clang  6.0.0svn
LiteralSupport.cpp
Go to the documentation of this file.
1 //===--- LiteralSupport.cpp - Code to parse and process literals ----------===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the NumericLiteralParser, CharLiteralParser, and
11 // StringLiteralParser interfaces.
12 //
13 //===----------------------------------------------------------------------===//
14 
16 #include "clang/Basic/CharInfo.h"
19 #include "clang/Basic/TargetInfo.h"
21 #include "clang/Lex/Lexer.h"
22 #include "clang/Lex/Preprocessor.h"
23 #include "clang/Lex/Token.h"
24 #include "llvm/ADT/APInt.h"
25 #include "llvm/ADT/SmallVector.h"
26 #include "llvm/ADT/StringExtras.h"
27 #include "llvm/ADT/StringSwitch.h"
28 #include "llvm/Support/ConvertUTF.h"
29 #include "llvm/Support/ErrorHandling.h"
30 #include <algorithm>
31 #include <cassert>
32 #include <cstddef>
33 #include <cstdint>
34 #include <cstring>
35 #include <string>
36 
37 using namespace clang;
38 
39 static unsigned getCharWidth(tok::TokenKind kind, const TargetInfo &Target) {
40  switch (kind) {
41  default: llvm_unreachable("Unknown token type!");
42  case tok::char_constant:
43  case tok::string_literal:
44  case tok::utf8_char_constant:
45  case tok::utf8_string_literal:
46  return Target.getCharWidth();
47  case tok::wide_char_constant:
48  case tok::wide_string_literal:
49  return Target.getWCharWidth();
50  case tok::utf16_char_constant:
51  case tok::utf16_string_literal:
52  return Target.getChar16Width();
53  case tok::utf32_char_constant:
54  case tok::utf32_string_literal:
55  return Target.getChar32Width();
56  }
57 }
58 
60  FullSourceLoc TokLoc,
61  const char *TokBegin,
62  const char *TokRangeBegin,
63  const char *TokRangeEnd) {
65  Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
66  TokLoc.getManager(), Features);
68  Lexer::AdvanceToTokenCharacter(Begin, TokRangeEnd - TokRangeBegin,
69  TokLoc.getManager(), Features);
70  return CharSourceRange::getCharRange(Begin, End);
71 }
72 
73 /// \brief Produce a diagnostic highlighting some portion of a literal.
74 ///
75 /// Emits the diagnostic \p DiagID, highlighting the range of characters from
76 /// \p TokRangeBegin (inclusive) to \p TokRangeEnd (exclusive), which must be
77 /// a substring of a spelling buffer for the token beginning at \p TokBegin.
79  const LangOptions &Features, FullSourceLoc TokLoc,
80  const char *TokBegin, const char *TokRangeBegin,
81  const char *TokRangeEnd, unsigned DiagID) {
83  Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
84  TokLoc.getManager(), Features);
85  return Diags->Report(Begin, DiagID) <<
86  MakeCharSourceRange(Features, TokLoc, TokBegin, TokRangeBegin, TokRangeEnd);
87 }
88 
89 /// ProcessCharEscape - Parse a standard C escape sequence, which can occur in
90 /// either a character or a string literal.
91 static unsigned ProcessCharEscape(const char *ThisTokBegin,
92  const char *&ThisTokBuf,
93  const char *ThisTokEnd, bool &HadError,
94  FullSourceLoc Loc, unsigned CharWidth,
95  DiagnosticsEngine *Diags,
96  const LangOptions &Features) {
97  const char *EscapeBegin = ThisTokBuf;
98 
99  // Skip the '\' char.
100  ++ThisTokBuf;
101 
102  // We know that this character can't be off the end of the buffer, because
103  // that would have been \", which would not have been the end of string.
104  unsigned ResultChar = *ThisTokBuf++;
105  switch (ResultChar) {
106  // These map to themselves.
107  case '\\': case '\'': case '"': case '?': break;
108 
109  // These have fixed mappings.
110  case 'a':
111  // TODO: K&R: the meaning of '\\a' is different in traditional C
112  ResultChar = 7;
113  break;
114  case 'b':
115  ResultChar = 8;
116  break;
117  case 'e':
118  if (Diags)
119  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
120  diag::ext_nonstandard_escape) << "e";
121  ResultChar = 27;
122  break;
123  case 'E':
124  if (Diags)
125  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
126  diag::ext_nonstandard_escape) << "E";
127  ResultChar = 27;
128  break;
129  case 'f':
130  ResultChar = 12;
131  break;
132  case 'n':
133  ResultChar = 10;
134  break;
135  case 'r':
136  ResultChar = 13;
137  break;
138  case 't':
139  ResultChar = 9;
140  break;
141  case 'v':
142  ResultChar = 11;
143  break;
144  case 'x': { // Hex escape.
145  ResultChar = 0;
146  if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
147  if (Diags)
148  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
149  diag::err_hex_escape_no_digits) << "x";
150  HadError = true;
151  break;
152  }
153 
154  // Hex escapes are a maximal series of hex digits.
155  bool Overflow = false;
156  for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) {
157  int CharVal = llvm::hexDigitValue(ThisTokBuf[0]);
158  if (CharVal == -1) break;
159  // About to shift out a digit?
160  if (ResultChar & 0xF0000000)
161  Overflow = true;
162  ResultChar <<= 4;
163  ResultChar |= CharVal;
164  }
165 
166  // See if any bits will be truncated when evaluated as a character.
167  if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
168  Overflow = true;
169  ResultChar &= ~0U >> (32-CharWidth);
170  }
171 
172  // Check for overflow.
173  if (Overflow && Diags) // Too many digits to fit in
174  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
175  diag::err_escape_too_large) << 0;
176  break;
177  }
178  case '0': case '1': case '2': case '3':
179  case '4': case '5': case '6': case '7': {
180  // Octal escapes.
181  --ThisTokBuf;
182  ResultChar = 0;
183 
184  // Octal escapes are a series of octal digits with maximum length 3.
185  // "\0123" is a two digit sequence equal to "\012" "3".
186  unsigned NumDigits = 0;
187  do {
188  ResultChar <<= 3;
189  ResultChar |= *ThisTokBuf++ - '0';
190  ++NumDigits;
191  } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 &&
192  ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7');
193 
194  // Check for overflow. Reject '\777', but not L'\777'.
195  if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
196  if (Diags)
197  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
198  diag::err_escape_too_large) << 1;
199  ResultChar &= ~0U >> (32-CharWidth);
200  }
201  break;
202  }
203 
204  // Otherwise, these are not valid escapes.
205  case '(': case '{': case '[': case '%':
206  // GCC accepts these as extensions. We warn about them as such though.
207  if (Diags)
208  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
209  diag::ext_nonstandard_escape)
210  << std::string(1, ResultChar);
211  break;
212  default:
213  if (!Diags)
214  break;
215 
216  if (isPrintable(ResultChar))
217  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
218  diag::ext_unknown_escape)
219  << std::string(1, ResultChar);
220  else
221  Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
222  diag::ext_unknown_escape)
223  << "x" + llvm::utohexstr(ResultChar);
224  break;
225  }
226 
227  return ResultChar;
228 }
229 
230 static void appendCodePoint(unsigned Codepoint,
232  char ResultBuf[4];
233  char *ResultPtr = ResultBuf;
234  bool Res = llvm::ConvertCodePointToUTF8(Codepoint, ResultPtr);
235  (void)Res;
236  assert(Res && "Unexpected conversion failure");
237  Str.append(ResultBuf, ResultPtr);
238 }
239 
240 void clang::expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
241  for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
242  if (*I != '\\') {
243  Buf.push_back(*I);
244  continue;
245  }
246 
247  ++I;
248  assert(*I == 'u' || *I == 'U');
249 
250  unsigned NumHexDigits;
251  if (*I == 'u')
252  NumHexDigits = 4;
253  else
254  NumHexDigits = 8;
255 
256  assert(I + NumHexDigits <= E);
257 
258  uint32_t CodePoint = 0;
259  for (++I; NumHexDigits != 0; ++I, --NumHexDigits) {
260  unsigned Value = llvm::hexDigitValue(*I);
261  assert(Value != -1U);
262 
263  CodePoint <<= 4;
264  CodePoint += Value;
265  }
266 
267  appendCodePoint(CodePoint, Buf);
268  --I;
269  }
270 }
271 
272 /// ProcessUCNEscape - Read the Universal Character Name, check constraints and
273 /// return the UTF32.
274 static bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
275  const char *ThisTokEnd,
276  uint32_t &UcnVal, unsigned short &UcnLen,
277  FullSourceLoc Loc, DiagnosticsEngine *Diags,
278  const LangOptions &Features,
279  bool in_char_string_literal = false) {
280  const char *UcnBegin = ThisTokBuf;
281 
282  // Skip the '\u' char's.
283  ThisTokBuf += 2;
284 
285  if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
286  if (Diags)
287  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
288  diag::err_hex_escape_no_digits) << StringRef(&ThisTokBuf[-1], 1);
289  return false;
290  }
291  UcnLen = (ThisTokBuf[-1] == 'u' ? 4 : 8);
292  unsigned short UcnLenSave = UcnLen;
293  for (; ThisTokBuf != ThisTokEnd && UcnLenSave; ++ThisTokBuf, UcnLenSave--) {
294  int CharVal = llvm::hexDigitValue(ThisTokBuf[0]);
295  if (CharVal == -1) break;
296  UcnVal <<= 4;
297  UcnVal |= CharVal;
298  }
299  // If we didn't consume the proper number of digits, there is a problem.
300  if (UcnLenSave) {
301  if (Diags)
302  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
303  diag::err_ucn_escape_incomplete);
304  return false;
305  }
306 
307  // Check UCN constraints (C99 6.4.3p2) [C++11 lex.charset p2]
308  if ((0xD800 <= UcnVal && UcnVal <= 0xDFFF) || // surrogate codepoints
309  UcnVal > 0x10FFFF) { // maximum legal UTF32 value
310  if (Diags)
311  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
312  diag::err_ucn_escape_invalid);
313  return false;
314  }
315 
316  // C++11 allows UCNs that refer to control characters and basic source
317  // characters inside character and string literals
318  if (UcnVal < 0xa0 &&
319  (UcnVal != 0x24 && UcnVal != 0x40 && UcnVal != 0x60)) { // $, @, `
320  bool IsError = (!Features.CPlusPlus11 || !in_char_string_literal);
321  if (Diags) {
322  char BasicSCSChar = UcnVal;
323  if (UcnVal >= 0x20 && UcnVal < 0x7f)
324  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
325  IsError ? diag::err_ucn_escape_basic_scs :
326  diag::warn_cxx98_compat_literal_ucn_escape_basic_scs)
327  << StringRef(&BasicSCSChar, 1);
328  else
329  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
330  IsError ? diag::err_ucn_control_character :
331  diag::warn_cxx98_compat_literal_ucn_control_character);
332  }
333  if (IsError)
334  return false;
335  }
336 
337  if (!Features.CPlusPlus && !Features.C99 && Diags)
338  Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
339  diag::warn_ucn_not_valid_in_c89_literal);
340 
341  return true;
342 }
343 
344 /// MeasureUCNEscape - Determine the number of bytes within the resulting string
345 /// which this UCN will occupy.
346 static int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
347  const char *ThisTokEnd, unsigned CharByteWidth,
348  const LangOptions &Features, bool &HadError) {
349  // UTF-32: 4 bytes per escape.
350  if (CharByteWidth == 4)
351  return 4;
352 
353  uint32_t UcnVal = 0;
354  unsigned short UcnLen = 0;
355  FullSourceLoc Loc;
356 
357  if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
358  UcnLen, Loc, nullptr, Features, true)) {
359  HadError = true;
360  return 0;
361  }
362 
363  // UTF-16: 2 bytes for BMP, 4 bytes otherwise.
364  if (CharByteWidth == 2)
365  return UcnVal <= 0xFFFF ? 2 : 4;
366 
367  // UTF-8.
368  if (UcnVal < 0x80)
369  return 1;
370  if (UcnVal < 0x800)
371  return 2;
372  if (UcnVal < 0x10000)
373  return 3;
374  return 4;
375 }
376 
377 /// EncodeUCNEscape - Read the Universal Character Name, check constraints and
378 /// convert the UTF32 to UTF8 or UTF16. This is a subroutine of
379 /// StringLiteralParser. When we decide to implement UCN's for identifiers,
380 /// we will likely rework our support for UCN's.
381 static void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
382  const char *ThisTokEnd,
383  char *&ResultBuf, bool &HadError,
384  FullSourceLoc Loc, unsigned CharByteWidth,
385  DiagnosticsEngine *Diags,
386  const LangOptions &Features) {
387  typedef uint32_t UTF32;
388  UTF32 UcnVal = 0;
389  unsigned short UcnLen = 0;
390  if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, UcnLen,
391  Loc, Diags, Features, true)) {
392  HadError = true;
393  return;
394  }
395 
396  assert((CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) &&
397  "only character widths of 1, 2, or 4 bytes supported");
398 
399  (void)UcnLen;
400  assert((UcnLen== 4 || UcnLen== 8) && "only ucn length of 4 or 8 supported");
401 
402  if (CharByteWidth == 4) {
403  // FIXME: Make the type of the result buffer correct instead of
404  // using reinterpret_cast.
405  llvm::UTF32 *ResultPtr = reinterpret_cast<llvm::UTF32*>(ResultBuf);
406  *ResultPtr = UcnVal;
407  ResultBuf += 4;
408  return;
409  }
410 
411  if (CharByteWidth == 2) {
412  // FIXME: Make the type of the result buffer correct instead of
413  // using reinterpret_cast.
414  llvm::UTF16 *ResultPtr = reinterpret_cast<llvm::UTF16*>(ResultBuf);
415 
416  if (UcnVal <= (UTF32)0xFFFF) {
417  *ResultPtr = UcnVal;
418  ResultBuf += 2;
419  return;
420  }
421 
422  // Convert to UTF16.
423  UcnVal -= 0x10000;
424  *ResultPtr = 0xD800 + (UcnVal >> 10);
425  *(ResultPtr+1) = 0xDC00 + (UcnVal & 0x3FF);
426  ResultBuf += 4;
427  return;
428  }
429 
430  assert(CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters");
431 
432  // Now that we've parsed/checked the UCN, we convert from UTF32->UTF8.
433  // The conversion below was inspired by:
434  // http://www.unicode.org/Public/PROGRAMS/CVTUTF/ConvertUTF.c
435  // First, we determine how many bytes the result will require.
436  typedef uint8_t UTF8;
437 
438  unsigned short bytesToWrite = 0;
439  if (UcnVal < (UTF32)0x80)
440  bytesToWrite = 1;
441  else if (UcnVal < (UTF32)0x800)
442  bytesToWrite = 2;
443  else if (UcnVal < (UTF32)0x10000)
444  bytesToWrite = 3;
445  else
446  bytesToWrite = 4;
447 
448  const unsigned byteMask = 0xBF;
449  const unsigned byteMark = 0x80;
450 
451  // Once the bits are split out into bytes of UTF8, this is a mask OR-ed
452  // into the first byte, depending on how many bytes follow.
453  static const UTF8 firstByteMark[5] = {
454  0x00, 0x00, 0xC0, 0xE0, 0xF0
455  };
456  // Finally, we write the bytes into ResultBuf.
457  ResultBuf += bytesToWrite;
458  switch (bytesToWrite) { // note: everything falls through.
459  case 4:
460  *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
461  LLVM_FALLTHROUGH;
462  case 3:
463  *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
464  LLVM_FALLTHROUGH;
465  case 2:
466  *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
467  LLVM_FALLTHROUGH;
468  case 1:
469  *--ResultBuf = (UTF8) (UcnVal | firstByteMark[bytesToWrite]);
470  }
471  // Update the buffer.
472  ResultBuf += bytesToWrite;
473 }
474 
475 /// integer-constant: [C99 6.4.4.1]
476 /// decimal-constant integer-suffix
477 /// octal-constant integer-suffix
478 /// hexadecimal-constant integer-suffix
479 /// binary-literal integer-suffix [GNU, C++1y]
480 /// user-defined-integer-literal: [C++11 lex.ext]
481 /// decimal-literal ud-suffix
482 /// octal-literal ud-suffix
483 /// hexadecimal-literal ud-suffix
484 /// binary-literal ud-suffix [GNU, C++1y]
485 /// decimal-constant:
486 /// nonzero-digit
487 /// decimal-constant digit
488 /// octal-constant:
489 /// 0
490 /// octal-constant octal-digit
491 /// hexadecimal-constant:
492 /// hexadecimal-prefix hexadecimal-digit
493 /// hexadecimal-constant hexadecimal-digit
494 /// hexadecimal-prefix: one of
495 /// 0x 0X
496 /// binary-literal:
497 /// 0b binary-digit
498 /// 0B binary-digit
499 /// binary-literal binary-digit
500 /// integer-suffix:
501 /// unsigned-suffix [long-suffix]
502 /// unsigned-suffix [long-long-suffix]
503 /// long-suffix [unsigned-suffix]
504 /// long-long-suffix [unsigned-sufix]
505 /// nonzero-digit:
506 /// 1 2 3 4 5 6 7 8 9
507 /// octal-digit:
508 /// 0 1 2 3 4 5 6 7
509 /// hexadecimal-digit:
510 /// 0 1 2 3 4 5 6 7 8 9
511 /// a b c d e f
512 /// A B C D E F
513 /// binary-digit:
514 /// 0
515 /// 1
516 /// unsigned-suffix: one of
517 /// u U
518 /// long-suffix: one of
519 /// l L
520 /// long-long-suffix: one of
521 /// ll LL
522 ///
523 /// floating-constant: [C99 6.4.4.2]
524 /// TODO: add rules...
525 ///
527  SourceLocation TokLoc,
528  Preprocessor &PP)
529  : PP(PP), ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) {
530 
531  // This routine assumes that the range begin/end matches the regex for integer
532  // and FP constants (specifically, the 'pp-number' regex), and assumes that
533  // the byte at "*end" is both valid and not part of the regex. Because of
534  // this, it doesn't have to check for 'overscan' in various places.
535  assert(!isPreprocessingNumberBody(*ThisTokEnd) && "didn't maximally munch?");
536 
537  s = DigitsBegin = ThisTokBegin;
538  saw_exponent = false;
539  saw_period = false;
540  saw_ud_suffix = false;
541  isLong = false;
542  isUnsigned = false;
543  isLongLong = false;
544  isHalf = false;
545  isFloat = false;
546  isImaginary = false;
547  isFloat16 = false;
548  isFloat128 = false;
549  MicrosoftInteger = 0;
550  hadError = false;
551 
552  if (*s == '0') { // parse radix
553  ParseNumberStartingWithZero(TokLoc);
554  if (hadError)
555  return;
556  } else { // the first digit is non-zero
557  radix = 10;
558  s = SkipDigits(s);
559  if (s == ThisTokEnd) {
560  // Done.
561  } else {
562  ParseDecimalOrOctalCommon(TokLoc);
563  if (hadError)
564  return;
565  }
566  }
567 
568  SuffixBegin = s;
569  checkSeparator(TokLoc, s, CSK_AfterDigits);
570 
571  // Parse the suffix. At this point we can classify whether we have an FP or
572  // integer constant.
573  bool isFPConstant = isFloatingLiteral();
574 
575  // Loop over all of the characters of the suffix. If we see something bad,
576  // we break out of the loop.
577  for (; s != ThisTokEnd; ++s) {
578  switch (*s) {
579  case 'h': // FP Suffix for "half".
580  case 'H':
581  // OpenCL Extension v1.2 s9.5 - h or H suffix for half type.
582  if (!PP.getLangOpts().Half) break;
583  if (!isFPConstant) break; // Error for integer constant.
584  if (isHalf || isFloat || isLong) break; // HH, FH, LH invalid.
585  isHalf = true;
586  continue; // Success.
587  case 'f': // FP Suffix for "float"
588  case 'F':
589  if (!isFPConstant) break; // Error for integer constant.
590  if (isHalf || isFloat || isLong || isFloat128)
591  break; // HF, FF, LF, QF invalid.
592 
593  if (s + 2 < ThisTokEnd && s[1] == '1' && s[2] == '6') {
594  s += 2; // success, eat up 2 characters.
595  isFloat16 = true;
596  continue;
597  }
598 
599  isFloat = true;
600  continue; // Success.
601  case 'q': // FP Suffix for "__float128"
602  case 'Q':
603  if (!isFPConstant) break; // Error for integer constant.
604  if (isHalf || isFloat || isLong || isFloat128)
605  break; // HQ, FQ, LQ, QQ invalid.
606  isFloat128 = true;
607  continue; // Success.
608  case 'u':
609  case 'U':
610  if (isFPConstant) break; // Error for floating constant.
611  if (isUnsigned) break; // Cannot be repeated.
612  isUnsigned = true;
613  continue; // Success.
614  case 'l':
615  case 'L':
616  if (isLong || isLongLong) break; // Cannot be repeated.
617  if (isHalf || isFloat || isFloat128) break; // LH, LF, LQ invalid.
618 
619  // Check for long long. The L's need to be adjacent and the same case.
620  if (s[1] == s[0]) {
621  assert(s + 1 < ThisTokEnd && "didn't maximally munch?");
622  if (isFPConstant) break; // long long invalid for floats.
623  isLongLong = true;
624  ++s; // Eat both of them.
625  } else {
626  isLong = true;
627  }
628  continue; // Success.
629  case 'i':
630  case 'I':
631  if (PP.getLangOpts().MicrosoftExt) {
633  break;
634 
635  if (!isFPConstant) {
636  // Allow i8, i16, i32, and i64.
637  switch (s[1]) {
638  case '8':
639  s += 2; // i8 suffix
640  MicrosoftInteger = 8;
641  break;
642  case '1':
643  if (s[2] == '6') {
644  s += 3; // i16 suffix
645  MicrosoftInteger = 16;
646  }
647  break;
648  case '3':
649  if (s[2] == '2') {
650  s += 3; // i32 suffix
651  MicrosoftInteger = 32;
652  }
653  break;
654  case '6':
655  if (s[2] == '4') {
656  s += 3; // i64 suffix
657  MicrosoftInteger = 64;
658  }
659  break;
660  default:
661  break;
662  }
663  }
664  if (MicrosoftInteger) {
665  assert(s <= ThisTokEnd && "didn't maximally munch?");
666  break;
667  }
668  }
669  // fall through.
670  case 'j':
671  case 'J':
672  if (isImaginary) break; // Cannot be repeated.
673  isImaginary = true;
674  continue; // Success.
675  }
676  // If we reached here, there was an error or a ud-suffix.
677  break;
678  }
679 
680  // "i", "if", and "il" are user-defined suffixes in C++1y.
681  if (s != ThisTokEnd || isImaginary) {
682  // FIXME: Don't bother expanding UCNs if !tok.hasUCN().
683  expandUCNs(UDSuffixBuf, StringRef(SuffixBegin, ThisTokEnd - SuffixBegin));
684  if (isValidUDSuffix(PP.getLangOpts(), UDSuffixBuf)) {
685  if (!isImaginary) {
686  // Any suffix pieces we might have parsed are actually part of the
687  // ud-suffix.
688  isLong = false;
689  isUnsigned = false;
690  isLongLong = false;
691  isFloat = false;
692  isFloat16 = false;
693  isHalf = false;
694  isImaginary = false;
695  MicrosoftInteger = 0;
696  }
697 
698  saw_ud_suffix = true;
699  return;
700  }
701 
702  if (s != ThisTokEnd) {
703  // Report an error if there are any.
704  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, SuffixBegin - ThisTokBegin),
705  diag::err_invalid_suffix_constant)
706  << StringRef(SuffixBegin, ThisTokEnd - SuffixBegin) << isFPConstant;
707  hadError = true;
708  }
709  }
710 }
711 
712 /// ParseDecimalOrOctalCommon - This method is called for decimal or octal
713 /// numbers. It issues an error for illegal digits, and handles floating point
714 /// parsing. If it detects a floating point number, the radix is set to 10.
715 void NumericLiteralParser::ParseDecimalOrOctalCommon(SourceLocation TokLoc){
716  assert((radix == 8 || radix == 10) && "Unexpected radix");
717 
718  // If we have a hex digit other than 'e' (which denotes a FP exponent) then
719  // the code is using an incorrect base.
720  if (isHexDigit(*s) && *s != 'e' && *s != 'E') {
721  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin),
722  diag::err_invalid_digit) << StringRef(s, 1) << (radix == 8 ? 1 : 0);
723  hadError = true;
724  return;
725  }
726 
727  if (*s == '.') {
728  checkSeparator(TokLoc, s, CSK_AfterDigits);
729  s++;
730  radix = 10;
731  saw_period = true;
732  checkSeparator(TokLoc, s, CSK_BeforeDigits);
733  s = SkipDigits(s); // Skip suffix.
734  }
735  if (*s == 'e' || *s == 'E') { // exponent
736  checkSeparator(TokLoc, s, CSK_AfterDigits);
737  const char *Exponent = s;
738  s++;
739  radix = 10;
740  saw_exponent = true;
741  if (*s == '+' || *s == '-') s++; // sign
742  const char *first_non_digit = SkipDigits(s);
743  if (containsDigits(s, first_non_digit)) {
744  checkSeparator(TokLoc, s, CSK_BeforeDigits);
745  s = first_non_digit;
746  } else {
747  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin),
748  diag::err_exponent_has_no_digits);
749  hadError = true;
750  return;
751  }
752  }
753 }
754 
755 /// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
756 /// suffixes as ud-suffixes, because the diagnostic experience is better if we
757 /// treat it as an invalid suffix.
759  StringRef Suffix) {
760  if (!LangOpts.CPlusPlus11 || Suffix.empty())
761  return false;
762 
763  // By C++11 [lex.ext]p10, ud-suffixes starting with an '_' are always valid.
764  if (Suffix[0] == '_')
765  return true;
766 
767  // In C++11, there are no library suffixes.
768  if (!LangOpts.CPlusPlus14)
769  return false;
770 
771  // In C++1y, "s", "h", "min", "ms", "us", and "ns" are used in the library.
772  // Per tweaked N3660, "il", "i", and "if" are also used in the library.
773  return llvm::StringSwitch<bool>(Suffix)
774  .Cases("h", "min", "s", true)
775  .Cases("ms", "us", "ns", true)
776  .Cases("il", "i", "if", true)
777  .Default(false);
778 }
779 
780 void NumericLiteralParser::checkSeparator(SourceLocation TokLoc,
781  const char *Pos,
782  CheckSeparatorKind IsAfterDigits) {
783  if (IsAfterDigits == CSK_AfterDigits) {
784  if (Pos == ThisTokBegin)
785  return;
786  --Pos;
787  } else if (Pos == ThisTokEnd)
788  return;
789 
790  if (isDigitSeparator(*Pos))
791  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Pos - ThisTokBegin),
792  diag::err_digit_separator_not_between_digits)
793  << IsAfterDigits;
794 }
795 
796 /// ParseNumberStartingWithZero - This method is called when the first character
797 /// of the number is found to be a zero. This means it is either an octal
798 /// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or
799 /// a floating point number (01239.123e4). Eat the prefix, determining the
800 /// radix etc.
801 void NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) {
802  assert(s[0] == '0' && "Invalid method call");
803  s++;
804 
805  int c1 = s[0];
806 
807  // Handle a hex number like 0x1234.
808  if ((c1 == 'x' || c1 == 'X') && (isHexDigit(s[1]) || s[1] == '.')) {
809  s++;
810  assert(s < ThisTokEnd && "didn't maximally munch?");
811  radix = 16;
812  DigitsBegin = s;
813  s = SkipHexDigits(s);
814  bool HasSignificandDigits = containsDigits(DigitsBegin, s);
815  if (s == ThisTokEnd) {
816  // Done.
817  } else if (*s == '.') {
818  s++;
819  saw_period = true;
820  const char *floatDigitsBegin = s;
821  s = SkipHexDigits(s);
822  if (containsDigits(floatDigitsBegin, s))
823  HasSignificandDigits = true;
824  if (HasSignificandDigits)
825  checkSeparator(TokLoc, floatDigitsBegin, CSK_BeforeDigits);
826  }
827 
828  if (!HasSignificandDigits) {
829  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin),
830  diag::err_hex_constant_requires)
831  << PP.getLangOpts().CPlusPlus << 1;
832  hadError = true;
833  return;
834  }
835 
836  // A binary exponent can appear with or with a '.'. If dotted, the
837  // binary exponent is required.
838  if (*s == 'p' || *s == 'P') {
839  checkSeparator(TokLoc, s, CSK_AfterDigits);
840  const char *Exponent = s;
841  s++;
842  saw_exponent = true;
843  if (*s == '+' || *s == '-') s++; // sign
844  const char *first_non_digit = SkipDigits(s);
845  if (!containsDigits(s, first_non_digit)) {
846  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin),
847  diag::err_exponent_has_no_digits);
848  hadError = true;
849  return;
850  }
851  checkSeparator(TokLoc, s, CSK_BeforeDigits);
852  s = first_non_digit;
853 
854  if (!PP.getLangOpts().HexFloats)
855  PP.Diag(TokLoc, PP.getLangOpts().CPlusPlus
856  ? diag::ext_hex_literal_invalid
857  : diag::ext_hex_constant_invalid);
858  else if (PP.getLangOpts().CPlusPlus1z)
859  PP.Diag(TokLoc, diag::warn_cxx17_hex_literal);
860  } else if (saw_period) {
861  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin),
862  diag::err_hex_constant_requires)
863  << PP.getLangOpts().CPlusPlus << 0;
864  hadError = true;
865  }
866  return;
867  }
868 
869  // Handle simple binary numbers 0b01010
870  if ((c1 == 'b' || c1 == 'B') && (s[1] == '0' || s[1] == '1')) {
871  // 0b101010 is a C++1y / GCC extension.
872  PP.Diag(TokLoc,
873  PP.getLangOpts().CPlusPlus14
874  ? diag::warn_cxx11_compat_binary_literal
875  : PP.getLangOpts().CPlusPlus
876  ? diag::ext_binary_literal_cxx14
877  : diag::ext_binary_literal);
878  ++s;
879  assert(s < ThisTokEnd && "didn't maximally munch?");
880  radix = 2;
881  DigitsBegin = s;
882  s = SkipBinaryDigits(s);
883  if (s == ThisTokEnd) {
884  // Done.
885  } else if (isHexDigit(*s)) {
886  PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin),
887  diag::err_invalid_digit) << StringRef(s, 1) << 2;
888  hadError = true;
889  }
890  // Other suffixes will be diagnosed by the caller.
891  return;
892  }
893 
894  // For now, the radix is set to 8. If we discover that we have a
895  // floating point constant, the radix will change to 10. Octal floating
896  // point constants are not permitted (only decimal and hexadecimal).
897  radix = 8;
898  DigitsBegin = s;
899  s = SkipOctalDigits(s);
900  if (s == ThisTokEnd)
901  return; // Done, simple octal number like 01234
902 
903  // If we have some other non-octal digit that *is* a decimal digit, see if
904  // this is part of a floating point number like 094.123 or 09e1.
905  if (isDigit(*s)) {
906  const char *EndDecimal = SkipDigits(s);
907  if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') {
908  s = EndDecimal;
909  radix = 10;
910  }
911  }
912 
913  ParseDecimalOrOctalCommon(TokLoc);
914 }
915 
916 static bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits) {
917  switch (Radix) {
918  case 2:
919  return NumDigits <= 64;
920  case 8:
921  return NumDigits <= 64 / 3; // Digits are groups of 3 bits.
922  case 10:
923  return NumDigits <= 19; // floor(log10(2^64))
924  case 16:
925  return NumDigits <= 64 / 4; // Digits are groups of 4 bits.
926  default:
927  llvm_unreachable("impossible Radix");
928  }
929 }
930 
931 /// GetIntegerValue - Convert this numeric literal value to an APInt that
932 /// matches Val's input width. If there is an overflow, set Val to the low bits
933 /// of the result and return true. Otherwise, return false.
935  // Fast path: Compute a conservative bound on the maximum number of
936  // bits per digit in this radix. If we can't possibly overflow a
937  // uint64 based on that bound then do the simple conversion to
938  // integer. This avoids the expensive overflow checking below, and
939  // handles the common cases that matter (small decimal integers and
940  // hex/octal values which don't overflow).
941  const unsigned NumDigits = SuffixBegin - DigitsBegin;
942  if (alwaysFitsInto64Bits(radix, NumDigits)) {
943  uint64_t N = 0;
944  for (const char *Ptr = DigitsBegin; Ptr != SuffixBegin; ++Ptr)
945  if (!isDigitSeparator(*Ptr))
946  N = N * radix + llvm::hexDigitValue(*Ptr);
947 
948  // This will truncate the value to Val's input width. Simply check
949  // for overflow by comparing.
950  Val = N;
951  return Val.getZExtValue() != N;
952  }
953 
954  Val = 0;
955  const char *Ptr = DigitsBegin;
956 
957  llvm::APInt RadixVal(Val.getBitWidth(), radix);
958  llvm::APInt CharVal(Val.getBitWidth(), 0);
959  llvm::APInt OldVal = Val;
960 
961  bool OverflowOccurred = false;
962  while (Ptr < SuffixBegin) {
963  if (isDigitSeparator(*Ptr)) {
964  ++Ptr;
965  continue;
966  }
967 
968  unsigned C = llvm::hexDigitValue(*Ptr++);
969 
970  // If this letter is out of bound for this radix, reject it.
971  assert(C < radix && "NumericLiteralParser ctor should have rejected this");
972 
973  CharVal = C;
974 
975  // Add the digit to the value in the appropriate radix. If adding in digits
976  // made the value smaller, then this overflowed.
977  OldVal = Val;
978 
979  // Multiply by radix, did overflow occur on the multiply?
980  Val *= RadixVal;
981  OverflowOccurred |= Val.udiv(RadixVal) != OldVal;
982 
983  // Add value, did overflow occur on the value?
984  // (a + b) ult b <=> overflow
985  Val += CharVal;
986  OverflowOccurred |= Val.ult(CharVal);
987  }
988  return OverflowOccurred;
989 }
990 
991 llvm::APFloat::opStatus
993  using llvm::APFloat;
994 
995  unsigned n = std::min(SuffixBegin - ThisTokBegin, ThisTokEnd - ThisTokBegin);
996 
997  llvm::SmallString<16> Buffer;
998  StringRef Str(ThisTokBegin, n);
999  if (Str.find('\'') != StringRef::npos) {
1000  Buffer.reserve(n);
1001  std::remove_copy_if(Str.begin(), Str.end(), std::back_inserter(Buffer),
1002  &isDigitSeparator);
1003  Str = Buffer;
1004  }
1005 
1006  return Result.convertFromString(Str, APFloat::rmNearestTiesToEven);
1007 }
1008 
1009 /// \verbatim
1010 /// user-defined-character-literal: [C++11 lex.ext]
1011 /// character-literal ud-suffix
1012 /// ud-suffix:
1013 /// identifier
1014 /// character-literal: [C++11 lex.ccon]
1015 /// ' c-char-sequence '
1016 /// u' c-char-sequence '
1017 /// U' c-char-sequence '
1018 /// L' c-char-sequence '
1019 /// u8' c-char-sequence ' [C++1z lex.ccon]
1020 /// c-char-sequence:
1021 /// c-char
1022 /// c-char-sequence c-char
1023 /// c-char:
1024 /// any member of the source character set except the single-quote ',
1025 /// backslash \, or new-line character
1026 /// escape-sequence
1027 /// universal-character-name
1028 /// escape-sequence:
1029 /// simple-escape-sequence
1030 /// octal-escape-sequence
1031 /// hexadecimal-escape-sequence
1032 /// simple-escape-sequence:
1033 /// one of \' \" \? \\ \a \b \f \n \r \t \v
1034 /// octal-escape-sequence:
1035 /// \ octal-digit
1036 /// \ octal-digit octal-digit
1037 /// \ octal-digit octal-digit octal-digit
1038 /// hexadecimal-escape-sequence:
1039 /// \x hexadecimal-digit
1040 /// hexadecimal-escape-sequence hexadecimal-digit
1041 /// universal-character-name: [C++11 lex.charset]
1042 /// \u hex-quad
1043 /// \U hex-quad hex-quad
1044 /// hex-quad:
1045 /// hex-digit hex-digit hex-digit hex-digit
1046 /// \endverbatim
1047 ///
1048 CharLiteralParser::CharLiteralParser(const char *begin, const char *end,
1049  SourceLocation Loc, Preprocessor &PP,
1050  tok::TokenKind kind) {
1051  // At this point we know that the character matches the regex "(L|u|U)?'.*'".
1052  HadError = false;
1053 
1054  Kind = kind;
1055 
1056  const char *TokBegin = begin;
1057 
1058  // Skip over wide character determinant.
1059  if (Kind != tok::char_constant)
1060  ++begin;
1061  if (Kind == tok::utf8_char_constant)
1062  ++begin;
1063 
1064  // Skip over the entry quote.
1065  assert(begin[0] == '\'' && "Invalid token lexed");
1066  ++begin;
1067 
1068  // Remove an optional ud-suffix.
1069  if (end[-1] != '\'') {
1070  const char *UDSuffixEnd = end;
1071  do {
1072  --end;
1073  } while (end[-1] != '\'');
1074  // FIXME: Don't bother with this if !tok.hasUCN().
1075  expandUCNs(UDSuffixBuf, StringRef(end, UDSuffixEnd - end));
1076  UDSuffixOffset = end - TokBegin;
1077  }
1078 
1079  // Trim the ending quote.
1080  assert(end != begin && "Invalid token lexed");
1081  --end;
1082 
1083  // FIXME: The "Value" is an uint64_t so we can handle char literals of
1084  // up to 64-bits.
1085  // FIXME: This extensively assumes that 'char' is 8-bits.
1086  assert(PP.getTargetInfo().getCharWidth() == 8 &&
1087  "Assumes char is 8 bits");
1088  assert(PP.getTargetInfo().getIntWidth() <= 64 &&
1089  (PP.getTargetInfo().getIntWidth() & 7) == 0 &&
1090  "Assumes sizeof(int) on target is <= 64 and a multiple of char");
1091  assert(PP.getTargetInfo().getWCharWidth() <= 64 &&
1092  "Assumes sizeof(wchar) on target is <= 64");
1093 
1094  SmallVector<uint32_t, 4> codepoint_buffer;
1095  codepoint_buffer.resize(end - begin);
1096  uint32_t *buffer_begin = &codepoint_buffer.front();
1097  uint32_t *buffer_end = buffer_begin + codepoint_buffer.size();
1098 
1099  // Unicode escapes representing characters that cannot be correctly
1100  // represented in a single code unit are disallowed in character literals
1101  // by this implementation.
1102  uint32_t largest_character_for_kind;
1103  if (tok::wide_char_constant == Kind) {
1104  largest_character_for_kind =
1105  0xFFFFFFFFu >> (32-PP.getTargetInfo().getWCharWidth());
1106  } else if (tok::utf8_char_constant == Kind) {
1107  largest_character_for_kind = 0x7F;
1108  } else if (tok::utf16_char_constant == Kind) {
1109  largest_character_for_kind = 0xFFFF;
1110  } else if (tok::utf32_char_constant == Kind) {
1111  largest_character_for_kind = 0x10FFFF;
1112  } else {
1113  largest_character_for_kind = 0x7Fu;
1114  }
1115 
1116  while (begin != end) {
1117  // Is this a span of non-escape characters?
1118  if (begin[0] != '\\') {
1119  char const *start = begin;
1120  do {
1121  ++begin;
1122  } while (begin != end && *begin != '\\');
1123 
1124  char const *tmp_in_start = start;
1125  uint32_t *tmp_out_start = buffer_begin;
1126  llvm::ConversionResult res =
1127  llvm::ConvertUTF8toUTF32(reinterpret_cast<llvm::UTF8 const **>(&start),
1128  reinterpret_cast<llvm::UTF8 const *>(begin),
1129  &buffer_begin, buffer_end, llvm::strictConversion);
1130  if (res != llvm::conversionOK) {
1131  // If we see bad encoding for unprefixed character literals, warn and
1132  // simply copy the byte values, for compatibility with gcc and
1133  // older versions of clang.
1134  bool NoErrorOnBadEncoding = isAscii();
1135  unsigned Msg = diag::err_bad_character_encoding;
1136  if (NoErrorOnBadEncoding)
1137  Msg = diag::warn_bad_character_encoding;
1138  PP.Diag(Loc, Msg);
1139  if (NoErrorOnBadEncoding) {
1140  start = tmp_in_start;
1141  buffer_begin = tmp_out_start;
1142  for (; start != begin; ++start, ++buffer_begin)
1143  *buffer_begin = static_cast<uint8_t>(*start);
1144  } else {
1145  HadError = true;
1146  }
1147  } else {
1148  for (; tmp_out_start < buffer_begin; ++tmp_out_start) {
1149  if (*tmp_out_start > largest_character_for_kind) {
1150  HadError = true;
1151  PP.Diag(Loc, diag::err_character_too_large);
1152  }
1153  }
1154  }
1155 
1156  continue;
1157  }
1158  // Is this a Universal Character Name escape?
1159  if (begin[1] == 'u' || begin[1] == 'U') {
1160  unsigned short UcnLen = 0;
1161  if (!ProcessUCNEscape(TokBegin, begin, end, *buffer_begin, UcnLen,
1162  FullSourceLoc(Loc, PP.getSourceManager()),
1163  &PP.getDiagnostics(), PP.getLangOpts(), true)) {
1164  HadError = true;
1165  } else if (*buffer_begin > largest_character_for_kind) {
1166  HadError = true;
1167  PP.Diag(Loc, diag::err_character_too_large);
1168  }
1169 
1170  ++buffer_begin;
1171  continue;
1172  }
1173  unsigned CharWidth = getCharWidth(Kind, PP.getTargetInfo());
1174  uint64_t result =
1175  ProcessCharEscape(TokBegin, begin, end, HadError,
1176  FullSourceLoc(Loc,PP.getSourceManager()),
1177  CharWidth, &PP.getDiagnostics(), PP.getLangOpts());
1178  *buffer_begin++ = result;
1179  }
1180 
1181  unsigned NumCharsSoFar = buffer_begin - &codepoint_buffer.front();
1182 
1183  if (NumCharsSoFar > 1) {
1184  if (isWide())
1185  PP.Diag(Loc, diag::warn_extraneous_char_constant);
1186  else if (isAscii() && NumCharsSoFar == 4)
1187  PP.Diag(Loc, diag::ext_four_char_character_literal);
1188  else if (isAscii())
1189  PP.Diag(Loc, diag::ext_multichar_character_literal);
1190  else
1191  PP.Diag(Loc, diag::err_multichar_utf_character_literal);
1192  IsMultiChar = true;
1193  } else {
1194  IsMultiChar = false;
1195  }
1196 
1197  llvm::APInt LitVal(PP.getTargetInfo().getIntWidth(), 0);
1198 
1199  // Narrow character literals act as though their value is concatenated
1200  // in this implementation, but warn on overflow.
1201  bool multi_char_too_long = false;
1202  if (isAscii() && isMultiChar()) {
1203  LitVal = 0;
1204  for (size_t i = 0; i < NumCharsSoFar; ++i) {
1205  // check for enough leading zeros to shift into
1206  multi_char_too_long |= (LitVal.countLeadingZeros() < 8);
1207  LitVal <<= 8;
1208  LitVal = LitVal + (codepoint_buffer[i] & 0xFF);
1209  }
1210  } else if (NumCharsSoFar > 0) {
1211  // otherwise just take the last character
1212  LitVal = buffer_begin[-1];
1213  }
1214 
1215  if (!HadError && multi_char_too_long) {
1216  PP.Diag(Loc, diag::warn_char_constant_too_large);
1217  }
1218 
1219  // Transfer the value from APInt to uint64_t
1220  Value = LitVal.getZExtValue();
1221 
1222  // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1")
1223  // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple
1224  // character constants are not sign extended in the this implementation:
1225  // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC.
1226  if (isAscii() && NumCharsSoFar == 1 && (Value & 128) &&
1227  PP.getLangOpts().CharIsSigned)
1228  Value = (signed char)Value;
1229 }
1230 
1231 /// \verbatim
1232 /// string-literal: [C++0x lex.string]
1233 /// encoding-prefix " [s-char-sequence] "
1234 /// encoding-prefix R raw-string
1235 /// encoding-prefix:
1236 /// u8
1237 /// u
1238 /// U
1239 /// L
1240 /// s-char-sequence:
1241 /// s-char
1242 /// s-char-sequence s-char
1243 /// s-char:
1244 /// any member of the source character set except the double-quote ",
1245 /// backslash \, or new-line character
1246 /// escape-sequence
1247 /// universal-character-name
1248 /// raw-string:
1249 /// " d-char-sequence ( r-char-sequence ) d-char-sequence "
1250 /// r-char-sequence:
1251 /// r-char
1252 /// r-char-sequence r-char
1253 /// r-char:
1254 /// any member of the source character set, except a right parenthesis )
1255 /// followed by the initial d-char-sequence (which may be empty)
1256 /// followed by a double quote ".
1257 /// d-char-sequence:
1258 /// d-char
1259 /// d-char-sequence d-char
1260 /// d-char:
1261 /// any member of the basic source character set except:
1262 /// space, the left parenthesis (, the right parenthesis ),
1263 /// the backslash \, and the control characters representing horizontal
1264 /// tab, vertical tab, form feed, and newline.
1265 /// escape-sequence: [C++0x lex.ccon]
1266 /// simple-escape-sequence
1267 /// octal-escape-sequence
1268 /// hexadecimal-escape-sequence
1269 /// simple-escape-sequence:
1270 /// one of \' \" \? \\ \a \b \f \n \r \t \v
1271 /// octal-escape-sequence:
1272 /// \ octal-digit
1273 /// \ octal-digit octal-digit
1274 /// \ octal-digit octal-digit octal-digit
1275 /// hexadecimal-escape-sequence:
1276 /// \x hexadecimal-digit
1277 /// hexadecimal-escape-sequence hexadecimal-digit
1278 /// universal-character-name:
1279 /// \u hex-quad
1280 /// \U hex-quad hex-quad
1281 /// hex-quad:
1282 /// hex-digit hex-digit hex-digit hex-digit
1283 /// \endverbatim
1284 ///
1287  Preprocessor &PP, bool Complain)
1288  : SM(PP.getSourceManager()), Features(PP.getLangOpts()),
1289  Target(PP.getTargetInfo()), Diags(Complain ? &PP.getDiagnostics() :nullptr),
1290  MaxTokenLength(0), SizeBound(0), CharByteWidth(0), Kind(tok::unknown),
1291  ResultPtr(ResultBuf.data()), hadError(false), Pascal(false) {
1292  init(StringToks);
1293 }
1294 
1295 void StringLiteralParser::init(ArrayRef<Token> StringToks){
1296  // The literal token may have come from an invalid source location (e.g. due
1297  // to a PCH error), in which case the token length will be 0.
1298  if (StringToks.empty() || StringToks[0].getLength() < 2)
1299  return DiagnoseLexingError(SourceLocation());
1300 
1301  // Scan all of the string portions, remember the max individual token length,
1302  // computing a bound on the concatenated string length, and see whether any
1303  // piece is a wide-string. If any of the string portions is a wide-string
1304  // literal, the result is a wide-string literal [C99 6.4.5p4].
1305  assert(!StringToks.empty() && "expected at least one token");
1306  MaxTokenLength = StringToks[0].getLength();
1307  assert(StringToks[0].getLength() >= 2 && "literal token is invalid!");
1308  SizeBound = StringToks[0].getLength()-2; // -2 for "".
1309  Kind = StringToks[0].getKind();
1310 
1311  hadError = false;
1312 
1313  // Implement Translation Phase #6: concatenation of string literals
1314  /// (C99 5.1.1.2p1). The common case is only one string fragment.
1315  for (unsigned i = 1; i != StringToks.size(); ++i) {
1316  if (StringToks[i].getLength() < 2)
1317  return DiagnoseLexingError(StringToks[i].getLocation());
1318 
1319  // The string could be shorter than this if it needs cleaning, but this is a
1320  // reasonable bound, which is all we need.
1321  assert(StringToks[i].getLength() >= 2 && "literal token is invalid!");
1322  SizeBound += StringToks[i].getLength()-2; // -2 for "".
1323 
1324  // Remember maximum string piece length.
1325  if (StringToks[i].getLength() > MaxTokenLength)
1326  MaxTokenLength = StringToks[i].getLength();
1327 
1328  // Remember if we see any wide or utf-8/16/32 strings.
1329  // Also check for illegal concatenations.
1330  if (StringToks[i].isNot(Kind) && StringToks[i].isNot(tok::string_literal)) {
1331  if (isAscii()) {
1332  Kind = StringToks[i].getKind();
1333  } else {
1334  if (Diags)
1335  Diags->Report(StringToks[i].getLocation(),
1336  diag::err_unsupported_string_concat);
1337  hadError = true;
1338  }
1339  }
1340  }
1341 
1342  // Include space for the null terminator.
1343  ++SizeBound;
1344 
1345  // TODO: K&R warning: "traditional C rejects string constant concatenation"
1346 
1347  // Get the width in bytes of char/wchar_t/char16_t/char32_t
1348  CharByteWidth = getCharWidth(Kind, Target);
1349  assert((CharByteWidth & 7) == 0 && "Assumes character size is byte multiple");
1350  CharByteWidth /= 8;
1351 
1352  // The output buffer size needs to be large enough to hold wide characters.
1353  // This is a worst-case assumption which basically corresponds to L"" "long".
1354  SizeBound *= CharByteWidth;
1355 
1356  // Size the temporary buffer to hold the result string data.
1357  ResultBuf.resize(SizeBound);
1358 
1359  // Likewise, but for each string piece.
1360  SmallString<512> TokenBuf;
1361  TokenBuf.resize(MaxTokenLength);
1362 
1363  // Loop over all the strings, getting their spelling, and expanding them to
1364  // wide strings as appropriate.
1365  ResultPtr = &ResultBuf[0]; // Next byte to fill in.
1366 
1367  Pascal = false;
1368 
1369  SourceLocation UDSuffixTokLoc;
1370 
1371  for (unsigned i = 0, e = StringToks.size(); i != e; ++i) {
1372  const char *ThisTokBuf = &TokenBuf[0];
1373  // Get the spelling of the token, which eliminates trigraphs, etc. We know
1374  // that ThisTokBuf points to a buffer that is big enough for the whole token
1375  // and 'spelled' tokens can only shrink.
1376  bool StringInvalid = false;
1377  unsigned ThisTokLen =
1378  Lexer::getSpelling(StringToks[i], ThisTokBuf, SM, Features,
1379  &StringInvalid);
1380  if (StringInvalid)
1381  return DiagnoseLexingError(StringToks[i].getLocation());
1382 
1383  const char *ThisTokBegin = ThisTokBuf;
1384  const char *ThisTokEnd = ThisTokBuf+ThisTokLen;
1385 
1386  // Remove an optional ud-suffix.
1387  if (ThisTokEnd[-1] != '"') {
1388  const char *UDSuffixEnd = ThisTokEnd;
1389  do {
1390  --ThisTokEnd;
1391  } while (ThisTokEnd[-1] != '"');
1392 
1393  StringRef UDSuffix(ThisTokEnd, UDSuffixEnd - ThisTokEnd);
1394 
1395  if (UDSuffixBuf.empty()) {
1396  if (StringToks[i].hasUCN())
1397  expandUCNs(UDSuffixBuf, UDSuffix);
1398  else
1399  UDSuffixBuf.assign(UDSuffix);
1400  UDSuffixToken = i;
1401  UDSuffixOffset = ThisTokEnd - ThisTokBuf;
1402  UDSuffixTokLoc = StringToks[i].getLocation();
1403  } else {
1404  SmallString<32> ExpandedUDSuffix;
1405  if (StringToks[i].hasUCN()) {
1406  expandUCNs(ExpandedUDSuffix, UDSuffix);
1407  UDSuffix = ExpandedUDSuffix;
1408  }
1409 
1410  // C++11 [lex.ext]p8: At the end of phase 6, if a string literal is the
1411  // result of a concatenation involving at least one user-defined-string-
1412  // literal, all the participating user-defined-string-literals shall
1413  // have the same ud-suffix.
1414  if (UDSuffixBuf != UDSuffix) {
1415  if (Diags) {
1416  SourceLocation TokLoc = StringToks[i].getLocation();
1417  Diags->Report(TokLoc, diag::err_string_concat_mixed_suffix)
1418  << UDSuffixBuf << UDSuffix
1419  << SourceRange(UDSuffixTokLoc, UDSuffixTokLoc)
1420  << SourceRange(TokLoc, TokLoc);
1421  }
1422  hadError = true;
1423  }
1424  }
1425  }
1426 
1427  // Strip the end quote.
1428  --ThisTokEnd;
1429 
1430  // TODO: Input character set mapping support.
1431 
1432  // Skip marker for wide or unicode strings.
1433  if (ThisTokBuf[0] == 'L' || ThisTokBuf[0] == 'u' || ThisTokBuf[0] == 'U') {
1434  ++ThisTokBuf;
1435  // Skip 8 of u8 marker for utf8 strings.
1436  if (ThisTokBuf[0] == '8')
1437  ++ThisTokBuf;
1438  }
1439 
1440  // Check for raw string
1441  if (ThisTokBuf[0] == 'R') {
1442  ThisTokBuf += 2; // skip R"
1443 
1444  const char *Prefix = ThisTokBuf;
1445  while (ThisTokBuf[0] != '(')
1446  ++ThisTokBuf;
1447  ++ThisTokBuf; // skip '('
1448 
1449  // Remove same number of characters from the end
1450  ThisTokEnd -= ThisTokBuf - Prefix;
1451  assert(ThisTokEnd >= ThisTokBuf && "malformed raw string literal");
1452 
1453  // C++14 [lex.string]p4: A source-file new-line in a raw string literal
1454  // results in a new-line in the resulting execution string-literal.
1455  StringRef RemainingTokenSpan(ThisTokBuf, ThisTokEnd - ThisTokBuf);
1456  while (!RemainingTokenSpan.empty()) {
1457  // Split the string literal on \r\n boundaries.
1458  size_t CRLFPos = RemainingTokenSpan.find("\r\n");
1459  StringRef BeforeCRLF = RemainingTokenSpan.substr(0, CRLFPos);
1460  StringRef AfterCRLF = RemainingTokenSpan.substr(CRLFPos);
1461 
1462  // Copy everything before the \r\n sequence into the string literal.
1463  if (CopyStringFragment(StringToks[i], ThisTokBegin, BeforeCRLF))
1464  hadError = true;
1465 
1466  // Point into the \n inside the \r\n sequence and operate on the
1467  // remaining portion of the literal.
1468  RemainingTokenSpan = AfterCRLF.substr(1);
1469  }
1470  } else {
1471  if (ThisTokBuf[0] != '"') {
1472  // The file may have come from PCH and then changed after loading the
1473  // PCH; Fail gracefully.
1474  return DiagnoseLexingError(StringToks[i].getLocation());
1475  }
1476  ++ThisTokBuf; // skip "
1477 
1478  // Check if this is a pascal string
1479  if (Features.PascalStrings && ThisTokBuf + 1 != ThisTokEnd &&
1480  ThisTokBuf[0] == '\\' && ThisTokBuf[1] == 'p') {
1481 
1482  // If the \p sequence is found in the first token, we have a pascal string
1483  // Otherwise, if we already have a pascal string, ignore the first \p
1484  if (i == 0) {
1485  ++ThisTokBuf;
1486  Pascal = true;
1487  } else if (Pascal)
1488  ThisTokBuf += 2;
1489  }
1490 
1491  while (ThisTokBuf != ThisTokEnd) {
1492  // Is this a span of non-escape characters?
1493  if (ThisTokBuf[0] != '\\') {
1494  const char *InStart = ThisTokBuf;
1495  do {
1496  ++ThisTokBuf;
1497  } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\');
1498 
1499  // Copy the character span over.
1500  if (CopyStringFragment(StringToks[i], ThisTokBegin,
1501  StringRef(InStart, ThisTokBuf - InStart)))
1502  hadError = true;
1503  continue;
1504  }
1505  // Is this a Universal Character Name escape?
1506  if (ThisTokBuf[1] == 'u' || ThisTokBuf[1] == 'U') {
1507  EncodeUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
1508  ResultPtr, hadError,
1509  FullSourceLoc(StringToks[i].getLocation(), SM),
1510  CharByteWidth, Diags, Features);
1511  continue;
1512  }
1513  // Otherwise, this is a non-UCN escape character. Process it.
1514  unsigned ResultChar =
1515  ProcessCharEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, hadError,
1516  FullSourceLoc(StringToks[i].getLocation(), SM),
1517  CharByteWidth*8, Diags, Features);
1518 
1519  if (CharByteWidth == 4) {
1520  // FIXME: Make the type of the result buffer correct instead of
1521  // using reinterpret_cast.
1522  llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultPtr);
1523  *ResultWidePtr = ResultChar;
1524  ResultPtr += 4;
1525  } else if (CharByteWidth == 2) {
1526  // FIXME: Make the type of the result buffer correct instead of
1527  // using reinterpret_cast.
1528  llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultPtr);
1529  *ResultWidePtr = ResultChar & 0xFFFF;
1530  ResultPtr += 2;
1531  } else {
1532  assert(CharByteWidth == 1 && "Unexpected char width");
1533  *ResultPtr++ = ResultChar & 0xFF;
1534  }
1535  }
1536  }
1537  }
1538 
1539  if (Pascal) {
1540  if (CharByteWidth == 4) {
1541  // FIXME: Make the type of the result buffer correct instead of
1542  // using reinterpret_cast.
1543  llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultBuf.data());
1544  ResultWidePtr[0] = GetNumStringChars() - 1;
1545  } else if (CharByteWidth == 2) {
1546  // FIXME: Make the type of the result buffer correct instead of
1547  // using reinterpret_cast.
1548  llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultBuf.data());
1549  ResultWidePtr[0] = GetNumStringChars() - 1;
1550  } else {
1551  assert(CharByteWidth == 1 && "Unexpected char width");
1552  ResultBuf[0] = GetNumStringChars() - 1;
1553  }
1554 
1555  // Verify that pascal strings aren't too large.
1556  if (GetStringLength() > 256) {
1557  if (Diags)
1558  Diags->Report(StringToks.front().getLocation(),
1559  diag::err_pascal_string_too_long)
1560  << SourceRange(StringToks.front().getLocation(),
1561  StringToks.back().getLocation());
1562  hadError = true;
1563  return;
1564  }
1565  } else if (Diags) {
1566  // Complain if this string literal has too many characters.
1567  unsigned MaxChars = Features.CPlusPlus? 65536 : Features.C99 ? 4095 : 509;
1568 
1569  if (GetNumStringChars() > MaxChars)
1570  Diags->Report(StringToks.front().getLocation(),
1571  diag::ext_string_too_long)
1572  << GetNumStringChars() << MaxChars
1573  << (Features.CPlusPlus ? 2 : Features.C99 ? 1 : 0)
1574  << SourceRange(StringToks.front().getLocation(),
1575  StringToks.back().getLocation());
1576  }
1577 }
1578 
1579 static const char *resyncUTF8(const char *Err, const char *End) {
1580  if (Err == End)
1581  return End;
1582  End = Err + std::min<unsigned>(llvm::getNumBytesForUTF8(*Err), End-Err);
1583  while (++Err != End && (*Err & 0xC0) == 0x80)
1584  ;
1585  return Err;
1586 }
1587 
1588 /// \brief This function copies from Fragment, which is a sequence of bytes
1589 /// within Tok's contents (which begin at TokBegin) into ResultPtr.
1590 /// Performs widening for multi-byte characters.
1591 bool StringLiteralParser::CopyStringFragment(const Token &Tok,
1592  const char *TokBegin,
1593  StringRef Fragment) {
1594  const llvm::UTF8 *ErrorPtrTmp;
1595  if (ConvertUTF8toWide(CharByteWidth, Fragment, ResultPtr, ErrorPtrTmp))
1596  return false;
1597 
1598  // If we see bad encoding for unprefixed string literals, warn and
1599  // simply copy the byte values, for compatibility with gcc and older
1600  // versions of clang.
1601  bool NoErrorOnBadEncoding = isAscii();
1602  if (NoErrorOnBadEncoding) {
1603  memcpy(ResultPtr, Fragment.data(), Fragment.size());
1604  ResultPtr += Fragment.size();
1605  }
1606 
1607  if (Diags) {
1608  const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
1609 
1610  FullSourceLoc SourceLoc(Tok.getLocation(), SM);
1611  const DiagnosticBuilder &Builder =
1612  Diag(Diags, Features, SourceLoc, TokBegin,
1613  ErrorPtr, resyncUTF8(ErrorPtr, Fragment.end()),
1614  NoErrorOnBadEncoding ? diag::warn_bad_string_encoding
1615  : diag::err_bad_string_encoding);
1616 
1617  const char *NextStart = resyncUTF8(ErrorPtr, Fragment.end());
1618  StringRef NextFragment(NextStart, Fragment.end()-NextStart);
1619 
1620  // Decode into a dummy buffer.
1621  SmallString<512> Dummy;
1622  Dummy.reserve(Fragment.size() * CharByteWidth);
1623  char *Ptr = Dummy.data();
1624 
1625  while (!ConvertUTF8toWide(CharByteWidth, NextFragment, Ptr, ErrorPtrTmp)) {
1626  const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
1627  NextStart = resyncUTF8(ErrorPtr, Fragment.end());
1628  Builder << MakeCharSourceRange(Features, SourceLoc, TokBegin,
1629  ErrorPtr, NextStart);
1630  NextFragment = StringRef(NextStart, Fragment.end()-NextStart);
1631  }
1632  }
1633  return !NoErrorOnBadEncoding;
1634 }
1635 
1636 void StringLiteralParser::DiagnoseLexingError(SourceLocation Loc) {
1637  hadError = true;
1638  if (Diags)
1639  Diags->Report(Loc, diag::err_lexing_string);
1640 }
1641 
1642 /// getOffsetOfStringByte - This function returns the offset of the
1643 /// specified byte of the string data represented by Token. This handles
1644 /// advancing over escape sequences in the string.
1646  unsigned ByteNo) const {
1647  // Get the spelling of the token.
1648  SmallString<32> SpellingBuffer;
1649  SpellingBuffer.resize(Tok.getLength());
1650 
1651  bool StringInvalid = false;
1652  const char *SpellingPtr = &SpellingBuffer[0];
1653  unsigned TokLen = Lexer::getSpelling(Tok, SpellingPtr, SM, Features,
1654  &StringInvalid);
1655  if (StringInvalid)
1656  return 0;
1657 
1658  const char *SpellingStart = SpellingPtr;
1659  const char *SpellingEnd = SpellingPtr+TokLen;
1660 
1661  // Handle UTF-8 strings just like narrow strings.
1662  if (SpellingPtr[0] == 'u' && SpellingPtr[1] == '8')
1663  SpellingPtr += 2;
1664 
1665  assert(SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' &&
1666  SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet");
1667 
1668  // For raw string literals, this is easy.
1669  if (SpellingPtr[0] == 'R') {
1670  assert(SpellingPtr[1] == '"' && "Should be a raw string literal!");
1671  // Skip 'R"'.
1672  SpellingPtr += 2;
1673  while (*SpellingPtr != '(') {
1674  ++SpellingPtr;
1675  assert(SpellingPtr < SpellingEnd && "Missing ( for raw string literal");
1676  }
1677  // Skip '('.
1678  ++SpellingPtr;
1679  return SpellingPtr - SpellingStart + ByteNo;
1680  }
1681 
1682  // Skip over the leading quote
1683  assert(SpellingPtr[0] == '"' && "Should be a string literal!");
1684  ++SpellingPtr;
1685 
1686  // Skip over bytes until we find the offset we're looking for.
1687  while (ByteNo) {
1688  assert(SpellingPtr < SpellingEnd && "Didn't find byte offset!");
1689 
1690  // Step over non-escapes simply.
1691  if (*SpellingPtr != '\\') {
1692  ++SpellingPtr;
1693  --ByteNo;
1694  continue;
1695  }
1696 
1697  // Otherwise, this is an escape character. Advance over it.
1698  bool HadError = false;
1699  if (SpellingPtr[1] == 'u' || SpellingPtr[1] == 'U') {
1700  const char *EscapePtr = SpellingPtr;
1701  unsigned Len = MeasureUCNEscape(SpellingStart, SpellingPtr, SpellingEnd,
1702  1, Features, HadError);
1703  if (Len > ByteNo) {
1704  // ByteNo is somewhere within the escape sequence.
1705  SpellingPtr = EscapePtr;
1706  break;
1707  }
1708  ByteNo -= Len;
1709  } else {
1710  ProcessCharEscape(SpellingStart, SpellingPtr, SpellingEnd, HadError,
1711  FullSourceLoc(Tok.getLocation(), SM),
1712  CharByteWidth*8, Diags, Features);
1713  --ByteNo;
1714  }
1715  assert(!HadError && "This method isn't valid on erroneous strings");
1716  }
1717 
1718  return SpellingPtr-SpellingStart;
1719 }
1720 
1721 /// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
1722 /// suffixes as ud-suffixes, because the diagnostic experience is better if we
1723 /// treat it as an invalid suffix.
1725  StringRef Suffix) {
1726  return NumericLiteralParser::isValidUDSuffix(LangOpts, Suffix) ||
1727  Suffix == "sv";
1728 }
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:370
static DiagnosticBuilder Diag(DiagnosticsEngine *Diags, const LangOptions &Features, FullSourceLoc TokLoc, const char *TokBegin, const char *TokRangeBegin, const char *TokRangeEnd, unsigned DiagID)
Produce a diagnostic highlighting some portion of a literal.
static LLVM_READONLY bool isDigit(unsigned char c)
Return true if this character is an ASCII digit: [0-9].
Definition: CharInfo.h:94
unsigned getOffsetOfStringByte(const Token &TheTok, unsigned ByteNo) const
getOffsetOfStringByte - This function returns the offset of the specified byte of the string data rep...
StringLiteralParser(ArrayRef< Token > StringToks, Preprocessor &PP, bool Complain=true)
unsigned GetNumStringChars() const
DiagnosticBuilder Report(SourceLocation Loc, unsigned DiagID)
Issue the message to the client.
Definition: Diagnostic.h:1205
unsigned getCharWidth() const
Definition: TargetInfo.h:333
static LLVM_READONLY bool isPreprocessingNumberBody(unsigned char c)
Return true if this is the body character of a C preprocessing number, which is [a-zA-Z0-9_.
Definition: CharInfo.h:148
unsigned GetStringLength() const
static bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits)
unsigned getChar32Width() const
getChar32Width/Align - Return the size of &#39;char32_t&#39; for this target, in bits.
Definition: TargetInfo.h:400
const TargetInfo & getTargetInfo() const
Definition: Preprocessor.h:729
Token - This structure provides full information about a lexed token.
Definition: Token.h:35
Keeps track of the various options that can be enabled, which controls the dialect of C or C++ that i...
Definition: LangOptions.h:48
const LangOptions & getLangOpts() const
Definition: Preprocessor.h:728
static int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, const char *ThisTokEnd, unsigned CharByteWidth, const LangOptions &Features, bool &HadError)
MeasureUCNEscape - Determine the number of bytes within the resulting string which this UCN will occu...
SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, unsigned Char) const
Given a location that specifies the start of a token, return a new location that specifies a characte...
const FormatToken & Tok
Concrete class used by the front-end to report problems and issues.
Definition: Diagnostic.h:147
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
NumericLiteralParser(StringRef TokSpelling, SourceLocation TokLoc, Preprocessor &PP)
integer-constant: [C99 6.4.4.1] decimal-constant integer-suffix octal-constant integer-suffix hexadec...
static SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, unsigned Character, const SourceManager &SM, const LangOptions &LangOpts)
AdvanceToTokenCharacter - If the current SourceLocation specifies a location at the start of a token...
Definition: Lexer.cpp:694
static bool isValidUDSuffix(const LangOptions &LangOpts, StringRef Suffix)
Determine whether a suffix is a valid ud-suffix.
A little helper class used to produce diagnostics.
Definition: Diagnostic.h:953
Exposes information about the current target.
Definition: TargetInfo.h:55
CharLiteralParser(const char *begin, const char *end, SourceLocation Loc, Preprocessor &PP, tok::TokenKind kind)
Defines the clang::LangOptions interface.
SourceLocation End
Represents a character-granular source range.
unsigned getIntWidth() const
getIntWidth/Align - Return the size of &#39;signed int&#39; and &#39;unsigned int&#39; for this target, in bits.
Definition: TargetInfo.h:346
SourceLocation getLocation() const
Return a source location identifier for the specified offset in the current file. ...
Definition: Token.h:124
Defines the clang::Preprocessor interface.
const SourceManager & getManager() const
SourceLocation Begin
static const char * resyncUTF8(const char *Err, const char *End)
char __ovld __cnfn min(char x, char y)
Returns y if y < x, otherwise it returns x.
static void appendCodePoint(unsigned Codepoint, llvm::SmallVectorImpl< char > &Str)
The result type of a method or function.
const SourceManager & SM
Definition: Format.cpp:1308
static CharSourceRange getCharRange(SourceRange R)
SourceManager & getSourceManager() const
Definition: Preprocessor.h:732
unsigned getWCharWidth() const
getWCharWidth/Align - Return the size of &#39;wchar_t&#39; for this target, in bits.
Definition: TargetInfo.h:390
bool GetIntegerValue(llvm::APInt &Val)
GetIntegerValue - Convert this numeric literal value to an APInt that matches Val&#39;s input width...
#define false
Definition: stdbool.h:33
Kind
Encodes a location in the source.
llvm::APFloat::opStatus GetFloatValue(llvm::APFloat &Result)
GetFloatValue - Convert this numeric literal to a floating value, using the specified APFloat fltSema...
unsigned getChar16Width() const
getChar16Width/Align - Return the size of &#39;char16_t&#39; for this target, in bits.
Definition: TargetInfo.h:395
static unsigned ProcessCharEscape(const char *ThisTokBegin, const char *&ThisTokBuf, const char *ThisTokEnd, bool &HadError, FullSourceLoc Loc, unsigned CharWidth, DiagnosticsEngine *Diags, const LangOptions &Features)
ProcessCharEscape - Parse a standard C escape sequence, which can occur in either a character or a st...
TokenKind
Provides a simple uniform namespace for tokens from all C languages.
Definition: TokenKinds.h:25
void expandUCNs(SmallVectorImpl< char > &Buf, StringRef Input)
Copy characters from Input to Buf, expanding any UCNs.
static void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, const char *ThisTokEnd, char *&ResultBuf, bool &HadError, FullSourceLoc Loc, unsigned CharByteWidth, DiagnosticsEngine *Diags, const LangOptions &Features)
EncodeUCNEscape - Read the Universal Character Name, check constraints and convert the UTF32 to UTF8 ...
Dataflow Directional Tag Classes.
static LLVM_READONLY bool isPrintable(unsigned char c)
Return true if this character is an ASCII printable character; that is, a character that should take ...
Definition: CharInfo.h:140
unsigned getLength() const
Definition: Token.h:127
Defines the clang::SourceLocation class and associated facilities.
DiagnosticsEngine & getDiagnostics() const
Definition: Preprocessor.h:725
static unsigned getCharWidth(tok::TokenKind kind, const TargetInfo &Target)
unsigned kind
All of the diagnostics that can be emitted by the frontend.
Definition: DiagnosticIDs.h:59
Defines the clang::TargetInfo interface.
A SourceLocation and its associated SourceManager.
A trivial tuple used to represent a source range.
static bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, const char *ThisTokEnd, uint32_t &UcnVal, unsigned short &UcnLen, FullSourceLoc Loc, DiagnosticsEngine *Diags, const LangOptions &Features, bool in_char_string_literal=false)
ProcessUCNEscape - Read the Universal Character Name, check constraints and return the UTF32...
DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const
Forwarding function for diagnostics.
static CharSourceRange MakeCharSourceRange(const LangOptions &Features, FullSourceLoc TokLoc, const char *TokBegin, const char *TokRangeBegin, const char *TokRangeEnd)
Engages in a tight little dance with the lexer to efficiently preprocess tokens.
Definition: Preprocessor.h:98
static LLVM_READONLY bool isHexDigit(unsigned char c)
Return true if this character is an ASCII hex digit: [0-9a-fA-F].
Definition: CharInfo.h:124