clang  10.0.0svn
Lexer.cpp
Go to the documentation of this file.
1 //===- Lexer.cpp - C Language Family Lexer --------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the Lexer and Token interfaces.
10 //
11 //===----------------------------------------------------------------------===//
12 
13 #include "clang/Lex/Lexer.h"
14 #include "UnicodeCharSets.h"
15 #include "clang/Basic/CharInfo.h"
20 #include "clang/Basic/TokenKinds.h"
24 #include "clang/Lex/Preprocessor.h"
26 #include "clang/Lex/Token.h"
27 #include "clang/Basic/Diagnostic.h"
28 #include "clang/Basic/LLVM.h"
29 #include "clang/Basic/TokenKinds.h"
30 #include "llvm/ADT/None.h"
31 #include "llvm/ADT/Optional.h"
32 #include "llvm/ADT/StringExtras.h"
33 #include "llvm/ADT/StringSwitch.h"
34 #include "llvm/ADT/StringRef.h"
35 #include "llvm/Support/Compiler.h"
36 #include "llvm/Support/ConvertUTF.h"
37 #include "llvm/Support/MathExtras.h"
38 #include "llvm/Support/MemoryBuffer.h"
39 #include "llvm/Support/NativeFormatting.h"
40 #include "llvm/Support/UnicodeCharRanges.h"
41 #include <algorithm>
42 #include <cassert>
43 #include <cstddef>
44 #include <cstdint>
45 #include <cstring>
46 #include <string>
47 #include <tuple>
48 #include <utility>
49 
50 using namespace clang;
51 
52 //===----------------------------------------------------------------------===//
53 // Token Class Implementation
54 //===----------------------------------------------------------------------===//
55 
56 /// isObjCAtKeyword - Return true if we have an ObjC keyword identifier.
58  if (isAnnotation())
59  return false;
61  return II->getObjCKeywordID() == objcKey;
62  return false;
63 }
64 
65 /// getObjCKeywordID - Return the ObjC keyword kind.
67  if (isAnnotation())
68  return tok::objc_not_keyword;
70  return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword;
71 }
72 
73 //===----------------------------------------------------------------------===//
74 // Lexer Class Implementation
75 //===----------------------------------------------------------------------===//
76 
77 void Lexer::anchor() {}
78 
79 void Lexer::InitLexer(const char *BufStart, const char *BufPtr,
80  const char *BufEnd) {
81  BufferStart = BufStart;
82  BufferPtr = BufPtr;
83  BufferEnd = BufEnd;
84 
85  assert(BufEnd[0] == 0 &&
86  "We assume that the input buffer has a null character at the end"
87  " to simplify lexing!");
88 
89  // Check whether we have a BOM in the beginning of the buffer. If yes - act
90  // accordingly. Right now we support only UTF-8 with and without BOM, so, just
91  // skip the UTF-8 BOM if it's present.
92  if (BufferStart == BufferPtr) {
93  // Determine the size of the BOM.
94  StringRef Buf(BufferStart, BufferEnd - BufferStart);
95  size_t BOMLength = llvm::StringSwitch<size_t>(Buf)
96  .StartsWith("\xEF\xBB\xBF", 3) // UTF-8 BOM
97  .Default(0);
98 
99  // Skip the BOM.
100  BufferPtr += BOMLength;
101  }
102 
103  Is_PragmaLexer = false;
104  CurrentConflictMarkerState = CMK_None;
105 
106  // Start of the file is a start of line.
107  IsAtStartOfLine = true;
108  IsAtPhysicalStartOfLine = true;
109 
110  HasLeadingSpace = false;
111  HasLeadingEmptyMacro = false;
112 
113  // We are not after parsing a #.
114  ParsingPreprocessorDirective = false;
115 
116  // We are not after parsing #include.
117  ParsingFilename = false;
118 
119  // We are not in raw mode. Raw mode disables diagnostics and interpretation
120  // of tokens (e.g. identifiers, thus disabling macro expansion). It is used
121  // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block
122  // or otherwise skipping over tokens.
123  LexingRawMode = false;
124 
125  // Default to not keeping comments.
126  ExtendedTokenMode = 0;
127 }
128 
129 /// Lexer constructor - Create a new lexer object for the specified buffer
130 /// with the specified preprocessor managing the lexing process. This lexer
131 /// assumes that the associated file buffer and Preprocessor objects will
132 /// outlive it, so it doesn't take ownership of either of them.
133 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *InputFile, Preprocessor &PP)
134  : PreprocessorLexer(&PP, FID),
135  FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)),
136  LangOpts(PP.getLangOpts()) {
137  InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(),
138  InputFile->getBufferEnd());
139 
141 }
142 
143 /// Lexer constructor - Create a new raw lexer object. This object is only
144 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
145 /// range will outlive it, so it doesn't take ownership of it.
146 Lexer::Lexer(SourceLocation fileloc, const LangOptions &langOpts,
147  const char *BufStart, const char *BufPtr, const char *BufEnd)
148  : FileLoc(fileloc), LangOpts(langOpts) {
149  InitLexer(BufStart, BufPtr, BufEnd);
150 
151  // We *are* in raw mode.
152  LexingRawMode = true;
153 }
154 
155 /// Lexer constructor - Create a new raw lexer object. This object is only
156 /// suitable for calls to 'LexFromRawLexer'. This lexer assumes that the text
157 /// range will outlive it, so it doesn't take ownership of it.
158 Lexer::Lexer(FileID FID, const llvm::MemoryBuffer *FromFile,
159  const SourceManager &SM, const LangOptions &langOpts)
160  : Lexer(SM.getLocForStartOfFile(FID), langOpts, FromFile->getBufferStart(),
161  FromFile->getBufferStart(), FromFile->getBufferEnd()) {}
162 
164  assert(PP && "Cannot reset token mode without a preprocessor");
165  if (LangOpts.TraditionalCPP)
166  SetKeepWhitespaceMode(true);
167  else
169 }
170 
171 /// Create_PragmaLexer: Lexer constructor - Create a new lexer object for
172 /// _Pragma expansion. This has a variety of magic semantics that this method
173 /// sets up. It returns a new'd Lexer that must be delete'd when done.
174 ///
175 /// On entrance to this routine, TokStartLoc is a macro location which has a
176 /// spelling loc that indicates the bytes to be lexed for the token and an
177 /// expansion location that indicates where all lexed tokens should be
178 /// "expanded from".
179 ///
180 /// TODO: It would really be nice to make _Pragma just be a wrapper around a
181 /// normal lexer that remaps tokens as they fly by. This would require making
182 /// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer
183 /// interface that could handle this stuff. This would pull GetMappedTokenLoc
184 /// out of the critical path of the lexer!
185 ///
187  SourceLocation ExpansionLocStart,
188  SourceLocation ExpansionLocEnd,
189  unsigned TokLen, Preprocessor &PP) {
191 
192  // Create the lexer as if we were going to lex the file normally.
193  FileID SpellingFID = SM.getFileID(SpellingLoc);
194  const llvm::MemoryBuffer *InputFile = SM.getBuffer(SpellingFID);
195  Lexer *L = new Lexer(SpellingFID, InputFile, PP);
196 
197  // Now that the lexer is created, change the start/end locations so that we
198  // just lex the subsection of the file that we want. This is lexing from a
199  // scratch buffer.
200  const char *StrData = SM.getCharacterData(SpellingLoc);
201 
202  L->BufferPtr = StrData;
203  L->BufferEnd = StrData+TokLen;
204  assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!");
205 
206  // Set the SourceLocation with the remapping information. This ensures that
207  // GetMappedTokenLoc will remap the tokens as they are lexed.
208  L->FileLoc = SM.createExpansionLoc(SM.getLocForStartOfFile(SpellingFID),
209  ExpansionLocStart,
210  ExpansionLocEnd, TokLen);
211 
212  // Ensure that the lexer thinks it is inside a directive, so that end \n will
213  // return an EOD token.
215 
216  // This lexer really is for _Pragma.
217  L->Is_PragmaLexer = true;
218  return L;
219 }
220 
221 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.CPlusPlus11 || LangOpts.C11) {
1435  static const llvm::sys::UnicodeCharSet C11AllowedIDChars(
1437  return C11AllowedIDChars.contains(C);
1438  } else if (LangOpts.CPlusPlus) {
1439  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1441  return CXX03AllowedIDChars.contains(C);
1442  } else {
1443  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1445  return C99AllowedIDChars.contains(C);
1446  }
1447 }
1448 
1449 static bool isAllowedInitiallyIDChar(uint32_t C, const LangOptions &LangOpts) {
1450  assert(isAllowedIDChar(C, LangOpts));
1451  if (LangOpts.AsmPreprocessor) {
1452  return false;
1453  } else if (LangOpts.CPlusPlus11 || LangOpts.C11) {
1454  static const llvm::sys::UnicodeCharSet C11DisallowedInitialIDChars(
1456  return !C11DisallowedInitialIDChars.contains(C);
1457  } else if (LangOpts.CPlusPlus) {
1458  return true;
1459  } else {
1460  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1462  return !C99DisallowedInitialIDChars.contains(C);
1463  }
1464 }
1465 
1466 static inline CharSourceRange makeCharRange(Lexer &L, const char *Begin,
1467  const char *End) {
1469  L.getSourceLocation(End));
1470 }
1471 
1472 static void maybeDiagnoseIDCharCompat(DiagnosticsEngine &Diags, uint32_t C,
1473  CharSourceRange Range, bool IsFirst) {
1474  // Check C99 compatibility.
1475  if (!Diags.isIgnored(diag::warn_c99_compat_unicode_id, Range.getBegin())) {
1476  enum {
1477  CannotAppearInIdentifier = 0,
1478  CannotStartIdentifier
1479  };
1480 
1481  static const llvm::sys::UnicodeCharSet C99AllowedIDChars(
1483  static const llvm::sys::UnicodeCharSet C99DisallowedInitialIDChars(
1485  if (!C99AllowedIDChars.contains(C)) {
1486  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1487  << Range
1488  << CannotAppearInIdentifier;
1489  } else if (IsFirst && C99DisallowedInitialIDChars.contains(C)) {
1490  Diags.Report(Range.getBegin(), diag::warn_c99_compat_unicode_id)
1491  << Range
1492  << CannotStartIdentifier;
1493  }
1494  }
1495 
1496  // Check C++98 compatibility.
1497  if (!Diags.isIgnored(diag::warn_cxx98_compat_unicode_id, Range.getBegin())) {
1498  static const llvm::sys::UnicodeCharSet CXX03AllowedIDChars(
1500  if (!CXX03AllowedIDChars.contains(C)) {
1501  Diags.Report(Range.getBegin(), diag::warn_cxx98_compat_unicode_id)
1502  << Range;
1503  }
1504  }
1505 }
1506 
1507 /// After encountering UTF-8 character C and interpreting it as an identifier
1508 /// character, check whether it's a homoglyph for a common non-identifier
1509 /// source character that is unlikely to be an intentional identifier
1510 /// character and warn if so.
1511 static void maybeDiagnoseUTF8Homoglyph(DiagnosticsEngine &Diags, uint32_t C,
1512  CharSourceRange Range) {
1513  // FIXME: Handle Unicode quotation marks (smart quotes, fullwidth quotes).
1514  struct HomoglyphPair {
1515  uint32_t Character;
1516  char LooksLike;
1517  bool operator<(HomoglyphPair R) const { return Character < R.Character; }
1518  };
1519  static constexpr HomoglyphPair SortedHomoglyphs[] = {
1520  {U'\u00ad', 0}, // SOFT HYPHEN
1521  {U'\u01c3', '!'}, // LATIN LETTER RETROFLEX CLICK
1522  {U'\u037e', ';'}, // GREEK QUESTION MARK
1523  {U'\u200b', 0}, // ZERO WIDTH SPACE
1524  {U'\u200c', 0}, // ZERO WIDTH NON-JOINER
1525  {U'\u200d', 0}, // ZERO WIDTH JOINER
1526  {U'\u2060', 0}, // WORD JOINER
1527  {U'\u2061', 0}, // FUNCTION APPLICATION
1528  {U'\u2062', 0}, // INVISIBLE TIMES
1529  {U'\u2063', 0}, // INVISIBLE SEPARATOR
1530  {U'\u2064', 0}, // INVISIBLE PLUS
1531  {U'\u2212', '-'}, // MINUS SIGN
1532  {U'\u2215', '/'}, // DIVISION SLASH
1533  {U'\u2216', '\\'}, // SET MINUS
1534  {U'\u2217', '*'}, // ASTERISK OPERATOR
1535  {U'\u2223', '|'}, // DIVIDES
1536  {U'\u2227', '^'}, // LOGICAL AND
1537  {U'\u2236', ':'}, // RATIO
1538  {U'\u223c', '~'}, // TILDE OPERATOR
1539  {U'\ua789', ':'}, // MODIFIER LETTER COLON
1540  {U'\ufeff', 0}, // ZERO WIDTH NO-BREAK SPACE
1541  {U'\uff01', '!'}, // FULLWIDTH EXCLAMATION MARK
1542  {U'\uff03', '#'}, // FULLWIDTH NUMBER SIGN
1543  {U'\uff04', '$'}, // FULLWIDTH DOLLAR SIGN
1544  {U'\uff05', '%'}, // FULLWIDTH PERCENT SIGN
1545  {U'\uff06', '&'}, // FULLWIDTH AMPERSAND
1546  {U'\uff08', '('}, // FULLWIDTH LEFT PARENTHESIS
1547  {U'\uff09', ')'}, // FULLWIDTH RIGHT PARENTHESIS
1548  {U'\uff0a', '*'}, // FULLWIDTH ASTERISK
1549  {U'\uff0b', '+'}, // FULLWIDTH ASTERISK
1550  {U'\uff0c', ','}, // FULLWIDTH COMMA
1551  {U'\uff0d', '-'}, // FULLWIDTH HYPHEN-MINUS
1552  {U'\uff0e', '.'}, // FULLWIDTH FULL STOP
1553  {U'\uff0f', '/'}, // FULLWIDTH SOLIDUS
1554  {U'\uff1a', ':'}, // FULLWIDTH COLON
1555  {U'\uff1b', ';'}, // FULLWIDTH SEMICOLON
1556  {U'\uff1c', '<'}, // FULLWIDTH LESS-THAN SIGN
1557  {U'\uff1d', '='}, // FULLWIDTH EQUALS SIGN
1558  {U'\uff1e', '>'}, // FULLWIDTH GREATER-THAN SIGN
1559  {U'\uff1f', '?'}, // FULLWIDTH QUESTION MARK
1560  {U'\uff20', '@'}, // FULLWIDTH COMMERCIAL AT
1561  {U'\uff3b', '['}, // FULLWIDTH LEFT SQUARE BRACKET
1562  {U'\uff3c', '\\'}, // FULLWIDTH REVERSE SOLIDUS
1563  {U'\uff3d', ']'}, // FULLWIDTH RIGHT SQUARE BRACKET
1564  {U'\uff3e', '^'}, // FULLWIDTH CIRCUMFLEX ACCENT
1565  {U'\uff5b', '{'}, // FULLWIDTH LEFT CURLY BRACKET
1566  {U'\uff5c', '|'}, // FULLWIDTH VERTICAL LINE
1567  {U'\uff5d', '}'}, // FULLWIDTH RIGHT CURLY BRACKET
1568  {U'\uff5e', '~'}, // FULLWIDTH TILDE
1569  {0, 0}
1570  };
1571  auto Homoglyph =
1572  std::lower_bound(std::begin(SortedHomoglyphs),
1573  std::end(SortedHomoglyphs) - 1, HomoglyphPair{C, '\0'});
1574  if (Homoglyph->Character == C) {
1575  llvm::SmallString<5> CharBuf;
1576  {
1577  llvm::raw_svector_ostream CharOS(CharBuf);
1578  llvm::write_hex(CharOS, C, llvm::HexPrintStyle::Upper, 4);
1579  }
1580  if (Homoglyph->LooksLike) {
1581  const char LooksLikeStr[] = {Homoglyph->LooksLike, 0};
1582  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_homoglyph)
1583  << Range << CharBuf << LooksLikeStr;
1584  } else {
1585  Diags.Report(Range.getBegin(), diag::warn_utf8_symbol_zero_width)
1586  << Range << CharBuf;
1587  }
1588  }
1589 }
1590 
1591 bool Lexer::tryConsumeIdentifierUCN(const char *&CurPtr, unsigned Size,
1592  Token &Result) {
1593  const char *UCNPtr = CurPtr + Size;
1594  uint32_t CodePoint = tryReadUCN(UCNPtr, CurPtr, /*Token=*/nullptr);
1595  if (CodePoint == 0 || !isAllowedIDChar(CodePoint, LangOpts))
1596  return false;
1597 
1598  if (!isLexingRawMode())
1599  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1600  makeCharRange(*this, CurPtr, UCNPtr),
1601  /*IsFirst=*/false);
1602 
1603  Result.setFlag(Token::HasUCN);
1604  if ((UCNPtr - CurPtr == 6 && CurPtr[1] == 'u') ||
1605  (UCNPtr - CurPtr == 10 && CurPtr[1] == 'U'))
1606  CurPtr = UCNPtr;
1607  else
1608  while (CurPtr != UCNPtr)
1609  (void)getAndAdvanceChar(CurPtr, Result);
1610  return true;
1611 }
1612 
1613 bool Lexer::tryConsumeIdentifierUTF8Char(const char *&CurPtr) {
1614  const char *UnicodePtr = CurPtr;
1615  llvm::UTF32 CodePoint;
1616  llvm::ConversionResult Result =
1617  llvm::convertUTF8Sequence((const llvm::UTF8 **)&UnicodePtr,
1618  (const llvm::UTF8 *)BufferEnd,
1619  &CodePoint,
1620  llvm::strictConversion);
1621  if (Result != llvm::conversionOK ||
1622  !isAllowedIDChar(static_cast<uint32_t>(CodePoint), LangOpts))
1623  return false;
1624 
1625  if (!isLexingRawMode()) {
1626  maybeDiagnoseIDCharCompat(PP->getDiagnostics(), CodePoint,
1627  makeCharRange(*this, CurPtr, UnicodePtr),
1628  /*IsFirst=*/false);
1629  maybeDiagnoseUTF8Homoglyph(PP->getDiagnostics(), CodePoint,
1630  makeCharRange(*this, CurPtr, UnicodePtr));
1631  }
1632 
1633  CurPtr = UnicodePtr;
1634  return true;
1635 }
1636 
1637 bool Lexer::LexIdentifier(Token &Result, const char *CurPtr) {
1638  // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$]
1639  unsigned Size;
1640  unsigned char C = *CurPtr++;
1641  while (isIdentifierBody(C))
1642  C = *CurPtr++;
1643 
1644  --CurPtr; // Back up over the skipped character.
1645 
1646  // Fast path, no $,\,? in identifier found. '\' might be an escaped newline
1647  // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN.
1648  //
1649  // TODO: Could merge these checks into an InfoTable flag to make the
1650  // comparison cheaper
1651  if (isASCII(C) && C != '\\' && C != '?' &&
1652  (C != '$' || !LangOpts.DollarIdents)) {
1653 FinishIdentifier:
1654  const char *IdStart = BufferPtr;
1655  FormTokenWithChars(Result, CurPtr, tok::raw_identifier);
1656  Result.setRawIdentifierData(IdStart);
1657 
1658  // If we are in raw mode, return this identifier raw. There is no need to
1659  // look up identifier information or attempt to macro expand it.
1660  if (LexingRawMode)
1661  return true;
1662 
1663  // Fill in Result.IdentifierInfo and update the token kind,
1664  // looking up the identifier in the identifier table.
1665  IdentifierInfo *II = PP->LookUpIdentifierInfo(Result);
1666  // Note that we have to call PP->LookUpIdentifierInfo() even for code
1667  // completion, it writes IdentifierInfo into Result, and callers rely on it.
1668 
1669  // If the completion point is at the end of an identifier, we want to treat
1670  // the identifier as incomplete even if it resolves to a macro or a keyword.
1671  // This allows e.g. 'class^' to complete to 'classifier'.
1672  if (isCodeCompletionPoint(CurPtr)) {
1673  // Return the code-completion token.
1674  Result.setKind(tok::code_completion);
1675  // Skip the code-completion char and all immediate identifier characters.
1676  // This ensures we get consistent behavior when completing at any point in
1677  // an identifier (i.e. at the start, in the middle, at the end). Note that
1678  // only simple cases (i.e. [a-zA-Z0-9_]) are supported to keep the code
1679  // simpler.
1680  assert(*CurPtr == 0 && "Completion character must be 0");
1681  ++CurPtr;
1682  // Note that code completion token is not added as a separate character
1683  // when the completion point is at the end of the buffer. Therefore, we need
1684  // to check if the buffer has ended.
1685  if (CurPtr < BufferEnd) {
1686  while (isIdentifierBody(*CurPtr))
1687  ++CurPtr;
1688  }
1689  BufferPtr = CurPtr;
1690  return true;
1691  }
1692 
1693  // Finally, now that we know we have an identifier, pass this off to the
1694  // preprocessor, which may macro expand it or something.
1695  if (II->isHandleIdentifierCase())
1696  return PP->HandleIdentifier(Result);
1697 
1698  return true;
1699  }
1700 
1701  // Otherwise, $,\,? in identifier found. Enter slower path.
1702 
1703  C = getCharAndSize(CurPtr, Size);
1704  while (true) {
1705  if (C == '$') {
1706  // If we hit a $ and they are not supported in identifiers, we are done.
1707  if (!LangOpts.DollarIdents) goto FinishIdentifier;
1708 
1709  // Otherwise, emit a diagnostic and continue.
1710  if (!isLexingRawMode())
1711  Diag(CurPtr, diag::ext_dollar_in_identifier);
1712  CurPtr = ConsumeChar(CurPtr, Size, Result);
1713  C = getCharAndSize(CurPtr, Size);
1714  continue;
1715  } else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {
1716  C = getCharAndSize(CurPtr, Size);
1717  continue;
1718  } else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {
1719  C = getCharAndSize(CurPtr, Size);
1720  continue;
1721  } else if (!isIdentifierBody(C)) {
1722  goto FinishIdentifier;
1723  }
1724 
1725  // Otherwise, this character is good, consume it.
1726  CurPtr = ConsumeChar(CurPtr, Size, Result);
1727 
1728  C = getCharAndSize(CurPtr, Size);
1729  while (isIdentifierBody(C)) {
1730  CurPtr = ConsumeChar(CurPtr, Size, Result);
1731  C = getCharAndSize(CurPtr, Size);
1732  }
1733  }
1734 }
1735 
1736 /// isHexaLiteral - Return true if Start points to a hex constant.
1737 /// in microsoft mode (where this is supposed to be several different tokens).
1738 bool Lexer::isHexaLiteral(const char *Start, const LangOptions &LangOpts) {
1739  unsigned Size;
1740  char C1 = Lexer::getCharAndSizeNoWarn(Start, Size, LangOpts);
1741  if (C1 != '0')
1742  return false;
1743  char C2 = Lexer::getCharAndSizeNoWarn(Start + Size, Size, LangOpts);
1744  return (C2 == 'x' || C2 == 'X');
1745 }
1746 
1747 /// LexNumericConstant - Lex the remainder of a integer or floating point
1748 /// constant. From[-1] is the first character lexed. Return the end of the
1749 /// constant.
1750 bool Lexer::LexNumericConstant(Token &Result, const char *CurPtr) {
1751  unsigned Size;
1752  char C = getCharAndSize(CurPtr, Size);
1753  char PrevCh = 0;
1754  while (isPreprocessingNumberBody(C)) {
1755  CurPtr = ConsumeChar(CurPtr, Size, Result);
1756  PrevCh = C;
1757  C = getCharAndSize(CurPtr, Size);
1758  }
1759 
1760  // If we fell out, check for a sign, due to 1e+12. If we have one, continue.
1761  if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) {
1762  // If we are in Microsoft mode, don't continue if the constant is hex.
1763  // For example, MSVC will accept the following as 3 tokens: 0x1234567e+1
1764  if (!LangOpts.MicrosoftExt || !isHexaLiteral(BufferPtr, LangOpts))
1765  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1766  }
1767 
1768  // If we have a hex FP constant, continue.
1769  if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p')) {
1770  // Outside C99 and C++17, we accept hexadecimal floating point numbers as a
1771  // not-quite-conforming extension. Only do so if this looks like it's
1772  // actually meant to be a hexfloat, and not if it has a ud-suffix.
1773  bool IsHexFloat = true;
1774  if (!LangOpts.C99) {
1775  if (!isHexaLiteral(BufferPtr, LangOpts))
1776  IsHexFloat = false;
1777  else if (!getLangOpts().CPlusPlus17 &&
1778  std::find(BufferPtr, CurPtr, '_') != CurPtr)
1779  IsHexFloat = false;
1780  }
1781  if (IsHexFloat)
1782  return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result));
1783  }
1784 
1785  // If we have a digit separator, continue.
1786  if (C == '\'' && getLangOpts().CPlusPlus14) {
1787  unsigned NextSize;
1788  char Next = getCharAndSizeNoWarn(CurPtr + Size, NextSize, getLangOpts());
1789  if (isIdentifierBody(Next)) {
1790  if (!isLexingRawMode())
1791  Diag(CurPtr, diag::warn_cxx11_compat_digit_separator);
1792  CurPtr = ConsumeChar(CurPtr, Size, Result);
1793  CurPtr = ConsumeChar(CurPtr, NextSize, Result);
1794  return LexNumericConstant(Result, CurPtr);
1795  }
1796  }
1797 
1798  // If we have a UCN or UTF-8 character (perhaps in a ud-suffix), continue.
1799  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1800  return LexNumericConstant(Result, CurPtr);
1801  if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1802  return LexNumericConstant(Result, CurPtr);
1803 
1804  // Update the location of token as well as BufferPtr.
1805  const char *TokStart = BufferPtr;
1806  FormTokenWithChars(Result, CurPtr, tok::numeric_constant);
1807  Result.setLiteralData(TokStart);
1808  return true;
1809 }
1810 
1811 /// LexUDSuffix - Lex the ud-suffix production for user-defined literal suffixes
1812 /// in C++11, or warn on a ud-suffix in C++98.
1813 const char *Lexer::LexUDSuffix(Token &Result, const char *CurPtr,
1814  bool IsStringLiteral) {
1815  assert(getLangOpts().CPlusPlus);
1816 
1817  // Maximally munch an identifier.
1818  unsigned Size;
1819  char C = getCharAndSize(CurPtr, Size);
1820  bool Consumed = false;
1821 
1822  if (!isIdentifierHead(C)) {
1823  if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result))
1824  Consumed = true;
1825  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr))
1826  Consumed = true;
1827  else
1828  return CurPtr;
1829  }
1830 
1831  if (!getLangOpts().CPlusPlus11) {
1832  if (!isLexingRawMode())
1833  Diag(CurPtr,
1834  C == '_' ? diag::warn_cxx11_compat_user_defined_literal
1835  : diag::warn_cxx11_compat_reserved_user_defined_literal)
1837  return CurPtr;
1838  }
1839 
1840  // C++11 [lex.ext]p10, [usrlit.suffix]p1: A program containing a ud-suffix
1841  // that does not start with an underscore is ill-formed. As a conforming
1842  // extension, we treat all such suffixes as if they had whitespace before
1843  // them. We assume a suffix beginning with a UCN or UTF-8 character is more
1844  // likely to be a ud-suffix than a macro, however, and accept that.
1845  if (!Consumed) {
1846  bool IsUDSuffix = false;
1847  if (C == '_')
1848  IsUDSuffix = true;
1849  else if (IsStringLiteral && getLangOpts().CPlusPlus14) {
1850  // In C++1y, we need to look ahead a few characters to see if this is a
1851  // valid suffix for a string literal or a numeric literal (this could be
1852  // the 'operator""if' defining a numeric literal operator).
1853  const unsigned MaxStandardSuffixLength = 3;
1854  char Buffer[MaxStandardSuffixLength] = { C };
1855  unsigned Consumed = Size;
1856  unsigned Chars = 1;
1857  while (true) {
1858  unsigned NextSize;
1859  char Next = getCharAndSizeNoWarn(CurPtr + Consumed, NextSize,
1860  getLangOpts());
1861  if (!isIdentifierBody(Next)) {
1862  // End of suffix. Check whether this is on the whitelist.
1863  const StringRef CompleteSuffix(Buffer, Chars);
1865  CompleteSuffix);
1866  break;
1867  }
1868 
1869  if (Chars == MaxStandardSuffixLength)
1870  // Too long: can't be a standard suffix.
1871  break;
1872 
1873  Buffer[Chars++] = Next;
1874  Consumed += NextSize;
1875  }
1876  }
1877 
1878  if (!IsUDSuffix) {
1879  if (!isLexingRawMode())
1880  Diag(CurPtr, getLangOpts().MSVCCompat
1881  ? diag::ext_ms_reserved_user_defined_literal
1882  : diag::ext_reserved_user_defined_literal)
1884  return CurPtr;
1885  }
1886 
1887  CurPtr = ConsumeChar(CurPtr, Size, Result);
1888  }
1889 
1890  Result.setFlag(Token::HasUDSuffix);
1891  while (true) {
1892  C = getCharAndSize(CurPtr, Size);
1893  if (isIdentifierBody(C)) { CurPtr = ConsumeChar(CurPtr, Size, Result); }
1894  else if (C == '\\' && tryConsumeIdentifierUCN(CurPtr, Size, Result)) {}
1895  else if (!isASCII(C) && tryConsumeIdentifierUTF8Char(CurPtr)) {}
1896  else break;
1897  }
1898 
1899  return CurPtr;
1900 }
1901 
1902 /// LexStringLiteral - Lex the remainder of a string literal, after having lexed
1903 /// either " or L" or u8" or u" or U".
1904 bool Lexer::LexStringLiteral(Token &Result, const char *CurPtr,
1905  tok::TokenKind Kind) {
1906  const char *AfterQuote = CurPtr;
1907  // Does this string contain the \0 character?
1908  const char *NulCharacter = nullptr;
1909 
1910  if (!isLexingRawMode() &&
1911  (Kind == tok::utf8_string_literal ||
1912  Kind == tok::utf16_string_literal ||
1913  Kind == tok::utf32_string_literal))
1914  Diag(BufferPtr, getLangOpts().CPlusPlus
1915  ? diag::warn_cxx98_compat_unicode_literal
1916  : diag::warn_c99_compat_unicode_literal);
1917 
1918  char C = getAndAdvanceChar(CurPtr, Result);
1919  while (C != '"') {
1920  // Skip escaped characters. Escaped newlines will already be processed by
1921  // getAndAdvanceChar.
1922  if (C == '\\')
1923  C = getAndAdvanceChar(CurPtr, Result);
1924 
1925  if (C == '\n' || C == '\r' || // Newline.
1926  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
1927  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
1928  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 1;
1929  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
1930  return true;
1931  }
1932 
1933  if (C == 0) {
1934  if (isCodeCompletionPoint(CurPtr-1)) {
1935  if (ParsingFilename)
1936  codeCompleteIncludedFile(AfterQuote, CurPtr - 1, /*IsAngled=*/false);
1937  else
1939  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
1940  cutOffLexing();
1941  return true;
1942  }
1943 
1944  NulCharacter = CurPtr-1;
1945  }
1946  C = getAndAdvanceChar(CurPtr, Result);
1947  }
1948 
1949  // If we are in C++11, lex the optional ud-suffix.
1950  if (getLangOpts().CPlusPlus)
1951  CurPtr = LexUDSuffix(Result, CurPtr, true);
1952 
1953  // If a nul character existed in the string, warn about it.
1954  if (NulCharacter && !isLexingRawMode())
1955  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
1956 
1957  // Update the location of the token as well as the BufferPtr instance var.
1958  const char *TokStart = BufferPtr;
1959  FormTokenWithChars(Result, CurPtr, Kind);
1960  Result.setLiteralData(TokStart);
1961  return true;
1962 }
1963 
1964 /// LexRawStringLiteral - Lex the remainder of a raw string literal, after
1965 /// having lexed R", LR", u8R", uR", or UR".
1966 bool Lexer::LexRawStringLiteral(Token &Result, const char *CurPtr,
1967  tok::TokenKind Kind) {
1968  // This function doesn't use getAndAdvanceChar because C++0x [lex.pptoken]p3:
1969  // Between the initial and final double quote characters of the raw string,
1970  // any transformations performed in phases 1 and 2 (trigraphs,
1971  // universal-character-names, and line splicing) are reverted.
1972 
1973  if (!isLexingRawMode())
1974  Diag(BufferPtr, diag::warn_cxx98_compat_raw_string_literal);
1975 
1976  unsigned PrefixLen = 0;
1977 
1978  while (PrefixLen != 16 && isRawStringDelimBody(CurPtr[PrefixLen]))
1979  ++PrefixLen;
1980 
1981  // If the last character was not a '(', then we didn't lex a valid delimiter.
1982  if (CurPtr[PrefixLen] != '(') {
1983  if (!isLexingRawMode()) {
1984  const char *PrefixEnd = &CurPtr[PrefixLen];
1985  if (PrefixLen == 16) {
1986  Diag(PrefixEnd, diag::err_raw_delim_too_long);
1987  } else {
1988  Diag(PrefixEnd, diag::err_invalid_char_raw_delim)
1989  << StringRef(PrefixEnd, 1);
1990  }
1991  }
1992 
1993  // Search for the next '"' in hopes of salvaging the lexer. Unfortunately,
1994  // it's possible the '"' was intended to be part of the raw string, but
1995  // there's not much we can do about that.
1996  while (true) {
1997  char C = *CurPtr++;
1998 
1999  if (C == '"')
2000  break;
2001  if (C == 0 && CurPtr-1 == BufferEnd) {
2002  --CurPtr;
2003  break;
2004  }
2005  }
2006 
2007  FormTokenWithChars(Result, CurPtr, tok::unknown);
2008  return true;
2009  }
2010 
2011  // Save prefix and move CurPtr past it
2012  const char *Prefix = CurPtr;
2013  CurPtr += PrefixLen + 1; // skip over prefix and '('
2014 
2015  while (true) {
2016  char C = *CurPtr++;
2017 
2018  if (C == ')') {
2019  // Check for prefix match and closing quote.
2020  if (strncmp(CurPtr, Prefix, PrefixLen) == 0 && CurPtr[PrefixLen] == '"') {
2021  CurPtr += PrefixLen + 1; // skip over prefix and '"'
2022  break;
2023  }
2024  } else if (C == 0 && CurPtr-1 == BufferEnd) { // End of file.
2025  if (!isLexingRawMode())
2026  Diag(BufferPtr, diag::err_unterminated_raw_string)
2027  << StringRef(Prefix, PrefixLen);
2028  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2029  return true;
2030  }
2031  }
2032 
2033  // If we are in C++11, lex the optional ud-suffix.
2034  if (getLangOpts().CPlusPlus)
2035  CurPtr = LexUDSuffix(Result, CurPtr, true);
2036 
2037  // Update the location of token as well as BufferPtr.
2038  const char *TokStart = BufferPtr;
2039  FormTokenWithChars(Result, CurPtr, Kind);
2040  Result.setLiteralData(TokStart);
2041  return true;
2042 }
2043 
2044 /// LexAngledStringLiteral - Lex the remainder of an angled string literal,
2045 /// after having lexed the '<' character. This is used for #include filenames.
2046 bool Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) {
2047  // Does this string contain the \0 character?
2048  const char *NulCharacter = nullptr;
2049  const char *AfterLessPos = CurPtr;
2050  char C = getAndAdvanceChar(CurPtr, Result);
2051  while (C != '>') {
2052  // Skip escaped characters. Escaped newlines will already be processed by
2053  // getAndAdvanceChar.
2054  if (C == '\\')
2055  C = getAndAdvanceChar(CurPtr, Result);
2056 
2057  if (C == '\n' || C == '\r' || // Newline.
2058  (C == 0 && (CurPtr - 1 == BufferEnd))) { // End of file.
2059  // If the filename is unterminated, then it must just be a lone <
2060  // character. Return this as such.
2061  FormTokenWithChars(Result, AfterLessPos, tok::less);
2062  return true;
2063  }
2064 
2065  if (C == 0) {
2066  if (isCodeCompletionPoint(CurPtr - 1)) {
2067  codeCompleteIncludedFile(AfterLessPos, CurPtr - 1, /*IsAngled=*/true);
2068  cutOffLexing();
2069  FormTokenWithChars(Result, CurPtr - 1, tok::unknown);
2070  return true;
2071  }
2072  NulCharacter = CurPtr-1;
2073  }
2074  C = getAndAdvanceChar(CurPtr, Result);
2075  }
2076 
2077  // If a nul character existed in the string, warn about it.
2078  if (NulCharacter && !isLexingRawMode())
2079  Diag(NulCharacter, diag::null_in_char_or_string) << 1;
2080 
2081  // Update the location of token as well as BufferPtr.
2082  const char *TokStart = BufferPtr;
2083  FormTokenWithChars(Result, CurPtr, tok::header_name);
2084  Result.setLiteralData(TokStart);
2085  return true;
2086 }
2087 
2088 void Lexer::codeCompleteIncludedFile(const char *PathStart,
2089  const char *CompletionPoint,
2090  bool IsAngled) {
2091  // Completion only applies to the filename, after the last slash.
2092  StringRef PartialPath(PathStart, CompletionPoint - PathStart);
2093  auto Slash = PartialPath.find_last_of(LangOpts.MSVCCompat ? "/\\" : "/");
2094  StringRef Dir =
2095  (Slash == StringRef::npos) ? "" : PartialPath.take_front(Slash);
2096  const char *StartOfFilename =
2097  (Slash == StringRef::npos) ? PathStart : PathStart + Slash + 1;
2098  // Code completion filter range is the filename only, up to completion point.
2100  StringRef(StartOfFilename, CompletionPoint - StartOfFilename)));
2101  // We should replace the characters up to the closing quote, if any.
2102  while (CompletionPoint < BufferEnd) {
2103  char Next = *(CompletionPoint + 1);
2104  if (Next == 0 || Next == '\r' || Next == '\n')
2105  break;
2106  ++CompletionPoint;
2107  if (Next == (IsAngled ? '>' : '"'))
2108  break;
2109  }
2111  FileLoc.getLocWithOffset(StartOfFilename - BufferStart),
2112  FileLoc.getLocWithOffset(CompletionPoint - BufferStart));
2113  PP->CodeCompleteIncludedFile(Dir, IsAngled);
2114 }
2115 
2116 /// LexCharConstant - Lex the remainder of a character constant, after having
2117 /// lexed either ' or L' or u8' or u' or U'.
2118 bool Lexer::LexCharConstant(Token &Result, const char *CurPtr,
2119  tok::TokenKind Kind) {
2120  // Does this character contain the \0 character?
2121  const char *NulCharacter = nullptr;
2122 
2123  if (!isLexingRawMode()) {
2124  if (Kind == tok::utf16_char_constant || Kind == tok::utf32_char_constant)
2125  Diag(BufferPtr, getLangOpts().CPlusPlus
2126  ? diag::warn_cxx98_compat_unicode_literal
2127  : diag::warn_c99_compat_unicode_literal);
2128  else if (Kind == tok::utf8_char_constant)
2129  Diag(BufferPtr, diag::warn_cxx14_compat_u8_character_literal);
2130  }
2131 
2132  char C = getAndAdvanceChar(CurPtr, Result);
2133  if (C == '\'') {
2134  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2135  Diag(BufferPtr, diag::ext_empty_character);
2136  FormTokenWithChars(Result, CurPtr, tok::unknown);
2137  return true;
2138  }
2139 
2140  while (C != '\'') {
2141  // Skip escaped characters.
2142  if (C == '\\')
2143  C = getAndAdvanceChar(CurPtr, Result);
2144 
2145  if (C == '\n' || C == '\r' || // Newline.
2146  (C == 0 && CurPtr-1 == BufferEnd)) { // End of file.
2147  if (!isLexingRawMode() && !LangOpts.AsmPreprocessor)
2148  Diag(BufferPtr, diag::ext_unterminated_char_or_string) << 0;
2149  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2150  return true;
2151  }
2152 
2153  if (C == 0) {
2154  if (isCodeCompletionPoint(CurPtr-1)) {
2156  FormTokenWithChars(Result, CurPtr-1, tok::unknown);
2157  cutOffLexing();
2158  return true;
2159  }
2160 
2161  NulCharacter = CurPtr-1;
2162  }
2163  C = getAndAdvanceChar(CurPtr, Result);
2164  }
2165 
2166  // If we are in C++11, lex the optional ud-suffix.
2167  if (getLangOpts().CPlusPlus)
2168  CurPtr = LexUDSuffix(Result, CurPtr, false);
2169 
2170  // If a nul character existed in the character, warn about it.
2171  if (NulCharacter && !isLexingRawMode())
2172  Diag(NulCharacter, diag::null_in_char_or_string) << 0;
2173 
2174  // Update the location of token as well as BufferPtr.
2175  const char *TokStart = BufferPtr;
2176  FormTokenWithChars(Result, CurPtr, Kind);
2177  Result.setLiteralData(TokStart);
2178  return true;
2179 }
2180 
2181 /// SkipWhitespace - Efficiently skip over a series of whitespace characters.
2182 /// Update BufferPtr to point to the next non-whitespace character and return.
2183 ///
2184 /// This method forms a token and returns true if KeepWhitespaceMode is enabled.
2185 bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr,
2186  bool &TokAtPhysicalStartOfLine) {
2187  // Whitespace - Skip it, then return the token after the whitespace.
2188  bool SawNewline = isVerticalWhitespace(CurPtr[-1]);
2189 
2190  unsigned char Char = *CurPtr;
2191 
2192  // Skip consecutive spaces efficiently.
2193  while (true) {
2194  // Skip horizontal whitespace very aggressively.
2195  while (isHorizontalWhitespace(Char))
2196  Char = *++CurPtr;
2197 
2198  // Otherwise if we have something other than whitespace, we're done.
2199  if (!isVerticalWhitespace(Char))
2200  break;
2201 
2203  // End of preprocessor directive line, let LexTokenInternal handle this.
2204  BufferPtr = CurPtr;
2205  return false;
2206  }
2207 
2208  // OK, but handle newline.
2209  SawNewline = true;
2210  Char = *++CurPtr;
2211  }
2212 
2213  // If the client wants us to return whitespace, return it now.
2214  if (isKeepWhitespaceMode()) {
2215  FormTokenWithChars(Result, CurPtr, tok::unknown);
2216  if (SawNewline) {
2217  IsAtStartOfLine = true;
2218  IsAtPhysicalStartOfLine = true;
2219  }
2220  // FIXME: The next token will not have LeadingSpace set.
2221  return true;
2222  }
2223 
2224  // If this isn't immediately after a newline, there is leading space.
2225  char PrevChar = CurPtr[-1];
2226  bool HasLeadingSpace = !isVerticalWhitespace(PrevChar);
2227 
2228  Result.setFlagValue(Token::LeadingSpace, HasLeadingSpace);
2229  if (SawNewline) {
2230  Result.setFlag(Token::StartOfLine);
2231  TokAtPhysicalStartOfLine = true;
2232  }
2233 
2234  BufferPtr = CurPtr;
2235  return false;
2236 }
2237 
2238 /// We have just read the // characters from input. Skip until we find the
2239 /// newline character that terminates the comment. Then update BufferPtr and
2240 /// return.
2241 ///
2242 /// If we're in KeepCommentMode or any CommentHandler has inserted
2243 /// some tokens, this will store the first token and return true.
2244 bool Lexer::SkipLineComment(Token &Result, const char *CurPtr,
2245  bool &TokAtPhysicalStartOfLine) {
2246  // If Line comments aren't explicitly enabled for this language, emit an
2247  // extension warning.
2248  if (!LangOpts.LineComment && !isLexingRawMode()) {
2249  Diag(BufferPtr, diag::ext_line_comment);
2250 
2251  // Mark them enabled so we only emit one warning for this translation
2252  // unit.
2253  LangOpts.LineComment = true;
2254  }
2255 
2256  // Scan over the body of the comment. The common case, when scanning, is that
2257  // the comment contains normal ascii characters with nothing interesting in
2258  // them. As such, optimize for this case with the inner loop.
2259  //
2260  // This loop terminates with CurPtr pointing at the newline (or end of buffer)
2261  // character that ends the line comment.
2262  char C;
2263  while (true) {
2264  C = *CurPtr;
2265  // Skip over characters in the fast loop.
2266  while (C != 0 && // Potentially EOF.
2267  C != '\n' && C != '\r') // Newline or DOS-style newline.
2268  C = *++CurPtr;
2269 
2270  const char *NextLine = CurPtr;
2271  if (C != 0) {
2272  // We found a newline, see if it's escaped.
2273  const char *EscapePtr = CurPtr-1;
2274  bool HasSpace = false;
2275  while (isHorizontalWhitespace(*EscapePtr)) { // Skip whitespace.
2276  --EscapePtr;
2277  HasSpace = true;
2278  }
2279 
2280  if (*EscapePtr == '\\')
2281  // Escaped newline.
2282  CurPtr = EscapePtr;
2283  else if (EscapePtr[0] == '/' && EscapePtr[-1] == '?' &&
2284  EscapePtr[-2] == '?' && LangOpts.Trigraphs)
2285  // Trigraph-escaped newline.
2286  CurPtr = EscapePtr-2;
2287  else
2288  break; // This is a newline, we're done.
2289 
2290  // If there was space between the backslash and newline, warn about it.
2291  if (HasSpace && !isLexingRawMode())
2292  Diag(EscapePtr, diag::backslash_newline_space);
2293  }
2294 
2295  // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to
2296  // properly decode the character. Read it in raw mode to avoid emitting
2297  // diagnostics about things like trigraphs. If we see an escaped newline,
2298  // we'll handle it below.
2299  const char *OldPtr = CurPtr;
2300  bool OldRawMode = isLexingRawMode();
2301  LexingRawMode = true;
2302  C = getAndAdvanceChar(CurPtr, Result);
2303  LexingRawMode = OldRawMode;
2304 
2305  // If we only read only one character, then no special handling is needed.
2306  // We're done and can skip forward to the newline.
2307  if (C != 0 && CurPtr == OldPtr+1) {
2308  CurPtr = NextLine;
2309  break;
2310  }
2311 
2312  // If we read multiple characters, and one of those characters was a \r or
2313  // \n, then we had an escaped newline within the comment. Emit diagnostic
2314  // unless the next line is also a // comment.
2315  if (CurPtr != OldPtr + 1 && C != '/' &&
2316  (CurPtr == BufferEnd + 1 || CurPtr[0] != '/')) {
2317  for (; OldPtr != CurPtr; ++OldPtr)
2318  if (OldPtr[0] == '\n' || OldPtr[0] == '\r') {
2319  // Okay, we found a // comment that ends in a newline, if the next
2320  // line is also a // comment, but has spaces, don't emit a diagnostic.
2321  if (isWhitespace(C)) {
2322  const char *ForwardPtr = CurPtr;
2323  while (isWhitespace(*ForwardPtr)) // Skip whitespace.
2324  ++ForwardPtr;
2325  if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/')
2326  break;
2327  }
2328 
2329  if (!isLexingRawMode())
2330  Diag(OldPtr-1, diag::ext_multi_line_line_comment);
2331  break;
2332  }
2333  }
2334 
2335  if (C == '\r' || C == '\n' || CurPtr == BufferEnd + 1) {
2336  --CurPtr;
2337  break;
2338  }
2339 
2340  if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2342  cutOffLexing();
2343  return false;
2344  }
2345  }
2346 
2347  // Found but did not consume the newline. Notify comment handlers about the
2348  // comment unless we're in a #if 0 block.
2349  if (PP && !isLexingRawMode() &&
2350  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2351  getSourceLocation(CurPtr)))) {
2352  BufferPtr = CurPtr;
2353  return true; // A token has to be returned.
2354  }
2355 
2356  // If we are returning comments as tokens, return this comment as a token.
2357  if (inKeepCommentMode())
2358  return SaveLineComment(Result, CurPtr);
2359 
2360  // If we are inside a preprocessor directive and we see the end of line,
2361  // return immediately, so that the lexer can return this as an EOD token.
2362  if (ParsingPreprocessorDirective || CurPtr == BufferEnd) {
2363  BufferPtr = CurPtr;
2364  return false;
2365  }
2366 
2367  // Otherwise, eat the \n character. We don't care if this is a \n\r or
2368  // \r\n sequence. This is an efficiency hack (because we know the \n can't
2369  // contribute to another token), it isn't needed for correctness. Note that
2370  // this is ok even in KeepWhitespaceMode, because we would have returned the
2371  /// comment above in that mode.
2372  ++CurPtr;
2373 
2374  // The next returned token is at the start of the line.
2375  Result.setFlag(Token::StartOfLine);
2376  TokAtPhysicalStartOfLine = true;
2377  // No leading whitespace seen so far.
2379  BufferPtr = CurPtr;
2380  return false;
2381 }
2382 
2383 /// If in save-comment mode, package up this Line comment in an appropriate
2384 /// way and return it.
2385 bool Lexer::SaveLineComment(Token &Result, const char *CurPtr) {
2386  // If we're not in a preprocessor directive, just return the // comment
2387  // directly.
2388  FormTokenWithChars(Result, CurPtr, tok::comment);
2389 
2391  return true;
2392 
2393  // If this Line-style comment is in a macro definition, transmogrify it into
2394  // a C-style block comment.
2395  bool Invalid = false;
2396  std::string Spelling = PP->getSpelling(Result, &Invalid);
2397  if (Invalid)
2398  return true;
2399 
2400  assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not line comment?");
2401  Spelling[1] = '*'; // Change prefix to "/*".
2402  Spelling += "*/"; // add suffix.
2403 
2404  Result.setKind(tok::comment);
2405  PP->CreateString(Spelling, Result,
2406  Result.getLocation(), Result.getLocation());
2407  return true;
2408 }
2409 
2410 /// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline
2411 /// character (either \\n or \\r) is part of an escaped newline sequence. Issue
2412 /// a diagnostic if so. We know that the newline is inside of a block comment.
2413 static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr,
2414  Lexer *L) {
2415  assert(CurPtr[0] == '\n' || CurPtr[0] == '\r');
2416 
2417  // Back up off the newline.
2418  --CurPtr;
2419 
2420  // If this is a two-character newline sequence, skip the other character.
2421  if (CurPtr[0] == '\n' || CurPtr[0] == '\r') {
2422  // \n\n or \r\r -> not escaped newline.
2423  if (CurPtr[0] == CurPtr[1])
2424  return false;
2425  // \n\r or \r\n -> skip the newline.
2426  --CurPtr;
2427  }
2428 
2429  // If we have horizontal whitespace, skip over it. We allow whitespace
2430  // between the slash and newline.
2431  bool HasSpace = false;
2432  while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) {
2433  --CurPtr;
2434  HasSpace = true;
2435  }
2436 
2437  // If we have a slash, we know this is an escaped newline.
2438  if (*CurPtr == '\\') {
2439  if (CurPtr[-1] != '*') return false;
2440  } else {
2441  // It isn't a slash, is it the ?? / trigraph?
2442  if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' ||
2443  CurPtr[-3] != '*')
2444  return false;
2445 
2446  // This is the trigraph ending the comment. Emit a stern warning!
2447  CurPtr -= 2;
2448 
2449  // If no trigraphs are enabled, warn that we ignored this trigraph and
2450  // ignore this * character.
2451  if (!L->getLangOpts().Trigraphs) {
2452  if (!L->isLexingRawMode())
2453  L->Diag(CurPtr, diag::trigraph_ignored_block_comment);
2454  return false;
2455  }
2456  if (!L->isLexingRawMode())
2457  L->Diag(CurPtr, diag::trigraph_ends_block_comment);
2458  }
2459 
2460  // Warn about having an escaped newline between the */ characters.
2461  if (!L->isLexingRawMode())
2462  L->Diag(CurPtr, diag::escaped_newline_block_comment_end);
2463 
2464  // If there was space between the backslash and newline, warn about it.
2465  if (HasSpace && !L->isLexingRawMode())
2466  L->Diag(CurPtr, diag::backslash_newline_space);
2467 
2468  return true;
2469 }
2470 
2471 #ifdef __SSE2__
2472 #include <emmintrin.h>
2473 #elif __ALTIVEC__
2474 #include <altivec.h>
2475 #undef bool
2476 #endif
2477 
2478 /// We have just read from input the / and * characters that started a comment.
2479 /// Read until we find the * and / characters that terminate the comment.
2480 /// Note that we don't bother decoding trigraphs or escaped newlines in block
2481 /// comments, because they cannot cause the comment to end. The only thing
2482 /// that can happen is the comment could end with an escaped newline between
2483 /// the terminating * and /.
2484 ///
2485 /// If we're in KeepCommentMode or any CommentHandler has inserted
2486 /// some tokens, this will store the first token and return true.
2487 bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr,
2488  bool &TokAtPhysicalStartOfLine) {
2489  // Scan one character past where we should, looking for a '/' character. Once
2490  // we find it, check to see if it was preceded by a *. This common
2491  // optimization helps people who like to put a lot of * characters in their
2492  // comments.
2493 
2494  // The first character we get with newlines and trigraphs skipped to handle
2495  // the degenerate /*/ case below correctly if the * has an escaped newline
2496  // after it.
2497  unsigned CharSize;
2498  unsigned char C = getCharAndSize(CurPtr, CharSize);
2499  CurPtr += CharSize;
2500  if (C == 0 && CurPtr == BufferEnd+1) {
2501  if (!isLexingRawMode())
2502  Diag(BufferPtr, diag::err_unterminated_block_comment);
2503  --CurPtr;
2504 
2505  // KeepWhitespaceMode should return this broken comment as a token. Since
2506  // it isn't a well formed comment, just return it as an 'unknown' token.
2507  if (isKeepWhitespaceMode()) {
2508  FormTokenWithChars(Result, CurPtr, tok::unknown);
2509  return true;
2510  }
2511 
2512  BufferPtr = CurPtr;
2513  return false;
2514  }
2515 
2516  // Check to see if the first character after the '/*' is another /. If so,
2517  // then this slash does not end the block comment, it is part of it.
2518  if (C == '/')
2519  C = *CurPtr++;
2520 
2521  while (true) {
2522  // Skip over all non-interesting characters until we find end of buffer or a
2523  // (probably ending) '/' character.
2524  if (CurPtr + 24 < BufferEnd &&
2525  // If there is a code-completion point avoid the fast scan because it
2526  // doesn't check for '\0'.
2527  !(PP && PP->getCodeCompletionFileLoc() == FileLoc)) {
2528  // While not aligned to a 16-byte boundary.
2529  while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0)
2530  C = *CurPtr++;
2531 
2532  if (C == '/') goto FoundSlash;
2533 
2534 #ifdef __SSE2__
2535  __m128i Slashes = _mm_set1_epi8('/');
2536  while (CurPtr+16 <= BufferEnd) {
2537  int cmp = _mm_movemask_epi8(_mm_cmpeq_epi8(*(const __m128i*)CurPtr,
2538  Slashes));
2539  if (cmp != 0) {
2540  // Adjust the pointer to point directly after the first slash. It's
2541  // not necessary to set C here, it will be overwritten at the end of
2542  // the outer loop.
2543  CurPtr += llvm::countTrailingZeros<unsigned>(cmp) + 1;
2544  goto FoundSlash;
2545  }
2546  CurPtr += 16;
2547  }
2548 #elif __ALTIVEC__
2549  __vector unsigned char Slashes = {
2550  '/', '/', '/', '/', '/', '/', '/', '/',
2551  '/', '/', '/', '/', '/', '/', '/', '/'
2552  };
2553  while (CurPtr+16 <= BufferEnd &&
2554  !vec_any_eq(*(const vector unsigned char*)CurPtr, Slashes))
2555  CurPtr += 16;
2556 #else
2557  // Scan for '/' quickly. Many block comments are very large.
2558  while (CurPtr[0] != '/' &&
2559  CurPtr[1] != '/' &&
2560  CurPtr[2] != '/' &&
2561  CurPtr[3] != '/' &&
2562  CurPtr+4 < BufferEnd) {
2563  CurPtr += 4;
2564  }
2565 #endif
2566 
2567  // It has to be one of the bytes scanned, increment to it and read one.
2568  C = *CurPtr++;
2569  }
2570 
2571  // Loop to scan the remainder.
2572  while (C != '/' && C != '\0')
2573  C = *CurPtr++;
2574 
2575  if (C == '/') {
2576  FoundSlash:
2577  if (CurPtr[-2] == '*') // We found the final */. We're done!
2578  break;
2579 
2580  if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) {
2581  if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) {
2582  // We found the final */, though it had an escaped newline between the
2583  // * and /. We're done!
2584  break;
2585  }
2586  }
2587  if (CurPtr[0] == '*' && CurPtr[1] != '/') {
2588  // If this is a /* inside of the comment, emit a warning. Don't do this
2589  // if this is a /*/, which will end the comment. This misses cases with
2590  // embedded escaped newlines, but oh well.
2591  if (!isLexingRawMode())
2592  Diag(CurPtr-1, diag::warn_nested_block_comment);
2593  }
2594  } else if (C == 0 && CurPtr == BufferEnd+1) {
2595  if (!isLexingRawMode())
2596  Diag(BufferPtr, diag::err_unterminated_block_comment);
2597  // Note: the user probably forgot a */. We could continue immediately
2598  // after the /*, but this would involve lexing a lot of what really is the
2599  // comment, which surely would confuse the parser.
2600  --CurPtr;
2601 
2602  // KeepWhitespaceMode should return this broken comment as a token. Since
2603  // it isn't a well formed comment, just return it as an 'unknown' token.
2604  if (isKeepWhitespaceMode()) {
2605  FormTokenWithChars(Result, CurPtr, tok::unknown);
2606  return true;
2607  }
2608 
2609  BufferPtr = CurPtr;
2610  return false;
2611  } else if (C == '\0' && isCodeCompletionPoint(CurPtr-1)) {
2613  cutOffLexing();
2614  return false;
2615  }
2616 
2617  C = *CurPtr++;
2618  }
2619 
2620  // Notify comment handlers about the comment unless we're in a #if 0 block.
2621  if (PP && !isLexingRawMode() &&
2622  PP->HandleComment(Result, SourceRange(getSourceLocation(BufferPtr),
2623  getSourceLocation(CurPtr)))) {
2624  BufferPtr = CurPtr;
2625  return true; // A token has to be returned.
2626  }
2627 
2628  // If we are returning comments as tokens, return this comment as a token.
2629  if (inKeepCommentMode()) {
2630  FormTokenWithChars(Result, CurPtr, tok::comment);
2631  return true;
2632  }
2633 
2634  // It is common for the tokens immediately after a /**/ comment to be
2635  // whitespace. Instead of going through the big switch, handle it
2636  // efficiently now. This is safe even in KeepWhitespaceMode because we would
2637  // have already returned above with the comment as a token.
2638  if (isHorizontalWhitespace(*CurPtr)) {
2639  SkipWhitespace(Result, CurPtr+1, TokAtPhysicalStartOfLine);
2640  return false;
2641  }
2642 
2643  // Otherwise, just return so that the next character will be lexed as a token.
2644  BufferPtr = CurPtr;
2645  Result.setFlag(Token::LeadingSpace);
2646  return false;
2647 }
2648 
2649 //===----------------------------------------------------------------------===//
2650 // Primary Lexing Entry Points
2651 //===----------------------------------------------------------------------===//
2652 
2653 /// ReadToEndOfLine - Read the rest of the current preprocessor line as an
2654 /// uninterpreted string. This switches the lexer out of directive mode.
2656  assert(ParsingPreprocessorDirective && ParsingFilename == false &&
2657  "Must be in a preprocessing directive!");
2658  Token Tmp;
2659 
2660  // CurPtr - Cache BufferPtr in an automatic variable.
2661  const char *CurPtr = BufferPtr;
2662  while (true) {
2663  char Char = getAndAdvanceChar(CurPtr, Tmp);
2664  switch (Char) {
2665  default:
2666  if (Result)
2667  Result->push_back(Char);
2668  break;
2669  case 0: // Null.
2670  // Found end of file?
2671  if (CurPtr-1 != BufferEnd) {
2672  if (isCodeCompletionPoint(CurPtr-1)) {
2674  cutOffLexing();
2675  return;
2676  }
2677 
2678  // Nope, normal character, continue.
2679  if (Result)
2680  Result->push_back(Char);
2681  break;
2682  }
2683  // FALL THROUGH.
2684  LLVM_FALLTHROUGH;
2685  case '\r':
2686  case '\n':
2687  // Okay, we found the end of the line. First, back up past the \0, \r, \n.
2688  assert(CurPtr[-1] == Char && "Trigraphs for newline?");
2689  BufferPtr = CurPtr-1;
2690 
2691  // Next, lex the character, which should handle the EOD transition.
2692  Lex(Tmp);
2693  if (Tmp.is(tok::code_completion)) {
2694  if (PP)
2696  Lex(Tmp);
2697  }
2698  assert(Tmp.is(tok::eod) && "Unexpected token!");
2699 
2700  // Finally, we're done;
2701  return;
2702  }
2703  }
2704 }
2705 
2706 /// LexEndOfFile - CurPtr points to the end of this file. Handle this
2707 /// condition, reporting diagnostics and handling other edge cases as required.
2708 /// This returns true if Result contains a token, false if PP.Lex should be
2709 /// called again.
2710 bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) {
2711  // If we hit the end of the file while parsing a preprocessor directive,
2712  // end the preprocessor directive first. The next token returned will
2713  // then be the end of file.
2715  // Done parsing the "line".
2717  // Update the location of token as well as BufferPtr.
2718  FormTokenWithChars(Result, CurPtr, tok::eod);
2719 
2720  // Restore comment saving mode, in case it was disabled for directive.
2721  if (PP)
2723  return true; // Have a token.
2724  }
2725 
2726  // If we are in raw mode, return this event as an EOF token. Let the caller
2727  // that put us in raw mode handle the event.
2728  if (isLexingRawMode()) {
2729  Result.startToken();
2730  BufferPtr = BufferEnd;
2731  FormTokenWithChars(Result, BufferEnd, tok::eof);
2732  return true;
2733  }
2734 
2735  if (PP->isRecordingPreamble() && PP->isInPrimaryFile()) {
2737  ConditionalStack.clear();
2738  }
2739 
2740  // Issue diagnostics for unterminated #if and missing newline.
2741 
2742  // If we are in a #if directive, emit an error.
2743  while (!ConditionalStack.empty()) {
2744  if (PP->getCodeCompletionFileLoc() != FileLoc)
2745  PP->Diag(ConditionalStack.back().IfLoc,
2746  diag::err_pp_unterminated_conditional);
2747  ConditionalStack.pop_back();
2748  }
2749 
2750  // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue
2751  // a pedwarn.
2752  if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) {
2753  DiagnosticsEngine &Diags = PP->getDiagnostics();
2754  SourceLocation EndLoc = getSourceLocation(BufferEnd);
2755  unsigned DiagID;
2756 
2757  if (LangOpts.CPlusPlus11) {
2758  // C++11 [lex.phases] 2.2 p2
2759  // Prefer the C++98 pedantic compatibility warning over the generic,
2760  // non-extension, user-requested "missing newline at EOF" warning.
2761  if (!Diags.isIgnored(diag::warn_cxx98_compat_no_newline_eof, EndLoc)) {
2762  DiagID = diag::warn_cxx98_compat_no_newline_eof;
2763  } else {
2764  DiagID = diag::warn_no_newline_eof;
2765  }
2766  } else {
2767  DiagID = diag::ext_no_newline_eof;
2768  }
2769 
2770  Diag(BufferEnd, DiagID)
2771  << FixItHint::CreateInsertion(EndLoc, "\n");
2772  }
2773 
2774  BufferPtr = CurPtr;
2775 
2776  // Finally, let the preprocessor handle this.
2777  return PP->HandleEndOfFile(Result, isPragmaLexer());
2778 }
2779 
2780 /// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from
2781 /// the specified lexer will return a tok::l_paren token, 0 if it is something
2782 /// else and 2 if there are no more tokens in the buffer controlled by the
2783 /// lexer.
2784 unsigned Lexer::isNextPPTokenLParen() {
2785  assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?");
2786 
2787  // Switch to 'skipping' mode. This will ensure that we can lex a token
2788  // without emitting diagnostics, disables macro expansion, and will cause EOF
2789  // to return an EOF token instead of popping the include stack.
2790  LexingRawMode = true;
2791 
2792  // Save state that can be changed while lexing so that we can restore it.
2793  const char *TmpBufferPtr = BufferPtr;
2794  bool inPPDirectiveMode = ParsingPreprocessorDirective;
2795  bool atStartOfLine = IsAtStartOfLine;
2796  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
2797  bool leadingSpace = HasLeadingSpace;
2798 
2799  Token Tok;
2800  Lex(Tok);
2801 
2802  // Restore state that may have changed.
2803  BufferPtr = TmpBufferPtr;
2804  ParsingPreprocessorDirective = inPPDirectiveMode;
2805  HasLeadingSpace = leadingSpace;
2806  IsAtStartOfLine = atStartOfLine;
2807  IsAtPhysicalStartOfLine = atPhysicalStartOfLine;
2808 
2809  // Restore the lexer back to non-skipping mode.
2810  LexingRawMode = false;
2811 
2812  if (Tok.is(tok::eof))
2813  return 2;
2814  return Tok.is(tok::l_paren);
2815 }
2816 
2817 /// Find the end of a version control conflict marker.
2818 static const char *FindConflictEnd(const char *CurPtr, const char *BufferEnd,
2819  ConflictMarkerKind CMK) {
2820  const char *Terminator = CMK == CMK_Perforce ? "<<<<\n" : ">>>>>>>";
2821  size_t TermLen = CMK == CMK_Perforce ? 5 : 7;
2822  auto RestOfBuffer = StringRef(CurPtr, BufferEnd - CurPtr).substr(TermLen);
2823  size_t Pos = RestOfBuffer.find(Terminator);
2824  while (Pos != StringRef::npos) {
2825  // Must occur at start of line.
2826  if (Pos == 0 ||
2827  (RestOfBuffer[Pos - 1] != '\r' && RestOfBuffer[Pos - 1] != '\n')) {
2828  RestOfBuffer = RestOfBuffer.substr(Pos+TermLen);
2829  Pos = RestOfBuffer.find(Terminator);
2830  continue;
2831  }
2832  return RestOfBuffer.data()+Pos;
2833  }
2834  return nullptr;
2835 }
2836 
2837 /// IsStartOfConflictMarker - If the specified pointer is the start of a version
2838 /// control conflict marker like '<<<<<<<', recognize it as such, emit an error
2839 /// and recover nicely. This returns true if it is a conflict marker and false
2840 /// if not.
2841 bool Lexer::IsStartOfConflictMarker(const char *CurPtr) {
2842  // Only a conflict marker if it starts at the beginning of a line.
2843  if (CurPtr != BufferStart &&
2844  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2845  return false;
2846 
2847  // Check to see if we have <<<<<<< or >>>>.
2848  if (!StringRef(CurPtr, BufferEnd - CurPtr).startswith("<<<<<<<") &&
2849  !StringRef(CurPtr, BufferEnd - CurPtr).startswith(">>>> "))
2850  return false;
2851 
2852  // If we have a situation where we don't care about conflict markers, ignore
2853  // it.
2854  if (CurrentConflictMarkerState || isLexingRawMode())
2855  return false;
2856 
2857  ConflictMarkerKind Kind = *CurPtr == '<' ? CMK_Normal : CMK_Perforce;
2858 
2859  // Check to see if there is an ending marker somewhere in the buffer at the
2860  // start of a line to terminate this conflict marker.
2861  if (FindConflictEnd(CurPtr, BufferEnd, Kind)) {
2862  // We found a match. We are really in a conflict marker.
2863  // Diagnose this, and ignore to the end of line.
2864  Diag(CurPtr, diag::err_conflict_marker);
2865  CurrentConflictMarkerState = Kind;
2866 
2867  // Skip ahead to the end of line. We know this exists because the
2868  // end-of-conflict marker starts with \r or \n.
2869  while (*CurPtr != '\r' && *CurPtr != '\n') {
2870  assert(CurPtr != BufferEnd && "Didn't find end of line");
2871  ++CurPtr;
2872  }
2873  BufferPtr = CurPtr;
2874  return true;
2875  }
2876 
2877  // No end of conflict marker found.
2878  return false;
2879 }
2880 
2881 /// HandleEndOfConflictMarker - If this is a '====' or '||||' or '>>>>', or if
2882 /// it is '<<<<' and the conflict marker started with a '>>>>' marker, then it
2883 /// is the end of a conflict marker. Handle it by ignoring up until the end of
2884 /// the line. This returns true if it is a conflict marker and false if not.
2885 bool Lexer::HandleEndOfConflictMarker(const char *CurPtr) {
2886  // Only a conflict marker if it starts at the beginning of a line.
2887  if (CurPtr != BufferStart &&
2888  CurPtr[-1] != '\n' && CurPtr[-1] != '\r')
2889  return false;
2890 
2891  // If we have a situation where we don't care about conflict markers, ignore
2892  // it.
2893  if (!CurrentConflictMarkerState || isLexingRawMode())
2894  return false;
2895 
2896  // Check to see if we have the marker (4 characters in a row).
2897  for (unsigned i = 1; i != 4; ++i)
2898  if (CurPtr[i] != CurPtr[0])
2899  return false;
2900 
2901  // If we do have it, search for the end of the conflict marker. This could
2902  // fail if it got skipped with a '#if 0' or something. Note that CurPtr might
2903  // be the end of conflict marker.
2904  if (const char *End = FindConflictEnd(CurPtr, BufferEnd,
2905  CurrentConflictMarkerState)) {
2906  CurPtr = End;
2907 
2908  // Skip ahead to the end of line.
2909  while (CurPtr != BufferEnd && *CurPtr != '\r' && *CurPtr != '\n')
2910  ++CurPtr;
2911 
2912  BufferPtr = CurPtr;
2913 
2914  // No longer in the conflict marker.
2915  CurrentConflictMarkerState = CMK_None;
2916  return true;
2917  }
2918 
2919  return false;
2920 }
2921 
2922 static const char *findPlaceholderEnd(const char *CurPtr,
2923  const char *BufferEnd) {
2924  if (CurPtr == BufferEnd)
2925  return nullptr;
2926  BufferEnd -= 1; // Scan until the second last character.
2927  for (; CurPtr != BufferEnd; ++CurPtr) {
2928  if (CurPtr[0] == '#' && CurPtr[1] == '>')
2929  return CurPtr + 2;
2930  }
2931  return nullptr;
2932 }
2933 
2934 bool Lexer::lexEditorPlaceholder(Token &Result, const char *CurPtr) {
2935  assert(CurPtr[-1] == '<' && CurPtr[0] == '#' && "Not a placeholder!");
2937  return false;
2938  const char *End = findPlaceholderEnd(CurPtr + 1, BufferEnd);
2939  if (!End)
2940  return false;
2941  const char *Start = CurPtr - 1;
2942  if (!LangOpts.AllowEditorPlaceholders)
2943  Diag(Start, diag::err_placeholder_in_source);
2944  Result.startToken();
2945  FormTokenWithChars(Result, End, tok::raw_identifier);
2946  Result.setRawIdentifierData(Start);
2947  PP->LookUpIdentifierInfo(Result);
2949  BufferPtr = End;
2950  return true;
2951 }
2952 
2953 bool Lexer::isCodeCompletionPoint(const char *CurPtr) const {
2954  if (PP && PP->isCodeCompletionEnabled()) {
2955  SourceLocation Loc = FileLoc.getLocWithOffset(CurPtr-BufferStart);
2956  return Loc == PP->getCodeCompletionLoc();
2957  }
2958 
2959  return false;
2960 }
2961 
2962 uint32_t Lexer::tryReadUCN(const char *&StartPtr, const char *SlashLoc,
2963  Token *Result) {
2964  unsigned CharSize;
2965  char Kind = getCharAndSize(StartPtr, CharSize);
2966 
2967  unsigned NumHexDigits;
2968  if (Kind == 'u')
2969  NumHexDigits = 4;
2970  else if (Kind == 'U')
2971  NumHexDigits = 8;
2972  else
2973  return 0;
2974 
2975  if (!LangOpts.CPlusPlus && !LangOpts.C99) {
2976  if (Result && !isLexingRawMode())
2977  Diag(SlashLoc, diag::warn_ucn_not_valid_in_c89);
2978  return 0;
2979  }
2980 
2981  const char *CurPtr = StartPtr + CharSize;
2982  const char *KindLoc = &CurPtr[-1];
2983 
2984  uint32_t CodePoint = 0;
2985  for (unsigned i = 0; i < NumHexDigits; ++i) {
2986  char C = getCharAndSize(CurPtr, CharSize);
2987 
2988  unsigned Value = llvm::hexDigitValue(C);
2989  if (Value == -1U) {
2990  if (Result && !isLexingRawMode()) {
2991  if (i == 0) {
2992  Diag(BufferPtr, diag::warn_ucn_escape_no_digits)
2993  << StringRef(KindLoc, 1);
2994  } else {
2995  Diag(BufferPtr, diag::warn_ucn_escape_incomplete);
2996 
2997  // If the user wrote \U1234, suggest a fixit to \u.
2998  if (i == 4 && NumHexDigits == 8) {
2999  CharSourceRange URange = makeCharRange(*this, KindLoc, KindLoc + 1);
3000  Diag(KindLoc, diag::note_ucn_four_not_eight)
3001  << FixItHint::CreateReplacement(URange, "u");
3002  }
3003  }
3004  }
3005 
3006  return 0;
3007  }
3008 
3009  CodePoint <<= 4;
3010  CodePoint += Value;
3011 
3012  CurPtr += CharSize;
3013  }
3014 
3015  if (Result) {
3016  Result->setFlag(Token::HasUCN);
3017  if (CurPtr - StartPtr == (ptrdiff_t)NumHexDigits + 2)
3018  StartPtr = CurPtr;
3019  else
3020  while (StartPtr != CurPtr)
3021  (void)getAndAdvanceChar(StartPtr, *Result);
3022  } else {
3023  StartPtr = CurPtr;
3024  }
3025 
3026  // Don't apply C family restrictions to UCNs in assembly mode
3027  if (LangOpts.AsmPreprocessor)
3028  return CodePoint;
3029 
3030  // C99 6.4.3p2: A universal character name shall not specify a character whose
3031  // short identifier is less than 00A0 other than 0024 ($), 0040 (@), or
3032  // 0060 (`), nor one in the range D800 through DFFF inclusive.)
3033  // C++11 [lex.charset]p2: If the hexadecimal value for a
3034  // universal-character-name corresponds to a surrogate code point (in the
3035  // range 0xD800-0xDFFF, inclusive), the program is ill-formed. Additionally,
3036  // if the hexadecimal value for a universal-character-name outside the
3037  // c-char-sequence, s-char-sequence, or r-char-sequence of a character or
3038  // string literal corresponds to a control character (in either of the
3039  // ranges 0x00-0x1F or 0x7F-0x9F, both inclusive) or to a character in the
3040  // basic source character set, the program is ill-formed.
3041  if (CodePoint < 0xA0) {
3042  if (CodePoint == 0x24 || CodePoint == 0x40 || CodePoint == 0x60)
3043  return CodePoint;
3044 
3045  // We don't use isLexingRawMode() here because we need to warn about bad
3046  // UCNs even when skipping preprocessing tokens in a #if block.
3047  if (Result && PP) {
3048  if (CodePoint < 0x20 || CodePoint >= 0x7F)
3049  Diag(BufferPtr, diag::err_ucn_control_character);
3050  else {
3051  char C = static_cast<char>(CodePoint);
3052  Diag(BufferPtr, diag::err_ucn_escape_basic_scs) << StringRef(&C, 1);
3053  }
3054  }
3055 
3056  return 0;
3057  } else if (CodePoint >= 0xD800 && CodePoint <= 0xDFFF) {
3058  // C++03 allows UCNs representing surrogate characters. C99 and C++11 don't.
3059  // We don't use isLexingRawMode() here because we need to diagnose bad
3060  // UCNs even when skipping preprocessing tokens in a #if block.
3061  if (Result && PP) {
3062  if (LangOpts.CPlusPlus && !LangOpts.CPlusPlus11)
3063  Diag(BufferPtr, diag::warn_ucn_escape_surrogate);
3064  else
3065  Diag(BufferPtr, diag::err_ucn_escape_invalid);
3066  }
3067  return 0;
3068  }
3069 
3070  return CodePoint;
3071 }
3072 
3073 bool Lexer::CheckUnicodeWhitespace(Token &Result, uint32_t C,
3074  const char *CurPtr) {
3075  static const llvm::sys::UnicodeCharSet UnicodeWhitespaceChars(
3077  if (!isLexingRawMode() && !PP->isPreprocessedOutput() &&
3078  UnicodeWhitespaceChars.contains(C)) {
3079  Diag(BufferPtr, diag::ext_unicode_whitespace)
3080  << makeCharRange(*this, BufferPtr, CurPtr);
3081 
3082  Result.setFlag(Token::LeadingSpace);
3083  return true;
3084  }
3085  return false;
3086 }
3087 
3088 bool Lexer::LexUnicode(Token &Result, uint32_t C, const char *CurPtr) {
3089  if (isAllowedIDChar(C, LangOpts) && isAllowedInitiallyIDChar(C, LangOpts)) {
3091  !PP->isPreprocessedOutput()) {
3093  makeCharRange(*this, BufferPtr, CurPtr),
3094  /*IsFirst=*/true);
3096  makeCharRange(*this, BufferPtr, CurPtr));
3097  }
3098 
3099  MIOpt.ReadToken();
3100  return LexIdentifier(Result, CurPtr);
3101  }
3102 
3104  !PP->isPreprocessedOutput() &&
3105  !isASCII(*BufferPtr) && !isAllowedIDChar(C, LangOpts)) {
3106  // Non-ASCII characters tend to creep into source code unintentionally.
3107  // Instead of letting the parser complain about the unknown token,
3108  // just drop the character.
3109  // Note that we can /only/ do this when the non-ASCII character is actually
3110  // spelled as Unicode, not written as a UCN. The standard requires that
3111  // we not throw away any possible preprocessor tokens, but there's a
3112  // loophole in the mapping of Unicode characters to basic character set
3113  // characters that allows us to map these particular characters to, say,
3114  // whitespace.
3115  Diag(BufferPtr, diag::err_non_ascii)
3116  << FixItHint::CreateRemoval(makeCharRange(*this, BufferPtr, CurPtr));
3117 
3118  BufferPtr = CurPtr;
3119  return false;
3120  }
3121 
3122  // Otherwise, we have an explicit UCN or a character that's unlikely to show
3123  // up by accident.
3124  MIOpt.ReadToken();
3125  FormTokenWithChars(Result, CurPtr, tok::unknown);
3126  return true;
3127 }
3128 
3129 void Lexer::PropagateLineStartLeadingSpaceInfo(Token &Result) {
3130  IsAtStartOfLine = Result.isAtStartOfLine();
3131  HasLeadingSpace = Result.hasLeadingSpace();
3132  HasLeadingEmptyMacro = Result.hasLeadingEmptyMacro();
3133  // Note that this doesn't affect IsAtPhysicalStartOfLine.
3134 }
3135 
3136 bool Lexer::Lex(Token &Result) {
3137  // Start a new token.
3138  Result.startToken();
3139 
3140  // Set up misc whitespace flags for LexTokenInternal.
3141  if (IsAtStartOfLine) {
3142  Result.setFlag(Token::StartOfLine);
3143  IsAtStartOfLine = false;
3144  }
3145 
3146  if (HasLeadingSpace) {
3147  Result.setFlag(Token::LeadingSpace);
3148  HasLeadingSpace = false;
3149  }
3150 
3151  if (HasLeadingEmptyMacro) {
3153  HasLeadingEmptyMacro = false;
3154  }
3155 
3156  bool atPhysicalStartOfLine = IsAtPhysicalStartOfLine;
3157  IsAtPhysicalStartOfLine = false;
3158  bool isRawLex = isLexingRawMode();
3159  (void) isRawLex;
3160  bool returnedToken = LexTokenInternal(Result, atPhysicalStartOfLine);
3161  // (After the LexTokenInternal call, the lexer might be destroyed.)
3162  assert((returnedToken || !isRawLex) && "Raw lex must succeed");
3163  return returnedToken;
3164 }
3165 
3166 /// LexTokenInternal - This implements a simple C family lexer. It is an
3167 /// extremely performance critical piece of code. This assumes that the buffer
3168 /// has a null character at the end of the file. This returns a preprocessing
3169 /// token, not a normal token, as such, it is an internal interface. It assumes
3170 /// that the Flags of result have been cleared before calling this.
3171 bool Lexer::LexTokenInternal(Token &Result, bool TokAtPhysicalStartOfLine) {
3172 LexNextToken:
3173  // New token, can't need cleaning yet.
3175  Result.setIdentifierInfo(nullptr);
3176 
3177  // CurPtr - Cache BufferPtr in an automatic variable.
3178  const char *CurPtr = BufferPtr;
3179 
3180  // Small amounts of horizontal whitespace is very common between tokens.
3181  if ((*CurPtr == ' ') || (*CurPtr == '\t')) {
3182  ++CurPtr;
3183  while ((*CurPtr == ' ') || (*CurPtr == '\t'))
3184  ++CurPtr;
3185 
3186  // If we are keeping whitespace and other tokens, just return what we just
3187  // skipped. The next lexer invocation will return the token after the
3188  // whitespace.
3189  if (isKeepWhitespaceMode()) {
3190  FormTokenWithChars(Result, CurPtr, tok::unknown);
3191  // FIXME: The next token will not have LeadingSpace set.
3192  return true;
3193  }
3194 
3195  BufferPtr = CurPtr;
3196  Result.setFlag(Token::LeadingSpace);
3197  }
3198 
3199  unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below.
3200 
3201  // Read a character, advancing over it.
3202  char Char = getAndAdvanceChar(CurPtr, Result);
3204 
3205  switch (Char) {
3206  case 0: // Null.
3207  // Found end of file?
3208  if (CurPtr-1 == BufferEnd)
3209  return LexEndOfFile(Result, CurPtr-1);
3210 
3211  // Check if we are performing code completion.
3212  if (isCodeCompletionPoint(CurPtr-1)) {
3213  // Return the code-completion token.
3214  Result.startToken();
3215  FormTokenWithChars(Result, CurPtr, tok::code_completion);
3216  return true;
3217  }
3218 
3219  if (!isLexingRawMode())
3220  Diag(CurPtr-1, diag::null_in_file);
3221  Result.setFlag(Token::LeadingSpace);
3222  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3223  return true; // KeepWhitespaceMode
3224 
3225  // We know the lexer hasn't changed, so just try again with this lexer.
3226  // (We manually eliminate the tail call to avoid recursion.)
3227  goto LexNextToken;
3228 
3229  case 26: // DOS & CP/M EOF: "^Z".
3230  // If we're in Microsoft extensions mode, treat this as end of file.
3231  if (LangOpts.MicrosoftExt) {
3232  if (!isLexingRawMode())
3233  Diag(CurPtr-1, diag::ext_ctrl_z_eof_microsoft);
3234  return LexEndOfFile(Result, CurPtr-1);
3235  }
3236 
3237  // If Microsoft extensions are disabled, this is just random garbage.
3238  Kind = tok::unknown;
3239  break;
3240 
3241  case '\r':
3242  if (CurPtr[0] == '\n')
3243  (void)getAndAdvanceChar(CurPtr, Result);
3244  LLVM_FALLTHROUGH;
3245  case '\n':
3246  // If we are inside a preprocessor directive and we see the end of line,
3247  // we know we are done with the directive, so return an EOD token.
3249  // Done parsing the "line".
3251 
3252  // Restore comment saving mode, in case it was disabled for directive.
3253  if (PP)
3255 
3256  // Since we consumed a newline, we are back at the start of a line.
3257  IsAtStartOfLine = true;
3258  IsAtPhysicalStartOfLine = true;
3259 
3260  Kind = tok::eod;
3261  break;
3262  }
3263 
3264  // No leading whitespace seen so far.
3266 
3267  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3268  return true; // KeepWhitespaceMode
3269 
3270  // We only saw whitespace, so just try again with this lexer.
3271  // (We manually eliminate the tail call to avoid recursion.)
3272  goto LexNextToken;
3273  case ' ':
3274  case '\t':
3275  case '\f':
3276  case '\v':
3277  SkipHorizontalWhitespace:
3278  Result.setFlag(Token::LeadingSpace);
3279  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3280  return true; // KeepWhitespaceMode
3281 
3282  SkipIgnoredUnits:
3283  CurPtr = BufferPtr;
3284 
3285  // If the next token is obviously a // or /* */ comment, skip it efficiently
3286  // too (without going through the big switch stmt).
3287  if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() &&
3288  LangOpts.LineComment &&
3289  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP)) {
3290  if (SkipLineComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3291  return true; // There is a token to return.
3292  goto SkipIgnoredUnits;
3293  } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) {
3294  if (SkipBlockComment(Result, CurPtr+2, TokAtPhysicalStartOfLine))
3295  return true; // There is a token to return.
3296  goto SkipIgnoredUnits;
3297  } else if (isHorizontalWhitespace(*CurPtr)) {
3298  goto SkipHorizontalWhitespace;
3299  }
3300  // We only saw whitespace, so just try again with this lexer.
3301  // (We manually eliminate the tail call to avoid recursion.)
3302  goto LexNextToken;
3303 
3304  // C99 6.4.4.1: Integer Constants.
3305  // C99 6.4.4.2: Floating Constants.
3306  case '0': case '1': case '2': case '3': case '4':
3307  case '5': case '6': case '7': case '8': case '9':
3308  // Notify MIOpt that we read a non-whitespace/non-comment token.
3309  MIOpt.ReadToken();
3310  return LexNumericConstant(Result, CurPtr);
3311 
3312  case 'u': // Identifier (uber) or C11/C++11 UTF-8 or UTF-16 string literal
3313  // Notify MIOpt that we read a non-whitespace/non-comment token.
3314  MIOpt.ReadToken();
3315 
3316  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3317  Char = getCharAndSize(CurPtr, SizeTmp);
3318 
3319  // UTF-16 string literal
3320  if (Char == '"')
3321  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3322  tok::utf16_string_literal);
3323 
3324  // UTF-16 character constant
3325  if (Char == '\'')
3326  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3327  tok::utf16_char_constant);
3328 
3329  // UTF-16 raw string literal
3330  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3331  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3332  return LexRawStringLiteral(Result,
3333  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3334  SizeTmp2, Result),
3335  tok::utf16_string_literal);
3336 
3337  if (Char == '8') {
3338  char Char2 = getCharAndSize(CurPtr + SizeTmp, SizeTmp2);
3339 
3340  // UTF-8 string literal
3341  if (Char2 == '"')
3342  return LexStringLiteral(Result,
3343  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3344  SizeTmp2, Result),
3345  tok::utf8_string_literal);
3346  if (Char2 == '\'' && LangOpts.CPlusPlus17)
3347  return LexCharConstant(
3348  Result, ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3349  SizeTmp2, Result),
3350  tok::utf8_char_constant);
3351 
3352  if (Char2 == 'R' && LangOpts.CPlusPlus11) {
3353  unsigned SizeTmp3;
3354  char Char3 = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3355  // UTF-8 raw string literal
3356  if (Char3 == '"') {
3357  return LexRawStringLiteral(Result,
3358  ConsumeChar(ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3359  SizeTmp2, Result),
3360  SizeTmp3, Result),
3361  tok::utf8_string_literal);
3362  }
3363  }
3364  }
3365  }
3366 
3367  // treat u like the start of an identifier.
3368  return LexIdentifier(Result, CurPtr);
3369 
3370  case 'U': // Identifier (Uber) or C11/C++11 UTF-32 string literal
3371  // Notify MIOpt that we read a non-whitespace/non-comment token.
3372  MIOpt.ReadToken();
3373 
3374  if (LangOpts.CPlusPlus11 || LangOpts.C11) {
3375  Char = getCharAndSize(CurPtr, SizeTmp);
3376 
3377  // UTF-32 string literal
3378  if (Char == '"')
3379  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3380  tok::utf32_string_literal);
3381 
3382  // UTF-32 character constant
3383  if (Char == '\'')
3384  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3385  tok::utf32_char_constant);
3386 
3387  // UTF-32 raw string literal
3388  if (Char == 'R' && LangOpts.CPlusPlus11 &&
3389  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3390  return LexRawStringLiteral(Result,
3391  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3392  SizeTmp2, Result),
3393  tok::utf32_string_literal);
3394  }
3395 
3396  // treat U like the start of an identifier.
3397  return LexIdentifier(Result, CurPtr);
3398 
3399  case 'R': // Identifier or C++0x raw string literal
3400  // Notify MIOpt that we read a non-whitespace/non-comment token.
3401  MIOpt.ReadToken();
3402 
3403  if (LangOpts.CPlusPlus11) {
3404  Char = getCharAndSize(CurPtr, SizeTmp);
3405 
3406  if (Char == '"')
3407  return LexRawStringLiteral(Result,
3408  ConsumeChar(CurPtr, SizeTmp, Result),
3409  tok::string_literal);
3410  }
3411 
3412  // treat R like the start of an identifier.
3413  return LexIdentifier(Result, CurPtr);
3414 
3415  case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz").
3416  // Notify MIOpt that we read a non-whitespace/non-comment token.
3417  MIOpt.ReadToken();
3418  Char = getCharAndSize(CurPtr, SizeTmp);
3419 
3420  // Wide string literal.
3421  if (Char == '"')
3422  return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3423  tok::wide_string_literal);
3424 
3425  // Wide raw string literal.
3426  if (LangOpts.CPlusPlus11 && Char == 'R' &&
3427  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == '"')
3428  return LexRawStringLiteral(Result,
3429  ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3430  SizeTmp2, Result),
3431  tok::wide_string_literal);
3432 
3433  // Wide character constant.
3434  if (Char == '\'')
3435  return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3436  tok::wide_char_constant);
3437  // FALL THROUGH, treating L like the start of an identifier.
3438  LLVM_FALLTHROUGH;
3439 
3440  // C99 6.4.2: Identifiers.
3441  case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G':
3442  case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N':
3443  case 'O': case 'P': case 'Q': /*'R'*/case 'S': case 'T': /*'U'*/
3444  case 'V': case 'W': case 'X': case 'Y': case 'Z':
3445  case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g':
3446  case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n':
3447  case 'o': case 'p': case 'q': case 'r': case 's': case 't': /*'u'*/
3448  case 'v': case 'w': case 'x': case 'y': case 'z':
3449  case '_':
3450  // Notify MIOpt that we read a non-whitespace/non-comment token.
3451  MIOpt.ReadToken();
3452  return LexIdentifier(Result, CurPtr);
3453 
3454  case '$': // $ in identifiers.
3455  if (LangOpts.DollarIdents) {
3456  if (!isLexingRawMode())
3457  Diag(CurPtr-1, diag::ext_dollar_in_identifier);
3458  // Notify MIOpt that we read a non-whitespace/non-comment token.
3459  MIOpt.ReadToken();
3460  return LexIdentifier(Result, CurPtr);
3461  }
3462 
3463  Kind = tok::unknown;
3464  break;
3465 
3466  // C99 6.4.4: Character Constants.
3467  case '\'':
3468  // Notify MIOpt that we read a non-whitespace/non-comment token.
3469  MIOpt.ReadToken();
3470  return LexCharConstant(Result, CurPtr, tok::char_constant);
3471 
3472  // C99 6.4.5: String Literals.
3473  case '"':
3474  // Notify MIOpt that we read a non-whitespace/non-comment token.
3475  MIOpt.ReadToken();
3476  return LexStringLiteral(Result, CurPtr,
3477  ParsingFilename ? tok::header_name
3478  : tok::string_literal);
3479 
3480  // C99 6.4.6: Punctuators.
3481  case '?':
3482  Kind = tok::question;
3483  break;
3484  case '[':
3485  Kind = tok::l_square;
3486  break;
3487  case ']':
3488  Kind = tok::r_square;
3489  break;
3490  case '(':
3491  Kind = tok::l_paren;
3492  break;
3493  case ')':
3494  Kind = tok::r_paren;
3495  break;
3496  case '{':
3497  Kind = tok::l_brace;
3498  break;
3499  case '}':
3500  Kind = tok::r_brace;
3501  break;
3502  case '.':
3503  Char = getCharAndSize(CurPtr, SizeTmp);
3504  if (Char >= '0' && Char <= '9') {
3505  // Notify MIOpt that we read a non-whitespace/non-comment token.
3506  MIOpt.ReadToken();
3507 
3508  return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result));
3509  } else if (LangOpts.CPlusPlus && Char == '*') {
3510  Kind = tok::periodstar;
3511  CurPtr += SizeTmp;
3512  } else if (Char == '.' &&
3513  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') {
3514  Kind = tok::ellipsis;
3515  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3516  SizeTmp2, Result);
3517  } else {
3518  Kind = tok::period;
3519  }
3520  break;
3521  case '&':
3522  Char = getCharAndSize(CurPtr, SizeTmp);
3523  if (Char == '&') {
3524  Kind = tok::ampamp;
3525  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3526  } else if (Char == '=') {
3527  Kind = tok::ampequal;
3528  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3529  } else {
3530  Kind = tok::amp;
3531  }
3532  break;
3533  case '*':
3534  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3535  Kind = tok::starequal;
3536  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3537  } else {
3538  Kind = tok::star;
3539  }
3540  break;
3541  case '+':
3542  Char = getCharAndSize(CurPtr, SizeTmp);
3543  if (Char == '+') {
3544  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3545  Kind = tok::plusplus;
3546  } else if (Char == '=') {
3547  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3548  Kind = tok::plusequal;
3549  } else {
3550  Kind = tok::plus;
3551  }
3552  break;
3553  case '-':
3554  Char = getCharAndSize(CurPtr, SizeTmp);
3555  if (Char == '-') { // --
3556  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3557  Kind = tok::minusminus;
3558  } else if (Char == '>' && LangOpts.CPlusPlus &&
3559  getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->*
3560  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3561  SizeTmp2, Result);
3562  Kind = tok::arrowstar;
3563  } else if (Char == '>') { // ->
3564  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3565  Kind = tok::arrow;
3566  } else if (Char == '=') { // -=
3567  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3568  Kind = tok::minusequal;
3569  } else {
3570  Kind = tok::minus;
3571  }
3572  break;
3573  case '~':
3574  Kind = tok::tilde;
3575  break;
3576  case '!':
3577  if (getCharAndSize(CurPtr, SizeTmp) == '=') {
3578  Kind = tok::exclaimequal;
3579  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3580  } else {
3581  Kind = tok::exclaim;
3582  }
3583  break;
3584  case '/':
3585  // 6.4.9: Comments
3586  Char = getCharAndSize(CurPtr, SizeTmp);
3587  if (Char == '/') { // Line comment.
3588  // Even if Line comments are disabled (e.g. in C89 mode), we generally
3589  // want to lex this as a comment. There is one problem with this though,
3590  // that in one particular corner case, this can change the behavior of the
3591  // resultant program. For example, In "foo //**/ bar", C89 would lex
3592  // this as "foo / bar" and languages with Line comments would lex it as
3593  // "foo". Check to see if the character after the second slash is a '*'.
3594  // If so, we will lex that as a "/" instead of the start of a comment.
3595  // However, we never do this if we are just preprocessing.
3596  bool TreatAsComment = LangOpts.LineComment &&
3597  (LangOpts.CPlusPlus || !LangOpts.TraditionalCPP);
3598  if (!TreatAsComment)
3599  if (!(PP && PP->isPreprocessedOutput()))
3600  TreatAsComment = getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*';
3601 
3602  if (TreatAsComment) {
3603  if (SkipLineComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3604  TokAtPhysicalStartOfLine))
3605  return true; // There is a token to return.
3606 
3607  // It is common for the tokens immediately after a // comment to be
3608  // whitespace (indentation for the next line). Instead of going through
3609  // the big switch, handle it efficiently now.
3610  goto SkipIgnoredUnits;
3611  }
3612  }
3613 
3614  if (Char == '*') { // /**/ comment.
3615  if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result),
3616  TokAtPhysicalStartOfLine))
3617  return true; // There is a token to return.
3618 
3619  // We only saw whitespace, so just try again with this lexer.
3620  // (We manually eliminate the tail call to avoid recursion.)
3621  goto LexNextToken;
3622  }
3623 
3624  if (Char == '=') {
3625  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3626  Kind = tok::slashequal;
3627  } else {
3628  Kind = tok::slash;
3629  }
3630  break;
3631  case '%':
3632  Char = getCharAndSize(CurPtr, SizeTmp);
3633  if (Char == '=') {
3634  Kind = tok::percentequal;
3635  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3636  } else if (LangOpts.Digraphs && Char == '>') {
3637  Kind = tok::r_brace; // '%>' -> '}'
3638  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3639  } else if (LangOpts.Digraphs && Char == ':') {
3640  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3641  Char = getCharAndSize(CurPtr, SizeTmp);
3642  if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') {
3643  Kind = tok::hashhash; // '%:%:' -> '##'
3644  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3645  SizeTmp2, Result);
3646  } else if (Char == '@' && LangOpts.MicrosoftExt) {// %:@ -> #@ -> Charize
3647  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3648  if (!isLexingRawMode())
3649  Diag(BufferPtr, diag::ext_charize_microsoft);
3650  Kind = tok::hashat;
3651  } else { // '%:' -> '#'
3652  // We parsed a # character. If this occurs at the start of the line,
3653  // it's actually the start of a preprocessing directive. Callback to
3654  // the preprocessor to handle it.
3655  // TODO: -fpreprocessed mode??
3656  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3657  goto HandleDirective;
3658 
3659  Kind = tok::hash;
3660  }
3661  } else {
3662  Kind = tok::percent;
3663  }
3664  break;
3665  case '<':
3666  Char = getCharAndSize(CurPtr, SizeTmp);
3667  if (ParsingFilename) {
3668  return LexAngledStringLiteral(Result, CurPtr);
3669  } else if (Char == '<') {
3670  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3671  if (After == '=') {
3672  Kind = tok::lesslessequal;
3673  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3674  SizeTmp2, Result);
3675  } else if (After == '<' && IsStartOfConflictMarker(CurPtr-1)) {
3676  // If this is actually a '<<<<<<<' version control conflict marker,
3677  // recognize it as such and recover nicely.
3678  goto LexNextToken;
3679  } else if (After == '<' && HandleEndOfConflictMarker(CurPtr-1)) {
3680  // If this is '<<<<' and we're in a Perforce-style conflict marker,
3681  // ignore it.
3682  goto LexNextToken;
3683  } else if (LangOpts.CUDA && After == '<') {
3684  Kind = tok::lesslessless;
3685  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3686  SizeTmp2, Result);
3687  } else {
3688  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3689  Kind = tok::lessless;
3690  }
3691  } else if (Char == '=') {
3692  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3693  if (After == '>') {
3694  if (getLangOpts().CPlusPlus2a) {
3695  if (!isLexingRawMode())
3696  Diag(BufferPtr, diag::warn_cxx17_compat_spaceship);
3697  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3698  SizeTmp2, Result);
3699  Kind = tok::spaceship;
3700  break;
3701  }
3702  // Suggest adding a space between the '<=' and the '>' to avoid a
3703  // change in semantics if this turns up in C++ <=17 mode.
3704  if (getLangOpts().CPlusPlus && !isLexingRawMode()) {
3705  Diag(BufferPtr, diag::warn_cxx2a_compat_spaceship)
3707  getSourceLocation(CurPtr + SizeTmp, SizeTmp2), " ");
3708  }
3709  }
3710  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3711  Kind = tok::lessequal;
3712  } else if (LangOpts.Digraphs && Char == ':') { // '<:' -> '['
3713  if (LangOpts.CPlusPlus11 &&
3714  getCharAndSize(CurPtr + SizeTmp, SizeTmp2) == ':') {
3715  // C++0x [lex.pptoken]p3:
3716  // Otherwise, if the next three characters are <:: and the subsequent
3717  // character is neither : nor >, the < is treated as a preprocessor
3718  // token by itself and not as the first character of the alternative
3719  // token <:.
3720  unsigned SizeTmp3;
3721  char After = getCharAndSize(CurPtr + SizeTmp + SizeTmp2, SizeTmp3);
3722  if (After != ':' && After != '>') {
3723  Kind = tok::less;
3724  if (!isLexingRawMode())
3725  Diag(BufferPtr, diag::warn_cxx98_compat_less_colon_colon);
3726  break;
3727  }
3728  }
3729 
3730  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3731  Kind = tok::l_square;
3732  } else if (LangOpts.Digraphs && Char == '%') { // '<%' -> '{'
3733  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3734  Kind = tok::l_brace;
3735  } else if (Char == '#' && /*Not a trigraph*/ SizeTmp == 1 &&
3736  lexEditorPlaceholder(Result, CurPtr)) {
3737  return true;
3738  } else {
3739  Kind = tok::less;
3740  }
3741  break;
3742  case '>':
3743  Char = getCharAndSize(CurPtr, SizeTmp);
3744  if (Char == '=') {
3745  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3746  Kind = tok::greaterequal;
3747  } else if (Char == '>') {
3748  char After = getCharAndSize(CurPtr+SizeTmp, SizeTmp2);
3749  if (After == '=') {
3750  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3751  SizeTmp2, Result);
3752  Kind = tok::greatergreaterequal;
3753  } else if (After == '>' && IsStartOfConflictMarker(CurPtr-1)) {
3754  // If this is actually a '>>>>' conflict marker, recognize it as such
3755  // and recover nicely.
3756  goto LexNextToken;
3757  } else if (After == '>' && HandleEndOfConflictMarker(CurPtr-1)) {
3758  // If this is '>>>>>>>' and we're in a conflict marker, ignore it.
3759  goto LexNextToken;
3760  } else if (LangOpts.CUDA && After == '>') {
3761  Kind = tok::greatergreatergreater;
3762  CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result),
3763  SizeTmp2, Result);
3764  } else {
3765  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3766  Kind = tok::greatergreater;
3767  }
3768  } else {
3769  Kind = tok::greater;
3770  }
3771  break;
3772  case '^':
3773  Char = getCharAndSize(CurPtr, SizeTmp);
3774  if (Char == '=') {
3775  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3776  Kind = tok::caretequal;
3777  } else if (LangOpts.OpenCL && Char == '^') {
3778  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3779  Kind = tok::caretcaret;
3780  } else {
3781  Kind = tok::caret;
3782  }
3783  break;
3784  case '|':
3785  Char = getCharAndSize(CurPtr, SizeTmp);
3786  if (Char == '=') {
3787  Kind = tok::pipeequal;
3788  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3789  } else if (Char == '|') {
3790  // If this is '|||||||' and we're in a conflict marker, ignore it.
3791  if (CurPtr[1] == '|' && HandleEndOfConflictMarker(CurPtr-1))
3792  goto LexNextToken;
3793  Kind = tok::pipepipe;
3794  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3795  } else {
3796  Kind = tok::pipe;
3797  }
3798  break;
3799  case ':':
3800  Char = getCharAndSize(CurPtr, SizeTmp);
3801  if (LangOpts.Digraphs && Char == '>') {
3802  Kind = tok::r_square; // ':>' -> ']'
3803  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3804  } else if ((LangOpts.CPlusPlus ||
3805  LangOpts.DoubleSquareBracketAttributes) &&
3806  Char == ':') {
3807  Kind = tok::coloncolon;
3808  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3809  } else {
3810  Kind = tok::colon;
3811  }
3812  break;
3813  case ';':
3814  Kind = tok::semi;
3815  break;
3816  case '=':
3817  Char = getCharAndSize(CurPtr, SizeTmp);
3818  if (Char == '=') {
3819  // If this is '====' and we're in a conflict marker, ignore it.
3820  if (CurPtr[1] == '=' && HandleEndOfConflictMarker(CurPtr-1))
3821  goto LexNextToken;
3822 
3823  Kind = tok::equalequal;
3824  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3825  } else {
3826  Kind = tok::equal;
3827  }
3828  break;
3829  case ',':
3830  Kind = tok::comma;
3831  break;
3832  case '#':
3833  Char = getCharAndSize(CurPtr, SizeTmp);
3834  if (Char == '#') {
3835  Kind = tok::hashhash;
3836  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3837  } else if (Char == '@' && LangOpts.MicrosoftExt) { // #@ -> Charize
3838  Kind = tok::hashat;
3839  if (!isLexingRawMode())
3840  Diag(BufferPtr, diag::ext_charize_microsoft);
3841  CurPtr = ConsumeChar(CurPtr, SizeTmp, Result);
3842  } else {
3843  // We parsed a # character. If this occurs at the start of the line,
3844  // it's actually the start of a preprocessing directive. Callback to
3845  // the preprocessor to handle it.
3846  // TODO: -fpreprocessed mode??
3847  if (TokAtPhysicalStartOfLine && !LexingRawMode && !Is_PragmaLexer)
3848  goto HandleDirective;
3849 
3850  Kind = tok::hash;
3851  }
3852  break;
3853 
3854  case '@':
3855  // Objective C support.
3856  if (CurPtr[-1] == '@' && LangOpts.ObjC)
3857  Kind = tok::at;
3858  else
3859  Kind = tok::unknown;
3860  break;
3861 
3862  // UCNs (C99 6.4.3, C++11 [lex.charset]p2)
3863  case '\\':
3864  if (!LangOpts.AsmPreprocessor) {
3865  if (uint32_t CodePoint = tryReadUCN(CurPtr, BufferPtr, &Result)) {
3866  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3867  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3868  return true; // KeepWhitespaceMode
3869 
3870  // We only saw whitespace, so just try again with this lexer.
3871  // (We manually eliminate the tail call to avoid recursion.)
3872  goto LexNextToken;
3873  }
3874 
3875  return LexUnicode(Result, CodePoint, CurPtr);
3876  }
3877  }
3878 
3879  Kind = tok::unknown;
3880  break;
3881 
3882  default: {
3883  if (isASCII(Char)) {
3884  Kind = tok::unknown;
3885  break;
3886  }
3887 
3888  llvm::UTF32 CodePoint;
3889 
3890  // We can't just reset CurPtr to BufferPtr because BufferPtr may point to
3891  // an escaped newline.
3892  --CurPtr;
3893  llvm::ConversionResult Status =
3894  llvm::convertUTF8Sequence((const llvm::UTF8 **)&CurPtr,
3895  (const llvm::UTF8 *)BufferEnd,
3896  &CodePoint,
3897  llvm::strictConversion);
3898  if (Status == llvm::conversionOK) {
3899  if (CheckUnicodeWhitespace(Result, CodePoint, CurPtr)) {
3900  if (SkipWhitespace(Result, CurPtr, TokAtPhysicalStartOfLine))
3901  return true; // KeepWhitespaceMode
3902 
3903  // We only saw whitespace, so just try again with this lexer.
3904  // (We manually eliminate the tail call to avoid recursion.)
3905  goto LexNextToken;
3906  }
3907  return LexUnicode(Result, CodePoint, CurPtr);
3908  }
3909 
3911  PP->isPreprocessedOutput()) {
3912  ++CurPtr;
3913  Kind = tok::unknown;
3914  break;
3915  }
3916 
3917  // Non-ASCII characters tend to creep into source code unintentionally.
3918  // Instead of letting the parser complain about the unknown token,
3919  // just diagnose the invalid UTF-8, then drop the character.
3920  Diag(CurPtr, diag::err_invalid_utf8);
3921 
3922  BufferPtr = CurPtr+1;
3923  // We're pretending the character didn't exist, so just try again with
3924  // this lexer.
3925  // (We manually eliminate the tail call to avoid recursion.)
3926  goto LexNextToken;
3927  }
3928  }
3929 
3930  // Notify MIOpt that we read a non-whitespace/non-comment token.
3931  MIOpt.ReadToken();
3932 
3933  // Update the location of token as well as BufferPtr.
3934  FormTokenWithChars(Result, CurPtr, Kind);
3935  return true;
3936 
3937 HandleDirective:
3938  // We parsed a # character and it's the start of a preprocessing directive.
3939 
3940  FormTokenWithChars(Result, CurPtr, tok::hash);
3941  PP->HandleDirective(Result);
3942 
3943  if (PP->hadModuleLoaderFatalFailure()) {
3944  // With a fatal failure in the module loader, we abort parsing.
3945  assert(Result.is(tok::eof) && "Preprocessor did not set tok:eof");
3946  return true;
3947  }
3948 
3949  // We parsed the directive; lex a token with the new state.
3950  return false;
3951 }
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:1290
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:928
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:49
bool isPragmaLexer() const
isPragmaLexer - Returns true if this Lexer is being used to lex a pragma.
Definition: Lexer.h:183
void resetExtendedTokenMode()
Sets the extended token mode back to its initial value, according to the language options and preproc...
Definition: Lexer.cpp:163
A Perforce-style conflict marker, initiated by 4 ">"s, separated by 4 "="s, and terminated by 4 "<"s...
Definition: Lexer.h:52
bool isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const
Return true if we have an ObjC keyword identifier.
Definition: Lexer.cpp:57
SourceLocation getSourceLocation() override
getSourceLocation - Return a source location for the next character in the current file...
Definition: Lexer.h:261
bool isInvalid() const
bool isAtEndOfImmediateMacroExpansion(SourceLocation Loc, SourceLocation *MacroEnd=nullptr) const
Returns true if the given MacroID location points at the character end of the immediate macro expansi...
static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_set1_epi8(char __b)
Initializes all values in a 128-bit vector of [16 x i8] with the specified 8-bit value.
Definition: emmintrin.h:3842
static SourceLocation getBeginningOfFileToken(SourceLocation Loc, const SourceManager &SM, const LangOptions &LangOpts)
Definition: Lexer.cpp: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:2413
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:1466
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:1043
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:1809
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:943
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:1667
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:908
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:1511
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:911
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:1472
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:2922
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:1449
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:899
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:2655
__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:901
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:2818
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:819
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:964
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:125
bool LexEditorPlaceholders
When enabled, the preprocessor will construct editor placeholder tokens.