1 //===- Preprocessor.cpp - C Language Family Preprocessor Implementation ---===//
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
7 //===----------------------------------------------------------------------===//
9 // This file implements the Preprocessor interface.
11 //===----------------------------------------------------------------------===//
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.
18 // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
23 // "Multiple include guards may be useful for:\n"
25 //===----------------------------------------------------------------------===//
27 #include "clang/Lex/Preprocessor.h"
28 #include "clang/Basic/FileManager.h"
29 #include "clang/Basic/FileSystemStatCache.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/SmallString.h"
57 #include "llvm/ADT/SmallVector.h"
58 #include "llvm/ADT/STLExtras.h"
59 #include "llvm/ADT/StringRef.h"
60 #include "llvm/ADT/StringSwitch.h"
61 #include "llvm/Support/Capacity.h"
62 #include "llvm/Support/ErrorHandling.h"
63 #include "llvm/Support/MemoryBuffer.h"
64 #include "llvm/Support/raw_ostream.h"
72 using namespace clang;
74 LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
76 ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
78 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
79 DiagnosticsEngine &diags, LangOptions &opts,
80 SourceManager &SM, HeaderSearch &Headers,
81 ModuleLoader &TheModuleLoader,
82 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
83 TranslationUnitKind TUKind)
84 : PPOpts(std::move(PPOpts)), Diags(&diags), LangOpts(opts),
85 FileMgr(Headers.getFileMgr()), SourceMgr(SM),
86 ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers),
87 TheModuleLoader(TheModuleLoader), ExternalSource(nullptr),
88 // As the language options may have not been loaded yet (when
89 // deserializing an ASTUnit), adding keywords to the identifier table is
90 // deferred to Preprocessor::Initialize().
91 Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())),
92 TUKind(TUKind), SkipMainFilePreamble(0, true),
93 CurSubmoduleState(&NullSubmoduleState) {
94 OwnsHeaderSearch = OwnsHeaders;
96 // Default to discarding comments.
98 KeepMacroComments = false;
99 SuppressIncludeNotFoundError = false;
101 // Macro expansion is enabled.
102 DisableMacroExpansion = false;
103 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,
571 /*IsFrameworkFound=*/nullptr);
573 Diag(SourceLocation(), diag::err_pp_through_header_not_found)
574 << PPOpts->PCHThroughHeader;
577 setPCHThroughHeaderFileID(
578 SourceMgr.createFileID(File, SourceLocation(), SrcMgr::C_User));
581 // Skip tokens from the Predefines and if needed the main file.
582 if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
583 (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
584 SkipTokensWhileUsingPCH();
587 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
588 assert(PCHThroughHeaderFileID.isInvalid() &&
589 "PCHThroughHeaderFileID already set!");
590 PCHThroughHeaderFileID = FID;
593 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
594 assert(PCHThroughHeaderFileID.isValid() &&
595 "Invalid PCH through header FileID");
596 return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
599 bool Preprocessor::creatingPCHWithThroughHeader() {
600 return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
601 PCHThroughHeaderFileID.isValid();
604 bool Preprocessor::usingPCHWithThroughHeader() {
605 return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
606 PCHThroughHeaderFileID.isValid();
609 bool Preprocessor::creatingPCHWithPragmaHdrStop() {
610 return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
613 bool Preprocessor::usingPCHWithPragmaHdrStop() {
614 return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
617 /// Skip tokens until after the #include of the through header or
618 /// until after a #pragma hdrstop is seen. Tokens in the predefines file
619 /// and the main file may be skipped. If the end of the predefines file
620 /// is reached, skipping continues into the main file. If the end of the
621 /// main file is reached, it's a fatal error.
622 void Preprocessor::SkipTokensWhileUsingPCH() {
623 bool ReachedMainFileEOF = false;
624 bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
625 bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
629 (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
630 switch (CurLexerKind) {
635 CurTokenLexer->Lex(Tok);
637 case CLK_CachingLexer:
640 case CLK_LexAfterModuleImport:
641 LexAfterModuleImport(Tok);
644 if (Tok.is(tok::eof) && !InPredefines) {
645 ReachedMainFileEOF = true;
648 if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
650 if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
653 if (ReachedMainFileEOF) {
654 if (UsingPCHThroughHeader)
655 Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
656 << PPOpts->PCHThroughHeader << 1;
657 else if (!PPOpts->PCHWithHdrStopCreate)
658 Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
662 void Preprocessor::replayPreambleConditionalStack() {
663 // Restore the conditional stack from the preamble, if there is one.
664 if (PreambleConditionalStack.isReplaying()) {
666 "CurPPLexer is null when calling replayPreambleConditionalStack.");
667 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
668 PreambleConditionalStack.doneReplaying();
669 if (PreambleConditionalStack.reachedEOFWhileSkipping())
670 SkipExcludedConditionalBlock(
671 PreambleConditionalStack.SkipInfo->HashTokenLoc,
672 PreambleConditionalStack.SkipInfo->IfTokenLoc,
673 PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
674 PreambleConditionalStack.SkipInfo->FoundElse,
675 PreambleConditionalStack.SkipInfo->ElseLoc);
679 void Preprocessor::EndSourceFile() {
680 // Notify the client that we reached the end of the source file.
682 Callbacks->EndOfMainFile();
685 //===----------------------------------------------------------------------===//
686 // Lexer Event Handling.
687 //===----------------------------------------------------------------------===//
689 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
690 /// identifier information for the token and install it into the token,
691 /// updating the token kind accordingly.
692 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
693 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
695 // Look up this token, see if it is a macro, or if it is a language keyword.
697 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
698 // No cleaning needed, just use the characters from the lexed buffer.
699 II = getIdentifierInfo(Identifier.getRawIdentifier());
701 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
702 SmallString<64> IdentifierBuffer;
703 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
705 if (Identifier.hasUCN()) {
706 SmallString<64> UCNIdentifierBuffer;
707 expandUCNs(UCNIdentifierBuffer, CleanedStr);
708 II = getIdentifierInfo(UCNIdentifierBuffer);
710 II = getIdentifierInfo(CleanedStr);
714 // Update the token info (identifier info and appropriate token kind).
715 Identifier.setIdentifierInfo(II);
716 if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
717 getSourceManager().isInSystemHeader(Identifier.getLocation()))
718 Identifier.setKind(tok::identifier);
720 Identifier.setKind(II->getTokenID());
725 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
726 PoisonReasons[II] = DiagID;
729 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
730 assert(Ident__exception_code && Ident__exception_info);
731 assert(Ident___exception_code && Ident___exception_info);
732 Ident__exception_code->setIsPoisoned(Poison);
733 Ident___exception_code->setIsPoisoned(Poison);
734 Ident_GetExceptionCode->setIsPoisoned(Poison);
735 Ident__exception_info->setIsPoisoned(Poison);
736 Ident___exception_info->setIsPoisoned(Poison);
737 Ident_GetExceptionInfo->setIsPoisoned(Poison);
738 Ident__abnormal_termination->setIsPoisoned(Poison);
739 Ident___abnormal_termination->setIsPoisoned(Poison);
740 Ident_AbnormalTermination->setIsPoisoned(Poison);
743 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
744 assert(Identifier.getIdentifierInfo() &&
745 "Can't handle identifiers without identifier info!");
746 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
747 PoisonReasons.find(Identifier.getIdentifierInfo());
748 if(it == PoisonReasons.end())
749 Diag(Identifier, diag::err_pp_used_poisoned_id);
751 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
754 /// Returns a diagnostic message kind for reporting a future keyword as
755 /// appropriate for the identifier and specified language.
756 static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
757 const LangOptions &LangOpts) {
758 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
760 if (LangOpts.CPlusPlus)
761 return llvm::StringSwitch<diag::kind>(II.getName())
762 #define CXX11_KEYWORD(NAME, FLAGS) \
763 .Case(#NAME, diag::warn_cxx11_keyword)
764 #define CXX2A_KEYWORD(NAME, FLAGS) \
765 .Case(#NAME, diag::warn_cxx2a_keyword)
766 #include "clang/Basic/TokenKinds.def"
770 "Keyword not known to come from a newer Standard or proposed Standard");
773 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
774 assert(II.isOutOfDate() && "not out of date");
775 getExternalSource()->updateOutOfDateIdentifier(II);
778 /// HandleIdentifier - This callback is invoked when the lexer reads an
779 /// identifier. This callback looks up the identifier in the map and/or
780 /// potentially macro expands it or turns it into a named token (like 'for').
782 /// Note that callers of this method are guarded by checking the
783 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
784 /// IdentifierInfo methods that compute these properties will need to change to
786 bool Preprocessor::HandleIdentifier(Token &Identifier) {
787 assert(Identifier.getIdentifierInfo() &&
788 "Can't handle identifiers without identifier info!");
790 IdentifierInfo &II = *Identifier.getIdentifierInfo();
792 // If the information about this identifier is out of date, update it from
793 // the external source.
794 // We have to treat __VA_ARGS__ in a special way, since it gets
795 // serialized with isPoisoned = true, but our preprocessor may have
796 // unpoisoned it if we're defining a C99 macro.
797 if (II.isOutOfDate()) {
798 bool CurrentIsPoisoned = false;
799 const bool IsSpecialVariadicMacro =
800 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
801 if (IsSpecialVariadicMacro)
802 CurrentIsPoisoned = II.isPoisoned();
804 updateOutOfDateIdentifier(II);
805 Identifier.setKind(II.getTokenID());
807 if (IsSpecialVariadicMacro)
808 II.setIsPoisoned(CurrentIsPoisoned);
811 // If this identifier was poisoned, and if it was not produced from a macro
812 // expansion, emit an error.
813 if (II.isPoisoned() && CurPPLexer) {
814 HandlePoisonedIdentifier(Identifier);
817 // If this is a macro to be expanded, do it.
818 if (MacroDefinition MD = getMacroDefinition(&II)) {
819 auto *MI = MD.getMacroInfo();
820 assert(MI && "macro definition with no macro info?");
821 if (!DisableMacroExpansion) {
822 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
823 // C99 6.10.3p10: If the preprocessing token immediately after the
824 // macro name isn't a '(', this macro should not be expanded.
825 if (!MI->isFunctionLike() || isNextPPTokenLParen())
826 return HandleMacroExpandedIdentifier(Identifier, MD);
828 // C99 6.10.3.4p2 says that a disabled macro may never again be
829 // expanded, even if it's in a context where it could be expanded in the
831 Identifier.setFlag(Token::DisableExpand);
832 if (MI->isObjectLike() || isNextPPTokenLParen())
833 Diag(Identifier, diag::pp_disabled_macro_expansion);
838 // If this identifier is a keyword in a newer Standard or proposed Standard,
839 // produce a warning. Don't warn if we're not considering macro expansion,
840 // since this identifier might be the name of a macro.
841 // FIXME: This warning is disabled in cases where it shouldn't be, like
842 // "#define constexpr constexpr", "int constexpr;"
843 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
844 Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
846 // Don't diagnose this keyword again in this translation unit.
847 II.setIsFutureCompatKeyword(false);
850 // If this is an extension token, diagnose its use.
851 // We avoid diagnosing tokens that originate from macro definitions.
852 // FIXME: This warning is disabled in cases where it shouldn't be,
853 // like "#define TY typeof", "TY(1) x".
854 if (II.isExtensionToken() && !DisableMacroExpansion)
855 Diag(Identifier, diag::ext_token_used);
857 // If this is the 'import' contextual keyword following an '@', note
858 // that the next token indicates a module name.
860 // Note that we do not treat 'import' as a contextual
861 // keyword when we're in a caching lexer, because caching lexers only get
862 // used in contexts where import declarations are disallowed.
864 // Likewise if this is the C++ Modules TS import keyword.
865 if (((LastTokenWasAt && II.isModulesImport()) ||
866 Identifier.is(tok::kw_import)) &&
867 !InMacroArgs && !DisableMacroExpansion &&
868 (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
869 CurLexerKind != CLK_CachingLexer) {
870 ModuleImportLoc = Identifier.getLocation();
871 ModuleImportPath.clear();
872 ModuleImportExpectsIdentifier = true;
873 CurLexerKind = CLK_LexAfterModuleImport;
878 void Preprocessor::Lex(Token &Result) {
881 // We loop here until a lex function returns a token; this avoids recursion.
884 switch (CurLexerKind) {
886 ReturnedToken = CurLexer->Lex(Result);
889 ReturnedToken = CurTokenLexer->Lex(Result);
891 case CLK_CachingLexer:
893 ReturnedToken = true;
895 case CLK_LexAfterModuleImport:
896 ReturnedToken = LexAfterModuleImport(Result);
899 } while (!ReturnedToken);
901 if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
902 // Remember the identifier before code completion token.
903 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
904 setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
905 // Set IdenfitierInfo to null to avoid confusing code that handles both
906 // identifiers and completion tokens.
907 Result.setIdentifierInfo(nullptr);
910 // Update ImportSeqState to track our position within a C++20 import-seq
911 // if this token is being produced as a result of phase 4 of translation.
912 if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
913 !Result.getFlag(Token::IsReinjected)) {
914 switch (Result.getKind()) {
915 case tok::l_paren: case tok::l_square: case tok::l_brace:
916 ImportSeqState.handleOpenBracket();
918 case tok::r_paren: case tok::r_square:
919 ImportSeqState.handleCloseBracket();
922 ImportSeqState.handleCloseBrace();
925 ImportSeqState.handleSemi();
927 case tok::header_name:
928 case tok::annot_header_unit:
929 ImportSeqState.handleHeaderName();
932 ImportSeqState.handleExport();
934 case tok::identifier:
935 if (Result.getIdentifierInfo()->isModulesImport()) {
936 ImportSeqState.handleImport();
937 if (ImportSeqState.afterImportSeq()) {
938 ModuleImportLoc = Result.getLocation();
939 ModuleImportPath.clear();
940 ModuleImportExpectsIdentifier = true;
941 CurLexerKind = CLK_LexAfterModuleImport;
947 ImportSeqState.handleMisc();
952 LastTokenWasAt = Result.is(tok::at);
954 if (OnToken && LexLevel == 0 && !Result.getFlag(Token::IsReinjected))
958 /// Lex a header-name token (including one formed from header-name-tokens if
959 /// \p AllowConcatenation is \c true).
961 /// \param FilenameTok Filled in with the next token. On success, this will
962 /// be either a header_name token. On failure, it will be whatever other
963 /// token was found instead.
964 /// \param AllowMacroExpansion If \c true, allow the header name to be formed
965 /// by macro expansion (concatenating tokens as necessary if the first
967 /// \return \c true if we reached EOD or EOF while looking for a > token in
968 /// a concatenated header name and diagnosed it. \c false otherwise.
969 bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
970 // Lex using header-name tokenization rules if tokens are being lexed from
971 // a file. Just grab a token normally if we're in a macro expansion.
973 CurPPLexer->LexIncludeFilename(FilenameTok);
977 // This could be a <foo/bar.h> file coming from a macro expansion. In this
978 // case, glue the tokens together into an angle_string_literal token.
979 SmallString<128> FilenameBuffer;
980 if (FilenameTok.is(tok::less) && AllowMacroExpansion) {
981 bool StartOfLine = FilenameTok.isAtStartOfLine();
982 bool LeadingSpace = FilenameTok.hasLeadingSpace();
983 bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
985 SourceLocation Start = FilenameTok.getLocation();
987 FilenameBuffer.push_back('<');
989 // Consume tokens until we find a '>'.
990 // FIXME: A header-name could be formed starting or ending with an
991 // alternative token. It's not clear whether that's ill-formed in all
993 while (FilenameTok.isNot(tok::greater)) {
995 if (FilenameTok.isOneOf(tok::eod, tok::eof)) {
996 Diag(FilenameTok.getLocation(), diag::err_expected) << tok::greater;
997 Diag(Start, diag::note_matching) << tok::less;
1001 End = FilenameTok.getLocation();
1003 // FIXME: Provide code completion for #includes.
1004 if (FilenameTok.is(tok::code_completion)) {
1005 setCodeCompletionReached();
1010 // Append the spelling of this token to the buffer. If there was a space
1011 // before it, add it now.
1012 if (FilenameTok.hasLeadingSpace())
1013 FilenameBuffer.push_back(' ');
1015 // Get the spelling of the token, directly into FilenameBuffer if
1017 size_t PreAppendSize = FilenameBuffer.size();
1018 FilenameBuffer.resize(PreAppendSize + FilenameTok.getLength());
1020 const char *BufPtr = &FilenameBuffer[PreAppendSize];
1021 unsigned ActualLen = getSpelling(FilenameTok, BufPtr);
1023 // If the token was spelled somewhere else, copy it into FilenameBuffer.
1024 if (BufPtr != &FilenameBuffer[PreAppendSize])
1025 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1027 // Resize FilenameBuffer to the correct size.
1028 if (FilenameTok.getLength() != ActualLen)
1029 FilenameBuffer.resize(PreAppendSize + ActualLen);
1032 FilenameTok.startToken();
1033 FilenameTok.setKind(tok::header_name);
1034 FilenameTok.setFlagValue(Token::StartOfLine, StartOfLine);
1035 FilenameTok.setFlagValue(Token::LeadingSpace, LeadingSpace);
1036 FilenameTok.setFlagValue(Token::LeadingEmptyMacro, LeadingEmptyMacro);
1037 CreateString(FilenameBuffer, FilenameTok, Start, End);
1038 } else if (FilenameTok.is(tok::string_literal) && AllowMacroExpansion) {
1039 // Convert a string-literal token of the form " h-char-sequence "
1040 // (produced by macro expansion) into a header-name token.
1042 // The rules for header-names don't quite match the rules for
1043 // string-literals, but all the places where they differ result in
1044 // undefined behavior, so we can and do treat them the same.
1046 // A string-literal with a prefix or suffix is not translated into a
1047 // header-name. This could theoretically be observable via the C++20
1048 // context-sensitive header-name formation rules.
1049 StringRef Str = getSpelling(FilenameTok, FilenameBuffer);
1050 if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1051 FilenameTok.setKind(tok::header_name);
1057 /// Collect the tokens of a C++20 pp-import-suffix.
1058 void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) {
1059 // FIXME: For error recovery, consider recognizing attribute syntax here
1060 // and terminating / diagnosing a missing semicolon if we find anything
1061 // else? (Can we leave that to the parser?)
1062 unsigned BracketDepth = 0;
1064 Toks.emplace_back();
1067 switch (Toks.back().getKind()) {
1068 case tok::l_paren: case tok::l_square: case tok::l_brace:
1072 case tok::r_paren: case tok::r_square: case tok::r_brace:
1073 if (BracketDepth == 0)
1079 if (BracketDepth == 0)
1093 /// Lex a token following the 'import' contextual keyword.
1095 /// pp-import: [C++20]
1096 /// import header-name pp-import-suffix[opt] ;
1097 /// import header-name-tokens pp-import-suffix[opt] ;
1098 /// [ObjC] @ import module-name ;
1099 /// [Clang] import module-name ;
1101 /// header-name-tokens:
1103 /// < [any sequence of preprocessing-tokens other than >] >
1106 /// module-name-qualifier[opt] identifier
1108 /// module-name-qualifier
1109 /// module-name-qualifier[opt] identifier .
1111 /// We respond to a pp-import by importing macros from the named module.
1112 bool Preprocessor::LexAfterModuleImport(Token &Result) {
1113 // Figure out what kind of lexer we actually have.
1114 recomputeCurLexerKind();
1116 // Lex the next token. The header-name lexing rules are used at the start of
1119 // For now, we only support header-name imports in C++20 mode.
1120 // FIXME: Should we allow this in all language modes that support an import
1121 // declaration as an extension?
1122 if (ModuleImportPath.empty() && getLangOpts().CPlusPlusModules) {
1123 if (LexHeaderName(Result))
1129 // Allocate a holding buffer for a sequence of tokens and introduce it into
1130 // the token stream.
1131 auto EnterTokens = [this](ArrayRef<Token> Toks) {
1132 auto ToksCopy = llvm::make_unique<Token[]>(Toks.size());
1133 std::copy(Toks.begin(), Toks.end(), ToksCopy.get());
1134 EnterTokenStream(std::move(ToksCopy), Toks.size(),
1135 /*DisableMacroExpansion*/ true, /*IsReinject*/ false);
1138 // Check for a header-name.
1139 SmallVector<Token, 32> Suffix;
1140 if (Result.is(tok::header_name)) {
1141 // Enter the header-name token into the token stream; a Lex action cannot
1142 // both return a token and cache tokens (doing so would corrupt the token
1143 // cache if the call to Lex comes from CachingLex / PeekAhead).
1144 Suffix.push_back(Result);
1146 // Consume the pp-import-suffix and expand any macros in it now. We'll add
1147 // it back into the token stream later.
1148 CollectPpImportSuffix(Suffix);
1149 if (Suffix.back().isNot(tok::semi)) {
1150 // This is not a pp-import after all.
1151 EnterTokens(Suffix);
1155 // C++2a [cpp.module]p1:
1156 // The ';' preprocessing-token terminating a pp-import shall not have
1157 // been produced by macro replacement.
1158 SourceLocation SemiLoc = Suffix.back().getLocation();
1159 if (SemiLoc.isMacroID())
1160 Diag(SemiLoc, diag::err_header_import_semi_in_macro);
1162 // Reconstitute the import token.
1164 ImportTok.startToken();
1165 ImportTok.setKind(tok::kw_import);
1166 ImportTok.setLocation(ModuleImportLoc);
1167 ImportTok.setIdentifierInfo(getIdentifierInfo("import"));
1168 ImportTok.setLength(6);
1170 auto Action = HandleHeaderIncludeOrImport(
1171 /*HashLoc*/ SourceLocation(), ImportTok, Suffix.front(), SemiLoc);
1172 switch (Action.Kind) {
1173 case ImportAction::None:
1176 case ImportAction::ModuleBegin:
1177 // Let the parser know we're textually entering the module.
1178 Suffix.emplace_back();
1179 Suffix.back().startToken();
1180 Suffix.back().setKind(tok::annot_module_begin);
1181 Suffix.back().setLocation(SemiLoc);
1182 Suffix.back().setAnnotationEndLoc(SemiLoc);
1183 Suffix.back().setAnnotationValue(Action.ModuleForHeader);
1186 case ImportAction::ModuleImport:
1187 case ImportAction::SkippedModuleImport:
1188 // We chose to import (or textually enter) the file. Convert the
1189 // header-name token into a header unit annotation token.
1190 Suffix[0].setKind(tok::annot_header_unit);
1191 Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation());
1192 Suffix[0].setAnnotationValue(Action.ModuleForHeader);
1193 // FIXME: Call the moduleImport callback?
1197 EnterTokens(Suffix);
1201 // The token sequence
1203 // import identifier (. identifier)*
1205 // indicates a module import directive. We already saw the 'import'
1206 // contextual keyword, so now we're looking for the identifiers.
1207 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
1208 // We expected to see an identifier here, and we did; continue handling
1210 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
1211 Result.getLocation()));
1212 ModuleImportExpectsIdentifier = false;
1213 CurLexerKind = CLK_LexAfterModuleImport;
1217 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
1218 // see the next identifier. (We can also see a '[[' that begins an
1219 // attribute-specifier-seq here under the C++ Modules TS.)
1220 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
1221 ModuleImportExpectsIdentifier = true;
1222 CurLexerKind = CLK_LexAfterModuleImport;
1226 // If we didn't recognize a module name at all, this is not a (valid) import.
1227 if (ModuleImportPath.empty() || Result.is(tok::eof))
1230 // Consume the pp-import-suffix and expand any macros in it now, if we're not
1231 // at the semicolon already.
1232 SourceLocation SemiLoc = Result.getLocation();
1233 if (Result.isNot(tok::semi)) {
1234 Suffix.push_back(Result);
1235 CollectPpImportSuffix(Suffix);
1236 if (Suffix.back().isNot(tok::semi)) {
1237 // This is not an import after all.
1238 EnterTokens(Suffix);
1241 SemiLoc = Suffix.back().getLocation();
1244 // Under the Modules TS, the dot is just part of the module name, and not
1245 // a real hierarchy separator. Flatten such module names now.
1247 // FIXME: Is this the right level to be performing this transformation?
1248 std::string FlatModuleName;
1249 if (getLangOpts().ModulesTS || getLangOpts().CPlusPlusModules) {
1250 for (auto &Piece : ModuleImportPath) {
1251 if (!FlatModuleName.empty())
1252 FlatModuleName += ".";
1253 FlatModuleName += Piece.first->getName();
1255 SourceLocation FirstPathLoc = ModuleImportPath[0].second;
1256 ModuleImportPath.clear();
1257 ModuleImportPath.push_back(
1258 std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
1261 Module *Imported = nullptr;
1262 if (getLangOpts().Modules) {
1263 Imported = TheModuleLoader.loadModule(ModuleImportLoc,
1266 /*IsInclusionDirective=*/false);
1268 makeModuleVisible(Imported, SemiLoc);
1271 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
1273 if (!Suffix.empty()) {
1274 EnterTokens(Suffix);
1280 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
1281 CurSubmoduleState->VisibleModules.setVisible(
1282 M, Loc, [](Module *) {},
1283 [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1284 // FIXME: Include the path in the diagnostic.
1285 // FIXME: Include the import location for the conflicting module.
1286 Diag(ModuleImportLoc, diag::warn_module_conflict)
1287 << Path[0]->getFullModuleName()
1288 << Conflict->getFullModuleName()
1292 // Add this module to the imports list of the currently-built submodule.
1293 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1294 BuildingSubmoduleStack.back().M->Imports.insert(M);
1297 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1298 const char *DiagnosticTag,
1299 bool AllowMacroExpansion) {
1300 // We need at least one string literal.
1301 if (Result.isNot(tok::string_literal)) {
1302 Diag(Result, diag::err_expected_string_literal)
1303 << /*Source='in...'*/0 << DiagnosticTag;
1307 // Lex string literal tokens, optionally with macro expansion.
1308 SmallVector<Token, 4> StrToks;
1310 StrToks.push_back(Result);
1312 if (Result.hasUDSuffix())
1313 Diag(Result, diag::err_invalid_string_udl);
1315 if (AllowMacroExpansion)
1318 LexUnexpandedToken(Result);
1319 } while (Result.is(tok::string_literal));
1321 // Concatenate and parse the strings.
1322 StringLiteralParser Literal(StrToks, *this);
1323 assert(Literal.isAscii() && "Didn't allow wide strings in");
1325 if (Literal.hadError)
1328 if (Literal.Pascal) {
1329 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1330 << /*Source='in...'*/0 << DiagnosticTag;
1334 String = Literal.GetString();
1338 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1339 assert(Tok.is(tok::numeric_constant));
1340 SmallString<8> IntegerBuffer;
1341 bool NumberInvalid = false;
1342 StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1345 NumericLiteralParser Literal(Spelling, Tok.getLocation(), *this);
1346 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1348 llvm::APInt APVal(64, 0);
1349 if (Literal.GetIntegerValue(APVal))
1352 Value = APVal.getLimitedValue();
1356 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1357 assert(Handler && "NULL comment handler");
1358 assert(llvm::find(CommentHandlers, Handler) == CommentHandlers.end() &&
1359 "Comment handler already registered");
1360 CommentHandlers.push_back(Handler);
1363 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1364 std::vector<CommentHandler *>::iterator Pos =
1365 llvm::find(CommentHandlers, Handler);
1366 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1367 CommentHandlers.erase(Pos);
1370 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1371 bool AnyPendingTokens = false;
1372 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1373 HEnd = CommentHandlers.end();
1375 if ((*H)->HandleComment(*this, Comment))
1376 AnyPendingTokens = true;
1378 if (!AnyPendingTokens || getCommentRetentionState())
1384 ModuleLoader::~ModuleLoader() = default;
1386 CommentHandler::~CommentHandler() = default;
1388 CodeCompletionHandler::~CodeCompletionHandler() = default;
1390 void Preprocessor::createPreprocessingRecord() {
1394 Record = new PreprocessingRecord(getSourceManager());
1395 addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));