Home | History | Annotate | Download | only in Lex 
      1 //===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===//
      2 //
      3 //                     The LLVM Compiler Infrastructure
      4 //
      5 // This file is distributed under the University of Illinois Open Source
      6 // License. See LICENSE.TXT for details.
      7 //
      8 //===----------------------------------------------------------------------===//
      9 //
     10 //  This file implements the Preprocessor interface.
     11 //
     12 //===----------------------------------------------------------------------===//
     13 //
     14 // Options to support:
     15 //   -H       - Print the name of each header file used.
     16 //   -d[DNI] - Dump various things.
     17 //   -fworking-directory - #line's with preprocessor's working dir.
     18 //   -fpreprocessed
     19 //   -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
     20 //   -W*
     21 //   -w
     22 //
     23 // Messages to emit:
     24 //   "Multiple include guards may be useful for:\n"
     25 //
     26 //===----------------------------------------------------------------------===//
     27 
     28 #include "clang/Lex/Preprocessor.h"
     29 #include "clang/Lex/MacroArgs.h"
     30 #include "clang/Basic/FileManager.h"
     31 #include "clang/Basic/SourceManager.h"
     32 #include "clang/Basic/TargetInfo.h"
     33 #include "clang/Lex/CodeCompletionHandler.h"
     34 #include "clang/Lex/ExternalPreprocessorSource.h"
     35 #include "clang/Lex/HeaderSearch.h"
     36 #include "clang/Lex/LexDiagnostic.h"
     37 #include "clang/Lex/LiteralSupport.h"
     38 #include "clang/Lex/MacroInfo.h"
     39 #include "clang/Lex/ModuleLoader.h"
     40 #include "clang/Lex/Pragma.h"
     41 #include "clang/Lex/PreprocessingRecord.h"
     42 #include "clang/Lex/PreprocessorOptions.h"
     43 #include "clang/Lex/ScratchBuffer.h"
     44 #include "llvm/ADT/APFloat.h"
     45 #include "llvm/ADT/SmallString.h"
     46 #include "llvm/ADT/STLExtras.h"
     47 #include "llvm/ADT/StringExtras.h"
     48 #include "llvm/Support/Capacity.h"
     49 #include "llvm/Support/ConvertUTF.h"
     50 #include "llvm/Support/MemoryBuffer.h"
     51 #include "llvm/Support/raw_ostream.h"
     52 using namespace clang;
     53 
     54 //===----------------------------------------------------------------------===//
     55 ExternalPreprocessorSource::~ExternalPreprocessorSource() { }
     56 
     57 Preprocessor::Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts,
     58                            DiagnosticsEngine &diags, LangOptions &opts,
     59                            const TargetInfo *target, SourceManager &SM,
     60                            HeaderSearch &Headers, ModuleLoader &TheModuleLoader,
     61                            IdentifierInfoLookup *IILookup, bool OwnsHeaders,
     62                            bool DelayInitialization, bool IncrProcessing)
     63     : PPOpts(PPOpts), Diags(&diags), LangOpts(opts), Target(target),
     64       FileMgr(Headers.getFileMgr()), SourceMgr(SM), HeaderInfo(Headers),
     65       TheModuleLoader(TheModuleLoader), ExternalSource(0),
     66       Identifiers(opts, IILookup), IncrementalProcessing(IncrProcessing),
     67       CodeComplete(0), CodeCompletionFile(0), CodeCompletionOffset(0),
     68       CodeCompletionReached(0), SkipMainFilePreamble(0, true), CurPPLexer(0),
     69       CurDirLookup(0), CurLexerKind(CLK_Lexer), Callbacks(0),
     70       MacroArgCache(0), Record(0), MIChainHead(0), MICache(0),
     71       DeserialMIChainHead(0) {
     72   OwnsHeaderSearch = OwnsHeaders;
     73 
     74   ScratchBuf = new ScratchBuffer(SourceMgr);
     75   CounterValue = 0; // __COUNTER__ starts at 0.
     76 
     77   // Clear stats.
     78   NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
     79   NumIf = NumElse = NumEndif = 0;
     80   NumEnteredSourceFiles = 0;
     81   NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
     82   NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
     83   MaxIncludeStackDepth = 0;
     84   NumSkipped = 0;
     85 
     86   // Default to discarding comments.
     87   KeepComments = false;
     88   KeepMacroComments = false;
     89   SuppressIncludeNotFoundError = false;
     90 
     91   // Macro expansion is enabled.
     92   DisableMacroExpansion = false;
     93   MacroExpansionInDirectivesOverride = false;
     94   InMacroArgs = false;
     95   InMacroArgPreExpansion = false;
     96   NumCachedTokenLexers = 0;
     97   PragmasEnabled = true;
     98   ParsingIfOrElifDirective = false;
     99   PreprocessedOutput = false;
    100 
    101   CachedLexPos = 0;
    102 
    103   // We haven't read anything from the external source.
    104   ReadMacrosFromExternalSource = false;
    105 
    106   // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
    107   // This gets unpoisoned where it is allowed.
    108   (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
    109   SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
    110 
    111   // Initialize the pragma handlers.
    112   PragmaHandlers = new PragmaNamespace(StringRef());
    113   RegisterBuiltinPragmas();
    114 
    115   // Initialize builtin macros like __LINE__ and friends.
    116   RegisterBuiltinMacros();
    117 
    118   if(LangOpts.Borland) {
    119     Ident__exception_info        = getIdentifierInfo("_exception_info");
    120     Ident___exception_info       = getIdentifierInfo("__exception_info");
    121     Ident_GetExceptionInfo       = getIdentifierInfo("GetExceptionInformation");
    122     Ident__exception_code        = getIdentifierInfo("_exception_code");
    123     Ident___exception_code       = getIdentifierInfo("__exception_code");
    124     Ident_GetExceptionCode       = getIdentifierInfo("GetExceptionCode");
    125     Ident__abnormal_termination  = getIdentifierInfo("_abnormal_termination");
    126     Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
    127     Ident_AbnormalTermination    = getIdentifierInfo("AbnormalTermination");
    128   } else {
    129     Ident__exception_info = Ident__exception_code = Ident__abnormal_termination = 0;
    130     Ident___exception_info = Ident___exception_code = Ident___abnormal_termination = 0;
    131     Ident_GetExceptionInfo = Ident_GetExceptionCode = Ident_AbnormalTermination = 0;
    132   }
    133 
    134   if (!DelayInitialization) {
    135     assert(Target && "Must provide target information for PP initialization");
    136     Initialize(*Target);
    137   }
    138 }
    139 
    140 Preprocessor::~Preprocessor() {
    141   assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
    142 
    143   while (!IncludeMacroStack.empty()) {
    144     delete IncludeMacroStack.back().TheLexer;
    145     delete IncludeMacroStack.back().TheTokenLexer;
    146     IncludeMacroStack.pop_back();
    147   }
    148 
    149   // Free any macro definitions.
    150   for (MacroInfoChain *I = MIChainHead ; I ; I = I->Next)
    151     I->MI.Destroy();
    152 
    153   // Free any cached macro expanders.
    154   for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
    155     delete TokenLexerCache[i];
    156 
    157   for (DeserializedMacroInfoChain *I = DeserialMIChainHead ; I ; I = I->Next)
    158     I->MI.Destroy();
    159 
    160   // Free any cached MacroArgs.
    161   for (MacroArgs *ArgList = MacroArgCache; ArgList; )
    162     ArgList = ArgList->deallocate();
    163 
    164   // Release pragma information.
    165   delete PragmaHandlers;
    166 
    167   // Delete the scratch buffer info.
    168   delete ScratchBuf;
    169 
    170   // Delete the header search info, if we own it.
    171   if (OwnsHeaderSearch)
    172     delete &HeaderInfo;
    173 
    174   delete Callbacks;
    175 }
    176 
    177 void Preprocessor::Initialize(const TargetInfo &Target) {
    178   assert((!this->Target || this->Target == &Target) &&
    179          "Invalid override of target information");
    180   this->Target = &Target;
    181 
    182   // Initialize information about built-ins.
    183   BuiltinInfo.InitializeTarget(Target);
    184   HeaderInfo.setTarget(Target);
    185 }
    186 
    187 void Preprocessor::setPTHManager(PTHManager* pm) {
    188   PTH.reset(pm);
    189   FileMgr.addStatCache(PTH->createStatCache());
    190 }
    191 
    192 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
    193   llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
    194                << getSpelling(Tok) << "'";
    195 
    196   if (!DumpFlags) return;
    197 
    198   llvm::errs() << "\t";
    199   if (Tok.isAtStartOfLine())
    200     llvm::errs() << " [StartOfLine]";
    201   if (Tok.hasLeadingSpace())
    202     llvm::errs() << " [LeadingSpace]";
    203   if (Tok.isExpandDisabled())
    204     llvm::errs() << " [ExpandDisabled]";
    205   if (Tok.needsCleaning()) {
    206     const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
    207     llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
    208                  << "']";
    209   }
    210 
    211   llvm::errs() << "\tLoc=<";
    212   DumpLocation(Tok.getLocation());
    213   llvm::errs() << ">";
    214 }
    215 
    216 void Preprocessor::DumpLocation(SourceLocation Loc) const {
    217   Loc.dump(SourceMgr);
    218 }
    219 
    220 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
    221   llvm::errs() << "MACRO: ";
    222   for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
    223     DumpToken(MI.getReplacementToken(i));
    224     llvm::errs() << "  ";
    225   }
    226   llvm::errs() << "\n";
    227 }
    228 
    229 void Preprocessor::PrintStats() {
    230   llvm::errs() << "\n*** Preprocessor Stats:\n";
    231   llvm::errs() << NumDirectives << " directives found:\n";
    232   llvm::errs() << "  " << NumDefined << " #define.\n";
    233   llvm::errs() << "  " << NumUndefined << " #undef.\n";
    234   llvm::errs() << "  #include/#include_next/#import:\n";
    235   llvm::errs() << "    " << NumEnteredSourceFiles << " source files entered.\n";
    236   llvm::errs() << "    " << MaxIncludeStackDepth << " max include stack depth\n";
    237   llvm::errs() << "  " << NumIf << " #if/#ifndef/#ifdef.\n";
    238   llvm::errs() << "  " << NumElse << " #else/#elif.\n";
    239   llvm::errs() << "  " << NumEndif << " #endif.\n";
    240   llvm::errs() << "  " << NumPragma << " #pragma.\n";
    241   llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
    242 
    243   llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
    244              << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
    245              << NumFastMacroExpanded << " on the fast path.\n";
    246   llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
    247              << " token paste (##) operations performed, "
    248              << NumFastTokenPaste << " on the fast path.\n";
    249 
    250   llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
    251 
    252   llvm::errs() << "\n  BumpPtr: " << BP.getTotalMemory();
    253   llvm::errs() << "\n  Macro Expanded Tokens: "
    254                << llvm::capacity_in_bytes(MacroExpandedTokens);
    255   llvm::errs() << "\n  Predefines Buffer: " << Predefines.capacity();
    256   llvm::errs() << "\n  Macros: " << llvm::capacity_in_bytes(Macros);
    257   llvm::errs() << "\n  #pragma push_macro Info: "
    258                << llvm::capacity_in_bytes(PragmaPushMacroInfo);
    259   llvm::errs() << "\n  Poison Reasons: "
    260                << llvm::capacity_in_bytes(PoisonReasons);
    261   llvm::errs() << "\n  Comment Handlers: "
    262                << llvm::capacity_in_bytes(CommentHandlers) << "\n";
    263 }
    264 
    265 Preprocessor::macro_iterator
    266 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
    267   if (IncludeExternalMacros && ExternalSource &&
    268       !ReadMacrosFromExternalSource) {
    269     ReadMacrosFromExternalSource = true;
    270     ExternalSource->ReadDefinedMacros();
    271   }
    272 
    273   return Macros.begin();
    274 }
    275 
    276 size_t Preprocessor::getTotalMemory() const {
    277   return BP.getTotalMemory()
    278     + llvm::capacity_in_bytes(MacroExpandedTokens)
    279     + Predefines.capacity() /* Predefines buffer. */
    280     + llvm::capacity_in_bytes(Macros)
    281     + llvm::capacity_in_bytes(PragmaPushMacroInfo)
    282     + llvm::capacity_in_bytes(PoisonReasons)
    283     + llvm::capacity_in_bytes(CommentHandlers);
    284 }
    285 
    286 Preprocessor::macro_iterator
    287 Preprocessor::macro_end(bool IncludeExternalMacros) const {
    288   if (IncludeExternalMacros && ExternalSource &&
    289       !ReadMacrosFromExternalSource) {
    290     ReadMacrosFromExternalSource = true;
    291     ExternalSource->ReadDefinedMacros();
    292   }
    293 
    294   return Macros.end();
    295 }
    296 
    297 /// \brief Compares macro tokens with a specified token value sequence.
    298 static bool MacroDefinitionEquals(const MacroInfo *MI,
    299                                   ArrayRef<TokenValue> Tokens) {
    300   return Tokens.size() == MI->getNumTokens() &&
    301       std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
    302 }
    303 
    304 StringRef Preprocessor::getLastMacroWithSpelling(
    305                                     SourceLocation Loc,
    306                                     ArrayRef<TokenValue> Tokens) const {
    307   SourceLocation BestLocation;
    308   StringRef BestSpelling;
    309   for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
    310        I != E; ++I) {
    311     if (!I->second->getMacroInfo()->isObjectLike())
    312       continue;
    313     const MacroDirective::DefInfo
    314       Def = I->second->findDirectiveAtLoc(Loc, SourceMgr);
    315     if (!Def)
    316       continue;
    317     if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
    318       continue;
    319     SourceLocation Location = Def.getLocation();
    320     // Choose the macro defined latest.
    321     if (BestLocation.isInvalid() ||
    322         (Location.isValid() &&
    323          SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
    324       BestLocation = Location;
    325       BestSpelling = I->first->getName();
    326     }
    327   }
    328   return BestSpelling;
    329 }
    330 
    331 void Preprocessor::recomputeCurLexerKind() {
    332   if (CurLexer)
    333     CurLexerKind = CLK_Lexer;
    334   else if (CurPTHLexer)
    335     CurLexerKind = CLK_PTHLexer;
    336   else if (CurTokenLexer)
    337     CurLexerKind = CLK_TokenLexer;
    338   else
    339     CurLexerKind = CLK_CachingLexer;
    340 }
    341 
    342 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
    343                                           unsigned CompleteLine,
    344                                           unsigned CompleteColumn) {
    345   assert(File);
    346   assert(CompleteLine && CompleteColumn && "Starts from 1:1");
    347   assert(!CodeCompletionFile && "Already set");
    348 
    349   using llvm::MemoryBuffer;
    350 
    351   // Load the actual file's contents.
    352   bool Invalid = false;
    353   const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
    354   if (Invalid)
    355     return true;
    356 
    357   // Find the byte position of the truncation point.
    358   const char *Position = Buffer->getBufferStart();
    359   for (unsigned Line = 1; Line < CompleteLine; ++Line) {
    360     for (; *Position; ++Position) {
    361       if (*Position != '\r' && *Position != '\n')
    362         continue;
    363 
    364       // Eat \r\n or \n\r as a single line.
    365       if ((Position[1] == '\r' || Position[1] == '\n') &&
    366           Position[0] != Position[1])
    367         ++Position;
    368       ++Position;
    369       break;
    370     }
    371   }
    372 
    373   Position += CompleteColumn - 1;
    374 
    375   // Insert '\0' at the code-completion point.
    376   if (Position < Buffer->getBufferEnd()) {
    377     CodeCompletionFile = File;
    378     CodeCompletionOffset = Position - Buffer->getBufferStart();
    379 
    380     MemoryBuffer *NewBuffer =
    381         MemoryBuffer::getNewUninitMemBuffer(Buffer->getBufferSize() + 1,
    382                                             Buffer->getBufferIdentifier());
    383     char *NewBuf = const_cast<char*>(NewBuffer->getBufferStart());
    384     char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
    385     *NewPos = '\0';
    386     std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
    387     SourceMgr.overrideFileContents(File, NewBuffer);
    388   }
    389 
    390   return false;
    391 }
    392 
    393 void Preprocessor::CodeCompleteNaturalLanguage() {
    394   if (CodeComplete)
    395     CodeComplete->CodeCompleteNaturalLanguage();
    396   setCodeCompletionReached();
    397 }
    398 
    399 /// getSpelling - This method is used to get the spelling of a token into a
    400 /// SmallVector. Note that the returned StringRef may not point to the
    401 /// supplied buffer if a copy can be avoided.
    402 StringRef Preprocessor::getSpelling(const Token &Tok,
    403                                           SmallVectorImpl<char> &Buffer,
    404                                           bool *Invalid) const {
    405   // NOTE: this has to be checked *before* testing for an IdentifierInfo.
    406   if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
    407     // Try the fast path.
    408     if (const IdentifierInfo *II = Tok.getIdentifierInfo())
    409       return II->getName();
    410   }
    411 
    412   // Resize the buffer if we need to copy into it.
    413   if (Tok.needsCleaning())
    414     Buffer.resize(Tok.getLength());
    415 
    416   const char *Ptr = Buffer.data();
    417   unsigned Len = getSpelling(Tok, Ptr, Invalid);
    418   return StringRef(Ptr, Len);
    419 }
    420 
    421 /// CreateString - Plop the specified string into a scratch buffer and return a
    422 /// location for it.  If specified, the source location provides a source
    423 /// location for the token.
    424 void Preprocessor::CreateString(StringRef Str, Token &Tok,
    425                                 SourceLocation ExpansionLocStart,
    426                                 SourceLocation ExpansionLocEnd) {
    427   Tok.setLength(Str.size());
    428 
    429   const char *DestPtr;
    430   SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
    431 
    432   if (ExpansionLocStart.isValid())
    433     Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
    434                                        ExpansionLocEnd, Str.size());
    435   Tok.setLocation(Loc);
    436 
    437   // If this is a raw identifier or a literal token, set the pointer data.
    438   if (Tok.is(tok::raw_identifier))
    439     Tok.setRawIdentifierData(DestPtr);
    440   else if (Tok.isLiteral())
    441     Tok.setLiteralData(DestPtr);
    442 }
    443 
    444 Module *Preprocessor::getCurrentModule() {
    445   if (getLangOpts().CurrentModule.empty())
    446     return 0;
    447 
    448   return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
    449 }
    450 
    451 //===----------------------------------------------------------------------===//
    452 // Preprocessor Initialization Methods
    453 //===----------------------------------------------------------------------===//
    454 
    455 
    456 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
    457 /// which implicitly adds the builtin defines etc.
    458 void Preprocessor::EnterMainSourceFile() {
    459   // We do not allow the preprocessor to reenter the main file.  Doing so will
    460   // cause FileID's to accumulate information from both runs (e.g. #line
    461   // information) and predefined macros aren't guaranteed to be set properly.
    462   assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
    463   FileID MainFileID = SourceMgr.getMainFileID();
    464 
    465   // If MainFileID is loaded it means we loaded an AST file, no need to enter
    466   // a main file.
    467   if (!SourceMgr.isLoadedFileID(MainFileID)) {
    468     // Enter the main file source buffer.
    469     EnterSourceFile(MainFileID, 0, SourceLocation());
    470 
    471     // If we've been asked to skip bytes in the main file (e.g., as part of a
    472     // precompiled preamble), do so now.
    473     if (SkipMainFilePreamble.first > 0)
    474       CurLexer->SkipBytes(SkipMainFilePreamble.first,
    475                           SkipMainFilePreamble.second);
    476 
    477     // Tell the header info that the main file was entered.  If the file is later
    478     // #imported, it won't be re-entered.
    479     if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
    480       HeaderInfo.IncrementIncludeCount(FE);
    481   }
    482 
    483   // Preprocess Predefines to populate the initial preprocessor state.
    484   llvm::MemoryBuffer *SB =
    485     llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
    486   assert(SB && "Cannot create predefined source buffer");
    487   FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
    488   assert(!FID.isInvalid() && "Could not create FileID for predefines?");
    489   setPredefinesFileID(FID);
    490 
    491   // Start parsing the predefines.
    492   EnterSourceFile(FID, 0, SourceLocation());
    493 }
    494 
    495 void Preprocessor::EndSourceFile() {
    496   // Notify the client that we reached the end of the source file.
    497   if (Callbacks)
    498     Callbacks->EndOfMainFile();
    499 }
    500 
    501 //===----------------------------------------------------------------------===//
    502 // Lexer Event Handling.
    503 //===----------------------------------------------------------------------===//
    504 
    505 static void appendCodePoint(unsigned Codepoint,
    506                             llvm::SmallVectorImpl<char> &Str) {
    507   char ResultBuf[4];
    508   char *ResultPtr = ResultBuf;
    509   bool Res = llvm::ConvertCodePointToUTF8(Codepoint, ResultPtr);
    510   (void)Res;
    511   assert(Res && "Unexpected conversion failure");
    512   Str.append(ResultBuf, ResultPtr);
    513 }
    514 
    515 static void expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
    516   for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
    517     if (*I != '\\') {
    518       Buf.push_back(*I);
    519       continue;
    520     }
    521 
    522     ++I;
    523     assert(*I == 'u' || *I == 'U');
    524 
    525     unsigned NumHexDigits;
    526     if (*I == 'u')
    527       NumHexDigits = 4;
    528     else
    529       NumHexDigits = 8;
    530 
    531     assert(I + NumHexDigits <= E);
    532 
    533     uint32_t CodePoint = 0;
    534     for (++I; NumHexDigits != 0; ++I, --NumHexDigits) {
    535       unsigned Value = llvm::hexDigitValue(*I);
    536       assert(Value != -1U);
    537 
    538       CodePoint <<= 4;
    539       CodePoint += Value;
    540     }
    541 
    542     appendCodePoint(CodePoint, Buf);
    543     --I;
    544   }
    545 }
    546 
    547 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
    548 /// identifier information for the token and install it into the token,
    549 /// updating the token kind accordingly.
    550 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
    551   assert(Identifier.getRawIdentifierData() != 0 && "No raw identifier data!");
    552 
    553   // Look up this token, see if it is a macro, or if it is a language keyword.
    554   IdentifierInfo *II;
    555   if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
    556     // No cleaning needed, just use the characters from the lexed buffer.
    557     II = getIdentifierInfo(StringRef(Identifier.getRawIdentifierData(),
    558                                      Identifier.getLength()));
    559   } else {
    560     // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
    561     SmallString<64> IdentifierBuffer;
    562     StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
    563 
    564     if (Identifier.hasUCN()) {
    565       SmallString<64> UCNIdentifierBuffer;
    566       expandUCNs(UCNIdentifierBuffer, CleanedStr);
    567       II = getIdentifierInfo(UCNIdentifierBuffer);
    568     } else {
    569       II = getIdentifierInfo(CleanedStr);
    570     }
    571   }
    572 
    573   // Update the token info (identifier info and appropriate token kind).
    574   Identifier.setIdentifierInfo(II);
    575   Identifier.setKind(II->getTokenID());
    576 
    577   return II;
    578 }
    579 
    580 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
    581   PoisonReasons[II] = DiagID;
    582 }
    583 
    584 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
    585   assert(Ident__exception_code && Ident__exception_info);
    586   assert(Ident___exception_code && Ident___exception_info);
    587   Ident__exception_code->setIsPoisoned(Poison);
    588   Ident___exception_code->setIsPoisoned(Poison);
    589   Ident_GetExceptionCode->setIsPoisoned(Poison);
    590   Ident__exception_info->setIsPoisoned(Poison);
    591   Ident___exception_info->setIsPoisoned(Poison);
    592   Ident_GetExceptionInfo->setIsPoisoned(Poison);
    593   Ident__abnormal_termination->setIsPoisoned(Poison);
    594   Ident___abnormal_termination->setIsPoisoned(Poison);
    595   Ident_AbnormalTermination->setIsPoisoned(Poison);
    596 }
    597 
    598 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
    599   assert(Identifier.getIdentifierInfo() &&
    600          "Can't handle identifiers without identifier info!");
    601   llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
    602     PoisonReasons.find(Identifier.getIdentifierInfo());
    603   if(it == PoisonReasons.end())
    604     Diag(Identifier, diag::err_pp_used_poisoned_id);
    605   else
    606     Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
    607 }
    608 
    609 /// HandleIdentifier - This callback is invoked when the lexer reads an
    610 /// identifier.  This callback looks up the identifier in the map and/or
    611 /// potentially macro expands it or turns it into a named token (like 'for').
    612 ///
    613 /// Note that callers of this method are guarded by checking the
    614 /// IdentifierInfo's 'isHandleIdentifierCase' bit.  If this method changes, the
    615 /// IdentifierInfo methods that compute these properties will need to change to
    616 /// match.
    617 void Preprocessor::HandleIdentifier(Token &Identifier) {
    618   assert(Identifier.getIdentifierInfo() &&
    619          "Can't handle identifiers without identifier info!");
    620 
    621   IdentifierInfo &II = *Identifier.getIdentifierInfo();
    622 
    623   // If the information about this identifier is out of date, update it from
    624   // the external source.
    625   // We have to treat __VA_ARGS__ in a special way, since it gets
    626   // serialized with isPoisoned = true, but our preprocessor may have
    627   // unpoisoned it if we're defining a C99 macro.
    628   if (II.isOutOfDate()) {
    629     bool CurrentIsPoisoned = false;
    630     if (&II == Ident__VA_ARGS__)
    631       CurrentIsPoisoned = Ident__VA_ARGS__->isPoisoned();
    632 
    633     ExternalSource->updateOutOfDateIdentifier(II);
    634     Identifier.setKind(II.getTokenID());
    635 
    636     if (&II == Ident__VA_ARGS__)
    637       II.setIsPoisoned(CurrentIsPoisoned);
    638   }
    639 
    640   // If this identifier was poisoned, and if it was not produced from a macro
    641   // expansion, emit an error.
    642   if (II.isPoisoned() && CurPPLexer) {
    643     HandlePoisonedIdentifier(Identifier);
    644   }
    645 
    646   // If this is a macro to be expanded, do it.
    647   if (MacroDirective *MD = getMacroDirective(&II)) {
    648     MacroInfo *MI = MD->getMacroInfo();
    649     if (!DisableMacroExpansion) {
    650       if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
    651         if (!HandleMacroExpandedIdentifier(Identifier, MD))
    652           return;
    653       } else {
    654         // C99 6.10.3.4p2 says that a disabled macro may never again be
    655         // expanded, even if it's in a context where it could be expanded in the
    656         // future.
    657         Identifier.setFlag(Token::DisableExpand);
    658         if (MI->isObjectLike() || isNextPPTokenLParen())
    659           Diag(Identifier, diag::pp_disabled_macro_expansion);
    660       }
    661     }
    662   }
    663 
    664   // If this identifier is a keyword in C++11, produce a warning. Don't warn if
    665   // we're not considering macro expansion, since this identifier might be the
    666   // name of a macro.
    667   // FIXME: This warning is disabled in cases where it shouldn't be, like
    668   //   "#define constexpr constexpr", "int constexpr;"
    669   if (II.isCXX11CompatKeyword() & !DisableMacroExpansion) {
    670     Diag(Identifier, diag::warn_cxx11_keyword) << II.getName();
    671     // Don't diagnose this keyword again in this translation unit.
    672     II.setIsCXX11CompatKeyword(false);
    673   }
    674 
    675   // C++ 2.11p2: If this is an alternative representation of a C++ operator,
    676   // then we act as if it is the actual operator and not the textual
    677   // representation of it.
    678   if (II.isCPlusPlusOperatorKeyword())
    679     Identifier.setIdentifierInfo(0);
    680 
    681   // If this is an extension token, diagnose its use.
    682   // We avoid diagnosing tokens that originate from macro definitions.
    683   // FIXME: This warning is disabled in cases where it shouldn't be,
    684   // like "#define TY typeof", "TY(1) x".
    685   if (II.isExtensionToken() && !DisableMacroExpansion)
    686     Diag(Identifier, diag::ext_token_used);
    687 
    688   // If this is the 'import' contextual keyword, note
    689   // that the next token indicates a module name.
    690   //
    691   // Note that we do not treat 'import' as a contextual
    692   // keyword when we're in a caching lexer, because caching lexers only get
    693   // used in contexts where import declarations are disallowed.
    694   if (II.isModulesImport() && !InMacroArgs && !DisableMacroExpansion &&
    695       getLangOpts().Modules && CurLexerKind != CLK_CachingLexer) {
    696     ModuleImportLoc = Identifier.getLocation();
    697     ModuleImportPath.clear();
    698     ModuleImportExpectsIdentifier = true;
    699     CurLexerKind = CLK_LexAfterModuleImport;
    700   }
    701 }
    702 
    703 /// \brief Lex a token following the 'import' contextual keyword.
    704 ///
    705 void Preprocessor::LexAfterModuleImport(Token &Result) {
    706   // Figure out what kind of lexer we actually have.
    707   recomputeCurLexerKind();
    708 
    709   // Lex the next token.
    710   Lex(Result);
    711 
    712   // The token sequence
    713   //
    714   //   import identifier (. identifier)*
    715   //
    716   // indicates a module import directive. We already saw the 'import'
    717   // contextual keyword, so now we're looking for the identifiers.
    718   if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
    719     // We expected to see an identifier here, and we did; continue handling
    720     // identifiers.
    721     ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
    722                                               Result.getLocation()));
    723     ModuleImportExpectsIdentifier = false;
    724     CurLexerKind = CLK_LexAfterModuleImport;
    725     return;
    726   }
    727 
    728   // If we're expecting a '.' or a ';', and we got a '.', then wait until we
    729   // see the next identifier.
    730   if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
    731     ModuleImportExpectsIdentifier = true;
    732     CurLexerKind = CLK_LexAfterModuleImport;
    733     return;
    734   }
    735 
    736   // If we have a non-empty module path, load the named module.
    737   if (!ModuleImportPath.empty() && getLangOpts().Modules) {
    738     Module *Imported = TheModuleLoader.loadModule(ModuleImportLoc,
    739                                                   ModuleImportPath,
    740                                                   Module::MacrosVisible,
    741                                                   /*IsIncludeDirective=*/false);
    742     if (Callbacks)
    743       Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
    744   }
    745 }
    746 
    747 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
    748                                           const char *DiagnosticTag,
    749                                           bool AllowMacroExpansion) {
    750   // We need at least one string literal.
    751   if (Result.isNot(tok::string_literal)) {
    752     Diag(Result, diag::err_expected_string_literal)
    753       << /*Source='in...'*/0 << DiagnosticTag;
    754     return false;
    755   }
    756 
    757   // Lex string literal tokens, optionally with macro expansion.
    758   SmallVector<Token, 4> StrToks;
    759   do {
    760     StrToks.push_back(Result);
    761 
    762     if (Result.hasUDSuffix())
    763       Diag(Result, diag::err_invalid_string_udl);
    764 
    765     if (AllowMacroExpansion)
    766       Lex(Result);
    767     else
    768       LexUnexpandedToken(Result);
    769   } while (Result.is(tok::string_literal));
    770 
    771   // Concatenate and parse the strings.
    772   StringLiteralParser Literal(&StrToks[0], StrToks.size(), *this);
    773   assert(Literal.isAscii() && "Didn't allow wide strings in");
    774 
    775   if (Literal.hadError)
    776     return false;
    777 
    778   if (Literal.Pascal) {
    779     Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
    780       << /*Source='in...'*/0 << DiagnosticTag;
    781     return false;
    782   }
    783 
    784   String = Literal.GetString();
    785   return true;
    786 }
    787 
    788 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
    789   assert(Handler && "NULL comment handler");
    790   assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
    791          CommentHandlers.end() && "Comment handler already registered");
    792   CommentHandlers.push_back(Handler);
    793 }
    794 
    795 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
    796   std::vector<CommentHandler *>::iterator Pos
    797   = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
    798   assert(Pos != CommentHandlers.end() && "Comment handler not registered");
    799   CommentHandlers.erase(Pos);
    800 }
    801 
    802 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
    803   bool AnyPendingTokens = false;
    804   for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
    805        HEnd = CommentHandlers.end();
    806        H != HEnd; ++H) {
    807     if ((*H)->HandleComment(*this, Comment))
    808       AnyPendingTokens = true;
    809   }
    810   if (!AnyPendingTokens || getCommentRetentionState())
    811     return false;
    812   Lex(result);
    813   return true;
    814 }
    815 
    816 ModuleLoader::~ModuleLoader() { }
    817 
    818 CommentHandler::~CommentHandler() { }
    819 
    820 CodeCompletionHandler::~CodeCompletionHandler() { }
    821 
    822 void Preprocessor::createPreprocessingRecord() {
    823   if (Record)
    824     return;
    825 
    826   Record = new PreprocessingRecord(getSourceManager());
    827   addPPCallbacks(Record);
    828 }
    829