiagBuf != CppHashBuf) {
2797 if (Parser->SavedDiagHandler)
2798 Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext);
2799 else
2800 Diag.print(nullptr, OS);
2801 return;
2802 }
2803
2804 // Use the CppHashFilename and calculate a line number based on the
2805 // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc
2806 // for the diagnostic.
2807 const std::string &Filename = std::string(Parser->CppHashInfo.Filename);
2808
2809 int DiagLocLineNo = DiagSrcMgr.FindLineNumber(DiagLoc, DiagBuf);
2810 int CppHashLocLineNo =
2811 Parser->SrcMgr.FindLineNumber(Parser->CppHashInfo.Loc, CppHashBuf);
2812 int LineNo =
2813 Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo);
2814
2815 SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo,
2816 Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(),
2817 Diag.getLineContents(), Diag.getRanges());
2818
2819 if (Parser->SavedDiagHandler)
2820 Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext);
2821 else
2822 NewDiag.print(nullptr, OS);
2823}
2824
2825// This is similar to the IsIdentifierChar function in AsmLexer.cpp, but does
2826// not accept '.'.
2827static bool isMacroParameterChar(char C) {
2828 return isAlnum(C) || C == '_' || C == '$' || C == '@' || C == '?';
2829}
2830
2831bool MasmParser::expandMacro(raw_svector_ostream &OS, StringRef Body,
2834 const std::vector<std::string> &Locals, SMLoc L) {
2835 unsigned NParameters = Parameters.size();
2836 if (NParameters != A.size())
2837 return Error(L, "Wrong number of arguments");
2838 StringMap<std::string> LocalSymbols;
2839 std::string Name;
2840 Name.reserve(6);
2841 for (StringRef Local : Locals) {
2843 LocalName << "??"
2844 << format_hex_no_prefix(LocalCounter++, 4, /*Upper=*/true);
2845 LocalSymbols.insert({Local, LocalName.str()});
2846 Name.clear();
2847 }
2848
2849 std::optional<char> CurrentQuote;
2850 while (!Body.empty()) {
2851 // Scan for the next substitution.
2852 std::size_t End = Body.size(), Pos = 0;
2853 std::size_t IdentifierPos = End;
2854 for (; Pos != End; ++Pos) {
2855 // Find the next possible macro parameter, including preceding a '&'
2856 // inside quotes.
2857 if (Body[Pos] == '&')
2858 break;
2859 if (isMacroParameterChar(Body[Pos])) {
2860 if (!CurrentQuote)
2861 break;
2862 if (IdentifierPos == End)
2863 IdentifierPos = Pos;
2864 } else {
2865 IdentifierPos = End;
2866 }
2867
2868 // Track quotation status
2869 if (!CurrentQuote) {
2870 if (Body[Pos] == '\'' || Body[Pos] == '"')
2871 CurrentQuote = Body[Pos];
2872 } else if (Body[Pos] == CurrentQuote) {
2873 if (Pos + 1 != End && Body[Pos + 1] == CurrentQuote) {
2874 // Escaped quote, and quotes aren't identifier chars; skip
2875 ++Pos;
2876 continue;
2877 } else {
2878 CurrentQuote.reset();
2879 }
2880 }
2881 }
2882 if (IdentifierPos != End) {
2883 // We've recognized an identifier before an apostrophe inside quotes;
2884 // check once to see if we can expand it.
2885 Pos = IdentifierPos;
2886 IdentifierPos = End;
2887 }
2888
2889 // Add the prefix.
2890 OS << Body.slice(0, Pos);
2891
2892 // Check if we reached the end.
2893 if (Pos == End)
2894 break;
2895
2896 unsigned I = Pos;
2897 bool InitialAmpersand = (Body[I] == '&');
2898 if (InitialAmpersand) {
2899 ++I;
2900 ++Pos;
2901 }
2902 while (I < End && isMacroParameterChar(Body[I]))
2903 ++I;
2904
2905 const char *Begin = Body.data() + Pos;
2906 StringRef Argument(Begin, I - Pos);
2907 const std::string ArgumentLower = Argument.lower();
2908 unsigned Index = 0;
2909
2910 for (; Index < NParameters; ++Index)
2911 if (Parameters[Index].Name.equals_insensitive(ArgumentLower))
2912 break;
2913
2914 if (Index == NParameters) {
2915 if (InitialAmpersand)
2916 OS << '&';
2917 auto it = LocalSymbols.find(ArgumentLower);
2918 if (it != LocalSymbols.end())
2919 OS << it->second;
2920 else
2921 OS << Argument;
2922 Pos = I;
2923 } else {
2924 for (const AsmToken &Token : A[Index]) {
2925 // In MASM, you can write '%expr'.
2926 // The prefix '%' evaluates the expression 'expr'
2927 // and uses the result as a string (e.g. replace %(1+2) with the
2928 // string "3").
2929 // Here, we identify the integer token which is the result of the
2930 // absolute expression evaluation and replace it with its string
2931 // representation.
2932 if (Token.getString().front() == '%' && Token.is(AsmToken::Integer))
2933 // Emit an integer value to the buffer.
2934 OS << Token.getIntVal();
2935 else
2936 OS << Token.getString();
2937 }
2938
2939 Pos += Argument.size();
2940 if (Pos < End && Body[Pos] == '&') {
2941 ++Pos;
2942 }
2943 }
2944 // Update the scan point.
2945 Body = Body.substr(Pos);
2946 }
2947
2948 return false;
2949}
2950
2952 switch (kind) {
2953 default:
2954 return false;
2955 case AsmToken::Plus:
2956 case AsmToken::Minus:
2957 case AsmToken::Tilde:
2958 case AsmToken::Slash:
2959 case AsmToken::Star:
2960 case AsmToken::Dot:
2961 case AsmToken::Equal:
2963 case AsmToken::Pipe:
2964 case AsmToken::PipePipe:
2965 case AsmToken::Caret:
2966 case AsmToken::Amp:
2967 case AsmToken::AmpAmp:
2968 case AsmToken::Exclaim:
2970 case AsmToken::Less:
2972 case AsmToken::LessLess:
2974 case AsmToken::Greater:
2977 return true;
2978 }
2979}
2980
2981namespace {
2982
2983class AsmLexerSkipSpaceRAII {
2984public:
2985 AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) {
2986 Lexer.setSkipSpace(SkipSpace);
2987 }
2988
2989 ~AsmLexerSkipSpaceRAII() {
2990 Lexer.setSkipSpace(true);
2991 }
2992
2993private:
2994 AsmLexer &Lexer;
2995};
2996
2997} // end anonymous namespace
2998
2999bool MasmParser::parseMacroArgument(const MCAsmMacroParameter *MP,
3000 MCAsmMacroArgument &MA,
3001 AsmToken::TokenKind EndTok) {
3002 if (MP && MP->Vararg) {
3003 if (Lexer.isNot(EndTok)) {
3004 SmallVector<StringRef, 1> Str = parseStringRefsTo(EndTok);
3005 for (StringRef S : Str) {
3006 MA.emplace_back(AsmToken::String, S);
3007 }
3008 }
3009 return false;
3010 }
3011
3012 SMLoc StrLoc = Lexer.getLoc(), EndLoc;
3013 if (Lexer.is(AsmToken::Less) && isAngleBracketString(StrLoc, EndLoc)) {
3014 const char *StrChar = StrLoc.getPointer() + 1;
3015 const char *EndChar = EndLoc.getPointer() - 1;
3016 jumpToLoc(EndLoc, CurBuffer, EndStatementAtEOFStack.back());
3017 /// Eat from '<' to '>'.
3018 Lex();
3019 MA.emplace_back(AsmToken::String, StringRef(StrChar, EndChar - StrChar));
3020 return false;
3021 }
3022
3023 unsigned ParenLevel = 0;
3024
3025 // Darwin doesn't use spaces to delmit arguments.
3026 AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin);
3027
3028 bool SpaceEaten;
3029
3030 while (true) {
3031 SpaceEaten = false;
3032 if (Lexer.is(AsmToken::Eof) || Lexer.is(AsmToken::Equal))
3033 return TokError("unexpected token");
3034
3035 if (ParenLevel == 0) {
3036 if (Lexer.is(AsmToken::Comma))
3037 break;
3038
3039 if (Lexer.is(AsmToken::Space)) {
3040 SpaceEaten = true;
3041 Lex(); // Eat spaces.
3042 }
3043
3044 // Spaces can delimit parameters, but could also be part an expression.
3045 // If the token after a space is an operator, add the token and the next
3046 // one into this argument
3047 if (!IsDarwin) {
3048 if (isOperator(Lexer.getKind()) && Lexer.isNot(EndTok)) {
3049 MA.push_back(getTok());
3050 Lex();
3051
3052 // Whitespace after an operator can be ignored.
3053 if (Lexer.is(AsmToken::Space))
3054 Lex();
3055
3056 continue;
3057 }
3058 }
3059 if (SpaceEaten)
3060 break;
3061 }
3062
3063 // handleMacroEntry relies on not advancing the lexer here
3064 // to be able to fill in the remaining default parameter values
3065 if (Lexer.is(EndTok) && (EndTok != AsmToken::RParen || ParenLevel == 0))
3066 break;
3067
3068 // Adjust the current parentheses level.
3069 if (Lexer.is(AsmToken::LParen))
3070 ++ParenLevel;
3071 else if (Lexer.is(AsmToken::RParen) && ParenLevel)
3072 --ParenLevel;
3073
3074 // Append the token to the current argument list.
3075 MA.push_back(getTok());
3076 Lex();
3077 }
3078
3079 if (ParenLevel != 0)
3080 return TokError("unbalanced parentheses in argument");
3081
3082 if (MA.empty() && MP) {
3083 if (MP->Required) {
3084 return TokError("missing value for required parameter '" + MP->Name +
3085 "'");
3086 } else {
3087 MA = MP->Value;
3088 }
3089 }
3090 return false;
3091}
3092
3093// Parse the macro instantiation arguments.
3094bool MasmParser::parseMacroArguments(const MCAsmMacro *M,
3095 MCAsmMacroArguments &A,
3096 AsmToken::TokenKind EndTok) {
3097 const unsigned NParameters = M ? M->Parameters.size() : 0;
3098 bool NamedParametersFound = false;
3099 SmallVector<SMLoc, 4> FALocs;
3100
3101 A.resize(NParameters);
3102 FALocs.resize(NParameters);
3103
3104 // Parse two kinds of macro invocations:
3105 // - macros defined without any parameters accept an arbitrary number of them
3106 // - macros defined with parameters accept at most that many of them
3107 for (unsigned Parameter = 0; !NParameters || Parameter < NParameters;
3108 ++Parameter) {
3109 SMLoc IDLoc = Lexer.getLoc();
3111
3112 if (Lexer.is(AsmToken::Identifier) && peekTok().is(AsmToken::Equal)) {
3113 if (parseIdentifier(FA.Name))
3114 return Error(IDLoc, "invalid argument identifier for formal argument");
3115
3116 if (Lexer.isNot(AsmToken::Equal))
3117 return TokError("expected '=' after formal parameter identifier");
3118
3119 Lex();
3120
3121 NamedParametersFound = true;
3122 }
3123
3124 if (NamedParametersFound && FA.Name.empty())
3125 return Error(IDLoc, "cannot mix positional and keyword arguments");
3126
3127 unsigned PI = Parameter;
3128 if (!FA.Name.empty()) {
3129 assert(M && "expected macro to be defined");
3130 unsigned FAI = 0;
3131 for (FAI = 0; FAI < NParameters; ++FAI)
3132 if (M->Parameters[FAI].Name == FA.Name)
3133 break;
3134
3135 if (FAI >= NParameters) {
3136 return Error(IDLoc, "parameter named '" + FA.Name +
3137 "' does not exist for macro '" + M->Name + "'");
3138 }
3139 PI = FAI;
3140 }
3141 const MCAsmMacroParameter *MP = nullptr;
3142 if (M && PI < NParameters)
3143 MP = &M->Parameters[PI];
3144
3145 SMLoc StrLoc = Lexer.getLoc();
3146 SMLoc EndLoc;
3147 if (Lexer.is(AsmToken::Percent)) {
3148 const MCExpr *AbsoluteExp;
3149 int64_t Value;
3150 /// Eat '%'.
3151 Lex();
3152 if (parseExpression(AbsoluteExp, EndLoc))
3153 return false;
3154 if (!AbsoluteExp->evaluateAsAbsolute(Value,
3155 getStreamer().getAssemblerPtr()))
3156 return Error(StrLoc, "expected absolute expression");
3157 const char *StrChar = StrLoc.getPointer();
3158 const char *EndChar = EndLoc.getPointer();
3159 AsmToken newToken(AsmToken::Integer,
3160 StringRef(StrChar, EndChar - StrChar), Value);
3161 FA.Value.push_back(newToken);
3162 } else if (parseMacroArgument(MP, FA.Value, EndTok)) {
3163 if (M)
3164 return addErrorSuffix(" in '" + M->Name + "' macro");
3165 else
3166 return true;
3167 }
3168
3169 if (!FA.Value.empty()) {
3170 if (A.size() <= PI)
3171 A.resize(PI + 1);
3172 A[PI] = FA.Value;
3173
3174 if (FALocs.size() <= PI)
3175 FALocs.resize(PI + 1);
3176
3177 FALocs[PI] = Lexer.getLoc();
3178 }
3179
3180 // At the end of the statement, fill in remaining arguments that have
3181 // default values. If there aren't any, then the next argument is
3182 // required but missing
3183 if (Lexer.is(EndTok)) {
3184 bool Failure = false;
3185 for (unsigned FAI = 0; FAI < NParameters; ++FAI) {
3186 if (A[FAI].empty()) {
3187 if (M->Parameters[FAI].Required) {
3188 Error(FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(),
3189 "missing value for required parameter "
3190 "'" +
3191 M->Parameters[FAI].Name + "' in macro '" + M->Name + "'");
3192 Failure = true;
3193 }
3194
3195 if (!M->Parameters[FAI].Value.empty())
3196 A[FAI] = M->Parameters[FAI].Value;
3197 }
3198 }
3199 return Failure;
3200 }
3201
3202 if (Lexer.is(AsmToken::Comma))
3203 Lex();
3204 }
3205
3206 return TokError("too many positional arguments");
3207}
3208
3209bool MasmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc,
3210 AsmToken::TokenKind ArgumentEndTok) {
3211 // Arbitrarily limit macro nesting depth (default matches 'as'). We can
3212 // eliminate this, although we should protect against infinite loops.
3213 unsigned MaxNestingDepth = AsmMacroMaxNestingDepth;
3214 if (ActiveMacros.size() == MaxNestingDepth) {
3215 std::ostringstream MaxNestingDepthError;
3216 MaxNestingDepthError << "macros cannot be nested more than "
3217 << MaxNestingDepth << " levels deep."
3218 << " Use -asm-macro-max-nesting-depth to increase "
3219 "this limit.";
3220 return TokError(MaxNestingDepthError.str());
3221 }
3222
3223 MCAsmMacroArguments A;
3224 if (parseMacroArguments(M, A, ArgumentEndTok))
3225 return true;
3226
3227 // Macro instantiation is lexical, unfortunately. We construct a new buffer
3228 // to hold the macro body with substitutions.
3229 SmallString<256> Buf;
3230 StringRef Body = M->Body;
3232
3233 if (expandMacro(OS, Body, M->Parameters, A, M->Locals, getTok().getLoc()))
3234 return true;
3235
3236 // We include the endm in the buffer as our cue to exit the macro
3237 // instantiation.
3238 OS << "endm\n";
3239
3240 std::unique_ptr<MemoryBuffer> Instantiation =
3241 MemoryBuffer::getMemBufferCopy(OS.str(), "<instantiation>");
3242
3243 // Create the macro instantiation object and add to the current macro
3244 // instantiation stack.
3245 MacroInstantiation *MI = new MacroInstantiation{
3246 NameLoc, CurBuffer, getTok().getLoc(), TheCondStack.size()};
3247 ActiveMacros.push_back(MI);
3248
3249 ++NumOfMacroInstantiations;
3250
3251 // Jump to the macro instantiation and prime the lexer.
3252 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(Instantiation), SMLoc());
3253 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer());
3254 EndStatementAtEOFStack.push_back(true);
3255 Lex();
3256
3257 return false;
3258}
3259
3260void MasmParser::handleMacroExit() {
3261 // Jump to the token we should return to, and consume it.
3262 EndStatementAtEOFStack.pop_back();
3263 jumpToLoc(ActiveMacros.back()->ExitLoc, ActiveMacros.back()->ExitBuffer,
3264 EndStatementAtEOFStack.back());
3265 Lex();
3266
3267 // Pop the instantiation entry.
3268 delete ActiveMacros.back();
3269 ActiveMacros.pop_back();
3270}
3271
3272bool MasmParser::handleMacroInvocation(const MCAsmMacro *M, SMLoc NameLoc) {
3273 if (!M->IsFunction)
3274 return Error(NameLoc, "cannot invoke macro procedure as function");
3275
3276 if (parseToken(AsmToken::LParen, "invoking macro function '" + M->Name +
3277 "' requires arguments in parentheses") ||
3278 handleMacroEntry(M, NameLoc, AsmToken::RParen))
3279 return true;
3280
3281 // Parse all statements in the macro, retrieving the exit value when it ends.
3282 std::string ExitValue;
3283 SmallVector<AsmRewrite, 4> AsmStrRewrites;
3284 while (Lexer.isNot(AsmToken::Eof)) {
3285 ParseStatementInfo Info(&AsmStrRewrites);
3286 bool Parsed = parseStatement(Info, nullptr);
3287
3288 if (!Parsed && Info.ExitValue) {
3289 ExitValue = std::move(*Info.ExitValue);
3290 break;
3291 }
3292
3293 // If we have a Lexer Error we are on an Error Token. Load in Lexer Error
3294 // for printing ErrMsg via Lex() only if no (presumably better) parser error
3295 // exists.
3296 if (Parsed && !hasPendingError() && Lexer.getTok().is(AsmToken::Error)) {
3297 Lex();
3298 }
3299
3300 // parseStatement returned true so may need to emit an error.
3301 printPendingErrors();
3302
3303 // Skipping to the next line if needed.
3304 if (Parsed && !getLexer().isAtStartOfStatement())
3305 eatToEndOfStatement();
3306 }
3307
3308 // Consume the right-parenthesis on the other side of the arguments.
3309 if (parseRParen())
3310 return true;
3311
3312 // Exit values may require lexing, unfortunately. We construct a new buffer to
3313 // hold the exit value.
3314 std::unique_ptr<MemoryBuffer> MacroValue =
3315 MemoryBuffer::getMemBufferCopy(ExitValue, "<macro-value>");
3316
3317 // Jump from this location to the instantiated exit value, and prime the
3318 // lexer.
3319 CurBuffer = SrcMgr.AddNewSourceBuffer(std::move(MacroValue), Lexer.getLoc());
3320 Lexer.setBuffer(SrcMgr.getMemoryBuffer(CurBuffer)->getBuffer(), nullptr,
3321 /*EndStatementAtEOF=*/false);
3322 EndStatementAtEOFStack.push_back(false);
3323 Lex();
3324
3325 return false;
3326}
3327
3328/// parseIdentifier:
3329/// ::= identifier
3330/// ::= string
3331bool MasmParser::parseIdentifier(StringRef &Res,
3332 IdentifierPositionKind Position) {
3333 // The assembler has relaxed rules for accepting identifiers, in particular we
3334 // allow things like '.globl $foo' and '.def @feat.00', which would normally
3335 // be separate tokens. At this level, we have already lexed so we cannot
3336 // (currently) handle this as a context dependent token, instead we detect
3337 // adjacent tokens and return the combined identifier.
3338 if (Lexer.is(AsmToken::Dollar) || Lexer.is(AsmToken::At)) {
3339 SMLoc PrefixLoc = getLexer().getLoc();
3340
3341 // Consume the prefix character, and check for a following identifier.
3342
3343 AsmToken nextTok = peekTok(false);
3344
3345 if (nextTok.isNot(AsmToken::Identifier))
3346 return true;
3347
3348 // We have a '$' or '@' followed by an identifier, make sure they are adjacent.
3349 if (PrefixLoc.getPointer() + 1 != nextTok.getLoc().getPointer())
3350 return true;
3351
3352 // eat $ or @
3353 Lexer.Lex(); // Lexer's Lex guarantees consecutive token.
3354 // Construct the joined identifier and consume the token.
3355 Res =
3356 StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1);
3357 Lex(); // Parser Lex to maintain invariants.
3358 return false;
3359 }
3360
3361 if (Lexer.isNot(AsmToken::Identifier) && Lexer.isNot(AsmToken::String))
3362 return true;
3363
3364 Res = getTok().getIdentifier();
3365
3366 // Consume the identifier token - but if parsing certain directives, avoid
3367 // lexical expansion of the next token.
3368 ExpandKind ExpandNextToken = ExpandMacros;
3369 if (Position == StartOfStatement &&
3371 .CaseLower("echo", true)
3372 .CasesLower("ifdef", "ifndef", "elseifdef", "elseifndef", true)
3373 .Default(false)) {
3374 ExpandNextToken = DoNotExpandMacros;
3375 }
3376 Lex(ExpandNextToken);
3377
3378 return false;
3379}
3380
3381/// parseDirectiveEquate:
3382/// ::= name "=" expression
3383/// | name "equ" expression (not redefinable)
3384/// | name "equ" text-list
3385/// | name "textequ" text-list (redefinability unspecified)
3386bool MasmParser::parseDirectiveEquate(StringRef IDVal, StringRef Name,
3387 DirectiveKind DirKind, SMLoc NameLoc) {
3388 auto BuiltinIt = BuiltinSymbolMap.find(Name.lower());
3389 if (BuiltinIt != BuiltinSymbolMap.end())
3390 return Error(NameLoc, "cannot redefine a built-in symbol");
3391
3392 Variable &Var = Variables[Name.lower()];
3393 if (Var.Name.empty()) {
3394 Var.Name = Name;
3395 }
3396
3397 SMLoc StartLoc = Lexer.getLoc();
3398 if (DirKind == DK_EQU || DirKind == DK_TEXTEQU) {
3399 // "equ" and "textequ" both allow text expressions.
3400 std::string Value;
3401 std::string TextItem;
3402 if (!parseTextItem(TextItem)) {
3403 Value += TextItem;
3404
3405 // Accept a text-list, not just one text-item.
3406 auto parseItem = [&]() -> bool {
3407 if (parseTextItem(TextItem))
3408 return TokError("expected text item");
3409 Value += TextItem;
3410 return false;
3411 };
3412 if (parseOptionalToken(AsmToken::Comma) && parseMany(parseItem))
3413 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3414
3415 if (!Var.IsText || Var.TextValue != Value) {
3416 switch (Var.Redefinable) {
3417 case Variable::NOT_REDEFINABLE:
3418 return Error(getTok().getLoc(), "invalid variable redefinition");
3419 case Variable::WARN_ON_REDEFINITION:
3420 if (Warning(NameLoc, "redefining '" + Name +
3421 "', already defined on the command line")) {
3422 return true;
3423 }
3424 break;
3425 default:
3426 break;
3427 }
3428 }
3429 Var.IsText = true;
3430 Var.TextValue = Value;
3431 Var.Redefinable = Variable::REDEFINABLE;
3432
3433 return false;
3434 }
3435 }
3436 if (DirKind == DK_TEXTEQU)
3437 return TokError("expected <text> in '" + Twine(IDVal) + "' directive");
3438
3439 // Parse as expression assignment.
3440 const MCExpr *Expr;
3441 SMLoc EndLoc;
3442 if (parseExpression(Expr, EndLoc))
3443 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3444 StringRef ExprAsString = StringRef(
3445 StartLoc.getPointer(), EndLoc.getPointer() - StartLoc.getPointer());
3446
3447 int64_t Value;
3448 if (!Expr->evaluateAsAbsolute(Value, getStreamer().getAssemblerPtr())) {
3449 if (DirKind == DK_ASSIGN)
3450 return Error(
3451 StartLoc,
3452 "expected absolute expression; not all symbols have known values",
3453 {StartLoc, EndLoc});
3454
3455 // Not an absolute expression; define as a text replacement.
3456 if (!Var.IsText || Var.TextValue != ExprAsString) {
3457 switch (Var.Redefinable) {
3458 case Variable::NOT_REDEFINABLE:
3459 return Error(getTok().getLoc(), "invalid variable redefinition");
3460 case Variable::WARN_ON_REDEFINITION:
3461 if (Warning(NameLoc, "redefining '" + Name +
3462 "', already defined on the command line")) {
3463 return true;
3464 }
3465 break;
3466 default:
3467 break;
3468 }
3469 }
3470
3471 Var.IsText = true;
3472 Var.TextValue = ExprAsString.str();
3473 Var.Redefinable = Variable::REDEFINABLE;
3474
3475 return false;
3476 }
3477
3478 MCSymbol *Sym = getContext().getOrCreateSymbol(Var.Name);
3479
3480 const MCConstantExpr *PrevValue =
3481 Sym->isVariable() ? dyn_cast_or_null<MCConstantExpr>(
3482 Sym->getVariableValue(/*SetUsed=*/false))
3483 : nullptr;
3484 if (Var.IsText || !PrevValue || PrevValue->getValue() != Value) {
3485 switch (Var.Redefinable) {
3486 case Variable::NOT_REDEFINABLE:
3487 return Error(getTok().getLoc(), "invalid variable redefinition");
3488 case Variable::WARN_ON_REDEFINITION:
3489 if (Warning(NameLoc, "redefining '" + Name +
3490 "', already defined on the command line")) {
3491 return true;
3492 }
3493 break;
3494 default:
3495 break;
3496 }
3497 }
3498
3499 Var.IsText = false;
3500 Var.TextValue.clear();
3501 Var.Redefinable = (DirKind == DK_ASSIGN) ? Variable::REDEFINABLE
3502 : Variable::NOT_REDEFINABLE;
3503
3504 Sym->setRedefinable(Var.Redefinable != Variable::NOT_REDEFINABLE);
3505 Sym->setVariableValue(Expr);
3506 Sym->setExternal(false);
3507
3508 return false;
3509}
3510
3511bool MasmParser::parseEscapedString(std::string &Data) {
3512 if (check(getTok().isNot(AsmToken::String), "expected string"))
3513 return true;
3514
3515 Data = "";
3516 char Quote = getTok().getString().front();
3517 StringRef Str = getTok().getStringContents();
3518 Data.reserve(Str.size());
3519 for (size_t i = 0, e = Str.size(); i != e; ++i) {
3520 Data.push_back(Str[i]);
3521 if (Str[i] == Quote) {
3522 // MASM treats doubled delimiting quotes as an escaped delimiting quote.
3523 // If we're escaping the string's trailing delimiter, we're definitely
3524 // missing a quotation mark.
3525 if (i + 1 == Str.size())
3526 return Error(getTok().getLoc(), "missing quotation mark in string");
3527 if (Str[i + 1] == Quote)
3528 ++i;
3529 }
3530 }
3531
3532 Lex();
3533 return false;
3534}
3535
3536bool MasmParser::parseAngleBracketString(std::string &Data) {
3537 SMLoc EndLoc, StartLoc = getTok().getLoc();
3538 if (isAngleBracketString(StartLoc, EndLoc)) {
3539 const char *StartChar = StartLoc.getPointer() + 1;
3540 const char *EndChar = EndLoc.getPointer() - 1;
3541 jumpToLoc(EndLoc, CurBuffer, EndStatementAtEOFStack.back());
3542 // Eat from '<' to '>'.
3543 Lex();
3544
3545 Data = angleBracketString(StringRef(StartChar, EndChar - StartChar));
3546 return false;
3547 }
3548 return true;
3549}
3550
3551/// textItem ::= textLiteral | textMacroID | % constExpr
3552bool MasmParser::parseTextItem(std::string &Data) {
3553 switch (getTok().getKind()) {
3554 default:
3555 return true;
3556 case AsmToken::Percent: {
3557 int64_t Res;
3558 if (parseToken(AsmToken::Percent) || parseAbsoluteExpression(Res))
3559 return true;
3560 Data = std::to_string(Res);
3561 return false;
3562 }
3563 case AsmToken::Less:
3565 case AsmToken::LessLess:
3567 return parseAngleBracketString(Data);
3568 case AsmToken::Identifier: {
3569 // This must be a text macro; we need to expand it accordingly.
3570 StringRef ID;
3571 SMLoc StartLoc = getTok().getLoc();
3572 if (parseIdentifier(ID))
3573 return true;
3574 Data = ID.str();
3575
3576 bool Expanded = false;
3577 while (true) {
3578 // Try to resolve as a built-in text macro
3579 auto BuiltinIt = BuiltinSymbolMap.find(ID.lower());
3580 if (BuiltinIt != BuiltinSymbolMap.end()) {
3581 std::optional<std::string> BuiltinText =
3582 evaluateBuiltinTextMacro(BuiltinIt->getValue(), StartLoc);
3583 if (!BuiltinText) {
3584 // Not a text macro; break without substituting
3585 break;
3586 }
3587 Data = std::move(*BuiltinText);
3588 ID = StringRef(Data);
3589 Expanded = true;
3590 continue;
3591 }
3592
3593 // Try to resolve as a variable text macro
3594 auto VarIt = Variables.find(ID.lower());
3595 if (VarIt != Variables.end()) {
3596 const Variable &Var = VarIt->getValue();
3597 if (!Var.IsText) {
3598 // Not a text macro; break without substituting
3599 break;
3600 }
3601 Data = Var.TextValue;
3602 ID = StringRef(Data);
3603 Expanded = true;
3604 continue;
3605 }
3606
3607 break;
3608 }
3609
3610 if (!Expanded) {
3611 // Not a text macro; not usable in TextItem context. Since we haven't used
3612 // the token, put it back for better error recovery.
3613 getLexer().UnLex(AsmToken(AsmToken::Identifier, ID));
3614 return true;
3615 }
3616 return false;
3617 }
3618 }
3619 llvm_unreachable("unhandled token kind");
3620}
3621
3622/// parseDirectiveAscii:
3623/// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ]
3624bool MasmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) {
3625 auto parseOp = [&]() -> bool {
3626 std::string Data;
3627 if (checkForValidSection() || parseEscapedString(Data))
3628 return true;
3629 getStreamer().emitBytes(Data);
3630 if (ZeroTerminated)
3631 getStreamer().emitBytes(StringRef("\0", 1));
3632 return false;
3633 };
3634
3635 if (parseMany(parseOp))
3636 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3637 return false;
3638}
3639
3640bool MasmParser::emitIntValue(const MCExpr *Value, unsigned Size) {
3641 // Special case constant expressions to match code generator.
3642 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
3643 assert(Size <= 8 && "Invalid size");
3644 int64_t IntValue = MCE->getValue();
3645 if (!isUIntN(8 * Size, IntValue) && !isIntN(8 * Size, IntValue))
3646 return Error(MCE->getLoc(), "out of range literal value");
3647 getStreamer().emitIntValue(IntValue, Size);
3648 } else {
3649 const MCSymbolRefExpr *MSE = dyn_cast<MCSymbolRefExpr>(Value);
3650 if (MSE && MSE->getSymbol().getName() == "?") {
3651 // ? initializer; treat as 0.
3652 getStreamer().emitIntValue(0, Size);
3653 } else {
3654 getStreamer().emitValue(Value, Size, Value->getLoc());
3655 }
3656 }
3657 return false;
3658}
3659
3660bool MasmParser::parseScalarInitializer(unsigned Size,
3662 unsigned StringPadLength) {
3663 if (Size == 1 && getTok().is(AsmToken::String)) {
3664 std::string Value;
3665 if (parseEscapedString(Value))
3666 return true;
3667 // Treat each character as an initializer.
3668 for (const unsigned char CharVal : Value)
3669 Values.push_back(MCConstantExpr::create(CharVal, getContext()));
3670
3671 // Pad the string with spaces to the specified length.
3672 for (size_t i = Value.size(); i < StringPadLength; ++i)
3673 Values.push_back(MCConstantExpr::create(' ', getContext()));
3674 } else {
3675 const MCExpr *Value;
3676 if (parseExpression(Value))
3677 return true;
3678 if (getTok().is(AsmToken::Identifier) &&
3679 getTok().getString().equals_insensitive("dup")) {
3680 Lex(); // Eat 'dup'.
3681 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
3682 if (!MCE)
3683 return Error(Value->getLoc(),
3684 "cannot repeat value a non-constant number of times");
3685 const int64_t Repetitions = MCE->getValue();
3686 if (Repetitions < 0)
3687 return Error(Value->getLoc(),
3688 "cannot repeat value a negative number of times");
3689
3690 SmallVector<const MCExpr *, 1> DuplicatedValues;
3691 if (parseToken(AsmToken::LParen,
3692 "parentheses required for 'dup' contents") ||
3693 parseScalarInstList(Size, DuplicatedValues) || parseRParen())
3694 return true;
3695
3696 for (int i = 0; i < Repetitions; ++i)
3697 Values.append(DuplicatedValues.begin(), DuplicatedValues.end());
3698 } else {
3699 Values.push_back(Value);
3700 }
3701 }
3702 return false;
3703}
3704
3705bool MasmParser::parseScalarInstList(unsigned Size,
3707 const AsmToken::TokenKind EndToken) {
3708 while (getTok().isNot(EndToken) &&
3709 (EndToken != AsmToken::Greater ||
3710 getTok().isNot(AsmToken::GreaterGreater))) {
3711 parseScalarInitializer(Size, Values);
3712
3713 // If we see a comma, continue, and allow line continuation.
3714 if (!parseOptionalToken(AsmToken::Comma))
3715 break;
3716 parseOptionalToken(AsmToken::EndOfStatement);
3717 }
3718 return false;
3719}
3720
3721bool MasmParser::emitIntegralValues(unsigned Size, unsigned *Count) {
3723 if (checkForValidSection() || parseScalarInstList(Size, Values))
3724 return true;
3725
3726 for (const auto *Value : Values) {
3727 emitIntValue(Value, Size);
3728 }
3729 if (Count)
3730 *Count = Values.size();
3731 return false;
3732}
3733
3734// Add a field to the current structure.
3735bool MasmParser::addIntegralField(StringRef Name, unsigned Size) {
3736 StructInfo &Struct = StructInProgress.back();
3737 FieldInfo &Field = Struct.addField(Name, FT_INTEGRAL, Size);
3738 IntFieldInfo &IntInfo = Field.Contents.IntInfo;
3739
3740 Field.Type = Size;
3741
3742 if (parseScalarInstList(Size, IntInfo.Values))
3743 return true;
3744
3745 Field.SizeOf = Field.Type * IntInfo.Values.size();
3746 Field.LengthOf = IntInfo.Values.size();
3747 const unsigned FieldEnd = Field.Offset + Field.SizeOf;
3748 if (!Struct.IsUnion) {
3749 Struct.NextOffset = FieldEnd;
3750 }
3751 Struct.Size = std::max(Struct.Size, FieldEnd);
3752 return false;
3753}
3754
3755/// parseDirectiveValue
3756/// ::= (byte | word | ... ) [ expression (, expression)* ]
3757bool MasmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) {
3758 if (StructInProgress.empty()) {
3759 // Initialize data value.
3760 if (emitIntegralValues(Size))
3761 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3762 } else if (addIntegralField("", Size)) {
3763 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3764 }
3765
3766 return false;
3767}
3768
3769/// parseDirectiveNamedValue
3770/// ::= name (byte | word | ... ) [ expression (, expression)* ]
3771bool MasmParser::parseDirectiveNamedValue(StringRef TypeName, unsigned Size,
3772 StringRef Name, SMLoc NameLoc) {
3773 if (StructInProgress.empty()) {
3774 // Initialize named data value.
3775 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3776 getStreamer().emitLabel(Sym);
3777 unsigned Count;
3778 if (emitIntegralValues(Size, &Count))
3779 return addErrorSuffix(" in '" + Twine(TypeName) + "' directive");
3780
3782 Type.Name = TypeName;
3783 Type.Size = Size * Count;
3784 Type.ElementSize = Size;
3785 Type.Length = Count;
3786 KnownType[Name.lower()] = Type;
3787 } else if (addIntegralField(Name, Size)) {
3788 return addErrorSuffix(" in '" + Twine(TypeName) + "' directive");
3789 }
3790
3791 return false;
3792}
3793
3794static bool parseHexOcta(MasmParser &Asm, uint64_t &hi, uint64_t &lo) {
3795 if (Asm.getTok().isNot(AsmToken::Integer) &&
3796 Asm.getTok().isNot(AsmToken::BigNum))
3797 return Asm.TokError("unknown token in expression");
3798 SMLoc ExprLoc = Asm.getTok().getLoc();
3799 APInt IntValue = Asm.getTok().getAPIntVal();
3800 Asm.Lex();
3801 if (!IntValue.isIntN(128))
3802 return Asm.Error(ExprLoc, "out of range literal value");
3803 if (!IntValue.isIntN(64)) {
3804 hi = IntValue.getHiBits(IntValue.getBitWidth() - 64).getZExtValue();
3805 lo = IntValue.getLoBits(64).getZExtValue();
3806 } else {
3807 hi = 0;
3808 lo = IntValue.getZExtValue();
3809 }
3810 return false;
3811}
3812
3813bool MasmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) {
3814 // We don't truly support arithmetic on floating point expressions, so we
3815 // have to manually parse unary prefixes.
3816 bool IsNeg = false;
3817 SMLoc SignLoc;
3818 if (getLexer().is(AsmToken::Minus)) {
3819 SignLoc = getLexer().getLoc();
3820 Lexer.Lex();
3821 IsNeg = true;
3822 } else if (getLexer().is(AsmToken::Plus)) {
3823 SignLoc = getLexer().getLoc();
3824 Lexer.Lex();
3825 }
3826
3827 if (Lexer.is(AsmToken::Error))
3828 return TokError(Lexer.getErr());
3829 if (Lexer.isNot(AsmToken::Integer) && Lexer.isNot(AsmToken::Real) &&
3831 return TokError("unexpected token in directive");
3832
3833 // Convert to an APFloat.
3834 APFloat Value(Semantics);
3835 StringRef IDVal = getTok().getString();
3836 if (getLexer().is(AsmToken::Identifier)) {
3837 if (IDVal.equals_insensitive("infinity") || IDVal.equals_insensitive("inf"))
3838 Value = APFloat::getInf(Semantics);
3839 else if (IDVal.equals_insensitive("nan"))
3840 Value = APFloat::getNaN(Semantics, false, ~0);
3841 else if (IDVal.equals_insensitive("?"))
3842 Value = APFloat::getZero(Semantics);
3843 else
3844 return TokError("invalid floating point literal");
3845 } else if (IDVal.consume_back("r") || IDVal.consume_back("R")) {
3846 // MASM hexadecimal floating-point literal; no APFloat conversion needed.
3847 // To match ML64.exe, ignore the initial sign.
3848 unsigned SizeInBits = Value.getSizeInBits(Semantics);
3849 if (SizeInBits != (IDVal.size() << 2))
3850 return TokError("invalid floating point literal");
3851
3852 // Consume the numeric token.
3853 Lex();
3854
3855 Res = APInt(SizeInBits, IDVal, 16);
3856 if (SignLoc.isValid())
3857 return Warning(SignLoc, "MASM-style hex floats ignore explicit sign");
3858 return false;
3859 } else if (errorToBool(
3860 Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven)
3861 .takeError())) {
3862 return TokError("invalid floating point literal");
3863 }
3864 if (IsNeg)
3865 Value.changeSign();
3866
3867 // Consume the numeric token.
3868 Lex();
3869
3870 Res = Value.bitcastToAPInt();
3871
3872 return false;
3873}
3874
3875bool MasmParser::parseRealInstList(const fltSemantics &Semantics,
3876 SmallVectorImpl<APInt> &ValuesAsInt,
3877 const AsmToken::TokenKind EndToken) {
3878 while (getTok().isNot(EndToken) ||
3879 (EndToken == AsmToken::Greater &&
3880 getTok().isNot(AsmToken::GreaterGreater))) {
3881 const AsmToken NextTok = peekTok();
3882 if (NextTok.is(AsmToken::Identifier) &&
3883 NextTok.getString().equals_insensitive("dup")) {
3884 const MCExpr *Value;
3885 if (parseExpression(Value) || parseToken(AsmToken::Identifier))
3886 return true;
3887 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
3888 if (!MCE)
3889 return Error(Value->getLoc(),
3890 "cannot repeat value a non-constant number of times");
3891 const int64_t Repetitions = MCE->getValue();
3892 if (Repetitions < 0)
3893 return Error(Value->getLoc(),
3894 "cannot repeat value a negative number of times");
3895
3896 SmallVector<APInt, 1> DuplicatedValues;
3897 if (parseToken(AsmToken::LParen,
3898 "parentheses required for 'dup' contents") ||
3899 parseRealInstList(Semantics, DuplicatedValues) || parseRParen())
3900 return true;
3901
3902 for (int i = 0; i < Repetitions; ++i)
3903 ValuesAsInt.append(DuplicatedValues.begin(), DuplicatedValues.end());
3904 } else {
3905 APInt AsInt;
3906 if (parseRealValue(Semantics, AsInt))
3907 return true;
3908 ValuesAsInt.push_back(AsInt);
3909 }
3910
3911 // Continue if we see a comma. (Also, allow line continuation.)
3912 if (!parseOptionalToken(AsmToken::Comma))
3913 break;
3914 parseOptionalToken(AsmToken::EndOfStatement);
3915 }
3916
3917 return false;
3918}
3919
3920// Initialize real data values.
3921bool MasmParser::emitRealValues(const fltSemantics &Semantics,
3922 unsigned *Count) {
3923 if (checkForValidSection())
3924 return true;
3925
3926 SmallVector<APInt, 1> ValuesAsInt;
3927 if (parseRealInstList(Semantics, ValuesAsInt))
3928 return true;
3929
3930 for (const APInt &AsInt : ValuesAsInt) {
3931 getStreamer().emitIntValue(AsInt);
3932 }
3933 if (Count)
3934 *Count = ValuesAsInt.size();
3935 return false;
3936}
3937
3938// Add a real field to the current struct.
3939bool MasmParser::addRealField(StringRef Name, const fltSemantics &Semantics,
3940 size_t Size) {
3941 StructInfo &Struct = StructInProgress.back();
3942 FieldInfo &Field = Struct.addField(Name, FT_REAL, Size);
3943 RealFieldInfo &RealInfo = Field.Contents.RealInfo;
3944
3945 Field.SizeOf = 0;
3946
3947 if (parseRealInstList(Semantics, RealInfo.AsIntValues))
3948 return true;
3949
3950 Field.Type = RealInfo.AsIntValues.back().getBitWidth() / 8;
3951 Field.LengthOf = RealInfo.AsIntValues.size();
3952 Field.SizeOf = Field.Type * Field.LengthOf;
3953
3954 const unsigned FieldEnd = Field.Offset + Field.SizeOf;
3955 if (!Struct.IsUnion) {
3956 Struct.NextOffset = FieldEnd;
3957 }
3958 Struct.Size = std::max(Struct.Size, FieldEnd);
3959 return false;
3960}
3961
3962/// parseDirectiveRealValue
3963/// ::= (real4 | real8 | real10) [ expression (, expression)* ]
3964bool MasmParser::parseDirectiveRealValue(StringRef IDVal,
3965 const fltSemantics &Semantics,
3966 size_t Size) {
3967 if (StructInProgress.empty()) {
3968 // Initialize data value.
3969 if (emitRealValues(Semantics))
3970 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3971 } else if (addRealField("", Semantics, Size)) {
3972 return addErrorSuffix(" in '" + Twine(IDVal) + "' directive");
3973 }
3974 return false;
3975}
3976
3977/// parseDirectiveNamedRealValue
3978/// ::= name (real4 | real8 | real10) [ expression (, expression)* ]
3979bool MasmParser::parseDirectiveNamedRealValue(StringRef TypeName,
3980 const fltSemantics &Semantics,
3981 unsigned Size, StringRef Name,
3982 SMLoc NameLoc) {
3983 if (StructInProgress.empty()) {
3984 // Initialize named data value.
3985 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
3986 getStreamer().emitLabel(Sym);
3987 unsigned Count;
3988 if (emitRealValues(Semantics, &Count))
3989 return addErrorSuffix(" in '" + TypeName + "' directive");
3990
3992 Type.Name = TypeName;
3993 Type.Size = Size * Count;
3994 Type.ElementSize = Size;
3995 Type.Length = Count;
3996 KnownType[Name.lower()] = Type;
3997 } else if (addRealField(Name, Semantics, Size)) {
3998 return addErrorSuffix(" in '" + TypeName + "' directive");
3999 }
4000 return false;
4001}
4002
4003bool MasmParser::parseOptionalAngleBracketOpen() {
4004 const AsmToken Tok = getTok();
4005 if (parseOptionalToken(AsmToken::LessLess)) {
4006 AngleBracketDepth++;
4007 Lexer.UnLex(AsmToken(AsmToken::Less, Tok.getString().substr(1)));
4008 return true;
4009 } else if (parseOptionalToken(AsmToken::LessGreater)) {
4010 AngleBracketDepth++;
4012 return true;
4013 } else if (parseOptionalToken(AsmToken::Less)) {
4014 AngleBracketDepth++;
4015 return true;
4016 }
4017
4018 return false;
4019}
4020
4021bool MasmParser::parseAngleBracketClose(const Twine &Msg) {
4022 const AsmToken Tok = getTok();
4023 if (parseOptionalToken(AsmToken::GreaterGreater)) {
4025 } else if (parseToken(AsmToken::Greater, Msg)) {
4026 return true;
4027 }
4028 AngleBracketDepth--;
4029 return false;
4030}
4031
4032bool MasmParser::parseFieldInitializer(const FieldInfo &Field,
4033 const IntFieldInfo &Contents,
4034 FieldInitializer &Initializer) {
4035 SMLoc Loc = getTok().getLoc();
4036
4038 if (parseOptionalToken(AsmToken::LCurly)) {
4039 if (Field.LengthOf == 1 && Field.Type > 1)
4040 return Error(Loc, "Cannot initialize scalar field with array value");
4041 if (parseScalarInstList(Field.Type, Values, AsmToken::RCurly) ||
4042 parseToken(AsmToken::RCurly))
4043 return true;
4044 } else if (parseOptionalAngleBracketOpen()) {
4045 if (Field.LengthOf == 1 && Field.Type > 1)
4046 return Error(Loc, "Cannot initialize scalar field with array value");
4047 if (parseScalarInstList(Field.Type, Values, AsmToken::Greater) ||
4048 parseAngleBracketClose())
4049 return true;
4050 } else if (Field.LengthOf > 1 && Field.Type > 1) {
4051 return Error(Loc, "Cannot initialize array field with scalar value");
4052 } else if (parseScalarInitializer(Field.Type, Values,
4053 /*StringPadLength=*/Field.LengthOf)) {
4054 return true;
4055 }
4056
4057 if (Values.size() > Field.LengthOf) {
4058 return Error(Loc, "Initializer too long for field; expected at most " +
4059 std::to_string(Field.LengthOf) + " elements, got " +
4060 std::to_string(Values.size()));
4061 }
4062 // Default-initialize all remaining values.
4063 Values.append(Contents.Values.begin() + Values.size(), Contents.Values.end());
4064
4065 Initializer = FieldInitializer(std::move(Values));
4066 return false;
4067}
4068
4069bool MasmParser::parseFieldInitializer(const FieldInfo &Field,
4070 const RealFieldInfo &Contents,
4071 FieldInitializer &Initializer) {
4072 const fltSemantics *Semantics;
4073 switch (Field.Type) {
4074 case 4:
4075 Semantics = &APFloat::IEEEsingle();
4076 break;
4077 case 8:
4078 Semantics = &APFloat::IEEEdouble();
4079 break;
4080 case 10:
4081 Semantics = &APFloat::x87DoubleExtended();
4082 break;
4083 default:
4084 llvm_unreachable("unknown real field type");
4085 }
4086
4087 SMLoc Loc = getTok().getLoc();
4088
4089 SmallVector<APInt, 1> AsIntValues;
4090 if (parseOptionalToken(AsmToken::LCurly)) {
4091 if (Field.LengthOf == 1)
4092 return Error(Loc, "Cannot initialize scalar field with array value");
4093 if (parseRealInstList(*Semantics, AsIntValues, AsmToken::RCurly) ||
4094 parseToken(AsmToken::RCurly))
4095 return true;
4096 } else if (parseOptionalAngleBracketOpen()) {
4097 if (Field.LengthOf == 1)
4098 return Error(Loc, "Cannot initialize scalar field with array value");
4099 if (parseRealInstList(*Semantics, AsIntValues, AsmToken::Greater) ||
4100 parseAngleBracketClose())
4101 return true;
4102 } else if (Field.LengthOf > 1) {
4103 return Error(Loc, "Cannot initialize array field with scalar value");
4104 } else {
4105 AsIntValues.emplace_back();
4106 if (parseRealValue(*Semantics, AsIntValues.back()))
4107 return true;
4108 }
4109
4110 if (AsIntValues.size() > Field.LengthOf) {
4111 return Error(Loc, "Initializer too long for field; expected at most " +
4112 std::to_string(Field.LengthOf) + " elements, got " +
4113 std::to_string(AsIntValues.size()));
4114 }
4115 // Default-initialize all remaining values.
4116 AsIntValues.append(Contents.AsIntValues.begin() + AsIntValues.size(),
4117 Contents.AsIntValues.end());
4118
4119 Initializer = FieldInitializer(std::move(AsIntValues));
4120 return false;
4121}
4122
4123bool MasmParser::parseFieldInitializer(const FieldInfo &Field,
4124 const StructFieldInfo &Contents,
4125 FieldInitializer &Initializer) {
4126 SMLoc Loc = getTok().getLoc();
4127
4128 std::vector<StructInitializer> Initializers;
4129 if (Field.LengthOf > 1) {
4130 if (parseOptionalToken(AsmToken::LCurly)) {
4131 if (parseStructInstList(Contents.Structure, Initializers,
4133 parseToken(AsmToken::RCurly))
4134 return true;
4135 } else if (parseOptionalAngleBracketOpen()) {
4136 if (parseStructInstList(Contents.Structure, Initializers,
4138 parseAngleBracketClose())
4139 return true;
4140 } else {
4141 return Error(Loc, "Cannot initialize array field with scalar value");
4142 }
4143 } else {
4144 Initializers.emplace_back();
4145 if (parseStructInitializer(Contents.Structure, Initializers.back()))
4146 return true;
4147 }
4148
4149 if (Initializers.size() > Field.LengthOf) {
4150 return Error(Loc, "Initializer too long for field; expected at most " +
4151 std::to_string(Field.LengthOf) + " elements, got " +
4152 std::to_string(Initializers.size()));
4153 }
4154 // Default-initialize all remaining values.
4155 Initializers.insert(Initializers.end(),
4156 Contents.Initializers.begin() + Initializers.size(),
4157 Contents.Initializers.end());
4158
4159 Initializer = FieldInitializer(std::move(Initializers), Contents.Structure);
4160 return false;
4161}
4162
4163bool MasmParser::parseFieldInitializer(const FieldInfo &Field,
4164 FieldInitializer &Initializer) {
4165 switch (Field.Contents.FT) {
4166 case FT_INTEGRAL:
4167 return parseFieldInitializer(Field, Field.Contents.IntInfo, Initializer);
4168 case FT_REAL:
4169 return parseFieldInitializer(Field, Field.Contents.RealInfo, Initializer);
4170 case FT_STRUCT:
4171 return parseFieldInitializer(Field, Field.Contents.StructInfo, Initializer);
4172 }
4173 llvm_unreachable("Unhandled FieldType enum");
4174}
4175
4176bool MasmParser::parseStructInitializer(const StructInfo &Structure,
4177 StructInitializer &Initializer) {
4178 const AsmToken FirstToken = getTok();
4179
4180 std::optional<AsmToken::TokenKind> EndToken;
4181 if (parseOptionalToken(AsmToken::LCurly)) {
4182 EndToken = AsmToken::RCurly;
4183 } else if (parseOptionalAngleBracketOpen()) {
4184 EndToken = AsmToken::Greater;
4185 AngleBracketDepth++;
4186 } else if (FirstToken.is(AsmToken::Identifier) &&
4187 FirstToken.getString() == "?") {
4188 // ? initializer; leave EndToken uninitialized to treat as empty.
4189 if (parseToken(AsmToken::Identifier))
4190 return true;
4191 } else {
4192 return Error(FirstToken.getLoc(), "Expected struct initializer");
4193 }
4194
4195 auto &FieldInitializers = Initializer.FieldInitializers;
4196 size_t FieldIndex = 0;
4197 if (EndToken) {
4198 // Initialize all fields with given initializers.
4199 while (getTok().isNot(*EndToken) && FieldIndex < Structure.Fields.size()) {
4200 const FieldInfo &Field = Structure.Fields[FieldIndex++];
4201 if (parseOptionalToken(AsmToken::Comma)) {
4202 // Empty initializer; use the default and continue. (Also, allow line
4203 // continuation.)
4204 FieldInitializers.push_back(Field.Contents);
4205 parseOptionalToken(AsmToken::EndOfStatement);
4206 continue;
4207 }
4208 FieldInitializers.emplace_back(Field.Contents.FT);
4209 if (parseFieldInitializer(Field, FieldInitializers.back()))
4210 return true;
4211
4212 // Continue if we see a comma. (Also, allow line continuation.)
4213 SMLoc CommaLoc = getTok().getLoc();
4214 if (!parseOptionalToken(AsmToken::Comma))
4215 break;
4216 if (FieldIndex == Structure.Fields.size())
4217 return Error(CommaLoc, "'" + Structure.Name +
4218 "' initializer initializes too many fields");
4219 parseOptionalToken(AsmToken::EndOfStatement);
4220 }
4221 }
4222 // Default-initialize all remaining fields.
4223 for (const FieldInfo &Field : llvm::drop_begin(Structure.Fields, FieldIndex))
4224 FieldInitializers.push_back(Field.Contents);
4225
4226 if (EndToken) {
4227 if (*EndToken == AsmToken::Greater)
4228 return parseAngleBracketClose();
4229
4230 return parseToken(*EndToken);
4231 }
4232
4233 return false;
4234}
4235
4236bool MasmParser::parseStructInstList(
4237 const StructInfo &Structure, std::vector<StructInitializer> &Initializers,
4238 const AsmToken::TokenKind EndToken) {
4239 while (getTok().isNot(EndToken) ||
4240 (EndToken == AsmToken::Greater &&
4241 getTok().isNot(AsmToken::GreaterGreater))) {
4242 const AsmToken NextTok = peekTok();
4243 if (NextTok.is(AsmToken::Identifier) &&
4244 NextTok.getString().equals_insensitive("dup")) {
4245 const MCExpr *Value;
4246 if (parseExpression(Value) || parseToken(AsmToken::Identifier))
4247 return true;
4248 const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value);
4249 if (!MCE)
4250 return Error(Value->getLoc(),
4251 "cannot repeat value a non-constant number of times");
4252 const int64_t Repetitions = MCE->getValue();
4253 if (Repetitions < 0)
4254 return Error(Value->getLoc(),
4255 "cannot repeat value a negative number of times");
4256
4257 std::vector<StructInitializer> DuplicatedValues;
4258 if (parseToken(AsmToken::LParen,
4259 "parentheses required for 'dup' contents") ||
4260 parseStructInstList(Structure, DuplicatedValues) || parseRParen())
4261 return true;
4262
4263 for (int i = 0; i < Repetitions; ++i)
4264 llvm::append_range(Initializers, DuplicatedValues);
4265 } else {
4266 Initializers.emplace_back();
4267 if (parseStructInitializer(Structure, Initializers.back()))
4268 return true;
4269 }
4270
4271 // Continue if we see a comma. (Also, allow line continuation.)
4272 if (!parseOptionalToken(AsmToken::Comma))
4273 break;
4274 parseOptionalToken(AsmToken::EndOfStatement);
4275 }
4276
4277 return false;
4278}
4279
4280bool MasmParser::emitFieldValue(const FieldInfo &Field,
4281 const IntFieldInfo &Contents) {
4282 // Default-initialize all values.
4283 for (const MCExpr *Value : Contents.Values) {
4284 if (emitIntValue(Value, Field.Type))
4285 return true;
4286 }
4287 return false;
4288}
4289
4290bool MasmParser::emitFieldValue(const FieldInfo &Field,
4291 const RealFieldInfo &Contents) {
4292 for (const APInt &AsInt : Contents.AsIntValues) {
4293 getStreamer().emitIntValue(AsInt.getLimitedValue(),
4294 AsInt.getBitWidth() / 8);
4295 }
4296 return false;
4297}
4298
4299bool MasmParser::emitFieldValue(const FieldInfo &Field,
4300 const StructFieldInfo &Contents) {
4301 for (const auto &Initializer : Contents.Initializers) {
4302 size_t Index = 0, Offset = 0;
4303 for (const auto &SubField : Contents.Structure.Fields) {
4304 getStreamer().emitZeros(SubField.Offset - Offset);
4305 Offset = SubField.Offset + SubField.SizeOf;
4306 emitFieldInitializer(SubField, Initializer.FieldInitializers[Index++]);
4307 }
4308 }
4309 return false;
4310}
4311
4312bool MasmParser::emitFieldValue(const FieldInfo &Field) {
4313 switch (Field.Contents.FT) {
4314 case FT_INTEGRAL:
4315 return emitFieldValue(Field, Field.Contents.IntInfo);
4316 case FT_REAL:
4317 return emitFieldValue(Field, Field.Contents.RealInfo);
4318 case FT_STRUCT:
4319 return emitFieldValue(Field, Field.Contents.StructInfo);
4320 }
4321 llvm_unreachable("Unhandled FieldType enum");
4322}
4323
4324bool MasmParser::emitFieldInitializer(const FieldInfo &Field,
4325 const IntFieldInfo &Contents,
4326 const IntFieldInfo &Initializer) {
4327 for (const auto &Value : Initializer.Values) {
4328 if (emitIntValue(Value, Field.Type))
4329 return true;
4330 }
4331 // Default-initialize all remaining values.
4332 for (const auto &Value :
4333 llvm::drop_begin(Contents.Values, Initializer.Values.size())) {
4334 if (emitIntValue(Value, Field.Type))
4335 return true;
4336 }
4337 return false;
4338}
4339
4340bool MasmParser::emitFieldInitializer(const FieldInfo &Field,
4341 const RealFieldInfo &Contents,
4342 const RealFieldInfo &Initializer) {
4343 for (const auto &AsInt : Initializer.AsIntValues) {
4344 getStreamer().emitIntValue(AsInt.getLimitedValue(),
4345 AsInt.getBitWidth() / 8);
4346 }
4347 // Default-initialize all remaining values.
4348 for (const auto &AsInt :
4349 llvm::drop_begin(Contents.AsIntValues, Initializer.AsIntValues.size())) {
4350 getStreamer().emitIntValue(AsInt.getLimitedValue(),
4351 AsInt.getBitWidth() / 8);
4352 }
4353 return false;
4354}
4355
4356bool MasmParser::emitFieldInitializer(const FieldInfo &Field,
4357 const StructFieldInfo &Contents,
4358 const StructFieldInfo &Initializer) {
4359 for (const auto &Init : Initializer.Initializers) {
4360 if (emitStructInitializer(Contents.Structure, Init))
4361 return true;
4362 }
4363 // Default-initialize all remaining values.
4364 for (const auto &Init : llvm::drop_begin(Contents.Initializers,
4365 Initializer.Initializers.size())) {
4366 if (emitStructInitializer(Contents.Structure, Init))
4367 return true;
4368 }
4369 return false;
4370}
4371
4372bool MasmParser::emitFieldInitializer(const FieldInfo &Field,
4373 const FieldInitializer &Initializer) {
4374 switch (Field.Contents.FT) {
4375 case FT_INTEGRAL:
4376 return emitFieldInitializer(Field, Field.Contents.IntInfo,
4377 Initializer.IntInfo);
4378 case FT_REAL:
4379 return emitFieldInitializer(Field, Field.Contents.RealInfo,
4380 Initializer.RealInfo);
4381 case FT_STRUCT:
4382 return emitFieldInitializer(Field, Field.Contents.StructInfo,
4383 Initializer.StructInfo);
4384 }
4385 llvm_unreachable("Unhandled FieldType enum");
4386}
4387
4388bool MasmParser::emitStructInitializer(const StructInfo &Structure,
4389 const StructInitializer &Initializer) {
4390 if (!Structure.Initializable)
4391 return Error(getLexer().getLoc(),
4392 "cannot initialize a value of type '" + Structure.Name +
4393 "'; 'org' was used in the type's declaration");
4394 size_t Index = 0, Offset = 0;
4395 for (const auto &Init : Initializer.FieldInitializers) {
4396 const auto &Field = Structure.Fields[Index++];
4397 getStreamer().emitZeros(Field.Offset - Offset);
4398 Offset = Field.Offset + Field.SizeOf;
4399 if (emitFieldInitializer(Field, Init))
4400 return true;
4401 }
4402 // Default-initialize all remaining fields.
4403 for (const auto &Field : llvm::drop_begin(
4404 Structure.Fields, Initializer.FieldInitializers.size())) {
4405 getStreamer().emitZeros(Field.Offset - Offset);
4406 Offset = Field.Offset + Field.SizeOf;
4407 if (emitFieldValue(Field))
4408 return true;
4409 }
4410 // Add final padding.
4411 if (Offset != Structure.Size)
4412 getStreamer().emitZeros(Structure.Size - Offset);
4413 return false;
4414}
4415
4416// Set data values from initializers.
4417bool MasmParser::emitStructValues(const StructInfo &Structure,
4418 unsigned *Count) {
4419 std::vector<StructInitializer> Initializers;
4420 if (parseStructInstList(Structure, Initializers))
4421 return true;
4422
4423 for (const auto &Initializer : Initializers) {
4424 if (emitStructInitializer(Structure, Initializer))
4425 return true;
4426 }
4427
4428 if (Count)
4429 *Count = Initializers.size();
4430 return false;
4431}
4432
4433// Declare a field in the current struct.
4434bool MasmParser::addStructField(StringRef Name, const StructInfo &Structure) {
4435 StructInfo &OwningStruct = StructInProgress.back();
4436 FieldInfo &Field =
4437 OwningStruct.addField(Name, FT_STRUCT, Structure.AlignmentSize);
4438 StructFieldInfo &StructInfo = Field.Contents.StructInfo;
4439
4440 StructInfo.Structure = Structure;
4441 Field.Type = Structure.Size;
4442
4443 if (parseStructInstList(Structure, StructInfo.Initializers))
4444 return true;
4445
4446 Field.LengthOf = StructInfo.Initializers.size();
4447 Field.SizeOf = Field.Type * Field.LengthOf;
4448
4449 const unsigned FieldEnd = Field.Offset + Field.SizeOf;
4450 if (!OwningStruct.IsUnion) {
4451 OwningStruct.NextOffset = FieldEnd;
4452 }
4453 OwningStruct.Size = std::max(OwningStruct.Size, FieldEnd);
4454
4455 return false;
4456}
4457
4458/// parseDirectiveStructValue
4459/// ::= struct-id (<struct-initializer> | {struct-initializer})
4460/// [, (<struct-initializer> | {struct-initializer})]*
4461bool MasmParser::parseDirectiveStructValue(const StructInfo &Structure,
4462 StringRef Directive, SMLoc DirLoc) {
4463 if (StructInProgress.empty()) {
4464 if (emitStructValues(Structure))
4465 return true;
4466 } else if (addStructField("", Structure)) {
4467 return addErrorSuffix(" in '" + Twine(Directive) + "' directive");
4468 }
4469
4470 return false;
4471}
4472
4473/// parseDirectiveNamedValue
4474/// ::= name (byte | word | ... ) [ expression (, expression)* ]
4475bool MasmParser::parseDirectiveNamedStructValue(const StructInfo &Structure,
4477 SMLoc DirLoc, StringRef Name) {
4478 if (StructInProgress.empty()) {
4479 // Initialize named data value.
4480 MCSymbol *Sym = getContext().getOrCreateSymbol(Name);
4481 getStreamer().emitLabel(Sym);
4482 unsigned Count;
4483 if (emitStructValues(Structure, &Count))
4484 return true;
4486 Type.Name = Structure.Name;
4487 Type.Size = Structure.Size * Count;
4488 Type.ElementSize = Structure.Size;
4489 Type.Length = Count;
4490 KnownType[Name.lower()] = Type;
4491 } else if (addStructField(Name, Structure)) {
4492 return addErrorSuffix(" in '" + Twine(Directive) + "' directive");
4493 }
4494
4495 return false;
4496}
4497
4498/// parseDirectiveStruct
4499/// ::= <name> (STRUC | STRUCT | UNION) [fieldAlign] [, NONUNIQUE]
4500/// (dataDir | generalDir | offsetDir | nestedStruct)+
4501/// <name> ENDS
4502////// dataDir = data declaration
4503////// offsetDir = EVEN, ORG, ALIGN
4504bool MasmParser::parseDirectiveStruct(StringRef Directive,
4505 DirectiveKind DirKind, StringRef Name,
4506 SMLoc NameLoc) {
4507 // We ignore NONUNIQUE; we do not support OPTION M510 or OPTION OLDSTRUCTS
4508 // anyway, so all field accesses must be qualified.
4509 AsmToken NextTok = getTok();
4510 int64_t AlignmentValue = 1;
4511 if (NextTok.isNot(AsmToken::Comma) &&
4513 parseAbsoluteExpression(AlignmentValue)) {
4514 return addErrorSuffix(" in alignment value for '" + Twine(Directive) +
4515 "' directive");
4516 }
4517 if (!isPowerOf2_64(AlignmentValue)) {
4518 return Error(NextTok.getLoc(), "alignment must be a power of two; was " +
4519 std::to_string(AlignmentValue));
4520 }
4521
4523 SMLoc QualifierLoc;
4524 if (parseOptionalToken(AsmToken::Comma)) {
4525 QualifierLoc = getTok().getLoc();
4526 if (parseIdentifier(Qualifier))
4527 return addErrorSuffix(" in '" + Twine(Directive) + "' directive");
4528 if (!Qualifier.equals_insensitive("nonunique"))
4529 return Error(QualifierLoc, "Unrecognized qualifier for '" +
4530 Twine(Directive) +
4531 "' directive; expected none or NONUNIQUE");
4532 }
4533
4534 if (parseEOL())
4535 return addErrorSuffix(" in '" + Twine(Directive) + "' directive");
4536
4537 StructInProgress.emplace_back(Name, DirKind == DK_UNION, AlignmentValue);
4538 return false;
4539}
4540
4541/// parseDirectiveNestedStruct
4542/// ::= (STRUC | STRUCT | UNION) [name]
4543/// (dataDir | generalDir | offsetDir | nestedStruct)+
4544/// ENDS
4545bool MasmParser::parseDirectiveNestedStruct(StringRef Directive,
4546 DirectiveKind DirKind) {
4547 if (StructInProgress.empty())
4548 return TokError("missing name in top-level '" + Twine(Directive) +
4549 "' directive");
4550
4552 if (getTok().is(AsmToken::Identifier)) {
4553 Name = getTok().getIdentifier();
4554 parseToken(AsmToken::Identifier);
4555 }
4556 if (parseEOL())
4557 return addErrorSuffix(" in '" + Twine(Directive) + "' directive");
4558
4559 // Reserve space to ensure Alignment doesn't get invalidated when
4560 // StructInProgress grows.
4561 StructInProgress.reserve(StructInProgress.size() + 1);
4562 StructInProgress.emplace_back(Name, DirKind == DK_UNION,
4563 StructInProgress.back().Alignment);
4564 return false;
4565}
4566
4567bool MasmParser::parseDirectiveEnds(StringRef Name, SMLoc NameLoc) {
4568 if (StructInProgress.empty())
4569 return Error(NameLoc, "ENDS directive without matching STRUC/STRUCT/UNION");
4570 if (StructInProgress.size() > 1)
4571 return Error(NameLoc, "unexpected name in nested ENDS directive");
4572 if (StructInProgress.back().Name.compare_insensitive(Name))
4573 return Error(NameLoc, "mismatched name in ENDS directive; expected '" +
4574 StructInProgress.back().Name + "'");
4575 StructInfo Structure = StructInProgress.pop_back_val();
4576 // Pad to make the structure's size divisible by the smaller of its alignment
4577 // and the size of its largest field.
4578 Structure.Size = llvm::alignTo(
4579 Structure.Size, std::min(Structure.Alignment, Structure.AlignmentSize));
4580 Structs[Name.lower()] = Structure;
4581
4582 if (parseEOL())
4583 return addErrorSuffix(" in ENDS directive");
4584
4585 return false;
4586}
4587
4588bool MasmParser::parseDirectiveNestedEnds() {
4589 if (StructInProgress.empty())
4590 return TokError("ENDS directive without matching STRUC/STRUCT/UNION");
4591 if (StructInProgress.size() == 1)
4592 return TokError("missing name in top-level ENDS directive");
4593
4594 if (parseEOL())
4595 return addErrorSuffix(" in nested ENDS directive");
4596
4597 StructInfo Structure = StructInProgress.pop_back_val();
4598 // Pad to make the structure's size divisible by its alignment.
4599 Structure.Size = llvm::alignTo(Structure.Size, Structure.Alignment);
4600
4601 StructInfo &ParentStruct = StructInProgress.back();
4602 if (Structure.Name.empty()) {
4603 // Anonymous substructures' fields are addressed as if they belong to the
4604 // parent structure - so we transfer them to the parent here.
4605 const size_t OldFields = ParentStruct.Fields.size();
4606 ParentStruct.Fields.insert(
4607 ParentStruct.Fields.end(),
4608 std::make_move_iterator(Structure.Fields.begin()),
4609 std::make_move_iterator(Structure.Fields.end()));
4610 for (const auto &FieldByName : Structure.FieldsByName) {
4611 ParentStruct.FieldsByName[FieldByName.getKey()] =
4612 FieldByName.getValue() + OldFields;
4613 }
4614
4615 unsigned FirstFieldOffset = 0;
4616 if (!Structure.Fields.empty() && !ParentStruct.IsUnion) {
4617 FirstFieldOffset = llvm::alignTo(
4618 ParentStruct.NextOffset,
4619 std::min(ParentStruct.Alignment, Structure.AlignmentSize));
4620 }
4621
4622 if (ParentStruct.IsUnion) {
4623 ParentStruct.Size = std::max(ParentStruct.Size, Structure.Size);
4624 } else {
4625 for (auto &Field : llvm::drop_begin(ParentStruct.Fields, OldFields))
4626 Field.Offset += FirstFieldOffset;
4627
4628 const unsigned StructureEnd = FirstFieldOffset + Structure.Size;
4629 if (!ParentStruct.IsUnion) {
4630 ParentStruct.NextOffset = StructureEnd;
4631 }
4632 ParentStruct.Size = std::max(ParentStruct.Size, StructureEnd);
4633 }
4634 } else {
4635 FieldInfo &Field = ParentStruct.addField(Structure.Name, FT_STRUCT,
4636 Structure.AlignmentSize);
4637 StructFieldInfo &StructInfo = Field.Contents.StructInfo;
4638 Field.Type = Structure.Size;
4639 Field.LengthOf = 1;
4640 Field.SizeOf = Structure.Size;
4641
4642 const unsigned StructureEnd = Field.Offset + Field.SizeOf;
4643 if (!ParentStruct.IsUnion) {
4644 ParentStruct.NextOffset = StructureEnd;
4645 }
4646 ParentStruct.Size = std::max(ParentStruct.Size, StructureEnd);
4647
4648 StructInfo.Structure = Structure;
4649 StructInfo.Initializers.emplace_back();
4650 auto &FieldInitializers = StructInfo.Initializers.back().FieldInitializers;
4651 for (const auto &SubField : Structure.Fields) {
4652 FieldInitializers.push_back(SubField.Contents);
4653 }
4654 }
4655
4656 return false;
4657}
4658
4659/// parseDirectiveOrg
4660/// ::= org expression
4661bool MasmParser::parseDirectiveOrg() {
4662 const MCExpr *Offset;
4663 SMLoc OffsetLoc = Lexer.getLoc();
4664 if (checkForValidSection() || parseExpression(Offset))
4665 return true;
4666 if (parseEOL())
4667 return addErrorSuffix(" in 'org' directive");
4668
4669 if (StructInProgress.empty()) {
4670 // Not in a struct; change the offset for the next instruction or data
4671 if (checkForValidSection())
4672 return addErrorSuffix(" in 'org' directive");
4673
4674 getStreamer().emitValueToOffset(Offset, 0, OffsetLoc);
4675 } else {
4676 // Offset the next field of this struct
4677 StructInfo &Structure = StructInProgress.back();
4678 int64_t OffsetRes;
4679 if (!Offset->evaluateAsAbsolute(OffsetRes, getStreamer().getAssemblerPtr()))
4680 return Error(OffsetLoc,
4681 "expected absolute expression in 'org' directive");
4682 if (OffsetRes < 0)
4683 return Error(
4684 OffsetLoc,
4685 "expected non-negative value in struct's 'org' directive; was " +
4686 std::to_string(OffsetRes));
4687 Structure.NextOffset = static_cast<unsigned>(OffsetRes);
4688
4689 // ORG-affected structures cannot be initialized
4690 Structure.Initializable = false;
4691 }
4692
4693 return false;
4694}
4695
4696bool MasmParser::emitAlignTo(int64_t Alignment) {
4697 if (StructInProgress.empty()) {
4698 // Not in a struct; align the next instruction or data
4699 if (checkForValidSection())
4700 return true;
4701
4702 // Check whether we should use optimal code alignment for this align
4703 // directive.
4704 const MCSection *Section = getStreamer().getCurrentSectionOnly();
4705 assert(Section && "must have section to emit alignment");
4706 if (Section->useCodeAlign()) {
4707 getStreamer().emitCodeAlignment(Align(Alignment),
4708 &getTargetParser().getSTI(),
4709 /*MaxBytesToEmit=*/0);
4710 } else {
4711 // FIXME: Target specific behavior about how the "extra" bytes are filled.
4712 getStreamer().emitValueToAlignment(Align(Alignment), /*Value=*/0,
4713 /*ValueSize=*/1,
4714 /*MaxBytesToEmit=*/0);
4715 }
4716 } else {
4717 // Align the next field of this struct
4718 StructInfo &Structure = StructInProgress.back();
4719 Structure.NextOffset = llvm::alignTo(Structure.NextOffset, Alignment);
4720 }
4721
4722 return false;
4723}
4724
4725/// parseDirectiveAlign
4726/// ::= align expression
4727bool MasmParser::parseDirectiveAlign() {
4728 SMLoc AlignmentLoc = getLexer().getLoc();
4729 int64_t Alignment;
4730
4731 // Ignore empty 'align' directives.
4732 if (getTok().is(AsmToken::EndOfStatement)) {
4733 return Warning(AlignmentLoc,
4734 "align directive with no operand is ignored") &&
4735 parseEOL();
4736 }
4737 if (parseAbsoluteExpression(Alignment) || parseEOL())
4738 return addErrorSuffix(" in align directive");
4739
4740 // Always emit an alignment here even if we throw an error.
4741 bool ReturnVal = false;
4742
4743 // Reject alignments that aren't either a power of two or zero, for ML.exe
4744 // compatibility. Alignment of zero is silently rounded up to one.
4745 if (Alignment == 0)
4746 Alignment = 1;
4747 if (!isPowerOf2_64(Alignment))
4748 ReturnVal |= Error(AlignmentLoc, "alignment must be a power of 2; was " +
4749 std::to_string(Alignment));
4750
4751 if (emitAlignTo(Alignment))
4752 ReturnVal |= addErrorSuffix(" in align directive");
4753
4754 return ReturnVal;
4755}
4756
4757/// parseDirectiveEven
4758/// ::= even
4759bool MasmParser::parseDirectiveEven() {
4760 if (parseEOL() || emitAlignTo(2))
4761 return addErrorSuffix(" in even directive");
4762
4763 return false;
4764}
4765
4766/// parseDirectiveFile
4767/// ::= .file filename
4768/// ::= .file number [directory] filename [md5 checksum] [source source-text]
4769bool MasmParser::parseDirectiveFile(SMLoc DirectiveLoc) {
4770 // FIXME: I'm not sure what this is.
4771 int64_t FileNumber = -1;
4772 if (getLexer().is(AsmToken::Integer)) {
4773 FileNumber = getTok().getIntVal();
4774 Lex();
4775
4776 if (FileNumber < 0)
4777 return TokError("negative file number");
4778 }
4779
4780 std::string Path;
4781
4782 // Usually the directory and filename together, otherwise just the directory.
4783 // Allow the strings to have escaped octal character sequence.
4784 if (check(getTok().isNot(AsmToken::String),
4785 "unexpected token in '.file' directive") ||
4786 parseEscapedString(Path))
4787 return true;
4788
4789 StringRef Directory;
4791 std::string FilenameData;
4792 if (getLexer().is(AsmToken::String)) {
4793 if (check(FileNumber == -1,
4794 "explicit path specified, but no file number") ||
4795 parseEscapedString(FilenameData))
4796 return true;
4797 Filename = FilenameData;
4798 Directory = Path;
4799 } else {
4800 Filename = Path;
4801 }
4802
4803 uint64_t MD5Hi, MD5Lo;
4804 bool HasMD5 = false;
4805
4806 std::optional<StringRef> Source;
4807 bool HasSource = false;
4808 std::string SourceString;
4809
4810 while (!parseOptionalToken(AsmToken::EndOfStatement)) {
4812 if (check(getTok().isNot(AsmToken::Identifier),
4813 "unexpected token in '.file' directive") ||
4814 parseIdentifier(Keyword))
4815 return true;
4816 if (Keyword == "md5") {
4817 HasMD5 = true;
4818 if (check(FileNumber == -1,
4819 "MD5 checksum specified, but no file number") ||
4820 parseHexOcta(*this, MD5Hi, MD5Lo))
4821 return true;
4822 } else if (Keyword == "source") {
4823 HasSource = true;
4824 if (check(FileNumber == -1,
4825 "source specified, but no file number") ||
4826 check(getTok().isNot(AsmToken::String),
4827 "unexpected token in '.file' directive") ||
4828 parseEscapedString(SourceString))
4829 return true;
4830 } else {
4831 return TokError("unexpected token in '.file' directive");
4832 }
4833 }
4834
4835 if (FileNumber == -1) {
4836 // Ignore the directive if there is no number and the target doesn't support
4837 // numberless .file directives. This allows some portability of assembler
4838 // between different object file formats.
4839 if (getContext().getAsmInfo()->hasSingleParameterDotFile())
4840 getStreamer().emitFileDirective(Filename);
4841 } else {
4842 // In case there is a -g option as well as debug info from directive .file,
4843 // we turn off the -g option, directly use the existing debug info instead.
4844 // Throw away any implicit file table for the assembler source.
4845 if (Ctx.getGenDwarfForAssembly()) {
4847 Ctx.setGenDwarfForAssembly(false);
4848 }
4849
4850 std::optional<MD5::MD5Result> CKMem;
4851 if (HasMD5) {
4852 MD5::MD5Result Sum;
4853 for (unsigned i = 0; i != 8; ++i) {
4854 Sum[i] = uint8_t(MD5Hi >> ((7 - i) * 8));
4855 Sum[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8));
4856 }
4857 CKMem = Sum;
4858 }
4859 if (HasSource) {
4860 char *SourceBuf = static_cast<char *>(Ctx.allocate(SourceString.size()));
4861 memcpy(SourceBuf, SourceString.data(), SourceString.size());
4862 Source = StringRef(SourceBuf, SourceString.size());
4863 }
4864 if (FileNumber == 0) {
4865 if (Ctx.getDwarfVersion() < 5)
4866 return Warning(DirectiveLoc, "file 0 not supported prior to DWARF-5");
4867 getStreamer().emitDwarfFile0Directive(Directory, Filename, CKMem, Source);
4868 } else {
4869 Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective(
4870 FileNumber, Directory, Filename, CKMem, Source);
4871 if (!FileNumOrErr)
4872 return Error(DirectiveLoc, toString(FileNumOrErr.takeError()));
4873 }
4874 // Alert the user if there are some .file directives with MD5 and some not.
4875 // But only do that once.
4876 if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(0)) {
4877 ReportedInconsistentMD5 = true;
4878 return Warning(DirectiveLoc, "inconsistent use of MD5 checksums");
4879 }
4880 }
4881
4882 return false;
4883}
4884
4885/// parseDirectiveLine
4886/// ::= .line [number]
4887bool MasmParser::parseDirectiveLine() {
4888 int64_t LineNumber;
4889 if (getLexer().is(AsmToken::Integer)) {
4890 if (parseIntToken(LineNumber, "unexpected token in '.line' directive"))
4891 return true;
4892 (void)LineNumber;
4893 // FIXME: Do something with the .line.
4894 }
4895 if (parseEOL())
4896 return true;
4897
4898 return false;
4899}
4900
4901/// parseDirectiveLoc
4902/// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end]
4903/// [epilogue_begin] [is_stmt VALUE] [isa VALUE]
4904/// The first number is a file number, must have been previously assigned with
4905/// a .file directive, the second number is the line number and optionally the
4906/// third number is a column position (zero if not specified). The remaining
4907/// optional items are .loc sub-directives.
4908bool MasmParser::parseDirectiveLoc() {
4909 int64_t FileNumber = 0, LineNumber = 0;
4910 SMLoc Loc = getTok().getLoc();
4911 if (parseIntToken(FileNumber, "unexpected token in '.loc' directive") ||
4912 check(FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc,
4913 "file number less than one in '.loc' directive") ||
4914 check(!getContext().isValidDwarfFileNumber(FileNumber), Loc,
4915 "unassigned file number in '.loc' directive"))
4916 return true;
4917
4918 // optional
4919 if (getLexer().is(AsmToken::Integer)) {
4920 LineNumber = getTok().getIntVal();
4921 if (LineNumber < 0)
4922 return TokError("line number less than zero in '.loc' directive");
4923 Lex();
4924 }
4925
4926 int64_t ColumnPos = 0;
4927 if (getLexer().is(AsmToken::Integer)) {
4928 ColumnPos = getTok().getIntVal();
4929 if (ColumnPos < 0)
4930 return TokError("column position less than zero in '.loc' directive");
4931 Lex();
4932 }
4933
4934 auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags();
4935 unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT;
4936 unsigned Isa = 0;
4937 int64_t Discriminator = 0;
4938
4939 auto parseLocOp = [&]() -> bool {
4941 SMLoc Loc = getTok().getLoc();
4942 if (parseIdentifier(Name))
4943 return TokError("unexpected token in '.loc' directive");
4944
4945 if (Name == "basic_block")
4947 else if (Name == "prologue_end")
4949 else if (Name == "epilogue_begin")
4951 else if (Name == "is_stmt") {
4952 Loc = getTok().getLoc();
4953 const MCExpr *Value;
4954 if (parseExpression(Value))
4955 return true;
4956 // The expression must be the constant 0 or 1.
4957 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
4958 int Value = MCE->getValue();
4959 if (Value == 0)
4960 Flags &= ~DWARF2_FLAG_IS_STMT;
4961 else if (Value == 1)
4963 else
4964 return Error(Loc, "is_stmt value not 0 or 1");
4965 } else {
4966 return Error(Loc, "is_stmt value not the constant value of 0 or 1");
4967 }
4968 } else if (Name == "isa") {
4969 Loc = getTok().getLoc();
4970 const MCExpr *Value;
4971 if (parseExpression(Value))
4972 return true;
4973 // The expression must be a constant greater or equal to 0.
4974 if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Value)) {
4975 int Value = MCE->getValue();
4976 if (Value < 0)
4977 return Error(Loc, "isa number less than zero");
4978 Isa = Value;
4979 } else {
4980 return Error(Loc, "isa number not a constant value");
4981 }
4982 } else if (Name == "discriminator") {
4983 if (parseAbsoluteExpression(Discriminator))
4984 return true;
4985 } else {
4986 return Error(Loc, "unknown sub-directive in '.loc' directive");
4987 }
4988 return false;
4989 };
4990
4991 if (parseMany(parseLocOp, false /*hasComma*/))
4992 return true;
4993
4994 getStreamer().emitDwarfLocDirective(FileNumber, LineNumber, ColumnPos, Flags,
4995 Isa, Discriminator, StringRef());
4996
4997 return false;
4998}
4999
5000/// parseDirectiveStabs
5001/// ::= .stabs string, number, number, number
5002bool MasmParser::parseDirectiveStabs() {
5003 return TokError("unsupported directive '.stabs'");
5004}
5005
5006/// parseDirectiveCVFile
5007/// ::= .cv_file number filename [checksum] [checksumkind]
5008bool MasmParser::parseDirectiveCVFile() {
5009 SMLoc FileNumberLoc = getTok().getLoc();
5010 int64_t FileNumber;
5011 std::string Filename;
5012 std::string Checksum;
5013 int64_t ChecksumKind = 0;
5014
5015 if (parseIntToken(FileNumber,
5016 "expected file number in '.cv_file' directive") ||
5017 check(FileNumber < 1, FileNumberLoc, "file number less than one") ||
5018 check(getTok().isNot(AsmToken::String),
5019 "unexpected token in '.cv_file' directive") ||
5020 parseEscapedString(Filename))
5021 return true;
5022 if (!parseOptionalToken(AsmToken::EndOfStatement)) {
5023 if (check(getTok().isNot(AsmToken::String),
5024 "unexpected token in '.cv_file' directive") ||
5025 parseEscapedString(Checksum) ||
5026 parseIntToken(ChecksumKind,
5027 "expected checksum kind in '.cv_file' directive") ||
5028 parseEOL())
5029 return true;
5030 }
5031
5032 Checksum = fromHex(Checksum);
5033 void *CKMem = Ctx.allocate(Checksum.size(), 1);
5034 memcpy(CKMem, Checksum.data(), Checksum.size());
5035 ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem),
5036 Checksum.size());
5037
5038 if (!getStreamer().emitCVFileDirective(FileNumber, Filename, ChecksumAsBytes,
5039 static_cast<uint8_t>(ChecksumKind)))
5040 return Error(FileNumberLoc, "file number already allocated");
5041
5042 return false;
5043}
5044
5045bool MasmParser::parseCVFunctionId(int64_t &FunctionId,
5046 StringRef DirectiveName) {
5047 SMLoc Loc;
5048 return parseTokenLoc(Loc) ||
5049 parseIntToken(FunctionId, "expected function id in '" + DirectiveName +
5050 "' directive") ||
5051 check(FunctionId < 0 || FunctionId >= UINT_MAX, Loc,
5052 "expected function id within range [0, UINT_MAX)");
5053}
5054
5055bool MasmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) {
5056 SMLoc Loc;
5057 return parseTokenLoc(Loc) ||
5058 parseIntToken(FileNumber, "expected integer in '" + DirectiveName +
5059 "' directive") ||
5060 check(FileNumber < 1, Loc, "file number less than one in '" +
5061 DirectiveName + "' directive") ||
5062 check(!getCVContext().isValidFileNumber(FileNumber), Loc,
5063 "unassigned file number in '" + DirectiveName + "' directive");
5064}
5065
5066/// parseDirectiveCVFuncId
5067/// ::= .cv_func_id FunctionId
5068///
5069/// Introduces a function ID that can be used with .cv_loc.
5070bool MasmParser::parseDirectiveCVFuncId() {
5071 SMLoc FunctionIdLoc = getTok().getLoc();
5072 int64_t FunctionId;
5073
5074 if (parseCVFunctionId(FunctionId, ".cv_func_id") || parseEOL())
5075 return true;
5076
5077 if (!getStreamer().emitCVFuncIdDirective(FunctionId))
5078 return Error(FunctionIdLoc, "function id already allocated");
5079
5080 return false;
5081}
5082
5083/// parseDirectiveCVInlineSiteId
5084/// ::= .cv_inline_site_id FunctionId
5085/// "within" IAFunc
5086/// "inlined_at" IAFile IALine [IACol]
5087///
5088/// Introduces a function ID that can be used with .cv_loc. Includes "inlined
5089/// at" source location information for use in the line table of the caller,
5090/// whether the caller is a real function or another inlined call site.
5091bool MasmParser::parseDirectiveCVInlineSiteId() {
5092 SMLoc FunctionIdLoc = getTok().getLoc();
5093 int64_t FunctionId;
5094 int64_t IAFunc;
5095 int64_t IAFile;
5096 int64_t IALine;
5097 int64_t IACol = 0;
5098
5099 // FunctionId
5100 if (parseCVFunctionId(FunctionId, ".cv_inline_site_id"))
5101 return true;
5102
5103 // "within"
5104 if (check((getLexer().isNot(AsmToken::Identifier) ||
5105 getTok().getIdentifier() != "within"),
5106 "expected 'within' identifier in '.cv_inline_site_id' directive"))
5107 return true;
5108 Lex();
5109
5110 // IAFunc
5111 if (parseCVFunctionId(IAFunc, ".cv_inline_site_id"))
5112 return true;
5113
5114 // "inlined_at"
5115 if (check((getLexer().isNot(AsmToken::Identifier) ||
5116 getTok().getIdentifier() != "inlined_at"),
5117 "expected 'inlined_at' identifier in '.cv_inline_site_id' "
5118 "directive") )
5119 return true;
5120 Lex();
5121
5122 // IAFile IALine
5123 if (parseCVFileId(IAFile, ".cv_inline_site_id") ||
5124 parseIntToken(IALine, "expected line number after 'inlined_at'"))
5125 return true;
5126
5127 // [IACol]
5128 if (getLexer().is(AsmToken::Integer)) {
5129 IACol = getTok().getIntVal();
5130 Lex();
5131 }
5132
5133 if (parseEOL())
5134 return true;
5135
5136 if (!getStreamer().emitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile,
5137 IALine, IACol, FunctionIdLoc))
5138 return Error(FunctionIdLoc, "function id already allocated");
5139
5140 return false;
5141}
5142
5143/// parseDirectiveCVLoc
5144/// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end]
5145/// [is_stmt VALUE]
5146/// The first number is a file number, must have been previously assigned with
5147/// a .file directive, the second number is the line number and optionally the
5148/// third number is a column position (zero if not specified). The remaining
5149/// optional items are .loc sub-directives.
5150bool MasmParser::parseDirectiveCVLoc() {
5151 SMLoc DirectiveLoc = getTok().getLoc();
5152 int64_t FunctionId, FileNumber;
5153 if (parseCVFunctionId(FunctionId, ".cv_loc") ||
5154 parseCVFileId(FileNumber, ".cv_loc"))
5155 return true;
5156
5157 int64_t LineNumber = 0;
5158 if (getLexer().is(AsmToken::Integer)) {
5159 LineNumber = getTok().getIntVal();
5160 if (LineNumber < 0)
5161 return TokError("line number less than zero in '.cv_loc' directive");
5162 Lex();
5163 }
5164
5165 int64_t ColumnPos = 0;
5166 if (getLexer().is(AsmToken::Integer)) {
5167 ColumnPos = getTok().getIntVal();
5168 if (ColumnPos < 0)
5169 return TokError("column position less than zero in '.cv_loc' directive");
5170 Lex();
5171 }
5172
5173 bool PrologueEnd = false;
5174 uint64_t IsStmt = 0;
5175
5176 auto parseOp = [&]() -> bool {
5178 SMLoc Loc = getTok().getLoc();
5179 if (parseIdentifier(Name))
5180 return TokError("unexpected token in '.cv_loc' directive");
5181 if (Name == "prologue_end")
5182 PrologueEnd = true;
5183 else if (Name == "is_stmt") {
5184 Loc = getTok().getLoc();
5185 const MCExpr *Value;
5186 if (parseExpression(Value))
5187 return true;
5188 // The expression must be the constant 0 or 1.
5189 IsStmt = ~0ULL;
5190 if (const auto *MCE = dyn_cast<MCConstantExpr>(Value))
5191 IsStmt = MCE->getValue();
5192
5193 if (IsStmt > 1)
5194 return Error(Loc, "is_stmt value not 0 or 1");
5195 } else {
5196 return Error(Loc, "unknown sub-directive in '.cv_loc' directive");
5197 }
5198 return false;
5199 };
5200
5201 if (parseMany(parseOp, false /*hasComma*/))
5202 return true;
5203
5204 getStreamer().emitCVLocDirective(FunctionId, FileNumber, LineNumber,
5205 ColumnPos, PrologueEnd, IsStmt, StringRef(),
5206 DirectiveLoc);
5207 return false;
5208}
5209
5210/// parseDirectiveCVLinetable
5211/// ::= .cv_linetable FunctionId, FnStart, FnEnd
5212bool MasmParser::parseDirectiveCVLinetable() {
5213 int64_t FunctionId;
5214 StringRef FnStartName, FnEndName;
5215 SMLoc Loc = getTok().getLoc();
5216 if (parseCVFunctionId(FunctionId, ".cv_linetable") ||
5217 parseToken(AsmToken::Comma,
5218 "unexpected token in '.cv_linetable' directive") ||
5219 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc,
5220 "expected identifier in directive") ||
5221 parseToken(AsmToken::Comma,
5222 "unexpected token in '.cv_linetable' directive") ||
5223 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc,
5224 "expected identifier in directive"))
5225 return true;
5226
5227 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
5228 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
5229
5230 getStreamer().emitCVLinetableDirective(FunctionId, FnStartSym, FnEndSym);
5231 return false;
5232}
5233
5234/// parseDirectiveCVInlineLinetable
5235/// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd
5236bool MasmParser::parseDirectiveCVInlineLinetable() {
5237 int64_t PrimaryFunctionId, SourceFileId, SourceLineNum;
5238 StringRef FnStartName, FnEndName;
5239 SMLoc Loc = getTok().getLoc();
5240 if (parseCVFunctionId(PrimaryFunctionId, ".cv_inline_linetable") ||
5241 parseTokenLoc(Loc) ||
5242 parseIntToken(
5243 SourceFileId,
5244 "expected SourceField in '.cv_inline_linetable' directive") ||
5245 check(SourceFileId <= 0, Loc,
5246 "File id less than zero in '.cv_inline_linetable' directive") ||
5247 parseTokenLoc(Loc) ||
5248 parseIntToken(
5249 SourceLineNum,
5250 "expected SourceLineNum in '.cv_inline_linetable' directive") ||
5251 check(SourceLineNum < 0, Loc,
5252 "Line number less than zero in '.cv_inline_linetable' directive") ||
5253 parseTokenLoc(Loc) || check(parseIdentifier(FnStartName), Loc,
5254 "expected identifier in directive") ||
5255 parseTokenLoc(Loc) || check(parseIdentifier(FnEndName), Loc,
5256 "expected identifier in directive"))
5257 return true;
5258
5259 if (parseEOL())
5260 return true;
5261
5262 MCSymbol *FnStartSym = getContext().getOrCreateSymbol(FnStartName);
5263 MCSymbol *FnEndSym = getContext().getOrCreateSymbol(FnEndName);
5264 getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId,
5265 SourceLineNum, FnStartSym,
5266 FnEndSym);
5267 return false;
5268}
5269
5270void MasmParser::initializeCVDefRangeTypeMap() {
5271 CVDefRangeTypeMap["reg"] = CVDR_DEFRANGE_REGISTER;
5272 CVDefRangeTypeMap["frame_ptr_rel"] = CVDR_DEFRANGE_FRAMEPOINTER_REL;
5273 CVDefRangeTypeMap["subfield_reg"] = CVDR_DEFRANGE_SUBFIELD_REGISTER;
5274 CVDefRangeTypeMap["reg_rel"] = CVDR_DEFRANGE_REGISTER_REL;
5275}
5276
5277/// parseDirectiveCVDefRange
5278/// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes*
5279bool MasmParser::parseDirectiveCVDefRange() {
5280 SMLoc Loc;
5281 std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges;
5282 while (getLexer().is(AsmToken::Identifier)) {
5283 Loc = getLexer().getLoc();
5284 StringRef GapStartName;
5285 if (parseIdentifier(GapStartName))
5286 return Error(Loc, "expected identifier in directive");
5287 MCSymbol *GapStartSym = getContext().getOrCreateSymbol(GapStartName);
5288
5289 Loc = getLexer().getLoc();
5290 StringRef GapEndName;
5291 if (parseIdentifier(GapEndName))
5292 return Error(Loc, "expected identifier in directive");
5293 MCSymbol *GapEndSym = getContext().getOrCreateSymbol(GapEndName);
5294
5295 Ranges.push_back({GapStartSym, GapEndSym});
5296 }
5297
5298 StringRef CVDefRangeTypeStr;
5299 if (parseToken(
5301 "expected comma before def_range type in .cv_def_range directive") ||
5302 parseIdentifier(CVDefRangeTypeStr))
5303 return Error(Loc, "expected def_range type in directive");
5304
5306 CVDefRangeTypeMap.find(CVDefRangeTypeStr);
5307 CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end())
5308 ? CVDR_DEFRANGE
5309 : CVTypeIt->getValue();
5310 switch (CVDRType) {
5311 case CVDR_DEFRANGE_REGISTER: {
5312 int64_t DRRegister;
5313 if (parseToken(AsmToken::Comma, "expected comma before register number in "
5314 ".cv_def_range directive") ||
5315 parseAbsoluteExpression(DRRegister))
5316 return Error(Loc, "expected register number");
5317
5319 DRHdr.Register = DRRegister;
5320 DRHdr.MayHaveNoName = 0;
5321 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
5322 break;
5323 }
5324 case CVDR_DEFRANGE_FRAMEPOINTER_REL: {
5325 int64_t DROffset;
5326 if (parseToken(AsmToken::Comma,
5327 "expected comma before offset in .cv_def_range directive") ||
5328 parseAbsoluteExpression(DROffset))
5329 return Error(Loc, "expected offset value");
5330
5332 DRHdr.Offset = DROffset;
5333 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
5334 break;
5335 }
5336 case CVDR_DEFRANGE_SUBFIELD_REGISTER: {
5337 int64_t DRRegister;
5338 int64_t DROffsetInParent;
5339 if (parseToken(AsmToken::Comma, "expected comma before register number in "
5340 ".cv_def_range directive") ||
5341 parseAbsoluteExpression(DRRegister))
5342 return Error(Loc, "expected register number");
5343 if (parseToken(AsmToken::Comma,
5344 "expected comma before offset in .cv_def_range directive") ||
5345 parseAbsoluteExpression(DROffsetInParent))
5346 return Error(Loc, "expected offset value");
5347
5349 DRHdr.Register = DRRegister;
5350 DRHdr.MayHaveNoName = 0;
5351 DRHdr.OffsetInParent = DROffsetInParent;
5352 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
5353 break;
5354 }
5355 case CVDR_DEFRANGE_REGISTER_REL: {
5356 int64_t DRRegister;
5357 int64_t DRFlags;
5358 int64_t DRBasePointerOffset;
5359 if (parseToken(AsmToken::Comma, "expected comma before register number in "
5360 ".cv_def_range directive") ||
5361 parseAbsoluteExpression(DRRegister))
5362 return Error(Loc, "expected register value");
5363 if (parseToken(
5365 "expected comma before flag value in .cv_def_range directive") ||
5366 parseAbsoluteExpression(DRFlags))
5367 return Error(Loc, "expected flag value");
5368 if (parseToken(AsmToken::Comma, "expected comma before base pointer offset "
5369 "in .cv_def_range directive") ||
5370 parseAbsoluteExpression(DRBasePointerOffset))
5371 return Error(Loc, "expected base pointer offset value");
5372
5374 DRHdr.Register = DRRegister;
5375 DRHdr.Flags = DRFlags;
5376 DRHdr.BasePointerOffset = DRBasePointerOffset;
5377 getStreamer().emitCVDefRangeDirective(Ranges, DRHdr);
5378 break;
5379 }
5380 default:
5381 return Error(Loc, "unexpected def_range type in .cv_def_range directive");
5382 }
5383 return true;
5384}
5385
5386/// parseDirectiveCVString
5387/// ::= .cv_stringtable "string"
5388bool MasmParser::parseDirectiveCVString() {
5389 std::string Data;
5390 if (checkForValidSection() || parseEscapedString(Data))
5391 return addErrorSuffix(" in '.cv_string' directive");
5392
5393 // Put the string in the table and emit the offset.
5394 std::pair<StringRef, unsigned> Insertion =
5395 getCVContext().addToStringTable(Data);
5396 getStreamer().emitIntValue(Insertion.second, 4);
5397 return false;
5398}
5399
5400/// parseDirectiveCVStringTable
5401/// ::= .cv_stringtable
5402bool MasmParser::parseDirectiveCVStringTable() {
5403 getStreamer().emitCVStringTableDirective();
5404 return false;
5405}
5406
5407/// parseDirectiveCVFileChecksums
5408/// ::= .cv_filechecksums
5409bool MasmParser::parseDirectiveCVFileChecksums() {
5410 getStreamer().emitCVFileChecksumsDirective();
5411 return false;
5412}
5413
5414/// parseDirectiveCVFileChecksumOffset
5415/// ::= .cv_filechecksumoffset fileno
5416bool MasmParser::parseDirectiveCVFileChecksumOffset() {
5417 int64_t FileNo;
5418 if (parseIntToken(FileNo, "expected identifier in directive"))
5419 return true;
5420 if (parseEOL())
5421 return true;
5422 getStreamer().emitCVFileChecksumOffsetDirective(FileNo);
5423 return false;
5424}
5425
5426/// parseDirectiveCVFPOData
5427/// ::= .cv_fpo_data procsym
5428bool MasmParser::parseDirectiveCVFPOData() {
5429 SMLoc DirLoc = getLexer().getLoc();
5430 StringRef ProcName;
5431 if (parseIdentifier(ProcName))
5432 return TokError("expected symbol name");
5433 if (parseEOL("unexpected tokens"))
5434 return addErrorSuffix(" in '.cv_fpo_data' directive");
5435 MCSymbol *ProcSym = getContext().getOrCreateSymbol(ProcName);
5436 getStreamer().emitCVFPOData(ProcSym, DirLoc);
5437 return false;
5438}
5439
5440/// parseDirectiveCFISections
5441/// ::= .cfi_sections section [, section]
5442bool MasmParser::parseDirectiveCFISections() {
5443