TokenLexer.cpp revision a64ccefdf0ea4e03ec88805d71b0af74950c7472
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" 21using 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. 26void 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. 81void 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 108void 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. 123void 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 hashInstLoc = 147 getExpansionLocForMacroDefLoc(CurTok.getLocation()); 148 149 Token Res; 150 if (CurTok.is(tok::hash)) // Stringify 151 Res = ActualArgs->getStringifiedArgument(ArgNo, PP, hashInstLoc); 152 else { 153 // 'charify': don't bother caching these. 154 Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo), 155 PP, true, hashInstLoc); 156 } 157 158 // The stringified/charified string leading space flag gets set to match 159 // the #/#@ operator. 160 if (CurTok.hasLeadingSpace() || NextTokGetsSpace) 161 Res.setFlag(Token::LeadingSpace); 162 163 ResultToks.push_back(Res); 164 MadeChange = true; 165 ++i; // Skip arg name. 166 NextTokGetsSpace = false; 167 continue; 168 } 169 170 // Otherwise, if this is not an argument token, just add the token to the 171 // output buffer. 172 IdentifierInfo *II = CurTok.getIdentifierInfo(); 173 int ArgNo = II ? Macro->getArgumentNum(II) : -1; 174 if (ArgNo == -1) { 175 // This isn't an argument, just add it. 176 ResultToks.push_back(CurTok); 177 178 if (NextTokGetsSpace) { 179 ResultToks.back().setFlag(Token::LeadingSpace); 180 NextTokGetsSpace = false; 181 } 182 continue; 183 } 184 185 // An argument is expanded somehow, the result is different than the 186 // input. 187 MadeChange = true; 188 189 // Otherwise, this is a use of the argument. Find out if there is a paste 190 // (##) operator before or after the argument. 191 bool PasteBefore = 192 !ResultToks.empty() && ResultToks.back().is(tok::hashhash); 193 bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash); 194 195 // If it is not the LHS/RHS of a ## operator, we must pre-expand the 196 // argument and substitute the expanded tokens into the result. This is 197 // C99 6.10.3.1p1. 198 if (!PasteBefore && !PasteAfter) { 199 const Token *ResultArgToks; 200 201 // Only preexpand the argument if it could possibly need it. This 202 // avoids some work in common cases. 203 const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo); 204 if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP)) 205 ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, Macro, PP)[0]; 206 else 207 ResultArgToks = ArgTok; // Use non-preexpanded tokens. 208 209 // If the arg token expanded into anything, append it. 210 if (ResultArgToks->isNot(tok::eof)) { 211 unsigned FirstResult = ResultToks.size(); 212 unsigned NumToks = MacroArgs::getArgLength(ResultArgToks); 213 ResultToks.append(ResultArgToks, ResultArgToks+NumToks); 214 215 // If the '##' came from expanding an argument, turn it into 'unknown' 216 // to avoid pasting. 217 for (unsigned i = FirstResult, e = ResultToks.size(); i != e; ++i) { 218 Token &Tok = ResultToks[i]; 219 if (Tok.is(tok::hashhash)) 220 Tok.setKind(tok::unknown); 221 } 222 223 if(ExpandLocStart.isValid()) { 224 updateLocForMacroArgTokens(CurTok.getLocation(), 225 ResultToks.begin()+FirstResult, 226 ResultToks.end()); 227 } 228 229 // If any tokens were substituted from the argument, the whitespace 230 // before the first token should match the whitespace of the arg 231 // identifier. 232 ResultToks[FirstResult].setFlagValue(Token::LeadingSpace, 233 CurTok.hasLeadingSpace() || 234 NextTokGetsSpace); 235 NextTokGetsSpace = false; 236 } else { 237 // If this is an empty argument, and if there was whitespace before the 238 // formal token, make sure the next token gets whitespace before it. 239 NextTokGetsSpace = CurTok.hasLeadingSpace(); 240 } 241 continue; 242 } 243 244 // Okay, we have a token that is either the LHS or RHS of a paste (##) 245 // argument. It gets substituted as its non-pre-expanded tokens. 246 const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo); 247 unsigned NumToks = MacroArgs::getArgLength(ArgToks); 248 if (NumToks) { // Not an empty argument? 249 // If this is the GNU ", ## __VA_ARG__" extension, and we just learned 250 // that __VA_ARG__ expands to multiple tokens, avoid a pasting error when 251 // the expander trys to paste ',' with the first token of the __VA_ARG__ 252 // expansion. 253 if (PasteBefore && ResultToks.size() >= 2 && 254 ResultToks[ResultToks.size()-2].is(tok::comma) && 255 (unsigned)ArgNo == Macro->getNumArgs()-1 && 256 Macro->isVariadic()) { 257 // Remove the paste operator, report use of the extension. 258 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); 259 ResultToks.pop_back(); 260 } 261 262 ResultToks.append(ArgToks, ArgToks+NumToks); 263 264 // If the '##' came from expanding an argument, turn it into 'unknown' 265 // to avoid pasting. 266 for (unsigned i = ResultToks.size() - NumToks, e = ResultToks.size(); 267 i != e; ++i) { 268 Token &Tok = ResultToks[i]; 269 if (Tok.is(tok::hashhash)) 270 Tok.setKind(tok::unknown); 271 } 272 273 if (ExpandLocStart.isValid()) { 274 updateLocForMacroArgTokens(CurTok.getLocation(), 275 ResultToks.end()-NumToks, ResultToks.end()); 276 } 277 278 // If this token (the macro argument) was supposed to get leading 279 // whitespace, transfer this information onto the first token of the 280 // expansion. 281 // 282 // Do not do this if the paste operator occurs before the macro argument, 283 // as in "A ## MACROARG". In valid code, the first token will get 284 // smooshed onto the preceding one anyway (forming AMACROARG). In 285 // assembler-with-cpp mode, invalid pastes are allowed through: in this 286 // case, we do not want the extra whitespace to be added. For example, 287 // we want ". ## foo" -> ".foo" not ". foo". 288 if ((CurTok.hasLeadingSpace() || NextTokGetsSpace) && 289 !PasteBefore) 290 ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace); 291 292 NextTokGetsSpace = false; 293 continue; 294 } 295 296 // If an empty argument is on the LHS or RHS of a paste, the standard (C99 297 // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We 298 // implement this by eating ## operators when a LHS or RHS expands to 299 // empty. 300 NextTokGetsSpace |= CurTok.hasLeadingSpace(); 301 if (PasteAfter) { 302 // Discard the argument token and skip (don't copy to the expansion 303 // buffer) the paste operator after it. 304 NextTokGetsSpace |= Tokens[i+1].hasLeadingSpace(); 305 ++i; 306 continue; 307 } 308 309 // If this is on the RHS of a paste operator, we've already copied the 310 // paste operator to the ResultToks list. Remove it. 311 assert(PasteBefore && ResultToks.back().is(tok::hashhash)); 312 NextTokGetsSpace |= ResultToks.back().hasLeadingSpace(); 313 ResultToks.pop_back(); 314 315 // If this is the __VA_ARGS__ token, and if the argument wasn't provided, 316 // and if the macro had at least one real argument, and if the token before 317 // the ## was a comma, remove the comma. 318 if ((unsigned)ArgNo == Macro->getNumArgs()-1 && // is __VA_ARGS__ 319 ActualArgs->isVarargsElidedUse() && // Argument elided. 320 !ResultToks.empty() && ResultToks.back().is(tok::comma)) { 321 // Never add a space, even if the comma, ##, or arg had a space. 322 NextTokGetsSpace = false; 323 // Remove the paste operator, report use of the extension. 324 PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); 325 ResultToks.pop_back(); 326 327 // If the comma was right after another paste (e.g. "X##,##__VA_ARGS__"), 328 // then removal of the comma should produce a placemarker token (in C99 329 // terms) which we model by popping off the previous ##, giving us a plain 330 // "X" when __VA_ARGS__ is empty. 331 if (!ResultToks.empty() && ResultToks.back().is(tok::hashhash)) 332 ResultToks.pop_back(); 333 } 334 continue; 335 } 336 337 // If anything changed, install this as the new Tokens list. 338 if (MadeChange) { 339 assert(!OwnsTokens && "This would leak if we already own the token list"); 340 // This is deleted in the dtor. 341 NumTokens = ResultToks.size(); 342 // The tokens will be added to Preprocessor's cache and will be removed 343 // when this TokenLexer finishes lexing them. 344 Tokens = PP.cacheMacroExpandedTokens(this, ResultToks); 345 346 // The preprocessor cache of macro expanded tokens owns these tokens,not us. 347 OwnsTokens = false; 348 } 349} 350 351/// Lex - Lex and return a token from this macro stream. 352/// 353void TokenLexer::Lex(Token &Tok) { 354 // Lexing off the end of the macro, pop this macro off the expansion stack. 355 if (isAtEnd()) { 356 // If this is a macro (not a token stream), mark the macro enabled now 357 // that it is no longer being expanded. 358 if (Macro) Macro->EnableMacro(); 359 360 // Pop this context off the preprocessors lexer stack and get the next 361 // token. This will delete "this" so remember the PP instance var. 362 Preprocessor &PPCache = PP; 363 if (PP.HandleEndOfTokenLexer(Tok)) 364 return; 365 366 // HandleEndOfTokenLexer may not return a token. If it doesn't, lex 367 // whatever is next. 368 return PPCache.Lex(Tok); 369 } 370 371 SourceManager &SM = PP.getSourceManager(); 372 373 // If this is the first token of the expanded result, we inherit spacing 374 // properties later. 375 bool isFirstToken = CurToken == 0; 376 377 // Get the next token to return. 378 Tok = Tokens[CurToken++]; 379 380 bool TokenIsFromPaste = false; 381 382 // If this token is followed by a token paste (##) operator, paste the tokens! 383 // Note that ## is a normal token when not expanding a macro. 384 if (!isAtEnd() && Tokens[CurToken].is(tok::hashhash) && Macro) { 385 // When handling the microsoft /##/ extension, the final token is 386 // returned by PasteTokens, not the pasted token. 387 if (PasteTokens(Tok)) 388 return; 389 390 TokenIsFromPaste = true; 391 } 392 393 // The token's current location indicate where the token was lexed from. We 394 // need this information to compute the spelling of the token, but any 395 // diagnostics for the expanded token should appear as if they came from 396 // ExpansionLoc. Pull this information together into a new SourceLocation 397 // that captures all of this. 398 if (ExpandLocStart.isValid() && // Don't do this for token streams. 399 // Check that the token's location was not already set properly. 400 SM.isBeforeInSLocAddrSpace(Tok.getLocation(), MacroStartSLocOffset)) { 401 SourceLocation instLoc; 402 if (Tok.is(tok::comment)) { 403 instLoc = SM.createExpansionLoc(Tok.getLocation(), 404 ExpandLocStart, 405 ExpandLocEnd, 406 Tok.getLength()); 407 } else { 408 instLoc = getExpansionLocForMacroDefLoc(Tok.getLocation()); 409 } 410 411 Tok.setLocation(instLoc); 412 } 413 414 // If this is the first token, set the lexical properties of the token to 415 // match the lexical properties of the macro identifier. 416 if (isFirstToken) { 417 Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); 418 Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace); 419 } 420 421 // Handle recursive expansion! 422 if (!Tok.isAnnotation() && Tok.getIdentifierInfo() != 0) { 423 // Change the kind of this identifier to the appropriate token kind, e.g. 424 // turning "for" into a keyword. 425 IdentifierInfo *II = Tok.getIdentifierInfo(); 426 Tok.setKind(II->getTokenID()); 427 428 // If this identifier was poisoned and from a paste, emit an error. This 429 // won't be handled by Preprocessor::HandleIdentifier because this is coming 430 // from a macro expansion. 431 if (II->isPoisoned() && TokenIsFromPaste) { 432 PP.HandlePoisonedIdentifier(Tok); 433 } 434 435 if (!DisableMacroExpansion && II->isHandleIdentifierCase()) 436 PP.HandleIdentifier(Tok); 437 } 438 439 // Otherwise, return a normal token. 440} 441 442/// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ## 443/// operator. Read the ## and RHS, and paste the LHS/RHS together. If there 444/// are more ## after it, chomp them iteratively. Return the result as Tok. 445/// If this returns true, the caller should immediately return the token. 446bool TokenLexer::PasteTokens(Token &Tok) { 447 llvm::SmallString<128> Buffer; 448 const char *ResultTokStrPtr = 0; 449 SourceLocation PasteOpLoc; 450 do { 451 // Consume the ## operator. 452 PasteOpLoc = Tokens[CurToken].getLocation(); 453 ++CurToken; 454 assert(!isAtEnd() && "No token on the RHS of a paste operator!"); 455 456 // Get the RHS token. 457 const Token &RHS = Tokens[CurToken]; 458 459 // Allocate space for the result token. This is guaranteed to be enough for 460 // the two tokens. 461 Buffer.resize(Tok.getLength() + RHS.getLength()); 462 463 // Get the spelling of the LHS token in Buffer. 464 const char *BufPtr = &Buffer[0]; 465 bool Invalid = false; 466 unsigned LHSLen = PP.getSpelling(Tok, BufPtr, &Invalid); 467 if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer! 468 memcpy(&Buffer[0], BufPtr, LHSLen); 469 if (Invalid) 470 return true; 471 472 BufPtr = &Buffer[LHSLen]; 473 unsigned RHSLen = PP.getSpelling(RHS, BufPtr, &Invalid); 474 if (Invalid) 475 return true; 476 if (BufPtr != &Buffer[LHSLen]) // Really, we want the chars in Buffer! 477 memcpy(&Buffer[LHSLen], BufPtr, RHSLen); 478 479 // Trim excess space. 480 Buffer.resize(LHSLen+RHSLen); 481 482 // Plop the pasted result (including the trailing newline and null) into a 483 // scratch buffer where we can lex it. 484 Token ResultTokTmp; 485 ResultTokTmp.startToken(); 486 487 // Claim that the tmp token is a string_literal so that we can get the 488 // character pointer back from CreateString in getLiteralData(). 489 ResultTokTmp.setKind(tok::string_literal); 490 PP.CreateString(&Buffer[0], Buffer.size(), ResultTokTmp); 491 SourceLocation ResultTokLoc = ResultTokTmp.getLocation(); 492 ResultTokStrPtr = ResultTokTmp.getLiteralData(); 493 494 // Lex the resultant pasted token into Result. 495 Token Result; 496 497 if (Tok.isAnyIdentifier() && RHS.isAnyIdentifier()) { 498 // Common paste case: identifier+identifier = identifier. Avoid creating 499 // a lexer and other overhead. 500 PP.IncrementPasteCounter(true); 501 Result.startToken(); 502 Result.setKind(tok::raw_identifier); 503 Result.setRawIdentifierData(ResultTokStrPtr); 504 Result.setLocation(ResultTokLoc); 505 Result.setLength(LHSLen+RHSLen); 506 } else { 507 PP.IncrementPasteCounter(false); 508 509 assert(ResultTokLoc.isFileID() && 510 "Should be a raw location into scratch buffer"); 511 SourceManager &SourceMgr = PP.getSourceManager(); 512 FileID LocFileID = SourceMgr.getFileID(ResultTokLoc); 513 514 bool Invalid = false; 515 const char *ScratchBufStart 516 = SourceMgr.getBufferData(LocFileID, &Invalid).data(); 517 if (Invalid) 518 return false; 519 520 // Make a lexer to lex this string from. Lex just this one token. 521 // Make a lexer object so that we lex and expand the paste result. 522 Lexer TL(SourceMgr.getLocForStartOfFile(LocFileID), 523 PP.getLangOptions(), ScratchBufStart, 524 ResultTokStrPtr, ResultTokStrPtr+LHSLen+RHSLen); 525 526 // Lex a token in raw mode. This way it won't look up identifiers 527 // automatically, lexing off the end will return an eof token, and 528 // warnings are disabled. This returns true if the result token is the 529 // entire buffer. 530 bool isInvalid = !TL.LexFromRawLexer(Result); 531 532 // If we got an EOF token, we didn't form even ONE token. For example, we 533 // did "/ ## /" to get "//". 534 isInvalid |= Result.is(tok::eof); 535 536 // If pasting the two tokens didn't form a full new token, this is an 537 // error. This occurs with "x ## +" and other stuff. Return with Tok 538 // unmodified and with RHS as the next token to lex. 539 if (isInvalid) { 540 // Test for the Microsoft extension of /##/ turning into // here on the 541 // error path. 542 if (PP.getLangOptions().MicrosoftExt && Tok.is(tok::slash) && 543 RHS.is(tok::slash)) { 544 HandleMicrosoftCommentPaste(Tok); 545 return true; 546 } 547 548 // Do not emit the error when preprocessing assembler code. 549 if (!PP.getLangOptions().AsmPreprocessor) { 550 // Explicitly convert the token location to have proper expansion 551 // information so that the user knows where it came from. 552 SourceManager &SM = PP.getSourceManager(); 553 SourceLocation Loc = 554 SM.createExpansionLoc(PasteOpLoc, ExpandLocStart, ExpandLocEnd, 2); 555 // If we're in microsoft extensions mode, downgrade this from a hard 556 // error to a warning that defaults to an error. This allows 557 // disabling it. 558 PP.Diag(Loc, 559 PP.getLangOptions().MicrosoftExt ? diag::err_pp_bad_paste_ms 560 : diag::err_pp_bad_paste) 561 << Buffer.str(); 562 } 563 564 // Do not consume the RHS. 565 --CurToken; 566 } 567 568 // Turn ## into 'unknown' to avoid # ## # from looking like a paste 569 // operator. 570 if (Result.is(tok::hashhash)) 571 Result.setKind(tok::unknown); 572 } 573 574 // Transfer properties of the LHS over the the Result. 575 Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine()); 576 Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace()); 577 578 // Finally, replace LHS with the result, consume the RHS, and iterate. 579 ++CurToken; 580 Tok = Result; 581 } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)); 582 583 // The token's current location indicate where the token was lexed from. We 584 // need this information to compute the spelling of the token, but any 585 // diagnostics for the expanded token should appear as if the token was 586 // expanded from the (##) operator. Pull this information together into 587 // a new SourceLocation that captures all of this. 588 SourceManager &SM = PP.getSourceManager(); 589 SourceLocation pasteLocInst = getExpansionLocForMacroDefLoc(PasteOpLoc); 590 Tok.setLocation(SM.createExpansionLoc(Tok.getLocation(), 591 pasteLocInst, 592 pasteLocInst, 593 Tok.getLength())); 594 595 // Now that we got the result token, it will be subject to expansion. Since 596 // token pasting re-lexes the result token in raw mode, identifier information 597 // isn't looked up. As such, if the result is an identifier, look up id info. 598 if (Tok.is(tok::raw_identifier)) { 599 // Look up the identifier info for the token. We disabled identifier lookup 600 // by saying we're skipping contents, so we need to do this manually. 601 PP.LookUpIdentifierInfo(Tok); 602 } 603 return false; 604} 605 606/// isNextTokenLParen - If the next token lexed will pop this macro off the 607/// expansion stack, return 2. If the next unexpanded token is a '(', return 608/// 1, otherwise return 0. 609unsigned TokenLexer::isNextTokenLParen() const { 610 // Out of tokens? 611 if (isAtEnd()) 612 return 2; 613 return Tokens[CurToken].is(tok::l_paren); 614} 615 616/// isParsingPreprocessorDirective - Return true if we are in the middle of a 617/// preprocessor directive. 618bool TokenLexer::isParsingPreprocessorDirective() const { 619 return Tokens[NumTokens-1].is(tok::eod) && !isAtEnd(); 620} 621 622/// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes 623/// together to form a comment that comments out everything in the current 624/// macro, other active macros, and anything left on the current physical 625/// source line of the expanded buffer. Handle this by returning the 626/// first token on the next line. 627void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) { 628 // We 'comment out' the rest of this macro by just ignoring the rest of the 629 // tokens that have not been lexed yet, if any. 630 631 // Since this must be a macro, mark the macro enabled now that it is no longer 632 // being expanded. 633 assert(Macro && "Token streams can't paste comments"); 634 Macro->EnableMacro(); 635 636 PP.HandleMicrosoftCommentPaste(Tok); 637} 638 639/// \brief If \arg loc is a file ID and points inside the current macro 640/// definition, returns the appropriate source location pointing at the 641/// macro expansion source location entry, otherwise it returns an invalid 642/// SourceLocation. 643SourceLocation 644TokenLexer::getExpansionLocForMacroDefLoc(SourceLocation loc) const { 645 assert(ExpandLocStart.isValid() && MacroExpansionStart.isValid() && 646 "Not appropriate for token streams"); 647 assert(loc.isValid() && loc.isFileID()); 648 649 SourceManager &SM = PP.getSourceManager(); 650 assert(SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength) && 651 "Expected loc to come from the macro definition"); 652 653 unsigned relativeOffset = 0; 654 SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength, &relativeOffset); 655 return MacroExpansionStart.getLocWithOffset(relativeOffset); 656} 657 658/// \brief Finds the tokens that are consecutive (from the same FileID) 659/// creates a single SLocEntry, and assigns SourceLocations to each token that 660/// point to that SLocEntry. e.g for 661/// assert(foo == bar); 662/// There will be a single SLocEntry for the "foo == bar" chunk and locations 663/// for the 'foo', '==', 'bar' tokens will point inside that chunk. 664/// 665/// \arg begin_tokens will be updated to a position past all the found 666/// consecutive tokens. 667static void updateConsecutiveMacroArgTokens(SourceManager &SM, 668 SourceLocation InstLoc, 669 Token *&begin_tokens, 670 Token * end_tokens) { 671 assert(begin_tokens < end_tokens); 672 673 SourceLocation FirstLoc = begin_tokens->getLocation(); 674 SourceLocation CurLoc = FirstLoc; 675 676 // Compare the source location offset of tokens and group together tokens that 677 // are close, even if their locations point to different FileIDs. e.g. 678 // 679 // |bar | foo | cake | (3 tokens from 3 consecutive FileIDs) 680 // ^ ^ 681 // |bar foo cake| (one SLocEntry chunk for all tokens) 682 // 683 // we can perform this "merge" since the token's spelling location depends 684 // on the relative offset. 685 686 Token *NextTok = begin_tokens + 1; 687 for (; NextTok < end_tokens; ++NextTok) { 688 int RelOffs; 689 if (!SM.isInSameSLocAddrSpace(CurLoc, NextTok->getLocation(), &RelOffs)) 690 break; // Token from different local/loaded location. 691 // Check that token is not before the previous token or more than 50 692 // "characters" away. 693 if (RelOffs < 0 || RelOffs > 50) 694 break; 695 CurLoc = NextTok->getLocation(); 696 } 697 698 // For the consecutive tokens, find the length of the SLocEntry to contain 699 // all of them. 700 Token &LastConsecutiveTok = *(NextTok-1); 701 int LastRelOffs = 0; 702 SM.isInSameSLocAddrSpace(FirstLoc, LastConsecutiveTok.getLocation(), 703 &LastRelOffs); 704 unsigned FullLength = LastRelOffs + LastConsecutiveTok.getLength(); 705 706 // Create a macro expansion SLocEntry that will "contain" all of the tokens. 707 SourceLocation Expansion = 708 SM.createMacroArgExpansionLoc(FirstLoc, InstLoc,FullLength); 709 710 // Change the location of the tokens from the spelling location to the new 711 // expanded location. 712 for (; begin_tokens < NextTok; ++begin_tokens) { 713 Token &Tok = *begin_tokens; 714 int RelOffs = 0; 715 SM.isInSameSLocAddrSpace(FirstLoc, Tok.getLocation(), &RelOffs); 716 Tok.setLocation(Expansion.getLocWithOffset(RelOffs)); 717 } 718} 719 720/// \brief Creates SLocEntries and updates the locations of macro argument 721/// tokens to their new expanded locations. 722/// 723/// \param ArgIdDefLoc the location of the macro argument id inside the macro 724/// definition. 725/// \param Tokens the macro argument tokens to update. 726void TokenLexer::updateLocForMacroArgTokens(SourceLocation ArgIdSpellLoc, 727 Token *begin_tokens, 728 Token *end_tokens) { 729 SourceManager &SM = PP.getSourceManager(); 730 731 SourceLocation InstLoc = 732 getExpansionLocForMacroDefLoc(ArgIdSpellLoc); 733 734 while (begin_tokens < end_tokens) { 735 // If there's only one token just create a SLocEntry for it. 736 if (end_tokens - begin_tokens == 1) { 737 Token &Tok = *begin_tokens; 738 Tok.setLocation(SM.createMacroArgExpansionLoc(Tok.getLocation(), 739 InstLoc, 740 Tok.getLength())); 741 return; 742 } 743 744 updateConsecutiveMacroArgTokens(SM, InstLoc, begin_tokens, end_tokens); 745 } 746} 747