clang  3.9.0
Lexer.cpp
Go to the documentation of this file.
1 //===--- Lexer.cpp - C Language Family Lexer ------------------------------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the Lexer and Token interfaces.
11 //
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Lex/Lexer.h"
15 #include "UnicodeCharSets.h"
16 #include "clang/Basic/CharInfo.h"
21 #include "clang/Lex/Preprocessor.h"
22 #include "llvm/ADT/STLExtras.h"
23 #include "llvm/ADT/StringExtras.h"
24 #include "llvm/ADT/StringSwitch.h"
25 #include "llvm/Support/Compiler.h"
26 #include "llvm/Support/ConvertUTF.h"
27 #include "llvm/Support/MemoryBuffer.h"
28 #include <cstring>
29 using namespace clang;
30 
31 //===----------------------------------------------------------------------===//
32 // Token Class Implementation
33 //===----------------------------------------------------------------------===//
34 
35 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
38  return II->getObjCKeywordID() == objcKey;
39  return false;
40 }
41 
42 /// getObjCKeywordID - Return the ObjC keyword kind.
45  return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
46 }
47 
48 
49 //===----------------------------------------------------------------------===//
50 // Lexer Class Implementation
51 //===----------------------------------------------------------------------===//
52 
53 void Lexer::anchor() { }
54 
55 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
56  const char *BufEnd) {
57  BufferStart = BufStart;
58  BufferPtr = BufPtr;
59  BufferEnd = BufEnd;
60 
61  assert(BufEnd[0] == 0 &&
62  "We assume that the input buffer has a null character at the end"
63  " to simplify lexing!");
64 
65  // Check whether we have a BOM in the beginning of the buffer. If yes - act
66  // accordingly. Right now we support only UTF-8 with and without BOM, so, just
67  // skip the UTF-8 BOM if it's present.
68  if (BufferStart == BufferPtr) {
69  // Determine the size of the BOM.
70  StringRef Buf(BufferStart, BufferEnd - BufferStart);
71  size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
72  .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
73  .Default(0);
74 
75  // Skip the BOM.
76  BufferPtr += BOMLength;
77  }
78 
79  Is_PragmaLexer = false;
80  CurrentConflictMarkerState = CMK_None;
81 
82  // Start of the file is a start of line.
83  IsAtStartOfLine = true;
84  IsAtPhysicalStartOfLine = true;
85 
86  HasLeadingSpace = false;
87  HasLeadingEmptyMacro = false;
88 
89  // We are not after parsing a #.
91 
92  // We are not after parsing #include.
93  ParsingFilename = false;
94 
95  // We are not in raw mode. Raw mode disables diagnostics and interpretation
96  // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
97  // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
98  // or otherwise skipping over tokens.
99  LexingRawMode = false;
100 
101  // Default to not keeping comments.
102  ExtendedTokenMode = 0;
103 }
104 
105 /// Lexer constructor - Create a new lexer object for the specified buffer
106 /// with the specified preprocessor managing the lexing process. This lexer
107 /// assumes that the associated file buffer and Preprocessor objects will
108 /// outlive it, so it doesn't take ownership of either of them.
109 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
110  : PreprocessorLexer(&PP, FID),
111  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
112  LangOpts(PP.getLangOpts()) {
113 
114  InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
115  InputFile->getBufferEnd());
116 
118 }
119 
121  assert(PP && "Cannot reset token mode without a preprocessor");
122  if (LangOpts.TraditionalCPP)
123  SetKeepWhitespaceMode(true);
124  else
126 }
127 
128 /// Lexer constructor - Create a new raw lexer object. This object is only
129 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
130 /// range will outlive it, so it doesn't take ownership of it.
131 Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
132  const char *BufStart, const char *BufPtr, const char *BufEnd)
133  : FileLoc(fileloc), LangOpts(langOpts) {
134 
135  InitLexer(BufStart, BufPtr, BufEnd);
136 
137  // We *are* in raw mode.
138  LexingRawMode = true;
139 }
140 
141 /// Lexer constructor - Create a new raw lexer object. This object is only
142 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
143 /// range will outlive it, so it doesn't take ownership of it.
144 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
145  const SourceManager &SM, const LangOptions &langOpts)
146  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile->getBufferStart(),
147  FromFile->getBufferStart(), FromFile->getBufferEnd()) {}
148 
149 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
150 /// _Pragma expansion. This has a variety of magic semantics that this method
151 /// sets up. It returns a new'd Lexer that must be delete'd when done.
152 ///
153 /// On entrance to this routine, TokStartLoc is a macro location which has a
154 /// spelling loc that indicates the bytes to be lexed for the token and an
155 /// expansion location that indicates where all lexed tokens should be
156 /// "expanded from".
157 ///
158 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
159 /// normal lexer that remaps tokens as they fly by. This would require making
160 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
161 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
162 /// out of the critical path of the lexer!
163 ///
165  SourceLocation ExpansionLocStart,
166  SourceLocation ExpansionLocEnd,
167  unsigned TokLen, Preprocessor &PP) {
169 
170  // Create the lexer as if we were going to lex the file normally.
171  FileID SpellingFID = SM.getFileID(SpellingLoc);
172  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
173  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
174 
175  // Now that the lexer is created, change the start/end locations so that we
176  // just lex the subsection of the file that we want. This is lexing from a
177  // scratch buffer.
178  const char *StrData = SM.getCharacterData(SpellingLoc);
179 
180  L->BufferPtr = StrData;
181  L->BufferEnd = StrData+TokLen;
182  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
183 
184  // Set the SourceLocation with the remapping information. This ensures that
185  // GetMappedTokenLoc will remap the tokens as they are lexed.
186  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
187  ExpansionLocStart,
188  ExpansionLocEnd, TokLen);
189 
190  // Ensure that the lexer thinks it is inside a directive, so that end \n will
191  // return an EOD token.
193 
194  // This lexer really is for _Pragma.
195  L->Is_PragmaLexer = true;
196  return L;
197 }
198 
199 
200 /// Stringify - Convert the specified string into a C string, with surrounding
201 /// ""'s, and with escaped \ and " characters.
202 std::string Lexer::Stringify(StringRef Str, bool Charify) {
203  std::string Result = Str;
204  char Quote = Charify ? '\'' : '"';
205  for (unsigned i = 0, e = Result.size(); i != e; ++i) {
206  if (Result[i] == '\\' || Result[i] == Quote) {
207  Result.insert(Result.begin()+i, '\\');
208  ++i; ++e;
209  }
210  }
211  return Result;
212 }
213 
214 /// Stringify - Convert the specified string into a C string by escaping '\'
215 /// and " characters. This does not add surrounding ""'s to the string.
217  for (unsigned i = 0, e = Str.size(); i != e; ++i) {
218  if (Str[i] == '\\' || Str[i] == '"') {
219  Str.insert(Str.begin()+i, '\\');
220  ++i; ++e;
221  }
222  }
223 }
224 
225 //===----------------------------------------------------------------------===//
226 // Token Spelling
227 //===----------------------------------------------------------------------===//
228 
229 /// \brief Slow case of getSpelling. Extract the characters comprising the
230 /// spelling of this token from the provided input buffer.
231 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
232  const LangOptions &LangOpts, char *Spelling) {
233  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
234 
235  size_t Length = 0;
236  const char *BufEnd = BufPtr + Tok.getLength();
237 
238  if (tok::isStringLiteral(Tok.getKind())) {
239  // Munch the encoding-prefix and opening double-quote.
240  while (BufPtr < BufEnd) {
241  unsigned Size;
242  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
243  BufPtr += Size;
244 
245  if (Spelling[Length - 1] == '"')
246  break;
247  }
248 
249  // Raw string literals need special handling; trigraph expansion and line
250  // splicing do not occur within their d-char-sequence nor within their
251  // r-char-sequence.
252  if (Length >= 2 &&
253  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
254  // Search backwards from the end of the token to find the matching closing
255  // quote.
256  const char *RawEnd = BufEnd;
257  do --RawEnd; while (*RawEnd != '"');
258  size_t RawLength = RawEnd - BufPtr + 1;
259 
260  // Everything between the quotes is included verbatim in the spelling.
261  memcpy(Spelling + Length, BufPtr, RawLength);
262  Length += RawLength;
263  BufPtr += RawLength;
264 
265  // The rest of the token is lexed normally.
266  }
267  }
268 
269  while (BufPtr < BufEnd) {
270  unsigned Size;
271  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
272  BufPtr += Size;
273  }
274 
275  assert(Length < Tok.getLength() &&
276  "NeedsCleaning flag set on token that didn't need cleaning!");
277  return Length;
278 }
279 
280 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
281 /// token are the characters used to represent the token in the source file
282 /// after trigraph expansion and escaped-newline folding. In particular, this
283 /// wants to get the true, uncanonicalized, spelling of things like digraphs
284 /// UCNs, etc.
286  SmallVectorImpl<char> &buffer,
287  const SourceManager &SM,
288  const LangOptions &options,
289  bool *invalid) {
290  // Break down the source location.
291  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
292 
293  // Try to the load the file buffer.
294  bool invalidTemp = false;
295  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
296  if (invalidTemp) {
297  if (invalid) *invalid = true;
298  return StringRef();
299  }
300 
301  const char *tokenBegin = file.data() + locInfo.second;
302 
303  // Lex from the start of the given location.
304  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
305  file.begin(), tokenBegin, file.end());
306  Token token;
307  lexer.LexFromRawLexer(token);
308 
309  unsigned length = token.getLength();
310 
311  // Common case: no need for cleaning.
312  if (!token.needsCleaning())
313  return StringRef(tokenBegin, length);
314 
315  // Hard case, we need to relex the characters into the string.
316  buffer.resize(length);
317  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
318  return StringRef(buffer.data(), buffer.size());
319 }
320 
321 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
322 /// token are the characters used to represent the token in the source file
323 /// after trigraph expansion and escaped-newline folding. In particular, this
324 /// wants to get the true, uncanonicalized, spelling of things like digraphs
325 /// UCNs, etc.
326 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
327  const LangOptions &LangOpts, bool *Invalid) {
328  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
329 
330  bool CharDataInvalid = false;
331  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
332  &CharDataInvalid);
333  if (Invalid)
334  *Invalid = CharDataInvalid;
335  if (CharDataInvalid)
336  return std::string();
337 
338  // If this token contains nothing interesting, return it directly.
339  if (!Tok.needsCleaning())
340  return std::string(TokStart, TokStart + Tok.getLength());
341 
342  std::string Result;
343  Result.resize(Tok.getLength());
344  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
345  return Result;
346 }
347 
348 /// getSpelling - This method is used to get the spelling of a token into a
349 /// preallocated buffer, instead of as an std::string. The caller is required
350 /// to allocate enough space for the token, which is guaranteed to be at least
351 /// Tok.getLength() bytes long. The actual length of the token is returned.
352 ///
353 /// Note that this method may do two possible things: it may either fill in
354 /// the buffer specified with characters, or it may *change the input pointer*
355 /// to point to a constant buffer with the data already in it (avoiding a
356 /// copy). The caller is not allowed to modify the returned buffer pointer
357 /// if an internal buffer is returned.
358 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
359  const SourceManager &SourceMgr,
360  const LangOptions &LangOpts, bool *Invalid) {
361  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
362 
363  const char *TokStart = nullptr;
364  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
365  if (Tok.is(tok::raw_identifier))
366  TokStart = Tok.getRawIdentifier().data();
367  else if (!Tok.hasUCN()) {
368  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
369  // Just return the string from the identifier table, which is very quick.
370  Buffer = II->getNameStart();
371  return II->getLength();
372  }
373  }
374 
375  // NOTE: this can be checked even after testing for an IdentifierInfo.
376  if (Tok.isLiteral())
377  TokStart = Tok.getLiteralData();
378 
379  if (!TokStart) {
380  // Compute the start of the token in the input lexer buffer.
381  bool CharDataInvalid = false;
382  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
383  if (Invalid)
384  *Invalid = CharDataInvalid;
385  if (CharDataInvalid) {
386  Buffer = "";
387  return 0;
388  }
389  }
390 
391  // If this token contains nothing interesting, return it directly.
392  if (!Tok.needsCleaning()) {
393  Buffer = TokStart;
394  return Tok.getLength();
395  }
396 
397  // Otherwise, hard case, relex the characters into the string.
398  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
399 }
400 
401 
402 /// MeasureTokenLength - Relex the token at the specified location and return
403 /// its length in bytes in the input file. If the token needs cleaning (e.g.
404 /// includes a trigraph or an escaped newline) then this count includes bytes
405 /// that are part of that.
407  const SourceManager &SM,
408  const LangOptions &LangOpts) {
409  Token TheTok;
410  if (getRawToken(Loc, TheTok, SM, LangOpts))
411  return 0;
412  return TheTok.getLength();
413 }
414 
415 /// \brief Relex the token at the specified location.
416 /// \returns true if there was a failure, false on success.
418  const SourceManager &SM,
419  const LangOptions &LangOpts,
420  bool IgnoreWhiteSpace) {
421  // TODO: this could be special cased for common tokens like identifiers, ')',
422  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
423  // all obviously single-char tokens. This could use
424  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
425  // something.
426 
427  // If this comes from a macro expansion, we really do want the macro name, not
428  // the token this macro expanded to.
429  Loc = SM.getExpansionLoc(Loc);
430  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
431  bool Invalid = false;
432  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
433  if (Invalid)
434  return true;
435 
436  const char *StrData = Buffer.data()+LocInfo.second;
437 
438  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
439  return true;
440 
441  // Create a lexer starting at the beginning of this token.
442  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
443  Buffer.begin(), StrData, Buffer.end());
444  TheLexer.SetCommentRetentionState(true);
445  TheLexer.LexFromRawLexer(Result);
446  return false;
447 }
448 
450  const SourceManager &SM,
451  const LangOptions &LangOpts) {
452  assert(Loc.isFileID());
453  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
454  if (LocInfo.first.isInvalid())
455  return Loc;
456 
457  bool Invalid = false;
458  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
459  if (Invalid)
460  return Loc;
461 
462  // Back up from the current location until we hit the beginning of a line
463  // (or the buffer). We'll relex from that point.
464  const char *BufStart = Buffer.data();
465  if (LocInfo.second >= Buffer.size())
466  return Loc;
467 
468  const char *StrData = BufStart+LocInfo.second;
469  if (StrData[0] == '\n' || StrData[0] == '\r')
470  return Loc;
471 
472  const char *LexStart = StrData;
473  while (LexStart != BufStart) {
474  if (LexStart[0] == '\n' || LexStart[0] == '\r') {
475  ++LexStart;
476  break;
477  }
478 
479  --LexStart;
480  }
481 
482  // Create a lexer starting at the beginning of this token.
483  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
484  Lexer TheLexer(LexerStartLoc, LangOpts, BufStart, LexStart, Buffer.end());
485  TheLexer.SetCommentRetentionState(true);
486 
487  // Lex tokens until we find the token that contains the source location.
488  Token TheTok;
489  do {
490  TheLexer.LexFromRawLexer(TheTok);
491 
492  if (TheLexer.getBufferLocation() > StrData) {
493  // Lexing this token has taken the lexer past the source location we're
494  // looking for. If the current token encompasses our source location,
495  // return the beginning of that token.
496  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
497  return TheTok.getLocation();
498 
499  // We ended up skipping over the source location entirely, which means
500  // that it points into whitespace. We're done here.
501  break;
502  }
503  } while (TheTok.getKind() != tok::eof);
504 
505  // We've passed our source location; just return the original source location.
506  return Loc;
507 }
508 
510  const SourceManager &SM,
511  const LangOptions &LangOpts) {
512  if (Loc.isFileID())
513  return getBeginningOfFileToken(Loc, SM, LangOpts);
514 
515  if (!SM.isMacroArgExpansion(Loc))
516  return Loc;
517 
518  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
519  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
520  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
521  std::pair<FileID, unsigned> BeginFileLocInfo
522  = SM.getDecomposedLoc(BeginFileLoc);
523  assert(FileLocInfo.first == BeginFileLocInfo.first &&
524  FileLocInfo.second >= BeginFileLocInfo.second);
525  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
526 }
527 
528 namespace {
530  PDK_Skipped,
531  PDK_StartIf,
532  PDK_EndIf,
533  PDK_Unknown
534  };
535 }
536 
537 std::pair<unsigned, bool> Lexer::ComputePreamble(StringRef Buffer,
538  const LangOptions &LangOpts,
539  unsigned MaxLines) {
540  // Create a lexer starting at the beginning of the file. Note that we use a
541  // "fake" file source location at offset 1 so that the lexer will track our
542  // position within the file.
543  const unsigned StartOffset = 1;
544  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
545  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
546  Buffer.end());
547  TheLexer.SetCommentRetentionState(true);
548 
549  // StartLoc will differ from FileLoc if there is a BOM that was skipped.
550  SourceLocation StartLoc = TheLexer.getSourceLocation();
551 
552  bool InPreprocessorDirective = false;
553  Token TheTok;
554  Token IfStartTok;
555  unsigned IfCount = 0;
556  SourceLocation ActiveCommentLoc;
557 
558  unsigned MaxLineOffset = 0;
559  if (MaxLines) {
560  const char *CurPtr = Buffer.begin();
561  unsigned CurLine = 0;
562  while (CurPtr != Buffer.end()) {
563  char ch = *CurPtr++;
564  if (ch == '\n') {
565  ++CurLine;
566  if (CurLine == MaxLines)
567  break;
568  }
569  }
570  if (CurPtr != Buffer.end())
571  MaxLineOffset = CurPtr - Buffer.begin();
572  }
573 
574  do {
575  TheLexer.LexFromRawLexer(TheTok);
576 
577  if (InPreprocessorDirective) {
578  // If we've hit the end of the file, we're done.
579  if (TheTok.getKind() == tok::eof) {
580  break;
581  }
582 
583  // If we haven't hit the end of the preprocessor directive, skip this
584  // token.
585  if (!TheTok.isAtStartOfLine())
586  continue;
587 
588  // We've passed the end of the preprocessor directive, and will look
589  // at this token again below.
590  InPreprocessorDirective = false;
591  }
592 
593  // Keep track of the # of lines in the preamble.
594  if (TheTok.isAtStartOfLine()) {
595  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
596 
597  // If we were asked to limit the number of lines in the preamble,
598  // and we're about to exceed that limit, we're done.
599  if (MaxLineOffset && TokOffset >= MaxLineOffset)
600  break;
601  }
602 
603  // Comments are okay; skip over them.
604  if (TheTok.getKind() == tok::comment) {
605  if (ActiveCommentLoc.isInvalid())
606  ActiveCommentLoc = TheTok.getLocation();
607  continue;
608  }
609 
610  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
611  // This is the start of a preprocessor directive.
612  Token HashTok = TheTok;
613  InPreprocessorDirective = true;
614  ActiveCommentLoc = SourceLocation();
615 
616  // Figure out which directive this is. Since we're lexing raw tokens,
617  // we don't have an identifier table available. Instead, just look at
618  // the raw identifier to recognize and categorize preprocessor directives.
619  TheLexer.LexFromRawLexer(TheTok);
620  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
621  StringRef Keyword = TheTok.getRawIdentifier();
623  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
624  .Case("include", PDK_Skipped)
625  .Case("__include_macros", PDK_Skipped)
626  .Case("define", PDK_Skipped)
627  .Case("undef", PDK_Skipped)
628  .Case("line", PDK_Skipped)
629  .Case("error", PDK_Skipped)
630  .Case("pragma", PDK_Skipped)
631  .Case("import", PDK_Skipped)
632  .Case("include_next", PDK_Skipped)
633  .Case("warning", PDK_Skipped)
634  .Case("ident", PDK_Skipped)
635  .Case("sccs", PDK_Skipped)
636  .Case("assert", PDK_Skipped)
637  .Case("unassert", PDK_Skipped)
638  .Case("if", PDK_StartIf)
639  .Case("ifdef", PDK_StartIf)
640  .Case("ifndef", PDK_StartIf)
641  .Case("elif", PDK_Skipped)
642  .Case("else", PDK_Skipped)
643  .Case("endif", PDK_EndIf)
644  .Default(PDK_Unknown);
645 
646  switch (PDK) {
647  case PDK_Skipped:
648  continue;
649 
650  case PDK_StartIf:
651  if (IfCount == 0)
652  IfStartTok = HashTok;
653 
654  ++IfCount;
655  continue;
656 
657  case PDK_EndIf:
658  // Mismatched #endif. The preamble ends here.
659  if (IfCount == 0)
660  break;
661 
662  --IfCount;
663  continue;
664 
665  case PDK_Unknown:
666  // We don't know what this directive is; stop at the '#'.
667  break;
668  }
669  }
670 
671  // We only end up here if we didn't recognize the preprocessor
672  // directive or it was one that can't occur in the preamble at this
673  // point. Roll back the current token to the location of the '#'.
674  InPreprocessorDirective = false;
675  TheTok = HashTok;
676  }
677 
678  // We hit a token that we don't recognize as being in the
679  // "preprocessing only" part of the file, so we're no longer in
680  // the preamble.
681  break;
682  } while (true);
683 
685  if (IfCount)
686  End = IfStartTok.getLocation();
687  else if (ActiveCommentLoc.isValid())
688  End = ActiveCommentLoc; // don't truncate a decl comment.
689  else
690  End = TheTok.getLocation();
691 
692  return std::make_pair(End.getRawEncoding() - StartLoc.getRawEncoding(),
693  IfCount? IfStartTok.isAtStartOfLine()
694  : TheTok.isAtStartOfLine());
695 }
696 
697 
698 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
699 /// token, return a new location that specifies a character within the token.
701  unsigned CharNo,
702  const SourceManager &SM,
703  const LangOptions &LangOpts) {
704  // Figure out how many physical characters away the specified expansion
705  // character is. This needs to take into consideration newlines and
706  // trigraphs.
707  bool Invalid = false;
708  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
709 
710  // If they request the first char of the token, we're trivially done.
711  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
712  return TokStart;
713 
714  unsigned PhysOffset = 0;
715 
716  // The usual case is that tokens don't contain anything interesting. Skip
717  // over the uninteresting characters. If a token only consists of simple
718  // chars, this method is extremely fast.
719  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
720  if (CharNo == 0)
721  return TokStart.getLocWithOffset(PhysOffset);
722  ++TokPtr;
723  --CharNo;
724  ++PhysOffset;
725  }
726 
727  // If we have a character that may be a trigraph or escaped newline, use a
728  // lexer to parse it correctly.
729  for (; CharNo; --CharNo) {
730  unsigned Size;
731  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
732  TokPtr += Size;
733  PhysOffset += Size;
734  }
735 
736  // Final detail: if we end up on an escaped newline, we want to return the
737  // location of the actual byte of the token. For example foo<newline>bar
738  // advanced by 3 should return the location of b, not of \\. One compounding
739  // detail of this is that the escape may be made by a trigraph.
740  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
741  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
742 
743  return TokStart.getLocWithOffset(PhysOffset);
744 }
745 
746 /// \brief Computes the source location just past the end of the
747 /// token at this source location.
748 ///
749 /// This routine can be used to produce a source location that
750 /// points just past the end of the token referenced by \p Loc, and
751 /// is generally used when a diagnostic needs to point just after a
752 /// token where it expected something different that it received. If
753 /// the returned source location would not be meaningful (e.g., if
754 /// it points into a macro), this routine returns an invalid
755 /// source location.
756 ///
757 /// \param Offset an offset from the end of the token, where the source
758 /// location should refer to. The default offset (0) produces a source
759 /// location pointing just past the end of the token; an offset of 1 produces
760 /// a source location pointing to the last character in the token, etc.
762  const SourceManager &SM,
763  const LangOptions &LangOpts) {
764  if (Loc.isInvalid())
765  return SourceLocation();
766 
767  if (Loc.isMacroID()) {
768  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
769  return SourceLocation(); // Points inside the macro expansion.
770  }
771 
772  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
773  if (Len > Offset)
774  Len = Len - Offset;
775  else
776  return Loc;
777 
778  return Loc.getLocWithOffset(Len);
779 }
780 
781 /// \brief Returns true if the given MacroID location points at the first
782 /// token of the macro expansion.
784  const SourceManager &SM,
785  const LangOptions &LangOpts,
786  SourceLocation *MacroBegin) {
787  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
788 
789  SourceLocation expansionLoc;
790  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
791  return false;
792 
793  if (expansionLoc.isFileID()) {
794  // No other macro expansions, this is the first.
795  if (MacroBegin)
796  *MacroBegin = expansionLoc;
797  return true;
798  }
799 
800  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
801 }
802 
803 /// \brief Returns true if the given MacroID location points at the last
804 /// token of the macro expansion.
806  const SourceManager &SM,
807  const LangOptions &LangOpts,
808  SourceLocation *MacroEnd) {
809  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
810 
811  SourceLocation spellLoc = SM.getSpellingLoc(loc);
812  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
813  if (tokLen == 0)
814  return false;
815 
816  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
817  SourceLocation expansionLoc;
818  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
819  return false;
820 
821  if (expansionLoc.isFileID()) {
822  // No other macro expansions.
823  if (MacroEnd)
824  *MacroEnd = expansionLoc;
825  return true;
826  }
827 
828  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
829 }
830 
832  const SourceManager &SM,
833  const LangOptions &LangOpts) {
834  SourceLocation Begin = Range.getBegin();
835  SourceLocation End = Range.getEnd();
836  assert(Begin.isFileID() && End.isFileID());
837  if (Range.isTokenRange()) {
838  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
839  if (End.isInvalid())
840  return CharSourceRange();
841  }
842 
843  // Break down the source locations.
844  FileID FID;
845  unsigned BeginOffs;
846  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
847  if (FID.isInvalid())
848  return CharSourceRange();
849 
850  unsigned EndOffs;
851  if (!SM.isInFileID(End, FID, &EndOffs) ||
852  BeginOffs > EndOffs)
853  return CharSourceRange();
854 
855  return CharSourceRange::getCharRange(Begin, End);
856 }
857 
859  const SourceManager &SM,
860  const LangOptions &LangOpts) {
861  SourceLocation Begin = Range.getBegin();
862  SourceLocation End = Range.getEnd();
863  if (Begin.isInvalid() || End.isInvalid())
864  return CharSourceRange();
865 
866  if (Begin.isFileID() && End.isFileID())
867  return makeRangeFromFileLocs(Range, SM, LangOpts);
868 
869  if (Begin.isMacroID() && End.isFileID()) {
870  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
871  return CharSourceRange();
872  Range.setBegin(Begin);
873  return makeRangeFromFileLocs(Range, SM, LangOpts);
874  }
875 
876  if (Begin.isFileID() && End.isMacroID()) {
877  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
878  &End)) ||
879  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
880  &End)))
881  return CharSourceRange();
882  Range.setEnd(End);
883  return makeRangeFromFileLocs(Range, SM, LangOpts);
884  }
885 
886  assert(Begin.isMacroID() && End.isMacroID());
887  SourceLocation MacroBegin, MacroEnd;
888  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
889  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
890  &MacroEnd)) ||
891  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
892  &MacroEnd)))) {
893  Range.setBegin(MacroBegin);
894  Range.setEnd(MacroEnd);
895  return makeRangeFromFileLocs(Range, SM, LangOpts);
896  }
897 
898  bool Invalid = false;
899  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
900  &Invalid);
901  if (Invalid)
902  return CharSourceRange();
903 
904  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
905  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
906  &Invalid);
907  if (Invalid)
908  return CharSourceRange();
909 
910  if (EndEntry.getExpansion().isMacroArgExpansion() &&
911  BeginEntry.getExpansion().getExpansionLocStart() ==
912  EndEntry.getExpansion().getExpansionLocStart()) {
913  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
914  Range.setEnd(SM.getImmediateSpellingLoc(End));
915  return makeFileCharRange(Range, SM, LangOpts);
916  }
917  }
918 
919  return CharSourceRange();
920 }
921 
923  const SourceManager &SM,
924  const LangOptions &LangOpts,
925  bool *Invalid) {
926  Range = makeFileCharRange(Range, SM, LangOpts);
927  if (Range.isInvalid()) {
928  if (Invalid) *Invalid = true;
929  return StringRef();
930  }
931 
932  // Break down the source location.
933  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
934  if (beginInfo.first.isInvalid()) {
935  if (Invalid) *Invalid = true;
936  return StringRef();
937  }
938 
939  unsigned EndOffs;
940  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
941  beginInfo.second > EndOffs) {
942  if (Invalid) *Invalid = true;
943  return StringRef();
944  }
945 
946  // Try to the load the file buffer.
947  bool invalidTemp = false;
948  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
949  if (invalidTemp) {
950  if (Invalid) *Invalid = true;
951  return StringRef();
952  }
953 
954  if (Invalid) *Invalid = false;
955  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
956 }
957 
959  const SourceManager &SM,
960  const LangOptions &LangOpts) {
961  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
962 
963  // Find the location of the immediate macro expansion.
964  while (1) {
965  FileID FID = SM.getFileID(Loc);
966  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
967  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
968  Loc = Expansion.getExpansionLocStart();
969  if (!Expansion.isMacroArgExpansion())
970  break;
971 
972  // For macro arguments we need to check that the argument did not come
973  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
974 
975  // Loc points to the argument id of the macro definition, move to the
976  // macro expansion.
977  Loc = SM.getImmediateExpansionRange(Loc).first;
978  SourceLocation SpellLoc = Expansion.getSpellingLoc();
979  if (SpellLoc.isFileID())
980  break; // No inner macro.
981 
982  // If spelling location resides in the same FileID as macro expansion
983  // location, it means there is no inner macro.
984  FileID MacroFID = SM.getFileID(Loc);
985  if (SM.isInFileID(SpellLoc, MacroFID))
986  break;
987 
988  // Argument came from inner macro.
989  Loc = SpellLoc;
990  }
991 
992  // Find the spelling location of the start of the non-argument expansion
993  // range. This is where the macro name was spelled in order to begin
994  // expanding this macro.
995  Loc = SM.getSpellingLoc(Loc);
996 
997  // Dig out the buffer where the macro name was spelled and the extents of the
998  // name so that we can render it into the expansion note.
999  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1000  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1001  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1002  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1003 }
1004 
1006  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1007  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
1008  // Walk past macro argument expanions.
1009  while (SM.isMacroArgExpansion(Loc))
1010  Loc = SM.getImmediateExpansionRange(Loc).first;
1011 
1012  // If the macro's spelling has no FileID, then it's actually a token paste
1013  // or stringization (or similar) and not a macro at all.
1014  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1015  return StringRef();
1016 
1017  // Find the spelling location of the start of the non-argument expansion
1018  // range. This is where the macro name was spelled in order to begin
1019  // expanding this macro.
1020  Loc = SM.getSpellingLoc(SM.getImmediateExpansionRange(Loc).first);
1021 
1022  // Dig out the buffer where the macro name was spelled and the extents of the
1023  // name so that we can render it into the expansion note.
1024  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1025  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1026  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1027  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1028 }
1029 
1030 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1031  return isIdentifierBody(c, LangOpts.DollarIdents);
1032 }
1033 
1034 
1035 //===----------------------------------------------------------------------===//
1036 // Diagnostics forwarding code.
1037 //===----------------------------------------------------------------------===//
1038 
1039 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1040 /// lexer buffer was all expanded at a single point, perform the mapping.
1041 /// This is currently only used for _Pragma implementation, so it is the slow
1042 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1043 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1044  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1046  SourceLocation FileLoc,
1047  unsigned CharNo, unsigned TokLen) {
1048  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1049 
1050  // Otherwise, we're lexing "mapped tokens". This is used for things like
1051  // _Pragma handling. Combine the expansion location of FileLoc with the
1052  // spelling location.
1054 
1055  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1056  // characters come from spelling(FileLoc)+Offset.
1057  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1058  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1059 
1060  // Figure out the expansion loc range, which is the range covered by the
1061  // original _Pragma(...) sequence.
1062  std::pair<SourceLocation,SourceLocation> II =
1063  SM.getImmediateExpansionRange(FileLoc);
1064 
1065  return SM.createExpansionLoc(SpellingLoc, II.first, II.second, TokLen);
1066 }
1067 
1068 /// getSourceLocation - Return a source location identifier for the specified
1069 /// offset in the current file.
1071  unsigned TokLen) const {
1072  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1073  "Location out of range for this buffer!");
1074 
1075  // In the normal case, we're just lexing from a simple file buffer, return
1076  // the file id from FileLoc with the offset specified.
1077  unsigned CharNo = Loc-BufferStart;
1078  if (FileLoc.isFileID())
1079  return FileLoc.getLocWithOffset(CharNo);
1080 
1081  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1082  // tokens are lexed from where the _Pragma was defined.
1083  assert(PP && "This doesn't work on raw lexers");
1084  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1085 }
1086 
1087 /// Diag - Forwarding function for diagnostics. This translate a source
1088 /// position in the current buffer into a SourceLocation object for rendering.
1089 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1090  return PP->Diag(getSourceLocation(Loc), DiagID);
1091 }
1092 
1093 //===----------------------------------------------------------------------===//
1094 // Trigraph and Escaped Newline Handling Code.
1095 //===----------------------------------------------------------------------===//
1096 
1097 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1098 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1099 static char GetTrigraphCharForLetter(char Letter) {
1100  switch (Letter) {
1101  default: return 0;
1102  case '=': return '#';
1103  case ')': return ']';
1104  case '(': return '[';
1105  case '!': return '|';
1106  case '\'': return '^';
1107  case '>': return '}';
1108  case '/': return '\\';
1109  case '<': return '{';
1110  case '-': return '~';
1111  }
1112 }
1113 
1114 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1115 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1116 /// return the result character. Finally, emit a warning about trigraph use
1117 /// whether trigraphs are enabled or not.
1118 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1119  char Res = GetTrigraphCharForLetter(*CP);
1120  if (!Res || !L) return Res;
1121 
1122  if (!L->getLangOpts().Trigraphs) {
1123  if (!L->isLexingRawMode())
1124  L->Diag(CP-2, diag::trigraph_ignored);
1125  return 0;
1126  }
1127 
1128  if (!L->isLexingRawMode())
1129  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1130  return Res;
1131 }
1132 
1133 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1134 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1135 /// trigraph equivalent on entry to this function.
1136 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1137  unsigned Size = 0;
1138  while (isWhitespace(Ptr[Size])) {
1139  ++Size;
1140 
1141  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1142  continue;
1143 
1144  // If this is a \r\n or \n\r, skip the other half.
1145  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1146  Ptr[Size-1] != Ptr[Size])
1147  ++Size;
1148 
1149  return Size;
1150  }
1151 
1152  // Not an escaped newline, must be a \t or something else.
1153  return 0;
1154 }
1155 
1156 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1157 /// them), skip over them and return the first non-escaped-newline found,
1158 /// otherwise return P.
1159 const char *Lexer::SkipEscapedNewLines(const char *P) {
1160  while (1) {
1161  const char *AfterEscape;
1162  if (*P == '\\') {
1163  AfterEscape = P+1;
1164  } else if (*P == '?') {
1165  // If not a trigraph for escape, bail out.
1166  if (P[1] != '?' || P[2] != '/')
1167  return P;
1168  AfterEscape = P+3;
1169  } else {
1170  return P;
1171  }
1172 
1173  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1174  if (NewLineSize == 0) return P;
1175  P = AfterEscape+NewLineSize;
1176  }
1177 }
1178 
1179 /// \brief Checks that the given token is the first token that occurs after the
1180 /// given location (this excludes comments and whitespace). Returns the location
1181 /// immediately after the specified token. If the token is not found or the
1182 /// location is inside a macro, the returned source location will be invalid.
1184  tok::TokenKind TKind,
1185  const SourceManager &SM,
1186  const LangOptions &LangOpts,
1187  bool SkipTrailingWhitespaceAndNewLine) {
1188  if (Loc.isMacroID()) {
1189  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1190  return SourceLocation();
1191  }
1192  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1193 
1194  // Break down the source location.
1195  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1196 
1197  // Try to load the file buffer.
1198  bool InvalidTemp = false;
1199  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1200  if (InvalidTemp)
1201  return SourceLocation();
1202 
1203  const char *TokenBegin = File.data() + LocInfo.second;
1204 
1205  // Lex from the start of the given location.
1206  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1207  TokenBegin, File.end());
1208  // Find the token.
1209  Token Tok;
1210  lexer.LexFromRawLexer(Tok);
1211  if (Tok.isNot(TKind))
1212  return SourceLocation();
1213  SourceLocation TokenLoc = Tok.getLocation();
1214 
1215  // Calculate how much whitespace needs to be skipped if any.
1216  unsigned NumWhitespaceChars = 0;
1217  if (SkipTrailingWhitespaceAndNewLine) {
1218  const char *TokenEnd = SM.getCharacterData(TokenLoc) +
1219  Tok.getLength();
1220  unsigned char C = *TokenEnd;
1221  while (isHorizontalWhitespace(C)) {
1222  C = *(++TokenEnd);
1223  NumWhitespaceChars++;
1224  }
1225 
1226  // Skip \r, \n, \r\n, or \n\r
1227  if (C == '\n' || C == '\r') {
1228  char PrevC = C;
1229  C = *(++TokenEnd);
1230  NumWhitespaceChars++;
1231  if ((C == '\n' || C == '\r') && C != PrevC)
1232  NumWhitespaceChars++;
1233  }
1234  }
1235 
1236  return TokenLoc.getLocWithOffset(Tok.getLength() + NumWhitespaceChars);
1237 }
1238 
1239 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1240 /// get its size, and return it. This is tricky in several cases:
1241 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1242 /// then either return the trigraph (skipping 3 chars) or the '?',
1243 /// depending on whether trigraphs are enabled or not.
1244 /// 2. If this is an escaped newline (potentially with whitespace between
1245 /// the backslash and newline), implicitly skip the newline and return
1246 /// the char after it.
1247 ///
1248 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1249 /// know that we can accumulate into Size, and that we have already incremented
1250 /// Ptr by Size bytes.
1251 ///
1252 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1253 /// be updated to match.
1254 ///
1255 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1256  Token *Tok) {
1257  // If we have a slash, look for an escaped newline.
1258  if (Ptr[0] == '\\') {
1259  ++Size;
1260  ++Ptr;
1261 Slash:
1262  // Common case, backslash-char where the char is not whitespace.
1263  if (!isWhitespace(Ptr[0])) return '\\';
1264 
1265  // See if we have optional whitespace characters between the slash and
1266  // newline.
1267  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1268  // Remember that this token needs to be cleaned.
1269  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1270 
1271  // Warn if there was whitespace between the backslash and newline.
1272  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1273  Diag(Ptr, diag::backslash_newline_space);
1274 
1275  // Found backslash<whitespace><newline>. Parse the char after it.
1276  Size += EscapedNewLineSize;
1277  Ptr += EscapedNewLineSize;
1278 
1279  // If the char that we finally got was a \n, then we must have had
1280  // something like <newline><newline>. We don't want to consume the
1281  // second newline.
1282  if (*Ptr == '\n' || *Ptr == '\r' || *Ptr == '\0')
1283  return ' ';
1284 
1285  // Use slow version to accumulate a correct size field.
1286  return getCharAndSizeSlow(Ptr, Size, Tok);
1287  }
1288 
1289  // Otherwise, this is not an escaped newline, just return the slash.
1290  return '\\';
1291  }
1292 
1293  // If this is a trigraph, process it.
1294  if (Ptr[0] == '?' && Ptr[1] == '?') {
1295  // If this is actually a legal trigraph (not something like "??x"), emit
1296  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1297  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1298  // Remember that this token needs to be cleaned.
1299  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1300 
1301  Ptr += 3;
1302  Size += 3;
1303  if (C == '\\') goto Slash;
1304  return C;
1305  }
1306  }
1307 
1308  // If this is neither, return a single character.
1309  ++Size;
1310  return *Ptr;
1311 }
1312 
1313 
1314 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1315 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1316 /// and that we have already incremented Ptr by Size bytes.
1317 ///
1318 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1319 /// be updated to match.
1320 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1321  const LangOptions &LangOpts) {
1322  // If we have a slash, look for an escaped newline.
1323  if (Ptr[0] == '\\') {
1324  ++Size;
1325  ++Ptr;
1326 Slash:
1327  // Common case, backslash-char where the char is not whitespace.
1328  if (!isWhitespace(Ptr[0])) return '\\';
1329 
1330  // See if we have optional whitespace characters followed by a newline.
1331  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1332  // Found backslash<whitespace><newline>. Parse the char after it.
1333  Size += EscapedNewLineSize;
1334  Ptr += EscapedNewLineSize;
1335 
1336  // If the char that we finally got was a \n, then we must have had
1337  // something like <newline><newline>. We don't want to consume the
1338  // second newline.
1339  if (*Ptr == '\n' || *Ptr == '\r' || *Ptr == '\0')
1340  return ' ';
1341 
1342  // Use slow version to accumulate a correct size field.
1343  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1344  }
1345 
1346  // Otherwise, this is not an escaped newline, just return the slash.
1347  return '\\';
1348  }
1349 
1350  // If this is a trigraph, process it.
1351  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1352  // If this is actually a legal trigraph (not something like "??x"), return
1353  // it.
1354  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1355  Ptr += 3;
1356  Size += 3;
1357  if (C == '\\') goto Slash;
1358  return C;
1359  }
1360  }
1361 
1362  // If this is neither, return a single character.
1363  ++Size;
1364  return *Ptr;
1365 }
1366 
1367 //===----------------------------------------------------------------------===//
1368 // Helper methods for lexing.
1369 //===----------------------------------------------------------------------===//
1370 
1371 /// \brief Routine that indiscriminately skips bytes in the source file.
1372 void Lexer::SkipBytes(unsigned Bytes, bool StartOfLine) {
1373  BufferPtr += Bytes;
1374  if (BufferPtr > BufferEnd)
1375  BufferPtr = BufferEnd;
1376  // FIXME: What exactly does the StartOfLine bit mean? There are two
1377  // possible meanings for the "start" of the line: the first token on the
1378  // unexpanded line, or the first token on the expanded line.
1379  IsAtStartOfLine = StartOfLine;
1380  IsAtPhysicalStartOfLine = StartOfLine;
1381 }
1382 
1383 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1384  if (LangOpts.AsmPreprocessor) {
1385  return false;
1386  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1387  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1389  return C11AllowedIDChars.contains(C);
1390  } else if (LangOpts.CPlusPlus) {
1391  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1393  return CXX03AllowedIDChars.contains(C);
1394  } else {
1395  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1397  return C99AllowedIDChars.contains(C);
1398  }
1399 }
1400 
1401 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1402  assert(isAllowedIDChar(C, LangOpts));
1403  if (LangOpts.AsmPreprocessor) {
1404  return false;
1405  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1406  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1408  return !C11DisallowedInitialIDChars.contains(C);
1409  } else if (LangOpts.CPlusPlus) {
1410  return true;
1411  } else {
1412  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1414  return !C99DisallowedInitialIDChars.contains(C);
1415  }
1416 }
1417 
1418 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1419  const char *End) {
1421  L.getSourceLocation(End));
1422 }
1423 
1424 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1425  CharSourceRange Range, bool IsFirst) {
1426  // Check C99 compatibility.
1427  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1428  enum {
1429  CannotAppearInIdentifier = 0,
1430  CannotStartIdentifier
1431  };
1432 
1433  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1435  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1437  if (!C99AllowedIDChars.contains(C)) {
1438  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1439  << Range
1440  << CannotAppearInIdentifier;
1441  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1442  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1443  << Range
1444  << CannotStartIdentifier;
1445  }
1446  }
1447 
1448  // Check C++98 compatibility.
1449  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1450  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1452  if (!CXX03AllowedIDChars.contains(C)) {
1453  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1454  << Range;
1455  }
1456  }
1457 }
1458 
1459 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1460  Token &Result) {
1461  const char *UCNPtr = CurPtr + Size;
1462  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1463  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1464  return false;
1465 
1466  if (!isLexingRawMode())
1468  makeCharRange(*this, CurPtr, UCNPtr),
1469  /*IsFirst=*/false);
1470 
1471  Result.setFlag(Token::HasUCN);
1472  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1473  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1474  CurPtr = UCNPtr;
1475  else
1476  while (CurPtr != UCNPtr)
1477  (void)getAndAdvanceChar(CurPtr, Result);
1478  return true;
1479 }
1480 
1481 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1482  const char *UnicodePtr = CurPtr;
1483  UTF32 CodePoint;
1484  ConversionResult Result =
1485  llvm::convertUTF8Sequence((const UTF8 **)&UnicodePtr,
1486  (const UTF8 *)BufferEnd,
1487  &CodePoint,
1488  strictConversion);
1489  if (Result != conversionOK ||
1490  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1491  return false;
1492 
1493  if (!isLexingRawMode())
1495  makeCharRange(*this, CurPtr, UnicodePtr),
1496  /*IsFirst=*/false);
1497 
1498  CurPtr = UnicodePtr;
1499  return true;
1500 }
1501 
1502 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1503  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1504  unsigned Size;
1505  unsigned char C = *CurPtr++;
1506  while (isIdentifierBody(C))
1507  C = *CurPtr++;
1508 
1509  --CurPtr; // Back up over the skipped character.
1510 
1511  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1512  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1513  //
1514  // TODO: Could merge these checks into an InfoTable flag to make the
1515  // comparison cheaper
1516  if (isASCII(C) && C != '\\' && C != '?' &&
1517  (C != '$' || !LangOpts.DollarIdents)) {
1518 FinishIdentifier:
1519  const char *IdStart = BufferPtr;
1520  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1521  Result.setRawIdentifierData(IdStart);
1522 
1523  // If we are in raw mode, return this identifier raw. There is no need to
1524  // look up identifier information or attempt to macro expand it.
1525  if (LexingRawMode)
1526  return true;
1527 
1528  // Fill in Result.IdentifierInfo and update the token kind,
1529  // looking up the identifier in the identifier table.
1530  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1531 
1532  // Finally, now that we know we have an identifier, pass this off to the
1533  // preprocessor, which may macro expand it or something.
1534  if (II->isHandleIdentifierCase())
1535  return PP->HandleIdentifier(Result);
1536 
1537  return true;
1538  }
1539 
1540  // Otherwise, $,\,? in identifier found. Enter slower path.
1541 
1542  C = getCharAndSize(CurPtr, Size);
1543  while (1) {
1544  if (C == '$') {
1545  // If we hit a $ and they are not supported in identifiers, we are done.
1546  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1547 
1548  // Otherwise, emit a diagnostic and continue.
1549  if (!isLexingRawMode())
1550  Diag(CurPtr, diag::ext_dollar_in_identifier);
1551  CurPtr = ConsumeChar(CurPtr, Size, Result);
1552  C = getCharAndSize(CurPtr, Size);
1553  continue;
1554 
1555  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1556  C = getCharAndSize(CurPtr, Size);
1557  continue;
1558  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1559  C = getCharAndSize(CurPtr, Size);
1560  continue;
1561  } else if (!isIdentifierBody(C)) {
1562  goto FinishIdentifier;
1563  }
1564 
1565  // Otherwise, this character is good, consume it.
1566  CurPtr = ConsumeChar(CurPtr, Size, Result);
1567 
1568  C = getCharAndSize(CurPtr, Size);
1569  while (isIdentifierBody(C)) {
1570  CurPtr = ConsumeChar(CurPtr, Size, Result);
1571  C = getCharAndSize(CurPtr, Size);
1572  }
1573  }
1574 }
1575 
1576 /// isHexaLiteral - Return true if Start points to a hex constant.
1577 /// in microsoft mode (where this is supposed to be several different tokens).
1578 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1579  unsigned Size;
1580  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1581  if (C1 != '0')
1582  return false;
1583  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1584  return (C2 == 'x' || C2 == 'X');
1585 }
1586 
1587 /// LexNumericConstant - Lex the remainder of a integer or floating point
1588 /// constant. From[-1] is the first character lexed. Return the end of the
1589 /// constant.
1590 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1591  unsigned Size;
1592  char C = getCharAndSize(CurPtr, Size);
1593  char PrevCh = 0;
1594  while (isPreprocessingNumberBody(C)) {
1595  CurPtr = ConsumeChar(CurPtr, Size, Result);
1596  PrevCh = C;
1597  C = getCharAndSize(CurPtr, Size);
1598  }
1599 
1600  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1601  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1602  // If we are in Microsoft mode, don't continue if the constant is hex.
1603  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1604  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1605  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1606  }
1607 
1608  // If we have a hex FP constant, continue.
1609  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1610  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1611  // not-quite-conforming extension. Only do so if this looks like it's
1612  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1613  bool IsHexFloat = true;
1614  if (!LangOpts.C99) {
1615  if (!isHexaLiteral(BufferPtr, LangOpts))
1616  IsHexFloat = false;
1617  else if (!getLangOpts().CPlusPlus1z &&
1618  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1619  IsHexFloat = false;
1620  }
1621  if (IsHexFloat)
1622  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1623  }
1624 
1625  // If we have a digit separator, continue.
1626  if (C == '\'' && getLangOpts().CPlusPlus14) {
1627  unsigned NextSize;
1628  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1629  if (isIdentifierBody(Next)) {
1630  if (!isLexingRawMode())
1631  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1632  CurPtr = ConsumeChar(CurPtr, Size, Result);
1633  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1634  return LexNumericConstant(Result, CurPtr);
1635  }
1636  }
1637 
1638  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1639  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1640  return LexNumericConstant(Result, CurPtr);
1641  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1642  return LexNumericConstant(Result, CurPtr);
1643 
1644  // Update the location of token as well as BufferPtr.
1645  const char *TokStart = BufferPtr;
1646  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1647  Result.setLiteralData(TokStart);
1648  return true;
1649 }
1650 
1651 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1652 /// in C++11, or warn on a ud-suffix in C++98.
1653 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1654  bool IsStringLiteral) {
1655  assert(getLangOpts().CPlusPlus);
1656 
1657  // Maximally munch an identifier.
1658  unsigned Size;
1659  char C = getCharAndSize(CurPtr, Size);
1660  bool Consumed = false;
1661 
1662  if (!isIdentifierHead(C)) {
1663  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1664  Consumed = true;
1665  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1666  Consumed = true;
1667  else
1668  return CurPtr;
1669  }
1670 
1671  if (!getLangOpts().CPlusPlus11) {
1672  if (!isLexingRawMode())
1673  Diag(CurPtr,
1674  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1675  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1677  return CurPtr;
1678  }
1679 
1680  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1681  // that does not start with an underscore is ill-formed. As a conforming
1682  // extension, we treat all such suffixes as if they had whitespace before
1683  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1684  // likely to be a ud-suffix than a macro, however, and accept that.
1685  if (!Consumed) {
1686  bool IsUDSuffix = false;
1687  if (C == '_')
1688  IsUDSuffix = true;
1689  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1690  // In C++1y, we need to look ahead a few characters to see if this is a
1691  // valid suffix for a string literal or a numeric literal (this could be
1692  // the 'operator""if' defining a numeric literal operator).
1693  const unsigned MaxStandardSuffixLength = 3;
1694  char Buffer[MaxStandardSuffixLength] = { C };
1695  unsigned Consumed = Size;
1696  unsigned Chars = 1;
1697  while (true) {
1698  unsigned NextSize;
1699  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1700  getLangOpts());
1701  if (!isIdentifierBody(Next)) {
1702  // End of suffix. Check whether this is on the whitelist.
1703  IsUDSuffix = (Chars == 1 && Buffer[0] == 's') ||
1705  getLangOpts(), StringRef(Buffer, Chars));
1706  break;
1707  }
1708 
1709  if (Chars == MaxStandardSuffixLength)
1710  // Too long: can't be a standard suffix.
1711  break;
1712 
1713  Buffer[Chars++] = Next;
1714  Consumed += NextSize;
1715  }
1716  }
1717 
1718  if (!IsUDSuffix) {
1719  if (!isLexingRawMode())
1720  Diag(CurPtr, getLangOpts().MSVCCompat
1721  ? diag::ext_ms_reserved_user_defined_literal
1722  : diag::ext_reserved_user_defined_literal)
1724  return CurPtr;
1725  }
1726 
1727  CurPtr = ConsumeChar(CurPtr, Size, Result);
1728  }
1729 
1730  Result.setFlag(Token::HasUDSuffix);
1731  while (true) {
1732  C = getCharAndSize(CurPtr, Size);
1733  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1734  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1735  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1736  else break;
1737  }
1738 
1739  return CurPtr;
1740 }
1741 
1742 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1743 /// either " or L" or u8" or u" or U".
1744 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1745  tok::TokenKind Kind) {
1746  // Does this string contain the \0 character?
1747  const char *NulCharacter = nullptr;
1748 
1749  if (!isLexingRawMode() &&
1750  (Kind == tok::utf8_string_literal ||
1751  Kind == tok::utf16_string_literal ||
1752  Kind == tok::utf32_string_literal))
1753  Diag(BufferPtr, getLangOpts().CPlusPlus
1754  ? diag::warn_cxx98_compat_unicode_literal
1755  : diag::warn_c99_compat_unicode_literal);
1756 
1757  char C = getAndAdvanceChar(CurPtr, Result);
1758  while (C != '"') {
1759  // Skip escaped characters. Escaped newlines will already be processed by
1760  // getAndAdvanceChar.
1761  if (C == '\\')
1762  C = getAndAdvanceChar(CurPtr, Result);
1763 
1764  if (C == '\n' || C == '\r' || // Newline.
1765  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1766  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1767  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1768  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1769  return true;
1770  }
1771 
1772  if (C == 0) {
1773  if (isCodeCompletionPoint(CurPtr-1)) {
1775  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1776  cutOffLexing();
1777  return true;
1778  }
1779 
1780  NulCharacter = CurPtr-1;
1781  }
1782  C = getAndAdvanceChar(CurPtr, Result);
1783  }
1784 
1785  // If we are in C++11, lex the optional ud-suffix.
1786  if (getLangOpts().CPlusPlus)
1787  CurPtr = LexUDSuffix(Result, CurPtr, true);
1788 
1789  // If a nul character existed in the string, warn about it.
1790  if (NulCharacter && !isLexingRawMode())
1791  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1792 
1793  // Update the location of the token as well as the BufferPtr instance var.
1794  const char *TokStart = BufferPtr;
1795  FormTokenWithChars(Result, CurPtr, Kind);
1796  Result.setLiteralData(TokStart);
1797  return true;
1798 }
1799 
1800 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1801 /// having lexed R", LR", u8R", uR", or UR".
1802 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1803  tok::TokenKind Kind) {
1804  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1805  // Between the initial and final double quote characters of the raw string,
1806  // any transformations performed in phases 1 and 2 (trigraphs,
1807  // universal-character-names, and line splicing) are reverted.
1808 
1809  if (!isLexingRawMode())
1810  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1811 
1812  unsigned PrefixLen = 0;
1813 
1814  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1815  ++PrefixLen;
1816 
1817  // If the last character was not a '(', then we didn't lex a valid delimiter.
1818  if (CurPtr[PrefixLen] != '(') {
1819  if (!isLexingRawMode()) {
1820  const char *PrefixEnd = &CurPtr[PrefixLen];
1821  if (PrefixLen == 16) {
1822  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1823  } else {
1824  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1825  << StringRef(PrefixEnd, 1);
1826  }
1827  }
1828 
1829  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1830  // it's possible the '"' was intended to be part of the raw string, but
1831  // there's not much we can do about that.
1832  while (1) {
1833  char C = *CurPtr++;
1834 
1835  if (C == '"')
1836  break;
1837  if (C == 0 && CurPtr-1 == BufferEnd) {
1838  --CurPtr;
1839  break;
1840  }
1841  }
1842 
1843  FormTokenWithChars(Result, CurPtr, tok::unknown);
1844  return true;
1845  }
1846 
1847  // Save prefix and move CurPtr past it
1848  const char *Prefix = CurPtr;
1849  CurPtr += PrefixLen + 1; // skip over prefix and '('
1850 
1851  while (1) {
1852  char C = *CurPtr++;
1853 
1854  if (C == ')') {
1855  // Check for prefix match and closing quote.
1856  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1857  CurPtr += PrefixLen + 1; // skip over prefix and '"'
1858  break;
1859  }
1860  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1861  if (!isLexingRawMode())
1862  Diag(BufferPtr, diag::err_unterminated_raw_string)
1863  << StringRef(Prefix, PrefixLen);
1864  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1865  return true;
1866  }
1867  }
1868 
1869  // If we are in C++11, lex the optional ud-suffix.
1870  if (getLangOpts().CPlusPlus)
1871  CurPtr = LexUDSuffix(Result, CurPtr, true);
1872 
1873  // Update the location of token as well as BufferPtr.
1874  const char *TokStart = BufferPtr;
1875  FormTokenWithChars(Result, CurPtr, Kind);
1876  Result.setLiteralData(TokStart);
1877  return true;
1878 }
1879 
1880 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
1881 /// after having lexed the '<' character. This is used for #include filenames.
1882 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
1883  // Does this string contain the \0 character?
1884  const char *NulCharacter = nullptr;
1885  const char *AfterLessPos = CurPtr;
1886  char C = getAndAdvanceChar(CurPtr, Result);
1887  while (C != '>') {
1888  // Skip escaped characters.
1889  if (C == '\\' && CurPtr < BufferEnd) {
1890  // Skip the escaped character.
1891  getAndAdvanceChar(CurPtr, Result);
1892  } else if (C == '\n' || C == '\r' || // Newline.
1893  (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
1894  isCodeCompletionPoint(CurPtr-1)))) {
1895  // If the filename is unterminated, then it must just be a lone <
1896  // character. Return this as such.
1897  FormTokenWithChars(Result, AfterLessPos, tok::less);
1898  return true;
1899  } else if (C == 0) {
1900  NulCharacter = CurPtr-1;
1901  }
1902  C = getAndAdvanceChar(CurPtr, Result);
1903  }
1904 
1905  // If a nul character existed in the string, warn about it.
1906  if (NulCharacter && !isLexingRawMode())
1907  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1908 
1909  // Update the location of token as well as BufferPtr.
1910  const char *TokStart = BufferPtr;
1911  FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
1912  Result.setLiteralData(TokStart);
1913  return true;
1914 }
1915 
1916 
1917 /// LexCharConstant - Lex the remainder of a character constant, after having
1918 /// lexed either ' or L' or u8' or u' or U'.
1919 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
1920  tok::TokenKind Kind) {
1921  // Does this character contain the \0 character?
1922  const char *NulCharacter = nullptr;
1923 
1924  if (!isLexingRawMode()) {
1925  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
1926  Diag(BufferPtr, getLangOpts().CPlusPlus
1927  ? diag::warn_cxx98_compat_unicode_literal
1928  : diag::warn_c99_compat_unicode_literal);
1929  else if (Kind == tok::utf8_char_constant)
1930  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
1931  }
1932 
1933  char C = getAndAdvanceChar(CurPtr, Result);
1934  if (C == '\'') {
1935  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1936  Diag(BufferPtr, diag::ext_empty_character);
1937  FormTokenWithChars(Result, CurPtr, tok::unknown);
1938  return true;
1939  }
1940 
1941  while (C != '\'') {
1942  // Skip escaped characters.
1943  if (C == '\\')
1944  C = getAndAdvanceChar(CurPtr, Result);
1945 
1946  if (C == '\n' || C == '\r' || // Newline.
1947  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1948  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1949  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
1950  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1951  return true;
1952  }
1953 
1954  if (C == 0) {
1955  if (isCodeCompletionPoint(CurPtr-1)) {
1957  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1958  cutOffLexing();
1959  return true;
1960  }
1961 
1962  NulCharacter = CurPtr-1;
1963  }
1964  C = getAndAdvanceChar(CurPtr, Result);
1965  }
1966 
1967  // If we are in C++11, lex the optional ud-suffix.
1968  if (getLangOpts().CPlusPlus)
1969  CurPtr = LexUDSuffix(Result, CurPtr, false);
1970 
1971  // If a nul character existed in the character, warn about it.
1972  if (NulCharacter && !isLexingRawMode())
1973  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
1974 
1975  // Update the location of token as well as BufferPtr.
1976  const char *TokStart = BufferPtr;
1977  FormTokenWithChars(Result, CurPtr, Kind);
1978  Result.setLiteralData(TokStart);
1979  return true;
1980 }
1981 
1982 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
1983 /// Update BufferPtr to point to the next non-whitespace character and return.
1984 ///
1985 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
1986 ///
1987 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
1988  bool &TokAtPhysicalStartOfLine) {
1989  // Whitespace - Skip it, then return the token after the whitespace.
1990  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
1991 
1992  unsigned char Char = *CurPtr;
1993 
1994  // Skip consecutive spaces efficiently.
1995  while (1) {
1996  // Skip horizontal whitespace very aggressively.
1997  while (isHorizontalWhitespace(Char))
1998  Char = *++CurPtr;
1999 
2000  // Otherwise if we have something other than whitespace, we're done.
2001  if (!isVerticalWhitespace(Char))
2002  break;
2003 
2005  // End of preprocessor directive line, let LexTokenInternal handle this.
2006  BufferPtr = CurPtr;
2007  return false;
2008  }
2009 
2010  // OK, but handle newline.
2011  SawNewline = true;
2012  Char = *++CurPtr;
2013  }
2014 
2015  // If the client wants us to return whitespace, return it now.
2016  if (isKeepWhitespaceMode()) {
2017  FormTokenWithChars(Result, CurPtr, tok::unknown);
2018  if (SawNewline) {
2019  IsAtStartOfLine = true;
2020  IsAtPhysicalStartOfLine = true;
2021  }
2022  // FIXME: The next token will not have LeadingSpace set.
2023  return true;
2024  }
2025 
2026  // If this isn't immediately after a newline, there is leading space.
2027  char PrevChar = CurPtr[-1];
2028  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2029 
2030  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2031  if (SawNewline) {
2032  Result.setFlag(Token::StartOfLine);
2033  TokAtPhysicalStartOfLine = true;
2034  }
2035 
2036  BufferPtr = CurPtr;
2037  return false;
2038 }
2039 
2040 /// We have just read the // characters from input. Skip until we find the
2041 /// newline character thats terminate the comment. Then update BufferPtr and
2042 /// return.
2043 ///
2044 /// If we're in KeepCommentMode or any CommentHandler has inserted
2045 /// some tokens, this will store the first token and return true.
2046 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2047  bool &TokAtPhysicalStartOfLine) {
2048  // If Line comments aren't explicitly enabled for this language, emit an
2049  // extension warning.
2050  if (!LangOpts.LineComment && !isLexingRawMode()) {
2051  Diag(BufferPtr, diag::ext_line_comment);
2052 
2053  // Mark them enabled so we only emit one warning for this translation
2054  // unit.
2055  LangOpts.LineComment = true;
2056  }
2057 
2058  // Scan over the body of the comment. The common case, when scanning, is that
2059  // the comment contains normal ascii characters with nothing interesting in
2060  // them. As such, optimize for this case with the inner loop.
2061  char C;
2062  do {
2063  C = *CurPtr;
2064  // Skip over characters in the fast loop.
2065  while (C != 0 && // Potentially EOF.
2066  C != '\n' && C != '\r') // Newline or DOS-style newline.
2067  C = *++CurPtr;
2068 
2069  const char *NextLine = CurPtr;
2070  if (C != 0) {
2071  // We found a newline, see if it's escaped.
2072  const char *EscapePtr = CurPtr-1;
2073  bool HasSpace = false;
2074  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2075  --EscapePtr;
2076  HasSpace = true;
2077  }
2078 
2079  if (*EscapePtr == '\\') // Escaped newline.
2080  CurPtr = EscapePtr;
2081  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2082  EscapePtr[-2] == '?') // Trigraph-escaped newline.
2083  CurPtr = EscapePtr-2;
2084  else
2085  break; // This is a newline, we're done.
2086 
2087  // If there was space between the backslash and newline, warn about it.
2088  if (HasSpace && !isLexingRawMode())
2089  Diag(EscapePtr, diag::backslash_newline_space);
2090  }
2091 
2092  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2093  // properly decode the character. Read it in raw mode to avoid emitting
2094  // diagnostics about things like trigraphs. If we see an escaped newline,
2095  // we'll handle it below.
2096  const char *OldPtr = CurPtr;
2097  bool OldRawMode = isLexingRawMode();
2098  LexingRawMode = true;
2099  C = getAndAdvanceChar(CurPtr, Result);
2100  LexingRawMode = OldRawMode;
2101 
2102  // If we only read only one character, then no special handling is needed.
2103  // We're done and can skip forward to the newline.
2104  if (C != 0 && CurPtr == OldPtr+1) {
2105  CurPtr = NextLine;
2106  break;
2107  }
2108 
2109  // If we read multiple characters, and one of those characters was a \r or
2110  // \n, then we had an escaped newline within the comment. Emit diagnostic
2111  // unless the next line is also a // comment.
2112  if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') {
2113  for (; OldPtr != CurPtr; ++OldPtr)
2114  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2115  // Okay, we found a // comment that ends in a newline, if the next
2116  // line is also a // comment, but has spaces, don't emit a diagnostic.
2117  if (isWhitespace(C)) {
2118  const char *ForwardPtr = CurPtr;
2119  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2120  ++ForwardPtr;
2121  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2122  break;
2123  }
2124 
2125  if (!isLexingRawMode())
2126  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2127  break;
2128  }
2129  }
2130 
2131  if (CurPtr == BufferEnd+1) {
2132  --CurPtr;
2133  break;
2134  }
2135 
2136  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2138  cutOffLexing();
2139  return false;
2140  }
2141 
2142  } while (C != '\n' && C != '\r');
2143 
2144  // Found but did not consume the newline. Notify comment handlers about the
2145  // comment unless we're in a #if 0 block.
2146  if (PP && !isLexingRawMode() &&
2147  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2148  getSourceLocation(CurPtr)))) {
2149  BufferPtr = CurPtr;
2150  return true; // A token has to be returned.
2151  }
2152 
2153  // If we are returning comments as tokens, return this comment as a token.
2154  if (inKeepCommentMode())
2155  return SaveLineComment(Result, CurPtr);
2156 
2157  // If we are inside a preprocessor directive and we see the end of line,
2158  // return immediately, so that the lexer can return this as an EOD token.
2159  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2160  BufferPtr = CurPtr;
2161  return false;
2162  }
2163 
2164  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2165  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2166  // contribute to another token), it isn't needed for correctness. Note that
2167  // this is ok even in KeepWhitespaceMode, because we would have returned the
2168  /// comment above in that mode.
2169  ++CurPtr;
2170 
2171  // The next returned token is at the start of the line.
2172  Result.setFlag(Token::StartOfLine);
2173  TokAtPhysicalStartOfLine = true;
2174  // No leading whitespace seen so far.
2176  BufferPtr = CurPtr;
2177  return false;
2178 }
2179 
2180 /// If in save-comment mode, package up this Line comment in an appropriate
2181 /// way and return it.
2182 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2183  // If we're not in a preprocessor directive, just return the // comment
2184  // directly.
2185  FormTokenWithChars(Result, CurPtr, tok::comment);
2186 
2188  return true;
2189 
2190  // If this Line-style comment is in a macro definition, transmogrify it into
2191  // a C-style block comment.
2192  bool Invalid = false;
2193  std::string Spelling = PP->getSpelling(Result, &Invalid);
2194  if (Invalid)
2195  return true;
2196 
2197  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2198  Spelling[1] = '*'; // Change prefix to "/*".
2199  Spelling += "*/"; // add suffix.
2200 
2201  Result.setKind(tok::comment);
2202  PP->CreateString(Spelling, Result,
2203  Result.getLocation(), Result.getLocation());
2204  return true;
2205 }
2206 
2207 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2208 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2209 /// a diagnostic if so. We know that the newline is inside of a block comment.
2210 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2211  Lexer *L) {
2212  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2213 
2214  // Back up off the newline.
2215  --CurPtr;
2216 
2217  // If this is a two-character newline sequence, skip the other character.
2218  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2219  // \n\n or \r\r -> not escaped newline.
2220  if (CurPtr[0] == CurPtr[1])
2221  return false;
2222  // \n\r or \r\n -> skip the newline.
2223  --CurPtr;
2224  }
2225 
2226  // If we have horizontal whitespace, skip over it. We allow whitespace
2227  // between the slash and newline.
2228  bool HasSpace = false;
2229  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2230  --CurPtr;
2231  HasSpace = true;
2232  }
2233 
2234  // If we have a slash, we know this is an escaped newline.
2235  if (*CurPtr == '\\') {
2236  if (CurPtr[-1] != '*') return false;
2237  } else {
2238  // It isn't a slash, is it the ?? / trigraph?
2239  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2240  CurPtr[-3] != '*')
2241  return false;
2242 
2243  // This is the trigraph ending the comment. Emit a stern warning!
2244  CurPtr -= 2;
2245 
2246  // If no trigraphs are enabled, warn that we ignored this trigraph and
2247  // ignore this * character.
2248  if (!L->getLangOpts().Trigraphs) {
2249  if (!L->isLexingRawMode())
2250  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2251  return false;
2252  }
2253  if (!L->isLexingRawMode())
2254  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2255  }
2256 
2257  // Warn about having an escaped newline between the */ characters.
2258  if (!L->isLexingRawMode())
2259  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2260 
2261  // If there was space between the backslash and newline, warn about it.
2262  if (HasSpace && !L->isLexingRawMode())
2263  L->Diag(CurPtr, diag::backslash_newline_space);
2264 
2265  return true;
2266 }
2267 
2268 #ifdef __SSE2__
2269 #include <emmintrin.h>
2270 #elif __ALTIVEC__
2271 #include <altivec.h>
2272 #undef bool
2273 #endif
2274 
2275 /// We have just read from input the / and * characters that started a comment.
2276 /// Read until we find the * and / characters that terminate the comment.
2277 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2278 /// comments, because they cannot cause the comment to end. The only thing
2279 /// that can happen is the comment could end with an escaped newline between
2280 /// the terminating * and /.
2281 ///
2282 /// If we're in KeepCommentMode or any CommentHandler has inserted
2283 /// some tokens, this will store the first token and return true.
2284 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2285  bool &TokAtPhysicalStartOfLine) {
2286  // Scan one character past where we should, looking for a '/' character. Once
2287  // we find it, check to see if it was preceded by a *. This common
2288  // optimization helps people who like to put a lot of * characters in their
2289  // comments.
2290 
2291  // The first character we get with newlines and trigraphs skipped to handle
2292  // the degenerate /*/ case below correctly if the * has an escaped newline
2293  // after it.
2294  unsigned CharSize;
2295  unsigned char C = getCharAndSize(CurPtr, CharSize);
2296  CurPtr += CharSize;
2297  if (C == 0 && CurPtr == BufferEnd+1) {
2298  if (!isLexingRawMode())
2299  Diag(BufferPtr, diag::err_unterminated_block_comment);
2300  --CurPtr;
2301 
2302  // KeepWhitespaceMode should return this broken comment as a token. Since
2303  // it isn't a well formed comment, just return it as an 'unknown' token.
2304  if (isKeepWhitespaceMode()) {
2305  FormTokenWithChars(Result, CurPtr, tok::unknown);
2306  return true;
2307  }
2308 
2309  BufferPtr = CurPtr;
2310  return false;
2311  }
2312 
2313  // Check to see if the first character after the '/*' is another /. If so,
2314  // then this slash does not end the block comment, it is part of it.
2315  if (C == '/')
2316  C = *CurPtr++;
2317 
2318  while (1) {
2319  // Skip over all non-interesting characters until we find end of buffer or a
2320  // (probably ending) '/' character.
2321  if (CurPtr + 24 < BufferEnd &&
2322  // If there is a code-completion point avoid the fast scan because it
2323  // doesn't check for '\0'.
2324  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2325  // While not aligned to a 16-byte boundary.
2326  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2327  C = *CurPtr++;
2328 
2329  if (C == '/') goto FoundSlash;
2330 
2331 #ifdef __SSE2__
2332  __m128i Slashes = _mm_set1_epi8('/');
2333  while (CurPtr+16 <= BufferEnd) {
2334  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2335  Slashes));
2336  if (cmp != 0) {
2337  // Adjust the pointer to point directly after the first slash. It's
2338  // not necessary to set C here, it will be overwritten at the end of
2339  // the outer loop.
2340  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2341  goto FoundSlash;
2342  }
2343  CurPtr += 16;
2344  }
2345 #elif __ALTIVEC__
2346  __vector unsigned char Slashes = {
2347  '/', '/', '/', '/', '/', '/', '/', '/',
2348  '/', '/', '/', '/', '/', '/', '/', '/'
2349  };
2350  while (CurPtr+16 <= BufferEnd &&
2351  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2352  CurPtr += 16;
2353 #else
2354  // Scan for '/' quickly. Many block comments are very large.
2355  while (CurPtr[0] != '/' &&
2356  CurPtr[1] != '/' &&
2357  CurPtr[2] != '/' &&
2358  CurPtr[3] != '/' &&
2359  CurPtr+4 < BufferEnd) {
2360  CurPtr += 4;
2361  }
2362 #endif
2363 
2364  // It has to be one of the bytes scanned, increment to it and read one.
2365  C = *CurPtr++;
2366  }
2367 
2368  // Loop to scan the remainder.
2369  while (C != '/' && C != '\0')
2370  C = *CurPtr++;
2371 
2372  if (C == '/') {
2373  FoundSlash:
2374  if (CurPtr[-2] == '*') // We found the final */. We're done!
2375  break;
2376 
2377  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2378  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2379  // We found the final */, though it had an escaped newline between the
2380  // * and /. We're done!
2381  break;
2382  }
2383  }
2384  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2385  // If this is a /* inside of the comment, emit a warning. Don't do this
2386  // if this is a /*/, which will end the comment. This misses cases with
2387  // embedded escaped newlines, but oh well.
2388  if (!isLexingRawMode())
2389  Diag(CurPtr-1, diag::warn_nested_block_comment);
2390  }
2391  } else if (C == 0 && CurPtr == BufferEnd+1) {
2392  if (!isLexingRawMode())
2393  Diag(BufferPtr, diag::err_unterminated_block_comment);
2394  // Note: the user probably forgot a */. We could continue immediately
2395  // after the /*, but this would involve lexing a lot of what really is the
2396  // comment, which surely would confuse the parser.
2397  --CurPtr;
2398 
2399  // KeepWhitespaceMode should return this broken comment as a token. Since
2400  // it isn't a well formed comment, just return it as an 'unknown' token.
2401  if (isKeepWhitespaceMode()) {
2402  FormTokenWithChars(Result, CurPtr, tok::unknown);
2403  return true;
2404  }
2405 
2406  BufferPtr = CurPtr;
2407  return false;
2408  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2410  cutOffLexing();
2411  return false;
2412  }
2413 
2414  C = *CurPtr++;
2415  }
2416 
2417  // Notify comment handlers about the comment unless we're in a #if 0 block.
2418  if (PP && !isLexingRawMode() &&
2419  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2420  getSourceLocation(CurPtr)))) {
2421  BufferPtr = CurPtr;
2422  return true; // A token has to be returned.
2423  }
2424 
2425  // If we are returning comments as tokens, return this comment as a token.
2426  if (inKeepCommentMode()) {
2427  FormTokenWithChars(Result, CurPtr, tok::comment);
2428  return true;
2429  }
2430 
2431  // It is common for the tokens immediately after a /**/ comment to be
2432  // whitespace. Instead of going through the big switch, handle it
2433  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2434  // have already returned above with the comment as a token.
2435  if (isHorizontalWhitespace(*CurPtr)) {
2436  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2437  return false;
2438  }
2439 
2440  // Otherwise, just return so that the next character will be lexed as a token.
2441  BufferPtr = CurPtr;
2442  Result.setFlag(Token::LeadingSpace);
2443  return false;
2444 }
2445 
2446 //===----------------------------------------------------------------------===//
2447 // Primary Lexing Entry Points
2448 //===----------------------------------------------------------------------===//
2449 
2450 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2451 /// uninterpreted string. This switches the lexer out of directive mode.
2453  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2454  "Must be in a preprocessing directive!");
2455  Token Tmp;
2456 
2457  // CurPtr - Cache BufferPtr in an automatic variable.
2458  const char *CurPtr = BufferPtr;
2459  while (1) {
2460  char Char = getAndAdvanceChar(CurPtr, Tmp);
2461  switch (Char) {
2462  default:
2463  if (Result)
2464  Result->push_back(Char);
2465  break;
2466  case 0: // Null.
2467  // Found end of file?
2468  if (CurPtr-1 != BufferEnd) {
2469  if (isCodeCompletionPoint(CurPtr-1)) {
2471  cutOffLexing();
2472  return;
2473  }
2474 
2475  // Nope, normal character, continue.
2476  if (Result)
2477  Result->push_back(Char);
2478  break;
2479  }
2480  // FALL THROUGH.
2481  case '\r':
2482  case '\n':
2483  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2484  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2485  BufferPtr = CurPtr-1;
2486 
2487  // Next, lex the character, which should handle the EOD transition.
2488  Lex(Tmp);
2489  if (Tmp.is(tok::code_completion)) {
2490  if (PP)
2492  Lex(Tmp);
2493  }
2494  assert(Tmp.is(tok::eod) && "Unexpected token!");
2495 
2496  // Finally, we're done;
2497  return;
2498  }
2499  }
2500 }
2501 
2502 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2503 /// condition, reporting diagnostics and handling other edge cases as required.
2504 /// This returns true if Result contains a token, false if PP.Lex should be
2505 /// called again.
2506 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2507  // If we hit the end of the file while parsing a preprocessor directive,
2508  // end the preprocessor directive first. The next token returned will
2509  // then be the end of file.
2511  // Done parsing the "line".
2513  // Update the location of token as well as BufferPtr.
2514  FormTokenWithChars(Result, CurPtr, tok::eod);
2515 
2516  // Restore comment saving mode, in case it was disabled for directive.
2517  if (PP)
2519  return true; // Have a token.
2520  }
2521 
2522  // If we are in raw mode, return this event as an EOF token. Let the caller
2523  // that put us in raw mode handle the event.
2524  if (isLexingRawMode()) {
2525  Result.startToken();
2526  BufferPtr = BufferEnd;
2527  FormTokenWithChars(Result, BufferEnd, tok::eof);
2528  return true;
2529  }
2530 
2531  // Issue diagnostics for unterminated #if and missing newline.
2532 
2533  // If we are in a #if directive, emit an error.
2534  while (!ConditionalStack.empty()) {
2535  if (PP->getCodeCompletionFileLoc() != FileLoc)
2536  PP->Diag(ConditionalStack.back().IfLoc,
2537  diag::err_pp_unterminated_conditional);
2538  ConditionalStack.pop_back();
2539  }
2540 
2541  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2542  // a pedwarn.
2543  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2544  DiagnosticsEngine &Diags = PP->getDiagnostics();
2545  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2546  unsigned DiagID;
2547 
2548  if (LangOpts.CPlusPlus11) {
2549  // C++11 [lex.phases] 2.2 p2
2550  // Prefer the C++98 pedantic compatibility warning over the generic,
2551  // non-extension, user-requested "missing newline at EOF" warning.
2552  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2553  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2554  } else {
2555  DiagID = diag::warn_no_newline_eof;
2556  }
2557  } else {
2558  DiagID = diag::ext_no_newline_eof;
2559  }
2560 
2561  Diag(BufferEnd, DiagID)
2562  << FixItHint::CreateInsertion(EndLoc, "\n");
2563  }
2564 
2565  BufferPtr = CurPtr;
2566 
2567  // Finally, let the preprocessor handle this.
2568  return PP->HandleEndOfFile(Result, isPragmaLexer());
2569 }
2570 
2571 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2572 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2573 /// else and 2 if there are no more tokens in the buffer controlled by the
2574 /// lexer.
2575 unsigned Lexer::isNextPPTokenLParen() {
2576  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2577 
2578  // Switch to 'skipping' mode. This will ensure that we can lex a token
2579  // without emitting diagnostics, disables macro expansion, and will cause EOF
2580  // to return an EOF token instead of popping the include stack.
2581  LexingRawMode = true;
2582 
2583  // Save state that can be changed while lexing so that we can restore it.
2584  const char *TmpBufferPtr = BufferPtr;
2585  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2586  bool atStartOfLine = IsAtStartOfLine;
2587  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2588  bool leadingSpace = HasLeadingSpace;
2589 
2590  Token Tok;
2591  Lex(Tok);
2592 
2593  // Restore state that may have changed.
2594  BufferPtr = TmpBufferPtr;
2595  ParsingPreprocessorDirective = inPPDirectiveMode;
2596  HasLeadingSpace = leadingSpace;
2597  IsAtStartOfLine = atStartOfLine;
2598  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2599 
2600  // Restore the lexer back to non-skipping mode.
2601  LexingRawMode = false;
2602 
2603  if (Tok.is(tok::eof))
2604  return 2;
2605  return Tok.is(tok::l_paren);
2606 }
2607 
2608 /// \brief Find the end of a version control conflict marker.
2609 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2610  ConflictMarkerKind CMK) {
2611  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2612  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2613  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2614  size_t Pos = RestOfBuffer.find(Terminator);
2615  while (Pos != StringRef::npos) {
2616  // Must occur at start of line.
2617  if (Pos == 0 ||
2618  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2619  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2620  Pos = RestOfBuffer.find(Terminator);
2621  continue;
2622  }
2623  return RestOfBuffer.data()+Pos;
2624  }
2625  return nullptr;
2626 }
2627 
2628 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2629 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2630 /// and recover nicely. This returns true if it is a conflict marker and false
2631 /// if not.
2632 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2633  // Only a conflict marker if it starts at the beginning of a line.
2634  if (CurPtr != BufferStart &&
2635  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2636  return false;
2637 
2638  // Check to see if we have <<<<<<< or >>>>.
2639  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2640  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2641  return false;
2642 
2643  // If we have a situation where we don't care about conflict markers, ignore
2644  // it.
2645  if (CurrentConflictMarkerState || isLexingRawMode())
2646  return false;
2647 
2648  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2649 
2650  // Check to see if there is an ending marker somewhere in the buffer at the
2651  // start of a line to terminate this conflict marker.
2652  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2653  // We found a match. We are really in a conflict marker.
2654  // Diagnose this, and ignore to the end of line.
2655  Diag(CurPtr, diag::err_conflict_marker);
2656  CurrentConflictMarkerState = Kind;
2657 
2658  // Skip ahead to the end of line. We know this exists because the
2659  // end-of-conflict marker starts with \r or \n.
2660  while (*CurPtr != '\r' && *CurPtr != '\n') {
2661  assert(CurPtr != BufferEnd && "Didn't find end of line");
2662  ++CurPtr;
2663  }
2664  BufferPtr = CurPtr;
2665  return true;
2666  }
2667 
2668  // No end of conflict marker found.
2669  return false;
2670 }
2671 
2672 
2673 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2674 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2675 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2676 /// the line. This returns true if it is a conflict marker and false if not.
2677 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2678  // Only a conflict marker if it starts at the beginning of a line.
2679  if (CurPtr != BufferStart &&
2680  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2681  return false;
2682 
2683  // If we have a situation where we don't care about conflict markers, ignore
2684  // it.
2685  if (!CurrentConflictMarkerState || isLexingRawMode())
2686  return false;
2687 
2688  // Check to see if we have the marker (4 characters in a row).
2689  for (unsigned i = 1; i != 4; ++i)
2690  if (CurPtr[i] != CurPtr[0])
2691  return false;
2692 
2693  // If we do have it, search for the end of the conflict marker. This could
2694  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2695  // be the end of conflict marker.
2696  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2697  CurrentConflictMarkerState)) {
2698  CurPtr = End;
2699 
2700  // Skip ahead to the end of line.
2701  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2702  ++CurPtr;
2703 
2704  BufferPtr = CurPtr;
2705 
2706  // No longer in the conflict marker.
2707  CurrentConflictMarkerState = CMK_None;
2708  return true;
2709  }
2710 
2711  return false;
2712 }
2713 
2714 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2715  if (PP && PP->isCodeCompletionEnabled()) {
2716  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2717  return Loc == PP->getCodeCompletionLoc();
2718  }
2719 
2720  return false;
2721 }
2722 
2723 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2724  Token *Result) {
2725  unsigned CharSize;
2726  char Kind = getCharAndSize(StartPtr, CharSize);
2727 
2728  unsigned NumHexDigits;
2729  if (Kind == 'u')
2730  NumHexDigits = 4;
2731  else if (Kind == 'U')
2732  NumHexDigits = 8;
2733  else
2734  return 0;
2735 
2736  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2737  if (Result && !isLexingRawMode())
2738  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2739  return 0;
2740  }
2741 
2742  const char *CurPtr = StartPtr + CharSize;
2743  const char *KindLoc = &CurPtr[-1];
2744 
2745  uint32_t CodePoint = 0;
2746  for (unsigned i = 0; i < NumHexDigits; ++i) {
2747  char C = getCharAndSize(CurPtr, CharSize);
2748 
2749  unsigned Value = llvm::hexDigitValue(C);
2750  if (Value == -1U) {
2751  if (Result && !isLexingRawMode()) {
2752  if (i == 0) {
2753  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2754  << StringRef(KindLoc, 1);
2755  } else {
2756  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2757 
2758  // If the user wrote \U1234, suggest a fixit to \u.
2759  if (i == 4 && NumHexDigits == 8) {
2760  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2761  Diag(KindLoc, diag::note_ucn_four_not_eight)
2762  << FixItHint::CreateReplacement(URange, "u");
2763  }
2764  }
2765  }
2766 
2767  return 0;
2768  }
2769 
2770  CodePoint <<= 4;
2771  CodePoint += Value;
2772 
2773  CurPtr += CharSize;
2774  }
2775 
2776  if (Result) {
2777  Result->setFlag(Token::HasUCN);
2778  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
2779  StartPtr = CurPtr;
2780  else
2781  while (StartPtr != CurPtr)
2782  (void)getAndAdvanceChar(StartPtr, *Result);
2783  } else {
2784  StartPtr = CurPtr;
2785  }
2786 
2787  // Don't apply C family restrictions to UCNs in assembly mode
2788  if (LangOpts.AsmPreprocessor)
2789  return CodePoint;
2790 
2791  // C99 6.4.3p2: A universal character name shall not specify a character whose
2792  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
2793  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
2794  // C++11 [lex.charset]p2: If the hexadecimal value for a
2795  // universal-character-name corresponds to a surrogate code point (in the
2796  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
2797  // if the hexadecimal value for a universal-character-name outside the
2798  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
2799  // string literal corresponds to a control character (in either of the
2800  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
2801  // basic source character set, the program is ill-formed.
2802  if (CodePoint < 0xA0) {
2803  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
2804  return CodePoint;
2805 
2806  // We don't use isLexingRawMode() here because we need to warn about bad
2807  // UCNs even when skipping preprocessing tokens in a #if block.
2808  if (Result && PP) {
2809  if (CodePoint < 0x20 || CodePoint >= 0x7F)
2810  Diag(BufferPtr, diag::err_ucn_control_character);
2811  else {
2812  char C = static_cast<char>(CodePoint);
2813  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
2814  }
2815  }
2816 
2817  return 0;
2818 
2819  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
2820  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
2821  // We don't use isLexingRawMode() here because we need to diagnose bad
2822  // UCNs even when skipping preprocessing tokens in a #if block.
2823  if (Result && PP) {
2824  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
2825  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
2826  else
2827  Diag(BufferPtr, diag::err_ucn_escape_invalid);
2828  }
2829  return 0;
2830  }
2831 
2832  return CodePoint;
2833 }
2834 
2835 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
2836  const char *CurPtr) {
2837  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
2839  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
2840  UnicodeWhitespaceChars.contains(C)) {
2841  Diag(BufferPtr, diag::ext_unicode_whitespace)
2842  << makeCharRange(*this, BufferPtr, CurPtr);
2843 
2844  Result.setFlag(Token::LeadingSpace);
2845  return true;
2846  }
2847  return false;
2848 }
2849 
2850 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
2851  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
2853  !PP->isPreprocessedOutput()) {
2855  makeCharRange(*this, BufferPtr, CurPtr),
2856  /*IsFirst=*/true);
2857  }
2858 
2859  MIOpt.ReadToken();
2860  return LexIdentifier(Result, CurPtr);
2861  }
2862 
2864  !PP->isPreprocessedOutput() &&
2865  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
2866  // Non-ASCII characters tend to creep into source code unintentionally.
2867  // Instead of letting the parser complain about the unknown token,
2868  // just drop the character.
2869  // Note that we can /only/ do this when the non-ASCII character is actually
2870  // spelled as Unicode, not written as a UCN. The standard requires that
2871  // we not throw away any possible preprocessor tokens, but there's a
2872  // loophole in the mapping of Unicode characters to basic character set
2873  // characters that allows us to map these particular characters to, say,
2874  // whitespace.
2875  Diag(BufferPtr, diag::err_non_ascii)
2876  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
2877 
2878  BufferPtr = CurPtr;
2879  return false;
2880  }
2881 
2882  // Otherwise, we have an explicit UCN or a character that's unlikely to show
2883  // up by accident.
2884  MIOpt.ReadToken();
2885  FormTokenWithChars(Result, CurPtr, tok::unknown);
2886  return true;
2887 }
2888 
2889 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
2890  IsAtStartOfLine = Result.isAtStartOfLine();
2891  HasLeadingSpace = Result.hasLeadingSpace();
2892  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
2893  // Note that this doesn't affect IsAtPhysicalStartOfLine.
2894 }
2895 
2896 bool Lexer::Lex(Token &Result) {
2897  // Start a new token.
2898  Result.startToken();
2899 
2900  // Set up misc whitespace flags for LexTokenInternal.
2901  if (IsAtStartOfLine) {
2902  Result.setFlag(Token::StartOfLine);
2903  IsAtStartOfLine = false;
2904  }
2905 
2906  if (HasLeadingSpace) {
2907  Result.setFlag(Token::LeadingSpace);
2908  HasLeadingSpace = false;
2909  }
2910 
2911  if (HasLeadingEmptyMacro) {
2913  HasLeadingEmptyMacro = false;
2914  }
2915 
2916  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2917  IsAtPhysicalStartOfLine = false;
2918  bool isRawLex = isLexingRawMode();
2919  (void) isRawLex;
2920  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
2921  // (After the LexTokenInternal call, the lexer might be destroyed.)
2922  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
2923  return returnedToken;
2924 }
2925 
2926 /// LexTokenInternal - This implements a simple C family lexer. It is an
2927 /// extremely performance critical piece of code. This assumes that the buffer
2928 /// has a null character at the end of the file. This returns a preprocessing
2929 /// token, not a normal token, as such, it is an internal interface. It assumes
2930 /// that the Flags of result have been cleared before calling this.
2931 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
2932 LexNextToken:
2933  // New token, can't need cleaning yet.
2935  Result.setIdentifierInfo(nullptr);
2936 
2937  // CurPtr - Cache BufferPtr in an automatic variable.
2938  const char *CurPtr = BufferPtr;
2939 
2940  // Small amounts of horizontal whitespace is very common between tokens.
2941  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
2942  ++CurPtr;
2943  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
2944  ++CurPtr;
2945 
2946  // If we are keeping whitespace and other tokens, just return what we just
2947  // skipped. The next lexer invocation will return the token after the
2948  // whitespace.
2949  if (isKeepWhitespaceMode()) {
2950  FormTokenWithChars(Result, CurPtr, tok::unknown);
2951  // FIXME: The next token will not have LeadingSpace set.
2952  return true;
2953  }
2954 
2955  BufferPtr = CurPtr;
2956  Result.setFlag(Token::LeadingSpace);
2957  }
2958 
2959  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
2960 
2961  // Read a character, advancing over it.
2962  char Char = getAndAdvanceChar(CurPtr, Result);
2964 
2965  switch (Char) {
2966  case 0: // Null.
2967  // Found end of file?
2968  if (CurPtr-1 == BufferEnd)
2969  return LexEndOfFile(Result, CurPtr-1);
2970 
2971  // Check if we are performing code completion.
2972  if (isCodeCompletionPoint(CurPtr-1)) {
2973  // Return the code-completion token.
2974  Result.startToken();
2975  FormTokenWithChars(Result, CurPtr, tok::code_completion);
2976  return true;
2977  }
2978 
2979  if (!isLexingRawMode())
2980  Diag(CurPtr-1, diag::null_in_file);
2981  Result.setFlag(Token::LeadingSpace);
2982  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
2983  return true; // KeepWhitespaceMode
2984 
2985  // We know the lexer hasn't changed, so just try again with this lexer.
2986  // (We manually eliminate the tail call to avoid recursion.)
2987  goto LexNextToken;
2988 
2989  case 26: // DOS & CP/M EOF: "^Z".
2990  // If we're in Microsoft extensions mode, treat this as end of file.
2991  if (LangOpts.MicrosoftExt) {
2992  if (!isLexingRawMode())
2993  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
2994  return LexEndOfFile(Result, CurPtr-1);
2995  }
2996 
2997  // If Microsoft extensions are disabled, this is just random garbage.
2998  Kind = tok::unknown;
2999  break;
3000 
3001  case '\n':
3002  case '\r':
3003  // If we are inside a preprocessor directive and we see the end of line,
3004  // we know we are done with the directive, so return an EOD token.
3006  // Done parsing the "line".
3008 
3009  // Restore comment saving mode, in case it was disabled for directive.
3010  if (PP)
3012 
3013  // Since we consumed a newline, we are back at the start of a line.
3014  IsAtStartOfLine = true;
3015  IsAtPhysicalStartOfLine = true;
3016 
3017  Kind = tok::eod;
3018  break;
3019  }
3020 
3021  // No leading whitespace seen so far.
3023 
3024  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3025  return true; // KeepWhitespaceMode
3026 
3027  // We only saw whitespace, so just try again with this lexer.
3028  // (We manually eliminate the tail call to avoid recursion.)
3029  goto LexNextToken;
3030  case ' ':
3031  case '\t':
3032  case '\f':
3033  case '\v':
3034  SkipHorizontalWhitespace:
3035  Result.setFlag(Token::LeadingSpace);
3036  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3037  return true; // KeepWhitespaceMode
3038 
3039  SkipIgnoredUnits:
3040  CurPtr = BufferPtr;
3041 
3042  // If the next token is obviously a // or /* */ comment, skip it efficiently
3043  // too (without going through the big switch stmt).
3044  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3045  LangOpts.LineComment &&
3046  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3047  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3048  return true; // There is a token to return.
3049  goto SkipIgnoredUnits;
3050  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3051  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3052  return true; // There is a token to return.
3053  goto SkipIgnoredUnits;
3054  } else if (isHorizontalWhitespace(*CurPtr)) {
3055  goto SkipHorizontalWhitespace;
3056  }
3057  // We only saw whitespace, so just try again with this lexer.
3058  // (We manually eliminate the tail call to avoid recursion.)
3059  goto LexNextToken;
3060 
3061  // C99 6.4.4.1: Integer Constants.
3062  // C99 6.4.4.2: Floating Constants.
3063  case '0': case '1': case '2': case '3': case '4':
3064  case '5': case '6': case '7': case '8': case '9':
3065  // Notify MIOpt that we read a non-whitespace/non-comment token.
3066  MIOpt.ReadToken();
3067  return LexNumericConstant(Result, CurPtr);
3068 
3069  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3070  // Notify MIOpt that we read a non-whitespace/non-comment token.
3071  MIOpt.ReadToken();
3072 
3073  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3074  Char = getCharAndSize(CurPtr, SizeTmp);
3075 
3076  // UTF-16 string literal
3077  if (Char == '"')
3078  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3079  tok::utf16_string_literal);
3080 
3081  // UTF-16 character constant
3082  if (Char == '\'')
3083  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3084  tok::utf16_char_constant);
3085 
3086  // UTF-16 raw string literal
3087  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3088  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3089  return LexRawStringLiteral(Result,
3090  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3091  SizeTmp2, Result),
3092  tok::utf16_string_literal);
3093 
3094  if (Char == '8') {
3095  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3096 
3097  // UTF-8 string literal
3098  if (Char2 == '"')
3099  return LexStringLiteral(Result,
3100  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3101  SizeTmp2, Result),
3102  tok::utf8_string_literal);
3103  if (Char2 == '\'' && LangOpts.CPlusPlus1z)
3104  return LexCharConstant(
3105  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3106  SizeTmp2, Result),
3107  tok::utf8_char_constant);
3108 
3109  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3110  unsigned SizeTmp3;
3111  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3112  // UTF-8 raw string literal
3113  if (Char3 == '"') {
3114  return LexRawStringLiteral(Result,
3115  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3116  SizeTmp2, Result),
3117  SizeTmp3, Result),
3118  tok::utf8_string_literal);
3119  }
3120  }
3121  }
3122  }
3123 
3124  // treat u like the start of an identifier.
3125  return LexIdentifier(Result, CurPtr);
3126 
3127  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3128  // Notify MIOpt that we read a non-whitespace/non-comment token.
3129  MIOpt.ReadToken();
3130 
3131  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3132  Char = getCharAndSize(CurPtr, SizeTmp);
3133 
3134  // UTF-32 string literal
3135  if (Char == '"')
3136  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3137  tok::utf32_string_literal);
3138 
3139  // UTF-32 character constant
3140  if (Char == '\'')
3141  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3142  tok::utf32_char_constant);
3143 
3144  // UTF-32 raw string literal
3145  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3146  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3147  return LexRawStringLiteral(Result,
3148  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3149  SizeTmp2, Result),
3150  tok::utf32_string_literal);
3151  }
3152 
3153  // treat U like the start of an identifier.
3154  return LexIdentifier(Result, CurPtr);
3155 
3156  case 'R': // Identifier or C++0x raw string literal
3157  // Notify MIOpt that we read a non-whitespace/non-comment token.
3158  MIOpt.ReadToken();
3159 
3160  if (LangOpts.CPlusPlus11) {
3161  Char = getCharAndSize(CurPtr, SizeTmp);
3162 
3163  if (Char == '"')
3164  return LexRawStringLiteral(Result,
3165  ConsumeChar(CurPtr, SizeTmp, Result),
3166  tok::string_literal);
3167  }
3168 
3169  // treat R like the start of an identifier.
3170  return LexIdentifier(Result, CurPtr);
3171 
3172  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3173  // Notify MIOpt that we read a non-whitespace/non-comment token.
3174  MIOpt.ReadToken();
3175  Char = getCharAndSize(CurPtr, SizeTmp);
3176 
3177  // Wide string literal.
3178  if (Char == '"')
3179  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3180  tok::wide_string_literal);
3181 
3182  // Wide raw string literal.
3183  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3184  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3185  return LexRawStringLiteral(Result,
3186  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3187  SizeTmp2, Result),
3188  tok::wide_string_literal);
3189 
3190  // Wide character constant.
3191  if (Char == '\'')
3192  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3193  tok::wide_char_constant);
3194  // FALL THROUGH, treating L like the start of an identifier.
3195 
3196  // C99 6.4.2: Identifiers.
3197  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3198  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3199  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3200  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3201  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3202  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3203  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3204  case 'v': case 'w': case 'x': case 'y': case 'z':
3205  case '_':
3206  // Notify MIOpt that we read a non-whitespace/non-comment token.
3207  MIOpt.ReadToken();
3208  return LexIdentifier(Result, CurPtr);
3209 
3210  case '$': // $ in identifiers.
3211  if (LangOpts.DollarIdents) {
3212  if (!isLexingRawMode())
3213  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3214  // Notify MIOpt that we read a non-whitespace/non-comment token.
3215  MIOpt.ReadToken();
3216  return LexIdentifier(Result, CurPtr);
3217  }
3218 
3219  Kind = tok::unknown;
3220  break;
3221 
3222  // C99 6.4.4: Character Constants.
3223  case '\'':
3224  // Notify MIOpt that we read a non-whitespace/non-comment token.
3225  MIOpt.ReadToken();
3226  return LexCharConstant(Result, CurPtr, tok::char_constant);
3227 
3228  // C99 6.4.5: String Literals.
3229  case '"':
3230  // Notify MIOpt that we read a non-whitespace/non-comment token.
3231  MIOpt.ReadToken();
3232  return LexStringLiteral(Result, CurPtr, tok::string_literal);
3233 
3234  // C99 6.4.6: Punctuators.
3235  case '?':
3236  Kind = tok::question;
3237  break;
3238  case '[':
3239  Kind = tok::l_square;
3240  break;
3241  case ']':
3242  Kind = tok::r_square;
3243  break;
3244  case '(':
3245  Kind = tok::l_paren;
3246  break;
3247  case ')':
3248  Kind = tok::r_paren;
3249  break;
3250  case '{':
3251  Kind = tok::l_brace;
3252  break;
3253  case '}':
3254  Kind = tok::r_brace;
3255  break;
3256  case '.':
3257  Char = getCharAndSize(CurPtr, SizeTmp);
3258  if (Char >= '0' && Char <= '9') {
3259  // Notify MIOpt that we read a non-whitespace/non-comment token.
3260  MIOpt.ReadToken();
3261 
3262  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3263  } else if (LangOpts.CPlusPlus && Char == '*') {
3264  Kind = tok::periodstar;
3265  CurPtr += SizeTmp;
3266  } else if (Char == '.' &&
3267  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3268  Kind = tok::ellipsis;
3269  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3270  SizeTmp2, Result);
3271  } else {
3272  Kind = tok::period;
3273  }
3274  break;
3275  case '&':
3276  Char = getCharAndSize(CurPtr, SizeTmp);
3277  if (Char == '&') {
3278  Kind = tok::ampamp;
3279  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3280  } else if (Char == '=') {
3281  Kind = tok::ampequal;
3282  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3283  } else {
3284  Kind = tok::amp;
3285  }
3286  break;
3287  case '*':
3288  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3289  Kind = tok::starequal;
3290  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3291  } else {
3292  Kind = tok::star;
3293  }
3294  break;
3295  case '+':
3296  Char = getCharAndSize(CurPtr, SizeTmp);
3297  if (Char == '+') {
3298  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3299  Kind = tok::plusplus;
3300  } else if (Char == '=') {
3301  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3302  Kind = tok::plusequal;
3303  } else {
3304  Kind = tok::plus;
3305  }
3306  break;
3307  case '-':
3308  Char = getCharAndSize(CurPtr, SizeTmp);
3309  if (Char == '-') { // --
3310  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3311  Kind = tok::minusminus;
3312  } else if (Char == '>' && LangOpts.CPlusPlus &&
3313  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3314  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3315  SizeTmp2, Result);
3316  Kind = tok::arrowstar;
3317  } else if (Char == '>') { // ->
3318  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3319  Kind = tok::arrow;
3320  } else if (Char == '=') { // -=
3321  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3322  Kind = tok::minusequal;
3323  } else {
3324  Kind = tok::minus;
3325  }
3326  break;
3327  case '~':
3328  Kind = tok::tilde;
3329  break;
3330  case '!':
3331  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3332  Kind = tok::exclaimequal;
3333  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3334  } else {
3335  Kind = tok::exclaim;
3336  }
3337  break;
3338  case '/':
3339  // 6.4.9: Comments
3340  Char = getCharAndSize(CurPtr, SizeTmp);
3341  if (Char == '/') { // Line comment.
3342  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3343  // want to lex this as a comment. There is one problem with this though,
3344  // that in one particular corner case, this can change the behavior of the
3345  // resultant program. For example, In "foo //**/ bar", C89 would lex
3346  // this as "foo / bar" and langauges with Line comments would lex it as
3347  // "foo". Check to see if the character after the second slash is a '*'.
3348  // If so, we will lex that as a "/" instead of the start of a comment.
3349  // However, we never do this if we are just preprocessing.
3350  bool TreatAsComment = LangOpts.LineComment &&
3351  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3352  if (!TreatAsComment)
3353  if (!(PP && PP->isPreprocessedOutput()))
3354  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3355 
3356  if (TreatAsComment) {
3357  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3358  TokAtPhysicalStartOfLine))
3359  return true; // There is a token to return.
3360 
3361  // It is common for the tokens immediately after a // comment to be
3362  // whitespace (indentation for the next line). Instead of going through
3363  // the big switch, handle it efficiently now.
3364  goto SkipIgnoredUnits;
3365  }
3366  }
3367 
3368  if (Char == '*') { // /**/ comment.
3369  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3370  TokAtPhysicalStartOfLine))
3371  return true; // There is a token to return.
3372 
3373  // We only saw whitespace, so just try again with this lexer.
3374  // (We manually eliminate the tail call to avoid recursion.)
3375  goto LexNextToken;
3376  }
3377 
3378  if (Char == '=') {
3379  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3380  Kind = tok::slashequal;
3381  } else {
3382  Kind = tok::slash;
3383  }
3384  break;
3385  case '%':
3386  Char = getCharAndSize(CurPtr, SizeTmp);
3387  if (Char == '=') {
3388  Kind = tok::percentequal;
3389  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3390  } else if (LangOpts.Digraphs && Char == '>') {
3391  Kind = tok::r_brace; // '%>' -> '}'
3392  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3393  } else if (LangOpts.Digraphs && Char == ':') {
3394  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3395  Char = getCharAndSize(CurPtr, SizeTmp);
3396  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3397  Kind = tok::hashhash; // '%:%:' -> '##'
3398  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3399  SizeTmp2, Result);
3400  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3401  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3402  if (!isLexingRawMode())
3403  Diag(BufferPtr, diag::ext_charize_microsoft);
3404  Kind = tok::hashat;
3405  } else { // '%:' -> '#'
3406  // We parsed a # character. If this occurs at the start of the line,
3407  // it's actually the start of a preprocessing directive. Callback to
3408  // the preprocessor to handle it.
3409  // TODO: -fpreprocessed mode??
3410  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3411  goto HandleDirective;
3412 
3413  Kind = tok::hash;
3414  }
3415  } else {
3416  Kind = tok::percent;
3417  }
3418  break;
3419  case '<':
3420  Char = getCharAndSize(CurPtr, SizeTmp);
3421  if (ParsingFilename) {
3422  return LexAngledStringLiteral(Result, CurPtr);
3423  } else if (Char == '<') {
3424  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3425  if (After == '=') {
3426  Kind = tok::lesslessequal;
3427  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3428  SizeTmp2, Result);
3429  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3430  // If this is actually a '<<<<<<<' version control conflict marker,
3431  // recognize it as such and recover nicely.
3432  goto LexNextToken;
3433  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3434  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3435  // ignore it.
3436  goto LexNextToken;
3437  } else if (LangOpts.CUDA && After == '<') {
3438  Kind = tok::lesslessless;
3439  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3440  SizeTmp2, Result);
3441  } else {
3442  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3443  Kind = tok::lessless;
3444  }
3445  } else if (Char == '=') {
3446  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3447  Kind = tok::lessequal;
3448  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3449  if (LangOpts.CPlusPlus11 &&
3450  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3451  // C++0x [lex.pptoken]p3:
3452  // Otherwise, if the next three characters are <:: and the subsequent
3453  // character is neither : nor >, the < is treated as a preprocessor
3454  // token by itself and not as the first character of the alternative
3455  // token <:.
3456  unsigned SizeTmp3;
3457  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3458  if (After != ':' && After != '>') {
3459  Kind = tok::less;
3460  if (!isLexingRawMode())
3461  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3462  break;
3463  }
3464  }
3465 
3466  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3467  Kind = tok::l_square;
3468  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3469  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3470  Kind = tok::l_brace;
3471  } else {
3472  Kind = tok::less;
3473  }
3474  break;
3475  case '>':
3476  Char = getCharAndSize(CurPtr, SizeTmp);
3477  if (Char == '=') {
3478  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3479  Kind = tok::greaterequal;
3480  } else if (Char == '>') {
3481  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3482  if (After == '=') {
3483  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3484  SizeTmp2, Result);
3485  Kind = tok::greatergreaterequal;
3486  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3487  // If this is actually a '>>>>' conflict marker, recognize it as such
3488  // and recover nicely.
3489  goto LexNextToken;
3490  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3491  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3492  goto LexNextToken;
3493  } else if (LangOpts.CUDA && After == '>') {
3494  Kind = tok::greatergreatergreater;
3495  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3496  SizeTmp2, Result);
3497  } else {
3498  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3499  Kind = tok::greatergreater;
3500  }
3501 
3502  } else {
3503  Kind = tok::greater;
3504  }
3505  break;
3506  case '^':
3507  Char = getCharAndSize(CurPtr, SizeTmp);
3508  if (Char == '=') {
3509  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3510  Kind = tok::caretequal;
3511  } else if (LangOpts.OpenCL && Char == '^') {
3512  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3513  Kind = tok::caretcaret;
3514  } else {
3515  Kind = tok::caret;
3516  }
3517  break;
3518  case '|':
3519  Char = getCharAndSize(CurPtr, SizeTmp);
3520  if (Char == '=') {
3521  Kind = tok::pipeequal;
3522  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3523  } else if (Char == '|') {
3524  // If this is '|||||||' and we're in a conflict marker, ignore it.
3525  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3526  goto LexNextToken;
3527  Kind = tok::pipepipe;
3528  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3529  } else {
3530  Kind = tok::pipe;
3531  }
3532  break;
3533  case ':':
3534  Char = getCharAndSize(CurPtr, SizeTmp);
3535  if (LangOpts.Digraphs && Char == '>') {
3536  Kind = tok::r_square; // ':>' -> ']'
3537  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3538  } else if (LangOpts.CPlusPlus && Char == ':') {
3539  Kind = tok::coloncolon;
3540  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3541  } else {
3542  Kind = tok::colon;
3543  }
3544  break;
3545  case ';':
3546  Kind = tok::semi;
3547  break;
3548  case '=':
3549  Char = getCharAndSize(CurPtr, SizeTmp);
3550  if (Char == '=') {
3551  // If this is '====' and we're in a conflict marker, ignore it.
3552  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3553  goto LexNextToken;
3554 
3555  Kind = tok::equalequal;
3556  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3557  } else {
3558  Kind = tok::equal;
3559  }
3560  break;
3561  case ',':
3562  Kind = tok::comma;
3563  break;
3564  case '#':
3565  Char = getCharAndSize(CurPtr, SizeTmp);
3566  if (Char == '#') {
3567  Kind = tok::hashhash;
3568  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3569  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3570  Kind = tok::hashat;
3571  if (!isLexingRawMode())
3572  Diag(BufferPtr, diag::ext_charize_microsoft);
3573  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3574  } else {
3575  // We parsed a # character. If this occurs at the start of the line,
3576  // it's actually the start of a preprocessing directive. Callback to
3577  // the preprocessor to handle it.
3578  // TODO: -fpreprocessed mode??
3579  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3580  goto HandleDirective;
3581 
3582  Kind = tok::hash;
3583  }
3584  break;
3585 
3586  case '@':
3587  // Objective C support.
3588  if (CurPtr[-1] == '@' && LangOpts.ObjC1)
3589  Kind = tok::at;
3590  else
3591  Kind = tok::unknown;
3592  break;
3593 
3594  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3595  case '\\':
3596  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3597  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3598  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3599  return true; // KeepWhitespaceMode
3600 
3601  // We only saw whitespace, so just try again with this lexer.
3602  // (We manually eliminate the tail call to avoid recursion.)
3603  goto LexNextToken;
3604  }
3605 
3606  return LexUnicode(Result, CodePoint, CurPtr);
3607  }
3608 
3609  Kind = tok::unknown;
3610  break;
3611 
3612  default: {
3613  if (isASCII(Char)) {
3614  Kind = tok::unknown;
3615  break;
3616  }
3617 
3618  UTF32 CodePoint;
3619 
3620  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3621  // an escaped newline.
3622  --CurPtr;
3623  ConversionResult Status =
3624  llvm::convertUTF8Sequence((const UTF8 **)&CurPtr,
3625  (const UTF8 *)BufferEnd,
3626  &CodePoint,
3627  strictConversion);
3628  if (Status == conversionOK) {
3629  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3630  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3631  return true; // KeepWhitespaceMode
3632 
3633  // We only saw whitespace, so just try again with this lexer.
3634  // (We manually eliminate the tail call to avoid recursion.)
3635  goto LexNextToken;
3636  }
3637  return LexUnicode(Result, CodePoint, CurPtr);
3638  }
3639 
3641  PP->isPreprocessedOutput()) {
3642  ++CurPtr;
3643  Kind = tok::unknown;
3644  break;
3645  }
3646 
3647  // Non-ASCII characters tend to creep into source code unintentionally.
3648  // Instead of letting the parser complain about the unknown token,
3649  // just diagnose the invalid UTF-8, then drop the character.
3650  Diag(CurPtr, diag::err_invalid_utf8);
3651 
3652  BufferPtr = CurPtr+1;
3653  // We're pretending the character didn't exist, so just try again with
3654  // this lexer.
3655  // (We manually eliminate the tail call to avoid recursion.)
3656  goto LexNextToken;
3657  }
3658  }
3659 
3660  // Notify MIOpt that we read a non-whitespace/non-comment token.
3661  MIOpt.ReadToken();
3662 
3663  // Update the location of token as well as BufferPtr.
3664  FormTokenWithChars(Result, CurPtr, Kind);
3665  return true;
3666 
3667 HandleDirective:
3668  // We parsed a # character and it's the start of a preprocessing directive.
3669 
3670  FormTokenWithChars(Result, CurPtr, tok::hash);
3671  PP->HandleDirective(Result);
3672 
3674  // With a fatal failure in the module loader, we abort parsing.
3675  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3676  return true;
3677  }
3678 
3679  // We parsed the directive; lex a token with the new state.
3680  return false;
3681 }
bool isAtStartOfLine() const
isAtStartOfLine - Return true if this token is at the start of a line.
Definition: Token.h:265
SourceManager & getSourceManager() const
Definition: Preprocessor.h:694
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:43
bool isMacroArgExpansion(SourceLocation Loc, SourceLocation *StartLoc=nullptr) const
Tests whether the given source location represents a macro argument's expansion into the function-lik...
static unsigned getSpelling(const Token &Tok, const char *&Buffer, const SourceManager &SourceMgr, const LangOptions &LangOpts, bool *Invalid=nullptr)
getSpelling - This method is used to get the spelling of a token into a preallocated buffer...
Definition: Lexer.cpp:358
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens...
Definition: Lexer.h:46
This is a discriminated union of FileInfo and ExpansionInfo.
unsigned Length
SourceLocation getBegin() const
bool isMacroID() const
static std::pair< unsigned, bool > ComputePreamble(StringRef Buffer, const LangOptions &LangOpts, unsigned MaxLines=0)
Compute the preamble of the given file.
Definition: Lexer.cpp:537
void setFlagValue(TokenFlags Flag, bool Val)
Set a flag to either true or false.
Definition: Token.h:256
static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[]
SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
static LLVM_READONLY bool isWhitespace(unsigned char c)
Return true if this character is horizontal or vertical ASCII whitespace: ' ', '\t', '\f', '\v', '\n', '\r'.
Definition: CharInfo.h:88
void setBegin(SourceLocation b)
SourceLocation getSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
static __inline__ int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:13174
Defines the SourceManager interface.
const SrcMgr::SLocEntry & getSLocEntry(FileID FID, bool *Invalid=nullptr) const
const char * getCharacterData(SourceLocation SL, bool *Invalid=nullptr) const
Return a pointer to the start of the specified location in the appropriate spelling MemoryBuffer...
static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1383
StringRef P
llvm::MemoryBuffer * getBuffer(FileID FID, SourceLocation Loc, bool *Invalid=nullptr) const
Return the buffer for the specified FileID.
bool hasLeadingSpace() const
Return true if this token has whitespace before it.
Definition: Token.h:269
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded...
const ExpansionInfo & getExpansion() const
std::unique_ptr< llvm::MemoryBuffer > Buffer
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1124
bool hasUCN() const
Returns true if this token contains a universal character name.
Definition: Token.h:293
void setFlag(TokenFlags Flag)
Set the specified flag.
Definition: Token.h:233
unsigned getRawEncoding() const
When a SourceLocation itself cannot be used, this returns an (opaque) 32-bit integer encoding for it...
bool needsCleaning() const
Return true if this token has trigraphs or escaped newlines in it.
Definition: Token.h:282
static char getCharAndSizeNoWarn(const char *Ptr, unsigned &Size, const LangOptions &LangOpts)
getCharAndSizeNoWarn - Like the getCharAndSize method, but does not ever emit a warning.
Definition: Lexer.h:468
static bool isAtStartOfMacroExpansion(SourceLocation loc, const SourceManager &SM, const LangOptions &LangOpts, SourceLocation *MacroBegin=nullptr)
Returns true if the given MacroID location points at the first token of the macro expansion...
Definition: Lexer.cpp:783
static LLVM_READNONE bool isASCII(char c)
Returns true if this is an ASCII character.
Definition: CharInfo.h:43
bool isStringLiteral(TokenKind K)
Return true if this is a C or C++ string-literal (or C++11 user-defined-string-literal) token...
Definition: TokenKinds.h:79
ConflictMarkerKind
ConflictMarkerKind - Kinds of conflict marker which the lexer might be recovering from...
Definition: Lexer.h:31
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the lexer buffer was all exp...
Definition: Lexer.cpp:1045
Like System, but searched after the system directories.
static Lexer * Create_PragmaLexer(SourceLocation SpellingLoc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLen, Preprocessor &PP)
Create_PragmaLexer: Lexer constructor - Create a new lexer object for _Pragma expansion.
Definition: Lexer.cpp:164
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
Return the 'spelling' of the token at the given location; does not go up to the spelling location or ...
static LLVM_READONLY bool isPreprocessingNumberBody(unsigned char c)
Return true if this is the body character of a C preprocessing number, which is [a-zA-Z0-9_.
Definition: CharInfo.h:148
One of these records is kept for each identifier that is lexed.
bool ParsingPreprocessorDirective
True when parsing #XXX; turns '\n' into a tok::eod token.
StringRef getBufferData(FileID FID, bool *Invalid=nullptr) const
Return a StringRef to the source buffer data for the specified FileID.
void setRawIdentifierData(const char *Ptr)
Definition: Token.h:206
bool isPragmaLexer() const
isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
Definition: Lexer.h:143
static SourceLocation getFromRawEncoding(unsigned Encoding)
Turn a raw encoding of a SourceLocation object into a real SourceLocation.
static LLVM_READONLY bool isHorizontalWhitespace(unsigned char c)
Returns true if this character is horizontal ASCII whitespace: ' ', '\t', '\f', '\v'.
Definition: CharInfo.h:71
bool isFileID() const
SmallVector< PPConditionalInfo, 4 > ConditionalStack
Information about the set of #if/#ifdef/#ifndef blocks we are currently in.
Token - This structure provides full information about a lexed token.
Definition: Token.h:35
void setKind(tok::TokenKind K)
Definition: Token.h:90
Keeps track of the various options that can be enabled, which controls the dialect of C or C++ that i...
Definition: LangOptions.h:48
void resetExtendedTokenMode()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:120
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s...
Definition: Lexer.h:39
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file...
Definition: Lexer.h:221
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi8(char __b)
Initializes all values in a 128-bit vector of [16 x i8] with the specified 8-bit value.
Definition: emmintrin.h:2152
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:449
SourceLocation getLocWithOffset(int Offset) const
Return a source location with the specified offset from this SourceLocation.
uint32_t Offset
Definition: CacheTokens.cpp:44
bool getCommentRetentionState() const
Definition: Preprocessor.h:733
static bool getRawToken(SourceLocation Loc, Token &Result, const SourceManager &SM, const LangOptions &LangOpts, bool IgnoreWhiteSpace=false)
Relex the token at the specified location.
Definition: Lexer.cpp:417
void HandleDirective(Token &Result)
Callback invoked when the lexer sees a # token at the start of a line.
Concrete class used by the front-end to report problems and issues.
Definition: Diagnostic.h:135
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:718
static LLVM_READONLY bool isRawStringDelimBody(unsigned char c)
Return true if this is the body character of a C++ raw string delimiter.
Definition: CharInfo.h:155
SourceLocation getCodeCompletionFileLoc() const
Returns the start location of the file of code-completion point.
tok::TokenKind getKind() const
Definition: Token.h:89
__INTPTR_TYPE__ intptr_t
A signed integer type with the property that any valid pointer to void can be converted to this type...
Definition: opencl-c.h:68
const FileID FID
The SourceManager FileID corresponding to the file being lexed.
static SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, unsigned Character, const SourceManager &SM, const LangOptions &LangOpts)
AdvanceToTokenCharacter - If the current SourceLocation specifies a location at the start of a token...
Definition: Lexer.cpp:700
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L)
isBlockCommentEndOfEscapedNewLine - Return true if the specified newline character (either \n or \r) ...
Definition: Lexer.cpp:2210
bool isInvalid() const
StringRef getRawIdentifier() const
getRawIdentifier - For a raw identifier token (i.e., an identifier lexed in raw mode), returns a reference to the text substring in the buffer if known.
Definition: Token.h:202
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1418
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:873
const FileEntry * getFileEntryForID(FileID FID) const
Returns the FileEntry record for the provided FileID.
bool ParsingFilename
True after #include; turns <xx> into a tok::angle_string_literal token.
FileID getFileID(SourceLocation SpellingLoc) const
Return the FileID for a SourceLocation.
static const llvm::sys::UnicodeCharRange C11DisallowedInitialIDCharRanges[]
const SmallVectorImpl< AnnotatedLine * >::const_iterator End
static StringRef getSourceText(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts, bool *Invalid=nullptr)
Returns a string for the source that the range encompasses.
Definition: Lexer.cpp:922
bool isInFileID(SourceLocation Loc, FileID FID, unsigned *RelativeOffset=nullptr) const
Given a specific FileID, returns true if Loc is inside that FileID chunk and sets relative offset (of...
static bool isAtEndOfMacroExpansion(SourceLocation loc, const SourceManager &SM, const LangOptions &LangOpts, SourceLocation *MacroEnd=nullptr)
Returns true if the given MacroID location points at the last token of the macro expansion.
Definition: Lexer.cpp:805
bool LexingRawMode
True if in raw mode.
static SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset, const SourceManager &SM, const LangOptions &LangOpts)
Computes the source location just past the end of the token at this source location.
Definition: Lexer.cpp:761
Represents a character-granular source range.
SourceLocation getEnd() const
static unsigned MeasureTokenLength(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
MeasureTokenLength - Relex the token at the specified location and return its length in bytes in the ...
Definition: Lexer.cpp:406
static SourceLocation findLocationAfterToken(SourceLocation loc, tok::TokenKind TKind, const SourceManager &SM, const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine)
Checks that the given token is the first token that occurs after the given location (this excludes co...
Definition: Lexer.cpp:1183
Defines the clang::Preprocessor interface.
PreambleDirectiveKind
Definition: Lexer.cpp:529
MultipleIncludeOpt MIOpt
A state machine that detects the #ifndef-wrapping a file idiom for the multiple-include optimization...
void setEnd(SourceLocation e)
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:123
bool HandleEndOfFile(Token &Result, bool isEndOfMacro=false)
Callback invoked when the lexer hits the end of the current file.
SourceLocation createExpansionLoc(SourceLocation Loc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLength, int LoadedID=0, unsigned LoadedOffset=0)
Return a new SourceLocation that encodes the fact that a token from SpellingLoc should actually be re...
SourceLocation getCodeCompletionLoc() const
Returns the location of the code-completion point.
The result type of a method or function.
float __ovld __cnfn length(float p)
Return the length of vector p, i.e., sqrt(p.x2 + p.y 2 + ...)
const SourceManager & SM
Definition: Format.cpp:1184
ObjCKeywordKind
Provides a namespace for Objective-C keywords which start with an '@'.
Definition: TokenKinds.h:41
static CharSourceRange getCharRange(SourceRange R)
const char * getLiteralData() const
getLiteralData - For a literal token (numeric constant, string, etc), this returns a pointer to the s...
Definition: Token.h:214
bool isHandleIdentifierCase() const
Return true if the Preprocessor::HandleIdentifier must be called on a token of this identifier...
Kind
bool isTokenRange() const
Return true if the end of this range specifies the start of the last token.
SmallVectorImpl< AnnotatedLine * >::const_iterator Next
Encodes a location in the source.
bool isValid() const
Return true if this is a valid SourceLocation object.
bool isCharRange() const
bool isAtEndOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroEnd=nullptr) const
Returns true if the given MacroID location points at the character end of the immediate macro expansi...
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1424
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:36
void setIdentifierInfo(IdentifierInfo *II)
Definition: Token.h:185
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1089
static const llvm::sys::UnicodeCharRange C99DisallowedInitialIDCharRanges[]
static SourceLocation GetBeginningOfToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Given a location any where in a source buffer, find the location that corresponds to the beginning of...
Definition: Lexer.cpp:509
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:831
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
bool is(tok::TokenKind K) const
is/isNot - Predicates to check if this token is a specific kind, as in "if (Tok.is(tok::l_brace)) {...
Definition: Token.h:94
static const llvm::sys::UnicodeCharRange C99AllowedIDCharRanges[]
bool isIgnored(unsigned DiagID, SourceLocation Loc) const
Determine whether the diagnostic is known to be ignored.
Definition: Diagnostic.h:652
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:687
__PTRDIFF_TYPE__ ptrdiff_t
A signed integer type that is the result of subtracting two pointers.
Definition: opencl-c.h:61
static StringRef getImmediateMacroName(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:958
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1005
static const llvm::sys::UnicodeCharRange UnicodeWhitespaceCharRanges[]
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
bool inKeepCommentMode() const
inKeepCommentMode - Return true if the lexer should return comments as tokens.
Definition: Lexer.h:180
static CharSourceRange makeFileCharRange(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Accepts a range and returns a character range with file locations.
Definition: Lexer.cpp:858
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:231
bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroBegin=nullptr) const
Returns true if the given MacroID location points at the beginning of the immediate macro expansion...
static FixItHint CreateRemoval(CharSourceRange RemoveRange)
Create a code modification hint that removes the given source range.
Definition: Diagnostic.h:104
std::pair< SourceLocation, SourceLocation > getImmediateExpansionRange(SourceLocation Loc) const
Return the start/end of the expansion information for an expansion location.
SourceLocation getSourceLocation(const char *Loc, unsigned TokLen=1) const
getSourceLocation - Return a source location identifier for the specified offset in the current file...
Definition: Lexer.cpp:1070
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
detail::InMemoryDirectory::const_iterator E
SourceLocation getExpansionLocStart() const
void setLiteralData(const char *Ptr)
Definition: Token.h:218
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:112
static const llvm::sys::UnicodeCharRange CXX03AllowedIDCharRanges[]
bool HandleIdentifier(Token &Identifier)
Callback invoked when the lexer reads an identifier and has filled in the tokens IdentifierInfo membe...
bool isInvalid() const
void CreateString(StringRef Str, Token &Tok, SourceLocation ExpansionLocStart=SourceLocation(), SourceLocation ExpansionLocEnd=SourceLocation())
Plop the specified string into a scratch buffer and set the specified token's location and length to ...
tok::ObjCKeywordKind getObjCKeywordID() const
Return the Objective-C keyword ID for the this identifier.
static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1401
bool hasLeadingEmptyMacro() const
Return true if this token has an empty macro before it.
Definition: Token.h:286
bool isCodeCompletionEnabled() const
Determine if we are performing code completion.
static FixItHint CreateInsertion(SourceLocation InsertionLoc, StringRef Code, bool BeforePreviousInsertions=false)
Create a code modification hint that inserts the given code string at a specific location.
Definition: Diagnostic.h:78
static char GetTrigraphCharForLetter(char Letter)
GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, return the decoded trigraph...
Definition: Lexer.cpp:1099
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1030
static LLVM_READONLY bool isIdentifierBody(unsigned char c, bool AllowDollar=false)
Returns true if this is a body character of a C identifier, which is [a-zA-Z0-9_].
Definition: CharInfo.h:59
bool HandleComment(Token &Token, SourceRange Comment)
const LangOptions & getLangOpts() const
getLangOpts - Return the language features currently enabled.
Definition: Lexer.h:128
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string...
Definition: Lexer.cpp:2452
Not within a conflict marker.
Definition: Lexer.h:33
static LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
Returns true if this character is vertical ASCII whitespace: '\n', '\r'.
Definition: CharInfo.h:79
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
static char DecodeTrigraphChar(const char *CP, Lexer *L)
DecodeTrigraphChar - If the specified character is a legal trigraph when prefixed with ...
Definition: Lexer.cpp:1118
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2609
static FixItHint CreateReplacement(CharSourceRange RemoveRange, StringRef Code)
Create a code modification hint that replaces the given source range with the given code string...
Definition: Diagnostic.h:115
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode...
Definition: Lexer.h:187
static __inline__ int __DEFAULT_FN_ATTRS _mm_movemask_epi8(__m128i __a)
Definition: emmintrin.h:2295
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
Given a tok::raw_identifier token, look up the identifier information for the token and install it in...
unsigned getLength() const
Definition: Token.h:126
SourceLocation getLocForStartOfFile(FileID FID) const
Return the source location corresponding to the first byte of the specified file. ...
bool isKeepWhitespaceMode() const
isKeepWhitespaceMode - Return true if the lexer should return tokens for every character in the file...
Definition: Lexer.h:166
bool isPreprocessedOutput() const
Returns true if the preprocessor is responsible for generating output, false if it is producing token...
Definition: Preprocessor.h:754
static LLVM_READONLY bool isIdentifierHead(unsigned char c, bool AllowDollar=false)
Returns true if this is a valid first character of a C identifier, which is [a-zA-Z_].
Definition: CharInfo.h:49
A normal or diff3 conflict marker, initiated by at least 7 "<"s, separated by at least 7 "="s or "|"s...
Definition: Lexer.h:36
A trivial tuple used to represent a source range.
void clearFlag(TokenFlags Flag)
Unset the specified flag.
Definition: Token.h:243
SourceLocation getExpansionLoc(SourceLocation Loc) const
Given a SourceLocation object Loc, return the expansion location referenced by the ID...
std::pair< FileID, unsigned > getDecomposedLoc(SourceLocation Loc) const
Decompose the specified location into a raw FileID + Offset pair.
void SetKeepWhitespaceMode(bool Val)
SetKeepWhitespaceMode - This method lets clients enable or disable whitespace retention mode...
Definition: Lexer.h:172
This class handles loading and caching of source files into memory.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding 8-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:1488
void startToken()
Reset all flags to cleared.
Definition: Token.h:168
static std::string Stringify(StringRef Str, bool Charify=false)
Stringify - Convert the specified string into a C string by escaping '\' and " characters. This does not add surrounding ""'s to the string.
Definition: Lexer.cpp:202
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:97
SourceLocation getSpellingLoc() const
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:176