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 /// 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 /// 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 unsigned Lexer::getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo,
712  const SourceManager &SM,
713  const LangOptions &LangOpts) {
714  // Figure out how many physical characters away the specified expansion
715  // character is. This needs to take into consideration newlines and
716  // trigraphs.
717  bool Invalid = false;
718  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
719 
720  // If they request the first char of the token, we're trivially done.
721  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
722  return 0;
723 
724  unsigned PhysOffset = 0;
725 
726  // The usual case is that tokens don't contain anything interesting. Skip
727  // over the uninteresting characters. If a token only consists of simple
728  // chars, this method is extremely fast.
729  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
730  if (CharNo == 0)
731  return PhysOffset;
732  ++TokPtr;
733  --CharNo;
734  ++PhysOffset;
735  }
736 
737  // If we have a character that may be a trigraph or escaped newline, use a
738  // lexer to parse it correctly.
739  for (; CharNo; --CharNo) {
740  unsigned Size;
741  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
742  TokPtr += Size;
743  PhysOffset += Size;
744  }
745 
746  // Final detail: if we end up on an escaped newline, we want to return the
747  // location of the actual byte of the token. For example foo<newline>bar
748  // advanced by 3 should return the location of b, not of \\. One compounding
749  // detail of this is that the escape may be made by a trigraph.
750  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
751  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
752 
753  return PhysOffset;
754 }
755 
756 /// Computes the source location just past the end of the
757 /// token at this source location.
758 ///
759 /// This routine can be used to produce a source location that
760 /// points just past the end of the token referenced by \p Loc, and
761 /// is generally used when a diagnostic needs to point just after a
762 /// token where it expected something different that it received. If
763 /// the returned source location would not be meaningful (e.g., if
764 /// it points into a macro), this routine returns an invalid
765 /// source location.
766 ///
767 /// \param Offset an offset from the end of the token, where the source
768 /// location should refer to. The default offset (0) produces a source
769 /// location pointing just past the end of the token; an offset of 1 produces
770 /// a source location pointing to the last character in the token, etc.
772  const SourceManager &SM,
773  const LangOptions &LangOpts) {
774  if (Loc.isInvalid())
775  return {};
776 
777  if (Loc.isMacroID()) {
778  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
779  return {}; // Points inside the macro expansion.
780  }
781 
782  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
783  if (Len > Offset)
784  Len = Len - Offset;
785  else
786  return Loc;
787 
788  return Loc.getLocWithOffset(Len);
789 }
790 
791 /// Returns true if the given MacroID location points at the first
792 /// token of the macro expansion.
794  const SourceManager &SM,
795  const LangOptions &LangOpts,
796  SourceLocation *MacroBegin) {
797  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
798 
799  SourceLocation expansionLoc;
800  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
801  return false;
802 
803  if (expansionLoc.isFileID()) {
804  // No other macro expansions, this is the first.
805  if (MacroBegin)
806  *MacroBegin = expansionLoc;
807  return true;
808  }
809 
810  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
811 }
812 
813 /// Returns true if the given MacroID location points at the last
814 /// token of the macro expansion.
816  const SourceManager &SM,
817  const LangOptions &LangOpts,
818  SourceLocation *MacroEnd) {
819  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
820 
821  SourceLocation spellLoc = SM.getSpellingLoc(loc);
822  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
823  if (tokLen == 0)
824  return false;
825 
826  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
827  SourceLocation expansionLoc;
828  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
829  return false;
830 
831  if (expansionLoc.isFileID()) {
832  // No other macro expansions.
833  if (MacroEnd)
834  *MacroEnd = expansionLoc;
835  return true;
836  }
837 
838  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
839 }
840 
842  const SourceManager &SM,
843  const LangOptions &LangOpts) {
844  SourceLocation Begin = Range.getBegin();
845  SourceLocation End = Range.getEnd();
846  assert(Begin.isFileID() && End.isFileID());
847  if (Range.isTokenRange()) {
848  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
849  if (End.isInvalid())
850  return {};
851  }
852 
853  // Break down the source locations.
854  FileID FID;
855  unsigned BeginOffs;
856  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
857  if (FID.isInvalid())
858  return {};
859 
860  unsigned EndOffs;
861  if (!SM.isInFileID(End, FID, &EndOffs) ||
862  BeginOffs > EndOffs)
863  return {};
864 
865  return CharSourceRange::getCharRange(Begin, End);
866 }
867 
869  const SourceManager &SM,
870  const LangOptions &LangOpts) {
871  SourceLocation Begin = Range.getBegin();
872  SourceLocation End = Range.getEnd();
873  if (Begin.isInvalid() || End.isInvalid())
874  return {};
875 
876  if (Begin.isFileID() && End.isFileID())
877  return makeRangeFromFileLocs(Range, SM, LangOpts);
878 
879  if (Begin.isMacroID() && End.isFileID()) {
880  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
881  return {};
882  Range.setBegin(Begin);
883  return makeRangeFromFileLocs(Range, SM, LangOpts);
884  }
885 
886  if (Begin.isFileID() && End.isMacroID()) {
887  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
888  &End)) ||
889  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
890  &End)))
891  return {};
892  Range.setEnd(End);
893  return makeRangeFromFileLocs(Range, SM, LangOpts);
894  }
895 
896  assert(Begin.isMacroID() && End.isMacroID());
897  SourceLocation MacroBegin, MacroEnd;
898  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
899  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
900  &MacroEnd)) ||
901  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
902  &MacroEnd)))) {
903  Range.setBegin(MacroBegin);
904  Range.setEnd(MacroEnd);
905  return makeRangeFromFileLocs(Range, SM, LangOpts);
906  }
907 
908  bool Invalid = false;
909  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
910  &Invalid);
911  if (Invalid)
912  return {};
913 
914  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
915  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
916  &Invalid);
917  if (Invalid)
918  return {};
919 
920  if (EndEntry.getExpansion().isMacroArgExpansion() &&
921  BeginEntry.getExpansion().getExpansionLocStart() ==
922  EndEntry.getExpansion().getExpansionLocStart()) {
923  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
924  Range.setEnd(SM.getImmediateSpellingLoc(End));
925  return makeFileCharRange(Range, SM, LangOpts);
926  }
927  }
928 
929  return {};
930 }
931 
933  const SourceManager &SM,
934  const LangOptions &LangOpts,
935  bool *Invalid) {
936  Range = makeFileCharRange(Range, SM, LangOpts);
937  if (Range.isInvalid()) {
938  if (Invalid) *Invalid = true;
939  return {};
940  }
941 
942  // Break down the source location.
943  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
944  if (beginInfo.first.isInvalid()) {
945  if (Invalid) *Invalid = true;
946  return {};
947  }
948 
949  unsigned EndOffs;
950  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
951  beginInfo.second > EndOffs) {
952  if (Invalid) *Invalid = true;
953  return {};
954  }
955 
956  // Try to the load the file buffer.
957  bool invalidTemp = false;
958  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
959  if (invalidTemp) {
960  if (Invalid) *Invalid = true;
961  return {};
962  }
963 
964  if (Invalid) *Invalid = false;
965  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
966 }
967 
969  const SourceManager &SM,
970  const LangOptions &LangOpts) {
971  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
972 
973  // Find the location of the immediate macro expansion.
974  while (true) {
975  FileID FID = SM.getFileID(Loc);
976  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
977  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
978  Loc = Expansion.getExpansionLocStart();
979  if (!Expansion.isMacroArgExpansion())
980  break;
981 
982  // For macro arguments we need to check that the argument did not come
983  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
984 
985  // Loc points to the argument id of the macro definition, move to the
986  // macro expansion.
987  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
988  SourceLocation SpellLoc = Expansion.getSpellingLoc();
989  if (SpellLoc.isFileID())
990  break; // No inner macro.
991 
992  // If spelling location resides in the same FileID as macro expansion
993  // location, it means there is no inner macro.
994  FileID MacroFID = SM.getFileID(Loc);
995  if (SM.isInFileID(SpellLoc, MacroFID))
996  break;
997 
998  // Argument came from inner macro.
999  Loc = SpellLoc;
1000  }
1001 
1002  // Find the spelling location of the start of the non-argument expansion
1003  // range. This is where the macro name was spelled in order to begin
1004  // expanding this macro.
1005  Loc = SM.getSpellingLoc(Loc);
1006 
1007  // Dig out the buffer where the macro name was spelled and the extents of the
1008  // name so that we can render it into the expansion note.
1009  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1010  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1011  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1012  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1013 }
1014 
1016  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1017  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
1018  // Walk past macro argument expanions.
1019  while (SM.isMacroArgExpansion(Loc))
1020  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1021 
1022  // If the macro's spelling has no FileID, then it's actually a token paste
1023  // or stringization (or similar) and not a macro at all.
1024  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1025  return {};
1026 
1027  // Find the spelling location of the start of the non-argument expansion
1028  // range. This is where the macro name was spelled in order to begin
1029  // expanding this macro.
1031 
1032  // Dig out the buffer where the macro name was spelled and the extents of the
1033  // name so that we can render it into the expansion note.
1034  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1035  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1036  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1037  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1038 }
1039 
1040 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1041  return isIdentifierBody(c, LangOpts.DollarIdents);
1042 }
1043 
1044 bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1045  assert(isVerticalWhitespace(Str[0]));
1046  if (Str - 1 < BufferStart)
1047  return false;
1048 
1049  if ((Str[0] == '\n' && Str[-1] == '\r') ||
1050  (Str[0] == '\r' && Str[-1] == '\n')) {
1051  if (Str - 2 < BufferStart)
1052  return false;
1053  --Str;
1054  }
1055  --Str;
1056 
1057  // Rewind to first non-space character:
1058  while (Str > BufferStart && isHorizontalWhitespace(*Str))
1059  --Str;
1060 
1061  return *Str == '\\';
1062 }
1063 
1065  const SourceManager &SM) {
1066  if (Loc.isInvalid() || Loc.isMacroID())
1067  return {};
1068  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1069  if (LocInfo.first.isInvalid())
1070  return {};
1071  bool Invalid = false;
1072  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1073  if (Invalid)
1074  return {};
1075  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1076  if (!Line)
1077  return {};
1078  StringRef Rest = Buffer.substr(Line - Buffer.data());
1079  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1080  return NumWhitespaceChars == StringRef::npos
1081  ? ""
1082  : Rest.take_front(NumWhitespaceChars);
1083 }
1084 
1085 //===----------------------------------------------------------------------===//
1086 // Diagnostics forwarding code.
1087 //===----------------------------------------------------------------------===//
1088 
1089 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1090 /// lexer buffer was all expanded at a single point, perform the mapping.
1091 /// This is currently only used for _Pragma implementation, so it is the slow
1092 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1093 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1094  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1096  SourceLocation FileLoc,
1097  unsigned CharNo, unsigned TokLen) {
1098  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1099 
1100  // Otherwise, we're lexing "mapped tokens". This is used for things like
1101  // _Pragma handling. Combine the expansion location of FileLoc with the
1102  // spelling location.
1104 
1105  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1106  // characters come from spelling(FileLoc)+Offset.
1107  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1108  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1109 
1110  // Figure out the expansion loc range, which is the range covered by the
1111  // original _Pragma(...) sequence.
1112  CharSourceRange II = SM.getImmediateExpansionRange(FileLoc);
1113 
1114  return SM.createExpansionLoc(SpellingLoc, II.getBegin(), II.getEnd(), TokLen);
1115 }
1116 
1117 /// getSourceLocation - Return a source location identifier for the specified
1118 /// offset in the current file.
1120  unsigned TokLen) const {
1121  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1122  "Location out of range for this buffer!");
1123 
1124  // In the normal case, we're just lexing from a simple file buffer, return
1125  // the file id from FileLoc with the offset specified.
1126  unsigned CharNo = Loc-BufferStart;
1127  if (FileLoc.isFileID())
1128  return FileLoc.getLocWithOffset(CharNo);
1129 
1130  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1131  // tokens are lexed from where the _Pragma was defined.
1132  assert(PP && "This doesn't work on raw lexers");
1133  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1134 }
1135 
1136 /// Diag - Forwarding function for diagnostics. This translate a source
1137 /// position in the current buffer into a SourceLocation object for rendering.
1138 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1139  return PP->Diag(getSourceLocation(Loc), DiagID);
1140 }
1141 
1142 //===----------------------------------------------------------------------===//
1143 // Trigraph and Escaped Newline Handling Code.
1144 //===----------------------------------------------------------------------===//
1145 
1146 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1147 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1148 static char GetTrigraphCharForLetter(char Letter) {
1149  switch (Letter) {
1150  default: return 0;
1151  case '=': return '#';
1152  case ')': return ']';
1153  case '(': return '[';
1154  case '!': return '|';
1155  case '\'': return '^';
1156  case '>': return '}';
1157  case '/': return '\\';
1158  case '<': return '{';
1159  case '-': return '~';
1160  }
1161 }
1162 
1163 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1164 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1165 /// return the result character. Finally, emit a warning about trigraph use
1166 /// whether trigraphs are enabled or not.
1167 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1168  char Res = GetTrigraphCharForLetter(*CP);
1169  if (!Res || !L) return Res;
1170 
1171  if (!L->getLangOpts().Trigraphs) {
1172  if (!L->isLexingRawMode())
1173  L->Diag(CP-2, diag::trigraph_ignored);
1174  return 0;
1175  }
1176 
1177  if (!L->isLexingRawMode())
1178  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1179  return Res;
1180 }
1181 
1182 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1183 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1184 /// trigraph equivalent on entry to this function.
1185 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1186  unsigned Size = 0;
1187  while (isWhitespace(Ptr[Size])) {
1188  ++Size;
1189 
1190  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1191  continue;
1192 
1193  // If this is a \r\n or \n\r, skip the other half.
1194  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1195  Ptr[Size-1] != Ptr[Size])
1196  ++Size;
1197 
1198  return Size;
1199  }
1200 
1201  // Not an escaped newline, must be a \t or something else.
1202  return 0;
1203 }
1204 
1205 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1206 /// them), skip over them and return the first non-escaped-newline found,
1207 /// otherwise return P.
1208 const char *Lexer::SkipEscapedNewLines(const char *P) {
1209  while (true) {
1210  const char *AfterEscape;
1211  if (*P == '\\') {
1212  AfterEscape = P+1;
1213  } else if (*P == '?') {
1214  // If not a trigraph for escape, bail out.
1215  if (P[1] != '?' || P[2] != '/')
1216  return P;
1217  // FIXME: Take LangOpts into account; the language might not
1218  // support trigraphs.
1219  AfterEscape = P+3;
1220  } else {
1221  return P;
1222  }
1223 
1224  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1225  if (NewLineSize == 0) return P;
1226  P = AfterEscape+NewLineSize;
1227  }
1228 }
1229 
1231  const SourceManager &SM,
1232  const LangOptions &LangOpts) {
1233  if (Loc.isMacroID()) {
1234  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1235  return None;
1236  }
1237  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1238 
1239  // Break down the source location.
1240  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1241 
1242  // Try to load the file buffer.
1243  bool InvalidTemp = false;
1244  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1245  if (InvalidTemp)
1246  return None;
1247 
1248  const char *TokenBegin = File.data() + LocInfo.second;
1249 
1250  // Lex from the start of the given location.
1251  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1252  TokenBegin, File.end());
1253  // Find the token.
1254  Token Tok;
1255  lexer.LexFromRawLexer(Tok);
1256  return Tok;
1257 }
1258 
1259 /// Checks that the given token is the first token that occurs after the
1260 /// given location (this excludes comments and whitespace). Returns the location
1261 /// immediately after the specified token. If the token is not found or the
1262 /// location is inside a macro, the returned source location will be invalid.
1264  SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1265  const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1266  Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1267  if (!Tok || Tok->isNot(TKind))
1268  return {};
1269  SourceLocation TokenLoc = Tok->getLocation();
1270 
1271  // Calculate how much whitespace needs to be skipped if any.
1272  unsigned NumWhitespaceChars = 0;
1273  if (SkipTrailingWhitespaceAndNewLine) {
1274  const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1275  unsigned char C = *TokenEnd;
1276  while (isHorizontalWhitespace(C)) {
1277  C = *(++TokenEnd);
1278  NumWhitespaceChars++;
1279  }
1280 
1281  // Skip \r, \n, \r\n, or \n\r
1282  if (C == '\n' || C == '\r') {
1283  char PrevC = C;
1284  C = *(++TokenEnd);
1285  NumWhitespaceChars++;
1286  if ((C == '\n' || C == '\r') && C != PrevC)
1287  NumWhitespaceChars++;
1288  }
1289  }
1290 
1291  return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1292 }
1293 
1294 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1295 /// get its size, and return it. This is tricky in several cases:
1296 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1297 /// then either return the trigraph (skipping 3 chars) or the '?',
1298 /// depending on whether trigraphs are enabled or not.
1299 /// 2. If this is an escaped newline (potentially with whitespace between
1300 /// the backslash and newline), implicitly skip the newline and return
1301 /// the char after it.
1302 ///
1303 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1304 /// know that we can accumulate into Size, and that we have already incremented
1305 /// Ptr by Size bytes.
1306 ///
1307 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1308 /// be updated to match.
1309 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1310  Token *Tok) {
1311  // If we have a slash, look for an escaped newline.
1312  if (Ptr[0] == '\\') {
1313  ++Size;
1314  ++Ptr;
1315 Slash:
1316  // Common case, backslash-char where the char is not whitespace.
1317  if (!isWhitespace(Ptr[0])) return '\\';
1318 
1319  // See if we have optional whitespace characters between the slash and
1320  // newline.
1321  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1322  // Remember that this token needs to be cleaned.
1323  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1324 
1325  // Warn if there was whitespace between the backslash and newline.
1326  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1327  Diag(Ptr, diag::backslash_newline_space);
1328 
1329  // Found backslash<whitespace><newline>. Parse the char after it.
1330  Size += EscapedNewLineSize;
1331  Ptr += EscapedNewLineSize;
1332 
1333  // Use slow version to accumulate a correct size field.
1334  return getCharAndSizeSlow(Ptr, Size, Tok);
1335  }
1336 
1337  // Otherwise, this is not an escaped newline, just return the slash.
1338  return '\\';
1339  }
1340 
1341  // If this is a trigraph, process it.
1342  if (Ptr[0] == '?' && Ptr[1] == '?') {
1343  // If this is actually a legal trigraph (not something like "??x"), emit
1344  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1345  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1346  // Remember that this token needs to be cleaned.
1347  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1348 
1349  Ptr += 3;
1350  Size += 3;
1351  if (C == '\\') goto Slash;
1352  return C;
1353  }
1354  }
1355 
1356  // If this is neither, return a single character.
1357  ++Size;
1358  return *Ptr;
1359 }
1360 
1361 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1362 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1363 /// and that we have already incremented Ptr by Size bytes.
1364 ///
1365 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1366 /// be updated to match.
1367 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1368  const LangOptions &LangOpts) {
1369  // If we have a slash, look for an escaped newline.
1370  if (Ptr[0] == '\\') {
1371  ++Size;
1372  ++Ptr;
1373 Slash:
1374  // Common case, backslash-char where the char is not whitespace.
1375  if (!isWhitespace(Ptr[0])) return '\\';
1376 
1377  // See if we have optional whitespace characters followed by a newline.
1378  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1379  // Found backslash<whitespace><newline>. Parse the char after it.
1380  Size += EscapedNewLineSize;
1381  Ptr += EscapedNewLineSize;
1382 
1383  // Use slow version to accumulate a correct size field.
1384  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1385  }
1386 
1387  // Otherwise, this is not an escaped newline, just return the slash.
1388  return '\\';
1389  }
1390 
1391  // If this is a trigraph, process it.
1392  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1393  // If this is actually a legal trigraph (not something like "??x"), return
1394  // it.
1395  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1396  Ptr += 3;
1397  Size += 3;
1398  if (C == '\\') goto Slash;
1399  return C;
1400  }
1401  }
1402 
1403  // If this is neither, return a single character.
1404  ++Size;
1405  return *Ptr;
1406 }
1407 
1408 //===----------------------------------------------------------------------===//
1409 // Helper methods for lexing.
1410 //===----------------------------------------------------------------------===//
1411 
1412 /// Routine that indiscriminately sets the offset into the source file.
1413 void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1414  BufferPtr = BufferStart + Offset;
1415  if (BufferPtr > BufferEnd)
1416  BufferPtr = BufferEnd;
1417  // FIXME: What exactly does the StartOfLine bit mean? There are two
1418  // possible meanings for the "start" of the line: the first token on the
1419  // unexpanded line, or the first token on the expanded line.
1420  IsAtStartOfLine = StartOfLine;
1421  IsAtPhysicalStartOfLine = StartOfLine;
1422 }
1423 
1424 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1425  if (LangOpts.AsmPreprocessor) {
1426  return false;
1427  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1428  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1430  return C11AllowedIDChars.contains(C);
1431  } else if (LangOpts.CPlusPlus) {
1432  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1434  return CXX03AllowedIDChars.contains(C);
1435  } else {
1436  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1438  return C99AllowedIDChars.contains(C);
1439  }
1440 }
1441 
1442 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1443  assert(isAllowedIDChar(C, LangOpts));
1444  if (LangOpts.AsmPreprocessor) {
1445  return false;
1446  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1447  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1449  return !C11DisallowedInitialIDChars.contains(C);
1450  } else if (LangOpts.CPlusPlus) {
1451  return true;
1452  } else {
1453  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1455  return !C99DisallowedInitialIDChars.contains(C);
1456  }
1457 }
1458 
1459 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1460  const char *End) {
1462  L.getSourceLocation(End));
1463 }
1464 
1465 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1466  CharSourceRange Range, bool IsFirst) {
1467  // Check C99 compatibility.
1468  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1469  enum {
1470  CannotAppearInIdentifier = 0,
1471  CannotStartIdentifier
1472  };
1473 
1474  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1476  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1478  if (!C99AllowedIDChars.contains(C)) {
1479  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1480  << Range
1481  << CannotAppearInIdentifier;
1482  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1483  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1484  << Range
1485  << CannotStartIdentifier;
1486  }
1487  }
1488 
1489  // Check C++98 compatibility.
1490  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1491  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1493  if (!CXX03AllowedIDChars.contains(C)) {
1494  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1495  << Range;
1496  }
1497  }
1498 }
1499 
1500 /// After encountering UTF-8 character C and interpreting it as an identifier
1501 /// character, check whether it's a homoglyph for a common non-identifier
1502 /// source character that is unlikely to be an intentional identifier
1503 /// character and warn if so.
1504 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1505  CharSourceRange Range) {
1506  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1507  struct HomoglyphPair {
1508  uint32_t Character;
1509  char LooksLike;
1510  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1511  };
1512  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1513  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1514  {U'\u037e', ';'}, // GREEK QUESTION MARK
1515  {U'\u2212', '-'}, // MINUS SIGN
1516  {U'\u2215', '/'}, // DIVISION SLASH
1517  {U'\u2216', '\\'}, // SET MINUS
1518  {U'\u2217', '*'}, // ASTERISK OPERATOR
1519  {U'\u2223', '|'}, // DIVIDES
1520  {U'\u2227', '^'}, // LOGICAL AND
1521  {U'\u2236', ':'}, // RATIO
1522  {U'\u223c', '~'}, // TILDE OPERATOR
1523  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1524  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1525  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1526  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1527  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1528  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1529  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1530  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1531  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1532  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1533  {U'\uff0c', ','}, // FULLWIDTH COMMA
1534  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1535  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1536  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1537  {U'\uff1a', ':'}, // FULLWIDTH COLON
1538  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1539  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1540  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1541  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1542  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1543  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1544  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1545  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1546  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1547  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1548  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1549  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1550  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1551  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1552  {0, 0}
1553  };
1554  auto Homoglyph =
1555  std::lower_bound(std::begin(SortedHomoglyphs),
1556  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1557  if (Homoglyph->Character == C) {
1558  llvm::SmallString<5> CharBuf;
1559  {
1560  llvm::raw_svector_ostream CharOS(CharBuf);
1561  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1562  }
1563  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1564  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1565  << Range << CharBuf << LooksLikeStr;
1566  }
1567 }
1568 
1569 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1570  Token &Result) {
1571  const char *UCNPtr = CurPtr + Size;
1572  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1573  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1574  return false;
1575 
1576  if (!isLexingRawMode())
1577  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1578  makeCharRange(*this, CurPtr, UCNPtr),
1579  /*IsFirst=*/false);
1580 
1581  Result.setFlag(Token::HasUCN);
1582  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1583  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1584  CurPtr = UCNPtr;
1585  else
1586  while (CurPtr != UCNPtr)
1587  (void)getAndAdvanceChar(CurPtr, Result);
1588  return true;
1589 }
1590 
1591 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1592  const char *UnicodePtr = CurPtr;
1593  llvm::UTF32 CodePoint;
1594  llvm::ConversionResult Result =
1595  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1596  (const llvm::UTF8 *)BufferEnd,
1597  &CodePoint,
1598  llvm::strictConversion);
1599  if (Result != llvm::conversionOK ||
1600  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1601  return false;
1602 
1603  if (!isLexingRawMode()) {
1604  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1605  makeCharRange(*this, CurPtr, UnicodePtr),
1606  /*IsFirst=*/false);
1607  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1608  makeCharRange(*this, CurPtr, UnicodePtr));
1609  }
1610 
1611  CurPtr = UnicodePtr;
1612  return true;
1613 }
1614 
1615 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1616  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1617  unsigned Size;
1618  unsigned char C = *CurPtr++;
1619  while (isIdentifierBody(C))
1620  C = *CurPtr++;
1621 
1622  --CurPtr; // Back up over the skipped character.
1623 
1624  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1625  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1626  //
1627  // TODO: Could merge these checks into an InfoTable flag to make the
1628  // comparison cheaper
1629  if (isASCII(C) && C != '\\' && C != '?' &&
1630  (C != '$' || !LangOpts.DollarIdents)) {
1631 FinishIdentifier:
1632  const char *IdStart = BufferPtr;
1633  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1634  Result.setRawIdentifierData(IdStart);
1635 
1636  // If we are in raw mode, return this identifier raw. There is no need to
1637  // look up identifier information or attempt to macro expand it.
1638  if (LexingRawMode)
1639  return true;
1640 
1641  // Fill in Result.IdentifierInfo and update the token kind,
1642  // looking up the identifier in the identifier table.
1643  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1644  // Note that we have to call PP->LookUpIdentifierInfo() even for code
1645  // completion, it writes IdentifierInfo into Result, and callers rely on it.
1646 
1647  // If the completion point is at the end of an identifier, we want to treat
1648  // the identifier as incomplete even if it resolves to a macro or a keyword.
1649  // This allows e.g. 'class^' to complete to 'classifier'.
1650  if (isCodeCompletionPoint(CurPtr)) {
1651  // Return the code-completion token.
1652  Result.setKind(tok::code_completion);
1653  // Skip the code-completion char and all immediate identifier characters.
1654  // This ensures we get consistent behavior when completing at any point in
1655  // an identifier (i.e. at the start, in the middle, at the end). Note that
1656  // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1657  // simpler.
1658  assert(*CurPtr == 0 && "Completion character must be 0");
1659  ++CurPtr;
1660  // Note that code completion token is not added as a separate character
1661  // when the completion point is at the end of the buffer. Therefore, we need
1662  // to check if the buffer has ended.
1663  if (CurPtr < BufferEnd) {
1664  while (isIdentifierBody(*CurPtr))
1665  ++CurPtr;
1666  }
1667  BufferPtr = CurPtr;
1668  return true;
1669  }
1670 
1671  // Finally, now that we know we have an identifier, pass this off to the
1672  // preprocessor, which may macro expand it or something.
1673  if (II->isHandleIdentifierCase())
1674  return PP->HandleIdentifier(Result);
1675 
1676  return true;
1677  }
1678 
1679  // Otherwise, $,\,? in identifier found. Enter slower path.
1680 
1681  C = getCharAndSize(CurPtr, Size);
1682  while (true) {
1683  if (C == '$') {
1684  // If we hit a $ and they are not supported in identifiers, we are done.
1685  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1686 
1687  // Otherwise, emit a diagnostic and continue.
1688  if (!isLexingRawMode())
1689  Diag(CurPtr, diag::ext_dollar_in_identifier);
1690  CurPtr = ConsumeChar(CurPtr, Size, Result);
1691  C = getCharAndSize(CurPtr, Size);
1692  continue;
1693  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1694  C = getCharAndSize(CurPtr, Size);
1695  continue;
1696  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1697  C = getCharAndSize(CurPtr, Size);
1698  continue;
1699  } else if (!isIdentifierBody(C)) {
1700  goto FinishIdentifier;
1701  }
1702 
1703  // Otherwise, this character is good, consume it.
1704  CurPtr = ConsumeChar(CurPtr, Size, Result);
1705 
1706  C = getCharAndSize(CurPtr, Size);
1707  while (isIdentifierBody(C)) {
1708  CurPtr = ConsumeChar(CurPtr, Size, Result);
1709  C = getCharAndSize(CurPtr, Size);
1710  }
1711  }
1712 }
1713 
1714 /// isHexaLiteral - Return true if Start points to a hex constant.
1715 /// in microsoft mode (where this is supposed to be several different tokens).
1716 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1717  unsigned Size;
1718  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1719  if (C1 != '0')
1720  return false;
1721  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1722  return (C2 == 'x' || C2 == 'X');
1723 }
1724 
1725 /// LexNumericConstant - Lex the remainder of a integer or floating point
1726 /// constant. From[-1] is the first character lexed. Return the end of the
1727 /// constant.
1728 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1729  unsigned Size;
1730  char C = getCharAndSize(CurPtr, Size);
1731  char PrevCh = 0;
1732  while (isPreprocessingNumberBody(C)) {
1733  CurPtr = ConsumeChar(CurPtr, Size, Result);
1734  PrevCh = C;
1735  C = getCharAndSize(CurPtr, Size);
1736  }
1737 
1738  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1739  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1740  // If we are in Microsoft mode, don't continue if the constant is hex.
1741  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1742  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1743  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1744  }
1745 
1746  // If we have a hex FP constant, continue.
1747  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1748  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1749  // not-quite-conforming extension. Only do so if this looks like it's
1750  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1751  bool IsHexFloat = true;
1752  if (!LangOpts.C99) {
1753  if (!isHexaLiteral(BufferPtr, LangOpts))
1754  IsHexFloat = false;
1755  else if (!getLangOpts().CPlusPlus17 &&
1756  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1757  IsHexFloat = false;
1758  }
1759  if (IsHexFloat)
1760  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1761  }
1762 
1763  // If we have a digit separator, continue.
1764  if (C == '\'' && getLangOpts().CPlusPlus14) {
1765  unsigned NextSize;
1766  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1767  if (isIdentifierBody(Next)) {
1768  if (!isLexingRawMode())
1769  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1770  CurPtr = ConsumeChar(CurPtr, Size, Result);
1771  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1772  return LexNumericConstant(Result, CurPtr);
1773  }
1774  }
1775 
1776  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1777  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1778  return LexNumericConstant(Result, CurPtr);
1779  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1780  return LexNumericConstant(Result, CurPtr);
1781 
1782  // Update the location of token as well as BufferPtr.
1783  const char *TokStart = BufferPtr;
1784  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1785  Result.setLiteralData(TokStart);
1786  return true;
1787 }
1788 
1789 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1790 /// in C++11, or warn on a ud-suffix in C++98.
1791 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1792  bool IsStringLiteral) {
1793  assert(getLangOpts().CPlusPlus);
1794 
1795  // Maximally munch an identifier.
1796  unsigned Size;
1797  char C = getCharAndSize(CurPtr, Size);
1798  bool Consumed = false;
1799 
1800  if (!isIdentifierHead(C)) {
1801  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1802  Consumed = true;
1803  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1804  Consumed = true;
1805  else
1806  return CurPtr;
1807  }
1808 
1809  if (!getLangOpts().CPlusPlus11) {
1810  if (!isLexingRawMode())
1811  Diag(CurPtr,
1812  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1813  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1815  return CurPtr;
1816  }
1817 
1818  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1819  // that does not start with an underscore is ill-formed. As a conforming
1820  // extension, we treat all such suffixes as if they had whitespace before
1821  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1822  // likely to be a ud-suffix than a macro, however, and accept that.
1823  if (!Consumed) {
1824  bool IsUDSuffix = false;
1825  if (C == '_')
1826  IsUDSuffix = true;
1827  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1828  // In C++1y, we need to look ahead a few characters to see if this is a
1829  // valid suffix for a string literal or a numeric literal (this could be
1830  // the 'operator""if' defining a numeric literal operator).
1831  const unsigned MaxStandardSuffixLength = 3;
1832  char Buffer[MaxStandardSuffixLength] = { C };
1833  unsigned Consumed = Size;
1834  unsigned Chars = 1;
1835  while (true) {
1836  unsigned NextSize;
1837  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1838  getLangOpts());
1839  if (!isIdentifierBody(Next)) {
1840  // End of suffix. Check whether this is on the whitelist.
1841  const StringRef CompleteSuffix(Buffer, Chars);
1843  CompleteSuffix);
1844  break;
1845  }
1846 
1847  if (Chars == MaxStandardSuffixLength)
1848  // Too long: can't be a standard suffix.
1849  break;
1850 
1851  Buffer[Chars++] = Next;
1852  Consumed += NextSize;
1853  }
1854  }
1855 
1856  if (!IsUDSuffix) {
1857  if (!isLexingRawMode())
1858  Diag(CurPtr, getLangOpts().MSVCCompat
1859  ? diag::ext_ms_reserved_user_defined_literal
1860  : diag::ext_reserved_user_defined_literal)
1862  return CurPtr;
1863  }
1864 
1865  CurPtr = ConsumeChar(CurPtr, Size, Result);
1866  }
1867 
1868  Result.setFlag(Token::HasUDSuffix);
1869  while (true) {
1870  C = getCharAndSize(CurPtr, Size);
1871  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1872  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1873  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1874  else break;
1875  }
1876 
1877  return CurPtr;
1878 }
1879 
1880 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1881 /// either " or L" or u8" or u" or U".
1882 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1883  tok::TokenKind Kind) {
1884  // Does this string contain the \0 character?
1885  const char *NulCharacter = nullptr;
1886 
1887  if (!isLexingRawMode() &&
1888  (Kind == tok::utf8_string_literal ||
1889  Kind == tok::utf16_string_literal ||
1890  Kind == tok::utf32_string_literal))
1891  Diag(BufferPtr, getLangOpts().CPlusPlus
1892  ? diag::warn_cxx98_compat_unicode_literal
1893  : diag::warn_c99_compat_unicode_literal);
1894 
1895  char C = getAndAdvanceChar(CurPtr, Result);
1896  while (C != '"') {
1897  // Skip escaped characters. Escaped newlines will already be processed by
1898  // getAndAdvanceChar.
1899  if (C == '\\')
1900  C = getAndAdvanceChar(CurPtr, Result);
1901 
1902  if (C == '\n' || C == '\r' || // Newline.
1903  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1904  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1905  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1906  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1907  return true;
1908  }
1909 
1910  if (C == 0) {
1911  if (isCodeCompletionPoint(CurPtr-1)) {
1913  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1914  cutOffLexing();
1915  return true;
1916  }
1917 
1918  NulCharacter = CurPtr-1;
1919  }
1920  C = getAndAdvanceChar(CurPtr, Result);
1921  }
1922 
1923  // If we are in C++11, lex the optional ud-suffix.
1924  if (getLangOpts().CPlusPlus)
1925  CurPtr = LexUDSuffix(Result, CurPtr, true);
1926 
1927  // If a nul character existed in the string, warn about it.
1928  if (NulCharacter && !isLexingRawMode())
1929  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1930 
1931  // Update the location of the token as well as the BufferPtr instance var.
1932  const char *TokStart = BufferPtr;
1933  FormTokenWithChars(Result, CurPtr, Kind);
1934  Result.setLiteralData(TokStart);
1935  return true;
1936 }
1937 
1938 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1939 /// having lexed R", LR", u8R", uR", or UR".
1940 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1941  tok::TokenKind Kind) {
1942  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1943  // Between the initial and final double quote characters of the raw string,
1944  // any transformations performed in phases 1 and 2 (trigraphs,
1945  // universal-character-names, and line splicing) are reverted.
1946 
1947  if (!isLexingRawMode())
1948  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1949 
1950  unsigned PrefixLen = 0;
1951 
1952  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1953  ++PrefixLen;
1954 
1955  // If the last character was not a '(', then we didn't lex a valid delimiter.
1956  if (CurPtr[PrefixLen] != '(') {
1957  if (!isLexingRawMode()) {
1958  const char *PrefixEnd = &CurPtr[PrefixLen];
1959  if (PrefixLen == 16) {
1960  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1961  } else {
1962  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1963  << StringRef(PrefixEnd, 1);
1964  }
1965  }
1966 
1967  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1968  // it's possible the '"' was intended to be part of the raw string, but
1969  // there's not much we can do about that.
1970  while (true) {
1971  char C = *CurPtr++;
1972 
1973  if (C == '"')
1974  break;
1975  if (C == 0 && CurPtr-1 == BufferEnd) {
1976  --CurPtr;
1977  break;
1978  }
1979  }
1980 
1981  FormTokenWithChars(Result, CurPtr, tok::unknown);
1982  return true;
1983  }
1984 
1985  // Save prefix and move CurPtr past it
1986  const char *Prefix = CurPtr;
1987  CurPtr += PrefixLen + 1; // skip over prefix and '('
1988 
1989  while (true) {
1990  char C = *CurPtr++;
1991 
1992  if (C == ')') {
1993  // Check for prefix match and closing quote.
1994  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
1995  CurPtr += PrefixLen + 1; // skip over prefix and '"'
1996  break;
1997  }
1998  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
1999  if (!isLexingRawMode())
2000  Diag(BufferPtr, diag::err_unterminated_raw_string)
2001  << StringRef(Prefix, PrefixLen);
2002  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2003  return true;
2004  }
2005  }
2006 
2007  // If we are in C++11, lex the optional ud-suffix.
2008  if (getLangOpts().CPlusPlus)
2009  CurPtr = LexUDSuffix(Result, CurPtr, true);
2010 
2011  // Update the location of token as well as BufferPtr.
2012  const char *TokStart = BufferPtr;
2013  FormTokenWithChars(Result, CurPtr, Kind);
2014  Result.setLiteralData(TokStart);
2015  return true;
2016 }
2017 
2018 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2019 /// after having lexed the '<' character. This is used for #include filenames.
2020 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2021  // Does this string contain the \0 character?
2022  const char *NulCharacter = nullptr;
2023  const char *AfterLessPos = CurPtr;
2024  char C = getAndAdvanceChar(CurPtr, Result);
2025  while (C != '>') {
2026  // Skip escaped characters. Escaped newlines will already be processed by
2027  // getAndAdvanceChar.
2028  if (C == '\\')
2029  C = getAndAdvanceChar(CurPtr, Result);
2030 
2031  if (C == '\n' || C == '\r' || // Newline.
2032  (C == 0 && (CurPtr-1 == BufferEnd || // End of file.
2033  isCodeCompletionPoint(CurPtr-1)))) {
2034  // If the filename is unterminated, then it must just be a lone <
2035  // character. Return this as such.
2036  FormTokenWithChars(Result, AfterLessPos, tok::less);
2037  return true;
2038  }
2039 
2040  if (C == 0) {
2041  NulCharacter = CurPtr-1;
2042  }
2043  C = getAndAdvanceChar(CurPtr, Result);
2044  }
2045 
2046  // If a nul character existed in the string, warn about it.
2047  if (NulCharacter && !isLexingRawMode())
2048  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2049 
2050  // Update the location of token as well as BufferPtr.
2051  const char *TokStart = BufferPtr;
2052  FormTokenWithChars(Result, CurPtr, tok::angle_string_literal);
2053  Result.setLiteralData(TokStart);
2054  return true;
2055 }
2056 
2057 /// LexCharConstant - Lex the remainder of a character constant, after having
2058 /// lexed either ' or L' or u8' or u' or U'.
2059 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2060  tok::TokenKind Kind) {
2061  // Does this character contain the \0 character?
2062  const char *NulCharacter = nullptr;
2063 
2064  if (!isLexingRawMode()) {
2065  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2066  Diag(BufferPtr, getLangOpts().CPlusPlus
2067  ? diag::warn_cxx98_compat_unicode_literal
2068  : diag::warn_c99_compat_unicode_literal);
2069  else if (Kind == tok::utf8_char_constant)
2070  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2071  }
2072 
2073  char C = getAndAdvanceChar(CurPtr, Result);
2074  if (C == '\'') {
2075  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2076  Diag(BufferPtr, diag::ext_empty_character);
2077  FormTokenWithChars(Result, CurPtr, tok::unknown);
2078  return true;
2079  }
2080 
2081  while (C != '\'') {
2082  // Skip escaped characters.
2083  if (C == '\\')
2084  C = getAndAdvanceChar(CurPtr, Result);
2085 
2086  if (C == '\n' || C == '\r' || // Newline.
2087  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2088  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2089  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2090  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2091  return true;
2092  }
2093 
2094  if (C == 0) {
2095  if (isCodeCompletionPoint(CurPtr-1)) {
2097  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2098  cutOffLexing();
2099  return true;
2100  }
2101 
2102  NulCharacter = CurPtr-1;
2103  }
2104  C = getAndAdvanceChar(CurPtr, Result);
2105  }
2106 
2107  // If we are in C++11, lex the optional ud-suffix.
2108  if (getLangOpts().CPlusPlus)
2109  CurPtr = LexUDSuffix(Result, CurPtr, false);
2110 
2111  // If a nul character existed in the character, warn about it.
2112  if (NulCharacter && !isLexingRawMode())
2113  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2114 
2115  // Update the location of token as well as BufferPtr.
2116  const char *TokStart = BufferPtr;
2117  FormTokenWithChars(Result, CurPtr, Kind);
2118  Result.setLiteralData(TokStart);
2119  return true;
2120 }
2121 
2122 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2123 /// Update BufferPtr to point to the next non-whitespace character and return.
2124 ///
2125 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2126 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2127  bool &TokAtPhysicalStartOfLine) {
2128  // Whitespace - Skip it, then return the token after the whitespace.
2129  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2130 
2131  unsigned char Char = *CurPtr;
2132 
2133  // Skip consecutive spaces efficiently.
2134  while (true) {
2135  // Skip horizontal whitespace very aggressively.
2136  while (isHorizontalWhitespace(Char))
2137  Char = *++CurPtr;
2138 
2139  // Otherwise if we have something other than whitespace, we're done.
2140  if (!isVerticalWhitespace(Char))
2141  break;
2142 
2144  // End of preprocessor directive line, let LexTokenInternal handle this.
2145  BufferPtr = CurPtr;
2146  return false;
2147  }
2148 
2149  // OK, but handle newline.
2150  SawNewline = true;
2151  Char = *++CurPtr;
2152  }
2153 
2154  // If the client wants us to return whitespace, return it now.
2155  if (isKeepWhitespaceMode()) {
2156  FormTokenWithChars(Result, CurPtr, tok::unknown);
2157  if (SawNewline) {
2158  IsAtStartOfLine = true;
2159  IsAtPhysicalStartOfLine = true;
2160  }
2161  // FIXME: The next token will not have LeadingSpace set.
2162  return true;
2163  }
2164 
2165  // If this isn't immediately after a newline, there is leading space.
2166  char PrevChar = CurPtr[-1];
2167  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2168 
2169  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2170  if (SawNewline) {
2171  Result.setFlag(Token::StartOfLine);
2172  TokAtPhysicalStartOfLine = true;
2173  }
2174 
2175  BufferPtr = CurPtr;
2176  return false;
2177 }
2178 
2179 /// We have just read the // characters from input. Skip until we find the
2180 /// newline character that terminates the comment. Then update BufferPtr and
2181 /// return.
2182 ///
2183 /// If we're in KeepCommentMode or any CommentHandler has inserted
2184 /// some tokens, this will store the first token and return true.
2185 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2186  bool &TokAtPhysicalStartOfLine) {
2187  // If Line comments aren't explicitly enabled for this language, emit an
2188  // extension warning.
2189  if (!LangOpts.LineComment && !isLexingRawMode()) {
2190  Diag(BufferPtr, diag::ext_line_comment);
2191 
2192  // Mark them enabled so we only emit one warning for this translation
2193  // unit.
2194  LangOpts.LineComment = true;
2195  }
2196 
2197  // Scan over the body of the comment. The common case, when scanning, is that
2198  // the comment contains normal ascii characters with nothing interesting in
2199  // them. As such, optimize for this case with the inner loop.
2200  //
2201  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2202  // character that ends the line comment.
2203  char C;
2204  while (true) {
2205  C = *CurPtr;
2206  // Skip over characters in the fast loop.
2207  while (C != 0 && // Potentially EOF.
2208  C != '\n' && C != '\r') // Newline or DOS-style newline.
2209  C = *++CurPtr;
2210 
2211  const char *NextLine = CurPtr;
2212  if (C != 0) {
2213  // We found a newline, see if it's escaped.
2214  const char *EscapePtr = CurPtr-1;
2215  bool HasSpace = false;
2216  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2217  --EscapePtr;
2218  HasSpace = true;
2219  }
2220 
2221  if (*EscapePtr == '\\')
2222  // Escaped newline.
2223  CurPtr = EscapePtr;
2224  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2225  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2226  // Trigraph-escaped newline.
2227  CurPtr = EscapePtr-2;
2228  else
2229  break; // This is a newline, we're done.
2230 
2231  // If there was space between the backslash and newline, warn about it.
2232  if (HasSpace && !isLexingRawMode())
2233  Diag(EscapePtr, diag::backslash_newline_space);
2234  }
2235 
2236  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2237  // properly decode the character. Read it in raw mode to avoid emitting
2238  // diagnostics about things like trigraphs. If we see an escaped newline,
2239  // we'll handle it below.
2240  const char *OldPtr = CurPtr;
2241  bool OldRawMode = isLexingRawMode();
2242  LexingRawMode = true;
2243  C = getAndAdvanceChar(CurPtr, Result);
2244  LexingRawMode = OldRawMode;
2245 
2246  // If we only read only one character, then no special handling is needed.
2247  // We're done and can skip forward to the newline.
2248  if (C != 0 && CurPtr == OldPtr+1) {
2249  CurPtr = NextLine;
2250  break;
2251  }
2252 
2253  // If we read multiple characters, and one of those characters was a \r or
2254  // \n, then we had an escaped newline within the comment. Emit diagnostic
2255  // unless the next line is also a // comment.
2256  if (CurPtr != OldPtr + 1 && C != '/' &&
2257  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2258  for (; OldPtr != CurPtr; ++OldPtr)
2259  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2260  // Okay, we found a // comment that ends in a newline, if the next
2261  // line is also a // comment, but has spaces, don't emit a diagnostic.
2262  if (isWhitespace(C)) {
2263  const char *ForwardPtr = CurPtr;
2264  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2265  ++ForwardPtr;
2266  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2267  break;
2268  }
2269 
2270  if (!isLexingRawMode())
2271  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2272  break;
2273  }
2274  }
2275 
2276  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2277  --CurPtr;
2278  break;
2279  }
2280 
2281  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2283  cutOffLexing();
2284  return false;
2285  }
2286  }
2287 
2288  // Found but did not consume the newline. Notify comment handlers about the
2289  // comment unless we're in a #if 0 block.
2290  if (PP && !isLexingRawMode() &&
2291  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2292  getSourceLocation(CurPtr)))) {
2293  BufferPtr = CurPtr;
2294  return true; // A token has to be returned.
2295  }
2296 
2297  // If we are returning comments as tokens, return this comment as a token.
2298  if (inKeepCommentMode())
2299  return SaveLineComment(Result, CurPtr);
2300 
2301  // If we are inside a preprocessor directive and we see the end of line,
2302  // return immediately, so that the lexer can return this as an EOD token.
2303  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2304  BufferPtr = CurPtr;
2305  return false;
2306  }
2307 
2308  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2309  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2310  // contribute to another token), it isn't needed for correctness. Note that
2311  // this is ok even in KeepWhitespaceMode, because we would have returned the
2312  /// comment above in that mode.
2313  ++CurPtr;
2314 
2315  // The next returned token is at the start of the line.
2316  Result.setFlag(Token::StartOfLine);
2317  TokAtPhysicalStartOfLine = true;
2318  // No leading whitespace seen so far.
2320  BufferPtr = CurPtr;
2321  return false;
2322 }
2323 
2324 /// If in save-comment mode, package up this Line comment in an appropriate
2325 /// way and return it.
2326 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2327  // If we're not in a preprocessor directive, just return the // comment
2328  // directly.
2329  FormTokenWithChars(Result, CurPtr, tok::comment);
2330 
2332  return true;
2333 
2334  // If this Line-style comment is in a macro definition, transmogrify it into
2335  // a C-style block comment.
2336  bool Invalid = false;
2337  std::string Spelling = PP->getSpelling(Result, &Invalid);
2338  if (Invalid)
2339  return true;
2340 
2341  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2342  Spelling[1] = '*'; // Change prefix to "/*".
2343  Spelling += "*/"; // add suffix.
2344 
2345  Result.setKind(tok::comment);
2346  PP->CreateString(Spelling, Result,
2347  Result.getLocation(), Result.getLocation());
2348  return true;
2349 }
2350 
2351 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2352 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2353 /// a diagnostic if so. We know that the newline is inside of a block comment.
2354 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2355  Lexer *L) {
2356  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2357 
2358  // Back up off the newline.
2359  --CurPtr;
2360 
2361  // If this is a two-character newline sequence, skip the other character.
2362  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2363  // \n\n or \r\r -> not escaped newline.
2364  if (CurPtr[0] == CurPtr[1])
2365  return false;
2366  // \n\r or \r\n -> skip the newline.
2367  --CurPtr;
2368  }
2369 
2370  // If we have horizontal whitespace, skip over it. We allow whitespace
2371  // between the slash and newline.
2372  bool HasSpace = false;
2373  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2374  --CurPtr;
2375  HasSpace = true;
2376  }
2377 
2378  // If we have a slash, we know this is an escaped newline.
2379  if (*CurPtr == '\\') {
2380  if (CurPtr[-1] != '*') return false;
2381  } else {
2382  // It isn't a slash, is it the ?? / trigraph?
2383  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2384  CurPtr[-3] != '*')
2385  return false;
2386 
2387  // This is the trigraph ending the comment. Emit a stern warning!
2388  CurPtr -= 2;
2389 
2390  // If no trigraphs are enabled, warn that we ignored this trigraph and
2391  // ignore this * character.
2392  if (!L->getLangOpts().Trigraphs) {
2393  if (!L->isLexingRawMode())
2394  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2395  return false;
2396  }
2397  if (!L->isLexingRawMode())
2398  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2399  }
2400 
2401  // Warn about having an escaped newline between the */ characters.
2402  if (!L->isLexingRawMode())
2403  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2404 
2405  // If there was space between the backslash and newline, warn about it.
2406  if (HasSpace && !L->isLexingRawMode())
2407  L->Diag(CurPtr, diag::backslash_newline_space);
2408 
2409  return true;
2410 }
2411 
2412 #ifdef __SSE2__
2413 #include <emmintrin.h>
2414 #elif __ALTIVEC__
2415 #include <altivec.h>
2416 #undef bool
2417 #endif
2418 
2419 /// We have just read from input the / and * characters that started a comment.
2420 /// Read until we find the * and / characters that terminate the comment.
2421 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2422 /// comments, because they cannot cause the comment to end. The only thing
2423 /// that can happen is the comment could end with an escaped newline between
2424 /// the terminating * and /.
2425 ///
2426 /// If we're in KeepCommentMode or any CommentHandler has inserted
2427 /// some tokens, this will store the first token and return true.
2428 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2429  bool &TokAtPhysicalStartOfLine) {
2430  // Scan one character past where we should, looking for a '/' character. Once
2431  // we find it, check to see if it was preceded by a *. This common
2432  // optimization helps people who like to put a lot of * characters in their
2433  // comments.
2434 
2435  // The first character we get with newlines and trigraphs skipped to handle
2436  // the degenerate /*/ case below correctly if the * has an escaped newline
2437  // after it.
2438  unsigned CharSize;
2439  unsigned char C = getCharAndSize(CurPtr, CharSize);
2440  CurPtr += CharSize;
2441  if (C == 0 && CurPtr == BufferEnd+1) {
2442  if (!isLexingRawMode())
2443  Diag(BufferPtr, diag::err_unterminated_block_comment);
2444  --CurPtr;
2445 
2446  // KeepWhitespaceMode should return this broken comment as a token. Since
2447  // it isn't a well formed comment, just return it as an 'unknown' token.
2448  if (isKeepWhitespaceMode()) {
2449  FormTokenWithChars(Result, CurPtr, tok::unknown);
2450  return true;
2451  }
2452 
2453  BufferPtr = CurPtr;
2454  return false;
2455  }
2456 
2457  // Check to see if the first character after the '/*' is another /. If so,
2458  // then this slash does not end the block comment, it is part of it.
2459  if (C == '/')
2460  C = *CurPtr++;
2461 
2462  while (true) {
2463  // Skip over all non-interesting characters until we find end of buffer or a
2464  // (probably ending) '/' character.
2465  if (CurPtr + 24 < BufferEnd &&
2466  // If there is a code-completion point avoid the fast scan because it
2467  // doesn't check for '\0'.
2468  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2469  // While not aligned to a 16-byte boundary.
2470  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2471  C = *CurPtr++;
2472 
2473  if (C == '/') goto FoundSlash;
2474 
2475 #ifdef __SSE2__
2476  __m128i Slashes = _mm_set1_epi8('/');
2477  while (CurPtr+16 <= BufferEnd) {
2478  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2479  Slashes));
2480  if (cmp != 0) {
2481  // Adjust the pointer to point directly after the first slash. It's
2482  // not necessary to set C here, it will be overwritten at the end of
2483  // the outer loop.
2484  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2485  goto FoundSlash;
2486  }
2487  CurPtr += 16;
2488  }
2489 #elif __ALTIVEC__
2490  __vector unsigned char Slashes = {
2491  '/', '/', '/', '/', '/', '/', '/', '/',
2492  '/', '/', '/', '/', '/', '/', '/', '/'
2493  };
2494  while (CurPtr+16 <= BufferEnd &&
2495  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2496  CurPtr += 16;
2497 #else
2498  // Scan for '/' quickly. Many block comments are very large.
2499  while (CurPtr[0] != '/' &&
2500  CurPtr[1] != '/' &&
2501  CurPtr[2] != '/' &&
2502  CurPtr[3] != '/' &&
2503  CurPtr+4 < BufferEnd) {
2504  CurPtr += 4;
2505  }
2506 #endif
2507 
2508  // It has to be one of the bytes scanned, increment to it and read one.
2509  C = *CurPtr++;
2510  }
2511 
2512  // Loop to scan the remainder.
2513  while (C != '/' && C != '\0')
2514  C = *CurPtr++;
2515 
2516  if (C == '/') {
2517  FoundSlash:
2518  if (CurPtr[-2] == '*') // We found the final */. We're done!
2519  break;
2520 
2521  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2522  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2523  // We found the final */, though it had an escaped newline between the
2524  // * and /. We're done!
2525  break;
2526  }
2527  }
2528  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2529  // If this is a /* inside of the comment, emit a warning. Don't do this
2530  // if this is a /*/, which will end the comment. This misses cases with
2531  // embedded escaped newlines, but oh well.
2532  if (!isLexingRawMode())
2533  Diag(CurPtr-1, diag::warn_nested_block_comment);
2534  }
2535  } else if (C == 0 && CurPtr == BufferEnd+1) {
2536  if (!isLexingRawMode())
2537  Diag(BufferPtr, diag::err_unterminated_block_comment);
2538  // Note: the user probably forgot a */. We could continue immediately
2539  // after the /*, but this would involve lexing a lot of what really is the
2540  // comment, which surely would confuse the parser.
2541  --CurPtr;
2542 
2543  // KeepWhitespaceMode should return this broken comment as a token. Since
2544  // it isn't a well formed comment, just return it as an 'unknown' token.
2545  if (isKeepWhitespaceMode()) {
2546  FormTokenWithChars(Result, CurPtr, tok::unknown);
2547  return true;
2548  }
2549 
2550  BufferPtr = CurPtr;
2551  return false;
2552  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2554  cutOffLexing();
2555  return false;
2556  }
2557 
2558  C = *CurPtr++;
2559  }
2560 
2561  // Notify comment handlers about the comment unless we're in a #if 0 block.
2562  if (PP && !isLexingRawMode() &&
2563  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2564  getSourceLocation(CurPtr)))) {
2565  BufferPtr = CurPtr;
2566  return true; // A token has to be returned.
2567  }
2568 
2569  // If we are returning comments as tokens, return this comment as a token.
2570  if (inKeepCommentMode()) {
2571  FormTokenWithChars(Result, CurPtr, tok::comment);
2572  return true;
2573  }
2574 
2575  // It is common for the tokens immediately after a /**/ comment to be
2576  // whitespace. Instead of going through the big switch, handle it
2577  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2578  // have already returned above with the comment as a token.
2579  if (isHorizontalWhitespace(*CurPtr)) {
2580  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2581  return false;
2582  }
2583 
2584  // Otherwise, just return so that the next character will be lexed as a token.
2585  BufferPtr = CurPtr;
2586  Result.setFlag(Token::LeadingSpace);
2587  return false;
2588 }
2589 
2590 //===----------------------------------------------------------------------===//
2591 // Primary Lexing Entry Points
2592 //===----------------------------------------------------------------------===//
2593 
2594 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2595 /// uninterpreted string. This switches the lexer out of directive mode.
2597  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2598  "Must be in a preprocessing directive!");
2599  Token Tmp;
2600 
2601  // CurPtr - Cache BufferPtr in an automatic variable.
2602  const char *CurPtr = BufferPtr;
2603  while (true) {
2604  char Char = getAndAdvanceChar(CurPtr, Tmp);
2605  switch (Char) {
2606  default:
2607  if (Result)
2608  Result->push_back(Char);
2609  break;
2610  case 0: // Null.
2611  // Found end of file?
2612  if (CurPtr-1 != BufferEnd) {
2613  if (isCodeCompletionPoint(CurPtr-1)) {
2615  cutOffLexing();
2616  return;
2617  }
2618 
2619  // Nope, normal character, continue.
2620  if (Result)
2621  Result->push_back(Char);
2622  break;
2623  }
2624  // FALL THROUGH.
2625  LLVM_FALLTHROUGH;
2626  case '\r':
2627  case '\n':
2628  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2629  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2630  BufferPtr = CurPtr-1;
2631 
2632  // Next, lex the character, which should handle the EOD transition.
2633  Lex(Tmp);
2634  if (Tmp.is(tok::code_completion)) {
2635  if (PP)
2637  Lex(Tmp);
2638  }
2639  assert(Tmp.is(tok::eod) && "Unexpected token!");
2640 
2641  // Finally, we're done;
2642  return;
2643  }
2644  }
2645 }
2646 
2647 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2648 /// condition, reporting diagnostics and handling other edge cases as required.
2649 /// This returns true if Result contains a token, false if PP.Lex should be
2650 /// called again.
2651 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2652  // If we hit the end of the file while parsing a preprocessor directive,
2653  // end the preprocessor directive first. The next token returned will
2654  // then be the end of file.
2656  // Done parsing the "line".
2658  // Update the location of token as well as BufferPtr.
2659  FormTokenWithChars(Result, CurPtr, tok::eod);
2660 
2661  // Restore comment saving mode, in case it was disabled for directive.
2662  if (PP)
2664  return true; // Have a token.
2665  }
2666 
2667  // If we are in raw mode, return this event as an EOF token. Let the caller
2668  // that put us in raw mode handle the event.
2669  if (isLexingRawMode()) {
2670  Result.startToken();
2671  BufferPtr = BufferEnd;
2672  FormTokenWithChars(Result, BufferEnd, tok::eof);
2673  return true;
2674  }
2675 
2676  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2678  ConditionalStack.clear();
2679  }
2680 
2681  // Issue diagnostics for unterminated #if and missing newline.
2682 
2683  // If we are in a #if directive, emit an error.
2684  while (!ConditionalStack.empty()) {
2685  if (PP->getCodeCompletionFileLoc() != FileLoc)
2686  PP->Diag(ConditionalStack.back().IfLoc,
2687  diag::err_pp_unterminated_conditional);
2688  ConditionalStack.pop_back();
2689  }
2690 
2691  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2692  // a pedwarn.
2693  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2694  DiagnosticsEngine &Diags = PP->getDiagnostics();
2695  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2696  unsigned DiagID;
2697 
2698  if (LangOpts.CPlusPlus11) {
2699  // C++11 [lex.phases] 2.2 p2
2700  // Prefer the C++98 pedantic compatibility warning over the generic,
2701  // non-extension, user-requested "missing newline at EOF" warning.
2702  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2703  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2704  } else {
2705  DiagID = diag::warn_no_newline_eof;
2706  }
2707  } else {
2708  DiagID = diag::ext_no_newline_eof;
2709  }
2710 
2711  Diag(BufferEnd, DiagID)
2712  << FixItHint::CreateInsertion(EndLoc, "\n");
2713  }
2714 
2715  BufferPtr = CurPtr;
2716 
2717  // Finally, let the preprocessor handle this.
2718  return PP->HandleEndOfFile(Result, isPragmaLexer());
2719 }
2720 
2721 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2722 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2723 /// else and 2 if there are no more tokens in the buffer controlled by the
2724 /// lexer.
2725 unsigned Lexer::isNextPPTokenLParen() {
2726  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2727 
2728  // Switch to 'skipping' mode. This will ensure that we can lex a token
2729  // without emitting diagnostics, disables macro expansion, and will cause EOF
2730  // to return an EOF token instead of popping the include stack.
2731  LexingRawMode = true;
2732 
2733  // Save state that can be changed while lexing so that we can restore it.
2734  const char *TmpBufferPtr = BufferPtr;
2735  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2736  bool atStartOfLine = IsAtStartOfLine;
2737  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2738  bool leadingSpace = HasLeadingSpace;
2739 
2740  Token Tok;
2741  Lex(Tok);
2742 
2743  // Restore state that may have changed.
2744  BufferPtr = TmpBufferPtr;
2745  ParsingPreprocessorDirective = inPPDirectiveMode;
2746  HasLeadingSpace = leadingSpace;
2747  IsAtStartOfLine = atStartOfLine;
2748  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2749 
2750  // Restore the lexer back to non-skipping mode.
2751  LexingRawMode = false;
2752 
2753  if (Tok.is(tok::eof))
2754  return 2;
2755  return Tok.is(tok::l_paren);
2756 }
2757 
2758 /// Find the end of a version control conflict marker.
2759 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2760  ConflictMarkerKind CMK) {
2761  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2762  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2763  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2764  size_t Pos = RestOfBuffer.find(Terminator);
2765  while (Pos != StringRef::npos) {
2766  // Must occur at start of line.
2767  if (Pos == 0 ||
2768  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2769  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2770  Pos = RestOfBuffer.find(Terminator);
2771  continue;
2772  }
2773  return RestOfBuffer.data()+Pos;
2774  }
2775  return nullptr;
2776 }
2777 
2778 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2779 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2780 /// and recover nicely. This returns true if it is a conflict marker and false
2781 /// if not.
2782 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2783  // Only a conflict marker if it starts at the beginning of a line.
2784  if (CurPtr != BufferStart &&
2785  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2786  return false;
2787 
2788  // Check to see if we have <<<<<<< or >>>>.
2789  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2790  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2791  return false;
2792 
2793  // If we have a situation where we don't care about conflict markers, ignore
2794  // it.
2795  if (CurrentConflictMarkerState || isLexingRawMode())
2796  return false;
2797 
2798  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2799 
2800  // Check to see if there is an ending marker somewhere in the buffer at the
2801  // start of a line to terminate this conflict marker.
2802  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2803  // We found a match. We are really in a conflict marker.
2804  // Diagnose this, and ignore to the end of line.
2805  Diag(CurPtr, diag::err_conflict_marker);
2806  CurrentConflictMarkerState = Kind;
2807 
2808  // Skip ahead to the end of line. We know this exists because the
2809  // end-of-conflict marker starts with \r or \n.
2810  while (*CurPtr != '\r' && *CurPtr != '\n') {
2811  assert(CurPtr != BufferEnd && "Didn't find end of line");
2812  ++CurPtr;
2813  }
2814  BufferPtr = CurPtr;
2815  return true;
2816  }
2817 
2818  // No end of conflict marker found.
2819  return false;
2820 }
2821 
2822 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2823 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2824 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2825 /// the line. This returns true if it is a conflict marker and false if not.
2826 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2827  // Only a conflict marker if it starts at the beginning of a line.
2828  if (CurPtr != BufferStart &&
2829  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2830  return false;
2831 
2832  // If we have a situation where we don't care about conflict markers, ignore
2833  // it.
2834  if (!CurrentConflictMarkerState || isLexingRawMode())
2835  return false;
2836 
2837  // Check to see if we have the marker (4 characters in a row).
2838  for (unsigned i = 1; i != 4; ++i)
2839  if (CurPtr[i] != CurPtr[0])
2840  return false;
2841 
2842  // If we do have it, search for the end of the conflict marker. This could
2843  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2844  // be the end of conflict marker.
2845  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2846  CurrentConflictMarkerState)) {
2847  CurPtr = End;
2848 
2849  // Skip ahead to the end of line.
2850  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2851  ++CurPtr;
2852 
2853  BufferPtr = CurPtr;
2854 
2855  // No longer in the conflict marker.
2856  CurrentConflictMarkerState = CMK_None;
2857  return true;
2858  }
2859 
2860  return false;
2861 }
2862 
2863 static const char *findPlaceholderEnd(const char *CurPtr,
2864  const char *BufferEnd) {
2865  if (CurPtr == BufferEnd)
2866  return nullptr;
2867  BufferEnd -= 1; // Scan until the second last character.
2868  for (; CurPtr != BufferEnd; ++CurPtr) {
2869  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2870  return CurPtr + 2;
2871  }
2872  return nullptr;
2873 }
2874 
2875 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2876  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2878  return false;
2879  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2880  if (!End)
2881  return false;
2882  const char *Start = CurPtr - 1;
2883  if (!LangOpts.AllowEditorPlaceholders)
2884  Diag(Start, diag::err_placeholder_in_source);
2885  Result.startToken();
2886  FormTokenWithChars(Result, End, tok::raw_identifier);
2887  Result.setRawIdentifierData(Start);
2888  PP->LookUpIdentifierInfo(Result);
2890  BufferPtr = End;
2891  return true;
2892 }
2893 
2894 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2895  if (PP && PP->isCodeCompletionEnabled()) {
2896  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2897  return Loc == PP->getCodeCompletionLoc();
2898  }
2899 
2900  return false;
2901 }
2902 
2903 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2904  Token *Result) {
2905  unsigned CharSize;
2906  char Kind = getCharAndSize(StartPtr, CharSize);
2907 
2908  unsigned NumHexDigits;
2909  if (Kind == 'u')
2910  NumHexDigits = 4;
2911  else if (Kind == 'U')
2912  NumHexDigits = 8;
2913  else
2914  return 0;
2915 
2916  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2917  if (Result && !isLexingRawMode())
2918  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2919  return 0;
2920  }
2921 
2922  const char *CurPtr = StartPtr + CharSize;
2923  const char *KindLoc = &CurPtr[-1];
2924 
2925  uint32_t CodePoint = 0;
2926  for (unsigned i = 0; i < NumHexDigits; ++i) {
2927  char C = getCharAndSize(CurPtr, CharSize);
2928 
2929  unsigned Value = llvm::hexDigitValue(C);
2930  if (Value == -1U) {
2931  if (Result && !isLexingRawMode()) {
2932  if (i == 0) {
2933  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2934  << StringRef(KindLoc, 1);
2935  } else {
2936  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2937 
2938  // If the user wrote \U1234, suggest a fixit to \u.
2939  if (i == 4 && NumHexDigits == 8) {
2940  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2941  Diag(KindLoc, diag::note_ucn_four_not_eight)
2942  << FixItHint::CreateReplacement(URange, "u");
2943  }
2944  }
2945  }
2946 
2947  return 0;
2948  }
2949 
2950  CodePoint <<= 4;
2951  CodePoint += Value;
2952 
2953  CurPtr += CharSize;
2954  }
2955 
2956  if (Result) {
2957  Result->setFlag(Token::HasUCN);
2958  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
2959  StartPtr = CurPtr;
2960  else
2961  while (StartPtr != CurPtr)
2962  (void)getAndAdvanceChar(StartPtr, *Result);
2963  } else {
2964  StartPtr = CurPtr;
2965  }
2966 
2967  // Don't apply C family restrictions to UCNs in assembly mode
2968  if (LangOpts.AsmPreprocessor)
2969  return CodePoint;
2970 
2971  // C99 6.4.3p2: A universal character name shall not specify a character whose
2972  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
2973  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
2974  // C++11 [lex.charset]p2: If the hexadecimal value for a
2975  // universal-character-name corresponds to a surrogate code point (in the
2976  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
2977  // if the hexadecimal value for a universal-character-name outside the
2978  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
2979  // string literal corresponds to a control character (in either of the
2980  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
2981  // basic source character set, the program is ill-formed.
2982  if (CodePoint < 0xA0) {
2983  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
2984  return CodePoint;
2985 
2986  // We don't use isLexingRawMode() here because we need to warn about bad
2987  // UCNs even when skipping preprocessing tokens in a #if block.
2988  if (Result && PP) {
2989  if (CodePoint < 0x20 || CodePoint >= 0x7F)
2990  Diag(BufferPtr, diag::err_ucn_control_character);
2991  else {
2992  char C = static_cast<char>(CodePoint);
2993  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
2994  }
2995  }
2996 
2997  return 0;
2998  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
2999  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3000  // We don't use isLexingRawMode() here because we need to diagnose bad
3001  // UCNs even when skipping preprocessing tokens in a #if block.
3002  if (Result && PP) {
3003  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3004  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3005  else
3006  Diag(BufferPtr, diag::err_ucn_escape_invalid);
3007  }
3008  return 0;
3009  }
3010 
3011  return CodePoint;
3012 }
3013 
3014 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3015  const char *CurPtr) {
3016  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3018  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3019  UnicodeWhitespaceChars.contains(C)) {
3020  Diag(BufferPtr, diag::ext_unicode_whitespace)
3021  << makeCharRange(*this, BufferPtr, CurPtr);
3022 
3023  Result.setFlag(Token::LeadingSpace);
3024  return true;
3025  }
3026  return false;
3027 }
3028 
3029 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3030  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3032  !PP->isPreprocessedOutput()) {
3034  makeCharRange(*this, BufferPtr, CurPtr),
3035  /*IsFirst=*/true);
3036  }
3037 
3038  MIOpt.ReadToken();
3039  return LexIdentifier(Result, CurPtr);
3040  }
3041 
3043  !PP->isPreprocessedOutput() &&
3044  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3045  // Non-ASCII characters tend to creep into source code unintentionally.
3046  // Instead of letting the parser complain about the unknown token,
3047  // just drop the character.
3048  // Note that we can /only/ do this when the non-ASCII character is actually
3049  // spelled as Unicode, not written as a UCN. The standard requires that
3050  // we not throw away any possible preprocessor tokens, but there's a
3051  // loophole in the mapping of Unicode characters to basic character set
3052  // characters that allows us to map these particular characters to, say,
3053  // whitespace.
3054  Diag(BufferPtr, diag::err_non_ascii)
3055  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3056 
3057  BufferPtr = CurPtr;
3058  return false;
3059  }
3060 
3061  // Otherwise, we have an explicit UCN or a character that's unlikely to show
3062  // up by accident.
3063  MIOpt.ReadToken();
3064  FormTokenWithChars(Result, CurPtr, tok::unknown);
3065  return true;
3066 }
3067 
3068 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3069  IsAtStartOfLine = Result.isAtStartOfLine();
3070  HasLeadingSpace = Result.hasLeadingSpace();
3071  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3072  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3073 }
3074 
3075 bool Lexer::Lex(Token &Result) {
3076  // Start a new token.
3077  Result.startToken();
3078 
3079  // Set up misc whitespace flags for LexTokenInternal.
3080  if (IsAtStartOfLine) {
3081  Result.setFlag(Token::StartOfLine);
3082  IsAtStartOfLine = false;
3083  }
3084 
3085  if (HasLeadingSpace) {
3086  Result.setFlag(Token::LeadingSpace);
3087  HasLeadingSpace = false;
3088  }
3089 
3090  if (HasLeadingEmptyMacro) {
3092  HasLeadingEmptyMacro = false;
3093  }
3094 
3095  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3096  IsAtPhysicalStartOfLine = false;
3097  bool isRawLex = isLexingRawMode();
3098  (void) isRawLex;
3099  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3100  // (After the LexTokenInternal call, the lexer might be destroyed.)
3101  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3102  return returnedToken;
3103 }
3104 
3105 /// LexTokenInternal - This implements a simple C family lexer. It is an
3106 /// extremely performance critical piece of code. This assumes that the buffer
3107 /// has a null character at the end of the file. This returns a preprocessing
3108 /// token, not a normal token, as such, it is an internal interface. It assumes
3109 /// that the Flags of result have been cleared before calling this.
3110 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3111 LexNextToken:
3112  // New token, can't need cleaning yet.
3114  Result.setIdentifierInfo(nullptr);
3115 
3116  // CurPtr - Cache BufferPtr in an automatic variable.
3117  const char *CurPtr = BufferPtr;
3118 
3119  // Small amounts of horizontal whitespace is very common between tokens.
3120  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
3121  ++CurPtr;
3122  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3123  ++CurPtr;
3124 
3125  // If we are keeping whitespace and other tokens, just return what we just
3126  // skipped. The next lexer invocation will return the token after the
3127  // whitespace.
3128  if (isKeepWhitespaceMode()) {
3129  FormTokenWithChars(Result, CurPtr, tok::unknown);
3130  // FIXME: The next token will not have LeadingSpace set.
3131  return true;
3132  }
3133 
3134  BufferPtr = CurPtr;
3135  Result.setFlag(Token::LeadingSpace);
3136  }
3137 
3138  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3139 
3140  // Read a character, advancing over it.
3141  char Char = getAndAdvanceChar(CurPtr, Result);
3143 
3144  switch (Char) {
3145  case 0: // Null.
3146  // Found end of file?
3147  if (CurPtr-1 == BufferEnd)
3148  return LexEndOfFile(Result, CurPtr-1);
3149 
3150  // Check if we are performing code completion.
3151  if (isCodeCompletionPoint(CurPtr-1)) {
3152  // Return the code-completion token.
3153  Result.startToken();
3154  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3155  return true;
3156  }
3157 
3158  if (!isLexingRawMode())
3159  Diag(CurPtr-1, diag::null_in_file);
3160  Result.setFlag(Token::LeadingSpace);
3161  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3162  return true; // KeepWhitespaceMode
3163 
3164  // We know the lexer hasn't changed, so just try again with this lexer.
3165  // (We manually eliminate the tail call to avoid recursion.)
3166  goto LexNextToken;
3167 
3168  case 26: // DOS & CP/M EOF: "^Z".
3169  // If we're in Microsoft extensions mode, treat this as end of file.
3170  if (LangOpts.MicrosoftExt) {
3171  if (!isLexingRawMode())
3172  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3173  return LexEndOfFile(Result, CurPtr-1);
3174  }
3175 
3176  // If Microsoft extensions are disabled, this is just random garbage.
3177  Kind = tok::unknown;
3178  break;
3179 
3180  case '\r':
3181  if (CurPtr[0] == '\n')
3182  Char = getAndAdvanceChar(CurPtr, Result);
3183  LLVM_FALLTHROUGH;
3184  case '\n':
3185  // If we are inside a preprocessor directive and we see the end of line,
3186  // we know we are done with the directive, so return an EOD token.
3188  // Done parsing the "line".
3190 
3191  // Restore comment saving mode, in case it was disabled for directive.
3192  if (PP)
3194 
3195  // Since we consumed a newline, we are back at the start of a line.
3196  IsAtStartOfLine = true;
3197  IsAtPhysicalStartOfLine = true;
3198 
3199  Kind = tok::eod;
3200  break;
3201  }
3202 
3203  // No leading whitespace seen so far.
3205 
3206  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3207  return true; // KeepWhitespaceMode
3208 
3209  // We only saw whitespace, so just try again with this lexer.
3210  // (We manually eliminate the tail call to avoid recursion.)
3211  goto LexNextToken;
3212  case ' ':
3213  case '\t':
3214  case '\f':
3215  case '\v':
3216  SkipHorizontalWhitespace:
3217  Result.setFlag(Token::LeadingSpace);
3218  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3219  return true; // KeepWhitespaceMode
3220 
3221  SkipIgnoredUnits:
3222  CurPtr = BufferPtr;
3223 
3224  // If the next token is obviously a // or /* */ comment, skip it efficiently
3225  // too (without going through the big switch stmt).
3226  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3227  LangOpts.LineComment &&
3228  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3229  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3230  return true; // There is a token to return.
3231  goto SkipIgnoredUnits;
3232  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3233  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3234  return true; // There is a token to return.
3235  goto SkipIgnoredUnits;
3236  } else if (isHorizontalWhitespace(*CurPtr)) {
3237  goto SkipHorizontalWhitespace;
3238  }
3239  // We only saw whitespace, so just try again with this lexer.
3240  // (We manually eliminate the tail call to avoid recursion.)
3241  goto LexNextToken;
3242 
3243  // C99 6.4.4.1: Integer Constants.
3244  // C99 6.4.4.2: Floating Constants.
3245  case '0': case '1': case '2': case '3': case '4':
3246  case '5': case '6': case '7': case '8': case '9':
3247  // Notify MIOpt that we read a non-whitespace/non-comment token.
3248  MIOpt.ReadToken();
3249  return LexNumericConstant(Result, CurPtr);
3250 
3251  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3252  // Notify MIOpt that we read a non-whitespace/non-comment token.
3253  MIOpt.ReadToken();
3254 
3255  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3256  Char = getCharAndSize(CurPtr, SizeTmp);
3257 
3258  // UTF-16 string literal
3259  if (Char == '"')
3260  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3261  tok::utf16_string_literal);
3262 
3263  // UTF-16 character constant
3264  if (Char == '\'')
3265  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3266  tok::utf16_char_constant);
3267 
3268  // UTF-16 raw string literal
3269  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3270  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3271  return LexRawStringLiteral(Result,
3272  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3273  SizeTmp2, Result),
3274  tok::utf16_string_literal);
3275 
3276  if (Char == '8') {
3277  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3278 
3279  // UTF-8 string literal
3280  if (Char2 == '"')
3281  return LexStringLiteral(Result,
3282  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3283  SizeTmp2, Result),
3284  tok::utf8_string_literal);
3285  if (Char2 == '\'' && LangOpts.CPlusPlus17)
3286  return LexCharConstant(
3287  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3288  SizeTmp2, Result),
3289  tok::utf8_char_constant);
3290 
3291  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3292  unsigned SizeTmp3;
3293  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3294  // UTF-8 raw string literal
3295  if (Char3 == '"') {
3296  return LexRawStringLiteral(Result,
3297  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3298  SizeTmp2, Result),
3299  SizeTmp3, Result),
3300  tok::utf8_string_literal);
3301  }
3302  }
3303  }
3304  }
3305 
3306  // treat u like the start of an identifier.
3307  return LexIdentifier(Result, CurPtr);
3308 
3309  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3310  // Notify MIOpt that we read a non-whitespace/non-comment token.
3311  MIOpt.ReadToken();
3312 
3313  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3314  Char = getCharAndSize(CurPtr, SizeTmp);
3315 
3316  // UTF-32 string literal
3317  if (Char == '"')
3318  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3319  tok::utf32_string_literal);
3320 
3321  // UTF-32 character constant
3322  if (Char == '\'')
3323  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3324  tok::utf32_char_constant);
3325 
3326  // UTF-32 raw string literal
3327  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3328  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3329  return LexRawStringLiteral(Result,
3330  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3331  SizeTmp2, Result),
3332  tok::utf32_string_literal);
3333  }
3334 
3335  // treat U like the start of an identifier.
3336  return LexIdentifier(Result, CurPtr);
3337 
3338  case 'R': // Identifier or C++0x raw string literal
3339  // Notify MIOpt that we read a non-whitespace/non-comment token.
3340  MIOpt.ReadToken();
3341 
3342  if (LangOpts.CPlusPlus11) {
3343  Char = getCharAndSize(CurPtr, SizeTmp);
3344 
3345  if (Char == '"')
3346  return LexRawStringLiteral(Result,
3347  ConsumeChar(CurPtr, SizeTmp, Result),
3348  tok::string_literal);
3349  }
3350 
3351  // treat R like the start of an identifier.
3352  return LexIdentifier(Result, CurPtr);
3353 
3354  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3355  // Notify MIOpt that we read a non-whitespace/non-comment token.
3356  MIOpt.ReadToken();
3357  Char = getCharAndSize(CurPtr, SizeTmp);
3358 
3359  // Wide string literal.
3360  if (Char == '"')
3361  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3362  tok::wide_string_literal);
3363 
3364  // Wide raw string literal.
3365  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3366  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3367  return LexRawStringLiteral(Result,
3368  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3369  SizeTmp2, Result),
3370  tok::wide_string_literal);
3371 
3372  // Wide character constant.
3373  if (Char == '\'')
3374  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3375  tok::wide_char_constant);
3376  // FALL THROUGH, treating L like the start of an identifier.
3377  LLVM_FALLTHROUGH;
3378 
3379  // C99 6.4.2: Identifiers.
3380  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3381  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3382  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3383  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3384  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3385  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3386  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3387  case 'v': case 'w': case 'x': case 'y': case 'z':
3388  case '_':
3389  // Notify MIOpt that we read a non-whitespace/non-comment token.
3390  MIOpt.ReadToken();
3391  return LexIdentifier(Result, CurPtr);
3392 
3393  case '$': // $ in identifiers.
3394  if (LangOpts.DollarIdents) {
3395  if (!isLexingRawMode())
3396  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3397  // Notify MIOpt that we read a non-whitespace/non-comment token.
3398  MIOpt.ReadToken();
3399  return LexIdentifier(Result, CurPtr);
3400  }
3401 
3402  Kind = tok::unknown;
3403  break;
3404 
3405  // C99 6.4.4: Character Constants.
3406  case '\'':
3407  // Notify MIOpt that we read a non-whitespace/non-comment token.
3408  MIOpt.ReadToken();
3409  return LexCharConstant(Result, CurPtr, tok::char_constant);
3410 
3411  // C99 6.4.5: String Literals.
3412  case '"':
3413  // Notify MIOpt that we read a non-whitespace/non-comment token.
3414  MIOpt.ReadToken();
3415  return LexStringLiteral(Result, CurPtr, tok::string_literal);
3416 
3417  // C99 6.4.6: Punctuators.
3418  case '?':
3419  Kind = tok::question;
3420  break;
3421  case '[':
3422  Kind = tok::l_square;
3423  break;
3424  case ']':
3425  Kind = tok::r_square;
3426  break;
3427  case '(':
3428  Kind = tok::l_paren;
3429  break;
3430  case ')':
3431  Kind = tok::r_paren;
3432  break;
3433  case '{':
3434  Kind = tok::l_brace;
3435  break;
3436  case '}':
3437  Kind = tok::r_brace;
3438  break;
3439  case '.':
3440  Char = getCharAndSize(CurPtr, SizeTmp);
3441  if (Char >= '0' && Char <= '9') {
3442  // Notify MIOpt that we read a non-whitespace/non-comment token.
3443  MIOpt.ReadToken();
3444 
3445  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3446  } else if (LangOpts.CPlusPlus && Char == '*') {
3447  Kind = tok::periodstar;
3448  CurPtr += SizeTmp;
3449  } else if (Char == '.' &&
3450  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3451  Kind = tok::ellipsis;
3452  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3453  SizeTmp2, Result);
3454  } else {
3455  Kind = tok::period;
3456  }
3457  break;
3458  case '&':
3459  Char = getCharAndSize(CurPtr, SizeTmp);
3460  if (Char == '&') {
3461  Kind = tok::ampamp;
3462  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3463  } else if (Char == '=') {
3464  Kind = tok::ampequal;
3465  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3466  } else {
3467  Kind = tok::amp;
3468  }
3469  break;
3470  case '*':
3471  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3472  Kind = tok::starequal;
3473  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3474  } else {
3475  Kind = tok::star;
3476  }
3477  break;
3478  case '+':
3479  Char = getCharAndSize(CurPtr, SizeTmp);
3480  if (Char == '+') {
3481  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3482  Kind = tok::plusplus;
3483  } else if (Char == '=') {
3484  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3485  Kind = tok::plusequal;
3486  } else {
3487  Kind = tok::plus;
3488  }
3489  break;
3490  case '-':
3491  Char = getCharAndSize(CurPtr, SizeTmp);
3492  if (Char == '-') { // --
3493  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3494  Kind = tok::minusminus;
3495  } else if (Char == '>' && LangOpts.CPlusPlus &&
3496  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3497  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3498  SizeTmp2, Result);
3499  Kind = tok::arrowstar;
3500  } else if (Char == '>') { // ->
3501  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3502  Kind = tok::arrow;
3503  } else if (Char == '=') { // -=
3504  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3505  Kind = tok::minusequal;
3506  } else {
3507  Kind = tok::minus;
3508  }
3509  break;
3510  case '~':
3511  Kind = tok::tilde;
3512  break;
3513  case '!':
3514  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3515  Kind = tok::exclaimequal;
3516  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3517  } else {
3518  Kind = tok::exclaim;
3519  }
3520  break;
3521  case '/':
3522  // 6.4.9: Comments
3523  Char = getCharAndSize(CurPtr, SizeTmp);
3524  if (Char == '/') { // Line comment.
3525  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3526  // want to lex this as a comment. There is one problem with this though,
3527  // that in one particular corner case, this can change the behavior of the
3528  // resultant program. For example, In "foo //**/ bar", C89 would lex
3529  // this as "foo / bar" and languages with Line comments would lex it as
3530  // "foo". Check to see if the character after the second slash is a '*'.
3531  // If so, we will lex that as a "/" instead of the start of a comment.
3532  // However, we never do this if we are just preprocessing.
3533  bool TreatAsComment = LangOpts.LineComment &&
3534  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3535  if (!TreatAsComment)
3536  if (!(PP && PP->isPreprocessedOutput()))
3537  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3538 
3539  if (TreatAsComment) {
3540  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3541  TokAtPhysicalStartOfLine))
3542  return true; // There is a token to return.
3543 
3544  // It is common for the tokens immediately after a // comment to be
3545  // whitespace (indentation for the next line). Instead of going through
3546  // the big switch, handle it efficiently now.
3547  goto SkipIgnoredUnits;
3548  }
3549  }
3550 
3551  if (Char == '*') { // /**/ comment.
3552  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3553  TokAtPhysicalStartOfLine))
3554  return true; // There is a token to return.
3555 
3556  // We only saw whitespace, so just try again with this lexer.
3557  // (We manually eliminate the tail call to avoid recursion.)
3558  goto LexNextToken;
3559  }
3560 
3561  if (Char == '=') {
3562  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3563  Kind = tok::slashequal;
3564  } else {
3565  Kind = tok::slash;
3566  }
3567  break;
3568  case '%':
3569  Char = getCharAndSize(CurPtr, SizeTmp);
3570  if (Char == '=') {
3571  Kind = tok::percentequal;
3572  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3573  } else if (LangOpts.Digraphs && Char == '>') {
3574  Kind = tok::r_brace; // '%>' -> '}'
3575  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3576  } else if (LangOpts.Digraphs && Char == ':') {
3577  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3578  Char = getCharAndSize(CurPtr, SizeTmp);
3579  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3580  Kind = tok::hashhash; // '%:%:' -> '##'
3581  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3582  SizeTmp2, Result);
3583  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3584  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3585  if (!isLexingRawMode())
3586  Diag(BufferPtr, diag::ext_charize_microsoft);
3587  Kind = tok::hashat;
3588  } else { // '%:' -> '#'
3589  // We parsed a # character. If this occurs at the start of the line,
3590  // it's actually the start of a preprocessing directive. Callback to
3591  // the preprocessor to handle it.
3592  // TODO: -fpreprocessed mode??
3593  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3594  goto HandleDirective;
3595 
3596  Kind = tok::hash;
3597  }
3598  } else {
3599  Kind = tok::percent;
3600  }
3601  break;
3602  case '<':
3603  Char = getCharAndSize(CurPtr, SizeTmp);
3604  if (ParsingFilename) {
3605  return LexAngledStringLiteral(Result, CurPtr);
3606  } else if (Char == '<') {
3607  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3608  if (After == '=') {
3609  Kind = tok::lesslessequal;
3610  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3611  SizeTmp2, Result);
3612  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3613  // If this is actually a '<<<<<<<' version control conflict marker,
3614  // recognize it as such and recover nicely.
3615  goto LexNextToken;
3616  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3617  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3618  // ignore it.
3619  goto LexNextToken;
3620  } else if (LangOpts.CUDA && After == '<') {
3621  Kind = tok::lesslessless;
3622  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3623  SizeTmp2, Result);
3624  } else {
3625  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3626  Kind = tok::lessless;
3627  }
3628  } else if (Char == '=') {
3629  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3630  if (After == '>') {
3631  if (getLangOpts().CPlusPlus2a) {
3632  if (!isLexingRawMode())
3633  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3634  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3635  SizeTmp2, Result);
3636  Kind = tok::spaceship;
3637  break;
3638  }
3639  // Suggest adding a space between the '<=' and the '>' to avoid a
3640  // change in semantics if this turns up in C++ <=17 mode.
3641  if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3642  Diag(BufferPtr, diag::warn_cxx2a_compat_spaceship)
3644  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3645  }
3646  }
3647  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3648  Kind = tok::lessequal;
3649  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3650  if (LangOpts.CPlusPlus11 &&
3651  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3652  // C++0x [lex.pptoken]p3:
3653  // Otherwise, if the next three characters are <:: and the subsequent
3654  // character is neither : nor >, the < is treated as a preprocessor
3655  // token by itself and not as the first character of the alternative
3656  // token <:.
3657  unsigned SizeTmp3;
3658  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3659  if (After != ':' && After != '>') {
3660  Kind = tok::less;
3661  if (!isLexingRawMode())
3662  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3663  break;
3664  }
3665  }
3666 
3667  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3668  Kind = tok::l_square;
3669  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3670  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3671  Kind = tok::l_brace;
3672  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3673  lexEditorPlaceholder(Result, CurPtr)) {
3674  return true;
3675  } else {
3676  Kind = tok::less;
3677  }
3678  break;
3679  case '>':
3680  Char = getCharAndSize(CurPtr, SizeTmp);
3681  if (Char == '=') {
3682  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3683  Kind = tok::greaterequal;
3684  } else if (Char == '>') {
3685  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3686  if (After == '=') {
3687  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3688  SizeTmp2, Result);
3689  Kind = tok::greatergreaterequal;
3690  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3691  // If this is actually a '>>>>' conflict marker, recognize it as such
3692  // and recover nicely.
3693  goto LexNextToken;
3694  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3695  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3696  goto LexNextToken;
3697  } else if (LangOpts.CUDA && After == '>') {
3698  Kind = tok::greatergreatergreater;
3699  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3700  SizeTmp2, Result);
3701  } else {
3702  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3703  Kind = tok::greatergreater;
3704  }
3705  } else {
3706  Kind = tok::greater;
3707  }
3708  break;
3709  case '^':
3710  Char = getCharAndSize(CurPtr, SizeTmp);
3711  if (Char == '=') {
3712  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3713  Kind = tok::caretequal;
3714  } else if (LangOpts.OpenCL && Char == '^') {
3715  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3716  Kind = tok::caretcaret;
3717  } else {
3718  Kind = tok::caret;
3719  }
3720  break;
3721  case '|':
3722  Char = getCharAndSize(CurPtr, SizeTmp);
3723  if (Char == '=') {
3724  Kind = tok::pipeequal;
3725  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3726  } else if (Char == '|') {
3727  // If this is '|||||||' and we're in a conflict marker, ignore it.
3728  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3729  goto LexNextToken;
3730  Kind = tok::pipepipe;
3731  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3732  } else {
3733  Kind = tok::pipe;
3734  }
3735  break;
3736  case ':':
3737  Char = getCharAndSize(CurPtr, SizeTmp);
3738  if (LangOpts.Digraphs && Char == '>') {
3739  Kind = tok::r_square; // ':>' -> ']'
3740  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3741  } else if ((LangOpts.CPlusPlus ||
3742  LangOpts.DoubleSquareBracketAttributes) &&
3743  Char == ':') {
3744  Kind = tok::coloncolon;
3745  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3746  } else {
3747  Kind = tok::colon;
3748  }
3749  break;
3750  case ';':
3751  Kind = tok::semi;
3752  break;
3753  case '=':
3754  Char = getCharAndSize(CurPtr, SizeTmp);
3755  if (Char == '=') {
3756  // If this is '====' and we're in a conflict marker, ignore it.
3757  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3758  goto LexNextToken;
3759 
3760  Kind = tok::equalequal;
3761  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3762  } else {
3763  Kind = tok::equal;
3764  }
3765  break;
3766  case ',':
3767  Kind = tok::comma;
3768  break;
3769  case '#':
3770  Char = getCharAndSize(CurPtr, SizeTmp);
3771  if (Char == '#') {
3772  Kind = tok::hashhash;
3773  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3774  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3775  Kind = tok::hashat;
3776  if (!isLexingRawMode())
3777  Diag(BufferPtr, diag::ext_charize_microsoft);
3778  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3779  } else {
3780  // We parsed a # character. If this occurs at the start of the line,
3781  // it's actually the start of a preprocessing directive. Callback to
3782  // the preprocessor to handle it.
3783  // TODO: -fpreprocessed mode??
3784  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3785  goto HandleDirective;
3786 
3787  Kind = tok::hash;
3788  }
3789  break;
3790 
3791  case '@':
3792  // Objective C support.
3793  if (CurPtr[-1] == '@' && LangOpts.ObjC1)
3794  Kind = tok::at;
3795  else
3796  Kind = tok::unknown;
3797  break;
3798 
3799  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3800  case '\\':
3801  if (!LangOpts.AsmPreprocessor) {
3802  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3803  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3804  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3805  return true; // KeepWhitespaceMode
3806 
3807  // We only saw whitespace, so just try again with this lexer.
3808  // (We manually eliminate the tail call to avoid recursion.)
3809  goto LexNextToken;
3810  }
3811 
3812  return LexUnicode(Result, CodePoint, CurPtr);
3813  }
3814  }
3815 
3816  Kind = tok::unknown;
3817  break;
3818 
3819  default: {
3820  if (isASCII(Char)) {
3821  Kind = tok::unknown;
3822  break;
3823  }
3824 
3825  llvm::UTF32 CodePoint;
3826 
3827  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3828  // an escaped newline.
3829  --CurPtr;
3830  const char *UTF8StartPtr = CurPtr;
3831  llvm::ConversionResult Status =
3832  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3833  (const llvm::UTF8 *)BufferEnd,
3834  &CodePoint,
3835  llvm::strictConversion);
3836  if (Status == llvm::conversionOK) {
3837  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3838  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3839  return true; // KeepWhitespaceMode
3840 
3841  // We only saw whitespace, so just try again with this lexer.
3842  // (We manually eliminate the tail call to avoid recursion.)
3843  goto LexNextToken;
3844  }
3845  if (!isLexingRawMode())
3846  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
3847  makeCharRange(*this, UTF8StartPtr, CurPtr));
3848  return LexUnicode(Result, CodePoint, CurPtr);
3849  }
3850 
3852  PP->isPreprocessedOutput()) {
3853  ++CurPtr;
3854  Kind = tok::unknown;
3855  break;
3856  }
3857 
3858  // Non-ASCII characters tend to creep into source code unintentionally.
3859  // Instead of letting the parser complain about the unknown token,
3860  // just diagnose the invalid UTF-8, then drop the character.
3861  Diag(CurPtr, diag::err_invalid_utf8);
3862 
3863  BufferPtr = CurPtr+1;
3864  // We're pretending the character didn't exist, so just try again with
3865  // this lexer.
3866  // (We manually eliminate the tail call to avoid recursion.)
3867  goto LexNextToken;
3868  }
3869  }
3870 
3871  // Notify MIOpt that we read a non-whitespace/non-comment token.
3872  MIOpt.ReadToken();
3873 
3874  // Update the location of token as well as BufferPtr.
3875  FormTokenWithChars(Result, CurPtr, Kind);
3876  return true;
3877 
3878 HandleDirective:
3879  // We parsed a # character and it's the start of a preprocessing directive.
3880 
3881  FormTokenWithChars(Result, CurPtr, tok::hash);
3882  PP->HandleDirective(Result);
3883 
3884  if (PP->hadModuleLoaderFatalFailure()) {
3885  // With a fatal failure in the module loader, we abort parsing.
3886  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3887  return true;
3888  }
3889 
3890  // We parsed the directive; lex a token with the new state.
3891  return false;
3892 }
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:1424
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:529
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:850
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:793
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:1095
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:1064
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:3819
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:1044
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 ...
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.
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:2354
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:1459
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:932
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:815
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:771
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
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:1263
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:1119
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:865
Defines the clang::IdentifierInfo, clang::IdentifierTable, and clang::Selector interfaces.
CharSourceRange getImmediateExpansionRange(SourceLocation Loc) const
Return the start/end of the expansion information for an expansion location.
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.
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:1474
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:825
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:1504
SourceLocation createExpansionLoc(SourceLocation Loc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLength, bool ExpansionIsTokenRange=true, int LoadedID=0, unsigned LoadedOffset=0)
Return a new SourceLocation that encodes the fact that a token from SpellingLoc should actually be re...
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:1230
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1465
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:841
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.
static unsigned getTokenPrefixLength(SourceLocation TokStart, unsigned Characters, const SourceManager &SM, const LangOptions &LangOpts)
Get the physical length (including trigraphs and escaped newlines) of the first Characters characters...
Definition: Lexer.cpp:711
SourceLocation getExpansionLocStart() const
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1138
__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:968
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1015
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:868
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
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:2863
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:1442
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:816
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:1148
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1040
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:2596
Defines the clang::SourceLocation class and associated facilities.
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:818
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:1167
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2759
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:4313
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:886
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:3148
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.