clang  7.0.0svn
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/Basic/TokenKinds.h"
25 #include "clang/Lex/Preprocessor.h"
27 #include "clang/Lex/Token.h"
28 #include "clang/Basic/Diagnostic.h"
29 #include "clang/Basic/LLVM.h"
30 #include "clang/Basic/TokenKinds.h"
31 #include "llvm/ADT/None.h"
32 #include "llvm/ADT/Optional.h"
33 #include "llvm/ADT/StringExtras.h"
34 #include "llvm/ADT/StringSwitch.h"
35 #include "llvm/ADT/StringRef.h"
36 #include "llvm/Support/Compiler.h"
37 #include "llvm/Support/ConvertUTF.h"
38 #include "llvm/Support/MathExtras.h"
39 #include "llvm/Support/MemoryBuffer.h"
40 #include "llvm/Support/NativeFormatting.h"
41 #include "llvm/Support/UnicodeCharRanges.h"
42 #include <algorithm>
43 #include <cassert>
44 #include <cstddef>
45 #include <cstdint>
46 #include <cstring>
47 #include <string>
48 #include <tuple>
49 #include <utility>
50 
51 using namespace clang;
52 
53 //===----------------------------------------------------------------------===//
54 // Token Class Implementation
55 //===----------------------------------------------------------------------===//
56 
57 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
59  if (isAnnotation())
60  return false;
62  return II->getObjCKeywordID() == objcKey;
63  return false;
64 }
65 
66 /// getObjCKeywordID - Return the ObjC keyword kind.
68  if (isAnnotation())
69  return tok::objc_not_keyword;
71  return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
72 }
73 
74 //===----------------------------------------------------------------------===//
75 // Lexer Class Implementation
76 //===----------------------------------------------------------------------===//
77 
78 void Lexer::anchor() {}
79 
80 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
81  const char *BufEnd) {
82  BufferStart = BufStart;
83  BufferPtr = BufPtr;
84  BufferEnd = BufEnd;
85 
86  assert(BufEnd[0] == 0 &&
87  "We assume that the input buffer has a null character at the end"
88  " to simplify lexing!");
89 
90  // Check whether we have a BOM in the beginning of the buffer. If yes - act
91  // accordingly. Right now we support only UTF-8 with and without BOM, so, just
92  // skip the UTF-8 BOM if it's present.
93  if (BufferStart == BufferPtr) {
94  // Determine the size of the BOM.
95  StringRef Buf(BufferStart, BufferEnd - BufferStart);
96  size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
97  .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
98  .Default(0);
99 
100  // Skip the BOM.
101  BufferPtr += BOMLength;
102  }
103 
104  Is_PragmaLexer = false;
105  CurrentConflictMarkerState = CMK_None;
106 
107  // Start of the file is a start of line.
108  IsAtStartOfLine = true;
109  IsAtPhysicalStartOfLine = true;
110 
111  HasLeadingSpace = false;
112  HasLeadingEmptyMacro = false;
113 
114  // We are not after parsing a #.
115  ParsingPreprocessorDirective = false;
116 
117  // We are not after parsing #include.
118  ParsingFilename = false;
119 
120  // We are not in raw mode. Raw mode disables diagnostics and interpretation
121  // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
122  // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
123  // or otherwise skipping over tokens.
124  LexingRawMode = false;
125 
126  // Default to not keeping comments.
127  ExtendedTokenMode = 0;
128 }
129 
130 /// Lexer constructor - Create a new lexer object for the specified buffer
131 /// with the specified preprocessor managing the lexing process. This lexer
132 /// assumes that the associated file buffer and Preprocessor objects will
133 /// outlive it, so it doesn't take ownership of either of them.
134 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
135  : PreprocessorLexer(&PP, FID),
136  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
137  LangOpts(PP.getLangOpts()) {
138  InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
139  InputFile->getBufferEnd());
140 
142 }
143 
144 /// Lexer constructor - Create a new raw lexer object. This object is only
145 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
146 /// range will outlive it, so it doesn't take ownership of it.
147 Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
148  const char *BufStart, const char *BufPtr, const char *BufEnd)
149  : FileLoc(fileloc), LangOpts(langOpts) {
150  InitLexer(BufStart, BufPtr, BufEnd);
151 
152  // We *are* in raw mode.
153  LexingRawMode = true;
154 }
155 
156 /// Lexer constructor - Create a new raw lexer object. This object is only
157 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
158 /// range will outlive it, so it doesn't take ownership of it.
159 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
160  const SourceManager &SM, const LangOptions &langOpts)
161  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile->getBufferStart(),
162  FromFile->getBufferStart(), FromFile->getBufferEnd()) {}
163 
165  assert(PP && "Cannot reset token mode without a preprocessor");
166  if (LangOpts.TraditionalCPP)
167  SetKeepWhitespaceMode(true);
168  else
170 }
171 
172 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
173 /// _Pragma expansion. This has a variety of magic semantics that this method
174 /// sets up. It returns a new'd Lexer that must be delete'd when done.
175 ///
176 /// On entrance to this routine, TokStartLoc is a macro location which has a
177 /// spelling loc that indicates the bytes to be lexed for the token and an
178 /// expansion location that indicates where all lexed tokens should be
179 /// "expanded from".
180 ///
181 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
182 /// normal lexer that remaps tokens as they fly by. This would require making
183 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
184 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
185 /// out of the critical path of the lexer!
186 ///
188  SourceLocation ExpansionLocStart,
189  SourceLocation ExpansionLocEnd,
190  unsigned TokLen, Preprocessor &PP) {
192 
193  // Create the lexer as if we were going to lex the file normally.
194  FileID SpellingFID = SM.getFileID(SpellingLoc);
195  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
196  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
197 
198  // Now that the lexer is created, change the start/end locations so that we
199  // just lex the subsection of the file that we want. This is lexing from a
200  // scratch buffer.
201  const char *StrData = SM.getCharacterData(SpellingLoc);
202 
203  L->BufferPtr = StrData;
204  L->BufferEnd = StrData+TokLen;
205  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
206 
207  // Set the SourceLocation with the remapping information. This ensures that
208  // GetMappedTokenLoc will remap the tokens as they are lexed.
209  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
210  ExpansionLocStart,
211  ExpansionLocEnd, TokLen);
212 
213  // Ensure that the lexer thinks it is inside a directive, so that end \n will
214  // return an EOD token.
216 
217  // This lexer really is for _Pragma.
218  L->Is_PragmaLexer = true;
219  return L;
220 }
221 
222 template <typename T> static void StringifyImpl(T &Str, char Quote) {
223  typename T::size_type i = 0, e = Str.size();
224  while (i < e) {
225  if (Str[i] == '\\' || Str[i] == Quote) {
226  Str.insert(Str.begin() + i, '\\');
227  i += 2;
228  ++e;
229  } else if (Str[i] == '\n' || Str[i] == '\r') {
230  // Replace '\r\n' and '\n\r' to '\\' followed by 'n'.
231  if ((i < e - 1) && (Str[i + 1] == '\n' || Str[i + 1] == '\r') &&
232  Str[i] != Str[i + 1]) {
233  Str[i] = '\\';
234  Str[i + 1] = 'n';
235  } else {
236  // Replace '\n' and '\r' to '\\' followed by 'n'.
237  Str[i] = '\\';
238  Str.insert(Str.begin() + i + 1, 'n');
239  ++e;
240  }
241  i += 2;
242  } else
243  ++i;
244  }
245 }
246 
247 std::string Lexer::Stringify(StringRef Str, bool Charify) {
248  std::string Result = Str;
249  char Quote = Charify ? '\'' : '"';
250  StringifyImpl(Result, Quote);
251  return Result;
252 }
253 
255 
256 //===----------------------------------------------------------------------===//
257 // Token Spelling
258 //===----------------------------------------------------------------------===//
259 
260 /// \brief Slow case of getSpelling. Extract the characters comprising the
261 /// spelling of this token from the provided input buffer.
262 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
263  const LangOptions &LangOpts, char *Spelling) {
264  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
265 
266  size_t Length = 0;
267  const char *BufEnd = BufPtr + Tok.getLength();
268 
269  if (tok::isStringLiteral(Tok.getKind())) {
270  // Munch the encoding-prefix and opening double-quote.
271  while (BufPtr < BufEnd) {
272  unsigned Size;
273  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
274  BufPtr += Size;
275 
276  if (Spelling[Length - 1] == '"')
277  break;
278  }
279 
280  // Raw string literals need special handling; trigraph expansion and line
281  // splicing do not occur within their d-char-sequence nor within their
282  // r-char-sequence.
283  if (Length >= 2 &&
284  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
285  // Search backwards from the end of the token to find the matching closing
286  // quote.
287  const char *RawEnd = BufEnd;
288  do --RawEnd; while (*RawEnd != '"');
289  size_t RawLength = RawEnd - BufPtr + 1;
290 
291  // Everything between the quotes is included verbatim in the spelling.
292  memcpy(Spelling + Length, BufPtr, RawLength);
293  Length += RawLength;
294  BufPtr += RawLength;
295 
296  // The rest of the token is lexed normally.
297  }
298  }
299 
300  while (BufPtr < BufEnd) {
301  unsigned Size;
302  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
303  BufPtr += Size;
304  }
305 
306  assert(Length < Tok.getLength() &&
307  "NeedsCleaning flag set on token that didn't need cleaning!");
308  return Length;
309 }
310 
311 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
312 /// token are the characters used to represent the token in the source file
313 /// after trigraph expansion and escaped-newline folding. In particular, this
314 /// wants to get the true, uncanonicalized, spelling of things like digraphs
315 /// UCNs, etc.
317  SmallVectorImpl<char> &buffer,
318  const SourceManager &SM,
319  const LangOptions &options,
320  bool *invalid) {
321  // Break down the source location.
322  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
323 
324  // Try to the load the file buffer.
325  bool invalidTemp = false;
326  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
327  if (invalidTemp) {
328  if (invalid) *invalid = true;
329  return {};
330  }
331 
332  const char *tokenBegin = file.data() + locInfo.second;
333 
334  // Lex from the start of the given location.
335  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
336  file.begin(), tokenBegin, file.end());
337  Token token;
338  lexer.LexFromRawLexer(token);
339 
340  unsigned length = token.getLength();
341 
342  // Common case: no need for cleaning.
343  if (!token.needsCleaning())
344  return StringRef(tokenBegin, length);
345 
346  // Hard case, we need to relex the characters into the string.
347  buffer.resize(length);
348  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
349  return StringRef(buffer.data(), buffer.size());
350 }
351 
352 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
353 /// token are the characters used to represent the token in the source file
354 /// after trigraph expansion and escaped-newline folding. In particular, this
355 /// wants to get the true, uncanonicalized, spelling of things like digraphs
356 /// UCNs, etc.
357 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
358  const LangOptions &LangOpts, bool *Invalid) {
359  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
360 
361  bool CharDataInvalid = false;
362  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
363  &CharDataInvalid);
364  if (Invalid)
365  *Invalid = CharDataInvalid;
366  if (CharDataInvalid)
367  return {};
368 
369  // If this token contains nothing interesting, return it directly.
370  if (!Tok.needsCleaning())
371  return std::string(TokStart, TokStart + Tok.getLength());
372 
373  std::string Result;
374  Result.resize(Tok.getLength());
375  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
376  return Result;
377 }
378 
379 /// getSpelling - This method is used to get the spelling of a token into a
380 /// preallocated buffer, instead of as an std::string. The caller is required
381 /// to allocate enough space for the token, which is guaranteed to be at least
382 /// Tok.getLength() bytes long. The actual length of the token is returned.
383 ///
384 /// Note that this method may do two possible things: it may either fill in
385 /// the buffer specified with characters, or it may *change the input pointer*
386 /// to point to a constant buffer with the data already in it (avoiding a
387 /// copy). The caller is not allowed to modify the returned buffer pointer
388 /// if an internal buffer is returned.
389 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
390  const SourceManager &SourceMgr,
391  const LangOptions &LangOpts, bool *Invalid) {
392  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
393 
394  const char *TokStart = nullptr;
395  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
396  if (Tok.is(tok::raw_identifier))
397  TokStart = Tok.getRawIdentifier().data();
398  else if (!Tok.hasUCN()) {
399  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
400  // Just return the string from the identifier table, which is very quick.
401  Buffer = II->getNameStart();
402  return II->getLength();
403  }
404  }
405 
406  // NOTE: this can be checked even after testing for an IdentifierInfo.
407  if (Tok.isLiteral())
408  TokStart = Tok.getLiteralData();
409 
410  if (!TokStart) {
411  // Compute the start of the token in the input lexer buffer.
412  bool CharDataInvalid = false;
413  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
414  if (Invalid)
415  *Invalid = CharDataInvalid;
416  if (CharDataInvalid) {
417  Buffer = "";
418  return 0;
419  }
420  }
421 
422  // If this token contains nothing interesting, return it directly.
423  if (!Tok.needsCleaning()) {
424  Buffer = TokStart;
425  return Tok.getLength();
426  }
427 
428  // Otherwise, hard case, relex the characters into the string.
429  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
430 }
431 
432 /// MeasureTokenLength - Relex the token at the specified location and return
433 /// its length in bytes in the input file. If the token needs cleaning (e.g.
434 /// includes a trigraph or an escaped newline) then this count includes bytes
435 /// that are part of that.
437  const SourceManager &SM,
438  const LangOptions &LangOpts) {
439  Token TheTok;
440  if (getRawToken(Loc, TheTok, SM, LangOpts))
441  return 0;
442  return TheTok.getLength();
443 }
444 
445 /// \brief Relex the token at the specified location.
446 /// \returns true if there was a failure, false on success.
448  const SourceManager &SM,
449  const LangOptions &LangOpts,
450  bool IgnoreWhiteSpace) {
451  // TODO: this could be special cased for common tokens like identifiers, ')',
452  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
453  // all obviously single-char tokens. This could use
454  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
455  // something.
456 
457  // If this comes from a macro expansion, we really do want the macro name, not
458  // the token this macro expanded to.
459  Loc = SM.getExpansionLoc(Loc);
460  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
461  bool Invalid = false;
462  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
463  if (Invalid)
464  return true;
465 
466  const char *StrData = Buffer.data()+LocInfo.second;
467 
468  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
469  return true;
470 
471  // Create a lexer starting at the beginning of this token.
472  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
473  Buffer.begin(), StrData, Buffer.end());
474  TheLexer.SetCommentRetentionState(true);
475  TheLexer.LexFromRawLexer(Result);
476  return false;
477 }
478 
479 /// Returns the pointer that points to the beginning of line that contains
480 /// the given offset, or null if the offset if invalid.
481 static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
482  const char *BufStart = Buffer.data();
483  if (Offset >= Buffer.size())
484  return nullptr;
485 
486  const char *LexStart = BufStart + Offset;
487  for (; LexStart != BufStart; --LexStart) {
488  if (isVerticalWhitespace(LexStart[0]) &&
489  !Lexer::isNewLineEscaped(BufStart, LexStart)) {
490  // LexStart should point at first character of logical line.
491  ++LexStart;
492  break;
493  }
494  }
495  return LexStart;
496 }
497 
499  const SourceManager &SM,
500  const LangOptions &LangOpts) {
501  assert(Loc.isFileID());
502  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
503  if (LocInfo.first.isInvalid())
504  return Loc;
505 
506  bool Invalid = false;
507  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
508  if (Invalid)
509  return Loc;
510 
511  // Back up from the current location until we hit the beginning of a line
512  // (or the buffer). We'll relex from that point.
513  const char *StrData = Buffer.data() + LocInfo.second;
514  const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
515  if (!LexStart || LexStart == StrData)
516  return Loc;
517 
518  // Create a lexer starting at the beginning of this token.
519  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
520  Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
521  Buffer.end());
522  TheLexer.SetCommentRetentionState(true);
523 
524  // Lex tokens until we find the token that contains the source location.
525  Token TheTok;
526  do {
527  TheLexer.LexFromRawLexer(TheTok);
528 
529  if (TheLexer.getBufferLocation() > StrData) {
530  // Lexing this token has taken the lexer past the source location we're
531  // looking for. If the current token encompasses our source location,
532  // return the beginning of that token.
533  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
534  return TheTok.getLocation();
535 
536  // We ended up skipping over the source location entirely, which means
537  // that it points into whitespace. We're done here.
538  break;
539  }
540  } while (TheTok.getKind() != tok::eof);
541 
542  // We've passed our source location; just return the original source location.
543  return Loc;
544 }
545 
547  const SourceManager &SM,
548  const LangOptions &LangOpts) {
549  if (Loc.isFileID())
550  return getBeginningOfFileToken(Loc, SM, LangOpts);
551 
552  if (!SM.isMacroArgExpansion(Loc))
553  return Loc;
554 
555  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
556  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
557  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
558  std::pair<FileID, unsigned> BeginFileLocInfo =
559  SM.getDecomposedLoc(BeginFileLoc);
560  assert(FileLocInfo.first == BeginFileLocInfo.first &&
561  FileLocInfo.second >= BeginFileLocInfo.second);
562  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
563 }
564 
565 namespace {
566 
568  PDK_Skipped,
569  PDK_Unknown
570 };
571 
572 } // namespace
573 
575  const LangOptions &LangOpts,
576  unsigned MaxLines) {
577  // Create a lexer starting at the beginning of the file. Note that we use a
578  // "fake" file source location at offset 1 so that the lexer will track our
579  // position within the file.
580  const unsigned StartOffset = 1;
581  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
582  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
583  Buffer.end());
584  TheLexer.SetCommentRetentionState(true);
585 
586  bool InPreprocessorDirective = false;
587  Token TheTok;
588  SourceLocation ActiveCommentLoc;
589 
590  unsigned MaxLineOffset = 0;
591  if (MaxLines) {
592  const char *CurPtr = Buffer.begin();
593  unsigned CurLine = 0;
594  while (CurPtr != Buffer.end()) {
595  char ch = *CurPtr++;
596  if (ch == '\n') {
597  ++CurLine;
598  if (CurLine == MaxLines)
599  break;
600  }
601  }
602  if (CurPtr != Buffer.end())
603  MaxLineOffset = CurPtr - Buffer.begin();
604  }
605 
606  do {
607  TheLexer.LexFromRawLexer(TheTok);
608 
609  if (InPreprocessorDirective) {
610  // If we've hit the end of the file, we're done.
611  if (TheTok.getKind() == tok::eof) {
612  break;
613  }
614 
615  // If we haven't hit the end of the preprocessor directive, skip this
616  // token.
617  if (!TheTok.isAtStartOfLine())
618  continue;
619 
620  // We've passed the end of the preprocessor directive, and will look
621  // at this token again below.
622  InPreprocessorDirective = false;
623  }
624 
625  // Keep track of the # of lines in the preamble.
626  if (TheTok.isAtStartOfLine()) {
627  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
628 
629  // If we were asked to limit the number of lines in the preamble,
630  // and we're about to exceed that limit, we're done.
631  if (MaxLineOffset && TokOffset >= MaxLineOffset)
632  break;
633  }
634 
635  // Comments are okay; skip over them.
636  if (TheTok.getKind() == tok::comment) {
637  if (ActiveCommentLoc.isInvalid())
638  ActiveCommentLoc = TheTok.getLocation();
639  continue;
640  }
641 
642  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
643  // This is the start of a preprocessor directive.
644  Token HashTok = TheTok;
645  InPreprocessorDirective = true;
646  ActiveCommentLoc = SourceLocation();
647 
648  // Figure out which directive this is. Since we're lexing raw tokens,
649  // we don't have an identifier table available. Instead, just look at
650  // the raw identifier to recognize and categorize preprocessor directives.
651  TheLexer.LexFromRawLexer(TheTok);
652  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
653  StringRef Keyword = TheTok.getRawIdentifier();
655  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
656  .Case("include", PDK_Skipped)
657  .Case("__include_macros", PDK_Skipped)
658  .Case("define", PDK_Skipped)
659  .Case("undef", PDK_Skipped)
660  .Case("line", PDK_Skipped)
661  .Case("error", PDK_Skipped)
662  .Case("pragma", PDK_Skipped)
663  .Case("import", PDK_Skipped)
664  .Case("include_next", PDK_Skipped)
665  .Case("warning", PDK_Skipped)
666  .Case("ident", PDK_Skipped)
667  .Case("sccs", PDK_Skipped)
668  .Case("assert", PDK_Skipped)
669  .Case("unassert", PDK_Skipped)
670  .Case("if", PDK_Skipped)
671  .Case("ifdef", PDK_Skipped)
672  .Case("ifndef", PDK_Skipped)
673  .Case("elif", PDK_Skipped)
674  .Case("else", PDK_Skipped)
675  .Case("endif", PDK_Skipped)
676  .Default(PDK_Unknown);
677 
678  switch (PDK) {
679  case PDK_Skipped:
680  continue;
681 
682  case PDK_Unknown:
683  // We don't know what this directive is; stop at the '#'.
684  break;
685  }
686  }
687 
688  // We only end up here if we didn't recognize the preprocessor
689  // directive or it was one that can't occur in the preamble at this
690  // point. Roll back the current token to the location of the '#'.
691  InPreprocessorDirective = false;
692  TheTok = HashTok;
693  }
694 
695  // We hit a token that we don't recognize as being in the
696  // "preprocessing only" part of the file, so we're no longer in
697  // the preamble.
698  break;
699  } while (true);
700 
702  if (ActiveCommentLoc.isValid())
703  End = ActiveCommentLoc; // don't truncate a decl comment.
704  else
705  End = TheTok.getLocation();
706 
707  return PreambleBounds(End.getRawEncoding() - FileLoc.getRawEncoding(),
708  TheTok.isAtStartOfLine());
709 }
710 
711 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
712 /// token, return a new location that specifies a character within the token.
714  unsigned CharNo,
715  const SourceManager &SM,
716  const LangOptions &LangOpts) {
717  // Figure out how many physical characters away the specified expansion
718  // character is. This needs to take into consideration newlines and
719  // trigraphs.
720  bool Invalid = false;
721  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
722 
723  // If they request the first char of the token, we're trivially done.
724  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
725  return TokStart;
726 
727  unsigned PhysOffset = 0;
728 
729  // The usual case is that tokens don't contain anything interesting. Skip
730  // over the uninteresting characters. If a token only consists of simple
731  // chars, this method is extremely fast.
732  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
733  if (CharNo == 0)
734  return TokStart.getLocWithOffset(PhysOffset);
735  ++TokPtr;
736  --CharNo;
737  ++PhysOffset;
738  }
739 
740  // If we have a character that may be a trigraph or escaped newline, use a
741  // lexer to parse it correctly.
742  for (; CharNo; --CharNo) {
743  unsigned Size;
744  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
745  TokPtr += Size;
746  PhysOffset += Size;
747  }
748 
749  // Final detail: if we end up on an escaped newline, we want to return the
750  // location of the actual byte of the token. For example foo<newline>bar
751  // advanced by 3 should return the location of b, not of \\. One compounding
752  // detail of this is that the escape may be made by a trigraph.
753  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
754  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
755 
756  return TokStart.getLocWithOffset(PhysOffset);
757 }
758 
759 /// \brief Computes the source location just past the end of the
760 /// token at this source location.
761 ///
762 /// This routine can be used to produce a source location that
763 /// points just past the end of the token referenced by \p Loc, and
764 /// is generally used when a diagnostic needs to point just after a
765 /// token where it expected something different that it received. If
766 /// the returned source location would not be meaningful (e.g., if
767 /// it points into a macro), this routine returns an invalid
768 /// source location.
769 ///
770 /// \param Offset an offset from the end of the token, where the source
771 /// location should refer to. The default offset (0) produces a source
772 /// location pointing just past the end of the token; an offset of 1 produces
773 /// a source location pointing to the last character in the token, etc.
775  const SourceManager &SM,
776  const LangOptions &LangOpts) {
777  if (Loc.isInvalid())
778  return {};
779 
780  if (Loc.isMacroID()) {
781  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
782  return {}; // Points inside the macro expansion.
783  }
784 
785  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
786  if (Len > Offset)
787  Len = Len - Offset;
788  else
789  return Loc;
790 
791  return Loc.getLocWithOffset(Len);
792 }
793 
794 /// \brief Returns true if the given MacroID location points at the first
795 /// token of the macro expansion.
797  const SourceManager &SM,
798  const LangOptions &LangOpts,
799  SourceLocation *MacroBegin) {
800  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
801 
802  SourceLocation expansionLoc;
803  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
804  return false;
805 
806  if (expansionLoc.isFileID()) {
807  // No other macro expansions, this is the first.
808  if (MacroBegin)
809  *MacroBegin = expansionLoc;
810  return true;
811  }
812 
813  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
814 }
815 
816 /// \brief Returns true if the given MacroID location points at the last
817 /// token of the macro expansion.
819  const SourceManager &SM,
820  const LangOptions &LangOpts,
821  SourceLocation *MacroEnd) {
822  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
823 
824  SourceLocation spellLoc = SM.getSpellingLoc(loc);
825  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
826  if (tokLen == 0)
827  return false;
828 
829  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
830  SourceLocation expansionLoc;
831  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
832  return false;
833 
834  if (expansionLoc.isFileID()) {
835  // No other macro expansions.
836  if (MacroEnd)
837  *MacroEnd = expansionLoc;
838  return true;
839  }
840 
841  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
842 }
843 
845  const SourceManager &SM,
846  const LangOptions &LangOpts) {
847  SourceLocation Begin = Range.getBegin();
848  SourceLocation End = Range.getEnd();
849  assert(Begin.isFileID() && End.isFileID());
850  if (Range.isTokenRange()) {
851  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
852  if (End.isInvalid())
853  return {};
854  }
855 
856  // Break down the source locations.
857  FileID FID;
858  unsigned BeginOffs;
859  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
860  if (FID.isInvalid())
861  return {};
862 
863  unsigned EndOffs;
864  if (!SM.isInFileID(End, FID, &EndOffs) ||
865  BeginOffs > EndOffs)
866  return {};
867 
868  return CharSourceRange::getCharRange(Begin, End);
869 }
870 
872  const SourceManager &SM,
873  const LangOptions &LangOpts) {
874  SourceLocation Begin = Range.getBegin();
875  SourceLocation End = Range.getEnd();
876  if (Begin.isInvalid() || End.isInvalid())
877  return {};
878 
879  if (Begin.isFileID() && End.isFileID())
880  return makeRangeFromFileLocs(Range, SM, LangOpts);
881 
882  if (Begin.isMacroID() && End.isFileID()) {
883  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
884  return {};
885  Range.setBegin(Begin);
886  return makeRangeFromFileLocs(Range, SM, LangOpts);
887  }
888 
889  if (Begin.isFileID() && End.isMacroID()) {
890  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
891  &End)) ||
892  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
893  &End)))
894  return {};
895  Range.setEnd(End);
896  return makeRangeFromFileLocs(Range, SM, LangOpts);
897  }
898 
899  assert(Begin.isMacroID() && End.isMacroID());
900  SourceLocation MacroBegin, MacroEnd;
901  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
902  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
903  &MacroEnd)) ||
904  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
905  &MacroEnd)))) {
906  Range.setBegin(MacroBegin);
907  Range.setEnd(MacroEnd);
908  return makeRangeFromFileLocs(Range, SM, LangOpts);
909  }
910 
911  bool Invalid = false;
912  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
913  &Invalid);
914  if (Invalid)
915  return {};
916 
917  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
918  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
919  &Invalid);
920  if (Invalid)
921  return {};
922 
923  if (EndEntry.getExpansion().isMacroArgExpansion() &&
924  BeginEntry.getExpansion().getExpansionLocStart() ==
925  EndEntry.getExpansion().getExpansionLocStart()) {
926  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
927  Range.setEnd(SM.getImmediateSpellingLoc(End));
928  return makeFileCharRange(Range, SM, LangOpts);
929  }
930  }
931 
932  return {};
933 }
934 
936  const SourceManager &SM,
937  const LangOptions &LangOpts,
938  bool *Invalid) {
939  Range = makeFileCharRange(Range, SM, LangOpts);
940  if (Range.isInvalid()) {
941  if (Invalid) *Invalid = true;
942  return {};
943  }
944 
945  // Break down the source location.
946  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
947  if (beginInfo.first.isInvalid()) {
948  if (Invalid) *Invalid = true;
949  return {};
950  }
951 
952  unsigned EndOffs;
953  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
954  beginInfo.second > EndOffs) {
955  if (Invalid) *Invalid = true;
956  return {};
957  }
958 
959  // Try to the load the file buffer.
960  bool invalidTemp = false;
961  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
962  if (invalidTemp) {
963  if (Invalid) *Invalid = true;
964  return {};
965  }
966 
967  if (Invalid) *Invalid = false;
968  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
969 }
970 
972  const SourceManager &SM,
973  const LangOptions &LangOpts) {
974  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
975 
976  // Find the location of the immediate macro expansion.
977  while (true) {
978  FileID FID = SM.getFileID(Loc);
979  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
980  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
981  Loc = Expansion.getExpansionLocStart();
982  if (!Expansion.isMacroArgExpansion())
983  break;
984 
985  // For macro arguments we need to check that the argument did not come
986  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
987 
988  // Loc points to the argument id of the macro definition, move to the
989  // macro expansion.
990  Loc = SM.getImmediateExpansionRange(Loc).first;
991  SourceLocation SpellLoc = Expansion.getSpellingLoc();
992  if (SpellLoc.isFileID())
993  break; // No inner macro.
994 
995  // If spelling location resides in the same FileID as macro expansion
996  // location, it means there is no inner macro.
997  FileID MacroFID = SM.getFileID(Loc);
998  if (SM.isInFileID(SpellLoc, MacroFID))
999  break;
1000 
1001  // Argument came from inner macro.
1002  Loc = SpellLoc;
1003  }
1004 
1005  // Find the spelling location of the start of the non-argument expansion
1006  // range. This is where the macro name was spelled in order to begin
1007  // expanding this macro.
1008  Loc = SM.getSpellingLoc(Loc);
1009 
1010  // Dig out the buffer where the macro name was spelled and the extents of the
1011  // name so that we can render it into the expansion note.
1012  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1013  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1014  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1015  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1016 }
1017 
1019  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1020  assert(Loc.isMacroID() && "Only reasonble to call this on macros");
1021  // Walk past macro argument expanions.
1022  while (SM.isMacroArgExpansion(Loc))
1023  Loc = SM.getImmediateExpansionRange(Loc).first;
1024 
1025  // If the macro's spelling has no FileID, then it's actually a token paste
1026  // or stringization (or similar) and not a macro at all.
1027  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1028  return {};
1029 
1030  // Find the spelling location of the start of the non-argument expansion
1031  // range. This is where the macro name was spelled in order to begin
1032  // expanding this macro.
1033  Loc = SM.getSpellingLoc(SM.getImmediateExpansionRange(Loc).first);
1034 
1035  // Dig out the buffer where the macro name was spelled and the extents of the
1036  // name so that we can render it into the expansion note.
1037  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1038  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1039  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1040  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1041 }
1042 
1043 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1044  return isIdentifierBody(c, LangOpts.DollarIdents);
1045 }
1046 
1047 bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1048  assert(isVerticalWhitespace(Str[0]));
1049  if (Str - 1 < BufferStart)
1050  return false;
1051 
1052  if ((Str[0] == '\n' && Str[-1] == '\r') ||
1053  (Str[0] == '\r' && Str[-1] == '\n')) {
1054  if (Str - 2 < BufferStart)
1055  return false;
1056  --Str;
1057  }
1058  --Str;
1059 
1060  // Rewind to first non-space character:
1061  while (Str > BufferStart && isHorizontalWhitespace(*Str))
1062  --Str;
1063 
1064  return *Str == '\\';
1065 }
1066 
1068  const SourceManager &SM) {
1069  if (Loc.isInvalid() || Loc.isMacroID())
1070  return {};
1071  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1072  if (LocInfo.first.isInvalid())
1073  return {};
1074  bool Invalid = false;
1075  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1076  if (Invalid)
1077  return {};
1078  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1079  if (!Line)
1080  return {};
1081  StringRef Rest = Buffer.substr(Line - Buffer.data());
1082  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1083  return NumWhitespaceChars == StringRef::npos
1084  ? ""
1085  : Rest.take_front(NumWhitespaceChars);
1086 }
1087 
1088 //===----------------------------------------------------------------------===//
1089 // Diagnostics forwarding code.
1090 //===----------------------------------------------------------------------===//
1091 
1092 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1093 /// lexer buffer was all expanded at a single point, perform the mapping.
1094 /// This is currently only used for _Pragma implementation, so it is the slow
1095 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1096 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1097  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1099  SourceLocation FileLoc,
1100  unsigned CharNo, unsigned TokLen) {
1101  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1102 
1103  // Otherwise, we're lexing "mapped tokens". This is used for things like
1104  // _Pragma handling. Combine the expansion location of FileLoc with the
1105  // spelling location.
1107 
1108  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1109  // characters come from spelling(FileLoc)+Offset.
1110  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1111  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1112 
1113  // Figure out the expansion loc range, which is the range covered by the
1114  // original _Pragma(...) sequence.
1115  std::pair<SourceLocation,SourceLocation> II =
1116  SM.getImmediateExpansionRange(FileLoc);
1117 
1118  return SM.createExpansionLoc(SpellingLoc, II.first, II.second, TokLen);
1119 }
1120 
1121 /// getSourceLocation - Return a source location identifier for the specified
1122 /// offset in the current file.
1124  unsigned TokLen) const {
1125  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1126  "Location out of range for this buffer!");
1127 
1128  // In the normal case, we're just lexing from a simple file buffer, return
1129  // the file id from FileLoc with the offset specified.
1130  unsigned CharNo = Loc-BufferStart;
1131  if (FileLoc.isFileID())
1132  return FileLoc.getLocWithOffset(CharNo);
1133 
1134  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1135  // tokens are lexed from where the _Pragma was defined.
1136  assert(PP && "This doesn't work on raw lexers");
1137  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1138 }
1139 
1140 /// Diag - Forwarding function for diagnostics. This translate a source
1141 /// position in the current buffer into a SourceLocation object for rendering.
1142 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1143  return PP->Diag(getSourceLocation(Loc), DiagID);
1144 }
1145 
1146 //===----------------------------------------------------------------------===//
1147 // Trigraph and Escaped Newline Handling Code.
1148 //===----------------------------------------------------------------------===//
1149 
1150 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1151 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1152 static char GetTrigraphCharForLetter(char Letter) {
1153  switch (Letter) {
1154  default: return 0;
1155  case '=': return '#';
1156  case ')': return ']';
1157  case '(': return '[';
1158  case '!': return '|';
1159  case '\'': return '^';
1160  case '>': return '}';
1161  case '/': return '\\';
1162  case '<': return '{';
1163  case '-': return '~';
1164  }
1165 }
1166 
1167 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1168 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1169 /// return the result character. Finally, emit a warning about trigraph use
1170 /// whether trigraphs are enabled or not.
1171 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1172  char Res = GetTrigraphCharForLetter(*CP);
1173  if (!Res || !L) return Res;
1174 
1175  if (!L->getLangOpts().Trigraphs) {
1176  if (!L->isLexingRawMode())
1177  L->Diag(CP-2, diag::trigraph_ignored);
1178  return 0;
1179  }
1180 
1181  if (!L->isLexingRawMode())
1182  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1183  return Res;
1184 }
1185 
1186 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1187 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1188 /// trigraph equivalent on entry to this function.
1189 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1190  unsigned Size = 0;
1191  while (isWhitespace(Ptr[Size])) {
1192  ++Size;
1193 
1194  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1195  continue;
1196 
1197  // If this is a \r\n or \n\r, skip the other half.
1198  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1199  Ptr[Size-1] != Ptr[Size])
1200  ++Size;
1201 
1202  return Size;
1203  }
1204 
1205  // Not an escaped newline, must be a \t or something else.
1206  return 0;
1207 }
1208 
1209 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1210 /// them), skip over them and return the first non-escaped-newline found,
1211 /// otherwise return P.
1212 const char *Lexer::SkipEscapedNewLines(const char *P) {
1213  while (true) {
1214  const char *AfterEscape;
1215  if (*P == '\\') {
1216  AfterEscape = P+1;
1217  } else if (*P == '?') {
1218  // If not a trigraph for escape, bail out.
1219  if (P[1] != '?' || P[2] != '/')
1220  return P;
1221  // FIXME: Take LangOpts into account; the language might not
1222  // support trigraphs.
1223  AfterEscape = P+3;
1224  } else {
1225  return P;
1226  }
1227 
1228  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1229  if (NewLineSize == 0) return P;
1230  P = AfterEscape+NewLineSize;
1231  }
1232 }
1233 
1235  const SourceManager &SM,
1236  const LangOptions &LangOpts) {
1237  if (Loc.isMacroID()) {
1238  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1239  return None;
1240  }
1241  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1242 
1243  // Break down the source location.
1244  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1245 
1246  // Try to load the file buffer.
1247  bool InvalidTemp = false;
1248  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1249  if (InvalidTemp)
1250  return None;
1251 
1252  const char *TokenBegin = File.data() + LocInfo.second;
1253 
1254  // Lex from the start of the given location.
1255  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1256  TokenBegin, File.end());
1257  // Find the token.
1258  Token Tok;
1259  lexer.LexFromRawLexer(Tok);
1260  return Tok;
1261 }
1262 
1263 /// \brief Checks that the given token is the first token that occurs after the
1264 /// given location (this excludes comments and whitespace). Returns the location
1265 /// immediately after the specified token. If the token is not found or the
1266 /// location is inside a macro, the returned source location will be invalid.
1268  SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1269  const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1270  Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1271  if (!Tok || Tok->isNot(TKind))
1272  return {};
1273  SourceLocation TokenLoc = Tok->getLocation();
1274 
1275  // Calculate how much whitespace needs to be skipped if any.
1276  unsigned NumWhitespaceChars = 0;
1277  if (SkipTrailingWhitespaceAndNewLine) {
1278  const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1279  unsigned char C = *TokenEnd;
1280  while (isHorizontalWhitespace(C)) {
1281  C = *(++TokenEnd);
1282  NumWhitespaceChars++;
1283  }
1284 
1285  // Skip \r, \n, \r\n, or \n\r
1286  if (C == '\n' || C == '\r') {
1287  char PrevC = C;
1288  C = *(++TokenEnd);
1289  NumWhitespaceChars++;
1290  if ((C == '\n' || C == '\r') && C != PrevC)
1291  NumWhitespaceChars++;
1292  }
1293  }
1294 
1295  return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1296 }
1297 
1298 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1299 /// get its size, and return it. This is tricky in several cases:
1300 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1301 /// then either return the trigraph (skipping 3 chars) or the '?',
1302 /// depending on whether trigraphs are enabled or not.
1303 /// 2. If this is an escaped newline (potentially with whitespace between
1304 /// the backslash and newline), implicitly skip the newline and return
1305 /// the char after it.
1306 ///
1307 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1308 /// know that we can accumulate into Size, and that we have already incremented
1309 /// Ptr by Size bytes.
1310 ///
1311 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1312 /// be updated to match.
1313 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1314  Token *Tok) {
1315  // If we have a slash, look for an escaped newline.
1316  if (Ptr[0] == '\\') {
1317  ++Size;
1318  ++Ptr;
1319 Slash:
1320  // Common case, backslash-char where the char is not whitespace.
1321  if (!isWhitespace(Ptr[0])) return '\\';
1322 
1323  // See if we have optional whitespace characters between the slash and
1324  // newline.
1325  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1326  // Remember that this token needs to be cleaned.
1327  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1328 
1329  // Warn if there was whitespace between the backslash and newline.
1330  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1331  Diag(Ptr, diag::backslash_newline_space);
1332 
1333  // Found backslash<whitespace><newline>. Parse the char after it.
1334  Size += EscapedNewLineSize;
1335  Ptr += EscapedNewLineSize;
1336 
1337  // Use slow version to accumulate a correct size field.
1338  return getCharAndSizeSlow(Ptr, Size, Tok);
1339  }
1340 
1341  // Otherwise, this is not an escaped newline, just return the slash.
1342  return '\\';
1343  }
1344 
1345  // If this is a trigraph, process it.
1346  if (Ptr[0] == '?' && Ptr[1] == '?') {
1347  // If this is actually a legal trigraph (not something like "??x"), emit
1348  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1349  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1350  // Remember that this token needs to be cleaned.
1351  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1352 
1353  Ptr += 3;
1354  Size += 3;
1355  if (C == '\\') goto Slash;
1356  return C;
1357  }
1358  }
1359 
1360  // If this is neither, return a single character.
1361  ++Size;
1362  return *Ptr;
1363 }
1364 
1365 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1366 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1367 /// and that we have already incremented Ptr by Size bytes.
1368 ///
1369 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1370 /// be updated to match.
1371 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1372  const LangOptions &LangOpts) {
1373  // If we have a slash, look for an escaped newline.
1374  if (Ptr[0] == '\\') {
1375  ++Size;
1376  ++Ptr;
1377 Slash:
1378  // Common case, backslash-char where the char is not whitespace.
1379  if (!isWhitespace(Ptr[0])) return '\\';
1380 
1381  // See if we have optional whitespace characters followed by a newline.
1382  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1383  // Found backslash<whitespace><newline>. Parse the char after it.
1384  Size += EscapedNewLineSize;
1385  Ptr += EscapedNewLineSize;
1386 
1387  // Use slow version to accumulate a correct size field.
1388  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1389  }
1390 
1391  // Otherwise, this is not an escaped newline, just return the slash.
1392  return '\\';
1393  }
1394 
1395  // If this is a trigraph, process it.
1396  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1397  // If this is actually a legal trigraph (not something like "??x"), return
1398  // it.
1399  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1400  Ptr += 3;
1401  Size += 3;
1402  if (C == '\\') goto Slash;
1403  return C;
1404  }
1405  }
1406 
1407  // If this is neither, return a single character.
1408  ++Size;
1409  return *Ptr;
1410 }
1411 
1412 //===----------------------------------------------------------------------===//
1413 // Helper methods for lexing.
1414 //===----------------------------------------------------------------------===//
1415 
1416 /// \brief Routine that indiscriminately sets the offset into the source file.
1417 void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1418  BufferPtr = BufferStart + Offset;
1419  if (BufferPtr > BufferEnd)
1420  BufferPtr = BufferEnd;
1421  // FIXME: What exactly does the StartOfLine bit mean? There are two
1422  // possible meanings for the "start" of the line: the first token on the
1423  // unexpanded line, or the first token on the expanded line.
1424  IsAtStartOfLine = StartOfLine;
1425  IsAtPhysicalStartOfLine = StartOfLine;
1426 }
1427 
1428 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1429  if (LangOpts.AsmPreprocessor) {
1430  return false;
1431  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1432  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1434  return C11AllowedIDChars.contains(C);
1435  } else if (LangOpts.CPlusPlus) {
1436  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1438  return CXX03AllowedIDChars.contains(C);
1439  } else {
1440  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1442  return C99AllowedIDChars.contains(C);
1443  }
1444 }
1445 
1446 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1447  assert(isAllowedIDChar(C, LangOpts));
1448  if (LangOpts.AsmPreprocessor) {
1449  return false;
1450  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1451  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1453  return !C11DisallowedInitialIDChars.contains(C);
1454  } else if (LangOpts.CPlusPlus) {
1455  return true;
1456  } else {
1457  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1459  return !C99DisallowedInitialIDChars.contains(C);
1460  }
1461 }
1462 
1463 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1464  const char *End) {
1466  L.getSourceLocation(End));
1467 }
1468 
1469 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1470  CharSourceRange Range, bool IsFirst) {
1471  // Check C99 compatibility.
1472  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1473  enum {
1474  CannotAppearInIdentifier = 0,
1475  CannotStartIdentifier
1476  };
1477 
1478  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1480  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1482  if (!C99AllowedIDChars.contains(C)) {
1483  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1484  << Range
1485  << CannotAppearInIdentifier;
1486  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1487  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1488  << Range
1489  << CannotStartIdentifier;
1490  }
1491  }
1492 
1493  // Check C++98 compatibility.
1494  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1495  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1497  if (!CXX03AllowedIDChars.contains(C)) {
1498  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1499  << Range;
1500  }
1501  }
1502 }
1503 
1504 /// After encountering UTF-8 character C and interpreting it as an identifier
1505 /// character, check whether it's a homoglyph for a common non-identifier
1506 /// source character that is unlikely to be an intentional identifier
1507 /// character and warn if so.
1508 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1509  CharSourceRange Range) {
1510  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1511  struct HomoglyphPair {
1512  uint32_t Character;
1513  char LooksLike;
1514  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1515  };
1516  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1517  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1518  {U'\u037e', ';'}, // GREEK QUESTION MARK
1519  {U'\u2212', '-'}, // MINUS SIGN
1520  {U'\u2215', '/'}, // DIVISION SLASH
1521  {U'\u2216', '\\'}, // SET MINUS
1522  {U'\u2217', '*'}, // ASTERISK OPERATOR
1523  {U'\u2223', '|'}, // DIVIDES
1524  {U'\u2227', '^'}, // LOGICAL AND
1525  {U'\u2236', ':'}, // RATIO
1526  {U'\u223c', '~'}, // TILDE OPERATOR
1527  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1528  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1529  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1530  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1531  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1532  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1533  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1534  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1535  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1536  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1537  {U'\uff0c', ','}, // FULLWIDTH COMMA
1538  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1539  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1540  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1541  {U'\uff1a', ':'}, // FULLWIDTH COLON
1542  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1543  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1544  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1545  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1546  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1547  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1548  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1549  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1550  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1551  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1552  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1553  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1554  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1555  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1556  {0, 0}
1557  };
1558  auto Homoglyph =
1559  std::lower_bound(std::begin(SortedHomoglyphs),
1560  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1561  if (Homoglyph->Character == C) {
1562  llvm::SmallString<5> CharBuf;
1563  {
1564  llvm::raw_svector_ostream CharOS(CharBuf);
1565  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1566  }
1567  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1568  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1569  << Range << CharBuf << LooksLikeStr;
1570  }
1571 }
1572 
1573 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1574  Token &Result) {
1575  const char *UCNPtr = CurPtr + Size;
1576  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1577  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1578  return false;
1579 
1580  if (!isLexingRawMode())
1581  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1582  makeCharRange(*this, CurPtr, UCNPtr),
1583  /*IsFirst=*/false);
1584 
1585  Result.setFlag(Token::HasUCN);
1586  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1587  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1588  CurPtr = UCNPtr;
1589  else
1590  while (CurPtr != UCNPtr)
1591  (void)getAndAdvanceChar(CurPtr, Result);
1592  return true;
1593 }
1594 
1595 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1596  const char *UnicodePtr = CurPtr;
1597  llvm::UTF32 CodePoint;
1598  llvm::ConversionResult Result =
1599  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1600  (const llvm::UTF8 *)BufferEnd,
1601  &CodePoint,
1602  llvm::strictConversion);
1603  if (Result != llvm::conversionOK ||
1604  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1605  return false;
1606 
1607  if (!isLexingRawMode()) {
1608  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1609  makeCharRange(*this, CurPtr, UnicodePtr),
1610  /*IsFirst=*/false);
1611  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1612  makeCharRange(*this, CurPtr, UnicodePtr));
1613  }
1614 
1615  CurPtr = UnicodePtr;
1616  return true;
1617 }
1618 
1619 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1620  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1621  unsigned Size;
1622  unsigned char C = *CurPtr++;
1623  while (isIdentifierBody(C))
1624  C = *CurPtr++;
1625 
1626  --CurPtr; // Back up over the skipped character.
1627 
1628  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1629  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1630  //
1631  // TODO: Could merge these checks into an InfoTable flag to make the
1632  // comparison cheaper
1633  if (isASCII(C) && C != '\\' && C != '?' &&
1634  (C != '$' || !LangOpts.DollarIdents)) {
1635 FinishIdentifier:
1636  const char *IdStart = BufferPtr;
1637  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1638  Result.setRawIdentifierData(IdStart);
1639 
1640  // If we are in raw mode, return this identifier raw. There is no need to
1641  // look up identifier information or attempt to macro expand it.
1642  if (LexingRawMode)
1643  return true;
1644 
1645  // Fill in Result.IdentifierInfo and update the token kind,
1646  // looking up the identifier in the identifier table.
1647  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1648 
1649  // Finally, now that we know we have an identifier, pass this off to the
1650  // preprocessor, which may macro expand it or something.
1651  if (II->isHandleIdentifierCase())
1652  return PP->HandleIdentifier(Result);
1653 
1654  if (II->getTokenID() == tok::identifier && isCodeCompletionPoint(CurPtr)
1655  && II->getPPKeywordID() == tok::pp_not_keyword
1656  && II->getObjCKeywordID() == tok::objc_not_keyword) {
1657  // Return the code-completion token.
1658  Result.setKind(tok::code_completion);
1659  cutOffLexing();
1660  return true;
1661  }
1662  return true;
1663  }
1664 
1665  // Otherwise, $,\,? in identifier found. Enter slower path.
1666 
1667  C = getCharAndSize(CurPtr, Size);
1668  while (true) {
1669  if (C == '$') {
1670  // If we hit a $ and they are not supported in identifiers, we are done.
1671  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1672 
1673  // Otherwise, emit a diagnostic and continue.
1674  if (!isLexingRawMode())
1675  Diag(CurPtr, diag::ext_dollar_in_identifier);
1676  CurPtr = ConsumeChar(CurPtr, Size, Result);
1677  C = getCharAndSize(CurPtr, Size);
1678  continue;
1679  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1680  C = getCharAndSize(CurPtr, Size);
1681  continue;
1682  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1683  C = getCharAndSize(CurPtr, Size);
1684  continue;
1685  } else if (!isIdentifierBody(C)) {
1686  goto FinishIdentifier;
1687  }
1688 
1689  // Otherwise, this character is good, consume it.
1690  CurPtr = ConsumeChar(CurPtr, Size, Result);
1691 
1692  C = getCharAndSize(CurPtr, Size);
1693  while (isIdentifierBody(C)) {
1694  CurPtr = ConsumeChar(CurPtr, Size, Result);
1695  C = getCharAndSize(CurPtr, Size);
1696  }
1697  }
1698 }
1699 
1700 /// isHexaLiteral - Return true if Start points to a hex constant.
1701 /// in microsoft mode (where this is supposed to be several different tokens).
1702 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1703  unsigned Size;
1704  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1705  if (C1 != '0')
1706  return false;
1707  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1708  return (C2 == 'x' || C2 == 'X');
1709 }
1710 
1711 /// LexNumericConstant - Lex the remainder of a integer or floating point
1712 /// constant. From[-1] is the first character lexed. Return the end of the
1713 /// constant.
1714 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1715  unsigned Size;
1716  char C = getCharAndSize(CurPtr, Size);
1717  char PrevCh = 0;
1718  while (isPreprocessingNumberBody(C)) {
1719  CurPtr = ConsumeChar(CurPtr, Size, Result);
1720  PrevCh = C;
1721  C = getCharAndSize(CurPtr, Size);
1722  }
1723 
1724  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1725  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1726  // If we are in Microsoft mode, don't continue if the constant is hex.
1727  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1728  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1729  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1730  }
1731 
1732  // If we have a hex FP constant, continue.
1733  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1734  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1735  // not-quite-conforming extension. Only do so if this looks like it's
1736  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1737  bool IsHexFloat = true;
1738  if (!LangOpts.C99) {
1739  if (!isHexaLiteral(BufferPtr, LangOpts))
1740  IsHexFloat = false;
1741  else if (!getLangOpts().CPlusPlus17 &&
1742  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1743  IsHexFloat = false;
1744  }
1745  if (IsHexFloat)
1746  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1747  }
1748 
1749  // If we have a digit separator, continue.
1750  if (C == '\'' && getLangOpts().CPlusPlus14) {
1751  unsigned NextSize;
1752  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1753  if (isIdentifierBody(Next)) {
1754  if (!isLexingRawMode())
1755  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1756  CurPtr = ConsumeChar(CurPtr, Size, Result);
1757  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1758  return LexNumericConstant(Result, CurPtr);
1759  }
1760  }
1761 
1762  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1763  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1764  return LexNumericConstant(Result, CurPtr);
1765  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1766  return LexNumericConstant(Result, CurPtr);
1767 
1768  // Update the location of token as well as BufferPtr.
1769  const char *TokStart = BufferPtr;
1770  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1771  Result.setLiteralData(TokStart);
1772  return true;
1773 }
1774 
1775 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1776 /// in C++11, or warn on a ud-suffix in C++98.
1777 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1778  bool IsStringLiteral) {
1779  assert(getLangOpts().CPlusPlus);
1780 
1781  // Maximally munch an identifier.
1782  unsigned Size;
1783  char C = getCharAndSize(CurPtr, Size);
1784  bool Consumed = false;
1785 
1786  if (!isIdentifierHead(C)) {
1787  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1788  Consumed = true;
1789  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1790  Consumed = true;
1791  else
1792  return CurPtr;
1793  }
1794 
1795  if (!getLangOpts().CPlusPlus11) {
1796  if (!isLexingRawMode())
1797  Diag(CurPtr,
1798  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1799  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1801  return CurPtr;
1802  }
1803 
1804  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1805  // that does not start with an underscore is ill-formed. As a conforming
1806  // extension, we treat all such suffixes as if they had whitespace before
1807  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1808  // likely to be a ud-suffix than a macro, however, and accept that.
1809  if (!Consumed) {
1810  bool IsUDSuffix = false;
1811  if (C == '_')
1812  IsUDSuffix = true;
1813  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1814  // In C++1y, we need to look ahead a few characters to see if this is a
1815  // valid suffix for a string literal or a numeric literal (this could be
1816  // the 'operator""if' defining a numeric literal operator).
1817  const unsigned MaxStandardSuffixLength = 3;
1818  char Buffer[MaxStandardSuffixLength] = { C };
1819  unsigned Consumed = Size;
1820  unsigned Chars = 1;
1821  while (true) {
1822  unsigned NextSize;
1823  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1824  getLangOpts());
1825  if (!isIdentifierBody(Next)) {
1826  // End of suffix. Check whether this is on the whitelist.
1827  const StringRef CompleteSuffix(Buffer, Chars);
1829  CompleteSuffix);
1830  break;
1831  }
1832 
1833  if (Chars == MaxStandardSuffixLength)
1834  // Too long: can't be a standard suffix.
1835  break;
1836 
1837  Buffer[Chars++] = Next;
1838  Consumed += NextSize;
1839  }
1840  }
1841 
1842  if (!IsUDSuffix) {
1843  if (!isLexingRawMode())
1844  Diag(CurPtr, getLangOpts().MSVCCompat
1845  ? diag::ext_ms_reserved_user_defined_literal
1846  : diag::ext_reserved_user_defined_literal)
1848  return CurPtr;
1849  }
1850 
1851  CurPtr = ConsumeChar(CurPtr, Size, Result);
1852  }
1853 
1854  Result.setFlag(Token::HasUDSuffix);
1855  while (true) {
1856  C = getCharAndSize(CurPtr, Size);
1857  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1858  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1859  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1860  else break;
1861  }
1862 
1863  return CurPtr;
1864 }
1865 
1866 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1867 /// either " or L" or u8" or u" or U".
1868 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1869  tok::TokenKind Kind) {
1870  // Does this string contain the \0 character?
1871  const char *NulCharacter = nullptr;
1872 
1873  if (!isLexingRawMode() &&
1874  (Kind == tok::utf8_string_literal ||
1875  Kind == tok::utf16_string_literal ||
1876  Kind == tok::utf32_string_literal))
1877  Diag(BufferPtr, getLangOpts().CPlusPlus
1878  ? diag::warn_cxx98_compat_unicode_literal
1879  : diag::warn_c99_compat_unicode_literal);
1880 
1881  char C = getAndAdvanceChar(CurPtr, Result);
1882  while (C != '"') {
1883  // Skip escaped characters. Escaped newlines will already be processed by
1884  // getAndAdvanceChar.
1885  if (C == '\\')
1886  C = getAndAdvanceChar(CurPtr, Result);
1887 
1888  if (C == '\n' || C == '\r' || // Newline.
1889  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1890  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1891  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1892  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1893  return true;
1894  }
1895 
1896  if (C == 0) {
1897  if (isCodeCompletionPoint(CurPtr-1)) {
1899  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1900  cutOffLexing();
1901  return true;
1902  }
1903 
1904  NulCharacter = CurPtr-1;
1905  }
1906  C = getAndAdvanceChar(CurPtr, Result);
1907  }
1908 
1909  // If we are in C++11, lex the optional ud-suffix.
1910  if (getLangOpts().CPlusPlus)
1911  CurPtr = LexUDSuffix(Result, CurPtr, true);
1912 
1913  // If a nul character existed in the string, warn about it.
1914  if (NulCharacter && !isLexingRawMode())
1915  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1916 
1917  // Update the location of the token as well as the BufferPtr instance var.
1918  const char *TokStart = BufferPtr;
1919  FormTokenWithChars(Result, CurPtr, Kind);
1920  Result.setLiteralData(TokStart);
1921  return true;
1922 }
1923 
1924 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1925 /// having lexed R", LR", u8R", uR", or UR".
1926 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1927  tok::TokenKind Kind) {
1928  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1929  // Between the initial and final double quote characters of the raw string,
1930  // any transformations performed in phases 1 and 2 (trigraphs,
1931  // universal-character-names, and line splicing) are reverted.
1932 
1933  if (!isLexingRawMode())
1934  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1935 
1936  unsigned PrefixLen = 0;
1937 
1938  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1939  ++PrefixLen;
1940 
1941  // If the last character was not a '(', then we didn't lex a valid delimiter.
1942  if (CurPtr[PrefixLen] != '(') {
1943  if (!isLexingRawMode()) {
1944  const char *PrefixEnd = &CurPtr[PrefixLen];
1945  if (PrefixLen == 16) {
1946  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1947  } else {
1948  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1949  << StringRef(PrefixEnd, 1);
1950  }
1951  }
1952 
1953  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1954  // it's possible the '"' was intended to be part of the raw string, but
1955  // there's not much we can do about that.
1956  while (true) {
1957  char C = *CurPtr++;
1958 
1959  if (C == '"')
1960  break;
1961  if (C == 0 && CurPtr-1 == BufferEnd) {
1962  --CurPtr;
1963  break;
1964  }
1965  }
1966 
1967  FormTokenWithChars(Result, CurPtr, tok::unknown);
1968  return true;
1969  }
1970 
1971  // Save prefix and move CurPtr past it
1972  const char *Prefix = CurPtr;
1973  CurPtr += PrefixLen + 1; // skip over prefix and '('
1974 
1975  while (true) {
1976  char C = *CurPtr++;
1977 
1978  if (C == ')') {
1979  // Check for prefix match and closing quote.
1980  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1981  CurPtr += PrefixLen + 1; // skip over prefix and '"'
1982  break;
1983  }
1984  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1985  if (!isLexingRawMode())
1986  Diag(BufferPtr, diag::err_unterminated_raw_string)
1987  << StringRef(Prefix, PrefixLen);
1988  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1989  return true;
1990  }
1991  }
1992 
1993  // If we are in C++11, lex the optional ud-suffix.
1994  if (getLangOpts().CPlusPlus)
1995  CurPtr = LexUDSuffix(Result, CurPtr, true);
1996 
1997  // Update the location of token as well as BufferPtr.
1998  const char *TokStart = BufferPtr;
1999  FormTokenWithChars(Result, CurPtr, Kind);
2000  Result.setLiteralData(TokStart);
2001  return true;
2002 }
2003 
2004 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2005 /// after having lexed the '<' character. This is used for #include filenames.
2006 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2007  // Does this string contain the \0 character?
2008  const char *NulCharacter = nullptr;
2009  const char *AfterLessPos = CurPtr;
2010  char C = getAndAdvanceChar(CurPtr, Result);
2011  while (C != '>') {
2012  // Skip escaped characters. Escaped newlines will already be processed by
2013  // getAndAdvanceChar.
2014  if (C == '\\')
2015  C = getAndAdvanceChar(CurPtr, Result);
2016 
2017  if (C == '\n' || C == '\r' || // Newline.
2018  (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
2019  isCodeCompletionPoint(CurPtr-1)))) {
2020  // If the filename is unterminated, then it must just be a lone <
2021  // character. Return this as such.
2022  FormTokenWithChars(Result, AfterLessPos, tok::less);
2023  return true;
2024  }
2025 
2026  if (C == 0) {
2027  NulCharacter = CurPtr-1;
2028  }
2029  C = getAndAdvanceChar(CurPtr, Result);
2030  }
2031 
2032  // If a nul character existed in the string, warn about it.
2033  if (NulCharacter && !isLexingRawMode())
2034  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2035 
2036  // Update the location of token as well as BufferPtr.
2037  const char *TokStart = BufferPtr;
2038  FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
2039  Result.setLiteralData(TokStart);
2040  return true;
2041 }
2042 
2043 /// LexCharConstant - Lex the remainder of a character constant, after having
2044 /// lexed either ' or L' or u8' or u' or U'.
2045 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2046  tok::TokenKind Kind) {
2047  // Does this character contain the \0 character?
2048  const char *NulCharacter = nullptr;
2049 
2050  if (!isLexingRawMode()) {
2051  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2052  Diag(BufferPtr, getLangOpts().CPlusPlus
2053  ? diag::warn_cxx98_compat_unicode_literal
2054  : diag::warn_c99_compat_unicode_literal);
2055  else if (Kind == tok::utf8_char_constant)
2056  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2057  }
2058 
2059  char C = getAndAdvanceChar(CurPtr, Result);
2060  if (C == '\'') {
2061  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2062  Diag(BufferPtr, diag::ext_empty_character);
2063  FormTokenWithChars(Result, CurPtr, tok::unknown);
2064  return true;
2065  }
2066 
2067  while (C != '\'') {
2068  // Skip escaped characters.
2069  if (C == '\\')
2070  C = getAndAdvanceChar(CurPtr, Result);
2071 
2072  if (C == '\n' || C == '\r' || // Newline.
2073  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2074  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2075  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2076  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2077  return true;
2078  }
2079 
2080  if (C == 0) {
2081  if (isCodeCompletionPoint(CurPtr-1)) {
2083  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2084  cutOffLexing();
2085  return true;
2086  }
2087 
2088  NulCharacter = CurPtr-1;
2089  }
2090  C = getAndAdvanceChar(CurPtr, Result);
2091  }
2092 
2093  // If we are in C++11, lex the optional ud-suffix.
2094  if (getLangOpts().CPlusPlus)
2095  CurPtr = LexUDSuffix(Result, CurPtr, false);
2096 
2097  // If a nul character existed in the character, warn about it.
2098  if (NulCharacter && !isLexingRawMode())
2099  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2100 
2101  // Update the location of token as well as BufferPtr.
2102  const char *TokStart = BufferPtr;
2103  FormTokenWithChars(Result, CurPtr, Kind);
2104  Result.setLiteralData(TokStart);
2105  return true;
2106 }
2107 
2108 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2109 /// Update BufferPtr to point to the next non-whitespace character and return.
2110 ///
2111 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2112 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2113  bool &TokAtPhysicalStartOfLine) {
2114  // Whitespace - Skip it, then return the token after the whitespace.
2115  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2116 
2117  unsigned char Char = *CurPtr;
2118 
2119  // Skip consecutive spaces efficiently.
2120  while (true) {
2121  // Skip horizontal whitespace very aggressively.
2122  while (isHorizontalWhitespace(Char))
2123  Char = *++CurPtr;
2124 
2125  // Otherwise if we have something other than whitespace, we're done.
2126  if (!isVerticalWhitespace(Char))
2127  break;
2128 
2130  // End of preprocessor directive line, let LexTokenInternal handle this.
2131  BufferPtr = CurPtr;
2132  return false;
2133  }
2134 
2135  // OK, but handle newline.
2136  SawNewline = true;
2137  Char = *++CurPtr;
2138  }
2139 
2140  // If the client wants us to return whitespace, return it now.
2141  if (isKeepWhitespaceMode()) {
2142  FormTokenWithChars(Result, CurPtr, tok::unknown);
2143  if (SawNewline) {
2144  IsAtStartOfLine = true;
2145  IsAtPhysicalStartOfLine = true;
2146  }
2147  // FIXME: The next token will not have LeadingSpace set.
2148  return true;
2149  }
2150 
2151  // If this isn't immediately after a newline, there is leading space.
2152  char PrevChar = CurPtr[-1];
2153  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2154 
2155  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2156  if (SawNewline) {
2157  Result.setFlag(Token::StartOfLine);
2158  TokAtPhysicalStartOfLine = true;
2159  }
2160 
2161  BufferPtr = CurPtr;
2162  return false;
2163 }
2164 
2165 /// We have just read the // characters from input. Skip until we find the
2166 /// newline character thats terminate the comment. Then update BufferPtr and
2167 /// return.
2168 ///
2169 /// If we're in KeepCommentMode or any CommentHandler has inserted
2170 /// some tokens, this will store the first token and return true.
2171 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2172  bool &TokAtPhysicalStartOfLine) {
2173  // If Line comments aren't explicitly enabled for this language, emit an
2174  // extension warning.
2175  if (!LangOpts.LineComment && !isLexingRawMode()) {
2176  Diag(BufferPtr, diag::ext_line_comment);
2177 
2178  // Mark them enabled so we only emit one warning for this translation
2179  // unit.
2180  LangOpts.LineComment = true;
2181  }
2182 
2183  // Scan over the body of the comment. The common case, when scanning, is that
2184  // the comment contains normal ascii characters with nothing interesting in
2185  // them. As such, optimize for this case with the inner loop.
2186  //
2187  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2188  // character that ends the line comment.
2189  char C;
2190  while (true) {
2191  C = *CurPtr;
2192  // Skip over characters in the fast loop.
2193  while (C != 0 && // Potentially EOF.
2194  C != '\n' && C != '\r') // Newline or DOS-style newline.
2195  C = *++CurPtr;
2196 
2197  const char *NextLine = CurPtr;
2198  if (C != 0) {
2199  // We found a newline, see if it's escaped.
2200  const char *EscapePtr = CurPtr-1;
2201  bool HasSpace = false;
2202  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2203  --EscapePtr;
2204  HasSpace = true;
2205  }
2206 
2207  if (*EscapePtr == '\\')
2208  // Escaped newline.
2209  CurPtr = EscapePtr;
2210  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2211  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2212  // Trigraph-escaped newline.
2213  CurPtr = EscapePtr-2;
2214  else
2215  break; // This is a newline, we're done.
2216 
2217  // If there was space between the backslash and newline, warn about it.
2218  if (HasSpace && !isLexingRawMode())
2219  Diag(EscapePtr, diag::backslash_newline_space);
2220  }
2221 
2222  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2223  // properly decode the character. Read it in raw mode to avoid emitting
2224  // diagnostics about things like trigraphs. If we see an escaped newline,
2225  // we'll handle it below.
2226  const char *OldPtr = CurPtr;
2227  bool OldRawMode = isLexingRawMode();
2228  LexingRawMode = true;
2229  C = getAndAdvanceChar(CurPtr, Result);
2230  LexingRawMode = OldRawMode;
2231 
2232  // If we only read only one character, then no special handling is needed.
2233  // We're done and can skip forward to the newline.
2234  if (C != 0 && CurPtr == OldPtr+1) {
2235  CurPtr = NextLine;
2236  break;
2237  }
2238 
2239  // If we read multiple characters, and one of those characters was a \r or
2240  // \n, then we had an escaped newline within the comment. Emit diagnostic
2241  // unless the next line is also a // comment.
2242  if (CurPtr != OldPtr + 1 && C != '/' &&
2243  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2244  for (; OldPtr != CurPtr; ++OldPtr)
2245  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2246  // Okay, we found a // comment that ends in a newline, if the next
2247  // line is also a // comment, but has spaces, don't emit a diagnostic.
2248  if (isWhitespace(C)) {
2249  const char *ForwardPtr = CurPtr;
2250  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2251  ++ForwardPtr;
2252  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2253  break;
2254  }
2255 
2256  if (!isLexingRawMode())
2257  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2258  break;
2259  }
2260  }
2261 
2262  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2263  --CurPtr;
2264  break;
2265  }
2266 
2267  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2269  cutOffLexing();
2270  return false;
2271  }
2272  }
2273 
2274  // Found but did not consume the newline. Notify comment handlers about the
2275  // comment unless we're in a #if 0 block.
2276  if (PP && !isLexingRawMode() &&
2277  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2278  getSourceLocation(CurPtr)))) {
2279  BufferPtr = CurPtr;
2280  return true; // A token has to be returned.
2281  }
2282 
2283  // If we are returning comments as tokens, return this comment as a token.
2284  if (inKeepCommentMode())
2285  return SaveLineComment(Result, CurPtr);
2286 
2287  // If we are inside a preprocessor directive and we see the end of line,
2288  // return immediately, so that the lexer can return this as an EOD token.
2289  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2290  BufferPtr = CurPtr;
2291  return false;
2292  }
2293 
2294  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2295  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2296  // contribute to another token), it isn't needed for correctness. Note that
2297  // this is ok even in KeepWhitespaceMode, because we would have returned the
2298  /// comment above in that mode.
2299  ++CurPtr;
2300 
2301  // The next returned token is at the start of the line.
2302  Result.setFlag(Token::StartOfLine);
2303  TokAtPhysicalStartOfLine = true;
2304  // No leading whitespace seen so far.
2306  BufferPtr = CurPtr;
2307  return false;
2308 }
2309 
2310 /// If in save-comment mode, package up this Line comment in an appropriate
2311 /// way and return it.
2312 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2313  // If we're not in a preprocessor directive, just return the // comment
2314  // directly.
2315  FormTokenWithChars(Result, CurPtr, tok::comment);
2316 
2318  return true;
2319 
2320  // If this Line-style comment is in a macro definition, transmogrify it into
2321  // a C-style block comment.
2322  bool Invalid = false;
2323  std::string Spelling = PP->getSpelling(Result, &Invalid);
2324  if (Invalid)
2325  return true;
2326 
2327  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2328  Spelling[1] = '*'; // Change prefix to "/*".
2329  Spelling += "*/"; // add suffix.
2330 
2331  Result.setKind(tok::comment);
2332  PP->CreateString(Spelling, Result,
2333  Result.getLocation(), Result.getLocation());
2334  return true;
2335 }
2336 
2337 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2338 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2339 /// a diagnostic if so. We know that the newline is inside of a block comment.
2340 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2341  Lexer *L) {
2342  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2343 
2344  // Back up off the newline.
2345  --CurPtr;
2346 
2347  // If this is a two-character newline sequence, skip the other character.
2348  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2349  // \n\n or \r\r -> not escaped newline.
2350  if (CurPtr[0] == CurPtr[1])
2351  return false;
2352  // \n\r or \r\n -> skip the newline.
2353  --CurPtr;
2354  }
2355 
2356  // If we have horizontal whitespace, skip over it. We allow whitespace
2357  // between the slash and newline.
2358  bool HasSpace = false;
2359  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2360  --CurPtr;
2361  HasSpace = true;
2362  }
2363 
2364  // If we have a slash, we know this is an escaped newline.
2365  if (*CurPtr == '\\') {
2366  if (CurPtr[-1] != '*') return false;
2367  } else {
2368  // It isn't a slash, is it the ?? / trigraph?
2369  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2370  CurPtr[-3] != '*')
2371  return false;
2372 
2373  // This is the trigraph ending the comment. Emit a stern warning!
2374  CurPtr -= 2;
2375 
2376  // If no trigraphs are enabled, warn that we ignored this trigraph and
2377  // ignore this * character.
2378  if (!L->getLangOpts().Trigraphs) {
2379  if (!L->isLexingRawMode())
2380  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2381  return false;
2382  }
2383  if (!L->isLexingRawMode())
2384  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2385  }
2386 
2387  // Warn about having an escaped newline between the */ characters.
2388  if (!L->isLexingRawMode())
2389  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2390 
2391  // If there was space between the backslash and newline, warn about it.
2392  if (HasSpace && !L->isLexingRawMode())
2393  L->Diag(CurPtr, diag::backslash_newline_space);
2394 
2395  return true;
2396 }
2397 
2398 #ifdef __SSE2__
2399 #include <emmintrin.h>
2400 #elif __ALTIVEC__
2401 #include <altivec.h>
2402 #undef bool
2403 #endif
2404 
2405 /// We have just read from input the / and * characters that started a comment.
2406 /// Read until we find the * and / characters that terminate the comment.
2407 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2408 /// comments, because they cannot cause the comment to end. The only thing
2409 /// that can happen is the comment could end with an escaped newline between
2410 /// the terminating * and /.
2411 ///
2412 /// If we're in KeepCommentMode or any CommentHandler has inserted
2413 /// some tokens, this will store the first token and return true.
2414 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2415  bool &TokAtPhysicalStartOfLine) {
2416  // Scan one character past where we should, looking for a '/' character. Once
2417  // we find it, check to see if it was preceded by a *. This common
2418  // optimization helps people who like to put a lot of * characters in their
2419  // comments.
2420 
2421  // The first character we get with newlines and trigraphs skipped to handle
2422  // the degenerate /*/ case below correctly if the * has an escaped newline
2423  // after it.
2424  unsigned CharSize;
2425  unsigned char C = getCharAndSize(CurPtr, CharSize);
2426  CurPtr += CharSize;
2427  if (C == 0 && CurPtr == BufferEnd+1) {
2428  if (!isLexingRawMode())
2429  Diag(BufferPtr, diag::err_unterminated_block_comment);
2430  --CurPtr;
2431 
2432  // KeepWhitespaceMode should return this broken comment as a token. Since
2433  // it isn't a well formed comment, just return it as an 'unknown' token.
2434  if (isKeepWhitespaceMode()) {
2435  FormTokenWithChars(Result, CurPtr, tok::unknown);
2436  return true;
2437  }
2438 
2439  BufferPtr = CurPtr;
2440  return false;
2441  }
2442 
2443  // Check to see if the first character after the '/*' is another /. If so,
2444  // then this slash does not end the block comment, it is part of it.
2445  if (C == '/')
2446  C = *CurPtr++;
2447 
2448  while (true) {
2449  // Skip over all non-interesting characters until we find end of buffer or a
2450  // (probably ending) '/' character.
2451  if (CurPtr + 24 < BufferEnd &&
2452  // If there is a code-completion point avoid the fast scan because it
2453  // doesn't check for '\0'.
2454  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2455  // While not aligned to a 16-byte boundary.
2456  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2457  C = *CurPtr++;
2458 
2459  if (C == '/') goto FoundSlash;
2460 
2461 #ifdef __SSE2__
2462  __m128i Slashes = _mm_set1_epi8('/');
2463  while (CurPtr+16 <= BufferEnd) {
2464  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2465  Slashes));
2466  if (cmp != 0) {
2467  // Adjust the pointer to point directly after the first slash. It's
2468  // not necessary to set C here, it will be overwritten at the end of
2469  // the outer loop.
2470  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2471  goto FoundSlash;
2472  }
2473  CurPtr += 16;
2474  }
2475 #elif __ALTIVEC__
2476  __vector unsigned char Slashes = {
2477  '/', '/', '/', '/', '/', '/', '/', '/',
2478  '/', '/', '/', '/', '/', '/', '/', '/'
2479  };
2480  while (CurPtr+16 <= BufferEnd &&
2481  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2482  CurPtr += 16;
2483 #else
2484  // Scan for '/' quickly. Many block comments are very large.
2485  while (CurPtr[0] != '/' &&
2486  CurPtr[1] != '/' &&
2487  CurPtr[2] != '/' &&
2488  CurPtr[3] != '/' &&
2489  CurPtr+4 < BufferEnd) {
2490  CurPtr += 4;
2491  }
2492 #endif
2493 
2494  // It has to be one of the bytes scanned, increment to it and read one.
2495  C = *CurPtr++;
2496  }
2497 
2498  // Loop to scan the remainder.
2499  while (C != '/' && C != '\0')
2500  C = *CurPtr++;
2501 
2502  if (C == '/') {
2503  FoundSlash:
2504  if (CurPtr[-2] == '*') // We found the final */. We're done!
2505  break;
2506 
2507  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2508  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2509  // We found the final */, though it had an escaped newline between the
2510  // * and /. We're done!
2511  break;
2512  }
2513  }
2514  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2515  // If this is a /* inside of the comment, emit a warning. Don't do this
2516  // if this is a /*/, which will end the comment. This misses cases with
2517  // embedded escaped newlines, but oh well.
2518  if (!isLexingRawMode())
2519  Diag(CurPtr-1, diag::warn_nested_block_comment);
2520  }
2521  } else if (C == 0 && CurPtr == BufferEnd+1) {
2522  if (!isLexingRawMode())
2523  Diag(BufferPtr, diag::err_unterminated_block_comment);
2524  // Note: the user probably forgot a */. We could continue immediately
2525  // after the /*, but this would involve lexing a lot of what really is the
2526  // comment, which surely would confuse the parser.
2527  --CurPtr;
2528 
2529  // KeepWhitespaceMode should return this broken comment as a token. Since
2530  // it isn't a well formed comment, just return it as an 'unknown' token.
2531  if (isKeepWhitespaceMode()) {
2532  FormTokenWithChars(Result, CurPtr, tok::unknown);
2533  return true;
2534  }
2535 
2536  BufferPtr = CurPtr;
2537  return false;
2538  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2540  cutOffLexing();
2541  return false;
2542  }
2543 
2544  C = *CurPtr++;
2545  }
2546 
2547  // Notify comment handlers about the comment unless we're in a #if 0 block.
2548  if (PP && !isLexingRawMode() &&
2549  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2550  getSourceLocation(CurPtr)))) {
2551  BufferPtr = CurPtr;
2552  return true; // A token has to be returned.
2553  }
2554 
2555  // If we are returning comments as tokens, return this comment as a token.
2556  if (inKeepCommentMode()) {
2557  FormTokenWithChars(Result, CurPtr, tok::comment);
2558  return true;
2559  }
2560 
2561  // It is common for the tokens immediately after a /**/ comment to be
2562  // whitespace. Instead of going through the big switch, handle it
2563  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2564  // have already returned above with the comment as a token.
2565  if (isHorizontalWhitespace(*CurPtr)) {
2566  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2567  return false;
2568  }
2569 
2570  // Otherwise, just return so that the next character will be lexed as a token.
2571  BufferPtr = CurPtr;
2572  Result.setFlag(Token::LeadingSpace);
2573  return false;
2574 }
2575 
2576 //===----------------------------------------------------------------------===//
2577 // Primary Lexing Entry Points
2578 //===----------------------------------------------------------------------===//
2579 
2580 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2581 /// uninterpreted string. This switches the lexer out of directive mode.
2583  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2584  "Must be in a preprocessing directive!");
2585  Token Tmp;
2586 
2587  // CurPtr - Cache BufferPtr in an automatic variable.
2588  const char *CurPtr = BufferPtr;
2589  while (true) {
2590  char Char = getAndAdvanceChar(CurPtr, Tmp);
2591  switch (Char) {
2592  default:
2593  if (Result)
2594  Result->push_back(Char);
2595  break;
2596  case 0: // Null.
2597  // Found end of file?
2598  if (CurPtr-1 != BufferEnd) {
2599  if (isCodeCompletionPoint(CurPtr-1)) {
2601  cutOffLexing();
2602  return;
2603  }
2604 
2605  // Nope, normal character, continue.
2606  if (Result)
2607  Result->push_back(Char);
2608  break;
2609  }
2610  // FALL THROUGH.
2611  LLVM_FALLTHROUGH;
2612  case '\r':
2613  case '\n':
2614  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2615  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2616  BufferPtr = CurPtr-1;
2617 
2618  // Next, lex the character, which should handle the EOD transition.
2619  Lex(Tmp);
2620  if (Tmp.is(tok::code_completion)) {
2621  if (PP)
2623  Lex(Tmp);
2624  }
2625  assert(Tmp.is(tok::eod) && "Unexpected token!");
2626 
2627  // Finally, we're done;
2628  return;
2629  }
2630  }
2631 }
2632 
2633 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2634 /// condition, reporting diagnostics and handling other edge cases as required.
2635 /// This returns true if Result contains a token, false if PP.Lex should be
2636 /// called again.
2637 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2638  // If we hit the end of the file while parsing a preprocessor directive,
2639  // end the preprocessor directive first. The next token returned will
2640  // then be the end of file.
2642  // Done parsing the "line".
2644  // Update the location of token as well as BufferPtr.
2645  FormTokenWithChars(Result, CurPtr, tok::eod);
2646 
2647  // Restore comment saving mode, in case it was disabled for directive.
2648  if (PP)
2650  return true; // Have a token.
2651  }
2652 
2653  // If we are in raw mode, return this event as an EOF token. Let the caller
2654  // that put us in raw mode handle the event.
2655  if (isLexingRawMode()) {
2656  Result.startToken();
2657  BufferPtr = BufferEnd;
2658  FormTokenWithChars(Result, BufferEnd, tok::eof);
2659  return true;
2660  }
2661 
2662  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2664  ConditionalStack.clear();
2665  }
2666 
2667  // Issue diagnostics for unterminated #if and missing newline.
2668 
2669  // If we are in a #if directive, emit an error.
2670  while (!ConditionalStack.empty()) {
2671  if (PP->getCodeCompletionFileLoc() != FileLoc)
2672  PP->Diag(ConditionalStack.back().IfLoc,
2673  diag::err_pp_unterminated_conditional);
2674  ConditionalStack.pop_back();
2675  }
2676 
2677  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2678  // a pedwarn.
2679  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2680  DiagnosticsEngine &Diags = PP->getDiagnostics();
2681  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2682  unsigned DiagID;
2683 
2684  if (LangOpts.CPlusPlus11) {
2685  // C++11 [lex.phases] 2.2 p2
2686  // Prefer the C++98 pedantic compatibility warning over the generic,
2687  // non-extension, user-requested "missing newline at EOF" warning.
2688  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2689  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2690  } else {
2691  DiagID = diag::warn_no_newline_eof;
2692  }
2693  } else {
2694  DiagID = diag::ext_no_newline_eof;
2695  }
2696 
2697  Diag(BufferEnd, DiagID)
2698  << FixItHint::CreateInsertion(EndLoc, "\n");
2699  }
2700 
2701  BufferPtr = CurPtr;
2702 
2703  // Finally, let the preprocessor handle this.
2704  return PP->HandleEndOfFile(Result, isPragmaLexer());
2705 }
2706 
2707 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2708 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2709 /// else and 2 if there are no more tokens in the buffer controlled by the
2710 /// lexer.
2711 unsigned Lexer::isNextPPTokenLParen() {
2712  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2713 
2714  // Switch to 'skipping' mode. This will ensure that we can lex a token
2715  // without emitting diagnostics, disables macro expansion, and will cause EOF
2716  // to return an EOF token instead of popping the include stack.
2717  LexingRawMode = true;
2718 
2719  // Save state that can be changed while lexing so that we can restore it.
2720  const char *TmpBufferPtr = BufferPtr;
2721  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2722  bool atStartOfLine = IsAtStartOfLine;
2723  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2724  bool leadingSpace = HasLeadingSpace;
2725 
2726  Token Tok;
2727  Lex(Tok);
2728 
2729  // Restore state that may have changed.
2730  BufferPtr = TmpBufferPtr;
2731  ParsingPreprocessorDirective = inPPDirectiveMode;
2732  HasLeadingSpace = leadingSpace;
2733  IsAtStartOfLine = atStartOfLine;
2734  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2735 
2736  // Restore the lexer back to non-skipping mode.
2737  LexingRawMode = false;
2738 
2739  if (Tok.is(tok::eof))
2740  return 2;
2741  return Tok.is(tok::l_paren);
2742 }
2743 
2744 /// \brief Find the end of a version control conflict marker.
2745 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2746  ConflictMarkerKind CMK) {
2747  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2748  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2749  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2750  size_t Pos = RestOfBuffer.find(Terminator);
2751  while (Pos != StringRef::npos) {
2752  // Must occur at start of line.
2753  if (Pos == 0 ||
2754  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2755  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2756  Pos = RestOfBuffer.find(Terminator);
2757  continue;
2758  }
2759  return RestOfBuffer.data()+Pos;
2760  }
2761  return nullptr;
2762 }
2763 
2764 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2765 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2766 /// and recover nicely. This returns true if it is a conflict marker and false
2767 /// if not.
2768 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2769  // Only a conflict marker if it starts at the beginning of a line.
2770  if (CurPtr != BufferStart &&
2771  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2772  return false;
2773 
2774  // Check to see if we have <<<<<<< or >>>>.
2775  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2776  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2777  return false;
2778 
2779  // If we have a situation where we don't care about conflict markers, ignore
2780  // it.
2781  if (CurrentConflictMarkerState || isLexingRawMode())
2782  return false;
2783 
2784  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2785 
2786  // Check to see if there is an ending marker somewhere in the buffer at the
2787  // start of a line to terminate this conflict marker.
2788  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2789  // We found a match. We are really in a conflict marker.
2790  // Diagnose this, and ignore to the end of line.
2791  Diag(CurPtr, diag::err_conflict_marker);
2792  CurrentConflictMarkerState = Kind;
2793 
2794  // Skip ahead to the end of line. We know this exists because the
2795  // end-of-conflict marker starts with \r or \n.
2796  while (*CurPtr != '\r' && *CurPtr != '\n') {
2797  assert(CurPtr != BufferEnd && "Didn't find end of line");
2798  ++CurPtr;
2799  }
2800  BufferPtr = CurPtr;
2801  return true;
2802  }
2803 
2804  // No end of conflict marker found.
2805  return false;
2806 }
2807 
2808 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2809 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2810 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2811 /// the line. This returns true if it is a conflict marker and false if not.
2812 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2813  // Only a conflict marker if it starts at the beginning of a line.
2814  if (CurPtr != BufferStart &&
2815  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2816  return false;
2817 
2818  // If we have a situation where we don't care about conflict markers, ignore
2819  // it.
2820  if (!CurrentConflictMarkerState || isLexingRawMode())
2821  return false;
2822 
2823  // Check to see if we have the marker (4 characters in a row).
2824  for (unsigned i = 1; i != 4; ++i)
2825  if (CurPtr[i] != CurPtr[0])
2826  return false;
2827 
2828  // If we do have it, search for the end of the conflict marker. This could
2829  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2830  // be the end of conflict marker.
2831  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2832  CurrentConflictMarkerState)) {
2833  CurPtr = End;
2834 
2835  // Skip ahead to the end of line.
2836  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2837  ++CurPtr;
2838 
2839  BufferPtr = CurPtr;
2840 
2841  // No longer in the conflict marker.
2842  CurrentConflictMarkerState = CMK_None;
2843  return true;
2844  }
2845 
2846  return false;
2847 }
2848 
2849 static const char *findPlaceholderEnd(const char *CurPtr,
2850  const char *BufferEnd) {
2851  if (CurPtr == BufferEnd)
2852  return nullptr;
2853  BufferEnd -= 1; // Scan until the second last character.
2854  for (; CurPtr != BufferEnd; ++CurPtr) {
2855  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2856  return CurPtr + 2;
2857  }
2858  return nullptr;
2859 }
2860 
2861 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2862  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2864  return false;
2865  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2866  if (!End)
2867  return false;
2868  const char *Start = CurPtr - 1;
2869  if (!LangOpts.AllowEditorPlaceholders)
2870  Diag(Start, diag::err_placeholder_in_source);
2871  Result.startToken();
2872  FormTokenWithChars(Result, End, tok::raw_identifier);
2873  Result.setRawIdentifierData(Start);
2874  PP->LookUpIdentifierInfo(Result);
2876  BufferPtr = End;
2877  return true;
2878 }
2879 
2880 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2881  if (PP && PP->isCodeCompletionEnabled()) {
2882  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2883  return Loc == PP->getCodeCompletionLoc();
2884  }
2885 
2886  return false;
2887 }
2888 
2889 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2890  Token *Result) {
2891  unsigned CharSize;
2892  char Kind = getCharAndSize(StartPtr, CharSize);
2893 
2894  unsigned NumHexDigits;
2895  if (Kind == 'u')
2896  NumHexDigits = 4;
2897  else if (Kind == 'U')
2898  NumHexDigits = 8;
2899  else
2900  return 0;
2901 
2902  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2903  if (Result && !isLexingRawMode())
2904  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2905  return 0;
2906  }
2907 
2908  const char *CurPtr = StartPtr + CharSize;
2909  const char *KindLoc = &CurPtr[-1];
2910 
2911  uint32_t CodePoint = 0;
2912  for (unsigned i = 0; i < NumHexDigits; ++i) {
2913  char C = getCharAndSize(CurPtr, CharSize);
2914 
2915  unsigned Value = llvm::hexDigitValue(C);
2916  if (Value == -1U) {
2917  if (Result && !isLexingRawMode()) {
2918  if (i == 0) {
2919  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2920  << StringRef(KindLoc, 1);
2921  } else {
2922  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2923 
2924  // If the user wrote \U1234, suggest a fixit to \u.
2925  if (i == 4 && NumHexDigits == 8) {
2926  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2927  Diag(KindLoc, diag::note_ucn_four_not_eight)
2928  << FixItHint::CreateReplacement(URange, "u");
2929  }
2930  }
2931  }
2932 
2933  return 0;
2934  }
2935 
2936  CodePoint <<= 4;
2937  CodePoint += Value;
2938 
2939  CurPtr += CharSize;
2940  }
2941 
2942  if (Result) {
2943  Result->setFlag(Token::HasUCN);
2944  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
2945  StartPtr = CurPtr;
2946  else
2947  while (StartPtr != CurPtr)
2948  (void)getAndAdvanceChar(StartPtr, *Result);
2949  } else {
2950  StartPtr = CurPtr;
2951  }
2952 
2953  // Don't apply C family restrictions to UCNs in assembly mode
2954  if (LangOpts.AsmPreprocessor)
2955  return CodePoint;
2956 
2957  // C99 6.4.3p2: A universal character name shall not specify a character whose
2958  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
2959  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
2960  // C++11 [lex.charset]p2: If the hexadecimal value for a
2961  // universal-character-name corresponds to a surrogate code point (in the
2962  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
2963  // if the hexadecimal value for a universal-character-name outside the
2964  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
2965  // string literal corresponds to a control character (in either of the
2966  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
2967  // basic source character set, the program is ill-formed.
2968  if (CodePoint < 0xA0) {
2969  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
2970  return CodePoint;
2971 
2972  // We don't use isLexingRawMode() here because we need to warn about bad
2973  // UCNs even when skipping preprocessing tokens in a #if block.
2974  if (Result && PP) {
2975  if (CodePoint < 0x20 || CodePoint >= 0x7F)
2976  Diag(BufferPtr, diag::err_ucn_control_character);
2977  else {
2978  char C = static_cast<char>(CodePoint);
2979  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
2980  }
2981  }
2982 
2983  return 0;
2984  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
2985  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
2986  // We don't use isLexingRawMode() here because we need to diagnose bad
2987  // UCNs even when skipping preprocessing tokens in a #if block.
2988  if (Result && PP) {
2989  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
2990  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
2991  else
2992  Diag(BufferPtr, diag::err_ucn_escape_invalid);
2993  }
2994  return 0;
2995  }
2996 
2997  return CodePoint;
2998 }
2999 
3000 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3001  const char *CurPtr) {
3002  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3004  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3005  UnicodeWhitespaceChars.contains(C)) {
3006  Diag(BufferPtr, diag::ext_unicode_whitespace)
3007  << makeCharRange(*this, BufferPtr, CurPtr);
3008 
3009  Result.setFlag(Token::LeadingSpace);
3010  return true;
3011  }
3012  return false;
3013 }
3014 
3015 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3016  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3018  !PP->isPreprocessedOutput()) {
3020  makeCharRange(*this, BufferPtr, CurPtr),
3021  /*IsFirst=*/true);
3022  }
3023 
3024  MIOpt.ReadToken();
3025  return LexIdentifier(Result, CurPtr);
3026  }
3027 
3029  !PP->isPreprocessedOutput() &&
3030  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3031  // Non-ASCII characters tend to creep into source code unintentionally.
3032  // Instead of letting the parser complain about the unknown token,
3033  // just drop the character.
3034  // Note that we can /only/ do this when the non-ASCII character is actually
3035  // spelled as Unicode, not written as a UCN. The standard requires that
3036  // we not throw away any possible preprocessor tokens, but there's a
3037  // loophole in the mapping of Unicode characters to basic character set
3038  // characters that allows us to map these particular characters to, say,
3039  // whitespace.
3040  Diag(BufferPtr, diag::err_non_ascii)
3041  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3042 
3043  BufferPtr = CurPtr;
3044  return false;
3045  }
3046 
3047  // Otherwise, we have an explicit UCN or a character that's unlikely to show
3048  // up by accident.
3049  MIOpt.ReadToken();
3050  FormTokenWithChars(Result, CurPtr, tok::unknown);
3051  return true;
3052 }
3053 
3054 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3055  IsAtStartOfLine = Result.isAtStartOfLine();
3056  HasLeadingSpace = Result.hasLeadingSpace();
3057  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3058  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3059 }
3060 
3061 bool Lexer::Lex(Token &Result) {
3062  // Start a new token.
3063  Result.startToken();
3064 
3065  // Set up misc whitespace flags for LexTokenInternal.
3066  if (IsAtStartOfLine) {
3067  Result.setFlag(Token::StartOfLine);
3068  IsAtStartOfLine = false;
3069  }
3070 
3071  if (HasLeadingSpace) {
3072  Result.setFlag(Token::LeadingSpace);
3073  HasLeadingSpace = false;
3074  }
3075 
3076  if (HasLeadingEmptyMacro) {
3078  HasLeadingEmptyMacro = false;
3079  }
3080 
3081  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3082  IsAtPhysicalStartOfLine = false;
3083  bool isRawLex = isLexingRawMode();
3084  (void) isRawLex;
3085  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3086  // (After the LexTokenInternal call, the lexer might be destroyed.)
3087  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3088  return returnedToken;
3089 }
3090 
3091 /// LexTokenInternal - This implements a simple C family lexer. It is an
3092 /// extremely performance critical piece of code. This assumes that the buffer
3093 /// has a null character at the end of the file. This returns a preprocessing
3094 /// token, not a normal token, as such, it is an internal interface. It assumes
3095 /// that the Flags of result have been cleared before calling this.
3096 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3097 LexNextToken:
3098  // New token, can't need cleaning yet.
3100  Result.setIdentifierInfo(nullptr);
3101 
3102  // CurPtr - Cache BufferPtr in an automatic variable.
3103  const char *CurPtr = BufferPtr;
3104 
3105  // Small amounts of horizontal whitespace is very common between tokens.
3106  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
3107  ++CurPtr;
3108  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3109  ++CurPtr;
3110 
3111  // If we are keeping whitespace and other tokens, just return what we just
3112  // skipped. The next lexer invocation will return the token after the
3113  // whitespace.
3114  if (isKeepWhitespaceMode()) {
3115  FormTokenWithChars(Result, CurPtr, tok::unknown);
3116  // FIXME: The next token will not have LeadingSpace set.
3117  return true;
3118  }
3119 
3120  BufferPtr = CurPtr;
3121  Result.setFlag(Token::LeadingSpace);
3122  }
3123 
3124  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3125 
3126  // Read a character, advancing over it.
3127  char Char = getAndAdvanceChar(CurPtr, Result);
3129 
3130  switch (Char) {
3131  case 0: // Null.
3132  // Found end of file?
3133  if (CurPtr-1 == BufferEnd)
3134  return LexEndOfFile(Result, CurPtr-1);
3135 
3136  // Check if we are performing code completion.
3137  if (isCodeCompletionPoint(CurPtr-1)) {
3138  // Return the code-completion token.
3139  Result.startToken();
3140  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3141  return true;
3142  }
3143 
3144  if (!isLexingRawMode())
3145  Diag(CurPtr-1, diag::null_in_file);
3146  Result.setFlag(Token::LeadingSpace);
3147  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3148  return true; // KeepWhitespaceMode
3149 
3150  // We know the lexer hasn't changed, so just try again with this lexer.
3151  // (We manually eliminate the tail call to avoid recursion.)
3152  goto LexNextToken;
3153 
3154  case 26: // DOS & CP/M EOF: "^Z".
3155  // If we're in Microsoft extensions mode, treat this as end of file.
3156  if (LangOpts.MicrosoftExt) {
3157  if (!isLexingRawMode())
3158  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3159  return LexEndOfFile(Result, CurPtr-1);
3160  }
3161 
3162  // If Microsoft extensions are disabled, this is just random garbage.
3163  Kind = tok::unknown;
3164  break;
3165 
3166  case '\r':
3167  if (CurPtr[0] == '\n')
3168  Char = getAndAdvanceChar(CurPtr, Result);
3169  LLVM_FALLTHROUGH;
3170  case '\n':
3171  // If we are inside a preprocessor directive and we see the end of line,
3172  // we know we are done with the directive, so return an EOD token.
3174  // Done parsing the "line".
3176 
3177  // Restore comment saving mode, in case it was disabled for directive.
3178  if (PP)
3180 
3181  // Since we consumed a newline, we are back at the start of a line.
3182  IsAtStartOfLine = true;
3183  IsAtPhysicalStartOfLine = true;
3184 
3185  Kind = tok::eod;
3186  break;
3187  }
3188 
3189  // No leading whitespace seen so far.
3191 
3192  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3193  return true; // KeepWhitespaceMode
3194 
3195  // We only saw whitespace, so just try again with this lexer.
3196  // (We manually eliminate the tail call to avoid recursion.)
3197  goto LexNextToken;
3198  case ' ':
3199  case '\t':
3200  case '\f':
3201  case '\v':
3202  SkipHorizontalWhitespace:
3203  Result.setFlag(Token::LeadingSpace);
3204  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3205  return true; // KeepWhitespaceMode
3206 
3207  SkipIgnoredUnits:
3208  CurPtr = BufferPtr;
3209 
3210  // If the next token is obviously a // or /* */ comment, skip it efficiently
3211  // too (without going through the big switch stmt).
3212  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3213  LangOpts.LineComment &&
3214  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3215  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3216  return true; // There is a token to return.
3217  goto SkipIgnoredUnits;
3218  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3219  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3220  return true; // There is a token to return.
3221  goto SkipIgnoredUnits;
3222  } else if (isHorizontalWhitespace(*CurPtr)) {
3223  goto SkipHorizontalWhitespace;
3224  }
3225  // We only saw whitespace, so just try again with this lexer.
3226  // (We manually eliminate the tail call to avoid recursion.)
3227  goto LexNextToken;
3228 
3229  // C99 6.4.4.1: Integer Constants.
3230  // C99 6.4.4.2: Floating Constants.
3231  case '0': case '1': case '2': case '3': case '4':
3232  case '5': case '6': case '7': case '8': case '9':
3233  // Notify MIOpt that we read a non-whitespace/non-comment token.
3234  MIOpt.ReadToken();
3235  return LexNumericConstant(Result, CurPtr);
3236 
3237  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3238  // Notify MIOpt that we read a non-whitespace/non-comment token.
3239  MIOpt.ReadToken();
3240 
3241  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3242  Char = getCharAndSize(CurPtr, SizeTmp);
3243 
3244  // UTF-16 string literal
3245  if (Char == '"')
3246  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3247  tok::utf16_string_literal);
3248 
3249  // UTF-16 character constant
3250  if (Char == '\'')
3251  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3252  tok::utf16_char_constant);
3253 
3254  // UTF-16 raw string literal
3255  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3256  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3257  return LexRawStringLiteral(Result,
3258  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3259  SizeTmp2, Result),
3260  tok::utf16_string_literal);
3261 
3262  if (Char == '8') {
3263  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3264 
3265  // UTF-8 string literal
3266  if (Char2 == '"')
3267  return LexStringLiteral(Result,
3268  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3269  SizeTmp2, Result),
3270  tok::utf8_string_literal);
3271  if (Char2 == '\'' && LangOpts.CPlusPlus17)
3272  return LexCharConstant(
3273  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3274  SizeTmp2, Result),
3275  tok::utf8_char_constant);
3276 
3277  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3278  unsigned SizeTmp3;
3279  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3280  // UTF-8 raw string literal
3281  if (Char3 == '"') {
3282  return LexRawStringLiteral(Result,
3283  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3284  SizeTmp2, Result),
3285  SizeTmp3, Result),
3286  tok::utf8_string_literal);
3287  }
3288  }
3289  }
3290  }
3291 
3292  // treat u like the start of an identifier.
3293  return LexIdentifier(Result, CurPtr);
3294 
3295  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3296  // Notify MIOpt that we read a non-whitespace/non-comment token.
3297  MIOpt.ReadToken();
3298 
3299  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3300  Char = getCharAndSize(CurPtr, SizeTmp);
3301 
3302  // UTF-32 string literal
3303  if (Char == '"')
3304  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3305  tok::utf32_string_literal);
3306 
3307  // UTF-32 character constant
3308  if (Char == '\'')
3309  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3310  tok::utf32_char_constant);
3311 
3312  // UTF-32 raw string literal
3313  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3314  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3315  return LexRawStringLiteral(Result,
3316  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3317  SizeTmp2, Result),
3318  tok::utf32_string_literal);
3319  }
3320 
3321  // treat U like the start of an identifier.
3322  return LexIdentifier(Result, CurPtr);
3323 
3324  case 'R': // Identifier or C++0x raw string literal
3325  // Notify MIOpt that we read a non-whitespace/non-comment token.
3326  MIOpt.ReadToken();
3327 
3328  if (LangOpts.CPlusPlus11) {
3329  Char = getCharAndSize(CurPtr, SizeTmp);
3330 
3331  if (Char == '"')
3332  return LexRawStringLiteral(Result,
3333  ConsumeChar(CurPtr, SizeTmp, Result),
3334  tok::string_literal);
3335  }
3336 
3337  // treat R like the start of an identifier.
3338  return LexIdentifier(Result, CurPtr);
3339 
3340  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3341  // Notify MIOpt that we read a non-whitespace/non-comment token.
3342  MIOpt.ReadToken();
3343  Char = getCharAndSize(CurPtr, SizeTmp);
3344 
3345  // Wide string literal.
3346  if (Char == '"')
3347  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3348  tok::wide_string_literal);
3349 
3350  // Wide raw string literal.
3351  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3352  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3353  return LexRawStringLiteral(Result,
3354  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3355  SizeTmp2, Result),
3356  tok::wide_string_literal);
3357 
3358  // Wide character constant.
3359  if (Char == '\'')
3360  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3361  tok::wide_char_constant);
3362  // FALL THROUGH, treating L like the start of an identifier.
3363  LLVM_FALLTHROUGH;
3364 
3365  // C99 6.4.2: Identifiers.
3366  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3367  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3368  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3369  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3370  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3371  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3372  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3373  case 'v': case 'w': case 'x': case 'y': case 'z':
3374  case '_':
3375  // Notify MIOpt that we read a non-whitespace/non-comment token.
3376  MIOpt.ReadToken();
3377  return LexIdentifier(Result, CurPtr);
3378 
3379  case '$': // $ in identifiers.
3380  if (LangOpts.DollarIdents) {
3381  if (!isLexingRawMode())
3382  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3383  // Notify MIOpt that we read a non-whitespace/non-comment token.
3384  MIOpt.ReadToken();
3385  return LexIdentifier(Result, CurPtr);
3386  }
3387 
3388  Kind = tok::unknown;
3389  break;
3390 
3391  // C99 6.4.4: Character Constants.
3392  case '\'':
3393  // Notify MIOpt that we read a non-whitespace/non-comment token.
3394  MIOpt.ReadToken();
3395  return LexCharConstant(Result, CurPtr, tok::char_constant);
3396 
3397  // C99 6.4.5: String Literals.
3398  case '"':
3399  // Notify MIOpt that we read a non-whitespace/non-comment token.
3400  MIOpt.ReadToken();
3401  return LexStringLiteral(Result, CurPtr, tok::string_literal);
3402 
3403  // C99 6.4.6: Punctuators.
3404  case '?':
3405  Kind = tok::question;
3406  break;
3407  case '[':
3408  Kind = tok::l_square;
3409  break;
3410  case ']':
3411  Kind = tok::r_square;
3412  break;
3413  case '(':
3414  Kind = tok::l_paren;
3415  break;
3416  case ')':
3417  Kind = tok::r_paren;
3418  break;
3419  case '{':
3420  Kind = tok::l_brace;
3421  break;
3422  case '}':
3423  Kind = tok::r_brace;
3424  break;
3425  case '.':
3426  Char = getCharAndSize(CurPtr, SizeTmp);
3427  if (Char >= '0' && Char <= '9') {
3428  // Notify MIOpt that we read a non-whitespace/non-comment token.
3429  MIOpt.ReadToken();
3430 
3431  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3432  } else if (LangOpts.CPlusPlus && Char == '*') {
3433  Kind = tok::periodstar;
3434  CurPtr += SizeTmp;
3435  } else if (Char == '.' &&
3436  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3437  Kind = tok::ellipsis;
3438  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3439  SizeTmp2, Result);
3440  } else {
3441  Kind = tok::period;
3442  }
3443  break;
3444  case '&':
3445  Char = getCharAndSize(CurPtr, SizeTmp);
3446  if (Char == '&') {
3447  Kind = tok::ampamp;
3448  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3449  } else if (Char == '=') {
3450  Kind = tok::ampequal;
3451  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3452  } else {
3453  Kind = tok::amp;
3454  }
3455  break;
3456  case '*':
3457  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3458  Kind = tok::starequal;
3459  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3460  } else {
3461  Kind = tok::star;
3462  }
3463  break;
3464  case '+':
3465  Char = getCharAndSize(CurPtr, SizeTmp);
3466  if (Char == '+') {
3467  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3468  Kind = tok::plusplus;
3469  } else if (Char == '=') {
3470  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3471  Kind = tok::plusequal;
3472  } else {
3473  Kind = tok::plus;
3474  }
3475  break;
3476  case '-':
3477  Char = getCharAndSize(CurPtr, SizeTmp);
3478  if (Char == '-') { // --
3479  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3480  Kind = tok::minusminus;
3481  } else if (Char == '>' && LangOpts.CPlusPlus &&
3482  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3483  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3484  SizeTmp2, Result);
3485  Kind = tok::arrowstar;
3486  } else if (Char == '>') { // ->
3487  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3488  Kind = tok::arrow;
3489  } else if (Char == '=') { // -=
3490  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3491  Kind = tok::minusequal;
3492  } else {
3493  Kind = tok::minus;
3494  }
3495  break;
3496  case '~':
3497  Kind = tok::tilde;
3498  break;
3499  case '!':
3500  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3501  Kind = tok::exclaimequal;
3502  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3503  } else {
3504  Kind = tok::exclaim;
3505  }
3506  break;
3507  case '/':
3508  // 6.4.9: Comments
3509  Char = getCharAndSize(CurPtr, SizeTmp);
3510  if (Char == '/') { // Line comment.
3511  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3512  // want to lex this as a comment. There is one problem with this though,
3513  // that in one particular corner case, this can change the behavior of the
3514  // resultant program. For example, In "foo //**/ bar", C89 would lex
3515  // this as "foo / bar" and langauges with Line comments would lex it as
3516  // "foo". Check to see if the character after the second slash is a '*'.
3517  // If so, we will lex that as a "/" instead of the start of a comment.
3518  // However, we never do this if we are just preprocessing.
3519  bool TreatAsComment = LangOpts.LineComment &&
3520  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3521  if (!TreatAsComment)
3522  if (!(PP && PP->isPreprocessedOutput()))
3523  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3524 
3525  if (TreatAsComment) {
3526  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3527  TokAtPhysicalStartOfLine))
3528  return true; // There is a token to return.
3529 
3530  // It is common for the tokens immediately after a // comment to be
3531  // whitespace (indentation for the next line). Instead of going through
3532  // the big switch, handle it efficiently now.
3533  goto SkipIgnoredUnits;
3534  }
3535  }
3536 
3537  if (Char == '*') { // /**/ comment.
3538  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3539  TokAtPhysicalStartOfLine))
3540  return true; // There is a token to return.
3541 
3542  // We only saw whitespace, so just try again with this lexer.
3543  // (We manually eliminate the tail call to avoid recursion.)
3544  goto LexNextToken;
3545  }
3546 
3547  if (Char == '=') {
3548  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3549  Kind = tok::slashequal;
3550  } else {
3551  Kind = tok::slash;
3552  }
3553  break;
3554  case '%':
3555  Char = getCharAndSize(CurPtr, SizeTmp);
3556  if (Char == '=') {
3557  Kind = tok::percentequal;
3558  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3559  } else if (LangOpts.Digraphs && Char == '>') {
3560  Kind = tok::r_brace; // '%>' -> '}'
3561  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3562  } else if (LangOpts.Digraphs && Char == ':') {
3563  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3564  Char = getCharAndSize(CurPtr, SizeTmp);
3565  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3566  Kind = tok::hashhash; // '%:%:' -> '##'
3567  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3568  SizeTmp2, Result);
3569  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3570  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3571  if (!isLexingRawMode())
3572  Diag(BufferPtr, diag::ext_charize_microsoft);
3573  Kind = tok::hashat;
3574  } else { // '%:' -> '#'
3575  // We parsed a # character. If this occurs at the start of the line,
3576  // it's actually the start of a preprocessing directive. Callback to
3577  // the preprocessor to handle it.
3578  // TODO: -fpreprocessed mode??
3579  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3580  goto HandleDirective;
3581 
3582  Kind = tok::hash;
3583  }
3584  } else {
3585  Kind = tok::percent;
3586  }
3587  break;
3588  case '<':
3589  Char = getCharAndSize(CurPtr, SizeTmp);
3590  if (ParsingFilename) {
3591  return LexAngledStringLiteral(Result, CurPtr);
3592  } else if (Char == '<') {
3593  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3594  if (After == '=') {
3595  Kind = tok::lesslessequal;
3596  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3597  SizeTmp2, Result);
3598  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3599  // If this is actually a '<<<<<<<' version control conflict marker,
3600  // recognize it as such and recover nicely.
3601  goto LexNextToken;
3602  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3603  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3604  // ignore it.
3605  goto LexNextToken;
3606  } else if (LangOpts.CUDA && After == '<') {
3607  Kind = tok::lesslessless;
3608  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3609  SizeTmp2, Result);
3610  } else {
3611  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3612  Kind = tok::lessless;
3613  }
3614  } else if (Char == '=') {
3615  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3616  if (After == '>') {
3617  if (getLangOpts().CPlusPlus2a) {
3618  if (!isLexingRawMode())
3619  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3620  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3621  SizeTmp2, Result);
3622  Kind = tok::spaceship;
3623  break;
3624  }
3625  // Suggest adding a space between the '<=' and the '>' to avoid a
3626  // change in semantics if this turns up in C++ <=17 mode.
3627  if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3628  Diag(BufferPtr, diag::warn_cxx2a_compat_spaceship)
3630  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3631  }
3632  }
3633  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3634  Kind = tok::lessequal;
3635  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3636  if (LangOpts.CPlusPlus11 &&
3637  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3638  // C++0x [lex.pptoken]p3:
3639  // Otherwise, if the next three characters are <:: and the subsequent
3640  // character is neither : nor >, the < is treated as a preprocessor
3641  // token by itself and not as the first character of the alternative
3642  // token <:.
3643  unsigned SizeTmp3;
3644  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3645  if (After != ':' && After != '>') {
3646  Kind = tok::less;
3647  if (!isLexingRawMode())
3648  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3649  break;
3650  }
3651  }
3652 
3653  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3654  Kind = tok::l_square;
3655  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3656  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3657  Kind = tok::l_brace;
3658  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3659  lexEditorPlaceholder(Result, CurPtr)) {
3660  return true;
3661  } else {
3662  Kind = tok::less;
3663  }
3664  break;
3665  case '>':
3666  Char = getCharAndSize(CurPtr, SizeTmp);
3667  if (Char == '=') {
3668  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3669  Kind = tok::greaterequal;
3670  } else if (Char == '>') {
3671  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3672  if (After == '=') {
3673  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3674  SizeTmp2, Result);
3675  Kind = tok::greatergreaterequal;
3676  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3677  // If this is actually a '>>>>' conflict marker, recognize it as such
3678  // and recover nicely.
3679  goto LexNextToken;
3680  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3681  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3682  goto LexNextToken;
3683  } else if (LangOpts.CUDA && After == '>') {
3684  Kind = tok::greatergreatergreater;
3685  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3686  SizeTmp2, Result);
3687  } else {
3688  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3689  Kind = tok::greatergreater;
3690  }
3691  } else {
3692  Kind = tok::greater;
3693  }
3694  break;
3695  case '^':
3696  Char = getCharAndSize(CurPtr, SizeTmp);
3697  if (Char == '=') {
3698  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3699  Kind = tok::caretequal;
3700  } else if (LangOpts.OpenCL && Char == '^') {
3701  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3702  Kind = tok::caretcaret;
3703  } else {
3704  Kind = tok::caret;
3705  }
3706  break;
3707  case '|':
3708  Char = getCharAndSize(CurPtr, SizeTmp);
3709  if (Char == '=') {
3710  Kind = tok::pipeequal;
3711  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3712  } else if (Char == '|') {
3713  // If this is '|||||||' and we're in a conflict marker, ignore it.
3714  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3715  goto LexNextToken;
3716  Kind = tok::pipepipe;
3717  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3718  } else {
3719  Kind = tok::pipe;
3720  }
3721  break;
3722  case ':':
3723  Char = getCharAndSize(CurPtr, SizeTmp);
3724  if (LangOpts.Digraphs && Char == '>') {
3725  Kind = tok::r_square; // ':>' -> ']'
3726  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3727  } else if ((LangOpts.CPlusPlus ||
3728  LangOpts.DoubleSquareBracketAttributes) &&
3729  Char == ':') {
3730  Kind = tok::coloncolon;
3731  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3732  } else {
3733  Kind = tok::colon;
3734  }
3735  break;
3736  case ';':
3737  Kind = tok::semi;
3738  break;
3739  case '=':
3740  Char = getCharAndSize(CurPtr, SizeTmp);
3741  if (Char == '=') {
3742  // If this is '====' and we're in a conflict marker, ignore it.
3743  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3744  goto LexNextToken;
3745 
3746  Kind = tok::equalequal;
3747  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3748  } else {
3749  Kind = tok::equal;
3750  }
3751  break;
3752  case ',':
3753  Kind = tok::comma;
3754  break;
3755  case '#':
3756  Char = getCharAndSize(CurPtr, SizeTmp);
3757  if (Char == '#') {
3758  Kind = tok::hashhash;
3759  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3760  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3761  Kind = tok::hashat;
3762  if (!isLexingRawMode())
3763  Diag(BufferPtr, diag::ext_charize_microsoft);
3764  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3765  } else {
3766  // We parsed a # character. If this occurs at the start of the line,
3767  // it's actually the start of a preprocessing directive. Callback to
3768  // the preprocessor to handle it.
3769  // TODO: -fpreprocessed mode??
3770  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3771  goto HandleDirective;
3772 
3773  Kind = tok::hash;
3774  }
3775  break;
3776 
3777  case '@':
3778  // Objective C support.
3779  if (CurPtr[-1] == '@' && LangOpts.ObjC1)
3780  Kind = tok::at;
3781  else
3782  Kind = tok::unknown;
3783  break;
3784 
3785  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3786  case '\\':
3787  if (!LangOpts.AsmPreprocessor) {
3788  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3789  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3790  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3791  return true; // KeepWhitespaceMode
3792 
3793  // We only saw whitespace, so just try again with this lexer.
3794  // (We manually eliminate the tail call to avoid recursion.)
3795  goto LexNextToken;
3796  }
3797 
3798  return LexUnicode(Result, CodePoint, CurPtr);
3799  }
3800  }
3801 
3802  Kind = tok::unknown;
3803  break;
3804 
3805  default: {
3806  if (isASCII(Char)) {
3807  Kind = tok::unknown;
3808  break;
3809  }
3810 
3811  llvm::UTF32 CodePoint;
3812 
3813  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3814  // an escaped newline.
3815  --CurPtr;
3816  const char *UTF8StartPtr = CurPtr;
3817  llvm::ConversionResult Status =
3818  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3819  (const llvm::UTF8 *)BufferEnd,
3820  &CodePoint,
3821  llvm::strictConversion);
3822  if (Status == llvm::conversionOK) {
3823  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3824  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3825  return true; // KeepWhitespaceMode
3826 
3827  // We only saw whitespace, so just try again with this lexer.
3828  // (We manually eliminate the tail call to avoid recursion.)
3829  goto LexNextToken;
3830  }
3831  if (!isLexingRawMode())
3832  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
3833  makeCharRange(*this, UTF8StartPtr, CurPtr));
3834  return LexUnicode(Result, CodePoint, CurPtr);
3835  }
3836 
3838  PP->isPreprocessedOutput()) {
3839  ++CurPtr;
3840  Kind = tok::unknown;
3841  break;
3842  }
3843 
3844  // Non-ASCII characters tend to creep into source code unintentionally.
3845  // Instead of letting the parser complain about the unknown token,
3846  // just diagnose the invalid UTF-8, then drop the character.
3847  Diag(CurPtr, diag::err_invalid_utf8);
3848 
3849  BufferPtr = CurPtr+1;
3850  // We're pretending the character didn't exist, so just try again with
3851  // this lexer.
3852  // (We manually eliminate the tail call to avoid recursion.)
3853  goto LexNextToken;
3854  }
3855  }
3856 
3857  // Notify MIOpt that we read a non-whitespace/non-comment token.
3858  MIOpt.ReadToken();
3859 
3860  // Update the location of token as well as BufferPtr.
3861  FormTokenWithChars(Result, CurPtr, Kind);
3862  return true;
3863 
3864 HandleDirective:
3865  // We parsed a # character and it's the start of a preprocessing directive.
3866 
3867  FormTokenWithChars(Result, CurPtr, tok::hash);
3868  PP->HandleDirective(Result);
3869 
3870  if (PP->hadModuleLoaderFatalFailure()) {
3871  // With a fatal failure in the module loader, we abort parsing.
3872  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3873  return true;
3874  }
3875 
3876  // We parsed the directive; lex a token with the new state.
3877  return false;
3878 }
SourceLocation getLocForStartOfFile(FileID FID) const
Return the source location corresponding to the first byte of the specified file. ...
Describes the bounds (start, size) of the preamble and a flag required by PreprocessorOptions::Precom...
Definition: Lexer.h:59
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:389
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens...
Definition: Lexer.h:77
SourceLocation getLocWithOffset(int Offset) const
Return a source location with the specified offset from this SourceLocation.
This is a discriminated union of FileInfo and ExpansionInfo.
unsigned getRawEncoding() const
When a SourceLocation itself cannot be used, this returns an (opaque) 32-bit integer encoding for it...
SourceLocation getSpellingLoc() const
void setFlagValue(TokenFlags Flag, bool Val)
Set a flag to either true or false.
Definition: Token.h:257
static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[]
void setBegin(SourceLocation b)
static __inline__ int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:14868
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
Defines the SourceManager interface.
LLVM_READNONE bool isASCII(char c)
Returns true if this is an ASCII character.
Definition: CharInfo.h:43
static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1428
bool isInPrimaryFile() const
Return true if we&#39;re in the top-level file, not in a #include.
StringRef P
const char * getCharacterData(SourceLocation SL, bool *Invalid=nullptr) const
Return a pointer to the start of the specified location in the appropriate spelling MemoryBuffer...
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded...
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1294
void setFlag(TokenFlags Flag)
Set the specified flag.
Definition: Token.h:234
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:519
StringRef getBufferData(FileID FID, bool *Invalid=nullptr) const
Return a StringRef to the source buffer data for the specified FileID.
Lexer(FileID FID, const llvm::MemoryBuffer *InputBuffer, Preprocessor &PP)
Lexer constructor - Create a new lexer object for the specified buffer with the specified preprocesso...
Definition: Lexer.cpp:134
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:844
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:796
LLVM_READONLY bool isHorizontalWhitespace(unsigned char c)
Returns true if this character is horizontal ASCII whitespace: &#39; &#39;, &#39;\t&#39;, &#39;\f&#39;, &#39;\v&#39;.
Definition: CharInfo.h:71
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:43
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
GetMappedTokenLoc - If lexing out of a &#39;mapped buffer&#39;, where we pretend the lexer buffer was all exp...
Definition: Lexer.cpp:1098
Like System, but searched after the system directories.
SourceLocation getCodeCompletionFileLoc() const
Returns the start location of the file of code-completion point.
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:187
Defines the MultipleIncludeOpt interface.
bool isAnnotation() const
Return true if this is any of tok::annot_* kind tokens.
Definition: Token.h:118
tok::TokenKind getKind() const
Definition: Token.h:90
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:113
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:1067
SourceLocation getBegin() const
bool ParsingPreprocessorDirective
True when parsing #XXX; turns &#39;\n&#39; into a tok::eod token.
void setRawIdentifierData(const char *Ptr)
Definition: Token.h:207
static SourceLocation getFromRawEncoding(unsigned Encoding)
Turn a raw encoding of a SourceLocation object into a real SourceLocation.
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:50
bool isPragmaLexer() const
isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
Definition: Lexer.h:184
void resetExtendedTokenMode()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:164
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s...
Definition: Lexer.h:53
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:58
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file...
Definition: Lexer.h:262
bool isInvalid() 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 __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:3855
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:498
static bool isNewLineEscaped(const char *BufferStart, const char *Str)
Checks whether new line pointed by Str is preceded by escape sequence.
Definition: Lexer.cpp:1047
LLVM_READONLY bool isWhitespace(unsigned char c)
Return true if this character is horizontal or vertical ASCII whitespace: &#39; &#39;, &#39;\t&#39;, &#39;\f&#39;, &#39;\v&#39;, &#39;\n&#39;, &#39;\r&#39;.
Definition: CharInfo.h:88
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
Return the &#39;spelling&#39; of the token at the given location; does not go up to the spelling location or ...
tok::TokenKind getTokenID() const
If this is a source-language token (e.g.
uint32_t Offset
Definition: CacheTokens.cpp:43
const FormatToken & Tok
Forward-declares and imports various common LLVM datatypes that clang wants to use unqualified...
SourceLocation getExpansionLoc(SourceLocation Loc) const
Given a SourceLocation object Loc, return the expansion location referenced by the ID...
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:447
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:149
Defines the Diagnostic-related interfaces.
SourceLocation getSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
__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:713
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
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L)
isBlockCommentEndOfEscapedNewLine - Return true if the specified newline character (either \n or \r) ...
Definition: Lexer.cpp:2340
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1463
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:1042
bool ParsingFilename
True after #include; turns <xx> into a tok::angle_string_literal token.
const LangOptions & getLangOpts() const
getLangOpts - Return the language features currently enabled.
Definition: Lexer.h:169
static const llvm::sys::UnicodeCharRange C11DisallowedInitialIDCharRanges[]
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...
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
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:935
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
Given a tok::raw_identifier token, look up the identifier information for the token and install it in...
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:818
bool isCodeCompletionEnabled() const
Determine if we are performing code completion.
SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
Defines the clang::LangOptions interface.
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:774
SourceLocation End
Represents a character-granular source range.
bool isKeepWhitespaceMode() const
isKeepWhitespaceMode - Return true if the lexer should return tokens for every character in the file...
Definition: Lexer.h:207
static PreambleBounds ComputePreamble(StringRef Buffer, const LangOptions &LangOpts, unsigned MaxLines=0)
Compute the preamble of the given file.
Definition: Lexer.cpp:574
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:436
const FileEntry * getFileEntryForID(FileID FID) const
Returns the FileEntry record for the provided FileID.
const AnnotatedLine * Line
const FunctionProtoType * T
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:1267
bool hasLeadingEmptyMacro() const
Return true if this token has an empty macro before it.
Definition: Token.h:287
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:1123
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:124
Defines the clang::Preprocessor interface.
tok::ObjCKeywordKind getObjCKeywordID() const
Return the Objective-C keyword ID for the this identifier.
PreambleDirectiveKind
Definition: Lexer.cpp:567
SourceLocation Begin
MultipleIncludeOpt MIOpt
A state machine that detects the #ifndef-wrapping a file idiom for the multiple-include optimization...
void setEnd(SourceLocation e)
bool getCommentRetentionState() const
Definition: Preprocessor.h:859
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:481
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...
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:1412
ObjCKeywordKind
Provides a namespace for Objective-C keywords which start with an &#39;@&#39;.
Definition: TokenKinds.h:41
const ExpansionInfo & getExpansion() const
bool isRecordingPreamble() const
static CharSourceRange getCharRange(SourceRange R)
SourceManager & getSourceManager() const
Definition: Preprocessor.h:819
bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroBegin=nullptr) const
Returns true if the given MacroID location points at the beginning of the immediate macro expansion...
Kind
llvm::MemoryBuffer * getBuffer(FileID FID, SourceLocation Loc, bool *Invalid=nullptr) const
Return the buffer for the specified FileID.
Encodes a location in the source.
static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range)
After encountering UTF-8 character C and interpreting it as an identifier character, check whether it&#39;s a homoglyph for a common non-identifier source character that is unlikely to be an intentional identifier character and warn if so.
Definition: Lexer.cpp:1508
std::pair< SourceLocation, SourceLocation > getImmediateExpansionRange(SourceLocation Loc) const
Return the start/end of the expansion information for an expansion location.
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:177
static Optional< Token > findNextToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Finds the token that comes right after the given location.
Definition: Lexer.cpp:1234
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1469
void setIdentifierInfo(IdentifierInfo *II)
Definition: Token.h:186
static const llvm::sys::UnicodeCharRange C99DisallowedInitialIDCharRanges[]
bool isAtStartOfLine() const
isAtStartOfLine - Return true if this token is at the start of a line.
Definition: Token.h:266
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:546
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:844
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
bool operator<(DeclarationName LHS, DeclarationName RHS)
Ordering on two declaration names.
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:67
SourceLocation getCodeCompletionLoc() const
Returns the location of the code-completion point.
SourceLocation getExpansionLocStart() const
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1142
__DEVICE__ void * memcpy(void *__a, const void *__b, size_t __c)
static const llvm::sys::UnicodeCharRange C99AllowedIDCharRanges[]
__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:971
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1018
bool inKeepCommentMode() const
inKeepCommentMode - Return true if the lexer should return comments as tokens.
Definition: Lexer.h:221
bool isTokenRange() const
Return true if the end of this range specifies the start of the last token.
static const llvm::sys::UnicodeCharRange UnicodeWhitespaceCharRanges[]
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
Dataflow Directional Tag Classes.
bool isValid() const
Return true if this is a valid SourceLocation object.
LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
Returns true if this character is vertical ASCII whitespace: &#39;\n&#39;, &#39;\r&#39;.
Definition: CharInfo.h:79
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:871
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:262
static FixItHint CreateRemoval(CharSourceRange RemoveRange)
Create a code modification hint that removes the given source range.
Definition: Diagnostic.h:118
bool isHandleIdentifierCase() const
Return true if the Preprocessor::HandleIdentifier must be called on a token of this identifier...
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
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
tok::PPKeywordKind getPPKeywordID() const
Return the preprocessor keyword ID for this identifier.
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
unsigned getLength() const
Definition: Token.h:127
static const char * findPlaceholderEnd(const char *CurPtr, const char *BufferEnd)
Definition: Lexer.cpp:2849
void setLiteralData(const char *Ptr)
Definition: Token.h:219
const char * getLiteralData() const
getLiteralData - For a literal token (numeric constant, string, etc), this returns a pointer to the s...
Definition: Token.h:215
FileID getFileID(SourceLocation SpellingLoc) const
Return the FileID for a SourceLocation.
static const llvm::sys::UnicodeCharRange CXX03AllowedIDCharRanges[]
bool isMacroArgExpansion(SourceLocation Loc, SourceLocation *StartLoc=nullptr) const
Tests whether the given source location represents a macro argument&#39;s expansion into the function-lik...
bool HandleIdentifier(Token &Identifier)
Callback invoked when the lexer reads an identifier and has filled in the tokens IdentifierInfo membe...
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&#39;s location and length to ...
static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1446
SourceLocation getEnd() const
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:92
PreprocessorOptions & getPreprocessorOpts() const
Retrieve the preprocessor options used to initialize this preprocessor.
Definition: Preprocessor.h:810
Defines the clang::TokenKind enum and support functions.
const SrcMgr::SLocEntry & getSLocEntry(FileID FID, bool *Invalid=nullptr) const
static char GetTrigraphCharForLetter(char Letter)
GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, return the decoded trigraph...
Definition: Lexer.cpp:1152
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1043
bool HandleComment(Token &Token, SourceRange Comment)
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string...
Definition: Lexer.cpp:2582
Defines the clang::SourceLocation class and associated facilities.
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:812
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
Not within a conflict marker.
Definition: Lexer.h:45
static char DecodeTrigraphChar(const char *CP, Lexer *L)
DecodeTrigraphChar - If the specified character is a legal trigraph when prefixed with ...
Definition: Lexer.cpp:1171
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2745
static void StringifyImpl(T &Str, char Quote)
Definition: Lexer.cpp:222
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:129
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode...
Definition: Lexer.h:228
bool needsCleaning() const
Return true if this token has trigraphs or escaped newlines in it.
Definition: Token.h:283
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:4356
bool isIgnored(unsigned DiagID, SourceLocation Loc) const
Determine whether the diagnostic is known to be ignored.
Definition: Diagnostic.h:818
A normal or diff3 conflict marker, initiated by at least 7 "<"s, separated by at least 7 "="s or "|"s...
Definition: Lexer.h:49
A trivial tuple used to represent a source range.
void clearFlag(TokenFlags Flag)
Unset the specified flag.
Definition: Token.h:244
bool hasUCN() const
Returns true if this token contains a universal character name.
Definition: Token.h:294
bool isPreprocessedOutput() const
Returns true if the preprocessor is responsible for generating output, false if it is producing token...
Definition: Preprocessor.h:880
void SetKeepWhitespaceMode(bool Val)
SetKeepWhitespaceMode - This method lets clients enable or disable whitespace retention mode...
Definition: Lexer.h:213
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
bool hasLeadingSpace() const
Return true if this token has whitespace before it.
Definition: Token.h:270
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
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:3184
void startToken()
Reset all flags to cleared.
Definition: Token.h:169
std::pair< FileID, unsigned > getDecomposedLoc(SourceLocation Loc) const
Decompose the specified location into a raw FileID + Offset pair.
static std::string Stringify(StringRef Str, bool Charify=false)
Stringify - Convert the specified string into a C string by i) escaping &#39;\&#39; and " characters and ii) ...
Definition: Lexer.cpp:247
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:127
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.