Lexer.cpp revision 24f0e48c0aa62f2268e061aad70f9b19a59e7b52
1//===--- Lexer.cpp - C Language Family Lexer ------------------------------===// 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 Lexer and Token interfaces. 11// 12//===----------------------------------------------------------------------===// 13// 14// TODO: GCC Diagnostics emitted by the lexer: 15// PEDWARN: (form feed|vertical tab) in preprocessing directive 16// 17// Universal characters, unicode, char mapping: 18// WARNING: `%.*s' is not in NFKC 19// WARNING: `%.*s' is not in NFC 20// 21// Other: 22// TODO: Options to support: 23// -fexec-charset,-fwide-exec-charset 24// 25//===----------------------------------------------------------------------===// 26 27#include "clang/Lex/Lexer.h" 28#include "clang/Lex/Preprocessor.h" 29#include "clang/Lex/LexDiagnostic.h" 30#include "clang/Basic/SourceManager.h" 31#include "llvm/Support/Compiler.h" 32#include "llvm/Support/MemoryBuffer.h" 33#include <cctype> 34using namespace clang; 35 36static void InitCharacterInfo(); 37 38//===----------------------------------------------------------------------===// 39// Token Class Implementation 40//===----------------------------------------------------------------------===// 41 42/// isObjCAtKeyword - Return true if we have an ObjC keyword identifier. 43bool Token::isObjCAtKeyword(tok::ObjCKeywordKind objcKey) const { 44 if (IdentifierInfo *II = getIdentifierInfo()) 45 return II->getObjCKeywordID() == objcKey; 46 return false; 47} 48 49/// getObjCKeywordID - Return the ObjC keyword kind. 50tok::ObjCKeywordKind Token::getObjCKeywordID() const { 51 IdentifierInfo *specId = getIdentifierInfo(); 52 return specId ? specId->getObjCKeywordID() : tok::objc_not_keyword; 53} 54 55 56//===----------------------------------------------------------------------===// 57// Lexer Class Implementation 58//===----------------------------------------------------------------------===// 59 60void Lexer::InitLexer(const char *BufStart, const char *BufPtr, 61 const char *BufEnd) { 62 InitCharacterInfo(); 63 64 BufferStart = BufStart; 65 BufferPtr = BufPtr; 66 BufferEnd = BufEnd; 67 68 assert(BufEnd[0] == 0 && 69 "We assume that the input buffer has a null character at the end" 70 " to simplify lexing!"); 71 72 Is_PragmaLexer = false; 73 74 // Start of the file is a start of line. 75 IsAtStartOfLine = true; 76 77 // We are not after parsing a #. 78 ParsingPreprocessorDirective = false; 79 80 // We are not after parsing #include. 81 ParsingFilename = false; 82 83 // We are not in raw mode. Raw mode disables diagnostics and interpretation 84 // of tokens (e.g. identifiers, thus disabling macro expansion). It is used 85 // to quickly lex the tokens of the buffer, e.g. when handling a "#if 0" block 86 // or otherwise skipping over tokens. 87 LexingRawMode = false; 88 89 // Default to not keeping comments. 90 ExtendedTokenMode = 0; 91} 92 93/// Lexer constructor - Create a new lexer object for the specified buffer 94/// with the specified preprocessor managing the lexing process. This lexer 95/// assumes that the associated file buffer and Preprocessor objects will 96/// outlive it, so it doesn't take ownership of either of them. 97Lexer::Lexer(FileID FID, Preprocessor &PP) 98 : PreprocessorLexer(&PP, FID), 99 FileLoc(PP.getSourceManager().getLocForStartOfFile(FID)), 100 Features(PP.getLangOptions()) { 101 102 const llvm::MemoryBuffer *InputFile = PP.getSourceManager().getBuffer(FID); 103 104 InitLexer(InputFile->getBufferStart(), InputFile->getBufferStart(), 105 InputFile->getBufferEnd()); 106 107 // Default to keeping comments if the preprocessor wants them. 108 SetCommentRetentionState(PP.getCommentRetentionState()); 109} 110 111/// Lexer constructor - Create a new raw lexer object. This object is only 112/// suitable for calls to 'LexRawToken'. This lexer assumes that the text 113/// range will outlive it, so it doesn't take ownership of it. 114Lexer::Lexer(SourceLocation fileloc, const LangOptions &features, 115 const char *BufStart, const char *BufPtr, const char *BufEnd) 116 : FileLoc(fileloc), Features(features) { 117 118 InitLexer(BufStart, BufPtr, BufEnd); 119 120 // We *are* in raw mode. 121 LexingRawMode = true; 122} 123 124/// Lexer constructor - Create a new raw lexer object. This object is only 125/// suitable for calls to 'LexRawToken'. This lexer assumes that the text 126/// range will outlive it, so it doesn't take ownership of it. 127Lexer::Lexer(FileID FID, const SourceManager &SM, const LangOptions &features) 128 : FileLoc(SM.getLocForStartOfFile(FID)), Features(features) { 129 const llvm::MemoryBuffer *FromFile = SM.getBuffer(FID); 130 131 InitLexer(FromFile->getBufferStart(), FromFile->getBufferStart(), 132 FromFile->getBufferEnd()); 133 134 // We *are* in raw mode. 135 LexingRawMode = true; 136} 137 138/// Create_PragmaLexer: Lexer constructor - Create a new lexer object for 139/// _Pragma expansion. This has a variety of magic semantics that this method 140/// sets up. It returns a new'd Lexer that must be delete'd when done. 141/// 142/// On entrance to this routine, TokStartLoc is a macro location which has a 143/// spelling loc that indicates the bytes to be lexed for the token and an 144/// instantiation location that indicates where all lexed tokens should be 145/// "expanded from". 146/// 147/// FIXME: It would really be nice to make _Pragma just be a wrapper around a 148/// normal lexer that remaps tokens as they fly by. This would require making 149/// Preprocessor::Lex virtual. Given that, we could just dump in a magic lexer 150/// interface that could handle this stuff. This would pull GetMappedTokenLoc 151/// out of the critical path of the lexer! 152/// 153Lexer *Lexer::Create_PragmaLexer(SourceLocation SpellingLoc, 154 SourceLocation InstantiationLocStart, 155 SourceLocation InstantiationLocEnd, 156 unsigned TokLen, Preprocessor &PP) { 157 SourceManager &SM = PP.getSourceManager(); 158 159 // Create the lexer as if we were going to lex the file normally. 160 FileID SpellingFID = SM.getFileID(SpellingLoc); 161 Lexer *L = new Lexer(SpellingFID, PP); 162 163 // Now that the lexer is created, change the start/end locations so that we 164 // just lex the subsection of the file that we want. This is lexing from a 165 // scratch buffer. 166 const char *StrData = SM.getCharacterData(SpellingLoc); 167 168 L->BufferPtr = StrData; 169 L->BufferEnd = StrData+TokLen; 170 assert(L->BufferEnd[0] == 0 && "Buffer is not nul terminated!"); 171 172 // Set the SourceLocation with the remapping information. This ensures that 173 // GetMappedTokenLoc will remap the tokens as they are lexed. 174 L->FileLoc = SM.createInstantiationLoc(SM.getLocForStartOfFile(SpellingFID), 175 InstantiationLocStart, 176 InstantiationLocEnd, TokLen); 177 178 // Ensure that the lexer thinks it is inside a directive, so that end \n will 179 // return an EOM token. 180 L->ParsingPreprocessorDirective = true; 181 182 // This lexer really is for _Pragma. 183 L->Is_PragmaLexer = true; 184 return L; 185} 186 187 188/// Stringify - Convert the specified string into a C string, with surrounding 189/// ""'s, and with escaped \ and " characters. 190std::string Lexer::Stringify(const std::string &Str, bool Charify) { 191 std::string Result = Str; 192 char Quote = Charify ? '\'' : '"'; 193 for (unsigned i = 0, e = Result.size(); i != e; ++i) { 194 if (Result[i] == '\\' || Result[i] == Quote) { 195 Result.insert(Result.begin()+i, '\\'); 196 ++i; ++e; 197 } 198 } 199 return Result; 200} 201 202/// Stringify - Convert the specified string into a C string by escaping '\' 203/// and " characters. This does not add surrounding ""'s to the string. 204void Lexer::Stringify(llvm::SmallVectorImpl<char> &Str) { 205 for (unsigned i = 0, e = Str.size(); i != e; ++i) { 206 if (Str[i] == '\\' || Str[i] == '"') { 207 Str.insert(Str.begin()+i, '\\'); 208 ++i; ++e; 209 } 210 } 211} 212 213 214/// MeasureTokenLength - Relex the token at the specified location and return 215/// its length in bytes in the input file. If the token needs cleaning (e.g. 216/// includes a trigraph or an escaped newline) then this count includes bytes 217/// that are part of that. 218unsigned Lexer::MeasureTokenLength(SourceLocation Loc, 219 const SourceManager &SM, 220 const LangOptions &LangOpts) { 221 // TODO: this could be special cased for common tokens like identifiers, ')', 222 // etc to make this faster, if it mattered. Just look at StrData[0] to handle 223 // all obviously single-char tokens. This could use 224 // Lexer::isObviouslySimpleCharacter for example to handle identifiers or 225 // something. 226 227 // If this comes from a macro expansion, we really do want the macro name, not 228 // the token this macro expanded to. 229 Loc = SM.getInstantiationLoc(Loc); 230 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(Loc); 231 std::pair<const char *,const char *> Buffer = SM.getBufferData(LocInfo.first); 232 const char *StrData = Buffer.first+LocInfo.second; 233 234 // Create a lexer starting at the beginning of this token. 235 Lexer TheLexer(Loc, LangOpts, Buffer.first, StrData, Buffer.second); 236 Token TheTok; 237 TheLexer.LexFromRawLexer(TheTok); 238 return TheTok.getLength(); 239} 240 241//===----------------------------------------------------------------------===// 242// Character information. 243//===----------------------------------------------------------------------===// 244 245static unsigned char CharInfo[256]; 246 247enum { 248 CHAR_HORZ_WS = 0x01, // ' ', '\t', '\f', '\v'. Note, no '\0' 249 CHAR_VERT_WS = 0x02, // '\r', '\n' 250 CHAR_LETTER = 0x04, // a-z,A-Z 251 CHAR_NUMBER = 0x08, // 0-9 252 CHAR_UNDER = 0x10, // _ 253 CHAR_PERIOD = 0x20 // . 254}; 255 256static void InitCharacterInfo() { 257 static bool isInited = false; 258 if (isInited) return; 259 isInited = true; 260 261 // Intiialize the CharInfo table. 262 // TODO: statically initialize this. 263 CharInfo[(int)' '] = CharInfo[(int)'\t'] = 264 CharInfo[(int)'\f'] = CharInfo[(int)'\v'] = CHAR_HORZ_WS; 265 CharInfo[(int)'\n'] = CharInfo[(int)'\r'] = CHAR_VERT_WS; 266 267 CharInfo[(int)'_'] = CHAR_UNDER; 268 CharInfo[(int)'.'] = CHAR_PERIOD; 269 for (unsigned i = 'a'; i <= 'z'; ++i) 270 CharInfo[i] = CharInfo[i+'A'-'a'] = CHAR_LETTER; 271 for (unsigned i = '0'; i <= '9'; ++i) 272 CharInfo[i] = CHAR_NUMBER; 273} 274 275/// isIdentifierBody - Return true if this is the body character of an 276/// identifier, which is [a-zA-Z0-9_]. 277static inline bool isIdentifierBody(unsigned char c) { 278 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER)) ? true : false; 279} 280 281/// isHorizontalWhitespace - Return true if this character is horizontal 282/// whitespace: ' ', '\t', '\f', '\v'. Note that this returns false for '\0'. 283static inline bool isHorizontalWhitespace(unsigned char c) { 284 return (CharInfo[c] & CHAR_HORZ_WS) ? true : false; 285} 286 287/// isWhitespace - Return true if this character is horizontal or vertical 288/// whitespace: ' ', '\t', '\f', '\v', '\n', '\r'. Note that this returns false 289/// for '\0'. 290static inline bool isWhitespace(unsigned char c) { 291 return (CharInfo[c] & (CHAR_HORZ_WS|CHAR_VERT_WS)) ? true : false; 292} 293 294/// isNumberBody - Return true if this is the body character of an 295/// preprocessing number, which is [a-zA-Z0-9_.]. 296static inline bool isNumberBody(unsigned char c) { 297 return (CharInfo[c] & (CHAR_LETTER|CHAR_NUMBER|CHAR_UNDER|CHAR_PERIOD)) ? 298 true : false; 299} 300 301 302//===----------------------------------------------------------------------===// 303// Diagnostics forwarding code. 304//===----------------------------------------------------------------------===// 305 306/// GetMappedTokenLoc - If lexing out of a 'mapped buffer', where we pretend the 307/// lexer buffer was all instantiated at a single point, perform the mapping. 308/// This is currently only used for _Pragma implementation, so it is the slow 309/// path of the hot getSourceLocation method. Do not allow it to be inlined. 310static SourceLocation GetMappedTokenLoc(Preprocessor &PP, 311 SourceLocation FileLoc, 312 unsigned CharNo, 313 unsigned TokLen) DISABLE_INLINE; 314static SourceLocation GetMappedTokenLoc(Preprocessor &PP, 315 SourceLocation FileLoc, 316 unsigned CharNo, unsigned TokLen) { 317 assert(FileLoc.isMacroID() && "Must be an instantiation"); 318 319 // Otherwise, we're lexing "mapped tokens". This is used for things like 320 // _Pragma handling. Combine the instantiation location of FileLoc with the 321 // spelling location. 322 SourceManager &SM = PP.getSourceManager(); 323 324 // Create a new SLoc which is expanded from Instantiation(FileLoc) but whose 325 // characters come from spelling(FileLoc)+Offset. 326 SourceLocation SpellingLoc = SM.getSpellingLoc(FileLoc); 327 SpellingLoc = SpellingLoc.getFileLocWithOffset(CharNo); 328 329 // Figure out the expansion loc range, which is the range covered by the 330 // original _Pragma(...) sequence. 331 std::pair<SourceLocation,SourceLocation> II = 332 SM.getImmediateInstantiationRange(FileLoc); 333 334 return SM.createInstantiationLoc(SpellingLoc, II.first, II.second, TokLen); 335} 336 337/// getSourceLocation - Return a source location identifier for the specified 338/// offset in the current file. 339SourceLocation Lexer::getSourceLocation(const char *Loc, 340 unsigned TokLen) const { 341 assert(Loc >= BufferStart && Loc <= BufferEnd && 342 "Location out of range for this buffer!"); 343 344 // In the normal case, we're just lexing from a simple file buffer, return 345 // the file id from FileLoc with the offset specified. 346 unsigned CharNo = Loc-BufferStart; 347 if (FileLoc.isFileID()) 348 return FileLoc.getFileLocWithOffset(CharNo); 349 350 // Otherwise, this is the _Pragma lexer case, which pretends that all of the 351 // tokens are lexed from where the _Pragma was defined. 352 assert(PP && "This doesn't work on raw lexers"); 353 return GetMappedTokenLoc(*PP, FileLoc, CharNo, TokLen); 354} 355 356/// Diag - Forwarding function for diagnostics. This translate a source 357/// position in the current buffer into a SourceLocation object for rendering. 358DiagnosticBuilder Lexer::Diag(const char *Loc, unsigned DiagID) const { 359 return PP->Diag(getSourceLocation(Loc), DiagID); 360} 361 362//===----------------------------------------------------------------------===// 363// Trigraph and Escaped Newline Handling Code. 364//===----------------------------------------------------------------------===// 365 366/// GetTrigraphCharForLetter - Given a character that occurs after a ?? pair, 367/// return the decoded trigraph letter it corresponds to, or '\0' if nothing. 368static char GetTrigraphCharForLetter(char Letter) { 369 switch (Letter) { 370 default: return 0; 371 case '=': return '#'; 372 case ')': return ']'; 373 case '(': return '['; 374 case '!': return '|'; 375 case '\'': return '^'; 376 case '>': return '}'; 377 case '/': return '\\'; 378 case '<': return '{'; 379 case '-': return '~'; 380 } 381} 382 383/// DecodeTrigraphChar - If the specified character is a legal trigraph when 384/// prefixed with ??, emit a trigraph warning. If trigraphs are enabled, 385/// return the result character. Finally, emit a warning about trigraph use 386/// whether trigraphs are enabled or not. 387static char DecodeTrigraphChar(const char *CP, Lexer *L) { 388 char Res = GetTrigraphCharForLetter(*CP); 389 if (!Res || !L) return Res; 390 391 if (!L->getFeatures().Trigraphs) { 392 if (!L->isLexingRawMode()) 393 L->Diag(CP-2, diag::trigraph_ignored); 394 return 0; 395 } 396 397 if (!L->isLexingRawMode()) 398 L->Diag(CP-2, diag::trigraph_converted) << std::string()+Res; 399 return Res; 400} 401 402/// getEscapedNewLineSize - Return the size of the specified escaped newline, 403/// or 0 if it is not an escaped newline. P[-1] is known to be a "\" or a 404/// trigraph equivalent on entry to this function. 405unsigned Lexer::getEscapedNewLineSize(const char *Ptr) { 406 unsigned Size = 0; 407 while (isWhitespace(Ptr[Size])) { 408 ++Size; 409 410 if (Ptr[Size-1] != '\n' && Ptr[Size-1] != '\r') 411 continue; 412 413 // If this is a \r\n or \n\r, skip the other half. 414 if ((Ptr[Size] == '\r' || Ptr[Size] == '\n') && 415 Ptr[Size-1] != Ptr[Size]) 416 ++Size; 417 418 return Size; 419 } 420 421 // Not an escaped newline, must be a \t or something else. 422 return 0; 423} 424 425 426/// getCharAndSizeSlow - Peek a single 'character' from the specified buffer, 427/// get its size, and return it. This is tricky in several cases: 428/// 1. If currently at the start of a trigraph, we warn about the trigraph, 429/// then either return the trigraph (skipping 3 chars) or the '?', 430/// depending on whether trigraphs are enabled or not. 431/// 2. If this is an escaped newline (potentially with whitespace between 432/// the backslash and newline), implicitly skip the newline and return 433/// the char after it. 434/// 3. If this is a UCN, return it. FIXME: C++ UCN's? 435/// 436/// This handles the slow/uncommon case of the getCharAndSize method. Here we 437/// know that we can accumulate into Size, and that we have already incremented 438/// Ptr by Size bytes. 439/// 440/// NOTE: When this method is updated, getCharAndSizeSlowNoWarn (below) should 441/// be updated to match. 442/// 443char Lexer::getCharAndSizeSlow(const char *Ptr, unsigned &Size, 444 Token *Tok) { 445 // If we have a slash, look for an escaped newline. 446 if (Ptr[0] == '\\') { 447 ++Size; 448 ++Ptr; 449Slash: 450 // Common case, backslash-char where the char is not whitespace. 451 if (!isWhitespace(Ptr[0])) return '\\'; 452 453 // See if we have optional whitespace characters followed by a newline. 454 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) { 455 // Remember that this token needs to be cleaned. 456 if (Tok) Tok->setFlag(Token::NeedsCleaning); 457 458 // Warn if there was whitespace between the backslash and newline. 459 if (EscapedNewLineSize != 1 && Tok && !isLexingRawMode()) 460 Diag(Ptr, diag::backslash_newline_space); 461 462 // Found backslash<whitespace><newline>. Parse the char after it. 463 Size += EscapedNewLineSize; 464 Ptr += EscapedNewLineSize; 465 // Use slow version to accumulate a correct size field. 466 return getCharAndSizeSlow(Ptr, Size, Tok); 467 } 468 469 // Otherwise, this is not an escaped newline, just return the slash. 470 return '\\'; 471 } 472 473 // If this is a trigraph, process it. 474 if (Ptr[0] == '?' && Ptr[1] == '?') { 475 // If this is actually a legal trigraph (not something like "??x"), emit 476 // a trigraph warning. If so, and if trigraphs are enabled, return it. 477 if (char C = DecodeTrigraphChar(Ptr+2, Tok ? this : 0)) { 478 // Remember that this token needs to be cleaned. 479 if (Tok) Tok->setFlag(Token::NeedsCleaning); 480 481 Ptr += 3; 482 Size += 3; 483 if (C == '\\') goto Slash; 484 return C; 485 } 486 } 487 488 // If this is neither, return a single character. 489 ++Size; 490 return *Ptr; 491} 492 493 494/// getCharAndSizeSlowNoWarn - Handle the slow/uncommon case of the 495/// getCharAndSizeNoWarn method. Here we know that we can accumulate into Size, 496/// and that we have already incremented Ptr by Size bytes. 497/// 498/// NOTE: When this method is updated, getCharAndSizeSlow (above) should 499/// be updated to match. 500char Lexer::getCharAndSizeSlowNoWarn(const char *Ptr, unsigned &Size, 501 const LangOptions &Features) { 502 // If we have a slash, look for an escaped newline. 503 if (Ptr[0] == '\\') { 504 ++Size; 505 ++Ptr; 506Slash: 507 // Common case, backslash-char where the char is not whitespace. 508 if (!isWhitespace(Ptr[0])) return '\\'; 509 510 // See if we have optional whitespace characters followed by a newline. 511 if (unsigned EscapedNewLineSize = getEscapedNewLineSize(Ptr)) { 512 // Found backslash<whitespace><newline>. Parse the char after it. 513 Size += EscapedNewLineSize; 514 Ptr += EscapedNewLineSize; 515 516 // Use slow version to accumulate a correct size field. 517 return getCharAndSizeSlowNoWarn(Ptr, Size, Features); 518 } 519 520 // Otherwise, this is not an escaped newline, just return the slash. 521 return '\\'; 522 } 523 524 // If this is a trigraph, process it. 525 if (Features.Trigraphs && Ptr[0] == '?' && Ptr[1] == '?') { 526 // If this is actually a legal trigraph (not something like "??x"), return 527 // it. 528 if (char C = GetTrigraphCharForLetter(Ptr[2])) { 529 Ptr += 3; 530 Size += 3; 531 if (C == '\\') goto Slash; 532 return C; 533 } 534 } 535 536 // If this is neither, return a single character. 537 ++Size; 538 return *Ptr; 539} 540 541//===----------------------------------------------------------------------===// 542// Helper methods for lexing. 543//===----------------------------------------------------------------------===// 544 545void Lexer::LexIdentifier(Token &Result, const char *CurPtr) { 546 // Match [_A-Za-z0-9]*, we have already matched [_A-Za-z$] 547 unsigned Size; 548 unsigned char C = *CurPtr++; 549 while (isIdentifierBody(C)) { 550 C = *CurPtr++; 551 } 552 --CurPtr; // Back up over the skipped character. 553 554 // Fast path, no $,\,? in identifier found. '\' might be an escaped newline 555 // or UCN, and ? might be a trigraph for '\', an escaped newline or UCN. 556 // FIXME: UCNs. 557 if (C != '\\' && C != '?' && (C != '$' || !Features.DollarIdents)) { 558FinishIdentifier: 559 const char *IdStart = BufferPtr; 560 FormTokenWithChars(Result, CurPtr, tok::identifier); 561 562 // If we are in raw mode, return this identifier raw. There is no need to 563 // look up identifier information or attempt to macro expand it. 564 if (LexingRawMode) return; 565 566 // Fill in Result.IdentifierInfo, looking up the identifier in the 567 // identifier table. 568 IdentifierInfo *II = PP->LookUpIdentifierInfo(Result, IdStart); 569 570 // Change the kind of this identifier to the appropriate token kind, e.g. 571 // turning "for" into a keyword. 572 Result.setKind(II->getTokenID()); 573 574 // Finally, now that we know we have an identifier, pass this off to the 575 // preprocessor, which may macro expand it or something. 576 if (II->isHandleIdentifierCase()) 577 PP->HandleIdentifier(Result); 578 return; 579 } 580 581 // Otherwise, $,\,? in identifier found. Enter slower path. 582 583 C = getCharAndSize(CurPtr, Size); 584 while (1) { 585 if (C == '$') { 586 // If we hit a $ and they are not supported in identifiers, we are done. 587 if (!Features.DollarIdents) goto FinishIdentifier; 588 589 // Otherwise, emit a diagnostic and continue. 590 if (!isLexingRawMode()) 591 Diag(CurPtr, diag::ext_dollar_in_identifier); 592 CurPtr = ConsumeChar(CurPtr, Size, Result); 593 C = getCharAndSize(CurPtr, Size); 594 continue; 595 } else if (!isIdentifierBody(C)) { // FIXME: UCNs. 596 // Found end of identifier. 597 goto FinishIdentifier; 598 } 599 600 // Otherwise, this character is good, consume it. 601 CurPtr = ConsumeChar(CurPtr, Size, Result); 602 603 C = getCharAndSize(CurPtr, Size); 604 while (isIdentifierBody(C)) { // FIXME: UCNs. 605 CurPtr = ConsumeChar(CurPtr, Size, Result); 606 C = getCharAndSize(CurPtr, Size); 607 } 608 } 609} 610 611 612/// LexNumericConstant - Lex the remainder of a integer or floating point 613/// constant. From[-1] is the first character lexed. Return the end of the 614/// constant. 615void Lexer::LexNumericConstant(Token &Result, const char *CurPtr) { 616 unsigned Size; 617 char C = getCharAndSize(CurPtr, Size); 618 char PrevCh = 0; 619 while (isNumberBody(C)) { // FIXME: UCNs? 620 CurPtr = ConsumeChar(CurPtr, Size, Result); 621 PrevCh = C; 622 C = getCharAndSize(CurPtr, Size); 623 } 624 625 // If we fell out, check for a sign, due to 1e+12. If we have one, continue. 626 if ((C == '-' || C == '+') && (PrevCh == 'E' || PrevCh == 'e')) 627 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result)); 628 629 // If we have a hex FP constant, continue. 630 if ((C == '-' || C == '+') && (PrevCh == 'P' || PrevCh == 'p') && 631 (Features.HexFloats || !Features.NoExtensions)) 632 return LexNumericConstant(Result, ConsumeChar(CurPtr, Size, Result)); 633 634 // Update the location of token as well as BufferPtr. 635 const char *TokStart = BufferPtr; 636 FormTokenWithChars(Result, CurPtr, tok::numeric_constant); 637 Result.setLiteralData(TokStart); 638} 639 640/// LexStringLiteral - Lex the remainder of a string literal, after having lexed 641/// either " or L". 642void Lexer::LexStringLiteral(Token &Result, const char *CurPtr, bool Wide) { 643 const char *NulCharacter = 0; // Does this string contain the \0 character? 644 645 char C = getAndAdvanceChar(CurPtr, Result); 646 while (C != '"') { 647 // Skip escaped characters. 648 if (C == '\\') { 649 // Skip the escaped character. 650 C = getAndAdvanceChar(CurPtr, Result); 651 } else if (C == '\n' || C == '\r' || // Newline. 652 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file. 653 if (!isLexingRawMode() && !Features.AsmPreprocessor) 654 Diag(BufferPtr, diag::err_unterminated_string); 655 FormTokenWithChars(Result, CurPtr-1, tok::unknown); 656 return; 657 } else if (C == 0) { 658 NulCharacter = CurPtr-1; 659 } 660 C = getAndAdvanceChar(CurPtr, Result); 661 } 662 663 // If a nul character existed in the string, warn about it. 664 if (NulCharacter && !isLexingRawMode()) 665 Diag(NulCharacter, diag::null_in_string); 666 667 // Update the location of the token as well as the BufferPtr instance var. 668 const char *TokStart = BufferPtr; 669 FormTokenWithChars(Result, CurPtr, 670 Wide ? tok::wide_string_literal : tok::string_literal); 671 Result.setLiteralData(TokStart); 672} 673 674/// LexAngledStringLiteral - Lex the remainder of an angled string literal, 675/// after having lexed the '<' character. This is used for #include filenames. 676void Lexer::LexAngledStringLiteral(Token &Result, const char *CurPtr) { 677 const char *NulCharacter = 0; // Does this string contain the \0 character? 678 const char *AfterLessPos = CurPtr; 679 char C = getAndAdvanceChar(CurPtr, Result); 680 while (C != '>') { 681 // Skip escaped characters. 682 if (C == '\\') { 683 // Skip the escaped character. 684 C = getAndAdvanceChar(CurPtr, Result); 685 } else if (C == '\n' || C == '\r' || // Newline. 686 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file. 687 // If the filename is unterminated, then it must just be a lone < 688 // character. Return this as such. 689 FormTokenWithChars(Result, AfterLessPos, tok::less); 690 return; 691 } else if (C == 0) { 692 NulCharacter = CurPtr-1; 693 } 694 C = getAndAdvanceChar(CurPtr, Result); 695 } 696 697 // If a nul character existed in the string, warn about it. 698 if (NulCharacter && !isLexingRawMode()) 699 Diag(NulCharacter, diag::null_in_string); 700 701 // Update the location of token as well as BufferPtr. 702 const char *TokStart = BufferPtr; 703 FormTokenWithChars(Result, CurPtr, tok::angle_string_literal); 704 Result.setLiteralData(TokStart); 705} 706 707 708/// LexCharConstant - Lex the remainder of a character constant, after having 709/// lexed either ' or L'. 710void Lexer::LexCharConstant(Token &Result, const char *CurPtr) { 711 const char *NulCharacter = 0; // Does this character contain the \0 character? 712 713 // Handle the common case of 'x' and '\y' efficiently. 714 char C = getAndAdvanceChar(CurPtr, Result); 715 if (C == '\'') { 716 if (!isLexingRawMode() && !Features.AsmPreprocessor) 717 Diag(BufferPtr, diag::err_empty_character); 718 FormTokenWithChars(Result, CurPtr, tok::unknown); 719 return; 720 } else if (C == '\\') { 721 // Skip the escaped character. 722 // FIXME: UCN's. 723 C = getAndAdvanceChar(CurPtr, Result); 724 } 725 726 if (C && C != '\n' && C != '\r' && CurPtr[0] == '\'') { 727 ++CurPtr; 728 } else { 729 // Fall back on generic code for embedded nulls, newlines, wide chars. 730 do { 731 // Skip escaped characters. 732 if (C == '\\') { 733 // Skip the escaped character. 734 C = getAndAdvanceChar(CurPtr, Result); 735 } else if (C == '\n' || C == '\r' || // Newline. 736 (C == 0 && CurPtr-1 == BufferEnd)) { // End of file. 737 if (!isLexingRawMode() && !Features.AsmPreprocessor) 738 Diag(BufferPtr, diag::err_unterminated_char); 739 FormTokenWithChars(Result, CurPtr-1, tok::unknown); 740 return; 741 } else if (C == 0) { 742 NulCharacter = CurPtr-1; 743 } 744 C = getAndAdvanceChar(CurPtr, Result); 745 } while (C != '\''); 746 } 747 748 if (NulCharacter && !isLexingRawMode()) 749 Diag(NulCharacter, diag::null_in_char); 750 751 // Update the location of token as well as BufferPtr. 752 const char *TokStart = BufferPtr; 753 FormTokenWithChars(Result, CurPtr, tok::char_constant); 754 Result.setLiteralData(TokStart); 755} 756 757/// SkipWhitespace - Efficiently skip over a series of whitespace characters. 758/// Update BufferPtr to point to the next non-whitespace character and return. 759/// 760/// This method forms a token and returns true if KeepWhitespaceMode is enabled. 761/// 762bool Lexer::SkipWhitespace(Token &Result, const char *CurPtr) { 763 // Whitespace - Skip it, then return the token after the whitespace. 764 unsigned char Char = *CurPtr; // Skip consequtive spaces efficiently. 765 while (1) { 766 // Skip horizontal whitespace very aggressively. 767 while (isHorizontalWhitespace(Char)) 768 Char = *++CurPtr; 769 770 // Otherwise if we have something other than whitespace, we're done. 771 if (Char != '\n' && Char != '\r') 772 break; 773 774 if (ParsingPreprocessorDirective) { 775 // End of preprocessor directive line, let LexTokenInternal handle this. 776 BufferPtr = CurPtr; 777 return false; 778 } 779 780 // ok, but handle newline. 781 // The returned token is at the start of the line. 782 Result.setFlag(Token::StartOfLine); 783 // No leading whitespace seen so far. 784 Result.clearFlag(Token::LeadingSpace); 785 Char = *++CurPtr; 786 } 787 788 // If this isn't immediately after a newline, there is leading space. 789 char PrevChar = CurPtr[-1]; 790 if (PrevChar != '\n' && PrevChar != '\r') 791 Result.setFlag(Token::LeadingSpace); 792 793 // If the client wants us to return whitespace, return it now. 794 if (isKeepWhitespaceMode()) { 795 FormTokenWithChars(Result, CurPtr, tok::unknown); 796 return true; 797 } 798 799 BufferPtr = CurPtr; 800 return false; 801} 802 803// SkipBCPLComment - We have just read the // characters from input. Skip until 804// we find the newline character thats terminate the comment. Then update 805/// BufferPtr and return. If we're in KeepCommentMode, this will form the token 806/// and return true. 807bool Lexer::SkipBCPLComment(Token &Result, const char *CurPtr) { 808 // If BCPL comments aren't explicitly enabled for this language, emit an 809 // extension warning. 810 if (!Features.BCPLComment && !isLexingRawMode()) { 811 Diag(BufferPtr, diag::ext_bcpl_comment); 812 813 // Mark them enabled so we only emit one warning for this translation 814 // unit. 815 Features.BCPLComment = true; 816 } 817 818 // Scan over the body of the comment. The common case, when scanning, is that 819 // the comment contains normal ascii characters with nothing interesting in 820 // them. As such, optimize for this case with the inner loop. 821 char C; 822 do { 823 C = *CurPtr; 824 // FIXME: Speedup BCPL comment lexing. Just scan for a \n or \r character. 825 // If we find a \n character, scan backwards, checking to see if it's an 826 // escaped newline, like we do for block comments. 827 828 // Skip over characters in the fast loop. 829 while (C != 0 && // Potentially EOF. 830 C != '\\' && // Potentially escaped newline. 831 C != '?' && // Potentially trigraph. 832 C != '\n' && C != '\r') // Newline or DOS-style newline. 833 C = *++CurPtr; 834 835 // If this is a newline, we're done. 836 if (C == '\n' || C == '\r') 837 break; // Found the newline? Break out! 838 839 // Otherwise, this is a hard case. Fall back on getAndAdvanceChar to 840 // properly decode the character. Read it in raw mode to avoid emitting 841 // diagnostics about things like trigraphs. If we see an escaped newline, 842 // we'll handle it below. 843 const char *OldPtr = CurPtr; 844 bool OldRawMode = isLexingRawMode(); 845 LexingRawMode = true; 846 C = getAndAdvanceChar(CurPtr, Result); 847 LexingRawMode = OldRawMode; 848 849 // If the char that we finally got was a \n, then we must have had something 850 // like \<newline><newline>. We don't want to have consumed the second 851 // newline, we want CurPtr, to end up pointing to it down below. 852 if (C == '\n' || C == '\r') { 853 --CurPtr; 854 C = 'x'; // doesn't matter what this is. 855 } 856 857 // If we read multiple characters, and one of those characters was a \r or 858 // \n, then we had an escaped newline within the comment. Emit diagnostic 859 // unless the next line is also a // comment. 860 if (CurPtr != OldPtr+1 && C != '/' && CurPtr[0] != '/') { 861 for (; OldPtr != CurPtr; ++OldPtr) 862 if (OldPtr[0] == '\n' || OldPtr[0] == '\r') { 863 // Okay, we found a // comment that ends in a newline, if the next 864 // line is also a // comment, but has spaces, don't emit a diagnostic. 865 if (isspace(C)) { 866 const char *ForwardPtr = CurPtr; 867 while (isspace(*ForwardPtr)) // Skip whitespace. 868 ++ForwardPtr; 869 if (ForwardPtr[0] == '/' && ForwardPtr[1] == '/') 870 break; 871 } 872 873 if (!isLexingRawMode()) 874 Diag(OldPtr-1, diag::ext_multi_line_bcpl_comment); 875 break; 876 } 877 } 878 879 if (CurPtr == BufferEnd+1) { --CurPtr; break; } 880 } while (C != '\n' && C != '\r'); 881 882 // Found but did not consume the newline. 883 884 // If we are returning comments as tokens, return this comment as a token. 885 if (inKeepCommentMode()) 886 return SaveBCPLComment(Result, CurPtr); 887 888 // If we are inside a preprocessor directive and we see the end of line, 889 // return immediately, so that the lexer can return this as an EOM token. 890 if (ParsingPreprocessorDirective || CurPtr == BufferEnd) { 891 BufferPtr = CurPtr; 892 return false; 893 } 894 895 // Otherwise, eat the \n character. We don't care if this is a \n\r or 896 // \r\n sequence. This is an efficiency hack (because we know the \n can't 897 // contribute to another token), it isn't needed for correctness. Note that 898 // this is ok even in KeepWhitespaceMode, because we would have returned the 899 /// comment above in that mode. 900 ++CurPtr; 901 902 // The next returned token is at the start of the line. 903 Result.setFlag(Token::StartOfLine); 904 // No leading whitespace seen so far. 905 Result.clearFlag(Token::LeadingSpace); 906 BufferPtr = CurPtr; 907 return false; 908} 909 910/// SaveBCPLComment - If in save-comment mode, package up this BCPL comment in 911/// an appropriate way and return it. 912bool Lexer::SaveBCPLComment(Token &Result, const char *CurPtr) { 913 // If we're not in a preprocessor directive, just return the // comment 914 // directly. 915 FormTokenWithChars(Result, CurPtr, tok::comment); 916 917 if (!ParsingPreprocessorDirective) 918 return true; 919 920 // If this BCPL-style comment is in a macro definition, transmogrify it into 921 // a C-style block comment. 922 std::string Spelling = PP->getSpelling(Result); 923 assert(Spelling[0] == '/' && Spelling[1] == '/' && "Not bcpl comment?"); 924 Spelling[1] = '*'; // Change prefix to "/*". 925 Spelling += "*/"; // add suffix. 926 927 Result.setKind(tok::comment); 928 PP->CreateString(&Spelling[0], Spelling.size(), Result, 929 Result.getLocation()); 930 return true; 931} 932 933/// isBlockCommentEndOfEscapedNewLine - Return true if the specified newline 934/// character (either \n or \r) is part of an escaped newline sequence. Issue a 935/// diagnostic if so. We know that the newline is inside of a block comment. 936static bool isEndOfBlockCommentWithEscapedNewLine(const char *CurPtr, 937 Lexer *L) { 938 assert(CurPtr[0] == '\n' || CurPtr[0] == '\r'); 939 940 // Back up off the newline. 941 --CurPtr; 942 943 // If this is a two-character newline sequence, skip the other character. 944 if (CurPtr[0] == '\n' || CurPtr[0] == '\r') { 945 // \n\n or \r\r -> not escaped newline. 946 if (CurPtr[0] == CurPtr[1]) 947 return false; 948 // \n\r or \r\n -> skip the newline. 949 --CurPtr; 950 } 951 952 // If we have horizontal whitespace, skip over it. We allow whitespace 953 // between the slash and newline. 954 bool HasSpace = false; 955 while (isHorizontalWhitespace(*CurPtr) || *CurPtr == 0) { 956 --CurPtr; 957 HasSpace = true; 958 } 959 960 // If we have a slash, we know this is an escaped newline. 961 if (*CurPtr == '\\') { 962 if (CurPtr[-1] != '*') return false; 963 } else { 964 // It isn't a slash, is it the ?? / trigraph? 965 if (CurPtr[0] != '/' || CurPtr[-1] != '?' || CurPtr[-2] != '?' || 966 CurPtr[-3] != '*') 967 return false; 968 969 // This is the trigraph ending the comment. Emit a stern warning! 970 CurPtr -= 2; 971 972 // If no trigraphs are enabled, warn that we ignored this trigraph and 973 // ignore this * character. 974 if (!L->getFeatures().Trigraphs) { 975 if (!L->isLexingRawMode()) 976 L->Diag(CurPtr, diag::trigraph_ignored_block_comment); 977 return false; 978 } 979 if (!L->isLexingRawMode()) 980 L->Diag(CurPtr, diag::trigraph_ends_block_comment); 981 } 982 983 // Warn about having an escaped newline between the */ characters. 984 if (!L->isLexingRawMode()) 985 L->Diag(CurPtr, diag::escaped_newline_block_comment_end); 986 987 // If there was space between the backslash and newline, warn about it. 988 if (HasSpace && !L->isLexingRawMode()) 989 L->Diag(CurPtr, diag::backslash_newline_space); 990 991 return true; 992} 993 994#ifdef __SSE2__ 995#include <emmintrin.h> 996#elif __ALTIVEC__ 997#include <altivec.h> 998#undef bool 999#endif 1000 1001/// SkipBlockComment - We have just read the /* characters from input. Read 1002/// until we find the */ characters that terminate the comment. Note that we 1003/// don't bother decoding trigraphs or escaped newlines in block comments, 1004/// because they cannot cause the comment to end. The only thing that can 1005/// happen is the comment could end with an escaped newline between the */ end 1006/// of comment. 1007/// 1008/// If KeepCommentMode is enabled, this forms a token from the comment and 1009/// returns true. 1010bool Lexer::SkipBlockComment(Token &Result, const char *CurPtr) { 1011 // Scan one character past where we should, looking for a '/' character. Once 1012 // we find it, check to see if it was preceeded by a *. This common 1013 // optimization helps people who like to put a lot of * characters in their 1014 // comments. 1015 1016 // The first character we get with newlines and trigraphs skipped to handle 1017 // the degenerate /*/ case below correctly if the * has an escaped newline 1018 // after it. 1019 unsigned CharSize; 1020 unsigned char C = getCharAndSize(CurPtr, CharSize); 1021 CurPtr += CharSize; 1022 if (C == 0 && CurPtr == BufferEnd+1) { 1023 if (!isLexingRawMode()) 1024 Diag(BufferPtr, diag::err_unterminated_block_comment); 1025 --CurPtr; 1026 1027 // KeepWhitespaceMode should return this broken comment as a token. Since 1028 // it isn't a well formed comment, just return it as an 'unknown' token. 1029 if (isKeepWhitespaceMode()) { 1030 FormTokenWithChars(Result, CurPtr, tok::unknown); 1031 return true; 1032 } 1033 1034 BufferPtr = CurPtr; 1035 return false; 1036 } 1037 1038 // Check to see if the first character after the '/*' is another /. If so, 1039 // then this slash does not end the block comment, it is part of it. 1040 if (C == '/') 1041 C = *CurPtr++; 1042 1043 while (1) { 1044 // Skip over all non-interesting characters until we find end of buffer or a 1045 // (probably ending) '/' character. 1046 if (CurPtr + 24 < BufferEnd) { 1047 // While not aligned to a 16-byte boundary. 1048 while (C != '/' && ((intptr_t)CurPtr & 0x0F) != 0) 1049 C = *CurPtr++; 1050 1051 if (C == '/') goto FoundSlash; 1052 1053#ifdef __SSE2__ 1054 __m128i Slashes = _mm_set_epi8('/', '/', '/', '/', '/', '/', '/', '/', 1055 '/', '/', '/', '/', '/', '/', '/', '/'); 1056 while (CurPtr+16 <= BufferEnd && 1057 _mm_movemask_epi8(_mm_cmpeq_epi8(*(__m128i*)CurPtr, Slashes)) == 0) 1058 CurPtr += 16; 1059#elif __ALTIVEC__ 1060 __vector unsigned char Slashes = { 1061 '/', '/', '/', '/', '/', '/', '/', '/', 1062 '/', '/', '/', '/', '/', '/', '/', '/' 1063 }; 1064 while (CurPtr+16 <= BufferEnd && 1065 !vec_any_eq(*(vector unsigned char*)CurPtr, Slashes)) 1066 CurPtr += 16; 1067#else 1068 // Scan for '/' quickly. Many block comments are very large. 1069 while (CurPtr[0] != '/' && 1070 CurPtr[1] != '/' && 1071 CurPtr[2] != '/' && 1072 CurPtr[3] != '/' && 1073 CurPtr+4 < BufferEnd) { 1074 CurPtr += 4; 1075 } 1076#endif 1077 1078 // It has to be one of the bytes scanned, increment to it and read one. 1079 C = *CurPtr++; 1080 } 1081 1082 // Loop to scan the remainder. 1083 while (C != '/' && C != '\0') 1084 C = *CurPtr++; 1085 1086 FoundSlash: 1087 if (C == '/') { 1088 if (CurPtr[-2] == '*') // We found the final */. We're done! 1089 break; 1090 1091 if ((CurPtr[-2] == '\n' || CurPtr[-2] == '\r')) { 1092 if (isEndOfBlockCommentWithEscapedNewLine(CurPtr-2, this)) { 1093 // We found the final */, though it had an escaped newline between the 1094 // * and /. We're done! 1095 break; 1096 } 1097 } 1098 if (CurPtr[0] == '*' && CurPtr[1] != '/') { 1099 // If this is a /* inside of the comment, emit a warning. Don't do this 1100 // if this is a /*/, which will end the comment. This misses cases with 1101 // embedded escaped newlines, but oh well. 1102 if (!isLexingRawMode()) 1103 Diag(CurPtr-1, diag::warn_nested_block_comment); 1104 } 1105 } else if (C == 0 && CurPtr == BufferEnd+1) { 1106 if (!isLexingRawMode()) 1107 Diag(BufferPtr, diag::err_unterminated_block_comment); 1108 // Note: the user probably forgot a */. We could continue immediately 1109 // after the /*, but this would involve lexing a lot of what really is the 1110 // comment, which surely would confuse the parser. 1111 --CurPtr; 1112 1113 // KeepWhitespaceMode should return this broken comment as a token. Since 1114 // it isn't a well formed comment, just return it as an 'unknown' token. 1115 if (isKeepWhitespaceMode()) { 1116 FormTokenWithChars(Result, CurPtr, tok::unknown); 1117 return true; 1118 } 1119 1120 BufferPtr = CurPtr; 1121 return false; 1122 } 1123 C = *CurPtr++; 1124 } 1125 1126 // If we are returning comments as tokens, return this comment as a token. 1127 if (inKeepCommentMode()) { 1128 FormTokenWithChars(Result, CurPtr, tok::comment); 1129 return true; 1130 } 1131 1132 // It is common for the tokens immediately after a /**/ comment to be 1133 // whitespace. Instead of going through the big switch, handle it 1134 // efficiently now. This is safe even in KeepWhitespaceMode because we would 1135 // have already returned above with the comment as a token. 1136 if (isHorizontalWhitespace(*CurPtr)) { 1137 Result.setFlag(Token::LeadingSpace); 1138 SkipWhitespace(Result, CurPtr+1); 1139 return false; 1140 } 1141 1142 // Otherwise, just return so that the next character will be lexed as a token. 1143 BufferPtr = CurPtr; 1144 Result.setFlag(Token::LeadingSpace); 1145 return false; 1146} 1147 1148//===----------------------------------------------------------------------===// 1149// Primary Lexing Entry Points 1150//===----------------------------------------------------------------------===// 1151 1152/// ReadToEndOfLine - Read the rest of the current preprocessor line as an 1153/// uninterpreted string. This switches the lexer out of directive mode. 1154std::string Lexer::ReadToEndOfLine() { 1155 assert(ParsingPreprocessorDirective && ParsingFilename == false && 1156 "Must be in a preprocessing directive!"); 1157 std::string Result; 1158 Token Tmp; 1159 1160 // CurPtr - Cache BufferPtr in an automatic variable. 1161 const char *CurPtr = BufferPtr; 1162 while (1) { 1163 char Char = getAndAdvanceChar(CurPtr, Tmp); 1164 switch (Char) { 1165 default: 1166 Result += Char; 1167 break; 1168 case 0: // Null. 1169 // Found end of file? 1170 if (CurPtr-1 != BufferEnd) { 1171 // Nope, normal character, continue. 1172 Result += Char; 1173 break; 1174 } 1175 // FALL THROUGH. 1176 case '\r': 1177 case '\n': 1178 // Okay, we found the end of the line. First, back up past the \0, \r, \n. 1179 assert(CurPtr[-1] == Char && "Trigraphs for newline?"); 1180 BufferPtr = CurPtr-1; 1181 1182 // Next, lex the character, which should handle the EOM transition. 1183 Lex(Tmp); 1184 assert(Tmp.is(tok::eom) && "Unexpected token!"); 1185 1186 // Finally, we're done, return the string we found. 1187 return Result; 1188 } 1189 } 1190} 1191 1192/// LexEndOfFile - CurPtr points to the end of this file. Handle this 1193/// condition, reporting diagnostics and handling other edge cases as required. 1194/// This returns true if Result contains a token, false if PP.Lex should be 1195/// called again. 1196bool Lexer::LexEndOfFile(Token &Result, const char *CurPtr) { 1197 // If we hit the end of the file while parsing a preprocessor directive, 1198 // end the preprocessor directive first. The next token returned will 1199 // then be the end of file. 1200 if (ParsingPreprocessorDirective) { 1201 // Done parsing the "line". 1202 ParsingPreprocessorDirective = false; 1203 // Update the location of token as well as BufferPtr. 1204 FormTokenWithChars(Result, CurPtr, tok::eom); 1205 1206 // Restore comment saving mode, in case it was disabled for directive. 1207 SetCommentRetentionState(PP->getCommentRetentionState()); 1208 return true; // Have a token. 1209 } 1210 1211 // If we are in raw mode, return this event as an EOF token. Let the caller 1212 // that put us in raw mode handle the event. 1213 if (isLexingRawMode()) { 1214 Result.startToken(); 1215 BufferPtr = BufferEnd; 1216 FormTokenWithChars(Result, BufferEnd, tok::eof); 1217 return true; 1218 } 1219 1220 // Otherwise, issue diagnostics for unterminated #if and missing newline. 1221 1222 // If we are in a #if directive, emit an error. 1223 while (!ConditionalStack.empty()) { 1224 PP->Diag(ConditionalStack.back().IfLoc, 1225 diag::err_pp_unterminated_conditional); 1226 ConditionalStack.pop_back(); 1227 } 1228 1229 // C99 5.1.1.2p2: If the file is non-empty and didn't end in a newline, issue 1230 // a pedwarn. 1231 if (CurPtr != BufferStart && (CurPtr[-1] != '\n' && CurPtr[-1] != '\r')) 1232 Diag(BufferEnd, diag::ext_no_newline_eof) 1233 << CodeModificationHint::CreateInsertion(getSourceLocation(BufferEnd), 1234 "\n"); 1235 1236 BufferPtr = CurPtr; 1237 1238 // Finally, let the preprocessor handle this. 1239 return PP->HandleEndOfFile(Result); 1240} 1241 1242/// isNextPPTokenLParen - Return 1 if the next unexpanded token lexed from 1243/// the specified lexer will return a tok::l_paren token, 0 if it is something 1244/// else and 2 if there are no more tokens in the buffer controlled by the 1245/// lexer. 1246unsigned Lexer::isNextPPTokenLParen() { 1247 assert(!LexingRawMode && "How can we expand a macro from a skipping buffer?"); 1248 1249 // Switch to 'skipping' mode. This will ensure that we can lex a token 1250 // without emitting diagnostics, disables macro expansion, and will cause EOF 1251 // to return an EOF token instead of popping the include stack. 1252 LexingRawMode = true; 1253 1254 // Save state that can be changed while lexing so that we can restore it. 1255 const char *TmpBufferPtr = BufferPtr; 1256 1257 Token Tok; 1258 Tok.startToken(); 1259 LexTokenInternal(Tok); 1260 1261 // Restore state that may have changed. 1262 BufferPtr = TmpBufferPtr; 1263 1264 // Restore the lexer back to non-skipping mode. 1265 LexingRawMode = false; 1266 1267 if (Tok.is(tok::eof)) 1268 return 2; 1269 return Tok.is(tok::l_paren); 1270} 1271 1272 1273/// LexTokenInternal - This implements a simple C family lexer. It is an 1274/// extremely performance critical piece of code. This assumes that the buffer 1275/// has a null character at the end of the file. Return true if an error 1276/// occurred and compilation should terminate, false if normal. This returns a 1277/// preprocessing token, not a normal token, as such, it is an internal 1278/// interface. It assumes that the Flags of result have been cleared before 1279/// calling this. 1280void Lexer::LexTokenInternal(Token &Result) { 1281LexNextToken: 1282 // New token, can't need cleaning yet. 1283 Result.clearFlag(Token::NeedsCleaning); 1284 Result.setIdentifierInfo(0); 1285 1286 // CurPtr - Cache BufferPtr in an automatic variable. 1287 const char *CurPtr = BufferPtr; 1288 1289 // Small amounts of horizontal whitespace is very common between tokens. 1290 if ((*CurPtr == ' ') || (*CurPtr == '\t')) { 1291 ++CurPtr; 1292 while ((*CurPtr == ' ') || (*CurPtr == '\t')) 1293 ++CurPtr; 1294 1295 // If we are keeping whitespace and other tokens, just return what we just 1296 // skipped. The next lexer invocation will return the token after the 1297 // whitespace. 1298 if (isKeepWhitespaceMode()) { 1299 FormTokenWithChars(Result, CurPtr, tok::unknown); 1300 return; 1301 } 1302 1303 BufferPtr = CurPtr; 1304 Result.setFlag(Token::LeadingSpace); 1305 } 1306 1307 unsigned SizeTmp, SizeTmp2; // Temporaries for use in cases below. 1308 1309 // Read a character, advancing over it. 1310 char Char = getAndAdvanceChar(CurPtr, Result); 1311 tok::TokenKind Kind; 1312 1313 switch (Char) { 1314 case 0: // Null. 1315 // Found end of file? 1316 if (CurPtr-1 == BufferEnd) { 1317 // Read the PP instance variable into an automatic variable, because 1318 // LexEndOfFile will often delete 'this'. 1319 Preprocessor *PPCache = PP; 1320 if (LexEndOfFile(Result, CurPtr-1)) // Retreat back into the file. 1321 return; // Got a token to return. 1322 assert(PPCache && "Raw buffer::LexEndOfFile should return a token"); 1323 return PPCache->Lex(Result); 1324 } 1325 1326 if (!isLexingRawMode()) 1327 Diag(CurPtr-1, diag::null_in_file); 1328 Result.setFlag(Token::LeadingSpace); 1329 if (SkipWhitespace(Result, CurPtr)) 1330 return; // KeepWhitespaceMode 1331 1332 goto LexNextToken; // GCC isn't tail call eliminating. 1333 case '\n': 1334 case '\r': 1335 // If we are inside a preprocessor directive and we see the end of line, 1336 // we know we are done with the directive, so return an EOM token. 1337 if (ParsingPreprocessorDirective) { 1338 // Done parsing the "line". 1339 ParsingPreprocessorDirective = false; 1340 1341 // Restore comment saving mode, in case it was disabled for directive. 1342 SetCommentRetentionState(PP->getCommentRetentionState()); 1343 1344 // Since we consumed a newline, we are back at the start of a line. 1345 IsAtStartOfLine = true; 1346 1347 Kind = tok::eom; 1348 break; 1349 } 1350 // The returned token is at the start of the line. 1351 Result.setFlag(Token::StartOfLine); 1352 // No leading whitespace seen so far. 1353 Result.clearFlag(Token::LeadingSpace); 1354 1355 if (SkipWhitespace(Result, CurPtr)) 1356 return; // KeepWhitespaceMode 1357 goto LexNextToken; // GCC isn't tail call eliminating. 1358 case ' ': 1359 case '\t': 1360 case '\f': 1361 case '\v': 1362 SkipHorizontalWhitespace: 1363 Result.setFlag(Token::LeadingSpace); 1364 if (SkipWhitespace(Result, CurPtr)) 1365 return; // KeepWhitespaceMode 1366 1367 SkipIgnoredUnits: 1368 CurPtr = BufferPtr; 1369 1370 // If the next token is obviously a // or /* */ comment, skip it efficiently 1371 // too (without going through the big switch stmt). 1372 if (CurPtr[0] == '/' && CurPtr[1] == '/' && !inKeepCommentMode() && 1373 Features.BCPLComment) { 1374 SkipBCPLComment(Result, CurPtr+2); 1375 goto SkipIgnoredUnits; 1376 } else if (CurPtr[0] == '/' && CurPtr[1] == '*' && !inKeepCommentMode()) { 1377 SkipBlockComment(Result, CurPtr+2); 1378 goto SkipIgnoredUnits; 1379 } else if (isHorizontalWhitespace(*CurPtr)) { 1380 goto SkipHorizontalWhitespace; 1381 } 1382 goto LexNextToken; // GCC isn't tail call eliminating. 1383 1384 // C99 6.4.4.1: Integer Constants. 1385 // C99 6.4.4.2: Floating Constants. 1386 case '0': case '1': case '2': case '3': case '4': 1387 case '5': case '6': case '7': case '8': case '9': 1388 // Notify MIOpt that we read a non-whitespace/non-comment token. 1389 MIOpt.ReadToken(); 1390 return LexNumericConstant(Result, CurPtr); 1391 1392 case 'L': // Identifier (Loony) or wide literal (L'x' or L"xyz"). 1393 // Notify MIOpt that we read a non-whitespace/non-comment token. 1394 MIOpt.ReadToken(); 1395 Char = getCharAndSize(CurPtr, SizeTmp); 1396 1397 // Wide string literal. 1398 if (Char == '"') 1399 return LexStringLiteral(Result, ConsumeChar(CurPtr, SizeTmp, Result), 1400 true); 1401 1402 // Wide character constant. 1403 if (Char == '\'') 1404 return LexCharConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result)); 1405 // FALL THROUGH, treating L like the start of an identifier. 1406 1407 // C99 6.4.2: Identifiers. 1408 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G': 1409 case 'H': case 'I': case 'J': case 'K': /*'L'*/case 'M': case 'N': 1410 case 'O': case 'P': case 'Q': case 'R': case 'S': case 'T': case 'U': 1411 case 'V': case 'W': case 'X': case 'Y': case 'Z': 1412 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g': 1413 case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n': 1414 case 'o': case 'p': case 'q': case 'r': case 's': case 't': case 'u': 1415 case 'v': case 'w': case 'x': case 'y': case 'z': 1416 case '_': 1417 // Notify MIOpt that we read a non-whitespace/non-comment token. 1418 MIOpt.ReadToken(); 1419 return LexIdentifier(Result, CurPtr); 1420 1421 case '$': // $ in identifiers. 1422 if (Features.DollarIdents) { 1423 if (!isLexingRawMode()) 1424 Diag(CurPtr-1, diag::ext_dollar_in_identifier); 1425 // Notify MIOpt that we read a non-whitespace/non-comment token. 1426 MIOpt.ReadToken(); 1427 return LexIdentifier(Result, CurPtr); 1428 } 1429 1430 Kind = tok::unknown; 1431 break; 1432 1433 // C99 6.4.4: Character Constants. 1434 case '\'': 1435 // Notify MIOpt that we read a non-whitespace/non-comment token. 1436 MIOpt.ReadToken(); 1437 return LexCharConstant(Result, CurPtr); 1438 1439 // C99 6.4.5: String Literals. 1440 case '"': 1441 // Notify MIOpt that we read a non-whitespace/non-comment token. 1442 MIOpt.ReadToken(); 1443 return LexStringLiteral(Result, CurPtr, false); 1444 1445 // C99 6.4.6: Punctuators. 1446 case '?': 1447 Kind = tok::question; 1448 break; 1449 case '[': 1450 Kind = tok::l_square; 1451 break; 1452 case ']': 1453 Kind = tok::r_square; 1454 break; 1455 case '(': 1456 Kind = tok::l_paren; 1457 break; 1458 case ')': 1459 Kind = tok::r_paren; 1460 break; 1461 case '{': 1462 Kind = tok::l_brace; 1463 break; 1464 case '}': 1465 Kind = tok::r_brace; 1466 break; 1467 case '.': 1468 Char = getCharAndSize(CurPtr, SizeTmp); 1469 if (Char >= '0' && Char <= '9') { 1470 // Notify MIOpt that we read a non-whitespace/non-comment token. 1471 MIOpt.ReadToken(); 1472 1473 return LexNumericConstant(Result, ConsumeChar(CurPtr, SizeTmp, Result)); 1474 } else if (Features.CPlusPlus && Char == '*') { 1475 Kind = tok::periodstar; 1476 CurPtr += SizeTmp; 1477 } else if (Char == '.' && 1478 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '.') { 1479 Kind = tok::ellipsis; 1480 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result), 1481 SizeTmp2, Result); 1482 } else { 1483 Kind = tok::period; 1484 } 1485 break; 1486 case '&': 1487 Char = getCharAndSize(CurPtr, SizeTmp); 1488 if (Char == '&') { 1489 Kind = tok::ampamp; 1490 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1491 } else if (Char == '=') { 1492 Kind = tok::ampequal; 1493 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1494 } else { 1495 Kind = tok::amp; 1496 } 1497 break; 1498 case '*': 1499 if (getCharAndSize(CurPtr, SizeTmp) == '=') { 1500 Kind = tok::starequal; 1501 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1502 } else { 1503 Kind = tok::star; 1504 } 1505 break; 1506 case '+': 1507 Char = getCharAndSize(CurPtr, SizeTmp); 1508 if (Char == '+') { 1509 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1510 Kind = tok::plusplus; 1511 } else if (Char == '=') { 1512 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1513 Kind = tok::plusequal; 1514 } else { 1515 Kind = tok::plus; 1516 } 1517 break; 1518 case '-': 1519 Char = getCharAndSize(CurPtr, SizeTmp); 1520 if (Char == '-') { // -- 1521 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1522 Kind = tok::minusminus; 1523 } else if (Char == '>' && Features.CPlusPlus && 1524 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '*') { // C++ ->* 1525 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result), 1526 SizeTmp2, Result); 1527 Kind = tok::arrowstar; 1528 } else if (Char == '>') { // -> 1529 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1530 Kind = tok::arrow; 1531 } else if (Char == '=') { // -= 1532 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1533 Kind = tok::minusequal; 1534 } else { 1535 Kind = tok::minus; 1536 } 1537 break; 1538 case '~': 1539 Kind = tok::tilde; 1540 break; 1541 case '!': 1542 if (getCharAndSize(CurPtr, SizeTmp) == '=') { 1543 Kind = tok::exclaimequal; 1544 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1545 } else { 1546 Kind = tok::exclaim; 1547 } 1548 break; 1549 case '/': 1550 // 6.4.9: Comments 1551 Char = getCharAndSize(CurPtr, SizeTmp); 1552 if (Char == '/') { // BCPL comment. 1553 // Even if BCPL comments are disabled (e.g. in C89 mode), we generally 1554 // want to lex this as a comment. There is one problem with this though, 1555 // that in one particular corner case, this can change the behavior of the 1556 // resultant program. For example, In "foo //**/ bar", C89 would lex 1557 // this as "foo / bar" and langauges with BCPL comments would lex it as 1558 // "foo". Check to see if the character after the second slash is a '*'. 1559 // If so, we will lex that as a "/" instead of the start of a comment. 1560 if (Features.BCPLComment || 1561 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) != '*') { 1562 if (SkipBCPLComment(Result, ConsumeChar(CurPtr, SizeTmp, Result))) 1563 return; // KeepCommentMode 1564 1565 // It is common for the tokens immediately after a // comment to be 1566 // whitespace (indentation for the next line). Instead of going through 1567 // the big switch, handle it efficiently now. 1568 goto SkipIgnoredUnits; 1569 } 1570 } 1571 1572 if (Char == '*') { // /**/ comment. 1573 if (SkipBlockComment(Result, ConsumeChar(CurPtr, SizeTmp, Result))) 1574 return; // KeepCommentMode 1575 goto LexNextToken; // GCC isn't tail call eliminating. 1576 } 1577 1578 if (Char == '=') { 1579 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1580 Kind = tok::slashequal; 1581 } else { 1582 Kind = tok::slash; 1583 } 1584 break; 1585 case '%': 1586 Char = getCharAndSize(CurPtr, SizeTmp); 1587 if (Char == '=') { 1588 Kind = tok::percentequal; 1589 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1590 } else if (Features.Digraphs && Char == '>') { 1591 Kind = tok::r_brace; // '%>' -> '}' 1592 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1593 } else if (Features.Digraphs && Char == ':') { 1594 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1595 Char = getCharAndSize(CurPtr, SizeTmp); 1596 if (Char == '%' && getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == ':') { 1597 Kind = tok::hashhash; // '%:%:' -> '##' 1598 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result), 1599 SizeTmp2, Result); 1600 } else if (Char == '@' && Features.Microsoft) { // %:@ -> #@ -> Charize 1601 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1602 if (!isLexingRawMode()) 1603 Diag(BufferPtr, diag::charize_microsoft_ext); 1604 Kind = tok::hashat; 1605 } else { // '%:' -> '#' 1606 // We parsed a # character. If this occurs at the start of the line, 1607 // it's actually the start of a preprocessing directive. Callback to 1608 // the preprocessor to handle it. 1609 // FIXME: -fpreprocessed mode?? 1610 if (Result.isAtStartOfLine() && !LexingRawMode) { 1611 FormTokenWithChars(Result, CurPtr, tok::hash); 1612 PP->HandleDirective(Result); 1613 1614 // As an optimization, if the preprocessor didn't switch lexers, tail 1615 // recurse. 1616 if (PP->isCurrentLexer(this)) { 1617 // Start a new token. If this is a #include or something, the PP may 1618 // want us starting at the beginning of the line again. If so, set 1619 // the StartOfLine flag. 1620 if (IsAtStartOfLine) { 1621 Result.setFlag(Token::StartOfLine); 1622 IsAtStartOfLine = false; 1623 } 1624 goto LexNextToken; // GCC isn't tail call eliminating. 1625 } 1626 1627 return PP->Lex(Result); 1628 } 1629 1630 Kind = tok::hash; 1631 } 1632 } else { 1633 Kind = tok::percent; 1634 } 1635 break; 1636 case '<': 1637 Char = getCharAndSize(CurPtr, SizeTmp); 1638 if (ParsingFilename) { 1639 return LexAngledStringLiteral(Result, CurPtr); 1640 } else if (Char == '<' && 1641 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '=') { 1642 Kind = tok::lesslessequal; 1643 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result), 1644 SizeTmp2, Result); 1645 } else if (Char == '<') { 1646 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1647 Kind = tok::lessless; 1648 } else if (Char == '=') { 1649 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1650 Kind = tok::lessequal; 1651 } else if (Features.Digraphs && Char == ':') { // '<:' -> '[' 1652 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1653 Kind = tok::l_square; 1654 } else if (Features.Digraphs && Char == '%') { // '<%' -> '{' 1655 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1656 Kind = tok::l_brace; 1657 } else { 1658 Kind = tok::less; 1659 } 1660 break; 1661 case '>': 1662 Char = getCharAndSize(CurPtr, SizeTmp); 1663 if (Char == '=') { 1664 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1665 Kind = tok::greaterequal; 1666 } else if (Char == '>' && 1667 getCharAndSize(CurPtr+SizeTmp, SizeTmp2) == '=') { 1668 CurPtr = ConsumeChar(ConsumeChar(CurPtr, SizeTmp, Result), 1669 SizeTmp2, Result); 1670 Kind = tok::greatergreaterequal; 1671 } else if (Char == '>') { 1672 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1673 Kind = tok::greatergreater; 1674 } else { 1675 Kind = tok::greater; 1676 } 1677 break; 1678 case '^': 1679 Char = getCharAndSize(CurPtr, SizeTmp); 1680 if (Char == '=') { 1681 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1682 Kind = tok::caretequal; 1683 } else { 1684 Kind = tok::caret; 1685 } 1686 break; 1687 case '|': 1688 Char = getCharAndSize(CurPtr, SizeTmp); 1689 if (Char == '=') { 1690 Kind = tok::pipeequal; 1691 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1692 } else if (Char == '|') { 1693 Kind = tok::pipepipe; 1694 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1695 } else { 1696 Kind = tok::pipe; 1697 } 1698 break; 1699 case ':': 1700 Char = getCharAndSize(CurPtr, SizeTmp); 1701 if (Features.Digraphs && Char == '>') { 1702 Kind = tok::r_square; // ':>' -> ']' 1703 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1704 } else if (Features.CPlusPlus && Char == ':') { 1705 Kind = tok::coloncolon; 1706 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1707 } else { 1708 Kind = tok::colon; 1709 } 1710 break; 1711 case ';': 1712 Kind = tok::semi; 1713 break; 1714 case '=': 1715 Char = getCharAndSize(CurPtr, SizeTmp); 1716 if (Char == '=') { 1717 Kind = tok::equalequal; 1718 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1719 } else { 1720 Kind = tok::equal; 1721 } 1722 break; 1723 case ',': 1724 Kind = tok::comma; 1725 break; 1726 case '#': 1727 Char = getCharAndSize(CurPtr, SizeTmp); 1728 if (Char == '#') { 1729 Kind = tok::hashhash; 1730 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1731 } else if (Char == '@' && Features.Microsoft) { // #@ -> Charize 1732 Kind = tok::hashat; 1733 if (!isLexingRawMode()) 1734 Diag(BufferPtr, diag::charize_microsoft_ext); 1735 CurPtr = ConsumeChar(CurPtr, SizeTmp, Result); 1736 } else { 1737 // We parsed a # character. If this occurs at the start of the line, 1738 // it's actually the start of a preprocessing directive. Callback to 1739 // the preprocessor to handle it. 1740 // FIXME: -fpreprocessed mode?? 1741 if (Result.isAtStartOfLine() && !LexingRawMode) { 1742 FormTokenWithChars(Result, CurPtr, tok::hash); 1743 PP->HandleDirective(Result); 1744 1745 // As an optimization, if the preprocessor didn't switch lexers, tail 1746 // recurse. 1747 if (PP->isCurrentLexer(this)) { 1748 // Start a new token. If this is a #include or something, the PP may 1749 // want us starting at the beginning of the line again. If so, set 1750 // the StartOfLine flag. 1751 if (IsAtStartOfLine) { 1752 Result.setFlag(Token::StartOfLine); 1753 IsAtStartOfLine = false; 1754 } 1755 goto LexNextToken; // GCC isn't tail call eliminating. 1756 } 1757 return PP->Lex(Result); 1758 } 1759 1760 Kind = tok::hash; 1761 } 1762 break; 1763 1764 case '@': 1765 // Objective C support. 1766 if (CurPtr[-1] == '@' && Features.ObjC1) 1767 Kind = tok::at; 1768 else 1769 Kind = tok::unknown; 1770 break; 1771 1772 case '\\': 1773 // FIXME: UCN's. 1774 // FALL THROUGH. 1775 default: 1776 Kind = tok::unknown; 1777 break; 1778 } 1779 1780 // Notify MIOpt that we read a non-whitespace/non-comment token. 1781 MIOpt.ReadToken(); 1782 1783 // Update the location of token as well as BufferPtr. 1784 FormTokenWithChars(Result, CurPtr, Kind); 1785} 1786