PPDirectives.cpp revision de076650fa79a02743b5fe57a72f991ba24779f9
1//===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===// 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 # directive processing for the Preprocessor. 11// 12//===----------------------------------------------------------------------===// 13 14#include "clang/Lex/Preprocessor.h" 15#include "clang/Lex/LiteralSupport.h" 16#include "clang/Lex/HeaderSearch.h" 17#include "clang/Lex/MacroInfo.h" 18#include "clang/Lex/LexDiagnostic.h" 19#include "clang/Basic/SourceManager.h" 20#include "llvm/ADT/APInt.h" 21using namespace clang; 22 23//===----------------------------------------------------------------------===// 24// Utility Methods for Preprocessor Directive Handling. 25//===----------------------------------------------------------------------===// 26 27MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 28 MacroInfo *MI; 29 30 if (!MICache.empty()) { 31 MI = MICache.back(); 32 MICache.pop_back(); 33 } else 34 MI = (MacroInfo*) BP.Allocate<MacroInfo>(); 35 new (MI) MacroInfo(L); 36 return MI; 37} 38 39/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 40/// be reused for allocating new MacroInfo objects. 41void Preprocessor::ReleaseMacroInfo(MacroInfo* MI) { 42 MICache.push_back(MI); 43 MI->FreeArgumentList(BP); 44} 45 46 47/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 48/// current line until the tok::eom token is found. 49void Preprocessor::DiscardUntilEndOfDirective() { 50 Token Tmp; 51 do { 52 LexUnexpandedToken(Tmp); 53 } while (Tmp.isNot(tok::eom)); 54} 55 56/// ReadMacroName - Lex and validate a macro name, which occurs after a 57/// #define or #undef. This sets the token kind to eom and discards the rest 58/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 59/// this is due to a a #define, 2 if #undef directive, 0 if it is something 60/// else (e.g. #ifdef). 61void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 62 // Read the token, don't allow macro expansion on it. 63 LexUnexpandedToken(MacroNameTok); 64 65 // Missing macro name? 66 if (MacroNameTok.is(tok::eom)) { 67 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 68 return; 69 } 70 71 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 72 if (II == 0) { 73 std::string Spelling = getSpelling(MacroNameTok); 74 const IdentifierInfo &Info = Identifiers.get(Spelling); 75 if (Info.isCPlusPlusOperatorKeyword()) 76 // C++ 2.5p2: Alternative tokens behave the same as its primary token 77 // except for their spellings. 78 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 79 else 80 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 81 // Fall through on error. 82 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 83 // Error if defining "defined": C99 6.10.8.4. 84 Diag(MacroNameTok, diag::err_defined_macro_name); 85 } else if (isDefineUndef && II->hasMacroDefinition() && 86 getMacroInfo(II)->isBuiltinMacro()) { 87 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 88 if (isDefineUndef == 1) 89 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 90 else 91 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 92 } else { 93 // Okay, we got a good identifier node. Return it. 94 return; 95 } 96 97 // Invalid macro name, read and discard the rest of the line. Then set the 98 // token kind to tok::eom. 99 MacroNameTok.setKind(tok::eom); 100 return DiscardUntilEndOfDirective(); 101} 102 103/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 104/// not, emit a diagnostic and consume up until the eom. 105void Preprocessor::CheckEndOfDirective(const char *DirType) { 106 Token Tmp; 107 // Lex unexpanded tokens: macros might expand to zero tokens, causing us to 108 // miss diagnosing invalid lines. 109 LexUnexpandedToken(Tmp); 110 111 // There should be no tokens after the directive, but we allow them as an 112 // extension. 113 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 114 LexUnexpandedToken(Tmp); 115 116 if (Tmp.isNot(tok::eom)) { 117 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType; 118 DiscardUntilEndOfDirective(); 119 } 120} 121 122 123 124/// SkipExcludedConditionalBlock - We just read a #if or related directive and 125/// decided that the subsequent tokens are in the #if'd out portion of the 126/// file. Lex the rest of the file, until we see an #endif. If 127/// FoundNonSkipPortion is true, then we have already emitted code for part of 128/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 129/// is true, then #else directives are ok, if not, then we have already seen one 130/// so a #else directive is a duplicate. When this returns, the caller can lex 131/// the first valid token. 132void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 133 bool FoundNonSkipPortion, 134 bool FoundElse) { 135 ++NumSkipped; 136 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 137 138 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 139 FoundNonSkipPortion, FoundElse); 140 141 if (CurPTHLexer) { 142 PTHSkipExcludedConditionalBlock(); 143 return; 144 } 145 146 // Enter raw mode to disable identifier lookup (and thus macro expansion), 147 // disabling warnings, etc. 148 CurPPLexer->LexingRawMode = true; 149 Token Tok; 150 while (1) { 151 if (CurLexer) 152 CurLexer->Lex(Tok); 153 else 154 CurPTHLexer->Lex(Tok); 155 156 // If this is the end of the buffer, we have an error. 157 if (Tok.is(tok::eof)) { 158 // Emit errors for each unterminated conditional on the stack, including 159 // the current one. 160 while (!CurPPLexer->ConditionalStack.empty()) { 161 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 162 diag::err_pp_unterminated_conditional); 163 CurPPLexer->ConditionalStack.pop_back(); 164 } 165 166 // Just return and let the caller lex after this #include. 167 break; 168 } 169 170 // If this token is not a preprocessor directive, just skip it. 171 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 172 continue; 173 174 // We just parsed a # character at the start of a line, so we're in 175 // directive mode. Tell the lexer this so any newlines we see will be 176 // converted into an EOM token (this terminates the macro). 177 CurPPLexer->ParsingPreprocessorDirective = true; 178 if (CurLexer) CurLexer->SetCommentRetentionState(false); 179 180 181 // Read the next token, the directive flavor. 182 LexUnexpandedToken(Tok); 183 184 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 185 // something bogus), skip it. 186 if (Tok.isNot(tok::identifier)) { 187 CurPPLexer->ParsingPreprocessorDirective = false; 188 // Restore comment saving mode. 189 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 190 continue; 191 } 192 193 // If the first letter isn't i or e, it isn't intesting to us. We know that 194 // this is safe in the face of spelling differences, because there is no way 195 // to spell an i/e in a strange way that is another letter. Skipping this 196 // allows us to avoid looking up the identifier info for #define/#undef and 197 // other common directives. 198 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation()); 199 char FirstChar = RawCharData[0]; 200 if (FirstChar >= 'a' && FirstChar <= 'z' && 201 FirstChar != 'i' && FirstChar != 'e') { 202 CurPPLexer->ParsingPreprocessorDirective = false; 203 // Restore comment saving mode. 204 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 205 continue; 206 } 207 208 // Get the identifier name without trigraphs or embedded newlines. Note 209 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 210 // when skipping. 211 // TODO: could do this with zero copies in the no-clean case by using 212 // strncmp below. 213 char Directive[20]; 214 unsigned IdLen; 215 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 216 IdLen = Tok.getLength(); 217 memcpy(Directive, RawCharData, IdLen); 218 Directive[IdLen] = 0; 219 } else { 220 std::string DirectiveStr = getSpelling(Tok); 221 IdLen = DirectiveStr.size(); 222 if (IdLen >= 20) { 223 CurPPLexer->ParsingPreprocessorDirective = false; 224 // Restore comment saving mode. 225 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 226 continue; 227 } 228 memcpy(Directive, &DirectiveStr[0], IdLen); 229 Directive[IdLen] = 0; 230 FirstChar = Directive[0]; 231 } 232 233 if (FirstChar == 'i' && Directive[1] == 'f') { 234 if ((IdLen == 2) || // "if" 235 (IdLen == 5 && !strcmp(Directive+2, "def")) || // "ifdef" 236 (IdLen == 6 && !strcmp(Directive+2, "ndef"))) { // "ifndef" 237 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 238 // bother parsing the condition. 239 DiscardUntilEndOfDirective(); 240 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 241 /*foundnonskip*/false, 242 /*fnddelse*/false); 243 } 244 } else if (FirstChar == 'e') { 245 if (IdLen == 5 && !strcmp(Directive+1, "ndif")) { // "endif" 246 CheckEndOfDirective("#endif"); 247 PPConditionalInfo CondInfo; 248 CondInfo.WasSkipping = true; // Silence bogus warning. 249 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 250 InCond = InCond; // Silence warning in no-asserts mode. 251 assert(!InCond && "Can't be skipping if not in a conditional!"); 252 253 // If we popped the outermost skipping block, we're done skipping! 254 if (!CondInfo.WasSkipping) 255 break; 256 } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else". 257 // #else directive in a skipping conditional. If not in some other 258 // skipping conditional, and if #else hasn't already been seen, enter it 259 // as a non-skipping conditional. 260 CheckEndOfDirective("#else"); 261 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 262 263 // If this is a #else with a #else before it, report the error. 264 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 265 266 // Note that we've seen a #else in this conditional. 267 CondInfo.FoundElse = true; 268 269 // If the conditional is at the top level, and the #if block wasn't 270 // entered, enter the #else block now. 271 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 272 CondInfo.FoundNonSkip = true; 273 break; 274 } 275 } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) { // "elif". 276 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 277 278 bool ShouldEnter; 279 // If this is in a skipping block or if we're already handled this #if 280 // block, don't bother parsing the condition. 281 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 282 DiscardUntilEndOfDirective(); 283 ShouldEnter = false; 284 } else { 285 // Restore the value of LexingRawMode so that identifiers are 286 // looked up, etc, inside the #elif expression. 287 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 288 CurPPLexer->LexingRawMode = false; 289 IdentifierInfo *IfNDefMacro = 0; 290 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 291 CurPPLexer->LexingRawMode = true; 292 } 293 294 // If this is a #elif with a #else before it, report the error. 295 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 296 297 // If this condition is true, enter it! 298 if (ShouldEnter) { 299 CondInfo.FoundNonSkip = true; 300 break; 301 } 302 } 303 } 304 305 CurPPLexer->ParsingPreprocessorDirective = false; 306 // Restore comment saving mode. 307 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 308 } 309 310 // Finally, if we are out of the conditional (saw an #endif or ran off the end 311 // of the file, just stop skipping and return to lexing whatever came after 312 // the #if block. 313 CurPPLexer->LexingRawMode = false; 314} 315 316void Preprocessor::PTHSkipExcludedConditionalBlock() { 317 318 while(1) { 319 assert(CurPTHLexer); 320 assert(CurPTHLexer->LexingRawMode == false); 321 322 // Skip to the next '#else', '#elif', or #endif. 323 if (CurPTHLexer->SkipBlock()) { 324 // We have reached an #endif. Both the '#' and 'endif' tokens 325 // have been consumed by the PTHLexer. Just pop off the condition level. 326 PPConditionalInfo CondInfo; 327 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 328 InCond = InCond; // Silence warning in no-asserts mode. 329 assert(!InCond && "Can't be skipping if not in a conditional!"); 330 break; 331 } 332 333 // We have reached a '#else' or '#elif'. Lex the next token to get 334 // the directive flavor. 335 Token Tok; 336 LexUnexpandedToken(Tok); 337 338 // We can actually look up the IdentifierInfo here since we aren't in 339 // raw mode. 340 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 341 342 if (K == tok::pp_else) { 343 // #else: Enter the else condition. We aren't in a nested condition 344 // since we skip those. We're always in the one matching the last 345 // blocked we skipped. 346 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 347 // Note that we've seen a #else in this conditional. 348 CondInfo.FoundElse = true; 349 350 // If the #if block wasn't entered then enter the #else block now. 351 if (!CondInfo.FoundNonSkip) { 352 CondInfo.FoundNonSkip = true; 353 354 // Consume the eom token. 355 CurPTHLexer->ParsingPreprocessorDirective = true; 356 LexUnexpandedToken(Tok); 357 assert(Tok.is(tok::eom)); 358 CurPTHLexer->ParsingPreprocessorDirective = false; 359 360 break; 361 } 362 363 // Otherwise skip this block. 364 continue; 365 } 366 367 assert(K == tok::pp_elif); 368 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 369 370 // If this is a #elif with a #else before it, report the error. 371 if (CondInfo.FoundElse) 372 Diag(Tok, diag::pp_err_elif_after_else); 373 374 // If this is in a skipping block or if we're already handled this #if 375 // block, don't bother parsing the condition. We just skip this block. 376 if (CondInfo.FoundNonSkip) 377 continue; 378 379 // Evaluate the condition of the #elif. 380 IdentifierInfo *IfNDefMacro = 0; 381 CurPTHLexer->ParsingPreprocessorDirective = true; 382 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 383 CurPTHLexer->ParsingPreprocessorDirective = false; 384 385 // If this condition is true, enter it! 386 if (ShouldEnter) { 387 CondInfo.FoundNonSkip = true; 388 break; 389 } 390 391 // Otherwise, skip this block and go to the next one. 392 continue; 393 } 394} 395 396/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 397/// return null on failure. isAngled indicates whether the file reference is 398/// for system #include's or not (i.e. using <> instead of ""). 399const FileEntry *Preprocessor::LookupFile(const char *FilenameStart, 400 const char *FilenameEnd, 401 bool isAngled, 402 const DirectoryLookup *FromDir, 403 const DirectoryLookup *&CurDir) { 404 // If the header lookup mechanism may be relative to the current file, pass in 405 // info about where the current file is. 406 const FileEntry *CurFileEnt = 0; 407 if (!FromDir) { 408 FileID FID = getCurrentFileLexer()->getFileID(); 409 CurFileEnt = SourceMgr.getFileEntryForID(FID); 410 411 // If there is no file entry associated with this file, it must be the 412 // predefines buffer. Any other file is not lexed with a normal lexer, so 413 // it won't be scanned for preprocessor directives. If we have the 414 // predefines buffer, resolve #include references (which come from the 415 // -include command line argument) as if they came from the main file, this 416 // affects file lookup etc. 417 if (CurFileEnt == 0) { 418 FID = SourceMgr.getMainFileID(); 419 CurFileEnt = SourceMgr.getFileEntryForID(FID); 420 } 421 } 422 423 // Do a standard file entry lookup. 424 CurDir = CurDirLookup; 425 const FileEntry *FE = 426 HeaderInfo.LookupFile(FilenameStart, FilenameEnd, 427 isAngled, FromDir, CurDir, CurFileEnt); 428 if (FE) return FE; 429 430 // Otherwise, see if this is a subframework header. If so, this is relative 431 // to one of the headers on the #include stack. Walk the list of the current 432 // headers on the #include stack and pass them to HeaderInfo. 433 if (IsFileLexer()) { 434 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 435 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd, 436 CurFileEnt))) 437 return FE; 438 } 439 440 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 441 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 442 if (IsFileLexer(ISEntry)) { 443 if ((CurFileEnt = 444 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 445 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, 446 FilenameEnd, CurFileEnt))) 447 return FE; 448 } 449 } 450 451 // Otherwise, we really couldn't find the file. 452 return 0; 453} 454 455 456//===----------------------------------------------------------------------===// 457// Preprocessor Directive Handling. 458//===----------------------------------------------------------------------===// 459 460/// HandleDirective - This callback is invoked when the lexer sees a # token 461/// at the start of a line. This consumes the directive, modifies the 462/// lexer/preprocessor state, and advances the lexer(s) so that the next token 463/// read is the correct one. 464void Preprocessor::HandleDirective(Token &Result) { 465 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 466 467 // We just parsed a # character at the start of a line, so we're in directive 468 // mode. Tell the lexer this so any newlines we see will be converted into an 469 // EOM token (which terminates the directive). 470 CurPPLexer->ParsingPreprocessorDirective = true; 471 472 ++NumDirectives; 473 474 // We are about to read a token. For the multiple-include optimization FA to 475 // work, we have to remember if we had read any tokens *before* this 476 // pp-directive. 477 bool ReadAnyTokensBeforeDirective = CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 478 479 // Save the '#' token in case we need to return it later. 480 Token SavedHash = Result; 481 482 // Read the next token, the directive flavor. This isn't expanded due to 483 // C99 6.10.3p8. 484 LexUnexpandedToken(Result); 485 486 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 487 // #define A(x) #x 488 // A(abc 489 // #warning blah 490 // def) 491 // If so, the user is relying on non-portable behavior, emit a diagnostic. 492 if (InMacroArgs) 493 Diag(Result, diag::ext_embedded_directive); 494 495TryAgain: 496 switch (Result.getKind()) { 497 case tok::eom: 498 return; // null directive. 499 case tok::comment: 500 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 501 LexUnexpandedToken(Result); 502 goto TryAgain; 503 504 case tok::numeric_constant: // # 7 GNU line marker directive. 505 if (getLangOptions().AsmPreprocessor) 506 break; // # 4 is not a preprocessor directive in .S files. 507 return HandleDigitDirective(Result); 508 default: 509 IdentifierInfo *II = Result.getIdentifierInfo(); 510 if (II == 0) break; // Not an identifier. 511 512 // Ask what the preprocessor keyword ID is. 513 switch (II->getPPKeywordID()) { 514 default: break; 515 // C99 6.10.1 - Conditional Inclusion. 516 case tok::pp_if: 517 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 518 case tok::pp_ifdef: 519 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 520 case tok::pp_ifndef: 521 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 522 case tok::pp_elif: 523 return HandleElifDirective(Result); 524 case tok::pp_else: 525 return HandleElseDirective(Result); 526 case tok::pp_endif: 527 return HandleEndifDirective(Result); 528 529 // C99 6.10.2 - Source File Inclusion. 530 case tok::pp_include: 531 return HandleIncludeDirective(Result); // Handle #include. 532 case tok::pp___include_macros: 533 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 534 535 // C99 6.10.3 - Macro Replacement. 536 case tok::pp_define: 537 return HandleDefineDirective(Result); 538 case tok::pp_undef: 539 return HandleUndefDirective(Result); 540 541 // C99 6.10.4 - Line Control. 542 case tok::pp_line: 543 return HandleLineDirective(Result); 544 545 // C99 6.10.5 - Error Directive. 546 case tok::pp_error: 547 return HandleUserDiagnosticDirective(Result, false); 548 549 // C99 6.10.6 - Pragma Directive. 550 case tok::pp_pragma: 551 return HandlePragmaDirective(); 552 553 // GNU Extensions. 554 case tok::pp_import: 555 return HandleImportDirective(Result); 556 case tok::pp_include_next: 557 return HandleIncludeNextDirective(Result); 558 559 case tok::pp_warning: 560 Diag(Result, diag::ext_pp_warning_directive); 561 return HandleUserDiagnosticDirective(Result, true); 562 case tok::pp_ident: 563 return HandleIdentSCCSDirective(Result); 564 case tok::pp_sccs: 565 return HandleIdentSCCSDirective(Result); 566 case tok::pp_assert: 567 //isExtension = true; // FIXME: implement #assert 568 break; 569 case tok::pp_unassert: 570 //isExtension = true; // FIXME: implement #unassert 571 break; 572 } 573 break; 574 } 575 576 // If this is a .S file, treat unknown # directives as non-preprocessor 577 // directives. This is important because # may be a comment or introduce 578 // various pseudo-ops. Just return the # token and push back the following 579 // token to be lexed next time. 580 if (getLangOptions().AsmPreprocessor) { 581 Token *Toks = new Token[2](); 582 // Return the # and the token after it. 583 Toks[0] = SavedHash; 584 Toks[1] = Result; 585 // Enter this token stream so that we re-lex the tokens. Make sure to 586 // enable macro expansion, in case the token after the # is an identifier 587 // that is expanded. 588 EnterTokenStream(Toks, 2, false, true); 589 return; 590 } 591 592 // If we reached here, the preprocessing token is not valid! 593 Diag(Result, diag::err_pp_invalid_directive); 594 595 // Read the rest of the PP line. 596 DiscardUntilEndOfDirective(); 597 598 // Okay, we're done parsing the directive. 599} 600 601/// GetLineValue - Convert a numeric token into an unsigned value, emitting 602/// Diagnostic DiagID if it is invalid, and returning the value in Val. 603static bool GetLineValue(Token &DigitTok, unsigned &Val, 604 unsigned DiagID, Preprocessor &PP) { 605 if (DigitTok.isNot(tok::numeric_constant)) { 606 PP.Diag(DigitTok, DiagID); 607 608 if (DigitTok.isNot(tok::eom)) 609 PP.DiscardUntilEndOfDirective(); 610 return true; 611 } 612 613 llvm::SmallString<64> IntegerBuffer; 614 IntegerBuffer.resize(DigitTok.getLength()); 615 const char *DigitTokBegin = &IntegerBuffer[0]; 616 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin); 617 NumericLiteralParser Literal(DigitTokBegin, DigitTokBegin+ActualLength, 618 DigitTok.getLocation(), PP); 619 if (Literal.hadError) 620 return true; // Error already emitted. 621 622 if (Literal.isFloatingLiteral() || Literal.isImaginary) { 623 PP.Diag(DigitTok, DiagID); 624 return true; 625 } 626 627 // Parse the integer literal into Result. 628 llvm::APInt APVal(32, 0); 629 if (Literal.GetIntegerValue(APVal)) { 630 // Overflow parsing integer literal. 631 PP.Diag(DigitTok, DiagID); 632 return true; 633 } 634 Val = APVal.getZExtValue(); 635 636 // Reject 0, this is needed both by #line numbers and flags. 637 if (Val == 0) { 638 PP.Diag(DigitTok, DiagID); 639 PP.DiscardUntilEndOfDirective(); 640 return true; 641 } 642 643 return false; 644} 645 646/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 647/// acceptable forms are: 648/// # line digit-sequence 649/// # line digit-sequence "s-char-sequence" 650void Preprocessor::HandleLineDirective(Token &Tok) { 651 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 652 // expanded. 653 Token DigitTok; 654 Lex(DigitTok); 655 656 // Validate the number and convert it to an unsigned. 657 unsigned LineNo; 658 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer, *this)) 659 return; 660 661 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 662 // number greater than 2147483647". C90 requires that the line # be <= 32767. 663 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 664 if (LineNo >= LineLimit) 665 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 666 667 int FilenameID = -1; 668 Token StrTok; 669 Lex(StrTok); 670 671 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 672 // string followed by eom. 673 if (StrTok.is(tok::eom)) 674 ; // ok 675 else if (StrTok.isNot(tok::string_literal)) { 676 Diag(StrTok, diag::err_pp_line_invalid_filename); 677 DiscardUntilEndOfDirective(); 678 return; 679 } else { 680 // Parse and validate the string, converting it into a unique ID. 681 StringLiteralParser Literal(&StrTok, 1, *this); 682 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 683 if (Literal.hadError) 684 return DiscardUntilEndOfDirective(); 685 if (Literal.Pascal) { 686 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 687 return DiscardUntilEndOfDirective(); 688 } 689 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 690 Literal.GetStringLength()); 691 692 // Verify that there is nothing after the string, other than EOM. 693 CheckEndOfDirective("#line"); 694 } 695 696 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 697 698 if (Callbacks) 699 Callbacks->FileChanged(DigitTok.getLocation(), PPCallbacks::RenameFile, 700 SrcMgr::C_User); 701} 702 703/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 704/// marker directive. 705static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 706 bool &IsSystemHeader, bool &IsExternCHeader, 707 Preprocessor &PP) { 708 unsigned FlagVal; 709 Token FlagTok; 710 PP.Lex(FlagTok); 711 if (FlagTok.is(tok::eom)) return false; 712 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 713 return true; 714 715 if (FlagVal == 1) { 716 IsFileEntry = true; 717 718 PP.Lex(FlagTok); 719 if (FlagTok.is(tok::eom)) return false; 720 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 721 return true; 722 } else if (FlagVal == 2) { 723 IsFileExit = true; 724 725 SourceManager &SM = PP.getSourceManager(); 726 // If we are leaving the current presumed file, check to make sure the 727 // presumed include stack isn't empty! 728 FileID CurFileID = 729 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 730 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 731 732 // If there is no include loc (main file) or if the include loc is in a 733 // different physical file, then we aren't in a "1" line marker flag region. 734 SourceLocation IncLoc = PLoc.getIncludeLoc(); 735 if (IncLoc.isInvalid() || 736 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 737 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 738 PP.DiscardUntilEndOfDirective(); 739 return true; 740 } 741 742 PP.Lex(FlagTok); 743 if (FlagTok.is(tok::eom)) return false; 744 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 745 return true; 746 } 747 748 // We must have 3 if there are still flags. 749 if (FlagVal != 3) { 750 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 751 PP.DiscardUntilEndOfDirective(); 752 return true; 753 } 754 755 IsSystemHeader = true; 756 757 PP.Lex(FlagTok); 758 if (FlagTok.is(tok::eom)) return false; 759 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 760 return true; 761 762 // We must have 4 if there is yet another flag. 763 if (FlagVal != 4) { 764 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 765 PP.DiscardUntilEndOfDirective(); 766 return true; 767 } 768 769 IsExternCHeader = true; 770 771 PP.Lex(FlagTok); 772 if (FlagTok.is(tok::eom)) return false; 773 774 // There are no more valid flags here. 775 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 776 PP.DiscardUntilEndOfDirective(); 777 return true; 778} 779 780/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 781/// one of the following forms: 782/// 783/// # 42 784/// # 42 "file" ('1' | '2')? 785/// # 42 "file" ('1' | '2')? '3' '4'? 786/// 787void Preprocessor::HandleDigitDirective(Token &DigitTok) { 788 // Validate the number and convert it to an unsigned. GNU does not have a 789 // line # limit other than it fit in 32-bits. 790 unsigned LineNo; 791 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 792 *this)) 793 return; 794 795 Token StrTok; 796 Lex(StrTok); 797 798 bool IsFileEntry = false, IsFileExit = false; 799 bool IsSystemHeader = false, IsExternCHeader = false; 800 int FilenameID = -1; 801 802 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 803 // string followed by eom. 804 if (StrTok.is(tok::eom)) 805 ; // ok 806 else if (StrTok.isNot(tok::string_literal)) { 807 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 808 return DiscardUntilEndOfDirective(); 809 } else { 810 // Parse and validate the string, converting it into a unique ID. 811 StringLiteralParser Literal(&StrTok, 1, *this); 812 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 813 if (Literal.hadError) 814 return DiscardUntilEndOfDirective(); 815 if (Literal.Pascal) { 816 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 817 return DiscardUntilEndOfDirective(); 818 } 819 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 820 Literal.GetStringLength()); 821 822 // If a filename was present, read any flags that are present. 823 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 824 IsSystemHeader, IsExternCHeader, *this)) 825 return; 826 } 827 828 // Create a line note with this information. 829 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 830 IsFileEntry, IsFileExit, 831 IsSystemHeader, IsExternCHeader); 832 833 // If the preprocessor has callbacks installed, notify them of the #line 834 // change. This is used so that the line marker comes out in -E mode for 835 // example. 836 if (Callbacks) { 837 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 838 if (IsFileEntry) 839 Reason = PPCallbacks::EnterFile; 840 else if (IsFileExit) 841 Reason = PPCallbacks::ExitFile; 842 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 843 if (IsExternCHeader) 844 FileKind = SrcMgr::C_ExternCSystem; 845 else if (IsSystemHeader) 846 FileKind = SrcMgr::C_System; 847 848 Callbacks->FileChanged(DigitTok.getLocation(), Reason, FileKind); 849 } 850} 851 852 853/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 854/// 855void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 856 bool isWarning) { 857 // PTH doesn't emit #warning or #error directives. 858 if (CurPTHLexer) 859 return CurPTHLexer->DiscardToEndOfLine(); 860 861 // Read the rest of the line raw. We do this because we don't want macros 862 // to be expanded and we don't require that the tokens be valid preprocessing 863 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 864 // collapse multiple consequtive white space between tokens, but this isn't 865 // specified by the standard. 866 std::string Message = CurLexer->ReadToEndOfLine(); 867 if (isWarning) 868 Diag(Tok, diag::pp_hash_warning) << Message; 869 else 870 Diag(Tok, diag::err_pp_hash_error) << Message; 871} 872 873/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 874/// 875void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 876 // Yes, this directive is an extension. 877 Diag(Tok, diag::ext_pp_ident_directive); 878 879 // Read the string argument. 880 Token StrTok; 881 Lex(StrTok); 882 883 // If the token kind isn't a string, it's a malformed directive. 884 if (StrTok.isNot(tok::string_literal) && 885 StrTok.isNot(tok::wide_string_literal)) { 886 Diag(StrTok, diag::err_pp_malformed_ident); 887 if (StrTok.isNot(tok::eom)) 888 DiscardUntilEndOfDirective(); 889 return; 890 } 891 892 // Verify that there is nothing after the string, other than EOM. 893 CheckEndOfDirective("#ident"); 894 895 if (Callbacks) 896 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok)); 897} 898 899//===----------------------------------------------------------------------===// 900// Preprocessor Include Directive Handling. 901//===----------------------------------------------------------------------===// 902 903/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 904/// checked and spelled filename, e.g. as an operand of #include. This returns 905/// true if the input filename was in <>'s or false if it were in ""'s. The 906/// caller is expected to provide a buffer that is large enough to hold the 907/// spelling of the filename, but is also expected to handle the case when 908/// this method decides to use a different buffer. 909bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 910 const char *&BufStart, 911 const char *&BufEnd) { 912 // Get the text form of the filename. 913 assert(BufStart != BufEnd && "Can't have tokens with empty spellings!"); 914 915 // Make sure the filename is <x> or "x". 916 bool isAngled; 917 if (BufStart[0] == '<') { 918 if (BufEnd[-1] != '>') { 919 Diag(Loc, diag::err_pp_expects_filename); 920 BufStart = 0; 921 return true; 922 } 923 isAngled = true; 924 } else if (BufStart[0] == '"') { 925 if (BufEnd[-1] != '"') { 926 Diag(Loc, diag::err_pp_expects_filename); 927 BufStart = 0; 928 return true; 929 } 930 isAngled = false; 931 } else { 932 Diag(Loc, diag::err_pp_expects_filename); 933 BufStart = 0; 934 return true; 935 } 936 937 // Diagnose #include "" as invalid. 938 if (BufEnd-BufStart <= 2) { 939 Diag(Loc, diag::err_pp_empty_filename); 940 BufStart = 0; 941 return ""; 942 } 943 944 // Skip the brackets. 945 ++BufStart; 946 --BufEnd; 947 return isAngled; 948} 949 950/// ConcatenateIncludeName - Handle cases where the #include name is expanded 951/// from a macro as multiple tokens, which need to be glued together. This 952/// occurs for code like: 953/// #define FOO <a/b.h> 954/// #include FOO 955/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 956/// 957/// This code concatenates and consumes tokens up to the '>' token. It returns 958/// false if the > was found, otherwise it returns true if it finds and consumes 959/// the EOM marker. 960static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer, 961 Preprocessor &PP) { 962 Token CurTok; 963 964 PP.Lex(CurTok); 965 while (CurTok.isNot(tok::eom)) { 966 // Append the spelling of this token to the buffer. If there was a space 967 // before it, add it now. 968 if (CurTok.hasLeadingSpace()) 969 FilenameBuffer.push_back(' '); 970 971 // Get the spelling of the token, directly into FilenameBuffer if possible. 972 unsigned PreAppendSize = FilenameBuffer.size(); 973 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 974 975 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 976 unsigned ActualLen = PP.getSpelling(CurTok, BufPtr); 977 978 // If the token was spelled somewhere else, copy it into FilenameBuffer. 979 if (BufPtr != &FilenameBuffer[PreAppendSize]) 980 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 981 982 // Resize FilenameBuffer to the correct size. 983 if (CurTok.getLength() != ActualLen) 984 FilenameBuffer.resize(PreAppendSize+ActualLen); 985 986 // If we found the '>' marker, return success. 987 if (CurTok.is(tok::greater)) 988 return false; 989 990 PP.Lex(CurTok); 991 } 992 993 // If we hit the eom marker, emit an error and return true so that the caller 994 // knows the EOM has been read. 995 PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 996 return true; 997} 998 999/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1000/// file to be included from the lexer, then include it! This is a common 1001/// routine with functionality shared between #include, #include_next and 1002/// #import. LookupFrom is set when this is a #include_next directive, it 1003/// specifies the file to start searching from. 1004void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1005 const DirectoryLookup *LookupFrom, 1006 bool isImport) { 1007 1008 Token FilenameTok; 1009 CurPPLexer->LexIncludeFilename(FilenameTok); 1010 1011 // Reserve a buffer to get the spelling. 1012 llvm::SmallVector<char, 128> FilenameBuffer; 1013 const char *FilenameStart, *FilenameEnd; 1014 1015 switch (FilenameTok.getKind()) { 1016 case tok::eom: 1017 // If the token kind is EOM, the error has already been diagnosed. 1018 return; 1019 1020 case tok::angle_string_literal: 1021 case tok::string_literal: { 1022 FilenameBuffer.resize(FilenameTok.getLength()); 1023 FilenameStart = &FilenameBuffer[0]; 1024 unsigned Len = getSpelling(FilenameTok, FilenameStart); 1025 FilenameEnd = FilenameStart+Len; 1026 break; 1027 } 1028 1029 case tok::less: 1030 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1031 // case, glue the tokens together into FilenameBuffer and interpret those. 1032 FilenameBuffer.push_back('<'); 1033 if (ConcatenateIncludeName(FilenameBuffer, *this)) 1034 return; // Found <eom> but no ">"? Diagnostic already emitted. 1035 FilenameStart = &FilenameBuffer[0]; 1036 FilenameEnd = &FilenameBuffer[FilenameBuffer.size()]; 1037 break; 1038 default: 1039 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1040 DiscardUntilEndOfDirective(); 1041 return; 1042 } 1043 1044 bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(), 1045 FilenameStart, FilenameEnd); 1046 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1047 // error. 1048 if (FilenameStart == 0) { 1049 DiscardUntilEndOfDirective(); 1050 return; 1051 } 1052 1053 // Verify that there is nothing after the filename, other than EOM. Use the 1054 // preprocessor to lex this in case lexing the filename entered a macro. 1055 CheckEndOfDirective("#include"); 1056 1057 // Check that we don't have infinite #include recursion. 1058 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1059 Diag(FilenameTok, diag::err_pp_include_too_deep); 1060 return; 1061 } 1062 1063 // Search include directories. 1064 const DirectoryLookup *CurDir; 1065 const FileEntry *File = LookupFile(FilenameStart, FilenameEnd, 1066 isAngled, LookupFrom, CurDir); 1067 if (File == 0) { 1068 Diag(FilenameTok, diag::err_pp_file_not_found) 1069 << std::string(FilenameStart, FilenameEnd); 1070 return; 1071 } 1072 1073 // Ask HeaderInfo if we should enter this #include file. If not, #including 1074 // this file will have no effect. 1075 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) 1076 return; 1077 1078 // The #included file will be considered to be a system header if either it is 1079 // in a system include directory, or if the #includer is a system include 1080 // header. 1081 SrcMgr::CharacteristicKind FileCharacter = 1082 std::max(HeaderInfo.getFileDirFlavor(File), 1083 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1084 1085 // Look up the file, create a File ID for it. 1086 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1087 FileCharacter); 1088 if (FID.isInvalid()) { 1089 Diag(FilenameTok, diag::err_pp_file_not_found) 1090 << std::string(FilenameStart, FilenameEnd); 1091 return; 1092 } 1093 1094 // Finally, if all is good, enter the new file! 1095 EnterSourceFile(FID, CurDir); 1096} 1097 1098/// HandleIncludeNextDirective - Implements #include_next. 1099/// 1100void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1101 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1102 1103 // #include_next is like #include, except that we start searching after 1104 // the current found directory. If we can't do this, issue a 1105 // diagnostic. 1106 const DirectoryLookup *Lookup = CurDirLookup; 1107 if (isInPrimaryFile()) { 1108 Lookup = 0; 1109 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1110 } else if (Lookup == 0) { 1111 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1112 } else { 1113 // Start looking up in the next directory. 1114 ++Lookup; 1115 } 1116 1117 return HandleIncludeDirective(IncludeNextTok, Lookup); 1118} 1119 1120/// HandleImportDirective - Implements #import. 1121/// 1122void Preprocessor::HandleImportDirective(Token &ImportTok) { 1123 if (!Features.ObjC1) // #import is standard for ObjC. 1124 Diag(ImportTok, diag::ext_pp_import_directive); 1125 1126 return HandleIncludeDirective(ImportTok, 0, true); 1127} 1128 1129/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1130/// pseudo directive in the predefines buffer. This handles it by sucking all 1131/// tokens through the preprocessor and discarding them (only keeping the side 1132/// effects on the preprocessor). 1133void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1134 // This directive should only occur in the predefines buffer. If not, emit an 1135 // error and reject it. 1136 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1137 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1138 Diag(IncludeMacrosTok.getLocation(), 1139 diag::pp_include_macros_out_of_predefines); 1140 DiscardUntilEndOfDirective(); 1141 return; 1142 } 1143 1144 // TODO: implement me :) 1145 DiscardUntilEndOfDirective(); 1146} 1147 1148//===----------------------------------------------------------------------===// 1149// Preprocessor Macro Directive Handling. 1150//===----------------------------------------------------------------------===// 1151 1152/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1153/// definition has just been read. Lex the rest of the arguments and the 1154/// closing ), updating MI with what we learn. Return true if an error occurs 1155/// parsing the arg list. 1156bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1157 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1158 1159 Token Tok; 1160 while (1) { 1161 LexUnexpandedToken(Tok); 1162 switch (Tok.getKind()) { 1163 case tok::r_paren: 1164 // Found the end of the argument list. 1165 if (Arguments.empty()) // #define FOO() 1166 return false; 1167 // Otherwise we have #define FOO(A,) 1168 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1169 return true; 1170 case tok::ellipsis: // #define X(... -> C99 varargs 1171 // Warn if use of C99 feature in non-C99 mode. 1172 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1173 1174 // Lex the token after the identifier. 1175 LexUnexpandedToken(Tok); 1176 if (Tok.isNot(tok::r_paren)) { 1177 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1178 return true; 1179 } 1180 // Add the __VA_ARGS__ identifier as an argument. 1181 Arguments.push_back(Ident__VA_ARGS__); 1182 MI->setIsC99Varargs(); 1183 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1184 return false; 1185 case tok::eom: // #define X( 1186 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1187 return true; 1188 default: 1189 // Handle keywords and identifiers here to accept things like 1190 // #define Foo(for) for. 1191 IdentifierInfo *II = Tok.getIdentifierInfo(); 1192 if (II == 0) { 1193 // #define X(1 1194 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1195 return true; 1196 } 1197 1198 // If this is already used as an argument, it is used multiple times (e.g. 1199 // #define X(A,A. 1200 if (std::find(Arguments.begin(), Arguments.end(), II) != 1201 Arguments.end()) { // C99 6.10.3p6 1202 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1203 return true; 1204 } 1205 1206 // Add the argument to the macro info. 1207 Arguments.push_back(II); 1208 1209 // Lex the token after the identifier. 1210 LexUnexpandedToken(Tok); 1211 1212 switch (Tok.getKind()) { 1213 default: // #define X(A B 1214 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1215 return true; 1216 case tok::r_paren: // #define X(A) 1217 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1218 return false; 1219 case tok::comma: // #define X(A, 1220 break; 1221 case tok::ellipsis: // #define X(A... -> GCC extension 1222 // Diagnose extension. 1223 Diag(Tok, diag::ext_named_variadic_macro); 1224 1225 // Lex the token after the identifier. 1226 LexUnexpandedToken(Tok); 1227 if (Tok.isNot(tok::r_paren)) { 1228 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1229 return true; 1230 } 1231 1232 MI->setIsGNUVarargs(); 1233 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1234 return false; 1235 } 1236 } 1237 } 1238} 1239 1240/// HandleDefineDirective - Implements #define. This consumes the entire macro 1241/// line then lets the caller lex the next real token. 1242void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1243 ++NumDefined; 1244 1245 Token MacroNameTok; 1246 ReadMacroName(MacroNameTok, 1); 1247 1248 // Error reading macro name? If so, diagnostic already issued. 1249 if (MacroNameTok.is(tok::eom)) 1250 return; 1251 1252 // If we are supposed to keep comments in #defines, reenable comment saving 1253 // mode. 1254 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1255 1256 // Create the new macro. 1257 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1258 1259 Token Tok; 1260 LexUnexpandedToken(Tok); 1261 1262 // If this is a function-like macro definition, parse the argument list, 1263 // marking each of the identifiers as being used as macro arguments. Also, 1264 // check other constraints on the first token of the macro body. 1265 if (Tok.is(tok::eom)) { 1266 // If there is no body to this macro, we have no special handling here. 1267 } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) { 1268 // This is a function-like macro definition. Read the argument list. 1269 MI->setIsFunctionLike(); 1270 if (ReadMacroDefinitionArgList(MI)) { 1271 // Forget about MI. 1272 ReleaseMacroInfo(MI); 1273 // Throw away the rest of the line. 1274 if (CurPPLexer->ParsingPreprocessorDirective) 1275 DiscardUntilEndOfDirective(); 1276 return; 1277 } 1278 1279 // Read the first token after the arg list for down below. 1280 LexUnexpandedToken(Tok); 1281 } else if (!Tok.hasLeadingSpace()) { 1282 // C99 requires whitespace between the macro definition and the body. Emit 1283 // a diagnostic for something like "#define X+". 1284 if (Features.C99) { 1285 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1286 } else { 1287 // FIXME: C90/C++ do not get this diagnostic, but it does get a similar 1288 // one in some cases! 1289 } 1290 } else { 1291 // This is a normal token with leading space. Clear the leading space 1292 // marker on the first token to get proper expansion. 1293 Tok.clearFlag(Token::LeadingSpace); 1294 } 1295 1296 // If this is a definition of a variadic C99 function-like macro, not using 1297 // the GNU named varargs extension, enabled __VA_ARGS__. 1298 1299 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1300 // This gets unpoisoned where it is allowed. 1301 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1302 if (MI->isC99Varargs()) 1303 Ident__VA_ARGS__->setIsPoisoned(false); 1304 1305 // Read the rest of the macro body. 1306 if (MI->isObjectLike()) { 1307 // Object-like macros are very simple, just read their body. 1308 while (Tok.isNot(tok::eom)) { 1309 MI->AddTokenToBody(Tok); 1310 // Get the next token of the macro. 1311 LexUnexpandedToken(Tok); 1312 } 1313 1314 } else { 1315 // Otherwise, read the body of a function-like macro. This has to validate 1316 // the # (stringize) operator. 1317 while (Tok.isNot(tok::eom)) { 1318 MI->AddTokenToBody(Tok); 1319 1320 // Check C99 6.10.3.2p1: ensure that # operators are followed by macro 1321 // parameters in function-like macro expansions. 1322 if (Tok.isNot(tok::hash)) { 1323 // Get the next token of the macro. 1324 LexUnexpandedToken(Tok); 1325 continue; 1326 } 1327 1328 // Get the next token of the macro. 1329 LexUnexpandedToken(Tok); 1330 1331 // Not a macro arg identifier? 1332 if (!Tok.getIdentifierInfo() || 1333 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1334 Diag(Tok, diag::err_pp_stringize_not_parameter); 1335 ReleaseMacroInfo(MI); 1336 1337 // Disable __VA_ARGS__ again. 1338 Ident__VA_ARGS__->setIsPoisoned(true); 1339 return; 1340 } 1341 1342 // Things look ok, add the param name token to the macro. 1343 MI->AddTokenToBody(Tok); 1344 1345 // Get the next token of the macro. 1346 LexUnexpandedToken(Tok); 1347 } 1348 } 1349 1350 1351 // Disable __VA_ARGS__ again. 1352 Ident__VA_ARGS__->setIsPoisoned(true); 1353 1354 // Check that there is no paste (##) operator at the begining or end of the 1355 // replacement list. 1356 unsigned NumTokens = MI->getNumTokens(); 1357 if (NumTokens != 0) { 1358 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1359 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1360 ReleaseMacroInfo(MI); 1361 return; 1362 } 1363 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1364 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1365 ReleaseMacroInfo(MI); 1366 return; 1367 } 1368 } 1369 1370 // If this is the primary source file, remember that this macro hasn't been 1371 // used yet. 1372 if (isInPrimaryFile()) 1373 MI->setIsUsed(false); 1374 1375 // Finally, if this identifier already had a macro defined for it, verify that 1376 // the macro bodies are identical and free the old definition. 1377 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1378 // It is very common for system headers to have tons of macro redefinitions 1379 // and for warnings to be disabled in system headers. If this is the case, 1380 // then don't bother calling MacroInfo::isIdenticalTo. 1381 if (!getDiagnostics().getSuppressSystemWarnings() || 1382 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1383 if (!OtherMI->isUsed()) 1384 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1385 1386 // Macros must be identical. This means all tokes and whitespace 1387 // separation must be the same. C99 6.10.3.2. 1388 if (!MI->isIdenticalTo(*OtherMI, *this)) { 1389 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1390 << MacroNameTok.getIdentifierInfo(); 1391 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1392 } 1393 } 1394 1395 ReleaseMacroInfo(OtherMI); 1396 } 1397 1398 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1399} 1400 1401/// HandleUndefDirective - Implements #undef. 1402/// 1403void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1404 ++NumUndefined; 1405 1406 Token MacroNameTok; 1407 ReadMacroName(MacroNameTok, 2); 1408 1409 // Error reading macro name? If so, diagnostic already issued. 1410 if (MacroNameTok.is(tok::eom)) 1411 return; 1412 1413 // Check to see if this is the last token on the #undef line. 1414 CheckEndOfDirective("#undef"); 1415 1416 // Okay, we finally have a valid identifier to undef. 1417 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1418 1419 // If the macro is not defined, this is a noop undef, just return. 1420 if (MI == 0) return; 1421 1422 if (!MI->isUsed()) 1423 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1424 1425 // Free macro definition. 1426 ReleaseMacroInfo(MI); 1427 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1428} 1429 1430 1431//===----------------------------------------------------------------------===// 1432// Preprocessor Conditional Directive Handling. 1433//===----------------------------------------------------------------------===// 1434 1435/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1436/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1437/// if any tokens have been returned or pp-directives activated before this 1438/// #ifndef has been lexed. 1439/// 1440void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1441 bool ReadAnyTokensBeforeDirective) { 1442 ++NumIf; 1443 Token DirectiveTok = Result; 1444 1445 Token MacroNameTok; 1446 ReadMacroName(MacroNameTok); 1447 1448 // Error reading macro name? If so, diagnostic already issued. 1449 if (MacroNameTok.is(tok::eom)) { 1450 // Skip code until we get to #endif. This helps with recovery by not 1451 // emitting an error when the #endif is reached. 1452 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1453 /*Foundnonskip*/false, /*FoundElse*/false); 1454 return; 1455 } 1456 1457 // Check to see if this is the last token on the #if[n]def line. 1458 CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef"); 1459 1460 if (CurPPLexer->getConditionalStackDepth() == 0) { 1461 // If the start of a top-level #ifdef, inform MIOpt. 1462 if (!ReadAnyTokensBeforeDirective) { 1463 assert(isIfndef && "#ifdef shouldn't reach here"); 1464 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo()); 1465 } else 1466 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1467 } 1468 1469 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1470 MacroInfo *MI = getMacroInfo(MII); 1471 1472 // If there is a macro, process it. 1473 if (MI) // Mark it used. 1474 MI->setIsUsed(true); 1475 1476 // Should we include the stuff contained by this directive? 1477 if (!MI == isIfndef) { 1478 // Yes, remember that we are inside a conditional, then lex the next token. 1479 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false, 1480 /*foundnonskip*/true, /*foundelse*/false); 1481 } else { 1482 // No, skip the contents of this block and return the first token after it. 1483 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1484 /*Foundnonskip*/false, 1485 /*FoundElse*/false); 1486 } 1487} 1488 1489/// HandleIfDirective - Implements the #if directive. 1490/// 1491void Preprocessor::HandleIfDirective(Token &IfToken, 1492 bool ReadAnyTokensBeforeDirective) { 1493 ++NumIf; 1494 1495 // Parse and evaluation the conditional expression. 1496 IdentifierInfo *IfNDefMacro = 0; 1497 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1498 1499 1500 // If this condition is equivalent to #ifndef X, and if this is the first 1501 // directive seen, handle it for the multiple-include optimization. 1502 if (CurPPLexer->getConditionalStackDepth() == 0) { 1503 if (!ReadAnyTokensBeforeDirective && IfNDefMacro) 1504 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1505 else 1506 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1507 } 1508 1509 // Should we include the stuff contained by this directive? 1510 if (ConditionalTrue) { 1511 // Yes, remember that we are inside a conditional, then lex the next token. 1512 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1513 /*foundnonskip*/true, /*foundelse*/false); 1514 } else { 1515 // No, skip the contents of this block and return the first token after it. 1516 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1517 /*FoundElse*/false); 1518 } 1519} 1520 1521/// HandleEndifDirective - Implements the #endif directive. 1522/// 1523void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1524 ++NumEndif; 1525 1526 // Check that this is the whole directive. 1527 CheckEndOfDirective("#endif"); 1528 1529 PPConditionalInfo CondInfo; 1530 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1531 // No conditionals on the stack: this is an #endif without an #if. 1532 Diag(EndifToken, diag::err_pp_endif_without_if); 1533 return; 1534 } 1535 1536 // If this the end of a top-level #endif, inform MIOpt. 1537 if (CurPPLexer->getConditionalStackDepth() == 0) 1538 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1539 1540 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1541 "This code should only be reachable in the non-skipping case!"); 1542} 1543 1544 1545void Preprocessor::HandleElseDirective(Token &Result) { 1546 ++NumElse; 1547 1548 // #else directive in a non-skipping conditional... start skipping. 1549 CheckEndOfDirective("#else"); 1550 1551 PPConditionalInfo CI; 1552 if (CurPPLexer->popConditionalLevel(CI)) { 1553 Diag(Result, diag::pp_err_else_without_if); 1554 return; 1555 } 1556 1557 // If this is a top-level #else, inform the MIOpt. 1558 if (CurPPLexer->getConditionalStackDepth() == 0) 1559 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1560 1561 // If this is a #else with a #else before it, report the error. 1562 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1563 1564 // Finally, skip the rest of the contents of this block and return the first 1565 // token after it. 1566 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1567 /*FoundElse*/true); 1568} 1569 1570void Preprocessor::HandleElifDirective(Token &ElifToken) { 1571 ++NumElse; 1572 1573 // #elif directive in a non-skipping conditional... start skipping. 1574 // We don't care what the condition is, because we will always skip it (since 1575 // the block immediately before it was included). 1576 DiscardUntilEndOfDirective(); 1577 1578 PPConditionalInfo CI; 1579 if (CurPPLexer->popConditionalLevel(CI)) { 1580 Diag(ElifToken, diag::pp_err_elif_without_if); 1581 return; 1582 } 1583 1584 // If this is a top-level #elif, inform the MIOpt. 1585 if (CurPPLexer->getConditionalStackDepth() == 0) 1586 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1587 1588 // If this is a #elif with a #else before it, report the error. 1589 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1590 1591 // Finally, skip the rest of the contents of this block and return the first 1592 // token after it. 1593 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1594 /*FoundElse*/CI.FoundElse); 1595} 1596 1597