clang  10.0.0git
Lexer.cpp
Go to the documentation of this file.
1 //===- Lexer.cpp - C Language Family Lexer --------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the Lexer and Token interfaces.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "clang/Lex/Lexer.h"
14 #include "UnicodeCharSets.h"
15 #include "clang/Basic/CharInfo.h"
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 bool Lexer::skipOver(unsigned NumBytes) {
222  IsAtPhysicalStartOfLine = true;
223  IsAtStartOfLine = true;
224  if ((BufferPtr + NumBytes) > BufferEnd)
225  return true;
226  BufferPtr += NumBytes;
227  return false;
228 }
229 
230 template <typename T> static void StringifyImpl(T &Str, char Quote) {
231  typename T::size_type i = 0, e = Str.size();
232  while (i < e) {
233  if (Str[i] == '\\' || Str[i] == Quote) {
234  Str.insert(Str.begin() + i, '\\');
235  i += 2;
236  ++e;
237  } else if (Str[i] == '\n' || Str[i] == '\r') {
238  // Replace '\r\n' and '\n\r' to '\\' followed by 'n'.
239  if ((i < e - 1) && (Str[i + 1] == '\n' || Str[i + 1] == '\r') &&
240  Str[i] != Str[i + 1]) {
241  Str[i] = '\\';
242  Str[i + 1] = 'n';
243  } else {
244  // Replace '\n' and '\r' to '\\' followed by 'n'.
245  Str[i] = '\\';
246  Str.insert(Str.begin() + i + 1, 'n');
247  ++e;
248  }
249  i += 2;
250  } else
251  ++i;
252  }
253 }
254 
255 std::string Lexer::Stringify(StringRef Str, bool Charify) {
256  std::string Result = Str;
257  char Quote = Charify ? '\'' : '"';
258  StringifyImpl(Result, Quote);
259  return Result;
260 }
261 
263 
264 //===----------------------------------------------------------------------===//
265 // Token Spelling
266 //===----------------------------------------------------------------------===//
267 
268 /// Slow case of getSpelling. Extract the characters comprising the
269 /// spelling of this token from the provided input buffer.
270 static size_t getSpellingSlow(const Token &Tok, const char *BufPtr,
271  const LangOptions &LangOpts, char *Spelling) {
272  assert(Tok.needsCleaning() && "getSpellingSlow called on simple token");
273 
274  size_t Length = 0;
275  const char *BufEnd = BufPtr + Tok.getLength();
276 
277  if (tok::isStringLiteral(Tok.getKind())) {
278  // Munch the encoding-prefix and opening double-quote.
279  while (BufPtr < BufEnd) {
280  unsigned Size;
281  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
282  BufPtr += Size;
283 
284  if (Spelling[Length - 1] == '"')
285  break;
286  }
287 
288  // Raw string literals need special handling; trigraph expansion and line
289  // splicing do not occur within their d-char-sequence nor within their
290  // r-char-sequence.
291  if (Length >= 2 &&
292  Spelling[Length - 2] == 'R' && Spelling[Length - 1] == '"') {
293  // Search backwards from the end of the token to find the matching closing
294  // quote.
295  const char *RawEnd = BufEnd;
296  do --RawEnd; while (*RawEnd != '"');
297  size_t RawLength = RawEnd - BufPtr + 1;
298 
299  // Everything between the quotes is included verbatim in the spelling.
300  memcpy(Spelling + Length, BufPtr, RawLength);
301  Length += RawLength;
302  BufPtr += RawLength;
303 
304  // The rest of the token is lexed normally.
305  }
306  }
307 
308  while (BufPtr < BufEnd) {
309  unsigned Size;
310  Spelling[Length++] = Lexer::getCharAndSizeNoWarn(BufPtr, Size, LangOpts);
311  BufPtr += Size;
312  }
313 
314  assert(Length < Tok.getLength() &&
315  "NeedsCleaning flag set on token that didn't need cleaning!");
316  return Length;
317 }
318 
319 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
320 /// token are the characters used to represent the token in the source file
321 /// after trigraph expansion and escaped-newline folding. In particular, this
322 /// wants to get the true, uncanonicalized, spelling of things like digraphs
323 /// UCNs, etc.
325  SmallVectorImpl<char> &buffer,
326  const SourceManager &SM,
327  const LangOptions &options,
328  bool *invalid) {
329  // Break down the source location.
330  std::pair<FileID, unsigned> locInfo = SM.getDecomposedLoc(loc);
331 
332  // Try to the load the file buffer.
333  bool invalidTemp = false;
334  StringRef file = SM.getBufferData(locInfo.first, &invalidTemp);
335  if (invalidTemp) {
336  if (invalid) *invalid = true;
337  return {};
338  }
339 
340  const char *tokenBegin = file.data() + locInfo.second;
341 
342  // Lex from the start of the given location.
343  Lexer lexer(SM.getLocForStartOfFile(locInfo.first), options,
344  file.begin(), tokenBegin, file.end());
345  Token token;
346  lexer.LexFromRawLexer(token);
347 
348  unsigned length = token.getLength();
349 
350  // Common case: no need for cleaning.
351  if (!token.needsCleaning())
352  return StringRef(tokenBegin, length);
353 
354  // Hard case, we need to relex the characters into the string.
355  buffer.resize(length);
356  buffer.resize(getSpellingSlow(token, tokenBegin, options, buffer.data()));
357  return StringRef(buffer.data(), buffer.size());
358 }
359 
360 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
361 /// token are the characters used to represent the token in the source file
362 /// after trigraph expansion and escaped-newline folding. In particular, this
363 /// wants to get the true, uncanonicalized, spelling of things like digraphs
364 /// UCNs, etc.
365 std::string Lexer::getSpelling(const Token &Tok, const SourceManager &SourceMgr,
366  const LangOptions &LangOpts, bool *Invalid) {
367  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
368 
369  bool CharDataInvalid = false;
370  const char *TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
371  &CharDataInvalid);
372  if (Invalid)
373  *Invalid = CharDataInvalid;
374  if (CharDataInvalid)
375  return {};
376 
377  // If this token contains nothing interesting, return it directly.
378  if (!Tok.needsCleaning())
379  return std::string(TokStart, TokStart + Tok.getLength());
380 
381  std::string Result;
382  Result.resize(Tok.getLength());
383  Result.resize(getSpellingSlow(Tok, TokStart, LangOpts, &*Result.begin()));
384  return Result;
385 }
386 
387 /// getSpelling - This method is used to get the spelling of a token into a
388 /// preallocated buffer, instead of as an std::string. The caller is required
389 /// to allocate enough space for the token, which is guaranteed to be at least
390 /// Tok.getLength() bytes long. The actual length of the token is returned.
391 ///
392 /// Note that this method may do two possible things: it may either fill in
393 /// the buffer specified with characters, or it may *change the input pointer*
394 /// to point to a constant buffer with the data already in it (avoiding a
395 /// copy). The caller is not allowed to modify the returned buffer pointer
396 /// if an internal buffer is returned.
397 unsigned Lexer::getSpelling(const Token &Tok, const char *&Buffer,
398  const SourceManager &SourceMgr,
399  const LangOptions &LangOpts, bool *Invalid) {
400  assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
401 
402  const char *TokStart = nullptr;
403  // NOTE: this has to be checked *before* testing for an IdentifierInfo.
404  if (Tok.is(tok::raw_identifier))
405  TokStart = Tok.getRawIdentifier().data();
406  else if (!Tok.hasUCN()) {
407  if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
408  // Just return the string from the identifier table, which is very quick.
409  Buffer = II->getNameStart();
410  return II->getLength();
411  }
412  }
413 
414  // NOTE: this can be checked even after testing for an IdentifierInfo.
415  if (Tok.isLiteral())
416  TokStart = Tok.getLiteralData();
417 
418  if (!TokStart) {
419  // Compute the start of the token in the input lexer buffer.
420  bool CharDataInvalid = false;
421  TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
422  if (Invalid)
423  *Invalid = CharDataInvalid;
424  if (CharDataInvalid) {
425  Buffer = "";
426  return 0;
427  }
428  }
429 
430  // If this token contains nothing interesting, return it directly.
431  if (!Tok.needsCleaning()) {
432  Buffer = TokStart;
433  return Tok.getLength();
434  }
435 
436  // Otherwise, hard case, relex the characters into the string.
437  return getSpellingSlow(Tok, TokStart, LangOpts, const_cast<char*>(Buffer));
438 }
439 
440 /// MeasureTokenLength - Relex the token at the specified location and return
441 /// its length in bytes in the input file. If the token needs cleaning (e.g.
442 /// includes a trigraph or an escaped newline) then this count includes bytes
443 /// that are part of that.
445  const SourceManager &SM,
446  const LangOptions &LangOpts) {
447  Token TheTok;
448  if (getRawToken(Loc, TheTok, SM, LangOpts))
449  return 0;
450  return TheTok.getLength();
451 }
452 
453 /// Relex the token at the specified location.
454 /// \returns true if there was a failure, false on success.
456  const SourceManager &SM,
457  const LangOptions &LangOpts,
458  bool IgnoreWhiteSpace) {
459  // TODO: this could be special cased for common tokens like identifiers, ')',
460  // etc to make this faster, if it mattered. Just look at StrData[0] to handle
461  // all obviously single-char tokens. This could use
462  // Lexer::isObviouslySimpleCharacter for example to handle identifiers or
463  // something.
464 
465  // If this comes from a macro expansion, we really do want the macro name, not
466  // the token this macro expanded to.
467  Loc = SM.getExpansionLoc(Loc);
468  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
469  bool Invalid = false;
470  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
471  if (Invalid)
472  return true;
473 
474  const char *StrData = Buffer.data()+LocInfo.second;
475 
476  if (!IgnoreWhiteSpace && isWhitespace(StrData[0]))
477  return true;
478 
479  // Create a lexer starting at the beginning of this token.
480  Lexer TheLexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts,
481  Buffer.begin(), StrData, Buffer.end());
482  TheLexer.SetCommentRetentionState(true);
483  TheLexer.LexFromRawLexer(Result);
484  return false;
485 }
486 
487 /// Returns the pointer that points to the beginning of line that contains
488 /// the given offset, or null if the offset if invalid.
489 static const char *findBeginningOfLine(StringRef Buffer, unsigned Offset) {
490  const char *BufStart = Buffer.data();
491  if (Offset >= Buffer.size())
492  return nullptr;
493 
494  const char *LexStart = BufStart + Offset;
495  for (; LexStart != BufStart; --LexStart) {
496  if (isVerticalWhitespace(LexStart[0]) &&
497  !Lexer::isNewLineEscaped(BufStart, LexStart)) {
498  // LexStart should point at first character of logical line.
499  ++LexStart;
500  break;
501  }
502  }
503  return LexStart;
504 }
505 
507  const SourceManager &SM,
508  const LangOptions &LangOpts) {
509  assert(Loc.isFileID());
510  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
511  if (LocInfo.first.isInvalid())
512  return Loc;
513 
514  bool Invalid = false;
515  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
516  if (Invalid)
517  return Loc;
518 
519  // Back up from the current location until we hit the beginning of a line
520  // (or the buffer). We'll relex from that point.
521  const char *StrData = Buffer.data() + LocInfo.second;
522  const char *LexStart = findBeginningOfLine(Buffer, LocInfo.second);
523  if (!LexStart || LexStart == StrData)
524  return Loc;
525 
526  // Create a lexer starting at the beginning of this token.
527  SourceLocation LexerStartLoc = Loc.getLocWithOffset(-LocInfo.second);
528  Lexer TheLexer(LexerStartLoc, LangOpts, Buffer.data(), LexStart,
529  Buffer.end());
530  TheLexer.SetCommentRetentionState(true);
531 
532  // Lex tokens until we find the token that contains the source location.
533  Token TheTok;
534  do {
535  TheLexer.LexFromRawLexer(TheTok);
536 
537  if (TheLexer.getBufferLocation() > StrData) {
538  // Lexing this token has taken the lexer past the source location we're
539  // looking for. If the current token encompasses our source location,
540  // return the beginning of that token.
541  if (TheLexer.getBufferLocation() - TheTok.getLength() <= StrData)
542  return TheTok.getLocation();
543 
544  // We ended up skipping over the source location entirely, which means
545  // that it points into whitespace. We're done here.
546  break;
547  }
548  } while (TheTok.getKind() != tok::eof);
549 
550  // We've passed our source location; just return the original source location.
551  return Loc;
552 }
553 
555  const SourceManager &SM,
556  const LangOptions &LangOpts) {
557  if (Loc.isFileID())
558  return getBeginningOfFileToken(Loc, SM, LangOpts);
559 
560  if (!SM.isMacroArgExpansion(Loc))
561  return Loc;
562 
563  SourceLocation FileLoc = SM.getSpellingLoc(Loc);
564  SourceLocation BeginFileLoc = getBeginningOfFileToken(FileLoc, SM, LangOpts);
565  std::pair<FileID, unsigned> FileLocInfo = SM.getDecomposedLoc(FileLoc);
566  std::pair<FileID, unsigned> BeginFileLocInfo =
567  SM.getDecomposedLoc(BeginFileLoc);
568  assert(FileLocInfo.first == BeginFileLocInfo.first &&
569  FileLocInfo.second >= BeginFileLocInfo.second);
570  return Loc.getLocWithOffset(BeginFileLocInfo.second - FileLocInfo.second);
571 }
572 
573 namespace {
574 
576  PDK_Skipped,
577  PDK_Unknown
578 };
579 
580 } // namespace
581 
583  const LangOptions &LangOpts,
584  unsigned MaxLines) {
585  // Create a lexer starting at the beginning of the file. Note that we use a
586  // "fake" file source location at offset 1 so that the lexer will track our
587  // position within the file.
588  const unsigned StartOffset = 1;
589  SourceLocation FileLoc = SourceLocation::getFromRawEncoding(StartOffset);
590  Lexer TheLexer(FileLoc, LangOpts, Buffer.begin(), Buffer.begin(),
591  Buffer.end());
592  TheLexer.SetCommentRetentionState(true);
593 
594  bool InPreprocessorDirective = false;
595  Token TheTok;
596  SourceLocation ActiveCommentLoc;
597 
598  unsigned MaxLineOffset = 0;
599  if (MaxLines) {
600  const char *CurPtr = Buffer.begin();
601  unsigned CurLine = 0;
602  while (CurPtr != Buffer.end()) {
603  char ch = *CurPtr++;
604  if (ch == '\n') {
605  ++CurLine;
606  if (CurLine == MaxLines)
607  break;
608  }
609  }
610  if (CurPtr != Buffer.end())
611  MaxLineOffset = CurPtr - Buffer.begin();
612  }
613 
614  do {
615  TheLexer.LexFromRawLexer(TheTok);
616 
617  if (InPreprocessorDirective) {
618  // If we've hit the end of the file, we're done.
619  if (TheTok.getKind() == tok::eof) {
620  break;
621  }
622 
623  // If we haven't hit the end of the preprocessor directive, skip this
624  // token.
625  if (!TheTok.isAtStartOfLine())
626  continue;
627 
628  // We've passed the end of the preprocessor directive, and will look
629  // at this token again below.
630  InPreprocessorDirective = false;
631  }
632 
633  // Keep track of the # of lines in the preamble.
634  if (TheTok.isAtStartOfLine()) {
635  unsigned TokOffset = TheTok.getLocation().getRawEncoding() - StartOffset;
636 
637  // If we were asked to limit the number of lines in the preamble,
638  // and we're about to exceed that limit, we're done.
639  if (MaxLineOffset && TokOffset >= MaxLineOffset)
640  break;
641  }
642 
643  // Comments are okay; skip over them.
644  if (TheTok.getKind() == tok::comment) {
645  if (ActiveCommentLoc.isInvalid())
646  ActiveCommentLoc = TheTok.getLocation();
647  continue;
648  }
649 
650  if (TheTok.isAtStartOfLine() && TheTok.getKind() == tok::hash) {
651  // This is the start of a preprocessor directive.
652  Token HashTok = TheTok;
653  InPreprocessorDirective = true;
654  ActiveCommentLoc = SourceLocation();
655 
656  // Figure out which directive this is. Since we're lexing raw tokens,
657  // we don't have an identifier table available. Instead, just look at
658  // the raw identifier to recognize and categorize preprocessor directives.
659  TheLexer.LexFromRawLexer(TheTok);
660  if (TheTok.getKind() == tok::raw_identifier && !TheTok.needsCleaning()) {
661  StringRef Keyword = TheTok.getRawIdentifier();
663  = llvm::StringSwitch<PreambleDirectiveKind>(Keyword)
664  .Case("include", PDK_Skipped)
665  .Case("__include_macros", PDK_Skipped)
666  .Case("define", PDK_Skipped)
667  .Case("undef", PDK_Skipped)
668  .Case("line", PDK_Skipped)
669  .Case("error", PDK_Skipped)
670  .Case("pragma", PDK_Skipped)
671  .Case("import", PDK_Skipped)
672  .Case("include_next", PDK_Skipped)
673  .Case("warning", PDK_Skipped)
674  .Case("ident", PDK_Skipped)
675  .Case("sccs", PDK_Skipped)
676  .Case("assert", PDK_Skipped)
677  .Case("unassert", PDK_Skipped)
678  .Case("if", PDK_Skipped)
679  .Case("ifdef", PDK_Skipped)
680  .Case("ifndef", PDK_Skipped)
681  .Case("elif", PDK_Skipped)
682  .Case("else", PDK_Skipped)
683  .Case("endif", PDK_Skipped)
684  .Default(PDK_Unknown);
685 
686  switch (PDK) {
687  case PDK_Skipped:
688  continue;
689 
690  case PDK_Unknown:
691  // We don't know what this directive is; stop at the '#'.
692  break;
693  }
694  }
695 
696  // We only end up here if we didn't recognize the preprocessor
697  // directive or it was one that can't occur in the preamble at this
698  // point. Roll back the current token to the location of the '#'.
699  TheTok = HashTok;
700  }
701 
702  // We hit a token that we don't recognize as being in the
703  // "preprocessing only" part of the file, so we're no longer in
704  // the preamble.
705  break;
706  } while (true);
707 
709  if (ActiveCommentLoc.isValid())
710  End = ActiveCommentLoc; // don't truncate a decl comment.
711  else
712  End = TheTok.getLocation();
713 
714  return PreambleBounds(End.getRawEncoding() - FileLoc.getRawEncoding(),
715  TheTok.isAtStartOfLine());
716 }
717 
718 unsigned Lexer::getTokenPrefixLength(SourceLocation TokStart, unsigned CharNo,
719  const SourceManager &SM,
720  const LangOptions &LangOpts) {
721  // Figure out how many physical characters away the specified expansion
722  // character is. This needs to take into consideration newlines and
723  // trigraphs.
724  bool Invalid = false;
725  const char *TokPtr = SM.getCharacterData(TokStart, &Invalid);
726 
727  // If they request the first char of the token, we're trivially done.
728  if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
729  return 0;
730 
731  unsigned PhysOffset = 0;
732 
733  // The usual case is that tokens don't contain anything interesting. Skip
734  // over the uninteresting characters. If a token only consists of simple
735  // chars, this method is extremely fast.
736  while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
737  if (CharNo == 0)
738  return PhysOffset;
739  ++TokPtr;
740  --CharNo;
741  ++PhysOffset;
742  }
743 
744  // If we have a character that may be a trigraph or escaped newline, use a
745  // lexer to parse it correctly.
746  for (; CharNo; --CharNo) {
747  unsigned Size;
748  Lexer::getCharAndSizeNoWarn(TokPtr, Size, LangOpts);
749  TokPtr += Size;
750  PhysOffset += Size;
751  }
752 
753  // Final detail: if we end up on an escaped newline, we want to return the
754  // location of the actual byte of the token. For example foo<newline>bar
755  // advanced by 3 should return the location of b, not of \\. One compounding
756  // detail of this is that the escape may be made by a trigraph.
757  if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
758  PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
759 
760  return PhysOffset;
761 }
762 
763 /// Computes the source location just past the end of the
764 /// token at this source location.
765 ///
766 /// This routine can be used to produce a source location that
767 /// points just past the end of the token referenced by \p Loc, and
768 /// is generally used when a diagnostic needs to point just after a
769 /// token where it expected something different that it received. If
770 /// the returned source location would not be meaningful (e.g., if
771 /// it points into a macro), this routine returns an invalid
772 /// source location.
773 ///
774 /// \param Offset an offset from the end of the token, where the source
775 /// location should refer to. The default offset (0) produces a source
776 /// location pointing just past the end of the token; an offset of 1 produces
777 /// a source location pointing to the last character in the token, etc.
779  const SourceManager &SM,
780  const LangOptions &LangOpts) {
781  if (Loc.isInvalid())
782  return {};
783 
784  if (Loc.isMacroID()) {
785  if (Offset > 0 || !isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
786  return {}; // Points inside the macro expansion.
787  }
788 
789  unsigned Len = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
790  if (Len > Offset)
791  Len = Len - Offset;
792  else
793  return Loc;
794 
795  return Loc.getLocWithOffset(Len);
796 }
797 
798 /// Returns true if the given MacroID location points at the first
799 /// token of the macro expansion.
801  const SourceManager &SM,
802  const LangOptions &LangOpts,
803  SourceLocation *MacroBegin) {
804  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
805 
806  SourceLocation expansionLoc;
807  if (!SM.isAtStartOfImmediateMacroExpansion(loc, &expansionLoc))
808  return false;
809 
810  if (expansionLoc.isFileID()) {
811  // No other macro expansions, this is the first.
812  if (MacroBegin)
813  *MacroBegin = expansionLoc;
814  return true;
815  }
816 
817  return isAtStartOfMacroExpansion(expansionLoc, SM, LangOpts, MacroBegin);
818 }
819 
820 /// Returns true if the given MacroID location points at the last
821 /// token of the macro expansion.
823  const SourceManager &SM,
824  const LangOptions &LangOpts,
825  SourceLocation *MacroEnd) {
826  assert(loc.isValid() && loc.isMacroID() && "Expected a valid macro loc");
827 
828  SourceLocation spellLoc = SM.getSpellingLoc(loc);
829  unsigned tokLen = MeasureTokenLength(spellLoc, SM, LangOpts);
830  if (tokLen == 0)
831  return false;
832 
833  SourceLocation afterLoc = loc.getLocWithOffset(tokLen);
834  SourceLocation expansionLoc;
835  if (!SM.isAtEndOfImmediateMacroExpansion(afterLoc, &expansionLoc))
836  return false;
837 
838  if (expansionLoc.isFileID()) {
839  // No other macro expansions.
840  if (MacroEnd)
841  *MacroEnd = expansionLoc;
842  return true;
843  }
844 
845  return isAtEndOfMacroExpansion(expansionLoc, SM, LangOpts, MacroEnd);
846 }
847 
849  const SourceManager &SM,
850  const LangOptions &LangOpts) {
851  SourceLocation Begin = Range.getBegin();
852  SourceLocation End = Range.getEnd();
853  assert(Begin.isFileID() && End.isFileID());
854  if (Range.isTokenRange()) {
855  End = Lexer::getLocForEndOfToken(End, 0, SM,LangOpts);
856  if (End.isInvalid())
857  return {};
858  }
859 
860  // Break down the source locations.
861  FileID FID;
862  unsigned BeginOffs;
863  std::tie(FID, BeginOffs) = SM.getDecomposedLoc(Begin);
864  if (FID.isInvalid())
865  return {};
866 
867  unsigned EndOffs;
868  if (!SM.isInFileID(End, FID, &EndOffs) ||
869  BeginOffs > EndOffs)
870  return {};
871 
872  return CharSourceRange::getCharRange(Begin, End);
873 }
874 
876  const SourceManager &SM,
877  const LangOptions &LangOpts) {
878  SourceLocation Begin = Range.getBegin();
879  SourceLocation End = Range.getEnd();
880  if (Begin.isInvalid() || End.isInvalid())
881  return {};
882 
883  if (Begin.isFileID() && End.isFileID())
884  return makeRangeFromFileLocs(Range, SM, LangOpts);
885 
886  if (Begin.isMacroID() && End.isFileID()) {
887  if (!isAtStartOfMacroExpansion(Begin, SM, LangOpts, &Begin))
888  return {};
889  Range.setBegin(Begin);
890  return makeRangeFromFileLocs(Range, SM, LangOpts);
891  }
892 
893  if (Begin.isFileID() && End.isMacroID()) {
894  if ((Range.isTokenRange() && !isAtEndOfMacroExpansion(End, SM, LangOpts,
895  &End)) ||
896  (Range.isCharRange() && !isAtStartOfMacroExpansion(End, SM, LangOpts,
897  &End)))
898  return {};
899  Range.setEnd(End);
900  return makeRangeFromFileLocs(Range, SM, LangOpts);
901  }
902 
903  assert(Begin.isMacroID() && End.isMacroID());
904  SourceLocation MacroBegin, MacroEnd;
905  if (isAtStartOfMacroExpansion(Begin, SM, LangOpts, &MacroBegin) &&
906  ((Range.isTokenRange() && isAtEndOfMacroExpansion(End, SM, LangOpts,
907  &MacroEnd)) ||
908  (Range.isCharRange() && isAtStartOfMacroExpansion(End, SM, LangOpts,
909  &MacroEnd)))) {
910  Range.setBegin(MacroBegin);
911  Range.setEnd(MacroEnd);
912  return makeRangeFromFileLocs(Range, SM, LangOpts);
913  }
914 
915  bool Invalid = false;
916  const SrcMgr::SLocEntry &BeginEntry = SM.getSLocEntry(SM.getFileID(Begin),
917  &Invalid);
918  if (Invalid)
919  return {};
920 
921  if (BeginEntry.getExpansion().isMacroArgExpansion()) {
922  const SrcMgr::SLocEntry &EndEntry = SM.getSLocEntry(SM.getFileID(End),
923  &Invalid);
924  if (Invalid)
925  return {};
926 
927  if (EndEntry.getExpansion().isMacroArgExpansion() &&
928  BeginEntry.getExpansion().getExpansionLocStart() ==
929  EndEntry.getExpansion().getExpansionLocStart()) {
930  Range.setBegin(SM.getImmediateSpellingLoc(Begin));
931  Range.setEnd(SM.getImmediateSpellingLoc(End));
932  return makeFileCharRange(Range, SM, LangOpts);
933  }
934  }
935 
936  return {};
937 }
938 
940  const SourceManager &SM,
941  const LangOptions &LangOpts,
942  bool *Invalid) {
943  Range = makeFileCharRange(Range, SM, LangOpts);
944  if (Range.isInvalid()) {
945  if (Invalid) *Invalid = true;
946  return {};
947  }
948 
949  // Break down the source location.
950  std::pair<FileID, unsigned> beginInfo = SM.getDecomposedLoc(Range.getBegin());
951  if (beginInfo.first.isInvalid()) {
952  if (Invalid) *Invalid = true;
953  return {};
954  }
955 
956  unsigned EndOffs;
957  if (!SM.isInFileID(Range.getEnd(), beginInfo.first, &EndOffs) ||
958  beginInfo.second > EndOffs) {
959  if (Invalid) *Invalid = true;
960  return {};
961  }
962 
963  // Try to the load the file buffer.
964  bool invalidTemp = false;
965  StringRef file = SM.getBufferData(beginInfo.first, &invalidTemp);
966  if (invalidTemp) {
967  if (Invalid) *Invalid = true;
968  return {};
969  }
970 
971  if (Invalid) *Invalid = false;
972  return file.substr(beginInfo.second, EndOffs - beginInfo.second);
973 }
974 
976  const SourceManager &SM,
977  const LangOptions &LangOpts) {
978  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
979 
980  // Find the location of the immediate macro expansion.
981  while (true) {
982  FileID FID = SM.getFileID(Loc);
983  const SrcMgr::SLocEntry *E = &SM.getSLocEntry(FID);
984  const SrcMgr::ExpansionInfo &Expansion = E->getExpansion();
985  Loc = Expansion.getExpansionLocStart();
986  if (!Expansion.isMacroArgExpansion())
987  break;
988 
989  // For macro arguments we need to check that the argument did not come
990  // from an inner macro, e.g: "MAC1( MAC2(foo) )"
991 
992  // Loc points to the argument id of the macro definition, move to the
993  // macro expansion.
994  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
995  SourceLocation SpellLoc = Expansion.getSpellingLoc();
996  if (SpellLoc.isFileID())
997  break; // No inner macro.
998 
999  // If spelling location resides in the same FileID as macro expansion
1000  // location, it means there is no inner macro.
1001  FileID MacroFID = SM.getFileID(Loc);
1002  if (SM.isInFileID(SpellLoc, MacroFID))
1003  break;
1004 
1005  // Argument came from inner macro.
1006  Loc = SpellLoc;
1007  }
1008 
1009  // Find the spelling location of the start of the non-argument expansion
1010  // range. This is where the macro name was spelled in order to begin
1011  // expanding this macro.
1012  Loc = SM.getSpellingLoc(Loc);
1013 
1014  // Dig out the buffer where the macro name was spelled and the extents of the
1015  // name so that we can render it into the expansion note.
1016  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1017  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1018  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1019  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1020 }
1021 
1023  SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts) {
1024  assert(Loc.isMacroID() && "Only reasonable to call this on macros");
1025  // Walk past macro argument expansions.
1026  while (SM.isMacroArgExpansion(Loc))
1027  Loc = SM.getImmediateExpansionRange(Loc).getBegin();
1028 
1029  // If the macro's spelling has no FileID, then it's actually a token paste
1030  // or stringization (or similar) and not a macro at all.
1031  if (!SM.getFileEntryForID(SM.getFileID(SM.getSpellingLoc(Loc))))
1032  return {};
1033 
1034  // Find the spelling location of the start of the non-argument expansion
1035  // range. This is where the macro name was spelled in order to begin
1036  // expanding this macro.
1038 
1039  // Dig out the buffer where the macro name was spelled and the extents of the
1040  // name so that we can render it into the expansion note.
1041  std::pair<FileID, unsigned> ExpansionInfo = SM.getDecomposedLoc(Loc);
1042  unsigned MacroTokenLength = Lexer::MeasureTokenLength(Loc, SM, LangOpts);
1043  StringRef ExpansionBuffer = SM.getBufferData(ExpansionInfo.first);
1044  return ExpansionBuffer.substr(ExpansionInfo.second, MacroTokenLength);
1045 }
1046 
1047 bool Lexer::isIdentifierBodyChar(char c, const LangOptions &LangOpts) {
1048  return isIdentifierBody(c, LangOpts.DollarIdents);
1049 }
1050 
1051 bool Lexer::isNewLineEscaped(const char *BufferStart, const char *Str) {
1052  assert(isVerticalWhitespace(Str[0]));
1053  if (Str - 1 < BufferStart)
1054  return false;
1055 
1056  if ((Str[0] == '\n' && Str[-1] == '\r') ||
1057  (Str[0] == '\r' && Str[-1] == '\n')) {
1058  if (Str - 2 < BufferStart)
1059  return false;
1060  --Str;
1061  }
1062  --Str;
1063 
1064  // Rewind to first non-space character:
1065  while (Str > BufferStart && isHorizontalWhitespace(*Str))
1066  --Str;
1067 
1068  return *Str == '\\';
1069 }
1070 
1072  const SourceManager &SM) {
1073  if (Loc.isInvalid() || Loc.isMacroID())
1074  return {};
1075  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1076  if (LocInfo.first.isInvalid())
1077  return {};
1078  bool Invalid = false;
1079  StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
1080  if (Invalid)
1081  return {};
1082  const char *Line = findBeginningOfLine(Buffer, LocInfo.second);
1083  if (!Line)
1084  return {};
1085  StringRef Rest = Buffer.substr(Line - Buffer.data());
1086  size_t NumWhitespaceChars = Rest.find_first_not_of(" \t");
1087  return NumWhitespaceChars == StringRef::npos
1088  ? ""
1089  : Rest.take_front(NumWhitespaceChars);
1090 }
1091 
1092 //===----------------------------------------------------------------------===//
1093 // Diagnostics forwarding code.
1094 //===----------------------------------------------------------------------===//
1095 
1096 /// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the
1097 /// lexer buffer was all expanded at a single point, perform the mapping.
1098 /// This is currently only used for _Pragma implementation, so it is the slow
1099 /// path of the hot getSourceLocation method. Do not allow it to be inlined.
1100 static LLVM_ATTRIBUTE_NOINLINE SourceLocation GetMappedTokenLoc(
1101  Preprocessor &PP, SourceLocation FileLoc, unsigned CharNo, unsigned TokLen);
1103  SourceLocation FileLoc,
1104  unsigned CharNo, unsigned TokLen) {
1105  assert(FileLoc.isMacroID() && "Must be a macro expansion");
1106 
1107  // Otherwise, we're lexing "mapped tokens". This is used for things like
1108  // _Pragma handling. Combine the expansion location of FileLoc with the
1109  // spelling location.
1111 
1112  // Create a new SLoc which is expanded from Expansion(FileLoc) but whose
1113  // characters come from spelling(FileLoc)+Offset.
1114  SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc);
1115  SpellingLoc = SpellingLoc.getLocWithOffset(CharNo);
1116 
1117  // Figure out the expansion loc range, which is the range covered by the
1118  // original _Pragma(...) sequence.
1119  CharSourceRange II = SM.getImmediateExpansionRange(FileLoc);
1120 
1121  return SM.createExpansionLoc(SpellingLoc, II.getBegin(), II.getEnd(), TokLen);
1122 }
1123 
1124 /// getSourceLocation - Return a source location identifier for the specified
1125 /// offset in the current file.
1127  unsigned TokLen) const {
1128  assert(Loc >= BufferStart && Loc <= BufferEnd &&
1129  "Location out of range for this buffer!");
1130 
1131  // In the normal case, we're just lexing from a simple file buffer, return
1132  // the file id from FileLoc with the offset specified.
1133  unsigned CharNo = Loc-BufferStart;
1134  if (FileLoc.isFileID())
1135  return FileLoc.getLocWithOffset(CharNo);
1136 
1137  // Otherwise, this is the _Pragma lexer case, which pretends that all of the
1138  // tokens are lexed from where the _Pragma was defined.
1139  assert(PP && "This doesn't work on raw lexers");
1140  return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen);
1141 }
1142 
1143 /// Diag - Forwarding function for diagnostics. This translate a source
1144 /// position in the current buffer into a SourceLocation object for rendering.
1145 DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const {
1146  return PP->Diag(getSourceLocation(Loc), DiagID);
1147 }
1148 
1149 //===----------------------------------------------------------------------===//
1150 // Trigraph and Escaped Newline Handling Code.
1151 //===----------------------------------------------------------------------===//
1152 
1153 /// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair,
1154 /// return the decoded trigraph letter it corresponds to, or '\0' if nothing.
1155 static char GetTrigraphCharForLetter(char Letter) {
1156  switch (Letter) {
1157  default: return 0;
1158  case '=': return '#';
1159  case ')': return ']';
1160  case '(': return '[';
1161  case '!': return '|';
1162  case '\'': return '^';
1163  case '>': return '}';
1164  case '/': return '\\';
1165  case '<': return '{';
1166  case '-': return '~';
1167  }
1168 }
1169 
1170 /// DecodeTrigraphChar - If the specified character is a legal trigraph when
1171 /// prefixed with ??, emit a trigraph warning. If trigraphs are enabled,
1172 /// return the result character. Finally, emit a warning about trigraph use
1173 /// whether trigraphs are enabled or not.
1174 static char DecodeTrigraphChar(const char *CP, Lexer *L) {
1175  char Res = GetTrigraphCharForLetter(*CP);
1176  if (!Res || !L) return Res;
1177 
1178  if (!L->getLangOpts().Trigraphs) {
1179  if (!L->isLexingRawMode())
1180  L->Diag(CP-2, diag::trigraph_ignored);
1181  return 0;
1182  }
1183 
1184  if (!L->isLexingRawMode())
1185  L->Diag(CP-2, diag::trigraph_converted) << StringRef(&Res, 1);
1186  return Res;
1187 }
1188 
1189 /// getEscapedNewLineSize - Return the size of the specified escaped newline,
1190 /// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a
1191 /// trigraph equivalent on entry to this function.
1192 unsigned Lexer::getEscapedNewLineSize(const char *Ptr) {
1193  unsigned Size = 0;
1194  while (isWhitespace(Ptr[Size])) {
1195  ++Size;
1196 
1197  if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r')
1198  continue;
1199 
1200  // If this is a \r\n or \n\r, skip the other half.
1201  if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') &&
1202  Ptr[Size-1] != Ptr[Size])
1203  ++Size;
1204 
1205  return Size;
1206  }
1207 
1208  // Not an escaped newline, must be a \t or something else.
1209  return 0;
1210 }
1211 
1212 /// SkipEscapedNewLines - If P points to an escaped newline (or a series of
1213 /// them), skip over them and return the first non-escaped-newline found,
1214 /// otherwise return P.
1215 const char *Lexer::SkipEscapedNewLines(const char *P) {
1216  while (true) {
1217  const char *AfterEscape;
1218  if (*P == '\\') {
1219  AfterEscape = P+1;
1220  } else if (*P == '?') {
1221  // If not a trigraph for escape, bail out.
1222  if (P[1] != '?' || P[2] != '/')
1223  return P;
1224  // FIXME: Take LangOpts into account; the language might not
1225  // support trigraphs.
1226  AfterEscape = P+3;
1227  } else {
1228  return P;
1229  }
1230 
1231  unsigned NewLineSize = Lexer::getEscapedNewLineSize(AfterEscape);
1232  if (NewLineSize == 0) return P;
1233  P = AfterEscape+NewLineSize;
1234  }
1235 }
1236 
1238  const SourceManager &SM,
1239  const LangOptions &LangOpts) {
1240  if (Loc.isMacroID()) {
1241  if (!Lexer::isAtEndOfMacroExpansion(Loc, SM, LangOpts, &Loc))
1242  return None;
1243  }
1244  Loc = Lexer::getLocForEndOfToken(Loc, 0, SM, LangOpts);
1245 
1246  // Break down the source location.
1247  std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc);
1248 
1249  // Try to load the file buffer.
1250  bool InvalidTemp = false;
1251  StringRef File = SM.getBufferData(LocInfo.first, &InvalidTemp);
1252  if (InvalidTemp)
1253  return None;
1254 
1255  const char *TokenBegin = File.data() + LocInfo.second;
1256 
1257  // Lex from the start of the given location.
1258  Lexer lexer(SM.getLocForStartOfFile(LocInfo.first), LangOpts, File.begin(),
1259  TokenBegin, File.end());
1260  // Find the token.
1261  Token Tok;
1262  lexer.LexFromRawLexer(Tok);
1263  return Tok;
1264 }
1265 
1266 /// Checks that the given token is the first token that occurs after the
1267 /// given location (this excludes comments and whitespace). Returns the location
1268 /// immediately after the specified token. If the token is not found or the
1269 /// location is inside a macro, the returned source location will be invalid.
1271  SourceLocation Loc, tok::TokenKind TKind, const SourceManager &SM,
1272  const LangOptions &LangOpts, bool SkipTrailingWhitespaceAndNewLine) {
1273  Optional<Token> Tok = findNextToken(Loc, SM, LangOpts);
1274  if (!Tok || Tok->isNot(TKind))
1275  return {};
1276  SourceLocation TokenLoc = Tok->getLocation();
1277 
1278  // Calculate how much whitespace needs to be skipped if any.
1279  unsigned NumWhitespaceChars = 0;
1280  if (SkipTrailingWhitespaceAndNewLine) {
1281  const char *TokenEnd = SM.getCharacterData(TokenLoc) + Tok->getLength();
1282  unsigned char C = *TokenEnd;
1283  while (isHorizontalWhitespace(C)) {
1284  C = *(++TokenEnd);
1285  NumWhitespaceChars++;
1286  }
1287 
1288  // Skip \r, \n, \r\n, or \n\r
1289  if (C == '\n' || C == '\r') {
1290  char PrevC = C;
1291  C = *(++TokenEnd);
1292  NumWhitespaceChars++;
1293  if ((C == '\n' || C == '\r') && C != PrevC)
1294  NumWhitespaceChars++;
1295  }
1296  }
1297 
1298  return TokenLoc.getLocWithOffset(Tok->getLength() + NumWhitespaceChars);
1299 }
1300 
1301 /// getCharAndSizeSlow - Peek a single 'character' from the specified buffer,
1302 /// get its size, and return it. This is tricky in several cases:
1303 /// 1. If currently at the start of a trigraph, we warn about the trigraph,
1304 /// then either return the trigraph (skipping 3 chars) or the '?',
1305 /// depending on whether trigraphs are enabled or not.
1306 /// 2. If this is an escaped newline (potentially with whitespace between
1307 /// the backslash and newline), implicitly skip the newline and return
1308 /// the char after it.
1309 ///
1310 /// This handles the slow/uncommon case of the getCharAndSize method. Here we
1311 /// know that we can accumulate into Size, and that we have already incremented
1312 /// Ptr by Size bytes.
1313 ///
1314 /// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should
1315 /// be updated to match.
1316 char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size,
1317  Token *Tok) {
1318  // If we have a slash, look for an escaped newline.
1319  if (Ptr[0] == '\\') {
1320  ++Size;
1321  ++Ptr;
1322 Slash:
1323  // Common case, backslash-char where the char is not whitespace.
1324  if (!isWhitespace(Ptr[0])) return '\\';
1325 
1326  // See if we have optional whitespace characters between the slash and
1327  // newline.
1328  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1329  // Remember that this token needs to be cleaned.
1330  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1331 
1332  // Warn if there was whitespace between the backslash and newline.
1333  if (Ptr[0] != '\n' && Ptr[0] != '\r' && Tok && !isLexingRawMode())
1334  Diag(Ptr, diag::backslash_newline_space);
1335 
1336  // Found backslash<whitespace><newline>. Parse the char after it.
1337  Size += EscapedNewLineSize;
1338  Ptr += EscapedNewLineSize;
1339 
1340  // Use slow version to accumulate a correct size field.
1341  return getCharAndSizeSlow(Ptr, Size, Tok);
1342  }
1343 
1344  // Otherwise, this is not an escaped newline, just return the slash.
1345  return '\\';
1346  }
1347 
1348  // If this is a trigraph, process it.
1349  if (Ptr[0] == '?' && Ptr[1] == '?') {
1350  // If this is actually a legal trigraph (not something like "??x"), emit
1351  // a trigraph warning. If so, and if trigraphs are enabled, return it.
1352  if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : nullptr)) {
1353  // Remember that this token needs to be cleaned.
1354  if (Tok) Tok->setFlag(Token::NeedsCleaning);
1355 
1356  Ptr += 3;
1357  Size += 3;
1358  if (C == '\\') goto Slash;
1359  return C;
1360  }
1361  }
1362 
1363  // If this is neither, return a single character.
1364  ++Size;
1365  return *Ptr;
1366 }
1367 
1368 /// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the
1369 /// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size,
1370 /// and that we have already incremented Ptr by Size bytes.
1371 ///
1372 /// NOTE: When this method is updated, getCharAndSizeSlow (above) should
1373 /// be updated to match.
1374 char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size,
1375  const LangOptions &LangOpts) {
1376  // If we have a slash, look for an escaped newline.
1377  if (Ptr[0] == '\\') {
1378  ++Size;
1379  ++Ptr;
1380 Slash:
1381  // Common case, backslash-char where the char is not whitespace.
1382  if (!isWhitespace(Ptr[0])) return '\\';
1383 
1384  // See if we have optional whitespace characters followed by a newline.
1385  if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) {
1386  // Found backslash<whitespace><newline>. Parse the char after it.
1387  Size += EscapedNewLineSize;
1388  Ptr += EscapedNewLineSize;
1389 
1390  // Use slow version to accumulate a correct size field.
1391  return getCharAndSizeSlowNoWarn(Ptr, Size, LangOpts);
1392  }
1393 
1394  // Otherwise, this is not an escaped newline, just return the slash.
1395  return '\\';
1396  }
1397 
1398  // If this is a trigraph, process it.
1399  if (LangOpts.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') {
1400  // If this is actually a legal trigraph (not something like "??x"), return
1401  // it.
1402  if (char C = GetTrigraphCharForLetter(Ptr[2])) {
1403  Ptr += 3;
1404  Size += 3;
1405  if (C == '\\') goto Slash;
1406  return C;
1407  }
1408  }
1409 
1410  // If this is neither, return a single character.
1411  ++Size;
1412  return *Ptr;
1413 }
1414 
1415 //===----------------------------------------------------------------------===//
1416 // Helper methods for lexing.
1417 //===----------------------------------------------------------------------===//
1418 
1419 /// Routine that indiscriminately sets the offset into the source file.
1420 void Lexer::SetByteOffset(unsigned Offset, bool StartOfLine) {
1421  BufferPtr = BufferStart + Offset;
1422  if (BufferPtr > BufferEnd)
1423  BufferPtr = BufferEnd;
1424  // FIXME: What exactly does the StartOfLine bit mean? There are two
1425  // possible meanings for the "start" of the line: the first token on the
1426  // unexpanded line, or the first token on the expanded line.
1427  IsAtStartOfLine = StartOfLine;
1428  IsAtPhysicalStartOfLine = StartOfLine;
1429 }
1430 
1431 static bool isAllowedIDChar(uint32_t C, const LangOptions &LangOpts) {
1432  if (LangOpts.AsmPreprocessor) {
1433  return false;
1434  } else if (LangOpts.DollarIdents && '$' == C) {
1435  return true;
1436  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1437  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1439  return C11AllowedIDChars.contains(C);
1440  } else if (LangOpts.CPlusPlus) {
1441  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1443  return CXX03AllowedIDChars.contains(C);
1444  } else {
1445  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1447  return C99AllowedIDChars.contains(C);
1448  }
1449 }
1450 
1451 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1452  assert(isAllowedIDChar(C, LangOpts));
1453  if (LangOpts.AsmPreprocessor) {
1454  return false;
1455  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1456  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1458  return !C11DisallowedInitialIDChars.contains(C);
1459  } else if (LangOpts.CPlusPlus) {
1460  return true;
1461  } else {
1462  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1464  return !C99DisallowedInitialIDChars.contains(C);
1465  }
1466 }
1467 
1468 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1469  const char *End) {
1471  L.getSourceLocation(End));
1472 }
1473 
1474 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1475  CharSourceRange Range, bool IsFirst) {
1476  // Check C99 compatibility.
1477  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1478  enum {
1479  CannotAppearInIdentifier = 0,
1480  CannotStartIdentifier
1481  };
1482 
1483  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1485  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1487  if (!C99AllowedIDChars.contains(C)) {
1488  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1489  << Range
1490  << CannotAppearInIdentifier;
1491  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1492  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1493  << Range
1494  << CannotStartIdentifier;
1495  }
1496  }
1497 
1498  // Check C++98 compatibility.
1499  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1500  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1502  if (!CXX03AllowedIDChars.contains(C)) {
1503  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1504  << Range;
1505  }
1506  }
1507 }
1508 
1509 /// After encountering UTF-8 character C and interpreting it as an identifier
1510 /// character, check whether it's a homoglyph for a common non-identifier
1511 /// source character that is unlikely to be an intentional identifier
1512 /// character and warn if so.
1513 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1514  CharSourceRange Range) {
1515  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1516  struct HomoglyphPair {
1517  uint32_t Character;
1518  char LooksLike;
1519  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1520  };
1521  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1522  {U'\u00ad', 0}, // SOFT HYPHEN
1523  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1524  {U'\u037e', ';'}, // GREEK QUESTION MARK
1525  {U'\u200b', 0}, // ZERO WIDTH SPACE
1526  {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1527  {U'\u200d', 0}, // ZERO WIDTH JOINER
1528  {U'\u2060', 0}, // WORD JOINER
1529  {U'\u2061', 0}, // FUNCTION APPLICATION
1530  {U'\u2062', 0}, // INVISIBLE TIMES
1531  {U'\u2063', 0}, // INVISIBLE SEPARATOR
1532  {U'\u2064', 0}, // INVISIBLE PLUS
1533  {U'\u2212', '-'}, // MINUS SIGN
1534  {U'\u2215', '/'}, // DIVISION SLASH
1535  {U'\u2216', '\\'}, // SET MINUS
1536  {U'\u2217', '*'}, // ASTERISK OPERATOR
1537  {U'\u2223', '|'}, // DIVIDES
1538  {U'\u2227', '^'}, // LOGICAL AND
1539  {U'\u2236', ':'}, // RATIO
1540  {U'\u223c', '~'}, // TILDE OPERATOR
1541  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1542  {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1543  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1544  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1545  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1546  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1547  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1548  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1549  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1550  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1551  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1552  {U'\uff0c', ','}, // FULLWIDTH COMMA
1553  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1554  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1555  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1556  {U'\uff1a', ':'}, // FULLWIDTH COLON
1557  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1558  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1559  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1560  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1561  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1562  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1563  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1564  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1565  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1566  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1567  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1568  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1569  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1570  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1571  {0, 0}
1572  };
1573  auto Homoglyph =
1574  std::lower_bound(std::begin(SortedHomoglyphs),
1575  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1576  if (Homoglyph->Character == C) {
1577  llvm::SmallString<5> CharBuf;
1578  {
1579  llvm::raw_svector_ostream CharOS(CharBuf);
1580  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1581  }
1582  if (Homoglyph->LooksLike) {
1583  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1584  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1585  << Range << CharBuf << LooksLikeStr;
1586  } else {
1587  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1588  << Range << CharBuf;
1589  }
1590  }
1591 }
1592 
1593 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1594  Token &Result) {
1595  const char *UCNPtr = CurPtr + Size;
1596  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1597  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1598  return false;
1599 
1600  if (!isLexingRawMode())
1601  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1602  makeCharRange(*this, CurPtr, UCNPtr),
1603  /*IsFirst=*/false);
1604 
1605  Result.setFlag(Token::HasUCN);
1606  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1607  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1608  CurPtr = UCNPtr;
1609  else
1610  while (CurPtr != UCNPtr)
1611  (void)getAndAdvanceChar(CurPtr, Result);
1612  return true;
1613 }
1614 
1615 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1616  const char *UnicodePtr = CurPtr;
1617  llvm::UTF32 CodePoint;
1618  llvm::ConversionResult Result =
1619  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1620  (const llvm::UTF8 *)BufferEnd,
1621  &CodePoint,
1622  llvm::strictConversion);
1623  if (Result != llvm::conversionOK ||
1624  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1625  return false;
1626 
1627  if (!isLexingRawMode()) {
1628  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1629  makeCharRange(*this, CurPtr, UnicodePtr),
1630  /*IsFirst=*/false);
1631  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1632  makeCharRange(*this, CurPtr, UnicodePtr));
1633  }
1634 
1635  CurPtr = UnicodePtr;
1636  return true;
1637 }
1638 
1639 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1640  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1641  unsigned Size;
1642  unsigned char C = *CurPtr++;
1643  while (isIdentifierBody(C))
1644  C = *CurPtr++;
1645 
1646  --CurPtr; // Back up over the skipped character.
1647 
1648  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1649  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1650  //
1651  // TODO: Could merge these checks into an InfoTable flag to make the
1652  // comparison cheaper
1653  if (isASCII(C) && C != '\\' && C != '?' &&
1654  (C != '$' || !LangOpts.DollarIdents)) {
1655 FinishIdentifier:
1656  const char *IdStart = BufferPtr;
1657  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1658  Result.setRawIdentifierData(IdStart);
1659 
1660  // If we are in raw mode, return this identifier raw. There is no need to
1661  // look up identifier information or attempt to macro expand it.
1662  if (LexingRawMode)
1663  return true;
1664 
1665  // Fill in Result.IdentifierInfo and update the token kind,
1666  // looking up the identifier in the identifier table.
1667  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1668  // Note that we have to call PP->LookUpIdentifierInfo() even for code
1669  // completion, it writes IdentifierInfo into Result, and callers rely on it.
1670 
1671  // If the completion point is at the end of an identifier, we want to treat
1672  // the identifier as incomplete even if it resolves to a macro or a keyword.
1673  // This allows e.g. 'class^' to complete to 'classifier'.
1674  if (isCodeCompletionPoint(CurPtr)) {
1675  // Return the code-completion token.
1676  Result.setKind(tok::code_completion);
1677  // Skip the code-completion char and all immediate identifier characters.
1678  // This ensures we get consistent behavior when completing at any point in
1679  // an identifier (i.e. at the start, in the middle, at the end). Note that
1680  // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1681  // simpler.
1682  assert(*CurPtr == 0 && "Completion character must be 0");
1683  ++CurPtr;
1684  // Note that code completion token is not added as a separate character
1685  // when the completion point is at the end of the buffer. Therefore, we need
1686  // to check if the buffer has ended.
1687  if (CurPtr < BufferEnd) {
1688  while (isIdentifierBody(*CurPtr))
1689  ++CurPtr;
1690  }
1691  BufferPtr = CurPtr;
1692  return true;
1693  }
1694 
1695  // Finally, now that we know we have an identifier, pass this off to the
1696  // preprocessor, which may macro expand it or something.
1697  if (II->isHandleIdentifierCase())
1698  return PP->HandleIdentifier(Result);
1699 
1700  return true;
1701  }
1702 
1703  // Otherwise, $,\,? in identifier found. Enter slower path.
1704 
1705  C = getCharAndSize(CurPtr, Size);
1706  while (true) {
1707  if (C == '$') {
1708  // If we hit a $ and they are not supported in identifiers, we are done.
1709  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1710 
1711  // Otherwise, emit a diagnostic and continue.
1712  if (!isLexingRawMode())
1713  Diag(CurPtr, diag::ext_dollar_in_identifier);
1714  CurPtr = ConsumeChar(CurPtr, Size, Result);
1715  C = getCharAndSize(CurPtr, Size);
1716  continue;
1717  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1718  C = getCharAndSize(CurPtr, Size);
1719  continue;
1720  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1721  C = getCharAndSize(CurPtr, Size);
1722  continue;
1723  } else if (!isIdentifierBody(C)) {
1724  goto FinishIdentifier;
1725  }
1726 
1727  // Otherwise, this character is good, consume it.
1728  CurPtr = ConsumeChar(CurPtr, Size, Result);
1729 
1730  C = getCharAndSize(CurPtr, Size);
1731  while (isIdentifierBody(C)) {
1732  CurPtr = ConsumeChar(CurPtr, Size, Result);
1733  C = getCharAndSize(CurPtr, Size);
1734  }
1735  }
1736 }
1737 
1738 /// isHexaLiteral - Return true if Start points to a hex constant.
1739 /// in microsoft mode (where this is supposed to be several different tokens).
1740 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1741  unsigned Size;
1742  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1743  if (C1 != '0')
1744  return false;
1745  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1746  return (C2 == 'x' || C2 == 'X');
1747 }
1748 
1749 /// LexNumericConstant - Lex the remainder of a integer or floating point
1750 /// constant. From[-1] is the first character lexed. Return the end of the
1751 /// constant.
1752 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1753  unsigned Size;
1754  char C = getCharAndSize(CurPtr, Size);
1755  char PrevCh = 0;
1756  while (isPreprocessingNumberBody(C)) {
1757  CurPtr = ConsumeChar(CurPtr, Size, Result);
1758  PrevCh = C;
1759  C = getCharAndSize(CurPtr, Size);
1760  }
1761 
1762  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1763  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1764  // If we are in Microsoft mode, don't continue if the constant is hex.
1765  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1766  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1767  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1768  }
1769 
1770  // If we have a hex FP constant, continue.
1771  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1772  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1773  // not-quite-conforming extension. Only do so if this looks like it's
1774  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1775  bool IsHexFloat = true;
1776  if (!LangOpts.C99) {
1777  if (!isHexaLiteral(BufferPtr, LangOpts))
1778  IsHexFloat = false;
1779  else if (!getLangOpts().CPlusPlus17 &&
1780  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1781  IsHexFloat = false;
1782  }
1783  if (IsHexFloat)
1784  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1785  }
1786 
1787  // If we have a digit separator, continue.
1788  if (C == '\'' && getLangOpts().CPlusPlus14) {
1789  unsigned NextSize;
1790  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1791  if (isIdentifierBody(Next)) {
1792  if (!isLexingRawMode())
1793  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1794  CurPtr = ConsumeChar(CurPtr, Size, Result);
1795  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1796  return LexNumericConstant(Result, CurPtr);
1797  }
1798  }
1799 
1800  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1801  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1802  return LexNumericConstant(Result, CurPtr);
1803  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1804  return LexNumericConstant(Result, CurPtr);
1805 
1806  // Update the location of token as well as BufferPtr.
1807  const char *TokStart = BufferPtr;
1808  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1809  Result.setLiteralData(TokStart);
1810  return true;
1811 }
1812 
1813 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1814 /// in C++11, or warn on a ud-suffix in C++98.
1815 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1816  bool IsStringLiteral) {
1817  assert(getLangOpts().CPlusPlus);
1818 
1819  // Maximally munch an identifier.
1820  unsigned Size;
1821  char C = getCharAndSize(CurPtr, Size);
1822  bool Consumed = false;
1823 
1824  if (!isIdentifierHead(C)) {
1825  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1826  Consumed = true;
1827  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1828  Consumed = true;
1829  else
1830  return CurPtr;
1831  }
1832 
1833  if (!getLangOpts().CPlusPlus11) {
1834  if (!isLexingRawMode())
1835  Diag(CurPtr,
1836  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1837  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1839  return CurPtr;
1840  }
1841 
1842  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1843  // that does not start with an underscore is ill-formed. As a conforming
1844  // extension, we treat all such suffixes as if they had whitespace before
1845  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1846  // likely to be a ud-suffix than a macro, however, and accept that.
1847  if (!Consumed) {
1848  bool IsUDSuffix = false;
1849  if (C == '_')
1850  IsUDSuffix = true;
1851  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1852  // In C++1y, we need to look ahead a few characters to see if this is a
1853  // valid suffix for a string literal or a numeric literal (this could be
1854  // the 'operator""if' defining a numeric literal operator).
1855  const unsigned MaxStandardSuffixLength = 3;
1856  char Buffer[MaxStandardSuffixLength] = { C };
1857  unsigned Consumed = Size;
1858  unsigned Chars = 1;
1859  while (true) {
1860  unsigned NextSize;
1861  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1862  getLangOpts());
1863  if (!isIdentifierBody(Next)) {
1864  // End of suffix. Check whether this is on the whitelist.
1865  const StringRef CompleteSuffix(Buffer, Chars);
1867  CompleteSuffix);
1868  break;
1869  }
1870 
1871  if (Chars == MaxStandardSuffixLength)
1872  // Too long: can't be a standard suffix.
1873  break;
1874 
1875  Buffer[Chars++] = Next;
1876  Consumed += NextSize;
1877  }
1878  }
1879 
1880  if (!IsUDSuffix) {
1881  if (!isLexingRawMode())
1882  Diag(CurPtr, getLangOpts().MSVCCompat
1883  ? diag::ext_ms_reserved_user_defined_literal
1884  : diag::ext_reserved_user_defined_literal)
1886  return CurPtr;
1887  }
1888 
1889  CurPtr = ConsumeChar(CurPtr, Size, Result);
1890  }
1891 
1892  Result.setFlag(Token::HasUDSuffix);
1893  while (true) {
1894  C = getCharAndSize(CurPtr, Size);
1895  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1896  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1897  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1898  else break;
1899  }
1900 
1901  return CurPtr;
1902 }
1903 
1904 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1905 /// either " or L" or u8" or u" or U".
1906 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1907  tok::TokenKind Kind) {
1908  const char *AfterQuote = CurPtr;
1909  // Does this string contain the \0 character?
1910  const char *NulCharacter = nullptr;
1911 
1912  if (!isLexingRawMode() &&
1913  (Kind == tok::utf8_string_literal ||
1914  Kind == tok::utf16_string_literal ||
1915  Kind == tok::utf32_string_literal))
1916  Diag(BufferPtr, getLangOpts().CPlusPlus
1917  ? diag::warn_cxx98_compat_unicode_literal
1918  : diag::warn_c99_compat_unicode_literal);
1919 
1920  char C = getAndAdvanceChar(CurPtr, Result);
1921  while (C != '"') {
1922  // Skip escaped characters. Escaped newlines will already be processed by
1923  // getAndAdvanceChar.
1924  if (C == '\\')
1925  C = getAndAdvanceChar(CurPtr, Result);
1926 
1927  if (C == '\n' || C == '\r' || // Newline.
1928  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1929  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1930  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1931  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1932  return true;
1933  }
1934 
1935  if (C == 0) {
1936  if (isCodeCompletionPoint(CurPtr-1)) {
1937  if (ParsingFilename)
1938  codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
1939  else
1941  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
1942  cutOffLexing();
1943  return true;
1944  }
1945 
1946  NulCharacter = CurPtr-1;
1947  }
1948  C = getAndAdvanceChar(CurPtr, Result);
1949  }
1950 
1951  // If we are in C++11, lex the optional ud-suffix.
1952  if (getLangOpts().CPlusPlus)
1953  CurPtr = LexUDSuffix(Result, CurPtr, true);
1954 
1955  // If a nul character existed in the string, warn about it.
1956  if (NulCharacter && !isLexingRawMode())
1957  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1958 
1959  // Update the location of the token as well as the BufferPtr instance var.
1960  const char *TokStart = BufferPtr;
1961  FormTokenWithChars(Result, CurPtr, Kind);
1962  Result.setLiteralData(TokStart);
1963  return true;
1964 }
1965 
1966 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1967 /// having lexed R", LR", u8R", uR", or UR".
1968 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1969  tok::TokenKind Kind) {
1970  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1971  // Between the initial and final double quote characters of the raw string,
1972  // any transformations performed in phases 1 and 2 (trigraphs,
1973  // universal-character-names, and line splicing) are reverted.
1974 
1975  if (!isLexingRawMode())
1976  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1977 
1978  unsigned PrefixLen = 0;
1979 
1980  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1981  ++PrefixLen;
1982 
1983  // If the last character was not a '(', then we didn't lex a valid delimiter.
1984  if (CurPtr[PrefixLen] != '(') {
1985  if (!isLexingRawMode()) {
1986  const char *PrefixEnd = &CurPtr[PrefixLen];
1987  if (PrefixLen == 16) {
1988  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1989  } else {
1990  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1991  << StringRef(PrefixEnd, 1);
1992  }
1993  }
1994 
1995  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1996  // it's possible the '"' was intended to be part of the raw string, but
1997  // there's not much we can do about that.
1998  while (true) {
1999  char C = *CurPtr++;
2000 
2001  if (C == '"')
2002  break;
2003  if (C == 0 && CurPtr-1 == BufferEnd) {
2004  --CurPtr;
2005  break;
2006  }
2007  }
2008 
2009  FormTokenWithChars(Result, CurPtr, tok::unknown);
2010  return true;
2011  }
2012 
2013  // Save prefix and move CurPtr past it
2014  const char *Prefix = CurPtr;
2015  CurPtr += PrefixLen + 1; // skip over prefix and '('
2016 
2017  while (true) {
2018  char C = *CurPtr++;
2019 
2020  if (C == ')') {
2021  // Check for prefix match and closing quote.
2022  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2023  CurPtr += PrefixLen + 1; // skip over prefix and '"'
2024  break;
2025  }
2026  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2027  if (!isLexingRawMode())
2028  Diag(BufferPtr, diag::err_unterminated_raw_string)
2029  << StringRef(Prefix, PrefixLen);
2030  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2031  return true;
2032  }
2033  }
2034 
2035  // If we are in C++11, lex the optional ud-suffix.
2036  if (getLangOpts().CPlusPlus)
2037  CurPtr = LexUDSuffix(Result, CurPtr, true);
2038 
2039  // Update the location of token as well as BufferPtr.
2040  const char *TokStart = BufferPtr;
2041  FormTokenWithChars(Result, CurPtr, Kind);
2042  Result.setLiteralData(TokStart);
2043  return true;
2044 }
2045 
2046 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2047 /// after having lexed the '<' character. This is used for #include filenames.
2048 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2049  // Does this string contain the \0 character?
2050  const char *NulCharacter = nullptr;
2051  const char *AfterLessPos = CurPtr;
2052  char C = getAndAdvanceChar(CurPtr, Result);
2053  while (C != '>') {
2054  // Skip escaped characters. Escaped newlines will already be processed by
2055  // getAndAdvanceChar.
2056  if (C == '\\')
2057  C = getAndAdvanceChar(CurPtr, Result);
2058 
2059  if (C == '\n' || C == '\r' || // Newline.
2060  (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2061  // If the filename is unterminated, then it must just be a lone <
2062  // character. Return this as such.
2063  FormTokenWithChars(Result, AfterLessPos, tok::less);
2064  return true;
2065  }
2066 
2067  if (C == 0) {
2068  if (isCodeCompletionPoint(CurPtr - 1)) {
2069  codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2070  cutOffLexing();
2071  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2072  return true;
2073  }
2074  NulCharacter = CurPtr-1;
2075  }
2076  C = getAndAdvanceChar(CurPtr, Result);
2077  }
2078 
2079  // If a nul character existed in the string, warn about it.
2080  if (NulCharacter && !isLexingRawMode())
2081  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2082 
2083  // Update the location of token as well as BufferPtr.
2084  const char *TokStart = BufferPtr;
2085  FormTokenWithChars(Result, CurPtr, tok::header_name);
2086  Result.setLiteralData(TokStart);
2087  return true;
2088 }
2089 
2090 void Lexer::codeCompleteIncludedFile(const char *PathStart,
2091  const char *CompletionPoint,
2092  bool IsAngled) {
2093  // Completion only applies to the filename, after the last slash.
2094  StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2095  auto Slash = PartialPath.find_last_of(LangOpts.MSVCCompat ? "/\\" : "/");
2096  StringRef Dir =
2097  (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2098  const char *StartOfFilename =
2099  (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2100  // Code completion filter range is the filename only, up to completion point.
2102  StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2103  // We should replace the characters up to the closing quote, if any.
2104  while (CompletionPoint < BufferEnd) {
2105  char Next = *(CompletionPoint + 1);
2106  if (Next == 0 || Next == '\r' || Next == '\n')
2107  break;
2108  ++CompletionPoint;
2109  if (Next == (IsAngled ? '>' : '"'))
2110  break;
2111  }
2113  FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2114  FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2115  PP->CodeCompleteIncludedFile(Dir, IsAngled);
2116 }
2117 
2118 /// LexCharConstant - Lex the remainder of a character constant, after having
2119 /// lexed either ' or L' or u8' or u' or U'.
2120 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2121  tok::TokenKind Kind) {
2122  // Does this character contain the \0 character?
2123  const char *NulCharacter = nullptr;
2124 
2125  if (!isLexingRawMode()) {
2126  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2127  Diag(BufferPtr, getLangOpts().CPlusPlus
2128  ? diag::warn_cxx98_compat_unicode_literal
2129  : diag::warn_c99_compat_unicode_literal);
2130  else if (Kind == tok::utf8_char_constant)
2131  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2132  }
2133 
2134  char C = getAndAdvanceChar(CurPtr, Result);
2135  if (C == '\'') {
2136  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2137  Diag(BufferPtr, diag::ext_empty_character);
2138  FormTokenWithChars(Result, CurPtr, tok::unknown);
2139  return true;
2140  }
2141 
2142  while (C != '\'') {
2143  // Skip escaped characters.
2144  if (C == '\\')
2145  C = getAndAdvanceChar(CurPtr, Result);
2146 
2147  if (C == '\n' || C == '\r' || // Newline.
2148  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2149  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2150  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2151  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2152  return true;
2153  }
2154 
2155  if (C == 0) {
2156  if (isCodeCompletionPoint(CurPtr-1)) {
2158  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2159  cutOffLexing();
2160  return true;
2161  }
2162 
2163  NulCharacter = CurPtr-1;
2164  }
2165  C = getAndAdvanceChar(CurPtr, Result);
2166  }
2167 
2168  // If we are in C++11, lex the optional ud-suffix.
2169  if (getLangOpts().CPlusPlus)
2170  CurPtr = LexUDSuffix(Result, CurPtr, false);
2171 
2172  // If a nul character existed in the character, warn about it.
2173  if (NulCharacter && !isLexingRawMode())
2174  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2175 
2176  // Update the location of token as well as BufferPtr.
2177  const char *TokStart = BufferPtr;
2178  FormTokenWithChars(Result, CurPtr, Kind);
2179  Result.setLiteralData(TokStart);
2180  return true;
2181 }
2182 
2183 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2184 /// Update BufferPtr to point to the next non-whitespace character and return.
2185 ///
2186 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2187 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2188  bool &TokAtPhysicalStartOfLine) {
2189  // Whitespace - Skip it, then return the token after the whitespace.
2190  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2191 
2192  unsigned char Char = *CurPtr;
2193 
2194  // Skip consecutive spaces efficiently.
2195  while (true) {
2196  // Skip horizontal whitespace very aggressively.
2197  while (isHorizontalWhitespace(Char))
2198  Char = *++CurPtr;
2199 
2200  // Otherwise if we have something other than whitespace, we're done.
2201  if (!isVerticalWhitespace(Char))
2202  break;
2203 
2205  // End of preprocessor directive line, let LexTokenInternal handle this.
2206  BufferPtr = CurPtr;
2207  return false;
2208  }
2209 
2210  // OK, but handle newline.
2211  SawNewline = true;
2212  Char = *++CurPtr;
2213  }
2214 
2215  // If the client wants us to return whitespace, return it now.
2216  if (isKeepWhitespaceMode()) {
2217  FormTokenWithChars(Result, CurPtr, tok::unknown);
2218  if (SawNewline) {
2219  IsAtStartOfLine = true;
2220  IsAtPhysicalStartOfLine = true;
2221  }
2222  // FIXME: The next token will not have LeadingSpace set.
2223  return true;
2224  }
2225 
2226  // If this isn't immediately after a newline, there is leading space.
2227  char PrevChar = CurPtr[-1];
2228  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2229 
2230  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2231  if (SawNewline) {
2232  Result.setFlag(Token::StartOfLine);
2233  TokAtPhysicalStartOfLine = true;
2234  }
2235 
2236  BufferPtr = CurPtr;
2237  return false;
2238 }
2239 
2240 /// We have just read the // characters from input. Skip until we find the
2241 /// newline character that terminates the comment. Then update BufferPtr and
2242 /// return.
2243 ///
2244 /// If we're in KeepCommentMode or any CommentHandler has inserted
2245 /// some tokens, this will store the first token and return true.
2246 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2247  bool &TokAtPhysicalStartOfLine) {
2248  // If Line comments aren't explicitly enabled for this language, emit an
2249  // extension warning.
2250  if (!LangOpts.LineComment && !isLexingRawMode()) {
2251  Diag(BufferPtr, diag::ext_line_comment);
2252 
2253  // Mark them enabled so we only emit one warning for this translation
2254  // unit.
2255  LangOpts.LineComment = true;
2256  }
2257 
2258  // Scan over the body of the comment. The common case, when scanning, is that
2259  // the comment contains normal ascii characters with nothing interesting in
2260  // them. As such, optimize for this case with the inner loop.
2261  //
2262  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2263  // character that ends the line comment.
2264  char C;
2265  while (true) {
2266  C = *CurPtr;
2267  // Skip over characters in the fast loop.
2268  while (C != 0 && // Potentially EOF.
2269  C != '\n' && C != '\r') // Newline or DOS-style newline.
2270  C = *++CurPtr;
2271 
2272  const char *NextLine = CurPtr;
2273  if (C != 0) {
2274  // We found a newline, see if it's escaped.
2275  const char *EscapePtr = CurPtr-1;
2276  bool HasSpace = false;
2277  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2278  --EscapePtr;
2279  HasSpace = true;
2280  }
2281 
2282  if (*EscapePtr == '\\')
2283  // Escaped newline.
2284  CurPtr = EscapePtr;
2285  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2286  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2287  // Trigraph-escaped newline.
2288  CurPtr = EscapePtr-2;
2289  else
2290  break; // This is a newline, we're done.
2291 
2292  // If there was space between the backslash and newline, warn about it.
2293  if (HasSpace && !isLexingRawMode())
2294  Diag(EscapePtr, diag::backslash_newline_space);
2295  }
2296 
2297  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2298  // properly decode the character. Read it in raw mode to avoid emitting
2299  // diagnostics about things like trigraphs. If we see an escaped newline,
2300  // we'll handle it below.
2301  const char *OldPtr = CurPtr;
2302  bool OldRawMode = isLexingRawMode();
2303  LexingRawMode = true;
2304  C = getAndAdvanceChar(CurPtr, Result);
2305  LexingRawMode = OldRawMode;
2306 
2307  // If we only read only one character, then no special handling is needed.
2308  // We're done and can skip forward to the newline.
2309  if (C != 0 && CurPtr == OldPtr+1) {
2310  CurPtr = NextLine;
2311  break;
2312  }
2313 
2314  // If we read multiple characters, and one of those characters was a \r or
2315  // \n, then we had an escaped newline within the comment. Emit diagnostic
2316  // unless the next line is also a // comment.
2317  if (CurPtr != OldPtr + 1 && C != '/' &&
2318  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2319  for (; OldPtr != CurPtr; ++OldPtr)
2320  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2321  // Okay, we found a // comment that ends in a newline, if the next
2322  // line is also a // comment, but has spaces, don't emit a diagnostic.
2323  if (isWhitespace(C)) {
2324  const char *ForwardPtr = CurPtr;
2325  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2326  ++ForwardPtr;
2327  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2328  break;
2329  }
2330 
2331  if (!isLexingRawMode())
2332  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2333  break;
2334  }
2335  }
2336 
2337  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2338  --CurPtr;
2339  break;
2340  }
2341 
2342  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2344  cutOffLexing();
2345  return false;
2346  }
2347  }
2348 
2349  // Found but did not consume the newline. Notify comment handlers about the
2350  // comment unless we're in a #if 0 block.
2351  if (PP && !isLexingRawMode() &&
2352  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2353  getSourceLocation(CurPtr)))) {
2354  BufferPtr = CurPtr;
2355  return true; // A token has to be returned.
2356  }
2357 
2358  // If we are returning comments as tokens, return this comment as a token.
2359  if (inKeepCommentMode())
2360  return SaveLineComment(Result, CurPtr);
2361 
2362  // If we are inside a preprocessor directive and we see the end of line,
2363  // return immediately, so that the lexer can return this as an EOD token.
2364  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2365  BufferPtr = CurPtr;
2366  return false;
2367  }
2368 
2369  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2370  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2371  // contribute to another token), it isn't needed for correctness. Note that
2372  // this is ok even in KeepWhitespaceMode, because we would have returned the
2373  /// comment above in that mode.
2374  ++CurPtr;
2375 
2376  // The next returned token is at the start of the line.
2377  Result.setFlag(Token::StartOfLine);
2378  TokAtPhysicalStartOfLine = true;
2379  // No leading whitespace seen so far.
2381  BufferPtr = CurPtr;
2382  return false;
2383 }
2384 
2385 /// If in save-comment mode, package up this Line comment in an appropriate
2386 /// way and return it.
2387 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2388  // If we're not in a preprocessor directive, just return the // comment
2389  // directly.
2390  FormTokenWithChars(Result, CurPtr, tok::comment);
2391 
2393  return true;
2394 
2395  // If this Line-style comment is in a macro definition, transmogrify it into
2396  // a C-style block comment.
2397  bool Invalid = false;
2398  std::string Spelling = PP->getSpelling(Result, &Invalid);
2399  if (Invalid)
2400  return true;
2401 
2402  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2403  Spelling[1] = '*'; // Change prefix to "/*".
2404  Spelling += "*/"; // add suffix.
2405 
2406  Result.setKind(tok::comment);
2407  PP->CreateString(Spelling, Result,
2408  Result.getLocation(), Result.getLocation());
2409  return true;
2410 }
2411 
2412 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2413 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2414 /// a diagnostic if so. We know that the newline is inside of a block comment.
2415 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2416  Lexer *L) {
2417  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2418 
2419  // Back up off the newline.
2420  --CurPtr;
2421 
2422  // If this is a two-character newline sequence, skip the other character.
2423  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2424  // \n\n or \r\r -> not escaped newline.
2425  if (CurPtr[0] == CurPtr[1])
2426  return false;
2427  // \n\r or \r\n -> skip the newline.
2428  --CurPtr;
2429  }
2430 
2431  // If we have horizontal whitespace, skip over it. We allow whitespace
2432  // between the slash and newline.
2433  bool HasSpace = false;
2434  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2435  --CurPtr;
2436  HasSpace = true;
2437  }
2438 
2439  // If we have a slash, we know this is an escaped newline.
2440  if (*CurPtr == '\\') {
2441  if (CurPtr[-1] != '*') return false;
2442  } else {
2443  // It isn't a slash, is it the ?? / trigraph?
2444  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2445  CurPtr[-3] != '*')
2446  return false;
2447 
2448  // This is the trigraph ending the comment. Emit a stern warning!
2449  CurPtr -= 2;
2450 
2451  // If no trigraphs are enabled, warn that we ignored this trigraph and
2452  // ignore this * character.
2453  if (!L->getLangOpts().Trigraphs) {
2454  if (!L->isLexingRawMode())
2455  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2456  return false;
2457  }
2458  if (!L->isLexingRawMode())
2459  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2460  }
2461 
2462  // Warn about having an escaped newline between the */ characters.
2463  if (!L->isLexingRawMode())
2464  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2465 
2466  // If there was space between the backslash and newline, warn about it.
2467  if (HasSpace && !L->isLexingRawMode())
2468  L->Diag(CurPtr, diag::backslash_newline_space);
2469 
2470  return true;
2471 }
2472 
2473 #ifdef __SSE2__
2474 #include <emmintrin.h>
2475 #elif __ALTIVEC__
2476 #include <altivec.h>
2477 #undef bool
2478 #endif
2479 
2480 /// We have just read from input the / and * characters that started a comment.
2481 /// Read until we find the * and / characters that terminate the comment.
2482 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2483 /// comments, because they cannot cause the comment to end. The only thing
2484 /// that can happen is the comment could end with an escaped newline between
2485 /// the terminating * and /.
2486 ///
2487 /// If we're in KeepCommentMode or any CommentHandler has inserted
2488 /// some tokens, this will store the first token and return true.
2489 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2490  bool &TokAtPhysicalStartOfLine) {
2491  // Scan one character past where we should, looking for a '/' character. Once
2492  // we find it, check to see if it was preceded by a *. This common
2493  // optimization helps people who like to put a lot of * characters in their
2494  // comments.
2495 
2496  // The first character we get with newlines and trigraphs skipped to handle
2497  // the degenerate /*/ case below correctly if the * has an escaped newline
2498  // after it.
2499  unsigned CharSize;
2500  unsigned char C = getCharAndSize(CurPtr, CharSize);
2501  CurPtr += CharSize;
2502  if (C == 0 && CurPtr == BufferEnd+1) {
2503  if (!isLexingRawMode())
2504  Diag(BufferPtr, diag::err_unterminated_block_comment);
2505  --CurPtr;
2506 
2507  // KeepWhitespaceMode should return this broken comment as a token. Since
2508  // it isn't a well formed comment, just return it as an 'unknown' token.
2509  if (isKeepWhitespaceMode()) {
2510  FormTokenWithChars(Result, CurPtr, tok::unknown);
2511  return true;
2512  }
2513 
2514  BufferPtr = CurPtr;
2515  return false;
2516  }
2517 
2518  // Check to see if the first character after the '/*' is another /. If so,
2519  // then this slash does not end the block comment, it is part of it.
2520  if (C == '/')
2521  C = *CurPtr++;
2522 
2523  while (true) {
2524  // Skip over all non-interesting characters until we find end of buffer or a
2525  // (probably ending) '/' character.
2526  if (CurPtr + 24 < BufferEnd &&
2527  // If there is a code-completion point avoid the fast scan because it
2528  // doesn't check for '\0'.
2529  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2530  // While not aligned to a 16-byte boundary.
2531  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2532  C = *CurPtr++;
2533 
2534  if (C == '/') goto FoundSlash;
2535 
2536 #ifdef __SSE2__
2537  __m128i Slashes = _mm_set1_epi8('/');
2538  while (CurPtr+16 <= BufferEnd) {
2539  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2540  Slashes));
2541  if (cmp != 0) {
2542  // Adjust the pointer to point directly after the first slash. It's
2543  // not necessary to set C here, it will be overwritten at the end of
2544  // the outer loop.
2545  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2546  goto FoundSlash;
2547  }
2548  CurPtr += 16;
2549  }
2550 #elif __ALTIVEC__
2551  __vector unsigned char Slashes = {
2552  '/', '/', '/', '/', '/', '/', '/', '/',
2553  '/', '/', '/', '/', '/', '/', '/', '/'
2554  };
2555  while (CurPtr + 16 <= BufferEnd &&
2556  !vec_any_eq(*(const __vector unsigned char *)CurPtr, Slashes))
2557  CurPtr += 16;
2558 #else
2559  // Scan for '/' quickly. Many block comments are very large.
2560  while (CurPtr[0] != '/' &&
2561  CurPtr[1] != '/' &&
2562  CurPtr[2] != '/' &&
2563  CurPtr[3] != '/' &&
2564  CurPtr+4 < BufferEnd) {
2565  CurPtr += 4;
2566  }
2567 #endif
2568 
2569  // It has to be one of the bytes scanned, increment to it and read one.
2570  C = *CurPtr++;
2571  }
2572 
2573  // Loop to scan the remainder.
2574  while (C != '/' && C != '\0')
2575  C = *CurPtr++;
2576 
2577  if (C == '/') {
2578  FoundSlash:
2579  if (CurPtr[-2] == '*') // We found the final */. We're done!
2580  break;
2581 
2582  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2583  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2584  // We found the final */, though it had an escaped newline between the
2585  // * and /. We're done!
2586  break;
2587  }
2588  }
2589  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2590  // If this is a /* inside of the comment, emit a warning. Don't do this
2591  // if this is a /*/, which will end the comment. This misses cases with
2592  // embedded escaped newlines, but oh well.
2593  if (!isLexingRawMode())
2594  Diag(CurPtr-1, diag::warn_nested_block_comment);
2595  }
2596  } else if (C == 0 && CurPtr == BufferEnd+1) {
2597  if (!isLexingRawMode())
2598  Diag(BufferPtr, diag::err_unterminated_block_comment);
2599  // Note: the user probably forgot a */. We could continue immediately
2600  // after the /*, but this would involve lexing a lot of what really is the
2601  // comment, which surely would confuse the parser.
2602  --CurPtr;
2603 
2604  // KeepWhitespaceMode should return this broken comment as a token. Since
2605  // it isn't a well formed comment, just return it as an 'unknown' token.
2606  if (isKeepWhitespaceMode()) {
2607  FormTokenWithChars(Result, CurPtr, tok::unknown);
2608  return true;
2609  }
2610 
2611  BufferPtr = CurPtr;
2612  return false;
2613  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2615  cutOffLexing();
2616  return false;
2617  }
2618 
2619  C = *CurPtr++;
2620  }
2621 
2622  // Notify comment handlers about the comment unless we're in a #if 0 block.
2623  if (PP && !isLexingRawMode() &&
2624  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2625  getSourceLocation(CurPtr)))) {
2626  BufferPtr = CurPtr;
2627  return true; // A token has to be returned.
2628  }
2629 
2630  // If we are returning comments as tokens, return this comment as a token.
2631  if (inKeepCommentMode()) {
2632  FormTokenWithChars(Result, CurPtr, tok::comment);
2633  return true;
2634  }
2635 
2636  // It is common for the tokens immediately after a /**/ comment to be
2637  // whitespace. Instead of going through the big switch, handle it
2638  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2639  // have already returned above with the comment as a token.
2640  if (isHorizontalWhitespace(*CurPtr)) {
2641  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2642  return false;
2643  }
2644 
2645  // Otherwise, just return so that the next character will be lexed as a token.
2646  BufferPtr = CurPtr;
2647  Result.setFlag(Token::LeadingSpace);
2648  return false;
2649 }
2650 
2651 //===----------------------------------------------------------------------===//
2652 // Primary Lexing Entry Points
2653 //===----------------------------------------------------------------------===//
2654 
2655 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2656 /// uninterpreted string. This switches the lexer out of directive mode.
2658  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2659  "Must be in a preprocessing directive!");
2660  Token Tmp;
2661  Tmp.startToken();
2662 
2663  // CurPtr - Cache BufferPtr in an automatic variable.
2664  const char *CurPtr = BufferPtr;
2665  while (true) {
2666  char Char = getAndAdvanceChar(CurPtr, Tmp);
2667  switch (Char) {
2668  default:
2669  if (Result)
2670  Result->push_back(Char);
2671  break;
2672  case 0: // Null.
2673  // Found end of file?
2674  if (CurPtr-1 != BufferEnd) {
2675  if (isCodeCompletionPoint(CurPtr-1)) {
2677  cutOffLexing();
2678  return;
2679  }
2680 
2681  // Nope, normal character, continue.
2682  if (Result)
2683  Result->push_back(Char);
2684  break;
2685  }
2686  // FALL THROUGH.
2687  LLVM_FALLTHROUGH;
2688  case '\r':
2689  case '\n':
2690  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2691  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2692  BufferPtr = CurPtr-1;
2693 
2694  // Next, lex the character, which should handle the EOD transition.
2695  Lex(Tmp);
2696  if (Tmp.is(tok::code_completion)) {
2697  if (PP)
2699  Lex(Tmp);
2700  }
2701  assert(Tmp.is(tok::eod) && "Unexpected token!");
2702 
2703  // Finally, we're done;
2704  return;
2705  }
2706  }
2707 }
2708 
2709 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2710 /// condition, reporting diagnostics and handling other edge cases as required.
2711 /// This returns true if Result contains a token, false if PP.Lex should be
2712 /// called again.
2713 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2714  // If we hit the end of the file while parsing a preprocessor directive,
2715  // end the preprocessor directive first. The next token returned will
2716  // then be the end of file.
2718  // Done parsing the "line".
2720  // Update the location of token as well as BufferPtr.
2721  FormTokenWithChars(Result, CurPtr, tok::eod);
2722 
2723  // Restore comment saving mode, in case it was disabled for directive.
2724  if (PP)
2726  return true; // Have a token.
2727  }
2728 
2729  // If we are in raw mode, return this event as an EOF token. Let the caller
2730  // that put us in raw mode handle the event.
2731  if (isLexingRawMode()) {
2732  Result.startToken();
2733  BufferPtr = BufferEnd;
2734  FormTokenWithChars(Result, BufferEnd, tok::eof);
2735  return true;
2736  }
2737 
2738  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2740  ConditionalStack.clear();
2741  }
2742 
2743  // Issue diagnostics for unterminated #if and missing newline.
2744 
2745  // If we are in a #if directive, emit an error.
2746  while (!ConditionalStack.empty()) {
2747  if (PP->getCodeCompletionFileLoc() != FileLoc)
2748  PP->Diag(ConditionalStack.back().IfLoc,
2749  diag::err_pp_unterminated_conditional);
2750  ConditionalStack.pop_back();
2751  }
2752 
2753  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2754  // a pedwarn.
2755  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2756  DiagnosticsEngine &Diags = PP->getDiagnostics();
2757  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2758  unsigned DiagID;
2759 
2760  if (LangOpts.CPlusPlus11) {
2761  // C++11 [lex.phases] 2.2 p2
2762  // Prefer the C++98 pedantic compatibility warning over the generic,
2763  // non-extension, user-requested "missing newline at EOF" warning.
2764  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2765  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2766  } else {
2767  DiagID = diag::warn_no_newline_eof;
2768  }
2769  } else {
2770  DiagID = diag::ext_no_newline_eof;
2771  }
2772 
2773  Diag(BufferEnd, DiagID)
2774  << FixItHint::CreateInsertion(EndLoc, "\n");
2775  }
2776 
2777  BufferPtr = CurPtr;
2778 
2779  // Finally, let the preprocessor handle this.
2780  return PP->HandleEndOfFile(Result, isPragmaLexer());
2781 }
2782 
2783 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2784 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2785 /// else and 2 if there are no more tokens in the buffer controlled by the
2786 /// lexer.
2787 unsigned Lexer::isNextPPTokenLParen() {
2788  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2789 
2790  // Switch to 'skipping' mode. This will ensure that we can lex a token
2791  // without emitting diagnostics, disables macro expansion, and will cause EOF
2792  // to return an EOF token instead of popping the include stack.
2793  LexingRawMode = true;
2794 
2795  // Save state that can be changed while lexing so that we can restore it.
2796  const char *TmpBufferPtr = BufferPtr;
2797  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2798  bool atStartOfLine = IsAtStartOfLine;
2799  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2800  bool leadingSpace = HasLeadingSpace;
2801 
2802  Token Tok;
2803  Lex(Tok);
2804 
2805  // Restore state that may have changed.
2806  BufferPtr = TmpBufferPtr;
2807  ParsingPreprocessorDirective = inPPDirectiveMode;
2808  HasLeadingSpace = leadingSpace;
2809  IsAtStartOfLine = atStartOfLine;
2810  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2811 
2812  // Restore the lexer back to non-skipping mode.
2813  LexingRawMode = false;
2814 
2815  if (Tok.is(tok::eof))
2816  return 2;
2817  return Tok.is(tok::l_paren);
2818 }
2819 
2820 /// Find the end of a version control conflict marker.
2821 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2822  ConflictMarkerKind CMK) {
2823  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2824  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2825  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2826  size_t Pos = RestOfBuffer.find(Terminator);
2827  while (Pos != StringRef::npos) {
2828  // Must occur at start of line.
2829  if (Pos == 0 ||
2830  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2831  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2832  Pos = RestOfBuffer.find(Terminator);
2833  continue;
2834  }
2835  return RestOfBuffer.data()+Pos;
2836  }
2837  return nullptr;
2838 }
2839 
2840 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2841 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2842 /// and recover nicely. This returns true if it is a conflict marker and false
2843 /// if not.
2844 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2845  // Only a conflict marker if it starts at the beginning of a line.
2846  if (CurPtr != BufferStart &&
2847  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2848  return false;
2849 
2850  // Check to see if we have <<<<<<< or >>>>.
2851  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2852  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2853  return false;
2854 
2855  // If we have a situation where we don't care about conflict markers, ignore
2856  // it.
2857  if (CurrentConflictMarkerState || isLexingRawMode())
2858  return false;
2859 
2860  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2861 
2862  // Check to see if there is an ending marker somewhere in the buffer at the
2863  // start of a line to terminate this conflict marker.
2864  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2865  // We found a match. We are really in a conflict marker.
2866  // Diagnose this, and ignore to the end of line.
2867  Diag(CurPtr, diag::err_conflict_marker);
2868  CurrentConflictMarkerState = Kind;
2869 
2870  // Skip ahead to the end of line. We know this exists because the
2871  // end-of-conflict marker starts with \r or \n.
2872  while (*CurPtr != '\r' && *CurPtr != '\n') {
2873  assert(CurPtr != BufferEnd && "Didn't find end of line");
2874  ++CurPtr;
2875  }
2876  BufferPtr = CurPtr;
2877  return true;
2878  }
2879 
2880  // No end of conflict marker found.
2881  return false;
2882 }
2883 
2884 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2885 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2886 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2887 /// the line. This returns true if it is a conflict marker and false if not.
2888 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2889  // Only a conflict marker if it starts at the beginning of a line.
2890  if (CurPtr != BufferStart &&
2891  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2892  return false;
2893 
2894  // If we have a situation where we don't care about conflict markers, ignore
2895  // it.
2896  if (!CurrentConflictMarkerState || isLexingRawMode())
2897  return false;
2898 
2899  // Check to see if we have the marker (4 characters in a row).
2900  for (unsigned i = 1; i != 4; ++i)
2901  if (CurPtr[i] != CurPtr[0])
2902  return false;
2903 
2904  // If we do have it, search for the end of the conflict marker. This could
2905  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2906  // be the end of conflict marker.
2907  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2908  CurrentConflictMarkerState)) {
2909  CurPtr = End;
2910 
2911  // Skip ahead to the end of line.
2912  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2913  ++CurPtr;
2914 
2915  BufferPtr = CurPtr;
2916 
2917  // No longer in the conflict marker.
2918  CurrentConflictMarkerState = CMK_None;
2919  return true;
2920  }
2921 
2922  return false;
2923 }
2924 
2925 static const char *findPlaceholderEnd(const char *CurPtr,
2926  const char *BufferEnd) {
2927  if (CurPtr == BufferEnd)
2928  return nullptr;
2929  BufferEnd -= 1; // Scan until the second last character.
2930  for (; CurPtr != BufferEnd; ++CurPtr) {
2931  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2932  return CurPtr + 2;
2933  }
2934  return nullptr;
2935 }
2936 
2937 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2938  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2940  return false;
2941  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2942  if (!End)
2943  return false;
2944  const char *Start = CurPtr - 1;
2945  if (!LangOpts.AllowEditorPlaceholders)
2946  Diag(Start, diag::err_placeholder_in_source);
2947  Result.startToken();
2948  FormTokenWithChars(Result, End, tok::raw_identifier);
2949  Result.setRawIdentifierData(Start);
2950  PP->LookUpIdentifierInfo(Result);
2952  BufferPtr = End;
2953  return true;
2954 }
2955 
2956 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2957  if (PP && PP->isCodeCompletionEnabled()) {
2958  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2959  return Loc == PP->getCodeCompletionLoc();
2960  }
2961 
2962  return false;
2963 }
2964 
2965 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2966  Token *Result) {
2967  unsigned CharSize;
2968  char Kind = getCharAndSize(StartPtr, CharSize);
2969 
2970  unsigned NumHexDigits;
2971  if (Kind == 'u')
2972  NumHexDigits = 4;
2973  else if (Kind == 'U')
2974  NumHexDigits = 8;
2975  else
2976  return 0;
2977 
2978  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2979  if (Result && !isLexingRawMode())
2980  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2981  return 0;
2982  }
2983 
2984  const char *CurPtr = StartPtr + CharSize;
2985  const char *KindLoc = &CurPtr[-1];
2986 
2987  uint32_t CodePoint = 0;
2988  for (unsigned i = 0; i < NumHexDigits; ++i) {
2989  char C = getCharAndSize(CurPtr, CharSize);
2990 
2991  unsigned Value = llvm::hexDigitValue(C);
2992  if (Value == -1U) {
2993  if (Result && !isLexingRawMode()) {
2994  if (i == 0) {
2995  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2996  << StringRef(KindLoc, 1);
2997  } else {
2998  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2999 
3000  // If the user wrote \U1234, suggest a fixit to \u.
3001  if (i == 4 && NumHexDigits == 8) {
3002  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
3003  Diag(KindLoc, diag::note_ucn_four_not_eight)
3004  << FixItHint::CreateReplacement(URange, "u");
3005  }
3006  }
3007  }
3008 
3009  return 0;
3010  }
3011 
3012  CodePoint <<= 4;
3013  CodePoint += Value;
3014 
3015  CurPtr += CharSize;
3016  }
3017 
3018  if (Result) {
3019  Result->setFlag(Token::HasUCN);
3020  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
3021  StartPtr = CurPtr;
3022  else
3023  while (StartPtr != CurPtr)
3024  (void)getAndAdvanceChar(StartPtr, *Result);
3025  } else {
3026  StartPtr = CurPtr;
3027  }
3028 
3029  // Don't apply C family restrictions to UCNs in assembly mode
3030  if (LangOpts.AsmPreprocessor)
3031  return CodePoint;
3032 
3033  // C99 6.4.3p2: A universal character name shall not specify a character whose
3034  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3035  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3036  // C++11 [lex.charset]p2: If the hexadecimal value for a
3037  // universal-character-name corresponds to a surrogate code point (in the
3038  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3039  // if the hexadecimal value for a universal-character-name outside the
3040  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3041  // string literal corresponds to a control character (in either of the
3042  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3043  // basic source character set, the program is ill-formed.
3044  if (CodePoint < 0xA0) {
3045  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3046  return CodePoint;
3047 
3048  // We don't use isLexingRawMode() here because we need to warn about bad
3049  // UCNs even when skipping preprocessing tokens in a #if block.
3050  if (Result && PP) {
3051  if (CodePoint < 0x20 || CodePoint >= 0x7F)
3052  Diag(BufferPtr, diag::err_ucn_control_character);
3053  else {
3054  char C = static_cast<char>(CodePoint);
3055  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3056  }
3057  }
3058 
3059  return 0;
3060  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3061  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3062  // We don't use isLexingRawMode() here because we need to diagnose bad
3063  // UCNs even when skipping preprocessing tokens in a #if block.
3064  if (Result && PP) {
3065  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3066  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3067  else
3068  Diag(BufferPtr, diag::err_ucn_escape_invalid);
3069  }
3070  return 0;
3071  }
3072 
3073  return CodePoint;
3074 }
3075 
3076 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3077  const char *CurPtr) {
3078  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3080  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3081  UnicodeWhitespaceChars.contains(C)) {
3082  Diag(BufferPtr, diag::ext_unicode_whitespace)
3083  << makeCharRange(*this, BufferPtr, CurPtr);
3084 
3085  Result.setFlag(Token::LeadingSpace);
3086  return true;
3087  }
3088  return false;
3089 }
3090 
3091 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3092  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3094  !PP->isPreprocessedOutput()) {
3096  makeCharRange(*this, BufferPtr, CurPtr),
3097  /*IsFirst=*/true);
3099  makeCharRange(*this, BufferPtr, CurPtr));
3100  }
3101 
3102  MIOpt.ReadToken();
3103  return LexIdentifier(Result, CurPtr);
3104  }
3105 
3107  !PP->isPreprocessedOutput() &&
3108  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3109  // Non-ASCII characters tend to creep into source code unintentionally.
3110  // Instead of letting the parser complain about the unknown token,
3111  // just drop the character.
3112  // Note that we can /only/ do this when the non-ASCII character is actually
3113  // spelled as Unicode, not written as a UCN. The standard requires that
3114  // we not throw away any possible preprocessor tokens, but there's a
3115  // loophole in the mapping of Unicode characters to basic character set
3116  // characters that allows us to map these particular characters to, say,
3117  // whitespace.
3118  Diag(BufferPtr, diag::err_non_ascii)
3119  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3120 
3121  BufferPtr = CurPtr;
3122  return false;
3123  }
3124 
3125  // Otherwise, we have an explicit UCN or a character that's unlikely to show
3126  // up by accident.
3127  MIOpt.ReadToken();
3128  FormTokenWithChars(Result, CurPtr, tok::unknown);
3129  return true;
3130 }
3131 
3132 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3133  IsAtStartOfLine = Result.isAtStartOfLine();
3134  HasLeadingSpace = Result.hasLeadingSpace();
3135  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3136  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3137 }
3138 
3139 bool Lexer::Lex(Token &Result) {
3140  // Start a new token.
3141  Result.startToken();
3142 
3143  // Set up misc whitespace flags for LexTokenInternal.
3144  if (IsAtStartOfLine) {
3145  Result.setFlag(Token::StartOfLine);
3146  IsAtStartOfLine = false;
3147  }
3148 
3149  if (HasLeadingSpace) {
3150  Result.setFlag(Token::LeadingSpace);
3151  HasLeadingSpace = false;
3152  }
3153 
3154  if (HasLeadingEmptyMacro) {
3156  HasLeadingEmptyMacro = false;
3157  }
3158 
3159  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3160  IsAtPhysicalStartOfLine = false;
3161  bool isRawLex = isLexingRawMode();
3162  (void) isRawLex;
3163  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3164  // (After the LexTokenInternal call, the lexer might be destroyed.)
3165  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3166  return returnedToken;
3167 }
3168 
3169 /// LexTokenInternal - This implements a simple C family lexer. It is an
3170 /// extremely performance critical piece of code. This assumes that the buffer
3171 /// has a null character at the end of the file. This returns a preprocessing
3172 /// token, not a normal token, as such, it is an internal interface. It assumes
3173 /// that the Flags of result have been cleared before calling this.
3174 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3175 LexNextToken:
3176  // New token, can't need cleaning yet.
3178  Result.setIdentifierInfo(nullptr);
3179 
3180  // CurPtr - Cache BufferPtr in an automatic variable.
3181  const char *CurPtr = BufferPtr;
3182 
3183  // Small amounts of horizontal whitespace is very common between tokens.
3184  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
3185  ++CurPtr;
3186  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3187  ++CurPtr;
3188 
3189  // If we are keeping whitespace and other tokens, just return what we just
3190  // skipped. The next lexer invocation will return the token after the
3191  // whitespace.
3192  if (isKeepWhitespaceMode()) {
3193  FormTokenWithChars(Result, CurPtr, tok::unknown);
3194  // FIXME: The next token will not have LeadingSpace set.
3195  return true;
3196  }
3197 
3198  BufferPtr = CurPtr;
3199  Result.setFlag(Token::LeadingSpace);
3200  }
3201 
3202  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3203 
3204  // Read a character, advancing over it.
3205  char Char = getAndAdvanceChar(CurPtr, Result);
3207 
3208  switch (Char) {
3209  case 0: // Null.
3210  // Found end of file?
3211  if (CurPtr-1 == BufferEnd)
3212  return LexEndOfFile(Result, CurPtr-1);
3213 
3214  // Check if we are performing code completion.
3215  if (isCodeCompletionPoint(CurPtr-1)) {
3216  // Return the code-completion token.
3217  Result.startToken();
3218  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3219  return true;
3220  }
3221 
3222  if (!isLexingRawMode())
3223  Diag(CurPtr-1, diag::null_in_file);
3224  Result.setFlag(Token::LeadingSpace);
3225  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3226  return true; // KeepWhitespaceMode
3227 
3228  // We know the lexer hasn't changed, so just try again with this lexer.
3229  // (We manually eliminate the tail call to avoid recursion.)
3230  goto LexNextToken;
3231 
3232  case 26: // DOS & CP/M EOF: "^Z".
3233  // If we're in Microsoft extensions mode, treat this as end of file.
3234  if (LangOpts.MicrosoftExt) {
3235  if (!isLexingRawMode())
3236  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3237  return LexEndOfFile(Result, CurPtr-1);
3238  }
3239 
3240  // If Microsoft extensions are disabled, this is just random garbage.
3241  Kind = tok::unknown;
3242  break;
3243 
3244  case '\r':
3245  if (CurPtr[0] == '\n')
3246  (void)getAndAdvanceChar(CurPtr, Result);
3247  LLVM_FALLTHROUGH;
3248  case '\n':
3249  // If we are inside a preprocessor directive and we see the end of line,
3250  // we know we are done with the directive, so return an EOD token.
3252  // Done parsing the "line".
3254 
3255  // Restore comment saving mode, in case it was disabled for directive.
3256  if (PP)
3258 
3259  // Since we consumed a newline, we are back at the start of a line.
3260  IsAtStartOfLine = true;
3261  IsAtPhysicalStartOfLine = true;
3262 
3263  Kind = tok::eod;
3264  break;
3265  }
3266 
3267  // No leading whitespace seen so far.
3269 
3270  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3271  return true; // KeepWhitespaceMode
3272 
3273  // We only saw whitespace, so just try again with this lexer.
3274  // (We manually eliminate the tail call to avoid recursion.)
3275  goto LexNextToken;
3276  case ' ':
3277  case '\t':
3278  case '\f':
3279  case '\v':
3280  SkipHorizontalWhitespace:
3281  Result.setFlag(Token::LeadingSpace);
3282  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3283  return true; // KeepWhitespaceMode
3284 
3285  SkipIgnoredUnits:
3286  CurPtr = BufferPtr;
3287 
3288  // If the next token is obviously a // or /* */ comment, skip it efficiently
3289  // too (without going through the big switch stmt).
3290  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3291  LangOpts.LineComment &&
3292  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3293  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3294  return true; // There is a token to return.
3295  goto SkipIgnoredUnits;
3296  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3297  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3298  return true; // There is a token to return.
3299  goto SkipIgnoredUnits;
3300  } else if (isHorizontalWhitespace(*CurPtr)) {
3301  goto SkipHorizontalWhitespace;
3302  }
3303  // We only saw whitespace, so just try again with this lexer.
3304  // (We manually eliminate the tail call to avoid recursion.)
3305  goto LexNextToken;
3306 
3307  // C99 6.4.4.1: Integer Constants.
3308  // C99 6.4.4.2: Floating Constants.
3309  case '0': case '1': case '2': case '3': case '4':
3310  case '5': case '6': case '7': case '8': case '9':
3311  // Notify MIOpt that we read a non-whitespace/non-comment token.
3312  MIOpt.ReadToken();
3313  return LexNumericConstant(Result, CurPtr);
3314 
3315  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3316  // Notify MIOpt that we read a non-whitespace/non-comment token.
3317  MIOpt.ReadToken();
3318 
3319  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3320  Char = getCharAndSize(CurPtr, SizeTmp);
3321 
3322  // UTF-16 string literal
3323  if (Char == '"')
3324  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3325  tok::utf16_string_literal);
3326 
3327  // UTF-16 character constant
3328  if (Char == '\'')
3329  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3330  tok::utf16_char_constant);
3331 
3332  // UTF-16 raw string literal
3333  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3334  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3335  return LexRawStringLiteral(Result,
3336  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3337  SizeTmp2, Result),
3338  tok::utf16_string_literal);
3339 
3340  if (Char == '8') {
3341  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3342 
3343  // UTF-8 string literal
3344  if (Char2 == '"')
3345  return LexStringLiteral(Result,
3346  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3347  SizeTmp2, Result),
3348  tok::utf8_string_literal);
3349  if (Char2 == '\'' && LangOpts.CPlusPlus17)
3350  return LexCharConstant(
3351  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3352  SizeTmp2, Result),
3353  tok::utf8_char_constant);
3354 
3355  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3356  unsigned SizeTmp3;
3357  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3358  // UTF-8 raw string literal
3359  if (Char3 == '"') {
3360  return LexRawStringLiteral(Result,
3361  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3362  SizeTmp2, Result),
3363  SizeTmp3, Result),
3364  tok::utf8_string_literal);
3365  }
3366  }
3367  }
3368  }
3369 
3370  // treat u like the start of an identifier.
3371  return LexIdentifier(Result, CurPtr);
3372 
3373  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3374  // Notify MIOpt that we read a non-whitespace/non-comment token.
3375  MIOpt.ReadToken();
3376 
3377  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3378  Char = getCharAndSize(CurPtr, SizeTmp);
3379 
3380  // UTF-32 string literal
3381  if (Char == '"')
3382  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3383  tok::utf32_string_literal);
3384 
3385  // UTF-32 character constant
3386  if (Char == '\'')
3387  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3388  tok::utf32_char_constant);
3389 
3390  // UTF-32 raw string literal
3391  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3392  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3393  return LexRawStringLiteral(Result,
3394  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3395  SizeTmp2, Result),
3396  tok::utf32_string_literal);
3397  }
3398 
3399  // treat U like the start of an identifier.
3400  return LexIdentifier(Result, CurPtr);
3401 
3402  case 'R': // Identifier or C++0x raw string literal
3403  // Notify MIOpt that we read a non-whitespace/non-comment token.
3404  MIOpt.ReadToken();
3405 
3406  if (LangOpts.CPlusPlus11) {
3407  Char = getCharAndSize(CurPtr, SizeTmp);
3408 
3409  if (Char == '"')
3410  return LexRawStringLiteral(Result,
3411  ConsumeChar(CurPtr, SizeTmp, Result),
3412  tok::string_literal);
3413  }
3414 
3415  // treat R like the start of an identifier.
3416  return LexIdentifier(Result, CurPtr);
3417 
3418  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3419  // Notify MIOpt that we read a non-whitespace/non-comment token.
3420  MIOpt.ReadToken();
3421  Char = getCharAndSize(CurPtr, SizeTmp);
3422 
3423  // Wide string literal.
3424  if (Char == '"')
3425  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3426  tok::wide_string_literal);
3427 
3428  // Wide raw string literal.
3429  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3430  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3431  return LexRawStringLiteral(Result,
3432  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3433  SizeTmp2, Result),
3434  tok::wide_string_literal);
3435 
3436  // Wide character constant.
3437  if (Char == '\'')
3438  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3439  tok::wide_char_constant);
3440  // FALL THROUGH, treating L like the start of an identifier.
3441  LLVM_FALLTHROUGH;
3442 
3443  // C99 6.4.2: Identifiers.
3444  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3445  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3446  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3447  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3448  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3449  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3450  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3451  case 'v': case 'w': case 'x': case 'y': case 'z':
3452  case '_':
3453  // Notify MIOpt that we read a non-whitespace/non-comment token.
3454  MIOpt.ReadToken();
3455  return LexIdentifier(Result, CurPtr);
3456 
3457  case '$': // $ in identifiers.
3458  if (LangOpts.DollarIdents) {
3459  if (!isLexingRawMode())
3460  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3461  // Notify MIOpt that we read a non-whitespace/non-comment token.
3462  MIOpt.ReadToken();
3463  return LexIdentifier(Result, CurPtr);
3464  }
3465 
3466  Kind = tok::unknown;
3467  break;
3468 
3469  // C99 6.4.4: Character Constants.
3470  case '\'':
3471  // Notify MIOpt that we read a non-whitespace/non-comment token.
3472  MIOpt.ReadToken();
3473  return LexCharConstant(Result, CurPtr, tok::char_constant);
3474 
3475  // C99 6.4.5: String Literals.
3476  case '"':
3477  // Notify MIOpt that we read a non-whitespace/non-comment token.
3478  MIOpt.ReadToken();
3479  return LexStringLiteral(Result, CurPtr,
3480  ParsingFilename ? tok::header_name
3481  : tok::string_literal);
3482 
3483  // C99 6.4.6: Punctuators.
3484  case '?':
3485  Kind = tok::question;
3486  break;
3487  case '[':
3488  Kind = tok::l_square;
3489  break;
3490  case ']':
3491  Kind = tok::r_square;
3492  break;
3493  case '(':
3494  Kind = tok::l_paren;
3495  break;
3496  case ')':
3497  Kind = tok::r_paren;
3498  break;
3499  case '{':
3500  Kind = tok::l_brace;
3501  break;
3502  case '}':
3503  Kind = tok::r_brace;
3504  break;
3505  case '.':
3506  Char = getCharAndSize(CurPtr, SizeTmp);
3507  if (Char >= '0' && Char <= '9') {
3508  // Notify MIOpt that we read a non-whitespace/non-comment token.
3509  MIOpt.ReadToken();
3510 
3511  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3512  } else if (LangOpts.CPlusPlus && Char == '*') {
3513  Kind = tok::periodstar;
3514  CurPtr += SizeTmp;
3515  } else if (Char == '.' &&
3516  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3517  Kind = tok::ellipsis;
3518  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3519  SizeTmp2, Result);
3520  } else {
3521  Kind = tok::period;
3522  }
3523  break;
3524  case '&':
3525  Char = getCharAndSize(CurPtr, SizeTmp);
3526  if (Char == '&') {
3527  Kind = tok::ampamp;
3528  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3529  } else if (Char == '=') {
3530  Kind = tok::ampequal;
3531  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3532  } else {
3533  Kind = tok::amp;
3534  }
3535  break;
3536  case '*':
3537  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3538  Kind = tok::starequal;
3539  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3540  } else {
3541  Kind = tok::star;
3542  }
3543  break;
3544  case '+':
3545  Char = getCharAndSize(CurPtr, SizeTmp);
3546  if (Char == '+') {
3547  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3548  Kind = tok::plusplus;
3549  } else if (Char == '=') {
3550  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3551  Kind = tok::plusequal;
3552  } else {
3553  Kind = tok::plus;
3554  }
3555  break;
3556  case '-':
3557  Char = getCharAndSize(CurPtr, SizeTmp);
3558  if (Char == '-') { // --
3559  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3560  Kind = tok::minusminus;
3561  } else if (Char == '>' && LangOpts.CPlusPlus &&
3562  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3563  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3564  SizeTmp2, Result);
3565  Kind = tok::arrowstar;
3566  } else if (Char == '>') { // ->
3567  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3568  Kind = tok::arrow;
3569  } else if (Char == '=') { // -=
3570  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3571  Kind = tok::minusequal;
3572  } else {
3573  Kind = tok::minus;
3574  }
3575  break;
3576  case '~':
3577  Kind = tok::tilde;
3578  break;
3579  case '!':
3580  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3581  Kind = tok::exclaimequal;
3582  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3583  } else {
3584  Kind = tok::exclaim;
3585  }
3586  break;
3587  case '/':
3588  // 6.4.9: Comments
3589  Char = getCharAndSize(CurPtr, SizeTmp);
3590  if (Char == '/') { // Line comment.
3591  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3592  // want to lex this as a comment. There is one problem with this though,
3593  // that in one particular corner case, this can change the behavior of the
3594  // resultant program. For example, In "foo //**/ bar", C89 would lex
3595  // this as "foo / bar" and languages with Line comments would lex it as
3596  // "foo". Check to see if the character after the second slash is a '*'.
3597  // If so, we will lex that as a "/" instead of the start of a comment.
3598  // However, we never do this if we are just preprocessing.
3599  bool TreatAsComment = LangOpts.LineComment &&
3600  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3601  if (!TreatAsComment)
3602  if (!(PP && PP->isPreprocessedOutput()))
3603  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3604 
3605  if (TreatAsComment) {
3606  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3607  TokAtPhysicalStartOfLine))
3608  return true; // There is a token to return.
3609 
3610  // It is common for the tokens immediately after a // comment to be
3611  // whitespace (indentation for the next line). Instead of going through
3612  // the big switch, handle it efficiently now.
3613  goto SkipIgnoredUnits;
3614  }
3615  }
3616 
3617  if (Char == '*') { // /**/ comment.
3618  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3619  TokAtPhysicalStartOfLine))
3620  return true; // There is a token to return.
3621 
3622  // We only saw whitespace, so just try again with this lexer.
3623  // (We manually eliminate the tail call to avoid recursion.)
3624  goto LexNextToken;
3625  }
3626 
3627  if (Char == '=') {
3628  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3629  Kind = tok::slashequal;
3630  } else {
3631  Kind = tok::slash;
3632  }
3633  break;
3634  case '%':
3635  Char = getCharAndSize(CurPtr, SizeTmp);
3636  if (Char == '=') {
3637  Kind = tok::percentequal;
3638  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3639  } else if (LangOpts.Digraphs && Char == '>') {
3640  Kind = tok::r_brace; // '%>' -> '}'
3641  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3642  } else if (LangOpts.Digraphs && Char == ':') {
3643  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3644  Char = getCharAndSize(CurPtr, SizeTmp);
3645  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3646  Kind = tok::hashhash; // '%:%:' -> '##'
3647  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3648  SizeTmp2, Result);
3649  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3650  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3651  if (!isLexingRawMode())
3652  Diag(BufferPtr, diag::ext_charize_microsoft);
3653  Kind = tok::hashat;
3654  } else { // '%:' -> '#'
3655  // We parsed a # character. If this occurs at the start of the line,
3656  // it's actually the start of a preprocessing directive. Callback to
3657  // the preprocessor to handle it.
3658  // TODO: -fpreprocessed mode??
3659  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3660  goto HandleDirective;
3661 
3662  Kind = tok::hash;
3663  }
3664  } else {
3665  Kind = tok::percent;
3666  }
3667  break;
3668  case '<':
3669  Char = getCharAndSize(CurPtr, SizeTmp);
3670  if (ParsingFilename) {
3671  return LexAngledStringLiteral(Result, CurPtr);
3672  } else if (Char == '<') {
3673  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3674  if (After == '=') {
3675  Kind = tok::lesslessequal;
3676  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3677  SizeTmp2, Result);
3678  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3679  // If this is actually a '<<<<<<<' version control conflict marker,
3680  // recognize it as such and recover nicely.
3681  goto LexNextToken;
3682  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3683  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3684  // ignore it.
3685  goto LexNextToken;
3686  } else if (LangOpts.CUDA && After == '<') {
3687  Kind = tok::lesslessless;
3688  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3689  SizeTmp2, Result);
3690  } else {
3691  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3692  Kind = tok::lessless;
3693  }
3694  } else if (Char == '=') {
3695  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3696  if (After == '>') {
3697  if (getLangOpts().CPlusPlus2a) {
3698  if (!isLexingRawMode())
3699  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3700  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3701  SizeTmp2, Result);
3702  Kind = tok::spaceship;
3703  break;
3704  }
3705  // Suggest adding a space between the '<=' and the '>' to avoid a
3706  // change in semantics if this turns up in C++ <=17 mode.
3707  if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3708  Diag(BufferPtr, diag::warn_cxx2a_compat_spaceship)
3710  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3711  }
3712  }
3713  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3714  Kind = tok::lessequal;
3715  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3716  if (LangOpts.CPlusPlus11 &&
3717  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3718  // C++0x [lex.pptoken]p3:
3719  // Otherwise, if the next three characters are <:: and the subsequent
3720  // character is neither : nor >, the < is treated as a preprocessor
3721  // token by itself and not as the first character of the alternative
3722  // token <:.
3723  unsigned SizeTmp3;
3724  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3725  if (After != ':' && After != '>') {
3726  Kind = tok::less;
3727  if (!isLexingRawMode())
3728  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3729  break;
3730  }
3731  }
3732 
3733  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3734  Kind = tok::l_square;
3735  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3736  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3737  Kind = tok::l_brace;
3738  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3739  lexEditorPlaceholder(Result, CurPtr)) {
3740  return true;
3741  } else {
3742  Kind = tok::less;
3743  }
3744  break;
3745  case '>':
3746  Char = getCharAndSize(CurPtr, SizeTmp);
3747  if (Char == '=') {
3748  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3749  Kind = tok::greaterequal;
3750  } else if (Char == '>') {
3751  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3752  if (After == '=') {
3753  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3754  SizeTmp2, Result);
3755  Kind = tok::greatergreaterequal;
3756  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3757  // If this is actually a '>>>>' conflict marker, recognize it as such
3758  // and recover nicely.
3759  goto LexNextToken;
3760  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3761  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3762  goto LexNextToken;
3763  } else if (LangOpts.CUDA && After == '>') {
3764  Kind = tok::greatergreatergreater;
3765  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3766  SizeTmp2, Result);
3767  } else {
3768  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3769  Kind = tok::greatergreater;
3770  }
3771  } else {
3772  Kind = tok::greater;
3773  }
3774  break;
3775  case '^':
3776  Char = getCharAndSize(CurPtr, SizeTmp);
3777  if (Char == '=') {
3778  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3779  Kind = tok::caretequal;
3780  } else if (LangOpts.OpenCL && Char == '^') {
3781  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3782  Kind = tok::caretcaret;
3783  } else {
3784  Kind = tok::caret;
3785  }
3786  break;
3787  case '|':
3788  Char = getCharAndSize(CurPtr, SizeTmp);
3789  if (Char == '=') {
3790  Kind = tok::pipeequal;
3791  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3792  } else if (Char == '|') {
3793  // If this is '|||||||' and we're in a conflict marker, ignore it.
3794  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3795  goto LexNextToken;
3796  Kind = tok::pipepipe;
3797  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3798  } else {
3799  Kind = tok::pipe;
3800  }
3801  break;
3802  case ':':
3803  Char = getCharAndSize(CurPtr, SizeTmp);
3804  if (LangOpts.Digraphs && Char == '>') {
3805  Kind = tok::r_square; // ':>' -> ']'
3806  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3807  } else if ((LangOpts.CPlusPlus ||
3808  LangOpts.DoubleSquareBracketAttributes) &&
3809  Char == ':') {
3810  Kind = tok::coloncolon;
3811  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3812  } else {
3813  Kind = tok::colon;
3814  }
3815  break;
3816  case ';':
3817  Kind = tok::semi;
3818  break;
3819  case '=':
3820  Char = getCharAndSize(CurPtr, SizeTmp);
3821  if (Char == '=') {
3822  // If this is '====' and we're in a conflict marker, ignore it.
3823  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3824  goto LexNextToken;
3825 
3826  Kind = tok::equalequal;
3827  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3828  } else {
3829  Kind = tok::equal;
3830  }
3831  break;
3832  case ',':
3833  Kind = tok::comma;
3834  break;
3835  case '#':
3836  Char = getCharAndSize(CurPtr, SizeTmp);
3837  if (Char == '#') {
3838  Kind = tok::hashhash;
3839  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3840  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3841  Kind = tok::hashat;
3842  if (!isLexingRawMode())
3843  Diag(BufferPtr, diag::ext_charize_microsoft);
3844  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3845  } else {
3846  // We parsed a # character. If this occurs at the start of the line,
3847  // it's actually the start of a preprocessing directive. Callback to
3848  // the preprocessor to handle it.
3849  // TODO: -fpreprocessed mode??
3850  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3851  goto HandleDirective;
3852 
3853  Kind = tok::hash;
3854  }
3855  break;
3856 
3857  case '@':
3858  // Objective C support.
3859  if (CurPtr[-1] == '@' && LangOpts.ObjC)
3860  Kind = tok::at;
3861  else
3862  Kind = tok::unknown;
3863  break;
3864 
3865  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3866  case '\\':
3867  if (!LangOpts.AsmPreprocessor) {
3868  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3869  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3870  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3871  return true; // KeepWhitespaceMode
3872 
3873  // We only saw whitespace, so just try again with this lexer.
3874  // (We manually eliminate the tail call to avoid recursion.)
3875  goto LexNextToken;
3876  }
3877 
3878  return LexUnicode(Result, CodePoint, CurPtr);
3879  }
3880  }
3881 
3882  Kind = tok::unknown;
3883  break;
3884 
3885  default: {
3886  if (isASCII(Char)) {
3887  Kind = tok::unknown;
3888  break;
3889  }
3890 
3891  llvm::UTF32 CodePoint;
3892 
3893  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3894  // an escaped newline.
3895  --CurPtr;
3896  llvm::ConversionResult Status =
3897  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3898  (const llvm::UTF8 *)BufferEnd,
3899  &CodePoint,
3900  llvm::strictConversion);
3901  if (Status == llvm::conversionOK) {
3902  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3903  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3904  return true; // KeepWhitespaceMode
3905 
3906  // We only saw whitespace, so just try again with this lexer.
3907  // (We manually eliminate the tail call to avoid recursion.)
3908  goto LexNextToken;
3909  }
3910  return LexUnicode(Result, CodePoint, CurPtr);
3911  }
3912 
3914  PP->isPreprocessedOutput()) {
3915  ++CurPtr;
3916  Kind = tok::unknown;
3917  break;
3918  }
3919 
3920  // Non-ASCII characters tend to creep into source code unintentionally.
3921  // Instead of letting the parser complain about the unknown token,
3922  // just diagnose the invalid UTF-8, then drop the character.
3923  Diag(CurPtr, diag::err_invalid_utf8);
3924 
3925  BufferPtr = CurPtr+1;
3926  // We're pretending the character didn't exist, so just try again with
3927  // this lexer.
3928  // (We manually eliminate the tail call to avoid recursion.)
3929  goto LexNextToken;
3930  }
3931  }
3932 
3933  // Notify MIOpt that we read a non-whitespace/non-comment token.
3934  MIOpt.ReadToken();
3935 
3936  // Update the location of token as well as BufferPtr.
3937  FormTokenWithChars(Result, CurPtr, Kind);
3938  return true;
3939 
3940 HandleDirective:
3941  // We parsed a # character and it's the start of a preprocessing directive.
3942 
3943  FormTokenWithChars(Result, CurPtr, tok::hash);
3944  PP->HandleDirective(Result);
3945 
3946  if (PP->hadModuleLoaderFatalFailure()) {
3947  // With a fatal failure in the module loader, we abort parsing.
3948  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3949  return true;
3950  }
3951 
3952  // We parsed the directive; lex a token with the new state.
3953  return false;
3954 }
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:397
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:14859
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:1431
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:1300
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:543
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:931
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:800
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:1102
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
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:1071
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
Keeps track of the various options that can be enabled, which controls the dialect of C or C++ that i...
Definition: LangOptions.h:53
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:506
static bool isNewLineEscaped(const char *BufferStart, const char *Str)
Checks whether new line pointed by Str is preceded by escape sequence.
Definition: Lexer.cpp:1051
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 ...
bool skipOver(unsigned NumBytes)
Skip over NumBytes bytes.
Definition: Lexer.cpp:221
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:455
void HandleDirective(Token &Result)
Callback invoked when the lexer sees a # token at the start of a line.
Concrete class used by the front-end to report problems and issues.
Definition: Diagnostic.h:149
Defines the Diagnostic-related interfaces.
SourceLocation getSpellingLoc(SourceLocation Loc) const
Given a SourceLocation object, return the spelling location referenced by the ID. ...
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:2415
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:1468
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:1053
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:1827
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:939
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:822
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:778
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:582
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:444
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:1270
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:1126
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:718
PreambleDirectiveKind
Definition: Lexer.cpp:575
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:952
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:489
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:1685
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:911
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:1513
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:914
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:1237
static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C, CharSourceRange Range, bool IsFirst)
Definition: Lexer.cpp:1474
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:554
static CharSourceRange makeRangeFromFileLocs(CharSourceRange Range, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp:848
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:1145
__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:975
static StringRef getImmediateMacroNameForDiagnostics(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Retrieve the name of the immediate macro expansion.
Definition: Lexer.cpp:1022
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:875
static size_t getSpellingSlow(const Token &Tok, const char *BufPtr, const LangOptions &LangOpts, char *Spelling)
Slow case of getSpelling.
Definition: Lexer.cpp:270
static FixItHint CreateRemoval(CharSourceRange RemoveRange)
Create a code modification hint that removes the given source range.
Definition: Diagnostic.h:118
bool isHandleIdentifierCase() const
Return true if the Preprocessor::HandleIdentifier must be called on a token of this identifier...
bool isLexingRawMode() const
Return true if this lexer is in raw mode or not.
LLVM_READONLY bool isIdentifierBody(unsigned char c, bool AllowDollar=false)
Returns true if this is a body character of a C identifier, which is [a-zA-Z0-9_].
Definition: CharInfo.h: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:2925
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:1451
SourceLocation getEnd() const
static FixItHint CreateInsertion(SourceLocation InsertionLoc, StringRef Code, bool BeforePreviousInsertions=false)
Create a code modification hint that inserts the given code string at a specific location.
Definition: Diagnostic.h:92
PreprocessorOptions & getPreprocessorOpts() const
Retrieve the preprocessor options used to initialize this preprocessor.
Definition: Preprocessor.h:902
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:1155
static bool isIdentifierBodyChar(char c, const LangOptions &LangOpts)
Returns true if the given character could appear in an identifier.
Definition: Lexer.cpp:1047
void ReadToEndOfLine(SmallVectorImpl< char > *Result=nullptr)
ReadToEndOfLine - Read the rest of the current preprocessor line as an uninterpreted string...
Definition: Lexer.cpp:2657
__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:904
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:1174
static const char * FindConflictEnd(const char *CurPtr, const char *BufferEnd, ConflictMarkerKind CMK)
Find the end of a version control conflict marker.
Definition: Lexer.cpp:2821
static void StringifyImpl(T &Str, char Quote)
Definition: Lexer.cpp:230
static FixItHint CreateReplacement(CharSourceRange RemoveRange, StringRef Code)
Create a code modification hint that replaces the given source range with the given code string...
Definition: Diagnostic.h:129
void SetCommentRetentionState(bool Mode)
SetCommentRetentionMode - Change the comment retention mode of the lexer to the specified mode...
Definition: Lexer.h: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:825
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:973
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:255
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:128
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.