1//===--- LiteralSupport.cpp - Code to parse and process literals ----------===//
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 NumericLiteralParser, CharLiteralParser, and
10// StringLiteralParser interfaces.
11//
12//===----------------------------------------------------------------------===//
13
14#include "clang/Lex/LiteralSupport.h"
15#include "clang/Basic/CharInfo.h"
16#include "clang/Basic/LangOptions.h"
17#include "clang/Basic/SourceLocation.h"
18#include "clang/Basic/TargetInfo.h"
19#include "clang/Lex/LexDiagnostic.h"
20#include "clang/Lex/Lexer.h"
21#include "clang/Lex/Preprocessor.h"
22#include "clang/Lex/Token.h"
23#include "llvm/ADT/APInt.h"
24#include "llvm/ADT/SmallVector.h"
25#include "llvm/ADT/StringExtras.h"
26#include "llvm/ADT/StringSwitch.h"
27#include "llvm/Support/ConvertUTF.h"
28#include "llvm/Support/Error.h"
29#include "llvm/Support/ErrorHandling.h"
30#include "llvm/Support/Unicode.h"
31#include <algorithm>
32#include <cassert>
33#include <cstddef>
34#include <cstdint>
35#include <cstring>
36#include <string>
37
38using namespace clang;
39
40static unsigned getCharWidth(tok::TokenKind kind, const TargetInfo &Target) {
41 switch (kind) {
42 default: llvm_unreachable("Unknown token type!");
43 case tok::char_constant:
44 case tok::string_literal:
45 case tok::utf8_char_constant:
46 case tok::utf8_string_literal:
47 return Target.getCharWidth();
48 case tok::wide_char_constant:
49 case tok::wide_string_literal:
50 return Target.getWCharWidth();
51 case tok::utf16_char_constant:
52 case tok::utf16_string_literal:
53 return Target.getChar16Width();
54 case tok::utf32_char_constant:
55 case tok::utf32_string_literal:
56 return Target.getChar32Width();
57 }
58}
59
60static unsigned getEncodingPrefixLen(tok::TokenKind kind) {
61 switch (kind) {
62 default:
63 llvm_unreachable("Unknown token type!");
64 case tok::char_constant:
65 case tok::string_literal:
66 return 0;
67 case tok::utf8_char_constant:
68 case tok::utf8_string_literal:
69 return 2;
70 case tok::wide_char_constant:
71 case tok::wide_string_literal:
72 case tok::utf16_char_constant:
73 case tok::utf16_string_literal:
74 case tok::utf32_char_constant:
75 case tok::utf32_string_literal:
76 return 1;
77 }
78}
79
80static CharSourceRange MakeCharSourceRange(const LangOptions &Features,
81 FullSourceLoc TokLoc,
82 const char *TokBegin,
83 const char *TokRangeBegin,
84 const char *TokRangeEnd) {
85 SourceLocation Begin =
86 Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: TokRangeBegin - TokBegin,
87 SM: TokLoc.getManager(), LangOpts: Features);
88 SourceLocation End =
89 Lexer::AdvanceToTokenCharacter(TokStart: Begin, Characters: TokRangeEnd - TokRangeBegin,
90 SM: TokLoc.getManager(), LangOpts: Features);
91 return CharSourceRange::getCharRange(B: Begin, E: End);
92}
93
94/// Produce a diagnostic highlighting some portion of a literal.
95///
96/// Emits the diagnostic \p DiagID, highlighting the range of characters from
97/// \p TokRangeBegin (inclusive) to \p TokRangeEnd (exclusive), which must be
98/// a substring of a spelling buffer for the token beginning at \p TokBegin.
99static DiagnosticBuilder Diag(DiagnosticsEngine *Diags,
100 const LangOptions &Features, FullSourceLoc TokLoc,
101 const char *TokBegin, const char *TokRangeBegin,
102 const char *TokRangeEnd, unsigned DiagID) {
103 SourceLocation Begin =
104 Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: TokRangeBegin - TokBegin,
105 SM: TokLoc.getManager(), LangOpts: Features);
106 return Diags->Report(Loc: Begin, DiagID) <<
107 MakeCharSourceRange(Features, TokLoc, TokBegin, TokRangeBegin, TokRangeEnd);
108}
109
110static bool IsEscapeValidInUnevaluatedStringLiteral(char Escape) {
111 switch (Escape) {
112 case '\'':
113 case '"':
114 case '?':
115 case '\\':
116 case 'a':
117 case 'b':
118 case 'f':
119 case 'n':
120 case 'r':
121 case 't':
122 case 'v':
123 return true;
124 }
125 return false;
126}
127
128/// ProcessCharEscape - Parse a standard C escape sequence, which can occur in
129/// either a character or a string literal.
130static unsigned ProcessCharEscape(const char *ThisTokBegin,
131 const char *&ThisTokBuf,
132 const char *ThisTokEnd, bool &HadError,
133 FullSourceLoc Loc, unsigned CharWidth,
134 DiagnosticsEngine *Diags,
135 const LangOptions &Features,
136 StringLiteralEvalMethod EvalMethod) {
137 const char *EscapeBegin = ThisTokBuf;
138 bool Delimited = false;
139 bool EndDelimiterFound = false;
140
141 // Skip the '\' char.
142 ++ThisTokBuf;
143
144 // We know that this character can't be off the end of the buffer, because
145 // that would have been \", which would not have been the end of string.
146 unsigned ResultChar = *ThisTokBuf++;
147 char Escape = ResultChar;
148 switch (ResultChar) {
149 // These map to themselves.
150 case '\\': case '\'': case '"': case '?': break;
151
152 // These have fixed mappings.
153 case 'a':
154 // TODO: K&R: the meaning of '\\a' is different in traditional C
155 ResultChar = 7;
156 break;
157 case 'b':
158 ResultChar = 8;
159 break;
160 case 'e':
161 if (Diags)
162 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
163 DiagID: diag::ext_nonstandard_escape) << "e";
164 ResultChar = 27;
165 break;
166 case 'E':
167 if (Diags)
168 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
169 DiagID: diag::ext_nonstandard_escape) << "E";
170 ResultChar = 27;
171 break;
172 case 'f':
173 ResultChar = 12;
174 break;
175 case 'n':
176 ResultChar = 10;
177 break;
178 case 'r':
179 ResultChar = 13;
180 break;
181 case 't':
182 ResultChar = 9;
183 break;
184 case 'v':
185 ResultChar = 11;
186 break;
187 case 'x': { // Hex escape.
188 ResultChar = 0;
189 if (ThisTokBuf != ThisTokEnd && *ThisTokBuf == '{') {
190 Delimited = true;
191 ThisTokBuf++;
192 if (*ThisTokBuf == '}') {
193 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
194 DiagID: diag::err_delimited_escape_empty);
195 return ResultChar;
196 }
197 } else if (ThisTokBuf == ThisTokEnd || !isHexDigit(c: *ThisTokBuf)) {
198 if (Diags)
199 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
200 DiagID: diag::err_hex_escape_no_digits) << "x";
201 return ResultChar;
202 }
203
204 // Hex escapes are a maximal series of hex digits.
205 bool Overflow = false;
206 for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) {
207 if (Delimited && *ThisTokBuf == '}') {
208 ThisTokBuf++;
209 EndDelimiterFound = true;
210 break;
211 }
212 int CharVal = llvm::hexDigitValue(C: *ThisTokBuf);
213 if (CharVal == -1) {
214 // Non delimited hex escape sequences stop at the first non-hex digit.
215 if (!Delimited)
216 break;
217 HadError = true;
218 if (Diags)
219 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
220 DiagID: diag::err_delimited_escape_invalid)
221 << StringRef(ThisTokBuf, 1);
222 continue;
223 }
224 // About to shift out a digit?
225 if (ResultChar & 0xF0000000)
226 Overflow = true;
227 ResultChar <<= 4;
228 ResultChar |= CharVal;
229 }
230 // See if any bits will be truncated when evaluated as a character.
231 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
232 Overflow = true;
233 ResultChar &= ~0U >> (32-CharWidth);
234 }
235
236 // Check for overflow.
237 if (!HadError && Overflow) { // Too many digits to fit in
238 HadError = true;
239 if (Diags)
240 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
241 DiagID: diag::err_escape_too_large)
242 << 0;
243 }
244 break;
245 }
246 case '0': case '1': case '2': case '3':
247 case '4': case '5': case '6': case '7': {
248 // Octal escapes.
249 --ThisTokBuf;
250 ResultChar = 0;
251
252 // Octal escapes are a series of octal digits with maximum length 3.
253 // "\0123" is a two digit sequence equal to "\012" "3".
254 unsigned NumDigits = 0;
255 do {
256 ResultChar <<= 3;
257 ResultChar |= *ThisTokBuf++ - '0';
258 ++NumDigits;
259 } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 &&
260 ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7');
261
262 // Check for overflow. Reject '\777', but not L'\777'.
263 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
264 if (Diags)
265 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
266 DiagID: diag::err_escape_too_large) << 1;
267 ResultChar &= ~0U >> (32-CharWidth);
268 }
269 break;
270 }
271 case 'o': {
272 bool Overflow = false;
273 if (ThisTokBuf == ThisTokEnd || *ThisTokBuf != '{') {
274 HadError = true;
275 if (Diags)
276 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
277 DiagID: diag::err_delimited_escape_missing_brace)
278 << "o";
279
280 break;
281 }
282 ResultChar = 0;
283 Delimited = true;
284 ++ThisTokBuf;
285 if (*ThisTokBuf == '}') {
286 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
287 DiagID: diag::err_delimited_escape_empty);
288 return ResultChar;
289 }
290
291 while (ThisTokBuf != ThisTokEnd) {
292 if (*ThisTokBuf == '}') {
293 EndDelimiterFound = true;
294 ThisTokBuf++;
295 break;
296 }
297 if (*ThisTokBuf < '0' || *ThisTokBuf > '7') {
298 HadError = true;
299 if (Diags)
300 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
301 DiagID: diag::err_delimited_escape_invalid)
302 << StringRef(ThisTokBuf, 1);
303 ThisTokBuf++;
304 continue;
305 }
306 // Check if one of the top three bits is set before shifting them out.
307 if (ResultChar & 0xE0000000)
308 Overflow = true;
309
310 ResultChar <<= 3;
311 ResultChar |= *ThisTokBuf++ - '0';
312 }
313 // Check for overflow. Reject '\777', but not L'\777'.
314 if (!HadError &&
315 (Overflow || (CharWidth != 32 && (ResultChar >> CharWidth) != 0))) {
316 HadError = true;
317 if (Diags)
318 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
319 DiagID: diag::err_escape_too_large)
320 << 1;
321 ResultChar &= ~0U >> (32 - CharWidth);
322 }
323 break;
324 }
325 // Otherwise, these are not valid escapes.
326 case '(': case '{': case '[': case '%':
327 // GCC accepts these as extensions. We warn about them as such though.
328 if (Diags)
329 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
330 DiagID: diag::ext_nonstandard_escape)
331 << std::string(1, ResultChar);
332 break;
333 default:
334 if (!Diags)
335 break;
336
337 if (isPrintable(c: ResultChar))
338 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
339 DiagID: diag::ext_unknown_escape)
340 << std::string(1, ResultChar);
341 else
342 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
343 DiagID: diag::ext_unknown_escape)
344 << "x" + llvm::utohexstr(X: ResultChar);
345 break;
346 }
347
348 if (Delimited && Diags) {
349 if (!EndDelimiterFound)
350 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
351 DiagID: diag::err_expected)
352 << tok::r_brace;
353 else if (!HadError) {
354 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
355 DiagID: Features.CPlusPlus23 ? diag::warn_cxx23_delimited_escape_sequence
356 : diag::ext_delimited_escape_sequence)
357 << /*delimited*/ 0 << (Features.CPlusPlus ? 1 : 0);
358 }
359 }
360
361 if (EvalMethod == StringLiteralEvalMethod::Unevaluated &&
362 !IsEscapeValidInUnevaluatedStringLiteral(Escape)) {
363 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: EscapeBegin, TokRangeEnd: ThisTokBuf,
364 DiagID: diag::err_unevaluated_string_invalid_escape_sequence)
365 << StringRef(EscapeBegin, ThisTokBuf - EscapeBegin);
366 HadError = true;
367 }
368
369 return ResultChar;
370}
371
372static void appendCodePoint(unsigned Codepoint,
373 llvm::SmallVectorImpl<char> &Str) {
374 char ResultBuf[4];
375 char *ResultPtr = ResultBuf;
376 if (llvm::ConvertCodePointToUTF8(Source: Codepoint, ResultPtr))
377 Str.append(in_start: ResultBuf, in_end: ResultPtr);
378}
379
380void clang::expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
381 for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
382 if (*I != '\\') {
383 Buf.push_back(Elt: *I);
384 continue;
385 }
386
387 ++I;
388 char Kind = *I;
389 ++I;
390
391 assert(Kind == 'u' || Kind == 'U' || Kind == 'N');
392 uint32_t CodePoint = 0;
393
394 if (Kind == 'u' && *I == '{') {
395 for (++I; *I != '}'; ++I) {
396 unsigned Value = llvm::hexDigitValue(C: *I);
397 assert(Value != -1U);
398 CodePoint <<= 4;
399 CodePoint += Value;
400 }
401 appendCodePoint(Codepoint: CodePoint, Str&: Buf);
402 continue;
403 }
404
405 if (Kind == 'N') {
406 assert(*I == '{');
407 ++I;
408 auto Delim = std::find(first: I, last: Input.end(), val: '}');
409 assert(Delim != Input.end());
410 StringRef Name(I, std::distance(first: I, last: Delim));
411 std::optional<llvm::sys::unicode::LooseMatchingResult> Res =
412 llvm::sys::unicode::nameToCodepointLooseMatching(Name);
413 assert(Res && "could not find a codepoint that was previously found");
414 CodePoint = Res->CodePoint;
415 assert(CodePoint != 0xFFFFFFFF);
416 appendCodePoint(Codepoint: CodePoint, Str&: Buf);
417 I = Delim;
418 continue;
419 }
420
421 unsigned NumHexDigits;
422 if (Kind == 'u')
423 NumHexDigits = 4;
424 else
425 NumHexDigits = 8;
426
427 assert(I + NumHexDigits <= E);
428
429 for (; NumHexDigits != 0; ++I, --NumHexDigits) {
430 unsigned Value = llvm::hexDigitValue(C: *I);
431 assert(Value != -1U);
432
433 CodePoint <<= 4;
434 CodePoint += Value;
435 }
436
437 appendCodePoint(Codepoint: CodePoint, Str&: Buf);
438 --I;
439 }
440}
441
442bool clang::isFunctionLocalStringLiteralMacro(tok::TokenKind K,
443 const LangOptions &LO) {
444 return LO.MicrosoftExt &&
445 (K == tok::kw___FUNCTION__ || K == tok::kw_L__FUNCTION__ ||
446 K == tok::kw___FUNCSIG__ || K == tok::kw_L__FUNCSIG__ ||
447 K == tok::kw___FUNCDNAME__);
448}
449
450bool clang::tokenIsLikeStringLiteral(const Token &Tok, const LangOptions &LO) {
451 return tok::isStringLiteral(K: Tok.getKind()) ||
452 isFunctionLocalStringLiteralMacro(K: Tok.getKind(), LO);
453}
454
455static bool ProcessNumericUCNEscape(const char *ThisTokBegin,
456 const char *&ThisTokBuf,
457 const char *ThisTokEnd, uint32_t &UcnVal,
458 unsigned short &UcnLen, bool &Delimited,
459 FullSourceLoc Loc, DiagnosticsEngine *Diags,
460 const LangOptions &Features,
461 bool in_char_string_literal = false) {
462 const char *UcnBegin = ThisTokBuf;
463 bool HasError = false;
464 bool EndDelimiterFound = false;
465
466 // Skip the '\u' char's.
467 ThisTokBuf += 2;
468 Delimited = false;
469 if (UcnBegin[1] == 'u' && in_char_string_literal &&
470 ThisTokBuf != ThisTokEnd && *ThisTokBuf == '{') {
471 Delimited = true;
472 ThisTokBuf++;
473 } else if (ThisTokBuf == ThisTokEnd || !isHexDigit(c: *ThisTokBuf)) {
474 if (Diags)
475 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
476 DiagID: diag::err_hex_escape_no_digits)
477 << StringRef(&ThisTokBuf[-1], 1);
478 return false;
479 }
480 UcnLen = (ThisTokBuf[-1] == 'u' ? 4 : 8);
481
482 bool Overflow = false;
483 unsigned short Count = 0;
484 for (; ThisTokBuf != ThisTokEnd && (Delimited || Count != UcnLen);
485 ++ThisTokBuf) {
486 if (Delimited && *ThisTokBuf == '}') {
487 ++ThisTokBuf;
488 EndDelimiterFound = true;
489 break;
490 }
491 int CharVal = llvm::hexDigitValue(C: *ThisTokBuf);
492 if (CharVal == -1) {
493 HasError = true;
494 if (!Delimited)
495 break;
496 if (Diags) {
497 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
498 DiagID: diag::err_delimited_escape_invalid)
499 << StringRef(ThisTokBuf, 1);
500 }
501 Count++;
502 continue;
503 }
504 if (UcnVal & 0xF0000000) {
505 Overflow = true;
506 continue;
507 }
508 UcnVal <<= 4;
509 UcnVal |= CharVal;
510 Count++;
511 }
512
513 if (Overflow) {
514 if (Diags)
515 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
516 DiagID: diag::err_escape_too_large)
517 << 0;
518 return false;
519 }
520
521 if (Delimited && !EndDelimiterFound) {
522 if (Diags) {
523 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
524 DiagID: diag::err_expected)
525 << tok::r_brace;
526 }
527 return false;
528 }
529
530 // If we didn't consume the proper number of digits, there is a problem.
531 if (Count == 0 || (!Delimited && Count != UcnLen)) {
532 if (Diags)
533 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
534 DiagID: Delimited ? diag::err_delimited_escape_empty
535 : diag::err_ucn_escape_incomplete);
536 return false;
537 }
538 return !HasError;
539}
540
541static void DiagnoseInvalidUnicodeCharacterName(
542 DiagnosticsEngine *Diags, const LangOptions &Features, FullSourceLoc Loc,
543 const char *TokBegin, const char *TokRangeBegin, const char *TokRangeEnd,
544 llvm::StringRef Name) {
545
546 Diag(Diags, Features, TokLoc: Loc, TokBegin, TokRangeBegin, TokRangeEnd,
547 DiagID: diag::err_invalid_ucn_name)
548 << Name;
549
550 namespace u = llvm::sys::unicode;
551
552 std::optional<u::LooseMatchingResult> Res =
553 u::nameToCodepointLooseMatching(Name);
554 if (Res) {
555 Diag(Diags, Features, TokLoc: Loc, TokBegin, TokRangeBegin, TokRangeEnd,
556 DiagID: diag::note_invalid_ucn_name_loose_matching)
557 << FixItHint::CreateReplacement(
558 RemoveRange: MakeCharSourceRange(Features, TokLoc: Loc, TokBegin, TokRangeBegin,
559 TokRangeEnd),
560 Code: Res->Name);
561 return;
562 }
563
564 unsigned Distance = 0;
565 SmallVector<u::MatchForCodepointName> Matches =
566 u::nearestMatchesForCodepointName(Pattern: Name, MaxMatchesCount: 5);
567 assert(!Matches.empty() && "No unicode characters found");
568
569 for (const auto &Match : Matches) {
570 if (Distance == 0)
571 Distance = Match.Distance;
572 if (std::max(a: Distance, b: Match.Distance) -
573 std::min(a: Distance, b: Match.Distance) >
574 3)
575 break;
576 Distance = Match.Distance;
577
578 std::string Str;
579 llvm::UTF32 V = Match.Value;
580 bool Converted =
581 llvm::convertUTF32ToUTF8String(Src: llvm::ArrayRef<llvm::UTF32>(&V, 1), Out&: Str);
582 (void)Converted;
583 assert(Converted && "Found a match wich is not a unicode character");
584
585 Diag(Diags, Features, TokLoc: Loc, TokBegin, TokRangeBegin, TokRangeEnd,
586 DiagID: diag::note_invalid_ucn_name_candidate)
587 << Match.Name << llvm::utohexstr(X: Match.Value)
588 << Str // FIXME: Fix the rendering of non printable characters
589 << FixItHint::CreateReplacement(
590 RemoveRange: MakeCharSourceRange(Features, TokLoc: Loc, TokBegin, TokRangeBegin,
591 TokRangeEnd),
592 Code: Match.Name);
593 }
594}
595
596static bool ProcessNamedUCNEscape(const char *ThisTokBegin,
597 const char *&ThisTokBuf,
598 const char *ThisTokEnd, uint32_t &UcnVal,
599 unsigned short &UcnLen, FullSourceLoc Loc,
600 DiagnosticsEngine *Diags,
601 const LangOptions &Features) {
602 const char *UcnBegin = ThisTokBuf;
603 assert(UcnBegin[0] == '\\' && UcnBegin[1] == 'N');
604 ThisTokBuf += 2;
605 if (ThisTokBuf == ThisTokEnd || *ThisTokBuf != '{') {
606 if (Diags) {
607 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
608 DiagID: diag::err_delimited_escape_missing_brace)
609 << StringRef(&ThisTokBuf[-1], 1);
610 }
611 return false;
612 }
613 ThisTokBuf++;
614 const char *ClosingBrace = std::find_if(first: ThisTokBuf, last: ThisTokEnd, pred: [](char C) {
615 return C == '}' || isVerticalWhitespace(c: C);
616 });
617 bool Incomplete = ClosingBrace == ThisTokEnd;
618 bool Empty = ClosingBrace == ThisTokBuf;
619 if (Incomplete || Empty) {
620 if (Diags) {
621 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
622 DiagID: Incomplete ? diag::err_ucn_escape_incomplete
623 : diag::err_delimited_escape_empty)
624 << StringRef(&UcnBegin[1], 1);
625 }
626 ThisTokBuf = ClosingBrace == ThisTokEnd ? ClosingBrace : ClosingBrace + 1;
627 return false;
628 }
629 StringRef Name(ThisTokBuf, ClosingBrace - ThisTokBuf);
630 ThisTokBuf = ClosingBrace + 1;
631 std::optional<char32_t> Res = llvm::sys::unicode::nameToCodepointStrict(Name);
632 if (!Res) {
633 if (Diags)
634 DiagnoseInvalidUnicodeCharacterName(Diags, Features, Loc, TokBegin: ThisTokBegin,
635 TokRangeBegin: &UcnBegin[3], TokRangeEnd: ClosingBrace, Name);
636 return false;
637 }
638 UcnVal = *Res;
639 UcnLen = UcnVal > 0xFFFF ? 8 : 4;
640 return true;
641}
642
643/// ProcessUCNEscape - Read the Universal Character Name, check constraints and
644/// return the UTF32.
645static bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
646 const char *ThisTokEnd, uint32_t &UcnVal,
647 unsigned short &UcnLen, FullSourceLoc Loc,
648 DiagnosticsEngine *Diags,
649 const LangOptions &Features,
650 bool in_char_string_literal = false) {
651
652 bool HasError;
653 const char *UcnBegin = ThisTokBuf;
654 bool IsDelimitedEscapeSequence = false;
655 bool IsNamedEscapeSequence = false;
656 if (ThisTokBuf[1] == 'N') {
657 IsNamedEscapeSequence = true;
658 HasError = !ProcessNamedUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
659 UcnVal, UcnLen, Loc, Diags, Features);
660 } else {
661 HasError =
662 !ProcessNumericUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
663 UcnLen, Delimited&: IsDelimitedEscapeSequence, Loc, Diags,
664 Features, in_char_string_literal);
665 }
666 if (HasError)
667 return false;
668
669 // Check UCN constraints (C99 6.4.3p2) [C++11 lex.charset p2]
670 if ((0xD800 <= UcnVal && UcnVal <= 0xDFFF) || // surrogate codepoints
671 UcnVal > 0x10FFFF) { // maximum legal UTF32 value
672 if (Diags)
673 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
674 DiagID: diag::err_ucn_escape_invalid);
675 return false;
676 }
677
678 // C23 and C++11 allow UCNs that refer to control characters
679 // and basic source characters inside character and string literals
680 if (UcnVal < 0xa0 &&
681 // $, @, ` are allowed in all language modes
682 (UcnVal != 0x24 && UcnVal != 0x40 && UcnVal != 0x60)) {
683 bool IsError =
684 (!(Features.CPlusPlus11 || Features.C23) || !in_char_string_literal);
685 if (Diags) {
686 char BasicSCSChar = UcnVal;
687 if (UcnVal >= 0x20 && UcnVal < 0x7f)
688 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
689 DiagID: IsError ? diag::err_ucn_escape_basic_scs
690 : Features.CPlusPlus
691 ? diag::warn_cxx98_compat_literal_ucn_escape_basic_scs
692 : diag::warn_c23_compat_literal_ucn_escape_basic_scs)
693 << StringRef(&BasicSCSChar, 1);
694 else
695 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
696 DiagID: IsError ? diag::err_ucn_control_character
697 : Features.CPlusPlus
698 ? diag::warn_cxx98_compat_literal_ucn_control_character
699 : diag::warn_c23_compat_literal_ucn_control_character);
700 }
701 if (IsError)
702 return false;
703 }
704
705 if (!Features.CPlusPlus && !Features.C99 && Diags)
706 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
707 DiagID: diag::warn_ucn_not_valid_in_c89_literal);
708
709 if ((IsDelimitedEscapeSequence || IsNamedEscapeSequence) && Diags)
710 Diag(Diags, Features, TokLoc: Loc, TokBegin: ThisTokBegin, TokRangeBegin: UcnBegin, TokRangeEnd: ThisTokBuf,
711 DiagID: Features.CPlusPlus23 ? diag::warn_cxx23_delimited_escape_sequence
712 : diag::ext_delimited_escape_sequence)
713 << (IsNamedEscapeSequence ? 1 : 0) << (Features.CPlusPlus ? 1 : 0);
714
715 return true;
716}
717
718/// MeasureUCNEscape - Determine the number of bytes within the resulting string
719/// which this UCN will occupy.
720static int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
721 const char *ThisTokEnd, unsigned CharByteWidth,
722 const LangOptions &Features, bool &HadError) {
723 // UTF-32: 4 bytes per escape.
724 if (CharByteWidth == 4)
725 return 4;
726
727 uint32_t UcnVal = 0;
728 unsigned short UcnLen = 0;
729 FullSourceLoc Loc;
730
731 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
732 UcnLen, Loc, Diags: nullptr, Features, in_char_string_literal: true)) {
733 HadError = true;
734 return 0;
735 }
736
737 // UTF-16: 2 bytes for BMP, 4 bytes otherwise.
738 if (CharByteWidth == 2)
739 return UcnVal <= 0xFFFF ? 2 : 4;
740
741 // UTF-8.
742 if (UcnVal < 0x80)
743 return 1;
744 if (UcnVal < 0x800)
745 return 2;
746 if (UcnVal < 0x10000)
747 return 3;
748 return 4;
749}
750
751/// EncodeUCNEscape - Read the Universal Character Name, check constraints and
752/// convert the UTF32 to UTF8 or UTF16. This is a subroutine of
753/// StringLiteralParser. When we decide to implement UCN's for identifiers,
754/// we will likely rework our support for UCN's.
755static void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
756 const char *ThisTokEnd,
757 char *&ResultBuf, bool &HadError,
758 FullSourceLoc Loc, unsigned CharByteWidth,
759 DiagnosticsEngine *Diags,
760 const LangOptions &Features) {
761 typedef uint32_t UTF32;
762 UTF32 UcnVal = 0;
763 unsigned short UcnLen = 0;
764 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, UcnLen,
765 Loc, Diags, Features, in_char_string_literal: true)) {
766 HadError = true;
767 return;
768 }
769
770 assert((CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) &&
771 "only character widths of 1, 2, or 4 bytes supported");
772
773 (void)UcnLen;
774 assert((UcnLen== 4 || UcnLen== 8) && "only ucn length of 4 or 8 supported");
775
776 if (CharByteWidth == 4) {
777 // FIXME: Make the type of the result buffer correct instead of
778 // using reinterpret_cast.
779 llvm::UTF32 *ResultPtr = reinterpret_cast<llvm::UTF32*>(ResultBuf);
780 *ResultPtr = UcnVal;
781 ResultBuf += 4;
782 return;
783 }
784
785 if (CharByteWidth == 2) {
786 // FIXME: Make the type of the result buffer correct instead of
787 // using reinterpret_cast.
788 llvm::UTF16 *ResultPtr = reinterpret_cast<llvm::UTF16*>(ResultBuf);
789
790 if (UcnVal <= (UTF32)0xFFFF) {
791 *ResultPtr = UcnVal;
792 ResultBuf += 2;
793 return;
794 }
795
796 // Convert to UTF16.
797 UcnVal -= 0x10000;
798 *ResultPtr = 0xD800 + (UcnVal >> 10);
799 *(ResultPtr+1) = 0xDC00 + (UcnVal & 0x3FF);
800 ResultBuf += 4;
801 return;
802 }
803
804 assert(CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters");
805
806 // Now that we've parsed/checked the UCN, we convert from UTF32->UTF8.
807 // The conversion below was inspired by:
808 // http://www.unicode.org/Public/PROGRAMS/CVTUTF/ConvertUTF.c
809 // First, we determine how many bytes the result will require.
810 typedef uint8_t UTF8;
811
812 unsigned short bytesToWrite = 0;
813 if (UcnVal < (UTF32)0x80)
814 bytesToWrite = 1;
815 else if (UcnVal < (UTF32)0x800)
816 bytesToWrite = 2;
817 else if (UcnVal < (UTF32)0x10000)
818 bytesToWrite = 3;
819 else
820 bytesToWrite = 4;
821
822 const unsigned byteMask = 0xBF;
823 const unsigned byteMark = 0x80;
824
825 // Once the bits are split out into bytes of UTF8, this is a mask OR-ed
826 // into the first byte, depending on how many bytes follow.
827 static const UTF8 firstByteMark[5] = {
828 0x00, 0x00, 0xC0, 0xE0, 0xF0
829 };
830 // Finally, we write the bytes into ResultBuf.
831 ResultBuf += bytesToWrite;
832 switch (bytesToWrite) { // note: everything falls through.
833 case 4:
834 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
835 [[fallthrough]];
836 case 3:
837 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
838 [[fallthrough]];
839 case 2:
840 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
841 [[fallthrough]];
842 case 1:
843 *--ResultBuf = (UTF8) (UcnVal | firstByteMark[bytesToWrite]);
844 }
845 // Update the buffer.
846 ResultBuf += bytesToWrite;
847}
848
849/// integer-constant: [C99 6.4.4.1]
850/// decimal-constant integer-suffix
851/// octal-constant integer-suffix
852/// hexadecimal-constant integer-suffix
853/// binary-literal integer-suffix [GNU, C++1y]
854/// user-defined-integer-literal: [C++11 lex.ext]
855/// decimal-literal ud-suffix
856/// octal-literal ud-suffix
857/// hexadecimal-literal ud-suffix
858/// binary-literal ud-suffix [GNU, C++1y]
859/// decimal-constant:
860/// nonzero-digit
861/// decimal-constant digit
862/// octal-constant:
863/// 0
864/// octal-constant octal-digit
865/// hexadecimal-constant:
866/// hexadecimal-prefix hexadecimal-digit
867/// hexadecimal-constant hexadecimal-digit
868/// hexadecimal-prefix: one of
869/// 0x 0X
870/// binary-literal:
871/// 0b binary-digit
872/// 0B binary-digit
873/// binary-literal binary-digit
874/// integer-suffix:
875/// unsigned-suffix [long-suffix]
876/// unsigned-suffix [long-long-suffix]
877/// long-suffix [unsigned-suffix]
878/// long-long-suffix [unsigned-sufix]
879/// nonzero-digit:
880/// 1 2 3 4 5 6 7 8 9
881/// octal-digit:
882/// 0 1 2 3 4 5 6 7
883/// hexadecimal-digit:
884/// 0 1 2 3 4 5 6 7 8 9
885/// a b c d e f
886/// A B C D E F
887/// binary-digit:
888/// 0
889/// 1
890/// unsigned-suffix: one of
891/// u U
892/// long-suffix: one of
893/// l L
894/// long-long-suffix: one of
895/// ll LL
896///
897/// floating-constant: [C99 6.4.4.2]
898/// TODO: add rules...
899///
900NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
901 SourceLocation TokLoc,
902 const SourceManager &SM,
903 const LangOptions &LangOpts,
904 const TargetInfo &Target,
905 DiagnosticsEngine &Diags)
906 : SM(SM), LangOpts(LangOpts), Diags(Diags),
907 ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) {
908
909 s = DigitsBegin = ThisTokBegin;
910 saw_exponent = false;
911 saw_period = false;
912 saw_ud_suffix = false;
913 saw_fixed_point_suffix = false;
914 isLong = false;
915 isUnsigned = false;
916 isLongLong = false;
917 isSizeT = false;
918 isHalf = false;
919 isFloat = false;
920 isImaginary = false;
921 isFloat16 = false;
922 isFloat128 = false;
923 MicrosoftInteger = 0;
924 isFract = false;
925 isAccum = false;
926 hadError = false;
927 isBitInt = false;
928
929 // This routine assumes that the range begin/end matches the regex for integer
930 // and FP constants (specifically, the 'pp-number' regex), and assumes that
931 // the byte at "*end" is both valid and not part of the regex. Because of
932 // this, it doesn't have to check for 'overscan' in various places.
933 // Note: For HLSL, the end token is allowed to be '.' which would be in the
934 // 'pp-number' regex. This is required to support vector swizzles on numeric
935 // constants (i.e. 1.xx or 1.5f.rrr).
936 if (isPreprocessingNumberBody(c: *ThisTokEnd) &&
937 !(LangOpts.HLSL && *ThisTokEnd == '.')) {
938 Diags.Report(Loc: TokLoc, DiagID: diag::err_lexing_numeric);
939 hadError = true;
940 return;
941 }
942
943 if (*s == '0') { // parse radix
944 ParseNumberStartingWithZero(TokLoc);
945 if (hadError)
946 return;
947 } else { // the first digit is non-zero
948 radix = 10;
949 s = SkipDigits(ptr: s);
950 if (s == ThisTokEnd) {
951 // Done.
952 } else {
953 ParseDecimalOrOctalCommon(TokLoc);
954 if (hadError)
955 return;
956 }
957 }
958
959 SuffixBegin = s;
960 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_AfterDigits);
961
962 // Initial scan to lookahead for fixed point suffix.
963 if (LangOpts.FixedPoint) {
964 for (const char *c = s; c != ThisTokEnd; ++c) {
965 if (*c == 'r' || *c == 'k' || *c == 'R' || *c == 'K') {
966 saw_fixed_point_suffix = true;
967 break;
968 }
969 }
970 }
971
972 // Parse the suffix. At this point we can classify whether we have an FP or
973 // integer constant.
974 bool isFixedPointConstant = isFixedPointLiteral();
975 bool isFPConstant = isFloatingLiteral();
976 bool HasSize = false;
977 bool DoubleUnderscore = false;
978
979 // Loop over all of the characters of the suffix. If we see something bad,
980 // we break out of the loop.
981 for (; s != ThisTokEnd; ++s) {
982 switch (*s) {
983 case 'R':
984 case 'r':
985 if (!LangOpts.FixedPoint)
986 break;
987 if (isFract || isAccum) break;
988 if (!(saw_period || saw_exponent)) break;
989 isFract = true;
990 continue;
991 case 'K':
992 case 'k':
993 if (!LangOpts.FixedPoint)
994 break;
995 if (isFract || isAccum) break;
996 if (!(saw_period || saw_exponent)) break;
997 isAccum = true;
998 continue;
999 case 'h': // FP Suffix for "half".
1000 case 'H':
1001 // OpenCL Extension v1.2 s9.5 - h or H suffix for half type.
1002 if (!(LangOpts.Half || LangOpts.FixedPoint))
1003 break;
1004 if (isIntegerLiteral()) break; // Error for integer constant.
1005 if (HasSize)
1006 break;
1007 HasSize = true;
1008 isHalf = true;
1009 continue; // Success.
1010 case 'f': // FP Suffix for "float"
1011 case 'F':
1012 if (!isFPConstant) break; // Error for integer constant.
1013 if (HasSize)
1014 break;
1015 HasSize = true;
1016
1017 // CUDA host and device may have different _Float16 support, therefore
1018 // allows f16 literals to avoid false alarm.
1019 // When we compile for OpenMP target offloading on NVPTX, f16 suffix
1020 // should also be supported.
1021 // ToDo: more precise check for CUDA.
1022 // TODO: AMDGPU might also support it in the future.
1023 if ((Target.hasFloat16Type() || LangOpts.CUDA ||
1024 (LangOpts.OpenMPIsTargetDevice && Target.getTriple().isNVPTX())) &&
1025 s + 2 < ThisTokEnd && s[1] == '1' && s[2] == '6') {
1026 s += 2; // success, eat up 2 characters.
1027 isFloat16 = true;
1028 continue;
1029 }
1030
1031 isFloat = true;
1032 continue; // Success.
1033 case 'q': // FP Suffix for "__float128"
1034 case 'Q':
1035 if (!isFPConstant) break; // Error for integer constant.
1036 if (HasSize)
1037 break;
1038 HasSize = true;
1039 isFloat128 = true;
1040 continue; // Success.
1041 case 'u':
1042 case 'U':
1043 if (isFPConstant) break; // Error for floating constant.
1044 if (isUnsigned) break; // Cannot be repeated.
1045 isUnsigned = true;
1046 continue; // Success.
1047 case 'l':
1048 case 'L':
1049 if (HasSize)
1050 break;
1051 HasSize = true;
1052
1053 // Check for long long. The L's need to be adjacent and the same case.
1054 if (s[1] == s[0]) {
1055 assert(s + 1 < ThisTokEnd && "didn't maximally munch?");
1056 if (isFPConstant) break; // long long invalid for floats.
1057 isLongLong = true;
1058 ++s; // Eat both of them.
1059 } else {
1060 isLong = true;
1061 }
1062 continue; // Success.
1063 case 'z':
1064 case 'Z':
1065 if (isFPConstant)
1066 break; // Invalid for floats.
1067 if (HasSize)
1068 break;
1069 HasSize = true;
1070 isSizeT = true;
1071 continue;
1072 case 'i':
1073 case 'I':
1074 if (LangOpts.MicrosoftExt && !isFPConstant) {
1075 // Allow i8, i16, i32, and i64. First, look ahead and check if
1076 // suffixes are Microsoft integers and not the imaginary unit.
1077 uint8_t Bits = 0;
1078 size_t ToSkip = 0;
1079 switch (s[1]) {
1080 case '8': // i8 suffix
1081 Bits = 8;
1082 ToSkip = 2;
1083 break;
1084 case '1':
1085 if (s[2] == '6') { // i16 suffix
1086 Bits = 16;
1087 ToSkip = 3;
1088 }
1089 break;
1090 case '3':
1091 if (s[2] == '2') { // i32 suffix
1092 Bits = 32;
1093 ToSkip = 3;
1094 }
1095 break;
1096 case '6':
1097 if (s[2] == '4') { // i64 suffix
1098 Bits = 64;
1099 ToSkip = 3;
1100 }
1101 break;
1102 default:
1103 break;
1104 }
1105 if (Bits) {
1106 if (HasSize)
1107 break;
1108 HasSize = true;
1109 MicrosoftInteger = Bits;
1110 s += ToSkip;
1111 assert(s <= ThisTokEnd && "didn't maximally munch?");
1112 break;
1113 }
1114 }
1115 [[fallthrough]];
1116 case 'j':
1117 case 'J':
1118 if (isImaginary) break; // Cannot be repeated.
1119 isImaginary = true;
1120 continue; // Success.
1121 case '_':
1122 if (isFPConstant)
1123 break; // Invalid for floats
1124 if (HasSize)
1125 break;
1126 // There is currently no way to reach this with DoubleUnderscore set.
1127 // If new double underscope literals are added handle it here as above.
1128 assert(!DoubleUnderscore && "unhandled double underscore case");
1129 if (LangOpts.CPlusPlus && s + 2 < ThisTokEnd &&
1130 s[1] == '_') { // s + 2 < ThisTokEnd to ensure some character exists
1131 // after __
1132 DoubleUnderscore = true;
1133 s += 2; // Skip both '_'
1134 if (s + 1 < ThisTokEnd &&
1135 (*s == 'u' || *s == 'U')) { // Ensure some character after 'u'/'U'
1136 isUnsigned = true;
1137 ++s;
1138 }
1139 if (s + 1 < ThisTokEnd &&
1140 ((*s == 'w' && *(++s) == 'b') || (*s == 'W' && *(++s) == 'B'))) {
1141 isBitInt = true;
1142 HasSize = true;
1143 continue;
1144 }
1145 }
1146 break;
1147 case 'w':
1148 case 'W':
1149 if (isFPConstant)
1150 break; // Invalid for floats.
1151 if (HasSize)
1152 break; // Invalid if we already have a size for the literal.
1153
1154 // wb and WB are allowed, but a mixture of cases like Wb or wB is not. We
1155 // explicitly do not support the suffix in C++ as an extension because a
1156 // library-based UDL that resolves to a library type may be more
1157 // appropriate there. The same rules apply for __wb/__WB.
1158 if ((!LangOpts.CPlusPlus || DoubleUnderscore) && s + 1 < ThisTokEnd &&
1159 ((s[0] == 'w' && s[1] == 'b') || (s[0] == 'W' && s[1] == 'B'))) {
1160 isBitInt = true;
1161 HasSize = true;
1162 ++s; // Skip both characters (2nd char skipped on continue).
1163 continue; // Success.
1164 }
1165 }
1166 // If we reached here, there was an error or a ud-suffix.
1167 break;
1168 }
1169
1170 // "i", "if", and "il" are user-defined suffixes in C++1y.
1171 if (s != ThisTokEnd || isImaginary) {
1172 // FIXME: Don't bother expanding UCNs if !tok.hasUCN().
1173 expandUCNs(Buf&: UDSuffixBuf, Input: StringRef(SuffixBegin, ThisTokEnd - SuffixBegin));
1174 if (isValidUDSuffix(LangOpts, Suffix: UDSuffixBuf)) {
1175 if (!isImaginary) {
1176 // Any suffix pieces we might have parsed are actually part of the
1177 // ud-suffix.
1178 isLong = false;
1179 isUnsigned = false;
1180 isLongLong = false;
1181 isSizeT = false;
1182 isFloat = false;
1183 isFloat16 = false;
1184 isHalf = false;
1185 isImaginary = false;
1186 isBitInt = false;
1187 MicrosoftInteger = 0;
1188 saw_fixed_point_suffix = false;
1189 isFract = false;
1190 isAccum = false;
1191 }
1192
1193 saw_ud_suffix = true;
1194 return;
1195 }
1196
1197 if (s != ThisTokEnd) {
1198 // Report an error if there are any.
1199 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(
1200 TokStart: TokLoc, Characters: SuffixBegin - ThisTokBegin, SM, LangOpts),
1201 DiagID: diag::err_invalid_suffix_constant)
1202 << StringRef(SuffixBegin, ThisTokEnd - SuffixBegin)
1203 << (isFixedPointConstant ? 2 : isFPConstant);
1204 hadError = true;
1205 }
1206 }
1207
1208 if (!hadError && saw_fixed_point_suffix) {
1209 assert(isFract || isAccum);
1210 }
1211}
1212
1213/// ParseDecimalOrOctalCommon - This method is called for decimal or octal
1214/// numbers. It issues an error for illegal digits, and handles floating point
1215/// parsing. If it detects a floating point number, the radix is set to 10.
1216void NumericLiteralParser::ParseDecimalOrOctalCommon(SourceLocation TokLoc){
1217 assert((radix == 8 || radix == 10) && "Unexpected radix");
1218
1219 // If we have a hex digit other than 'e' (which denotes a FP exponent) then
1220 // the code is using an incorrect base.
1221 if (isHexDigit(c: *s) && *s != 'e' && *s != 'E' &&
1222 !isValidUDSuffix(LangOpts, Suffix: StringRef(s, ThisTokEnd - s))) {
1223 Diags.Report(
1224 Loc: Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: s - ThisTokBegin, SM, LangOpts),
1225 DiagID: diag::err_invalid_digit)
1226 << StringRef(s, 1) << (radix == 8 ? 1 : 0);
1227 hadError = true;
1228 return;
1229 }
1230
1231 if (*s == '.') {
1232 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_AfterDigits);
1233 s++;
1234 radix = 10;
1235 saw_period = true;
1236 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_BeforeDigits);
1237 s = SkipDigits(ptr: s); // Skip suffix.
1238 }
1239 if (*s == 'e' || *s == 'E') { // exponent
1240 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_AfterDigits);
1241 const char *Exponent = s;
1242 s++;
1243 radix = 10;
1244 saw_exponent = true;
1245 if (s != ThisTokEnd && (*s == '+' || *s == '-')) s++; // sign
1246 const char *first_non_digit = SkipDigits(ptr: s);
1247 if (containsDigits(Start: s, End: first_non_digit)) {
1248 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_BeforeDigits);
1249 s = first_non_digit;
1250 } else {
1251 if (!hadError) {
1252 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(
1253 TokStart: TokLoc, Characters: Exponent - ThisTokBegin, SM, LangOpts),
1254 DiagID: diag::err_exponent_has_no_digits);
1255 hadError = true;
1256 }
1257 return;
1258 }
1259 }
1260}
1261
1262/// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
1263/// suffixes as ud-suffixes, because the diagnostic experience is better if we
1264/// treat it as an invalid suffix.
1265bool NumericLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
1266 StringRef Suffix) {
1267 if (!LangOpts.CPlusPlus11 || Suffix.empty())
1268 return false;
1269
1270 // By C++11 [lex.ext]p10, ud-suffixes starting with an '_' are always valid.
1271 // Suffixes starting with '__' (double underscore) are for use by
1272 // the implementation.
1273 if (Suffix.starts_with(Prefix: "_") && !Suffix.starts_with(Prefix: "__"))
1274 return true;
1275
1276 // In C++11, there are no library suffixes.
1277 if (!LangOpts.CPlusPlus14)
1278 return false;
1279
1280 // In C++14, "s", "h", "min", "ms", "us", and "ns" are used in the library.
1281 // Per tweaked N3660, "il", "i", and "if" are also used in the library.
1282 // In C++2a "d" and "y" are used in the library.
1283 return llvm::StringSwitch<bool>(Suffix)
1284 .Cases(S0: "h", S1: "min", S2: "s", Value: true)
1285 .Cases(S0: "ms", S1: "us", S2: "ns", Value: true)
1286 .Cases(S0: "il", S1: "i", S2: "if", Value: true)
1287 .Cases(S0: "d", S1: "y", Value: LangOpts.CPlusPlus20)
1288 .Default(Value: false);
1289}
1290
1291void NumericLiteralParser::checkSeparator(SourceLocation TokLoc,
1292 const char *Pos,
1293 CheckSeparatorKind IsAfterDigits) {
1294 if (IsAfterDigits == CSK_AfterDigits) {
1295 if (Pos == ThisTokBegin)
1296 return;
1297 --Pos;
1298 } else if (Pos == ThisTokEnd)
1299 return;
1300
1301 if (isDigitSeparator(C: *Pos)) {
1302 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: Pos - ThisTokBegin, SM,
1303 LangOpts),
1304 DiagID: diag::err_digit_separator_not_between_digits)
1305 << IsAfterDigits;
1306 hadError = true;
1307 }
1308}
1309
1310/// ParseNumberStartingWithZero - This method is called when the first character
1311/// of the number is found to be a zero. This means it is either an octal
1312/// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or
1313/// a floating point number (01239.123e4). Eat the prefix, determining the
1314/// radix etc.
1315void NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) {
1316 assert(s[0] == '0' && "Invalid method call");
1317 s++;
1318
1319 int c1 = s[0];
1320
1321 // Handle a hex number like 0x1234.
1322 if ((c1 == 'x' || c1 == 'X') && (isHexDigit(c: s[1]) || s[1] == '.')) {
1323 s++;
1324 assert(s < ThisTokEnd && "didn't maximally munch?");
1325 radix = 16;
1326 DigitsBegin = s;
1327 s = SkipHexDigits(ptr: s);
1328 bool HasSignificandDigits = containsDigits(Start: DigitsBegin, End: s);
1329 if (s == ThisTokEnd) {
1330 // Done.
1331 } else if (*s == '.') {
1332 s++;
1333 saw_period = true;
1334 const char *floatDigitsBegin = s;
1335 s = SkipHexDigits(ptr: s);
1336 if (containsDigits(Start: floatDigitsBegin, End: s))
1337 HasSignificandDigits = true;
1338 if (HasSignificandDigits)
1339 checkSeparator(TokLoc, Pos: floatDigitsBegin, IsAfterDigits: CSK_BeforeDigits);
1340 }
1341
1342 if (!HasSignificandDigits) {
1343 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: s - ThisTokBegin, SM,
1344 LangOpts),
1345 DiagID: diag::err_hex_constant_requires)
1346 << LangOpts.CPlusPlus << 1;
1347 hadError = true;
1348 return;
1349 }
1350
1351 // A binary exponent can appear with or with a '.'. If dotted, the
1352 // binary exponent is required.
1353 if (*s == 'p' || *s == 'P') {
1354 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_AfterDigits);
1355 const char *Exponent = s;
1356 s++;
1357 saw_exponent = true;
1358 if (s != ThisTokEnd && (*s == '+' || *s == '-')) s++; // sign
1359 const char *first_non_digit = SkipDigits(ptr: s);
1360 if (!containsDigits(Start: s, End: first_non_digit)) {
1361 if (!hadError) {
1362 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(
1363 TokStart: TokLoc, Characters: Exponent - ThisTokBegin, SM, LangOpts),
1364 DiagID: diag::err_exponent_has_no_digits);
1365 hadError = true;
1366 }
1367 return;
1368 }
1369 checkSeparator(TokLoc, Pos: s, IsAfterDigits: CSK_BeforeDigits);
1370 s = first_non_digit;
1371
1372 if (!LangOpts.HexFloats)
1373 Diags.Report(Loc: TokLoc, DiagID: LangOpts.CPlusPlus
1374 ? diag::ext_hex_literal_invalid
1375 : diag::ext_hex_constant_invalid);
1376 else if (LangOpts.CPlusPlus17)
1377 Diags.Report(Loc: TokLoc, DiagID: diag::warn_cxx17_hex_literal);
1378 } else if (saw_period) {
1379 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: s - ThisTokBegin, SM,
1380 LangOpts),
1381 DiagID: diag::err_hex_constant_requires)
1382 << LangOpts.CPlusPlus << 0;
1383 hadError = true;
1384 }
1385 return;
1386 }
1387
1388 // Handle simple binary numbers 0b01010
1389 if ((c1 == 'b' || c1 == 'B') && (s[1] == '0' || s[1] == '1')) {
1390 // 0b101010 is a C++14 and C23 extension.
1391 unsigned DiagId;
1392 if (LangOpts.CPlusPlus14)
1393 DiagId = diag::warn_cxx11_compat_binary_literal;
1394 else if (LangOpts.C23)
1395 DiagId = diag::warn_c23_compat_binary_literal;
1396 else if (LangOpts.CPlusPlus)
1397 DiagId = diag::ext_binary_literal_cxx14;
1398 else
1399 DiagId = diag::ext_binary_literal;
1400 Diags.Report(Loc: TokLoc, DiagID: DiagId);
1401 ++s;
1402 assert(s < ThisTokEnd && "didn't maximally munch?");
1403 radix = 2;
1404 DigitsBegin = s;
1405 s = SkipBinaryDigits(ptr: s);
1406 if (s == ThisTokEnd) {
1407 // Done.
1408 } else if (isHexDigit(c: *s) &&
1409 !isValidUDSuffix(LangOpts, Suffix: StringRef(s, ThisTokEnd - s))) {
1410 Diags.Report(Loc: Lexer::AdvanceToTokenCharacter(TokStart: TokLoc, Characters: s - ThisTokBegin, SM,
1411 LangOpts),
1412 DiagID: diag::err_invalid_digit)
1413 << StringRef(s, 1) << 2;
1414 hadError = true;
1415 }
1416 // Other suffixes will be diagnosed by the caller.
1417 return;
1418 }
1419
1420 // For now, the radix is set to 8. If we discover that we have a
1421 // floating point constant, the radix will change to 10. Octal floating
1422 // point constants are not permitted (only decimal and hexadecimal).
1423 radix = 8;
1424 const char *PossibleNewDigitStart = s;
1425 s = SkipOctalDigits(ptr: s);
1426 // When the value is 0 followed by a suffix (like 0wb), we want to leave 0
1427 // as the start of the digits. So if skipping octal digits does not skip
1428 // anything, we leave the digit start where it was.
1429 if (s != PossibleNewDigitStart)
1430 DigitsBegin = PossibleNewDigitStart;
1431
1432 if (s == ThisTokEnd)
1433 return; // Done, simple octal number like 01234
1434
1435 // If we have some other non-octal digit that *is* a decimal digit, see if
1436 // this is part of a floating point number like 094.123 or 09e1.
1437 if (isDigit(c: *s)) {
1438 const char *EndDecimal = SkipDigits(ptr: s);
1439 if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') {
1440 s = EndDecimal;
1441 radix = 10;
1442 }
1443 }
1444
1445 ParseDecimalOrOctalCommon(TokLoc);
1446}
1447
1448static bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits) {
1449 switch (Radix) {
1450 case 2:
1451 return NumDigits <= 64;
1452 case 8:
1453 return NumDigits <= 64 / 3; // Digits are groups of 3 bits.
1454 case 10:
1455 return NumDigits <= 19; // floor(log10(2^64))
1456 case 16:
1457 return NumDigits <= 64 / 4; // Digits are groups of 4 bits.
1458 default:
1459 llvm_unreachable("impossible Radix");
1460 }
1461}
1462
1463/// GetIntegerValue - Convert this numeric literal value to an APInt that
1464/// matches Val's input width. If there is an overflow, set Val to the low bits
1465/// of the result and return true. Otherwise, return false.
1466bool NumericLiteralParser::GetIntegerValue(llvm::APInt &Val) {
1467 // Fast path: Compute a conservative bound on the maximum number of
1468 // bits per digit in this radix. If we can't possibly overflow a
1469 // uint64 based on that bound then do the simple conversion to
1470 // integer. This avoids the expensive overflow checking below, and
1471 // handles the common cases that matter (small decimal integers and
1472 // hex/octal values which don't overflow).
1473 const unsigned NumDigits = SuffixBegin - DigitsBegin;
1474 if (alwaysFitsInto64Bits(Radix: radix, NumDigits)) {
1475 uint64_t N = 0;
1476 for (const char *Ptr = DigitsBegin; Ptr != SuffixBegin; ++Ptr)
1477 if (!isDigitSeparator(C: *Ptr))
1478 N = N * radix + llvm::hexDigitValue(C: *Ptr);
1479
1480 // This will truncate the value to Val's input width. Simply check
1481 // for overflow by comparing.
1482 Val = N;
1483 return Val.getZExtValue() != N;
1484 }
1485
1486 Val = 0;
1487 const char *Ptr = DigitsBegin;
1488
1489 llvm::APInt RadixVal(Val.getBitWidth(), radix);
1490 llvm::APInt CharVal(Val.getBitWidth(), 0);
1491 llvm::APInt OldVal = Val;
1492
1493 bool OverflowOccurred = false;
1494 while (Ptr < SuffixBegin) {
1495 if (isDigitSeparator(C: *Ptr)) {
1496 ++Ptr;
1497 continue;
1498 }
1499
1500 unsigned C = llvm::hexDigitValue(C: *Ptr++);
1501
1502 // If this letter is out of bound for this radix, reject it.
1503 assert(C < radix && "NumericLiteralParser ctor should have rejected this");
1504
1505 CharVal = C;
1506
1507 // Add the digit to the value in the appropriate radix. If adding in digits
1508 // made the value smaller, then this overflowed.
1509 OldVal = Val;
1510
1511 // Multiply by radix, did overflow occur on the multiply?
1512 Val *= RadixVal;
1513 OverflowOccurred |= Val.udiv(RHS: RadixVal) != OldVal;
1514
1515 // Add value, did overflow occur on the value?
1516 // (a + b) ult b <=> overflow
1517 Val += CharVal;
1518 OverflowOccurred |= Val.ult(RHS: CharVal);
1519 }
1520 return OverflowOccurred;
1521}
1522
1523llvm::APFloat::opStatus
1524NumericLiteralParser::GetFloatValue(llvm::APFloat &Result,
1525 llvm::RoundingMode RM) {
1526 using llvm::APFloat;
1527
1528 unsigned n = std::min(a: SuffixBegin - ThisTokBegin, b: ThisTokEnd - ThisTokBegin);
1529
1530 llvm::SmallString<16> Buffer;
1531 StringRef Str(ThisTokBegin, n);
1532 if (Str.contains(C: '\'')) {
1533 Buffer.reserve(N: n);
1534 std::remove_copy_if(first: Str.begin(), last: Str.end(), result: std::back_inserter(x&: Buffer),
1535 pred: &isDigitSeparator);
1536 Str = Buffer;
1537 }
1538
1539 auto StatusOrErr = Result.convertFromString(Str, RM);
1540 assert(StatusOrErr && "Invalid floating point representation");
1541 return !errorToBool(Err: StatusOrErr.takeError()) ? *StatusOrErr
1542 : APFloat::opInvalidOp;
1543}
1544
1545static inline bool IsExponentPart(char c, bool isHex) {
1546 if (isHex)
1547 return c == 'p' || c == 'P';
1548 return c == 'e' || c == 'E';
1549}
1550
1551bool NumericLiteralParser::GetFixedPointValue(llvm::APInt &StoreVal, unsigned Scale) {
1552 assert(radix == 16 || radix == 10);
1553
1554 // Find how many digits are needed to store the whole literal.
1555 unsigned NumDigits = SuffixBegin - DigitsBegin;
1556 if (saw_period) --NumDigits;
1557
1558 // Initial scan of the exponent if it exists
1559 bool ExpOverflowOccurred = false;
1560 bool NegativeExponent = false;
1561 const char *ExponentBegin;
1562 uint64_t Exponent = 0;
1563 int64_t BaseShift = 0;
1564 if (saw_exponent) {
1565 const char *Ptr = DigitsBegin;
1566
1567 while (!IsExponentPart(c: *Ptr, isHex: radix == 16))
1568 ++Ptr;
1569 ExponentBegin = Ptr;
1570 ++Ptr;
1571 NegativeExponent = *Ptr == '-';
1572 if (NegativeExponent) ++Ptr;
1573
1574 unsigned NumExpDigits = SuffixBegin - Ptr;
1575 if (alwaysFitsInto64Bits(Radix: radix, NumDigits: NumExpDigits)) {
1576 llvm::StringRef ExpStr(Ptr, NumExpDigits);
1577 llvm::APInt ExpInt(/*numBits=*/64, ExpStr, /*radix=*/10);
1578 Exponent = ExpInt.getZExtValue();
1579 } else {
1580 ExpOverflowOccurred = true;
1581 }
1582
1583 if (NegativeExponent) BaseShift -= Exponent;
1584 else BaseShift += Exponent;
1585 }
1586
1587 // Number of bits needed for decimal literal is
1588 // ceil(NumDigits * log2(10)) Integral part
1589 // + Scale Fractional part
1590 // + ceil(Exponent * log2(10)) Exponent
1591 // --------------------------------------------------
1592 // ceil((NumDigits + Exponent) * log2(10)) + Scale
1593 //
1594 // But for simplicity in handling integers, we can round up log2(10) to 4,
1595 // making:
1596 // 4 * (NumDigits + Exponent) + Scale
1597 //
1598 // Number of digits needed for hexadecimal literal is
1599 // 4 * NumDigits Integral part
1600 // + Scale Fractional part
1601 // + Exponent Exponent
1602 // --------------------------------------------------
1603 // (4 * NumDigits) + Scale + Exponent
1604 uint64_t NumBitsNeeded;
1605 if (radix == 10)
1606 NumBitsNeeded = 4 * (NumDigits + Exponent) + Scale;
1607 else
1608 NumBitsNeeded = 4 * NumDigits + Exponent + Scale;
1609
1610 if (NumBitsNeeded > std::numeric_limits<unsigned>::max())
1611 ExpOverflowOccurred = true;
1612 llvm::APInt Val(static_cast<unsigned>(NumBitsNeeded), 0, /*isSigned=*/false);
1613
1614 bool FoundDecimal = false;
1615
1616 int64_t FractBaseShift = 0;
1617 const char *End = saw_exponent ? ExponentBegin : SuffixBegin;
1618 for (const char *Ptr = DigitsBegin; Ptr < End; ++Ptr) {
1619 if (*Ptr == '.') {
1620 FoundDecimal = true;
1621 continue;
1622 }
1623
1624 // Normal reading of an integer
1625 unsigned C = llvm::hexDigitValue(C: *Ptr);
1626 assert(C < radix && "NumericLiteralParser ctor should have rejected this");
1627
1628 Val *= radix;
1629 Val += C;
1630
1631 if (FoundDecimal)
1632 // Keep track of how much we will need to adjust this value by from the
1633 // number of digits past the radix point.
1634 --FractBaseShift;
1635 }
1636
1637 // For a radix of 16, we will be multiplying by 2 instead of 16.
1638 if (radix == 16) FractBaseShift *= 4;
1639 BaseShift += FractBaseShift;
1640
1641 Val <<= Scale;
1642
1643 uint64_t Base = (radix == 16) ? 2 : 10;
1644 if (BaseShift > 0) {
1645 for (int64_t i = 0; i < BaseShift; ++i) {
1646 Val *= Base;
1647 }
1648 } else if (BaseShift < 0) {
1649 for (int64_t i = BaseShift; i < 0 && !Val.isZero(); ++i)
1650 Val = Val.udiv(RHS: Base);
1651 }
1652
1653 bool IntOverflowOccurred = false;
1654 auto MaxVal = llvm::APInt::getMaxValue(numBits: StoreVal.getBitWidth());
1655 if (Val.getBitWidth() > StoreVal.getBitWidth()) {
1656 IntOverflowOccurred |= Val.ugt(RHS: MaxVal.zext(width: Val.getBitWidth()));
1657 StoreVal = Val.trunc(width: StoreVal.getBitWidth());
1658 } else if (Val.getBitWidth() < StoreVal.getBitWidth()) {
1659 IntOverflowOccurred |= Val.zext(width: MaxVal.getBitWidth()).ugt(RHS: MaxVal);
1660 StoreVal = Val.zext(width: StoreVal.getBitWidth());
1661 } else {
1662 StoreVal = Val;
1663 }
1664
1665 return IntOverflowOccurred || ExpOverflowOccurred;
1666}
1667
1668/// \verbatim
1669/// user-defined-character-literal: [C++11 lex.ext]
1670/// character-literal ud-suffix
1671/// ud-suffix:
1672/// identifier
1673/// character-literal: [C++11 lex.ccon]
1674/// ' c-char-sequence '
1675/// u' c-char-sequence '
1676/// U' c-char-sequence '
1677/// L' c-char-sequence '
1678/// u8' c-char-sequence ' [C++1z lex.ccon]
1679/// c-char-sequence:
1680/// c-char
1681/// c-char-sequence c-char
1682/// c-char:
1683/// any member of the source character set except the single-quote ',
1684/// backslash \, or new-line character
1685/// escape-sequence
1686/// universal-character-name
1687/// escape-sequence:
1688/// simple-escape-sequence
1689/// octal-escape-sequence
1690/// hexadecimal-escape-sequence
1691/// simple-escape-sequence:
1692/// one of \' \" \? \\ \a \b \f \n \r \t \v
1693/// octal-escape-sequence:
1694/// \ octal-digit
1695/// \ octal-digit octal-digit
1696/// \ octal-digit octal-digit octal-digit
1697/// hexadecimal-escape-sequence:
1698/// \x hexadecimal-digit
1699/// hexadecimal-escape-sequence hexadecimal-digit
1700/// universal-character-name: [C++11 lex.charset]
1701/// \u hex-quad
1702/// \U hex-quad hex-quad
1703/// hex-quad:
1704/// hex-digit hex-digit hex-digit hex-digit
1705/// \endverbatim
1706///
1707CharLiteralParser::CharLiteralParser(const char *begin, const char *end,
1708 SourceLocation Loc, Preprocessor &PP,
1709 tok::TokenKind kind) {
1710 // At this point we know that the character matches the regex "(L|u|U)?'.*'".
1711 HadError = false;
1712
1713 Kind = kind;
1714
1715 const char *TokBegin = begin;
1716
1717 // Skip over wide character determinant.
1718 if (Kind != tok::char_constant)
1719 ++begin;
1720 if (Kind == tok::utf8_char_constant)
1721 ++begin;
1722
1723 // Skip over the entry quote.
1724 if (begin[0] != '\'') {
1725 PP.Diag(Loc, DiagID: diag::err_lexing_char);
1726 HadError = true;
1727 return;
1728 }
1729
1730 ++begin;
1731
1732 // Remove an optional ud-suffix.
1733 if (end[-1] != '\'') {
1734 const char *UDSuffixEnd = end;
1735 do {
1736 --end;
1737 } while (end[-1] != '\'');
1738 // FIXME: Don't bother with this if !tok.hasUCN().
1739 expandUCNs(Buf&: UDSuffixBuf, Input: StringRef(end, UDSuffixEnd - end));
1740 UDSuffixOffset = end - TokBegin;
1741 }
1742
1743 // Trim the ending quote.
1744 assert(end != begin && "Invalid token lexed");
1745 --end;
1746
1747 // FIXME: The "Value" is an uint64_t so we can handle char literals of
1748 // up to 64-bits.
1749 // FIXME: This extensively assumes that 'char' is 8-bits.
1750 assert(PP.getTargetInfo().getCharWidth() == 8 &&
1751 "Assumes char is 8 bits");
1752 assert(PP.getTargetInfo().getIntWidth() <= 64 &&
1753 (PP.getTargetInfo().getIntWidth() & 7) == 0 &&
1754 "Assumes sizeof(int) on target is <= 64 and a multiple of char");
1755 assert(PP.getTargetInfo().getWCharWidth() <= 64 &&
1756 "Assumes sizeof(wchar) on target is <= 64");
1757
1758 SmallVector<uint32_t, 4> codepoint_buffer;
1759 codepoint_buffer.resize(N: end - begin);
1760 uint32_t *buffer_begin = &codepoint_buffer.front();
1761 uint32_t *buffer_end = buffer_begin + codepoint_buffer.size();
1762
1763 // Unicode escapes representing characters that cannot be correctly
1764 // represented in a single code unit are disallowed in character literals
1765 // by this implementation.
1766 uint32_t largest_character_for_kind;
1767 if (tok::wide_char_constant == Kind) {
1768 largest_character_for_kind =
1769 0xFFFFFFFFu >> (32-PP.getTargetInfo().getWCharWidth());
1770 } else if (tok::utf8_char_constant == Kind) {
1771 largest_character_for_kind = 0x7F;
1772 } else if (tok::utf16_char_constant == Kind) {
1773 largest_character_for_kind = 0xFFFF;
1774 } else if (tok::utf32_char_constant == Kind) {
1775 largest_character_for_kind = 0x10FFFF;
1776 } else {
1777 largest_character_for_kind = 0x7Fu;
1778 }
1779
1780 while (begin != end) {
1781 // Is this a span of non-escape characters?
1782 if (begin[0] != '\\') {
1783 char const *start = begin;
1784 do {
1785 ++begin;
1786 } while (begin != end && *begin != '\\');
1787
1788 char const *tmp_in_start = start;
1789 uint32_t *tmp_out_start = buffer_begin;
1790 llvm::ConversionResult res =
1791 llvm::ConvertUTF8toUTF32(sourceStart: reinterpret_cast<llvm::UTF8 const **>(&start),
1792 sourceEnd: reinterpret_cast<llvm::UTF8 const *>(begin),
1793 targetStart: &buffer_begin, targetEnd: buffer_end, flags: llvm::strictConversion);
1794 if (res != llvm::conversionOK) {
1795 // If we see bad encoding for unprefixed character literals, warn and
1796 // simply copy the byte values, for compatibility with gcc and
1797 // older versions of clang.
1798 bool NoErrorOnBadEncoding = isOrdinary();
1799 unsigned Msg = diag::err_bad_character_encoding;
1800 if (NoErrorOnBadEncoding)
1801 Msg = diag::warn_bad_character_encoding;
1802 PP.Diag(Loc, DiagID: Msg);
1803 if (NoErrorOnBadEncoding) {
1804 start = tmp_in_start;
1805 buffer_begin = tmp_out_start;
1806 for (; start != begin; ++start, ++buffer_begin)
1807 *buffer_begin = static_cast<uint8_t>(*start);
1808 } else {
1809 HadError = true;
1810 }
1811 } else {
1812 for (; tmp_out_start < buffer_begin; ++tmp_out_start) {
1813 if (*tmp_out_start > largest_character_for_kind) {
1814 HadError = true;
1815 PP.Diag(Loc, DiagID: diag::err_character_too_large);
1816 }
1817 }
1818 }
1819
1820 continue;
1821 }
1822 // Is this a Universal Character Name escape?
1823 if (begin[1] == 'u' || begin[1] == 'U' || begin[1] == 'N') {
1824 unsigned short UcnLen = 0;
1825 if (!ProcessUCNEscape(ThisTokBegin: TokBegin, ThisTokBuf&: begin, ThisTokEnd: end, UcnVal&: *buffer_begin, UcnLen,
1826 Loc: FullSourceLoc(Loc, PP.getSourceManager()),
1827 Diags: &PP.getDiagnostics(), Features: PP.getLangOpts(), in_char_string_literal: true)) {
1828 HadError = true;
1829 } else if (*buffer_begin > largest_character_for_kind) {
1830 HadError = true;
1831 PP.Diag(Loc, DiagID: diag::err_character_too_large);
1832 }
1833
1834 ++buffer_begin;
1835 continue;
1836 }
1837 unsigned CharWidth = getCharWidth(kind: Kind, Target: PP.getTargetInfo());
1838 uint64_t result =
1839 ProcessCharEscape(ThisTokBegin: TokBegin, ThisTokBuf&: begin, ThisTokEnd: end, HadError,
1840 Loc: FullSourceLoc(Loc, PP.getSourceManager()), CharWidth,
1841 Diags: &PP.getDiagnostics(), Features: PP.getLangOpts(),
1842 EvalMethod: StringLiteralEvalMethod::Evaluated);
1843 *buffer_begin++ = result;
1844 }
1845
1846 unsigned NumCharsSoFar = buffer_begin - &codepoint_buffer.front();
1847
1848 if (NumCharsSoFar > 1) {
1849 if (isOrdinary() && NumCharsSoFar == 4)
1850 PP.Diag(Loc, DiagID: diag::warn_four_char_character_literal);
1851 else if (isOrdinary())
1852 PP.Diag(Loc, DiagID: diag::warn_multichar_character_literal);
1853 else {
1854 PP.Diag(Loc, DiagID: diag::err_multichar_character_literal) << (isWide() ? 0 : 1);
1855 HadError = true;
1856 }
1857 IsMultiChar = true;
1858 } else {
1859 IsMultiChar = false;
1860 }
1861
1862 llvm::APInt LitVal(PP.getTargetInfo().getIntWidth(), 0);
1863
1864 // Narrow character literals act as though their value is concatenated
1865 // in this implementation, but warn on overflow.
1866 bool multi_char_too_long = false;
1867 if (isOrdinary() && isMultiChar()) {
1868 LitVal = 0;
1869 for (size_t i = 0; i < NumCharsSoFar; ++i) {
1870 // check for enough leading zeros to shift into
1871 multi_char_too_long |= (LitVal.countl_zero() < 8);
1872 LitVal <<= 8;
1873 LitVal = LitVal + (codepoint_buffer[i] & 0xFF);
1874 }
1875 } else if (NumCharsSoFar > 0) {
1876 // otherwise just take the last character
1877 LitVal = buffer_begin[-1];
1878 }
1879
1880 if (!HadError && multi_char_too_long) {
1881 PP.Diag(Loc, DiagID: diag::warn_char_constant_too_large);
1882 }
1883
1884 // Transfer the value from APInt to uint64_t
1885 Value = LitVal.getZExtValue();
1886
1887 // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1")
1888 // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple
1889 // character constants are not sign extended in the this implementation:
1890 // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC.
1891 if (isOrdinary() && NumCharsSoFar == 1 && (Value & 128) &&
1892 PP.getLangOpts().CharIsSigned)
1893 Value = (signed char)Value;
1894}
1895
1896/// \verbatim
1897/// string-literal: [C++0x lex.string]
1898/// encoding-prefix " [s-char-sequence] "
1899/// encoding-prefix R raw-string
1900/// encoding-prefix:
1901/// u8
1902/// u
1903/// U
1904/// L
1905/// s-char-sequence:
1906/// s-char
1907/// s-char-sequence s-char
1908/// s-char:
1909/// any member of the source character set except the double-quote ",
1910/// backslash \, or new-line character
1911/// escape-sequence
1912/// universal-character-name
1913/// raw-string:
1914/// " d-char-sequence ( r-char-sequence ) d-char-sequence "
1915/// r-char-sequence:
1916/// r-char
1917/// r-char-sequence r-char
1918/// r-char:
1919/// any member of the source character set, except a right parenthesis )
1920/// followed by the initial d-char-sequence (which may be empty)
1921/// followed by a double quote ".
1922/// d-char-sequence:
1923/// d-char
1924/// d-char-sequence d-char
1925/// d-char:
1926/// any member of the basic source character set except:
1927/// space, the left parenthesis (, the right parenthesis ),
1928/// the backslash \, and the control characters representing horizontal
1929/// tab, vertical tab, form feed, and newline.
1930/// escape-sequence: [C++0x lex.ccon]
1931/// simple-escape-sequence
1932/// octal-escape-sequence
1933/// hexadecimal-escape-sequence
1934/// simple-escape-sequence:
1935/// one of \' \" \? \\ \a \b \f \n \r \t \v
1936/// octal-escape-sequence:
1937/// \ octal-digit
1938/// \ octal-digit octal-digit
1939/// \ octal-digit octal-digit octal-digit
1940/// hexadecimal-escape-sequence:
1941/// \x hexadecimal-digit
1942/// hexadecimal-escape-sequence hexadecimal-digit
1943/// universal-character-name:
1944/// \u hex-quad
1945/// \U hex-quad hex-quad
1946/// hex-quad:
1947/// hex-digit hex-digit hex-digit hex-digit
1948/// \endverbatim
1949///
1950StringLiteralParser::StringLiteralParser(ArrayRef<Token> StringToks,
1951 Preprocessor &PP,
1952 StringLiteralEvalMethod EvalMethod)
1953 : SM(PP.getSourceManager()), Features(PP.getLangOpts()),
1954 Target(PP.getTargetInfo()), Diags(&PP.getDiagnostics()),
1955 MaxTokenLength(0), SizeBound(0), CharByteWidth(0), Kind(tok::unknown),
1956 ResultPtr(ResultBuf.data()), EvalMethod(EvalMethod), hadError(false),
1957 Pascal(false) {
1958 init(StringToks);
1959}
1960
1961void StringLiteralParser::init(ArrayRef<Token> StringToks){
1962 // The literal token may have come from an invalid source location (e.g. due
1963 // to a PCH error), in which case the token length will be 0.
1964 if (StringToks.empty() || StringToks[0].getLength() < 2)
1965 return DiagnoseLexingError(Loc: SourceLocation());
1966
1967 // Scan all of the string portions, remember the max individual token length,
1968 // computing a bound on the concatenated string length, and see whether any
1969 // piece is a wide-string. If any of the string portions is a wide-string
1970 // literal, the result is a wide-string literal [C99 6.4.5p4].
1971 assert(!StringToks.empty() && "expected at least one token");
1972 MaxTokenLength = StringToks[0].getLength();
1973 assert(StringToks[0].getLength() >= 2 && "literal token is invalid!");
1974 SizeBound = StringToks[0].getLength() - 2; // -2 for "".
1975 hadError = false;
1976
1977 // Determines the kind of string from the prefix
1978 Kind = tok::string_literal;
1979
1980 /// (C99 5.1.1.2p1). The common case is only one string fragment.
1981 for (const Token &Tok : StringToks) {
1982 if (Tok.getLength() < 2)
1983 return DiagnoseLexingError(Loc: Tok.getLocation());
1984
1985 // The string could be shorter than this if it needs cleaning, but this is a
1986 // reasonable bound, which is all we need.
1987 assert(Tok.getLength() >= 2 && "literal token is invalid!");
1988 SizeBound += Tok.getLength() - 2; // -2 for "".
1989
1990 // Remember maximum string piece length.
1991 if (Tok.getLength() > MaxTokenLength)
1992 MaxTokenLength = Tok.getLength();
1993
1994 // Remember if we see any wide or utf-8/16/32 strings.
1995 // Also check for illegal concatenations.
1996 if (isUnevaluated() && Tok.getKind() != tok::string_literal) {
1997 if (Diags) {
1998 SourceLocation PrefixEndLoc = Lexer::AdvanceToTokenCharacter(
1999 TokStart: Tok.getLocation(), Characters: getEncodingPrefixLen(kind: Tok.getKind()), SM,
2000 LangOpts: Features);
2001 CharSourceRange Range =
2002 CharSourceRange::getCharRange(R: {Tok.getLocation(), PrefixEndLoc});
2003 StringRef Prefix(SM.getCharacterData(SL: Tok.getLocation()),
2004 getEncodingPrefixLen(kind: Tok.getKind()));
2005 Diags->Report(Loc: Tok.getLocation(),
2006 DiagID: Features.CPlusPlus26
2007 ? diag::err_unevaluated_string_prefix
2008 : diag::warn_unevaluated_string_prefix)
2009 << Prefix << Features.CPlusPlus << FixItHint::CreateRemoval(RemoveRange: Range);
2010 }
2011 if (Features.CPlusPlus26)
2012 hadError = true;
2013 } else if (Tok.isNot(K: Kind) && Tok.isNot(K: tok::string_literal)) {
2014 if (isOrdinary()) {
2015 Kind = Tok.getKind();
2016 } else {
2017 if (Diags)
2018 Diags->Report(Loc: Tok.getLocation(), DiagID: diag::err_unsupported_string_concat);
2019 hadError = true;
2020 }
2021 }
2022 }
2023
2024 // Include space for the null terminator.
2025 ++SizeBound;
2026
2027 // TODO: K&R warning: "traditional C rejects string constant concatenation"
2028
2029 // Get the width in bytes of char/wchar_t/char16_t/char32_t
2030 CharByteWidth = getCharWidth(kind: Kind, Target);
2031 assert((CharByteWidth & 7) == 0 && "Assumes character size is byte multiple");
2032 CharByteWidth /= 8;
2033
2034 // The output buffer size needs to be large enough to hold wide characters.
2035 // This is a worst-case assumption which basically corresponds to L"" "long".
2036 SizeBound *= CharByteWidth;
2037
2038 // Size the temporary buffer to hold the result string data.
2039 ResultBuf.resize(N: SizeBound);
2040
2041 // Likewise, but for each string piece.
2042 SmallString<512> TokenBuf;
2043 TokenBuf.resize(N: MaxTokenLength);
2044
2045 // Loop over all the strings, getting their spelling, and expanding them to
2046 // wide strings as appropriate.
2047 ResultPtr = &ResultBuf[0]; // Next byte to fill in.
2048
2049 Pascal = false;
2050
2051 SourceLocation UDSuffixTokLoc;
2052
2053 for (unsigned i = 0, e = StringToks.size(); i != e; ++i) {
2054 const char *ThisTokBuf = &TokenBuf[0];
2055 // Get the spelling of the token, which eliminates trigraphs, etc. We know
2056 // that ThisTokBuf points to a buffer that is big enough for the whole token
2057 // and 'spelled' tokens can only shrink.
2058 bool StringInvalid = false;
2059 unsigned ThisTokLen =
2060 Lexer::getSpelling(Tok: StringToks[i], Buffer&: ThisTokBuf, SourceMgr: SM, LangOpts: Features,
2061 Invalid: &StringInvalid);
2062 if (StringInvalid)
2063 return DiagnoseLexingError(Loc: StringToks[i].getLocation());
2064
2065 const char *ThisTokBegin = ThisTokBuf;
2066 const char *ThisTokEnd = ThisTokBuf+ThisTokLen;
2067
2068 // Remove an optional ud-suffix.
2069 if (ThisTokEnd[-1] != '"') {
2070 const char *UDSuffixEnd = ThisTokEnd;
2071 do {
2072 --ThisTokEnd;
2073 } while (ThisTokEnd[-1] != '"');
2074
2075 StringRef UDSuffix(ThisTokEnd, UDSuffixEnd - ThisTokEnd);
2076
2077 if (UDSuffixBuf.empty()) {
2078 if (StringToks[i].hasUCN())
2079 expandUCNs(Buf&: UDSuffixBuf, Input: UDSuffix);
2080 else
2081 UDSuffixBuf.assign(RHS: UDSuffix);
2082 UDSuffixToken = i;
2083 UDSuffixOffset = ThisTokEnd - ThisTokBuf;
2084 UDSuffixTokLoc = StringToks[i].getLocation();
2085 } else {
2086 SmallString<32> ExpandedUDSuffix;
2087 if (StringToks[i].hasUCN()) {
2088 expandUCNs(Buf&: ExpandedUDSuffix, Input: UDSuffix);
2089 UDSuffix = ExpandedUDSuffix;
2090 }
2091
2092 // C++11 [lex.ext]p8: At the end of phase 6, if a string literal is the
2093 // result of a concatenation involving at least one user-defined-string-
2094 // literal, all the participating user-defined-string-literals shall
2095 // have the same ud-suffix.
2096 bool UnevaluatedStringHasUDL = isUnevaluated() && !UDSuffix.empty();
2097 if (UDSuffixBuf != UDSuffix || UnevaluatedStringHasUDL) {
2098 if (Diags) {
2099 SourceLocation TokLoc = StringToks[i].getLocation();
2100 if (UnevaluatedStringHasUDL) {
2101 Diags->Report(Loc: TokLoc, DiagID: diag::err_unevaluated_string_udl)
2102 << SourceRange(TokLoc, TokLoc);
2103 } else {
2104 Diags->Report(Loc: TokLoc, DiagID: diag::err_string_concat_mixed_suffix)
2105 << UDSuffixBuf << UDSuffix
2106 << SourceRange(UDSuffixTokLoc, UDSuffixTokLoc);
2107 }
2108 }
2109 hadError = true;
2110 }
2111 }
2112 }
2113
2114 // Strip the end quote.
2115 --ThisTokEnd;
2116
2117 // TODO: Input character set mapping support.
2118
2119 // Skip marker for wide or unicode strings.
2120 if (ThisTokBuf[0] == 'L' || ThisTokBuf[0] == 'u' || ThisTokBuf[0] == 'U') {
2121 ++ThisTokBuf;
2122 // Skip 8 of u8 marker for utf8 strings.
2123 if (ThisTokBuf[0] == '8')
2124 ++ThisTokBuf;
2125 }
2126
2127 // Check for raw string
2128 if (ThisTokBuf[0] == 'R') {
2129 if (ThisTokBuf[1] != '"') {
2130 // The file may have come from PCH and then changed after loading the
2131 // PCH; Fail gracefully.
2132 return DiagnoseLexingError(Loc: StringToks[i].getLocation());
2133 }
2134 ThisTokBuf += 2; // skip R"
2135
2136 // C++11 [lex.string]p2: A `d-char-sequence` shall consist of at most 16
2137 // characters.
2138 constexpr unsigned MaxRawStrDelimLen = 16;
2139
2140 const char *Prefix = ThisTokBuf;
2141 while (static_cast<unsigned>(ThisTokBuf - Prefix) < MaxRawStrDelimLen &&
2142 ThisTokBuf[0] != '(')
2143 ++ThisTokBuf;
2144 if (ThisTokBuf[0] != '(')
2145 return DiagnoseLexingError(Loc: StringToks[i].getLocation());
2146 ++ThisTokBuf; // skip '('
2147
2148 // Remove same number of characters from the end
2149 ThisTokEnd -= ThisTokBuf - Prefix;
2150 if (ThisTokEnd < ThisTokBuf)
2151 return DiagnoseLexingError(Loc: StringToks[i].getLocation());
2152
2153 // C++14 [lex.string]p4: A source-file new-line in a raw string literal
2154 // results in a new-line in the resulting execution string-literal.
2155 StringRef RemainingTokenSpan(ThisTokBuf, ThisTokEnd - ThisTokBuf);
2156 while (!RemainingTokenSpan.empty()) {
2157 // Split the string literal on \r\n boundaries.
2158 size_t CRLFPos = RemainingTokenSpan.find(Str: "\r\n");
2159 StringRef BeforeCRLF = RemainingTokenSpan.substr(Start: 0, N: CRLFPos);
2160 StringRef AfterCRLF = RemainingTokenSpan.substr(Start: CRLFPos);
2161
2162 // Copy everything before the \r\n sequence into the string literal.
2163 if (CopyStringFragment(Tok: StringToks[i], TokBegin: ThisTokBegin, Fragment: BeforeCRLF))
2164 hadError = true;
2165
2166 // Point into the \n inside the \r\n sequence and operate on the
2167 // remaining portion of the literal.
2168 RemainingTokenSpan = AfterCRLF.substr(Start: 1);
2169 }
2170 } else {
2171 if (ThisTokBuf[0] != '"') {
2172 // The file may have come from PCH and then changed after loading the
2173 // PCH; Fail gracefully.
2174 return DiagnoseLexingError(Loc: StringToks[i].getLocation());
2175 }
2176 ++ThisTokBuf; // skip "
2177
2178 // Check if this is a pascal string
2179 if (!isUnevaluated() && Features.PascalStrings &&
2180 ThisTokBuf + 1 != ThisTokEnd && ThisTokBuf[0] == '\\' &&
2181 ThisTokBuf[1] == 'p') {
2182
2183 // If the \p sequence is found in the first token, we have a pascal string
2184 // Otherwise, if we already have a pascal string, ignore the first \p
2185 if (i == 0) {
2186 ++ThisTokBuf;
2187 Pascal = true;
2188 } else if (Pascal)
2189 ThisTokBuf += 2;
2190 }
2191
2192 while (ThisTokBuf != ThisTokEnd) {
2193 // Is this a span of non-escape characters?
2194 if (ThisTokBuf[0] != '\\') {
2195 const char *InStart = ThisTokBuf;
2196 do {
2197 ++ThisTokBuf;
2198 } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\');
2199
2200 // Copy the character span over.
2201 if (CopyStringFragment(Tok: StringToks[i], TokBegin: ThisTokBegin,
2202 Fragment: StringRef(InStart, ThisTokBuf - InStart)))
2203 hadError = true;
2204 continue;
2205 }
2206 // Is this a Universal Character Name escape?
2207 if (ThisTokBuf[1] == 'u' || ThisTokBuf[1] == 'U' ||
2208 ThisTokBuf[1] == 'N') {
2209 EncodeUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
2210 ResultBuf&: ResultPtr, HadError&: hadError,
2211 Loc: FullSourceLoc(StringToks[i].getLocation(), SM),
2212 CharByteWidth, Diags, Features);
2213 continue;
2214 }
2215 // Otherwise, this is a non-UCN escape character. Process it.
2216 unsigned ResultChar =
2217 ProcessCharEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, HadError&: hadError,
2218 Loc: FullSourceLoc(StringToks[i].getLocation(), SM),
2219 CharWidth: CharByteWidth * 8, Diags, Features, EvalMethod);
2220
2221 if (CharByteWidth == 4) {
2222 // FIXME: Make the type of the result buffer correct instead of
2223 // using reinterpret_cast.
2224 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultPtr);
2225 *ResultWidePtr = ResultChar;
2226 ResultPtr += 4;
2227 } else if (CharByteWidth == 2) {
2228 // FIXME: Make the type of the result buffer correct instead of
2229 // using reinterpret_cast.
2230 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultPtr);
2231 *ResultWidePtr = ResultChar & 0xFFFF;
2232 ResultPtr += 2;
2233 } else {
2234 assert(CharByteWidth == 1 && "Unexpected char width");
2235 *ResultPtr++ = ResultChar & 0xFF;
2236 }
2237 }
2238 }
2239 }
2240
2241 assert((!Pascal || !isUnevaluated()) &&
2242 "Pascal string in unevaluated context");
2243 if (Pascal) {
2244 if (CharByteWidth == 4) {
2245 // FIXME: Make the type of the result buffer correct instead of
2246 // using reinterpret_cast.
2247 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultBuf.data());
2248 ResultWidePtr[0] = GetNumStringChars() - 1;
2249 } else if (CharByteWidth == 2) {
2250 // FIXME: Make the type of the result buffer correct instead of
2251 // using reinterpret_cast.
2252 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultBuf.data());
2253 ResultWidePtr[0] = GetNumStringChars() - 1;
2254 } else {
2255 assert(CharByteWidth == 1 && "Unexpected char width");
2256 ResultBuf[0] = GetNumStringChars() - 1;
2257 }
2258
2259 // Verify that pascal strings aren't too large.
2260 if (GetStringLength() > 256) {
2261 if (Diags)
2262 Diags->Report(Loc: StringToks.front().getLocation(),
2263 DiagID: diag::err_pascal_string_too_long)
2264 << SourceRange(StringToks.front().getLocation(),
2265 StringToks.back().getLocation());
2266 hadError = true;
2267 return;
2268 }
2269 } else if (Diags) {
2270 // Complain if this string literal has too many characters.
2271 unsigned MaxChars = Features.CPlusPlus? 65536 : Features.C99 ? 4095 : 509;
2272
2273 if (GetNumStringChars() > MaxChars)
2274 Diags->Report(Loc: StringToks.front().getLocation(),
2275 DiagID: diag::ext_string_too_long)
2276 << GetNumStringChars() << MaxChars
2277 << (Features.CPlusPlus ? 2 : Features.C99 ? 1 : 0)
2278 << SourceRange(StringToks.front().getLocation(),
2279 StringToks.back().getLocation());
2280 }
2281}
2282
2283static const char *resyncUTF8(const char *Err, const char *End) {
2284 if (Err == End)
2285 return End;
2286 End = Err + std::min<unsigned>(a: llvm::getNumBytesForUTF8(firstByte: *Err), b: End-Err);
2287 while (++Err != End && (*Err & 0xC0) == 0x80)
2288 ;
2289 return Err;
2290}
2291
2292/// This function copies from Fragment, which is a sequence of bytes
2293/// within Tok's contents (which begin at TokBegin) into ResultPtr.
2294/// Performs widening for multi-byte characters.
2295bool StringLiteralParser::CopyStringFragment(const Token &Tok,
2296 const char *TokBegin,
2297 StringRef Fragment) {
2298 const llvm::UTF8 *ErrorPtrTmp;
2299 if (ConvertUTF8toWide(WideCharWidth: CharByteWidth, Source: Fragment, ResultPtr, ErrorPtr&: ErrorPtrTmp))
2300 return false;
2301
2302 // If we see bad encoding for unprefixed string literals, warn and
2303 // simply copy the byte values, for compatibility with gcc and older
2304 // versions of clang.
2305 bool NoErrorOnBadEncoding = isOrdinary();
2306 if (NoErrorOnBadEncoding) {
2307 memcpy(dest: ResultPtr, src: Fragment.data(), n: Fragment.size());
2308 ResultPtr += Fragment.size();
2309 }
2310
2311 if (Diags) {
2312 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
2313
2314 FullSourceLoc SourceLoc(Tok.getLocation(), SM);
2315 const DiagnosticBuilder &Builder =
2316 Diag(Diags, Features, TokLoc: SourceLoc, TokBegin,
2317 TokRangeBegin: ErrorPtr, TokRangeEnd: resyncUTF8(Err: ErrorPtr, End: Fragment.end()),
2318 DiagID: NoErrorOnBadEncoding ? diag::warn_bad_string_encoding
2319 : diag::err_bad_string_encoding);
2320
2321 const char *NextStart = resyncUTF8(Err: ErrorPtr, End: Fragment.end());
2322 StringRef NextFragment(NextStart, Fragment.end()-NextStart);
2323
2324 // Decode into a dummy buffer.
2325 SmallString<512> Dummy;
2326 Dummy.reserve(N: Fragment.size() * CharByteWidth);
2327 char *Ptr = Dummy.data();
2328
2329 while (!ConvertUTF8toWide(WideCharWidth: CharByteWidth, Source: NextFragment, ResultPtr&: Ptr, ErrorPtr&: ErrorPtrTmp)) {
2330 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
2331 NextStart = resyncUTF8(Err: ErrorPtr, End: Fragment.end());
2332 Builder << MakeCharSourceRange(Features, TokLoc: SourceLoc, TokBegin,
2333 TokRangeBegin: ErrorPtr, TokRangeEnd: NextStart);
2334 NextFragment = StringRef(NextStart, Fragment.end()-NextStart);
2335 }
2336 }
2337 return !NoErrorOnBadEncoding;
2338}
2339
2340void StringLiteralParser::DiagnoseLexingError(SourceLocation Loc) {
2341 hadError = true;
2342 if (Diags)
2343 Diags->Report(Loc, DiagID: diag::err_lexing_string);
2344}
2345
2346/// getOffsetOfStringByte - This function returns the offset of the
2347/// specified byte of the string data represented by Token. This handles
2348/// advancing over escape sequences in the string.
2349unsigned StringLiteralParser::getOffsetOfStringByte(const Token &Tok,
2350 unsigned ByteNo) const {
2351 // Get the spelling of the token.
2352 SmallString<32> SpellingBuffer;
2353 SpellingBuffer.resize(N: Tok.getLength());
2354
2355 bool StringInvalid = false;
2356 const char *SpellingPtr = &SpellingBuffer[0];
2357 unsigned TokLen = Lexer::getSpelling(Tok, Buffer&: SpellingPtr, SourceMgr: SM, LangOpts: Features,
2358 Invalid: &StringInvalid);
2359 if (StringInvalid)
2360 return 0;
2361
2362 const char *SpellingStart = SpellingPtr;
2363 const char *SpellingEnd = SpellingPtr+TokLen;
2364
2365 // Handle UTF-8 strings just like narrow strings.
2366 if (SpellingPtr[0] == 'u' && SpellingPtr[1] == '8')
2367 SpellingPtr += 2;
2368
2369 assert(SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' &&
2370 SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet");
2371
2372 // For raw string literals, this is easy.
2373 if (SpellingPtr[0] == 'R') {
2374 assert(SpellingPtr[1] == '"' && "Should be a raw string literal!");
2375 // Skip 'R"'.
2376 SpellingPtr += 2;
2377 while (*SpellingPtr != '(') {
2378 ++SpellingPtr;
2379 assert(SpellingPtr < SpellingEnd && "Missing ( for raw string literal");
2380 }
2381 // Skip '('.
2382 ++SpellingPtr;
2383 return SpellingPtr - SpellingStart + ByteNo;
2384 }
2385
2386 // Skip over the leading quote
2387 assert(SpellingPtr[0] == '"' && "Should be a string literal!");
2388 ++SpellingPtr;
2389
2390 // Skip over bytes until we find the offset we're looking for.
2391 while (ByteNo) {
2392 assert(SpellingPtr < SpellingEnd && "Didn't find byte offset!");
2393
2394 // Step over non-escapes simply.
2395 if (*SpellingPtr != '\\') {
2396 ++SpellingPtr;
2397 --ByteNo;
2398 continue;
2399 }
2400
2401 // Otherwise, this is an escape character. Advance over it.
2402 bool HadError = false;
2403 if (SpellingPtr[1] == 'u' || SpellingPtr[1] == 'U' ||
2404 SpellingPtr[1] == 'N') {
2405 const char *EscapePtr = SpellingPtr;
2406 unsigned Len = MeasureUCNEscape(ThisTokBegin: SpellingStart, ThisTokBuf&: SpellingPtr, ThisTokEnd: SpellingEnd,
2407 CharByteWidth: 1, Features, HadError);
2408 if (Len > ByteNo) {
2409 // ByteNo is somewhere within the escape sequence.
2410 SpellingPtr = EscapePtr;
2411 break;
2412 }
2413 ByteNo -= Len;
2414 } else {
2415 ProcessCharEscape(ThisTokBegin: SpellingStart, ThisTokBuf&: SpellingPtr, ThisTokEnd: SpellingEnd, HadError,
2416 Loc: FullSourceLoc(Tok.getLocation(), SM), CharWidth: CharByteWidth * 8,
2417 Diags, Features, EvalMethod: StringLiteralEvalMethod::Evaluated);
2418 --ByteNo;
2419 }
2420 assert(!HadError && "This method isn't valid on erroneous strings");
2421 }
2422
2423 return SpellingPtr-SpellingStart;
2424}
2425
2426/// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
2427/// suffixes as ud-suffixes, because the diagnostic experience is better if we
2428/// treat it as an invalid suffix.
2429bool StringLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
2430 StringRef Suffix) {
2431 return NumericLiteralParser::isValidUDSuffix(LangOpts, Suffix) ||
2432 Suffix == "sv";
2433}
2434