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),
89 PCMCache(PCMCache), ScratchBuf(new ScratchBuffer(SourceMgr)),
90 HeaderInfo(Headers), TheModuleLoader(TheModuleLoader),
91 ExternalSource(nullptr), Identifiers(opts, IILookup),
92 PragmaHandlers(new PragmaNamespace(StringRef())), TUKind(TUKind),
93 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 (this->PPOpts->GeneratePreamble)
151 PreambleConditionalStack.startRecording();
154 Preprocessor::~Preprocessor() {
155 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
157 IncludeMacroStack.clear();
159 // Destroy any macro definitions.
160 while (MacroInfoChain *I = MIChainHead) {
161 MIChainHead = I->Next;
162 I->~MacroInfoChain();
165 // Free any cached macro expanders.
166 // This populates MacroArgCache, so all TokenLexers need to be destroyed
167 // before the code below that frees up the MacroArgCache list.
168 std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
169 CurTokenLexer.reset();
171 // Free any cached MacroArgs.
172 for (MacroArgs *ArgList = MacroArgCache; ArgList;)
173 ArgList = ArgList->deallocate();
175 // Delete the header search info, if we own it.
176 if (OwnsHeaderSearch)
180 void Preprocessor::Initialize(const TargetInfo &Target,
181 const TargetInfo *AuxTarget) {
182 assert((!this->Target || this->Target == &Target) &&
183 "Invalid override of target information");
184 this->Target = &Target;
186 assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
187 "Invalid override of aux target information.");
188 this->AuxTarget = AuxTarget;
190 // Initialize information about built-ins.
191 BuiltinInfo.InitializeTarget(Target, AuxTarget);
192 HeaderInfo.setTarget(Target);
195 void Preprocessor::InitializeForModelFile() {
196 NumEnteredSourceFiles = 0;
199 PragmaHandlersBackup = std::move(PragmaHandlers);
200 PragmaHandlers = llvm::make_unique<PragmaNamespace>(StringRef());
201 RegisterBuiltinPragmas();
203 // Reset PredefinesFileID
204 PredefinesFileID = FileID();
207 void Preprocessor::FinalizeForModelFile() {
208 NumEnteredSourceFiles = 1;
210 PragmaHandlers = std::move(PragmaHandlersBackup);
213 void Preprocessor::setPTHManager(PTHManager* pm) {
215 FileMgr.addStatCache(PTH->createStatCache());
218 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
219 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
220 << getSpelling(Tok) << "'";
222 if (!DumpFlags) return;
224 llvm::errs() << "\t";
225 if (Tok.isAtStartOfLine())
226 llvm::errs() << " [StartOfLine]";
227 if (Tok.hasLeadingSpace())
228 llvm::errs() << " [LeadingSpace]";
229 if (Tok.isExpandDisabled())
230 llvm::errs() << " [ExpandDisabled]";
231 if (Tok.needsCleaning()) {
232 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
233 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
237 llvm::errs() << "\tLoc=<";
238 DumpLocation(Tok.getLocation());
242 void Preprocessor::DumpLocation(SourceLocation Loc) const {
246 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
247 llvm::errs() << "MACRO: ";
248 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
249 DumpToken(MI.getReplacementToken(i));
252 llvm::errs() << "\n";
255 void Preprocessor::PrintStats() {
256 llvm::errs() << "\n*** Preprocessor Stats:\n";
257 llvm::errs() << NumDirectives << " directives found:\n";
258 llvm::errs() << " " << NumDefined << " #define.\n";
259 llvm::errs() << " " << NumUndefined << " #undef.\n";
260 llvm::errs() << " #include/#include_next/#import:\n";
261 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
262 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
263 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
264 llvm::errs() << " " << NumElse << " #else/#elif.\n";
265 llvm::errs() << " " << NumEndif << " #endif.\n";
266 llvm::errs() << " " << NumPragma << " #pragma.\n";
267 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
269 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
270 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
271 << NumFastMacroExpanded << " on the fast path.\n";
272 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
273 << " token paste (##) operations performed, "
274 << NumFastTokenPaste << " on the fast path.\n";
276 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
278 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
279 llvm::errs() << "\n Macro Expanded Tokens: "
280 << llvm::capacity_in_bytes(MacroExpandedTokens);
281 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
282 // FIXME: List information for all submodules.
283 llvm::errs() << "\n Macros: "
284 << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
285 llvm::errs() << "\n #pragma push_macro Info: "
286 << llvm::capacity_in_bytes(PragmaPushMacroInfo);
287 llvm::errs() << "\n Poison Reasons: "
288 << llvm::capacity_in_bytes(PoisonReasons);
289 llvm::errs() << "\n Comment Handlers: "
290 << llvm::capacity_in_bytes(CommentHandlers) << "\n";
293 Preprocessor::macro_iterator
294 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
295 if (IncludeExternalMacros && ExternalSource &&
296 !ReadMacrosFromExternalSource) {
297 ReadMacrosFromExternalSource = true;
298 ExternalSource->ReadDefinedMacros();
301 // Make sure we cover all macros in visible modules.
302 for (const ModuleMacro &Macro : ModuleMacros)
303 CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
305 return CurSubmoduleState->Macros.begin();
308 size_t Preprocessor::getTotalMemory() const {
309 return BP.getTotalMemory()
310 + llvm::capacity_in_bytes(MacroExpandedTokens)
311 + Predefines.capacity() /* Predefines buffer. */
312 // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
314 + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
315 + llvm::capacity_in_bytes(PragmaPushMacroInfo)
316 + llvm::capacity_in_bytes(PoisonReasons)
317 + llvm::capacity_in_bytes(CommentHandlers);
320 Preprocessor::macro_iterator
321 Preprocessor::macro_end(bool IncludeExternalMacros) const {
322 if (IncludeExternalMacros && ExternalSource &&
323 !ReadMacrosFromExternalSource) {
324 ReadMacrosFromExternalSource = true;
325 ExternalSource->ReadDefinedMacros();
328 return CurSubmoduleState->Macros.end();
331 /// \brief Compares macro tokens with a specified token value sequence.
332 static bool MacroDefinitionEquals(const MacroInfo *MI,
333 ArrayRef<TokenValue> Tokens) {
334 return Tokens.size() == MI->getNumTokens() &&
335 std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
338 StringRef Preprocessor::getLastMacroWithSpelling(
340 ArrayRef<TokenValue> Tokens) const {
341 SourceLocation BestLocation;
342 StringRef BestSpelling;
343 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
345 const MacroDirective::DefInfo
346 Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
347 if (!Def || !Def.getMacroInfo())
349 if (!Def.getMacroInfo()->isObjectLike())
351 if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
353 SourceLocation Location = Def.getLocation();
354 // Choose the macro defined latest.
355 if (BestLocation.isInvalid() ||
356 (Location.isValid() &&
357 SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
358 BestLocation = Location;
359 BestSpelling = I->first->getName();
365 void Preprocessor::recomputeCurLexerKind() {
367 CurLexerKind = CLK_Lexer;
368 else if (CurPTHLexer)
369 CurLexerKind = CLK_PTHLexer;
370 else if (CurTokenLexer)
371 CurLexerKind = CLK_TokenLexer;
373 CurLexerKind = CLK_CachingLexer;
376 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
377 unsigned CompleteLine,
378 unsigned CompleteColumn) {
380 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
381 assert(!CodeCompletionFile && "Already set");
383 using llvm::MemoryBuffer;
385 // Load the actual file's contents.
386 bool Invalid = false;
387 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
391 // Find the byte position of the truncation point.
392 const char *Position = Buffer->getBufferStart();
393 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
394 for (; *Position; ++Position) {
395 if (*Position != '\r' && *Position != '\n')
398 // Eat \r\n or \n\r as a single line.
399 if ((Position[1] == '\r' || Position[1] == '\n') &&
400 Position[0] != Position[1])
407 Position += CompleteColumn - 1;
409 // If pointing inside the preamble, adjust the position at the beginning of
410 // the file after the preamble.
411 if (SkipMainFilePreamble.first &&
412 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
413 if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
414 Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
417 if (Position > Buffer->getBufferEnd())
418 Position = Buffer->getBufferEnd();
420 CodeCompletionFile = File;
421 CodeCompletionOffset = Position - Buffer->getBufferStart();
423 std::unique_ptr<MemoryBuffer> NewBuffer =
424 MemoryBuffer::getNewUninitMemBuffer(Buffer->getBufferSize() + 1,
425 Buffer->getBufferIdentifier());
426 char *NewBuf = const_cast<char*>(NewBuffer->getBufferStart());
427 char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
429 std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
430 SourceMgr.overrideFileContents(File, std::move(NewBuffer));
435 void Preprocessor::CodeCompleteNaturalLanguage() {
437 CodeComplete->CodeCompleteNaturalLanguage();
438 setCodeCompletionReached();
441 /// getSpelling - This method is used to get the spelling of a token into a
442 /// SmallVector. Note that the returned StringRef may not point to the
443 /// supplied buffer if a copy can be avoided.
444 StringRef Preprocessor::getSpelling(const Token &Tok,
445 SmallVectorImpl<char> &Buffer,
446 bool *Invalid) const {
447 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
448 if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
449 // Try the fast path.
450 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
451 return II->getName();
454 // Resize the buffer if we need to copy into it.
455 if (Tok.needsCleaning())
456 Buffer.resize(Tok.getLength());
458 const char *Ptr = Buffer.data();
459 unsigned Len = getSpelling(Tok, Ptr, Invalid);
460 return StringRef(Ptr, Len);
463 /// CreateString - Plop the specified string into a scratch buffer and return a
464 /// location for it. If specified, the source location provides a source
465 /// location for the token.
466 void Preprocessor::CreateString(StringRef Str, Token &Tok,
467 SourceLocation ExpansionLocStart,
468 SourceLocation ExpansionLocEnd) {
469 Tok.setLength(Str.size());
472 SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
474 if (ExpansionLocStart.isValid())
475 Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
476 ExpansionLocEnd, Str.size());
477 Tok.setLocation(Loc);
479 // If this is a raw identifier or a literal token, set the pointer data.
480 if (Tok.is(tok::raw_identifier))
481 Tok.setRawIdentifierData(DestPtr);
482 else if (Tok.isLiteral())
483 Tok.setLiteralData(DestPtr);
486 Module *Preprocessor::getCurrentModule() {
487 if (!getLangOpts().isCompilingModule())
490 return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
493 //===----------------------------------------------------------------------===//
494 // Preprocessor Initialization Methods
495 //===----------------------------------------------------------------------===//
497 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
498 /// which implicitly adds the builtin defines etc.
499 void Preprocessor::EnterMainSourceFile() {
500 // We do not allow the preprocessor to reenter the main file. Doing so will
501 // cause FileID's to accumulate information from both runs (e.g. #line
502 // information) and predefined macros aren't guaranteed to be set properly.
503 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
504 FileID MainFileID = SourceMgr.getMainFileID();
506 // If MainFileID is loaded it means we loaded an AST file, no need to enter
508 if (!SourceMgr.isLoadedFileID(MainFileID)) {
509 // Enter the main file source buffer.
510 EnterSourceFile(MainFileID, nullptr, SourceLocation());
512 // If we've been asked to skip bytes in the main file (e.g., as part of a
513 // precompiled preamble), do so now.
514 if (SkipMainFilePreamble.first > 0)
515 CurLexer->SetByteOffset(SkipMainFilePreamble.first,
516 SkipMainFilePreamble.second);
518 // Tell the header info that the main file was entered. If the file is later
519 // #imported, it won't be re-entered.
520 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
521 HeaderInfo.IncrementIncludeCount(FE);
524 // Preprocess Predefines to populate the initial preprocessor state.
525 std::unique_ptr<llvm::MemoryBuffer> SB =
526 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
527 assert(SB && "Cannot create predefined source buffer");
528 FileID FID = SourceMgr.createFileID(std::move(SB));
529 assert(FID.isValid() && "Could not create FileID for predefines?");
530 setPredefinesFileID(FID);
532 // Start parsing the predefines.
533 EnterSourceFile(FID, nullptr, SourceLocation());
536 void Preprocessor::replayPreambleConditionalStack() {
537 // Restore the conditional stack from the preamble, if there is one.
538 if (PreambleConditionalStack.isReplaying()) {
540 "CurPPLexer is null when calling replayPreambleConditionalStack.");
541 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
542 PreambleConditionalStack.doneReplaying();
543 if (PreambleConditionalStack.reachedEOFWhileSkipping())
544 SkipExcludedConditionalBlock(
545 PreambleConditionalStack.SkipInfo->HashTokenLoc,
546 PreambleConditionalStack.SkipInfo->IfTokenLoc,
547 PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
548 PreambleConditionalStack.SkipInfo->FoundElse,
549 PreambleConditionalStack.SkipInfo->ElseLoc);
553 void Preprocessor::EndSourceFile() {
554 // Notify the client that we reached the end of the source file.
556 Callbacks->EndOfMainFile();
559 //===----------------------------------------------------------------------===//
560 // Lexer Event Handling.
561 //===----------------------------------------------------------------------===//
563 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
564 /// identifier information for the token and install it into the token,
565 /// updating the token kind accordingly.
566 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
567 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
569 // Look up this token, see if it is a macro, or if it is a language keyword.
571 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
572 // No cleaning needed, just use the characters from the lexed buffer.
573 II = getIdentifierInfo(Identifier.getRawIdentifier());
575 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
576 SmallString<64> IdentifierBuffer;
577 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
579 if (Identifier.hasUCN()) {
580 SmallString<64> UCNIdentifierBuffer;
581 expandUCNs(UCNIdentifierBuffer, CleanedStr);
582 II = getIdentifierInfo(UCNIdentifierBuffer);
584 II = getIdentifierInfo(CleanedStr);
588 // Update the token info (identifier info and appropriate token kind).
589 Identifier.setIdentifierInfo(II);
590 if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
591 getSourceManager().isInSystemHeader(Identifier.getLocation()))
592 Identifier.setKind(tok::identifier);
594 Identifier.setKind(II->getTokenID());
599 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
600 PoisonReasons[II] = DiagID;
603 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
604 assert(Ident__exception_code && Ident__exception_info);
605 assert(Ident___exception_code && Ident___exception_info);
606 Ident__exception_code->setIsPoisoned(Poison);
607 Ident___exception_code->setIsPoisoned(Poison);
608 Ident_GetExceptionCode->setIsPoisoned(Poison);
609 Ident__exception_info->setIsPoisoned(Poison);
610 Ident___exception_info->setIsPoisoned(Poison);
611 Ident_GetExceptionInfo->setIsPoisoned(Poison);
612 Ident__abnormal_termination->setIsPoisoned(Poison);
613 Ident___abnormal_termination->setIsPoisoned(Poison);
614 Ident_AbnormalTermination->setIsPoisoned(Poison);
617 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
618 assert(Identifier.getIdentifierInfo() &&
619 "Can't handle identifiers without identifier info!");
620 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
621 PoisonReasons.find(Identifier.getIdentifierInfo());
622 if(it == PoisonReasons.end())
623 Diag(Identifier, diag::err_pp_used_poisoned_id);
625 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
628 /// \brief Returns a diagnostic message kind for reporting a future keyword as
629 /// appropriate for the identifier and specified language.
630 static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
631 const LangOptions &LangOpts) {
632 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
634 if (LangOpts.CPlusPlus)
635 return llvm::StringSwitch<diag::kind>(II.getName())
636 #define CXX11_KEYWORD(NAME, FLAGS) \
637 .Case(#NAME, diag::warn_cxx11_keyword)
638 #define CXX2A_KEYWORD(NAME, FLAGS) \
639 .Case(#NAME, diag::warn_cxx2a_keyword)
640 #include "clang/Basic/TokenKinds.def"
644 "Keyword not known to come from a newer Standard or proposed Standard");
647 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
648 assert(II.isOutOfDate() && "not out of date");
649 getExternalSource()->updateOutOfDateIdentifier(II);
652 /// HandleIdentifier - This callback is invoked when the lexer reads an
653 /// identifier. This callback looks up the identifier in the map and/or
654 /// potentially macro expands it or turns it into a named token (like 'for').
656 /// Note that callers of this method are guarded by checking the
657 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
658 /// IdentifierInfo methods that compute these properties will need to change to
660 bool Preprocessor::HandleIdentifier(Token &Identifier) {
661 assert(Identifier.getIdentifierInfo() &&
662 "Can't handle identifiers without identifier info!");
664 IdentifierInfo &II = *Identifier.getIdentifierInfo();
666 // If the information about this identifier is out of date, update it from
667 // the external source.
668 // We have to treat __VA_ARGS__ in a special way, since it gets
669 // serialized with isPoisoned = true, but our preprocessor may have
670 // unpoisoned it if we're defining a C99 macro.
671 if (II.isOutOfDate()) {
672 bool CurrentIsPoisoned = false;
673 const bool IsSpecialVariadicMacro =
674 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
675 if (IsSpecialVariadicMacro)
676 CurrentIsPoisoned = II.isPoisoned();
678 updateOutOfDateIdentifier(II);
679 Identifier.setKind(II.getTokenID());
681 if (IsSpecialVariadicMacro)
682 II.setIsPoisoned(CurrentIsPoisoned);
685 // If this identifier was poisoned, and if it was not produced from a macro
686 // expansion, emit an error.
687 if (II.isPoisoned() && CurPPLexer) {
688 HandlePoisonedIdentifier(Identifier);
691 // If this is a macro to be expanded, do it.
692 if (MacroDefinition MD = getMacroDefinition(&II)) {
693 auto *MI = MD.getMacroInfo();
694 assert(MI && "macro definition with no macro info?");
695 if (!DisableMacroExpansion) {
696 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
697 // C99 6.10.3p10: If the preprocessing token immediately after the
698 // macro name isn't a '(', this macro should not be expanded.
699 if (!MI->isFunctionLike() || isNextPPTokenLParen())
700 return HandleMacroExpandedIdentifier(Identifier, MD);
702 // C99 6.10.3.4p2 says that a disabled macro may never again be
703 // expanded, even if it's in a context where it could be expanded in the
705 Identifier.setFlag(Token::DisableExpand);
706 if (MI->isObjectLike() || isNextPPTokenLParen())
707 Diag(Identifier, diag::pp_disabled_macro_expansion);
712 // If this identifier is a keyword in a newer Standard or proposed Standard,
713 // produce a warning. Don't warn if we're not considering macro expansion,
714 // since this identifier might be the name of a macro.
715 // FIXME: This warning is disabled in cases where it shouldn't be, like
716 // "#define constexpr constexpr", "int constexpr;"
717 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
718 Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
720 // Don't diagnose this keyword again in this translation unit.
721 II.setIsFutureCompatKeyword(false);
724 // If this is an extension token, diagnose its use.
725 // We avoid diagnosing tokens that originate from macro definitions.
726 // FIXME: This warning is disabled in cases where it shouldn't be,
727 // like "#define TY typeof", "TY(1) x".
728 if (II.isExtensionToken() && !DisableMacroExpansion)
729 Diag(Identifier, diag::ext_token_used);
731 // If this is the 'import' contextual keyword following an '@', note
732 // that the next token indicates a module name.
734 // Note that we do not treat 'import' as a contextual
735 // keyword when we're in a caching lexer, because caching lexers only get
736 // used in contexts where import declarations are disallowed.
738 // Likewise if this is the C++ Modules TS import keyword.
739 if (((LastTokenWasAt && II.isModulesImport()) ||
740 Identifier.is(tok::kw_import)) &&
741 !InMacroArgs && !DisableMacroExpansion &&
742 (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
743 CurLexerKind != CLK_CachingLexer) {
744 ModuleImportLoc = Identifier.getLocation();
745 ModuleImportPath.clear();
746 ModuleImportExpectsIdentifier = true;
747 CurLexerKind = CLK_LexAfterModuleImport;
752 void Preprocessor::Lex(Token &Result) {
753 // We loop here until a lex function returns a token; this avoids recursion.
756 switch (CurLexerKind) {
758 ReturnedToken = CurLexer->Lex(Result);
761 ReturnedToken = CurPTHLexer->Lex(Result);
764 ReturnedToken = CurTokenLexer->Lex(Result);
766 case CLK_CachingLexer:
768 ReturnedToken = true;
770 case CLK_LexAfterModuleImport:
771 LexAfterModuleImport(Result);
772 ReturnedToken = true;
775 } while (!ReturnedToken);
777 if (Result.is(tok::code_completion))
778 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
780 LastTokenWasAt = Result.is(tok::at);
783 /// \brief Lex a token following the 'import' contextual keyword.
785 void Preprocessor::LexAfterModuleImport(Token &Result) {
786 // Figure out what kind of lexer we actually have.
787 recomputeCurLexerKind();
789 // Lex the next token.
792 // The token sequence
794 // import identifier (. identifier)*
796 // indicates a module import directive. We already saw the 'import'
797 // contextual keyword, so now we're looking for the identifiers.
798 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
799 // We expected to see an identifier here, and we did; continue handling
801 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
802 Result.getLocation()));
803 ModuleImportExpectsIdentifier = false;
804 CurLexerKind = CLK_LexAfterModuleImport;
808 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
809 // see the next identifier. (We can also see a '[[' that begins an
810 // attribute-specifier-seq here under the C++ Modules TS.)
811 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
812 ModuleImportExpectsIdentifier = true;
813 CurLexerKind = CLK_LexAfterModuleImport;
817 // If we have a non-empty module path, load the named module.
818 if (!ModuleImportPath.empty()) {
819 // Under the Modules TS, the dot is just part of the module name, and not
820 // a real hierarachy separator. Flatten such module names now.
822 // FIXME: Is this the right level to be performing this transformation?
823 std::string FlatModuleName;
824 if (getLangOpts().ModulesTS) {
825 for (auto &Piece : ModuleImportPath) {
826 if (!FlatModuleName.empty())
827 FlatModuleName += ".";
828 FlatModuleName += Piece.first->getName();
830 SourceLocation FirstPathLoc = ModuleImportPath[0].second;
831 ModuleImportPath.clear();
832 ModuleImportPath.push_back(
833 std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
836 Module *Imported = nullptr;
837 if (getLangOpts().Modules) {
838 Imported = TheModuleLoader.loadModule(ModuleImportLoc,
841 /*IsIncludeDirective=*/false);
843 makeModuleVisible(Imported, ModuleImportLoc);
845 if (Callbacks && (getLangOpts().Modules || getLangOpts().DebuggerSupport))
846 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
850 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
851 CurSubmoduleState->VisibleModules.setVisible(
852 M, Loc, [](Module *) {},
853 [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
854 // FIXME: Include the path in the diagnostic.
855 // FIXME: Include the import location for the conflicting module.
856 Diag(ModuleImportLoc, diag::warn_module_conflict)
857 << Path[0]->getFullModuleName()
858 << Conflict->getFullModuleName()
862 // Add this module to the imports list of the currently-built submodule.
863 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
864 BuildingSubmoduleStack.back().M->Imports.insert(M);
867 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
868 const char *DiagnosticTag,
869 bool AllowMacroExpansion) {
870 // We need at least one string literal.
871 if (Result.isNot(tok::string_literal)) {
872 Diag(Result, diag::err_expected_string_literal)
873 << /*Source='in...'*/0 << DiagnosticTag;
877 // Lex string literal tokens, optionally with macro expansion.
878 SmallVector<Token, 4> StrToks;
880 StrToks.push_back(Result);
882 if (Result.hasUDSuffix())
883 Diag(Result, diag::err_invalid_string_udl);
885 if (AllowMacroExpansion)
888 LexUnexpandedToken(Result);
889 } while (Result.is(tok::string_literal));
891 // Concatenate and parse the strings.
892 StringLiteralParser Literal(StrToks, *this);
893 assert(Literal.isAscii() && "Didn't allow wide strings in");
895 if (Literal.hadError)
898 if (Literal.Pascal) {
899 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
900 << /*Source='in...'*/0 << DiagnosticTag;
904 String = Literal.GetString();
908 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
909 assert(Tok.is(tok::numeric_constant));
910 SmallString<8> IntegerBuffer;
911 bool NumberInvalid = false;
912 StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
915 NumericLiteralParser Literal(Spelling, Tok.getLocation(), *this);
916 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
918 llvm::APInt APVal(64, 0);
919 if (Literal.GetIntegerValue(APVal))
922 Value = APVal.getLimitedValue();
926 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
927 assert(Handler && "NULL comment handler");
928 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
929 CommentHandlers.end() && "Comment handler already registered");
930 CommentHandlers.push_back(Handler);
933 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
934 std::vector<CommentHandler *>::iterator Pos =
935 std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
936 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
937 CommentHandlers.erase(Pos);
940 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
941 bool AnyPendingTokens = false;
942 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
943 HEnd = CommentHandlers.end();
945 if ((*H)->HandleComment(*this, Comment))
946 AnyPendingTokens = true;
948 if (!AnyPendingTokens || getCommentRetentionState())
954 ModuleLoader::~ModuleLoader() = default;
956 CommentHandler::~CommentHandler() = default;
958 CodeCompletionHandler::~CodeCompletionHandler() = default;
960 void Preprocessor::createPreprocessingRecord() {
964 Record = new PreprocessingRecord(getSourceManager());
965 addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));