PPDirectives.cpp revision c4bf2b9afb7d47445a9dc6bc848657098a4e3851
1//===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9// 10// This file implements # directive processing for the Preprocessor. 11// 12//===----------------------------------------------------------------------===// 13 14#include "clang/Lex/Preprocessor.h" 15#include "clang/Lex/LiteralSupport.h" 16#include "clang/Lex/HeaderSearch.h" 17#include "clang/Lex/MacroInfo.h" 18#include "clang/Lex/LexDiagnostic.h" 19#include "clang/Lex/CodeCompletionHandler.h" 20#include "clang/Lex/Pragma.h" 21#include "clang/Basic/FileManager.h" 22#include "clang/Basic/SourceManager.h" 23#include "llvm/ADT/APInt.h" 24using namespace clang; 25 26//===----------------------------------------------------------------------===// 27// Utility Methods for Preprocessor Directive Handling. 28//===----------------------------------------------------------------------===// 29 30MacroInfo *Preprocessor::AllocateMacroInfo() { 31 MacroInfoChain *MIChain; 32 33 if (MICache) { 34 MIChain = MICache; 35 MICache = MICache->Next; 36 } 37 else { 38 MIChain = BP.Allocate<MacroInfoChain>(); 39 } 40 41 MIChain->Next = MIChainHead; 42 MIChain->Prev = 0; 43 if (MIChainHead) 44 MIChainHead->Prev = MIChain; 45 MIChainHead = MIChain; 46 47 return &(MIChain->MI); 48} 49 50MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 51 MacroInfo *MI = AllocateMacroInfo(); 52 new (MI) MacroInfo(L); 53 return MI; 54} 55 56MacroInfo *Preprocessor::CloneMacroInfo(const MacroInfo &MacroToClone) { 57 MacroInfo *MI = AllocateMacroInfo(); 58 new (MI) MacroInfo(MacroToClone, BP); 59 return MI; 60} 61 62/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 63/// be reused for allocating new MacroInfo objects. 64void Preprocessor::ReleaseMacroInfo(MacroInfo *MI) { 65 MacroInfoChain *MIChain = (MacroInfoChain*) MI; 66 if (MacroInfoChain *Prev = MIChain->Prev) { 67 MacroInfoChain *Next = MIChain->Next; 68 Prev->Next = Next; 69 if (Next) 70 Next->Prev = Prev; 71 } 72 else { 73 assert(MIChainHead == MIChain); 74 MIChainHead = MIChain->Next; 75 MIChainHead->Prev = 0; 76 } 77 MIChain->Next = MICache; 78 MICache = MIChain; 79 80 MI->Destroy(); 81} 82 83/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 84/// current line until the tok::eom token is found. 85void Preprocessor::DiscardUntilEndOfDirective() { 86 Token Tmp; 87 do { 88 LexUnexpandedToken(Tmp); 89 } while (Tmp.isNot(tok::eom)); 90} 91 92/// ReadMacroName - Lex and validate a macro name, which occurs after a 93/// #define or #undef. This sets the token kind to eom and discards the rest 94/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 95/// this is due to a a #define, 2 if #undef directive, 0 if it is something 96/// else (e.g. #ifdef). 97void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 98 // Read the token, don't allow macro expansion on it. 99 LexUnexpandedToken(MacroNameTok); 100 101 if (MacroNameTok.is(tok::code_completion)) { 102 if (CodeComplete) 103 CodeComplete->CodeCompleteMacroName(isDefineUndef == 1); 104 LexUnexpandedToken(MacroNameTok); 105 return; 106 } 107 108 // Missing macro name? 109 if (MacroNameTok.is(tok::eom)) { 110 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 111 return; 112 } 113 114 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 115 if (II == 0) { 116 bool Invalid = false; 117 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 118 if (Invalid) 119 return; 120 121 const IdentifierInfo &Info = Identifiers.get(Spelling); 122 if (Info.isCPlusPlusOperatorKeyword()) 123 // C++ 2.5p2: Alternative tokens behave the same as its primary token 124 // except for their spellings. 125 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 126 else 127 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 128 // Fall through on error. 129 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 130 // Error if defining "defined": C99 6.10.8.4. 131 Diag(MacroNameTok, diag::err_defined_macro_name); 132 } else if (isDefineUndef && II->hasMacroDefinition() && 133 getMacroInfo(II)->isBuiltinMacro()) { 134 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 135 if (isDefineUndef == 1) 136 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 137 else 138 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 139 } else { 140 // Okay, we got a good identifier node. Return it. 141 return; 142 } 143 144 // Invalid macro name, read and discard the rest of the line. Then set the 145 // token kind to tok::eom. 146 MacroNameTok.setKind(tok::eom); 147 return DiscardUntilEndOfDirective(); 148} 149 150/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 151/// not, emit a diagnostic and consume up until the eom. If EnableMacros is 152/// true, then we consider macros that expand to zero tokens as being ok. 153void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 154 Token Tmp; 155 // Lex unexpanded tokens for most directives: macros might expand to zero 156 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 157 // #line) allow empty macros. 158 if (EnableMacros) 159 Lex(Tmp); 160 else 161 LexUnexpandedToken(Tmp); 162 163 // There should be no tokens after the directive, but we allow them as an 164 // extension. 165 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 166 LexUnexpandedToken(Tmp); 167 168 if (Tmp.isNot(tok::eom)) { 169 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 170 // because it is more trouble than it is worth to insert /**/ and check that 171 // there is no /**/ in the range also. 172 FixItHint Hint; 173 if (Features.GNUMode || Features.C99 || Features.CPlusPlus) 174 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 175 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 176 DiscardUntilEndOfDirective(); 177 } 178} 179 180 181 182/// SkipExcludedConditionalBlock - We just read a #if or related directive and 183/// decided that the subsequent tokens are in the #if'd out portion of the 184/// file. Lex the rest of the file, until we see an #endif. If 185/// FoundNonSkipPortion is true, then we have already emitted code for part of 186/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 187/// is true, then #else directives are ok, if not, then we have already seen one 188/// so a #else directive is a duplicate. When this returns, the caller can lex 189/// the first valid token. 190void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 191 bool FoundNonSkipPortion, 192 bool FoundElse) { 193 ++NumSkipped; 194 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 195 196 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 197 FoundNonSkipPortion, FoundElse); 198 199 if (CurPTHLexer) { 200 PTHSkipExcludedConditionalBlock(); 201 return; 202 } 203 204 // Enter raw mode to disable identifier lookup (and thus macro expansion), 205 // disabling warnings, etc. 206 CurPPLexer->LexingRawMode = true; 207 Token Tok; 208 while (1) { 209 CurLexer->Lex(Tok); 210 211 if (Tok.is(tok::code_completion)) { 212 if (CodeComplete) 213 CodeComplete->CodeCompleteInConditionalExclusion(); 214 continue; 215 } 216 217 // If this is the end of the buffer, we have an error. 218 if (Tok.is(tok::eof)) { 219 // Emit errors for each unterminated conditional on the stack, including 220 // the current one. 221 while (!CurPPLexer->ConditionalStack.empty()) { 222 if (!isCodeCompletionFile(Tok.getLocation())) 223 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 224 diag::err_pp_unterminated_conditional); 225 CurPPLexer->ConditionalStack.pop_back(); 226 } 227 228 // Just return and let the caller lex after this #include. 229 break; 230 } 231 232 // If this token is not a preprocessor directive, just skip it. 233 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 234 continue; 235 236 // We just parsed a # character at the start of a line, so we're in 237 // directive mode. Tell the lexer this so any newlines we see will be 238 // converted into an EOM token (this terminates the macro). 239 CurPPLexer->ParsingPreprocessorDirective = true; 240 if (CurLexer) CurLexer->SetCommentRetentionState(false); 241 242 243 // Read the next token, the directive flavor. 244 LexUnexpandedToken(Tok); 245 246 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 247 // something bogus), skip it. 248 if (Tok.isNot(tok::raw_identifier)) { 249 CurPPLexer->ParsingPreprocessorDirective = false; 250 // Restore comment saving mode. 251 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 252 continue; 253 } 254 255 // If the first letter isn't i or e, it isn't intesting to us. We know that 256 // this is safe in the face of spelling differences, because there is no way 257 // to spell an i/e in a strange way that is another letter. Skipping this 258 // allows us to avoid looking up the identifier info for #define/#undef and 259 // other common directives. 260 const char *RawCharData = Tok.getRawIdentifierData(); 261 262 char FirstChar = RawCharData[0]; 263 if (FirstChar >= 'a' && FirstChar <= 'z' && 264 FirstChar != 'i' && FirstChar != 'e') { 265 CurPPLexer->ParsingPreprocessorDirective = false; 266 // Restore comment saving mode. 267 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 268 continue; 269 } 270 271 // Get the identifier name without trigraphs or embedded newlines. Note 272 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 273 // when skipping. 274 char DirectiveBuf[20]; 275 llvm::StringRef Directive; 276 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 277 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 278 } else { 279 std::string DirectiveStr = getSpelling(Tok); 280 unsigned IdLen = DirectiveStr.size(); 281 if (IdLen >= 20) { 282 CurPPLexer->ParsingPreprocessorDirective = false; 283 // Restore comment saving mode. 284 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 285 continue; 286 } 287 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 288 Directive = llvm::StringRef(DirectiveBuf, IdLen); 289 } 290 291 if (Directive.startswith("if")) { 292 llvm::StringRef Sub = Directive.substr(2); 293 if (Sub.empty() || // "if" 294 Sub == "def" || // "ifdef" 295 Sub == "ndef") { // "ifndef" 296 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 297 // bother parsing the condition. 298 DiscardUntilEndOfDirective(); 299 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 300 /*foundnonskip*/false, 301 /*fnddelse*/false); 302 } 303 } else if (Directive[0] == 'e') { 304 llvm::StringRef Sub = Directive.substr(1); 305 if (Sub == "ndif") { // "endif" 306 CheckEndOfDirective("endif"); 307 PPConditionalInfo CondInfo; 308 CondInfo.WasSkipping = true; // Silence bogus warning. 309 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 310 InCond = InCond; // Silence warning in no-asserts mode. 311 assert(!InCond && "Can't be skipping if not in a conditional!"); 312 313 // If we popped the outermost skipping block, we're done skipping! 314 if (!CondInfo.WasSkipping) 315 break; 316 } else if (Sub == "lse") { // "else". 317 // #else directive in a skipping conditional. If not in some other 318 // skipping conditional, and if #else hasn't already been seen, enter it 319 // as a non-skipping conditional. 320 DiscardUntilEndOfDirective(); // C99 6.10p4. 321 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 322 323 // If this is a #else with a #else before it, report the error. 324 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 325 326 // Note that we've seen a #else in this conditional. 327 CondInfo.FoundElse = true; 328 329 // If the conditional is at the top level, and the #if block wasn't 330 // entered, enter the #else block now. 331 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 332 CondInfo.FoundNonSkip = true; 333 break; 334 } 335 } else if (Sub == "lif") { // "elif". 336 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 337 338 bool ShouldEnter; 339 // If this is in a skipping block or if we're already handled this #if 340 // block, don't bother parsing the condition. 341 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 342 DiscardUntilEndOfDirective(); 343 ShouldEnter = false; 344 } else { 345 // Restore the value of LexingRawMode so that identifiers are 346 // looked up, etc, inside the #elif expression. 347 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 348 CurPPLexer->LexingRawMode = false; 349 IdentifierInfo *IfNDefMacro = 0; 350 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 351 CurPPLexer->LexingRawMode = true; 352 } 353 354 // If this is a #elif with a #else before it, report the error. 355 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 356 357 // If this condition is true, enter it! 358 if (ShouldEnter) { 359 CondInfo.FoundNonSkip = true; 360 break; 361 } 362 } 363 } 364 365 CurPPLexer->ParsingPreprocessorDirective = false; 366 // Restore comment saving mode. 367 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 368 } 369 370 // Finally, if we are out of the conditional (saw an #endif or ran off the end 371 // of the file, just stop skipping and return to lexing whatever came after 372 // the #if block. 373 CurPPLexer->LexingRawMode = false; 374} 375 376void Preprocessor::PTHSkipExcludedConditionalBlock() { 377 378 while (1) { 379 assert(CurPTHLexer); 380 assert(CurPTHLexer->LexingRawMode == false); 381 382 // Skip to the next '#else', '#elif', or #endif. 383 if (CurPTHLexer->SkipBlock()) { 384 // We have reached an #endif. Both the '#' and 'endif' tokens 385 // have been consumed by the PTHLexer. Just pop off the condition level. 386 PPConditionalInfo CondInfo; 387 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 388 InCond = InCond; // Silence warning in no-asserts mode. 389 assert(!InCond && "Can't be skipping if not in a conditional!"); 390 break; 391 } 392 393 // We have reached a '#else' or '#elif'. Lex the next token to get 394 // the directive flavor. 395 Token Tok; 396 LexUnexpandedToken(Tok); 397 398 // We can actually look up the IdentifierInfo here since we aren't in 399 // raw mode. 400 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 401 402 if (K == tok::pp_else) { 403 // #else: Enter the else condition. We aren't in a nested condition 404 // since we skip those. We're always in the one matching the last 405 // blocked we skipped. 406 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 407 // Note that we've seen a #else in this conditional. 408 CondInfo.FoundElse = true; 409 410 // If the #if block wasn't entered then enter the #else block now. 411 if (!CondInfo.FoundNonSkip) { 412 CondInfo.FoundNonSkip = true; 413 414 // Scan until the eom token. 415 CurPTHLexer->ParsingPreprocessorDirective = true; 416 DiscardUntilEndOfDirective(); 417 CurPTHLexer->ParsingPreprocessorDirective = false; 418 419 break; 420 } 421 422 // Otherwise skip this block. 423 continue; 424 } 425 426 assert(K == tok::pp_elif); 427 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 428 429 // If this is a #elif with a #else before it, report the error. 430 if (CondInfo.FoundElse) 431 Diag(Tok, diag::pp_err_elif_after_else); 432 433 // If this is in a skipping block or if we're already handled this #if 434 // block, don't bother parsing the condition. We just skip this block. 435 if (CondInfo.FoundNonSkip) 436 continue; 437 438 // Evaluate the condition of the #elif. 439 IdentifierInfo *IfNDefMacro = 0; 440 CurPTHLexer->ParsingPreprocessorDirective = true; 441 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 442 CurPTHLexer->ParsingPreprocessorDirective = false; 443 444 // If this condition is true, enter it! 445 if (ShouldEnter) { 446 CondInfo.FoundNonSkip = true; 447 break; 448 } 449 450 // Otherwise, skip this block and go to the next one. 451 continue; 452 } 453} 454 455/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 456/// return null on failure. isAngled indicates whether the file reference is 457/// for system #include's or not (i.e. using <> instead of ""). 458const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 459 bool isAngled, 460 const DirectoryLookup *FromDir, 461 const DirectoryLookup *&CurDir) { 462 // If the header lookup mechanism may be relative to the current file, pass in 463 // info about where the current file is. 464 const FileEntry *CurFileEnt = 0; 465 if (!FromDir) { 466 FileID FID = getCurrentFileLexer()->getFileID(); 467 CurFileEnt = SourceMgr.getFileEntryForID(FID); 468 469 // If there is no file entry associated with this file, it must be the 470 // predefines buffer. Any other file is not lexed with a normal lexer, so 471 // it won't be scanned for preprocessor directives. If we have the 472 // predefines buffer, resolve #include references (which come from the 473 // -include command line argument) as if they came from the main file, this 474 // affects file lookup etc. 475 if (CurFileEnt == 0) { 476 FID = SourceMgr.getMainFileID(); 477 CurFileEnt = SourceMgr.getFileEntryForID(FID); 478 } 479 } 480 481 // Do a standard file entry lookup. 482 CurDir = CurDirLookup; 483 const FileEntry *FE = 484 HeaderInfo.LookupFile(Filename, isAngled, FromDir, CurDir, CurFileEnt); 485 if (FE) return FE; 486 487 // Otherwise, see if this is a subframework header. If so, this is relative 488 // to one of the headers on the #include stack. Walk the list of the current 489 // headers on the #include stack and pass them to HeaderInfo. 490 if (IsFileLexer()) { 491 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 492 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 493 return FE; 494 } 495 496 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 497 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 498 if (IsFileLexer(ISEntry)) { 499 if ((CurFileEnt = 500 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 501 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 502 return FE; 503 } 504 } 505 506 // Otherwise, we really couldn't find the file. 507 return 0; 508} 509 510 511//===----------------------------------------------------------------------===// 512// Preprocessor Directive Handling. 513//===----------------------------------------------------------------------===// 514 515/// HandleDirective - This callback is invoked when the lexer sees a # token 516/// at the start of a line. This consumes the directive, modifies the 517/// lexer/preprocessor state, and advances the lexer(s) so that the next token 518/// read is the correct one. 519void Preprocessor::HandleDirective(Token &Result) { 520 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 521 522 // We just parsed a # character at the start of a line, so we're in directive 523 // mode. Tell the lexer this so any newlines we see will be converted into an 524 // EOM token (which terminates the directive). 525 CurPPLexer->ParsingPreprocessorDirective = true; 526 527 ++NumDirectives; 528 529 // We are about to read a token. For the multiple-include optimization FA to 530 // work, we have to remember if we had read any tokens *before* this 531 // pp-directive. 532 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 533 534 // Save the '#' token in case we need to return it later. 535 Token SavedHash = Result; 536 537 // Read the next token, the directive flavor. This isn't expanded due to 538 // C99 6.10.3p8. 539 LexUnexpandedToken(Result); 540 541 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 542 // #define A(x) #x 543 // A(abc 544 // #warning blah 545 // def) 546 // If so, the user is relying on non-portable behavior, emit a diagnostic. 547 if (InMacroArgs) 548 Diag(Result, diag::ext_embedded_directive); 549 550TryAgain: 551 switch (Result.getKind()) { 552 case tok::eom: 553 return; // null directive. 554 case tok::comment: 555 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 556 LexUnexpandedToken(Result); 557 goto TryAgain; 558 case tok::code_completion: 559 if (CodeComplete) 560 CodeComplete->CodeCompleteDirective( 561 CurPPLexer->getConditionalStackDepth() > 0); 562 return; 563 case tok::numeric_constant: // # 7 GNU line marker directive. 564 if (getLangOptions().AsmPreprocessor) 565 break; // # 4 is not a preprocessor directive in .S files. 566 return HandleDigitDirective(Result); 567 default: 568 IdentifierInfo *II = Result.getIdentifierInfo(); 569 if (II == 0) break; // Not an identifier. 570 571 // Ask what the preprocessor keyword ID is. 572 switch (II->getPPKeywordID()) { 573 default: break; 574 // C99 6.10.1 - Conditional Inclusion. 575 case tok::pp_if: 576 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 577 case tok::pp_ifdef: 578 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 579 case tok::pp_ifndef: 580 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 581 case tok::pp_elif: 582 return HandleElifDirective(Result); 583 case tok::pp_else: 584 return HandleElseDirective(Result); 585 case tok::pp_endif: 586 return HandleEndifDirective(Result); 587 588 // C99 6.10.2 - Source File Inclusion. 589 case tok::pp_include: 590 // Handle #include. 591 return HandleIncludeDirective(SavedHash.getLocation(), Result); 592 case tok::pp___include_macros: 593 // Handle -imacros. 594 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result); 595 596 // C99 6.10.3 - Macro Replacement. 597 case tok::pp_define: 598 return HandleDefineDirective(Result); 599 case tok::pp_undef: 600 return HandleUndefDirective(Result); 601 602 // C99 6.10.4 - Line Control. 603 case tok::pp_line: 604 return HandleLineDirective(Result); 605 606 // C99 6.10.5 - Error Directive. 607 case tok::pp_error: 608 return HandleUserDiagnosticDirective(Result, false); 609 610 // C99 6.10.6 - Pragma Directive. 611 case tok::pp_pragma: 612 return HandlePragmaDirective(PIK_HashPragma); 613 614 // GNU Extensions. 615 case tok::pp_import: 616 return HandleImportDirective(SavedHash.getLocation(), Result); 617 case tok::pp_include_next: 618 return HandleIncludeNextDirective(SavedHash.getLocation(), Result); 619 620 case tok::pp_warning: 621 Diag(Result, diag::ext_pp_warning_directive); 622 return HandleUserDiagnosticDirective(Result, true); 623 case tok::pp_ident: 624 return HandleIdentSCCSDirective(Result); 625 case tok::pp_sccs: 626 return HandleIdentSCCSDirective(Result); 627 case tok::pp_assert: 628 //isExtension = true; // FIXME: implement #assert 629 break; 630 case tok::pp_unassert: 631 //isExtension = true; // FIXME: implement #unassert 632 break; 633 } 634 break; 635 } 636 637 // If this is a .S file, treat unknown # directives as non-preprocessor 638 // directives. This is important because # may be a comment or introduce 639 // various pseudo-ops. Just return the # token and push back the following 640 // token to be lexed next time. 641 if (getLangOptions().AsmPreprocessor) { 642 Token *Toks = new Token[2]; 643 // Return the # and the token after it. 644 Toks[0] = SavedHash; 645 Toks[1] = Result; 646 // Enter this token stream so that we re-lex the tokens. Make sure to 647 // enable macro expansion, in case the token after the # is an identifier 648 // that is expanded. 649 EnterTokenStream(Toks, 2, false, true); 650 return; 651 } 652 653 // If we reached here, the preprocessing token is not valid! 654 Diag(Result, diag::err_pp_invalid_directive); 655 656 // Read the rest of the PP line. 657 DiscardUntilEndOfDirective(); 658 659 // Okay, we're done parsing the directive. 660} 661 662/// GetLineValue - Convert a numeric token into an unsigned value, emitting 663/// Diagnostic DiagID if it is invalid, and returning the value in Val. 664static bool GetLineValue(Token &DigitTok, unsigned &Val, 665 unsigned DiagID, Preprocessor &PP) { 666 if (DigitTok.isNot(tok::numeric_constant)) { 667 PP.Diag(DigitTok, DiagID); 668 669 if (DigitTok.isNot(tok::eom)) 670 PP.DiscardUntilEndOfDirective(); 671 return true; 672 } 673 674 llvm::SmallString<64> IntegerBuffer; 675 IntegerBuffer.resize(DigitTok.getLength()); 676 const char *DigitTokBegin = &IntegerBuffer[0]; 677 bool Invalid = false; 678 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 679 if (Invalid) 680 return true; 681 682 // Verify that we have a simple digit-sequence, and compute the value. This 683 // is always a simple digit string computed in decimal, so we do this manually 684 // here. 685 Val = 0; 686 for (unsigned i = 0; i != ActualLength; ++i) { 687 if (!isdigit(DigitTokBegin[i])) { 688 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 689 diag::err_pp_line_digit_sequence); 690 PP.DiscardUntilEndOfDirective(); 691 return true; 692 } 693 694 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 695 if (NextVal < Val) { // overflow. 696 PP.Diag(DigitTok, DiagID); 697 PP.DiscardUntilEndOfDirective(); 698 return true; 699 } 700 Val = NextVal; 701 } 702 703 // Reject 0, this is needed both by #line numbers and flags. 704 if (Val == 0) { 705 PP.Diag(DigitTok, DiagID); 706 PP.DiscardUntilEndOfDirective(); 707 return true; 708 } 709 710 if (DigitTokBegin[0] == '0') 711 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 712 713 return false; 714} 715 716/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 717/// acceptable forms are: 718/// # line digit-sequence 719/// # line digit-sequence "s-char-sequence" 720void Preprocessor::HandleLineDirective(Token &Tok) { 721 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 722 // expanded. 723 Token DigitTok; 724 Lex(DigitTok); 725 726 // Validate the number and convert it to an unsigned. 727 unsigned LineNo; 728 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 729 return; 730 731 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 732 // number greater than 2147483647". C90 requires that the line # be <= 32767. 733 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 734 if (LineNo >= LineLimit) 735 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 736 737 int FilenameID = -1; 738 Token StrTok; 739 Lex(StrTok); 740 741 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 742 // string followed by eom. 743 if (StrTok.is(tok::eom)) 744 ; // ok 745 else if (StrTok.isNot(tok::string_literal)) { 746 Diag(StrTok, diag::err_pp_line_invalid_filename); 747 DiscardUntilEndOfDirective(); 748 return; 749 } else { 750 // Parse and validate the string, converting it into a unique ID. 751 StringLiteralParser Literal(&StrTok, 1, *this); 752 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 753 if (Literal.hadError) 754 return DiscardUntilEndOfDirective(); 755 if (Literal.Pascal) { 756 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 757 return DiscardUntilEndOfDirective(); 758 } 759 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 760 Literal.GetStringLength()); 761 762 // Verify that there is nothing after the string, other than EOM. Because 763 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 764 CheckEndOfDirective("line", true); 765 } 766 767 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 768 769 if (Callbacks) 770 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 771 PPCallbacks::RenameFile, 772 SrcMgr::C_User); 773} 774 775/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 776/// marker directive. 777static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 778 bool &IsSystemHeader, bool &IsExternCHeader, 779 Preprocessor &PP) { 780 unsigned FlagVal; 781 Token FlagTok; 782 PP.Lex(FlagTok); 783 if (FlagTok.is(tok::eom)) return false; 784 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 785 return true; 786 787 if (FlagVal == 1) { 788 IsFileEntry = true; 789 790 PP.Lex(FlagTok); 791 if (FlagTok.is(tok::eom)) return false; 792 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 793 return true; 794 } else if (FlagVal == 2) { 795 IsFileExit = true; 796 797 SourceManager &SM = PP.getSourceManager(); 798 // If we are leaving the current presumed file, check to make sure the 799 // presumed include stack isn't empty! 800 FileID CurFileID = 801 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 802 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 803 if (PLoc.isInvalid()) 804 return true; 805 806 // If there is no include loc (main file) or if the include loc is in a 807 // different physical file, then we aren't in a "1" line marker flag region. 808 SourceLocation IncLoc = PLoc.getIncludeLoc(); 809 if (IncLoc.isInvalid() || 810 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 811 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 812 PP.DiscardUntilEndOfDirective(); 813 return true; 814 } 815 816 PP.Lex(FlagTok); 817 if (FlagTok.is(tok::eom)) return false; 818 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 819 return true; 820 } 821 822 // We must have 3 if there are still flags. 823 if (FlagVal != 3) { 824 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 825 PP.DiscardUntilEndOfDirective(); 826 return true; 827 } 828 829 IsSystemHeader = true; 830 831 PP.Lex(FlagTok); 832 if (FlagTok.is(tok::eom)) return false; 833 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 834 return true; 835 836 // We must have 4 if there is yet another flag. 837 if (FlagVal != 4) { 838 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 839 PP.DiscardUntilEndOfDirective(); 840 return true; 841 } 842 843 IsExternCHeader = true; 844 845 PP.Lex(FlagTok); 846 if (FlagTok.is(tok::eom)) return false; 847 848 // There are no more valid flags here. 849 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 850 PP.DiscardUntilEndOfDirective(); 851 return true; 852} 853 854/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 855/// one of the following forms: 856/// 857/// # 42 858/// # 42 "file" ('1' | '2')? 859/// # 42 "file" ('1' | '2')? '3' '4'? 860/// 861void Preprocessor::HandleDigitDirective(Token &DigitTok) { 862 // Validate the number and convert it to an unsigned. GNU does not have a 863 // line # limit other than it fit in 32-bits. 864 unsigned LineNo; 865 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 866 *this)) 867 return; 868 869 Token StrTok; 870 Lex(StrTok); 871 872 bool IsFileEntry = false, IsFileExit = false; 873 bool IsSystemHeader = false, IsExternCHeader = false; 874 int FilenameID = -1; 875 876 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 877 // string followed by eom. 878 if (StrTok.is(tok::eom)) 879 ; // ok 880 else if (StrTok.isNot(tok::string_literal)) { 881 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 882 return DiscardUntilEndOfDirective(); 883 } else { 884 // Parse and validate the string, converting it into a unique ID. 885 StringLiteralParser Literal(&StrTok, 1, *this); 886 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 887 if (Literal.hadError) 888 return DiscardUntilEndOfDirective(); 889 if (Literal.Pascal) { 890 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 891 return DiscardUntilEndOfDirective(); 892 } 893 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 894 Literal.GetStringLength()); 895 896 // If a filename was present, read any flags that are present. 897 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 898 IsSystemHeader, IsExternCHeader, *this)) 899 return; 900 } 901 902 // Create a line note with this information. 903 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 904 IsFileEntry, IsFileExit, 905 IsSystemHeader, IsExternCHeader); 906 907 // If the preprocessor has callbacks installed, notify them of the #line 908 // change. This is used so that the line marker comes out in -E mode for 909 // example. 910 if (Callbacks) { 911 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 912 if (IsFileEntry) 913 Reason = PPCallbacks::EnterFile; 914 else if (IsFileExit) 915 Reason = PPCallbacks::ExitFile; 916 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 917 if (IsExternCHeader) 918 FileKind = SrcMgr::C_ExternCSystem; 919 else if (IsSystemHeader) 920 FileKind = SrcMgr::C_System; 921 922 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 923 } 924} 925 926 927/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 928/// 929void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 930 bool isWarning) { 931 // PTH doesn't emit #warning or #error directives. 932 if (CurPTHLexer) 933 return CurPTHLexer->DiscardToEndOfLine(); 934 935 // Read the rest of the line raw. We do this because we don't want macros 936 // to be expanded and we don't require that the tokens be valid preprocessing 937 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 938 // collapse multiple consequtive white space between tokens, but this isn't 939 // specified by the standard. 940 std::string Message = CurLexer->ReadToEndOfLine(); 941 if (isWarning) 942 Diag(Tok, diag::pp_hash_warning) << Message; 943 else 944 Diag(Tok, diag::err_pp_hash_error) << Message; 945} 946 947/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 948/// 949void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 950 // Yes, this directive is an extension. 951 Diag(Tok, diag::ext_pp_ident_directive); 952 953 // Read the string argument. 954 Token StrTok; 955 Lex(StrTok); 956 957 // If the token kind isn't a string, it's a malformed directive. 958 if (StrTok.isNot(tok::string_literal) && 959 StrTok.isNot(tok::wide_string_literal)) { 960 Diag(StrTok, diag::err_pp_malformed_ident); 961 if (StrTok.isNot(tok::eom)) 962 DiscardUntilEndOfDirective(); 963 return; 964 } 965 966 // Verify that there is nothing after the string, other than EOM. 967 CheckEndOfDirective("ident"); 968 969 if (Callbacks) { 970 bool Invalid = false; 971 std::string Str = getSpelling(StrTok, &Invalid); 972 if (!Invalid) 973 Callbacks->Ident(Tok.getLocation(), Str); 974 } 975} 976 977//===----------------------------------------------------------------------===// 978// Preprocessor Include Directive Handling. 979//===----------------------------------------------------------------------===// 980 981/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 982/// checked and spelled filename, e.g. as an operand of #include. This returns 983/// true if the input filename was in <>'s or false if it were in ""'s. The 984/// caller is expected to provide a buffer that is large enough to hold the 985/// spelling of the filename, but is also expected to handle the case when 986/// this method decides to use a different buffer. 987bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 988 llvm::StringRef &Buffer) { 989 // Get the text form of the filename. 990 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 991 992 // Make sure the filename is <x> or "x". 993 bool isAngled; 994 if (Buffer[0] == '<') { 995 if (Buffer.back() != '>') { 996 Diag(Loc, diag::err_pp_expects_filename); 997 Buffer = llvm::StringRef(); 998 return true; 999 } 1000 isAngled = true; 1001 } else if (Buffer[0] == '"') { 1002 if (Buffer.back() != '"') { 1003 Diag(Loc, diag::err_pp_expects_filename); 1004 Buffer = llvm::StringRef(); 1005 return true; 1006 } 1007 isAngled = false; 1008 } else { 1009 Diag(Loc, diag::err_pp_expects_filename); 1010 Buffer = llvm::StringRef(); 1011 return true; 1012 } 1013 1014 // Diagnose #include "" as invalid. 1015 if (Buffer.size() <= 2) { 1016 Diag(Loc, diag::err_pp_empty_filename); 1017 Buffer = llvm::StringRef(); 1018 return true; 1019 } 1020 1021 // Skip the brackets. 1022 Buffer = Buffer.substr(1, Buffer.size()-2); 1023 return isAngled; 1024} 1025 1026/// ConcatenateIncludeName - Handle cases where the #include name is expanded 1027/// from a macro as multiple tokens, which need to be glued together. This 1028/// occurs for code like: 1029/// #define FOO <a/b.h> 1030/// #include FOO 1031/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1032/// 1033/// This code concatenates and consumes tokens up to the '>' token. It returns 1034/// false if the > was found, otherwise it returns true if it finds and consumes 1035/// the EOM marker. 1036bool Preprocessor::ConcatenateIncludeName( 1037 llvm::SmallString<128> &FilenameBuffer, 1038 SourceLocation &End) { 1039 Token CurTok; 1040 1041 Lex(CurTok); 1042 while (CurTok.isNot(tok::eom)) { 1043 End = CurTok.getLocation(); 1044 1045 // FIXME: Provide code completion for #includes. 1046 if (CurTok.is(tok::code_completion)) { 1047 Lex(CurTok); 1048 continue; 1049 } 1050 1051 // Append the spelling of this token to the buffer. If there was a space 1052 // before it, add it now. 1053 if (CurTok.hasLeadingSpace()) 1054 FilenameBuffer.push_back(' '); 1055 1056 // Get the spelling of the token, directly into FilenameBuffer if possible. 1057 unsigned PreAppendSize = FilenameBuffer.size(); 1058 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1059 1060 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1061 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1062 1063 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1064 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1065 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1066 1067 // Resize FilenameBuffer to the correct size. 1068 if (CurTok.getLength() != ActualLen) 1069 FilenameBuffer.resize(PreAppendSize+ActualLen); 1070 1071 // If we found the '>' marker, return success. 1072 if (CurTok.is(tok::greater)) 1073 return false; 1074 1075 Lex(CurTok); 1076 } 1077 1078 // If we hit the eom marker, emit an error and return true so that the caller 1079 // knows the EOM has been read. 1080 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1081 return true; 1082} 1083 1084/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1085/// file to be included from the lexer, then include it! This is a common 1086/// routine with functionality shared between #include, #include_next and 1087/// #import. LookupFrom is set when this is a #include_next directive, it 1088/// specifies the file to start searching from. 1089void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc, 1090 Token &IncludeTok, 1091 const DirectoryLookup *LookupFrom, 1092 bool isImport) { 1093 1094 Token FilenameTok; 1095 CurPPLexer->LexIncludeFilename(FilenameTok); 1096 1097 // Reserve a buffer to get the spelling. 1098 llvm::SmallString<128> FilenameBuffer; 1099 llvm::StringRef Filename; 1100 SourceLocation End; 1101 1102 switch (FilenameTok.getKind()) { 1103 case tok::eom: 1104 // If the token kind is EOM, the error has already been diagnosed. 1105 return; 1106 1107 case tok::angle_string_literal: 1108 case tok::string_literal: 1109 Filename = getSpelling(FilenameTok, FilenameBuffer); 1110 End = FilenameTok.getLocation(); 1111 break; 1112 1113 case tok::less: 1114 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1115 // case, glue the tokens together into FilenameBuffer and interpret those. 1116 FilenameBuffer.push_back('<'); 1117 if (ConcatenateIncludeName(FilenameBuffer, End)) 1118 return; // Found <eom> but no ">"? Diagnostic already emitted. 1119 Filename = FilenameBuffer.str(); 1120 break; 1121 default: 1122 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1123 DiscardUntilEndOfDirective(); 1124 return; 1125 } 1126 1127 bool isAngled = 1128 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1129 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1130 // error. 1131 if (Filename.empty()) { 1132 DiscardUntilEndOfDirective(); 1133 return; 1134 } 1135 1136 // Verify that there is nothing after the filename, other than EOM. Note that 1137 // we allow macros that expand to nothing after the filename, because this 1138 // falls into the category of "#include pp-tokens new-line" specified in 1139 // C99 6.10.2p4. 1140 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1141 1142 // Check that we don't have infinite #include recursion. 1143 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1144 Diag(FilenameTok, diag::err_pp_include_too_deep); 1145 return; 1146 } 1147 1148 // Search include directories. 1149 const DirectoryLookup *CurDir; 1150 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1151 if (File == 0) { 1152 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1153 return; 1154 } 1155 1156 // Notify the callback object that we've seen an inclusion directive. 1157 if (Callbacks) 1158 Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, File, 1159 End); 1160 1161 // The #included file will be considered to be a system header if either it is 1162 // in a system include directory, or if the #includer is a system include 1163 // header. 1164 SrcMgr::CharacteristicKind FileCharacter = 1165 std::max(HeaderInfo.getFileDirFlavor(File), 1166 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1167 1168 // Ask HeaderInfo if we should enter this #include file. If not, #including 1169 // this file will have no effect. 1170 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1171 if (Callbacks) 1172 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1173 return; 1174 } 1175 1176 // Look up the file, create a File ID for it. 1177 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1178 FileCharacter); 1179 if (FID.isInvalid()) { 1180 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1181 return; 1182 } 1183 1184 // Finally, if all is good, enter the new file! 1185 EnterSourceFile(FID, CurDir, FilenameTok.getLocation()); 1186} 1187 1188/// HandleIncludeNextDirective - Implements #include_next. 1189/// 1190void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc, 1191 Token &IncludeNextTok) { 1192 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1193 1194 // #include_next is like #include, except that we start searching after 1195 // the current found directory. If we can't do this, issue a 1196 // diagnostic. 1197 const DirectoryLookup *Lookup = CurDirLookup; 1198 if (isInPrimaryFile()) { 1199 Lookup = 0; 1200 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1201 } else if (Lookup == 0) { 1202 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1203 } else { 1204 // Start looking up in the next directory. 1205 ++Lookup; 1206 } 1207 1208 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup); 1209} 1210 1211/// HandleImportDirective - Implements #import. 1212/// 1213void Preprocessor::HandleImportDirective(SourceLocation HashLoc, 1214 Token &ImportTok) { 1215 if (!Features.ObjC1) // #import is standard for ObjC. 1216 Diag(ImportTok, diag::ext_pp_import_directive); 1217 1218 return HandleIncludeDirective(HashLoc, ImportTok, 0, true); 1219} 1220 1221/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1222/// pseudo directive in the predefines buffer. This handles it by sucking all 1223/// tokens through the preprocessor and discarding them (only keeping the side 1224/// effects on the preprocessor). 1225void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc, 1226 Token &IncludeMacrosTok) { 1227 // This directive should only occur in the predefines buffer. If not, emit an 1228 // error and reject it. 1229 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1230 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1231 Diag(IncludeMacrosTok.getLocation(), 1232 diag::pp_include_macros_out_of_predefines); 1233 DiscardUntilEndOfDirective(); 1234 return; 1235 } 1236 1237 // Treat this as a normal #include for checking purposes. If this is 1238 // successful, it will push a new lexer onto the include stack. 1239 HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false); 1240 1241 Token TmpTok; 1242 do { 1243 Lex(TmpTok); 1244 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1245 } while (TmpTok.isNot(tok::hashhash)); 1246} 1247 1248//===----------------------------------------------------------------------===// 1249// Preprocessor Macro Directive Handling. 1250//===----------------------------------------------------------------------===// 1251 1252/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1253/// definition has just been read. Lex the rest of the arguments and the 1254/// closing ), updating MI with what we learn. Return true if an error occurs 1255/// parsing the arg list. 1256bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1257 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1258 1259 Token Tok; 1260 while (1) { 1261 LexUnexpandedToken(Tok); 1262 switch (Tok.getKind()) { 1263 case tok::r_paren: 1264 // Found the end of the argument list. 1265 if (Arguments.empty()) // #define FOO() 1266 return false; 1267 // Otherwise we have #define FOO(A,) 1268 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1269 return true; 1270 case tok::ellipsis: // #define X(... -> C99 varargs 1271 // Warn if use of C99 feature in non-C99 mode. 1272 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1273 1274 // Lex the token after the identifier. 1275 LexUnexpandedToken(Tok); 1276 if (Tok.isNot(tok::r_paren)) { 1277 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1278 return true; 1279 } 1280 // Add the __VA_ARGS__ identifier as an argument. 1281 Arguments.push_back(Ident__VA_ARGS__); 1282 MI->setIsC99Varargs(); 1283 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1284 return false; 1285 case tok::eom: // #define X( 1286 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1287 return true; 1288 default: 1289 // Handle keywords and identifiers here to accept things like 1290 // #define Foo(for) for. 1291 IdentifierInfo *II = Tok.getIdentifierInfo(); 1292 if (II == 0) { 1293 // #define X(1 1294 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1295 return true; 1296 } 1297 1298 // If this is already used as an argument, it is used multiple times (e.g. 1299 // #define X(A,A. 1300 if (std::find(Arguments.begin(), Arguments.end(), II) != 1301 Arguments.end()) { // C99 6.10.3p6 1302 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1303 return true; 1304 } 1305 1306 // Add the argument to the macro info. 1307 Arguments.push_back(II); 1308 1309 // Lex the token after the identifier. 1310 LexUnexpandedToken(Tok); 1311 1312 switch (Tok.getKind()) { 1313 default: // #define X(A B 1314 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1315 return true; 1316 case tok::r_paren: // #define X(A) 1317 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1318 return false; 1319 case tok::comma: // #define X(A, 1320 break; 1321 case tok::ellipsis: // #define X(A... -> GCC extension 1322 // Diagnose extension. 1323 Diag(Tok, diag::ext_named_variadic_macro); 1324 1325 // Lex the token after the identifier. 1326 LexUnexpandedToken(Tok); 1327 if (Tok.isNot(tok::r_paren)) { 1328 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1329 return true; 1330 } 1331 1332 MI->setIsGNUVarargs(); 1333 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1334 return false; 1335 } 1336 } 1337 } 1338} 1339 1340/// HandleDefineDirective - Implements #define. This consumes the entire macro 1341/// line then lets the caller lex the next real token. 1342void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1343 ++NumDefined; 1344 1345 Token MacroNameTok; 1346 ReadMacroName(MacroNameTok, 1); 1347 1348 // Error reading macro name? If so, diagnostic already issued. 1349 if (MacroNameTok.is(tok::eom)) 1350 return; 1351 1352 Token LastTok = MacroNameTok; 1353 1354 // If we are supposed to keep comments in #defines, reenable comment saving 1355 // mode. 1356 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1357 1358 // Create the new macro. 1359 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1360 1361 Token Tok; 1362 LexUnexpandedToken(Tok); 1363 1364 // If this is a function-like macro definition, parse the argument list, 1365 // marking each of the identifiers as being used as macro arguments. Also, 1366 // check other constraints on the first token of the macro body. 1367 if (Tok.is(tok::eom)) { 1368 // If there is no body to this macro, we have no special handling here. 1369 } else if (Tok.hasLeadingSpace()) { 1370 // This is a normal token with leading space. Clear the leading space 1371 // marker on the first token to get proper expansion. 1372 Tok.clearFlag(Token::LeadingSpace); 1373 } else if (Tok.is(tok::l_paren)) { 1374 // This is a function-like macro definition. Read the argument list. 1375 MI->setIsFunctionLike(); 1376 if (ReadMacroDefinitionArgList(MI)) { 1377 // Forget about MI. 1378 ReleaseMacroInfo(MI); 1379 // Throw away the rest of the line. 1380 if (CurPPLexer->ParsingPreprocessorDirective) 1381 DiscardUntilEndOfDirective(); 1382 return; 1383 } 1384 1385 // If this is a definition of a variadic C99 function-like macro, not using 1386 // the GNU named varargs extension, enabled __VA_ARGS__. 1387 1388 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1389 // This gets unpoisoned where it is allowed. 1390 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1391 if (MI->isC99Varargs()) 1392 Ident__VA_ARGS__->setIsPoisoned(false); 1393 1394 // Read the first token after the arg list for down below. 1395 LexUnexpandedToken(Tok); 1396 } else if (Features.C99) { 1397 // C99 requires whitespace between the macro definition and the body. Emit 1398 // a diagnostic for something like "#define X+". 1399 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1400 } else { 1401 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1402 // first character of a replacement list is not a character required by 1403 // subclause 5.2.1, then there shall be white-space separation between the 1404 // identifier and the replacement list.". 5.2.1 lists this set: 1405 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1406 // is irrelevant here. 1407 bool isInvalid = false; 1408 if (Tok.is(tok::at)) // @ is not in the list above. 1409 isInvalid = true; 1410 else if (Tok.is(tok::unknown)) { 1411 // If we have an unknown token, it is something strange like "`". Since 1412 // all of valid characters would have lexed into a single character 1413 // token of some sort, we know this is not a valid case. 1414 isInvalid = true; 1415 } 1416 if (isInvalid) 1417 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1418 else 1419 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1420 } 1421 1422 if (!Tok.is(tok::eom)) 1423 LastTok = Tok; 1424 1425 // Read the rest of the macro body. 1426 if (MI->isObjectLike()) { 1427 // Object-like macros are very simple, just read their body. 1428 while (Tok.isNot(tok::eom)) { 1429 LastTok = Tok; 1430 MI->AddTokenToBody(Tok); 1431 // Get the next token of the macro. 1432 LexUnexpandedToken(Tok); 1433 } 1434 1435 } else { 1436 // Otherwise, read the body of a function-like macro. While we are at it, 1437 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1438 // parameters in function-like macro expansions. 1439 while (Tok.isNot(tok::eom)) { 1440 LastTok = Tok; 1441 1442 if (Tok.isNot(tok::hash)) { 1443 MI->AddTokenToBody(Tok); 1444 1445 // Get the next token of the macro. 1446 LexUnexpandedToken(Tok); 1447 continue; 1448 } 1449 1450 // Get the next token of the macro. 1451 LexUnexpandedToken(Tok); 1452 1453 // Check for a valid macro arg identifier. 1454 if (Tok.getIdentifierInfo() == 0 || 1455 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1456 1457 // If this is assembler-with-cpp mode, we accept random gibberish after 1458 // the '#' because '#' is often a comment character. However, change 1459 // the kind of the token to tok::unknown so that the preprocessor isn't 1460 // confused. 1461 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1462 LastTok.setKind(tok::unknown); 1463 } else { 1464 Diag(Tok, diag::err_pp_stringize_not_parameter); 1465 ReleaseMacroInfo(MI); 1466 1467 // Disable __VA_ARGS__ again. 1468 Ident__VA_ARGS__->setIsPoisoned(true); 1469 return; 1470 } 1471 } 1472 1473 // Things look ok, add the '#' and param name tokens to the macro. 1474 MI->AddTokenToBody(LastTok); 1475 MI->AddTokenToBody(Tok); 1476 LastTok = Tok; 1477 1478 // Get the next token of the macro. 1479 LexUnexpandedToken(Tok); 1480 } 1481 } 1482 1483 1484 // Disable __VA_ARGS__ again. 1485 Ident__VA_ARGS__->setIsPoisoned(true); 1486 1487 // Check that there is no paste (##) operator at the begining or end of the 1488 // replacement list. 1489 unsigned NumTokens = MI->getNumTokens(); 1490 if (NumTokens != 0) { 1491 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1492 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1493 ReleaseMacroInfo(MI); 1494 return; 1495 } 1496 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1497 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1498 ReleaseMacroInfo(MI); 1499 return; 1500 } 1501 } 1502 1503 MI->setDefinitionEndLoc(LastTok.getLocation()); 1504 1505 // Finally, if this identifier already had a macro defined for it, verify that 1506 // the macro bodies are identical and free the old definition. 1507 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1508 // It is very common for system headers to have tons of macro redefinitions 1509 // and for warnings to be disabled in system headers. If this is the case, 1510 // then don't bother calling MacroInfo::isIdenticalTo. 1511 if (!getDiagnostics().getSuppressSystemWarnings() || 1512 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1513 if (!OtherMI->isUsed()) 1514 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1515 1516 // Macros must be identical. This means all tokens and whitespace 1517 // separation must be the same. C99 6.10.3.2. 1518 if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 1519 !MI->isIdenticalTo(*OtherMI, *this)) { 1520 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1521 << MacroNameTok.getIdentifierInfo(); 1522 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1523 } 1524 } 1525 ReleaseMacroInfo(OtherMI); 1526 } 1527 1528 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1529 1530 assert(!MI->isUsed()); 1531 // If we need warning for not using the macro, add its location in the 1532 // warn-because-unused-macro set. If it gets used it will be removed from set. 1533 if (isInPrimaryFile() && // don't warn for include'd macros. 1534 Diags->getDiagnosticLevel(diag::pp_macro_not_used, 1535 MI->getDefinitionLoc()) != Diagnostic::Ignored) { 1536 MI->setIsWarnIfUnused(true); 1537 WarnUnusedMacroLocs.insert(MI->getDefinitionLoc()); 1538 } 1539 1540 // If the callbacks want to know, tell them about the macro definition. 1541 if (Callbacks) 1542 Callbacks->MacroDefined(MacroNameTok, MI); 1543} 1544 1545/// HandleUndefDirective - Implements #undef. 1546/// 1547void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1548 ++NumUndefined; 1549 1550 Token MacroNameTok; 1551 ReadMacroName(MacroNameTok, 2); 1552 1553 // Error reading macro name? If so, diagnostic already issued. 1554 if (MacroNameTok.is(tok::eom)) 1555 return; 1556 1557 // Check to see if this is the last token on the #undef line. 1558 CheckEndOfDirective("undef"); 1559 1560 // Okay, we finally have a valid identifier to undef. 1561 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1562 1563 // If the macro is not defined, this is a noop undef, just return. 1564 if (MI == 0) return; 1565 1566 if (!MI->isUsed()) 1567 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1568 1569 // If the callbacks want to know, tell them about the macro #undef. 1570 if (Callbacks) 1571 Callbacks->MacroUndefined(MacroNameTok, MI); 1572 1573 if (MI->isWarnIfUnused()) 1574 WarnUnusedMacroLocs.erase(MI->getDefinitionLoc()); 1575 1576 // Free macro definition. 1577 ReleaseMacroInfo(MI); 1578 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1579} 1580 1581 1582//===----------------------------------------------------------------------===// 1583// Preprocessor Conditional Directive Handling. 1584//===----------------------------------------------------------------------===// 1585 1586/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1587/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1588/// if any tokens have been returned or pp-directives activated before this 1589/// #ifndef has been lexed. 1590/// 1591void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1592 bool ReadAnyTokensBeforeDirective) { 1593 ++NumIf; 1594 Token DirectiveTok = Result; 1595 1596 Token MacroNameTok; 1597 ReadMacroName(MacroNameTok); 1598 1599 // Error reading macro name? If so, diagnostic already issued. 1600 if (MacroNameTok.is(tok::eom)) { 1601 // Skip code until we get to #endif. This helps with recovery by not 1602 // emitting an error when the #endif is reached. 1603 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1604 /*Foundnonskip*/false, /*FoundElse*/false); 1605 return; 1606 } 1607 1608 // Check to see if this is the last token on the #if[n]def line. 1609 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1610 1611 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1612 MacroInfo *MI = getMacroInfo(MII); 1613 1614 if (CurPPLexer->getConditionalStackDepth() == 0) { 1615 // If the start of a top-level #ifdef and if the macro is not defined, 1616 // inform MIOpt that this might be the start of a proper include guard. 1617 // Otherwise it is some other form of unknown conditional which we can't 1618 // handle. 1619 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1620 assert(isIfndef && "#ifdef shouldn't reach here"); 1621 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1622 } else 1623 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1624 } 1625 1626 // If there is a macro, process it. 1627 if (MI) // Mark it used. 1628 markMacroAsUsed(MI); 1629 1630 // Should we include the stuff contained by this directive? 1631 if (!MI == isIfndef) { 1632 // Yes, remember that we are inside a conditional, then lex the next token. 1633 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1634 /*wasskip*/false, /*foundnonskip*/true, 1635 /*foundelse*/false); 1636 } else { 1637 // No, skip the contents of this block. 1638 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1639 /*Foundnonskip*/false, 1640 /*FoundElse*/false); 1641 } 1642 1643 if (Callbacks) { 1644 if (isIfndef) 1645 Callbacks->Ifndef(MacroNameTok); 1646 else 1647 Callbacks->Ifdef(MacroNameTok); 1648 } 1649} 1650 1651/// HandleIfDirective - Implements the #if directive. 1652/// 1653void Preprocessor::HandleIfDirective(Token &IfToken, 1654 bool ReadAnyTokensBeforeDirective) { 1655 ++NumIf; 1656 1657 // Parse and evaluate the conditional expression. 1658 IdentifierInfo *IfNDefMacro = 0; 1659 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1660 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1661 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1662 1663 // If this condition is equivalent to #ifndef X, and if this is the first 1664 // directive seen, handle it for the multiple-include optimization. 1665 if (CurPPLexer->getConditionalStackDepth() == 0) { 1666 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1667 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1668 else 1669 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1670 } 1671 1672 // Should we include the stuff contained by this directive? 1673 if (ConditionalTrue) { 1674 // Yes, remember that we are inside a conditional, then lex the next token. 1675 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1676 /*foundnonskip*/true, /*foundelse*/false); 1677 } else { 1678 // No, skip the contents of this block. 1679 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1680 /*FoundElse*/false); 1681 } 1682 1683 if (Callbacks) 1684 Callbacks->If(SourceRange(ConditionalBegin, ConditionalEnd)); 1685} 1686 1687/// HandleEndifDirective - Implements the #endif directive. 1688/// 1689void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1690 ++NumEndif; 1691 1692 // Check that this is the whole directive. 1693 CheckEndOfDirective("endif"); 1694 1695 PPConditionalInfo CondInfo; 1696 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1697 // No conditionals on the stack: this is an #endif without an #if. 1698 Diag(EndifToken, diag::err_pp_endif_without_if); 1699 return; 1700 } 1701 1702 // If this the end of a top-level #endif, inform MIOpt. 1703 if (CurPPLexer->getConditionalStackDepth() == 0) 1704 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1705 1706 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1707 "This code should only be reachable in the non-skipping case!"); 1708 1709 if (Callbacks) 1710 Callbacks->Endif(); 1711} 1712 1713/// HandleElseDirective - Implements the #else directive. 1714/// 1715void Preprocessor::HandleElseDirective(Token &Result) { 1716 ++NumElse; 1717 1718 // #else directive in a non-skipping conditional... start skipping. 1719 CheckEndOfDirective("else"); 1720 1721 PPConditionalInfo CI; 1722 if (CurPPLexer->popConditionalLevel(CI)) { 1723 Diag(Result, diag::pp_err_else_without_if); 1724 return; 1725 } 1726 1727 // If this is a top-level #else, inform the MIOpt. 1728 if (CurPPLexer->getConditionalStackDepth() == 0) 1729 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1730 1731 // If this is a #else with a #else before it, report the error. 1732 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1733 1734 // Finally, skip the rest of the contents of this block. 1735 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1736 /*FoundElse*/true); 1737 1738 if (Callbacks) 1739 Callbacks->Else(); 1740} 1741 1742/// HandleElifDirective - Implements the #elif directive. 1743/// 1744void Preprocessor::HandleElifDirective(Token &ElifToken) { 1745 ++NumElse; 1746 1747 // #elif directive in a non-skipping conditional... start skipping. 1748 // We don't care what the condition is, because we will always skip it (since 1749 // the block immediately before it was included). 1750 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1751 DiscardUntilEndOfDirective(); 1752 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1753 1754 PPConditionalInfo CI; 1755 if (CurPPLexer->popConditionalLevel(CI)) { 1756 Diag(ElifToken, diag::pp_err_elif_without_if); 1757 return; 1758 } 1759 1760 // If this is a top-level #elif, inform the MIOpt. 1761 if (CurPPLexer->getConditionalStackDepth() == 0) 1762 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1763 1764 // If this is a #elif with a #else before it, report the error. 1765 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1766 1767 // Finally, skip the rest of the contents of this block. 1768 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1769 /*FoundElse*/CI.FoundElse); 1770 1771 if (Callbacks) 1772 Callbacks->Elif(SourceRange(ConditionalBegin, ConditionalEnd)); 1773} 1774