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/IdentifierTable.h" |
31 | #include "clang/Basic/LLVM.h" |
32 | #include "clang/Basic/LangOptions.h" |
33 | #include "clang/Basic/Module.h" |
34 | #include "clang/Basic/SourceLocation.h" |
35 | #include "clang/Basic/SourceManager.h" |
36 | #include "clang/Basic/TargetInfo.h" |
37 | #include "clang/Lex/CodeCompletionHandler.h" |
38 | #include "clang/Lex/ExternalPreprocessorSource.h" |
39 | #include "clang/Lex/HeaderSearch.h" |
40 | #include "clang/Lex/LexDiagnostic.h" |
41 | #include "clang/Lex/Lexer.h" |
42 | #include "clang/Lex/LiteralSupport.h" |
43 | #include "clang/Lex/MacroArgs.h" |
44 | #include "clang/Lex/MacroInfo.h" |
45 | #include "clang/Lex/ModuleLoader.h" |
46 | #include "clang/Lex/Pragma.h" |
47 | #include "clang/Lex/PreprocessingRecord.h" |
48 | #include "clang/Lex/PreprocessorLexer.h" |
49 | #include "clang/Lex/PreprocessorOptions.h" |
50 | #include "clang/Lex/ScratchBuffer.h" |
51 | #include "clang/Lex/Token.h" |
52 | #include "clang/Lex/TokenLexer.h" |
53 | #include "llvm/ADT/APInt.h" |
54 | #include "llvm/ADT/ArrayRef.h" |
55 | #include "llvm/ADT/DenseMap.h" |
56 | #include "llvm/ADT/STLExtras.h" |
57 | #include "llvm/ADT/SmallVector.h" |
58 | #include "llvm/ADT/StringRef.h" |
59 | #include "llvm/Support/Capacity.h" |
60 | #include "llvm/Support/ErrorHandling.h" |
61 | #include "llvm/Support/MemoryBuffer.h" |
62 | #include "llvm/Support/raw_ostream.h" |
63 | #include <algorithm> |
64 | #include <cassert> |
65 | #include <memory> |
66 | #include <optional> |
67 | #include <string> |
68 | #include <utility> |
69 | #include <vector> |
70 | |
71 | using namespace clang; |
72 | |
73 | /// Minimum distance between two check points, in tokens. |
74 | static constexpr unsigned CheckPointStepSize = 1024; |
75 | |
76 | LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry) |
77 | |
78 | ExternalPreprocessorSource::~ExternalPreprocessorSource() = default; |
79 | |
80 | Preprocessor::(const PreprocessorOptions &PPOpts, |
81 | DiagnosticsEngine &diags, const LangOptions &opts, |
82 | SourceManager &SM, HeaderSearch &, |
83 | ModuleLoader &TheModuleLoader, |
84 | IdentifierInfoLookup *IILookup, bool , |
85 | TranslationUnitKind TUKind) |
86 | : PPOpts(PPOpts), Diags(&diags), LangOpts(opts), |
87 | FileMgr(Headers.getFileMgr()), SourceMgr(SM), |
88 | ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers), |
89 | TheModuleLoader(TheModuleLoader), ExternalSource(nullptr), |
90 | // As the language options may have not been loaded yet (when |
91 | // deserializing an ASTUnit), adding keywords to the identifier table is |
92 | // deferred to Preprocessor::Initialize(). |
93 | Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())), |
94 | TUKind(TUKind), SkipMainFilePreamble(0, true), |
95 | CurSubmoduleState(&NullSubmoduleState) { |
96 | OwnsHeaderSearch = OwnsHeaders; |
97 | |
98 | // Default to discarding comments. |
99 | KeepComments = false; |
100 | KeepMacroComments = false; |
101 | SuppressIncludeNotFoundError = false; |
102 | |
103 | // Macro expansion is enabled. |
104 | DisableMacroExpansion = false; |
105 | MacroExpansionInDirectivesOverride = false; |
106 | InMacroArgs = false; |
107 | ArgMacro = nullptr; |
108 | InMacroArgPreExpansion = false; |
109 | NumCachedTokenLexers = 0; |
110 | PragmasEnabled = true; |
111 | ParsingIfOrElifDirective = false; |
112 | PreprocessedOutput = false; |
113 | |
114 | // We haven't read anything from the external source. |
115 | ReadMacrosFromExternalSource = false; |
116 | |
117 | BuiltinInfo = std::make_unique<Builtin::Context>(); |
118 | |
119 | // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of |
120 | // a macro. They get unpoisoned where it is allowed. |
121 | (Ident__VA_ARGS__ = getIdentifierInfo(Name: "__VA_ARGS__" ))->setIsPoisoned(); |
122 | SetPoisonReason(II: Ident__VA_ARGS__,DiagID: diag::ext_pp_bad_vaargs_use); |
123 | (Ident__VA_OPT__ = getIdentifierInfo(Name: "__VA_OPT__" ))->setIsPoisoned(); |
124 | SetPoisonReason(II: Ident__VA_OPT__,DiagID: diag::ext_pp_bad_vaopt_use); |
125 | |
126 | // Initialize the pragma handlers. |
127 | RegisterBuiltinPragmas(); |
128 | |
129 | // Initialize builtin macros like __LINE__ and friends. |
130 | RegisterBuiltinMacros(); |
131 | |
132 | if(LangOpts.Borland) { |
133 | Ident__exception_info = getIdentifierInfo(Name: "_exception_info" ); |
134 | Ident___exception_info = getIdentifierInfo(Name: "__exception_info" ); |
135 | Ident_GetExceptionInfo = getIdentifierInfo(Name: "GetExceptionInformation" ); |
136 | Ident__exception_code = getIdentifierInfo(Name: "_exception_code" ); |
137 | Ident___exception_code = getIdentifierInfo(Name: "__exception_code" ); |
138 | Ident_GetExceptionCode = getIdentifierInfo(Name: "GetExceptionCode" ); |
139 | Ident__abnormal_termination = getIdentifierInfo(Name: "_abnormal_termination" ); |
140 | Ident___abnormal_termination = getIdentifierInfo(Name: "__abnormal_termination" ); |
141 | Ident_AbnormalTermination = getIdentifierInfo(Name: "AbnormalTermination" ); |
142 | } else { |
143 | Ident__exception_info = Ident__exception_code = nullptr; |
144 | Ident__abnormal_termination = Ident___exception_info = nullptr; |
145 | Ident___exception_code = Ident___abnormal_termination = nullptr; |
146 | Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr; |
147 | Ident_AbnormalTermination = nullptr; |
148 | } |
149 | |
150 | // Default incremental processing to -fincremental-extensions, clients can |
151 | // override with `enableIncrementalProcessing` if desired. |
152 | IncrementalProcessing = LangOpts.IncrementalExtensions; |
153 | |
154 | // If using a PCH where a #pragma hdrstop is expected, start skipping tokens. |
155 | if (usingPCHWithPragmaHdrStop()) |
156 | SkippingUntilPragmaHdrStop = true; |
157 | |
158 | // If using a PCH with a through header, start skipping tokens. |
159 | if (!this->PPOpts.PCHThroughHeader.empty() && |
160 | !this->PPOpts.ImplicitPCHInclude.empty()) |
161 | SkippingUntilPCHThroughHeader = true; |
162 | |
163 | if (this->PPOpts.GeneratePreamble) |
164 | PreambleConditionalStack.startRecording(); |
165 | |
166 | MaxTokens = LangOpts.MaxTokens; |
167 | } |
168 | |
169 | Preprocessor::~Preprocessor() { |
170 | assert(!isBacktrackEnabled() && "EnableBacktrack/Backtrack imbalance!" ); |
171 | |
172 | IncludeMacroStack.clear(); |
173 | |
174 | // Free any cached macro expanders. |
175 | // This populates MacroArgCache, so all TokenLexers need to be destroyed |
176 | // before the code below that frees up the MacroArgCache list. |
177 | std::fill(first: TokenLexerCache, last: TokenLexerCache + NumCachedTokenLexers, value: nullptr); |
178 | CurTokenLexer.reset(); |
179 | |
180 | // Free any cached MacroArgs. |
181 | for (MacroArgs *ArgList = MacroArgCache; ArgList;) |
182 | ArgList = ArgList->deallocate(); |
183 | |
184 | // Delete the header search info, if we own it. |
185 | if (OwnsHeaderSearch) |
186 | delete &HeaderInfo; |
187 | } |
188 | |
189 | void Preprocessor::Initialize(const TargetInfo &Target, |
190 | const TargetInfo *AuxTarget) { |
191 | assert((!this->Target || this->Target == &Target) && |
192 | "Invalid override of target information" ); |
193 | this->Target = &Target; |
194 | |
195 | assert((!this->AuxTarget || this->AuxTarget == AuxTarget) && |
196 | "Invalid override of aux target information." ); |
197 | this->AuxTarget = AuxTarget; |
198 | |
199 | // Initialize information about built-ins. |
200 | BuiltinInfo->InitializeTarget(Target, AuxTarget); |
201 | HeaderInfo.setTarget(Target); |
202 | |
203 | // Populate the identifier table with info about keywords for the current language. |
204 | Identifiers.AddKeywords(LangOpts); |
205 | |
206 | // Initialize the __FTL_EVAL_METHOD__ macro to the TargetInfo. |
207 | setTUFPEvalMethod(getTargetInfo().getFPEvalMethod()); |
208 | |
209 | if (getLangOpts().getFPEvalMethod() == LangOptions::FEM_UnsetOnCommandLine) |
210 | // Use setting from TargetInfo. |
211 | setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: Target.getFPEvalMethod()); |
212 | else |
213 | // Set initial value of __FLT_EVAL_METHOD__ from the command line. |
214 | setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: getLangOpts().getFPEvalMethod()); |
215 | } |
216 | |
217 | void Preprocessor::InitializeForModelFile() { |
218 | NumEnteredSourceFiles = 0; |
219 | |
220 | // Reset pragmas |
221 | PragmaHandlersBackup = std::move(PragmaHandlers); |
222 | PragmaHandlers = std::make_unique<PragmaNamespace>(args: StringRef()); |
223 | RegisterBuiltinPragmas(); |
224 | |
225 | // Reset PredefinesFileID |
226 | PredefinesFileID = FileID(); |
227 | } |
228 | |
229 | void Preprocessor::FinalizeForModelFile() { |
230 | NumEnteredSourceFiles = 1; |
231 | |
232 | PragmaHandlers = std::move(PragmaHandlersBackup); |
233 | } |
234 | |
235 | void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const { |
236 | llvm::errs() << tok::getTokenName(Kind: Tok.getKind()); |
237 | |
238 | if (!Tok.isAnnotation()) |
239 | llvm::errs() << " '" << getSpelling(Tok) << "'" ; |
240 | |
241 | if (!DumpFlags) return; |
242 | |
243 | llvm::errs() << "\t" ; |
244 | if (Tok.isAtStartOfLine()) |
245 | llvm::errs() << " [StartOfLine]" ; |
246 | if (Tok.hasLeadingSpace()) |
247 | llvm::errs() << " [LeadingSpace]" ; |
248 | if (Tok.isExpandDisabled()) |
249 | llvm::errs() << " [ExpandDisabled]" ; |
250 | if (Tok.isFirstPPToken()) |
251 | llvm::errs() << " [First pp-token]" ; |
252 | if (Tok.needsCleaning()) { |
253 | const char *Start = SourceMgr.getCharacterData(SL: Tok.getLocation()); |
254 | llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength()) |
255 | << "']" ; |
256 | } |
257 | |
258 | llvm::errs() << "\tLoc=<" ; |
259 | DumpLocation(Loc: Tok.getLocation()); |
260 | llvm::errs() << ">" ; |
261 | } |
262 | |
263 | void Preprocessor::DumpLocation(SourceLocation Loc) const { |
264 | Loc.print(OS&: llvm::errs(), SM: SourceMgr); |
265 | } |
266 | |
267 | void Preprocessor::DumpMacro(const MacroInfo &MI) const { |
268 | llvm::errs() << "MACRO: " ; |
269 | for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) { |
270 | DumpToken(Tok: MI.getReplacementToken(Tok: i)); |
271 | llvm::errs() << " " ; |
272 | } |
273 | llvm::errs() << "\n" ; |
274 | } |
275 | |
276 | void Preprocessor::PrintStats() { |
277 | llvm::errs() << "\n*** Preprocessor Stats:\n" ; |
278 | llvm::errs() << NumDirectives << " directives found:\n" ; |
279 | llvm::errs() << " " << NumDefined << " #define.\n" ; |
280 | llvm::errs() << " " << NumUndefined << " #undef.\n" ; |
281 | llvm::errs() << " #include/#include_next/#import:\n" ; |
282 | llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n" ; |
283 | llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n" ; |
284 | llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n" ; |
285 | llvm::errs() << " " << NumElse << " #else/#elif/#elifdef/#elifndef.\n" ; |
286 | llvm::errs() << " " << NumEndif << " #endif.\n" ; |
287 | llvm::errs() << " " << NumPragma << " #pragma.\n" ; |
288 | llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n" ; |
289 | |
290 | llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/" |
291 | << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, " |
292 | << NumFastMacroExpanded << " on the fast path.\n" ; |
293 | llvm::errs() << (NumFastTokenPaste+NumTokenPaste) |
294 | << " token paste (##) operations performed, " |
295 | << NumFastTokenPaste << " on the fast path.\n" ; |
296 | |
297 | llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total" ; |
298 | |
299 | llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory(); |
300 | llvm::errs() << "\n Macro Expanded Tokens: " |
301 | << llvm::capacity_in_bytes(X: MacroExpandedTokens); |
302 | llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity(); |
303 | // FIXME: List information for all submodules. |
304 | llvm::errs() << "\n Macros: " |
305 | << llvm::capacity_in_bytes(X: CurSubmoduleState->Macros); |
306 | llvm::errs() << "\n #pragma push_macro Info: " |
307 | << llvm::capacity_in_bytes(X: PragmaPushMacroInfo); |
308 | llvm::errs() << "\n Poison Reasons: " |
309 | << llvm::capacity_in_bytes(X: PoisonReasons); |
310 | llvm::errs() << "\n Comment Handlers: " |
311 | << llvm::capacity_in_bytes(x: CommentHandlers) << "\n" ; |
312 | } |
313 | |
314 | Preprocessor::macro_iterator |
315 | Preprocessor::macro_begin(bool IncludeExternalMacros) const { |
316 | if (IncludeExternalMacros && ExternalSource && |
317 | !ReadMacrosFromExternalSource) { |
318 | ReadMacrosFromExternalSource = true; |
319 | ExternalSource->ReadDefinedMacros(); |
320 | } |
321 | |
322 | // Make sure we cover all macros in visible modules. |
323 | for (const ModuleMacro &Macro : ModuleMacros) |
324 | CurSubmoduleState->Macros.try_emplace(Key: Macro.II); |
325 | |
326 | return CurSubmoduleState->Macros.begin(); |
327 | } |
328 | |
329 | size_t Preprocessor::getTotalMemory() const { |
330 | return BP.getTotalMemory() |
331 | + llvm::capacity_in_bytes(X: MacroExpandedTokens) |
332 | + Predefines.capacity() /* Predefines buffer. */ |
333 | // FIXME: Include sizes from all submodules, and include MacroInfo sizes, |
334 | // and ModuleMacros. |
335 | + llvm::capacity_in_bytes(X: CurSubmoduleState->Macros) |
336 | + llvm::capacity_in_bytes(X: PragmaPushMacroInfo) |
337 | + llvm::capacity_in_bytes(X: PoisonReasons) |
338 | + llvm::capacity_in_bytes(x: CommentHandlers); |
339 | } |
340 | |
341 | Preprocessor::macro_iterator |
342 | Preprocessor::macro_end(bool IncludeExternalMacros) const { |
343 | if (IncludeExternalMacros && ExternalSource && |
344 | !ReadMacrosFromExternalSource) { |
345 | ReadMacrosFromExternalSource = true; |
346 | ExternalSource->ReadDefinedMacros(); |
347 | } |
348 | |
349 | return CurSubmoduleState->Macros.end(); |
350 | } |
351 | |
352 | /// Compares macro tokens with a specified token value sequence. |
353 | static bool MacroDefinitionEquals(const MacroInfo *MI, |
354 | ArrayRef<TokenValue> Tokens) { |
355 | return Tokens.size() == MI->getNumTokens() && |
356 | std::equal(first1: Tokens.begin(), last1: Tokens.end(), first2: MI->tokens_begin()); |
357 | } |
358 | |
359 | StringRef Preprocessor::getLastMacroWithSpelling( |
360 | SourceLocation Loc, |
361 | ArrayRef<TokenValue> Tokens) const { |
362 | SourceLocation BestLocation; |
363 | StringRef BestSpelling; |
364 | for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end(); |
365 | I != E; ++I) { |
366 | const MacroDirective::DefInfo |
367 | Def = I->second.findDirectiveAtLoc(Loc, SourceMgr); |
368 | if (!Def || !Def.getMacroInfo()) |
369 | continue; |
370 | if (!Def.getMacroInfo()->isObjectLike()) |
371 | continue; |
372 | if (!MacroDefinitionEquals(MI: Def.getMacroInfo(), Tokens)) |
373 | continue; |
374 | SourceLocation Location = Def.getLocation(); |
375 | // Choose the macro defined latest. |
376 | if (BestLocation.isInvalid() || |
377 | (Location.isValid() && |
378 | SourceMgr.isBeforeInTranslationUnit(LHS: BestLocation, RHS: Location))) { |
379 | BestLocation = Location; |
380 | BestSpelling = I->first->getName(); |
381 | } |
382 | } |
383 | return BestSpelling; |
384 | } |
385 | |
386 | void Preprocessor::recomputeCurLexerKind() { |
387 | if (CurLexer) |
388 | CurLexerCallback = CurLexer->isDependencyDirectivesLexer() |
389 | ? CLK_DependencyDirectivesLexer |
390 | : CLK_Lexer; |
391 | else if (CurTokenLexer) |
392 | CurLexerCallback = CLK_TokenLexer; |
393 | else |
394 | CurLexerCallback = CLK_CachingLexer; |
395 | } |
396 | |
397 | bool Preprocessor::SetCodeCompletionPoint(FileEntryRef File, |
398 | unsigned CompleteLine, |
399 | unsigned CompleteColumn) { |
400 | assert(CompleteLine && CompleteColumn && "Starts from 1:1" ); |
401 | assert(!CodeCompletionFile && "Already set" ); |
402 | |
403 | // Load the actual file's contents. |
404 | std::optional<llvm::MemoryBufferRef> Buffer = |
405 | SourceMgr.getMemoryBufferForFileOrNone(File); |
406 | if (!Buffer) |
407 | return true; |
408 | |
409 | // Find the byte position of the truncation point. |
410 | const char *Position = Buffer->getBufferStart(); |
411 | for (unsigned Line = 1; Line < CompleteLine; ++Line) { |
412 | for (; *Position; ++Position) { |
413 | if (*Position != '\r' && *Position != '\n') |
414 | continue; |
415 | |
416 | // Eat \r\n or \n\r as a single line. |
417 | if ((Position[1] == '\r' || Position[1] == '\n') && |
418 | Position[0] != Position[1]) |
419 | ++Position; |
420 | ++Position; |
421 | break; |
422 | } |
423 | } |
424 | |
425 | Position += CompleteColumn - 1; |
426 | |
427 | // If pointing inside the preamble, adjust the position at the beginning of |
428 | // the file after the preamble. |
429 | if (SkipMainFilePreamble.first && |
430 | SourceMgr.getFileEntryForID(FID: SourceMgr.getMainFileID()) == File) { |
431 | if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first) |
432 | Position = Buffer->getBufferStart() + SkipMainFilePreamble.first; |
433 | } |
434 | |
435 | if (Position > Buffer->getBufferEnd()) |
436 | Position = Buffer->getBufferEnd(); |
437 | |
438 | CodeCompletionFile = File; |
439 | CodeCompletionOffset = Position - Buffer->getBufferStart(); |
440 | |
441 | auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer( |
442 | Size: Buffer->getBufferSize() + 1, BufferName: Buffer->getBufferIdentifier()); |
443 | char *NewBuf = NewBuffer->getBufferStart(); |
444 | char *NewPos = std::copy(first: Buffer->getBufferStart(), last: Position, result: NewBuf); |
445 | *NewPos = '\0'; |
446 | std::copy(first: Position, last: Buffer->getBufferEnd(), result: NewPos+1); |
447 | SourceMgr.overrideFileContents(SourceFile: File, Buffer: std::move(NewBuffer)); |
448 | |
449 | return false; |
450 | } |
451 | |
452 | void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir, |
453 | bool IsAngled) { |
454 | setCodeCompletionReached(); |
455 | if (CodeComplete) |
456 | CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled); |
457 | } |
458 | |
459 | void Preprocessor::CodeCompleteNaturalLanguage() { |
460 | setCodeCompletionReached(); |
461 | if (CodeComplete) |
462 | CodeComplete->CodeCompleteNaturalLanguage(); |
463 | } |
464 | |
465 | /// getSpelling - This method is used to get the spelling of a token into a |
466 | /// SmallVector. Note that the returned StringRef may not point to the |
467 | /// supplied buffer if a copy can be avoided. |
468 | StringRef Preprocessor::getSpelling(const Token &Tok, |
469 | SmallVectorImpl<char> &Buffer, |
470 | bool *Invalid) const { |
471 | // NOTE: this has to be checked *before* testing for an IdentifierInfo. |
472 | if (Tok.isNot(K: tok::raw_identifier) && !Tok.hasUCN()) { |
473 | // Try the fast path. |
474 | if (const IdentifierInfo *II = Tok.getIdentifierInfo()) |
475 | return II->getName(); |
476 | } |
477 | |
478 | // Resize the buffer if we need to copy into it. |
479 | if (Tok.needsCleaning()) |
480 | Buffer.resize(N: Tok.getLength()); |
481 | |
482 | const char *Ptr = Buffer.data(); |
483 | unsigned Len = getSpelling(Tok, Buffer&: Ptr, Invalid); |
484 | return StringRef(Ptr, Len); |
485 | } |
486 | |
487 | /// CreateString - Plop the specified string into a scratch buffer and return a |
488 | /// location for it. If specified, the source location provides a source |
489 | /// location for the token. |
490 | void Preprocessor::CreateString(StringRef Str, Token &Tok, |
491 | SourceLocation ExpansionLocStart, |
492 | SourceLocation ExpansionLocEnd) { |
493 | Tok.setLength(Str.size()); |
494 | |
495 | const char *DestPtr; |
496 | SourceLocation Loc = ScratchBuf->getToken(Buf: Str.data(), Len: Str.size(), DestPtr); |
497 | |
498 | if (ExpansionLocStart.isValid()) |
499 | Loc = SourceMgr.createExpansionLoc(SpellingLoc: Loc, ExpansionLocStart, |
500 | ExpansionLocEnd, Length: Str.size()); |
501 | Tok.setLocation(Loc); |
502 | |
503 | // If this is a raw identifier or a literal token, set the pointer data. |
504 | if (Tok.is(K: tok::raw_identifier)) |
505 | Tok.setRawIdentifierData(DestPtr); |
506 | else if (Tok.isLiteral()) |
507 | Tok.setLiteralData(DestPtr); |
508 | } |
509 | |
510 | SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) { |
511 | auto &SM = getSourceManager(); |
512 | SourceLocation SpellingLoc = SM.getSpellingLoc(Loc); |
513 | FileIDAndOffset LocInfo = SM.getDecomposedLoc(Loc: SpellingLoc); |
514 | bool Invalid = false; |
515 | StringRef Buffer = SM.getBufferData(FID: LocInfo.first, Invalid: &Invalid); |
516 | if (Invalid) |
517 | return SourceLocation(); |
518 | |
519 | // FIXME: We could consider re-using spelling for tokens we see repeatedly. |
520 | const char *DestPtr; |
521 | SourceLocation Spelling = |
522 | ScratchBuf->getToken(Buf: Buffer.data() + LocInfo.second, Len: Length, DestPtr); |
523 | return SM.createTokenSplitLoc(SpellingLoc: Spelling, TokenStart: Loc, TokenEnd: Loc.getLocWithOffset(Offset: Length)); |
524 | } |
525 | |
526 | Module *Preprocessor::getCurrentModule() { |
527 | if (!getLangOpts().isCompilingModule()) |
528 | return nullptr; |
529 | |
530 | return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().CurrentModule); |
531 | } |
532 | |
533 | Module *Preprocessor::getCurrentModuleImplementation() { |
534 | if (!getLangOpts().isCompilingModuleImplementation()) |
535 | return nullptr; |
536 | |
537 | return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().ModuleName); |
538 | } |
539 | |
540 | //===----------------------------------------------------------------------===// |
541 | // Preprocessor Initialization Methods |
542 | //===----------------------------------------------------------------------===// |
543 | |
544 | /// EnterMainSourceFile - Enter the specified FileID as the main source file, |
545 | /// which implicitly adds the builtin defines etc. |
546 | void Preprocessor::EnterMainSourceFile() { |
547 | // We do not allow the preprocessor to reenter the main file. Doing so will |
548 | // cause FileID's to accumulate information from both runs (e.g. #line |
549 | // information) and predefined macros aren't guaranteed to be set properly. |
550 | assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!" ); |
551 | FileID MainFileID = SourceMgr.getMainFileID(); |
552 | |
553 | // If MainFileID is loaded it means we loaded an AST file, no need to enter |
554 | // a main file. |
555 | if (!SourceMgr.isLoadedFileID(FID: MainFileID)) { |
556 | // Enter the main file source buffer. |
557 | EnterSourceFile(FID: MainFileID, Dir: nullptr, Loc: SourceLocation()); |
558 | |
559 | // If we've been asked to skip bytes in the main file (e.g., as part of a |
560 | // precompiled preamble), do so now. |
561 | if (SkipMainFilePreamble.first > 0) |
562 | CurLexer->SetByteOffset(Offset: SkipMainFilePreamble.first, |
563 | StartOfLine: SkipMainFilePreamble.second); |
564 | |
565 | // Tell the header info that the main file was entered. If the file is later |
566 | // #imported, it won't be re-entered. |
567 | if (OptionalFileEntryRef FE = SourceMgr.getFileEntryRefForID(FID: MainFileID)) |
568 | markIncluded(File: *FE); |
569 | |
570 | // Record the first PP token in the main file. This is used to generate |
571 | // better diagnostics for C++ modules. |
572 | // |
573 | // // This is a comment. |
574 | // #define FOO int // note: add 'module;' to the start of the file |
575 | // ^ FirstPPToken // to introduce a global module fragment. |
576 | // |
577 | // export module M; // error: module declaration must occur |
578 | // // at the start of the translation unit. |
579 | if (getLangOpts().CPlusPlusModules) { |
580 | std::optional<Token> FirstPPTok = CurLexer->peekNextPPToken(); |
581 | if (FirstPPTok && FirstPPTok->isFirstPPToken()) |
582 | FirstPPTokenLoc = FirstPPTok->getLocation(); |
583 | } |
584 | } |
585 | |
586 | // Preprocess Predefines to populate the initial preprocessor state. |
587 | std::unique_ptr<llvm::MemoryBuffer> SB = |
588 | llvm::MemoryBuffer::getMemBufferCopy(InputData: Predefines, BufferName: "<built-in>" ); |
589 | assert(SB && "Cannot create predefined source buffer" ); |
590 | FileID FID = SourceMgr.createFileID(Buffer: std::move(SB)); |
591 | assert(FID.isValid() && "Could not create FileID for predefines?" ); |
592 | setPredefinesFileID(FID); |
593 | |
594 | // Start parsing the predefines. |
595 | EnterSourceFile(FID, Dir: nullptr, Loc: SourceLocation()); |
596 | |
597 | if (!PPOpts.PCHThroughHeader.empty()) { |
598 | // Lookup and save the FileID for the through header. If it isn't found |
599 | // in the search path, it's a fatal error. |
600 | OptionalFileEntryRef File = LookupFile( |
601 | FilenameLoc: SourceLocation(), Filename: PPOpts.PCHThroughHeader, |
602 | /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, |
603 | /*CurDir=*/nullptr, /*SearchPath=*/nullptr, /*RelativePath=*/nullptr, |
604 | /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr, |
605 | /*IsFrameworkFound=*/nullptr); |
606 | if (!File) { |
607 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_found) |
608 | << PPOpts.PCHThroughHeader; |
609 | return; |
610 | } |
611 | setPCHThroughHeaderFileID( |
612 | SourceMgr.createFileID(SourceFile: *File, IncludePos: SourceLocation(), FileCharacter: SrcMgr::C_User)); |
613 | } |
614 | |
615 | // Skip tokens from the Predefines and if needed the main file. |
616 | if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) || |
617 | (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop)) |
618 | SkipTokensWhileUsingPCH(); |
619 | } |
620 | |
621 | void Preprocessor::(FileID FID) { |
622 | assert(PCHThroughHeaderFileID.isInvalid() && |
623 | "PCHThroughHeaderFileID already set!" ); |
624 | PCHThroughHeaderFileID = FID; |
625 | } |
626 | |
627 | bool Preprocessor::(const FileEntry *FE) { |
628 | assert(PCHThroughHeaderFileID.isValid() && |
629 | "Invalid PCH through header FileID" ); |
630 | return FE == SourceMgr.getFileEntryForID(FID: PCHThroughHeaderFileID); |
631 | } |
632 | |
633 | bool Preprocessor::() { |
634 | return TUKind == TU_Prefix && !PPOpts.PCHThroughHeader.empty() && |
635 | PCHThroughHeaderFileID.isValid(); |
636 | } |
637 | |
638 | bool Preprocessor::() { |
639 | return TUKind != TU_Prefix && !PPOpts.PCHThroughHeader.empty() && |
640 | PCHThroughHeaderFileID.isValid(); |
641 | } |
642 | |
643 | bool Preprocessor::creatingPCHWithPragmaHdrStop() { |
644 | return TUKind == TU_Prefix && PPOpts.PCHWithHdrStop; |
645 | } |
646 | |
647 | bool Preprocessor::usingPCHWithPragmaHdrStop() { |
648 | return TUKind != TU_Prefix && PPOpts.PCHWithHdrStop; |
649 | } |
650 | |
651 | /// Skip tokens until after the #include of the through header or |
652 | /// until after a #pragma hdrstop is seen. Tokens in the predefines file |
653 | /// and the main file may be skipped. If the end of the predefines file |
654 | /// is reached, skipping continues into the main file. If the end of the |
655 | /// main file is reached, it's a fatal error. |
656 | void Preprocessor::SkipTokensWhileUsingPCH() { |
657 | bool ReachedMainFileEOF = false; |
658 | bool = SkippingUntilPCHThroughHeader; |
659 | bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop; |
660 | Token Tok; |
661 | while (true) { |
662 | bool InPredefines = |
663 | (CurLexer && CurLexer->getFileID() == getPredefinesFileID()); |
664 | CurLexerCallback(*this, Tok); |
665 | if (Tok.is(K: tok::eof) && !InPredefines) { |
666 | ReachedMainFileEOF = true; |
667 | break; |
668 | } |
669 | if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader) |
670 | break; |
671 | if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop) |
672 | break; |
673 | } |
674 | if (ReachedMainFileEOF) { |
675 | if (UsingPCHThroughHeader) |
676 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_seen) |
677 | << PPOpts.PCHThroughHeader << 1; |
678 | else if (!PPOpts.PCHWithHdrStopCreate) |
679 | Diag(Loc: SourceLocation(), DiagID: diag::err_pp_pragma_hdrstop_not_seen); |
680 | } |
681 | } |
682 | |
683 | void Preprocessor::replayPreambleConditionalStack() { |
684 | // Restore the conditional stack from the preamble, if there is one. |
685 | if (PreambleConditionalStack.isReplaying()) { |
686 | assert(CurPPLexer && |
687 | "CurPPLexer is null when calling replayPreambleConditionalStack." ); |
688 | CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack()); |
689 | PreambleConditionalStack.doneReplaying(); |
690 | if (PreambleConditionalStack.reachedEOFWhileSkipping()) |
691 | SkipExcludedConditionalBlock( |
692 | HashTokenLoc: PreambleConditionalStack.SkipInfo->HashTokenLoc, |
693 | IfTokenLoc: PreambleConditionalStack.SkipInfo->IfTokenLoc, |
694 | FoundNonSkipPortion: PreambleConditionalStack.SkipInfo->FoundNonSkipPortion, |
695 | FoundElse: PreambleConditionalStack.SkipInfo->FoundElse, |
696 | ElseLoc: PreambleConditionalStack.SkipInfo->ElseLoc); |
697 | } |
698 | } |
699 | |
700 | void Preprocessor::EndSourceFile() { |
701 | // Notify the client that we reached the end of the source file. |
702 | if (Callbacks) |
703 | Callbacks->EndOfMainFile(); |
704 | } |
705 | |
706 | //===----------------------------------------------------------------------===// |
707 | // Lexer Event Handling. |
708 | //===----------------------------------------------------------------------===// |
709 | |
710 | /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the |
711 | /// identifier information for the token and install it into the token, |
712 | /// updating the token kind accordingly. |
713 | IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const { |
714 | assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!" ); |
715 | |
716 | // Look up this token, see if it is a macro, or if it is a language keyword. |
717 | IdentifierInfo *II; |
718 | if (!Identifier.needsCleaning() && !Identifier.hasUCN()) { |
719 | // No cleaning needed, just use the characters from the lexed buffer. |
720 | II = getIdentifierInfo(Name: Identifier.getRawIdentifier()); |
721 | } else { |
722 | // Cleaning needed, alloca a buffer, clean into it, then use the buffer. |
723 | SmallString<64> IdentifierBuffer; |
724 | StringRef CleanedStr = getSpelling(Tok: Identifier, Buffer&: IdentifierBuffer); |
725 | |
726 | if (Identifier.hasUCN()) { |
727 | SmallString<64> UCNIdentifierBuffer; |
728 | expandUCNs(Buf&: UCNIdentifierBuffer, Input: CleanedStr); |
729 | II = getIdentifierInfo(Name: UCNIdentifierBuffer); |
730 | } else { |
731 | II = getIdentifierInfo(Name: CleanedStr); |
732 | } |
733 | } |
734 | |
735 | // Update the token info (identifier info and appropriate token kind). |
736 | // FIXME: the raw_identifier may contain leading whitespace which is removed |
737 | // from the cleaned identifier token. The SourceLocation should be updated to |
738 | // refer to the non-whitespace character. For instance, the text "\\\nB" (a |
739 | // line continuation before 'B') is parsed as a single tok::raw_identifier and |
740 | // is cleaned to tok::identifier "B". After cleaning the token's length is |
741 | // still 3 and the SourceLocation refers to the location of the backslash. |
742 | Identifier.setIdentifierInfo(II); |
743 | Identifier.setKind(II->getTokenID()); |
744 | |
745 | return II; |
746 | } |
747 | |
748 | void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) { |
749 | PoisonReasons[II] = DiagID; |
750 | } |
751 | |
752 | void Preprocessor::PoisonSEHIdentifiers(bool Poison) { |
753 | assert(Ident__exception_code && Ident__exception_info); |
754 | assert(Ident___exception_code && Ident___exception_info); |
755 | Ident__exception_code->setIsPoisoned(Poison); |
756 | Ident___exception_code->setIsPoisoned(Poison); |
757 | Ident_GetExceptionCode->setIsPoisoned(Poison); |
758 | Ident__exception_info->setIsPoisoned(Poison); |
759 | Ident___exception_info->setIsPoisoned(Poison); |
760 | Ident_GetExceptionInfo->setIsPoisoned(Poison); |
761 | Ident__abnormal_termination->setIsPoisoned(Poison); |
762 | Ident___abnormal_termination->setIsPoisoned(Poison); |
763 | Ident_AbnormalTermination->setIsPoisoned(Poison); |
764 | } |
765 | |
766 | void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) { |
767 | assert(Identifier.getIdentifierInfo() && |
768 | "Can't handle identifiers without identifier info!" ); |
769 | llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it = |
770 | PoisonReasons.find(Val: Identifier.getIdentifierInfo()); |
771 | if(it == PoisonReasons.end()) |
772 | Diag(Tok: Identifier, DiagID: diag::err_pp_used_poisoned_id); |
773 | else |
774 | Diag(Tok: Identifier,DiagID: it->second) << Identifier.getIdentifierInfo(); |
775 | } |
776 | |
777 | void Preprocessor::updateOutOfDateIdentifier(const IdentifierInfo &II) const { |
778 | assert(II.isOutOfDate() && "not out of date" ); |
779 | assert(getExternalSource() && |
780 | "getExternalSource() should not return nullptr" ); |
781 | getExternalSource()->updateOutOfDateIdentifier(II); |
782 | } |
783 | |
784 | /// HandleIdentifier - This callback is invoked when the lexer reads an |
785 | /// identifier. This callback looks up the identifier in the map and/or |
786 | /// potentially macro expands it or turns it into a named token (like 'for'). |
787 | /// |
788 | /// Note that callers of this method are guarded by checking the |
789 | /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the |
790 | /// IdentifierInfo methods that compute these properties will need to change to |
791 | /// match. |
792 | bool Preprocessor::HandleIdentifier(Token &Identifier) { |
793 | assert(Identifier.getIdentifierInfo() && |
794 | "Can't handle identifiers without identifier info!" ); |
795 | |
796 | IdentifierInfo &II = *Identifier.getIdentifierInfo(); |
797 | |
798 | // If the information about this identifier is out of date, update it from |
799 | // the external source. |
800 | // We have to treat __VA_ARGS__ in a special way, since it gets |
801 | // serialized with isPoisoned = true, but our preprocessor may have |
802 | // unpoisoned it if we're defining a C99 macro. |
803 | if (II.isOutOfDate()) { |
804 | bool CurrentIsPoisoned = false; |
805 | const bool IsSpecialVariadicMacro = |
806 | &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__; |
807 | if (IsSpecialVariadicMacro) |
808 | CurrentIsPoisoned = II.isPoisoned(); |
809 | |
810 | updateOutOfDateIdentifier(II); |
811 | Identifier.setKind(II.getTokenID()); |
812 | |
813 | if (IsSpecialVariadicMacro) |
814 | II.setIsPoisoned(CurrentIsPoisoned); |
815 | } |
816 | |
817 | // If this identifier was poisoned, and if it was not produced from a macro |
818 | // expansion, emit an error. |
819 | if (II.isPoisoned() && CurPPLexer) { |
820 | HandlePoisonedIdentifier(Identifier); |
821 | } |
822 | |
823 | // If this is a macro to be expanded, do it. |
824 | if (const MacroDefinition MD = getMacroDefinition(II: &II)) { |
825 | const auto *MI = MD.getMacroInfo(); |
826 | assert(MI && "macro definition with no macro info?" ); |
827 | if (!DisableMacroExpansion) { |
828 | if (!Identifier.isExpandDisabled() && MI->isEnabled()) { |
829 | // C99 6.10.3p10: If the preprocessing token immediately after the |
830 | // macro name isn't a '(', this macro should not be expanded. |
831 | if (!MI->isFunctionLike() || isNextPPTokenOneOf(Ks: tok::l_paren)) |
832 | return HandleMacroExpandedIdentifier(Identifier, MD); |
833 | } else { |
834 | // C99 6.10.3.4p2 says that a disabled macro may never again be |
835 | // expanded, even if it's in a context where it could be expanded in the |
836 | // future. |
837 | Identifier.setFlag(Token::DisableExpand); |
838 | if (MI->isObjectLike() || isNextPPTokenOneOf(Ks: tok::l_paren)) |
839 | Diag(Tok: Identifier, DiagID: diag::pp_disabled_macro_expansion); |
840 | } |
841 | } |
842 | } |
843 | |
844 | // If this identifier is a keyword in a newer Standard or proposed Standard, |
845 | // produce a warning. Don't warn if we're not considering macro expansion, |
846 | // since this identifier might be the name of a macro. |
847 | // FIXME: This warning is disabled in cases where it shouldn't be, like |
848 | // "#define constexpr constexpr", "int constexpr;" |
849 | if (II.isFutureCompatKeyword() && !DisableMacroExpansion) { |
850 | Diag(Tok: Identifier, DiagID: getIdentifierTable().getFutureCompatDiagKind(II, LangOpts: getLangOpts())) |
851 | << II.getName(); |
852 | // Don't diagnose this keyword again in this translation unit. |
853 | II.setIsFutureCompatKeyword(false); |
854 | } |
855 | |
856 | // If this identifier would be a keyword in C++, diagnose as a compatibility |
857 | // issue. |
858 | if (II.IsKeywordInCPlusPlus() && !DisableMacroExpansion) |
859 | Diag(Tok: Identifier, DiagID: diag::warn_pp_identifier_is_cpp_keyword) << &II; |
860 | |
861 | // If this is an extension token, diagnose its use. |
862 | // We avoid diagnosing tokens that originate from macro definitions. |
863 | // FIXME: This warning is disabled in cases where it shouldn't be, |
864 | // like "#define TY typeof", "TY(1) x". |
865 | if (II.isExtensionToken() && !DisableMacroExpansion) |
866 | Diag(Tok: Identifier, DiagID: diag::ext_token_used); |
867 | |
868 | // If this is the 'import' contextual keyword following an '@', note |
869 | // that the next token indicates a module name. |
870 | // |
871 | // Note that we do not treat 'import' as a contextual |
872 | // keyword when we're in a caching lexer, because caching lexers only get |
873 | // used in contexts where import declarations are disallowed. |
874 | // |
875 | // Likewise if this is the standard C++ import keyword. |
876 | if (((LastTokenWasAt && II.isModulesImport()) || |
877 | Identifier.is(K: tok::kw_import)) && |
878 | !InMacroArgs && !DisableMacroExpansion && |
879 | (getLangOpts().Modules || getLangOpts().DebuggerSupport) && |
880 | CurLexerCallback != CLK_CachingLexer) { |
881 | ModuleImportLoc = Identifier.getLocation(); |
882 | NamedModuleImportPath.clear(); |
883 | IsAtImport = true; |
884 | ModuleImportExpectsIdentifier = true; |
885 | CurLexerCallback = CLK_LexAfterModuleImport; |
886 | } |
887 | return true; |
888 | } |
889 | |
890 | void Preprocessor::Lex(Token &Result) { |
891 | ++LexLevel; |
892 | |
893 | // We loop here until a lex function returns a token; this avoids recursion. |
894 | while (!CurLexerCallback(*this, Result)) |
895 | ; |
896 | |
897 | if (Result.is(K: tok::unknown) && TheModuleLoader.HadFatalFailure) |
898 | return; |
899 | |
900 | if (Result.is(K: tok::code_completion) && Result.getIdentifierInfo()) { |
901 | // Remember the identifier before code completion token. |
902 | setCodeCompletionIdentifierInfo(Result.getIdentifierInfo()); |
903 | setCodeCompletionTokenRange(Start: Result.getLocation(), End: Result.getEndLoc()); |
904 | // Set IdenfitierInfo to null to avoid confusing code that handles both |
905 | // identifiers and completion tokens. |
906 | Result.setIdentifierInfo(nullptr); |
907 | } |
908 | |
909 | // Update StdCXXImportSeqState to track our position within a C++20 import-seq |
910 | // if this token is being produced as a result of phase 4 of translation. |
911 | // Update TrackGMFState to decide if we are currently in a Global Module |
912 | // Fragment. GMF state updates should precede StdCXXImportSeq ones, since GMF state |
913 | // depends on the prevailing StdCXXImportSeq state in two cases. |
914 | if (getLangOpts().CPlusPlusModules && LexLevel == 1 && |
915 | !Result.getFlag(Flag: Token::IsReinjected)) { |
916 | switch (Result.getKind()) { |
917 | case tok::l_paren: case tok::l_square: case tok::l_brace: |
918 | StdCXXImportSeqState.handleOpenBracket(); |
919 | break; |
920 | case tok::r_paren: case tok::r_square: |
921 | StdCXXImportSeqState.handleCloseBracket(); |
922 | break; |
923 | case tok::r_brace: |
924 | StdCXXImportSeqState.handleCloseBrace(); |
925 | break; |
926 | #define PRAGMA_ANNOTATION(X) case tok::annot_##X: |
927 | // For `#pragma ...` mimic ';'. |
928 | #include "clang/Basic/TokenKinds.def" |
929 | #undef PRAGMA_ANNOTATION |
930 | // This token is injected to represent the translation of '#include "a.h"' |
931 | // into "import a.h;". Mimic the notional ';'. |
932 | case tok::annot_module_include: |
933 | case tok::semi: |
934 | TrackGMFState.handleSemi(); |
935 | StdCXXImportSeqState.handleSemi(); |
936 | ModuleDeclState.handleSemi(); |
937 | break; |
938 | case tok::header_name: |
939 | case tok::annot_header_unit: |
940 | StdCXXImportSeqState.handleHeaderName(); |
941 | break; |
942 | case tok::kw_export: |
943 | TrackGMFState.handleExport(); |
944 | StdCXXImportSeqState.handleExport(); |
945 | ModuleDeclState.handleExport(); |
946 | break; |
947 | case tok::colon: |
948 | ModuleDeclState.handleColon(); |
949 | break; |
950 | case tok::period: |
951 | ModuleDeclState.handlePeriod(); |
952 | break; |
953 | case tok::identifier: |
954 | // Check "import" and "module" when there is no open bracket. The two |
955 | // identifiers are not meaningful with open brackets. |
956 | if (StdCXXImportSeqState.atTopLevel()) { |
957 | if (Result.getIdentifierInfo()->isModulesImport()) { |
958 | TrackGMFState.handleImport(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq()); |
959 | StdCXXImportSeqState.handleImport(); |
960 | if (StdCXXImportSeqState.afterImportSeq()) { |
961 | ModuleImportLoc = Result.getLocation(); |
962 | NamedModuleImportPath.clear(); |
963 | IsAtImport = false; |
964 | ModuleImportExpectsIdentifier = true; |
965 | CurLexerCallback = CLK_LexAfterModuleImport; |
966 | } |
967 | break; |
968 | } else if (Result.getIdentifierInfo() == getIdentifierInfo(Name: "module" )) { |
969 | TrackGMFState.handleModule(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq()); |
970 | ModuleDeclState.handleModule(); |
971 | break; |
972 | } |
973 | } |
974 | ModuleDeclState.handleIdentifier(Identifier: Result.getIdentifierInfo()); |
975 | if (ModuleDeclState.isModuleCandidate()) |
976 | break; |
977 | [[fallthrough]]; |
978 | default: |
979 | TrackGMFState.handleMisc(); |
980 | StdCXXImportSeqState.handleMisc(); |
981 | ModuleDeclState.handleMisc(); |
982 | break; |
983 | } |
984 | } |
985 | |
986 | if (CurLexer && ++CheckPointCounter == CheckPointStepSize) { |
987 | CheckPoints[CurLexer->getFileID()].push_back(Elt: CurLexer->BufferPtr); |
988 | CheckPointCounter = 0; |
989 | } |
990 | |
991 | LastTokenWasAt = Result.is(K: tok::at); |
992 | --LexLevel; |
993 | |
994 | if ((LexLevel == 0 || PreprocessToken) && |
995 | !Result.getFlag(Flag: Token::IsReinjected)) { |
996 | if (LexLevel == 0) |
997 | ++TokenCount; |
998 | if (OnToken) |
999 | OnToken(Result); |
1000 | } |
1001 | } |
1002 | |
1003 | void Preprocessor::LexTokensUntilEOF(std::vector<Token> *Tokens) { |
1004 | while (1) { |
1005 | Token Tok; |
1006 | Lex(Result&: Tok); |
1007 | if (Tok.isOneOf(Ks: tok::unknown, Ks: tok::eof, Ks: tok::eod, |
1008 | Ks: tok::annot_repl_input_end)) |
1009 | break; |
1010 | if (Tokens != nullptr) |
1011 | Tokens->push_back(x: Tok); |
1012 | } |
1013 | } |
1014 | |
1015 | /// Lex a header-name token (including one formed from header-name-tokens if |
1016 | /// \p AllowMacroExpansion is \c true). |
1017 | /// |
1018 | /// \param FilenameTok Filled in with the next token. On success, this will |
1019 | /// be either a header_name token. On failure, it will be whatever other |
1020 | /// token was found instead. |
1021 | /// \param AllowMacroExpansion If \c true, allow the header name to be formed |
1022 | /// by macro expansion (concatenating tokens as necessary if the first |
1023 | /// token is a '<'). |
1024 | /// \return \c true if we reached EOD or EOF while looking for a > token in |
1025 | /// a concatenated header name and diagnosed it. \c false otherwise. |
1026 | bool Preprocessor::(Token &FilenameTok, bool AllowMacroExpansion) { |
1027 | // Lex using header-name tokenization rules if tokens are being lexed from |
1028 | // a file. Just grab a token normally if we're in a macro expansion. |
1029 | if (CurPPLexer) |
1030 | CurPPLexer->LexIncludeFilename(FilenameTok); |
1031 | else |
1032 | Lex(Result&: FilenameTok); |
1033 | |
1034 | // This could be a <foo/bar.h> file coming from a macro expansion. In this |
1035 | // case, glue the tokens together into an angle_string_literal token. |
1036 | SmallString<128> FilenameBuffer; |
1037 | if (FilenameTok.is(K: tok::less) && AllowMacroExpansion) { |
1038 | bool StartOfLine = FilenameTok.isAtStartOfLine(); |
1039 | bool LeadingSpace = FilenameTok.hasLeadingSpace(); |
1040 | bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro(); |
1041 | |
1042 | SourceLocation Start = FilenameTok.getLocation(); |
1043 | SourceLocation End; |
1044 | FilenameBuffer.push_back(Elt: '<'); |
1045 | |
1046 | // Consume tokens until we find a '>'. |
1047 | // FIXME: A header-name could be formed starting or ending with an |
1048 | // alternative token. It's not clear whether that's ill-formed in all |
1049 | // cases. |
1050 | while (FilenameTok.isNot(K: tok::greater)) { |
1051 | Lex(Result&: FilenameTok); |
1052 | if (FilenameTok.isOneOf(Ks: tok::eod, Ks: tok::eof)) { |
1053 | Diag(Loc: FilenameTok.getLocation(), DiagID: diag::err_expected) << tok::greater; |
1054 | Diag(Loc: Start, DiagID: diag::note_matching) << tok::less; |
1055 | return true; |
1056 | } |
1057 | |
1058 | End = FilenameTok.getLocation(); |
1059 | |
1060 | // FIXME: Provide code completion for #includes. |
1061 | if (FilenameTok.is(K: tok::code_completion)) { |
1062 | setCodeCompletionReached(); |
1063 | Lex(Result&: FilenameTok); |
1064 | continue; |
1065 | } |
1066 | |
1067 | // Append the spelling of this token to the buffer. If there was a space |
1068 | // before it, add it now. |
1069 | if (FilenameTok.hasLeadingSpace()) |
1070 | FilenameBuffer.push_back(Elt: ' '); |
1071 | |
1072 | // Get the spelling of the token, directly into FilenameBuffer if |
1073 | // possible. |
1074 | size_t PreAppendSize = FilenameBuffer.size(); |
1075 | FilenameBuffer.resize(N: PreAppendSize + FilenameTok.getLength()); |
1076 | |
1077 | const char *BufPtr = &FilenameBuffer[PreAppendSize]; |
1078 | unsigned ActualLen = getSpelling(Tok: FilenameTok, Buffer&: BufPtr); |
1079 | |
1080 | // If the token was spelled somewhere else, copy it into FilenameBuffer. |
1081 | if (BufPtr != &FilenameBuffer[PreAppendSize]) |
1082 | memcpy(dest: &FilenameBuffer[PreAppendSize], src: BufPtr, n: ActualLen); |
1083 | |
1084 | // Resize FilenameBuffer to the correct size. |
1085 | if (FilenameTok.getLength() != ActualLen) |
1086 | FilenameBuffer.resize(N: PreAppendSize + ActualLen); |
1087 | } |
1088 | |
1089 | FilenameTok.startToken(); |
1090 | FilenameTok.setKind(tok::header_name); |
1091 | FilenameTok.setFlagValue(Flag: Token::StartOfLine, Val: StartOfLine); |
1092 | FilenameTok.setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace); |
1093 | FilenameTok.setFlagValue(Flag: Token::LeadingEmptyMacro, Val: LeadingEmptyMacro); |
1094 | CreateString(Str: FilenameBuffer, Tok&: FilenameTok, ExpansionLocStart: Start, ExpansionLocEnd: End); |
1095 | } else if (FilenameTok.is(K: tok::string_literal) && AllowMacroExpansion) { |
1096 | // Convert a string-literal token of the form " h-char-sequence " |
1097 | // (produced by macro expansion) into a header-name token. |
1098 | // |
1099 | // The rules for header-names don't quite match the rules for |
1100 | // string-literals, but all the places where they differ result in |
1101 | // undefined behavior, so we can and do treat them the same. |
1102 | // |
1103 | // A string-literal with a prefix or suffix is not translated into a |
1104 | // header-name. This could theoretically be observable via the C++20 |
1105 | // context-sensitive header-name formation rules. |
1106 | StringRef Str = getSpelling(Tok: FilenameTok, Buffer&: FilenameBuffer); |
1107 | if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"') |
1108 | FilenameTok.setKind(tok::header_name); |
1109 | } |
1110 | |
1111 | return false; |
1112 | } |
1113 | |
1114 | /// Collect the tokens of a C++20 pp-import-suffix. |
1115 | void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) { |
1116 | // FIXME: For error recovery, consider recognizing attribute syntax here |
1117 | // and terminating / diagnosing a missing semicolon if we find anything |
1118 | // else? (Can we leave that to the parser?) |
1119 | unsigned BracketDepth = 0; |
1120 | while (true) { |
1121 | Toks.emplace_back(); |
1122 | Lex(Result&: Toks.back()); |
1123 | |
1124 | switch (Toks.back().getKind()) { |
1125 | case tok::l_paren: case tok::l_square: case tok::l_brace: |
1126 | ++BracketDepth; |
1127 | break; |
1128 | |
1129 | case tok::r_paren: case tok::r_square: case tok::r_brace: |
1130 | if (BracketDepth == 0) |
1131 | return; |
1132 | --BracketDepth; |
1133 | break; |
1134 | |
1135 | case tok::semi: |
1136 | if (BracketDepth == 0) |
1137 | return; |
1138 | break; |
1139 | |
1140 | case tok::eof: |
1141 | return; |
1142 | |
1143 | default: |
1144 | break; |
1145 | } |
1146 | } |
1147 | } |
1148 | |
1149 | |
1150 | /// Lex a token following the 'import' contextual keyword. |
1151 | /// |
1152 | /// pp-import: [C++20] |
1153 | /// import header-name pp-import-suffix[opt] ; |
1154 | /// import header-name-tokens pp-import-suffix[opt] ; |
1155 | /// [ObjC] @ import module-name ; |
1156 | /// [Clang] import module-name ; |
1157 | /// |
1158 | /// header-name-tokens: |
1159 | /// string-literal |
1160 | /// < [any sequence of preprocessing-tokens other than >] > |
1161 | /// |
1162 | /// module-name: |
1163 | /// module-name-qualifier[opt] identifier |
1164 | /// |
1165 | /// module-name-qualifier |
1166 | /// module-name-qualifier[opt] identifier . |
1167 | /// |
1168 | /// We respond to a pp-import by importing macros from the named module. |
1169 | bool Preprocessor::LexAfterModuleImport(Token &Result) { |
1170 | // Figure out what kind of lexer we actually have. |
1171 | recomputeCurLexerKind(); |
1172 | |
1173 | // Lex the next token. The header-name lexing rules are used at the start of |
1174 | // a pp-import. |
1175 | // |
1176 | // For now, we only support header-name imports in C++20 mode. |
1177 | // FIXME: Should we allow this in all language modes that support an import |
1178 | // declaration as an extension? |
1179 | if (NamedModuleImportPath.empty() && getLangOpts().CPlusPlusModules) { |
1180 | if (LexHeaderName(FilenameTok&: Result)) |
1181 | return true; |
1182 | |
1183 | if (Result.is(K: tok::colon) && ModuleDeclState.isNamedModule()) { |
1184 | std::string Name = ModuleDeclState.getPrimaryName().str(); |
1185 | Name += ":" ; |
1186 | NamedModuleImportPath.emplace_back(Args: Result.getLocation(), |
1187 | Args: getIdentifierInfo(Name)); |
1188 | CurLexerCallback = CLK_LexAfterModuleImport; |
1189 | return true; |
1190 | } |
1191 | } else { |
1192 | Lex(Result); |
1193 | } |
1194 | |
1195 | // Allocate a holding buffer for a sequence of tokens and introduce it into |
1196 | // the token stream. |
1197 | auto EnterTokens = [this](ArrayRef<Token> Toks) { |
1198 | auto ToksCopy = std::make_unique<Token[]>(num: Toks.size()); |
1199 | std::copy(first: Toks.begin(), last: Toks.end(), result: ToksCopy.get()); |
1200 | EnterTokenStream(Toks: std::move(ToksCopy), NumToks: Toks.size(), |
1201 | /*DisableMacroExpansion*/ true, /*IsReinject*/ false); |
1202 | }; |
1203 | |
1204 | bool = Result.is(K: tok::header_name); |
1205 | // Check for a header-name. |
1206 | SmallVector<Token, 32> Suffix; |
1207 | if (ImportingHeader) { |
1208 | // Enter the header-name token into the token stream; a Lex action cannot |
1209 | // both return a token and cache tokens (doing so would corrupt the token |
1210 | // cache if the call to Lex comes from CachingLex / PeekAhead). |
1211 | Suffix.push_back(Elt: Result); |
1212 | |
1213 | // Consume the pp-import-suffix and expand any macros in it now. We'll add |
1214 | // it back into the token stream later. |
1215 | CollectPpImportSuffix(Toks&: Suffix); |
1216 | if (Suffix.back().isNot(K: tok::semi)) { |
1217 | // This is not a pp-import after all. |
1218 | EnterTokens(Suffix); |
1219 | return false; |
1220 | } |
1221 | |
1222 | // C++2a [cpp.module]p1: |
1223 | // The ';' preprocessing-token terminating a pp-import shall not have |
1224 | // been produced by macro replacement. |
1225 | SourceLocation SemiLoc = Suffix.back().getLocation(); |
1226 | if (SemiLoc.isMacroID()) |
1227 | Diag(Loc: SemiLoc, DiagID: diag::err_header_import_semi_in_macro); |
1228 | |
1229 | // Reconstitute the import token. |
1230 | Token ImportTok; |
1231 | ImportTok.startToken(); |
1232 | ImportTok.setKind(tok::kw_import); |
1233 | ImportTok.setLocation(ModuleImportLoc); |
1234 | ImportTok.setIdentifierInfo(getIdentifierInfo(Name: "import" )); |
1235 | ImportTok.setLength(6); |
1236 | |
1237 | auto Action = HandleHeaderIncludeOrImport( |
1238 | /*HashLoc*/ SourceLocation(), IncludeTok&: ImportTok, FilenameTok&: Suffix.front(), EndLoc: SemiLoc); |
1239 | switch (Action.Kind) { |
1240 | case ImportAction::None: |
1241 | break; |
1242 | |
1243 | case ImportAction::ModuleBegin: |
1244 | // Let the parser know we're textually entering the module. |
1245 | Suffix.emplace_back(); |
1246 | Suffix.back().startToken(); |
1247 | Suffix.back().setKind(tok::annot_module_begin); |
1248 | Suffix.back().setLocation(SemiLoc); |
1249 | Suffix.back().setAnnotationEndLoc(SemiLoc); |
1250 | Suffix.back().setAnnotationValue(Action.ModuleForHeader); |
1251 | [[fallthrough]]; |
1252 | |
1253 | case ImportAction::ModuleImport: |
1254 | case ImportAction::HeaderUnitImport: |
1255 | case ImportAction::SkippedModuleImport: |
1256 | // We chose to import (or textually enter) the file. Convert the |
1257 | // header-name token into a header unit annotation token. |
1258 | Suffix[0].setKind(tok::annot_header_unit); |
1259 | Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation()); |
1260 | Suffix[0].setAnnotationValue(Action.ModuleForHeader); |
1261 | // FIXME: Call the moduleImport callback? |
1262 | break; |
1263 | case ImportAction::Failure: |
1264 | assert(TheModuleLoader.HadFatalFailure && |
1265 | "This should be an early exit only to a fatal error" ); |
1266 | Result.setKind(tok::eof); |
1267 | CurLexer->cutOffLexing(); |
1268 | EnterTokens(Suffix); |
1269 | return true; |
1270 | } |
1271 | |
1272 | EnterTokens(Suffix); |
1273 | return false; |
1274 | } |
1275 | |
1276 | // The token sequence |
1277 | // |
1278 | // import identifier (. identifier)* |
1279 | // |
1280 | // indicates a module import directive. We already saw the 'import' |
1281 | // contextual keyword, so now we're looking for the identifiers. |
1282 | if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) { |
1283 | // We expected to see an identifier here, and we did; continue handling |
1284 | // identifiers. |
1285 | NamedModuleImportPath.emplace_back(Args: Result.getLocation(), |
1286 | Args: Result.getIdentifierInfo()); |
1287 | ModuleImportExpectsIdentifier = false; |
1288 | CurLexerCallback = CLK_LexAfterModuleImport; |
1289 | return true; |
1290 | } |
1291 | |
1292 | // If we're expecting a '.' or a ';', and we got a '.', then wait until we |
1293 | // see the next identifier. (We can also see a '[[' that begins an |
1294 | // attribute-specifier-seq here under the Standard C++ Modules.) |
1295 | if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) { |
1296 | ModuleImportExpectsIdentifier = true; |
1297 | CurLexerCallback = CLK_LexAfterModuleImport; |
1298 | return true; |
1299 | } |
1300 | |
1301 | // If we didn't recognize a module name at all, this is not a (valid) import. |
1302 | if (NamedModuleImportPath.empty() || Result.is(K: tok::eof)) |
1303 | return true; |
1304 | |
1305 | // Consume the pp-import-suffix and expand any macros in it now, if we're not |
1306 | // at the semicolon already. |
1307 | SourceLocation SemiLoc = Result.getLocation(); |
1308 | if (Result.isNot(K: tok::semi)) { |
1309 | Suffix.push_back(Elt: Result); |
1310 | CollectPpImportSuffix(Toks&: Suffix); |
1311 | if (Suffix.back().isNot(K: tok::semi)) { |
1312 | // This is not an import after all. |
1313 | EnterTokens(Suffix); |
1314 | return false; |
1315 | } |
1316 | SemiLoc = Suffix.back().getLocation(); |
1317 | } |
1318 | |
1319 | // Under the standard C++ Modules, the dot is just part of the module name, |
1320 | // and not a real hierarchy separator. Flatten such module names now. |
1321 | // |
1322 | // FIXME: Is this the right level to be performing this transformation? |
1323 | std::string FlatModuleName; |
1324 | if (getLangOpts().CPlusPlusModules) { |
1325 | for (auto &Piece : NamedModuleImportPath) { |
1326 | // If the FlatModuleName ends with colon, it implies it is a partition. |
1327 | if (!FlatModuleName.empty() && FlatModuleName.back() != ':') |
1328 | FlatModuleName += "." ; |
1329 | FlatModuleName += Piece.getIdentifierInfo()->getName(); |
1330 | } |
1331 | SourceLocation FirstPathLoc = NamedModuleImportPath[0].getLoc(); |
1332 | NamedModuleImportPath.clear(); |
1333 | NamedModuleImportPath.emplace_back(Args&: FirstPathLoc, |
1334 | Args: getIdentifierInfo(Name: FlatModuleName)); |
1335 | } |
1336 | |
1337 | Module *Imported = nullptr; |
1338 | // We don't/shouldn't load the standard c++20 modules when preprocessing. |
1339 | if (getLangOpts().Modules && !isInImportingCXXNamedModules()) { |
1340 | Imported = TheModuleLoader.loadModule(ImportLoc: ModuleImportLoc, |
1341 | Path: NamedModuleImportPath, |
1342 | Visibility: Module::Hidden, |
1343 | /*IsInclusionDirective=*/false); |
1344 | if (Imported) |
1345 | makeModuleVisible(M: Imported, Loc: SemiLoc); |
1346 | } |
1347 | |
1348 | if (Callbacks) |
1349 | Callbacks->moduleImport(ImportLoc: ModuleImportLoc, Path: NamedModuleImportPath, Imported); |
1350 | |
1351 | if (!Suffix.empty()) { |
1352 | EnterTokens(Suffix); |
1353 | return false; |
1354 | } |
1355 | return true; |
1356 | } |
1357 | |
1358 | void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc, |
1359 | bool IncludeExports) { |
1360 | CurSubmoduleState->VisibleModules.setVisible( |
1361 | M, Loc, IncludeExports, Vis: [](Module *) {}, |
1362 | Cb: [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) { |
1363 | // FIXME: Include the path in the diagnostic. |
1364 | // FIXME: Include the import location for the conflicting module. |
1365 | Diag(Loc: ModuleImportLoc, DiagID: diag::warn_module_conflict) |
1366 | << Path[0]->getFullModuleName() |
1367 | << Conflict->getFullModuleName() |
1368 | << Message; |
1369 | }); |
1370 | |
1371 | // Add this module to the imports list of the currently-built submodule. |
1372 | if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M) |
1373 | BuildingSubmoduleStack.back().M->Imports.insert(X: M); |
1374 | } |
1375 | |
1376 | bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String, |
1377 | const char *DiagnosticTag, |
1378 | bool AllowMacroExpansion) { |
1379 | // We need at least one string literal. |
1380 | if (Result.isNot(K: tok::string_literal)) { |
1381 | Diag(Tok: Result, DiagID: diag::err_expected_string_literal) |
1382 | << /*Source='in...'*/0 << DiagnosticTag; |
1383 | return false; |
1384 | } |
1385 | |
1386 | // Lex string literal tokens, optionally with macro expansion. |
1387 | SmallVector<Token, 4> StrToks; |
1388 | do { |
1389 | StrToks.push_back(Elt: Result); |
1390 | |
1391 | if (Result.hasUDSuffix()) |
1392 | Diag(Tok: Result, DiagID: diag::err_invalid_string_udl); |
1393 | |
1394 | if (AllowMacroExpansion) |
1395 | Lex(Result); |
1396 | else |
1397 | LexUnexpandedToken(Result); |
1398 | } while (Result.is(K: tok::string_literal)); |
1399 | |
1400 | // Concatenate and parse the strings. |
1401 | StringLiteralParser Literal(StrToks, *this); |
1402 | assert(Literal.isOrdinary() && "Didn't allow wide strings in" ); |
1403 | |
1404 | if (Literal.hadError) |
1405 | return false; |
1406 | |
1407 | if (Literal.Pascal) { |
1408 | Diag(Loc: StrToks[0].getLocation(), DiagID: diag::err_expected_string_literal) |
1409 | << /*Source='in...'*/0 << DiagnosticTag; |
1410 | return false; |
1411 | } |
1412 | |
1413 | String = std::string(Literal.GetString()); |
1414 | return true; |
1415 | } |
1416 | |
1417 | bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) { |
1418 | assert(Tok.is(tok::numeric_constant)); |
1419 | SmallString<8> IntegerBuffer; |
1420 | bool NumberInvalid = false; |
1421 | StringRef Spelling = getSpelling(Tok, Buffer&: IntegerBuffer, Invalid: &NumberInvalid); |
1422 | if (NumberInvalid) |
1423 | return false; |
1424 | NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(), |
1425 | getLangOpts(), getTargetInfo(), |
1426 | getDiagnostics()); |
1427 | if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix()) |
1428 | return false; |
1429 | llvm::APInt APVal(64, 0); |
1430 | if (Literal.GetIntegerValue(Val&: APVal)) |
1431 | return false; |
1432 | Lex(Result&: Tok); |
1433 | Value = APVal.getLimitedValue(); |
1434 | return true; |
1435 | } |
1436 | |
1437 | void Preprocessor::addCommentHandler(CommentHandler *Handler) { |
1438 | assert(Handler && "NULL comment handler" ); |
1439 | assert(!llvm::is_contained(CommentHandlers, Handler) && |
1440 | "Comment handler already registered" ); |
1441 | CommentHandlers.push_back(x: Handler); |
1442 | } |
1443 | |
1444 | void Preprocessor::removeCommentHandler(CommentHandler *Handler) { |
1445 | std::vector<CommentHandler *>::iterator Pos = |
1446 | llvm::find(Range&: CommentHandlers, Val: Handler); |
1447 | assert(Pos != CommentHandlers.end() && "Comment handler not registered" ); |
1448 | CommentHandlers.erase(position: Pos); |
1449 | } |
1450 | |
1451 | bool Preprocessor::HandleComment(Token &result, SourceRange ) { |
1452 | bool AnyPendingTokens = false; |
1453 | for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(), |
1454 | HEnd = CommentHandlers.end(); |
1455 | H != HEnd; ++H) { |
1456 | if ((*H)->HandleComment(PP&: *this, Comment)) |
1457 | AnyPendingTokens = true; |
1458 | } |
1459 | if (!AnyPendingTokens || getCommentRetentionState()) |
1460 | return false; |
1461 | Lex(Result&: result); |
1462 | return true; |
1463 | } |
1464 | |
1465 | void Preprocessor::emitMacroDeprecationWarning(const Token &Identifier) const { |
1466 | const MacroAnnotations &A = |
1467 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1468 | assert(A.DeprecationInfo && |
1469 | "Macro deprecation warning without recorded annotation!" ); |
1470 | const MacroAnnotationInfo &Info = *A.DeprecationInfo; |
1471 | if (Info.Message.empty()) |
1472 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use) |
1473 | << Identifier.getIdentifierInfo() << 0; |
1474 | else |
1475 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use) |
1476 | << Identifier.getIdentifierInfo() << 1 << Info.Message; |
1477 | Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 0; |
1478 | } |
1479 | |
1480 | void Preprocessor::emitRestrictExpansionWarning(const Token &Identifier) const { |
1481 | const MacroAnnotations &A = |
1482 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1483 | assert(A.RestrictExpansionInfo && |
1484 | "Macro restricted expansion warning without recorded annotation!" ); |
1485 | const MacroAnnotationInfo &Info = *A.RestrictExpansionInfo; |
1486 | if (Info.Message.empty()) |
1487 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use) |
1488 | << Identifier.getIdentifierInfo() << 0; |
1489 | else |
1490 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use) |
1491 | << Identifier.getIdentifierInfo() << 1 << Info.Message; |
1492 | Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 1; |
1493 | } |
1494 | |
1495 | void Preprocessor::emitRestrictInfNaNWarning(const Token &Identifier, |
1496 | unsigned DiagSelection) const { |
1497 | Diag(Tok: Identifier, DiagID: diag::warn_fp_nan_inf_when_disabled) << DiagSelection << 1; |
1498 | } |
1499 | |
1500 | void Preprocessor::emitFinalMacroWarning(const Token &Identifier, |
1501 | bool IsUndef) const { |
1502 | const MacroAnnotations &A = |
1503 | getMacroAnnotations(II: Identifier.getIdentifierInfo()); |
1504 | assert(A.FinalAnnotationLoc && |
1505 | "Final macro warning without recorded annotation!" ); |
1506 | |
1507 | Diag(Tok: Identifier, DiagID: diag::warn_pragma_final_macro) |
1508 | << Identifier.getIdentifierInfo() << (IsUndef ? 0 : 1); |
1509 | Diag(Loc: *A.FinalAnnotationLoc, DiagID: diag::note_pp_macro_annotation) << 2; |
1510 | } |
1511 | |
1512 | bool Preprocessor::isSafeBufferOptOut(const SourceManager &SourceMgr, |
1513 | const SourceLocation &Loc) const { |
1514 | // The lambda that tests if a `Loc` is in an opt-out region given one opt-out |
1515 | // region map: |
1516 | auto TestInMap = [&SourceMgr](const SafeBufferOptOutRegionsTy &Map, |
1517 | const SourceLocation &Loc) -> bool { |
1518 | // Try to find a region in `SafeBufferOptOutMap` where `Loc` is in: |
1519 | auto FirstRegionEndingAfterLoc = llvm::partition_point( |
1520 | Range: Map, P: [&SourceMgr, |
1521 | &Loc](const std::pair<SourceLocation, SourceLocation> &Region) { |
1522 | return SourceMgr.isBeforeInTranslationUnit(LHS: Region.second, RHS: Loc); |
1523 | }); |
1524 | |
1525 | if (FirstRegionEndingAfterLoc != Map.end()) { |
1526 | // To test if the start location of the found region precedes `Loc`: |
1527 | return SourceMgr.isBeforeInTranslationUnit( |
1528 | LHS: FirstRegionEndingAfterLoc->first, RHS: Loc); |
1529 | } |
1530 | // If we do not find a region whose end location passes `Loc`, we want to |
1531 | // check if the current region is still open: |
1532 | if (!Map.empty() && Map.back().first == Map.back().second) |
1533 | return SourceMgr.isBeforeInTranslationUnit(LHS: Map.back().first, RHS: Loc); |
1534 | return false; |
1535 | }; |
1536 | |
1537 | // What the following does: |
1538 | // |
1539 | // If `Loc` belongs to the local TU, we just look up `SafeBufferOptOutMap`. |
1540 | // Otherwise, `Loc` is from a loaded AST. We look up the |
1541 | // `LoadedSafeBufferOptOutMap` first to get the opt-out region map of the |
1542 | // loaded AST where `Loc` is at. Then we find if `Loc` is in an opt-out |
1543 | // region w.r.t. the region map. If the region map is absent, it means there |
1544 | // is no opt-out pragma in that loaded AST. |
1545 | // |
1546 | // Opt-out pragmas in the local TU or a loaded AST is not visible to another |
1547 | // one of them. That means if you put the pragmas around a `#include |
1548 | // "module.h"`, where module.h is a module, it is not actually suppressing |
1549 | // warnings in module.h. This is fine because warnings in module.h will be |
1550 | // reported when module.h is compiled in isolation and nothing in module.h |
1551 | // will be analyzed ever again. So you will not see warnings from the file |
1552 | // that imports module.h anyway. And you can't even do the same thing for PCHs |
1553 | // because they can only be included from the command line. |
1554 | |
1555 | if (SourceMgr.isLocalSourceLocation(Loc)) |
1556 | return TestInMap(SafeBufferOptOutMap, Loc); |
1557 | |
1558 | const SafeBufferOptOutRegionsTy *LoadedRegions = |
1559 | LoadedSafeBufferOptOutMap.lookupLoadedOptOutMap(Loc, SrcMgr: SourceMgr); |
1560 | |
1561 | if (LoadedRegions) |
1562 | return TestInMap(*LoadedRegions, Loc); |
1563 | return false; |
1564 | } |
1565 | |
1566 | bool Preprocessor::enterOrExitSafeBufferOptOutRegion( |
1567 | bool isEnter, const SourceLocation &Loc) { |
1568 | if (isEnter) { |
1569 | if (isPPInSafeBufferOptOutRegion()) |
1570 | return true; // invalid enter action |
1571 | InSafeBufferOptOutRegion = true; |
1572 | CurrentSafeBufferOptOutStart = Loc; |
1573 | |
1574 | // To set the start location of a new region: |
1575 | |
1576 | if (!SafeBufferOptOutMap.empty()) { |
1577 | [[maybe_unused]] auto *PrevRegion = &SafeBufferOptOutMap.back(); |
1578 | assert(PrevRegion->first != PrevRegion->second && |
1579 | "Shall not begin a safe buffer opt-out region before closing the " |
1580 | "previous one." ); |
1581 | } |
1582 | // If the start location equals to the end location, we call the region a |
1583 | // open region or a unclosed region (i.e., end location has not been set |
1584 | // yet). |
1585 | SafeBufferOptOutMap.emplace_back(Args: Loc, Args: Loc); |
1586 | } else { |
1587 | if (!isPPInSafeBufferOptOutRegion()) |
1588 | return true; // invalid enter action |
1589 | InSafeBufferOptOutRegion = false; |
1590 | |
1591 | // To set the end location of the current open region: |
1592 | |
1593 | assert(!SafeBufferOptOutMap.empty() && |
1594 | "Misordered safe buffer opt-out regions" ); |
1595 | auto *CurrRegion = &SafeBufferOptOutMap.back(); |
1596 | assert(CurrRegion->first == CurrRegion->second && |
1597 | "Set end location to a closed safe buffer opt-out region" ); |
1598 | CurrRegion->second = Loc; |
1599 | } |
1600 | return false; |
1601 | } |
1602 | |
1603 | bool Preprocessor::isPPInSafeBufferOptOutRegion() { |
1604 | return InSafeBufferOptOutRegion; |
1605 | } |
1606 | bool Preprocessor::isPPInSafeBufferOptOutRegion(SourceLocation &StartLoc) { |
1607 | StartLoc = CurrentSafeBufferOptOutStart; |
1608 | return InSafeBufferOptOutRegion; |
1609 | } |
1610 | |
1611 | SmallVector<SourceLocation, 64> |
1612 | Preprocessor::serializeSafeBufferOptOutMap() const { |
1613 | assert(!InSafeBufferOptOutRegion && |
1614 | "Attempt to serialize safe buffer opt-out regions before file being " |
1615 | "completely preprocessed" ); |
1616 | |
1617 | SmallVector<SourceLocation, 64> SrcSeq; |
1618 | |
1619 | for (const auto &[begin, end] : SafeBufferOptOutMap) { |
1620 | SrcSeq.push_back(Elt: begin); |
1621 | SrcSeq.push_back(Elt: end); |
1622 | } |
1623 | // Only `SafeBufferOptOutMap` gets serialized. No need to serialize |
1624 | // `LoadedSafeBufferOptOutMap` because if this TU loads a pch/module, every |
1625 | // pch/module in the pch-chain/module-DAG will be loaded one by one in order. |
1626 | // It means that for each loading pch/module m, it just needs to load m's own |
1627 | // `SafeBufferOptOutMap`. |
1628 | return SrcSeq; |
1629 | } |
1630 | |
1631 | bool Preprocessor::setDeserializedSafeBufferOptOutMap( |
1632 | const SmallVectorImpl<SourceLocation> &SourceLocations) { |
1633 | if (SourceLocations.size() == 0) |
1634 | return false; |
1635 | |
1636 | assert(SourceLocations.size() % 2 == 0 && |
1637 | "ill-formed SourceLocation sequence" ); |
1638 | |
1639 | auto It = SourceLocations.begin(); |
1640 | SafeBufferOptOutRegionsTy &Regions = |
1641 | LoadedSafeBufferOptOutMap.findAndConsLoadedOptOutMap(Loc: *It, SrcMgr&: SourceMgr); |
1642 | |
1643 | do { |
1644 | SourceLocation Begin = *It++; |
1645 | SourceLocation End = *It++; |
1646 | |
1647 | Regions.emplace_back(Args&: Begin, Args&: End); |
1648 | } while (It != SourceLocations.end()); |
1649 | return true; |
1650 | } |
1651 | |
1652 | ModuleLoader::~ModuleLoader() = default; |
1653 | |
1654 | CommentHandler::~CommentHandler() = default; |
1655 | |
1656 | EmptylineHandler::~EmptylineHandler() = default; |
1657 | |
1658 | CodeCompletionHandler::~CodeCompletionHandler() = default; |
1659 | |
1660 | void Preprocessor::createPreprocessingRecord() { |
1661 | if (Record) |
1662 | return; |
1663 | |
1664 | Record = new PreprocessingRecord(getSourceManager()); |
1665 | addPPCallbacks(C: std::unique_ptr<PPCallbacks>(Record)); |
1666 | } |
1667 | |
1668 | const char *Preprocessor::getCheckPoint(FileID FID, const char *Start) const { |
1669 | if (auto It = CheckPoints.find(Val: FID); It != CheckPoints.end()) { |
1670 | const SmallVector<const char *> &FileCheckPoints = It->second; |
1671 | const char *Last = nullptr; |
1672 | // FIXME: Do better than a linear search. |
1673 | for (const char *P : FileCheckPoints) { |
1674 | if (P > Start) |
1675 | break; |
1676 | Last = P; |
1677 | } |
1678 | return Last; |
1679 | } |
1680 | |
1681 | return nullptr; |
1682 | } |
1683 | |