1 //===--- LiteralSupport.cpp - Code to parse and process literals ----------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the NumericLiteralParser, CharLiteralParser, and
11 // StringLiteralParser interfaces.
13 //===----------------------------------------------------------------------===//
15 #include "clang/Lex/LiteralSupport.h"
16 #include "clang/Basic/CharInfo.h"
17 #include "clang/Basic/LangOptions.h"
18 #include "clang/Basic/SourceLocation.h"
19 #include "clang/Basic/TargetInfo.h"
20 #include "clang/Lex/LexDiagnostic.h"
21 #include "clang/Lex/Lexer.h"
22 #include "clang/Lex/Preprocessor.h"
23 #include "clang/Lex/Token.h"
24 #include "llvm/ADT/APInt.h"
25 #include "llvm/ADT/SmallVector.h"
26 #include "llvm/ADT/StringExtras.h"
27 #include "llvm/ADT/StringSwitch.h"
28 #include "llvm/Support/ConvertUTF.h"
29 #include "llvm/Support/ErrorHandling.h"
37 using namespace clang;
39 static unsigned getCharWidth(tok::TokenKind kind, const TargetInfo &Target) {
41 default: llvm_unreachable("Unknown token type!");
42 case tok::char_constant:
43 case tok::string_literal:
44 case tok::utf8_char_constant:
45 case tok::utf8_string_literal:
46 return Target.getCharWidth();
47 case tok::wide_char_constant:
48 case tok::wide_string_literal:
49 return Target.getWCharWidth();
50 case tok::utf16_char_constant:
51 case tok::utf16_string_literal:
52 return Target.getChar16Width();
53 case tok::utf32_char_constant:
54 case tok::utf32_string_literal:
55 return Target.getChar32Width();
59 static CharSourceRange MakeCharSourceRange(const LangOptions &Features,
62 const char *TokRangeBegin,
63 const char *TokRangeEnd) {
64 SourceLocation Begin =
65 Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
66 TokLoc.getManager(), Features);
68 Lexer::AdvanceToTokenCharacter(Begin, TokRangeEnd - TokRangeBegin,
69 TokLoc.getManager(), Features);
70 return CharSourceRange::getCharRange(Begin, End);
73 /// \brief Produce a diagnostic highlighting some portion of a literal.
75 /// Emits the diagnostic \p DiagID, highlighting the range of characters from
76 /// \p TokRangeBegin (inclusive) to \p TokRangeEnd (exclusive), which must be
77 /// a substring of a spelling buffer for the token beginning at \p TokBegin.
78 static DiagnosticBuilder Diag(DiagnosticsEngine *Diags,
79 const LangOptions &Features, FullSourceLoc TokLoc,
80 const char *TokBegin, const char *TokRangeBegin,
81 const char *TokRangeEnd, unsigned DiagID) {
82 SourceLocation Begin =
83 Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin,
84 TokLoc.getManager(), Features);
85 return Diags->Report(Begin, DiagID) <<
86 MakeCharSourceRange(Features, TokLoc, TokBegin, TokRangeBegin, TokRangeEnd);
89 /// ProcessCharEscape - Parse a standard C escape sequence, which can occur in
90 /// either a character or a string literal.
91 static unsigned ProcessCharEscape(const char *ThisTokBegin,
92 const char *&ThisTokBuf,
93 const char *ThisTokEnd, bool &HadError,
94 FullSourceLoc Loc, unsigned CharWidth,
95 DiagnosticsEngine *Diags,
96 const LangOptions &Features) {
97 const char *EscapeBegin = ThisTokBuf;
102 // We know that this character can't be off the end of the buffer, because
103 // that would have been \", which would not have been the end of string.
104 unsigned ResultChar = *ThisTokBuf++;
105 switch (ResultChar) {
106 // These map to themselves.
107 case '\\': case '\'': case '"': case '?': break;
109 // These have fixed mappings.
111 // TODO: K&R: the meaning of '\\a' is different in traditional C
119 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
120 diag::ext_nonstandard_escape) << "e";
125 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
126 diag::ext_nonstandard_escape) << "E";
144 case 'x': { // Hex escape.
146 if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
148 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
149 diag::err_hex_escape_no_digits) << "x";
154 // Hex escapes are a maximal series of hex digits.
155 bool Overflow = false;
156 for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) {
157 int CharVal = llvm::hexDigitValue(ThisTokBuf[0]);
158 if (CharVal == -1) break;
159 // About to shift out a digit?
160 if (ResultChar & 0xF0000000)
163 ResultChar |= CharVal;
166 // See if any bits will be truncated when evaluated as a character.
167 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
169 ResultChar &= ~0U >> (32-CharWidth);
172 // Check for overflow.
173 if (Overflow && Diags) // Too many digits to fit in
174 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
175 diag::err_escape_too_large) << 0;
178 case '0': case '1': case '2': case '3':
179 case '4': case '5': case '6': case '7': {
184 // Octal escapes are a series of octal digits with maximum length 3.
185 // "\0123" is a two digit sequence equal to "\012" "3".
186 unsigned NumDigits = 0;
189 ResultChar |= *ThisTokBuf++ - '0';
191 } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 &&
192 ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7');
194 // Check for overflow. Reject '\777', but not L'\777'.
195 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) {
197 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
198 diag::err_escape_too_large) << 1;
199 ResultChar &= ~0U >> (32-CharWidth);
204 // Otherwise, these are not valid escapes.
205 case '(': case '{': case '[': case '%':
206 // GCC accepts these as extensions. We warn about them as such though.
208 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
209 diag::ext_nonstandard_escape)
210 << std::string(1, ResultChar);
216 if (isPrintable(ResultChar))
217 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
218 diag::ext_unknown_escape)
219 << std::string(1, ResultChar);
221 Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf,
222 diag::ext_unknown_escape)
223 << "x" + llvm::utohexstr(ResultChar);
230 static void appendCodePoint(unsigned Codepoint,
231 llvm::SmallVectorImpl<char> &Str) {
233 char *ResultPtr = ResultBuf;
234 bool Res = llvm::ConvertCodePointToUTF8(Codepoint, ResultPtr);
236 assert(Res && "Unexpected conversion failure");
237 Str.append(ResultBuf, ResultPtr);
240 void clang::expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
241 for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
248 assert(*I == 'u' || *I == 'U');
250 unsigned NumHexDigits;
256 assert(I + NumHexDigits <= E);
258 uint32_t CodePoint = 0;
259 for (++I; NumHexDigits != 0; ++I, --NumHexDigits) {
260 unsigned Value = llvm::hexDigitValue(*I);
261 assert(Value != -1U);
267 appendCodePoint(CodePoint, Buf);
272 /// ProcessUCNEscape - Read the Universal Character Name, check constraints and
273 /// return the UTF32.
274 static bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
275 const char *ThisTokEnd,
276 uint32_t &UcnVal, unsigned short &UcnLen,
277 FullSourceLoc Loc, DiagnosticsEngine *Diags,
278 const LangOptions &Features,
279 bool in_char_string_literal = false) {
280 const char *UcnBegin = ThisTokBuf;
282 // Skip the '\u' char's.
285 if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) {
287 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
288 diag::err_hex_escape_no_digits) << StringRef(&ThisTokBuf[-1], 1);
291 UcnLen = (ThisTokBuf[-1] == 'u' ? 4 : 8);
292 unsigned short UcnLenSave = UcnLen;
293 for (; ThisTokBuf != ThisTokEnd && UcnLenSave; ++ThisTokBuf, UcnLenSave--) {
294 int CharVal = llvm::hexDigitValue(ThisTokBuf[0]);
295 if (CharVal == -1) break;
299 // If we didn't consume the proper number of digits, there is a problem.
302 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
303 diag::err_ucn_escape_incomplete);
307 // Check UCN constraints (C99 6.4.3p2) [C++11 lex.charset p2]
308 if ((0xD800 <= UcnVal && UcnVal <= 0xDFFF) || // surrogate codepoints
309 UcnVal > 0x10FFFF) { // maximum legal UTF32 value
311 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
312 diag::err_ucn_escape_invalid);
316 // C++11 allows UCNs that refer to control characters and basic source
317 // characters inside character and string literals
319 (UcnVal != 0x24 && UcnVal != 0x40 && UcnVal != 0x60)) { // $, @, `
320 bool IsError = (!Features.CPlusPlus11 || !in_char_string_literal);
322 char BasicSCSChar = UcnVal;
323 if (UcnVal >= 0x20 && UcnVal < 0x7f)
324 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
325 IsError ? diag::err_ucn_escape_basic_scs :
326 diag::warn_cxx98_compat_literal_ucn_escape_basic_scs)
327 << StringRef(&BasicSCSChar, 1);
329 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
330 IsError ? diag::err_ucn_control_character :
331 diag::warn_cxx98_compat_literal_ucn_control_character);
337 if (!Features.CPlusPlus && !Features.C99 && Diags)
338 Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf,
339 diag::warn_ucn_not_valid_in_c89_literal);
344 /// MeasureUCNEscape - Determine the number of bytes within the resulting string
345 /// which this UCN will occupy.
346 static int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
347 const char *ThisTokEnd, unsigned CharByteWidth,
348 const LangOptions &Features, bool &HadError) {
349 // UTF-32: 4 bytes per escape.
350 if (CharByteWidth == 4)
354 unsigned short UcnLen = 0;
357 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal,
358 UcnLen, Loc, nullptr, Features, true)) {
363 // UTF-16: 2 bytes for BMP, 4 bytes otherwise.
364 if (CharByteWidth == 2)
365 return UcnVal <= 0xFFFF ? 2 : 4;
372 if (UcnVal < 0x10000)
377 /// EncodeUCNEscape - Read the Universal Character Name, check constraints and
378 /// convert the UTF32 to UTF8 or UTF16. This is a subroutine of
379 /// StringLiteralParser. When we decide to implement UCN's for identifiers,
380 /// we will likely rework our support for UCN's.
381 static void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf,
382 const char *ThisTokEnd,
383 char *&ResultBuf, bool &HadError,
384 FullSourceLoc Loc, unsigned CharByteWidth,
385 DiagnosticsEngine *Diags,
386 const LangOptions &Features) {
387 typedef uint32_t UTF32;
389 unsigned short UcnLen = 0;
390 if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, UcnLen,
391 Loc, Diags, Features, true)) {
396 assert((CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) &&
397 "only character widths of 1, 2, or 4 bytes supported");
400 assert((UcnLen== 4 || UcnLen== 8) && "only ucn length of 4 or 8 supported");
402 if (CharByteWidth == 4) {
403 // FIXME: Make the type of the result buffer correct instead of
404 // using reinterpret_cast.
405 llvm::UTF32 *ResultPtr = reinterpret_cast<llvm::UTF32*>(ResultBuf);
411 if (CharByteWidth == 2) {
412 // FIXME: Make the type of the result buffer correct instead of
413 // using reinterpret_cast.
414 llvm::UTF16 *ResultPtr = reinterpret_cast<llvm::UTF16*>(ResultBuf);
416 if (UcnVal <= (UTF32)0xFFFF) {
424 *ResultPtr = 0xD800 + (UcnVal >> 10);
425 *(ResultPtr+1) = 0xDC00 + (UcnVal & 0x3FF);
430 assert(CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters");
432 // Now that we've parsed/checked the UCN, we convert from UTF32->UTF8.
433 // The conversion below was inspired by:
434 // http://www.unicode.org/Public/PROGRAMS/CVTUTF/ConvertUTF.c
435 // First, we determine how many bytes the result will require.
436 typedef uint8_t UTF8;
438 unsigned short bytesToWrite = 0;
439 if (UcnVal < (UTF32)0x80)
441 else if (UcnVal < (UTF32)0x800)
443 else if (UcnVal < (UTF32)0x10000)
448 const unsigned byteMask = 0xBF;
449 const unsigned byteMark = 0x80;
451 // Once the bits are split out into bytes of UTF8, this is a mask OR-ed
452 // into the first byte, depending on how many bytes follow.
453 static const UTF8 firstByteMark[5] = {
454 0x00, 0x00, 0xC0, 0xE0, 0xF0
456 // Finally, we write the bytes into ResultBuf.
457 ResultBuf += bytesToWrite;
458 switch (bytesToWrite) { // note: everything falls through.
459 case 4: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
460 case 3: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
461 case 2: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6;
462 case 1: *--ResultBuf = (UTF8) (UcnVal | firstByteMark[bytesToWrite]);
464 // Update the buffer.
465 ResultBuf += bytesToWrite;
468 /// integer-constant: [C99 6.4.4.1]
469 /// decimal-constant integer-suffix
470 /// octal-constant integer-suffix
471 /// hexadecimal-constant integer-suffix
472 /// binary-literal integer-suffix [GNU, C++1y]
473 /// user-defined-integer-literal: [C++11 lex.ext]
474 /// decimal-literal ud-suffix
475 /// octal-literal ud-suffix
476 /// hexadecimal-literal ud-suffix
477 /// binary-literal ud-suffix [GNU, C++1y]
478 /// decimal-constant:
480 /// decimal-constant digit
483 /// octal-constant octal-digit
484 /// hexadecimal-constant:
485 /// hexadecimal-prefix hexadecimal-digit
486 /// hexadecimal-constant hexadecimal-digit
487 /// hexadecimal-prefix: one of
492 /// binary-literal binary-digit
494 /// unsigned-suffix [long-suffix]
495 /// unsigned-suffix [long-long-suffix]
496 /// long-suffix [unsigned-suffix]
497 /// long-long-suffix [unsigned-sufix]
499 /// 1 2 3 4 5 6 7 8 9
502 /// hexadecimal-digit:
503 /// 0 1 2 3 4 5 6 7 8 9
509 /// unsigned-suffix: one of
511 /// long-suffix: one of
513 /// long-long-suffix: one of
516 /// floating-constant: [C99 6.4.4.2]
517 /// TODO: add rules...
519 NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
520 SourceLocation TokLoc,
522 : PP(PP), ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) {
524 // This routine assumes that the range begin/end matches the regex for integer
525 // and FP constants (specifically, the 'pp-number' regex), and assumes that
526 // the byte at "*end" is both valid and not part of the regex. Because of
527 // this, it doesn't have to check for 'overscan' in various places.
528 assert(!isPreprocessingNumberBody(*ThisTokEnd) && "didn't maximally munch?");
530 s = DigitsBegin = ThisTokBegin;
531 saw_exponent = false;
533 saw_ud_suffix = false;
541 MicrosoftInteger = 0;
544 if (*s == '0') { // parse radix
545 ParseNumberStartingWithZero(TokLoc);
548 } else { // the first digit is non-zero
551 if (s == ThisTokEnd) {
554 ParseDecimalOrOctalCommon(TokLoc);
561 checkSeparator(TokLoc, s, CSK_AfterDigits);
563 // Parse the suffix. At this point we can classify whether we have an FP or
565 bool isFPConstant = isFloatingLiteral();
567 // Loop over all of the characters of the suffix. If we see something bad,
568 // we break out of the loop.
569 for (; s != ThisTokEnd; ++s) {
571 case 'h': // FP Suffix for "half".
573 // OpenCL Extension v1.2 s9.5 - h or H suffix for half type.
574 if (!PP.getLangOpts().Half) break;
575 if (!isFPConstant) break; // Error for integer constant.
576 if (isHalf || isFloat || isLong) break; // HH, FH, LH invalid.
578 continue; // Success.
579 case 'f': // FP Suffix for "float"
581 if (!isFPConstant) break; // Error for integer constant.
582 if (isHalf || isFloat || isLong || isFloat128)
583 break; // HF, FF, LF, QF invalid.
585 continue; // Success.
586 case 'q': // FP Suffix for "__float128"
588 if (!isFPConstant) break; // Error for integer constant.
589 if (isHalf || isFloat || isLong || isFloat128)
590 break; // HQ, FQ, LQ, QQ invalid.
592 continue; // Success.
595 if (isFPConstant) break; // Error for floating constant.
596 if (isUnsigned) break; // Cannot be repeated.
598 continue; // Success.
601 if (isLong || isLongLong) break; // Cannot be repeated.
602 if (isHalf || isFloat || isFloat128) break; // LH, LF, LQ invalid.
604 // Check for long long. The L's need to be adjacent and the same case.
606 assert(s + 1 < ThisTokEnd && "didn't maximally munch?");
607 if (isFPConstant) break; // long long invalid for floats.
609 ++s; // Eat both of them.
613 continue; // Success.
616 if (PP.getLangOpts().MicrosoftExt) {
617 if (isLong || isLongLong || MicrosoftInteger)
621 // Allow i8, i16, i32, and i64.
625 MicrosoftInteger = 8;
629 s += 3; // i16 suffix
630 MicrosoftInteger = 16;
635 s += 3; // i32 suffix
636 MicrosoftInteger = 32;
641 s += 3; // i64 suffix
642 MicrosoftInteger = 64;
649 if (MicrosoftInteger) {
650 assert(s <= ThisTokEnd && "didn't maximally munch?");
654 // "i", "if", and "il" are user-defined suffixes in C++1y.
655 if (*s == 'i' && PP.getLangOpts().CPlusPlus14)
660 if (isImaginary) break; // Cannot be repeated.
662 continue; // Success.
664 // If we reached here, there was an error or a ud-suffix.
668 if (s != ThisTokEnd) {
669 // FIXME: Don't bother expanding UCNs if !tok.hasUCN().
670 expandUCNs(UDSuffixBuf, StringRef(SuffixBegin, ThisTokEnd - SuffixBegin));
671 if (isValidUDSuffix(PP.getLangOpts(), UDSuffixBuf)) {
672 // Any suffix pieces we might have parsed are actually part of the
680 MicrosoftInteger = 0;
682 saw_ud_suffix = true;
686 // Report an error if there are any.
687 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, SuffixBegin - ThisTokBegin),
688 diag::err_invalid_suffix_constant)
689 << StringRef(SuffixBegin, ThisTokEnd-SuffixBegin) << isFPConstant;
695 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, SuffixBegin - ThisTokBegin),
696 diag::ext_imaginary_constant);
700 /// ParseDecimalOrOctalCommon - This method is called for decimal or octal
701 /// numbers. It issues an error for illegal digits, and handles floating point
702 /// parsing. If it detects a floating point number, the radix is set to 10.
703 void NumericLiteralParser::ParseDecimalOrOctalCommon(SourceLocation TokLoc){
704 assert((radix == 8 || radix == 10) && "Unexpected radix");
706 // If we have a hex digit other than 'e' (which denotes a FP exponent) then
707 // the code is using an incorrect base.
708 if (isHexDigit(*s) && *s != 'e' && *s != 'E') {
709 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin),
710 diag::err_invalid_digit) << StringRef(s, 1) << (radix == 8 ? 1 : 0);
716 checkSeparator(TokLoc, s, CSK_AfterDigits);
720 checkSeparator(TokLoc, s, CSK_BeforeDigits);
721 s = SkipDigits(s); // Skip suffix.
723 if (*s == 'e' || *s == 'E') { // exponent
724 checkSeparator(TokLoc, s, CSK_AfterDigits);
725 const char *Exponent = s;
729 if (*s == '+' || *s == '-') s++; // sign
730 const char *first_non_digit = SkipDigits(s);
731 if (containsDigits(s, first_non_digit)) {
732 checkSeparator(TokLoc, s, CSK_BeforeDigits);
735 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin),
736 diag::err_exponent_has_no_digits);
743 /// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
744 /// suffixes as ud-suffixes, because the diagnostic experience is better if we
745 /// treat it as an invalid suffix.
746 bool NumericLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
748 if (!LangOpts.CPlusPlus11 || Suffix.empty())
751 // By C++11 [lex.ext]p10, ud-suffixes starting with an '_' are always valid.
752 if (Suffix[0] == '_')
755 // In C++11, there are no library suffixes.
756 if (!LangOpts.CPlusPlus14)
759 // In C++1y, "s", "h", "min", "ms", "us", and "ns" are used in the library.
760 // Per tweaked N3660, "il", "i", and "if" are also used in the library.
761 return llvm::StringSwitch<bool>(Suffix)
762 .Cases("h", "min", "s", true)
763 .Cases("ms", "us", "ns", true)
764 .Cases("il", "i", "if", true)
768 void NumericLiteralParser::checkSeparator(SourceLocation TokLoc,
770 CheckSeparatorKind IsAfterDigits) {
771 if (IsAfterDigits == CSK_AfterDigits) {
772 if (Pos == ThisTokBegin)
775 } else if (Pos == ThisTokEnd)
778 if (isDigitSeparator(*Pos))
779 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Pos - ThisTokBegin),
780 diag::err_digit_separator_not_between_digits)
784 /// ParseNumberStartingWithZero - This method is called when the first character
785 /// of the number is found to be a zero. This means it is either an octal
786 /// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or
787 /// a floating point number (01239.123e4). Eat the prefix, determining the
789 void NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) {
790 assert(s[0] == '0' && "Invalid method call");
795 // Handle a hex number like 0x1234.
796 if ((c1 == 'x' || c1 == 'X') && (isHexDigit(s[1]) || s[1] == '.')) {
798 assert(s < ThisTokEnd && "didn't maximally munch?");
801 s = SkipHexDigits(s);
802 bool HasSignificandDigits = containsDigits(DigitsBegin, s);
803 if (s == ThisTokEnd) {
805 } else if (*s == '.') {
808 const char *floatDigitsBegin = s;
809 s = SkipHexDigits(s);
810 if (containsDigits(floatDigitsBegin, s))
811 HasSignificandDigits = true;
812 if (HasSignificandDigits)
813 checkSeparator(TokLoc, floatDigitsBegin, CSK_BeforeDigits);
816 if (!HasSignificandDigits) {
817 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin),
818 diag::err_hex_constant_requires)
819 << PP.getLangOpts().CPlusPlus << 1;
824 // A binary exponent can appear with or with a '.'. If dotted, the
825 // binary exponent is required.
826 if (*s == 'p' || *s == 'P') {
827 checkSeparator(TokLoc, s, CSK_AfterDigits);
828 const char *Exponent = s;
831 if (*s == '+' || *s == '-') s++; // sign
832 const char *first_non_digit = SkipDigits(s);
833 if (!containsDigits(s, first_non_digit)) {
834 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin),
835 diag::err_exponent_has_no_digits);
839 checkSeparator(TokLoc, s, CSK_BeforeDigits);
842 if (!PP.getLangOpts().HexFloats)
843 PP.Diag(TokLoc, PP.getLangOpts().CPlusPlus
844 ? diag::ext_hex_literal_invalid
845 : diag::ext_hex_constant_invalid);
846 else if (PP.getLangOpts().CPlusPlus1z)
847 PP.Diag(TokLoc, diag::warn_cxx1z_hex_literal);
848 } else if (saw_period) {
849 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin),
850 diag::err_hex_constant_requires)
851 << PP.getLangOpts().CPlusPlus << 0;
857 // Handle simple binary numbers 0b01010
858 if ((c1 == 'b' || c1 == 'B') && (s[1] == '0' || s[1] == '1')) {
859 // 0b101010 is a C++1y / GCC extension.
861 PP.getLangOpts().CPlusPlus14
862 ? diag::warn_cxx11_compat_binary_literal
863 : PP.getLangOpts().CPlusPlus
864 ? diag::ext_binary_literal_cxx14
865 : diag::ext_binary_literal);
867 assert(s < ThisTokEnd && "didn't maximally munch?");
870 s = SkipBinaryDigits(s);
871 if (s == ThisTokEnd) {
873 } else if (isHexDigit(*s)) {
874 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin),
875 diag::err_invalid_digit) << StringRef(s, 1) << 2;
878 // Other suffixes will be diagnosed by the caller.
882 // For now, the radix is set to 8. If we discover that we have a
883 // floating point constant, the radix will change to 10. Octal floating
884 // point constants are not permitted (only decimal and hexadecimal).
887 s = SkipOctalDigits(s);
889 return; // Done, simple octal number like 01234
891 // If we have some other non-octal digit that *is* a decimal digit, see if
892 // this is part of a floating point number like 094.123 or 09e1.
894 const char *EndDecimal = SkipDigits(s);
895 if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') {
901 ParseDecimalOrOctalCommon(TokLoc);
904 static bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits) {
907 return NumDigits <= 64;
909 return NumDigits <= 64 / 3; // Digits are groups of 3 bits.
911 return NumDigits <= 19; // floor(log10(2^64))
913 return NumDigits <= 64 / 4; // Digits are groups of 4 bits.
915 llvm_unreachable("impossible Radix");
919 /// GetIntegerValue - Convert this numeric literal value to an APInt that
920 /// matches Val's input width. If there is an overflow, set Val to the low bits
921 /// of the result and return true. Otherwise, return false.
922 bool NumericLiteralParser::GetIntegerValue(llvm::APInt &Val) {
923 // Fast path: Compute a conservative bound on the maximum number of
924 // bits per digit in this radix. If we can't possibly overflow a
925 // uint64 based on that bound then do the simple conversion to
926 // integer. This avoids the expensive overflow checking below, and
927 // handles the common cases that matter (small decimal integers and
928 // hex/octal values which don't overflow).
929 const unsigned NumDigits = SuffixBegin - DigitsBegin;
930 if (alwaysFitsInto64Bits(radix, NumDigits)) {
932 for (const char *Ptr = DigitsBegin; Ptr != SuffixBegin; ++Ptr)
933 if (!isDigitSeparator(*Ptr))
934 N = N * radix + llvm::hexDigitValue(*Ptr);
936 // This will truncate the value to Val's input width. Simply check
937 // for overflow by comparing.
939 return Val.getZExtValue() != N;
943 const char *Ptr = DigitsBegin;
945 llvm::APInt RadixVal(Val.getBitWidth(), radix);
946 llvm::APInt CharVal(Val.getBitWidth(), 0);
947 llvm::APInt OldVal = Val;
949 bool OverflowOccurred = false;
950 while (Ptr < SuffixBegin) {
951 if (isDigitSeparator(*Ptr)) {
956 unsigned C = llvm::hexDigitValue(*Ptr++);
958 // If this letter is out of bound for this radix, reject it.
959 assert(C < radix && "NumericLiteralParser ctor should have rejected this");
963 // Add the digit to the value in the appropriate radix. If adding in digits
964 // made the value smaller, then this overflowed.
967 // Multiply by radix, did overflow occur on the multiply?
969 OverflowOccurred |= Val.udiv(RadixVal) != OldVal;
971 // Add value, did overflow occur on the value?
972 // (a + b) ult b <=> overflow
974 OverflowOccurred |= Val.ult(CharVal);
976 return OverflowOccurred;
979 llvm::APFloat::opStatus
980 NumericLiteralParser::GetFloatValue(llvm::APFloat &Result) {
983 unsigned n = std::min(SuffixBegin - ThisTokBegin, ThisTokEnd - ThisTokBegin);
985 llvm::SmallString<16> Buffer;
986 StringRef Str(ThisTokBegin, n);
987 if (Str.find('\'') != StringRef::npos) {
989 std::remove_copy_if(Str.begin(), Str.end(), std::back_inserter(Buffer),
994 return Result.convertFromString(Str, APFloat::rmNearestTiesToEven);
998 /// user-defined-character-literal: [C++11 lex.ext]
999 /// character-literal ud-suffix
1002 /// character-literal: [C++11 lex.ccon]
1003 /// ' c-char-sequence '
1004 /// u' c-char-sequence '
1005 /// U' c-char-sequence '
1006 /// L' c-char-sequence '
1007 /// u8' c-char-sequence ' [C++1z lex.ccon]
1008 /// c-char-sequence:
1010 /// c-char-sequence c-char
1012 /// any member of the source character set except the single-quote ',
1013 /// backslash \, or new-line character
1015 /// universal-character-name
1016 /// escape-sequence:
1017 /// simple-escape-sequence
1018 /// octal-escape-sequence
1019 /// hexadecimal-escape-sequence
1020 /// simple-escape-sequence:
1021 /// one of \' \" \? \\ \a \b \f \n \r \t \v
1022 /// octal-escape-sequence:
1024 /// \ octal-digit octal-digit
1025 /// \ octal-digit octal-digit octal-digit
1026 /// hexadecimal-escape-sequence:
1027 /// \x hexadecimal-digit
1028 /// hexadecimal-escape-sequence hexadecimal-digit
1029 /// universal-character-name: [C++11 lex.charset]
1031 /// \U hex-quad hex-quad
1033 /// hex-digit hex-digit hex-digit hex-digit
1036 CharLiteralParser::CharLiteralParser(const char *begin, const char *end,
1037 SourceLocation Loc, Preprocessor &PP,
1038 tok::TokenKind kind) {
1039 // At this point we know that the character matches the regex "(L|u|U)?'.*'".
1044 const char *TokBegin = begin;
1046 // Skip over wide character determinant.
1047 if (Kind != tok::char_constant)
1049 if (Kind == tok::utf8_char_constant)
1052 // Skip over the entry quote.
1053 assert(begin[0] == '\'' && "Invalid token lexed");
1056 // Remove an optional ud-suffix.
1057 if (end[-1] != '\'') {
1058 const char *UDSuffixEnd = end;
1061 } while (end[-1] != '\'');
1062 // FIXME: Don't bother with this if !tok.hasUCN().
1063 expandUCNs(UDSuffixBuf, StringRef(end, UDSuffixEnd - end));
1064 UDSuffixOffset = end - TokBegin;
1067 // Trim the ending quote.
1068 assert(end != begin && "Invalid token lexed");
1071 // FIXME: The "Value" is an uint64_t so we can handle char literals of
1073 // FIXME: This extensively assumes that 'char' is 8-bits.
1074 assert(PP.getTargetInfo().getCharWidth() == 8 &&
1075 "Assumes char is 8 bits");
1076 assert(PP.getTargetInfo().getIntWidth() <= 64 &&
1077 (PP.getTargetInfo().getIntWidth() & 7) == 0 &&
1078 "Assumes sizeof(int) on target is <= 64 and a multiple of char");
1079 assert(PP.getTargetInfo().getWCharWidth() <= 64 &&
1080 "Assumes sizeof(wchar) on target is <= 64");
1082 SmallVector<uint32_t, 4> codepoint_buffer;
1083 codepoint_buffer.resize(end - begin);
1084 uint32_t *buffer_begin = &codepoint_buffer.front();
1085 uint32_t *buffer_end = buffer_begin + codepoint_buffer.size();
1087 // Unicode escapes representing characters that cannot be correctly
1088 // represented in a single code unit are disallowed in character literals
1089 // by this implementation.
1090 uint32_t largest_character_for_kind;
1091 if (tok::wide_char_constant == Kind) {
1092 largest_character_for_kind =
1093 0xFFFFFFFFu >> (32-PP.getTargetInfo().getWCharWidth());
1094 } else if (tok::utf8_char_constant == Kind) {
1095 largest_character_for_kind = 0x7F;
1096 } else if (tok::utf16_char_constant == Kind) {
1097 largest_character_for_kind = 0xFFFF;
1098 } else if (tok::utf32_char_constant == Kind) {
1099 largest_character_for_kind = 0x10FFFF;
1101 largest_character_for_kind = 0x7Fu;
1104 while (begin != end) {
1105 // Is this a span of non-escape characters?
1106 if (begin[0] != '\\') {
1107 char const *start = begin;
1110 } while (begin != end && *begin != '\\');
1112 char const *tmp_in_start = start;
1113 uint32_t *tmp_out_start = buffer_begin;
1114 llvm::ConversionResult res =
1115 llvm::ConvertUTF8toUTF32(reinterpret_cast<llvm::UTF8 const **>(&start),
1116 reinterpret_cast<llvm::UTF8 const *>(begin),
1117 &buffer_begin, buffer_end, llvm::strictConversion);
1118 if (res != llvm::conversionOK) {
1119 // If we see bad encoding for unprefixed character literals, warn and
1120 // simply copy the byte values, for compatibility with gcc and
1121 // older versions of clang.
1122 bool NoErrorOnBadEncoding = isAscii();
1123 unsigned Msg = diag::err_bad_character_encoding;
1124 if (NoErrorOnBadEncoding)
1125 Msg = diag::warn_bad_character_encoding;
1127 if (NoErrorOnBadEncoding) {
1128 start = tmp_in_start;
1129 buffer_begin = tmp_out_start;
1130 for (; start != begin; ++start, ++buffer_begin)
1131 *buffer_begin = static_cast<uint8_t>(*start);
1136 for (; tmp_out_start < buffer_begin; ++tmp_out_start) {
1137 if (*tmp_out_start > largest_character_for_kind) {
1139 PP.Diag(Loc, diag::err_character_too_large);
1146 // Is this a Universal Character Name escape?
1147 if (begin[1] == 'u' || begin[1] == 'U') {
1148 unsigned short UcnLen = 0;
1149 if (!ProcessUCNEscape(TokBegin, begin, end, *buffer_begin, UcnLen,
1150 FullSourceLoc(Loc, PP.getSourceManager()),
1151 &PP.getDiagnostics(), PP.getLangOpts(), true)) {
1153 } else if (*buffer_begin > largest_character_for_kind) {
1155 PP.Diag(Loc, diag::err_character_too_large);
1161 unsigned CharWidth = getCharWidth(Kind, PP.getTargetInfo());
1163 ProcessCharEscape(TokBegin, begin, end, HadError,
1164 FullSourceLoc(Loc,PP.getSourceManager()),
1165 CharWidth, &PP.getDiagnostics(), PP.getLangOpts());
1166 *buffer_begin++ = result;
1169 unsigned NumCharsSoFar = buffer_begin - &codepoint_buffer.front();
1171 if (NumCharsSoFar > 1) {
1173 PP.Diag(Loc, diag::warn_extraneous_char_constant);
1174 else if (isAscii() && NumCharsSoFar == 4)
1175 PP.Diag(Loc, diag::ext_four_char_character_literal);
1177 PP.Diag(Loc, diag::ext_multichar_character_literal);
1179 PP.Diag(Loc, diag::err_multichar_utf_character_literal);
1182 IsMultiChar = false;
1185 llvm::APInt LitVal(PP.getTargetInfo().getIntWidth(), 0);
1187 // Narrow character literals act as though their value is concatenated
1188 // in this implementation, but warn on overflow.
1189 bool multi_char_too_long = false;
1190 if (isAscii() && isMultiChar()) {
1192 for (size_t i = 0; i < NumCharsSoFar; ++i) {
1193 // check for enough leading zeros to shift into
1194 multi_char_too_long |= (LitVal.countLeadingZeros() < 8);
1196 LitVal = LitVal + (codepoint_buffer[i] & 0xFF);
1198 } else if (NumCharsSoFar > 0) {
1199 // otherwise just take the last character
1200 LitVal = buffer_begin[-1];
1203 if (!HadError && multi_char_too_long) {
1204 PP.Diag(Loc, diag::warn_char_constant_too_large);
1207 // Transfer the value from APInt to uint64_t
1208 Value = LitVal.getZExtValue();
1210 // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1")
1211 // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple
1212 // character constants are not sign extended in the this implementation:
1213 // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC.
1214 if (isAscii() && NumCharsSoFar == 1 && (Value & 128) &&
1215 PP.getLangOpts().CharIsSigned)
1216 Value = (signed char)Value;
1220 /// string-literal: [C++0x lex.string]
1221 /// encoding-prefix " [s-char-sequence] "
1222 /// encoding-prefix R raw-string
1223 /// encoding-prefix:
1228 /// s-char-sequence:
1230 /// s-char-sequence s-char
1232 /// any member of the source character set except the double-quote ",
1233 /// backslash \, or new-line character
1235 /// universal-character-name
1237 /// " d-char-sequence ( r-char-sequence ) d-char-sequence "
1238 /// r-char-sequence:
1240 /// r-char-sequence r-char
1242 /// any member of the source character set, except a right parenthesis )
1243 /// followed by the initial d-char-sequence (which may be empty)
1244 /// followed by a double quote ".
1245 /// d-char-sequence:
1247 /// d-char-sequence d-char
1249 /// any member of the basic source character set except:
1250 /// space, the left parenthesis (, the right parenthesis ),
1251 /// the backslash \, and the control characters representing horizontal
1252 /// tab, vertical tab, form feed, and newline.
1253 /// escape-sequence: [C++0x lex.ccon]
1254 /// simple-escape-sequence
1255 /// octal-escape-sequence
1256 /// hexadecimal-escape-sequence
1257 /// simple-escape-sequence:
1258 /// one of \' \" \? \\ \a \b \f \n \r \t \v
1259 /// octal-escape-sequence:
1261 /// \ octal-digit octal-digit
1262 /// \ octal-digit octal-digit octal-digit
1263 /// hexadecimal-escape-sequence:
1264 /// \x hexadecimal-digit
1265 /// hexadecimal-escape-sequence hexadecimal-digit
1266 /// universal-character-name:
1268 /// \U hex-quad hex-quad
1270 /// hex-digit hex-digit hex-digit hex-digit
1273 StringLiteralParser::
1274 StringLiteralParser(ArrayRef<Token> StringToks,
1275 Preprocessor &PP, bool Complain)
1276 : SM(PP.getSourceManager()), Features(PP.getLangOpts()),
1277 Target(PP.getTargetInfo()), Diags(Complain ? &PP.getDiagnostics() :nullptr),
1278 MaxTokenLength(0), SizeBound(0), CharByteWidth(0), Kind(tok::unknown),
1279 ResultPtr(ResultBuf.data()), hadError(false), Pascal(false) {
1283 void StringLiteralParser::init(ArrayRef<Token> StringToks){
1284 // The literal token may have come from an invalid source location (e.g. due
1285 // to a PCH error), in which case the token length will be 0.
1286 if (StringToks.empty() || StringToks[0].getLength() < 2)
1287 return DiagnoseLexingError(SourceLocation());
1289 // Scan all of the string portions, remember the max individual token length,
1290 // computing a bound on the concatenated string length, and see whether any
1291 // piece is a wide-string. If any of the string portions is a wide-string
1292 // literal, the result is a wide-string literal [C99 6.4.5p4].
1293 assert(!StringToks.empty() && "expected at least one token");
1294 MaxTokenLength = StringToks[0].getLength();
1295 assert(StringToks[0].getLength() >= 2 && "literal token is invalid!");
1296 SizeBound = StringToks[0].getLength()-2; // -2 for "".
1297 Kind = StringToks[0].getKind();
1301 // Implement Translation Phase #6: concatenation of string literals
1302 /// (C99 5.1.1.2p1). The common case is only one string fragment.
1303 for (unsigned i = 1; i != StringToks.size(); ++i) {
1304 if (StringToks[i].getLength() < 2)
1305 return DiagnoseLexingError(StringToks[i].getLocation());
1307 // The string could be shorter than this if it needs cleaning, but this is a
1308 // reasonable bound, which is all we need.
1309 assert(StringToks[i].getLength() >= 2 && "literal token is invalid!");
1310 SizeBound += StringToks[i].getLength()-2; // -2 for "".
1312 // Remember maximum string piece length.
1313 if (StringToks[i].getLength() > MaxTokenLength)
1314 MaxTokenLength = StringToks[i].getLength();
1316 // Remember if we see any wide or utf-8/16/32 strings.
1317 // Also check for illegal concatenations.
1318 if (StringToks[i].isNot(Kind) && StringToks[i].isNot(tok::string_literal)) {
1320 Kind = StringToks[i].getKind();
1323 Diags->Report(StringToks[i].getLocation(),
1324 diag::err_unsupported_string_concat);
1330 // Include space for the null terminator.
1333 // TODO: K&R warning: "traditional C rejects string constant concatenation"
1335 // Get the width in bytes of char/wchar_t/char16_t/char32_t
1336 CharByteWidth = getCharWidth(Kind, Target);
1337 assert((CharByteWidth & 7) == 0 && "Assumes character size is byte multiple");
1340 // The output buffer size needs to be large enough to hold wide characters.
1341 // This is a worst-case assumption which basically corresponds to L"" "long".
1342 SizeBound *= CharByteWidth;
1344 // Size the temporary buffer to hold the result string data.
1345 ResultBuf.resize(SizeBound);
1347 // Likewise, but for each string piece.
1348 SmallString<512> TokenBuf;
1349 TokenBuf.resize(MaxTokenLength);
1351 // Loop over all the strings, getting their spelling, and expanding them to
1352 // wide strings as appropriate.
1353 ResultPtr = &ResultBuf[0]; // Next byte to fill in.
1357 SourceLocation UDSuffixTokLoc;
1359 for (unsigned i = 0, e = StringToks.size(); i != e; ++i) {
1360 const char *ThisTokBuf = &TokenBuf[0];
1361 // Get the spelling of the token, which eliminates trigraphs, etc. We know
1362 // that ThisTokBuf points to a buffer that is big enough for the whole token
1363 // and 'spelled' tokens can only shrink.
1364 bool StringInvalid = false;
1365 unsigned ThisTokLen =
1366 Lexer::getSpelling(StringToks[i], ThisTokBuf, SM, Features,
1369 return DiagnoseLexingError(StringToks[i].getLocation());
1371 const char *ThisTokBegin = ThisTokBuf;
1372 const char *ThisTokEnd = ThisTokBuf+ThisTokLen;
1374 // Remove an optional ud-suffix.
1375 if (ThisTokEnd[-1] != '"') {
1376 const char *UDSuffixEnd = ThisTokEnd;
1379 } while (ThisTokEnd[-1] != '"');
1381 StringRef UDSuffix(ThisTokEnd, UDSuffixEnd - ThisTokEnd);
1383 if (UDSuffixBuf.empty()) {
1384 if (StringToks[i].hasUCN())
1385 expandUCNs(UDSuffixBuf, UDSuffix);
1387 UDSuffixBuf.assign(UDSuffix);
1389 UDSuffixOffset = ThisTokEnd - ThisTokBuf;
1390 UDSuffixTokLoc = StringToks[i].getLocation();
1392 SmallString<32> ExpandedUDSuffix;
1393 if (StringToks[i].hasUCN()) {
1394 expandUCNs(ExpandedUDSuffix, UDSuffix);
1395 UDSuffix = ExpandedUDSuffix;
1398 // C++11 [lex.ext]p8: At the end of phase 6, if a string literal is the
1399 // result of a concatenation involving at least one user-defined-string-
1400 // literal, all the participating user-defined-string-literals shall
1401 // have the same ud-suffix.
1402 if (UDSuffixBuf != UDSuffix) {
1404 SourceLocation TokLoc = StringToks[i].getLocation();
1405 Diags->Report(TokLoc, diag::err_string_concat_mixed_suffix)
1406 << UDSuffixBuf << UDSuffix
1407 << SourceRange(UDSuffixTokLoc, UDSuffixTokLoc)
1408 << SourceRange(TokLoc, TokLoc);
1415 // Strip the end quote.
1418 // TODO: Input character set mapping support.
1420 // Skip marker for wide or unicode strings.
1421 if (ThisTokBuf[0] == 'L' || ThisTokBuf[0] == 'u' || ThisTokBuf[0] == 'U') {
1423 // Skip 8 of u8 marker for utf8 strings.
1424 if (ThisTokBuf[0] == '8')
1428 // Check for raw string
1429 if (ThisTokBuf[0] == 'R') {
1430 ThisTokBuf += 2; // skip R"
1432 const char *Prefix = ThisTokBuf;
1433 while (ThisTokBuf[0] != '(')
1435 ++ThisTokBuf; // skip '('
1437 // Remove same number of characters from the end
1438 ThisTokEnd -= ThisTokBuf - Prefix;
1439 assert(ThisTokEnd >= ThisTokBuf && "malformed raw string literal");
1441 // C++14 [lex.string]p4: A source-file new-line in a raw string literal
1442 // results in a new-line in the resulting execution string-literal.
1443 StringRef RemainingTokenSpan(ThisTokBuf, ThisTokEnd - ThisTokBuf);
1444 while (!RemainingTokenSpan.empty()) {
1445 // Split the string literal on \r\n boundaries.
1446 size_t CRLFPos = RemainingTokenSpan.find("\r\n");
1447 StringRef BeforeCRLF = RemainingTokenSpan.substr(0, CRLFPos);
1448 StringRef AfterCRLF = RemainingTokenSpan.substr(CRLFPos);
1450 // Copy everything before the \r\n sequence into the string literal.
1451 if (CopyStringFragment(StringToks[i], ThisTokBegin, BeforeCRLF))
1454 // Point into the \n inside the \r\n sequence and operate on the
1455 // remaining portion of the literal.
1456 RemainingTokenSpan = AfterCRLF.substr(1);
1459 if (ThisTokBuf[0] != '"') {
1460 // The file may have come from PCH and then changed after loading the
1461 // PCH; Fail gracefully.
1462 return DiagnoseLexingError(StringToks[i].getLocation());
1464 ++ThisTokBuf; // skip "
1466 // Check if this is a pascal string
1467 if (Features.PascalStrings && ThisTokBuf + 1 != ThisTokEnd &&
1468 ThisTokBuf[0] == '\\' && ThisTokBuf[1] == 'p') {
1470 // If the \p sequence is found in the first token, we have a pascal string
1471 // Otherwise, if we already have a pascal string, ignore the first \p
1479 while (ThisTokBuf != ThisTokEnd) {
1480 // Is this a span of non-escape characters?
1481 if (ThisTokBuf[0] != '\\') {
1482 const char *InStart = ThisTokBuf;
1485 } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\');
1487 // Copy the character span over.
1488 if (CopyStringFragment(StringToks[i], ThisTokBegin,
1489 StringRef(InStart, ThisTokBuf - InStart)))
1493 // Is this a Universal Character Name escape?
1494 if (ThisTokBuf[1] == 'u' || ThisTokBuf[1] == 'U') {
1495 EncodeUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd,
1496 ResultPtr, hadError,
1497 FullSourceLoc(StringToks[i].getLocation(), SM),
1498 CharByteWidth, Diags, Features);
1501 // Otherwise, this is a non-UCN escape character. Process it.
1502 unsigned ResultChar =
1503 ProcessCharEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, hadError,
1504 FullSourceLoc(StringToks[i].getLocation(), SM),
1505 CharByteWidth*8, Diags, Features);
1507 if (CharByteWidth == 4) {
1508 // FIXME: Make the type of the result buffer correct instead of
1509 // using reinterpret_cast.
1510 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultPtr);
1511 *ResultWidePtr = ResultChar;
1513 } else if (CharByteWidth == 2) {
1514 // FIXME: Make the type of the result buffer correct instead of
1515 // using reinterpret_cast.
1516 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultPtr);
1517 *ResultWidePtr = ResultChar & 0xFFFF;
1520 assert(CharByteWidth == 1 && "Unexpected char width");
1521 *ResultPtr++ = ResultChar & 0xFF;
1528 if (CharByteWidth == 4) {
1529 // FIXME: Make the type of the result buffer correct instead of
1530 // using reinterpret_cast.
1531 llvm::UTF32 *ResultWidePtr = reinterpret_cast<llvm::UTF32*>(ResultBuf.data());
1532 ResultWidePtr[0] = GetNumStringChars() - 1;
1533 } else if (CharByteWidth == 2) {
1534 // FIXME: Make the type of the result buffer correct instead of
1535 // using reinterpret_cast.
1536 llvm::UTF16 *ResultWidePtr = reinterpret_cast<llvm::UTF16*>(ResultBuf.data());
1537 ResultWidePtr[0] = GetNumStringChars() - 1;
1539 assert(CharByteWidth == 1 && "Unexpected char width");
1540 ResultBuf[0] = GetNumStringChars() - 1;
1543 // Verify that pascal strings aren't too large.
1544 if (GetStringLength() > 256) {
1546 Diags->Report(StringToks.front().getLocation(),
1547 diag::err_pascal_string_too_long)
1548 << SourceRange(StringToks.front().getLocation(),
1549 StringToks.back().getLocation());
1554 // Complain if this string literal has too many characters.
1555 unsigned MaxChars = Features.CPlusPlus? 65536 : Features.C99 ? 4095 : 509;
1557 if (GetNumStringChars() > MaxChars)
1558 Diags->Report(StringToks.front().getLocation(),
1559 diag::ext_string_too_long)
1560 << GetNumStringChars() << MaxChars
1561 << (Features.CPlusPlus ? 2 : Features.C99 ? 1 : 0)
1562 << SourceRange(StringToks.front().getLocation(),
1563 StringToks.back().getLocation());
1567 static const char *resyncUTF8(const char *Err, const char *End) {
1570 End = Err + std::min<unsigned>(llvm::getNumBytesForUTF8(*Err), End-Err);
1571 while (++Err != End && (*Err & 0xC0) == 0x80)
1576 /// \brief This function copies from Fragment, which is a sequence of bytes
1577 /// within Tok's contents (which begin at TokBegin) into ResultPtr.
1578 /// Performs widening for multi-byte characters.
1579 bool StringLiteralParser::CopyStringFragment(const Token &Tok,
1580 const char *TokBegin,
1581 StringRef Fragment) {
1582 const llvm::UTF8 *ErrorPtrTmp;
1583 if (ConvertUTF8toWide(CharByteWidth, Fragment, ResultPtr, ErrorPtrTmp))
1586 // If we see bad encoding for unprefixed string literals, warn and
1587 // simply copy the byte values, for compatibility with gcc and older
1588 // versions of clang.
1589 bool NoErrorOnBadEncoding = isAscii();
1590 if (NoErrorOnBadEncoding) {
1591 memcpy(ResultPtr, Fragment.data(), Fragment.size());
1592 ResultPtr += Fragment.size();
1596 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
1598 FullSourceLoc SourceLoc(Tok.getLocation(), SM);
1599 const DiagnosticBuilder &Builder =
1600 Diag(Diags, Features, SourceLoc, TokBegin,
1601 ErrorPtr, resyncUTF8(ErrorPtr, Fragment.end()),
1602 NoErrorOnBadEncoding ? diag::warn_bad_string_encoding
1603 : diag::err_bad_string_encoding);
1605 const char *NextStart = resyncUTF8(ErrorPtr, Fragment.end());
1606 StringRef NextFragment(NextStart, Fragment.end()-NextStart);
1608 // Decode into a dummy buffer.
1609 SmallString<512> Dummy;
1610 Dummy.reserve(Fragment.size() * CharByteWidth);
1611 char *Ptr = Dummy.data();
1613 while (!ConvertUTF8toWide(CharByteWidth, NextFragment, Ptr, ErrorPtrTmp)) {
1614 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp);
1615 NextStart = resyncUTF8(ErrorPtr, Fragment.end());
1616 Builder << MakeCharSourceRange(Features, SourceLoc, TokBegin,
1617 ErrorPtr, NextStart);
1618 NextFragment = StringRef(NextStart, Fragment.end()-NextStart);
1621 return !NoErrorOnBadEncoding;
1624 void StringLiteralParser::DiagnoseLexingError(SourceLocation Loc) {
1627 Diags->Report(Loc, diag::err_lexing_string);
1630 /// getOffsetOfStringByte - This function returns the offset of the
1631 /// specified byte of the string data represented by Token. This handles
1632 /// advancing over escape sequences in the string.
1633 unsigned StringLiteralParser::getOffsetOfStringByte(const Token &Tok,
1634 unsigned ByteNo) const {
1635 // Get the spelling of the token.
1636 SmallString<32> SpellingBuffer;
1637 SpellingBuffer.resize(Tok.getLength());
1639 bool StringInvalid = false;
1640 const char *SpellingPtr = &SpellingBuffer[0];
1641 unsigned TokLen = Lexer::getSpelling(Tok, SpellingPtr, SM, Features,
1646 const char *SpellingStart = SpellingPtr;
1647 const char *SpellingEnd = SpellingPtr+TokLen;
1649 // Handle UTF-8 strings just like narrow strings.
1650 if (SpellingPtr[0] == 'u' && SpellingPtr[1] == '8')
1653 assert(SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' &&
1654 SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet");
1656 // For raw string literals, this is easy.
1657 if (SpellingPtr[0] == 'R') {
1658 assert(SpellingPtr[1] == '"' && "Should be a raw string literal!");
1661 while (*SpellingPtr != '(') {
1663 assert(SpellingPtr < SpellingEnd && "Missing ( for raw string literal");
1667 return SpellingPtr - SpellingStart + ByteNo;
1670 // Skip over the leading quote
1671 assert(SpellingPtr[0] == '"' && "Should be a string literal!");
1674 // Skip over bytes until we find the offset we're looking for.
1676 assert(SpellingPtr < SpellingEnd && "Didn't find byte offset!");
1678 // Step over non-escapes simply.
1679 if (*SpellingPtr != '\\') {
1685 // Otherwise, this is an escape character. Advance over it.
1686 bool HadError = false;
1687 if (SpellingPtr[1] == 'u' || SpellingPtr[1] == 'U') {
1688 const char *EscapePtr = SpellingPtr;
1689 unsigned Len = MeasureUCNEscape(SpellingStart, SpellingPtr, SpellingEnd,
1690 1, Features, HadError);
1692 // ByteNo is somewhere within the escape sequence.
1693 SpellingPtr = EscapePtr;
1698 ProcessCharEscape(SpellingStart, SpellingPtr, SpellingEnd, HadError,
1699 FullSourceLoc(Tok.getLocation(), SM),
1700 CharByteWidth*8, Diags, Features);
1703 assert(!HadError && "This method isn't valid on erroneous strings");
1706 return SpellingPtr-SpellingStart;
1709 /// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved
1710 /// suffixes as ud-suffixes, because the diagnostic experience is better if we
1711 /// treat it as an invalid suffix.
1712 bool StringLiteralParser::isValidUDSuffix(const LangOptions &LangOpts,
1714 return NumericLiteralParser::isValidUDSuffix(LangOpts, Suffix) ||