1 //===--- TokenLexer.cpp - Lex from a token stream -------------------------===// 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 TokenLexer interface. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "clang/Lex/TokenLexer.h" 15 #include "MacroArgs.h" 16 #include "clang/Lex/MacroInfo.h" 17 #include "clang/Lex/Preprocessor.h" 18 #include "clang/Basic/SourceManager.h" 19 #include "clang/Lex/LexDiagnostic.h" 20 #include "llvm/ADT/SmallVector.h" 21 using namespace clang; 22 23 24 /// Create a TokenLexer for the specified macro with the specified actual 25 /// arguments. Note that this ctor takes ownership of the ActualArgs pointer. 26 void TokenLexer::Init(Token &Tok, SourceLocation ELEnd, MacroArgs *Actuals) { 27 // If the client is reusing a TokenLexer, make sure to free any memory 28 // associated with it. 29 destroy(); 30 31 Macro = PP.getMacroInfo(Tok.getIdentifierInfo()); 32 ActualArgs = Actuals; 33 CurToken = 0; 34 35 ExpandLocStart = Tok.getLocation(); 36 ExpandLocEnd = ELEnd; 37 AtStartOfLine = Tok.isAtStartOfLine(); 38 HasLeadingSpace = Tok.hasLeadingSpace(); 39 Tokens = &*Macro->tokens_begin(); 40 OwnsTokens = false; 41 DisableMacroExpansion = false; 42 NumTokens = Macro->tokens_end()-Macro->tokens_begin(); 43 MacroExpansionStart = SourceLocation(); 44 45 SourceManager &SM = PP.getSourceManager(); 46 MacroStartSLocOffset = SM.getNextLocalOffset(); 47 48 if (NumTokens > 0) { 49 assert(Tokens[0].getLocation().isValid()); 50 assert((Tokens[0].getLocation().isFileID() || Tokens[0].is(tok::comment)) && 51 "Macro defined in macro?"); 52 assert(ExpandLocStart.isValid()); 53 54 // Reserve a source location entry chunk for the length of the macro 55 // definition. Tokens that get lexed directly from the definition will 56 // have their locations pointing inside this chunk. This is to avoid 57 // creating separate source location entries for each token. 58 MacroDefStart = SM.getExpansionLoc(Tokens[0].getLocation()); 59 MacroDefLength = Macro->getDefinitionLength(SM); 60 MacroExpansionStart = SM.createExpansionLoc(MacroDefStart, 61 ExpandLocStart, 62 ExpandLocEnd, 63 MacroDefLength); 64 } 65 66 // If this is a function-like macro, expand the arguments and change 67 // Tokens to point to the expanded tokens. 68 if (Macro->isFunctionLike() && Macro->getNumArgs()) 69 ExpandFunctionArguments(); 70 71 // Mark the macro as currently disabled, so that it is not recursively 72 // expanded. The macro must be disabled only after argument pre-expansion of 73 // function-like macro arguments occurs. 74 Macro->DisableMacro(); 75 } 76 77 78 79 /// Create a TokenLexer for the specified token stream. This does not 80 /// take ownership of the specified token vector. 81 void TokenLexer::Init(const Token *TokArray, unsigned NumToks, 82 bool disableMacroExpansion, bool ownsTokens) { 83 // If the client is reusing a TokenLexer, make sure to free any memory 84 // associated with it. 85 destroy(); 86 87 Macro = 0; 88 ActualArgs = 0; 89 Tokens = TokArray; 90 OwnsTokens = ownsTokens; 91 DisableMacroExpansion = disableMacroExpansion; 92 NumTokens = NumToks; 93 CurToken = 0; 94 ExpandLocStart = ExpandLocEnd = SourceLocation(); 95 AtStartOfLine = false; 96 HasLeadingSpace = false; 97 MacroExpansionStart = SourceLocation(); 98 99 // Set HasLeadingSpace/AtStartOfLine so that the first token will be 100 // returned unmodified. 101 if (NumToks != 0) { 102 AtStartOfLine = TokArray[0].isAtStartOfLine(); 103 HasLeadingSpace = TokArray[0].hasLeadingSpace(); 104 } 105 } 106 107 108 void TokenLexer::destroy() { 109 // If this was a function-like macro that actually uses its arguments, delete 110 // the expanded tokens. 111 if (OwnsTokens) { 112 delete [] Tokens; 113 Tokens = 0; 114 OwnsTokens = false; 115 } 116 117 // TokenLexer owns its formal arguments. 118 if (ActualArgs) ActualArgs->destroy(PP); 119 } 120 121 /// Expand the arguments of a function-like macro so that we can quickly 122 /// return preexpanded tokens from Tokens. 123 void TokenLexer::ExpandFunctionArguments() { 124 125 SmallVector<Token, 128> ResultToks; 126 127 // Loop through 'Tokens', expanding them into ResultToks. Keep 128 // track of whether we change anything. If not, no need to keep them. If so, 129 // we install the newly expanded sequence as the new 'Tokens' list. 130 bool MadeChange = false; 131 132 // NextTokGetsSpace - When this is true, the next token appended to the 133 // output list will get a leading space, regardless of whether it had one to 134 // begin with or not. This is used for placemarker support. 135 bool NextTokGetsSpace = false; 136 137 for (unsigned i = 0, e = NumTokens; i != e; ++i) { 138 // If we found the stringify operator, get the argument stringified. The 139 // preprocessor already verified that the following token is a macro name 140 // when the #define was parsed. 141 const Token &CurTok = Tokens[i]; 142 if (CurTok.is(tok::hash) || CurTok.is(tok::hashat)) { 143 int ArgNo = Macro->getArgumentNum(Tokens[i+1].getIdentifierInfo()); 144 assert(ArgNo != -1 && "Token following # is not an argument?"); 145 146 SourceLocation ExpansionLocStart = 147 getExpansionLocForMacroDefLoc(CurTok.getLocation()); 148 SourceLocation ExpansionLocEnd = 149 getExpansionLocForMacroDefLoc(Tokens[i+1].getLocation()); 150 151 Token Res; 152 if (CurTok.is(tok::hash)) // Stringify 153 Res = ActualArgs->getStringifiedArgument(ArgNo, PP, 154 ExpansionLocStart, 155 ExpansionLocEnd); 156 else { 157 // 'charify': don't bother caching these. 158 Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo), 159 PP, true, 160 ExpansionLocStart, 161 ExpansionLocEnd); 162 } 163 164 // The stringified/charified string leading space flag gets set to match 165 // the #/#@ operator. 166 if (CurTok.hasLeadingSpace() || NextTokGetsSpace) 167 Res.setFlag(Token::LeadingSpace); 168 169 ResultToks.push_back(Res); 170 MadeChange = true; 171 ++i; // Skip arg name. 172 NextTokGetsSpace = false; 173 continue; 174 } 175 176 // Otherwise, if this is not an argument token, just add the token to the 177 // output buffer. 178 IdentifierInfo *II = CurTok.getIdentifierInfo(); 179 int ArgNo = II ? Macro->getArgumentNum(II) : -1; 180 if (ArgNo == -1) { 181 // This isn't an argument, just add it. 182 ResultToks.push_back(CurTok); 183 184 if (NextTokGetsSpace) { 185 ResultToks.back().setFlag(Token::LeadingSpace); 186 NextTokGetsSpace = false; 187 } 188 continue; 189 } 190 191 // An argument is expanded somehow, the result is different than the 192 // input. 193 MadeChange = true; 194 195 // Otherwise, this is a use of the argument. Find out if there is a paste 196 // (##) operator before or after the argument. 197 bool PasteBefore = 198 !ResultToks.empty() && ResultToks.back().is(tok::hashhash); 199 bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash); 200 201 // If it is not the LHS/RHS of a ## operator, we must pre-expand the 202 // argument and substitute the expanded tokens into the result. This is 203 // C99 6.10.3.1p1. 204 if (!PasteBefore && !PasteAfter) { 205 const Token *ResultArgToks; 206 207 // Only preexpand the argument if it could possibly need it. This 208 // avoids some work in common cases. 209 const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo); 210 if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP)) 211 ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, Macro, PP)[0]; 212 else 213 ResultArgToks = ArgTok; // Use non-preexpanded tokens. 214 215 // If the arg token expanded into anything, append it. 216 if (ResultArgToks->isNot(tok::eof)) { 217 unsigned FirstResult = ResultToks.size(); 218 unsigned NumToks = MacroArgs::getArgLength(ResultArgToks); 219 ResultToks.append(ResultArgToks, ResultArgToks+NumToks); 220 221 // If the '##' came from expanding an argument, turn it into 'unknown' 222 // to avoid pasting. 223 for (unsigned i = FirstResult, e = ResultToks.size(); i != e; ++i) { 224 Token &Tok = ResultToks[i]; 225 if (Tok.is(tok::hashhash)) 226 Tok.setKind(tok::unknown); 227 } 228 229 if(ExpandLocStart.isValid()) { 230 updateLocForMacroArgTokens(CurTok.getLocation(), 231 ResultToks.begin()+FirstResult, 232 ResultToks.end()); 233 } 234 235 // If any tokens were substituted from the argument, the whitespace 236 // before the first token should match the whitespace of the arg 237 // identifier. 238 ResultToks[FirstResult].setFlagValue(Token::LeadingSpace, 239 CurTok.hasLeadingSpace() || 240 NextTokGetsSpace); 241 NextTokGetsSpace = false; 242 } else { 243 // If this is an empty argument, and if there was whitespace before the 244 // formal token, make sure the next token gets whitespace before it. 245 NextTokGetsSpace = CurTok.hasLeadingSpace(); 246 } 247 continue; 248 } 249 250 // Okay, we have a token that is either the LHS or RHS of a paste (##) 251 // argument. It gets substituted as its non-pre-expanded tokens. 252 const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo); 253 unsigned NumToks = MacroArgs::getArgLength(ArgToks); 254 if (NumToks) { // Not an empty argument? 255 // If this is the GNU ", ## __VA_ARG__" extension, and we just learned 256 // that __VA_ARG__ expands to multiple tokens, avoid a pasting error when 257 // the expander trys to paste ',' with the first token of the __VA_ARG__ 258 // expansion. 259 if (PasteBefore && ResultToks.size() >= 2 && 260 ResultToks[ResultToks.size()-2].is(tok::comma) && 261 (unsigned)ArgNo == Macro->getNumArgs()-1 && 262 Macro->isVariadic()) { 263 // Remove the paste operator, report use of the extension. 264 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); 265 ResultToks.pop_back(); 266 } 267 268 ResultToks.append(ArgToks, ArgToks+NumToks); 269 270 // If the '##' came from expanding an argument, turn it into 'unknown' 271 // to avoid pasting. 272 for (unsigned i = ResultToks.size() - NumToks, e = ResultToks.size(); 273 i != e; ++i) { 274 Token &Tok = ResultToks[i]; 275 if (Tok.is(tok::hashhash)) 276 Tok.setKind(tok::unknown); 277 } 278 279 if (ExpandLocStart.isValid()) { 280 updateLocForMacroArgTokens(CurTok.getLocation(), 281 ResultToks.end()-NumToks, ResultToks.end()); 282 } 283 284 // If this token (the macro argument) was supposed to get leading 285 // whitespace, transfer this information onto the first token of the 286 // expansion. 287 // 288 // Do not do this if the paste operator occurs before the macro argument, 289 // as in "A ## MACROARG". In valid code, the first token will get 290 // smooshed onto the preceding one anyway (forming AMACROARG). In 291 // assembler-with-cpp mode, invalid pastes are allowed through: in this 292 // case, we do not want the extra whitespace to be added. For example, 293 // we want ". ## foo" -> ".foo" not ". foo". 294 if ((CurTok.hasLeadingSpace() || NextTokGetsSpace) && 295 !PasteBefore) 296 ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace); 297 298 NextTokGetsSpace = false; 299 continue; 300 } 301 302 // If an empty argument is on the LHS or RHS of a paste, the standard (C99 303 // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We 304 // implement this by eating ## operators when a LHS or RHS expands to 305 // empty. 306 NextTokGetsSpace |= CurTok.hasLeadingSpace(); 307 if (PasteAfter) { 308 // Discard the argument token and skip (don't copy to the expansion 309 // buffer) the paste operator after it. 310 NextTokGetsSpace |= Tokens[i+1].hasLeadingSpace(); 311 ++i; 312 continue; 313 } 314 315 // If this is on the RHS of a paste operator, we've already copied the 316 // paste operator to the ResultToks list. Remove it. 317 assert(PasteBefore && ResultToks.back().is(tok::hashhash)); 318 NextTokGetsSpace |= ResultToks.back().hasLeadingSpace(); 319 ResultToks.pop_back(); 320 321 // If this is the __VA_ARGS__ token, and if the argument wasn't provided, 322 // and if the macro had at least one real argument, and if the token before 323 // the ## was a comma, remove the comma. 324 if ((unsigned)ArgNo == Macro->getNumArgs()-1 && // is __VA_ARGS__ 325 ActualArgs->isVarargsElidedUse() && // Argument elided. 326 !ResultToks.empty() && ResultToks.back().is(tok::comma)) { 327 // Never add a space, even if the comma, ##, or arg had a space. 328 NextTokGetsSpace = false; 329 // Remove the paste operator, report use of the extension. 330 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); 331 ResultToks.pop_back(); 332 333 // If the comma was right after another paste (e.g. "X##,##__VA_ARGS__"), 334 // then removal of the comma should produce a placemarker token (in C99 335 // terms) which we model by popping off the previous ##, giving us a plain 336 // "X" when __VA_ARGS__ is empty. 337 if (!ResultToks.empty() && ResultToks.back().is(tok::hashhash)) 338 ResultToks.pop_back(); 339 } 340 continue; 341 } 342 343 // If anything changed, install this as the new Tokens list. 344 if (MadeChange) { 345 assert(!OwnsTokens && "This would leak if we already own the token list"); 346 // This is deleted in the dtor. 347 NumTokens = ResultToks.size(); 348 // The tokens will be added to Preprocessor's cache and will be removed 349 // when this TokenLexer finishes lexing them. 350 Tokens = PP.cacheMacroExpandedTokens(this, ResultToks); 351 352 // The preprocessor cache of macro expanded tokens owns these tokens,not us. 353 OwnsTokens = false; 354 } 355 } 356 357 /// Lex - Lex and return a token from this macro stream. 358 /// 359 void TokenLexer::Lex(Token &Tok) { 360 // Lexing off the end of the macro, pop this macro off the expansion stack. 361 if (isAtEnd()) { 362 // If this is a macro (not a token stream), mark the macro enabled now 363 // that it is no longer being expanded. 364 if (Macro) Macro->EnableMacro(); 365 366 // Pop this context off the preprocessors lexer stack and get the next 367 // token. This will delete "this" so remember the PP instance var. 368 Preprocessor &PPCache = PP; 369 if (PP.HandleEndOfTokenLexer(Tok)) 370 return; 371 372 // HandleEndOfTokenLexer may not return a token. If it doesn't, lex 373 // whatever is next. 374 return PPCache.Lex(Tok); 375 } 376 377 SourceManager &SM = PP.getSourceManager(); 378 379 // If this is the first token of the expanded result, we inherit spacing 380 // properties later. 381 bool isFirstToken = CurToken == 0; 382 383 // Get the next token to return. 384 Tok = Tokens[CurToken++]; 385 386 bool TokenIsFromPaste = false; 387 388 // If this token is followed by a token paste (##) operator, paste the tokens! 389 // Note that ## is a normal token when not expanding a macro. 390 if (!isAtEnd() && Tokens[CurToken].is(tok::hashhash) && Macro) { 391 // When handling the microsoft /##/ extension, the final token is 392 // returned by PasteTokens, not the pasted token. 393 if (PasteTokens(Tok)) 394 return; 395 396 TokenIsFromPaste = true; 397 } 398 399 // The token's current location indicate where the token was lexed from. We 400 // need this information to compute the spelling of the token, but any 401 // diagnostics for the expanded token should appear as if they came from 402 // ExpansionLoc. Pull this information together into a new SourceLocation 403 // that captures all of this. 404 if (ExpandLocStart.isValid() && // Don't do this for token streams. 405 // Check that the token's location was not already set properly. 406 SM.isBeforeInSLocAddrSpace(Tok.getLocation(), MacroStartSLocOffset)) { 407 SourceLocation instLoc; 408 if (Tok.is(tok::comment)) { 409 instLoc = SM.createExpansionLoc(Tok.getLocation(), 410 ExpandLocStart, 411 ExpandLocEnd, 412 Tok.getLength()); 413 } else { 414 instLoc = getExpansionLocForMacroDefLoc(Tok.getLocation()); 415 } 416 417 Tok.setLocation(instLoc); 418 } 419 420 // If this is the first token, set the lexical properties of the token to 421 // match the lexical properties of the macro identifier. 422 if (isFirstToken) { 423 Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); 424 Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace); 425 } 426 427 // Handle recursive expansion! 428 if (!Tok.isAnnotation() && Tok.getIdentifierInfo() != 0) { 429 // Change the kind of this identifier to the appropriate token kind, e.g. 430 // turning "for" into a keyword. 431 IdentifierInfo *II = Tok.getIdentifierInfo(); 432 Tok.setKind(II->getTokenID()); 433 434 // If this identifier was poisoned and from a paste, emit an error. This 435 // won't be handled by Preprocessor::HandleIdentifier because this is coming 436 // from a macro expansion. 437 if (II->isPoisoned() && TokenIsFromPaste) { 438 PP.HandlePoisonedIdentifier(Tok); 439 } 440 441 if (!DisableMacroExpansion && II->isHandleIdentifierCase()) 442 PP.HandleIdentifier(Tok); 443 } 444 445 // Otherwise, return a normal token. 446 } 447 448 /// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ## 449 /// operator. Read the ## and RHS, and paste the LHS/RHS together. If there 450 /// are more ## after it, chomp them iteratively. Return the result as Tok. 451 /// If this returns true, the caller should immediately return the token. 452 bool TokenLexer::PasteTokens(Token &Tok) { 453 llvm::SmallString<128> Buffer; 454 const char *ResultTokStrPtr = 0; 455 SourceLocation StartLoc = Tok.getLocation(); 456 SourceLocation PasteOpLoc; 457 do { 458 // Consume the ## operator. 459 PasteOpLoc = Tokens[CurToken].getLocation(); 460 ++CurToken; 461 assert(!isAtEnd() && "No token on the RHS of a paste operator!"); 462 463 // Get the RHS token. 464 const Token &RHS = Tokens[CurToken]; 465 466 // Allocate space for the result token. This is guaranteed to be enough for 467 // the two tokens. 468 Buffer.resize(Tok.getLength() + RHS.getLength()); 469 470 // Get the spelling of the LHS token in Buffer. 471 const char *BufPtr = &Buffer[0]; 472 bool Invalid = false; 473 unsigned LHSLen = PP.getSpelling(Tok, BufPtr, &Invalid); 474 if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer! 475 memcpy(&Buffer[0], BufPtr, LHSLen); 476 if (Invalid) 477 return true; 478 479 BufPtr = &Buffer[LHSLen]; 480 unsigned RHSLen = PP.getSpelling(RHS, BufPtr, &Invalid); 481 if (Invalid) 482 return true; 483 if (BufPtr != &Buffer[LHSLen]) // Really, we want the chars in Buffer! 484 memcpy(&Buffer[LHSLen], BufPtr, RHSLen); 485 486 // Trim excess space. 487 Buffer.resize(LHSLen+RHSLen); 488 489 // Plop the pasted result (including the trailing newline and null) into a 490 // scratch buffer where we can lex it. 491 Token ResultTokTmp; 492 ResultTokTmp.startToken(); 493 494 // Claim that the tmp token is a string_literal so that we can get the 495 // character pointer back from CreateString in getLiteralData(). 496 ResultTokTmp.setKind(tok::string_literal); 497 PP.CreateString(&Buffer[0], Buffer.size(), ResultTokTmp); 498 SourceLocation ResultTokLoc = ResultTokTmp.getLocation(); 499 ResultTokStrPtr = ResultTokTmp.getLiteralData(); 500 501 // Lex the resultant pasted token into Result. 502 Token Result; 503 504 if (Tok.isAnyIdentifier() && RHS.isAnyIdentifier()) { 505 // Common paste case: identifier+identifier = identifier. Avoid creating 506 // a lexer and other overhead. 507 PP.IncrementPasteCounter(true); 508 Result.startToken(); 509 Result.setKind(tok::raw_identifier); 510 Result.setRawIdentifierData(ResultTokStrPtr); 511 Result.setLocation(ResultTokLoc); 512 Result.setLength(LHSLen+RHSLen); 513 } else { 514 PP.IncrementPasteCounter(false); 515 516 assert(ResultTokLoc.isFileID() && 517 "Should be a raw location into scratch buffer"); 518 SourceManager &SourceMgr = PP.getSourceManager(); 519 FileID LocFileID = SourceMgr.getFileID(ResultTokLoc); 520 521 bool Invalid = false; 522 const char *ScratchBufStart 523 = SourceMgr.getBufferData(LocFileID, &Invalid).data(); 524 if (Invalid) 525 return false; 526 527 // Make a lexer to lex this string from. Lex just this one token. 528 // Make a lexer object so that we lex and expand the paste result. 529 Lexer TL(SourceMgr.getLocForStartOfFile(LocFileID), 530 PP.getLangOptions(), ScratchBufStart, 531 ResultTokStrPtr, ResultTokStrPtr+LHSLen+RHSLen); 532 533 // Lex a token in raw mode. This way it won't look up identifiers 534 // automatically, lexing off the end will return an eof token, and 535 // warnings are disabled. This returns true if the result token is the 536 // entire buffer. 537 bool isInvalid = !TL.LexFromRawLexer(Result); 538 539 // If we got an EOF token, we didn't form even ONE token. For example, we 540 // did "/ ## /" to get "//". 541 isInvalid |= Result.is(tok::eof); 542 543 // If pasting the two tokens didn't form a full new token, this is an 544 // error. This occurs with "x ## +" and other stuff. Return with Tok 545 // unmodified and with RHS as the next token to lex. 546 if (isInvalid) { 547 // Test for the Microsoft extension of /##/ turning into // here on the 548 // error path. 549 if (PP.getLangOptions().MicrosoftExt && Tok.is(tok::slash) && 550 RHS.is(tok::slash)) { 551 HandleMicrosoftCommentPaste(Tok); 552 return true; 553 } 554 555 // Do not emit the error when preprocessing assembler code. 556 if (!PP.getLangOptions().AsmPreprocessor) { 557 // Explicitly convert the token location to have proper expansion 558 // information so that the user knows where it came from. 559 SourceManager &SM = PP.getSourceManager(); 560 SourceLocation Loc = 561 SM.createExpansionLoc(PasteOpLoc, ExpandLocStart, ExpandLocEnd, 2); 562 // If we're in microsoft extensions mode, downgrade this from a hard 563 // error to a warning that defaults to an error. This allows 564 // disabling it. 565 PP.Diag(Loc, 566 PP.getLangOptions().MicrosoftExt ? diag::err_pp_bad_paste_ms 567 : diag::err_pp_bad_paste) 568 << Buffer.str(); 569 } 570 571 // Do not consume the RHS. 572 --CurToken; 573 } 574 575 // Turn ## into 'unknown' to avoid # ## # from looking like a paste 576 // operator. 577 if (Result.is(tok::hashhash)) 578 Result.setKind(tok::unknown); 579 } 580 581 // Transfer properties of the LHS over the the Result. 582 Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine()); 583 Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace()); 584 585 // Finally, replace LHS with the result, consume the RHS, and iterate. 586 ++CurToken; 587 Tok = Result; 588 } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)); 589 590 SourceLocation EndLoc = Tokens[CurToken - 1].getLocation(); 591 592 // The token's current location indicate where the token was lexed from. We 593 // need this information to compute the spelling of the token, but any 594 // diagnostics for the expanded token should appear as if the token was 595 // expanded from the full ## expression. Pull this information together into 596 // a new SourceLocation that captures all of this. 597 SourceManager &SM = PP.getSourceManager(); 598 if (StartLoc.isFileID()) 599 StartLoc = getExpansionLocForMacroDefLoc(StartLoc); 600 if (EndLoc.isFileID()) 601 EndLoc = getExpansionLocForMacroDefLoc(EndLoc); 602 Tok.setLocation(SM.createExpansionLoc(Tok.getLocation(), StartLoc, EndLoc, 603 Tok.getLength())); 604 605 // Now that we got the result token, it will be subject to expansion. Since 606 // token pasting re-lexes the result token in raw mode, identifier information 607 // isn't looked up. As such, if the result is an identifier, look up id info. 608 if (Tok.is(tok::raw_identifier)) { 609 // Look up the identifier info for the token. We disabled identifier lookup 610 // by saying we're skipping contents, so we need to do this manually. 611 PP.LookUpIdentifierInfo(Tok); 612 } 613 return false; 614 } 615 616 /// isNextTokenLParen - If the next token lexed will pop this macro off the 617 /// expansion stack, return 2. If the next unexpanded token is a '(', return 618 /// 1, otherwise return 0. 619 unsigned TokenLexer::isNextTokenLParen() const { 620 // Out of tokens? 621 if (isAtEnd()) 622 return 2; 623 return Tokens[CurToken].is(tok::l_paren); 624 } 625 626 /// isParsingPreprocessorDirective - Return true if we are in the middle of a 627 /// preprocessor directive. 628 bool TokenLexer::isParsingPreprocessorDirective() const { 629 return Tokens[NumTokens-1].is(tok::eod) && !isAtEnd(); 630 } 631 632 /// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes 633 /// together to form a comment that comments out everything in the current 634 /// macro, other active macros, and anything left on the current physical 635 /// source line of the expanded buffer. Handle this by returning the 636 /// first token on the next line. 637 void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) { 638 // We 'comment out' the rest of this macro by just ignoring the rest of the 639 // tokens that have not been lexed yet, if any. 640 641 // Since this must be a macro, mark the macro enabled now that it is no longer 642 // being expanded. 643 assert(Macro && "Token streams can't paste comments"); 644 Macro->EnableMacro(); 645 646 PP.HandleMicrosoftCommentPaste(Tok); 647 } 648 649 /// \brief If \arg loc is a file ID and points inside the current macro 650 /// definition, returns the appropriate source location pointing at the 651 /// macro expansion source location entry, otherwise it returns an invalid 652 /// SourceLocation. 653 SourceLocation 654 TokenLexer::getExpansionLocForMacroDefLoc(SourceLocation loc) const { 655 assert(ExpandLocStart.isValid() && MacroExpansionStart.isValid() && 656 "Not appropriate for token streams"); 657 assert(loc.isValid() && loc.isFileID()); 658 659 SourceManager &SM = PP.getSourceManager(); 660 assert(SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength) && 661 "Expected loc to come from the macro definition"); 662 663 unsigned relativeOffset = 0; 664 SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength, &relativeOffset); 665 return MacroExpansionStart.getLocWithOffset(relativeOffset); 666 } 667 668 /// \brief Finds the tokens that are consecutive (from the same FileID) 669 /// creates a single SLocEntry, and assigns SourceLocations to each token that 670 /// point to that SLocEntry. e.g for 671 /// assert(foo == bar); 672 /// There will be a single SLocEntry for the "foo == bar" chunk and locations 673 /// for the 'foo', '==', 'bar' tokens will point inside that chunk. 674 /// 675 /// \arg begin_tokens will be updated to a position past all the found 676 /// consecutive tokens. 677 static void updateConsecutiveMacroArgTokens(SourceManager &SM, 678 SourceLocation InstLoc, 679 Token *&begin_tokens, 680 Token * end_tokens) { 681 assert(begin_tokens < end_tokens); 682 683 SourceLocation FirstLoc = begin_tokens->getLocation(); 684 SourceLocation CurLoc = FirstLoc; 685 686 // Compare the source location offset of tokens and group together tokens that 687 // are close, even if their locations point to different FileIDs. e.g. 688 // 689 // |bar | foo | cake | (3 tokens from 3 consecutive FileIDs) 690 // ^ ^ 691 // |bar foo cake| (one SLocEntry chunk for all tokens) 692 // 693 // we can perform this "merge" since the token's spelling location depends 694 // on the relative offset. 695 696 Token *NextTok = begin_tokens + 1; 697 for (; NextTok < end_tokens; ++NextTok) { 698 int RelOffs; 699 if (!SM.isInSameSLocAddrSpace(CurLoc, NextTok->getLocation(), &RelOffs)) 700 break; // Token from different local/loaded location. 701 // Check that token is not before the previous token or more than 50 702 // "characters" away. 703 if (RelOffs < 0 || RelOffs > 50) 704 break; 705 CurLoc = NextTok->getLocation(); 706 } 707 708 // For the consecutive tokens, find the length of the SLocEntry to contain 709 // all of them. 710 Token &LastConsecutiveTok = *(NextTok-1); 711 int LastRelOffs = 0; 712 SM.isInSameSLocAddrSpace(FirstLoc, LastConsecutiveTok.getLocation(), 713 &LastRelOffs); 714 unsigned FullLength = LastRelOffs + LastConsecutiveTok.getLength(); 715 716 // Create a macro expansion SLocEntry that will "contain" all of the tokens. 717 SourceLocation Expansion = 718 SM.createMacroArgExpansionLoc(FirstLoc, InstLoc,FullLength); 719 720 // Change the location of the tokens from the spelling location to the new 721 // expanded location. 722 for (; begin_tokens < NextTok; ++begin_tokens) { 723 Token &Tok = *begin_tokens; 724 int RelOffs = 0; 725 SM.isInSameSLocAddrSpace(FirstLoc, Tok.getLocation(), &RelOffs); 726 Tok.setLocation(Expansion.getLocWithOffset(RelOffs)); 727 } 728 } 729 730 /// \brief Creates SLocEntries and updates the locations of macro argument 731 /// tokens to their new expanded locations. 732 /// 733 /// \param ArgIdDefLoc the location of the macro argument id inside the macro 734 /// definition. 735 /// \param Tokens the macro argument tokens to update. 736 void TokenLexer::updateLocForMacroArgTokens(SourceLocation ArgIdSpellLoc, 737 Token *begin_tokens, 738 Token *end_tokens) { 739 SourceManager &SM = PP.getSourceManager(); 740 741 SourceLocation InstLoc = 742 getExpansionLocForMacroDefLoc(ArgIdSpellLoc); 743 744 while (begin_tokens < end_tokens) { 745 // If there's only one token just create a SLocEntry for it. 746 if (end_tokens - begin_tokens == 1) { 747 Token &Tok = *begin_tokens; 748 Tok.setLocation(SM.createMacroArgExpansionLoc(Tok.getLocation(), 749 InstLoc, 750 Tok.getLength())); 751 return; 752 } 753 754 updateConsecutiveMacroArgTokens(SM, InstLoc, begin_tokens, end_tokens); 755 } 756 } 757