clang  9.0.0svn
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"
20 #include "clang/Basic/TokenKinds.h"
24 #include "clang/Lex/Preprocessor.h"
26 #include "clang/Lex/Token.h"
27 #include "clang/Basic/Diagnostic.h"
28 #include "clang/Basic/LLVM.h"
29 #include "clang/Basic/TokenKinds.h"
30 #include "llvm/ADT/None.h"
31 #include "llvm/ADT/Optional.h"
32 #include "llvm/ADT/StringExtras.h"
33 #include "llvm/ADT/StringSwitch.h"
34 #include "llvm/ADT/StringRef.h"
35 #include "llvm/Support/Compiler.h"
36 #include "llvm/Support/ConvertUTF.h"
37 #include "llvm/Support/MathExtras.h"
38 #include "llvm/Support/MemoryBuffer.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 #.
114  ParsingPreprocessorDirective = false;
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 
129 /// Lexer constructor - Create a new lexer object for the specified buffer
130 /// with the specified preprocessor managing the lexing process. This lexer
131 /// assumes that the associated file buffer and Preprocessor objects will
132 /// outlive it, so it doesn't take ownership of either of them.
133 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
134  : PreprocessorLexer(&PP, FID),
135  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
136  LangOpts(PP.getLangOpts()) {
137  InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
138  InputFile->getBufferEnd());
139 
141 }
142 
143 /// Lexer constructor - Create a new raw lexer object. This object is only
144 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
145 /// range will outlive it, so it doesn't take ownership of it.
146 Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
147  const char *BufStart, const char *BufPtr, const char *BufEnd)
148  : FileLoc(fileloc), LangOpts(langOpts) {
149  InitLexer(BufStart, BufPtr, BufEnd);
150 
151  // We *are* in raw mode.
152  LexingRawMode = true;
153 }
154 
155 /// Lexer constructor - Create a new raw lexer object. This object is only
156 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
157 /// range will outlive it, so it doesn't take ownership of it.
158 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
159  const SourceManager &SM, const LangOptions &langOpts)
160  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile->getBufferStart(),
161  FromFile->getBufferStart(), FromFile->getBufferEnd()) {}
162 
164  assert(PP && "Cannot reset token mode without a preprocessor");
165  if (LangOpts.TraditionalCPP)
166  SetKeepWhitespaceMode(true);
167  else
169 }
170 
171 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
172 /// _Pragma expansion. This has a variety of magic semantics that this method
173 /// sets up. It returns a new'd Lexer that must be delete'd when done.
174 ///
175 /// On entrance to this routine, TokStartLoc is a macro location which has a
176 /// spelling loc that indicates the bytes to be lexed for the token and an
177 /// expansion location that indicates where all lexed tokens should be
178 /// "expanded from".
179 ///
180 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
181 /// normal lexer that remaps tokens as they fly by. This would require making
182 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
183 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
184 /// out of the critical path of the lexer!
185 ///
187  SourceLocation ExpansionLocStart,
188  SourceLocation ExpansionLocEnd,
189  unsigned TokLen, Preprocessor &PP) {
191 
192  // Create the lexer as if we were going to lex the file normally.
193  FileID SpellingFID = SM.getFileID(SpellingLoc);
194  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
195  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
196 
197  // Now that the lexer is created, change the start/end locations so that we
198  // just lex the subsection of the file that we want. This is lexing from a
199  // scratch buffer.
200  const char *StrData = SM.getCharacterData(SpellingLoc);
201 
202  L->BufferPtr = StrData;
203  L->BufferEnd = StrData+TokLen;
204  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
205 
206  // Set the SourceLocation with the remapping information. This ensures that
207  // GetMappedTokenLoc will remap the tokens as they are lexed.
208  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
209  ExpansionLocStart,
210  ExpansionLocEnd, TokLen);
211 
212  // Ensure that the lexer thinks it is inside a directive, so that end \n will
213  // return an EOD token.
215 
216  // This lexer really is for _Pragma.
217  L->Is_PragmaLexer = true;
218  return L;
219 }
220 
221 template <typename T> static void StringifyImpl(T &Str, char Quote) {
222  typename T::size_type i = 0, e = Str.size();
223  while (i < e) {
224  if (Str[i] == '\\' || Str[i] == Quote) {
225  Str.insert(Str.begin() + i, '\\');
226  i += 2;
227  ++e;
228  } else if (Str[i] == '\n' || Str[i] == '\r') {
229  // Replace '\r\n' and '\n\r' to '\\' followed by 'n'.
230  if ((i < e - 1) && (Str[i + 1] == '\n' || Str[i + 1] == '\r') &&
231  Str[i] != Str[i + 1]) {
232  Str[i] = '\\';
233  Str[i + 1] = 'n';
234  } else {
235  // Replace '\n' and '\r' to '\\' followed by 'n'.
236  Str[i] = '\\';
237  Str.insert(Str.begin() + i + 1, 'n');
238  ++e;
239  }
240  i += 2;
241  } else
242  ++i;
243  }
244 }
245 
246 std::string Lexer::Stringify(StringRef Str, bool Charify) {
247  std::string Result = Str;
248  char Quote = Charify ? '\'' : '"';
249  StringifyImpl(Result, Quote);
250  return Result;
251 }
252 
254 
255 //===----------------------------------------------------------------------===//
256 // Token Spelling
257 //===----------------------------------------------------------------------===//
258 
259 /// Slow case of getSpelling. Extract the characters comprising the
260 /// spelling of this token from the provided input buffer.
261 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
262  const LangOptions &LangOpts, char *Spelling) {
263  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
264 
265  size_t Length = 0;
266  const char *BufEnd = BufPtr + Tok.getLength();
267 
268  if (tok::isStringLiteral(Tok.getKind())) {
269  // Munch the encoding-prefix and opening double-quote.
270  while (BufPtr < BufEnd) {
271  unsigned Size;
272  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
273  BufPtr += Size;
274 
275  if (Spelling[Length - 1] == '"')
276  break;
277  }
278 
279  // Raw string literals need special handling; trigraph expansion and line
280  // splicing do not occur within their d-char-sequence nor within their
281  // r-char-sequence.
282  if (Length >= 2 &&
283  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
284  // Search backwards from the end of the token to find the matching closing
285  // quote.
286  const char *RawEnd = BufEnd;
287  do --RawEnd; while (*RawEnd != '"');
288  size_t RawLength = RawEnd - BufPtr + 1;
289 
290  // Everything between the quotes is included verbatim in the spelling.
291  memcpy(Spelling + Length, BufPtr, RawLength);
292  Length += RawLength;
293  BufPtr += RawLength;
294 
295  // The rest of the token is lexed normally.
296  }
297  }
298 
299  while (BufPtr < BufEnd) {
300  unsigned Size;
301  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
302  BufPtr += Size;
303  }
304 
305  assert(Length < Tok.getLength() &&
306  "NeedsCleaning flag set on token that didn't need cleaning!");
307  return Length;
308 }
309 
310 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
311 /// token are the characters used to represent the token in the source file
312 /// after trigraph expansion and escaped-newline folding. In particular, this
313 /// wants to get the true, uncanonicalized, spelling of things like digraphs
314 /// UCNs, etc.
316  SmallVectorImpl<char> &buffer,
317  const SourceManager &SM,
318  const LangOptions &options,
319  bool *invalid) {
320  // Break down the source location.
321  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
322 
323  // Try to the load the file buffer.
324  bool invalidTemp = false;
325  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
326  if (invalidTemp) {
327  if (invalid) *invalid = true;
328  return {};
329  }
330 
331  const char *tokenBegin = file.data() + locInfo.second;
332 
333  // Lex from the start of the given location.
334  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
335  file.begin(), tokenBegin, file.end());
336  Token token;
337  lexer.LexFromRawLexer(token);
338 
339  unsigned length = token.getLength();
340 
341  // Common case: no need for cleaning.
342  if (!token.needsCleaning())
343  return StringRef(tokenBegin, length);
344 
345  // Hard case, we need to relex the characters into the string.
346  buffer.resize(length);
347  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
348  return StringRef(buffer.data(), buffer.size());
349 }
350 
351 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
352 /// token are the characters used to represent the token in the source file
353 /// after trigraph expansion and escaped-newline folding. In particular, this
354 /// wants to get the true, uncanonicalized, spelling of things like digraphs
355 /// UCNs, etc.
356 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
357  const LangOptions &LangOpts, bool *Invalid) {
358  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
359 
360  bool CharDataInvalid = false;
361  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
362  &CharDataInvalid);
363  if (Invalid)
364  *Invalid = CharDataInvalid;
365  if (CharDataInvalid)
366  return {};
367 
368  // If this token contains nothing interesting, return it directly.
369  if (!Tok.needsCleaning())
370  return std::string(TokStart, TokStart + Tok.getLength());
371 
372  std::string Result;
373  Result.resize(Tok.getLength());
374  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
375  return Result;
376 }
377 
378 /// getSpelling - This method is used to get the spelling of a token into a
379 /// preallocated buffer, instead of as an std::string. The caller is required
380 /// to allocate enough space for the token, which is guaranteed to be at least
381 /// Tok.getLength() bytes long. The actual length of the token is returned.
382 ///
383 /// Note that this method may do two possible things: it may either fill in
384 /// the buffer specified with characters, or it may *change the input pointer*
385 /// to point to a constant buffer with the data already in it (avoiding a
386 /// copy). The caller is not allowed to modify the returned buffer pointer
387 /// if an internal buffer is returned.
388 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
389  const SourceManager &SourceMgr,
390  const LangOptions &LangOpts, bool *Invalid) {
391  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
392 
393  const char *TokStart = nullptr;
394  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
395  if (Tok.is(tok::raw_identifier))
396  TokStart = Tok.getRawIdentifier().data();
397  else if (!Tok.hasUCN()) {
398  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
399  // Just return the string from the identifier table, which is very quick.
400  Buffer = II->getNameStart();
401  return II->getLength();
402  }
403  }
404 
405  // NOTE: this can be checked even after testing for an IdentifierInfo.
406  if (Tok.isLiteral())
407  TokStart = Tok.getLiteralData();
408 
409  if (!TokStart) {
410  // Compute the start of the token in the input lexer buffer.
411  bool CharDataInvalid = false;
412  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
413  if (Invalid)
414  *Invalid = CharDataInvalid;
415  if (CharDataInvalid) {
416  Buffer = "";
417  return 0;
418  }
419  }
420 
421  // If this token contains nothing interesting, return it directly.
422  if (!Tok.needsCleaning()) {
423  Buffer = TokStart;
424  return Tok.getLength();
425  }
426 
427  // Otherwise, hard case, relex the characters into the string.
428  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
429 }
430 
431 /// MeasureTokenLength - Relex the token at the specified location and return
432 /// its length in bytes in the input file. If the token needs cleaning (e.g.
433 /// includes a trigraph or an escaped newline) then this count includes bytes
434 /// that are part of that.
436  const SourceManager &SM,
437  const LangOptions &LangOpts) {
438  Token TheTok;
439  if (getRawToken(Loc, TheTok, SM, LangOpts))
440  return 0;
441  return TheTok.getLength();
442 }
443 
444 /// Relex the token at the specified location.
445 /// \returns true if there was a failure, false on success.
447  const SourceManager &SM,
448  const LangOptions &LangOpts,
449  bool IgnoreWhiteSpace) {
450  // TODO: this could be special cased for common tokens like identifiers, ')',
451  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
452  // all obviously single-char tokens. This could use
453  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
454  // something.
455 
456  // If this comes from a macro expansion, we really do want the macro name, not
457  // the token this macro expanded to.
458  Loc = SM.getExpansionLoc(Loc);
459  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
460  bool Invalid = false;
461  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
462  if (Invalid)
463  return true;
464 
465  const char *StrData = Buffer.data()+LocInfo.second;
466 
467  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
468  return true;
469 
470  // Create a lexer starting at the beginning of this token.
471  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
472  Buffer.begin(), StrData, Buffer.end());
473  TheLexer.SetCommentRetentionState(true);
474  TheLexer.LexFromRawLexer(Result);
475  return false;
476 }
477 
478 /// Returns the pointer that points to the beginning of line that contains
479 /// the given offset, or null if the offset if invalid.
480 static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
481  const char *BufStart = Buffer.data();
482  if (Offset >= Buffer.size())
483  return nullptr;
484 
485  const char *LexStart = BufStart + Offset;
486  for (; LexStart != BufStart; --LexStart) {
487  if (isVerticalWhitespace(LexStart[0]) &&
488  !Lexer::isNewLineEscaped(BufStart, LexStart)) {
489  // LexStart should point at first character of logical line.
490  ++LexStart;
491  break;
492  }
493  }
494  return LexStart;
495 }
496 
498  const SourceManager &SM,
499  const LangOptions &LangOpts) {
500  assert(Loc.isFileID());
501  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
502  if (LocInfo.first.isInvalid())
503  return Loc;
504 
505  bool Invalid = false;
506  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
507  if (Invalid)
508  return Loc;
509 
510  // Back up from the current location until we hit the beginning of a line
511  // (or the buffer). We'll relex from that point.
512  const char *StrData = Buffer.data() + LocInfo.second;
513  const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
514  if (!LexStart || LexStart == StrData)
515  return Loc;
516 
517  // Create a lexer starting at the beginning of this token.
518  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
519  Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
520  Buffer.end());
521  TheLexer.SetCommentRetentionState(true);
522 
523  // Lex tokens until we find the token that contains the source location.
524  Token TheTok;
525  do {
526  TheLexer.LexFromRawLexer(TheTok);
527 
528  if (TheLexer.getBufferLocation() > StrData) {
529  // Lexing this token has taken the lexer past the source location we're
530  // looking for. If the current token encompasses our source location,
531  // return the beginning of that token.
532  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
533  return TheTok.getLocation();
534 
535  // We ended up skipping over the source location entirely, which means
536  // that it points into whitespace. We're done here.
537  break;
538  }
539  } while (TheTok.getKind() != tok::eof);
540 
541  // We've passed our source location; just return the original source location.
542  return Loc;
543 }
544 
546  const SourceManager &SM,
547  const LangOptions &LangOpts) {
548  if (Loc.isFileID())
549  return getBeginningOfFileToken(Loc, SM, LangOpts);
550 
551  if (!SM.isMacroArgExpansion(Loc))
552  return Loc;
553 
554  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
555  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
556  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
557  std::pair<FileID, unsigned> BeginFileLocInfo =
558  SM.getDecomposedLoc(BeginFileLoc);
559  assert(FileLocInfo.first == BeginFileLocInfo.first &&
560  FileLocInfo.second >= BeginFileLocInfo.second);
561  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
562 }
563 
564 namespace {
565 
567  PDK_Skipped,
568  PDK_Unknown
569 };
570 
571 } // namespace
572 
574  const LangOptions &LangOpts,
575  unsigned MaxLines) {
576  // Create a lexer starting at the beginning of the file. Note that we use a
577  // "fake" file source location at offset 1 so that the lexer will track our
578  // position within the file.
579  const unsigned StartOffset = 1;
580  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
581  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
582  Buffer.end());
583  TheLexer.SetCommentRetentionState(true);
584 
585  bool InPreprocessorDirective = false;
586  Token TheTok;
587  SourceLocation ActiveCommentLoc;
588 
589  unsigned MaxLineOffset = 0;
590  if (MaxLines) {
591  const char *CurPtr = Buffer.begin();
592  unsigned CurLine = 0;
593  while (CurPtr != Buffer.end()) {
594  char ch = *CurPtr++;
595  if (ch == '\n') {
596  ++CurLine;
597  if (CurLine == MaxLines)
598  break;
599  }
600  }
601  if (CurPtr != Buffer.end())
602  MaxLineOffset = CurPtr - Buffer.begin();
603  }
604 
605  do {
606  TheLexer.LexFromRawLexer(TheTok);
607 
608  if (InPreprocessorDirective) {
609  // If we've hit the end of the file, we're done.
610  if (TheTok.getKind() == tok::eof) {
611  break;
612  }
613 
614  // If we haven't hit the end of the preprocessor directive, skip this
615  // token.
616  if (!TheTok.isAtStartOfLine())
617  continue;
618 
619  // We've passed the end of the preprocessor directive, and will look
620  // at this token again below.
621  InPreprocessorDirective = false;
622  }
623 
624  // Keep track of the # of lines in the preamble.
625  if (TheTok.isAtStartOfLine()) {
626  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
627 
628  // If we were asked to limit the number of lines in the preamble,
629  // and we're about to exceed that limit, we're done.
630  if (MaxLineOffset && TokOffset >= MaxLineOffset)
631  break;
632  }
633 
634  // Comments are okay; skip over them.
635  if (TheTok.getKind() == tok::comment) {
636  if (ActiveCommentLoc.isInvalid())
637  ActiveCommentLoc = TheTok.getLocation();
638  continue;
639  }
640 
641  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
642  // This is the start of a preprocessor directive.
643  Token HashTok = TheTok;
644  InPreprocessorDirective = true;
645  ActiveCommentLoc = SourceLocation();
646 
647  // Figure out which directive this is. Since we're lexing raw tokens,
648  // we don't have an identifier table available. Instead, just look at
649  // the raw identifier to recognize and categorize preprocessor directives.
650  TheLexer.LexFromRawLexer(TheTok);
651  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
652  StringRef Keyword = TheTok.getRawIdentifier();
654  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
655  .Case("include", PDK_Skipped)
656  .Case("__include_macros", PDK_Skipped)
657  .Case("define", PDK_Skipped)
658  .Case("undef", PDK_Skipped)
659  .Case("line", PDK_Skipped)
660  .Case("error", PDK_Skipped)
661  .Case("pragma", PDK_Skipped)
662  .Case("import", PDK_Skipped)
663  .Case("include_next", PDK_Skipped)
664  .Case("warning", PDK_Skipped)
665  .Case("ident", PDK_Skipped)
666  .Case("sccs", PDK_Skipped)
667  .Case("assert", PDK_Skipped)
668  .Case("unassert", PDK_Skipped)
669  .Case("if", PDK_Skipped)
670  .Case("ifdef", PDK_Skipped)
671  .Case("ifndef", PDK_Skipped)
672  .Case("elif", PDK_Skipped)
673  .Case("else", PDK_Skipped)
674  .Case("endif", PDK_Skipped)
675  .Default(PDK_Unknown);
676 
677  switch (PDK) {
678  case PDK_Skipped:
679  continue;
680 
681  case PDK_Unknown:
682  // We don't know what this directive is; stop at the '#'.
683  break;
684  }
685  }
686 
687  // We only end up here if we didn't recognize the preprocessor
688  // directive or it was one that can't occur in the preamble at this
689  // point. Roll back the current token to the location of the '#'.
690  InPreprocessorDirective = false;
691  TheTok = HashTok;
692  }
693 
694  // We hit a token that we don't recognize as being in the
695  // "preprocessing only" part of the file, so we're no longer in
696  // the preamble.
697  break;
698  } while (true);
699 
701  if (ActiveCommentLoc.isValid())
702  End = ActiveCommentLoc; // don't truncate a decl comment.
703  else
704  End = TheTok.getLocation();
705 
706  return PreambleBounds(End.getRawEncoding() - FileLoc.getRawEncoding(),
707  TheTok.isAtStartOfLine());
708 }
709 
710 unsigned Lexer::getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo,
711  const SourceManager &SM,
712  const LangOptions &LangOpts) {
713  // Figure out how many physical characters away the specified expansion
714  // character is. This needs to take into consideration newlines and
715  // trigraphs.
716  bool Invalid = false;
717  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
718 
719  // If they request the first char of the token, we're trivially done.
720  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
721  return 0;
722 
723  unsigned PhysOffset = 0;
724 
725  // The usual case is that tokens don't contain anything interesting. Skip
726  // over the uninteresting characters. If a token only consists of simple
727  // chars, this method is extremely fast.
728  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
729  if (CharNo == 0)
730  return PhysOffset;
731  ++TokPtr;
732  --CharNo;
733  ++PhysOffset;
734  }
735 
736  // If we have a character that may be a trigraph or escaped newline, use a
737  // lexer to parse it correctly.
738  for (; CharNo; --CharNo) {
739  unsigned Size;
740  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
741  TokPtr += Size;
742  PhysOffset += Size;
743  }
744 
745  // Final detail: if we end up on an escaped newline, we want to return the
746  // location of the actual byte of the token. For example foo<newline>bar
747  // advanced by 3 should return the location of b, not of \\. One compounding
748  // detail of this is that the escape may be made by a trigraph.
749  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
750  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
751 
752  return PhysOffset;
753 }
754 
755 /// Computes the source location just past the end of the
756 /// token at this source location.
757 ///
758 /// This routine can be used to produce a source location that
759 /// points just past the end of the token referenced by \p Loc, and
760 /// is generally used when a diagnostic needs to point just after a
761 /// token where it expected something different that it received. If
762 /// the returned source location would not be meaningful (e.g., if
763 /// it points into a macro), this routine returns an invalid
764 /// source location.
765 ///
766 /// \param Offset an offset from the end of the token, where the source
767 /// location should refer to. The default offset (0) produces a source
768 /// location pointing just past the end of the token; an offset of 1 produces
769 /// a source location pointing to the last character in the token, etc.
771  const SourceManager &SM,
772  const LangOptions &LangOpts) {
773  if (Loc.isInvalid())
774  return {};
775 
776  if (Loc.isMacroID()) {
777  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
778  return {}; // Points inside the macro expansion.
779  }
780 
781  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
782  if (Len > Offset)
783  Len = Len - Offset;
784  else
785  return Loc;
786 
787  return Loc.getLocWithOffset(Len);
788 }
789 
790 /// Returns true if the given MacroID location points at the first
791 /// token of the macro expansion.
793  const SourceManager &SM,
794  const LangOptions &LangOpts,
795  SourceLocation *MacroBegin) {
796  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
797 
798  SourceLocation expansionLoc;
799  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
800  return false;
801 
802  if (expansionLoc.isFileID()) {
803  // No other macro expansions, this is the first.
804  if (MacroBegin)
805  *MacroBegin = expansionLoc;
806  return true;
807  }
808 
809  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
810 }
811 
812 /// Returns true if the given MacroID location points at the last
813 /// token of the macro expansion.
815  const SourceManager &SM,
816  const LangOptions &LangOpts,
817  SourceLocation *MacroEnd) {
818  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
819 
820  SourceLocation spellLoc = SM.getSpellingLoc(loc);
821  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
822  if (tokLen == 0)
823  return false;
824 
825  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
826  SourceLocation expansionLoc;
827  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
828  return false;
829 
830  if (expansionLoc.isFileID()) {
831  // No other macro expansions.
832  if (MacroEnd)
833  *MacroEnd = expansionLoc;
834  return true;
835  }
836 
837  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
838 }
839 
841  const SourceManager &SM,
842  const LangOptions &LangOpts) {
843  SourceLocation Begin = Range.getBegin();
844  SourceLocation End = Range.getEnd();
845  assert(Begin.isFileID() && End.isFileID());
846  if (Range.isTokenRange()) {
847  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
848  if (End.isInvalid())
849  return {};
850  }
851 
852  // Break down the source locations.
853  FileID FID;
854  unsigned BeginOffs;
855  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
856  if (FID.isInvalid())
857  return {};
858 
859  unsigned EndOffs;
860  if (!SM.isInFileID(End, FID, &EndOffs) ||
861  BeginOffs > EndOffs)
862  return {};
863 
864  return CharSourceRange::getCharRange(Begin, End);
865 }
866 
868  const SourceManager &SM,
869  const LangOptions &LangOpts) {
870  SourceLocation Begin = Range.getBegin();
871  SourceLocation End = Range.getEnd();
872  if (Begin.isInvalid() || End.isInvalid())
873  return {};
874 
875  if (Begin.isFileID() && End.isFileID())
876  return makeRangeFromFileLocs(Range, SM, LangOpts);
877 
878  if (Begin.isMacroID() && End.isFileID()) {
879  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
880  return {};
881  Range.setBegin(Begin);
882  return makeRangeFromFileLocs(Range, SM, LangOpts);
883  }
884 
885  if (Begin.isFileID() && End.isMacroID()) {
886  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
887  &End)) ||
888  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
889  &End)))
890  return {};
891  Range.setEnd(End);
892  return makeRangeFromFileLocs(Range, SM, LangOpts);
893  }
894 
895  assert(Begin.isMacroID() && End.isMacroID());
896  SourceLocation MacroBegin, MacroEnd;
897  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
898  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
899  &MacroEnd)) ||
900  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
901  &MacroEnd)))) {
902  Range.setBegin(MacroBegin);
903  Range.setEnd(MacroEnd);
904  return makeRangeFromFileLocs(Range, SM, LangOpts);
905  }
906 
907  bool Invalid = false;
908  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
909  &Invalid);
910  if (Invalid)
911  return {};
912 
913  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
914  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
915  &Invalid);
916  if (Invalid)
917  return {};
918 
919  if (EndEntry.getExpansion().isMacroArgExpansion() &&
920  BeginEntry.getExpansion().getExpansionLocStart() ==
921  EndEntry.getExpansion().getExpansionLocStart()) {
922  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
923  Range.setEnd(SM.getImmediateSpellingLoc(End));
924  return makeFileCharRange(Range, SM, LangOpts);
925  }
926  }
927 
928  return {};
929 }
930 
932  const SourceManager &SM,
933  const LangOptions &LangOpts,
934  bool *Invalid) {
935  Range = makeFileCharRange(Range, SM, LangOpts);
936  if (Range.isInvalid()) {
937  if (Invalid) *Invalid = true;
938  return {};
939  }
940 
941  // Break down the source location.
942  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
943  if (beginInfo.first.isInvalid()) {
944  if (Invalid) *Invalid = true;
945  return {};
946  }
947 
948  unsigned EndOffs;
949  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
950  beginInfo.second > EndOffs) {
951  if (Invalid) *Invalid = true;
952  return {};
953  }
954 
955  // Try to the load the file buffer.
956  bool invalidTemp = false;
957  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
958  if (invalidTemp) {
959  if (Invalid) *Invalid = true;
960  return {};
961  }
962 
963  if (Invalid) *Invalid = false;
964  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
965 }
966 
968  const SourceManager &SM,
969  const LangOptions &LangOpts) {
970  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
971 
972  // Find the location of the immediate macro expansion.
973  while (true) {
974  FileID FID = SM.getFileID(Loc);
975  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
976  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
977  Loc = Expansion.getExpansionLocStart();
978  if (!Expansion.isMacroArgExpansion())
979  break;
980 
981  // For macro arguments we need to check that the argument did not come
982  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
983 
984  // Loc points to the argument id of the macro definition, move to the
985  // macro expansion.
986  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
987  SourceLocation SpellLoc = Expansion.getSpellingLoc();
988  if (SpellLoc.isFileID())
989  break; // No inner macro.
990 
991  // If spelling location resides in the same FileID as macro expansion
992  // location, it means there is no inner macro.
993  FileID MacroFID = SM.getFileID(Loc);
994  if (SM.isInFileID(SpellLoc, MacroFID))
995  break;
996 
997  // Argument came from inner macro.
998  Loc = SpellLoc;
999  }
1000 
1001  // Find the spelling location of the start of the non-argument expansion
1002  // range. This is where the macro name was spelled in order to begin
1003  // expanding this macro.
1004  Loc = SM.getSpellingLoc(Loc);
1005 
1006  // Dig out the buffer where the macro name was spelled and the extents of the
1007  // name so that we can render it into the expansion note.
1008  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1009  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1010  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1011  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1012 }
1013 
1015  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1016  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
1017  // Walk past macro argument expansions.
1018  while (SM.isMacroArgExpansion(Loc))
1019  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1020 
1021  // If the macro's spelling has no FileID, then it's actually a token paste
1022  // or stringization (or similar) and not a macro at all.
1023  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1024  return {};
1025 
1026  // Find the spelling location of the start of the non-argument expansion
1027  // range. This is where the macro name was spelled in order to begin
1028  // expanding this macro.
1030 
1031  // Dig out the buffer where the macro name was spelled and the extents of the
1032  // name so that we can render it into the expansion note.
1033  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1034  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1035  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1036  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1037 }
1038 
1039 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1040  return isIdentifierBody(c, LangOpts.DollarIdents);
1041 }
1042 
1043 bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1044  assert(isVerticalWhitespace(Str[0]));
1045  if (Str - 1 < BufferStart)
1046  return false;
1047 
1048  if ((Str[0] == '\n' && Str[-1] == '\r') ||
1049  (Str[0] == '\r' && Str[-1] == '\n')) {
1050  if (Str - 2 < BufferStart)
1051  return false;
1052  --Str;
1053  }
1054  --Str;
1055 
1056  // Rewind to first non-space character:
1057  while (Str > BufferStart && isHorizontalWhitespace(*Str))
1058  --Str;
1059 
1060  return *Str == '\\';
1061 }
1062 
1064  const SourceManager &SM) {
1065  if (Loc.isInvalid() || Loc.isMacroID())
1066  return {};
1067  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1068  if (LocInfo.first.isInvalid())
1069  return {};
1070  bool Invalid = false;
1071  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1072  if (Invalid)
1073  return {};
1074  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1075  if (!Line)
1076  return {};
1077  StringRef Rest = Buffer.substr(Line - Buffer.data());
1078  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1079  return NumWhitespaceChars == StringRef::npos
1080  ? ""
1081  : Rest.take_front(NumWhitespaceChars);
1082 }
1083 
1084 //===----------------------------------------------------------------------===//
1085 // Diagnostics forwarding code.
1086 //===----------------------------------------------------------------------===//
1087 
1088 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1089 /// lexer buffer was all expanded at a single point, perform the mapping.
1090 /// This is currently only used for _Pragma implementation, so it is the slow
1091 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1092 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1093  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1095  SourceLocation FileLoc,
1096  unsigned CharNo, unsigned TokLen) {
1097  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1098 
1099  // Otherwise, we're lexing "mapped tokens". This is used for things like
1100  // _Pragma handling. Combine the expansion location of FileLoc with the
1101  // spelling location.
1103 
1104  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1105  // characters come from spelling(FileLoc)+Offset.
1106  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1107  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1108 
1109  // Figure out the expansion loc range, which is the range covered by the
1110  // original _Pragma(...) sequence.
1111  CharSourceRange II = SM.getImmediateExpansionRange(FileLoc);
1112 
1113  return SM.createExpansionLoc(SpellingLoc, II.getBegin(), II.getEnd(), TokLen);
1114 }
1115 
1116 /// getSourceLocation - Return a source location identifier for the specified
1117 /// offset in the current file.
1119  unsigned TokLen) const {
1120  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1121  "Location out of range for this buffer!");
1122 
1123  // In the normal case, we're just lexing from a simple file buffer, return
1124  // the file id from FileLoc with the offset specified.
1125  unsigned CharNo = Loc-BufferStart;
1126  if (FileLoc.isFileID())
1127  return FileLoc.getLocWithOffset(CharNo);
1128 
1129  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1130  // tokens are lexed from where the _Pragma was defined.
1131  assert(PP && "This doesn't work on raw lexers");
1132  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1133 }
1134 
1135 /// Diag - Forwarding function for diagnostics. This translate a source
1136 /// position in the current buffer into a SourceLocation object for rendering.
1137 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1138  return PP->Diag(getSourceLocation(Loc), DiagID);
1139 }
1140 
1141 //===----------------------------------------------------------------------===//
1142 // Trigraph and Escaped Newline Handling Code.
1143 //===----------------------------------------------------------------------===//
1144 
1145 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1146 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1147 static char GetTrigraphCharForLetter(char Letter) {
1148  switch (Letter) {
1149  default: return 0;
1150  case '=': return '#';
1151  case ')': return ']';
1152  case '(': return '[';
1153  case '!': return '|';
1154  case '\'': return '^';
1155  case '>': return '}';
1156  case '/': return '\\';
1157  case '<': return '{';
1158  case '-': return '~';
1159  }
1160 }
1161 
1162 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1163 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1164 /// return the result character. Finally, emit a warning about trigraph use
1165 /// whether trigraphs are enabled or not.
1166 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1167  char Res = GetTrigraphCharForLetter(*CP);
1168  if (!Res || !L) return Res;
1169 
1170  if (!L->getLangOpts().Trigraphs) {
1171  if (!L->isLexingRawMode())
1172  L->Diag(CP-2, diag::trigraph_ignored);
1173  return 0;
1174  }
1175 
1176  if (!L->isLexingRawMode())
1177  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1178  return Res;
1179 }
1180 
1181 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1182 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1183 /// trigraph equivalent on entry to this function.
1184 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1185  unsigned Size = 0;
1186  while (isWhitespace(Ptr[Size])) {
1187  ++Size;
1188 
1189  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1190  continue;
1191 
1192  // If this is a \r\n or \n\r, skip the other half.
1193  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1194  Ptr[Size-1] != Ptr[Size])
1195  ++Size;
1196 
1197  return Size;
1198  }
1199 
1200  // Not an escaped newline, must be a \t or something else.
1201  return 0;
1202 }
1203 
1204 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1205 /// them), skip over them and return the first non-escaped-newline found,
1206 /// otherwise return P.
1207 const char *Lexer::SkipEscapedNewLines(const char *P) {
1208  while (true) {
1209  const char *AfterEscape;
1210  if (*P == '\\') {
1211  AfterEscape = P+1;
1212  } else if (*P == '?') {
1213  // If not a trigraph for escape, bail out.
1214  if (P[1] != '?' || P[2] != '/')
1215  return P;
1216  // FIXME: Take LangOpts into account; the language might not
1217  // support trigraphs.
1218  AfterEscape = P+3;
1219  } else {
1220  return P;
1221  }
1222 
1223  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1224  if (NewLineSize == 0) return P;
1225  P = AfterEscape+NewLineSize;
1226  }
1227 }
1228 
1230  const SourceManager &SM,
1231  const LangOptions &LangOpts) {
1232  if (Loc.isMacroID()) {
1233  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1234  return None;
1235  }
1236  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1237 
1238  // Break down the source location.
1239  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1240 
1241  // Try to load the file buffer.
1242  bool InvalidTemp = false;
1243  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1244  if (InvalidTemp)
1245  return None;
1246 
1247  const char *TokenBegin = File.data() + LocInfo.second;
1248 
1249  // Lex from the start of the given location.
1250  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1251  TokenBegin, File.end());
1252  // Find the token.
1253  Token Tok;
1254  lexer.LexFromRawLexer(Tok);
1255  return Tok;
1256 }
1257 
1258 /// Checks that the given token is the first token that occurs after the
1259 /// given location (this excludes comments and whitespace). Returns the location
1260 /// immediately after the specified token. If the token is not found or the
1261 /// location is inside a macro, the returned source location will be invalid.
1263  SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1264  const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1265  Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1266  if (!Tok || Tok->isNot(TKind))
1267  return {};
1268  SourceLocation TokenLoc = Tok->getLocation();
1269 
1270  // Calculate how much whitespace needs to be skipped if any.
1271  unsigned NumWhitespaceChars = 0;
1272  if (SkipTrailingWhitespaceAndNewLine) {
1273  const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1274  unsigned char C = *TokenEnd;
1275  while (isHorizontalWhitespace(C)) {
1276  C = *(++TokenEnd);
1277  NumWhitespaceChars++;
1278  }
1279 
1280  // Skip \r, \n, \r\n, or \n\r
1281  if (C == '\n' || C == '\r') {
1282  char PrevC = C;
1283  C = *(++TokenEnd);
1284  NumWhitespaceChars++;
1285  if ((C == '\n' || C == '\r') && C != PrevC)
1286  NumWhitespaceChars++;
1287  }
1288  }
1289 
1290  return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1291 }
1292 
1293 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1294 /// get its size, and return it. This is tricky in several cases:
1295 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1296 /// then either return the trigraph (skipping 3 chars) or the '?',
1297 /// depending on whether trigraphs are enabled or not.
1298 /// 2. If this is an escaped newline (potentially with whitespace between
1299 /// the backslash and newline), implicitly skip the newline and return
1300 /// the char after it.
1301 ///
1302 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1303 /// know that we can accumulate into Size, and that we have already incremented
1304 /// Ptr by Size bytes.
1305 ///
1306 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1307 /// be updated to match.
1308 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1309  Token *Tok) {
1310  // If we have a slash, look for an escaped newline.
1311  if (Ptr[0] == '\\') {
1312  ++Size;
1313  ++Ptr;
1314 Slash:
1315  // Common case, backslash-char where the char is not whitespace.
1316  if (!isWhitespace(Ptr[0])) return '\\';
1317 
1318  // See if we have optional whitespace characters between the slash and
1319  // newline.
1320  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1321  // Remember that this token needs to be cleaned.
1322  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1323 
1324  // Warn if there was whitespace between the backslash and newline.
1325  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1326  Diag(Ptr, diag::backslash_newline_space);
1327 
1328  // Found backslash<whitespace><newline>. Parse the char after it.
1329  Size += EscapedNewLineSize;
1330  Ptr += EscapedNewLineSize;
1331 
1332  // Use slow version to accumulate a correct size field.
1333  return getCharAndSizeSlow(Ptr, Size, Tok);
1334  }
1335 
1336  // Otherwise, this is not an escaped newline, just return the slash.
1337  return '\\';
1338  }
1339 
1340  // If this is a trigraph, process it.
1341  if (Ptr[0] == '?' && Ptr[1] == '?') {
1342  // If this is actually a legal trigraph (not something like "??x"), emit
1343  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1344  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1345  // Remember that this token needs to be cleaned.
1346  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1347 
1348  Ptr += 3;
1349  Size += 3;
1350  if (C == '\\') goto Slash;
1351  return C;
1352  }
1353  }
1354 
1355  // If this is neither, return a single character.
1356  ++Size;
1357  return *Ptr;
1358 }
1359 
1360 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1361 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1362 /// and that we have already incremented Ptr by Size bytes.
1363 ///
1364 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1365 /// be updated to match.
1366 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1367  const LangOptions &LangOpts) {
1368  // If we have a slash, look for an escaped newline.
1369  if (Ptr[0] == '\\') {
1370  ++Size;
1371  ++Ptr;
1372 Slash:
1373  // Common case, backslash-char where the char is not whitespace.
1374  if (!isWhitespace(Ptr[0])) return '\\';
1375 
1376  // See if we have optional whitespace characters followed by a newline.
1377  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1378  // Found backslash<whitespace><newline>. Parse the char after it.
1379  Size += EscapedNewLineSize;
1380  Ptr += EscapedNewLineSize;
1381 
1382  // Use slow version to accumulate a correct size field.
1383  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1384  }
1385 
1386  // Otherwise, this is not an escaped newline, just return the slash.
1387  return '\\';
1388  }
1389 
1390  // If this is a trigraph, process it.
1391  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1392  // If this is actually a legal trigraph (not something like "??x"), return
1393  // it.
1394  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1395  Ptr += 3;
1396  Size += 3;
1397  if (C == '\\') goto Slash;
1398  return C;
1399  }
1400  }
1401 
1402  // If this is neither, return a single character.
1403  ++Size;
1404  return *Ptr;
1405 }
1406 
1407 //===----------------------------------------------------------------------===//
1408 // Helper methods for lexing.
1409 //===----------------------------------------------------------------------===//
1410 
1411 /// Routine that indiscriminately sets the offset into the source file.
1412 void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1413  BufferPtr = BufferStart + Offset;
1414  if (BufferPtr > BufferEnd)
1415  BufferPtr = BufferEnd;
1416  // FIXME: What exactly does the StartOfLine bit mean? There are two
1417  // possible meanings for the "start" of the line: the first token on the
1418  // unexpanded line, or the first token on the expanded line.
1419  IsAtStartOfLine = StartOfLine;
1420  IsAtPhysicalStartOfLine = StartOfLine;
1421 }
1422 
1423 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1424  if (LangOpts.AsmPreprocessor) {
1425  return false;
1426  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1427  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1429  return C11AllowedIDChars.contains(C);
1430  } else if (LangOpts.CPlusPlus) {
1431  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1433  return CXX03AllowedIDChars.contains(C);
1434  } else {
1435  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1437  return C99AllowedIDChars.contains(C);
1438  }
1439 }
1440 
1441 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1442  assert(isAllowedIDChar(C, LangOpts));
1443  if (LangOpts.AsmPreprocessor) {
1444  return false;
1445  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1446  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1448  return !C11DisallowedInitialIDChars.contains(C);
1449  } else if (LangOpts.CPlusPlus) {
1450  return true;
1451  } else {
1452  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1454  return !C99DisallowedInitialIDChars.contains(C);
1455  }
1456 }
1457 
1458 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1459  const char *End) {
1461  L.getSourceLocation(End));
1462 }
1463 
1464 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1465  CharSourceRange Range, bool IsFirst) {
1466  // Check C99 compatibility.
1467  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1468  enum {
1469  CannotAppearInIdentifier = 0,
1470  CannotStartIdentifier
1471  };
1472 
1473  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1475  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1477  if (!C99AllowedIDChars.contains(C)) {
1478  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1479  << Range
1480  << CannotAppearInIdentifier;
1481  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1482  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1483  << Range
1484  << CannotStartIdentifier;
1485  }
1486  }
1487 
1488  // Check C++98 compatibility.
1489  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1490  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1492  if (!CXX03AllowedIDChars.contains(C)) {
1493  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1494  << Range;
1495  }
1496  }
1497 }
1498 
1499 /// After encountering UTF-8 character C and interpreting it as an identifier
1500 /// character, check whether it's a homoglyph for a common non-identifier
1501 /// source character that is unlikely to be an intentional identifier
1502 /// character and warn if so.
1503 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1504  CharSourceRange Range) {
1505  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1506  struct HomoglyphPair {
1507  uint32_t Character;
1508  char LooksLike;
1509  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1510  };
1511  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1512  {U'\u00ad', 0}, // SOFT HYPHEN
1513  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1514  {U'\u037e', ';'}, // GREEK QUESTION MARK
1515  {U'\u200b', 0}, // ZERO WIDTH SPACE
1516  {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1517  {U'\u200d', 0}, // ZERO WIDTH JOINER
1518  {U'\u2060', 0}, // WORD JOINER
1519  {U'\u2061', 0}, // FUNCTION APPLICATION
1520  {U'\u2062', 0}, // INVISIBLE TIMES
1521  {U'\u2063', 0}, // INVISIBLE SEPARATOR
1522  {U'\u2064', 0}, // INVISIBLE PLUS
1523  {U'\u2212', '-'}, // MINUS SIGN
1524  {U'\u2215', '/'}, // DIVISION SLASH
1525  {U'\u2216', '\\'}, // SET MINUS
1526  {U'\u2217', '*'}, // ASTERISK OPERATOR
1527  {U'\u2223', '|'}, // DIVIDES
1528  {U'\u2227', '^'}, // LOGICAL AND
1529  {U'\u2236', ':'}, // RATIO
1530  {U'\u223c', '~'}, // TILDE OPERATOR
1531  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1532  {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1533  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1534  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1535  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1536  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1537  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1538  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1539  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1540  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1541  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1542  {U'\uff0c', ','}, // FULLWIDTH COMMA
1543  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1544  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1545  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1546  {U'\uff1a', ':'}, // FULLWIDTH COLON
1547  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1548  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1549  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1550  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1551  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1552  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1553  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1554  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1555  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1556  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1557  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1558  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1559  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1560  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1561  {0, 0}
1562  };
1563  auto Homoglyph =
1564  std::lower_bound(std::begin(SortedHomoglyphs),
1565  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1566  if (Homoglyph->Character == C) {
1567  llvm::SmallString<5> CharBuf;
1568  {
1569  llvm::raw_svector_ostream CharOS(CharBuf);
1570  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1571  }
1572  if (Homoglyph->LooksLike) {
1573  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1574  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1575  << Range << CharBuf << LooksLikeStr;
1576  } else {
1577  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1578  << Range << CharBuf;
1579  }
1580  }
1581 }
1582 
1583 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1584  Token &Result) {
1585  const char *UCNPtr = CurPtr + Size;
1586  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1587  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1588  return false;
1589 
1590  if (!isLexingRawMode())
1591  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1592  makeCharRange(*this, CurPtr, UCNPtr),
1593  /*IsFirst=*/false);
1594 
1595  Result.setFlag(Token::HasUCN);
1596  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1597  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1598  CurPtr = UCNPtr;
1599  else
1600  while (CurPtr != UCNPtr)
1601  (void)getAndAdvanceChar(CurPtr, Result);
1602  return true;
1603 }
1604 
1605 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1606  const char *UnicodePtr = CurPtr;
1607  llvm::UTF32 CodePoint;
1608  llvm::ConversionResult Result =
1609  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1610  (const llvm::UTF8 *)BufferEnd,
1611  &CodePoint,
1612  llvm::strictConversion);
1613  if (Result != llvm::conversionOK ||
1614  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1615  return false;
1616 
1617  if (!isLexingRawMode()) {
1618  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1619  makeCharRange(*this, CurPtr, UnicodePtr),
1620  /*IsFirst=*/false);
1621  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1622  makeCharRange(*this, CurPtr, UnicodePtr));
1623  }
1624 
1625  CurPtr = UnicodePtr;
1626  return true;
1627 }
1628 
1629 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1630  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1631  unsigned Size;
1632  unsigned char C = *CurPtr++;
1633  while (isIdentifierBody(C))
1634  C = *CurPtr++;
1635 
1636  --CurPtr; // Back up over the skipped character.
1637 
1638  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1639  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1640  //
1641  // TODO: Could merge these checks into an InfoTable flag to make the
1642  // comparison cheaper
1643  if (isASCII(C) && C != '\\' && C != '?' &&
1644  (C != '$' || !LangOpts.DollarIdents)) {
1645 FinishIdentifier:
1646  const char *IdStart = BufferPtr;
1647  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1648  Result.setRawIdentifierData(IdStart);
1649 
1650  // If we are in raw mode, return this identifier raw. There is no need to
1651  // look up identifier information or attempt to macro expand it.
1652  if (LexingRawMode)
1653  return true;
1654 
1655  // Fill in Result.IdentifierInfo and update the token kind,
1656  // looking up the identifier in the identifier table.
1657  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1658  // Note that we have to call PP->LookUpIdentifierInfo() even for code
1659  // completion, it writes IdentifierInfo into Result, and callers rely on it.
1660 
1661  // If the completion point is at the end of an identifier, we want to treat
1662  // the identifier as incomplete even if it resolves to a macro or a keyword.
1663  // This allows e.g. 'class^' to complete to 'classifier'.
1664  if (isCodeCompletionPoint(CurPtr)) {
1665  // Return the code-completion token.
1666  Result.setKind(tok::code_completion);
1667  // Skip the code-completion char and all immediate identifier characters.
1668  // This ensures we get consistent behavior when completing at any point in
1669  // an identifier (i.e. at the start, in the middle, at the end). Note that
1670  // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1671  // simpler.
1672  assert(*CurPtr == 0 && "Completion character must be 0");
1673  ++CurPtr;
1674  // Note that code completion token is not added as a separate character
1675  // when the completion point is at the end of the buffer. Therefore, we need
1676  // to check if the buffer has ended.
1677  if (CurPtr < BufferEnd) {
1678  while (isIdentifierBody(*CurPtr))
1679  ++CurPtr;
1680  }
1681  BufferPtr = CurPtr;
1682  return true;
1683  }
1684 
1685  // Finally, now that we know we have an identifier, pass this off to the
1686  // preprocessor, which may macro expand it or something.
1687  if (II->isHandleIdentifierCase())
1688  return PP->HandleIdentifier(Result);
1689 
1690  return true;
1691  }
1692 
1693  // Otherwise, $,\,? in identifier found. Enter slower path.
1694 
1695  C = getCharAndSize(CurPtr, Size);
1696  while (true) {
1697  if (C == '$') {
1698  // If we hit a $ and they are not supported in identifiers, we are done.
1699  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1700 
1701  // Otherwise, emit a diagnostic and continue.
1702  if (!isLexingRawMode())
1703  Diag(CurPtr, diag::ext_dollar_in_identifier);
1704  CurPtr = ConsumeChar(CurPtr, Size, Result);
1705  C = getCharAndSize(CurPtr, Size);
1706  continue;
1707  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1708  C = getCharAndSize(CurPtr, Size);
1709  continue;
1710  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1711  C = getCharAndSize(CurPtr, Size);
1712  continue;
1713  } else if (!isIdentifierBody(C)) {
1714  goto FinishIdentifier;
1715  }
1716 
1717  // Otherwise, this character is good, consume it.
1718  CurPtr = ConsumeChar(CurPtr, Size, Result);
1719 
1720  C = getCharAndSize(CurPtr, Size);
1721  while (isIdentifierBody(C)) {
1722  CurPtr = ConsumeChar(CurPtr, Size, Result);
1723  C = getCharAndSize(CurPtr, Size);
1724  }
1725  }
1726 }
1727 
1728 /// isHexaLiteral - Return true if Start points to a hex constant.
1729 /// in microsoft mode (where this is supposed to be several different tokens).
1730 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1731  unsigned Size;
1732  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1733  if (C1 != '0')
1734  return false;
1735  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1736  return (C2 == 'x' || C2 == 'X');
1737 }
1738 
1739 /// LexNumericConstant - Lex the remainder of a integer or floating point
1740 /// constant. From[-1] is the first character lexed. Return the end of the
1741 /// constant.
1742 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1743  unsigned Size;
1744  char C = getCharAndSize(CurPtr, Size);
1745  char PrevCh = 0;
1746  while (isPreprocessingNumberBody(C)) {
1747  CurPtr = ConsumeChar(CurPtr, Size, Result);
1748  PrevCh = C;
1749  C = getCharAndSize(CurPtr, Size);
1750  }
1751 
1752  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1753  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1754  // If we are in Microsoft mode, don't continue if the constant is hex.
1755  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1756  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1757  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1758  }
1759 
1760  // If we have a hex FP constant, continue.
1761  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1762  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1763  // not-quite-conforming extension. Only do so if this looks like it's
1764  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1765  bool IsHexFloat = true;
1766  if (!LangOpts.C99) {
1767  if (!isHexaLiteral(BufferPtr, LangOpts))
1768  IsHexFloat = false;
1769  else if (!getLangOpts().CPlusPlus17 &&
1770  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1771  IsHexFloat = false;
1772  }
1773  if (IsHexFloat)
1774  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1775  }
1776 
1777  // If we have a digit separator, continue.
1778  if (C == '\'' && getLangOpts().CPlusPlus14) {
1779  unsigned NextSize;
1780  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1781  if (isIdentifierBody(Next)) {
1782  if (!isLexingRawMode())
1783  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1784  CurPtr = ConsumeChar(CurPtr, Size, Result);
1785  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1786  return LexNumericConstant(Result, CurPtr);
1787  }
1788  }
1789 
1790  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1791  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1792  return LexNumericConstant(Result, CurPtr);
1793  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1794  return LexNumericConstant(Result, CurPtr);
1795 
1796  // Update the location of token as well as BufferPtr.
1797  const char *TokStart = BufferPtr;
1798  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1799  Result.setLiteralData(TokStart);
1800  return true;
1801 }
1802 
1803 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1804 /// in C++11, or warn on a ud-suffix in C++98.
1805 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1806  bool IsStringLiteral) {
1807  assert(getLangOpts().CPlusPlus);
1808 
1809  // Maximally munch an identifier.
1810  unsigned Size;
1811  char C = getCharAndSize(CurPtr, Size);
1812  bool Consumed = false;
1813 
1814  if (!isIdentifierHead(C)) {
1815  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1816  Consumed = true;
1817  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1818  Consumed = true;
1819  else
1820  return CurPtr;
1821  }
1822 
1823  if (!getLangOpts().CPlusPlus11) {
1824  if (!isLexingRawMode())
1825  Diag(CurPtr,
1826  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1827  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1829  return CurPtr;
1830  }
1831 
1832  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1833  // that does not start with an underscore is ill-formed. As a conforming
1834  // extension, we treat all such suffixes as if they had whitespace before
1835  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1836  // likely to be a ud-suffix than a macro, however, and accept that.
1837  if (!Consumed) {
1838  bool IsUDSuffix = false;
1839  if (C == '_')
1840  IsUDSuffix = true;
1841  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1842  // In C++1y, we need to look ahead a few characters to see if this is a
1843  // valid suffix for a string literal or a numeric literal (this could be
1844  // the 'operator""if' defining a numeric literal operator).
1845  const unsigned MaxStandardSuffixLength = 3;
1846  char Buffer[MaxStandardSuffixLength] = { C };
1847  unsigned Consumed = Size;
1848  unsigned Chars = 1;
1849  while (true) {
1850  unsigned NextSize;
1851  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1852  getLangOpts());
1853  if (!isIdentifierBody(Next)) {
1854  // End of suffix. Check whether this is on the whitelist.
1855  const StringRef CompleteSuffix(Buffer, Chars);
1857  CompleteSuffix);
1858  break;
1859  }
1860 
1861  if (Chars == MaxStandardSuffixLength)
1862  // Too long: can't be a standard suffix.
1863  break;
1864 
1865  Buffer[Chars++] = Next;
1866  Consumed += NextSize;
1867  }
1868  }
1869 
1870  if (!IsUDSuffix) {
1871  if (!isLexingRawMode())
1872  Diag(CurPtr, getLangOpts().MSVCCompat
1873  ? diag::ext_ms_reserved_user_defined_literal
1874  : diag::ext_reserved_user_defined_literal)
1876  return CurPtr;
1877  }
1878 
1879  CurPtr = ConsumeChar(CurPtr, Size, Result);
1880  }
1881 
1882  Result.setFlag(Token::HasUDSuffix);
1883  while (true) {
1884  C = getCharAndSize(CurPtr, Size);
1885  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1886  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1887  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1888  else break;
1889  }
1890 
1891  return CurPtr;
1892 }
1893 
1894 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1895 /// either " or L" or u8" or u" or U".
1896 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1897  tok::TokenKind Kind) {
1898  const char *AfterQuote = CurPtr;
1899  // Does this string contain the \0 character?
1900  const char *NulCharacter = nullptr;
1901 
1902  if (!isLexingRawMode() &&
1903  (Kind == tok::utf8_string_literal ||
1904  Kind == tok::utf16_string_literal ||
1905  Kind == tok::utf32_string_literal))
1906  Diag(BufferPtr, getLangOpts().CPlusPlus
1907  ? diag::warn_cxx98_compat_unicode_literal
1908  : diag::warn_c99_compat_unicode_literal);
1909 
1910  char C = getAndAdvanceChar(CurPtr, Result);
1911  while (C != '"') {
1912  // Skip escaped characters. Escaped newlines will already be processed by
1913  // getAndAdvanceChar.
1914  if (C == '\\')
1915  C = getAndAdvanceChar(CurPtr, Result);
1916 
1917  if (C == '\n' || C == '\r' || // Newline.
1918  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1919  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1920  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1921  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1922  return true;
1923  }
1924 
1925  if (C == 0) {
1926  if (isCodeCompletionPoint(CurPtr-1)) {
1927  if (ParsingFilename)
1928  codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
1929  else
1931  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
1932  cutOffLexing();
1933  return true;
1934  }
1935 
1936  NulCharacter = CurPtr-1;
1937  }
1938  C = getAndAdvanceChar(CurPtr, Result);
1939  }
1940 
1941  // If we are in C++11, lex the optional ud-suffix.
1942  if (getLangOpts().CPlusPlus)
1943  CurPtr = LexUDSuffix(Result, CurPtr, true);
1944 
1945  // If a nul character existed in the string, warn about it.
1946  if (NulCharacter && !isLexingRawMode())
1947  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1948 
1949  // Update the location of the token as well as the BufferPtr instance var.
1950  const char *TokStart = BufferPtr;
1951  FormTokenWithChars(Result, CurPtr, Kind);
1952  Result.setLiteralData(TokStart);
1953  return true;
1954 }
1955 
1956 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1957 /// having lexed R", LR", u8R", uR", or UR".
1958 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1959  tok::TokenKind Kind) {
1960  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1961  // Between the initial and final double quote characters of the raw string,
1962  // any transformations performed in phases 1 and 2 (trigraphs,
1963  // universal-character-names, and line splicing) are reverted.
1964 
1965  if (!isLexingRawMode())
1966  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1967 
1968  unsigned PrefixLen = 0;
1969 
1970  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1971  ++PrefixLen;
1972 
1973  // If the last character was not a '(', then we didn't lex a valid delimiter.
1974  if (CurPtr[PrefixLen] != '(') {
1975  if (!isLexingRawMode()) {
1976  const char *PrefixEnd = &CurPtr[PrefixLen];
1977  if (PrefixLen == 16) {
1978  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1979  } else {
1980  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1981  << StringRef(PrefixEnd, 1);
1982  }
1983  }
1984 
1985  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1986  // it's possible the '"' was intended to be part of the raw string, but
1987  // there's not much we can do about that.
1988  while (true) {
1989  char C = *CurPtr++;
1990 
1991  if (C == '"')
1992  break;
1993  if (C == 0 && CurPtr-1 == BufferEnd) {
1994  --CurPtr;
1995  break;
1996  }
1997  }
1998 
1999  FormTokenWithChars(Result, CurPtr, tok::unknown);
2000  return true;
2001  }
2002 
2003  // Save prefix and move CurPtr past it
2004  const char *Prefix = CurPtr;
2005  CurPtr += PrefixLen + 1; // skip over prefix and '('
2006 
2007  while (true) {
2008  char C = *CurPtr++;
2009 
2010  if (C == ')') {
2011  // Check for prefix match and closing quote.
2012  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2013  CurPtr += PrefixLen + 1; // skip over prefix and '"'
2014  break;
2015  }
2016  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2017  if (!isLexingRawMode())
2018  Diag(BufferPtr, diag::err_unterminated_raw_string)
2019  << StringRef(Prefix, PrefixLen);
2020  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2021  return true;
2022  }
2023  }
2024 
2025  // If we are in C++11, lex the optional ud-suffix.
2026  if (getLangOpts().CPlusPlus)
2027  CurPtr = LexUDSuffix(Result, CurPtr, true);
2028 
2029  // Update the location of token as well as BufferPtr.
2030  const char *TokStart = BufferPtr;
2031  FormTokenWithChars(Result, CurPtr, Kind);
2032  Result.setLiteralData(TokStart);
2033  return true;
2034 }
2035 
2036 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2037 /// after having lexed the '<' character. This is used for #include filenames.
2038 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2039  // Does this string contain the \0 character?
2040  const char *NulCharacter = nullptr;
2041  const char *AfterLessPos = CurPtr;
2042  char C = getAndAdvanceChar(CurPtr, Result);
2043  while (C != '>') {
2044  // Skip escaped characters. Escaped newlines will already be processed by
2045  // getAndAdvanceChar.
2046  if (C == '\\')
2047  C = getAndAdvanceChar(CurPtr, Result);
2048 
2049  if (C == '\n' || C == '\r' || // Newline.
2050  (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2051  // If the filename is unterminated, then it must just be a lone <
2052  // character. Return this as such.
2053  FormTokenWithChars(Result, AfterLessPos, tok::less);
2054  return true;
2055  }
2056 
2057  if (C == 0) {
2058  if (isCodeCompletionPoint(CurPtr - 1)) {
2059  codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2060  cutOffLexing();
2061  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2062  return true;
2063  }
2064  NulCharacter = CurPtr-1;
2065  }
2066  C = getAndAdvanceChar(CurPtr, Result);
2067  }
2068 
2069  // If a nul character existed in the string, warn about it.
2070  if (NulCharacter && !isLexingRawMode())
2071  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2072 
2073  // Update the location of token as well as BufferPtr.
2074  const char *TokStart = BufferPtr;
2075  FormTokenWithChars(Result, CurPtr, tok::header_name);
2076  Result.setLiteralData(TokStart);
2077  return true;
2078 }
2079 
2080 void Lexer::codeCompleteIncludedFile(const char *PathStart,
2081  const char *CompletionPoint,
2082  bool IsAngled) {
2083  // Completion only applies to the filename, after the last slash.
2084  StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2085  auto Slash = PartialPath.find_last_of(LangOpts.MSVCCompat ? "/\\" : "/");
2086  StringRef Dir =
2087  (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2088  const char *StartOfFilename =
2089  (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2090  // Code completion filter range is the filename only, up to completion point.
2092  StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2093  // We should replace the characters up to the closing quote, if any.
2094  while (CompletionPoint < BufferEnd) {
2095  char Next = *(CompletionPoint + 1);
2096  if (Next == 0 || Next == '\r' || Next == '\n')
2097  break;
2098  ++CompletionPoint;
2099  if (Next == (IsAngled ? '>' : '"'))
2100  break;
2101  }
2103  FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2104  FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2105  PP->CodeCompleteIncludedFile(Dir, IsAngled);
2106 }
2107 
2108 /// LexCharConstant - Lex the remainder of a character constant, after having
2109 /// lexed either ' or L' or u8' or u' or U'.
2110 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2111  tok::TokenKind Kind) {
2112  // Does this character contain the \0 character?
2113  const char *NulCharacter = nullptr;
2114 
2115  if (!isLexingRawMode()) {
2116  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2117  Diag(BufferPtr, getLangOpts().CPlusPlus
2118  ? diag::warn_cxx98_compat_unicode_literal
2119  : diag::warn_c99_compat_unicode_literal);
2120  else if (Kind == tok::utf8_char_constant)
2121  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2122  }
2123 
2124  char C = getAndAdvanceChar(CurPtr, Result);
2125  if (C == '\'') {
2126  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2127  Diag(BufferPtr, diag::ext_empty_character);
2128  FormTokenWithChars(Result, CurPtr, tok::unknown);
2129  return true;
2130  }
2131 
2132  while (C != '\'') {
2133  // Skip escaped characters.
2134  if (C == '\\')
2135  C = getAndAdvanceChar(CurPtr, Result);
2136 
2137  if (C == '\n' || C == '\r' || // Newline.
2138  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2139  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2140  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2141  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2142  return true;
2143  }
2144 
2145  if (C == 0) {
2146  if (isCodeCompletionPoint(CurPtr-1)) {
2148  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2149  cutOffLexing();
2150  return true;
2151  }
2152 
2153  NulCharacter = CurPtr-1;
2154  }
2155  C = getAndAdvanceChar(CurPtr, Result);
2156  }
2157 
2158  // If we are in C++11, lex the optional ud-suffix.
2159  if (getLangOpts().CPlusPlus)
2160  CurPtr = LexUDSuffix(Result, CurPtr, false);
2161 
2162  // If a nul character existed in the character, warn about it.
2163  if (NulCharacter && !isLexingRawMode())
2164  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2165 
2166  // Update the location of token as well as BufferPtr.
2167  const char *TokStart = BufferPtr;
2168  FormTokenWithChars(Result, CurPtr, Kind);
2169  Result.setLiteralData(TokStart);
2170  return true;
2171 }
2172 
2173 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2174 /// Update BufferPtr to point to the next non-whitespace character and return.
2175 ///
2176 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2177 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2178  bool &TokAtPhysicalStartOfLine) {
2179  // Whitespace - Skip it, then return the token after the whitespace.
2180  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2181 
2182  unsigned char Char = *CurPtr;
2183 
2184  // Skip consecutive spaces efficiently.
2185  while (true) {
2186  // Skip horizontal whitespace very aggressively.
2187  while (isHorizontalWhitespace(Char))
2188  Char = *++CurPtr;
2189 
2190  // Otherwise if we have something other than whitespace, we're done.
2191  if (!isVerticalWhitespace(Char))
2192  break;
2193 
2195  // End of preprocessor directive line, let LexTokenInternal handle this.
2196  BufferPtr = CurPtr;
2197  return false;
2198  }
2199 
2200  // OK, but handle newline.
2201  SawNewline = true;
2202  Char = *++CurPtr;
2203  }
2204 
2205  // If the client wants us to return whitespace, return it now.
2206  if (isKeepWhitespaceMode()) {
2207  FormTokenWithChars(Result, CurPtr, tok::unknown);
2208  if (SawNewline) {
2209  IsAtStartOfLine = true;
2210  IsAtPhysicalStartOfLine = true;
2211  }
2212  // FIXME: The next token will not have LeadingSpace set.
2213  return true;
2214  }
2215 
2216  // If this isn't immediately after a newline, there is leading space.
2217  char PrevChar = CurPtr[-1];
2218  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2219 
2220  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2221  if (SawNewline) {
2222  Result.setFlag(Token::StartOfLine);
2223  TokAtPhysicalStartOfLine = true;
2224  }
2225 
2226  BufferPtr = CurPtr;
2227  return false;
2228 }
2229 
2230 /// We have just read the // characters from input. Skip until we find the
2231 /// newline character that terminates the comment. Then update BufferPtr and
2232 /// return.
2233 ///
2234 /// If we're in KeepCommentMode or any CommentHandler has inserted
2235 /// some tokens, this will store the first token and return true.
2236 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2237  bool &TokAtPhysicalStartOfLine) {
2238  // If Line comments aren't explicitly enabled for this language, emit an
2239  // extension warning.
2240  if (!LangOpts.LineComment && !isLexingRawMode()) {
2241  Diag(BufferPtr, diag::ext_line_comment);
2242 
2243  // Mark them enabled so we only emit one warning for this translation
2244  // unit.
2245  LangOpts.LineComment = true;
2246  }
2247 
2248  // Scan over the body of the comment. The common case, when scanning, is that
2249  // the comment contains normal ascii characters with nothing interesting in
2250  // them. As such, optimize for this case with the inner loop.
2251  //
2252  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2253  // character that ends the line comment.
2254  char C;
2255  while (true) {
2256  C = *CurPtr;
2257  // Skip over characters in the fast loop.
2258  while (C != 0 && // Potentially EOF.
2259  C != '\n' && C != '\r') // Newline or DOS-style newline.
2260  C = *++CurPtr;
2261 
2262  const char *NextLine = CurPtr;
2263  if (C != 0) {
2264  // We found a newline, see if it's escaped.
2265  const char *EscapePtr = CurPtr-1;
2266  bool HasSpace = false;
2267  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2268  --EscapePtr;
2269  HasSpace = true;
2270  }
2271 
2272  if (*EscapePtr == '\\')
2273  // Escaped newline.
2274  CurPtr = EscapePtr;
2275  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2276  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2277  // Trigraph-escaped newline.
2278  CurPtr = EscapePtr-2;
2279  else
2280  break; // This is a newline, we're done.
2281 
2282  // If there was space between the backslash and newline, warn about it.
2283  if (HasSpace && !isLexingRawMode())
2284  Diag(EscapePtr, diag::backslash_newline_space);
2285  }
2286 
2287  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2288  // properly decode the character. Read it in raw mode to avoid emitting
2289  // diagnostics about things like trigraphs. If we see an escaped newline,
2290  // we'll handle it below.
2291  const char *OldPtr = CurPtr;
2292  bool OldRawMode = isLexingRawMode();
2293  LexingRawMode = true;
2294  C = getAndAdvanceChar(CurPtr, Result);
2295  LexingRawMode = OldRawMode;
2296 
2297  // If we only read only one character, then no special handling is needed.
2298  // We're done and can skip forward to the newline.
2299  if (C != 0 && CurPtr == OldPtr+1) {
2300  CurPtr = NextLine;
2301  break;
2302  }
2303 
2304  // If we read multiple characters, and one of those characters was a \r or
2305  // \n, then we had an escaped newline within the comment. Emit diagnostic
2306  // unless the next line is also a // comment.
2307  if (CurPtr != OldPtr + 1 && C != '/' &&
2308  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2309  for (; OldPtr != CurPtr; ++OldPtr)
2310  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2311  // Okay, we found a // comment that ends in a newline, if the next
2312  // line is also a // comment, but has spaces, don't emit a diagnostic.
2313  if (isWhitespace(C)) {
2314  const char *ForwardPtr = CurPtr;
2315  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2316  ++ForwardPtr;
2317  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2318  break;
2319  }
2320 
2321  if (!isLexingRawMode())
2322  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2323  break;
2324  }
2325  }
2326 
2327  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2328  --CurPtr;
2329  break;
2330  }
2331 
2332  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2334  cutOffLexing();
2335  return false;
2336  }
2337  }
2338 
2339  // Found but did not consume the newline. Notify comment handlers about the
2340  // comment unless we're in a #if 0 block.
2341  if (PP && !isLexingRawMode() &&
2342  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2343  getSourceLocation(CurPtr)))) {
2344  BufferPtr = CurPtr;
2345  return true; // A token has to be returned.
2346  }
2347 
2348  // If we are returning comments as tokens, return this comment as a token.
2349  if (inKeepCommentMode())
2350  return SaveLineComment(Result, CurPtr);
2351 
2352  // If we are inside a preprocessor directive and we see the end of line,
2353  // return immediately, so that the lexer can return this as an EOD token.
2354  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2355  BufferPtr = CurPtr;
2356  return false;
2357  }
2358 
2359  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2360  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2361  // contribute to another token), it isn't needed for correctness. Note that
2362  // this is ok even in KeepWhitespaceMode, because we would have returned the
2363  /// comment above in that mode.
2364  ++CurPtr;
2365 
2366  // The next returned token is at the start of the line.
2367  Result.setFlag(Token::StartOfLine);
2368  TokAtPhysicalStartOfLine = true;
2369  // No leading whitespace seen so far.
2371  BufferPtr = CurPtr;
2372  return false;
2373 }
2374 
2375 /// If in save-comment mode, package up this Line comment in an appropriate
2376 /// way and return it.
2377 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2378  // If we're not in a preprocessor directive, just return the // comment
2379  // directly.
2380  FormTokenWithChars(Result, CurPtr, tok::comment);
2381 
2383  return true;
2384 
2385  // If this Line-style comment is in a macro definition, transmogrify it into
2386  // a C-style block comment.
2387  bool Invalid = false;
2388  std::string Spelling = PP->getSpelling(Result, &Invalid);
2389  if (Invalid)
2390  return true;
2391 
2392  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2393  Spelling[1] = '*'; // Change prefix to "/*".
2394  Spelling += "*/"; // add suffix.
2395 
2396  Result.setKind(tok::comment);
2397  PP->CreateString(Spelling, Result,
2398  Result.getLocation(), Result.getLocation());
2399  return true;
2400 }
2401 
2402 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2403 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2404 /// a diagnostic if so. We know that the newline is inside of a block comment.
2405 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2406  Lexer *L) {
2407  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2408 
2409  // Back up off the newline.
2410  --CurPtr;
2411 
2412  // If this is a two-character newline sequence, skip the other character.
2413  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2414  // \n\n or \r\r -> not escaped newline.
2415  if (CurPtr[0] == CurPtr[1])
2416  return false;
2417  // \n\r or \r\n -> skip the newline.
2418  --CurPtr;
2419  }
2420 
2421  // If we have horizontal whitespace, skip over it. We allow whitespace
2422  // between the slash and newline.
2423  bool HasSpace = false;
2424  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2425  --CurPtr;
2426  HasSpace = true;
2427  }
2428 
2429  // If we have a slash, we know this is an escaped newline.
2430  if (*CurPtr == '\\') {
2431  if (CurPtr[-1] != '*') return false;
2432  } else {
2433  // It isn't a slash, is it the ?? / trigraph?
2434  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2435  CurPtr[-3] != '*')
2436  return false;
2437 
2438  // This is the trigraph ending the comment. Emit a stern warning!
2439  CurPtr -= 2;
2440 
2441  // If no trigraphs are enabled, warn that we ignored this trigraph and
2442  // ignore this * character.
2443  if (!L->getLangOpts().Trigraphs) {
2444  if (!L->isLexingRawMode())
2445  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2446  return false;
2447  }
2448  if (!L->isLexingRawMode())
2449  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2450  }
2451 
2452  // Warn about having an escaped newline between the */ characters.
2453  if (!L->isLexingRawMode())
2454  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2455 
2456  // If there was space between the backslash and newline, warn about it.
2457  if (HasSpace && !L->isLexingRawMode())
2458  L->Diag(CurPtr, diag::backslash_newline_space);
2459 
2460  return true;
2461 }
2462 
2463 #ifdef __SSE2__
2464 #include <emmintrin.h>
2465 #elif __ALTIVEC__
2466 #include <altivec.h>
2467 #undef bool
2468 #endif
2469 
2470 /// We have just read from input the / and * characters that started a comment.
2471 /// Read until we find the * and / characters that terminate the comment.
2472 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2473 /// comments, because they cannot cause the comment to end. The only thing
2474 /// that can happen is the comment could end with an escaped newline between
2475 /// the terminating * and /.
2476 ///
2477 /// If we're in KeepCommentMode or any CommentHandler has inserted
2478 /// some tokens, this will store the first token and return true.
2479 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2480  bool &TokAtPhysicalStartOfLine) {
2481  // Scan one character past where we should, looking for a '/' character. Once
2482  // we find it, check to see if it was preceded by a *. This common
2483  // optimization helps people who like to put a lot of * characters in their
2484  // comments.
2485 
2486  // The first character we get with newlines and trigraphs skipped to handle
2487  // the degenerate /*/ case below correctly if the * has an escaped newline
2488  // after it.
2489  unsigned CharSize;
2490  unsigned char C = getCharAndSize(CurPtr, CharSize);
2491  CurPtr += CharSize;
2492  if (C == 0 && CurPtr == BufferEnd+1) {
2493  if (!isLexingRawMode())
2494  Diag(BufferPtr, diag::err_unterminated_block_comment);
2495  --CurPtr;
2496 
2497  // KeepWhitespaceMode should return this broken comment as a token. Since
2498  // it isn't a well formed comment, just return it as an 'unknown' token.
2499  if (isKeepWhitespaceMode()) {
2500  FormTokenWithChars(Result, CurPtr, tok::unknown);
2501  return true;
2502  }
2503 
2504  BufferPtr = CurPtr;
2505  return false;
2506  }
2507 
2508  // Check to see if the first character after the '/*' is another /. If so,
2509  // then this slash does not end the block comment, it is part of it.
2510  if (C == '/')
2511  C = *CurPtr++;
2512 
2513  while (true) {
2514  // Skip over all non-interesting characters until we find end of buffer or a
2515  // (probably ending) '/' character.
2516  if (CurPtr + 24 < BufferEnd &&
2517  // If there is a code-completion point avoid the fast scan because it
2518  // doesn't check for '\0'.
2519  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2520  // While not aligned to a 16-byte boundary.
2521  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2522  C = *CurPtr++;
2523 
2524  if (C == '/') goto FoundSlash;
2525 
2526 #ifdef __SSE2__
2527  __m128i Slashes = _mm_set1_epi8('/');
2528  while (CurPtr+16 <= BufferEnd) {
2529  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2530  Slashes));
2531  if (cmp != 0) {
2532  // Adjust the pointer to point directly after the first slash. It's
2533  // not necessary to set C here, it will be overwritten at the end of
2534  // the outer loop.
2535  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2536  goto FoundSlash;
2537  }
2538  CurPtr += 16;
2539  }
2540 #elif __ALTIVEC__
2541  __vector unsigned char Slashes = {
2542  '/', '/', '/', '/', '/', '/', '/', '/',
2543  '/', '/', '/', '/', '/', '/', '/', '/'
2544  };
2545  while (CurPtr+16 <= BufferEnd &&
2546  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2547  CurPtr += 16;
2548 #else
2549  // Scan for '/' quickly. Many block comments are very large.
2550  while (CurPtr[0] != '/' &&
2551  CurPtr[1] != '/' &&
2552  CurPtr[2] != '/' &&
2553  CurPtr[3] != '/' &&
2554  CurPtr+4 < BufferEnd) {
2555  CurPtr += 4;
2556  }
2557 #endif
2558 
2559  // It has to be one of the bytes scanned, increment to it and read one.
2560  C = *CurPtr++;
2561  }
2562 
2563  // Loop to scan the remainder.
2564  while (C != '/' && C != '\0')
2565  C = *CurPtr++;
2566 
2567  if (C == '/') {
2568  FoundSlash:
2569  if (CurPtr[-2] == '*') // We found the final */. We're done!
2570  break;
2571 
2572  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2573  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2574  // We found the final */, though it had an escaped newline between the
2575  // * and /. We're done!
2576  break;
2577  }
2578  }
2579  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2580  // If this is a /* inside of the comment, emit a warning. Don't do this
2581  // if this is a /*/, which will end the comment. This misses cases with
2582  // embedded escaped newlines, but oh well.
2583  if (!isLexingRawMode())
2584  Diag(CurPtr-1, diag::warn_nested_block_comment);
2585  }
2586  } else if (C == 0 && CurPtr == BufferEnd+1) {
2587  if (!isLexingRawMode())
2588  Diag(BufferPtr, diag::err_unterminated_block_comment);
2589  // Note: the user probably forgot a */. We could continue immediately
2590  // after the /*, but this would involve lexing a lot of what really is the
2591  // comment, which surely would confuse the parser.
2592  --CurPtr;
2593 
2594  // KeepWhitespaceMode should return this broken comment as a token. Since
2595  // it isn't a well formed comment, just return it as an 'unknown' token.
2596  if (isKeepWhitespaceMode()) {
2597  FormTokenWithChars(Result, CurPtr, tok::unknown);
2598  return true;
2599  }
2600 
2601  BufferPtr = CurPtr;
2602  return false;
2603  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2605  cutOffLexing();
2606  return false;
2607  }
2608 
2609  C = *CurPtr++;
2610  }
2611 
2612  // Notify comment handlers about the comment unless we're in a #if 0 block.
2613  if (PP && !isLexingRawMode() &&
2614  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2615  getSourceLocation(CurPtr)))) {
2616  BufferPtr = CurPtr;
2617  return true; // A token has to be returned.
2618  }
2619 
2620  // If we are returning comments as tokens, return this comment as a token.
2621  if (inKeepCommentMode()) {
2622  FormTokenWithChars(Result, CurPtr, tok::comment);
2623  return true;
2624  }
2625 
2626  // It is common for the tokens immediately after a /**/ comment to be
2627  // whitespace. Instead of going through the big switch, handle it
2628  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2629  // have already returned above with the comment as a token.
2630  if (isHorizontalWhitespace(*CurPtr)) {
2631  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2632  return false;
2633  }
2634 
2635  // Otherwise, just return so that the next character will be lexed as a token.
2636  BufferPtr = CurPtr;
2637  Result.setFlag(Token::LeadingSpace);
2638  return false;
2639 }
2640 
2641 //===----------------------------------------------------------------------===//
2642 // Primary Lexing Entry Points
2643 //===----------------------------------------------------------------------===//
2644 
2645 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2646 /// uninterpreted string. This switches the lexer out of directive mode.
2648  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2649  "Must be in a preprocessing directive!");
2650  Token Tmp;
2651 
2652  // CurPtr - Cache BufferPtr in an automatic variable.
2653  const char *CurPtr = BufferPtr;
2654  while (true) {
2655  char Char = getAndAdvanceChar(CurPtr, Tmp);
2656  switch (Char) {
2657  default:
2658  if (Result)
2659  Result->push_back(Char);
2660  break;
2661  case 0: // Null.
2662  // Found end of file?
2663  if (CurPtr-1 != BufferEnd) {
2664  if (isCodeCompletionPoint(CurPtr-1)) {
2666  cutOffLexing();
2667  return;
2668  }
2669 
2670  // Nope, normal character, continue.
2671  if (Result)
2672  Result->push_back(Char);
2673  break;
2674  }
2675  // FALL THROUGH.
2676  LLVM_FALLTHROUGH;
2677  case '\r':
2678  case '\n':
2679  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2680  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2681  BufferPtr = CurPtr-1;
2682 
2683  // Next, lex the character, which should handle the EOD transition.
2684  Lex(Tmp);
2685  if (Tmp.is(tok::code_completion)) {
2686  if (PP)
2688  Lex(Tmp);
2689  }
2690  assert(Tmp.is(tok::eod) && "Unexpected token!");
2691 
2692  // Finally, we're done;
2693  return;
2694  }
2695  }
2696 }
2697 
2698 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2699 /// condition, reporting diagnostics and handling other edge cases as required.
2700 /// This returns true if Result contains a token, false if PP.Lex should be
2701 /// called again.
2702 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2703  // If we hit the end of the file while parsing a preprocessor directive,
2704  // end the preprocessor directive first. The next token returned will
2705  // then be the end of file.
2707  // Done parsing the "line".
2709  // Update the location of token as well as BufferPtr.
2710  FormTokenWithChars(Result, CurPtr, tok::eod);
2711 
2712  // Restore comment saving mode, in case it was disabled for directive.
2713  if (PP)
2715  return true; // Have a token.
2716  }
2717 
2718  // If we are in raw mode, return this event as an EOF token. Let the caller
2719  // that put us in raw mode handle the event.
2720  if (isLexingRawMode()) {
2721  Result.startToken();
2722  BufferPtr = BufferEnd;
2723  FormTokenWithChars(Result, BufferEnd, tok::eof);
2724  return true;
2725  }
2726 
2727  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2729  ConditionalStack.clear();
2730  }
2731 
2732  // Issue diagnostics for unterminated #if and missing newline.
2733 
2734  // If we are in a #if directive, emit an error.
2735  while (!ConditionalStack.empty()) {
2736  if (PP->getCodeCompletionFileLoc() != FileLoc)
2737  PP->Diag(ConditionalStack.back().IfLoc,
2738  diag::err_pp_unterminated_conditional);
2739  ConditionalStack.pop_back();
2740  }
2741 
2742  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2743  // a pedwarn.
2744  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2745  DiagnosticsEngine &Diags = PP->getDiagnostics();
2746  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2747  unsigned DiagID;
2748 
2749  if (LangOpts.CPlusPlus11) {
2750  // C++11 [lex.phases] 2.2 p2
2751  // Prefer the C++98 pedantic compatibility warning over the generic,
2752  // non-extension, user-requested "missing newline at EOF" warning.
2753  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2754  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2755  } else {
2756  DiagID = diag::warn_no_newline_eof;
2757  }
2758  } else {
2759  DiagID = diag::ext_no_newline_eof;
2760  }
2761 
2762  Diag(BufferEnd, DiagID)
2763  << FixItHint::CreateInsertion(EndLoc, "\n");
2764  }
2765 
2766  BufferPtr = CurPtr;
2767 
2768  // Finally, let the preprocessor handle this.
2769  return PP->HandleEndOfFile(Result, isPragmaLexer());
2770 }
2771 
2772 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2773 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2774 /// else and 2 if there are no more tokens in the buffer controlled by the
2775 /// lexer.
2776 unsigned Lexer::isNextPPTokenLParen() {
2777  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2778 
2779  // Switch to 'skipping' mode. This will ensure that we can lex a token
2780  // without emitting diagnostics, disables macro expansion, and will cause EOF
2781  // to return an EOF token instead of popping the include stack.
2782  LexingRawMode = true;
2783 
2784  // Save state that can be changed while lexing so that we can restore it.
2785  const char *TmpBufferPtr = BufferPtr;
2786  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2787  bool atStartOfLine = IsAtStartOfLine;
2788  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2789  bool leadingSpace = HasLeadingSpace;
2790 
2791  Token Tok;
2792  Lex(Tok);
2793 
2794  // Restore state that may have changed.
2795  BufferPtr = TmpBufferPtr;
2796  ParsingPreprocessorDirective = inPPDirectiveMode;
2797  HasLeadingSpace = leadingSpace;
2798  IsAtStartOfLine = atStartOfLine;
2799  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2800 
2801  // Restore the lexer back to non-skipping mode.
2802  LexingRawMode = false;
2803 
2804  if (Tok.is(tok::eof))
2805  return 2;
2806  return Tok.is(tok::l_paren);
2807 }
2808 
2809 /// Find the end of a version control conflict marker.
2810 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2811  ConflictMarkerKind CMK) {
2812  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2813  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2814  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2815  size_t Pos = RestOfBuffer.find(Terminator);
2816  while (Pos != StringRef::npos) {
2817  // Must occur at start of line.
2818  if (Pos == 0 ||
2819  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2820  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2821  Pos = RestOfBuffer.find(Terminator);
2822  continue;
2823  }
2824  return RestOfBuffer.data()+Pos;
2825  }
2826  return nullptr;
2827 }
2828 
2829 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2830 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2831 /// and recover nicely. This returns true if it is a conflict marker and false
2832 /// if not.
2833 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2834  // Only a conflict marker if it starts at the beginning of a line.
2835  if (CurPtr != BufferStart &&
2836  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2837  return false;
2838 
2839  // Check to see if we have <<<<<<< or >>>>.
2840  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2841  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2842  return false;
2843 
2844  // If we have a situation where we don't care about conflict markers, ignore
2845  // it.
2846  if (CurrentConflictMarkerState || isLexingRawMode())
2847  return false;
2848 
2849  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2850 
2851  // Check to see if there is an ending marker somewhere in the buffer at the
2852  // start of a line to terminate this conflict marker.
2853  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2854  // We found a match. We are really in a conflict marker.
2855  // Diagnose this, and ignore to the end of line.
2856  Diag(CurPtr, diag::err_conflict_marker);
2857  CurrentConflictMarkerState = Kind;
2858 
2859  // Skip ahead to the end of line. We know this exists because the
2860  // end-of-conflict marker starts with \r or \n.
2861  while (*CurPtr != '\r' && *CurPtr != '\n') {
2862  assert(CurPtr != BufferEnd && "Didn't find end of line");
2863  ++CurPtr;
2864  }
2865  BufferPtr = CurPtr;
2866  return true;
2867  }
2868 
2869  // No end of conflict marker found.
2870  return false;
2871 }
2872 
2873 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2874 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2875 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2876 /// the line. This returns true if it is a conflict marker and false if not.
2877 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2878  // Only a conflict marker if it starts at the beginning of a line.
2879  if (CurPtr != BufferStart &&
2880  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2881  return false;
2882 
2883  // If we have a situation where we don't care about conflict markers, ignore
2884  // it.
2885  if (!CurrentConflictMarkerState || isLexingRawMode())
2886  return false;
2887 
2888  // Check to see if we have the marker (4 characters in a row).
2889  for (unsigned i = 1; i != 4; ++i)
2890  if (CurPtr[i] != CurPtr[0])
2891  return false;
2892 
2893  // If we do have it, search for the end of the conflict marker. This could
2894  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2895  // be the end of conflict marker.
2896  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2897  CurrentConflictMarkerState)) {
2898  CurPtr = End;
2899 
2900  // Skip ahead to the end of line.
2901  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2902  ++CurPtr;
2903 
2904  BufferPtr = CurPtr;
2905 
2906  // No longer in the conflict marker.
2907  CurrentConflictMarkerState = CMK_None;
2908  return true;
2909  }
2910 
2911  return false;
2912 }
2913 
2914 static const char *findPlaceholderEnd(const char *CurPtr,
2915  const char *BufferEnd) {
2916  if (CurPtr == BufferEnd)
2917  return nullptr;
2918  BufferEnd -= 1; // Scan until the second last character.
2919  for (; CurPtr != BufferEnd; ++CurPtr) {
2920  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2921  return CurPtr + 2;
2922  }
2923  return nullptr;
2924 }
2925 
2926 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2927  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2929  return false;
2930  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2931  if (!End)
2932  return false;
2933  const char *Start = CurPtr - 1;
2934  if (!LangOpts.AllowEditorPlaceholders)
2935  Diag(Start, diag::err_placeholder_in_source);
2936  Result.startToken();
2937  FormTokenWithChars(Result, End, tok::raw_identifier);
2938  Result.setRawIdentifierData(Start);
2939  PP->LookUpIdentifierInfo(Result);
2941  BufferPtr = End;
2942  return true;
2943 }
2944 
2945 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2946  if (PP && PP->isCodeCompletionEnabled()) {
2947  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2948  return Loc == PP->getCodeCompletionLoc();
2949  }
2950 
2951  return false;
2952 }
2953 
2954 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2955  Token *Result) {
2956  unsigned CharSize;
2957  char Kind = getCharAndSize(StartPtr, CharSize);
2958 
2959  unsigned NumHexDigits;
2960  if (Kind == 'u')
2961  NumHexDigits = 4;
2962  else if (Kind == 'U')
2963  NumHexDigits = 8;
2964  else
2965  return 0;
2966 
2967  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2968  if (Result && !isLexingRawMode())
2969  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2970  return 0;
2971  }
2972 
2973  const char *CurPtr = StartPtr + CharSize;
2974  const char *KindLoc = &CurPtr[-1];
2975 
2976  uint32_t CodePoint = 0;
2977  for (unsigned i = 0; i < NumHexDigits; ++i) {
2978  char C = getCharAndSize(CurPtr, CharSize);
2979 
2980  unsigned Value = llvm::hexDigitValue(C);
2981  if (Value == -1U) {
2982  if (Result && !isLexingRawMode()) {
2983  if (i == 0) {
2984  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2985  << StringRef(KindLoc, 1);
2986  } else {
2987  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2988 
2989  // If the user wrote \U1234, suggest a fixit to \u.
2990  if (i == 4 && NumHexDigits == 8) {
2991  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
2992  Diag(KindLoc, diag::note_ucn_four_not_eight)
2993  << FixItHint::CreateReplacement(URange, "u");
2994  }
2995  }
2996  }
2997 
2998  return 0;
2999  }
3000 
3001  CodePoint <<= 4;
3002  CodePoint += Value;
3003 
3004  CurPtr += CharSize;
3005  }
3006 
3007  if (Result) {
3008  Result->setFlag(Token::HasUCN);
3009  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
3010  StartPtr = CurPtr;
3011  else
3012  while (StartPtr != CurPtr)
3013  (void)getAndAdvanceChar(StartPtr, *Result);
3014  } else {
3015  StartPtr = CurPtr;
3016  }
3017 
3018  // Don't apply C family restrictions to UCNs in assembly mode
3019  if (LangOpts.AsmPreprocessor)
3020  return CodePoint;
3021 
3022  // C99 6.4.3p2: A universal character name shall not specify a character whose
3023  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3024  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3025  // C++11 [lex.charset]p2: If the hexadecimal value for a
3026  // universal-character-name corresponds to a surrogate code point (in the
3027  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3028  // if the hexadecimal value for a universal-character-name outside the
3029  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3030  // string literal corresponds to a control character (in either of the
3031  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3032  // basic source character set, the program is ill-formed.
3033  if (CodePoint < 0xA0) {
3034  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3035  return CodePoint;
3036 
3037  // We don't use isLexingRawMode() here because we need to warn about bad
3038  // UCNs even when skipping preprocessing tokens in a #if block.
3039  if (Result && PP) {
3040  if (CodePoint < 0x20 || CodePoint >= 0x7F)
3041  Diag(BufferPtr, diag::err_ucn_control_character);
3042  else {
3043  char C = static_cast<char>(CodePoint);
3044  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3045  }
3046  }
3047 
3048  return 0;
3049  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3050  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3051  // We don't use isLexingRawMode() here because we need to diagnose bad
3052  // UCNs even when skipping preprocessing tokens in a #if block.
3053  if (Result && PP) {
3054  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3055  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3056  else
3057  Diag(BufferPtr, diag::err_ucn_escape_invalid);
3058  }
3059  return 0;
3060  }
3061 
3062  return CodePoint;
3063 }
3064 
3065 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3066  const char *CurPtr) {
3067  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3069  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3070  UnicodeWhitespaceChars.contains(C)) {
3071  Diag(BufferPtr, diag::ext_unicode_whitespace)
3072  << makeCharRange(*this, BufferPtr, CurPtr);
3073 
3074  Result.setFlag(Token::LeadingSpace);
3075  return true;
3076  }
3077  return false;
3078 }
3079 
3080 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3081  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3083  !PP->isPreprocessedOutput()) {
3085  makeCharRange(*this, BufferPtr, CurPtr),
3086  /*IsFirst=*/true);
3088  makeCharRange(*this, BufferPtr, CurPtr));
3089  }
3090 
3091  MIOpt.ReadToken();
3092  return LexIdentifier(Result, CurPtr);
3093  }
3094 
3096  !PP->isPreprocessedOutput() &&
3097  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3098  // Non-ASCII characters tend to creep into source code unintentionally.
3099  // Instead of letting the parser complain about the unknown token,
3100  // just drop the character.
3101  // Note that we can /only/ do this when the non-ASCII character is actually
3102  // spelled as Unicode, not written as a UCN. The standard requires that
3103  // we not throw away any possible preprocessor tokens, but there's a
3104  // loophole in the mapping of Unicode characters to basic character set
3105  // characters that allows us to map these particular characters to, say,
3106  // whitespace.
3107  Diag(BufferPtr, diag::err_non_ascii)
3108  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3109 
3110  BufferPtr = CurPtr;
3111  return false;
3112  }
3113 
3114  // Otherwise, we have an explicit UCN or a character that's unlikely to show
3115  // up by accident.
3116  MIOpt.ReadToken();
3117  FormTokenWithChars(Result, CurPtr, tok::unknown);
3118  return true;
3119 }
3120 
3121 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3122  IsAtStartOfLine = Result.isAtStartOfLine();
3123  HasLeadingSpace = Result.hasLeadingSpace();
3124  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3125  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3126 }
3127 
3128 bool Lexer::Lex(Token &Result) {
3129  // Start a new token.
3130  Result.startToken();
3131 
3132  // Set up misc whitespace flags for LexTokenInternal.
3133  if (IsAtStartOfLine) {
3134  Result.setFlag(Token::StartOfLine);
3135  IsAtStartOfLine = false;
3136  }
3137 
3138  if (HasLeadingSpace) {
3139  Result.setFlag(Token::LeadingSpace);
3140  HasLeadingSpace = false;
3141  }
3142 
3143  if (HasLeadingEmptyMacro) {
3145  HasLeadingEmptyMacro = false;
3146  }
3147 
3148  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3149  IsAtPhysicalStartOfLine = false;
3150  bool isRawLex = isLexingRawMode();
3151  (void) isRawLex;
3152  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3153  // (After the LexTokenInternal call, the lexer might be destroyed.)
3154  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3155  return returnedToken;
3156 }
3157 
3158 /// LexTokenInternal - This implements a simple C family lexer. It is an
3159 /// extremely performance critical piece of code. This assumes that the buffer
3160 /// has a null character at the end of the file. This returns a preprocessing
3161 /// token, not a normal token, as such, it is an internal interface. It assumes
3162 /// that the Flags of result have been cleared before calling this.
3163 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3164 LexNextToken:
3165  // New token, can't need cleaning yet.
3167  Result.setIdentifierInfo(nullptr);
3168 
3169  // CurPtr - Cache BufferPtr in an automatic variable.
3170  const char *CurPtr = BufferPtr;
3171 
3172  // Small amounts of horizontal whitespace is very common between tokens.
3173  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
3174  ++CurPtr;
3175  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3176  ++CurPtr;
3177 
3178  // If we are keeping whitespace and other tokens, just return what we just
3179  // skipped. The next lexer invocation will return the token after the
3180  // whitespace.
3181  if (isKeepWhitespaceMode()) {
3182  FormTokenWithChars(Result, CurPtr, tok::unknown);
3183  // FIXME: The next token will not have LeadingSpace set.
3184  return true;
3185  }
3186 
3187  BufferPtr = CurPtr;
3188  Result.setFlag(Token::LeadingSpace);
3189  }
3190 
3191  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3192 
3193  // Read a character, advancing over it.
3194  char Char = getAndAdvanceChar(CurPtr, Result);
3196 
3197  switch (Char) {
3198  case 0: // Null.
3199  // Found end of file?
3200  if (CurPtr-1 == BufferEnd)
3201  return LexEndOfFile(Result, CurPtr-1);
3202 
3203  // Check if we are performing code completion.
3204  if (isCodeCompletionPoint(CurPtr-1)) {
3205  // Return the code-completion token.
3206  Result.startToken();
3207  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3208  return true;
3209  }
3210 
3211  if (!isLexingRawMode())
3212  Diag(CurPtr-1, diag::null_in_file);
3213  Result.setFlag(Token::LeadingSpace);
3214  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3215  return true; // KeepWhitespaceMode
3216 
3217  // We know the lexer hasn't changed, so just try again with this lexer.
3218  // (We manually eliminate the tail call to avoid recursion.)
3219  goto LexNextToken;
3220 
3221  case 26: // DOS & CP/M EOF: "^Z".
3222  // If we're in Microsoft extensions mode, treat this as end of file.
3223  if (LangOpts.MicrosoftExt) {
3224  if (!isLexingRawMode())
3225  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3226  return LexEndOfFile(Result, CurPtr-1);
3227  }
3228 
3229  // If Microsoft extensions are disabled, this is just random garbage.
3230  Kind = tok::unknown;
3231  break;
3232 
3233  case '\r':
3234  if (CurPtr[0] == '\n')
3235  Char = getAndAdvanceChar(CurPtr, Result);
3236  LLVM_FALLTHROUGH;
3237  case '\n':
3238  // If we are inside a preprocessor directive and we see the end of line,
3239  // we know we are done with the directive, so return an EOD token.
3241  // Done parsing the "line".
3243 
3244  // Restore comment saving mode, in case it was disabled for directive.
3245  if (PP)
3247 
3248  // Since we consumed a newline, we are back at the start of a line.
3249  IsAtStartOfLine = true;
3250  IsAtPhysicalStartOfLine = true;
3251 
3252  Kind = tok::eod;
3253  break;
3254  }
3255 
3256  // No leading whitespace seen so far.
3258 
3259  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3260  return true; // KeepWhitespaceMode
3261 
3262  // We only saw whitespace, so just try again with this lexer.
3263  // (We manually eliminate the tail call to avoid recursion.)
3264  goto LexNextToken;
3265  case ' ':
3266  case '\t':
3267  case '\f':
3268  case '\v':
3269  SkipHorizontalWhitespace:
3270  Result.setFlag(Token::LeadingSpace);
3271  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3272  return true; // KeepWhitespaceMode
3273 
3274  SkipIgnoredUnits:
3275  CurPtr = BufferPtr;
3276 
3277  // If the next token is obviously a // or /* */ comment, skip it efficiently
3278  // too (without going through the big switch stmt).
3279  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3280  LangOpts.LineComment &&
3281  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3282  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3283  return true; // There is a token to return.
3284  goto SkipIgnoredUnits;
3285  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3286  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3287  return true; // There is a token to return.
3288  goto SkipIgnoredUnits;
3289  } else if (isHorizontalWhitespace(*CurPtr)) {
3290  goto SkipHorizontalWhitespace;
3291  }
3292  // We only saw whitespace, so just try again with this lexer.
3293  // (We manually eliminate the tail call to avoid recursion.)
3294  goto LexNextToken;
3295 
3296  // C99 6.4.4.1: Integer Constants.
3297  // C99 6.4.4.2: Floating Constants.
3298  case '0': case '1': case '2': case '3': case '4':
3299  case '5': case '6': case '7': case '8': case '9':
3300  // Notify MIOpt that we read a non-whitespace/non-comment token.
3301  MIOpt.ReadToken();
3302  return LexNumericConstant(Result, CurPtr);
3303 
3304  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3305  // Notify MIOpt that we read a non-whitespace/non-comment token.
3306  MIOpt.ReadToken();
3307 
3308  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3309  Char = getCharAndSize(CurPtr, SizeTmp);
3310 
3311  // UTF-16 string literal
3312  if (Char == '"')
3313  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3314  tok::utf16_string_literal);
3315 
3316  // UTF-16 character constant
3317  if (Char == '\'')
3318  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3319  tok::utf16_char_constant);
3320 
3321  // UTF-16 raw string literal
3322  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3323  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3324  return LexRawStringLiteral(Result,
3325  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3326  SizeTmp2, Result),
3327  tok::utf16_string_literal);
3328 
3329  if (Char == '8') {
3330  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3331 
3332  // UTF-8 string literal
3333  if (Char2 == '"')
3334  return LexStringLiteral(Result,
3335  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3336  SizeTmp2, Result),
3337  tok::utf8_string_literal);
3338  if (Char2 == '\'' && LangOpts.CPlusPlus17)
3339  return LexCharConstant(
3340  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3341  SizeTmp2, Result),
3342  tok::utf8_char_constant);
3343 
3344  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3345  unsigned SizeTmp3;
3346  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3347  // UTF-8 raw string literal
3348  if (Char3 == '"') {
3349  return LexRawStringLiteral(Result,
3350  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3351  SizeTmp2, Result),
3352  SizeTmp3, Result),
3353  tok::utf8_string_literal);
3354  }
3355  }
3356  }
3357  }
3358 
3359  // treat u like the start of an identifier.
3360  return LexIdentifier(Result, CurPtr);
3361 
3362  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3363  // Notify MIOpt that we read a non-whitespace/non-comment token.
3364  MIOpt.ReadToken();
3365 
3366  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3367  Char = getCharAndSize(CurPtr, SizeTmp);
3368 
3369  // UTF-32 string literal
3370  if (Char == '"')
3371  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3372  tok::utf32_string_literal);
3373 
3374  // UTF-32 character constant
3375  if (Char == '\'')
3376  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3377  tok::utf32_char_constant);
3378 
3379  // UTF-32 raw string literal
3380  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3381  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3382  return LexRawStringLiteral(Result,
3383  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3384  SizeTmp2, Result),
3385  tok::utf32_string_literal);
3386  }
3387 
3388  // treat U like the start of an identifier.
3389  return LexIdentifier(Result, CurPtr);
3390 
3391  case 'R': // Identifier or C++0x raw string literal
3392  // Notify MIOpt that we read a non-whitespace/non-comment token.
3393  MIOpt.ReadToken();
3394 
3395  if (LangOpts.CPlusPlus11) {
3396  Char = getCharAndSize(CurPtr, SizeTmp);
3397 
3398  if (Char == '"')
3399  return LexRawStringLiteral(Result,
3400  ConsumeChar(CurPtr, SizeTmp, Result),
3401  tok::string_literal);
3402  }
3403 
3404  // treat R like the start of an identifier.
3405  return LexIdentifier(Result, CurPtr);
3406 
3407  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3408  // Notify MIOpt that we read a non-whitespace/non-comment token.
3409  MIOpt.ReadToken();
3410  Char = getCharAndSize(CurPtr, SizeTmp);
3411 
3412  // Wide string literal.
3413  if (Char == '"')
3414  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3415  tok::wide_string_literal);
3416 
3417  // Wide raw string literal.
3418  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3419  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3420  return LexRawStringLiteral(Result,
3421  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3422  SizeTmp2, Result),
3423  tok::wide_string_literal);
3424 
3425  // Wide character constant.
3426  if (Char == '\'')
3427  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3428  tok::wide_char_constant);
3429  // FALL THROUGH, treating L like the start of an identifier.
3430  LLVM_FALLTHROUGH;
3431 
3432  // C99 6.4.2: Identifiers.
3433  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3434  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3435  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3436  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3437  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3438  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3439  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3440  case 'v': case 'w': case 'x': case 'y': case 'z':
3441  case '_':
3442  // Notify MIOpt that we read a non-whitespace/non-comment token.
3443  MIOpt.ReadToken();
3444  return LexIdentifier(Result, CurPtr);
3445 
3446  case '$': // $ in identifiers.
3447  if (LangOpts.DollarIdents) {
3448  if (!isLexingRawMode())
3449  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3450  // Notify MIOpt that we read a non-whitespace/non-comment token.
3451  MIOpt.ReadToken();
3452  return LexIdentifier(Result, CurPtr);
3453  }
3454 
3455  Kind = tok::unknown;
3456  break;
3457 
3458  // C99 6.4.4: Character Constants.
3459  case '\'':
3460  // Notify MIOpt that we read a non-whitespace/non-comment token.
3461  MIOpt.ReadToken();
3462  return LexCharConstant(Result, CurPtr, tok::char_constant);
3463 
3464  // C99 6.4.5: String Literals.
3465  case '"':
3466  // Notify MIOpt that we read a non-whitespace/non-comment token.
3467  MIOpt.ReadToken();
3468  return LexStringLiteral(Result, CurPtr,
3469  ParsingFilename ? tok::header_name
3470  : tok::string_literal);
3471 
3472  // C99 6.4.6: Punctuators.
3473  case '?':
3474  Kind = tok::question;
3475  break;
3476  case '[':
3477  Kind = tok::l_square;
3478  break;
3479  case ']':
3480  Kind = tok::r_square;
3481  break;
3482  case '(':
3483  Kind = tok::l_paren;
3484  break;
3485  case ')':
3486  Kind = tok::r_paren;
3487  break;
3488  case '{':
3489  Kind = tok::l_brace;
3490  break;
3491  case '}':
3492  Kind = tok::r_brace;
3493  break;
3494  case '.':
3495  Char = getCharAndSize(CurPtr, SizeTmp);
3496  if (Char >= '0' && Char <= '9') {
3497  // Notify MIOpt that we read a non-whitespace/non-comment token.
3498  MIOpt.ReadToken();
3499 
3500  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3501  } else if (LangOpts.CPlusPlus && Char == '*') {
3502  Kind = tok::periodstar;
3503  CurPtr += SizeTmp;
3504  } else if (Char == '.' &&
3505  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3506  Kind = tok::ellipsis;
3507  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3508  SizeTmp2, Result);
3509  } else {
3510  Kind = tok::period;
3511  }
3512  break;
3513  case '&':
3514  Char = getCharAndSize(CurPtr, SizeTmp);
3515  if (Char == '&') {
3516  Kind = tok::ampamp;
3517  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3518  } else if (Char == '=') {
3519  Kind = tok::ampequal;
3520  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3521  } else {
3522  Kind = tok::amp;
3523  }
3524  break;
3525  case '*':
3526  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3527  Kind = tok::starequal;
3528  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3529  } else {
3530  Kind = tok::star;
3531  }
3532  break;
3533  case '+':
3534  Char = getCharAndSize(CurPtr, SizeTmp);
3535  if (Char == '+') {
3536  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3537  Kind = tok::plusplus;
3538  } else if (Char == '=') {
3539  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3540  Kind = tok::plusequal;
3541  } else {
3542  Kind = tok::plus;
3543  }
3544  break;
3545  case '-':
3546  Char = getCharAndSize(CurPtr, SizeTmp);
3547  if (Char == '-') { // --
3548  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3549  Kind = tok::minusminus;
3550  } else if (Char == '>' && LangOpts.CPlusPlus &&
3551  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3552  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3553  SizeTmp2, Result);
3554  Kind = tok::arrowstar;
3555  } else if (Char == '>') { // ->
3556  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3557  Kind = tok::arrow;
3558  } else if (Char == '=') { // -=
3559  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3560  Kind = tok::minusequal;
3561  } else {
3562  Kind = tok::minus;
3563  }
3564  break;
3565  case '~':
3566  Kind = tok::tilde;
3567  break;
3568  case '!':
3569  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3570  Kind = tok::exclaimequal;
3571  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3572  } else {
3573  Kind = tok::exclaim;
3574  }
3575  break;
3576  case '/':
3577  // 6.4.9: Comments
3578  Char = getCharAndSize(CurPtr, SizeTmp);
3579  if (Char == '/') { // Line comment.
3580  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3581  // want to lex this as a comment. There is one problem with this though,
3582  // that in one particular corner case, this can change the behavior of the
3583  // resultant program. For example, In "foo //**/ bar", C89 would lex
3584  // this as "foo / bar" and languages with Line comments would lex it as
3585  // "foo". Check to see if the character after the second slash is a '*'.
3586  // If so, we will lex that as a "/" instead of the start of a comment.
3587  // However, we never do this if we are just preprocessing.
3588  bool TreatAsComment = LangOpts.LineComment &&
3589  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3590  if (!TreatAsComment)
3591  if (!(PP && PP->isPreprocessedOutput()))
3592  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3593 
3594  if (TreatAsComment) {
3595  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3596  TokAtPhysicalStartOfLine))
3597  return true; // There is a token to return.
3598 
3599  // It is common for the tokens immediately after a // comment to be
3600  // whitespace (indentation for the next line). Instead of going through
3601  // the big switch, handle it efficiently now.
3602  goto SkipIgnoredUnits;
3603  }
3604  }
3605 
3606  if (Char == '*') { // /**/ comment.
3607  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3608  TokAtPhysicalStartOfLine))
3609  return true; // There is a token to return.
3610 
3611  // We only saw whitespace, so just try again with this lexer.
3612  // (We manually eliminate the tail call to avoid recursion.)
3613  goto LexNextToken;
3614  }
3615 
3616  if (Char == '=') {
3617  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3618  Kind = tok::slashequal;
3619  } else {
3620  Kind = tok::slash;
3621  }
3622  break;
3623  case '%':
3624  Char = getCharAndSize(CurPtr, SizeTmp);
3625  if (Char == '=') {
3626  Kind = tok::percentequal;
3627  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3628  } else if (LangOpts.Digraphs && Char == '>') {
3629  Kind = tok::r_brace; // '%>' -> '}'
3630  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3631  } else if (LangOpts.Digraphs && Char == ':') {
3632  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3633  Char = getCharAndSize(CurPtr, SizeTmp);
3634  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3635  Kind = tok::hashhash; // '%:%:' -> '##'
3636  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3637  SizeTmp2, Result);
3638  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3639  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3640  if (!isLexingRawMode())
3641  Diag(BufferPtr, diag::ext_charize_microsoft);
3642  Kind = tok::hashat;
3643  } else { // '%:' -> '#'
3644  // We parsed a # character. If this occurs at the start of the line,
3645  // it's actually the start of a preprocessing directive. Callback to
3646  // the preprocessor to handle it.
3647  // TODO: -fpreprocessed mode??
3648  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3649  goto HandleDirective;
3650 
3651  Kind = tok::hash;
3652  }
3653  } else {
3654  Kind = tok::percent;
3655  }
3656  break;
3657  case '<':
3658  Char = getCharAndSize(CurPtr, SizeTmp);
3659  if (ParsingFilename) {
3660  return LexAngledStringLiteral(Result, CurPtr);
3661  } else if (Char == '<') {
3662  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3663  if (After == '=') {
3664  Kind = tok::lesslessequal;
3665  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3666  SizeTmp2, Result);
3667  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3668  // If this is actually a '<<<<<<<' version control conflict marker,
3669  // recognize it as such and recover nicely.
3670  goto LexNextToken;
3671  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3672  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3673  // ignore it.
3674  goto LexNextToken;
3675  } else if (LangOpts.CUDA && After == '<') {
3676  Kind = tok::lesslessless;
3677  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3678  SizeTmp2, Result);
3679  } else {
3680  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3681  Kind = tok::lessless;
3682  }
3683  } else if (Char == '=') {
3684  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3685  if (After == '>') {
3686  if (getLangOpts().CPlusPlus2a) {
3687  if (!isLexingRawMode())
3688  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3689  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3690  SizeTmp2, Result);
3691  Kind = tok::spaceship;
3692  break;
3693  }
3694  // Suggest adding a space between the '<=' and the '>' to avoid a
3695  // change in semantics if this turns up in C++ <=17 mode.
3696  if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3697  Diag(BufferPtr, diag::warn_cxx2a_compat_spaceship)
3699  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3700  }
3701  }
3702  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3703  Kind = tok::lessequal;
3704  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3705  if (LangOpts.CPlusPlus11 &&
3706  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3707  // C++0x [lex.pptoken]p3:
3708  // Otherwise, if the next three characters are <:: and the subsequent
3709  // character is neither : nor >, the < is treated as a preprocessor
3710  // token by itself and not as the first character of the alternative
3711  // token <:.
3712  unsigned SizeTmp3;
3713  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3714  if (After != ':' && After != '>') {
3715  Kind = tok::less;
3716  if (!isLexingRawMode())
3717  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3718  break;
3719  }
3720  }
3721 
3722  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3723  Kind = tok::l_square;
3724  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3725  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3726  Kind = tok::l_brace;
3727  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3728  lexEditorPlaceholder(Result, CurPtr)) {
3729  return true;
3730  } else {
3731  Kind = tok::less;
3732  }
3733  break;
3734  case '>':
3735  Char = getCharAndSize(CurPtr, SizeTmp);
3736  if (Char == '=') {
3737  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3738  Kind = tok::greaterequal;
3739  } else if (Char == '>') {
3740  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3741  if (After == '=') {
3742  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3743  SizeTmp2, Result);
3744  Kind = tok::greatergreaterequal;
3745  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3746  // If this is actually a '>>>>' conflict marker, recognize it as such
3747  // and recover nicely.
3748  goto LexNextToken;
3749  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3750  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3751  goto LexNextToken;
3752  } else if (LangOpts.CUDA && After == '>') {
3753  Kind = tok::greatergreatergreater;
3754  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3755  SizeTmp2, Result);
3756  } else {
3757  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3758  Kind = tok::greatergreater;
3759  }
3760  } else {
3761  Kind = tok::greater;
3762  }
3763  break;
3764  case '^':
3765  Char = getCharAndSize(CurPtr, SizeTmp);
3766  if (Char == '=') {
3767  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3768  Kind = tok::caretequal;
3769  } else if (LangOpts.OpenCL && Char == '^') {
3770  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3771  Kind = tok::caretcaret;
3772  } else {
3773  Kind = tok::caret;
3774  }
3775  break;
3776  case '|':
3777  Char = getCharAndSize(CurPtr, SizeTmp);
3778  if (Char == '=') {
3779  Kind = tok::pipeequal;
3780  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3781  } else if (Char == '|') {
3782  // If this is '|||||||' and we're in a conflict marker, ignore it.
3783  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3784  goto LexNextToken;
3785  Kind = tok::pipepipe;
3786  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3787  } else {
3788  Kind = tok::pipe;
3789  }
3790  break;
3791  case ':':
3792  Char = getCharAndSize(CurPtr, SizeTmp);
3793  if (LangOpts.Digraphs && Char == '>') {
3794  Kind = tok::r_square; // ':>' -> ']'
3795  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3796  } else if ((LangOpts.CPlusPlus ||
3797  LangOpts.DoubleSquareBracketAttributes) &&
3798  Char == ':') {
3799  Kind = tok::coloncolon;
3800  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3801  } else {
3802  Kind = tok::colon;
3803  }
3804  break;
3805  case ';':
3806  Kind = tok::semi;
3807  break;
3808  case '=':
3809  Char = getCharAndSize(CurPtr, SizeTmp);
3810  if (Char == '=') {
3811  // If this is '====' and we're in a conflict marker, ignore it.
3812  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3813  goto LexNextToken;
3814 
3815  Kind = tok::equalequal;
3816  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3817  } else {
3818  Kind = tok::equal;
3819  }
3820  break;
3821  case ',':
3822  Kind = tok::comma;
3823  break;
3824  case '#':
3825  Char = getCharAndSize(CurPtr, SizeTmp);
3826  if (Char == '#') {
3827  Kind = tok::hashhash;
3828  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3829  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3830  Kind = tok::hashat;
3831  if (!isLexingRawMode())
3832  Diag(BufferPtr, diag::ext_charize_microsoft);
3833  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3834  } else {
3835  // We parsed a # character. If this occurs at the start of the line,
3836  // it's actually the start of a preprocessing directive. Callback to
3837  // the preprocessor to handle it.
3838  // TODO: -fpreprocessed mode??
3839  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3840  goto HandleDirective;
3841 
3842  Kind = tok::hash;
3843  }
3844  break;
3845 
3846  case '@':
3847  // Objective C support.
3848  if (CurPtr[-1] == '@' && LangOpts.ObjC)
3849  Kind = tok::at;
3850  else
3851  Kind = tok::unknown;
3852  break;
3853 
3854  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3855  case '\\':
3856  if (!LangOpts.AsmPreprocessor) {
3857  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3858  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3859  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3860  return true; // KeepWhitespaceMode
3861 
3862  // We only saw whitespace, so just try again with this lexer.
3863  // (We manually eliminate the tail call to avoid recursion.)
3864  goto LexNextToken;
3865  }
3866 
3867  return LexUnicode(Result, CodePoint, CurPtr);
3868  }
3869  }
3870 
3871  Kind = tok::unknown;
3872  break;
3873 
3874  default: {
3875  if (isASCII(Char)) {
3876  Kind = tok::unknown;
3877  break;
3878  }
3879 
3880  llvm::UTF32 CodePoint;
3881 
3882  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3883  // an escaped newline.
3884  --CurPtr;
3885  llvm::ConversionResult Status =
3886  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3887  (const llvm::UTF8 *)BufferEnd,
3888  &CodePoint,
3889  llvm::strictConversion);
3890  if (Status == llvm::conversionOK) {
3891  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3892  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3893  return true; // KeepWhitespaceMode
3894 
3895  // We only saw whitespace, so just try again with this lexer.
3896  // (We manually eliminate the tail call to avoid recursion.)
3897  goto LexNextToken;
3898  }
3899  return LexUnicode(Result, CodePoint, CurPtr);
3900  }
3901 
3903  PP->isPreprocessedOutput()) {
3904  ++CurPtr;
3905  Kind = tok::unknown;
3906  break;
3907  }
3908 
3909  // Non-ASCII characters tend to creep into source code unintentionally.
3910  // Instead of letting the parser complain about the unknown token,
3911  // just diagnose the invalid UTF-8, then drop the character.
3912  Diag(CurPtr, diag::err_invalid_utf8);
3913 
3914  BufferPtr = CurPtr+1;
3915  // We're pretending the character didn't exist, so just try again with
3916  // this lexer.
3917  // (We manually eliminate the tail call to avoid recursion.)
3918  goto LexNextToken;
3919  }
3920  }
3921 
3922  // Notify MIOpt that we read a non-whitespace/non-comment token.
3923  MIOpt.ReadToken();
3924 
3925  // Update the location of token as well as BufferPtr.
3926  FormTokenWithChars(Result, CurPtr, Kind);
3927  return true;
3928 
3929 HandleDirective:
3930  // We parsed a # character and it's the start of a preprocessing directive.
3931 
3932  FormTokenWithChars(Result, CurPtr, tok::hash);
3933  PP->HandleDirective(Result);
3934 
3935  if (PP->hadModuleLoaderFatalFailure()) {
3936  // With a fatal failure in the module loader, we abort parsing.
3937  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3938  return true;
3939  }
3940 
3941  // We parsed the directive; lex a token with the new state.
3942  return false;
3943 }
SourceLocation getLocForStartOfFile(FileID FID) const
Return the source location corresponding to the first byte of the specified file. ...
Describes the bounds (start, size) of the preamble and a flag required by PreprocessorOptions::Precom...
Definition: Lexer.h:58
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:388
void setCodeCompletionTokenRange(const SourceLocation Start, const SourceLocation End)
Set the code completion token range for detecting replacement range later on.
Lexer - This provides a simple interface that turns a text buffer into a stream of tokens...
Definition: Lexer.h:76
SourceLocation getLocWithOffset(int Offset) const
Return a source location with the specified offset from this SourceLocation.
This is a discriminated union of FileInfo and ExpansionInfo.
unsigned getRawEncoding() const
When a SourceLocation itself cannot be used, this returns an (opaque) 32-bit integer encoding for it...
SourceLocation getSpellingLoc() const
void setFlagValue(TokenFlags Flag, bool Val)
Set a flag to either true or false.
Definition: Token.h:259
static const llvm::sys::UnicodeCharRange C11AllowedIDCharRanges[]
void setBegin(SourceLocation b)
static __inline__ int __ATTRS_o_ai vec_any_eq(vector signed char __a, vector signed char __b)
Definition: altivec.h:14856
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
Defines the SourceManager interface.
LLVM_READNONE bool isASCII(char c)
Returns true if this is an ASCII character.
Definition: CharInfo.h:42
static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1423
bool isInPrimaryFile() const
Return true if we&#39;re in the top-level file, not in a #include.
StringRef P
const char * getCharacterData(SourceLocation SL, bool *Invalid=nullptr) const
Return a pointer to the start of the specified location in the appropriate spelling MemoryBuffer...
Each ExpansionInfo encodes the expansion location - where the token was ultimately expanded...
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1296
void setFlag(TokenFlags Flag)
Set the specified flag.
Definition: Token.h:236
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:528
void setCodeCompletionIdentifierInfo(IdentifierInfo *Filter)
Set the code completion token for filtering purposes.
StringRef getBufferData(FileID FID, bool *Invalid=nullptr) const
Return a StringRef to the source buffer data for the specified FileID.
bool hadModuleLoaderFatalFailure() const
Definition: Preprocessor.h:927
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:792
LLVM_READONLY bool isHorizontalWhitespace(unsigned char c)
Returns true if this character is horizontal ASCII whitespace: &#39; &#39;, &#39;\t&#39;, &#39;\f&#39;, &#39;\v&#39;.
Definition: CharInfo.h:70
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:77
ConflictMarkerKind
ConflictMarkerKind - Kinds of conflict marker which the lexer might be recovering from...
Definition: Lexer.h:42
static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen)
GetMappedTokenLoc - If lexing out of a &#39;mapped buffer&#39;, where we pretend the lexer buffer was all exp...
Definition: Lexer.cpp:1094
Like System, but searched after the system directories.
SourceLocation getCodeCompletionFileLoc() const
Returns the start location of the file of code-completion point.
static Lexer * Create_PragmaLexer(SourceLocation SpellingLoc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLen, Preprocessor &PP)
Create_PragmaLexer: Lexer constructor - Create a new lexer object for _Pragma expansion.
Definition: Lexer.cpp:186
Defines the MultipleIncludeOpt interface.
bool isAnnotation() const
Return true if this is any of tok::annot_* kind tokens.
Definition: Token.h:120
long i
Definition: xmmintrin.h:1456
tok::TokenKind getKind() const
Definition: Token.h:92
bool isLiteral() const
Return true if this is a "literal", like a numeric constant, string, etc.
Definition: Token.h:115
Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
Lexer constructor - Create a new lexer object for the specified buffer with the specified preprocesso...
Definition: Lexer.cpp:133
One of these records is kept for each identifier that is lexed.
static StringRef getIndentationForLine(SourceLocation Loc, const SourceManager &SM)
Returns the leading whitespace for line that corresponds to the given location Loc.
Definition: Lexer.cpp:1063
SourceLocation getBegin() const
bool ParsingPreprocessorDirective
True when parsing #XXX; turns &#39;\n&#39; into a tok::eod token.
void setRawIdentifierData(const char *Ptr)
Definition: Token.h:209
static SourceLocation getFromRawEncoding(unsigned Encoding)
Turn a raw encoding of a SourceLocation object into a real SourceLocation.
SmallVector< PPConditionalInfo, 4 > ConditionalStack
Information about the set of #if/#ifdef/#ifndef blocks we are currently in.
Token - This structure provides full information about a lexed token.
Definition: Token.h:34
void setKind(tok::TokenKind K)
Definition: Token.h:93
c
Definition: emmintrin.h:306
Keeps track of the various options that can be enabled, which controls the dialect of C or C++ that i...
Definition: LangOptions.h:49
bool isPragmaLexer() const
isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
Definition: Lexer.h:183
void resetExtendedTokenMode()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:163
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s...
Definition: Lexer.h:52
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:57
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file...
Definition: Lexer.h:261
bool isInvalid() const
bool isAtEndOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroEnd=nullptr) const
Returns true if the given MacroID location points at the character end of the immediate macro expansi...
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi8(char __b)
Initializes all values in a 128-bit vector of [16 x i8] with the specified 8-bit value.
Definition: emmintrin.h:3842
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:497
static bool isNewLineEscaped(const char *BufferStart, const char *Str)
Checks whether new line pointed by Str is preceded by escape sequence.
Definition: Lexer.cpp:1043
LLVM_READONLY bool isWhitespace(unsigned char c)
Return true if this character is horizontal or vertical ASCII whitespace: &#39; &#39;, &#39;\t&#39;, &#39;\f&#39;, &#39;\v&#39;, &#39;\n&#39;, &#39;\r&#39;.
Definition: CharInfo.h:87
StringRef getSpelling(SourceLocation loc, SmallVectorImpl< char > &buffer, bool *invalid=nullptr) const
Return the &#39;spelling&#39; of the token at the given location; does not go up to the spelling location or ...
const FormatToken & Tok
Forward-declares and imports various common LLVM datatypes that clang wants to use unqualified...
SourceLocation getExpansionLoc(SourceLocation Loc) const
Given a SourceLocation object Loc, return the expansion location referenced by the ID...
static bool getRawToken(SourceLocation Loc, Token &Result, const SourceManager &SM, const LangOptions &LangOpts, bool IgnoreWhiteSpace=false)
Relex the token at the specified location.
Definition: Lexer.cpp:446
void HandleDirective(Token &Result)
Callback invoked when the lexer sees a # token at the start of a line.
Concrete class used by the front-end to report problems and issues.
Definition: Diagnostic.h:148
Defines the Diagnostic-related interfaces.
SourceLocation getSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
const FileID FID
The SourceManager FileID corresponding to the file being lexed.
LLVM_READONLY bool isRawStringDelimBody(unsigned char c)
Return true if this is the body character of a C++ raw string delimiter.
Definition: CharInfo.h:154
static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, Lexer *L)
isBlockCommentEndOfEscapedNewLine - Return true if the specified newline character (either \n or \r) ...
Definition: Lexer.cpp:2405
bool HandleComment(Token &result, SourceRange Comment)
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
static CharSourceRange makeCharRange(Lexer &L, const char *Begin, const char *End)
Definition: Lexer.cpp:1458
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:1044
bool ParsingFilename
True after #include; turns <xx> or "xxx" into a tok::header_name token.
const LangOptions & getLangOpts() const
getLangOpts - Return the language features currently enabled.
Definition: Lexer.h:168
static const llvm::sys::UnicodeCharRange C11DisallowedInitialIDCharRanges[]
bool isInFileID(SourceLocation Loc, FileID FID, unsigned *RelativeOffset=nullptr) const
Given a specific FileID, returns true if Loc is inside that FileID chunk and sets relative offset (of...
LLVM_READONLY bool isIdentifierHead(unsigned char c, bool AllowDollar=false)
Returns true if this is a valid first character of a C identifier, which is [a-zA-Z_].
Definition: CharInfo.h:48
unsigned Offset
Definition: Format.cpp:1711
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:931
IdentifierInfo * LookUpIdentifierInfo(Token &Identifier) const
Given a tok::raw_identifier token, look up the identifier information for the token and install it in...
static bool isAtEndOfMacroExpansion(SourceLocation loc, const SourceManager &SM, const LangOptions &LangOpts, SourceLocation *MacroEnd=nullptr)
Returns true if the given MacroID location points at the last token of the macro expansion.
Definition: Lexer.cpp:814
bool isCodeCompletionEnabled() const
Determine if we are performing code completion.
SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
Defines the clang::LangOptions interface.
bool LexingRawMode
True if in raw mode.
static SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset, const SourceManager &SM, const LangOptions &LangOpts)
Computes the source location just past the end of the token at this source location.
Definition: Lexer.cpp:770
SourceLocation End
Represents a character-granular source range.
bool isKeepWhitespaceMode() const
isKeepWhitespaceMode - Return true if the lexer should return tokens for every character in the file...
Definition: Lexer.h:206
static PreambleBounds ComputePreamble(StringRef Buffer, const LangOptions &LangOpts, unsigned MaxLines=0)
Compute the preamble of the given file.
Definition: Lexer.cpp:573
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:435
const FileEntry * getFileEntryForID(FileID FID) const
Returns the FileEntry record for the provided FileID.
const AnnotatedLine * Line
static SourceLocation findLocationAfterToken(SourceLocation loc, tok::TokenKind TKind, const SourceManager &SM, const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine)
Checks that the given token is the first token that occurs after the given location (this excludes co...
Definition: Lexer.cpp:1262
bool hasLeadingEmptyMacro() const
Return true if this token has an empty macro before it.
Definition: Token.h:289
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:1118
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:126
Defines the clang::Preprocessor interface.
tok::ObjCKeywordKind getObjCKeywordID() const
Return the Objective-C keyword ID for the this identifier.
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:710
PreambleDirectiveKind
Definition: Lexer.cpp:566
SourceLocation Begin
MultipleIncludeOpt MIOpt
A state machine that detects the #ifndef-wrapping a file idiom for the multiple-include optimization...
void setEnd(SourceLocation e)
bool getCommentRetentionState() const
Definition: Preprocessor.h:942
Defines the clang::IdentifierInfo, clang::IdentifierTable, and clang::Selector interfaces.
CharSourceRange getImmediateExpansionRange(SourceLocation Loc) const
Return the start/end of the expansion information for an expansion location.
static const char * findBeginningOfLine(StringRef Buffer, unsigned Offset)
Returns the pointer that points to the beginning of line that contains the given offset, or null if the offset if invalid.
Definition: Lexer.cpp:480
bool HandleEndOfFile(Token &Result, bool isEndOfMacro=false)
Callback invoked when the lexer hits the end of the current file.
The result type of a method or function.
float __ovld __cnfn length(float p)
Return the length of vector p, i.e., sqrt(p.x2 + p.y 2 + ...)
const SourceManager & SM
Definition: Format.cpp:1570
ObjCKeywordKind
Provides a namespace for Objective-C keywords which start with an &#39;@&#39;.
Definition: TokenKinds.h:40
const ExpansionInfo & getExpansion() const
bool isRecordingPreamble() const
static CharSourceRange getCharRange(SourceRange R)
SourceManager & getSourceManager() const
Definition: Preprocessor.h:907
bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroBegin=nullptr) const
Returns true if the given MacroID location points at the beginning of the immediate macro expansion...
__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:55
Kind
Encodes a location in the source.
IdentifierInfo & get(StringRef Name)
Return the identifier token info for the specified named identifier.
static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range)
After encountering UTF-8 character C and interpreting it as an identifier character, check whether it&#39;s a homoglyph for a common non-identifier source character that is unlikely to be an intentional identifier character and warn if so.
Definition: Lexer.cpp:1503
SourceLocation createExpansionLoc(SourceLocation Loc, SourceLocation ExpansionLocStart, SourceLocation ExpansionLocEnd, unsigned TokLength, bool ExpansionIsTokenRange=true, int LoadedID=0, unsigned LoadedOffset=0)
Return a new SourceLocation that encodes the fact that a token from SpellingLoc should actually be re...
IdentifierInfo * getIdentifierInfo() const
Definition: Token.h:179
IdentifierTable & getIdentifierTable()
Definition: Preprocessor.h:910
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:1229
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1464
void setIdentifierInfo(IdentifierInfo *II)
Definition: Token.h:188
static const llvm::sys::UnicodeCharRange C99DisallowedInitialIDCharRanges[]
bool isAtStartOfLine() const
isAtStartOfLine - Return true if this token is at the start of a line.
Definition: Token.h:268
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:545
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:840
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:24
bool operator<(DeclarationName LHS, DeclarationName RHS)
Ordering on two declaration names.
tok::ObjCKeywordKind getObjCKeywordID() const
Return the ObjC keyword kind.
Definition: Lexer.cpp:66
SourceLocation getCodeCompletionLoc() const
Returns the location of the code-completion point.
SourceLocation getExpansionLocStart() const
DiagnosticBuilder Diag(const char *Loc, unsigned DiagID) const
Diag - Forwarding function for diagnostics.
Definition: Lexer.cpp:1137
__DEVICE__ void * memcpy(void *__a, const void *__b, size_t __c)
static const llvm::sys::UnicodeCharRange C99AllowedIDCharRanges[]
static StringRef getImmediateMacroName(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:967
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1014
bool inKeepCommentMode() const
inKeepCommentMode - Return true if the lexer should return comments as tokens.
Definition: Lexer.h:220
bool isTokenRange() const
Return true if the end of this range specifies the start of the last token.
static const llvm::sys::UnicodeCharRange UnicodeWhitespaceCharRanges[]
An opaque identifier used by SourceManager which refers to a source file (MemoryBuffer) along with it...
const llvm::MemoryBuffer * getBuffer(FileID FID, SourceLocation Loc, bool *Invalid=nullptr) const
Return the buffer for the specified FileID.
Dataflow Directional Tag Classes.
bool isValid() const
Return true if this is a valid SourceLocation object.
LLVM_READONLY bool isVerticalWhitespace(unsigned char c)
Returns true if this character is vertical ASCII whitespace: &#39;\n&#39;, &#39;\r&#39;.
Definition: CharInfo.h:78
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:867
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:261
static FixItHint CreateRemoval(CharSourceRange RemoveRange)
Create a code modification hint that removes the given source range.
Definition: Diagnostic.h:117
bool isHandleIdentifierCase() const
Return true if the Preprocessor::HandleIdentifier must be called on a token of this identifier...
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
LLVM_READONLY bool isIdentifierBody(unsigned char c, bool AllowDollar=false)
Returns true if this is a body character of a C identifier, which is [a-zA-Z0-9_].
Definition: CharInfo.h:58
void CodeCompleteNaturalLanguage()
Hook used by the lexer to invoke the "natural language" code completion point.
unsigned getLength() const
Definition: Token.h:129
static const char * findPlaceholderEnd(const char *CurPtr, const char *BufferEnd)
Definition: Lexer.cpp:2914
bool isMacroID() const
void setLiteralData(const char *Ptr)
Definition: Token.h:221
const char * getLiteralData() const
getLiteralData - For a literal token (numeric constant, string, etc), this returns a pointer to the s...
Definition: Token.h:217
FileID getFileID(SourceLocation SpellingLoc) const
Return the FileID for a SourceLocation.
static const llvm::sys::UnicodeCharRange CXX03AllowedIDCharRanges[]
bool isMacroArgExpansion(SourceLocation Loc, SourceLocation *StartLoc=nullptr) const
Tests whether the given source location represents a macro argument&#39;s expansion into the function-lik...
bool HandleIdentifier(Token &Identifier)
Callback invoked when the lexer reads an identifier and has filled in the tokens IdentifierInfo membe...
void CreateString(StringRef Str, Token &Tok, SourceLocation ExpansionLocStart=SourceLocation(), SourceLocation ExpansionLocEnd=SourceLocation())
Plop the specified string into a scratch buffer and set the specified token&#39;s location and length to ...
static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts)
Definition: Lexer.cpp:1441
SourceLocation getEnd() const
static FixItHint CreateInsertion(SourceLocation InsertionLoc, StringRef Code, bool BeforePreviousInsertions=false)
Create a code modification hint that inserts the given code string at a specific location.
Definition: Diagnostic.h:91
PreprocessorOptions & getPreprocessorOpts() const
Retrieve the preprocessor options used to initialize this preprocessor.
Definition: Preprocessor.h:898
Defines the clang::TokenKind enum and support functions.
const SrcMgr::SLocEntry & getSLocEntry(FileID FID, bool *Invalid=nullptr) const
static char GetTrigraphCharForLetter(char Letter)
GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, return the decoded trigraph...
Definition: Lexer.cpp:1147
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1039
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string...
Definition: Lexer.cpp:2647
__PTRDIFF_TYPE__ ptrdiff_t
A signed integer type that is the result of subtracting two pointers.
Definition: opencl-c-base.h:48
Defines the clang::SourceLocation class and associated facilities.
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:900
StringRef getRawIdentifier() const
getRawIdentifier - For a raw identifier token (i.e., an identifier lexed in raw mode), returns a reference to the text substring in the buffer if known.
Definition: Token.h:205
Not within a conflict marker.
Definition: Lexer.h:44
static char DecodeTrigraphChar(const char *CP, Lexer *L)
DecodeTrigraphChar - If the specified character is a legal trigraph when prefixed with ...
Definition: Lexer.cpp:1166
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2810
static void StringifyImpl(T &Str, char Quote)
Definition: Lexer.cpp:221
static FixItHint CreateReplacement(CharSourceRange RemoveRange, StringRef Code)
Create a code modification hint that replaces the given source range with the given code string...
Definition: Diagnostic.h:128
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode...
Definition: Lexer.h:227
bool needsCleaning() const
Return true if this token has trigraphs or escaped newlines in it.
Definition: Token.h:285
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:4399
bool isIgnored(unsigned DiagID, SourceLocation Loc) const
Determine whether the diagnostic is known to be ignored.
Definition: Diagnostic.h:820
void CodeCompleteIncludedFile(llvm::StringRef Dir, bool IsAngled)
Hook used by the lexer to invoke the "included file" code completion point.
A normal or diff3 conflict marker, initiated by at least 7 "<"s, separated by at least 7 "="s or "|"s...
Definition: Lexer.h:48
A trivial tuple used to represent a source range.
void clearFlag(TokenFlags Flag)
Unset the specified flag.
Definition: Token.h:246
bool hasUCN() const
Returns true if this token contains a universal character name.
Definition: Token.h:296
bool isPreprocessedOutput() const
Returns true if the preprocessor is responsible for generating output, false if it is producing token...
Definition: Preprocessor.h:963
void SetKeepWhitespaceMode(bool Val)
SetKeepWhitespaceMode - This method lets clients enable or disable whitespace retention mode...
Definition: Lexer.h:212
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:147
bool hasLeadingSpace() const
Return true if this token has whitespace before it.
Definition: Token.h:272
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
void setRecordedPreambleConditionalStack(ArrayRef< PPConditionalInfo > s)
This class handles loading and caching of source files into memory.
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a, __m128i __b)
Compares each of the corresponding 8-bit values of the 128-bit integer vectors for equality...
Definition: emmintrin.h:3171
void startToken()
Reset all flags to cleared.
Definition: Token.h:171
std::pair< FileID, unsigned > getDecomposedLoc(SourceLocation Loc) const
Decompose the specified location into a raw FileID + Offset pair.
static std::string Stringify(StringRef Str, bool Charify=false)
Stringify - Convert the specified string into a C string by i) escaping &#39;\&#39; and " characters and ii) ...
Definition: Lexer.cpp:246
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:124
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.