1 | //===- Preprocessor.cpp - C Language Family Preprocessor Implementation ---===// |
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 file implements the Preprocessor interface. |
10 | // |
11 | //===----------------------------------------------------------------------===// |
12 | // |
13 | // Options to support: |
14 | // -H - Print the name of each header file used. |
15 | // -d[DNI] - Dump various things. |
16 | // -fworking-directory - #line's with preprocessor's working dir. |
17 | // -fpreprocessed |
18 | // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD |
19 | // -W* |
20 | // -w |
21 | // |
22 | // Messages to emit: |
23 | // "Multiple include guards may be useful for:\n" |
24 | // |
25 | //===----------------------------------------------------------------------===// |
26 | |
27 | #include "clang/Lex/Preprocessor.h" |
28 | #include "clang/Basic/Builtins.h" |
29 | #include "clang/Basic/FileManager.h" |
30 | #include "clang/Basic/FileSystemStatCache.h" |
31 | #include "clang/Basic/IdentifierTable.h" |
32 | #include "clang/Basic/LLVM.h" |
33 | #include "clang/Basic/LangOptions.h" |
34 | #include "clang/Basic/Module.h" |
35 | #include "clang/Basic/SourceLocation.h" |
36 | #include "clang/Basic/SourceManager.h" |
37 | #include "clang/Basic/TargetInfo.h" |
38 | #include "clang/Lex/CodeCompletionHandler.h" |
39 | #include "clang/Lex/ExternalPreprocessorSource.h" |
40 | #include "clang/Lex/HeaderSearch.h" |
41 | #include "clang/Lex/LexDiagnostic.h" |
42 | #include "clang/Lex/Lexer.h" |
43 | #include "clang/Lex/LiteralSupport.h" |
44 | #include "clang/Lex/MacroArgs.h" |
45 | #include "clang/Lex/MacroInfo.h" |
46 | #include "clang/Lex/ModuleLoader.h" |
47 | #include "clang/Lex/Pragma.h" |
48 | #include "clang/Lex/PreprocessingRecord.h" |
49 | #include "clang/Lex/PreprocessorLexer.h" |
50 | #include "clang/Lex/PreprocessorOptions.h" |
51 | #include "clang/Lex/ScratchBuffer.h" |
52 | #include "clang/Lex/Token.h" |
53 | #include "clang/Lex/TokenLexer.h" |
54 | #include "llvm/ADT/APInt.h" |
55 | #include "llvm/ADT/ArrayRef.h" |
56 | #include "llvm/ADT/DenseMap.h" |
57 | #include "llvm/ADT/STLExtras.h" |
58 | #include "llvm/ADT/SmallString.h" |
59 | #include "llvm/ADT/SmallVector.h" |
60 | #include "llvm/ADT/StringRef.h" |
61 | #include "llvm/ADT/iterator_range.h" |
62 | #include "llvm/Support/Capacity.h" |
63 | #include "llvm/Support/ErrorHandling.h" |
64 | #include "llvm/Support/MemoryBuffer.h" |
65 | #include "llvm/Support/raw_ostream.h" |
66 | #include <algorithm> |
67 | #include <cassert> |
68 | #include <memory> |
69 | #include <optional> |
70 | #include <string> |
71 | #include <utility> |
72 | #include <vector> |
73 | |
74 | using namespace clang; |
75 | |
76 | /// Minimum distance between two check points, in tokens. |
77 | static constexpr unsigned CheckPointStepSize = 1024; |
78 | |
79 | LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry) |
80 | |
81 | ExternalPreprocessorSource::~ExternalPreprocessorSource() = default; |
82 | |
83 | Preprocessor::(std::shared_ptr<PreprocessorOptions> PPOpts, |
84 | DiagnosticsEngine &diags, const LangOptions &opts, |
85 | SourceManager &SM, HeaderSearch &, |
86 | ModuleLoader &TheModuleLoader, |
87 | IdentifierInfoLookup *IILookup, bool , |
88 | TranslationUnitKind TUKind) |
89 | : PPOpts(std::move(PPOpts)), Diags(&diags), LangOpts(opts), |
90 | FileMgr(Headers.getFileMgr()), SourceMgr(SM), |
91 | ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers), |
92 | TheModuleLoader(TheModuleLoader), ExternalSource(nullptr), |
93 | // As the language options may have not been loaded yet (when |
94 | // deserializing an ASTUnit), adding keywords to the identifier table is |
95 | // deferred to Preprocessor::Initialize(). |
96 | Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())), |
97 | TUKind(TUKind), SkipMainFilePreamble(0, true), |
98 | CurSubmoduleState(&NullSubmoduleState) { |
99 | OwnsHeaderSearch = OwnsHeaders; |
100 | |
101 | // Default to discarding comments. |
102 | KeepComments = false; |
103 | KeepMacroComments = false; |
104 | SuppressIncludeNotFoundError = false; |
105 | |
106 | // Macro expansion is enabled. |
107 | DisableMacroExpansion = false; |
108 | MacroExpansionInDirectivesOverride = false; |
109 | InMacroArgs = false; |
110 | ArgMacro = nullptr; |
111 | InMacroArgPreExpansion = false; |
112 | NumCachedTokenLexers = 0; |
113 | PragmasEnabled = true; |
114 | ParsingIfOrElifDirective = false; |
115 | PreprocessedOutput = false; |
116 | |
117 | // We haven't read anything from the external source. |
118 | ReadMacrosFromExternalSource = false; |
119 | |
120 | BuiltinInfo = std::make_unique<Builtin::Context>(); |
121 | |
122 | // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of |
123 | // a macro. They get unpoisoned where it is allowed. |
124 | (Ident__VA_ARGS__ = getIdentifierInfo(Name: "__VA_ARGS__" ))->setIsPoisoned(); |
125 | SetPoisonReason(II: Ident__VA_ARGS__,DiagID: diag::ext_pp_bad_vaargs_use); |
126 | (Ident__VA_OPT__ = getIdentifierInfo(Name: "__VA_OPT__" ))->setIsPoisoned(); |
127 | SetPoisonReason(II: Ident__VA_OPT__,DiagID: diag::ext_pp_bad_vaopt_use); |
128 | |
129 | // Initialize the pragma handlers. |
130 | RegisterBuiltinPragmas(); |
131 | |
132 | // Initialize builtin macros like __LINE__ and friends. |
133 | RegisterBuiltinMacros(); |
134 | |
135 | if(LangOpts.Borland) { |
136 | Ident__exception_info = getIdentifierInfo(Name: "_exception_info" ); |
137 | Ident___exception_info = getIdentifierInfo(Name: "__exception_info" ); |
138 | Ident_GetExceptionInfo = getIdentifierInfo(Name: "GetExceptionInformation" ); |
139 | Ident__exception_code = getIdentifierInfo(Name: "_exception_code" ); |
140 | Ident___exception_code = getIdentifierInfo(Name: "__exception_code" ); |
141 | Ident_GetExceptionCode = getIdentifierInfo(Name: "GetExceptionCode" ); |
142 | Ident__abnormal_termination = getIdentifierInfo(Name: "_abnormal_termination" ); |
143 | Ident___abnormal_termination = getIdentifierInfo(Name: "__abnormal_termination" ); |
144 | Ident_AbnormalTermination = getIdentifierInfo(Name: "AbnormalTermination" ); |
145 | } else { |
146 | Ident__exception_info = Ident__exception_code = nullptr; |
147 | Ident__abnormal_termination = Ident___exception_info = nullptr; |
148 | Ident___exception_code = Ident___abnormal_termination = nullptr; |
149 | Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr; |
150 | Ident_AbnormalTermination = nullptr; |
151 | } |
152 | |
153 | // Default incremental processing to -fincremental-extensions, clients can |
154 | // override with `enableIncrementalProcessing` if desired. |
155 | IncrementalProcessing = LangOpts.IncrementalExtensions; |
156 | |
157 | // If using a PCH where a #pragma hdrstop is expected, start skipping tokens. |
158 | if (usingPCHWithPragmaHdrStop()) |
159 | SkippingUntilPragmaHdrStop = true; |
160 | |
161 | // If using a PCH with a through header, start skipping tokens. |
162 | if (!this->PPOpts->PCHThroughHeader.empty() && |
163 | !this->PPOpts->ImplicitPCHInclude.empty()) |
164 | SkippingUntilPCHThroughHeader = true; |
165 | |
166 | if (this->PPOpts->GeneratePreamble) |
167 | PreambleConditionalStack.startRecording(); |
168 | |
169 | MaxTokens = LangOpts.MaxTokens; |
170 | } |
171 | |
172 | Preprocessor::~Preprocessor() { |
173 | assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!" ); |
174 | |
175 | IncludeMacroStack.clear(); |
176 | |
177 | // Free any cached macro expanders. |
178 | // This populates MacroArgCache, so all TokenLexers need to be destroyed |
179 | // before the code below that frees up the MacroArgCache list. |
180 | std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr); |
181 | CurTokenLexer.reset(); |
182 | |
183 | // Free any cached MacroArgs. |
184 | for (MacroArgs *ArgList = MacroArgCache; ArgList;) |
185 | ArgList = ArgList->deallocate(); |
186 | |
187 | // Delete the header search info, if we own it. |
188 | if (OwnsHeaderSearch) |
189 | delete &HeaderInfo; |
190 | } |
191 | |
192 | void Preprocessor::Initialize(const TargetInfo &Target, |
193 | const TargetInfo *AuxTarget) { |
194 | assert((!this->Target || this->Target == &Target) && |
195 | "Invalid override of target information" ); |
196 | this->Target = &Target; |
197 | |
198 | assert((!this->AuxTarget || this->AuxTarget == AuxTarget) && |
199 | "Invalid override of aux target information." ); |
200 | this->AuxTarget = AuxTarget; |
201 | |
202 | // Initialize information about built-ins. |
203 | BuiltinInfo->InitializeTarget(Target, AuxTarget); |
204 | HeaderInfo.setTarget(Target); |
205 | |
206 | // Populate the identifier table with info about keywords for the current language. |
207 | Identifiers.AddKeywords(LangOpts); |
208 | |
209 | // Initialize the __FTL_EVAL_METHOD__ macro to the TargetInfo. |
210 | setTUFPEvalMethod(getTargetInfo().getFPEvalMethod()); |
211 | |
212 | if (getLangOpts().getFPEvalMethod() == LangOptions::FEM_UnsetOnCommandLine) |
213 | // Use setting from TargetInfo. |
214 | setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: Target.getFPEvalMethod()); |
215 | else |
216 | // Set initial value of __FLT_EVAL_METHOD__ from the command line. |
217 | setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: getLangOpts().getFPEvalMethod()); |
218 | } |
219 | |
220 | void Preprocessor::InitializeForModelFile() { |
221 | NumEnteredSourceFiles = 0; |
222 | |
223 | // Reset pragmas |
224 | PragmaHandlersBackup = std::move(PragmaHandlers); |
225 | PragmaHandlers = std::make_unique<PragmaNamespace>(args: StringRef()); |
226 | RegisterBuiltinPragmas(); |
227 | |
228 | // Reset PredefinesFileID |
229 | PredefinesFileID = FileID(); |
230 | } |
231 | |
232 | void Preprocessor::FinalizeForModelFile() { |
233 | NumEnteredSourceFiles = 1; |
234 | |
235 | PragmaHandlers = std::move(PragmaHandlersBackup); |
236 | } |
237 | |
238 | void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const { |
239 | llvm::errs() << tok::getTokenName(Kind: Tok.getKind()); |
240 | |
241 | if (!Tok.isAnnotation()) |
242 | llvm::errs() << " '" << getSpelling(Tok) << "'" ; |
243 | |
244 | if (!DumpFlags) return; |
245 | |
246 | llvm::errs() << "\t" ; |
247 | if (Tok.isAtStartOfLine()) |
248 | llvm::errs() << " [StartOfLine]" ; |
249 | if (Tok.hasLeadingSpace()) |
250 | llvm::errs() << " [LeadingSpace]" ; |
251 | if (Tok.isExpandDisabled()) |
252 | llvm::errs() << " [ExpandDisabled]" ; |
253 | if (Tok.needsCleaning()) { |
254 | const char *Start = SourceMgr.getCharacterData(SL: Tok.getLocation()); |
255 | llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength()) |
256 | << "']" ; |
257 | } |
258 | |
259 | llvm::errs() << "\tLoc=<" ; |
260 | DumpLocation(Loc: Tok.getLocation()); |
261 | llvm::errs() << ">" ; |
262 | } |
263 | |
264 | void Preprocessor::DumpLocation(SourceLocation Loc) const { |
265 | Loc.print(OS&: llvm::errs(), SM: SourceMgr); |
266 | } |
267 | |
268 | void Preprocessor::DumpMacro(const MacroInfo &MI) const { |
269 | llvm::errs() << "MACRO: " ; |
270 | for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) { |
271 | DumpToken(Tok: MI.getReplacementToken(Tok: i)); |
272 | llvm::errs() << " " ; |
273 | } |
274 | llvm::errs() << "\n" ; |
275 | } |
276 | |
277 | void Preprocessor::PrintStats() { |
278 | llvm::errs() << "\n*** Preprocessor Stats:\n" ; |
279 | llvm::errs() << NumDirectives << " directives found:\n" ; |
280 | llvm::errs() << " " << NumDefined << " #define.\n" ; |
281 | llvm::errs() << " " << NumUndefined << " #undef.\n" ; |
282 | llvm::errs() << " #include/#include_next/#import:\n" ; |
283 | llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n" ; |
284 | llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n" ; |
285 | llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n" ; |
286 | llvm::errs() << " " << NumElse << " #else/#elif/#elifdef/#elifndef.\n" ; |
287 | llvm::errs() << " " << NumEndif << " #endif.\n" ; |
288 | llvm::errs() << " " << NumPragma << " #pragma.\n" ; |
289 | llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n" ; |
290 | |
291 | llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/" |
292 | << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, " |
293 | << NumFastMacroExpanded << " on the fast path.\n" ; |
294 | llvm::errs() << (NumFastTokenPaste+NumTokenPaste) |
295 | << " token paste (##) operations performed, " |
296 | << NumFastTokenPaste << " on the fast path.\n" ; |
297 | |
298 | llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total" ; |
299 | |
300 | llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory(); |
301 | llvm::errs() << "\n Macro Expanded Tokens: " |
302 | << llvm::capacity_in_bytes(X: MacroExpandedTokens); |
303 | llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity(); |
304 | // FIXME: List information for all submodules. |
305 | llvm::errs() << "\n Macros: " |
306 | << llvm::capacity_in_bytes(X: CurSubmoduleState->Macros); |
307 | llvm::errs() << "\n #pragma push_macro Info: " |
308 | << llvm::capacity_in_bytes(X: PragmaPushMacroInfo); |
309 | llvm::errs() << "\n Poison Reasons: " |
310 | << llvm::capacity_in_bytes(X: PoisonReasons); |
311 | llvm::errs() << "\n Comment Handlers: " |
312 | << llvm::capacity_in_bytes(x: CommentHandlers) << "\n" ; |
313 | } |
314 | |
315 | Preprocessor::macro_iterator |
316 | Preprocessor::macro_begin(bool IncludeExternalMacros) const { |
317 | if (IncludeExternalMacros && ExternalSource && |
318 | !ReadMacrosFromExternalSource) { |
319 | ReadMacrosFromExternalSource = true; |
320 | ExternalSource->ReadDefinedMacros(); |
321 | } |
322 | |
323 | // Make sure we cover all macros in visible modules. |
324 | for (const ModuleMacro &Macro : ModuleMacros) |
325 | CurSubmoduleState->Macros.insert(KV: std::make_pair(x: Macro.II, y: MacroState())); |
326 | |
327 | return CurSubmoduleState->Macros.begin(); |
328 | } |
329 | |
330 | size_t Preprocessor::getTotalMemory() const { |
331 | return BP.getTotalMemory() |
332 | + llvm::capacity_in_bytes(X: MacroExpandedTokens) |
333 | + Predefines.capacity() /* Predefines buffer. */ |
334 | // FIXME: Include sizes from all submodules, and include MacroInfo sizes, |
335 | // and ModuleMacros. |
336 | + llvm::capacity_in_bytes(X: CurSubmoduleState->Macros) |
337 | + llvm::capacity_in_bytes(X: PragmaPushMacroInfo) |
338 | + llvm::capacity_in_bytes(X: PoisonReasons) |
339 | + llvm::capacity_in_bytes(x: CommentHandlers); |
340 | } |
341 | |
342 | Preprocessor::macro_iterator |
343 | Preprocessor::macro_end(bool IncludeExternalMacros) const { |
344 | if (IncludeExternalMacros && ExternalSource && |
345 | !ReadMacrosFromExternalSource) { |
346 | ReadMacrosFromExternalSource = true; |
347 | ExternalSource->ReadDefinedMacros(); |
348 | } |
349 | |
350 | return CurSubmoduleState->Macros.end(); |
351 | } |
352 | |
353 | /// Compares macro tokens with a specified token value sequence. |
354 | static bool MacroDefinitionEquals(const MacroInfo *MI, |
355 | ArrayRef<TokenValue> Tokens) { |
356 | return Tokens.size() == MI->getNumTokens() && |
357 | std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin()); |
358 | } |
359 | |
360 | StringRef Preprocessor::getLastMacroWithSpelling( |
361 | SourceLocation Loc, |
362 | ArrayRef<TokenValue> Tokens) const { |
363 | SourceLocation BestLocation; |
364 | StringRef BestSpelling; |
365 | for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end(); |
366 | I != E; ++I) { |
367 | const MacroDirective::DefInfo |
368 | Def = I->second.findDirectiveAtLoc(Loc, SourceMgr); |
369 | if (!Def || !Def.getMacroInfo()) |
370 | continue; |
371 | if (!Def.getMacroInfo()->isObjectLike()) |
372 | continue; |
373 | if (!MacroDefinitionEquals(MI: Def.getMacroInfo(), Tokens)) |
374 | continue; |
375 | SourceLocation Location = Def.getLocation(); |
376 | // Choose the macro defined latest. |
377 | if (BestLocation.isInvalid() || |
378 | (Location.isValid() && |
379 | SourceMgr.isBeforeInTranslationUnit(LHS: BestLocation, RHS: Location))) { |
380 | BestLocation = Location; |
381 | BestSpelling = I->first->getName(); |
382 | } |
383 | } |
384 | return BestSpelling; |
385 | } |
386 | |
387 | void Preprocessor::recomputeCurLexerKind() { |
388 | if (CurLexer) |
389 | CurLexerCallback = CurLexer->isDependencyDirectivesLexer() |
390 | ? CLK_DependencyDirectivesLexer |
391 | : CLK_Lexer; |
392 | else if (CurTokenLexer) |
393 | CurLexerCallback = CLK_TokenLexer; |
394 | else |
395 | CurLexerCallback = CLK_CachingLexer; |
396 | } |
397 | |
398 | bool Preprocessor::SetCodeCompletionPoint(FileEntryRef File, |
399 | unsigned CompleteLine, |
400 | unsigned CompleteColumn) { |
401 | assert(CompleteLine && CompleteColumn && "Starts from 1:1" ); |
402 | assert(!CodeCompletionFile && "Already set" ); |
403 | |
404 | // Load the actual file's contents. |
405 | std::optional<llvm::MemoryBufferRef> Buffer = |
406 | SourceMgr.getMemoryBufferForFileOrNone(File); |
407 | if (!Buffer) |
408 | return true; |
409 | |
410 | // Find the byte position of the truncation point. |
411 | const char *Position = Buffer->getBufferStart(); |
412 | for (unsigned Line = 1; Line < CompleteLine; ++Line) { |
413 | for (; *Position; ++Position) { |
414 | if (*Position != '\r' && *Position != '\n') |
415 | continue; |
416 | |
417 | // Eat \r\n or \n\r as a single line. |
418 | if ((Position[1] == '\r' || Position[1] == '\n') && |
419 | Position[0] != Position[1]) |
420 | ++Position; |
421 | ++Position; |
422 | break; |
423 | } |
424 | } |
425 | |
426 | Position += CompleteColumn - 1; |
427 | |
428 | // If pointing inside the preamble, adjust the position at the beginning of |
429 | // the file after the preamble. |
430 | if (SkipMainFilePreamble.first && |
431 | SourceMgr.getFileEntryForID(FID: SourceMgr.getMainFileID()) == File) { |
432 | if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first) |
433 | Position = Buffer->getBufferStart() + SkipMainFilePreamble.first; |
434 | } |
435 | |
436 | if (Position > Buffer->getBufferEnd()) |
437 | Position = Buffer->getBufferEnd(); |
438 | |
439 | CodeCompletionFile = File; |
440 | CodeCompletionOffset = Position - Buffer->getBufferStart(); |
441 | |
442 | auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer( |
443 | Size: Buffer->getBufferSize() + 1, BufferName: Buffer->getBufferIdentifier()); |
444 | char *NewBuf = NewBuffer->getBufferStart(); |
445 | char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf); |
446 | *NewPos = '\0'; |
447 | std::copy(Position, Buffer->getBufferEnd(), NewPos+1); |
448 | SourceMgr.overrideFileContents(SourceFile: File, Buffer: std::move(NewBuffer)); |
449 | |
450 | return false; |
451 | } |
452 | |
453 | void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir, |
454 | bool IsAngled) { |
455 | setCodeCompletionReached(); |
456 | if (CodeComplete) |
457 | CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled); |
458 | } |
459 | |
460 | void Preprocessor::CodeCompleteNaturalLanguage() { |
461 | setCodeCompletionReached(); |
462 | if (CodeComplete) |
463 | CodeComplete->CodeCompleteNaturalLanguage(); |
464 | } |
465 | |
466 | /// getSpelling - This method is used to get the spelling of a token into a |
467 | /// SmallVector. Note that the returned StringRef may not point to the |
468 | /// supplied buffer if a copy can be avoided. |
469 | StringRef Preprocessor::getSpelling(const Token &Tok, |
470 | SmallVectorImpl<char> &Buffer, |
471 | bool *Invalid) const { |
472 | // NOTE: this has to be checked *before* testing for an IdentifierInfo. |
473 | if (Tok.isNot(K: tok::raw_identifier) && !Tok.hasUCN()) { |
474 | // Try the fast path. |
475 | if (const IdentifierInfo *II = Tok.getIdentifierInfo()) |
476 | return II->getName(); |
477 | } |
478 | |
479 | // Resize the buffer if we need to copy into it. |
480 | if (Tok.needsCleaning()) |
481 | Buffer.resize(N: Tok.getLength()); |
482 | |
483 | const char *Ptr = Buffer.data(); |
484 | unsigned Len = getSpelling(Tok, Buffer&: Ptr, Invalid); |
485 | return StringRef(Ptr, Len); |
486 | } |
487 | |
488 | /// CreateString - Plop the specified string into a scratch buffer and return a |
489 | /// location for it. If specified, the source location provides a source |
490 | /// location for the token. |
491 | void Preprocessor::CreateString(StringRef Str, Token &Tok, |
492 | SourceLocation ExpansionLocStart, |
493 | SourceLocation ExpansionLocEnd) { |
494 | Tok.setLength(Str.size()); |
495 | |
496 | const char *DestPtr; |
497 | SourceLocation Loc = ScratchBuf->getToken(Buf: Str.data(), Len: Str.size(), DestPtr); |
498 | |
499 | if (ExpansionLocStart.isValid()) |
500 | Loc = SourceMgr.createExpansionLoc(SpellingLoc: Loc, ExpansionLocStart, |
501 | ExpansionLocEnd, Length: Str.size()); |
502 | Tok.setLocation(Loc); |
503 | |
504 | // If this is a raw identifier or a literal token, set the pointer data. |
505 | if (Tok.is(K: tok::raw_identifier)) |
506 | Tok.setRawIdentifierData(DestPtr); |
507 | else if (Tok.isLiteral()) |
508 | Tok.setLiteralData(DestPtr); |
509 | } |
510 | |
511 | SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) { |
512 | auto &SM = getSourceManager(); |
513 | SourceLocation SpellingLoc = SM.getSpellingLoc(Loc); |
514 | std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc: SpellingLoc); |
515 | bool Invalid = false; |
516 | StringRef Buffer = SM.getBufferData(FID: LocInfo.first, Invalid: &Invalid); |
517 | if (Invalid) |
518 | return SourceLocation(); |
519 | |
520 | // FIXME: We could consider re-using spelling for tokens we see repeatedly. |
521 | const char *DestPtr; |
522 | SourceLocation Spelling = |
523 | ScratchBuf->getToken(Buf: Buffer.data() + LocInfo.second, Len: Length, DestPtr); |
524 | return SM.createTokenSplitLoc(SpellingLoc: Spelling, TokenStart: Loc, TokenEnd: Loc.getLocWithOffset(Offset: Length)); |
525 | } |
526 | |
527 | Module *Preprocessor::getCurrentModule() { |
528 | if (!getLangOpts().isCompilingModule()) |
529 | return nullptr; |
530 | |
531 | return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().CurrentModule); |
532 | } |
533 | |
534 | Module *Preprocessor::getCurrentModuleImplementation() { |
535 | if (!getLangOpts().isCompilingModuleImplementation()) |
536 | return nullptr; |
537 | |
538 | return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().ModuleName); |
539 | } |
540 | |
541 | //===----------------------------------------------------------------------===// |
542 | // Preprocessor Initialization Methods |
543 | //===----------------------------------------------------------------------===// |
544 | |
545 | /// EnterMainSourceFile - Enter the specified FileID as the main source file, |
546 | /// which implicitly adds the builtin defines etc. |
547 | void Preprocessor::EnterMainSourceFile() { |
548 | // We do not allow the preprocessor to reenter the main file. Doing so will |
549 | // cause FileID's to accumulate information from both runs (e.g. #line |
550 | // information) and predefined macros aren't guaranteed to be set properly. |
551 | assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!" ); |
552 | FileID MainFileID = SourceMgr.getMainFileID(); |
553 | |
554 | // If MainFileID is loaded it means we loaded an AST file, no need to enter |
555 | // a main file. |
556 | if (!SourceMgr.isLoadedFileID(FID: MainFileID)) { |
557 | // Enter the main file source buffer. |
558 | EnterSourceFile(FID: MainFileID, Dir: nullptr, Loc: SourceLocation()); |
559 | |
560 | // If we've been asked to skip bytes in the main file (e.g., as part of a |
561 | // precompiled preamble), do so now. |
562 | if (SkipMainFilePreamble.first > 0) |
563 | CurLexer->SetByteOffset(Offset: SkipMainFilePreamble.first, |
564 | StartOfLine: SkipMainFilePreamble.second); |
565 | |
566 | // Tell the header info that the main file was entered. If the file is later |
567 | // #imported, it won't be re-entered. |
568 | if (OptionalFileEntryRef FE = SourceMgr.getFileEntryRefForID(FID: MainFileID)) |
569 | markIncluded(File: *FE); |
570 | } |
571 | |
572 | // Preprocess Predefines to populate the initial preprocessor state. |
573 | std::unique_ptr<llvm::MemoryBuffer> SB = |
574 | llvm::MemoryBuffer::getMemBufferCopy(InputData: Predefines, BufferName: "<built-in>" ); |
575 | assert(SB && "Cannot create predefined source buffer" ); |
576 | FileID FID = SourceMgr.createFileID(Buffer: std::move(SB)); |
577 | assert(FID.isValid() && "Could not create FileID for predefines?" ); |
578 | setPredefinesFileID(FID); |
579 | |
580 | // Start parsing the predefines. |
581 | EnterSourceFile(FID, Dir: nullptr, Loc: SourceLocation()); |
582 | |
583 | if (!PPOpts->PCHThroughHeader.empty()) { |
584 | // Lookup and save the FileID for the through header. If it isn't found |
585 | // in the search path, it's a fatal error. |
586 | OptionalFileEntryRef File = LookupFile( |
587 | FilenameLoc: SourceLocation(), Filename: PPOpts->PCHThroughHeader, |
588 | /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, |
589 | /*CurDir=*/nullptr, /*SearchPath=*/nullptr, /*RelativePath=*/nullptr, |
590 | /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr, |
591 | /*IsFrameworkFound=*/nullptr); |
592 | if (!File) { |
593 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_found) |
594 | << PPOpts->PCHThroughHeader; |
595 | return; |
596 | } |
597 | setPCHThroughHeaderFileID( |
598 | SourceMgr.createFileID(SourceFile: *File, IncludePos: SourceLocation(), FileCharacter: SrcMgr::C_User)); |
599 | } |
600 | |
601 | // Skip tokens from the Predefines and if needed the main file. |
602 | if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) || |
603 | (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop)) |
604 | SkipTokensWhileUsingPCH(); |
605 | } |
606 | |
607 | void Preprocessor::(FileID FID) { |
608 | assert(PCHThroughHeaderFileID.isInvalid() && |
609 | "PCHThroughHeaderFileID already set!" ); |
610 | PCHThroughHeaderFileID = FID; |
611 | } |
612 | |
613 | bool Preprocessor::(const FileEntry *FE) { |
614 | assert(PCHThroughHeaderFileID.isValid() && |
615 | "Invalid PCH through header FileID" ); |
616 | return FE == SourceMgr.getFileEntryForID(FID: PCHThroughHeaderFileID); |
617 | } |
618 | |
619 | bool Preprocessor::() { |
620 | return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() && |
621 | PCHThroughHeaderFileID.isValid(); |
622 | } |
623 | |
624 | bool Preprocessor::() { |
625 | return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() && |
626 | PCHThroughHeaderFileID.isValid(); |
627 | } |
628 | |
629 | bool Preprocessor::creatingPCHWithPragmaHdrStop() { |
630 | return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop; |
631 | } |
632 | |
633 | bool Preprocessor::usingPCHWithPragmaHdrStop() { |
634 | return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop; |
635 | } |
636 | |
637 | /// Skip tokens until after the #include of the through header or |
638 | /// until after a #pragma hdrstop is seen. Tokens in the predefines file |
639 | /// and the main file may be skipped. If the end of the predefines file |
640 | /// is reached, skipping continues into the main file. If the end of the |
641 | /// main file is reached, it's a fatal error. |
642 | void Preprocessor::SkipTokensWhileUsingPCH() { |
643 | bool ReachedMainFileEOF = false; |
644 | bool = SkippingUntilPCHThroughHeader; |
645 | bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop; |
646 | Token Tok; |
647 | while (true) { |
648 | bool InPredefines = |
649 | (CurLexer && CurLexer->getFileID() == getPredefinesFileID()); |
650 | CurLexerCallback(*this, Tok); |
651 | if (Tok.is(K: tok::eof) && !InPredefines) { |
652 | ReachedMainFileEOF = true; |
653 | break; |
654 | } |
655 | if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader) |
656 | break; |
657 | if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop) |
658 | break; |
659 | } |
660 | if (ReachedMainFileEOF) { |
661 | if (UsingPCHThroughHeader) |
662 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_seen) |
663 | << PPOpts->PCHThroughHeader << 1; |
664 | else if (!PPOpts->PCHWithHdrStopCreate) |
665 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_pragma_hdrstop_not_seen); |
666 | } |
667 | } |
668 | |
669 | void Preprocessor::replayPreambleConditionalStack() { |
670 | // Restore the conditional stack from the preamble, if there is one. |
671 | if (PreambleConditionalStack.isReplaying()) { |
672 | assert(CurPPLexer && |
673 | "CurPPLexer is null when calling replayPreambleConditionalStack." ); |
674 | CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack()); |
675 | PreambleConditionalStack.doneReplaying(); |
676 | if (PreambleConditionalStack.reachedEOFWhileSkipping()) |
677 | SkipExcludedConditionalBlock( |
678 | HashTokenLoc: PreambleConditionalStack.SkipInfo->HashTokenLoc, |
679 | IfTokenLoc: PreambleConditionalStack.SkipInfo->IfTokenLoc, |
680 | FoundNonSkipPortion: PreambleConditionalStack.SkipInfo->FoundNonSkipPortion, |
681 | FoundElse: PreambleConditionalStack.SkipInfo->FoundElse, |
682 | ElseLoc: PreambleConditionalStack.SkipInfo->ElseLoc); |
683 | } |
684 | } |
685 | |
686 | void Preprocessor::EndSourceFile() { |
687 | // Notify the client that we reached the end of the source file. |
688 | if (Callbacks) |
689 | Callbacks->EndOfMainFile(); |
690 | } |
691 | |
692 | //===----------------------------------------------------------------------===// |
693 | // Lexer Event Handling. |
694 | //===----------------------------------------------------------------------===// |
695 | |
696 | /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the |
697 | /// identifier information for the token and install it into the token, |
698 | /// updating the token kind accordingly. |
699 | IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const { |
700 | assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!" ); |
701 | |
702 | // Look up this token, see if it is a macro, or if it is a language keyword. |
703 | IdentifierInfo *II; |
704 | if (!Identifier.needsCleaning() && !Identifier.hasUCN()) { |
705 | // No cleaning needed, just use the characters from the lexed buffer. |
706 | II = getIdentifierInfo(Name: Identifier.getRawIdentifier()); |
707 | } else { |
708 | // Cleaning needed, alloca a buffer, clean into it, then use the buffer. |
709 | SmallString<64> IdentifierBuffer; |
710 | StringRef CleanedStr = getSpelling(Tok: Identifier, Buffer&: IdentifierBuffer); |
711 | |
712 | if (Identifier.hasUCN()) { |
713 | SmallString<64> UCNIdentifierBuffer; |
714 | expandUCNs(Buf&: UCNIdentifierBuffer, Input: CleanedStr); |
715 | II = getIdentifierInfo(Name: UCNIdentifierBuffer); |
716 | } else { |
717 | II = getIdentifierInfo(Name: CleanedStr); |
718 | } |
719 | } |
720 | |
721 | // Update the token info (identifier info and appropriate token kind). |
722 | // FIXME: the raw_identifier may contain leading whitespace which is removed |
723 | // from the cleaned identifier token. The SourceLocation should be updated to |
724 | // refer to the non-whitespace character. For instance, the text "\\\nB" (a |
725 | // line continuation before 'B') is parsed as a single tok::raw_identifier and |
726 | // is cleaned to tok::identifier "B". After cleaning the token's length is |
727 | // still 3 and the SourceLocation refers to the location of the backslash. |
728 | Identifier.setIdentifierInfo(II); |
729 | Identifier.setKind(II->getTokenID()); |
730 | |
731 | return II; |
732 | } |
733 | |
734 | void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) { |
735 | PoisonReasons[II] = DiagID; |
736 | } |
737 | |
738 | void Preprocessor::PoisonSEHIdentifiers(bool Poison) { |
739 | assert(Ident__exception_code && Ident__exception_info); |
740 | assert(Ident___exception_code && Ident___exception_info); |
741 | Ident__exception_code->setIsPoisoned(Poison); |
742 | Ident___exception_code->setIsPoisoned(Poison); |
743 | Ident_GetExceptionCode->setIsPoisoned(Poison); |
744 | Ident__exception_info->setIsPoisoned(Poison); |
745 | Ident___exception_info->setIsPoisoned(Poison); |
746 | Ident_GetExceptionInfo->setIsPoisoned(Poison); |
747 | Ident__abnormal_termination->setIsPoisoned(Poison); |
748 | Ident___abnormal_termination->setIsPoisoned(Poison); |
749 | Ident_AbnormalTermination->setIsPoisoned(Poison); |
750 | } |
751 | |
752 | void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) { |
753 | assert(Identifier.getIdentifierInfo() && |
754 | "Can't handle identifiers without identifier info!" ); |
755 | llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it = |
756 | PoisonReasons.find(Val: Identifier.getIdentifierInfo()); |
757 | if(it == PoisonReasons.end()) |
758 | Diag(Tok: Identifier, DiagID: diag::err_pp_used_poisoned_id); |
759 | else |
760 | Diag(Tok: Identifier,DiagID: it->second) << Identifier.getIdentifierInfo(); |
761 | } |
762 | |
763 | void Preprocessor::updateOutOfDateIdentifier(const IdentifierInfo &II) const { |
764 | assert(II.isOutOfDate() && "not out of date" ); |
765 | getExternalSource()->updateOutOfDateIdentifier(II); |
766 | } |
767 | |
768 | /// HandleIdentifier - This callback is invoked when the lexer reads an |
769 | /// identifier. This callback looks up the identifier in the map and/or |
770 | /// potentially macro expands it or turns it into a named token (like 'for'). |
771 | /// |
772 | /// Note that callers of this method are guarded by checking the |
773 | /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the |
774 | /// IdentifierInfo methods that compute these properties will need to change to |
775 | /// match. |
776 | bool Preprocessor::HandleIdentifier(Token &Identifier) { |
777 | assert(Identifier.getIdentifierInfo() && |
778 | "Can't handle identifiers without identifier info!" ); |
779 | |
780 | IdentifierInfo &II = *Identifier.getIdentifierInfo(); |
781 | |
782 | // If the information about this identifier is out of date, update it from |
783 | // the external source. |
784 | // We have to treat __VA_ARGS__ in a special way, since it gets |
785 | // serialized with isPoisoned = true, but our preprocessor may have |
786 | // unpoisoned it if we're defining a C99 macro. |
787 | if (II.isOutOfDate()) { |
788 | bool CurrentIsPoisoned = false; |
789 | const bool IsSpecialVariadicMacro = |
790 | &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__; |
791 | if (IsSpecialVariadicMacro) |
792 | CurrentIsPoisoned = II.isPoisoned(); |
793 | |
794 | updateOutOfDateIdentifier(II); |
795 | Identifier.setKind(II.getTokenID()); |
796 | |
797 | if (IsSpecialVariadicMacro) |
798 | II.setIsPoisoned(CurrentIsPoisoned); |
799 | } |
800 | |
801 | // If this identifier was poisoned, and if it was not produced from a macro |
802 | // expansion, emit an error. |
803 | if (II.isPoisoned() && CurPPLexer) { |
804 | HandlePoisonedIdentifier(Identifier); |
805 | } |
806 | |
807 | // If this is a macro to be expanded, do it. |
808 | if (const MacroDefinition MD = getMacroDefinition(II: &II)) { |
809 | const auto *MI = MD.getMacroInfo(); |
810 | assert(MI && "macro definition with no macro info?" ); |
811 | if (!DisableMacroExpansion) { |
812 | if (!Identifier.isExpandDisabled() && MI->isEnabled()) { |
813 | // C99 6.10.3p10: If the preprocessing token immediately after the |
814 | // macro name isn't a '(', this macro should not be expanded. |
815 | if (!MI->isFunctionLike() || isNextPPTokenLParen()) |
816 | return HandleMacroExpandedIdentifier(Identifier, MD); |
817 | } else { |
818 | // C99 6.10.3.4p2 says that a disabled macro may never again be |
819 | // expanded, even if it's in a context where it could be expanded in the |
820 | // future. |
821 | Identifier.setFlag(Token::DisableExpand); |
822 | if (MI->isObjectLike() || isNextPPTokenLParen()) |
823 | Diag(Tok: Identifier, DiagID: diag::pp_disabled_macro_expansion); |
824 | } |
825 | } |
826 | } |
827 | |
828 | // If this identifier is a keyword in a newer Standard or proposed Standard, |
829 | // produce a warning. Don't warn if we're not considering macro expansion, |
830 | // since this identifier might be the name of a macro. |
831 | // FIXME: This warning is disabled in cases where it shouldn't be, like |
832 | // "#define constexpr constexpr", "int constexpr;" |
833 | if (II.isFutureCompatKeyword() && !DisableMacroExpansion) { |
834 | Diag(Tok: Identifier, DiagID: getIdentifierTable().getFutureCompatDiagKind(II, LangOpts: getLangOpts())) |
835 | << II.getName(); |
836 | // Don't diagnose this keyword again in this translation unit. |
837 | II.setIsFutureCompatKeyword(false); |
838 | } |
839 | |
840 | // If this is an extension token, diagnose its use. |
841 | // We avoid diagnosing tokens that originate from macro definitions. |
842 | // FIXME: This warning is disabled in cases where it shouldn't be, |
843 | // like "#define TY typeof", "TY(1) x". |
844 | if (II.isExtensionToken() && !DisableMacroExpansion) |
845 | Diag(Tok: Identifier, DiagID: diag::ext_token_used); |
846 | |
847 | // If this is the 'import' contextual keyword following an '@', note |
848 | // that the next token indicates a module name. |
849 | // |
850 | // Note that we do not treat 'import' as a contextual |
851 | // keyword when we're in a caching lexer, because caching lexers only get |
852 | // used in contexts where import declarations are disallowed. |
853 | // |
854 | // Likewise if this is the standard C++ import keyword. |
855 | if (((LastTokenWasAt && II.isModulesImport()) || |
856 | Identifier.is(K: tok::kw_import)) && |
857 | !InMacroArgs && !DisableMacroExpansion && |
858 | (getLangOpts().Modules || getLangOpts().DebuggerSupport) && |
859 | CurLexerCallback != CLK_CachingLexer) { |
860 | ModuleImportLoc = Identifier.getLocation(); |
861 | NamedModuleImportPath.clear(); |
862 | IsAtImport = true; |
863 | ModuleImportExpectsIdentifier = true; |
864 | CurLexerCallback = CLK_LexAfterModuleImport; |
865 | } |
866 | return true; |
867 | } |
868 | |
869 | void Preprocessor::Lex(Token &Result) { |
870 | ++LexLevel; |
871 | |
872 | // We loop here until a lex function returns a token; this avoids recursion. |
873 | while (!CurLexerCallback(*this, Result)) |
874 | ; |
875 | |
876 | if (Result.is(K: tok::unknown) && TheModuleLoader.HadFatalFailure) |
877 | return; |
878 | |
879 | if (Result.is(K: tok::code_completion) && Result.getIdentifierInfo()) { |
880 | // Remember the identifier before code completion token. |
881 | setCodeCompletionIdentifierInfo(Result.getIdentifierInfo()); |
882 | setCodeCompletionTokenRange(Start: Result.getLocation(), End: Result.getEndLoc()); |
883 | // Set IdenfitierInfo to null to avoid confusing code that handles both |
884 | // identifiers and completion tokens. |
885 | Result.setIdentifierInfo(nullptr); |
886 | } |
887 | |
888 | // Update StdCXXImportSeqState to track our position within a C++20 import-seq |
889 | // if this token is being produced as a result of phase 4 of translation. |
890 | // Update TrackGMFState to decide if we are currently in a Global Module |
891 | // Fragment. GMF state updates should precede StdCXXImportSeq ones, since GMF state |
892 | // depends on the prevailing StdCXXImportSeq state in two cases. |
893 | if (getLangOpts().CPlusPlusModules && LexLevel == 1 && |
894 | !Result.getFlag(Flag: Token::IsReinjected)) { |
895 | switch (Result.getKind()) { |
896 | case tok::l_paren: case tok::l_square: case tok::l_brace: |
897 | StdCXXImportSeqState.handleOpenBracket(); |
898 | break; |
899 | case tok::r_paren: case tok::r_square: |
900 | StdCXXImportSeqState.handleCloseBracket(); |
901 | break; |
902 | case tok::r_brace: |
903 | StdCXXImportSeqState.handleCloseBrace(); |
904 | break; |
905 | // This token is injected to represent the translation of '#include "a.h"' |
906 | // into "import a.h;". Mimic the notional ';'. |
907 | case tok::annot_module_include: |
908 | case tok::semi: |
909 | TrackGMFState.handleSemi(); |
910 | StdCXXImportSeqState.handleSemi(); |
911 | ModuleDeclState.handleSemi(); |
912 | break; |
913 | case tok::header_name: |
914 | case tok::annot_header_unit: |
915 | StdCXXImportSeqState.handleHeaderName(); |
916 | break; |
917 | case tok::kw_export: |
918 | TrackGMFState.handleExport(); |
919 | StdCXXImportSeqState.handleExport(); |
920 | ModuleDeclState.handleExport(); |
921 | break; |
922 | case tok::colon: |
923 | ModuleDeclState.handleColon(); |
924 | break; |
925 | case tok::period: |
926 | ModuleDeclState.handlePeriod(); |
927 | break; |
928 | case tok::identifier: |
929 | // Check "import" and "module" when there is no open bracket. The two |
930 | // identifiers are not meaningful with open brackets. |
931 | if (StdCXXImportSeqState.atTopLevel()) { |
932 | if (Result.getIdentifierInfo()->isModulesImport()) { |
933 | TrackGMFState.handleImport(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq()); |
934 | StdCXXImportSeqState.handleImport(); |
935 | if (StdCXXImportSeqState.afterImportSeq()) { |
936 | ModuleImportLoc = Result.getLocation(); |
937 | NamedModuleImportPath.clear(); |
938 | IsAtImport = false; |
939 | ModuleImportExpectsIdentifier = true; |
940 | CurLexerCallback = CLK_LexAfterModuleImport; |
941 | } |
942 | break; |
943 | } else if (Result.getIdentifierInfo() == getIdentifierInfo(Name: "module" )) { |
944 | TrackGMFState.handleModule(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq()); |
945 | ModuleDeclState.handleModule(); |
946 | break; |
947 | } |
948 | } |
949 | ModuleDeclState.handleIdentifier(Identifier: Result.getIdentifierInfo()); |
950 | if (ModuleDeclState.isModuleCandidate()) |
951 | break; |
952 | [[fallthrough]]; |
953 | default: |
954 | TrackGMFState.handleMisc(); |
955 | StdCXXImportSeqState.handleMisc(); |
956 | ModuleDeclState.handleMisc(); |
957 | break; |
958 | } |
959 | } |
960 | |
961 | if (CurLexer && ++CheckPointCounter == CheckPointStepSize) { |
962 | CheckPoints[CurLexer->getFileID()].push_back(Elt: CurLexer->BufferPtr); |
963 | CheckPointCounter = 0; |
964 | } |
965 | |
966 | LastTokenWasAt = Result.is(K: tok::at); |
967 | --LexLevel; |
968 | |
969 | if ((LexLevel == 0 || PreprocessToken) && |
970 | !Result.getFlag(Flag: Token::IsReinjected)) { |
971 | if (LexLevel == 0) |
972 | ++TokenCount; |
973 | if (OnToken) |
974 | OnToken(Result); |
975 | } |
976 | } |
977 | |
978 | void Preprocessor::LexTokensUntilEOF(std::vector<Token> *Tokens) { |
979 | while (1) { |
980 | Token Tok; |
981 | Lex(Result&: Tok); |
982 | if (Tok.isOneOf(K1: tok::unknown, Ks: tok::eof, Ks: tok::eod, |
983 | Ks: tok::annot_repl_input_end)) |
984 | break; |
985 | if (Tokens != nullptr) |
986 | Tokens->push_back(x: Tok); |
987 | } |
988 | } |
989 | |
990 | /// Lex a header-name token (including one formed from header-name-tokens if |
991 | /// \p AllowMacroExpansion is \c true). |
992 | /// |
993 | /// \param FilenameTok Filled in with the next token. On success, this will |
994 | /// be either a header_name token. On failure, it will be whatever other |
995 | /// token was found instead. |
996 | /// \param AllowMacroExpansion If \c true, allow the header name to be formed |
997 | /// by macro expansion (concatenating tokens as necessary if the first |
998 | /// token is a '<'). |
999 | /// \return \c true if we reached EOD or EOF while looking for a > token in |
1000 | /// a concatenated header name and diagnosed it. \c false otherwise. |
1001 | bool Preprocessor::(Token &FilenameTok, bool AllowMacroExpansion) { |
1002 | // Lex using header-name tokenization rules if tokens are being lexed from |
1003 | // a file. Just grab a token normally if we're in a macro expansion. |
1004 | if (CurPPLexer) |
1005 | CurPPLexer->LexIncludeFilename(FilenameTok); |
1006 | else |
1007 | Lex(Result&: FilenameTok); |
1008 | |
1009 | // This could be a <foo/bar.h> file coming from a macro expansion. In this |
1010 | // case, glue the tokens together into an angle_string_literal token. |
1011 | SmallString<128> FilenameBuffer; |
1012 | if (FilenameTok.is(K: tok::less) && AllowMacroExpansion) { |
1013 | bool StartOfLine = FilenameTok.isAtStartOfLine(); |
1014 | bool LeadingSpace = FilenameTok.hasLeadingSpace(); |
1015 | bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro(); |
1016 | |
1017 | SourceLocation Start = FilenameTok.getLocation(); |
1018 | SourceLocation End; |
1019 | FilenameBuffer.push_back(Elt: '<'); |
1020 | |
1021 | // Consume tokens until we find a '>'. |
1022 | // FIXME: A header-name could be formed starting or ending with an |
1023 | // alternative token. It's not clear whether that's ill-formed in all |
1024 | // cases. |
1025 | while (FilenameTok.isNot(K: tok::greater)) { |
1026 | Lex(Result&: FilenameTok); |
1027 | if (FilenameTok.isOneOf(K1: tok::eod, K2: tok::eof)) { |
1028 | Diag(Loc: FilenameTok.getLocation(), DiagID: diag::err_expected) << tok::greater; |
1029 | Diag(Loc: Start, DiagID: diag::note_matching) << tok::less; |
1030 | return true; |
1031 | } |
1032 | |
1033 | End = FilenameTok.getLocation(); |
1034 | |
1035 | // FIXME: Provide code completion for #includes. |
1036 | if (FilenameTok.is(K: tok::code_completion)) { |
1037 | setCodeCompletionReached(); |
1038 | Lex(Result&: FilenameTok); |
1039 | continue; |
1040 | } |
1041 | |
1042 | // Append the spelling of this token to the buffer. If there was a space |
1043 | // before it, add it now. |
1044 | if (FilenameTok.hasLeadingSpace()) |
1045 | FilenameBuffer.push_back(Elt: ' '); |
1046 | |
1047 | // Get the spelling of the token, directly into FilenameBuffer if |
1048 | // possible. |
1049 | size_t PreAppendSize = FilenameBuffer.size(); |
1050 | FilenameBuffer.resize(N: PreAppendSize + FilenameTok.getLength()); |
1051 | |
1052 | const char *BufPtr = &FilenameBuffer[PreAppendSize]; |
1053 | unsigned ActualLen = getSpelling(Tok: FilenameTok, Buffer&: BufPtr); |
1054 | |
1055 | // If the token was spelled somewhere else, copy it into FilenameBuffer. |
1056 | if (BufPtr != &FilenameBuffer[PreAppendSize]) |
1057 | memcpy(dest: &FilenameBuffer[PreAppendSize], src: BufPtr, n: ActualLen); |
1058 | |
1059 | // Resize FilenameBuffer to the correct size. |
1060 | if (FilenameTok.getLength() != ActualLen) |
1061 | FilenameBuffer.resize(N: PreAppendSize + ActualLen); |
1062 | } |
1063 | |
1064 | FilenameTok.startToken(); |
1065 | FilenameTok.setKind(tok::header_name); |
1066 | FilenameTok.setFlagValue(Flag: Token::StartOfLine, Val: StartOfLine); |
1067 | FilenameTok.setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace); |
1068 | FilenameTok.setFlagValue(Flag: Token::LeadingEmptyMacro, Val: LeadingEmptyMacro); |
1069 | CreateString(Str: FilenameBuffer, Tok&: FilenameTok, ExpansionLocStart: Start, ExpansionLocEnd: End); |
1070 | } else if (FilenameTok.is(K: tok::string_literal) && AllowMacroExpansion) { |
1071 | // Convert a string-literal token of the form " h-char-sequence " |
1072 | // (produced by macro expansion) into a header-name token. |
1073 | // |
1074 | // The rules for header-names don't quite match the rules for |
1075 | // string-literals, but all the places where they differ result in |
1076 | // undefined behavior, so we can and do treat them the same. |
1077 | // |
1078 | // A string-literal with a prefix or suffix is not translated into a |
1079 | // header-name. This could theoretically be observable via the C++20 |
1080 | // context-sensitive header-name formation rules. |
1081 | StringRef Str = getSpelling(Tok: FilenameTok, Buffer&: FilenameBuffer); |
1082 | if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"') |
1083 | FilenameTok.setKind(tok::header_name); |
1084 | } |
1085 | |
1086 | return false; |
1087 | } |
1088 | |
1089 | /// Collect the tokens of a C++20 pp-import-suffix. |
1090 | void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) { |
1091 | // FIXME: For error recovery, consider recognizing attribute syntax here |
1092 | // and terminating / diagnosing a missing semicolon if we find anything |
1093 | // else? (Can we leave that to the parser?) |
1094 | unsigned BracketDepth = 0; |
1095 | while (true) { |
1096 | Toks.emplace_back(); |
1097 | Lex(Result&: Toks.back()); |
1098 | |
1099 | switch (Toks.back().getKind()) { |
1100 | case tok::l_paren: case tok::l_square: case tok::l_brace: |
1101 | ++BracketDepth; |
1102 | break; |
1103 | |
1104 | case tok::r_paren: case tok::r_square: case tok::r_brace: |
1105 | if (BracketDepth == 0) |
1106 | return; |
1107 | --BracketDepth; |
1108 | break; |
1109 | |
1110 | case tok::semi: |
1111 | if (BracketDepth == 0) |
1112 | return; |
1113 | break; |
1114 | |
1115 | case tok::eof: |
1116 | return; |
1117 | |
1118 | default: |
1119 | break; |
1120 | } |
1121 | } |
1122 | } |
1123 | |
1124 | |
1125 | /// Lex a token following the 'import' contextual keyword. |
1126 | /// |
1127 | /// pp-import: [C++20] |
1128 | /// import header-name pp-import-suffix[opt] ; |
1129 | /// import header-name-tokens pp-import-suffix[opt] ; |
1130 | /// [ObjC] @ import module-name ; |
1131 | /// [Clang] import module-name ; |
1132 | /// |
1133 | /// header-name-tokens: |
1134 | /// string-literal |
1135 | /// < [any sequence of preprocessing-tokens other than >] > |
1136 | /// |
1137 | /// module-name: |
1138 | /// module-name-qualifier[opt] identifier |
1139 | /// |
1140 | /// module-name-qualifier |
1141 | /// module-name-qualifier[opt] identifier . |
1142 | /// |
1143 | /// We respond to a pp-import by importing macros from the named module. |
1144 | bool Preprocessor::LexAfterModuleImport(Token &Result) { |
1145 | // Figure out what kind of lexer we actually have. |
1146 | recomputeCurLexerKind(); |
1147 | |
1148 | // Lex the next token. The header-name lexing rules are used at the start of |
1149 | // a pp-import. |
1150 | // |
1151 | // For now, we only support header-name imports in C++20 mode. |
1152 | // FIXME: Should we allow this in all language modes that support an import |
1153 | // declaration as an extension? |
1154 | if (NamedModuleImportPath.empty() && getLangOpts().CPlusPlusModules) { |
1155 | if (LexHeaderName(FilenameTok&: Result)) |
1156 | return true; |
1157 | |
1158 | if (Result.is(K: tok::colon) && ModuleDeclState.isNamedModule()) { |
1159 | std::string Name = ModuleDeclState.getPrimaryName().str(); |
1160 | Name += ":" ; |
1161 | NamedModuleImportPath.push_back( |
1162 | Elt: {getIdentifierInfo(Name), Result.getLocation()}); |
1163 | CurLexerCallback = CLK_LexAfterModuleImport; |
1164 | return true; |
1165 | } |
1166 | } else { |
1167 | Lex(Result); |
1168 | } |
1169 | |
1170 | // Allocate a holding buffer for a sequence of tokens and introduce it into |
1171 | // the token stream. |
1172 | auto EnterTokens = [this](ArrayRef<Token> Toks) { |
1173 | auto ToksCopy = std::make_unique<Token[]>(num: Toks.size()); |
1174 | std::copy(Toks.begin(), Toks.end(), ToksCopy.get()); |
1175 | EnterTokenStream(Toks: std::move(ToksCopy), NumToks: Toks.size(), |
1176 | /*DisableMacroExpansion*/ true, /*IsReinject*/ false); |
1177 | }; |
1178 | |
1179 | bool = Result.is(K: tok::header_name); |
1180 | // Check for a header-name. |
1181 | SmallVector<Token, 32> Suffix; |
1182 | if (ImportingHeader) { |
1183 | // Enter the header-name token into the token stream; a Lex action cannot |
1184 | // both return a token and cache tokens (doing so would corrupt the token |
1185 | // cache if the call to Lex comes from CachingLex / PeekAhead). |
1186 | Suffix.push_back(Elt: Result); |
1187 | |
1188 | // Consume the pp-import-suffix and expand any macros in it now. We'll add |
1189 | // it back into the token stream later. |
1190 | CollectPpImportSuffix(Toks&: Suffix); |
1191 | if (Suffix.back().isNot(K: tok::semi)) { |
1192 | // This is not a pp-import after all. |
1193 | EnterTokens(Suffix); |
1194 | return false; |
1195 | } |
1196 | |
1197 | // C++2a [cpp.module]p1: |
1198 | // The ';' preprocessing-token terminating a pp-import shall not have |
1199 | // been produced by macro replacement. |
1200 | SourceLocation SemiLoc = Suffix.back().getLocation(); |
1201 | if (SemiLoc.isMacroID()) |
1202 | Diag(Loc: SemiLoc, DiagID: diag::err_header_import_semi_in_macro); |
1203 | |
1204 | // Reconstitute the import token. |
1205 | Token ImportTok; |
1206 | ImportTok.startToken(); |
1207 | ImportTok.setKind(tok::kw_import); |
1208 | ImportTok.setLocation(ModuleImportLoc); |
1209 | ImportTok.setIdentifierInfo(getIdentifierInfo(Name: "import" )); |
1210 | ImportTok.setLength(6); |
1211 | |
1212 | auto Action = HandleHeaderIncludeOrImport( |
1213 | /*HashLoc*/ SourceLocation(), IncludeTok&: ImportTok, FilenameTok&: Suffix.front(), EndLoc: SemiLoc); |
1214 | switch (Action.Kind) { |
1215 | case ImportAction::None: |
1216 | break; |
1217 | |
1218 | case ImportAction::ModuleBegin: |
1219 | // Let the parser know we're textually entering the module. |
1220 | Suffix.emplace_back(); |
1221 | Suffix.back().startToken(); |
1222 | Suffix.back().setKind(tok::annot_module_begin); |
1223 | Suffix.back().setLocation(SemiLoc); |
1224 | Suffix.back().setAnnotationEndLoc(SemiLoc); |
1225 | Suffix.back().setAnnotationValue(Action.ModuleForHeader); |
1226 | [[fallthrough]]; |
1227 | |
1228 | case ImportAction::ModuleImport: |
1229 | case ImportAction::HeaderUnitImport: |
1230 | case ImportAction::SkippedModuleImport: |
1231 | // We chose to import (or textually enter) the file. Convert the |
1232 | // header-name token into a header unit annotation token. |
1233 | Suffix[0].setKind(tok::annot_header_unit); |
1234 | Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation()); |
1235 | Suffix[0].setAnnotationValue(Action.ModuleForHeader); |
1236 | // FIXME: Call the moduleImport callback? |
1237 | break; |
1238 | case ImportAction::Failure: |
1239 | assert(TheModuleLoader.HadFatalFailure && |
1240 | "This should be an early exit only to a fatal error" ); |
1241 | Result.setKind(tok::eof); |
1242 | CurLexer->cutOffLexing(); |
1243 | EnterTokens(Suffix); |
1244 | return true; |
1245 | } |
1246 | |
1247 | EnterTokens(Suffix); |
1248 | return false; |
1249 | } |
1250 | |
1251 | // The token sequence |
1252 | // |
1253 | // import identifier (. identifier)* |
1254 | // |
1255 | // indicates a module import directive. We already saw the 'import' |
1256 | // contextual keyword, so now we're looking for the identifiers. |
1257 | if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) { |
1258 | // We expected to see an identifier here, and we did; continue handling |
1259 | // identifiers. |
1260 | NamedModuleImportPath.push_back( |
1261 | Elt: std::make_pair(x: Result.getIdentifierInfo(), y: Result.getLocation())); |
1262 | ModuleImportExpectsIdentifier = false; |
1263 | CurLexerCallback = CLK_LexAfterModuleImport; |
1264 | return true; |
1265 | } |
1266 | |
1267 | // If we're expecting a '.' or a ';', and we got a '.', then wait until we |
1268 | // see the next identifier. (We can also see a '[[' that begins an |
1269 | // attribute-specifier-seq here under the Standard C++ Modules.) |
1270 | if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) { |
1271 | ModuleImportExpectsIdentifier = true; |
1272 | CurLexerCallback = CLK_LexAfterModuleImport; |
1273 | return true; |
1274 | } |
1275 | |
1276 | // If we didn't recognize a module name at all, this is not a (valid) import. |
1277 | if (NamedModuleImportPath.empty() || Result.is(K: tok::eof)) |
1278 | return true; |
1279 | |
1280 | // Consume the pp-import-suffix and expand any macros in it now, if we're not |
1281 | // at the semicolon already. |
1282 | SourceLocation SemiLoc = Result.getLocation(); |
1283 | if (Result.isNot(K: tok::semi)) { |
1284 | Suffix.push_back(Elt: Result); |
1285 | CollectPpImportSuffix(Toks&: Suffix); |
1286 | if (Suffix.back().isNot(K: tok::semi)) { |
1287 | // This is not an import after all. |
1288 | EnterTokens(Suffix); |
1289 | return false; |
1290 | } |
1291 | SemiLoc = Suffix.back().getLocation(); |
1292 | } |
1293 | |
1294 | // Under the standard C++ Modules, the dot is just part of the module name, |
1295 | // and not a real hierarchy separator. Flatten such module names now. |
1296 | // |
1297 | // FIXME: Is this the right level to be performing this transformation? |
1298 | std::string FlatModuleName; |
1299 | if (getLangOpts().CPlusPlusModules) { |
1300 | for (auto &Piece : NamedModuleImportPath) { |
1301 | // If the FlatModuleName ends with colon, it implies it is a partition. |
1302 | if (!FlatModuleName.empty() && FlatModuleName.back() != ':') |
1303 | FlatModuleName += "." ; |
1304 | FlatModuleName += Piece.first->getName(); |
1305 | } |
1306 | SourceLocation FirstPathLoc = NamedModuleImportPath[0].second; |
1307 | NamedModuleImportPath.clear(); |
1308 | NamedModuleImportPath.push_back( |
1309 | Elt: std::make_pair(x: getIdentifierInfo(Name: FlatModuleName), y&: FirstPathLoc)); |
1310 | } |
1311 | |
1312 | Module *Imported = nullptr; |
1313 | // We don't/shouldn't load the standard c++20 modules when preprocessing. |
1314 | if (getLangOpts().Modules && !isInImportingCXXNamedModules()) { |
1315 | Imported = TheModuleLoader.loadModule(ImportLoc: ModuleImportLoc, |
1316 | Path: NamedModuleImportPath, |
1317 | Visibility: Module::Hidden, |
1318 | /*IsInclusionDirective=*/false); |
1319 | if (Imported) |
1320 | makeModuleVisible(M: Imported, Loc: SemiLoc); |
1321 | } |
1322 | |
1323 | if (Callbacks) |
1324 | Callbacks->moduleImport(ImportLoc: ModuleImportLoc, Path: NamedModuleImportPath, Imported); |
1325 | |
1326 | if (!Suffix.empty()) { |
1327 | EnterTokens(Suffix); |
1328 | return false; |
1329 | } |
1330 | return true; |
1331 | } |
1332 | |
1333 | void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) { |
1334 | CurSubmoduleState->VisibleModules.setVisible( |
1335 | M, Loc, Vis: [](Module *) {}, |
1336 | Cb: [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) { |
1337 | // FIXME: Include the path in the diagnostic. |
1338 | // FIXME: Include the import location for the conflicting module. |
1339 | Diag(Loc: ModuleImportLoc, DiagID: diag::warn_module_conflict) |
1340 | << Path[0]->getFullModuleName() |
1341 | << Conflict->getFullModuleName() |
1342 | << Message; |
1343 | }); |
1344 | |
1345 | // Add this module to the imports list of the currently-built submodule. |
1346 | if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M) |
1347 | BuildingSubmoduleStack.back().M->Imports.insert(X: M); |
1348 | } |
1349 | |
1350 | bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String, |
1351 | const char *DiagnosticTag, |
1352 | bool AllowMacroExpansion) { |
1353 | // We need at least one string literal. |
1354 | if (Result.isNot(K: tok::string_literal)) { |
1355 | Diag(Tok: Result, DiagID: diag::err_expected_string_literal) |
1356 | << /*Source='in...'*/0 << DiagnosticTag; |
1357 | return false; |
1358 | } |
1359 | |
1360 | // Lex string literal tokens, optionally with macro expansion. |
1361 | SmallVector<Token, 4> StrToks; |
1362 | do { |
1363 | StrToks.push_back(Elt: Result); |
1364 | |
1365 | if (Result.hasUDSuffix()) |
1366 | Diag(Tok: Result, DiagID: diag::err_invalid_string_udl); |
1367 | |
1368 | if (AllowMacroExpansion) |
1369 | Lex(Result); |
1370 | else |
1371 | LexUnexpandedToken(Result); |
1372 | } while (Result.is(K: tok::string_literal)); |
1373 | |
1374 | // Concatenate and parse the strings. |
1375 | StringLiteralParser Literal(StrToks, *this); |
1376 | assert(Literal.isOrdinary() && "Didn't allow wide strings in" ); |
1377 | |
1378 | if (Literal.hadError) |
1379 | return false; |
1380 | |
1381 | if (Literal.Pascal) { |
1382 | Diag(Loc: StrToks[0].getLocation(), DiagID: diag::err_expected_string_literal) |
1383 | << /*Source='in...'*/0 << DiagnosticTag; |
1384 | return false; |
1385 | } |
1386 | |
1387 | String = std::string(Literal.GetString()); |
1388 | return true; |
1389 | } |
1390 | |
1391 | bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) { |
1392 | assert(Tok.is(tok::numeric_constant)); |
1393 | SmallString<8> IntegerBuffer; |
1394 | bool NumberInvalid = false; |
1395 | StringRef Spelling = getSpelling(Tok, Buffer&: IntegerBuffer, Invalid: &NumberInvalid); |
1396 | if (NumberInvalid) |
1397 | return false; |
1398 | NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(), |
1399 | getLangOpts(), getTargetInfo(), |
1400 | getDiagnostics()); |
1401 | if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix()) |
1402 | return false; |
1403 | llvm::APInt APVal(64, 0); |
1404 | if (Literal.GetIntegerValue(Val&: APVal)) |
1405 | return false; |
1406 | Lex(Result&: Tok); |
1407 | Value = APVal.getLimitedValue(); |
1408 | return true; |
1409 | } |
1410 | |
1411 | void Preprocessor::addCommentHandler(CommentHandler *Handler) { |
1412 | assert(Handler && "NULL comment handler" ); |
1413 | assert(!llvm::is_contained(CommentHandlers, Handler) && |
1414 | "Comment handler already registered" ); |
1415 | CommentHandlers.push_back(x: Handler); |
1416 | } |
1417 | |
1418 | void Preprocessor::removeCommentHandler(CommentHandler *Handler) { |
1419 | std::vector<CommentHandler *>::iterator Pos = |
1420 | llvm::find(Range&: CommentHandlers, Val: Handler); |
1421 | assert(Pos != CommentHandlers.end() && "Comment handler not registered" ); |
1422 | CommentHandlers.erase(position: Pos); |
1423 | } |
1424 | |
1425 | bool Preprocessor::HandleComment(Token &result, SourceRange ) { |
1426 | bool AnyPendingTokens = false; |
1427 | for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(), |
1428 | HEnd = CommentHandlers.end(); |
1429 | H != HEnd; ++H) { |
1430 | if ((*H)->HandleComment(PP&: *this, Comment)) |
1431 | AnyPendingTokens = true; |
1432 | } |
1433 | if (!AnyPendingTokens || getCommentRetentionState()) |
1434 | return false; |
1435 | Lex(Result&: result); |
1436 | return true; |
1437 | } |
1438 | |
1439 | void Preprocessor::emitMacroDeprecationWarning(const Token &Identifier) const { |
1440 | const MacroAnnotations &A = |
1441 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1442 | assert(A.DeprecationInfo && |
1443 | "Macro deprecation warning without recorded annotation!" ); |
1444 | const MacroAnnotationInfo &Info = *A.DeprecationInfo; |
1445 | if (Info.Message.empty()) |
1446 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use) |
1447 | << Identifier.getIdentifierInfo() << 0; |
1448 | else |
1449 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use) |
1450 | << Identifier.getIdentifierInfo() << 1 << Info.Message; |
1451 | Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 0; |
1452 | } |
1453 | |
1454 | void Preprocessor::emitRestrictExpansionWarning(const Token &Identifier) const { |
1455 | const MacroAnnotations &A = |
1456 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1457 | assert(A.RestrictExpansionInfo && |
1458 | "Macro restricted expansion warning without recorded annotation!" ); |
1459 | const MacroAnnotationInfo &Info = *A.RestrictExpansionInfo; |
1460 | if (Info.Message.empty()) |
1461 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use) |
1462 | << Identifier.getIdentifierInfo() << 0; |
1463 | else |
1464 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use) |
1465 | << Identifier.getIdentifierInfo() << 1 << Info.Message; |
1466 | Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 1; |
1467 | } |
1468 | |
1469 | void Preprocessor::emitRestrictInfNaNWarning(const Token &Identifier, |
1470 | unsigned DiagSelection) const { |
1471 | Diag(Tok: Identifier, DiagID: diag::warn_fp_nan_inf_when_disabled) << DiagSelection << 1; |
1472 | } |
1473 | |
1474 | void Preprocessor::emitFinalMacroWarning(const Token &Identifier, |
1475 | bool IsUndef) const { |
1476 | const MacroAnnotations &A = |
1477 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1478 | assert(A.FinalAnnotationLoc && |
1479 | "Final macro warning without recorded annotation!" ); |
1480 | |
1481 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_final_macro) |
1482 | << Identifier.getIdentifierInfo() << (IsUndef ? 0 : 1); |
1483 | Diag(Loc: *A.FinalAnnotationLoc, DiagID: diag::note_pp_macro_annotation) << 2; |
1484 | } |
1485 | |
1486 | bool Preprocessor::isSafeBufferOptOut(const SourceManager &SourceMgr, |
1487 | const SourceLocation &Loc) const { |
1488 | // The lambda that tests if a `Loc` is in an opt-out region given one opt-out |
1489 | // region map: |
1490 | auto TestInMap = [&SourceMgr](const SafeBufferOptOutRegionsTy &Map, |
1491 | const SourceLocation &Loc) -> bool { |
1492 | // Try to find a region in `SafeBufferOptOutMap` where `Loc` is in: |
1493 | auto FirstRegionEndingAfterLoc = llvm::partition_point( |
1494 | Range: Map, P: [&SourceMgr, |
1495 | &Loc](const std::pair<SourceLocation, SourceLocation> &Region) { |
1496 | return SourceMgr.isBeforeInTranslationUnit(LHS: Region.second, RHS: Loc); |
1497 | }); |
1498 | |
1499 | if (FirstRegionEndingAfterLoc != Map.end()) { |
1500 | // To test if the start location of the found region precedes `Loc`: |
1501 | return SourceMgr.isBeforeInTranslationUnit( |
1502 | LHS: FirstRegionEndingAfterLoc->first, RHS: Loc); |
1503 | } |
1504 | // If we do not find a region whose end location passes `Loc`, we want to |
1505 | // check if the current region is still open: |
1506 | if (!Map.empty() && Map.back().first == Map.back().second) |
1507 | return SourceMgr.isBeforeInTranslationUnit(LHS: Map.back().first, RHS: Loc); |
1508 | return false; |
1509 | }; |
1510 | |
1511 | // What the following does: |
1512 | // |
1513 | // If `Loc` belongs to the local TU, we just look up `SafeBufferOptOutMap`. |
1514 | // Otherwise, `Loc` is from a loaded AST. We look up the |
1515 | // `LoadedSafeBufferOptOutMap` first to get the opt-out region map of the |
1516 | // loaded AST where `Loc` is at. Then we find if `Loc` is in an opt-out |
1517 | // region w.r.t. the region map. If the region map is absent, it means there |
1518 | // is no opt-out pragma in that loaded AST. |
1519 | // |
1520 | // Opt-out pragmas in the local TU or a loaded AST is not visible to another |
1521 | // one of them. That means if you put the pragmas around a `#include |
1522 | // "module.h"`, where module.h is a module, it is not actually suppressing |
1523 | // warnings in module.h. This is fine because warnings in module.h will be |
1524 | // reported when module.h is compiled in isolation and nothing in module.h |
1525 | // will be analyzed ever again. So you will not see warnings from the file |
1526 | // that imports module.h anyway. And you can't even do the same thing for PCHs |
1527 | // because they can only be included from the command line. |
1528 | |
1529 | if (SourceMgr.isLocalSourceLocation(Loc)) |
1530 | return TestInMap(SafeBufferOptOutMap, Loc); |
1531 | |
1532 | const SafeBufferOptOutRegionsTy *LoadedRegions = |
1533 | LoadedSafeBufferOptOutMap.lookupLoadedOptOutMap(Loc, SrcMgr: SourceMgr); |
1534 | |
1535 | if (LoadedRegions) |
1536 | return TestInMap(*LoadedRegions, Loc); |
1537 | return false; |
1538 | } |
1539 | |
1540 | bool Preprocessor::enterOrExitSafeBufferOptOutRegion( |
1541 | bool isEnter, const SourceLocation &Loc) { |
1542 | if (isEnter) { |
1543 | if (isPPInSafeBufferOptOutRegion()) |
1544 | return true; // invalid enter action |
1545 | InSafeBufferOptOutRegion = true; |
1546 | CurrentSafeBufferOptOutStart = Loc; |
1547 | |
1548 | // To set the start location of a new region: |
1549 | |
1550 | if (!SafeBufferOptOutMap.empty()) { |
1551 | [[maybe_unused]] auto *PrevRegion = &SafeBufferOptOutMap.back(); |
1552 | assert(PrevRegion->first != PrevRegion->second && |
1553 | "Shall not begin a safe buffer opt-out region before closing the " |
1554 | "previous one." ); |
1555 | } |
1556 | // If the start location equals to the end location, we call the region a |
1557 | // open region or a unclosed region (i.e., end location has not been set |
1558 | // yet). |
1559 | SafeBufferOptOutMap.emplace_back(Args: Loc, Args: Loc); |
1560 | } else { |
1561 | if (!isPPInSafeBufferOptOutRegion()) |
1562 | return true; // invalid enter action |
1563 | InSafeBufferOptOutRegion = false; |
1564 | |
1565 | // To set the end location of the current open region: |
1566 | |
1567 | assert(!SafeBufferOptOutMap.empty() && |
1568 | "Misordered safe buffer opt-out regions" ); |
1569 | auto *CurrRegion = &SafeBufferOptOutMap.back(); |
1570 | assert(CurrRegion->first == CurrRegion->second && |
1571 | "Set end location to a closed safe buffer opt-out region" ); |
1572 | CurrRegion->second = Loc; |
1573 | } |
1574 | return false; |
1575 | } |
1576 | |
1577 | bool Preprocessor::isPPInSafeBufferOptOutRegion() { |
1578 | return InSafeBufferOptOutRegion; |
1579 | } |
1580 | bool Preprocessor::isPPInSafeBufferOptOutRegion(SourceLocation &StartLoc) { |
1581 | StartLoc = CurrentSafeBufferOptOutStart; |
1582 | return InSafeBufferOptOutRegion; |
1583 | } |
1584 | |
1585 | SmallVector<SourceLocation, 64> |
1586 | Preprocessor::serializeSafeBufferOptOutMap() const { |
1587 | assert(!InSafeBufferOptOutRegion && |
1588 | "Attempt to serialize safe buffer opt-out regions before file being " |
1589 | "completely preprocessed" ); |
1590 | |
1591 | SmallVector<SourceLocation, 64> SrcSeq; |
1592 | |
1593 | for (const auto &[begin, end] : SafeBufferOptOutMap) { |
1594 | SrcSeq.push_back(Elt: begin); |
1595 | SrcSeq.push_back(Elt: end); |
1596 | } |
1597 | // Only `SafeBufferOptOutMap` gets serialized. No need to serialize |
1598 | // `LoadedSafeBufferOptOutMap` because if this TU loads a pch/module, every |
1599 | // pch/module in the pch-chain/module-DAG will be loaded one by one in order. |
1600 | // It means that for each loading pch/module m, it just needs to load m's own |
1601 | // `SafeBufferOptOutMap`. |
1602 | return SrcSeq; |
1603 | } |
1604 | |
1605 | bool Preprocessor::setDeserializedSafeBufferOptOutMap( |
1606 | const SmallVectorImpl<SourceLocation> &SourceLocations) { |
1607 | if (SourceLocations.size() == 0) |
1608 | return false; |
1609 | |
1610 | assert(SourceLocations.size() % 2 == 0 && |
1611 | "ill-formed SourceLocation sequence" ); |
1612 | |
1613 | auto It = SourceLocations.begin(); |
1614 | SafeBufferOptOutRegionsTy &Regions = |
1615 | LoadedSafeBufferOptOutMap.findAndConsLoadedOptOutMap(Loc: *It, SrcMgr&: SourceMgr); |
1616 | |
1617 | do { |
1618 | SourceLocation Begin = *It++; |
1619 | SourceLocation End = *It++; |
1620 | |
1621 | Regions.emplace_back(Args&: Begin, Args&: End); |
1622 | } while (It != SourceLocations.end()); |
1623 | return true; |
1624 | } |
1625 | |
1626 | ModuleLoader::~ModuleLoader() = default; |
1627 | |
1628 | CommentHandler::~CommentHandler() = default; |
1629 | |
1630 | EmptylineHandler::~EmptylineHandler() = default; |
1631 | |
1632 | CodeCompletionHandler::~CodeCompletionHandler() = default; |
1633 | |
1634 | void Preprocessor::createPreprocessingRecord() { |
1635 | if (Record) |
1636 | return; |
1637 | |
1638 | Record = new PreprocessingRecord(getSourceManager()); |
1639 | addPPCallbacks(C: std::unique_ptr<PPCallbacks>(Record)); |
1640 | } |
1641 | |
1642 | const char *Preprocessor::getCheckPoint(FileID FID, const char *Start) const { |
1643 | if (auto It = CheckPoints.find(Val: FID); It != CheckPoints.end()) { |
1644 | const SmallVector<const char *> &FileCheckPoints = It->second; |
1645 | const char *Last = nullptr; |
1646 | // FIXME: Do better than a linear search. |
1647 | for (const char *P : FileCheckPoints) { |
1648 | if (P > Start) |
1649 | break; |
1650 | Last = P; |
1651 | } |
1652 | return Last; |
1653 | } |
1654 | |
1655 | return nullptr; |
1656 | } |
1657 | |