clang  5.0.0
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"
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/MathExtras.h"
28 #include "llvm/Support/MemoryBuffer.h"
29 #include "llvm/Support/UnicodeCharRanges.h"
30 #include <algorithm>
31 #include <cassert>
32 #include <cstddef>
33 #include <cstdint>
34 #include <cstring>
35 #include <string>
36 #include <tuple>
37 #include <utility>
38 
39 using namespace clang;
40 
41 //===----------------------------------------------------------------------===//
42 // Token Class Implementation
43 //===----------------------------------------------------------------------===//
44 
45 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
47  if (isAnnotation())
48  return false;
50  return II->getObjCKeywordID() == objcKey;
51  return false;
52 }
53 
54 /// getObjCKeywordID - Return the ObjC keyword kind.
56  if (isAnnotation())
57  return tok::objc_not_keyword;
59  return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
60 }
61 
62 //===----------------------------------------------------------------------===//
63 // Lexer Class Implementation
64 //===----------------------------------------------------------------------===//
65 
66 void Lexer::anchor() { }
67 
68 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
69  const char *BufEnd) {
70  BufferStart = BufStart;
71  BufferPtr = BufPtr;
72  BufferEnd = BufEnd;
73 
74  assert(BufEnd[0] == 0 &&
75  "We assume that the input buffer has a null character at the end"
76  " to simplify lexing!");
77 
78  // Check whether we have a BOM in the beginning of the buffer. If yes - act
79  // accordingly. Right now we support only UTF-8 with and without BOM, so, just
80  // skip the UTF-8 BOM if it's present.
81  if (BufferStart == BufferPtr) {
82  // Determine the size of the BOM.
83  StringRef Buf(BufferStart, BufferEnd - BufferStart);
84  size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
85  .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
86  .Default(0);
87 
88  // Skip the BOM.
89  BufferPtr += BOMLength;
90  }
91 
92  Is_PragmaLexer = false;
93  CurrentConflictMarkerState = CMK_None;
94 
95  // Start of the file is a start of line.
96  IsAtStartOfLine = true;
97  IsAtPhysicalStartOfLine = true;
98 
99  HasLeadingSpace = false;
100  HasLeadingEmptyMacro = false;
101 
102  // We are not after parsing a #.
104 
105  // We are not after parsing #include.
106  ParsingFilename = false;
107 
108  // We are not in raw mode. Raw mode disables diagnostics and interpretation
109  // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
110  // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
111  // or otherwise skipping over tokens.
112  LexingRawMode = false;
113 
114  // Default to not keeping comments.
115  ExtendedTokenMode = 0;
116 }
117 
118 /// Lexer constructor - Create a new lexer object for the specified buffer
119 /// with the specified preprocessor managing the lexing process. This lexer
120 /// assumes that the associated file buffer and Preprocessor objects will
121 /// outlive it, so it doesn't take ownership of either of them.
122 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
123  : PreprocessorLexer(&PP, FID),
124  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
125  LangOpts(PP.getLangOpts()) {
126 
127  InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
128  InputFile->getBufferEnd());
129 
131 }
132 
134  assert(PP && "Cannot reset token mode without a preprocessor");
135  if (LangOpts.TraditionalCPP)
136  SetKeepWhitespaceMode(true);
137  else
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(SourceLocation fileloc, const LangOptions &langOpts,
145  const char *BufStart, const char *BufPtr, const char *BufEnd)
146  : FileLoc(fileloc), LangOpts(langOpts) {
147 
148  InitLexer(BufStart, BufPtr, BufEnd);
149 
150  // We *are* in raw mode.
151  LexingRawMode = true;
152 }
153 
154 /// Lexer constructor - Create a new raw lexer object. This object is only
155 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
156 /// range will outlive it, so it doesn't take ownership of it.
157 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
158  const SourceManager &SM, const LangOptions &langOpts)
159  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile->getBufferStart(),
160  FromFile->getBufferStart(), FromFile->getBufferEnd()) {}
161 
162 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
163 /// _Pragma expansion. This has a variety of magic semantics that this method
164 /// sets up. It returns a new'd Lexer that must be delete'd when done.
165 ///
166 /// On entrance to this routine, TokStartLoc is a macro location which has a
167 /// spelling loc that indicates the bytes to be lexed for the token and an
168 /// expansion location that indicates where all lexed tokens should be
169 /// "expanded from".
170 ///
171 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
172 /// normal lexer that remaps tokens as they fly by. This would require making
173 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
174 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
175 /// out of the critical path of the lexer!
176 ///
178  SourceLocation ExpansionLocStart,
179  SourceLocation ExpansionLocEnd,
180  unsigned TokLen, Preprocessor &PP) {
182 
183  // Create the lexer as if we were going to lex the file normally.
184  FileID SpellingFID = SM.getFileID(SpellingLoc);
185  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
186  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
187 
188  // Now that the lexer is created, change the start/end locations so that we
189  // just lex the subsection of the file that we want. This is lexing from a
190  // scratch buffer.
191  const char *StrData = SM.getCharacterData(SpellingLoc);
192 
193  L->BufferPtr = StrData;
194  L->BufferEnd = StrData+TokLen;
195  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
196 
197  // Set the SourceLocation with the remapping information. This ensures that
198  // GetMappedTokenLoc will remap the tokens as they are lexed.
199  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
200  ExpansionLocStart,
201  ExpansionLocEnd, TokLen);
202 
203  // Ensure that the lexer thinks it is inside a directive, so that end \n will
204  // return an EOD token.
206 
207  // This lexer really is for _Pragma.
208  L->Is_PragmaLexer = true;
209  return L;
210 }
211 
212 /// Stringify - Convert the specified string into a C string, with surrounding
213 /// ""'s, and with escaped \ and " characters.
214 std::string Lexer::Stringify(StringRef Str, bool Charify) {
215  std::string Result = Str;
216  char Quote = Charify ? '\'' : '"';
217  for (unsigned i = 0, e = Result.size(); i != e; ++i) {
218  if (Result[i] == '\\' || Result[i] == Quote) {
219  Result.insert(Result.begin()+i, '\\');
220  ++i; ++e;
221  }
222  }
223  return Result;
224 }
225 
226 /// Stringify - Convert the specified string into a C string by escaping '\'
227 /// and " characters. This does not add surrounding ""'s to the string.
229  for (unsigned i = 0, e = Str.size(); i != e; ++i) {
230  if (Str[i] == '\\' || Str[i] == '"') {
231  Str.insert(Str.begin()+i, '\\');
232  ++i; ++e;
233  }
234  }
235 }
236 
237 //===----------------------------------------------------------------------===//
238 // Token Spelling
239 //===----------------------------------------------------------------------===//
240 
241 /// \brief Slow case of getSpelling. Extract the characters comprising the
242 /// spelling of this token from the provided input buffer.
243 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
244  const LangOptions &LangOpts, char *Spelling) {
245  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
246 
247  size_t Length = 0;
248  const char *BufEnd = BufPtr + Tok.getLength();
249 
250  if (tok::isStringLiteral(Tok.getKind())) {
251  // Munch the encoding-prefix and opening double-quote.
252  while (BufPtr < BufEnd) {
253  unsigned Size;
254  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
255  BufPtr += Size;
256 
257  if (Spelling[Length - 1] == '"')
258  break;
259  }
260 
261  // Raw string literals need special handling; trigraph expansion and line
262  // splicing do not occur within their d-char-sequence nor within their
263  // r-char-sequence.
264  if (Length >= 2 &&
265  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
266  // Search backwards from the end of the token to find the matching closing
267  // quote.
268  const char *RawEnd = BufEnd;
269  do --RawEnd; while (*RawEnd != '"');
270  size_t RawLength = RawEnd - BufPtr + 1;
271 
272  // Everything between the quotes is included verbatim in the spelling.
273  memcpy(Spelling + Length, BufPtr, RawLength);
274  Length += RawLength;
275  BufPtr += RawLength;
276 
277  // The rest of the token is lexed normally.
278  }
279  }
280 
281  while (BufPtr < BufEnd) {
282  unsigned Size;
283  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
284  BufPtr += Size;
285  }
286 
287  assert(Length < Tok.getLength() &&
288  "NeedsCleaning flag set on token that didn't need cleaning!");
289  return Length;
290 }
291 
292 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
293 /// token are the characters used to represent the token in the source file
294 /// after trigraph expansion and escaped-newline folding. In particular, this
295 /// wants to get the true, uncanonicalized, spelling of things like digraphs
296 /// UCNs, etc.
298  SmallVectorImpl<char> &buffer,
299  const SourceManager &SM,
300  const LangOptions &options,
301  bool *invalid) {
302  // Break down the source location.
303  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
304 
305  // Try to the load the file buffer.
306  bool invalidTemp = false;
307  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
308  if (invalidTemp) {
309  if (invalid) *invalid = true;
310  return StringRef();
311  }
312 
313  const char *tokenBegin = file.data() + locInfo.second;
314 
315  // Lex from the start of the given location.
316  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
317  file.begin(), tokenBegin, file.end());
318  Token token;
319  lexer.LexFromRawLexer(token);
320 
321  unsigned length = token.getLength();
322 
323  // Common case: no need for cleaning.
324  if (!token.needsCleaning())
325  return StringRef(tokenBegin, length);
326 
327  // Hard case, we need to relex the characters into the string.
328  buffer.resize(length);
329  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
330  return StringRef(buffer.data(), buffer.size());
331 }
332 
333 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
334 /// token are the characters used to represent the token in the source file
335 /// after trigraph expansion and escaped-newline folding. In particular, this
336 /// wants to get the true, uncanonicalized, spelling of things like digraphs
337 /// UCNs, etc.
338 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
339  const LangOptions &LangOpts, bool *Invalid) {
340  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
341 
342  bool CharDataInvalid = false;
343  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
344  &CharDataInvalid);
345  if (Invalid)
346  *Invalid = CharDataInvalid;
347  if (CharDataInvalid)
348  return std::string();
349 
350  // If this token contains nothing interesting, return it directly.
351  if (!Tok.needsCleaning())
352  return std::string(TokStart, TokStart + Tok.getLength());
353 
354  std::string Result;
355  Result.resize(Tok.getLength());
356  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
357  return Result;
358 }
359 
360 /// getSpelling - This method is used to get the spelling of a token into a
361 /// preallocated buffer, instead of as an std::string. The caller is required
362 /// to allocate enough space for the token, which is guaranteed to be at least
363 /// Tok.getLength() bytes long. The actual length of the token is returned.
364 ///
365 /// Note that this method may do two possible things: it may either fill in
366 /// the buffer specified with characters, or it may *change the input pointer*
367 /// to point to a constant buffer with the data already in it (avoiding a
368 /// copy). The caller is not allowed to modify the returned buffer pointer
369 /// if an internal buffer is returned.
370 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
371  const SourceManager &SourceMgr,
372  const LangOptions &LangOpts, bool *Invalid) {
373  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
374 
375  const char *TokStart = nullptr;
376  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
377  if (Tok.is(tok::raw_identifier))
378  TokStart = Tok.getRawIdentifier().data();
379  else if (!Tok.hasUCN()) {
380  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
381  // Just return the string from the identifier table, which is very quick.
382  Buffer = II->getNameStart();
383  return II->getLength();
384  }
385  }
386 
387  // NOTE: this can be checked even after testing for an IdentifierInfo.
388  if (Tok.isLiteral())
389  TokStart = Tok.getLiteralData();
390 
391  if (!TokStart) {
392  // Compute the start of the token in the input lexer buffer.
393  bool CharDataInvalid = false;
394  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
395  if (Invalid)
396  *Invalid = CharDataInvalid;
397  if (CharDataInvalid) {
398  Buffer = "";
399  return 0;
400  }
401  }
402 
403  // If this token contains nothing interesting, return it directly.
404  if (!Tok.needsCleaning()) {
405  Buffer = TokStart;
406  return Tok.getLength();
407  }
408 
409  // Otherwise, hard case, relex the characters into the string.
410  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
411 }
412 
413 /// MeasureTokenLength - Relex the token at the specified location and return
414 /// its length in bytes in the input file. If the token needs cleaning (e.g.
415 /// includes a trigraph or an escaped newline) then this count includes bytes
416 /// that are part of that.
418  const SourceManager &SM,
419  const LangOptions &LangOpts) {
420  Token TheTok;
421  if (getRawToken(Loc, TheTok, SM, LangOpts))
422  return 0;
423  return TheTok.getLength();
424 }
425 
426 /// \brief Relex the token at the specified location.
427 /// \returns true if there was a failure, false on success.
429  const SourceManager &SM,
430  const LangOptions &LangOpts,
431  bool IgnoreWhiteSpace) {
432  // TODO: this could be special cased for common tokens like identifiers, ')',
433  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
434  // all obviously single-char tokens. This could use
435  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
436  // something.
437 
438  // If this comes from a macro expansion, we really do want the macro name, not
439  // the token this macro expanded to.
440  Loc = SM.getExpansionLoc(Loc);
441  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
442  bool Invalid = false;
443  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
444  if (Invalid)
445  return true;
446 
447  const char *StrData = Buffer.data()+LocInfo.second;
448 
449  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
450  return true;
451 
452  // Create a lexer starting at the beginning of this token.
453  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
454  Buffer.begin(), StrData, Buffer.end());
455  TheLexer.SetCommentRetentionState(true);
456  TheLexer.LexFromRawLexer(Result);
457  return false;
458 }
459 
460 /// Returns the pointer that points to the beginning of line that contains
461 /// the given offset, or null if the offset if invalid.
462 static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
463  const char *BufStart = Buffer.data();
464  if (Offset >= Buffer.size())
465  return nullptr;
466  const char *StrData = BufStart + Offset;
467 
468  if (StrData[0] == '\n' || StrData[0] == '\r')
469  return StrData;
470 
471  const char *LexStart = StrData;
472  while (LexStart != BufStart) {
473  if (LexStart[0] == '\n' || LexStart[0] == '\r') {
474  ++LexStart;
475  break;
476  }
477 
478  --LexStart;
479  }
480  return LexStart;
481 }
482 
484  const SourceManager &SM,
485  const LangOptions &LangOpts) {
486  assert(Loc.isFileID());
487  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
488  if (LocInfo.first.isInvalid())
489  return Loc;
490 
491  bool Invalid = false;
492  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
493  if (Invalid)
494  return Loc;
495 
496  // Back up from the current location until we hit the beginning of a line
497  // (or the buffer). We'll relex from that point.
498  const char *StrData = Buffer.data() + LocInfo.second;
499  const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
500  if (!LexStart || LexStart == StrData)
501  return Loc;
502 
503  // Create a lexer starting at the beginning of this token.
504  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
505  Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
506  Buffer.end());
507  TheLexer.SetCommentRetentionState(true);
508 
509  // Lex tokens until we find the token that contains the source location.
510  Token TheTok;
511  do {
512  TheLexer.LexFromRawLexer(TheTok);
513 
514  if (TheLexer.getBufferLocation() > StrData) {
515  // Lexing this token has taken the lexer past the source location we're
516  // looking for. If the current token encompasses our source location,
517  // return the beginning of that token.
518  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
519  return TheTok.getLocation();
520 
521  // We ended up skipping over the source location entirely, which means
522  // that it points into whitespace. We're done here.
523  break;
524  }
525  } while (TheTok.getKind() != tok::eof);
526 
527  // We've passed our source location; just return the original source location.
528  return Loc;
529 }
530 
532  const SourceManager &SM,
533  const LangOptions &LangOpts) {
534  if (Loc.isFileID())
535  return getBeginningOfFileToken(Loc, SM, LangOpts);
536 
537  if (!SM.isMacroArgExpansion(Loc))
538  return Loc;
539 
540  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
541  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
542  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
543  std::pair<FileID, unsigned> BeginFileLocInfo
544  = SM.getDecomposedLoc(BeginFileLoc);
545  assert(FileLocInfo.first == BeginFileLocInfo.first &&
546  FileLocInfo.second >= BeginFileLocInfo.second);
547  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
548 }
549 
550 namespace {
551 
553  PDK_Skipped,
554  PDK_Unknown
555  };
556 
557 } // end anonymous namespace
558 
559 std::pair<unsigned, bool> Lexer::ComputePreamble(StringRef Buffer,
560  const LangOptions &LangOpts,
561  unsigned MaxLines) {
562  // Create a lexer starting at the beginning of the file. Note that we use a
563  // "fake" file source location at offset 1 so that the lexer will track our
564  // position within the file.
565  const unsigned StartOffset = 1;
566  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
567  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
568  Buffer.end());
569  TheLexer.SetCommentRetentionState(true);
570 
571  // StartLoc will differ from FileLoc if there is a BOM that was skipped.
572  SourceLocation StartLoc = TheLexer.getSourceLocation();
573 
574  bool InPreprocessorDirective = false;
575  Token TheTok;
576  SourceLocation ActiveCommentLoc;
577 
578  unsigned MaxLineOffset = 0;
579  if (MaxLines) {
580  const char *CurPtr = Buffer.begin();
581  unsigned CurLine = 0;
582  while (CurPtr != Buffer.end()) {
583  char ch = *CurPtr++;
584  if (ch == '\n') {
585  ++CurLine;
586  if (CurLine == MaxLines)
587  break;
588  }
589  }
590  if (CurPtr != Buffer.end())
591  MaxLineOffset = CurPtr - Buffer.begin();
592  }
593 
594  do {
595  TheLexer.LexFromRawLexer(TheTok);
596 
597  if (InPreprocessorDirective) {
598  // If we've hit the end of the file, we're done.
599  if (TheTok.getKind() == tok::eof) {
600  break;
601  }
602 
603  // If we haven't hit the end of the preprocessor directive, skip this
604  // token.
605  if (!TheTok.isAtStartOfLine())
606  continue;
607 
608  // We've passed the end of the preprocessor directive, and will look
609  // at this token again below.
610  InPreprocessorDirective = false;
611  }
612 
613  // Keep track of the # of lines in the preamble.
614  if (TheTok.isAtStartOfLine()) {
615  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
616 
617  // If we were asked to limit the number of lines in the preamble,
618  // and we're about to exceed that limit, we're done.
619  if (MaxLineOffset && TokOffset >= MaxLineOffset)
620  break;
621  }
622 
623  // Comments are okay; skip over them.
624  if (TheTok.getKind() == tok::comment) {
625  if (ActiveCommentLoc.isInvalid())
626  ActiveCommentLoc = TheTok.getLocation();
627  continue;
628  }
629 
630  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
631  // This is the start of a preprocessor directive.
632  Token HashTok = TheTok;
633  InPreprocessorDirective = true;
634  ActiveCommentLoc = SourceLocation();
635 
636  // Figure out which directive this is. Since we're lexing raw tokens,
637  // we don't have an identifier table available. Instead, just look at
638  // the raw identifier to recognize and categorize preprocessor directives.
639  TheLexer.LexFromRawLexer(TheTok);
640  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
641  StringRef Keyword = TheTok.getRawIdentifier();
643  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
644  .Case("include", PDK_Skipped)
645  .Case("__include_macros", PDK_Skipped)
646  .Case("define", PDK_Skipped)
647  .Case("undef", PDK_Skipped)
648  .Case("line", PDK_Skipped)
649  .Case("error", PDK_Skipped)
650  .Case("pragma", PDK_Skipped)
651  .Case("import", PDK_Skipped)
652  .Case("include_next", PDK_Skipped)
653  .Case("warning", PDK_Skipped)
654  .Case("ident", PDK_Skipped)
655  .Case("sccs", PDK_Skipped)
656  .Case("assert", PDK_Skipped)
657  .Case("unassert", PDK_Skipped)
658  .Case("if", PDK_Skipped)
659  .Case("ifdef", PDK_Skipped)
660  .Case("ifndef", PDK_Skipped)
661  .Case("elif", PDK_Skipped)
662  .Case("else", PDK_Skipped)
663  .Case("endif", PDK_Skipped)
664  .Default(PDK_Unknown);
665 
666  switch (PDK) {
667  case PDK_Skipped:
668  continue;
669 
670  case PDK_Unknown:
671  // We don't know what this directive is; stop at the '#'.
672  break;
673  }
674  }
675 
676  // We only end up here if we didn't recognize the preprocessor
677  // directive or it was one that can't occur in the preamble at this
678  // point. Roll back the current token to the location of the '#'.
679  InPreprocessorDirective = false;
680  TheTok = HashTok;
681  }
682 
683  // We hit a token that we don't recognize as being in the
684  // "preprocessing only" part of the file, so we're no longer in
685  // the preamble.
686  break;
687  } while (true);
688 
690  if (ActiveCommentLoc.isValid())
691  End = ActiveCommentLoc; // don't truncate a decl comment.
692  else
693  End = TheTok.getLocation();
694 
695  return std::make_pair(End.getRawEncoding() - StartLoc.getRawEncoding(),
696  TheTok.isAtStartOfLine());
697 }
698 
699 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
700 /// token, return a new location that specifies a character within the token.
702  unsigned CharNo,
703  const SourceManager &SM,
704  const LangOptions &LangOpts) {
705  // Figure out how many physical characters away the specified expansion
706  // character is. This needs to take into consideration newlines and
707  // trigraphs.
708  bool Invalid = false;
709  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
710 
711  // If they request the first char of the token, we're trivially done.
712  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
713  return TokStart;
714 
715  unsigned PhysOffset = 0;
716 
717  // The usual case is that tokens don't contain anything interesting. Skip
718  // over the uninteresting characters. If a token only consists of simple
719  // chars, this method is extremely fast.
720  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
721  if (CharNo == 0)
722  return TokStart.getLocWithOffset(PhysOffset);
723  ++TokPtr;
724  --CharNo;
725  ++PhysOffset;
726  }
727 
728  // If we have a character that may be a trigraph or escaped newline, use a
729  // lexer to parse it correctly.
730  for (; CharNo; --CharNo) {
731  unsigned Size;
732  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
733  TokPtr += Size;
734  PhysOffset += Size;
735  }
736 
737  // Final detail: if we end up on an escaped newline, we want to return the
738  // location of the actual byte of the token. For example foo<newline>bar
739  // advanced by 3 should return the location of b, not of \\. One compounding
740  // detail of this is that the escape may be made by a trigraph.
741  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
742  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
743 
744  return TokStart.getLocWithOffset(PhysOffset);
745 }
746 
747 /// \brief Computes the source location just past the end of the
748 /// token at this source location.
749 ///
750 /// This routine can be used to produce a source location that
751 /// points just past the end of the token referenced by \p Loc, and
752 /// is generally used when a diagnostic needs to point just after a
753 /// token where it expected something different that it received. If
754 /// the returned source location would not be meaningful (e.g., if
755 /// it points into a macro), this routine returns an invalid
756 /// source location.
757 ///
758 /// \param Offset an offset from the end of the token, where the source
759 /// location should refer to. The default offset (0) produces a source
760 /// location pointing just past the end of the token; an offset of 1 produces
761 /// a source location pointing to the last character in the token, etc.
763  const SourceManager &SM,
764  const LangOptions &LangOpts) {
765  if (Loc.isInvalid())
766  return SourceLocation();
767 
768  if (Loc.isMacroID()) {
769  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
770  return SourceLocation(); // Points inside the macro expansion.
771  }
772 
773  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
774  if (Len > Offset)
775  Len = Len - Offset;
776  else
777  return Loc;
778 
779  return Loc.getLocWithOffset(Len);
780 }
781 
782 /// \brief Returns true if the given MacroID location points at the first
783 /// token of the macro expansion.
785  const SourceManager &SM,
786  const LangOptions &LangOpts,
787  SourceLocation *MacroBegin) {
788  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
789 
790  SourceLocation expansionLoc;
791  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
792  return false;
793 
794  if (expansionLoc.isFileID()) {
795  // No other macro expansions, this is the first.
796  if (MacroBegin)
797  *MacroBegin = expansionLoc;
798  return true;
799  }
800 
801  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
802 }
803 
804 /// \brief Returns true if the given MacroID location points at the last
805 /// token of the macro expansion.
807  const SourceManager &SM,
808  const LangOptions &LangOpts,
809  SourceLocation *MacroEnd) {
810  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
811 
812  SourceLocation spellLoc = SM.getSpellingLoc(loc);
813  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
814  if (tokLen == 0)
815  return false;
816 
817  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
818  SourceLocation expansionLoc;
819  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
820  return false;
821 
822  if (expansionLoc.isFileID()) {
823  // No other macro expansions.
824  if (MacroEnd)
825  *MacroEnd = expansionLoc;
826  return true;
827  }
828 
829  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
830 }
831 
833  const SourceManager &SM,
834  const LangOptions &LangOpts) {
835  SourceLocation Begin = Range.getBegin();
836  SourceLocation End = Range.getEnd();
837  assert(Begin.isFileID() && End.isFileID());
838  if (Range.isTokenRange()) {
839  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
840  if (End.isInvalid())
841  return CharSourceRange();
842  }
843 
844  // Break down the source locations.
845  FileID FID;
846  unsigned BeginOffs;
847  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
848  if (FID.isInvalid())
849  return CharSourceRange();
850 
851  unsigned EndOffs;
852  if (!SM.isInFileID(End, FID, &EndOffs) ||
853  BeginOffs > EndOffs)
854  return CharSourceRange();
855 
856  return CharSourceRange::getCharRange(Begin, End);
857 }
858 
860  const SourceManager &SM,
861  const LangOptions &LangOpts) {
862  SourceLocation Begin = Range.getBegin();
863  SourceLocation End = Range.getEnd();
864  if (Begin.isInvalid() || End.isInvalid())
865  return CharSourceRange();
866 
867  if (Begin.isFileID() && End.isFileID())
868  return makeRangeFromFileLocs(Range, SM, LangOpts);
869 
870  if (Begin.isMacroID() && End.isFileID()) {
871  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
872  return CharSourceRange();
873  Range.setBegin(Begin);
874  return makeRangeFromFileLocs(Range, SM, LangOpts);
875  }
876 
877  if (Begin.isFileID() && End.isMacroID()) {
878  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
879  &End)) ||
880  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
881  &End)))
882  return CharSourceRange();
883  Range.setEnd(End);
884  return makeRangeFromFileLocs(Range, SM, LangOpts);
885  }
886 
887  assert(Begin.isMacroID() && End.isMacroID());
888  SourceLocation MacroBegin, MacroEnd;
889  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
890  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
891  &MacroEnd)) ||
892  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
893  &MacroEnd)))) {
894  Range.setBegin(MacroBegin);
895  Range.setEnd(MacroEnd);
896  return makeRangeFromFileLocs(Range, SM, LangOpts);
897  }
898 
899  bool Invalid = false;
900  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
901  &Invalid);
902  if (Invalid)
903  return CharSourceRange();
904 
905  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
906  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
907  &Invalid);
908  if (Invalid)
909  return CharSourceRange();
910 
911  if (EndEntry.getExpansion().isMacroArgExpansion() &&
912  BeginEntry.getExpansion().getExpansionLocStart() ==
913  EndEntry.getExpansion().getExpansionLocStart()) {
914  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
915  Range.setEnd(SM.getImmediateSpellingLoc(End));
916  return makeFileCharRange(Range, SM, LangOpts);
917  }
918  }
919 
920  return CharSourceRange();
921 }
922 
924  const SourceManager &SM,
925  const LangOptions &LangOpts,
926  bool *Invalid) {
927  Range = makeFileCharRange(Range, SM, LangOpts);
928  if (Range.isInvalid()) {
929  if (Invalid) *Invalid = true;
930  return StringRef();
931  }
932 
933  // Break down the source location.
934  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
935  if (beginInfo.first.isInvalid()) {
936  if (Invalid) *Invalid = true;
937  return StringRef();
938  }
939 
940  unsigned EndOffs;
941  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
942  beginInfo.second > EndOffs) {
943  if (Invalid) *Invalid = true;
944  return StringRef();
945  }
946 
947  // Try to the load the file buffer.
948  bool invalidTemp = false;
949  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
950  if (invalidTemp) {
951  if (Invalid) *Invalid = true;
952  return StringRef();
953  }
954 
955  if (Invalid) *Invalid = false;
956  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
957 }
958 
960  const SourceManager &SM,
961  const LangOptions &LangOpts) {
962  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
963 
964  // Find the location of the immediate macro expansion.
965  while (true) {
966  FileID FID = SM.getFileID(Loc);
967  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
968  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
969  Loc = Expansion.getExpansionLocStart();
970  if (!Expansion.isMacroArgExpansion())
971  break;
972 
973  // For macro arguments we need to check that the argument did not come
974  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
975 
976  // Loc points to the argument id of the macro definition, move to the
977  // macro expansion.
978  Loc = SM.getImmediateExpansionRange(Loc).first;
979  SourceLocation SpellLoc = Expansion.getSpellingLoc();
980  if (SpellLoc.isFileID())
981  break; // No inner macro.
982 
983  // If spelling location resides in the same FileID as macro expansion
984  // location, it means there is no inner macro.
985  FileID MacroFID = SM.getFileID(Loc);
986  if (SM.isInFileID(SpellLoc, MacroFID))
987  break;
988 
989  // Argument came from inner macro.
990  Loc = SpellLoc;
991  }
992 
993  // Find the spelling location of the start of the non-argument expansion
994  // range. This is where the macro name was spelled in order to begin
995  // expanding this macro.
996  Loc = SM.getSpellingLoc(Loc);
997 
998  // Dig out the buffer where the macro name was spelled and the extents of the
999  // name so that we can render it into the expansion note.
1000  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1001  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1002  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1003  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1004 }
1005 
1007  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1008  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
1009  // Walk past macro argument expanions.
1010  while (SM.isMacroArgExpansion(Loc))
1011  Loc = SM.getImmediateExpansionRange(Loc).first;
1012 
1013  // If the macro's spelling has no FileID, then it's actually a token paste
1014  // or stringization (or similar) and not a macro at all.
1015  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1016  return StringRef();
1017 
1018  // Find the spelling location of the start of the non-argument expansion
1019  // range. This is where the macro name was spelled in order to begin
1020  // expanding this macro.
1021  Loc = SM.getSpellingLoc(SM.getImmediateExpansionRange(Loc).first);
1022 
1023  // Dig out the buffer where the macro name was spelled and the extents of the
1024  // name so that we can render it into the expansion note.
1025  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1026  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1027  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1028  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1029 }
1030 
1031 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1032  return isIdentifierBody(c, LangOpts.DollarIdents);
1033 }
1034 
1036  const SourceManager &SM) {
1037  if (Loc.isInvalid() || Loc.isMacroID())
1038  return "";
1039  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1040  if (LocInfo.first.isInvalid())
1041  return "";
1042  bool Invalid = false;
1043  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1044  if (Invalid)
1045  return "";
1046  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1047  if (!Line)
1048  return "";
1049  StringRef Rest = Buffer.substr(Line - Buffer.data());
1050  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1051  return NumWhitespaceChars == StringRef::npos
1052  ? ""
1053  : Rest.take_front(NumWhitespaceChars);
1054 }
1055 
1056 //===----------------------------------------------------------------------===//
1057 // Diagnostics forwarding code.
1058 //===----------------------------------------------------------------------===//
1059 
1060 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1061 /// lexer buffer was all expanded at a single point, perform the mapping.
1062 /// This is currently only used for _Pragma implementation, so it is the slow
1063 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1064 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1065  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1067  SourceLocation FileLoc,
1068  unsigned CharNo, unsigned TokLen) {
1069  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1070 
1071  // Otherwise, we're lexing "mapped tokens". This is used for things like
1072  // _Pragma handling. Combine the expansion location of FileLoc with the
1073  // spelling location.
1075 
1076  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1077  // characters come from spelling(FileLoc)+Offset.
1078  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1079  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1080 
1081  // Figure out the expansion loc range, which is the range covered by the
1082  // original _Pragma(...) sequence.
1083  std::pair<SourceLocation,SourceLocation> II =
1084  SM.getImmediateExpansionRange(FileLoc);
1085 
1086  return SM.createExpansionLoc(SpellingLoc, II.first, II.second, TokLen);
1087 }
1088 
1089 /// getSourceLocation - Return a source location identifier for the specified
1090 /// offset in the current file.
1092  unsigned TokLen) const {
1093  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1094  "Location out of range for this buffer!");
1095 
1096  // In the normal case, we're just lexing from a simple file buffer, return
1097  // the file id from FileLoc with the offset specified.
1098  unsigned CharNo = Loc-BufferStart;
1099  if (FileLoc.isFileID())
1100  return FileLoc.getLocWithOffset(CharNo);
1101 
1102  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1103  // tokens are lexed from where the _Pragma was defined.
1104  assert(PP && "This doesn't work on raw lexers");
1105  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1106 }
1107 
1108 /// Diag - Forwarding function for diagnostics. This translate a source
1109 /// position in the current buffer into a SourceLocation object for rendering.
1110 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1111  return PP->Diag(getSourceLocation(Loc), DiagID);
1112 }
1113 
1114 //===----------------------------------------------------------------------===//
1115 // Trigraph and Escaped Newline Handling Code.
1116 //===----------------------------------------------------------------------===//
1117 
1118 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1119 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1120 static char GetTrigraphCharForLetter(char Letter) {
1121  switch (Letter) {
1122  default: return 0;
1123  case '=': return '#';
1124  case ')': return ']';
1125  case '(': return '[';
1126  case '!': return '|';
1127  case '\'': return '^';
1128  case '>': return '}';
1129  case '/': return '\\';
1130  case '<': return '{';
1131  case '-': return '~';
1132  }
1133 }
1134 
1135 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1136 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1137 /// return the result character. Finally, emit a warning about trigraph use
1138 /// whether trigraphs are enabled or not.
1139 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1140  char Res = GetTrigraphCharForLetter(*CP);
1141  if (!Res || !L) return Res;
1142 
1143  if (!L->getLangOpts().Trigraphs) {
1144  if (!L->isLexingRawMode())
1145  L->Diag(CP-2, diag::trigraph_ignored);
1146  return 0;
1147  }
1148 
1149  if (!L->isLexingRawMode())
1150  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1151  return Res;
1152 }
1153 
1154 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1155 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1156 /// trigraph equivalent on entry to this function.
1157 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1158  unsigned Size = 0;
1159  while (isWhitespace(Ptr[Size])) {
1160  ++Size;
1161 
1162  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1163  continue;
1164 
1165  // If this is a \r\n or \n\r, skip the other half.
1166  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1167  Ptr[Size-1] != Ptr[Size])
1168  ++Size;
1169 
1170  return Size;
1171  }
1172 
1173  // Not an escaped newline, must be a \t or something else.
1174  return 0;
1175 }
1176 
1177 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1178 /// them), skip over them and return the first non-escaped-newline found,
1179 /// otherwise return P.
1180 const char *Lexer::SkipEscapedNewLines(const char *P) {
1181  while (true) {
1182  const char *AfterEscape;
1183  if (*P == '\\') {
1184  AfterEscape = P+1;
1185  } else if (*P == '?') {
1186  // If not a trigraph for escape, bail out.
1187  if (P[1] != '?' || P[2] != '/')
1188  return P;
1189  // FIXME: Take LangOpts into account; the language might not
1190  // support trigraphs.
1191  AfterEscape = P+3;
1192  } else {
1193  return P;
1194  }
1195 
1196  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1197  if (NewLineSize == 0) return P;
1198  P = AfterEscape+NewLineSize;
1199  }
1200 }
1201 
1202 /// \brief Checks that the given token is the first token that occurs after the
1203 /// given location (this excludes comments and whitespace). Returns the location
1204 /// immediately after the specified token. If the token is not found or the
1205 /// location is inside a macro, the returned source location will be invalid.
1207  tok::TokenKind TKind,
1208  const SourceManager &SM,
1209  const LangOptions &LangOpts,
1210  bool SkipTrailingWhitespaceAndNewLine) {
1211  if (Loc.isMacroID()) {
1212  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1213  return SourceLocation();
1214  }
1215  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1216 
1217  // Break down the source location.
1218  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1219 
1220  // Try to load the file buffer.
1221  bool InvalidTemp = false;
1222  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1223  if (InvalidTemp)
1224  return SourceLocation();
1225 
1226  const char *TokenBegin = File.data() + LocInfo.second;
1227 
1228  // Lex from the start of the given location.
1229  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1230  TokenBegin, File.end());
1231  // Find the token.
1232  Token Tok;
1233  lexer.LexFromRawLexer(Tok);
1234  if (Tok.isNot(TKind))
1235  return SourceLocation();
1236  SourceLocation TokenLoc = Tok.getLocation();
1237 
1238  // Calculate how much whitespace needs to be skipped if any.
1239  unsigned NumWhitespaceChars = 0;
1240  if (SkipTrailingWhitespaceAndNewLine) {
1241  const char *TokenEnd = SM.getCharacterData(TokenLoc) +
1242  Tok.getLength();
1243  unsigned char C = *TokenEnd;
1244  while (isHorizontalWhitespace(C)) {
1245  C = *(++TokenEnd);
1246  NumWhitespaceChars++;
1247  }
1248 
1249  // Skip \r, \n, \r\n, or \n\r
1250  if (C == '\n' || C == '\r') {
1251  char PrevC = C;
1252  C = *(++TokenEnd);
1253  NumWhitespaceChars++;
1254  if ((C == '\n' || C == '\r') && C != PrevC)
1255  NumWhitespaceChars++;
1256  }
1257  }
1258 
1259  return TokenLoc.getLocWithOffset(Tok.getLength() + NumWhitespaceChars);
1260 }
1261 
1262 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1263 /// get its size, and return it. This is tricky in several cases:
1264 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1265 /// then either return the trigraph (skipping 3 chars) or the '?',
1266 /// depending on whether trigraphs are enabled or not.
1267 /// 2. If this is an escaped newline (potentially with whitespace between
1268 /// the backslash and newline), implicitly skip the newline and return
1269 /// the char after it.
1270 ///
1271 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1272 /// know that we can accumulate into Size, and that we have already incremented
1273 /// Ptr by Size bytes.
1274 ///
1275 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1276 /// be updated to match.
1277 ///
1278 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1279  Token *Tok) {
1280  // If we have a slash, look for an escaped newline.
1281  if (Ptr[0] == '\\') {
1282  ++Size;
1283  ++Ptr;
1284 Slash:
1285  // Common case, backslash-char where the char is not whitespace.
1286  if (!isWhitespace(Ptr[0])) return '\\';
1287 
1288  // See if we have optional whitespace characters between the slash and
1289  // newline.
1290  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1291  // Remember that this token needs to be cleaned.
1292  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1293 
1294  // Warn if there was whitespace between the backslash and newline.
1295  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1296  Diag(Ptr, diag::backslash_newline_space);
1297 
1298  // Found backslash<whitespace><newline>. Parse the char after it.
1299  Size += EscapedNewLineSize;
1300  Ptr += EscapedNewLineSize;
1301 
1302  // Use slow version to accumulate a correct size field.
1303  return getCharAndSizeSlow(Ptr, Size, Tok);
1304  }
1305 
1306  // Otherwise, this is not an escaped newline, just return the slash.
1307  return '\\';
1308  }
1309 
1310  // If this is a trigraph, process it.
1311  if (Ptr[0] == '?' && Ptr[1] == '?') {
1312  // If this is actually a legal trigraph (not something like "??x"), emit
1313  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1314  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1315  // Remember that this token needs to be cleaned.
1316  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1317 
1318  Ptr += 3;
1319  Size += 3;
1320  if (C == '\\') goto Slash;
1321  return C;
1322  }
1323  }
1324 
1325  // If this is neither, return a single character.
1326  ++Size;
1327  return *Ptr;
1328 }
1329 
1330 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1331 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1332 /// and that we have already incremented Ptr by Size bytes.
1333 ///
1334 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1335 /// be updated to match.
1336 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1337  const LangOptions &LangOpts) {
1338  // If we have a slash, look for an escaped newline.
1339  if (Ptr[0] == '\\') {
1340  ++Size;
1341  ++Ptr;
1342 Slash:
1343  // Common case, backslash-char where the char is not whitespace.
1344  if (!isWhitespace(Ptr[0])) return '\\';
1345 
1346  // See if we have optional whitespace characters followed by a newline.
1347  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1348  // Found backslash<whitespace><newline>. Parse the char after it.
1349  Size += EscapedNewLineSize;
1350  Ptr += EscapedNewLineSize;
1351 
1352  // Use slow version to accumulate a correct size field.
1353  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1354  }
1355 
1356  // Otherwise, this is not an escaped newline, just return the slash.
1357  return '\\';
1358  }
1359 
1360  // If this is a trigraph, process it.
1361  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1362  // If this is actually a legal trigraph (not something like "??x"), return
1363  // it.
1364  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1365  Ptr += 3;
1366  Size += 3;
1367  if (C == '\\') goto Slash;
1368  return C;
1369  }
1370  }
1371 
1372  // If this is neither, return a single character.
1373  ++Size;
1374  return *Ptr;
1375 }
1376 
1377 //===----------------------------------------------------------------------===//
1378 // Helper methods for lexing.
1379 //===----------------------------------------------------------------------===//
1380 
1381 /// \brief Routine that indiscriminately skips bytes in the source file.
1382 void Lexer::SkipBytes(unsigned Bytes, bool StartOfLine) {
1383  BufferPtr += Bytes;
1384  if (BufferPtr > BufferEnd)
1385  BufferPtr = BufferEnd;
1386  // FIXME: What exactly does the StartOfLine bit mean? There are two
1387  // possible meanings for the "start" of the line: the first token on the
1388  // unexpanded line, or the first token on the expanded line.
1389  IsAtStartOfLine = StartOfLine;
1390  IsAtPhysicalStartOfLine = StartOfLine;
1391 }
1392 
1393 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1394  if (LangOpts.AsmPreprocessor) {
1395  return false;
1396  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1397  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1399  return C11AllowedIDChars.contains(C);
1400  } else if (LangOpts.CPlusPlus) {
1401  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1403  return CXX03AllowedIDChars.contains(C);
1404  } else {
1405  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1407  return C99AllowedIDChars.contains(C);
1408  }
1409 }
1410 
1411 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1412  assert(isAllowedIDChar(C, LangOpts));
1413  if (LangOpts.AsmPreprocessor) {
1414  return false;
1415  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1416  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1418  return !C11DisallowedInitialIDChars.contains(C);
1419  } else if (LangOpts.CPlusPlus) {
1420  return true;
1421  } else {
1422  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1424  return !C99DisallowedInitialIDChars.contains(C);
1425  }
1426 }
1427 
1428 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1429  const char *End) {
1431  L.getSourceLocation(End));
1432 }
1433 
1434 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1435  CharSourceRange Range, bool IsFirst) {
1436  // Check C99 compatibility.
1437  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1438  enum {
1439  CannotAppearInIdentifier = 0,
1440  CannotStartIdentifier
1441  };
1442 
1443  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1445  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1447  if (!C99AllowedIDChars.contains(C)) {
1448  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1449  << Range
1450  << CannotAppearInIdentifier;
1451  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1452  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1453  << Range
1454  << CannotStartIdentifier;
1455  }
1456  }
1457 
1458  // Check C++98 compatibility.
1459  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1460  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1462  if (!CXX03AllowedIDChars.contains(C)) {
1463  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1464  << Range;
1465  }
1466  }
1467 }
1468 
1469 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1470  Token &Result) {
1471  const char *UCNPtr = CurPtr + Size;
1472  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1473  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1474  return false;
1475 
1476  if (!isLexingRawMode())
1478  makeCharRange(*this, CurPtr, UCNPtr),
1479  /*IsFirst=*/false);
1480 
1481  Result.setFlag(Token::HasUCN);
1482  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1483  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1484  CurPtr = UCNPtr;
1485  else
1486  while (CurPtr != UCNPtr)
1487  (void)getAndAdvanceChar(CurPtr, Result);
1488  return true;
1489 }
1490 
1491 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1492  const char *UnicodePtr = CurPtr;
1493  llvm::UTF32 CodePoint;
1494  llvm::ConversionResult Result =
1495  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1496  (const llvm::UTF8 *)BufferEnd,
1497  &CodePoint,
1498  llvm::strictConversion);
1499  if (Result != llvm::conversionOK ||
1500  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1501  return false;
1502 
1503  if (!isLexingRawMode())
1505  makeCharRange(*this, CurPtr, UnicodePtr),
1506  /*IsFirst=*/false);
1507 
1508  CurPtr = UnicodePtr;
1509  return true;
1510 }
1511 
1512 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1513  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1514  unsigned Size;
1515  unsigned char C = *CurPtr++;
1516  while (isIdentifierBody(C))
1517  C = *CurPtr++;
1518 
1519  --CurPtr; // Back up over the skipped character.
1520 
1521  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1522  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1523  //
1524  // TODO: Could merge these checks into an InfoTable flag to make the
1525  // comparison cheaper
1526  if (isASCII(C) && C != '\\' && C != '?' &&
1527  (C != '$' || !LangOpts.DollarIdents)) {
1528 FinishIdentifier:
1529  const char *IdStart = BufferPtr;
1530  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1531  Result.setRawIdentifierData(IdStart);
1532 
1533  // If we are in raw mode, return this identifier raw. There is no need to
1534  // look up identifier information or attempt to macro expand it.
1535  if (LexingRawMode)
1536  return true;
1537 
1538  // Fill in Result.IdentifierInfo and update the token kind,
1539  // looking up the identifier in the identifier table.
1540  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1541 
1542  // Finally, now that we know we have an identifier, pass this off to the
1543  // preprocessor, which may macro expand it or something.
1544  if (II->isHandleIdentifierCase())
1545  return PP->HandleIdentifier(Result);
1546 
1547  if (II->getTokenID() == tok::identifier && isCodeCompletionPoint(CurPtr)
1548  && II->getPPKeywordID() == tok::pp_not_keyword
1549  && II->getObjCKeywordID() == tok::objc_not_keyword) {
1550  // Return the code-completion token.
1551  Result.setKind(tok::code_completion);
1552  cutOffLexing();
1553  return true;
1554  }
1555  return true;
1556  }
1557 
1558  // Otherwise, $,\,? in identifier found. Enter slower path.
1559 
1560  C = getCharAndSize(CurPtr, Size);
1561  while (true) {
1562  if (C == '$') {
1563  // If we hit a $ and they are not supported in identifiers, we are done.
1564  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1565 
1566  // Otherwise, emit a diagnostic and continue.
1567  if (!isLexingRawMode())
1568  Diag(CurPtr, diag::ext_dollar_in_identifier);
1569  CurPtr = ConsumeChar(CurPtr, Size, Result);
1570  C = getCharAndSize(CurPtr, Size);
1571  continue;
1572 
1573  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1574  C = getCharAndSize(CurPtr, Size);
1575  continue;
1576  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1577  C = getCharAndSize(CurPtr, Size);
1578  continue;
1579  } else if (!isIdentifierBody(C)) {
1580  goto FinishIdentifier;
1581  }
1582 
1583  // Otherwise, this character is good, consume it.
1584  CurPtr = ConsumeChar(CurPtr, Size, Result);
1585 
1586  C = getCharAndSize(CurPtr, Size);
1587  while (isIdentifierBody(C)) {
1588  CurPtr = ConsumeChar(CurPtr, Size, Result);
1589  C = getCharAndSize(CurPtr, Size);
1590  }
1591  }
1592 }
1593 
1594 /// isHexaLiteral - Return true if Start points to a hex constant.
1595 /// in microsoft mode (where this is supposed to be several different tokens).
1596 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1597  unsigned Size;
1598  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1599  if (C1 != '0')
1600  return false;
1601  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1602  return (C2 == 'x' || C2 == 'X');
1603 }
1604 
1605 /// LexNumericConstant - Lex the remainder of a integer or floating point
1606 /// constant. From[-1] is the first character lexed. Return the end of the
1607 /// constant.
1608 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1609  unsigned Size;
1610  char C = getCharAndSize(CurPtr, Size);
1611  char PrevCh = 0;
1612  while (isPreprocessingNumberBody(C)) {
1613  CurPtr = ConsumeChar(CurPtr, Size, Result);
1614  PrevCh = C;
1615  C = getCharAndSize(CurPtr, Size);
1616  }
1617 
1618  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1619  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1620  // If we are in Microsoft mode, don't continue if the constant is hex.
1621  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1622  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1623  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1624  }
1625 
1626  // If we have a hex FP constant, continue.
1627  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1628  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1629  // not-quite-conforming extension. Only do so if this looks like it's
1630  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1631  bool IsHexFloat = true;
1632  if (!LangOpts.C99) {
1633  if (!isHexaLiteral(BufferPtr, LangOpts))
1634  IsHexFloat = false;
1635  else if (!getLangOpts().CPlusPlus1z &&
1636  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1637  IsHexFloat = false;
1638  }
1639  if (IsHexFloat)
1640  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1641  }
1642 
1643  // If we have a digit separator, continue.
1644  if (C == '\'' && getLangOpts().CPlusPlus14) {
1645  unsigned NextSize;
1646  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1647  if (isIdentifierBody(Next)) {
1648  if (!isLexingRawMode())
1649  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1650  CurPtr = ConsumeChar(CurPtr, Size, Result);
1651  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1652  return LexNumericConstant(Result, CurPtr);
1653  }
1654  }
1655 
1656  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1657  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1658  return LexNumericConstant(Result, CurPtr);
1659  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1660  return LexNumericConstant(Result, CurPtr);
1661 
1662  // Update the location of token as well as BufferPtr.
1663  const char *TokStart = BufferPtr;
1664  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1665  Result.setLiteralData(TokStart);
1666  return true;
1667 }
1668 
1669 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1670 /// in C++11, or warn on a ud-suffix in C++98.
1671 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1672  bool IsStringLiteral) {
1673  assert(getLangOpts().CPlusPlus);
1674 
1675  // Maximally munch an identifier.
1676  unsigned Size;
1677  char C = getCharAndSize(CurPtr, Size);
1678  bool Consumed = false;
1679 
1680  if (!isIdentifierHead(C)) {
1681  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1682  Consumed = true;
1683  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1684  Consumed = true;
1685  else
1686  return CurPtr;
1687  }
1688 
1689  if (!getLangOpts().CPlusPlus11) {
1690  if (!isLexingRawMode())
1691  Diag(CurPtr,
1692  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1693  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1695  return CurPtr;
1696  }
1697 
1698  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1699  // that does not start with an underscore is ill-formed. As a conforming
1700  // extension, we treat all such suffixes as if they had whitespace before
1701  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1702  // likely to be a ud-suffix than a macro, however, and accept that.
1703  if (!Consumed) {
1704  bool IsUDSuffix = false;
1705  if (C == '_')
1706  IsUDSuffix = true;
1707  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1708  // In C++1y, we need to look ahead a few characters to see if this is a
1709  // valid suffix for a string literal or a numeric literal (this could be
1710  // the 'operator""if' defining a numeric literal operator).
1711  const unsigned MaxStandardSuffixLength = 3;
1712  char Buffer[MaxStandardSuffixLength] = { C };
1713  unsigned Consumed = Size;
1714  unsigned Chars = 1;
1715  while (true) {
1716  unsigned NextSize;
1717  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1718  getLangOpts());
1719  if (!isIdentifierBody(Next)) {
1720  // End of suffix. Check whether this is on the whitelist.
1721  const StringRef CompleteSuffix(Buffer, Chars);
1723  CompleteSuffix);
1724  break;
1725  }
1726 
1727  if (Chars == MaxStandardSuffixLength)
1728  // Too long: can't be a standard suffix.
1729  break;
1730 
1731  Buffer[Chars++] = Next;
1732  Consumed += NextSize;
1733  }
1734  }
1735 
1736  if (!IsUDSuffix) {
1737  if (!isLexingRawMode())
1738  Diag(CurPtr, getLangOpts().MSVCCompat
1739  ? diag::ext_ms_reserved_user_defined_literal
1740  : diag::ext_reserved_user_defined_literal)
1742  return CurPtr;
1743  }
1744 
1745  CurPtr = ConsumeChar(CurPtr, Size, Result);
1746  }
1747 
1748  Result.setFlag(Token::HasUDSuffix);
1749  while (true) {
1750  C = getCharAndSize(CurPtr, Size);
1751  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1752  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1753  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1754  else break;
1755  }
1756 
1757  return CurPtr;
1758 }
1759 
1760 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1761 /// either " or L" or u8" or u" or U".
1762 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1763  tok::TokenKind Kind) {
1764  // Does this string contain the \0 character?
1765  const char *NulCharacter = nullptr;
1766 
1767  if (!isLexingRawMode() &&
1768  (Kind == tok::utf8_string_literal ||
1769  Kind == tok::utf16_string_literal ||
1770  Kind == tok::utf32_string_literal))
1771  Diag(BufferPtr, getLangOpts().CPlusPlus
1772  ? diag::warn_cxx98_compat_unicode_literal
1773  : diag::warn_c99_compat_unicode_literal);
1774 
1775  char C = getAndAdvanceChar(CurPtr, Result);
1776  while (C != '"') {
1777  // Skip escaped characters. Escaped newlines will already be processed by
1778  // getAndAdvanceChar.
1779  if (C == '\\')
1780  C = getAndAdvanceChar(CurPtr, Result);
1781 
1782  if (C == '\n' || C == '\r' || // Newline.
1783  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1784  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1785  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1786  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1787  return true;
1788  }
1789 
1790  if (C == 0) {
1791  if (isCodeCompletionPoint(CurPtr-1)) {
1793  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1794  cutOffLexing();
1795  return true;
1796  }
1797 
1798  NulCharacter = CurPtr-1;
1799  }
1800  C = getAndAdvanceChar(CurPtr, Result);
1801  }
1802 
1803  // If we are in C++11, lex the optional ud-suffix.
1804  if (getLangOpts().CPlusPlus)
1805  CurPtr = LexUDSuffix(Result, CurPtr, true);
1806 
1807  // If a nul character existed in the string, warn about it.
1808  if (NulCharacter && !isLexingRawMode())
1809  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1810 
1811  // Update the location of the token as well as the BufferPtr instance var.
1812  const char *TokStart = BufferPtr;
1813  FormTokenWithChars(Result, CurPtr, Kind);
1814  Result.setLiteralData(TokStart);
1815  return true;
1816 }
1817 
1818 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1819 /// having lexed R", LR", u8R", uR", or UR".
1820 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1821  tok::TokenKind Kind) {
1822  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1823  // Between the initial and final double quote characters of the raw string,
1824  // any transformations performed in phases 1 and 2 (trigraphs,
1825  // universal-character-names, and line splicing) are reverted.
1826 
1827  if (!isLexingRawMode())
1828  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1829 
1830  unsigned PrefixLen = 0;
1831 
1832  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1833  ++PrefixLen;
1834 
1835  // If the last character was not a '(', then we didn't lex a valid delimiter.
1836  if (CurPtr[PrefixLen] != '(') {
1837  if (!isLexingRawMode()) {
1838  const char *PrefixEnd = &CurPtr[PrefixLen];
1839  if (PrefixLen == 16) {
1840  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1841  } else {
1842  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1843  << StringRef(PrefixEnd, 1);
1844  }
1845  }
1846 
1847  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1848  // it's possible the '"' was intended to be part of the raw string, but
1849  // there's not much we can do about that.
1850  while (true) {
1851  char C = *CurPtr++;
1852 
1853  if (C == '"')
1854  break;
1855  if (C == 0 && CurPtr-1 == BufferEnd) {
1856  --CurPtr;
1857  break;
1858  }
1859  }
1860 
1861  FormTokenWithChars(Result, CurPtr, tok::unknown);
1862  return true;
1863  }
1864 
1865  // Save prefix and move CurPtr past it
1866  const char *Prefix = CurPtr;
1867  CurPtr += PrefixLen + 1; // skip over prefix and '('
1868 
1869  while (true) {
1870  char C = *CurPtr++;
1871 
1872  if (C == ')') {
1873  // Check for prefix match and closing quote.
1874  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1875  CurPtr += PrefixLen + 1; // skip over prefix and '"'
1876  break;
1877  }
1878  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1879  if (!isLexingRawMode())
1880  Diag(BufferPtr, diag::err_unterminated_raw_string)
1881  << StringRef(Prefix, PrefixLen);
1882  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1883  return true;
1884  }
1885  }
1886 
1887  // If we are in C++11, lex the optional ud-suffix.
1888  if (getLangOpts().CPlusPlus)
1889  CurPtr = LexUDSuffix(Result, CurPtr, true);
1890 
1891  // Update the location of token as well as BufferPtr.
1892  const char *TokStart = BufferPtr;
1893  FormTokenWithChars(Result, CurPtr, Kind);
1894  Result.setLiteralData(TokStart);
1895  return true;
1896 }
1897 
1898 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
1899 /// after having lexed the '<' character. This is used for #include filenames.
1900 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
1901  // Does this string contain the \0 character?
1902  const char *NulCharacter = nullptr;
1903  const char *AfterLessPos = CurPtr;
1904  char C = getAndAdvanceChar(CurPtr, Result);
1905  while (C != '>') {
1906  // Skip escaped characters.
1907  if (C == '\\' && CurPtr < BufferEnd) {
1908  // Skip the escaped character.
1909  getAndAdvanceChar(CurPtr, Result);
1910  } else if (C == '\n' || C == '\r' || // Newline.
1911  (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
1912  isCodeCompletionPoint(CurPtr-1)))) {
1913  // If the filename is unterminated, then it must just be a lone <
1914  // character. Return this as such.
1915  FormTokenWithChars(Result, AfterLessPos, tok::less);
1916  return true;
1917  } else if (C == 0) {
1918  NulCharacter = CurPtr-1;
1919  }
1920  C = getAndAdvanceChar(CurPtr, Result);
1921  }
1922 
1923  // If a nul character existed in the string, warn about it.
1924  if (NulCharacter && !isLexingRawMode())
1925  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1926 
1927  // Update the location of token as well as BufferPtr.
1928  const char *TokStart = BufferPtr;
1929  FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
1930  Result.setLiteralData(TokStart);
1931  return true;
1932 }
1933 
1934 /// LexCharConstant - Lex the remainder of a character constant, after having
1935 /// lexed either ' or L' or u8' or u' or U'.
1936 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
1937  tok::TokenKind Kind) {
1938  // Does this character contain the \0 character?
1939  const char *NulCharacter = nullptr;
1940 
1941  if (!isLexingRawMode()) {
1942  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
1943  Diag(BufferPtr, getLangOpts().CPlusPlus
1944  ? diag::warn_cxx98_compat_unicode_literal
1945  : diag::warn_c99_compat_unicode_literal);
1946  else if (Kind == tok::utf8_char_constant)
1947  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
1948  }
1949 
1950  char C = getAndAdvanceChar(CurPtr, Result);
1951  if (C == '\'') {
1952  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1953  Diag(BufferPtr, diag::ext_empty_character);
1954  FormTokenWithChars(Result, CurPtr, tok::unknown);
1955  return true;
1956  }
1957 
1958  while (C != '\'') {
1959  // Skip escaped characters.
1960  if (C == '\\')
1961  C = getAndAdvanceChar(CurPtr, Result);
1962 
1963  if (C == '\n' || C == '\r' || // Newline.
1964  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1965  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1966  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
1967  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1968  return true;
1969  }
1970 
1971  if (C == 0) {
1972  if (isCodeCompletionPoint(CurPtr-1)) {
1974  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1975  cutOffLexing();
1976  return true;
1977  }
1978 
1979  NulCharacter = CurPtr-1;
1980  }
1981  C = getAndAdvanceChar(CurPtr, Result);
1982  }
1983 
1984  // If we are in C++11, lex the optional ud-suffix.
1985  if (getLangOpts().CPlusPlus)
1986  CurPtr = LexUDSuffix(Result, CurPtr, false);
1987 
1988  // If a nul character existed in the character, warn about it.
1989  if (NulCharacter && !isLexingRawMode())
1990  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
1991 
1992  // Update the location of token as well as BufferPtr.
1993  const char *TokStart = BufferPtr;
1994  FormTokenWithChars(Result, CurPtr, Kind);
1995  Result.setLiteralData(TokStart);
1996  return true;
1997 }
1998 
1999 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2000 /// Update BufferPtr to point to the next non-whitespace character and return.
2001 ///
2002 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2003 ///
2004 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2005  bool &TokAtPhysicalStartOfLine) {
2006  // Whitespace - Skip it, then return the token after the whitespace.
2007  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2008 
2009  unsigned char Char = *CurPtr;
2010 
2011  // Skip consecutive spaces efficiently.
2012  while (true) {
2013  // Skip horizontal whitespace very aggressively.
2014  while (isHorizontalWhitespace(Char))
2015  Char = *++CurPtr;
2016 
2017  // Otherwise if we have something other than whitespace, we're done.
2018  if (!isVerticalWhitespace(Char))
2019  break;
2020 
2022  // End of preprocessor directive line, let LexTokenInternal handle this.
2023  BufferPtr = CurPtr;
2024  return false;
2025  }
2026 
2027  // OK, but handle newline.
2028  SawNewline = true;
2029  Char = *++CurPtr;
2030  }
2031 
2032  // If the client wants us to return whitespace, return it now.
2033  if (isKeepWhitespaceMode()) {
2034  FormTokenWithChars(Result, CurPtr, tok::unknown);
2035  if (SawNewline) {
2036  IsAtStartOfLine = true;
2037  IsAtPhysicalStartOfLine = true;
2038  }
2039  // FIXME: The next token will not have LeadingSpace set.
2040  return true;
2041  }
2042 
2043  // If this isn't immediately after a newline, there is leading space.
2044  char PrevChar = CurPtr[-1];
2045  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2046 
2047  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2048  if (SawNewline) {
2049  Result.setFlag(Token::StartOfLine);
2050  TokAtPhysicalStartOfLine = true;
2051  }
2052 
2053  BufferPtr = CurPtr;
2054  return false;
2055 }
2056 
2057 /// We have just read the // characters from input. Skip until we find the
2058 /// newline character thats terminate the comment. Then update BufferPtr and
2059 /// return.
2060 ///
2061 /// If we're in KeepCommentMode or any CommentHandler has inserted
2062 /// some tokens, this will store the first token and return true.
2063 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2064  bool &TokAtPhysicalStartOfLine) {
2065  // If Line comments aren't explicitly enabled for this language, emit an
2066  // extension warning.
2067  if (!LangOpts.LineComment && !isLexingRawMode()) {
2068  Diag(BufferPtr, diag::ext_line_comment);
2069 
2070  // Mark them enabled so we only emit one warning for this translation
2071  // unit.
2072  LangOpts.LineComment = true;
2073  }
2074 
2075  // Scan over the body of the comment. The common case, when scanning, is that
2076  // the comment contains normal ascii characters with nothing interesting in
2077  // them. As such, optimize for this case with the inner loop.
2078  //
2079  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2080  // character that ends the line comment.
2081  char C;
2082  while (true) {
2083  C = *CurPtr;
2084  // Skip over characters in the fast loop.
2085  while (C != 0 && // Potentially EOF.
2086  C != '\n' && C != '\r') // Newline or DOS-style newline.
2087  C = *++CurPtr;
2088 
2089  const char *NextLine = CurPtr;
2090  if (C != 0) {
2091  // We found a newline, see if it's escaped.
2092  const char *EscapePtr = CurPtr-1;
2093  bool HasSpace = false;
2094  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2095  --EscapePtr;
2096  HasSpace = true;
2097  }
2098 
2099  if (*EscapePtr == '\\')
2100  // Escaped newline.
2101  CurPtr = EscapePtr;
2102  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2103  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2104  // Trigraph-escaped newline.
2105  CurPtr = EscapePtr-2;
2106  else
2107  break; // This is a newline, we're done.
2108 
2109  // If there was space between the backslash and newline, warn about it.
2110  if (HasSpace && !isLexingRawMode())
2111  Diag(EscapePtr, diag::backslash_newline_space);
2112  }
2113 
2114  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2115  // properly decode the character. Read it in raw mode to avoid emitting
2116  // diagnostics about things like trigraphs. If we see an escaped newline,
2117  // we'll handle it below.
2118  const char *OldPtr = CurPtr;
2119  bool OldRawMode = isLexingRawMode();
2120  LexingRawMode = true;
2121  C = getAndAdvanceChar(CurPtr, Result);
2122  LexingRawMode = OldRawMode;
2123 
2124  // If we only read only one character, then no special handling is needed.
2125  // We're done and can skip forward to the newline.
2126  if (C != 0 && CurPtr == OldPtr+1) {
2127  CurPtr = NextLine;
2128  break;
2129  }
2130 
2131  // If we read multiple characters, and one of those characters was a \r or
2132  // \n, then we had an escaped newline within the comment. Emit diagnostic
2133  // unless the next line is also a // comment.
2134  if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') {
2135  for (; OldPtr != CurPtr; ++OldPtr)
2136  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2137  // Okay, we found a // comment that ends in a newline, if the next
2138  // line is also a // comment, but has spaces, don't emit a diagnostic.
2139  if (isWhitespace(C)) {
2140  const char *ForwardPtr = CurPtr;
2141  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2142  ++ForwardPtr;
2143  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2144  break;
2145  }
2146 
2147  if (!isLexingRawMode())
2148  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2149  break;
2150  }
2151  }
2152 
2153  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2154  --CurPtr;
2155  break;
2156  }
2157 
2158  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2160  cutOffLexing();
2161  return false;
2162  }
2163  }
2164 
2165  // Found but did not consume the newline. Notify comment handlers about the
2166  // comment unless we're in a #if 0 block.
2167  if (PP && !isLexingRawMode() &&
2168  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2169  getSourceLocation(CurPtr)))) {
2170  BufferPtr = CurPtr;
2171  return true; // A token has to be returned.
2172  }
2173 
2174  // If we are returning comments as tokens, return this comment as a token.
2175  if (inKeepCommentMode())
2176  return SaveLineComment(Result, CurPtr);
2177 
2178  // If we are inside a preprocessor directive and we see the end of line,
2179  // return immediately, so that the lexer can return this as an EOD token.
2180  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2181  BufferPtr = CurPtr;
2182  return false;
2183  }
2184 
2185  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2186  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2187  // contribute to another token), it isn't needed for correctness. Note that
2188  // this is ok even in KeepWhitespaceMode, because we would have returned the
2189  /// comment above in that mode.
2190  ++CurPtr;
2191 
2192  // The next returned token is at the start of the line.
2193  Result.setFlag(Token::StartOfLine);
2194  TokAtPhysicalStartOfLine = true;
2195  // No leading whitespace seen so far.
2197  BufferPtr = CurPtr;
2198  return false;
2199 }
2200 
2201 /// If in save-comment mode, package up this Line comment in an appropriate
2202 /// way and return it.
2203 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2204  // If we're not in a preprocessor directive, just return the // comment
2205  // directly.
2206  FormTokenWithChars(Result, CurPtr, tok::comment);
2207 
2209  return true;
2210 
2211  // If this Line-style comment is in a macro definition, transmogrify it into
2212  // a C-style block comment.
2213  bool Invalid = false;
2214  std::string Spelling = PP->getSpelling(Result, &Invalid);
2215  if (Invalid)
2216  return true;
2217 
2218  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2219  Spelling[1] = '*'; // Change prefix to "/*".
2220  Spelling += "*/"; // add suffix.
2221 
2222  Result.setKind(tok::comment);
2223  PP->CreateString(Spelling, Result,
2224  Result.getLocation(), Result.getLocation());
2225  return true;
2226 }
2227 
2228 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2229 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2230 /// a diagnostic if so. We know that the newline is inside of a block comment.
2231 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2232  Lexer *L) {
2233  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2234 
2235  // Back up off the newline.
2236  --CurPtr;
2237 
2238  // If this is a two-character newline sequence, skip the other character.
2239  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2240  // \n\n or \r\r -> not escaped newline.
2241  if (CurPtr[0] == CurPtr[1])
2242  return false;
2243  // \n\r or \r\n -> skip the newline.
2244  --CurPtr;
2245  }
2246 
2247  // If we have horizontal whitespace, skip over it. We allow whitespace
2248  // between the slash and newline.
2249  bool HasSpace = false;
2250  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2251  --CurPtr;
2252  HasSpace = true;
2253  }
2254 
2255  // If we have a slash, we know this is an escaped newline.
2256  if (*CurPtr == '\\') {
2257  if (CurPtr[-1] != '*') return false;
2258  } else {
2259  // It isn't a slash, is it the ?? / trigraph?
2260  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2261  CurPtr[-3] != '*')
2262  return false;
2263 
2264  // This is the trigraph ending the comment. Emit a stern warning!
2265  CurPtr -= 2;
2266 
2267  // If no trigraphs are enabled, warn that we ignored this trigraph and
2268  // ignore this * character.
2269  if (!L->getLangOpts().Trigraphs) {
2270  if (!L->isLexingRawMode())
2271  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2272  return false;
2273  }
2274  if (!L->isLexingRawMode())
2275  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2276  }
2277 
2278  // Warn about having an escaped newline between the */ characters.
2279  if (!L->isLexingRawMode())
2280  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2281 
2282  // If there was space between the backslash and newline, warn about it.
2283  if (HasSpace && !L->isLexingRawMode())
2284  L->Diag(CurPtr, diag::backslash_newline_space);
2285 
2286  return true;
2287 }
2288 
2289 #ifdef __SSE2__
2290 #include <emmintrin.h>
2291 #elif __ALTIVEC__
2292 #include <altivec.h>
2293 #undef bool
2294 #endif
2295 
2296 /// We have just read from input the / and * characters that started a comment.
2297 /// Read until we find the * and / characters that terminate the comment.
2298 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2299 /// comments, because they cannot cause the comment to end. The only thing
2300 /// that can happen is the comment could end with an escaped newline between
2301 /// the terminating * and /.
2302 ///
2303 /// If we're in KeepCommentMode or any CommentHandler has inserted
2304 /// some tokens, this will store the first token and return true.
2305 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2306  bool &TokAtPhysicalStartOfLine) {
2307  // Scan one character past where we should, looking for a '/' character. Once
2308  // we find it, check to see if it was preceded by a *. This common
2309  // optimization helps people who like to put a lot of * characters in their
2310  // comments.
2311 
2312  // The first character we get with newlines and trigraphs skipped to handle
2313  // the degenerate /*/ case below correctly if the * has an escaped newline
2314  // after it.
2315  unsigned CharSize;
2316  unsigned char C = getCharAndSize(CurPtr, CharSize);
2317  CurPtr += CharSize;
2318  if (C == 0 && CurPtr == BufferEnd+1) {
2319  if (!isLexingRawMode())
2320  Diag(BufferPtr, diag::err_unterminated_block_comment);
2321  --CurPtr;
2322 
2323  // KeepWhitespaceMode should return this broken comment as a token. Since
2324  // it isn't a well formed comment, just return it as an 'unknown' token.
2325  if (isKeepWhitespaceMode()) {
2326  FormTokenWithChars(Result, CurPtr, tok::unknown);
2327  return true;
2328  }
2329 
2330  BufferPtr = CurPtr;
2331  return false;
2332  }
2333 
2334  // Check to see if the first character after the '/*' is another /. If so,
2335  // then this slash does not end the block comment, it is part of it.
2336  if (C == '/')
2337  C = *CurPtr++;
2338 
2339  while (true) {
2340  // Skip over all non-interesting characters until we find end of buffer or a
2341  // (probably ending) '/' character.
2342  if (CurPtr + 24 < BufferEnd &&
2343  // If there is a code-completion point avoid the fast scan because it
2344  // doesn't check for '\0'.
2345  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2346  // While not aligned to a 16-byte boundary.
2347  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2348  C = *CurPtr++;
2349 
2350  if (C == '/') goto FoundSlash;
2351 
2352 #ifdef __SSE2__
2353  __m128i Slashes = _mm_set1_epi8('/');
2354  while (CurPtr+16 <= BufferEnd) {
2355  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2356  Slashes));
2357  if (cmp != 0) {
2358  // Adjust the pointer to point directly after the first slash. It's
2359  // not necessary to set C here, it will be overwritten at the end of
2360  // the outer loop.
2361  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2362  goto FoundSlash;
2363  }
2364  CurPtr += 16;
2365  }
2366 #elif __ALTIVEC__
2367  __vector unsigned char Slashes = {
2368  '/', '/', '/', '/', '/', '/', '/', '/',
2369  '/', '/', '/', '/', '/', '/', '/', '/'
2370  };
2371  while (CurPtr+16 <= BufferEnd &&
2372  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2373  CurPtr += 16;
2374 #else
2375  // Scan for '/' quickly. Many block comments are very large.
2376  while (CurPtr[0] != '/' &&
2377  CurPtr[1] != '/' &&
2378  CurPtr[2] != '/' &&
2379  CurPtr[3] != '/' &&
2380  CurPtr+4 < BufferEnd) {
2381  CurPtr += 4;
2382  }
2383 #endif
2384 
2385  // It has to be one of the bytes scanned, increment to it and read one.
2386  C = *CurPtr++;
2387  }
2388 
2389  // Loop to scan the remainder.
2390  while (C != '/' && C != '\0')
2391  C = *CurPtr++;
2392 
2393  if (C == '/') {
2394  FoundSlash:
2395  if (CurPtr[-2] == '*') // We found the final */. We're done!
2396  break;
2397 
2398  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2399  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2400  // We found the final */, though it had an escaped newline between the
2401  // * and /. We're done!
2402  break;
2403  }
2404  }
2405  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2406  // If this is a /* inside of the comment, emit a warning. Don't do this
2407  // if this is a /*/, which will end the comment. This misses cases with
2408  // embedded escaped newlines, but oh well.
2409  if (!isLexingRawMode())
2410  Diag(CurPtr-1, diag::warn_nested_block_comment);
2411  }
2412  } else if (C == 0 && CurPtr == BufferEnd+1) {
2413  if (!isLexingRawMode())
2414  Diag(BufferPtr, diag::err_unterminated_block_comment);
2415  // Note: the user probably forgot a */. We could continue immediately
2416  // after the /*, but this would involve lexing a lot of what really is the
2417  // comment, which surely would confuse the parser.
2418  --CurPtr;
2419 
2420  // KeepWhitespaceMode should return this broken comment as a token. Since
2421  // it isn't a well formed comment, just return it as an 'unknown' token.
2422  if (isKeepWhitespaceMode()) {
2423  FormTokenWithChars(Result, CurPtr, tok::unknown);
2424  return true;
2425  }
2426 
2427  BufferPtr = CurPtr;
2428  return false;
2429  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2431  cutOffLexing();
2432  return false;
2433  }
2434 
2435  C = *CurPtr++;
2436  }
2437 
2438  // Notify comment handlers about the comment unless we're in a #if 0 block.
2439  if (PP && !isLexingRawMode() &&
2440  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2441  getSourceLocation(CurPtr)))) {
2442  BufferPtr = CurPtr;
2443  return true; // A token has to be returned.
2444  }
2445 
2446  // If we are returning comments as tokens, return this comment as a token.
2447  if (inKeepCommentMode()) {
2448  FormTokenWithChars(Result, CurPtr, tok::comment);
2449  return true;
2450  }
2451 
2452  // It is common for the tokens immediately after a /**/ comment to be
2453  // whitespace. Instead of going through the big switch, handle it
2454  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2455  // have already returned above with the comment as a token.
2456  if (isHorizontalWhitespace(*CurPtr)) {
2457  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2458  return false;
2459  }
2460 
2461  // Otherwise, just return so that the next character will be lexed as a token.
2462  BufferPtr = CurPtr;
2463  Result.setFlag(Token::LeadingSpace);
2464  return false;
2465 }
2466 
2467 //===----------------------------------------------------------------------===//
2468 // Primary Lexing Entry Points
2469 //===----------------------------------------------------------------------===//
2470 
2471 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2472 /// uninterpreted string. This switches the lexer out of directive mode.
2474  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2475  "Must be in a preprocessing directive!");
2476  Token Tmp;
2477 
2478  // CurPtr - Cache BufferPtr in an automatic variable.
2479  const char *CurPtr = BufferPtr;
2480  while (true) {
2481  char Char = getAndAdvanceChar(CurPtr, Tmp);
2482  switch (Char) {
2483  default:
2484  if (Result)
2485  Result->push_back(Char);
2486  break;
2487  case 0: // Null.
2488  // Found end of file?
2489  if (CurPtr-1 != BufferEnd) {
2490  if (isCodeCompletionPoint(CurPtr-1)) {
2492  cutOffLexing();
2493  return;
2494  }
2495 
2496  // Nope, normal character, continue.
2497  if (Result)
2498  Result->push_back(Char);
2499  break;
2500  }
2501  // FALL THROUGH.
2502  LLVM_FALLTHROUGH;
2503  case '\r':
2504  case '\n':
2505  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2506  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2507  BufferPtr = CurPtr-1;
2508 
2509  // Next, lex the character, which should handle the EOD transition.
2510  Lex(Tmp);
2511  if (Tmp.is(tok::code_completion)) {
2512  if (PP)
2514  Lex(Tmp);
2515  }
2516  assert(Tmp.is(tok::eod) && "Unexpected token!");
2517 
2518  // Finally, we're done;
2519  return;
2520  }
2521  }
2522 }
2523 
2524 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2525 /// condition, reporting diagnostics and handling other edge cases as required.
2526 /// This returns true if Result contains a token, false if PP.Lex should be
2527 /// called again.
2528 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2529  // If we hit the end of the file while parsing a preprocessor directive,
2530  // end the preprocessor directive first. The next token returned will
2531  // then be the end of file.
2533  // Done parsing the "line".
2535  // Update the location of token as well as BufferPtr.
2536  FormTokenWithChars(Result, CurPtr, tok::eod);
2537 
2538  // Restore comment saving mode, in case it was disabled for directive.
2539  if (PP)
2541  return true; // Have a token.
2542  }
2543 
2544  // If we are in raw mode, return this event as an EOF token. Let the caller
2545  // that put us in raw mode handle the event.
2546  if (isLexingRawMode()) {
2547  Result.startToken();
2548  BufferPtr = BufferEnd;
2549  FormTokenWithChars(Result, BufferEnd, tok::eof);
2550  return true;
2551  }
2552 
2553  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2555  ConditionalStack.clear();
2556  }
2557 
2558  // Issue diagnostics for unterminated #if and missing newline.
2559 
2560  // If we are in a #if directive, emit an error.
2561  while (!ConditionalStack.empty()) {
2562  if (PP->getCodeCompletionFileLoc() != FileLoc)
2563  PP->Diag(ConditionalStack.back().IfLoc,
2564  diag::err_pp_unterminated_conditional);
2565  ConditionalStack.pop_back();
2566  }
2567 
2568  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2569  // a pedwarn.
2570  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2571  DiagnosticsEngine &Diags = PP->getDiagnostics();
2572  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2573  unsigned DiagID;
2574 
2575  if (LangOpts.CPlusPlus11) {
2576  // C++11 [lex.phases] 2.2 p2
2577  // Prefer the C++98 pedantic compatibility warning over the generic,
2578  // non-extension, user-requested "missing newline at EOF" warning.
2579  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2580  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2581  } else {
2582  DiagID = diag::warn_no_newline_eof;
2583  }
2584  } else {
2585  DiagID = diag::ext_no_newline_eof;
2586  }
2587 
2588  Diag(BufferEnd, DiagID)
2589  << FixItHint::CreateInsertion(EndLoc, "\n");
2590  }
2591 
2592  BufferPtr = CurPtr;
2593 
2594  // Finally, let the preprocessor handle this.
2595  return PP->HandleEndOfFile(Result, isPragmaLexer());
2596 }
2597 
2598 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2599 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2600 /// else and 2 if there are no more tokens in the buffer controlled by the
2601 /// lexer.
2602 unsigned Lexer::isNextPPTokenLParen() {
2603  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2604 
2605  // Switch to 'skipping' mode. This will ensure that we can lex a token
2606  // without emitting diagnostics, disables macro expansion, and will cause EOF
2607  // to return an EOF token instead of popping the include stack.
2608  LexingRawMode = true;
2609 
2610  // Save state that can be changed while lexing so that we can restore it.
2611  const char *TmpBufferPtr = BufferPtr;
2612  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2613  bool atStartOfLine = IsAtStartOfLine;
2614  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2615  bool leadingSpace = HasLeadingSpace;
2616 
2617  Token Tok;
2618  Lex(Tok);
2619 
2620  // Restore state that may have changed.
2621  BufferPtr = TmpBufferPtr;
2622  ParsingPreprocessorDirective = inPPDirectiveMode;
2623  HasLeadingSpace = leadingSpace;
2624  IsAtStartOfLine = atStartOfLine;
2625  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2626 
2627  // Restore the lexer back to non-skipping mode.
2628  LexingRawMode = false;
2629 
2630  if (Tok.is(tok::eof))
2631  return 2;
2632  return Tok.is(tok::l_paren);
2633 }
2634 
2635 /// \brief Find the end of a version control conflict marker.
2636 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2637  ConflictMarkerKind CMK) {
2638  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2639  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2640  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2641  size_t Pos = RestOfBuffer.find(Terminator);
2642  while (Pos != StringRef::npos) {
2643  // Must occur at start of line.
2644  if (Pos == 0 ||
2645  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2646  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2647  Pos = RestOfBuffer.find(Terminator);
2648  continue;
2649  }
2650  return RestOfBuffer.data()+Pos;
2651  }
2652  return nullptr;
2653 }
2654 
2655 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2656 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2657 /// and recover nicely. This returns true if it is a conflict marker and false
2658 /// if not.
2659 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2660  // Only a conflict marker if it starts at the beginning of a line.
2661  if (CurPtr != BufferStart &&
2662  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2663  return false;
2664 
2665  // Check to see if we have <<<<<<< or >>>>.
2666  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2667  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2668  return false;
2669 
2670  // If we have a situation where we don't care about conflict markers, ignore
2671  // it.
2672  if (CurrentConflictMarkerState || isLexingRawMode())
2673  return false;
2674 
2675  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2676 
2677  // Check to see if there is an ending marker somewhere in the buffer at the
2678  // start of a line to terminate this conflict marker.
2679  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2680  // We found a match. We are really in a conflict marker.
2681  // Diagnose this, and ignore to the end of line.
2682  Diag(CurPtr, diag::err_conflict_marker);
2683  CurrentConflictMarkerState = Kind;
2684 
2685  // Skip ahead to the end of line. We know this exists because the
2686  // end-of-conflict marker starts with \r or \n.
2687  while (*CurPtr != '\r' && *CurPtr != '\n') {
2688  assert(CurPtr != BufferEnd && "Didn't find end of line");
2689  ++CurPtr;
2690  }
2691  BufferPtr = CurPtr;
2692  return true;
2693  }
2694 
2695  // No end of conflict marker found.
2696  return false;
2697 }
2698 
2699 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2700 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2701 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2702 /// the line. This returns true if it is a conflict marker and false if not.
2703 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2704  // Only a conflict marker if it starts at the beginning of a line.
2705  if (CurPtr != BufferStart &&
2706  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2707  return false;
2708 
2709  // If we have a situation where we don't care about conflict markers, ignore
2710  // it.
2711  if (!CurrentConflictMarkerState || isLexingRawMode())
2712  return false;
2713 
2714  // Check to see if we have the marker (4 characters in a row).
2715  for (unsigned i = 1; i != 4; ++i)
2716  if (CurPtr[i] != CurPtr[0])
2717  return false;
2718 
2719  // If we do have it, search for the end of the conflict marker. This could
2720  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2721  // be the end of conflict marker.
2722  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2723  CurrentConflictMarkerState)) {
2724  CurPtr = End;
2725 
2726  // Skip ahead to the end of line.
2727  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2728  ++CurPtr;
2729 
2730  BufferPtr = CurPtr;
2731 
2732  // No longer in the conflict marker.
2733  CurrentConflictMarkerState = CMK_None;
2734  return true;
2735  }
2736 
2737  return false;
2738 }
2739 
2740 static const char *findPlaceholderEnd(const char *CurPtr,
2741  const char *BufferEnd) {
2742  if (CurPtr == BufferEnd)
2743  return nullptr;
2744  BufferEnd -= 1; // Scan until the second last character.
2745  for (; CurPtr != BufferEnd; ++CurPtr) {
2746  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2747  return CurPtr + 2;
2748  }
2749  return nullptr;
2750 }
2751 
2752 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2753  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2755  return false;
2756  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2757  if (!End)
2758  return false;
2759  const char *Start = CurPtr - 1;
2760  if (!LangOpts.AllowEditorPlaceholders)
2761  Diag(Start, diag::err_placeholder_in_source);
2762  Result.startToken();
2763  FormTokenWithChars(Result, End, tok::raw_identifier);
2764  Result.setRawIdentifierData(Start);
2765  PP->LookUpIdentifierInfo(Result);
2767  BufferPtr = End;
2768  return true;
2769 }
2770 
2771 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2772  if (PP && PP->isCodeCompletionEnabled()) {
2773  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2774  return Loc == PP->getCodeCompletionLoc();
2775  }
2776 
2777  return false;
2778 }
2779 
2780 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2781  Token *Result) {
2782  unsigned CharSize;
2783  char Kind = getCharAndSize(StartPtr, CharSize);
2784 
2785  unsigned NumHexDigits;
2786  if (Kind == 'u')
2787  NumHexDigits = 4;
2788  else if (Kind == 'U')
2789  NumHexDigits = 8;
2790  else
2791  return 0;
2792 
2793  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2794  if (Result && !isLexingRawMode())
2795  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2796  return 0;
2797  }
2798 
2799  const char *CurPtr = StartPtr + CharSize;
2800  const char *KindLoc = &CurPtr[-1];
2801 
2802  uint32_t CodePoint = 0;
2803  for (unsigned i = 0; i < NumHexDigits; ++i) {
2804  char C = getCharAndSize(CurPtr, CharSize);
2805 
2806  unsigned Value = llvm::hexDigitValue(C);
2807  if (Value == -1U) {
2808  if (Result && !isLexingRawMode()) {
2809  if (i == 0) {
2810  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2811  << StringRef(KindLoc, 1);
2812  } else {
2813  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2814 
2815  // If the user wrote \U1234, suggest a fixit to \u.
2816  if (i == 4 && NumHexDigits == 8) {
2817  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2818  Diag(KindLoc, diag::note_ucn_four_not_eight)
2819  << FixItHint::CreateReplacement(URange, "u");
2820  }
2821  }
2822  }
2823 
2824  return 0;
2825  }
2826 
2827  CodePoint <<= 4;
2828  CodePoint += Value;
2829 
2830  CurPtr += CharSize;
2831  }
2832 
2833  if (Result) {
2834  Result->setFlag(Token::HasUCN);
2835  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
2836  StartPtr = CurPtr;
2837  else
2838  while (StartPtr != CurPtr)
2839  (void)getAndAdvanceChar(StartPtr, *Result);
2840  } else {
2841  StartPtr = CurPtr;
2842  }
2843 
2844  // Don't apply C family restrictions to UCNs in assembly mode
2845  if (LangOpts.AsmPreprocessor)
2846  return CodePoint;
2847 
2848  // C99 6.4.3p2: A universal character name shall not specify a character whose
2849  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
2850  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
2851  // C++11 [lex.charset]p2: If the hexadecimal value for a
2852  // universal-character-name corresponds to a surrogate code point (in the
2853  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
2854  // if the hexadecimal value for a universal-character-name outside the
2855  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
2856  // string literal corresponds to a control character (in either of the
2857  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
2858  // basic source character set, the program is ill-formed.
2859  if (CodePoint < 0xA0) {
2860  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
2861  return CodePoint;
2862 
2863  // We don't use isLexingRawMode() here because we need to warn about bad
2864  // UCNs even when skipping preprocessing tokens in a #if block.
2865  if (Result && PP) {
2866  if (CodePoint < 0x20 || CodePoint >= 0x7F)
2867  Diag(BufferPtr, diag::err_ucn_control_character);
2868  else {
2869  char C = static_cast<char>(CodePoint);
2870  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
2871  }
2872  }
2873 
2874  return 0;
2875 
2876  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
2877  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
2878  // We don't use isLexingRawMode() here because we need to diagnose bad
2879  // UCNs even when skipping preprocessing tokens in a #if block.
2880  if (Result && PP) {
2881  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
2882  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
2883  else
2884  Diag(BufferPtr, diag::err_ucn_escape_invalid);
2885  }
2886  return 0;
2887  }
2888 
2889  return CodePoint;
2890 }
2891 
2892 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
2893  const char *CurPtr) {
2894  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
2896  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
2897  UnicodeWhitespaceChars.contains(C)) {
2898  Diag(BufferPtr, diag::ext_unicode_whitespace)
2899  << makeCharRange(*this, BufferPtr, CurPtr);
2900 
2901  Result.setFlag(Token::LeadingSpace);
2902  return true;
2903  }
2904  return false;
2905 }
2906 
2907 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
2908  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
2910  !PP->isPreprocessedOutput()) {
2912  makeCharRange(*this, BufferPtr, CurPtr),
2913  /*IsFirst=*/true);
2914  }
2915 
2916  MIOpt.ReadToken();
2917  return LexIdentifier(Result, CurPtr);
2918  }
2919 
2921  !PP->isPreprocessedOutput() &&
2922  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
2923  // Non-ASCII characters tend to creep into source code unintentionally.
2924  // Instead of letting the parser complain about the unknown token,
2925  // just drop the character.
2926  // Note that we can /only/ do this when the non-ASCII character is actually
2927  // spelled as Unicode, not written as a UCN. The standard requires that
2928  // we not throw away any possible preprocessor tokens, but there's a
2929  // loophole in the mapping of Unicode characters to basic character set
2930  // characters that allows us to map these particular characters to, say,
2931  // whitespace.
2932  Diag(BufferPtr, diag::err_non_ascii)
2933  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
2934 
2935  BufferPtr = CurPtr;
2936  return false;
2937  }
2938 
2939  // Otherwise, we have an explicit UCN or a character that's unlikely to show
2940  // up by accident.
2941  MIOpt.ReadToken();
2942  FormTokenWithChars(Result, CurPtr, tok::unknown);
2943  return true;
2944 }
2945 
2946 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
2947  IsAtStartOfLine = Result.isAtStartOfLine();
2948  HasLeadingSpace = Result.hasLeadingSpace();
2949  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
2950  // Note that this doesn't affect IsAtPhysicalStartOfLine.
2951 }
2952 
2953 bool Lexer::Lex(Token &Result) {
2954  // Start a new token.
2955  Result.startToken();
2956 
2957  // Set up misc whitespace flags for LexTokenInternal.
2958  if (IsAtStartOfLine) {
2959  Result.setFlag(Token::StartOfLine);
2960  IsAtStartOfLine = false;
2961  }
2962 
2963  if (HasLeadingSpace) {
2964  Result.setFlag(Token::LeadingSpace);
2965  HasLeadingSpace = false;
2966  }
2967 
2968  if (HasLeadingEmptyMacro) {
2970  HasLeadingEmptyMacro = false;
2971  }
2972 
2973  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2974  IsAtPhysicalStartOfLine = false;
2975  bool isRawLex = isLexingRawMode();
2976  (void) isRawLex;
2977  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
2978  // (After the LexTokenInternal call, the lexer might be destroyed.)
2979  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
2980  return returnedToken;
2981 }
2982 
2983 /// LexTokenInternal - This implements a simple C family lexer. It is an
2984 /// extremely performance critical piece of code. This assumes that the buffer
2985 /// has a null character at the end of the file. This returns a preprocessing
2986 /// token, not a normal token, as such, it is an internal interface. It assumes
2987 /// that the Flags of result have been cleared before calling this.
2988 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
2989 LexNextToken:
2990  // New token, can't need cleaning yet.
2992  Result.setIdentifierInfo(nullptr);
2993 
2994  // CurPtr - Cache BufferPtr in an automatic variable.
2995  const char *CurPtr = BufferPtr;
2996 
2997  // Small amounts of horizontal whitespace is very common between tokens.
2998  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
2999  ++CurPtr;
3000  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3001  ++CurPtr;
3002 
3003  // If we are keeping whitespace and other tokens, just return what we just
3004  // skipped. The next lexer invocation will return the token after the
3005  // whitespace.
3006  if (isKeepWhitespaceMode()) {
3007  FormTokenWithChars(Result, CurPtr, tok::unknown);
3008  // FIXME: The next token will not have LeadingSpace set.
3009  return true;
3010  }
3011 
3012  BufferPtr = CurPtr;
3013  Result.setFlag(Token::LeadingSpace);
3014  }
3015 
3016  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3017 
3018  // Read a character, advancing over it.
3019  char Char = getAndAdvanceChar(CurPtr, Result);
3021 
3022  switch (Char) {
3023  case 0: // Null.
3024  // Found end of file?
3025  if (CurPtr-1 == BufferEnd)
3026  return LexEndOfFile(Result, CurPtr-1);
3027 
3028  // Check if we are performing code completion.
3029  if (isCodeCompletionPoint(CurPtr-1)) {
3030  // Return the code-completion token.
3031  Result.startToken();
3032  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3033  return true;
3034  }
3035 
3036  if (!isLexingRawMode())
3037  Diag(CurPtr-1, diag::null_in_file);
3038  Result.setFlag(Token::LeadingSpace);
3039  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3040  return true; // KeepWhitespaceMode
3041 
3042  // We know the lexer hasn't changed, so just try again with this lexer.
3043  // (We manually eliminate the tail call to avoid recursion.)
3044  goto LexNextToken;
3045 
3046  case 26: // DOS & CP/M EOF: "^Z".
3047  // If we're in Microsoft extensions mode, treat this as end of file.
3048  if (LangOpts.MicrosoftExt) {
3049  if (!isLexingRawMode())
3050  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3051  return LexEndOfFile(Result, CurPtr-1);
3052  }
3053 
3054  // If Microsoft extensions are disabled, this is just random garbage.
3055  Kind = tok::unknown;
3056  break;
3057 
3058  case '\n':
3059  case '\r':
3060  // If we are inside a preprocessor directive and we see the end of line,
3061  // we know we are done with the directive, so return an EOD token.
3063  // Done parsing the "line".
3065 
3066  // Restore comment saving mode, in case it was disabled for directive.
3067  if (PP)
3069 
3070  // Since we consumed a newline, we are back at the start of a line.
3071  IsAtStartOfLine = true;
3072  IsAtPhysicalStartOfLine = true;
3073 
3074  Kind = tok::eod;
3075  break;
3076  }
3077 
3078  // No leading whitespace seen so far.
3080 
3081  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3082  return true; // KeepWhitespaceMode
3083 
3084  // We only saw whitespace, so just try again with this lexer.
3085  // (We manually eliminate the tail call to avoid recursion.)
3086  goto LexNextToken;
3087  case ' ':
3088  case '\t':
3089  case '\f':
3090  case '\v':
3091  SkipHorizontalWhitespace:
3092  Result.setFlag(Token::LeadingSpace);
3093  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3094  return true; // KeepWhitespaceMode
3095 
3096  SkipIgnoredUnits:
3097  CurPtr = BufferPtr;
3098 
3099  // If the next token is obviously a // or /* */ comment, skip it efficiently
3100  // too (without going through the big switch stmt).
3101  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3102  LangOpts.LineComment &&
3103  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3104  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3105  return true; // There is a token to return.
3106  goto SkipIgnoredUnits;
3107  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3108  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3109  return true; // There is a token to return.
3110  goto SkipIgnoredUnits;
3111  } else if (isHorizontalWhitespace(*CurPtr)) {
3112  goto SkipHorizontalWhitespace;
3113  }
3114  // We only saw whitespace, so just try again with this lexer.
3115  // (We manually eliminate the tail call to avoid recursion.)
3116  goto LexNextToken;
3117 
3118  // C99 6.4.4.1: Integer Constants.
3119  // C99 6.4.4.2: Floating Constants.
3120  case '0': case '1': case '2': case '3': case '4':
3121  case '5': case '6': case '7': case '8': case '9':
3122  // Notify MIOpt that we read a non-whitespace/non-comment token.
3123  MIOpt.ReadToken();
3124  return LexNumericConstant(Result, CurPtr);
3125 
3126  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3127  // Notify MIOpt that we read a non-whitespace/non-comment token.
3128  MIOpt.ReadToken();
3129 
3130  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3131  Char = getCharAndSize(CurPtr, SizeTmp);
3132 
3133  // UTF-16 string literal
3134  if (Char == '"')
3135  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3136  tok::utf16_string_literal);
3137 
3138  // UTF-16 character constant
3139  if (Char == '\'')
3140  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3141  tok::utf16_char_constant);
3142 
3143  // UTF-16 raw string literal
3144  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3145  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3146  return LexRawStringLiteral(Result,
3147  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3148  SizeTmp2, Result),
3149  tok::utf16_string_literal);
3150 
3151  if (Char == '8') {
3152  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3153 
3154  // UTF-8 string literal
3155  if (Char2 == '"')
3156  return LexStringLiteral(Result,
3157  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3158  SizeTmp2, Result),
3159  tok::utf8_string_literal);
3160  if (Char2 == '\'' && LangOpts.CPlusPlus1z)
3161  return LexCharConstant(
3162  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3163  SizeTmp2, Result),
3164  tok::utf8_char_constant);
3165 
3166  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3167  unsigned SizeTmp3;
3168  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3169  // UTF-8 raw string literal
3170  if (Char3 == '"') {
3171  return LexRawStringLiteral(Result,
3172  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3173  SizeTmp2, Result),
3174  SizeTmp3, Result),
3175  tok::utf8_string_literal);
3176  }
3177  }
3178  }
3179  }
3180 
3181  // treat u like the start of an identifier.
3182  return LexIdentifier(Result, CurPtr);
3183 
3184  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3185  // Notify MIOpt that we read a non-whitespace/non-comment token.
3186  MIOpt.ReadToken();
3187 
3188  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3189  Char = getCharAndSize(CurPtr, SizeTmp);
3190 
3191  // UTF-32 string literal
3192  if (Char == '"')
3193  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3194  tok::utf32_string_literal);
3195 
3196  // UTF-32 character constant
3197  if (Char == '\'')
3198  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3199  tok::utf32_char_constant);
3200 
3201  // UTF-32 raw string literal
3202  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3203  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3204  return LexRawStringLiteral(Result,
3205  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3206  SizeTmp2, Result),
3207  tok::utf32_string_literal);
3208  }
3209 
3210  // treat U like the start of an identifier.
3211  return LexIdentifier(Result, CurPtr);
3212 
3213  case 'R': // Identifier or C++0x raw string literal
3214  // Notify MIOpt that we read a non-whitespace/non-comment token.
3215  MIOpt.ReadToken();
3216 
3217  if (LangOpts.CPlusPlus11) {
3218  Char = getCharAndSize(CurPtr, SizeTmp);
3219 
3220  if (Char == '"')
3221  return LexRawStringLiteral(Result,
3222  ConsumeChar(CurPtr, SizeTmp, Result),
3223  tok::string_literal);
3224  }
3225 
3226  // treat R like the start of an identifier.
3227  return LexIdentifier(Result, CurPtr);
3228 
3229  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3230  // Notify MIOpt that we read a non-whitespace/non-comment token.
3231  MIOpt.ReadToken();
3232  Char = getCharAndSize(CurPtr, SizeTmp);
3233 
3234  // Wide string literal.
3235  if (Char == '"')
3236  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3237  tok::wide_string_literal);
3238 
3239  // Wide raw string literal.
3240  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3241  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3242  return LexRawStringLiteral(Result,
3243  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3244  SizeTmp2, Result),
3245  tok::wide_string_literal);
3246 
3247  // Wide character constant.
3248  if (Char == '\'')
3249  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3250  tok::wide_char_constant);
3251  // FALL THROUGH, treating L like the start of an identifier.
3252  LLVM_FALLTHROUGH;
3253 
3254  // C99 6.4.2: Identifiers.
3255  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3256  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3257  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3258  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3259  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3260  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3261  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3262  case 'v': case 'w': case 'x': case 'y': case 'z':
3263  case '_':
3264  // Notify MIOpt that we read a non-whitespace/non-comment token.
3265  MIOpt.ReadToken();
3266  return LexIdentifier(Result, CurPtr);
3267 
3268  case '$': // $ in identifiers.
3269  if (LangOpts.DollarIdents) {
3270  if (!isLexingRawMode())
3271  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3272  // Notify MIOpt that we read a non-whitespace/non-comment token.
3273  MIOpt.ReadToken();
3274  return LexIdentifier(Result, CurPtr);
3275  }
3276 
3277  Kind = tok::unknown;
3278  break;
3279 
3280  // C99 6.4.4: Character Constants.
3281  case '\'':
3282  // Notify MIOpt that we read a non-whitespace/non-comment token.
3283  MIOpt.ReadToken();
3284  return LexCharConstant(Result, CurPtr, tok::char_constant);
3285 
3286  // C99 6.4.5: String Literals.
3287  case '"':
3288  // Notify MIOpt that we read a non-whitespace/non-comment token.
3289  MIOpt.ReadToken();
3290  return LexStringLiteral(Result, CurPtr, tok::string_literal);
3291 
3292  // C99 6.4.6: Punctuators.
3293  case '?':
3294  Kind = tok::question;
3295  break;
3296  case '[':
3297  Kind = tok::l_square;
3298  break;
3299  case ']':
3300  Kind = tok::r_square;
3301  break;
3302  case '(':
3303  Kind = tok::l_paren;
3304  break;
3305  case ')':
3306  Kind = tok::r_paren;
3307  break;
3308  case '{':
3309  Kind = tok::l_brace;
3310  break;
3311  case '}':
3312  Kind = tok::r_brace;
3313  break;
3314  case '.':
3315  Char = getCharAndSize(CurPtr, SizeTmp);
3316  if (Char >= '0' && Char <= '9') {
3317  // Notify MIOpt that we read a non-whitespace/non-comment token.
3318  MIOpt.ReadToken();
3319 
3320  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3321  } else if (LangOpts.CPlusPlus && Char == '*') {
3322  Kind = tok::periodstar;
3323  CurPtr += SizeTmp;
3324  } else if (Char == '.' &&
3325  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3326  Kind = tok::ellipsis;
3327  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3328  SizeTmp2, Result);
3329  } else {
3330  Kind = tok::period;
3331  }
3332  break;
3333  case '&':
3334  Char = getCharAndSize(CurPtr, SizeTmp);
3335  if (Char == '&') {
3336  Kind = tok::ampamp;
3337  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3338  } else if (Char == '=') {
3339  Kind = tok::ampequal;
3340  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3341  } else {
3342  Kind = tok::amp;
3343  }
3344  break;
3345  case '*':
3346  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3347  Kind = tok::starequal;
3348  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3349  } else {
3350  Kind = tok::star;
3351  }
3352  break;
3353  case '+':
3354  Char = getCharAndSize(CurPtr, SizeTmp);
3355  if (Char == '+') {
3356  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3357  Kind = tok::plusplus;
3358  } else if (Char == '=') {
3359  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3360  Kind = tok::plusequal;
3361  } else {
3362  Kind = tok::plus;
3363  }
3364  break;
3365  case '-':
3366  Char = getCharAndSize(CurPtr, SizeTmp);
3367  if (Char == '-') { // --
3368  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3369  Kind = tok::minusminus;
3370  } else if (Char == '>' && LangOpts.CPlusPlus &&
3371  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3372  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3373  SizeTmp2, Result);
3374  Kind = tok::arrowstar;
3375  } else if (Char == '>') { // ->
3376  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3377  Kind = tok::arrow;
3378  } else if (Char == '=') { // -=
3379  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3380  Kind = tok::minusequal;
3381  } else {
3382  Kind = tok::minus;
3383  }
3384  break;
3385  case '~':
3386  Kind = tok::tilde;
3387  break;
3388  case '!':
3389  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3390  Kind = tok::exclaimequal;
3391  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3392  } else {
3393  Kind = tok::exclaim;
3394  }
3395  break;
3396  case '/':
3397  // 6.4.9: Comments
3398  Char = getCharAndSize(CurPtr, SizeTmp);
3399  if (Char == '/') { // Line comment.
3400  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3401  // want to lex this as a comment. There is one problem with this though,
3402  // that in one particular corner case, this can change the behavior of the
3403  // resultant program. For example, In "foo //**/ bar", C89 would lex
3404  // this as "foo / bar" and langauges with Line comments would lex it as
3405  // "foo". Check to see if the character after the second slash is a '*'.
3406  // If so, we will lex that as a "/" instead of the start of a comment.
3407  // However, we never do this if we are just preprocessing.
3408  bool TreatAsComment = LangOpts.LineComment &&
3409  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3410  if (!TreatAsComment)
3411  if (!(PP && PP->isPreprocessedOutput()))
3412  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3413 
3414  if (TreatAsComment) {
3415  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3416  TokAtPhysicalStartOfLine))
3417  return true; // There is a token to return.
3418 
3419  // It is common for the tokens immediately after a // comment to be
3420  // whitespace (indentation for the next line). Instead of going through
3421  // the big switch, handle it efficiently now.
3422  goto SkipIgnoredUnits;
3423  }
3424  }
3425 
3426  if (Char == '*') { // /**/ comment.
3427  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3428  TokAtPhysicalStartOfLine))
3429  return true; // There is a token to return.
3430 
3431  // We only saw whitespace, so just try again with this lexer.
3432  // (We manually eliminate the tail call to avoid recursion.)
3433  goto LexNextToken;
3434  }
3435 
3436  if (Char == '=') {
3437  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3438  Kind = tok::slashequal;
3439  } else {
3440  Kind = tok::slash;
3441  }
3442  break;
3443  case '%':
3444  Char = getCharAndSize(CurPtr, SizeTmp);
3445  if (Char == '=') {
3446  Kind = tok::percentequal;
3447  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3448  } else if (LangOpts.Digraphs && Char == '>') {
3449  Kind = tok::r_brace; // '%>' -> '}'
3450  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3451  } else if (LangOpts.Digraphs && Char == ':') {
3452  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3453  Char = getCharAndSize(CurPtr, SizeTmp);
3454  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3455  Kind = tok::hashhash; // '%:%:' -> '##'
3456  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3457  SizeTmp2, Result);
3458  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3459  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3460  if (!isLexingRawMode())
3461  Diag(BufferPtr, diag::ext_charize_microsoft);
3462  Kind = tok::hashat;
3463  } else { // '%:' -> '#'
3464  // We parsed a # character. If this occurs at the start of the line,
3465  // it's actually the start of a preprocessing directive. Callback to
3466  // the preprocessor to handle it.
3467  // TODO: -fpreprocessed mode??
3468  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3469  goto HandleDirective;
3470 
3471  Kind = tok::hash;
3472  }
3473  } else {
3474  Kind = tok::percent;
3475  }
3476  break;
3477  case '<':
3478  Char = getCharAndSize(CurPtr, SizeTmp);
3479  if (ParsingFilename) {
3480  return LexAngledStringLiteral(Result, CurPtr);
3481  } else if (Char == '<') {
3482  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3483  if (After == '=') {
3484  Kind = tok::lesslessequal;
3485  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3486  SizeTmp2, Result);
3487  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3488  // If this is actually a '<<<<<<<' version control conflict marker,
3489  // recognize it as such and recover nicely.
3490  goto LexNextToken;
3491  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3492  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3493  // ignore it.
3494  goto LexNextToken;
3495  } else if (LangOpts.CUDA && After == '<') {
3496  Kind = tok::lesslessless;
3497  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3498  SizeTmp2, Result);
3499  } else {
3500  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3501  Kind = tok::lessless;
3502  }
3503  } else if (Char == '=') {
3504  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3505  Kind = tok::lessequal;
3506  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3507  if (LangOpts.CPlusPlus11 &&
3508  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3509  // C++0x [lex.pptoken]p3:
3510  // Otherwise, if the next three characters are <:: and the subsequent
3511  // character is neither : nor >, the < is treated as a preprocessor
3512  // token by itself and not as the first character of the alternative
3513  // token <:.
3514  unsigned SizeTmp3;
3515  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3516  if (After != ':' && After != '>') {
3517  Kind = tok::less;
3518  if (!isLexingRawMode())
3519  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3520  break;
3521  }
3522  }
3523 
3524  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3525  Kind = tok::l_square;
3526  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3527  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3528  Kind = tok::l_brace;
3529  } else if (Char == '#' && lexEditorPlaceholder(Result, CurPtr)) {
3530  return true;
3531  } else {
3532  Kind = tok::less;
3533  }
3534  break;
3535  case '>':
3536  Char = getCharAndSize(CurPtr, SizeTmp);
3537  if (Char == '=') {
3538  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3539  Kind = tok::greaterequal;
3540  } else if (Char == '>') {
3541  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3542  if (After == '=') {
3543  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3544  SizeTmp2, Result);
3545  Kind = tok::greatergreaterequal;
3546  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3547  // If this is actually a '>>>>' conflict marker, recognize it as such
3548  // and recover nicely.
3549  goto LexNextToken;
3550  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3551  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3552  goto LexNextToken;
3553  } else if (LangOpts.CUDA && After == '>') {
3554  Kind = tok::greatergreatergreater;
3555  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3556  SizeTmp2, Result);
3557  } else {
3558  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3559  Kind = tok::greatergreater;
3560  }
3561  } else {
3562  Kind = tok::greater;
3563  }
3564  break;
3565  case '^':
3566  Char = getCharAndSize(CurPtr, SizeTmp);
3567  if (Char == '=') {
3568  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3569  Kind = tok::caretequal;
3570  } else if (LangOpts.OpenCL && Char == '^') {
3571  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3572  Kind = tok::caretcaret;
3573  } else {
3574  Kind = tok::caret;
3575  }
3576  break;
3577  case '|':
3578  Char = getCharAndSize(CurPtr, SizeTmp);
3579  if (Char == '=') {
3580  Kind = tok::pipeequal;
3581  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3582  } else if (Char == '|') {
3583  // If this is '|||||||' and we're in a conflict marker, ignore it.
3584  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3585  goto LexNextToken;
3586  Kind = tok::pipepipe;
3587  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3588  } else {
3589  Kind = tok::pipe;
3590  }
3591  break;
3592  case ':':
3593  Char = getCharAndSize(CurPtr, SizeTmp);
3594  if (LangOpts.Digraphs && Char == '>') {
3595  Kind = tok::r_square; // ':>' -> ']'
3596  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3597  } else if (LangOpts.CPlusPlus && Char == ':') {
3598  Kind = tok::coloncolon;
3599  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3600  } else {
3601  Kind = tok::colon;
3602  }
3603  break;
3604  case ';':
3605  Kind = tok::semi;
3606  break;
3607  case '=':
3608  Char = getCharAndSize(CurPtr, SizeTmp);
3609  if (Char == '=') {
3610  // If this is '====' and we're in a conflict marker, ignore it.
3611  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3612  goto LexNextToken;
3613 
3614  Kind = tok::equalequal;
3615  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3616  } else {
3617  Kind = tok::equal;
3618  }
3619  break;
3620  case ',':
3621  Kind = tok::comma;
3622  break;
3623  case '#':
3624  Char = getCharAndSize(CurPtr, SizeTmp);
3625  if (Char == '#') {
3626  Kind = tok::hashhash;
3627  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3628  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3629  Kind = tok::hashat;
3630  if (!isLexingRawMode())
3631  Diag(BufferPtr, diag::ext_charize_microsoft);
3632  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3633  } else {
3634  // We parsed a # character. If this occurs at the start of the line,
3635  // it's actually the start of a preprocessing directive. Callback to
3636  // the preprocessor to handle it.
3637  // TODO: -fpreprocessed mode??
3638  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3639  goto HandleDirective;
3640 
3641  Kind = tok::hash;
3642  }
3643  break;
3644 
3645  case '@':
3646  // Objective C support.
3647  if (CurPtr[-1] == '@' && LangOpts.ObjC1)
3648  Kind = tok::at;
3649  else
3650  Kind = tok::unknown;
3651  break;
3652 
3653  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3654  case '\\':
3655  if (!LangOpts.AsmPreprocessor) {
3656  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3657  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3658  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3659  return true; // KeepWhitespaceMode
3660 
3661  // We only saw whitespace, so just try again with this lexer.
3662  // (We manually eliminate the tail call to avoid recursion.)
3663  goto LexNextToken;
3664  }
3665 
3666  return LexUnicode(Result, CodePoint, CurPtr);
3667  }
3668  }
3669 
3670  Kind = tok::unknown;
3671  break;
3672 
3673  default: {
3674  if (isASCII(Char)) {
3675  Kind = tok::unknown;
3676  break;
3677  }
3678 
3679  llvm::UTF32 CodePoint;
3680 
3681  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3682  // an escaped newline.
3683  --CurPtr;
3684  llvm::ConversionResult Status =
3685  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3686  (const llvm::UTF8 *)BufferEnd,
3687  &CodePoint,
3688  llvm::strictConversion);
3689  if (Status == llvm::conversionOK) {
3690  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3691  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3692  return true; // KeepWhitespaceMode
3693 
3694  // We only saw whitespace, so just try again with this lexer.
3695  // (We manually eliminate the tail call to avoid recursion.)
3696  goto LexNextToken;
3697  }
3698  return LexUnicode(Result, CodePoint, CurPtr);
3699  }
3700 
3702  PP->isPreprocessedOutput()) {
3703  ++CurPtr;
3704  Kind = tok::unknown;
3705  break;
3706  }
3707 
3708  // Non-ASCII characters tend to creep into source code unintentionally.
3709  // Instead of letting the parser complain about the unknown token,
3710  // just diagnose the invalid UTF-8, then drop the character.
3711  Diag(CurPtr, diag::err_invalid_utf8);
3712 
3713  BufferPtr = CurPtr+1;
3714  // We're pretending the character didn't exist, so just try again with
3715  // this lexer.
3716  // (We manually eliminate the tail call to avoid recursion.)
3717  goto LexNextToken;
3718  }
3719  }
3720 
3721  // Notify MIOpt that we read a non-whitespace/non-comment token.
3722  MIOpt.ReadToken();
3723 
3724  // Update the location of token as well as BufferPtr.
3725  FormTokenWithChars(Result, CurPtr, Kind);
3726  return true;
3727 
3728 HandleDirective:
3729  // We parsed a # character and it's the start of a preprocessing directive.
3730 
3731  FormTokenWithChars(Result, CurPtr, tok::hash);
3732  PP->HandleDirective(Result);
3733 
3735  // With a fatal failure in the module loader, we abort parsing.
3736  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3737  return true;
3738  }
3739 
3740  // We parsed the directive; lex a token with the new state.
3741  return false;
3742 }
bool isAtStartOfLine() const
isAtStartOfLine - Return true if this token is at the start of a line.
Definition: Token.h:266
SourceManager & getSourceManager() const
Definition: Preprocessor.h:729
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:55
bool isMacroArgExpansion(SourceLocation Loc, SourceLocation *StartLoc=nullptr) const
Tests whether the given source location represents a macro argument's expansion into the function-lik...
static unsigned getSpelling(const Token &Tok, const char *&Buffer, const SourceManager &SourceMgr, const LangOptions &LangOpts, bool *Invalid=nullptr)
getSpelling - This method is used to get the spelling of a token into a preallocated buffer...
Definition: Lexer.cpp:370
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens...
Definition: Lexer.h:46
This is a discriminated union of FileInfo and ExpansionInfo.
unsigned Length
SourceLocation getBegin() const
bool isMacroID() const
static std::pair< unsigned, bool > ComputePreamble(StringRef Buffer, const LangOptions &LangOpts, unsigned MaxLines=0)
Compute the preamble of the given file.
Definition: Lexer.cpp:559
void setFlagValue(TokenFlags Flag, bool Val)
Set a flag to either true or false.
Definition: Token.h:257
static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[]
SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
static LLVM_READONLY bool isWhitespace(unsigned char c)
Return true if this character is horizontal or vertical ASCII whitespace: ' ', '\t', '\f', '\v', '\n', '\r'.
Definition: CharInfo.h:88
void setBegin(SourceLocation b)
SourceLocation getSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
static __inline__ int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:14868
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:1393
StringRef P
llvm::MemoryBuffer * getBuffer(FileID FID, SourceLocation Loc, bool *Invalid=nullptr) const
Return the buffer for the specified FileID.
bool hasLeadingSpace() const
Return true if this token has whitespace before it.
Definition: Token.h:270
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded...
const ExpansionInfo & getExpansion() const
std::unique_ptr< llvm::MemoryBuffer > Buffer
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1205
bool hasUCN() const
Returns true if this token contains a universal character name.
Definition: Token.h:294
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:283
static char getCharAndSizeNoWarn(const char *Ptr, unsigned &Size, const LangOptions &LangOpts)
getCharAndSizeNoWarn - Like the getCharAndSize method, but does not ever emit a warning.
Definition: Lexer.h:468
static bool isAtStartOfMacroExpansion(SourceLocation loc, const SourceManager &SM, const LangOptions &LangOpts, SourceLocation *MacroBegin=nullptr)
Returns true if the given MacroID location points at the first token of the macro expansion...
Definition: Lexer.cpp:784
static LLVM_READNONE bool isASCII(char c)
Returns true if this is an ASCII character.
Definition: CharInfo.h:43
bool isStringLiteral(TokenKind K)
Return true if this is a C or C++ string-literal (or C++11 user-defined-string-literal) token...
Definition: TokenKinds.h:79
ConflictMarkerKind
ConflictMarkerKind - Kinds of conflict marker which the lexer might be recovering from...
Definition: Lexer.h:31
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the lexer buffer was all exp...
Definition: Lexer.cpp:1066
Like System, but searched after the system directories.
bool isInPrimaryFile() const
Return true if we're in the top-level file, not in a #include.
static Lexer * Create_PragmaLexer(SourceLocation SpellingLoc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLen, Preprocessor &PP)
Create_PragmaLexer: Lexer constructor - Create a new lexer object for _Pragma expansion.
Definition: Lexer.cpp:177
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
Return the 'spelling' of the token at the given location; does not go up to the spelling location or ...
static LLVM_READONLY bool isPreprocessingNumberBody(unsigned char c)
Return true if this is the body character of a C preprocessing number, which is [a-zA-Z0-9_.
Definition: CharInfo.h:148
One of these records is kept for each identifier that is lexed.
static StringRef getIndentationForLine(SourceLocation Loc, const SourceManager &SM)
Returns the leading whitespace for line that corresponds to the given location Loc.
Definition: Lexer.cpp:1035
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)
Returns true if this character is horizontal ASCII whitespace: ' ', '\t', '\f', '\v'.
Definition: CharInfo.h:71
bool isFileID() const
SmallVector< PPConditionalInfo, 4 > ConditionalStack
Information about the set of #if/#ifdef/#ifndef blocks we are currently in.
Token - This structure provides full information about a lexed token.
Definition: Token.h:35
void setKind(tok::TokenKind K)
Definition: Token.h: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()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:133
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s...
Definition: Lexer.h:39
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file...
Definition: Lexer.h:221
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi8(char __b)
Initializes all values in a 128-bit vector of [16 x i8] with the specified 8-bit value.
Definition: emmintrin.h:3831
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:483
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:769
tok::TokenKind getTokenID() const
If this is a source-language token (e.g.
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:428
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:147
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:754
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
__INTPTR_TYPE__ intptr_t
A signed integer type with the property that any valid pointer to void can be converted to this type...
Definition: opencl-c.h:75
const FileID FID
The SourceManager FileID corresponding to the file being lexed.
static SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, unsigned Character, const SourceManager &SM, const LangOptions &LangOpts)
AdvanceToTokenCharacter - If the current SourceLocation specifies a location at the start of a token...
Definition: Lexer.cpp:701
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L)
isBlockCommentEndOfEscapedNewLine - Return true if the specified newline character (either \n or \r) ...
Definition: Lexer.cpp:2231
bool isInvalid() const
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
StringRef getRawIdentifier() const
getRawIdentifier - For a raw identifier token (i.e., an identifier lexed in raw mode), returns a reference to the text substring in the buffer if known.
Definition: Token.h:203
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1428
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:953
const FileEntry * getFileEntryForID(FileID FID) const
Returns the FileEntry record for the provided FileID.
bool ParsingFilename
True after #include; turns <xx> into a tok::angle_string_literal token.
FileID getFileID(SourceLocation SpellingLoc) const
Return the FileID for a SourceLocation.
static const llvm::sys::UnicodeCharRange C11DisallowedInitialIDCharRanges[]
const SmallVectorImpl< AnnotatedLine * >::const_iterator End
bool isRecordingPreamble() const
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:923
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:806
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:762
Represents a character-granular source range.
SourceLocation getEnd() const
static unsigned MeasureTokenLength(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
MeasureTokenLength - Relex the token at the specified location and return its length in bytes in the ...
Definition: Lexer.cpp:417
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:1206
Defines the clang::Preprocessor interface.
PreambleDirectiveKind
Definition: Lexer.cpp:552
SourceLocation Begin
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
Defines the clang::IdentifierInfo, clang::IdentifierTable, and clang::Selector interfaces.
static const char * findBeginningOfLine(StringRef Buffer, unsigned Offset)
Returns the pointer that points to the beginning of line that contains the given offset, or null if the offset if invalid.
Definition: Lexer.cpp:462
bool HandleEndOfFile(Token &Result, bool isEndOfMacro=false)
Callback invoked when the lexer hits the end of the current file.
SourceLocation createExpansionLoc(SourceLocation Loc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLength, int LoadedID=0, unsigned LoadedOffset=0)
Return a new SourceLocation that encodes the fact that a token from SpellingLoc should actually be re...
SourceLocation getCodeCompletionLoc() const
Returns the location of the code-completion point.
The result type of a method or function.
float __ovld __cnfn length(float p)
Return the length of vector p, i.e., sqrt(p.x2 + p.y 2 + ...)
const SourceManager & SM
Definition: Format.cpp:1293
ObjCKeywordKind
Provides a namespace for Objective-C keywords which start with an '@'.
Definition: TokenKinds.h:41
static CharSourceRange getCharRange(SourceRange R)
const char * getLiteralData() const
getLiteralData - For a literal token (numeric constant, string, etc), this returns a pointer to the s...
Definition: Token.h: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.
SmallVectorImpl< AnnotatedLine * >::const_iterator Next
Encodes a location in the source.
AnnotatedLine & Line
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:1434
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:46
void setIdentifierInfo(IdentifierInfo *II)
Definition: Token.h:186
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1110
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:531
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:832
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
bool is(tok::TokenKind K) const
is/isNot - Predicates to check if this token is a specific kind, as in "if (Tok.is(tok::l_brace)) {...
Definition: Token.h: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:732
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:722
__PTRDIFF_TYPE__ ptrdiff_t
A signed integer type that is the result of subtracting two pointers.
Definition: opencl-c.h:68
static StringRef getImmediateMacroName(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:959
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1006
static const llvm::sys::UnicodeCharRange UnicodeWhitespaceCharRanges[]
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
bool inKeepCommentMode() const
inKeepCommentMode - Return true if the lexer should return comments as tokens.
Definition: Lexer.h:180
static CharSourceRange makeFileCharRange(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Accepts a range and returns a character range with file locations.
Definition: Lexer.cpp:859
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:243
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:116
std::pair< SourceLocation, SourceLocation > getImmediateExpansionRange(SourceLocation Loc) const
Return the start/end of the expansion information for an expansion location.
SourceLocation getSourceLocation(const char *Loc, unsigned TokLen=1) const
getSourceLocation - Return a source location identifier for the specified offset in the current file...
Definition: Lexer.cpp:1091
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
detail::InMemoryDirectory::const_iterator E
SourceLocation getExpansionLocStart() const
SourceMgr(SourceMgr)
static const char * findPlaceholderEnd(const char *CurPtr, const char *BufferEnd)
Definition: Lexer.cpp:2740
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:1411
bool hasLeadingEmptyMacro() const
Return true if this token has an empty macro before it.
Definition: Token.h:287
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:90
static char GetTrigraphCharForLetter(char Letter)
GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, return the decoded trigraph...
Definition: Lexer.cpp:1120
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1031
static LLVM_READONLY bool isIdentifierBody(unsigned char c, bool AllowDollar=false)
Returns true if this is a body character of a C identifier, which is [a-zA-Z0-9_].
Definition: CharInfo.h:59
bool HandleComment(Token &Token, SourceRange Comment)
const LangOptions & getLangOpts() const
getLangOpts - Return the language features currently enabled.
Definition: Lexer.h:128
PreprocessorOptions & getPreprocessorOpts() const
Retrieve the preprocessor options used to initialize this preprocessor.
Definition: Preprocessor.h:720
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string...
Definition: Lexer.cpp:2473
Not within a conflict marker.
Definition: Lexer.h:33
static LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
Returns true if this character is vertical ASCII whitespace: '\n', '\r'.
Definition: CharInfo.h:79
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
static char DecodeTrigraphChar(const char *CP, Lexer *L)
DecodeTrigraphChar - If the specified character is a legal trigraph when prefixed with ...
Definition: Lexer.cpp:1139
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2636
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:127
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode...
Definition: Lexer.h:187
static __inline__ int __DEFAULT_FN_ATTRS _mm_movemask_epi8(__m128i __a)
Copies the values of the most significant bits from each 8-bit element in a 128-bit integer vector of...
Definition: emmintrin.h:4333
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
Given a tok::raw_identifier token, look up the identifier information for the token and install it in...
unsigned getLength() const
Definition: Token.h:127
SourceLocation getLocForStartOfFile(FileID FID) const
Return the source location corresponding to the first byte of the specified file. ...
bool isKeepWhitespaceMode() const
isKeepWhitespaceMode - Return true if the lexer should return tokens for every character in the file...
Definition: Lexer.h:166
bool isPreprocessedOutput() const
Returns true if the preprocessor is responsible for generating output, false if it is producing token...
Definition: Preprocessor.h:790
static LLVM_READONLY bool isIdentifierHead(unsigned char c, bool AllowDollar=false)
Returns true if this is a valid first character of a C identifier, which is [a-zA-Z_].
Definition: CharInfo.h:49
FormattingAttemptStatus * Status
Definition: Format.cpp:1073
A normal or diff3 conflict marker, initiated by at least 7 "<"s, separated by at least 7 "="s or "|"s...
Definition: Lexer.h:36
A trivial tuple used to represent a source range.
void clearFlag(TokenFlags Flag)
Unset the specified flag.
Definition: Token.h:244
SourceLocation getExpansionLoc(SourceLocation Loc) const
Given a SourceLocation object Loc, return the expansion location referenced by the ID...
std::pair< FileID, unsigned > getDecomposedLoc(SourceLocation Loc) const
Decompose the specified location into a raw FileID + Offset pair.
void SetKeepWhitespaceMode(bool Val)
SetKeepWhitespaceMode - This method lets clients enable or disable whitespace retention mode...
Definition: Lexer.h:172
bool isAnnotation() const
Return true if this is any of tok::annot_* kind tokens.
Definition: Token.h:118
void setRecordedPreambleConditionalStack(ArrayRef< PPConditionalInfo > s)
This class handles loading and caching of source files into memory.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding 8-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:3160
void startToken()
Reset all flags to cleared.
Definition: Token.h:169
static std::string Stringify(StringRef Str, bool Charify=false)
Stringify - Convert the specified string into a C string by escaping '\' and " characters. This does not add surrounding ""'s to the string.
Definition: Lexer.cpp:214
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:98
SourceLocation getSpellingLoc() const
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:177
tok::PPKeywordKind getPPKeywordID() const
Return the preprocessor keyword ID for this identifier.
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.