LiteralSupport.cpp revision 263509
1193326Sed//===--- LiteralSupport.cpp - Code to parse and process literals ----------===// 2193326Sed// 3193326Sed// The LLVM Compiler Infrastructure 4193326Sed// 5193326Sed// This file is distributed under the University of Illinois Open Source 6193326Sed// License. See LICENSE.TXT for details. 7193326Sed// 8193326Sed//===----------------------------------------------------------------------===// 9193326Sed// 10193326Sed// This file implements the NumericLiteralParser, CharLiteralParser, and 11193326Sed// StringLiteralParser interfaces. 12193326Sed// 13193326Sed//===----------------------------------------------------------------------===// 14193326Sed 15193326Sed#include "clang/Lex/LiteralSupport.h" 16193326Sed#include "clang/Basic/CharInfo.h" 17193326Sed#include "clang/Basic/TargetInfo.h" 18193326Sed#include "clang/Lex/LexDiagnostic.h" 19198092Srdivacky#include "clang/Lex/Preprocessor.h" 20193326Sed#include "llvm/ADT/StringExtras.h" 21193326Sed#include "llvm/Support/ConvertUTF.h" 22193326Sed#include "llvm/Support/ErrorHandling.h" 23193326Sed 24193326Sedusing namespace clang; 25193326Sed 26193326Sedstatic unsigned getCharWidth(tok::TokenKind kind, const TargetInfo &Target) { 27193326Sed switch (kind) { 28193326Sed default: llvm_unreachable("Unknown token type!"); 29193326Sed case tok::char_constant: 30193326Sed case tok::string_literal: 31193326Sed case tok::utf8_string_literal: 32193326Sed return Target.getCharWidth(); 33193326Sed case tok::wide_char_constant: 34193326Sed case tok::wide_string_literal: 35193326Sed return Target.getWCharWidth(); 36218893Sdim case tok::utf16_char_constant: 37218893Sdim case tok::utf16_string_literal: 38193326Sed return Target.getChar16Width(); 39193326Sed case tok::utf32_char_constant: 40193326Sed case tok::utf32_string_literal: 41193326Sed return Target.getChar32Width(); 42193326Sed } 43193326Sed} 44193326Sed 45193326Sedstatic CharSourceRange MakeCharSourceRange(const LangOptions &Features, 46193326Sed FullSourceLoc TokLoc, 47198092Srdivacky const char *TokBegin, 48193326Sed const char *TokRangeBegin, 49193326Sed const char *TokRangeEnd) { 50193326Sed SourceLocation Begin = 51193326Sed Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin, 52193326Sed TokLoc.getManager(), Features); 53193326Sed SourceLocation End = 54193326Sed Lexer::AdvanceToTokenCharacter(Begin, TokRangeEnd - TokRangeBegin, 55193326Sed TokLoc.getManager(), Features); 56193326Sed return CharSourceRange::getCharRange(Begin, End); 57218893Sdim} 58218893Sdim 59193326Sed/// \brief Produce a diagnostic highlighting some portion of a literal. 60193326Sed/// 61194179Sed/// Emits the diagnostic \p DiagID, highlighting the range of characters from 62218893Sdim/// \p TokRangeBegin (inclusive) to \p TokRangeEnd (exclusive), which must be 63218893Sdim/// a substring of a spelling buffer for the token beginning at \p TokBegin. 64194179Sedstatic DiagnosticBuilder Diag(DiagnosticsEngine *Diags, 65194179Sed const LangOptions &Features, FullSourceLoc TokLoc, 66193326Sed const char *TokBegin, const char *TokRangeBegin, 67193326Sed const char *TokRangeEnd, unsigned DiagID) { 68193326Sed SourceLocation Begin = 69193326Sed Lexer::AdvanceToTokenCharacter(TokLoc, TokRangeBegin - TokBegin, 70193326Sed TokLoc.getManager(), Features); 71193326Sed return Diags->Report(Begin, DiagID) << 72193326Sed MakeCharSourceRange(Features, TokLoc, TokBegin, TokRangeBegin, TokRangeEnd); 73193326Sed} 74193326Sed 75193326Sed/// ProcessCharEscape - Parse a standard C escape sequence, which can occur in 76193326Sed/// either a character or a string literal. 77193326Sedstatic unsigned ProcessCharEscape(const char *ThisTokBegin, 78193326Sed const char *&ThisTokBuf, 79193326Sed const char *ThisTokEnd, bool &HadError, 80193326Sed FullSourceLoc Loc, unsigned CharWidth, 81193326Sed DiagnosticsEngine *Diags, 82193326Sed const LangOptions &Features) { 83193326Sed const char *EscapeBegin = ThisTokBuf; 84218893Sdim 85218893Sdim // Skip the '\' char. 86193326Sed ++ThisTokBuf; 87193326Sed 88193326Sed // We know that this character can't be off the end of the buffer, because 89198092Srdivacky // that would have been \", which would not have been the end of string. 90193326Sed unsigned ResultChar = *ThisTokBuf++; 91193326Sed switch (ResultChar) { 92193326Sed // These map to themselves. 93193326Sed case '\\': case '\'': case '"': case '?': break; 94193326Sed 95193326Sed // These have fixed mappings. 96193326Sed case 'a': 97193326Sed // TODO: K&R: the meaning of '\\a' is different in traditional C 98193326Sed ResultChar = 7; 99193326Sed break; 100193326Sed case 'b': 101193326Sed ResultChar = 8; 102218893Sdim break; 103218893Sdim case 'e': 104198092Srdivacky if (Diags) 105193326Sed Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 106193326Sed diag::ext_nonstandard_escape) << "e"; 107193326Sed ResultChar = 27; 108193326Sed break; 109198092Srdivacky case 'E': 110193326Sed if (Diags) 111218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 112218893Sdim diag::ext_nonstandard_escape) << "E"; 113193326Sed ResultChar = 27; 114193326Sed break; 115193326Sed case 'f': 116193326Sed ResultChar = 12; 117193326Sed break; 118193326Sed case 'n': 119193326Sed ResultChar = 10; 120193326Sed break; 121193326Sed case 'r': 122193326Sed ResultChar = 13; 123193326Sed break; 124193326Sed case 't': 125193326Sed ResultChar = 9; 126193326Sed break; 127193326Sed case 'v': 128193326Sed ResultChar = 11; 129193326Sed break; 130198092Srdivacky case 'x': { // Hex escape. 131193326Sed ResultChar = 0; 132218893Sdim if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) { 133218893Sdim if (Diags) 134198092Srdivacky Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 135193326Sed diag::err_hex_escape_no_digits) << "x"; 136218893Sdim HadError = 1; 137218893Sdim break; 138193326Sed } 139193326Sed 140193326Sed // Hex escapes are a maximal series of hex digits. 141193326Sed bool Overflow = false; 142198092Srdivacky for (; ThisTokBuf != ThisTokEnd; ++ThisTokBuf) { 143193326Sed int CharVal = llvm::hexDigitValue(ThisTokBuf[0]); 144193326Sed if (CharVal == -1) break; 145193326Sed // About to shift out a digit? 146218893Sdim Overflow |= (ResultChar & 0xF0000000) ? true : false; 147218893Sdim ResultChar <<= 4; 148208600Srdivacky ResultChar |= CharVal; 149193326Sed } 150193326Sed 151218893Sdim // See if any bits will be truncated when evaluated as a character. 152208600Srdivacky if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) { 153208600Srdivacky Overflow = true; 154218893Sdim ResultChar &= ~0U >> (32-CharWidth); 155218893Sdim } 156218893Sdim 157193326Sed // Check for overflow. 158218893Sdim if (Overflow && Diags) // Too many digits to fit in 159218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 160193326Sed diag::err_hex_escape_too_large); 161193326Sed break; 162198092Srdivacky } 163193326Sed case '0': case '1': case '2': case '3': 164193326Sed case '4': case '5': case '6': case '7': { 165193326Sed // Octal escapes. 166193326Sed --ThisTokBuf; 167218893Sdim ResultChar = 0; 168218893Sdim 169218893Sdim // Octal escapes are a series of octal digits with maximum length 3. 170218893Sdim // "\0123" is a two digit sequence equal to "\012" "3". 171218893Sdim unsigned NumDigits = 0; 172218893Sdim do { 173218893Sdim ResultChar <<= 3; 174198092Srdivacky ResultChar |= *ThisTokBuf++ - '0'; 175193326Sed ++NumDigits; 176193326Sed } while (ThisTokBuf != ThisTokEnd && NumDigits < 3 && 177198092Srdivacky ThisTokBuf[0] >= '0' && ThisTokBuf[0] <= '7'); 178193326Sed 179193326Sed // Check for overflow. Reject '\777', but not L'\777'. 180193326Sed if (CharWidth != 32 && (ResultChar >> CharWidth) != 0) { 181193326Sed if (Diags) 182218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 183218893Sdim diag::err_octal_escape_too_large); 184218893Sdim ResultChar &= ~0U >> (32-CharWidth); 185193326Sed } 186218893Sdim break; 187212904Sdim } 188218893Sdim 189193326Sed // Otherwise, these are not valid escapes. 190193326Sed case '(': case '{': case '[': case '%': 191193326Sed // GCC accepts these as extensions. We warn about them as such though. 192193326Sed if (Diags) 193193326Sed Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 194193326Sed diag::ext_nonstandard_escape) 195218893Sdim << std::string(1, ResultChar); 196218893Sdim break; 197218893Sdim default: 198218893Sdim if (Diags == 0) 199218893Sdim break; 200218893Sdim 201218893Sdim if (isPrintable(ResultChar)) 202218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 203218893Sdim diag::ext_unknown_escape) 204193326Sed << std::string(1, ResultChar); 205198092Srdivacky else 206193326Sed Diag(Diags, Features, Loc, ThisTokBegin, EscapeBegin, ThisTokBuf, 207193326Sed diag::ext_unknown_escape) 208198092Srdivacky << "x" + llvm::utohexstr(ResultChar); 209193326Sed break; 210218893Sdim } 211218893Sdim 212218893Sdim return ResultChar; 213218893Sdim} 214218893Sdim 215218893Sdim/// ProcessUCNEscape - Read the Universal Character Name, check constraints and 216218893Sdim/// return the UTF32. 217218893Sdimstatic bool ProcessUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, 218218893Sdim const char *ThisTokEnd, 219218893Sdim uint32_t &UcnVal, unsigned short &UcnLen, 220218893Sdim FullSourceLoc Loc, DiagnosticsEngine *Diags, 221218893Sdim const LangOptions &Features, 222218893Sdim bool in_char_string_literal = false) { 223218893Sdim const char *UcnBegin = ThisTokBuf; 224218893Sdim 225218893Sdim // Skip the '\u' char's. 226218893Sdim ThisTokBuf += 2; 227218893Sdim 228218893Sdim if (ThisTokBuf == ThisTokEnd || !isHexDigit(*ThisTokBuf)) { 229218893Sdim if (Diags) 230193326Sed Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 231193326Sed diag::err_hex_escape_no_digits) << StringRef(&ThisTokBuf[-1], 1); 232193326Sed return false; 233218893Sdim } 234212904Sdim UcnLen = (ThisTokBuf[-1] == 'u' ? 4 : 8); 235218893Sdim unsigned short UcnLenSave = UcnLen; 236218893Sdim for (; ThisTokBuf != ThisTokEnd && UcnLenSave; ++ThisTokBuf, UcnLenSave--) { 237218893Sdim int CharVal = llvm::hexDigitValue(ThisTokBuf[0]); 238218893Sdim if (CharVal == -1) break; 239218893Sdim UcnVal <<= 4; 240218893Sdim UcnVal |= CharVal; 241218893Sdim } 242218893Sdim // If we didn't consume the proper number of digits, there is a problem. 243218893Sdim if (UcnLenSave) { 244218893Sdim if (Diags) 245218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 246218893Sdim diag::err_ucn_escape_incomplete); 247218893Sdim return false; 248218893Sdim } 249218893Sdim 250218893Sdim // Check UCN constraints (C99 6.4.3p2) [C++11 lex.charset p2] 251218893Sdim if ((0xD800 <= UcnVal && UcnVal <= 0xDFFF) || // surrogate codepoints 252218893Sdim UcnVal > 0x10FFFF) { // maximum legal UTF32 value 253218893Sdim if (Diags) 254218893Sdim Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 255218893Sdim diag::err_ucn_escape_invalid); 256218893Sdim return false; 257218893Sdim } 258218893Sdim 259218893Sdim // C++11 allows UCNs that refer to control characters and basic source 260218893Sdim // characters inside character and string literals 261218893Sdim if (UcnVal < 0xa0 && 262218893Sdim (UcnVal != 0x24 && UcnVal != 0x40 && UcnVal != 0x60)) { // $, @, ` 263212904Sdim bool IsError = (!Features.CPlusPlus11 || !in_char_string_literal); 264212904Sdim if (Diags) { 265193326Sed char BasicSCSChar = UcnVal; 266193326Sed if (UcnVal >= 0x20 && UcnVal < 0x7f) 267193326Sed Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 268198092Srdivacky IsError ? diag::err_ucn_escape_basic_scs : 269193326Sed diag::warn_cxx98_compat_literal_ucn_escape_basic_scs) 270193326Sed << StringRef(&BasicSCSChar, 1); 271193326Sed else 272193326Sed Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 273193326Sed IsError ? diag::err_ucn_control_character : 274193326Sed diag::warn_cxx98_compat_literal_ucn_control_character); 275193326Sed } 276193326Sed if (IsError) 277193326Sed return false; 278193326Sed } 279193326Sed 280198092Srdivacky if (!Features.CPlusPlus && !Features.C99 && Diags) 281193326Sed Diag(Diags, Features, Loc, ThisTokBegin, UcnBegin, ThisTokBuf, 282193326Sed diag::warn_ucn_not_valid_in_c89_literal); 283198092Srdivacky 284193326Sed return true; 285193326Sed} 286198092Srdivacky 287193326Sed/// MeasureUCNEscape - Determine the number of bytes within the resulting string 288193326Sed/// which this UCN will occupy. 289193326Sedstatic int MeasureUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, 290193326Sed const char *ThisTokEnd, unsigned CharByteWidth, 291193326Sed const LangOptions &Features, bool &HadError) { 292193326Sed // UTF-32: 4 bytes per escape. 293193326Sed if (CharByteWidth == 4) 294193326Sed return 4; 295193326Sed 296193326Sed uint32_t UcnVal = 0; 297193326Sed unsigned short UcnLen = 0; 298193326Sed FullSourceLoc Loc; 299193326Sed 300193326Sed if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, 301193326Sed UcnLen, Loc, 0, Features, true)) { 302193326Sed HadError = true; 303193326Sed return 0; 304193326Sed } 305193326Sed 306198092Srdivacky // UTF-16: 2 bytes for BMP, 4 bytes otherwise. 307193326Sed if (CharByteWidth == 2) 308193326Sed return UcnVal <= 0xFFFF ? 2 : 4; 309198092Srdivacky 310193326Sed // UTF-8. 311193326Sed if (UcnVal < 0x80) 312198092Srdivacky return 1; 313193326Sed if (UcnVal < 0x800) 314193326Sed return 2; 315193326Sed if (UcnVal < 0x10000) 316193326Sed return 3; 317193326Sed return 4; 318193326Sed} 319193326Sed 320193326Sed/// EncodeUCNEscape - Read the Universal Character Name, check constraints and 321193326Sed/// convert the UTF32 to UTF8 or UTF16. This is a subroutine of 322193326Sed/// StringLiteralParser. When we decide to implement UCN's for identifiers, 323193326Sed/// we will likely rework our support for UCN's. 324193326Sedstatic void EncodeUCNEscape(const char *ThisTokBegin, const char *&ThisTokBuf, 325193326Sed const char *ThisTokEnd, 326193326Sed char *&ResultBuf, bool &HadError, 327193326Sed FullSourceLoc Loc, unsigned CharByteWidth, 328193326Sed DiagnosticsEngine *Diags, 329193326Sed const LangOptions &Features) { 330193326Sed typedef uint32_t UTF32; 331193326Sed UTF32 UcnVal = 0; 332193326Sed unsigned short UcnLen = 0; 333193326Sed if (!ProcessUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, UcnVal, UcnLen, 334198092Srdivacky Loc, Diags, Features, true)) { 335193326Sed HadError = true; 336193326Sed return; 337193326Sed } 338193326Sed 339193326Sed assert((CharByteWidth == 1 || CharByteWidth == 2 || CharByteWidth == 4) && 340193326Sed "only character widths of 1, 2, or 4 bytes supported"); 341193326Sed 342193326Sed (void)UcnLen; 343193326Sed assert((UcnLen== 4 || UcnLen== 8) && "only ucn length of 4 or 8 supported"); 344198092Srdivacky 345193326Sed if (CharByteWidth == 4) { 346193326Sed // FIXME: Make the type of the result buffer correct instead of 347193326Sed // using reinterpret_cast. 348193326Sed UTF32 *ResultPtr = reinterpret_cast<UTF32*>(ResultBuf); 349193326Sed *ResultPtr = UcnVal; 350193326Sed ResultBuf += 4; 351198092Srdivacky return; 352193326Sed } 353193326Sed 354193326Sed if (CharByteWidth == 2) { 355193326Sed // FIXME: Make the type of the result buffer correct instead of 356193326Sed // using reinterpret_cast. 357193326Sed UTF16 *ResultPtr = reinterpret_cast<UTF16*>(ResultBuf); 358193326Sed 359193326Sed if (UcnVal <= (UTF32)0xFFFF) { 360198092Srdivacky *ResultPtr = UcnVal; 361193326Sed ResultBuf += 2; 362198092Srdivacky return; 363193326Sed } 364193326Sed 365193326Sed // Convert to UTF16. 366193326Sed UcnVal -= 0x10000; 367193326Sed *ResultPtr = 0xD800 + (UcnVal >> 10); 368193326Sed *(ResultPtr+1) = 0xDC00 + (UcnVal & 0x3FF); 369193326Sed ResultBuf += 4; 370193326Sed return; 371193326Sed } 372193326Sed 373193326Sed assert(CharByteWidth == 1 && "UTF-8 encoding is only for 1 byte characters"); 374212904Sdim 375193326Sed // Now that we've parsed/checked the UCN, we convert from UTF32->UTF8. 376193326Sed // The conversion below was inspired by: 377193326Sed // http://www.unicode.org/Public/PROGRAMS/CVTUTF/ConvertUTF.c 378193326Sed // First, we determine how many bytes the result will require. 379193326Sed typedef uint8_t UTF8; 380193326Sed 381198092Srdivacky unsigned short bytesToWrite = 0; 382193326Sed if (UcnVal < (UTF32)0x80) 383193326Sed bytesToWrite = 1; 384193326Sed else if (UcnVal < (UTF32)0x800) 385193326Sed bytesToWrite = 2; 386193326Sed else if (UcnVal < (UTF32)0x10000) 387193326Sed bytesToWrite = 3; 388193326Sed else 389193326Sed bytesToWrite = 4; 390193326Sed 391193326Sed const unsigned byteMask = 0xBF; 392193326Sed const unsigned byteMark = 0x80; 393193326Sed 394193326Sed // Once the bits are split out into bytes of UTF8, this is a mask OR-ed 395193326Sed // into the first byte, depending on how many bytes follow. 396193326Sed static const UTF8 firstByteMark[5] = { 397193326Sed 0x00, 0x00, 0xC0, 0xE0, 0xF0 398193326Sed }; 399193326Sed // Finally, we write the bytes into ResultBuf. 400198092Srdivacky ResultBuf += bytesToWrite; 401193326Sed switch (bytesToWrite) { // note: everything falls through. 402193326Sed case 4: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6; 403193326Sed case 3: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6; 404198092Srdivacky case 2: *--ResultBuf = (UTF8)((UcnVal | byteMark) & byteMask); UcnVal >>= 6; 405193326Sed case 1: *--ResultBuf = (UTF8) (UcnVal | firstByteMark[bytesToWrite]); 406193326Sed } 407193326Sed // Update the buffer. 408193326Sed ResultBuf += bytesToWrite; 409193326Sed} 410193326Sed 411193326Sed 412193326Sed/// integer-constant: [C99 6.4.4.1] 413193326Sed/// decimal-constant integer-suffix 414193326Sed/// octal-constant integer-suffix 415193326Sed/// hexadecimal-constant integer-suffix 416193326Sed/// binary-literal integer-suffix [GNU, C++1y] 417193326Sed/// user-defined-integer-literal: [C++11 lex.ext] 418193326Sed/// decimal-literal ud-suffix 419193326Sed/// octal-literal ud-suffix 420193326Sed/// hexadecimal-literal ud-suffix 421193326Sed/// binary-literal ud-suffix [GNU, C++1y] 422193326Sed/// decimal-constant: 423193326Sed/// nonzero-digit 424193326Sed/// decimal-constant digit 425198092Srdivacky/// octal-constant: 426193326Sed/// 0 427193326Sed/// octal-constant octal-digit 428193326Sed/// hexadecimal-constant: 429193326Sed/// hexadecimal-prefix hexadecimal-digit 430193326Sed/// hexadecimal-constant hexadecimal-digit 431193326Sed/// hexadecimal-prefix: one of 432193326Sed/// 0x 0X 433193326Sed/// binary-literal: 434193326Sed/// 0b binary-digit 435193326Sed/// 0B binary-digit 436218893Sdim/// binary-literal binary-digit 437193326Sed/// integer-suffix: 438202879Srdivacky/// unsigned-suffix [long-suffix] 439199990Srdivacky/// unsigned-suffix [long-long-suffix] 440193326Sed/// long-suffix [unsigned-suffix] 441198092Srdivacky/// long-long-suffix [unsigned-sufix] 442198092Srdivacky/// nonzero-digit: 443198092Srdivacky/// 1 2 3 4 5 6 7 8 9 444198092Srdivacky/// octal-digit: 445198092Srdivacky/// 0 1 2 3 4 5 6 7 446199990Srdivacky/// hexadecimal-digit: 447198092Srdivacky/// 0 1 2 3 4 5 6 7 8 9 448199990Srdivacky/// a b c d e f 449218893Sdim/// A B C D E F 450218893Sdim/// binary-digit: 451218893Sdim/// 0 452218893Sdim/// 1 453199990Srdivacky/// unsigned-suffix: one of 454199990Srdivacky/// u U 455218893Sdim/// long-suffix: one of 456218893Sdim/// l L 457218893Sdim/// long-long-suffix: one of 458218893Sdim/// ll LL 459198092Srdivacky/// 460199990Srdivacky/// floating-constant: [C99 6.4.4.2] 461198092Srdivacky/// TODO: add rules... 462199990Srdivacky/// 463218893SdimNumericLiteralParser::NumericLiteralParser(StringRef TokSpelling, 464218893Sdim SourceLocation TokLoc, 465218893Sdim Preprocessor &PP) 466218893Sdim : PP(PP), ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) { 467218893Sdim 468199990Srdivacky // This routine assumes that the range begin/end matches the regex for integer 469198092Srdivacky // and FP constants (specifically, the 'pp-number' regex), and assumes that 470199990Srdivacky // the byte at "*end" is both valid and not part of the regex. Because of 471218893Sdim // this, it doesn't have to check for 'overscan' in various places. 472218893Sdim assert(!isPreprocessingNumberBody(*ThisTokEnd) && "didn't maximally munch?"); 473218893Sdim 474218893Sdim s = DigitsBegin = ThisTokBegin; 475218893Sdim saw_exponent = false; 476199990Srdivacky saw_period = false; 477198092Srdivacky saw_ud_suffix = false; 478198092Srdivacky isLong = false; 479198092Srdivacky isUnsigned = false; 480198092Srdivacky isLongLong = false; 481193326Sed isFloat = false; 482193326Sed isImaginary = false; 483193326Sed isMicrosoftInteger = false; 484193326Sed hadError = false; 485193326Sed 486193326Sed if (*s == '0') { // parse radix 487193326Sed ParseNumberStartingWithZero(TokLoc); 488193326Sed if (hadError) 489193326Sed return; 490193326Sed } else { // the first digit is non-zero 491193326Sed radix = 10; 492193326Sed s = SkipDigits(s); 493193326Sed if (s == ThisTokEnd) { 494193326Sed // Done. 495198092Srdivacky } else if (isHexDigit(*s) && !(*s == 'e' || *s == 'E')) { 496193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin), 497193326Sed diag::err_invalid_decimal_digit) << StringRef(s, 1); 498193326Sed hadError = true; 499193326Sed return; 500193326Sed } else if (*s == '.') { 501212904Sdim checkSeparator(TokLoc, s, CSK_AfterDigits); 502193326Sed s++; 503193326Sed saw_period = true; 504193326Sed checkSeparator(TokLoc, s, CSK_BeforeDigits); 505193326Sed s = SkipDigits(s); 506193326Sed } 507193326Sed if ((*s == 'e' || *s == 'E')) { // exponent 508193326Sed checkSeparator(TokLoc, s, CSK_AfterDigits); 509193326Sed const char *Exponent = s; 510198092Srdivacky s++; 511193326Sed saw_exponent = true; 512193326Sed if (*s == '+' || *s == '-') s++; // sign 513193326Sed checkSeparator(TokLoc, s, CSK_BeforeDigits); 514193326Sed const char *first_non_digit = SkipDigits(s); 515198092Srdivacky if (first_non_digit != s) { 516193326Sed s = first_non_digit; 517193326Sed } else { 518193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent - ThisTokBegin), 519193326Sed diag::err_exponent_has_no_digits); 520193326Sed hadError = true; 521193326Sed return; 522193326Sed } 523193326Sed } 524193326Sed } 525193326Sed 526193326Sed SuffixBegin = s; 527193326Sed checkSeparator(TokLoc, s, CSK_AfterDigits); 528193326Sed 529193326Sed // Parse the suffix. At this point we can classify whether we have an FP or 530198092Srdivacky // integer constant. 531202379Srdivacky bool isFPConstant = isFloatingLiteral(); 532193326Sed const char *ImaginarySuffixLoc = 0; 533193326Sed 534193326Sed // Loop over all of the characters of the suffix. If we see something bad, 535193326Sed // we break out of the loop. 536193326Sed for (; s != ThisTokEnd; ++s) { 537193326Sed switch (*s) { 538193326Sed case 'f': // FP Suffix for "float" 539193326Sed case 'F': 540193326Sed if (!isFPConstant) break; // Error for integer constant. 541193326Sed if (isFloat || isLong) break; // FF, LF invalid. 542193326Sed isFloat = true; 543193326Sed continue; // Success. 544198092Srdivacky case 'u': 545202379Srdivacky case 'U': 546202379Srdivacky if (isFPConstant) break; // Error for floating constant. 547202379Srdivacky if (isUnsigned) break; // Cannot be repeated. 548202379Srdivacky isUnsigned = true; 549202379Srdivacky continue; // Success. 550202379Srdivacky case 'l': 551193326Sed case 'L': 552193326Sed if (isLong || isLongLong) break; // Cannot be repeated. 553193326Sed if (isFloat) break; // LF invalid. 554193326Sed 555193326Sed // Check for long long. The L's need to be adjacent and the same case. 556193326Sed if (s+1 != ThisTokEnd && s[1] == s[0]) { 557193326Sed if (isFPConstant) break; // long long invalid for floats. 558193326Sed isLongLong = true; 559198092Srdivacky ++s; // Eat both of them. 560193326Sed } else { 561193326Sed isLong = true; 562193326Sed } 563193326Sed continue; // Success. 564193326Sed case 'i': 565193326Sed case 'I': 566193326Sed if (PP.getLangOpts().MicrosoftExt) { 567193326Sed if (isFPConstant || isLong || isLongLong) break; 568193326Sed 569193326Sed // Allow i8, i16, i32, i64, and i128. 570193326Sed if (s + 1 != ThisTokEnd) { 571193326Sed switch (s[1]) { 572212904Sdim case '8': 573193326Sed s += 2; // i8 suffix 574193326Sed isMicrosoftInteger = true; 575193326Sed break; 576193326Sed case '1': 577193326Sed if (s + 2 == ThisTokEnd) break; 578198092Srdivacky if (s[2] == '6') { 579193326Sed s += 3; // i16 suffix 580193326Sed isMicrosoftInteger = true; 581198092Srdivacky } 582193326Sed else if (s[2] == '2') { 583193326Sed if (s + 3 == ThisTokEnd) break; 584193326Sed if (s[3] == '8') { 585193326Sed s += 4; // i128 suffix 586193326Sed isMicrosoftInteger = true; 587198092Srdivacky } 588193326Sed } 589193326Sed break; 590193326Sed case '3': 591193326Sed if (s + 2 == ThisTokEnd) break; 592193326Sed if (s[2] == '2') { 593193326Sed s += 3; // i32 suffix 594193326Sed isLong = true; 595193326Sed isMicrosoftInteger = true; 596193326Sed } 597198092Srdivacky break; 598193326Sed case '6': 599193326Sed if (s + 2 == ThisTokEnd) break; 600193326Sed if (s[2] == '4') { 601193326Sed s += 3; // i64 suffix 602212904Sdim isLongLong = true; 603193326Sed isMicrosoftInteger = true; 604193326Sed } 605193326Sed break; 606198092Srdivacky default: 607193326Sed break; 608193326Sed } 609193326Sed break; 610193326Sed } 611193326Sed } 612193326Sed // "i", "if", and "il" are user-defined suffixes in C++1y. 613193326Sed if (PP.getLangOpts().CPlusPlus1y && *s == 'i') 614193326Sed break; 615193326Sed // fall through. 616193326Sed case 'j': 617193326Sed case 'J': 618193326Sed if (isImaginary) break; // Cannot be repeated. 619193326Sed isImaginary = true; 620193326Sed ImaginarySuffixLoc = s; 621193326Sed continue; // Success. 622193326Sed } 623198092Srdivacky // If we reached here, there was an error or a ud-suffix. 624193326Sed break; 625193326Sed } 626193326Sed 627193326Sed if (s != ThisTokEnd) { 628193326Sed if (isValidUDSuffix(PP.getLangOpts(), 629193326Sed StringRef(SuffixBegin, ThisTokEnd - SuffixBegin))) { 630193326Sed // Any suffix pieces we might have parsed are actually part of the 631193326Sed // ud-suffix. 632193326Sed isLong = false; 633193326Sed isUnsigned = false; 634193326Sed isLongLong = false; 635193326Sed isFloat = false; 636193326Sed isImaginary = false; 637193326Sed isMicrosoftInteger = false; 638193326Sed 639193326Sed saw_ud_suffix = true; 640193326Sed return; 641193326Sed } 642193326Sed 643198092Srdivacky // Report an error if there are any. 644193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, SuffixBegin - ThisTokBegin), 645193326Sed isFPConstant ? diag::err_invalid_suffix_float_constant : 646193326Sed diag::err_invalid_suffix_integer_constant) 647193326Sed << StringRef(SuffixBegin, ThisTokEnd-SuffixBegin); 648193326Sed hadError = true; 649193326Sed return; 650193326Sed } 651193326Sed 652193326Sed if (isImaginary) { 653193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, 654193326Sed ImaginarySuffixLoc - ThisTokBegin), 655193326Sed diag::ext_imaginary_constant); 656193326Sed } 657193326Sed} 658193326Sed 659193326Sed/// Determine whether a suffix is a valid ud-suffix. We avoid treating reserved 660193326Sed/// suffixes as ud-suffixes, because the diagnostic experience is better if we 661193326Sed/// treat it as an invalid suffix. 662198092Srdivackybool NumericLiteralParser::isValidUDSuffix(const LangOptions &LangOpts, 663193326Sed StringRef Suffix) { 664193326Sed if (!LangOpts.CPlusPlus11 || Suffix.empty()) 665193326Sed return false; 666198092Srdivacky 667193326Sed // By C++11 [lex.ext]p10, ud-suffixes starting with an '_' are always valid. 668193326Sed if (Suffix[0] == '_') 669198092Srdivacky return true; 670193326Sed 671198092Srdivacky // In C++11, there are no library suffixes. 672193326Sed if (!LangOpts.CPlusPlus1y) 673193326Sed return false; 674193326Sed 675193326Sed // In C++1y, "s", "h", "min", "ms", "us", and "ns" are used in the library. 676193326Sed // Per tweaked N3660, "il", "i", and "if" are also used in the library. 677193326Sed return llvm::StringSwitch<bool>(Suffix) 678193326Sed .Cases("h", "min", "s", true) 679193326Sed .Cases("ms", "us", "ns", true) 680193326Sed .Cases("il", "i", "if", true) 681193326Sed .Default(false); 682193326Sed} 683193326Sed 684193326Sedvoid NumericLiteralParser::checkSeparator(SourceLocation TokLoc, 685193326Sed const char *Pos, 686193326Sed CheckSeparatorKind IsAfterDigits) { 687193326Sed if (IsAfterDigits == CSK_AfterDigits) { 688201361Srdivacky if (Pos == ThisTokBegin) 689201361Srdivacky return; 690193326Sed --Pos; 691198092Srdivacky } else if (Pos == ThisTokEnd) 692198092Srdivacky return; 693198092Srdivacky 694201361Srdivacky if (isDigitSeparator(*Pos)) 695201361Srdivacky PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Pos - ThisTokBegin), 696193326Sed diag::err_digit_separator_not_between_digits) 697193326Sed << IsAfterDigits; 698193326Sed} 699193326Sed 700193326Sed/// ParseNumberStartingWithZero - This method is called when the first character 701193326Sed/// of the number is found to be a zero. This means it is either an octal 702193326Sed/// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or 703198092Srdivacky/// a floating point number (01239.123e4). Eat the prefix, determining the 704193326Sed/// radix etc. 705193326Sedvoid NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) { 706193326Sed assert(s[0] == '0' && "Invalid method call"); 707198092Srdivacky s++; 708193326Sed 709193326Sed int c1 = s[0]; 710193326Sed int c2 = s[1]; 711193326Sed 712198092Srdivacky // Handle a hex number like 0x1234. 713221345Sdim if ((c1 == 'x' || c1 == 'X') && (isHexDigit(c2) || c2 == '.')) { 714193326Sed s++; 715193326Sed radix = 16; 716193326Sed DigitsBegin = s; 717193326Sed s = SkipHexDigits(s); 718193326Sed bool noSignificand = (s == DigitsBegin); 719193326Sed if (s == ThisTokEnd) { 720193326Sed // Done. 721193326Sed } else if (*s == '.') { 722193326Sed s++; 723198092Srdivacky saw_period = true; 724193326Sed const char *floatDigitsBegin = s; 725198092Srdivacky s = SkipHexDigits(s); 726193326Sed noSignificand &= (floatDigitsBegin == s); 727207619Srdivacky } 728193326Sed 729193326Sed if (noSignificand) { 730218893Sdim PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin), 731218893Sdim diag::err_hexconstant_requires_digits); 732193326Sed hadError = true; 733193326Sed return; 734218893Sdim } 735218893Sdim 736218893Sdim // A binary exponent can appear with or with a '.'. If dotted, the 737218893Sdim // binary exponent is required. 738218893Sdim if (*s == 'p' || *s == 'P') { 739218893Sdim const char *Exponent = s; 740218893Sdim s++; 741218893Sdim saw_exponent = true; 742218893Sdim if (*s == '+' || *s == '-') s++; // sign 743218893Sdim const char *first_non_digit = SkipDigits(s); 744218893Sdim if (first_non_digit == s) { 745218893Sdim PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin), 746218893Sdim diag::err_exponent_has_no_digits); 747218893Sdim hadError = true; 748218893Sdim return; 749218893Sdim } 750218893Sdim s = first_non_digit; 751218893Sdim 752218893Sdim if (!PP.getLangOpts().HexFloats) 753193326Sed PP.Diag(TokLoc, diag::ext_hexconstant_invalid); 754193326Sed } else if (saw_period) { 755193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), 756193326Sed diag::err_hexconstant_requires_exponent); 757193326Sed hadError = true; 758193326Sed } 759193326Sed return; 760193326Sed } 761193326Sed 762193326Sed // Handle simple binary numbers 0b01010 763207619Srdivacky if ((c1 == 'b' || c1 == 'B') && (c2 == '0' || c2 == '1')) { 764193326Sed // 0b101010 is a C++1y / GCC extension. 765207619Srdivacky PP.Diag(TokLoc, 766207619Srdivacky PP.getLangOpts().CPlusPlus1y 767193326Sed ? diag::warn_cxx11_compat_binary_literal 768193326Sed : PP.getLangOpts().CPlusPlus 769193326Sed ? diag::ext_binary_literal_cxx1y 770198092Srdivacky : diag::ext_binary_literal); 771193326Sed ++s; 772193326Sed radix = 2; 773193326Sed DigitsBegin = s; 774193326Sed s = SkipBinaryDigits(s); 775193326Sed if (s == ThisTokEnd) { 776193326Sed // Done. 777193326Sed } else if (isHexDigit(*s)) { 778193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), 779193326Sed diag::err_invalid_binary_digit) << StringRef(s, 1); 780193326Sed hadError = true; 781193326Sed } 782193326Sed // Other suffixes will be diagnosed by the caller. 783193326Sed return; 784193326Sed } 785193326Sed 786198092Srdivacky // For now, the radix is set to 8. If we discover that we have a 787198092Srdivacky // floating point constant, the radix will change to 10. Octal floating 788193326Sed // point constants are not permitted (only decimal and hexadecimal). 789193326Sed radix = 8; 790193326Sed DigitsBegin = s; 791198092Srdivacky s = SkipOctalDigits(s); 792218893Sdim if (s == ThisTokEnd) 793218893Sdim return; // Done, simple octal number like 01234 794218893Sdim 795193326Sed // If we have some other non-octal digit that *is* a decimal digit, see if 796193326Sed // this is part of a floating point number like 094.123 or 09e1. 797193326Sed if (isDigit(*s)) { 798193326Sed const char *EndDecimal = SkipDigits(s); 799193326Sed if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') { 800193576Sed s = EndDecimal; 801193326Sed radix = 10; 802193326Sed } 803193326Sed } 804193326Sed 805193326Sed // If we have a hex digit other than 'e' (which denotes a FP exponent) then 806193326Sed // the code is using an incorrect base. 807193326Sed if (isHexDigit(*s) && *s != 'e' && *s != 'E') { 808193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), 809193326Sed diag::err_invalid_octal_digit) << StringRef(s, 1); 810193326Sed hadError = true; 811193326Sed return; 812193326Sed } 813193326Sed 814193326Sed if (*s == '.') { 815193326Sed s++; 816193326Sed radix = 10; 817193326Sed saw_period = true; 818193326Sed s = SkipDigits(s); // Skip suffix. 819193326Sed } 820193326Sed if (*s == 'e' || *s == 'E') { // exponent 821193326Sed const char *Exponent = s; 822193326Sed s++; 823193326Sed radix = 10; 824193326Sed saw_exponent = true; 825193326Sed if (*s == '+' || *s == '-') s++; // sign 826193326Sed const char *first_non_digit = SkipDigits(s); 827193326Sed if (first_non_digit != s) { 828193326Sed s = first_non_digit; 829193326Sed } else { 830193326Sed PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin), 831193326Sed diag::err_exponent_has_no_digits); 832193326Sed hadError = true; 833193326Sed return; 834193326Sed } 835193326Sed } 836193326Sed} 837193326Sed 838193326Sedstatic bool alwaysFitsInto64Bits(unsigned Radix, unsigned NumDigits) { 839193326Sed switch (Radix) { 840193326Sed case 2: 841218893Sdim return NumDigits <= 64; 842218893Sdim case 8: 843218893Sdim return NumDigits <= 64 / 3; // Digits are groups of 3 bits. 844218893Sdim case 10: 845218893Sdim return NumDigits <= 19; // floor(log10(2^64)) 846218893Sdim case 16: 847218893Sdim return NumDigits <= 64 / 4; // Digits are groups of 4 bits. 848193326Sed default: 849193326Sed llvm_unreachable("impossible Radix"); 850193326Sed } 851193326Sed} 852193326Sed 853193326Sed/// GetIntegerValue - Convert this numeric literal value to an APInt that 854193326Sed/// matches Val's input width. If there is an overflow, set Val to the low bits 855198092Srdivacky/// of the result and return true. Otherwise, return false. 856193326Sedbool NumericLiteralParser::GetIntegerValue(llvm::APInt &Val) { 857193326Sed // Fast path: Compute a conservative bound on the maximum number of 858193326Sed // bits per digit in this radix. If we can't possibly overflow a 859193326Sed // uint64 based on that bound then do the simple conversion to 860193326Sed // integer. This avoids the expensive overflow checking below, and 861193326Sed // handles the common cases that matter (small decimal integers and 862193326Sed // hex/octal values which don't overflow). 863193326Sed const unsigned NumDigits = SuffixBegin - DigitsBegin; 864198092Srdivacky if (alwaysFitsInto64Bits(radix, NumDigits)) { 865193326Sed uint64_t N = 0; 866198092Srdivacky for (const char *Ptr = DigitsBegin; Ptr != SuffixBegin; ++Ptr) 867193326Sed if (!isDigitSeparator(*Ptr)) 868198092Srdivacky N = N * radix + llvm::hexDigitValue(*Ptr); 869193326Sed 870193326Sed // This will truncate the value to Val's input width. Simply check 871193326Sed // for overflow by comparing. 872193326Sed Val = N; 873193326Sed return Val.getZExtValue() != N; 874193326Sed } 875198092Srdivacky 876193326Sed Val = 0; 877198092Srdivacky const char *Ptr = DigitsBegin; 878193326Sed 879193326Sed llvm::APInt RadixVal(Val.getBitWidth(), radix); 880193326Sed llvm::APInt CharVal(Val.getBitWidth(), 0); 881193326Sed llvm::APInt OldVal = Val; 882218893Sdim 883193326Sed bool OverflowOccurred = false; 884193326Sed while (Ptr < SuffixBegin) { 885193326Sed if (isDigitSeparator(*Ptr)) { 886198092Srdivacky ++Ptr; 887193326Sed continue; 888193326Sed } 889193326Sed 890193326Sed unsigned C = llvm::hexDigitValue(*Ptr++); 891198092Srdivacky 892193326Sed // If this letter is out of bound for this radix, reject it. 893193326Sed assert(C < radix && "NumericLiteralParser ctor should have rejected this"); 894198092Srdivacky 895193326Sed CharVal = C; 896193326Sed 897193326Sed // Add the digit to the value in the appropriate radix. If adding in digits 898198092Srdivacky // made the value smaller, then this overflowed. 899193326Sed OldVal = Val; 900193326Sed 901193326Sed // Multiply by radix, did overflow occur on the multiply? 902198092Srdivacky Val *= RadixVal; 903193326Sed OverflowOccurred |= Val.udiv(RadixVal) != OldVal; 904198092Srdivacky 905193326Sed // Add value, did overflow occur on the value? 906193326Sed // (a + b) ult b <=> overflow 907193326Sed Val += CharVal; 908193326Sed OverflowOccurred |= Val.ult(CharVal); 909193326Sed } 910205219Srdivacky return OverflowOccurred; 911218893Sdim} 912218893Sdim 913218893Sdimllvm::APFloat::opStatus 914205219SrdivackyNumericLiteralParser::GetFloatValue(llvm::APFloat &Result) { 915205219Srdivacky using llvm::APFloat; 916205219Srdivacky 917205219Srdivacky unsigned n = std::min(SuffixBegin - ThisTokBegin, ThisTokEnd - ThisTokBegin); 918205219Srdivacky 919193326Sed llvm::SmallString<16> Buffer; 920212904Sdim StringRef Str(ThisTokBegin, n); 921193326Sed if (Str.find('\'') != StringRef::npos) { 922198092Srdivacky Buffer.reserve(n); 923193326Sed std::remove_copy_if(Str.begin(), Str.end(), std::back_inserter(Buffer), 924212904Sdim &isDigitSeparator); 925212904Sdim Str = Buffer; 926193326Sed } 927212904Sdim 928198092Srdivacky return Result.convertFromString(Str, APFloat::rmNearestTiesToEven); 929193326Sed} 930193326Sed 931198092Srdivacky 932193326Sed/// \verbatim 933218893Sdim/// user-defined-character-literal: [C++11 lex.ext] 934193326Sed/// character-literal ud-suffix 935198092Srdivacky/// ud-suffix: 936193326Sed/// identifier 937193326Sed/// character-literal: [C++11 lex.ccon] 938193326Sed/// ' c-char-sequence ' 939193326Sed/// u' c-char-sequence ' 940193326Sed/// U' c-char-sequence ' 941193326Sed/// L' c-char-sequence ' 942193326Sed/// c-char-sequence: 943193326Sed/// c-char 944198092Srdivacky/// c-char-sequence c-char 945193326Sed/// c-char: 946193326Sed/// any member of the source character set except the single-quote ', 947193326Sed/// backslash \, or new-line character 948193326Sed/// escape-sequence 949193326Sed/// universal-character-name 950193326Sed/// escape-sequence: 951193326Sed/// simple-escape-sequence 952198092Srdivacky/// octal-escape-sequence 953193326Sed/// hexadecimal-escape-sequence 954193326Sed/// simple-escape-sequence: 955193326Sed/// one of \' \" \? \\ \a \b \f \n \r \t \v 956193326Sed/// octal-escape-sequence: 957193326Sed/// \ octal-digit 958193326Sed/// \ octal-digit octal-digit 959193326Sed/// \ octal-digit octal-digit octal-digit 960193326Sed/// hexadecimal-escape-sequence: 961193326Sed/// \x hexadecimal-digit 962193326Sed/// hexadecimal-escape-sequence hexadecimal-digit 963193326Sed/// universal-character-name: [C++11 lex.charset] 964193326Sed/// \u hex-quad 965193326Sed/// \U hex-quad hex-quad 966193326Sed/// hex-quad: 967193326Sed/// hex-digit hex-digit hex-digit hex-digit 968193326Sed/// \endverbatim 969193326Sed/// 970193326SedCharLiteralParser::CharLiteralParser(const char *begin, const char *end, 971218893Sdim SourceLocation Loc, Preprocessor &PP, 972218893Sdim tok::TokenKind kind) { 973218893Sdim // At this point we know that the character matches the regex "(L|u|U)?'.*'". 974193326Sed HadError = false; 975193326Sed 976193326Sed Kind = kind; 977218893Sdim 978218893Sdim const char *TokBegin = begin; 979218893Sdim 980218893Sdim // Skip over wide character determinant. 981198092Srdivacky if (Kind != tok::char_constant) { 982193326Sed ++begin; 983193326Sed } 984198092Srdivacky 985193326Sed // Skip over the entry quote. 986193326Sed assert(begin[0] == '\'' && "Invalid token lexed"); 987193326Sed ++begin; 988193326Sed 989193326Sed // Remove an optional ud-suffix. 990193326Sed if (end[-1] != '\'') { 991198092Srdivacky const char *UDSuffixEnd = end; 992193326Sed do { 993193326Sed --end; 994210299Sed } while (end[-1] != '\''); 995210299Sed UDSuffixBuf.assign(end, UDSuffixEnd); 996193326Sed UDSuffixOffset = end - TokBegin; 997193326Sed } 998218893Sdim 999218893Sdim // Trim the ending quote. 1000218893Sdim assert(end != begin && "Invalid token lexed"); 1001218893Sdim --end; 1002218893Sdim 1003218893Sdim // FIXME: The "Value" is an uint64_t so we can handle char literals of 1004193326Sed // up to 64-bits. 1005193326Sed // FIXME: This extensively assumes that 'char' is 8-bits. 1006193326Sed assert(PP.getTargetInfo().getCharWidth() == 8 && 1007218893Sdim "Assumes char is 8 bits"); 1008212904Sdim assert(PP.getTargetInfo().getIntWidth() <= 64 && 1009218893Sdim (PP.getTargetInfo().getIntWidth() & 7) == 0 && 1010212904Sdim "Assumes sizeof(int) on target is <= 64 and a multiple of char"); 1011212904Sdim assert(PP.getTargetInfo().getWCharWidth() <= 64 && 1012218893Sdim "Assumes sizeof(wchar) on target is <= 64"); 1013218893Sdim 1014212904Sdim SmallVector<uint32_t, 4> codepoint_buffer; 1015218893Sdim codepoint_buffer.resize(end - begin); 1016212904Sdim uint32_t *buffer_begin = &codepoint_buffer.front(); 1017212904Sdim uint32_t *buffer_end = buffer_begin + codepoint_buffer.size(); 1018193326Sed 1019193326Sed // Unicode escapes representing characters that cannot be correctly 1020193326Sed // represented in a single code unit are disallowed in character literals 1021193326Sed // by this implementation. 1022193326Sed uint32_t largest_character_for_kind; 1023193326Sed if (tok::wide_char_constant == Kind) { 1024193326Sed largest_character_for_kind = 1025193326Sed 0xFFFFFFFFu >> (32-PP.getTargetInfo().getWCharWidth()); 1026218893Sdim } else if (tok::utf16_char_constant == Kind) { 1027193326Sed largest_character_for_kind = 0xFFFF; 1028218893Sdim } else if (tok::utf32_char_constant == Kind) { 1029193326Sed largest_character_for_kind = 0x10FFFF; 1030198092Srdivacky } else { 1031205219Srdivacky largest_character_for_kind = 0x7Fu; 1032193326Sed } 1033218893Sdim 1034218893Sdim while (begin != end) { 1035218893Sdim // Is this a span of non-escape characters? 1036205219Srdivacky if (begin[0] != '\\') { 1037193326Sed char const *start = begin; 1038193326Sed do { 1039193326Sed ++begin; 1040198092Srdivacky } while (begin != end && *begin != '\\'); 1041193326Sed 1042193326Sed char const *tmp_in_start = start; 1043193326Sed uint32_t *tmp_out_start = buffer_begin; 1044193326Sed ConversionResult res = 1045193326Sed ConvertUTF8toUTF32(reinterpret_cast<UTF8 const **>(&start), 1046193326Sed reinterpret_cast<UTF8 const *>(begin), 1047198092Srdivacky &buffer_begin, buffer_end, strictConversion); 1048193326Sed if (res != conversionOK) { 1049193326Sed // If we see bad encoding for unprefixed character literals, warn and 1050193326Sed // simply copy the byte values, for compatibility with gcc and 1051198092Srdivacky // older versions of clang. 1052193326Sed bool NoErrorOnBadEncoding = isAscii(); 1053193326Sed unsigned Msg = diag::err_bad_character_encoding; 1054193326Sed if (NoErrorOnBadEncoding) 1055193326Sed Msg = diag::warn_bad_character_encoding; 1056193326Sed PP.Diag(Loc, Msg); 1057193326Sed if (NoErrorOnBadEncoding) { 1058198092Srdivacky start = tmp_in_start; 1059193326Sed buffer_begin = tmp_out_start; 1060193326Sed for (; start != begin; ++start, ++buffer_begin) 1061193326Sed *buffer_begin = static_cast<uint8_t>(*start); 1062218893Sdim } else { 1063218893Sdim HadError = true; 1064193326Sed } 1065193326Sed } else { 1066193326Sed for (; tmp_out_start < buffer_begin; ++tmp_out_start) { 1067198092Srdivacky if (*tmp_out_start > largest_character_for_kind) { 1068193326Sed HadError = true; 1069193326Sed PP.Diag(Loc, diag::err_character_too_large); 1070 } 1071 } 1072 } 1073 1074 continue; 1075 } 1076 // Is this a Universal Character Name escape? 1077 if (begin[1] == 'u' || begin[1] == 'U') { 1078 unsigned short UcnLen = 0; 1079 if (!ProcessUCNEscape(TokBegin, begin, end, *buffer_begin, UcnLen, 1080 FullSourceLoc(Loc, PP.getSourceManager()), 1081 &PP.getDiagnostics(), PP.getLangOpts(), true)) { 1082 HadError = true; 1083 } else if (*buffer_begin > largest_character_for_kind) { 1084 HadError = true; 1085 PP.Diag(Loc, diag::err_character_too_large); 1086 } 1087 1088 ++buffer_begin; 1089 continue; 1090 } 1091 unsigned CharWidth = getCharWidth(Kind, PP.getTargetInfo()); 1092 uint64_t result = 1093 ProcessCharEscape(TokBegin, begin, end, HadError, 1094 FullSourceLoc(Loc,PP.getSourceManager()), 1095 CharWidth, &PP.getDiagnostics(), PP.getLangOpts()); 1096 *buffer_begin++ = result; 1097 } 1098 1099 unsigned NumCharsSoFar = buffer_begin - &codepoint_buffer.front(); 1100 1101 if (NumCharsSoFar > 1) { 1102 if (isWide()) 1103 PP.Diag(Loc, diag::warn_extraneous_char_constant); 1104 else if (isAscii() && NumCharsSoFar == 4) 1105 PP.Diag(Loc, diag::ext_four_char_character_literal); 1106 else if (isAscii()) 1107 PP.Diag(Loc, diag::ext_multichar_character_literal); 1108 else 1109 PP.Diag(Loc, diag::err_multichar_utf_character_literal); 1110 IsMultiChar = true; 1111 } else { 1112 IsMultiChar = false; 1113 } 1114 1115 llvm::APInt LitVal(PP.getTargetInfo().getIntWidth(), 0); 1116 1117 // Narrow character literals act as though their value is concatenated 1118 // in this implementation, but warn on overflow. 1119 bool multi_char_too_long = false; 1120 if (isAscii() && isMultiChar()) { 1121 LitVal = 0; 1122 for (size_t i = 0; i < NumCharsSoFar; ++i) { 1123 // check for enough leading zeros to shift into 1124 multi_char_too_long |= (LitVal.countLeadingZeros() < 8); 1125 LitVal <<= 8; 1126 LitVal = LitVal + (codepoint_buffer[i] & 0xFF); 1127 } 1128 } else if (NumCharsSoFar > 0) { 1129 // otherwise just take the last character 1130 LitVal = buffer_begin[-1]; 1131 } 1132 1133 if (!HadError && multi_char_too_long) { 1134 PP.Diag(Loc, diag::warn_char_constant_too_large); 1135 } 1136 1137 // Transfer the value from APInt to uint64_t 1138 Value = LitVal.getZExtValue(); 1139 1140 // If this is a single narrow character, sign extend it (e.g. '\xFF' is "-1") 1141 // if 'char' is signed for this target (C99 6.4.4.4p10). Note that multiple 1142 // character constants are not sign extended in the this implementation: 1143 // '\xFF\xFF' = 65536 and '\x0\xFF' = 255, which matches GCC. 1144 if (isAscii() && NumCharsSoFar == 1 && (Value & 128) && 1145 PP.getLangOpts().CharIsSigned) 1146 Value = (signed char)Value; 1147} 1148 1149/// \verbatim 1150/// string-literal: [C++0x lex.string] 1151/// encoding-prefix " [s-char-sequence] " 1152/// encoding-prefix R raw-string 1153/// encoding-prefix: 1154/// u8 1155/// u 1156/// U 1157/// L 1158/// s-char-sequence: 1159/// s-char 1160/// s-char-sequence s-char 1161/// s-char: 1162/// any member of the source character set except the double-quote ", 1163/// backslash \, or new-line character 1164/// escape-sequence 1165/// universal-character-name 1166/// raw-string: 1167/// " d-char-sequence ( r-char-sequence ) d-char-sequence " 1168/// r-char-sequence: 1169/// r-char 1170/// r-char-sequence r-char 1171/// r-char: 1172/// any member of the source character set, except a right parenthesis ) 1173/// followed by the initial d-char-sequence (which may be empty) 1174/// followed by a double quote ". 1175/// d-char-sequence: 1176/// d-char 1177/// d-char-sequence d-char 1178/// d-char: 1179/// any member of the basic source character set except: 1180/// space, the left parenthesis (, the right parenthesis ), 1181/// the backslash \, and the control characters representing horizontal 1182/// tab, vertical tab, form feed, and newline. 1183/// escape-sequence: [C++0x lex.ccon] 1184/// simple-escape-sequence 1185/// octal-escape-sequence 1186/// hexadecimal-escape-sequence 1187/// simple-escape-sequence: 1188/// one of \' \" \? \\ \a \b \f \n \r \t \v 1189/// octal-escape-sequence: 1190/// \ octal-digit 1191/// \ octal-digit octal-digit 1192/// \ octal-digit octal-digit octal-digit 1193/// hexadecimal-escape-sequence: 1194/// \x hexadecimal-digit 1195/// hexadecimal-escape-sequence hexadecimal-digit 1196/// universal-character-name: 1197/// \u hex-quad 1198/// \U hex-quad hex-quad 1199/// hex-quad: 1200/// hex-digit hex-digit hex-digit hex-digit 1201/// \endverbatim 1202/// 1203StringLiteralParser:: 1204StringLiteralParser(const Token *StringToks, unsigned NumStringToks, 1205 Preprocessor &PP, bool Complain) 1206 : SM(PP.getSourceManager()), Features(PP.getLangOpts()), 1207 Target(PP.getTargetInfo()), Diags(Complain ? &PP.getDiagnostics() : 0), 1208 MaxTokenLength(0), SizeBound(0), CharByteWidth(0), Kind(tok::unknown), 1209 ResultPtr(ResultBuf.data()), hadError(false), Pascal(false) { 1210 init(StringToks, NumStringToks); 1211} 1212 1213void StringLiteralParser::init(const Token *StringToks, unsigned NumStringToks){ 1214 // The literal token may have come from an invalid source location (e.g. due 1215 // to a PCH error), in which case the token length will be 0. 1216 if (NumStringToks == 0 || StringToks[0].getLength() < 2) 1217 return DiagnoseLexingError(SourceLocation()); 1218 1219 // Scan all of the string portions, remember the max individual token length, 1220 // computing a bound on the concatenated string length, and see whether any 1221 // piece is a wide-string. If any of the string portions is a wide-string 1222 // literal, the result is a wide-string literal [C99 6.4.5p4]. 1223 assert(NumStringToks && "expected at least one token"); 1224 MaxTokenLength = StringToks[0].getLength(); 1225 assert(StringToks[0].getLength() >= 2 && "literal token is invalid!"); 1226 SizeBound = StringToks[0].getLength()-2; // -2 for "". 1227 Kind = StringToks[0].getKind(); 1228 1229 hadError = false; 1230 1231 // Implement Translation Phase #6: concatenation of string literals 1232 /// (C99 5.1.1.2p1). The common case is only one string fragment. 1233 for (unsigned i = 1; i != NumStringToks; ++i) { 1234 if (StringToks[i].getLength() < 2) 1235 return DiagnoseLexingError(StringToks[i].getLocation()); 1236 1237 // The string could be shorter than this if it needs cleaning, but this is a 1238 // reasonable bound, which is all we need. 1239 assert(StringToks[i].getLength() >= 2 && "literal token is invalid!"); 1240 SizeBound += StringToks[i].getLength()-2; // -2 for "". 1241 1242 // Remember maximum string piece length. 1243 if (StringToks[i].getLength() > MaxTokenLength) 1244 MaxTokenLength = StringToks[i].getLength(); 1245 1246 // Remember if we see any wide or utf-8/16/32 strings. 1247 // Also check for illegal concatenations. 1248 if (StringToks[i].isNot(Kind) && StringToks[i].isNot(tok::string_literal)) { 1249 if (isAscii()) { 1250 Kind = StringToks[i].getKind(); 1251 } else { 1252 if (Diags) 1253 Diags->Report(StringToks[i].getLocation(), 1254 diag::err_unsupported_string_concat); 1255 hadError = true; 1256 } 1257 } 1258 } 1259 1260 // Include space for the null terminator. 1261 ++SizeBound; 1262 1263 // TODO: K&R warning: "traditional C rejects string constant concatenation" 1264 1265 // Get the width in bytes of char/wchar_t/char16_t/char32_t 1266 CharByteWidth = getCharWidth(Kind, Target); 1267 assert((CharByteWidth & 7) == 0 && "Assumes character size is byte multiple"); 1268 CharByteWidth /= 8; 1269 1270 // The output buffer size needs to be large enough to hold wide characters. 1271 // This is a worst-case assumption which basically corresponds to L"" "long". 1272 SizeBound *= CharByteWidth; 1273 1274 // Size the temporary buffer to hold the result string data. 1275 ResultBuf.resize(SizeBound); 1276 1277 // Likewise, but for each string piece. 1278 SmallString<512> TokenBuf; 1279 TokenBuf.resize(MaxTokenLength); 1280 1281 // Loop over all the strings, getting their spelling, and expanding them to 1282 // wide strings as appropriate. 1283 ResultPtr = &ResultBuf[0]; // Next byte to fill in. 1284 1285 Pascal = false; 1286 1287 SourceLocation UDSuffixTokLoc; 1288 1289 for (unsigned i = 0, e = NumStringToks; i != e; ++i) { 1290 const char *ThisTokBuf = &TokenBuf[0]; 1291 // Get the spelling of the token, which eliminates trigraphs, etc. We know 1292 // that ThisTokBuf points to a buffer that is big enough for the whole token 1293 // and 'spelled' tokens can only shrink. 1294 bool StringInvalid = false; 1295 unsigned ThisTokLen = 1296 Lexer::getSpelling(StringToks[i], ThisTokBuf, SM, Features, 1297 &StringInvalid); 1298 if (StringInvalid) 1299 return DiagnoseLexingError(StringToks[i].getLocation()); 1300 1301 const char *ThisTokBegin = ThisTokBuf; 1302 const char *ThisTokEnd = ThisTokBuf+ThisTokLen; 1303 1304 // Remove an optional ud-suffix. 1305 if (ThisTokEnd[-1] != '"') { 1306 const char *UDSuffixEnd = ThisTokEnd; 1307 do { 1308 --ThisTokEnd; 1309 } while (ThisTokEnd[-1] != '"'); 1310 1311 StringRef UDSuffix(ThisTokEnd, UDSuffixEnd - ThisTokEnd); 1312 1313 if (UDSuffixBuf.empty()) { 1314 UDSuffixBuf.assign(UDSuffix); 1315 UDSuffixToken = i; 1316 UDSuffixOffset = ThisTokEnd - ThisTokBuf; 1317 UDSuffixTokLoc = StringToks[i].getLocation(); 1318 } else if (!UDSuffixBuf.equals(UDSuffix)) { 1319 // C++11 [lex.ext]p8: At the end of phase 6, if a string literal is the 1320 // result of a concatenation involving at least one user-defined-string- 1321 // literal, all the participating user-defined-string-literals shall 1322 // have the same ud-suffix. 1323 if (Diags) { 1324 SourceLocation TokLoc = StringToks[i].getLocation(); 1325 Diags->Report(TokLoc, diag::err_string_concat_mixed_suffix) 1326 << UDSuffixBuf << UDSuffix 1327 << SourceRange(UDSuffixTokLoc, UDSuffixTokLoc) 1328 << SourceRange(TokLoc, TokLoc); 1329 } 1330 hadError = true; 1331 } 1332 } 1333 1334 // Strip the end quote. 1335 --ThisTokEnd; 1336 1337 // TODO: Input character set mapping support. 1338 1339 // Skip marker for wide or unicode strings. 1340 if (ThisTokBuf[0] == 'L' || ThisTokBuf[0] == 'u' || ThisTokBuf[0] == 'U') { 1341 ++ThisTokBuf; 1342 // Skip 8 of u8 marker for utf8 strings. 1343 if (ThisTokBuf[0] == '8') 1344 ++ThisTokBuf; 1345 } 1346 1347 // Check for raw string 1348 if (ThisTokBuf[0] == 'R') { 1349 ThisTokBuf += 2; // skip R" 1350 1351 const char *Prefix = ThisTokBuf; 1352 while (ThisTokBuf[0] != '(') 1353 ++ThisTokBuf; 1354 ++ThisTokBuf; // skip '(' 1355 1356 // Remove same number of characters from the end 1357 ThisTokEnd -= ThisTokBuf - Prefix; 1358 assert(ThisTokEnd >= ThisTokBuf && "malformed raw string literal"); 1359 1360 // Copy the string over 1361 if (CopyStringFragment(StringToks[i], ThisTokBegin, 1362 StringRef(ThisTokBuf, ThisTokEnd - ThisTokBuf))) 1363 hadError = true; 1364 } else { 1365 if (ThisTokBuf[0] != '"') { 1366 // The file may have come from PCH and then changed after loading the 1367 // PCH; Fail gracefully. 1368 return DiagnoseLexingError(StringToks[i].getLocation()); 1369 } 1370 ++ThisTokBuf; // skip " 1371 1372 // Check if this is a pascal string 1373 if (Features.PascalStrings && ThisTokBuf + 1 != ThisTokEnd && 1374 ThisTokBuf[0] == '\\' && ThisTokBuf[1] == 'p') { 1375 1376 // If the \p sequence is found in the first token, we have a pascal string 1377 // Otherwise, if we already have a pascal string, ignore the first \p 1378 if (i == 0) { 1379 ++ThisTokBuf; 1380 Pascal = true; 1381 } else if (Pascal) 1382 ThisTokBuf += 2; 1383 } 1384 1385 while (ThisTokBuf != ThisTokEnd) { 1386 // Is this a span of non-escape characters? 1387 if (ThisTokBuf[0] != '\\') { 1388 const char *InStart = ThisTokBuf; 1389 do { 1390 ++ThisTokBuf; 1391 } while (ThisTokBuf != ThisTokEnd && ThisTokBuf[0] != '\\'); 1392 1393 // Copy the character span over. 1394 if (CopyStringFragment(StringToks[i], ThisTokBegin, 1395 StringRef(InStart, ThisTokBuf - InStart))) 1396 hadError = true; 1397 continue; 1398 } 1399 // Is this a Universal Character Name escape? 1400 if (ThisTokBuf[1] == 'u' || ThisTokBuf[1] == 'U') { 1401 EncodeUCNEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, 1402 ResultPtr, hadError, 1403 FullSourceLoc(StringToks[i].getLocation(), SM), 1404 CharByteWidth, Diags, Features); 1405 continue; 1406 } 1407 // Otherwise, this is a non-UCN escape character. Process it. 1408 unsigned ResultChar = 1409 ProcessCharEscape(ThisTokBegin, ThisTokBuf, ThisTokEnd, hadError, 1410 FullSourceLoc(StringToks[i].getLocation(), SM), 1411 CharByteWidth*8, Diags, Features); 1412 1413 if (CharByteWidth == 4) { 1414 // FIXME: Make the type of the result buffer correct instead of 1415 // using reinterpret_cast. 1416 UTF32 *ResultWidePtr = reinterpret_cast<UTF32*>(ResultPtr); 1417 *ResultWidePtr = ResultChar; 1418 ResultPtr += 4; 1419 } else if (CharByteWidth == 2) { 1420 // FIXME: Make the type of the result buffer correct instead of 1421 // using reinterpret_cast. 1422 UTF16 *ResultWidePtr = reinterpret_cast<UTF16*>(ResultPtr); 1423 *ResultWidePtr = ResultChar & 0xFFFF; 1424 ResultPtr += 2; 1425 } else { 1426 assert(CharByteWidth == 1 && "Unexpected char width"); 1427 *ResultPtr++ = ResultChar & 0xFF; 1428 } 1429 } 1430 } 1431 } 1432 1433 if (Pascal) { 1434 if (CharByteWidth == 4) { 1435 // FIXME: Make the type of the result buffer correct instead of 1436 // using reinterpret_cast. 1437 UTF32 *ResultWidePtr = reinterpret_cast<UTF32*>(ResultBuf.data()); 1438 ResultWidePtr[0] = GetNumStringChars() - 1; 1439 } else if (CharByteWidth == 2) { 1440 // FIXME: Make the type of the result buffer correct instead of 1441 // using reinterpret_cast. 1442 UTF16 *ResultWidePtr = reinterpret_cast<UTF16*>(ResultBuf.data()); 1443 ResultWidePtr[0] = GetNumStringChars() - 1; 1444 } else { 1445 assert(CharByteWidth == 1 && "Unexpected char width"); 1446 ResultBuf[0] = GetNumStringChars() - 1; 1447 } 1448 1449 // Verify that pascal strings aren't too large. 1450 if (GetStringLength() > 256) { 1451 if (Diags) 1452 Diags->Report(StringToks[0].getLocation(), 1453 diag::err_pascal_string_too_long) 1454 << SourceRange(StringToks[0].getLocation(), 1455 StringToks[NumStringToks-1].getLocation()); 1456 hadError = true; 1457 return; 1458 } 1459 } else if (Diags) { 1460 // Complain if this string literal has too many characters. 1461 unsigned MaxChars = Features.CPlusPlus? 65536 : Features.C99 ? 4095 : 509; 1462 1463 if (GetNumStringChars() > MaxChars) 1464 Diags->Report(StringToks[0].getLocation(), 1465 diag::ext_string_too_long) 1466 << GetNumStringChars() << MaxChars 1467 << (Features.CPlusPlus ? 2 : Features.C99 ? 1 : 0) 1468 << SourceRange(StringToks[0].getLocation(), 1469 StringToks[NumStringToks-1].getLocation()); 1470 } 1471} 1472 1473static const char *resyncUTF8(const char *Err, const char *End) { 1474 if (Err == End) 1475 return End; 1476 End = Err + std::min<unsigned>(getNumBytesForUTF8(*Err), End-Err); 1477 while (++Err != End && (*Err & 0xC0) == 0x80) 1478 ; 1479 return Err; 1480} 1481 1482/// \brief This function copies from Fragment, which is a sequence of bytes 1483/// within Tok's contents (which begin at TokBegin) into ResultPtr. 1484/// Performs widening for multi-byte characters. 1485bool StringLiteralParser::CopyStringFragment(const Token &Tok, 1486 const char *TokBegin, 1487 StringRef Fragment) { 1488 const UTF8 *ErrorPtrTmp; 1489 if (ConvertUTF8toWide(CharByteWidth, Fragment, ResultPtr, ErrorPtrTmp)) 1490 return false; 1491 1492 // If we see bad encoding for unprefixed string literals, warn and 1493 // simply copy the byte values, for compatibility with gcc and older 1494 // versions of clang. 1495 bool NoErrorOnBadEncoding = isAscii(); 1496 if (NoErrorOnBadEncoding) { 1497 memcpy(ResultPtr, Fragment.data(), Fragment.size()); 1498 ResultPtr += Fragment.size(); 1499 } 1500 1501 if (Diags) { 1502 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp); 1503 1504 FullSourceLoc SourceLoc(Tok.getLocation(), SM); 1505 const DiagnosticBuilder &Builder = 1506 Diag(Diags, Features, SourceLoc, TokBegin, 1507 ErrorPtr, resyncUTF8(ErrorPtr, Fragment.end()), 1508 NoErrorOnBadEncoding ? diag::warn_bad_string_encoding 1509 : diag::err_bad_string_encoding); 1510 1511 const char *NextStart = resyncUTF8(ErrorPtr, Fragment.end()); 1512 StringRef NextFragment(NextStart, Fragment.end()-NextStart); 1513 1514 // Decode into a dummy buffer. 1515 SmallString<512> Dummy; 1516 Dummy.reserve(Fragment.size() * CharByteWidth); 1517 char *Ptr = Dummy.data(); 1518 1519 while (!Builder.hasMaxRanges() && 1520 !ConvertUTF8toWide(CharByteWidth, NextFragment, Ptr, ErrorPtrTmp)) { 1521 const char *ErrorPtr = reinterpret_cast<const char *>(ErrorPtrTmp); 1522 NextStart = resyncUTF8(ErrorPtr, Fragment.end()); 1523 Builder << MakeCharSourceRange(Features, SourceLoc, TokBegin, 1524 ErrorPtr, NextStart); 1525 NextFragment = StringRef(NextStart, Fragment.end()-NextStart); 1526 } 1527 } 1528 return !NoErrorOnBadEncoding; 1529} 1530 1531void StringLiteralParser::DiagnoseLexingError(SourceLocation Loc) { 1532 hadError = true; 1533 if (Diags) 1534 Diags->Report(Loc, diag::err_lexing_string); 1535} 1536 1537/// getOffsetOfStringByte - This function returns the offset of the 1538/// specified byte of the string data represented by Token. This handles 1539/// advancing over escape sequences in the string. 1540unsigned StringLiteralParser::getOffsetOfStringByte(const Token &Tok, 1541 unsigned ByteNo) const { 1542 // Get the spelling of the token. 1543 SmallString<32> SpellingBuffer; 1544 SpellingBuffer.resize(Tok.getLength()); 1545 1546 bool StringInvalid = false; 1547 const char *SpellingPtr = &SpellingBuffer[0]; 1548 unsigned TokLen = Lexer::getSpelling(Tok, SpellingPtr, SM, Features, 1549 &StringInvalid); 1550 if (StringInvalid) 1551 return 0; 1552 1553 const char *SpellingStart = SpellingPtr; 1554 const char *SpellingEnd = SpellingPtr+TokLen; 1555 1556 // Handle UTF-8 strings just like narrow strings. 1557 if (SpellingPtr[0] == 'u' && SpellingPtr[1] == '8') 1558 SpellingPtr += 2; 1559 1560 assert(SpellingPtr[0] != 'L' && SpellingPtr[0] != 'u' && 1561 SpellingPtr[0] != 'U' && "Doesn't handle wide or utf strings yet"); 1562 1563 // For raw string literals, this is easy. 1564 if (SpellingPtr[0] == 'R') { 1565 assert(SpellingPtr[1] == '"' && "Should be a raw string literal!"); 1566 // Skip 'R"'. 1567 SpellingPtr += 2; 1568 while (*SpellingPtr != '(') { 1569 ++SpellingPtr; 1570 assert(SpellingPtr < SpellingEnd && "Missing ( for raw string literal"); 1571 } 1572 // Skip '('. 1573 ++SpellingPtr; 1574 return SpellingPtr - SpellingStart + ByteNo; 1575 } 1576 1577 // Skip over the leading quote 1578 assert(SpellingPtr[0] == '"' && "Should be a string literal!"); 1579 ++SpellingPtr; 1580 1581 // Skip over bytes until we find the offset we're looking for. 1582 while (ByteNo) { 1583 assert(SpellingPtr < SpellingEnd && "Didn't find byte offset!"); 1584 1585 // Step over non-escapes simply. 1586 if (*SpellingPtr != '\\') { 1587 ++SpellingPtr; 1588 --ByteNo; 1589 continue; 1590 } 1591 1592 // Otherwise, this is an escape character. Advance over it. 1593 bool HadError = false; 1594 if (SpellingPtr[1] == 'u' || SpellingPtr[1] == 'U') { 1595 const char *EscapePtr = SpellingPtr; 1596 unsigned Len = MeasureUCNEscape(SpellingStart, SpellingPtr, SpellingEnd, 1597 1, Features, HadError); 1598 if (Len > ByteNo) { 1599 // ByteNo is somewhere within the escape sequence. 1600 SpellingPtr = EscapePtr; 1601 break; 1602 } 1603 ByteNo -= Len; 1604 } else { 1605 ProcessCharEscape(SpellingStart, SpellingPtr, SpellingEnd, HadError, 1606 FullSourceLoc(Tok.getLocation(), SM), 1607 CharByteWidth*8, Diags, Features); 1608 --ByteNo; 1609 } 1610 assert(!HadError && "This method isn't valid on erroneous strings"); 1611 } 1612 1613 return SpellingPtr-SpellingStart; 1614} 1615