1 //===- Preprocess.cpp - C Language Family Preprocessor Implementation -----===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the Preprocessor interface.
12 //===----------------------------------------------------------------------===//
14 // Options to support:
15 // -H - Print the name of each header file used.
16 // -d[DNI] - Dump various things.
17 // -fworking-directory - #line's with preprocessor's working dir.
19 // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
24 // "Multiple include guards may be useful for:\n"
26 //===----------------------------------------------------------------------===//
28 #include "clang/Lex/Preprocessor.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/PTHLexer.h"
48 #include "clang/Lex/PTHManager.h"
49 #include "clang/Lex/Pragma.h"
50 #include "clang/Lex/PreprocessingRecord.h"
51 #include "clang/Lex/PreprocessorLexer.h"
52 #include "clang/Lex/PreprocessorOptions.h"
53 #include "clang/Lex/ScratchBuffer.h"
54 #include "clang/Lex/Token.h"
55 #include "clang/Lex/TokenLexer.h"
56 #include "llvm/ADT/APInt.h"
57 #include "llvm/ADT/ArrayRef.h"
58 #include "llvm/ADT/DenseMap.h"
59 #include "llvm/ADT/SmallString.h"
60 #include "llvm/ADT/SmallVector.h"
61 #include "llvm/ADT/STLExtras.h"
62 #include "llvm/ADT/StringRef.h"
63 #include "llvm/ADT/StringSwitch.h"
64 #include "llvm/Support/Capacity.h"
65 #include "llvm/Support/ErrorHandling.h"
66 #include "llvm/Support/MemoryBuffer.h"
67 #include "llvm/Support/raw_ostream.h"
75 using namespace clang;
77 LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
79 ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
81 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
82 DiagnosticsEngine &diags, LangOptions &opts,
83 SourceManager &SM, MemoryBufferCache &PCMCache,
84 HeaderSearch &Headers, ModuleLoader &TheModuleLoader,
85 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
86 TranslationUnitKind TUKind)
87 : PPOpts(std::move(PPOpts)), Diags(&diags), LangOpts(opts),
88 FileMgr(Headers.getFileMgr()), SourceMgr(SM), PCMCache(PCMCache),
89 ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers),
90 TheModuleLoader(TheModuleLoader), ExternalSource(nullptr),
91 // As the language options may have not been loaded yet (when
92 // deserializing an ASTUnit), adding keywords to the identifier table is
93 // deferred to Preprocessor::Initialize().
94 Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())),
95 TUKind(TUKind), SkipMainFilePreamble(0, true),
96 CurSubmoduleState(&NullSubmoduleState) {
97 OwnsHeaderSearch = OwnsHeaders;
99 // Default to discarding comments.
100 KeepComments = false;
101 KeepMacroComments = false;
102 SuppressIncludeNotFoundError = false;
104 // Macro expansion is enabled.
105 DisableMacroExpansion = false;
106 MacroExpansionInDirectivesOverride = false;
108 InMacroArgPreExpansion = false;
109 NumCachedTokenLexers = 0;
110 PragmasEnabled = true;
111 ParsingIfOrElifDirective = false;
112 PreprocessedOutput = false;
114 // We haven't read anything from the external source.
115 ReadMacrosFromExternalSource = false;
117 // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of
118 // a macro. They get unpoisoned where it is allowed.
119 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
120 SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
121 if (getLangOpts().CPlusPlus2a) {
122 (Ident__VA_OPT__ = getIdentifierInfo("__VA_OPT__"))->setIsPoisoned();
123 SetPoisonReason(Ident__VA_OPT__,diag::ext_pp_bad_vaopt_use);
125 Ident__VA_OPT__ = nullptr;
128 // Initialize the pragma handlers.
129 RegisterBuiltinPragmas();
131 // Initialize builtin macros like __LINE__ and friends.
132 RegisterBuiltinMacros();
134 if(LangOpts.Borland) {
135 Ident__exception_info = getIdentifierInfo("_exception_info");
136 Ident___exception_info = getIdentifierInfo("__exception_info");
137 Ident_GetExceptionInfo = getIdentifierInfo("GetExceptionInformation");
138 Ident__exception_code = getIdentifierInfo("_exception_code");
139 Ident___exception_code = getIdentifierInfo("__exception_code");
140 Ident_GetExceptionCode = getIdentifierInfo("GetExceptionCode");
141 Ident__abnormal_termination = getIdentifierInfo("_abnormal_termination");
142 Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
143 Ident_AbnormalTermination = getIdentifierInfo("AbnormalTermination");
145 Ident__exception_info = Ident__exception_code = nullptr;
146 Ident__abnormal_termination = Ident___exception_info = nullptr;
147 Ident___exception_code = Ident___abnormal_termination = nullptr;
148 Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr;
149 Ident_AbnormalTermination = nullptr;
152 // If using a PCH with a through header, start skipping tokens.
153 if (!this->PPOpts->PCHThroughHeader.empty() &&
154 !this->PPOpts->ImplicitPCHInclude.empty())
155 SkippingUntilPCHThroughHeader = true;
157 if (this->PPOpts->GeneratePreamble)
158 PreambleConditionalStack.startRecording();
161 Preprocessor::~Preprocessor() {
162 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
164 IncludeMacroStack.clear();
166 // Destroy any macro definitions.
167 while (MacroInfoChain *I = MIChainHead) {
168 MIChainHead = I->Next;
169 I->~MacroInfoChain();
172 // Free any cached macro expanders.
173 // This populates MacroArgCache, so all TokenLexers need to be destroyed
174 // before the code below that frees up the MacroArgCache list.
175 std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
176 CurTokenLexer.reset();
178 // Free any cached MacroArgs.
179 for (MacroArgs *ArgList = MacroArgCache; ArgList;)
180 ArgList = ArgList->deallocate();
182 // Delete the header search info, if we own it.
183 if (OwnsHeaderSearch)
187 void Preprocessor::Initialize(const TargetInfo &Target,
188 const TargetInfo *AuxTarget) {
189 assert((!this->Target || this->Target == &Target) &&
190 "Invalid override of target information");
191 this->Target = &Target;
193 assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
194 "Invalid override of aux target information.");
195 this->AuxTarget = AuxTarget;
197 // Initialize information about built-ins.
198 BuiltinInfo.InitializeTarget(Target, AuxTarget);
199 HeaderInfo.setTarget(Target);
201 // Populate the identifier table with info about keywords for the current language.
202 Identifiers.AddKeywords(LangOpts);
205 void Preprocessor::InitializeForModelFile() {
206 NumEnteredSourceFiles = 0;
209 PragmaHandlersBackup = std::move(PragmaHandlers);
210 PragmaHandlers = llvm::make_unique<PragmaNamespace>(StringRef());
211 RegisterBuiltinPragmas();
213 // Reset PredefinesFileID
214 PredefinesFileID = FileID();
217 void Preprocessor::FinalizeForModelFile() {
218 NumEnteredSourceFiles = 1;
220 PragmaHandlers = std::move(PragmaHandlersBackup);
223 void Preprocessor::setPTHManager(PTHManager* pm) {
225 FileMgr.addStatCache(PTH->createStatCache());
228 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
229 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
230 << getSpelling(Tok) << "'";
232 if (!DumpFlags) return;
234 llvm::errs() << "\t";
235 if (Tok.isAtStartOfLine())
236 llvm::errs() << " [StartOfLine]";
237 if (Tok.hasLeadingSpace())
238 llvm::errs() << " [LeadingSpace]";
239 if (Tok.isExpandDisabled())
240 llvm::errs() << " [ExpandDisabled]";
241 if (Tok.needsCleaning()) {
242 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
243 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
247 llvm::errs() << "\tLoc=<";
248 DumpLocation(Tok.getLocation());
252 void Preprocessor::DumpLocation(SourceLocation Loc) const {
256 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
257 llvm::errs() << "MACRO: ";
258 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
259 DumpToken(MI.getReplacementToken(i));
262 llvm::errs() << "\n";
265 void Preprocessor::PrintStats() {
266 llvm::errs() << "\n*** Preprocessor Stats:\n";
267 llvm::errs() << NumDirectives << " directives found:\n";
268 llvm::errs() << " " << NumDefined << " #define.\n";
269 llvm::errs() << " " << NumUndefined << " #undef.\n";
270 llvm::errs() << " #include/#include_next/#import:\n";
271 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
272 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
273 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
274 llvm::errs() << " " << NumElse << " #else/#elif.\n";
275 llvm::errs() << " " << NumEndif << " #endif.\n";
276 llvm::errs() << " " << NumPragma << " #pragma.\n";
277 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
279 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
280 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
281 << NumFastMacroExpanded << " on the fast path.\n";
282 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
283 << " token paste (##) operations performed, "
284 << NumFastTokenPaste << " on the fast path.\n";
286 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
288 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
289 llvm::errs() << "\n Macro Expanded Tokens: "
290 << llvm::capacity_in_bytes(MacroExpandedTokens);
291 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
292 // FIXME: List information for all submodules.
293 llvm::errs() << "\n Macros: "
294 << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
295 llvm::errs() << "\n #pragma push_macro Info: "
296 << llvm::capacity_in_bytes(PragmaPushMacroInfo);
297 llvm::errs() << "\n Poison Reasons: "
298 << llvm::capacity_in_bytes(PoisonReasons);
299 llvm::errs() << "\n Comment Handlers: "
300 << llvm::capacity_in_bytes(CommentHandlers) << "\n";
303 Preprocessor::macro_iterator
304 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
305 if (IncludeExternalMacros && ExternalSource &&
306 !ReadMacrosFromExternalSource) {
307 ReadMacrosFromExternalSource = true;
308 ExternalSource->ReadDefinedMacros();
311 // Make sure we cover all macros in visible modules.
312 for (const ModuleMacro &Macro : ModuleMacros)
313 CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
315 return CurSubmoduleState->Macros.begin();
318 size_t Preprocessor::getTotalMemory() const {
319 return BP.getTotalMemory()
320 + llvm::capacity_in_bytes(MacroExpandedTokens)
321 + Predefines.capacity() /* Predefines buffer. */
322 // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
324 + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
325 + llvm::capacity_in_bytes(PragmaPushMacroInfo)
326 + llvm::capacity_in_bytes(PoisonReasons)
327 + llvm::capacity_in_bytes(CommentHandlers);
330 Preprocessor::macro_iterator
331 Preprocessor::macro_end(bool IncludeExternalMacros) const {
332 if (IncludeExternalMacros && ExternalSource &&
333 !ReadMacrosFromExternalSource) {
334 ReadMacrosFromExternalSource = true;
335 ExternalSource->ReadDefinedMacros();
338 return CurSubmoduleState->Macros.end();
341 /// Compares macro tokens with a specified token value sequence.
342 static bool MacroDefinitionEquals(const MacroInfo *MI,
343 ArrayRef<TokenValue> Tokens) {
344 return Tokens.size() == MI->getNumTokens() &&
345 std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
348 StringRef Preprocessor::getLastMacroWithSpelling(
350 ArrayRef<TokenValue> Tokens) const {
351 SourceLocation BestLocation;
352 StringRef BestSpelling;
353 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
355 const MacroDirective::DefInfo
356 Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
357 if (!Def || !Def.getMacroInfo())
359 if (!Def.getMacroInfo()->isObjectLike())
361 if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
363 SourceLocation Location = Def.getLocation();
364 // Choose the macro defined latest.
365 if (BestLocation.isInvalid() ||
366 (Location.isValid() &&
367 SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
368 BestLocation = Location;
369 BestSpelling = I->first->getName();
375 void Preprocessor::recomputeCurLexerKind() {
377 CurLexerKind = CLK_Lexer;
378 else if (CurPTHLexer)
379 CurLexerKind = CLK_PTHLexer;
380 else if (CurTokenLexer)
381 CurLexerKind = CLK_TokenLexer;
383 CurLexerKind = CLK_CachingLexer;
386 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
387 unsigned CompleteLine,
388 unsigned CompleteColumn) {
390 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
391 assert(!CodeCompletionFile && "Already set");
393 using llvm::MemoryBuffer;
395 // Load the actual file's contents.
396 bool Invalid = false;
397 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
401 // Find the byte position of the truncation point.
402 const char *Position = Buffer->getBufferStart();
403 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
404 for (; *Position; ++Position) {
405 if (*Position != '\r' && *Position != '\n')
408 // Eat \r\n or \n\r as a single line.
409 if ((Position[1] == '\r' || Position[1] == '\n') &&
410 Position[0] != Position[1])
417 Position += CompleteColumn - 1;
419 // If pointing inside the preamble, adjust the position at the beginning of
420 // the file after the preamble.
421 if (SkipMainFilePreamble.first &&
422 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
423 if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
424 Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
427 if (Position > Buffer->getBufferEnd())
428 Position = Buffer->getBufferEnd();
430 CodeCompletionFile = File;
431 CodeCompletionOffset = Position - Buffer->getBufferStart();
433 auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
434 Buffer->getBufferSize() + 1, Buffer->getBufferIdentifier());
435 char *NewBuf = NewBuffer->getBufferStart();
436 char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
438 std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
439 SourceMgr.overrideFileContents(File, std::move(NewBuffer));
444 void Preprocessor::CodeCompleteNaturalLanguage() {
446 CodeComplete->CodeCompleteNaturalLanguage();
447 setCodeCompletionReached();
450 /// getSpelling - This method is used to get the spelling of a token into a
451 /// SmallVector. Note that the returned StringRef may not point to the
452 /// supplied buffer if a copy can be avoided.
453 StringRef Preprocessor::getSpelling(const Token &Tok,
454 SmallVectorImpl<char> &Buffer,
455 bool *Invalid) const {
456 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
457 if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
458 // Try the fast path.
459 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
460 return II->getName();
463 // Resize the buffer if we need to copy into it.
464 if (Tok.needsCleaning())
465 Buffer.resize(Tok.getLength());
467 const char *Ptr = Buffer.data();
468 unsigned Len = getSpelling(Tok, Ptr, Invalid);
469 return StringRef(Ptr, Len);
472 /// CreateString - Plop the specified string into a scratch buffer and return a
473 /// location for it. If specified, the source location provides a source
474 /// location for the token.
475 void Preprocessor::CreateString(StringRef Str, Token &Tok,
476 SourceLocation ExpansionLocStart,
477 SourceLocation ExpansionLocEnd) {
478 Tok.setLength(Str.size());
481 SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
483 if (ExpansionLocStart.isValid())
484 Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
485 ExpansionLocEnd, Str.size());
486 Tok.setLocation(Loc);
488 // If this is a raw identifier or a literal token, set the pointer data.
489 if (Tok.is(tok::raw_identifier))
490 Tok.setRawIdentifierData(DestPtr);
491 else if (Tok.isLiteral())
492 Tok.setLiteralData(DestPtr);
495 SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
496 auto &SM = getSourceManager();
497 SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
498 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(SpellingLoc);
499 bool Invalid = false;
500 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
502 return SourceLocation();
504 // FIXME: We could consider re-using spelling for tokens we see repeatedly.
506 SourceLocation Spelling =
507 ScratchBuf->getToken(Buffer.data() + LocInfo.second, Length, DestPtr);
508 return SM.createTokenSplitLoc(Spelling, Loc, Loc.getLocWithOffset(Length));
511 Module *Preprocessor::getCurrentModule() {
512 if (!getLangOpts().isCompilingModule())
515 return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
518 //===----------------------------------------------------------------------===//
519 // Preprocessor Initialization Methods
520 //===----------------------------------------------------------------------===//
522 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
523 /// which implicitly adds the builtin defines etc.
524 void Preprocessor::EnterMainSourceFile() {
525 // We do not allow the preprocessor to reenter the main file. Doing so will
526 // cause FileID's to accumulate information from both runs (e.g. #line
527 // information) and predefined macros aren't guaranteed to be set properly.
528 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
529 FileID MainFileID = SourceMgr.getMainFileID();
531 // If MainFileID is loaded it means we loaded an AST file, no need to enter
533 if (!SourceMgr.isLoadedFileID(MainFileID)) {
534 // Enter the main file source buffer.
535 EnterSourceFile(MainFileID, nullptr, SourceLocation());
537 // If we've been asked to skip bytes in the main file (e.g., as part of a
538 // precompiled preamble), do so now.
539 if (SkipMainFilePreamble.first > 0)
540 CurLexer->SetByteOffset(SkipMainFilePreamble.first,
541 SkipMainFilePreamble.second);
543 // Tell the header info that the main file was entered. If the file is later
544 // #imported, it won't be re-entered.
545 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
546 HeaderInfo.IncrementIncludeCount(FE);
549 // Preprocess Predefines to populate the initial preprocessor state.
550 std::unique_ptr<llvm::MemoryBuffer> SB =
551 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
552 assert(SB && "Cannot create predefined source buffer");
553 FileID FID = SourceMgr.createFileID(std::move(SB));
554 assert(FID.isValid() && "Could not create FileID for predefines?");
555 setPredefinesFileID(FID);
557 // Start parsing the predefines.
558 EnterSourceFile(FID, nullptr, SourceLocation());
560 if (!PPOpts->PCHThroughHeader.empty()) {
561 // Lookup and save the FileID for the through header. If it isn't found
562 // in the search path, it's a fatal error.
563 const DirectoryLookup *CurDir;
564 const FileEntry *File = LookupFile(
565 SourceLocation(), PPOpts->PCHThroughHeader,
566 /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, CurDir,
567 /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
568 /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr);
570 Diag(SourceLocation(), diag::err_pp_through_header_not_found)
571 << PPOpts->PCHThroughHeader;
574 setPCHThroughHeaderFileID(
575 SourceMgr.createFileID(File, SourceLocation(), SrcMgr::C_User));
578 // Skip tokens from the Predefines and if needed the main file.
579 if (usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader)
580 SkipTokensUntilPCHThroughHeader();
583 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
584 assert(PCHThroughHeaderFileID.isInvalid() &&
585 "PCHThroughHeaderFileID already set!");
586 PCHThroughHeaderFileID = FID;
589 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
590 assert(PCHThroughHeaderFileID.isValid() &&
591 "Invalid PCH through header FileID");
592 return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
595 bool Preprocessor::creatingPCHWithThroughHeader() {
596 return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
597 PCHThroughHeaderFileID.isValid();
600 bool Preprocessor::usingPCHWithThroughHeader() {
601 return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
602 PCHThroughHeaderFileID.isValid();
605 /// Skip tokens until after the #include of the through header.
606 /// Tokens in the predefines file and the main file may be skipped. If the end
607 /// of the predefines file is reached, skipping continues into the main file.
608 /// If the end of the main file is reached, it's a fatal error.
609 void Preprocessor::SkipTokensUntilPCHThroughHeader() {
610 bool ReachedMainFileEOF = false;
613 bool InPredefines = (CurLexer->getFileID() == getPredefinesFileID());
615 if (Tok.is(tok::eof) && !InPredefines) {
616 ReachedMainFileEOF = true;
619 if (!SkippingUntilPCHThroughHeader)
622 if (ReachedMainFileEOF)
623 Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
624 << PPOpts->PCHThroughHeader << 1;
627 void Preprocessor::replayPreambleConditionalStack() {
628 // Restore the conditional stack from the preamble, if there is one.
629 if (PreambleConditionalStack.isReplaying()) {
631 "CurPPLexer is null when calling replayPreambleConditionalStack.");
632 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
633 PreambleConditionalStack.doneReplaying();
634 if (PreambleConditionalStack.reachedEOFWhileSkipping())
635 SkipExcludedConditionalBlock(
636 PreambleConditionalStack.SkipInfo->HashTokenLoc,
637 PreambleConditionalStack.SkipInfo->IfTokenLoc,
638 PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
639 PreambleConditionalStack.SkipInfo->FoundElse,
640 PreambleConditionalStack.SkipInfo->ElseLoc);
644 void Preprocessor::EndSourceFile() {
645 // Notify the client that we reached the end of the source file.
647 Callbacks->EndOfMainFile();
650 //===----------------------------------------------------------------------===//
651 // Lexer Event Handling.
652 //===----------------------------------------------------------------------===//
654 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
655 /// identifier information for the token and install it into the token,
656 /// updating the token kind accordingly.
657 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
658 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
660 // Look up this token, see if it is a macro, or if it is a language keyword.
662 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
663 // No cleaning needed, just use the characters from the lexed buffer.
664 II = getIdentifierInfo(Identifier.getRawIdentifier());
666 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
667 SmallString<64> IdentifierBuffer;
668 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
670 if (Identifier.hasUCN()) {
671 SmallString<64> UCNIdentifierBuffer;
672 expandUCNs(UCNIdentifierBuffer, CleanedStr);
673 II = getIdentifierInfo(UCNIdentifierBuffer);
675 II = getIdentifierInfo(CleanedStr);
679 // Update the token info (identifier info and appropriate token kind).
680 Identifier.setIdentifierInfo(II);
681 if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
682 getSourceManager().isInSystemHeader(Identifier.getLocation()))
683 Identifier.setKind(tok::identifier);
685 Identifier.setKind(II->getTokenID());
690 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
691 PoisonReasons[II] = DiagID;
694 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
695 assert(Ident__exception_code && Ident__exception_info);
696 assert(Ident___exception_code && Ident___exception_info);
697 Ident__exception_code->setIsPoisoned(Poison);
698 Ident___exception_code->setIsPoisoned(Poison);
699 Ident_GetExceptionCode->setIsPoisoned(Poison);
700 Ident__exception_info->setIsPoisoned(Poison);
701 Ident___exception_info->setIsPoisoned(Poison);
702 Ident_GetExceptionInfo->setIsPoisoned(Poison);
703 Ident__abnormal_termination->setIsPoisoned(Poison);
704 Ident___abnormal_termination->setIsPoisoned(Poison);
705 Ident_AbnormalTermination->setIsPoisoned(Poison);
708 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
709 assert(Identifier.getIdentifierInfo() &&
710 "Can't handle identifiers without identifier info!");
711 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
712 PoisonReasons.find(Identifier.getIdentifierInfo());
713 if(it == PoisonReasons.end())
714 Diag(Identifier, diag::err_pp_used_poisoned_id);
716 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
719 /// Returns a diagnostic message kind for reporting a future keyword as
720 /// appropriate for the identifier and specified language.
721 static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
722 const LangOptions &LangOpts) {
723 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
725 if (LangOpts.CPlusPlus)
726 return llvm::StringSwitch<diag::kind>(II.getName())
727 #define CXX11_KEYWORD(NAME, FLAGS) \
728 .Case(#NAME, diag::warn_cxx11_keyword)
729 #define CXX2A_KEYWORD(NAME, FLAGS) \
730 .Case(#NAME, diag::warn_cxx2a_keyword)
731 #include "clang/Basic/TokenKinds.def"
735 "Keyword not known to come from a newer Standard or proposed Standard");
738 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
739 assert(II.isOutOfDate() && "not out of date");
740 getExternalSource()->updateOutOfDateIdentifier(II);
743 /// HandleIdentifier - This callback is invoked when the lexer reads an
744 /// identifier. This callback looks up the identifier in the map and/or
745 /// potentially macro expands it or turns it into a named token (like 'for').
747 /// Note that callers of this method are guarded by checking the
748 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
749 /// IdentifierInfo methods that compute these properties will need to change to
751 bool Preprocessor::HandleIdentifier(Token &Identifier) {
752 assert(Identifier.getIdentifierInfo() &&
753 "Can't handle identifiers without identifier info!");
755 IdentifierInfo &II = *Identifier.getIdentifierInfo();
757 // If the information about this identifier is out of date, update it from
758 // the external source.
759 // We have to treat __VA_ARGS__ in a special way, since it gets
760 // serialized with isPoisoned = true, but our preprocessor may have
761 // unpoisoned it if we're defining a C99 macro.
762 if (II.isOutOfDate()) {
763 bool CurrentIsPoisoned = false;
764 const bool IsSpecialVariadicMacro =
765 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
766 if (IsSpecialVariadicMacro)
767 CurrentIsPoisoned = II.isPoisoned();
769 updateOutOfDateIdentifier(II);
770 Identifier.setKind(II.getTokenID());
772 if (IsSpecialVariadicMacro)
773 II.setIsPoisoned(CurrentIsPoisoned);
776 // If this identifier was poisoned, and if it was not produced from a macro
777 // expansion, emit an error.
778 if (II.isPoisoned() && CurPPLexer) {
779 HandlePoisonedIdentifier(Identifier);
782 // If this is a macro to be expanded, do it.
783 if (MacroDefinition MD = getMacroDefinition(&II)) {
784 auto *MI = MD.getMacroInfo();
785 assert(MI && "macro definition with no macro info?");
786 if (!DisableMacroExpansion) {
787 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
788 // C99 6.10.3p10: If the preprocessing token immediately after the
789 // macro name isn't a '(', this macro should not be expanded.
790 if (!MI->isFunctionLike() || isNextPPTokenLParen())
791 return HandleMacroExpandedIdentifier(Identifier, MD);
793 // C99 6.10.3.4p2 says that a disabled macro may never again be
794 // expanded, even if it's in a context where it could be expanded in the
796 Identifier.setFlag(Token::DisableExpand);
797 if (MI->isObjectLike() || isNextPPTokenLParen())
798 Diag(Identifier, diag::pp_disabled_macro_expansion);
803 // If this identifier is a keyword in a newer Standard or proposed Standard,
804 // produce a warning. Don't warn if we're not considering macro expansion,
805 // since this identifier might be the name of a macro.
806 // FIXME: This warning is disabled in cases where it shouldn't be, like
807 // "#define constexpr constexpr", "int constexpr;"
808 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
809 Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
811 // Don't diagnose this keyword again in this translation unit.
812 II.setIsFutureCompatKeyword(false);
815 // If this is an extension token, diagnose its use.
816 // We avoid diagnosing tokens that originate from macro definitions.
817 // FIXME: This warning is disabled in cases where it shouldn't be,
818 // like "#define TY typeof", "TY(1) x".
819 if (II.isExtensionToken() && !DisableMacroExpansion)
820 Diag(Identifier, diag::ext_token_used);
822 // If this is the 'import' contextual keyword following an '@', note
823 // that the next token indicates a module name.
825 // Note that we do not treat 'import' as a contextual
826 // keyword when we're in a caching lexer, because caching lexers only get
827 // used in contexts where import declarations are disallowed.
829 // Likewise if this is the C++ Modules TS import keyword.
830 if (((LastTokenWasAt && II.isModulesImport()) ||
831 Identifier.is(tok::kw_import)) &&
832 !InMacroArgs && !DisableMacroExpansion &&
833 (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
834 CurLexerKind != CLK_CachingLexer) {
835 ModuleImportLoc = Identifier.getLocation();
836 ModuleImportPath.clear();
837 ModuleImportExpectsIdentifier = true;
838 CurLexerKind = CLK_LexAfterModuleImport;
843 void Preprocessor::Lex(Token &Result) {
844 // We loop here until a lex function returns a token; this avoids recursion.
847 switch (CurLexerKind) {
849 ReturnedToken = CurLexer->Lex(Result);
852 ReturnedToken = CurPTHLexer->Lex(Result);
855 ReturnedToken = CurTokenLexer->Lex(Result);
857 case CLK_CachingLexer:
859 ReturnedToken = true;
861 case CLK_LexAfterModuleImport:
862 LexAfterModuleImport(Result);
863 ReturnedToken = true;
866 } while (!ReturnedToken);
868 if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
869 // Remember the identifier before code completion token.
870 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
871 // Set IdenfitierInfo to null to avoid confusing code that handles both
872 // identifiers and completion tokens.
873 Result.setIdentifierInfo(nullptr);
876 LastTokenWasAt = Result.is(tok::at);
879 /// Lex a token following the 'import' contextual keyword.
881 void Preprocessor::LexAfterModuleImport(Token &Result) {
882 // Figure out what kind of lexer we actually have.
883 recomputeCurLexerKind();
885 // Lex the next token.
888 // The token sequence
890 // import identifier (. identifier)*
892 // indicates a module import directive. We already saw the 'import'
893 // contextual keyword, so now we're looking for the identifiers.
894 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
895 // We expected to see an identifier here, and we did; continue handling
897 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
898 Result.getLocation()));
899 ModuleImportExpectsIdentifier = false;
900 CurLexerKind = CLK_LexAfterModuleImport;
904 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
905 // see the next identifier. (We can also see a '[[' that begins an
906 // attribute-specifier-seq here under the C++ Modules TS.)
907 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
908 ModuleImportExpectsIdentifier = true;
909 CurLexerKind = CLK_LexAfterModuleImport;
913 // If we have a non-empty module path, load the named module.
914 if (!ModuleImportPath.empty()) {
915 // Under the Modules TS, the dot is just part of the module name, and not
916 // a real hierarachy separator. Flatten such module names now.
918 // FIXME: Is this the right level to be performing this transformation?
919 std::string FlatModuleName;
920 if (getLangOpts().ModulesTS) {
921 for (auto &Piece : ModuleImportPath) {
922 if (!FlatModuleName.empty())
923 FlatModuleName += ".";
924 FlatModuleName += Piece.first->getName();
926 SourceLocation FirstPathLoc = ModuleImportPath[0].second;
927 ModuleImportPath.clear();
928 ModuleImportPath.push_back(
929 std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
932 Module *Imported = nullptr;
933 if (getLangOpts().Modules) {
934 Imported = TheModuleLoader.loadModule(ModuleImportLoc,
937 /*IsIncludeDirective=*/false);
939 makeModuleVisible(Imported, ModuleImportLoc);
941 if (Callbacks && (getLangOpts().Modules || getLangOpts().DebuggerSupport))
942 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
946 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
947 CurSubmoduleState->VisibleModules.setVisible(
948 M, Loc, [](Module *) {},
949 [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
950 // FIXME: Include the path in the diagnostic.
951 // FIXME: Include the import location for the conflicting module.
952 Diag(ModuleImportLoc, diag::warn_module_conflict)
953 << Path[0]->getFullModuleName()
954 << Conflict->getFullModuleName()
958 // Add this module to the imports list of the currently-built submodule.
959 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
960 BuildingSubmoduleStack.back().M->Imports.insert(M);
963 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
964 const char *DiagnosticTag,
965 bool AllowMacroExpansion) {
966 // We need at least one string literal.
967 if (Result.isNot(tok::string_literal)) {
968 Diag(Result, diag::err_expected_string_literal)
969 << /*Source='in...'*/0 << DiagnosticTag;
973 // Lex string literal tokens, optionally with macro expansion.
974 SmallVector<Token, 4> StrToks;
976 StrToks.push_back(Result);
978 if (Result.hasUDSuffix())
979 Diag(Result, diag::err_invalid_string_udl);
981 if (AllowMacroExpansion)
984 LexUnexpandedToken(Result);
985 } while (Result.is(tok::string_literal));
987 // Concatenate and parse the strings.
988 StringLiteralParser Literal(StrToks, *this);
989 assert(Literal.isAscii() && "Didn't allow wide strings in");
991 if (Literal.hadError)
994 if (Literal.Pascal) {
995 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
996 << /*Source='in...'*/0 << DiagnosticTag;
1000 String = Literal.GetString();
1004 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1005 assert(Tok.is(tok::numeric_constant));
1006 SmallString<8> IntegerBuffer;
1007 bool NumberInvalid = false;
1008 StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1011 NumericLiteralParser Literal(Spelling, Tok.getLocation(), *this);
1012 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1014 llvm::APInt APVal(64, 0);
1015 if (Literal.GetIntegerValue(APVal))
1018 Value = APVal.getLimitedValue();
1022 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1023 assert(Handler && "NULL comment handler");
1024 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
1025 CommentHandlers.end() && "Comment handler already registered");
1026 CommentHandlers.push_back(Handler);
1029 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1030 std::vector<CommentHandler *>::iterator Pos =
1031 std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
1032 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1033 CommentHandlers.erase(Pos);
1036 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1037 bool AnyPendingTokens = false;
1038 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1039 HEnd = CommentHandlers.end();
1041 if ((*H)->HandleComment(*this, Comment))
1042 AnyPendingTokens = true;
1044 if (!AnyPendingTokens || getCommentRetentionState())
1050 ModuleLoader::~ModuleLoader() = default;
1052 CommentHandler::~CommentHandler() = default;
1054 CodeCompletionHandler::~CodeCompletionHandler() = default;
1056 void Preprocessor::createPreprocessingRecord() {
1060 Record = new PreprocessingRecord(getSourceManager());
1061 addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));