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" |
18 | #include "llvm/ADT/SmallString.h" |
19 | #include "llvm/ADT/SmallVector.h" |
20 | #include "llvm/ADT/StringExtras.h" |
21 | #include "llvm/ADT/StringMap.h" |
22 | #include "llvm/ADT/StringRef.h" |
23 | #include "llvm/ADT/StringSwitch.h" |
24 | #include "llvm/ADT/Twine.h" |
25 | #include "llvm/BinaryFormat/Dwarf.h" |
26 | #include "llvm/DebugInfo/CodeView/SymbolRecord.h" |
27 | #include "llvm/MC/MCAsmInfo.h" |
28 | #include "llvm/MC/MCCodeView.h" |
29 | #include "llvm/MC/MCContext.h" |
30 | #include "llvm/MC/MCDirectives.h" |
31 | #include "llvm/MC/MCDwarf.h" |
32 | #include "llvm/MC/MCExpr.h" |
33 | #include "llvm/MC/MCInstPrinter.h" |
34 | #include "llvm/MC/MCInstrDesc.h" |
35 | #include "llvm/MC/MCInstrInfo.h" |
36 | #include "llvm/MC/MCParser/AsmCond.h" |
37 | #include "llvm/MC/MCParser/AsmLexer.h" |
38 | #include "llvm/MC/MCParser/MCAsmLexer.h" |
39 | #include "llvm/MC/MCParser/MCAsmParser.h" |
40 | #include "llvm/MC/MCParser/MCAsmParserExtension.h" |
41 | #include "llvm/MC/MCParser/MCParsedAsmOperand.h" |
42 | #include "llvm/MC/MCParser/MCTargetAsmParser.h" |
43 | #include "llvm/MC/MCRegisterInfo.h" |
44 | #include "llvm/MC/MCSection.h" |
45 | #include "llvm/MC/MCStreamer.h" |
46 | #include "llvm/MC/MCSubtargetInfo.h" |
47 | #include "llvm/MC/MCSymbol.h" |
48 | #include "llvm/MC/MCTargetOptions.h" |
49 | #include "llvm/Support/Casting.h" |
50 | #include "llvm/Support/CommandLine.h" |
51 | #include "llvm/Support/ErrorHandling.h" |
52 | #include "llvm/Support/Format.h" |
53 | #include "llvm/Support/MD5.h" |
54 | #include "llvm/Support/MathExtras.h" |
55 | #include "llvm/Support/MemoryBuffer.h" |
56 | #include "llvm/Support/Path.h" |
57 | #include "llvm/Support/SMLoc.h" |
58 | #include "llvm/Support/SourceMgr.h" |
59 | #include "llvm/Support/raw_ostream.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 | |
75 | using namespace llvm; |
76 | |
77 | namespace { |
78 | |
79 | /// Helper types for tracking macro definitions. |
80 | typedef std::vector<AsmToken> MCAsmMacroArgument; |
81 | typedef std::vector<MCAsmMacroArgument> MCAsmMacroArguments; |
82 | |
83 | /// Helper class for storing information about an active macro instantiation. |
84 | struct 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 | |
98 | struct ParseStatementInfo { |
99 | /// The parsed operands from the last parsed statement. |
100 | SmallVector<std::unique_ptr<MCParsedAsmOperand>, 8> ParsedOperands; |
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 | |
118 | enum 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 | |
124 | struct FieldInfo; |
125 | struct StructInfo { |
126 | StringRef Name; |
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 | |
148 | struct StructInitializer; |
149 | struct IntFieldInfo { |
150 | SmallVector<const MCExpr *, 1> Values; |
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 | }; |
156 | struct 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 | }; |
163 | struct StructFieldInfo { |
164 | std::vector<StructInitializer> Initializers; |
165 | StructInfo Structure; |
166 | |
167 | StructFieldInfo() = default; |
168 | StructFieldInfo(std::vector<StructInitializer> V, StructInfo S); |
169 | }; |
170 | |
171 | class FieldInitializer { |
172 | public: |
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 | |
195 | struct StructInitializer { |
196 | std::vector<FieldInitializer> FieldInitializers; |
197 | }; |
198 | |
199 | struct 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 | |
217 | StructFieldInfo::StructFieldInfo(std::vector<StructInitializer> V, |
218 | StructInfo S) { |
219 | Initializers = std::move(V); |
220 | Structure = S; |
221 | } |
222 | |
223 | StructInfo::StructInfo(StringRef StructName, bool Union, |
224 | unsigned AlignmentValue) |
225 | : Name(StructName), IsUnion(Union), Alignment(AlignmentValue) {} |
226 | |
227 | FieldInfo &StructInfo::addField(StringRef FieldName, FieldType FT, |
228 | unsigned FieldAlignmentSize) { |
229 | if (!FieldName.empty()) |
230 | FieldsByName[FieldName.lower()] = Fields.size(); |
231 | Fields.emplace_back(args&: FT); |
232 | FieldInfo &Field = Fields.back(); |
233 | Field.Offset = |
234 | llvm::alignTo(Value: NextOffset, Align: std::min(a: Alignment, b: FieldAlignmentSize)); |
235 | if (!IsUnion) { |
236 | NextOffset = std::max(a: NextOffset, b: Field.Offset); |
237 | } |
238 | AlignmentSize = std::max(a: AlignmentSize, b: FieldAlignmentSize); |
239 | return Field; |
240 | } |
241 | |
242 | FieldInitializer::~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 | |
256 | FieldInitializer::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 | |
270 | FieldInitializer::FieldInitializer(SmallVector<const MCExpr *, 1> &&Values) |
271 | : FT(FT_INTEGRAL) { |
272 | new (&IntInfo) IntFieldInfo(std::move(Values)); |
273 | } |
274 | |
275 | FieldInitializer::FieldInitializer(SmallVector<APInt, 1> &&AsIntValues) |
276 | : FT(FT_REAL) { |
277 | new (&RealInfo) RealFieldInfo(std::move(AsIntValues)); |
278 | } |
279 | |
280 | FieldInitializer::FieldInitializer( |
281 | std::vector<StructInitializer> &&Initializers, struct StructInfo Structure) |
282 | : FT(FT_STRUCT) { |
283 | new (&StructInfo) StructFieldInfo(std::move(Initializers), Structure); |
284 | } |
285 | |
286 | FieldInitializer::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 | |
301 | FieldInitializer::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 | |
316 | FieldInitializer & |
317 | FieldInitializer::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 | |
346 | FieldInitializer &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. |
378 | class MasmParser : public MCAsmParser { |
379 | private: |
380 | AsmLexer Lexer; |
381 | MCContext &Ctx; |
382 | MCStreamer &Out; |
383 | const MCAsmInfo &MAI; |
384 | SourceMgr &SrcMgr; |
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 | |
410 | StringRef Name; |
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 { |
437 | StringRef Filename; |
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. |
449 | SmallVector<std::tuple<SMLoc, CppHashInfoTy, MCSymbol *>, 4> DirLabels; |
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 | |
470 | public: |
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 | |
479 | void addDirectiveHandler(StringRef Directive, |
480 | ExtensionDirectiveHandler Handler) override { |
481 | ExtensionDirectiveMap[Directive] = Handler; |
482 | if (!DirectiveKindMap.contains(Key: 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(ExpandNextToken: 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, |
543 | SmallVectorImpl<std::string> &Clobbers, |
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, Position: StandardPosition); |
566 | } |
567 | void eatToEndOfStatement() override; |
568 | |
569 | bool checkForValidSection() override; |
570 | |
571 | /// } |
572 | |
573 | private: |
574 | bool expandMacros(); |
575 | const AsmToken peekTok(bool ShouldSkipSpace = true); |
576 | |
577 | bool parseStatement(ParseStatementInfo &Info, |
578 | MCAsmParserSemaCallback *SI); |
579 | bool parseCurlyBlockScope(SmallVectorImpl<AsmRewrite>& AsmStrRewrites); |
580 | bool parseCppHashLineFilenameComment(SMLoc L); |
581 | |
582 | bool expandMacro(raw_svector_ostream &OS, StringRef Body, |
583 | ArrayRef<MCAsmMacroParameter> Parameters, |
584 | ArrayRef<MCAsmMacroArgument> A, |
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, |
596 | AsmToken::TokenKind ArgumentEndTok = AsmToken::EndOfStatement); |
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, |
610 | AsmToken::TokenKind EndTok = AsmToken::EndOfStatement); |
611 | |
612 | /// Parse all macro arguments for a given macro. |
613 | bool |
614 | parseMacroArguments(const MCAsmMacro *M, MCAsmMacroArguments &A, |
615 | AsmToken::TokenKind EndTok = AsmToken::EndOfStatement); |
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 { |
623 | ArrayRef<SMRange> Ranges(Range); |
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. |
658 | StringRef parseStringToEndOfStatement() override; |
659 | |
660 | bool parseTextItem(std::string &Data); |
661 | |
662 | unsigned getBinOpPrecedence(AsmToken::TokenKind K, |
663 | MCBinaryExpr::Opcode &Kind); |
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 | , |
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, |
858 | SmallVectorImpl<const MCExpr *> &Values, |
859 | unsigned StringPadLength = 0); |
860 | bool parseScalarInstList( |
861 | unsigned Size, SmallVectorImpl<const MCExpr *> &Values, |
862 | const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); |
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, |
876 | const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); |
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, |
901 | const AsmToken::TokenKind EndToken = AsmToken::EndOfStatement); |
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, |
930 | StringRef Name); |
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, |
1038 | raw_svector_ostream &OS); |
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 | |
1082 | namespace llvm { |
1083 | |
1084 | extern cl::opt<unsigned> AsmMacroMaxNestingDepth; |
1085 | |
1086 | extern MCAsmParserExtension *createCOFFMasmParser(); |
1087 | |
1088 | } // end namespace llvm |
1089 | |
1090 | enum { DEFAULT_ADDRSPACE = 0 }; |
1091 | |
1092 | MasmParser::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. |
1101 | SrcMgr.setDiagHandler(DH: DiagHandler, Ctx: this); |
1102 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer()); |
1103 | EndStatementAtEOFStack.push_back(Val: true); |
1104 | |
1105 | // Initialize the platform / file format parser. |
1106 | switch (Ctx.getObjectFileType()) { |
1107 | case MCContext::IsCOFF: |
1108 | PlatformParser.reset(p: createCOFFMasmParser()); |
1109 | break; |
1110 | default: |
1111 | report_fatal_error(reason: "llvm-ml currently supports only COFF output." ); |
1112 | break; |
1113 | } |
1114 | |
1115 | initializeDirectiveKindMap(); |
1116 | PlatformParser->Initialize(Parser&: *this); |
1117 | initializeCVDefRangeTypeMap(); |
1118 | initializeBuiltinSymbolMap(); |
1119 | |
1120 | NumOfMacroInstantiations = 0; |
1121 | } |
1122 | |
1123 | MasmParser::~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(DH: SavedDiagHandler, Ctx: SavedDiagContext); |
1130 | } |
1131 | |
1132 | void 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(Loc: (*it)->InstantiationLoc, Kind: SourceMgr::DK_Note, |
1139 | Msg: "while in macro instantiation" ); |
1140 | } |
1141 | |
1142 | void MasmParser::Note(SMLoc L, const Twine &Msg, SMRange Range) { |
1143 | printPendingErrors(); |
1144 | printMessage(Loc: L, Kind: SourceMgr::DK_Note, Msg, Range); |
1145 | printMacroInstantiations(); |
1146 | } |
1147 | |
1148 | bool 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(Loc: L, Kind: SourceMgr::DK_Warning, Msg, Range); |
1154 | printMacroInstantiations(); |
1155 | return false; |
1156 | } |
1157 | |
1158 | bool MasmParser::printError(SMLoc L, const Twine &Msg, SMRange Range) { |
1159 | HadError = true; |
1160 | printMessage(Loc: L, Kind: SourceMgr::DK_Error, Msg, Range); |
1161 | printMacroInstantiations(); |
1162 | return true; |
1163 | } |
1164 | |
1165 | bool MasmParser::enterIncludeFile(const std::string &Filename) { |
1166 | std::string IncludedFile; |
1167 | unsigned NewBuf = |
1168 | SrcMgr.AddIncludeFile(Filename, IncludeLoc: Lexer.getLoc(), IncludedFile); |
1169 | if (!NewBuf) |
1170 | return true; |
1171 | |
1172 | CurBuffer = NewBuf; |
1173 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer()); |
1174 | EndStatementAtEOFStack.push_back(Val: true); |
1175 | return false; |
1176 | } |
1177 | |
1178 | void MasmParser::jumpToLoc(SMLoc Loc, unsigned InBuffer, |
1179 | bool EndStatementAtEOF) { |
1180 | CurBuffer = InBuffer ? InBuffer : SrcMgr.FindBufferContainingLoc(Loc); |
1181 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer(), |
1182 | ptr: Loc.getPointer(), EndStatementAtEOF); |
1183 | } |
1184 | |
1185 | bool MasmParser::expandMacros() { |
1186 | const AsmToken &Tok = getTok(); |
1187 | const std::string IDLower = Tok.getIdentifier().lower(); |
1188 | |
1189 | const llvm::MCAsmMacro *M = getContext().lookupMacro(Name: IDLower); |
1190 | if (M && M->IsFunction && peekTok().is(K: 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, NameLoc: MacroLoc)) { |
1196 | Lexer.UnLex(Token: AsmToken(AsmToken::Error, MacroId)); |
1197 | Lexer.Lex(); |
1198 | } |
1199 | return false; |
1200 | } |
1201 | |
1202 | std::optional<std::string> ExpandedValue; |
1203 | auto BuiltinIt = BuiltinSymbolMap.find(Key: IDLower); |
1204 | if (BuiltinIt != BuiltinSymbolMap.end()) { |
1205 | ExpandedValue = |
1206 | evaluateBuiltinTextMacro(Symbol: BuiltinIt->getValue(), StartLoc: Tok.getLoc()); |
1207 | } else { |
1208 | auto VarIt = Variables.find(Key: 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(InputData: *ExpandedValue, BufferName: "<instantiation>" ); |
1218 | |
1219 | // Jump to the macro instantiation and prime the lexer. |
1220 | CurBuffer = |
1221 | SrcMgr.AddNewSourceBuffer(F: std::move(Instantiation), IncludeLoc: Tok.getEndLoc()); |
1222 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer(), ptr: nullptr, |
1223 | /*EndStatementAtEOF=*/false); |
1224 | EndStatementAtEOFStack.push_back(Val: false); |
1225 | Lexer.Lex(); |
1226 | return false; |
1227 | } |
1228 | |
1229 | const AsmToken &MasmParser::Lex(ExpandKind ExpandNextToken) { |
1230 | if (Lexer.getTok().is(K: AsmToken::Error)) |
1231 | Error(L: Lexer.getErrLoc(), Msg: Lexer.getErr()); |
1232 | |
1233 | // if it's a end of statement with a comment in it |
1234 | if (getTok().is(K: 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(T: Twine(getTok().getString())); |
1239 | } |
1240 | |
1241 | const AsmToken *tok = &Lexer.Lex(); |
1242 | bool StartOfStatement = Lexer.isAtStartOfStatement(); |
1243 | |
1244 | while (ExpandNextToken == ExpandMacros && tok->is(K: 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(K: AsmToken::Identifier) && |
1250 | (NextTok.getString().equals_insensitive(RHS: "equ" ) || |
1251 | NextTok.getString().equals_insensitive(RHS: "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(K: AsmToken::Comment)) { |
1263 | if (MAI.preserveAsmComments()) |
1264 | Out.addExplicitComment(T: Twine(tok->getString())); |
1265 | tok = &Lexer.Lex(); |
1266 | } |
1267 | |
1268 | // Recognize and bypass line continuations. |
1269 | while (tok->is(K: AsmToken::BackSlash) && |
1270 | peekTok().is(K: 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(K: 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(i: CurBuffer); |
1280 | if (ParentIncludeLoc != SMLoc()) { |
1281 | EndStatementAtEOFStack.pop_back(); |
1282 | jumpToLoc(Loc: ParentIncludeLoc, InBuffer: 0, EndStatementAtEOF: EndStatementAtEOFStack.back()); |
1283 | return Lex(); |
1284 | } |
1285 | EndStatementAtEOFStack.pop_back(); |
1286 | assert(EndStatementAtEOFStack.empty()); |
1287 | } |
1288 | |
1289 | return *tok; |
1290 | } |
1291 | |
1292 | const AsmToken MasmParser::peekTok(bool ShouldSkipSpace) { |
1293 | AsmToken Tok; |
1294 | |
1295 | MutableArrayRef<AsmToken> Buf(Tok); |
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(i: CurBuffer); |
1302 | if (ParentIncludeLoc != SMLoc()) { |
1303 | EndStatementAtEOFStack.pop_back(); |
1304 | jumpToLoc(Loc: ParentIncludeLoc, InBuffer: 0, EndStatementAtEOF: 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 | |
1315 | bool 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, CompilationDir: getContext().getCompilationDir(), Filename: FirstCppHashFilename, |
1328 | /*Cksum=*/Checksum: std::nullopt, /*Source=*/std::nullopt); |
1329 | const MCDwarfFile &RootFile = |
1330 | getContext().getMCDwarfLineTable(/*CUID=*/0).getRootFile(); |
1331 | getContext().setGenDwarfFileNumber(getStreamer().emitDwarfFileDirective( |
1332 | /*CUID=*/FileNo: 0, Directory: getContext().getCompilationDir(), Filename: RootFile.Name, |
1333 | Checksum: RootFile.Checksum, Source: RootFile.Source)); |
1334 | } |
1335 | return true; |
1336 | } |
1337 | |
1338 | bool MasmParser::Run(bool NoInitialTextSection, bool NoFinalize) { |
1339 | // Create the initial section, if requested. |
1340 | if (!NoInitialTextSection) |
1341 | Out.initSections(NoExecStack: false, STI: 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(Symbol: 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(K: AsmToken::Eof) || |
1370 | SrcMgr.getParentIncludeLoc(i: CurBuffer) != SMLoc()) { |
1371 | // Skip through the EOF at the end of an inclusion. |
1372 | if (Lexer.is(K: AsmToken::Eof)) |
1373 | Lex(); |
1374 | |
1375 | ParseStatementInfo Info(&AsmStrRewrites); |
1376 | bool Parsed = parseStatement(Info, SI: 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(K: 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(Out&: getStreamer()); |
1400 | |
1401 | if (TheCondState.TheCond != StartingCondState.TheCond || |
1402 | TheCondState.Ignore != StartingCondState.Ignore) |
1403 | printError(L: getTok().getLoc(), Msg: "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(L: getTok().getLoc(), Msg: "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().Symbol; |
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 && Sym->isTemporary() && !Sym->isVariable() && |
1429 | !Sym->isDefined()) |
1430 | // FIXME: We would really like to refer back to where the symbol was |
1431 | // first referenced for a source location. We need to add something |
1432 | // to track that. Currently, we just point to the end of the file. |
1433 | printError(L: getTok().getLoc(), Msg: "assembler local symbol '" + |
1434 | Sym->getName() + "' not defined" ); |
1435 | } |
1436 | } |
1437 | |
1438 | // Temporary symbols like the ones for directional jumps don't go in the |
1439 | // symbol table. They also need to be diagnosed in all (final) cases. |
1440 | for (std::tuple<SMLoc, CppHashInfoTy, MCSymbol *> &LocSym : DirLabels) { |
1441 | if (std::get<2>(t&: LocSym)->isUndefined()) { |
1442 | // Reset the state of any "# line file" directives we've seen to the |
1443 | // context as it was at the diagnostic site. |
1444 | CppHashInfo = std::get<1>(t&: LocSym); |
1445 | printError(L: std::get<0>(t&: LocSym), Msg: "directional label undefined" ); |
1446 | } |
1447 | } |
1448 | } |
1449 | |
1450 | // Finalize the output stream if there are no errors and if the client wants |
1451 | // us to. |
1452 | if (!HadError && !NoFinalize) |
1453 | Out.finish(EndLoc: Lexer.getLoc()); |
1454 | |
1455 | return HadError || getContext().hadError(); |
1456 | } |
1457 | |
1458 | bool MasmParser::checkForValidSection() { |
1459 | if (!ParsingMSInlineAsm && !getStreamer().getCurrentSectionOnly()) { |
1460 | Out.initSections(NoExecStack: false, STI: getTargetParser().getSTI()); |
1461 | return Error(L: getTok().getLoc(), |
1462 | Msg: "expected section directive before assembly directive" ); |
1463 | } |
1464 | return false; |
1465 | } |
1466 | |
1467 | /// Throw away the rest of the line for testing purposes. |
1468 | void MasmParser::eatToEndOfStatement() { |
1469 | while (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
1470 | if (Lexer.is(K: AsmToken::Eof)) { |
1471 | SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(i: CurBuffer); |
1472 | if (ParentIncludeLoc == SMLoc()) { |
1473 | break; |
1474 | } |
1475 | |
1476 | EndStatementAtEOFStack.pop_back(); |
1477 | jumpToLoc(Loc: ParentIncludeLoc, InBuffer: 0, EndStatementAtEOF: EndStatementAtEOFStack.back()); |
1478 | } |
1479 | |
1480 | Lexer.Lex(); |
1481 | } |
1482 | |
1483 | // Eat EOL. |
1484 | if (Lexer.is(K: AsmToken::EndOfStatement)) |
1485 | Lexer.Lex(); |
1486 | } |
1487 | |
1488 | SmallVector<StringRef, 1> |
1489 | MasmParser::parseStringRefsTo(AsmToken::TokenKind EndTok) { |
1490 | SmallVector<StringRef, 1> Refs; |
1491 | const char *Start = getTok().getLoc().getPointer(); |
1492 | while (Lexer.isNot(K: EndTok)) { |
1493 | if (Lexer.is(K: AsmToken::Eof)) { |
1494 | SMLoc ParentIncludeLoc = SrcMgr.getParentIncludeLoc(i: CurBuffer); |
1495 | if (ParentIncludeLoc == SMLoc()) { |
1496 | break; |
1497 | } |
1498 | Refs.emplace_back(Args&: Start, Args: getTok().getLoc().getPointer() - Start); |
1499 | |
1500 | EndStatementAtEOFStack.pop_back(); |
1501 | jumpToLoc(Loc: ParentIncludeLoc, InBuffer: 0, EndStatementAtEOF: EndStatementAtEOFStack.back()); |
1502 | Lexer.Lex(); |
1503 | Start = getTok().getLoc().getPointer(); |
1504 | } else { |
1505 | Lexer.Lex(); |
1506 | } |
1507 | } |
1508 | Refs.emplace_back(Args&: Start, Args: getTok().getLoc().getPointer() - Start); |
1509 | return Refs; |
1510 | } |
1511 | |
1512 | std::string MasmParser::parseStringTo(AsmToken::TokenKind EndTok) { |
1513 | SmallVector<StringRef, 1> Refs = parseStringRefsTo(EndTok); |
1514 | std::string Str; |
1515 | for (StringRef S : Refs) { |
1516 | Str.append(str: S.str()); |
1517 | } |
1518 | return Str; |
1519 | } |
1520 | |
1521 | StringRef MasmParser::parseStringToEndOfStatement() { |
1522 | const char *Start = getTok().getLoc().getPointer(); |
1523 | |
1524 | while (Lexer.isNot(K: AsmToken::EndOfStatement) && Lexer.isNot(K: AsmToken::Eof)) |
1525 | Lexer.Lex(); |
1526 | |
1527 | const char *End = getTok().getLoc().getPointer(); |
1528 | return StringRef(Start, End - Start); |
1529 | } |
1530 | |
1531 | /// Parse a paren expression and return it. |
1532 | /// NOTE: This assumes the leading '(' has already been consumed. |
1533 | /// |
1534 | /// parenexpr ::= expr) |
1535 | /// |
1536 | bool MasmParser::parseParenExpr(const MCExpr *&Res, SMLoc &EndLoc) { |
1537 | if (parseExpression(Res)) |
1538 | return true; |
1539 | EndLoc = Lexer.getTok().getEndLoc(); |
1540 | return parseRParen(); |
1541 | } |
1542 | |
1543 | /// Parse a bracket expression and return it. |
1544 | /// NOTE: This assumes the leading '[' has already been consumed. |
1545 | /// |
1546 | /// bracketexpr ::= expr] |
1547 | /// |
1548 | bool MasmParser::parseBracketExpr(const MCExpr *&Res, SMLoc &EndLoc) { |
1549 | if (parseExpression(Res)) |
1550 | return true; |
1551 | EndLoc = getTok().getEndLoc(); |
1552 | if (parseToken(T: AsmToken::RBrac, Msg: "expected ']' in brackets expression" )) |
1553 | return true; |
1554 | return false; |
1555 | } |
1556 | |
1557 | /// Parse a primary expression and return it. |
1558 | /// primaryexpr ::= (parenexpr |
1559 | /// primaryexpr ::= symbol |
1560 | /// primaryexpr ::= number |
1561 | /// primaryexpr ::= '.' |
1562 | /// primaryexpr ::= ~,+,-,'not' primaryexpr |
1563 | /// primaryexpr ::= string |
1564 | /// (a string is interpreted as a 64-bit number in big-endian base-256) |
1565 | bool MasmParser::parsePrimaryExpr(const MCExpr *&Res, SMLoc &EndLoc, |
1566 | AsmTypeInfo *TypeInfo) { |
1567 | SMLoc FirstTokenLoc = getLexer().getLoc(); |
1568 | AsmToken::TokenKind FirstTokenKind = Lexer.getKind(); |
1569 | switch (FirstTokenKind) { |
1570 | default: |
1571 | return TokError(Msg: "unknown token in expression" ); |
1572 | // If we have an error assume that we've already handled it. |
1573 | case AsmToken::Error: |
1574 | return true; |
1575 | case AsmToken::Exclaim: |
1576 | Lex(); // Eat the operator. |
1577 | if (parsePrimaryExpr(Res, EndLoc, TypeInfo: nullptr)) |
1578 | return true; |
1579 | Res = MCUnaryExpr::createLNot(Expr: Res, Ctx&: getContext(), Loc: FirstTokenLoc); |
1580 | return false; |
1581 | case AsmToken::Dollar: |
1582 | case AsmToken::At: |
1583 | case AsmToken::Identifier: { |
1584 | StringRef Identifier; |
1585 | if (parseIdentifier(Res&: Identifier)) { |
1586 | // We may have failed but $ may be a valid token. |
1587 | if (getTok().is(K: AsmToken::Dollar)) { |
1588 | if (Lexer.getMAI().getDollarIsPC()) { |
1589 | Lex(); |
1590 | // This is a '$' reference, which references the current PC. Emit a |
1591 | // temporary label to the streamer and refer to it. |
1592 | MCSymbol *Sym = Ctx.createTempSymbol(); |
1593 | Out.emitLabel(Symbol: Sym); |
1594 | Res = MCSymbolRefExpr::create(Symbol: Sym, Kind: MCSymbolRefExpr::VK_None, |
1595 | Ctx&: getContext()); |
1596 | EndLoc = FirstTokenLoc; |
1597 | return false; |
1598 | } |
1599 | return Error(L: FirstTokenLoc, Msg: "invalid token in expression" ); |
1600 | } |
1601 | } |
1602 | // Parse named bitwise negation. |
1603 | if (Identifier.equals_insensitive(RHS: "not" )) { |
1604 | if (parsePrimaryExpr(Res, EndLoc, TypeInfo: nullptr)) |
1605 | return true; |
1606 | Res = MCUnaryExpr::createNot(Expr: Res, Ctx&: getContext(), Loc: FirstTokenLoc); |
1607 | return false; |
1608 | } |
1609 | // Parse directional local label references. |
1610 | if (Identifier.equals_insensitive(RHS: "@b" ) || |
1611 | Identifier.equals_insensitive(RHS: "@f" )) { |
1612 | bool Before = Identifier.equals_insensitive(RHS: "@b" ); |
1613 | MCSymbol *Sym = getContext().getDirectionalLocalSymbol(LocalLabelVal: 0, Before); |
1614 | if (Before && Sym->isUndefined()) |
1615 | return Error(L: FirstTokenLoc, Msg: "Expected @@ label before @B reference" ); |
1616 | Res = MCSymbolRefExpr::create(Symbol: Sym, Ctx&: getContext()); |
1617 | return false; |
1618 | } |
1619 | // Parse symbol variant. |
1620 | std::pair<StringRef, StringRef> Split; |
1621 | if (!MAI.useParensForSymbolVariant()) { |
1622 | Split = Identifier.split(Separator: '@'); |
1623 | } else if (Lexer.is(K: AsmToken::LParen)) { |
1624 | Lex(); // eat '('. |
1625 | StringRef VName; |
1626 | parseIdentifier(Res&: VName); |
1627 | // eat ')'. |
1628 | if (parseToken(T: AsmToken::RParen, |
1629 | Msg: "unexpected token in variant, expected ')'" )) |
1630 | return true; |
1631 | Split = std::make_pair(x&: Identifier, y&: VName); |
1632 | } |
1633 | |
1634 | EndLoc = SMLoc::getFromPointer(Ptr: Identifier.end()); |
1635 | |
1636 | // This is a symbol reference. |
1637 | StringRef SymbolName = Identifier; |
1638 | if (SymbolName.empty()) |
1639 | return Error(L: getLexer().getLoc(), Msg: "expected a symbol reference" ); |
1640 | |
1641 | MCSymbolRefExpr::VariantKind Variant = MCSymbolRefExpr::VK_None; |
1642 | |
1643 | // Look up the symbol variant if used. |
1644 | if (!Split.second.empty()) { |
1645 | Variant = MCSymbolRefExpr::getVariantKindForName(Name: Split.second); |
1646 | if (Variant != MCSymbolRefExpr::VK_Invalid) { |
1647 | SymbolName = Split.first; |
1648 | } else if (MAI.doesAllowAtInName() && !MAI.useParensForSymbolVariant()) { |
1649 | Variant = MCSymbolRefExpr::VK_None; |
1650 | } else { |
1651 | return Error(L: SMLoc::getFromPointer(Ptr: Split.second.begin()), |
1652 | Msg: "invalid variant '" + Split.second + "'" ); |
1653 | } |
1654 | } |
1655 | |
1656 | // Find the field offset if used. |
1657 | AsmFieldInfo Info; |
1658 | Split = SymbolName.split(Separator: '.'); |
1659 | if (Split.second.empty()) { |
1660 | } else { |
1661 | SymbolName = Split.first; |
1662 | if (lookUpField(Base: SymbolName, Member: Split.second, Info)) { |
1663 | std::pair<StringRef, StringRef> BaseMember = Split.second.split(Separator: '.'); |
1664 | StringRef Base = BaseMember.first, Member = BaseMember.second; |
1665 | lookUpField(Base, Member, Info); |
1666 | } else if (Structs.count(Key: SymbolName.lower())) { |
1667 | // This is actually a reference to a field offset. |
1668 | Res = MCConstantExpr::create(Value: Info.Offset, Ctx&: getContext()); |
1669 | return false; |
1670 | } |
1671 | } |
1672 | |
1673 | MCSymbol *Sym = getContext().getInlineAsmLabel(Name: SymbolName); |
1674 | if (!Sym) { |
1675 | // If this is a built-in numeric value, treat it as a constant. |
1676 | auto BuiltinIt = BuiltinSymbolMap.find(Key: SymbolName.lower()); |
1677 | const BuiltinSymbol Symbol = (BuiltinIt == BuiltinSymbolMap.end()) |
1678 | ? BI_NO_SYMBOL |
1679 | : BuiltinIt->getValue(); |
1680 | if (Symbol != BI_NO_SYMBOL) { |
1681 | const MCExpr *Value = evaluateBuiltinValue(Symbol, StartLoc: FirstTokenLoc); |
1682 | if (Value) { |
1683 | Res = Value; |
1684 | return false; |
1685 | } |
1686 | } |
1687 | |
1688 | // Variables use case-insensitive symbol names; if this is a variable, we |
1689 | // find the symbol using its canonical name. |
1690 | auto VarIt = Variables.find(Key: SymbolName.lower()); |
1691 | if (VarIt != Variables.end()) |
1692 | SymbolName = VarIt->second.Name; |
1693 | Sym = getContext().getOrCreateSymbol(Name: SymbolName); |
1694 | } |
1695 | |
1696 | // If this is an absolute variable reference, substitute it now to preserve |
1697 | // semantics in the face of reassignment. |
1698 | if (Sym->isVariable()) { |
1699 | auto V = Sym->getVariableValue(/*SetUsed=*/false); |
1700 | bool DoInline = isa<MCConstantExpr>(Val: V) && !Variant; |
1701 | if (auto TV = dyn_cast<MCTargetExpr>(Val: V)) |
1702 | DoInline = TV->inlineAssignedExpr(); |
1703 | if (DoInline) { |
1704 | if (Variant) |
1705 | return Error(L: EndLoc, Msg: "unexpected modifier on variable reference" ); |
1706 | Res = Sym->getVariableValue(/*SetUsed=*/false); |
1707 | return false; |
1708 | } |
1709 | } |
1710 | |
1711 | // Otherwise create a symbol ref. |
1712 | const MCExpr *SymRef = |
1713 | MCSymbolRefExpr::create(Symbol: Sym, Kind: Variant, Ctx&: getContext(), Loc: FirstTokenLoc); |
1714 | if (Info.Offset) { |
1715 | Res = MCBinaryExpr::create( |
1716 | Op: MCBinaryExpr::Add, LHS: SymRef, |
1717 | RHS: MCConstantExpr::create(Value: Info.Offset, Ctx&: getContext()), Ctx&: getContext()); |
1718 | } else { |
1719 | Res = SymRef; |
1720 | } |
1721 | if (TypeInfo) { |
1722 | if (Info.Type.Name.empty()) { |
1723 | auto TypeIt = KnownType.find(Key: Identifier.lower()); |
1724 | if (TypeIt != KnownType.end()) { |
1725 | Info.Type = TypeIt->second; |
1726 | } |
1727 | } |
1728 | |
1729 | *TypeInfo = Info.Type; |
1730 | } |
1731 | return false; |
1732 | } |
1733 | case AsmToken::BigNum: |
1734 | return TokError(Msg: "literal value out of range for directive" ); |
1735 | case AsmToken::Integer: { |
1736 | int64_t IntVal = getTok().getIntVal(); |
1737 | Res = MCConstantExpr::create(Value: IntVal, Ctx&: getContext()); |
1738 | EndLoc = Lexer.getTok().getEndLoc(); |
1739 | Lex(); // Eat token. |
1740 | return false; |
1741 | } |
1742 | case AsmToken::String: { |
1743 | // MASM strings (used as constants) are interpreted as big-endian base-256. |
1744 | SMLoc ValueLoc = getTok().getLoc(); |
1745 | std::string Value; |
1746 | if (parseEscapedString(Data&: Value)) |
1747 | return true; |
1748 | if (Value.size() > 8) |
1749 | return Error(L: ValueLoc, Msg: "literal value out of range" ); |
1750 | uint64_t IntValue = 0; |
1751 | for (const unsigned char CharVal : Value) |
1752 | IntValue = (IntValue << 8) | CharVal; |
1753 | Res = MCConstantExpr::create(Value: IntValue, Ctx&: getContext()); |
1754 | return false; |
1755 | } |
1756 | case AsmToken::Real: { |
1757 | APFloat RealVal(APFloat::IEEEdouble(), getTok().getString()); |
1758 | uint64_t IntVal = RealVal.bitcastToAPInt().getZExtValue(); |
1759 | Res = MCConstantExpr::create(Value: IntVal, Ctx&: getContext()); |
1760 | EndLoc = Lexer.getTok().getEndLoc(); |
1761 | Lex(); // Eat token. |
1762 | return false; |
1763 | } |
1764 | case AsmToken::Dot: { |
1765 | // This is a '.' reference, which references the current PC. Emit a |
1766 | // temporary label to the streamer and refer to it. |
1767 | MCSymbol *Sym = Ctx.createTempSymbol(); |
1768 | Out.emitLabel(Symbol: Sym); |
1769 | Res = MCSymbolRefExpr::create(Symbol: Sym, Kind: MCSymbolRefExpr::VK_None, Ctx&: getContext()); |
1770 | EndLoc = Lexer.getTok().getEndLoc(); |
1771 | Lex(); // Eat identifier. |
1772 | return false; |
1773 | } |
1774 | case AsmToken::LParen: |
1775 | Lex(); // Eat the '('. |
1776 | return parseParenExpr(Res, EndLoc); |
1777 | case AsmToken::LBrac: |
1778 | if (!PlatformParser->HasBracketExpressions()) |
1779 | return TokError(Msg: "brackets expression not supported on this target" ); |
1780 | Lex(); // Eat the '['. |
1781 | return parseBracketExpr(Res, EndLoc); |
1782 | case AsmToken::Minus: |
1783 | Lex(); // Eat the operator. |
1784 | if (parsePrimaryExpr(Res, EndLoc, TypeInfo: nullptr)) |
1785 | return true; |
1786 | Res = MCUnaryExpr::createMinus(Expr: Res, Ctx&: getContext(), Loc: FirstTokenLoc); |
1787 | return false; |
1788 | case AsmToken::Plus: |
1789 | Lex(); // Eat the operator. |
1790 | if (parsePrimaryExpr(Res, EndLoc, TypeInfo: nullptr)) |
1791 | return true; |
1792 | Res = MCUnaryExpr::createPlus(Expr: Res, Ctx&: getContext(), Loc: FirstTokenLoc); |
1793 | return false; |
1794 | case AsmToken::Tilde: |
1795 | Lex(); // Eat the operator. |
1796 | if (parsePrimaryExpr(Res, EndLoc, TypeInfo: nullptr)) |
1797 | return true; |
1798 | Res = MCUnaryExpr::createNot(Expr: Res, Ctx&: getContext(), Loc: FirstTokenLoc); |
1799 | return false; |
1800 | // MIPS unary expression operators. The lexer won't generate these tokens if |
1801 | // MCAsmInfo::HasMipsExpressions is false for the target. |
1802 | case AsmToken::PercentCall16: |
1803 | case AsmToken::PercentCall_Hi: |
1804 | case AsmToken::PercentCall_Lo: |
1805 | case AsmToken::PercentDtprel_Hi: |
1806 | case AsmToken::PercentDtprel_Lo: |
1807 | case AsmToken::PercentGot: |
1808 | case AsmToken::PercentGot_Disp: |
1809 | case AsmToken::PercentGot_Hi: |
1810 | case AsmToken::PercentGot_Lo: |
1811 | case AsmToken::PercentGot_Ofst: |
1812 | case AsmToken::PercentGot_Page: |
1813 | case AsmToken::PercentGottprel: |
1814 | case AsmToken::PercentGp_Rel: |
1815 | case AsmToken::PercentHi: |
1816 | case AsmToken::PercentHigher: |
1817 | case AsmToken::PercentHighest: |
1818 | case AsmToken::PercentLo: |
1819 | case AsmToken::PercentNeg: |
1820 | case AsmToken::PercentPcrel_Hi: |
1821 | case AsmToken::PercentPcrel_Lo: |
1822 | case AsmToken::PercentTlsgd: |
1823 | case AsmToken::PercentTlsldm: |
1824 | case AsmToken::PercentTprel_Hi: |
1825 | case AsmToken::PercentTprel_Lo: |
1826 | Lex(); // Eat the operator. |
1827 | if (Lexer.isNot(K: AsmToken::LParen)) |
1828 | return TokError(Msg: "expected '(' after operator" ); |
1829 | Lex(); // Eat the operator. |
1830 | if (parseExpression(Res, EndLoc)) |
1831 | return true; |
1832 | if (parseRParen()) |
1833 | return true; |
1834 | Res = getTargetParser().createTargetUnaryExpr(E: Res, OperatorToken: FirstTokenKind, Ctx); |
1835 | return !Res; |
1836 | } |
1837 | } |
1838 | |
1839 | bool MasmParser::parseExpression(const MCExpr *&Res) { |
1840 | SMLoc EndLoc; |
1841 | return parseExpression(Res, EndLoc); |
1842 | } |
1843 | |
1844 | /// This function checks if the next token is <string> type or arithmetic. |
1845 | /// string that begin with character '<' must end with character '>'. |
1846 | /// otherwise it is arithmetics. |
1847 | /// If the function returns a 'true' value, |
1848 | /// the End argument will be filled with the last location pointed to the '>' |
1849 | /// character. |
1850 | static bool isAngleBracketString(SMLoc &StrLoc, SMLoc &EndLoc) { |
1851 | assert((StrLoc.getPointer() != nullptr) && |
1852 | "Argument to the function cannot be a NULL value" ); |
1853 | const char *CharPtr = StrLoc.getPointer(); |
1854 | while ((*CharPtr != '>') && (*CharPtr != '\n') && (*CharPtr != '\r') && |
1855 | (*CharPtr != '\0')) { |
1856 | if (*CharPtr == '!') |
1857 | CharPtr++; |
1858 | CharPtr++; |
1859 | } |
1860 | if (*CharPtr == '>') { |
1861 | EndLoc = StrLoc.getFromPointer(Ptr: CharPtr + 1); |
1862 | return true; |
1863 | } |
1864 | return false; |
1865 | } |
1866 | |
1867 | /// creating a string without the escape characters '!'. |
1868 | static std::string angleBracketString(StringRef BracketContents) { |
1869 | std::string Res; |
1870 | for (size_t Pos = 0; Pos < BracketContents.size(); Pos++) { |
1871 | if (BracketContents[Pos] == '!') |
1872 | Pos++; |
1873 | Res += BracketContents[Pos]; |
1874 | } |
1875 | return Res; |
1876 | } |
1877 | |
1878 | /// Parse an expression and return it. |
1879 | /// |
1880 | /// expr ::= expr &&,|| expr -> lowest. |
1881 | /// expr ::= expr |,^,&,! expr |
1882 | /// expr ::= expr ==,!=,<>,<,<=,>,>= expr |
1883 | /// expr ::= expr <<,>> expr |
1884 | /// expr ::= expr +,- expr |
1885 | /// expr ::= expr *,/,% expr -> highest. |
1886 | /// expr ::= primaryexpr |
1887 | /// |
1888 | bool MasmParser::parseExpression(const MCExpr *&Res, SMLoc &EndLoc) { |
1889 | // Parse the expression. |
1890 | Res = nullptr; |
1891 | if (getTargetParser().parsePrimaryExpr(Res, EndLoc) || |
1892 | parseBinOpRHS(Precedence: 1, Res, EndLoc)) |
1893 | return true; |
1894 | |
1895 | // Try to constant fold it up front, if possible. Do not exploit |
1896 | // assembler here. |
1897 | int64_t Value; |
1898 | if (Res->evaluateAsAbsolute(Res&: Value)) |
1899 | Res = MCConstantExpr::create(Value, Ctx&: getContext()); |
1900 | |
1901 | return false; |
1902 | } |
1903 | |
1904 | bool MasmParser::parseParenExpression(const MCExpr *&Res, SMLoc &EndLoc) { |
1905 | Res = nullptr; |
1906 | return parseParenExpr(Res, EndLoc) || parseBinOpRHS(Precedence: 1, Res, EndLoc); |
1907 | } |
1908 | |
1909 | bool MasmParser::parseParenExprOfDepth(unsigned ParenDepth, const MCExpr *&Res, |
1910 | SMLoc &EndLoc) { |
1911 | if (parseParenExpr(Res, EndLoc)) |
1912 | return true; |
1913 | |
1914 | for (; ParenDepth > 0; --ParenDepth) { |
1915 | if (parseBinOpRHS(Precedence: 1, Res, EndLoc)) |
1916 | return true; |
1917 | |
1918 | // We don't Lex() the last RParen. |
1919 | // This is the same behavior as parseParenExpression(). |
1920 | if (ParenDepth - 1 > 0) { |
1921 | EndLoc = getTok().getEndLoc(); |
1922 | if (parseRParen()) |
1923 | return true; |
1924 | } |
1925 | } |
1926 | return false; |
1927 | } |
1928 | |
1929 | bool MasmParser::parseAbsoluteExpression(int64_t &Res) { |
1930 | const MCExpr *Expr; |
1931 | |
1932 | SMLoc StartLoc = Lexer.getLoc(); |
1933 | if (parseExpression(Res&: Expr)) |
1934 | return true; |
1935 | |
1936 | if (!Expr->evaluateAsAbsolute(Res, Asm: getStreamer().getAssemblerPtr())) |
1937 | return Error(L: StartLoc, Msg: "expected absolute expression" ); |
1938 | |
1939 | return false; |
1940 | } |
1941 | |
1942 | static unsigned getGNUBinOpPrecedence(AsmToken::TokenKind K, |
1943 | MCBinaryExpr::Opcode &Kind, |
1944 | bool ShouldUseLogicalShr, |
1945 | bool EndExpressionAtGreater) { |
1946 | switch (K) { |
1947 | default: |
1948 | return 0; // not a binop. |
1949 | |
1950 | // Lowest Precedence: &&, || |
1951 | case AsmToken::AmpAmp: |
1952 | Kind = MCBinaryExpr::LAnd; |
1953 | return 2; |
1954 | case AsmToken::PipePipe: |
1955 | Kind = MCBinaryExpr::LOr; |
1956 | return 1; |
1957 | |
1958 | // Low Precedence: ==, !=, <>, <, <=, >, >= |
1959 | case AsmToken::EqualEqual: |
1960 | Kind = MCBinaryExpr::EQ; |
1961 | return 3; |
1962 | case AsmToken::ExclaimEqual: |
1963 | case AsmToken::LessGreater: |
1964 | Kind = MCBinaryExpr::NE; |
1965 | return 3; |
1966 | case AsmToken::Less: |
1967 | Kind = MCBinaryExpr::LT; |
1968 | return 3; |
1969 | case AsmToken::LessEqual: |
1970 | Kind = MCBinaryExpr::LTE; |
1971 | return 3; |
1972 | case AsmToken::Greater: |
1973 | if (EndExpressionAtGreater) |
1974 | return 0; |
1975 | Kind = MCBinaryExpr::GT; |
1976 | return 3; |
1977 | case AsmToken::GreaterEqual: |
1978 | Kind = MCBinaryExpr::GTE; |
1979 | return 3; |
1980 | |
1981 | // Low Intermediate Precedence: +, - |
1982 | case AsmToken::Plus: |
1983 | Kind = MCBinaryExpr::Add; |
1984 | return 4; |
1985 | case AsmToken::Minus: |
1986 | Kind = MCBinaryExpr::Sub; |
1987 | return 4; |
1988 | |
1989 | // High Intermediate Precedence: |, &, ^ |
1990 | case AsmToken::Pipe: |
1991 | Kind = MCBinaryExpr::Or; |
1992 | return 5; |
1993 | case AsmToken::Caret: |
1994 | Kind = MCBinaryExpr::Xor; |
1995 | return 5; |
1996 | case AsmToken::Amp: |
1997 | Kind = MCBinaryExpr::And; |
1998 | return 5; |
1999 | |
2000 | // Highest Precedence: *, /, %, <<, >> |
2001 | case AsmToken::Star: |
2002 | Kind = MCBinaryExpr::Mul; |
2003 | return 6; |
2004 | case AsmToken::Slash: |
2005 | Kind = MCBinaryExpr::Div; |
2006 | return 6; |
2007 | case AsmToken::Percent: |
2008 | Kind = MCBinaryExpr::Mod; |
2009 | return 6; |
2010 | case AsmToken::LessLess: |
2011 | Kind = MCBinaryExpr::Shl; |
2012 | return 6; |
2013 | case AsmToken::GreaterGreater: |
2014 | if (EndExpressionAtGreater) |
2015 | return 0; |
2016 | Kind = ShouldUseLogicalShr ? MCBinaryExpr::LShr : MCBinaryExpr::AShr; |
2017 | return 6; |
2018 | } |
2019 | } |
2020 | |
2021 | unsigned MasmParser::getBinOpPrecedence(AsmToken::TokenKind K, |
2022 | MCBinaryExpr::Opcode &Kind) { |
2023 | bool ShouldUseLogicalShr = MAI.shouldUseLogicalShr(); |
2024 | return getGNUBinOpPrecedence(K, Kind, ShouldUseLogicalShr, |
2025 | EndExpressionAtGreater: AngleBracketDepth > 0); |
2026 | } |
2027 | |
2028 | /// Parse all binary operators with precedence >= 'Precedence'. |
2029 | /// Res contains the LHS of the expression on input. |
2030 | bool MasmParser::parseBinOpRHS(unsigned Precedence, const MCExpr *&Res, |
2031 | SMLoc &EndLoc) { |
2032 | SMLoc StartLoc = Lexer.getLoc(); |
2033 | while (true) { |
2034 | AsmToken::TokenKind TokKind = Lexer.getKind(); |
2035 | if (Lexer.getKind() == AsmToken::Identifier) { |
2036 | TokKind = StringSwitch<AsmToken::TokenKind>(Lexer.getTok().getString()) |
2037 | .CaseLower(S: "and" , Value: AsmToken::Amp) |
2038 | .CaseLower(S: "not" , Value: AsmToken::Exclaim) |
2039 | .CaseLower(S: "or" , Value: AsmToken::Pipe) |
2040 | .CaseLower(S: "xor" , Value: AsmToken::Caret) |
2041 | .CaseLower(S: "shl" , Value: AsmToken::LessLess) |
2042 | .CaseLower(S: "shr" , Value: AsmToken::GreaterGreater) |
2043 | .CaseLower(S: "eq" , Value: AsmToken::EqualEqual) |
2044 | .CaseLower(S: "ne" , Value: AsmToken::ExclaimEqual) |
2045 | .CaseLower(S: "lt" , Value: AsmToken::Less) |
2046 | .CaseLower(S: "le" , Value: AsmToken::LessEqual) |
2047 | .CaseLower(S: "gt" , Value: AsmToken::Greater) |
2048 | .CaseLower(S: "ge" , Value: AsmToken::GreaterEqual) |
2049 | .Default(Value: TokKind); |
2050 | } |
2051 | MCBinaryExpr::Opcode Kind = MCBinaryExpr::Add; |
2052 | unsigned TokPrec = getBinOpPrecedence(K: TokKind, Kind); |
2053 | |
2054 | // If the next token is lower precedence than we are allowed to eat, return |
2055 | // successfully with what we ate already. |
2056 | if (TokPrec < Precedence) |
2057 | return false; |
2058 | |
2059 | Lex(); |
2060 | |
2061 | // Eat the next primary expression. |
2062 | const MCExpr *RHS; |
2063 | if (getTargetParser().parsePrimaryExpr(Res&: RHS, EndLoc)) |
2064 | return true; |
2065 | |
2066 | // If BinOp binds less tightly with RHS than the operator after RHS, let |
2067 | // the pending operator take RHS as its LHS. |
2068 | MCBinaryExpr::Opcode Dummy; |
2069 | unsigned NextTokPrec = getBinOpPrecedence(K: Lexer.getKind(), Kind&: Dummy); |
2070 | if (TokPrec < NextTokPrec && parseBinOpRHS(Precedence: TokPrec + 1, Res&: RHS, EndLoc)) |
2071 | return true; |
2072 | |
2073 | // Merge LHS and RHS according to operator. |
2074 | Res = MCBinaryExpr::create(Op: Kind, LHS: Res, RHS, Ctx&: getContext(), Loc: StartLoc); |
2075 | } |
2076 | } |
2077 | |
2078 | /// ParseStatement: |
2079 | /// ::= % statement |
2080 | /// ::= EndOfStatement |
2081 | /// ::= Label* Directive ...Operands... EndOfStatement |
2082 | /// ::= Label* Identifier OperandList* EndOfStatement |
2083 | bool MasmParser::parseStatement(ParseStatementInfo &Info, |
2084 | MCAsmParserSemaCallback *SI) { |
2085 | assert(!hasPendingError() && "parseStatement started with pending error" ); |
2086 | // Eat initial spaces and comments. |
2087 | while (Lexer.is(K: AsmToken::Space)) |
2088 | Lex(); |
2089 | if (Lexer.is(K: AsmToken::EndOfStatement)) { |
2090 | // If this is a line comment we can drop it safely. |
2091 | if (getTok().getString().empty() || getTok().getString().front() == '\r' || |
2092 | getTok().getString().front() == '\n') |
2093 | Out.addBlankLine(); |
2094 | Lex(); |
2095 | return false; |
2096 | } |
2097 | |
2098 | // If preceded by an expansion operator, first expand all text macros and |
2099 | // macro functions. |
2100 | if (getTok().is(K: AsmToken::Percent)) { |
2101 | SMLoc ExpansionLoc = getTok().getLoc(); |
2102 | if (parseToken(T: AsmToken::Percent) || expandStatement(Loc: ExpansionLoc)) |
2103 | return true; |
2104 | } |
2105 | |
2106 | // Statements always start with an identifier, unless we're dealing with a |
2107 | // processor directive (.386, .686, etc.) that lexes as a real. |
2108 | AsmToken ID = getTok(); |
2109 | SMLoc IDLoc = ID.getLoc(); |
2110 | StringRef IDVal; |
2111 | if (Lexer.is(K: AsmToken::HashDirective)) |
2112 | return parseCppHashLineFilenameComment(L: IDLoc); |
2113 | if (Lexer.is(K: AsmToken::Dot)) { |
2114 | // Treat '.' as a valid identifier in this context. |
2115 | Lex(); |
2116 | IDVal = "." ; |
2117 | } else if (Lexer.is(K: AsmToken::Real)) { |
2118 | // Treat ".<number>" as a valid identifier in this context. |
2119 | IDVal = getTok().getString(); |
2120 | Lex(); // always eat a token |
2121 | if (!IDVal.starts_with(Prefix: "." )) |
2122 | return Error(L: IDLoc, Msg: "unexpected token at start of statement" ); |
2123 | } else if (parseIdentifier(Res&: IDVal, Position: StartOfStatement)) { |
2124 | if (!TheCondState.Ignore) { |
2125 | Lex(); // always eat a token |
2126 | return Error(L: IDLoc, Msg: "unexpected token at start of statement" ); |
2127 | } |
2128 | IDVal = "" ; |
2129 | } |
2130 | |
2131 | // Handle conditional assembly here before checking for skipping. We |
2132 | // have to do this so that .endif isn't skipped in a ".if 0" block for |
2133 | // example. |
2134 | StringMap<DirectiveKind>::const_iterator DirKindIt = |
2135 | DirectiveKindMap.find(Key: IDVal.lower()); |
2136 | DirectiveKind DirKind = (DirKindIt == DirectiveKindMap.end()) |
2137 | ? DK_NO_DIRECTIVE |
2138 | : DirKindIt->getValue(); |
2139 | switch (DirKind) { |
2140 | default: |
2141 | break; |
2142 | case DK_IF: |
2143 | case DK_IFE: |
2144 | return parseDirectiveIf(DirectiveLoc: IDLoc, DirKind); |
2145 | case DK_IFB: |
2146 | return parseDirectiveIfb(DirectiveLoc: IDLoc, ExpectBlank: true); |
2147 | case DK_IFNB: |
2148 | return parseDirectiveIfb(DirectiveLoc: IDLoc, ExpectBlank: false); |
2149 | case DK_IFDEF: |
2150 | return parseDirectiveIfdef(DirectiveLoc: IDLoc, expect_defined: true); |
2151 | case DK_IFNDEF: |
2152 | return parseDirectiveIfdef(DirectiveLoc: IDLoc, expect_defined: false); |
2153 | case DK_IFDIF: |
2154 | return parseDirectiveIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2155 | /*CaseInsensitive=*/false); |
2156 | case DK_IFDIFI: |
2157 | return parseDirectiveIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2158 | /*CaseInsensitive=*/true); |
2159 | case DK_IFIDN: |
2160 | return parseDirectiveIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2161 | /*CaseInsensitive=*/false); |
2162 | case DK_IFIDNI: |
2163 | return parseDirectiveIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2164 | /*CaseInsensitive=*/true); |
2165 | case DK_ELSEIF: |
2166 | case DK_ELSEIFE: |
2167 | return parseDirectiveElseIf(DirectiveLoc: IDLoc, DirKind); |
2168 | case DK_ELSEIFB: |
2169 | return parseDirectiveElseIfb(DirectiveLoc: IDLoc, ExpectBlank: true); |
2170 | case DK_ELSEIFNB: |
2171 | return parseDirectiveElseIfb(DirectiveLoc: IDLoc, ExpectBlank: false); |
2172 | case DK_ELSEIFDEF: |
2173 | return parseDirectiveElseIfdef(DirectiveLoc: IDLoc, expect_defined: true); |
2174 | case DK_ELSEIFNDEF: |
2175 | return parseDirectiveElseIfdef(DirectiveLoc: IDLoc, expect_defined: false); |
2176 | case DK_ELSEIFDIF: |
2177 | return parseDirectiveElseIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2178 | /*CaseInsensitive=*/false); |
2179 | case DK_ELSEIFDIFI: |
2180 | return parseDirectiveElseIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2181 | /*CaseInsensitive=*/true); |
2182 | case DK_ELSEIFIDN: |
2183 | return parseDirectiveElseIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2184 | /*CaseInsensitive=*/false); |
2185 | case DK_ELSEIFIDNI: |
2186 | return parseDirectiveElseIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2187 | /*CaseInsensitive=*/true); |
2188 | case DK_ELSE: |
2189 | return parseDirectiveElse(DirectiveLoc: IDLoc); |
2190 | case DK_ENDIF: |
2191 | return parseDirectiveEndIf(DirectiveLoc: IDLoc); |
2192 | } |
2193 | |
2194 | // Ignore the statement if in the middle of inactive conditional |
2195 | // (e.g. ".if 0"). |
2196 | if (TheCondState.Ignore) { |
2197 | eatToEndOfStatement(); |
2198 | return false; |
2199 | } |
2200 | |
2201 | // FIXME: Recurse on local labels? |
2202 | |
2203 | // Check for a label. |
2204 | // ::= identifier ':' |
2205 | // ::= number ':' |
2206 | if (Lexer.is(K: AsmToken::Colon) && getTargetParser().isLabel(Token&: ID)) { |
2207 | if (checkForValidSection()) |
2208 | return true; |
2209 | |
2210 | // identifier ':' -> Label. |
2211 | Lex(); |
2212 | |
2213 | // Diagnose attempt to use '.' as a label. |
2214 | if (IDVal == "." ) |
2215 | return Error(L: IDLoc, Msg: "invalid use of pseudo-symbol '.' as a label" ); |
2216 | |
2217 | // Diagnose attempt to use a variable as a label. |
2218 | // |
2219 | // FIXME: Diagnostics. Note the location of the definition as a label. |
2220 | // FIXME: This doesn't diagnose assignment to a symbol which has been |
2221 | // implicitly marked as external. |
2222 | MCSymbol *Sym; |
2223 | if (ParsingMSInlineAsm && SI) { |
2224 | StringRef RewrittenLabel = |
2225 | SI->LookupInlineAsmLabel(Identifier: IDVal, SM&: getSourceManager(), Location: IDLoc, Create: true); |
2226 | assert(!RewrittenLabel.empty() && |
2227 | "We should have an internal name here." ); |
2228 | Info.AsmRewrites->emplace_back(Args: AOK_Label, Args&: IDLoc, Args: IDVal.size(), |
2229 | Args&: RewrittenLabel); |
2230 | IDVal = RewrittenLabel; |
2231 | } |
2232 | // Handle directional local labels |
2233 | if (IDVal == "@@" ) { |
2234 | Sym = Ctx.createDirectionalLocalSymbol(LocalLabelVal: 0); |
2235 | } else { |
2236 | Sym = getContext().getOrCreateSymbol(Name: IDVal); |
2237 | } |
2238 | |
2239 | // End of Labels should be treated as end of line for lexing |
2240 | // purposes but that information is not available to the Lexer who |
2241 | // does not understand Labels. This may cause us to see a Hash |
2242 | // here instead of a preprocessor line comment. |
2243 | if (getTok().is(K: AsmToken::Hash)) { |
2244 | std::string = parseStringTo(EndTok: AsmToken::EndOfStatement); |
2245 | Lexer.Lex(); |
2246 | Lexer.UnLex(Token: AsmToken(AsmToken::EndOfStatement, CommentStr)); |
2247 | } |
2248 | |
2249 | // Consume any end of statement token, if present, to avoid spurious |
2250 | // addBlankLine calls(). |
2251 | if (getTok().is(K: AsmToken::EndOfStatement)) { |
2252 | Lex(); |
2253 | } |
2254 | |
2255 | getTargetParser().doBeforeLabelEmit(Symbol: Sym, IDLoc); |
2256 | |
2257 | // Emit the label. |
2258 | if (!getTargetParser().isParsingMSInlineAsm()) |
2259 | Out.emitLabel(Symbol: Sym, Loc: IDLoc); |
2260 | |
2261 | // If we are generating dwarf for assembly source files then gather the |
2262 | // info to make a dwarf label entry for this label if needed. |
2263 | if (enabledGenDwarfForAssembly()) |
2264 | MCGenDwarfLabelEntry::Make(Symbol: Sym, MCOS: &getStreamer(), SrcMgr&: getSourceManager(), |
2265 | Loc&: IDLoc); |
2266 | |
2267 | getTargetParser().onLabelParsed(Symbol: Sym); |
2268 | |
2269 | return false; |
2270 | } |
2271 | |
2272 | // If macros are enabled, check to see if this is a macro instantiation. |
2273 | if (const MCAsmMacro *M = getContext().lookupMacro(Name: IDVal.lower())) { |
2274 | return handleMacroEntry(M, NameLoc: IDLoc); |
2275 | } |
2276 | |
2277 | // Otherwise, we have a normal instruction or directive. |
2278 | |
2279 | if (DirKind != DK_NO_DIRECTIVE) { |
2280 | // There are several entities interested in parsing directives: |
2281 | // |
2282 | // 1. Asm parser extensions. For example, platform-specific parsers |
2283 | // (like the ELF parser) register themselves as extensions. |
2284 | // 2. The target-specific assembly parser. Some directives are target |
2285 | // specific or may potentially behave differently on certain targets. |
2286 | // 3. The generic directive parser implemented by this class. These are |
2287 | // all the directives that behave in a target and platform independent |
2288 | // manner, or at least have a default behavior that's shared between |
2289 | // all targets and platforms. |
2290 | |
2291 | getTargetParser().flushPendingInstructions(Out&: getStreamer()); |
2292 | |
2293 | // Special-case handling of structure-end directives at higher priority, |
2294 | // since ENDS is overloaded as a segment-end directive. |
2295 | if (IDVal.equals_insensitive(RHS: "ends" ) && StructInProgress.size() > 1 && |
2296 | getTok().is(K: AsmToken::EndOfStatement)) { |
2297 | return parseDirectiveNestedEnds(); |
2298 | } |
2299 | |
2300 | // First, check the extension directive map to see if any extension has |
2301 | // registered itself to parse this directive. |
2302 | std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = |
2303 | ExtensionDirectiveMap.lookup(Key: IDVal.lower()); |
2304 | if (Handler.first) |
2305 | return (*Handler.second)(Handler.first, IDVal, IDLoc); |
2306 | |
2307 | // Next, let the target-specific assembly parser try. |
2308 | if (ID.isNot(K: AsmToken::Identifier)) |
2309 | return false; |
2310 | |
2311 | ParseStatus TPDirectiveReturn = getTargetParser().parseDirective(DirectiveID: ID); |
2312 | assert(TPDirectiveReturn.isFailure() == hasPendingError() && |
2313 | "Should only return Failure iff there was an error" ); |
2314 | if (TPDirectiveReturn.isFailure()) |
2315 | return true; |
2316 | if (TPDirectiveReturn.isSuccess()) |
2317 | return false; |
2318 | |
2319 | // Finally, if no one else is interested in this directive, it must be |
2320 | // generic and familiar to this class. |
2321 | switch (DirKind) { |
2322 | default: |
2323 | break; |
2324 | case DK_ASCII: |
2325 | return parseDirectiveAscii(IDVal, ZeroTerminated: false); |
2326 | case DK_ASCIZ: |
2327 | case DK_STRING: |
2328 | return parseDirectiveAscii(IDVal, ZeroTerminated: true); |
2329 | case DK_BYTE: |
2330 | case DK_SBYTE: |
2331 | case DK_DB: |
2332 | return parseDirectiveValue(IDVal, Size: 1); |
2333 | case DK_WORD: |
2334 | case DK_SWORD: |
2335 | case DK_DW: |
2336 | return parseDirectiveValue(IDVal, Size: 2); |
2337 | case DK_DWORD: |
2338 | case DK_SDWORD: |
2339 | case DK_DD: |
2340 | return parseDirectiveValue(IDVal, Size: 4); |
2341 | case DK_FWORD: |
2342 | case DK_DF: |
2343 | return parseDirectiveValue(IDVal, Size: 6); |
2344 | case DK_QWORD: |
2345 | case DK_SQWORD: |
2346 | case DK_DQ: |
2347 | return parseDirectiveValue(IDVal, Size: 8); |
2348 | case DK_REAL4: |
2349 | return parseDirectiveRealValue(IDVal, Semantics: APFloat::IEEEsingle(), Size: 4); |
2350 | case DK_REAL8: |
2351 | return parseDirectiveRealValue(IDVal, Semantics: APFloat::IEEEdouble(), Size: 8); |
2352 | case DK_REAL10: |
2353 | return parseDirectiveRealValue(IDVal, Semantics: APFloat::x87DoubleExtended(), Size: 10); |
2354 | case DK_STRUCT: |
2355 | case DK_UNION: |
2356 | return parseDirectiveNestedStruct(Directive: IDVal, DirKind); |
2357 | case DK_ENDS: |
2358 | return parseDirectiveNestedEnds(); |
2359 | case DK_ALIGN: |
2360 | return parseDirectiveAlign(); |
2361 | case DK_EVEN: |
2362 | return parseDirectiveEven(); |
2363 | case DK_ORG: |
2364 | return parseDirectiveOrg(); |
2365 | case DK_EXTERN: |
2366 | return parseDirectiveExtern(); |
2367 | case DK_PUBLIC: |
2368 | return parseDirectiveSymbolAttribute(Attr: MCSA_Global); |
2369 | case DK_COMM: |
2370 | return parseDirectiveComm(/*IsLocal=*/false); |
2371 | case DK_COMMENT: |
2372 | return parseDirectiveComment(DirectiveLoc: IDLoc); |
2373 | case DK_INCLUDE: |
2374 | return parseDirectiveInclude(); |
2375 | case DK_REPEAT: |
2376 | return parseDirectiveRepeat(DirectiveLoc: IDLoc, Directive: IDVal); |
2377 | case DK_WHILE: |
2378 | return parseDirectiveWhile(DirectiveLoc: IDLoc); |
2379 | case DK_FOR: |
2380 | return parseDirectiveFor(DirectiveLoc: IDLoc, Directive: IDVal); |
2381 | case DK_FORC: |
2382 | return parseDirectiveForc(DirectiveLoc: IDLoc, Directive: IDVal); |
2383 | case DK_FILE: |
2384 | return parseDirectiveFile(DirectiveLoc: IDLoc); |
2385 | case DK_LINE: |
2386 | return parseDirectiveLine(); |
2387 | case DK_LOC: |
2388 | return parseDirectiveLoc(); |
2389 | case DK_STABS: |
2390 | return parseDirectiveStabs(); |
2391 | case DK_CV_FILE: |
2392 | return parseDirectiveCVFile(); |
2393 | case DK_CV_FUNC_ID: |
2394 | return parseDirectiveCVFuncId(); |
2395 | case DK_CV_INLINE_SITE_ID: |
2396 | return parseDirectiveCVInlineSiteId(); |
2397 | case DK_CV_LOC: |
2398 | return parseDirectiveCVLoc(); |
2399 | case DK_CV_LINETABLE: |
2400 | return parseDirectiveCVLinetable(); |
2401 | case DK_CV_INLINE_LINETABLE: |
2402 | return parseDirectiveCVInlineLinetable(); |
2403 | case DK_CV_DEF_RANGE: |
2404 | return parseDirectiveCVDefRange(); |
2405 | case DK_CV_STRING: |
2406 | return parseDirectiveCVString(); |
2407 | case DK_CV_STRINGTABLE: |
2408 | return parseDirectiveCVStringTable(); |
2409 | case DK_CV_FILECHECKSUMS: |
2410 | return parseDirectiveCVFileChecksums(); |
2411 | case DK_CV_FILECHECKSUM_OFFSET: |
2412 | return parseDirectiveCVFileChecksumOffset(); |
2413 | case DK_CV_FPO_DATA: |
2414 | return parseDirectiveCVFPOData(); |
2415 | case DK_CFI_SECTIONS: |
2416 | return parseDirectiveCFISections(); |
2417 | case DK_CFI_STARTPROC: |
2418 | return parseDirectiveCFIStartProc(); |
2419 | case DK_CFI_ENDPROC: |
2420 | return parseDirectiveCFIEndProc(); |
2421 | case DK_CFI_DEF_CFA: |
2422 | return parseDirectiveCFIDefCfa(DirectiveLoc: IDLoc); |
2423 | case DK_CFI_DEF_CFA_OFFSET: |
2424 | return parseDirectiveCFIDefCfaOffset(DirectiveLoc: IDLoc); |
2425 | case DK_CFI_ADJUST_CFA_OFFSET: |
2426 | return parseDirectiveCFIAdjustCfaOffset(DirectiveLoc: IDLoc); |
2427 | case DK_CFI_DEF_CFA_REGISTER: |
2428 | return parseDirectiveCFIDefCfaRegister(DirectiveLoc: IDLoc); |
2429 | case DK_CFI_OFFSET: |
2430 | return parseDirectiveCFIOffset(DirectiveLoc: IDLoc); |
2431 | case DK_CFI_REL_OFFSET: |
2432 | return parseDirectiveCFIRelOffset(DirectiveLoc: IDLoc); |
2433 | case DK_CFI_PERSONALITY: |
2434 | return parseDirectiveCFIPersonalityOrLsda(IsPersonality: true); |
2435 | case DK_CFI_LSDA: |
2436 | return parseDirectiveCFIPersonalityOrLsda(IsPersonality: false); |
2437 | case DK_CFI_REMEMBER_STATE: |
2438 | return parseDirectiveCFIRememberState(DirectiveLoc: IDLoc); |
2439 | case DK_CFI_RESTORE_STATE: |
2440 | return parseDirectiveCFIRestoreState(DirectiveLoc: IDLoc); |
2441 | case DK_CFI_SAME_VALUE: |
2442 | return parseDirectiveCFISameValue(DirectiveLoc: IDLoc); |
2443 | case DK_CFI_RESTORE: |
2444 | return parseDirectiveCFIRestore(DirectiveLoc: IDLoc); |
2445 | case DK_CFI_ESCAPE: |
2446 | return parseDirectiveCFIEscape(DirectiveLoc: IDLoc); |
2447 | case DK_CFI_RETURN_COLUMN: |
2448 | return parseDirectiveCFIReturnColumn(DirectiveLoc: IDLoc); |
2449 | case DK_CFI_SIGNAL_FRAME: |
2450 | return parseDirectiveCFISignalFrame(); |
2451 | case DK_CFI_UNDEFINED: |
2452 | return parseDirectiveCFIUndefined(DirectiveLoc: IDLoc); |
2453 | case DK_CFI_REGISTER: |
2454 | return parseDirectiveCFIRegister(DirectiveLoc: IDLoc); |
2455 | case DK_CFI_WINDOW_SAVE: |
2456 | return parseDirectiveCFIWindowSave(DirectiveLoc: IDLoc); |
2457 | case DK_EXITM: |
2458 | Info.ExitValue = "" ; |
2459 | return parseDirectiveExitMacro(DirectiveLoc: IDLoc, Directive: IDVal, Value&: *Info.ExitValue); |
2460 | case DK_ENDM: |
2461 | Info.ExitValue = "" ; |
2462 | return parseDirectiveEndMacro(Directive: IDVal); |
2463 | case DK_PURGE: |
2464 | return parseDirectivePurgeMacro(DirectiveLoc: IDLoc); |
2465 | case DK_END: |
2466 | return parseDirectiveEnd(DirectiveLoc: IDLoc); |
2467 | case DK_ERR: |
2468 | return parseDirectiveError(DirectiveLoc: IDLoc); |
2469 | case DK_ERRB: |
2470 | return parseDirectiveErrorIfb(DirectiveLoc: IDLoc, ExpectBlank: true); |
2471 | case DK_ERRNB: |
2472 | return parseDirectiveErrorIfb(DirectiveLoc: IDLoc, ExpectBlank: false); |
2473 | case DK_ERRDEF: |
2474 | return parseDirectiveErrorIfdef(DirectiveLoc: IDLoc, ExpectDefined: true); |
2475 | case DK_ERRNDEF: |
2476 | return parseDirectiveErrorIfdef(DirectiveLoc: IDLoc, ExpectDefined: false); |
2477 | case DK_ERRDIF: |
2478 | return parseDirectiveErrorIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2479 | /*CaseInsensitive=*/false); |
2480 | case DK_ERRDIFI: |
2481 | return parseDirectiveErrorIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/false, |
2482 | /*CaseInsensitive=*/true); |
2483 | case DK_ERRIDN: |
2484 | return parseDirectiveErrorIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2485 | /*CaseInsensitive=*/false); |
2486 | case DK_ERRIDNI: |
2487 | return parseDirectiveErrorIfidn(DirectiveLoc: IDLoc, /*ExpectEqual=*/true, |
2488 | /*CaseInsensitive=*/true); |
2489 | case DK_ERRE: |
2490 | return parseDirectiveErrorIfe(DirectiveLoc: IDLoc, ExpectZero: true); |
2491 | case DK_ERRNZ: |
2492 | return parseDirectiveErrorIfe(DirectiveLoc: IDLoc, ExpectZero: false); |
2493 | case DK_RADIX: |
2494 | return parseDirectiveRadix(DirectiveLoc: IDLoc); |
2495 | case DK_ECHO: |
2496 | return parseDirectiveEcho(DirectiveLoc: IDLoc); |
2497 | } |
2498 | |
2499 | return Error(L: IDLoc, Msg: "unknown directive" ); |
2500 | } |
2501 | |
2502 | // We also check if this is allocating memory with user-defined type. |
2503 | auto IDIt = Structs.find(Key: IDVal.lower()); |
2504 | if (IDIt != Structs.end()) |
2505 | return parseDirectiveStructValue(/*Structure=*/IDIt->getValue(), Directive: IDVal, |
2506 | DirLoc: IDLoc); |
2507 | |
2508 | // Non-conditional Microsoft directives sometimes follow their first argument. |
2509 | const AsmToken nextTok = getTok(); |
2510 | const StringRef nextVal = nextTok.getString(); |
2511 | const SMLoc nextLoc = nextTok.getLoc(); |
2512 | |
2513 | const AsmToken afterNextTok = peekTok(); |
2514 | |
2515 | // There are several entities interested in parsing infix directives: |
2516 | // |
2517 | // 1. Asm parser extensions. For example, platform-specific parsers |
2518 | // (like the ELF parser) register themselves as extensions. |
2519 | // 2. The generic directive parser implemented by this class. These are |
2520 | // all the directives that behave in a target and platform independent |
2521 | // manner, or at least have a default behavior that's shared between |
2522 | // all targets and platforms. |
2523 | |
2524 | getTargetParser().flushPendingInstructions(Out&: getStreamer()); |
2525 | |
2526 | // Special-case handling of structure-end directives at higher priority, since |
2527 | // ENDS is overloaded as a segment-end directive. |
2528 | if (nextVal.equals_insensitive(RHS: "ends" ) && StructInProgress.size() == 1) { |
2529 | Lex(); |
2530 | return parseDirectiveEnds(Name: IDVal, NameLoc: IDLoc); |
2531 | } |
2532 | |
2533 | // First, check the extension directive map to see if any extension has |
2534 | // registered itself to parse this directive. |
2535 | std::pair<MCAsmParserExtension *, DirectiveHandler> Handler = |
2536 | ExtensionDirectiveMap.lookup(Key: nextVal.lower()); |
2537 | if (Handler.first) { |
2538 | Lex(); |
2539 | Lexer.UnLex(Token: ID); |
2540 | return (*Handler.second)(Handler.first, nextVal, nextLoc); |
2541 | } |
2542 | |
2543 | // If no one else is interested in this directive, it must be |
2544 | // generic and familiar to this class. |
2545 | DirKindIt = DirectiveKindMap.find(Key: nextVal.lower()); |
2546 | DirKind = (DirKindIt == DirectiveKindMap.end()) |
2547 | ? DK_NO_DIRECTIVE |
2548 | : DirKindIt->getValue(); |
2549 | switch (DirKind) { |
2550 | default: |
2551 | break; |
2552 | case DK_ASSIGN: |
2553 | case DK_EQU: |
2554 | case DK_TEXTEQU: |
2555 | Lex(); |
2556 | return parseDirectiveEquate(IDVal: nextVal, Name: IDVal, DirKind, NameLoc: IDLoc); |
2557 | case DK_BYTE: |
2558 | if (afterNextTok.is(K: AsmToken::Identifier) && |
2559 | afterNextTok.getString().equals_insensitive(RHS: "ptr" )) { |
2560 | // Size directive; part of an instruction. |
2561 | break; |
2562 | } |
2563 | [[fallthrough]]; |
2564 | case DK_SBYTE: |
2565 | case DK_DB: |
2566 | Lex(); |
2567 | return parseDirectiveNamedValue(TypeName: nextVal, Size: 1, Name: IDVal, NameLoc: IDLoc); |
2568 | case DK_WORD: |
2569 | if (afterNextTok.is(K: AsmToken::Identifier) && |
2570 | afterNextTok.getString().equals_insensitive(RHS: "ptr" )) { |
2571 | // Size directive; part of an instruction. |
2572 | break; |
2573 | } |
2574 | [[fallthrough]]; |
2575 | case DK_SWORD: |
2576 | case DK_DW: |
2577 | Lex(); |
2578 | return parseDirectiveNamedValue(TypeName: nextVal, Size: 2, Name: IDVal, NameLoc: IDLoc); |
2579 | case DK_DWORD: |
2580 | if (afterNextTok.is(K: AsmToken::Identifier) && |
2581 | afterNextTok.getString().equals_insensitive(RHS: "ptr" )) { |
2582 | // Size directive; part of an instruction. |
2583 | break; |
2584 | } |
2585 | [[fallthrough]]; |
2586 | case DK_SDWORD: |
2587 | case DK_DD: |
2588 | Lex(); |
2589 | return parseDirectiveNamedValue(TypeName: nextVal, Size: 4, Name: IDVal, NameLoc: IDLoc); |
2590 | case DK_FWORD: |
2591 | if (afterNextTok.is(K: AsmToken::Identifier) && |
2592 | afterNextTok.getString().equals_insensitive(RHS: "ptr" )) { |
2593 | // Size directive; part of an instruction. |
2594 | break; |
2595 | } |
2596 | [[fallthrough]]; |
2597 | case DK_DF: |
2598 | Lex(); |
2599 | return parseDirectiveNamedValue(TypeName: nextVal, Size: 6, Name: IDVal, NameLoc: IDLoc); |
2600 | case DK_QWORD: |
2601 | if (afterNextTok.is(K: AsmToken::Identifier) && |
2602 | afterNextTok.getString().equals_insensitive(RHS: "ptr" )) { |
2603 | // Size directive; part of an instruction. |
2604 | break; |
2605 | } |
2606 | [[fallthrough]]; |
2607 | case DK_SQWORD: |
2608 | case DK_DQ: |
2609 | Lex(); |
2610 | return parseDirectiveNamedValue(TypeName: nextVal, Size: 8, Name: IDVal, NameLoc: IDLoc); |
2611 | case DK_REAL4: |
2612 | Lex(); |
2613 | return parseDirectiveNamedRealValue(TypeName: nextVal, Semantics: APFloat::IEEEsingle(), Size: 4, |
2614 | Name: IDVal, NameLoc: IDLoc); |
2615 | case DK_REAL8: |
2616 | Lex(); |
2617 | return parseDirectiveNamedRealValue(TypeName: nextVal, Semantics: APFloat::IEEEdouble(), Size: 8, |
2618 | Name: IDVal, NameLoc: IDLoc); |
2619 | case DK_REAL10: |
2620 | Lex(); |
2621 | return parseDirectiveNamedRealValue(TypeName: nextVal, Semantics: APFloat::x87DoubleExtended(), |
2622 | Size: 10, Name: IDVal, NameLoc: IDLoc); |
2623 | case DK_STRUCT: |
2624 | case DK_UNION: |
2625 | Lex(); |
2626 | return parseDirectiveStruct(Directive: nextVal, DirKind, Name: IDVal, NameLoc: IDLoc); |
2627 | case DK_ENDS: |
2628 | Lex(); |
2629 | return parseDirectiveEnds(Name: IDVal, NameLoc: IDLoc); |
2630 | case DK_MACRO: |
2631 | Lex(); |
2632 | return parseDirectiveMacro(Name: IDVal, NameLoc: IDLoc); |
2633 | } |
2634 | |
2635 | // Finally, we check if this is allocating a variable with user-defined type. |
2636 | auto NextIt = Structs.find(Key: nextVal.lower()); |
2637 | if (NextIt != Structs.end()) { |
2638 | Lex(); |
2639 | return parseDirectiveNamedStructValue(/*Structure=*/NextIt->getValue(), |
2640 | Directive: nextVal, DirLoc: nextLoc, Name: IDVal); |
2641 | } |
2642 | |
2643 | // __asm _emit or __asm __emit |
2644 | if (ParsingMSInlineAsm && (IDVal == "_emit" || IDVal == "__emit" || |
2645 | IDVal == "_EMIT" || IDVal == "__EMIT" )) |
2646 | return parseDirectiveMSEmit(DirectiveLoc: IDLoc, Info, Len: IDVal.size()); |
2647 | |
2648 | // __asm align |
2649 | if (ParsingMSInlineAsm && (IDVal == "align" || IDVal == "ALIGN" )) |
2650 | return parseDirectiveMSAlign(DirectiveLoc: IDLoc, Info); |
2651 | |
2652 | if (ParsingMSInlineAsm && (IDVal == "even" || IDVal == "EVEN" )) |
2653 | Info.AsmRewrites->emplace_back(Args: AOK_EVEN, Args&: IDLoc, Args: 4); |
2654 | if (checkForValidSection()) |
2655 | return true; |
2656 | |
2657 | // Canonicalize the opcode to lower case. |
2658 | std::string OpcodeStr = IDVal.lower(); |
2659 | ParseInstructionInfo IInfo(Info.AsmRewrites); |
2660 | bool ParseHadError = getTargetParser().ParseInstruction(Info&: IInfo, Name: OpcodeStr, Token: ID, |
2661 | Operands&: Info.ParsedOperands); |
2662 | Info.ParseError = ParseHadError; |
2663 | |
2664 | // Dump the parsed representation, if requested. |
2665 | if (getShowParsedOperands()) { |
2666 | SmallString<256> Str; |
2667 | raw_svector_ostream OS(Str); |
2668 | OS << "parsed instruction: [" ; |
2669 | for (unsigned i = 0; i != Info.ParsedOperands.size(); ++i) { |
2670 | if (i != 0) |
2671 | OS << ", " ; |
2672 | Info.ParsedOperands[i]->print(OS); |
2673 | } |
2674 | OS << "]" ; |
2675 | |
2676 | printMessage(Loc: IDLoc, Kind: SourceMgr::DK_Note, Msg: OS.str()); |
2677 | } |
2678 | |
2679 | // Fail even if ParseInstruction erroneously returns false. |
2680 | if (hasPendingError() || ParseHadError) |
2681 | return true; |
2682 | |
2683 | // If we are generating dwarf for the current section then generate a .loc |
2684 | // directive for the instruction. |
2685 | if (!ParseHadError && enabledGenDwarfForAssembly() && |
2686 | getContext().getGenDwarfSectionSyms().count( |
2687 | key: getStreamer().getCurrentSectionOnly())) { |
2688 | unsigned Line; |
2689 | if (ActiveMacros.empty()) |
2690 | Line = SrcMgr.FindLineNumber(Loc: IDLoc, BufferID: CurBuffer); |
2691 | else |
2692 | Line = SrcMgr.FindLineNumber(Loc: ActiveMacros.front()->InstantiationLoc, |
2693 | BufferID: ActiveMacros.front()->ExitBuffer); |
2694 | |
2695 | // If we previously parsed a cpp hash file line comment then make sure the |
2696 | // current Dwarf File is for the CppHashFilename if not then emit the |
2697 | // Dwarf File table for it and adjust the line number for the .loc. |
2698 | if (!CppHashInfo.Filename.empty()) { |
2699 | unsigned FileNumber = getStreamer().emitDwarfFileDirective( |
2700 | FileNo: 0, Directory: StringRef(), Filename: CppHashInfo.Filename); |
2701 | getContext().setGenDwarfFileNumber(FileNumber); |
2702 | |
2703 | unsigned CppHashLocLineNo = |
2704 | SrcMgr.FindLineNumber(Loc: CppHashInfo.Loc, BufferID: CppHashInfo.Buf); |
2705 | Line = CppHashInfo.LineNumber - 1 + (Line - CppHashLocLineNo); |
2706 | } |
2707 | |
2708 | getStreamer().emitDwarfLocDirective( |
2709 | FileNo: getContext().getGenDwarfFileNumber(), Line, Column: 0, |
2710 | DWARF2_LINE_DEFAULT_IS_STMT ? DWARF2_FLAG_IS_STMT : 0, Isa: 0, Discriminator: 0, |
2711 | FileName: StringRef()); |
2712 | } |
2713 | |
2714 | // If parsing succeeded, match the instruction. |
2715 | if (!ParseHadError) { |
2716 | uint64_t ErrorInfo; |
2717 | if (getTargetParser().MatchAndEmitInstruction( |
2718 | IDLoc, Opcode&: Info.Opcode, Operands&: Info.ParsedOperands, Out, ErrorInfo, |
2719 | MatchingInlineAsm: getTargetParser().isParsingMSInlineAsm())) |
2720 | return true; |
2721 | } |
2722 | return false; |
2723 | } |
2724 | |
2725 | // Parse and erase curly braces marking block start/end. |
2726 | bool MasmParser::parseCurlyBlockScope( |
2727 | SmallVectorImpl<AsmRewrite> &AsmStrRewrites) { |
2728 | // Identify curly brace marking block start/end. |
2729 | if (Lexer.isNot(K: AsmToken::LCurly) && Lexer.isNot(K: AsmToken::RCurly)) |
2730 | return false; |
2731 | |
2732 | SMLoc StartLoc = Lexer.getLoc(); |
2733 | Lex(); // Eat the brace. |
2734 | if (Lexer.is(K: AsmToken::EndOfStatement)) |
2735 | Lex(); // Eat EndOfStatement following the brace. |
2736 | |
2737 | // Erase the block start/end brace from the output asm string. |
2738 | AsmStrRewrites.emplace_back(Args: AOK_Skip, Args&: StartLoc, Args: Lexer.getLoc().getPointer() - |
2739 | StartLoc.getPointer()); |
2740 | return true; |
2741 | } |
2742 | |
2743 | /// parseCppHashLineFilenameComment as this: |
2744 | /// ::= # number "filename" |
2745 | bool MasmParser::(SMLoc L) { |
2746 | Lex(); // Eat the hash token. |
2747 | // Lexer only ever emits HashDirective if it fully formed if it's |
2748 | // done the checking already so this is an internal error. |
2749 | assert(getTok().is(AsmToken::Integer) && |
2750 | "Lexing Cpp line comment: Expected Integer" ); |
2751 | int64_t LineNumber = getTok().getIntVal(); |
2752 | Lex(); |
2753 | assert(getTok().is(AsmToken::String) && |
2754 | "Lexing Cpp line comment: Expected String" ); |
2755 | StringRef Filename = getTok().getString(); |
2756 | Lex(); |
2757 | |
2758 | // Get rid of the enclosing quotes. |
2759 | Filename = Filename.substr(Start: 1, N: Filename.size() - 2); |
2760 | |
2761 | // Save the SMLoc, Filename and LineNumber for later use by diagnostics |
2762 | // and possibly DWARF file info. |
2763 | CppHashInfo.Loc = L; |
2764 | CppHashInfo.Filename = Filename; |
2765 | CppHashInfo.LineNumber = LineNumber; |
2766 | CppHashInfo.Buf = CurBuffer; |
2767 | if (FirstCppHashFilename.empty()) |
2768 | FirstCppHashFilename = Filename; |
2769 | return false; |
2770 | } |
2771 | |
2772 | /// will use the last parsed cpp hash line filename comment |
2773 | /// for the Filename and LineNo if any in the diagnostic. |
2774 | void MasmParser::DiagHandler(const SMDiagnostic &Diag, void *Context) { |
2775 | const MasmParser *Parser = static_cast<const MasmParser *>(Context); |
2776 | raw_ostream &OS = errs(); |
2777 | |
2778 | const SourceMgr &DiagSrcMgr = *Diag.getSourceMgr(); |
2779 | SMLoc DiagLoc = Diag.getLoc(); |
2780 | unsigned DiagBuf = DiagSrcMgr.FindBufferContainingLoc(Loc: DiagLoc); |
2781 | unsigned CppHashBuf = |
2782 | Parser->SrcMgr.FindBufferContainingLoc(Loc: Parser->CppHashInfo.Loc); |
2783 | |
2784 | // Like SourceMgr::printMessage() we need to print the include stack if any |
2785 | // before printing the message. |
2786 | unsigned DiagCurBuffer = DiagSrcMgr.FindBufferContainingLoc(Loc: DiagLoc); |
2787 | if (!Parser->SavedDiagHandler && DiagCurBuffer && |
2788 | DiagCurBuffer != DiagSrcMgr.getMainFileID()) { |
2789 | SMLoc ParentIncludeLoc = DiagSrcMgr.getParentIncludeLoc(i: DiagCurBuffer); |
2790 | DiagSrcMgr.PrintIncludeStack(IncludeLoc: ParentIncludeLoc, OS); |
2791 | } |
2792 | |
2793 | // If we have not parsed a cpp hash line filename comment or the source |
2794 | // manager changed or buffer changed (like in a nested include) then just |
2795 | // print the normal diagnostic using its Filename and LineNo. |
2796 | if (!Parser->CppHashInfo.LineNumber || &DiagSrcMgr != &Parser->SrcMgr || |
2797 | DiagBuf != CppHashBuf) { |
2798 | if (Parser->SavedDiagHandler) |
2799 | Parser->SavedDiagHandler(Diag, Parser->SavedDiagContext); |
2800 | else |
2801 | Diag.print(ProgName: nullptr, S&: OS); |
2802 | return; |
2803 | } |
2804 | |
2805 | // Use the CppHashFilename and calculate a line number based on the |
2806 | // CppHashInfo.Loc and CppHashInfo.LineNumber relative to this Diag's SMLoc |
2807 | // for the diagnostic. |
2808 | const std::string &Filename = std::string(Parser->CppHashInfo.Filename); |
2809 | |
2810 | int DiagLocLineNo = DiagSrcMgr.FindLineNumber(Loc: DiagLoc, BufferID: DiagBuf); |
2811 | int CppHashLocLineNo = |
2812 | Parser->SrcMgr.FindLineNumber(Loc: Parser->CppHashInfo.Loc, BufferID: CppHashBuf); |
2813 | int LineNo = |
2814 | Parser->CppHashInfo.LineNumber - 1 + (DiagLocLineNo - CppHashLocLineNo); |
2815 | |
2816 | SMDiagnostic NewDiag(*Diag.getSourceMgr(), Diag.getLoc(), Filename, LineNo, |
2817 | Diag.getColumnNo(), Diag.getKind(), Diag.getMessage(), |
2818 | Diag.getLineContents(), Diag.getRanges()); |
2819 | |
2820 | if (Parser->SavedDiagHandler) |
2821 | Parser->SavedDiagHandler(NewDiag, Parser->SavedDiagContext); |
2822 | else |
2823 | NewDiag.print(ProgName: nullptr, S&: OS); |
2824 | } |
2825 | |
2826 | // This is similar to the IsIdentifierChar function in AsmLexer.cpp, but does |
2827 | // not accept '.'. |
2828 | static bool isMacroParameterChar(char C) { |
2829 | return isAlnum(C) || C == '_' || C == '$' || C == '@' || C == '?'; |
2830 | } |
2831 | |
2832 | bool MasmParser::expandMacro(raw_svector_ostream &OS, StringRef Body, |
2833 | ArrayRef<MCAsmMacroParameter> Parameters, |
2834 | ArrayRef<MCAsmMacroArgument> A, |
2835 | const std::vector<std::string> &Locals, SMLoc L) { |
2836 | unsigned NParameters = Parameters.size(); |
2837 | if (NParameters != A.size()) |
2838 | return Error(L, Msg: "Wrong number of arguments" ); |
2839 | StringMap<std::string> LocalSymbols; |
2840 | std::string Name; |
2841 | Name.reserve(res_arg: 6); |
2842 | for (StringRef Local : Locals) { |
2843 | raw_string_ostream LocalName(Name); |
2844 | LocalName << "??" |
2845 | << format_hex_no_prefix(N: LocalCounter++, Width: 4, /*Upper=*/true); |
2846 | LocalSymbols.insert(KV: {Local, Name}); |
2847 | Name.clear(); |
2848 | } |
2849 | |
2850 | std::optional<char> CurrentQuote; |
2851 | while (!Body.empty()) { |
2852 | // Scan for the next substitution. |
2853 | std::size_t End = Body.size(), Pos = 0; |
2854 | std::size_t IdentifierPos = End; |
2855 | for (; Pos != End; ++Pos) { |
2856 | // Find the next possible macro parameter, including preceding a '&' |
2857 | // inside quotes. |
2858 | if (Body[Pos] == '&') |
2859 | break; |
2860 | if (isMacroParameterChar(C: Body[Pos])) { |
2861 | if (!CurrentQuote) |
2862 | break; |
2863 | if (IdentifierPos == End) |
2864 | IdentifierPos = Pos; |
2865 | } else { |
2866 | IdentifierPos = End; |
2867 | } |
2868 | |
2869 | // Track quotation status |
2870 | if (!CurrentQuote) { |
2871 | if (Body[Pos] == '\'' || Body[Pos] == '"') |
2872 | CurrentQuote = Body[Pos]; |
2873 | } else if (Body[Pos] == CurrentQuote) { |
2874 | if (Pos + 1 != End && Body[Pos + 1] == CurrentQuote) { |
2875 | // Escaped quote, and quotes aren't identifier chars; skip |
2876 | ++Pos; |
2877 | continue; |
2878 | } else { |
2879 | CurrentQuote.reset(); |
2880 | } |
2881 | } |
2882 | } |
2883 | if (IdentifierPos != End) { |
2884 | // We've recognized an identifier before an apostrophe inside quotes; |
2885 | // check once to see if we can expand it. |
2886 | Pos = IdentifierPos; |
2887 | IdentifierPos = End; |
2888 | } |
2889 | |
2890 | // Add the prefix. |
2891 | OS << Body.slice(Start: 0, End: Pos); |
2892 | |
2893 | // Check if we reached the end. |
2894 | if (Pos == End) |
2895 | break; |
2896 | |
2897 | unsigned I = Pos; |
2898 | bool InitialAmpersand = (Body[I] == '&'); |
2899 | if (InitialAmpersand) { |
2900 | ++I; |
2901 | ++Pos; |
2902 | } |
2903 | while (I < End && isMacroParameterChar(C: Body[I])) |
2904 | ++I; |
2905 | |
2906 | const char *Begin = Body.data() + Pos; |
2907 | StringRef Argument(Begin, I - Pos); |
2908 | const std::string ArgumentLower = Argument.lower(); |
2909 | unsigned Index = 0; |
2910 | |
2911 | for (; Index < NParameters; ++Index) |
2912 | if (Parameters[Index].Name.equals_insensitive(RHS: ArgumentLower)) |
2913 | break; |
2914 | |
2915 | if (Index == NParameters) { |
2916 | if (InitialAmpersand) |
2917 | OS << '&'; |
2918 | auto it = LocalSymbols.find(Key: ArgumentLower); |
2919 | if (it != LocalSymbols.end()) |
2920 | OS << it->second; |
2921 | else |
2922 | OS << Argument; |
2923 | Pos = I; |
2924 | } else { |
2925 | for (const AsmToken &Token : A[Index]) { |
2926 | // In MASM, you can write '%expr'. |
2927 | // The prefix '%' evaluates the expression 'expr' |
2928 | // and uses the result as a string (e.g. replace %(1+2) with the |
2929 | // string "3"). |
2930 | // Here, we identify the integer token which is the result of the |
2931 | // absolute expression evaluation and replace it with its string |
2932 | // representation. |
2933 | if (Token.getString().front() == '%' && Token.is(K: AsmToken::Integer)) |
2934 | // Emit an integer value to the buffer. |
2935 | OS << Token.getIntVal(); |
2936 | else |
2937 | OS << Token.getString(); |
2938 | } |
2939 | |
2940 | Pos += Argument.size(); |
2941 | if (Pos < End && Body[Pos] == '&') { |
2942 | ++Pos; |
2943 | } |
2944 | } |
2945 | // Update the scan point. |
2946 | Body = Body.substr(Start: Pos); |
2947 | } |
2948 | |
2949 | return false; |
2950 | } |
2951 | |
2952 | static bool isOperator(AsmToken::TokenKind kind) { |
2953 | switch (kind) { |
2954 | default: |
2955 | return false; |
2956 | case AsmToken::Plus: |
2957 | case AsmToken::Minus: |
2958 | case AsmToken::Tilde: |
2959 | case AsmToken::Slash: |
2960 | case AsmToken::Star: |
2961 | case AsmToken::Dot: |
2962 | case AsmToken::Equal: |
2963 | case AsmToken::EqualEqual: |
2964 | case AsmToken::Pipe: |
2965 | case AsmToken::PipePipe: |
2966 | case AsmToken::Caret: |
2967 | case AsmToken::Amp: |
2968 | case AsmToken::AmpAmp: |
2969 | case AsmToken::Exclaim: |
2970 | case AsmToken::ExclaimEqual: |
2971 | case AsmToken::Less: |
2972 | case AsmToken::LessEqual: |
2973 | case AsmToken::LessLess: |
2974 | case AsmToken::LessGreater: |
2975 | case AsmToken::Greater: |
2976 | case AsmToken::GreaterEqual: |
2977 | case AsmToken::GreaterGreater: |
2978 | return true; |
2979 | } |
2980 | } |
2981 | |
2982 | namespace { |
2983 | |
2984 | class AsmLexerSkipSpaceRAII { |
2985 | public: |
2986 | AsmLexerSkipSpaceRAII(AsmLexer &Lexer, bool SkipSpace) : Lexer(Lexer) { |
2987 | Lexer.setSkipSpace(SkipSpace); |
2988 | } |
2989 | |
2990 | ~AsmLexerSkipSpaceRAII() { |
2991 | Lexer.setSkipSpace(true); |
2992 | } |
2993 | |
2994 | private: |
2995 | AsmLexer &Lexer; |
2996 | }; |
2997 | |
2998 | } // end anonymous namespace |
2999 | |
3000 | bool MasmParser::parseMacroArgument(const MCAsmMacroParameter *MP, |
3001 | MCAsmMacroArgument &MA, |
3002 | AsmToken::TokenKind EndTok) { |
3003 | if (MP && MP->Vararg) { |
3004 | if (Lexer.isNot(K: EndTok)) { |
3005 | SmallVector<StringRef, 1> Str = parseStringRefsTo(EndTok); |
3006 | for (StringRef S : Str) { |
3007 | MA.emplace_back(args: AsmToken::String, args&: S); |
3008 | } |
3009 | } |
3010 | return false; |
3011 | } |
3012 | |
3013 | SMLoc StrLoc = Lexer.getLoc(), EndLoc; |
3014 | if (Lexer.is(K: AsmToken::Less) && isAngleBracketString(StrLoc, EndLoc)) { |
3015 | const char *StrChar = StrLoc.getPointer() + 1; |
3016 | const char *EndChar = EndLoc.getPointer() - 1; |
3017 | jumpToLoc(Loc: EndLoc, InBuffer: CurBuffer, EndStatementAtEOF: EndStatementAtEOFStack.back()); |
3018 | /// Eat from '<' to '>'. |
3019 | Lex(); |
3020 | MA.emplace_back(args: AsmToken::String, args: StringRef(StrChar, EndChar - StrChar)); |
3021 | return false; |
3022 | } |
3023 | |
3024 | unsigned ParenLevel = 0; |
3025 | |
3026 | // Darwin doesn't use spaces to delmit arguments. |
3027 | AsmLexerSkipSpaceRAII ScopedSkipSpace(Lexer, IsDarwin); |
3028 | |
3029 | bool SpaceEaten; |
3030 | |
3031 | while (true) { |
3032 | SpaceEaten = false; |
3033 | if (Lexer.is(K: AsmToken::Eof) || Lexer.is(K: AsmToken::Equal)) |
3034 | return TokError(Msg: "unexpected token" ); |
3035 | |
3036 | if (ParenLevel == 0) { |
3037 | if (Lexer.is(K: AsmToken::Comma)) |
3038 | break; |
3039 | |
3040 | if (Lexer.is(K: AsmToken::Space)) { |
3041 | SpaceEaten = true; |
3042 | Lex(); // Eat spaces. |
3043 | } |
3044 | |
3045 | // Spaces can delimit parameters, but could also be part an expression. |
3046 | // If the token after a space is an operator, add the token and the next |
3047 | // one into this argument |
3048 | if (!IsDarwin) { |
3049 | if (isOperator(kind: Lexer.getKind()) && Lexer.isNot(K: EndTok)) { |
3050 | MA.push_back(x: getTok()); |
3051 | Lex(); |
3052 | |
3053 | // Whitespace after an operator can be ignored. |
3054 | if (Lexer.is(K: AsmToken::Space)) |
3055 | Lex(); |
3056 | |
3057 | continue; |
3058 | } |
3059 | } |
3060 | if (SpaceEaten) |
3061 | break; |
3062 | } |
3063 | |
3064 | // handleMacroEntry relies on not advancing the lexer here |
3065 | // to be able to fill in the remaining default parameter values |
3066 | if (Lexer.is(K: EndTok) && (EndTok != AsmToken::RParen || ParenLevel == 0)) |
3067 | break; |
3068 | |
3069 | // Adjust the current parentheses level. |
3070 | if (Lexer.is(K: AsmToken::LParen)) |
3071 | ++ParenLevel; |
3072 | else if (Lexer.is(K: AsmToken::RParen) && ParenLevel) |
3073 | --ParenLevel; |
3074 | |
3075 | // Append the token to the current argument list. |
3076 | MA.push_back(x: getTok()); |
3077 | Lex(); |
3078 | } |
3079 | |
3080 | if (ParenLevel != 0) |
3081 | return TokError(Msg: "unbalanced parentheses in argument" ); |
3082 | |
3083 | if (MA.empty() && MP) { |
3084 | if (MP->Required) { |
3085 | return TokError(Msg: "missing value for required parameter '" + MP->Name + |
3086 | "'" ); |
3087 | } else { |
3088 | MA = MP->Value; |
3089 | } |
3090 | } |
3091 | return false; |
3092 | } |
3093 | |
3094 | // Parse the macro instantiation arguments. |
3095 | bool MasmParser::parseMacroArguments(const MCAsmMacro *M, |
3096 | MCAsmMacroArguments &A, |
3097 | AsmToken::TokenKind EndTok) { |
3098 | const unsigned NParameters = M ? M->Parameters.size() : 0; |
3099 | bool NamedParametersFound = false; |
3100 | SmallVector<SMLoc, 4> FALocs; |
3101 | |
3102 | A.resize(new_size: NParameters); |
3103 | FALocs.resize(N: NParameters); |
3104 | |
3105 | // Parse two kinds of macro invocations: |
3106 | // - macros defined without any parameters accept an arbitrary number of them |
3107 | // - macros defined with parameters accept at most that many of them |
3108 | for (unsigned Parameter = 0; !NParameters || Parameter < NParameters; |
3109 | ++Parameter) { |
3110 | SMLoc IDLoc = Lexer.getLoc(); |
3111 | MCAsmMacroParameter FA; |
3112 | |
3113 | if (Lexer.is(K: AsmToken::Identifier) && peekTok().is(K: AsmToken::Equal)) { |
3114 | if (parseIdentifier(Res&: FA.Name)) |
3115 | return Error(L: IDLoc, Msg: "invalid argument identifier for formal argument" ); |
3116 | |
3117 | if (Lexer.isNot(K: AsmToken::Equal)) |
3118 | return TokError(Msg: "expected '=' after formal parameter identifier" ); |
3119 | |
3120 | Lex(); |
3121 | |
3122 | NamedParametersFound = true; |
3123 | } |
3124 | |
3125 | if (NamedParametersFound && FA.Name.empty()) |
3126 | return Error(L: IDLoc, Msg: "cannot mix positional and keyword arguments" ); |
3127 | |
3128 | unsigned PI = Parameter; |
3129 | if (!FA.Name.empty()) { |
3130 | assert(M && "expected macro to be defined" ); |
3131 | unsigned FAI = 0; |
3132 | for (FAI = 0; FAI < NParameters; ++FAI) |
3133 | if (M->Parameters[FAI].Name == FA.Name) |
3134 | break; |
3135 | |
3136 | if (FAI >= NParameters) { |
3137 | return Error(L: IDLoc, Msg: "parameter named '" + FA.Name + |
3138 | "' does not exist for macro '" + M->Name + "'" ); |
3139 | } |
3140 | PI = FAI; |
3141 | } |
3142 | const MCAsmMacroParameter *MP = nullptr; |
3143 | if (M && PI < NParameters) |
3144 | MP = &M->Parameters[PI]; |
3145 | |
3146 | SMLoc StrLoc = Lexer.getLoc(); |
3147 | SMLoc EndLoc; |
3148 | if (Lexer.is(K: AsmToken::Percent)) { |
3149 | const MCExpr *AbsoluteExp; |
3150 | int64_t Value; |
3151 | /// Eat '%'. |
3152 | Lex(); |
3153 | if (parseExpression(Res&: AbsoluteExp, EndLoc)) |
3154 | return false; |
3155 | if (!AbsoluteExp->evaluateAsAbsolute(Res&: Value, |
3156 | Asm: getStreamer().getAssemblerPtr())) |
3157 | return Error(L: StrLoc, Msg: "expected absolute expression" ); |
3158 | const char *StrChar = StrLoc.getPointer(); |
3159 | const char *EndChar = EndLoc.getPointer(); |
3160 | AsmToken newToken(AsmToken::Integer, |
3161 | StringRef(StrChar, EndChar - StrChar), Value); |
3162 | FA.Value.push_back(x: newToken); |
3163 | } else if (parseMacroArgument(MP, MA&: FA.Value, EndTok)) { |
3164 | if (M) |
3165 | return addErrorSuffix(Suffix: " in '" + M->Name + "' macro" ); |
3166 | else |
3167 | return true; |
3168 | } |
3169 | |
3170 | if (!FA.Value.empty()) { |
3171 | if (A.size() <= PI) |
3172 | A.resize(new_size: PI + 1); |
3173 | A[PI] = FA.Value; |
3174 | |
3175 | if (FALocs.size() <= PI) |
3176 | FALocs.resize(N: PI + 1); |
3177 | |
3178 | FALocs[PI] = Lexer.getLoc(); |
3179 | } |
3180 | |
3181 | // At the end of the statement, fill in remaining arguments that have |
3182 | // default values. If there aren't any, then the next argument is |
3183 | // required but missing |
3184 | if (Lexer.is(K: EndTok)) { |
3185 | bool Failure = false; |
3186 | for (unsigned FAI = 0; FAI < NParameters; ++FAI) { |
3187 | if (A[FAI].empty()) { |
3188 | if (M->Parameters[FAI].Required) { |
3189 | Error(L: FALocs[FAI].isValid() ? FALocs[FAI] : Lexer.getLoc(), |
3190 | Msg: "missing value for required parameter " |
3191 | "'" + |
3192 | M->Parameters[FAI].Name + "' in macro '" + M->Name + "'" ); |
3193 | Failure = true; |
3194 | } |
3195 | |
3196 | if (!M->Parameters[FAI].Value.empty()) |
3197 | A[FAI] = M->Parameters[FAI].Value; |
3198 | } |
3199 | } |
3200 | return Failure; |
3201 | } |
3202 | |
3203 | if (Lexer.is(K: AsmToken::Comma)) |
3204 | Lex(); |
3205 | } |
3206 | |
3207 | return TokError(Msg: "too many positional arguments" ); |
3208 | } |
3209 | |
3210 | bool MasmParser::handleMacroEntry(const MCAsmMacro *M, SMLoc NameLoc, |
3211 | AsmToken::TokenKind ArgumentEndTok) { |
3212 | // Arbitrarily limit macro nesting depth (default matches 'as'). We can |
3213 | // eliminate this, although we should protect against infinite loops. |
3214 | unsigned MaxNestingDepth = AsmMacroMaxNestingDepth; |
3215 | if (ActiveMacros.size() == MaxNestingDepth) { |
3216 | std::ostringstream MaxNestingDepthError; |
3217 | MaxNestingDepthError << "macros cannot be nested more than " |
3218 | << MaxNestingDepth << " levels deep." |
3219 | << " Use -asm-macro-max-nesting-depth to increase " |
3220 | "this limit." ; |
3221 | return TokError(Msg: MaxNestingDepthError.str()); |
3222 | } |
3223 | |
3224 | MCAsmMacroArguments A; |
3225 | if (parseMacroArguments(M, A, EndTok: ArgumentEndTok)) |
3226 | return true; |
3227 | |
3228 | // Macro instantiation is lexical, unfortunately. We construct a new buffer |
3229 | // to hold the macro body with substitutions. |
3230 | SmallString<256> Buf; |
3231 | StringRef Body = M->Body; |
3232 | raw_svector_ostream OS(Buf); |
3233 | |
3234 | if (expandMacro(OS, Body, Parameters: M->Parameters, A, Locals: M->Locals, L: getTok().getLoc())) |
3235 | return true; |
3236 | |
3237 | // We include the endm in the buffer as our cue to exit the macro |
3238 | // instantiation. |
3239 | OS << "endm\n" ; |
3240 | |
3241 | std::unique_ptr<MemoryBuffer> Instantiation = |
3242 | MemoryBuffer::getMemBufferCopy(InputData: OS.str(), BufferName: "<instantiation>" ); |
3243 | |
3244 | // Create the macro instantiation object and add to the current macro |
3245 | // instantiation stack. |
3246 | MacroInstantiation *MI = new MacroInstantiation{ |
3247 | .InstantiationLoc: NameLoc, .ExitBuffer: CurBuffer, .ExitLoc: getTok().getLoc(), .CondStackDepth: TheCondStack.size()}; |
3248 | ActiveMacros.push_back(x: MI); |
3249 | |
3250 | ++NumOfMacroInstantiations; |
3251 | |
3252 | // Jump to the macro instantiation and prime the lexer. |
3253 | CurBuffer = SrcMgr.AddNewSourceBuffer(F: std::move(Instantiation), IncludeLoc: SMLoc()); |
3254 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer()); |
3255 | EndStatementAtEOFStack.push_back(Val: true); |
3256 | Lex(); |
3257 | |
3258 | return false; |
3259 | } |
3260 | |
3261 | void MasmParser::handleMacroExit() { |
3262 | // Jump to the token we should return to, and consume it. |
3263 | EndStatementAtEOFStack.pop_back(); |
3264 | jumpToLoc(Loc: ActiveMacros.back()->ExitLoc, InBuffer: ActiveMacros.back()->ExitBuffer, |
3265 | EndStatementAtEOF: EndStatementAtEOFStack.back()); |
3266 | Lex(); |
3267 | |
3268 | // Pop the instantiation entry. |
3269 | delete ActiveMacros.back(); |
3270 | ActiveMacros.pop_back(); |
3271 | } |
3272 | |
3273 | bool MasmParser::handleMacroInvocation(const MCAsmMacro *M, SMLoc NameLoc) { |
3274 | if (!M->IsFunction) |
3275 | return Error(L: NameLoc, Msg: "cannot invoke macro procedure as function" ); |
3276 | |
3277 | if (parseToken(T: AsmToken::LParen, Msg: "invoking macro function '" + M->Name + |
3278 | "' requires arguments in parentheses" ) || |
3279 | handleMacroEntry(M, NameLoc, ArgumentEndTok: AsmToken::RParen)) |
3280 | return true; |
3281 | |
3282 | // Parse all statements in the macro, retrieving the exit value when it ends. |
3283 | std::string ExitValue; |
3284 | SmallVector<AsmRewrite, 4> AsmStrRewrites; |
3285 | while (Lexer.isNot(K: AsmToken::Eof)) { |
3286 | ParseStatementInfo Info(&AsmStrRewrites); |
3287 | bool Parsed = parseStatement(Info, SI: nullptr); |
3288 | |
3289 | if (!Parsed && Info.ExitValue) { |
3290 | ExitValue = std::move(*Info.ExitValue); |
3291 | break; |
3292 | } |
3293 | |
3294 | // If we have a Lexer Error we are on an Error Token. Load in Lexer Error |
3295 | // for printing ErrMsg via Lex() only if no (presumably better) parser error |
3296 | // exists. |
3297 | if (Parsed && !hasPendingError() && Lexer.getTok().is(K: AsmToken::Error)) { |
3298 | Lex(); |
3299 | } |
3300 | |
3301 | // parseStatement returned true so may need to emit an error. |
3302 | printPendingErrors(); |
3303 | |
3304 | // Skipping to the next line if needed. |
3305 | if (Parsed && !getLexer().isAtStartOfStatement()) |
3306 | eatToEndOfStatement(); |
3307 | } |
3308 | |
3309 | // Consume the right-parenthesis on the other side of the arguments. |
3310 | if (parseRParen()) |
3311 | return true; |
3312 | |
3313 | // Exit values may require lexing, unfortunately. We construct a new buffer to |
3314 | // hold the exit value. |
3315 | std::unique_ptr<MemoryBuffer> MacroValue = |
3316 | MemoryBuffer::getMemBufferCopy(InputData: ExitValue, BufferName: "<macro-value>" ); |
3317 | |
3318 | // Jump from this location to the instantiated exit value, and prime the |
3319 | // lexer. |
3320 | CurBuffer = SrcMgr.AddNewSourceBuffer(F: std::move(MacroValue), IncludeLoc: Lexer.getLoc()); |
3321 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer(), ptr: nullptr, |
3322 | /*EndStatementAtEOF=*/false); |
3323 | EndStatementAtEOFStack.push_back(Val: false); |
3324 | Lex(); |
3325 | |
3326 | return false; |
3327 | } |
3328 | |
3329 | /// parseIdentifier: |
3330 | /// ::= identifier |
3331 | /// ::= string |
3332 | bool MasmParser::parseIdentifier(StringRef &Res, |
3333 | IdentifierPositionKind Position) { |
3334 | // The assembler has relaxed rules for accepting identifiers, in particular we |
3335 | // allow things like '.globl $foo' and '.def @feat.00', which would normally |
3336 | // be separate tokens. At this level, we have already lexed so we cannot |
3337 | // (currently) handle this as a context dependent token, instead we detect |
3338 | // adjacent tokens and return the combined identifier. |
3339 | if (Lexer.is(K: AsmToken::Dollar) || Lexer.is(K: AsmToken::At)) { |
3340 | SMLoc PrefixLoc = getLexer().getLoc(); |
3341 | |
3342 | // Consume the prefix character, and check for a following identifier. |
3343 | |
3344 | AsmToken nextTok = peekTok(ShouldSkipSpace: false); |
3345 | |
3346 | if (nextTok.isNot(K: AsmToken::Identifier)) |
3347 | return true; |
3348 | |
3349 | // We have a '$' or '@' followed by an identifier, make sure they are adjacent. |
3350 | if (PrefixLoc.getPointer() + 1 != nextTok.getLoc().getPointer()) |
3351 | return true; |
3352 | |
3353 | // eat $ or @ |
3354 | Lexer.Lex(); // Lexer's Lex guarantees consecutive token. |
3355 | // Construct the joined identifier and consume the token. |
3356 | Res = |
3357 | StringRef(PrefixLoc.getPointer(), getTok().getIdentifier().size() + 1); |
3358 | Lex(); // Parser Lex to maintain invariants. |
3359 | return false; |
3360 | } |
3361 | |
3362 | if (Lexer.isNot(K: AsmToken::Identifier) && Lexer.isNot(K: AsmToken::String)) |
3363 | return true; |
3364 | |
3365 | Res = getTok().getIdentifier(); |
3366 | |
3367 | // Consume the identifier token - but if parsing certain directives, avoid |
3368 | // lexical expansion of the next token. |
3369 | ExpandKind ExpandNextToken = ExpandMacros; |
3370 | if (Position == StartOfStatement && |
3371 | StringSwitch<bool>(Res) |
3372 | .CaseLower(S: "echo" , Value: true) |
3373 | .CasesLower(S0: "ifdef" , S1: "ifndef" , S2: "elseifdef" , S3: "elseifndef" , Value: true) |
3374 | .Default(Value: false)) { |
3375 | ExpandNextToken = DoNotExpandMacros; |
3376 | } |
3377 | Lex(ExpandNextToken); |
3378 | |
3379 | return false; |
3380 | } |
3381 | |
3382 | /// parseDirectiveEquate: |
3383 | /// ::= name "=" expression |
3384 | /// | name "equ" expression (not redefinable) |
3385 | /// | name "equ" text-list |
3386 | /// | name "textequ" text-list (redefinability unspecified) |
3387 | bool MasmParser::parseDirectiveEquate(StringRef IDVal, StringRef Name, |
3388 | DirectiveKind DirKind, SMLoc NameLoc) { |
3389 | auto BuiltinIt = BuiltinSymbolMap.find(Key: Name.lower()); |
3390 | if (BuiltinIt != BuiltinSymbolMap.end()) |
3391 | return Error(L: NameLoc, Msg: "cannot redefine a built-in symbol" ); |
3392 | |
3393 | Variable &Var = Variables[Name.lower()]; |
3394 | if (Var.Name.empty()) { |
3395 | Var.Name = Name; |
3396 | } |
3397 | |
3398 | SMLoc StartLoc = Lexer.getLoc(); |
3399 | if (DirKind == DK_EQU || DirKind == DK_TEXTEQU) { |
3400 | // "equ" and "textequ" both allow text expressions. |
3401 | std::string Value; |
3402 | std::string TextItem; |
3403 | if (!parseTextItem(Data&: TextItem)) { |
3404 | Value += TextItem; |
3405 | |
3406 | // Accept a text-list, not just one text-item. |
3407 | auto parseItem = [&]() -> bool { |
3408 | if (parseTextItem(Data&: TextItem)) |
3409 | return TokError(Msg: "expected text item" ); |
3410 | Value += TextItem; |
3411 | return false; |
3412 | }; |
3413 | if (parseOptionalToken(T: AsmToken::Comma) && parseMany(parseOne: parseItem)) |
3414 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3415 | |
3416 | if (!Var.IsText || Var.TextValue != Value) { |
3417 | switch (Var.Redefinable) { |
3418 | case Variable::NOT_REDEFINABLE: |
3419 | return Error(L: getTok().getLoc(), Msg: "invalid variable redefinition" ); |
3420 | case Variable::WARN_ON_REDEFINITION: |
3421 | if (Warning(L: NameLoc, Msg: "redefining '" + Name + |
3422 | "', already defined on the command line" )) { |
3423 | return true; |
3424 | } |
3425 | break; |
3426 | default: |
3427 | break; |
3428 | } |
3429 | } |
3430 | Var.IsText = true; |
3431 | Var.TextValue = Value; |
3432 | Var.Redefinable = Variable::REDEFINABLE; |
3433 | |
3434 | return false; |
3435 | } |
3436 | } |
3437 | if (DirKind == DK_TEXTEQU) |
3438 | return TokError(Msg: "expected <text> in '" + Twine(IDVal) + "' directive" ); |
3439 | |
3440 | // Parse as expression assignment. |
3441 | const MCExpr *Expr; |
3442 | SMLoc EndLoc; |
3443 | if (parseExpression(Res&: Expr, EndLoc)) |
3444 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3445 | StringRef ExprAsString = StringRef( |
3446 | StartLoc.getPointer(), EndLoc.getPointer() - StartLoc.getPointer()); |
3447 | |
3448 | int64_t Value; |
3449 | if (!Expr->evaluateAsAbsolute(Res&: Value, Asm: getStreamer().getAssemblerPtr())) { |
3450 | if (DirKind == DK_ASSIGN) |
3451 | return Error( |
3452 | L: StartLoc, |
3453 | Msg: "expected absolute expression; not all symbols have known values" , |
3454 | Range: {StartLoc, EndLoc}); |
3455 | |
3456 | // Not an absolute expression; define as a text replacement. |
3457 | if (!Var.IsText || Var.TextValue != ExprAsString) { |
3458 | switch (Var.Redefinable) { |
3459 | case Variable::NOT_REDEFINABLE: |
3460 | return Error(L: getTok().getLoc(), Msg: "invalid variable redefinition" ); |
3461 | case Variable::WARN_ON_REDEFINITION: |
3462 | if (Warning(L: NameLoc, Msg: "redefining '" + Name + |
3463 | "', already defined on the command line" )) { |
3464 | return true; |
3465 | } |
3466 | break; |
3467 | default: |
3468 | break; |
3469 | } |
3470 | } |
3471 | |
3472 | Var.IsText = true; |
3473 | Var.TextValue = ExprAsString.str(); |
3474 | Var.Redefinable = Variable::REDEFINABLE; |
3475 | |
3476 | return false; |
3477 | } |
3478 | |
3479 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name: Var.Name); |
3480 | |
3481 | const MCConstantExpr *PrevValue = |
3482 | Sym->isVariable() ? dyn_cast_or_null<MCConstantExpr>( |
3483 | Val: Sym->getVariableValue(/*SetUsed=*/false)) |
3484 | : nullptr; |
3485 | if (Var.IsText || !PrevValue || PrevValue->getValue() != Value) { |
3486 | switch (Var.Redefinable) { |
3487 | case Variable::NOT_REDEFINABLE: |
3488 | return Error(L: getTok().getLoc(), Msg: "invalid variable redefinition" ); |
3489 | case Variable::WARN_ON_REDEFINITION: |
3490 | if (Warning(L: NameLoc, Msg: "redefining '" + Name + |
3491 | "', already defined on the command line" )) { |
3492 | return true; |
3493 | } |
3494 | break; |
3495 | default: |
3496 | break; |
3497 | } |
3498 | } |
3499 | |
3500 | Var.IsText = false; |
3501 | Var.TextValue.clear(); |
3502 | Var.Redefinable = (DirKind == DK_ASSIGN) ? Variable::REDEFINABLE |
3503 | : Variable::NOT_REDEFINABLE; |
3504 | |
3505 | Sym->setRedefinable(Var.Redefinable != Variable::NOT_REDEFINABLE); |
3506 | Sym->setVariableValue(Expr); |
3507 | Sym->setExternal(false); |
3508 | |
3509 | return false; |
3510 | } |
3511 | |
3512 | bool MasmParser::parseEscapedString(std::string &Data) { |
3513 | if (check(P: getTok().isNot(K: AsmToken::String), Msg: "expected string" )) |
3514 | return true; |
3515 | |
3516 | Data = "" ; |
3517 | char Quote = getTok().getString().front(); |
3518 | StringRef Str = getTok().getStringContents(); |
3519 | Data.reserve(res_arg: Str.size()); |
3520 | for (size_t i = 0, e = Str.size(); i != e; ++i) { |
3521 | Data.push_back(c: Str[i]); |
3522 | if (Str[i] == Quote) { |
3523 | // MASM treats doubled delimiting quotes as an escaped delimiting quote. |
3524 | // If we're escaping the string's trailing delimiter, we're definitely |
3525 | // missing a quotation mark. |
3526 | if (i + 1 == Str.size()) |
3527 | return Error(L: getTok().getLoc(), Msg: "missing quotation mark in string" ); |
3528 | if (Str[i + 1] == Quote) |
3529 | ++i; |
3530 | } |
3531 | } |
3532 | |
3533 | Lex(); |
3534 | return false; |
3535 | } |
3536 | |
3537 | bool MasmParser::parseAngleBracketString(std::string &Data) { |
3538 | SMLoc EndLoc, StartLoc = getTok().getLoc(); |
3539 | if (isAngleBracketString(StrLoc&: StartLoc, EndLoc)) { |
3540 | const char *StartChar = StartLoc.getPointer() + 1; |
3541 | const char *EndChar = EndLoc.getPointer() - 1; |
3542 | jumpToLoc(Loc: EndLoc, InBuffer: CurBuffer, EndStatementAtEOF: EndStatementAtEOFStack.back()); |
3543 | // Eat from '<' to '>'. |
3544 | Lex(); |
3545 | |
3546 | Data = angleBracketString(BracketContents: StringRef(StartChar, EndChar - StartChar)); |
3547 | return false; |
3548 | } |
3549 | return true; |
3550 | } |
3551 | |
3552 | /// textItem ::= textLiteral | textMacroID | % constExpr |
3553 | bool MasmParser::parseTextItem(std::string &Data) { |
3554 | switch (getTok().getKind()) { |
3555 | default: |
3556 | return true; |
3557 | case AsmToken::Percent: { |
3558 | int64_t Res; |
3559 | if (parseToken(T: AsmToken::Percent) || parseAbsoluteExpression(Res)) |
3560 | return true; |
3561 | Data = std::to_string(val: Res); |
3562 | return false; |
3563 | } |
3564 | case AsmToken::Less: |
3565 | case AsmToken::LessEqual: |
3566 | case AsmToken::LessLess: |
3567 | case AsmToken::LessGreater: |
3568 | return parseAngleBracketString(Data); |
3569 | case AsmToken::Identifier: { |
3570 | // This must be a text macro; we need to expand it accordingly. |
3571 | StringRef ID; |
3572 | SMLoc StartLoc = getTok().getLoc(); |
3573 | if (parseIdentifier(Res&: ID)) |
3574 | return true; |
3575 | Data = ID.str(); |
3576 | |
3577 | bool Expanded = false; |
3578 | while (true) { |
3579 | // Try to resolve as a built-in text macro |
3580 | auto BuiltinIt = BuiltinSymbolMap.find(Key: ID.lower()); |
3581 | if (BuiltinIt != BuiltinSymbolMap.end()) { |
3582 | std::optional<std::string> BuiltinText = |
3583 | evaluateBuiltinTextMacro(Symbol: BuiltinIt->getValue(), StartLoc); |
3584 | if (!BuiltinText) { |
3585 | // Not a text macro; break without substituting |
3586 | break; |
3587 | } |
3588 | Data = std::move(*BuiltinText); |
3589 | ID = StringRef(Data); |
3590 | Expanded = true; |
3591 | continue; |
3592 | } |
3593 | |
3594 | // Try to resolve as a variable text macro |
3595 | auto VarIt = Variables.find(Key: ID.lower()); |
3596 | if (VarIt != Variables.end()) { |
3597 | const Variable &Var = VarIt->getValue(); |
3598 | if (!Var.IsText) { |
3599 | // Not a text macro; break without substituting |
3600 | break; |
3601 | } |
3602 | Data = Var.TextValue; |
3603 | ID = StringRef(Data); |
3604 | Expanded = true; |
3605 | continue; |
3606 | } |
3607 | |
3608 | break; |
3609 | } |
3610 | |
3611 | if (!Expanded) { |
3612 | // Not a text macro; not usable in TextItem context. Since we haven't used |
3613 | // the token, put it back for better error recovery. |
3614 | getLexer().UnLex(Token: AsmToken(AsmToken::Identifier, ID)); |
3615 | return true; |
3616 | } |
3617 | return false; |
3618 | } |
3619 | } |
3620 | llvm_unreachable("unhandled token kind" ); |
3621 | } |
3622 | |
3623 | /// parseDirectiveAscii: |
3624 | /// ::= ( .ascii | .asciz | .string ) [ "string" ( , "string" )* ] |
3625 | bool MasmParser::parseDirectiveAscii(StringRef IDVal, bool ZeroTerminated) { |
3626 | auto parseOp = [&]() -> bool { |
3627 | std::string Data; |
3628 | if (checkForValidSection() || parseEscapedString(Data)) |
3629 | return true; |
3630 | getStreamer().emitBytes(Data); |
3631 | if (ZeroTerminated) |
3632 | getStreamer().emitBytes(Data: StringRef("\0" , 1)); |
3633 | return false; |
3634 | }; |
3635 | |
3636 | if (parseMany(parseOne: parseOp)) |
3637 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3638 | return false; |
3639 | } |
3640 | |
3641 | bool MasmParser::emitIntValue(const MCExpr *Value, unsigned Size) { |
3642 | // Special case constant expressions to match code generator. |
3643 | if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value)) { |
3644 | assert(Size <= 8 && "Invalid size" ); |
3645 | int64_t IntValue = MCE->getValue(); |
3646 | if (!isUIntN(N: 8 * Size, x: IntValue) && !isIntN(N: 8 * Size, x: IntValue)) |
3647 | return Error(L: MCE->getLoc(), Msg: "out of range literal value" ); |
3648 | getStreamer().emitIntValue(Value: IntValue, Size); |
3649 | } else { |
3650 | const MCSymbolRefExpr *MSE = dyn_cast<MCSymbolRefExpr>(Val: Value); |
3651 | if (MSE && MSE->getSymbol().getName() == "?" ) { |
3652 | // ? initializer; treat as 0. |
3653 | getStreamer().emitIntValue(Value: 0, Size); |
3654 | } else { |
3655 | getStreamer().emitValue(Value, Size, Loc: Value->getLoc()); |
3656 | } |
3657 | } |
3658 | return false; |
3659 | } |
3660 | |
3661 | bool MasmParser::parseScalarInitializer(unsigned Size, |
3662 | SmallVectorImpl<const MCExpr *> &Values, |
3663 | unsigned StringPadLength) { |
3664 | if (Size == 1 && getTok().is(K: AsmToken::String)) { |
3665 | std::string Value; |
3666 | if (parseEscapedString(Data&: Value)) |
3667 | return true; |
3668 | // Treat each character as an initializer. |
3669 | for (const unsigned char CharVal : Value) |
3670 | Values.push_back(Elt: MCConstantExpr::create(Value: CharVal, Ctx&: getContext())); |
3671 | |
3672 | // Pad the string with spaces to the specified length. |
3673 | for (size_t i = Value.size(); i < StringPadLength; ++i) |
3674 | Values.push_back(Elt: MCConstantExpr::create(Value: ' ', Ctx&: getContext())); |
3675 | } else { |
3676 | const MCExpr *Value; |
3677 | if (parseExpression(Res&: Value)) |
3678 | return true; |
3679 | if (getTok().is(K: AsmToken::Identifier) && |
3680 | getTok().getString().equals_insensitive(RHS: "dup" )) { |
3681 | Lex(); // Eat 'dup'. |
3682 | const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value); |
3683 | if (!MCE) |
3684 | return Error(L: Value->getLoc(), |
3685 | Msg: "cannot repeat value a non-constant number of times" ); |
3686 | const int64_t Repetitions = MCE->getValue(); |
3687 | if (Repetitions < 0) |
3688 | return Error(L: Value->getLoc(), |
3689 | Msg: "cannot repeat value a negative number of times" ); |
3690 | |
3691 | SmallVector<const MCExpr *, 1> DuplicatedValues; |
3692 | if (parseToken(T: AsmToken::LParen, |
3693 | Msg: "parentheses required for 'dup' contents" ) || |
3694 | parseScalarInstList(Size, Values&: DuplicatedValues) || parseRParen()) |
3695 | return true; |
3696 | |
3697 | for (int i = 0; i < Repetitions; ++i) |
3698 | Values.append(in_start: DuplicatedValues.begin(), in_end: DuplicatedValues.end()); |
3699 | } else { |
3700 | Values.push_back(Elt: Value); |
3701 | } |
3702 | } |
3703 | return false; |
3704 | } |
3705 | |
3706 | bool MasmParser::parseScalarInstList(unsigned Size, |
3707 | SmallVectorImpl<const MCExpr *> &Values, |
3708 | const AsmToken::TokenKind EndToken) { |
3709 | while (getTok().isNot(K: EndToken) && |
3710 | (EndToken != AsmToken::Greater || |
3711 | getTok().isNot(K: AsmToken::GreaterGreater))) { |
3712 | parseScalarInitializer(Size, Values); |
3713 | |
3714 | // If we see a comma, continue, and allow line continuation. |
3715 | if (!parseOptionalToken(T: AsmToken::Comma)) |
3716 | break; |
3717 | parseOptionalToken(T: AsmToken::EndOfStatement); |
3718 | } |
3719 | return false; |
3720 | } |
3721 | |
3722 | bool MasmParser::emitIntegralValues(unsigned Size, unsigned *Count) { |
3723 | SmallVector<const MCExpr *, 1> Values; |
3724 | if (checkForValidSection() || parseScalarInstList(Size, Values)) |
3725 | return true; |
3726 | |
3727 | for (const auto *Value : Values) { |
3728 | emitIntValue(Value, Size); |
3729 | } |
3730 | if (Count) |
3731 | *Count = Values.size(); |
3732 | return false; |
3733 | } |
3734 | |
3735 | // Add a field to the current structure. |
3736 | bool MasmParser::addIntegralField(StringRef Name, unsigned Size) { |
3737 | StructInfo &Struct = StructInProgress.back(); |
3738 | FieldInfo &Field = Struct.addField(FieldName: Name, FT: FT_INTEGRAL, FieldAlignmentSize: Size); |
3739 | IntFieldInfo &IntInfo = Field.Contents.IntInfo; |
3740 | |
3741 | Field.Type = Size; |
3742 | |
3743 | if (parseScalarInstList(Size, Values&: IntInfo.Values)) |
3744 | return true; |
3745 | |
3746 | Field.SizeOf = Field.Type * IntInfo.Values.size(); |
3747 | Field.LengthOf = IntInfo.Values.size(); |
3748 | const unsigned FieldEnd = Field.Offset + Field.SizeOf; |
3749 | if (!Struct.IsUnion) { |
3750 | Struct.NextOffset = FieldEnd; |
3751 | } |
3752 | Struct.Size = std::max(a: Struct.Size, b: FieldEnd); |
3753 | return false; |
3754 | } |
3755 | |
3756 | /// parseDirectiveValue |
3757 | /// ::= (byte | word | ... ) [ expression (, expression)* ] |
3758 | bool MasmParser::parseDirectiveValue(StringRef IDVal, unsigned Size) { |
3759 | if (StructInProgress.empty()) { |
3760 | // Initialize data value. |
3761 | if (emitIntegralValues(Size)) |
3762 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3763 | } else if (addIntegralField(Name: "" , Size)) { |
3764 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3765 | } |
3766 | |
3767 | return false; |
3768 | } |
3769 | |
3770 | /// parseDirectiveNamedValue |
3771 | /// ::= name (byte | word | ... ) [ expression (, expression)* ] |
3772 | bool MasmParser::parseDirectiveNamedValue(StringRef TypeName, unsigned Size, |
3773 | StringRef Name, SMLoc NameLoc) { |
3774 | if (StructInProgress.empty()) { |
3775 | // Initialize named data value. |
3776 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
3777 | getStreamer().emitLabel(Symbol: Sym); |
3778 | unsigned Count; |
3779 | if (emitIntegralValues(Size, Count: &Count)) |
3780 | return addErrorSuffix(Suffix: " in '" + Twine(TypeName) + "' directive" ); |
3781 | |
3782 | AsmTypeInfo Type; |
3783 | Type.Name = TypeName; |
3784 | Type.Size = Size * Count; |
3785 | Type.ElementSize = Size; |
3786 | Type.Length = Count; |
3787 | KnownType[Name.lower()] = Type; |
3788 | } else if (addIntegralField(Name, Size)) { |
3789 | return addErrorSuffix(Suffix: " in '" + Twine(TypeName) + "' directive" ); |
3790 | } |
3791 | |
3792 | return false; |
3793 | } |
3794 | |
3795 | static bool parseHexOcta(MasmParser &Asm, uint64_t &hi, uint64_t &lo) { |
3796 | if (Asm.getTok().isNot(K: AsmToken::Integer) && |
3797 | Asm.getTok().isNot(K: AsmToken::BigNum)) |
3798 | return Asm.TokError(Msg: "unknown token in expression" ); |
3799 | SMLoc ExprLoc = Asm.getTok().getLoc(); |
3800 | APInt IntValue = Asm.getTok().getAPIntVal(); |
3801 | Asm.Lex(); |
3802 | if (!IntValue.isIntN(N: 128)) |
3803 | return Asm.Error(L: ExprLoc, Msg: "out of range literal value" ); |
3804 | if (!IntValue.isIntN(N: 64)) { |
3805 | hi = IntValue.getHiBits(numBits: IntValue.getBitWidth() - 64).getZExtValue(); |
3806 | lo = IntValue.getLoBits(numBits: 64).getZExtValue(); |
3807 | } else { |
3808 | hi = 0; |
3809 | lo = IntValue.getZExtValue(); |
3810 | } |
3811 | return false; |
3812 | } |
3813 | |
3814 | bool MasmParser::parseRealValue(const fltSemantics &Semantics, APInt &Res) { |
3815 | // We don't truly support arithmetic on floating point expressions, so we |
3816 | // have to manually parse unary prefixes. |
3817 | bool IsNeg = false; |
3818 | SMLoc SignLoc; |
3819 | if (getLexer().is(K: AsmToken::Minus)) { |
3820 | SignLoc = getLexer().getLoc(); |
3821 | Lexer.Lex(); |
3822 | IsNeg = true; |
3823 | } else if (getLexer().is(K: AsmToken::Plus)) { |
3824 | SignLoc = getLexer().getLoc(); |
3825 | Lexer.Lex(); |
3826 | } |
3827 | |
3828 | if (Lexer.is(K: AsmToken::Error)) |
3829 | return TokError(Msg: Lexer.getErr()); |
3830 | if (Lexer.isNot(K: AsmToken::Integer) && Lexer.isNot(K: AsmToken::Real) && |
3831 | Lexer.isNot(K: AsmToken::Identifier)) |
3832 | return TokError(Msg: "unexpected token in directive" ); |
3833 | |
3834 | // Convert to an APFloat. |
3835 | APFloat Value(Semantics); |
3836 | StringRef IDVal = getTok().getString(); |
3837 | if (getLexer().is(K: AsmToken::Identifier)) { |
3838 | if (IDVal.equals_insensitive(RHS: "infinity" ) || IDVal.equals_insensitive(RHS: "inf" )) |
3839 | Value = APFloat::getInf(Sem: Semantics); |
3840 | else if (IDVal.equals_insensitive(RHS: "nan" )) |
3841 | Value = APFloat::getNaN(Sem: Semantics, Negative: false, payload: ~0); |
3842 | else if (IDVal.equals_insensitive(RHS: "?" )) |
3843 | Value = APFloat::getZero(Sem: Semantics); |
3844 | else |
3845 | return TokError(Msg: "invalid floating point literal" ); |
3846 | } else if (IDVal.consume_back(Suffix: "r" ) || IDVal.consume_back(Suffix: "R" )) { |
3847 | // MASM hexadecimal floating-point literal; no APFloat conversion needed. |
3848 | // To match ML64.exe, ignore the initial sign. |
3849 | unsigned SizeInBits = Value.getSizeInBits(Sem: Semantics); |
3850 | if (SizeInBits != (IDVal.size() << 2)) |
3851 | return TokError(Msg: "invalid floating point literal" ); |
3852 | |
3853 | // Consume the numeric token. |
3854 | Lex(); |
3855 | |
3856 | Res = APInt(SizeInBits, IDVal, 16); |
3857 | if (SignLoc.isValid()) |
3858 | return Warning(L: SignLoc, Msg: "MASM-style hex floats ignore explicit sign" ); |
3859 | return false; |
3860 | } else if (errorToBool( |
3861 | Err: Value.convertFromString(IDVal, APFloat::rmNearestTiesToEven) |
3862 | .takeError())) { |
3863 | return TokError(Msg: "invalid floating point literal" ); |
3864 | } |
3865 | if (IsNeg) |
3866 | Value.changeSign(); |
3867 | |
3868 | // Consume the numeric token. |
3869 | Lex(); |
3870 | |
3871 | Res = Value.bitcastToAPInt(); |
3872 | |
3873 | return false; |
3874 | } |
3875 | |
3876 | bool MasmParser::parseRealInstList(const fltSemantics &Semantics, |
3877 | SmallVectorImpl<APInt> &ValuesAsInt, |
3878 | const AsmToken::TokenKind EndToken) { |
3879 | while (getTok().isNot(K: EndToken) || |
3880 | (EndToken == AsmToken::Greater && |
3881 | getTok().isNot(K: AsmToken::GreaterGreater))) { |
3882 | const AsmToken NextTok = peekTok(); |
3883 | if (NextTok.is(K: AsmToken::Identifier) && |
3884 | NextTok.getString().equals_insensitive(RHS: "dup" )) { |
3885 | const MCExpr *Value; |
3886 | if (parseExpression(Res&: Value) || parseToken(T: AsmToken::Identifier)) |
3887 | return true; |
3888 | const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value); |
3889 | if (!MCE) |
3890 | return Error(L: Value->getLoc(), |
3891 | Msg: "cannot repeat value a non-constant number of times" ); |
3892 | const int64_t Repetitions = MCE->getValue(); |
3893 | if (Repetitions < 0) |
3894 | return Error(L: Value->getLoc(), |
3895 | Msg: "cannot repeat value a negative number of times" ); |
3896 | |
3897 | SmallVector<APInt, 1> DuplicatedValues; |
3898 | if (parseToken(T: AsmToken::LParen, |
3899 | Msg: "parentheses required for 'dup' contents" ) || |
3900 | parseRealInstList(Semantics, ValuesAsInt&: DuplicatedValues) || parseRParen()) |
3901 | return true; |
3902 | |
3903 | for (int i = 0; i < Repetitions; ++i) |
3904 | ValuesAsInt.append(in_start: DuplicatedValues.begin(), in_end: DuplicatedValues.end()); |
3905 | } else { |
3906 | APInt AsInt; |
3907 | if (parseRealValue(Semantics, Res&: AsInt)) |
3908 | return true; |
3909 | ValuesAsInt.push_back(Elt: AsInt); |
3910 | } |
3911 | |
3912 | // Continue if we see a comma. (Also, allow line continuation.) |
3913 | if (!parseOptionalToken(T: AsmToken::Comma)) |
3914 | break; |
3915 | parseOptionalToken(T: AsmToken::EndOfStatement); |
3916 | } |
3917 | |
3918 | return false; |
3919 | } |
3920 | |
3921 | // Initialize real data values. |
3922 | bool MasmParser::emitRealValues(const fltSemantics &Semantics, |
3923 | unsigned *Count) { |
3924 | if (checkForValidSection()) |
3925 | return true; |
3926 | |
3927 | SmallVector<APInt, 1> ValuesAsInt; |
3928 | if (parseRealInstList(Semantics, ValuesAsInt)) |
3929 | return true; |
3930 | |
3931 | for (const APInt &AsInt : ValuesAsInt) { |
3932 | getStreamer().emitIntValue(Value: AsInt); |
3933 | } |
3934 | if (Count) |
3935 | *Count = ValuesAsInt.size(); |
3936 | return false; |
3937 | } |
3938 | |
3939 | // Add a real field to the current struct. |
3940 | bool MasmParser::addRealField(StringRef Name, const fltSemantics &Semantics, |
3941 | size_t Size) { |
3942 | StructInfo &Struct = StructInProgress.back(); |
3943 | FieldInfo &Field = Struct.addField(FieldName: Name, FT: FT_REAL, FieldAlignmentSize: Size); |
3944 | RealFieldInfo &RealInfo = Field.Contents.RealInfo; |
3945 | |
3946 | Field.SizeOf = 0; |
3947 | |
3948 | if (parseRealInstList(Semantics, ValuesAsInt&: RealInfo.AsIntValues)) |
3949 | return true; |
3950 | |
3951 | Field.Type = RealInfo.AsIntValues.back().getBitWidth() / 8; |
3952 | Field.LengthOf = RealInfo.AsIntValues.size(); |
3953 | Field.SizeOf = Field.Type * Field.LengthOf; |
3954 | |
3955 | const unsigned FieldEnd = Field.Offset + Field.SizeOf; |
3956 | if (!Struct.IsUnion) { |
3957 | Struct.NextOffset = FieldEnd; |
3958 | } |
3959 | Struct.Size = std::max(a: Struct.Size, b: FieldEnd); |
3960 | return false; |
3961 | } |
3962 | |
3963 | /// parseDirectiveRealValue |
3964 | /// ::= (real4 | real8 | real10) [ expression (, expression)* ] |
3965 | bool MasmParser::parseDirectiveRealValue(StringRef IDVal, |
3966 | const fltSemantics &Semantics, |
3967 | size_t Size) { |
3968 | if (StructInProgress.empty()) { |
3969 | // Initialize data value. |
3970 | if (emitRealValues(Semantics)) |
3971 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3972 | } else if (addRealField(Name: "" , Semantics, Size)) { |
3973 | return addErrorSuffix(Suffix: " in '" + Twine(IDVal) + "' directive" ); |
3974 | } |
3975 | return false; |
3976 | } |
3977 | |
3978 | /// parseDirectiveNamedRealValue |
3979 | /// ::= name (real4 | real8 | real10) [ expression (, expression)* ] |
3980 | bool MasmParser::parseDirectiveNamedRealValue(StringRef TypeName, |
3981 | const fltSemantics &Semantics, |
3982 | unsigned Size, StringRef Name, |
3983 | SMLoc NameLoc) { |
3984 | if (StructInProgress.empty()) { |
3985 | // Initialize named data value. |
3986 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
3987 | getStreamer().emitLabel(Symbol: Sym); |
3988 | unsigned Count; |
3989 | if (emitRealValues(Semantics, Count: &Count)) |
3990 | return addErrorSuffix(Suffix: " in '" + TypeName + "' directive" ); |
3991 | |
3992 | AsmTypeInfo Type; |
3993 | Type.Name = TypeName; |
3994 | Type.Size = Size * Count; |
3995 | Type.ElementSize = Size; |
3996 | Type.Length = Count; |
3997 | KnownType[Name.lower()] = Type; |
3998 | } else if (addRealField(Name, Semantics, Size)) { |
3999 | return addErrorSuffix(Suffix: " in '" + TypeName + "' directive" ); |
4000 | } |
4001 | return false; |
4002 | } |
4003 | |
4004 | bool MasmParser::parseOptionalAngleBracketOpen() { |
4005 | const AsmToken Tok = getTok(); |
4006 | if (parseOptionalToken(T: AsmToken::LessLess)) { |
4007 | AngleBracketDepth++; |
4008 | Lexer.UnLex(Token: AsmToken(AsmToken::Less, Tok.getString().substr(Start: 1))); |
4009 | return true; |
4010 | } else if (parseOptionalToken(T: AsmToken::LessGreater)) { |
4011 | AngleBracketDepth++; |
4012 | Lexer.UnLex(Token: AsmToken(AsmToken::Greater, Tok.getString().substr(Start: 1))); |
4013 | return true; |
4014 | } else if (parseOptionalToken(T: AsmToken::Less)) { |
4015 | AngleBracketDepth++; |
4016 | return true; |
4017 | } |
4018 | |
4019 | return false; |
4020 | } |
4021 | |
4022 | bool MasmParser::parseAngleBracketClose(const Twine &Msg) { |
4023 | const AsmToken Tok = getTok(); |
4024 | if (parseOptionalToken(T: AsmToken::GreaterGreater)) { |
4025 | Lexer.UnLex(Token: AsmToken(AsmToken::Greater, Tok.getString().substr(Start: 1))); |
4026 | } else if (parseToken(T: AsmToken::Greater, Msg)) { |
4027 | return true; |
4028 | } |
4029 | AngleBracketDepth--; |
4030 | return false; |
4031 | } |
4032 | |
4033 | bool MasmParser::parseFieldInitializer(const FieldInfo &Field, |
4034 | const IntFieldInfo &Contents, |
4035 | FieldInitializer &Initializer) { |
4036 | SMLoc Loc = getTok().getLoc(); |
4037 | |
4038 | SmallVector<const MCExpr *, 1> Values; |
4039 | if (parseOptionalToken(T: AsmToken::LCurly)) { |
4040 | if (Field.LengthOf == 1 && Field.Type > 1) |
4041 | return Error(L: Loc, Msg: "Cannot initialize scalar field with array value" ); |
4042 | if (parseScalarInstList(Size: Field.Type, Values, EndToken: AsmToken::RCurly) || |
4043 | parseToken(T: AsmToken::RCurly)) |
4044 | return true; |
4045 | } else if (parseOptionalAngleBracketOpen()) { |
4046 | if (Field.LengthOf == 1 && Field.Type > 1) |
4047 | return Error(L: Loc, Msg: "Cannot initialize scalar field with array value" ); |
4048 | if (parseScalarInstList(Size: Field.Type, Values, EndToken: AsmToken::Greater) || |
4049 | parseAngleBracketClose()) |
4050 | return true; |
4051 | } else if (Field.LengthOf > 1 && Field.Type > 1) { |
4052 | return Error(L: Loc, Msg: "Cannot initialize array field with scalar value" ); |
4053 | } else if (parseScalarInitializer(Size: Field.Type, Values, |
4054 | /*StringPadLength=*/Field.LengthOf)) { |
4055 | return true; |
4056 | } |
4057 | |
4058 | if (Values.size() > Field.LengthOf) { |
4059 | return Error(L: Loc, Msg: "Initializer too long for field; expected at most " + |
4060 | std::to_string(val: Field.LengthOf) + " elements, got " + |
4061 | std::to_string(val: Values.size())); |
4062 | } |
4063 | // Default-initialize all remaining values. |
4064 | Values.append(in_start: Contents.Values.begin() + Values.size(), in_end: Contents.Values.end()); |
4065 | |
4066 | Initializer = FieldInitializer(std::move(Values)); |
4067 | return false; |
4068 | } |
4069 | |
4070 | bool MasmParser::parseFieldInitializer(const FieldInfo &Field, |
4071 | const RealFieldInfo &Contents, |
4072 | FieldInitializer &Initializer) { |
4073 | const fltSemantics *Semantics; |
4074 | switch (Field.Type) { |
4075 | case 4: |
4076 | Semantics = &APFloat::IEEEsingle(); |
4077 | break; |
4078 | case 8: |
4079 | Semantics = &APFloat::IEEEdouble(); |
4080 | break; |
4081 | case 10: |
4082 | Semantics = &APFloat::x87DoubleExtended(); |
4083 | break; |
4084 | default: |
4085 | llvm_unreachable("unknown real field type" ); |
4086 | } |
4087 | |
4088 | SMLoc Loc = getTok().getLoc(); |
4089 | |
4090 | SmallVector<APInt, 1> AsIntValues; |
4091 | if (parseOptionalToken(T: AsmToken::LCurly)) { |
4092 | if (Field.LengthOf == 1) |
4093 | return Error(L: Loc, Msg: "Cannot initialize scalar field with array value" ); |
4094 | if (parseRealInstList(Semantics: *Semantics, ValuesAsInt&: AsIntValues, EndToken: AsmToken::RCurly) || |
4095 | parseToken(T: AsmToken::RCurly)) |
4096 | return true; |
4097 | } else if (parseOptionalAngleBracketOpen()) { |
4098 | if (Field.LengthOf == 1) |
4099 | return Error(L: Loc, Msg: "Cannot initialize scalar field with array value" ); |
4100 | if (parseRealInstList(Semantics: *Semantics, ValuesAsInt&: AsIntValues, EndToken: AsmToken::Greater) || |
4101 | parseAngleBracketClose()) |
4102 | return true; |
4103 | } else if (Field.LengthOf > 1) { |
4104 | return Error(L: Loc, Msg: "Cannot initialize array field with scalar value" ); |
4105 | } else { |
4106 | AsIntValues.emplace_back(); |
4107 | if (parseRealValue(Semantics: *Semantics, Res&: AsIntValues.back())) |
4108 | return true; |
4109 | } |
4110 | |
4111 | if (AsIntValues.size() > Field.LengthOf) { |
4112 | return Error(L: Loc, Msg: "Initializer too long for field; expected at most " + |
4113 | std::to_string(val: Field.LengthOf) + " elements, got " + |
4114 | std::to_string(val: AsIntValues.size())); |
4115 | } |
4116 | // Default-initialize all remaining values. |
4117 | AsIntValues.append(in_start: Contents.AsIntValues.begin() + AsIntValues.size(), |
4118 | in_end: Contents.AsIntValues.end()); |
4119 | |
4120 | Initializer = FieldInitializer(std::move(AsIntValues)); |
4121 | return false; |
4122 | } |
4123 | |
4124 | bool MasmParser::parseFieldInitializer(const FieldInfo &Field, |
4125 | const StructFieldInfo &Contents, |
4126 | FieldInitializer &Initializer) { |
4127 | SMLoc Loc = getTok().getLoc(); |
4128 | |
4129 | std::vector<StructInitializer> Initializers; |
4130 | if (Field.LengthOf > 1) { |
4131 | if (parseOptionalToken(T: AsmToken::LCurly)) { |
4132 | if (parseStructInstList(Structure: Contents.Structure, Initializers, |
4133 | EndToken: AsmToken::RCurly) || |
4134 | parseToken(T: AsmToken::RCurly)) |
4135 | return true; |
4136 | } else if (parseOptionalAngleBracketOpen()) { |
4137 | if (parseStructInstList(Structure: Contents.Structure, Initializers, |
4138 | EndToken: AsmToken::Greater) || |
4139 | parseAngleBracketClose()) |
4140 | return true; |
4141 | } else { |
4142 | return Error(L: Loc, Msg: "Cannot initialize array field with scalar value" ); |
4143 | } |
4144 | } else { |
4145 | Initializers.emplace_back(); |
4146 | if (parseStructInitializer(Structure: Contents.Structure, Initializer&: Initializers.back())) |
4147 | return true; |
4148 | } |
4149 | |
4150 | if (Initializers.size() > Field.LengthOf) { |
4151 | return Error(L: Loc, Msg: "Initializer too long for field; expected at most " + |
4152 | std::to_string(val: Field.LengthOf) + " elements, got " + |
4153 | std::to_string(val: Initializers.size())); |
4154 | } |
4155 | // Default-initialize all remaining values. |
4156 | Initializers.insert(position: Initializers.end(), |
4157 | first: Contents.Initializers.begin() + Initializers.size(), |
4158 | last: Contents.Initializers.end()); |
4159 | |
4160 | Initializer = FieldInitializer(std::move(Initializers), Contents.Structure); |
4161 | return false; |
4162 | } |
4163 | |
4164 | bool MasmParser::parseFieldInitializer(const FieldInfo &Field, |
4165 | FieldInitializer &Initializer) { |
4166 | switch (Field.Contents.FT) { |
4167 | case FT_INTEGRAL: |
4168 | return parseFieldInitializer(Field, Contents: Field.Contents.IntInfo, Initializer); |
4169 | case FT_REAL: |
4170 | return parseFieldInitializer(Field, Contents: Field.Contents.RealInfo, Initializer); |
4171 | case FT_STRUCT: |
4172 | return parseFieldInitializer(Field, Contents: Field.Contents.StructInfo, Initializer); |
4173 | } |
4174 | llvm_unreachable("Unhandled FieldType enum" ); |
4175 | } |
4176 | |
4177 | bool MasmParser::parseStructInitializer(const StructInfo &Structure, |
4178 | StructInitializer &Initializer) { |
4179 | const AsmToken FirstToken = getTok(); |
4180 | |
4181 | std::optional<AsmToken::TokenKind> EndToken; |
4182 | if (parseOptionalToken(T: AsmToken::LCurly)) { |
4183 | EndToken = AsmToken::RCurly; |
4184 | } else if (parseOptionalAngleBracketOpen()) { |
4185 | EndToken = AsmToken::Greater; |
4186 | AngleBracketDepth++; |
4187 | } else if (FirstToken.is(K: AsmToken::Identifier) && |
4188 | FirstToken.getString() == "?" ) { |
4189 | // ? initializer; leave EndToken uninitialized to treat as empty. |
4190 | if (parseToken(T: AsmToken::Identifier)) |
4191 | return true; |
4192 | } else { |
4193 | return Error(L: FirstToken.getLoc(), Msg: "Expected struct initializer" ); |
4194 | } |
4195 | |
4196 | auto &FieldInitializers = Initializer.FieldInitializers; |
4197 | size_t FieldIndex = 0; |
4198 | if (EndToken) { |
4199 | // Initialize all fields with given initializers. |
4200 | while (getTok().isNot(K: *EndToken) && FieldIndex < Structure.Fields.size()) { |
4201 | const FieldInfo &Field = Structure.Fields[FieldIndex++]; |
4202 | if (parseOptionalToken(T: AsmToken::Comma)) { |
4203 | // Empty initializer; use the default and continue. (Also, allow line |
4204 | // continuation.) |
4205 | FieldInitializers.push_back(x: Field.Contents); |
4206 | parseOptionalToken(T: AsmToken::EndOfStatement); |
4207 | continue; |
4208 | } |
4209 | FieldInitializers.emplace_back(args: Field.Contents.FT); |
4210 | if (parseFieldInitializer(Field, Initializer&: FieldInitializers.back())) |
4211 | return true; |
4212 | |
4213 | // Continue if we see a comma. (Also, allow line continuation.) |
4214 | SMLoc CommaLoc = getTok().getLoc(); |
4215 | if (!parseOptionalToken(T: AsmToken::Comma)) |
4216 | break; |
4217 | if (FieldIndex == Structure.Fields.size()) |
4218 | return Error(L: CommaLoc, Msg: "'" + Structure.Name + |
4219 | "' initializer initializes too many fields" ); |
4220 | parseOptionalToken(T: AsmToken::EndOfStatement); |
4221 | } |
4222 | } |
4223 | // Default-initialize all remaining fields. |
4224 | for (const FieldInfo &Field : llvm::drop_begin(RangeOrContainer: Structure.Fields, N: FieldIndex)) |
4225 | FieldInitializers.push_back(x: Field.Contents); |
4226 | |
4227 | if (EndToken) { |
4228 | if (*EndToken == AsmToken::Greater) |
4229 | return parseAngleBracketClose(); |
4230 | |
4231 | return parseToken(T: *EndToken); |
4232 | } |
4233 | |
4234 | return false; |
4235 | } |
4236 | |
4237 | bool MasmParser::parseStructInstList( |
4238 | const StructInfo &Structure, std::vector<StructInitializer> &Initializers, |
4239 | const AsmToken::TokenKind EndToken) { |
4240 | while (getTok().isNot(K: EndToken) || |
4241 | (EndToken == AsmToken::Greater && |
4242 | getTok().isNot(K: AsmToken::GreaterGreater))) { |
4243 | const AsmToken NextTok = peekTok(); |
4244 | if (NextTok.is(K: AsmToken::Identifier) && |
4245 | NextTok.getString().equals_insensitive(RHS: "dup" )) { |
4246 | const MCExpr *Value; |
4247 | if (parseExpression(Res&: Value) || parseToken(T: AsmToken::Identifier)) |
4248 | return true; |
4249 | const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value); |
4250 | if (!MCE) |
4251 | return Error(L: Value->getLoc(), |
4252 | Msg: "cannot repeat value a non-constant number of times" ); |
4253 | const int64_t Repetitions = MCE->getValue(); |
4254 | if (Repetitions < 0) |
4255 | return Error(L: Value->getLoc(), |
4256 | Msg: "cannot repeat value a negative number of times" ); |
4257 | |
4258 | std::vector<StructInitializer> DuplicatedValues; |
4259 | if (parseToken(T: AsmToken::LParen, |
4260 | Msg: "parentheses required for 'dup' contents" ) || |
4261 | parseStructInstList(Structure, Initializers&: DuplicatedValues) || parseRParen()) |
4262 | return true; |
4263 | |
4264 | for (int i = 0; i < Repetitions; ++i) |
4265 | llvm::append_range(C&: Initializers, R&: DuplicatedValues); |
4266 | } else { |
4267 | Initializers.emplace_back(); |
4268 | if (parseStructInitializer(Structure, Initializer&: Initializers.back())) |
4269 | return true; |
4270 | } |
4271 | |
4272 | // Continue if we see a comma. (Also, allow line continuation.) |
4273 | if (!parseOptionalToken(T: AsmToken::Comma)) |
4274 | break; |
4275 | parseOptionalToken(T: AsmToken::EndOfStatement); |
4276 | } |
4277 | |
4278 | return false; |
4279 | } |
4280 | |
4281 | bool MasmParser::emitFieldValue(const FieldInfo &Field, |
4282 | const IntFieldInfo &Contents) { |
4283 | // Default-initialize all values. |
4284 | for (const MCExpr *Value : Contents.Values) { |
4285 | if (emitIntValue(Value, Size: Field.Type)) |
4286 | return true; |
4287 | } |
4288 | return false; |
4289 | } |
4290 | |
4291 | bool MasmParser::emitFieldValue(const FieldInfo &Field, |
4292 | const RealFieldInfo &Contents) { |
4293 | for (const APInt &AsInt : Contents.AsIntValues) { |
4294 | getStreamer().emitIntValue(Value: AsInt.getLimitedValue(), |
4295 | Size: AsInt.getBitWidth() / 8); |
4296 | } |
4297 | return false; |
4298 | } |
4299 | |
4300 | bool MasmParser::emitFieldValue(const FieldInfo &Field, |
4301 | const StructFieldInfo &Contents) { |
4302 | for (const auto &Initializer : Contents.Initializers) { |
4303 | size_t Index = 0, Offset = 0; |
4304 | for (const auto &SubField : Contents.Structure.Fields) { |
4305 | getStreamer().emitZeros(NumBytes: SubField.Offset - Offset); |
4306 | Offset = SubField.Offset + SubField.SizeOf; |
4307 | emitFieldInitializer(Field: SubField, Initializer: Initializer.FieldInitializers[Index++]); |
4308 | } |
4309 | } |
4310 | return false; |
4311 | } |
4312 | |
4313 | bool MasmParser::emitFieldValue(const FieldInfo &Field) { |
4314 | switch (Field.Contents.FT) { |
4315 | case FT_INTEGRAL: |
4316 | return emitFieldValue(Field, Contents: Field.Contents.IntInfo); |
4317 | case FT_REAL: |
4318 | return emitFieldValue(Field, Contents: Field.Contents.RealInfo); |
4319 | case FT_STRUCT: |
4320 | return emitFieldValue(Field, Contents: Field.Contents.StructInfo); |
4321 | } |
4322 | llvm_unreachable("Unhandled FieldType enum" ); |
4323 | } |
4324 | |
4325 | bool MasmParser::emitFieldInitializer(const FieldInfo &Field, |
4326 | const IntFieldInfo &Contents, |
4327 | const IntFieldInfo &Initializer) { |
4328 | for (const auto &Value : Initializer.Values) { |
4329 | if (emitIntValue(Value, Size: Field.Type)) |
4330 | return true; |
4331 | } |
4332 | // Default-initialize all remaining values. |
4333 | for (const auto &Value : |
4334 | llvm::drop_begin(RangeOrContainer: Contents.Values, N: Initializer.Values.size())) { |
4335 | if (emitIntValue(Value, Size: Field.Type)) |
4336 | return true; |
4337 | } |
4338 | return false; |
4339 | } |
4340 | |
4341 | bool MasmParser::emitFieldInitializer(const FieldInfo &Field, |
4342 | const RealFieldInfo &Contents, |
4343 | const RealFieldInfo &Initializer) { |
4344 | for (const auto &AsInt : Initializer.AsIntValues) { |
4345 | getStreamer().emitIntValue(Value: AsInt.getLimitedValue(), |
4346 | Size: AsInt.getBitWidth() / 8); |
4347 | } |
4348 | // Default-initialize all remaining values. |
4349 | for (const auto &AsInt : |
4350 | llvm::drop_begin(RangeOrContainer: Contents.AsIntValues, N: Initializer.AsIntValues.size())) { |
4351 | getStreamer().emitIntValue(Value: AsInt.getLimitedValue(), |
4352 | Size: AsInt.getBitWidth() / 8); |
4353 | } |
4354 | return false; |
4355 | } |
4356 | |
4357 | bool MasmParser::emitFieldInitializer(const FieldInfo &Field, |
4358 | const StructFieldInfo &Contents, |
4359 | const StructFieldInfo &Initializer) { |
4360 | for (const auto &Init : Initializer.Initializers) { |
4361 | if (emitStructInitializer(Structure: Contents.Structure, Initializer: Init)) |
4362 | return true; |
4363 | } |
4364 | // Default-initialize all remaining values. |
4365 | for (const auto &Init : llvm::drop_begin(RangeOrContainer: Contents.Initializers, |
4366 | N: Initializer.Initializers.size())) { |
4367 | if (emitStructInitializer(Structure: Contents.Structure, Initializer: Init)) |
4368 | return true; |
4369 | } |
4370 | return false; |
4371 | } |
4372 | |
4373 | bool MasmParser::emitFieldInitializer(const FieldInfo &Field, |
4374 | const FieldInitializer &Initializer) { |
4375 | switch (Field.Contents.FT) { |
4376 | case FT_INTEGRAL: |
4377 | return emitFieldInitializer(Field, Contents: Field.Contents.IntInfo, |
4378 | Initializer: Initializer.IntInfo); |
4379 | case FT_REAL: |
4380 | return emitFieldInitializer(Field, Contents: Field.Contents.RealInfo, |
4381 | Initializer: Initializer.RealInfo); |
4382 | case FT_STRUCT: |
4383 | return emitFieldInitializer(Field, Contents: Field.Contents.StructInfo, |
4384 | Initializer: Initializer.StructInfo); |
4385 | } |
4386 | llvm_unreachable("Unhandled FieldType enum" ); |
4387 | } |
4388 | |
4389 | bool MasmParser::emitStructInitializer(const StructInfo &Structure, |
4390 | const StructInitializer &Initializer) { |
4391 | if (!Structure.Initializable) |
4392 | return Error(L: getLexer().getLoc(), |
4393 | Msg: "cannot initialize a value of type '" + Structure.Name + |
4394 | "'; 'org' was used in the type's declaration" ); |
4395 | size_t Index = 0, Offset = 0; |
4396 | for (const auto &Init : Initializer.FieldInitializers) { |
4397 | const auto &Field = Structure.Fields[Index++]; |
4398 | getStreamer().emitZeros(NumBytes: Field.Offset - Offset); |
4399 | Offset = Field.Offset + Field.SizeOf; |
4400 | if (emitFieldInitializer(Field, Initializer: Init)) |
4401 | return true; |
4402 | } |
4403 | // Default-initialize all remaining fields. |
4404 | for (const auto &Field : llvm::drop_begin( |
4405 | RangeOrContainer: Structure.Fields, N: Initializer.FieldInitializers.size())) { |
4406 | getStreamer().emitZeros(NumBytes: Field.Offset - Offset); |
4407 | Offset = Field.Offset + Field.SizeOf; |
4408 | if (emitFieldValue(Field)) |
4409 | return true; |
4410 | } |
4411 | // Add final padding. |
4412 | if (Offset != Structure.Size) |
4413 | getStreamer().emitZeros(NumBytes: Structure.Size - Offset); |
4414 | return false; |
4415 | } |
4416 | |
4417 | // Set data values from initializers. |
4418 | bool MasmParser::emitStructValues(const StructInfo &Structure, |
4419 | unsigned *Count) { |
4420 | std::vector<StructInitializer> Initializers; |
4421 | if (parseStructInstList(Structure, Initializers)) |
4422 | return true; |
4423 | |
4424 | for (const auto &Initializer : Initializers) { |
4425 | if (emitStructInitializer(Structure, Initializer)) |
4426 | return true; |
4427 | } |
4428 | |
4429 | if (Count) |
4430 | *Count = Initializers.size(); |
4431 | return false; |
4432 | } |
4433 | |
4434 | // Declare a field in the current struct. |
4435 | bool MasmParser::addStructField(StringRef Name, const StructInfo &Structure) { |
4436 | StructInfo &OwningStruct = StructInProgress.back(); |
4437 | FieldInfo &Field = |
4438 | OwningStruct.addField(FieldName: Name, FT: FT_STRUCT, FieldAlignmentSize: Structure.AlignmentSize); |
4439 | StructFieldInfo &StructInfo = Field.Contents.StructInfo; |
4440 | |
4441 | StructInfo.Structure = Structure; |
4442 | Field.Type = Structure.Size; |
4443 | |
4444 | if (parseStructInstList(Structure, Initializers&: StructInfo.Initializers)) |
4445 | return true; |
4446 | |
4447 | Field.LengthOf = StructInfo.Initializers.size(); |
4448 | Field.SizeOf = Field.Type * Field.LengthOf; |
4449 | |
4450 | const unsigned FieldEnd = Field.Offset + Field.SizeOf; |
4451 | if (!OwningStruct.IsUnion) { |
4452 | OwningStruct.NextOffset = FieldEnd; |
4453 | } |
4454 | OwningStruct.Size = std::max(a: OwningStruct.Size, b: FieldEnd); |
4455 | |
4456 | return false; |
4457 | } |
4458 | |
4459 | /// parseDirectiveStructValue |
4460 | /// ::= struct-id (<struct-initializer> | {struct-initializer}) |
4461 | /// [, (<struct-initializer> | {struct-initializer})]* |
4462 | bool MasmParser::parseDirectiveStructValue(const StructInfo &Structure, |
4463 | StringRef Directive, SMLoc DirLoc) { |
4464 | if (StructInProgress.empty()) { |
4465 | if (emitStructValues(Structure)) |
4466 | return true; |
4467 | } else if (addStructField(Name: "" , Structure)) { |
4468 | return addErrorSuffix(Suffix: " in '" + Twine(Directive) + "' directive" ); |
4469 | } |
4470 | |
4471 | return false; |
4472 | } |
4473 | |
4474 | /// parseDirectiveNamedValue |
4475 | /// ::= name (byte | word | ... ) [ expression (, expression)* ] |
4476 | bool MasmParser::parseDirectiveNamedStructValue(const StructInfo &Structure, |
4477 | StringRef Directive, |
4478 | SMLoc DirLoc, StringRef Name) { |
4479 | if (StructInProgress.empty()) { |
4480 | // Initialize named data value. |
4481 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
4482 | getStreamer().emitLabel(Symbol: Sym); |
4483 | unsigned Count; |
4484 | if (emitStructValues(Structure, Count: &Count)) |
4485 | return true; |
4486 | AsmTypeInfo Type; |
4487 | Type.Name = Structure.Name; |
4488 | Type.Size = Structure.Size * Count; |
4489 | Type.ElementSize = Structure.Size; |
4490 | Type.Length = Count; |
4491 | KnownType[Name.lower()] = Type; |
4492 | } else if (addStructField(Name, Structure)) { |
4493 | return addErrorSuffix(Suffix: " in '" + Twine(Directive) + "' directive" ); |
4494 | } |
4495 | |
4496 | return false; |
4497 | } |
4498 | |
4499 | /// parseDirectiveStruct |
4500 | /// ::= <name> (STRUC | STRUCT | UNION) [fieldAlign] [, NONUNIQUE] |
4501 | /// (dataDir | generalDir | offsetDir | nestedStruct)+ |
4502 | /// <name> ENDS |
4503 | ////// dataDir = data declaration |
4504 | ////// offsetDir = EVEN, ORG, ALIGN |
4505 | bool MasmParser::parseDirectiveStruct(StringRef Directive, |
4506 | DirectiveKind DirKind, StringRef Name, |
4507 | SMLoc NameLoc) { |
4508 | // We ignore NONUNIQUE; we do not support OPTION M510 or OPTION OLDSTRUCTS |
4509 | // anyway, so all field accesses must be qualified. |
4510 | AsmToken NextTok = getTok(); |
4511 | int64_t AlignmentValue = 1; |
4512 | if (NextTok.isNot(K: AsmToken::Comma) && |
4513 | NextTok.isNot(K: AsmToken::EndOfStatement) && |
4514 | parseAbsoluteExpression(Res&: AlignmentValue)) { |
4515 | return addErrorSuffix(Suffix: " in alignment value for '" + Twine(Directive) + |
4516 | "' directive" ); |
4517 | } |
4518 | if (!isPowerOf2_64(Value: AlignmentValue)) { |
4519 | return Error(L: NextTok.getLoc(), Msg: "alignment must be a power of two; was " + |
4520 | std::to_string(val: AlignmentValue)); |
4521 | } |
4522 | |
4523 | StringRef Qualifier; |
4524 | SMLoc QualifierLoc; |
4525 | if (parseOptionalToken(T: AsmToken::Comma)) { |
4526 | QualifierLoc = getTok().getLoc(); |
4527 | if (parseIdentifier(Res&: Qualifier)) |
4528 | return addErrorSuffix(Suffix: " in '" + Twine(Directive) + "' directive" ); |
4529 | if (!Qualifier.equals_insensitive(RHS: "nonunique" )) |
4530 | return Error(L: QualifierLoc, Msg: "Unrecognized qualifier for '" + |
4531 | Twine(Directive) + |
4532 | "' directive; expected none or NONUNIQUE" ); |
4533 | } |
4534 | |
4535 | if (parseEOL()) |
4536 | return addErrorSuffix(Suffix: " in '" + Twine(Directive) + "' directive" ); |
4537 | |
4538 | StructInProgress.emplace_back(Args&: Name, Args: DirKind == DK_UNION, Args&: AlignmentValue); |
4539 | return false; |
4540 | } |
4541 | |
4542 | /// parseDirectiveNestedStruct |
4543 | /// ::= (STRUC | STRUCT | UNION) [name] |
4544 | /// (dataDir | generalDir | offsetDir | nestedStruct)+ |
4545 | /// ENDS |
4546 | bool MasmParser::parseDirectiveNestedStruct(StringRef Directive, |
4547 | DirectiveKind DirKind) { |
4548 | if (StructInProgress.empty()) |
4549 | return TokError(Msg: "missing name in top-level '" + Twine(Directive) + |
4550 | "' directive" ); |
4551 | |
4552 | StringRef Name; |
4553 | if (getTok().is(K: AsmToken::Identifier)) { |
4554 | Name = getTok().getIdentifier(); |
4555 | parseToken(T: AsmToken::Identifier); |
4556 | } |
4557 | if (parseEOL()) |
4558 | return addErrorSuffix(Suffix: " in '" + Twine(Directive) + "' directive" ); |
4559 | |
4560 | // Reserve space to ensure Alignment doesn't get invalidated when |
4561 | // StructInProgress grows. |
4562 | StructInProgress.reserve(N: StructInProgress.size() + 1); |
4563 | StructInProgress.emplace_back(Args&: Name, Args: DirKind == DK_UNION, |
4564 | Args&: StructInProgress.back().Alignment); |
4565 | return false; |
4566 | } |
4567 | |
4568 | bool MasmParser::parseDirectiveEnds(StringRef Name, SMLoc NameLoc) { |
4569 | if (StructInProgress.empty()) |
4570 | return Error(L: NameLoc, Msg: "ENDS directive without matching STRUC/STRUCT/UNION" ); |
4571 | if (StructInProgress.size() > 1) |
4572 | return Error(L: NameLoc, Msg: "unexpected name in nested ENDS directive" ); |
4573 | if (StructInProgress.back().Name.compare_insensitive(RHS: Name)) |
4574 | return Error(L: NameLoc, Msg: "mismatched name in ENDS directive; expected '" + |
4575 | StructInProgress.back().Name + "'" ); |
4576 | StructInfo Structure = StructInProgress.pop_back_val(); |
4577 | // Pad to make the structure's size divisible by the smaller of its alignment |
4578 | // and the size of its largest field. |
4579 | Structure.Size = llvm::alignTo( |
4580 | Value: Structure.Size, Align: std::min(a: Structure.Alignment, b: Structure.AlignmentSize)); |
4581 | Structs[Name.lower()] = Structure; |
4582 | |
4583 | if (parseEOL()) |
4584 | return addErrorSuffix(Suffix: " in ENDS directive" ); |
4585 | |
4586 | return false; |
4587 | } |
4588 | |
4589 | bool MasmParser::parseDirectiveNestedEnds() { |
4590 | if (StructInProgress.empty()) |
4591 | return TokError(Msg: "ENDS directive without matching STRUC/STRUCT/UNION" ); |
4592 | if (StructInProgress.size() == 1) |
4593 | return TokError(Msg: "missing name in top-level ENDS directive" ); |
4594 | |
4595 | if (parseEOL()) |
4596 | return addErrorSuffix(Suffix: " in nested ENDS directive" ); |
4597 | |
4598 | StructInfo Structure = StructInProgress.pop_back_val(); |
4599 | // Pad to make the structure's size divisible by its alignment. |
4600 | Structure.Size = llvm::alignTo(Value: Structure.Size, Align: Structure.Alignment); |
4601 | |
4602 | StructInfo &ParentStruct = StructInProgress.back(); |
4603 | if (Structure.Name.empty()) { |
4604 | // Anonymous substructures' fields are addressed as if they belong to the |
4605 | // parent structure - so we transfer them to the parent here. |
4606 | const size_t OldFields = ParentStruct.Fields.size(); |
4607 | ParentStruct.Fields.insert( |
4608 | position: ParentStruct.Fields.end(), |
4609 | first: std::make_move_iterator(i: Structure.Fields.begin()), |
4610 | last: std::make_move_iterator(i: Structure.Fields.end())); |
4611 | for (const auto &FieldByName : Structure.FieldsByName) { |
4612 | ParentStruct.FieldsByName[FieldByName.getKey()] = |
4613 | FieldByName.getValue() + OldFields; |
4614 | } |
4615 | |
4616 | unsigned FirstFieldOffset = 0; |
4617 | if (!Structure.Fields.empty() && !ParentStruct.IsUnion) { |
4618 | FirstFieldOffset = llvm::alignTo( |
4619 | Value: ParentStruct.NextOffset, |
4620 | Align: std::min(a: ParentStruct.Alignment, b: Structure.AlignmentSize)); |
4621 | } |
4622 | |
4623 | if (ParentStruct.IsUnion) { |
4624 | ParentStruct.Size = std::max(a: ParentStruct.Size, b: Structure.Size); |
4625 | } else { |
4626 | for (auto &Field : llvm::drop_begin(RangeOrContainer&: ParentStruct.Fields, N: OldFields)) |
4627 | Field.Offset += FirstFieldOffset; |
4628 | |
4629 | const unsigned StructureEnd = FirstFieldOffset + Structure.Size; |
4630 | if (!ParentStruct.IsUnion) { |
4631 | ParentStruct.NextOffset = StructureEnd; |
4632 | } |
4633 | ParentStruct.Size = std::max(a: ParentStruct.Size, b: StructureEnd); |
4634 | } |
4635 | } else { |
4636 | FieldInfo &Field = ParentStruct.addField(FieldName: Structure.Name, FT: FT_STRUCT, |
4637 | FieldAlignmentSize: Structure.AlignmentSize); |
4638 | StructFieldInfo &StructInfo = Field.Contents.StructInfo; |
4639 | Field.Type = Structure.Size; |
4640 | Field.LengthOf = 1; |
4641 | Field.SizeOf = Structure.Size; |
4642 | |
4643 | const unsigned StructureEnd = Field.Offset + Field.SizeOf; |
4644 | if (!ParentStruct.IsUnion) { |
4645 | ParentStruct.NextOffset = StructureEnd; |
4646 | } |
4647 | ParentStruct.Size = std::max(a: ParentStruct.Size, b: StructureEnd); |
4648 | |
4649 | StructInfo.Structure = Structure; |
4650 | StructInfo.Initializers.emplace_back(); |
4651 | auto &FieldInitializers = StructInfo.Initializers.back().FieldInitializers; |
4652 | for (const auto &SubField : Structure.Fields) { |
4653 | FieldInitializers.push_back(x: SubField.Contents); |
4654 | } |
4655 | } |
4656 | |
4657 | return false; |
4658 | } |
4659 | |
4660 | /// parseDirectiveOrg |
4661 | /// ::= org expression |
4662 | bool MasmParser::parseDirectiveOrg() { |
4663 | const MCExpr *Offset; |
4664 | SMLoc OffsetLoc = Lexer.getLoc(); |
4665 | if (checkForValidSection() || parseExpression(Res&: Offset)) |
4666 | return true; |
4667 | if (parseEOL()) |
4668 | return addErrorSuffix(Suffix: " in 'org' directive" ); |
4669 | |
4670 | if (StructInProgress.empty()) { |
4671 | // Not in a struct; change the offset for the next instruction or data |
4672 | if (checkForValidSection()) |
4673 | return addErrorSuffix(Suffix: " in 'org' directive" ); |
4674 | |
4675 | getStreamer().emitValueToOffset(Offset, Value: 0, Loc: OffsetLoc); |
4676 | } else { |
4677 | // Offset the next field of this struct |
4678 | StructInfo &Structure = StructInProgress.back(); |
4679 | int64_t OffsetRes; |
4680 | if (!Offset->evaluateAsAbsolute(Res&: OffsetRes, Asm: getStreamer().getAssemblerPtr())) |
4681 | return Error(L: OffsetLoc, |
4682 | Msg: "expected absolute expression in 'org' directive" ); |
4683 | if (OffsetRes < 0) |
4684 | return Error( |
4685 | L: OffsetLoc, |
4686 | Msg: "expected non-negative value in struct's 'org' directive; was " + |
4687 | std::to_string(val: OffsetRes)); |
4688 | Structure.NextOffset = static_cast<unsigned>(OffsetRes); |
4689 | |
4690 | // ORG-affected structures cannot be initialized |
4691 | Structure.Initializable = false; |
4692 | } |
4693 | |
4694 | return false; |
4695 | } |
4696 | |
4697 | bool MasmParser::emitAlignTo(int64_t Alignment) { |
4698 | if (StructInProgress.empty()) { |
4699 | // Not in a struct; align the next instruction or data |
4700 | if (checkForValidSection()) |
4701 | return true; |
4702 | |
4703 | // Check whether we should use optimal code alignment for this align |
4704 | // directive. |
4705 | const MCSection *Section = getStreamer().getCurrentSectionOnly(); |
4706 | assert(Section && "must have section to emit alignment" ); |
4707 | if (Section->useCodeAlign()) { |
4708 | getStreamer().emitCodeAlignment(Alignment: Align(Alignment), |
4709 | STI: &getTargetParser().getSTI(), |
4710 | /*MaxBytesToEmit=*/0); |
4711 | } else { |
4712 | // FIXME: Target specific behavior about how the "extra" bytes are filled. |
4713 | getStreamer().emitValueToAlignment(Alignment: Align(Alignment), /*Value=*/0, |
4714 | /*ValueSize=*/1, |
4715 | /*MaxBytesToEmit=*/0); |
4716 | } |
4717 | } else { |
4718 | // Align the next field of this struct |
4719 | StructInfo &Structure = StructInProgress.back(); |
4720 | Structure.NextOffset = llvm::alignTo(Value: Structure.NextOffset, Align: Alignment); |
4721 | } |
4722 | |
4723 | return false; |
4724 | } |
4725 | |
4726 | /// parseDirectiveAlign |
4727 | /// ::= align expression |
4728 | bool MasmParser::parseDirectiveAlign() { |
4729 | SMLoc AlignmentLoc = getLexer().getLoc(); |
4730 | int64_t Alignment; |
4731 | |
4732 | // Ignore empty 'align' directives. |
4733 | if (getTok().is(K: AsmToken::EndOfStatement)) { |
4734 | return Warning(L: AlignmentLoc, |
4735 | Msg: "align directive with no operand is ignored" ) && |
4736 | parseEOL(); |
4737 | } |
4738 | if (parseAbsoluteExpression(Res&: Alignment) || parseEOL()) |
4739 | return addErrorSuffix(Suffix: " in align directive" ); |
4740 | |
4741 | // Always emit an alignment here even if we throw an error. |
4742 | bool ReturnVal = false; |
4743 | |
4744 | // Reject alignments that aren't either a power of two or zero, for ML.exe |
4745 | // compatibility. Alignment of zero is silently rounded up to one. |
4746 | if (Alignment == 0) |
4747 | Alignment = 1; |
4748 | if (!isPowerOf2_64(Value: Alignment)) |
4749 | ReturnVal |= Error(L: AlignmentLoc, Msg: "alignment must be a power of 2; was " + |
4750 | std::to_string(val: Alignment)); |
4751 | |
4752 | if (emitAlignTo(Alignment)) |
4753 | ReturnVal |= addErrorSuffix(Suffix: " in align directive" ); |
4754 | |
4755 | return ReturnVal; |
4756 | } |
4757 | |
4758 | /// parseDirectiveEven |
4759 | /// ::= even |
4760 | bool MasmParser::parseDirectiveEven() { |
4761 | if (parseEOL() || emitAlignTo(Alignment: 2)) |
4762 | return addErrorSuffix(Suffix: " in even directive" ); |
4763 | |
4764 | return false; |
4765 | } |
4766 | |
4767 | /// parseDirectiveFile |
4768 | /// ::= .file filename |
4769 | /// ::= .file number [directory] filename [md5 checksum] [source source-text] |
4770 | bool MasmParser::parseDirectiveFile(SMLoc DirectiveLoc) { |
4771 | // FIXME: I'm not sure what this is. |
4772 | int64_t FileNumber = -1; |
4773 | if (getLexer().is(K: AsmToken::Integer)) { |
4774 | FileNumber = getTok().getIntVal(); |
4775 | Lex(); |
4776 | |
4777 | if (FileNumber < 0) |
4778 | return TokError(Msg: "negative file number" ); |
4779 | } |
4780 | |
4781 | std::string Path; |
4782 | |
4783 | // Usually the directory and filename together, otherwise just the directory. |
4784 | // Allow the strings to have escaped octal character sequence. |
4785 | if (check(P: getTok().isNot(K: AsmToken::String), |
4786 | Msg: "unexpected token in '.file' directive" ) || |
4787 | parseEscapedString(Data&: Path)) |
4788 | return true; |
4789 | |
4790 | StringRef Directory; |
4791 | StringRef Filename; |
4792 | std::string FilenameData; |
4793 | if (getLexer().is(K: AsmToken::String)) { |
4794 | if (check(P: FileNumber == -1, |
4795 | Msg: "explicit path specified, but no file number" ) || |
4796 | parseEscapedString(Data&: FilenameData)) |
4797 | return true; |
4798 | Filename = FilenameData; |
4799 | Directory = Path; |
4800 | } else { |
4801 | Filename = Path; |
4802 | } |
4803 | |
4804 | uint64_t MD5Hi, MD5Lo; |
4805 | bool HasMD5 = false; |
4806 | |
4807 | std::optional<StringRef> Source; |
4808 | bool HasSource = false; |
4809 | std::string SourceString; |
4810 | |
4811 | while (!parseOptionalToken(T: AsmToken::EndOfStatement)) { |
4812 | StringRef Keyword; |
4813 | if (check(P: getTok().isNot(K: AsmToken::Identifier), |
4814 | Msg: "unexpected token in '.file' directive" ) || |
4815 | parseIdentifier(Res&: Keyword)) |
4816 | return true; |
4817 | if (Keyword == "md5" ) { |
4818 | HasMD5 = true; |
4819 | if (check(P: FileNumber == -1, |
4820 | Msg: "MD5 checksum specified, but no file number" ) || |
4821 | parseHexOcta(Asm&: *this, hi&: MD5Hi, lo&: MD5Lo)) |
4822 | return true; |
4823 | } else if (Keyword == "source" ) { |
4824 | HasSource = true; |
4825 | if (check(P: FileNumber == -1, |
4826 | Msg: "source specified, but no file number" ) || |
4827 | check(P: getTok().isNot(K: AsmToken::String), |
4828 | Msg: "unexpected token in '.file' directive" ) || |
4829 | parseEscapedString(Data&: SourceString)) |
4830 | return true; |
4831 | } else { |
4832 | return TokError(Msg: "unexpected token in '.file' directive" ); |
4833 | } |
4834 | } |
4835 | |
4836 | if (FileNumber == -1) { |
4837 | // Ignore the directive if there is no number and the target doesn't support |
4838 | // numberless .file directives. This allows some portability of assembler |
4839 | // between different object file formats. |
4840 | if (getContext().getAsmInfo()->hasSingleParameterDotFile()) |
4841 | getStreamer().emitFileDirective(Filename); |
4842 | } else { |
4843 | // In case there is a -g option as well as debug info from directive .file, |
4844 | // we turn off the -g option, directly use the existing debug info instead. |
4845 | // Throw away any implicit file table for the assembler source. |
4846 | if (Ctx.getGenDwarfForAssembly()) { |
4847 | Ctx.getMCDwarfLineTable(CUID: 0).resetFileTable(); |
4848 | Ctx.setGenDwarfForAssembly(false); |
4849 | } |
4850 | |
4851 | std::optional<MD5::MD5Result> CKMem; |
4852 | if (HasMD5) { |
4853 | MD5::MD5Result Sum; |
4854 | for (unsigned i = 0; i != 8; ++i) { |
4855 | Sum[i] = uint8_t(MD5Hi >> ((7 - i) * 8)); |
4856 | Sum[i + 8] = uint8_t(MD5Lo >> ((7 - i) * 8)); |
4857 | } |
4858 | CKMem = Sum; |
4859 | } |
4860 | if (HasSource) { |
4861 | char *SourceBuf = static_cast<char *>(Ctx.allocate(Size: SourceString.size())); |
4862 | memcpy(dest: SourceBuf, src: SourceString.data(), n: SourceString.size()); |
4863 | Source = StringRef(SourceBuf, SourceString.size()); |
4864 | } |
4865 | if (FileNumber == 0) { |
4866 | if (Ctx.getDwarfVersion() < 5) |
4867 | return Warning(L: DirectiveLoc, Msg: "file 0 not supported prior to DWARF-5" ); |
4868 | getStreamer().emitDwarfFile0Directive(Directory, Filename, Checksum: CKMem, Source); |
4869 | } else { |
4870 | Expected<unsigned> FileNumOrErr = getStreamer().tryEmitDwarfFileDirective( |
4871 | FileNo: FileNumber, Directory, Filename, Checksum: CKMem, Source); |
4872 | if (!FileNumOrErr) |
4873 | return Error(L: DirectiveLoc, Msg: toString(E: FileNumOrErr.takeError())); |
4874 | } |
4875 | // Alert the user if there are some .file directives with MD5 and some not. |
4876 | // But only do that once. |
4877 | if (!ReportedInconsistentMD5 && !Ctx.isDwarfMD5UsageConsistent(CUID: 0)) { |
4878 | ReportedInconsistentMD5 = true; |
4879 | return Warning(L: DirectiveLoc, Msg: "inconsistent use of MD5 checksums" ); |
4880 | } |
4881 | } |
4882 | |
4883 | return false; |
4884 | } |
4885 | |
4886 | /// parseDirectiveLine |
4887 | /// ::= .line [number] |
4888 | bool MasmParser::parseDirectiveLine() { |
4889 | int64_t LineNumber; |
4890 | if (getLexer().is(K: AsmToken::Integer)) { |
4891 | if (parseIntToken(V&: LineNumber, ErrMsg: "unexpected token in '.line' directive" )) |
4892 | return true; |
4893 | (void)LineNumber; |
4894 | // FIXME: Do something with the .line. |
4895 | } |
4896 | if (parseEOL()) |
4897 | return true; |
4898 | |
4899 | return false; |
4900 | } |
4901 | |
4902 | /// parseDirectiveLoc |
4903 | /// ::= .loc FileNumber [LineNumber] [ColumnPos] [basic_block] [prologue_end] |
4904 | /// [epilogue_begin] [is_stmt VALUE] [isa VALUE] |
4905 | /// The first number is a file number, must have been previously assigned with |
4906 | /// a .file directive, the second number is the line number and optionally the |
4907 | /// third number is a column position (zero if not specified). The remaining |
4908 | /// optional items are .loc sub-directives. |
4909 | bool MasmParser::parseDirectiveLoc() { |
4910 | int64_t FileNumber = 0, LineNumber = 0; |
4911 | SMLoc Loc = getTok().getLoc(); |
4912 | if (parseIntToken(V&: FileNumber, ErrMsg: "unexpected token in '.loc' directive" ) || |
4913 | check(P: FileNumber < 1 && Ctx.getDwarfVersion() < 5, Loc, |
4914 | Msg: "file number less than one in '.loc' directive" ) || |
4915 | check(P: !getContext().isValidDwarfFileNumber(FileNumber), Loc, |
4916 | Msg: "unassigned file number in '.loc' directive" )) |
4917 | return true; |
4918 | |
4919 | // optional |
4920 | if (getLexer().is(K: AsmToken::Integer)) { |
4921 | LineNumber = getTok().getIntVal(); |
4922 | if (LineNumber < 0) |
4923 | return TokError(Msg: "line number less than zero in '.loc' directive" ); |
4924 | Lex(); |
4925 | } |
4926 | |
4927 | int64_t ColumnPos = 0; |
4928 | if (getLexer().is(K: AsmToken::Integer)) { |
4929 | ColumnPos = getTok().getIntVal(); |
4930 | if (ColumnPos < 0) |
4931 | return TokError(Msg: "column position less than zero in '.loc' directive" ); |
4932 | Lex(); |
4933 | } |
4934 | |
4935 | auto PrevFlags = getContext().getCurrentDwarfLoc().getFlags(); |
4936 | unsigned Flags = PrevFlags & DWARF2_FLAG_IS_STMT; |
4937 | unsigned Isa = 0; |
4938 | int64_t Discriminator = 0; |
4939 | |
4940 | auto parseLocOp = [&]() -> bool { |
4941 | StringRef Name; |
4942 | SMLoc Loc = getTok().getLoc(); |
4943 | if (parseIdentifier(Res&: Name)) |
4944 | return TokError(Msg: "unexpected token in '.loc' directive" ); |
4945 | |
4946 | if (Name == "basic_block" ) |
4947 | Flags |= DWARF2_FLAG_BASIC_BLOCK; |
4948 | else if (Name == "prologue_end" ) |
4949 | Flags |= DWARF2_FLAG_PROLOGUE_END; |
4950 | else if (Name == "epilogue_begin" ) |
4951 | Flags |= DWARF2_FLAG_EPILOGUE_BEGIN; |
4952 | else if (Name == "is_stmt" ) { |
4953 | Loc = getTok().getLoc(); |
4954 | const MCExpr *Value; |
4955 | if (parseExpression(Res&: Value)) |
4956 | return true; |
4957 | // The expression must be the constant 0 or 1. |
4958 | if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value)) { |
4959 | int Value = MCE->getValue(); |
4960 | if (Value == 0) |
4961 | Flags &= ~DWARF2_FLAG_IS_STMT; |
4962 | else if (Value == 1) |
4963 | Flags |= DWARF2_FLAG_IS_STMT; |
4964 | else |
4965 | return Error(L: Loc, Msg: "is_stmt value not 0 or 1" ); |
4966 | } else { |
4967 | return Error(L: Loc, Msg: "is_stmt value not the constant value of 0 or 1" ); |
4968 | } |
4969 | } else if (Name == "isa" ) { |
4970 | Loc = getTok().getLoc(); |
4971 | const MCExpr *Value; |
4972 | if (parseExpression(Res&: Value)) |
4973 | return true; |
4974 | // The expression must be a constant greater or equal to 0. |
4975 | if (const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value)) { |
4976 | int Value = MCE->getValue(); |
4977 | if (Value < 0) |
4978 | return Error(L: Loc, Msg: "isa number less than zero" ); |
4979 | Isa = Value; |
4980 | } else { |
4981 | return Error(L: Loc, Msg: "isa number not a constant value" ); |
4982 | } |
4983 | } else if (Name == "discriminator" ) { |
4984 | if (parseAbsoluteExpression(Res&: Discriminator)) |
4985 | return true; |
4986 | } else { |
4987 | return Error(L: Loc, Msg: "unknown sub-directive in '.loc' directive" ); |
4988 | } |
4989 | return false; |
4990 | }; |
4991 | |
4992 | if (parseMany(parseOne: parseLocOp, hasComma: false /*hasComma*/)) |
4993 | return true; |
4994 | |
4995 | getStreamer().emitDwarfLocDirective(FileNo: FileNumber, Line: LineNumber, Column: ColumnPos, Flags, |
4996 | Isa, Discriminator, FileName: StringRef()); |
4997 | |
4998 | return false; |
4999 | } |
5000 | |
5001 | /// parseDirectiveStabs |
5002 | /// ::= .stabs string, number, number, number |
5003 | bool MasmParser::parseDirectiveStabs() { |
5004 | return TokError(Msg: "unsupported directive '.stabs'" ); |
5005 | } |
5006 | |
5007 | /// parseDirectiveCVFile |
5008 | /// ::= .cv_file number filename [checksum] [checksumkind] |
5009 | bool MasmParser::parseDirectiveCVFile() { |
5010 | SMLoc FileNumberLoc = getTok().getLoc(); |
5011 | int64_t FileNumber; |
5012 | std::string Filename; |
5013 | std::string Checksum; |
5014 | int64_t ChecksumKind = 0; |
5015 | |
5016 | if (parseIntToken(V&: FileNumber, |
5017 | ErrMsg: "expected file number in '.cv_file' directive" ) || |
5018 | check(P: FileNumber < 1, Loc: FileNumberLoc, Msg: "file number less than one" ) || |
5019 | check(P: getTok().isNot(K: AsmToken::String), |
5020 | Msg: "unexpected token in '.cv_file' directive" ) || |
5021 | parseEscapedString(Data&: Filename)) |
5022 | return true; |
5023 | if (!parseOptionalToken(T: AsmToken::EndOfStatement)) { |
5024 | if (check(P: getTok().isNot(K: AsmToken::String), |
5025 | Msg: "unexpected token in '.cv_file' directive" ) || |
5026 | parseEscapedString(Data&: Checksum) || |
5027 | parseIntToken(V&: ChecksumKind, |
5028 | ErrMsg: "expected checksum kind in '.cv_file' directive" ) || |
5029 | parseEOL()) |
5030 | return true; |
5031 | } |
5032 | |
5033 | Checksum = fromHex(Input: Checksum); |
5034 | void *CKMem = Ctx.allocate(Size: Checksum.size(), Align: 1); |
5035 | memcpy(dest: CKMem, src: Checksum.data(), n: Checksum.size()); |
5036 | ArrayRef<uint8_t> ChecksumAsBytes(reinterpret_cast<const uint8_t *>(CKMem), |
5037 | Checksum.size()); |
5038 | |
5039 | if (!getStreamer().emitCVFileDirective(FileNo: FileNumber, Filename, Checksum: ChecksumAsBytes, |
5040 | ChecksumKind: static_cast<uint8_t>(ChecksumKind))) |
5041 | return Error(L: FileNumberLoc, Msg: "file number already allocated" ); |
5042 | |
5043 | return false; |
5044 | } |
5045 | |
5046 | bool MasmParser::parseCVFunctionId(int64_t &FunctionId, |
5047 | StringRef DirectiveName) { |
5048 | SMLoc Loc; |
5049 | return parseTokenLoc(Loc) || |
5050 | parseIntToken(V&: FunctionId, ErrMsg: "expected function id in '" + DirectiveName + |
5051 | "' directive" ) || |
5052 | check(P: FunctionId < 0 || FunctionId >= UINT_MAX, Loc, |
5053 | Msg: "expected function id within range [0, UINT_MAX)" ); |
5054 | } |
5055 | |
5056 | bool MasmParser::parseCVFileId(int64_t &FileNumber, StringRef DirectiveName) { |
5057 | SMLoc Loc; |
5058 | return parseTokenLoc(Loc) || |
5059 | parseIntToken(V&: FileNumber, ErrMsg: "expected integer in '" + DirectiveName + |
5060 | "' directive" ) || |
5061 | check(P: FileNumber < 1, Loc, Msg: "file number less than one in '" + |
5062 | DirectiveName + "' directive" ) || |
5063 | check(P: !getCVContext().isValidFileNumber(FileNumber), Loc, |
5064 | Msg: "unassigned file number in '" + DirectiveName + "' directive" ); |
5065 | } |
5066 | |
5067 | /// parseDirectiveCVFuncId |
5068 | /// ::= .cv_func_id FunctionId |
5069 | /// |
5070 | /// Introduces a function ID that can be used with .cv_loc. |
5071 | bool MasmParser::parseDirectiveCVFuncId() { |
5072 | SMLoc FunctionIdLoc = getTok().getLoc(); |
5073 | int64_t FunctionId; |
5074 | |
5075 | if (parseCVFunctionId(FunctionId, DirectiveName: ".cv_func_id" ) || parseEOL()) |
5076 | return true; |
5077 | |
5078 | if (!getStreamer().emitCVFuncIdDirective(FunctionId)) |
5079 | return Error(L: FunctionIdLoc, Msg: "function id already allocated" ); |
5080 | |
5081 | return false; |
5082 | } |
5083 | |
5084 | /// parseDirectiveCVInlineSiteId |
5085 | /// ::= .cv_inline_site_id FunctionId |
5086 | /// "within" IAFunc |
5087 | /// "inlined_at" IAFile IALine [IACol] |
5088 | /// |
5089 | /// Introduces a function ID that can be used with .cv_loc. Includes "inlined |
5090 | /// at" source location information for use in the line table of the caller, |
5091 | /// whether the caller is a real function or another inlined call site. |
5092 | bool MasmParser::parseDirectiveCVInlineSiteId() { |
5093 | SMLoc FunctionIdLoc = getTok().getLoc(); |
5094 | int64_t FunctionId; |
5095 | int64_t IAFunc; |
5096 | int64_t IAFile; |
5097 | int64_t IALine; |
5098 | int64_t IACol = 0; |
5099 | |
5100 | // FunctionId |
5101 | if (parseCVFunctionId(FunctionId, DirectiveName: ".cv_inline_site_id" )) |
5102 | return true; |
5103 | |
5104 | // "within" |
5105 | if (check(P: (getLexer().isNot(K: AsmToken::Identifier) || |
5106 | getTok().getIdentifier() != "within" ), |
5107 | Msg: "expected 'within' identifier in '.cv_inline_site_id' directive" )) |
5108 | return true; |
5109 | Lex(); |
5110 | |
5111 | // IAFunc |
5112 | if (parseCVFunctionId(FunctionId&: IAFunc, DirectiveName: ".cv_inline_site_id" )) |
5113 | return true; |
5114 | |
5115 | // "inlined_at" |
5116 | if (check(P: (getLexer().isNot(K: AsmToken::Identifier) || |
5117 | getTok().getIdentifier() != "inlined_at" ), |
5118 | Msg: "expected 'inlined_at' identifier in '.cv_inline_site_id' " |
5119 | "directive" ) ) |
5120 | return true; |
5121 | Lex(); |
5122 | |
5123 | // IAFile IALine |
5124 | if (parseCVFileId(FileNumber&: IAFile, DirectiveName: ".cv_inline_site_id" ) || |
5125 | parseIntToken(V&: IALine, ErrMsg: "expected line number after 'inlined_at'" )) |
5126 | return true; |
5127 | |
5128 | // [IACol] |
5129 | if (getLexer().is(K: AsmToken::Integer)) { |
5130 | IACol = getTok().getIntVal(); |
5131 | Lex(); |
5132 | } |
5133 | |
5134 | if (parseEOL()) |
5135 | return true; |
5136 | |
5137 | if (!getStreamer().emitCVInlineSiteIdDirective(FunctionId, IAFunc, IAFile, |
5138 | IALine, IACol, Loc: FunctionIdLoc)) |
5139 | return Error(L: FunctionIdLoc, Msg: "function id already allocated" ); |
5140 | |
5141 | return false; |
5142 | } |
5143 | |
5144 | /// parseDirectiveCVLoc |
5145 | /// ::= .cv_loc FunctionId FileNumber [LineNumber] [ColumnPos] [prologue_end] |
5146 | /// [is_stmt VALUE] |
5147 | /// The first number is a file number, must have been previously assigned with |
5148 | /// a .file directive, the second number is the line number and optionally the |
5149 | /// third number is a column position (zero if not specified). The remaining |
5150 | /// optional items are .loc sub-directives. |
5151 | bool MasmParser::parseDirectiveCVLoc() { |
5152 | SMLoc DirectiveLoc = getTok().getLoc(); |
5153 | int64_t FunctionId, FileNumber; |
5154 | if (parseCVFunctionId(FunctionId, DirectiveName: ".cv_loc" ) || |
5155 | parseCVFileId(FileNumber, DirectiveName: ".cv_loc" )) |
5156 | return true; |
5157 | |
5158 | int64_t LineNumber = 0; |
5159 | if (getLexer().is(K: AsmToken::Integer)) { |
5160 | LineNumber = getTok().getIntVal(); |
5161 | if (LineNumber < 0) |
5162 | return TokError(Msg: "line number less than zero in '.cv_loc' directive" ); |
5163 | Lex(); |
5164 | } |
5165 | |
5166 | int64_t ColumnPos = 0; |
5167 | if (getLexer().is(K: AsmToken::Integer)) { |
5168 | ColumnPos = getTok().getIntVal(); |
5169 | if (ColumnPos < 0) |
5170 | return TokError(Msg: "column position less than zero in '.cv_loc' directive" ); |
5171 | Lex(); |
5172 | } |
5173 | |
5174 | bool PrologueEnd = false; |
5175 | uint64_t IsStmt = 0; |
5176 | |
5177 | auto parseOp = [&]() -> bool { |
5178 | StringRef Name; |
5179 | SMLoc Loc = getTok().getLoc(); |
5180 | if (parseIdentifier(Res&: Name)) |
5181 | return TokError(Msg: "unexpected token in '.cv_loc' directive" ); |
5182 | if (Name == "prologue_end" ) |
5183 | PrologueEnd = true; |
5184 | else if (Name == "is_stmt" ) { |
5185 | Loc = getTok().getLoc(); |
5186 | const MCExpr *Value; |
5187 | if (parseExpression(Res&: Value)) |
5188 | return true; |
5189 | // The expression must be the constant 0 or 1. |
5190 | IsStmt = ~0ULL; |
5191 | if (const auto *MCE = dyn_cast<MCConstantExpr>(Val: Value)) |
5192 | IsStmt = MCE->getValue(); |
5193 | |
5194 | if (IsStmt > 1) |
5195 | return Error(L: Loc, Msg: "is_stmt value not 0 or 1" ); |
5196 | } else { |
5197 | return Error(L: Loc, Msg: "unknown sub-directive in '.cv_loc' directive" ); |
5198 | } |
5199 | return false; |
5200 | }; |
5201 | |
5202 | if (parseMany(parseOne: parseOp, hasComma: false /*hasComma*/)) |
5203 | return true; |
5204 | |
5205 | getStreamer().emitCVLocDirective(FunctionId, FileNo: FileNumber, Line: LineNumber, |
5206 | Column: ColumnPos, PrologueEnd, IsStmt, FileName: StringRef(), |
5207 | Loc: DirectiveLoc); |
5208 | return false; |
5209 | } |
5210 | |
5211 | /// parseDirectiveCVLinetable |
5212 | /// ::= .cv_linetable FunctionId, FnStart, FnEnd |
5213 | bool MasmParser::parseDirectiveCVLinetable() { |
5214 | int64_t FunctionId; |
5215 | StringRef FnStartName, FnEndName; |
5216 | SMLoc Loc = getTok().getLoc(); |
5217 | if (parseCVFunctionId(FunctionId, DirectiveName: ".cv_linetable" ) || |
5218 | parseToken(T: AsmToken::Comma, |
5219 | Msg: "unexpected token in '.cv_linetable' directive" ) || |
5220 | parseTokenLoc(Loc) || check(P: parseIdentifier(Res&: FnStartName), Loc, |
5221 | Msg: "expected identifier in directive" ) || |
5222 | parseToken(T: AsmToken::Comma, |
5223 | Msg: "unexpected token in '.cv_linetable' directive" ) || |
5224 | parseTokenLoc(Loc) || check(P: parseIdentifier(Res&: FnEndName), Loc, |
5225 | Msg: "expected identifier in directive" )) |
5226 | return true; |
5227 | |
5228 | MCSymbol *FnStartSym = getContext().getOrCreateSymbol(Name: FnStartName); |
5229 | MCSymbol *FnEndSym = getContext().getOrCreateSymbol(Name: FnEndName); |
5230 | |
5231 | getStreamer().emitCVLinetableDirective(FunctionId, FnStart: FnStartSym, FnEnd: FnEndSym); |
5232 | return false; |
5233 | } |
5234 | |
5235 | /// parseDirectiveCVInlineLinetable |
5236 | /// ::= .cv_inline_linetable PrimaryFunctionId FileId LineNum FnStart FnEnd |
5237 | bool MasmParser::parseDirectiveCVInlineLinetable() { |
5238 | int64_t PrimaryFunctionId, SourceFileId, SourceLineNum; |
5239 | StringRef FnStartName, FnEndName; |
5240 | SMLoc Loc = getTok().getLoc(); |
5241 | if (parseCVFunctionId(FunctionId&: PrimaryFunctionId, DirectiveName: ".cv_inline_linetable" ) || |
5242 | parseTokenLoc(Loc) || |
5243 | parseIntToken( |
5244 | V&: SourceFileId, |
5245 | ErrMsg: "expected SourceField in '.cv_inline_linetable' directive" ) || |
5246 | check(P: SourceFileId <= 0, Loc, |
5247 | Msg: "File id less than zero in '.cv_inline_linetable' directive" ) || |
5248 | parseTokenLoc(Loc) || |
5249 | parseIntToken( |
5250 | V&: SourceLineNum, |
5251 | ErrMsg: "expected SourceLineNum in '.cv_inline_linetable' directive" ) || |
5252 | check(P: SourceLineNum < 0, Loc, |
5253 | Msg: "Line number less than zero in '.cv_inline_linetable' directive" ) || |
5254 | parseTokenLoc(Loc) || check(P: parseIdentifier(Res&: FnStartName), Loc, |
5255 | Msg: "expected identifier in directive" ) || |
5256 | parseTokenLoc(Loc) || check(P: parseIdentifier(Res&: FnEndName), Loc, |
5257 | Msg: "expected identifier in directive" )) |
5258 | return true; |
5259 | |
5260 | if (parseEOL()) |
5261 | return true; |
5262 | |
5263 | MCSymbol *FnStartSym = getContext().getOrCreateSymbol(Name: FnStartName); |
5264 | MCSymbol *FnEndSym = getContext().getOrCreateSymbol(Name: FnEndName); |
5265 | getStreamer().emitCVInlineLinetableDirective(PrimaryFunctionId, SourceFileId, |
5266 | SourceLineNum, FnStartSym, |
5267 | FnEndSym); |
5268 | return false; |
5269 | } |
5270 | |
5271 | void MasmParser::initializeCVDefRangeTypeMap() { |
5272 | CVDefRangeTypeMap["reg" ] = CVDR_DEFRANGE_REGISTER; |
5273 | CVDefRangeTypeMap["frame_ptr_rel" ] = CVDR_DEFRANGE_FRAMEPOINTER_REL; |
5274 | CVDefRangeTypeMap["subfield_reg" ] = CVDR_DEFRANGE_SUBFIELD_REGISTER; |
5275 | CVDefRangeTypeMap["reg_rel" ] = CVDR_DEFRANGE_REGISTER_REL; |
5276 | } |
5277 | |
5278 | /// parseDirectiveCVDefRange |
5279 | /// ::= .cv_def_range RangeStart RangeEnd (GapStart GapEnd)*, bytes* |
5280 | bool MasmParser::parseDirectiveCVDefRange() { |
5281 | SMLoc Loc; |
5282 | std::vector<std::pair<const MCSymbol *, const MCSymbol *>> Ranges; |
5283 | while (getLexer().is(K: AsmToken::Identifier)) { |
5284 | Loc = getLexer().getLoc(); |
5285 | StringRef GapStartName; |
5286 | if (parseIdentifier(Res&: GapStartName)) |
5287 | return Error(L: Loc, Msg: "expected identifier in directive" ); |
5288 | MCSymbol *GapStartSym = getContext().getOrCreateSymbol(Name: GapStartName); |
5289 | |
5290 | Loc = getLexer().getLoc(); |
5291 | StringRef GapEndName; |
5292 | if (parseIdentifier(Res&: GapEndName)) |
5293 | return Error(L: Loc, Msg: "expected identifier in directive" ); |
5294 | MCSymbol *GapEndSym = getContext().getOrCreateSymbol(Name: GapEndName); |
5295 | |
5296 | Ranges.push_back(x: {GapStartSym, GapEndSym}); |
5297 | } |
5298 | |
5299 | StringRef CVDefRangeTypeStr; |
5300 | if (parseToken( |
5301 | T: AsmToken::Comma, |
5302 | Msg: "expected comma before def_range type in .cv_def_range directive" ) || |
5303 | parseIdentifier(Res&: CVDefRangeTypeStr)) |
5304 | return Error(L: Loc, Msg: "expected def_range type in directive" ); |
5305 | |
5306 | StringMap<CVDefRangeType>::const_iterator CVTypeIt = |
5307 | CVDefRangeTypeMap.find(Key: CVDefRangeTypeStr); |
5308 | CVDefRangeType CVDRType = (CVTypeIt == CVDefRangeTypeMap.end()) |
5309 | ? CVDR_DEFRANGE |
5310 | : CVTypeIt->getValue(); |
5311 | switch (CVDRType) { |
5312 | case CVDR_DEFRANGE_REGISTER: { |
5313 | int64_t DRRegister; |
5314 | if (parseToken(T: AsmToken::Comma, Msg: "expected comma before register number in " |
5315 | ".cv_def_range directive" ) || |
5316 | parseAbsoluteExpression(Res&: DRRegister)) |
5317 | return Error(L: Loc, Msg: "expected register number" ); |
5318 | |
5319 | codeview::DefRangeRegisterHeader DRHdr; |
5320 | DRHdr.Register = DRRegister; |
5321 | DRHdr.MayHaveNoName = 0; |
5322 | getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); |
5323 | break; |
5324 | } |
5325 | case CVDR_DEFRANGE_FRAMEPOINTER_REL: { |
5326 | int64_t DROffset; |
5327 | if (parseToken(T: AsmToken::Comma, |
5328 | Msg: "expected comma before offset in .cv_def_range directive" ) || |
5329 | parseAbsoluteExpression(Res&: DROffset)) |
5330 | return Error(L: Loc, Msg: "expected offset value" ); |
5331 | |
5332 | codeview::DefRangeFramePointerRelHeader DRHdr; |
5333 | DRHdr.Offset = DROffset; |
5334 | getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); |
5335 | break; |
5336 | } |
5337 | case CVDR_DEFRANGE_SUBFIELD_REGISTER: { |
5338 | int64_t DRRegister; |
5339 | int64_t DROffsetInParent; |
5340 | if (parseToken(T: AsmToken::Comma, Msg: "expected comma before register number in " |
5341 | ".cv_def_range directive" ) || |
5342 | parseAbsoluteExpression(Res&: DRRegister)) |
5343 | return Error(L: Loc, Msg: "expected register number" ); |
5344 | if (parseToken(T: AsmToken::Comma, |
5345 | Msg: "expected comma before offset in .cv_def_range directive" ) || |
5346 | parseAbsoluteExpression(Res&: DROffsetInParent)) |
5347 | return Error(L: Loc, Msg: "expected offset value" ); |
5348 | |
5349 | codeview::DefRangeSubfieldRegisterHeader DRHdr; |
5350 | DRHdr.Register = DRRegister; |
5351 | DRHdr.MayHaveNoName = 0; |
5352 | DRHdr.OffsetInParent = DROffsetInParent; |
5353 | getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); |
5354 | break; |
5355 | } |
5356 | case CVDR_DEFRANGE_REGISTER_REL: { |
5357 | int64_t DRRegister; |
5358 | int64_t DRFlags; |
5359 | int64_t DRBasePointerOffset; |
5360 | if (parseToken(T: AsmToken::Comma, Msg: "expected comma before register number in " |
5361 | ".cv_def_range directive" ) || |
5362 | parseAbsoluteExpression(Res&: DRRegister)) |
5363 | return Error(L: Loc, Msg: "expected register value" ); |
5364 | if (parseToken( |
5365 | T: AsmToken::Comma, |
5366 | Msg: "expected comma before flag value in .cv_def_range directive" ) || |
5367 | parseAbsoluteExpression(Res&: DRFlags)) |
5368 | return Error(L: Loc, Msg: "expected flag value" ); |
5369 | if (parseToken(T: AsmToken::Comma, Msg: "expected comma before base pointer offset " |
5370 | "in .cv_def_range directive" ) || |
5371 | parseAbsoluteExpression(Res&: DRBasePointerOffset)) |
5372 | return Error(L: Loc, Msg: "expected base pointer offset value" ); |
5373 | |
5374 | codeview::DefRangeRegisterRelHeader DRHdr; |
5375 | DRHdr.Register = DRRegister; |
5376 | DRHdr.Flags = DRFlags; |
5377 | DRHdr.BasePointerOffset = DRBasePointerOffset; |
5378 | getStreamer().emitCVDefRangeDirective(Ranges, DRHdr); |
5379 | break; |
5380 | } |
5381 | default: |
5382 | return Error(L: Loc, Msg: "unexpected def_range type in .cv_def_range directive" ); |
5383 | } |
5384 | return true; |
5385 | } |
5386 | |
5387 | /// parseDirectiveCVString |
5388 | /// ::= .cv_stringtable "string" |
5389 | bool MasmParser::parseDirectiveCVString() { |
5390 | std::string Data; |
5391 | if (checkForValidSection() || parseEscapedString(Data)) |
5392 | return addErrorSuffix(Suffix: " in '.cv_string' directive" ); |
5393 | |
5394 | // Put the string in the table and emit the offset. |
5395 | std::pair<StringRef, unsigned> Insertion = |
5396 | getCVContext().addToStringTable(S: Data); |
5397 | getStreamer().emitIntValue(Value: Insertion.second, Size: 4); |
5398 | return false; |
5399 | } |
5400 | |
5401 | /// parseDirectiveCVStringTable |
5402 | /// ::= .cv_stringtable |
5403 | bool MasmParser::parseDirectiveCVStringTable() { |
5404 | getStreamer().emitCVStringTableDirective(); |
5405 | return false; |
5406 | } |
5407 | |
5408 | /// parseDirectiveCVFileChecksums |
5409 | /// ::= .cv_filechecksums |
5410 | bool MasmParser::parseDirectiveCVFileChecksums() { |
5411 | getStreamer().emitCVFileChecksumsDirective(); |
5412 | return false; |
5413 | } |
5414 | |
5415 | /// parseDirectiveCVFileChecksumOffset |
5416 | /// ::= .cv_filechecksumoffset fileno |
5417 | bool MasmParser::parseDirectiveCVFileChecksumOffset() { |
5418 | int64_t FileNo; |
5419 | if (parseIntToken(V&: FileNo, ErrMsg: "expected identifier in directive" )) |
5420 | return true; |
5421 | if (parseEOL()) |
5422 | return true; |
5423 | getStreamer().emitCVFileChecksumOffsetDirective(FileNo); |
5424 | return false; |
5425 | } |
5426 | |
5427 | /// parseDirectiveCVFPOData |
5428 | /// ::= .cv_fpo_data procsym |
5429 | bool MasmParser::parseDirectiveCVFPOData() { |
5430 | SMLoc DirLoc = getLexer().getLoc(); |
5431 | StringRef ProcName; |
5432 | if (parseIdentifier(Res&: ProcName)) |
5433 | return TokError(Msg: "expected symbol name" ); |
5434 | if (parseEOL(ErrMsg: "unexpected tokens" )) |
5435 | return addErrorSuffix(Suffix: " in '.cv_fpo_data' directive" ); |
5436 | MCSymbol *ProcSym = getContext().getOrCreateSymbol(Name: ProcName); |
5437 | getStreamer().emitCVFPOData(ProcSym, Loc: DirLoc); |
5438 | return false; |
5439 | } |
5440 | |
5441 | /// parseDirectiveCFISections |
5442 | /// ::= .cfi_sections section [, section] |
5443 | bool MasmParser::parseDirectiveCFISections() { |
5444 | StringRef Name; |
5445 | bool EH = false; |
5446 | bool Debug = false; |
5447 | |
5448 | if (parseIdentifier(Res&: Name)) |
5449 | return TokError(Msg: "Expected an identifier" ); |
5450 | |
5451 | if (Name == ".eh_frame" ) |
5452 | EH = true; |
5453 | else if (Name == ".debug_frame" ) |
5454 | Debug = true; |
5455 | |
5456 | if (getLexer().is(K: AsmToken::Comma)) { |
5457 | Lex(); |
5458 | |
5459 | if (parseIdentifier(Res&: Name)) |
5460 | return TokError(Msg: "Expected an identifier" ); |
5461 | |
5462 | if (Name == ".eh_frame" ) |
5463 | EH = true; |
5464 | else if (Name == ".debug_frame" ) |
5465 | Debug = true; |
5466 | } |
5467 | |
5468 | getStreamer().emitCFISections(EH, Debug); |
5469 | return false; |
5470 | } |
5471 | |
5472 | /// parseDirectiveCFIStartProc |
5473 | /// ::= .cfi_startproc [simple] |
5474 | bool MasmParser::parseDirectiveCFIStartProc() { |
5475 | StringRef Simple; |
5476 | if (!parseOptionalToken(T: AsmToken::EndOfStatement)) { |
5477 | if (check(P: parseIdentifier(Res&: Simple) || Simple != "simple" , |
5478 | Msg: "unexpected token" ) || |
5479 | parseEOL()) |
5480 | return addErrorSuffix(Suffix: " in '.cfi_startproc' directive" ); |
5481 | } |
5482 | |
5483 | // TODO(kristina): Deal with a corner case of incorrect diagnostic context |
5484 | // being produced if this directive is emitted as part of preprocessor macro |
5485 | // expansion which can *ONLY* happen if Clang's cc1as is the API consumer. |
5486 | // Tools like llvm-mc on the other hand are not affected by it, and report |
5487 | // correct context information. |
5488 | getStreamer().emitCFIStartProc(IsSimple: !Simple.empty(), Loc: Lexer.getLoc()); |
5489 | return false; |
5490 | } |
5491 | |
5492 | /// parseDirectiveCFIEndProc |
5493 | /// ::= .cfi_endproc |
5494 | bool MasmParser::parseDirectiveCFIEndProc() { |
5495 | getStreamer().emitCFIEndProc(); |
5496 | return false; |
5497 | } |
5498 | |
5499 | /// parse register name or number. |
5500 | bool MasmParser::parseRegisterOrRegisterNumber(int64_t &Register, |
5501 | SMLoc DirectiveLoc) { |
5502 | MCRegister RegNo; |
5503 | |
5504 | if (getLexer().isNot(K: AsmToken::Integer)) { |
5505 | if (getTargetParser().parseRegister(Reg&: RegNo, StartLoc&: DirectiveLoc, EndLoc&: DirectiveLoc)) |
5506 | return true; |
5507 | Register = getContext().getRegisterInfo()->getDwarfRegNum(RegNum: RegNo, isEH: true); |
5508 | } else |
5509 | return parseAbsoluteExpression(Res&: Register); |
5510 | |
5511 | return false; |
5512 | } |
5513 | |
5514 | /// parseDirectiveCFIDefCfa |
5515 | /// ::= .cfi_def_cfa register, offset |
5516 | bool MasmParser::parseDirectiveCFIDefCfa(SMLoc DirectiveLoc) { |
5517 | int64_t Register = 0, Offset = 0; |
5518 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || |
5519 | parseToken(T: AsmToken::Comma, Msg: "unexpected token in directive" ) || |
5520 | parseAbsoluteExpression(Res&: Offset)) |
5521 | return true; |
5522 | |
5523 | getStreamer().emitCFIDefCfa(Register, Offset); |
5524 | return false; |
5525 | } |
5526 | |
5527 | /// parseDirectiveCFIDefCfaOffset |
5528 | /// ::= .cfi_def_cfa_offset offset |
5529 | bool MasmParser::parseDirectiveCFIDefCfaOffset(SMLoc DirectiveLoc) { |
5530 | int64_t Offset = 0; |
5531 | if (parseAbsoluteExpression(Res&: Offset)) |
5532 | return true; |
5533 | |
5534 | getStreamer().emitCFIDefCfaOffset(Offset, Loc: DirectiveLoc); |
5535 | return false; |
5536 | } |
5537 | |
5538 | /// parseDirectiveCFIRegister |
5539 | /// ::= .cfi_register register, register |
5540 | bool MasmParser::parseDirectiveCFIRegister(SMLoc DirectiveLoc) { |
5541 | int64_t Register1 = 0, Register2 = 0; |
5542 | if (parseRegisterOrRegisterNumber(Register&: Register1, DirectiveLoc) || |
5543 | parseToken(T: AsmToken::Comma, Msg: "unexpected token in directive" ) || |
5544 | parseRegisterOrRegisterNumber(Register&: Register2, DirectiveLoc)) |
5545 | return true; |
5546 | |
5547 | getStreamer().emitCFIRegister(Register1, Register2, Loc: DirectiveLoc); |
5548 | return false; |
5549 | } |
5550 | |
5551 | /// parseDirectiveCFIWindowSave |
5552 | /// ::= .cfi_window_save |
5553 | bool MasmParser::parseDirectiveCFIWindowSave(SMLoc DirectiveLoc) { |
5554 | getStreamer().emitCFIWindowSave(Loc: DirectiveLoc); |
5555 | return false; |
5556 | } |
5557 | |
5558 | /// parseDirectiveCFIAdjustCfaOffset |
5559 | /// ::= .cfi_adjust_cfa_offset adjustment |
5560 | bool MasmParser::parseDirectiveCFIAdjustCfaOffset(SMLoc DirectiveLoc) { |
5561 | int64_t Adjustment = 0; |
5562 | if (parseAbsoluteExpression(Res&: Adjustment)) |
5563 | return true; |
5564 | |
5565 | getStreamer().emitCFIAdjustCfaOffset(Adjustment, Loc: DirectiveLoc); |
5566 | return false; |
5567 | } |
5568 | |
5569 | /// parseDirectiveCFIDefCfaRegister |
5570 | /// ::= .cfi_def_cfa_register register |
5571 | bool MasmParser::parseDirectiveCFIDefCfaRegister(SMLoc DirectiveLoc) { |
5572 | int64_t Register = 0; |
5573 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) |
5574 | return true; |
5575 | |
5576 | getStreamer().emitCFIDefCfaRegister(Register); |
5577 | return false; |
5578 | } |
5579 | |
5580 | /// parseDirectiveCFIOffset |
5581 | /// ::= .cfi_offset register, offset |
5582 | bool MasmParser::parseDirectiveCFIOffset(SMLoc DirectiveLoc) { |
5583 | int64_t Register = 0; |
5584 | int64_t Offset = 0; |
5585 | |
5586 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || |
5587 | parseToken(T: AsmToken::Comma, Msg: "unexpected token in directive" ) || |
5588 | parseAbsoluteExpression(Res&: Offset)) |
5589 | return true; |
5590 | |
5591 | getStreamer().emitCFIOffset(Register, Offset); |
5592 | return false; |
5593 | } |
5594 | |
5595 | /// parseDirectiveCFIRelOffset |
5596 | /// ::= .cfi_rel_offset register, offset |
5597 | bool MasmParser::parseDirectiveCFIRelOffset(SMLoc DirectiveLoc) { |
5598 | int64_t Register = 0, Offset = 0; |
5599 | |
5600 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc) || |
5601 | parseToken(T: AsmToken::Comma, Msg: "unexpected token in directive" ) || |
5602 | parseAbsoluteExpression(Res&: Offset)) |
5603 | return true; |
5604 | |
5605 | getStreamer().emitCFIRelOffset(Register, Offset, Loc: DirectiveLoc); |
5606 | return false; |
5607 | } |
5608 | |
5609 | static bool isValidEncoding(int64_t Encoding) { |
5610 | if (Encoding & ~0xff) |
5611 | return false; |
5612 | |
5613 | if (Encoding == dwarf::DW_EH_PE_omit) |
5614 | return true; |
5615 | |
5616 | const unsigned Format = Encoding & 0xf; |
5617 | if (Format != dwarf::DW_EH_PE_absptr && Format != dwarf::DW_EH_PE_udata2 && |
5618 | Format != dwarf::DW_EH_PE_udata4 && Format != dwarf::DW_EH_PE_udata8 && |
5619 | Format != dwarf::DW_EH_PE_sdata2 && Format != dwarf::DW_EH_PE_sdata4 && |
5620 | Format != dwarf::DW_EH_PE_sdata8 && Format != dwarf::DW_EH_PE_signed) |
5621 | return false; |
5622 | |
5623 | const unsigned Application = Encoding & 0x70; |
5624 | if (Application != dwarf::DW_EH_PE_absptr && |
5625 | Application != dwarf::DW_EH_PE_pcrel) |
5626 | return false; |
5627 | |
5628 | return true; |
5629 | } |
5630 | |
5631 | /// parseDirectiveCFIPersonalityOrLsda |
5632 | /// IsPersonality true for cfi_personality, false for cfi_lsda |
5633 | /// ::= .cfi_personality encoding, [symbol_name] |
5634 | /// ::= .cfi_lsda encoding, [symbol_name] |
5635 | bool MasmParser::parseDirectiveCFIPersonalityOrLsda(bool IsPersonality) { |
5636 | int64_t Encoding = 0; |
5637 | if (parseAbsoluteExpression(Res&: Encoding)) |
5638 | return true; |
5639 | if (Encoding == dwarf::DW_EH_PE_omit) |
5640 | return false; |
5641 | |
5642 | StringRef Name; |
5643 | if (check(P: !isValidEncoding(Encoding), Msg: "unsupported encoding." ) || |
5644 | parseToken(T: AsmToken::Comma, Msg: "unexpected token in directive" ) || |
5645 | check(P: parseIdentifier(Res&: Name), Msg: "expected identifier in directive" )) |
5646 | return true; |
5647 | |
5648 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
5649 | |
5650 | if (IsPersonality) |
5651 | getStreamer().emitCFIPersonality(Sym, Encoding); |
5652 | else |
5653 | getStreamer().emitCFILsda(Sym, Encoding); |
5654 | return false; |
5655 | } |
5656 | |
5657 | /// parseDirectiveCFIRememberState |
5658 | /// ::= .cfi_remember_state |
5659 | bool MasmParser::parseDirectiveCFIRememberState(SMLoc DirectiveLoc) { |
5660 | getStreamer().emitCFIRememberState(Loc: DirectiveLoc); |
5661 | return false; |
5662 | } |
5663 | |
5664 | /// parseDirectiveCFIRestoreState |
5665 | /// ::= .cfi_remember_state |
5666 | bool MasmParser::parseDirectiveCFIRestoreState(SMLoc DirectiveLoc) { |
5667 | getStreamer().emitCFIRestoreState(Loc: DirectiveLoc); |
5668 | return false; |
5669 | } |
5670 | |
5671 | /// parseDirectiveCFISameValue |
5672 | /// ::= .cfi_same_value register |
5673 | bool MasmParser::parseDirectiveCFISameValue(SMLoc DirectiveLoc) { |
5674 | int64_t Register = 0; |
5675 | |
5676 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) |
5677 | return true; |
5678 | |
5679 | getStreamer().emitCFISameValue(Register, Loc: DirectiveLoc); |
5680 | return false; |
5681 | } |
5682 | |
5683 | /// parseDirectiveCFIRestore |
5684 | /// ::= .cfi_restore register |
5685 | bool MasmParser::parseDirectiveCFIRestore(SMLoc DirectiveLoc) { |
5686 | int64_t Register = 0; |
5687 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) |
5688 | return true; |
5689 | |
5690 | getStreamer().emitCFIRestore(Register); |
5691 | return false; |
5692 | } |
5693 | |
5694 | /// parseDirectiveCFIEscape |
5695 | /// ::= .cfi_escape expression[,...] |
5696 | bool MasmParser::parseDirectiveCFIEscape(SMLoc DirectiveLoc) { |
5697 | std::string Values; |
5698 | int64_t CurrValue; |
5699 | if (parseAbsoluteExpression(Res&: CurrValue)) |
5700 | return true; |
5701 | |
5702 | Values.push_back(c: (uint8_t)CurrValue); |
5703 | |
5704 | while (getLexer().is(K: AsmToken::Comma)) { |
5705 | Lex(); |
5706 | |
5707 | if (parseAbsoluteExpression(Res&: CurrValue)) |
5708 | return true; |
5709 | |
5710 | Values.push_back(c: (uint8_t)CurrValue); |
5711 | } |
5712 | |
5713 | getStreamer().emitCFIEscape(Values, Loc: DirectiveLoc); |
5714 | return false; |
5715 | } |
5716 | |
5717 | /// parseDirectiveCFIReturnColumn |
5718 | /// ::= .cfi_return_column register |
5719 | bool MasmParser::parseDirectiveCFIReturnColumn(SMLoc DirectiveLoc) { |
5720 | int64_t Register = 0; |
5721 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) |
5722 | return true; |
5723 | getStreamer().emitCFIReturnColumn(Register); |
5724 | return false; |
5725 | } |
5726 | |
5727 | /// parseDirectiveCFISignalFrame |
5728 | /// ::= .cfi_signal_frame |
5729 | bool MasmParser::parseDirectiveCFISignalFrame() { |
5730 | if (parseEOL()) |
5731 | return true; |
5732 | |
5733 | getStreamer().emitCFISignalFrame(); |
5734 | return false; |
5735 | } |
5736 | |
5737 | /// parseDirectiveCFIUndefined |
5738 | /// ::= .cfi_undefined register |
5739 | bool MasmParser::parseDirectiveCFIUndefined(SMLoc DirectiveLoc) { |
5740 | int64_t Register = 0; |
5741 | |
5742 | if (parseRegisterOrRegisterNumber(Register, DirectiveLoc)) |
5743 | return true; |
5744 | |
5745 | getStreamer().emitCFIUndefined(Register); |
5746 | return false; |
5747 | } |
5748 | |
5749 | /// parseDirectiveMacro |
5750 | /// ::= name macro [parameters] |
5751 | /// ["LOCAL" identifiers] |
5752 | /// parameters ::= parameter [, parameter]* |
5753 | /// parameter ::= name ":" qualifier |
5754 | /// qualifier ::= "req" | "vararg" | "=" macro_argument |
5755 | bool MasmParser::parseDirectiveMacro(StringRef Name, SMLoc NameLoc) { |
5756 | MCAsmMacroParameters Parameters; |
5757 | while (getLexer().isNot(K: AsmToken::EndOfStatement)) { |
5758 | if (!Parameters.empty() && Parameters.back().Vararg) |
5759 | return Error(L: Lexer.getLoc(), |
5760 | Msg: "Vararg parameter '" + Parameters.back().Name + |
5761 | "' should be last in the list of parameters" ); |
5762 | |
5763 | MCAsmMacroParameter Parameter; |
5764 | if (parseIdentifier(Res&: Parameter.Name)) |
5765 | return TokError(Msg: "expected identifier in 'macro' directive" ); |
5766 | |
5767 | // Emit an error if two (or more) named parameters share the same name. |
5768 | for (const MCAsmMacroParameter& CurrParam : Parameters) |
5769 | if (CurrParam.Name.equals_insensitive(RHS: Parameter.Name)) |
5770 | return TokError(Msg: "macro '" + Name + "' has multiple parameters" |
5771 | " named '" + Parameter.Name + "'" ); |
5772 | |
5773 | if (Lexer.is(K: AsmToken::Colon)) { |
5774 | Lex(); // consume ':' |
5775 | |
5776 | if (parseOptionalToken(T: AsmToken::Equal)) { |
5777 | // Default value |
5778 | SMLoc ParamLoc; |
5779 | |
5780 | ParamLoc = Lexer.getLoc(); |
5781 | if (parseMacroArgument(MP: nullptr, MA&: Parameter.Value)) |
5782 | return true; |
5783 | } else { |
5784 | SMLoc QualLoc; |
5785 | StringRef Qualifier; |
5786 | |
5787 | QualLoc = Lexer.getLoc(); |
5788 | if (parseIdentifier(Res&: Qualifier)) |
5789 | return Error(L: QualLoc, Msg: "missing parameter qualifier for " |
5790 | "'" + |
5791 | Parameter.Name + "' in macro '" + Name + |
5792 | "'" ); |
5793 | |
5794 | if (Qualifier.equals_insensitive(RHS: "req" )) |
5795 | Parameter.Required = true; |
5796 | else if (Qualifier.equals_insensitive(RHS: "vararg" )) |
5797 | Parameter.Vararg = true; |
5798 | else |
5799 | return Error(L: QualLoc, |
5800 | Msg: Qualifier + " is not a valid parameter qualifier for '" + |
5801 | Parameter.Name + "' in macro '" + Name + "'" ); |
5802 | } |
5803 | } |
5804 | |
5805 | Parameters.push_back(x: std::move(Parameter)); |
5806 | |
5807 | if (getLexer().is(K: AsmToken::Comma)) |
5808 | Lex(); |
5809 | } |
5810 | |
5811 | // Eat just the end of statement. |
5812 | Lexer.Lex(); |
5813 | |
5814 | std::vector<std::string> Locals; |
5815 | if (getTok().is(K: AsmToken::Identifier) && |
5816 | getTok().getIdentifier().equals_insensitive(RHS: "local" )) { |
5817 | Lex(); // Eat the LOCAL directive. |
5818 | |
5819 | StringRef ID; |
5820 | while (true) { |
5821 | if (parseIdentifier(Res&: ID)) |
5822 | return true; |
5823 | Locals.push_back(x: ID.lower()); |
5824 | |
5825 | // If we see a comma, continue (and allow line continuation). |
5826 | if (!parseOptionalToken(T: AsmToken::Comma)) |
5827 | break; |
5828 | parseOptionalToken(T: AsmToken::EndOfStatement); |
5829 | } |
5830 | } |
5831 | |
5832 | // Consuming deferred text, so use Lexer.Lex to ignore Lexing Errors. |
5833 | AsmToken EndToken, StartToken = getTok(); |
5834 | unsigned MacroDepth = 0; |
5835 | bool IsMacroFunction = false; |
5836 | // Lex the macro definition. |
5837 | while (true) { |
5838 | // Ignore Lexing errors in macros. |
5839 | while (Lexer.is(K: AsmToken::Error)) { |
5840 | Lexer.Lex(); |
5841 | } |
5842 | |
5843 | // Check whether we have reached the end of the file. |
5844 | if (getLexer().is(K: AsmToken::Eof)) |
5845 | return Error(L: NameLoc, Msg: "no matching 'endm' in definition" ); |
5846 | |
5847 | // Otherwise, check whether we have reached the 'endm'... and determine if |
5848 | // this is a macro function. |
5849 | if (getLexer().is(K: AsmToken::Identifier)) { |
5850 | if (getTok().getIdentifier().equals_insensitive(RHS: "endm" )) { |
5851 | if (MacroDepth == 0) { // Outermost macro. |
5852 | EndToken = getTok(); |
5853 | Lexer.Lex(); |
5854 | if (getLexer().isNot(K: AsmToken::EndOfStatement)) |
5855 | return TokError(Msg: "unexpected token in '" + EndToken.getIdentifier() + |
5856 | "' directive" ); |
5857 | break; |
5858 | } else { |
5859 | // Otherwise we just found the end of an inner macro. |
5860 | --MacroDepth; |
5861 | } |
5862 | } else if (getTok().getIdentifier().equals_insensitive(RHS: "exitm" )) { |
5863 | if (MacroDepth == 0 && peekTok().isNot(K: AsmToken::EndOfStatement)) { |
5864 | IsMacroFunction = true; |
5865 | } |
5866 | } else if (isMacroLikeDirective()) { |
5867 | // We allow nested macros. Those aren't instantiated until the |
5868 | // outermost macro is expanded so just ignore them for now. |
5869 | ++MacroDepth; |
5870 | } |
5871 | } |
5872 | |
5873 | // Otherwise, scan til the end of the statement. |
5874 | eatToEndOfStatement(); |
5875 | } |
5876 | |
5877 | if (getContext().lookupMacro(Name: Name.lower())) { |
5878 | return Error(L: NameLoc, Msg: "macro '" + Name + "' is already defined" ); |
5879 | } |
5880 | |
5881 | const char *BodyStart = StartToken.getLoc().getPointer(); |
5882 | const char *BodyEnd = EndToken.getLoc().getPointer(); |
5883 | StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); |
5884 | MCAsmMacro Macro(Name, Body, std::move(Parameters), std::move(Locals), |
5885 | IsMacroFunction); |
5886 | DEBUG_WITH_TYPE("asm-macros" , dbgs() << "Defining new macro:\n" ; |
5887 | Macro.dump()); |
5888 | getContext().defineMacro(Name: Name.lower(), Macro: std::move(Macro)); |
5889 | return false; |
5890 | } |
5891 | |
5892 | /// parseDirectiveExitMacro |
5893 | /// ::= "exitm" [textitem] |
5894 | bool MasmParser::parseDirectiveExitMacro(SMLoc DirectiveLoc, |
5895 | StringRef Directive, |
5896 | std::string &Value) { |
5897 | SMLoc EndLoc = getTok().getLoc(); |
5898 | if (getTok().isNot(K: AsmToken::EndOfStatement) && parseTextItem(Data&: Value)) |
5899 | return Error(L: EndLoc, |
5900 | Msg: "unable to parse text item in '" + Directive + "' directive" ); |
5901 | eatToEndOfStatement(); |
5902 | |
5903 | if (!isInsideMacroInstantiation()) |
5904 | return TokError(Msg: "unexpected '" + Directive + "' in file, " |
5905 | "no current macro definition" ); |
5906 | |
5907 | // Exit all conditionals that are active in the current macro. |
5908 | while (TheCondStack.size() != ActiveMacros.back()->CondStackDepth) { |
5909 | TheCondState = TheCondStack.back(); |
5910 | TheCondStack.pop_back(); |
5911 | } |
5912 | |
5913 | handleMacroExit(); |
5914 | return false; |
5915 | } |
5916 | |
5917 | /// parseDirectiveEndMacro |
5918 | /// ::= endm |
5919 | bool MasmParser::parseDirectiveEndMacro(StringRef Directive) { |
5920 | if (getLexer().isNot(K: AsmToken::EndOfStatement)) |
5921 | return TokError(Msg: "unexpected token in '" + Directive + "' directive" ); |
5922 | |
5923 | // If we are inside a macro instantiation, terminate the current |
5924 | // instantiation. |
5925 | if (isInsideMacroInstantiation()) { |
5926 | handleMacroExit(); |
5927 | return false; |
5928 | } |
5929 | |
5930 | // Otherwise, this .endmacro is a stray entry in the file; well formed |
5931 | // .endmacro directives are handled during the macro definition parsing. |
5932 | return TokError(Msg: "unexpected '" + Directive + "' in file, " |
5933 | "no current macro definition" ); |
5934 | } |
5935 | |
5936 | /// parseDirectivePurgeMacro |
5937 | /// ::= purge identifier ( , identifier )* |
5938 | bool MasmParser::parseDirectivePurgeMacro(SMLoc DirectiveLoc) { |
5939 | StringRef Name; |
5940 | while (true) { |
5941 | SMLoc NameLoc; |
5942 | if (parseTokenLoc(Loc&: NameLoc) || |
5943 | check(P: parseIdentifier(Res&: Name), Loc: NameLoc, |
5944 | Msg: "expected identifier in 'purge' directive" )) |
5945 | return true; |
5946 | |
5947 | DEBUG_WITH_TYPE("asm-macros" , dbgs() |
5948 | << "Un-defining macro: " << Name << "\n" ); |
5949 | if (!getContext().lookupMacro(Name: Name.lower())) |
5950 | return Error(L: NameLoc, Msg: "macro '" + Name + "' is not defined" ); |
5951 | getContext().undefineMacro(Name: Name.lower()); |
5952 | |
5953 | if (!parseOptionalToken(T: AsmToken::Comma)) |
5954 | break; |
5955 | parseOptionalToken(T: AsmToken::EndOfStatement); |
5956 | } |
5957 | |
5958 | return false; |
5959 | } |
5960 | |
5961 | bool MasmParser::parseDirectiveExtern() { |
5962 | // .extern is the default - but we still need to take any provided type info. |
5963 | auto parseOp = [&]() -> bool { |
5964 | StringRef Name; |
5965 | SMLoc NameLoc = getTok().getLoc(); |
5966 | if (parseIdentifier(Res&: Name)) |
5967 | return Error(L: NameLoc, Msg: "expected name" ); |
5968 | if (parseToken(T: AsmToken::Colon)) |
5969 | return true; |
5970 | |
5971 | StringRef TypeName; |
5972 | SMLoc TypeLoc = getTok().getLoc(); |
5973 | if (parseIdentifier(Res&: TypeName)) |
5974 | return Error(L: TypeLoc, Msg: "expected type" ); |
5975 | if (!TypeName.equals_insensitive(RHS: "proc" )) { |
5976 | AsmTypeInfo Type; |
5977 | if (lookUpType(Name: TypeName, Info&: Type)) |
5978 | return Error(L: TypeLoc, Msg: "unrecognized type" ); |
5979 | KnownType[Name.lower()] = Type; |
5980 | } |
5981 | |
5982 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
5983 | Sym->setExternal(true); |
5984 | getStreamer().emitSymbolAttribute(Symbol: Sym, Attribute: MCSA_Extern); |
5985 | |
5986 | return false; |
5987 | }; |
5988 | |
5989 | if (parseMany(parseOne: parseOp)) |
5990 | return addErrorSuffix(Suffix: " in directive 'extern'" ); |
5991 | return false; |
5992 | } |
5993 | |
5994 | /// parseDirectiveSymbolAttribute |
5995 | /// ::= { ".globl", ".weak", ... } [ identifier ( , identifier )* ] |
5996 | bool MasmParser::parseDirectiveSymbolAttribute(MCSymbolAttr Attr) { |
5997 | auto parseOp = [&]() -> bool { |
5998 | StringRef Name; |
5999 | SMLoc Loc = getTok().getLoc(); |
6000 | if (parseIdentifier(Res&: Name)) |
6001 | return Error(L: Loc, Msg: "expected identifier" ); |
6002 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
6003 | |
6004 | // Assembler local symbols don't make any sense here. Complain loudly. |
6005 | if (Sym->isTemporary()) |
6006 | return Error(L: Loc, Msg: "non-local symbol required" ); |
6007 | |
6008 | if (!getStreamer().emitSymbolAttribute(Symbol: Sym, Attribute: Attr)) |
6009 | return Error(L: Loc, Msg: "unable to emit symbol attribute" ); |
6010 | return false; |
6011 | }; |
6012 | |
6013 | if (parseMany(parseOne: parseOp)) |
6014 | return addErrorSuffix(Suffix: " in directive" ); |
6015 | return false; |
6016 | } |
6017 | |
6018 | /// parseDirectiveComm |
6019 | /// ::= ( .comm | .lcomm ) identifier , size_expression [ , align_expression ] |
6020 | bool MasmParser::parseDirectiveComm(bool IsLocal) { |
6021 | if (checkForValidSection()) |
6022 | return true; |
6023 | |
6024 | SMLoc IDLoc = getLexer().getLoc(); |
6025 | StringRef Name; |
6026 | if (parseIdentifier(Res&: Name)) |
6027 | return TokError(Msg: "expected identifier in directive" ); |
6028 | |
6029 | // Handle the identifier as the key symbol. |
6030 | MCSymbol *Sym = getContext().getOrCreateSymbol(Name); |
6031 | |
6032 | if (getLexer().isNot(K: AsmToken::Comma)) |
6033 | return TokError(Msg: "unexpected token in directive" ); |
6034 | Lex(); |
6035 | |
6036 | int64_t Size; |
6037 | SMLoc SizeLoc = getLexer().getLoc(); |
6038 | if (parseAbsoluteExpression(Res&: Size)) |
6039 | return true; |
6040 | |
6041 | int64_t Pow2Alignment = 0; |
6042 | SMLoc Pow2AlignmentLoc; |
6043 | if (getLexer().is(K: AsmToken::Comma)) { |
6044 | Lex(); |
6045 | Pow2AlignmentLoc = getLexer().getLoc(); |
6046 | if (parseAbsoluteExpression(Res&: Pow2Alignment)) |
6047 | return true; |
6048 | |
6049 | LCOMM::LCOMMType LCOMM = Lexer.getMAI().getLCOMMDirectiveAlignmentType(); |
6050 | if (IsLocal && LCOMM == LCOMM::NoAlignment) |
6051 | return Error(L: Pow2AlignmentLoc, Msg: "alignment not supported on this target" ); |
6052 | |
6053 | // If this target takes alignments in bytes (not log) validate and convert. |
6054 | if ((!IsLocal && Lexer.getMAI().getCOMMDirectiveAlignmentIsInBytes()) || |
6055 | (IsLocal && LCOMM == LCOMM::ByteAlignment)) { |
6056 | if (!isPowerOf2_64(Value: Pow2Alignment)) |
6057 | return Error(L: Pow2AlignmentLoc, Msg: "alignment must be a power of 2" ); |
6058 | Pow2Alignment = Log2_64(Value: Pow2Alignment); |
6059 | } |
6060 | } |
6061 | |
6062 | if (parseEOL()) |
6063 | return true; |
6064 | |
6065 | // NOTE: a size of zero for a .comm should create a undefined symbol |
6066 | // but a size of .lcomm creates a bss symbol of size zero. |
6067 | if (Size < 0) |
6068 | return Error(L: SizeLoc, Msg: "invalid '.comm' or '.lcomm' directive size, can't " |
6069 | "be less than zero" ); |
6070 | |
6071 | // NOTE: The alignment in the directive is a power of 2 value, the assembler |
6072 | // may internally end up wanting an alignment in bytes. |
6073 | // FIXME: Diagnose overflow. |
6074 | if (Pow2Alignment < 0) |
6075 | return Error(L: Pow2AlignmentLoc, Msg: "invalid '.comm' or '.lcomm' directive " |
6076 | "alignment, can't be less than zero" ); |
6077 | |
6078 | Sym->redefineIfPossible(); |
6079 | if (!Sym->isUndefined()) |
6080 | return Error(L: IDLoc, Msg: "invalid symbol redefinition" ); |
6081 | |
6082 | // Create the Symbol as a common or local common with Size and Pow2Alignment. |
6083 | if (IsLocal) { |
6084 | getStreamer().emitLocalCommonSymbol(Symbol: Sym, Size, |
6085 | ByteAlignment: Align(1ULL << Pow2Alignment)); |
6086 | return false; |
6087 | } |
6088 | |
6089 | getStreamer().emitCommonSymbol(Symbol: Sym, Size, ByteAlignment: Align(1ULL << Pow2Alignment)); |
6090 | return false; |
6091 | } |
6092 | |
6093 | /// parseDirectiveComment |
6094 | /// ::= comment delimiter [[text]] |
6095 | /// [[text]] |
6096 | /// [[text]] delimiter [[text]] |
6097 | bool MasmParser::(SMLoc DirectiveLoc) { |
6098 | std::string FirstLine = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6099 | size_t DelimiterEnd = FirstLine.find_first_of(s: "\b\t\v\f\r\x1A " ); |
6100 | assert(DelimiterEnd != std::string::npos); |
6101 | StringRef Delimiter = StringRef(FirstLine).take_front(N: DelimiterEnd); |
6102 | if (Delimiter.empty()) |
6103 | return Error(L: DirectiveLoc, Msg: "no delimiter in 'comment' directive" ); |
6104 | do { |
6105 | if (getTok().is(K: AsmToken::Eof)) |
6106 | return Error(L: DirectiveLoc, Msg: "unmatched delimiter in 'comment' directive" ); |
6107 | Lex(); // eat end of statement |
6108 | } while ( |
6109 | !StringRef(parseStringTo(EndTok: AsmToken::EndOfStatement)).contains(Other: Delimiter)); |
6110 | return parseEOL(); |
6111 | } |
6112 | |
6113 | /// parseDirectiveInclude |
6114 | /// ::= include <filename> |
6115 | /// | include filename |
6116 | bool MasmParser::parseDirectiveInclude() { |
6117 | // Allow the strings to have escaped octal character sequence. |
6118 | std::string Filename; |
6119 | SMLoc IncludeLoc = getTok().getLoc(); |
6120 | |
6121 | if (parseAngleBracketString(Data&: Filename)) |
6122 | Filename = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6123 | if (check(P: Filename.empty(), Msg: "missing filename in 'include' directive" ) || |
6124 | check(P: getTok().isNot(K: AsmToken::EndOfStatement), |
6125 | Msg: "unexpected token in 'include' directive" ) || |
6126 | // Attempt to switch the lexer to the included file before consuming the |
6127 | // end of statement to avoid losing it when we switch. |
6128 | check(P: enterIncludeFile(Filename), Loc: IncludeLoc, |
6129 | Msg: "Could not find include file '" + Filename + "'" )) |
6130 | return true; |
6131 | |
6132 | return false; |
6133 | } |
6134 | |
6135 | /// parseDirectiveIf |
6136 | /// ::= .if{,eq,ge,gt,le,lt,ne} expression |
6137 | bool MasmParser::parseDirectiveIf(SMLoc DirectiveLoc, DirectiveKind DirKind) { |
6138 | TheCondStack.push_back(x: TheCondState); |
6139 | TheCondState.TheCond = AsmCond::IfCond; |
6140 | if (TheCondState.Ignore) { |
6141 | eatToEndOfStatement(); |
6142 | } else { |
6143 | int64_t ExprValue; |
6144 | if (parseAbsoluteExpression(Res&: ExprValue) || parseEOL()) |
6145 | return true; |
6146 | |
6147 | switch (DirKind) { |
6148 | default: |
6149 | llvm_unreachable("unsupported directive" ); |
6150 | case DK_IF: |
6151 | break; |
6152 | case DK_IFE: |
6153 | ExprValue = ExprValue == 0; |
6154 | break; |
6155 | } |
6156 | |
6157 | TheCondState.CondMet = ExprValue; |
6158 | TheCondState.Ignore = !TheCondState.CondMet; |
6159 | } |
6160 | |
6161 | return false; |
6162 | } |
6163 | |
6164 | /// parseDirectiveIfb |
6165 | /// ::= .ifb textitem |
6166 | bool MasmParser::parseDirectiveIfb(SMLoc DirectiveLoc, bool ExpectBlank) { |
6167 | TheCondStack.push_back(x: TheCondState); |
6168 | TheCondState.TheCond = AsmCond::IfCond; |
6169 | |
6170 | if (TheCondState.Ignore) { |
6171 | eatToEndOfStatement(); |
6172 | } else { |
6173 | std::string Str; |
6174 | if (parseTextItem(Data&: Str)) |
6175 | return TokError(Msg: "expected text item parameter for 'ifb' directive" ); |
6176 | |
6177 | if (parseEOL()) |
6178 | return true; |
6179 | |
6180 | TheCondState.CondMet = ExpectBlank == Str.empty(); |
6181 | TheCondState.Ignore = !TheCondState.CondMet; |
6182 | } |
6183 | |
6184 | return false; |
6185 | } |
6186 | |
6187 | /// parseDirectiveIfidn |
6188 | /// ::= ifidn textitem, textitem |
6189 | bool MasmParser::parseDirectiveIfidn(SMLoc DirectiveLoc, bool ExpectEqual, |
6190 | bool CaseInsensitive) { |
6191 | std::string String1, String2; |
6192 | |
6193 | if (parseTextItem(Data&: String1)) { |
6194 | if (ExpectEqual) |
6195 | return TokError(Msg: "expected text item parameter for 'ifidn' directive" ); |
6196 | return TokError(Msg: "expected text item parameter for 'ifdif' directive" ); |
6197 | } |
6198 | |
6199 | if (Lexer.isNot(K: AsmToken::Comma)) { |
6200 | if (ExpectEqual) |
6201 | return TokError( |
6202 | Msg: "expected comma after first string for 'ifidn' directive" ); |
6203 | return TokError(Msg: "expected comma after first string for 'ifdif' directive" ); |
6204 | } |
6205 | Lex(); |
6206 | |
6207 | if (parseTextItem(Data&: String2)) { |
6208 | if (ExpectEqual) |
6209 | return TokError(Msg: "expected text item parameter for 'ifidn' directive" ); |
6210 | return TokError(Msg: "expected text item parameter for 'ifdif' directive" ); |
6211 | } |
6212 | |
6213 | TheCondStack.push_back(x: TheCondState); |
6214 | TheCondState.TheCond = AsmCond::IfCond; |
6215 | if (CaseInsensitive) |
6216 | TheCondState.CondMet = |
6217 | ExpectEqual == (StringRef(String1).equals_insensitive(RHS: String2)); |
6218 | else |
6219 | TheCondState.CondMet = ExpectEqual == (String1 == String2); |
6220 | TheCondState.Ignore = !TheCondState.CondMet; |
6221 | |
6222 | return false; |
6223 | } |
6224 | |
6225 | /// parseDirectiveIfdef |
6226 | /// ::= ifdef symbol |
6227 | /// | ifdef variable |
6228 | bool MasmParser::parseDirectiveIfdef(SMLoc DirectiveLoc, bool expect_defined) { |
6229 | TheCondStack.push_back(x: TheCondState); |
6230 | TheCondState.TheCond = AsmCond::IfCond; |
6231 | |
6232 | if (TheCondState.Ignore) { |
6233 | eatToEndOfStatement(); |
6234 | } else { |
6235 | bool is_defined = false; |
6236 | MCRegister Reg; |
6237 | SMLoc StartLoc, EndLoc; |
6238 | is_defined = |
6239 | getTargetParser().tryParseRegister(Reg, StartLoc, EndLoc).isSuccess(); |
6240 | if (!is_defined) { |
6241 | StringRef Name; |
6242 | if (check(P: parseIdentifier(Res&: Name), Msg: "expected identifier after 'ifdef'" ) || |
6243 | parseEOL()) |
6244 | return true; |
6245 | |
6246 | if (BuiltinSymbolMap.contains(Key: Name.lower())) { |
6247 | is_defined = true; |
6248 | } else if (Variables.contains(Key: Name.lower())) { |
6249 | is_defined = true; |
6250 | } else { |
6251 | MCSymbol *Sym = getContext().lookupSymbol(Name: Name.lower()); |
6252 | is_defined = (Sym && !Sym->isUndefined(SetUsed: false)); |
6253 | } |
6254 | } |
6255 | |
6256 | TheCondState.CondMet = (is_defined == expect_defined); |
6257 | TheCondState.Ignore = !TheCondState.CondMet; |
6258 | } |
6259 | |
6260 | return false; |
6261 | } |
6262 | |
6263 | /// parseDirectiveElseIf |
6264 | /// ::= elseif expression |
6265 | bool MasmParser::parseDirectiveElseIf(SMLoc DirectiveLoc, |
6266 | DirectiveKind DirKind) { |
6267 | if (TheCondState.TheCond != AsmCond::IfCond && |
6268 | TheCondState.TheCond != AsmCond::ElseIfCond) |
6269 | return Error(L: DirectiveLoc, Msg: "Encountered a .elseif that doesn't follow an" |
6270 | " .if or an .elseif" ); |
6271 | TheCondState.TheCond = AsmCond::ElseIfCond; |
6272 | |
6273 | bool LastIgnoreState = false; |
6274 | if (!TheCondStack.empty()) |
6275 | LastIgnoreState = TheCondStack.back().Ignore; |
6276 | if (LastIgnoreState || TheCondState.CondMet) { |
6277 | TheCondState.Ignore = true; |
6278 | eatToEndOfStatement(); |
6279 | } else { |
6280 | int64_t ExprValue; |
6281 | if (parseAbsoluteExpression(Res&: ExprValue)) |
6282 | return true; |
6283 | |
6284 | if (parseEOL()) |
6285 | return true; |
6286 | |
6287 | switch (DirKind) { |
6288 | default: |
6289 | llvm_unreachable("unsupported directive" ); |
6290 | case DK_ELSEIF: |
6291 | break; |
6292 | case DK_ELSEIFE: |
6293 | ExprValue = ExprValue == 0; |
6294 | break; |
6295 | } |
6296 | |
6297 | TheCondState.CondMet = ExprValue; |
6298 | TheCondState.Ignore = !TheCondState.CondMet; |
6299 | } |
6300 | |
6301 | return false; |
6302 | } |
6303 | |
6304 | /// parseDirectiveElseIfb |
6305 | /// ::= elseifb textitem |
6306 | bool MasmParser::parseDirectiveElseIfb(SMLoc DirectiveLoc, bool ExpectBlank) { |
6307 | if (TheCondState.TheCond != AsmCond::IfCond && |
6308 | TheCondState.TheCond != AsmCond::ElseIfCond) |
6309 | return Error(L: DirectiveLoc, Msg: "Encountered an elseif that doesn't follow an" |
6310 | " if or an elseif" ); |
6311 | TheCondState.TheCond = AsmCond::ElseIfCond; |
6312 | |
6313 | bool LastIgnoreState = false; |
6314 | if (!TheCondStack.empty()) |
6315 | LastIgnoreState = TheCondStack.back().Ignore; |
6316 | if (LastIgnoreState || TheCondState.CondMet) { |
6317 | TheCondState.Ignore = true; |
6318 | eatToEndOfStatement(); |
6319 | } else { |
6320 | std::string Str; |
6321 | if (parseTextItem(Data&: Str)) { |
6322 | if (ExpectBlank) |
6323 | return TokError(Msg: "expected text item parameter for 'elseifb' directive" ); |
6324 | return TokError(Msg: "expected text item parameter for 'elseifnb' directive" ); |
6325 | } |
6326 | |
6327 | if (parseEOL()) |
6328 | return true; |
6329 | |
6330 | TheCondState.CondMet = ExpectBlank == Str.empty(); |
6331 | TheCondState.Ignore = !TheCondState.CondMet; |
6332 | } |
6333 | |
6334 | return false; |
6335 | } |
6336 | |
6337 | /// parseDirectiveElseIfdef |
6338 | /// ::= elseifdef symbol |
6339 | /// | elseifdef variable |
6340 | bool MasmParser::parseDirectiveElseIfdef(SMLoc DirectiveLoc, |
6341 | bool expect_defined) { |
6342 | if (TheCondState.TheCond != AsmCond::IfCond && |
6343 | TheCondState.TheCond != AsmCond::ElseIfCond) |
6344 | return Error(L: DirectiveLoc, Msg: "Encountered an elseif that doesn't follow an" |
6345 | " if or an elseif" ); |
6346 | TheCondState.TheCond = AsmCond::ElseIfCond; |
6347 | |
6348 | bool LastIgnoreState = false; |
6349 | if (!TheCondStack.empty()) |
6350 | LastIgnoreState = TheCondStack.back().Ignore; |
6351 | if (LastIgnoreState || TheCondState.CondMet) { |
6352 | TheCondState.Ignore = true; |
6353 | eatToEndOfStatement(); |
6354 | } else { |
6355 | bool is_defined = false; |
6356 | MCRegister Reg; |
6357 | SMLoc StartLoc, EndLoc; |
6358 | is_defined = |
6359 | getTargetParser().tryParseRegister(Reg, StartLoc, EndLoc).isSuccess(); |
6360 | if (!is_defined) { |
6361 | StringRef Name; |
6362 | if (check(P: parseIdentifier(Res&: Name), |
6363 | Msg: "expected identifier after 'elseifdef'" ) || |
6364 | parseEOL()) |
6365 | return true; |
6366 | |
6367 | if (BuiltinSymbolMap.contains(Key: Name.lower())) { |
6368 | is_defined = true; |
6369 | } else if (Variables.contains(Key: Name.lower())) { |
6370 | is_defined = true; |
6371 | } else { |
6372 | MCSymbol *Sym = getContext().lookupSymbol(Name); |
6373 | is_defined = (Sym && !Sym->isUndefined(SetUsed: false)); |
6374 | } |
6375 | } |
6376 | |
6377 | TheCondState.CondMet = (is_defined == expect_defined); |
6378 | TheCondState.Ignore = !TheCondState.CondMet; |
6379 | } |
6380 | |
6381 | return false; |
6382 | } |
6383 | |
6384 | /// parseDirectiveElseIfidn |
6385 | /// ::= elseifidn textitem, textitem |
6386 | bool MasmParser::parseDirectiveElseIfidn(SMLoc DirectiveLoc, bool ExpectEqual, |
6387 | bool CaseInsensitive) { |
6388 | if (TheCondState.TheCond != AsmCond::IfCond && |
6389 | TheCondState.TheCond != AsmCond::ElseIfCond) |
6390 | return Error(L: DirectiveLoc, Msg: "Encountered an elseif that doesn't follow an" |
6391 | " if or an elseif" ); |
6392 | TheCondState.TheCond = AsmCond::ElseIfCond; |
6393 | |
6394 | bool LastIgnoreState = false; |
6395 | if (!TheCondStack.empty()) |
6396 | LastIgnoreState = TheCondStack.back().Ignore; |
6397 | if (LastIgnoreState || TheCondState.CondMet) { |
6398 | TheCondState.Ignore = true; |
6399 | eatToEndOfStatement(); |
6400 | } else { |
6401 | std::string String1, String2; |
6402 | |
6403 | if (parseTextItem(Data&: String1)) { |
6404 | if (ExpectEqual) |
6405 | return TokError( |
6406 | Msg: "expected text item parameter for 'elseifidn' directive" ); |
6407 | return TokError(Msg: "expected text item parameter for 'elseifdif' directive" ); |
6408 | } |
6409 | |
6410 | if (Lexer.isNot(K: AsmToken::Comma)) { |
6411 | if (ExpectEqual) |
6412 | return TokError( |
6413 | Msg: "expected comma after first string for 'elseifidn' directive" ); |
6414 | return TokError( |
6415 | Msg: "expected comma after first string for 'elseifdif' directive" ); |
6416 | } |
6417 | Lex(); |
6418 | |
6419 | if (parseTextItem(Data&: String2)) { |
6420 | if (ExpectEqual) |
6421 | return TokError( |
6422 | Msg: "expected text item parameter for 'elseifidn' directive" ); |
6423 | return TokError(Msg: "expected text item parameter for 'elseifdif' directive" ); |
6424 | } |
6425 | |
6426 | if (CaseInsensitive) |
6427 | TheCondState.CondMet = |
6428 | ExpectEqual == (StringRef(String1).equals_insensitive(RHS: String2)); |
6429 | else |
6430 | TheCondState.CondMet = ExpectEqual == (String1 == String2); |
6431 | TheCondState.Ignore = !TheCondState.CondMet; |
6432 | } |
6433 | |
6434 | return false; |
6435 | } |
6436 | |
6437 | /// parseDirectiveElse |
6438 | /// ::= else |
6439 | bool MasmParser::parseDirectiveElse(SMLoc DirectiveLoc) { |
6440 | if (parseEOL()) |
6441 | return true; |
6442 | |
6443 | if (TheCondState.TheCond != AsmCond::IfCond && |
6444 | TheCondState.TheCond != AsmCond::ElseIfCond) |
6445 | return Error(L: DirectiveLoc, Msg: "Encountered an else that doesn't follow an if" |
6446 | " or an elseif" ); |
6447 | TheCondState.TheCond = AsmCond::ElseCond; |
6448 | bool LastIgnoreState = false; |
6449 | if (!TheCondStack.empty()) |
6450 | LastIgnoreState = TheCondStack.back().Ignore; |
6451 | if (LastIgnoreState || TheCondState.CondMet) |
6452 | TheCondState.Ignore = true; |
6453 | else |
6454 | TheCondState.Ignore = false; |
6455 | |
6456 | return false; |
6457 | } |
6458 | |
6459 | /// parseDirectiveEnd |
6460 | /// ::= end |
6461 | bool MasmParser::parseDirectiveEnd(SMLoc DirectiveLoc) { |
6462 | if (parseEOL()) |
6463 | return true; |
6464 | |
6465 | while (Lexer.isNot(K: AsmToken::Eof)) |
6466 | Lexer.Lex(); |
6467 | |
6468 | return false; |
6469 | } |
6470 | |
6471 | /// parseDirectiveError |
6472 | /// ::= .err [message] |
6473 | bool MasmParser::parseDirectiveError(SMLoc DirectiveLoc) { |
6474 | if (!TheCondStack.empty()) { |
6475 | if (TheCondStack.back().Ignore) { |
6476 | eatToEndOfStatement(); |
6477 | return false; |
6478 | } |
6479 | } |
6480 | |
6481 | std::string Message = ".err directive invoked in source file" ; |
6482 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) |
6483 | Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6484 | Lex(); |
6485 | |
6486 | return Error(L: DirectiveLoc, Msg: Message); |
6487 | } |
6488 | |
6489 | /// parseDirectiveErrorIfb |
6490 | /// ::= .errb textitem[, message] |
6491 | bool MasmParser::parseDirectiveErrorIfb(SMLoc DirectiveLoc, bool ExpectBlank) { |
6492 | if (!TheCondStack.empty()) { |
6493 | if (TheCondStack.back().Ignore) { |
6494 | eatToEndOfStatement(); |
6495 | return false; |
6496 | } |
6497 | } |
6498 | |
6499 | std::string Text; |
6500 | if (parseTextItem(Data&: Text)) |
6501 | return Error(L: getTok().getLoc(), Msg: "missing text item in '.errb' directive" ); |
6502 | |
6503 | std::string Message = ".errb directive invoked in source file" ; |
6504 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
6505 | if (parseToken(T: AsmToken::Comma)) |
6506 | return addErrorSuffix(Suffix: " in '.errb' directive" ); |
6507 | Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6508 | } |
6509 | Lex(); |
6510 | |
6511 | if (Text.empty() == ExpectBlank) |
6512 | return Error(L: DirectiveLoc, Msg: Message); |
6513 | return false; |
6514 | } |
6515 | |
6516 | /// parseDirectiveErrorIfdef |
6517 | /// ::= .errdef name[, message] |
6518 | bool MasmParser::parseDirectiveErrorIfdef(SMLoc DirectiveLoc, |
6519 | bool ExpectDefined) { |
6520 | if (!TheCondStack.empty()) { |
6521 | if (TheCondStack.back().Ignore) { |
6522 | eatToEndOfStatement(); |
6523 | return false; |
6524 | } |
6525 | } |
6526 | |
6527 | bool IsDefined = false; |
6528 | MCRegister Reg; |
6529 | SMLoc StartLoc, EndLoc; |
6530 | IsDefined = |
6531 | getTargetParser().tryParseRegister(Reg, StartLoc, EndLoc).isSuccess(); |
6532 | if (!IsDefined) { |
6533 | StringRef Name; |
6534 | if (check(P: parseIdentifier(Res&: Name), Msg: "expected identifier after '.errdef'" )) |
6535 | return true; |
6536 | |
6537 | if (BuiltinSymbolMap.contains(Key: Name.lower())) { |
6538 | IsDefined = true; |
6539 | } else if (Variables.contains(Key: Name.lower())) { |
6540 | IsDefined = true; |
6541 | } else { |
6542 | MCSymbol *Sym = getContext().lookupSymbol(Name); |
6543 | IsDefined = (Sym && !Sym->isUndefined(SetUsed: false)); |
6544 | } |
6545 | } |
6546 | |
6547 | std::string Message = ".errdef directive invoked in source file" ; |
6548 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
6549 | if (parseToken(T: AsmToken::Comma)) |
6550 | return addErrorSuffix(Suffix: " in '.errdef' directive" ); |
6551 | Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6552 | } |
6553 | Lex(); |
6554 | |
6555 | if (IsDefined == ExpectDefined) |
6556 | return Error(L: DirectiveLoc, Msg: Message); |
6557 | return false; |
6558 | } |
6559 | |
6560 | /// parseDirectiveErrorIfidn |
6561 | /// ::= .erridn textitem, textitem[, message] |
6562 | bool MasmParser::parseDirectiveErrorIfidn(SMLoc DirectiveLoc, bool ExpectEqual, |
6563 | bool CaseInsensitive) { |
6564 | if (!TheCondStack.empty()) { |
6565 | if (TheCondStack.back().Ignore) { |
6566 | eatToEndOfStatement(); |
6567 | return false; |
6568 | } |
6569 | } |
6570 | |
6571 | std::string String1, String2; |
6572 | |
6573 | if (parseTextItem(Data&: String1)) { |
6574 | if (ExpectEqual) |
6575 | return TokError(Msg: "expected string parameter for '.erridn' directive" ); |
6576 | return TokError(Msg: "expected string parameter for '.errdif' directive" ); |
6577 | } |
6578 | |
6579 | if (Lexer.isNot(K: AsmToken::Comma)) { |
6580 | if (ExpectEqual) |
6581 | return TokError( |
6582 | Msg: "expected comma after first string for '.erridn' directive" ); |
6583 | return TokError( |
6584 | Msg: "expected comma after first string for '.errdif' directive" ); |
6585 | } |
6586 | Lex(); |
6587 | |
6588 | if (parseTextItem(Data&: String2)) { |
6589 | if (ExpectEqual) |
6590 | return TokError(Msg: "expected string parameter for '.erridn' directive" ); |
6591 | return TokError(Msg: "expected string parameter for '.errdif' directive" ); |
6592 | } |
6593 | |
6594 | std::string Message; |
6595 | if (ExpectEqual) |
6596 | Message = ".erridn directive invoked in source file" ; |
6597 | else |
6598 | Message = ".errdif directive invoked in source file" ; |
6599 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
6600 | if (parseToken(T: AsmToken::Comma)) |
6601 | return addErrorSuffix(Suffix: " in '.erridn' directive" ); |
6602 | Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6603 | } |
6604 | Lex(); |
6605 | |
6606 | if (CaseInsensitive) |
6607 | TheCondState.CondMet = |
6608 | ExpectEqual == (StringRef(String1).equals_insensitive(RHS: String2)); |
6609 | else |
6610 | TheCondState.CondMet = ExpectEqual == (String1 == String2); |
6611 | TheCondState.Ignore = !TheCondState.CondMet; |
6612 | |
6613 | if ((CaseInsensitive && |
6614 | ExpectEqual == StringRef(String1).equals_insensitive(RHS: String2)) || |
6615 | (ExpectEqual == (String1 == String2))) |
6616 | return Error(L: DirectiveLoc, Msg: Message); |
6617 | return false; |
6618 | } |
6619 | |
6620 | /// parseDirectiveErrorIfe |
6621 | /// ::= .erre expression[, message] |
6622 | bool MasmParser::parseDirectiveErrorIfe(SMLoc DirectiveLoc, bool ExpectZero) { |
6623 | if (!TheCondStack.empty()) { |
6624 | if (TheCondStack.back().Ignore) { |
6625 | eatToEndOfStatement(); |
6626 | return false; |
6627 | } |
6628 | } |
6629 | |
6630 | int64_t ExprValue; |
6631 | if (parseAbsoluteExpression(Res&: ExprValue)) |
6632 | return addErrorSuffix(Suffix: " in '.erre' directive" ); |
6633 | |
6634 | std::string Message = ".erre directive invoked in source file" ; |
6635 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
6636 | if (parseToken(T: AsmToken::Comma)) |
6637 | return addErrorSuffix(Suffix: " in '.erre' directive" ); |
6638 | Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6639 | } |
6640 | Lex(); |
6641 | |
6642 | if ((ExprValue == 0) == ExpectZero) |
6643 | return Error(L: DirectiveLoc, Msg: Message); |
6644 | return false; |
6645 | } |
6646 | |
6647 | /// parseDirectiveEndIf |
6648 | /// ::= .endif |
6649 | bool MasmParser::parseDirectiveEndIf(SMLoc DirectiveLoc) { |
6650 | if (parseEOL()) |
6651 | return true; |
6652 | |
6653 | if ((TheCondState.TheCond == AsmCond::NoCond) || TheCondStack.empty()) |
6654 | return Error(L: DirectiveLoc, Msg: "Encountered a .endif that doesn't follow " |
6655 | "an .if or .else" ); |
6656 | if (!TheCondStack.empty()) { |
6657 | TheCondState = TheCondStack.back(); |
6658 | TheCondStack.pop_back(); |
6659 | } |
6660 | |
6661 | return false; |
6662 | } |
6663 | |
6664 | void MasmParser::initializeDirectiveKindMap() { |
6665 | DirectiveKindMap["=" ] = DK_ASSIGN; |
6666 | DirectiveKindMap["equ" ] = DK_EQU; |
6667 | DirectiveKindMap["textequ" ] = DK_TEXTEQU; |
6668 | // DirectiveKindMap[".ascii"] = DK_ASCII; |
6669 | // DirectiveKindMap[".asciz"] = DK_ASCIZ; |
6670 | // DirectiveKindMap[".string"] = DK_STRING; |
6671 | DirectiveKindMap["byte" ] = DK_BYTE; |
6672 | DirectiveKindMap["sbyte" ] = DK_SBYTE; |
6673 | DirectiveKindMap["word" ] = DK_WORD; |
6674 | DirectiveKindMap["sword" ] = DK_SWORD; |
6675 | DirectiveKindMap["dword" ] = DK_DWORD; |
6676 | DirectiveKindMap["sdword" ] = DK_SDWORD; |
6677 | DirectiveKindMap["fword" ] = DK_FWORD; |
6678 | DirectiveKindMap["qword" ] = DK_QWORD; |
6679 | DirectiveKindMap["sqword" ] = DK_SQWORD; |
6680 | DirectiveKindMap["real4" ] = DK_REAL4; |
6681 | DirectiveKindMap["real8" ] = DK_REAL8; |
6682 | DirectiveKindMap["real10" ] = DK_REAL10; |
6683 | DirectiveKindMap["align" ] = DK_ALIGN; |
6684 | DirectiveKindMap["even" ] = DK_EVEN; |
6685 | DirectiveKindMap["org" ] = DK_ORG; |
6686 | DirectiveKindMap["extern" ] = DK_EXTERN; |
6687 | DirectiveKindMap["extrn" ] = DK_EXTERN; |
6688 | DirectiveKindMap["public" ] = DK_PUBLIC; |
6689 | // DirectiveKindMap[".comm"] = DK_COMM; |
6690 | DirectiveKindMap["comment" ] = DK_COMMENT; |
6691 | DirectiveKindMap["include" ] = DK_INCLUDE; |
6692 | DirectiveKindMap["repeat" ] = DK_REPEAT; |
6693 | DirectiveKindMap["rept" ] = DK_REPEAT; |
6694 | DirectiveKindMap["while" ] = DK_WHILE; |
6695 | DirectiveKindMap["for" ] = DK_FOR; |
6696 | DirectiveKindMap["irp" ] = DK_FOR; |
6697 | DirectiveKindMap["forc" ] = DK_FORC; |
6698 | DirectiveKindMap["irpc" ] = DK_FORC; |
6699 | DirectiveKindMap["if" ] = DK_IF; |
6700 | DirectiveKindMap["ife" ] = DK_IFE; |
6701 | DirectiveKindMap["ifb" ] = DK_IFB; |
6702 | DirectiveKindMap["ifnb" ] = DK_IFNB; |
6703 | DirectiveKindMap["ifdef" ] = DK_IFDEF; |
6704 | DirectiveKindMap["ifndef" ] = DK_IFNDEF; |
6705 | DirectiveKindMap["ifdif" ] = DK_IFDIF; |
6706 | DirectiveKindMap["ifdifi" ] = DK_IFDIFI; |
6707 | DirectiveKindMap["ifidn" ] = DK_IFIDN; |
6708 | DirectiveKindMap["ifidni" ] = DK_IFIDNI; |
6709 | DirectiveKindMap["elseif" ] = DK_ELSEIF; |
6710 | DirectiveKindMap["elseifdef" ] = DK_ELSEIFDEF; |
6711 | DirectiveKindMap["elseifndef" ] = DK_ELSEIFNDEF; |
6712 | DirectiveKindMap["elseifdif" ] = DK_ELSEIFDIF; |
6713 | DirectiveKindMap["elseifidn" ] = DK_ELSEIFIDN; |
6714 | DirectiveKindMap["else" ] = DK_ELSE; |
6715 | DirectiveKindMap["end" ] = DK_END; |
6716 | DirectiveKindMap["endif" ] = DK_ENDIF; |
6717 | // DirectiveKindMap[".file"] = DK_FILE; |
6718 | // DirectiveKindMap[".line"] = DK_LINE; |
6719 | // DirectiveKindMap[".loc"] = DK_LOC; |
6720 | // DirectiveKindMap[".stabs"] = DK_STABS; |
6721 | // DirectiveKindMap[".cv_file"] = DK_CV_FILE; |
6722 | // DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID; |
6723 | // DirectiveKindMap[".cv_loc"] = DK_CV_LOC; |
6724 | // DirectiveKindMap[".cv_linetable"] = DK_CV_LINETABLE; |
6725 | // DirectiveKindMap[".cv_inline_linetable"] = DK_CV_INLINE_LINETABLE; |
6726 | // DirectiveKindMap[".cv_inline_site_id"] = DK_CV_INLINE_SITE_ID; |
6727 | // DirectiveKindMap[".cv_def_range"] = DK_CV_DEF_RANGE; |
6728 | // DirectiveKindMap[".cv_string"] = DK_CV_STRING; |
6729 | // DirectiveKindMap[".cv_stringtable"] = DK_CV_STRINGTABLE; |
6730 | // DirectiveKindMap[".cv_filechecksums"] = DK_CV_FILECHECKSUMS; |
6731 | // DirectiveKindMap[".cv_filechecksumoffset"] = DK_CV_FILECHECKSUM_OFFSET; |
6732 | // DirectiveKindMap[".cv_fpo_data"] = DK_CV_FPO_DATA; |
6733 | // DirectiveKindMap[".cfi_sections"] = DK_CFI_SECTIONS; |
6734 | // DirectiveKindMap[".cfi_startproc"] = DK_CFI_STARTPROC; |
6735 | // DirectiveKindMap[".cfi_endproc"] = DK_CFI_ENDPROC; |
6736 | // DirectiveKindMap[".cfi_def_cfa"] = DK_CFI_DEF_CFA; |
6737 | // DirectiveKindMap[".cfi_def_cfa_offset"] = DK_CFI_DEF_CFA_OFFSET; |
6738 | // DirectiveKindMap[".cfi_adjust_cfa_offset"] = DK_CFI_ADJUST_CFA_OFFSET; |
6739 | // DirectiveKindMap[".cfi_def_cfa_register"] = DK_CFI_DEF_CFA_REGISTER; |
6740 | // DirectiveKindMap[".cfi_offset"] = DK_CFI_OFFSET; |
6741 | // DirectiveKindMap[".cfi_rel_offset"] = DK_CFI_REL_OFFSET; |
6742 | // DirectiveKindMap[".cfi_personality"] = DK_CFI_PERSONALITY; |
6743 | // DirectiveKindMap[".cfi_lsda"] = DK_CFI_LSDA; |
6744 | // DirectiveKindMap[".cfi_remember_state"] = DK_CFI_REMEMBER_STATE; |
6745 | // DirectiveKindMap[".cfi_restore_state"] = DK_CFI_RESTORE_STATE; |
6746 | // DirectiveKindMap[".cfi_same_value"] = DK_CFI_SAME_VALUE; |
6747 | // DirectiveKindMap[".cfi_restore"] = DK_CFI_RESTORE; |
6748 | // DirectiveKindMap[".cfi_escape"] = DK_CFI_ESCAPE; |
6749 | // DirectiveKindMap[".cfi_return_column"] = DK_CFI_RETURN_COLUMN; |
6750 | // DirectiveKindMap[".cfi_signal_frame"] = DK_CFI_SIGNAL_FRAME; |
6751 | // DirectiveKindMap[".cfi_undefined"] = DK_CFI_UNDEFINED; |
6752 | // DirectiveKindMap[".cfi_register"] = DK_CFI_REGISTER; |
6753 | // DirectiveKindMap[".cfi_window_save"] = DK_CFI_WINDOW_SAVE; |
6754 | // DirectiveKindMap[".cfi_b_key_frame"] = DK_CFI_B_KEY_FRAME; |
6755 | DirectiveKindMap["macro" ] = DK_MACRO; |
6756 | DirectiveKindMap["exitm" ] = DK_EXITM; |
6757 | DirectiveKindMap["endm" ] = DK_ENDM; |
6758 | DirectiveKindMap["purge" ] = DK_PURGE; |
6759 | DirectiveKindMap[".err" ] = DK_ERR; |
6760 | DirectiveKindMap[".errb" ] = DK_ERRB; |
6761 | DirectiveKindMap[".errnb" ] = DK_ERRNB; |
6762 | DirectiveKindMap[".errdef" ] = DK_ERRDEF; |
6763 | DirectiveKindMap[".errndef" ] = DK_ERRNDEF; |
6764 | DirectiveKindMap[".errdif" ] = DK_ERRDIF; |
6765 | DirectiveKindMap[".errdifi" ] = DK_ERRDIFI; |
6766 | DirectiveKindMap[".erridn" ] = DK_ERRIDN; |
6767 | DirectiveKindMap[".erridni" ] = DK_ERRIDNI; |
6768 | DirectiveKindMap[".erre" ] = DK_ERRE; |
6769 | DirectiveKindMap[".errnz" ] = DK_ERRNZ; |
6770 | DirectiveKindMap[".pushframe" ] = DK_PUSHFRAME; |
6771 | DirectiveKindMap[".pushreg" ] = DK_PUSHREG; |
6772 | DirectiveKindMap[".savereg" ] = DK_SAVEREG; |
6773 | DirectiveKindMap[".savexmm128" ] = DK_SAVEXMM128; |
6774 | DirectiveKindMap[".setframe" ] = DK_SETFRAME; |
6775 | DirectiveKindMap[".radix" ] = DK_RADIX; |
6776 | DirectiveKindMap["db" ] = DK_DB; |
6777 | DirectiveKindMap["dd" ] = DK_DD; |
6778 | DirectiveKindMap["df" ] = DK_DF; |
6779 | DirectiveKindMap["dq" ] = DK_DQ; |
6780 | DirectiveKindMap["dw" ] = DK_DW; |
6781 | DirectiveKindMap["echo" ] = DK_ECHO; |
6782 | DirectiveKindMap["struc" ] = DK_STRUCT; |
6783 | DirectiveKindMap["struct" ] = DK_STRUCT; |
6784 | DirectiveKindMap["union" ] = DK_UNION; |
6785 | DirectiveKindMap["ends" ] = DK_ENDS; |
6786 | } |
6787 | |
6788 | bool MasmParser::isMacroLikeDirective() { |
6789 | if (getLexer().is(K: AsmToken::Identifier)) { |
6790 | bool IsMacroLike = StringSwitch<bool>(getTok().getIdentifier()) |
6791 | .CasesLower(S0: "repeat" , S1: "rept" , Value: true) |
6792 | .CaseLower(S: "while" , Value: true) |
6793 | .CasesLower(S0: "for" , S1: "irp" , Value: true) |
6794 | .CasesLower(S0: "forc" , S1: "irpc" , Value: true) |
6795 | .Default(Value: false); |
6796 | if (IsMacroLike) |
6797 | return true; |
6798 | } |
6799 | if (peekTok().is(K: AsmToken::Identifier) && |
6800 | peekTok().getIdentifier().equals_insensitive(RHS: "macro" )) |
6801 | return true; |
6802 | |
6803 | return false; |
6804 | } |
6805 | |
6806 | MCAsmMacro *MasmParser::parseMacroLikeBody(SMLoc DirectiveLoc) { |
6807 | AsmToken EndToken, StartToken = getTok(); |
6808 | |
6809 | unsigned NestLevel = 0; |
6810 | while (true) { |
6811 | // Check whether we have reached the end of the file. |
6812 | if (getLexer().is(K: AsmToken::Eof)) { |
6813 | printError(L: DirectiveLoc, Msg: "no matching 'endm' in definition" ); |
6814 | return nullptr; |
6815 | } |
6816 | |
6817 | if (isMacroLikeDirective()) |
6818 | ++NestLevel; |
6819 | |
6820 | // Otherwise, check whether we have reached the endm. |
6821 | if (Lexer.is(K: AsmToken::Identifier) && |
6822 | getTok().getIdentifier().equals_insensitive(RHS: "endm" )) { |
6823 | if (NestLevel == 0) { |
6824 | EndToken = getTok(); |
6825 | Lex(); |
6826 | if (Lexer.isNot(K: AsmToken::EndOfStatement)) { |
6827 | printError(L: getTok().getLoc(), Msg: "unexpected token in 'endm' directive" ); |
6828 | return nullptr; |
6829 | } |
6830 | break; |
6831 | } |
6832 | --NestLevel; |
6833 | } |
6834 | |
6835 | // Otherwise, scan till the end of the statement. |
6836 | eatToEndOfStatement(); |
6837 | } |
6838 | |
6839 | const char *BodyStart = StartToken.getLoc().getPointer(); |
6840 | const char *BodyEnd = EndToken.getLoc().getPointer(); |
6841 | StringRef Body = StringRef(BodyStart, BodyEnd - BodyStart); |
6842 | |
6843 | // We Are Anonymous. |
6844 | MacroLikeBodies.emplace_back(args: StringRef(), args&: Body, args: MCAsmMacroParameters()); |
6845 | return &MacroLikeBodies.back(); |
6846 | } |
6847 | |
6848 | bool MasmParser::expandStatement(SMLoc Loc) { |
6849 | std::string Body = parseStringTo(EndTok: AsmToken::EndOfStatement); |
6850 | SMLoc EndLoc = getTok().getLoc(); |
6851 | |
6852 | MCAsmMacroParameters Parameters; |
6853 | MCAsmMacroArguments Arguments; |
6854 | |
6855 | StringMap<std::string> BuiltinValues; |
6856 | for (const auto &S : BuiltinSymbolMap) { |
6857 | const BuiltinSymbol &Sym = S.getValue(); |
6858 | if (std::optional<std::string> Text = evaluateBuiltinTextMacro(Symbol: Sym, StartLoc: Loc)) { |
6859 | BuiltinValues[S.getKey().lower()] = std::move(*Text); |
6860 | } |
6861 | } |
6862 | for (const auto &B : BuiltinValues) { |
6863 | MCAsmMacroParameter P; |
6864 | MCAsmMacroArgument A; |
6865 | P.Name = B.getKey(); |
6866 | P.Required = true; |
6867 | A.push_back(x: AsmToken(AsmToken::String, B.getValue())); |
6868 | |
6869 | Parameters.push_back(x: std::move(P)); |
6870 | Arguments.push_back(x: std::move(A)); |
6871 | } |
6872 | |
6873 | for (const auto &V : Variables) { |
6874 | const Variable &Var = V.getValue(); |
6875 | if (Var.IsText) { |
6876 | MCAsmMacroParameter P; |
6877 | MCAsmMacroArgument A; |
6878 | P.Name = Var.Name; |
6879 | P.Required = true; |
6880 | A.push_back(x: AsmToken(AsmToken::String, Var.TextValue)); |
6881 | |
6882 | Parameters.push_back(x: std::move(P)); |
6883 | Arguments.push_back(x: std::move(A)); |
6884 | } |
6885 | } |
6886 | MacroLikeBodies.emplace_back(args: StringRef(), args&: Body, args&: Parameters); |
6887 | MCAsmMacro M = MacroLikeBodies.back(); |
6888 | |
6889 | // Expand the statement in a new buffer. |
6890 | SmallString<80> Buf; |
6891 | raw_svector_ostream OS(Buf); |
6892 | if (expandMacro(OS, Body: M.Body, Parameters: M.Parameters, A: Arguments, Locals: M.Locals, L: EndLoc)) |
6893 | return true; |
6894 | std::unique_ptr<MemoryBuffer> Expansion = |
6895 | MemoryBuffer::getMemBufferCopy(InputData: OS.str(), BufferName: "<expansion>" ); |
6896 | |
6897 | // Jump to the expanded statement and prime the lexer. |
6898 | CurBuffer = SrcMgr.AddNewSourceBuffer(F: std::move(Expansion), IncludeLoc: EndLoc); |
6899 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer()); |
6900 | EndStatementAtEOFStack.push_back(Val: false); |
6901 | Lex(); |
6902 | return false; |
6903 | } |
6904 | |
6905 | void MasmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, |
6906 | raw_svector_ostream &OS) { |
6907 | instantiateMacroLikeBody(M, DirectiveLoc, /*ExitLoc=*/getTok().getLoc(), OS); |
6908 | } |
6909 | void MasmParser::instantiateMacroLikeBody(MCAsmMacro *M, SMLoc DirectiveLoc, |
6910 | SMLoc ExitLoc, |
6911 | raw_svector_ostream &OS) { |
6912 | OS << "endm\n" ; |
6913 | |
6914 | std::unique_ptr<MemoryBuffer> Instantiation = |
6915 | MemoryBuffer::getMemBufferCopy(InputData: OS.str(), BufferName: "<instantiation>" ); |
6916 | |
6917 | // Create the macro instantiation object and add to the current macro |
6918 | // instantiation stack. |
6919 | MacroInstantiation *MI = new MacroInstantiation{.InstantiationLoc: DirectiveLoc, .ExitBuffer: CurBuffer, |
6920 | .ExitLoc: ExitLoc, .CondStackDepth: TheCondStack.size()}; |
6921 | ActiveMacros.push_back(x: MI); |
6922 | |
6923 | // Jump to the macro instantiation and prime the lexer. |
6924 | CurBuffer = SrcMgr.AddNewSourceBuffer(F: std::move(Instantiation), IncludeLoc: SMLoc()); |
6925 | Lexer.setBuffer(Buf: SrcMgr.getMemoryBuffer(i: CurBuffer)->getBuffer()); |
6926 | EndStatementAtEOFStack.push_back(Val: true); |
6927 | Lex(); |
6928 | } |
6929 | |
6930 | /// parseDirectiveRepeat |
6931 | /// ::= ("repeat" | "rept") count |
6932 | /// body |
6933 | /// endm |
6934 | bool MasmParser::parseDirectiveRepeat(SMLoc DirectiveLoc, StringRef Dir) { |
6935 | const MCExpr *CountExpr; |
6936 | SMLoc CountLoc = getTok().getLoc(); |
6937 | if (parseExpression(Res&: CountExpr)) |
6938 | return true; |
6939 | |
6940 | int64_t Count; |
6941 | if (!CountExpr->evaluateAsAbsolute(Res&: Count, Asm: getStreamer().getAssemblerPtr())) { |
6942 | return Error(L: CountLoc, Msg: "unexpected token in '" + Dir + "' directive" ); |
6943 | } |
6944 | |
6945 | if (check(P: Count < 0, Loc: CountLoc, Msg: "Count is negative" ) || parseEOL()) |
6946 | return true; |
6947 | |
6948 | // Lex the repeat definition. |
6949 | MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); |
6950 | if (!M) |
6951 | return true; |
6952 | |
6953 | // Macro instantiation is lexical, unfortunately. We construct a new buffer |
6954 | // to hold the macro body with substitutions. |
6955 | SmallString<256> Buf; |
6956 | raw_svector_ostream OS(Buf); |
6957 | while (Count--) { |
6958 | if (expandMacro(OS, Body: M->Body, Parameters: std::nullopt, A: std::nullopt, Locals: M->Locals, |
6959 | L: getTok().getLoc())) |
6960 | return true; |
6961 | } |
6962 | instantiateMacroLikeBody(M, DirectiveLoc, OS); |
6963 | |
6964 | return false; |
6965 | } |
6966 | |
6967 | /// parseDirectiveWhile |
6968 | /// ::= "while" expression |
6969 | /// body |
6970 | /// endm |
6971 | bool MasmParser::parseDirectiveWhile(SMLoc DirectiveLoc) { |
6972 | const MCExpr *CondExpr; |
6973 | SMLoc CondLoc = getTok().getLoc(); |
6974 | if (parseExpression(Res&: CondExpr)) |
6975 | return true; |
6976 | |
6977 | // Lex the repeat definition. |
6978 | MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); |
6979 | if (!M) |
6980 | return true; |
6981 | |
6982 | // Macro instantiation is lexical, unfortunately. We construct a new buffer |
6983 | // to hold the macro body with substitutions. |
6984 | SmallString<256> Buf; |
6985 | raw_svector_ostream OS(Buf); |
6986 | int64_t Condition; |
6987 | if (!CondExpr->evaluateAsAbsolute(Res&: Condition, Asm: getStreamer().getAssemblerPtr())) |
6988 | return Error(L: CondLoc, Msg: "expected absolute expression in 'while' directive" ); |
6989 | if (Condition) { |
6990 | // Instantiate the macro, then resume at this directive to recheck the |
6991 | // condition. |
6992 | if (expandMacro(OS, Body: M->Body, Parameters: std::nullopt, A: std::nullopt, Locals: M->Locals, |
6993 | L: getTok().getLoc())) |
6994 | return true; |
6995 | instantiateMacroLikeBody(M, DirectiveLoc, /*ExitLoc=*/DirectiveLoc, OS); |
6996 | } |
6997 | |
6998 | return false; |
6999 | } |
7000 | |
7001 | /// parseDirectiveFor |
7002 | /// ::= ("for" | "irp") symbol [":" qualifier], <values> |
7003 | /// body |
7004 | /// endm |
7005 | bool MasmParser::parseDirectiveFor(SMLoc DirectiveLoc, StringRef Dir) { |
7006 | MCAsmMacroParameter Parameter; |
7007 | MCAsmMacroArguments A; |
7008 | if (check(P: parseIdentifier(Res&: Parameter.Name), |
7009 | Msg: "expected identifier in '" + Dir + "' directive" )) |
7010 | return true; |
7011 | |
7012 | // Parse optional qualifier (default value, or "req") |
7013 | if (parseOptionalToken(T: AsmToken::Colon)) { |
7014 | if (parseOptionalToken(T: AsmToken::Equal)) { |
7015 | // Default value |
7016 | SMLoc ParamLoc; |
7017 | |
7018 | ParamLoc = Lexer.getLoc(); |
7019 | if (parseMacroArgument(MP: nullptr, MA&: Parameter.Value)) |
7020 | return true; |
7021 | } else { |
7022 | SMLoc QualLoc; |
7023 | StringRef Qualifier; |
7024 | |
7025 | QualLoc = Lexer.getLoc(); |
7026 | if (parseIdentifier(Res&: Qualifier)) |
7027 | return Error(L: QualLoc, Msg: "missing parameter qualifier for " |
7028 | "'" + |
7029 | Parameter.Name + "' in '" + Dir + |
7030 | "' directive" ); |
7031 | |
7032 | if (Qualifier.equals_insensitive(RHS: "req" )) |
7033 | Parameter.Required = true; |
7034 | else |
7035 | return Error(L: QualLoc, |
7036 | Msg: Qualifier + " is not a valid parameter qualifier for '" + |
7037 | Parameter.Name + "' in '" + Dir + "' directive" ); |
7038 | } |
7039 | } |
7040 | |
7041 | if (parseToken(T: AsmToken::Comma, |
7042 | Msg: "expected comma in '" + Dir + "' directive" ) || |
7043 | parseToken(T: AsmToken::Less, |
7044 | Msg: "values in '" + Dir + |
7045 | "' directive must be enclosed in angle brackets" )) |
7046 | return true; |
7047 | |
7048 | while (true) { |
7049 | A.emplace_back(); |
7050 | if (parseMacroArgument(MP: &Parameter, MA&: A.back(), /*EndTok=*/AsmToken::Greater)) |
7051 | return addErrorSuffix(Suffix: " in arguments for '" + Dir + "' directive" ); |
7052 | |
7053 | // If we see a comma, continue, and allow line continuation. |
7054 | if (!parseOptionalToken(T: AsmToken::Comma)) |
7055 | break; |
7056 | parseOptionalToken(T: AsmToken::EndOfStatement); |
7057 | } |
7058 | |
7059 | if (parseToken(T: AsmToken::Greater, |
7060 | Msg: "values in '" + Dir + |
7061 | "' directive must be enclosed in angle brackets" ) || |
7062 | parseEOL()) |
7063 | return true; |
7064 | |
7065 | // Lex the for definition. |
7066 | MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); |
7067 | if (!M) |
7068 | return true; |
7069 | |
7070 | // Macro instantiation is lexical, unfortunately. We construct a new buffer |
7071 | // to hold the macro body with substitutions. |
7072 | SmallString<256> Buf; |
7073 | raw_svector_ostream OS(Buf); |
7074 | |
7075 | for (const MCAsmMacroArgument &Arg : A) { |
7076 | if (expandMacro(OS, Body: M->Body, Parameters: Parameter, A: Arg, Locals: M->Locals, L: getTok().getLoc())) |
7077 | return true; |
7078 | } |
7079 | |
7080 | instantiateMacroLikeBody(M, DirectiveLoc, OS); |
7081 | |
7082 | return false; |
7083 | } |
7084 | |
7085 | /// parseDirectiveForc |
7086 | /// ::= ("forc" | "irpc") symbol, <string> |
7087 | /// body |
7088 | /// endm |
7089 | bool MasmParser::parseDirectiveForc(SMLoc DirectiveLoc, StringRef Directive) { |
7090 | MCAsmMacroParameter Parameter; |
7091 | |
7092 | std::string Argument; |
7093 | if (check(P: parseIdentifier(Res&: Parameter.Name), |
7094 | Msg: "expected identifier in '" + Directive + "' directive" ) || |
7095 | parseToken(T: AsmToken::Comma, |
7096 | Msg: "expected comma in '" + Directive + "' directive" )) |
7097 | return true; |
7098 | if (parseAngleBracketString(Data&: Argument)) { |
7099 | // Match ml64.exe; treat all characters to end of statement as a string, |
7100 | // ignoring comment markers, then discard anything following a space (using |
7101 | // the C locale). |
7102 | Argument = parseStringTo(EndTok: AsmToken::EndOfStatement); |
7103 | if (getTok().is(K: AsmToken::EndOfStatement)) |
7104 | Argument += getTok().getString(); |
7105 | size_t End = 0; |
7106 | for (; End < Argument.size(); ++End) { |
7107 | if (isSpace(C: Argument[End])) |
7108 | break; |
7109 | } |
7110 | Argument.resize(n: End); |
7111 | } |
7112 | if (parseEOL()) |
7113 | return true; |
7114 | |
7115 | // Lex the irpc definition. |
7116 | MCAsmMacro *M = parseMacroLikeBody(DirectiveLoc); |
7117 | if (!M) |
7118 | return true; |
7119 | |
7120 | // Macro instantiation is lexical, unfortunately. We construct a new buffer |
7121 | // to hold the macro body with substitutions. |
7122 | SmallString<256> Buf; |
7123 | raw_svector_ostream OS(Buf); |
7124 | |
7125 | StringRef Values(Argument); |
7126 | for (std::size_t I = 0, End = Values.size(); I != End; ++I) { |
7127 | MCAsmMacroArgument Arg; |
7128 | Arg.emplace_back(args: AsmToken::Identifier, args: Values.slice(Start: I, End: I + 1)); |
7129 | |
7130 | if (expandMacro(OS, Body: M->Body, Parameters: Parameter, A: Arg, Locals: M->Locals, L: getTok().getLoc())) |
7131 | return true; |
7132 | } |
7133 | |
7134 | instantiateMacroLikeBody(M, DirectiveLoc, OS); |
7135 | |
7136 | return false; |
7137 | } |
7138 | |
7139 | bool MasmParser::parseDirectiveMSEmit(SMLoc IDLoc, ParseStatementInfo &Info, |
7140 | size_t Len) { |
7141 | const MCExpr *Value; |
7142 | SMLoc ExprLoc = getLexer().getLoc(); |
7143 | if (parseExpression(Res&: Value)) |
7144 | return true; |
7145 | const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value); |
7146 | if (!MCE) |
7147 | return Error(L: ExprLoc, Msg: "unexpected expression in _emit" ); |
7148 | uint64_t IntValue = MCE->getValue(); |
7149 | if (!isUInt<8>(x: IntValue) && !isInt<8>(x: IntValue)) |
7150 | return Error(L: ExprLoc, Msg: "literal value out of range for directive" ); |
7151 | |
7152 | Info.AsmRewrites->emplace_back(Args: AOK_Emit, Args&: IDLoc, Args&: Len); |
7153 | return false; |
7154 | } |
7155 | |
7156 | bool MasmParser::parseDirectiveMSAlign(SMLoc IDLoc, ParseStatementInfo &Info) { |
7157 | const MCExpr *Value; |
7158 | SMLoc ExprLoc = getLexer().getLoc(); |
7159 | if (parseExpression(Res&: Value)) |
7160 | return true; |
7161 | const MCConstantExpr *MCE = dyn_cast<MCConstantExpr>(Val: Value); |
7162 | if (!MCE) |
7163 | return Error(L: ExprLoc, Msg: "unexpected expression in align" ); |
7164 | uint64_t IntValue = MCE->getValue(); |
7165 | if (!isPowerOf2_64(Value: IntValue)) |
7166 | return Error(L: ExprLoc, Msg: "literal value not a power of two greater then zero" ); |
7167 | |
7168 | Info.AsmRewrites->emplace_back(Args: AOK_Align, Args&: IDLoc, Args: 5, Args: Log2_64(Value: IntValue)); |
7169 | return false; |
7170 | } |
7171 | |
7172 | bool MasmParser::parseDirectiveRadix(SMLoc DirectiveLoc) { |
7173 | const SMLoc Loc = getLexer().getLoc(); |
7174 | std::string RadixStringRaw = parseStringTo(EndTok: AsmToken::EndOfStatement); |
7175 | StringRef RadixString = StringRef(RadixStringRaw).trim(); |
7176 | unsigned Radix; |
7177 | if (RadixString.getAsInteger(Radix: 10, Result&: Radix)) { |
7178 | return Error(L: Loc, |
7179 | Msg: "radix must be a decimal number in the range 2 to 16; was " + |
7180 | RadixString); |
7181 | } |
7182 | if (Radix < 2 || Radix > 16) |
7183 | return Error(L: Loc, Msg: "radix must be in the range 2 to 16; was " + |
7184 | std::to_string(val: Radix)); |
7185 | getLexer().setMasmDefaultRadix(Radix); |
7186 | return false; |
7187 | } |
7188 | |
7189 | /// parseDirectiveEcho |
7190 | /// ::= "echo" message |
7191 | bool MasmParser::parseDirectiveEcho(SMLoc DirectiveLoc) { |
7192 | std::string Message = parseStringTo(EndTok: AsmToken::EndOfStatement); |
7193 | llvm::outs() << Message; |
7194 | if (!StringRef(Message).ends_with(Suffix: "\n" )) |
7195 | llvm::outs() << '\n'; |
7196 | return false; |
7197 | } |
7198 | |
7199 | // We are comparing pointers, but the pointers are relative to a single string. |
7200 | // Thus, this should always be deterministic. |
7201 | static int rewritesSort(const AsmRewrite *AsmRewriteA, |
7202 | const AsmRewrite *AsmRewriteB) { |
7203 | if (AsmRewriteA->Loc.getPointer() < AsmRewriteB->Loc.getPointer()) |
7204 | return -1; |
7205 | if (AsmRewriteB->Loc.getPointer() < AsmRewriteA->Loc.getPointer()) |
7206 | return 1; |
7207 | |
7208 | // It's possible to have a SizeDirective, Imm/ImmPrefix and an Input/Output |
7209 | // rewrite to the same location. Make sure the SizeDirective rewrite is |
7210 | // performed first, then the Imm/ImmPrefix and finally the Input/Output. This |
7211 | // ensures the sort algorithm is stable. |
7212 | if (AsmRewritePrecedence[AsmRewriteA->Kind] > |
7213 | AsmRewritePrecedence[AsmRewriteB->Kind]) |
7214 | return -1; |
7215 | |
7216 | if (AsmRewritePrecedence[AsmRewriteA->Kind] < |
7217 | AsmRewritePrecedence[AsmRewriteB->Kind]) |
7218 | return 1; |
7219 | llvm_unreachable("Unstable rewrite sort." ); |
7220 | } |
7221 | |
7222 | bool MasmParser::defineMacro(StringRef Name, StringRef Value) { |
7223 | Variable &Var = Variables[Name.lower()]; |
7224 | if (Var.Name.empty()) { |
7225 | Var.Name = Name; |
7226 | } else if (Var.Redefinable == Variable::NOT_REDEFINABLE) { |
7227 | return Error(L: SMLoc(), Msg: "invalid variable redefinition" ); |
7228 | } else if (Var.Redefinable == Variable::WARN_ON_REDEFINITION && |
7229 | Warning(L: SMLoc(), Msg: "redefining '" + Name + |
7230 | "', already defined on the command line" )) { |
7231 | return true; |
7232 | } |
7233 | Var.Redefinable = Variable::WARN_ON_REDEFINITION; |
7234 | Var.IsText = true; |
7235 | Var.TextValue = Value.str(); |
7236 | return false; |
7237 | } |
7238 | |
7239 | bool MasmParser::lookUpField(StringRef Name, AsmFieldInfo &Info) const { |
7240 | const std::pair<StringRef, StringRef> BaseMember = Name.split(Separator: '.'); |
7241 | const StringRef Base = BaseMember.first, Member = BaseMember.second; |
7242 | return lookUpField(Base, Member, Info); |
7243 | } |
7244 | |
7245 | bool MasmParser::lookUpField(StringRef Base, StringRef Member, |
7246 | AsmFieldInfo &Info) const { |
7247 | if (Base.empty()) |
7248 | return true; |
7249 | |
7250 | AsmFieldInfo BaseInfo; |
7251 | if (Base.contains(C: '.') && !lookUpField(Name: Base, Info&: BaseInfo)) |
7252 | Base = BaseInfo.Type.Name; |
7253 | |
7254 | auto StructIt = Structs.find(Key: Base.lower()); |
7255 | auto TypeIt = KnownType.find(Key: Base.lower()); |
7256 | if (TypeIt != KnownType.end()) { |
7257 | StructIt = Structs.find(Key: TypeIt->second.Name.lower()); |
7258 | } |
7259 | if (StructIt != Structs.end()) |
7260 | return lookUpField(Structure: StructIt->second, Member, Info); |
7261 | |
7262 | return true; |
7263 | } |
7264 | |
7265 | bool MasmParser::lookUpField(const StructInfo &Structure, StringRef Member, |
7266 | AsmFieldInfo &Info) const { |
7267 | if (Member.empty()) { |
7268 | Info.Type.Name = Structure.Name; |
7269 | Info.Type.Size = Structure.Size; |
7270 | Info.Type.ElementSize = Structure.Size; |
7271 | Info.Type.Length = 1; |
7272 | return false; |
7273 | } |
7274 | |
7275 | std::pair<StringRef, StringRef> Split = Member.split(Separator: '.'); |
7276 | const StringRef FieldName = Split.first, FieldMember = Split.second; |
7277 | |
7278 | auto StructIt = Structs.find(Key: FieldName.lower()); |
7279 | if (StructIt != Structs.end()) |
7280 | return lookUpField(Structure: StructIt->second, Member: FieldMember, Info); |
7281 | |
7282 | auto FieldIt = Structure.FieldsByName.find(Key: FieldName.lower()); |
7283 | if (FieldIt == Structure.FieldsByName.end()) |
7284 | return true; |
7285 | |
7286 | const FieldInfo &Field = Structure.Fields[FieldIt->second]; |
7287 | if (FieldMember.empty()) { |
7288 | Info.Offset += Field.Offset; |
7289 | Info.Type.Size = Field.SizeOf; |
7290 | Info.Type.ElementSize = Field.Type; |
7291 | Info.Type.Length = Field.LengthOf; |
7292 | if (Field.Contents.FT == FT_STRUCT) |
7293 | Info.Type.Name = Field.Contents.StructInfo.Structure.Name; |
7294 | else |
7295 | Info.Type.Name = "" ; |
7296 | return false; |
7297 | } |
7298 | |
7299 | if (Field.Contents.FT != FT_STRUCT) |
7300 | return true; |
7301 | const StructFieldInfo &StructInfo = Field.Contents.StructInfo; |
7302 | |
7303 | if (lookUpField(Structure: StructInfo.Structure, Member: FieldMember, Info)) |
7304 | return true; |
7305 | |
7306 | Info.Offset += Field.Offset; |
7307 | return false; |
7308 | } |
7309 | |
7310 | bool MasmParser::lookUpType(StringRef Name, AsmTypeInfo &Info) const { |
7311 | unsigned Size = StringSwitch<unsigned>(Name) |
7312 | .CasesLower(S0: "byte" , S1: "db" , S2: "sbyte" , Value: 1) |
7313 | .CasesLower(S0: "word" , S1: "dw" , S2: "sword" , Value: 2) |
7314 | .CasesLower(S0: "dword" , S1: "dd" , S2: "sdword" , Value: 4) |
7315 | .CasesLower(S0: "fword" , S1: "df" , Value: 6) |
7316 | .CasesLower(S0: "qword" , S1: "dq" , S2: "sqword" , Value: 8) |
7317 | .CaseLower(S: "real4" , Value: 4) |
7318 | .CaseLower(S: "real8" , Value: 8) |
7319 | .CaseLower(S: "real10" , Value: 10) |
7320 | .Default(Value: 0); |
7321 | if (Size) { |
7322 | Info.Name = Name; |
7323 | Info.ElementSize = Size; |
7324 | Info.Length = 1; |
7325 | Info.Size = Size; |
7326 | return false; |
7327 | } |
7328 | |
7329 | auto StructIt = Structs.find(Key: Name.lower()); |
7330 | if (StructIt != Structs.end()) { |
7331 | const StructInfo &Structure = StructIt->second; |
7332 | Info.Name = Name; |
7333 | Info.ElementSize = Structure.Size; |
7334 | Info.Length = 1; |
7335 | Info.Size = Structure.Size; |
7336 | return false; |
7337 | } |
7338 | |
7339 | return true; |
7340 | } |
7341 | |
7342 | bool MasmParser::parseMSInlineAsm( |
7343 | std::string &AsmString, unsigned &NumOutputs, unsigned &NumInputs, |
7344 | SmallVectorImpl<std::pair<void *, bool>> &OpDecls, |
7345 | SmallVectorImpl<std::string> &Constraints, |
7346 | SmallVectorImpl<std::string> &Clobbers, const MCInstrInfo *MII, |
7347 | const MCInstPrinter *IP, MCAsmParserSemaCallback &SI) { |
7348 | SmallVector<void *, 4> InputDecls; |
7349 | SmallVector<void *, 4> OutputDecls; |
7350 | SmallVector<bool, 4> InputDeclsAddressOf; |
7351 | SmallVector<bool, 4> OutputDeclsAddressOf; |
7352 | SmallVector<std::string, 4> InputConstraints; |
7353 | SmallVector<std::string, 4> OutputConstraints; |
7354 | SmallVector<unsigned, 4> ClobberRegs; |
7355 | |
7356 | SmallVector<AsmRewrite, 4> AsmStrRewrites; |
7357 | |
7358 | // Prime the lexer. |
7359 | Lex(); |
7360 | |
7361 | // While we have input, parse each statement. |
7362 | unsigned InputIdx = 0; |
7363 | unsigned OutputIdx = 0; |
7364 | while (getLexer().isNot(K: AsmToken::Eof)) { |
7365 | // Parse curly braces marking block start/end. |
7366 | if (parseCurlyBlockScope(AsmStrRewrites)) |
7367 | continue; |
7368 | |
7369 | ParseStatementInfo Info(&AsmStrRewrites); |
7370 | bool StatementErr = parseStatement(Info, SI: &SI); |
7371 | |
7372 | if (StatementErr || Info.ParseError) { |
7373 | // Emit pending errors if any exist. |
7374 | printPendingErrors(); |
7375 | return true; |
7376 | } |
7377 | |
7378 | // No pending error should exist here. |
7379 | assert(!hasPendingError() && "unexpected error from parseStatement" ); |
7380 | |
7381 | if (Info.Opcode == ~0U) |
7382 | continue; |
7383 | |
7384 | const MCInstrDesc &Desc = MII->get(Opcode: Info.Opcode); |
7385 | |
7386 | // Build the list of clobbers, outputs and inputs. |
7387 | for (unsigned i = 1, e = Info.ParsedOperands.size(); i != e; ++i) { |
7388 | MCParsedAsmOperand &Operand = *Info.ParsedOperands[i]; |
7389 | |
7390 | // Register operand. |
7391 | if (Operand.isReg() && !Operand.needAddressOf() && |
7392 | !getTargetParser().OmitRegisterFromClobberLists(RegNo: Operand.getReg())) { |
7393 | unsigned NumDefs = Desc.getNumDefs(); |
7394 | // Clobber. |
7395 | if (NumDefs && Operand.getMCOperandNum() < NumDefs) |
7396 | ClobberRegs.push_back(Elt: Operand.getReg()); |
7397 | continue; |
7398 | } |
7399 | |
7400 | // Expr/Input or Output. |
7401 | StringRef SymName = Operand.getSymName(); |
7402 | if (SymName.empty()) |
7403 | continue; |
7404 | |
7405 | void *OpDecl = Operand.getOpDecl(); |
7406 | if (!OpDecl) |
7407 | continue; |
7408 | |
7409 | StringRef Constraint = Operand.getConstraint(); |
7410 | if (Operand.isImm()) { |
7411 | // Offset as immediate. |
7412 | if (Operand.isOffsetOfLocal()) |
7413 | Constraint = "r" ; |
7414 | else |
7415 | Constraint = "i" ; |
7416 | } |
7417 | |
7418 | bool isOutput = (i == 1) && Desc.mayStore(); |
7419 | SMLoc Start = SMLoc::getFromPointer(Ptr: SymName.data()); |
7420 | if (isOutput) { |
7421 | ++InputIdx; |
7422 | OutputDecls.push_back(Elt: OpDecl); |
7423 | OutputDeclsAddressOf.push_back(Elt: Operand.needAddressOf()); |
7424 | OutputConstraints.push_back(Elt: ("=" + Constraint).str()); |
7425 | AsmStrRewrites.emplace_back(Args: AOK_Output, Args&: Start, Args: SymName.size()); |
7426 | } else { |
7427 | InputDecls.push_back(Elt: OpDecl); |
7428 | InputDeclsAddressOf.push_back(Elt: Operand.needAddressOf()); |
7429 | InputConstraints.push_back(Elt: Constraint.str()); |
7430 | if (Desc.operands()[i - 1].isBranchTarget()) |
7431 | AsmStrRewrites.emplace_back(Args: AOK_CallInput, Args&: Start, Args: SymName.size()); |
7432 | else |
7433 | AsmStrRewrites.emplace_back(Args: AOK_Input, Args&: Start, Args: SymName.size()); |
7434 | } |
7435 | } |
7436 | |
7437 | // Consider implicit defs to be clobbers. Think of cpuid and push. |
7438 | llvm::append_range(C&: ClobberRegs, R: Desc.implicit_defs()); |
7439 | } |
7440 | |
7441 | // Set the number of Outputs and Inputs. |
7442 | NumOutputs = OutputDecls.size(); |
7443 | NumInputs = InputDecls.size(); |
7444 | |
7445 | // Set the unique clobbers. |
7446 | array_pod_sort(Start: ClobberRegs.begin(), End: ClobberRegs.end()); |
7447 | ClobberRegs.erase(CS: llvm::unique(R&: ClobberRegs), CE: ClobberRegs.end()); |
7448 | Clobbers.assign(NumElts: ClobberRegs.size(), Elt: std::string()); |
7449 | for (unsigned I = 0, E = ClobberRegs.size(); I != E; ++I) { |
7450 | raw_string_ostream OS(Clobbers[I]); |
7451 | IP->printRegName(OS, Reg: ClobberRegs[I]); |
7452 | } |
7453 | |
7454 | // Merge the various outputs and inputs. Output are expected first. |
7455 | if (NumOutputs || NumInputs) { |
7456 | unsigned NumExprs = NumOutputs + NumInputs; |
7457 | OpDecls.resize(N: NumExprs); |
7458 | Constraints.resize(N: NumExprs); |
7459 | for (unsigned i = 0; i < NumOutputs; ++i) { |
7460 | OpDecls[i] = std::make_pair(x&: OutputDecls[i], y&: OutputDeclsAddressOf[i]); |
7461 | Constraints[i] = OutputConstraints[i]; |
7462 | } |
7463 | for (unsigned i = 0, j = NumOutputs; i < NumInputs; ++i, ++j) { |
7464 | OpDecls[j] = std::make_pair(x&: InputDecls[i], y&: InputDeclsAddressOf[i]); |
7465 | Constraints[j] = InputConstraints[i]; |
7466 | } |
7467 | } |
7468 | |
7469 | // Build the IR assembly string. |
7470 | std::string AsmStringIR; |
7471 | raw_string_ostream OS(AsmStringIR); |
7472 | StringRef ASMString = |
7473 | SrcMgr.getMemoryBuffer(i: SrcMgr.getMainFileID())->getBuffer(); |
7474 | const char *AsmStart = ASMString.begin(); |
7475 | const char *AsmEnd = ASMString.end(); |
7476 | array_pod_sort(Start: AsmStrRewrites.begin(), End: AsmStrRewrites.end(), Compare: rewritesSort); |
7477 | for (auto I = AsmStrRewrites.begin(), E = AsmStrRewrites.end(); I != E; ++I) { |
7478 | const AsmRewrite &AR = *I; |
7479 | // Check if this has already been covered by another rewrite... |
7480 | if (AR.Done) |
7481 | continue; |
7482 | AsmRewriteKind Kind = AR.Kind; |
7483 | |
7484 | const char *Loc = AR.Loc.getPointer(); |
7485 | assert(Loc >= AsmStart && "Expected Loc to be at or after Start!" ); |
7486 | |
7487 | // Emit everything up to the immediate/expression. |
7488 | if (unsigned Len = Loc - AsmStart) |
7489 | OS << StringRef(AsmStart, Len); |
7490 | |
7491 | // Skip the original expression. |
7492 | if (Kind == AOK_Skip) { |
7493 | AsmStart = Loc + AR.Len; |
7494 | continue; |
7495 | } |
7496 | |
7497 | unsigned AdditionalSkip = 0; |
7498 | // Rewrite expressions in $N notation. |
7499 | switch (Kind) { |
7500 | default: |
7501 | break; |
7502 | case AOK_IntelExpr: |
7503 | assert(AR.IntelExp.isValid() && "cannot write invalid intel expression" ); |
7504 | if (AR.IntelExp.NeedBracs) |
7505 | OS << "[" ; |
7506 | if (AR.IntelExp.hasBaseReg()) |
7507 | OS << AR.IntelExp.BaseReg; |
7508 | if (AR.IntelExp.hasIndexReg()) |
7509 | OS << (AR.IntelExp.hasBaseReg() ? " + " : "" ) |
7510 | << AR.IntelExp.IndexReg; |
7511 | if (AR.IntelExp.Scale > 1) |
7512 | OS << " * $$" << AR.IntelExp.Scale; |
7513 | if (AR.IntelExp.hasOffset()) { |
7514 | if (AR.IntelExp.hasRegs()) |
7515 | OS << " + " ; |
7516 | // Fuse this rewrite with a rewrite of the offset name, if present. |
7517 | StringRef OffsetName = AR.IntelExp.OffsetName; |
7518 | SMLoc OffsetLoc = SMLoc::getFromPointer(Ptr: AR.IntelExp.OffsetName.data()); |
7519 | size_t OffsetLen = OffsetName.size(); |
7520 | auto rewrite_it = std::find_if( |
7521 | first: I, last: AsmStrRewrites.end(), pred: [&](const AsmRewrite &FusingAR) { |
7522 | return FusingAR.Loc == OffsetLoc && FusingAR.Len == OffsetLen && |
7523 | (FusingAR.Kind == AOK_Input || |
7524 | FusingAR.Kind == AOK_CallInput); |
7525 | }); |
7526 | if (rewrite_it == AsmStrRewrites.end()) { |
7527 | OS << "offset " << OffsetName; |
7528 | } else if (rewrite_it->Kind == AOK_CallInput) { |
7529 | OS << "${" << InputIdx++ << ":P}" ; |
7530 | rewrite_it->Done = true; |
7531 | } else { |
7532 | OS << '$' << InputIdx++; |
7533 | rewrite_it->Done = true; |
7534 | } |
7535 | } |
7536 | if (AR.IntelExp.Imm || AR.IntelExp.emitImm()) |
7537 | OS << (AR.IntelExp.emitImm() ? "$$" : " + $$" ) << AR.IntelExp.Imm; |
7538 | if (AR.IntelExp.NeedBracs) |
7539 | OS << "]" ; |
7540 | break; |
7541 | case AOK_Label: |
7542 | OS << Ctx.getAsmInfo()->getPrivateLabelPrefix() << AR.Label; |
7543 | break; |
7544 | case AOK_Input: |
7545 | OS << '$' << InputIdx++; |
7546 | break; |
7547 | case AOK_CallInput: |
7548 | OS << "${" << InputIdx++ << ":P}" ; |
7549 | break; |
7550 | case AOK_Output: |
7551 | OS << '$' << OutputIdx++; |
7552 | break; |
7553 | case AOK_SizeDirective: |
7554 | switch (AR.Val) { |
7555 | default: break; |
7556 | case 8: OS << "byte ptr " ; break; |
7557 | case 16: OS << "word ptr " ; break; |
7558 | case 32: OS << "dword ptr " ; break; |
7559 | case 64: OS << "qword ptr " ; break; |
7560 | case 80: OS << "xword ptr " ; break; |
7561 | case 128: OS << "xmmword ptr " ; break; |
7562 | case 256: OS << "ymmword ptr " ; break; |
7563 | } |
7564 | break; |
7565 | case AOK_Emit: |
7566 | OS << ".byte" ; |
7567 | break; |
7568 | case AOK_Align: { |
7569 | // MS alignment directives are measured in bytes. If the native assembler |
7570 | // measures alignment in bytes, we can pass it straight through. |
7571 | OS << ".align" ; |
7572 | if (getContext().getAsmInfo()->getAlignmentIsInBytes()) |
7573 | break; |
7574 | |
7575 | // Alignment is in log2 form, so print that instead and skip the original |
7576 | // immediate. |
7577 | unsigned Val = AR.Val; |
7578 | OS << ' ' << Val; |
7579 | assert(Val < 10 && "Expected alignment less then 2^10." ); |
7580 | AdditionalSkip = (Val < 4) ? 2 : Val < 7 ? 3 : 4; |
7581 | break; |
7582 | } |
7583 | case AOK_EVEN: |
7584 | OS << ".even" ; |
7585 | break; |
7586 | case AOK_EndOfStatement: |
7587 | OS << "\n\t" ; |
7588 | break; |
7589 | } |
7590 | |
7591 | // Skip the original expression. |
7592 | AsmStart = Loc + AR.Len + AdditionalSkip; |
7593 | } |
7594 | |
7595 | // Emit the remainder of the asm string. |
7596 | if (AsmStart != AsmEnd) |
7597 | OS << StringRef(AsmStart, AsmEnd - AsmStart); |
7598 | |
7599 | AsmString = OS.str(); |
7600 | return false; |
7601 | } |
7602 | |
7603 | void MasmParser::initializeBuiltinSymbolMap() { |
7604 | // Numeric built-ins (supported in all versions) |
7605 | BuiltinSymbolMap["@version" ] = BI_VERSION; |
7606 | BuiltinSymbolMap["@line" ] = BI_LINE; |
7607 | |
7608 | // Text built-ins (supported in all versions) |
7609 | BuiltinSymbolMap["@date" ] = BI_DATE; |
7610 | BuiltinSymbolMap["@time" ] = BI_TIME; |
7611 | BuiltinSymbolMap["@filecur" ] = BI_FILECUR; |
7612 | BuiltinSymbolMap["@filename" ] = BI_FILENAME; |
7613 | BuiltinSymbolMap["@curseg" ] = BI_CURSEG; |
7614 | |
7615 | // Some built-ins exist only for MASM32 (32-bit x86) |
7616 | if (getContext().getSubtargetInfo()->getTargetTriple().getArch() == |
7617 | Triple::x86) { |
7618 | // Numeric built-ins |
7619 | // BuiltinSymbolMap["@cpu"] = BI_CPU; |
7620 | // BuiltinSymbolMap["@interface"] = BI_INTERFACE; |
7621 | // BuiltinSymbolMap["@wordsize"] = BI_WORDSIZE; |
7622 | // BuiltinSymbolMap["@codesize"] = BI_CODESIZE; |
7623 | // BuiltinSymbolMap["@datasize"] = BI_DATASIZE; |
7624 | // BuiltinSymbolMap["@model"] = BI_MODEL; |
7625 | |
7626 | // Text built-ins |
7627 | // BuiltinSymbolMap["@code"] = BI_CODE; |
7628 | // BuiltinSymbolMap["@data"] = BI_DATA; |
7629 | // BuiltinSymbolMap["@fardata?"] = BI_FARDATA; |
7630 | // BuiltinSymbolMap["@stack"] = BI_STACK; |
7631 | } |
7632 | } |
7633 | |
7634 | const MCExpr *MasmParser::evaluateBuiltinValue(BuiltinSymbol Symbol, |
7635 | SMLoc StartLoc) { |
7636 | switch (Symbol) { |
7637 | default: |
7638 | return nullptr; |
7639 | case BI_VERSION: |
7640 | // Match a recent version of ML.EXE. |
7641 | return MCConstantExpr::create(Value: 1427, Ctx&: getContext()); |
7642 | case BI_LINE: { |
7643 | int64_t Line; |
7644 | if (ActiveMacros.empty()) |
7645 | Line = SrcMgr.FindLineNumber(Loc: StartLoc, BufferID: CurBuffer); |
7646 | else |
7647 | Line = SrcMgr.FindLineNumber(Loc: ActiveMacros.front()->InstantiationLoc, |
7648 | BufferID: ActiveMacros.front()->ExitBuffer); |
7649 | return MCConstantExpr::create(Value: Line, Ctx&: getContext()); |
7650 | } |
7651 | } |
7652 | llvm_unreachable("unhandled built-in symbol" ); |
7653 | } |
7654 | |
7655 | std::optional<std::string> |
7656 | MasmParser::evaluateBuiltinTextMacro(BuiltinSymbol Symbol, SMLoc StartLoc) { |
7657 | switch (Symbol) { |
7658 | default: |
7659 | return {}; |
7660 | case BI_DATE: { |
7661 | // Current local date, formatted MM/DD/YY |
7662 | char TmpBuffer[sizeof("mm/dd/yy" )]; |
7663 | const size_t Len = strftime(s: TmpBuffer, maxsize: sizeof(TmpBuffer), format: "%D" , tp: &TM); |
7664 | return std::string(TmpBuffer, Len); |
7665 | } |
7666 | case BI_TIME: { |
7667 | // Current local time, formatted HH:MM:SS (24-hour clock) |
7668 | char TmpBuffer[sizeof("hh:mm:ss" )]; |
7669 | const size_t Len = strftime(s: TmpBuffer, maxsize: sizeof(TmpBuffer), format: "%T" , tp: &TM); |
7670 | return std::string(TmpBuffer, Len); |
7671 | } |
7672 | case BI_FILECUR: |
7673 | return SrcMgr |
7674 | .getMemoryBuffer( |
7675 | i: ActiveMacros.empty() ? CurBuffer : ActiveMacros.front()->ExitBuffer) |
7676 | ->getBufferIdentifier() |
7677 | .str(); |
7678 | case BI_FILENAME: |
7679 | return sys::path::stem(path: SrcMgr.getMemoryBuffer(i: SrcMgr.getMainFileID()) |
7680 | ->getBufferIdentifier()) |
7681 | .upper(); |
7682 | case BI_CURSEG: |
7683 | return getStreamer().getCurrentSectionOnly()->getName().str(); |
7684 | } |
7685 | llvm_unreachable("unhandled built-in symbol" ); |
7686 | } |
7687 | |
7688 | /// Create an MCAsmParser instance. |
7689 | MCAsmParser *llvm::createMCMasmParser(SourceMgr &SM, MCContext &C, |
7690 | MCStreamer &Out, const MCAsmInfo &MAI, |
7691 | struct tm TM, unsigned CB) { |
7692 | return new MasmParser(SM, C, Out, MAI, TM, CB); |
7693 | } |
7694 | |