xref: /openbsd/gnu/llvm/clang/lib/Lex/Preprocessor.cpp (revision 12c85518)
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 
Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,DiagnosticsEngine & diags,LangOptions & opts,SourceManager & SM,HeaderSearch & Headers,ModuleLoader & TheModuleLoader,IdentifierInfoLookup * IILookup,bool OwnsHeaders,TranslationUnitKind TUKind)79 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
80                            DiagnosticsEngine &diags, 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   // If using a PCH where a #pragma hdrstop is expected, start skipping tokens.
150   if (usingPCHWithPragmaHdrStop())
151     SkippingUntilPragmaHdrStop = true;
152 
153   // If using a PCH with a through header, start skipping tokens.
154   if (!this->PPOpts->PCHThroughHeader.empty() &&
155       !this->PPOpts->ImplicitPCHInclude.empty())
156     SkippingUntilPCHThroughHeader = true;
157 
158   if (this->PPOpts->GeneratePreamble)
159     PreambleConditionalStack.startRecording();
160 
161   MaxTokens = LangOpts.MaxTokens;
162 }
163 
~Preprocessor()164 Preprocessor::~Preprocessor() {
165   assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
166 
167   IncludeMacroStack.clear();
168 
169   // Free any cached macro expanders.
170   // This populates MacroArgCache, so all TokenLexers need to be destroyed
171   // before the code below that frees up the MacroArgCache list.
172   std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
173   CurTokenLexer.reset();
174 
175   // Free any cached MacroArgs.
176   for (MacroArgs *ArgList = MacroArgCache; ArgList;)
177     ArgList = ArgList->deallocate();
178 
179   // Delete the header search info, if we own it.
180   if (OwnsHeaderSearch)
181     delete &HeaderInfo;
182 }
183 
Initialize(const TargetInfo & Target,const TargetInfo * AuxTarget)184 void Preprocessor::Initialize(const TargetInfo &Target,
185                               const TargetInfo *AuxTarget) {
186   assert((!this->Target || this->Target == &Target) &&
187          "Invalid override of target information");
188   this->Target = &Target;
189 
190   assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
191          "Invalid override of aux target information.");
192   this->AuxTarget = AuxTarget;
193 
194   // Initialize information about built-ins.
195   BuiltinInfo->InitializeTarget(Target, AuxTarget);
196   HeaderInfo.setTarget(Target);
197 
198   // Populate the identifier table with info about keywords for the current language.
199   Identifiers.AddKeywords(LangOpts);
200 
201   // Initialize the __FTL_EVAL_METHOD__ macro to the TargetInfo.
202   setTUFPEvalMethod(getTargetInfo().getFPEvalMethod());
203 
204   if (getLangOpts().getFPEvalMethod() == LangOptions::FEM_UnsetOnCommandLine)
205     // Use setting from TargetInfo.
206     setCurrentFPEvalMethod(SourceLocation(), Target.getFPEvalMethod());
207   else
208     // Set initial value of __FLT_EVAL_METHOD__ from the command line.
209     setCurrentFPEvalMethod(SourceLocation(), getLangOpts().getFPEvalMethod());
210 }
211 
InitializeForModelFile()212 void Preprocessor::InitializeForModelFile() {
213   NumEnteredSourceFiles = 0;
214 
215   // Reset pragmas
216   PragmaHandlersBackup = std::move(PragmaHandlers);
217   PragmaHandlers = std::make_unique<PragmaNamespace>(StringRef());
218   RegisterBuiltinPragmas();
219 
220   // Reset PredefinesFileID
221   PredefinesFileID = FileID();
222 }
223 
FinalizeForModelFile()224 void Preprocessor::FinalizeForModelFile() {
225   NumEnteredSourceFiles = 1;
226 
227   PragmaHandlers = std::move(PragmaHandlersBackup);
228 }
229 
DumpToken(const Token & Tok,bool DumpFlags) const230 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
231   llvm::errs() << tok::getTokenName(Tok.getKind());
232 
233   if (!Tok.isAnnotation())
234     llvm::errs() << " '" << getSpelling(Tok) << "'";
235 
236   if (!DumpFlags) return;
237 
238   llvm::errs() << "\t";
239   if (Tok.isAtStartOfLine())
240     llvm::errs() << " [StartOfLine]";
241   if (Tok.hasLeadingSpace())
242     llvm::errs() << " [LeadingSpace]";
243   if (Tok.isExpandDisabled())
244     llvm::errs() << " [ExpandDisabled]";
245   if (Tok.needsCleaning()) {
246     const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
247     llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
248                  << "']";
249   }
250 
251   llvm::errs() << "\tLoc=<";
252   DumpLocation(Tok.getLocation());
253   llvm::errs() << ">";
254 }
255 
DumpLocation(SourceLocation Loc) const256 void Preprocessor::DumpLocation(SourceLocation Loc) const {
257   Loc.print(llvm::errs(), SourceMgr);
258 }
259 
DumpMacro(const MacroInfo & MI) const260 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
261   llvm::errs() << "MACRO: ";
262   for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
263     DumpToken(MI.getReplacementToken(i));
264     llvm::errs() << "  ";
265   }
266   llvm::errs() << "\n";
267 }
268 
PrintStats()269 void Preprocessor::PrintStats() {
270   llvm::errs() << "\n*** Preprocessor Stats:\n";
271   llvm::errs() << NumDirectives << " directives found:\n";
272   llvm::errs() << "  " << NumDefined << " #define.\n";
273   llvm::errs() << "  " << NumUndefined << " #undef.\n";
274   llvm::errs() << "  #include/#include_next/#import:\n";
275   llvm::errs() << "    " << NumEnteredSourceFiles << " source files entered.\n";
276   llvm::errs() << "    " << MaxIncludeStackDepth << " max include stack depth\n";
277   llvm::errs() << "  " << NumIf << " #if/#ifndef/#ifdef.\n";
278   llvm::errs() << "  " << NumElse << " #else/#elif/#elifdef/#elifndef.\n";
279   llvm::errs() << "  " << NumEndif << " #endif.\n";
280   llvm::errs() << "  " << NumPragma << " #pragma.\n";
281   llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
282 
283   llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
284              << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
285              << NumFastMacroExpanded << " on the fast path.\n";
286   llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
287              << " token paste (##) operations performed, "
288              << NumFastTokenPaste << " on the fast path.\n";
289 
290   llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
291 
292   llvm::errs() << "\n  BumpPtr: " << BP.getTotalMemory();
293   llvm::errs() << "\n  Macro Expanded Tokens: "
294                << llvm::capacity_in_bytes(MacroExpandedTokens);
295   llvm::errs() << "\n  Predefines Buffer: " << Predefines.capacity();
296   // FIXME: List information for all submodules.
297   llvm::errs() << "\n  Macros: "
298                << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
299   llvm::errs() << "\n  #pragma push_macro Info: "
300                << llvm::capacity_in_bytes(PragmaPushMacroInfo);
301   llvm::errs() << "\n  Poison Reasons: "
302                << llvm::capacity_in_bytes(PoisonReasons);
303   llvm::errs() << "\n  Comment Handlers: "
304                << llvm::capacity_in_bytes(CommentHandlers) << "\n";
305 }
306 
307 Preprocessor::macro_iterator
macro_begin(bool IncludeExternalMacros) const308 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
309   if (IncludeExternalMacros && ExternalSource &&
310       !ReadMacrosFromExternalSource) {
311     ReadMacrosFromExternalSource = true;
312     ExternalSource->ReadDefinedMacros();
313   }
314 
315   // Make sure we cover all macros in visible modules.
316   for (const ModuleMacro &Macro : ModuleMacros)
317     CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
318 
319   return CurSubmoduleState->Macros.begin();
320 }
321 
getTotalMemory() const322 size_t Preprocessor::getTotalMemory() const {
323   return BP.getTotalMemory()
324     + llvm::capacity_in_bytes(MacroExpandedTokens)
325     + Predefines.capacity() /* Predefines buffer. */
326     // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
327     // and ModuleMacros.
328     + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
329     + llvm::capacity_in_bytes(PragmaPushMacroInfo)
330     + llvm::capacity_in_bytes(PoisonReasons)
331     + llvm::capacity_in_bytes(CommentHandlers);
332 }
333 
334 Preprocessor::macro_iterator
macro_end(bool IncludeExternalMacros) const335 Preprocessor::macro_end(bool IncludeExternalMacros) const {
336   if (IncludeExternalMacros && ExternalSource &&
337       !ReadMacrosFromExternalSource) {
338     ReadMacrosFromExternalSource = true;
339     ExternalSource->ReadDefinedMacros();
340   }
341 
342   return CurSubmoduleState->Macros.end();
343 }
344 
345 /// Compares macro tokens with a specified token value sequence.
MacroDefinitionEquals(const MacroInfo * MI,ArrayRef<TokenValue> Tokens)346 static bool MacroDefinitionEquals(const MacroInfo *MI,
347                                   ArrayRef<TokenValue> Tokens) {
348   return Tokens.size() == MI->getNumTokens() &&
349       std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
350 }
351 
getLastMacroWithSpelling(SourceLocation Loc,ArrayRef<TokenValue> Tokens) const352 StringRef Preprocessor::getLastMacroWithSpelling(
353                                     SourceLocation Loc,
354                                     ArrayRef<TokenValue> Tokens) const {
355   SourceLocation BestLocation;
356   StringRef BestSpelling;
357   for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
358        I != E; ++I) {
359     const MacroDirective::DefInfo
360       Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
361     if (!Def || !Def.getMacroInfo())
362       continue;
363     if (!Def.getMacroInfo()->isObjectLike())
364       continue;
365     if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
366       continue;
367     SourceLocation Location = Def.getLocation();
368     // Choose the macro defined latest.
369     if (BestLocation.isInvalid() ||
370         (Location.isValid() &&
371          SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
372       BestLocation = Location;
373       BestSpelling = I->first->getName();
374     }
375   }
376   return BestSpelling;
377 }
378 
recomputeCurLexerKind()379 void Preprocessor::recomputeCurLexerKind() {
380   if (CurLexer)
381     CurLexerKind = CurLexer->isDependencyDirectivesLexer()
382                        ? CLK_DependencyDirectivesLexer
383                        : CLK_Lexer;
384   else if (CurTokenLexer)
385     CurLexerKind = CLK_TokenLexer;
386   else
387     CurLexerKind = CLK_CachingLexer;
388 }
389 
SetCodeCompletionPoint(const FileEntry * File,unsigned CompleteLine,unsigned CompleteColumn)390 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
391                                           unsigned CompleteLine,
392                                           unsigned CompleteColumn) {
393   assert(File);
394   assert(CompleteLine && CompleteColumn && "Starts from 1:1");
395   assert(!CodeCompletionFile && "Already set");
396 
397   // Load the actual file's contents.
398   std::optional<llvm::MemoryBufferRef> Buffer =
399       SourceMgr.getMemoryBufferForFileOrNone(File);
400   if (!Buffer)
401     return true;
402 
403   // Find the byte position of the truncation point.
404   const char *Position = Buffer->getBufferStart();
405   for (unsigned Line = 1; Line < CompleteLine; ++Line) {
406     for (; *Position; ++Position) {
407       if (*Position != '\r' && *Position != '\n')
408         continue;
409 
410       // Eat \r\n or \n\r as a single line.
411       if ((Position[1] == '\r' || Position[1] == '\n') &&
412           Position[0] != Position[1])
413         ++Position;
414       ++Position;
415       break;
416     }
417   }
418 
419   Position += CompleteColumn - 1;
420 
421   // If pointing inside the preamble, adjust the position at the beginning of
422   // the file after the preamble.
423   if (SkipMainFilePreamble.first &&
424       SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
425     if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
426       Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
427   }
428 
429   if (Position > Buffer->getBufferEnd())
430     Position = Buffer->getBufferEnd();
431 
432   CodeCompletionFile = File;
433   CodeCompletionOffset = Position - Buffer->getBufferStart();
434 
435   auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
436       Buffer->getBufferSize() + 1, Buffer->getBufferIdentifier());
437   char *NewBuf = NewBuffer->getBufferStart();
438   char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
439   *NewPos = '\0';
440   std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
441   SourceMgr.overrideFileContents(File, std::move(NewBuffer));
442 
443   return false;
444 }
445 
CodeCompleteIncludedFile(llvm::StringRef Dir,bool IsAngled)446 void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
447                                             bool IsAngled) {
448   setCodeCompletionReached();
449   if (CodeComplete)
450     CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
451 }
452 
CodeCompleteNaturalLanguage()453 void Preprocessor::CodeCompleteNaturalLanguage() {
454   setCodeCompletionReached();
455   if (CodeComplete)
456     CodeComplete->CodeCompleteNaturalLanguage();
457 }
458 
459 /// getSpelling - This method is used to get the spelling of a token into a
460 /// SmallVector. Note that the returned StringRef may not point to the
461 /// supplied buffer if a copy can be avoided.
getSpelling(const Token & Tok,SmallVectorImpl<char> & Buffer,bool * Invalid) const462 StringRef Preprocessor::getSpelling(const Token &Tok,
463                                           SmallVectorImpl<char> &Buffer,
464                                           bool *Invalid) const {
465   // NOTE: this has to be checked *before* testing for an IdentifierInfo.
466   if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
467     // Try the fast path.
468     if (const IdentifierInfo *II = Tok.getIdentifierInfo())
469       return II->getName();
470   }
471 
472   // Resize the buffer if we need to copy into it.
473   if (Tok.needsCleaning())
474     Buffer.resize(Tok.getLength());
475 
476   const char *Ptr = Buffer.data();
477   unsigned Len = getSpelling(Tok, Ptr, Invalid);
478   return StringRef(Ptr, Len);
479 }
480 
481 /// CreateString - Plop the specified string into a scratch buffer and return a
482 /// location for it.  If specified, the source location provides a source
483 /// location for the token.
CreateString(StringRef Str,Token & Tok,SourceLocation ExpansionLocStart,SourceLocation ExpansionLocEnd)484 void Preprocessor::CreateString(StringRef Str, Token &Tok,
485                                 SourceLocation ExpansionLocStart,
486                                 SourceLocation ExpansionLocEnd) {
487   Tok.setLength(Str.size());
488 
489   const char *DestPtr;
490   SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
491 
492   if (ExpansionLocStart.isValid())
493     Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
494                                        ExpansionLocEnd, Str.size());
495   Tok.setLocation(Loc);
496 
497   // If this is a raw identifier or a literal token, set the pointer data.
498   if (Tok.is(tok::raw_identifier))
499     Tok.setRawIdentifierData(DestPtr);
500   else if (Tok.isLiteral())
501     Tok.setLiteralData(DestPtr);
502 }
503 
SplitToken(SourceLocation Loc,unsigned Length)504 SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
505   auto &SM = getSourceManager();
506   SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
507   std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(SpellingLoc);
508   bool Invalid = false;
509   StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
510   if (Invalid)
511     return SourceLocation();
512 
513   // FIXME: We could consider re-using spelling for tokens we see repeatedly.
514   const char *DestPtr;
515   SourceLocation Spelling =
516       ScratchBuf->getToken(Buffer.data() + LocInfo.second, Length, DestPtr);
517   return SM.createTokenSplitLoc(Spelling, Loc, Loc.getLocWithOffset(Length));
518 }
519 
getCurrentModule()520 Module *Preprocessor::getCurrentModule() {
521   if (!getLangOpts().isCompilingModule())
522     return nullptr;
523 
524   return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
525 }
526 
getCurrentModuleImplementation()527 Module *Preprocessor::getCurrentModuleImplementation() {
528   if (!getLangOpts().isCompilingModuleImplementation())
529     return nullptr;
530 
531   return getHeaderSearchInfo().lookupModule(getLangOpts().ModuleName);
532 }
533 
534 //===----------------------------------------------------------------------===//
535 // Preprocessor Initialization Methods
536 //===----------------------------------------------------------------------===//
537 
538 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
539 /// which implicitly adds the builtin defines etc.
EnterMainSourceFile()540 void Preprocessor::EnterMainSourceFile() {
541   // We do not allow the preprocessor to reenter the main file.  Doing so will
542   // cause FileID's to accumulate information from both runs (e.g. #line
543   // information) and predefined macros aren't guaranteed to be set properly.
544   assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
545   FileID MainFileID = SourceMgr.getMainFileID();
546 
547   // If MainFileID is loaded it means we loaded an AST file, no need to enter
548   // a main file.
549   if (!SourceMgr.isLoadedFileID(MainFileID)) {
550     // Enter the main file source buffer.
551     EnterSourceFile(MainFileID, nullptr, SourceLocation());
552 
553     // If we've been asked to skip bytes in the main file (e.g., as part of a
554     // precompiled preamble), do so now.
555     if (SkipMainFilePreamble.first > 0)
556       CurLexer->SetByteOffset(SkipMainFilePreamble.first,
557                               SkipMainFilePreamble.second);
558 
559     // Tell the header info that the main file was entered.  If the file is later
560     // #imported, it won't be re-entered.
561     if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
562       markIncluded(FE);
563   }
564 
565   // Preprocess Predefines to populate the initial preprocessor state.
566   std::unique_ptr<llvm::MemoryBuffer> SB =
567     llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
568   assert(SB && "Cannot create predefined source buffer");
569   FileID FID = SourceMgr.createFileID(std::move(SB));
570   assert(FID.isValid() && "Could not create FileID for predefines?");
571   setPredefinesFileID(FID);
572 
573   // Start parsing the predefines.
574   EnterSourceFile(FID, nullptr, SourceLocation());
575 
576   if (!PPOpts->PCHThroughHeader.empty()) {
577     // Lookup and save the FileID for the through header. If it isn't found
578     // in the search path, it's a fatal error.
579     OptionalFileEntryRef File = LookupFile(
580         SourceLocation(), PPOpts->PCHThroughHeader,
581         /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr,
582         /*CurDir=*/nullptr, /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
583         /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr,
584         /*IsFrameworkFound=*/nullptr);
585     if (!File) {
586       Diag(SourceLocation(), diag::err_pp_through_header_not_found)
587           << PPOpts->PCHThroughHeader;
588       return;
589     }
590     setPCHThroughHeaderFileID(
591         SourceMgr.createFileID(*File, SourceLocation(), SrcMgr::C_User));
592   }
593 
594   // Skip tokens from the Predefines and if needed the main file.
595   if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
596       (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
597     SkipTokensWhileUsingPCH();
598 }
599 
setPCHThroughHeaderFileID(FileID FID)600 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
601   assert(PCHThroughHeaderFileID.isInvalid() &&
602          "PCHThroughHeaderFileID already set!");
603   PCHThroughHeaderFileID = FID;
604 }
605 
isPCHThroughHeader(const FileEntry * FE)606 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
607   assert(PCHThroughHeaderFileID.isValid() &&
608          "Invalid PCH through header FileID");
609   return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
610 }
611 
creatingPCHWithThroughHeader()612 bool Preprocessor::creatingPCHWithThroughHeader() {
613   return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
614          PCHThroughHeaderFileID.isValid();
615 }
616 
usingPCHWithThroughHeader()617 bool Preprocessor::usingPCHWithThroughHeader() {
618   return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
619          PCHThroughHeaderFileID.isValid();
620 }
621 
creatingPCHWithPragmaHdrStop()622 bool Preprocessor::creatingPCHWithPragmaHdrStop() {
623   return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
624 }
625 
usingPCHWithPragmaHdrStop()626 bool Preprocessor::usingPCHWithPragmaHdrStop() {
627   return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
628 }
629 
630 /// Skip tokens until after the #include of the through header or
631 /// until after a #pragma hdrstop is seen. Tokens in the predefines file
632 /// and the main file may be skipped. If the end of the predefines file
633 /// is reached, skipping continues into the main file. If the end of the
634 /// main file is reached, it's a fatal error.
SkipTokensWhileUsingPCH()635 void Preprocessor::SkipTokensWhileUsingPCH() {
636   bool ReachedMainFileEOF = false;
637   bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
638   bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
639   Token Tok;
640   while (true) {
641     bool InPredefines =
642         (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
643     switch (CurLexerKind) {
644     case CLK_Lexer:
645       CurLexer->Lex(Tok);
646      break;
647     case CLK_TokenLexer:
648       CurTokenLexer->Lex(Tok);
649       break;
650     case CLK_CachingLexer:
651       CachingLex(Tok);
652       break;
653     case CLK_DependencyDirectivesLexer:
654       CurLexer->LexDependencyDirectiveToken(Tok);
655       break;
656     case CLK_LexAfterModuleImport:
657       LexAfterModuleImport(Tok);
658       break;
659     }
660     if (Tok.is(tok::eof) && !InPredefines) {
661       ReachedMainFileEOF = true;
662       break;
663     }
664     if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
665       break;
666     if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
667       break;
668   }
669   if (ReachedMainFileEOF) {
670     if (UsingPCHThroughHeader)
671       Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
672           << PPOpts->PCHThroughHeader << 1;
673     else if (!PPOpts->PCHWithHdrStopCreate)
674       Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
675   }
676 }
677 
replayPreambleConditionalStack()678 void Preprocessor::replayPreambleConditionalStack() {
679   // Restore the conditional stack from the preamble, if there is one.
680   if (PreambleConditionalStack.isReplaying()) {
681     assert(CurPPLexer &&
682            "CurPPLexer is null when calling replayPreambleConditionalStack.");
683     CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
684     PreambleConditionalStack.doneReplaying();
685     if (PreambleConditionalStack.reachedEOFWhileSkipping())
686       SkipExcludedConditionalBlock(
687           PreambleConditionalStack.SkipInfo->HashTokenLoc,
688           PreambleConditionalStack.SkipInfo->IfTokenLoc,
689           PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
690           PreambleConditionalStack.SkipInfo->FoundElse,
691           PreambleConditionalStack.SkipInfo->ElseLoc);
692   }
693 }
694 
EndSourceFile()695 void Preprocessor::EndSourceFile() {
696   // Notify the client that we reached the end of the source file.
697   if (Callbacks)
698     Callbacks->EndOfMainFile();
699 }
700 
701 //===----------------------------------------------------------------------===//
702 // Lexer Event Handling.
703 //===----------------------------------------------------------------------===//
704 
705 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
706 /// identifier information for the token and install it into the token,
707 /// updating the token kind accordingly.
LookUpIdentifierInfo(Token & Identifier) const708 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
709   assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
710 
711   // Look up this token, see if it is a macro, or if it is a language keyword.
712   IdentifierInfo *II;
713   if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
714     // No cleaning needed, just use the characters from the lexed buffer.
715     II = getIdentifierInfo(Identifier.getRawIdentifier());
716   } else {
717     // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
718     SmallString<64> IdentifierBuffer;
719     StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
720 
721     if (Identifier.hasUCN()) {
722       SmallString<64> UCNIdentifierBuffer;
723       expandUCNs(UCNIdentifierBuffer, CleanedStr);
724       II = getIdentifierInfo(UCNIdentifierBuffer);
725     } else {
726       II = getIdentifierInfo(CleanedStr);
727     }
728   }
729 
730   // Update the token info (identifier info and appropriate token kind).
731   // FIXME: the raw_identifier may contain leading whitespace which is removed
732   // from the cleaned identifier token. The SourceLocation should be updated to
733   // refer to the non-whitespace character. For instance, the text "\\\nB" (a
734   // line continuation before 'B') is parsed as a single tok::raw_identifier and
735   // is cleaned to tok::identifier "B". After cleaning the token's length is
736   // still 3 and the SourceLocation refers to the location of the backslash.
737   Identifier.setIdentifierInfo(II);
738   Identifier.setKind(II->getTokenID());
739 
740   return II;
741 }
742 
SetPoisonReason(IdentifierInfo * II,unsigned DiagID)743 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
744   PoisonReasons[II] = DiagID;
745 }
746 
PoisonSEHIdentifiers(bool Poison)747 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
748   assert(Ident__exception_code && Ident__exception_info);
749   assert(Ident___exception_code && Ident___exception_info);
750   Ident__exception_code->setIsPoisoned(Poison);
751   Ident___exception_code->setIsPoisoned(Poison);
752   Ident_GetExceptionCode->setIsPoisoned(Poison);
753   Ident__exception_info->setIsPoisoned(Poison);
754   Ident___exception_info->setIsPoisoned(Poison);
755   Ident_GetExceptionInfo->setIsPoisoned(Poison);
756   Ident__abnormal_termination->setIsPoisoned(Poison);
757   Ident___abnormal_termination->setIsPoisoned(Poison);
758   Ident_AbnormalTermination->setIsPoisoned(Poison);
759 }
760 
HandlePoisonedIdentifier(Token & Identifier)761 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
762   assert(Identifier.getIdentifierInfo() &&
763          "Can't handle identifiers without identifier info!");
764   llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
765     PoisonReasons.find(Identifier.getIdentifierInfo());
766   if(it == PoisonReasons.end())
767     Diag(Identifier, diag::err_pp_used_poisoned_id);
768   else
769     Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
770 }
771 
updateOutOfDateIdentifier(IdentifierInfo & II) const772 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
773   assert(II.isOutOfDate() && "not out of date");
774   getExternalSource()->updateOutOfDateIdentifier(II);
775 }
776 
777 /// HandleIdentifier - This callback is invoked when the lexer reads an
778 /// identifier.  This callback looks up the identifier in the map and/or
779 /// potentially macro expands it or turns it into a named token (like 'for').
780 ///
781 /// Note that callers of this method are guarded by checking the
782 /// IdentifierInfo's 'isHandleIdentifierCase' bit.  If this method changes, the
783 /// IdentifierInfo methods that compute these properties will need to change to
784 /// match.
HandleIdentifier(Token & Identifier)785 bool Preprocessor::HandleIdentifier(Token &Identifier) {
786   assert(Identifier.getIdentifierInfo() &&
787          "Can't handle identifiers without identifier info!");
788 
789   IdentifierInfo &II = *Identifier.getIdentifierInfo();
790 
791   // If the information about this identifier is out of date, update it from
792   // the external source.
793   // We have to treat __VA_ARGS__ in a special way, since it gets
794   // serialized with isPoisoned = true, but our preprocessor may have
795   // unpoisoned it if we're defining a C99 macro.
796   if (II.isOutOfDate()) {
797     bool CurrentIsPoisoned = false;
798     const bool IsSpecialVariadicMacro =
799         &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
800     if (IsSpecialVariadicMacro)
801       CurrentIsPoisoned = II.isPoisoned();
802 
803     updateOutOfDateIdentifier(II);
804     Identifier.setKind(II.getTokenID());
805 
806     if (IsSpecialVariadicMacro)
807       II.setIsPoisoned(CurrentIsPoisoned);
808   }
809 
810   // If this identifier was poisoned, and if it was not produced from a macro
811   // expansion, emit an error.
812   if (II.isPoisoned() && CurPPLexer) {
813     HandlePoisonedIdentifier(Identifier);
814   }
815 
816   // If this is a macro to be expanded, do it.
817   if (MacroDefinition MD = getMacroDefinition(&II)) {
818     auto *MI = MD.getMacroInfo();
819     assert(MI && "macro definition with no macro info?");
820     if (!DisableMacroExpansion) {
821       if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
822         // C99 6.10.3p10: If the preprocessing token immediately after the
823         // macro name isn't a '(', this macro should not be expanded.
824         if (!MI->isFunctionLike() || isNextPPTokenLParen())
825           return HandleMacroExpandedIdentifier(Identifier, MD);
826       } else {
827         // C99 6.10.3.4p2 says that a disabled macro may never again be
828         // expanded, even if it's in a context where it could be expanded in the
829         // future.
830         Identifier.setFlag(Token::DisableExpand);
831         if (MI->isObjectLike() || isNextPPTokenLParen())
832           Diag(Identifier, diag::pp_disabled_macro_expansion);
833       }
834     }
835   }
836 
837   // If this identifier is a keyword in a newer Standard or proposed Standard,
838   // produce a warning. Don't warn if we're not considering macro expansion,
839   // since this identifier might be the name of a macro.
840   // FIXME: This warning is disabled in cases where it shouldn't be, like
841   //   "#define constexpr constexpr", "int constexpr;"
842   if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
843     Diag(Identifier, getIdentifierTable().getFutureCompatDiagKind(II, getLangOpts()))
844         << II.getName();
845     // Don't diagnose this keyword again in this translation unit.
846     II.setIsFutureCompatKeyword(false);
847   }
848 
849   // If this is an extension token, diagnose its use.
850   // We avoid diagnosing tokens that originate from macro definitions.
851   // FIXME: This warning is disabled in cases where it shouldn't be,
852   // like "#define TY typeof", "TY(1) x".
853   if (II.isExtensionToken() && !DisableMacroExpansion)
854     Diag(Identifier, diag::ext_token_used);
855 
856   // If this is the 'import' contextual keyword following an '@', note
857   // that the next token indicates a module name.
858   //
859   // Note that we do not treat 'import' as a contextual
860   // keyword when we're in a caching lexer, because caching lexers only get
861   // used in contexts where import declarations are disallowed.
862   //
863   // Likewise if this is the C++ Modules TS import keyword.
864   if (((LastTokenWasAt && II.isModulesImport()) ||
865        Identifier.is(tok::kw_import)) &&
866       !InMacroArgs && !DisableMacroExpansion &&
867       (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
868       CurLexerKind != CLK_CachingLexer) {
869     ModuleImportLoc = Identifier.getLocation();
870     NamedModuleImportPath.clear();
871     IsAtImport = true;
872     ModuleImportExpectsIdentifier = true;
873     CurLexerKind = CLK_LexAfterModuleImport;
874   }
875   return true;
876 }
877 
Lex(Token & Result)878 void Preprocessor::Lex(Token &Result) {
879   ++LexLevel;
880 
881   // We loop here until a lex function returns a token; this avoids recursion.
882   bool ReturnedToken;
883   do {
884     switch (CurLexerKind) {
885     case CLK_Lexer:
886       ReturnedToken = CurLexer->Lex(Result);
887       break;
888     case CLK_TokenLexer:
889       ReturnedToken = CurTokenLexer->Lex(Result);
890       break;
891     case CLK_CachingLexer:
892       CachingLex(Result);
893       ReturnedToken = true;
894       break;
895     case CLK_DependencyDirectivesLexer:
896       ReturnedToken = CurLexer->LexDependencyDirectiveToken(Result);
897       break;
898     case CLK_LexAfterModuleImport:
899       ReturnedToken = LexAfterModuleImport(Result);
900       break;
901     }
902   } while (!ReturnedToken);
903 
904   if (Result.is(tok::unknown) && TheModuleLoader.HadFatalFailure)
905     return;
906 
907   if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
908     // Remember the identifier before code completion token.
909     setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
910     setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
911     // Set IdenfitierInfo to null to avoid confusing code that handles both
912     // identifiers and completion tokens.
913     Result.setIdentifierInfo(nullptr);
914   }
915 
916   // Update StdCXXImportSeqState to track our position within a C++20 import-seq
917   // if this token is being produced as a result of phase 4 of translation.
918   // Update TrackGMFState to decide if we are currently in a Global Module
919   // Fragment. GMF state updates should precede StdCXXImportSeq ones, since GMF state
920   // depends on the prevailing StdCXXImportSeq state in two cases.
921   if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
922       !Result.getFlag(Token::IsReinjected)) {
923     switch (Result.getKind()) {
924     case tok::l_paren: case tok::l_square: case tok::l_brace:
925       StdCXXImportSeqState.handleOpenBracket();
926       break;
927     case tok::r_paren: case tok::r_square:
928       StdCXXImportSeqState.handleCloseBracket();
929       break;
930     case tok::r_brace:
931       StdCXXImportSeqState.handleCloseBrace();
932       break;
933     // This token is injected to represent the translation of '#include "a.h"'
934     // into "import a.h;". Mimic the notional ';'.
935     case tok::annot_module_include:
936     case tok::semi:
937       TrackGMFState.handleSemi();
938       StdCXXImportSeqState.handleSemi();
939       ModuleDeclState.handleSemi();
940       break;
941     case tok::header_name:
942     case tok::annot_header_unit:
943       StdCXXImportSeqState.handleHeaderName();
944       break;
945     case tok::kw_export:
946       TrackGMFState.handleExport();
947       StdCXXImportSeqState.handleExport();
948       ModuleDeclState.handleExport();
949       break;
950     case tok::colon:
951       ModuleDeclState.handleColon();
952       break;
953     case tok::period:
954       ModuleDeclState.handlePeriod();
955       break;
956     case tok::identifier:
957       if (Result.getIdentifierInfo()->isModulesImport()) {
958         TrackGMFState.handleImport(StdCXXImportSeqState.afterTopLevelSeq());
959         StdCXXImportSeqState.handleImport();
960         if (StdCXXImportSeqState.afterImportSeq()) {
961           ModuleImportLoc = Result.getLocation();
962           NamedModuleImportPath.clear();
963           IsAtImport = false;
964           ModuleImportExpectsIdentifier = true;
965           CurLexerKind = CLK_LexAfterModuleImport;
966         }
967         break;
968       } else if (Result.getIdentifierInfo() == getIdentifierInfo("module")) {
969         TrackGMFState.handleModule(StdCXXImportSeqState.afterTopLevelSeq());
970         ModuleDeclState.handleModule();
971         break;
972       } else {
973         ModuleDeclState.handleIdentifier(Result.getIdentifierInfo());
974         if (ModuleDeclState.isModuleCandidate())
975           break;
976       }
977       [[fallthrough]];
978     default:
979       TrackGMFState.handleMisc();
980       StdCXXImportSeqState.handleMisc();
981       ModuleDeclState.handleMisc();
982       break;
983     }
984   }
985 
986   LastTokenWasAt = Result.is(tok::at);
987   --LexLevel;
988 
989   if ((LexLevel == 0 || PreprocessToken) &&
990       !Result.getFlag(Token::IsReinjected)) {
991     if (LexLevel == 0)
992       ++TokenCount;
993     if (OnToken)
994       OnToken(Result);
995   }
996 }
997 
998 /// Lex a header-name token (including one formed from header-name-tokens if
999 /// \p AllowConcatenation is \c true).
1000 ///
1001 /// \param FilenameTok Filled in with the next token. On success, this will
1002 ///        be either a header_name token. On failure, it will be whatever other
1003 ///        token was found instead.
1004 /// \param AllowMacroExpansion If \c true, allow the header name to be formed
1005 ///        by macro expansion (concatenating tokens as necessary if the first
1006 ///        token is a '<').
1007 /// \return \c true if we reached EOD or EOF while looking for a > token in
1008 ///         a concatenated header name and diagnosed it. \c false otherwise.
LexHeaderName(Token & FilenameTok,bool AllowMacroExpansion)1009 bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
1010   // Lex using header-name tokenization rules if tokens are being lexed from
1011   // a file. Just grab a token normally if we're in a macro expansion.
1012   if (CurPPLexer)
1013     CurPPLexer->LexIncludeFilename(FilenameTok);
1014   else
1015     Lex(FilenameTok);
1016 
1017   // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1018   // case, glue the tokens together into an angle_string_literal token.
1019   SmallString<128> FilenameBuffer;
1020   if (FilenameTok.is(tok::less) && AllowMacroExpansion) {
1021     bool StartOfLine = FilenameTok.isAtStartOfLine();
1022     bool LeadingSpace = FilenameTok.hasLeadingSpace();
1023     bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
1024 
1025     SourceLocation Start = FilenameTok.getLocation();
1026     SourceLocation End;
1027     FilenameBuffer.push_back('<');
1028 
1029     // Consume tokens until we find a '>'.
1030     // FIXME: A header-name could be formed starting or ending with an
1031     // alternative token. It's not clear whether that's ill-formed in all
1032     // cases.
1033     while (FilenameTok.isNot(tok::greater)) {
1034       Lex(FilenameTok);
1035       if (FilenameTok.isOneOf(tok::eod, tok::eof)) {
1036         Diag(FilenameTok.getLocation(), diag::err_expected) << tok::greater;
1037         Diag(Start, diag::note_matching) << tok::less;
1038         return true;
1039       }
1040 
1041       End = FilenameTok.getLocation();
1042 
1043       // FIXME: Provide code completion for #includes.
1044       if (FilenameTok.is(tok::code_completion)) {
1045         setCodeCompletionReached();
1046         Lex(FilenameTok);
1047         continue;
1048       }
1049 
1050       // Append the spelling of this token to the buffer. If there was a space
1051       // before it, add it now.
1052       if (FilenameTok.hasLeadingSpace())
1053         FilenameBuffer.push_back(' ');
1054 
1055       // Get the spelling of the token, directly into FilenameBuffer if
1056       // possible.
1057       size_t PreAppendSize = FilenameBuffer.size();
1058       FilenameBuffer.resize(PreAppendSize + FilenameTok.getLength());
1059 
1060       const char *BufPtr = &FilenameBuffer[PreAppendSize];
1061       unsigned ActualLen = getSpelling(FilenameTok, BufPtr);
1062 
1063       // If the token was spelled somewhere else, copy it into FilenameBuffer.
1064       if (BufPtr != &FilenameBuffer[PreAppendSize])
1065         memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1066 
1067       // Resize FilenameBuffer to the correct size.
1068       if (FilenameTok.getLength() != ActualLen)
1069         FilenameBuffer.resize(PreAppendSize + ActualLen);
1070     }
1071 
1072     FilenameTok.startToken();
1073     FilenameTok.setKind(tok::header_name);
1074     FilenameTok.setFlagValue(Token::StartOfLine, StartOfLine);
1075     FilenameTok.setFlagValue(Token::LeadingSpace, LeadingSpace);
1076     FilenameTok.setFlagValue(Token::LeadingEmptyMacro, LeadingEmptyMacro);
1077     CreateString(FilenameBuffer, FilenameTok, Start, End);
1078   } else if (FilenameTok.is(tok::string_literal) && AllowMacroExpansion) {
1079     // Convert a string-literal token of the form " h-char-sequence "
1080     // (produced by macro expansion) into a header-name token.
1081     //
1082     // The rules for header-names don't quite match the rules for
1083     // string-literals, but all the places where they differ result in
1084     // undefined behavior, so we can and do treat them the same.
1085     //
1086     // A string-literal with a prefix or suffix is not translated into a
1087     // header-name. This could theoretically be observable via the C++20
1088     // context-sensitive header-name formation rules.
1089     StringRef Str = getSpelling(FilenameTok, FilenameBuffer);
1090     if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1091       FilenameTok.setKind(tok::header_name);
1092   }
1093 
1094   return false;
1095 }
1096 
1097 /// Collect the tokens of a C++20 pp-import-suffix.
CollectPpImportSuffix(SmallVectorImpl<Token> & Toks)1098 void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) {
1099   // FIXME: For error recovery, consider recognizing attribute syntax here
1100   // and terminating / diagnosing a missing semicolon if we find anything
1101   // else? (Can we leave that to the parser?)
1102   unsigned BracketDepth = 0;
1103   while (true) {
1104     Toks.emplace_back();
1105     Lex(Toks.back());
1106 
1107     switch (Toks.back().getKind()) {
1108     case tok::l_paren: case tok::l_square: case tok::l_brace:
1109       ++BracketDepth;
1110       break;
1111 
1112     case tok::r_paren: case tok::r_square: case tok::r_brace:
1113       if (BracketDepth == 0)
1114         return;
1115       --BracketDepth;
1116       break;
1117 
1118     case tok::semi:
1119       if (BracketDepth == 0)
1120         return;
1121     break;
1122 
1123     case tok::eof:
1124       return;
1125 
1126     default:
1127       break;
1128     }
1129   }
1130 }
1131 
1132 
1133 /// Lex a token following the 'import' contextual keyword.
1134 ///
1135 ///     pp-import: [C++20]
1136 ///           import header-name pp-import-suffix[opt] ;
1137 ///           import header-name-tokens pp-import-suffix[opt] ;
1138 /// [ObjC]    @ import module-name ;
1139 /// [Clang]   import module-name ;
1140 ///
1141 ///     header-name-tokens:
1142 ///           string-literal
1143 ///           < [any sequence of preprocessing-tokens other than >] >
1144 ///
1145 ///     module-name:
1146 ///           module-name-qualifier[opt] identifier
1147 ///
1148 ///     module-name-qualifier
1149 ///           module-name-qualifier[opt] identifier .
1150 ///
1151 /// We respond to a pp-import by importing macros from the named module.
LexAfterModuleImport(Token & Result)1152 bool Preprocessor::LexAfterModuleImport(Token &Result) {
1153   // Figure out what kind of lexer we actually have.
1154   recomputeCurLexerKind();
1155 
1156   // Lex the next token. The header-name lexing rules are used at the start of
1157   // a pp-import.
1158   //
1159   // For now, we only support header-name imports in C++20 mode.
1160   // FIXME: Should we allow this in all language modes that support an import
1161   // declaration as an extension?
1162   if (NamedModuleImportPath.empty() && getLangOpts().CPlusPlusModules) {
1163     if (LexHeaderName(Result))
1164       return true;
1165 
1166     if (Result.is(tok::colon) && ModuleDeclState.isNamedModule()) {
1167       std::string Name = ModuleDeclState.getPrimaryName().str();
1168       Name += ":";
1169       NamedModuleImportPath.push_back(
1170           {getIdentifierInfo(Name), Result.getLocation()});
1171       CurLexerKind = CLK_LexAfterModuleImport;
1172       return true;
1173     }
1174   } else {
1175     Lex(Result);
1176   }
1177 
1178   // Allocate a holding buffer for a sequence of tokens and introduce it into
1179   // the token stream.
1180   auto EnterTokens = [this](ArrayRef<Token> Toks) {
1181     auto ToksCopy = std::make_unique<Token[]>(Toks.size());
1182     std::copy(Toks.begin(), Toks.end(), ToksCopy.get());
1183     EnterTokenStream(std::move(ToksCopy), Toks.size(),
1184                      /*DisableMacroExpansion*/ true, /*IsReinject*/ false);
1185   };
1186 
1187   bool ImportingHeader = Result.is(tok::header_name);
1188   // Check for a header-name.
1189   SmallVector<Token, 32> Suffix;
1190   if (ImportingHeader) {
1191     // Enter the header-name token into the token stream; a Lex action cannot
1192     // both return a token and cache tokens (doing so would corrupt the token
1193     // cache if the call to Lex comes from CachingLex / PeekAhead).
1194     Suffix.push_back(Result);
1195 
1196     // Consume the pp-import-suffix and expand any macros in it now. We'll add
1197     // it back into the token stream later.
1198     CollectPpImportSuffix(Suffix);
1199     if (Suffix.back().isNot(tok::semi)) {
1200       // This is not a pp-import after all.
1201       EnterTokens(Suffix);
1202       return false;
1203     }
1204 
1205     // C++2a [cpp.module]p1:
1206     //   The ';' preprocessing-token terminating a pp-import shall not have
1207     //   been produced by macro replacement.
1208     SourceLocation SemiLoc = Suffix.back().getLocation();
1209     if (SemiLoc.isMacroID())
1210       Diag(SemiLoc, diag::err_header_import_semi_in_macro);
1211 
1212     // Reconstitute the import token.
1213     Token ImportTok;
1214     ImportTok.startToken();
1215     ImportTok.setKind(tok::kw_import);
1216     ImportTok.setLocation(ModuleImportLoc);
1217     ImportTok.setIdentifierInfo(getIdentifierInfo("import"));
1218     ImportTok.setLength(6);
1219 
1220     auto Action = HandleHeaderIncludeOrImport(
1221         /*HashLoc*/ SourceLocation(), ImportTok, Suffix.front(), SemiLoc);
1222     switch (Action.Kind) {
1223     case ImportAction::None:
1224       break;
1225 
1226     case ImportAction::ModuleBegin:
1227       // Let the parser know we're textually entering the module.
1228       Suffix.emplace_back();
1229       Suffix.back().startToken();
1230       Suffix.back().setKind(tok::annot_module_begin);
1231       Suffix.back().setLocation(SemiLoc);
1232       Suffix.back().setAnnotationEndLoc(SemiLoc);
1233       Suffix.back().setAnnotationValue(Action.ModuleForHeader);
1234       [[fallthrough]];
1235 
1236     case ImportAction::ModuleImport:
1237     case ImportAction::HeaderUnitImport:
1238     case ImportAction::SkippedModuleImport:
1239       // We chose to import (or textually enter) the file. Convert the
1240       // header-name token into a header unit annotation token.
1241       Suffix[0].setKind(tok::annot_header_unit);
1242       Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation());
1243       Suffix[0].setAnnotationValue(Action.ModuleForHeader);
1244       // FIXME: Call the moduleImport callback?
1245       break;
1246     case ImportAction::Failure:
1247       assert(TheModuleLoader.HadFatalFailure &&
1248              "This should be an early exit only to a fatal error");
1249       Result.setKind(tok::eof);
1250       CurLexer->cutOffLexing();
1251       EnterTokens(Suffix);
1252       return true;
1253     }
1254 
1255     EnterTokens(Suffix);
1256     return false;
1257   }
1258 
1259   // The token sequence
1260   //
1261   //   import identifier (. identifier)*
1262   //
1263   // indicates a module import directive. We already saw the 'import'
1264   // contextual keyword, so now we're looking for the identifiers.
1265   if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
1266     // We expected to see an identifier here, and we did; continue handling
1267     // identifiers.
1268     NamedModuleImportPath.push_back(
1269         std::make_pair(Result.getIdentifierInfo(), Result.getLocation()));
1270     ModuleImportExpectsIdentifier = false;
1271     CurLexerKind = CLK_LexAfterModuleImport;
1272     return true;
1273   }
1274 
1275   // If we're expecting a '.' or a ';', and we got a '.', then wait until we
1276   // see the next identifier. (We can also see a '[[' that begins an
1277   // attribute-specifier-seq here under the C++ Modules TS.)
1278   if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
1279     ModuleImportExpectsIdentifier = true;
1280     CurLexerKind = CLK_LexAfterModuleImport;
1281     return true;
1282   }
1283 
1284   // If we didn't recognize a module name at all, this is not a (valid) import.
1285   if (NamedModuleImportPath.empty() || Result.is(tok::eof))
1286     return true;
1287 
1288   // Consume the pp-import-suffix and expand any macros in it now, if we're not
1289   // at the semicolon already.
1290   SourceLocation SemiLoc = Result.getLocation();
1291   if (Result.isNot(tok::semi)) {
1292     Suffix.push_back(Result);
1293     CollectPpImportSuffix(Suffix);
1294     if (Suffix.back().isNot(tok::semi)) {
1295       // This is not an import after all.
1296       EnterTokens(Suffix);
1297       return false;
1298     }
1299     SemiLoc = Suffix.back().getLocation();
1300   }
1301 
1302   // Under the Modules TS, the dot is just part of the module name, and not
1303   // a real hierarchy separator. Flatten such module names now.
1304   //
1305   // FIXME: Is this the right level to be performing this transformation?
1306   std::string FlatModuleName;
1307   if (getLangOpts().ModulesTS || getLangOpts().CPlusPlusModules) {
1308     for (auto &Piece : NamedModuleImportPath) {
1309       // If the FlatModuleName ends with colon, it implies it is a partition.
1310       if (!FlatModuleName.empty() && FlatModuleName.back() != ':')
1311         FlatModuleName += ".";
1312       FlatModuleName += Piece.first->getName();
1313     }
1314     SourceLocation FirstPathLoc = NamedModuleImportPath[0].second;
1315     NamedModuleImportPath.clear();
1316     NamedModuleImportPath.push_back(
1317         std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
1318   }
1319 
1320   Module *Imported = nullptr;
1321   // We don't/shouldn't load the standard c++20 modules when preprocessing.
1322   if (getLangOpts().Modules && !isInImportingCXXNamedModules()) {
1323     Imported = TheModuleLoader.loadModule(ModuleImportLoc,
1324                                           NamedModuleImportPath,
1325                                           Module::Hidden,
1326                                           /*IsInclusionDirective=*/false);
1327     if (Imported)
1328       makeModuleVisible(Imported, SemiLoc);
1329   }
1330 
1331   if (Callbacks)
1332     Callbacks->moduleImport(ModuleImportLoc, NamedModuleImportPath, Imported);
1333 
1334   if (!Suffix.empty()) {
1335     EnterTokens(Suffix);
1336     return false;
1337   }
1338   return true;
1339 }
1340 
makeModuleVisible(Module * M,SourceLocation Loc)1341 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
1342   CurSubmoduleState->VisibleModules.setVisible(
1343       M, Loc, [](Module *) {},
1344       [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1345         // FIXME: Include the path in the diagnostic.
1346         // FIXME: Include the import location for the conflicting module.
1347         Diag(ModuleImportLoc, diag::warn_module_conflict)
1348             << Path[0]->getFullModuleName()
1349             << Conflict->getFullModuleName()
1350             << Message;
1351       });
1352 
1353   // Add this module to the imports list of the currently-built submodule.
1354   if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1355     BuildingSubmoduleStack.back().M->Imports.insert(M);
1356 }
1357 
FinishLexStringLiteral(Token & Result,std::string & String,const char * DiagnosticTag,bool AllowMacroExpansion)1358 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1359                                           const char *DiagnosticTag,
1360                                           bool AllowMacroExpansion) {
1361   // We need at least one string literal.
1362   if (Result.isNot(tok::string_literal)) {
1363     Diag(Result, diag::err_expected_string_literal)
1364       << /*Source='in...'*/0 << DiagnosticTag;
1365     return false;
1366   }
1367 
1368   // Lex string literal tokens, optionally with macro expansion.
1369   SmallVector<Token, 4> StrToks;
1370   do {
1371     StrToks.push_back(Result);
1372 
1373     if (Result.hasUDSuffix())
1374       Diag(Result, diag::err_invalid_string_udl);
1375 
1376     if (AllowMacroExpansion)
1377       Lex(Result);
1378     else
1379       LexUnexpandedToken(Result);
1380   } while (Result.is(tok::string_literal));
1381 
1382   // Concatenate and parse the strings.
1383   StringLiteralParser Literal(StrToks, *this);
1384   assert(Literal.isOrdinary() && "Didn't allow wide strings in");
1385 
1386   if (Literal.hadError)
1387     return false;
1388 
1389   if (Literal.Pascal) {
1390     Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1391       << /*Source='in...'*/0 << DiagnosticTag;
1392     return false;
1393   }
1394 
1395   String = std::string(Literal.GetString());
1396   return true;
1397 }
1398 
parseSimpleIntegerLiteral(Token & Tok,uint64_t & Value)1399 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1400   assert(Tok.is(tok::numeric_constant));
1401   SmallString<8> IntegerBuffer;
1402   bool NumberInvalid = false;
1403   StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1404   if (NumberInvalid)
1405     return false;
1406   NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(),
1407                                getLangOpts(), getTargetInfo(),
1408                                getDiagnostics());
1409   if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1410     return false;
1411   llvm::APInt APVal(64, 0);
1412   if (Literal.GetIntegerValue(APVal))
1413     return false;
1414   Lex(Tok);
1415   Value = APVal.getLimitedValue();
1416   return true;
1417 }
1418 
addCommentHandler(CommentHandler * Handler)1419 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1420   assert(Handler && "NULL comment handler");
1421   assert(!llvm::is_contained(CommentHandlers, Handler) &&
1422          "Comment handler already registered");
1423   CommentHandlers.push_back(Handler);
1424 }
1425 
removeCommentHandler(CommentHandler * Handler)1426 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1427   std::vector<CommentHandler *>::iterator Pos =
1428       llvm::find(CommentHandlers, Handler);
1429   assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1430   CommentHandlers.erase(Pos);
1431 }
1432 
HandleComment(Token & result,SourceRange Comment)1433 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1434   bool AnyPendingTokens = false;
1435   for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1436        HEnd = CommentHandlers.end();
1437        H != HEnd; ++H) {
1438     if ((*H)->HandleComment(*this, Comment))
1439       AnyPendingTokens = true;
1440   }
1441   if (!AnyPendingTokens || getCommentRetentionState())
1442     return false;
1443   Lex(result);
1444   return true;
1445 }
1446 
emitMacroDeprecationWarning(const Token & Identifier) const1447 void Preprocessor::emitMacroDeprecationWarning(const Token &Identifier) const {
1448   const MacroAnnotations &A =
1449       getMacroAnnotations(Identifier.getIdentifierInfo());
1450   assert(A.DeprecationInfo &&
1451          "Macro deprecation warning without recorded annotation!");
1452   const MacroAnnotationInfo &Info = *A.DeprecationInfo;
1453   if (Info.Message.empty())
1454     Diag(Identifier, diag::warn_pragma_deprecated_macro_use)
1455         << Identifier.getIdentifierInfo() << 0;
1456   else
1457     Diag(Identifier, diag::warn_pragma_deprecated_macro_use)
1458         << Identifier.getIdentifierInfo() << 1 << Info.Message;
1459   Diag(Info.Location, diag::note_pp_macro_annotation) << 0;
1460 }
1461 
emitRestrictExpansionWarning(const Token & Identifier) const1462 void Preprocessor::emitRestrictExpansionWarning(const Token &Identifier) const {
1463   const MacroAnnotations &A =
1464       getMacroAnnotations(Identifier.getIdentifierInfo());
1465   assert(A.RestrictExpansionInfo &&
1466          "Macro restricted expansion warning without recorded annotation!");
1467   const MacroAnnotationInfo &Info = *A.RestrictExpansionInfo;
1468   if (Info.Message.empty())
1469     Diag(Identifier, diag::warn_pragma_restrict_expansion_macro_use)
1470         << Identifier.getIdentifierInfo() << 0;
1471   else
1472     Diag(Identifier, diag::warn_pragma_restrict_expansion_macro_use)
1473         << Identifier.getIdentifierInfo() << 1 << Info.Message;
1474   Diag(Info.Location, diag::note_pp_macro_annotation) << 1;
1475 }
1476 
emitFinalMacroWarning(const Token & Identifier,bool IsUndef) const1477 void Preprocessor::emitFinalMacroWarning(const Token &Identifier,
1478                                          bool IsUndef) const {
1479   const MacroAnnotations &A =
1480       getMacroAnnotations(Identifier.getIdentifierInfo());
1481   assert(A.FinalAnnotationLoc &&
1482          "Final macro warning without recorded annotation!");
1483 
1484   Diag(Identifier, diag::warn_pragma_final_macro)
1485       << Identifier.getIdentifierInfo() << (IsUndef ? 0 : 1);
1486   Diag(*A.FinalAnnotationLoc, diag::note_pp_macro_annotation) << 2;
1487 }
1488 
1489 ModuleLoader::~ModuleLoader() = default;
1490 
1491 CommentHandler::~CommentHandler() = default;
1492 
1493 EmptylineHandler::~EmptylineHandler() = default;
1494 
1495 CodeCompletionHandler::~CodeCompletionHandler() = default;
1496 
createPreprocessingRecord()1497 void Preprocessor::createPreprocessingRecord() {
1498   if (Record)
1499     return;
1500 
1501   Record = new PreprocessingRecord(getSourceManager());
1502   addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));
1503 }
1504