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/ADT/StringSwitch.h"
62#include "llvm/Support/Capacity.h"
63#include "llvm/Support/ErrorHandling.h"
64#include "llvm/Support/MemoryBuffer.h"
65#include "llvm/Support/raw_ostream.h"
66#include <algorithm>
67#include <cassert>
68#include <memory>
69#include <string>
70#include <utility>
71#include <vector>
72
73using namespace clang;
74
75LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
76
77ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
78
79Preprocessor::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  if (getLangOpts().CPlusPlus20) {
123    (Ident__VA_OPT__ = getIdentifierInfo("__VA_OPT__"))->setIsPoisoned();
124    SetPoisonReason(Ident__VA_OPT__,diag::ext_pp_bad_vaopt_use);
125  } else {
126    Ident__VA_OPT__ = nullptr;
127  }
128
129  // Initialize the pragma handlers.
130  RegisterBuiltinPragmas();
131
132  // Initialize builtin macros like __LINE__ and friends.
133  RegisterBuiltinMacros();
134
135  if(LangOpts.Borland) {
136    Ident__exception_info        = getIdentifierInfo("_exception_info");
137    Ident___exception_info       = getIdentifierInfo("__exception_info");
138    Ident_GetExceptionInfo       = getIdentifierInfo("GetExceptionInformation");
139    Ident__exception_code        = getIdentifierInfo("_exception_code");
140    Ident___exception_code       = getIdentifierInfo("__exception_code");
141    Ident_GetExceptionCode       = getIdentifierInfo("GetExceptionCode");
142    Ident__abnormal_termination  = getIdentifierInfo("_abnormal_termination");
143    Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
144    Ident_AbnormalTermination    = getIdentifierInfo("AbnormalTermination");
145  } else {
146    Ident__exception_info = Ident__exception_code = nullptr;
147    Ident__abnormal_termination = Ident___exception_info = nullptr;
148    Ident___exception_code = Ident___abnormal_termination = nullptr;
149    Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr;
150    Ident_AbnormalTermination = nullptr;
151  }
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  ExcludedConditionalDirectiveSkipMappings =
166      this->PPOpts->ExcludedConditionalDirectiveSkipMappings;
167  if (ExcludedConditionalDirectiveSkipMappings)
168    ExcludedConditionalDirectiveSkipMappings->clear();
169
170  MaxTokens = LangOpts.MaxTokens;
171}
172
173Preprocessor::~Preprocessor() {
174  assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
175
176  IncludeMacroStack.clear();
177
178  // Destroy any macro definitions.
179  while (MacroInfoChain *I = MIChainHead) {
180    MIChainHead = I->Next;
181    I->~MacroInfoChain();
182  }
183
184  // Free any cached macro expanders.
185  // This populates MacroArgCache, so all TokenLexers need to be destroyed
186  // before the code below that frees up the MacroArgCache list.
187  std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
188  CurTokenLexer.reset();
189
190  // Free any cached MacroArgs.
191  for (MacroArgs *ArgList = MacroArgCache; ArgList;)
192    ArgList = ArgList->deallocate();
193
194  // Delete the header search info, if we own it.
195  if (OwnsHeaderSearch)
196    delete &HeaderInfo;
197}
198
199void Preprocessor::Initialize(const TargetInfo &Target,
200                              const TargetInfo *AuxTarget) {
201  assert((!this->Target || this->Target == &Target) &&
202         "Invalid override of target information");
203  this->Target = &Target;
204
205  assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
206         "Invalid override of aux target information.");
207  this->AuxTarget = AuxTarget;
208
209  // Initialize information about built-ins.
210  BuiltinInfo->InitializeTarget(Target, AuxTarget);
211  HeaderInfo.setTarget(Target);
212
213  // Populate the identifier table with info about keywords for the current language.
214  Identifiers.AddKeywords(LangOpts);
215}
216
217void Preprocessor::InitializeForModelFile() {
218  NumEnteredSourceFiles = 0;
219
220  // Reset pragmas
221  PragmaHandlersBackup = std::move(PragmaHandlers);
222  PragmaHandlers = std::make_unique<PragmaNamespace>(StringRef());
223  RegisterBuiltinPragmas();
224
225  // Reset PredefinesFileID
226  PredefinesFileID = FileID();
227}
228
229void Preprocessor::FinalizeForModelFile() {
230  NumEnteredSourceFiles = 1;
231
232  PragmaHandlers = std::move(PragmaHandlersBackup);
233}
234
235void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
236  llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
237               << getSpelling(Tok) << "'";
238
239  if (!DumpFlags) return;
240
241  llvm::errs() << "\t";
242  if (Tok.isAtStartOfLine())
243    llvm::errs() << " [StartOfLine]";
244  if (Tok.hasLeadingSpace())
245    llvm::errs() << " [LeadingSpace]";
246  if (Tok.isExpandDisabled())
247    llvm::errs() << " [ExpandDisabled]";
248  if (Tok.needsCleaning()) {
249    const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
250    llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
251                 << "']";
252  }
253
254  llvm::errs() << "\tLoc=<";
255  DumpLocation(Tok.getLocation());
256  llvm::errs() << ">";
257}
258
259void Preprocessor::DumpLocation(SourceLocation Loc) const {
260  Loc.print(llvm::errs(), SourceMgr);
261}
262
263void Preprocessor::DumpMacro(const MacroInfo &MI) const {
264  llvm::errs() << "MACRO: ";
265  for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
266    DumpToken(MI.getReplacementToken(i));
267    llvm::errs() << "  ";
268  }
269  llvm::errs() << "\n";
270}
271
272void Preprocessor::PrintStats() {
273  llvm::errs() << "\n*** Preprocessor Stats:\n";
274  llvm::errs() << NumDirectives << " directives found:\n";
275  llvm::errs() << "  " << NumDefined << " #define.\n";
276  llvm::errs() << "  " << NumUndefined << " #undef.\n";
277  llvm::errs() << "  #include/#include_next/#import:\n";
278  llvm::errs() << "    " << NumEnteredSourceFiles << " source files entered.\n";
279  llvm::errs() << "    " << MaxIncludeStackDepth << " max include stack depth\n";
280  llvm::errs() << "  " << NumIf << " #if/#ifndef/#ifdef.\n";
281  llvm::errs() << "  " << NumElse << " #else/#elif.\n";
282  llvm::errs() << "  " << NumEndif << " #endif.\n";
283  llvm::errs() << "  " << NumPragma << " #pragma.\n";
284  llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
285
286  llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
287             << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
288             << NumFastMacroExpanded << " on the fast path.\n";
289  llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
290             << " token paste (##) operations performed, "
291             << NumFastTokenPaste << " on the fast path.\n";
292
293  llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
294
295  llvm::errs() << "\n  BumpPtr: " << BP.getTotalMemory();
296  llvm::errs() << "\n  Macro Expanded Tokens: "
297               << llvm::capacity_in_bytes(MacroExpandedTokens);
298  llvm::errs() << "\n  Predefines Buffer: " << Predefines.capacity();
299  // FIXME: List information for all submodules.
300  llvm::errs() << "\n  Macros: "
301               << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
302  llvm::errs() << "\n  #pragma push_macro Info: "
303               << llvm::capacity_in_bytes(PragmaPushMacroInfo);
304  llvm::errs() << "\n  Poison Reasons: "
305               << llvm::capacity_in_bytes(PoisonReasons);
306  llvm::errs() << "\n  Comment Handlers: "
307               << llvm::capacity_in_bytes(CommentHandlers) << "\n";
308}
309
310Preprocessor::macro_iterator
311Preprocessor::macro_begin(bool IncludeExternalMacros) const {
312  if (IncludeExternalMacros && ExternalSource &&
313      !ReadMacrosFromExternalSource) {
314    ReadMacrosFromExternalSource = true;
315    ExternalSource->ReadDefinedMacros();
316  }
317
318  // Make sure we cover all macros in visible modules.
319  for (const ModuleMacro &Macro : ModuleMacros)
320    CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
321
322  return CurSubmoduleState->Macros.begin();
323}
324
325size_t Preprocessor::getTotalMemory() const {
326  return BP.getTotalMemory()
327    + llvm::capacity_in_bytes(MacroExpandedTokens)
328    + Predefines.capacity() /* Predefines buffer. */
329    // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
330    // and ModuleMacros.
331    + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
332    + llvm::capacity_in_bytes(PragmaPushMacroInfo)
333    + llvm::capacity_in_bytes(PoisonReasons)
334    + llvm::capacity_in_bytes(CommentHandlers);
335}
336
337Preprocessor::macro_iterator
338Preprocessor::macro_end(bool IncludeExternalMacros) const {
339  if (IncludeExternalMacros && ExternalSource &&
340      !ReadMacrosFromExternalSource) {
341    ReadMacrosFromExternalSource = true;
342    ExternalSource->ReadDefinedMacros();
343  }
344
345  return CurSubmoduleState->Macros.end();
346}
347
348/// Compares macro tokens with a specified token value sequence.
349static bool MacroDefinitionEquals(const MacroInfo *MI,
350                                  ArrayRef<TokenValue> Tokens) {
351  return Tokens.size() == MI->getNumTokens() &&
352      std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
353}
354
355StringRef Preprocessor::getLastMacroWithSpelling(
356                                    SourceLocation Loc,
357                                    ArrayRef<TokenValue> Tokens) const {
358  SourceLocation BestLocation;
359  StringRef BestSpelling;
360  for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
361       I != E; ++I) {
362    const MacroDirective::DefInfo
363      Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
364    if (!Def || !Def.getMacroInfo())
365      continue;
366    if (!Def.getMacroInfo()->isObjectLike())
367      continue;
368    if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
369      continue;
370    SourceLocation Location = Def.getLocation();
371    // Choose the macro defined latest.
372    if (BestLocation.isInvalid() ||
373        (Location.isValid() &&
374         SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
375      BestLocation = Location;
376      BestSpelling = I->first->getName();
377    }
378  }
379  return BestSpelling;
380}
381
382void Preprocessor::recomputeCurLexerKind() {
383  if (CurLexer)
384    CurLexerKind = CLK_Lexer;
385  else if (CurTokenLexer)
386    CurLexerKind = CLK_TokenLexer;
387  else
388    CurLexerKind = CLK_CachingLexer;
389}
390
391bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
392                                          unsigned CompleteLine,
393                                          unsigned CompleteColumn) {
394  assert(File);
395  assert(CompleteLine && CompleteColumn && "Starts from 1:1");
396  assert(!CodeCompletionFile && "Already set");
397
398  using llvm::MemoryBuffer;
399
400  // Load the actual file's contents.
401  bool Invalid = false;
402  const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
403  if (Invalid)
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
449void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
450                                            bool IsAngled) {
451  if (CodeComplete)
452    CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
453  setCodeCompletionReached();
454}
455
456void Preprocessor::CodeCompleteNaturalLanguage() {
457  if (CodeComplete)
458    CodeComplete->CodeCompleteNaturalLanguage();
459  setCodeCompletionReached();
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.
465StringRef 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.
487void 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
507SourceLocation 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
523Module *Preprocessor::getCurrentModule() {
524  if (!getLangOpts().isCompilingModule())
525    return nullptr;
526
527  return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
528}
529
530//===----------------------------------------------------------------------===//
531// Preprocessor Initialization Methods
532//===----------------------------------------------------------------------===//
533
534/// EnterMainSourceFile - Enter the specified FileID as the main source file,
535/// which implicitly adds the builtin defines etc.
536void Preprocessor::EnterMainSourceFile() {
537  // We do not allow the preprocessor to reenter the main file.  Doing so will
538  // cause FileID's to accumulate information from both runs (e.g. #line
539  // information) and predefined macros aren't guaranteed to be set properly.
540  assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
541  FileID MainFileID = SourceMgr.getMainFileID();
542
543  // If MainFileID is loaded it means we loaded an AST file, no need to enter
544  // a main file.
545  if (!SourceMgr.isLoadedFileID(MainFileID)) {
546    // Enter the main file source buffer.
547    EnterSourceFile(MainFileID, nullptr, SourceLocation());
548
549    // If we've been asked to skip bytes in the main file (e.g., as part of a
550    // precompiled preamble), do so now.
551    if (SkipMainFilePreamble.first > 0)
552      CurLexer->SetByteOffset(SkipMainFilePreamble.first,
553                              SkipMainFilePreamble.second);
554
555    // Tell the header info that the main file was entered.  If the file is later
556    // #imported, it won't be re-entered.
557    if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
558      HeaderInfo.IncrementIncludeCount(FE);
559  }
560
561  // Preprocess Predefines to populate the initial preprocessor state.
562  std::unique_ptr<llvm::MemoryBuffer> SB =
563    llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
564  assert(SB && "Cannot create predefined source buffer");
565  FileID FID = SourceMgr.createFileID(std::move(SB));
566  assert(FID.isValid() && "Could not create FileID for predefines?");
567  setPredefinesFileID(FID);
568
569  // Start parsing the predefines.
570  EnterSourceFile(FID, nullptr, SourceLocation());
571
572  if (!PPOpts->PCHThroughHeader.empty()) {
573    // Lookup and save the FileID for the through header. If it isn't found
574    // in the search path, it's a fatal error.
575    const DirectoryLookup *CurDir;
576    Optional<FileEntryRef> File = LookupFile(
577        SourceLocation(), PPOpts->PCHThroughHeader,
578        /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, CurDir,
579        /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
580        /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr,
581        /*IsFrameworkFound=*/nullptr);
582    if (!File) {
583      Diag(SourceLocation(), diag::err_pp_through_header_not_found)
584          << PPOpts->PCHThroughHeader;
585      return;
586    }
587    setPCHThroughHeaderFileID(
588        SourceMgr.createFileID(*File, SourceLocation(), SrcMgr::C_User));
589  }
590
591  // Skip tokens from the Predefines and if needed the main file.
592  if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
593      (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
594    SkipTokensWhileUsingPCH();
595}
596
597void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
598  assert(PCHThroughHeaderFileID.isInvalid() &&
599         "PCHThroughHeaderFileID already set!");
600  PCHThroughHeaderFileID = FID;
601}
602
603bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
604  assert(PCHThroughHeaderFileID.isValid() &&
605         "Invalid PCH through header FileID");
606  return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
607}
608
609bool Preprocessor::creatingPCHWithThroughHeader() {
610  return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
611         PCHThroughHeaderFileID.isValid();
612}
613
614bool Preprocessor::usingPCHWithThroughHeader() {
615  return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
616         PCHThroughHeaderFileID.isValid();
617}
618
619bool Preprocessor::creatingPCHWithPragmaHdrStop() {
620  return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
621}
622
623bool Preprocessor::usingPCHWithPragmaHdrStop() {
624  return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
625}
626
627/// Skip tokens until after the #include of the through header or
628/// until after a #pragma hdrstop is seen. Tokens in the predefines file
629/// and the main file may be skipped. If the end of the predefines file
630/// is reached, skipping continues into the main file. If the end of the
631/// main file is reached, it's a fatal error.
632void Preprocessor::SkipTokensWhileUsingPCH() {
633  bool ReachedMainFileEOF = false;
634  bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
635  bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
636  Token Tok;
637  while (true) {
638    bool InPredefines =
639        (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
640    switch (CurLexerKind) {
641    case CLK_Lexer:
642      CurLexer->Lex(Tok);
643     break;
644    case CLK_TokenLexer:
645      CurTokenLexer->Lex(Tok);
646      break;
647    case CLK_CachingLexer:
648      CachingLex(Tok);
649      break;
650    case CLK_LexAfterModuleImport:
651      LexAfterModuleImport(Tok);
652      break;
653    }
654    if (Tok.is(tok::eof) && !InPredefines) {
655      ReachedMainFileEOF = true;
656      break;
657    }
658    if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
659      break;
660    if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
661      break;
662  }
663  if (ReachedMainFileEOF) {
664    if (UsingPCHThroughHeader)
665      Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
666          << PPOpts->PCHThroughHeader << 1;
667    else if (!PPOpts->PCHWithHdrStopCreate)
668      Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
669  }
670}
671
672void Preprocessor::replayPreambleConditionalStack() {
673  // Restore the conditional stack from the preamble, if there is one.
674  if (PreambleConditionalStack.isReplaying()) {
675    assert(CurPPLexer &&
676           "CurPPLexer is null when calling replayPreambleConditionalStack.");
677    CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
678    PreambleConditionalStack.doneReplaying();
679    if (PreambleConditionalStack.reachedEOFWhileSkipping())
680      SkipExcludedConditionalBlock(
681          PreambleConditionalStack.SkipInfo->HashTokenLoc,
682          PreambleConditionalStack.SkipInfo->IfTokenLoc,
683          PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
684          PreambleConditionalStack.SkipInfo->FoundElse,
685          PreambleConditionalStack.SkipInfo->ElseLoc);
686  }
687}
688
689void Preprocessor::EndSourceFile() {
690  // Notify the client that we reached the end of the source file.
691  if (Callbacks)
692    Callbacks->EndOfMainFile();
693}
694
695//===----------------------------------------------------------------------===//
696// Lexer Event Handling.
697//===----------------------------------------------------------------------===//
698
699/// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
700/// identifier information for the token and install it into the token,
701/// updating the token kind accordingly.
702IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
703  assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
704
705  // Look up this token, see if it is a macro, or if it is a language keyword.
706  IdentifierInfo *II;
707  if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
708    // No cleaning needed, just use the characters from the lexed buffer.
709    II = getIdentifierInfo(Identifier.getRawIdentifier());
710  } else {
711    // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
712    SmallString<64> IdentifierBuffer;
713    StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
714
715    if (Identifier.hasUCN()) {
716      SmallString<64> UCNIdentifierBuffer;
717      expandUCNs(UCNIdentifierBuffer, CleanedStr);
718      II = getIdentifierInfo(UCNIdentifierBuffer);
719    } else {
720      II = getIdentifierInfo(CleanedStr);
721    }
722  }
723
724  // Update the token info (identifier info and appropriate token kind).
725  Identifier.setIdentifierInfo(II);
726  if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
727      getSourceManager().isInSystemHeader(Identifier.getLocation()))
728    Identifier.setKind(tok::identifier);
729  else
730    Identifier.setKind(II->getTokenID());
731
732  return II;
733}
734
735void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
736  PoisonReasons[II] = DiagID;
737}
738
739void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
740  assert(Ident__exception_code && Ident__exception_info);
741  assert(Ident___exception_code && Ident___exception_info);
742  Ident__exception_code->setIsPoisoned(Poison);
743  Ident___exception_code->setIsPoisoned(Poison);
744  Ident_GetExceptionCode->setIsPoisoned(Poison);
745  Ident__exception_info->setIsPoisoned(Poison);
746  Ident___exception_info->setIsPoisoned(Poison);
747  Ident_GetExceptionInfo->setIsPoisoned(Poison);
748  Ident__abnormal_termination->setIsPoisoned(Poison);
749  Ident___abnormal_termination->setIsPoisoned(Poison);
750  Ident_AbnormalTermination->setIsPoisoned(Poison);
751}
752
753void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
754  assert(Identifier.getIdentifierInfo() &&
755         "Can't handle identifiers without identifier info!");
756  llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
757    PoisonReasons.find(Identifier.getIdentifierInfo());
758  if(it == PoisonReasons.end())
759    Diag(Identifier, diag::err_pp_used_poisoned_id);
760  else
761    Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
762}
763
764/// Returns a diagnostic message kind for reporting a future keyword as
765/// appropriate for the identifier and specified language.
766static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
767                                          const LangOptions &LangOpts) {
768  assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
769
770  if (LangOpts.CPlusPlus)
771    return llvm::StringSwitch<diag::kind>(II.getName())
772#define CXX11_KEYWORD(NAME, FLAGS)                                             \
773        .Case(#NAME, diag::warn_cxx11_keyword)
774#define CXX20_KEYWORD(NAME, FLAGS)                                             \
775        .Case(#NAME, diag::warn_cxx20_keyword)
776#include "clang/Basic/TokenKinds.def"
777        // char8_t is not modeled as a CXX20_KEYWORD because it's not
778        // unconditionally enabled in C++20 mode. (It can be disabled
779        // by -fno-char8_t.)
780        .Case("char8_t", diag::warn_cxx20_keyword)
781        ;
782
783  llvm_unreachable(
784      "Keyword not known to come from a newer Standard or proposed Standard");
785}
786
787void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
788  assert(II.isOutOfDate() && "not out of date");
789  getExternalSource()->updateOutOfDateIdentifier(II);
790}
791
792/// HandleIdentifier - This callback is invoked when the lexer reads an
793/// identifier.  This callback looks up the identifier in the map and/or
794/// potentially macro expands it or turns it into a named token (like 'for').
795///
796/// Note that callers of this method are guarded by checking the
797/// IdentifierInfo's 'isHandleIdentifierCase' bit.  If this method changes, the
798/// IdentifierInfo methods that compute these properties will need to change to
799/// match.
800bool Preprocessor::HandleIdentifier(Token &Identifier) {
801  assert(Identifier.getIdentifierInfo() &&
802         "Can't handle identifiers without identifier info!");
803
804  IdentifierInfo &II = *Identifier.getIdentifierInfo();
805
806  // If the information about this identifier is out of date, update it from
807  // the external source.
808  // We have to treat __VA_ARGS__ in a special way, since it gets
809  // serialized with isPoisoned = true, but our preprocessor may have
810  // unpoisoned it if we're defining a C99 macro.
811  if (II.isOutOfDate()) {
812    bool CurrentIsPoisoned = false;
813    const bool IsSpecialVariadicMacro =
814        &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
815    if (IsSpecialVariadicMacro)
816      CurrentIsPoisoned = II.isPoisoned();
817
818    updateOutOfDateIdentifier(II);
819    Identifier.setKind(II.getTokenID());
820
821    if (IsSpecialVariadicMacro)
822      II.setIsPoisoned(CurrentIsPoisoned);
823  }
824
825  // If this identifier was poisoned, and if it was not produced from a macro
826  // expansion, emit an error.
827  if (II.isPoisoned() && CurPPLexer) {
828    HandlePoisonedIdentifier(Identifier);
829  }
830
831  // If this is a macro to be expanded, do it.
832  if (MacroDefinition MD = getMacroDefinition(&II)) {
833    auto *MI = MD.getMacroInfo();
834    assert(MI && "macro definition with no macro info?");
835    if (!DisableMacroExpansion) {
836      if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
837        // C99 6.10.3p10: If the preprocessing token immediately after the
838        // macro name isn't a '(', this macro should not be expanded.
839        if (!MI->isFunctionLike() || isNextPPTokenLParen())
840          return HandleMacroExpandedIdentifier(Identifier, MD);
841      } else {
842        // C99 6.10.3.4p2 says that a disabled macro may never again be
843        // expanded, even if it's in a context where it could be expanded in the
844        // future.
845        Identifier.setFlag(Token::DisableExpand);
846        if (MI->isObjectLike() || isNextPPTokenLParen())
847          Diag(Identifier, diag::pp_disabled_macro_expansion);
848      }
849    }
850  }
851
852  // If this identifier is a keyword in a newer Standard or proposed Standard,
853  // produce a warning. Don't warn if we're not considering macro expansion,
854  // since this identifier might be the name of a macro.
855  // FIXME: This warning is disabled in cases where it shouldn't be, like
856  //   "#define constexpr constexpr", "int constexpr;"
857  if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
858    Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
859        << II.getName();
860    // Don't diagnose this keyword again in this translation unit.
861    II.setIsFutureCompatKeyword(false);
862  }
863
864  // If this is an extension token, diagnose its use.
865  // We avoid diagnosing tokens that originate from macro definitions.
866  // FIXME: This warning is disabled in cases where it shouldn't be,
867  // like "#define TY typeof", "TY(1) x".
868  if (II.isExtensionToken() && !DisableMacroExpansion)
869    Diag(Identifier, diag::ext_token_used);
870
871  // If this is the 'import' contextual keyword following an '@', note
872  // that the next token indicates a module name.
873  //
874  // Note that we do not treat 'import' as a contextual
875  // keyword when we're in a caching lexer, because caching lexers only get
876  // used in contexts where import declarations are disallowed.
877  //
878  // Likewise if this is the C++ Modules TS import keyword.
879  if (((LastTokenWasAt && II.isModulesImport()) ||
880       Identifier.is(tok::kw_import)) &&
881      !InMacroArgs && !DisableMacroExpansion &&
882      (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
883      CurLexerKind != CLK_CachingLexer) {
884    ModuleImportLoc = Identifier.getLocation();
885    ModuleImportPath.clear();
886    ModuleImportExpectsIdentifier = true;
887    CurLexerKind = CLK_LexAfterModuleImport;
888  }
889  return true;
890}
891
892void Preprocessor::Lex(Token &Result) {
893  ++LexLevel;
894
895  // We loop here until a lex function returns a token; this avoids recursion.
896  bool ReturnedToken;
897  do {
898    switch (CurLexerKind) {
899    case CLK_Lexer:
900      ReturnedToken = CurLexer->Lex(Result);
901      break;
902    case CLK_TokenLexer:
903      ReturnedToken = CurTokenLexer->Lex(Result);
904      break;
905    case CLK_CachingLexer:
906      CachingLex(Result);
907      ReturnedToken = true;
908      break;
909    case CLK_LexAfterModuleImport:
910      ReturnedToken = LexAfterModuleImport(Result);
911      break;
912    }
913  } while (!ReturnedToken);
914
915  if (Result.is(tok::unknown) && TheModuleLoader.HadFatalFailure)
916    return;
917
918  if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
919    // Remember the identifier before code completion token.
920    setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
921    setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
922    // Set IdenfitierInfo to null to avoid confusing code that handles both
923    // identifiers and completion tokens.
924    Result.setIdentifierInfo(nullptr);
925  }
926
927  // Update ImportSeqState to track our position within a C++20 import-seq
928  // if this token is being produced as a result of phase 4 of translation.
929  if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
930      !Result.getFlag(Token::IsReinjected)) {
931    switch (Result.getKind()) {
932    case tok::l_paren: case tok::l_square: case tok::l_brace:
933      ImportSeqState.handleOpenBracket();
934      break;
935    case tok::r_paren: case tok::r_square:
936      ImportSeqState.handleCloseBracket();
937      break;
938    case tok::r_brace:
939      ImportSeqState.handleCloseBrace();
940      break;
941    case tok::semi:
942      ImportSeqState.handleSemi();
943      break;
944    case tok::header_name:
945    case tok::annot_header_unit:
946      ImportSeqState.handleHeaderName();
947      break;
948    case tok::kw_export:
949      ImportSeqState.handleExport();
950      break;
951    case tok::identifier:
952      if (Result.getIdentifierInfo()->isModulesImport()) {
953        ImportSeqState.handleImport();
954        if (ImportSeqState.afterImportSeq()) {
955          ModuleImportLoc = Result.getLocation();
956          ModuleImportPath.clear();
957          ModuleImportExpectsIdentifier = true;
958          CurLexerKind = CLK_LexAfterModuleImport;
959        }
960        break;
961      }
962      LLVM_FALLTHROUGH;
963    default:
964      ImportSeqState.handleMisc();
965      break;
966    }
967  }
968
969  LastTokenWasAt = Result.is(tok::at);
970  --LexLevel;
971
972  if (LexLevel == 0 && !Result.getFlag(Token::IsReinjected)) {
973    ++TokenCount;
974    if (OnToken)
975      OnToken(Result);
976  }
977}
978
979/// Lex a header-name token (including one formed from header-name-tokens if
980/// \p AllowConcatenation is \c true).
981///
982/// \param FilenameTok Filled in with the next token. On success, this will
983///        be either a header_name token. On failure, it will be whatever other
984///        token was found instead.
985/// \param AllowMacroExpansion If \c true, allow the header name to be formed
986///        by macro expansion (concatenating tokens as necessary if the first
987///        token is a '<').
988/// \return \c true if we reached EOD or EOF while looking for a > token in
989///         a concatenated header name and diagnosed it. \c false otherwise.
990bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
991  // Lex using header-name tokenization rules if tokens are being lexed from
992  // a file. Just grab a token normally if we're in a macro expansion.
993  if (CurPPLexer)
994    CurPPLexer->LexIncludeFilename(FilenameTok);
995  else
996    Lex(FilenameTok);
997
998  // This could be a <foo/bar.h> file coming from a macro expansion.  In this
999  // case, glue the tokens together into an angle_string_literal token.
1000  SmallString<128> FilenameBuffer;
1001  if (FilenameTok.is(tok::less) && AllowMacroExpansion) {
1002    bool StartOfLine = FilenameTok.isAtStartOfLine();
1003    bool LeadingSpace = FilenameTok.hasLeadingSpace();
1004    bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
1005
1006    SourceLocation Start = FilenameTok.getLocation();
1007    SourceLocation End;
1008    FilenameBuffer.push_back('<');
1009
1010    // Consume tokens until we find a '>'.
1011    // FIXME: A header-name could be formed starting or ending with an
1012    // alternative token. It's not clear whether that's ill-formed in all
1013    // cases.
1014    while (FilenameTok.isNot(tok::greater)) {
1015      Lex(FilenameTok);
1016      if (FilenameTok.isOneOf(tok::eod, tok::eof)) {
1017        Diag(FilenameTok.getLocation(), diag::err_expected) << tok::greater;
1018        Diag(Start, diag::note_matching) << tok::less;
1019        return true;
1020      }
1021
1022      End = FilenameTok.getLocation();
1023
1024      // FIXME: Provide code completion for #includes.
1025      if (FilenameTok.is(tok::code_completion)) {
1026        setCodeCompletionReached();
1027        Lex(FilenameTok);
1028        continue;
1029      }
1030
1031      // Append the spelling of this token to the buffer. If there was a space
1032      // before it, add it now.
1033      if (FilenameTok.hasLeadingSpace())
1034        FilenameBuffer.push_back(' ');
1035
1036      // Get the spelling of the token, directly into FilenameBuffer if
1037      // possible.
1038      size_t PreAppendSize = FilenameBuffer.size();
1039      FilenameBuffer.resize(PreAppendSize + FilenameTok.getLength());
1040
1041      const char *BufPtr = &FilenameBuffer[PreAppendSize];
1042      unsigned ActualLen = getSpelling(FilenameTok, BufPtr);
1043
1044      // If the token was spelled somewhere else, copy it into FilenameBuffer.
1045      if (BufPtr != &FilenameBuffer[PreAppendSize])
1046        memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1047
1048      // Resize FilenameBuffer to the correct size.
1049      if (FilenameTok.getLength() != ActualLen)
1050        FilenameBuffer.resize(PreAppendSize + ActualLen);
1051    }
1052
1053    FilenameTok.startToken();
1054    FilenameTok.setKind(tok::header_name);
1055    FilenameTok.setFlagValue(Token::StartOfLine, StartOfLine);
1056    FilenameTok.setFlagValue(Token::LeadingSpace, LeadingSpace);
1057    FilenameTok.setFlagValue(Token::LeadingEmptyMacro, LeadingEmptyMacro);
1058    CreateString(FilenameBuffer, FilenameTok, Start, End);
1059  } else if (FilenameTok.is(tok::string_literal) && AllowMacroExpansion) {
1060    // Convert a string-literal token of the form " h-char-sequence "
1061    // (produced by macro expansion) into a header-name token.
1062    //
1063    // The rules for header-names don't quite match the rules for
1064    // string-literals, but all the places where they differ result in
1065    // undefined behavior, so we can and do treat them the same.
1066    //
1067    // A string-literal with a prefix or suffix is not translated into a
1068    // header-name. This could theoretically be observable via the C++20
1069    // context-sensitive header-name formation rules.
1070    StringRef Str = getSpelling(FilenameTok, FilenameBuffer);
1071    if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1072      FilenameTok.setKind(tok::header_name);
1073  }
1074
1075  return false;
1076}
1077
1078/// Collect the tokens of a C++20 pp-import-suffix.
1079void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) {
1080  // FIXME: For error recovery, consider recognizing attribute syntax here
1081  // and terminating / diagnosing a missing semicolon if we find anything
1082  // else? (Can we leave that to the parser?)
1083  unsigned BracketDepth = 0;
1084  while (true) {
1085    Toks.emplace_back();
1086    Lex(Toks.back());
1087
1088    switch (Toks.back().getKind()) {
1089    case tok::l_paren: case tok::l_square: case tok::l_brace:
1090      ++BracketDepth;
1091      break;
1092
1093    case tok::r_paren: case tok::r_square: case tok::r_brace:
1094      if (BracketDepth == 0)
1095        return;
1096      --BracketDepth;
1097      break;
1098
1099    case tok::semi:
1100      if (BracketDepth == 0)
1101        return;
1102    break;
1103
1104    case tok::eof:
1105      return;
1106
1107    default:
1108      break;
1109    }
1110  }
1111}
1112
1113
1114/// Lex a token following the 'import' contextual keyword.
1115///
1116///     pp-import: [C++20]
1117///           import header-name pp-import-suffix[opt] ;
1118///           import header-name-tokens pp-import-suffix[opt] ;
1119/// [ObjC]    @ import module-name ;
1120/// [Clang]   import module-name ;
1121///
1122///     header-name-tokens:
1123///           string-literal
1124///           < [any sequence of preprocessing-tokens other than >] >
1125///
1126///     module-name:
1127///           module-name-qualifier[opt] identifier
1128///
1129///     module-name-qualifier
1130///           module-name-qualifier[opt] identifier .
1131///
1132/// We respond to a pp-import by importing macros from the named module.
1133bool Preprocessor::LexAfterModuleImport(Token &Result) {
1134  // Figure out what kind of lexer we actually have.
1135  recomputeCurLexerKind();
1136
1137  // Lex the next token. The header-name lexing rules are used at the start of
1138  // a pp-import.
1139  //
1140  // For now, we only support header-name imports in C++20 mode.
1141  // FIXME: Should we allow this in all language modes that support an import
1142  // declaration as an extension?
1143  if (ModuleImportPath.empty() && getLangOpts().CPlusPlusModules) {
1144    if (LexHeaderName(Result))
1145      return true;
1146  } else {
1147    Lex(Result);
1148  }
1149
1150  // Allocate a holding buffer for a sequence of tokens and introduce it into
1151  // the token stream.
1152  auto EnterTokens = [this](ArrayRef<Token> Toks) {
1153    auto ToksCopy = std::make_unique<Token[]>(Toks.size());
1154    std::copy(Toks.begin(), Toks.end(), ToksCopy.get());
1155    EnterTokenStream(std::move(ToksCopy), Toks.size(),
1156                     /*DisableMacroExpansion*/ true, /*IsReinject*/ false);
1157  };
1158
1159  // Check for a header-name.
1160  SmallVector<Token, 32> Suffix;
1161  if (Result.is(tok::header_name)) {
1162    // Enter the header-name token into the token stream; a Lex action cannot
1163    // both return a token and cache tokens (doing so would corrupt the token
1164    // cache if the call to Lex comes from CachingLex / PeekAhead).
1165    Suffix.push_back(Result);
1166
1167    // Consume the pp-import-suffix and expand any macros in it now. We'll add
1168    // it back into the token stream later.
1169    CollectPpImportSuffix(Suffix);
1170    if (Suffix.back().isNot(tok::semi)) {
1171      // This is not a pp-import after all.
1172      EnterTokens(Suffix);
1173      return false;
1174    }
1175
1176    // C++2a [cpp.module]p1:
1177    //   The ';' preprocessing-token terminating a pp-import shall not have
1178    //   been produced by macro replacement.
1179    SourceLocation SemiLoc = Suffix.back().getLocation();
1180    if (SemiLoc.isMacroID())
1181      Diag(SemiLoc, diag::err_header_import_semi_in_macro);
1182
1183    // Reconstitute the import token.
1184    Token ImportTok;
1185    ImportTok.startToken();
1186    ImportTok.setKind(tok::kw_import);
1187    ImportTok.setLocation(ModuleImportLoc);
1188    ImportTok.setIdentifierInfo(getIdentifierInfo("import"));
1189    ImportTok.setLength(6);
1190
1191    auto Action = HandleHeaderIncludeOrImport(
1192        /*HashLoc*/ SourceLocation(), ImportTok, Suffix.front(), SemiLoc);
1193    switch (Action.Kind) {
1194    case ImportAction::None:
1195      break;
1196
1197    case ImportAction::ModuleBegin:
1198      // Let the parser know we're textually entering the module.
1199      Suffix.emplace_back();
1200      Suffix.back().startToken();
1201      Suffix.back().setKind(tok::annot_module_begin);
1202      Suffix.back().setLocation(SemiLoc);
1203      Suffix.back().setAnnotationEndLoc(SemiLoc);
1204      Suffix.back().setAnnotationValue(Action.ModuleForHeader);
1205      LLVM_FALLTHROUGH;
1206
1207    case ImportAction::ModuleImport:
1208    case ImportAction::SkippedModuleImport:
1209      // We chose to import (or textually enter) the file. Convert the
1210      // header-name token into a header unit annotation token.
1211      Suffix[0].setKind(tok::annot_header_unit);
1212      Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation());
1213      Suffix[0].setAnnotationValue(Action.ModuleForHeader);
1214      // FIXME: Call the moduleImport callback?
1215      break;
1216    case ImportAction::Failure:
1217      assert(TheModuleLoader.HadFatalFailure &&
1218             "This should be an early exit only to a fatal error");
1219      Result.setKind(tok::eof);
1220      CurLexer->cutOffLexing();
1221      EnterTokens(Suffix);
1222      return true;
1223    }
1224
1225    EnterTokens(Suffix);
1226    return false;
1227  }
1228
1229  // The token sequence
1230  //
1231  //   import identifier (. identifier)*
1232  //
1233  // indicates a module import directive. We already saw the 'import'
1234  // contextual keyword, so now we're looking for the identifiers.
1235  if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
1236    // We expected to see an identifier here, and we did; continue handling
1237    // identifiers.
1238    ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
1239                                              Result.getLocation()));
1240    ModuleImportExpectsIdentifier = false;
1241    CurLexerKind = CLK_LexAfterModuleImport;
1242    return true;
1243  }
1244
1245  // If we're expecting a '.' or a ';', and we got a '.', then wait until we
1246  // see the next identifier. (We can also see a '[[' that begins an
1247  // attribute-specifier-seq here under the C++ Modules TS.)
1248  if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
1249    ModuleImportExpectsIdentifier = true;
1250    CurLexerKind = CLK_LexAfterModuleImport;
1251    return true;
1252  }
1253
1254  // If we didn't recognize a module name at all, this is not a (valid) import.
1255  if (ModuleImportPath.empty() || Result.is(tok::eof))
1256    return true;
1257
1258  // Consume the pp-import-suffix and expand any macros in it now, if we're not
1259  // at the semicolon already.
1260  SourceLocation SemiLoc = Result.getLocation();
1261  if (Result.isNot(tok::semi)) {
1262    Suffix.push_back(Result);
1263    CollectPpImportSuffix(Suffix);
1264    if (Suffix.back().isNot(tok::semi)) {
1265      // This is not an import after all.
1266      EnterTokens(Suffix);
1267      return false;
1268    }
1269    SemiLoc = Suffix.back().getLocation();
1270  }
1271
1272  // Under the Modules TS, the dot is just part of the module name, and not
1273  // a real hierarchy separator. Flatten such module names now.
1274  //
1275  // FIXME: Is this the right level to be performing this transformation?
1276  std::string FlatModuleName;
1277  if (getLangOpts().ModulesTS || getLangOpts().CPlusPlusModules) {
1278    for (auto &Piece : ModuleImportPath) {
1279      if (!FlatModuleName.empty())
1280        FlatModuleName += ".";
1281      FlatModuleName += Piece.first->getName();
1282    }
1283    SourceLocation FirstPathLoc = ModuleImportPath[0].second;
1284    ModuleImportPath.clear();
1285    ModuleImportPath.push_back(
1286        std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
1287  }
1288
1289  Module *Imported = nullptr;
1290  if (getLangOpts().Modules) {
1291    Imported = TheModuleLoader.loadModule(ModuleImportLoc,
1292                                          ModuleImportPath,
1293                                          Module::Hidden,
1294                                          /*IsInclusionDirective=*/false);
1295    if (Imported)
1296      makeModuleVisible(Imported, SemiLoc);
1297  }
1298  if (Callbacks)
1299    Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
1300
1301  if (!Suffix.empty()) {
1302    EnterTokens(Suffix);
1303    return false;
1304  }
1305  return true;
1306}
1307
1308void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
1309  CurSubmoduleState->VisibleModules.setVisible(
1310      M, Loc, [](Module *) {},
1311      [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1312        // FIXME: Include the path in the diagnostic.
1313        // FIXME: Include the import location for the conflicting module.
1314        Diag(ModuleImportLoc, diag::warn_module_conflict)
1315            << Path[0]->getFullModuleName()
1316            << Conflict->getFullModuleName()
1317            << Message;
1318      });
1319
1320  // Add this module to the imports list of the currently-built submodule.
1321  if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1322    BuildingSubmoduleStack.back().M->Imports.insert(M);
1323}
1324
1325bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1326                                          const char *DiagnosticTag,
1327                                          bool AllowMacroExpansion) {
1328  // We need at least one string literal.
1329  if (Result.isNot(tok::string_literal)) {
1330    Diag(Result, diag::err_expected_string_literal)
1331      << /*Source='in...'*/0 << DiagnosticTag;
1332    return false;
1333  }
1334
1335  // Lex string literal tokens, optionally with macro expansion.
1336  SmallVector<Token, 4> StrToks;
1337  do {
1338    StrToks.push_back(Result);
1339
1340    if (Result.hasUDSuffix())
1341      Diag(Result, diag::err_invalid_string_udl);
1342
1343    if (AllowMacroExpansion)
1344      Lex(Result);
1345    else
1346      LexUnexpandedToken(Result);
1347  } while (Result.is(tok::string_literal));
1348
1349  // Concatenate and parse the strings.
1350  StringLiteralParser Literal(StrToks, *this);
1351  assert(Literal.isAscii() && "Didn't allow wide strings in");
1352
1353  if (Literal.hadError)
1354    return false;
1355
1356  if (Literal.Pascal) {
1357    Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1358      << /*Source='in...'*/0 << DiagnosticTag;
1359    return false;
1360  }
1361
1362  String = std::string(Literal.GetString());
1363  return true;
1364}
1365
1366bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1367  assert(Tok.is(tok::numeric_constant));
1368  SmallString<8> IntegerBuffer;
1369  bool NumberInvalid = false;
1370  StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1371  if (NumberInvalid)
1372    return false;
1373  NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(),
1374                               getLangOpts(), getTargetInfo(),
1375                               getDiagnostics());
1376  if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1377    return false;
1378  llvm::APInt APVal(64, 0);
1379  if (Literal.GetIntegerValue(APVal))
1380    return false;
1381  Lex(Tok);
1382  Value = APVal.getLimitedValue();
1383  return true;
1384}
1385
1386void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1387  assert(Handler && "NULL comment handler");
1388  assert(llvm::find(CommentHandlers, Handler) == CommentHandlers.end() &&
1389         "Comment handler already registered");
1390  CommentHandlers.push_back(Handler);
1391}
1392
1393void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1394  std::vector<CommentHandler *>::iterator Pos =
1395      llvm::find(CommentHandlers, Handler);
1396  assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1397  CommentHandlers.erase(Pos);
1398}
1399
1400bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1401  bool AnyPendingTokens = false;
1402  for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1403       HEnd = CommentHandlers.end();
1404       H != HEnd; ++H) {
1405    if ((*H)->HandleComment(*this, Comment))
1406      AnyPendingTokens = true;
1407  }
1408  if (!AnyPendingTokens || getCommentRetentionState())
1409    return false;
1410  Lex(result);
1411  return true;
1412}
1413
1414ModuleLoader::~ModuleLoader() = default;
1415
1416CommentHandler::~CommentHandler() = default;
1417
1418CodeCompletionHandler::~CodeCompletionHandler() = default;
1419
1420void Preprocessor::createPreprocessingRecord() {
1421  if (Record)
1422    return;
1423
1424  Record = new PreprocessingRecord(getSourceManager());
1425  addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));
1426}
1427