clang  3.7.0
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.is(tok::string_literal)) {
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, --CharNo, ++PhysOffset;
723  }
724 
725  // If we have a character that may be a trigraph or escaped newline, use a
726  // lexer to parse it correctly.
727  for (; CharNo; --CharNo) {
728  unsigned Size;
729  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
730  TokPtr += Size;
731  PhysOffset += Size;
732  }
733 
734  // Final detail: if we end up on an escaped newline, we want to return the
735  // location of the actual byte of the token. For example foo<newline>bar
736  // advanced by 3 should return the location of b, not of \\. One compounding
737  // detail of this is that the escape may be made by a trigraph.
738  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
739  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
740 
741  return TokStart.getLocWithOffset(PhysOffset);
742 }
743 
744 /// \brief Computes the source location just past the end of the
745 /// token at this source location.
746 ///
747 /// This routine can be used to produce a source location that
748 /// points just past the end of the token referenced by \p Loc, and
749 /// is generally used when a diagnostic needs to point just after a
750 /// token where it expected something different that it received. If
751 /// the returned source location would not be meaningful (e.g., if
752 /// it points into a macro), this routine returns an invalid
753 /// source location.
754 ///
755 /// \param Offset an offset from the end of the token, where the source
756 /// location should refer to. The default offset (0) produces a source
757 /// location pointing just past the end of the token; an offset of 1 produces
758 /// a source location pointing to the last character in the token, etc.
760  const SourceManager &SM,
761  const LangOptions &LangOpts) {
762  if (Loc.isInvalid())
763  return SourceLocation();
764 
765  if (Loc.isMacroID()) {
766  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
767  return SourceLocation(); // Points inside the macro expansion.
768  }
769 
770  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
771  if (Len > Offset)
772  Len = Len - Offset;
773  else
774  return Loc;
775 
776  return Loc.getLocWithOffset(Len);
777 }
778 
779 /// \brief Returns true if the given MacroID location points at the first
780 /// token of the macro expansion.
782  const SourceManager &SM,
783  const LangOptions &LangOpts,
784  SourceLocation *MacroBegin) {
785  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
786 
787  SourceLocation expansionLoc;
788  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
789  return false;
790 
791  if (expansionLoc.isFileID()) {
792  // No other macro expansions, this is the first.
793  if (MacroBegin)
794  *MacroBegin = expansionLoc;
795  return true;
796  }
797 
798  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
799 }
800 
801 /// \brief Returns true if the given MacroID location points at the last
802 /// token of the macro expansion.
804  const SourceManager &SM,
805  const LangOptions &LangOpts,
806  SourceLocation *MacroEnd) {
807  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
808 
809  SourceLocation spellLoc = SM.getSpellingLoc(loc);
810  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
811  if (tokLen == 0)
812  return false;
813 
814  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
815  SourceLocation expansionLoc;
816  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
817  return false;
818 
819  if (expansionLoc.isFileID()) {
820  // No other macro expansions.
821  if (MacroEnd)
822  *MacroEnd = expansionLoc;
823  return true;
824  }
825 
826  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
827 }
828 
830  const SourceManager &SM,
831  const LangOptions &LangOpts) {
832  SourceLocation Begin = Range.getBegin();
833  SourceLocation End = Range.getEnd();
834  assert(Begin.isFileID() && End.isFileID());
835  if (Range.isTokenRange()) {
836  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
837  if (End.isInvalid())
838  return CharSourceRange();
839  }
840 
841  // Break down the source locations.
842  FileID FID;
843  unsigned BeginOffs;
844  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
845  if (FID.isInvalid())
846  return CharSourceRange();
847 
848  unsigned EndOffs;
849  if (!SM.isInFileID(End, FID, &EndOffs) ||
850  BeginOffs > EndOffs)
851  return CharSourceRange();
852 
853  return CharSourceRange::getCharRange(Begin, End);
854 }
855 
857  const SourceManager &SM,
858  const LangOptions &LangOpts) {
859  SourceLocation Begin = Range.getBegin();
860  SourceLocation End = Range.getEnd();
861  if (Begin.isInvalid() || End.isInvalid())
862  return CharSourceRange();
863 
864  if (Begin.isFileID() && End.isFileID())
865  return makeRangeFromFileLocs(Range, SM, LangOpts);
866 
867  if (Begin.isMacroID() && End.isFileID()) {
868  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
869  return CharSourceRange();
870  Range.setBegin(Begin);
871  return makeRangeFromFileLocs(Range, SM, LangOpts);
872  }
873 
874  if (Begin.isFileID() && End.isMacroID()) {
875  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
876  &End)) ||
877  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
878  &End)))
879  return CharSourceRange();
880  Range.setEnd(End);
881  return makeRangeFromFileLocs(Range, SM, LangOpts);
882  }
883 
884  assert(Begin.isMacroID() && End.isMacroID());
885  SourceLocation MacroBegin, MacroEnd;
886  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
887  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
888  &MacroEnd)) ||
889  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
890  &MacroEnd)))) {
891  Range.setBegin(MacroBegin);
892  Range.setEnd(MacroEnd);
893  return makeRangeFromFileLocs(Range, SM, LangOpts);
894  }
895 
896  bool Invalid = false;
897  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
898  &Invalid);
899  if (Invalid)
900  return CharSourceRange();
901 
902  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
903  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
904  &Invalid);
905  if (Invalid)
906  return CharSourceRange();
907 
908  if (EndEntry.getExpansion().isMacroArgExpansion() &&
909  BeginEntry.getExpansion().getExpansionLocStart() ==
910  EndEntry.getExpansion().getExpansionLocStart()) {
911  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
912  Range.setEnd(SM.getImmediateSpellingLoc(End));
913  return makeFileCharRange(Range, SM, LangOpts);
914  }
915  }
916 
917  return CharSourceRange();
918 }
919 
921  const SourceManager &SM,
922  const LangOptions &LangOpts,
923  bool *Invalid) {
924  Range = makeFileCharRange(Range, SM, LangOpts);
925  if (Range.isInvalid()) {
926  if (Invalid) *Invalid = true;
927  return StringRef();
928  }
929 
930  // Break down the source location.
931  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
932  if (beginInfo.first.isInvalid()) {
933  if (Invalid) *Invalid = true;
934  return StringRef();
935  }
936 
937  unsigned EndOffs;
938  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
939  beginInfo.second > EndOffs) {
940  if (Invalid) *Invalid = true;
941  return StringRef();
942  }
943 
944  // Try to the load the file buffer.
945  bool invalidTemp = false;
946  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
947  if (invalidTemp) {
948  if (Invalid) *Invalid = true;
949  return StringRef();
950  }
951 
952  if (Invalid) *Invalid = false;
953  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
954 }
955 
957  const SourceManager &SM,
958  const LangOptions &LangOpts) {
959  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
960 
961  // Find the location of the immediate macro expansion.
962  while (1) {
963  FileID FID = SM.getFileID(Loc);
964  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
965  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
966  Loc = Expansion.getExpansionLocStart();
967  if (!Expansion.isMacroArgExpansion())
968  break;
969 
970  // For macro arguments we need to check that the argument did not come
971  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
972 
973  // Loc points to the argument id of the macro definition, move to the
974  // macro expansion.
975  Loc = SM.getImmediateExpansionRange(Loc).first;
976  SourceLocation SpellLoc = Expansion.getSpellingLoc();
977  if (SpellLoc.isFileID())
978  break; // No inner macro.
979 
980  // If spelling location resides in the same FileID as macro expansion
981  // location, it means there is no inner macro.
982  FileID MacroFID = SM.getFileID(Loc);
983  if (SM.isInFileID(SpellLoc, MacroFID))
984  break;
985 
986  // Argument came from inner macro.
987  Loc = SpellLoc;
988  }
989 
990  // Find the spelling location of the start of the non-argument expansion
991  // range. This is where the macro name was spelled in order to begin
992  // expanding this macro.
993  Loc = SM.getSpellingLoc(Loc);
994 
995  // Dig out the buffer where the macro name was spelled and the extents of the
996  // name so that we can render it into the expansion note.
997  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
998  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
999  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1000  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1001 }
1002 
1003 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1004  return isIdentifierBody(c, LangOpts.DollarIdents);
1005 }
1006 
1007 
1008 //===----------------------------------------------------------------------===//
1009 // Diagnostics forwarding code.
1010 //===----------------------------------------------------------------------===//
1011 
1012 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1013 /// lexer buffer was all expanded at a single point, perform the mapping.
1014 /// This is currently only used for _Pragma implementation, so it is the slow
1015 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1016 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1017  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1019  SourceLocation FileLoc,
1020  unsigned CharNo, unsigned TokLen) {
1021  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1022 
1023  // Otherwise, we're lexing "mapped tokens". This is used for things like
1024  // _Pragma handling. Combine the expansion location of FileLoc with the
1025  // spelling location.
1027 
1028  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1029  // characters come from spelling(FileLoc)+Offset.
1030  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1031  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1032 
1033  // Figure out the expansion loc range, which is the range covered by the
1034  // original _Pragma(...) sequence.
1035  std::pair<SourceLocation,SourceLocation> II =
1036  SM.getImmediateExpansionRange(FileLoc);
1037 
1038  return SM.createExpansionLoc(SpellingLoc, II.first, II.second, TokLen);
1039 }
1040 
1041 /// getSourceLocation - Return a source location identifier for the specified
1042 /// offset in the current file.
1044  unsigned TokLen) const {
1045  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1046  "Location out of range for this buffer!");
1047 
1048  // In the normal case, we're just lexing from a simple file buffer, return
1049  // the file id from FileLoc with the offset specified.
1050  unsigned CharNo = Loc-BufferStart;
1051  if (FileLoc.isFileID())
1052  return FileLoc.getLocWithOffset(CharNo);
1053 
1054  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1055  // tokens are lexed from where the _Pragma was defined.
1056  assert(PP && "This doesn't work on raw lexers");
1057  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1058 }
1059 
1060 /// Diag - Forwarding function for diagnostics. This translate a source
1061 /// position in the current buffer into a SourceLocation object for rendering.
1062 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1063  return PP->Diag(getSourceLocation(Loc), DiagID);
1064 }
1065 
1066 //===----------------------------------------------------------------------===//
1067 // Trigraph and Escaped Newline Handling Code.
1068 //===----------------------------------------------------------------------===//
1069 
1070 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1071 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1072 static char GetTrigraphCharForLetter(char Letter) {
1073  switch (Letter) {
1074  default: return 0;
1075  case '=': return '#';
1076  case ')': return ']';
1077  case '(': return '[';
1078  case '!': return '|';
1079  case '\'': return '^';
1080  case '>': return '}';
1081  case '/': return '\\';
1082  case '<': return '{';
1083  case '-': return '~';
1084  }
1085 }
1086 
1087 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1088 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1089 /// return the result character. Finally, emit a warning about trigraph use
1090 /// whether trigraphs are enabled or not.
1091 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1092  char Res = GetTrigraphCharForLetter(*CP);
1093  if (!Res || !L) return Res;
1094 
1095  if (!L->getLangOpts().Trigraphs) {
1096  if (!L->isLexingRawMode())
1097  L->Diag(CP-2, diag::trigraph_ignored);
1098  return 0;
1099  }
1100 
1101  if (!L->isLexingRawMode())
1102  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1103  return Res;
1104 }
1105 
1106 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1107 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1108 /// trigraph equivalent on entry to this function.
1109 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1110  unsigned Size = 0;
1111  while (isWhitespace(Ptr[Size])) {
1112  ++Size;
1113 
1114  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1115  continue;
1116 
1117  // If this is a \r\n or \n\r, skip the other half.
1118  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1119  Ptr[Size-1] != Ptr[Size])
1120  ++Size;
1121 
1122  return Size;
1123  }
1124 
1125  // Not an escaped newline, must be a \t or something else.
1126  return 0;
1127 }
1128 
1129 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1130 /// them), skip over them and return the first non-escaped-newline found,
1131 /// otherwise return P.
1132 const char *Lexer::SkipEscapedNewLines(const char *P) {
1133  while (1) {
1134  const char *AfterEscape;
1135  if (*P == '\\') {
1136  AfterEscape = P+1;
1137  } else if (*P == '?') {
1138  // If not a trigraph for escape, bail out.
1139  if (P[1] != '?' || P[2] != '/')
1140  return P;
1141  AfterEscape = P+3;
1142  } else {
1143  return P;
1144  }
1145 
1146  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1147  if (NewLineSize == 0) return P;
1148  P = AfterEscape+NewLineSize;
1149  }
1150 }
1151 
1152 /// \brief Checks that the given token is the first token that occurs after the
1153 /// given location (this excludes comments and whitespace). Returns the location
1154 /// immediately after the specified token. If the token is not found or the
1155 /// location is inside a macro, the returned source location will be invalid.
1157  tok::TokenKind TKind,
1158  const SourceManager &SM,
1159  const LangOptions &LangOpts,
1160  bool SkipTrailingWhitespaceAndNewLine) {
1161  if (Loc.isMacroID()) {
1162  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1163  return SourceLocation();
1164  }
1165  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1166 
1167  // Break down the source location.
1168  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1169 
1170  // Try to load the file buffer.
1171  bool InvalidTemp = false;
1172  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1173  if (InvalidTemp)
1174  return SourceLocation();
1175 
1176  const char *TokenBegin = File.data() + LocInfo.second;
1177 
1178  // Lex from the start of the given location.
1179  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1180  TokenBegin, File.end());
1181  // Find the token.
1182  Token Tok;
1183  lexer.LexFromRawLexer(Tok);
1184  if (Tok.isNot(TKind))
1185  return SourceLocation();
1186  SourceLocation TokenLoc = Tok.getLocation();
1187 
1188  // Calculate how much whitespace needs to be skipped if any.
1189  unsigned NumWhitespaceChars = 0;
1190  if (SkipTrailingWhitespaceAndNewLine) {
1191  const char *TokenEnd = SM.getCharacterData(TokenLoc) +
1192  Tok.getLength();
1193  unsigned char C = *TokenEnd;
1194  while (isHorizontalWhitespace(C)) {
1195  C = *(++TokenEnd);
1196  NumWhitespaceChars++;
1197  }
1198 
1199  // Skip \r, \n, \r\n, or \n\r
1200  if (C == '\n' || C == '\r') {
1201  char PrevC = C;
1202  C = *(++TokenEnd);
1203  NumWhitespaceChars++;
1204  if ((C == '\n' || C == '\r') && C != PrevC)
1205  NumWhitespaceChars++;
1206  }
1207  }
1208 
1209  return TokenLoc.getLocWithOffset(Tok.getLength() + NumWhitespaceChars);
1210 }
1211 
1212 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1213 /// get its size, and return it. This is tricky in several cases:
1214 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1215 /// then either return the trigraph (skipping 3 chars) or the '?',
1216 /// depending on whether trigraphs are enabled or not.
1217 /// 2. If this is an escaped newline (potentially with whitespace between
1218 /// the backslash and newline), implicitly skip the newline and return
1219 /// the char after it.
1220 ///
1221 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1222 /// know that we can accumulate into Size, and that we have already incremented
1223 /// Ptr by Size bytes.
1224 ///
1225 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1226 /// be updated to match.
1227 ///
1228 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1229  Token *Tok) {
1230  // If we have a slash, look for an escaped newline.
1231  if (Ptr[0] == '\\') {
1232  ++Size;
1233  ++Ptr;
1234 Slash:
1235  // Common case, backslash-char where the char is not whitespace.
1236  if (!isWhitespace(Ptr[0])) return '\\';
1237 
1238  // See if we have optional whitespace characters between the slash and
1239  // newline.
1240  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1241  // Remember that this token needs to be cleaned.
1242  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1243 
1244  // Warn if there was whitespace between the backslash and newline.
1245  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1246  Diag(Ptr, diag::backslash_newline_space);
1247 
1248  // Found backslash<whitespace><newline>. Parse the char after it.
1249  Size += EscapedNewLineSize;
1250  Ptr += EscapedNewLineSize;
1251 
1252  // If the char that we finally got was a \n, then we must have had
1253  // something like <newline><newline>. We don't want to consume the
1254  // second newline.
1255  if (*Ptr == '\n' || *Ptr == '\r' || *Ptr == '\0')
1256  return ' ';
1257 
1258  // Use slow version to accumulate a correct size field.
1259  return getCharAndSizeSlow(Ptr, Size, Tok);
1260  }
1261 
1262  // Otherwise, this is not an escaped newline, just return the slash.
1263  return '\\';
1264  }
1265 
1266  // If this is a trigraph, process it.
1267  if (Ptr[0] == '?' && Ptr[1] == '?') {
1268  // If this is actually a legal trigraph (not something like "??x"), emit
1269  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1270  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1271  // Remember that this token needs to be cleaned.
1272  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1273 
1274  Ptr += 3;
1275  Size += 3;
1276  if (C == '\\') goto Slash;
1277  return C;
1278  }
1279  }
1280 
1281  // If this is neither, return a single character.
1282  ++Size;
1283  return *Ptr;
1284 }
1285 
1286 
1287 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1288 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1289 /// and that we have already incremented Ptr by Size bytes.
1290 ///
1291 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1292 /// be updated to match.
1293 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1294  const LangOptions &LangOpts) {
1295  // If we have a slash, look for an escaped newline.
1296  if (Ptr[0] == '\\') {
1297  ++Size;
1298  ++Ptr;
1299 Slash:
1300  // Common case, backslash-char where the char is not whitespace.
1301  if (!isWhitespace(Ptr[0])) return '\\';
1302 
1303  // See if we have optional whitespace characters followed by a newline.
1304  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1305  // Found backslash<whitespace><newline>. Parse the char after it.
1306  Size += EscapedNewLineSize;
1307  Ptr += EscapedNewLineSize;
1308 
1309  // If the char that we finally got was a \n, then we must have had
1310  // something like <newline><newline>. We don't want to consume the
1311  // second newline.
1312  if (*Ptr == '\n' || *Ptr == '\r' || *Ptr == '\0')
1313  return ' ';
1314 
1315  // Use slow version to accumulate a correct size field.
1316  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1317  }
1318 
1319  // Otherwise, this is not an escaped newline, just return the slash.
1320  return '\\';
1321  }
1322 
1323  // If this is a trigraph, process it.
1324  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1325  // If this is actually a legal trigraph (not something like "??x"), return
1326  // it.
1327  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1328  Ptr += 3;
1329  Size += 3;
1330  if (C == '\\') goto Slash;
1331  return C;
1332  }
1333  }
1334 
1335  // If this is neither, return a single character.
1336  ++Size;
1337  return *Ptr;
1338 }
1339 
1340 //===----------------------------------------------------------------------===//
1341 // Helper methods for lexing.
1342 //===----------------------------------------------------------------------===//
1343 
1344 /// \brief Routine that indiscriminately skips bytes in the source file.
1345 void Lexer::SkipBytes(unsigned Bytes, bool StartOfLine) {
1346  BufferPtr += Bytes;
1347  if (BufferPtr > BufferEnd)
1348  BufferPtr = BufferEnd;
1349  // FIXME: What exactly does the StartOfLine bit mean? There are two
1350  // possible meanings for the "start" of the line: the first token on the
1351  // unexpanded line, or the first token on the expanded line.
1352  IsAtStartOfLine = StartOfLine;
1353  IsAtPhysicalStartOfLine = StartOfLine;
1354 }
1355 
1356 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1357  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1358  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1360  return C11AllowedIDChars.contains(C);
1361  } else if (LangOpts.CPlusPlus) {
1362  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1364  return CXX03AllowedIDChars.contains(C);
1365  } else {
1366  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1368  return C99AllowedIDChars.contains(C);
1369  }
1370 }
1371 
1372 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1373  assert(isAllowedIDChar(C, LangOpts));
1374  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1375  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1377  return !C11DisallowedInitialIDChars.contains(C);
1378  } else if (LangOpts.CPlusPlus) {
1379  return true;
1380  } else {
1381  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1383  return !C99DisallowedInitialIDChars.contains(C);
1384  }
1385 }
1386 
1387 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1388  const char *End) {
1390  L.getSourceLocation(End));
1391 }
1392 
1393 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1394  CharSourceRange Range, bool IsFirst) {
1395  // Check C99 compatibility.
1396  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1397  enum {
1398  CannotAppearInIdentifier = 0,
1399  CannotStartIdentifier
1400  };
1401 
1402  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1404  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1406  if (!C99AllowedIDChars.contains(C)) {
1407  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1408  << Range
1409  << CannotAppearInIdentifier;
1410  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1411  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1412  << Range
1413  << CannotStartIdentifier;
1414  }
1415  }
1416 
1417  // Check C++98 compatibility.
1418  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1419  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1421  if (!CXX03AllowedIDChars.contains(C)) {
1422  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1423  << Range;
1424  }
1425  }
1426 }
1427 
1428 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1429  Token &Result) {
1430  const char *UCNPtr = CurPtr + Size;
1431  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1432  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1433  return false;
1434 
1435  if (!isLexingRawMode())
1437  makeCharRange(*this, CurPtr, UCNPtr),
1438  /*IsFirst=*/false);
1439 
1440  Result.setFlag(Token::HasUCN);
1441  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1442  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1443  CurPtr = UCNPtr;
1444  else
1445  while (CurPtr != UCNPtr)
1446  (void)getAndAdvanceChar(CurPtr, Result);
1447  return true;
1448 }
1449 
1450 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1451  const char *UnicodePtr = CurPtr;
1452  UTF32 CodePoint;
1453  ConversionResult Result =
1454  llvm::convertUTF8Sequence((const UTF8 **)&UnicodePtr,
1455  (const UTF8 *)BufferEnd,
1456  &CodePoint,
1457  strictConversion);
1458  if (Result != conversionOK ||
1459  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1460  return false;
1461 
1462  if (!isLexingRawMode())
1464  makeCharRange(*this, CurPtr, UnicodePtr),
1465  /*IsFirst=*/false);
1466 
1467  CurPtr = UnicodePtr;
1468  return true;
1469 }
1470 
1471 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1472  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1473  unsigned Size;
1474  unsigned char C = *CurPtr++;
1475  while (isIdentifierBody(C))
1476  C = *CurPtr++;
1477 
1478  --CurPtr; // Back up over the skipped character.
1479 
1480  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1481  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1482  //
1483  // TODO: Could merge these checks into an InfoTable flag to make the
1484  // comparison cheaper
1485  if (isASCII(C) && C != '\\' && C != '?' &&
1486  (C != '$' || !LangOpts.DollarIdents)) {
1487 FinishIdentifier:
1488  const char *IdStart = BufferPtr;
1489  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1490  Result.setRawIdentifierData(IdStart);
1491 
1492  // If we are in raw mode, return this identifier raw. There is no need to
1493  // look up identifier information or attempt to macro expand it.
1494  if (LexingRawMode)
1495  return true;
1496 
1497  // Fill in Result.IdentifierInfo and update the token kind,
1498  // looking up the identifier in the identifier table.
1499  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1500 
1501  // Finally, now that we know we have an identifier, pass this off to the
1502  // preprocessor, which may macro expand it or something.
1503  if (II->isHandleIdentifierCase())
1504  return PP->HandleIdentifier(Result);
1505 
1506  return true;
1507  }
1508 
1509  // Otherwise, $,\,? in identifier found. Enter slower path.
1510 
1511  C = getCharAndSize(CurPtr, Size);
1512  while (1) {
1513  if (C == '$') {
1514  // If we hit a $ and they are not supported in identifiers, we are done.
1515  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1516 
1517  // Otherwise, emit a diagnostic and continue.
1518  if (!isLexingRawMode())
1519  Diag(CurPtr, diag::ext_dollar_in_identifier);
1520  CurPtr = ConsumeChar(CurPtr, Size, Result);
1521  C = getCharAndSize(CurPtr, Size);
1522  continue;
1523 
1524  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1525  C = getCharAndSize(CurPtr, Size);
1526  continue;
1527  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1528  C = getCharAndSize(CurPtr, Size);
1529  continue;
1530  } else if (!isIdentifierBody(C)) {
1531  goto FinishIdentifier;
1532  }
1533 
1534  // Otherwise, this character is good, consume it.
1535  CurPtr = ConsumeChar(CurPtr, Size, Result);
1536 
1537  C = getCharAndSize(CurPtr, Size);
1538  while (isIdentifierBody(C)) {
1539  CurPtr = ConsumeChar(CurPtr, Size, Result);
1540  C = getCharAndSize(CurPtr, Size);
1541  }
1542  }
1543 }
1544 
1545 /// isHexaLiteral - Return true if Start points to a hex constant.
1546 /// in microsoft mode (where this is supposed to be several different tokens).
1547 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1548  unsigned Size;
1549  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1550  if (C1 != '0')
1551  return false;
1552  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1553  return (C2 == 'x' || C2 == 'X');
1554 }
1555 
1556 /// LexNumericConstant - Lex the remainder of a integer or floating point
1557 /// constant. From[-1] is the first character lexed. Return the end of the
1558 /// constant.
1559 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1560  unsigned Size;
1561  char C = getCharAndSize(CurPtr, Size);
1562  char PrevCh = 0;
1563  while (isPreprocessingNumberBody(C)) {
1564  CurPtr = ConsumeChar(CurPtr, Size, Result);
1565  PrevCh = C;
1566  C = getCharAndSize(CurPtr, Size);
1567  }
1568 
1569  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1570  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1571  // If we are in Microsoft mode, don't continue if the constant is hex.
1572  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1573  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1574  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1575  }
1576 
1577  // If we have a hex FP constant, continue.
1578  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1579  // Outside C99, we accept hexadecimal floating point numbers as a
1580  // not-quite-conforming extension. Only do so if this looks like it's
1581  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1582  bool IsHexFloat = true;
1583  if (!LangOpts.C99) {
1584  if (!isHexaLiteral(BufferPtr, LangOpts))
1585  IsHexFloat = false;
1586  else if (std::find(BufferPtr, CurPtr, '_') != CurPtr)
1587  IsHexFloat = false;
1588  }
1589  if (IsHexFloat)
1590  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1591  }
1592 
1593  // If we have a digit separator, continue.
1594  if (C == '\'' && getLangOpts().CPlusPlus14) {
1595  unsigned NextSize;
1596  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1597  if (isIdentifierBody(Next)) {
1598  if (!isLexingRawMode())
1599  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1600  CurPtr = ConsumeChar(CurPtr, Size, Result);
1601  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1602  return LexNumericConstant(Result, CurPtr);
1603  }
1604  }
1605 
1606  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1607  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1608  return LexNumericConstant(Result, CurPtr);
1609  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1610  return LexNumericConstant(Result, CurPtr);
1611 
1612  // Update the location of token as well as BufferPtr.
1613  const char *TokStart = BufferPtr;
1614  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1615  Result.setLiteralData(TokStart);
1616  return true;
1617 }
1618 
1619 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1620 /// in C++11, or warn on a ud-suffix in C++98.
1621 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1622  bool IsStringLiteral) {
1623  assert(getLangOpts().CPlusPlus);
1624 
1625  // Maximally munch an identifier.
1626  unsigned Size;
1627  char C = getCharAndSize(CurPtr, Size);
1628  bool Consumed = false;
1629 
1630  if (!isIdentifierHead(C)) {
1631  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1632  Consumed = true;
1633  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1634  Consumed = true;
1635  else
1636  return CurPtr;
1637  }
1638 
1639  if (!getLangOpts().CPlusPlus11) {
1640  if (!isLexingRawMode())
1641  Diag(CurPtr,
1642  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1643  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1645  return CurPtr;
1646  }
1647 
1648  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1649  // that does not start with an underscore is ill-formed. As a conforming
1650  // extension, we treat all such suffixes as if they had whitespace before
1651  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1652  // likely to be a ud-suffix than a macro, however, and accept that.
1653  if (!Consumed) {
1654  bool IsUDSuffix = false;
1655  if (C == '_')
1656  IsUDSuffix = true;
1657  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1658  // In C++1y, we need to look ahead a few characters to see if this is a
1659  // valid suffix for a string literal or a numeric literal (this could be
1660  // the 'operator""if' defining a numeric literal operator).
1661  const unsigned MaxStandardSuffixLength = 3;
1662  char Buffer[MaxStandardSuffixLength] = { C };
1663  unsigned Consumed = Size;
1664  unsigned Chars = 1;
1665  while (true) {
1666  unsigned NextSize;
1667  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1668  getLangOpts());
1669  if (!isIdentifierBody(Next)) {
1670  // End of suffix. Check whether this is on the whitelist.
1671  IsUDSuffix = (Chars == 1 && Buffer[0] == 's') ||
1673  getLangOpts(), StringRef(Buffer, Chars));
1674  break;
1675  }
1676 
1677  if (Chars == MaxStandardSuffixLength)
1678  // Too long: can't be a standard suffix.
1679  break;
1680 
1681  Buffer[Chars++] = Next;
1682  Consumed += NextSize;
1683  }
1684  }
1685 
1686  if (!IsUDSuffix) {
1687  if (!isLexingRawMode())
1688  Diag(CurPtr, getLangOpts().MSVCCompat
1689  ? diag::ext_ms_reserved_user_defined_literal
1690  : diag::ext_reserved_user_defined_literal)
1692  return CurPtr;
1693  }
1694 
1695  CurPtr = ConsumeChar(CurPtr, Size, Result);
1696  }
1697 
1698  Result.setFlag(Token::HasUDSuffix);
1699  while (true) {
1700  C = getCharAndSize(CurPtr, Size);
1701  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1702  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1703  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1704  else break;
1705  }
1706 
1707  return CurPtr;
1708 }
1709 
1710 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1711 /// either " or L" or u8" or u" or U".
1712 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1713  tok::TokenKind Kind) {
1714  // Does this string contain the \0 character?
1715  const char *NulCharacter = nullptr;
1716 
1717  if (!isLexingRawMode() &&
1718  (Kind == tok::utf8_string_literal ||
1719  Kind == tok::utf16_string_literal ||
1720  Kind == tok::utf32_string_literal))
1721  Diag(BufferPtr, getLangOpts().CPlusPlus
1722  ? diag::warn_cxx98_compat_unicode_literal
1723  : diag::warn_c99_compat_unicode_literal);
1724 
1725  char C = getAndAdvanceChar(CurPtr, Result);
1726  while (C != '"') {
1727  // Skip escaped characters. Escaped newlines will already be processed by
1728  // getAndAdvanceChar.
1729  if (C == '\\')
1730  C = getAndAdvanceChar(CurPtr, Result);
1731 
1732  if (C == '\n' || C == '\r' || // Newline.
1733  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1734  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1735  Diag(BufferPtr, diag::ext_unterminated_string);
1736  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1737  return true;
1738  }
1739 
1740  if (C == 0) {
1741  if (isCodeCompletionPoint(CurPtr-1)) {
1743  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1744  cutOffLexing();
1745  return true;
1746  }
1747 
1748  NulCharacter = CurPtr-1;
1749  }
1750  C = getAndAdvanceChar(CurPtr, Result);
1751  }
1752 
1753  // If we are in C++11, lex the optional ud-suffix.
1754  if (getLangOpts().CPlusPlus)
1755  CurPtr = LexUDSuffix(Result, CurPtr, true);
1756 
1757  // If a nul character existed in the string, warn about it.
1758  if (NulCharacter && !isLexingRawMode())
1759  Diag(NulCharacter, diag::null_in_string);
1760 
1761  // Update the location of the token as well as the BufferPtr instance var.
1762  const char *TokStart = BufferPtr;
1763  FormTokenWithChars(Result, CurPtr, Kind);
1764  Result.setLiteralData(TokStart);
1765  return true;
1766 }
1767 
1768 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1769 /// having lexed R", LR", u8R", uR", or UR".
1770 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1771  tok::TokenKind Kind) {
1772  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1773  // Between the initial and final double quote characters of the raw string,
1774  // any transformations performed in phases 1 and 2 (trigraphs,
1775  // universal-character-names, and line splicing) are reverted.
1776 
1777  if (!isLexingRawMode())
1778  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1779 
1780  unsigned PrefixLen = 0;
1781 
1782  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1783  ++PrefixLen;
1784 
1785  // If the last character was not a '(', then we didn't lex a valid delimiter.
1786  if (CurPtr[PrefixLen] != '(') {
1787  if (!isLexingRawMode()) {
1788  const char *PrefixEnd = &CurPtr[PrefixLen];
1789  if (PrefixLen == 16) {
1790  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1791  } else {
1792  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1793  << StringRef(PrefixEnd, 1);
1794  }
1795  }
1796 
1797  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1798  // it's possible the '"' was intended to be part of the raw string, but
1799  // there's not much we can do about that.
1800  while (1) {
1801  char C = *CurPtr++;
1802 
1803  if (C == '"')
1804  break;
1805  if (C == 0 && CurPtr-1 == BufferEnd) {
1806  --CurPtr;
1807  break;
1808  }
1809  }
1810 
1811  FormTokenWithChars(Result, CurPtr, tok::unknown);
1812  return true;
1813  }
1814 
1815  // Save prefix and move CurPtr past it
1816  const char *Prefix = CurPtr;
1817  CurPtr += PrefixLen + 1; // skip over prefix and '('
1818 
1819  while (1) {
1820  char C = *CurPtr++;
1821 
1822  if (C == ')') {
1823  // Check for prefix match and closing quote.
1824  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1825  CurPtr += PrefixLen + 1; // skip over prefix and '"'
1826  break;
1827  }
1828  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1829  if (!isLexingRawMode())
1830  Diag(BufferPtr, diag::err_unterminated_raw_string)
1831  << StringRef(Prefix, PrefixLen);
1832  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1833  return true;
1834  }
1835  }
1836 
1837  // If we are in C++11, lex the optional ud-suffix.
1838  if (getLangOpts().CPlusPlus)
1839  CurPtr = LexUDSuffix(Result, CurPtr, true);
1840 
1841  // Update the location of token as well as BufferPtr.
1842  const char *TokStart = BufferPtr;
1843  FormTokenWithChars(Result, CurPtr, Kind);
1844  Result.setLiteralData(TokStart);
1845  return true;
1846 }
1847 
1848 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
1849 /// after having lexed the '<' character. This is used for #include filenames.
1850 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
1851  // Does this string contain the \0 character?
1852  const char *NulCharacter = nullptr;
1853  const char *AfterLessPos = CurPtr;
1854  char C = getAndAdvanceChar(CurPtr, Result);
1855  while (C != '>') {
1856  // Skip escaped characters.
1857  if (C == '\\' && CurPtr < BufferEnd) {
1858  // Skip the escaped character.
1859  getAndAdvanceChar(CurPtr, Result);
1860  } else if (C == '\n' || C == '\r' || // Newline.
1861  (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
1862  isCodeCompletionPoint(CurPtr-1)))) {
1863  // If the filename is unterminated, then it must just be a lone <
1864  // character. Return this as such.
1865  FormTokenWithChars(Result, AfterLessPos, tok::less);
1866  return true;
1867  } else if (C == 0) {
1868  NulCharacter = CurPtr-1;
1869  }
1870  C = getAndAdvanceChar(CurPtr, Result);
1871  }
1872 
1873  // If a nul character existed in the string, warn about it.
1874  if (NulCharacter && !isLexingRawMode())
1875  Diag(NulCharacter, diag::null_in_string);
1876 
1877  // Update the location of token as well as BufferPtr.
1878  const char *TokStart = BufferPtr;
1879  FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
1880  Result.setLiteralData(TokStart);
1881  return true;
1882 }
1883 
1884 
1885 /// LexCharConstant - Lex the remainder of a character constant, after having
1886 /// lexed either ' or L' or u8' or u' or U'.
1887 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
1888  tok::TokenKind Kind) {
1889  // Does this character contain the \0 character?
1890  const char *NulCharacter = nullptr;
1891 
1892  if (!isLexingRawMode()) {
1893  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
1894  Diag(BufferPtr, getLangOpts().CPlusPlus
1895  ? diag::warn_cxx98_compat_unicode_literal
1896  : diag::warn_c99_compat_unicode_literal);
1897  else if (Kind == tok::utf8_char_constant)
1898  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
1899  }
1900 
1901  char C = getAndAdvanceChar(CurPtr, Result);
1902  if (C == '\'') {
1903  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1904  Diag(BufferPtr, diag::ext_empty_character);
1905  FormTokenWithChars(Result, CurPtr, tok::unknown);
1906  return true;
1907  }
1908 
1909  while (C != '\'') {
1910  // Skip escaped characters.
1911  if (C == '\\')
1912  C = getAndAdvanceChar(CurPtr, Result);
1913 
1914  if (C == '\n' || C == '\r' || // Newline.
1915  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1916  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1917  Diag(BufferPtr, diag::ext_unterminated_char);
1918  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1919  return true;
1920  }
1921 
1922  if (C == 0) {
1923  if (isCodeCompletionPoint(CurPtr-1)) {
1925  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1926  cutOffLexing();
1927  return true;
1928  }
1929 
1930  NulCharacter = CurPtr-1;
1931  }
1932  C = getAndAdvanceChar(CurPtr, Result);
1933  }
1934 
1935  // If we are in C++11, lex the optional ud-suffix.
1936  if (getLangOpts().CPlusPlus)
1937  CurPtr = LexUDSuffix(Result, CurPtr, false);
1938 
1939  // If a nul character existed in the character, warn about it.
1940  if (NulCharacter && !isLexingRawMode())
1941  Diag(NulCharacter, diag::null_in_char);
1942 
1943  // Update the location of token as well as BufferPtr.
1944  const char *TokStart = BufferPtr;
1945  FormTokenWithChars(Result, CurPtr, Kind);
1946  Result.setLiteralData(TokStart);
1947  return true;
1948 }
1949 
1950 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
1951 /// Update BufferPtr to point to the next non-whitespace character and return.
1952 ///
1953 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
1954 ///
1955 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
1956  bool &TokAtPhysicalStartOfLine) {
1957  // Whitespace - Skip it, then return the token after the whitespace.
1958  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
1959 
1960  unsigned char Char = *CurPtr;
1961 
1962  // Skip consecutive spaces efficiently.
1963  while (1) {
1964  // Skip horizontal whitespace very aggressively.
1965  while (isHorizontalWhitespace(Char))
1966  Char = *++CurPtr;
1967 
1968  // Otherwise if we have something other than whitespace, we're done.
1969  if (!isVerticalWhitespace(Char))
1970  break;
1971 
1973  // End of preprocessor directive line, let LexTokenInternal handle this.
1974  BufferPtr = CurPtr;
1975  return false;
1976  }
1977 
1978  // OK, but handle newline.
1979  SawNewline = true;
1980  Char = *++CurPtr;
1981  }
1982 
1983  // If the client wants us to return whitespace, return it now.
1984  if (isKeepWhitespaceMode()) {
1985  FormTokenWithChars(Result, CurPtr, tok::unknown);
1986  if (SawNewline) {
1987  IsAtStartOfLine = true;
1988  IsAtPhysicalStartOfLine = true;
1989  }
1990  // FIXME: The next token will not have LeadingSpace set.
1991  return true;
1992  }
1993 
1994  // If this isn't immediately after a newline, there is leading space.
1995  char PrevChar = CurPtr[-1];
1996  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
1997 
1998  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
1999  if (SawNewline) {
2000  Result.setFlag(Token::StartOfLine);
2001  TokAtPhysicalStartOfLine = true;
2002  }
2003 
2004  BufferPtr = CurPtr;
2005  return false;
2006 }
2007 
2008 /// We have just read the // characters from input. Skip until we find the
2009 /// newline character thats terminate the comment. Then update BufferPtr and
2010 /// return.
2011 ///
2012 /// If we're in KeepCommentMode or any CommentHandler has inserted
2013 /// some tokens, this will store the first token and return true.
2014 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2015  bool &TokAtPhysicalStartOfLine) {
2016  // If Line comments aren't explicitly enabled for this language, emit an
2017  // extension warning.
2018  if (!LangOpts.LineComment && !isLexingRawMode()) {
2019  Diag(BufferPtr, diag::ext_line_comment);
2020 
2021  // Mark them enabled so we only emit one warning for this translation
2022  // unit.
2023  LangOpts.LineComment = true;
2024  }
2025 
2026  // Scan over the body of the comment. The common case, when scanning, is that
2027  // the comment contains normal ascii characters with nothing interesting in
2028  // them. As such, optimize for this case with the inner loop.
2029  char C;
2030  do {
2031  C = *CurPtr;
2032  // Skip over characters in the fast loop.
2033  while (C != 0 && // Potentially EOF.
2034  C != '\n' && C != '\r') // Newline or DOS-style newline.
2035  C = *++CurPtr;
2036 
2037  const char *NextLine = CurPtr;
2038  if (C != 0) {
2039  // We found a newline, see if it's escaped.
2040  const char *EscapePtr = CurPtr-1;
2041  bool HasSpace = false;
2042  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2043  --EscapePtr;
2044  HasSpace = true;
2045  }
2046 
2047  if (*EscapePtr == '\\') // Escaped newline.
2048  CurPtr = EscapePtr;
2049  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2050  EscapePtr[-2] == '?') // Trigraph-escaped newline.
2051  CurPtr = EscapePtr-2;
2052  else
2053  break; // This is a newline, we're done.
2054 
2055  // If there was space between the backslash and newline, warn about it.
2056  if (HasSpace && !isLexingRawMode())
2057  Diag(EscapePtr, diag::backslash_newline_space);
2058  }
2059 
2060  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2061  // properly decode the character. Read it in raw mode to avoid emitting
2062  // diagnostics about things like trigraphs. If we see an escaped newline,
2063  // we'll handle it below.
2064  const char *OldPtr = CurPtr;
2065  bool OldRawMode = isLexingRawMode();
2066  LexingRawMode = true;
2067  C = getAndAdvanceChar(CurPtr, Result);
2068  LexingRawMode = OldRawMode;
2069 
2070  // If we only read only one character, then no special handling is needed.
2071  // We're done and can skip forward to the newline.
2072  if (C != 0 && CurPtr == OldPtr+1) {
2073  CurPtr = NextLine;
2074  break;
2075  }
2076 
2077  // If we read multiple characters, and one of those characters was a \r or
2078  // \n, then we had an escaped newline within the comment. Emit diagnostic
2079  // unless the next line is also a // comment.
2080  if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') {
2081  for (; OldPtr != CurPtr; ++OldPtr)
2082  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2083  // Okay, we found a // comment that ends in a newline, if the next
2084  // line is also a // comment, but has spaces, don't emit a diagnostic.
2085  if (isWhitespace(C)) {
2086  const char *ForwardPtr = CurPtr;
2087  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2088  ++ForwardPtr;
2089  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2090  break;
2091  }
2092 
2093  if (!isLexingRawMode())
2094  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2095  break;
2096  }
2097  }
2098 
2099  if (CurPtr == BufferEnd+1) {
2100  --CurPtr;
2101  break;
2102  }
2103 
2104  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2106  cutOffLexing();
2107  return false;
2108  }
2109 
2110  } while (C != '\n' && C != '\r');
2111 
2112  // Found but did not consume the newline. Notify comment handlers about the
2113  // comment unless we're in a #if 0 block.
2114  if (PP && !isLexingRawMode() &&
2115  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2116  getSourceLocation(CurPtr)))) {
2117  BufferPtr = CurPtr;
2118  return true; // A token has to be returned.
2119  }
2120 
2121  // If we are returning comments as tokens, return this comment as a token.
2122  if (inKeepCommentMode())
2123  return SaveLineComment(Result, CurPtr);
2124 
2125  // If we are inside a preprocessor directive and we see the end of line,
2126  // return immediately, so that the lexer can return this as an EOD token.
2127  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2128  BufferPtr = CurPtr;
2129  return false;
2130  }
2131 
2132  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2133  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2134  // contribute to another token), it isn't needed for correctness. Note that
2135  // this is ok even in KeepWhitespaceMode, because we would have returned the
2136  /// comment above in that mode.
2137  ++CurPtr;
2138 
2139  // The next returned token is at the start of the line.
2140  Result.setFlag(Token::StartOfLine);
2141  TokAtPhysicalStartOfLine = true;
2142  // No leading whitespace seen so far.
2144  BufferPtr = CurPtr;
2145  return false;
2146 }
2147 
2148 /// If in save-comment mode, package up this Line comment in an appropriate
2149 /// way and return it.
2150 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2151  // If we're not in a preprocessor directive, just return the // comment
2152  // directly.
2153  FormTokenWithChars(Result, CurPtr, tok::comment);
2154 
2156  return true;
2157 
2158  // If this Line-style comment is in a macro definition, transmogrify it into
2159  // a C-style block comment.
2160  bool Invalid = false;
2161  std::string Spelling = PP->getSpelling(Result, &Invalid);
2162  if (Invalid)
2163  return true;
2164 
2165  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2166  Spelling[1] = '*'; // Change prefix to "/*".
2167  Spelling += "*/"; // add suffix.
2168 
2169  Result.setKind(tok::comment);
2170  PP->CreateString(Spelling, Result,
2171  Result.getLocation(), Result.getLocation());
2172  return true;
2173 }
2174 
2175 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2176 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2177 /// a diagnostic if so. We know that the newline is inside of a block comment.
2178 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2179  Lexer *L) {
2180  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2181 
2182  // Back up off the newline.
2183  --CurPtr;
2184 
2185  // If this is a two-character newline sequence, skip the other character.
2186  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2187  // \n\n or \r\r -> not escaped newline.
2188  if (CurPtr[0] == CurPtr[1])
2189  return false;
2190  // \n\r or \r\n -> skip the newline.
2191  --CurPtr;
2192  }
2193 
2194  // If we have horizontal whitespace, skip over it. We allow whitespace
2195  // between the slash and newline.
2196  bool HasSpace = false;
2197  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2198  --CurPtr;
2199  HasSpace = true;
2200  }
2201 
2202  // If we have a slash, we know this is an escaped newline.
2203  if (*CurPtr == '\\') {
2204  if (CurPtr[-1] != '*') return false;
2205  } else {
2206  // It isn't a slash, is it the ?? / trigraph?
2207  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2208  CurPtr[-3] != '*')
2209  return false;
2210 
2211  // This is the trigraph ending the comment. Emit a stern warning!
2212  CurPtr -= 2;
2213 
2214  // If no trigraphs are enabled, warn that we ignored this trigraph and
2215  // ignore this * character.
2216  if (!L->getLangOpts().Trigraphs) {
2217  if (!L->isLexingRawMode())
2218  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2219  return false;
2220  }
2221  if (!L->isLexingRawMode())
2222  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2223  }
2224 
2225  // Warn about having an escaped newline between the */ characters.
2226  if (!L->isLexingRawMode())
2227  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2228 
2229  // If there was space between the backslash and newline, warn about it.
2230  if (HasSpace && !L->isLexingRawMode())
2231  L->Diag(CurPtr, diag::backslash_newline_space);
2232 
2233  return true;
2234 }
2235 
2236 #ifdef __SSE2__
2237 #include <emmintrin.h>
2238 #elif __ALTIVEC__
2239 #include <altivec.h>
2240 #undef bool
2241 #endif
2242 
2243 /// We have just read from input the / and * characters that started a comment.
2244 /// Read until we find the * and / characters that terminate the comment.
2245 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2246 /// comments, because they cannot cause the comment to end. The only thing
2247 /// that can happen is the comment could end with an escaped newline between
2248 /// the terminating * and /.
2249 ///
2250 /// If we're in KeepCommentMode or any CommentHandler has inserted
2251 /// some tokens, this will store the first token and return true.
2252 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2253  bool &TokAtPhysicalStartOfLine) {
2254  // Scan one character past where we should, looking for a '/' character. Once
2255  // we find it, check to see if it was preceded by a *. This common
2256  // optimization helps people who like to put a lot of * characters in their
2257  // comments.
2258 
2259  // The first character we get with newlines and trigraphs skipped to handle
2260  // the degenerate /*/ case below correctly if the * has an escaped newline
2261  // after it.
2262  unsigned CharSize;
2263  unsigned char C = getCharAndSize(CurPtr, CharSize);
2264  CurPtr += CharSize;
2265  if (C == 0 && CurPtr == BufferEnd+1) {
2266  if (!isLexingRawMode())
2267  Diag(BufferPtr, diag::err_unterminated_block_comment);
2268  --CurPtr;
2269 
2270  // KeepWhitespaceMode should return this broken comment as a token. Since
2271  // it isn't a well formed comment, just return it as an 'unknown' token.
2272  if (isKeepWhitespaceMode()) {
2273  FormTokenWithChars(Result, CurPtr, tok::unknown);
2274  return true;
2275  }
2276 
2277  BufferPtr = CurPtr;
2278  return false;
2279  }
2280 
2281  // Check to see if the first character after the '/*' is another /. If so,
2282  // then this slash does not end the block comment, it is part of it.
2283  if (C == '/')
2284  C = *CurPtr++;
2285 
2286  while (1) {
2287  // Skip over all non-interesting characters until we find end of buffer or a
2288  // (probably ending) '/' character.
2289  if (CurPtr + 24 < BufferEnd &&
2290  // If there is a code-completion point avoid the fast scan because it
2291  // doesn't check for '\0'.
2292  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2293  // While not aligned to a 16-byte boundary.
2294  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2295  C = *CurPtr++;
2296 
2297  if (C == '/') goto FoundSlash;
2298 
2299 #ifdef __SSE2__
2300  __m128i Slashes = _mm_set1_epi8('/');
2301  while (CurPtr+16 <= BufferEnd) {
2302  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2303  Slashes));
2304  if (cmp != 0) {
2305  // Adjust the pointer to point directly after the first slash. It's
2306  // not necessary to set C here, it will be overwritten at the end of
2307  // the outer loop.
2308  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2309  goto FoundSlash;
2310  }
2311  CurPtr += 16;
2312  }
2313 #elif __ALTIVEC__
2314  __vector unsigned char Slashes = {
2315  '/', '/', '/', '/', '/', '/', '/', '/',
2316  '/', '/', '/', '/', '/', '/', '/', '/'
2317  };
2318  while (CurPtr+16 <= BufferEnd &&
2319  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2320  CurPtr += 16;
2321 #else
2322  // Scan for '/' quickly. Many block comments are very large.
2323  while (CurPtr[0] != '/' &&
2324  CurPtr[1] != '/' &&
2325  CurPtr[2] != '/' &&
2326  CurPtr[3] != '/' &&
2327  CurPtr+4 < BufferEnd) {
2328  CurPtr += 4;
2329  }
2330 #endif
2331 
2332  // It has to be one of the bytes scanned, increment to it and read one.
2333  C = *CurPtr++;
2334  }
2335 
2336  // Loop to scan the remainder.
2337  while (C != '/' && C != '\0')
2338  C = *CurPtr++;
2339 
2340  if (C == '/') {
2341  FoundSlash:
2342  if (CurPtr[-2] == '*') // We found the final */. We're done!
2343  break;
2344 
2345  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2346  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2347  // We found the final */, though it had an escaped newline between the
2348  // * and /. We're done!
2349  break;
2350  }
2351  }
2352  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2353  // If this is a /* inside of the comment, emit a warning. Don't do this
2354  // if this is a /*/, which will end the comment. This misses cases with
2355  // embedded escaped newlines, but oh well.
2356  if (!isLexingRawMode())
2357  Diag(CurPtr-1, diag::warn_nested_block_comment);
2358  }
2359  } else if (C == 0 && CurPtr == BufferEnd+1) {
2360  if (!isLexingRawMode())
2361  Diag(BufferPtr, diag::err_unterminated_block_comment);
2362  // Note: the user probably forgot a */. We could continue immediately
2363  // after the /*, but this would involve lexing a lot of what really is the
2364  // comment, which surely would confuse the parser.
2365  --CurPtr;
2366 
2367  // KeepWhitespaceMode should return this broken comment as a token. Since
2368  // it isn't a well formed comment, just return it as an 'unknown' token.
2369  if (isKeepWhitespaceMode()) {
2370  FormTokenWithChars(Result, CurPtr, tok::unknown);
2371  return true;
2372  }
2373 
2374  BufferPtr = CurPtr;
2375  return false;
2376  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2378  cutOffLexing();
2379  return false;
2380  }
2381 
2382  C = *CurPtr++;
2383  }
2384 
2385  // Notify comment handlers about the comment unless we're in a #if 0 block.
2386  if (PP && !isLexingRawMode() &&
2387  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2388  getSourceLocation(CurPtr)))) {
2389  BufferPtr = CurPtr;
2390  return true; // A token has to be returned.
2391  }
2392 
2393  // If we are returning comments as tokens, return this comment as a token.
2394  if (inKeepCommentMode()) {
2395  FormTokenWithChars(Result, CurPtr, tok::comment);
2396  return true;
2397  }
2398 
2399  // It is common for the tokens immediately after a /**/ comment to be
2400  // whitespace. Instead of going through the big switch, handle it
2401  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2402  // have already returned above with the comment as a token.
2403  if (isHorizontalWhitespace(*CurPtr)) {
2404  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2405  return false;
2406  }
2407 
2408  // Otherwise, just return so that the next character will be lexed as a token.
2409  BufferPtr = CurPtr;
2410  Result.setFlag(Token::LeadingSpace);
2411  return false;
2412 }
2413 
2414 //===----------------------------------------------------------------------===//
2415 // Primary Lexing Entry Points
2416 //===----------------------------------------------------------------------===//
2417 
2418 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2419 /// uninterpreted string. This switches the lexer out of directive mode.
2421  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2422  "Must be in a preprocessing directive!");
2423  Token Tmp;
2424 
2425  // CurPtr - Cache BufferPtr in an automatic variable.
2426  const char *CurPtr = BufferPtr;
2427  while (1) {
2428  char Char = getAndAdvanceChar(CurPtr, Tmp);
2429  switch (Char) {
2430  default:
2431  if (Result)
2432  Result->push_back(Char);
2433  break;
2434  case 0: // Null.
2435  // Found end of file?
2436  if (CurPtr-1 != BufferEnd) {
2437  if (isCodeCompletionPoint(CurPtr-1)) {
2439  cutOffLexing();
2440  return;
2441  }
2442 
2443  // Nope, normal character, continue.
2444  if (Result)
2445  Result->push_back(Char);
2446  break;
2447  }
2448  // FALL THROUGH.
2449  case '\r':
2450  case '\n':
2451  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2452  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2453  BufferPtr = CurPtr-1;
2454 
2455  // Next, lex the character, which should handle the EOD transition.
2456  Lex(Tmp);
2457  if (Tmp.is(tok::code_completion)) {
2458  if (PP)
2460  Lex(Tmp);
2461  }
2462  assert(Tmp.is(tok::eod) && "Unexpected token!");
2463 
2464  // Finally, we're done;
2465  return;
2466  }
2467  }
2468 }
2469 
2470 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2471 /// condition, reporting diagnostics and handling other edge cases as required.
2472 /// This returns true if Result contains a token, false if PP.Lex should be
2473 /// called again.
2474 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2475  // If we hit the end of the file while parsing a preprocessor directive,
2476  // end the preprocessor directive first. The next token returned will
2477  // then be the end of file.
2479  // Done parsing the "line".
2481  // Update the location of token as well as BufferPtr.
2482  FormTokenWithChars(Result, CurPtr, tok::eod);
2483 
2484  // Restore comment saving mode, in case it was disabled for directive.
2485  if (PP)
2487  return true; // Have a token.
2488  }
2489 
2490  // If we are in raw mode, return this event as an EOF token. Let the caller
2491  // that put us in raw mode handle the event.
2492  if (isLexingRawMode()) {
2493  Result.startToken();
2494  BufferPtr = BufferEnd;
2495  FormTokenWithChars(Result, BufferEnd, tok::eof);
2496  return true;
2497  }
2498 
2499  // Issue diagnostics for unterminated #if and missing newline.
2500 
2501  // If we are in a #if directive, emit an error.
2502  while (!ConditionalStack.empty()) {
2503  if (PP->getCodeCompletionFileLoc() != FileLoc)
2504  PP->Diag(ConditionalStack.back().IfLoc,
2505  diag::err_pp_unterminated_conditional);
2506  ConditionalStack.pop_back();
2507  }
2508 
2509  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2510  // a pedwarn.
2511  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2512  DiagnosticsEngine &Diags = PP->getDiagnostics();
2513  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2514  unsigned DiagID;
2515 
2516  if (LangOpts.CPlusPlus11) {
2517  // C++11 [lex.phases] 2.2 p2
2518  // Prefer the C++98 pedantic compatibility warning over the generic,
2519  // non-extension, user-requested "missing newline at EOF" warning.
2520  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2521  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2522  } else {
2523  DiagID = diag::warn_no_newline_eof;
2524  }
2525  } else {
2526  DiagID = diag::ext_no_newline_eof;
2527  }
2528 
2529  Diag(BufferEnd, DiagID)
2530  << FixItHint::CreateInsertion(EndLoc, "\n");
2531  }
2532 
2533  BufferPtr = CurPtr;
2534 
2535  // Finally, let the preprocessor handle this.
2536  return PP->HandleEndOfFile(Result, isPragmaLexer());
2537 }
2538 
2539 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2540 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2541 /// else and 2 if there are no more tokens in the buffer controlled by the
2542 /// lexer.
2543 unsigned Lexer::isNextPPTokenLParen() {
2544  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2545 
2546  // Switch to 'skipping' mode. This will ensure that we can lex a token
2547  // without emitting diagnostics, disables macro expansion, and will cause EOF
2548  // to return an EOF token instead of popping the include stack.
2549  LexingRawMode = true;
2550 
2551  // Save state that can be changed while lexing so that we can restore it.
2552  const char *TmpBufferPtr = BufferPtr;
2553  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2554  bool atStartOfLine = IsAtStartOfLine;
2555  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2556  bool leadingSpace = HasLeadingSpace;
2557 
2558  Token Tok;
2559  Lex(Tok);
2560 
2561  // Restore state that may have changed.
2562  BufferPtr = TmpBufferPtr;
2563  ParsingPreprocessorDirective = inPPDirectiveMode;
2564  HasLeadingSpace = leadingSpace;
2565  IsAtStartOfLine = atStartOfLine;
2566  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2567 
2568  // Restore the lexer back to non-skipping mode.
2569  LexingRawMode = false;
2570 
2571  if (Tok.is(tok::eof))
2572  return 2;
2573  return Tok.is(tok::l_paren);
2574 }
2575 
2576 /// \brief Find the end of a version control conflict marker.
2577 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2578  ConflictMarkerKind CMK) {
2579  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2580  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2581  StringRef RestOfBuffer(CurPtr+TermLen, BufferEnd-CurPtr-TermLen);
2582  size_t Pos = RestOfBuffer.find(Terminator);
2583  while (Pos != StringRef::npos) {
2584  // Must occur at start of line.
2585  if (Pos == 0 ||
2586  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2587  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2588  Pos = RestOfBuffer.find(Terminator);
2589  continue;
2590  }
2591  return RestOfBuffer.data()+Pos;
2592  }
2593  return nullptr;
2594 }
2595 
2596 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2597 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2598 /// and recover nicely. This returns true if it is a conflict marker and false
2599 /// if not.
2600 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2601  // Only a conflict marker if it starts at the beginning of a line.
2602  if (CurPtr != BufferStart &&
2603  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2604  return false;
2605 
2606  // Check to see if we have <<<<<<< or >>>>.
2607  if ((BufferEnd-CurPtr < 8 || StringRef(CurPtr, 7) != "<<<<<<<") &&
2608  (BufferEnd-CurPtr < 6 || StringRef(CurPtr, 5) != ">>>> "))
2609  return false;
2610 
2611  // If we have a situation where we don't care about conflict markers, ignore
2612  // it.
2613  if (CurrentConflictMarkerState || isLexingRawMode())
2614  return false;
2615 
2616  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2617 
2618  // Check to see if there is an ending marker somewhere in the buffer at the
2619  // start of a line to terminate this conflict marker.
2620  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2621  // We found a match. We are really in a conflict marker.
2622  // Diagnose this, and ignore to the end of line.
2623  Diag(CurPtr, diag::err_conflict_marker);
2624  CurrentConflictMarkerState = Kind;
2625 
2626  // Skip ahead to the end of line. We know this exists because the
2627  // end-of-conflict marker starts with \r or \n.
2628  while (*CurPtr != '\r' && *CurPtr != '\n') {
2629  assert(CurPtr != BufferEnd && "Didn't find end of line");
2630  ++CurPtr;
2631  }
2632  BufferPtr = CurPtr;
2633  return true;
2634  }
2635 
2636  // No end of conflict marker found.
2637  return false;
2638 }
2639 
2640 
2641 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2642 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2643 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2644 /// the line. This returns true if it is a conflict marker and false if not.
2645 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2646  // Only a conflict marker if it starts at the beginning of a line.
2647  if (CurPtr != BufferStart &&
2648  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2649  return false;
2650 
2651  // If we have a situation where we don't care about conflict markers, ignore
2652  // it.
2653  if (!CurrentConflictMarkerState || isLexingRawMode())
2654  return false;
2655 
2656  // Check to see if we have the marker (4 characters in a row).
2657  for (unsigned i = 1; i != 4; ++i)
2658  if (CurPtr[i] != CurPtr[0])
2659  return false;
2660 
2661  // If we do have it, search for the end of the conflict marker. This could
2662  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2663  // be the end of conflict marker.
2664  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2665  CurrentConflictMarkerState)) {
2666  CurPtr = End;
2667 
2668  // Skip ahead to the end of line.
2669  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2670  ++CurPtr;
2671 
2672  BufferPtr = CurPtr;
2673 
2674  // No longer in the conflict marker.
2675  CurrentConflictMarkerState = CMK_None;
2676  return true;
2677  }
2678 
2679  return false;
2680 }
2681 
2682 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2683  if (PP && PP->isCodeCompletionEnabled()) {
2684  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2685  return Loc == PP->getCodeCompletionLoc();
2686  }
2687 
2688  return false;
2689 }
2690 
2691 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2692  Token *Result) {
2693  unsigned CharSize;
2694  char Kind = getCharAndSize(StartPtr, CharSize);
2695 
2696  unsigned NumHexDigits;
2697  if (Kind == 'u')
2698  NumHexDigits = 4;
2699  else if (Kind == 'U')
2700  NumHexDigits = 8;
2701  else
2702  return 0;
2703 
2704  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2705  if (Result && !isLexingRawMode())
2706  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2707  return 0;
2708  }
2709 
2710  const char *CurPtr = StartPtr + CharSize;
2711  const char *KindLoc = &CurPtr[-1];
2712 
2713  uint32_t CodePoint = 0;
2714  for (unsigned i = 0; i < NumHexDigits; ++i) {
2715  char C = getCharAndSize(CurPtr, CharSize);
2716 
2717  unsigned Value = llvm::hexDigitValue(C);
2718  if (Value == -1U) {
2719  if (Result && !isLexingRawMode()) {
2720  if (i == 0) {
2721  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2722  << StringRef(KindLoc, 1);
2723  } else {
2724  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2725 
2726  // If the user wrote \U1234, suggest a fixit to \u.
2727  if (i == 4 && NumHexDigits == 8) {
2728  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2729  Diag(KindLoc, diag::note_ucn_four_not_eight)
2730  << FixItHint::CreateReplacement(URange, "u");
2731  }
2732  }
2733  }
2734 
2735  return 0;
2736  }
2737 
2738  CodePoint <<= 4;
2739  CodePoint += Value;
2740 
2741  CurPtr += CharSize;
2742  }
2743 
2744  if (Result) {
2745  Result->setFlag(Token::HasUCN);
2746  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
2747  StartPtr = CurPtr;
2748  else
2749  while (StartPtr != CurPtr)
2750  (void)getAndAdvanceChar(StartPtr, *Result);
2751  } else {
2752  StartPtr = CurPtr;
2753  }
2754 
2755  // Don't apply C family restrictions to UCNs in assembly mode
2756  if (LangOpts.AsmPreprocessor)
2757  return CodePoint;
2758 
2759  // C99 6.4.3p2: A universal character name shall not specify a character whose
2760  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
2761  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
2762  // C++11 [lex.charset]p2: If the hexadecimal value for a
2763  // universal-character-name corresponds to a surrogate code point (in the
2764  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
2765  // if the hexadecimal value for a universal-character-name outside the
2766  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
2767  // string literal corresponds to a control character (in either of the
2768  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
2769  // basic source character set, the program is ill-formed.
2770  if (CodePoint < 0xA0) {
2771  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
2772  return CodePoint;
2773 
2774  // We don't use isLexingRawMode() here because we need to warn about bad
2775  // UCNs even when skipping preprocessing tokens in a #if block.
2776  if (Result && PP) {
2777  if (CodePoint < 0x20 || CodePoint >= 0x7F)
2778  Diag(BufferPtr, diag::err_ucn_control_character);
2779  else {
2780  char C = static_cast<char>(CodePoint);
2781  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
2782  }
2783  }
2784 
2785  return 0;
2786 
2787  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
2788  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
2789  // We don't use isLexingRawMode() here because we need to diagnose bad
2790  // UCNs even when skipping preprocessing tokens in a #if block.
2791  if (Result && PP) {
2792  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
2793  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
2794  else
2795  Diag(BufferPtr, diag::err_ucn_escape_invalid);
2796  }
2797  return 0;
2798  }
2799 
2800  return CodePoint;
2801 }
2802 
2803 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
2804  const char *CurPtr) {
2805  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
2807  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
2808  UnicodeWhitespaceChars.contains(C)) {
2809  Diag(BufferPtr, diag::ext_unicode_whitespace)
2810  << makeCharRange(*this, BufferPtr, CurPtr);
2811 
2812  Result.setFlag(Token::LeadingSpace);
2813  return true;
2814  }
2815  return false;
2816 }
2817 
2818 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
2819  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
2821  !PP->isPreprocessedOutput()) {
2823  makeCharRange(*this, BufferPtr, CurPtr),
2824  /*IsFirst=*/true);
2825  }
2826 
2827  MIOpt.ReadToken();
2828  return LexIdentifier(Result, CurPtr);
2829  }
2830 
2832  !PP->isPreprocessedOutput() &&
2833  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
2834  // Non-ASCII characters tend to creep into source code unintentionally.
2835  // Instead of letting the parser complain about the unknown token,
2836  // just drop the character.
2837  // Note that we can /only/ do this when the non-ASCII character is actually
2838  // spelled as Unicode, not written as a UCN. The standard requires that
2839  // we not throw away any possible preprocessor tokens, but there's a
2840  // loophole in the mapping of Unicode characters to basic character set
2841  // characters that allows us to map these particular characters to, say,
2842  // whitespace.
2843  Diag(BufferPtr, diag::err_non_ascii)
2844  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
2845 
2846  BufferPtr = CurPtr;
2847  return false;
2848  }
2849 
2850  // Otherwise, we have an explicit UCN or a character that's unlikely to show
2851  // up by accident.
2852  MIOpt.ReadToken();
2853  FormTokenWithChars(Result, CurPtr, tok::unknown);
2854  return true;
2855 }
2856 
2857 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
2858  IsAtStartOfLine = Result.isAtStartOfLine();
2859  HasLeadingSpace = Result.hasLeadingSpace();
2860  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
2861  // Note that this doesn't affect IsAtPhysicalStartOfLine.
2862 }
2863 
2864 bool Lexer::Lex(Token &Result) {
2865  // Start a new token.
2866  Result.startToken();
2867 
2868  // Set up misc whitespace flags for LexTokenInternal.
2869  if (IsAtStartOfLine) {
2870  Result.setFlag(Token::StartOfLine);
2871  IsAtStartOfLine = false;
2872  }
2873 
2874  if (HasLeadingSpace) {
2875  Result.setFlag(Token::LeadingSpace);
2876  HasLeadingSpace = false;
2877  }
2878 
2879  if (HasLeadingEmptyMacro) {
2881  HasLeadingEmptyMacro = false;
2882  }
2883 
2884  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2885  IsAtPhysicalStartOfLine = false;
2886  bool isRawLex = isLexingRawMode();
2887  (void) isRawLex;
2888  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
2889  // (After the LexTokenInternal call, the lexer might be destroyed.)
2890  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
2891  return returnedToken;
2892 }
2893 
2894 /// LexTokenInternal - This implements a simple C family lexer. It is an
2895 /// extremely performance critical piece of code. This assumes that the buffer
2896 /// has a null character at the end of the file. This returns a preprocessing
2897 /// token, not a normal token, as such, it is an internal interface. It assumes
2898 /// that the Flags of result have been cleared before calling this.
2899 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
2900 LexNextToken:
2901  // New token, can't need cleaning yet.
2903  Result.setIdentifierInfo(nullptr);
2904 
2905  // CurPtr - Cache BufferPtr in an automatic variable.
2906  const char *CurPtr = BufferPtr;
2907 
2908  // Small amounts of horizontal whitespace is very common between tokens.
2909  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
2910  ++CurPtr;
2911  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
2912  ++CurPtr;
2913 
2914  // If we are keeping whitespace and other tokens, just return what we just
2915  // skipped. The next lexer invocation will return the token after the
2916  // whitespace.
2917  if (isKeepWhitespaceMode()) {
2918  FormTokenWithChars(Result, CurPtr, tok::unknown);
2919  // FIXME: The next token will not have LeadingSpace set.
2920  return true;
2921  }
2922 
2923  BufferPtr = CurPtr;
2924  Result.setFlag(Token::LeadingSpace);
2925  }
2926 
2927  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
2928 
2929  // Read a character, advancing over it.
2930  char Char = getAndAdvanceChar(CurPtr, Result);
2932 
2933  switch (Char) {
2934  case 0: // Null.
2935  // Found end of file?
2936  if (CurPtr-1 == BufferEnd)
2937  return LexEndOfFile(Result, CurPtr-1);
2938 
2939  // Check if we are performing code completion.
2940  if (isCodeCompletionPoint(CurPtr-1)) {
2941  // Return the code-completion token.
2942  Result.startToken();
2943  FormTokenWithChars(Result, CurPtr, tok::code_completion);
2944  return true;
2945  }
2946 
2947  if (!isLexingRawMode())
2948  Diag(CurPtr-1, diag::null_in_file);
2949  Result.setFlag(Token::LeadingSpace);
2950  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
2951  return true; // KeepWhitespaceMode
2952 
2953  // We know the lexer hasn't changed, so just try again with this lexer.
2954  // (We manually eliminate the tail call to avoid recursion.)
2955  goto LexNextToken;
2956 
2957  case 26: // DOS & CP/M EOF: "^Z".
2958  // If we're in Microsoft extensions mode, treat this as end of file.
2959  if (LangOpts.MicrosoftExt)
2960  return LexEndOfFile(Result, CurPtr-1);
2961 
2962  // If Microsoft extensions are disabled, this is just random garbage.
2963  Kind = tok::unknown;
2964  break;
2965 
2966  case '\n':
2967  case '\r':
2968  // If we are inside a preprocessor directive and we see the end of line,
2969  // we know we are done with the directive, so return an EOD token.
2971  // Done parsing the "line".
2973 
2974  // Restore comment saving mode, in case it was disabled for directive.
2975  if (PP)
2977 
2978  // Since we consumed a newline, we are back at the start of a line.
2979  IsAtStartOfLine = true;
2980  IsAtPhysicalStartOfLine = true;
2981 
2982  Kind = tok::eod;
2983  break;
2984  }
2985 
2986  // No leading whitespace seen so far.
2988 
2989  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
2990  return true; // KeepWhitespaceMode
2991 
2992  // We only saw whitespace, so just try again with this lexer.
2993  // (We manually eliminate the tail call to avoid recursion.)
2994  goto LexNextToken;
2995  case ' ':
2996  case '\t':
2997  case '\f':
2998  case '\v':
2999  SkipHorizontalWhitespace:
3000  Result.setFlag(Token::LeadingSpace);
3001  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3002  return true; // KeepWhitespaceMode
3003 
3004  SkipIgnoredUnits:
3005  CurPtr = BufferPtr;
3006 
3007  // If the next token is obviously a // or /* */ comment, skip it efficiently
3008  // too (without going through the big switch stmt).
3009  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3010  LangOpts.LineComment &&
3011  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3012  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3013  return true; // There is a token to return.
3014  goto SkipIgnoredUnits;
3015  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3016  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3017  return true; // There is a token to return.
3018  goto SkipIgnoredUnits;
3019  } else if (isHorizontalWhitespace(*CurPtr)) {
3020  goto SkipHorizontalWhitespace;
3021  }
3022  // We only saw whitespace, so just try again with this lexer.
3023  // (We manually eliminate the tail call to avoid recursion.)
3024  goto LexNextToken;
3025 
3026  // C99 6.4.4.1: Integer Constants.
3027  // C99 6.4.4.2: Floating Constants.
3028  case '0': case '1': case '2': case '3': case '4':
3029  case '5': case '6': case '7': case '8': case '9':
3030  // Notify MIOpt that we read a non-whitespace/non-comment token.
3031  MIOpt.ReadToken();
3032  return LexNumericConstant(Result, CurPtr);
3033 
3034  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3035  // Notify MIOpt that we read a non-whitespace/non-comment token.
3036  MIOpt.ReadToken();
3037 
3038  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3039  Char = getCharAndSize(CurPtr, SizeTmp);
3040 
3041  // UTF-16 string literal
3042  if (Char == '"')
3043  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3044  tok::utf16_string_literal);
3045 
3046  // UTF-16 character constant
3047  if (Char == '\'')
3048  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3049  tok::utf16_char_constant);
3050 
3051  // UTF-16 raw string literal
3052  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3053  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3054  return LexRawStringLiteral(Result,
3055  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3056  SizeTmp2, Result),
3057  tok::utf16_string_literal);
3058 
3059  if (Char == '8') {
3060  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3061 
3062  // UTF-8 string literal
3063  if (Char2 == '"')
3064  return LexStringLiteral(Result,
3065  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3066  SizeTmp2, Result),
3067  tok::utf8_string_literal);
3068  if (Char2 == '\'' && LangOpts.CPlusPlus1z)
3069  return LexCharConstant(
3070  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3071  SizeTmp2, Result),
3072  tok::utf8_char_constant);
3073 
3074  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3075  unsigned SizeTmp3;
3076  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3077  // UTF-8 raw string literal
3078  if (Char3 == '"') {
3079  return LexRawStringLiteral(Result,
3080  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3081  SizeTmp2, Result),
3082  SizeTmp3, Result),
3083  tok::utf8_string_literal);
3084  }
3085  }
3086  }
3087  }
3088 
3089  // treat u like the start of an identifier.
3090  return LexIdentifier(Result, CurPtr);
3091 
3092  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3093  // Notify MIOpt that we read a non-whitespace/non-comment token.
3094  MIOpt.ReadToken();
3095 
3096  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3097  Char = getCharAndSize(CurPtr, SizeTmp);
3098 
3099  // UTF-32 string literal
3100  if (Char == '"')
3101  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3102  tok::utf32_string_literal);
3103 
3104  // UTF-32 character constant
3105  if (Char == '\'')
3106  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3107  tok::utf32_char_constant);
3108 
3109  // UTF-32 raw string literal
3110  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3111  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3112  return LexRawStringLiteral(Result,
3113  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3114  SizeTmp2, Result),
3115  tok::utf32_string_literal);
3116  }
3117 
3118  // treat U like the start of an identifier.
3119  return LexIdentifier(Result, CurPtr);
3120 
3121  case 'R': // Identifier or C++0x raw string literal
3122  // Notify MIOpt that we read a non-whitespace/non-comment token.
3123  MIOpt.ReadToken();
3124 
3125  if (LangOpts.CPlusPlus11) {
3126  Char = getCharAndSize(CurPtr, SizeTmp);
3127 
3128  if (Char == '"')
3129  return LexRawStringLiteral(Result,
3130  ConsumeChar(CurPtr, SizeTmp, Result),
3131  tok::string_literal);
3132  }
3133 
3134  // treat R like the start of an identifier.
3135  return LexIdentifier(Result, CurPtr);
3136 
3137  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3138  // Notify MIOpt that we read a non-whitespace/non-comment token.
3139  MIOpt.ReadToken();
3140  Char = getCharAndSize(CurPtr, SizeTmp);
3141 
3142  // Wide string literal.
3143  if (Char == '"')
3144  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3145  tok::wide_string_literal);
3146 
3147  // Wide raw string literal.
3148  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3149  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3150  return LexRawStringLiteral(Result,
3151  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3152  SizeTmp2, Result),
3153  tok::wide_string_literal);
3154 
3155  // Wide character constant.
3156  if (Char == '\'')
3157  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3158  tok::wide_char_constant);
3159  // FALL THROUGH, treating L like the start of an identifier.
3160 
3161  // C99 6.4.2: Identifiers.
3162  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3163  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3164  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3165  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3166  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3167  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3168  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3169  case 'v': case 'w': case 'x': case 'y': case 'z':
3170  case '_':
3171  // Notify MIOpt that we read a non-whitespace/non-comment token.
3172  MIOpt.ReadToken();
3173  return LexIdentifier(Result, CurPtr);
3174 
3175  case '$': // $ in identifiers.
3176  if (LangOpts.DollarIdents) {
3177  if (!isLexingRawMode())
3178  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3179  // Notify MIOpt that we read a non-whitespace/non-comment token.
3180  MIOpt.ReadToken();
3181  return LexIdentifier(Result, CurPtr);
3182  }
3183 
3184  Kind = tok::unknown;
3185  break;
3186 
3187  // C99 6.4.4: Character Constants.
3188  case '\'':
3189  // Notify MIOpt that we read a non-whitespace/non-comment token.
3190  MIOpt.ReadToken();
3191  return LexCharConstant(Result, CurPtr, tok::char_constant);
3192 
3193  // C99 6.4.5: String Literals.
3194  case '"':
3195  // Notify MIOpt that we read a non-whitespace/non-comment token.
3196  MIOpt.ReadToken();
3197  return LexStringLiteral(Result, CurPtr, tok::string_literal);
3198 
3199  // C99 6.4.6: Punctuators.
3200  case '?':
3201  Kind = tok::question;
3202  break;
3203  case '[':
3204  Kind = tok::l_square;
3205  break;
3206  case ']':
3207  Kind = tok::r_square;
3208  break;
3209  case '(':
3210  Kind = tok::l_paren;
3211  break;
3212  case ')':
3213  Kind = tok::r_paren;
3214  break;
3215  case '{':
3216  Kind = tok::l_brace;
3217  break;
3218  case '}':
3219  Kind = tok::r_brace;
3220  break;
3221  case '.':
3222  Char = getCharAndSize(CurPtr, SizeTmp);
3223  if (Char >= '0' && Char <= '9') {
3224  // Notify MIOpt that we read a non-whitespace/non-comment token.
3225  MIOpt.ReadToken();
3226 
3227  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3228  } else if (LangOpts.CPlusPlus && Char == '*') {
3229  Kind = tok::periodstar;
3230  CurPtr += SizeTmp;
3231  } else if (Char == '.' &&
3232  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3233  Kind = tok::ellipsis;
3234  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3235  SizeTmp2, Result);
3236  } else {
3237  Kind = tok::period;
3238  }
3239  break;
3240  case '&':
3241  Char = getCharAndSize(CurPtr, SizeTmp);
3242  if (Char == '&') {
3243  Kind = tok::ampamp;
3244  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3245  } else if (Char == '=') {
3246  Kind = tok::ampequal;
3247  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3248  } else {
3249  Kind = tok::amp;
3250  }
3251  break;
3252  case '*':
3253  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3254  Kind = tok::starequal;
3255  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3256  } else {
3257  Kind = tok::star;
3258  }
3259  break;
3260  case '+':
3261  Char = getCharAndSize(CurPtr, SizeTmp);
3262  if (Char == '+') {
3263  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3264  Kind = tok::plusplus;
3265  } else if (Char == '=') {
3266  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3267  Kind = tok::plusequal;
3268  } else {
3269  Kind = tok::plus;
3270  }
3271  break;
3272  case '-':
3273  Char = getCharAndSize(CurPtr, SizeTmp);
3274  if (Char == '-') { // --
3275  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3276  Kind = tok::minusminus;
3277  } else if (Char == '>' && LangOpts.CPlusPlus &&
3278  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3279  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3280  SizeTmp2, Result);
3281  Kind = tok::arrowstar;
3282  } else if (Char == '>') { // ->
3283  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3284  Kind = tok::arrow;
3285  } else if (Char == '=') { // -=
3286  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3287  Kind = tok::minusequal;
3288  } else {
3289  Kind = tok::minus;
3290  }
3291  break;
3292  case '~':
3293  Kind = tok::tilde;
3294  break;
3295  case '!':
3296  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3297  Kind = tok::exclaimequal;
3298  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3299  } else {
3300  Kind = tok::exclaim;
3301  }
3302  break;
3303  case '/':
3304  // 6.4.9: Comments
3305  Char = getCharAndSize(CurPtr, SizeTmp);
3306  if (Char == '/') { // Line comment.
3307  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3308  // want to lex this as a comment. There is one problem with this though,
3309  // that in one particular corner case, this can change the behavior of the
3310  // resultant program. For example, In "foo //**/ bar", C89 would lex
3311  // this as "foo / bar" and langauges with Line comments would lex it as
3312  // "foo". Check to see if the character after the second slash is a '*'.
3313  // If so, we will lex that as a "/" instead of the start of a comment.
3314  // However, we never do this if we are just preprocessing.
3315  bool TreatAsComment = LangOpts.LineComment &&
3316  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3317  if (!TreatAsComment)
3318  if (!(PP && PP->isPreprocessedOutput()))
3319  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3320 
3321  if (TreatAsComment) {
3322  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3323  TokAtPhysicalStartOfLine))
3324  return true; // There is a token to return.
3325 
3326  // It is common for the tokens immediately after a // comment to be
3327  // whitespace (indentation for the next line). Instead of going through
3328  // the big switch, handle it efficiently now.
3329  goto SkipIgnoredUnits;
3330  }
3331  }
3332 
3333  if (Char == '*') { // /**/ comment.
3334  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3335  TokAtPhysicalStartOfLine))
3336  return true; // There is a token to return.
3337 
3338  // We only saw whitespace, so just try again with this lexer.
3339  // (We manually eliminate the tail call to avoid recursion.)
3340  goto LexNextToken;
3341  }
3342 
3343  if (Char == '=') {
3344  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3345  Kind = tok::slashequal;
3346  } else {
3347  Kind = tok::slash;
3348  }
3349  break;
3350  case '%':
3351  Char = getCharAndSize(CurPtr, SizeTmp);
3352  if (Char == '=') {
3353  Kind = tok::percentequal;
3354  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3355  } else if (LangOpts.Digraphs && Char == '>') {
3356  Kind = tok::r_brace; // '%>' -> '}'
3357  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3358  } else if (LangOpts.Digraphs && Char == ':') {
3359  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3360  Char = getCharAndSize(CurPtr, SizeTmp);
3361  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3362  Kind = tok::hashhash; // '%:%:' -> '##'
3363  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3364  SizeTmp2, Result);
3365  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3366  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3367  if (!isLexingRawMode())
3368  Diag(BufferPtr, diag::ext_charize_microsoft);
3369  Kind = tok::hashat;
3370  } else { // '%:' -> '#'
3371  // We parsed a # character. If this occurs at the start of the line,
3372  // it's actually the start of a preprocessing directive. Callback to
3373  // the preprocessor to handle it.
3374  // TODO: -fpreprocessed mode??
3375  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3376  goto HandleDirective;
3377 
3378  Kind = tok::hash;
3379  }
3380  } else {
3381  Kind = tok::percent;
3382  }
3383  break;
3384  case '<':
3385  Char = getCharAndSize(CurPtr, SizeTmp);
3386  if (ParsingFilename) {
3387  return LexAngledStringLiteral(Result, CurPtr);
3388  } else if (Char == '<') {
3389  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3390  if (After == '=') {
3391  Kind = tok::lesslessequal;
3392  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3393  SizeTmp2, Result);
3394  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3395  // If this is actually a '<<<<<<<' version control conflict marker,
3396  // recognize it as such and recover nicely.
3397  goto LexNextToken;
3398  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3399  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3400  // ignore it.
3401  goto LexNextToken;
3402  } else if (LangOpts.CUDA && After == '<') {
3403  Kind = tok::lesslessless;
3404  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3405  SizeTmp2, Result);
3406  } else {
3407  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3408  Kind = tok::lessless;
3409  }
3410  } else if (Char == '=') {
3411  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3412  Kind = tok::lessequal;
3413  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3414  if (LangOpts.CPlusPlus11 &&
3415  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3416  // C++0x [lex.pptoken]p3:
3417  // Otherwise, if the next three characters are <:: and the subsequent
3418  // character is neither : nor >, the < is treated as a preprocessor
3419  // token by itself and not as the first character of the alternative
3420  // token <:.
3421  unsigned SizeTmp3;
3422  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3423  if (After != ':' && After != '>') {
3424  Kind = tok::less;
3425  if (!isLexingRawMode())
3426  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3427  break;
3428  }
3429  }
3430 
3431  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3432  Kind = tok::l_square;
3433  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3434  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3435  Kind = tok::l_brace;
3436  } else {
3437  Kind = tok::less;
3438  }
3439  break;
3440  case '>':
3441  Char = getCharAndSize(CurPtr, SizeTmp);
3442  if (Char == '=') {
3443  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3444  Kind = tok::greaterequal;
3445  } else if (Char == '>') {
3446  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3447  if (After == '=') {
3448  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3449  SizeTmp2, Result);
3450  Kind = tok::greatergreaterequal;
3451  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3452  // If this is actually a '>>>>' conflict marker, recognize it as such
3453  // and recover nicely.
3454  goto LexNextToken;
3455  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3456  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3457  goto LexNextToken;
3458  } else if (LangOpts.CUDA && After == '>') {
3459  Kind = tok::greatergreatergreater;
3460  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3461  SizeTmp2, Result);
3462  } else {
3463  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3464  Kind = tok::greatergreater;
3465  }
3466 
3467  } else {
3468  Kind = tok::greater;
3469  }
3470  break;
3471  case '^':
3472  Char = getCharAndSize(CurPtr, SizeTmp);
3473  if (Char == '=') {
3474  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3475  Kind = tok::caretequal;
3476  } else {
3477  Kind = tok::caret;
3478  }
3479  break;
3480  case '|':
3481  Char = getCharAndSize(CurPtr, SizeTmp);
3482  if (Char == '=') {
3483  Kind = tok::pipeequal;
3484  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3485  } else if (Char == '|') {
3486  // If this is '|||||||' and we're in a conflict marker, ignore it.
3487  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3488  goto LexNextToken;
3489  Kind = tok::pipepipe;
3490  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3491  } else {
3492  Kind = tok::pipe;
3493  }
3494  break;
3495  case ':':
3496  Char = getCharAndSize(CurPtr, SizeTmp);
3497  if (LangOpts.Digraphs && Char == '>') {
3498  Kind = tok::r_square; // ':>' -> ']'
3499  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3500  } else if (LangOpts.CPlusPlus && Char == ':') {
3501  Kind = tok::coloncolon;
3502  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3503  } else {
3504  Kind = tok::colon;
3505  }
3506  break;
3507  case ';':
3508  Kind = tok::semi;
3509  break;
3510  case '=':
3511  Char = getCharAndSize(CurPtr, SizeTmp);
3512  if (Char == '=') {
3513  // If this is '====' and we're in a conflict marker, ignore it.
3514  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3515  goto LexNextToken;
3516 
3517  Kind = tok::equalequal;
3518  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3519  } else {
3520  Kind = tok::equal;
3521  }
3522  break;
3523  case ',':
3524  Kind = tok::comma;
3525  break;
3526  case '#':
3527  Char = getCharAndSize(CurPtr, SizeTmp);
3528  if (Char == '#') {
3529  Kind = tok::hashhash;
3530  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3531  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3532  Kind = tok::hashat;
3533  if (!isLexingRawMode())
3534  Diag(BufferPtr, diag::ext_charize_microsoft);
3535  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3536  } else {
3537  // We parsed a # character. If this occurs at the start of the line,
3538  // it's actually the start of a preprocessing directive. Callback to
3539  // the preprocessor to handle it.
3540  // TODO: -fpreprocessed mode??
3541  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3542  goto HandleDirective;
3543 
3544  Kind = tok::hash;
3545  }
3546  break;
3547 
3548  case '@':
3549  // Objective C support.
3550  if (CurPtr[-1] == '@' && LangOpts.ObjC1)
3551  Kind = tok::at;
3552  else
3553  Kind = tok::unknown;
3554  break;
3555 
3556  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3557  case '\\':
3558  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3559  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3560  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3561  return true; // KeepWhitespaceMode
3562 
3563  // We only saw whitespace, so just try again with this lexer.
3564  // (We manually eliminate the tail call to avoid recursion.)
3565  goto LexNextToken;
3566  }
3567 
3568  return LexUnicode(Result, CodePoint, CurPtr);
3569  }
3570 
3571  Kind = tok::unknown;
3572  break;
3573 
3574  default: {
3575  if (isASCII(Char)) {
3576  Kind = tok::unknown;
3577  break;
3578  }
3579 
3580  UTF32 CodePoint;
3581 
3582  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3583  // an escaped newline.
3584  --CurPtr;
3585  ConversionResult Status =
3586  llvm::convertUTF8Sequence((const UTF8 **)&CurPtr,
3587  (const UTF8 *)BufferEnd,
3588  &CodePoint,
3589  strictConversion);
3590  if (Status == conversionOK) {
3591  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3592  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3593  return true; // KeepWhitespaceMode
3594 
3595  // We only saw whitespace, so just try again with this lexer.
3596  // (We manually eliminate the tail call to avoid recursion.)
3597  goto LexNextToken;
3598  }
3599  return LexUnicode(Result, CodePoint, CurPtr);
3600  }
3601 
3603  PP->isPreprocessedOutput()) {
3604  ++CurPtr;
3605  Kind = tok::unknown;
3606  break;
3607  }
3608 
3609  // Non-ASCII characters tend to creep into source code unintentionally.
3610  // Instead of letting the parser complain about the unknown token,
3611  // just diagnose the invalid UTF-8, then drop the character.
3612  Diag(CurPtr, diag::err_invalid_utf8);
3613 
3614  BufferPtr = CurPtr+1;
3615  // We're pretending the character didn't exist, so just try again with
3616  // this lexer.
3617  // (We manually eliminate the tail call to avoid recursion.)
3618  goto LexNextToken;
3619  }
3620  }
3621 
3622  // Notify MIOpt that we read a non-whitespace/non-comment token.
3623  MIOpt.ReadToken();
3624 
3625  // Update the location of token as well as BufferPtr.
3626  FormTokenWithChars(Result, CurPtr, Kind);
3627  return true;
3628 
3629 HandleDirective:
3630  // We parsed a # character and it's the start of a preprocessing directive.
3631 
3632  FormTokenWithChars(Result, CurPtr, tok::hash);
3633  PP->HandleDirective(Result);
3634 
3636  // With a fatal failure in the module loader, we abort parsing.
3637  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3638  return true;
3639  }
3640 
3641  // We parsed the directive; lex a token with the new state.
3642  return false;
3643 }
bool isAtStartOfLine() const
Definition: Token.h:261
SourceManager & getSourceManager() const
Definition: Preprocessor.h:682
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:43
static unsigned getSpelling(const Token &Tok, const char *&Buffer, const SourceManager &SourceMgr, const LangOptions &LangOpts, bool *Invalid=nullptr)
Definition: Lexer.cpp:358
This is a discriminated union of FileInfo and ExpansionInfo.
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:252
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)
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. ...
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:1356
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:265
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded...
const ExpansionInfo & getExpansion() const
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1118
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:234
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:280
static char getCharAndSizeNoWarn(const char *Ptr, unsigned &Size, const LangOptions &LangOpts)
Definition: Lexer.h:448
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:781
static LLVM_READNONE bool isASCII(char c)
Returns true if this is an ASCII character.
Definition: CharInfo.h:43
ConflictMarkerKind
Definition: Lexer.h:31
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
Definition: Lexer.cpp:1018
Like System, but searched after the system directories.
static Lexer * Create_PragmaLexer(SourceLocation SpellingLoc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLen, Preprocessor &PP)
Definition: Lexer.cpp:164
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
static LLVM_READONLY bool isPreprocessingNumberBody(unsigned char c)
Definition: CharInfo.h:148
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:207
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)
Definition: CharInfo.h:71
bool isFileID() const
SmallVector< PPConditionalInfo, 4 > ConditionalStack
Information about the set of #if/#ifdef/#ifndef blocks we are currently in.
void setKind(tok::TokenKind K)
Definition: Token.h:91
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()
Definition: Lexer.cpp:120
SourceLocation getSourceLocation() override
Definition: Lexer.h:221
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:43
bool getCommentRetentionState() const
Definition: Preprocessor.h:721
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)
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:706
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:90
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)
Definition: Lexer.cpp:700
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L)
Definition: Lexer.cpp:2178
bool isInvalid() const
AnnotatingParser & P
StringRef getRawIdentifier() const
Definition: Token.h:203
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1387
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:866
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
SourceManager & SM
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:920
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:803
static int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:12380
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:759
Represents a character-granular source range.
SourceLocation getEnd() const
static unsigned MeasureTokenLength(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
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:1156
SourceManager & SourceMgr
Definition: Format.cpp:1205
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:124
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.
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
Definition: Token.h:215
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. Return false if the end o...
SmallVectorImpl< AnnotatedLine * >::const_iterator Next
Encodes a location in the source. The SourceManager can decode this to get at the full include stack...
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:1393
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:186
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Definition: Lexer.cpp:1062
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:829
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
bool is(tok::TokenKind K) const
Definition: Token.h:95
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:645
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:676
static StringRef getImmediateMacroName(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:956
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
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:856
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling. Extract the characters comprising the spelling of this token from the provi...
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
Definition: Lexer.cpp:1043
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
SourceLocation getExpansionLocStart() const
void setLiteralData(const char *Ptr)
Definition: Token.h:219
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:113
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:1372
bool hasLeadingEmptyMacro() const
Return true if this token has an empty macro before it.
Definition: Token.h:284
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)
Definition: Lexer.cpp:1072
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1003
__PTRDIFF_TYPE__ ptrdiff_t
Definition: stddef.h:51
static LLVM_READONLY bool isIdentifierBody(unsigned char c, bool AllowDollar=false)
Definition: CharInfo.h:59
bool HandleComment(Token &Token, SourceRange Comment)
const LangOptions & getLangOpts() const
Definition: Lexer.h:128
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
Definition: Lexer.cpp:2420
Not within a conflict marker.
Definition: Lexer.h:33
static LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
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)
Definition: Lexer.cpp:1091
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2577
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)
Definition: Lexer.h:187
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
unsigned getLength() const
Definition: Token.h:127
SourceLocation getLocForStartOfFile(FileID FID) const
Return the source location corresponding to the first byte of the specified file. ...
bool isKeepWhitespaceMode() const
Definition: Lexer.h:166
bool isPreprocessedOutput() const
Definition: Preprocessor.h:742
static LLVM_READONLY bool isIdentifierHead(unsigned char c, bool AllowDollar=false)
Definition: CharInfo.h:49
A trivial tuple used to represent a source range.
void clearFlag(TokenFlags Flag)
Unset the specified flag.
Definition: Token.h:239
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)
Definition: Lexer.h:172
This class handles loading and caching of source files into memory.
void startToken()
Reset all flags to cleared.
Definition: Token.h:169
static std::string Stringify(StringRef Str, bool Charify=false)
Definition: Lexer.cpp:202
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:96
SourceLocation getSpellingLoc() const
bool isMacroArgExpansion(SourceLocation Loc) const
Tests whether the given source location represents a macro argument's expansion into the function-lik...
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:177