xref: /freebsd/contrib/llvm-project/clang/lib/Lex/Preprocessor.cpp (revision a90b9d0159070121c221b966469c3e36d912bf82)
1 //===- Preprocessor.cpp - C Language Family Preprocessor Implementation ---===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 //  This file implements the Preprocessor interface.
10 //
11 //===----------------------------------------------------------------------===//
12 //
13 // Options to support:
14 //   -H       - Print the name of each header file used.
15 //   -d[DNI] - Dump various things.
16 //   -fworking-directory - #line's with preprocessor's working dir.
17 //   -fpreprocessed
18 //   -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
19 //   -W*
20 //   -w
21 //
22 // Messages to emit:
23 //   "Multiple include guards may be useful for:\n"
24 //
25 //===----------------------------------------------------------------------===//
26 
27 #include "clang/Lex/Preprocessor.h"
28 #include "clang/Basic/Builtins.h"
29 #include "clang/Basic/FileManager.h"
30 #include "clang/Basic/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/STLExtras.h"
58 #include "llvm/ADT/SmallString.h"
59 #include "llvm/ADT/SmallVector.h"
60 #include "llvm/ADT/StringRef.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"
65 #include <algorithm>
66 #include <cassert>
67 #include <memory>
68 #include <optional>
69 #include <string>
70 #include <utility>
71 #include <vector>
72 
73 using namespace clang;
74 
75 LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
76 
77 ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
78 
79 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
80                            DiagnosticsEngine &diags, const LangOptions &opts,
81                            SourceManager &SM, HeaderSearch &Headers,
82                            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),
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;
96 
97   // Default to discarding comments.
98   KeepComments = false;
99   KeepMacroComments = false;
100   SuppressIncludeNotFoundError = false;
101 
102   // Macro expansion is enabled.
103   DisableMacroExpansion = false;
104   MacroExpansionInDirectivesOverride = false;
105   InMacroArgs = false;
106   ArgMacro = nullptr;
107   InMacroArgPreExpansion = false;
108   NumCachedTokenLexers = 0;
109   PragmasEnabled = true;
110   ParsingIfOrElifDirective = false;
111   PreprocessedOutput = false;
112 
113   // We haven't read anything from the external source.
114   ReadMacrosFromExternalSource = false;
115 
116   BuiltinInfo = std::make_unique<Builtin::Context>();
117 
118   // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of
119   // a macro. They get unpoisoned where it is allowed.
120   (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
121   SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
122   (Ident__VA_OPT__ = getIdentifierInfo("__VA_OPT__"))->setIsPoisoned();
123   SetPoisonReason(Ident__VA_OPT__,diag::ext_pp_bad_vaopt_use);
124 
125   // Initialize the pragma handlers.
126   RegisterBuiltinPragmas();
127 
128   // Initialize builtin macros like __LINE__ and friends.
129   RegisterBuiltinMacros();
130 
131   if(LangOpts.Borland) {
132     Ident__exception_info        = getIdentifierInfo("_exception_info");
133     Ident___exception_info       = getIdentifierInfo("__exception_info");
134     Ident_GetExceptionInfo       = getIdentifierInfo("GetExceptionInformation");
135     Ident__exception_code        = getIdentifierInfo("_exception_code");
136     Ident___exception_code       = getIdentifierInfo("__exception_code");
137     Ident_GetExceptionCode       = getIdentifierInfo("GetExceptionCode");
138     Ident__abnormal_termination  = getIdentifierInfo("_abnormal_termination");
139     Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
140     Ident_AbnormalTermination    = getIdentifierInfo("AbnormalTermination");
141   } else {
142     Ident__exception_info = Ident__exception_code = nullptr;
143     Ident__abnormal_termination = Ident___exception_info = nullptr;
144     Ident___exception_code = Ident___abnormal_termination = nullptr;
145     Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr;
146     Ident_AbnormalTermination = nullptr;
147   }
148 
149   // Default incremental processing to -fincremental-extensions, clients can
150   // override with `enableIncrementalProcessing` if desired.
151   IncrementalProcessing = LangOpts.IncrementalExtensions;
152 
153   // If using a PCH where a #pragma hdrstop is expected, start skipping tokens.
154   if (usingPCHWithPragmaHdrStop())
155     SkippingUntilPragmaHdrStop = true;
156 
157   // If using a PCH with a through header, start skipping tokens.
158   if (!this->PPOpts->PCHThroughHeader.empty() &&
159       !this->PPOpts->ImplicitPCHInclude.empty())
160     SkippingUntilPCHThroughHeader = true;
161 
162   if (this->PPOpts->GeneratePreamble)
163     PreambleConditionalStack.startRecording();
164 
165   MaxTokens = LangOpts.MaxTokens;
166 }
167 
168 Preprocessor::~Preprocessor() {
169   assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
170 
171   IncludeMacroStack.clear();
172 
173   // Free any cached macro expanders.
174   // This populates MacroArgCache, so all TokenLexers need to be destroyed
175   // before the code below that frees up the MacroArgCache list.
176   std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
177   CurTokenLexer.reset();
178 
179   // Free any cached MacroArgs.
180   for (MacroArgs *ArgList = MacroArgCache; ArgList;)
181     ArgList = ArgList->deallocate();
182 
183   // Delete the header search info, if we own it.
184   if (OwnsHeaderSearch)
185     delete &HeaderInfo;
186 }
187 
188 void Preprocessor::Initialize(const TargetInfo &Target,
189                               const TargetInfo *AuxTarget) {
190   assert((!this->Target || this->Target == &Target) &&
191          "Invalid override of target information");
192   this->Target = &Target;
193 
194   assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
195          "Invalid override of aux target information.");
196   this->AuxTarget = AuxTarget;
197 
198   // Initialize information about built-ins.
199   BuiltinInfo->InitializeTarget(Target, AuxTarget);
200   HeaderInfo.setTarget(Target);
201 
202   // Populate the identifier table with info about keywords for the current language.
203   Identifiers.AddKeywords(LangOpts);
204 
205   // Initialize the __FTL_EVAL_METHOD__ macro to the TargetInfo.
206   setTUFPEvalMethod(getTargetInfo().getFPEvalMethod());
207 
208   if (getLangOpts().getFPEvalMethod() == LangOptions::FEM_UnsetOnCommandLine)
209     // Use setting from TargetInfo.
210     setCurrentFPEvalMethod(SourceLocation(), Target.getFPEvalMethod());
211   else
212     // Set initial value of __FLT_EVAL_METHOD__ from the command line.
213     setCurrentFPEvalMethod(SourceLocation(), getLangOpts().getFPEvalMethod());
214 }
215 
216 void Preprocessor::InitializeForModelFile() {
217   NumEnteredSourceFiles = 0;
218 
219   // Reset pragmas
220   PragmaHandlersBackup = std::move(PragmaHandlers);
221   PragmaHandlers = std::make_unique<PragmaNamespace>(StringRef());
222   RegisterBuiltinPragmas();
223 
224   // Reset PredefinesFileID
225   PredefinesFileID = FileID();
226 }
227 
228 void Preprocessor::FinalizeForModelFile() {
229   NumEnteredSourceFiles = 1;
230 
231   PragmaHandlers = std::move(PragmaHandlersBackup);
232 }
233 
234 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
235   llvm::errs() << tok::getTokenName(Tok.getKind());
236 
237   if (!Tok.isAnnotation())
238     llvm::errs() << " '" << getSpelling(Tok) << "'";
239 
240   if (!DumpFlags) return;
241 
242   llvm::errs() << "\t";
243   if (Tok.isAtStartOfLine())
244     llvm::errs() << " [StartOfLine]";
245   if (Tok.hasLeadingSpace())
246     llvm::errs() << " [LeadingSpace]";
247   if (Tok.isExpandDisabled())
248     llvm::errs() << " [ExpandDisabled]";
249   if (Tok.needsCleaning()) {
250     const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
251     llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
252                  << "']";
253   }
254 
255   llvm::errs() << "\tLoc=<";
256   DumpLocation(Tok.getLocation());
257   llvm::errs() << ">";
258 }
259 
260 void Preprocessor::DumpLocation(SourceLocation Loc) const {
261   Loc.print(llvm::errs(), SourceMgr);
262 }
263 
264 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
265   llvm::errs() << "MACRO: ";
266   for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
267     DumpToken(MI.getReplacementToken(i));
268     llvm::errs() << "  ";
269   }
270   llvm::errs() << "\n";
271 }
272 
273 void Preprocessor::PrintStats() {
274   llvm::errs() << "\n*** Preprocessor Stats:\n";
275   llvm::errs() << NumDirectives << " directives found:\n";
276   llvm::errs() << "  " << NumDefined << " #define.\n";
277   llvm::errs() << "  " << NumUndefined << " #undef.\n";
278   llvm::errs() << "  #include/#include_next/#import:\n";
279   llvm::errs() << "    " << NumEnteredSourceFiles << " source files entered.\n";
280   llvm::errs() << "    " << MaxIncludeStackDepth << " max include stack depth\n";
281   llvm::errs() << "  " << NumIf << " #if/#ifndef/#ifdef.\n";
282   llvm::errs() << "  " << NumElse << " #else/#elif/#elifdef/#elifndef.\n";
283   llvm::errs() << "  " << NumEndif << " #endif.\n";
284   llvm::errs() << "  " << NumPragma << " #pragma.\n";
285   llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
286 
287   llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
288              << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
289              << NumFastMacroExpanded << " on the fast path.\n";
290   llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
291              << " token paste (##) operations performed, "
292              << NumFastTokenPaste << " on the fast path.\n";
293 
294   llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
295 
296   llvm::errs() << "\n  BumpPtr: " << BP.getTotalMemory();
297   llvm::errs() << "\n  Macro Expanded Tokens: "
298                << llvm::capacity_in_bytes(MacroExpandedTokens);
299   llvm::errs() << "\n  Predefines Buffer: " << Predefines.capacity();
300   // FIXME: List information for all submodules.
301   llvm::errs() << "\n  Macros: "
302                << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
303   llvm::errs() << "\n  #pragma push_macro Info: "
304                << llvm::capacity_in_bytes(PragmaPushMacroInfo);
305   llvm::errs() << "\n  Poison Reasons: "
306                << llvm::capacity_in_bytes(PoisonReasons);
307   llvm::errs() << "\n  Comment Handlers: "
308                << llvm::capacity_in_bytes(CommentHandlers) << "\n";
309 }
310 
311 Preprocessor::macro_iterator
312 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
313   if (IncludeExternalMacros && ExternalSource &&
314       !ReadMacrosFromExternalSource) {
315     ReadMacrosFromExternalSource = true;
316     ExternalSource->ReadDefinedMacros();
317   }
318 
319   // Make sure we cover all macros in visible modules.
320   for (const ModuleMacro &Macro : ModuleMacros)
321     CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
322 
323   return CurSubmoduleState->Macros.begin();
324 }
325 
326 size_t Preprocessor::getTotalMemory() const {
327   return BP.getTotalMemory()
328     + llvm::capacity_in_bytes(MacroExpandedTokens)
329     + Predefines.capacity() /* Predefines buffer. */
330     // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
331     // and ModuleMacros.
332     + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
333     + llvm::capacity_in_bytes(PragmaPushMacroInfo)
334     + llvm::capacity_in_bytes(PoisonReasons)
335     + llvm::capacity_in_bytes(CommentHandlers);
336 }
337 
338 Preprocessor::macro_iterator
339 Preprocessor::macro_end(bool IncludeExternalMacros) const {
340   if (IncludeExternalMacros && ExternalSource &&
341       !ReadMacrosFromExternalSource) {
342     ReadMacrosFromExternalSource = true;
343     ExternalSource->ReadDefinedMacros();
344   }
345 
346   return CurSubmoduleState->Macros.end();
347 }
348 
349 /// Compares macro tokens with a specified token value sequence.
350 static bool MacroDefinitionEquals(const MacroInfo *MI,
351                                   ArrayRef<TokenValue> Tokens) {
352   return Tokens.size() == MI->getNumTokens() &&
353       std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
354 }
355 
356 StringRef Preprocessor::getLastMacroWithSpelling(
357                                     SourceLocation Loc,
358                                     ArrayRef<TokenValue> Tokens) const {
359   SourceLocation BestLocation;
360   StringRef BestSpelling;
361   for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
362        I != E; ++I) {
363     const MacroDirective::DefInfo
364       Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
365     if (!Def || !Def.getMacroInfo())
366       continue;
367     if (!Def.getMacroInfo()->isObjectLike())
368       continue;
369     if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
370       continue;
371     SourceLocation Location = Def.getLocation();
372     // Choose the macro defined latest.
373     if (BestLocation.isInvalid() ||
374         (Location.isValid() &&
375          SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
376       BestLocation = Location;
377       BestSpelling = I->first->getName();
378     }
379   }
380   return BestSpelling;
381 }
382 
383 void Preprocessor::recomputeCurLexerKind() {
384   if (CurLexer)
385     CurLexerCallback = CurLexer->isDependencyDirectivesLexer()
386                            ? CLK_DependencyDirectivesLexer
387                            : CLK_Lexer;
388   else if (CurTokenLexer)
389     CurLexerCallback = CLK_TokenLexer;
390   else
391     CurLexerCallback = CLK_CachingLexer;
392 }
393 
394 bool Preprocessor::SetCodeCompletionPoint(FileEntryRef File,
395                                           unsigned CompleteLine,
396                                           unsigned CompleteColumn) {
397   assert(CompleteLine && CompleteColumn && "Starts from 1:1");
398   assert(!CodeCompletionFile && "Already set");
399 
400   // Load the actual file's contents.
401   std::optional<llvm::MemoryBufferRef> Buffer =
402       SourceMgr.getMemoryBufferForFileOrNone(File);
403   if (!Buffer)
404     return true;
405 
406   // Find the byte position of the truncation point.
407   const char *Position = Buffer->getBufferStart();
408   for (unsigned Line = 1; Line < CompleteLine; ++Line) {
409     for (; *Position; ++Position) {
410       if (*Position != '\r' && *Position != '\n')
411         continue;
412 
413       // Eat \r\n or \n\r as a single line.
414       if ((Position[1] == '\r' || Position[1] == '\n') &&
415           Position[0] != Position[1])
416         ++Position;
417       ++Position;
418       break;
419     }
420   }
421 
422   Position += CompleteColumn - 1;
423 
424   // If pointing inside the preamble, adjust the position at the beginning of
425   // the file after the preamble.
426   if (SkipMainFilePreamble.first &&
427       SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
428     if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
429       Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
430   }
431 
432   if (Position > Buffer->getBufferEnd())
433     Position = Buffer->getBufferEnd();
434 
435   CodeCompletionFile = File;
436   CodeCompletionOffset = Position - Buffer->getBufferStart();
437 
438   auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
439       Buffer->getBufferSize() + 1, Buffer->getBufferIdentifier());
440   char *NewBuf = NewBuffer->getBufferStart();
441   char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
442   *NewPos = '\0';
443   std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
444   SourceMgr.overrideFileContents(File, std::move(NewBuffer));
445 
446   return false;
447 }
448 
449 void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
450                                             bool IsAngled) {
451   setCodeCompletionReached();
452   if (CodeComplete)
453     CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
454 }
455 
456 void Preprocessor::CodeCompleteNaturalLanguage() {
457   setCodeCompletionReached();
458   if (CodeComplete)
459     CodeComplete->CodeCompleteNaturalLanguage();
460 }
461 
462 /// getSpelling - This method is used to get the spelling of a token into a
463 /// SmallVector. Note that the returned StringRef may not point to the
464 /// supplied buffer if a copy can be avoided.
465 StringRef Preprocessor::getSpelling(const Token &Tok,
466                                           SmallVectorImpl<char> &Buffer,
467                                           bool *Invalid) const {
468   // NOTE: this has to be checked *before* testing for an IdentifierInfo.
469   if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
470     // Try the fast path.
471     if (const IdentifierInfo *II = Tok.getIdentifierInfo())
472       return II->getName();
473   }
474 
475   // Resize the buffer if we need to copy into it.
476   if (Tok.needsCleaning())
477     Buffer.resize(Tok.getLength());
478 
479   const char *Ptr = Buffer.data();
480   unsigned Len = getSpelling(Tok, Ptr, Invalid);
481   return StringRef(Ptr, Len);
482 }
483 
484 /// CreateString - Plop the specified string into a scratch buffer and return a
485 /// location for it.  If specified, the source location provides a source
486 /// location for the token.
487 void Preprocessor::CreateString(StringRef Str, Token &Tok,
488                                 SourceLocation ExpansionLocStart,
489                                 SourceLocation ExpansionLocEnd) {
490   Tok.setLength(Str.size());
491 
492   const char *DestPtr;
493   SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
494 
495   if (ExpansionLocStart.isValid())
496     Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
497                                        ExpansionLocEnd, Str.size());
498   Tok.setLocation(Loc);
499 
500   // If this is a raw identifier or a literal token, set the pointer data.
501   if (Tok.is(tok::raw_identifier))
502     Tok.setRawIdentifierData(DestPtr);
503   else if (Tok.isLiteral())
504     Tok.setLiteralData(DestPtr);
505 }
506 
507 SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
508   auto &SM = getSourceManager();
509   SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
510   std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(SpellingLoc);
511   bool Invalid = false;
512   StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
513   if (Invalid)
514     return SourceLocation();
515 
516   // FIXME: We could consider re-using spelling for tokens we see repeatedly.
517   const char *DestPtr;
518   SourceLocation Spelling =
519       ScratchBuf->getToken(Buffer.data() + LocInfo.second, Length, DestPtr);
520   return SM.createTokenSplitLoc(Spelling, Loc, Loc.getLocWithOffset(Length));
521 }
522 
523 Module *Preprocessor::getCurrentModule() {
524   if (!getLangOpts().isCompilingModule())
525     return nullptr;
526 
527   return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
528 }
529 
530 Module *Preprocessor::getCurrentModuleImplementation() {
531   if (!getLangOpts().isCompilingModuleImplementation())
532     return nullptr;
533 
534   return getHeaderSearchInfo().lookupModule(getLangOpts().ModuleName);
535 }
536 
537 //===----------------------------------------------------------------------===//
538 // Preprocessor Initialization Methods
539 //===----------------------------------------------------------------------===//
540 
541 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
542 /// which implicitly adds the builtin defines etc.
543 void Preprocessor::EnterMainSourceFile() {
544   // We do not allow the preprocessor to reenter the main file.  Doing so will
545   // cause FileID's to accumulate information from both runs (e.g. #line
546   // information) and predefined macros aren't guaranteed to be set properly.
547   assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
548   FileID MainFileID = SourceMgr.getMainFileID();
549 
550   // If MainFileID is loaded it means we loaded an AST file, no need to enter
551   // a main file.
552   if (!SourceMgr.isLoadedFileID(MainFileID)) {
553     // Enter the main file source buffer.
554     EnterSourceFile(MainFileID, nullptr, SourceLocation());
555 
556     // If we've been asked to skip bytes in the main file (e.g., as part of a
557     // precompiled preamble), do so now.
558     if (SkipMainFilePreamble.first > 0)
559       CurLexer->SetByteOffset(SkipMainFilePreamble.first,
560                               SkipMainFilePreamble.second);
561 
562     // Tell the header info that the main file was entered.  If the file is later
563     // #imported, it won't be re-entered.
564     if (OptionalFileEntryRef FE = SourceMgr.getFileEntryRefForID(MainFileID))
565       markIncluded(*FE);
566   }
567 
568   // Preprocess Predefines to populate the initial preprocessor state.
569   std::unique_ptr<llvm::MemoryBuffer> SB =
570     llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
571   assert(SB && "Cannot create predefined source buffer");
572   FileID FID = SourceMgr.createFileID(std::move(SB));
573   assert(FID.isValid() && "Could not create FileID for predefines?");
574   setPredefinesFileID(FID);
575 
576   // Start parsing the predefines.
577   EnterSourceFile(FID, nullptr, SourceLocation());
578 
579   if (!PPOpts->PCHThroughHeader.empty()) {
580     // Lookup and save the FileID for the through header. If it isn't found
581     // in the search path, it's a fatal error.
582     OptionalFileEntryRef File = LookupFile(
583         SourceLocation(), PPOpts->PCHThroughHeader,
584         /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr,
585         /*CurDir=*/nullptr, /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
586         /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr,
587         /*IsFrameworkFound=*/nullptr);
588     if (!File) {
589       Diag(SourceLocation(), diag::err_pp_through_header_not_found)
590           << PPOpts->PCHThroughHeader;
591       return;
592     }
593     setPCHThroughHeaderFileID(
594         SourceMgr.createFileID(*File, SourceLocation(), SrcMgr::C_User));
595   }
596 
597   // Skip tokens from the Predefines and if needed the main file.
598   if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
599       (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
600     SkipTokensWhileUsingPCH();
601 }
602 
603 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
604   assert(PCHThroughHeaderFileID.isInvalid() &&
605          "PCHThroughHeaderFileID already set!");
606   PCHThroughHeaderFileID = FID;
607 }
608 
609 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
610   assert(PCHThroughHeaderFileID.isValid() &&
611          "Invalid PCH through header FileID");
612   return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
613 }
614 
615 bool Preprocessor::creatingPCHWithThroughHeader() {
616   return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
617          PCHThroughHeaderFileID.isValid();
618 }
619 
620 bool Preprocessor::usingPCHWithThroughHeader() {
621   return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
622          PCHThroughHeaderFileID.isValid();
623 }
624 
625 bool Preprocessor::creatingPCHWithPragmaHdrStop() {
626   return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
627 }
628 
629 bool Preprocessor::usingPCHWithPragmaHdrStop() {
630   return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
631 }
632 
633 /// Skip tokens until after the #include of the through header or
634 /// until after a #pragma hdrstop is seen. Tokens in the predefines file
635 /// and the main file may be skipped. If the end of the predefines file
636 /// is reached, skipping continues into the main file. If the end of the
637 /// main file is reached, it's a fatal error.
638 void Preprocessor::SkipTokensWhileUsingPCH() {
639   bool ReachedMainFileEOF = false;
640   bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
641   bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
642   Token Tok;
643   while (true) {
644     bool InPredefines =
645         (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
646     CurLexerCallback(*this, Tok);
647     if (Tok.is(tok::eof) && !InPredefines) {
648       ReachedMainFileEOF = true;
649       break;
650     }
651     if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
652       break;
653     if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
654       break;
655   }
656   if (ReachedMainFileEOF) {
657     if (UsingPCHThroughHeader)
658       Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
659           << PPOpts->PCHThroughHeader << 1;
660     else if (!PPOpts->PCHWithHdrStopCreate)
661       Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
662   }
663 }
664 
665 void Preprocessor::replayPreambleConditionalStack() {
666   // Restore the conditional stack from the preamble, if there is one.
667   if (PreambleConditionalStack.isReplaying()) {
668     assert(CurPPLexer &&
669            "CurPPLexer is null when calling replayPreambleConditionalStack.");
670     CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
671     PreambleConditionalStack.doneReplaying();
672     if (PreambleConditionalStack.reachedEOFWhileSkipping())
673       SkipExcludedConditionalBlock(
674           PreambleConditionalStack.SkipInfo->HashTokenLoc,
675           PreambleConditionalStack.SkipInfo->IfTokenLoc,
676           PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
677           PreambleConditionalStack.SkipInfo->FoundElse,
678           PreambleConditionalStack.SkipInfo->ElseLoc);
679   }
680 }
681 
682 void Preprocessor::EndSourceFile() {
683   // Notify the client that we reached the end of the source file.
684   if (Callbacks)
685     Callbacks->EndOfMainFile();
686 }
687 
688 //===----------------------------------------------------------------------===//
689 // Lexer Event Handling.
690 //===----------------------------------------------------------------------===//
691 
692 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
693 /// identifier information for the token and install it into the token,
694 /// updating the token kind accordingly.
695 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
696   assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
697 
698   // Look up this token, see if it is a macro, or if it is a language keyword.
699   IdentifierInfo *II;
700   if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
701     // No cleaning needed, just use the characters from the lexed buffer.
702     II = getIdentifierInfo(Identifier.getRawIdentifier());
703   } else {
704     // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
705     SmallString<64> IdentifierBuffer;
706     StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
707 
708     if (Identifier.hasUCN()) {
709       SmallString<64> UCNIdentifierBuffer;
710       expandUCNs(UCNIdentifierBuffer, CleanedStr);
711       II = getIdentifierInfo(UCNIdentifierBuffer);
712     } else {
713       II = getIdentifierInfo(CleanedStr);
714     }
715   }
716 
717   // Update the token info (identifier info and appropriate token kind).
718   // FIXME: the raw_identifier may contain leading whitespace which is removed
719   // from the cleaned identifier token. The SourceLocation should be updated to
720   // refer to the non-whitespace character. For instance, the text "\\\nB" (a
721   // line continuation before 'B') is parsed as a single tok::raw_identifier and
722   // is cleaned to tok::identifier "B". After cleaning the token's length is
723   // still 3 and the SourceLocation refers to the location of the backslash.
724   Identifier.setIdentifierInfo(II);
725   Identifier.setKind(II->getTokenID());
726 
727   return II;
728 }
729 
730 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
731   PoisonReasons[II] = DiagID;
732 }
733 
734 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
735   assert(Ident__exception_code && Ident__exception_info);
736   assert(Ident___exception_code && Ident___exception_info);
737   Ident__exception_code->setIsPoisoned(Poison);
738   Ident___exception_code->setIsPoisoned(Poison);
739   Ident_GetExceptionCode->setIsPoisoned(Poison);
740   Ident__exception_info->setIsPoisoned(Poison);
741   Ident___exception_info->setIsPoisoned(Poison);
742   Ident_GetExceptionInfo->setIsPoisoned(Poison);
743   Ident__abnormal_termination->setIsPoisoned(Poison);
744   Ident___abnormal_termination->setIsPoisoned(Poison);
745   Ident_AbnormalTermination->setIsPoisoned(Poison);
746 }
747 
748 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
749   assert(Identifier.getIdentifierInfo() &&
750          "Can't handle identifiers without identifier info!");
751   llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
752     PoisonReasons.find(Identifier.getIdentifierInfo());
753   if(it == PoisonReasons.end())
754     Diag(Identifier, diag::err_pp_used_poisoned_id);
755   else
756     Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
757 }
758 
759 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
760   assert(II.isOutOfDate() && "not out of date");
761   getExternalSource()->updateOutOfDateIdentifier(II);
762 }
763 
764 /// HandleIdentifier - This callback is invoked when the lexer reads an
765 /// identifier.  This callback looks up the identifier in the map and/or
766 /// potentially macro expands it or turns it into a named token (like 'for').
767 ///
768 /// Note that callers of this method are guarded by checking the
769 /// IdentifierInfo's 'isHandleIdentifierCase' bit.  If this method changes, the
770 /// IdentifierInfo methods that compute these properties will need to change to
771 /// match.
772 bool Preprocessor::HandleIdentifier(Token &Identifier) {
773   assert(Identifier.getIdentifierInfo() &&
774          "Can't handle identifiers without identifier info!");
775 
776   IdentifierInfo &II = *Identifier.getIdentifierInfo();
777 
778   // If the information about this identifier is out of date, update it from
779   // the external source.
780   // We have to treat __VA_ARGS__ in a special way, since it gets
781   // serialized with isPoisoned = true, but our preprocessor may have
782   // unpoisoned it if we're defining a C99 macro.
783   if (II.isOutOfDate()) {
784     bool CurrentIsPoisoned = false;
785     const bool IsSpecialVariadicMacro =
786         &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
787     if (IsSpecialVariadicMacro)
788       CurrentIsPoisoned = II.isPoisoned();
789 
790     updateOutOfDateIdentifier(II);
791     Identifier.setKind(II.getTokenID());
792 
793     if (IsSpecialVariadicMacro)
794       II.setIsPoisoned(CurrentIsPoisoned);
795   }
796 
797   // If this identifier was poisoned, and if it was not produced from a macro
798   // expansion, emit an error.
799   if (II.isPoisoned() && CurPPLexer) {
800     HandlePoisonedIdentifier(Identifier);
801   }
802 
803   // If this is a macro to be expanded, do it.
804   if (const MacroDefinition MD = getMacroDefinition(&II)) {
805     const auto *MI = MD.getMacroInfo();
806     assert(MI && "macro definition with no macro info?");
807     if (!DisableMacroExpansion) {
808       if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
809         // C99 6.10.3p10: If the preprocessing token immediately after the
810         // macro name isn't a '(', this macro should not be expanded.
811         if (!MI->isFunctionLike() || isNextPPTokenLParen())
812           return HandleMacroExpandedIdentifier(Identifier, MD);
813       } else {
814         // C99 6.10.3.4p2 says that a disabled macro may never again be
815         // expanded, even if it's in a context where it could be expanded in the
816         // future.
817         Identifier.setFlag(Token::DisableExpand);
818         if (MI->isObjectLike() || isNextPPTokenLParen())
819           Diag(Identifier, diag::pp_disabled_macro_expansion);
820       }
821     }
822   }
823 
824   // If this identifier is a keyword in a newer Standard or proposed Standard,
825   // produce a warning. Don't warn if we're not considering macro expansion,
826   // since this identifier might be the name of a macro.
827   // FIXME: This warning is disabled in cases where it shouldn't be, like
828   //   "#define constexpr constexpr", "int constexpr;"
829   if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
830     Diag(Identifier, getIdentifierTable().getFutureCompatDiagKind(II, getLangOpts()))
831         << II.getName();
832     // Don't diagnose this keyword again in this translation unit.
833     II.setIsFutureCompatKeyword(false);
834   }
835 
836   // If this is an extension token, diagnose its use.
837   // We avoid diagnosing tokens that originate from macro definitions.
838   // FIXME: This warning is disabled in cases where it shouldn't be,
839   // like "#define TY typeof", "TY(1) x".
840   if (II.isExtensionToken() && !DisableMacroExpansion)
841     Diag(Identifier, diag::ext_token_used);
842 
843   // If this is the 'import' contextual keyword following an '@', note
844   // that the next token indicates a module name.
845   //
846   // Note that we do not treat 'import' as a contextual
847   // keyword when we're in a caching lexer, because caching lexers only get
848   // used in contexts where import declarations are disallowed.
849   //
850   // Likewise if this is the standard C++ import keyword.
851   if (((LastTokenWasAt && II.isModulesImport()) ||
852        Identifier.is(tok::kw_import)) &&
853       !InMacroArgs && !DisableMacroExpansion &&
854       (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
855       CurLexerCallback != CLK_CachingLexer) {
856     ModuleImportLoc = Identifier.getLocation();
857     NamedModuleImportPath.clear();
858     IsAtImport = true;
859     ModuleImportExpectsIdentifier = true;
860     CurLexerCallback = CLK_LexAfterModuleImport;
861   }
862   return true;
863 }
864 
865 void Preprocessor::Lex(Token &Result) {
866   ++LexLevel;
867 
868   // We loop here until a lex function returns a token; this avoids recursion.
869   while (!CurLexerCallback(*this, Result))
870     ;
871 
872   if (Result.is(tok::unknown) && TheModuleLoader.HadFatalFailure)
873     return;
874 
875   if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
876     // Remember the identifier before code completion token.
877     setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
878     setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
879     // Set IdenfitierInfo to null to avoid confusing code that handles both
880     // identifiers and completion tokens.
881     Result.setIdentifierInfo(nullptr);
882   }
883 
884   // Update StdCXXImportSeqState to track our position within a C++20 import-seq
885   // if this token is being produced as a result of phase 4 of translation.
886   // Update TrackGMFState to decide if we are currently in a Global Module
887   // Fragment. GMF state updates should precede StdCXXImportSeq ones, since GMF state
888   // depends on the prevailing StdCXXImportSeq state in two cases.
889   if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
890       !Result.getFlag(Token::IsReinjected)) {
891     switch (Result.getKind()) {
892     case tok::l_paren: case tok::l_square: case tok::l_brace:
893       StdCXXImportSeqState.handleOpenBracket();
894       break;
895     case tok::r_paren: case tok::r_square:
896       StdCXXImportSeqState.handleCloseBracket();
897       break;
898     case tok::r_brace:
899       StdCXXImportSeqState.handleCloseBrace();
900       break;
901     // This token is injected to represent the translation of '#include "a.h"'
902     // into "import a.h;". Mimic the notional ';'.
903     case tok::annot_module_include:
904     case tok::semi:
905       TrackGMFState.handleSemi();
906       StdCXXImportSeqState.handleSemi();
907       ModuleDeclState.handleSemi();
908       break;
909     case tok::header_name:
910     case tok::annot_header_unit:
911       StdCXXImportSeqState.handleHeaderName();
912       break;
913     case tok::kw_export:
914       TrackGMFState.handleExport();
915       StdCXXImportSeqState.handleExport();
916       ModuleDeclState.handleExport();
917       break;
918     case tok::colon:
919       ModuleDeclState.handleColon();
920       break;
921     case tok::period:
922       ModuleDeclState.handlePeriod();
923       break;
924     case tok::identifier:
925       // Check "import" and "module" when there is no open bracket. The two
926       // identifiers are not meaningful with open brackets.
927       if (StdCXXImportSeqState.atTopLevel()) {
928         if (Result.getIdentifierInfo()->isModulesImport()) {
929           TrackGMFState.handleImport(StdCXXImportSeqState.afterTopLevelSeq());
930           StdCXXImportSeqState.handleImport();
931           if (StdCXXImportSeqState.afterImportSeq()) {
932             ModuleImportLoc = Result.getLocation();
933             NamedModuleImportPath.clear();
934             IsAtImport = false;
935             ModuleImportExpectsIdentifier = true;
936             CurLexerCallback = CLK_LexAfterModuleImport;
937           }
938           break;
939         } else if (Result.getIdentifierInfo() == getIdentifierInfo("module")) {
940           TrackGMFState.handleModule(StdCXXImportSeqState.afterTopLevelSeq());
941           ModuleDeclState.handleModule();
942           break;
943         }
944       }
945       ModuleDeclState.handleIdentifier(Result.getIdentifierInfo());
946       if (ModuleDeclState.isModuleCandidate())
947         break;
948       [[fallthrough]];
949     default:
950       TrackGMFState.handleMisc();
951       StdCXXImportSeqState.handleMisc();
952       ModuleDeclState.handleMisc();
953       break;
954     }
955   }
956 
957   LastTokenWasAt = Result.is(tok::at);
958   --LexLevel;
959 
960   if ((LexLevel == 0 || PreprocessToken) &&
961       !Result.getFlag(Token::IsReinjected)) {
962     if (LexLevel == 0)
963       ++TokenCount;
964     if (OnToken)
965       OnToken(Result);
966   }
967 }
968 
969 void Preprocessor::LexTokensUntilEOF(std::vector<Token> *Tokens) {
970   while (1) {
971     Token Tok;
972     Lex(Tok);
973     if (Tok.isOneOf(tok::unknown, tok::eof, tok::eod,
974                     tok::annot_repl_input_end))
975       break;
976     if (Tokens != nullptr)
977       Tokens->push_back(Tok);
978   }
979 }
980 
981 /// Lex a header-name token (including one formed from header-name-tokens if
982 /// \p AllowConcatenation is \c true).
983 ///
984 /// \param FilenameTok Filled in with the next token. On success, this will
985 ///        be either a header_name token. On failure, it will be whatever other
986 ///        token was found instead.
987 /// \param AllowMacroExpansion If \c true, allow the header name to be formed
988 ///        by macro expansion (concatenating tokens as necessary if the first
989 ///        token is a '<').
990 /// \return \c true if we reached EOD or EOF while looking for a > token in
991 ///         a concatenated header name and diagnosed it. \c false otherwise.
992 bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
993   // Lex using header-name tokenization rules if tokens are being lexed from
994   // a file. Just grab a token normally if we're in a macro expansion.
995   if (CurPPLexer)
996     CurPPLexer->LexIncludeFilename(FilenameTok);
997   else
998     Lex(FilenameTok);
999 
1000   // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1001   // case, glue the tokens together into an angle_string_literal token.
1002   SmallString<128> FilenameBuffer;
1003   if (FilenameTok.is(tok::less) && AllowMacroExpansion) {
1004     bool StartOfLine = FilenameTok.isAtStartOfLine();
1005     bool LeadingSpace = FilenameTok.hasLeadingSpace();
1006     bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
1007 
1008     SourceLocation Start = FilenameTok.getLocation();
1009     SourceLocation End;
1010     FilenameBuffer.push_back('<');
1011 
1012     // Consume tokens until we find a '>'.
1013     // FIXME: A header-name could be formed starting or ending with an
1014     // alternative token. It's not clear whether that's ill-formed in all
1015     // cases.
1016     while (FilenameTok.isNot(tok::greater)) {
1017       Lex(FilenameTok);
1018       if (FilenameTok.isOneOf(tok::eod, tok::eof)) {
1019         Diag(FilenameTok.getLocation(), diag::err_expected) << tok::greater;
1020         Diag(Start, diag::note_matching) << tok::less;
1021         return true;
1022       }
1023 
1024       End = FilenameTok.getLocation();
1025 
1026       // FIXME: Provide code completion for #includes.
1027       if (FilenameTok.is(tok::code_completion)) {
1028         setCodeCompletionReached();
1029         Lex(FilenameTok);
1030         continue;
1031       }
1032 
1033       // Append the spelling of this token to the buffer. If there was a space
1034       // before it, add it now.
1035       if (FilenameTok.hasLeadingSpace())
1036         FilenameBuffer.push_back(' ');
1037 
1038       // Get the spelling of the token, directly into FilenameBuffer if
1039       // possible.
1040       size_t PreAppendSize = FilenameBuffer.size();
1041       FilenameBuffer.resize(PreAppendSize + FilenameTok.getLength());
1042 
1043       const char *BufPtr = &FilenameBuffer[PreAppendSize];
1044       unsigned ActualLen = getSpelling(FilenameTok, BufPtr);
1045 
1046       // If the token was spelled somewhere else, copy it into FilenameBuffer.
1047       if (BufPtr != &FilenameBuffer[PreAppendSize])
1048         memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1049 
1050       // Resize FilenameBuffer to the correct size.
1051       if (FilenameTok.getLength() != ActualLen)
1052         FilenameBuffer.resize(PreAppendSize + ActualLen);
1053     }
1054 
1055     FilenameTok.startToken();
1056     FilenameTok.setKind(tok::header_name);
1057     FilenameTok.setFlagValue(Token::StartOfLine, StartOfLine);
1058     FilenameTok.setFlagValue(Token::LeadingSpace, LeadingSpace);
1059     FilenameTok.setFlagValue(Token::LeadingEmptyMacro, LeadingEmptyMacro);
1060     CreateString(FilenameBuffer, FilenameTok, Start, End);
1061   } else if (FilenameTok.is(tok::string_literal) && AllowMacroExpansion) {
1062     // Convert a string-literal token of the form " h-char-sequence "
1063     // (produced by macro expansion) into a header-name token.
1064     //
1065     // The rules for header-names don't quite match the rules for
1066     // string-literals, but all the places where they differ result in
1067     // undefined behavior, so we can and do treat them the same.
1068     //
1069     // A string-literal with a prefix or suffix is not translated into a
1070     // header-name. This could theoretically be observable via the C++20
1071     // context-sensitive header-name formation rules.
1072     StringRef Str = getSpelling(FilenameTok, FilenameBuffer);
1073     if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1074       FilenameTok.setKind(tok::header_name);
1075   }
1076 
1077   return false;
1078 }
1079 
1080 /// Collect the tokens of a C++20 pp-import-suffix.
1081 void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) {
1082   // FIXME: For error recovery, consider recognizing attribute syntax here
1083   // and terminating / diagnosing a missing semicolon if we find anything
1084   // else? (Can we leave that to the parser?)
1085   unsigned BracketDepth = 0;
1086   while (true) {
1087     Toks.emplace_back();
1088     Lex(Toks.back());
1089 
1090     switch (Toks.back().getKind()) {
1091     case tok::l_paren: case tok::l_square: case tok::l_brace:
1092       ++BracketDepth;
1093       break;
1094 
1095     case tok::r_paren: case tok::r_square: case tok::r_brace:
1096       if (BracketDepth == 0)
1097         return;
1098       --BracketDepth;
1099       break;
1100 
1101     case tok::semi:
1102       if (BracketDepth == 0)
1103         return;
1104     break;
1105 
1106     case tok::eof:
1107       return;
1108 
1109     default:
1110       break;
1111     }
1112   }
1113 }
1114 
1115 
1116 /// Lex a token following the 'import' contextual keyword.
1117 ///
1118 ///     pp-import: [C++20]
1119 ///           import header-name pp-import-suffix[opt] ;
1120 ///           import header-name-tokens pp-import-suffix[opt] ;
1121 /// [ObjC]    @ import module-name ;
1122 /// [Clang]   import module-name ;
1123 ///
1124 ///     header-name-tokens:
1125 ///           string-literal
1126 ///           < [any sequence of preprocessing-tokens other than >] >
1127 ///
1128 ///     module-name:
1129 ///           module-name-qualifier[opt] identifier
1130 ///
1131 ///     module-name-qualifier
1132 ///           module-name-qualifier[opt] identifier .
1133 ///
1134 /// We respond to a pp-import by importing macros from the named module.
1135 bool Preprocessor::LexAfterModuleImport(Token &Result) {
1136   // Figure out what kind of lexer we actually have.
1137   recomputeCurLexerKind();
1138 
1139   // Lex the next token. The header-name lexing rules are used at the start of
1140   // a pp-import.
1141   //
1142   // For now, we only support header-name imports in C++20 mode.
1143   // FIXME: Should we allow this in all language modes that support an import
1144   // declaration as an extension?
1145   if (NamedModuleImportPath.empty() && getLangOpts().CPlusPlusModules) {
1146     if (LexHeaderName(Result))
1147       return true;
1148 
1149     if (Result.is(tok::colon) && ModuleDeclState.isNamedModule()) {
1150       std::string Name = ModuleDeclState.getPrimaryName().str();
1151       Name += ":";
1152       NamedModuleImportPath.push_back(
1153           {getIdentifierInfo(Name), Result.getLocation()});
1154       CurLexerCallback = CLK_LexAfterModuleImport;
1155       return true;
1156     }
1157   } else {
1158     Lex(Result);
1159   }
1160 
1161   // Allocate a holding buffer for a sequence of tokens and introduce it into
1162   // the token stream.
1163   auto EnterTokens = [this](ArrayRef<Token> Toks) {
1164     auto ToksCopy = std::make_unique<Token[]>(Toks.size());
1165     std::copy(Toks.begin(), Toks.end(), ToksCopy.get());
1166     EnterTokenStream(std::move(ToksCopy), Toks.size(),
1167                      /*DisableMacroExpansion*/ true, /*IsReinject*/ false);
1168   };
1169 
1170   bool ImportingHeader = Result.is(tok::header_name);
1171   // Check for a header-name.
1172   SmallVector<Token, 32> Suffix;
1173   if (ImportingHeader) {
1174     // Enter the header-name token into the token stream; a Lex action cannot
1175     // both return a token and cache tokens (doing so would corrupt the token
1176     // cache if the call to Lex comes from CachingLex / PeekAhead).
1177     Suffix.push_back(Result);
1178 
1179     // Consume the pp-import-suffix and expand any macros in it now. We'll add
1180     // it back into the token stream later.
1181     CollectPpImportSuffix(Suffix);
1182     if (Suffix.back().isNot(tok::semi)) {
1183       // This is not a pp-import after all.
1184       EnterTokens(Suffix);
1185       return false;
1186     }
1187 
1188     // C++2a [cpp.module]p1:
1189     //   The ';' preprocessing-token terminating a pp-import shall not have
1190     //   been produced by macro replacement.
1191     SourceLocation SemiLoc = Suffix.back().getLocation();
1192     if (SemiLoc.isMacroID())
1193       Diag(SemiLoc, diag::err_header_import_semi_in_macro);
1194 
1195     // Reconstitute the import token.
1196     Token ImportTok;
1197     ImportTok.startToken();
1198     ImportTok.setKind(tok::kw_import);
1199     ImportTok.setLocation(ModuleImportLoc);
1200     ImportTok.setIdentifierInfo(getIdentifierInfo("import"));
1201     ImportTok.setLength(6);
1202 
1203     auto Action = HandleHeaderIncludeOrImport(
1204         /*HashLoc*/ SourceLocation(), ImportTok, Suffix.front(), SemiLoc);
1205     switch (Action.Kind) {
1206     case ImportAction::None:
1207       break;
1208 
1209     case ImportAction::ModuleBegin:
1210       // Let the parser know we're textually entering the module.
1211       Suffix.emplace_back();
1212       Suffix.back().startToken();
1213       Suffix.back().setKind(tok::annot_module_begin);
1214       Suffix.back().setLocation(SemiLoc);
1215       Suffix.back().setAnnotationEndLoc(SemiLoc);
1216       Suffix.back().setAnnotationValue(Action.ModuleForHeader);
1217       [[fallthrough]];
1218 
1219     case ImportAction::ModuleImport:
1220     case ImportAction::HeaderUnitImport:
1221     case ImportAction::SkippedModuleImport:
1222       // We chose to import (or textually enter) the file. Convert the
1223       // header-name token into a header unit annotation token.
1224       Suffix[0].setKind(tok::annot_header_unit);
1225       Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation());
1226       Suffix[0].setAnnotationValue(Action.ModuleForHeader);
1227       // FIXME: Call the moduleImport callback?
1228       break;
1229     case ImportAction::Failure:
1230       assert(TheModuleLoader.HadFatalFailure &&
1231              "This should be an early exit only to a fatal error");
1232       Result.setKind(tok::eof);
1233       CurLexer->cutOffLexing();
1234       EnterTokens(Suffix);
1235       return true;
1236     }
1237 
1238     EnterTokens(Suffix);
1239     return false;
1240   }
1241 
1242   // The token sequence
1243   //
1244   //   import identifier (. identifier)*
1245   //
1246   // indicates a module import directive. We already saw the 'import'
1247   // contextual keyword, so now we're looking for the identifiers.
1248   if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
1249     // We expected to see an identifier here, and we did; continue handling
1250     // identifiers.
1251     NamedModuleImportPath.push_back(
1252         std::make_pair(Result.getIdentifierInfo(), Result.getLocation()));
1253     ModuleImportExpectsIdentifier = false;
1254     CurLexerCallback = CLK_LexAfterModuleImport;
1255     return true;
1256   }
1257 
1258   // If we're expecting a '.' or a ';', and we got a '.', then wait until we
1259   // see the next identifier. (We can also see a '[[' that begins an
1260   // attribute-specifier-seq here under the Standard C++ Modules.)
1261   if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
1262     ModuleImportExpectsIdentifier = true;
1263     CurLexerCallback = CLK_LexAfterModuleImport;
1264     return true;
1265   }
1266 
1267   // If we didn't recognize a module name at all, this is not a (valid) import.
1268   if (NamedModuleImportPath.empty() || Result.is(tok::eof))
1269     return true;
1270 
1271   // Consume the pp-import-suffix and expand any macros in it now, if we're not
1272   // at the semicolon already.
1273   SourceLocation SemiLoc = Result.getLocation();
1274   if (Result.isNot(tok::semi)) {
1275     Suffix.push_back(Result);
1276     CollectPpImportSuffix(Suffix);
1277     if (Suffix.back().isNot(tok::semi)) {
1278       // This is not an import after all.
1279       EnterTokens(Suffix);
1280       return false;
1281     }
1282     SemiLoc = Suffix.back().getLocation();
1283   }
1284 
1285   // Under the standard C++ Modules, the dot is just part of the module name,
1286   // and not a real hierarchy separator. Flatten such module names now.
1287   //
1288   // FIXME: Is this the right level to be performing this transformation?
1289   std::string FlatModuleName;
1290   if (getLangOpts().CPlusPlusModules) {
1291     for (auto &Piece : NamedModuleImportPath) {
1292       // If the FlatModuleName ends with colon, it implies it is a partition.
1293       if (!FlatModuleName.empty() && FlatModuleName.back() != ':')
1294         FlatModuleName += ".";
1295       FlatModuleName += Piece.first->getName();
1296     }
1297     SourceLocation FirstPathLoc = NamedModuleImportPath[0].second;
1298     NamedModuleImportPath.clear();
1299     NamedModuleImportPath.push_back(
1300         std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
1301   }
1302 
1303   Module *Imported = nullptr;
1304   // We don't/shouldn't load the standard c++20 modules when preprocessing.
1305   if (getLangOpts().Modules && !isInImportingCXXNamedModules()) {
1306     Imported = TheModuleLoader.loadModule(ModuleImportLoc,
1307                                           NamedModuleImportPath,
1308                                           Module::Hidden,
1309                                           /*IsInclusionDirective=*/false);
1310     if (Imported)
1311       makeModuleVisible(Imported, SemiLoc);
1312   }
1313 
1314   if (Callbacks)
1315     Callbacks->moduleImport(ModuleImportLoc, NamedModuleImportPath, Imported);
1316 
1317   if (!Suffix.empty()) {
1318     EnterTokens(Suffix);
1319     return false;
1320   }
1321   return true;
1322 }
1323 
1324 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
1325   CurSubmoduleState->VisibleModules.setVisible(
1326       M, Loc, [](Module *) {},
1327       [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1328         // FIXME: Include the path in the diagnostic.
1329         // FIXME: Include the import location for the conflicting module.
1330         Diag(ModuleImportLoc, diag::warn_module_conflict)
1331             << Path[0]->getFullModuleName()
1332             << Conflict->getFullModuleName()
1333             << Message;
1334       });
1335 
1336   // Add this module to the imports list of the currently-built submodule.
1337   if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1338     BuildingSubmoduleStack.back().M->Imports.insert(M);
1339 }
1340 
1341 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1342                                           const char *DiagnosticTag,
1343                                           bool AllowMacroExpansion) {
1344   // We need at least one string literal.
1345   if (Result.isNot(tok::string_literal)) {
1346     Diag(Result, diag::err_expected_string_literal)
1347       << /*Source='in...'*/0 << DiagnosticTag;
1348     return false;
1349   }
1350 
1351   // Lex string literal tokens, optionally with macro expansion.
1352   SmallVector<Token, 4> StrToks;
1353   do {
1354     StrToks.push_back(Result);
1355 
1356     if (Result.hasUDSuffix())
1357       Diag(Result, diag::err_invalid_string_udl);
1358 
1359     if (AllowMacroExpansion)
1360       Lex(Result);
1361     else
1362       LexUnexpandedToken(Result);
1363   } while (Result.is(tok::string_literal));
1364 
1365   // Concatenate and parse the strings.
1366   StringLiteralParser Literal(StrToks, *this);
1367   assert(Literal.isOrdinary() && "Didn't allow wide strings in");
1368 
1369   if (Literal.hadError)
1370     return false;
1371 
1372   if (Literal.Pascal) {
1373     Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1374       << /*Source='in...'*/0 << DiagnosticTag;
1375     return false;
1376   }
1377 
1378   String = std::string(Literal.GetString());
1379   return true;
1380 }
1381 
1382 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1383   assert(Tok.is(tok::numeric_constant));
1384   SmallString<8> IntegerBuffer;
1385   bool NumberInvalid = false;
1386   StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1387   if (NumberInvalid)
1388     return false;
1389   NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(),
1390                                getLangOpts(), getTargetInfo(),
1391                                getDiagnostics());
1392   if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1393     return false;
1394   llvm::APInt APVal(64, 0);
1395   if (Literal.GetIntegerValue(APVal))
1396     return false;
1397   Lex(Tok);
1398   Value = APVal.getLimitedValue();
1399   return true;
1400 }
1401 
1402 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1403   assert(Handler && "NULL comment handler");
1404   assert(!llvm::is_contained(CommentHandlers, Handler) &&
1405          "Comment handler already registered");
1406   CommentHandlers.push_back(Handler);
1407 }
1408 
1409 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1410   std::vector<CommentHandler *>::iterator Pos =
1411       llvm::find(CommentHandlers, Handler);
1412   assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1413   CommentHandlers.erase(Pos);
1414 }
1415 
1416 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1417   bool AnyPendingTokens = false;
1418   for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1419        HEnd = CommentHandlers.end();
1420        H != HEnd; ++H) {
1421     if ((*H)->HandleComment(*this, Comment))
1422       AnyPendingTokens = true;
1423   }
1424   if (!AnyPendingTokens || getCommentRetentionState())
1425     return false;
1426   Lex(result);
1427   return true;
1428 }
1429 
1430 void Preprocessor::emitMacroDeprecationWarning(const Token &Identifier) const {
1431   const MacroAnnotations &A =
1432       getMacroAnnotations(Identifier.getIdentifierInfo());
1433   assert(A.DeprecationInfo &&
1434          "Macro deprecation warning without recorded annotation!");
1435   const MacroAnnotationInfo &Info = *A.DeprecationInfo;
1436   if (Info.Message.empty())
1437     Diag(Identifier, diag::warn_pragma_deprecated_macro_use)
1438         << Identifier.getIdentifierInfo() << 0;
1439   else
1440     Diag(Identifier, diag::warn_pragma_deprecated_macro_use)
1441         << Identifier.getIdentifierInfo() << 1 << Info.Message;
1442   Diag(Info.Location, diag::note_pp_macro_annotation) << 0;
1443 }
1444 
1445 void Preprocessor::emitRestrictExpansionWarning(const Token &Identifier) const {
1446   const MacroAnnotations &A =
1447       getMacroAnnotations(Identifier.getIdentifierInfo());
1448   assert(A.RestrictExpansionInfo &&
1449          "Macro restricted expansion warning without recorded annotation!");
1450   const MacroAnnotationInfo &Info = *A.RestrictExpansionInfo;
1451   if (Info.Message.empty())
1452     Diag(Identifier, diag::warn_pragma_restrict_expansion_macro_use)
1453         << Identifier.getIdentifierInfo() << 0;
1454   else
1455     Diag(Identifier, diag::warn_pragma_restrict_expansion_macro_use)
1456         << Identifier.getIdentifierInfo() << 1 << Info.Message;
1457   Diag(Info.Location, diag::note_pp_macro_annotation) << 1;
1458 }
1459 
1460 void Preprocessor::emitRestrictInfNaNWarning(const Token &Identifier,
1461                                              unsigned DiagSelection) const {
1462   Diag(Identifier, diag::warn_fp_nan_inf_when_disabled) << DiagSelection << 1;
1463 }
1464 
1465 void Preprocessor::emitFinalMacroWarning(const Token &Identifier,
1466                                          bool IsUndef) const {
1467   const MacroAnnotations &A =
1468       getMacroAnnotations(Identifier.getIdentifierInfo());
1469   assert(A.FinalAnnotationLoc &&
1470          "Final macro warning without recorded annotation!");
1471 
1472   Diag(Identifier, diag::warn_pragma_final_macro)
1473       << Identifier.getIdentifierInfo() << (IsUndef ? 0 : 1);
1474   Diag(*A.FinalAnnotationLoc, diag::note_pp_macro_annotation) << 2;
1475 }
1476 
1477 bool Preprocessor::isSafeBufferOptOut(const SourceManager &SourceMgr,
1478                                            const SourceLocation &Loc) const {
1479   // Try to find a region in `SafeBufferOptOutMap` where `Loc` is in:
1480   auto FirstRegionEndingAfterLoc = llvm::partition_point(
1481       SafeBufferOptOutMap,
1482       [&SourceMgr,
1483        &Loc](const std::pair<SourceLocation, SourceLocation> &Region) {
1484         return SourceMgr.isBeforeInTranslationUnit(Region.second, Loc);
1485       });
1486 
1487   if (FirstRegionEndingAfterLoc != SafeBufferOptOutMap.end()) {
1488     // To test if the start location of the found region precedes `Loc`:
1489     return SourceMgr.isBeforeInTranslationUnit(FirstRegionEndingAfterLoc->first,
1490                                                Loc);
1491   }
1492   // If we do not find a region whose end location passes `Loc`, we want to
1493   // check if the current region is still open:
1494   if (!SafeBufferOptOutMap.empty() &&
1495       SafeBufferOptOutMap.back().first == SafeBufferOptOutMap.back().second)
1496     return SourceMgr.isBeforeInTranslationUnit(SafeBufferOptOutMap.back().first,
1497                                                Loc);
1498   return false;
1499 }
1500 
1501 bool Preprocessor::enterOrExitSafeBufferOptOutRegion(
1502     bool isEnter, const SourceLocation &Loc) {
1503   if (isEnter) {
1504     if (isPPInSafeBufferOptOutRegion())
1505       return true; // invalid enter action
1506     InSafeBufferOptOutRegion = true;
1507     CurrentSafeBufferOptOutStart = Loc;
1508 
1509     // To set the start location of a new region:
1510 
1511     if (!SafeBufferOptOutMap.empty()) {
1512       [[maybe_unused]] auto *PrevRegion = &SafeBufferOptOutMap.back();
1513       assert(PrevRegion->first != PrevRegion->second &&
1514              "Shall not begin a safe buffer opt-out region before closing the "
1515              "previous one.");
1516     }
1517     // If the start location equals to the end location, we call the region a
1518     // open region or a unclosed region (i.e., end location has not been set
1519     // yet).
1520     SafeBufferOptOutMap.emplace_back(Loc, Loc);
1521   } else {
1522     if (!isPPInSafeBufferOptOutRegion())
1523       return true; // invalid enter action
1524     InSafeBufferOptOutRegion = false;
1525 
1526     // To set the end location of the current open region:
1527 
1528     assert(!SafeBufferOptOutMap.empty() &&
1529            "Misordered safe buffer opt-out regions");
1530     auto *CurrRegion = &SafeBufferOptOutMap.back();
1531     assert(CurrRegion->first == CurrRegion->second &&
1532            "Set end location to a closed safe buffer opt-out region");
1533     CurrRegion->second = Loc;
1534   }
1535   return false;
1536 }
1537 
1538 bool Preprocessor::isPPInSafeBufferOptOutRegion() {
1539   return InSafeBufferOptOutRegion;
1540 }
1541 bool Preprocessor::isPPInSafeBufferOptOutRegion(SourceLocation &StartLoc) {
1542   StartLoc = CurrentSafeBufferOptOutStart;
1543   return InSafeBufferOptOutRegion;
1544 }
1545 
1546 ModuleLoader::~ModuleLoader() = default;
1547 
1548 CommentHandler::~CommentHandler() = default;
1549 
1550 EmptylineHandler::~EmptylineHandler() = default;
1551 
1552 CodeCompletionHandler::~CodeCompletionHandler() = default;
1553 
1554 void Preprocessor::createPreprocessingRecord() {
1555   if (Record)
1556     return;
1557 
1558   Record = new PreprocessingRecord(getSourceManager());
1559   addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));
1560 }
1561