PPDirectives.cpp revision 6fbe3ebeaef08665a37423f8425314c90b8b5bcf
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/FileManager.h" 20#include "clang/Basic/SourceManager.h" 21#include "llvm/ADT/APInt.h" 22using namespace clang; 23 24//===----------------------------------------------------------------------===// 25// Utility Methods for Preprocessor Directive Handling. 26//===----------------------------------------------------------------------===// 27 28MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 29 MacroInfo *MI; 30 31 if (!MICache.empty()) { 32 MI = MICache.back(); 33 MICache.pop_back(); 34 } else 35 MI = (MacroInfo*) BP.Allocate<MacroInfo>(); 36 new (MI) MacroInfo(L); 37 return MI; 38} 39 40/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 41/// be reused for allocating new MacroInfo objects. 42void Preprocessor::ReleaseMacroInfo(MacroInfo* MI) { 43 MICache.push_back(MI); 44 MI->FreeArgumentList(BP); 45} 46 47 48/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 49/// current line until the tok::eom token is found. 50void Preprocessor::DiscardUntilEndOfDirective() { 51 Token Tmp; 52 do { 53 LexUnexpandedToken(Tmp); 54 } while (Tmp.isNot(tok::eom)); 55} 56 57/// ReadMacroName - Lex and validate a macro name, which occurs after a 58/// #define or #undef. This sets the token kind to eom and discards the rest 59/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 60/// this is due to a a #define, 2 if #undef directive, 0 if it is something 61/// else (e.g. #ifdef). 62void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 63 // Read the token, don't allow macro expansion on it. 64 LexUnexpandedToken(MacroNameTok); 65 66 // Missing macro name? 67 if (MacroNameTok.is(tok::eom)) { 68 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 69 return; 70 } 71 72 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 73 if (II == 0) { 74 bool Invalid = false; 75 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 76 if (Invalid) 77 return; 78 79 const IdentifierInfo &Info = Identifiers.get(Spelling); 80 if (Info.isCPlusPlusOperatorKeyword()) 81 // C++ 2.5p2: Alternative tokens behave the same as its primary token 82 // except for their spellings. 83 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 84 else 85 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 86 // Fall through on error. 87 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 88 // Error if defining "defined": C99 6.10.8.4. 89 Diag(MacroNameTok, diag::err_defined_macro_name); 90 } else if (isDefineUndef && II->hasMacroDefinition() && 91 getMacroInfo(II)->isBuiltinMacro()) { 92 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 93 if (isDefineUndef == 1) 94 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 95 else 96 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 97 } else { 98 // Okay, we got a good identifier node. Return it. 99 return; 100 } 101 102 // Invalid macro name, read and discard the rest of the line. Then set the 103 // token kind to tok::eom. 104 MacroNameTok.setKind(tok::eom); 105 return DiscardUntilEndOfDirective(); 106} 107 108/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 109/// not, emit a diagnostic and consume up until the eom. If EnableMacros is 110/// true, then we consider macros that expand to zero tokens as being ok. 111void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 112 Token Tmp; 113 // Lex unexpanded tokens for most directives: macros might expand to zero 114 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 115 // #line) allow empty macros. 116 if (EnableMacros) 117 Lex(Tmp); 118 else 119 LexUnexpandedToken(Tmp); 120 121 // There should be no tokens after the directive, but we allow them as an 122 // extension. 123 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 124 LexUnexpandedToken(Tmp); 125 126 if (Tmp.isNot(tok::eom)) { 127 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 128 // because it is more trouble than it is worth to insert /**/ and check that 129 // there is no /**/ in the range also. 130 FixItHint Hint; 131 if (Features.GNUMode || Features.C99 || Features.CPlusPlus) 132 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 133 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 134 DiscardUntilEndOfDirective(); 135 } 136} 137 138 139 140/// SkipExcludedConditionalBlock - We just read a #if or related directive and 141/// decided that the subsequent tokens are in the #if'd out portion of the 142/// file. Lex the rest of the file, until we see an #endif. If 143/// FoundNonSkipPortion is true, then we have already emitted code for part of 144/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 145/// is true, then #else directives are ok, if not, then we have already seen one 146/// so a #else directive is a duplicate. When this returns, the caller can lex 147/// the first valid token. 148void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 149 bool FoundNonSkipPortion, 150 bool FoundElse) { 151 ++NumSkipped; 152 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 153 154 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 155 FoundNonSkipPortion, FoundElse); 156 157 if (CurPTHLexer) { 158 PTHSkipExcludedConditionalBlock(); 159 return; 160 } 161 162 // Enter raw mode to disable identifier lookup (and thus macro expansion), 163 // disabling warnings, etc. 164 CurPPLexer->LexingRawMode = true; 165 Token Tok; 166 while (1) { 167 CurLexer->Lex(Tok); 168 169 // If this is the end of the buffer, we have an error. 170 if (Tok.is(tok::eof)) { 171 // Emit errors for each unterminated conditional on the stack, including 172 // the current one. 173 while (!CurPPLexer->ConditionalStack.empty()) { 174 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 175 diag::err_pp_unterminated_conditional); 176 CurPPLexer->ConditionalStack.pop_back(); 177 } 178 179 // Just return and let the caller lex after this #include. 180 break; 181 } 182 183 // If this token is not a preprocessor directive, just skip it. 184 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 185 continue; 186 187 // We just parsed a # character at the start of a line, so we're in 188 // directive mode. Tell the lexer this so any newlines we see will be 189 // converted into an EOM token (this terminates the macro). 190 CurPPLexer->ParsingPreprocessorDirective = true; 191 if (CurLexer) CurLexer->SetCommentRetentionState(false); 192 193 194 // Read the next token, the directive flavor. 195 LexUnexpandedToken(Tok); 196 197 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 198 // something bogus), skip it. 199 if (Tok.isNot(tok::identifier)) { 200 CurPPLexer->ParsingPreprocessorDirective = false; 201 // Restore comment saving mode. 202 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 203 continue; 204 } 205 206 // If the first letter isn't i or e, it isn't intesting to us. We know that 207 // this is safe in the face of spelling differences, because there is no way 208 // to spell an i/e in a strange way that is another letter. Skipping this 209 // allows us to avoid looking up the identifier info for #define/#undef and 210 // other common directives. 211 bool Invalid = false; 212 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation(), 213 &Invalid); 214 if (Invalid) 215 return; 216 217 char FirstChar = RawCharData[0]; 218 if (FirstChar >= 'a' && FirstChar <= 'z' && 219 FirstChar != 'i' && FirstChar != 'e') { 220 CurPPLexer->ParsingPreprocessorDirective = false; 221 // Restore comment saving mode. 222 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 223 continue; 224 } 225 226 // Get the identifier name without trigraphs or embedded newlines. Note 227 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 228 // when skipping. 229 char DirectiveBuf[20]; 230 llvm::StringRef Directive; 231 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 232 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 233 } else { 234 std::string DirectiveStr = getSpelling(Tok); 235 unsigned IdLen = DirectiveStr.size(); 236 if (IdLen >= 20) { 237 CurPPLexer->ParsingPreprocessorDirective = false; 238 // Restore comment saving mode. 239 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 240 continue; 241 } 242 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 243 Directive = llvm::StringRef(DirectiveBuf, IdLen); 244 } 245 246 if (Directive.startswith("if")) { 247 llvm::StringRef Sub = Directive.substr(2); 248 if (Sub.empty() || // "if" 249 Sub == "def" || // "ifdef" 250 Sub == "ndef") { // "ifndef" 251 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 252 // bother parsing the condition. 253 DiscardUntilEndOfDirective(); 254 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 255 /*foundnonskip*/false, 256 /*fnddelse*/false); 257 } 258 } else if (Directive[0] == 'e') { 259 llvm::StringRef Sub = Directive.substr(1); 260 if (Sub == "ndif") { // "endif" 261 CheckEndOfDirective("endif"); 262 PPConditionalInfo CondInfo; 263 CondInfo.WasSkipping = true; // Silence bogus warning. 264 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 265 InCond = InCond; // Silence warning in no-asserts mode. 266 assert(!InCond && "Can't be skipping if not in a conditional!"); 267 268 // If we popped the outermost skipping block, we're done skipping! 269 if (!CondInfo.WasSkipping) 270 break; 271 } else if (Sub == "lse") { // "else". 272 // #else directive in a skipping conditional. If not in some other 273 // skipping conditional, and if #else hasn't already been seen, enter it 274 // as a non-skipping conditional. 275 DiscardUntilEndOfDirective(); // C99 6.10p4. 276 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 277 278 // If this is a #else with a #else before it, report the error. 279 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 280 281 // Note that we've seen a #else in this conditional. 282 CondInfo.FoundElse = true; 283 284 // If the conditional is at the top level, and the #if block wasn't 285 // entered, enter the #else block now. 286 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 287 CondInfo.FoundNonSkip = true; 288 break; 289 } 290 } else if (Sub == "lif") { // "elif". 291 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 292 293 bool ShouldEnter; 294 // If this is in a skipping block or if we're already handled this #if 295 // block, don't bother parsing the condition. 296 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 297 DiscardUntilEndOfDirective(); 298 ShouldEnter = false; 299 } else { 300 // Restore the value of LexingRawMode so that identifiers are 301 // looked up, etc, inside the #elif expression. 302 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 303 CurPPLexer->LexingRawMode = false; 304 IdentifierInfo *IfNDefMacro = 0; 305 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 306 CurPPLexer->LexingRawMode = true; 307 } 308 309 // If this is a #elif with a #else before it, report the error. 310 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 311 312 // If this condition is true, enter it! 313 if (ShouldEnter) { 314 CondInfo.FoundNonSkip = true; 315 break; 316 } 317 } 318 } 319 320 CurPPLexer->ParsingPreprocessorDirective = false; 321 // Restore comment saving mode. 322 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 323 } 324 325 // Finally, if we are out of the conditional (saw an #endif or ran off the end 326 // of the file, just stop skipping and return to lexing whatever came after 327 // the #if block. 328 CurPPLexer->LexingRawMode = false; 329} 330 331void Preprocessor::PTHSkipExcludedConditionalBlock() { 332 333 while (1) { 334 assert(CurPTHLexer); 335 assert(CurPTHLexer->LexingRawMode == false); 336 337 // Skip to the next '#else', '#elif', or #endif. 338 if (CurPTHLexer->SkipBlock()) { 339 // We have reached an #endif. Both the '#' and 'endif' tokens 340 // have been consumed by the PTHLexer. Just pop off the condition level. 341 PPConditionalInfo CondInfo; 342 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 343 InCond = InCond; // Silence warning in no-asserts mode. 344 assert(!InCond && "Can't be skipping if not in a conditional!"); 345 break; 346 } 347 348 // We have reached a '#else' or '#elif'. Lex the next token to get 349 // the directive flavor. 350 Token Tok; 351 LexUnexpandedToken(Tok); 352 353 // We can actually look up the IdentifierInfo here since we aren't in 354 // raw mode. 355 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 356 357 if (K == tok::pp_else) { 358 // #else: Enter the else condition. We aren't in a nested condition 359 // since we skip those. We're always in the one matching the last 360 // blocked we skipped. 361 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 362 // Note that we've seen a #else in this conditional. 363 CondInfo.FoundElse = true; 364 365 // If the #if block wasn't entered then enter the #else block now. 366 if (!CondInfo.FoundNonSkip) { 367 CondInfo.FoundNonSkip = true; 368 369 // Scan until the eom token. 370 CurPTHLexer->ParsingPreprocessorDirective = true; 371 DiscardUntilEndOfDirective(); 372 CurPTHLexer->ParsingPreprocessorDirective = false; 373 374 break; 375 } 376 377 // Otherwise skip this block. 378 continue; 379 } 380 381 assert(K == tok::pp_elif); 382 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 383 384 // If this is a #elif with a #else before it, report the error. 385 if (CondInfo.FoundElse) 386 Diag(Tok, diag::pp_err_elif_after_else); 387 388 // If this is in a skipping block or if we're already handled this #if 389 // block, don't bother parsing the condition. We just skip this block. 390 if (CondInfo.FoundNonSkip) 391 continue; 392 393 // Evaluate the condition of the #elif. 394 IdentifierInfo *IfNDefMacro = 0; 395 CurPTHLexer->ParsingPreprocessorDirective = true; 396 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 397 CurPTHLexer->ParsingPreprocessorDirective = false; 398 399 // If this condition is true, enter it! 400 if (ShouldEnter) { 401 CondInfo.FoundNonSkip = true; 402 break; 403 } 404 405 // Otherwise, skip this block and go to the next one. 406 continue; 407 } 408} 409 410/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 411/// return null on failure. isAngled indicates whether the file reference is 412/// for system #include's or not (i.e. using <> instead of ""). 413const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 414 bool isAngled, 415 const DirectoryLookup *FromDir, 416 const DirectoryLookup *&CurDir) { 417 // If the header lookup mechanism may be relative to the current file, pass in 418 // info about where the current file is. 419 const FileEntry *CurFileEnt = 0; 420 if (!FromDir) { 421 FileID FID = getCurrentFileLexer()->getFileID(); 422 CurFileEnt = SourceMgr.getFileEntryForID(FID); 423 424 // If there is no file entry associated with this file, it must be the 425 // predefines buffer. Any other file is not lexed with a normal lexer, so 426 // it won't be scanned for preprocessor directives. If we have the 427 // predefines buffer, resolve #include references (which come from the 428 // -include command line argument) as if they came from the main file, this 429 // affects file lookup etc. 430 if (CurFileEnt == 0) { 431 FID = SourceMgr.getMainFileID(); 432 CurFileEnt = SourceMgr.getFileEntryForID(FID); 433 } 434 } 435 436 // Do a standard file entry lookup. 437 CurDir = CurDirLookup; 438 const FileEntry *FE = 439 HeaderInfo.LookupFile(Filename, isAngled, FromDir, CurDir, CurFileEnt); 440 if (FE) return FE; 441 442 // Otherwise, see if this is a subframework header. If so, this is relative 443 // to one of the headers on the #include stack. Walk the list of the current 444 // headers on the #include stack and pass them to HeaderInfo. 445 if (IsFileLexer()) { 446 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 447 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 448 return FE; 449 } 450 451 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 452 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 453 if (IsFileLexer(ISEntry)) { 454 if ((CurFileEnt = 455 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 456 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 457 return FE; 458 } 459 } 460 461 // Otherwise, we really couldn't find the file. 462 return 0; 463} 464 465 466//===----------------------------------------------------------------------===// 467// Preprocessor Directive Handling. 468//===----------------------------------------------------------------------===// 469 470/// HandleDirective - This callback is invoked when the lexer sees a # token 471/// at the start of a line. This consumes the directive, modifies the 472/// lexer/preprocessor state, and advances the lexer(s) so that the next token 473/// read is the correct one. 474void Preprocessor::HandleDirective(Token &Result) { 475 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 476 477 // We just parsed a # character at the start of a line, so we're in directive 478 // mode. Tell the lexer this so any newlines we see will be converted into an 479 // EOM token (which terminates the directive). 480 CurPPLexer->ParsingPreprocessorDirective = true; 481 482 ++NumDirectives; 483 484 // We are about to read a token. For the multiple-include optimization FA to 485 // work, we have to remember if we had read any tokens *before* this 486 // pp-directive. 487 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 488 489 // Save the '#' token in case we need to return it later. 490 Token SavedHash = Result; 491 492 // Read the next token, the directive flavor. This isn't expanded due to 493 // C99 6.10.3p8. 494 LexUnexpandedToken(Result); 495 496 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 497 // #define A(x) #x 498 // A(abc 499 // #warning blah 500 // def) 501 // If so, the user is relying on non-portable behavior, emit a diagnostic. 502 if (InMacroArgs) 503 Diag(Result, diag::ext_embedded_directive); 504 505TryAgain: 506 switch (Result.getKind()) { 507 case tok::eom: 508 return; // null directive. 509 case tok::comment: 510 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 511 LexUnexpandedToken(Result); 512 goto TryAgain; 513 514 case tok::numeric_constant: // # 7 GNU line marker directive. 515 if (getLangOptions().AsmPreprocessor) 516 break; // # 4 is not a preprocessor directive in .S files. 517 return HandleDigitDirective(Result); 518 default: 519 IdentifierInfo *II = Result.getIdentifierInfo(); 520 if (II == 0) break; // Not an identifier. 521 522 // Ask what the preprocessor keyword ID is. 523 switch (II->getPPKeywordID()) { 524 default: break; 525 // C99 6.10.1 - Conditional Inclusion. 526 case tok::pp_if: 527 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 528 case tok::pp_ifdef: 529 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 530 case tok::pp_ifndef: 531 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 532 case tok::pp_elif: 533 return HandleElifDirective(Result); 534 case tok::pp_else: 535 return HandleElseDirective(Result); 536 case tok::pp_endif: 537 return HandleEndifDirective(Result); 538 539 // C99 6.10.2 - Source File Inclusion. 540 case tok::pp_include: 541 return HandleIncludeDirective(Result); // Handle #include. 542 case tok::pp___include_macros: 543 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 544 545 // C99 6.10.3 - Macro Replacement. 546 case tok::pp_define: 547 return HandleDefineDirective(Result); 548 case tok::pp_undef: 549 return HandleUndefDirective(Result); 550 551 // C99 6.10.4 - Line Control. 552 case tok::pp_line: 553 return HandleLineDirective(Result); 554 555 // C99 6.10.5 - Error Directive. 556 case tok::pp_error: 557 return HandleUserDiagnosticDirective(Result, false); 558 559 // C99 6.10.6 - Pragma Directive. 560 case tok::pp_pragma: 561 return HandlePragmaDirective(); 562 563 // GNU Extensions. 564 case tok::pp_import: 565 return HandleImportDirective(Result); 566 case tok::pp_include_next: 567 return HandleIncludeNextDirective(Result); 568 569 case tok::pp_warning: 570 Diag(Result, diag::ext_pp_warning_directive); 571 return HandleUserDiagnosticDirective(Result, true); 572 case tok::pp_ident: 573 return HandleIdentSCCSDirective(Result); 574 case tok::pp_sccs: 575 return HandleIdentSCCSDirective(Result); 576 case tok::pp_assert: 577 //isExtension = true; // FIXME: implement #assert 578 break; 579 case tok::pp_unassert: 580 //isExtension = true; // FIXME: implement #unassert 581 break; 582 } 583 break; 584 } 585 586 // If this is a .S file, treat unknown # directives as non-preprocessor 587 // directives. This is important because # may be a comment or introduce 588 // various pseudo-ops. Just return the # token and push back the following 589 // token to be lexed next time. 590 if (getLangOptions().AsmPreprocessor) { 591 Token *Toks = new Token[2]; 592 // Return the # and the token after it. 593 Toks[0] = SavedHash; 594 Toks[1] = Result; 595 // Enter this token stream so that we re-lex the tokens. Make sure to 596 // enable macro expansion, in case the token after the # is an identifier 597 // that is expanded. 598 EnterTokenStream(Toks, 2, false, true); 599 return; 600 } 601 602 // If we reached here, the preprocessing token is not valid! 603 Diag(Result, diag::err_pp_invalid_directive); 604 605 // Read the rest of the PP line. 606 DiscardUntilEndOfDirective(); 607 608 // Okay, we're done parsing the directive. 609} 610 611/// GetLineValue - Convert a numeric token into an unsigned value, emitting 612/// Diagnostic DiagID if it is invalid, and returning the value in Val. 613static bool GetLineValue(Token &DigitTok, unsigned &Val, 614 unsigned DiagID, Preprocessor &PP) { 615 if (DigitTok.isNot(tok::numeric_constant)) { 616 PP.Diag(DigitTok, DiagID); 617 618 if (DigitTok.isNot(tok::eom)) 619 PP.DiscardUntilEndOfDirective(); 620 return true; 621 } 622 623 llvm::SmallString<64> IntegerBuffer; 624 IntegerBuffer.resize(DigitTok.getLength()); 625 const char *DigitTokBegin = &IntegerBuffer[0]; 626 bool Invalid = false; 627 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 628 if (Invalid) 629 return true; 630 631 // Verify that we have a simple digit-sequence, and compute the value. This 632 // is always a simple digit string computed in decimal, so we do this manually 633 // here. 634 Val = 0; 635 for (unsigned i = 0; i != ActualLength; ++i) { 636 if (!isdigit(DigitTokBegin[i])) { 637 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 638 diag::err_pp_line_digit_sequence); 639 PP.DiscardUntilEndOfDirective(); 640 return true; 641 } 642 643 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 644 if (NextVal < Val) { // overflow. 645 PP.Diag(DigitTok, DiagID); 646 PP.DiscardUntilEndOfDirective(); 647 return true; 648 } 649 Val = NextVal; 650 } 651 652 // Reject 0, this is needed both by #line numbers and flags. 653 if (Val == 0) { 654 PP.Diag(DigitTok, DiagID); 655 PP.DiscardUntilEndOfDirective(); 656 return true; 657 } 658 659 if (DigitTokBegin[0] == '0') 660 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 661 662 return false; 663} 664 665/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 666/// acceptable forms are: 667/// # line digit-sequence 668/// # line digit-sequence "s-char-sequence" 669void Preprocessor::HandleLineDirective(Token &Tok) { 670 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 671 // expanded. 672 Token DigitTok; 673 Lex(DigitTok); 674 675 // Validate the number and convert it to an unsigned. 676 unsigned LineNo; 677 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 678 return; 679 680 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 681 // number greater than 2147483647". C90 requires that the line # be <= 32767. 682 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 683 if (LineNo >= LineLimit) 684 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 685 686 int FilenameID = -1; 687 Token StrTok; 688 Lex(StrTok); 689 690 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 691 // string followed by eom. 692 if (StrTok.is(tok::eom)) 693 ; // ok 694 else if (StrTok.isNot(tok::string_literal)) { 695 Diag(StrTok, diag::err_pp_line_invalid_filename); 696 DiscardUntilEndOfDirective(); 697 return; 698 } else { 699 // Parse and validate the string, converting it into a unique ID. 700 StringLiteralParser Literal(&StrTok, 1, *this); 701 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 702 if (Literal.hadError) 703 return DiscardUntilEndOfDirective(); 704 if (Literal.Pascal) { 705 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 706 return DiscardUntilEndOfDirective(); 707 } 708 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 709 Literal.GetStringLength()); 710 711 // Verify that there is nothing after the string, other than EOM. Because 712 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 713 CheckEndOfDirective("line", true); 714 } 715 716 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 717 718 if (Callbacks) 719 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 720 PPCallbacks::RenameFile, 721 SrcMgr::C_User); 722} 723 724/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 725/// marker directive. 726static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 727 bool &IsSystemHeader, bool &IsExternCHeader, 728 Preprocessor &PP) { 729 unsigned FlagVal; 730 Token FlagTok; 731 PP.Lex(FlagTok); 732 if (FlagTok.is(tok::eom)) return false; 733 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 734 return true; 735 736 if (FlagVal == 1) { 737 IsFileEntry = true; 738 739 PP.Lex(FlagTok); 740 if (FlagTok.is(tok::eom)) return false; 741 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 742 return true; 743 } else if (FlagVal == 2) { 744 IsFileExit = true; 745 746 SourceManager &SM = PP.getSourceManager(); 747 // If we are leaving the current presumed file, check to make sure the 748 // presumed include stack isn't empty! 749 FileID CurFileID = 750 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 751 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 752 753 // If there is no include loc (main file) or if the include loc is in a 754 // different physical file, then we aren't in a "1" line marker flag region. 755 SourceLocation IncLoc = PLoc.getIncludeLoc(); 756 if (IncLoc.isInvalid() || 757 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 758 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 759 PP.DiscardUntilEndOfDirective(); 760 return true; 761 } 762 763 PP.Lex(FlagTok); 764 if (FlagTok.is(tok::eom)) return false; 765 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 766 return true; 767 } 768 769 // We must have 3 if there are still flags. 770 if (FlagVal != 3) { 771 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 772 PP.DiscardUntilEndOfDirective(); 773 return true; 774 } 775 776 IsSystemHeader = true; 777 778 PP.Lex(FlagTok); 779 if (FlagTok.is(tok::eom)) return false; 780 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 781 return true; 782 783 // We must have 4 if there is yet another flag. 784 if (FlagVal != 4) { 785 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 786 PP.DiscardUntilEndOfDirective(); 787 return true; 788 } 789 790 IsExternCHeader = true; 791 792 PP.Lex(FlagTok); 793 if (FlagTok.is(tok::eom)) return false; 794 795 // There are no more valid flags here. 796 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 797 PP.DiscardUntilEndOfDirective(); 798 return true; 799} 800 801/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 802/// one of the following forms: 803/// 804/// # 42 805/// # 42 "file" ('1' | '2')? 806/// # 42 "file" ('1' | '2')? '3' '4'? 807/// 808void Preprocessor::HandleDigitDirective(Token &DigitTok) { 809 // Validate the number and convert it to an unsigned. GNU does not have a 810 // line # limit other than it fit in 32-bits. 811 unsigned LineNo; 812 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 813 *this)) 814 return; 815 816 Token StrTok; 817 Lex(StrTok); 818 819 bool IsFileEntry = false, IsFileExit = false; 820 bool IsSystemHeader = false, IsExternCHeader = false; 821 int FilenameID = -1; 822 823 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 824 // string followed by eom. 825 if (StrTok.is(tok::eom)) 826 ; // ok 827 else if (StrTok.isNot(tok::string_literal)) { 828 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 829 return DiscardUntilEndOfDirective(); 830 } else { 831 // Parse and validate the string, converting it into a unique ID. 832 StringLiteralParser Literal(&StrTok, 1, *this); 833 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 834 if (Literal.hadError) 835 return DiscardUntilEndOfDirective(); 836 if (Literal.Pascal) { 837 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 838 return DiscardUntilEndOfDirective(); 839 } 840 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 841 Literal.GetStringLength()); 842 843 // If a filename was present, read any flags that are present. 844 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 845 IsSystemHeader, IsExternCHeader, *this)) 846 return; 847 } 848 849 // Create a line note with this information. 850 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 851 IsFileEntry, IsFileExit, 852 IsSystemHeader, IsExternCHeader); 853 854 // If the preprocessor has callbacks installed, notify them of the #line 855 // change. This is used so that the line marker comes out in -E mode for 856 // example. 857 if (Callbacks) { 858 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 859 if (IsFileEntry) 860 Reason = PPCallbacks::EnterFile; 861 else if (IsFileExit) 862 Reason = PPCallbacks::ExitFile; 863 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 864 if (IsExternCHeader) 865 FileKind = SrcMgr::C_ExternCSystem; 866 else if (IsSystemHeader) 867 FileKind = SrcMgr::C_System; 868 869 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 870 } 871} 872 873 874/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 875/// 876void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 877 bool isWarning) { 878 // PTH doesn't emit #warning or #error directives. 879 if (CurPTHLexer) 880 return CurPTHLexer->DiscardToEndOfLine(); 881 882 // Read the rest of the line raw. We do this because we don't want macros 883 // to be expanded and we don't require that the tokens be valid preprocessing 884 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 885 // collapse multiple consequtive white space between tokens, but this isn't 886 // specified by the standard. 887 std::string Message = CurLexer->ReadToEndOfLine(); 888 if (isWarning) 889 Diag(Tok, diag::pp_hash_warning) << Message; 890 else 891 Diag(Tok, diag::err_pp_hash_error) << Message; 892} 893 894/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 895/// 896void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 897 // Yes, this directive is an extension. 898 Diag(Tok, diag::ext_pp_ident_directive); 899 900 // Read the string argument. 901 Token StrTok; 902 Lex(StrTok); 903 904 // If the token kind isn't a string, it's a malformed directive. 905 if (StrTok.isNot(tok::string_literal) && 906 StrTok.isNot(tok::wide_string_literal)) { 907 Diag(StrTok, diag::err_pp_malformed_ident); 908 if (StrTok.isNot(tok::eom)) 909 DiscardUntilEndOfDirective(); 910 return; 911 } 912 913 // Verify that there is nothing after the string, other than EOM. 914 CheckEndOfDirective("ident"); 915 916 if (Callbacks) { 917 bool Invalid = false; 918 std::string Str = getSpelling(StrTok, &Invalid); 919 if (!Invalid) 920 Callbacks->Ident(Tok.getLocation(), Str); 921 } 922} 923 924//===----------------------------------------------------------------------===// 925// Preprocessor Include Directive Handling. 926//===----------------------------------------------------------------------===// 927 928/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 929/// checked and spelled filename, e.g. as an operand of #include. This returns 930/// true if the input filename was in <>'s or false if it were in ""'s. The 931/// caller is expected to provide a buffer that is large enough to hold the 932/// spelling of the filename, but is also expected to handle the case when 933/// this method decides to use a different buffer. 934bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 935 llvm::StringRef &Buffer) { 936 // Get the text form of the filename. 937 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 938 939 // Make sure the filename is <x> or "x". 940 bool isAngled; 941 if (Buffer[0] == '<') { 942 if (Buffer.back() != '>') { 943 Diag(Loc, diag::err_pp_expects_filename); 944 Buffer = llvm::StringRef(); 945 return true; 946 } 947 isAngled = true; 948 } else if (Buffer[0] == '"') { 949 if (Buffer.back() != '"') { 950 Diag(Loc, diag::err_pp_expects_filename); 951 Buffer = llvm::StringRef(); 952 return true; 953 } 954 isAngled = false; 955 } else { 956 Diag(Loc, diag::err_pp_expects_filename); 957 Buffer = llvm::StringRef(); 958 return true; 959 } 960 961 // Diagnose #include "" as invalid. 962 if (Buffer.size() <= 2) { 963 Diag(Loc, diag::err_pp_empty_filename); 964 Buffer = llvm::StringRef(); 965 return true; 966 } 967 968 // Skip the brackets. 969 Buffer = Buffer.substr(1, Buffer.size()-2); 970 return isAngled; 971} 972 973/// ConcatenateIncludeName - Handle cases where the #include name is expanded 974/// from a macro as multiple tokens, which need to be glued together. This 975/// occurs for code like: 976/// #define FOO <a/b.h> 977/// #include FOO 978/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 979/// 980/// This code concatenates and consumes tokens up to the '>' token. It returns 981/// false if the > was found, otherwise it returns true if it finds and consumes 982/// the EOM marker. 983bool Preprocessor::ConcatenateIncludeName( 984 llvm::SmallString<128> &FilenameBuffer) { 985 Token CurTok; 986 987 Lex(CurTok); 988 while (CurTok.isNot(tok::eom)) { 989 // Append the spelling of this token to the buffer. If there was a space 990 // before it, add it now. 991 if (CurTok.hasLeadingSpace()) 992 FilenameBuffer.push_back(' '); 993 994 // Get the spelling of the token, directly into FilenameBuffer if possible. 995 unsigned PreAppendSize = FilenameBuffer.size(); 996 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 997 998 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 999 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1000 1001 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1002 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1003 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1004 1005 // Resize FilenameBuffer to the correct size. 1006 if (CurTok.getLength() != ActualLen) 1007 FilenameBuffer.resize(PreAppendSize+ActualLen); 1008 1009 // If we found the '>' marker, return success. 1010 if (CurTok.is(tok::greater)) 1011 return false; 1012 1013 Lex(CurTok); 1014 } 1015 1016 // If we hit the eom marker, emit an error and return true so that the caller 1017 // knows the EOM has been read. 1018 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1019 return true; 1020} 1021 1022/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1023/// file to be included from the lexer, then include it! This is a common 1024/// routine with functionality shared between #include, #include_next and 1025/// #import. LookupFrom is set when this is a #include_next directive, it 1026/// specifies the file to start searching from. 1027void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1028 const DirectoryLookup *LookupFrom, 1029 bool isImport) { 1030 1031 Token FilenameTok; 1032 CurPPLexer->LexIncludeFilename(FilenameTok); 1033 1034 // Reserve a buffer to get the spelling. 1035 llvm::SmallString<128> FilenameBuffer; 1036 llvm::StringRef Filename; 1037 1038 switch (FilenameTok.getKind()) { 1039 case tok::eom: 1040 // If the token kind is EOM, the error has already been diagnosed. 1041 return; 1042 1043 case tok::angle_string_literal: 1044 case tok::string_literal: 1045 Filename = getSpelling(FilenameTok, FilenameBuffer); 1046 break; 1047 1048 case tok::less: 1049 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1050 // case, glue the tokens together into FilenameBuffer and interpret those. 1051 FilenameBuffer.push_back('<'); 1052 if (ConcatenateIncludeName(FilenameBuffer)) 1053 return; // Found <eom> but no ">"? Diagnostic already emitted. 1054 Filename = FilenameBuffer.str(); 1055 break; 1056 default: 1057 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1058 DiscardUntilEndOfDirective(); 1059 return; 1060 } 1061 1062 bool isAngled = 1063 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1064 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1065 // error. 1066 if (Filename.empty()) { 1067 DiscardUntilEndOfDirective(); 1068 return; 1069 } 1070 1071 // Verify that there is nothing after the filename, other than EOM. Note that 1072 // we allow macros that expand to nothing after the filename, because this 1073 // falls into the category of "#include pp-tokens new-line" specified in 1074 // C99 6.10.2p4. 1075 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1076 1077 // Check that we don't have infinite #include recursion. 1078 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1079 Diag(FilenameTok, diag::err_pp_include_too_deep); 1080 return; 1081 } 1082 1083 // Search include directories. 1084 const DirectoryLookup *CurDir; 1085 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1086 if (File == 0) { 1087 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1088 return; 1089 } 1090 1091 // The #included file will be considered to be a system header if either it is 1092 // in a system include directory, or if the #includer is a system include 1093 // header. 1094 SrcMgr::CharacteristicKind FileCharacter = 1095 std::max(HeaderInfo.getFileDirFlavor(File), 1096 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1097 1098 // Ask HeaderInfo if we should enter this #include file. If not, #including 1099 // this file will have no effect. 1100 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1101 if (Callbacks) { 1102 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1103 } 1104 return; 1105 } 1106 1107 // Look up the file, create a File ID for it. 1108 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1109 FileCharacter); 1110 if (FID.isInvalid()) { 1111 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1112 return; 1113 } 1114 1115 // Finally, if all is good, enter the new file! 1116 std::string ErrorStr; 1117 if (EnterSourceFile(FID, CurDir, ErrorStr)) 1118 Diag(FilenameTok, diag::err_pp_error_opening_file) 1119 << std::string(SourceMgr.getFileEntryForID(FID)->getName()) << ErrorStr; 1120} 1121 1122/// HandleIncludeNextDirective - Implements #include_next. 1123/// 1124void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1125 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1126 1127 // #include_next is like #include, except that we start searching after 1128 // the current found directory. If we can't do this, issue a 1129 // diagnostic. 1130 const DirectoryLookup *Lookup = CurDirLookup; 1131 if (isInPrimaryFile()) { 1132 Lookup = 0; 1133 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1134 } else if (Lookup == 0) { 1135 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1136 } else { 1137 // Start looking up in the next directory. 1138 ++Lookup; 1139 } 1140 1141 return HandleIncludeDirective(IncludeNextTok, Lookup); 1142} 1143 1144/// HandleImportDirective - Implements #import. 1145/// 1146void Preprocessor::HandleImportDirective(Token &ImportTok) { 1147 if (!Features.ObjC1) // #import is standard for ObjC. 1148 Diag(ImportTok, diag::ext_pp_import_directive); 1149 1150 return HandleIncludeDirective(ImportTok, 0, true); 1151} 1152 1153/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1154/// pseudo directive in the predefines buffer. This handles it by sucking all 1155/// tokens through the preprocessor and discarding them (only keeping the side 1156/// effects on the preprocessor). 1157void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1158 // This directive should only occur in the predefines buffer. If not, emit an 1159 // error and reject it. 1160 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1161 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1162 Diag(IncludeMacrosTok.getLocation(), 1163 diag::pp_include_macros_out_of_predefines); 1164 DiscardUntilEndOfDirective(); 1165 return; 1166 } 1167 1168 // Treat this as a normal #include for checking purposes. If this is 1169 // successful, it will push a new lexer onto the include stack. 1170 HandleIncludeDirective(IncludeMacrosTok, 0, false); 1171 1172 Token TmpTok; 1173 do { 1174 Lex(TmpTok); 1175 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1176 } while (TmpTok.isNot(tok::hashhash)); 1177} 1178 1179//===----------------------------------------------------------------------===// 1180// Preprocessor Macro Directive Handling. 1181//===----------------------------------------------------------------------===// 1182 1183/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1184/// definition has just been read. Lex the rest of the arguments and the 1185/// closing ), updating MI with what we learn. Return true if an error occurs 1186/// parsing the arg list. 1187bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1188 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1189 1190 Token Tok; 1191 while (1) { 1192 LexUnexpandedToken(Tok); 1193 switch (Tok.getKind()) { 1194 case tok::r_paren: 1195 // Found the end of the argument list. 1196 if (Arguments.empty()) // #define FOO() 1197 return false; 1198 // Otherwise we have #define FOO(A,) 1199 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1200 return true; 1201 case tok::ellipsis: // #define X(... -> C99 varargs 1202 // Warn if use of C99 feature in non-C99 mode. 1203 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1204 1205 // Lex the token after the identifier. 1206 LexUnexpandedToken(Tok); 1207 if (Tok.isNot(tok::r_paren)) { 1208 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1209 return true; 1210 } 1211 // Add the __VA_ARGS__ identifier as an argument. 1212 Arguments.push_back(Ident__VA_ARGS__); 1213 MI->setIsC99Varargs(); 1214 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1215 return false; 1216 case tok::eom: // #define X( 1217 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1218 return true; 1219 default: 1220 // Handle keywords and identifiers here to accept things like 1221 // #define Foo(for) for. 1222 IdentifierInfo *II = Tok.getIdentifierInfo(); 1223 if (II == 0) { 1224 // #define X(1 1225 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1226 return true; 1227 } 1228 1229 // If this is already used as an argument, it is used multiple times (e.g. 1230 // #define X(A,A. 1231 if (std::find(Arguments.begin(), Arguments.end(), II) != 1232 Arguments.end()) { // C99 6.10.3p6 1233 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1234 return true; 1235 } 1236 1237 // Add the argument to the macro info. 1238 Arguments.push_back(II); 1239 1240 // Lex the token after the identifier. 1241 LexUnexpandedToken(Tok); 1242 1243 switch (Tok.getKind()) { 1244 default: // #define X(A B 1245 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1246 return true; 1247 case tok::r_paren: // #define X(A) 1248 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1249 return false; 1250 case tok::comma: // #define X(A, 1251 break; 1252 case tok::ellipsis: // #define X(A... -> GCC extension 1253 // Diagnose extension. 1254 Diag(Tok, diag::ext_named_variadic_macro); 1255 1256 // Lex the token after the identifier. 1257 LexUnexpandedToken(Tok); 1258 if (Tok.isNot(tok::r_paren)) { 1259 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1260 return true; 1261 } 1262 1263 MI->setIsGNUVarargs(); 1264 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1265 return false; 1266 } 1267 } 1268 } 1269} 1270 1271/// HandleDefineDirective - Implements #define. This consumes the entire macro 1272/// line then lets the caller lex the next real token. 1273void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1274 ++NumDefined; 1275 1276 Token MacroNameTok; 1277 ReadMacroName(MacroNameTok, 1); 1278 1279 // Error reading macro name? If so, diagnostic already issued. 1280 if (MacroNameTok.is(tok::eom)) 1281 return; 1282 1283 Token LastTok = MacroNameTok; 1284 1285 // If we are supposed to keep comments in #defines, reenable comment saving 1286 // mode. 1287 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1288 1289 // Create the new macro. 1290 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1291 1292 Token Tok; 1293 LexUnexpandedToken(Tok); 1294 1295 // If this is a function-like macro definition, parse the argument list, 1296 // marking each of the identifiers as being used as macro arguments. Also, 1297 // check other constraints on the first token of the macro body. 1298 if (Tok.is(tok::eom)) { 1299 // If there is no body to this macro, we have no special handling here. 1300 } else if (Tok.hasLeadingSpace()) { 1301 // This is a normal token with leading space. Clear the leading space 1302 // marker on the first token to get proper expansion. 1303 Tok.clearFlag(Token::LeadingSpace); 1304 } else if (Tok.is(tok::l_paren)) { 1305 // This is a function-like macro definition. Read the argument list. 1306 MI->setIsFunctionLike(); 1307 if (ReadMacroDefinitionArgList(MI)) { 1308 // Forget about MI. 1309 ReleaseMacroInfo(MI); 1310 // Throw away the rest of the line. 1311 if (CurPPLexer->ParsingPreprocessorDirective) 1312 DiscardUntilEndOfDirective(); 1313 return; 1314 } 1315 1316 // If this is a definition of a variadic C99 function-like macro, not using 1317 // the GNU named varargs extension, enabled __VA_ARGS__. 1318 1319 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1320 // This gets unpoisoned where it is allowed. 1321 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1322 if (MI->isC99Varargs()) 1323 Ident__VA_ARGS__->setIsPoisoned(false); 1324 1325 // Read the first token after the arg list for down below. 1326 LexUnexpandedToken(Tok); 1327 } else if (Features.C99) { 1328 // C99 requires whitespace between the macro definition and the body. Emit 1329 // a diagnostic for something like "#define X+". 1330 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1331 } else { 1332 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1333 // first character of a replacement list is not a character required by 1334 // subclause 5.2.1, then there shall be white-space separation between the 1335 // identifier and the replacement list.". 5.2.1 lists this set: 1336 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1337 // is irrelevant here. 1338 bool isInvalid = false; 1339 if (Tok.is(tok::at)) // @ is not in the list above. 1340 isInvalid = true; 1341 else if (Tok.is(tok::unknown)) { 1342 // If we have an unknown token, it is something strange like "`". Since 1343 // all of valid characters would have lexed into a single character 1344 // token of some sort, we know this is not a valid case. 1345 isInvalid = true; 1346 } 1347 if (isInvalid) 1348 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1349 else 1350 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1351 } 1352 1353 if (!Tok.is(tok::eom)) 1354 LastTok = Tok; 1355 1356 // Read the rest of the macro body. 1357 if (MI->isObjectLike()) { 1358 // Object-like macros are very simple, just read their body. 1359 while (Tok.isNot(tok::eom)) { 1360 LastTok = Tok; 1361 MI->AddTokenToBody(Tok); 1362 // Get the next token of the macro. 1363 LexUnexpandedToken(Tok); 1364 } 1365 1366 } else { 1367 // Otherwise, read the body of a function-like macro. While we are at it, 1368 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1369 // parameters in function-like macro expansions. 1370 while (Tok.isNot(tok::eom)) { 1371 LastTok = Tok; 1372 1373 if (Tok.isNot(tok::hash)) { 1374 MI->AddTokenToBody(Tok); 1375 1376 // Get the next token of the macro. 1377 LexUnexpandedToken(Tok); 1378 continue; 1379 } 1380 1381 // Get the next token of the macro. 1382 LexUnexpandedToken(Tok); 1383 1384 // Check for a valid macro arg identifier. 1385 if (Tok.getIdentifierInfo() == 0 || 1386 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1387 1388 // If this is assembler-with-cpp mode, we accept random gibberish after 1389 // the '#' because '#' is often a comment character. However, change 1390 // the kind of the token to tok::unknown so that the preprocessor isn't 1391 // confused. 1392 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1393 LastTok.setKind(tok::unknown); 1394 } else { 1395 Diag(Tok, diag::err_pp_stringize_not_parameter); 1396 ReleaseMacroInfo(MI); 1397 1398 // Disable __VA_ARGS__ again. 1399 Ident__VA_ARGS__->setIsPoisoned(true); 1400 return; 1401 } 1402 } 1403 1404 // Things look ok, add the '#' and param name tokens to the macro. 1405 MI->AddTokenToBody(LastTok); 1406 MI->AddTokenToBody(Tok); 1407 LastTok = Tok; 1408 1409 // Get the next token of the macro. 1410 LexUnexpandedToken(Tok); 1411 } 1412 } 1413 1414 1415 // Disable __VA_ARGS__ again. 1416 Ident__VA_ARGS__->setIsPoisoned(true); 1417 1418 // Check that there is no paste (##) operator at the begining or end of the 1419 // replacement list. 1420 unsigned NumTokens = MI->getNumTokens(); 1421 if (NumTokens != 0) { 1422 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1423 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1424 ReleaseMacroInfo(MI); 1425 return; 1426 } 1427 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1428 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1429 ReleaseMacroInfo(MI); 1430 return; 1431 } 1432 } 1433 1434 // If this is the primary source file, remember that this macro hasn't been 1435 // used yet. 1436 if (isInPrimaryFile()) 1437 MI->setIsUsed(false); 1438 1439 MI->setDefinitionEndLoc(LastTok.getLocation()); 1440 1441 // Finally, if this identifier already had a macro defined for it, verify that 1442 // the macro bodies are identical and free the old definition. 1443 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1444 // It is very common for system headers to have tons of macro redefinitions 1445 // and for warnings to be disabled in system headers. If this is the case, 1446 // then don't bother calling MacroInfo::isIdenticalTo. 1447 if (!getDiagnostics().getSuppressSystemWarnings() || 1448 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1449 if (!OtherMI->isUsed()) 1450 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1451 1452 // Macros must be identical. This means all tokes and whitespace 1453 // separation must be the same. C99 6.10.3.2. 1454 if (!MI->isIdenticalTo(*OtherMI, *this)) { 1455 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1456 << MacroNameTok.getIdentifierInfo(); 1457 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1458 } 1459 } 1460 1461 ReleaseMacroInfo(OtherMI); 1462 } 1463 1464 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1465 1466 // If the callbacks want to know, tell them about the macro definition. 1467 if (Callbacks) 1468 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1469} 1470 1471/// HandleUndefDirective - Implements #undef. 1472/// 1473void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1474 ++NumUndefined; 1475 1476 Token MacroNameTok; 1477 ReadMacroName(MacroNameTok, 2); 1478 1479 // Error reading macro name? If so, diagnostic already issued. 1480 if (MacroNameTok.is(tok::eom)) 1481 return; 1482 1483 // Check to see if this is the last token on the #undef line. 1484 CheckEndOfDirective("undef"); 1485 1486 // Okay, we finally have a valid identifier to undef. 1487 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1488 1489 // If the macro is not defined, this is a noop undef, just return. 1490 if (MI == 0) return; 1491 1492 if (!MI->isUsed()) 1493 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1494 1495 // If the callbacks want to know, tell them about the macro #undef. 1496 if (Callbacks) 1497 Callbacks->MacroUndefined(MacroNameTok.getIdentifierInfo(), MI); 1498 1499 // Free macro definition. 1500 ReleaseMacroInfo(MI); 1501 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1502} 1503 1504 1505//===----------------------------------------------------------------------===// 1506// Preprocessor Conditional Directive Handling. 1507//===----------------------------------------------------------------------===// 1508 1509/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1510/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1511/// if any tokens have been returned or pp-directives activated before this 1512/// #ifndef has been lexed. 1513/// 1514void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1515 bool ReadAnyTokensBeforeDirective) { 1516 ++NumIf; 1517 Token DirectiveTok = Result; 1518 1519 Token MacroNameTok; 1520 ReadMacroName(MacroNameTok); 1521 1522 // Error reading macro name? If so, diagnostic already issued. 1523 if (MacroNameTok.is(tok::eom)) { 1524 // Skip code until we get to #endif. This helps with recovery by not 1525 // emitting an error when the #endif is reached. 1526 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1527 /*Foundnonskip*/false, /*FoundElse*/false); 1528 return; 1529 } 1530 1531 // Check to see if this is the last token on the #if[n]def line. 1532 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1533 1534 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1535 MacroInfo *MI = getMacroInfo(MII); 1536 1537 if (CurPPLexer->getConditionalStackDepth() == 0) { 1538 // If the start of a top-level #ifdef and if the macro is not defined, 1539 // inform MIOpt that this might be the start of a proper include guard. 1540 // Otherwise it is some other form of unknown conditional which we can't 1541 // handle. 1542 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1543 assert(isIfndef && "#ifdef shouldn't reach here"); 1544 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1545 } else 1546 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1547 } 1548 1549 // If there is a macro, process it. 1550 if (MI) // Mark it used. 1551 MI->setIsUsed(true); 1552 1553 // Should we include the stuff contained by this directive? 1554 if (!MI == isIfndef) { 1555 // Yes, remember that we are inside a conditional, then lex the next token. 1556 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1557 /*wasskip*/false, /*foundnonskip*/true, 1558 /*foundelse*/false); 1559 } else { 1560 // No, skip the contents of this block and return the first token after it. 1561 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1562 /*Foundnonskip*/false, 1563 /*FoundElse*/false); 1564 } 1565} 1566 1567/// HandleIfDirective - Implements the #if directive. 1568/// 1569void Preprocessor::HandleIfDirective(Token &IfToken, 1570 bool ReadAnyTokensBeforeDirective) { 1571 ++NumIf; 1572 1573 // Parse and evaluation the conditional expression. 1574 IdentifierInfo *IfNDefMacro = 0; 1575 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1576 1577 1578 // If this condition is equivalent to #ifndef X, and if this is the first 1579 // directive seen, handle it for the multiple-include optimization. 1580 if (CurPPLexer->getConditionalStackDepth() == 0) { 1581 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1582 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1583 else 1584 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1585 } 1586 1587 // Should we include the stuff contained by this directive? 1588 if (ConditionalTrue) { 1589 // Yes, remember that we are inside a conditional, then lex the next token. 1590 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1591 /*foundnonskip*/true, /*foundelse*/false); 1592 } else { 1593 // No, skip the contents of this block and return the first token after it. 1594 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1595 /*FoundElse*/false); 1596 } 1597} 1598 1599/// HandleEndifDirective - Implements the #endif directive. 1600/// 1601void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1602 ++NumEndif; 1603 1604 // Check that this is the whole directive. 1605 CheckEndOfDirective("endif"); 1606 1607 PPConditionalInfo CondInfo; 1608 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1609 // No conditionals on the stack: this is an #endif without an #if. 1610 Diag(EndifToken, diag::err_pp_endif_without_if); 1611 return; 1612 } 1613 1614 // If this the end of a top-level #endif, inform MIOpt. 1615 if (CurPPLexer->getConditionalStackDepth() == 0) 1616 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1617 1618 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1619 "This code should only be reachable in the non-skipping case!"); 1620} 1621 1622 1623void Preprocessor::HandleElseDirective(Token &Result) { 1624 ++NumElse; 1625 1626 // #else directive in a non-skipping conditional... start skipping. 1627 CheckEndOfDirective("else"); 1628 1629 PPConditionalInfo CI; 1630 if (CurPPLexer->popConditionalLevel(CI)) { 1631 Diag(Result, diag::pp_err_else_without_if); 1632 return; 1633 } 1634 1635 // If this is a top-level #else, inform the MIOpt. 1636 if (CurPPLexer->getConditionalStackDepth() == 0) 1637 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1638 1639 // If this is a #else with a #else before it, report the error. 1640 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1641 1642 // Finally, skip the rest of the contents of this block and return the first 1643 // token after it. 1644 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1645 /*FoundElse*/true); 1646} 1647 1648void Preprocessor::HandleElifDirective(Token &ElifToken) { 1649 ++NumElse; 1650 1651 // #elif directive in a non-skipping conditional... start skipping. 1652 // We don't care what the condition is, because we will always skip it (since 1653 // the block immediately before it was included). 1654 DiscardUntilEndOfDirective(); 1655 1656 PPConditionalInfo CI; 1657 if (CurPPLexer->popConditionalLevel(CI)) { 1658 Diag(ElifToken, diag::pp_err_elif_without_if); 1659 return; 1660 } 1661 1662 // If this is a top-level #elif, inform the MIOpt. 1663 if (CurPPLexer->getConditionalStackDepth() == 0) 1664 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1665 1666 // If this is a #elif with a #else before it, report the error. 1667 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1668 1669 // Finally, skip the rest of the contents of this block and return the first 1670 // token after it. 1671 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1672 /*FoundElse*/CI.FoundElse); 1673} 1674