clang  15.0.0git
Lexer.cpp
Go to the documentation of this file.
1 //===- Lexer.cpp - C Language Family Lexer --------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the Lexer and Token interfaces.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "clang/Lex/Lexer.h"
14 #include "UnicodeCharSets.h"
15 #include "clang/Basic/CharInfo.h"
16 #include "clang/Basic/Diagnostic.h"
18 #include "clang/Basic/LLVM.h"
22 #include "clang/Basic/TokenKinds.h"
26 #include "clang/Lex/Preprocessor.h"
28 #include "clang/Lex/Token.h"
29 #include "llvm/ADT/None.h"
30 #include "llvm/ADT/Optional.h"
31 #include "llvm/ADT/STLExtras.h"
32 #include "llvm/ADT/StringExtras.h"
33 #include "llvm/ADT/StringRef.h"
34 #include "llvm/ADT/StringSwitch.h"
35 #include "llvm/Support/Compiler.h"
36 #include "llvm/Support/ConvertUTF.h"
37 #include "llvm/Support/MathExtras.h"
38 #include "llvm/Support/MemoryBufferRef.h"
39 #include "llvm/Support/NativeFormatting.h"
40 #include "llvm/Support/UnicodeCharRanges.h"
41 #include <algorithm>
42 #include <cassert>
43 #include <cstddef>
44 #include <cstdint>
45 #include <cstring>
46 #include <string>
47 #include <tuple>
48 #include <utility>
49 
50 using namespace clang;
51 
52 //===----------------------------------------------------------------------===//
53 // Token Class Implementation
54 //===----------------------------------------------------------------------===//
55 
56 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
58  if (isAnnotation())
59  return false;
61  return II->getObjCKeywordID() == objcKey;
62  return false;
63 }
64 
65 /// getObjCKeywordID - Return the ObjC keyword kind.
67  if (isAnnotation())
68  return tok::objc_not_keyword;
70  return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
71 }
72 
73 //===----------------------------------------------------------------------===//
74 // Lexer Class Implementation
75 //===----------------------------------------------------------------------===//
76 
77 void Lexer::anchor() {}
78 
79 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
80  const char *BufEnd) {
81  BufferStart = BufStart;
82  BufferPtr = BufPtr;
83  BufferEnd = BufEnd;
84 
85  assert(BufEnd[0] == 0 &&
86  "We assume that the input buffer has a null character at the end"
87  " to simplify lexing!");
88 
89  // Check whether we have a BOM in the beginning of the buffer. If yes - act
90  // accordingly. Right now we support only UTF-8 with and without BOM, so, just
91  // skip the UTF-8 BOM if it's present.
92  if (BufferStart == BufferPtr) {
93  // Determine the size of the BOM.
94  StringRef Buf(BufferStart, BufferEnd - BufferStart);
95  size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
96  .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
97  .Default(0);
98 
99  // Skip the BOM.
100  BufferPtr += BOMLength;
101  }
102 
103  Is_PragmaLexer = false;
104  CurrentConflictMarkerState = CMK_None;
105 
106  // Start of the file is a start of line.
107  IsAtStartOfLine = true;
108  IsAtPhysicalStartOfLine = true;
109 
110  HasLeadingSpace = false;
111  HasLeadingEmptyMacro = false;
112 
113  // We are not after parsing a #.
115 
116  // We are not after parsing #include.
117  ParsingFilename = false;
118 
119  // We are not in raw mode. Raw mode disables diagnostics and interpretation
120  // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
121  // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
122  // or otherwise skipping over tokens.
123  LexingRawMode = false;
124 
125  // Default to not keeping comments.
126  ExtendedTokenMode = 0;
127 
128  NewLinePtr = nullptr;
129 }
130 
131 /// Lexer constructor - Create a new lexer object for the specified buffer
132 /// with the specified preprocessor managing the lexing process. This lexer
133 /// assumes that the associated file buffer and Preprocessor objects will
134 /// outlive it, so it doesn't take ownership of either of them.
135 Lexer::Lexer(FileID FID, const llvm::MemoryBufferRef &InputFile,
136  Preprocessor &PP, bool IsFirstIncludeOfFile)
137  : PreprocessorLexer(&PP, FID),
138  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
139  LangOpts(PP.getLangOpts()), LineComment(LangOpts.LineComment),
140  IsFirstTimeLexingFile(IsFirstIncludeOfFile) {
141  InitLexer(InputFile.getBufferStart(), InputFile.getBufferStart(),
142  InputFile.getBufferEnd());
143 
145 }
146 
147 /// Lexer constructor - Create a new raw lexer object. This object is only
148 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
149 /// range will outlive it, so it doesn't take ownership of it.
150 Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
151  const char *BufStart, const char *BufPtr, const char *BufEnd,
152  bool IsFirstIncludeOfFile)
153  : FileLoc(fileloc), LangOpts(langOpts), LineComment(LangOpts.LineComment),
154  IsFirstTimeLexingFile(IsFirstIncludeOfFile) {
155  InitLexer(BufStart, BufPtr, BufEnd);
156 
157  // We *are* in raw mode.
158  LexingRawMode = true;
159 }
160 
161 /// Lexer constructor - Create a new raw lexer object. This object is only
162 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
163 /// range will outlive it, so it doesn't take ownership of it.
164 Lexer::Lexer(FileID FID, const llvm::MemoryBufferRef &FromFile,
165  const SourceManager &SM, const LangOptions &langOpts,
166  bool IsFirstIncludeOfFile)
167  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile.getBufferStart(),
168  FromFile.getBufferStart(), FromFile.getBufferEnd(),
169  IsFirstIncludeOfFile) {}
170 
172  assert(PP && "Cannot reset token mode without a preprocessor");
173  if (LangOpts.TraditionalCPP)
174  SetKeepWhitespaceMode(true);
175  else
177 }
178 
179 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
180 /// _Pragma expansion. This has a variety of magic semantics that this method
181 /// sets up. It returns a new'd Lexer that must be delete'd when done.
182 ///
183 /// On entrance to this routine, TokStartLoc is a macro location which has a
184 /// spelling loc that indicates the bytes to be lexed for the token and an
185 /// expansion location that indicates where all lexed tokens should be
186 /// "expanded from".
187 ///
188 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
189 /// normal lexer that remaps tokens as they fly by. This would require making
190 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
191 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
192 /// out of the critical path of the lexer!
193 ///
195  SourceLocation ExpansionLocStart,
196  SourceLocation ExpansionLocEnd,
197  unsigned TokLen, Preprocessor &PP) {
199 
200  // Create the lexer as if we were going to lex the file normally.
201  FileID SpellingFID = SM.getFileID(SpellingLoc);
202  llvm::MemoryBufferRef InputFile = SM.getBufferOrFake(SpellingFID);
203  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
204 
205  // Now that the lexer is created, change the start/end locations so that we
206  // just lex the subsection of the file that we want. This is lexing from a
207  // scratch buffer.
208  const char *StrData = SM.getCharacterData(SpellingLoc);
209 
210  L->BufferPtr = StrData;
211  L->BufferEnd = StrData+TokLen;
212  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
213 
214  // Set the SourceLocation with the remapping information. This ensures that
215  // GetMappedTokenLoc will remap the tokens as they are lexed.
216  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
217  ExpansionLocStart,
218  ExpansionLocEnd, TokLen);
219 
220  // Ensure that the lexer thinks it is inside a directive, so that end \n will
221  // return an EOD token.
223 
224  // This lexer really is for _Pragma.
225  L->Is_PragmaLexer = true;
226  return L;
227 }
228 
229 bool Lexer::skipOver(unsigned NumBytes) {
230  IsAtPhysicalStartOfLine = true;
231  IsAtStartOfLine = true;
232  if ((BufferPtr + NumBytes) > BufferEnd)
233  return true;
234  BufferPtr += NumBytes;
235  return false;
236 }
237 
238 template <typename T> static void StringifyImpl(T &Str, char Quote) {
239  typename T::size_type i = 0, e = Str.size();
240  while (i < e) {
241  if (Str[i] == '\\' || Str[i] == Quote) {
242  Str.insert(Str.begin() + i, '\\');
243  i += 2;
244  ++e;
245  } else if (Str[i] == '\n' || Str[i] == '\r') {
246  // Replace '\r\n' and '\n\r' to '\\' followed by 'n'.
247  if ((i < e - 1) && (Str[i + 1] == '\n' || Str[i + 1] == '\r') &&
248  Str[i] != Str[i + 1]) {
249  Str[i] = '\\';
250  Str[i + 1] = 'n';
251  } else {
252  // Replace '\n' and '\r' to '\\' followed by 'n'.
253  Str[i] = '\\';
254  Str.insert(Str.begin() + i + 1, 'n');
255  ++e;
256  }
257  i += 2;
258  } else
259  ++i;
260  }
261 }
262 
263 std::string Lexer::Stringify(StringRef Str, bool Charify) {
264  std::string Result = std::string(Str);
265  char Quote = Charify ? '\'' : '"';
266  StringifyImpl(Result, Quote);
267  return Result;
268 }
269 
271 
272 //===----------------------------------------------------------------------===//
273 // Token Spelling
274 //===----------------------------------------------------------------------===//
275 
276 /// Slow case of getSpelling. Extract the characters comprising the
277 /// spelling of this token from the provided input buffer.
278 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
279  const LangOptions &LangOpts, char *Spelling) {
280  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
281 
282  size_t Length = 0;
283  const char *BufEnd = BufPtr + Tok.getLength();
284 
285  if (tok::isStringLiteral(Tok.getKind())) {
286  // Munch the encoding-prefix and opening double-quote.
287  while (BufPtr < BufEnd) {
288  unsigned Size;
289  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
290  BufPtr += Size;
291 
292  if (Spelling[Length - 1] == '"')
293  break;
294  }
295 
296  // Raw string literals need special handling; trigraph expansion and line
297  // splicing do not occur within their d-char-sequence nor within their
298  // r-char-sequence.
299  if (Length >= 2 &&
300  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
301  // Search backwards from the end of the token to find the matching closing
302  // quote.
303  const char *RawEnd = BufEnd;
304  do --RawEnd; while (*RawEnd != '"');
305  size_t RawLength = RawEnd - BufPtr + 1;
306 
307  // Everything between the quotes is included verbatim in the spelling.
308  memcpy(Spelling + Length, BufPtr, RawLength);
309  Length += RawLength;
310  BufPtr += RawLength;
311 
312  // The rest of the token is lexed normally.
313  }
314  }
315 
316  while (BufPtr < BufEnd) {
317  unsigned Size;
318  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
319  BufPtr += Size;
320  }
321 
322  assert(Length < Tok.getLength() &&
323  "NeedsCleaning flag set on token that didn't need cleaning!");
324  return Length;
325 }
326 
327 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
328 /// token are the characters used to represent the token in the source file
329 /// after trigraph expansion and escaped-newline folding. In particular, this
330 /// wants to get the true, uncanonicalized, spelling of things like digraphs
331 /// UCNs, etc.
333  SmallVectorImpl<char> &buffer,
334  const SourceManager &SM,
335  const LangOptions &options,
336  bool *invalid) {
337  // Break down the source location.
338  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
339 
340  // Try to the load the file buffer.
341  bool invalidTemp = false;
342  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
343  if (invalidTemp) {
344  if (invalid) *invalid = true;
345  return {};
346  }
347 
348  const char *tokenBegin = file.data() + locInfo.second;
349 
350  // Lex from the start of the given location.
351  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
352  file.begin(), tokenBegin, file.end());
353  Token token;
354  lexer.LexFromRawLexer(token);
355 
356  unsigned length = token.getLength();
357 
358  // Common case: no need for cleaning.
359  if (!token.needsCleaning())
360  return StringRef(tokenBegin, length);
361 
362  // Hard case, we need to relex the characters into the string.
363  buffer.resize(length);
364  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
365  return StringRef(buffer.data(), buffer.size());
366 }
367 
368 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
369 /// token are the characters used to represent the token in the source file
370 /// after trigraph expansion and escaped-newline folding. In particular, this
371 /// wants to get the true, uncanonicalized, spelling of things like digraphs
372 /// UCNs, etc.
373 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
374  const LangOptions &LangOpts, bool *Invalid) {
375  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
376 
377  bool CharDataInvalid = false;
378  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
379  &CharDataInvalid);
380  if (Invalid)
381  *Invalid = CharDataInvalid;
382  if (CharDataInvalid)
383  return {};
384 
385  // If this token contains nothing interesting, return it directly.
386  if (!Tok.needsCleaning())
387  return std::string(TokStart, TokStart + Tok.getLength());
388 
389  std::string Result;
390  Result.resize(Tok.getLength());
391  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
392  return Result;
393 }
394 
395 /// getSpelling - This method is used to get the spelling of a token into a
396 /// preallocated buffer, instead of as an std::string. The caller is required
397 /// to allocate enough space for the token, which is guaranteed to be at least
398 /// Tok.getLength() bytes long. The actual length of the token is returned.
399 ///
400 /// Note that this method may do two possible things: it may either fill in
401 /// the buffer specified with characters, or it may *change the input pointer*
402 /// to point to a constant buffer with the data already in it (avoiding a
403 /// copy). The caller is not allowed to modify the returned buffer pointer
404 /// if an internal buffer is returned.
405 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
406  const SourceManager &SourceMgr,
407  const LangOptions &LangOpts, bool *Invalid) {
408  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
409 
410  const char *TokStart = nullptr;
411  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
412  if (Tok.is(tok::raw_identifier))
413  TokStart = Tok.getRawIdentifier().data();
414  else if (!Tok.hasUCN()) {
415  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
416  // Just return the string from the identifier table, which is very quick.
417  Buffer = II->getNameStart();
418  return II->getLength();
419  }
420  }
421 
422  // NOTE: this can be checked even after testing for an IdentifierInfo.
423  if (Tok.isLiteral())
424  TokStart = Tok.getLiteralData();
425 
426  if (!TokStart) {
427  // Compute the start of the token in the input lexer buffer.
428  bool CharDataInvalid = false;
429  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
430  if (Invalid)
431  *Invalid = CharDataInvalid;
432  if (CharDataInvalid) {
433  Buffer = "";
434  return 0;
435  }
436  }
437 
438  // If this token contains nothing interesting, return it directly.
439  if (!Tok.needsCleaning()) {
440  Buffer = TokStart;
441  return Tok.getLength();
442  }
443 
444  // Otherwise, hard case, relex the characters into the string.
445  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
446 }
447 
448 /// MeasureTokenLength - Relex the token at the specified location and return
449 /// its length in bytes in the input file. If the token needs cleaning (e.g.
450 /// includes a trigraph or an escaped newline) then this count includes bytes
451 /// that are part of that.
453  const SourceManager &SM,
454  const LangOptions &LangOpts) {
455  Token TheTok;
456  if (getRawToken(Loc, TheTok, SM, LangOpts))
457  return 0;
458  return TheTok.getLength();
459 }
460 
461 /// Relex the token at the specified location.
462 /// \returns true if there was a failure, false on success.
464  const SourceManager &SM,
465  const LangOptions &LangOpts,
466  bool IgnoreWhiteSpace) {
467  // TODO: this could be special cased for common tokens like identifiers, ')',
468  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
469  // all obviously single-char tokens. This could use
470  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
471  // something.
472 
473  // If this comes from a macro expansion, we really do want the macro name, not
474  // the token this macro expanded to.
475  Loc = SM.getExpansionLoc(Loc);
476  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
477  bool Invalid = false;
478  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
479  if (Invalid)
480  return true;
481 
482  const char *StrData = Buffer.data()+LocInfo.second;
483 
484  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
485  return true;
486 
487  // Create a lexer starting at the beginning of this token.
488  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
489  Buffer.begin(), StrData, Buffer.end());
490  TheLexer.SetCommentRetentionState(true);
491  TheLexer.LexFromRawLexer(Result);
492  return false;
493 }
494 
495 /// Returns the pointer that points to the beginning of line that contains
496 /// the given offset, or null if the offset if invalid.
497 static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
498  const char *BufStart = Buffer.data();
499  if (Offset >= Buffer.size())
500  return nullptr;
501 
502  const char *LexStart = BufStart + Offset;
503  for (; LexStart != BufStart; --LexStart) {
504  if (isVerticalWhitespace(LexStart[0]) &&
505  !Lexer::isNewLineEscaped(BufStart, LexStart)) {
506  // LexStart should point at first character of logical line.
507  ++LexStart;
508  break;
509  }
510  }
511  return LexStart;
512 }
513 
515  const SourceManager &SM,
516  const LangOptions &LangOpts) {
517  assert(Loc.isFileID());
518  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
519  if (LocInfo.first.isInvalid())
520  return Loc;
521 
522  bool Invalid = false;
523  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
524  if (Invalid)
525  return Loc;
526 
527  // Back up from the current location until we hit the beginning of a line
528  // (or the buffer). We'll relex from that point.
529  const char *StrData = Buffer.data() + LocInfo.second;
530  const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
531  if (!LexStart || LexStart == StrData)
532  return Loc;
533 
534  // Create a lexer starting at the beginning of this token.
535  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
536  Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
537  Buffer.end());
538  TheLexer.SetCommentRetentionState(true);
539 
540  // Lex tokens until we find the token that contains the source location.
541  Token TheTok;
542  do {
543  TheLexer.LexFromRawLexer(TheTok);
544 
545  if (TheLexer.getBufferLocation() > StrData) {
546  // Lexing this token has taken the lexer past the source location we're
547  // looking for. If the current token encompasses our source location,
548  // return the beginning of that token.
549  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
550  return TheTok.getLocation();
551 
552  // We ended up skipping over the source location entirely, which means
553  // that it points into whitespace. We're done here.
554  break;
555  }
556  } while (TheTok.getKind() != tok::eof);
557 
558  // We've passed our source location; just return the original source location.
559  return Loc;
560 }
561 
563  const SourceManager &SM,
564  const LangOptions &LangOpts) {
565  if (Loc.isFileID())
566  return getBeginningOfFileToken(Loc, SM, LangOpts);
567 
568  if (!SM.isMacroArgExpansion(Loc))
569  return Loc;
570 
571  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
572  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
573  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
574  std::pair<FileID, unsigned> BeginFileLocInfo =
575  SM.getDecomposedLoc(BeginFileLoc);
576  assert(FileLocInfo.first == BeginFileLocInfo.first &&
577  FileLocInfo.second >= BeginFileLocInfo.second);
578  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
579 }
580 
581 namespace {
582 
583 enum PreambleDirectiveKind {
584  PDK_Skipped,
585  PDK_Unknown
586 };
587 
588 } // namespace
589 
591  const LangOptions &LangOpts,
592  unsigned MaxLines) {
593  // Create a lexer starting at the beginning of the file. Note that we use a
594  // "fake" file source location at offset 1 so that the lexer will track our
595  // position within the file.
596  const SourceLocation::UIntTy StartOffset = 1;
597  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
598  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
599  Buffer.end());
600  TheLexer.SetCommentRetentionState(true);
601 
602  bool InPreprocessorDirective = false;
603  Token TheTok;
604  SourceLocation ActiveCommentLoc;
605 
606  unsigned MaxLineOffset = 0;
607  if (MaxLines) {
608  const char *CurPtr = Buffer.begin();
609  unsigned CurLine = 0;
610  while (CurPtr != Buffer.end()) {
611  char ch = *CurPtr++;
612  if (ch == '\n') {
613  ++CurLine;
614  if (CurLine == MaxLines)
615  break;
616  }
617  }
618  if (CurPtr != Buffer.end())
619  MaxLineOffset = CurPtr - Buffer.begin();
620  }
621 
622  do {
623  TheLexer.LexFromRawLexer(TheTok);
624 
625  if (InPreprocessorDirective) {
626  // If we've hit the end of the file, we're done.
627  if (TheTok.getKind() == tok::eof) {
628  break;
629  }
630 
631  // If we haven't hit the end of the preprocessor directive, skip this
632  // token.
633  if (!TheTok.isAtStartOfLine())
634  continue;
635 
636  // We've passed the end of the preprocessor directive, and will look
637  // at this token again below.
638  InPreprocessorDirective = false;
639  }
640 
641  // Keep track of the # of lines in the preamble.
642  if (TheTok.isAtStartOfLine()) {
643  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
644 
645  // If we were asked to limit the number of lines in the preamble,
646  // and we're about to exceed that limit, we're done.
647  if (MaxLineOffset && TokOffset >= MaxLineOffset)
648  break;
649  }
650 
651  // Comments are okay; skip over them.
652  if (TheTok.getKind() == tok::comment) {
653  if (ActiveCommentLoc.isInvalid())
654  ActiveCommentLoc = TheTok.getLocation();
655  continue;
656  }
657 
658  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
659  // This is the start of a preprocessor directive.
660  Token HashTok = TheTok;
661  InPreprocessorDirective = true;
662  ActiveCommentLoc = SourceLocation();
663 
664  // Figure out which directive this is. Since we're lexing raw tokens,
665  // we don't have an identifier table available. Instead, just look at
666  // the raw identifier to recognize and categorize preprocessor directives.
667  TheLexer.LexFromRawLexer(TheTok);
668  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
669  StringRef Keyword = TheTok.getRawIdentifier();
670  PreambleDirectiveKind PDK
671  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
672  .Case("include", PDK_Skipped)
673  .Case("__include_macros", PDK_Skipped)
674  .Case("define", PDK_Skipped)
675  .Case("undef", PDK_Skipped)
676  .Case("line", PDK_Skipped)
677  .Case("error", PDK_Skipped)
678  .Case("pragma", PDK_Skipped)
679  .Case("import", PDK_Skipped)
680  .Case("include_next", PDK_Skipped)
681  .Case("warning", PDK_Skipped)
682  .Case("ident", PDK_Skipped)
683  .Case("sccs", PDK_Skipped)
684  .Case("assert", PDK_Skipped)
685  .Case("unassert", PDK_Skipped)
686  .Case("if", PDK_Skipped)
687  .Case("ifdef", PDK_Skipped)
688  .Case("ifndef", PDK_Skipped)
689  .Case("elif", PDK_Skipped)
690  .Case("elifdef", PDK_Skipped)
691  .Case("elifndef", PDK_Skipped)
692  .Case("else", PDK_Skipped)
693  .Case("endif", PDK_Skipped)
694  .Default(PDK_Unknown);
695 
696  switch (PDK) {
697  case PDK_Skipped:
698  continue;
699 
700  case PDK_Unknown:
701  // We don't know what this directive is; stop at the '#'.
702  break;
703  }
704  }
705 
706  // We only end up here if we didn't recognize the preprocessor
707  // directive or it was one that can't occur in the preamble at this
708  // point. Roll back the current token to the location of the '#'.
709  TheTok = HashTok;
710  }
711 
712  // We hit a token that we don't recognize as being in the
713  // "preprocessing only" part of the file, so we're no longer in
714  // the preamble.
715  break;
716  } while (true);
717 
719  if (ActiveCommentLoc.isValid())
720  End = ActiveCommentLoc; // don't truncate a decl comment.
721  else
722  End = TheTok.getLocation();
723 
724  return PreambleBounds(End.getRawEncoding() - FileLoc.getRawEncoding(),
725  TheTok.isAtStartOfLine());
726 }
727 
728 unsigned Lexer::getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo,
729  const SourceManager &SM,
730  const LangOptions &LangOpts) {
731  // Figure out how many physical characters away the specified expansion
732  // character is. This needs to take into consideration newlines and
733  // trigraphs.
734  bool Invalid = false;
735  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
736 
737  // If they request the first char of the token, we're trivially done.
738  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
739  return 0;
740 
741  unsigned PhysOffset = 0;
742 
743  // The usual case is that tokens don't contain anything interesting. Skip
744  // over the uninteresting characters. If a token only consists of simple
745  // chars, this method is extremely fast.
746  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
747  if (CharNo == 0)
748  return PhysOffset;
749  ++TokPtr;
750  --CharNo;
751  ++PhysOffset;
752  }
753 
754  // If we have a character that may be a trigraph or escaped newline, use a
755  // lexer to parse it correctly.
756  for (; CharNo; --CharNo) {
757  unsigned Size;
758  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
759  TokPtr += Size;
760  PhysOffset += Size;
761  }
762 
763  // Final detail: if we end up on an escaped newline, we want to return the
764  // location of the actual byte of the token. For example foo<newline>bar
765  // advanced by 3 should return the location of b, not of \\. One compounding
766  // detail of this is that the escape may be made by a trigraph.
767  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
768  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
769 
770  return PhysOffset;
771 }
772 
773 /// Computes the source location just past the end of the
774 /// token at this source location.
775 ///
776 /// This routine can be used to produce a source location that
777 /// points just past the end of the token referenced by \p Loc, and
778 /// is generally used when a diagnostic needs to point just after a
779 /// token where it expected something different that it received. If
780 /// the returned source location would not be meaningful (e.g., if
781 /// it points into a macro), this routine returns an invalid
782 /// source location.
783 ///
784 /// \param Offset an offset from the end of the token, where the source
785 /// location should refer to. The default offset (0) produces a source
786 /// location pointing just past the end of the token; an offset of 1 produces
787 /// a source location pointing to the last character in the token, etc.
789  const SourceManager &SM,
790  const LangOptions &LangOpts) {
791  if (Loc.isInvalid())
792  return {};
793 
794  if (Loc.isMacroID()) {
795  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
796  return {}; // Points inside the macro expansion.
797  }
798 
799  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
800  if (Len > Offset)
801  Len = Len - Offset;
802  else
803  return Loc;
804 
805  return Loc.getLocWithOffset(Len);
806 }
807 
808 /// Returns true if the given MacroID location points at the first
809 /// token of the macro expansion.
811  const SourceManager &SM,
812  const LangOptions &LangOpts,
813  SourceLocation *MacroBegin) {
814  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
815 
816  SourceLocation expansionLoc;
817  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
818  return false;
819 
820  if (expansionLoc.isFileID()) {
821  // No other macro expansions, this is the first.
822  if (MacroBegin)
823  *MacroBegin = expansionLoc;
824  return true;
825  }
826 
827  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
828 }
829 
830 /// Returns true if the given MacroID location points at the last
831 /// token of the macro expansion.
833  const SourceManager &SM,
834  const LangOptions &LangOpts,
835  SourceLocation *MacroEnd) {
836  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
837 
838  SourceLocation spellLoc = SM.getSpellingLoc(loc);
839  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
840  if (tokLen == 0)
841  return false;
842 
843  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
844  SourceLocation expansionLoc;
845  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
846  return false;
847 
848  if (expansionLoc.isFileID()) {
849  // No other macro expansions.
850  if (MacroEnd)
851  *MacroEnd = expansionLoc;
852  return true;
853  }
854 
855  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
856 }
857 
859  const SourceManager &SM,
860  const LangOptions &LangOpts) {
861  SourceLocation Begin = Range.getBegin();
862  SourceLocation End = Range.getEnd();
863  assert(Begin.isFileID() && End.isFileID());
864  if (Range.isTokenRange()) {
865  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
866  if (End.isInvalid())
867  return {};
868  }
869 
870  // Break down the source locations.
871  FileID FID;
872  unsigned BeginOffs;
873  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
874  if (FID.isInvalid())
875  return {};
876 
877  unsigned EndOffs;
878  if (!SM.isInFileID(End, FID, &EndOffs) ||
879  BeginOffs > EndOffs)
880  return {};
881 
883 }
884 
885 // Assumes that `Loc` is in an expansion.
887  const SourceManager &SM) {
888  return SM.getSLocEntry(SM.getFileID(Loc))
889  .getExpansion()
890  .isExpansionTokenRange();
891 }
892 
894  const SourceManager &SM,
895  const LangOptions &LangOpts) {
896  SourceLocation Begin = Range.getBegin();
897  SourceLocation End = Range.getEnd();
898  if (Begin.isInvalid() || End.isInvalid())
899  return {};
900 
901  if (Begin.isFileID() && End.isFileID())
902  return makeRangeFromFileLocs(Range, SM, LangOpts);
903 
904  if (Begin.isMacroID() && End.isFileID()) {
905  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
906  return {};
907  Range.setBegin(Begin);
908  return makeRangeFromFileLocs(Range, SM, LangOpts);
909  }
910 
911  if (Begin.isFileID() && End.isMacroID()) {
912  if (Range.isTokenRange()) {
913  if (!isAtEndOfMacroExpansion(End, SM, LangOpts, &End))
914  return {};
915  // Use the *original* end, not the expanded one in `End`.
916  Range.setTokenRange(isInExpansionTokenRange(Range.getEnd(), SM));
917  } else if (!isAtStartOfMacroExpansion(End, SM, LangOpts, &End))
918  return {};
919  Range.setEnd(End);
920  return makeRangeFromFileLocs(Range, SM, LangOpts);
921  }
922 
923  assert(Begin.isMacroID() && End.isMacroID());
924  SourceLocation MacroBegin, MacroEnd;
925  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
926  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
927  &MacroEnd)) ||
928  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
929  &MacroEnd)))) {
930  Range.setBegin(MacroBegin);
931  Range.setEnd(MacroEnd);
932  // Use the *original* `End`, not the expanded one in `MacroEnd`.
933  if (Range.isTokenRange())
934  Range.setTokenRange(isInExpansionTokenRange(End, SM));
935  return makeRangeFromFileLocs(Range, SM, LangOpts);
936  }
937 
938  bool Invalid = false;
939  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
940  &Invalid);
941  if (Invalid)
942  return {};
943 
944  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
945  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
946  &Invalid);
947  if (Invalid)
948  return {};
949 
950  if (EndEntry.getExpansion().isMacroArgExpansion() &&
951  BeginEntry.getExpansion().getExpansionLocStart() ==
952  EndEntry.getExpansion().getExpansionLocStart()) {
953  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
954  Range.setEnd(SM.getImmediateSpellingLoc(End));
955  return makeFileCharRange(Range, SM, LangOpts);
956  }
957  }
958 
959  return {};
960 }
961 
963  const SourceManager &SM,
964  const LangOptions &LangOpts,
965  bool *Invalid) {
966  Range = makeFileCharRange(Range, SM, LangOpts);
967  if (Range.isInvalid()) {
968  if (Invalid) *Invalid = true;
969  return {};
970  }
971 
972  // Break down the source location.
973  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
974  if (beginInfo.first.isInvalid()) {
975  if (Invalid) *Invalid = true;
976  return {};
977  }
978 
979  unsigned EndOffs;
980  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
981  beginInfo.second > EndOffs) {
982  if (Invalid) *Invalid = true;
983  return {};
984  }
985 
986  // Try to the load the file buffer.
987  bool invalidTemp = false;
988  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
989  if (invalidTemp) {
990  if (Invalid) *Invalid = true;
991  return {};
992  }
993 
994  if (Invalid) *Invalid = false;
995  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
996 }
997 
999  const SourceManager &SM,
1000  const LangOptions &LangOpts) {
1001  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
1002 
1003  // Find the location of the immediate macro expansion.
1004  while (true) {
1005  FileID FID = SM.getFileID(Loc);
1006  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
1007  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
1008  Loc = Expansion.getExpansionLocStart();
1009  if (!Expansion.isMacroArgExpansion())
1010  break;
1011 
1012  // For macro arguments we need to check that the argument did not come
1013  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
1014 
1015  // Loc points to the argument id of the macro definition, move to the
1016  // macro expansion.
1017  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1018  SourceLocation SpellLoc = Expansion.getSpellingLoc();
1019  if (SpellLoc.isFileID())
1020  break; // No inner macro.
1021 
1022  // If spelling location resides in the same FileID as macro expansion
1023  // location, it means there is no inner macro.
1024  FileID MacroFID = SM.getFileID(Loc);
1025  if (SM.isInFileID(SpellLoc, MacroFID))
1026  break;
1027 
1028  // Argument came from inner macro.
1029  Loc = SpellLoc;
1030  }
1031 
1032  // Find the spelling location of the start of the non-argument expansion
1033  // range. This is where the macro name was spelled in order to begin
1034  // expanding this macro.
1035  Loc = SM.getSpellingLoc(Loc);
1036 
1037  // Dig out the buffer where the macro name was spelled and the extents of the
1038  // name so that we can render it into the expansion note.
1039  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1040  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1041  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1042  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1043 }
1044 
1046  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1047  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
1048  // Walk past macro argument expansions.
1049  while (SM.isMacroArgExpansion(Loc))
1050  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1051 
1052  // If the macro's spelling has no FileID, then it's actually a token paste
1053  // or stringization (or similar) and not a macro at all.
1054  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1055  return {};
1056 
1057  // Find the spelling location of the start of the non-argument expansion
1058  // range. This is where the macro name was spelled in order to begin
1059  // expanding this macro.
1060  Loc = SM.getSpellingLoc(SM.getImmediateExpansionRange(Loc).getBegin());
1061 
1062  // Dig out the buffer where the macro name was spelled and the extents of the
1063  // name so that we can render it into the expansion note.
1064  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1065  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1066  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1067  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1068 }
1069 
1071  return isAsciiIdentifierContinue(c, LangOpts.DollarIdents);
1072 }
1073 
1074 bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1075  assert(isVerticalWhitespace(Str[0]));
1076  if (Str - 1 < BufferStart)
1077  return false;
1078 
1079  if ((Str[0] == '\n' && Str[-1] == '\r') ||
1080  (Str[0] == '\r' && Str[-1] == '\n')) {
1081  if (Str - 2 < BufferStart)
1082  return false;
1083  --Str;
1084  }
1085  --Str;
1086 
1087  // Rewind to first non-space character:
1088  while (Str > BufferStart && isHorizontalWhitespace(*Str))
1089  --Str;
1090 
1091  return *Str == '\\';
1092 }
1093 
1095  const SourceManager &SM) {
1096  if (Loc.isInvalid() || Loc.isMacroID())
1097  return {};
1098  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1099  if (LocInfo.first.isInvalid())
1100  return {};
1101  bool Invalid = false;
1102  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1103  if (Invalid)
1104  return {};
1105  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1106  if (!Line)
1107  return {};
1108  StringRef Rest = Buffer.substr(Line - Buffer.data());
1109  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1110  return NumWhitespaceChars == StringRef::npos
1111  ? ""
1112  : Rest.take_front(NumWhitespaceChars);
1113 }
1114 
1115 //===----------------------------------------------------------------------===//
1116 // Diagnostics forwarding code.
1117 //===----------------------------------------------------------------------===//
1118 
1119 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1120 /// lexer buffer was all expanded at a single point, perform the mapping.
1121 /// This is currently only used for _Pragma implementation, so it is the slow
1122 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1123 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1124  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1126  SourceLocation FileLoc,
1127  unsigned CharNo, unsigned TokLen) {
1128  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1129 
1130  // Otherwise, we're lexing "mapped tokens". This is used for things like
1131  // _Pragma handling. Combine the expansion location of FileLoc with the
1132  // spelling location.
1134 
1135  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1136  // characters come from spelling(FileLoc)+Offset.
1137  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1138  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1139 
1140  // Figure out the expansion loc range, which is the range covered by the
1141  // original _Pragma(...) sequence.
1142  CharSourceRange II = SM.getImmediateExpansionRange(FileLoc);
1143 
1144  return SM.createExpansionLoc(SpellingLoc, II.getBegin(), II.getEnd(), TokLen);
1145 }
1146 
1147 /// getSourceLocation - Return a source location identifier for the specified
1148 /// offset in the current file.
1150  unsigned TokLen) const {
1151  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1152  "Location out of range for this buffer!");
1153 
1154  // In the normal case, we're just lexing from a simple file buffer, return
1155  // the file id from FileLoc with the offset specified.
1156  unsigned CharNo = Loc-BufferStart;
1157  if (FileLoc.isFileID())
1158  return FileLoc.getLocWithOffset(CharNo);
1159 
1160  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1161  // tokens are lexed from where the _Pragma was defined.
1162  assert(PP && "This doesn't work on raw lexers");
1163  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1164 }
1165 
1166 /// Diag - Forwarding function for diagnostics. This translate a source
1167 /// position in the current buffer into a SourceLocation object for rendering.
1168 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1169  return PP->Diag(getSourceLocation(Loc), DiagID);
1170 }
1171 
1172 //===----------------------------------------------------------------------===//
1173 // Trigraph and Escaped Newline Handling Code.
1174 //===----------------------------------------------------------------------===//
1175 
1176 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1177 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1178 static char GetTrigraphCharForLetter(char Letter) {
1179  switch (Letter) {
1180  default: return 0;
1181  case '=': return '#';
1182  case ')': return ']';
1183  case '(': return '[';
1184  case '!': return '|';
1185  case '\'': return '^';
1186  case '>': return '}';
1187  case '/': return '\\';
1188  case '<': return '{';
1189  case '-': return '~';
1190  }
1191 }
1192 
1193 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1194 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1195 /// return the result character. Finally, emit a warning about trigraph use
1196 /// whether trigraphs are enabled or not.
1197 static char DecodeTrigraphChar(const char *CP, Lexer *L, bool Trigraphs) {
1198  char Res = GetTrigraphCharForLetter(*CP);
1199  if (!Res || !L) return Res;
1200 
1201  if (!Trigraphs) {
1202  if (!L->isLexingRawMode())
1203  L->Diag(CP-2, diag::trigraph_ignored);
1204  return 0;
1205  }
1206 
1207  if (!L->isLexingRawMode())
1208  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1209  return Res;
1210 }
1211 
1212 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1213 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1214 /// trigraph equivalent on entry to this function.
1215 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1216  unsigned Size = 0;
1217  while (isWhitespace(Ptr[Size])) {
1218  ++Size;
1219 
1220  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1221  continue;
1222 
1223  // If this is a \r\n or \n\r, skip the other half.
1224  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1225  Ptr[Size-1] != Ptr[Size])
1226  ++Size;
1227 
1228  return Size;
1229  }
1230 
1231  // Not an escaped newline, must be a \t or something else.
1232  return 0;
1233 }
1234 
1235 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1236 /// them), skip over them and return the first non-escaped-newline found,
1237 /// otherwise return P.
1238 const char *Lexer::SkipEscapedNewLines(const char *P) {
1239  while (true) {
1240  const char *AfterEscape;
1241  if (*P == '\\') {
1242  AfterEscape = P+1;
1243  } else if (*P == '?') {
1244  // If not a trigraph for escape, bail out.
1245  if (P[1] != '?' || P[2] != '/')
1246  return P;
1247  // FIXME: Take LangOpts into account; the language might not
1248  // support trigraphs.
1249  AfterEscape = P+3;
1250  } else {
1251  return P;
1252  }
1253 
1254  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1255  if (NewLineSize == 0) return P;
1256  P = AfterEscape+NewLineSize;
1257  }
1258 }
1259 
1261  const SourceManager &SM,
1262  const LangOptions &LangOpts) {
1263  if (Loc.isMacroID()) {
1264  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1265  return None;
1266  }
1267  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1268 
1269  // Break down the source location.
1270  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1271 
1272  // Try to load the file buffer.
1273  bool InvalidTemp = false;
1274  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1275  if (InvalidTemp)
1276  return None;
1277 
1278  const char *TokenBegin = File.data() + LocInfo.second;
1279 
1280  // Lex from the start of the given location.
1281  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1282  TokenBegin, File.end());
1283  // Find the token.
1284  Token Tok;
1285  lexer.LexFromRawLexer(Tok);
1286  return Tok;
1287 }
1288 
1289 /// Checks that the given token is the first token that occurs after the
1290 /// given location (this excludes comments and whitespace). Returns the location
1291 /// immediately after the specified token. If the token is not found or the
1292 /// location is inside a macro, the returned source location will be invalid.
1294  SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1295  const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1296  Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1297  if (!Tok || Tok->isNot(TKind))
1298  return {};
1299  SourceLocation TokenLoc = Tok->getLocation();
1300 
1301  // Calculate how much whitespace needs to be skipped if any.
1302  unsigned NumWhitespaceChars = 0;
1303  if (SkipTrailingWhitespaceAndNewLine) {
1304  const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1305  unsigned char C = *TokenEnd;
1306  while (isHorizontalWhitespace(C)) {
1307  C = *(++TokenEnd);
1308  NumWhitespaceChars++;
1309  }
1310 
1311  // Skip \r, \n, \r\n, or \n\r
1312  if (C == '\n' || C == '\r') {
1313  char PrevC = C;
1314  C = *(++TokenEnd);
1315  NumWhitespaceChars++;
1316  if ((C == '\n' || C == '\r') && C != PrevC)
1317  NumWhitespaceChars++;
1318  }
1319  }
1320 
1321  return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1322 }
1323 
1324 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1325 /// get its size, and return it. This is tricky in several cases:
1326 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1327 /// then either return the trigraph (skipping 3 chars) or the '?',
1328 /// depending on whether trigraphs are enabled or not.
1329 /// 2. If this is an escaped newline (potentially with whitespace between
1330 /// the backslash and newline), implicitly skip the newline and return
1331 /// the char after it.
1332 ///
1333 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1334 /// know that we can accumulate into Size, and that we have already incremented
1335 /// Ptr by Size bytes.
1336 ///
1337 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1338 /// be updated to match.
1339 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1340  Token *Tok) {
1341  // If we have a slash, look for an escaped newline.
1342  if (Ptr[0] == '\\') {
1343  ++Size;
1344  ++Ptr;
1345 Slash:
1346  // Common case, backslash-char where the char is not whitespace.
1347  if (!isWhitespace(Ptr[0])) return '\\';
1348 
1349  // See if we have optional whitespace characters between the slash and
1350  // newline.
1351  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1352  // Remember that this token needs to be cleaned.
1353  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1354 
1355  // Warn if there was whitespace between the backslash and newline.
1356  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1357  Diag(Ptr, diag::backslash_newline_space);
1358 
1359  // Found backslash<whitespace><newline>. Parse the char after it.
1360  Size += EscapedNewLineSize;
1361  Ptr += EscapedNewLineSize;
1362 
1363  // Use slow version to accumulate a correct size field.
1364  return getCharAndSizeSlow(Ptr, Size, Tok);
1365  }
1366 
1367  // Otherwise, this is not an escaped newline, just return the slash.
1368  return '\\';
1369  }
1370 
1371  // If this is a trigraph, process it.
1372  if (Ptr[0] == '?' && Ptr[1] == '?') {
1373  // If this is actually a legal trigraph (not something like "??x"), emit
1374  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1375  if (char C = DecodeTrigraphChar(Ptr + 2, Tok ? this : nullptr,
1376  LangOpts.Trigraphs)) {
1377  // Remember that this token needs to be cleaned.
1378  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1379 
1380  Ptr += 3;
1381  Size += 3;
1382  if (C == '\\') goto Slash;
1383  return C;
1384  }
1385  }
1386 
1387  // If this is neither, return a single character.
1388  ++Size;
1389  return *Ptr;
1390 }
1391 
1392 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1393 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1394 /// and that we have already incremented Ptr by Size bytes.
1395 ///
1396 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1397 /// be updated to match.
1398 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1399  const LangOptions &LangOpts) {
1400  // If we have a slash, look for an escaped newline.
1401  if (Ptr[0] == '\\') {
1402  ++Size;
1403  ++Ptr;
1404 Slash:
1405  // Common case, backslash-char where the char is not whitespace.
1406  if (!isWhitespace(Ptr[0])) return '\\';
1407 
1408  // See if we have optional whitespace characters followed by a newline.
1409  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1410  // Found backslash<whitespace><newline>. Parse the char after it.
1411  Size += EscapedNewLineSize;
1412  Ptr += EscapedNewLineSize;
1413 
1414  // Use slow version to accumulate a correct size field.
1415  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1416  }
1417 
1418  // Otherwise, this is not an escaped newline, just return the slash.
1419  return '\\';
1420  }
1421 
1422  // If this is a trigraph, process it.
1423  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1424  // If this is actually a legal trigraph (not something like "??x"), return
1425  // it.
1426  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1427  Ptr += 3;
1428  Size += 3;
1429  if (C == '\\') goto Slash;
1430  return C;
1431  }
1432  }
1433 
1434  // If this is neither, return a single character.
1435  ++Size;
1436  return *Ptr;
1437 }
1438 
1439 //===----------------------------------------------------------------------===//
1440 // Helper methods for lexing.
1441 //===----------------------------------------------------------------------===//
1442 
1443 /// Routine that indiscriminately sets the offset into the source file.
1444 void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1445  BufferPtr = BufferStart + Offset;
1446  if (BufferPtr > BufferEnd)
1447  BufferPtr = BufferEnd;
1448  // FIXME: What exactly does the StartOfLine bit mean? There are two
1449  // possible meanings for the "start" of the line: the first token on the
1450  // unexpanded line, or the first token on the expanded line.
1451  IsAtStartOfLine = StartOfLine;
1452  IsAtPhysicalStartOfLine = StartOfLine;
1453 }
1454 
1455 static bool isUnicodeWhitespace(uint32_t Codepoint) {
1456  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
1458  return UnicodeWhitespaceChars.contains(Codepoint);
1459 }
1460 
1461 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1462  if (LangOpts.AsmPreprocessor) {
1463  return false;
1464  } else if (LangOpts.DollarIdents && '$' == C) {
1465  return true;
1466  } else if (LangOpts.CPlusPlus) {
1467  // A non-leading codepoint must have the XID_Continue property.
1468  // XIDContinueRanges doesn't contains characters also in XIDStartRanges,
1469  // so we need to check both tables.
1470  // '_' doesn't have the XID_Continue property but is allowed in C++.
1471  static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
1472  static const llvm::sys::UnicodeCharSet XIDContinueChars(XIDContinueRanges);
1473  return C == '_' || XIDStartChars.contains(C) ||
1474  XIDContinueChars.contains(C);
1475  } else if (LangOpts.C11) {
1476  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1478  return C11AllowedIDChars.contains(C);
1479  } else {
1480  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1482  return C99AllowedIDChars.contains(C);
1483  }
1484 }
1485 
1486 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1487  if (LangOpts.AsmPreprocessor) {
1488  return false;
1489  }
1490  if (LangOpts.CPlusPlus) {
1491  static const llvm::sys::UnicodeCharSet XIDStartChars(XIDStartRanges);
1492  // '_' doesn't have the XID_Start property but is allowed in C++.
1493  return C == '_' || XIDStartChars.contains(C);
1494  }
1495  if (!isAllowedIDChar(C, LangOpts))
1496  return false;
1497  if (LangOpts.C11) {
1498  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1500  return !C11DisallowedInitialIDChars.contains(C);
1501  }
1502  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1504  return !C99DisallowedInitialIDChars.contains(C);
1505 }
1506 
1507 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1508  const char *End) {
1510  L.getSourceLocation(End));
1511 }
1512 
1513 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1514  CharSourceRange Range, bool IsFirst) {
1515  // Check C99 compatibility.
1516  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1517  enum {
1518  CannotAppearInIdentifier = 0,
1519  CannotStartIdentifier
1520  };
1521 
1522  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1524  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1526  if (!C99AllowedIDChars.contains(C)) {
1527  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1528  << Range
1529  << CannotAppearInIdentifier;
1530  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1531  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1532  << Range
1533  << CannotStartIdentifier;
1534  }
1535  }
1536 }
1537 
1538 /// After encountering UTF-8 character C and interpreting it as an identifier
1539 /// character, check whether it's a homoglyph for a common non-identifier
1540 /// source character that is unlikely to be an intentional identifier
1541 /// character and warn if so.
1542 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1543  CharSourceRange Range) {
1544  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1545  struct HomoglyphPair {
1546  uint32_t Character;
1547  char LooksLike;
1548  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1549  };
1550  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1551  {U'\u00ad', 0}, // SOFT HYPHEN
1552  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1553  {U'\u037e', ';'}, // GREEK QUESTION MARK
1554  {U'\u200b', 0}, // ZERO WIDTH SPACE
1555  {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1556  {U'\u200d', 0}, // ZERO WIDTH JOINER
1557  {U'\u2060', 0}, // WORD JOINER
1558  {U'\u2061', 0}, // FUNCTION APPLICATION
1559  {U'\u2062', 0}, // INVISIBLE TIMES
1560  {U'\u2063', 0}, // INVISIBLE SEPARATOR
1561  {U'\u2064', 0}, // INVISIBLE PLUS
1562  {U'\u2212', '-'}, // MINUS SIGN
1563  {U'\u2215', '/'}, // DIVISION SLASH
1564  {U'\u2216', '\\'}, // SET MINUS
1565  {U'\u2217', '*'}, // ASTERISK OPERATOR
1566  {U'\u2223', '|'}, // DIVIDES
1567  {U'\u2227', '^'}, // LOGICAL AND
1568  {U'\u2236', ':'}, // RATIO
1569  {U'\u223c', '~'}, // TILDE OPERATOR
1570  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1571  {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1572  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1573  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1574  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1575  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1576  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1577  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1578  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1579  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1580  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1581  {U'\uff0c', ','}, // FULLWIDTH COMMA
1582  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1583  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1584  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1585  {U'\uff1a', ':'}, // FULLWIDTH COLON
1586  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1587  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1588  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1589  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1590  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1591  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1592  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1593  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1594  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1595  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1596  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1597  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1598  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1599  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1600  {0, 0}
1601  };
1602  auto Homoglyph =
1603  std::lower_bound(std::begin(SortedHomoglyphs),
1604  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1605  if (Homoglyph->Character == C) {
1606  llvm::SmallString<5> CharBuf;
1607  {
1608  llvm::raw_svector_ostream CharOS(CharBuf);
1609  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1610  }
1611  if (Homoglyph->LooksLike) {
1612  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1613  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1614  << Range << CharBuf << LooksLikeStr;
1615  } else {
1616  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1617  << Range << CharBuf;
1618  }
1619  }
1620 }
1621 
1623  DiagnosticsEngine &Diags, const LangOptions &LangOpts, uint32_t CodePoint,
1624  CharSourceRange Range, bool IsFirst) {
1625  if (isASCII(CodePoint))
1626  return;
1627 
1628  bool IsIDStart = isAllowedInitiallyIDChar(CodePoint, LangOpts);
1629  bool IsIDContinue = IsIDStart || isAllowedIDChar(CodePoint, LangOpts);
1630 
1631  if ((IsFirst && IsIDStart) || (!IsFirst && IsIDContinue))
1632  return;
1633 
1634  bool InvalidOnlyAtStart = IsFirst && !IsIDStart && IsIDContinue;
1635 
1636  llvm::SmallString<5> CharBuf;
1637  llvm::raw_svector_ostream CharOS(CharBuf);
1638  llvm::write_hex(CharOS, CodePoint, llvm::HexPrintStyle::Upper, 4);
1639 
1640  if (!IsFirst || InvalidOnlyAtStart) {
1641  Diags.Report(Range.getBegin(), diag::err_character_not_allowed_identifier)
1642  << Range << CharBuf << int(InvalidOnlyAtStart)
1643  << FixItHint::CreateRemoval(Range);
1644  } else {
1645  Diags.Report(Range.getBegin(), diag::err_character_not_allowed)
1646  << Range << CharBuf << FixItHint::CreateRemoval(Range);
1647  }
1648 }
1649 
1650 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1651  Token &Result) {
1652  const char *UCNPtr = CurPtr + Size;
1653  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1654  if (CodePoint == 0) {
1655  return false;
1656  }
1657 
1658  if (!isAllowedIDChar(CodePoint, LangOpts)) {
1659  if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
1660  return false;
1664  PP->getDiagnostics(), LangOpts, CodePoint,
1665  makeCharRange(*this, CurPtr, UCNPtr),
1666  /*IsFirst=*/false);
1667 
1668  // We got a unicode codepoint that is neither a space nor a
1669  // a valid identifier part.
1670  // Carry on as if the codepoint was valid for recovery purposes.
1671  } else if (!isLexingRawMode())
1673  makeCharRange(*this, CurPtr, UCNPtr),
1674  /*IsFirst=*/false);
1675 
1676  Result.setFlag(Token::HasUCN);
1677  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1678  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1679  CurPtr = UCNPtr;
1680  else
1681  while (CurPtr != UCNPtr)
1682  (void)getAndAdvanceChar(CurPtr, Result);
1683  return true;
1684 }
1685 
1686 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1687  const char *UnicodePtr = CurPtr;
1688  llvm::UTF32 CodePoint;
1689  llvm::ConversionResult Result =
1690  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1691  (const llvm::UTF8 *)BufferEnd,
1692  &CodePoint,
1693  llvm::strictConversion);
1694  if (Result != llvm::conversionOK)
1695  return false;
1696 
1697  if (!isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts)) {
1698  if (isASCII(CodePoint) || isUnicodeWhitespace(CodePoint))
1699  return false;
1700 
1704  PP->getDiagnostics(), LangOpts, CodePoint,
1705  makeCharRange(*this, CurPtr, UnicodePtr), /*IsFirst=*/false);
1706  // We got a unicode codepoint that is neither a space nor a
1707  // a valid identifier part. Carry on as if the codepoint was
1708  // valid for recovery purposes.
1709  } else if (!isLexingRawMode()) {
1711  makeCharRange(*this, CurPtr, UnicodePtr),
1712  /*IsFirst=*/false);
1714  makeCharRange(*this, CurPtr, UnicodePtr));
1715  }
1716 
1717  CurPtr = UnicodePtr;
1718  return true;
1719 }
1720 
1721 bool Lexer::LexUnicodeIdentifierStart(Token &Result, uint32_t C,
1722  const char *CurPtr) {
1723  if (isAllowedInitiallyIDChar(C, LangOpts)) {
1725  !PP->isPreprocessedOutput()) {
1727  makeCharRange(*this, BufferPtr, CurPtr),
1728  /*IsFirst=*/true);
1730  makeCharRange(*this, BufferPtr, CurPtr));
1731  }
1732 
1733  MIOpt.ReadToken();
1734  return LexIdentifierContinue(Result, CurPtr);
1735  }
1736 
1738  !PP->isPreprocessedOutput() && !isASCII(*BufferPtr) &&
1739  !isAllowedInitiallyIDChar(C, LangOpts) && !isUnicodeWhitespace(C)) {
1740  // Non-ASCII characters tend to creep into source code unintentionally.
1741  // Instead of letting the parser complain about the unknown token,
1742  // just drop the character.
1743  // Note that we can /only/ do this when the non-ASCII character is actually
1744  // spelled as Unicode, not written as a UCN. The standard requires that
1745  // we not throw away any possible preprocessor tokens, but there's a
1746  // loophole in the mapping of Unicode characters to basic character set
1747  // characters that allows us to map these particular characters to, say,
1748  // whitespace.
1750  PP->getDiagnostics(), LangOpts, C,
1751  makeCharRange(*this, BufferPtr, CurPtr), /*IsStart*/ true);
1752  BufferPtr = CurPtr;
1753  return false;
1754  }
1755 
1756  // Otherwise, we have an explicit UCN or a character that's unlikely to show
1757  // up by accident.
1758  MIOpt.ReadToken();
1759  FormTokenWithChars(Result, CurPtr, tok::unknown);
1760  return true;
1761 }
1762 
1763 bool Lexer::LexIdentifierContinue(Token &Result, const char *CurPtr) {
1764  // Match [_A-Za-z0-9]*, we have already matched an identifier start.
1765  while (true) {
1766  unsigned char C = *CurPtr;
1767  // Fast path.
1768  if (isAsciiIdentifierContinue(C)) {
1769  ++CurPtr;
1770  continue;
1771  }
1772 
1773  unsigned Size;
1774  // Slow path: handle trigraph, unicode codepoints, UCNs.
1775  C = getCharAndSize(CurPtr, Size);
1776  if (isAsciiIdentifierContinue(C)) {
1777  CurPtr = ConsumeChar(CurPtr, Size, Result);
1778  continue;
1779  }
1780  if (C == '$') {
1781  // If we hit a $ and they are not supported in identifiers, we are done.
1782  if (!LangOpts.DollarIdents)
1783  break;
1784  // Otherwise, emit a diagnostic and continue.
1785  if (!isLexingRawMode())
1786  Diag(CurPtr, diag::ext_dollar_in_identifier);
1787  CurPtr = ConsumeChar(CurPtr, Size, Result);
1788  continue;
1789  }
1790  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1791  continue;
1792  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1793  continue;
1794  // Neither an expected Unicode codepoint nor a UCN.
1795  break;
1796  }
1797 
1798  const char *IdStart = BufferPtr;
1799  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1800  Result.setRawIdentifierData(IdStart);
1801 
1802  // If we are in raw mode, return this identifier raw. There is no need to
1803  // look up identifier information or attempt to macro expand it.
1804  if (LexingRawMode)
1805  return true;
1806 
1807  // Fill in Result.IdentifierInfo and update the token kind,
1808  // looking up the identifier in the identifier table.
1809  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1810  // Note that we have to call PP->LookUpIdentifierInfo() even for code
1811  // completion, it writes IdentifierInfo into Result, and callers rely on it.
1812 
1813  // If the completion point is at the end of an identifier, we want to treat
1814  // the identifier as incomplete even if it resolves to a macro or a keyword.
1815  // This allows e.g. 'class^' to complete to 'classifier'.
1816  if (isCodeCompletionPoint(CurPtr)) {
1817  // Return the code-completion token.
1818  Result.setKind(tok::code_completion);
1819  // Skip the code-completion char and all immediate identifier characters.
1820  // This ensures we get consistent behavior when completing at any point in
1821  // an identifier (i.e. at the start, in the middle, at the end). Note that
1822  // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1823  // simpler.
1824  assert(*CurPtr == 0 && "Completion character must be 0");
1825  ++CurPtr;
1826  // Note that code completion token is not added as a separate character
1827  // when the completion point is at the end of the buffer. Therefore, we need
1828  // to check if the buffer has ended.
1829  if (CurPtr < BufferEnd) {
1830  while (isAsciiIdentifierContinue(*CurPtr))
1831  ++CurPtr;
1832  }
1833  BufferPtr = CurPtr;
1834  return true;
1835  }
1836 
1837  // Finally, now that we know we have an identifier, pass this off to the
1838  // preprocessor, which may macro expand it or something.
1839  if (II->isHandleIdentifierCase())
1840  return PP->HandleIdentifier(Result);
1841 
1842  return true;
1843 }
1844 
1845 /// isHexaLiteral - Return true if Start points to a hex constant.
1846 /// in microsoft mode (where this is supposed to be several different tokens).
1847 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1848  unsigned Size;
1849  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1850  if (C1 != '0')
1851  return false;
1852  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1853  return (C2 == 'x' || C2 == 'X');
1854 }
1855 
1856 /// LexNumericConstant - Lex the remainder of a integer or floating point
1857 /// constant. From[-1] is the first character lexed. Return the end of the
1858 /// constant.
1859 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1860  unsigned Size;
1861  char C = getCharAndSize(CurPtr, Size);
1862  char PrevCh = 0;
1863  while (isPreprocessingNumberBody(C)) {
1864  CurPtr = ConsumeChar(CurPtr, Size, Result);
1865  PrevCh = C;
1866  C = getCharAndSize(CurPtr, Size);
1867  }
1868 
1869  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1870  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1871  // If we are in Microsoft mode, don't continue if the constant is hex.
1872  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1873  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1874  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1875  }
1876 
1877  // If we have a hex FP constant, continue.
1878  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1879  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1880  // not-quite-conforming extension. Only do so if this looks like it's
1881  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1882  bool IsHexFloat = true;
1883  if (!LangOpts.C99) {
1884  if (!isHexaLiteral(BufferPtr, LangOpts))
1885  IsHexFloat = false;
1886  else if (!LangOpts.CPlusPlus17 &&
1887  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1888  IsHexFloat = false;
1889  }
1890  if (IsHexFloat)
1891  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1892  }
1893 
1894  // If we have a digit separator, continue.
1895  if (C == '\'' && (LangOpts.CPlusPlus14 || LangOpts.C2x)) {
1896  unsigned NextSize;
1897  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, LangOpts);
1898  if (isAsciiIdentifierContinue(Next)) {
1899  if (!isLexingRawMode())
1900  Diag(CurPtr, LangOpts.CPlusPlus
1901  ? diag::warn_cxx11_compat_digit_separator
1902  : diag::warn_c2x_compat_digit_separator);
1903  CurPtr = ConsumeChar(CurPtr, Size, Result);
1904  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1905  return LexNumericConstant(Result, CurPtr);
1906  }
1907  }
1908 
1909  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1910  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1911  return LexNumericConstant(Result, CurPtr);
1912  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1913  return LexNumericConstant(Result, CurPtr);
1914 
1915  // Update the location of token as well as BufferPtr.
1916  const char *TokStart = BufferPtr;
1917  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1918  Result.setLiteralData(TokStart);
1919  return true;
1920 }
1921 
1922 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1923 /// in C++11, or warn on a ud-suffix in C++98.
1924 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1925  bool IsStringLiteral) {
1926  assert(LangOpts.CPlusPlus);
1927 
1928  // Maximally munch an identifier.
1929  unsigned Size;
1930  char C = getCharAndSize(CurPtr, Size);
1931  bool Consumed = false;
1932 
1933  if (!isAsciiIdentifierStart(C)) {
1934  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1935  Consumed = true;
1936  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1937  Consumed = true;
1938  else
1939  return CurPtr;
1940  }
1941 
1942  if (!LangOpts.CPlusPlus11) {
1943  if (!isLexingRawMode())
1944  Diag(CurPtr,
1945  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1946  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1948  return CurPtr;
1949  }
1950 
1951  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1952  // that does not start with an underscore is ill-formed. As a conforming
1953  // extension, we treat all such suffixes as if they had whitespace before
1954  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1955  // likely to be a ud-suffix than a macro, however, and accept that.
1956  if (!Consumed) {
1957  bool IsUDSuffix = false;
1958  if (C == '_')
1959  IsUDSuffix = true;
1960  else if (IsStringLiteral && LangOpts.CPlusPlus14) {
1961  // In C++1y, we need to look ahead a few characters to see if this is a
1962  // valid suffix for a string literal or a numeric literal (this could be
1963  // the 'operator""if' defining a numeric literal operator).
1964  const unsigned MaxStandardSuffixLength = 3;
1965  char Buffer[MaxStandardSuffixLength] = { C };
1966  unsigned Consumed = Size;
1967  unsigned Chars = 1;
1968  while (true) {
1969  unsigned NextSize;
1970  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize, LangOpts);
1971  if (!isAsciiIdentifierContinue(Next)) {
1972  // End of suffix. Check whether this is on the allowed list.
1973  const StringRef CompleteSuffix(Buffer, Chars);
1974  IsUDSuffix =
1975  StringLiteralParser::isValidUDSuffix(LangOpts, CompleteSuffix);
1976  break;
1977  }
1978 
1979  if (Chars == MaxStandardSuffixLength)
1980  // Too long: can't be a standard suffix.
1981  break;
1982 
1983  Buffer[Chars++] = Next;
1984  Consumed += NextSize;
1985  }
1986  }
1987 
1988  if (!IsUDSuffix) {
1989  if (!isLexingRawMode())
1990  Diag(CurPtr, LangOpts.MSVCCompat
1991  ? diag::ext_ms_reserved_user_defined_literal
1992  : diag::ext_reserved_user_defined_literal)
1994  return CurPtr;
1995  }
1996 
1997  CurPtr = ConsumeChar(CurPtr, Size, Result);
1998  }
1999 
2000  Result.setFlag(Token::HasUDSuffix);
2001  while (true) {
2002  C = getCharAndSize(CurPtr, Size);
2003  if (isAsciiIdentifierContinue(C)) {
2004  CurPtr = ConsumeChar(CurPtr, Size, Result);
2005  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
2006  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
2007  } else
2008  break;
2009  }
2010 
2011  return CurPtr;
2012 }
2013 
2014 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
2015 /// either " or L" or u8" or u" or U".
2016 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
2017  tok::TokenKind Kind) {
2018  const char *AfterQuote = CurPtr;
2019  // Does this string contain the \0 character?
2020  const char *NulCharacter = nullptr;
2021 
2022  if (!isLexingRawMode() &&
2023  (Kind == tok::utf8_string_literal ||
2024  Kind == tok::utf16_string_literal ||
2025  Kind == tok::utf32_string_literal))
2026  Diag(BufferPtr, LangOpts.CPlusPlus ? diag::warn_cxx98_compat_unicode_literal
2027  : diag::warn_c99_compat_unicode_literal);
2028 
2029  char C = getAndAdvanceChar(CurPtr, Result);
2030  while (C != '"') {
2031  // Skip escaped characters. Escaped newlines will already be processed by
2032  // getAndAdvanceChar.
2033  if (C == '\\')
2034  C = getAndAdvanceChar(CurPtr, Result);
2035 
2036  if (C == '\n' || C == '\r' || // Newline.
2037  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2038  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2039  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
2040  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2041  return true;
2042  }
2043 
2044  if (C == 0) {
2045  if (isCodeCompletionPoint(CurPtr-1)) {
2046  if (ParsingFilename)
2047  codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
2048  else
2050  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2051  cutOffLexing();
2052  return true;
2053  }
2054 
2055  NulCharacter = CurPtr-1;
2056  }
2057  C = getAndAdvanceChar(CurPtr, Result);
2058  }
2059 
2060  // If we are in C++11, lex the optional ud-suffix.
2061  if (LangOpts.CPlusPlus)
2062  CurPtr = LexUDSuffix(Result, CurPtr, true);
2063 
2064  // If a nul character existed in the string, warn about it.
2065  if (NulCharacter && !isLexingRawMode())
2066  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2067 
2068  // Update the location of the token as well as the BufferPtr instance var.
2069  const char *TokStart = BufferPtr;
2070  FormTokenWithChars(Result, CurPtr, Kind);
2071  Result.setLiteralData(TokStart);
2072  return true;
2073 }
2074 
2075 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
2076 /// having lexed R", LR", u8R", uR", or UR".
2077 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
2078  tok::TokenKind Kind) {
2079  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
2080  // Between the initial and final double quote characters of the raw string,
2081  // any transformations performed in phases 1 and 2 (trigraphs,
2082  // universal-character-names, and line splicing) are reverted.
2083 
2084  if (!isLexingRawMode())
2085  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
2086 
2087  unsigned PrefixLen = 0;
2088 
2089  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
2090  ++PrefixLen;
2091 
2092  // If the last character was not a '(', then we didn't lex a valid delimiter.
2093  if (CurPtr[PrefixLen] != '(') {
2094  if (!isLexingRawMode()) {
2095  const char *PrefixEnd = &CurPtr[PrefixLen];
2096  if (PrefixLen == 16) {
2097  Diag(PrefixEnd, diag::err_raw_delim_too_long);
2098  } else {
2099  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
2100  << StringRef(PrefixEnd, 1);
2101  }
2102  }
2103 
2104  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
2105  // it's possible the '"' was intended to be part of the raw string, but
2106  // there's not much we can do about that.
2107  while (true) {
2108  char C = *CurPtr++;
2109 
2110  if (C == '"')
2111  break;
2112  if (C == 0 && CurPtr-1 == BufferEnd) {
2113  --CurPtr;
2114  break;
2115  }
2116  }
2117 
2118  FormTokenWithChars(Result, CurPtr, tok::unknown);
2119  return true;
2120  }
2121 
2122  // Save prefix and move CurPtr past it
2123  const char *Prefix = CurPtr;
2124  CurPtr += PrefixLen + 1; // skip over prefix and '('
2125 
2126  while (true) {
2127  char C = *CurPtr++;
2128 
2129  if (C == ')') {
2130  // Check for prefix match and closing quote.
2131  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2132  CurPtr += PrefixLen + 1; // skip over prefix and '"'
2133  break;
2134  }
2135  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2136  if (!isLexingRawMode())
2137  Diag(BufferPtr, diag::err_unterminated_raw_string)
2138  << StringRef(Prefix, PrefixLen);
2139  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2140  return true;
2141  }
2142  }
2143 
2144  // If we are in C++11, lex the optional ud-suffix.
2145  if (LangOpts.CPlusPlus)
2146  CurPtr = LexUDSuffix(Result, CurPtr, true);
2147 
2148  // Update the location of token as well as BufferPtr.
2149  const char *TokStart = BufferPtr;
2150  FormTokenWithChars(Result, CurPtr, Kind);
2151  Result.setLiteralData(TokStart);
2152  return true;
2153 }
2154 
2155 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2156 /// after having lexed the '<' character. This is used for #include filenames.
2157 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2158  // Does this string contain the \0 character?
2159  const char *NulCharacter = nullptr;
2160  const char *AfterLessPos = CurPtr;
2161  char C = getAndAdvanceChar(CurPtr, Result);
2162  while (C != '>') {
2163  // Skip escaped characters. Escaped newlines will already be processed by
2164  // getAndAdvanceChar.
2165  if (C == '\\')
2166  C = getAndAdvanceChar(CurPtr, Result);
2167 
2168  if (isVerticalWhitespace(C) || // Newline.
2169  (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2170  // If the filename is unterminated, then it must just be a lone <
2171  // character. Return this as such.
2172  FormTokenWithChars(Result, AfterLessPos, tok::less);
2173  return true;
2174  }
2175 
2176  if (C == 0) {
2177  if (isCodeCompletionPoint(CurPtr - 1)) {
2178  codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2179  cutOffLexing();
2180  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2181  return true;
2182  }
2183  NulCharacter = CurPtr-1;
2184  }
2185  C = getAndAdvanceChar(CurPtr, Result);
2186  }
2187 
2188  // If a nul character existed in the string, warn about it.
2189  if (NulCharacter && !isLexingRawMode())
2190  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2191 
2192  // Update the location of token as well as BufferPtr.
2193  const char *TokStart = BufferPtr;
2194  FormTokenWithChars(Result, CurPtr, tok::header_name);
2195  Result.setLiteralData(TokStart);
2196  return true;
2197 }
2198 
2199 void Lexer::codeCompleteIncludedFile(const char *PathStart,
2200  const char *CompletionPoint,
2201  bool IsAngled) {
2202  // Completion only applies to the filename, after the last slash.
2203  StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2204  llvm::StringRef SlashChars = LangOpts.MSVCCompat ? "/\\" : "/";
2205  auto Slash = PartialPath.find_last_of(SlashChars);
2206  StringRef Dir =
2207  (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2208  const char *StartOfFilename =
2209  (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2210  // Code completion filter range is the filename only, up to completion point.
2212  StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2213  // We should replace the characters up to the closing quote or closest slash,
2214  // if any.
2215  while (CompletionPoint < BufferEnd) {
2216  char Next = *(CompletionPoint + 1);
2217  if (Next == 0 || Next == '\r' || Next == '\n')
2218  break;
2219  ++CompletionPoint;
2220  if (Next == (IsAngled ? '>' : '"'))
2221  break;
2222  if (llvm::is_contained(SlashChars, Next))
2223  break;
2224  }
2225 
2227  FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2228  FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2229  PP->CodeCompleteIncludedFile(Dir, IsAngled);
2230 }
2231 
2232 /// LexCharConstant - Lex the remainder of a character constant, after having
2233 /// lexed either ' or L' or u8' or u' or U'.
2234 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2235  tok::TokenKind Kind) {
2236  // Does this character contain the \0 character?
2237  const char *NulCharacter = nullptr;
2238 
2239  if (!isLexingRawMode()) {
2240  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2241  Diag(BufferPtr, LangOpts.CPlusPlus
2242  ? diag::warn_cxx98_compat_unicode_literal
2243  : diag::warn_c99_compat_unicode_literal);
2244  else if (Kind == tok::utf8_char_constant)
2245  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2246  }
2247 
2248  char C = getAndAdvanceChar(CurPtr, Result);
2249  if (C == '\'') {
2250  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2251  Diag(BufferPtr, diag::ext_empty_character);
2252  FormTokenWithChars(Result, CurPtr, tok::unknown);
2253  return true;
2254  }
2255 
2256  while (C != '\'') {
2257  // Skip escaped characters.
2258  if (C == '\\')
2259  C = getAndAdvanceChar(CurPtr, Result);
2260 
2261  if (C == '\n' || C == '\r' || // Newline.
2262  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2263  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2264  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2265  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2266  return true;
2267  }
2268 
2269  if (C == 0) {
2270  if (isCodeCompletionPoint(CurPtr-1)) {
2272  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2273  cutOffLexing();
2274  return true;
2275  }
2276 
2277  NulCharacter = CurPtr-1;
2278  }
2279  C = getAndAdvanceChar(CurPtr, Result);
2280  }
2281 
2282  // If we are in C++11, lex the optional ud-suffix.
2283  if (LangOpts.CPlusPlus)
2284  CurPtr = LexUDSuffix(Result, CurPtr, false);
2285 
2286  // If a nul character existed in the character, warn about it.
2287  if (NulCharacter && !isLexingRawMode())
2288  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2289 
2290  // Update the location of token as well as BufferPtr.
2291  const char *TokStart = BufferPtr;
2292  FormTokenWithChars(Result, CurPtr, Kind);
2293  Result.setLiteralData(TokStart);
2294  return true;
2295 }
2296 
2297 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2298 /// Update BufferPtr to point to the next non-whitespace character and return.
2299 ///
2300 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2301 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2302  bool &TokAtPhysicalStartOfLine) {
2303  // Whitespace - Skip it, then return the token after the whitespace.
2304  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2305 
2306  unsigned char Char = *CurPtr;
2307 
2308  const char *lastNewLine = nullptr;
2309  auto setLastNewLine = [&](const char *Ptr) {
2310  lastNewLine = Ptr;
2311  if (!NewLinePtr)
2312  NewLinePtr = Ptr;
2313  };
2314  if (SawNewline)
2315  setLastNewLine(CurPtr - 1);
2316 
2317  // Skip consecutive spaces efficiently.
2318  while (true) {
2319  // Skip horizontal whitespace very aggressively.
2320  while (isHorizontalWhitespace(Char))
2321  Char = *++CurPtr;
2322 
2323  // Otherwise if we have something other than whitespace, we're done.
2324  if (!isVerticalWhitespace(Char))
2325  break;
2326 
2328  // End of preprocessor directive line, let LexTokenInternal handle this.
2329  BufferPtr = CurPtr;
2330  return false;
2331  }
2332 
2333  // OK, but handle newline.
2334  if (*CurPtr == '\n')
2335  setLastNewLine(CurPtr);
2336  SawNewline = true;
2337  Char = *++CurPtr;
2338  }
2339 
2340  // If the client wants us to return whitespace, return it now.
2341  if (isKeepWhitespaceMode()) {
2342  FormTokenWithChars(Result, CurPtr, tok::unknown);
2343  if (SawNewline) {
2344  IsAtStartOfLine = true;
2345  IsAtPhysicalStartOfLine = true;
2346  }
2347  // FIXME: The next token will not have LeadingSpace set.
2348  return true;
2349  }
2350 
2351  // If this isn't immediately after a newline, there is leading space.
2352  char PrevChar = CurPtr[-1];
2353  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2354 
2355  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2356  if (SawNewline) {
2357  Result.setFlag(Token::StartOfLine);
2358  TokAtPhysicalStartOfLine = true;
2359 
2360  if (NewLinePtr && lastNewLine && NewLinePtr != lastNewLine && PP) {
2361  if (auto *Handler = PP->getEmptylineHandler())
2362  Handler->HandleEmptyline(SourceRange(getSourceLocation(NewLinePtr + 1),
2363  getSourceLocation(lastNewLine)));
2364  }
2365  }
2366 
2367  BufferPtr = CurPtr;
2368  return false;
2369 }
2370 
2371 /// We have just read the // characters from input. Skip until we find the
2372 /// newline character that terminates the comment. Then update BufferPtr and
2373 /// return.
2374 ///
2375 /// If we're in KeepCommentMode or any CommentHandler has inserted
2376 /// some tokens, this will store the first token and return true.
2377 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2378  bool &TokAtPhysicalStartOfLine) {
2379  // If Line comments aren't explicitly enabled for this language, emit an
2380  // extension warning.
2381  if (!LineComment) {
2382  if (!isLexingRawMode()) // There's no PP in raw mode, so can't emit diags.
2383  Diag(BufferPtr, diag::ext_line_comment);
2384 
2385  // Mark them enabled so we only emit one warning for this translation
2386  // unit.
2387  LineComment = true;
2388  }
2389 
2390  // Scan over the body of the comment. The common case, when scanning, is that
2391  // the comment contains normal ascii characters with nothing interesting in
2392  // them. As such, optimize for this case with the inner loop.
2393  //
2394  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2395  // character that ends the line comment.
2396  char C;
2397  while (true) {
2398  C = *CurPtr;
2399  // Skip over characters in the fast loop.
2400  while (C != 0 && // Potentially EOF.
2401  C != '\n' && C != '\r') // Newline or DOS-style newline.
2402  C = *++CurPtr;
2403 
2404  const char *NextLine = CurPtr;
2405  if (C != 0) {
2406  // We found a newline, see if it's escaped.
2407  const char *EscapePtr = CurPtr-1;
2408  bool HasSpace = false;
2409  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2410  --EscapePtr;
2411  HasSpace = true;
2412  }
2413 
2414  if (*EscapePtr == '\\')
2415  // Escaped newline.
2416  CurPtr = EscapePtr;
2417  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2418  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2419  // Trigraph-escaped newline.
2420  CurPtr = EscapePtr-2;
2421  else
2422  break; // This is a newline, we're done.
2423 
2424  // If there was space between the backslash and newline, warn about it.
2425  if (HasSpace && !isLexingRawMode())
2426  Diag(EscapePtr, diag::backslash_newline_space);
2427  }
2428 
2429  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2430  // properly decode the character. Read it in raw mode to avoid emitting
2431  // diagnostics about things like trigraphs. If we see an escaped newline,
2432  // we'll handle it below.
2433  const char *OldPtr = CurPtr;
2434  bool OldRawMode = isLexingRawMode();
2435  LexingRawMode = true;
2436  C = getAndAdvanceChar(CurPtr, Result);
2437  LexingRawMode = OldRawMode;
2438 
2439  // If we only read only one character, then no special handling is needed.
2440  // We're done and can skip forward to the newline.
2441  if (C != 0 && CurPtr == OldPtr+1) {
2442  CurPtr = NextLine;
2443  break;
2444  }
2445 
2446  // If we read multiple characters, and one of those characters was a \r or
2447  // \n, then we had an escaped newline within the comment. Emit diagnostic
2448  // unless the next line is also a // comment.
2449  if (CurPtr != OldPtr + 1 && C != '/' &&
2450  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2451  for (; OldPtr != CurPtr; ++OldPtr)
2452  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2453  // Okay, we found a // comment that ends in a newline, if the next
2454  // line is also a // comment, but has spaces, don't emit a diagnostic.
2455  if (isWhitespace(C)) {
2456  const char *ForwardPtr = CurPtr;
2457  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2458  ++ForwardPtr;
2459  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2460  break;
2461  }
2462 
2463  if (!isLexingRawMode())
2464  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2465  break;
2466  }
2467  }
2468 
2469  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2470  --CurPtr;
2471  break;
2472  }
2473 
2474  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2476  cutOffLexing();
2477  return false;
2478  }
2479  }
2480 
2481  // Found but did not consume the newline. Notify comment handlers about the
2482  // comment unless we're in a #if 0 block.
2483  if (PP && !isLexingRawMode() &&
2484  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2485  getSourceLocation(CurPtr)))) {
2486  BufferPtr = CurPtr;
2487  return true; // A token has to be returned.
2488  }
2489 
2490  // If we are returning comments as tokens, return this comment as a token.
2491  if (inKeepCommentMode())
2492  return SaveLineComment(Result, CurPtr);
2493 
2494  // If we are inside a preprocessor directive and we see the end of line,
2495  // return immediately, so that the lexer can return this as an EOD token.
2496  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2497  BufferPtr = CurPtr;
2498  return false;
2499  }
2500 
2501  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2502  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2503  // contribute to another token), it isn't needed for correctness. Note that
2504  // this is ok even in KeepWhitespaceMode, because we would have returned the
2505  /// comment above in that mode.
2506  NewLinePtr = CurPtr++;
2507 
2508  // The next returned token is at the start of the line.
2509  Result.setFlag(Token::StartOfLine);
2510  TokAtPhysicalStartOfLine = true;
2511  // No leading whitespace seen so far.
2512  Result.clearFlag(Token::LeadingSpace);
2513  BufferPtr = CurPtr;
2514  return false;
2515 }
2516 
2517 /// If in save-comment mode, package up this Line comment in an appropriate
2518 /// way and return it.
2519 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2520  // If we're not in a preprocessor directive, just return the // comment
2521  // directly.
2522  FormTokenWithChars(Result, CurPtr, tok::comment);
2523 
2525  return true;
2526 
2527  // If this Line-style comment is in a macro definition, transmogrify it into
2528  // a C-style block comment.
2529  bool Invalid = false;
2530  std::string Spelling = PP->getSpelling(Result, &Invalid);
2531  if (Invalid)
2532  return true;
2533 
2534  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2535  Spelling[1] = '*'; // Change prefix to "/*".
2536  Spelling += "*/"; // add suffix.
2537 
2538  Result.setKind(tok::comment);
2539  PP->CreateString(Spelling, Result,
2540  Result.getLocation(), Result.getLocation());
2541  return true;
2542 }
2543 
2544 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2545 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2546 /// a diagnostic if so. We know that the newline is inside of a block comment.
2547 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L,
2548  bool Trigraphs) {
2549  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2550 
2551  // Position of the first trigraph in the ending sequence.
2552  const char *TrigraphPos = nullptr;
2553  // Position of the first whitespace after a '\' in the ending sequence.
2554  const char *SpacePos = nullptr;
2555 
2556  while (true) {
2557  // Back up off the newline.
2558  --CurPtr;
2559 
2560  // If this is a two-character newline sequence, skip the other character.
2561  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2562  // \n\n or \r\r -> not escaped newline.
2563  if (CurPtr[0] == CurPtr[1])
2564  return false;
2565  // \n\r or \r\n -> skip the newline.
2566  --CurPtr;
2567  }
2568 
2569  // If we have horizontal whitespace, skip over it. We allow whitespace
2570  // between the slash and newline.
2571  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2572  SpacePos = CurPtr;
2573  --CurPtr;
2574  }
2575 
2576  // If we have a slash, this is an escaped newline.
2577  if (*CurPtr == '\\') {
2578  --CurPtr;
2579  } else if (CurPtr[0] == '/' && CurPtr[-1] == '?' && CurPtr[-2] == '?') {
2580  // This is a trigraph encoding of a slash.
2581  TrigraphPos = CurPtr - 2;
2582  CurPtr -= 3;
2583  } else {
2584  return false;
2585  }
2586 
2587  // If the character preceding the escaped newline is a '*', then after line
2588  // splicing we have a '*/' ending the comment.
2589  if (*CurPtr == '*')
2590  break;
2591 
2592  if (*CurPtr != '\n' && *CurPtr != '\r')
2593  return false;
2594  }
2595 
2596  if (TrigraphPos) {
2597  // If no trigraphs are enabled, warn that we ignored this trigraph and
2598  // ignore this * character.
2599  if (!Trigraphs) {
2600  if (!L->isLexingRawMode())
2601  L->Diag(TrigraphPos, diag::trigraph_ignored_block_comment);
2602  return false;
2603  }
2604  if (!L->isLexingRawMode())
2605  L->Diag(TrigraphPos, diag::trigraph_ends_block_comment);
2606  }
2607 
2608  // Warn about having an escaped newline between the */ characters.
2609  if (!L->isLexingRawMode())
2610  L->Diag(CurPtr + 1, diag::escaped_newline_block_comment_end);
2611 
2612  // If there was space between the backslash and newline, warn about it.
2613  if (SpacePos && !L->isLexingRawMode())
2614  L->Diag(SpacePos, diag::backslash_newline_space);
2615 
2616  return true;
2617 }
2618 
2619 #ifdef __SSE2__
2620 #include <emmintrin.h>
2621 #elif __ALTIVEC__
2622 #include <altivec.h>
2623 #undef bool
2624 #endif
2625 
2626 /// We have just read from input the / and * characters that started a comment.
2627 /// Read until we find the * and / characters that terminate the comment.
2628 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2629 /// comments, because they cannot cause the comment to end. The only thing
2630 /// that can happen is the comment could end with an escaped newline between
2631 /// the terminating * and /.
2632 ///
2633 /// If we're in KeepCommentMode or any CommentHandler has inserted
2634 /// some tokens, this will store the first token and return true.
2635 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2636  bool &TokAtPhysicalStartOfLine) {
2637  // Scan one character past where we should, looking for a '/' character. Once
2638  // we find it, check to see if it was preceded by a *. This common
2639  // optimization helps people who like to put a lot of * characters in their
2640  // comments.
2641 
2642  // The first character we get with newlines and trigraphs skipped to handle
2643  // the degenerate /*/ case below correctly if the * has an escaped newline
2644  // after it.
2645  unsigned CharSize;
2646  unsigned char C = getCharAndSize(CurPtr, CharSize);
2647  CurPtr += CharSize;
2648  if (C == 0 && CurPtr == BufferEnd+1) {
2649  if (!isLexingRawMode())
2650  Diag(BufferPtr, diag::err_unterminated_block_comment);
2651  --CurPtr;
2652 
2653  // KeepWhitespaceMode should return this broken comment as a token. Since
2654  // it isn't a well formed comment, just return it as an 'unknown' token.
2655  if (isKeepWhitespaceMode()) {
2656  FormTokenWithChars(Result, CurPtr, tok::unknown);
2657  return true;
2658  }
2659 
2660  BufferPtr = CurPtr;
2661  return false;
2662  }
2663 
2664  // Check to see if the first character after the '/*' is another /. If so,
2665  // then this slash does not end the block comment, it is part of it.
2666  if (C == '/')
2667  C = *CurPtr++;
2668 
2669  while (true) {
2670  // Skip over all non-interesting characters until we find end of buffer or a
2671  // (probably ending) '/' character.
2672  if (CurPtr + 24 < BufferEnd &&
2673  // If there is a code-completion point avoid the fast scan because it
2674  // doesn't check for '\0'.
2675  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2676  // While not aligned to a 16-byte boundary.
2677  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2678  C = *CurPtr++;
2679 
2680  if (C == '/') goto FoundSlash;
2681 
2682 #ifdef __SSE2__
2683  __m128i Slashes = _mm_set1_epi8('/');
2684  while (CurPtr+16 <= BufferEnd) {
2685  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2686  Slashes));
2687  if (cmp != 0) {
2688  // Adjust the pointer to point directly after the first slash. It's
2689  // not necessary to set C here, it will be overwritten at the end of
2690  // the outer loop.
2691  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2692  goto FoundSlash;
2693  }
2694  CurPtr += 16;
2695  }
2696 #elif __ALTIVEC__
2697  __vector unsigned char Slashes = {
2698  '/', '/', '/', '/', '/', '/', '/', '/',
2699  '/', '/', '/', '/', '/', '/', '/', '/'
2700  };
2701  while (CurPtr + 16 <= BufferEnd &&
2702  !vec_any_eq(*(const __vector unsigned char *)CurPtr, Slashes))
2703  CurPtr += 16;
2704 #else
2705  // Scan for '/' quickly. Many block comments are very large.
2706  while (CurPtr[0] != '/' &&
2707  CurPtr[1] != '/' &&
2708  CurPtr[2] != '/' &&
2709  CurPtr[3] != '/' &&
2710  CurPtr+4 < BufferEnd) {
2711  CurPtr += 4;
2712  }
2713 #endif
2714 
2715  // It has to be one of the bytes scanned, increment to it and read one.
2716  C = *CurPtr++;
2717  }
2718 
2719  // Loop to scan the remainder.
2720  while (C != '/' && C != '\0')
2721  C = *CurPtr++;
2722 
2723  if (C == '/') {
2724  FoundSlash:
2725  if (CurPtr[-2] == '*') // We found the final */. We're done!
2726  break;
2727 
2728  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2729  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr - 2, this,
2730  LangOpts.Trigraphs)) {
2731  // We found the final */, though it had an escaped newline between the
2732  // * and /. We're done!
2733  break;
2734  }
2735  }
2736  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2737  // If this is a /* inside of the comment, emit a warning. Don't do this
2738  // if this is a /*/, which will end the comment. This misses cases with
2739  // embedded escaped newlines, but oh well.
2740  if (!isLexingRawMode())
2741  Diag(CurPtr-1, diag::warn_nested_block_comment);
2742  }
2743  } else if (C == 0 && CurPtr == BufferEnd+1) {
2744  if (!isLexingRawMode())
2745  Diag(BufferPtr, diag::err_unterminated_block_comment);
2746  // Note: the user probably forgot a */. We could continue immediately
2747  // after the /*, but this would involve lexing a lot of what really is the
2748  // comment, which surely would confuse the parser.
2749  --CurPtr;
2750 
2751  // KeepWhitespaceMode should return this broken comment as a token. Since
2752  // it isn't a well formed comment, just return it as an 'unknown' token.
2753  if (isKeepWhitespaceMode()) {
2754  FormTokenWithChars(Result, CurPtr, tok::unknown);
2755  return true;
2756  }
2757 
2758  BufferPtr = CurPtr;
2759  return false;
2760  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2762  cutOffLexing();
2763  return false;
2764  }
2765 
2766  C = *CurPtr++;
2767  }
2768 
2769  // Notify comment handlers about the comment unless we're in a #if 0 block.
2770  if (PP && !isLexingRawMode() &&
2771  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2772  getSourceLocation(CurPtr)))) {
2773  BufferPtr = CurPtr;
2774  return true; // A token has to be returned.
2775  }
2776 
2777  // If we are returning comments as tokens, return this comment as a token.
2778  if (inKeepCommentMode()) {
2779  FormTokenWithChars(Result, CurPtr, tok::comment);
2780  return true;
2781  }
2782 
2783  // It is common for the tokens immediately after a /**/ comment to be
2784  // whitespace. Instead of going through the big switch, handle it
2785  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2786  // have already returned above with the comment as a token.
2787  if (isHorizontalWhitespace(*CurPtr)) {
2788  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2789  return false;
2790  }
2791 
2792  // Otherwise, just return so that the next character will be lexed as a token.
2793  BufferPtr = CurPtr;
2794  Result.setFlag(Token::LeadingSpace);
2795  return false;
2796 }
2797 
2798 //===----------------------------------------------------------------------===//
2799 // Primary Lexing Entry Points
2800 //===----------------------------------------------------------------------===//
2801 
2802 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2803 /// uninterpreted string. This switches the lexer out of directive mode.
2805  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2806  "Must be in a preprocessing directive!");
2807  Token Tmp;
2808  Tmp.startToken();
2809 
2810  // CurPtr - Cache BufferPtr in an automatic variable.
2811  const char *CurPtr = BufferPtr;
2812  while (true) {
2813  char Char = getAndAdvanceChar(CurPtr, Tmp);
2814  switch (Char) {
2815  default:
2816  if (Result)
2817  Result->push_back(Char);
2818  break;
2819  case 0: // Null.
2820  // Found end of file?
2821  if (CurPtr-1 != BufferEnd) {
2822  if (isCodeCompletionPoint(CurPtr-1)) {
2824  cutOffLexing();
2825  return;
2826  }
2827 
2828  // Nope, normal character, continue.
2829  if (Result)
2830  Result->push_back(Char);
2831  break;
2832  }
2833  // FALL THROUGH.
2834  LLVM_FALLTHROUGH;
2835  case '\r':
2836  case '\n':
2837  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2838  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2839  BufferPtr = CurPtr-1;
2840 
2841  // Next, lex the character, which should handle the EOD transition.
2842  Lex(Tmp);
2843  if (Tmp.is(tok::code_completion)) {
2844  if (PP)
2846  Lex(Tmp);
2847  }
2848  assert(Tmp.is(tok::eod) && "Unexpected token!");
2849 
2850  // Finally, we're done;
2851  return;
2852  }
2853  }
2854 }
2855 
2856 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2857 /// condition, reporting diagnostics and handling other edge cases as required.
2858 /// This returns true if Result contains a token, false if PP.Lex should be
2859 /// called again.
2860 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2861  // If we hit the end of the file while parsing a preprocessor directive,
2862  // end the preprocessor directive first. The next token returned will
2863  // then be the end of file.
2865  // Done parsing the "line".
2867  // Update the location of token as well as BufferPtr.
2868  FormTokenWithChars(Result, CurPtr, tok::eod);
2869 
2870  // Restore comment saving mode, in case it was disabled for directive.
2871  if (PP)
2873  return true; // Have a token.
2874  }
2875 
2876  // If we are in raw mode, return this event as an EOF token. Let the caller
2877  // that put us in raw mode handle the event.
2878  if (isLexingRawMode()) {
2879  Result.startToken();
2880  BufferPtr = BufferEnd;
2881  FormTokenWithChars(Result, BufferEnd, tok::eof);
2882  return true;
2883  }
2884 
2885  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2887  // If the preamble cuts off the end of a header guard, consider it guarded.
2888  // The guard is valid for the preamble content itself, and for tools the
2889  // most useful answer is "yes, this file has a header guard".
2890  if (!ConditionalStack.empty())
2892  ConditionalStack.clear();
2893  }
2894 
2895  // Issue diagnostics for unterminated #if and missing newline.
2896 
2897  // If we are in a #if directive, emit an error.
2898  while (!ConditionalStack.empty()) {
2899  if (PP->getCodeCompletionFileLoc() != FileLoc)
2900  PP->Diag(ConditionalStack.back().IfLoc,
2901  diag::err_pp_unterminated_conditional);
2902  ConditionalStack.pop_back();
2903  }
2904 
2905  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2906  // a pedwarn.
2907  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2908  DiagnosticsEngine &Diags = PP->getDiagnostics();
2909  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2910  unsigned DiagID;
2911 
2912  if (LangOpts.CPlusPlus11) {
2913  // C++11 [lex.phases] 2.2 p2
2914  // Prefer the C++98 pedantic compatibility warning over the generic,
2915  // non-extension, user-requested "missing newline at EOF" warning.
2916  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2917  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2918  } else {
2919  DiagID = diag::warn_no_newline_eof;
2920  }
2921  } else {
2922  DiagID = diag::ext_no_newline_eof;
2923  }
2924 
2925  Diag(BufferEnd, DiagID)
2926  << FixItHint::CreateInsertion(EndLoc, "\n");
2927  }
2928 
2929  BufferPtr = CurPtr;
2930 
2931  // Finally, let the preprocessor handle this.
2932  return PP->HandleEndOfFile(Result, isPragmaLexer());
2933 }
2934 
2935 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2936 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2937 /// else and 2 if there are no more tokens in the buffer controlled by the
2938 /// lexer.
2939 unsigned Lexer::isNextPPTokenLParen() {
2940  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2941 
2942  // Switch to 'skipping' mode. This will ensure that we can lex a token
2943  // without emitting diagnostics, disables macro expansion, and will cause EOF
2944  // to return an EOF token instead of popping the include stack.
2945  LexingRawMode = true;
2946 
2947  // Save state that can be changed while lexing so that we can restore it.
2948  const char *TmpBufferPtr = BufferPtr;
2949  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2950  bool atStartOfLine = IsAtStartOfLine;
2951  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2952  bool leadingSpace = HasLeadingSpace;
2953 
2954  Token Tok;
2955  Lex(Tok);
2956 
2957  // Restore state that may have changed.
2958  BufferPtr = TmpBufferPtr;
2959  ParsingPreprocessorDirective = inPPDirectiveMode;
2960  HasLeadingSpace = leadingSpace;
2961  IsAtStartOfLine = atStartOfLine;
2962  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2963 
2964  // Restore the lexer back to non-skipping mode.
2965  LexingRawMode = false;
2966 
2967  if (Tok.is(tok::eof))
2968  return 2;
2969  return Tok.is(tok::l_paren);
2970 }
2971 
2972 /// Find the end of a version control conflict marker.
2973 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2974  ConflictMarkerKind CMK) {
2975  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2976  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2977  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2978  size_t Pos = RestOfBuffer.find(Terminator);
2979  while (Pos != StringRef::npos) {
2980  // Must occur at start of line.
2981  if (Pos == 0 ||
2982  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2983  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2984  Pos = RestOfBuffer.find(Terminator);
2985  continue;
2986  }
2987  return RestOfBuffer.data()+Pos;
2988  }
2989  return nullptr;
2990 }
2991 
2992 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2993 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2994 /// and recover nicely. This returns true if it is a conflict marker and false
2995 /// if not.
2996 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2997  // Only a conflict marker if it starts at the beginning of a line.
2998  if (CurPtr != BufferStart &&
2999  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
3000  return false;
3001 
3002  // Check to see if we have <<<<<<< or >>>>.
3003  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
3004  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
3005  return false;
3006 
3007  // If we have a situation where we don't care about conflict markers, ignore
3008  // it.
3009  if (CurrentConflictMarkerState || isLexingRawMode())
3010  return false;
3011 
3012  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
3013 
3014  // Check to see if there is an ending marker somewhere in the buffer at the
3015  // start of a line to terminate this conflict marker.
3016  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
3017  // We found a match. We are really in a conflict marker.
3018  // Diagnose this, and ignore to the end of line.
3019  Diag(CurPtr, diag::err_conflict_marker);
3020  CurrentConflictMarkerState = Kind;
3021 
3022  // Skip ahead to the end of line. We know this exists because the
3023  // end-of-conflict marker starts with \r or \n.
3024  while (*CurPtr != '\r' && *CurPtr != '\n') {
3025  assert(CurPtr != BufferEnd && "Didn't find end of line");
3026  ++CurPtr;
3027  }
3028  BufferPtr = CurPtr;
3029  return true;
3030  }
3031 
3032  // No end of conflict marker found.
3033  return false;
3034 }
3035 
3036 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
3037 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
3038 /// is the end of a conflict marker. Handle it by ignoring up until the end of
3039 /// the line. This returns true if it is a conflict marker and false if not.
3040 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
3041  // Only a conflict marker if it starts at the beginning of a line.
3042  if (CurPtr != BufferStart &&
3043  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
3044  return false;
3045 
3046  // If we have a situation where we don't care about conflict markers, ignore
3047  // it.
3048  if (!CurrentConflictMarkerState || isLexingRawMode())
3049  return false;
3050 
3051  // Check to see if we have the marker (4 characters in a row).
3052  for (unsigned i = 1; i != 4; ++i)
3053  if (CurPtr[i] != CurPtr[0])
3054  return false;
3055 
3056  // If we do have it, search for the end of the conflict marker. This could
3057  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
3058  // be the end of conflict marker.
3059  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
3060  CurrentConflictMarkerState)) {
3061  CurPtr = End;
3062 
3063  // Skip ahead to the end of line.
3064  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
3065  ++CurPtr;
3066 
3067  BufferPtr = CurPtr;
3068 
3069  // No longer in the conflict marker.
3070  CurrentConflictMarkerState = CMK_None;
3071  return true;
3072  }
3073 
3074  return false;
3075 }
3076 
3077 static const char *findPlaceholderEnd(const char *CurPtr,
3078  const char *BufferEnd) {
3079  if (CurPtr == BufferEnd)
3080  return nullptr;
3081  BufferEnd -= 1; // Scan until the second last character.
3082  for (; CurPtr != BufferEnd; ++CurPtr) {
3083  if (CurPtr[0] == '#' && CurPtr[1] == '>')
3084  return CurPtr + 2;
3085  }
3086  return nullptr;
3087 }
3088 
3089 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
3090  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
3092  return false;
3093  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
3094  if (!End)
3095  return false;
3096  const char *Start = CurPtr - 1;
3097  if (!LangOpts.AllowEditorPlaceholders)
3098  Diag(Start, diag::err_placeholder_in_source);
3099  Result.startToken();
3100  FormTokenWithChars(Result, End, tok::raw_identifier);
3101  Result.setRawIdentifierData(Start);
3102  PP->LookUpIdentifierInfo(Result);
3103  Result.setFlag(Token::IsEditorPlaceholder);
3104  BufferPtr = End;
3105  return true;
3106 }
3107 
3108 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
3109  if (PP && PP->isCodeCompletionEnabled()) {
3110  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
3111  return Loc == PP->getCodeCompletionLoc();
3112  }
3113 
3114  return false;
3115 }
3116 
3117 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
3118  Token *Result) {
3119  unsigned CharSize;
3120  char Kind = getCharAndSize(StartPtr, CharSize);
3121  bool Delimited = false;
3122  bool FoundEndDelimiter = false;
3123  unsigned Count = 0;
3124  bool Diagnose = Result && !isLexingRawMode();
3125 
3126  unsigned NumHexDigits;
3127  if (Kind == 'u')
3128  NumHexDigits = 4;
3129  else if (Kind == 'U')
3130  NumHexDigits = 8;
3131  else
3132  return 0;
3133 
3134  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
3135  if (Diagnose)
3136  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
3137  return 0;
3138  }
3139 
3140  const char *CurPtr = StartPtr + CharSize;
3141  const char *KindLoc = &CurPtr[-1];
3142 
3143  uint32_t CodePoint = 0;
3144  while (Count != NumHexDigits || Delimited) {
3145  char C = getCharAndSize(CurPtr, CharSize);
3146  if (!Delimited && C == '{') {
3147  Delimited = true;
3148  CurPtr += CharSize;
3149  continue;
3150  }
3151 
3152  if (Delimited && C == '}') {
3153  CurPtr += CharSize;
3154  FoundEndDelimiter = true;
3155  break;
3156  }
3157 
3158  unsigned Value = llvm::hexDigitValue(C);
3159  if (Value == -1U) {
3160  if (!Delimited)
3161  break;
3162  if (Diagnose)
3163  Diag(BufferPtr, diag::warn_delimited_ucn_incomplete)
3164  << StringRef(&C, 1);
3165  return 0;
3166  }
3167 
3168  if (CodePoint & 0xF000'0000) {
3169  if (Diagnose)
3170  Diag(KindLoc, diag::err_escape_too_large) << 0;
3171  return 0;
3172  }
3173 
3174  CodePoint <<= 4;
3175  CodePoint |= Value;
3176  CurPtr += CharSize;
3177  Count++;
3178  }
3179 
3180  if (Count == 0) {
3181  if (Diagnose)
3182  Diag(StartPtr, FoundEndDelimiter ? diag::warn_delimited_ucn_empty
3183  : diag::warn_ucn_escape_no_digits)
3184  << StringRef(KindLoc, 1);
3185  return 0;
3186  }
3187 
3188  if (!Delimited && Count != NumHexDigits) {
3189  if (Diagnose) {
3190  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
3191  // If the user wrote \U1234, suggest a fixit to \u.
3192  if (Count == 4 && NumHexDigits == 8) {
3193  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
3194  Diag(KindLoc, diag::note_ucn_four_not_eight)
3195  << FixItHint::CreateReplacement(URange, "u");
3196  }
3197  }
3198  return 0;
3199  }
3200 
3201  if (Delimited && PP) {
3202  Diag(BufferPtr, diag::ext_delimited_escape_sequence);
3203  }
3204 
3205  if (Result) {
3206  Result->setFlag(Token::HasUCN);
3207  if (CurPtr - StartPtr == (ptrdiff_t)(Count + 2 + (Delimited ? 2 : 0)))
3208  StartPtr = CurPtr;
3209  else
3210  while (StartPtr != CurPtr)
3211  (void)getAndAdvanceChar(StartPtr, *Result);
3212  } else {
3213  StartPtr = CurPtr;
3214  }
3215 
3216  // Don't apply C family restrictions to UCNs in assembly mode
3217  if (LangOpts.AsmPreprocessor)
3218  return CodePoint;
3219 
3220  // C99 6.4.3p2: A universal character name shall not specify a character whose
3221  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3222  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3223  // C++11 [lex.charset]p2: If the hexadecimal value for a
3224  // universal-character-name corresponds to a surrogate code point (in the
3225  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3226  // if the hexadecimal value for a universal-character-name outside the
3227  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3228  // string literal corresponds to a control character (in either of the
3229  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3230  // basic source character set, the program is ill-formed.
3231  if (CodePoint < 0xA0) {
3232  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3233  return CodePoint;
3234 
3235  // We don't use isLexingRawMode() here because we need to warn about bad
3236  // UCNs even when skipping preprocessing tokens in a #if block.
3237  if (Result && PP) {
3238  if (CodePoint < 0x20 || CodePoint >= 0x7F)
3239  Diag(BufferPtr, diag::err_ucn_control_character);
3240  else {
3241  char C = static_cast<char>(CodePoint);
3242  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3243  }
3244  }
3245 
3246  return 0;
3247  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3248  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3249  // We don't use isLexingRawMode() here because we need to diagnose bad
3250  // UCNs even when skipping preprocessing tokens in a #if block.
3251  if (Result && PP) {
3252  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3253  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3254  else
3255  Diag(BufferPtr, diag::err_ucn_escape_invalid);
3256  }
3257  return 0;
3258  }
3259 
3260  return CodePoint;
3261 }
3262 
3263 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3264  const char *CurPtr) {
3265  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3266  isUnicodeWhitespace(C)) {
3267  Diag(BufferPtr, diag::ext_unicode_whitespace)
3268  << makeCharRange(*this, BufferPtr, CurPtr);
3269 
3270  Result.setFlag(Token::LeadingSpace);
3271  return true;
3272  }
3273  return false;
3274 }
3275 
3276 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3277  IsAtStartOfLine = Result.isAtStartOfLine();
3278  HasLeadingSpace = Result.hasLeadingSpace();
3279  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3280  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3281 }
3282 
3283 bool Lexer::Lex(Token &Result) {
3284  // Start a new token.
3285  Result.startToken();
3286 
3287  // Set up misc whitespace flags for LexTokenInternal.
3288  if (IsAtStartOfLine) {
3289  Result.setFlag(Token::StartOfLine);
3290  IsAtStartOfLine = false;
3291  }
3292 
3293  if (HasLeadingSpace) {
3294  Result.setFlag(Token::LeadingSpace);
3295  HasLeadingSpace = false;
3296  }
3297 
3298  if (HasLeadingEmptyMacro) {
3299  Result.setFlag(Token::LeadingEmptyMacro);
3300  HasLeadingEmptyMacro = false;
3301  }
3302 
3303  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3304  IsAtPhysicalStartOfLine = false;
3305  bool isRawLex = isLexingRawMode();
3306  (void) isRawLex;
3307  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3308  // (After the LexTokenInternal call, the lexer might be destroyed.)
3309  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3310  return returnedToken;
3311 }
3312 
3313 /// LexTokenInternal - This implements a simple C family lexer. It is an
3314 /// extremely performance critical piece of code. This assumes that the buffer
3315 /// has a null character at the end of the file. This returns a preprocessing
3316 /// token, not a normal token, as such, it is an internal interface. It assumes
3317 /// that the Flags of result have been cleared before calling this.
3318 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3319 LexNextToken:
3320  // New token, can't need cleaning yet.
3321  Result.clearFlag(Token::NeedsCleaning);
3322  Result.setIdentifierInfo(nullptr);
3323 
3324  // CurPtr - Cache BufferPtr in an automatic variable.
3325  const char *CurPtr = BufferPtr;
3326 
3327  // Small amounts of horizontal whitespace is very common between tokens.
3328  if (isHorizontalWhitespace(*CurPtr)) {
3329  do {
3330  ++CurPtr;
3331  } while (isHorizontalWhitespace(*CurPtr));
3332 
3333  // If we are keeping whitespace and other tokens, just return what we just
3334  // skipped. The next lexer invocation will return the token after the
3335  // whitespace.
3336  if (isKeepWhitespaceMode()) {
3337  FormTokenWithChars(Result, CurPtr, tok::unknown);
3338  // FIXME: The next token will not have LeadingSpace set.
3339  return true;
3340  }
3341 
3342  BufferPtr = CurPtr;
3343  Result.setFlag(Token::LeadingSpace);
3344  }
3345 
3346  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3347 
3348  // Read a character, advancing over it.
3349  char Char = getAndAdvanceChar(CurPtr, Result);
3351 
3352  if (!isVerticalWhitespace(Char))
3353  NewLinePtr = nullptr;
3354 
3355  switch (Char) {
3356  case 0: // Null.
3357  // Found end of file?
3358  if (CurPtr-1 == BufferEnd)
3359  return LexEndOfFile(Result, CurPtr-1);
3360 
3361  // Check if we are performing code completion.
3362  if (isCodeCompletionPoint(CurPtr-1)) {
3363  // Return the code-completion token.
3364  Result.startToken();
3365  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3366  return true;
3367  }
3368 
3369  if (!isLexingRawMode())
3370  Diag(CurPtr-1, diag::null_in_file);
3371  Result.setFlag(Token::LeadingSpace);
3372  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3373  return true; // KeepWhitespaceMode
3374 
3375  // We know the lexer hasn't changed, so just try again with this lexer.
3376  // (We manually eliminate the tail call to avoid recursion.)
3377  goto LexNextToken;
3378 
3379  case 26: // DOS & CP/M EOF: "^Z".
3380  // If we're in Microsoft extensions mode, treat this as end of file.
3381  if (LangOpts.MicrosoftExt) {
3382  if (!isLexingRawMode())
3383  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3384  return LexEndOfFile(Result, CurPtr-1);
3385  }
3386 
3387  // If Microsoft extensions are disabled, this is just random garbage.
3388  Kind = tok::unknown;
3389  break;
3390 
3391  case '\r':
3392  if (CurPtr[0] == '\n')
3393  (void)getAndAdvanceChar(CurPtr, Result);
3394  LLVM_FALLTHROUGH;
3395  case '\n':
3396  // If we are inside a preprocessor directive and we see the end of line,
3397  // we know we are done with the directive, so return an EOD token.
3399  // Done parsing the "line".
3401 
3402  // Restore comment saving mode, in case it was disabled for directive.
3403  if (PP)
3405 
3406  // Since we consumed a newline, we are back at the start of a line.
3407  IsAtStartOfLine = true;
3408  IsAtPhysicalStartOfLine = true;
3409  NewLinePtr = CurPtr - 1;
3410 
3411  Kind = tok::eod;
3412  break;
3413  }
3414 
3415  // No leading whitespace seen so far.
3416  Result.clearFlag(Token::LeadingSpace);
3417 
3418  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3419  return true; // KeepWhitespaceMode
3420 
3421  // We only saw whitespace, so just try again with this lexer.
3422  // (We manually eliminate the tail call to avoid recursion.)
3423  goto LexNextToken;
3424  case ' ':
3425  case '\t':
3426  case '\f':
3427  case '\v':
3428  SkipHorizontalWhitespace:
3429  Result.setFlag(Token::LeadingSpace);
3430  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3431  return true; // KeepWhitespaceMode
3432 
3433  SkipIgnoredUnits:
3434  CurPtr = BufferPtr;
3435 
3436  // If the next token is obviously a // or /* */ comment, skip it efficiently
3437  // too (without going through the big switch stmt).
3438  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3439  LineComment && (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3440  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3441  return true; // There is a token to return.
3442  goto SkipIgnoredUnits;
3443  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3444  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3445  return true; // There is a token to return.
3446  goto SkipIgnoredUnits;
3447  } else if (isHorizontalWhitespace(*CurPtr)) {
3448  goto SkipHorizontalWhitespace;
3449  }
3450  // We only saw whitespace, so just try again with this lexer.
3451  // (We manually eliminate the tail call to avoid recursion.)
3452  goto LexNextToken;
3453 
3454  // C99 6.4.4.1: Integer Constants.
3455  // C99 6.4.4.2: Floating Constants.
3456  case '0': case '1': case '2': case '3': case '4':
3457  case '5': case '6': case '7': case '8': case '9':
3458  // Notify MIOpt that we read a non-whitespace/non-comment token.
3459  MIOpt.ReadToken();
3460  return LexNumericConstant(Result, CurPtr);
3461 
3462  // Identifier (e.g., uber), or
3463  // UTF-8 (C2x/C++17) or UTF-16 (C11/C++11) character literal, or
3464  // UTF-8 or UTF-16 string literal (C11/C++11).
3465  case 'u':
3466  // Notify MIOpt that we read a non-whitespace/non-comment token.
3467  MIOpt.ReadToken();
3468 
3469  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3470  Char = getCharAndSize(CurPtr, SizeTmp);
3471 
3472  // UTF-16 string literal
3473  if (Char == '"')
3474  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3475  tok::utf16_string_literal);
3476 
3477  // UTF-16 character constant
3478  if (Char == '\'')
3479  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3480  tok::utf16_char_constant);
3481 
3482  // UTF-16 raw string literal
3483  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3484  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3485  return LexRawStringLiteral(Result,
3486  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3487  SizeTmp2, Result),
3488  tok::utf16_string_literal);
3489 
3490  if (Char == '8') {
3491  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3492 
3493  // UTF-8 string literal
3494  if (Char2 == '"')
3495  return LexStringLiteral(Result,
3496  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3497  SizeTmp2, Result),
3498  tok::utf8_string_literal);
3499  if (Char2 == '\'' && (LangOpts.CPlusPlus17 || LangOpts.C2x))
3500  return LexCharConstant(
3501  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3502  SizeTmp2, Result),
3503  tok::utf8_char_constant);
3504 
3505  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3506  unsigned SizeTmp3;
3507  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3508  // UTF-8 raw string literal
3509  if (Char3 == '"') {
3510  return LexRawStringLiteral(Result,
3511  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3512  SizeTmp2, Result),
3513  SizeTmp3, Result),
3514  tok::utf8_string_literal);
3515  }
3516  }
3517  }
3518  }
3519 
3520  // treat u like the start of an identifier.
3521  return LexIdentifierContinue(Result, CurPtr);
3522 
3523  case 'U': // Identifier (e.g. Uber) or C11/C++11 UTF-32 string literal
3524  // Notify MIOpt that we read a non-whitespace/non-comment token.
3525  MIOpt.ReadToken();
3526 
3527  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3528  Char = getCharAndSize(CurPtr, SizeTmp);
3529 
3530  // UTF-32 string literal
3531  if (Char == '"')
3532  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3533  tok::utf32_string_literal);
3534 
3535  // UTF-32 character constant
3536  if (Char == '\'')
3537  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3538  tok::utf32_char_constant);
3539 
3540  // UTF-32 raw string literal
3541  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3542  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3543  return LexRawStringLiteral(Result,
3544  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3545  SizeTmp2, Result),
3546  tok::utf32_string_literal);
3547  }
3548 
3549  // treat U like the start of an identifier.
3550  return LexIdentifierContinue(Result, CurPtr);
3551 
3552  case 'R': // Identifier or C++0x raw string literal
3553  // Notify MIOpt that we read a non-whitespace/non-comment token.
3554  MIOpt.ReadToken();
3555 
3556  if (LangOpts.CPlusPlus11) {
3557  Char = getCharAndSize(CurPtr, SizeTmp);
3558 
3559  if (Char == '"')
3560  return LexRawStringLiteral(Result,
3561  ConsumeChar(CurPtr, SizeTmp, Result),
3562  tok::string_literal);
3563  }
3564 
3565  // treat R like the start of an identifier.
3566  return LexIdentifierContinue(Result, CurPtr);
3567 
3568  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3569  // Notify MIOpt that we read a non-whitespace/non-comment token.
3570  MIOpt.ReadToken();
3571  Char = getCharAndSize(CurPtr, SizeTmp);
3572 
3573  // Wide string literal.
3574  if (Char == '"')
3575  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3576  tok::wide_string_literal);
3577 
3578  // Wide raw string literal.
3579  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3580  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3581  return LexRawStringLiteral(Result,
3582  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3583  SizeTmp2, Result),
3584  tok::wide_string_literal);
3585 
3586  // Wide character constant.
3587  if (Char == '\'')
3588  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3589  tok::wide_char_constant);
3590  // FALL THROUGH, treating L like the start of an identifier.
3591  LLVM_FALLTHROUGH;
3592 
3593  // C99 6.4.2: Identifiers.
3594  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3595  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3596  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3597  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3598  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3599  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3600  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3601  case 'v': case 'w': case 'x': case 'y': case 'z':
3602  case '_':
3603  // Notify MIOpt that we read a non-whitespace/non-comment token.
3604  MIOpt.ReadToken();
3605  return LexIdentifierContinue(Result, CurPtr);
3606 
3607  case '$': // $ in identifiers.
3608  if (LangOpts.DollarIdents) {
3609  if (!isLexingRawMode())
3610  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3611  // Notify MIOpt that we read a non-whitespace/non-comment token.
3612  MIOpt.ReadToken();
3613  return LexIdentifierContinue(Result, CurPtr);
3614  }
3615 
3616  Kind = tok::unknown;
3617  break;
3618 
3619  // C99 6.4.4: Character Constants.
3620  case '\'':
3621  // Notify MIOpt that we read a non-whitespace/non-comment token.
3622  MIOpt.ReadToken();
3623  return LexCharConstant(Result, CurPtr, tok::char_constant);
3624 
3625  // C99 6.4.5: String Literals.
3626  case '"':
3627  // Notify MIOpt that we read a non-whitespace/non-comment token.
3628  MIOpt.ReadToken();
3629  return LexStringLiteral(Result, CurPtr,
3630  ParsingFilename ? tok::header_name
3631  : tok::string_literal);
3632 
3633  // C99 6.4.6: Punctuators.
3634  case '?':
3635  Kind = tok::question;
3636  break;
3637  case '[':
3638  Kind = tok::l_square;
3639  break;
3640  case ']':
3641  Kind = tok::r_square;
3642  break;
3643  case '(':
3644  Kind = tok::l_paren;
3645  break;
3646  case ')':
3647  Kind = tok::r_paren;
3648  break;
3649  case '{':
3650  Kind = tok::l_brace;
3651  break;
3652  case '}':
3653  Kind = tok::r_brace;
3654  break;
3655  case '.':
3656  Char = getCharAndSize(CurPtr, SizeTmp);
3657  if (Char >= '0' && Char <= '9') {
3658  // Notify MIOpt that we read a non-whitespace/non-comment token.
3659  MIOpt.ReadToken();
3660 
3661  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3662  } else if (LangOpts.CPlusPlus && Char == '*') {
3663  Kind = tok::periodstar;
3664  CurPtr += SizeTmp;
3665  } else if (Char == '.' &&
3666  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3667  Kind = tok::ellipsis;
3668  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3669  SizeTmp2, Result);
3670  } else {
3671  Kind = tok::period;
3672  }
3673  break;
3674  case '&':
3675  Char = getCharAndSize(CurPtr, SizeTmp);
3676  if (Char == '&') {
3677  Kind = tok::ampamp;
3678  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3679  } else if (Char == '=') {
3680  Kind = tok::ampequal;
3681  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3682  } else {
3683  Kind = tok::amp;
3684  }
3685  break;
3686  case '*':
3687  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3688  Kind = tok::starequal;
3689  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3690  } else {
3691  Kind = tok::star;
3692  }
3693  break;
3694  case '+':
3695  Char = getCharAndSize(CurPtr, SizeTmp);
3696  if (Char == '+') {
3697  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3698  Kind = tok::plusplus;
3699  } else if (Char == '=') {
3700  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3701  Kind = tok::plusequal;
3702  } else {
3703  Kind = tok::plus;
3704  }
3705  break;
3706  case '-':
3707  Char = getCharAndSize(CurPtr, SizeTmp);
3708  if (Char == '-') { // --
3709  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3710  Kind = tok::minusminus;
3711  } else if (Char == '>' && LangOpts.CPlusPlus &&
3712  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3713  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3714  SizeTmp2, Result);
3715  Kind = tok::arrowstar;
3716  } else if (Char == '>') { // ->
3717  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3718  Kind = tok::arrow;
3719  } else if (Char == '=') { // -=
3720  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3721  Kind = tok::minusequal;
3722  } else {
3723  Kind = tok::minus;
3724  }
3725  break;
3726  case '~':
3727  Kind = tok::tilde;
3728  break;
3729  case '!':
3730  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3731  Kind = tok::exclaimequal;
3732  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3733  } else {
3734  Kind = tok::exclaim;
3735  }
3736  break;
3737  case '/':
3738  // 6.4.9: Comments
3739  Char = getCharAndSize(CurPtr, SizeTmp);
3740  if (Char == '/') { // Line comment.
3741  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3742  // want to lex this as a comment. There is one problem with this though,
3743  // that in one particular corner case, this can change the behavior of the
3744  // resultant program. For example, In "foo //**/ bar", C89 would lex
3745  // this as "foo / bar" and languages with Line comments would lex it as
3746  // "foo". Check to see if the character after the second slash is a '*'.
3747  // If so, we will lex that as a "/" instead of the start of a comment.
3748  // However, we never do this if we are just preprocessing.
3749  bool TreatAsComment =
3750  LineComment && (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3751  if (!TreatAsComment)
3752  if (!(PP && PP->isPreprocessedOutput()))
3753  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3754 
3755  if (TreatAsComment) {
3756  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3757  TokAtPhysicalStartOfLine))
3758  return true; // There is a token to return.
3759 
3760  // It is common for the tokens immediately after a // comment to be
3761  // whitespace (indentation for the next line). Instead of going through
3762  // the big switch, handle it efficiently now.
3763  goto SkipIgnoredUnits;
3764  }
3765  }
3766 
3767  if (Char == '*') { // /**/ comment.
3768  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3769  TokAtPhysicalStartOfLine))
3770  return true; // There is a token to return.
3771 
3772  // We only saw whitespace, so just try again with this lexer.
3773  // (We manually eliminate the tail call to avoid recursion.)
3774  goto LexNextToken;
3775  }
3776 
3777  if (Char == '=') {
3778  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3779  Kind = tok::slashequal;
3780  } else {
3781  Kind = tok::slash;
3782  }
3783  break;
3784  case '%':
3785  Char = getCharAndSize(CurPtr, SizeTmp);
3786  if (Char == '=') {
3787  Kind = tok::percentequal;
3788  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3789  } else if (LangOpts.Digraphs && Char == '>') {
3790  Kind = tok::r_brace; // '%>' -> '}'
3791  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3792  } else if (LangOpts.Digraphs && Char == ':') {
3793  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3794  Char = getCharAndSize(CurPtr, SizeTmp);
3795  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3796  Kind = tok::hashhash; // '%:%:' -> '##'
3797  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3798  SizeTmp2, Result);
3799  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3800  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3801  if (!isLexingRawMode())
3802  Diag(BufferPtr, diag::ext_charize_microsoft);
3803  Kind = tok::hashat;
3804  } else { // '%:' -> '#'
3805  // We parsed a # character. If this occurs at the start of the line,
3806  // it's actually the start of a preprocessing directive. Callback to
3807  // the preprocessor to handle it.
3808  // TODO: -fpreprocessed mode??
3809  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3810  goto HandleDirective;
3811 
3812  Kind = tok::hash;
3813  }
3814  } else {
3815  Kind = tok::percent;
3816  }
3817  break;
3818  case '<':
3819  Char = getCharAndSize(CurPtr, SizeTmp);
3820  if (ParsingFilename) {
3821  return LexAngledStringLiteral(Result, CurPtr);
3822  } else if (Char == '<') {
3823  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3824  if (After == '=') {
3825  Kind = tok::lesslessequal;
3826  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3827  SizeTmp2, Result);
3828  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3829  // If this is actually a '<<<<<<<' version control conflict marker,
3830  // recognize it as such and recover nicely.
3831  goto LexNextToken;
3832  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3833  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3834  // ignore it.
3835  goto LexNextToken;
3836  } else if (LangOpts.CUDA && After == '<') {
3837  Kind = tok::lesslessless;
3838  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3839  SizeTmp2, Result);
3840  } else {
3841  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3842  Kind = tok::lessless;
3843  }
3844  } else if (Char == '=') {
3845  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3846  if (After == '>') {
3847  if (LangOpts.CPlusPlus20) {
3848  if (!isLexingRawMode())
3849  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3850  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3851  SizeTmp2, Result);
3852  Kind = tok::spaceship;
3853  break;
3854  }
3855  // Suggest adding a space between the '<=' and the '>' to avoid a
3856  // change in semantics if this turns up in C++ <=17 mode.
3857  if (LangOpts.CPlusPlus && !isLexingRawMode()) {
3858  Diag(BufferPtr, diag::warn_cxx20_compat_spaceship)
3860  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3861  }
3862  }
3863  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3864  Kind = tok::lessequal;
3865  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3866  if (LangOpts.CPlusPlus11 &&
3867  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3868  // C++0x [lex.pptoken]p3:
3869  // Otherwise, if the next three characters are <:: and the subsequent
3870  // character is neither : nor >, the < is treated as a preprocessor
3871  // token by itself and not as the first character of the alternative
3872  // token <:.
3873  unsigned SizeTmp3;
3874  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3875  if (After != ':' && After != '>') {
3876  Kind = tok::less;
3877  if (!isLexingRawMode())
3878  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3879  break;
3880  }
3881  }
3882 
3883  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3884  Kind = tok::l_square;
3885  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3886  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3887  Kind = tok::l_brace;
3888  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3889  lexEditorPlaceholder(Result, CurPtr)) {
3890  return true;
3891  } else {
3892  Kind = tok::less;
3893  }
3894  break;
3895  case '>':
3896  Char = getCharAndSize(CurPtr, SizeTmp);
3897  if (Char == '=') {
3898  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3899  Kind = tok::greaterequal;
3900  } else if (Char == '>') {
3901  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3902  if (After == '=') {
3903  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3904  SizeTmp2, Result);
3905  Kind = tok::greatergreaterequal;
3906  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3907  // If this is actually a '>>>>' conflict marker, recognize it as such
3908  // and recover nicely.
3909  goto LexNextToken;
3910  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3911  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3912  goto LexNextToken;
3913  } else if (LangOpts.CUDA && After == '>') {
3914  Kind = tok::greatergreatergreater;
3915  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3916  SizeTmp2, Result);
3917  } else {
3918  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3919  Kind = tok::greatergreater;
3920  }
3921  } else {
3922  Kind = tok::greater;
3923  }
3924  break;
3925  case '^':
3926  Char = getCharAndSize(CurPtr, SizeTmp);
3927  if (Char == '=') {
3928  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3929  Kind = tok::caretequal;
3930  } else if (LangOpts.OpenCL && Char == '^') {
3931  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3932  Kind = tok::caretcaret;
3933  } else {
3934  Kind = tok::caret;
3935  }
3936  break;
3937  case '|':
3938  Char = getCharAndSize(CurPtr, SizeTmp);
3939  if (Char == '=') {
3940  Kind = tok::pipeequal;
3941  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3942  } else if (Char == '|') {
3943  // If this is '|||||||' and we're in a conflict marker, ignore it.
3944  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3945  goto LexNextToken;
3946  Kind = tok::pipepipe;
3947  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3948  } else {
3949  Kind = tok::pipe;
3950  }
3951  break;
3952  case ':':
3953  Char = getCharAndSize(CurPtr, SizeTmp);
3954  if (LangOpts.Digraphs && Char == '>') {
3955  Kind = tok::r_square; // ':>' -> ']'
3956  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3957  } else if ((LangOpts.CPlusPlus ||
3958  LangOpts.DoubleSquareBracketAttributes) &&
3959  Char == ':') {
3960  Kind = tok::coloncolon;
3961  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3962  } else {
3963  Kind = tok::colon;
3964  }
3965  break;
3966  case ';':
3967  Kind = tok::semi;
3968  break;
3969  case '=':
3970  Char = getCharAndSize(CurPtr, SizeTmp);
3971  if (Char == '=') {
3972  // If this is '====' and we're in a conflict marker, ignore it.
3973  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3974  goto LexNextToken;
3975 
3976  Kind = tok::equalequal;
3977  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3978  } else {
3979  Kind = tok::equal;
3980  }
3981  break;
3982  case ',':
3983  Kind = tok::comma;
3984  break;
3985  case '#':
3986  Char = getCharAndSize(CurPtr, SizeTmp);
3987  if (Char == '#') {
3988  Kind = tok::hashhash;
3989  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3990  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3991  Kind = tok::hashat;
3992  if (!isLexingRawMode())
3993  Diag(BufferPtr, diag::ext_charize_microsoft);
3994  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3995  } else {
3996  // We parsed a # character. If this occurs at the start of the line,
3997  // it's actually the start of a preprocessing directive. Callback to
3998  // the preprocessor to handle it.
3999  // TODO: -fpreprocessed mode??
4000  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
4001  goto HandleDirective;
4002 
4003  Kind = tok::hash;
4004  }
4005  break;
4006 
4007  case '@':
4008  // Objective C support.
4009  if (CurPtr[-1] == '@' && LangOpts.ObjC)
4010  Kind = tok::at;
4011  else
4012  Kind = tok::unknown;
4013  break;
4014 
4015  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
4016  case '\\':
4017  if (!LangOpts.AsmPreprocessor) {
4018  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
4019  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
4020  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
4021  return true; // KeepWhitespaceMode
4022 
4023  // We only saw whitespace, so just try again with this lexer.
4024  // (We manually eliminate the tail call to avoid recursion.)
4025  goto LexNextToken;
4026  }
4027 
4028  return LexUnicodeIdentifierStart(Result, CodePoint, CurPtr);
4029  }
4030  }
4031 
4032  Kind = tok::unknown;
4033  break;
4034 
4035  default: {
4036  if (isASCII(Char)) {
4037  Kind = tok::unknown;
4038  break;
4039  }
4040 
4041  llvm::UTF32 CodePoint;
4042 
4043  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
4044  // an escaped newline.
4045  --CurPtr;
4046  llvm::ConversionResult Status =
4047  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
4048  (const llvm::UTF8 *)BufferEnd,
4049  &CodePoint,
4050  llvm::strictConversion);
4051  if (Status == llvm::conversionOK) {
4052  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
4053  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
4054  return true; // KeepWhitespaceMode
4055 
4056  // We only saw whitespace, so just try again with this lexer.
4057  // (We manually eliminate the tail call to avoid recursion.)
4058  goto LexNextToken;
4059  }
4060  return LexUnicodeIdentifierStart(Result, CodePoint, CurPtr);
4061  }
4062 
4064  PP->isPreprocessedOutput()) {
4065  ++CurPtr;
4066  Kind = tok::unknown;
4067  break;
4068  }
4069 
4070  // Non-ASCII characters tend to creep into source code unintentionally.
4071  // Instead of letting the parser complain about the unknown token,
4072  // just diagnose the invalid UTF-8, then drop the character.
4073  Diag(CurPtr, diag::err_invalid_utf8);
4074 
4075  BufferPtr = CurPtr+1;
4076  // We're pretending the character didn't exist, so just try again with
4077  // this lexer.
4078  // (We manually eliminate the tail call to avoid recursion.)
4079  goto LexNextToken;
4080  }
4081  }
4082 
4083  // Notify MIOpt that we read a non-whitespace/non-comment token.
4084  MIOpt.ReadToken();
4085 
4086  // Update the location of token as well as BufferPtr.
4087  FormTokenWithChars(Result, CurPtr, Kind);
4088  return true;
4089 
4090 HandleDirective:
4091  // We parsed a # character and it's the start of a preprocessing directive.
4092 
4093  FormTokenWithChars(Result, CurPtr, tok::hash);
4094  PP->HandleDirective(Result);
4095 
4097  // With a fatal failure in the module loader, we abort parsing.
4098  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
4099  return true;
4100  }
4101 
4102  // We parsed the directive; lex a token with the new state.
4103  return false;
4104 }
clang::Token::startToken
void startToken()
Reset all flags to cleared.
Definition: Token.h:170
clang::Lexer::isAtEndOfMacroExpansion
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:832
clang::Lexer::makeFileCharRange
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:893
clang::Preprocessor::CreateString
void CreateString(StringRef Str, Token &Tok, SourceLocation ExpansionLocStart=SourceLocation(), SourceLocation ExpansionLocEnd=SourceLocation())
Plop the specified string into a scratch buffer and set the specified token's location and length to ...
Definition: Preprocessor.cpp:486
getBeginningOfFileToken
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:514
_mm_cmpeq_epi8
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:3013
clang::Token::hasUCN
bool hasUCN() const
Returns true if this token contains a universal character name.
Definition: Token.h:295
isEndOfBlockCommentWithEscapedNewLine
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L, bool Trigraphs)
isBlockCommentEndOfEscapedNewLine - Return true if the specified newline character (either \n or \r) ...
Definition: Lexer.cpp:2547
clang::Lexer::SetCommentRetentionState
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode.
Definition: Lexer.h:244
clang::DiagnosticBuilder
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:1265
isInExpansionTokenRange
static bool isInExpansionTokenRange(const SourceLocation Loc, const SourceManager &SM)
Definition: Lexer.cpp:886
clang::CMK_Perforce
@ CMK_Perforce
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s.
Definition: Lexer.h:53
clang::CharSourceRange::getBegin
SourceLocation getBegin() const
Definition: SourceLocation.h:283
clang::Preprocessor::setRecordedPreambleConditionalStack
void setRecordedPreambleConditionalStack(ArrayRef< PPConditionalInfo > s)
Definition: Preprocessor.h:2486
clang::Token::isObjCAtKeyword
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:57
clang::DeclaratorContext::File
@ File
clang::SourceRange
A trivial tuple used to represent a source range.
Definition: SourceLocation.h:210
string
string(SUBSTRING ${CMAKE_CURRENT_BINARY_DIR} 0 ${PATH_LIB_START} PATH_HEAD) string(SUBSTRING $
Definition: CMakeLists.txt:22
clang::Lexer::getRawToken
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:463
clang::Lexer::isAsciiIdentifierContinueChar
static bool isAsciiIdentifierContinueChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1070
clang::Preprocessor::setCodeCompletionTokenRange
void setCodeCompletionTokenRange(const SourceLocation Start, const SourceLocation End)
Set the code completion token range for detecting replacement range later on.
Definition: Preprocessor.h:1359
clang::FixItHint::CreateInsertion
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:97
clang::Preprocessor::HandleIdentifier
bool HandleIdentifier(Token &Identifier)
Callback invoked when the lexer reads an identifier and has filled in the tokens IdentifierInfo membe...
Definition: Preprocessor.cpp:800
clang::ConflictMarkerKind
ConflictMarkerKind
ConflictMarkerKind - Kinds of conflict marker which the lexer might be recovering from.
Definition: Lexer.h:43
clang::SrcMgr::ExpansionInfo::isMacroArgExpansion
bool isMacroArgExpansion() const
Definition: SourceManager.h:387
GetTrigraphCharForLetter
static char GetTrigraphCharForLetter(char Letter)
GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, return the decoded trigraph...
Definition: Lexer.cpp:1178
clang::Token::getLiteralData
const char * getLiteralData() const
getLiteralData - For a literal token (numeric constant, string, etc), this returns a pointer to the s...
Definition: Token.h:216
clang::Lexer::getSpelling
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:405
clang::IdentifierTable::get
IdentifierInfo & get(StringRef Name)
Return the identifier token info for the specified named identifier.
Definition: IdentifierTable.h:596
clang::MultipleIncludeOpt::ReadToken
void ReadToken()
Definition: MultipleIncludeOpt.h:106
clang::if
if(T->getSizeExpr()) TRY_TO(TraverseStmt(T -> getSizeExpr()))
Definition: RecursiveASTVisitor.h:1098
clang::SourceLocation
Encodes a location in the source.
Definition: SourceLocation.h:86
clang::DiagnosticsEngine::isIgnored
bool isIgnored(unsigned DiagID, SourceLocation Loc) const
Determine whether the diagnostic is known to be ignored.
Definition: Diagnostic.h:910
clang::SourceLocation::getLocWithOffset
SourceLocation getLocWithOffset(IntTy Offset) const
Return a source location with the specified offset from this SourceLocation.
Definition: SourceLocation.h:134
clang::Preprocessor::getCommentRetentionState
bool getCommentRetentionState() const
Definition: Preprocessor.h:1039
clang::Lexer::inKeepCommentMode
bool inKeepCommentMode() const
inKeepCommentMode - Return true if the lexer should return comments as tokens.
Definition: Lexer.h:237
clang::PreprocessorLexer
Definition: PreprocessorLexer.h:30
clang::Lexer::Create_PragmaLexer
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:194
clang::PreprocessorLexer::ConditionalStack
SmallVector< PPConditionalInfo, 4 > ConditionalStack
Information about the set of #if/#ifdef/#ifndef blocks we are currently in.
Definition: PreprocessorLexer.h:76
FindConflictEnd
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2973
AttributeLangSupport::C
@ C
Definition: SemaDeclAttr.cpp:55
clang::Token::getIdentifierInfo
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:178
clang::PreprocessorLexer::MIOpt
MultipleIncludeOpt MIOpt
A state machine that detects the #ifndef-wrapping a file idiom for the multiple-include optimization.
Definition: PreprocessorLexer.h:72
clang::DiagnosticsEngine
Concrete class used by the front-end to report problems and issues.
Definition: Diagnostic.h:192
memcpy
__DEVICE__ void * memcpy(void *__a, const void *__b, size_t __c)
Definition: __clang_cuda_device_functions.h:1549
int
__device__ int
Definition: __clang_hip_libdevice_declares.h:63
clang::PreprocessorLexer::ParsingFilename
bool ParsingFilename
True after #include; turns <xx> or "xxx" into a tok::header_name token.
Definition: PreprocessorLexer.h:53
clang::Token::isAtStartOfLine
bool isAtStartOfLine() const
isAtStartOfLine - Return true if this token is at the start of a line.
Definition: Token.h:267
clang::Preprocessor::CodeCompleteIncludedFile
void CodeCompleteIncludedFile(llvm::StringRef Dir, bool IsAngled)
Hook used by the lexer to invoke the "included file" code completion point.
Definition: Preprocessor.cpp:448
llvm::Optional
Definition: LLVM.h:40
clang::Lexer::getSourceText
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:962
SourceManager.h
clang::Lexer::skipOver
bool skipOver(unsigned NumBytes)
Skip over NumBytes bytes.
Definition: Lexer.cpp:229
XIDContinueRanges
static const llvm::sys::UnicodeCharRange XIDContinueRanges[]
Definition: UnicodeCharSets.h:239
clang::PreprocessorLexer::LexingRawMode
bool LexingRawMode
True if in raw mode.
Definition: PreprocessorLexer.h:68
clang::Token::IsEditorPlaceholder
@ IsEditorPlaceholder
Definition: Token.h:86
vec_any_eq
static __inline__ int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:15958
clang::Lexer::isKeepWhitespaceMode
bool isKeepWhitespaceMode() const
isKeepWhitespaceMode - Return true if the lexer should return tokens for every character in the file,...
Definition: Lexer.h:223
clang::Lexer::MeasureTokenLength
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:452
clang::Lexer::LexFromRawLexer
bool LexFromRawLexer(Token &Result)
LexFromRawLexer - Lex a token from a designated raw lexer (one with no associated preprocessor object...
Definition: Lexer.h:211
clang::Token
Token - This structure provides full information about a lexed token.
Definition: Token.h:34
clang::tok::isStringLiteral
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:81
LiteralSupport.h
End
SourceLocation End
Definition: USRLocFinder.cpp:167
clang::StringLiteralParser::isValidUDSuffix
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
Definition: LiteralSupport.cpp:2140
clang::Preprocessor::getIdentifierTable
IdentifierTable & getIdentifierTable()
Definition: Preprocessor.h:1001
clang::isPreprocessingNumberBody
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:153
clang::SourceManager
This class handles loading and caching of source files into memory.
Definition: SourceManager.h:627
clang::Lexer::Stringify
static std::string Stringify(StringRef Str, bool Charify=false)
Stringify - Convert the specified string into a C string by i) escaping '\' and " characters and ii) ...
Definition: Lexer.cpp:263
Preprocessor.h
clang::Lexer
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens.
Definition: Lexer.h:77
Offset
unsigned Offset
Definition: Format.cpp:2553
TokenKinds.h
U
clang::CharSourceRange::getCharRange
static CharSourceRange getCharRange(SourceRange R)
Definition: SourceLocation.h:265
diagnoseInvalidUnicodeCodepointInIdentifier
static void diagnoseInvalidUnicodeCodepointInIdentifier(DiagnosticsEngine &Diags, const LangOptions &LangOpts, uint32_t CodePoint, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1622
length
float __ovld __cnfn length(float)
Return the length of vector p, i.e., sqrt(p.x2 + p.y 2 + ...)
clang::CMK_Normal
@ CMK_Normal
A normal or diff3 conflict marker, initiated by at least 7 "<"s, separated by at least 7 "="s or "|"s...
Definition: Lexer.h:49
clang::Lexer::Diag
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1168
isAllowedIDChar
static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1461
clang::SrcMgr::ExpansionInfo
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded,...
Definition: SourceManager.h:348
MultipleIncludeOpt.h
clang::Token::getKind
tok::TokenKind getKind() const
Definition: Token.h:92
LangOptions.h
clang::Lexer::isNewLineEscaped
static bool isNewLineEscaped(const char *BufferStart, const char *Str)
Checks whether new line pointed by Str is preceded by escape sequence.
Definition: Lexer.cpp:1074
makeRangeFromFileLocs
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:858
Diagnostic.h
LexDiagnostic.h
_mm_set1_epi8
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:3642
intptr_t
__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-base.h:132
clang::Preprocessor::Diag
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
Definition: Preprocessor.h:1806
GetMappedTokenLoc
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the lexer buffer was all exp...
Definition: Lexer.cpp:1125
clang::Token::is
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:97
clang::Lexer::Lexer
Lexer(FileID FID, const llvm::MemoryBufferRef &InputFile, Preprocessor &PP, bool IsFirstIncludeOfFile=true)
Lexer constructor - Create a new lexer object for the specified buffer with the specified preprocesso...
Definition: Lexer.cpp:135
IdentifierTable.h
clang::SrcMgr::ExpansionInfo::getSpellingLoc
SourceLocation getSpellingLoc() const
Definition: SourceManager.h:366
clang::Lexer::getTokenPrefixLength
static unsigned getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo, const SourceManager &SM, const LangOptions &LangOpts)
Get the physical length (including trigraphs and escaped newlines) of the first Characters characters...
Definition: Lexer.cpp:728
llvm::SmallString
Definition: LLVM.h:37
clang::Token::needsCleaning
bool needsCleaning() const
Return true if this token has trigraphs or escaped newlines in it.
Definition: Token.h:284
clang::Lexer::getIndentationForLine
static StringRef getIndentationForLine(SourceLocation Loc, const SourceManager &SM)
Returns the leading whitespace for line that corresponds to the given location Loc.
Definition: Lexer.cpp:1094
clang::Token::LeadingEmptyMacro
@ LeadingEmptyMacro
Definition: Token.h:79
clang::PreprocessorLexer::ParsingPreprocessorDirective
bool ParsingPreprocessorDirective
True when parsing #XXX; turns '\n' into a tok::eod token.
Definition: PreprocessorLexer.h:50
clang::isRawStringDelimBody
LLVM_READONLY bool isRawStringDelimBody(unsigned char c)
Return true if this is the body character of a C++ raw string delimiter.
Definition: CharInfo.h:160
clang::PreambleBounds
Describes the bounds (start, size) of the preamble and a flag required by PreprocessorOptions::Precom...
Definition: Lexer.h:59
clang::PreprocessorLexer::FID
const FileID FID
The SourceManager FileID corresponding to the file being lexed.
Definition: PreprocessorLexer.h:40
C11DisallowedInitialIDCharRanges
static const llvm::sys::UnicodeCharRange C11DisallowedInitialIDCharRanges[]
Definition: UnicodeCharSets.h:612
clang::SourceLocation::isFileID
bool isFileID() const
Definition: SourceLocation.h:102
clang::tok::ObjCKeywordKind
ObjCKeywordKind
Provides a namespace for Objective-C keywords which start with an '@'.
Definition: TokenKinds.h:41
clang::Lexer::ReadToEndOfLine
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string.
Definition: Lexer.cpp:2804
clang::PreprocessorLexer::PP
Preprocessor * PP
Definition: PreprocessorLexer.h:37
getSpellingSlow
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:278
clang::Lexer::findNextToken
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:1260
altivec.h
clang::Lexer::getImmediateMacroName
static StringRef getImmediateMacroName(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:998
clang::IdentifierInfo::getObjCKeywordID
tok::ObjCKeywordKind getObjCKeywordID() const
Return the Objective-C keyword ID for the this identifier.
Definition: IdentifierTable.h:291
Line
const AnnotatedLine * Line
Definition: UsingDeclarationsSorter.cpp:68
clang::PreprocessorLexer::isLexingRawMode
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
Definition: PreprocessorLexer.h:142
clang::Lexer::getSourceLocation
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:1149
findPlaceholderEnd
static const char * findPlaceholderEnd(const char *CurPtr, const char *BufferEnd)
Definition: Lexer.cpp:3077
clang::Token::getRawIdentifier
StringRef getRawIdentifier() const
getRawIdentifier - For a raw identifier token (i.e., an identifier lexed in raw mode),...
Definition: Token.h:204
clang::Token::isLiteral
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:114
clang::isAsciiIdentifierContinue
LLVM_READONLY bool isAsciiIdentifierContinue(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:64
clang::Token::getLength
unsigned getLength() const
Definition: Token.h:128
clang::Token::LeadingSpace
@ LeadingSpace
Definition: Token.h:75
makeCharRange
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1507
clang::Preprocessor::getSourceManager
SourceManager & getSourceManager() const
Definition: Preprocessor.h:998
clang::Preprocessor::HandleComment
bool HandleComment(Token &result, SourceRange Comment)
Definition: Preprocessor.cpp:1402
clang::Preprocessor::getSpelling
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
Return the 'spelling' of the token at the given location; does not go up to the spelling location or ...
Definition: Preprocessor.h:1821
clang::isWhitespace
LLVM_READONLY bool isWhitespace(unsigned char c)
Return true if this character is horizontal or vertical ASCII whitespace: ' ', '\t',...
Definition: CharInfo.h:93
UnicodeWhitespaceCharRanges
static const llvm::sys::UnicodeCharRange UnicodeWhitespaceCharRanges[]
Definition: UnicodeCharSets.h:629
clang::SrcMgr::SLocEntry::getExpansion
const ExpansionInfo & getExpansion() const
Definition: SourceManager.h:490
SourceLocation.h
clang::tok::TokenKind
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
clang::Token::HasUDSuffix
@ HasUDSuffix
Definition: Token.h:80
clang::Preprocessor::CodeCompleteNaturalLanguage
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
Definition: Preprocessor.cpp:455
P
StringRef P
Definition: ASTMatchersInternal.cpp:563
clang::Preprocessor::HandleDirective
void HandleDirective(Token &Result)
Callback invoked when the lexer sees a # token at the start of a line.
Definition: PPDirectives.cpp:1087
CharInfo.h
clang::Lexer::getBufferLocation
const char * getBufferLocation() const
Return the current location in the buffer.
Definition: Lexer.h:283
clang::syntax::NodeRole::Size
@ Size
clang::Preprocessor::hadModuleLoaderFatalFailure
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:1018
clang::Preprocessor::getCodeCompletionFileLoc
SourceLocation getCodeCompletionFileLoc() const
Returns the start location of the file of code-completion point.
Definition: Preprocessor.h:1725
StringifyImpl
static void StringifyImpl(T &Str, char Quote)
Definition: Lexer.cpp:238
clang::Preprocessor::isCodeCompletionEnabled
bool isCodeCompletionEnabled() const
Determine if we are performing code completion.
Definition: Preprocessor.h:1713
clang::SrcMgr::ExpansionInfo::getExpansionLocStart
SourceLocation getExpansionLocStart() const
Definition: SourceManager.h:370
Begin
SourceLocation Begin
Definition: USRLocFinder.cpp:165
Lexer.h
Value
Value
Definition: UninitializedValues.cpp:102
clang::Preprocessor::setCodeCompletionIdentifierInfo
void setCodeCompletionIdentifierInfo(IdentifierInfo *Filter)
Set the code completion token for filtering purposes.
Definition: Preprocessor.h:1353
clang::SourceManager::getCharacterData
const char * getCharacterData(SourceLocation SL, bool *Invalid=nullptr) const
Return a pointer to the start of the specified location in the appropriate spelling MemoryBuffer.
Definition: SourceManager.cpp:1152
isAllowedInitiallyIDChar
static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1486
clang::frontend::After
@ After
Like System, but searched after the system directories.
Definition: HeaderSearchOptions.h:61
clang::CharSourceRange
Represents a character-granular source range.
Definition: SourceLocation.h:253
findBeginningOfLine
static const char * findBeginningOfLine(StringRef Buffer, unsigned Offset)
Returns the pointer that points to the beginning of line that contains the given offset,...
Definition: Lexer.cpp:497
LLVM.h
clang::Token::StartOfLine
@ StartOfLine
Definition: Token.h:73
C99DisallowedInitialIDCharRanges
static const llvm::sys::UnicodeCharRange C99DisallowedInitialIDCharRanges[]
Definition: UnicodeCharSets.h:620
clang::CharSourceRange::getEnd
SourceLocation getEnd() const
Definition: SourceLocation.h:284
clang::Preprocessor::isPreprocessedOutput
bool isPreprocessedOutput() const
Returns true if the preprocessor is responsible for generating output, false if it is producing token...
Definition: Preprocessor.h:1060
clang::IdentifierInfo
One of these records is kept for each identifier that is lexed.
Definition: IdentifierTable.h:84
clang::Token::getLocation
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file.
Definition: Token.h:125
clang::SourceLocation::getFromRawEncoding
static SourceLocation getFromRawEncoding(UIntTy Encoding)
Turn a raw encoding of a SourceLocation object into a real SourceLocation.
Definition: SourceLocation.h:152
clang::SourceLocation::isMacroID
bool isMacroID() const
Definition: SourceLocation.h:103
clang::IdentifierInfo::isHandleIdentifierCase
bool isHandleIdentifierCase() const
Return true if the Preprocessor::HandleIdentifier must be called on a token of this identifier.
Definition: IdentifierTable.h:380
clang::LangOptions
Keeps track of the various options that can be enabled, which controls the dialect of C or C++ that i...
Definition: LangOptions.h:78
clang::Lexer::SetKeepWhitespaceMode
void SetKeepWhitespaceMode(bool Val)
SetKeepWhitespaceMode - This method lets clients enable or disable whitespace retention mode.
Definition: Lexer.h:229
clang::ObjCPropertyAttribute::Kind
Kind
Definition: DeclObjCCommon.h:22
clang::Lexer::getImmediateMacroNameForDiagnostics
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1045
clang::Token::isAnnotation
bool isAnnotation() const
Return true if this is any of tok::annot_* kind tokens.
Definition: Token.h:119
clang::Preprocessor::getPreprocessorOpts
PreprocessorOptions & getPreprocessorOpts() const
Retrieve the preprocessor options used to initialize this preprocessor.
Definition: Preprocessor.h:989
maybeDiagnoseUTF8Homoglyph
static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range)
After encountering UTF-8 character C and interpreting it as an identifier character,...
Definition: Lexer.cpp:1542
clang::SourceLocation::isInvalid
bool isInvalid() const
Definition: SourceLocation.h:111
clang::Preprocessor::getCodeCompletionLoc
SourceLocation getCodeCompletionLoc() const
Returns the location of the code-completion point.
Definition: Preprocessor.h:1719
clang::isVerticalWhitespace
LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
Returns true if this character is vertical ASCII whitespace: '\n', '\r'.
Definition: CharInfo.h:84
clang
Definition: CalledOnceCheck.h:17
clang::Lexer::isAtStartOfMacroExpansion
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:810
clang::PreprocessorOptions::LexEditorPlaceholders
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.
Definition: PreprocessorOptions.h:148
clang::Preprocessor::LookUpIdentifierInfo
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
Given a tok::raw_identifier token, look up the identifier information for the token and install it in...
Definition: Preprocessor.cpp:700
_mm_movemask_epi8
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:4194
maybeDiagnoseIDCharCompat
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1513
isUnicodeWhitespace
static bool isUnicodeWhitespace(uint32_t Codepoint)
Definition: Lexer.cpp:1455
clang::Preprocessor::getDiagnostics
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:991
clang::Lexer::GetBeginningOfToken
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:562
clang::Lexer::isPragmaLexer
bool isPragmaLexer() const
isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
Definition: Lexer.h:200
clang::operator<
bool operator<(DeclarationName LHS, DeclarationName RHS)
Ordering on two declaration names.
Definition: DeclarationName.h:549
clang::FileID
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
Definition: SourceLocation.h:38
C99AllowedIDCharRanges
static const llvm::sys::UnicodeCharRange C99AllowedIDCharRanges[]
Definition: UnicodeCharSets.h:393
clang::SourceLocation::isValid
bool isValid() const
Return true if this is a valid SourceLocation object.
Definition: SourceLocation.h:110
clang::FileID::isInvalid
bool isInvalid() const
Definition: SourceLocation.h:45
clang::FixItHint::CreateRemoval
static FixItHint CreateRemoval(CharSourceRange RemoveRange)
Create a code modification hint that removes the given source range.
Definition: Diagnostic.h:123
clang::Preprocessor::isInPrimaryFile
bool isInPrimaryFile() const
Return true if we're in the top-level file, not in a #include.
Definition: PPLexerChange.cpp:34
clang::Token::NeedsCleaning
@ NeedsCleaning
Definition: Token.h:78
UnicodeCharSets.h
clang::MultipleIncludeOpt::ExitTopLevelConditional
void ExitTopLevelConditional()
Called when the lexer exits the top-level conditional.
Definition: MultipleIncludeOpt.h:149
clang::Lexer::resetExtendedTokenMode
void resetExtendedTokenMode()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:171
c
__device__ __2f16 float c
Definition: __clang_hip_libdevice_declares.h:315
clang::isHorizontalWhitespace
LLVM_READONLY bool isHorizontalWhitespace(unsigned char c)
Returns true if this character is horizontal ASCII whitespace: ' ', '\t', '\f', '\v'.
Definition: CharInfo.h:76
DecodeTrigraphChar
static char DecodeTrigraphChar(const char *CP, Lexer *L, bool Trigraphs)
DecodeTrigraphChar - If the specified character is a legal trigraph when prefixed with ?...
Definition: Lexer.cpp:1197
clang::Token::setFlag
void setFlag(TokenFlags Flag)
Set the specified flag.
Definition: Token.h:235
clang::Lexer::getLocForEndOfToken
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:788
clang::comments::tok::eof
@ eof
Definition: CommentLexer.h:33
clang::Lexer::ComputePreamble
static PreambleBounds ComputePreamble(StringRef Buffer, const LangOptions &LangOpts, unsigned MaxLines=0)
Compute the preamble of the given file.
Definition: Lexer.cpp:590
llvm::SmallVectorImpl< char >
clang::Token::HasUCN
@ HasUCN
Definition: Token.h:81
clang::Lexer::findLocationAfterToken
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:1293
clang::Lexer::getCharAndSizeNoWarn
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:561
PreprocessorOptions.h
clang::Preprocessor
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:131
clang::Preprocessor::isRecordingPreamble
bool isRecordingPreamble() const
Definition: Preprocessor.h:2474
clang::LineComment
@ LineComment
Definition: LangStandard.h:48
clang::Token::getObjCKeywordID
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:66
SM
#define SM(sm)
Definition: Cuda.cpp:81
C11AllowedIDCharRanges
static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[]
Definition: UnicodeCharSets.h:363
clang::Lexer::getSourceLocation
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file.
Definition: Lexer.h:278
Token.h
clang::Preprocessor::getEmptylineHandler
EmptylineHandler * getEmptylineHandler() const
Definition: Preprocessor.h:1319
XIDStartRanges
static const llvm::sys::UnicodeCharRange XIDStartRanges[]
Definition: UnicodeCharSets.h:14
clang::isAsciiIdentifierStart
LLVM_READONLY bool isAsciiIdentifierStart(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:54
clang::Decl::getLocation
SourceLocation getLocation() const
Definition: DeclBase.h:425
clang::RISCV::Invalid
@ Invalid
Definition: RISCVVIntrinsicUtils.h:148
clang::SourceLocation::UIntTy
uint32_t UIntTy
Definition: SourceLocation.h:93
clang::isASCII
LLVM_READNONE bool isASCII(char c)
Returns true if a byte is an ASCII character.
Definition: CharInfo.h:42
clang::SourceLocation::getRawEncoding
UIntTy getRawEncoding() const
When a SourceLocation itself cannot be used, this returns an (opaque) 32-bit integer encoding for it.
Definition: SourceLocation.h:146
clang::DiagnosticsEngine::Report
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1536
clang::CMK_None
@ CMK_None
Not within a conflict marker.
Definition: Lexer.h:45
clang::Preprocessor::HandleEndOfFile
bool HandleEndOfFile(Token &Result, bool isEndOfMacro=false)
Callback invoked when the lexer hits the end of the current file.
Definition: PPLexerChange.cpp:308
clang::SrcMgr::SLocEntry
This is a discriminated union of FileInfo and ExpansionInfo.
Definition: SourceManager.h:468