1 //===--- Lexer.cpp - C Language Family Lexer ------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the Lexer and Token interfaces.
12 //===----------------------------------------------------------------------===//
14 // TODO: GCC Diagnostics emitted by the lexer:
15 // PEDWARN: (form feed|vertical tab) in preprocessing directive
17 // Universal characters, unicode, char mapping:
18 // WARNING: `%.*s' is not in NFKC
19 // WARNING: `%.*s' is not in NFC
22 // TODO: Options to support:
23 // -fexec-charset,-fwide-exec-charset
25 //===----------------------------------------------------------------------===//
27 #include "clang/Lex/Lexer.h"
28 #include "clang/Lex/Preprocessor.h"
29 #include "clang/Lex/LexDiagnostic.h"
30 #include "clang/Lex/CodeCompletionHandler.h"
31 #include "clang/Basic/SourceManager.h"
32 #include "llvm/ADT/StringSwitch.h"
33 #include "llvm/Support/Compiler.h"
34 #include "llvm/Support/MemoryBuffer.h"
36 using namespace clang;
38 static void InitCharacterInfo();
40 //===----------------------------------------------------------------------===//
41 // Token Class Implementation
42 //===----------------------------------------------------------------------===//
44 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
45 bool Token::isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const {
46 if (IdentifierInfo *II = getIdentifierInfo())
47 return II->getObjCKeywordID() == objcKey;
51 /// getObjCKeywordID - Return the ObjC keyword kind.
52 tok::ObjCKeywordKind Token::getObjCKeywordID() const {
53 IdentifierInfo *specId = getIdentifierInfo();
54 return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
58 //===----------------------------------------------------------------------===//
59 // Lexer Class Implementation
60 //===----------------------------------------------------------------------===//
62 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
66 BufferStart = BufStart;
70 assert(BufEnd[0] == 0 &&
71 "We assume that the input buffer has a null character at the end"
72 " to simplify lexing!");
74 // Check whether we have a BOM in the beginning of the buffer. If yes - act
75 // accordingly. Right now we support only UTF-8 with and without BOM, so, just
76 // skip the UTF-8 BOM if it's present.
77 if (BufferStart == BufferPtr) {
78 // Determine the size of the BOM.
79 StringRef Buf(BufferStart, BufferEnd - BufferStart);
80 size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
81 .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
85 BufferPtr += BOMLength;
88 Is_PragmaLexer = false;
89 CurrentConflictMarkerState = CMK_None;
91 // Start of the file is a start of line.
92 IsAtStartOfLine = true;
94 // We are not after parsing a #.
95 ParsingPreprocessorDirective = false;
97 // We are not after parsing #include.
98 ParsingFilename = false;
100 // We are not in raw mode. Raw mode disables diagnostics and interpretation
101 // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
102 // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
103 // or otherwise skipping over tokens.
104 LexingRawMode = false;
106 // Default to not keeping comments.
107 ExtendedTokenMode = 0;
110 /// Lexer constructor - Create a new lexer object for the specified buffer
111 /// with the specified preprocessor managing the lexing process. This lexer
112 /// assumes that the associated file buffer and Preprocessor objects will
113 /// outlive it, so it doesn't take ownership of either of them.
114 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
115 : PreprocessorLexer(&PP, FID),
116 FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
117 Features(PP.getLangOptions()) {
119 InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
120 InputFile->getBufferEnd());
122 // Default to keeping comments if the preprocessor wants them.
123 SetCommentRetentionState(PP.getCommentRetentionState());
126 /// Lexer constructor - Create a new raw lexer object. This object is only
127 /// suitable for calls to 'LexRawToken'. This lexer assumes that the text
128 /// range will outlive it, so it doesn't take ownership of it.
129 Lexer::Lexer(SourceLocation fileloc, const LangOptions &features,
130 const char *BufStart, const char *BufPtr, const char *BufEnd)
131 : FileLoc(fileloc), Features(features) {
133 InitLexer(BufStart, BufPtr, BufEnd);
135 // We *are* in raw mode.
136 LexingRawMode = true;
139 /// Lexer constructor - Create a new raw lexer object. This object is only
140 /// suitable for calls to 'LexRawToken'. This lexer assumes that the text
141 /// range will outlive it, so it doesn't take ownership of it.
142 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
143 const SourceManager &SM, const LangOptions &features)
144 : FileLoc(SM.getLocForStartOfFile(FID)), Features(features) {
146 InitLexer(FromFile->getBufferStart(), FromFile->getBufferStart(),
147 FromFile->getBufferEnd());
149 // We *are* in raw mode.
150 LexingRawMode = true;
153 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
154 /// _Pragma expansion. This has a variety of magic semantics that this method
155 /// sets up. It returns a new'd Lexer that must be delete'd when done.
157 /// On entrance to this routine, TokStartLoc is a macro location which has a
158 /// spelling loc that indicates the bytes to be lexed for the token and an
159 /// expansion location that indicates where all lexed tokens should be
162 /// FIXME: It would really be nice to make _Pragma just be a wrapper around a
163 /// normal lexer that remaps tokens as they fly by. This would require making
164 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
165 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
166 /// out of the critical path of the lexer!
168 Lexer *Lexer::Create_PragmaLexer(SourceLocation SpellingLoc,
169 SourceLocation ExpansionLocStart,
170 SourceLocation ExpansionLocEnd,
171 unsigned TokLen, Preprocessor &PP) {
172 SourceManager &SM = PP.getSourceManager();
174 // Create the lexer as if we were going to lex the file normally.
175 FileID SpellingFID = SM.getFileID(SpellingLoc);
176 const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
177 Lexer *L = new Lexer(SpellingFID, InputFile, PP);
179 // Now that the lexer is created, change the start/end locations so that we
180 // just lex the subsection of the file that we want. This is lexing from a
182 const char *StrData = SM.getCharacterData(SpellingLoc);
184 L->BufferPtr = StrData;
185 L->BufferEnd = StrData+TokLen;
186 assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
188 // Set the SourceLocation with the remapping information. This ensures that
189 // GetMappedTokenLoc will remap the tokens as they are lexed.
190 L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
192 ExpansionLocEnd, TokLen);
194 // Ensure that the lexer thinks it is inside a directive, so that end \n will
195 // return an EOD token.
196 L->ParsingPreprocessorDirective = true;
198 // This lexer really is for _Pragma.
199 L->Is_PragmaLexer = true;
204 /// Stringify - Convert the specified string into a C string, with surrounding
205 /// ""'s, and with escaped \ and " characters.
206 std::string Lexer::Stringify(const std::string &Str, bool Charify) {
207 std::string Result = Str;
208 char Quote = Charify ? '\'' : '"';
209 for (unsigned i = 0, e = Result.size(); i != e; ++i) {
210 if (Result[i] == '\\' || Result[i] == Quote) {
211 Result.insert(Result.begin()+i, '\\');
218 /// Stringify - Convert the specified string into a C string by escaping '\'
219 /// and " characters. This does not add surrounding ""'s to the string.
220 void Lexer::Stringify(SmallVectorImpl<char> &Str) {
221 for (unsigned i = 0, e = Str.size(); i != e; ++i) {
222 if (Str[i] == '\\' || Str[i] == '"') {
223 Str.insert(Str.begin()+i, '\\');
229 //===----------------------------------------------------------------------===//
231 //===----------------------------------------------------------------------===//
233 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
234 /// token are the characters used to represent the token in the source file
235 /// after trigraph expansion and escaped-newline folding. In particular, this
236 /// wants to get the true, uncanonicalized, spelling of things like digraphs
238 StringRef Lexer::getSpelling(SourceLocation loc,
239 SmallVectorImpl<char> &buffer,
240 const SourceManager &SM,
241 const LangOptions &options,
243 // Break down the source location.
244 std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
246 // Try to the load the file buffer.
247 bool invalidTemp = false;
248 StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
250 if (invalid) *invalid = true;
254 const char *tokenBegin = file.data() + locInfo.second;
256 // Lex from the start of the given location.
257 Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
258 file.begin(), tokenBegin, file.end());
260 lexer.LexFromRawLexer(token);
262 unsigned length = token.getLength();
264 // Common case: no need for cleaning.
265 if (!token.needsCleaning())
266 return StringRef(tokenBegin, length);
268 // Hard case, we need to relex the characters into the string.
270 buffer.reserve(length);
272 for (const char *ti = tokenBegin, *te = ti + length; ti != te; ) {
274 buffer.push_back(Lexer::getCharAndSizeNoWarn(ti, charSize, options));
278 return StringRef(buffer.data(), buffer.size());
281 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
282 /// token are the characters used to represent the token in the source file
283 /// after trigraph expansion and escaped-newline folding. In particular, this
284 /// wants to get the true, uncanonicalized, spelling of things like digraphs
286 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
287 const LangOptions &Features, bool *Invalid) {
288 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
290 // If this token contains nothing interesting, return it directly.
291 bool CharDataInvalid = false;
292 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
295 *Invalid = CharDataInvalid;
297 return std::string();
299 if (!Tok.needsCleaning())
300 return std::string(TokStart, TokStart+Tok.getLength());
303 Result.reserve(Tok.getLength());
305 // Otherwise, hard case, relex the characters into the string.
306 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
309 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
312 assert(Result.size() != unsigned(Tok.getLength()) &&
313 "NeedsCleaning flag set on something that didn't need cleaning!");
317 /// getSpelling - This method is used to get the spelling of a token into a
318 /// preallocated buffer, instead of as an std::string. The caller is required
319 /// to allocate enough space for the token, which is guaranteed to be at least
320 /// Tok.getLength() bytes long. The actual length of the token is returned.
322 /// Note that this method may do two possible things: it may either fill in
323 /// the buffer specified with characters, or it may *change the input pointer*
324 /// to point to a constant buffer with the data already in it (avoiding a
325 /// copy). The caller is not allowed to modify the returned buffer pointer
326 /// if an internal buffer is returned.
327 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
328 const SourceManager &SourceMgr,
329 const LangOptions &Features, bool *Invalid) {
330 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
332 const char *TokStart = 0;
333 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
334 if (Tok.is(tok::raw_identifier))
335 TokStart = Tok.getRawIdentifierData();
336 else if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
337 // Just return the string from the identifier table, which is very quick.
338 Buffer = II->getNameStart();
339 return II->getLength();
342 // NOTE: this can be checked even after testing for an IdentifierInfo.
344 TokStart = Tok.getLiteralData();
347 // Compute the start of the token in the input lexer buffer.
348 bool CharDataInvalid = false;
349 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
351 *Invalid = CharDataInvalid;
352 if (CharDataInvalid) {
358 // If this token contains nothing interesting, return it directly.
359 if (!Tok.needsCleaning()) {
361 return Tok.getLength();
364 // Otherwise, hard case, relex the characters into the string.
365 char *OutBuf = const_cast<char*>(Buffer);
366 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
369 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
372 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
373 "NeedsCleaning flag set on something that didn't need cleaning!");
375 return OutBuf-Buffer;
380 static bool isWhitespace(unsigned char c);
382 /// MeasureTokenLength - Relex the token at the specified location and return
383 /// its length in bytes in the input file. If the token needs cleaning (e.g.
384 /// includes a trigraph or an escaped newline) then this count includes bytes
385 /// that are part of that.
386 unsigned Lexer::MeasureTokenLength(SourceLocation Loc,
387 const SourceManager &SM,
388 const LangOptions &LangOpts) {
389 // TODO: this could be special cased for common tokens like identifiers, ')',
390 // etc to make this faster, if it mattered. Just look at StrData[0] to handle
391 // all obviously single-char tokens. This could use
392 // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
395 // If this comes from a macro expansion, we really do want the macro name, not
396 // the token this macro expanded to.
397 Loc = SM.getExpansionLoc(Loc);
398 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
399 bool Invalid = false;
400 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
404 const char *StrData = Buffer.data()+LocInfo.second;
406 if (isWhitespace(StrData[0]))
409 // Create a lexer starting at the beginning of this token.
410 Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
411 Buffer.begin(), StrData, Buffer.end());
412 TheLexer.SetCommentRetentionState(true);
414 TheLexer.LexFromRawLexer(TheTok);
415 return TheTok.getLength();
418 static SourceLocation getBeginningOfFileToken(SourceLocation Loc,
419 const SourceManager &SM,
420 const LangOptions &LangOpts) {
421 assert(Loc.isFileID());
422 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
423 if (LocInfo.first.isInvalid())
426 bool Invalid = false;
427 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
431 // Back up from the current location until we hit the beginning of a line
432 // (or the buffer). We'll relex from that point.
433 const char *BufStart = Buffer.data();
434 if (LocInfo.second >= Buffer.size())
437 const char *StrData = BufStart+LocInfo.second;
438 if (StrData[0] == '\n' || StrData[0] == '\r')
441 const char *LexStart = StrData;
442 while (LexStart != BufStart) {
443 if (LexStart[0] == '\n' || LexStart[0] == '\r') {
451 // Create a lexer starting at the beginning of this token.
452 SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
453 Lexer TheLexer(LexerStartLoc, LangOpts, BufStart, LexStart, Buffer.end());
454 TheLexer.SetCommentRetentionState(true);
456 // Lex tokens until we find the token that contains the source location.
459 TheLexer.LexFromRawLexer(TheTok);
461 if (TheLexer.getBufferLocation() > StrData) {
462 // Lexing this token has taken the lexer past the source location we're
463 // looking for. If the current token encompasses our source location,
464 // return the beginning of that token.
465 if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
466 return TheTok.getLocation();
468 // We ended up skipping over the source location entirely, which means
469 // that it points into whitespace. We're done here.
472 } while (TheTok.getKind() != tok::eof);
474 // We've passed our source location; just return the original source location.
478 SourceLocation Lexer::GetBeginningOfToken(SourceLocation Loc,
479 const SourceManager &SM,
480 const LangOptions &LangOpts) {
482 return getBeginningOfFileToken(Loc, SM, LangOpts);
484 if (!SM.isMacroArgExpansion(Loc))
487 SourceLocation FileLoc = SM.getSpellingLoc(Loc);
488 SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
489 std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
490 std::pair<FileID, unsigned> BeginFileLocInfo= SM.getDecomposedLoc(BeginFileLoc);
491 assert(FileLocInfo.first == BeginFileLocInfo.first &&
492 FileLocInfo.second >= BeginFileLocInfo.second);
493 return Loc.getLocWithOffset(SM.getDecomposedLoc(BeginFileLoc).second -
494 SM.getDecomposedLoc(FileLoc).second);
498 enum PreambleDirectiveKind {
506 std::pair<unsigned, bool>
507 Lexer::ComputePreamble(const llvm::MemoryBuffer *Buffer,
508 const LangOptions &Features, unsigned MaxLines) {
509 // Create a lexer starting at the beginning of the file. Note that we use a
510 // "fake" file source location at offset 1 so that the lexer will track our
511 // position within the file.
512 const unsigned StartOffset = 1;
513 SourceLocation StartLoc = SourceLocation::getFromRawEncoding(StartOffset);
514 Lexer TheLexer(StartLoc, Features, Buffer->getBufferStart(),
515 Buffer->getBufferStart(), Buffer->getBufferEnd());
517 bool InPreprocessorDirective = false;
520 unsigned IfCount = 0;
522 unsigned MaxLineOffset = 0;
524 const char *CurPtr = Buffer->getBufferStart();
525 unsigned CurLine = 0;
526 while (CurPtr != Buffer->getBufferEnd()) {
530 if (CurLine == MaxLines)
534 if (CurPtr != Buffer->getBufferEnd())
535 MaxLineOffset = CurPtr - Buffer->getBufferStart();
539 TheLexer.LexFromRawLexer(TheTok);
541 if (InPreprocessorDirective) {
542 // If we've hit the end of the file, we're done.
543 if (TheTok.getKind() == tok::eof) {
544 InPreprocessorDirective = false;
548 // If we haven't hit the end of the preprocessor directive, skip this
550 if (!TheTok.isAtStartOfLine())
553 // We've passed the end of the preprocessor directive, and will look
554 // at this token again below.
555 InPreprocessorDirective = false;
558 // Keep track of the # of lines in the preamble.
559 if (TheTok.isAtStartOfLine()) {
560 unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
562 // If we were asked to limit the number of lines in the preamble,
563 // and we're about to exceed that limit, we're done.
564 if (MaxLineOffset && TokOffset >= MaxLineOffset)
568 // Comments are okay; skip over them.
569 if (TheTok.getKind() == tok::comment)
572 if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
573 // This is the start of a preprocessor directive.
574 Token HashTok = TheTok;
575 InPreprocessorDirective = true;
577 // Figure out which directive this is. Since we're lexing raw tokens,
578 // we don't have an identifier table available. Instead, just look at
579 // the raw identifier to recognize and categorize preprocessor directives.
580 TheLexer.LexFromRawLexer(TheTok);
581 if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
582 StringRef Keyword(TheTok.getRawIdentifierData(),
584 PreambleDirectiveKind PDK
585 = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
586 .Case("include", PDK_Skipped)
587 .Case("__include_macros", PDK_Skipped)
588 .Case("define", PDK_Skipped)
589 .Case("undef", PDK_Skipped)
590 .Case("line", PDK_Skipped)
591 .Case("error", PDK_Skipped)
592 .Case("pragma", PDK_Skipped)
593 .Case("import", PDK_Skipped)
594 .Case("include_next", PDK_Skipped)
595 .Case("warning", PDK_Skipped)
596 .Case("ident", PDK_Skipped)
597 .Case("sccs", PDK_Skipped)
598 .Case("assert", PDK_Skipped)
599 .Case("unassert", PDK_Skipped)
600 .Case("if", PDK_StartIf)
601 .Case("ifdef", PDK_StartIf)
602 .Case("ifndef", PDK_StartIf)
603 .Case("elif", PDK_Skipped)
604 .Case("else", PDK_Skipped)
605 .Case("endif", PDK_EndIf)
606 .Default(PDK_Unknown);
614 IfStartTok = HashTok;
620 // Mismatched #endif. The preamble ends here.
628 // We don't know what this directive is; stop at the '#'.
633 // We only end up here if we didn't recognize the preprocessor
634 // directive or it was one that can't occur in the preamble at this
635 // point. Roll back the current token to the location of the '#'.
636 InPreprocessorDirective = false;
640 // We hit a token that we don't recognize as being in the
641 // "preprocessing only" part of the file, so we're no longer in
646 SourceLocation End = IfCount? IfStartTok.getLocation() : TheTok.getLocation();
647 return std::make_pair(End.getRawEncoding() - StartLoc.getRawEncoding(),
648 IfCount? IfStartTok.isAtStartOfLine()
649 : TheTok.isAtStartOfLine());
653 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
654 /// token, return a new location that specifies a character within the token.
655 SourceLocation Lexer::AdvanceToTokenCharacter(SourceLocation TokStart,
657 const SourceManager &SM,
658 const LangOptions &Features) {
659 // Figure out how many physical characters away the specified expansion
660 // character is. This needs to take into consideration newlines and
662 bool Invalid = false;
663 const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
665 // If they request the first char of the token, we're trivially done.
666 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
669 unsigned PhysOffset = 0;
671 // The usual case is that tokens don't contain anything interesting. Skip
672 // over the uninteresting characters. If a token only consists of simple
673 // chars, this method is extremely fast.
674 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
676 return TokStart.getLocWithOffset(PhysOffset);
677 ++TokPtr, --CharNo, ++PhysOffset;
680 // If we have a character that may be a trigraph or escaped newline, use a
681 // lexer to parse it correctly.
682 for (; CharNo; --CharNo) {
684 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
689 // Final detail: if we end up on an escaped newline, we want to return the
690 // location of the actual byte of the token. For example foo\<newline>bar
691 // advanced by 3 should return the location of b, not of \\. One compounding
692 // detail of this is that the escape may be made by a trigraph.
693 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
694 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
696 return TokStart.getLocWithOffset(PhysOffset);
699 /// \brief Computes the source location just past the end of the
700 /// token at this source location.
702 /// This routine can be used to produce a source location that
703 /// points just past the end of the token referenced by \p Loc, and
704 /// is generally used when a diagnostic needs to point just after a
705 /// token where it expected something different that it received. If
706 /// the returned source location would not be meaningful (e.g., if
707 /// it points into a macro), this routine returns an invalid
710 /// \param Offset an offset from the end of the token, where the source
711 /// location should refer to. The default offset (0) produces a source
712 /// location pointing just past the end of the token; an offset of 1 produces
713 /// a source location pointing to the last character in the token, etc.
714 SourceLocation Lexer::getLocForEndOfToken(SourceLocation Loc, unsigned Offset,
715 const SourceManager &SM,
716 const LangOptions &Features) {
718 return SourceLocation();
720 if (Loc.isMacroID()) {
721 if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, Features))
722 return SourceLocation(); // Points inside the macro expansion.
724 // Continue and find the location just after the macro expansion.
725 Loc = SM.getExpansionRange(Loc).second;
728 unsigned Len = Lexer::MeasureTokenLength(Loc, SM, Features);
734 return Loc.getLocWithOffset(Len);
737 /// \brief Returns true if the given MacroID location points at the first
738 /// token of the macro expansion.
739 bool Lexer::isAtStartOfMacroExpansion(SourceLocation loc,
740 const SourceManager &SM,
741 const LangOptions &LangOpts) {
742 assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
744 std::pair<FileID, unsigned> infoLoc = SM.getDecomposedLoc(loc);
745 // FIXME: If the token comes from the macro token paste operator ('##')
746 // this function will always return false;
747 if (infoLoc.second > 0)
748 return false; // Does not point at the start of token.
750 SourceLocation expansionLoc =
751 SM.getSLocEntry(infoLoc.first).getExpansion().getExpansionLocStart();
752 if (expansionLoc.isFileID())
753 return true; // No other macro expansions, this is the first.
755 return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts);
758 /// \brief Returns true if the given MacroID location points at the last
759 /// token of the macro expansion.
760 bool Lexer::isAtEndOfMacroExpansion(SourceLocation loc,
761 const SourceManager &SM,
762 const LangOptions &LangOpts) {
763 assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
765 SourceLocation spellLoc = SM.getSpellingLoc(loc);
766 unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
770 FileID FID = SM.getFileID(loc);
771 SourceLocation afterLoc = loc.getLocWithOffset(tokLen+1);
772 if (SM.isInFileID(afterLoc, FID))
773 return false; // Still in the same FileID, does not point to the last token.
775 // FIXME: If the token comes from the macro token paste operator ('##')
776 // or the stringify operator ('#') this function will always return false;
778 SourceLocation expansionLoc =
779 SM.getSLocEntry(FID).getExpansion().getExpansionLocEnd();
780 if (expansionLoc.isFileID())
781 return true; // No other macro expansions.
783 return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts);
786 //===----------------------------------------------------------------------===//
787 // Character information.
788 //===----------------------------------------------------------------------===//
791 CHAR_HORZ_WS = 0x01, // ' ', '\t', '\f', '\v'. Note, no '\0'
792 CHAR_VERT_WS = 0x02, // '\r', '\n'
793 CHAR_LETTER = 0x04, // a-z,A-Z
794 CHAR_NUMBER = 0x08, // 0-9
795 CHAR_UNDER = 0x10, // _
796 CHAR_PERIOD = 0x20, // .
797 CHAR_RAWDEL = 0x40 // {}[]#<>%:;?*+-/^&|~!=,"'
800 // Statically initialize CharInfo table based on ASCII character set
801 // Reference: FreeBSD 7.2 /usr/share/misc/ascii
802 static const unsigned char CharInfo[256] =
804 // 0 NUL 1 SOH 2 STX 3 ETX
805 // 4 EOT 5 ENQ 6 ACK 7 BEL
808 // 8 BS 9 HT 10 NL 11 VT
809 //12 NP 13 CR 14 SO 15 SI
810 0 , CHAR_HORZ_WS, CHAR_VERT_WS, CHAR_HORZ_WS,
811 CHAR_HORZ_WS, CHAR_VERT_WS, 0 , 0 ,
812 //16 DLE 17 DC1 18 DC2 19 DC3
813 //20 DC4 21 NAK 22 SYN 23 ETB
816 //24 CAN 25 EM 26 SUB 27 ESC
817 //28 FS 29 GS 30 RS 31 US
820 //32 SP 33 ! 34 " 35 #
821 //36 $ 37 % 38 & 39 '
822 CHAR_HORZ_WS, CHAR_RAWDEL , CHAR_RAWDEL , CHAR_RAWDEL ,
823 0 , CHAR_RAWDEL , CHAR_RAWDEL , CHAR_RAWDEL ,
824 //40 ( 41 ) 42 * 43 +
825 //44 , 45 - 46 . 47 /
826 0 , 0 , CHAR_RAWDEL , CHAR_RAWDEL ,
827 CHAR_RAWDEL , CHAR_RAWDEL , CHAR_PERIOD , CHAR_RAWDEL ,
828 //48 0 49 1 50 2 51 3
829 //52 4 53 5 54 6 55 7
830 CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER ,
831 CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER , CHAR_NUMBER ,
832 //56 8 57 9 58 : 59 ;
833 //60 < 61 = 62 > 63 ?
834 CHAR_NUMBER , CHAR_NUMBER , CHAR_RAWDEL , CHAR_RAWDEL ,
835 CHAR_RAWDEL , CHAR_RAWDEL , CHAR_RAWDEL , CHAR_RAWDEL ,
836 //64 @ 65 A 66 B 67 C
837 //68 D 69 E 70 F 71 G
838 0 , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
839 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
840 //72 H 73 I 74 J 75 K
841 //76 L 77 M 78 N 79 O
842 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
843 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
844 //80 P 81 Q 82 R 83 S
845 //84 T 85 U 86 V 87 W
846 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
847 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
848 //88 X 89 Y 90 Z 91 [
849 //92 \ 93 ] 94 ^ 95 _
850 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_RAWDEL ,
851 0 , CHAR_RAWDEL , CHAR_RAWDEL , CHAR_UNDER ,
852 //96 ` 97 a 98 b 99 c
853 //100 d 101 e 102 f 103 g
854 0 , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
855 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
856 //104 h 105 i 106 j 107 k
857 //108 l 109 m 110 n 111 o
858 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
859 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
860 //112 p 113 q 114 r 115 s
861 //116 t 117 u 118 v 119 w
862 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
863 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_LETTER ,
864 //120 x 121 y 122 z 123 {
865 //124 | 125 } 126 ~ 127 DEL
866 CHAR_LETTER , CHAR_LETTER , CHAR_LETTER , CHAR_RAWDEL ,
867 CHAR_RAWDEL , CHAR_RAWDEL , CHAR_RAWDEL , 0
870 static void InitCharacterInfo() {
871 static bool isInited = false;
872 if (isInited) return;
873 // check the statically-initialized CharInfo table
874 assert(CHAR_HORZ_WS == CharInfo[(int)' ']);
875 assert(CHAR_HORZ_WS == CharInfo[(int)'\t']);
876 assert(CHAR_HORZ_WS == CharInfo[(int)'\f']);
877 assert(CHAR_HORZ_WS == CharInfo[(int)'\v']);
878 assert(CHAR_VERT_WS == CharInfo[(int)'\n']);
879 assert(CHAR_VERT_WS == CharInfo[(int)'\r']);
880 assert(CHAR_UNDER == CharInfo[(int)'_']);
881 assert(CHAR_PERIOD == CharInfo[(int)'.']);
882 for (unsigned i = 'a'; i <= 'z'; ++i) {
883 assert(CHAR_LETTER == CharInfo[i]);
884 assert(CHAR_LETTER == CharInfo[i+'A'-'a']);
886 for (unsigned i = '0'; i <= '9'; ++i)
887 assert(CHAR_NUMBER == CharInfo[i]);
893 /// isIdentifierBody - Return true if this is the body character of an
894 /// identifier, which is [a-zA-Z0-9_].
895 static inline bool isIdentifierBody(unsigned char c) {
896 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER)) ? true : false;
899 /// isHorizontalWhitespace - Return true if this character is horizontal
900 /// whitespace: ' ', '\t', '\f', '\v'. Note that this returns false for '\0'.
901 static inline bool isHorizontalWhitespace(unsigned char c) {
902 return (CharInfo[c] & CHAR_HORZ_WS) ? true : false;
905 /// isVerticalWhitespace - Return true if this character is vertical
906 /// whitespace: '\n', '\r'. Note that this returns false for '\0'.
907 static inline bool isVerticalWhitespace(unsigned char c) {
908 return (CharInfo[c] & CHAR_VERT_WS) ? true : false;
911 /// isWhitespace - Return true if this character is horizontal or vertical
912 /// whitespace: ' ', '\t', '\f', '\v', '\n', '\r'. Note that this returns false
914 static inline bool isWhitespace(unsigned char c) {
915 return (CharInfo[c] & (CHAR_HORZ_WS|CHAR_VERT_WS)) ? true : false;
918 /// isNumberBody - Return true if this is the body character of an
919 /// preprocessing number, which is [a-zA-Z0-9_.].
920 static inline bool isNumberBody(unsigned char c) {
921 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER|CHAR_PERIOD)) ?
925 /// isRawStringDelimBody - Return true if this is the body character of a
926 /// raw string delimiter.
927 static inline bool isRawStringDelimBody(unsigned char c) {
928 return (CharInfo[c] &
929 (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER|CHAR_PERIOD|CHAR_RAWDEL)) ?
934 //===----------------------------------------------------------------------===//
935 // Diagnostics forwarding code.
936 //===----------------------------------------------------------------------===//
938 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
939 /// lexer buffer was all expanded at a single point, perform the mapping.
940 /// This is currently only used for _Pragma implementation, so it is the slow
941 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
942 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
943 Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
944 static SourceLocation GetMappedTokenLoc(Preprocessor &PP,
945 SourceLocation FileLoc,
946 unsigned CharNo, unsigned TokLen) {
947 assert(FileLoc.isMacroID() && "Must be a macro expansion");
949 // Otherwise, we're lexing "mapped tokens". This is used for things like
950 // _Pragma handling. Combine the expansion location of FileLoc with the
951 // spelling location.
952 SourceManager &SM = PP.getSourceManager();
954 // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
955 // characters come from spelling(FileLoc)+Offset.
956 SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
957 SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
959 // Figure out the expansion loc range, which is the range covered by the
960 // original _Pragma(...) sequence.
961 std::pair<SourceLocation,SourceLocation> II =
962 SM.getImmediateExpansionRange(FileLoc);
964 return SM.createExpansionLoc(SpellingLoc, II.first, II.second, TokLen);
967 /// getSourceLocation - Return a source location identifier for the specified
968 /// offset in the current file.
969 SourceLocation Lexer::getSourceLocation(const char *Loc,
970 unsigned TokLen) const {
971 assert(Loc >= BufferStart && Loc <= BufferEnd &&
972 "Location out of range for this buffer!");
974 // In the normal case, we're just lexing from a simple file buffer, return
975 // the file id from FileLoc with the offset specified.
976 unsigned CharNo = Loc-BufferStart;
977 if (FileLoc.isFileID())
978 return FileLoc.getLocWithOffset(CharNo);
980 // Otherwise, this is the _Pragma lexer case, which pretends that all of the
981 // tokens are lexed from where the _Pragma was defined.
982 assert(PP && "This doesn't work on raw lexers");
983 return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
986 /// Diag - Forwarding function for diagnostics. This translate a source
987 /// position in the current buffer into a SourceLocation object for rendering.
988 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
989 return PP->Diag(getSourceLocation(Loc), DiagID);
992 //===----------------------------------------------------------------------===//
993 // Trigraph and Escaped Newline Handling Code.
994 //===----------------------------------------------------------------------===//
996 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
997 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
998 static char GetTrigraphCharForLetter(char Letter) {
1001 case '=': return '#';
1002 case ')': return ']';
1003 case '(': return '[';
1004 case '!': return '|';
1005 case '\'': return '^';
1006 case '>': return '}';
1007 case '/': return '\\';
1008 case '<': return '{';
1009 case '-': return '~';
1013 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1014 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1015 /// return the result character. Finally, emit a warning about trigraph use
1016 /// whether trigraphs are enabled or not.
1017 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1018 char Res = GetTrigraphCharForLetter(*CP);
1019 if (!Res || !L) return Res;
1021 if (!L->getFeatures().Trigraphs) {
1022 if (!L->isLexingRawMode())
1023 L->Diag(CP-2, diag::trigraph_ignored);
1027 if (!L->isLexingRawMode())
1028 L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1032 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1033 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1034 /// trigraph equivalent on entry to this function.
1035 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1037 while (isWhitespace(Ptr[Size])) {
1040 if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1043 // If this is a \r\n or \n\r, skip the other half.
1044 if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1045 Ptr[Size-1] != Ptr[Size])
1051 // Not an escaped newline, must be a \t or something else.
1055 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1056 /// them), skip over them and return the first non-escaped-newline found,
1057 /// otherwise return P.
1058 const char *Lexer::SkipEscapedNewLines(const char *P) {
1060 const char *AfterEscape;
1063 } else if (*P == '?') {
1064 // If not a trigraph for escape, bail out.
1065 if (P[1] != '?' || P[2] != '/')
1072 unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1073 if (NewLineSize == 0) return P;
1074 P = AfterEscape+NewLineSize;
1078 /// \brief Checks that the given token is the first token that occurs after the
1079 /// given location (this excludes comments and whitespace). Returns the location
1080 /// immediately after the specified token. If the token is not found or the
1081 /// location is inside a macro, the returned source location will be invalid.
1082 SourceLocation Lexer::findLocationAfterToken(SourceLocation Loc,
1083 tok::TokenKind TKind,
1084 const SourceManager &SM,
1085 const LangOptions &LangOpts,
1086 bool SkipTrailingWhitespaceAndNewLine) {
1087 if (Loc.isMacroID()) {
1088 if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts))
1089 return SourceLocation();
1090 Loc = SM.getExpansionRange(Loc).second;
1092 Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1094 // Break down the source location.
1095 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1097 // Try to load the file buffer.
1098 bool InvalidTemp = false;
1099 llvm::StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1101 return SourceLocation();
1103 const char *TokenBegin = File.data() + LocInfo.second;
1105 // Lex from the start of the given location.
1106 Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1107 TokenBegin, File.end());
1110 lexer.LexFromRawLexer(Tok);
1111 if (Tok.isNot(TKind))
1112 return SourceLocation();
1113 SourceLocation TokenLoc = Tok.getLocation();
1115 // Calculate how much whitespace needs to be skipped if any.
1116 unsigned NumWhitespaceChars = 0;
1117 if (SkipTrailingWhitespaceAndNewLine) {
1118 const char *TokenEnd = SM.getCharacterData(TokenLoc) +
1120 unsigned char C = *TokenEnd;
1121 while (isHorizontalWhitespace(C)) {
1123 NumWhitespaceChars++;
1125 if (isVerticalWhitespace(C))
1126 NumWhitespaceChars++;
1129 return TokenLoc.getLocWithOffset(Tok.getLength() + NumWhitespaceChars);
1132 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1133 /// get its size, and return it. This is tricky in several cases:
1134 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1135 /// then either return the trigraph (skipping 3 chars) or the '?',
1136 /// depending on whether trigraphs are enabled or not.
1137 /// 2. If this is an escaped newline (potentially with whitespace between
1138 /// the backslash and newline), implicitly skip the newline and return
1139 /// the char after it.
1140 /// 3. If this is a UCN, return it. FIXME: C++ UCN's?
1142 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1143 /// know that we can accumulate into Size, and that we have already incremented
1144 /// Ptr by Size bytes.
1146 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1147 /// be updated to match.
1149 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1151 // If we have a slash, look for an escaped newline.
1152 if (Ptr[0] == '\\') {
1156 // Common case, backslash-char where the char is not whitespace.
1157 if (!isWhitespace(Ptr[0])) return '\\';
1159 // See if we have optional whitespace characters between the slash and
1161 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1162 // Remember that this token needs to be cleaned.
1163 if (Tok) Tok->setFlag(Token::NeedsCleaning);
1165 // Warn if there was whitespace between the backslash and newline.
1166 if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1167 Diag(Ptr, diag::backslash_newline_space);
1169 // Found backslash<whitespace><newline>. Parse the char after it.
1170 Size += EscapedNewLineSize;
1171 Ptr += EscapedNewLineSize;
1172 // Use slow version to accumulate a correct size field.
1173 return getCharAndSizeSlow(Ptr, Size, Tok);
1176 // Otherwise, this is not an escaped newline, just return the slash.
1180 // If this is a trigraph, process it.
1181 if (Ptr[0] == '?' && Ptr[1] == '?') {
1182 // If this is actually a legal trigraph (not something like "??x"), emit
1183 // a trigraph warning. If so, and if trigraphs are enabled, return it.
1184 if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : 0)) {
1185 // Remember that this token needs to be cleaned.
1186 if (Tok) Tok->setFlag(Token::NeedsCleaning);
1190 if (C == '\\') goto Slash;
1195 // If this is neither, return a single character.
1201 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1202 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1203 /// and that we have already incremented Ptr by Size bytes.
1205 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1206 /// be updated to match.
1207 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1208 const LangOptions &Features) {
1209 // If we have a slash, look for an escaped newline.
1210 if (Ptr[0] == '\\') {
1214 // Common case, backslash-char where the char is not whitespace.
1215 if (!isWhitespace(Ptr[0])) return '\\';
1217 // See if we have optional whitespace characters followed by a newline.
1218 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1219 // Found backslash<whitespace><newline>. Parse the char after it.
1220 Size += EscapedNewLineSize;
1221 Ptr += EscapedNewLineSize;
1223 // Use slow version to accumulate a correct size field.
1224 return getCharAndSizeSlowNoWarn(Ptr, Size, Features);
1227 // Otherwise, this is not an escaped newline, just return the slash.
1231 // If this is a trigraph, process it.
1232 if (Features.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1233 // If this is actually a legal trigraph (not something like "??x"), return
1235 if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1238 if (C == '\\') goto Slash;
1243 // If this is neither, return a single character.
1248 //===----------------------------------------------------------------------===//
1249 // Helper methods for lexing.
1250 //===----------------------------------------------------------------------===//
1252 /// \brief Routine that indiscriminately skips bytes in the source file.
1253 void Lexer::SkipBytes(unsigned Bytes, bool StartOfLine) {
1255 if (BufferPtr > BufferEnd)
1256 BufferPtr = BufferEnd;
1257 IsAtStartOfLine = StartOfLine;
1260 void Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1261 // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1263 unsigned char C = *CurPtr++;
1264 while (isIdentifierBody(C))
1267 --CurPtr; // Back up over the skipped character.
1269 // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1270 // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1273 // TODO: Could merge these checks into a CharInfo flag to make the comparison
1275 if (C != '\\' && C != '?' && (C != '$' || !Features.DollarIdents)) {
1277 const char *IdStart = BufferPtr;
1278 FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1279 Result.setRawIdentifierData(IdStart);
1281 // If we are in raw mode, return this identifier raw. There is no need to
1282 // look up identifier information or attempt to macro expand it.
1286 // Fill in Result.IdentifierInfo and update the token kind,
1287 // looking up the identifier in the identifier table.
1288 IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1290 // Finally, now that we know we have an identifier, pass this off to the
1291 // preprocessor, which may macro expand it or something.
1292 if (II->isHandleIdentifierCase())
1293 PP->HandleIdentifier(Result);
1298 // Otherwise, $,\,? in identifier found. Enter slower path.
1300 C = getCharAndSize(CurPtr, Size);
1303 // If we hit a $ and they are not supported in identifiers, we are done.
1304 if (!Features.DollarIdents) goto FinishIdentifier;
1306 // Otherwise, emit a diagnostic and continue.
1307 if (!isLexingRawMode())
1308 Diag(CurPtr, diag::ext_dollar_in_identifier);
1309 CurPtr = ConsumeChar(CurPtr, Size, Result);
1310 C = getCharAndSize(CurPtr, Size);
1312 } else if (!isIdentifierBody(C)) { // FIXME: UCNs.
1313 // Found end of identifier.
1314 goto FinishIdentifier;
1317 // Otherwise, this character is good, consume it.
1318 CurPtr = ConsumeChar(CurPtr, Size, Result);
1320 C = getCharAndSize(CurPtr, Size);
1321 while (isIdentifierBody(C)) { // FIXME: UCNs.
1322 CurPtr = ConsumeChar(CurPtr, Size, Result);
1323 C = getCharAndSize(CurPtr, Size);
1328 /// isHexaLiteral - Return true if Start points to a hex constant.
1329 /// in microsoft mode (where this is supposed to be several different tokens).
1330 static bool isHexaLiteral(const char *Start, const LangOptions &Features) {
1332 char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, Features);
1335 char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, Features);
1336 return (C2 == 'x' || C2 == 'X');
1339 /// LexNumericConstant - Lex the remainder of a integer or floating point
1340 /// constant. From[-1] is the first character lexed. Return the end of the
1342 void Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1344 char C = getCharAndSize(CurPtr, Size);
1346 while (isNumberBody(C)) { // FIXME: UCNs?
1347 CurPtr = ConsumeChar(CurPtr, Size, Result);
1349 C = getCharAndSize(CurPtr, Size);
1352 // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1353 if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1354 // If we are in Microsoft mode, don't continue if the constant is hex.
1355 // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1356 if (!Features.MicrosoftExt || !isHexaLiteral(BufferPtr, Features))
1357 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1360 // If we have a hex FP constant, continue.
1361 if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p'))
1362 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1364 // Update the location of token as well as BufferPtr.
1365 const char *TokStart = BufferPtr;
1366 FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1367 Result.setLiteralData(TokStart);
1370 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1371 /// either " or L" or u8" or u" or U".
1372 void Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1373 tok::TokenKind Kind) {
1374 const char *NulCharacter = 0; // Does this string contain the \0 character?
1376 if (!isLexingRawMode() &&
1377 (Kind == tok::utf8_string_literal ||
1378 Kind == tok::utf16_string_literal ||
1379 Kind == tok::utf32_string_literal))
1380 Diag(BufferPtr, diag::warn_cxx98_compat_unicode_literal);
1382 char C = getAndAdvanceChar(CurPtr, Result);
1384 // Skip escaped characters. Escaped newlines will already be processed by
1385 // getAndAdvanceChar.
1387 C = getAndAdvanceChar(CurPtr, Result);
1389 if (C == '\n' || C == '\r' || // Newline.
1390 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1391 if (!isLexingRawMode() && !Features.AsmPreprocessor)
1392 Diag(BufferPtr, diag::warn_unterminated_string);
1393 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1398 if (isCodeCompletionPoint(CurPtr-1)) {
1399 PP->CodeCompleteNaturalLanguage();
1400 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1401 return cutOffLexing();
1404 NulCharacter = CurPtr-1;
1406 C = getAndAdvanceChar(CurPtr, Result);
1409 // If a nul character existed in the string, warn about it.
1410 if (NulCharacter && !isLexingRawMode())
1411 Diag(NulCharacter, diag::null_in_string);
1413 // Update the location of the token as well as the BufferPtr instance var.
1414 const char *TokStart = BufferPtr;
1415 FormTokenWithChars(Result, CurPtr, Kind);
1416 Result.setLiteralData(TokStart);
1419 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1420 /// having lexed R", LR", u8R", uR", or UR".
1421 void Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1422 tok::TokenKind Kind) {
1423 // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1424 // Between the initial and final double quote characters of the raw string,
1425 // any transformations performed in phases 1 and 2 (trigraphs,
1426 // universal-character-names, and line splicing) are reverted.
1428 if (!isLexingRawMode())
1429 Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1431 unsigned PrefixLen = 0;
1433 while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1436 // If the last character was not a '(', then we didn't lex a valid delimiter.
1437 if (CurPtr[PrefixLen] != '(') {
1438 if (!isLexingRawMode()) {
1439 const char *PrefixEnd = &CurPtr[PrefixLen];
1440 if (PrefixLen == 16) {
1441 Diag(PrefixEnd, diag::err_raw_delim_too_long);
1443 Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1444 << StringRef(PrefixEnd, 1);
1448 // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1449 // it's possible the '"' was intended to be part of the raw string, but
1450 // there's not much we can do about that.
1456 if (C == 0 && CurPtr-1 == BufferEnd) {
1462 FormTokenWithChars(Result, CurPtr, tok::unknown);
1466 // Save prefix and move CurPtr past it
1467 const char *Prefix = CurPtr;
1468 CurPtr += PrefixLen + 1; // skip over prefix and '('
1474 // Check for prefix match and closing quote.
1475 if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1476 CurPtr += PrefixLen + 1; // skip over prefix and '"'
1479 } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1480 if (!isLexingRawMode())
1481 Diag(BufferPtr, diag::err_unterminated_raw_string)
1482 << StringRef(Prefix, PrefixLen);
1483 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1488 // Update the location of token as well as BufferPtr.
1489 const char *TokStart = BufferPtr;
1490 FormTokenWithChars(Result, CurPtr, Kind);
1491 Result.setLiteralData(TokStart);
1494 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
1495 /// after having lexed the '<' character. This is used for #include filenames.
1496 void Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
1497 const char *NulCharacter = 0; // Does this string contain the \0 character?
1498 const char *AfterLessPos = CurPtr;
1499 char C = getAndAdvanceChar(CurPtr, Result);
1501 // Skip escaped characters.
1503 // Skip the escaped character.
1504 C = getAndAdvanceChar(CurPtr, Result);
1505 } else if (C == '\n' || C == '\r' || // Newline.
1506 (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
1507 isCodeCompletionPoint(CurPtr-1)))) {
1508 // If the filename is unterminated, then it must just be a lone <
1509 // character. Return this as such.
1510 FormTokenWithChars(Result, AfterLessPos, tok::less);
1512 } else if (C == 0) {
1513 NulCharacter = CurPtr-1;
1515 C = getAndAdvanceChar(CurPtr, Result);
1518 // If a nul character existed in the string, warn about it.
1519 if (NulCharacter && !isLexingRawMode())
1520 Diag(NulCharacter, diag::null_in_string);
1522 // Update the location of token as well as BufferPtr.
1523 const char *TokStart = BufferPtr;
1524 FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
1525 Result.setLiteralData(TokStart);
1529 /// LexCharConstant - Lex the remainder of a character constant, after having
1530 /// lexed either ' or L' or u' or U'.
1531 void Lexer::LexCharConstant(Token &Result, const char *CurPtr,
1532 tok::TokenKind Kind) {
1533 const char *NulCharacter = 0; // Does this character contain the \0 character?
1535 if (!isLexingRawMode() &&
1536 (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant))
1537 Diag(BufferPtr, diag::warn_cxx98_compat_unicode_literal);
1539 char C = getAndAdvanceChar(CurPtr, Result);
1541 if (!isLexingRawMode() && !Features.AsmPreprocessor)
1542 Diag(BufferPtr, diag::err_empty_character);
1543 FormTokenWithChars(Result, CurPtr, tok::unknown);
1548 // Skip escaped characters.
1550 // Skip the escaped character.
1552 C = getAndAdvanceChar(CurPtr, Result);
1553 } else if (C == '\n' || C == '\r' || // Newline.
1554 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1555 if (!isLexingRawMode() && !Features.AsmPreprocessor)
1556 Diag(BufferPtr, diag::warn_unterminated_char);
1557 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1559 } else if (C == 0) {
1560 if (isCodeCompletionPoint(CurPtr-1)) {
1561 PP->CodeCompleteNaturalLanguage();
1562 FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1563 return cutOffLexing();
1566 NulCharacter = CurPtr-1;
1568 C = getAndAdvanceChar(CurPtr, Result);
1571 // If a nul character existed in the character, warn about it.
1572 if (NulCharacter && !isLexingRawMode())
1573 Diag(NulCharacter, diag::null_in_char);
1575 // Update the location of token as well as BufferPtr.
1576 const char *TokStart = BufferPtr;
1577 FormTokenWithChars(Result, CurPtr, Kind);
1578 Result.setLiteralData(TokStart);
1581 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
1582 /// Update BufferPtr to point to the next non-whitespace character and return.
1584 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
1586 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr) {
1587 // Whitespace - Skip it, then return the token after the whitespace.
1588 unsigned char Char = *CurPtr; // Skip consequtive spaces efficiently.
1590 // Skip horizontal whitespace very aggressively.
1591 while (isHorizontalWhitespace(Char))
1594 // Otherwise if we have something other than whitespace, we're done.
1595 if (Char != '\n' && Char != '\r')
1598 if (ParsingPreprocessorDirective) {
1599 // End of preprocessor directive line, let LexTokenInternal handle this.
1604 // ok, but handle newline.
1605 // The returned token is at the start of the line.
1606 Result.setFlag(Token::StartOfLine);
1607 // No leading whitespace seen so far.
1608 Result.clearFlag(Token::LeadingSpace);
1612 // If this isn't immediately after a newline, there is leading space.
1613 char PrevChar = CurPtr[-1];
1614 if (PrevChar != '\n' && PrevChar != '\r')
1615 Result.setFlag(Token::LeadingSpace);
1617 // If the client wants us to return whitespace, return it now.
1618 if (isKeepWhitespaceMode()) {
1619 FormTokenWithChars(Result, CurPtr, tok::unknown);
1627 // SkipBCPLComment - We have just read the // characters from input. Skip until
1628 // we find the newline character thats terminate the comment. Then update
1629 /// BufferPtr and return.
1631 /// If we're in KeepCommentMode or any CommentHandler has inserted
1632 /// some tokens, this will store the first token and return true.
1633 bool Lexer::SkipBCPLComment(Token &Result, const char *CurPtr) {
1634 // If BCPL comments aren't explicitly enabled for this language, emit an
1635 // extension warning.
1636 if (!Features.BCPLComment && !isLexingRawMode()) {
1637 Diag(BufferPtr, diag::ext_bcpl_comment);
1639 // Mark them enabled so we only emit one warning for this translation
1641 Features.BCPLComment = true;
1644 // Scan over the body of the comment. The common case, when scanning, is that
1645 // the comment contains normal ascii characters with nothing interesting in
1646 // them. As such, optimize for this case with the inner loop.
1650 // Skip over characters in the fast loop.
1651 while (C != 0 && // Potentially EOF.
1652 C != '\n' && C != '\r') // Newline or DOS-style newline.
1655 const char *NextLine = CurPtr;
1657 // We found a newline, see if it's escaped.
1658 const char *EscapePtr = CurPtr-1;
1659 while (isHorizontalWhitespace(*EscapePtr)) // Skip whitespace.
1662 if (*EscapePtr == '\\') // Escaped newline.
1664 else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
1665 EscapePtr[-2] == '?') // Trigraph-escaped newline.
1666 CurPtr = EscapePtr-2;
1668 break; // This is a newline, we're done.
1673 // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
1674 // properly decode the character. Read it in raw mode to avoid emitting
1675 // diagnostics about things like trigraphs. If we see an escaped newline,
1676 // we'll handle it below.
1677 const char *OldPtr = CurPtr;
1678 bool OldRawMode = isLexingRawMode();
1679 LexingRawMode = true;
1680 C = getAndAdvanceChar(CurPtr, Result);
1681 LexingRawMode = OldRawMode;
1683 // If we only read only one character, then no special handling is needed.
1684 // We're done and can skip forward to the newline.
1685 if (C != 0 && CurPtr == OldPtr+1) {
1690 // If the char that we finally got was a \n, then we must have had something
1691 // like \<newline><newline>. We don't want to have consumed the second
1692 // newline, we want CurPtr, to end up pointing to it down below.
1693 if (C == '\n' || C == '\r') {
1695 C = 'x'; // doesn't matter what this is.
1698 // If we read multiple characters, and one of those characters was a \r or
1699 // \n, then we had an escaped newline within the comment. Emit diagnostic
1700 // unless the next line is also a // comment.
1701 if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') {
1702 for (; OldPtr != CurPtr; ++OldPtr)
1703 if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
1704 // Okay, we found a // comment that ends in a newline, if the next
1705 // line is also a // comment, but has spaces, don't emit a diagnostic.
1706 if (isWhitespace(C)) {
1707 const char *ForwardPtr = CurPtr;
1708 while (isWhitespace(*ForwardPtr)) // Skip whitespace.
1710 if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
1714 if (!isLexingRawMode())
1715 Diag(OldPtr-1, diag::ext_multi_line_bcpl_comment);
1720 if (CurPtr == BufferEnd+1) {
1725 if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
1726 PP->CodeCompleteNaturalLanguage();
1731 } while (C != '\n' && C != '\r');
1733 // Found but did not consume the newline. Notify comment handlers about the
1734 // comment unless we're in a #if 0 block.
1735 if (PP && !isLexingRawMode() &&
1736 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
1737 getSourceLocation(CurPtr)))) {
1739 return true; // A token has to be returned.
1742 // If we are returning comments as tokens, return this comment as a token.
1743 if (inKeepCommentMode())
1744 return SaveBCPLComment(Result, CurPtr);
1746 // If we are inside a preprocessor directive and we see the end of line,
1747 // return immediately, so that the lexer can return this as an EOD token.
1748 if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
1753 // Otherwise, eat the \n character. We don't care if this is a \n\r or
1754 // \r\n sequence. This is an efficiency hack (because we know the \n can't
1755 // contribute to another token), it isn't needed for correctness. Note that
1756 // this is ok even in KeepWhitespaceMode, because we would have returned the
1757 /// comment above in that mode.
1760 // The next returned token is at the start of the line.
1761 Result.setFlag(Token::StartOfLine);
1762 // No leading whitespace seen so far.
1763 Result.clearFlag(Token::LeadingSpace);
1768 /// SaveBCPLComment - If in save-comment mode, package up this BCPL comment in
1769 /// an appropriate way and return it.
1770 bool Lexer::SaveBCPLComment(Token &Result, const char *CurPtr) {
1771 // If we're not in a preprocessor directive, just return the // comment
1773 FormTokenWithChars(Result, CurPtr, tok::comment);
1775 if (!ParsingPreprocessorDirective)
1778 // If this BCPL-style comment is in a macro definition, transmogrify it into
1779 // a C-style block comment.
1780 bool Invalid = false;
1781 std::string Spelling = PP->getSpelling(Result, &Invalid);
1785 assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not bcpl comment?");
1786 Spelling[1] = '*'; // Change prefix to "/*".
1787 Spelling += "*/"; // add suffix.
1789 Result.setKind(tok::comment);
1790 PP->CreateString(&Spelling[0], Spelling.size(), Result,
1791 Result.getLocation(), Result.getLocation());
1795 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
1796 /// character (either \n or \r) is part of an escaped newline sequence. Issue a
1797 /// diagnostic if so. We know that the newline is inside of a block comment.
1798 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
1800 assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
1802 // Back up off the newline.
1805 // If this is a two-character newline sequence, skip the other character.
1806 if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
1807 // \n\n or \r\r -> not escaped newline.
1808 if (CurPtr[0] == CurPtr[1])
1810 // \n\r or \r\n -> skip the newline.
1814 // If we have horizontal whitespace, skip over it. We allow whitespace
1815 // between the slash and newline.
1816 bool HasSpace = false;
1817 while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
1822 // If we have a slash, we know this is an escaped newline.
1823 if (*CurPtr == '\\') {
1824 if (CurPtr[-1] != '*') return false;
1826 // It isn't a slash, is it the ?? / trigraph?
1827 if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
1831 // This is the trigraph ending the comment. Emit a stern warning!
1834 // If no trigraphs are enabled, warn that we ignored this trigraph and
1835 // ignore this * character.
1836 if (!L->getFeatures().Trigraphs) {
1837 if (!L->isLexingRawMode())
1838 L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
1841 if (!L->isLexingRawMode())
1842 L->Diag(CurPtr, diag::trigraph_ends_block_comment);
1845 // Warn about having an escaped newline between the */ characters.
1846 if (!L->isLexingRawMode())
1847 L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
1849 // If there was space between the backslash and newline, warn about it.
1850 if (HasSpace && !L->isLexingRawMode())
1851 L->Diag(CurPtr, diag::backslash_newline_space);
1857 #include <emmintrin.h>
1859 #include <altivec.h>
1863 /// SkipBlockComment - We have just read the /* characters from input. Read
1864 /// until we find the */ characters that terminate the comment. Note that we
1865 /// don't bother decoding trigraphs or escaped newlines in block comments,
1866 /// because they cannot cause the comment to end. The only thing that can
1867 /// happen is the comment could end with an escaped newline between the */ end
1870 /// If we're in KeepCommentMode or any CommentHandler has inserted
1871 /// some tokens, this will store the first token and return true.
1872 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr) {
1873 // Scan one character past where we should, looking for a '/' character. Once
1874 // we find it, check to see if it was preceded by a *. This common
1875 // optimization helps people who like to put a lot of * characters in their
1878 // The first character we get with newlines and trigraphs skipped to handle
1879 // the degenerate /*/ case below correctly if the * has an escaped newline
1882 unsigned char C = getCharAndSize(CurPtr, CharSize);
1884 if (C == 0 && CurPtr == BufferEnd+1) {
1885 if (!isLexingRawMode())
1886 Diag(BufferPtr, diag::err_unterminated_block_comment);
1889 // KeepWhitespaceMode should return this broken comment as a token. Since
1890 // it isn't a well formed comment, just return it as an 'unknown' token.
1891 if (isKeepWhitespaceMode()) {
1892 FormTokenWithChars(Result, CurPtr, tok::unknown);
1900 // Check to see if the first character after the '/*' is another /. If so,
1901 // then this slash does not end the block comment, it is part of it.
1906 // Skip over all non-interesting characters until we find end of buffer or a
1907 // (probably ending) '/' character.
1908 if (CurPtr + 24 < BufferEnd &&
1909 // If there is a code-completion point avoid the fast scan because it
1910 // doesn't check for '\0'.
1911 !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
1912 // While not aligned to a 16-byte boundary.
1913 while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
1916 if (C == '/') goto FoundSlash;
1919 __m128i Slashes = _mm_set_epi8('/', '/', '/', '/', '/', '/', '/', '/',
1920 '/', '/', '/', '/', '/', '/', '/', '/');
1921 while (CurPtr+16 <= BufferEnd &&
1922 _mm_movemask_epi8(_mm_cmpeq_epi8(*(__m128i*)CurPtr, Slashes)) == 0)
1925 __vector unsigned char Slashes = {
1926 '/', '/', '/', '/', '/', '/', '/', '/',
1927 '/', '/', '/', '/', '/', '/', '/', '/'
1929 while (CurPtr+16 <= BufferEnd &&
1930 !vec_any_eq(*(vector unsigned char*)CurPtr, Slashes))
1933 // Scan for '/' quickly. Many block comments are very large.
1934 while (CurPtr[0] != '/' &&
1938 CurPtr+4 < BufferEnd) {
1943 // It has to be one of the bytes scanned, increment to it and read one.
1947 // Loop to scan the remainder.
1948 while (C != '/' && C != '\0')
1953 if (CurPtr[-2] == '*') // We found the final */. We're done!
1956 if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
1957 if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
1958 // We found the final */, though it had an escaped newline between the
1959 // * and /. We're done!
1963 if (CurPtr[0] == '*' && CurPtr[1] != '/') {
1964 // If this is a /* inside of the comment, emit a warning. Don't do this
1965 // if this is a /*/, which will end the comment. This misses cases with
1966 // embedded escaped newlines, but oh well.
1967 if (!isLexingRawMode())
1968 Diag(CurPtr-1, diag::warn_nested_block_comment);
1970 } else if (C == 0 && CurPtr == BufferEnd+1) {
1971 if (!isLexingRawMode())
1972 Diag(BufferPtr, diag::err_unterminated_block_comment);
1973 // Note: the user probably forgot a */. We could continue immediately
1974 // after the /*, but this would involve lexing a lot of what really is the
1975 // comment, which surely would confuse the parser.
1978 // KeepWhitespaceMode should return this broken comment as a token. Since
1979 // it isn't a well formed comment, just return it as an 'unknown' token.
1980 if (isKeepWhitespaceMode()) {
1981 FormTokenWithChars(Result, CurPtr, tok::unknown);
1987 } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
1988 PP->CodeCompleteNaturalLanguage();
1996 // Notify comment handlers about the comment unless we're in a #if 0 block.
1997 if (PP && !isLexingRawMode() &&
1998 PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
1999 getSourceLocation(CurPtr)))) {
2001 return true; // A token has to be returned.
2004 // If we are returning comments as tokens, return this comment as a token.
2005 if (inKeepCommentMode()) {
2006 FormTokenWithChars(Result, CurPtr, tok::comment);
2010 // It is common for the tokens immediately after a /**/ comment to be
2011 // whitespace. Instead of going through the big switch, handle it
2012 // efficiently now. This is safe even in KeepWhitespaceMode because we would
2013 // have already returned above with the comment as a token.
2014 if (isHorizontalWhitespace(*CurPtr)) {
2015 Result.setFlag(Token::LeadingSpace);
2016 SkipWhitespace(Result, CurPtr+1);
2020 // Otherwise, just return so that the next character will be lexed as a token.
2022 Result.setFlag(Token::LeadingSpace);
2026 //===----------------------------------------------------------------------===//
2027 // Primary Lexing Entry Points
2028 //===----------------------------------------------------------------------===//
2030 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2031 /// uninterpreted string. This switches the lexer out of directive mode.
2032 std::string Lexer::ReadToEndOfLine() {
2033 assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2034 "Must be in a preprocessing directive!");
2038 // CurPtr - Cache BufferPtr in an automatic variable.
2039 const char *CurPtr = BufferPtr;
2041 char Char = getAndAdvanceChar(CurPtr, Tmp);
2047 // Found end of file?
2048 if (CurPtr-1 != BufferEnd) {
2049 if (isCodeCompletionPoint(CurPtr-1)) {
2050 PP->CodeCompleteNaturalLanguage();
2055 // Nope, normal character, continue.
2062 // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2063 assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2064 BufferPtr = CurPtr-1;
2066 // Next, lex the character, which should handle the EOD transition.
2068 if (Tmp.is(tok::code_completion)) {
2070 PP->CodeCompleteNaturalLanguage();
2073 assert(Tmp.is(tok::eod) && "Unexpected token!");
2075 // Finally, we're done, return the string we found.
2081 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2082 /// condition, reporting diagnostics and handling other edge cases as required.
2083 /// This returns true if Result contains a token, false if PP.Lex should be
2085 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2086 // If we hit the end of the file while parsing a preprocessor directive,
2087 // end the preprocessor directive first. The next token returned will
2088 // then be the end of file.
2089 if (ParsingPreprocessorDirective) {
2090 // Done parsing the "line".
2091 ParsingPreprocessorDirective = false;
2092 // Update the location of token as well as BufferPtr.
2093 FormTokenWithChars(Result, CurPtr, tok::eod);
2095 // Restore comment saving mode, in case it was disabled for directive.
2096 SetCommentRetentionState(PP->getCommentRetentionState());
2097 return true; // Have a token.
2100 // If we are in raw mode, return this event as an EOF token. Let the caller
2101 // that put us in raw mode handle the event.
2102 if (isLexingRawMode()) {
2103 Result.startToken();
2104 BufferPtr = BufferEnd;
2105 FormTokenWithChars(Result, BufferEnd, tok::eof);
2109 // Issue diagnostics for unterminated #if and missing newline.
2111 // If we are in a #if directive, emit an error.
2112 while (!ConditionalStack.empty()) {
2113 if (PP->getCodeCompletionFileLoc() != FileLoc)
2114 PP->Diag(ConditionalStack.back().IfLoc,
2115 diag::err_pp_unterminated_conditional);
2116 ConditionalStack.pop_back();
2119 // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2121 if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r'))
2122 Diag(BufferEnd, diag::ext_no_newline_eof)
2123 << FixItHint::CreateInsertion(getSourceLocation(BufferEnd), "\n");
2127 // Finally, let the preprocessor handle this.
2128 return PP->HandleEndOfFile(Result);
2131 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2132 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2133 /// else and 2 if there are no more tokens in the buffer controlled by the
2135 unsigned Lexer::isNextPPTokenLParen() {
2136 assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2138 // Switch to 'skipping' mode. This will ensure that we can lex a token
2139 // without emitting diagnostics, disables macro expansion, and will cause EOF
2140 // to return an EOF token instead of popping the include stack.
2141 LexingRawMode = true;
2143 // Save state that can be changed while lexing so that we can restore it.
2144 const char *TmpBufferPtr = BufferPtr;
2145 bool inPPDirectiveMode = ParsingPreprocessorDirective;
2149 LexTokenInternal(Tok);
2151 // Restore state that may have changed.
2152 BufferPtr = TmpBufferPtr;
2153 ParsingPreprocessorDirective = inPPDirectiveMode;
2155 // Restore the lexer back to non-skipping mode.
2156 LexingRawMode = false;
2158 if (Tok.is(tok::eof))
2160 return Tok.is(tok::l_paren);
2163 /// FindConflictEnd - Find the end of a version control conflict marker.
2164 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2165 ConflictMarkerKind CMK) {
2166 const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2167 size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2168 StringRef RestOfBuffer(CurPtr+TermLen, BufferEnd-CurPtr-TermLen);
2169 size_t Pos = RestOfBuffer.find(Terminator);
2170 while (Pos != StringRef::npos) {
2171 // Must occur at start of line.
2172 if (RestOfBuffer[Pos-1] != '\r' &&
2173 RestOfBuffer[Pos-1] != '\n') {
2174 RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2175 Pos = RestOfBuffer.find(Terminator);
2178 return RestOfBuffer.data()+Pos;
2183 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2184 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2185 /// and recover nicely. This returns true if it is a conflict marker and false
2187 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2188 // Only a conflict marker if it starts at the beginning of a line.
2189 if (CurPtr != BufferStart &&
2190 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2193 // Check to see if we have <<<<<<< or >>>>.
2194 if ((BufferEnd-CurPtr < 8 || StringRef(CurPtr, 7) != "<<<<<<<") &&
2195 (BufferEnd-CurPtr < 6 || StringRef(CurPtr, 5) != ">>>> "))
2198 // If we have a situation where we don't care about conflict markers, ignore
2200 if (CurrentConflictMarkerState || isLexingRawMode())
2203 ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2205 // Check to see if there is an ending marker somewhere in the buffer at the
2206 // start of a line to terminate this conflict marker.
2207 if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2208 // We found a match. We are really in a conflict marker.
2209 // Diagnose this, and ignore to the end of line.
2210 Diag(CurPtr, diag::err_conflict_marker);
2211 CurrentConflictMarkerState = Kind;
2213 // Skip ahead to the end of line. We know this exists because the
2214 // end-of-conflict marker starts with \r or \n.
2215 while (*CurPtr != '\r' && *CurPtr != '\n') {
2216 assert(CurPtr != BufferEnd && "Didn't find end of line");
2223 // No end of conflict marker found.
2228 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2229 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2230 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2231 /// the line. This returns true if it is a conflict marker and false if not.
2232 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2233 // Only a conflict marker if it starts at the beginning of a line.
2234 if (CurPtr != BufferStart &&
2235 CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2238 // If we have a situation where we don't care about conflict markers, ignore
2240 if (!CurrentConflictMarkerState || isLexingRawMode())
2243 // Check to see if we have the marker (4 characters in a row).
2244 for (unsigned i = 1; i != 4; ++i)
2245 if (CurPtr[i] != CurPtr[0])
2248 // If we do have it, search for the end of the conflict marker. This could
2249 // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2250 // be the end of conflict marker.
2251 if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2252 CurrentConflictMarkerState)) {
2255 // Skip ahead to the end of line.
2256 while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2261 // No longer in the conflict marker.
2262 CurrentConflictMarkerState = CMK_None;
2269 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2270 if (PP && PP->isCodeCompletionEnabled()) {
2271 SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2272 return Loc == PP->getCodeCompletionLoc();
2279 /// LexTokenInternal - This implements a simple C family lexer. It is an
2280 /// extremely performance critical piece of code. This assumes that the buffer
2281 /// has a null character at the end of the file. This returns a preprocessing
2282 /// token, not a normal token, as such, it is an internal interface. It assumes
2283 /// that the Flags of result have been cleared before calling this.
2284 void Lexer::LexTokenInternal(Token &Result) {
2286 // New token, can't need cleaning yet.
2287 Result.clearFlag(Token::NeedsCleaning);
2288 Result.setIdentifierInfo(0);
2290 // CurPtr - Cache BufferPtr in an automatic variable.
2291 const char *CurPtr = BufferPtr;
2293 // Small amounts of horizontal whitespace is very common between tokens.
2294 if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
2296 while ((*CurPtr == ' ') || (*CurPtr == '\t'))
2299 // If we are keeping whitespace and other tokens, just return what we just
2300 // skipped. The next lexer invocation will return the token after the
2302 if (isKeepWhitespaceMode()) {
2303 FormTokenWithChars(Result, CurPtr, tok::unknown);
2308 Result.setFlag(Token::LeadingSpace);
2311 unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
2313 // Read a character, advancing over it.
2314 char Char = getAndAdvanceChar(CurPtr, Result);
2315 tok::TokenKind Kind;
2319 // Found end of file?
2320 if (CurPtr-1 == BufferEnd) {
2321 // Read the PP instance variable into an automatic variable, because
2322 // LexEndOfFile will often delete 'this'.
2323 Preprocessor *PPCache = PP;
2324 if (LexEndOfFile(Result, CurPtr-1)) // Retreat back into the file.
2325 return; // Got a token to return.
2326 assert(PPCache && "Raw buffer::LexEndOfFile should return a token");
2327 return PPCache->Lex(Result);
2330 // Check if we are performing code completion.
2331 if (isCodeCompletionPoint(CurPtr-1)) {
2332 // Return the code-completion token.
2333 Result.startToken();
2334 FormTokenWithChars(Result, CurPtr, tok::code_completion);
2338 if (!isLexingRawMode())
2339 Diag(CurPtr-1, diag::null_in_file);
2340 Result.setFlag(Token::LeadingSpace);
2341 if (SkipWhitespace(Result, CurPtr))
2342 return; // KeepWhitespaceMode
2344 goto LexNextToken; // GCC isn't tail call eliminating.
2346 case 26: // DOS & CP/M EOF: "^Z".
2347 // If we're in Microsoft extensions mode, treat this as end of file.
2348 if (Features.MicrosoftExt) {
2349 // Read the PP instance variable into an automatic variable, because
2350 // LexEndOfFile will often delete 'this'.
2351 Preprocessor *PPCache = PP;
2352 if (LexEndOfFile(Result, CurPtr-1)) // Retreat back into the file.
2353 return; // Got a token to return.
2354 assert(PPCache && "Raw buffer::LexEndOfFile should return a token");
2355 return PPCache->Lex(Result);
2357 // If Microsoft extensions are disabled, this is just random garbage.
2358 Kind = tok::unknown;
2363 // If we are inside a preprocessor directive and we see the end of line,
2364 // we know we are done with the directive, so return an EOD token.
2365 if (ParsingPreprocessorDirective) {
2366 // Done parsing the "line".
2367 ParsingPreprocessorDirective = false;
2369 // Restore comment saving mode, in case it was disabled for directive.
2370 SetCommentRetentionState(PP->getCommentRetentionState());
2372 // Since we consumed a newline, we are back at the start of a line.
2373 IsAtStartOfLine = true;
2378 // The returned token is at the start of the line.
2379 Result.setFlag(Token::StartOfLine);
2380 // No leading whitespace seen so far.
2381 Result.clearFlag(Token::LeadingSpace);
2383 if (SkipWhitespace(Result, CurPtr))
2384 return; // KeepWhitespaceMode
2385 goto LexNextToken; // GCC isn't tail call eliminating.
2390 SkipHorizontalWhitespace:
2391 Result.setFlag(Token::LeadingSpace);
2392 if (SkipWhitespace(Result, CurPtr))
2393 return; // KeepWhitespaceMode
2398 // If the next token is obviously a // or /* */ comment, skip it efficiently
2399 // too (without going through the big switch stmt).
2400 if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
2401 Features.BCPLComment && !Features.TraditionalCPP) {
2402 if (SkipBCPLComment(Result, CurPtr+2))
2403 return; // There is a token to return.
2404 goto SkipIgnoredUnits;
2405 } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
2406 if (SkipBlockComment(Result, CurPtr+2))
2407 return; // There is a token to return.
2408 goto SkipIgnoredUnits;
2409 } else if (isHorizontalWhitespace(*CurPtr)) {
2410 goto SkipHorizontalWhitespace;
2412 goto LexNextToken; // GCC isn't tail call eliminating.
2414 // C99 6.4.4.1: Integer Constants.
2415 // C99 6.4.4.2: Floating Constants.
2416 case '0': case '1': case '2': case '3': case '4':
2417 case '5': case '6': case '7': case '8': case '9':
2418 // Notify MIOpt that we read a non-whitespace/non-comment token.
2420 return LexNumericConstant(Result, CurPtr);
2422 case 'u': // Identifier (uber) or C++0x UTF-8 or UTF-16 string literal
2423 // Notify MIOpt that we read a non-whitespace/non-comment token.
2426 if (Features.CPlusPlus0x) {
2427 Char = getCharAndSize(CurPtr, SizeTmp);
2429 // UTF-16 string literal
2431 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2432 tok::utf16_string_literal);
2434 // UTF-16 character constant
2436 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2437 tok::utf16_char_constant);
2439 // UTF-16 raw string literal
2440 if (Char == 'R' && getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
2441 return LexRawStringLiteral(Result,
2442 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2444 tok::utf16_string_literal);
2447 char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
2449 // UTF-8 string literal
2451 return LexStringLiteral(Result,
2452 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2454 tok::utf8_string_literal);
2458 char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
2459 // UTF-8 raw string literal
2461 return LexRawStringLiteral(Result,
2462 ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2465 tok::utf8_string_literal);
2471 // treat u like the start of an identifier.
2472 return LexIdentifier(Result, CurPtr);
2474 case 'U': // Identifier (Uber) or C++0x UTF-32 string literal
2475 // Notify MIOpt that we read a non-whitespace/non-comment token.
2478 if (Features.CPlusPlus0x) {
2479 Char = getCharAndSize(CurPtr, SizeTmp);
2481 // UTF-32 string literal
2483 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2484 tok::utf32_string_literal);
2486 // UTF-32 character constant
2488 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2489 tok::utf32_char_constant);
2491 // UTF-32 raw string literal
2492 if (Char == 'R' && getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
2493 return LexRawStringLiteral(Result,
2494 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2496 tok::utf32_string_literal);
2499 // treat U like the start of an identifier.
2500 return LexIdentifier(Result, CurPtr);
2502 case 'R': // Identifier or C++0x raw string literal
2503 // Notify MIOpt that we read a non-whitespace/non-comment token.
2506 if (Features.CPlusPlus0x) {
2507 Char = getCharAndSize(CurPtr, SizeTmp);
2510 return LexRawStringLiteral(Result,
2511 ConsumeChar(CurPtr, SizeTmp, Result),
2512 tok::string_literal);
2515 // treat R like the start of an identifier.
2516 return LexIdentifier(Result, CurPtr);
2518 case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
2519 // Notify MIOpt that we read a non-whitespace/non-comment token.
2521 Char = getCharAndSize(CurPtr, SizeTmp);
2523 // Wide string literal.
2525 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2526 tok::wide_string_literal);
2528 // Wide raw string literal.
2529 if (Features.CPlusPlus0x && Char == 'R' &&
2530 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
2531 return LexRawStringLiteral(Result,
2532 ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2534 tok::wide_string_literal);
2536 // Wide character constant.
2538 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
2539 tok::wide_char_constant);
2540 // FALL THROUGH, treating L like the start of an identifier.
2542 // C99 6.4.2: Identifiers.
2543 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
2544 case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
2545 case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
2546 case 'V': case 'W': case 'X': case 'Y': case 'Z':
2547 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
2548 case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
2549 case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
2550 case 'v': case 'w': case 'x': case 'y': case 'z':
2552 // Notify MIOpt that we read a non-whitespace/non-comment token.
2554 return LexIdentifier(Result, CurPtr);
2556 case '$': // $ in identifiers.
2557 if (Features.DollarIdents) {
2558 if (!isLexingRawMode())
2559 Diag(CurPtr-1, diag::ext_dollar_in_identifier);
2560 // Notify MIOpt that we read a non-whitespace/non-comment token.
2562 return LexIdentifier(Result, CurPtr);
2565 Kind = tok::unknown;
2568 // C99 6.4.4: Character Constants.
2570 // Notify MIOpt that we read a non-whitespace/non-comment token.
2572 return LexCharConstant(Result, CurPtr, tok::char_constant);
2574 // C99 6.4.5: String Literals.
2576 // Notify MIOpt that we read a non-whitespace/non-comment token.
2578 return LexStringLiteral(Result, CurPtr, tok::string_literal);
2580 // C99 6.4.6: Punctuators.
2582 Kind = tok::question;
2585 Kind = tok::l_square;
2588 Kind = tok::r_square;
2591 Kind = tok::l_paren;
2594 Kind = tok::r_paren;
2597 Kind = tok::l_brace;
2600 Kind = tok::r_brace;
2603 Char = getCharAndSize(CurPtr, SizeTmp);
2604 if (Char >= '0' && Char <= '9') {
2605 // Notify MIOpt that we read a non-whitespace/non-comment token.
2608 return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
2609 } else if (Features.CPlusPlus && Char == '*') {
2610 Kind = tok::periodstar;
2612 } else if (Char == '.' &&
2613 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
2614 Kind = tok::ellipsis;
2615 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2622 Char = getCharAndSize(CurPtr, SizeTmp);
2625 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2626 } else if (Char == '=') {
2627 Kind = tok::ampequal;
2628 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2634 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
2635 Kind = tok::starequal;
2636 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2642 Char = getCharAndSize(CurPtr, SizeTmp);
2644 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2645 Kind = tok::plusplus;
2646 } else if (Char == '=') {
2647 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2648 Kind = tok::plusequal;
2654 Char = getCharAndSize(CurPtr, SizeTmp);
2655 if (Char == '-') { // --
2656 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2657 Kind = tok::minusminus;
2658 } else if (Char == '>' && Features.CPlusPlus &&
2659 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
2660 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2662 Kind = tok::arrowstar;
2663 } else if (Char == '>') { // ->
2664 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2666 } else if (Char == '=') { // -=
2667 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2668 Kind = tok::minusequal;
2677 if (getCharAndSize(CurPtr, SizeTmp) == '=') {
2678 Kind = tok::exclaimequal;
2679 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2681 Kind = tok::exclaim;
2686 Char = getCharAndSize(CurPtr, SizeTmp);
2687 if (Char == '/') { // BCPL comment.
2688 // Even if BCPL comments are disabled (e.g. in C89 mode), we generally
2689 // want to lex this as a comment. There is one problem with this though,
2690 // that in one particular corner case, this can change the behavior of the
2691 // resultant program. For example, In "foo //**/ bar", C89 would lex
2692 // this as "foo / bar" and langauges with BCPL comments would lex it as
2693 // "foo". Check to see if the character after the second slash is a '*'.
2694 // If so, we will lex that as a "/" instead of the start of a comment.
2695 // However, we never do this in -traditional-cpp mode.
2696 if ((Features.BCPLComment ||
2697 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*') &&
2698 !Features.TraditionalCPP) {
2699 if (SkipBCPLComment(Result, ConsumeChar(CurPtr, SizeTmp, Result)))
2700 return; // There is a token to return.
2702 // It is common for the tokens immediately after a // comment to be
2703 // whitespace (indentation for the next line). Instead of going through
2704 // the big switch, handle it efficiently now.
2705 goto SkipIgnoredUnits;
2709 if (Char == '*') { // /**/ comment.
2710 if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result)))
2711 return; // There is a token to return.
2712 goto LexNextToken; // GCC isn't tail call eliminating.
2716 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2717 Kind = tok::slashequal;
2723 Char = getCharAndSize(CurPtr, SizeTmp);
2725 Kind = tok::percentequal;
2726 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2727 } else if (Features.Digraphs && Char == '>') {
2728 Kind = tok::r_brace; // '%>' -> '}'
2729 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2730 } else if (Features.Digraphs && Char == ':') {
2731 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2732 Char = getCharAndSize(CurPtr, SizeTmp);
2733 if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
2734 Kind = tok::hashhash; // '%:%:' -> '##'
2735 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2737 } else if (Char == '@' && Features.MicrosoftExt) {// %:@ -> #@ -> Charize
2738 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2739 if (!isLexingRawMode())
2740 Diag(BufferPtr, diag::charize_microsoft_ext);
2742 } else { // '%:' -> '#'
2743 // We parsed a # character. If this occurs at the start of the line,
2744 // it's actually the start of a preprocessing directive. Callback to
2745 // the preprocessor to handle it.
2746 // FIXME: -fpreprocessed mode??
2747 if (Result.isAtStartOfLine() && !LexingRawMode && !Is_PragmaLexer) {
2748 FormTokenWithChars(Result, CurPtr, tok::hash);
2749 PP->HandleDirective(Result);
2751 // As an optimization, if the preprocessor didn't switch lexers, tail
2753 if (PP->isCurrentLexer(this)) {
2754 // Start a new token. If this is a #include or something, the PP may
2755 // want us starting at the beginning of the line again. If so, set
2756 // the StartOfLine flag and clear LeadingSpace.
2757 if (IsAtStartOfLine) {
2758 Result.setFlag(Token::StartOfLine);
2759 Result.clearFlag(Token::LeadingSpace);
2760 IsAtStartOfLine = false;
2762 goto LexNextToken; // GCC isn't tail call eliminating.
2765 return PP->Lex(Result);
2771 Kind = tok::percent;
2775 Char = getCharAndSize(CurPtr, SizeTmp);
2776 if (ParsingFilename) {
2777 return LexAngledStringLiteral(Result, CurPtr);
2778 } else if (Char == '<') {
2779 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
2781 Kind = tok::lesslessequal;
2782 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2784 } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
2785 // If this is actually a '<<<<<<<' version control conflict marker,
2786 // recognize it as such and recover nicely.
2788 } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
2789 // If this is '<<<<' and we're in a Perforce-style conflict marker,
2792 } else if (Features.CUDA && After == '<') {
2793 Kind = tok::lesslessless;
2794 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2797 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2798 Kind = tok::lessless;
2800 } else if (Char == '=') {
2801 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2802 Kind = tok::lessequal;
2803 } else if (Features.Digraphs && Char == ':') { // '<:' -> '['
2804 if (Features.CPlusPlus0x &&
2805 getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
2806 // C++0x [lex.pptoken]p3:
2807 // Otherwise, if the next three characters are <:: and the subsequent
2808 // character is neither : nor >, the < is treated as a preprocessor
2809 // token by itself and not as the first character of the alternative
2812 char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
2813 if (After != ':' && After != '>') {
2815 if (!isLexingRawMode())
2816 Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
2821 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2822 Kind = tok::l_square;
2823 } else if (Features.Digraphs && Char == '%') { // '<%' -> '{'
2824 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2825 Kind = tok::l_brace;
2831 Char = getCharAndSize(CurPtr, SizeTmp);
2833 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2834 Kind = tok::greaterequal;
2835 } else if (Char == '>') {
2836 char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
2838 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2840 Kind = tok::greatergreaterequal;
2841 } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
2842 // If this is actually a '>>>>' conflict marker, recognize it as such
2843 // and recover nicely.
2845 } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
2846 // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
2848 } else if (Features.CUDA && After == '>') {
2849 Kind = tok::greatergreatergreater;
2850 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
2853 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2854 Kind = tok::greatergreater;
2858 Kind = tok::greater;
2862 Char = getCharAndSize(CurPtr, SizeTmp);
2864 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2865 Kind = tok::caretequal;
2871 Char = getCharAndSize(CurPtr, SizeTmp);
2873 Kind = tok::pipeequal;
2874 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2875 } else if (Char == '|') {
2876 // If this is '|||||||' and we're in a conflict marker, ignore it.
2877 if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
2879 Kind = tok::pipepipe;
2880 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2886 Char = getCharAndSize(CurPtr, SizeTmp);
2887 if (Features.Digraphs && Char == '>') {
2888 Kind = tok::r_square; // ':>' -> ']'
2889 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2890 } else if (Features.CPlusPlus && Char == ':') {
2891 Kind = tok::coloncolon;
2892 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2901 Char = getCharAndSize(CurPtr, SizeTmp);
2903 // If this is '====' and we're in a conflict marker, ignore it.
2904 if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
2907 Kind = tok::equalequal;
2908 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2917 Char = getCharAndSize(CurPtr, SizeTmp);
2919 Kind = tok::hashhash;
2920 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2921 } else if (Char == '@' && Features.MicrosoftExt) { // #@ -> Charize
2923 if (!isLexingRawMode())
2924 Diag(BufferPtr, diag::charize_microsoft_ext);
2925 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
2927 // We parsed a # character. If this occurs at the start of the line,
2928 // it's actually the start of a preprocessing directive. Callback to
2929 // the preprocessor to handle it.
2930 // FIXME: -fpreprocessed mode??
2931 if (Result.isAtStartOfLine() && !LexingRawMode && !Is_PragmaLexer) {
2932 FormTokenWithChars(Result, CurPtr, tok::hash);
2933 PP->HandleDirective(Result);
2935 // As an optimization, if the preprocessor didn't switch lexers, tail
2937 if (PP->isCurrentLexer(this)) {
2938 // Start a new token. If this is a #include or something, the PP may
2939 // want us starting at the beginning of the line again. If so, set
2940 // the StartOfLine flag and clear LeadingSpace.
2941 if (IsAtStartOfLine) {
2942 Result.setFlag(Token::StartOfLine);
2943 Result.clearFlag(Token::LeadingSpace);
2944 IsAtStartOfLine = false;
2946 goto LexNextToken; // GCC isn't tail call eliminating.
2948 return PP->Lex(Result);
2956 // Objective C support.
2957 if (CurPtr[-1] == '@' && Features.ObjC1)
2960 Kind = tok::unknown;
2967 Kind = tok::unknown;
2971 // Notify MIOpt that we read a non-whitespace/non-comment token.
2974 // Update the location of token as well as BufferPtr.
2975 FormTokenWithChars(Result, CurPtr, Kind);