PPDirectives.cpp revision 52a4ff6eae5d5e303c8b43022211e6d3ed429254
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 MacroInfo *MI; 32 33 if (!MICache.empty()) { 34 MI = MICache.back(); 35 MICache.pop_back(); 36 } else { 37 MacroInfoChain *MIChain = BP.Allocate<MacroInfoChain>(); 38 MIChain->Next = MIChainHead; 39 MIChainHead = MIChain; 40 MI = &(MIChain->MI); 41 } 42 return MI; 43} 44 45MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 46 MacroInfo *MI = AllocateMacroInfo(); 47 new (MI) MacroInfo(L); 48 return MI; 49} 50 51MacroInfo *Preprocessor::CloneMacroInfo(const MacroInfo &MacroToClone) { 52 MacroInfo *MI = AllocateMacroInfo(); 53 new (MI) MacroInfo(MacroToClone, BP); 54 return MI; 55} 56 57/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 58/// be reused for allocating new MacroInfo objects. 59void Preprocessor::ReleaseMacroInfo(MacroInfo *MI) { 60 MICache.push_back(MI); 61 MI->FreeArgumentList(); 62} 63 64 65/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 66/// current line until the tok::eom token is found. 67void Preprocessor::DiscardUntilEndOfDirective() { 68 Token Tmp; 69 do { 70 LexUnexpandedToken(Tmp); 71 } while (Tmp.isNot(tok::eom)); 72} 73 74/// ReadMacroName - Lex and validate a macro name, which occurs after a 75/// #define or #undef. This sets the token kind to eom and discards the rest 76/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 77/// this is due to a a #define, 2 if #undef directive, 0 if it is something 78/// else (e.g. #ifdef). 79void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 80 // Read the token, don't allow macro expansion on it. 81 LexUnexpandedToken(MacroNameTok); 82 83 if (MacroNameTok.is(tok::code_completion)) { 84 if (CodeComplete) 85 CodeComplete->CodeCompleteMacroName(isDefineUndef == 1); 86 LexUnexpandedToken(MacroNameTok); 87 return; 88 } 89 90 // Missing macro name? 91 if (MacroNameTok.is(tok::eom)) { 92 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 93 return; 94 } 95 96 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 97 if (II == 0) { 98 bool Invalid = false; 99 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 100 if (Invalid) 101 return; 102 103 const IdentifierInfo &Info = Identifiers.get(Spelling); 104 if (Info.isCPlusPlusOperatorKeyword()) 105 // C++ 2.5p2: Alternative tokens behave the same as its primary token 106 // except for their spellings. 107 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 108 else 109 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 110 // Fall through on error. 111 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 112 // Error if defining "defined": C99 6.10.8.4. 113 Diag(MacroNameTok, diag::err_defined_macro_name); 114 } else if (isDefineUndef && II->hasMacroDefinition() && 115 getMacroInfo(II)->isBuiltinMacro()) { 116 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 117 if (isDefineUndef == 1) 118 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 119 else 120 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 121 } else { 122 // Okay, we got a good identifier node. Return it. 123 return; 124 } 125 126 // Invalid macro name, read and discard the rest of the line. Then set the 127 // token kind to tok::eom. 128 MacroNameTok.setKind(tok::eom); 129 return DiscardUntilEndOfDirective(); 130} 131 132/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 133/// not, emit a diagnostic and consume up until the eom. If EnableMacros is 134/// true, then we consider macros that expand to zero tokens as being ok. 135void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 136 Token Tmp; 137 // Lex unexpanded tokens for most directives: macros might expand to zero 138 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 139 // #line) allow empty macros. 140 if (EnableMacros) 141 Lex(Tmp); 142 else 143 LexUnexpandedToken(Tmp); 144 145 // There should be no tokens after the directive, but we allow them as an 146 // extension. 147 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 148 LexUnexpandedToken(Tmp); 149 150 if (Tmp.isNot(tok::eom)) { 151 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 152 // because it is more trouble than it is worth to insert /**/ and check that 153 // there is no /**/ in the range also. 154 FixItHint Hint; 155 if (Features.GNUMode || Features.C99 || Features.CPlusPlus) 156 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 157 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 158 DiscardUntilEndOfDirective(); 159 } 160} 161 162 163 164/// SkipExcludedConditionalBlock - We just read a #if or related directive and 165/// decided that the subsequent tokens are in the #if'd out portion of the 166/// file. Lex the rest of the file, until we see an #endif. If 167/// FoundNonSkipPortion is true, then we have already emitted code for part of 168/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 169/// is true, then #else directives are ok, if not, then we have already seen one 170/// so a #else directive is a duplicate. When this returns, the caller can lex 171/// the first valid token. 172void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 173 bool FoundNonSkipPortion, 174 bool FoundElse) { 175 ++NumSkipped; 176 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 177 178 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 179 FoundNonSkipPortion, FoundElse); 180 181 if (CurPTHLexer) { 182 PTHSkipExcludedConditionalBlock(); 183 return; 184 } 185 186 // Enter raw mode to disable identifier lookup (and thus macro expansion), 187 // disabling warnings, etc. 188 CurPPLexer->LexingRawMode = true; 189 Token Tok; 190 while (1) { 191 CurLexer->Lex(Tok); 192 193 if (Tok.is(tok::code_completion)) { 194 if (CodeComplete) 195 CodeComplete->CodeCompleteInConditionalExclusion(); 196 continue; 197 } 198 199 // If this is the end of the buffer, we have an error. 200 if (Tok.is(tok::eof)) { 201 // Emit errors for each unterminated conditional on the stack, including 202 // the current one. 203 while (!CurPPLexer->ConditionalStack.empty()) { 204 if (!isCodeCompletionFile(Tok.getLocation())) 205 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 206 diag::err_pp_unterminated_conditional); 207 CurPPLexer->ConditionalStack.pop_back(); 208 } 209 210 // Just return and let the caller lex after this #include. 211 break; 212 } 213 214 // If this token is not a preprocessor directive, just skip it. 215 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 216 continue; 217 218 // We just parsed a # character at the start of a line, so we're in 219 // directive mode. Tell the lexer this so any newlines we see will be 220 // converted into an EOM token (this terminates the macro). 221 CurPPLexer->ParsingPreprocessorDirective = true; 222 if (CurLexer) CurLexer->SetCommentRetentionState(false); 223 224 225 // Read the next token, the directive flavor. 226 LexUnexpandedToken(Tok); 227 228 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 229 // something bogus), skip it. 230 if (Tok.isNot(tok::identifier)) { 231 CurPPLexer->ParsingPreprocessorDirective = false; 232 // Restore comment saving mode. 233 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 234 continue; 235 } 236 237 // If the first letter isn't i or e, it isn't intesting to us. We know that 238 // this is safe in the face of spelling differences, because there is no way 239 // to spell an i/e in a strange way that is another letter. Skipping this 240 // allows us to avoid looking up the identifier info for #define/#undef and 241 // other common directives. 242 bool Invalid = false; 243 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation(), 244 &Invalid); 245 if (Invalid) 246 return; 247 248 char FirstChar = RawCharData[0]; 249 if (FirstChar >= 'a' && FirstChar <= 'z' && 250 FirstChar != 'i' && FirstChar != 'e') { 251 CurPPLexer->ParsingPreprocessorDirective = false; 252 // Restore comment saving mode. 253 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 254 continue; 255 } 256 257 // Get the identifier name without trigraphs or embedded newlines. Note 258 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 259 // when skipping. 260 char DirectiveBuf[20]; 261 llvm::StringRef Directive; 262 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 263 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 264 } else { 265 std::string DirectiveStr = getSpelling(Tok); 266 unsigned IdLen = DirectiveStr.size(); 267 if (IdLen >= 20) { 268 CurPPLexer->ParsingPreprocessorDirective = false; 269 // Restore comment saving mode. 270 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 271 continue; 272 } 273 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 274 Directive = llvm::StringRef(DirectiveBuf, IdLen); 275 } 276 277 if (Directive.startswith("if")) { 278 llvm::StringRef Sub = Directive.substr(2); 279 if (Sub.empty() || // "if" 280 Sub == "def" || // "ifdef" 281 Sub == "ndef") { // "ifndef" 282 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 283 // bother parsing the condition. 284 DiscardUntilEndOfDirective(); 285 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 286 /*foundnonskip*/false, 287 /*fnddelse*/false); 288 } 289 } else if (Directive[0] == 'e') { 290 llvm::StringRef Sub = Directive.substr(1); 291 if (Sub == "ndif") { // "endif" 292 CheckEndOfDirective("endif"); 293 PPConditionalInfo CondInfo; 294 CondInfo.WasSkipping = true; // Silence bogus warning. 295 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 296 InCond = InCond; // Silence warning in no-asserts mode. 297 assert(!InCond && "Can't be skipping if not in a conditional!"); 298 299 // If we popped the outermost skipping block, we're done skipping! 300 if (!CondInfo.WasSkipping) 301 break; 302 } else if (Sub == "lse") { // "else". 303 // #else directive in a skipping conditional. If not in some other 304 // skipping conditional, and if #else hasn't already been seen, enter it 305 // as a non-skipping conditional. 306 DiscardUntilEndOfDirective(); // C99 6.10p4. 307 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 308 309 // If this is a #else with a #else before it, report the error. 310 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 311 312 // Note that we've seen a #else in this conditional. 313 CondInfo.FoundElse = true; 314 315 // If the conditional is at the top level, and the #if block wasn't 316 // entered, enter the #else block now. 317 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 318 CondInfo.FoundNonSkip = true; 319 break; 320 } 321 } else if (Sub == "lif") { // "elif". 322 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 323 324 bool ShouldEnter; 325 // If this is in a skipping block or if we're already handled this #if 326 // block, don't bother parsing the condition. 327 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 328 DiscardUntilEndOfDirective(); 329 ShouldEnter = false; 330 } else { 331 // Restore the value of LexingRawMode so that identifiers are 332 // looked up, etc, inside the #elif expression. 333 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 334 CurPPLexer->LexingRawMode = false; 335 IdentifierInfo *IfNDefMacro = 0; 336 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 337 CurPPLexer->LexingRawMode = true; 338 } 339 340 // If this is a #elif with a #else before it, report the error. 341 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 342 343 // If this condition is true, enter it! 344 if (ShouldEnter) { 345 CondInfo.FoundNonSkip = true; 346 break; 347 } 348 } 349 } 350 351 CurPPLexer->ParsingPreprocessorDirective = false; 352 // Restore comment saving mode. 353 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 354 } 355 356 // Finally, if we are out of the conditional (saw an #endif or ran off the end 357 // of the file, just stop skipping and return to lexing whatever came after 358 // the #if block. 359 CurPPLexer->LexingRawMode = false; 360} 361 362void Preprocessor::PTHSkipExcludedConditionalBlock() { 363 364 while (1) { 365 assert(CurPTHLexer); 366 assert(CurPTHLexer->LexingRawMode == false); 367 368 // Skip to the next '#else', '#elif', or #endif. 369 if (CurPTHLexer->SkipBlock()) { 370 // We have reached an #endif. Both the '#' and 'endif' tokens 371 // have been consumed by the PTHLexer. Just pop off the condition level. 372 PPConditionalInfo CondInfo; 373 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 374 InCond = InCond; // Silence warning in no-asserts mode. 375 assert(!InCond && "Can't be skipping if not in a conditional!"); 376 break; 377 } 378 379 // We have reached a '#else' or '#elif'. Lex the next token to get 380 // the directive flavor. 381 Token Tok; 382 LexUnexpandedToken(Tok); 383 384 // We can actually look up the IdentifierInfo here since we aren't in 385 // raw mode. 386 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 387 388 if (K == tok::pp_else) { 389 // #else: Enter the else condition. We aren't in a nested condition 390 // since we skip those. We're always in the one matching the last 391 // blocked we skipped. 392 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 393 // Note that we've seen a #else in this conditional. 394 CondInfo.FoundElse = true; 395 396 // If the #if block wasn't entered then enter the #else block now. 397 if (!CondInfo.FoundNonSkip) { 398 CondInfo.FoundNonSkip = true; 399 400 // Scan until the eom token. 401 CurPTHLexer->ParsingPreprocessorDirective = true; 402 DiscardUntilEndOfDirective(); 403 CurPTHLexer->ParsingPreprocessorDirective = false; 404 405 break; 406 } 407 408 // Otherwise skip this block. 409 continue; 410 } 411 412 assert(K == tok::pp_elif); 413 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 414 415 // If this is a #elif with a #else before it, report the error. 416 if (CondInfo.FoundElse) 417 Diag(Tok, diag::pp_err_elif_after_else); 418 419 // If this is in a skipping block or if we're already handled this #if 420 // block, don't bother parsing the condition. We just skip this block. 421 if (CondInfo.FoundNonSkip) 422 continue; 423 424 // Evaluate the condition of the #elif. 425 IdentifierInfo *IfNDefMacro = 0; 426 CurPTHLexer->ParsingPreprocessorDirective = true; 427 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 428 CurPTHLexer->ParsingPreprocessorDirective = false; 429 430 // If this condition is true, enter it! 431 if (ShouldEnter) { 432 CondInfo.FoundNonSkip = true; 433 break; 434 } 435 436 // Otherwise, skip this block and go to the next one. 437 continue; 438 } 439} 440 441/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 442/// return null on failure. isAngled indicates whether the file reference is 443/// for system #include's or not (i.e. using <> instead of ""). 444const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 445 bool isAngled, 446 const DirectoryLookup *FromDir, 447 const DirectoryLookup *&CurDir) { 448 // If the header lookup mechanism may be relative to the current file, pass in 449 // info about where the current file is. 450 const FileEntry *CurFileEnt = 0; 451 if (!FromDir) { 452 FileID FID = getCurrentFileLexer()->getFileID(); 453 CurFileEnt = SourceMgr.getFileEntryForID(FID); 454 455 // If there is no file entry associated with this file, it must be the 456 // predefines buffer. Any other file is not lexed with a normal lexer, so 457 // it won't be scanned for preprocessor directives. If we have the 458 // predefines buffer, resolve #include references (which come from the 459 // -include command line argument) as if they came from the main file, this 460 // affects file lookup etc. 461 if (CurFileEnt == 0) { 462 FID = SourceMgr.getMainFileID(); 463 CurFileEnt = SourceMgr.getFileEntryForID(FID); 464 } 465 } 466 467 // Do a standard file entry lookup. 468 CurDir = CurDirLookup; 469 const FileEntry *FE = 470 HeaderInfo.LookupFile(Filename, isAngled, FromDir, CurDir, CurFileEnt); 471 if (FE) return FE; 472 473 // Otherwise, see if this is a subframework header. If so, this is relative 474 // to one of the headers on the #include stack. Walk the list of the current 475 // headers on the #include stack and pass them to HeaderInfo. 476 if (IsFileLexer()) { 477 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 478 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 479 return FE; 480 } 481 482 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 483 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 484 if (IsFileLexer(ISEntry)) { 485 if ((CurFileEnt = 486 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 487 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 488 return FE; 489 } 490 } 491 492 // Otherwise, we really couldn't find the file. 493 return 0; 494} 495 496 497//===----------------------------------------------------------------------===// 498// Preprocessor Directive Handling. 499//===----------------------------------------------------------------------===// 500 501/// HandleDirective - This callback is invoked when the lexer sees a # token 502/// at the start of a line. This consumes the directive, modifies the 503/// lexer/preprocessor state, and advances the lexer(s) so that the next token 504/// read is the correct one. 505void Preprocessor::HandleDirective(Token &Result) { 506 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 507 508 // We just parsed a # character at the start of a line, so we're in directive 509 // mode. Tell the lexer this so any newlines we see will be converted into an 510 // EOM token (which terminates the directive). 511 CurPPLexer->ParsingPreprocessorDirective = true; 512 513 ++NumDirectives; 514 515 // We are about to read a token. For the multiple-include optimization FA to 516 // work, we have to remember if we had read any tokens *before* this 517 // pp-directive. 518 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 519 520 // Save the '#' token in case we need to return it later. 521 Token SavedHash = Result; 522 523 // Read the next token, the directive flavor. This isn't expanded due to 524 // C99 6.10.3p8. 525 LexUnexpandedToken(Result); 526 527 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 528 // #define A(x) #x 529 // A(abc 530 // #warning blah 531 // def) 532 // If so, the user is relying on non-portable behavior, emit a diagnostic. 533 if (InMacroArgs) 534 Diag(Result, diag::ext_embedded_directive); 535 536TryAgain: 537 switch (Result.getKind()) { 538 case tok::eom: 539 return; // null directive. 540 case tok::comment: 541 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 542 LexUnexpandedToken(Result); 543 goto TryAgain; 544 case tok::code_completion: 545 if (CodeComplete) 546 CodeComplete->CodeCompleteDirective( 547 CurPPLexer->getConditionalStackDepth() > 0); 548 return; 549 case tok::numeric_constant: // # 7 GNU line marker directive. 550 if (getLangOptions().AsmPreprocessor) 551 break; // # 4 is not a preprocessor directive in .S files. 552 return HandleDigitDirective(Result); 553 default: 554 IdentifierInfo *II = Result.getIdentifierInfo(); 555 if (II == 0) break; // Not an identifier. 556 557 // Ask what the preprocessor keyword ID is. 558 switch (II->getPPKeywordID()) { 559 default: break; 560 // C99 6.10.1 - Conditional Inclusion. 561 case tok::pp_if: 562 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 563 case tok::pp_ifdef: 564 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 565 case tok::pp_ifndef: 566 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 567 case tok::pp_elif: 568 return HandleElifDirective(Result); 569 case tok::pp_else: 570 return HandleElseDirective(Result); 571 case tok::pp_endif: 572 return HandleEndifDirective(Result); 573 574 // C99 6.10.2 - Source File Inclusion. 575 case tok::pp_include: 576 return HandleIncludeDirective(Result); // Handle #include. 577 case tok::pp___include_macros: 578 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 579 580 // C99 6.10.3 - Macro Replacement. 581 case tok::pp_define: 582 return HandleDefineDirective(Result); 583 case tok::pp_undef: 584 return HandleUndefDirective(Result); 585 586 // C99 6.10.4 - Line Control. 587 case tok::pp_line: 588 return HandleLineDirective(Result); 589 590 // C99 6.10.5 - Error Directive. 591 case tok::pp_error: 592 return HandleUserDiagnosticDirective(Result, false); 593 594 // C99 6.10.6 - Pragma Directive. 595 case tok::pp_pragma: 596 return HandlePragmaDirective(PIK_HashPragma); 597 598 // GNU Extensions. 599 case tok::pp_import: 600 return HandleImportDirective(Result); 601 case tok::pp_include_next: 602 return HandleIncludeNextDirective(Result); 603 604 case tok::pp_warning: 605 Diag(Result, diag::ext_pp_warning_directive); 606 return HandleUserDiagnosticDirective(Result, true); 607 case tok::pp_ident: 608 return HandleIdentSCCSDirective(Result); 609 case tok::pp_sccs: 610 return HandleIdentSCCSDirective(Result); 611 case tok::pp_assert: 612 //isExtension = true; // FIXME: implement #assert 613 break; 614 case tok::pp_unassert: 615 //isExtension = true; // FIXME: implement #unassert 616 break; 617 } 618 break; 619 } 620 621 // If this is a .S file, treat unknown # directives as non-preprocessor 622 // directives. This is important because # may be a comment or introduce 623 // various pseudo-ops. Just return the # token and push back the following 624 // token to be lexed next time. 625 if (getLangOptions().AsmPreprocessor) { 626 Token *Toks = new Token[2]; 627 // Return the # and the token after it. 628 Toks[0] = SavedHash; 629 Toks[1] = Result; 630 // Enter this token stream so that we re-lex the tokens. Make sure to 631 // enable macro expansion, in case the token after the # is an identifier 632 // that is expanded. 633 EnterTokenStream(Toks, 2, false, true); 634 return; 635 } 636 637 // If we reached here, the preprocessing token is not valid! 638 Diag(Result, diag::err_pp_invalid_directive); 639 640 // Read the rest of the PP line. 641 DiscardUntilEndOfDirective(); 642 643 // Okay, we're done parsing the directive. 644} 645 646/// GetLineValue - Convert a numeric token into an unsigned value, emitting 647/// Diagnostic DiagID if it is invalid, and returning the value in Val. 648static bool GetLineValue(Token &DigitTok, unsigned &Val, 649 unsigned DiagID, Preprocessor &PP) { 650 if (DigitTok.isNot(tok::numeric_constant)) { 651 PP.Diag(DigitTok, DiagID); 652 653 if (DigitTok.isNot(tok::eom)) 654 PP.DiscardUntilEndOfDirective(); 655 return true; 656 } 657 658 llvm::SmallString<64> IntegerBuffer; 659 IntegerBuffer.resize(DigitTok.getLength()); 660 const char *DigitTokBegin = &IntegerBuffer[0]; 661 bool Invalid = false; 662 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 663 if (Invalid) 664 return true; 665 666 // Verify that we have a simple digit-sequence, and compute the value. This 667 // is always a simple digit string computed in decimal, so we do this manually 668 // here. 669 Val = 0; 670 for (unsigned i = 0; i != ActualLength; ++i) { 671 if (!isdigit(DigitTokBegin[i])) { 672 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 673 diag::err_pp_line_digit_sequence); 674 PP.DiscardUntilEndOfDirective(); 675 return true; 676 } 677 678 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 679 if (NextVal < Val) { // overflow. 680 PP.Diag(DigitTok, DiagID); 681 PP.DiscardUntilEndOfDirective(); 682 return true; 683 } 684 Val = NextVal; 685 } 686 687 // Reject 0, this is needed both by #line numbers and flags. 688 if (Val == 0) { 689 PP.Diag(DigitTok, DiagID); 690 PP.DiscardUntilEndOfDirective(); 691 return true; 692 } 693 694 if (DigitTokBegin[0] == '0') 695 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 696 697 return false; 698} 699 700/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 701/// acceptable forms are: 702/// # line digit-sequence 703/// # line digit-sequence "s-char-sequence" 704void Preprocessor::HandleLineDirective(Token &Tok) { 705 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 706 // expanded. 707 Token DigitTok; 708 Lex(DigitTok); 709 710 // Validate the number and convert it to an unsigned. 711 unsigned LineNo; 712 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 713 return; 714 715 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 716 // number greater than 2147483647". C90 requires that the line # be <= 32767. 717 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 718 if (LineNo >= LineLimit) 719 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 720 721 int FilenameID = -1; 722 Token StrTok; 723 Lex(StrTok); 724 725 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 726 // string followed by eom. 727 if (StrTok.is(tok::eom)) 728 ; // ok 729 else if (StrTok.isNot(tok::string_literal)) { 730 Diag(StrTok, diag::err_pp_line_invalid_filename); 731 DiscardUntilEndOfDirective(); 732 return; 733 } else { 734 // Parse and validate the string, converting it into a unique ID. 735 StringLiteralParser Literal(&StrTok, 1, *this); 736 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 737 if (Literal.hadError) 738 return DiscardUntilEndOfDirective(); 739 if (Literal.Pascal) { 740 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 741 return DiscardUntilEndOfDirective(); 742 } 743 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 744 Literal.GetStringLength()); 745 746 // Verify that there is nothing after the string, other than EOM. Because 747 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 748 CheckEndOfDirective("line", true); 749 } 750 751 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 752 753 if (Callbacks) 754 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 755 PPCallbacks::RenameFile, 756 SrcMgr::C_User); 757} 758 759/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 760/// marker directive. 761static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 762 bool &IsSystemHeader, bool &IsExternCHeader, 763 Preprocessor &PP) { 764 unsigned FlagVal; 765 Token FlagTok; 766 PP.Lex(FlagTok); 767 if (FlagTok.is(tok::eom)) return false; 768 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 769 return true; 770 771 if (FlagVal == 1) { 772 IsFileEntry = true; 773 774 PP.Lex(FlagTok); 775 if (FlagTok.is(tok::eom)) return false; 776 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 777 return true; 778 } else if (FlagVal == 2) { 779 IsFileExit = true; 780 781 SourceManager &SM = PP.getSourceManager(); 782 // If we are leaving the current presumed file, check to make sure the 783 // presumed include stack isn't empty! 784 FileID CurFileID = 785 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 786 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 787 788 // If there is no include loc (main file) or if the include loc is in a 789 // different physical file, then we aren't in a "1" line marker flag region. 790 SourceLocation IncLoc = PLoc.getIncludeLoc(); 791 if (IncLoc.isInvalid() || 792 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 793 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 794 PP.DiscardUntilEndOfDirective(); 795 return true; 796 } 797 798 PP.Lex(FlagTok); 799 if (FlagTok.is(tok::eom)) return false; 800 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 801 return true; 802 } 803 804 // We must have 3 if there are still flags. 805 if (FlagVal != 3) { 806 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 807 PP.DiscardUntilEndOfDirective(); 808 return true; 809 } 810 811 IsSystemHeader = true; 812 813 PP.Lex(FlagTok); 814 if (FlagTok.is(tok::eom)) return false; 815 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 816 return true; 817 818 // We must have 4 if there is yet another flag. 819 if (FlagVal != 4) { 820 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 821 PP.DiscardUntilEndOfDirective(); 822 return true; 823 } 824 825 IsExternCHeader = true; 826 827 PP.Lex(FlagTok); 828 if (FlagTok.is(tok::eom)) return false; 829 830 // There are no more valid flags here. 831 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 832 PP.DiscardUntilEndOfDirective(); 833 return true; 834} 835 836/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 837/// one of the following forms: 838/// 839/// # 42 840/// # 42 "file" ('1' | '2')? 841/// # 42 "file" ('1' | '2')? '3' '4'? 842/// 843void Preprocessor::HandleDigitDirective(Token &DigitTok) { 844 // Validate the number and convert it to an unsigned. GNU does not have a 845 // line # limit other than it fit in 32-bits. 846 unsigned LineNo; 847 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 848 *this)) 849 return; 850 851 Token StrTok; 852 Lex(StrTok); 853 854 bool IsFileEntry = false, IsFileExit = false; 855 bool IsSystemHeader = false, IsExternCHeader = false; 856 int FilenameID = -1; 857 858 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 859 // string followed by eom. 860 if (StrTok.is(tok::eom)) 861 ; // ok 862 else if (StrTok.isNot(tok::string_literal)) { 863 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 864 return DiscardUntilEndOfDirective(); 865 } else { 866 // Parse and validate the string, converting it into a unique ID. 867 StringLiteralParser Literal(&StrTok, 1, *this); 868 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 869 if (Literal.hadError) 870 return DiscardUntilEndOfDirective(); 871 if (Literal.Pascal) { 872 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 873 return DiscardUntilEndOfDirective(); 874 } 875 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 876 Literal.GetStringLength()); 877 878 // If a filename was present, read any flags that are present. 879 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 880 IsSystemHeader, IsExternCHeader, *this)) 881 return; 882 } 883 884 // Create a line note with this information. 885 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 886 IsFileEntry, IsFileExit, 887 IsSystemHeader, IsExternCHeader); 888 889 // If the preprocessor has callbacks installed, notify them of the #line 890 // change. This is used so that the line marker comes out in -E mode for 891 // example. 892 if (Callbacks) { 893 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 894 if (IsFileEntry) 895 Reason = PPCallbacks::EnterFile; 896 else if (IsFileExit) 897 Reason = PPCallbacks::ExitFile; 898 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 899 if (IsExternCHeader) 900 FileKind = SrcMgr::C_ExternCSystem; 901 else if (IsSystemHeader) 902 FileKind = SrcMgr::C_System; 903 904 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 905 } 906} 907 908 909/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 910/// 911void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 912 bool isWarning) { 913 // PTH doesn't emit #warning or #error directives. 914 if (CurPTHLexer) 915 return CurPTHLexer->DiscardToEndOfLine(); 916 917 // Read the rest of the line raw. We do this because we don't want macros 918 // to be expanded and we don't require that the tokens be valid preprocessing 919 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 920 // collapse multiple consequtive white space between tokens, but this isn't 921 // specified by the standard. 922 std::string Message = CurLexer->ReadToEndOfLine(); 923 if (isWarning) 924 Diag(Tok, diag::pp_hash_warning) << Message; 925 else 926 Diag(Tok, diag::err_pp_hash_error) << Message; 927} 928 929/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 930/// 931void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 932 // Yes, this directive is an extension. 933 Diag(Tok, diag::ext_pp_ident_directive); 934 935 // Read the string argument. 936 Token StrTok; 937 Lex(StrTok); 938 939 // If the token kind isn't a string, it's a malformed directive. 940 if (StrTok.isNot(tok::string_literal) && 941 StrTok.isNot(tok::wide_string_literal)) { 942 Diag(StrTok, diag::err_pp_malformed_ident); 943 if (StrTok.isNot(tok::eom)) 944 DiscardUntilEndOfDirective(); 945 return; 946 } 947 948 // Verify that there is nothing after the string, other than EOM. 949 CheckEndOfDirective("ident"); 950 951 if (Callbacks) { 952 bool Invalid = false; 953 std::string Str = getSpelling(StrTok, &Invalid); 954 if (!Invalid) 955 Callbacks->Ident(Tok.getLocation(), Str); 956 } 957} 958 959//===----------------------------------------------------------------------===// 960// Preprocessor Include Directive Handling. 961//===----------------------------------------------------------------------===// 962 963/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 964/// checked and spelled filename, e.g. as an operand of #include. This returns 965/// true if the input filename was in <>'s or false if it were in ""'s. The 966/// caller is expected to provide a buffer that is large enough to hold the 967/// spelling of the filename, but is also expected to handle the case when 968/// this method decides to use a different buffer. 969bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 970 llvm::StringRef &Buffer) { 971 // Get the text form of the filename. 972 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 973 974 // Make sure the filename is <x> or "x". 975 bool isAngled; 976 if (Buffer[0] == '<') { 977 if (Buffer.back() != '>') { 978 Diag(Loc, diag::err_pp_expects_filename); 979 Buffer = llvm::StringRef(); 980 return true; 981 } 982 isAngled = true; 983 } else if (Buffer[0] == '"') { 984 if (Buffer.back() != '"') { 985 Diag(Loc, diag::err_pp_expects_filename); 986 Buffer = llvm::StringRef(); 987 return true; 988 } 989 isAngled = false; 990 } else { 991 Diag(Loc, diag::err_pp_expects_filename); 992 Buffer = llvm::StringRef(); 993 return true; 994 } 995 996 // Diagnose #include "" as invalid. 997 if (Buffer.size() <= 2) { 998 Diag(Loc, diag::err_pp_empty_filename); 999 Buffer = llvm::StringRef(); 1000 return true; 1001 } 1002 1003 // Skip the brackets. 1004 Buffer = Buffer.substr(1, Buffer.size()-2); 1005 return isAngled; 1006} 1007 1008/// ConcatenateIncludeName - Handle cases where the #include name is expanded 1009/// from a macro as multiple tokens, which need to be glued together. This 1010/// occurs for code like: 1011/// #define FOO <a/b.h> 1012/// #include FOO 1013/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 1014/// 1015/// This code concatenates and consumes tokens up to the '>' token. It returns 1016/// false if the > was found, otherwise it returns true if it finds and consumes 1017/// the EOM marker. 1018bool Preprocessor::ConcatenateIncludeName( 1019 llvm::SmallString<128> &FilenameBuffer) { 1020 Token CurTok; 1021 1022 Lex(CurTok); 1023 while (CurTok.isNot(tok::eom)) { 1024 // Append the spelling of this token to the buffer. If there was a space 1025 // before it, add it now. 1026 if (CurTok.hasLeadingSpace()) 1027 FilenameBuffer.push_back(' '); 1028 1029 // Get the spelling of the token, directly into FilenameBuffer if possible. 1030 unsigned PreAppendSize = FilenameBuffer.size(); 1031 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1032 1033 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1034 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1035 1036 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1037 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1038 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1039 1040 // Resize FilenameBuffer to the correct size. 1041 if (CurTok.getLength() != ActualLen) 1042 FilenameBuffer.resize(PreAppendSize+ActualLen); 1043 1044 // If we found the '>' marker, return success. 1045 if (CurTok.is(tok::greater)) 1046 return false; 1047 1048 Lex(CurTok); 1049 } 1050 1051 // If we hit the eom marker, emit an error and return true so that the caller 1052 // knows the EOM has been read. 1053 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1054 return true; 1055} 1056 1057/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1058/// file to be included from the lexer, then include it! This is a common 1059/// routine with functionality shared between #include, #include_next and 1060/// #import. LookupFrom is set when this is a #include_next directive, it 1061/// specifies the file to start searching from. 1062void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1063 const DirectoryLookup *LookupFrom, 1064 bool isImport) { 1065 1066 Token FilenameTok; 1067 CurPPLexer->LexIncludeFilename(FilenameTok); 1068 1069 // Reserve a buffer to get the spelling. 1070 llvm::SmallString<128> FilenameBuffer; 1071 llvm::StringRef Filename; 1072 1073 switch (FilenameTok.getKind()) { 1074 case tok::eom: 1075 // If the token kind is EOM, the error has already been diagnosed. 1076 return; 1077 1078 case tok::angle_string_literal: 1079 case tok::string_literal: 1080 Filename = getSpelling(FilenameTok, FilenameBuffer); 1081 break; 1082 1083 case tok::less: 1084 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1085 // case, glue the tokens together into FilenameBuffer and interpret those. 1086 FilenameBuffer.push_back('<'); 1087 if (ConcatenateIncludeName(FilenameBuffer)) 1088 return; // Found <eom> but no ">"? Diagnostic already emitted. 1089 Filename = FilenameBuffer.str(); 1090 break; 1091 default: 1092 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1093 DiscardUntilEndOfDirective(); 1094 return; 1095 } 1096 1097 bool isAngled = 1098 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1099 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1100 // error. 1101 if (Filename.empty()) { 1102 DiscardUntilEndOfDirective(); 1103 return; 1104 } 1105 1106 // Verify that there is nothing after the filename, other than EOM. Note that 1107 // we allow macros that expand to nothing after the filename, because this 1108 // falls into the category of "#include pp-tokens new-line" specified in 1109 // C99 6.10.2p4. 1110 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1111 1112 // Check that we don't have infinite #include recursion. 1113 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1114 Diag(FilenameTok, diag::err_pp_include_too_deep); 1115 return; 1116 } 1117 1118 // Search include directories. 1119 const DirectoryLookup *CurDir; 1120 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1121 if (File == 0) { 1122 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1123 return; 1124 } 1125 1126 // The #included file will be considered to be a system header if either it is 1127 // in a system include directory, or if the #includer is a system include 1128 // header. 1129 SrcMgr::CharacteristicKind FileCharacter = 1130 std::max(HeaderInfo.getFileDirFlavor(File), 1131 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1132 1133 // Ask HeaderInfo if we should enter this #include file. If not, #including 1134 // this file will have no effect. 1135 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1136 if (Callbacks) 1137 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1138 return; 1139 } 1140 1141 // Look up the file, create a File ID for it. 1142 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1143 FileCharacter); 1144 if (FID.isInvalid()) { 1145 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1146 return; 1147 } 1148 1149 // Finally, if all is good, enter the new file! 1150 EnterSourceFile(FID, CurDir, FilenameTok.getLocation()); 1151} 1152 1153/// HandleIncludeNextDirective - Implements #include_next. 1154/// 1155void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1156 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1157 1158 // #include_next is like #include, except that we start searching after 1159 // the current found directory. If we can't do this, issue a 1160 // diagnostic. 1161 const DirectoryLookup *Lookup = CurDirLookup; 1162 if (isInPrimaryFile()) { 1163 Lookup = 0; 1164 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1165 } else if (Lookup == 0) { 1166 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1167 } else { 1168 // Start looking up in the next directory. 1169 ++Lookup; 1170 } 1171 1172 return HandleIncludeDirective(IncludeNextTok, Lookup); 1173} 1174 1175/// HandleImportDirective - Implements #import. 1176/// 1177void Preprocessor::HandleImportDirective(Token &ImportTok) { 1178 if (!Features.ObjC1) // #import is standard for ObjC. 1179 Diag(ImportTok, diag::ext_pp_import_directive); 1180 1181 return HandleIncludeDirective(ImportTok, 0, true); 1182} 1183 1184/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1185/// pseudo directive in the predefines buffer. This handles it by sucking all 1186/// tokens through the preprocessor and discarding them (only keeping the side 1187/// effects on the preprocessor). 1188void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1189 // This directive should only occur in the predefines buffer. If not, emit an 1190 // error and reject it. 1191 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1192 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1193 Diag(IncludeMacrosTok.getLocation(), 1194 diag::pp_include_macros_out_of_predefines); 1195 DiscardUntilEndOfDirective(); 1196 return; 1197 } 1198 1199 // Treat this as a normal #include for checking purposes. If this is 1200 // successful, it will push a new lexer onto the include stack. 1201 HandleIncludeDirective(IncludeMacrosTok, 0, false); 1202 1203 Token TmpTok; 1204 do { 1205 Lex(TmpTok); 1206 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1207 } while (TmpTok.isNot(tok::hashhash)); 1208} 1209 1210//===----------------------------------------------------------------------===// 1211// Preprocessor Macro Directive Handling. 1212//===----------------------------------------------------------------------===// 1213 1214/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1215/// definition has just been read. Lex the rest of the arguments and the 1216/// closing ), updating MI with what we learn. Return true if an error occurs 1217/// parsing the arg list. 1218bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1219 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1220 1221 Token Tok; 1222 while (1) { 1223 LexUnexpandedToken(Tok); 1224 switch (Tok.getKind()) { 1225 case tok::r_paren: 1226 // Found the end of the argument list. 1227 if (Arguments.empty()) // #define FOO() 1228 return false; 1229 // Otherwise we have #define FOO(A,) 1230 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1231 return true; 1232 case tok::ellipsis: // #define X(... -> C99 varargs 1233 // Warn if use of C99 feature in non-C99 mode. 1234 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1235 1236 // Lex the token after the identifier. 1237 LexUnexpandedToken(Tok); 1238 if (Tok.isNot(tok::r_paren)) { 1239 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1240 return true; 1241 } 1242 // Add the __VA_ARGS__ identifier as an argument. 1243 Arguments.push_back(Ident__VA_ARGS__); 1244 MI->setIsC99Varargs(); 1245 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1246 return false; 1247 case tok::eom: // #define X( 1248 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1249 return true; 1250 default: 1251 // Handle keywords and identifiers here to accept things like 1252 // #define Foo(for) for. 1253 IdentifierInfo *II = Tok.getIdentifierInfo(); 1254 if (II == 0) { 1255 // #define X(1 1256 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1257 return true; 1258 } 1259 1260 // If this is already used as an argument, it is used multiple times (e.g. 1261 // #define X(A,A. 1262 if (std::find(Arguments.begin(), Arguments.end(), II) != 1263 Arguments.end()) { // C99 6.10.3p6 1264 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1265 return true; 1266 } 1267 1268 // Add the argument to the macro info. 1269 Arguments.push_back(II); 1270 1271 // Lex the token after the identifier. 1272 LexUnexpandedToken(Tok); 1273 1274 switch (Tok.getKind()) { 1275 default: // #define X(A B 1276 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1277 return true; 1278 case tok::r_paren: // #define X(A) 1279 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1280 return false; 1281 case tok::comma: // #define X(A, 1282 break; 1283 case tok::ellipsis: // #define X(A... -> GCC extension 1284 // Diagnose extension. 1285 Diag(Tok, diag::ext_named_variadic_macro); 1286 1287 // Lex the token after the identifier. 1288 LexUnexpandedToken(Tok); 1289 if (Tok.isNot(tok::r_paren)) { 1290 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1291 return true; 1292 } 1293 1294 MI->setIsGNUVarargs(); 1295 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1296 return false; 1297 } 1298 } 1299 } 1300} 1301 1302/// HandleDefineDirective - Implements #define. This consumes the entire macro 1303/// line then lets the caller lex the next real token. 1304void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1305 ++NumDefined; 1306 1307 Token MacroNameTok; 1308 ReadMacroName(MacroNameTok, 1); 1309 1310 // Error reading macro name? If so, diagnostic already issued. 1311 if (MacroNameTok.is(tok::eom)) 1312 return; 1313 1314 Token LastTok = MacroNameTok; 1315 1316 // If we are supposed to keep comments in #defines, reenable comment saving 1317 // mode. 1318 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1319 1320 // Create the new macro. 1321 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1322 1323 Token Tok; 1324 LexUnexpandedToken(Tok); 1325 1326 // If this is a function-like macro definition, parse the argument list, 1327 // marking each of the identifiers as being used as macro arguments. Also, 1328 // check other constraints on the first token of the macro body. 1329 if (Tok.is(tok::eom)) { 1330 // If there is no body to this macro, we have no special handling here. 1331 } else if (Tok.hasLeadingSpace()) { 1332 // This is a normal token with leading space. Clear the leading space 1333 // marker on the first token to get proper expansion. 1334 Tok.clearFlag(Token::LeadingSpace); 1335 } else if (Tok.is(tok::l_paren)) { 1336 // This is a function-like macro definition. Read the argument list. 1337 MI->setIsFunctionLike(); 1338 if (ReadMacroDefinitionArgList(MI)) { 1339 // Forget about MI. 1340 ReleaseMacroInfo(MI); 1341 // Throw away the rest of the line. 1342 if (CurPPLexer->ParsingPreprocessorDirective) 1343 DiscardUntilEndOfDirective(); 1344 return; 1345 } 1346 1347 // If this is a definition of a variadic C99 function-like macro, not using 1348 // the GNU named varargs extension, enabled __VA_ARGS__. 1349 1350 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1351 // This gets unpoisoned where it is allowed. 1352 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1353 if (MI->isC99Varargs()) 1354 Ident__VA_ARGS__->setIsPoisoned(false); 1355 1356 // Read the first token after the arg list for down below. 1357 LexUnexpandedToken(Tok); 1358 } else if (Features.C99) { 1359 // C99 requires whitespace between the macro definition and the body. Emit 1360 // a diagnostic for something like "#define X+". 1361 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1362 } else { 1363 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1364 // first character of a replacement list is not a character required by 1365 // subclause 5.2.1, then there shall be white-space separation between the 1366 // identifier and the replacement list.". 5.2.1 lists this set: 1367 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1368 // is irrelevant here. 1369 bool isInvalid = false; 1370 if (Tok.is(tok::at)) // @ is not in the list above. 1371 isInvalid = true; 1372 else if (Tok.is(tok::unknown)) { 1373 // If we have an unknown token, it is something strange like "`". Since 1374 // all of valid characters would have lexed into a single character 1375 // token of some sort, we know this is not a valid case. 1376 isInvalid = true; 1377 } 1378 if (isInvalid) 1379 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1380 else 1381 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1382 } 1383 1384 if (!Tok.is(tok::eom)) 1385 LastTok = Tok; 1386 1387 // Read the rest of the macro body. 1388 if (MI->isObjectLike()) { 1389 // Object-like macros are very simple, just read their body. 1390 while (Tok.isNot(tok::eom)) { 1391 LastTok = Tok; 1392 MI->AddTokenToBody(Tok); 1393 // Get the next token of the macro. 1394 LexUnexpandedToken(Tok); 1395 } 1396 1397 } else { 1398 // Otherwise, read the body of a function-like macro. While we are at it, 1399 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1400 // parameters in function-like macro expansions. 1401 while (Tok.isNot(tok::eom)) { 1402 LastTok = Tok; 1403 1404 if (Tok.isNot(tok::hash)) { 1405 MI->AddTokenToBody(Tok); 1406 1407 // Get the next token of the macro. 1408 LexUnexpandedToken(Tok); 1409 continue; 1410 } 1411 1412 // Get the next token of the macro. 1413 LexUnexpandedToken(Tok); 1414 1415 // Check for a valid macro arg identifier. 1416 if (Tok.getIdentifierInfo() == 0 || 1417 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1418 1419 // If this is assembler-with-cpp mode, we accept random gibberish after 1420 // the '#' because '#' is often a comment character. However, change 1421 // the kind of the token to tok::unknown so that the preprocessor isn't 1422 // confused. 1423 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1424 LastTok.setKind(tok::unknown); 1425 } else { 1426 Diag(Tok, diag::err_pp_stringize_not_parameter); 1427 ReleaseMacroInfo(MI); 1428 1429 // Disable __VA_ARGS__ again. 1430 Ident__VA_ARGS__->setIsPoisoned(true); 1431 return; 1432 } 1433 } 1434 1435 // Things look ok, add the '#' and param name tokens to the macro. 1436 MI->AddTokenToBody(LastTok); 1437 MI->AddTokenToBody(Tok); 1438 LastTok = Tok; 1439 1440 // Get the next token of the macro. 1441 LexUnexpandedToken(Tok); 1442 } 1443 } 1444 1445 1446 // Disable __VA_ARGS__ again. 1447 Ident__VA_ARGS__->setIsPoisoned(true); 1448 1449 // Check that there is no paste (##) operator at the begining or end of the 1450 // replacement list. 1451 unsigned NumTokens = MI->getNumTokens(); 1452 if (NumTokens != 0) { 1453 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1454 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1455 ReleaseMacroInfo(MI); 1456 return; 1457 } 1458 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1459 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1460 ReleaseMacroInfo(MI); 1461 return; 1462 } 1463 } 1464 1465 // If this is the primary source file, remember that this macro hasn't been 1466 // used yet. 1467 if (isInPrimaryFile()) 1468 MI->setIsUsed(false); 1469 1470 MI->setDefinitionEndLoc(LastTok.getLocation()); 1471 1472 // Finally, if this identifier already had a macro defined for it, verify that 1473 // the macro bodies are identical and free the old definition. 1474 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1475 // It is very common for system headers to have tons of macro redefinitions 1476 // and for warnings to be disabled in system headers. If this is the case, 1477 // then don't bother calling MacroInfo::isIdenticalTo. 1478 if (!getDiagnostics().getSuppressSystemWarnings() || 1479 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1480 if (!OtherMI->isUsed()) 1481 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1482 1483 // Macros must be identical. This means all tokens and whitespace 1484 // separation must be the same. C99 6.10.3.2. 1485 if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 1486 !MI->isIdenticalTo(*OtherMI, *this)) { 1487 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1488 << MacroNameTok.getIdentifierInfo(); 1489 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1490 } 1491 } 1492 ReleaseMacroInfo(OtherMI); 1493 } 1494 1495 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1496 1497 // If the callbacks want to know, tell them about the macro definition. 1498 if (Callbacks) 1499 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1500} 1501 1502/// HandleUndefDirective - Implements #undef. 1503/// 1504void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1505 ++NumUndefined; 1506 1507 Token MacroNameTok; 1508 ReadMacroName(MacroNameTok, 2); 1509 1510 // Error reading macro name? If so, diagnostic already issued. 1511 if (MacroNameTok.is(tok::eom)) 1512 return; 1513 1514 // Check to see if this is the last token on the #undef line. 1515 CheckEndOfDirective("undef"); 1516 1517 // Okay, we finally have a valid identifier to undef. 1518 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1519 1520 // If the macro is not defined, this is a noop undef, just return. 1521 if (MI == 0) return; 1522 1523 if (!MI->isUsed()) 1524 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1525 1526 // If the callbacks want to know, tell them about the macro #undef. 1527 if (Callbacks) 1528 Callbacks->MacroUndefined(MacroNameTok.getLocation(), 1529 MacroNameTok.getIdentifierInfo(), MI); 1530 1531 // Free macro definition. 1532 ReleaseMacroInfo(MI); 1533 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1534} 1535 1536 1537//===----------------------------------------------------------------------===// 1538// Preprocessor Conditional Directive Handling. 1539//===----------------------------------------------------------------------===// 1540 1541/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1542/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1543/// if any tokens have been returned or pp-directives activated before this 1544/// #ifndef has been lexed. 1545/// 1546void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1547 bool ReadAnyTokensBeforeDirective) { 1548 ++NumIf; 1549 Token DirectiveTok = Result; 1550 1551 Token MacroNameTok; 1552 ReadMacroName(MacroNameTok); 1553 1554 // Error reading macro name? If so, diagnostic already issued. 1555 if (MacroNameTok.is(tok::eom)) { 1556 // Skip code until we get to #endif. This helps with recovery by not 1557 // emitting an error when the #endif is reached. 1558 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1559 /*Foundnonskip*/false, /*FoundElse*/false); 1560 return; 1561 } 1562 1563 // Check to see if this is the last token on the #if[n]def line. 1564 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1565 1566 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1567 MacroInfo *MI = getMacroInfo(MII); 1568 1569 if (CurPPLexer->getConditionalStackDepth() == 0) { 1570 // If the start of a top-level #ifdef and if the macro is not defined, 1571 // inform MIOpt that this might be the start of a proper include guard. 1572 // Otherwise it is some other form of unknown conditional which we can't 1573 // handle. 1574 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1575 assert(isIfndef && "#ifdef shouldn't reach here"); 1576 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1577 } else 1578 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1579 } 1580 1581 // If there is a macro, process it. 1582 if (MI) // Mark it used. 1583 MI->setIsUsed(true); 1584 1585 // Should we include the stuff contained by this directive? 1586 if (!MI == isIfndef) { 1587 // Yes, remember that we are inside a conditional, then lex the next token. 1588 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1589 /*wasskip*/false, /*foundnonskip*/true, 1590 /*foundelse*/false); 1591 } else { 1592 // No, skip the contents of this block and return the first token after it. 1593 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1594 /*Foundnonskip*/false, 1595 /*FoundElse*/false); 1596 } 1597} 1598 1599/// HandleIfDirective - Implements the #if directive. 1600/// 1601void Preprocessor::HandleIfDirective(Token &IfToken, 1602 bool ReadAnyTokensBeforeDirective) { 1603 ++NumIf; 1604 1605 // Parse and evaluation the conditional expression. 1606 IdentifierInfo *IfNDefMacro = 0; 1607 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1608 1609 1610 // If this condition is equivalent to #ifndef X, and if this is the first 1611 // directive seen, handle it for the multiple-include optimization. 1612 if (CurPPLexer->getConditionalStackDepth() == 0) { 1613 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1614 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1615 else 1616 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1617 } 1618 1619 // Should we include the stuff contained by this directive? 1620 if (ConditionalTrue) { 1621 // Yes, remember that we are inside a conditional, then lex the next token. 1622 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1623 /*foundnonskip*/true, /*foundelse*/false); 1624 } else { 1625 // No, skip the contents of this block and return the first token after it. 1626 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1627 /*FoundElse*/false); 1628 } 1629} 1630 1631/// HandleEndifDirective - Implements the #endif directive. 1632/// 1633void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1634 ++NumEndif; 1635 1636 // Check that this is the whole directive. 1637 CheckEndOfDirective("endif"); 1638 1639 PPConditionalInfo CondInfo; 1640 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1641 // No conditionals on the stack: this is an #endif without an #if. 1642 Diag(EndifToken, diag::err_pp_endif_without_if); 1643 return; 1644 } 1645 1646 // If this the end of a top-level #endif, inform MIOpt. 1647 if (CurPPLexer->getConditionalStackDepth() == 0) 1648 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1649 1650 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1651 "This code should only be reachable in the non-skipping case!"); 1652} 1653 1654 1655void Preprocessor::HandleElseDirective(Token &Result) { 1656 ++NumElse; 1657 1658 // #else directive in a non-skipping conditional... start skipping. 1659 CheckEndOfDirective("else"); 1660 1661 PPConditionalInfo CI; 1662 if (CurPPLexer->popConditionalLevel(CI)) { 1663 Diag(Result, diag::pp_err_else_without_if); 1664 return; 1665 } 1666 1667 // If this is a top-level #else, inform the MIOpt. 1668 if (CurPPLexer->getConditionalStackDepth() == 0) 1669 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1670 1671 // If this is a #else with a #else before it, report the error. 1672 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1673 1674 // Finally, skip the rest of the contents of this block and return the first 1675 // token after it. 1676 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1677 /*FoundElse*/true); 1678} 1679 1680void Preprocessor::HandleElifDirective(Token &ElifToken) { 1681 ++NumElse; 1682 1683 // #elif directive in a non-skipping conditional... start skipping. 1684 // We don't care what the condition is, because we will always skip it (since 1685 // the block immediately before it was included). 1686 DiscardUntilEndOfDirective(); 1687 1688 PPConditionalInfo CI; 1689 if (CurPPLexer->popConditionalLevel(CI)) { 1690 Diag(ElifToken, diag::pp_err_elif_without_if); 1691 return; 1692 } 1693 1694 // If this is a top-level #elif, inform the MIOpt. 1695 if (CurPPLexer->getConditionalStackDepth() == 0) 1696 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1697 1698 // If this is a #elif with a #else before it, report the error. 1699 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1700 1701 // Finally, skip the rest of the contents of this block and return the first 1702 // token after it. 1703 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1704 /*FoundElse*/CI.FoundElse); 1705} 1706