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/Pragma.h"
48 #include "clang/Lex/PreprocessingRecord.h"
49 #include "clang/Lex/PreprocessorLexer.h"
50 #include "clang/Lex/PreprocessorOptions.h"
51 #include "clang/Lex/ScratchBuffer.h"
52 #include "clang/Lex/Token.h"
53 #include "clang/Lex/TokenLexer.h"
54 #include "llvm/ADT/APInt.h"
55 #include "llvm/ADT/ArrayRef.h"
56 #include "llvm/ADT/DenseMap.h"
57 #include "llvm/ADT/SmallString.h"
58 #include "llvm/ADT/SmallVector.h"
59 #include "llvm/ADT/STLExtras.h"
60 #include "llvm/ADT/StringRef.h"
61 #include "llvm/ADT/StringSwitch.h"
62 #include "llvm/Support/Capacity.h"
63 #include "llvm/Support/ErrorHandling.h"
64 #include "llvm/Support/MemoryBuffer.h"
65 #include "llvm/Support/raw_ostream.h"
73 using namespace clang;
75 LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
77 ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
79 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
80 DiagnosticsEngine &diags, LangOptions &opts,
81 SourceManager &SM, MemoryBufferCache &PCMCache,
82 HeaderSearch &Headers, ModuleLoader &TheModuleLoader,
83 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
84 TranslationUnitKind TUKind)
85 : PPOpts(std::move(PPOpts)), Diags(&diags), LangOpts(opts),
86 FileMgr(Headers.getFileMgr()), SourceMgr(SM), PCMCache(PCMCache),
87 ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers),
88 TheModuleLoader(TheModuleLoader), ExternalSource(nullptr),
89 // As the language options may have not been loaded yet (when
90 // deserializing an ASTUnit), adding keywords to the identifier table is
91 // deferred to Preprocessor::Initialize().
92 Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())),
93 TUKind(TUKind), SkipMainFilePreamble(0, true),
94 CurSubmoduleState(&NullSubmoduleState) {
95 OwnsHeaderSearch = OwnsHeaders;
97 // Default to discarding comments.
99 KeepMacroComments = false;
100 SuppressIncludeNotFoundError = false;
102 // Macro expansion is enabled.
103 DisableMacroExpansion = false;
104 MacroExpansionInDirectivesOverride = false;
106 InMacroArgPreExpansion = false;
107 NumCachedTokenLexers = 0;
108 PragmasEnabled = true;
109 ParsingIfOrElifDirective = false;
110 PreprocessedOutput = false;
112 // We haven't read anything from the external source.
113 ReadMacrosFromExternalSource = false;
115 // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of
116 // a macro. They get unpoisoned where it is allowed.
117 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
118 SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
119 if (getLangOpts().CPlusPlus2a) {
120 (Ident__VA_OPT__ = getIdentifierInfo("__VA_OPT__"))->setIsPoisoned();
121 SetPoisonReason(Ident__VA_OPT__,diag::ext_pp_bad_vaopt_use);
123 Ident__VA_OPT__ = nullptr;
126 // Initialize the pragma handlers.
127 RegisterBuiltinPragmas();
129 // Initialize builtin macros like __LINE__ and friends.
130 RegisterBuiltinMacros();
132 if(LangOpts.Borland) {
133 Ident__exception_info = getIdentifierInfo("_exception_info");
134 Ident___exception_info = getIdentifierInfo("__exception_info");
135 Ident_GetExceptionInfo = getIdentifierInfo("GetExceptionInformation");
136 Ident__exception_code = getIdentifierInfo("_exception_code");
137 Ident___exception_code = getIdentifierInfo("__exception_code");
138 Ident_GetExceptionCode = getIdentifierInfo("GetExceptionCode");
139 Ident__abnormal_termination = getIdentifierInfo("_abnormal_termination");
140 Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
141 Ident_AbnormalTermination = getIdentifierInfo("AbnormalTermination");
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;
150 // If using a PCH where a #pragma hdrstop is expected, start skipping tokens.
151 if (usingPCHWithPragmaHdrStop())
152 SkippingUntilPragmaHdrStop = true;
154 // If using a PCH with a through header, start skipping tokens.
155 if (!this->PPOpts->PCHThroughHeader.empty() &&
156 !this->PPOpts->ImplicitPCHInclude.empty())
157 SkippingUntilPCHThroughHeader = true;
159 if (this->PPOpts->GeneratePreamble)
160 PreambleConditionalStack.startRecording();
163 Preprocessor::~Preprocessor() {
164 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
166 IncludeMacroStack.clear();
168 // Destroy any macro definitions.
169 while (MacroInfoChain *I = MIChainHead) {
170 MIChainHead = I->Next;
171 I->~MacroInfoChain();
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(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
178 CurTokenLexer.reset();
180 // Free any cached MacroArgs.
181 for (MacroArgs *ArgList = MacroArgCache; ArgList;)
182 ArgList = ArgList->deallocate();
184 // Delete the header search info, if we own it.
185 if (OwnsHeaderSearch)
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;
195 assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
196 "Invalid override of aux target information.");
197 this->AuxTarget = AuxTarget;
199 // Initialize information about built-ins.
200 BuiltinInfo.InitializeTarget(Target, AuxTarget);
201 HeaderInfo.setTarget(Target);
203 // Populate the identifier table with info about keywords for the current language.
204 Identifiers.AddKeywords(LangOpts);
207 void Preprocessor::InitializeForModelFile() {
208 NumEnteredSourceFiles = 0;
211 PragmaHandlersBackup = std::move(PragmaHandlers);
212 PragmaHandlers = llvm::make_unique<PragmaNamespace>(StringRef());
213 RegisterBuiltinPragmas();
215 // Reset PredefinesFileID
216 PredefinesFileID = FileID();
219 void Preprocessor::FinalizeForModelFile() {
220 NumEnteredSourceFiles = 1;
222 PragmaHandlers = std::move(PragmaHandlersBackup);
225 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
226 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
227 << getSpelling(Tok) << "'";
229 if (!DumpFlags) return;
231 llvm::errs() << "\t";
232 if (Tok.isAtStartOfLine())
233 llvm::errs() << " [StartOfLine]";
234 if (Tok.hasLeadingSpace())
235 llvm::errs() << " [LeadingSpace]";
236 if (Tok.isExpandDisabled())
237 llvm::errs() << " [ExpandDisabled]";
238 if (Tok.needsCleaning()) {
239 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
240 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
244 llvm::errs() << "\tLoc=<";
245 DumpLocation(Tok.getLocation());
249 void Preprocessor::DumpLocation(SourceLocation Loc) const {
250 Loc.print(llvm::errs(), SourceMgr);
253 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
254 llvm::errs() << "MACRO: ";
255 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
256 DumpToken(MI.getReplacementToken(i));
259 llvm::errs() << "\n";
262 void Preprocessor::PrintStats() {
263 llvm::errs() << "\n*** Preprocessor Stats:\n";
264 llvm::errs() << NumDirectives << " directives found:\n";
265 llvm::errs() << " " << NumDefined << " #define.\n";
266 llvm::errs() << " " << NumUndefined << " #undef.\n";
267 llvm::errs() << " #include/#include_next/#import:\n";
268 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
269 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
270 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
271 llvm::errs() << " " << NumElse << " #else/#elif.\n";
272 llvm::errs() << " " << NumEndif << " #endif.\n";
273 llvm::errs() << " " << NumPragma << " #pragma.\n";
274 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
276 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
277 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
278 << NumFastMacroExpanded << " on the fast path.\n";
279 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
280 << " token paste (##) operations performed, "
281 << NumFastTokenPaste << " on the fast path.\n";
283 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
285 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
286 llvm::errs() << "\n Macro Expanded Tokens: "
287 << llvm::capacity_in_bytes(MacroExpandedTokens);
288 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
289 // FIXME: List information for all submodules.
290 llvm::errs() << "\n Macros: "
291 << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
292 llvm::errs() << "\n #pragma push_macro Info: "
293 << llvm::capacity_in_bytes(PragmaPushMacroInfo);
294 llvm::errs() << "\n Poison Reasons: "
295 << llvm::capacity_in_bytes(PoisonReasons);
296 llvm::errs() << "\n Comment Handlers: "
297 << llvm::capacity_in_bytes(CommentHandlers) << "\n";
300 Preprocessor::macro_iterator
301 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
302 if (IncludeExternalMacros && ExternalSource &&
303 !ReadMacrosFromExternalSource) {
304 ReadMacrosFromExternalSource = true;
305 ExternalSource->ReadDefinedMacros();
308 // Make sure we cover all macros in visible modules.
309 for (const ModuleMacro &Macro : ModuleMacros)
310 CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
312 return CurSubmoduleState->Macros.begin();
315 size_t Preprocessor::getTotalMemory() const {
316 return BP.getTotalMemory()
317 + llvm::capacity_in_bytes(MacroExpandedTokens)
318 + Predefines.capacity() /* Predefines buffer. */
319 // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
321 + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
322 + llvm::capacity_in_bytes(PragmaPushMacroInfo)
323 + llvm::capacity_in_bytes(PoisonReasons)
324 + llvm::capacity_in_bytes(CommentHandlers);
327 Preprocessor::macro_iterator
328 Preprocessor::macro_end(bool IncludeExternalMacros) const {
329 if (IncludeExternalMacros && ExternalSource &&
330 !ReadMacrosFromExternalSource) {
331 ReadMacrosFromExternalSource = true;
332 ExternalSource->ReadDefinedMacros();
335 return CurSubmoduleState->Macros.end();
338 /// Compares macro tokens with a specified token value sequence.
339 static bool MacroDefinitionEquals(const MacroInfo *MI,
340 ArrayRef<TokenValue> Tokens) {
341 return Tokens.size() == MI->getNumTokens() &&
342 std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
345 StringRef Preprocessor::getLastMacroWithSpelling(
347 ArrayRef<TokenValue> Tokens) const {
348 SourceLocation BestLocation;
349 StringRef BestSpelling;
350 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
352 const MacroDirective::DefInfo
353 Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
354 if (!Def || !Def.getMacroInfo())
356 if (!Def.getMacroInfo()->isObjectLike())
358 if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
360 SourceLocation Location = Def.getLocation();
361 // Choose the macro defined latest.
362 if (BestLocation.isInvalid() ||
363 (Location.isValid() &&
364 SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
365 BestLocation = Location;
366 BestSpelling = I->first->getName();
372 void Preprocessor::recomputeCurLexerKind() {
374 CurLexerKind = CLK_Lexer;
375 else if (CurTokenLexer)
376 CurLexerKind = CLK_TokenLexer;
378 CurLexerKind = CLK_CachingLexer;
381 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
382 unsigned CompleteLine,
383 unsigned CompleteColumn) {
385 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
386 assert(!CodeCompletionFile && "Already set");
388 using llvm::MemoryBuffer;
390 // Load the actual file's contents.
391 bool Invalid = false;
392 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
396 // Find the byte position of the truncation point.
397 const char *Position = Buffer->getBufferStart();
398 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
399 for (; *Position; ++Position) {
400 if (*Position != '\r' && *Position != '\n')
403 // Eat \r\n or \n\r as a single line.
404 if ((Position[1] == '\r' || Position[1] == '\n') &&
405 Position[0] != Position[1])
412 Position += CompleteColumn - 1;
414 // If pointing inside the preamble, adjust the position at the beginning of
415 // the file after the preamble.
416 if (SkipMainFilePreamble.first &&
417 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
418 if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
419 Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
422 if (Position > Buffer->getBufferEnd())
423 Position = Buffer->getBufferEnd();
425 CodeCompletionFile = File;
426 CodeCompletionOffset = Position - Buffer->getBufferStart();
428 auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
429 Buffer->getBufferSize() + 1, Buffer->getBufferIdentifier());
430 char *NewBuf = NewBuffer->getBufferStart();
431 char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
433 std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
434 SourceMgr.overrideFileContents(File, std::move(NewBuffer));
439 void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
442 CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
443 setCodeCompletionReached();
446 void Preprocessor::CodeCompleteNaturalLanguage() {
448 CodeComplete->CodeCompleteNaturalLanguage();
449 setCodeCompletionReached();
452 /// getSpelling - This method is used to get the spelling of a token into a
453 /// SmallVector. Note that the returned StringRef may not point to the
454 /// supplied buffer if a copy can be avoided.
455 StringRef Preprocessor::getSpelling(const Token &Tok,
456 SmallVectorImpl<char> &Buffer,
457 bool *Invalid) const {
458 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
459 if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
460 // Try the fast path.
461 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
462 return II->getName();
465 // Resize the buffer if we need to copy into it.
466 if (Tok.needsCleaning())
467 Buffer.resize(Tok.getLength());
469 const char *Ptr = Buffer.data();
470 unsigned Len = getSpelling(Tok, Ptr, Invalid);
471 return StringRef(Ptr, Len);
474 /// CreateString - Plop the specified string into a scratch buffer and return a
475 /// location for it. If specified, the source location provides a source
476 /// location for the token.
477 void Preprocessor::CreateString(StringRef Str, Token &Tok,
478 SourceLocation ExpansionLocStart,
479 SourceLocation ExpansionLocEnd) {
480 Tok.setLength(Str.size());
483 SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
485 if (ExpansionLocStart.isValid())
486 Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
487 ExpansionLocEnd, Str.size());
488 Tok.setLocation(Loc);
490 // If this is a raw identifier or a literal token, set the pointer data.
491 if (Tok.is(tok::raw_identifier))
492 Tok.setRawIdentifierData(DestPtr);
493 else if (Tok.isLiteral())
494 Tok.setLiteralData(DestPtr);
497 SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
498 auto &SM = getSourceManager();
499 SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
500 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(SpellingLoc);
501 bool Invalid = false;
502 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
504 return SourceLocation();
506 // FIXME: We could consider re-using spelling for tokens we see repeatedly.
508 SourceLocation Spelling =
509 ScratchBuf->getToken(Buffer.data() + LocInfo.second, Length, DestPtr);
510 return SM.createTokenSplitLoc(Spelling, Loc, Loc.getLocWithOffset(Length));
513 Module *Preprocessor::getCurrentModule() {
514 if (!getLangOpts().isCompilingModule())
517 return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
520 //===----------------------------------------------------------------------===//
521 // Preprocessor Initialization Methods
522 //===----------------------------------------------------------------------===//
524 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
525 /// which implicitly adds the builtin defines etc.
526 void Preprocessor::EnterMainSourceFile() {
527 // We do not allow the preprocessor to reenter the main file. Doing so will
528 // cause FileID's to accumulate information from both runs (e.g. #line
529 // information) and predefined macros aren't guaranteed to be set properly.
530 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
531 FileID MainFileID = SourceMgr.getMainFileID();
533 // If MainFileID is loaded it means we loaded an AST file, no need to enter
535 if (!SourceMgr.isLoadedFileID(MainFileID)) {
536 // Enter the main file source buffer.
537 EnterSourceFile(MainFileID, nullptr, SourceLocation());
539 // If we've been asked to skip bytes in the main file (e.g., as part of a
540 // precompiled preamble), do so now.
541 if (SkipMainFilePreamble.first > 0)
542 CurLexer->SetByteOffset(SkipMainFilePreamble.first,
543 SkipMainFilePreamble.second);
545 // Tell the header info that the main file was entered. If the file is later
546 // #imported, it won't be re-entered.
547 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
548 HeaderInfo.IncrementIncludeCount(FE);
551 // Preprocess Predefines to populate the initial preprocessor state.
552 std::unique_ptr<llvm::MemoryBuffer> SB =
553 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
554 assert(SB && "Cannot create predefined source buffer");
555 FileID FID = SourceMgr.createFileID(std::move(SB));
556 assert(FID.isValid() && "Could not create FileID for predefines?");
557 setPredefinesFileID(FID);
559 // Start parsing the predefines.
560 EnterSourceFile(FID, nullptr, SourceLocation());
562 if (!PPOpts->PCHThroughHeader.empty()) {
563 // Lookup and save the FileID for the through header. If it isn't found
564 // in the search path, it's a fatal error.
565 const DirectoryLookup *CurDir;
566 const FileEntry *File = LookupFile(
567 SourceLocation(), PPOpts->PCHThroughHeader,
568 /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, CurDir,
569 /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
570 /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr);
572 Diag(SourceLocation(), diag::err_pp_through_header_not_found)
573 << PPOpts->PCHThroughHeader;
576 setPCHThroughHeaderFileID(
577 SourceMgr.createFileID(File, SourceLocation(), SrcMgr::C_User));
580 // Skip tokens from the Predefines and if needed the main file.
581 if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
582 (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
583 SkipTokensWhileUsingPCH();
586 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
587 assert(PCHThroughHeaderFileID.isInvalid() &&
588 "PCHThroughHeaderFileID already set!");
589 PCHThroughHeaderFileID = FID;
592 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
593 assert(PCHThroughHeaderFileID.isValid() &&
594 "Invalid PCH through header FileID");
595 return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
598 bool Preprocessor::creatingPCHWithThroughHeader() {
599 return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
600 PCHThroughHeaderFileID.isValid();
603 bool Preprocessor::usingPCHWithThroughHeader() {
604 return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
605 PCHThroughHeaderFileID.isValid();
608 bool Preprocessor::creatingPCHWithPragmaHdrStop() {
609 return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
612 bool Preprocessor::usingPCHWithPragmaHdrStop() {
613 return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
616 /// Skip tokens until after the #include of the through header or
617 /// until after a #pragma hdrstop is seen. Tokens in the predefines file
618 /// and the main file may be skipped. If the end of the predefines file
619 /// is reached, skipping continues into the main file. If the end of the
620 /// main file is reached, it's a fatal error.
621 void Preprocessor::SkipTokensWhileUsingPCH() {
622 bool ReachedMainFileEOF = false;
623 bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
624 bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
627 bool InPredefines = (CurLexer->getFileID() == getPredefinesFileID());
629 if (Tok.is(tok::eof) && !InPredefines) {
630 ReachedMainFileEOF = true;
633 if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
635 if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
638 if (ReachedMainFileEOF) {
639 if (UsingPCHThroughHeader)
640 Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
641 << PPOpts->PCHThroughHeader << 1;
642 else if (!PPOpts->PCHWithHdrStopCreate)
643 Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
647 void Preprocessor::replayPreambleConditionalStack() {
648 // Restore the conditional stack from the preamble, if there is one.
649 if (PreambleConditionalStack.isReplaying()) {
651 "CurPPLexer is null when calling replayPreambleConditionalStack.");
652 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
653 PreambleConditionalStack.doneReplaying();
654 if (PreambleConditionalStack.reachedEOFWhileSkipping())
655 SkipExcludedConditionalBlock(
656 PreambleConditionalStack.SkipInfo->HashTokenLoc,
657 PreambleConditionalStack.SkipInfo->IfTokenLoc,
658 PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
659 PreambleConditionalStack.SkipInfo->FoundElse,
660 PreambleConditionalStack.SkipInfo->ElseLoc);
664 void Preprocessor::EndSourceFile() {
665 // Notify the client that we reached the end of the source file.
667 Callbacks->EndOfMainFile();
670 //===----------------------------------------------------------------------===//
671 // Lexer Event Handling.
672 //===----------------------------------------------------------------------===//
674 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
675 /// identifier information for the token and install it into the token,
676 /// updating the token kind accordingly.
677 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
678 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
680 // Look up this token, see if it is a macro, or if it is a language keyword.
682 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
683 // No cleaning needed, just use the characters from the lexed buffer.
684 II = getIdentifierInfo(Identifier.getRawIdentifier());
686 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
687 SmallString<64> IdentifierBuffer;
688 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
690 if (Identifier.hasUCN()) {
691 SmallString<64> UCNIdentifierBuffer;
692 expandUCNs(UCNIdentifierBuffer, CleanedStr);
693 II = getIdentifierInfo(UCNIdentifierBuffer);
695 II = getIdentifierInfo(CleanedStr);
699 // Update the token info (identifier info and appropriate token kind).
700 Identifier.setIdentifierInfo(II);
701 if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
702 getSourceManager().isInSystemHeader(Identifier.getLocation()))
703 Identifier.setKind(tok::identifier);
705 Identifier.setKind(II->getTokenID());
710 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
711 PoisonReasons[II] = DiagID;
714 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
715 assert(Ident__exception_code && Ident__exception_info);
716 assert(Ident___exception_code && Ident___exception_info);
717 Ident__exception_code->setIsPoisoned(Poison);
718 Ident___exception_code->setIsPoisoned(Poison);
719 Ident_GetExceptionCode->setIsPoisoned(Poison);
720 Ident__exception_info->setIsPoisoned(Poison);
721 Ident___exception_info->setIsPoisoned(Poison);
722 Ident_GetExceptionInfo->setIsPoisoned(Poison);
723 Ident__abnormal_termination->setIsPoisoned(Poison);
724 Ident___abnormal_termination->setIsPoisoned(Poison);
725 Ident_AbnormalTermination->setIsPoisoned(Poison);
728 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
729 assert(Identifier.getIdentifierInfo() &&
730 "Can't handle identifiers without identifier info!");
731 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
732 PoisonReasons.find(Identifier.getIdentifierInfo());
733 if(it == PoisonReasons.end())
734 Diag(Identifier, diag::err_pp_used_poisoned_id);
736 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
739 /// Returns a diagnostic message kind for reporting a future keyword as
740 /// appropriate for the identifier and specified language.
741 static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
742 const LangOptions &LangOpts) {
743 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
745 if (LangOpts.CPlusPlus)
746 return llvm::StringSwitch<diag::kind>(II.getName())
747 #define CXX11_KEYWORD(NAME, FLAGS) \
748 .Case(#NAME, diag::warn_cxx11_keyword)
749 #define CXX2A_KEYWORD(NAME, FLAGS) \
750 .Case(#NAME, diag::warn_cxx2a_keyword)
751 #include "clang/Basic/TokenKinds.def"
755 "Keyword not known to come from a newer Standard or proposed Standard");
758 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
759 assert(II.isOutOfDate() && "not out of date");
760 getExternalSource()->updateOutOfDateIdentifier(II);
763 /// HandleIdentifier - This callback is invoked when the lexer reads an
764 /// identifier. This callback looks up the identifier in the map and/or
765 /// potentially macro expands it or turns it into a named token (like 'for').
767 /// Note that callers of this method are guarded by checking the
768 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
769 /// IdentifierInfo methods that compute these properties will need to change to
771 bool Preprocessor::HandleIdentifier(Token &Identifier) {
772 assert(Identifier.getIdentifierInfo() &&
773 "Can't handle identifiers without identifier info!");
775 IdentifierInfo &II = *Identifier.getIdentifierInfo();
777 // If the information about this identifier is out of date, update it from
778 // the external source.
779 // We have to treat __VA_ARGS__ in a special way, since it gets
780 // serialized with isPoisoned = true, but our preprocessor may have
781 // unpoisoned it if we're defining a C99 macro.
782 if (II.isOutOfDate()) {
783 bool CurrentIsPoisoned = false;
784 const bool IsSpecialVariadicMacro =
785 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
786 if (IsSpecialVariadicMacro)
787 CurrentIsPoisoned = II.isPoisoned();
789 updateOutOfDateIdentifier(II);
790 Identifier.setKind(II.getTokenID());
792 if (IsSpecialVariadicMacro)
793 II.setIsPoisoned(CurrentIsPoisoned);
796 // If this identifier was poisoned, and if it was not produced from a macro
797 // expansion, emit an error.
798 if (II.isPoisoned() && CurPPLexer) {
799 HandlePoisonedIdentifier(Identifier);
802 // If this is a macro to be expanded, do it.
803 if (MacroDefinition MD = getMacroDefinition(&II)) {
804 auto *MI = MD.getMacroInfo();
805 assert(MI && "macro definition with no macro info?");
806 if (!DisableMacroExpansion) {
807 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
808 // C99 6.10.3p10: If the preprocessing token immediately after the
809 // macro name isn't a '(', this macro should not be expanded.
810 if (!MI->isFunctionLike() || isNextPPTokenLParen())
811 return HandleMacroExpandedIdentifier(Identifier, MD);
813 // C99 6.10.3.4p2 says that a disabled macro may never again be
814 // expanded, even if it's in a context where it could be expanded in the
816 Identifier.setFlag(Token::DisableExpand);
817 if (MI->isObjectLike() || isNextPPTokenLParen())
818 Diag(Identifier, diag::pp_disabled_macro_expansion);
823 // If this identifier is a keyword in a newer Standard or proposed Standard,
824 // produce a warning. Don't warn if we're not considering macro expansion,
825 // since this identifier might be the name of a macro.
826 // FIXME: This warning is disabled in cases where it shouldn't be, like
827 // "#define constexpr constexpr", "int constexpr;"
828 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
829 Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
831 // Don't diagnose this keyword again in this translation unit.
832 II.setIsFutureCompatKeyword(false);
835 // If this is an extension token, diagnose its use.
836 // We avoid diagnosing tokens that originate from macro definitions.
837 // FIXME: This warning is disabled in cases where it shouldn't be,
838 // like "#define TY typeof", "TY(1) x".
839 if (II.isExtensionToken() && !DisableMacroExpansion)
840 Diag(Identifier, diag::ext_token_used);
842 // If this is the 'import' contextual keyword following an '@', note
843 // that the next token indicates a module name.
845 // Note that we do not treat 'import' as a contextual
846 // keyword when we're in a caching lexer, because caching lexers only get
847 // used in contexts where import declarations are disallowed.
849 // Likewise if this is the C++ Modules TS import keyword.
850 if (((LastTokenWasAt && II.isModulesImport()) ||
851 Identifier.is(tok::kw_import)) &&
852 !InMacroArgs && !DisableMacroExpansion &&
853 (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
854 CurLexerKind != CLK_CachingLexer) {
855 ModuleImportLoc = Identifier.getLocation();
856 ModuleImportPath.clear();
857 ModuleImportExpectsIdentifier = true;
858 CurLexerKind = CLK_LexAfterModuleImport;
863 void Preprocessor::Lex(Token &Result) {
864 // We loop here until a lex function returns a token; this avoids recursion.
867 switch (CurLexerKind) {
869 ReturnedToken = CurLexer->Lex(Result);
872 ReturnedToken = CurTokenLexer->Lex(Result);
874 case CLK_CachingLexer:
876 ReturnedToken = true;
878 case CLK_LexAfterModuleImport:
879 LexAfterModuleImport(Result);
880 ReturnedToken = true;
883 } while (!ReturnedToken);
885 if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
886 // Remember the identifier before code completion token.
887 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
888 setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
889 // Set IdenfitierInfo to null to avoid confusing code that handles both
890 // identifiers and completion tokens.
891 Result.setIdentifierInfo(nullptr);
894 LastTokenWasAt = Result.is(tok::at);
897 /// Lex a token following the 'import' contextual keyword.
899 void Preprocessor::LexAfterModuleImport(Token &Result) {
900 // Figure out what kind of lexer we actually have.
901 recomputeCurLexerKind();
903 // Lex the next token.
906 // The token sequence
908 // import identifier (. identifier)*
910 // indicates a module import directive. We already saw the 'import'
911 // contextual keyword, so now we're looking for the identifiers.
912 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
913 // We expected to see an identifier here, and we did; continue handling
915 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
916 Result.getLocation()));
917 ModuleImportExpectsIdentifier = false;
918 CurLexerKind = CLK_LexAfterModuleImport;
922 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
923 // see the next identifier. (We can also see a '[[' that begins an
924 // attribute-specifier-seq here under the C++ Modules TS.)
925 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
926 ModuleImportExpectsIdentifier = true;
927 CurLexerKind = CLK_LexAfterModuleImport;
931 // If we have a non-empty module path, load the named module.
932 if (!ModuleImportPath.empty()) {
933 // Under the Modules TS, the dot is just part of the module name, and not
934 // a real hierarchy separator. Flatten such module names now.
936 // FIXME: Is this the right level to be performing this transformation?
937 std::string FlatModuleName;
938 if (getLangOpts().ModulesTS) {
939 for (auto &Piece : ModuleImportPath) {
940 if (!FlatModuleName.empty())
941 FlatModuleName += ".";
942 FlatModuleName += Piece.first->getName();
944 SourceLocation FirstPathLoc = ModuleImportPath[0].second;
945 ModuleImportPath.clear();
946 ModuleImportPath.push_back(
947 std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
950 Module *Imported = nullptr;
951 if (getLangOpts().Modules) {
952 Imported = TheModuleLoader.loadModule(ModuleImportLoc,
955 /*IsIncludeDirective=*/false);
957 makeModuleVisible(Imported, ModuleImportLoc);
959 if (Callbacks && (getLangOpts().Modules || getLangOpts().DebuggerSupport))
960 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
964 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
965 CurSubmoduleState->VisibleModules.setVisible(
966 M, Loc, [](Module *) {},
967 [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
968 // FIXME: Include the path in the diagnostic.
969 // FIXME: Include the import location for the conflicting module.
970 Diag(ModuleImportLoc, diag::warn_module_conflict)
971 << Path[0]->getFullModuleName()
972 << Conflict->getFullModuleName()
976 // Add this module to the imports list of the currently-built submodule.
977 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
978 BuildingSubmoduleStack.back().M->Imports.insert(M);
981 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
982 const char *DiagnosticTag,
983 bool AllowMacroExpansion) {
984 // We need at least one string literal.
985 if (Result.isNot(tok::string_literal)) {
986 Diag(Result, diag::err_expected_string_literal)
987 << /*Source='in...'*/0 << DiagnosticTag;
991 // Lex string literal tokens, optionally with macro expansion.
992 SmallVector<Token, 4> StrToks;
994 StrToks.push_back(Result);
996 if (Result.hasUDSuffix())
997 Diag(Result, diag::err_invalid_string_udl);
999 if (AllowMacroExpansion)
1002 LexUnexpandedToken(Result);
1003 } while (Result.is(tok::string_literal));
1005 // Concatenate and parse the strings.
1006 StringLiteralParser Literal(StrToks, *this);
1007 assert(Literal.isAscii() && "Didn't allow wide strings in");
1009 if (Literal.hadError)
1012 if (Literal.Pascal) {
1013 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1014 << /*Source='in...'*/0 << DiagnosticTag;
1018 String = Literal.GetString();
1022 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1023 assert(Tok.is(tok::numeric_constant));
1024 SmallString<8> IntegerBuffer;
1025 bool NumberInvalid = false;
1026 StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1029 NumericLiteralParser Literal(Spelling, Tok.getLocation(), *this);
1030 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1032 llvm::APInt APVal(64, 0);
1033 if (Literal.GetIntegerValue(APVal))
1036 Value = APVal.getLimitedValue();
1040 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1041 assert(Handler && "NULL comment handler");
1042 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
1043 CommentHandlers.end() && "Comment handler already registered");
1044 CommentHandlers.push_back(Handler);
1047 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1048 std::vector<CommentHandler *>::iterator Pos =
1049 std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
1050 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1051 CommentHandlers.erase(Pos);
1054 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1055 bool AnyPendingTokens = false;
1056 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1057 HEnd = CommentHandlers.end();
1059 if ((*H)->HandleComment(*this, Comment))
1060 AnyPendingTokens = true;
1062 if (!AnyPendingTokens || getCommentRetentionState())
1068 ModuleLoader::~ModuleLoader() = default;
1070 CommentHandler::~CommentHandler() = default;
1072 CodeCompletionHandler::~CodeCompletionHandler() = default;
1074 void Preprocessor::createPreprocessingRecord() {
1078 Record = new PreprocessingRecord(getSourceManager());
1079 addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));