Bug Summary

File:build/source/clang/lib/Lex/LiteralSupport.cpp
Warning:line 1036, column 11
Value stored to 'HasSize' is never read

Annotated Source Code

Press '?' to see keyboard shortcuts

clang -cc1 -cc1 -triple x86_64-pc-linux-gnu -analyze -disable-free -clear-ast-before-backend -disable-llvm-verifier -discard-value-names -main-file-name LiteralSupport.cpp -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -analyzer-config-compatibility-mode=true -mrelocation-model pic -pic-level 2 -mframe-pointer=none -relaxed-aliasing -fmath-errno -ffp-contract=on -fno-rounding-math -mconstructor-aliases -funwind-tables=2 -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -ffunction-sections -fdata-sections -fcoverage-compilation-dir=/build/source/build-llvm/tools/clang/stage2-bins -resource-dir /usr/lib/llvm-17/lib/clang/17 -D CLANG_REPOSITORY_STRING="++20230510111145+7df43bdb42ae-1~exp1~20230510111303.1288" -D _DEBUG -D _GLIBCXX_ASSERTIONS -D _GNU_SOURCE -D _LIBCPP_ENABLE_ASSERTIONS -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D __STDC_LIMIT_MACROS -I tools/clang/lib/Lex -I /build/source/clang/lib/Lex -I /build/source/clang/include -I tools/clang/include -I include -I /build/source/llvm/include -D _FORTIFY_SOURCE=2 -D NDEBUG -U NDEBUG -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/x86_64-linux-gnu/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10/backward -internal-isystem /usr/lib/llvm-17/lib/clang/17/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -fmacro-prefix-map=/build/source/build-llvm/tools/clang/stage2-bins=build-llvm/tools/clang/stage2-bins -fmacro-prefix-map=/build/source/= -fcoverage-prefix-map=/build/source/build-llvm/tools/clang/stage2-bins=build-llvm/tools/clang/stage2-bins -fcoverage-prefix-map=/build/source/= -source-date-epoch 1683717183 -O2 -Wno-unused-command-line-argument -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-maybe-uninitialized -Wno-class-memaccess -Wno-redundant-move -Wno-pessimizing-move -Wno-noexcept-type -Wno-comment -Wno-misleading-indentation -std=c++17 -fdeprecated-macro -fdebug-compilation-dir=/build/source/build-llvm/tools/clang/stage2-bins -fdebug-prefix-map=/build/source/build-llvm/tools/clang/stage2-bins=build-llvm/tools/clang/stage2-bins -fdebug-prefix-map=/build/source/= -ferror-limit 19 -fvisibility-inlines-hidden -stack-protector 2 -fgnuc-version=4.2.1 -fcolor-diagnostics -vectorize-loops -vectorize-slp -analyzer-output=html -analyzer-config stable-report-filename=true -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/scan-build-2023-05-10-133810-16478-1 -x c++ /build/source/clang/lib/Lex/LiteralSupport.cpp
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!")::llvm::llvm_unreachable_internal("Unknown token type!", "clang/lib/Lex/LiteralSupport.cpp"
, 42)
;
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 CharSourceRange MakeCharSourceRange(const LangOptions &Features,
61 FullSourceLoc TokLoc,
62 const char *TokBegin,
63 const char *TokRangeBegin,
64 const char *TokRangeEnd) {
65 SourceLocation Begin =
66 Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
67 TokLoc.getManager(), Features);
68 SourceLocation End =
69 Lexer::AdvanceToTokenCharacter(Begin, TokRangeEnd - TokRangeBegin,
70 TokLoc.getManager(), Features);
71 return CharSourceRange::getCharRange(Begin, End);
72}
73
74/// Produce a diagnostic highlighting some portion of a literal.
75///
76/// Emits the diagnostic \p DiagID, highlighting the range of characters from
77/// \p TokRangeBegin (inclusive) to \p TokRangeEnd (exclusive), which must be
78/// a substring of a spelling buffer for the token beginning at \p TokBegin.
79static DiagnosticBuilder Diag(DiagnosticsEngine *Diags,
80 const LangOptions &Features, FullSourceLoc TokLoc,
81 const char *TokBegin, const char *TokRangeBegin,
82 const char *TokRangeEnd, unsigned DiagID) {
83 SourceLocation Begin =
84 Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
85 TokLoc.getManager(), Features);
86 return Diags->Report(Begin, DiagID) <<
87 MakeCharSourceRange(Features, TokLoc, TokBegin, TokRangeBegin, TokRangeEnd);
88}
89
90/// ProcessCharEscape - Parse a standard C escape sequence, which can occur in
91/// either a character or a string literal.
92static unsigned ProcessCharEscape(const char *ThisTokBegin,
93 const char *&ThisTokBuf,
94 const char *ThisTokEnd, bool &HadError,
95 FullSourceLoc Loc, unsigned CharWidth,
96 DiagnosticsEngine *Diags,
97 const LangOptions &Features) {
98 const char *EscapeBegin = ThisTokBuf;
99 bool Delimited = false;
100 bool EndDelimiterFound = false;
101
102 // Skip the '\' char.
103 ++ThisTokBuf;
104
105 // We know that this character can't be off the end of the buffer, because
106 // that would have been \", which would not have been the end of string.
107 unsigned ResultChar = *ThisTokBuf++;
108 switch (ResultChar) {
109 // These map to themselves.
110 case '\\': case '\'': case '"': case '?': break;
111
112 // These have fixed mappings.
113 case 'a':
114 // TODO: K&R: the meaning of '\\a' is different in traditional C
115 ResultChar = 7;
116 break;
117 case 'b':
118 ResultChar = 8;
119 break;
120 case 'e':
121 if (Diags)
122 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
123 diag::ext_nonstandard_escape) << "e";
124 ResultChar = 27;
125 break;
126 case 'E':
127 if (Diags)
128 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
129 diag::ext_nonstandard_escape) << "E";
130 ResultChar = 27;
131 break;
132 case 'f':
133 ResultChar = 12;
134 break;
135 case 'n':
136 ResultChar = 10;
137 break;
138 case 'r':
139 ResultChar = 13;
140 break;
141 case 't':
142 ResultChar = 9;
143 break;
144 case 'v':
145 ResultChar = 11;
146 break;
147 case 'x': { // Hex escape.
148 ResultChar = 0;
149 if (ThisTokBuf != ThisTokEnd && *ThisTokBuf == '{') {
150 Delimited = true;
151 ThisTokBuf++;
152 if (*ThisTokBuf == '}') {
153 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
154 diag::err_delimited_escape_empty);
155 return ResultChar;
156 }
157 } else if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
158 if (Diags)
159 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
160 diag::err_hex_escape_no_digits) << "x";
161 return ResultChar;
162 }
163
164 // Hex escapes are a maximal series of hex digits.
165 bool Overflow = false;
166 for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) {
167 if (Delimited && *ThisTokBuf == '}') {
168 ThisTokBuf++;
169 EndDelimiterFound = true;
170 break;
171 }
172 int CharVal = llvm::hexDigitValue(*ThisTokBuf);
173 if (CharVal == -1) {
174 // Non delimited hex escape sequences stop at the first non-hex digit.
175 if (!Delimited)
176 break;
177 HadError = true;
178 if (Diags)
179 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
180 diag::err_delimited_escape_invalid)
181 << StringRef(ThisTokBuf, 1);
182 continue;
183 }
184 // About to shift out a digit?
185 if (ResultChar & 0xF0000000)
186 Overflow = true;
187 ResultChar <<= 4;
188 ResultChar |= CharVal;
189 }
190 // See if any bits will be truncated when evaluated as a character.
191 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
192 Overflow = true;
193 ResultChar &= ~0U >> (32-CharWidth);
194 }
195
196 // Check for overflow.
197 if (!HadError && Overflow) { // Too many digits to fit in
198 HadError = true;
199 if (Diags)
200 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
201 diag::err_escape_too_large)
202 << 0;
203 }
204 break;
205 }
206 case '0': case '1': case '2': case '3':
207 case '4': case '5': case '6': case '7': {
208 // Octal escapes.
209 --ThisTokBuf;
210 ResultChar = 0;
211
212 // Octal escapes are a series of octal digits with maximum length 3.
213 // "\0123" is a two digit sequence equal to "\012" "3".
214 unsigned NumDigits = 0;
215 do {
216 ResultChar <<= 3;
217 ResultChar |= *ThisTokBuf++ - '0';
218 ++NumDigits;
219 } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 &&
220 ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7');
221
222 // Check for overflow. Reject '\777', but not L'\777'.
223 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
224 if (Diags)
225 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
226 diag::err_escape_too_large) << 1;
227 ResultChar &= ~0U >> (32-CharWidth);
228 }
229 break;
230 }
231 case 'o': {
232 bool Overflow = false;
233 if (ThisTokBuf == ThisTokEnd || *ThisTokBuf != '{') {
234 HadError = true;
235 if (Diags)
236 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
237 diag::err_delimited_escape_missing_brace)
238 << "o";
239
240 break;
241 }
242 ResultChar = 0;
243 Delimited = true;
244 ++ThisTokBuf;
245 if (*ThisTokBuf == '}') {
246 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
247 diag::err_delimited_escape_empty);
248 return ResultChar;
249 }
250
251 while (ThisTokBuf != ThisTokEnd) {
252 if (*ThisTokBuf == '}') {
253 EndDelimiterFound = true;
254 ThisTokBuf++;
255 break;
256 }
257 if (*ThisTokBuf < '0' || *ThisTokBuf > '7') {
258 HadError = true;
259 if (Diags)
260 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
261 diag::err_delimited_escape_invalid)
262 << StringRef(ThisTokBuf, 1);
263 ThisTokBuf++;
264 continue;
265 }
266 // Check if one of the top three bits is set before shifting them out.
267 if (ResultChar & 0xE0000000)
268 Overflow = true;
269
270 ResultChar <<= 3;
271 ResultChar |= *ThisTokBuf++ - '0';
272 }
273 // Check for overflow. Reject '\777', but not L'\777'.
274 if (!HadError &&
275 (Overflow || (CharWidth != 32 && (ResultChar >> CharWidth) != 0))) {
276 HadError = true;
277 if (Diags)
278 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
279 diag::err_escape_too_large)
280 << 1;
281 ResultChar &= ~0U >> (32 - CharWidth);
282 }
283 break;
284 }
285 // Otherwise, these are not valid escapes.
286 case '(': case '{': case '[': case '%':
287 // GCC accepts these as extensions. We warn about them as such though.
288 if (Diags)
289 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
290 diag::ext_nonstandard_escape)
291 << std::string(1, ResultChar);
292 break;
293 default:
294 if (!Diags)
295 break;
296
297 if (isPrintable(ResultChar))
298 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
299 diag::ext_unknown_escape)
300 << std::string(1, ResultChar);
301 else
302 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
303 diag::ext_unknown_escape)
304 << "x" + llvm::utohexstr(ResultChar);
305 break;
306 }
307
308 if (Delimited && Diags) {
309 if (!EndDelimiterFound)
310 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
311 diag::err_expected)
312 << tok::r_brace;
313 else if (!HadError) {
314 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
315 Features.CPlusPlus23 ? diag::warn_cxx23_delimited_escape_sequence
316 : diag::ext_delimited_escape_sequence)
317 << /*delimited*/ 0 << (Features.CPlusPlus ? 1 : 0);
318 }
319 }
320
321 return ResultChar;
322}
323
324static void appendCodePoint(unsigned Codepoint,
325 llvm::SmallVectorImpl<char> &Str) {
326 char ResultBuf[4];
327 char *ResultPtr = ResultBuf;
328 if (llvm::ConvertCodePointToUTF8(Codepoint, ResultPtr))
329 Str.append(ResultBuf, ResultPtr);
330}
331
332void clang::expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
333 for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
334 if (*I != '\\') {
335 Buf.push_back(*I);
336 continue;
337 }
338
339 ++I;
340 char Kind = *I;
341 ++I;
342
343 assert(Kind == 'u' || Kind == 'U' || Kind == 'N')(static_cast <bool> (Kind == 'u' || Kind == 'U' || Kind
== 'N') ? void (0) : __assert_fail ("Kind == 'u' || Kind == 'U' || Kind == 'N'"
, "clang/lib/Lex/LiteralSupport.cpp", 343, __extension__ __PRETTY_FUNCTION__
))
;
344 uint32_t CodePoint = 0;
345
346 if (Kind == 'u' && *I == '{') {
347 for (++I; *I != '}'; ++I) {
348 unsigned Value = llvm::hexDigitValue(*I);
349 assert(Value != -1U)(static_cast <bool> (Value != -1U) ? void (0) : __assert_fail
("Value != -1U", "clang/lib/Lex/LiteralSupport.cpp", 349, __extension__
__PRETTY_FUNCTION__))
;
350 CodePoint <<= 4;
351 CodePoint += Value;
352 }
353 appendCodePoint(CodePoint, Buf);
354 continue;
355 }
356
357 if (Kind == 'N') {
358 assert(*I == '{')(static_cast <bool> (*I == '{') ? void (0) : __assert_fail
("*I == '{'", "clang/lib/Lex/LiteralSupport.cpp", 358, __extension__
__PRETTY_FUNCTION__))
;
359 ++I;
360 auto Delim = std::find(I, Input.end(), '}');
361 assert(Delim != Input.end())(static_cast <bool> (Delim != Input.end()) ? void (0) :
__assert_fail ("Delim != Input.end()", "clang/lib/Lex/LiteralSupport.cpp"
, 361, __extension__ __PRETTY_FUNCTION__))
;
362 std::optional<llvm::sys::unicode::LooseMatchingResult> Res =
363 llvm::sys::unicode::nameToCodepointLooseMatching(
364 StringRef(I, std::distance(I, Delim)));
365 assert(Res)(static_cast <bool> (Res) ? void (0) : __assert_fail ("Res"
, "clang/lib/Lex/LiteralSupport.cpp", 365, __extension__ __PRETTY_FUNCTION__
))
;
366 CodePoint = Res->CodePoint;
367 assert(CodePoint != 0xFFFFFFFF)(static_cast <bool> (CodePoint != 0xFFFFFFFF) ? void (0
) : __assert_fail ("CodePoint != 0xFFFFFFFF", "clang/lib/Lex/LiteralSupport.cpp"
, 367, __extension__ __PRETTY_FUNCTION__))
;
368 appendCodePoint(CodePoint, Buf);
369 I = Delim;
370 continue;
371 }
372
373 unsigned NumHexDigits;
374 if (Kind == 'u')
375 NumHexDigits = 4;
376 else
377 NumHexDigits = 8;
378
379 assert(I + NumHexDigits <= E)(static_cast <bool> (I + NumHexDigits <= E) ? void (
0) : __assert_fail ("I + NumHexDigits <= E", "clang/lib/Lex/LiteralSupport.cpp"
, 379, __extension__ __PRETTY_FUNCTION__))
;
380
381 for (; NumHexDigits != 0; ++I, --NumHexDigits) {
382 unsigned Value = llvm::hexDigitValue(*I);
383 assert(Value != -1U)(static_cast <bool> (Value != -1U) ? void (0) : __assert_fail
("Value != -1U", "clang/lib/Lex/LiteralSupport.cpp", 383, __extension__
__PRETTY_FUNCTION__))
;
384
385 CodePoint <<= 4;
386 CodePoint += Value;
387 }
388
389 appendCodePoint(CodePoint, Buf);
390 --I;
391 }
392}
393
394static bool ProcessNumericUCNEscape(const char *ThisTokBegin,
395 const char *&ThisTokBuf,
396 const char *ThisTokEnd, uint32_t &UcnVal,
397 unsigned short &UcnLen, bool &Delimited,
398 FullSourceLoc Loc, DiagnosticsEngine *Diags,
399 const LangOptions &Features,
400 bool in_char_string_literal = false) {
401 const char *UcnBegin = ThisTokBuf;
402 bool HasError = false;
403 bool EndDelimiterFound = false;
404
405 // Skip the '\u' char's.
406 ThisTokBuf += 2;
407 Delimited = false;
408 if (UcnBegin[1] == 'u' && in_char_string_literal &&
409 ThisTokBuf != ThisTokEnd && *ThisTokBuf == '{') {
410 Delimited = true;
411 ThisTokBuf++;
412 } else if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
413 if (Diags)
414 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
415 diag::err_hex_escape_no_digits)
416 << StringRef(&ThisTokBuf[-1], 1);
417 return false;
418 }
419 UcnLen = (ThisTokBuf[-1] == 'u' ? 4 : 8);
420
421 bool Overflow = false;
422 unsigned short Count = 0;
423 for (; ThisTokBuf != ThisTokEnd && (Delimited || Count != UcnLen);
424 ++ThisTokBuf) {
425 if (Delimited && *ThisTokBuf == '}') {
426 ++ThisTokBuf;
427 EndDelimiterFound = true;
428 break;
429 }
430 int CharVal = llvm::hexDigitValue(*ThisTokBuf);
431 if (CharVal == -1) {
432 HasError = true;
433 if (!Delimited)
434 break;
435 if (Diags) {
436 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
437 diag::err_delimited_escape_invalid)
438 << StringRef(ThisTokBuf, 1);
439 }
440 Count++;
441 continue;
442 }
443 if (UcnVal & 0xF0000000) {
444 Overflow = true;
445 continue;
446 }
447 UcnVal <<= 4;
448 UcnVal |= CharVal;
449 Count++;
450 }
451
452 if (Overflow) {
453 if (Diags)
454 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
455 diag::err_escape_too_large)
456 << 0;
457 return false;
458 }
459
460 if (Delimited && !EndDelimiterFound) {
461 if (Diags) {
462 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
463 diag::err_expected)
464 << tok::r_brace;
465 }
466 return false;
467 }
468
469 // If we didn't consume the proper number of digits, there is a problem.
470 if (Count == 0 || (!Delimited && Count != UcnLen)) {
471 if (Diags)
472 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
473 Delimited ? diag::err_delimited_escape_empty
474 : diag::err_ucn_escape_incomplete);
475 return false;
476 }
477 return !HasError;
478}
479
480static void DiagnoseInvalidUnicodeCharacterName(
481 DiagnosticsEngine *Diags, const LangOptions &Features, FullSourceLoc Loc,
482 const char *TokBegin, const char *TokRangeBegin, const char *TokRangeEnd,
483 llvm::StringRef Name) {
484
485 Diag(Diags, Features, Loc, TokBegin, TokRangeBegin, TokRangeEnd,
486 diag::err_invalid_ucn_name)
487 << Name;
488
489 namespace u = llvm::sys::unicode;
490
491 std::optional<u::LooseMatchingResult> Res =
492 u::nameToCodepointLooseMatching(Name);
493 if (Res) {
494 Diag(Diags, Features, Loc, TokBegin, TokRangeBegin, TokRangeEnd,
495 diag::note_invalid_ucn_name_loose_matching)
496 << FixItHint::CreateReplacement(
497 MakeCharSourceRange(Features, Loc, TokBegin, TokRangeBegin,
498 TokRangeEnd),
499 Res->Name);
500 return;
501 }
502
503 unsigned Distance = 0;
504 SmallVector<u::MatchForCodepointName> Matches =
505 u::nearestMatchesForCodepointName(Name, 5);
506 assert(!Matches.empty() && "No unicode characters found")(static_cast <bool> (!Matches.empty() && "No unicode characters found"
) ? void (0) : __assert_fail ("!Matches.empty() && \"No unicode characters found\""
, "clang/lib/Lex/LiteralSupport.cpp", 506, __extension__ __PRETTY_FUNCTION__
))
;
507
508 for (const auto &Match : Matches) {
509 if (Distance == 0)
510 Distance = Match.Distance;
511 if (std::max(Distance, Match.Distance) -
512 std::min(Distance, Match.Distance) >
513 3)
514 break;
515 Distance = Match.Distance;
516
517 std::string Str;
518 llvm::UTF32 V = Match.Value;
519 bool Converted =
520 llvm::convertUTF32ToUTF8String(llvm::ArrayRef<llvm::UTF32>(&V, 1), Str);
521 (void)Converted;
522 assert(Converted && "Found a match wich is not a unicode character")(static_cast <bool> (Converted && "Found a match wich is not a unicode character"
) ? void (0) : __assert_fail ("Converted && \"Found a match wich is not a unicode character\""
, "clang/lib/Lex/LiteralSupport.cpp", 522, __extension__ __PRETTY_FUNCTION__
))
;
523
524 Diag(Diags, Features, Loc, TokBegin, TokRangeBegin, TokRangeEnd,
525 diag::note_invalid_ucn_name_candidate)
526 << Match.Name << llvm::utohexstr(Match.Value)
527 << Str // FIXME: Fix the rendering of non printable characters
528 << FixItHint::CreateReplacement(
529 MakeCharSourceRange(Features, Loc, TokBegin, TokRangeBegin,
530 TokRangeEnd),
531 Match.Name);
532 }
533}
534
535static bool ProcessNamedUCNEscape(const char *ThisTokBegin,
536 const char *&ThisTokBuf,
537 const char *ThisTokEnd, uint32_t &UcnVal,
538 unsigned short &UcnLen, FullSourceLoc Loc,
539 DiagnosticsEngine *Diags,
540 const LangOptions &Features) {
541 const char *UcnBegin = ThisTokBuf;
542 assert(UcnBegin[0] == '\\' && UcnBegin[1] == 'N')(static_cast <bool> (UcnBegin[0] == '\\' && UcnBegin
[1] == 'N') ? void (0) : __assert_fail ("UcnBegin[0] == '\\\\' && UcnBegin[1] == 'N'"
, "clang/lib/Lex/LiteralSupport.cpp", 542, __extension__ __PRETTY_FUNCTION__
))
;
543 ThisTokBuf += 2;
544 if (ThisTokBuf == ThisTokEnd || *ThisTokBuf != '{') {
545 if (Diags) {
546 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
547 diag::err_delimited_escape_missing_brace)
548 << StringRef(&ThisTokBuf[-1], 1);
549 }
550 return false;
551 }
552 ThisTokBuf++;
553 const char *ClosingBrace = std::find_if(ThisTokBuf, ThisTokEnd, [](char C) {
554 return C == '}' || isVerticalWhitespace(C);
555 });
556 bool Incomplete = ClosingBrace == ThisTokEnd;
557 bool Empty = ClosingBrace == ThisTokBuf;
558 if (Incomplete || Empty) {
559 if (Diags) {
560 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
561 Incomplete ? diag::err_ucn_escape_incomplete
562 : diag::err_delimited_escape_empty)
563 << StringRef(&UcnBegin[1], 1);
564 }
565 ThisTokBuf = ClosingBrace == ThisTokEnd ? ClosingBrace : ClosingBrace + 1;
566 return false;
567 }
568 StringRef Name(ThisTokBuf, ClosingBrace - ThisTokBuf);
569 ThisTokBuf = ClosingBrace + 1;
570 std::optional<char32_t> Res = llvm::sys::unicode::nameToCodepointStrict(Name);
571 if (!Res) {
572 if (Diags)
573 DiagnoseInvalidUnicodeCharacterName(Diags, Features, Loc, ThisTokBegin,
574 &UcnBegin[3], ClosingBrace, Name);
575 return false;
576 }
577 UcnVal = *Res;
578 UcnLen = UcnVal > 0xFFFF ? 8 : 4;
579 return true;
580}
581
582/// ProcessUCNEscape - Read the Universal Character Name, check constraints and
583/// return the UTF32.
584static bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
585 const char *ThisTokEnd, uint32_t &UcnVal,
586 unsigned short &UcnLen, FullSourceLoc Loc,
587 DiagnosticsEngine *Diags,
588 const LangOptions &Features,
589 bool in_char_string_literal = false) {
590
591 bool HasError;
592 const char *UcnBegin = ThisTokBuf;
593 bool IsDelimitedEscapeSequence = false;
594 bool IsNamedEscapeSequence = false;
595 if (ThisTokBuf[1] == 'N') {
596 IsNamedEscapeSequence = true;
597 HasError = !ProcessNamedUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
598 UcnVal, UcnLen, Loc, Diags, Features);
599 } else {
600 HasError =
601 !ProcessNumericUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
602 UcnLen, IsDelimitedEscapeSequence, Loc, Diags,
603 Features, in_char_string_literal);
604 }
605 if (HasError)
606 return false;
607
608 // Check UCN constraints (C99 6.4.3p2) [C++11 lex.charset p2]
609 if ((0xD800 <= UcnVal && UcnVal <= 0xDFFF) || // surrogate codepoints
610 UcnVal > 0x10FFFF) { // maximum legal UTF32 value
611 if (Diags)
612 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
613 diag::err_ucn_escape_invalid);
614 return false;
615 }
616
617 // C++11 allows UCNs that refer to control characters and basic source
618 // characters inside character and string literals
619 if (UcnVal < 0xa0 &&
620 (UcnVal != 0x24 && UcnVal != 0x40 && UcnVal != 0x60)) { // $, @, `
621 bool IsError = (!Features.CPlusPlus11 || !in_char_string_literal);
622 if (Diags) {
623 char BasicSCSChar = UcnVal;
624 if (UcnVal >= 0x20 && UcnVal < 0x7f)
625 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
626 IsError ? diag::err_ucn_escape_basic_scs :
627 diag::warn_cxx98_compat_literal_ucn_escape_basic_scs)
628 << StringRef(&BasicSCSChar, 1);
629 else
630 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
631 IsError ? diag::err_ucn_control_character :
632 diag::warn_cxx98_compat_literal_ucn_control_character);
633 }
634 if (IsError)
635 return false;
636 }
637
638 if (!Features.CPlusPlus && !Features.C99 && Diags)
639 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
640 diag::warn_ucn_not_valid_in_c89_literal);
641
642 if ((IsDelimitedEscapeSequence || IsNamedEscapeSequence) && Diags)
643 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
644 Features.CPlusPlus23 ? diag::warn_cxx23_delimited_escape_sequence
645 : diag::ext_delimited_escape_sequence)
646 << (IsNamedEscapeSequence ? 1 : 0) << (Features.CPlusPlus ? 1 : 0);
647
648 return true;
649}
650
651/// MeasureUCNEscape - Determine the number of bytes within the resulting string
652/// which this UCN will occupy.
653static int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
654 const char *ThisTokEnd, unsigned CharByteWidth,
655 const LangOptions &Features, bool &HadError) {
656 // UTF-32: 4 bytes per escape.
657 if (CharByteWidth == 4)
658 return 4;
659
660 uint32_t UcnVal = 0;
661 unsigned short UcnLen = 0;
662 FullSourceLoc Loc;
663
664 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
665 UcnLen, Loc, nullptr, Features, true)) {
666 HadError = true;
667 return 0;
668 }
669
670 // UTF-16: 2 bytes for BMP, 4 bytes otherwise.
671 if (CharByteWidth == 2)
672 return UcnVal <= 0xFFFF ? 2 : 4;
673
674 // UTF-8.
675 if (UcnVal < 0x80)
676 return 1;
677 if (UcnVal < 0x800)
678 return 2;
679 if (UcnVal < 0x10000)
680 return 3;
681 return 4;
682}
683
684/// EncodeUCNEscape - Read the Universal Character Name, check constraints and
685/// convert the UTF32 to UTF8 or UTF16. This is a subroutine of
686/// StringLiteralParser. When we decide to implement UCN's for identifiers,
687/// we will likely rework our support for UCN's.
688static void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
689 const char *ThisTokEnd,
690 char *&ResultBuf, bool &HadError,
691 FullSourceLoc Loc, unsigned CharByteWidth,
692 DiagnosticsEngine *Diags,
693 const LangOptions &Features) {
694 typedef uint32_t UTF32;
695 UTF32 UcnVal = 0;
696 unsigned short UcnLen = 0;
697 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, UcnLen,
698 Loc, Diags, Features, true)) {
699 HadError = true;
700 return;
701 }
702
703 assert((CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) &&(static_cast <bool> ((CharByteWidth == 1 || CharByteWidth
== 2 || CharByteWidth == 4) && "only character widths of 1, 2, or 4 bytes supported"
) ? void (0) : __assert_fail ("(CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) && \"only character widths of 1, 2, or 4 bytes supported\""
, "clang/lib/Lex/LiteralSupport.cpp", 704, __extension__ __PRETTY_FUNCTION__
))
704 "only character widths of 1, 2, or 4 bytes supported")(static_cast <bool> ((CharByteWidth == 1 || CharByteWidth
== 2 || CharByteWidth == 4) && "only character widths of 1, 2, or 4 bytes supported"
) ? void (0) : __assert_fail ("(CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) && \"only character widths of 1, 2, or 4 bytes supported\""
, "clang/lib/Lex/LiteralSupport.cpp", 704, __extension__ __PRETTY_FUNCTION__
))
;
705
706 (void)UcnLen;
707 assert((UcnLen== 4 || UcnLen== 8) && "only ucn length of 4 or 8 supported")(static_cast <bool> ((UcnLen== 4 || UcnLen== 8) &&
"only ucn length of 4 or 8 supported") ? void (0) : __assert_fail
("(UcnLen== 4 || UcnLen== 8) && \"only ucn length of 4 or 8 supported\""
, "clang/lib/Lex/LiteralSupport.cpp", 707, __extension__ __PRETTY_FUNCTION__
))
;
708
709 if (CharByteWidth == 4) {
710 // FIXME: Make the type of the result buffer correct instead of
711 // using reinterpret_cast.
712 llvm::UTF32 *ResultPtr = reinterpret_cast<llvm::UTF32*>(ResultBuf);
713 *ResultPtr = UcnVal;
714 ResultBuf += 4;
715 return;
716 }
717
718 if (CharByteWidth == 2) {
719 // FIXME: Make the type of the result buffer correct instead of
720 // using reinterpret_cast.
721 llvm::UTF16 *ResultPtr = reinterpret_cast<llvm::UTF16*>(ResultBuf);
722
723 if (UcnVal <= (UTF32)0xFFFF) {
724 *ResultPtr = UcnVal;
725 ResultBuf += 2;
726 return;
727 }
728
729 // Convert to UTF16.
730 UcnVal -= 0x10000;
731 *ResultPtr = 0xD800 + (UcnVal >> 10);
732 *(ResultPtr+1) = 0xDC00 + (UcnVal & 0x3FF);
733 ResultBuf += 4;
734 return;
735 }
736
737 assert(CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters")(static_cast <bool> (CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters"
) ? void (0) : __assert_fail ("CharByteWidth == 1 && \"UTF-8 encoding is only for 1 byte characters\""
, "clang/lib/Lex/LiteralSupport.cpp", 737, __extension__ __PRETTY_FUNCTION__
))
;
738
739 // Now that we've parsed/checked the UCN, we convert from UTF32->UTF8.
740 // The conversion below was inspired by:
741 // http://www.unicode.org/Public/PROGRAMS/CVTUTF/ConvertUTF.c
742 // First, we determine how many bytes the result will require.
743 typedef uint8_t UTF8;
744
745 unsigned short bytesToWrite = 0;
746 if (UcnVal < (UTF32)0x80)
747 bytesToWrite = 1;
748 else if (UcnVal < (UTF32)0x800)
749 bytesToWrite = 2;
750 else if (UcnVal < (UTF32)0x10000)
751 bytesToWrite = 3;
752 else
753 bytesToWrite = 4;
754
755 const unsigned byteMask = 0xBF;
756 const unsigned byteMark = 0x80;
757
758 // Once the bits are split out into bytes of UTF8, this is a mask OR-ed
759 // into the first byte, depending on how many bytes follow.
760 static const UTF8 firstByteMark[5] = {
761 0x00, 0x00, 0xC0, 0xE0, 0xF0
762 };
763 // Finally, we write the bytes into ResultBuf.
764 ResultBuf += bytesToWrite;
765 switch (bytesToWrite) { // note: everything falls through.
766 case 4:
767 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
768 [[fallthrough]];
769 case 3:
770 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
771 [[fallthrough]];
772 case 2:
773 *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
774 [[fallthrough]];
775 case 1:
776 *--ResultBuf = (UTF8) (UcnVal | firstByteMark[bytesToWrite]);
777 }
778 // Update the buffer.
779 ResultBuf += bytesToWrite;
780}
781
782/// integer-constant: [C99 6.4.4.1]
783/// decimal-constant integer-suffix
784/// octal-constant integer-suffix
785/// hexadecimal-constant integer-suffix
786/// binary-literal integer-suffix [GNU, C++1y]
787/// user-defined-integer-literal: [C++11 lex.ext]
788/// decimal-literal ud-suffix
789/// octal-literal ud-suffix
790/// hexadecimal-literal ud-suffix
791/// binary-literal ud-suffix [GNU, C++1y]
792/// decimal-constant:
793/// nonzero-digit
794/// decimal-constant digit
795/// octal-constant:
796/// 0
797/// octal-constant octal-digit
798/// hexadecimal-constant:
799/// hexadecimal-prefix hexadecimal-digit
800/// hexadecimal-constant hexadecimal-digit
801/// hexadecimal-prefix: one of
802/// 0x 0X
803/// binary-literal:
804/// 0b binary-digit
805/// 0B binary-digit
806/// binary-literal binary-digit
807/// integer-suffix:
808/// unsigned-suffix [long-suffix]
809/// unsigned-suffix [long-long-suffix]
810/// long-suffix [unsigned-suffix]
811/// long-long-suffix [unsigned-sufix]
812/// nonzero-digit:
813/// 1 2 3 4 5 6 7 8 9
814/// octal-digit:
815/// 0 1 2 3 4 5 6 7
816/// hexadecimal-digit:
817/// 0 1 2 3 4 5 6 7 8 9
818/// a b c d e f
819/// A B C D E F
820/// binary-digit:
821/// 0
822/// 1
823/// unsigned-suffix: one of
824/// u U
825/// long-suffix: one of
826/// l L
827/// long-long-suffix: one of
828/// ll LL
829///
830/// floating-constant: [C99 6.4.4.2]
831/// TODO: add rules...
832///
833NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
834 SourceLocation TokLoc,
835 const SourceManager &SM,
836 const LangOptions &LangOpts,
837 const TargetInfo &Target,
838 DiagnosticsEngine &Diags)
839 : SM(SM), LangOpts(LangOpts), Diags(Diags),
840 ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) {
841
842 s = DigitsBegin = ThisTokBegin;
843 saw_exponent = false;
844 saw_period = false;
845 saw_ud_suffix = false;
846 saw_fixed_point_suffix = false;
847 isLong = false;
848 isUnsigned = false;
849 isLongLong = false;
850 isSizeT = false;
851 isHalf = false;
852 isFloat = false;
853 isImaginary = false;
854 isFloat16 = false;
855 isFloat128 = false;
856 MicrosoftInteger = 0;
857 isFract = false;
858 isAccum = false;
859 hadError = false;
860 isBitInt = false;
861
862 // This routine assumes that the range begin/end matches the regex for integer
863 // and FP constants (specifically, the 'pp-number' regex), and assumes that
864 // the byte at "*end" is both valid and not part of the regex. Because of
865 // this, it doesn't have to check for 'overscan' in various places.
866 if (isPreprocessingNumberBody(*ThisTokEnd)) {
867 Diags.Report(TokLoc, diag::err_lexing_numeric);
868 hadError = true;
869 return;
870 }
871
872 if (*s == '0') { // parse radix
873 ParseNumberStartingWithZero(TokLoc);
874 if (hadError)
875 return;
876 } else { // the first digit is non-zero
877 radix = 10;
878 s = SkipDigits(s);
879 if (s == ThisTokEnd) {
880 // Done.
881 } else {
882 ParseDecimalOrOctalCommon(TokLoc);
883 if (hadError)
884 return;
885 }
886 }
887
888 SuffixBegin = s;
889 checkSeparator(TokLoc, s, CSK_AfterDigits);
890
891 // Initial scan to lookahead for fixed point suffix.
892 if (LangOpts.FixedPoint) {
893 for (const char *c = s; c != ThisTokEnd; ++c) {
894 if (*c == 'r' || *c == 'k' || *c == 'R' || *c == 'K') {
895 saw_fixed_point_suffix = true;
896 break;
897 }
898 }
899 }
900
901 // Parse the suffix. At this point we can classify whether we have an FP or
902 // integer constant.
903 bool isFixedPointConstant = isFixedPointLiteral();
904 bool isFPConstant = isFloatingLiteral();
905 bool HasSize = false;
906
907 // Loop over all of the characters of the suffix. If we see something bad,
908 // we break out of the loop.
909 for (; s != ThisTokEnd; ++s) {
910 switch (*s) {
911 case 'R':
912 case 'r':
913 if (!LangOpts.FixedPoint)
914 break;
915 if (isFract || isAccum) break;
916 if (!(saw_period || saw_exponent)) break;
917 isFract = true;
918 continue;
919 case 'K':
920 case 'k':
921 if (!LangOpts.FixedPoint)
922 break;
923 if (isFract || isAccum) break;
924 if (!(saw_period || saw_exponent)) break;
925 isAccum = true;
926 continue;
927 case 'h': // FP Suffix for "half".
928 case 'H':
929 // OpenCL Extension v1.2 s9.5 - h or H suffix for half type.
930 if (!(LangOpts.Half || LangOpts.FixedPoint))
931 break;
932 if (isIntegerLiteral()) break; // Error for integer constant.
933 if (HasSize)
934 break;
935 HasSize = true;
936 isHalf = true;
937 continue; // Success.
938 case 'f': // FP Suffix for "float"
939 case 'F':
940 if (!isFPConstant) break; // Error for integer constant.
941 if (HasSize)
942 break;
943 HasSize = true;
944
945 // CUDA host and device may have different _Float16 support, therefore
946 // allows f16 literals to avoid false alarm.
947 // When we compile for OpenMP target offloading on NVPTX, f16 suffix
948 // should also be supported.
949 // ToDo: more precise check for CUDA.
950 // TODO: AMDGPU might also support it in the future.
951 if ((Target.hasFloat16Type() || LangOpts.CUDA ||
952 (LangOpts.OpenMPIsDevice && Target.getTriple().isNVPTX())) &&
953 s + 2 < ThisTokEnd && s[1] == '1' && s[2] == '6') {
954 s += 2; // success, eat up 2 characters.
955 isFloat16 = true;
956 continue;
957 }
958
959 isFloat = true;
960 continue; // Success.
961 case 'q': // FP Suffix for "__float128"
962 case 'Q':
963 if (!isFPConstant) break; // Error for integer constant.
964 if (HasSize)
965 break;
966 HasSize = true;
967 isFloat128 = true;
968 continue; // Success.
969 case 'u':
970 case 'U':
971 if (isFPConstant) break; // Error for floating constant.
972 if (isUnsigned) break; // Cannot be repeated.
973 isUnsigned = true;
974 continue; // Success.
975 case 'l':
976 case 'L':
977 if (HasSize)
978 break;
979 HasSize = true;
980
981 // Check for long long. The L's need to be adjacent and the same case.
982 if (s[1] == s[0]) {
983 assert(s + 1 < ThisTokEnd && "didn't maximally munch?")(static_cast <bool> (s + 1 < ThisTokEnd && "didn't maximally munch?"
) ? void (0) : __assert_fail ("s + 1 < ThisTokEnd && \"didn't maximally munch?\""
, "clang/lib/Lex/LiteralSupport.cpp", 983, __extension__ __PRETTY_FUNCTION__
))
;
984 if (isFPConstant) break; // long long invalid for floats.
985 isLongLong = true;
986 ++s; // Eat both of them.
987 } else {
988 isLong = true;
989 }
990 continue; // Success.
991 case 'z':
992 case 'Z':
993 if (isFPConstant)
994 break; // Invalid for floats.
995 if (HasSize)
996 break;
997 HasSize = true;
998 isSizeT = true;
999 continue;
1000 case 'i':
1001 case 'I':
1002 if (LangOpts.MicrosoftExt && !isFPConstant) {
1003 // Allow i8, i16, i32, and i64. First, look ahead and check if
1004 // suffixes are Microsoft integers and not the imaginary unit.
1005 uint8_t Bits = 0;
1006 size_t ToSkip = 0;
1007 switch (s[1]) {
1008 case '8': // i8 suffix
1009 Bits = 8;
1010 ToSkip = 2;
1011 break;
1012 case '1':
1013 if (s[2] == '6') { // i16 suffix
1014 Bits = 16;
1015 ToSkip = 3;
1016 }
1017 break;
1018 case '3':
1019 if (s[2] == '2') { // i32 suffix
1020 Bits = 32;
1021 ToSkip = 3;
1022 }
1023 break;
1024 case '6':
1025 if (s[2] == '4') { // i64 suffix
1026 Bits = 64;
1027 ToSkip = 3;
1028 }
1029 break;
1030 default:
1031 break;
1032 }
1033 if (Bits) {
1034 if (HasSize)
1035 break;
1036 HasSize = true;
Value stored to 'HasSize' is never read
1037 MicrosoftInteger = Bits;
1038 s += ToSkip;
1039 assert(s <= ThisTokEnd && "didn't maximally munch?")(static_cast <bool> (s <= ThisTokEnd && "didn't maximally munch?"
) ? void (0) : __assert_fail ("s <= ThisTokEnd && \"didn't maximally munch?\""
, "clang/lib/Lex/LiteralSupport.cpp", 1039, __extension__ __PRETTY_FUNCTION__
))
;
1040 break;
1041 }
1042 }
1043 [[fallthrough]];
1044 case 'j':
1045 case 'J':
1046 if (isImaginary) break; // Cannot be repeated.
1047 isImaginary = true;
1048 continue; // Success.
1049 case 'w':
1050 case 'W':
1051 if (isFPConstant)
1052 break; // Invalid for floats.
1053 if (HasSize)
1054 break; // Invalid if we already have a size for the literal.
1055
1056 // wb and WB are allowed, but a mixture of cases like Wb or wB is not. We
1057 // explicitly do not support the suffix in C++ as an extension because a
1058 // library-based UDL that resolves to a library type may be more
1059 // appropriate there.
1060 if (!LangOpts.CPlusPlus && ((s[0] == 'w' && s[1] == 'b') ||
1061 (s[0] == 'W' && s[1] == 'B'))) {
1062 isBitInt = true;
1063 HasSize = true;
1064 ++s; // Skip both characters (2nd char skipped on continue).
1065 continue; // Success.
1066 }
1067 }
1068 // If we reached here, there was an error or a ud-suffix.
1069 break;
1070 }
1071
1072 // "i", "if", and "il" are user-defined suffixes in C++1y.
1073 if (s != ThisTokEnd || isImaginary) {
1074 // FIXME: Don't bother expanding UCNs if !tok.hasUCN().
1075 expandUCNs(UDSuffixBuf, StringRef(SuffixBegin, ThisTokEnd - SuffixBegin));
1076 if (isValidUDSuffix(LangOpts, UDSuffixBuf)) {
1077 if (!isImaginary) {
1078 // Any suffix pieces we might have parsed are actually part of the
1079 // ud-suffix.
1080 isLong = false;
1081 isUnsigned = false;
1082 isLongLong = false;
1083 isSizeT = false;
1084 isFloat = false;
1085 isFloat16 = false;
1086 isHalf = false;
1087 isImaginary = false;
1088 isBitInt = false;
1089 MicrosoftInteger = 0;
1090 saw_fixed_point_suffix = false;
1091 isFract = false;
1092 isAccum = false;
1093 }
1094
1095 saw_ud_suffix = true;
1096 return;
1097 }
1098
1099 if (s != ThisTokEnd) {
1100 // Report an error if there are any.
1101 Diags.Report(Lexer::AdvanceToTokenCharacter(
1102 TokLoc, SuffixBegin - ThisTokBegin, SM, LangOpts),
1103 diag::err_invalid_suffix_constant)
1104 << StringRef(SuffixBegin, ThisTokEnd - SuffixBegin)
1105 << (isFixedPointConstant ? 2 : isFPConstant);
1106 hadError = true;
1107 }
1108 }
1109
1110 if (!hadError && saw_fixed_point_suffix) {
1111 assert(isFract || isAccum)(static_cast <bool> (isFract || isAccum) ? void (0) : __assert_fail
("isFract || isAccum", "clang/lib/Lex/LiteralSupport.cpp", 1111
, __extension__ __PRETTY_FUNCTION__))
;
1112 }
1113}
1114
1115/// ParseDecimalOrOctalCommon - This method is called for decimal or octal
1116/// numbers. It issues an error for illegal digits, and handles floating point
1117/// parsing. If it detects a floating point number, the radix is set to 10.
1118void NumericLiteralParser::ParseDecimalOrOctalCommon(SourceLocation TokLoc){
1119 assert((radix == 8 || radix == 10) && "Unexpected radix")(static_cast <bool> ((radix == 8 || radix == 10) &&
"Unexpected radix") ? void (0) : __assert_fail ("(radix == 8 || radix == 10) && \"Unexpected radix\""
, "clang/lib/Lex/LiteralSupport.cpp", 1119, __extension__ __PRETTY_FUNCTION__
))
;
1120
1121 // If we have a hex digit other than 'e' (which denotes a FP exponent) then
1122 // the code is using an incorrect base.
1123 if (isHexDigit(*s) && *s != 'e' && *s != 'E' &&
1124 !isValidUDSuffix(LangOpts, StringRef(s, ThisTokEnd - s))) {
1125 Diags.Report(
1126 Lexer::AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin, SM, LangOpts),
1127 diag::err_invalid_digit)
1128 << StringRef(s, 1) << (radix == 8 ? 1 : 0);
1129 hadError = true;
1130 return;
1131 }
1132
1133 if (*s == '.') {
1134 checkSeparator(TokLoc, s, CSK_AfterDigits);
1135 s++;
1136 radix = 10;
1137 saw_period = true;
1138 checkSeparator(TokLoc, s, CSK_BeforeDigits);
1139 s = SkipDigits(s); // Skip suffix.
1140 }
1141 if (*s == 'e' || *s == 'E') { // exponent
1142 checkSeparator(TokLoc, s, CSK_AfterDigits);
1143 const char *Exponent = s;
1144 s++;
1145 radix = 10;
1146 saw_exponent = true;
1147 if (s != ThisTokEnd && (*s == '+' || *s == '-')) s++; // sign
1148 const char *first_non_digit = SkipDigits(s);
1149 if (containsDigits(s, first_non_digit)) {
1150 checkSeparator(TokLoc, s, CSK_BeforeDigits);
1151 s = first_non_digit;
1152 } else {
1153 if (!hadError) {
1154 Diags.Report(Lexer::AdvanceToTokenCharacter(
1155 TokLoc, Exponent - ThisTokBegin, SM, LangOpts),
1156 diag::err_exponent_has_no_digits);
1157 hadError = true;
1158 }
1159 return;
1160 }
1161 }
1162}
1163
1164/// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
1165/// suffixes as ud-suffixes, because the diagnostic experience is better if we
1166/// treat it as an invalid suffix.
1167bool NumericLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
1168 StringRef Suffix) {
1169 if (!LangOpts.CPlusPlus11 || Suffix.empty())
1170 return false;
1171
1172 // By C++11 [lex.ext]p10, ud-suffixes starting with an '_' are always valid.
1173 if (Suffix[0] == '_')
1174 return true;
1175
1176 // In C++11, there are no library suffixes.
1177 if (!LangOpts.CPlusPlus14)
1178 return false;
1179
1180 // In C++14, "s", "h", "min", "ms", "us", and "ns" are used in the library.
1181 // Per tweaked N3660, "il", "i", and "if" are also used in the library.
1182 // In C++2a "d" and "y" are used in the library.
1183 return llvm::StringSwitch<bool>(Suffix)
1184 .Cases("h", "min", "s", true)
1185 .Cases("ms", "us", "ns", true)
1186 .Cases("il", "i", "if", true)
1187 .Cases("d", "y", LangOpts.CPlusPlus20)
1188 .Default(false);
1189}
1190
1191void NumericLiteralParser::checkSeparator(SourceLocation TokLoc,
1192 const char *Pos,
1193 CheckSeparatorKind IsAfterDigits) {
1194 if (IsAfterDigits == CSK_AfterDigits) {
1195 if (Pos == ThisTokBegin)
1196 return;
1197 --Pos;
1198 } else if (Pos == ThisTokEnd)
1199 return;
1200
1201 if (isDigitSeparator(*Pos)) {
1202 Diags.Report(Lexer::AdvanceToTokenCharacter(TokLoc, Pos - ThisTokBegin, SM,
1203 LangOpts),
1204 diag::err_digit_separator_not_between_digits)
1205 << IsAfterDigits;
1206 hadError = true;
1207 }
1208}
1209
1210/// ParseNumberStartingWithZero - This method is called when the first character
1211/// of the number is found to be a zero. This means it is either an octal
1212/// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or
1213/// a floating point number (01239.123e4). Eat the prefix, determining the
1214/// radix etc.
1215void NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) {
1216 assert(s[0] == '0' && "Invalid method call")(static_cast <bool> (s[0] == '0' && "Invalid method call"
) ? void (0) : __assert_fail ("s[0] == '0' && \"Invalid method call\""
, "clang/lib/Lex/LiteralSupport.cpp", 1216, __extension__ __PRETTY_FUNCTION__
))
;
1217 s++;
1218
1219 int c1 = s[0];
1220
1221 // Handle a hex number like 0x1234.
1222 if ((c1 == 'x' || c1 == 'X') && (isHexDigit(s[1]) || s[1] == '.')) {
1223 s++;
1224 assert(s < ThisTokEnd && "didn't maximally munch?")(static_cast <bool> (s < ThisTokEnd && "didn't maximally munch?"
) ? void (0) : __assert_fail ("s < ThisTokEnd && \"didn't maximally munch?\""
, "clang/lib/Lex/LiteralSupport.cpp", 1224, __extension__ __PRETTY_FUNCTION__
))
;
1225 radix = 16;
1226 DigitsBegin = s;
1227 s = SkipHexDigits(s);
1228 bool HasSignificandDigits = containsDigits(DigitsBegin, s);
1229 if (s == ThisTokEnd) {
1230 // Done.
1231 } else if (*s == '.') {
1232 s++;
1233 saw_period = true;
1234 const char *floatDigitsBegin = s;
1235 s = SkipHexDigits(s);
1236 if (containsDigits(floatDigitsBegin, s))
1237 HasSignificandDigits = true;
1238 if (HasSignificandDigits)
1239 checkSeparator(TokLoc, floatDigitsBegin, CSK_BeforeDigits);
1240 }
1241
1242 if (!HasSignificandDigits) {
1243 Diags.Report(Lexer::AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin, SM,
1244 LangOpts),
1245 diag::err_hex_constant_requires)
1246 << LangOpts.CPlusPlus << 1;
1247 hadError = true;
1248 return;
1249 }
1250
1251 // A binary exponent can appear with or with a '.'. If dotted, the
1252 // binary exponent is required.
1253 if (*s == 'p' || *s == 'P') {
1254 checkSeparator(TokLoc, s, CSK_AfterDigits);
1255 const char *Exponent = s;
1256 s++;
1257 saw_exponent = true;
1258 if (s != ThisTokEnd && (*s == '+' || *s == '-')) s++; // sign
1259 const char *first_non_digit = SkipDigits(s);
1260 if (!containsDigits(s, first_non_digit)) {
1261 if (!hadError) {
1262 Diags.Report(Lexer::AdvanceToTokenCharacter(
1263 TokLoc, Exponent - ThisTokBegin, SM, LangOpts),
1264 diag::err_exponent_has_no_digits);
1265 hadError = true;
1266 }
1267 return;
1268 }
1269 checkSeparator(TokLoc, s, CSK_BeforeDigits);
1270 s = first_non_digit;
1271
1272 if (!LangOpts.HexFloats)
1273 Diags.Report(TokLoc, LangOpts.CPlusPlus
1274 ? diag::ext_hex_literal_invalid
1275 : diag::ext_hex_constant_invalid);
1276 else if (LangOpts.CPlusPlus17)
1277 Diags.Report(TokLoc, diag::warn_cxx17_hex_literal);
1278 } else if (saw_period) {
1279 Diags.Report(Lexer::AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin, SM,
1280 LangOpts),
1281 diag::err_hex_constant_requires)
1282 << LangOpts.CPlusPlus << 0;
1283 hadError = true;
1284 }
1285 return;
1286 }
1287
1288 // Handle simple binary numbers 0b01010
1289 if ((c1 == 'b' || c1 == 'B') && (s[1] == '0' || s[1] == '1')) {
1290 // 0b101010 is a C++1y / GCC extension.
1291 Diags.Report(TokLoc, LangOpts.CPlusPlus14
1292 ? diag::warn_cxx11_compat_binary_literal
1293 : LangOpts.CPlusPlus ? diag::ext_binary_literal_cxx14
1294 : diag::ext_binary_literal);
1295 ++s;
1296 assert(s < ThisTokEnd && "didn't maximally munch?")(static_cast <bool> (s < ThisTokEnd && "didn't maximally munch?"
) ? void (0) : __assert_fail ("s < ThisTokEnd && \"didn't maximally munch?\""
, "clang/lib/Lex/LiteralSupport.cpp", 1296, __extension__ __PRETTY_FUNCTION__
))
;
1297 radix = 2;
1298 DigitsBegin = s;
1299 s = SkipBinaryDigits(s);
1300 if (s == ThisTokEnd) {
1301 // Done.
1302 } else if (isHexDigit(*s) &&
1303 !isValidUDSuffix(LangOpts, StringRef(s, ThisTokEnd - s))) {
1304 Diags.Report(Lexer::AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin, SM,
1305 LangOpts),
1306 diag::err_invalid_digit)
1307 << StringRef(s, 1) << 2;
1308 hadError = true;
1309 }
1310 // Other suffixes will be diagnosed by the caller.
1311 return;
1312 }
1313
1314 // For now, the radix is set to 8. If we discover that we have a
1315 // floating point constant, the radix will change to 10. Octal floating
1316 // point constants are not permitted (only decimal and hexadecimal).
1317 radix = 8;
1318 const char *PossibleNewDigitStart = s;
1319 s = SkipOctalDigits(s);
1320 // When the value is 0 followed by a suffix (like 0wb), we want to leave 0
1321 // as the start of the digits. So if skipping octal digits does not skip
1322 // anything, we leave the digit start where it was.
1323 if (s != PossibleNewDigitStart)
1324 DigitsBegin = PossibleNewDigitStart;
1325
1326 if (s == ThisTokEnd)
1327 return; // Done, simple octal number like 01234
1328
1329 // If we have some other non-octal digit that *is* a decimal digit, see if
1330 // this is part of a floating point number like 094.123 or 09e1.
1331 if (isDigit(*s)) {
1332 const char *EndDecimal = SkipDigits(s);
1333 if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') {
1334 s = EndDecimal;
1335 radix = 10;
1336 }
1337 }
1338
1339 ParseDecimalOrOctalCommon(TokLoc);
1340}
1341
1342static bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits) {
1343 switch (Radix) {
1344 case 2:
1345 return NumDigits <= 64;
1346 case 8:
1347 return NumDigits <= 64 / 3; // Digits are groups of 3 bits.
1348 case 10:
1349 return NumDigits <= 19; // floor(log10(2^64))
1350 case 16:
1351 return NumDigits <= 64 / 4; // Digits are groups of 4 bits.
1352 default:
1353 llvm_unreachable("impossible Radix")::llvm::llvm_unreachable_internal("impossible Radix", "clang/lib/Lex/LiteralSupport.cpp"
, 1353)
;
1354 }
1355}
1356
1357/// GetIntegerValue - Convert this numeric literal value to an APInt that
1358/// matches Val's input width. If there is an overflow, set Val to the low bits
1359/// of the result and return true. Otherwise, return false.
1360bool NumericLiteralParser::GetIntegerValue(llvm::APInt &Val) {
1361 // Fast path: Compute a conservative bound on the maximum number of
1362 // bits per digit in this radix. If we can't possibly overflow a
1363 // uint64 based on that bound then do the simple conversion to
1364 // integer. This avoids the expensive overflow checking below, and
1365 // handles the common cases that matter (small decimal integers and
1366 // hex/octal values which don't overflow).
1367 const unsigned NumDigits = SuffixBegin - DigitsBegin;
1368 if (alwaysFitsInto64Bits(radix, NumDigits)) {
1369 uint64_t N = 0;
1370 for (const char *Ptr = DigitsBegin; Ptr != SuffixBegin; ++Ptr)
1371 if (!isDigitSeparator(*Ptr))
1372 N = N * radix + llvm::hexDigitValue(*Ptr);
1373
1374 // This will truncate the value to Val's input width. Simply check
1375 // for overflow by comparing.
1376 Val = N;
1377 return Val.getZExtValue() != N;
1378 }
1379
1380 Val = 0;
1381 const char *Ptr = DigitsBegin;
1382
1383 llvm::APInt RadixVal(Val.getBitWidth(), radix);
1384 llvm::APInt CharVal(Val.getBitWidth(), 0);
1385 llvm::APInt OldVal = Val;
1386
1387 bool OverflowOccurred = false;
1388 while (Ptr < SuffixBegin) {
1389 if (isDigitSeparator(*Ptr)) {
1390 ++Ptr;
1391 continue;
1392 }
1393
1394 unsigned C = llvm::hexDigitValue(*Ptr++);
1395
1396 // If this letter is out of bound for this radix, reject it.
1397 assert(C < radix && "NumericLiteralParser ctor should have rejected this")(static_cast <bool> (C < radix && "NumericLiteralParser ctor should have rejected this"
) ? void (0) : __assert_fail ("C < radix && \"NumericLiteralParser ctor should have rejected this\""
, "clang/lib/Lex/LiteralSupport.cpp", 1397, __extension__ __PRETTY_FUNCTION__
))
;
1398
1399 CharVal = C;
1400
1401 // Add the digit to the value in the appropriate radix. If adding in digits
1402 // made the value smaller, then this overflowed.
1403 OldVal = Val;
1404
1405 // Multiply by radix, did overflow occur on the multiply?
1406 Val *= RadixVal;
1407 OverflowOccurred |= Val.udiv(RadixVal) != OldVal;
1408
1409 // Add value, did overflow occur on the value?
1410 // (a + b) ult b <=> overflow
1411 Val += CharVal;
1412 OverflowOccurred |= Val.ult(CharVal);
1413 }
1414 return OverflowOccurred;
1415}
1416
1417llvm::APFloat::opStatus
1418NumericLiteralParser::GetFloatValue(llvm::APFloat &Result) {
1419 using llvm::APFloat;
1420
1421 unsigned n = std::min(SuffixBegin - ThisTokBegin, ThisTokEnd - ThisTokBegin);
1422
1423 llvm::SmallString<16> Buffer;
1424 StringRef Str(ThisTokBegin, n);
1425 if (Str.contains('\'')) {
1426 Buffer.reserve(n);
1427 std::remove_copy_if(Str.begin(), Str.end(), std::back_inserter(Buffer),
1428 &isDigitSeparator);
1429 Str = Buffer;
1430 }
1431
1432 auto StatusOrErr =
1433 Result.convertFromString(Str, APFloat::rmNearestTiesToEven);
1434 assert(StatusOrErr && "Invalid floating point representation")(static_cast <bool> (StatusOrErr && "Invalid floating point representation"
) ? void (0) : __assert_fail ("StatusOrErr && \"Invalid floating point representation\""
, "clang/lib/Lex/LiteralSupport.cpp", 1434, __extension__ __PRETTY_FUNCTION__
))
;
1435 return !errorToBool(StatusOrErr.takeError()) ? *StatusOrErr
1436 : APFloat::opInvalidOp;
1437}
1438
1439static inline bool IsExponentPart(char c) {
1440 return c == 'p' || c == 'P' || c == 'e' || c == 'E';
1441}
1442
1443bool NumericLiteralParser::GetFixedPointValue(llvm::APInt &StoreVal, unsigned Scale) {
1444 assert(radix == 16 || radix == 10)(static_cast <bool> (radix == 16 || radix == 10) ? void
(0) : __assert_fail ("radix == 16 || radix == 10", "clang/lib/Lex/LiteralSupport.cpp"
, 1444, __extension__ __PRETTY_FUNCTION__))
;
1445
1446 // Find how many digits are needed to store the whole literal.
1447 unsigned NumDigits = SuffixBegin - DigitsBegin;
1448 if (saw_period) --NumDigits;
1449
1450 // Initial scan of the exponent if it exists
1451 bool ExpOverflowOccurred = false;
1452 bool NegativeExponent = false;
1453 const char *ExponentBegin;
1454 uint64_t Exponent = 0;
1455 int64_t BaseShift = 0;
1456 if (saw_exponent) {
1457 const char *Ptr = DigitsBegin;
1458
1459 while (!IsExponentPart(*Ptr)) ++Ptr;
1460 ExponentBegin = Ptr;
1461 ++Ptr;
1462 NegativeExponent = *Ptr == '-';
1463 if (NegativeExponent) ++Ptr;
1464
1465 unsigned NumExpDigits = SuffixBegin - Ptr;
1466 if (alwaysFitsInto64Bits(radix, NumExpDigits)) {
1467 llvm::StringRef ExpStr(Ptr, NumExpDigits);
1468 llvm::APInt ExpInt(/*numBits=*/64, ExpStr, /*radix=*/10);
1469 Exponent = ExpInt.getZExtValue();
1470 } else {
1471 ExpOverflowOccurred = true;
1472 }
1473
1474 if (NegativeExponent) BaseShift -= Exponent;
1475 else BaseShift += Exponent;
1476 }
1477
1478 // Number of bits needed for decimal literal is
1479 // ceil(NumDigits * log2(10)) Integral part
1480 // + Scale Fractional part
1481 // + ceil(Exponent * log2(10)) Exponent
1482 // --------------------------------------------------
1483 // ceil((NumDigits + Exponent) * log2(10)) + Scale
1484 //
1485 // But for simplicity in handling integers, we can round up log2(10) to 4,
1486 // making:
1487 // 4 * (NumDigits + Exponent) + Scale
1488 //
1489 // Number of digits needed for hexadecimal literal is
1490 // 4 * NumDigits Integral part
1491 // + Scale Fractional part
1492 // + Exponent Exponent
1493 // --------------------------------------------------
1494 // (4 * NumDigits) + Scale + Exponent
1495 uint64_t NumBitsNeeded;
1496 if (radix == 10)
1497 NumBitsNeeded = 4 * (NumDigits + Exponent) + Scale;
1498 else
1499 NumBitsNeeded = 4 * NumDigits + Exponent + Scale;
1500
1501 if (NumBitsNeeded > std::numeric_limits<unsigned>::max())
1502 ExpOverflowOccurred = true;
1503 llvm::APInt Val(static_cast<unsigned>(NumBitsNeeded), 0, /*isSigned=*/false);
1504
1505 bool FoundDecimal = false;
1506
1507 int64_t FractBaseShift = 0;
1508 const char *End = saw_exponent ? ExponentBegin : SuffixBegin;
1509 for (const char *Ptr = DigitsBegin; Ptr < End; ++Ptr) {
1510 if (*Ptr == '.') {
1511 FoundDecimal = true;
1512 continue;
1513 }
1514
1515 // Normal reading of an integer
1516 unsigned C = llvm::hexDigitValue(*Ptr);
1517 assert(C < radix && "NumericLiteralParser ctor should have rejected this")(static_cast <bool> (C < radix && "NumericLiteralParser ctor should have rejected this"
) ? void (0) : __assert_fail ("C < radix && \"NumericLiteralParser ctor should have rejected this\""
, "clang/lib/Lex/LiteralSupport.cpp", 1517, __extension__ __PRETTY_FUNCTION__
))
;
1518
1519 Val *= radix;
1520 Val += C;
1521
1522 if (FoundDecimal)
1523 // Keep track of how much we will need to adjust this value by from the
1524 // number of digits past the radix point.
1525 --FractBaseShift;
1526 }
1527
1528 // For a radix of 16, we will be multiplying by 2 instead of 16.
1529 if (radix == 16) FractBaseShift *= 4;
1530 BaseShift += FractBaseShift;
1531
1532 Val <<= Scale;
1533
1534 uint64_t Base = (radix == 16) ? 2 : 10;
1535 if (BaseShift > 0) {
1536 for (int64_t i = 0; i < BaseShift; ++i) {
1537 Val *= Base;
1538 }
1539 } else if (BaseShift < 0) {
1540 for (int64_t i = BaseShift; i < 0 && !Val.isZero(); ++i)
1541 Val = Val.udiv(Base);
1542 }
1543
1544 bool IntOverflowOccurred = false;
1545 auto MaxVal = llvm::APInt::getMaxValue(StoreVal.getBitWidth());
1546 if (Val.getBitWidth() > StoreVal.getBitWidth()) {
1547 IntOverflowOccurred |= Val.ugt(MaxVal.zext(Val.getBitWidth()));
1548 StoreVal = Val.trunc(StoreVal.getBitWidth());
1549 } else if (Val.getBitWidth() < StoreVal.getBitWidth()) {
1550 IntOverflowOccurred |= Val.zext(MaxVal.getBitWidth()).ugt(MaxVal);
1551 StoreVal = Val.zext(StoreVal.getBitWidth());
1552 } else {
1553 StoreVal = Val;
1554 }
1555
1556 return IntOverflowOccurred || ExpOverflowOccurred;
1557}
1558
1559/// \verbatim
1560/// user-defined-character-literal: [C++11 lex.ext]
1561/// character-literal ud-suffix
1562/// ud-suffix:
1563/// identifier
1564/// character-literal: [C++11 lex.ccon]
1565/// ' c-char-sequence '
1566/// u' c-char-sequence '
1567/// U' c-char-sequence '
1568/// L' c-char-sequence '
1569/// u8' c-char-sequence ' [C++1z lex.ccon]
1570/// c-char-sequence:
1571/// c-char
1572/// c-char-sequence c-char
1573/// c-char:
1574/// any member of the source character set except the single-quote ',
1575/// backslash \, or new-line character
1576/// escape-sequence
1577/// universal-character-name
1578/// escape-sequence:
1579/// simple-escape-sequence
1580/// octal-escape-sequence
1581/// hexadecimal-escape-sequence
1582/// simple-escape-sequence:
1583/// one of \' \" \? \\ \a \b \f \n \r \t \v
1584/// octal-escape-sequence:
1585/// \ octal-digit
1586/// \ octal-digit octal-digit
1587/// \ octal-digit octal-digit octal-digit
1588/// hexadecimal-escape-sequence:
1589/// \x hexadecimal-digit
1590/// hexadecimal-escape-sequence hexadecimal-digit
1591/// universal-character-name: [C++11 lex.charset]
1592/// \u hex-quad
1593/// \U hex-quad hex-quad
1594/// hex-quad:
1595/// hex-digit hex-digit hex-digit hex-digit
1596/// \endverbatim
1597///
1598CharLiteralParser::CharLiteralParser(const char *begin, const char *end,
1599 SourceLocation Loc, Preprocessor &PP,
1600 tok::TokenKind kind) {
1601 // At this point we know that the character matches the regex "(L|u|U)?'.*'".
1602 HadError = false;
1603
1604 Kind = kind;
1605
1606 const char *TokBegin = begin;
1607
1608 // Skip over wide character determinant.
1609 if (Kind != tok::char_constant)
1610 ++begin;
1611 if (Kind == tok::utf8_char_constant)
1612 ++begin;
1613
1614 // Skip over the entry quote.
1615 if (begin[0] != '\'') {
1616 PP.Diag(Loc, diag::err_lexing_char);
1617 HadError = true;
1618 return;
1619 }
1620
1621 ++begin;
1622
1623 // Remove an optional ud-suffix.
1624 if (end[-1] != '\'') {
1625 const char *UDSuffixEnd = end;
1626 do {
1627 --end;
1628 } while (end[-1] != '\'');
1629 // FIXME: Don't bother with this if !tok.hasUCN().
1630 expandUCNs(UDSuffixBuf, StringRef(end, UDSuffixEnd - end));
1631 UDSuffixOffset = end - TokBegin;
1632 }
1633
1634 // Trim the ending quote.
1635 assert(end != begin && "Invalid token lexed")(static_cast <bool> (end != begin && "Invalid token lexed"
) ? void (0) : __assert_fail ("end != begin && \"Invalid token lexed\""
, "clang/lib/Lex/LiteralSupport.cpp", 1635, __extension__ __PRETTY_FUNCTION__
))
;
1636 --end;
1637
1638 // FIXME: The "Value" is an uint64_t so we can handle char literals of
1639 // up to 64-bits.
1640 // FIXME: This extensively assumes that 'char' is 8-bits.
1641 assert(PP.getTargetInfo().getCharWidth() == 8 &&(static_cast <bool> (PP.getTargetInfo().getCharWidth() ==
8 && "Assumes char is 8 bits") ? void (0) : __assert_fail
("PP.getTargetInfo().getCharWidth() == 8 && \"Assumes char is 8 bits\""
, "clang/lib/Lex/LiteralSupport.cpp", 1642, __extension__ __PRETTY_FUNCTION__
))
1642 "Assumes char is 8 bits")(static_cast <bool> (PP.getTargetInfo().getCharWidth() ==
8 && "Assumes char is 8 bits") ? void (0) : __assert_fail
("PP.getTargetInfo().getCharWidth() == 8 && \"Assumes char is 8 bits\""
, "clang/lib/Lex/LiteralSupport.cpp", 1642, __extension__ __PRETTY_FUNCTION__
))
;
1643 assert(PP.getTargetInfo().getIntWidth() <= 64 &&(static_cast <bool> (PP.getTargetInfo().getIntWidth() <=
64 && (PP.getTargetInfo().getIntWidth() & 7) == 0
&& "Assumes sizeof(int) on target is <= 64 and a multiple of char"
) ? void (0) : __assert_fail ("PP.getTargetInfo().getIntWidth() <= 64 && (PP.getTargetInfo().getIntWidth() & 7) == 0 && \"Assumes sizeof(int) on target is <= 64 and a multiple of char\""
, "clang/lib/Lex/LiteralSupport.cpp", 1645, __extension__ __PRETTY_FUNCTION__
))
1644 (PP.getTargetInfo().getIntWidth() & 7) == 0 &&(static_cast <bool> (PP.getTargetInfo().getIntWidth() <=
64 && (PP.getTargetInfo().getIntWidth() & 7) == 0
&& "Assumes sizeof(int) on target is <= 64 and a multiple of char"
) ? void (0) : __assert_fail ("PP.getTargetInfo().getIntWidth() <= 64 && (PP.getTargetInfo().getIntWidth() & 7) == 0 && \"Assumes sizeof(int) on target is <= 64 and a multiple of char\""
, "clang/lib/Lex/LiteralSupport.cpp", 1645, __extension__ __PRETTY_FUNCTION__
))
1645 "Assumes sizeof(int) on target is <= 64 and a multiple of char")(static_cast <bool> (PP.getTargetInfo().getIntWidth() <=
64 && (PP.getTargetInfo().getIntWidth() & 7) == 0
&& "Assumes sizeof(int) on target is <= 64 and a multiple of char"
) ? void (0) : __assert_fail ("PP.getTargetInfo().getIntWidth() <= 64 && (PP.getTargetInfo().getIntWidth() & 7) == 0 && \"Assumes sizeof(int) on target is <= 64 and a multiple of char\""
, "clang/lib/Lex/LiteralSupport.cpp", 1645, __extension__ __PRETTY_FUNCTION__
))
;
1646 assert(PP.getTargetInfo().getWCharWidth() <= 64 &&(static_cast <bool> (PP.getTargetInfo().getWCharWidth()
<= 64 && "Assumes sizeof(wchar) on target is <= 64"
) ? void (0) : __assert_fail ("PP.getTargetInfo().getWCharWidth() <= 64 && \"Assumes sizeof(wchar) on target is <= 64\""
, "clang/lib/Lex/LiteralSupport.cpp", 1647, __extension__ __PRETTY_FUNCTION__
))
1647 "Assumes sizeof(wchar) on target is <= 64")(static_cast <bool> (PP.getTargetInfo().getWCharWidth()
<= 64 && "Assumes sizeof(wchar) on target is <= 64"
) ? void (0) : __assert_fail ("PP.getTargetInfo().getWCharWidth() <= 64 && \"Assumes sizeof(wchar) on target is <= 64\""
, "clang/lib/Lex/LiteralSupport.cpp", 1647, __extension__ __PRETTY_FUNCTION__
))
;
1648
1649 SmallVector<uint32_t, 4> codepoint_buffer;
1650 codepoint_buffer.resize(end - begin);
1651 uint32_t *buffer_begin = &codepoint_buffer.front();
1652 uint32_t *buffer_end = buffer_begin + codepoint_buffer.size();
1653
1654 // Unicode escapes representing characters that cannot be correctly
1655 // represented in a single code unit are disallowed in character literals
1656 // by this implementation.
1657 uint32_t largest_character_for_kind;
1658 if (tok::wide_char_constant == Kind) {
1659 largest_character_for_kind =
1660 0xFFFFFFFFu >> (32-PP.getTargetInfo().getWCharWidth());
1661 } else if (tok::utf8_char_constant == Kind) {
1662 largest_character_for_kind = 0x7F;
1663 } else if (tok::utf16_char_constant == Kind) {
1664 largest_character_for_kind = 0xFFFF;
1665 } else if (tok::utf32_char_constant == Kind) {
1666 largest_character_for_kind = 0x10FFFF;
1667 } else {
1668 largest_character_for_kind = 0x7Fu;
1669 }
1670
1671 while (begin != end) {
1672 // Is this a span of non-escape characters?
1673 if (begin[0] != '\\') {
1674 char const *start = begin;
1675 do {
1676 ++begin;
1677 } while (begin != end && *begin != '\\');
1678
1679 char const *tmp_in_start = start;
1680 uint32_t *tmp_out_start = buffer_begin;
1681 llvm::ConversionResult res =
1682 llvm::ConvertUTF8toUTF32(reinterpret_cast<llvm::UTF8 const **>(&start),
1683 reinterpret_cast<llvm::UTF8 const *>(begin),
1684 &buffer_begin, buffer_end, llvm::strictConversion);
1685 if (res != llvm::conversionOK) {
1686 // If we see bad encoding for unprefixed character literals, warn and
1687 // simply copy the byte values, for compatibility with gcc and
1688 // older versions of clang.
1689 bool NoErrorOnBadEncoding = isOrdinary();
1690 unsigned Msg = diag::err_bad_character_encoding;
1691 if (NoErrorOnBadEncoding)
1692 Msg = diag::warn_bad_character_encoding;
1693 PP.Diag(Loc, Msg);
1694 if (NoErrorOnBadEncoding) {
1695 start = tmp_in_start;
1696 buffer_begin = tmp_out_start;
1697 for (; start != begin; ++start, ++buffer_begin)
1698 *buffer_begin = static_cast<uint8_t>(*start);
1699 } else {
1700 HadError = true;
1701 }
1702 } else {
1703 for (; tmp_out_start < buffer_begin; ++tmp_out_start) {
1704 if (*tmp_out_start > largest_character_for_kind) {
1705 HadError = true;
1706 PP.Diag(Loc, diag::err_character_too_large);
1707 }
1708 }
1709 }
1710
1711 continue;
1712 }
1713 // Is this a Universal Character Name escape?
1714 if (begin[1] == 'u' || begin[1] == 'U' || begin[1] == 'N') {
1715 unsigned short UcnLen = 0;
1716 if (!ProcessUCNEscape(TokBegin, begin, end, *buffer_begin, UcnLen,
1717 FullSourceLoc(Loc, PP.getSourceManager()),
1718 &PP.getDiagnostics(), PP.getLangOpts(), true)) {
1719 HadError = true;
1720 } else if (*buffer_begin > largest_character_for_kind) {
1721 HadError = true;
1722 PP.Diag(Loc, diag::err_character_too_large);
1723 }
1724
1725 ++buffer_begin;
1726 continue;
1727 }
1728 unsigned CharWidth = getCharWidth(Kind, PP.getTargetInfo());
1729 uint64_t result =
1730 ProcessCharEscape(TokBegin, begin, end, HadError,
1731 FullSourceLoc(Loc,PP.getSourceManager()),
1732 CharWidth, &PP.getDiagnostics(), PP.getLangOpts());
1733 *buffer_begin++ = result;
1734 }
1735
1736 unsigned NumCharsSoFar = buffer_begin - &codepoint_buffer.front();
1737
1738 if (NumCharsSoFar > 1) {
1739 if (isOrdinary() && NumCharsSoFar == 4)
1740 PP.Diag(Loc, diag::warn_four_char_character_literal);
1741 else if (isOrdinary())
1742 PP.Diag(Loc, diag::warn_multichar_character_literal);
1743 else {
1744 PP.Diag(Loc, diag::err_multichar_character_literal) << (isWide() ? 0 : 1);
1745 HadError = true;
1746 }
1747 IsMultiChar = true;
1748 } else {
1749 IsMultiChar = false;
1750 }
1751
1752 llvm::APInt LitVal(PP.getTargetInfo().getIntWidth(), 0);
1753
1754 // Narrow character literals act as though their value is concatenated
1755 // in this implementation, but warn on overflow.
1756 bool multi_char_too_long = false;
1757 if (isOrdinary() && isMultiChar()) {
1758 LitVal = 0;
1759 for (size_t i = 0; i < NumCharsSoFar; ++i) {
1760 // check for enough leading zeros to shift into
1761 multi_char_too_long |= (LitVal.countl_zero() < 8);
1762 LitVal <<= 8;
1763 LitVal = LitVal + (codepoint_buffer[i] & 0xFF);
1764 }
1765 } else if (NumCharsSoFar > 0) {
1766 // otherwise just take the last character
1767 LitVal = buffer_begin[-1];
1768 }
1769
1770 if (!HadError && multi_char_too_long) {
1771 PP.Diag(Loc, diag::warn_char_constant_too_large);
1772 }
1773
1774 // Transfer the value from APInt to uint64_t
1775 Value = LitVal.getZExtValue();
1776
1777 // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1")
1778 // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple
1779 // character constants are not sign extended in the this implementation:
1780 // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC.
1781 if (isOrdinary() && NumCharsSoFar == 1 && (Value & 128) &&
1782 PP.getLangOpts().CharIsSigned)
1783 Value = (signed char)Value;
1784}
1785
1786/// \verbatim
1787/// string-literal: [C++0x lex.string]
1788/// encoding-prefix " [s-char-sequence] "
1789/// encoding-prefix R raw-string
1790/// encoding-prefix:
1791/// u8
1792/// u
1793/// U
1794/// L
1795/// s-char-sequence:
1796/// s-char
1797/// s-char-sequence s-char
1798/// s-char:
1799/// any member of the source character set except the double-quote ",
1800/// backslash \, or new-line character
1801/// escape-sequence
1802/// universal-character-name
1803/// raw-string:
1804/// " d-char-sequence ( r-char-sequence ) d-char-sequence "
1805/// r-char-sequence:
1806/// r-char
1807/// r-char-sequence r-char
1808/// r-char:
1809/// any member of the source character set, except a right parenthesis )
1810/// followed by the initial d-char-sequence (which may be empty)
1811/// followed by a double quote ".
1812/// d-char-sequence:
1813/// d-char
1814/// d-char-sequence d-char
1815/// d-char:
1816/// any member of the basic source character set except:
1817/// space, the left parenthesis (, the right parenthesis ),
1818/// the backslash \, and the control characters representing horizontal
1819/// tab, vertical tab, form feed, and newline.
1820/// escape-sequence: [C++0x lex.ccon]
1821/// simple-escape-sequence
1822/// octal-escape-sequence
1823/// hexadecimal-escape-sequence
1824/// simple-escape-sequence:
1825/// one of \' \" \? \\ \a \b \f \n \r \t \v
1826/// octal-escape-sequence:
1827/// \ octal-digit
1828/// \ octal-digit octal-digit
1829/// \ octal-digit octal-digit octal-digit
1830/// hexadecimal-escape-sequence:
1831/// \x hexadecimal-digit
1832/// hexadecimal-escape-sequence hexadecimal-digit
1833/// universal-character-name:
1834/// \u hex-quad
1835/// \U hex-quad hex-quad
1836/// hex-quad:
1837/// hex-digit hex-digit hex-digit hex-digit
1838/// \endverbatim
1839///
1840StringLiteralParser::
1841StringLiteralParser(ArrayRef<Token> StringToks,
1842 Preprocessor &PP)
1843 : SM(PP.getSourceManager()), Features(PP.getLangOpts()),
1844 Target(PP.getTargetInfo()), Diags(&PP.getDiagnostics()),
1845 MaxTokenLength(0), SizeBound(0), CharByteWidth(0), Kind(tok::unknown),
1846 ResultPtr(ResultBuf.data()), hadError(false), Pascal(false) {
1847 init(StringToks);
1848}
1849
1850void StringLiteralParser::init(ArrayRef<Token> StringToks){
1851 // The literal token may have come from an invalid source location (e.g. due
1852 // to a PCH error), in which case the token length will be 0.
1853 if (StringToks.empty() || StringToks[0].getLength() < 2)
1854 return DiagnoseLexingError(SourceLocation());
1855
1856 // Scan all of the string portions, remember the max individual token length,
1857 // computing a bound on the concatenated string length, and see whether any
1858 // piece is a wide-string. If any of the string portions is a wide-string
1859 // literal, the result is a wide-string literal [C99 6.4.5p4].
1860 assert(!StringToks.empty() && "expected at least one token")(static_cast <bool> (!StringToks.empty() && "expected at least one token"
) ? void (0) : __assert_fail ("!StringToks.empty() && \"expected at least one token\""
, "clang/lib/Lex/LiteralSupport.cpp", 1860, __extension__ __PRETTY_FUNCTION__
))
;
1861 MaxTokenLength = StringToks[0].getLength();
1862 assert(StringToks[0].getLength() >= 2 && "literal token is invalid!")(static_cast <bool> (StringToks[0].getLength() >= 2 &&
"literal token is invalid!") ? void (0) : __assert_fail ("StringToks[0].getLength() >= 2 && \"literal token is invalid!\""
, "clang/lib/Lex/LiteralSupport.cpp", 1862, __extension__ __PRETTY_FUNCTION__
))
;
1863 SizeBound = StringToks[0].getLength()-2; // -2 for "".
1864 Kind = StringToks[0].getKind();
1865
1866 hadError = false;
1867
1868 // Implement Translation Phase #6: concatenation of string literals
1869 /// (C99 5.1.1.2p1). The common case is only one string fragment.
1870 for (const Token &Tok : StringToks) {
1871 if (Tok.getLength() < 2)
1872 return DiagnoseLexingError(Tok.getLocation());
1873
1874 // The string could be shorter than this if it needs cleaning, but this is a
1875 // reasonable bound, which is all we need.
1876 assert(Tok.getLength() >= 2 && "literal token is invalid!")(static_cast <bool> (Tok.getLength() >= 2 &&
"literal token is invalid!") ? void (0) : __assert_fail ("Tok.getLength() >= 2 && \"literal token is invalid!\""
, "clang/lib/Lex/LiteralSupport.cpp", 1876, __extension__ __PRETTY_FUNCTION__
))
;
1877 SizeBound += Tok.getLength() - 2; // -2 for "".
1878
1879 // Remember maximum string piece length.
1880 if (Tok.getLength() > MaxTokenLength)
1881 MaxTokenLength = Tok.getLength();
1882
1883 // Remember if we see any wide or utf-8/16/32 strings.
1884 // Also check for illegal concatenations.
1885 if (Tok.isNot(Kind) && Tok.isNot(tok::string_literal)) {
1886 if (isOrdinary()) {
1887 Kind = Tok.getKind();
1888 } else {
1889 if (Diags)
1890 Diags->Report(Tok.getLocation(), diag::err_unsupported_string_concat);
1891 hadError = true;
1892 }
1893 }
1894 }
1895
1896 // Include space for the null terminator.
1897 ++SizeBound;
1898
1899 // TODO: K&R warning: "traditional C rejects string constant concatenation"
1900
1901 // Get the width in bytes of char/wchar_t/char16_t/char32_t
1902 CharByteWidth = getCharWidth(Kind, Target);
1903 assert((CharByteWidth & 7) == 0 && "Assumes character size is byte multiple")(static_cast <bool> ((CharByteWidth & 7) == 0 &&
"Assumes character size is byte multiple") ? void (0) : __assert_fail
("(CharByteWidth & 7) == 0 && \"Assumes character size is byte multiple\""
, "clang/lib/Lex/LiteralSupport.cpp", 1903, __extension__ __PRETTY_FUNCTION__
))
;
1904 CharByteWidth /= 8;
1905
1906 // The output buffer size needs to be large enough to hold wide characters.
1907 // This is a worst-case assumption which basically corresponds to L"" "long".
1908 SizeBound *= CharByteWidth;
1909
1910 // Size the temporary buffer to hold the result string data.
1911 ResultBuf.resize(SizeBound);
1912
1913 // Likewise, but for each string piece.
1914 SmallString<512> TokenBuf;
1915 TokenBuf.resize(MaxTokenLength);
1916
1917 // Loop over all the strings, getting their spelling, and expanding them to
1918 // wide strings as appropriate.
1919 ResultPtr = &ResultBuf[0]; // Next byte to fill in.
1920
1921 Pascal = false;
1922
1923 SourceLocation UDSuffixTokLoc;
1924
1925 for (unsigned i = 0, e = StringToks.size(); i != e; ++i) {
1926 const char *ThisTokBuf = &TokenBuf[0];
1927 // Get the spelling of the token, which eliminates trigraphs, etc. We know
1928 // that ThisTokBuf points to a buffer that is big enough for the whole token
1929 // and 'spelled' tokens can only shrink.
1930 bool StringInvalid = false;
1931 unsigned ThisTokLen =
1932 Lexer::getSpelling(StringToks[i], ThisTokBuf, SM, Features,
1933 &StringInvalid);
1934 if (StringInvalid)
1935 return DiagnoseLexingError(StringToks[i].getLocation());
1936
1937 const char *ThisTokBegin = ThisTokBuf;
1938 const char *ThisTokEnd = ThisTokBuf+ThisTokLen;
1939
1940 // Remove an optional ud-suffix.
1941 if (ThisTokEnd[-1] != '"') {
1942 const char *UDSuffixEnd = ThisTokEnd;
1943 do {
1944 --ThisTokEnd;
1945 } while (ThisTokEnd[-1] != '"');
1946
1947 StringRef UDSuffix(ThisTokEnd, UDSuffixEnd - ThisTokEnd);
1948
1949 if (UDSuffixBuf.empty()) {
1950 if (StringToks[i].hasUCN())
1951 expandUCNs(UDSuffixBuf, UDSuffix);
1952 else
1953 UDSuffixBuf.assign(UDSuffix);
1954 UDSuffixToken = i;
1955 UDSuffixOffset = ThisTokEnd - ThisTokBuf;
1956 UDSuffixTokLoc = StringToks[i].getLocation();
1957 } else {
1958 SmallString<32> ExpandedUDSuffix;
1959 if (StringToks[i].hasUCN()) {
1960 expandUCNs(ExpandedUDSuffix, UDSuffix);
1961 UDSuffix = ExpandedUDSuffix;
1962 }
1963
1964 // C++11 [lex.ext]p8: At the end of phase 6, if a string literal is the
1965 // result of a concatenation involving at least one user-defined-string-
1966 // literal, all the participating user-defined-string-literals shall
1967 // have the same ud-suffix.
1968 if (UDSuffixBuf != UDSuffix) {
1969 if (Diags) {
1970 SourceLocation TokLoc = StringToks[i].getLocation();
1971 Diags->Report(TokLoc, diag::err_string_concat_mixed_suffix)
1972 << UDSuffixBuf << UDSuffix
1973 << SourceRange(UDSuffixTokLoc, UDSuffixTokLoc)
1974 << SourceRange(TokLoc, TokLoc);
1975 }
1976 hadError = true;
1977 }
1978 }
1979 }
1980
1981 // Strip the end quote.
1982 --ThisTokEnd;
1983
1984 // TODO: Input character set mapping support.
1985
1986 // Skip marker for wide or unicode strings.
1987 if (ThisTokBuf[0] == 'L' || ThisTokBuf[0] == 'u' || ThisTokBuf[0] == 'U') {
1988 ++ThisTokBuf;
1989 // Skip 8 of u8 marker for utf8 strings.
1990 if (ThisTokBuf[0] == '8')
1991 ++ThisTokBuf;
1992 }
1993
1994 // Check for raw string
1995 if (ThisTokBuf[0] == 'R') {
1996 if (ThisTokBuf[1] != '"') {
1997 // The file may have come from PCH and then changed after loading the
1998 // PCH; Fail gracefully.
1999 return DiagnoseLexingError(StringToks[i].getLocation());
2000 }
2001 ThisTokBuf += 2; // skip R"
2002
2003 // C++11 [lex.string]p2: A `d-char-sequence` shall consist of at most 16
2004 // characters.
2005 constexpr unsigned MaxRawStrDelimLen = 16;
2006
2007 const char *Prefix = ThisTokBuf;
2008 while (static_cast<unsigned>(ThisTokBuf - Prefix) < MaxRawStrDelimLen &&
2009 ThisTokBuf[0] != '(')
2010 ++ThisTokBuf;
2011 if (ThisTokBuf[0] != '(')
2012 return DiagnoseLexingError(StringToks[i].getLocation());
2013 ++ThisTokBuf; // skip '('
2014
2015 // Remove same number of characters from the end
2016 ThisTokEnd -= ThisTokBuf - Prefix;
2017 if (ThisTokEnd < ThisTokBuf)
2018 return DiagnoseLexingError(StringToks[i].getLocation());
2019
2020 // C++14 [lex.string]p4: A source-file new-line in a raw string literal
2021 // results in a new-line in the resulting execution string-literal.
2022 StringRef RemainingTokenSpan(ThisTokBuf, ThisTokEnd - ThisTokBuf);
2023 while (!RemainingTokenSpan.empty()) {
2024 // Split the string literal on \r\n boundaries.
2025 size_t CRLFPos = RemainingTokenSpan.find("\r\n");
2026 StringRef BeforeCRLF = RemainingTokenSpan.substr(0, CRLFPos);
2027 StringRef AfterCRLF = RemainingTokenSpan.substr(CRLFPos);
2028
2029 // Copy everything before the \r\n sequence into the string literal.
2030 if (CopyStringFragment(StringToks[i], ThisTokBegin, BeforeCRLF))
2031 hadError = true;
2032
2033 // Point into the \n inside the \r\n sequence and operate on the
2034 // remaining portion of the literal.
2035 RemainingTokenSpan = AfterCRLF.substr(1);
2036 }
2037 } else {
2038 if (ThisTokBuf[0] != '"') {
2039 // The file may have come from PCH and then changed after loading the
2040 // PCH; Fail gracefully.
2041 return DiagnoseLexingError(StringToks[i].getLocation());
2042 }
2043 ++ThisTokBuf; // skip "
2044
2045 // Check if this is a pascal string
2046 if (Features.PascalStrings && ThisTokBuf + 1 != ThisTokEnd &&
2047 ThisTokBuf[0] == '\\' && ThisTokBuf[1] == 'p') {
2048
2049 // If the \p sequence is found in the first token, we have a pascal string
2050 // Otherwise, if we already have a pascal string, ignore the first \p
2051 if (i == 0) {
2052 ++ThisTokBuf;
2053 Pascal = true;
2054 } else if (Pascal)
2055 ThisTokBuf += 2;
2056 }
2057
2058 while (ThisTokBuf != ThisTokEnd) {
2059 // Is this a span of non-escape characters?
2060 if (ThisTokBuf[0] != '\\') {
2061 const char *InStart = ThisTokBuf;
2062 do {
2063 ++ThisTokBuf;
2064 } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\');
2065
2066 // Copy the character span over.
2067 if (CopyStringFragment(StringToks[i], ThisTokBegin,
2068 StringRef(InStart, ThisTokBuf - InStart)))
2069 hadError = true;
2070 continue;
2071 }
2072 // Is this a Universal Character Name escape?
2073 if (ThisTokBuf[1] == 'u' || ThisTokBuf[1] == 'U' ||
2074 ThisTokBuf[1] == 'N') {
2075 EncodeUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
2076 ResultPtr, hadError,
2077 FullSourceLoc(StringToks[i].getLocation(), SM),
2078 CharByteWidth, Diags, Features);
2079 continue;
2080 }
2081 // Otherwise, this is a non-UCN escape character. Process it.
2082 unsigned ResultChar =
2083 ProcessCharEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, hadError,
2084 FullSourceLoc(StringToks[i].getLocation(), SM),
2085 CharByteWidth*8, Diags, Features);
2086
2087 if (CharByteWidth == 4) {
2088 // FIXME: Make the type of the result buffer correct instead of
2089 // using reinterpret_cast.
2090 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultPtr);
2091 *ResultWidePtr = ResultChar;
2092 ResultPtr += 4;
2093 } else if (CharByteWidth == 2) {
2094 // FIXME: Make the type of the result buffer correct instead of
2095 // using reinterpret_cast.
2096 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultPtr);
2097 *ResultWidePtr = ResultChar & 0xFFFF;
2098 ResultPtr += 2;
2099 } else {
2100 assert(CharByteWidth == 1 && "Unexpected char width")(static_cast <bool> (CharByteWidth == 1 && "Unexpected char width"
) ? void (0) : __assert_fail ("CharByteWidth == 1 && \"Unexpected char width\""
, "clang/lib/Lex/LiteralSupport.cpp", 2100, __extension__ __PRETTY_FUNCTION__
))
;
2101 *ResultPtr++ = ResultChar & 0xFF;
2102 }
2103 }
2104 }
2105 }
2106
2107 if (Pascal) {
2108 if (CharByteWidth == 4) {
2109 // FIXME: Make the type of the result buffer correct instead of
2110 // using reinterpret_cast.
2111 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultBuf.data());
2112 ResultWidePtr[0] = GetNumStringChars() - 1;
2113 } else if (CharByteWidth == 2) {
2114 // FIXME: Make the type of the result buffer correct instead of
2115 // using reinterpret_cast.
2116 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultBuf.data());
2117 ResultWidePtr[0] = GetNumStringChars() - 1;
2118 } else {
2119 assert(CharByteWidth == 1 && "Unexpected char width")(static_cast <bool> (CharByteWidth == 1 && "Unexpected char width"
) ? void (0) : __assert_fail ("CharByteWidth == 1 && \"Unexpected char width\""
, "clang/lib/Lex/LiteralSupport.cpp", 2119, __extension__ __PRETTY_FUNCTION__
))
;
2120 ResultBuf[0] = GetNumStringChars() - 1;
2121 }
2122
2123 // Verify that pascal strings aren't too large.
2124 if (GetStringLength() > 256) {
2125 if (Diags)
2126 Diags->Report(StringToks.front().getLocation(),
2127 diag::err_pascal_string_too_long)
2128 << SourceRange(StringToks.front().getLocation(),
2129 StringToks.back().getLocation());
2130 hadError = true;
2131 return;
2132 }
2133 } else if (Diags) {
2134 // Complain if this string literal has too many characters.
2135 unsigned MaxChars = Features.CPlusPlus? 65536 : Features.C99 ? 4095 : 509;
2136
2137 if (GetNumStringChars() > MaxChars)
2138 Diags->Report(StringToks.front().getLocation(),
2139 diag::ext_string_too_long)
2140 << GetNumStringChars() << MaxChars
2141 << (Features.CPlusPlus ? 2 : Features.C99 ? 1 : 0)
2142 << SourceRange(StringToks.front().getLocation(),
2143 StringToks.back().getLocation());
2144 }
2145}
2146
2147static const char *resyncUTF8(const char *Err, const char *End) {
2148 if (Err == End)
2149 return End;
2150 End = Err + std::min<unsigned>(llvm::getNumBytesForUTF8(*Err), End-Err);
2151 while (++Err != End && (*Err & 0xC0) == 0x80)
2152 ;
2153 return Err;
2154}
2155
2156/// This function copies from Fragment, which is a sequence of bytes
2157/// within Tok's contents (which begin at TokBegin) into ResultPtr.
2158/// Performs widening for multi-byte characters.
2159bool StringLiteralParser::CopyStringFragment(const Token &Tok,
2160 const char *TokBegin,
2161 StringRef Fragment) {
2162 const llvm::UTF8 *ErrorPtrTmp;
2163 if (ConvertUTF8toWide(CharByteWidth, Fragment, ResultPtr, ErrorPtrTmp))
2164 return false;
2165
2166 // If we see bad encoding for unprefixed string literals, warn and
2167 // simply copy the byte values, for compatibility with gcc and older
2168 // versions of clang.
2169 bool NoErrorOnBadEncoding = isOrdinary();
2170 if (NoErrorOnBadEncoding) {
2171 memcpy(ResultPtr, Fragment.data(), Fragment.size());
2172 ResultPtr += Fragment.size();
2173 }
2174
2175 if (Diags) {
2176 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
2177
2178 FullSourceLoc SourceLoc(Tok.getLocation(), SM);
2179 const DiagnosticBuilder &Builder =
2180 Diag(Diags, Features, SourceLoc, TokBegin,
2181 ErrorPtr, resyncUTF8(ErrorPtr, Fragment.end()),
2182 NoErrorOnBadEncoding ? diag::warn_bad_string_encoding
2183 : diag::err_bad_string_encoding);
2184
2185 const char *NextStart = resyncUTF8(ErrorPtr, Fragment.end());
2186 StringRef NextFragment(NextStart, Fragment.end()-NextStart);
2187
2188 // Decode into a dummy buffer.
2189 SmallString<512> Dummy;
2190 Dummy.reserve(Fragment.size() * CharByteWidth);
2191 char *Ptr = Dummy.data();
2192
2193 while (!ConvertUTF8toWide(CharByteWidth, NextFragment, Ptr, ErrorPtrTmp)) {
2194 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
2195 NextStart = resyncUTF8(ErrorPtr, Fragment.end());
2196 Builder << MakeCharSourceRange(Features, SourceLoc, TokBegin,
2197 ErrorPtr, NextStart);
2198 NextFragment = StringRef(NextStart, Fragment.end()-NextStart);
2199 }
2200 }
2201 return !NoErrorOnBadEncoding;
2202}
2203
2204void StringLiteralParser::DiagnoseLexingError(SourceLocation Loc) {
2205 hadError = true;
2206 if (Diags)
2207 Diags->Report(Loc, diag::err_lexing_string);
2208}
2209
2210/// getOffsetOfStringByte - This function returns the offset of the
2211/// specified byte of the string data represented by Token. This handles
2212/// advancing over escape sequences in the string.
2213unsigned StringLiteralParser::getOffsetOfStringByte(const Token &Tok,
2214 unsigned ByteNo) const {
2215 // Get the spelling of the token.
2216 SmallString<32> SpellingBuffer;
2217 SpellingBuffer.resize(Tok.getLength());
2218
2219 bool StringInvalid = false;
2220 const char *SpellingPtr = &SpellingBuffer[0];
2221 unsigned TokLen = Lexer::getSpelling(Tok, SpellingPtr, SM, Features,
2222 &StringInvalid);
2223 if (StringInvalid)
2224 return 0;
2225
2226 const char *SpellingStart = SpellingPtr;
2227 const char *SpellingEnd = SpellingPtr+TokLen;
2228
2229 // Handle UTF-8 strings just like narrow strings.
2230 if (SpellingPtr[0] == 'u' && SpellingPtr[1] == '8')
2231 SpellingPtr += 2;
2232
2233 assert(SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' &&(static_cast <bool> (SpellingPtr[0] != 'L' && SpellingPtr
[0] != 'u' && SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet"
) ? void (0) : __assert_fail ("SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' && SpellingPtr[0] != 'U' && \"Doesn't handle wide or utf strings yet\""
, "clang/lib/Lex/LiteralSupport.cpp", 2234, __extension__ __PRETTY_FUNCTION__
))
2234 SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet")(static_cast <bool> (SpellingPtr[0] != 'L' && SpellingPtr
[0] != 'u' && SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet"
) ? void (0) : __assert_fail ("SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' && SpellingPtr[0] != 'U' && \"Doesn't handle wide or utf strings yet\""
, "clang/lib/Lex/LiteralSupport.cpp", 2234, __extension__ __PRETTY_FUNCTION__
))
;
2235
2236 // For raw string literals, this is easy.
2237 if (SpellingPtr[0] == 'R') {
2238 assert(SpellingPtr[1] == '"' && "Should be a raw string literal!")(static_cast <bool> (SpellingPtr[1] == '"' && "Should be a raw string literal!"
) ? void (0) : __assert_fail ("SpellingPtr[1] == '\"' && \"Should be a raw string literal!\""
, "clang/lib/Lex/LiteralSupport.cpp", 2238, __extension__ __PRETTY_FUNCTION__
))
;
2239 // Skip 'R"'.
2240 SpellingPtr += 2;
2241 while (*SpellingPtr != '(') {
2242 ++SpellingPtr;
2243 assert(SpellingPtr < SpellingEnd && "Missing ( for raw string literal")(static_cast <bool> (SpellingPtr < SpellingEnd &&
"Missing ( for raw string literal") ? void (0) : __assert_fail
("SpellingPtr < SpellingEnd && \"Missing ( for raw string literal\""
, "clang/lib/Lex/LiteralSupport.cpp", 2243, __extension__ __PRETTY_FUNCTION__
))
;
2244 }
2245 // Skip '('.
2246 ++SpellingPtr;
2247 return SpellingPtr - SpellingStart + ByteNo;
2248 }
2249
2250 // Skip over the leading quote
2251 assert(SpellingPtr[0] == '"' && "Should be a string literal!")(static_cast <bool> (SpellingPtr[0] == '"' && "Should be a string literal!"
) ? void (0) : __assert_fail ("SpellingPtr[0] == '\"' && \"Should be a string literal!\""
, "clang/lib/Lex/LiteralSupport.cpp", 2251, __extension__ __PRETTY_FUNCTION__
))
;
2252 ++SpellingPtr;
2253
2254 // Skip over bytes until we find the offset we're looking for.
2255 while (ByteNo) {
2256 assert(SpellingPtr < SpellingEnd && "Didn't find byte offset!")(static_cast <bool> (SpellingPtr < SpellingEnd &&
"Didn't find byte offset!") ? void (0) : __assert_fail ("SpellingPtr < SpellingEnd && \"Didn't find byte offset!\""
, "clang/lib/Lex/LiteralSupport.cpp", 2256, __extension__ __PRETTY_FUNCTION__
))
;
2257
2258 // Step over non-escapes simply.
2259 if (*SpellingPtr != '\\') {
2260 ++SpellingPtr;
2261 --ByteNo;
2262 continue;
2263 }
2264
2265 // Otherwise, this is an escape character. Advance over it.
2266 bool HadError = false;
2267 if (SpellingPtr[1] == 'u' || SpellingPtr[1] == 'U' ||
2268 SpellingPtr[1] == 'N') {
2269 const char *EscapePtr = SpellingPtr;
2270 unsigned Len = MeasureUCNEscape(SpellingStart, SpellingPtr, SpellingEnd,
2271 1, Features, HadError);
2272 if (Len > ByteNo) {
2273 // ByteNo is somewhere within the escape sequence.
2274 SpellingPtr = EscapePtr;
2275 break;
2276 }
2277 ByteNo -= Len;
2278 } else {
2279 ProcessCharEscape(SpellingStart, SpellingPtr, SpellingEnd, HadError,
2280 FullSourceLoc(Tok.getLocation(), SM),
2281 CharByteWidth*8, Diags, Features);
2282 --ByteNo;
2283 }
2284 assert(!HadError && "This method isn't valid on erroneous strings")(static_cast <bool> (!HadError && "This method isn't valid on erroneous strings"
) ? void (0) : __assert_fail ("!HadError && \"This method isn't valid on erroneous strings\""
, "clang/lib/Lex/LiteralSupport.cpp", 2284, __extension__ __PRETTY_FUNCTION__
))
;
2285 }
2286
2287 return SpellingPtr-SpellingStart;
2288}
2289
2290/// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
2291/// suffixes as ud-suffixes, because the diagnostic experience is better if we
2292/// treat it as an invalid suffix.
2293bool StringLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
2294 StringRef Suffix) {
2295 return NumericLiteralParser::isValidUDSuffix(LangOpts, Suffix) ||
2296 Suffix == "sv";
2297}