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 /// \file 11 /// \brief Implements # directive processing for the Preprocessor. 12 /// 13 //===----------------------------------------------------------------------===// 14 15 #include "clang/Lex/Preprocessor.h" 16 #include "clang/Basic/FileManager.h" 17 #include "clang/Basic/SourceManager.h" 18 #include "clang/Lex/CodeCompletionHandler.h" 19 #include "clang/Lex/HeaderSearch.h" 20 #include "clang/Lex/HeaderSearchOptions.h" 21 #include "clang/Lex/LexDiagnostic.h" 22 #include "clang/Lex/LiteralSupport.h" 23 #include "clang/Lex/MacroInfo.h" 24 #include "clang/Lex/ModuleLoader.h" 25 #include "clang/Lex/Pragma.h" 26 #include "llvm/ADT/APInt.h" 27 #include "llvm/Support/ErrorHandling.h" 28 #include "llvm/Support/Path.h" 29 #include "llvm/Support/SaveAndRestore.h" 30 using namespace clang; 31 32 //===----------------------------------------------------------------------===// 33 // Utility Methods for Preprocessor Directive Handling. 34 //===----------------------------------------------------------------------===// 35 36 MacroInfo *Preprocessor::AllocateMacroInfo() { 37 MacroInfoChain *MIChain = BP.Allocate<MacroInfoChain>(); 38 MIChain->Next = MIChainHead; 39 MIChainHead = MIChain; 40 return &MIChain->MI; 41 } 42 43 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 44 MacroInfo *MI = AllocateMacroInfo(); 45 new (MI) MacroInfo(L); 46 return MI; 47 } 48 49 MacroInfo *Preprocessor::AllocateDeserializedMacroInfo(SourceLocation L, 50 unsigned SubModuleID) { 51 static_assert(llvm::AlignOf<MacroInfo>::Alignment >= sizeof(SubModuleID), 52 "alignment for MacroInfo is less than the ID"); 53 DeserializedMacroInfoChain *MIChain = 54 BP.Allocate<DeserializedMacroInfoChain>(); 55 MIChain->Next = DeserialMIChainHead; 56 DeserialMIChainHead = MIChain; 57 58 MacroInfo *MI = &MIChain->MI; 59 new (MI) MacroInfo(L); 60 MI->FromASTFile = true; 61 MI->setOwningModuleID(SubModuleID); 62 return MI; 63 } 64 65 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI, 66 SourceLocation Loc) { 67 return new (BP) DefMacroDirective(MI, Loc); 68 } 69 70 UndefMacroDirective * 71 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) { 72 return new (BP) UndefMacroDirective(UndefLoc); 73 } 74 75 VisibilityMacroDirective * 76 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc, 77 bool isPublic) { 78 return new (BP) VisibilityMacroDirective(Loc, isPublic); 79 } 80 81 /// \brief Read and discard all tokens remaining on the current line until 82 /// the tok::eod token is found. 83 void Preprocessor::DiscardUntilEndOfDirective() { 84 Token Tmp; 85 do { 86 LexUnexpandedToken(Tmp); 87 assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens"); 88 } while (Tmp.isNot(tok::eod)); 89 } 90 91 /// \brief Enumerates possible cases of #define/#undef a reserved identifier. 92 enum MacroDiag { 93 MD_NoWarn, //> Not a reserved identifier 94 MD_KeywordDef, //> Macro hides keyword, enabled by default 95 MD_ReservedMacro //> #define of #undef reserved id, disabled by default 96 }; 97 98 /// \brief Checks if the specified identifier is reserved in the specified 99 /// language. 100 /// This function does not check if the identifier is a keyword. 101 static bool isReservedId(StringRef Text, const LangOptions &Lang) { 102 // C++ [macro.names], C11 7.1.3: 103 // All identifiers that begin with an underscore and either an uppercase 104 // letter or another underscore are always reserved for any use. 105 if (Text.size() >= 2 && Text[0] == '_' && 106 (isUppercase(Text[1]) || Text[1] == '_')) 107 return true; 108 // C++ [global.names] 109 // Each name that contains a double underscore ... is reserved to the 110 // implementation for any use. 111 if (Lang.CPlusPlus) { 112 if (Text.find("__") != StringRef::npos) 113 return true; 114 } 115 return false; 116 } 117 118 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) { 119 const LangOptions &Lang = PP.getLangOpts(); 120 StringRef Text = II->getName(); 121 if (isReservedId(Text, Lang)) 122 return MD_ReservedMacro; 123 if (II->isKeyword(Lang)) 124 return MD_KeywordDef; 125 if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final"))) 126 return MD_KeywordDef; 127 return MD_NoWarn; 128 } 129 130 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) { 131 const LangOptions &Lang = PP.getLangOpts(); 132 StringRef Text = II->getName(); 133 // Do not warn on keyword undef. It is generally harmless and widely used. 134 if (isReservedId(Text, Lang)) 135 return MD_ReservedMacro; 136 return MD_NoWarn; 137 } 138 139 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef, 140 bool *ShadowFlag) { 141 // Missing macro name? 142 if (MacroNameTok.is(tok::eod)) 143 return Diag(MacroNameTok, diag::err_pp_missing_macro_name); 144 145 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 146 if (!II) { 147 bool Invalid = false; 148 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 149 if (Invalid) 150 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 151 II = getIdentifierInfo(Spelling); 152 153 if (!II->isCPlusPlusOperatorKeyword()) 154 return Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 155 156 // C++ 2.5p2: Alternative tokens behave the same as its primary token 157 // except for their spellings. 158 Diag(MacroNameTok, getLangOpts().MicrosoftExt 159 ? diag::ext_pp_operator_used_as_macro_name 160 : diag::err_pp_operator_used_as_macro_name) 161 << II << MacroNameTok.getKind(); 162 163 // Allow #defining |and| and friends for Microsoft compatibility or 164 // recovery when legacy C headers are included in C++. 165 MacroNameTok.setIdentifierInfo(II); 166 } 167 168 if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) { 169 // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4. 170 return Diag(MacroNameTok, diag::err_defined_macro_name); 171 } 172 173 if (isDefineUndef == MU_Undef) { 174 auto *MI = getMacroInfo(II); 175 if (MI && MI->isBuiltinMacro()) { 176 // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4 177 // and C++ [cpp.predefined]p4], but allow it as an extension. 178 Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro); 179 } 180 } 181 182 // If defining/undefining reserved identifier or a keyword, we need to issue 183 // a warning. 184 SourceLocation MacroNameLoc = MacroNameTok.getLocation(); 185 if (ShadowFlag) 186 *ShadowFlag = false; 187 if (!SourceMgr.isInSystemHeader(MacroNameLoc) && 188 (strcmp(SourceMgr.getBufferName(MacroNameLoc), "<built-in>") != 0)) { 189 MacroDiag D = MD_NoWarn; 190 if (isDefineUndef == MU_Define) { 191 D = shouldWarnOnMacroDef(*this, II); 192 } 193 else if (isDefineUndef == MU_Undef) 194 D = shouldWarnOnMacroUndef(*this, II); 195 if (D == MD_KeywordDef) { 196 // We do not want to warn on some patterns widely used in configuration 197 // scripts. This requires analyzing next tokens, so do not issue warnings 198 // now, only inform caller. 199 if (ShadowFlag) 200 *ShadowFlag = true; 201 } 202 if (D == MD_ReservedMacro) 203 Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id); 204 } 205 206 // Okay, we got a good identifier. 207 return false; 208 } 209 210 /// \brief Lex and validate a macro name, which occurs after a 211 /// \#define or \#undef. 212 /// 213 /// This sets the token kind to eod and discards the rest of the macro line if 214 /// the macro name is invalid. 215 /// 216 /// \param MacroNameTok Token that is expected to be a macro name. 217 /// \param isDefineUndef Context in which macro is used. 218 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword. 219 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef, 220 bool *ShadowFlag) { 221 // Read the token, don't allow macro expansion on it. 222 LexUnexpandedToken(MacroNameTok); 223 224 if (MacroNameTok.is(tok::code_completion)) { 225 if (CodeComplete) 226 CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define); 227 setCodeCompletionReached(); 228 LexUnexpandedToken(MacroNameTok); 229 } 230 231 if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag)) 232 return; 233 234 // Invalid macro name, read and discard the rest of the line and set the 235 // token kind to tok::eod if necessary. 236 if (MacroNameTok.isNot(tok::eod)) { 237 MacroNameTok.setKind(tok::eod); 238 DiscardUntilEndOfDirective(); 239 } 240 } 241 242 /// \brief Ensure that the next token is a tok::eod token. 243 /// 244 /// If not, emit a diagnostic and consume up until the eod. If EnableMacros is 245 /// true, then we consider macros that expand to zero tokens as being ok. 246 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 247 Token Tmp; 248 // Lex unexpanded tokens for most directives: macros might expand to zero 249 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 250 // #line) allow empty macros. 251 if (EnableMacros) 252 Lex(Tmp); 253 else 254 LexUnexpandedToken(Tmp); 255 256 // There should be no tokens after the directive, but we allow them as an 257 // extension. 258 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 259 LexUnexpandedToken(Tmp); 260 261 if (Tmp.isNot(tok::eod)) { 262 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 263 // or if this is a macro-style preprocessing directive, because it is more 264 // trouble than it is worth to insert /**/ and check that there is no /**/ 265 // in the range also. 266 FixItHint Hint; 267 if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) && 268 !CurTokenLexer) 269 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 270 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 271 DiscardUntilEndOfDirective(); 272 } 273 } 274 275 276 277 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and 278 /// decided that the subsequent tokens are in the \#if'd out portion of the 279 /// file. Lex the rest of the file, until we see an \#endif. If 280 /// FoundNonSkipPortion is true, then we have already emitted code for part of 281 /// this \#if directive, so \#else/\#elif blocks should never be entered. 282 /// If ElseOk is true, then \#else directives are ok, if not, then we have 283 /// already seen one so a \#else directive is a duplicate. When this returns, 284 /// the caller can lex the first valid token. 285 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 286 bool FoundNonSkipPortion, 287 bool FoundElse, 288 SourceLocation ElseLoc) { 289 ++NumSkipped; 290 assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?"); 291 292 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 293 FoundNonSkipPortion, FoundElse); 294 295 if (CurPTHLexer) { 296 PTHSkipExcludedConditionalBlock(); 297 return; 298 } 299 300 // Enter raw mode to disable identifier lookup (and thus macro expansion), 301 // disabling warnings, etc. 302 CurPPLexer->LexingRawMode = true; 303 Token Tok; 304 while (1) { 305 CurLexer->Lex(Tok); 306 307 if (Tok.is(tok::code_completion)) { 308 if (CodeComplete) 309 CodeComplete->CodeCompleteInConditionalExclusion(); 310 setCodeCompletionReached(); 311 continue; 312 } 313 314 // If this is the end of the buffer, we have an error. 315 if (Tok.is(tok::eof)) { 316 // Emit errors for each unterminated conditional on the stack, including 317 // the current one. 318 while (!CurPPLexer->ConditionalStack.empty()) { 319 if (CurLexer->getFileLoc() != CodeCompletionFileLoc) 320 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 321 diag::err_pp_unterminated_conditional); 322 CurPPLexer->ConditionalStack.pop_back(); 323 } 324 325 // Just return and let the caller lex after this #include. 326 break; 327 } 328 329 // If this token is not a preprocessor directive, just skip it. 330 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 331 continue; 332 333 // We just parsed a # character at the start of a line, so we're in 334 // directive mode. Tell the lexer this so any newlines we see will be 335 // converted into an EOD token (this terminates the macro). 336 CurPPLexer->ParsingPreprocessorDirective = true; 337 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false); 338 339 340 // Read the next token, the directive flavor. 341 LexUnexpandedToken(Tok); 342 343 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 344 // something bogus), skip it. 345 if (Tok.isNot(tok::raw_identifier)) { 346 CurPPLexer->ParsingPreprocessorDirective = false; 347 // Restore comment saving mode. 348 if (CurLexer) CurLexer->resetExtendedTokenMode(); 349 continue; 350 } 351 352 // If the first letter isn't i or e, it isn't intesting to us. We know that 353 // this is safe in the face of spelling differences, because there is no way 354 // to spell an i/e in a strange way that is another letter. Skipping this 355 // allows us to avoid looking up the identifier info for #define/#undef and 356 // other common directives. 357 StringRef RI = Tok.getRawIdentifier(); 358 359 char FirstChar = RI[0]; 360 if (FirstChar >= 'a' && FirstChar <= 'z' && 361 FirstChar != 'i' && FirstChar != 'e') { 362 CurPPLexer->ParsingPreprocessorDirective = false; 363 // Restore comment saving mode. 364 if (CurLexer) CurLexer->resetExtendedTokenMode(); 365 continue; 366 } 367 368 // Get the identifier name without trigraphs or embedded newlines. Note 369 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 370 // when skipping. 371 char DirectiveBuf[20]; 372 StringRef Directive; 373 if (!Tok.needsCleaning() && RI.size() < 20) { 374 Directive = RI; 375 } else { 376 std::string DirectiveStr = getSpelling(Tok); 377 unsigned IdLen = DirectiveStr.size(); 378 if (IdLen >= 20) { 379 CurPPLexer->ParsingPreprocessorDirective = false; 380 // Restore comment saving mode. 381 if (CurLexer) CurLexer->resetExtendedTokenMode(); 382 continue; 383 } 384 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 385 Directive = StringRef(DirectiveBuf, IdLen); 386 } 387 388 if (Directive.startswith("if")) { 389 StringRef Sub = Directive.substr(2); 390 if (Sub.empty() || // "if" 391 Sub == "def" || // "ifdef" 392 Sub == "ndef") { // "ifndef" 393 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 394 // bother parsing the condition. 395 DiscardUntilEndOfDirective(); 396 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 397 /*foundnonskip*/false, 398 /*foundelse*/false); 399 } 400 } else if (Directive[0] == 'e') { 401 StringRef Sub = Directive.substr(1); 402 if (Sub == "ndif") { // "endif" 403 PPConditionalInfo CondInfo; 404 CondInfo.WasSkipping = true; // Silence bogus warning. 405 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 406 (void)InCond; // Silence warning in no-asserts mode. 407 assert(!InCond && "Can't be skipping if not in a conditional!"); 408 409 // If we popped the outermost skipping block, we're done skipping! 410 if (!CondInfo.WasSkipping) { 411 // Restore the value of LexingRawMode so that trailing comments 412 // are handled correctly, if we've reached the outermost block. 413 CurPPLexer->LexingRawMode = false; 414 CheckEndOfDirective("endif"); 415 CurPPLexer->LexingRawMode = true; 416 if (Callbacks) 417 Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc); 418 break; 419 } else { 420 DiscardUntilEndOfDirective(); 421 } 422 } else if (Sub == "lse") { // "else". 423 // #else directive in a skipping conditional. If not in some other 424 // skipping conditional, and if #else hasn't already been seen, enter it 425 // as a non-skipping conditional. 426 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 427 428 // If this is a #else with a #else before it, report the error. 429 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 430 431 // Note that we've seen a #else in this conditional. 432 CondInfo.FoundElse = true; 433 434 // If the conditional is at the top level, and the #if block wasn't 435 // entered, enter the #else block now. 436 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 437 CondInfo.FoundNonSkip = true; 438 // Restore the value of LexingRawMode so that trailing comments 439 // are handled correctly. 440 CurPPLexer->LexingRawMode = false; 441 CheckEndOfDirective("else"); 442 CurPPLexer->LexingRawMode = true; 443 if (Callbacks) 444 Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc); 445 break; 446 } else { 447 DiscardUntilEndOfDirective(); // C99 6.10p4. 448 } 449 } else if (Sub == "lif") { // "elif". 450 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 451 452 // If this is a #elif with a #else before it, report the error. 453 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 454 455 // If this is in a skipping block or if we're already handled this #if 456 // block, don't bother parsing the condition. 457 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 458 DiscardUntilEndOfDirective(); 459 } else { 460 const SourceLocation CondBegin = CurPPLexer->getSourceLocation(); 461 // Restore the value of LexingRawMode so that identifiers are 462 // looked up, etc, inside the #elif expression. 463 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 464 CurPPLexer->LexingRawMode = false; 465 IdentifierInfo *IfNDefMacro = nullptr; 466 const bool CondValue = EvaluateDirectiveExpression(IfNDefMacro); 467 CurPPLexer->LexingRawMode = true; 468 if (Callbacks) { 469 const SourceLocation CondEnd = CurPPLexer->getSourceLocation(); 470 Callbacks->Elif(Tok.getLocation(), 471 SourceRange(CondBegin, CondEnd), 472 (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False), CondInfo.IfLoc); 473 } 474 // If this condition is true, enter it! 475 if (CondValue) { 476 CondInfo.FoundNonSkip = true; 477 break; 478 } 479 } 480 } 481 } 482 483 CurPPLexer->ParsingPreprocessorDirective = false; 484 // Restore comment saving mode. 485 if (CurLexer) CurLexer->resetExtendedTokenMode(); 486 } 487 488 // Finally, if we are out of the conditional (saw an #endif or ran off the end 489 // of the file, just stop skipping and return to lexing whatever came after 490 // the #if block. 491 CurPPLexer->LexingRawMode = false; 492 493 if (Callbacks) { 494 SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc; 495 Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation())); 496 } 497 } 498 499 void Preprocessor::PTHSkipExcludedConditionalBlock() { 500 501 while (1) { 502 assert(CurPTHLexer); 503 assert(CurPTHLexer->LexingRawMode == false); 504 505 // Skip to the next '#else', '#elif', or #endif. 506 if (CurPTHLexer->SkipBlock()) { 507 // We have reached an #endif. Both the '#' and 'endif' tokens 508 // have been consumed by the PTHLexer. Just pop off the condition level. 509 PPConditionalInfo CondInfo; 510 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 511 (void)InCond; // Silence warning in no-asserts mode. 512 assert(!InCond && "Can't be skipping if not in a conditional!"); 513 break; 514 } 515 516 // We have reached a '#else' or '#elif'. Lex the next token to get 517 // the directive flavor. 518 Token Tok; 519 LexUnexpandedToken(Tok); 520 521 // We can actually look up the IdentifierInfo here since we aren't in 522 // raw mode. 523 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 524 525 if (K == tok::pp_else) { 526 // #else: Enter the else condition. We aren't in a nested condition 527 // since we skip those. We're always in the one matching the last 528 // blocked we skipped. 529 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 530 // Note that we've seen a #else in this conditional. 531 CondInfo.FoundElse = true; 532 533 // If the #if block wasn't entered then enter the #else block now. 534 if (!CondInfo.FoundNonSkip) { 535 CondInfo.FoundNonSkip = true; 536 537 // Scan until the eod token. 538 CurPTHLexer->ParsingPreprocessorDirective = true; 539 DiscardUntilEndOfDirective(); 540 CurPTHLexer->ParsingPreprocessorDirective = false; 541 542 break; 543 } 544 545 // Otherwise skip this block. 546 continue; 547 } 548 549 assert(K == tok::pp_elif); 550 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 551 552 // If this is a #elif with a #else before it, report the error. 553 if (CondInfo.FoundElse) 554 Diag(Tok, diag::pp_err_elif_after_else); 555 556 // If this is in a skipping block or if we're already handled this #if 557 // block, don't bother parsing the condition. We just skip this block. 558 if (CondInfo.FoundNonSkip) 559 continue; 560 561 // Evaluate the condition of the #elif. 562 IdentifierInfo *IfNDefMacro = nullptr; 563 CurPTHLexer->ParsingPreprocessorDirective = true; 564 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 565 CurPTHLexer->ParsingPreprocessorDirective = false; 566 567 // If this condition is true, enter it! 568 if (ShouldEnter) { 569 CondInfo.FoundNonSkip = true; 570 break; 571 } 572 573 // Otherwise, skip this block and go to the next one. 574 continue; 575 } 576 } 577 578 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) { 579 ModuleMap &ModMap = HeaderInfo.getModuleMap(); 580 if (SourceMgr.isInMainFile(Loc)) { 581 if (Module *CurMod = getCurrentModule()) 582 return CurMod; // Compiling a module. 583 return HeaderInfo.getModuleMap().SourceModule; // Compiling a source. 584 } 585 // Try to determine the module of the include directive. 586 // FIXME: Look into directly passing the FileEntry from LookupFile instead. 587 FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc)); 588 if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) { 589 // The include comes from a file. 590 return ModMap.findModuleForHeader(EntryOfIncl).getModule(); 591 } else { 592 // The include does not come from a file, 593 // so it is probably a module compilation. 594 return getCurrentModule(); 595 } 596 } 597 598 Module *Preprocessor::getModuleContainingLocation(SourceLocation Loc) { 599 return HeaderInfo.getModuleMap().inferModuleFromLocation( 600 FullSourceLoc(Loc, SourceMgr)); 601 } 602 603 const FileEntry *Preprocessor::LookupFile( 604 SourceLocation FilenameLoc, 605 StringRef Filename, 606 bool isAngled, 607 const DirectoryLookup *FromDir, 608 const FileEntry *FromFile, 609 const DirectoryLookup *&CurDir, 610 SmallVectorImpl<char> *SearchPath, 611 SmallVectorImpl<char> *RelativePath, 612 ModuleMap::KnownHeader *SuggestedModule, 613 bool SkipCache) { 614 // If the header lookup mechanism may be relative to the current inclusion 615 // stack, record the parent #includes. 616 SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16> 617 Includers; 618 if (!FromDir && !FromFile) { 619 FileID FID = getCurrentFileLexer()->getFileID(); 620 const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID); 621 622 // If there is no file entry associated with this file, it must be the 623 // predefines buffer or the module includes buffer. Any other file is not 624 // lexed with a normal lexer, so it won't be scanned for preprocessor 625 // directives. 626 // 627 // If we have the predefines buffer, resolve #include references (which come 628 // from the -include command line argument) from the current working 629 // directory instead of relative to the main file. 630 // 631 // If we have the module includes buffer, resolve #include references (which 632 // come from header declarations in the module map) relative to the module 633 // map file. 634 if (!FileEnt) { 635 if (FID == SourceMgr.getMainFileID() && MainFileDir) 636 Includers.push_back(std::make_pair(nullptr, MainFileDir)); 637 else if ((FileEnt = 638 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()))) 639 Includers.push_back(std::make_pair(FileEnt, FileMgr.getDirectory("."))); 640 } else { 641 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir())); 642 } 643 644 // MSVC searches the current include stack from top to bottom for 645 // headers included by quoted include directives. 646 // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx 647 if (LangOpts.MSVCCompat && !isAngled) { 648 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 649 IncludeStackInfo &ISEntry = IncludeMacroStack[e - i - 1]; 650 if (IsFileLexer(ISEntry)) 651 if ((FileEnt = SourceMgr.getFileEntryForID( 652 ISEntry.ThePPLexer->getFileID()))) 653 Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir())); 654 } 655 } 656 } 657 658 CurDir = CurDirLookup; 659 660 if (FromFile) { 661 // We're supposed to start looking from after a particular file. Search 662 // the include path until we find that file or run out of files. 663 const DirectoryLookup *TmpCurDir = CurDir; 664 const DirectoryLookup *TmpFromDir = nullptr; 665 while (const FileEntry *FE = HeaderInfo.LookupFile( 666 Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir, 667 Includers, SearchPath, RelativePath, SuggestedModule, 668 SkipCache)) { 669 // Keep looking as if this file did a #include_next. 670 TmpFromDir = TmpCurDir; 671 ++TmpFromDir; 672 if (FE == FromFile) { 673 // Found it. 674 FromDir = TmpFromDir; 675 CurDir = TmpCurDir; 676 break; 677 } 678 } 679 } 680 681 // Do a standard file entry lookup. 682 const FileEntry *FE = HeaderInfo.LookupFile( 683 Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath, 684 RelativePath, SuggestedModule, SkipCache); 685 if (FE) { 686 if (SuggestedModule && !LangOpts.AsmPreprocessor) 687 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 688 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE); 689 return FE; 690 } 691 692 const FileEntry *CurFileEnt; 693 // Otherwise, see if this is a subframework header. If so, this is relative 694 // to one of the headers on the #include stack. Walk the list of the current 695 // headers on the #include stack and pass them to HeaderInfo. 696 if (IsFileLexer()) { 697 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) { 698 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt, 699 SearchPath, RelativePath, 700 SuggestedModule))) { 701 if (SuggestedModule && !LangOpts.AsmPreprocessor) 702 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 703 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE); 704 return FE; 705 } 706 } 707 } 708 709 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 710 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 711 if (IsFileLexer(ISEntry)) { 712 if ((CurFileEnt = 713 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) { 714 if ((FE = HeaderInfo.LookupSubframeworkHeader( 715 Filename, CurFileEnt, SearchPath, RelativePath, 716 SuggestedModule))) { 717 if (SuggestedModule && !LangOpts.AsmPreprocessor) 718 HeaderInfo.getModuleMap().diagnoseHeaderInclusion( 719 getModuleForLocation(FilenameLoc), FilenameLoc, Filename, FE); 720 return FE; 721 } 722 } 723 } 724 } 725 726 // Otherwise, we really couldn't find the file. 727 return nullptr; 728 } 729 730 731 //===----------------------------------------------------------------------===// 732 // Preprocessor Directive Handling. 733 //===----------------------------------------------------------------------===// 734 735 class Preprocessor::ResetMacroExpansionHelper { 736 public: 737 ResetMacroExpansionHelper(Preprocessor *pp) 738 : PP(pp), save(pp->DisableMacroExpansion) { 739 if (pp->MacroExpansionInDirectivesOverride) 740 pp->DisableMacroExpansion = false; 741 } 742 ~ResetMacroExpansionHelper() { 743 PP->DisableMacroExpansion = save; 744 } 745 private: 746 Preprocessor *PP; 747 bool save; 748 }; 749 750 /// HandleDirective - This callback is invoked when the lexer sees a # token 751 /// at the start of a line. This consumes the directive, modifies the 752 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 753 /// read is the correct one. 754 void Preprocessor::HandleDirective(Token &Result) { 755 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 756 757 // We just parsed a # character at the start of a line, so we're in directive 758 // mode. Tell the lexer this so any newlines we see will be converted into an 759 // EOD token (which terminates the directive). 760 CurPPLexer->ParsingPreprocessorDirective = true; 761 if (CurLexer) CurLexer->SetKeepWhitespaceMode(false); 762 763 bool ImmediatelyAfterTopLevelIfndef = 764 CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef(); 765 CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef(); 766 767 ++NumDirectives; 768 769 // We are about to read a token. For the multiple-include optimization FA to 770 // work, we have to remember if we had read any tokens *before* this 771 // pp-directive. 772 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 773 774 // Save the '#' token in case we need to return it later. 775 Token SavedHash = Result; 776 777 // Read the next token, the directive flavor. This isn't expanded due to 778 // C99 6.10.3p8. 779 LexUnexpandedToken(Result); 780 781 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 782 // #define A(x) #x 783 // A(abc 784 // #warning blah 785 // def) 786 // If so, the user is relying on undefined behavior, emit a diagnostic. Do 787 // not support this for #include-like directives, since that can result in 788 // terrible diagnostics, and does not work in GCC. 789 if (InMacroArgs) { 790 if (IdentifierInfo *II = Result.getIdentifierInfo()) { 791 switch (II->getPPKeywordID()) { 792 case tok::pp_include: 793 case tok::pp_import: 794 case tok::pp_include_next: 795 case tok::pp___include_macros: 796 case tok::pp_pragma: 797 Diag(Result, diag::err_embedded_directive) << II->getName(); 798 DiscardUntilEndOfDirective(); 799 return; 800 default: 801 break; 802 } 803 } 804 Diag(Result, diag::ext_embedded_directive); 805 } 806 807 // Temporarily enable macro expansion if set so 808 // and reset to previous state when returning from this function. 809 ResetMacroExpansionHelper helper(this); 810 811 switch (Result.getKind()) { 812 case tok::eod: 813 return; // null directive. 814 case tok::code_completion: 815 if (CodeComplete) 816 CodeComplete->CodeCompleteDirective( 817 CurPPLexer->getConditionalStackDepth() > 0); 818 setCodeCompletionReached(); 819 return; 820 case tok::numeric_constant: // # 7 GNU line marker directive. 821 if (getLangOpts().AsmPreprocessor) 822 break; // # 4 is not a preprocessor directive in .S files. 823 return HandleDigitDirective(Result); 824 default: 825 IdentifierInfo *II = Result.getIdentifierInfo(); 826 if (!II) break; // Not an identifier. 827 828 // Ask what the preprocessor keyword ID is. 829 switch (II->getPPKeywordID()) { 830 default: break; 831 // C99 6.10.1 - Conditional Inclusion. 832 case tok::pp_if: 833 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 834 case tok::pp_ifdef: 835 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 836 case tok::pp_ifndef: 837 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 838 case tok::pp_elif: 839 return HandleElifDirective(Result); 840 case tok::pp_else: 841 return HandleElseDirective(Result); 842 case tok::pp_endif: 843 return HandleEndifDirective(Result); 844 845 // C99 6.10.2 - Source File Inclusion. 846 case tok::pp_include: 847 // Handle #include. 848 return HandleIncludeDirective(SavedHash.getLocation(), Result); 849 case tok::pp___include_macros: 850 // Handle -imacros. 851 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result); 852 853 // C99 6.10.3 - Macro Replacement. 854 case tok::pp_define: 855 return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef); 856 case tok::pp_undef: 857 return HandleUndefDirective(Result); 858 859 // C99 6.10.4 - Line Control. 860 case tok::pp_line: 861 return HandleLineDirective(Result); 862 863 // C99 6.10.5 - Error Directive. 864 case tok::pp_error: 865 return HandleUserDiagnosticDirective(Result, false); 866 867 // C99 6.10.6 - Pragma Directive. 868 case tok::pp_pragma: 869 return HandlePragmaDirective(SavedHash.getLocation(), PIK_HashPragma); 870 871 // GNU Extensions. 872 case tok::pp_import: 873 return HandleImportDirective(SavedHash.getLocation(), Result); 874 case tok::pp_include_next: 875 return HandleIncludeNextDirective(SavedHash.getLocation(), Result); 876 877 case tok::pp_warning: 878 Diag(Result, diag::ext_pp_warning_directive); 879 return HandleUserDiagnosticDirective(Result, true); 880 case tok::pp_ident: 881 return HandleIdentSCCSDirective(Result); 882 case tok::pp_sccs: 883 return HandleIdentSCCSDirective(Result); 884 case tok::pp_assert: 885 //isExtension = true; // FIXME: implement #assert 886 break; 887 case tok::pp_unassert: 888 //isExtension = true; // FIXME: implement #unassert 889 break; 890 891 case tok::pp___public_macro: 892 if (getLangOpts().Modules) 893 return HandleMacroPublicDirective(Result); 894 break; 895 896 case tok::pp___private_macro: 897 if (getLangOpts().Modules) 898 return HandleMacroPrivateDirective(Result); 899 break; 900 } 901 break; 902 } 903 904 // If this is a .S file, treat unknown # directives as non-preprocessor 905 // directives. This is important because # may be a comment or introduce 906 // various pseudo-ops. Just return the # token and push back the following 907 // token to be lexed next time. 908 if (getLangOpts().AsmPreprocessor) { 909 Token *Toks = new Token[2]; 910 // Return the # and the token after it. 911 Toks[0] = SavedHash; 912 Toks[1] = Result; 913 914 // If the second token is a hashhash token, then we need to translate it to 915 // unknown so the token lexer doesn't try to perform token pasting. 916 if (Result.is(tok::hashhash)) 917 Toks[1].setKind(tok::unknown); 918 919 // Enter this token stream so that we re-lex the tokens. Make sure to 920 // enable macro expansion, in case the token after the # is an identifier 921 // that is expanded. 922 EnterTokenStream(Toks, 2, false, true); 923 return; 924 } 925 926 // If we reached here, the preprocessing token is not valid! 927 Diag(Result, diag::err_pp_invalid_directive); 928 929 // Read the rest of the PP line. 930 DiscardUntilEndOfDirective(); 931 932 // Okay, we're done parsing the directive. 933 } 934 935 /// GetLineValue - Convert a numeric token into an unsigned value, emitting 936 /// Diagnostic DiagID if it is invalid, and returning the value in Val. 937 static bool GetLineValue(Token &DigitTok, unsigned &Val, 938 unsigned DiagID, Preprocessor &PP, 939 bool IsGNULineDirective=false) { 940 if (DigitTok.isNot(tok::numeric_constant)) { 941 PP.Diag(DigitTok, DiagID); 942 943 if (DigitTok.isNot(tok::eod)) 944 PP.DiscardUntilEndOfDirective(); 945 return true; 946 } 947 948 SmallString<64> IntegerBuffer; 949 IntegerBuffer.resize(DigitTok.getLength()); 950 const char *DigitTokBegin = &IntegerBuffer[0]; 951 bool Invalid = false; 952 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 953 if (Invalid) 954 return true; 955 956 // Verify that we have a simple digit-sequence, and compute the value. This 957 // is always a simple digit string computed in decimal, so we do this manually 958 // here. 959 Val = 0; 960 for (unsigned i = 0; i != ActualLength; ++i) { 961 // C++1y [lex.fcon]p1: 962 // Optional separating single quotes in a digit-sequence are ignored 963 if (DigitTokBegin[i] == '\'') 964 continue; 965 966 if (!isDigit(DigitTokBegin[i])) { 967 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 968 diag::err_pp_line_digit_sequence) << IsGNULineDirective; 969 PP.DiscardUntilEndOfDirective(); 970 return true; 971 } 972 973 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 974 if (NextVal < Val) { // overflow. 975 PP.Diag(DigitTok, DiagID); 976 PP.DiscardUntilEndOfDirective(); 977 return true; 978 } 979 Val = NextVal; 980 } 981 982 if (DigitTokBegin[0] == '0' && Val) 983 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal) 984 << IsGNULineDirective; 985 986 return false; 987 } 988 989 /// \brief Handle a \#line directive: C99 6.10.4. 990 /// 991 /// The two acceptable forms are: 992 /// \verbatim 993 /// # line digit-sequence 994 /// # line digit-sequence "s-char-sequence" 995 /// \endverbatim 996 void Preprocessor::HandleLineDirective(Token &Tok) { 997 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 998 // expanded. 999 Token DigitTok; 1000 Lex(DigitTok); 1001 1002 // Validate the number and convert it to an unsigned. 1003 unsigned LineNo; 1004 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 1005 return; 1006 1007 if (LineNo == 0) 1008 Diag(DigitTok, diag::ext_pp_line_zero); 1009 1010 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 1011 // number greater than 2147483647". C90 requires that the line # be <= 32767. 1012 unsigned LineLimit = 32768U; 1013 if (LangOpts.C99 || LangOpts.CPlusPlus11) 1014 LineLimit = 2147483648U; 1015 if (LineNo >= LineLimit) 1016 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 1017 else if (LangOpts.CPlusPlus11 && LineNo >= 32768U) 1018 Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big); 1019 1020 int FilenameID = -1; 1021 Token StrTok; 1022 Lex(StrTok); 1023 1024 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a 1025 // string followed by eod. 1026 if (StrTok.is(tok::eod)) 1027 ; // ok 1028 else if (StrTok.isNot(tok::string_literal)) { 1029 Diag(StrTok, diag::err_pp_line_invalid_filename); 1030 return DiscardUntilEndOfDirective(); 1031 } else if (StrTok.hasUDSuffix()) { 1032 Diag(StrTok, diag::err_invalid_string_udl); 1033 return DiscardUntilEndOfDirective(); 1034 } else { 1035 // Parse and validate the string, converting it into a unique ID. 1036 StringLiteralParser Literal(StrTok, *this); 1037 assert(Literal.isAscii() && "Didn't allow wide strings in"); 1038 if (Literal.hadError) 1039 return DiscardUntilEndOfDirective(); 1040 if (Literal.Pascal) { 1041 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1042 return DiscardUntilEndOfDirective(); 1043 } 1044 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString()); 1045 1046 // Verify that there is nothing after the string, other than EOD. Because 1047 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 1048 CheckEndOfDirective("line", true); 1049 } 1050 1051 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 1052 1053 if (Callbacks) 1054 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 1055 PPCallbacks::RenameFile, 1056 SrcMgr::C_User); 1057 } 1058 1059 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 1060 /// marker directive. 1061 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 1062 bool &IsSystemHeader, bool &IsExternCHeader, 1063 Preprocessor &PP) { 1064 unsigned FlagVal; 1065 Token FlagTok; 1066 PP.Lex(FlagTok); 1067 if (FlagTok.is(tok::eod)) return false; 1068 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 1069 return true; 1070 1071 if (FlagVal == 1) { 1072 IsFileEntry = true; 1073 1074 PP.Lex(FlagTok); 1075 if (FlagTok.is(tok::eod)) return false; 1076 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 1077 return true; 1078 } else if (FlagVal == 2) { 1079 IsFileExit = true; 1080 1081 SourceManager &SM = PP.getSourceManager(); 1082 // If we are leaving the current presumed file, check to make sure the 1083 // presumed include stack isn't empty! 1084 FileID CurFileID = 1085 SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first; 1086 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 1087 if (PLoc.isInvalid()) 1088 return true; 1089 1090 // If there is no include loc (main file) or if the include loc is in a 1091 // different physical file, then we aren't in a "1" line marker flag region. 1092 SourceLocation IncLoc = PLoc.getIncludeLoc(); 1093 if (IncLoc.isInvalid() || 1094 SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) { 1095 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 1096 PP.DiscardUntilEndOfDirective(); 1097 return true; 1098 } 1099 1100 PP.Lex(FlagTok); 1101 if (FlagTok.is(tok::eod)) return false; 1102 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 1103 return true; 1104 } 1105 1106 // We must have 3 if there are still flags. 1107 if (FlagVal != 3) { 1108 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1109 PP.DiscardUntilEndOfDirective(); 1110 return true; 1111 } 1112 1113 IsSystemHeader = true; 1114 1115 PP.Lex(FlagTok); 1116 if (FlagTok.is(tok::eod)) return false; 1117 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 1118 return true; 1119 1120 // We must have 4 if there is yet another flag. 1121 if (FlagVal != 4) { 1122 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1123 PP.DiscardUntilEndOfDirective(); 1124 return true; 1125 } 1126 1127 IsExternCHeader = true; 1128 1129 PP.Lex(FlagTok); 1130 if (FlagTok.is(tok::eod)) return false; 1131 1132 // There are no more valid flags here. 1133 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 1134 PP.DiscardUntilEndOfDirective(); 1135 return true; 1136 } 1137 1138 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 1139 /// one of the following forms: 1140 /// 1141 /// # 42 1142 /// # 42 "file" ('1' | '2')? 1143 /// # 42 "file" ('1' | '2')? '3' '4'? 1144 /// 1145 void Preprocessor::HandleDigitDirective(Token &DigitTok) { 1146 // Validate the number and convert it to an unsigned. GNU does not have a 1147 // line # limit other than it fit in 32-bits. 1148 unsigned LineNo; 1149 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 1150 *this, true)) 1151 return; 1152 1153 Token StrTok; 1154 Lex(StrTok); 1155 1156 bool IsFileEntry = false, IsFileExit = false; 1157 bool IsSystemHeader = false, IsExternCHeader = false; 1158 int FilenameID = -1; 1159 1160 // If the StrTok is "eod", then it wasn't present. Otherwise, it must be a 1161 // string followed by eod. 1162 if (StrTok.is(tok::eod)) 1163 ; // ok 1164 else if (StrTok.isNot(tok::string_literal)) { 1165 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1166 return DiscardUntilEndOfDirective(); 1167 } else if (StrTok.hasUDSuffix()) { 1168 Diag(StrTok, diag::err_invalid_string_udl); 1169 return DiscardUntilEndOfDirective(); 1170 } else { 1171 // Parse and validate the string, converting it into a unique ID. 1172 StringLiteralParser Literal(StrTok, *this); 1173 assert(Literal.isAscii() && "Didn't allow wide strings in"); 1174 if (Literal.hadError) 1175 return DiscardUntilEndOfDirective(); 1176 if (Literal.Pascal) { 1177 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 1178 return DiscardUntilEndOfDirective(); 1179 } 1180 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString()); 1181 1182 // If a filename was present, read any flags that are present. 1183 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 1184 IsSystemHeader, IsExternCHeader, *this)) 1185 return; 1186 } 1187 1188 // Create a line note with this information. 1189 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 1190 IsFileEntry, IsFileExit, 1191 IsSystemHeader, IsExternCHeader); 1192 1193 // If the preprocessor has callbacks installed, notify them of the #line 1194 // change. This is used so that the line marker comes out in -E mode for 1195 // example. 1196 if (Callbacks) { 1197 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 1198 if (IsFileEntry) 1199 Reason = PPCallbacks::EnterFile; 1200 else if (IsFileExit) 1201 Reason = PPCallbacks::ExitFile; 1202 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 1203 if (IsExternCHeader) 1204 FileKind = SrcMgr::C_ExternCSystem; 1205 else if (IsSystemHeader) 1206 FileKind = SrcMgr::C_System; 1207 1208 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 1209 } 1210 } 1211 1212 1213 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 1214 /// 1215 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 1216 bool isWarning) { 1217 // PTH doesn't emit #warning or #error directives. 1218 if (CurPTHLexer) 1219 return CurPTHLexer->DiscardToEndOfLine(); 1220 1221 // Read the rest of the line raw. We do this because we don't want macros 1222 // to be expanded and we don't require that the tokens be valid preprocessing 1223 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 1224 // collapse multiple consequtive white space between tokens, but this isn't 1225 // specified by the standard. 1226 SmallString<128> Message; 1227 CurLexer->ReadToEndOfLine(&Message); 1228 1229 // Find the first non-whitespace character, so that we can make the 1230 // diagnostic more succinct. 1231 StringRef Msg = StringRef(Message).ltrim(" "); 1232 1233 if (isWarning) 1234 Diag(Tok, diag::pp_hash_warning) << Msg; 1235 else 1236 Diag(Tok, diag::err_pp_hash_error) << Msg; 1237 } 1238 1239 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 1240 /// 1241 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 1242 // Yes, this directive is an extension. 1243 Diag(Tok, diag::ext_pp_ident_directive); 1244 1245 // Read the string argument. 1246 Token StrTok; 1247 Lex(StrTok); 1248 1249 // If the token kind isn't a string, it's a malformed directive. 1250 if (StrTok.isNot(tok::string_literal) && 1251 StrTok.isNot(tok::wide_string_literal)) { 1252 Diag(StrTok, diag::err_pp_malformed_ident); 1253 if (StrTok.isNot(tok::eod)) 1254 DiscardUntilEndOfDirective(); 1255 return; 1256 } 1257 1258 if (StrTok.hasUDSuffix()) { 1259 Diag(StrTok, diag::err_invalid_string_udl); 1260 return DiscardUntilEndOfDirective(); 1261 } 1262 1263 // Verify that there is nothing after the string, other than EOD. 1264 CheckEndOfDirective("ident"); 1265 1266 if (Callbacks) { 1267 bool Invalid = false; 1268 std::string Str = getSpelling(StrTok, &Invalid); 1269 if (!Invalid) 1270 Callbacks->Ident(Tok.getLocation(), Str); 1271 } 1272 } 1273 1274 /// \brief Handle a #public directive. 1275 void Preprocessor::HandleMacroPublicDirective(Token &Tok) { 1276 Token MacroNameTok; 1277 ReadMacroName(MacroNameTok, MU_Undef); 1278 1279 // Error reading macro name? If so, diagnostic already issued. 1280 if (MacroNameTok.is(tok::eod)) 1281 return; 1282 1283 // Check to see if this is the last token on the #__public_macro line. 1284 CheckEndOfDirective("__public_macro"); 1285 1286 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 1287 // Okay, we finally have a valid identifier to undef. 1288 MacroDirective *MD = getLocalMacroDirective(II); 1289 1290 // If the macro is not defined, this is an error. 1291 if (!MD) { 1292 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II; 1293 return; 1294 } 1295 1296 // Note that this macro has now been exported. 1297 appendMacroDirective(II, AllocateVisibilityMacroDirective( 1298 MacroNameTok.getLocation(), /*IsPublic=*/true)); 1299 } 1300 1301 /// \brief Handle a #private directive. 1302 void Preprocessor::HandleMacroPrivateDirective(Token &Tok) { 1303 Token MacroNameTok; 1304 ReadMacroName(MacroNameTok, MU_Undef); 1305 1306 // Error reading macro name? If so, diagnostic already issued. 1307 if (MacroNameTok.is(tok::eod)) 1308 return; 1309 1310 // Check to see if this is the last token on the #__private_macro line. 1311 CheckEndOfDirective("__private_macro"); 1312 1313 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 1314 // Okay, we finally have a valid identifier to undef. 1315 MacroDirective *MD = getLocalMacroDirective(II); 1316 1317 // If the macro is not defined, this is an error. 1318 if (!MD) { 1319 Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II; 1320 return; 1321 } 1322 1323 // Note that this macro has now been marked private. 1324 appendMacroDirective(II, AllocateVisibilityMacroDirective( 1325 MacroNameTok.getLocation(), /*IsPublic=*/false)); 1326 } 1327 1328 //===----------------------------------------------------------------------===// 1329 // Preprocessor Include Directive Handling. 1330 //===----------------------------------------------------------------------===// 1331 1332 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 1333 /// checked and spelled filename, e.g. as an operand of \#include. This returns 1334 /// true if the input filename was in <>'s or false if it were in ""'s. The 1335 /// caller is expected to provide a buffer that is large enough to hold the 1336 /// spelling of the filename, but is also expected to handle the case when 1337 /// this method decides to use a different buffer. 1338 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 1339 StringRef &Buffer) { 1340 // Get the text form of the filename. 1341 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 1342 1343 // Make sure the filename is <x> or "x". 1344 bool isAngled; 1345 if (Buffer[0] == '<') { 1346 if (Buffer.back() != '>') { 1347 Diag(Loc, diag::err_pp_expects_filename); 1348 Buffer = StringRef(); 1349 return true; 1350 } 1351 isAngled = true; 1352 } else if (Buffer[0] == '"') { 1353 if (Buffer.back() != '"') { 1354 Diag(Loc, diag::err_pp_expects_filename); 1355 Buffer = StringRef(); 1356 return true; 1357 } 1358 isAngled = false; 1359 } else { 1360 Diag(Loc, diag::err_pp_expects_filename); 1361 Buffer = StringRef(); 1362 return true; 1363 } 1364 1365 // Diagnose #include "" as invalid. 1366 if (Buffer.size() <= 2) { 1367 Diag(Loc, diag::err_pp_empty_filename); 1368 Buffer = StringRef(); 1369 return true; 1370 } 1371 1372 // Skip the brackets. 1373 Buffer = Buffer.substr(1, Buffer.size()-2); 1374 return isAngled; 1375 } 1376 1377 // \brief Handle cases where the \#include name is expanded from a macro 1378 // as multiple tokens, which need to be glued together. 1379 // 1380 // This occurs for code like: 1381 // \code 1382 // \#define FOO <a/b.h> 1383 // \#include FOO 1384 // \endcode 1385 // because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1386 // 1387 // This code concatenates and consumes tokens up to the '>' token. It returns 1388 // false if the > was found, otherwise it returns true if it finds and consumes 1389 // the EOD marker. 1390 bool Preprocessor::ConcatenateIncludeName(SmallString<128> &FilenameBuffer, 1391 SourceLocation &End) { 1392 Token CurTok; 1393 1394 Lex(CurTok); 1395 while (CurTok.isNot(tok::eod)) { 1396 End = CurTok.getLocation(); 1397 1398 // FIXME: Provide code completion for #includes. 1399 if (CurTok.is(tok::code_completion)) { 1400 setCodeCompletionReached(); 1401 Lex(CurTok); 1402 continue; 1403 } 1404 1405 // Append the spelling of this token to the buffer. If there was a space 1406 // before it, add it now. 1407 if (CurTok.hasLeadingSpace()) 1408 FilenameBuffer.push_back(' '); 1409 1410 // Get the spelling of the token, directly into FilenameBuffer if possible. 1411 unsigned PreAppendSize = FilenameBuffer.size(); 1412 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1413 1414 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1415 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1416 1417 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1418 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1419 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1420 1421 // Resize FilenameBuffer to the correct size. 1422 if (CurTok.getLength() != ActualLen) 1423 FilenameBuffer.resize(PreAppendSize+ActualLen); 1424 1425 // If we found the '>' marker, return success. 1426 if (CurTok.is(tok::greater)) 1427 return false; 1428 1429 Lex(CurTok); 1430 } 1431 1432 // If we hit the eod marker, emit an error and return true so that the caller 1433 // knows the EOD has been read. 1434 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1435 return true; 1436 } 1437 1438 /// \brief Push a token onto the token stream containing an annotation. 1439 static void EnterAnnotationToken(Preprocessor &PP, 1440 SourceLocation Begin, SourceLocation End, 1441 tok::TokenKind Kind, void *AnnotationVal) { 1442 Token *Tok = new Token[1]; 1443 Tok[0].startToken(); 1444 Tok[0].setKind(Kind); 1445 Tok[0].setLocation(Begin); 1446 Tok[0].setAnnotationEndLoc(End); 1447 Tok[0].setAnnotationValue(AnnotationVal); 1448 PP.EnterTokenStream(Tok, 1, true, true); 1449 } 1450 1451 /// HandleIncludeDirective - The "\#include" tokens have just been read, read 1452 /// the file to be included from the lexer, then include it! This is a common 1453 /// routine with functionality shared between \#include, \#include_next and 1454 /// \#import. LookupFrom is set when this is a \#include_next directive, it 1455 /// specifies the file to start searching from. 1456 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc, 1457 Token &IncludeTok, 1458 const DirectoryLookup *LookupFrom, 1459 const FileEntry *LookupFromFile, 1460 bool isImport) { 1461 1462 Token FilenameTok; 1463 CurPPLexer->LexIncludeFilename(FilenameTok); 1464 1465 // Reserve a buffer to get the spelling. 1466 SmallString<128> FilenameBuffer; 1467 StringRef Filename; 1468 SourceLocation End; 1469 SourceLocation CharEnd; // the end of this directive, in characters 1470 1471 switch (FilenameTok.getKind()) { 1472 case tok::eod: 1473 // If the token kind is EOD, the error has already been diagnosed. 1474 return; 1475 1476 case tok::angle_string_literal: 1477 case tok::string_literal: 1478 Filename = getSpelling(FilenameTok, FilenameBuffer); 1479 End = FilenameTok.getLocation(); 1480 CharEnd = End.getLocWithOffset(FilenameTok.getLength()); 1481 break; 1482 1483 case tok::less: 1484 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1485 // case, glue the tokens together into FilenameBuffer and interpret those. 1486 FilenameBuffer.push_back('<'); 1487 if (ConcatenateIncludeName(FilenameBuffer, End)) 1488 return; // Found <eod> but no ">"? Diagnostic already emitted. 1489 Filename = FilenameBuffer; 1490 CharEnd = End.getLocWithOffset(1); 1491 break; 1492 default: 1493 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1494 DiscardUntilEndOfDirective(); 1495 return; 1496 } 1497 1498 CharSourceRange FilenameRange 1499 = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd); 1500 StringRef OriginalFilename = Filename; 1501 bool isAngled = 1502 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1503 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1504 // error. 1505 if (Filename.empty()) { 1506 DiscardUntilEndOfDirective(); 1507 return; 1508 } 1509 1510 // Verify that there is nothing after the filename, other than EOD. Note that 1511 // we allow macros that expand to nothing after the filename, because this 1512 // falls into the category of "#include pp-tokens new-line" specified in 1513 // C99 6.10.2p4. 1514 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1515 1516 // Check that we don't have infinite #include recursion. 1517 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1518 Diag(FilenameTok, diag::err_pp_include_too_deep); 1519 return; 1520 } 1521 1522 // Complain about attempts to #include files in an audit pragma. 1523 if (PragmaARCCFCodeAuditedLoc.isValid()) { 1524 Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited); 1525 Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here); 1526 1527 // Immediately leave the pragma. 1528 PragmaARCCFCodeAuditedLoc = SourceLocation(); 1529 } 1530 1531 if (HeaderInfo.HasIncludeAliasMap()) { 1532 // Map the filename with the brackets still attached. If the name doesn't 1533 // map to anything, fall back on the filename we've already gotten the 1534 // spelling for. 1535 StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename); 1536 if (!NewName.empty()) 1537 Filename = NewName; 1538 } 1539 1540 // Search include directories. 1541 const DirectoryLookup *CurDir; 1542 SmallString<1024> SearchPath; 1543 SmallString<1024> RelativePath; 1544 // We get the raw path only if we have 'Callbacks' to which we later pass 1545 // the path. 1546 ModuleMap::KnownHeader SuggestedModule; 1547 SourceLocation FilenameLoc = FilenameTok.getLocation(); 1548 SmallString<128> NormalizedPath; 1549 if (LangOpts.MSVCCompat) { 1550 NormalizedPath = Filename.str(); 1551 #ifndef LLVM_ON_WIN32 1552 llvm::sys::path::native(NormalizedPath); 1553 #endif 1554 } 1555 const FileEntry *File = LookupFile( 1556 FilenameLoc, LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, 1557 isAngled, LookupFrom, LookupFromFile, CurDir, 1558 Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr, 1559 HeaderInfo.getHeaderSearchOpts().ModuleMaps ? &SuggestedModule : nullptr); 1560 1561 if (Callbacks) { 1562 if (!File) { 1563 // Give the clients a chance to recover. 1564 SmallString<128> RecoveryPath; 1565 if (Callbacks->FileNotFound(Filename, RecoveryPath)) { 1566 if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) { 1567 // Add the recovery path to the list of search paths. 1568 DirectoryLookup DL(DE, SrcMgr::C_User, false); 1569 HeaderInfo.AddSearchPath(DL, isAngled); 1570 1571 // Try the lookup again, skipping the cache. 1572 File = LookupFile( 1573 FilenameLoc, 1574 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled, 1575 LookupFrom, LookupFromFile, CurDir, nullptr, nullptr, 1576 HeaderInfo.getHeaderSearchOpts().ModuleMaps ? &SuggestedModule 1577 : nullptr, 1578 /*SkipCache*/ true); 1579 } 1580 } 1581 } 1582 1583 if (!SuggestedModule || !getLangOpts().Modules) { 1584 // Notify the callback object that we've seen an inclusion directive. 1585 Callbacks->InclusionDirective(HashLoc, IncludeTok, 1586 LangOpts.MSVCCompat ? NormalizedPath.c_str() 1587 : Filename, 1588 isAngled, FilenameRange, File, SearchPath, 1589 RelativePath, /*ImportedModule=*/nullptr); 1590 } 1591 } 1592 1593 if (!File) { 1594 if (!SuppressIncludeNotFoundError) { 1595 // If the file could not be located and it was included via angle 1596 // brackets, we can attempt a lookup as though it were a quoted path to 1597 // provide the user with a possible fixit. 1598 if (isAngled) { 1599 File = LookupFile( 1600 FilenameLoc, 1601 LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, false, 1602 LookupFrom, LookupFromFile, CurDir, 1603 Callbacks ? &SearchPath : nullptr, 1604 Callbacks ? &RelativePath : nullptr, 1605 HeaderInfo.getHeaderSearchOpts().ModuleMaps ? &SuggestedModule 1606 : nullptr); 1607 if (File) { 1608 SourceRange Range(FilenameTok.getLocation(), CharEnd); 1609 Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) << 1610 Filename << 1611 FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\""); 1612 } 1613 } 1614 // If the file is still not found, just go with the vanilla diagnostic 1615 if (!File) 1616 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1617 } 1618 if (!File) 1619 return; 1620 } 1621 1622 // If we are supposed to import a module rather than including the header, 1623 // do so now. 1624 if (SuggestedModule && getLangOpts().Modules && 1625 SuggestedModule.getModule()->getTopLevelModuleName() != 1626 getLangOpts().ImplementationOfModule) { 1627 // Compute the module access path corresponding to this module. 1628 // FIXME: Should we have a second loadModule() overload to avoid this 1629 // extra lookup step? 1630 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path; 1631 for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent) 1632 Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name), 1633 FilenameTok.getLocation())); 1634 std::reverse(Path.begin(), Path.end()); 1635 1636 // Warn that we're replacing the include/import with a module import. 1637 SmallString<128> PathString; 1638 for (unsigned I = 0, N = Path.size(); I != N; ++I) { 1639 if (I) 1640 PathString += '.'; 1641 PathString += Path[I].first->getName(); 1642 } 1643 int IncludeKind = 0; 1644 1645 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) { 1646 case tok::pp_include: 1647 IncludeKind = 0; 1648 break; 1649 1650 case tok::pp_import: 1651 IncludeKind = 1; 1652 break; 1653 1654 case tok::pp_include_next: 1655 IncludeKind = 2; 1656 break; 1657 1658 case tok::pp___include_macros: 1659 IncludeKind = 3; 1660 break; 1661 1662 default: 1663 llvm_unreachable("unknown include directive kind"); 1664 } 1665 1666 // Determine whether we are actually building the module that this 1667 // include directive maps to. 1668 bool BuildingImportedModule 1669 = Path[0].first->getName() == getLangOpts().CurrentModule; 1670 1671 if (!BuildingImportedModule && getLangOpts().ObjC2) { 1672 // If we're not building the imported module, warn that we're going 1673 // to automatically turn this inclusion directive into a module import. 1674 // We only do this in Objective-C, where we have a module-import syntax. 1675 CharSourceRange ReplaceRange(SourceRange(HashLoc, CharEnd), 1676 /*IsTokenRange=*/false); 1677 Diag(HashLoc, diag::warn_auto_module_import) 1678 << IncludeKind << PathString 1679 << FixItHint::CreateReplacement( 1680 ReplaceRange, ("@import " + PathString + ";").str()); 1681 } 1682 1683 // Load the module to import its macros. We'll make the declarations 1684 // visible when the parser gets here. 1685 ModuleLoadResult Imported = TheModuleLoader.loadModule( 1686 IncludeTok.getLocation(), Path, Module::Hidden, 1687 /*IsIncludeDirective=*/true); 1688 if (Imported) 1689 makeModuleVisible(Imported, IncludeTok.getLocation()); 1690 assert((Imported == nullptr || Imported == SuggestedModule.getModule()) && 1691 "the imported module is different than the suggested one"); 1692 1693 if (!Imported && hadModuleLoaderFatalFailure()) { 1694 // With a fatal failure in the module loader, we abort parsing. 1695 Token &Result = IncludeTok; 1696 if (CurLexer) { 1697 Result.startToken(); 1698 CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof); 1699 CurLexer->cutOffLexing(); 1700 } else { 1701 assert(CurPTHLexer && "#include but no current lexer set!"); 1702 CurPTHLexer->getEOF(Result); 1703 } 1704 return; 1705 } 1706 1707 // If this header isn't part of the module we're building, we're done. 1708 if (!BuildingImportedModule && Imported) { 1709 if (Callbacks) { 1710 Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, 1711 FilenameRange, File, 1712 SearchPath, RelativePath, Imported); 1713 } 1714 1715 if (IncludeKind != 3) { 1716 // Let the parser know that we hit a module import, and it should 1717 // make the module visible. 1718 // FIXME: Produce this as the current token directly, rather than 1719 // allocating a new token for it. 1720 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_include, 1721 Imported); 1722 } 1723 return; 1724 } 1725 1726 // If we failed to find a submodule that we expected to find, we can 1727 // continue. Otherwise, there's an error in the included file, so we 1728 // don't want to include it. 1729 if (!BuildingImportedModule && !Imported.isMissingExpected()) { 1730 return; 1731 } 1732 } 1733 1734 if (Callbacks && SuggestedModule) { 1735 // We didn't notify the callback object that we've seen an inclusion 1736 // directive before. Now that we are parsing the include normally and not 1737 // turning it to a module import, notify the callback object. 1738 Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, 1739 FilenameRange, File, 1740 SearchPath, RelativePath, 1741 /*ImportedModule=*/nullptr); 1742 } 1743 1744 // The #included file will be considered to be a system header if either it is 1745 // in a system include directory, or if the #includer is a system include 1746 // header. 1747 SrcMgr::CharacteristicKind FileCharacter = 1748 std::max(HeaderInfo.getFileDirFlavor(File), 1749 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1750 1751 // Ask HeaderInfo if we should enter this #include file. If not, #including 1752 // this file will have no effect. 1753 if (!HeaderInfo.ShouldEnterIncludeFile(*this, File, isImport)) { 1754 if (Callbacks) 1755 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1756 return; 1757 } 1758 1759 // Look up the file, create a File ID for it. 1760 SourceLocation IncludePos = End; 1761 // If the filename string was the result of macro expansions, set the include 1762 // position on the file where it will be included and after the expansions. 1763 if (IncludePos.isMacroID()) 1764 IncludePos = SourceMgr.getExpansionRange(IncludePos).second; 1765 FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter); 1766 assert(!FID.isInvalid() && "Expected valid file ID"); 1767 1768 // Determine if we're switching to building a new submodule, and which one. 1769 // 1770 // FIXME: If we've already processed this header, just make it visible rather 1771 // than entering it again. 1772 ModuleMap::KnownHeader BuildingModule; 1773 if (getLangOpts().Modules && !getLangOpts().CurrentModule.empty()) { 1774 Module *RequestingModule = getModuleForLocation(FilenameLoc); 1775 BuildingModule = 1776 HeaderInfo.getModuleMap().findModuleForHeader(File, RequestingModule); 1777 } 1778 1779 // If all is good, enter the new file! 1780 if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation())) 1781 return; 1782 1783 // If we're walking into another part of the same module, let the parser 1784 // know that any future declarations are within that other submodule. 1785 if (BuildingModule) { 1786 assert(!CurSubmodule && "should not have marked this as a module yet"); 1787 CurSubmodule = BuildingModule.getModule(); 1788 1789 EnterSubmodule(CurSubmodule, HashLoc); 1790 1791 EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_begin, 1792 CurSubmodule); 1793 } 1794 } 1795 1796 /// HandleIncludeNextDirective - Implements \#include_next. 1797 /// 1798 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc, 1799 Token &IncludeNextTok) { 1800 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1801 1802 // #include_next is like #include, except that we start searching after 1803 // the current found directory. If we can't do this, issue a 1804 // diagnostic. 1805 const DirectoryLookup *Lookup = CurDirLookup; 1806 const FileEntry *LookupFromFile = nullptr; 1807 if (isInPrimaryFile()) { 1808 Lookup = nullptr; 1809 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1810 } else if (CurSubmodule) { 1811 // Start looking up in the directory *after* the one in which the current 1812 // file would be found, if any. 1813 assert(CurPPLexer && "#include_next directive in macro?"); 1814 LookupFromFile = CurPPLexer->getFileEntry(); 1815 Lookup = nullptr; 1816 } else if (!Lookup) { 1817 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1818 } else { 1819 // Start looking up in the next directory. 1820 ++Lookup; 1821 } 1822 1823 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup, 1824 LookupFromFile); 1825 } 1826 1827 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode 1828 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) { 1829 // The Microsoft #import directive takes a type library and generates header 1830 // files from it, and includes those. This is beyond the scope of what clang 1831 // does, so we ignore it and error out. However, #import can optionally have 1832 // trailing attributes that span multiple lines. We're going to eat those 1833 // so we can continue processing from there. 1834 Diag(Tok, diag::err_pp_import_directive_ms ); 1835 1836 // Read tokens until we get to the end of the directive. Note that the 1837 // directive can be split over multiple lines using the backslash character. 1838 DiscardUntilEndOfDirective(); 1839 } 1840 1841 /// HandleImportDirective - Implements \#import. 1842 /// 1843 void Preprocessor::HandleImportDirective(SourceLocation HashLoc, 1844 Token &ImportTok) { 1845 if (!LangOpts.ObjC1) { // #import is standard for ObjC. 1846 if (LangOpts.MSVCCompat) 1847 return HandleMicrosoftImportDirective(ImportTok); 1848 Diag(ImportTok, diag::ext_pp_import_directive); 1849 } 1850 return HandleIncludeDirective(HashLoc, ImportTok, nullptr, nullptr, true); 1851 } 1852 1853 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1854 /// pseudo directive in the predefines buffer. This handles it by sucking all 1855 /// tokens through the preprocessor and discarding them (only keeping the side 1856 /// effects on the preprocessor). 1857 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc, 1858 Token &IncludeMacrosTok) { 1859 // This directive should only occur in the predefines buffer. If not, emit an 1860 // error and reject it. 1861 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1862 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1863 Diag(IncludeMacrosTok.getLocation(), 1864 diag::pp_include_macros_out_of_predefines); 1865 DiscardUntilEndOfDirective(); 1866 return; 1867 } 1868 1869 // Treat this as a normal #include for checking purposes. If this is 1870 // successful, it will push a new lexer onto the include stack. 1871 HandleIncludeDirective(HashLoc, IncludeMacrosTok); 1872 1873 Token TmpTok; 1874 do { 1875 Lex(TmpTok); 1876 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1877 } while (TmpTok.isNot(tok::hashhash)); 1878 } 1879 1880 //===----------------------------------------------------------------------===// 1881 // Preprocessor Macro Directive Handling. 1882 //===----------------------------------------------------------------------===// 1883 1884 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1885 /// definition has just been read. Lex the rest of the arguments and the 1886 /// closing ), updating MI with what we learn. Return true if an error occurs 1887 /// parsing the arg list. 1888 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) { 1889 SmallVector<IdentifierInfo*, 32> Arguments; 1890 1891 while (1) { 1892 LexUnexpandedToken(Tok); 1893 switch (Tok.getKind()) { 1894 case tok::r_paren: 1895 // Found the end of the argument list. 1896 if (Arguments.empty()) // #define FOO() 1897 return false; 1898 // Otherwise we have #define FOO(A,) 1899 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1900 return true; 1901 case tok::ellipsis: // #define X(... -> C99 varargs 1902 if (!LangOpts.C99) 1903 Diag(Tok, LangOpts.CPlusPlus11 ? 1904 diag::warn_cxx98_compat_variadic_macro : 1905 diag::ext_variadic_macro); 1906 1907 // OpenCL v1.2 s6.9.e: variadic macros are not supported. 1908 if (LangOpts.OpenCL) { 1909 Diag(Tok, diag::err_pp_opencl_variadic_macros); 1910 return true; 1911 } 1912 1913 // Lex the token after the identifier. 1914 LexUnexpandedToken(Tok); 1915 if (Tok.isNot(tok::r_paren)) { 1916 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1917 return true; 1918 } 1919 // Add the __VA_ARGS__ identifier as an argument. 1920 Arguments.push_back(Ident__VA_ARGS__); 1921 MI->setIsC99Varargs(); 1922 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1923 return false; 1924 case tok::eod: // #define X( 1925 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1926 return true; 1927 default: 1928 // Handle keywords and identifiers here to accept things like 1929 // #define Foo(for) for. 1930 IdentifierInfo *II = Tok.getIdentifierInfo(); 1931 if (!II) { 1932 // #define X(1 1933 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1934 return true; 1935 } 1936 1937 // If this is already used as an argument, it is used multiple times (e.g. 1938 // #define X(A,A. 1939 if (std::find(Arguments.begin(), Arguments.end(), II) != 1940 Arguments.end()) { // C99 6.10.3p6 1941 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1942 return true; 1943 } 1944 1945 // Add the argument to the macro info. 1946 Arguments.push_back(II); 1947 1948 // Lex the token after the identifier. 1949 LexUnexpandedToken(Tok); 1950 1951 switch (Tok.getKind()) { 1952 default: // #define X(A B 1953 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1954 return true; 1955 case tok::r_paren: // #define X(A) 1956 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1957 return false; 1958 case tok::comma: // #define X(A, 1959 break; 1960 case tok::ellipsis: // #define X(A... -> GCC extension 1961 // Diagnose extension. 1962 Diag(Tok, diag::ext_named_variadic_macro); 1963 1964 // Lex the token after the identifier. 1965 LexUnexpandedToken(Tok); 1966 if (Tok.isNot(tok::r_paren)) { 1967 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1968 return true; 1969 } 1970 1971 MI->setIsGNUVarargs(); 1972 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1973 return false; 1974 } 1975 } 1976 } 1977 } 1978 1979 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI, 1980 const LangOptions &LOptions) { 1981 if (MI->getNumTokens() == 1) { 1982 const Token &Value = MI->getReplacementToken(0); 1983 1984 // Macro that is identity, like '#define inline inline' is a valid pattern. 1985 if (MacroName.getKind() == Value.getKind()) 1986 return true; 1987 1988 // Macro that maps a keyword to the same keyword decorated with leading/ 1989 // trailing underscores is a valid pattern: 1990 // #define inline __inline 1991 // #define inline __inline__ 1992 // #define inline _inline (in MS compatibility mode) 1993 StringRef MacroText = MacroName.getIdentifierInfo()->getName(); 1994 if (IdentifierInfo *II = Value.getIdentifierInfo()) { 1995 if (!II->isKeyword(LOptions)) 1996 return false; 1997 StringRef ValueText = II->getName(); 1998 StringRef TrimmedValue = ValueText; 1999 if (!ValueText.startswith("__")) { 2000 if (ValueText.startswith("_")) 2001 TrimmedValue = TrimmedValue.drop_front(1); 2002 else 2003 return false; 2004 } else { 2005 TrimmedValue = TrimmedValue.drop_front(2); 2006 if (TrimmedValue.endswith("__")) 2007 TrimmedValue = TrimmedValue.drop_back(2); 2008 } 2009 return TrimmedValue.equals(MacroText); 2010 } else { 2011 return false; 2012 } 2013 } 2014 2015 // #define inline 2016 if ((MacroName.is(tok::kw_extern) || MacroName.is(tok::kw_inline) || 2017 MacroName.is(tok::kw_static) || MacroName.is(tok::kw_const)) && 2018 MI->getNumTokens() == 0) { 2019 return true; 2020 } 2021 2022 return false; 2023 } 2024 2025 /// HandleDefineDirective - Implements \#define. This consumes the entire macro 2026 /// line then lets the caller lex the next real token. 2027 void Preprocessor::HandleDefineDirective(Token &DefineTok, 2028 bool ImmediatelyAfterHeaderGuard) { 2029 ++NumDefined; 2030 2031 Token MacroNameTok; 2032 bool MacroShadowsKeyword; 2033 ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword); 2034 2035 // Error reading macro name? If so, diagnostic already issued. 2036 if (MacroNameTok.is(tok::eod)) 2037 return; 2038 2039 Token LastTok = MacroNameTok; 2040 2041 // If we are supposed to keep comments in #defines, reenable comment saving 2042 // mode. 2043 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 2044 2045 // Create the new macro. 2046 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 2047 2048 Token Tok; 2049 LexUnexpandedToken(Tok); 2050 2051 // If this is a function-like macro definition, parse the argument list, 2052 // marking each of the identifiers as being used as macro arguments. Also, 2053 // check other constraints on the first token of the macro body. 2054 if (Tok.is(tok::eod)) { 2055 if (ImmediatelyAfterHeaderGuard) { 2056 // Save this macro information since it may part of a header guard. 2057 CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(), 2058 MacroNameTok.getLocation()); 2059 } 2060 // If there is no body to this macro, we have no special handling here. 2061 } else if (Tok.hasLeadingSpace()) { 2062 // This is a normal token with leading space. Clear the leading space 2063 // marker on the first token to get proper expansion. 2064 Tok.clearFlag(Token::LeadingSpace); 2065 } else if (Tok.is(tok::l_paren)) { 2066 // This is a function-like macro definition. Read the argument list. 2067 MI->setIsFunctionLike(); 2068 if (ReadMacroDefinitionArgList(MI, LastTok)) { 2069 // Throw away the rest of the line. 2070 if (CurPPLexer->ParsingPreprocessorDirective) 2071 DiscardUntilEndOfDirective(); 2072 return; 2073 } 2074 2075 // If this is a definition of a variadic C99 function-like macro, not using 2076 // the GNU named varargs extension, enabled __VA_ARGS__. 2077 2078 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 2079 // This gets unpoisoned where it is allowed. 2080 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 2081 if (MI->isC99Varargs()) 2082 Ident__VA_ARGS__->setIsPoisoned(false); 2083 2084 // Read the first token after the arg list for down below. 2085 LexUnexpandedToken(Tok); 2086 } else if (LangOpts.C99 || LangOpts.CPlusPlus11) { 2087 // C99 requires whitespace between the macro definition and the body. Emit 2088 // a diagnostic for something like "#define X+". 2089 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 2090 } else { 2091 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 2092 // first character of a replacement list is not a character required by 2093 // subclause 5.2.1, then there shall be white-space separation between the 2094 // identifier and the replacement list.". 5.2.1 lists this set: 2095 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 2096 // is irrelevant here. 2097 bool isInvalid = false; 2098 if (Tok.is(tok::at)) // @ is not in the list above. 2099 isInvalid = true; 2100 else if (Tok.is(tok::unknown)) { 2101 // If we have an unknown token, it is something strange like "`". Since 2102 // all of valid characters would have lexed into a single character 2103 // token of some sort, we know this is not a valid case. 2104 isInvalid = true; 2105 } 2106 if (isInvalid) 2107 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 2108 else 2109 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 2110 } 2111 2112 if (!Tok.is(tok::eod)) 2113 LastTok = Tok; 2114 2115 // Read the rest of the macro body. 2116 if (MI->isObjectLike()) { 2117 // Object-like macros are very simple, just read their body. 2118 while (Tok.isNot(tok::eod)) { 2119 LastTok = Tok; 2120 MI->AddTokenToBody(Tok); 2121 // Get the next token of the macro. 2122 LexUnexpandedToken(Tok); 2123 } 2124 2125 } else { 2126 // Otherwise, read the body of a function-like macro. While we are at it, 2127 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 2128 // parameters in function-like macro expansions. 2129 while (Tok.isNot(tok::eod)) { 2130 LastTok = Tok; 2131 2132 if (Tok.isNot(tok::hash) && Tok.isNot(tok::hashhash)) { 2133 MI->AddTokenToBody(Tok); 2134 2135 // Get the next token of the macro. 2136 LexUnexpandedToken(Tok); 2137 continue; 2138 } 2139 2140 // If we're in -traditional mode, then we should ignore stringification 2141 // and token pasting. Mark the tokens as unknown so as not to confuse 2142 // things. 2143 if (getLangOpts().TraditionalCPP) { 2144 Tok.setKind(tok::unknown); 2145 MI->AddTokenToBody(Tok); 2146 2147 // Get the next token of the macro. 2148 LexUnexpandedToken(Tok); 2149 continue; 2150 } 2151 2152 if (Tok.is(tok::hashhash)) { 2153 2154 // If we see token pasting, check if it looks like the gcc comma 2155 // pasting extension. We'll use this information to suppress 2156 // diagnostics later on. 2157 2158 // Get the next token of the macro. 2159 LexUnexpandedToken(Tok); 2160 2161 if (Tok.is(tok::eod)) { 2162 MI->AddTokenToBody(LastTok); 2163 break; 2164 } 2165 2166 unsigned NumTokens = MI->getNumTokens(); 2167 if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ && 2168 MI->getReplacementToken(NumTokens-1).is(tok::comma)) 2169 MI->setHasCommaPasting(); 2170 2171 // Things look ok, add the '##' token to the macro. 2172 MI->AddTokenToBody(LastTok); 2173 continue; 2174 } 2175 2176 // Get the next token of the macro. 2177 LexUnexpandedToken(Tok); 2178 2179 // Check for a valid macro arg identifier. 2180 if (Tok.getIdentifierInfo() == nullptr || 2181 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 2182 2183 // If this is assembler-with-cpp mode, we accept random gibberish after 2184 // the '#' because '#' is often a comment character. However, change 2185 // the kind of the token to tok::unknown so that the preprocessor isn't 2186 // confused. 2187 if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) { 2188 LastTok.setKind(tok::unknown); 2189 MI->AddTokenToBody(LastTok); 2190 continue; 2191 } else { 2192 Diag(Tok, diag::err_pp_stringize_not_parameter); 2193 2194 // Disable __VA_ARGS__ again. 2195 Ident__VA_ARGS__->setIsPoisoned(true); 2196 return; 2197 } 2198 } 2199 2200 // Things look ok, add the '#' and param name tokens to the macro. 2201 MI->AddTokenToBody(LastTok); 2202 MI->AddTokenToBody(Tok); 2203 LastTok = Tok; 2204 2205 // Get the next token of the macro. 2206 LexUnexpandedToken(Tok); 2207 } 2208 } 2209 2210 if (MacroShadowsKeyword && 2211 !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) { 2212 Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword); 2213 } 2214 2215 // Disable __VA_ARGS__ again. 2216 Ident__VA_ARGS__->setIsPoisoned(true); 2217 2218 // Check that there is no paste (##) operator at the beginning or end of the 2219 // replacement list. 2220 unsigned NumTokens = MI->getNumTokens(); 2221 if (NumTokens != 0) { 2222 if (MI->getReplacementToken(0).is(tok::hashhash)) { 2223 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 2224 return; 2225 } 2226 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 2227 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 2228 return; 2229 } 2230 } 2231 2232 MI->setDefinitionEndLoc(LastTok.getLocation()); 2233 2234 // Finally, if this identifier already had a macro defined for it, verify that 2235 // the macro bodies are identical, and issue diagnostics if they are not. 2236 if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) { 2237 // It is very common for system headers to have tons of macro redefinitions 2238 // and for warnings to be disabled in system headers. If this is the case, 2239 // then don't bother calling MacroInfo::isIdenticalTo. 2240 if (!getDiagnostics().getSuppressSystemWarnings() || 2241 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 2242 if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused()) 2243 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 2244 2245 // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and 2246 // C++ [cpp.predefined]p4, but allow it as an extension. 2247 if (OtherMI->isBuiltinMacro()) 2248 Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro); 2249 // Macros must be identical. This means all tokens and whitespace 2250 // separation must be the same. C99 6.10.3p2. 2251 else if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 2252 !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) { 2253 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 2254 << MacroNameTok.getIdentifierInfo(); 2255 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 2256 } 2257 } 2258 if (OtherMI->isWarnIfUnused()) 2259 WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc()); 2260 } 2261 2262 DefMacroDirective *MD = 2263 appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI); 2264 2265 assert(!MI->isUsed()); 2266 // If we need warning for not using the macro, add its location in the 2267 // warn-because-unused-macro set. If it gets used it will be removed from set. 2268 if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) && 2269 !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc())) { 2270 MI->setIsWarnIfUnused(true); 2271 WarnUnusedMacroLocs.insert(MI->getDefinitionLoc()); 2272 } 2273 2274 // If the callbacks want to know, tell them about the macro definition. 2275 if (Callbacks) 2276 Callbacks->MacroDefined(MacroNameTok, MD); 2277 } 2278 2279 /// HandleUndefDirective - Implements \#undef. 2280 /// 2281 void Preprocessor::HandleUndefDirective(Token &UndefTok) { 2282 ++NumUndefined; 2283 2284 Token MacroNameTok; 2285 ReadMacroName(MacroNameTok, MU_Undef); 2286 2287 // Error reading macro name? If so, diagnostic already issued. 2288 if (MacroNameTok.is(tok::eod)) 2289 return; 2290 2291 // Check to see if this is the last token on the #undef line. 2292 CheckEndOfDirective("undef"); 2293 2294 // Okay, we have a valid identifier to undef. 2295 auto *II = MacroNameTok.getIdentifierInfo(); 2296 auto MD = getMacroDefinition(II); 2297 2298 // If the callbacks want to know, tell them about the macro #undef. 2299 // Note: no matter if the macro was defined or not. 2300 if (Callbacks) 2301 Callbacks->MacroUndefined(MacroNameTok, MD); 2302 2303 // If the macro is not defined, this is a noop undef, just return. 2304 const MacroInfo *MI = MD.getMacroInfo(); 2305 if (!MI) 2306 return; 2307 2308 if (!MI->isUsed() && MI->isWarnIfUnused()) 2309 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 2310 2311 if (MI->isWarnIfUnused()) 2312 WarnUnusedMacroLocs.erase(MI->getDefinitionLoc()); 2313 2314 appendMacroDirective(MacroNameTok.getIdentifierInfo(), 2315 AllocateUndefMacroDirective(MacroNameTok.getLocation())); 2316 } 2317 2318 2319 //===----------------------------------------------------------------------===// 2320 // Preprocessor Conditional Directive Handling. 2321 //===----------------------------------------------------------------------===// 2322 2323 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive. isIfndef 2324 /// is true when this is a \#ifndef directive. ReadAnyTokensBeforeDirective is 2325 /// true if any tokens have been returned or pp-directives activated before this 2326 /// \#ifndef has been lexed. 2327 /// 2328 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 2329 bool ReadAnyTokensBeforeDirective) { 2330 ++NumIf; 2331 Token DirectiveTok = Result; 2332 2333 Token MacroNameTok; 2334 ReadMacroName(MacroNameTok); 2335 2336 // Error reading macro name? If so, diagnostic already issued. 2337 if (MacroNameTok.is(tok::eod)) { 2338 // Skip code until we get to #endif. This helps with recovery by not 2339 // emitting an error when the #endif is reached. 2340 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 2341 /*Foundnonskip*/false, /*FoundElse*/false); 2342 return; 2343 } 2344 2345 // Check to see if this is the last token on the #if[n]def line. 2346 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 2347 2348 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 2349 auto MD = getMacroDefinition(MII); 2350 MacroInfo *MI = MD.getMacroInfo(); 2351 2352 if (CurPPLexer->getConditionalStackDepth() == 0) { 2353 // If the start of a top-level #ifdef and if the macro is not defined, 2354 // inform MIOpt that this might be the start of a proper include guard. 2355 // Otherwise it is some other form of unknown conditional which we can't 2356 // handle. 2357 if (!ReadAnyTokensBeforeDirective && !MI) { 2358 assert(isIfndef && "#ifdef shouldn't reach here"); 2359 CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation()); 2360 } else 2361 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2362 } 2363 2364 // If there is a macro, process it. 2365 if (MI) // Mark it used. 2366 markMacroAsUsed(MI); 2367 2368 if (Callbacks) { 2369 if (isIfndef) 2370 Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD); 2371 else 2372 Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD); 2373 } 2374 2375 // Should we include the stuff contained by this directive? 2376 if (!MI == isIfndef) { 2377 // Yes, remember that we are inside a conditional, then lex the next token. 2378 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 2379 /*wasskip*/false, /*foundnonskip*/true, 2380 /*foundelse*/false); 2381 } else { 2382 // No, skip the contents of this block. 2383 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 2384 /*Foundnonskip*/false, 2385 /*FoundElse*/false); 2386 } 2387 } 2388 2389 /// HandleIfDirective - Implements the \#if directive. 2390 /// 2391 void Preprocessor::HandleIfDirective(Token &IfToken, 2392 bool ReadAnyTokensBeforeDirective) { 2393 ++NumIf; 2394 2395 // Parse and evaluate the conditional expression. 2396 IdentifierInfo *IfNDefMacro = nullptr; 2397 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 2398 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 2399 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 2400 2401 // If this condition is equivalent to #ifndef X, and if this is the first 2402 // directive seen, handle it for the multiple-include optimization. 2403 if (CurPPLexer->getConditionalStackDepth() == 0) { 2404 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 2405 // FIXME: Pass in the location of the macro name, not the 'if' token. 2406 CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation()); 2407 else 2408 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2409 } 2410 2411 if (Callbacks) 2412 Callbacks->If(IfToken.getLocation(), 2413 SourceRange(ConditionalBegin, ConditionalEnd), 2414 (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False)); 2415 2416 // Should we include the stuff contained by this directive? 2417 if (ConditionalTrue) { 2418 // Yes, remember that we are inside a conditional, then lex the next token. 2419 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 2420 /*foundnonskip*/true, /*foundelse*/false); 2421 } else { 2422 // No, skip the contents of this block. 2423 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 2424 /*FoundElse*/false); 2425 } 2426 } 2427 2428 /// HandleEndifDirective - Implements the \#endif directive. 2429 /// 2430 void Preprocessor::HandleEndifDirective(Token &EndifToken) { 2431 ++NumEndif; 2432 2433 // Check that this is the whole directive. 2434 CheckEndOfDirective("endif"); 2435 2436 PPConditionalInfo CondInfo; 2437 if (CurPPLexer->popConditionalLevel(CondInfo)) { 2438 // No conditionals on the stack: this is an #endif without an #if. 2439 Diag(EndifToken, diag::err_pp_endif_without_if); 2440 return; 2441 } 2442 2443 // If this the end of a top-level #endif, inform MIOpt. 2444 if (CurPPLexer->getConditionalStackDepth() == 0) 2445 CurPPLexer->MIOpt.ExitTopLevelConditional(); 2446 2447 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 2448 "This code should only be reachable in the non-skipping case!"); 2449 2450 if (Callbacks) 2451 Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc); 2452 } 2453 2454 /// HandleElseDirective - Implements the \#else directive. 2455 /// 2456 void Preprocessor::HandleElseDirective(Token &Result) { 2457 ++NumElse; 2458 2459 // #else directive in a non-skipping conditional... start skipping. 2460 CheckEndOfDirective("else"); 2461 2462 PPConditionalInfo CI; 2463 if (CurPPLexer->popConditionalLevel(CI)) { 2464 Diag(Result, diag::pp_err_else_without_if); 2465 return; 2466 } 2467 2468 // If this is a top-level #else, inform the MIOpt. 2469 if (CurPPLexer->getConditionalStackDepth() == 0) 2470 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2471 2472 // If this is a #else with a #else before it, report the error. 2473 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 2474 2475 if (Callbacks) 2476 Callbacks->Else(Result.getLocation(), CI.IfLoc); 2477 2478 // Finally, skip the rest of the contents of this block. 2479 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 2480 /*FoundElse*/true, Result.getLocation()); 2481 } 2482 2483 /// HandleElifDirective - Implements the \#elif directive. 2484 /// 2485 void Preprocessor::HandleElifDirective(Token &ElifToken) { 2486 ++NumElse; 2487 2488 // #elif directive in a non-skipping conditional... start skipping. 2489 // We don't care what the condition is, because we will always skip it (since 2490 // the block immediately before it was included). 2491 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 2492 DiscardUntilEndOfDirective(); 2493 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 2494 2495 PPConditionalInfo CI; 2496 if (CurPPLexer->popConditionalLevel(CI)) { 2497 Diag(ElifToken, diag::pp_err_elif_without_if); 2498 return; 2499 } 2500 2501 // If this is a top-level #elif, inform the MIOpt. 2502 if (CurPPLexer->getConditionalStackDepth() == 0) 2503 CurPPLexer->MIOpt.EnterTopLevelConditional(); 2504 2505 // If this is a #elif with a #else before it, report the error. 2506 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 2507 2508 if (Callbacks) 2509 Callbacks->Elif(ElifToken.getLocation(), 2510 SourceRange(ConditionalBegin, ConditionalEnd), 2511 PPCallbacks::CVK_NotEvaluated, CI.IfLoc); 2512 2513 // Finally, skip the rest of the contents of this block. 2514 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 2515 /*FoundElse*/CI.FoundElse, 2516 ElifToken.getLocation()); 2517 } 2518