LiteralSupport.cpp revision 70f66ab053f36ab3df7a778d09bcb2b4b0fec1f8
1//===--- LiteralSupport.cpp - Code to parse and process literals ----------===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9// 10// This file implements the NumericLiteralParser, CharLiteralParser, and 11// StringLiteralParser interfaces. 12// 13//===----------------------------------------------------------------------===// 14 15#include "clang/Lex/LiteralSupport.h" 16#include "clang/Lex/Preprocessor.h" 17#include "clang/Basic/Diagnostic.h" 18#include "clang/Basic/SourceManager.h" 19#include "clang/Basic/TargetInfo.h" 20#include "llvm/ADT/StringExtras.h" 21using namespace clang; 22 23/// HexDigitValue - Return the value of the specified hex digit, or -1 if it's 24/// not valid. 25static int HexDigitValue(char C) { 26 if (C >= '0' && C <= '9') return C-'0'; 27 if (C >= 'a' && C <= 'f') return C-'a'+10; 28 if (C >= 'A' && C <= 'F') return C-'A'+10; 29 return -1; 30} 31 32/// ProcessCharEscape - Parse a standard C escape sequence, which can occur in 33/// either a character or a string literal. 34static unsigned ProcessCharEscape(const char *&ThisTokBuf, 35 const char *ThisTokEnd, bool &HadError, 36 SourceLocation Loc, bool IsWide, 37 Preprocessor &PP) { 38 // Skip the '\' char. 39 ++ThisTokBuf; 40 41 // We know that this character can't be off the end of the buffer, because 42 // that would have been \", which would not have been the end of string. 43 unsigned ResultChar = *ThisTokBuf++; 44 switch (ResultChar) { 45 // These map to themselves. 46 case '\\': case '\'': case '"': case '?': break; 47 48 // These have fixed mappings. 49 case 'a': 50 // TODO: K&R: the meaning of '\\a' is different in traditional C 51 ResultChar = 7; 52 break; 53 case 'b': 54 ResultChar = 8; 55 break; 56 case 'e': 57 PP.Diag(Loc, diag::ext_nonstandard_escape, "e"); 58 ResultChar = 27; 59 break; 60 case 'f': 61 ResultChar = 12; 62 break; 63 case 'n': 64 ResultChar = 10; 65 break; 66 case 'r': 67 ResultChar = 13; 68 break; 69 case 't': 70 ResultChar = 9; 71 break; 72 case 'v': 73 ResultChar = 11; 74 break; 75 76 //case 'u': case 'U': // FIXME: UCNs. 77 case 'x': { // Hex escape. 78 ResultChar = 0; 79 if (ThisTokBuf == ThisTokEnd || !isxdigit(*ThisTokBuf)) { 80 PP.Diag(Loc, diag::err_hex_escape_no_digits); 81 HadError = 1; 82 break; 83 } 84 85 // Hex escapes are a maximal series of hex digits. 86 bool Overflow = false; 87 for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) { 88 int CharVal = HexDigitValue(ThisTokBuf[0]); 89 if (CharVal == -1) break; 90 Overflow |= (ResultChar & 0xF0000000) ? true : false; // About to shift out a digit? 91 ResultChar <<= 4; 92 ResultChar |= CharVal; 93 } 94 95 // See if any bits will be truncated when evaluated as a character. 96 unsigned CharWidth = PP.getTargetInfo().getCharWidth(IsWide); 97 98 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) { 99 Overflow = true; 100 ResultChar &= ~0U >> (32-CharWidth); 101 } 102 103 // Check for overflow. 104 if (Overflow) // Too many digits to fit in 105 PP.Diag(Loc, diag::warn_hex_escape_too_large); 106 break; 107 } 108 case '0': case '1': case '2': case '3': 109 case '4': case '5': case '6': case '7': { 110 // Octal escapes. 111 --ThisTokBuf; 112 ResultChar = 0; 113 114 // Octal escapes are a series of octal digits with maximum length 3. 115 // "\0123" is a two digit sequence equal to "\012" "3". 116 unsigned NumDigits = 0; 117 do { 118 ResultChar <<= 3; 119 ResultChar |= *ThisTokBuf++ - '0'; 120 ++NumDigits; 121 } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 && 122 ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7'); 123 124 // Check for overflow. Reject '\777', but not L'\777'. 125 unsigned CharWidth = PP.getTargetInfo().getCharWidth(IsWide); 126 127 if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) { 128 PP.Diag(Loc, diag::warn_octal_escape_too_large); 129 ResultChar &= ~0U >> (32-CharWidth); 130 } 131 break; 132 } 133 134 // Otherwise, these are not valid escapes. 135 case '(': case '{': case '[': case '%': 136 // GCC accepts these as extensions. We warn about them as such though. 137 if (!PP.getLangOptions().NoExtensions) { 138 PP.Diag(Loc, diag::ext_nonstandard_escape, 139 std::string()+(char)ResultChar); 140 break; 141 } 142 // FALL THROUGH. 143 default: 144 if (isgraph(ThisTokBuf[0])) { 145 PP.Diag(Loc, diag::ext_unknown_escape, std::string()+(char)ResultChar); 146 } else { 147 PP.Diag(Loc, diag::ext_unknown_escape, "x"+llvm::utohexstr(ResultChar)); 148 } 149 break; 150 } 151 152 return ResultChar; 153} 154 155 156 157 158/// integer-constant: [C99 6.4.4.1] 159/// decimal-constant integer-suffix 160/// octal-constant integer-suffix 161/// hexadecimal-constant integer-suffix 162/// decimal-constant: 163/// nonzero-digit 164/// decimal-constant digit 165/// octal-constant: 166/// 0 167/// octal-constant octal-digit 168/// hexadecimal-constant: 169/// hexadecimal-prefix hexadecimal-digit 170/// hexadecimal-constant hexadecimal-digit 171/// hexadecimal-prefix: one of 172/// 0x 0X 173/// integer-suffix: 174/// unsigned-suffix [long-suffix] 175/// unsigned-suffix [long-long-suffix] 176/// long-suffix [unsigned-suffix] 177/// long-long-suffix [unsigned-sufix] 178/// nonzero-digit: 179/// 1 2 3 4 5 6 7 8 9 180/// octal-digit: 181/// 0 1 2 3 4 5 6 7 182/// hexadecimal-digit: 183/// 0 1 2 3 4 5 6 7 8 9 184/// a b c d e f 185/// A B C D E F 186/// unsigned-suffix: one of 187/// u U 188/// long-suffix: one of 189/// l L 190/// long-long-suffix: one of 191/// ll LL 192/// 193/// floating-constant: [C99 6.4.4.2] 194/// TODO: add rules... 195/// 196 197NumericLiteralParser:: 198NumericLiteralParser(const char *begin, const char *end, 199 SourceLocation TokLoc, Preprocessor &pp) 200 : PP(pp), ThisTokBegin(begin), ThisTokEnd(end) { 201 s = DigitsBegin = begin; 202 saw_exponent = false; 203 saw_period = false; 204 isLong = false; 205 isUnsigned = false; 206 isLongLong = false; 207 isFloat = false; 208 isImaginary = false; 209 hadError = false; 210 211 if (*s == '0') { // parse radix 212 s++; 213 if ((*s == 'x' || *s == 'X') && (isxdigit(s[1]) || s[1] == '.')) { 214 s++; 215 radix = 16; 216 DigitsBegin = s; 217 s = SkipHexDigits(s); 218 if (s == ThisTokEnd) { 219 // Done. 220 } else if (*s == '.') { 221 s++; 222 saw_period = true; 223 s = SkipHexDigits(s); 224 } 225 // A binary exponent can appear with or with a '.'. If dotted, the 226 // binary exponent is required. 227 if ((*s == 'p' || *s == 'P') && PP.getLangOptions().HexFloats) { 228 const char *Exponent = s; 229 s++; 230 saw_exponent = true; 231 if (*s == '+' || *s == '-') s++; // sign 232 const char *first_non_digit = SkipDigits(s); 233 if (first_non_digit != s) { 234 s = first_non_digit; 235 } else { 236 Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-begin), 237 diag::err_exponent_has_no_digits); 238 return; 239 } 240 } else if (saw_period) { 241 Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 242 diag::err_hexconstant_requires_exponent); 243 return; 244 } 245 } else if (*s == 'b' || *s == 'B') { 246 // 0b101010 is a GCC extension. 247 ++s; 248 radix = 2; 249 DigitsBegin = s; 250 s = SkipBinaryDigits(s); 251 if (s == ThisTokEnd) { 252 // Done. 253 } else if (isxdigit(*s)) { 254 Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 255 diag::err_invalid_binary_digit, std::string(s, s+1)); 256 return; 257 } 258 PP.Diag(TokLoc, diag::ext_binary_literal); 259 } else { 260 // For now, the radix is set to 8. If we discover that we have a 261 // floating point constant, the radix will change to 10. Octal floating 262 // point constants are not permitted (only decimal and hexadecimal). 263 radix = 8; 264 DigitsBegin = s; 265 s = SkipOctalDigits(s); 266 if (s == ThisTokEnd) { 267 // Done. 268 } else if (isxdigit(*s) && !(*s == 'e' || *s == 'E')) { 269 Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 270 diag::err_invalid_octal_digit, std::string(s, s+1)); 271 return; 272 } else if (*s == '.') { 273 s++; 274 radix = 10; 275 saw_period = true; 276 s = SkipDigits(s); 277 } 278 if (*s == 'e' || *s == 'E') { // exponent 279 const char *Exponent = s; 280 s++; 281 radix = 10; 282 saw_exponent = true; 283 if (*s == '+' || *s == '-') s++; // sign 284 const char *first_non_digit = SkipDigits(s); 285 if (first_non_digit != s) { 286 s = first_non_digit; 287 } else { 288 Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-begin), 289 diag::err_exponent_has_no_digits); 290 return; 291 } 292 } 293 } 294 } else { // the first digit is non-zero 295 radix = 10; 296 s = SkipDigits(s); 297 if (s == ThisTokEnd) { 298 // Done. 299 } else if (isxdigit(*s) && !(*s == 'e' || *s == 'E')) { 300 Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 301 diag::err_invalid_decimal_digit, std::string(s, s+1)); 302 return; 303 } else if (*s == '.') { 304 s++; 305 saw_period = true; 306 s = SkipDigits(s); 307 } 308 if (*s == 'e' || *s == 'E') { // exponent 309 const char *Exponent = s; 310 s++; 311 saw_exponent = true; 312 if (*s == '+' || *s == '-') s++; // sign 313 const char *first_non_digit = SkipDigits(s); 314 if (first_non_digit != s) { 315 s = first_non_digit; 316 } else { 317 Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-begin), 318 diag::err_exponent_has_no_digits); 319 return; 320 } 321 } 322 } 323 324 SuffixBegin = s; 325 326 // Parse the suffix. At this point we can classify whether we have an FP or 327 // integer constant. 328 bool isFPConstant = isFloatingLiteral(); 329 330 // Loop over all of the characters of the suffix. If we see something bad, 331 // we break out of the loop. 332 for (; s != ThisTokEnd; ++s) { 333 switch (*s) { 334 case 'f': // FP Suffix for "float" 335 case 'F': 336 if (!isFPConstant) break; // Error for integer constant. 337 if (isFloat || isLong) break; // FF, LF invalid. 338 isFloat = true; 339 continue; // Success. 340 case 'u': 341 case 'U': 342 if (isFPConstant) break; // Error for floating constant. 343 if (isUnsigned) break; // Cannot be repeated. 344 isUnsigned = true; 345 continue; // Success. 346 case 'l': 347 case 'L': 348 if (isLong || isLongLong) break; // Cannot be repeated. 349 if (isFloat) break; // LF invalid. 350 351 // Check for long long. The L's need to be adjacent and the same case. 352 if (s+1 != ThisTokEnd && s[1] == s[0]) { 353 if (isFPConstant) break; // long long invalid for floats. 354 isLongLong = true; 355 ++s; // Eat both of them. 356 } else { 357 isLong = true; 358 } 359 continue; // Success. 360 case 'i': 361 if (PP.getLangOptions().Microsoft) { 362 // Allow i8, i16, i32, i64, and i128. 363 if (++s == ThisTokEnd) break; 364 switch (*s) { 365 case '8': 366 s++; // i8 suffix 367 break; 368 case '1': 369 if (++s == ThisTokEnd) break; 370 if (*s == '6') s++; // i16 suffix 371 else if (*s == '2') { 372 if (++s == ThisTokEnd) break; 373 if (*s == '8') s++; // i128 suffix 374 } 375 break; 376 case '3': 377 if (++s == ThisTokEnd) break; 378 if (*s == '2') s++; // i32 suffix 379 break; 380 case '6': 381 if (++s == ThisTokEnd) break; 382 if (*s == '4') s++; // i64 suffix 383 break; 384 default: 385 break; 386 } 387 break; 388 } 389 // fall through. 390 case 'I': 391 case 'j': 392 case 'J': 393 if (isImaginary) break; // Cannot be repeated. 394 PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 395 diag::ext_imaginary_constant); 396 isImaginary = true; 397 continue; // Success. 398 } 399 // If we reached here, there was an error. 400 break; 401 } 402 403 // Report an error if there are any. 404 if (s != ThisTokEnd) { 405 Diag(PP.AdvanceToTokenCharacter(TokLoc, s-begin), 406 isFPConstant ? diag::err_invalid_suffix_float_constant : 407 diag::err_invalid_suffix_integer_constant, 408 std::string(SuffixBegin, ThisTokEnd)); 409 return; 410 } 411} 412 413/// GetIntegerValue - Convert this numeric literal value to an APInt that 414/// matches Val's input width. If there is an overflow, set Val to the low bits 415/// of the result and return true. Otherwise, return false. 416bool NumericLiteralParser::GetIntegerValue(llvm::APInt &Val) { 417 Val = 0; 418 s = DigitsBegin; 419 420 llvm::APInt RadixVal(Val.getBitWidth(), radix); 421 llvm::APInt CharVal(Val.getBitWidth(), 0); 422 llvm::APInt OldVal = Val; 423 424 bool OverflowOccurred = false; 425 while (s < SuffixBegin) { 426 unsigned C = HexDigitValue(*s++); 427 428 // If this letter is out of bound for this radix, reject it. 429 assert(C < radix && "NumericLiteralParser ctor should have rejected this"); 430 431 CharVal = C; 432 433 // Add the digit to the value in the appropriate radix. If adding in digits 434 // made the value smaller, then this overflowed. 435 OldVal = Val; 436 437 // Multiply by radix, did overflow occur on the multiply? 438 Val *= RadixVal; 439 OverflowOccurred |= Val.udiv(RadixVal) != OldVal; 440 441 OldVal = Val; 442 // Add value, did overflow occur on the value? 443 Val += CharVal; 444 OverflowOccurred |= Val.ult(OldVal); 445 OverflowOccurred |= Val.ult(CharVal); 446 } 447 return OverflowOccurred; 448} 449 450llvm::APFloat NumericLiteralParser:: 451GetFloatValue(const llvm::fltSemantics &Format, bool* isExact) { 452 using llvm::APFloat; 453 454 llvm::SmallVector<char,256> floatChars; 455 for (unsigned i = 0, n = ThisTokEnd-ThisTokBegin; i != n; ++i) 456 floatChars.push_back(ThisTokBegin[i]); 457 458 floatChars.push_back('\0'); 459 460 APFloat V (Format, APFloat::fcZero, false); 461 APFloat::opStatus status; 462 463 status = V.convertFromString(&floatChars[0],APFloat::rmNearestTiesToEven); 464 465 if (isExact) 466 *isExact = status == APFloat::opOK; 467 468 return V; 469} 470 471void NumericLiteralParser::Diag(SourceLocation Loc, unsigned DiagID, 472 const std::string &M) { 473 PP.Diag(Loc, DiagID, M); 474 hadError = true; 475} 476 477 478CharLiteralParser::CharLiteralParser(const char *begin, const char *end, 479 SourceLocation Loc, Preprocessor &PP) { 480 // At this point we know that the character matches the regex "L?'.*'". 481 HadError = false; 482 Value = 0; 483 484 // Determine if this is a wide character. 485 IsWide = begin[0] == 'L'; 486 if (IsWide) ++begin; 487 488 // Skip over the entry quote. 489 assert(begin[0] == '\'' && "Invalid token lexed"); 490 ++begin; 491 492 // FIXME: This assumes that 'int' is 32-bits in overflow calculation, and the 493 // size of "value". 494 assert(PP.getTargetInfo().getIntWidth() == 32 && 495 "Assumes sizeof(int) == 4 for now"); 496 // FIXME: This assumes that wchar_t is 32-bits for now. 497 assert(PP.getTargetInfo().getWCharWidth() == 32 && 498 "Assumes sizeof(wchar_t) == 4 for now"); 499 // FIXME: This extensively assumes that 'char' is 8-bits. 500 assert(PP.getTargetInfo().getCharWidth() == 8 && 501 "Assumes char is 8 bits"); 502 503 bool isFirstChar = true; 504 bool isMultiChar = false; 505 while (begin[0] != '\'') { 506 unsigned ResultChar; 507 if (begin[0] != '\\') // If this is a normal character, consume it. 508 ResultChar = *begin++; 509 else // Otherwise, this is an escape character. 510 ResultChar = ProcessCharEscape(begin, end, HadError, Loc, IsWide, PP); 511 512 // If this is a multi-character constant (e.g. 'abc'), handle it. These are 513 // implementation defined (C99 6.4.4.4p10). 514 if (!isFirstChar) { 515 // If this is the second character being processed, do special handling. 516 if (!isMultiChar) { 517 isMultiChar = true; 518 519 // Warn about discarding the top bits for multi-char wide-character 520 // constants (L'abcd'). 521 if (IsWide) 522 PP.Diag(Loc, diag::warn_extraneous_wide_char_constant); 523 } 524 525 if (IsWide) { 526 // Emulate GCC's (unintentional?) behavior: L'ab' -> L'b'. 527 Value = 0; 528 } else { 529 // Narrow character literals act as though their value is concatenated 530 // in this implementation. 531 if (((Value << 8) >> 8) != Value) 532 PP.Diag(Loc, diag::warn_char_constant_too_large); 533 Value <<= 8; 534 } 535 } 536 537 Value += ResultChar; 538 isFirstChar = false; 539 } 540 541 // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1") 542 // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple 543 // character constants are not sign extended in the this implementation: 544 // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC. 545 if (!IsWide && !isMultiChar && (Value & 128) && 546 PP.getTargetInfo().isCharSigned()) 547 Value = (signed char)Value; 548} 549 550 551/// string-literal: [C99 6.4.5] 552/// " [s-char-sequence] " 553/// L" [s-char-sequence] " 554/// s-char-sequence: 555/// s-char 556/// s-char-sequence s-char 557/// s-char: 558/// any source character except the double quote ", 559/// backslash \, or newline character 560/// escape-character 561/// universal-character-name 562/// escape-character: [C99 6.4.4.4] 563/// \ escape-code 564/// universal-character-name 565/// escape-code: 566/// character-escape-code 567/// octal-escape-code 568/// hex-escape-code 569/// character-escape-code: one of 570/// n t b r f v a 571/// \ ' " ? 572/// octal-escape-code: 573/// octal-digit 574/// octal-digit octal-digit 575/// octal-digit octal-digit octal-digit 576/// hex-escape-code: 577/// x hex-digit 578/// hex-escape-code hex-digit 579/// universal-character-name: 580/// \u hex-quad 581/// \U hex-quad hex-quad 582/// hex-quad: 583/// hex-digit hex-digit hex-digit hex-digit 584/// 585StringLiteralParser:: 586StringLiteralParser(const Token *StringToks, unsigned NumStringToks, 587 Preprocessor &pp, TargetInfo &t) 588 : PP(pp), Target(t) { 589 // Scan all of the string portions, remember the max individual token length, 590 // computing a bound on the concatenated string length, and see whether any 591 // piece is a wide-string. If any of the string portions is a wide-string 592 // literal, the result is a wide-string literal [C99 6.4.5p4]. 593 MaxTokenLength = StringToks[0].getLength(); 594 SizeBound = StringToks[0].getLength()-2; // -2 for "". 595 AnyWide = StringToks[0].is(tok::wide_string_literal); 596 597 hadError = false; 598 599 // Implement Translation Phase #6: concatenation of string literals 600 /// (C99 5.1.1.2p1). The common case is only one string fragment. 601 for (unsigned i = 1; i != NumStringToks; ++i) { 602 // The string could be shorter than this if it needs cleaning, but this is a 603 // reasonable bound, which is all we need. 604 SizeBound += StringToks[i].getLength()-2; // -2 for "". 605 606 // Remember maximum string piece length. 607 if (StringToks[i].getLength() > MaxTokenLength) 608 MaxTokenLength = StringToks[i].getLength(); 609 610 // Remember if we see any wide strings. 611 AnyWide |= StringToks[i].is(tok::wide_string_literal); 612 } 613 614 615 // Include space for the null terminator. 616 ++SizeBound; 617 618 // TODO: K&R warning: "traditional C rejects string constant concatenation" 619 620 // Get the width in bytes of wchar_t. If no wchar_t strings are used, do not 621 // query the target. As such, wchar_tByteWidth is only valid if AnyWide=true. 622 wchar_tByteWidth = ~0U; 623 if (AnyWide) { 624 wchar_tByteWidth = Target.getWCharWidth(); 625 assert((wchar_tByteWidth & 7) == 0 && "Assumes wchar_t is byte multiple!"); 626 wchar_tByteWidth /= 8; 627 } 628 629 // The output buffer size needs to be large enough to hold wide characters. 630 // This is a worst-case assumption which basically corresponds to L"" "long". 631 if (AnyWide) 632 SizeBound *= wchar_tByteWidth; 633 634 // Size the temporary buffer to hold the result string data. 635 ResultBuf.resize(SizeBound); 636 637 // Likewise, but for each string piece. 638 llvm::SmallString<512> TokenBuf; 639 TokenBuf.resize(MaxTokenLength); 640 641 // Loop over all the strings, getting their spelling, and expanding them to 642 // wide strings as appropriate. 643 ResultPtr = &ResultBuf[0]; // Next byte to fill in. 644 645 Pascal = false; 646 647 for (unsigned i = 0, e = NumStringToks; i != e; ++i) { 648 const char *ThisTokBuf = &TokenBuf[0]; 649 // Get the spelling of the token, which eliminates trigraphs, etc. We know 650 // that ThisTokBuf points to a buffer that is big enough for the whole token 651 // and 'spelled' tokens can only shrink. 652 unsigned ThisTokLen = PP.getSpelling(StringToks[i], ThisTokBuf); 653 const char *ThisTokEnd = ThisTokBuf+ThisTokLen-1; // Skip end quote. 654 655 // TODO: Input character set mapping support. 656 657 // Skip L marker for wide strings. 658 bool ThisIsWide = false; 659 if (ThisTokBuf[0] == 'L') { 660 ++ThisTokBuf; 661 ThisIsWide = true; 662 } 663 664 assert(ThisTokBuf[0] == '"' && "Expected quote, lexer broken?"); 665 ++ThisTokBuf; 666 667 // Check if this is a pascal string 668 if (pp.getLangOptions().PascalStrings && ThisTokBuf + 1 != ThisTokEnd && 669 ThisTokBuf[0] == '\\' && ThisTokBuf[1] == 'p') { 670 671 // If the \p sequence is found in the first token, we have a pascal string 672 // Otherwise, if we already have a pascal string, ignore the first \p 673 if (i == 0) { 674 ++ThisTokBuf; 675 Pascal = true; 676 } else if (Pascal) 677 ThisTokBuf += 2; 678 } 679 680 while (ThisTokBuf != ThisTokEnd) { 681 // Is this a span of non-escape characters? 682 if (ThisTokBuf[0] != '\\') { 683 const char *InStart = ThisTokBuf; 684 do { 685 ++ThisTokBuf; 686 } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\'); 687 688 // Copy the character span over. 689 unsigned Len = ThisTokBuf-InStart; 690 if (!AnyWide) { 691 memcpy(ResultPtr, InStart, Len); 692 ResultPtr += Len; 693 } else { 694 // Note: our internal rep of wide char tokens is always little-endian. 695 for (; Len; --Len, ++InStart) { 696 *ResultPtr++ = InStart[0]; 697 // Add zeros at the end. 698 for (unsigned i = 1, e = wchar_tByteWidth; i != e; ++i) 699 *ResultPtr++ = 0; 700 } 701 } 702 continue; 703 } 704 705 // Otherwise, this is an escape character. Process it. 706 unsigned ResultChar = ProcessCharEscape(ThisTokBuf, ThisTokEnd, hadError, 707 StringToks[i].getLocation(), 708 ThisIsWide, PP); 709 710 // Note: our internal rep of wide char tokens is always little-endian. 711 *ResultPtr++ = ResultChar & 0xFF; 712 713 if (AnyWide) { 714 for (unsigned i = 1, e = wchar_tByteWidth; i != e; ++i) 715 *ResultPtr++ = ResultChar >> i*8; 716 } 717 } 718 } 719 720 // Add zero terminator. 721 *ResultPtr = 0; 722 if (AnyWide) { 723 for (unsigned i = 1, e = wchar_tByteWidth; i != e; ++i) 724 *ResultPtr++ = 0; 725 } 726 727 if (Pascal) 728 ResultBuf[0] = ResultPtr-&ResultBuf[0]-1; 729} 730