PPDirectives.cpp revision 08985b91e5a87e7731ba24c675f2dc9f9cb645d4
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()); 807 808 // If there is no include loc (main file) or if the include loc is in a 809 // different physical file, then we aren't in a "1" line marker flag region. 810 SourceLocation IncLoc = PLoc.getIncludeLoc(); 811 if (IncLoc.isInvalid() || 812 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 813 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 814 PP.DiscardUntilEndOfDirective(); 815 return true; 816 } 817 818 PP.Lex(FlagTok); 819 if (FlagTok.is(tok::eom)) return false; 820 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 821 return true; 822 } 823 824 // We must have 3 if there are still flags. 825 if (FlagVal != 3) { 826 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 827 PP.DiscardUntilEndOfDirective(); 828 return true; 829 } 830 831 IsSystemHeader = true; 832 833 PP.Lex(FlagTok); 834 if (FlagTok.is(tok::eom)) return false; 835 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 836 return true; 837 838 // We must have 4 if there is yet another flag. 839 if (FlagVal != 4) { 840 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 841 PP.DiscardUntilEndOfDirective(); 842 return true; 843 } 844 845 IsExternCHeader = true; 846 847 PP.Lex(FlagTok); 848 if (FlagTok.is(tok::eom)) return false; 849 850 // There are no more valid flags here. 851 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 852 PP.DiscardUntilEndOfDirective(); 853 return true; 854} 855 856/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 857/// one of the following forms: 858/// 859/// # 42 860/// # 42 "file" ('1' | '2')? 861/// # 42 "file" ('1' | '2')? '3' '4'? 862/// 863void Preprocessor::HandleDigitDirective(Token &DigitTok) { 864 // Validate the number and convert it to an unsigned. GNU does not have a 865 // line # limit other than it fit in 32-bits. 866 unsigned LineNo; 867 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 868 *this)) 869 return; 870 871 Token StrTok; 872 Lex(StrTok); 873 874 bool IsFileEntry = false, IsFileExit = false; 875 bool IsSystemHeader = false, IsExternCHeader = false; 876 int FilenameID = -1; 877 878 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 879 // string followed by eom. 880 if (StrTok.is(tok::eom)) 881 ; // ok 882 else if (StrTok.isNot(tok::string_literal)) { 883 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 884 return DiscardUntilEndOfDirective(); 885 } else { 886 // Parse and validate the string, converting it into a unique ID. 887 StringLiteralParser Literal(&StrTok, 1, *this); 888 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 889 if (Literal.hadError) 890 return DiscardUntilEndOfDirective(); 891 if (Literal.Pascal) { 892 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 893 return DiscardUntilEndOfDirective(); 894 } 895 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 896 Literal.GetStringLength()); 897 898 // If a filename was present, read any flags that are present. 899 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 900 IsSystemHeader, IsExternCHeader, *this)) 901 return; 902 } 903 904 // Create a line note with this information. 905 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 906 IsFileEntry, IsFileExit, 907 IsSystemHeader, IsExternCHeader); 908 909 // If the preprocessor has callbacks installed, notify them of the #line 910 // change. This is used so that the line marker comes out in -E mode for 911 // example. 912 if (Callbacks) { 913 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 914 if (IsFileEntry) 915 Reason = PPCallbacks::EnterFile; 916 else if (IsFileExit) 917 Reason = PPCallbacks::ExitFile; 918 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 919 if (IsExternCHeader) 920 FileKind = SrcMgr::C_ExternCSystem; 921 else if (IsSystemHeader) 922 FileKind = SrcMgr::C_System; 923 924 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 925 } 926} 927 928 929/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 930/// 931void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 932 bool isWarning) { 933 // PTH doesn't emit #warning or #error directives. 934 if (CurPTHLexer) 935 return CurPTHLexer->DiscardToEndOfLine(); 936 937 // Read the rest of the line raw. We do this because we don't want macros 938 // to be expanded and we don't require that the tokens be valid preprocessing 939 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 940 // collapse multiple consequtive white space between tokens, but this isn't 941 // specified by the standard. 942 std::string Message = CurLexer->ReadToEndOfLine(); 943 if (isWarning) 944 Diag(Tok, diag::pp_hash_warning) << Message; 945 else 946 Diag(Tok, diag::err_pp_hash_error) << Message; 947} 948 949/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 950/// 951void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 952 // Yes, this directive is an extension. 953 Diag(Tok, diag::ext_pp_ident_directive); 954 955 // Read the string argument. 956 Token StrTok; 957 Lex(StrTok); 958 959 // If the token kind isn't a string, it's a malformed directive. 960 if (StrTok.isNot(tok::string_literal) && 961 StrTok.isNot(tok::wide_string_literal)) { 962 Diag(StrTok, diag::err_pp_malformed_ident); 963 if (StrTok.isNot(tok::eom)) 964 DiscardUntilEndOfDirective(); 965 return; 966 } 967 968 // Verify that there is nothing after the string, other than EOM. 969 CheckEndOfDirective("ident"); 970 971 if (Callbacks) { 972 bool Invalid = false; 973 std::string Str = getSpelling(StrTok, &Invalid); 974 if (!Invalid) 975 Callbacks->Ident(Tok.getLocation(), Str); 976 } 977} 978 979//===----------------------------------------------------------------------===// 980// Preprocessor Include Directive Handling. 981//===----------------------------------------------------------------------===// 982 983/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 984/// checked and spelled filename, e.g. as an operand of #include. This returns 985/// true if the input filename was in <>'s or false if it were in ""'s. The 986/// caller is expected to provide a buffer that is large enough to hold the 987/// spelling of the filename, but is also expected to handle the case when 988/// this method decides to use a different buffer. 989bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 990 llvm::StringRef &Buffer) { 991 // Get the text form of the filename. 992 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 993 994 // Make sure the filename is <x> or "x". 995 bool isAngled; 996 if (Buffer[0] == '<') { 997 if (Buffer.back() != '>') { 998 Diag(Loc, diag::err_pp_expects_filename); 999 Buffer = llvm::StringRef(); 1000 return true; 1001 } 1002 isAngled = true; 1003 } else if (Buffer[0] == '"') { 1004 if (Buffer.back() != '"') { 1005 Diag(Loc, diag::err_pp_expects_filename); 1006 Buffer = llvm::StringRef(); 1007 return true; 1008 } 1009 isAngled = false; 1010 } else { 1011 Diag(Loc, diag::err_pp_expects_filename); 1012 Buffer = llvm::StringRef(); 1013 return true; 1014 } 1015 1016 // Diagnose #include "" as invalid. 1017 if (Buffer.size() <= 2) { 1018 Diag(Loc, diag::err_pp_empty_filename); 1019 Buffer = llvm::StringRef(); 1020 return true; 1021 } 1022 1023 // Skip the brackets. 1024 Buffer = Buffer.substr(1, Buffer.size()-2); 1025 return isAngled; 1026} 1027 1028/// ConcatenateIncludeName - Handle cases where the #include name is expanded 1029/// from a macro as multiple tokens, which need to be glued together. This 1030/// occurs for code like: 1031/// #define FOO <a/b.h> 1032/// #include FOO 1033/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1034/// 1035/// This code concatenates and consumes tokens up to the '>' token. It returns 1036/// false if the > was found, otherwise it returns true if it finds and consumes 1037/// the EOM marker. 1038bool Preprocessor::ConcatenateIncludeName( 1039 llvm::SmallString<128> &FilenameBuffer, 1040 SourceLocation &End) { 1041 Token CurTok; 1042 1043 Lex(CurTok); 1044 while (CurTok.isNot(tok::eom)) { 1045 End = CurTok.getLocation(); 1046 1047 // Append the spelling of this token to the buffer. If there was a space 1048 // before it, add it now. 1049 if (CurTok.hasLeadingSpace()) 1050 FilenameBuffer.push_back(' '); 1051 1052 // Get the spelling of the token, directly into FilenameBuffer if possible. 1053 unsigned PreAppendSize = FilenameBuffer.size(); 1054 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1055 1056 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1057 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1058 1059 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1060 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1061 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1062 1063 // Resize FilenameBuffer to the correct size. 1064 if (CurTok.getLength() != ActualLen) 1065 FilenameBuffer.resize(PreAppendSize+ActualLen); 1066 1067 // If we found the '>' marker, return success. 1068 if (CurTok.is(tok::greater)) 1069 return false; 1070 1071 Lex(CurTok); 1072 } 1073 1074 // If we hit the eom marker, emit an error and return true so that the caller 1075 // knows the EOM has been read. 1076 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1077 return true; 1078} 1079 1080/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1081/// file to be included from the lexer, then include it! This is a common 1082/// routine with functionality shared between #include, #include_next and 1083/// #import. LookupFrom is set when this is a #include_next directive, it 1084/// specifies the file to start searching from. 1085void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc, 1086 Token &IncludeTok, 1087 const DirectoryLookup *LookupFrom, 1088 bool isImport) { 1089 1090 Token FilenameTok; 1091 CurPPLexer->LexIncludeFilename(FilenameTok); 1092 1093 // Reserve a buffer to get the spelling. 1094 llvm::SmallString<128> FilenameBuffer; 1095 llvm::StringRef Filename; 1096 SourceLocation End; 1097 1098 switch (FilenameTok.getKind()) { 1099 case tok::eom: 1100 // If the token kind is EOM, the error has already been diagnosed. 1101 return; 1102 1103 case tok::angle_string_literal: 1104 case tok::string_literal: 1105 Filename = getSpelling(FilenameTok, FilenameBuffer); 1106 End = FilenameTok.getLocation(); 1107 break; 1108 1109 case tok::less: 1110 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1111 // case, glue the tokens together into FilenameBuffer and interpret those. 1112 FilenameBuffer.push_back('<'); 1113 if (ConcatenateIncludeName(FilenameBuffer, End)) 1114 return; // Found <eom> but no ">"? Diagnostic already emitted. 1115 Filename = FilenameBuffer.str(); 1116 break; 1117 default: 1118 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1119 DiscardUntilEndOfDirective(); 1120 return; 1121 } 1122 1123 bool isAngled = 1124 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1125 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1126 // error. 1127 if (Filename.empty()) { 1128 DiscardUntilEndOfDirective(); 1129 return; 1130 } 1131 1132 // Verify that there is nothing after the filename, other than EOM. Note that 1133 // we allow macros that expand to nothing after the filename, because this 1134 // falls into the category of "#include pp-tokens new-line" specified in 1135 // C99 6.10.2p4. 1136 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1137 1138 // Check that we don't have infinite #include recursion. 1139 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1140 Diag(FilenameTok, diag::err_pp_include_too_deep); 1141 return; 1142 } 1143 1144 // Search include directories. 1145 const DirectoryLookup *CurDir; 1146 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1147 if (File == 0) { 1148 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1149 return; 1150 } 1151 1152 // Notify the callback object that we've seen an inclusion directive. 1153 if (Callbacks) 1154 Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled, File, 1155 End); 1156 1157 // The #included file will be considered to be a system header if either it is 1158 // in a system include directory, or if the #includer is a system include 1159 // header. 1160 SrcMgr::CharacteristicKind FileCharacter = 1161 std::max(HeaderInfo.getFileDirFlavor(File), 1162 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1163 1164 // Ask HeaderInfo if we should enter this #include file. If not, #including 1165 // this file will have no effect. 1166 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1167 if (Callbacks) 1168 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1169 return; 1170 } 1171 1172 // Look up the file, create a File ID for it. 1173 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1174 FileCharacter); 1175 if (FID.isInvalid()) { 1176 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1177 return; 1178 } 1179 1180 // Finally, if all is good, enter the new file! 1181 EnterSourceFile(FID, CurDir, FilenameTok.getLocation()); 1182} 1183 1184/// HandleIncludeNextDirective - Implements #include_next. 1185/// 1186void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc, 1187 Token &IncludeNextTok) { 1188 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1189 1190 // #include_next is like #include, except that we start searching after 1191 // the current found directory. If we can't do this, issue a 1192 // diagnostic. 1193 const DirectoryLookup *Lookup = CurDirLookup; 1194 if (isInPrimaryFile()) { 1195 Lookup = 0; 1196 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1197 } else if (Lookup == 0) { 1198 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1199 } else { 1200 // Start looking up in the next directory. 1201 ++Lookup; 1202 } 1203 1204 return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup); 1205} 1206 1207/// HandleImportDirective - Implements #import. 1208/// 1209void Preprocessor::HandleImportDirective(SourceLocation HashLoc, 1210 Token &ImportTok) { 1211 if (!Features.ObjC1) // #import is standard for ObjC. 1212 Diag(ImportTok, diag::ext_pp_import_directive); 1213 1214 return HandleIncludeDirective(HashLoc, ImportTok, 0, true); 1215} 1216 1217/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1218/// pseudo directive in the predefines buffer. This handles it by sucking all 1219/// tokens through the preprocessor and discarding them (only keeping the side 1220/// effects on the preprocessor). 1221void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc, 1222 Token &IncludeMacrosTok) { 1223 // This directive should only occur in the predefines buffer. If not, emit an 1224 // error and reject it. 1225 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1226 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1227 Diag(IncludeMacrosTok.getLocation(), 1228 diag::pp_include_macros_out_of_predefines); 1229 DiscardUntilEndOfDirective(); 1230 return; 1231 } 1232 1233 // Treat this as a normal #include for checking purposes. If this is 1234 // successful, it will push a new lexer onto the include stack. 1235 HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false); 1236 1237 Token TmpTok; 1238 do { 1239 Lex(TmpTok); 1240 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1241 } while (TmpTok.isNot(tok::hashhash)); 1242} 1243 1244//===----------------------------------------------------------------------===// 1245// Preprocessor Macro Directive Handling. 1246//===----------------------------------------------------------------------===// 1247 1248/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1249/// definition has just been read. Lex the rest of the arguments and the 1250/// closing ), updating MI with what we learn. Return true if an error occurs 1251/// parsing the arg list. 1252bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1253 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1254 1255 Token Tok; 1256 while (1) { 1257 LexUnexpandedToken(Tok); 1258 switch (Tok.getKind()) { 1259 case tok::r_paren: 1260 // Found the end of the argument list. 1261 if (Arguments.empty()) // #define FOO() 1262 return false; 1263 // Otherwise we have #define FOO(A,) 1264 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1265 return true; 1266 case tok::ellipsis: // #define X(... -> C99 varargs 1267 // Warn if use of C99 feature in non-C99 mode. 1268 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1269 1270 // Lex the token after the identifier. 1271 LexUnexpandedToken(Tok); 1272 if (Tok.isNot(tok::r_paren)) { 1273 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1274 return true; 1275 } 1276 // Add the __VA_ARGS__ identifier as an argument. 1277 Arguments.push_back(Ident__VA_ARGS__); 1278 MI->setIsC99Varargs(); 1279 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1280 return false; 1281 case tok::eom: // #define X( 1282 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1283 return true; 1284 default: 1285 // Handle keywords and identifiers here to accept things like 1286 // #define Foo(for) for. 1287 IdentifierInfo *II = Tok.getIdentifierInfo(); 1288 if (II == 0) { 1289 // #define X(1 1290 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1291 return true; 1292 } 1293 1294 // If this is already used as an argument, it is used multiple times (e.g. 1295 // #define X(A,A. 1296 if (std::find(Arguments.begin(), Arguments.end(), II) != 1297 Arguments.end()) { // C99 6.10.3p6 1298 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1299 return true; 1300 } 1301 1302 // Add the argument to the macro info. 1303 Arguments.push_back(II); 1304 1305 // Lex the token after the identifier. 1306 LexUnexpandedToken(Tok); 1307 1308 switch (Tok.getKind()) { 1309 default: // #define X(A B 1310 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1311 return true; 1312 case tok::r_paren: // #define X(A) 1313 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1314 return false; 1315 case tok::comma: // #define X(A, 1316 break; 1317 case tok::ellipsis: // #define X(A... -> GCC extension 1318 // Diagnose extension. 1319 Diag(Tok, diag::ext_named_variadic_macro); 1320 1321 // Lex the token after the identifier. 1322 LexUnexpandedToken(Tok); 1323 if (Tok.isNot(tok::r_paren)) { 1324 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1325 return true; 1326 } 1327 1328 MI->setIsGNUVarargs(); 1329 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1330 return false; 1331 } 1332 } 1333 } 1334} 1335 1336/// HandleDefineDirective - Implements #define. This consumes the entire macro 1337/// line then lets the caller lex the next real token. 1338void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1339 ++NumDefined; 1340 1341 Token MacroNameTok; 1342 ReadMacroName(MacroNameTok, 1); 1343 1344 // Error reading macro name? If so, diagnostic already issued. 1345 if (MacroNameTok.is(tok::eom)) 1346 return; 1347 1348 Token LastTok = MacroNameTok; 1349 1350 // If we are supposed to keep comments in #defines, reenable comment saving 1351 // mode. 1352 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1353 1354 // Create the new macro. 1355 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1356 1357 Token Tok; 1358 LexUnexpandedToken(Tok); 1359 1360 // If this is a function-like macro definition, parse the argument list, 1361 // marking each of the identifiers as being used as macro arguments. Also, 1362 // check other constraints on the first token of the macro body. 1363 if (Tok.is(tok::eom)) { 1364 // If there is no body to this macro, we have no special handling here. 1365 } else if (Tok.hasLeadingSpace()) { 1366 // This is a normal token with leading space. Clear the leading space 1367 // marker on the first token to get proper expansion. 1368 Tok.clearFlag(Token::LeadingSpace); 1369 } else if (Tok.is(tok::l_paren)) { 1370 // This is a function-like macro definition. Read the argument list. 1371 MI->setIsFunctionLike(); 1372 if (ReadMacroDefinitionArgList(MI)) { 1373 // Forget about MI. 1374 ReleaseMacroInfo(MI); 1375 // Throw away the rest of the line. 1376 if (CurPPLexer->ParsingPreprocessorDirective) 1377 DiscardUntilEndOfDirective(); 1378 return; 1379 } 1380 1381 // If this is a definition of a variadic C99 function-like macro, not using 1382 // the GNU named varargs extension, enabled __VA_ARGS__. 1383 1384 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1385 // This gets unpoisoned where it is allowed. 1386 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1387 if (MI->isC99Varargs()) 1388 Ident__VA_ARGS__->setIsPoisoned(false); 1389 1390 // Read the first token after the arg list for down below. 1391 LexUnexpandedToken(Tok); 1392 } else if (Features.C99) { 1393 // C99 requires whitespace between the macro definition and the body. Emit 1394 // a diagnostic for something like "#define X+". 1395 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1396 } else { 1397 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1398 // first character of a replacement list is not a character required by 1399 // subclause 5.2.1, then there shall be white-space separation between the 1400 // identifier and the replacement list.". 5.2.1 lists this set: 1401 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1402 // is irrelevant here. 1403 bool isInvalid = false; 1404 if (Tok.is(tok::at)) // @ is not in the list above. 1405 isInvalid = true; 1406 else if (Tok.is(tok::unknown)) { 1407 // If we have an unknown token, it is something strange like "`". Since 1408 // all of valid characters would have lexed into a single character 1409 // token of some sort, we know this is not a valid case. 1410 isInvalid = true; 1411 } 1412 if (isInvalid) 1413 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1414 else 1415 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1416 } 1417 1418 if (!Tok.is(tok::eom)) 1419 LastTok = Tok; 1420 1421 // Read the rest of the macro body. 1422 if (MI->isObjectLike()) { 1423 // Object-like macros are very simple, just read their body. 1424 while (Tok.isNot(tok::eom)) { 1425 LastTok = Tok; 1426 MI->AddTokenToBody(Tok); 1427 // Get the next token of the macro. 1428 LexUnexpandedToken(Tok); 1429 } 1430 1431 } else { 1432 // Otherwise, read the body of a function-like macro. While we are at it, 1433 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1434 // parameters in function-like macro expansions. 1435 while (Tok.isNot(tok::eom)) { 1436 LastTok = Tok; 1437 1438 if (Tok.isNot(tok::hash)) { 1439 MI->AddTokenToBody(Tok); 1440 1441 // Get the next token of the macro. 1442 LexUnexpandedToken(Tok); 1443 continue; 1444 } 1445 1446 // Get the next token of the macro. 1447 LexUnexpandedToken(Tok); 1448 1449 // Check for a valid macro arg identifier. 1450 if (Tok.getIdentifierInfo() == 0 || 1451 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1452 1453 // If this is assembler-with-cpp mode, we accept random gibberish after 1454 // the '#' because '#' is often a comment character. However, change 1455 // the kind of the token to tok::unknown so that the preprocessor isn't 1456 // confused. 1457 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1458 LastTok.setKind(tok::unknown); 1459 } else { 1460 Diag(Tok, diag::err_pp_stringize_not_parameter); 1461 ReleaseMacroInfo(MI); 1462 1463 // Disable __VA_ARGS__ again. 1464 Ident__VA_ARGS__->setIsPoisoned(true); 1465 return; 1466 } 1467 } 1468 1469 // Things look ok, add the '#' and param name tokens to the macro. 1470 MI->AddTokenToBody(LastTok); 1471 MI->AddTokenToBody(Tok); 1472 LastTok = Tok; 1473 1474 // Get the next token of the macro. 1475 LexUnexpandedToken(Tok); 1476 } 1477 } 1478 1479 1480 // Disable __VA_ARGS__ again. 1481 Ident__VA_ARGS__->setIsPoisoned(true); 1482 1483 // Check that there is no paste (##) operator at the begining or end of the 1484 // replacement list. 1485 unsigned NumTokens = MI->getNumTokens(); 1486 if (NumTokens != 0) { 1487 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1488 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1489 ReleaseMacroInfo(MI); 1490 return; 1491 } 1492 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1493 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1494 ReleaseMacroInfo(MI); 1495 return; 1496 } 1497 } 1498 1499 // If this is the primary source file, remember that this macro hasn't been 1500 // used yet. 1501 if (isInPrimaryFile()) 1502 MI->setIsUsed(false); 1503 1504 MI->setDefinitionEndLoc(LastTok.getLocation()); 1505 1506 // Finally, if this identifier already had a macro defined for it, verify that 1507 // the macro bodies are identical and free the old definition. 1508 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1509 // It is very common for system headers to have tons of macro redefinitions 1510 // and for warnings to be disabled in system headers. If this is the case, 1511 // then don't bother calling MacroInfo::isIdenticalTo. 1512 if (!getDiagnostics().getSuppressSystemWarnings() || 1513 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1514 if (!OtherMI->isUsed()) 1515 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1516 1517 // Macros must be identical. This means all tokens and whitespace 1518 // separation must be the same. C99 6.10.3.2. 1519 if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 1520 !MI->isIdenticalTo(*OtherMI, *this)) { 1521 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1522 << MacroNameTok.getIdentifierInfo(); 1523 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1524 } 1525 } 1526 ReleaseMacroInfo(OtherMI); 1527 } 1528 1529 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1530 1531 // If the callbacks want to know, tell them about the macro definition. 1532 if (Callbacks) 1533 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1534} 1535 1536/// HandleUndefDirective - Implements #undef. 1537/// 1538void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1539 ++NumUndefined; 1540 1541 Token MacroNameTok; 1542 ReadMacroName(MacroNameTok, 2); 1543 1544 // Error reading macro name? If so, diagnostic already issued. 1545 if (MacroNameTok.is(tok::eom)) 1546 return; 1547 1548 // Check to see if this is the last token on the #undef line. 1549 CheckEndOfDirective("undef"); 1550 1551 // Okay, we finally have a valid identifier to undef. 1552 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1553 1554 // If the macro is not defined, this is a noop undef, just return. 1555 if (MI == 0) return; 1556 1557 if (!MI->isUsed()) 1558 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1559 1560 // If the callbacks want to know, tell them about the macro #undef. 1561 if (Callbacks) 1562 Callbacks->MacroUndefined(MacroNameTok.getLocation(), 1563 MacroNameTok.getIdentifierInfo(), MI); 1564 1565 // Free macro definition. 1566 ReleaseMacroInfo(MI); 1567 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1568} 1569 1570 1571//===----------------------------------------------------------------------===// 1572// Preprocessor Conditional Directive Handling. 1573//===----------------------------------------------------------------------===// 1574 1575/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1576/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1577/// if any tokens have been returned or pp-directives activated before this 1578/// #ifndef has been lexed. 1579/// 1580void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1581 bool ReadAnyTokensBeforeDirective) { 1582 ++NumIf; 1583 Token DirectiveTok = Result; 1584 1585 Token MacroNameTok; 1586 ReadMacroName(MacroNameTok); 1587 1588 // Error reading macro name? If so, diagnostic already issued. 1589 if (MacroNameTok.is(tok::eom)) { 1590 // Skip code until we get to #endif. This helps with recovery by not 1591 // emitting an error when the #endif is reached. 1592 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1593 /*Foundnonskip*/false, /*FoundElse*/false); 1594 return; 1595 } 1596 1597 // Check to see if this is the last token on the #if[n]def line. 1598 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1599 1600 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1601 MacroInfo *MI = getMacroInfo(MII); 1602 1603 if (CurPPLexer->getConditionalStackDepth() == 0) { 1604 // If the start of a top-level #ifdef and if the macro is not defined, 1605 // inform MIOpt that this might be the start of a proper include guard. 1606 // Otherwise it is some other form of unknown conditional which we can't 1607 // handle. 1608 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1609 assert(isIfndef && "#ifdef shouldn't reach here"); 1610 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1611 } else 1612 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1613 } 1614 1615 // If there is a macro, process it. 1616 if (MI) // Mark it used. 1617 MI->setIsUsed(true); 1618 1619 // Should we include the stuff contained by this directive? 1620 if (!MI == isIfndef) { 1621 // Yes, remember that we are inside a conditional, then lex the next token. 1622 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1623 /*wasskip*/false, /*foundnonskip*/true, 1624 /*foundelse*/false); 1625 } else { 1626 // No, skip the contents of this block. 1627 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1628 /*Foundnonskip*/false, 1629 /*FoundElse*/false); 1630 } 1631 1632 if (Callbacks) { 1633 if (isIfndef) 1634 Callbacks->Ifndef(MacroNameTok.getLocation(), MII); 1635 else 1636 Callbacks->Ifdef(MacroNameTok.getLocation(), MII); 1637 } 1638} 1639 1640/// HandleIfDirective - Implements the #if directive. 1641/// 1642void Preprocessor::HandleIfDirective(Token &IfToken, 1643 bool ReadAnyTokensBeforeDirective) { 1644 ++NumIf; 1645 1646 // Parse and evaluate the conditional expression. 1647 IdentifierInfo *IfNDefMacro = 0; 1648 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1649 const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1650 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1651 1652 // If this condition is equivalent to #ifndef X, and if this is the first 1653 // directive seen, handle it for the multiple-include optimization. 1654 if (CurPPLexer->getConditionalStackDepth() == 0) { 1655 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1656 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1657 else 1658 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1659 } 1660 1661 // Should we include the stuff contained by this directive? 1662 if (ConditionalTrue) { 1663 // Yes, remember that we are inside a conditional, then lex the next token. 1664 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1665 /*foundnonskip*/true, /*foundelse*/false); 1666 } else { 1667 // No, skip the contents of this block. 1668 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1669 /*FoundElse*/false); 1670 } 1671 1672 if (Callbacks) 1673 Callbacks->If(SourceRange(ConditionalBegin, ConditionalEnd)); 1674} 1675 1676/// HandleEndifDirective - Implements the #endif directive. 1677/// 1678void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1679 ++NumEndif; 1680 1681 // Check that this is the whole directive. 1682 CheckEndOfDirective("endif"); 1683 1684 PPConditionalInfo CondInfo; 1685 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1686 // No conditionals on the stack: this is an #endif without an #if. 1687 Diag(EndifToken, diag::err_pp_endif_without_if); 1688 return; 1689 } 1690 1691 // If this the end of a top-level #endif, inform MIOpt. 1692 if (CurPPLexer->getConditionalStackDepth() == 0) 1693 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1694 1695 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1696 "This code should only be reachable in the non-skipping case!"); 1697 1698 if (Callbacks) 1699 Callbacks->Endif(); 1700} 1701 1702/// HandleElseDirective - Implements the #else directive. 1703/// 1704void Preprocessor::HandleElseDirective(Token &Result) { 1705 ++NumElse; 1706 1707 // #else directive in a non-skipping conditional... start skipping. 1708 CheckEndOfDirective("else"); 1709 1710 PPConditionalInfo CI; 1711 if (CurPPLexer->popConditionalLevel(CI)) { 1712 Diag(Result, diag::pp_err_else_without_if); 1713 return; 1714 } 1715 1716 // If this is a top-level #else, inform the MIOpt. 1717 if (CurPPLexer->getConditionalStackDepth() == 0) 1718 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1719 1720 // If this is a #else with a #else before it, report the error. 1721 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1722 1723 // Finally, skip the rest of the contents of this block. 1724 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1725 /*FoundElse*/true); 1726 1727 if (Callbacks) 1728 Callbacks->Else(); 1729} 1730 1731/// HandleElifDirective - Implements the #elif directive. 1732/// 1733void Preprocessor::HandleElifDirective(Token &ElifToken) { 1734 ++NumElse; 1735 1736 // #elif directive in a non-skipping conditional... start skipping. 1737 // We don't care what the condition is, because we will always skip it (since 1738 // the block immediately before it was included). 1739 const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation(); 1740 DiscardUntilEndOfDirective(); 1741 const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation(); 1742 1743 PPConditionalInfo CI; 1744 if (CurPPLexer->popConditionalLevel(CI)) { 1745 Diag(ElifToken, diag::pp_err_elif_without_if); 1746 return; 1747 } 1748 1749 // If this is a top-level #elif, inform the MIOpt. 1750 if (CurPPLexer->getConditionalStackDepth() == 0) 1751 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1752 1753 // If this is a #elif with a #else before it, report the error. 1754 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1755 1756 // Finally, skip the rest of the contents of this block. 1757 SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1758 /*FoundElse*/CI.FoundElse); 1759 1760 if (Callbacks) 1761 Callbacks->Elif(SourceRange(ConditionalBegin, ConditionalEnd)); 1762} 1763