PPDirectives.cpp revision 25bb03bc0f2561e830d866d5905b20c5475d6762
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::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 bool Invalid = false; 261 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation(), 262 &Invalid); 263 if (Invalid) 264 return; 265 266 char FirstChar = RawCharData[0]; 267 if (FirstChar >= 'a' && FirstChar <= 'z' && 268 FirstChar != 'i' && FirstChar != 'e') { 269 CurPPLexer->ParsingPreprocessorDirective = false; 270 // Restore comment saving mode. 271 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 272 continue; 273 } 274 275 // Get the identifier name without trigraphs or embedded newlines. Note 276 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 277 // when skipping. 278 char DirectiveBuf[20]; 279 llvm::StringRef Directive; 280 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 281 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 282 } else { 283 std::string DirectiveStr = getSpelling(Tok); 284 unsigned IdLen = DirectiveStr.size(); 285 if (IdLen >= 20) { 286 CurPPLexer->ParsingPreprocessorDirective = false; 287 // Restore comment saving mode. 288 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 289 continue; 290 } 291 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 292 Directive = llvm::StringRef(DirectiveBuf, IdLen); 293 } 294 295 if (Directive.startswith("if")) { 296 llvm::StringRef Sub = Directive.substr(2); 297 if (Sub.empty() || // "if" 298 Sub == "def" || // "ifdef" 299 Sub == "ndef") { // "ifndef" 300 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 301 // bother parsing the condition. 302 DiscardUntilEndOfDirective(); 303 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 304 /*foundnonskip*/false, 305 /*fnddelse*/false); 306 } 307 } else if (Directive[0] == 'e') { 308 llvm::StringRef Sub = Directive.substr(1); 309 if (Sub == "ndif") { // "endif" 310 CheckEndOfDirective("endif"); 311 PPConditionalInfo CondInfo; 312 CondInfo.WasSkipping = true; // Silence bogus warning. 313 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 314 InCond = InCond; // Silence warning in no-asserts mode. 315 assert(!InCond && "Can't be skipping if not in a conditional!"); 316 317 // If we popped the outermost skipping block, we're done skipping! 318 if (!CondInfo.WasSkipping) 319 break; 320 } else if (Sub == "lse") { // "else". 321 // #else directive in a skipping conditional. If not in some other 322 // skipping conditional, and if #else hasn't already been seen, enter it 323 // as a non-skipping conditional. 324 DiscardUntilEndOfDirective(); // C99 6.10p4. 325 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 326 327 // If this is a #else with a #else before it, report the error. 328 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 329 330 // Note that we've seen a #else in this conditional. 331 CondInfo.FoundElse = true; 332 333 // If the conditional is at the top level, and the #if block wasn't 334 // entered, enter the #else block now. 335 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 336 CondInfo.FoundNonSkip = true; 337 break; 338 } 339 } else if (Sub == "lif") { // "elif". 340 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 341 342 bool ShouldEnter; 343 // If this is in a skipping block or if we're already handled this #if 344 // block, don't bother parsing the condition. 345 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 346 DiscardUntilEndOfDirective(); 347 ShouldEnter = false; 348 } else { 349 // Restore the value of LexingRawMode so that identifiers are 350 // looked up, etc, inside the #elif expression. 351 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 352 CurPPLexer->LexingRawMode = false; 353 IdentifierInfo *IfNDefMacro = 0; 354 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 355 CurPPLexer->LexingRawMode = true; 356 } 357 358 // If this is a #elif with a #else before it, report the error. 359 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 360 361 // If this condition is true, enter it! 362 if (ShouldEnter) { 363 CondInfo.FoundNonSkip = true; 364 break; 365 } 366 } 367 } 368 369 CurPPLexer->ParsingPreprocessorDirective = false; 370 // Restore comment saving mode. 371 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 372 } 373 374 // Finally, if we are out of the conditional (saw an #endif or ran off the end 375 // of the file, just stop skipping and return to lexing whatever came after 376 // the #if block. 377 CurPPLexer->LexingRawMode = false; 378} 379 380void Preprocessor::PTHSkipExcludedConditionalBlock() { 381 382 while (1) { 383 assert(CurPTHLexer); 384 assert(CurPTHLexer->LexingRawMode == false); 385 386 // Skip to the next '#else', '#elif', or #endif. 387 if (CurPTHLexer->SkipBlock()) { 388 // We have reached an #endif. Both the '#' and 'endif' tokens 389 // have been consumed by the PTHLexer. Just pop off the condition level. 390 PPConditionalInfo CondInfo; 391 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 392 InCond = InCond; // Silence warning in no-asserts mode. 393 assert(!InCond && "Can't be skipping if not in a conditional!"); 394 break; 395 } 396 397 // We have reached a '#else' or '#elif'. Lex the next token to get 398 // the directive flavor. 399 Token Tok; 400 LexUnexpandedToken(Tok); 401 402 // We can actually look up the IdentifierInfo here since we aren't in 403 // raw mode. 404 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 405 406 if (K == tok::pp_else) { 407 // #else: Enter the else condition. We aren't in a nested condition 408 // since we skip those. We're always in the one matching the last 409 // blocked we skipped. 410 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 411 // Note that we've seen a #else in this conditional. 412 CondInfo.FoundElse = true; 413 414 // If the #if block wasn't entered then enter the #else block now. 415 if (!CondInfo.FoundNonSkip) { 416 CondInfo.FoundNonSkip = true; 417 418 // Scan until the eom token. 419 CurPTHLexer->ParsingPreprocessorDirective = true; 420 DiscardUntilEndOfDirective(); 421 CurPTHLexer->ParsingPreprocessorDirective = false; 422 423 break; 424 } 425 426 // Otherwise skip this block. 427 continue; 428 } 429 430 assert(K == tok::pp_elif); 431 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 432 433 // If this is a #elif with a #else before it, report the error. 434 if (CondInfo.FoundElse) 435 Diag(Tok, diag::pp_err_elif_after_else); 436 437 // If this is in a skipping block or if we're already handled this #if 438 // block, don't bother parsing the condition. We just skip this block. 439 if (CondInfo.FoundNonSkip) 440 continue; 441 442 // Evaluate the condition of the #elif. 443 IdentifierInfo *IfNDefMacro = 0; 444 CurPTHLexer->ParsingPreprocessorDirective = true; 445 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 446 CurPTHLexer->ParsingPreprocessorDirective = false; 447 448 // If this condition is true, enter it! 449 if (ShouldEnter) { 450 CondInfo.FoundNonSkip = true; 451 break; 452 } 453 454 // Otherwise, skip this block and go to the next one. 455 continue; 456 } 457} 458 459/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 460/// return null on failure. isAngled indicates whether the file reference is 461/// for system #include's or not (i.e. using <> instead of ""). 462const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 463 bool isAngled, 464 const DirectoryLookup *FromDir, 465 const DirectoryLookup *&CurDir) { 466 // If the header lookup mechanism may be relative to the current file, pass in 467 // info about where the current file is. 468 const FileEntry *CurFileEnt = 0; 469 if (!FromDir) { 470 FileID FID = getCurrentFileLexer()->getFileID(); 471 CurFileEnt = SourceMgr.getFileEntryForID(FID); 472 473 // If there is no file entry associated with this file, it must be the 474 // predefines buffer. Any other file is not lexed with a normal lexer, so 475 // it won't be scanned for preprocessor directives. If we have the 476 // predefines buffer, resolve #include references (which come from the 477 // -include command line argument) as if they came from the main file, this 478 // affects file lookup etc. 479 if (CurFileEnt == 0) { 480 FID = SourceMgr.getMainFileID(); 481 CurFileEnt = SourceMgr.getFileEntryForID(FID); 482 } 483 } 484 485 // Do a standard file entry lookup. 486 CurDir = CurDirLookup; 487 const FileEntry *FE = 488 HeaderInfo.LookupFile(Filename, isAngled, FromDir, CurDir, CurFileEnt); 489 if (FE) return FE; 490 491 // Otherwise, see if this is a subframework header. If so, this is relative 492 // to one of the headers on the #include stack. Walk the list of the current 493 // headers on the #include stack and pass them to HeaderInfo. 494 if (IsFileLexer()) { 495 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 496 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 497 return FE; 498 } 499 500 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 501 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 502 if (IsFileLexer(ISEntry)) { 503 if ((CurFileEnt = 504 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 505 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 506 return FE; 507 } 508 } 509 510 // Otherwise, we really couldn't find the file. 511 return 0; 512} 513 514 515//===----------------------------------------------------------------------===// 516// Preprocessor Directive Handling. 517//===----------------------------------------------------------------------===// 518 519/// HandleDirective - This callback is invoked when the lexer sees a # token 520/// at the start of a line. This consumes the directive, modifies the 521/// lexer/preprocessor state, and advances the lexer(s) so that the next token 522/// read is the correct one. 523void Preprocessor::HandleDirective(Token &Result) { 524 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 525 526 // We just parsed a # character at the start of a line, so we're in directive 527 // mode. Tell the lexer this so any newlines we see will be converted into an 528 // EOM token (which terminates the directive). 529 CurPPLexer->ParsingPreprocessorDirective = true; 530 531 ++NumDirectives; 532 533 // We are about to read a token. For the multiple-include optimization FA to 534 // work, we have to remember if we had read any tokens *before* this 535 // pp-directive. 536 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 537 538 // Save the '#' token in case we need to return it later. 539 Token SavedHash = Result; 540 541 // Read the next token, the directive flavor. This isn't expanded due to 542 // C99 6.10.3p8. 543 LexUnexpandedToken(Result); 544 545 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 546 // #define A(x) #x 547 // A(abc 548 // #warning blah 549 // def) 550 // If so, the user is relying on non-portable behavior, emit a diagnostic. 551 if (InMacroArgs) 552 Diag(Result, diag::ext_embedded_directive); 553 554TryAgain: 555 switch (Result.getKind()) { 556 case tok::eom: 557 return; // null directive. 558 case tok::comment: 559 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 560 LexUnexpandedToken(Result); 561 goto TryAgain; 562 case tok::code_completion: 563 if (CodeComplete) 564 CodeComplete->CodeCompleteDirective( 565 CurPPLexer->getConditionalStackDepth() > 0); 566 return; 567 case tok::numeric_constant: // # 7 GNU line marker directive. 568 if (getLangOptions().AsmPreprocessor) 569 break; // # 4 is not a preprocessor directive in .S files. 570 return HandleDigitDirective(Result); 571 default: 572 IdentifierInfo *II = Result.getIdentifierInfo(); 573 if (II == 0) break; // Not an identifier. 574 575 // Ask what the preprocessor keyword ID is. 576 switch (II->getPPKeywordID()) { 577 default: break; 578 // C99 6.10.1 - Conditional Inclusion. 579 case tok::pp_if: 580 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 581 case tok::pp_ifdef: 582 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 583 case tok::pp_ifndef: 584 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 585 case tok::pp_elif: 586 return HandleElifDirective(Result); 587 case tok::pp_else: 588 return HandleElseDirective(Result); 589 case tok::pp_endif: 590 return HandleEndifDirective(Result); 591 592 // C99 6.10.2 - Source File Inclusion. 593 case tok::pp_include: 594 // Handle #include. 595 return HandleIncludeDirective(SavedHash.getLocation(), Result); 596 case tok::pp___include_macros: 597 // Handle -imacros. 598 return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result); 599 600 // C99 6.10.3 - Macro Replacement. 601 case tok::pp_define: 602 return HandleDefineDirective(Result); 603 case tok::pp_undef: 604 return HandleUndefDirective(Result); 605 606 // C99 6.10.4 - Line Control. 607 case tok::pp_line: 608 return HandleLineDirective(Result); 609 610 // C99 6.10.5 - Error Directive. 611 case tok::pp_error: 612 return HandleUserDiagnosticDirective(Result, false); 613 614 // C99 6.10.6 - Pragma Directive. 615 case tok::pp_pragma: 616 return HandlePragmaDirective(PIK_HashPragma); 617 618 // GNU Extensions. 619 case tok::pp_import: 620 return HandleImportDirective(SavedHash.getLocation(), Result); 621 case tok::pp_include_next: 622 return HandleIncludeNextDirective(SavedHash.getLocation(), Result); 623 624 case tok::pp_warning: 625 Diag(Result, diag::ext_pp_warning_directive); 626 return HandleUserDiagnosticDirective(Result, true); 627 case tok::pp_ident: 628 return HandleIdentSCCSDirective(Result); 629 case tok::pp_sccs: 630 return HandleIdentSCCSDirective(Result); 631 case tok::pp_assert: 632 //isExtension = true; // FIXME: implement #assert 633 break; 634 case tok::pp_unassert: 635 //isExtension = true; // FIXME: implement #unassert 636 break; 637 } 638 break; 639 } 640 641 // If this is a .S file, treat unknown # directives as non-preprocessor 642 // directives. This is important because # may be a comment or introduce 643 // various pseudo-ops. Just return the # token and push back the following 644 // token to be lexed next time. 645 if (getLangOptions().AsmPreprocessor) { 646 Token *Toks = new Token[2]; 647 // Return the # and the token after it. 648 Toks[0] = SavedHash; 649 Toks[1] = Result; 650 // Enter this token stream so that we re-lex the tokens. Make sure to 651 // enable macro expansion, in case the token after the # is an identifier 652 // that is expanded. 653 EnterTokenStream(Toks, 2, false, true); 654 return; 655 } 656 657 // If we reached here, the preprocessing token is not valid! 658 Diag(Result, diag::err_pp_invalid_directive); 659 660 // Read the rest of the PP line. 661 DiscardUntilEndOfDirective(); 662 663 // Okay, we're done parsing the directive. 664} 665 666/// GetLineValue - Convert a numeric token into an unsigned value, emitting 667/// Diagnostic DiagID if it is invalid, and returning the value in Val. 668static bool GetLineValue(Token &DigitTok, unsigned &Val, 669 unsigned DiagID, Preprocessor &PP) { 670 if (DigitTok.isNot(tok::numeric_constant)) { 671 PP.Diag(DigitTok, DiagID); 672 673 if (DigitTok.isNot(tok::eom)) 674 PP.DiscardUntilEndOfDirective(); 675 return true; 676 } 677 678 llvm::SmallString<64> IntegerBuffer; 679 IntegerBuffer.resize(DigitTok.getLength()); 680 const char *DigitTokBegin = &IntegerBuffer[0]; 681 bool Invalid = false; 682 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 683 if (Invalid) 684 return true; 685 686 // Verify that we have a simple digit-sequence, and compute the value. This 687 // is always a simple digit string computed in decimal, so we do this manually 688 // here. 689 Val = 0; 690 for (unsigned i = 0; i != ActualLength; ++i) { 691 if (!isdigit(DigitTokBegin[i])) { 692 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 693 diag::err_pp_line_digit_sequence); 694 PP.DiscardUntilEndOfDirective(); 695 return true; 696 } 697 698 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 699 if (NextVal < Val) { // overflow. 700 PP.Diag(DigitTok, DiagID); 701 PP.DiscardUntilEndOfDirective(); 702 return true; 703 } 704 Val = NextVal; 705 } 706 707 // Reject 0, this is needed both by #line numbers and flags. 708 if (Val == 0) { 709 PP.Diag(DigitTok, DiagID); 710 PP.DiscardUntilEndOfDirective(); 711 return true; 712 } 713 714 if (DigitTokBegin[0] == '0') 715 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 716 717 return false; 718} 719 720/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 721/// acceptable forms are: 722/// # line digit-sequence 723/// # line digit-sequence "s-char-sequence" 724void Preprocessor::HandleLineDirective(Token &Tok) { 725 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 726 // expanded. 727 Token DigitTok; 728 Lex(DigitTok); 729 730 // Validate the number and convert it to an unsigned. 731 unsigned LineNo; 732 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 733 return; 734 735 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 736 // number greater than 2147483647". C90 requires that the line # be <= 32767. 737 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 738 if (LineNo >= LineLimit) 739 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 740 741 int FilenameID = -1; 742 Token StrTok; 743 Lex(StrTok); 744 745 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 746 // string followed by eom. 747 if (StrTok.is(tok::eom)) 748 ; // ok 749 else if (StrTok.isNot(tok::string_literal)) { 750 Diag(StrTok, diag::err_pp_line_invalid_filename); 751 DiscardUntilEndOfDirective(); 752 return; 753 } else { 754 // Parse and validate the string, converting it into a unique ID. 755 StringLiteralParser Literal(&StrTok, 1, *this); 756 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 757 if (Literal.hadError) 758 return DiscardUntilEndOfDirective(); 759 if (Literal.Pascal) { 760 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 761 return DiscardUntilEndOfDirective(); 762 } 763 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 764 Literal.GetStringLength()); 765 766 // Verify that there is nothing after the string, other than EOM. Because 767 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 768 CheckEndOfDirective("line", true); 769 } 770 771 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 772 773 if (Callbacks) 774 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 775 PPCallbacks::RenameFile, 776 SrcMgr::C_User); 777} 778 779/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 780/// marker directive. 781static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 782 bool &IsSystemHeader, bool &IsExternCHeader, 783 Preprocessor &PP) { 784 unsigned FlagVal; 785 Token FlagTok; 786 PP.Lex(FlagTok); 787 if (FlagTok.is(tok::eom)) return false; 788 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 789 return true; 790 791 if (FlagVal == 1) { 792 IsFileEntry = true; 793 794 PP.Lex(FlagTok); 795 if (FlagTok.is(tok::eom)) return false; 796 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 797 return true; 798 } else if (FlagVal == 2) { 799 IsFileExit = true; 800 801 SourceManager &SM = PP.getSourceManager(); 802 // If we are leaving the current presumed file, check to make sure the 803 // presumed include stack isn't empty! 804 FileID CurFileID = 805 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 806 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 807 if (PLoc.isInvalid()) 808 return true; 809 810 // If there is no include loc (main file) or if the include loc is in a 811 // different physical file, then we aren't in a "1" line marker flag region. 812 SourceLocation IncLoc = PLoc.getIncludeLoc(); 813 if (IncLoc.isInvalid() || 814 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 815 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 816 PP.DiscardUntilEndOfDirective(); 817 return true; 818 } 819 820 PP.Lex(FlagTok); 821 if (FlagTok.is(tok::eom)) return false; 822 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 823 return true; 824 } 825 826 // We must have 3 if there are still flags. 827 if (FlagVal != 3) { 828 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 829 PP.DiscardUntilEndOfDirective(); 830 return true; 831 } 832 833 IsSystemHeader = true; 834 835 PP.Lex(FlagTok); 836 if (FlagTok.is(tok::eom)) return false; 837 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 838 return true; 839 840 // We must have 4 if there is yet another flag. 841 if (FlagVal != 4) { 842 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 843 PP.DiscardUntilEndOfDirective(); 844 return true; 845 } 846 847 IsExternCHeader = true; 848 849 PP.Lex(FlagTok); 850 if (FlagTok.is(tok::eom)) return false; 851 852 // There are no more valid flags here. 853 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 854 PP.DiscardUntilEndOfDirective(); 855 return true; 856} 857 858/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 859/// one of the following forms: 860/// 861/// # 42 862/// # 42 "file" ('1' | '2')? 863/// # 42 "file" ('1' | '2')? '3' '4'? 864/// 865void Preprocessor::HandleDigitDirective(Token &DigitTok) { 866 // Validate the number and convert it to an unsigned. GNU does not have a 867 // line # limit other than it fit in 32-bits. 868 unsigned LineNo; 869 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 870 *this)) 871 return; 872 873 Token StrTok; 874 Lex(StrTok); 875 876 bool IsFileEntry = false, IsFileExit = false; 877 bool IsSystemHeader = false, IsExternCHeader = false; 878 int FilenameID = -1; 879 880 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 881 // string followed by eom. 882 if (StrTok.is(tok::eom)) 883 ; // ok 884 else if (StrTok.isNot(tok::string_literal)) { 885 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 886 return DiscardUntilEndOfDirective(); 887 } else { 888 // Parse and validate the string, converting it into a unique ID. 889 StringLiteralParser Literal(&StrTok, 1, *this); 890 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 891 if (Literal.hadError) 892 return DiscardUntilEndOfDirective(); 893 if (Literal.Pascal) { 894 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 895 return DiscardUntilEndOfDirective(); 896 } 897 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 898 Literal.GetStringLength()); 899 900 // If a filename was present, read any flags that are present. 901 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 902 IsSystemHeader, IsExternCHeader, *this)) 903 return; 904 } 905 906 // Create a line note with this information. 907 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 908 IsFileEntry, IsFileExit, 909 IsSystemHeader, IsExternCHeader); 910 911 // If the preprocessor has callbacks installed, notify them of the #line 912 // change. This is used so that the line marker comes out in -E mode for 913 // example. 914 if (Callbacks) { 915 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 916 if (IsFileEntry) 917 Reason = PPCallbacks::EnterFile; 918 else if (IsFileExit) 919 Reason = PPCallbacks::ExitFile; 920 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 921 if (IsExternCHeader) 922 FileKind = SrcMgr::C_ExternCSystem; 923 else if (IsSystemHeader) 924 FileKind = SrcMgr::C_System; 925 926 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 927 } 928} 929 930 931/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 932/// 933void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 934 bool isWarning) { 935 // PTH doesn't emit #warning or #error directives. 936 if (CurPTHLexer) 937 return CurPTHLexer->DiscardToEndOfLine(); 938 939 // Read the rest of the line raw. We do this because we don't want macros 940 // to be expanded and we don't require that the tokens be valid preprocessing 941 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 942 // collapse multiple consequtive white space between tokens, but this isn't 943 // specified by the standard. 944 std::string Message = CurLexer->ReadToEndOfLine(); 945 if (isWarning) 946 Diag(Tok, diag::pp_hash_warning) << Message; 947 else 948 Diag(Tok, diag::err_pp_hash_error) << Message; 949} 950 951/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 952/// 953void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 954 // Yes, this directive is an extension. 955 Diag(Tok, diag::ext_pp_ident_directive); 956 957 // Read the string argument. 958 Token StrTok; 959 Lex(StrTok); 960 961 // If the token kind isn't a string, it's a malformed directive. 962 if (StrTok.isNot(tok::string_literal) && 963 StrTok.isNot(tok::wide_string_literal)) { 964 Diag(StrTok, diag::err_pp_malformed_ident); 965 if (StrTok.isNot(tok::eom)) 966 DiscardUntilEndOfDirective(); 967 return; 968 } 969 970 // Verify that there is nothing after the string, other than EOM. 971 CheckEndOfDirective("ident"); 972 973 if (Callbacks) { 974 bool Invalid = false; 975 std::string Str = getSpelling(StrTok, &Invalid); 976 if (!Invalid) 977 Callbacks->Ident(Tok.getLocation(), Str); 978 } 979} 980 981//===----------------------------------------------------------------------===// 982// Preprocessor Include Directive Handling. 983//===----------------------------------------------------------------------===// 984 985/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 986/// checked and spelled filename, e.g. as an operand of #include. This returns 987/// true if the input filename was in <>'s or false if it were in ""'s. The 988/// caller is expected to provide a buffer that is large enough to hold the 989/// spelling of the filename, but is also expected to handle the case when 990/// this method decides to use a different buffer. 991bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 992 llvm::StringRef &Buffer) { 993 // Get the text form of the filename. 994 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 995 996 // Make sure the filename is <x> or "x". 997 bool isAngled; 998 if (Buffer[0] == '<') { 999 if (Buffer.back() != '>') { 1000 Diag(Loc, diag::err_pp_expects_filename); 1001 Buffer = llvm::StringRef(); 1002 return true; 1003 } 1004 isAngled = true; 1005 } else if (Buffer[0] == '"') { 1006 if (Buffer.back() != '"') { 1007 Diag(Loc, diag::err_pp_expects_filename); 1008 Buffer = llvm::StringRef(); 1009 return true; 1010 } 1011 isAngled = false; 1012 } else { 1013 Diag(Loc, diag::err_pp_expects_filename); 1014 Buffer = llvm::StringRef(); 1015 return true; 1016 } 1017 1018 // Diagnose #include "" as invalid. 1019 if (Buffer.size() <= 2) { 1020 Diag(Loc, diag::err_pp_empty_filename); 1021 Buffer = llvm::StringRef(); 1022 return true; 1023 } 1024 1025 // Skip the brackets. 1026 Buffer = Buffer.substr(1, Buffer.size()-2); 1027 return isAngled; 1028} 1029 1030/// ConcatenateIncludeName - Handle cases where the #include name is expanded 1031/// from a macro as multiple tokens, which need to be glued together. This 1032/// occurs for code like: 1033/// #define FOO <a/b.h> 1034/// #include FOO 1035/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1036/// 1037/// This code concatenates and consumes tokens up to the '>' token. It returns 1038/// false if the > was found, otherwise it returns true if it finds and consumes 1039/// the EOM marker. 1040bool Preprocessor::ConcatenateIncludeName( 1041 llvm::SmallString<128> &FilenameBuffer, 1042 SourceLocation &End) { 1043 Token CurTok; 1044 1045 Lex(CurTok); 1046 while (CurTok.isNot(tok::eom)) { 1047 End = CurTok.getLocation(); 1048 1049 // FIXME: Provide code completion for #includes. 1050 if (CurTok.is(tok::code_completion)) { 1051 Lex(CurTok); 1052 continue; 1053 } 1054 1055 // Append the spelling of this token to the buffer. If there was a space 1056 // before it, add it now. 1057 if (CurTok.hasLeadingSpace()) 1058 FilenameBuffer.push_back(' '); 1059 1060 // Get the spelling of the token, directly into FilenameBuffer if possible. 1061 unsigned PreAppendSize = FilenameBuffer.size(); 1062 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1063 1064 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1065 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1066 1067 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1068 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1069 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1070 1071 // Resize FilenameBuffer to the correct size. 1072 if (CurTok.getLength() != ActualLen) 1073 FilenameBuffer.resize(PreAppendSize+ActualLen); 1074 1075 // If we found the '>' marker, return success. 1076 if (CurTok.is(tok::greater)) 1077 return false; 1078 1079 Lex(CurTok); 1080 } 1081 1082 // If we hit the eom marker, emit an error and return true so that the caller 1083 // knows the EOM has been read. 1084 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1085 return true; 1086} 1087 1088/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1089/// file to be included from the lexer, then include it! This is a common 1090/// routine with functionality shared between #include, #include_next and 1091/// #import. LookupFrom is set when this is a #include_next directive, it 1092/// specifies the file to start searching from. 1093void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc, 1094 Token &IncludeTok, 1095 const DirectoryLookup *LookupFrom, 1096 bool isImport) { 1097 1098 Token FilenameTok; 1099 CurPPLexer->LexIncludeFilename(FilenameTok); 1100 1101 // Reserve a buffer to get the spelling. 1102 llvm::SmallString<128> FilenameBuffer; 1103 llvm::StringRef Filename; 1104 SourceLocation End; 1105 1106 switch (FilenameTok.getKind()) { 1107 case tok::eom: 1108 // If the token kind is EOM, the error has already been diagnosed. 1109 return; 1110 1111 case tok::angle_string_literal: 1112 case tok::string_literal: 1113 Filename = getSpelling(FilenameTok, FilenameBuffer); 1114 End = FilenameTok.getLocation(); 1115 break; 1116 1117 case tok::less: 1118 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1119 // case, glue the tokens together into FilenameBuffer and interpret those. 1120 FilenameBuffer.push_back('<'); 1121 if (ConcatenateIncludeName(FilenameBuffer, End)) 1122 return; // Found <eom> but no ">"? Diagnostic already emitted. 1123 Filename = FilenameBuffer.str(); 1124 break; 1125 default: 1126 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1127 DiscardUntilEndOfDirective(); 1128 return; 1129 } 1130 1131 bool isAngled = 1132 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1133 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1134 // error. 1135 if (Filename.empty()) { 1136 DiscardUntilEndOfDirective(); 1137 return; 1138 } 1139 1140 // Verify that there is nothing after the filename, other than EOM. Note that 1141 // we allow macros that expand to nothing after the filename, because this 1142 // falls into the category of "#include pp-tokens new-line" specified in 1143 // C99 6.10.2p4. 1144 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1145 1146 // Check that we don't have infinite #include recursion. 1147 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1148 Diag(FilenameTok, diag::err_pp_include_too_deep); 1149 return; 1150 } 1151 1152 // Search include directories. 1153 const DirectoryLookup *CurDir; 1154 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1155 if (File == 0) { 1156 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1157 return; 1158 } 1159 1160 // Notify the callback object that we've seen an inclusion directive. 1161 if (Callbacks) 1162 Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, File, 1163 End); 1164 1165 // The #included file will be considered to be a system header if either it is 1166 // in a system include directory, or if the #includer is a system include 1167 // header. 1168 SrcMgr::CharacteristicKind FileCharacter = 1169 std::max(HeaderInfo.getFileDirFlavor(File), 1170 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1171 1172 // Ask HeaderInfo if we should enter this #include file. If not, #including 1173 // this file will have no effect. 1174 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1175 if (Callbacks) 1176 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1177 return; 1178 } 1179 1180 // Look up the file, create a File ID for it. 1181 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1182 FileCharacter); 1183 if (FID.isInvalid()) { 1184 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1185 return; 1186 } 1187 1188 // Finally, if all is good, enter the new file! 1189 EnterSourceFile(FID, CurDir, FilenameTok.getLocation()); 1190} 1191 1192/// HandleIncludeNextDirective - Implements #include_next. 1193/// 1194void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc, 1195 Token &IncludeNextTok) { 1196 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1197 1198 // #include_next is like #include, except that we start searching after 1199 // the current found directory. If we can't do this, issue a 1200 // diagnostic. 1201 const DirectoryLookup *Lookup = CurDirLookup; 1202 if (isInPrimaryFile()) { 1203 Lookup = 0; 1204 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1205 } else if (Lookup == 0) { 1206 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1207 } else { 1208 // Start looking up in the next directory. 1209 ++Lookup; 1210 } 1211 1212 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup); 1213} 1214 1215/// HandleImportDirective - Implements #import. 1216/// 1217void Preprocessor::HandleImportDirective(SourceLocation HashLoc, 1218 Token &ImportTok) { 1219 if (!Features.ObjC1) // #import is standard for ObjC. 1220 Diag(ImportTok, diag::ext_pp_import_directive); 1221 1222 return HandleIncludeDirective(HashLoc, ImportTok, 0, true); 1223} 1224 1225/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1226/// pseudo directive in the predefines buffer. This handles it by sucking all 1227/// tokens through the preprocessor and discarding them (only keeping the side 1228/// effects on the preprocessor). 1229void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc, 1230 Token &IncludeMacrosTok) { 1231 // This directive should only occur in the predefines buffer. If not, emit an 1232 // error and reject it. 1233 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1234 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1235 Diag(IncludeMacrosTok.getLocation(), 1236 diag::pp_include_macros_out_of_predefines); 1237 DiscardUntilEndOfDirective(); 1238 return; 1239 } 1240 1241 // Treat this as a normal #include for checking purposes. If this is 1242 // successful, it will push a new lexer onto the include stack. 1243 HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false); 1244 1245 Token TmpTok; 1246 do { 1247 Lex(TmpTok); 1248 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1249 } while (TmpTok.isNot(tok::hashhash)); 1250} 1251 1252//===----------------------------------------------------------------------===// 1253// Preprocessor Macro Directive Handling. 1254//===----------------------------------------------------------------------===// 1255 1256/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1257/// definition has just been read. Lex the rest of the arguments and the 1258/// closing ), updating MI with what we learn. Return true if an error occurs 1259/// parsing the arg list. 1260bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1261 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1262 1263 Token Tok; 1264 while (1) { 1265 LexUnexpandedToken(Tok); 1266 switch (Tok.getKind()) { 1267 case tok::r_paren: 1268 // Found the end of the argument list. 1269 if (Arguments.empty()) // #define FOO() 1270 return false; 1271 // Otherwise we have #define FOO(A,) 1272 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1273 return true; 1274 case tok::ellipsis: // #define X(... -> C99 varargs 1275 // Warn if use of C99 feature in non-C99 mode. 1276 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1277 1278 // Lex the token after the identifier. 1279 LexUnexpandedToken(Tok); 1280 if (Tok.isNot(tok::r_paren)) { 1281 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1282 return true; 1283 } 1284 // Add the __VA_ARGS__ identifier as an argument. 1285 Arguments.push_back(Ident__VA_ARGS__); 1286 MI->setIsC99Varargs(); 1287 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1288 return false; 1289 case tok::eom: // #define X( 1290 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1291 return true; 1292 default: 1293 // Handle keywords and identifiers here to accept things like 1294 // #define Foo(for) for. 1295 IdentifierInfo *II = Tok.getIdentifierInfo(); 1296 if (II == 0) { 1297 // #define X(1 1298 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1299 return true; 1300 } 1301 1302 // If this is already used as an argument, it is used multiple times (e.g. 1303 // #define X(A,A. 1304 if (std::find(Arguments.begin(), Arguments.end(), II) != 1305 Arguments.end()) { // C99 6.10.3p6 1306 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1307 return true; 1308 } 1309 1310 // Add the argument to the macro info. 1311 Arguments.push_back(II); 1312 1313 // Lex the token after the identifier. 1314 LexUnexpandedToken(Tok); 1315 1316 switch (Tok.getKind()) { 1317 default: // #define X(A B 1318 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1319 return true; 1320 case tok::r_paren: // #define X(A) 1321 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1322 return false; 1323 case tok::comma: // #define X(A, 1324 break; 1325 case tok::ellipsis: // #define X(A... -> GCC extension 1326 // Diagnose extension. 1327 Diag(Tok, diag::ext_named_variadic_macro); 1328 1329 // Lex the token after the identifier. 1330 LexUnexpandedToken(Tok); 1331 if (Tok.isNot(tok::r_paren)) { 1332 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1333 return true; 1334 } 1335 1336 MI->setIsGNUVarargs(); 1337 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1338 return false; 1339 } 1340 } 1341 } 1342} 1343 1344/// HandleDefineDirective - Implements #define. This consumes the entire macro 1345/// line then lets the caller lex the next real token. 1346void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1347 ++NumDefined; 1348 1349 Token MacroNameTok; 1350 ReadMacroName(MacroNameTok, 1); 1351 1352 // Error reading macro name? If so, diagnostic already issued. 1353 if (MacroNameTok.is(tok::eom)) 1354 return; 1355 1356 Token LastTok = MacroNameTok; 1357 1358 // If we are supposed to keep comments in #defines, reenable comment saving 1359 // mode. 1360 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1361 1362 // Create the new macro. 1363 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1364 1365 Token Tok; 1366 LexUnexpandedToken(Tok); 1367 1368 // If this is a function-like macro definition, parse the argument list, 1369 // marking each of the identifiers as being used as macro arguments. Also, 1370 // check other constraints on the first token of the macro body. 1371 if (Tok.is(tok::eom)) { 1372 // If there is no body to this macro, we have no special handling here. 1373 } else if (Tok.hasLeadingSpace()) { 1374 // This is a normal token with leading space. Clear the leading space 1375 // marker on the first token to get proper expansion. 1376 Tok.clearFlag(Token::LeadingSpace); 1377 } else if (Tok.is(tok::l_paren)) { 1378 // This is a function-like macro definition. Read the argument list. 1379 MI->setIsFunctionLike(); 1380 if (ReadMacroDefinitionArgList(MI)) { 1381 // Forget about MI. 1382 ReleaseMacroInfo(MI); 1383 // Throw away the rest of the line. 1384 if (CurPPLexer->ParsingPreprocessorDirective) 1385 DiscardUntilEndOfDirective(); 1386 return; 1387 } 1388 1389 // If this is a definition of a variadic C99 function-like macro, not using 1390 // the GNU named varargs extension, enabled __VA_ARGS__. 1391 1392 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1393 // This gets unpoisoned where it is allowed. 1394 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1395 if (MI->isC99Varargs()) 1396 Ident__VA_ARGS__->setIsPoisoned(false); 1397 1398 // Read the first token after the arg list for down below. 1399 LexUnexpandedToken(Tok); 1400 } else if (Features.C99) { 1401 // C99 requires whitespace between the macro definition and the body. Emit 1402 // a diagnostic for something like "#define X+". 1403 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1404 } else { 1405 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1406 // first character of a replacement list is not a character required by 1407 // subclause 5.2.1, then there shall be white-space separation between the 1408 // identifier and the replacement list.". 5.2.1 lists this set: 1409 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1410 // is irrelevant here. 1411 bool isInvalid = false; 1412 if (Tok.is(tok::at)) // @ is not in the list above. 1413 isInvalid = true; 1414 else if (Tok.is(tok::unknown)) { 1415 // If we have an unknown token, it is something strange like "`". Since 1416 // all of valid characters would have lexed into a single character 1417 // token of some sort, we know this is not a valid case. 1418 isInvalid = true; 1419 } 1420 if (isInvalid) 1421 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1422 else 1423 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1424 } 1425 1426 if (!Tok.is(tok::eom)) 1427 LastTok = Tok; 1428 1429 // Read the rest of the macro body. 1430 if (MI->isObjectLike()) { 1431 // Object-like macros are very simple, just read their body. 1432 while (Tok.isNot(tok::eom)) { 1433 LastTok = Tok; 1434 MI->AddTokenToBody(Tok); 1435 // Get the next token of the macro. 1436 LexUnexpandedToken(Tok); 1437 } 1438 1439 } else { 1440 // Otherwise, read the body of a function-like macro. While we are at it, 1441 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1442 // parameters in function-like macro expansions. 1443 while (Tok.isNot(tok::eom)) { 1444 LastTok = Tok; 1445 1446 if (Tok.isNot(tok::hash)) { 1447 MI->AddTokenToBody(Tok); 1448 1449 // Get the next token of the macro. 1450 LexUnexpandedToken(Tok); 1451 continue; 1452 } 1453 1454 // Get the next token of the macro. 1455 LexUnexpandedToken(Tok); 1456 1457 // Check for a valid macro arg identifier. 1458 if (Tok.getIdentifierInfo() == 0 || 1459 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1460 1461 // If this is assembler-with-cpp mode, we accept random gibberish after 1462 // the '#' because '#' is often a comment character. However, change 1463 // the kind of the token to tok::unknown so that the preprocessor isn't 1464 // confused. 1465 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1466 LastTok.setKind(tok::unknown); 1467 } else { 1468 Diag(Tok, diag::err_pp_stringize_not_parameter); 1469 ReleaseMacroInfo(MI); 1470 1471 // Disable __VA_ARGS__ again. 1472 Ident__VA_ARGS__->setIsPoisoned(true); 1473 return; 1474 } 1475 } 1476 1477 // Things look ok, add the '#' and param name tokens to the macro. 1478 MI->AddTokenToBody(LastTok); 1479 MI->AddTokenToBody(Tok); 1480 LastTok = Tok; 1481 1482 // Get the next token of the macro. 1483 LexUnexpandedToken(Tok); 1484 } 1485 } 1486 1487 1488 // Disable __VA_ARGS__ again. 1489 Ident__VA_ARGS__->setIsPoisoned(true); 1490 1491 // Check that there is no paste (##) operator at the begining or end of the 1492 // replacement list. 1493 unsigned NumTokens = MI->getNumTokens(); 1494 if (NumTokens != 0) { 1495 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1496 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1497 ReleaseMacroInfo(MI); 1498 return; 1499 } 1500 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1501 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1502 ReleaseMacroInfo(MI); 1503 return; 1504 } 1505 } 1506 1507 // If this is the primary source file, remember that this macro hasn't been 1508 // used yet. 1509 if (isInPrimaryFile()) 1510 MI->setIsUsed(false); 1511 1512 MI->setDefinitionEndLoc(LastTok.getLocation()); 1513 1514 // Finally, if this identifier already had a macro defined for it, verify that 1515 // the macro bodies are identical and free the old definition. 1516 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1517 // It is very common for system headers to have tons of macro redefinitions 1518 // and for warnings to be disabled in system headers. If this is the case, 1519 // then don't bother calling MacroInfo::isIdenticalTo. 1520 if (!getDiagnostics().getSuppressSystemWarnings() || 1521 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1522 if (!OtherMI->isUsed()) 1523 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1524 1525 // Macros must be identical. This means all tokens and whitespace 1526 // separation must be the same. C99 6.10.3.2. 1527 if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 1528 !MI->isIdenticalTo(*OtherMI, *this)) { 1529 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1530 << MacroNameTok.getIdentifierInfo(); 1531 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1532 } 1533 } 1534 ReleaseMacroInfo(OtherMI); 1535 } 1536 1537 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1538 1539 // If the callbacks want to know, tell them about the macro definition. 1540 if (Callbacks) 1541 Callbacks->MacroDefined(MacroNameTok, MI); 1542} 1543 1544/// HandleUndefDirective - Implements #undef. 1545/// 1546void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1547 ++NumUndefined; 1548 1549 Token MacroNameTok; 1550 ReadMacroName(MacroNameTok, 2); 1551 1552 // Error reading macro name? If so, diagnostic already issued. 1553 if (MacroNameTok.is(tok::eom)) 1554 return; 1555 1556 // Check to see if this is the last token on the #undef line. 1557 CheckEndOfDirective("undef"); 1558 1559 // Okay, we finally have a valid identifier to undef. 1560 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1561 1562 // If the macro is not defined, this is a noop undef, just return. 1563 if (MI == 0) return; 1564 1565 if (!MI->isUsed()) 1566 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1567 1568 // If the callbacks want to know, tell them about the macro #undef. 1569 if (Callbacks) 1570 Callbacks->MacroUndefined(MacroNameTok, MI); 1571 1572 // Free macro definition. 1573 ReleaseMacroInfo(MI); 1574 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1575} 1576 1577 1578//===----------------------------------------------------------------------===// 1579// Preprocessor Conditional Directive Handling. 1580//===----------------------------------------------------------------------===// 1581 1582/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1583/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1584/// if any tokens have been returned or pp-directives activated before this 1585/// #ifndef has been lexed. 1586/// 1587void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1588 bool ReadAnyTokensBeforeDirective) { 1589 ++NumIf; 1590 Token DirectiveTok = Result; 1591 1592 Token MacroNameTok; 1593 ReadMacroName(MacroNameTok); 1594 1595 // Error reading macro name? If so, diagnostic already issued. 1596 if (MacroNameTok.is(tok::eom)) { 1597 // Skip code until we get to #endif. This helps with recovery by not 1598 // emitting an error when the #endif is reached. 1599 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1600 /*Foundnonskip*/false, /*FoundElse*/false); 1601 return; 1602 } 1603 1604 // Check to see if this is the last token on the #if[n]def line. 1605 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1606 1607 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1608 MacroInfo *MI = getMacroInfo(MII); 1609 1610 if (CurPPLexer->getConditionalStackDepth() == 0) { 1611 // If the start of a top-level #ifdef and if the macro is not defined, 1612 // inform MIOpt that this might be the start of a proper include guard. 1613 // Otherwise it is some other form of unknown conditional which we can't 1614 // handle. 1615 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1616 assert(isIfndef && "#ifdef shouldn't reach here"); 1617 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1618 } else 1619 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1620 } 1621 1622 // If there is a macro, process it. 1623 if (MI) // Mark it used. 1624 MI->setIsUsed(true); 1625 1626 // Should we include the stuff contained by this directive? 1627 if (!MI == isIfndef) { 1628 // Yes, remember that we are inside a conditional, then lex the next token. 1629 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1630 /*wasskip*/false, /*foundnonskip*/true, 1631 /*foundelse*/false); 1632 } else { 1633 // No, skip the contents of this block. 1634 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1635 /*Foundnonskip*/false, 1636 /*FoundElse*/false); 1637 } 1638 1639 if (Callbacks) { 1640 if (isIfndef) 1641 Callbacks->Ifndef(MacroNameTok); 1642 else 1643 Callbacks->Ifdef(MacroNameTok); 1644 } 1645} 1646 1647/// HandleIfDirective - Implements the #if directive. 1648/// 1649void Preprocessor::HandleIfDirective(Token &IfToken, 1650 bool ReadAnyTokensBeforeDirective) { 1651 ++NumIf; 1652 1653 // Parse and evaluate the conditional expression. 1654 IdentifierInfo *IfNDefMacro = 0; 1655 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1656 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1657 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1658 1659 // If this condition is equivalent to #ifndef X, and if this is the first 1660 // directive seen, handle it for the multiple-include optimization. 1661 if (CurPPLexer->getConditionalStackDepth() == 0) { 1662 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1663 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1664 else 1665 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1666 } 1667 1668 // Should we include the stuff contained by this directive? 1669 if (ConditionalTrue) { 1670 // Yes, remember that we are inside a conditional, then lex the next token. 1671 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1672 /*foundnonskip*/true, /*foundelse*/false); 1673 } else { 1674 // No, skip the contents of this block. 1675 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1676 /*FoundElse*/false); 1677 } 1678 1679 if (Callbacks) 1680 Callbacks->If(SourceRange(ConditionalBegin, ConditionalEnd)); 1681} 1682 1683/// HandleEndifDirective - Implements the #endif directive. 1684/// 1685void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1686 ++NumEndif; 1687 1688 // Check that this is the whole directive. 1689 CheckEndOfDirective("endif"); 1690 1691 PPConditionalInfo CondInfo; 1692 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1693 // No conditionals on the stack: this is an #endif without an #if. 1694 Diag(EndifToken, diag::err_pp_endif_without_if); 1695 return; 1696 } 1697 1698 // If this the end of a top-level #endif, inform MIOpt. 1699 if (CurPPLexer->getConditionalStackDepth() == 0) 1700 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1701 1702 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1703 "This code should only be reachable in the non-skipping case!"); 1704 1705 if (Callbacks) 1706 Callbacks->Endif(); 1707} 1708 1709/// HandleElseDirective - Implements the #else directive. 1710/// 1711void Preprocessor::HandleElseDirective(Token &Result) { 1712 ++NumElse; 1713 1714 // #else directive in a non-skipping conditional... start skipping. 1715 CheckEndOfDirective("else"); 1716 1717 PPConditionalInfo CI; 1718 if (CurPPLexer->popConditionalLevel(CI)) { 1719 Diag(Result, diag::pp_err_else_without_if); 1720 return; 1721 } 1722 1723 // If this is a top-level #else, inform the MIOpt. 1724 if (CurPPLexer->getConditionalStackDepth() == 0) 1725 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1726 1727 // If this is a #else with a #else before it, report the error. 1728 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1729 1730 // Finally, skip the rest of the contents of this block. 1731 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1732 /*FoundElse*/true); 1733 1734 if (Callbacks) 1735 Callbacks->Else(); 1736} 1737 1738/// HandleElifDirective - Implements the #elif directive. 1739/// 1740void Preprocessor::HandleElifDirective(Token &ElifToken) { 1741 ++NumElse; 1742 1743 // #elif directive in a non-skipping conditional... start skipping. 1744 // We don't care what the condition is, because we will always skip it (since 1745 // the block immediately before it was included). 1746 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1747 DiscardUntilEndOfDirective(); 1748 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1749 1750 PPConditionalInfo CI; 1751 if (CurPPLexer->popConditionalLevel(CI)) { 1752 Diag(ElifToken, diag::pp_err_elif_without_if); 1753 return; 1754 } 1755 1756 // If this is a top-level #elif, inform the MIOpt. 1757 if (CurPPLexer->getConditionalStackDepth() == 0) 1758 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1759 1760 // If this is a #elif with a #else before it, report the error. 1761 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1762 1763 // Finally, skip the rest of the contents of this block. 1764 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1765 /*FoundElse*/CI.FoundElse); 1766 1767 if (Callbacks) 1768 Callbacks->Elif(SourceRange(ConditionalBegin, ConditionalEnd)); 1769} 1770