PPDirectives.cpp revision c1abafc4fad5fc5ae0d428d0074495e3d2164f98
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/Basic/FileManager.h" 20#include "clang/Basic/SourceManager.h" 21#include "llvm/ADT/APInt.h" 22using namespace clang; 23 24//===----------------------------------------------------------------------===// 25// Utility Methods for Preprocessor Directive Handling. 26//===----------------------------------------------------------------------===// 27 28MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 29 MacroInfo *MI; 30 31 if (!MICache.empty()) { 32 MI = MICache.back(); 33 MICache.pop_back(); 34 } else 35 MI = (MacroInfo*) BP.Allocate<MacroInfo>(); 36 new (MI) MacroInfo(L); 37 return MI; 38} 39 40/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 41/// be reused for allocating new MacroInfo objects. 42void Preprocessor::ReleaseMacroInfo(MacroInfo* MI) { 43 MICache.push_back(MI); 44 MI->FreeArgumentList(BP); 45} 46 47 48/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 49/// current line until the tok::eom token is found. 50void Preprocessor::DiscardUntilEndOfDirective() { 51 Token Tmp; 52 do { 53 LexUnexpandedToken(Tmp); 54 } while (Tmp.isNot(tok::eom)); 55} 56 57/// ReadMacroName - Lex and validate a macro name, which occurs after a 58/// #define or #undef. This sets the token kind to eom and discards the rest 59/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 60/// this is due to a a #define, 2 if #undef directive, 0 if it is something 61/// else (e.g. #ifdef). 62void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 63 // Read the token, don't allow macro expansion on it. 64 LexUnexpandedToken(MacroNameTok); 65 66 // Missing macro name? 67 if (MacroNameTok.is(tok::eom)) { 68 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 69 return; 70 } 71 72 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 73 if (II == 0) { 74 std::string Spelling = getSpelling(MacroNameTok); 75 const IdentifierInfo &Info = Identifiers.get(Spelling); 76 if (Info.isCPlusPlusOperatorKeyword()) 77 // C++ 2.5p2: Alternative tokens behave the same as its primary token 78 // except for their spellings. 79 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 80 else 81 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 82 // Fall through on error. 83 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 84 // Error if defining "defined": C99 6.10.8.4. 85 Diag(MacroNameTok, diag::err_defined_macro_name); 86 } else if (isDefineUndef && II->hasMacroDefinition() && 87 getMacroInfo(II)->isBuiltinMacro()) { 88 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 89 if (isDefineUndef == 1) 90 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 91 else 92 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 93 } else { 94 // Okay, we got a good identifier node. Return it. 95 return; 96 } 97 98 // Invalid macro name, read and discard the rest of the line. Then set the 99 // token kind to tok::eom. 100 MacroNameTok.setKind(tok::eom); 101 return DiscardUntilEndOfDirective(); 102} 103 104/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 105/// not, emit a diagnostic and consume up until the eom. If EnableMacros is 106/// true, then we consider macros that expand to zero tokens as being ok. 107void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 108 Token Tmp; 109 // Lex unexpanded tokens for most directives: macros might expand to zero 110 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 111 // #line) allow empty macros. 112 if (EnableMacros) 113 Lex(Tmp); 114 else 115 LexUnexpandedToken(Tmp); 116 117 // There should be no tokens after the directive, but we allow them as an 118 // extension. 119 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 120 LexUnexpandedToken(Tmp); 121 122 if (Tmp.isNot(tok::eom)) { 123 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 124 // because it is more trouble than it is worth to insert /**/ and check that 125 // there is no /**/ in the range also. 126 CodeModificationHint FixItHint; 127 if (Features.GNUMode || Features.C99 || Features.CPlusPlus) 128 FixItHint = CodeModificationHint::CreateInsertion(Tmp.getLocation(),"//"); 129 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << FixItHint; 130 DiscardUntilEndOfDirective(); 131 } 132} 133 134 135 136/// SkipExcludedConditionalBlock - We just read a #if or related directive and 137/// decided that the subsequent tokens are in the #if'd out portion of the 138/// file. Lex the rest of the file, until we see an #endif. If 139/// FoundNonSkipPortion is true, then we have already emitted code for part of 140/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 141/// is true, then #else directives are ok, if not, then we have already seen one 142/// so a #else directive is a duplicate. When this returns, the caller can lex 143/// the first valid token. 144void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 145 bool FoundNonSkipPortion, 146 bool FoundElse) { 147 ++NumSkipped; 148 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 149 150 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 151 FoundNonSkipPortion, FoundElse); 152 153 if (CurPTHLexer) { 154 PTHSkipExcludedConditionalBlock(); 155 return; 156 } 157 158 // Enter raw mode to disable identifier lookup (and thus macro expansion), 159 // disabling warnings, etc. 160 CurPPLexer->LexingRawMode = true; 161 Token Tok; 162 while (1) { 163 if (CurLexer) 164 CurLexer->Lex(Tok); 165 else 166 CurPTHLexer->Lex(Tok); 167 168 // If this is the end of the buffer, we have an error. 169 if (Tok.is(tok::eof)) { 170 // Emit errors for each unterminated conditional on the stack, including 171 // the current one. 172 while (!CurPPLexer->ConditionalStack.empty()) { 173 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 174 diag::err_pp_unterminated_conditional); 175 CurPPLexer->ConditionalStack.pop_back(); 176 } 177 178 // Just return and let the caller lex after this #include. 179 break; 180 } 181 182 // If this token is not a preprocessor directive, just skip it. 183 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 184 continue; 185 186 // We just parsed a # character at the start of a line, so we're in 187 // directive mode. Tell the lexer this so any newlines we see will be 188 // converted into an EOM token (this terminates the macro). 189 CurPPLexer->ParsingPreprocessorDirective = true; 190 if (CurLexer) CurLexer->SetCommentRetentionState(false); 191 192 193 // Read the next token, the directive flavor. 194 LexUnexpandedToken(Tok); 195 196 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 197 // something bogus), skip it. 198 if (Tok.isNot(tok::identifier)) { 199 CurPPLexer->ParsingPreprocessorDirective = false; 200 // Restore comment saving mode. 201 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 202 continue; 203 } 204 205 // If the first letter isn't i or e, it isn't intesting to us. We know that 206 // this is safe in the face of spelling differences, because there is no way 207 // to spell an i/e in a strange way that is another letter. Skipping this 208 // allows us to avoid looking up the identifier info for #define/#undef and 209 // other common directives. 210 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation()); 211 char FirstChar = RawCharData[0]; 212 if (FirstChar >= 'a' && FirstChar <= 'z' && 213 FirstChar != 'i' && FirstChar != 'e') { 214 CurPPLexer->ParsingPreprocessorDirective = false; 215 // Restore comment saving mode. 216 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 217 continue; 218 } 219 220 // Get the identifier name without trigraphs or embedded newlines. Note 221 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 222 // when skipping. 223 char DirectiveBuf[20]; 224 llvm::StringRef Directive; 225 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 226 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 227 } else { 228 std::string DirectiveStr = getSpelling(Tok); 229 unsigned IdLen = DirectiveStr.size(); 230 if (IdLen >= 20) { 231 CurPPLexer->ParsingPreprocessorDirective = false; 232 // Restore comment saving mode. 233 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 234 continue; 235 } 236 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 237 Directive = llvm::StringRef(DirectiveBuf, IdLen); 238 } 239 240 if (Directive.startswith("if")) { 241 llvm::StringRef Sub = Directive.substr(2); 242 if (Sub.empty() || // "if" 243 Sub == "def" || // "ifdef" 244 Sub == "ndef") { // "ifndef" 245 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 246 // bother parsing the condition. 247 DiscardUntilEndOfDirective(); 248 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 249 /*foundnonskip*/false, 250 /*fnddelse*/false); 251 } 252 } else if (Directive[0] == 'e') { 253 llvm::StringRef Sub = Directive.substr(1); 254 if (Sub == "ndif") { // "endif" 255 CheckEndOfDirective("endif"); 256 PPConditionalInfo CondInfo; 257 CondInfo.WasSkipping = true; // Silence bogus warning. 258 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 259 InCond = InCond; // Silence warning in no-asserts mode. 260 assert(!InCond && "Can't be skipping if not in a conditional!"); 261 262 // If we popped the outermost skipping block, we're done skipping! 263 if (!CondInfo.WasSkipping) 264 break; 265 } else if (Sub == "lse") { // "else". 266 // #else directive in a skipping conditional. If not in some other 267 // skipping conditional, and if #else hasn't already been seen, enter it 268 // as a non-skipping conditional. 269 DiscardUntilEndOfDirective(); // C99 6.10p4. 270 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 271 272 // If this is a #else with a #else before it, report the error. 273 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 274 275 // Note that we've seen a #else in this conditional. 276 CondInfo.FoundElse = true; 277 278 // If the conditional is at the top level, and the #if block wasn't 279 // entered, enter the #else block now. 280 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 281 CondInfo.FoundNonSkip = true; 282 break; 283 } 284 } else if (Sub == "lif") { // "elif". 285 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 286 287 bool ShouldEnter; 288 // If this is in a skipping block or if we're already handled this #if 289 // block, don't bother parsing the condition. 290 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 291 DiscardUntilEndOfDirective(); 292 ShouldEnter = false; 293 } else { 294 // Restore the value of LexingRawMode so that identifiers are 295 // looked up, etc, inside the #elif expression. 296 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 297 CurPPLexer->LexingRawMode = false; 298 IdentifierInfo *IfNDefMacro = 0; 299 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 300 CurPPLexer->LexingRawMode = true; 301 } 302 303 // If this is a #elif with a #else before it, report the error. 304 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 305 306 // If this condition is true, enter it! 307 if (ShouldEnter) { 308 CondInfo.FoundNonSkip = true; 309 break; 310 } 311 } 312 } 313 314 CurPPLexer->ParsingPreprocessorDirective = false; 315 // Restore comment saving mode. 316 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 317 } 318 319 // Finally, if we are out of the conditional (saw an #endif or ran off the end 320 // of the file, just stop skipping and return to lexing whatever came after 321 // the #if block. 322 CurPPLexer->LexingRawMode = false; 323} 324 325void Preprocessor::PTHSkipExcludedConditionalBlock() { 326 327 while (1) { 328 assert(CurPTHLexer); 329 assert(CurPTHLexer->LexingRawMode == false); 330 331 // Skip to the next '#else', '#elif', or #endif. 332 if (CurPTHLexer->SkipBlock()) { 333 // We have reached an #endif. Both the '#' and 'endif' tokens 334 // have been consumed by the PTHLexer. Just pop off the condition level. 335 PPConditionalInfo CondInfo; 336 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 337 InCond = InCond; // Silence warning in no-asserts mode. 338 assert(!InCond && "Can't be skipping if not in a conditional!"); 339 break; 340 } 341 342 // We have reached a '#else' or '#elif'. Lex the next token to get 343 // the directive flavor. 344 Token Tok; 345 LexUnexpandedToken(Tok); 346 347 // We can actually look up the IdentifierInfo here since we aren't in 348 // raw mode. 349 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 350 351 if (K == tok::pp_else) { 352 // #else: Enter the else condition. We aren't in a nested condition 353 // since we skip those. We're always in the one matching the last 354 // blocked we skipped. 355 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 356 // Note that we've seen a #else in this conditional. 357 CondInfo.FoundElse = true; 358 359 // If the #if block wasn't entered then enter the #else block now. 360 if (!CondInfo.FoundNonSkip) { 361 CondInfo.FoundNonSkip = true; 362 363 // Scan until the eom token. 364 CurPTHLexer->ParsingPreprocessorDirective = true; 365 DiscardUntilEndOfDirective(); 366 CurPTHLexer->ParsingPreprocessorDirective = false; 367 368 break; 369 } 370 371 // Otherwise skip this block. 372 continue; 373 } 374 375 assert(K == tok::pp_elif); 376 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 377 378 // If this is a #elif with a #else before it, report the error. 379 if (CondInfo.FoundElse) 380 Diag(Tok, diag::pp_err_elif_after_else); 381 382 // If this is in a skipping block or if we're already handled this #if 383 // block, don't bother parsing the condition. We just skip this block. 384 if (CondInfo.FoundNonSkip) 385 continue; 386 387 // Evaluate the condition of the #elif. 388 IdentifierInfo *IfNDefMacro = 0; 389 CurPTHLexer->ParsingPreprocessorDirective = true; 390 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 391 CurPTHLexer->ParsingPreprocessorDirective = false; 392 393 // If this condition is true, enter it! 394 if (ShouldEnter) { 395 CondInfo.FoundNonSkip = true; 396 break; 397 } 398 399 // Otherwise, skip this block and go to the next one. 400 continue; 401 } 402} 403 404/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 405/// return null on failure. isAngled indicates whether the file reference is 406/// for system #include's or not (i.e. using <> instead of ""). 407const FileEntry *Preprocessor::LookupFile(const char *FilenameStart, 408 const char *FilenameEnd, 409 SourceLocation FilenameTokLoc, 410 bool isAngled, 411 const DirectoryLookup *FromDir, 412 const DirectoryLookup *&CurDir) { 413 // If the header lookup mechanism may be relative to the current file, pass in 414 // info about where the current file is. 415 const FileEntry *CurFileEnt = 0; 416 if (!FromDir) { 417 FileID FID = getCurrentFileLexer()->getFileID(); 418 CurFileEnt = SourceMgr.getFileEntryForID(FID); 419 420 // If there is no file entry associated with this file, it must be the 421 // predefines buffer. Any other file is not lexed with a normal lexer, so 422 // it won't be scanned for preprocessor directives. If we have the 423 // predefines buffer, resolve #include references (which come from the 424 // -include command line argument) as if they came from the main file, this 425 // affects file lookup etc. 426 if (CurFileEnt == 0) { 427 FID = SourceMgr.getMainFileID(); 428 CurFileEnt = SourceMgr.getFileEntryForID(FID); 429 } 430 } 431 432 // Do a standard file entry lookup. 433 CurDir = CurDirLookup; 434 const FileEntry *FE = 435 HeaderInfo.LookupFile(FilenameStart, FilenameEnd, 436 isAngled, FromDir, CurDir, CurFileEnt); 437 if (FE) { 438 // Warn about normal quoted #include from framework headers. Since 439 // framework headers are published (both public and private ones) they 440 // should not do relative searches, they should do an include with the 441 // framework path included. 442 if (!isAngled && CurDir && FilenameTokLoc.isValid() && 443 CurDir->isFramework() && CurDir == CurDirLookup) 444 Diag(FilenameTokLoc, diag::warn_pp_relative_include_from_framework); 445 return FE; 446 } 447 448 // Otherwise, see if this is a subframework header. If so, this is relative 449 // to one of the headers on the #include stack. Walk the list of the current 450 // headers on the #include stack and pass them to HeaderInfo. 451 if (IsFileLexer()) { 452 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 453 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd, 454 CurFileEnt))) 455 return FE; 456 } 457 458 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 459 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 460 if (IsFileLexer(ISEntry)) { 461 if ((CurFileEnt = 462 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 463 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, 464 FilenameEnd, CurFileEnt))) 465 return FE; 466 } 467 } 468 469 // Otherwise, we really couldn't find the file. 470 return 0; 471} 472 473 474//===----------------------------------------------------------------------===// 475// Preprocessor Directive Handling. 476//===----------------------------------------------------------------------===// 477 478/// HandleDirective - This callback is invoked when the lexer sees a # token 479/// at the start of a line. This consumes the directive, modifies the 480/// lexer/preprocessor state, and advances the lexer(s) so that the next token 481/// read is the correct one. 482void Preprocessor::HandleDirective(Token &Result) { 483 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 484 485 // We just parsed a # character at the start of a line, so we're in directive 486 // mode. Tell the lexer this so any newlines we see will be converted into an 487 // EOM token (which terminates the directive). 488 CurPPLexer->ParsingPreprocessorDirective = true; 489 490 ++NumDirectives; 491 492 // We are about to read a token. For the multiple-include optimization FA to 493 // work, we have to remember if we had read any tokens *before* this 494 // pp-directive. 495 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 496 497 // Save the '#' token in case we need to return it later. 498 Token SavedHash = Result; 499 500 // Read the next token, the directive flavor. This isn't expanded due to 501 // C99 6.10.3p8. 502 LexUnexpandedToken(Result); 503 504 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 505 // #define A(x) #x 506 // A(abc 507 // #warning blah 508 // def) 509 // If so, the user is relying on non-portable behavior, emit a diagnostic. 510 if (InMacroArgs) 511 Diag(Result, diag::ext_embedded_directive); 512 513TryAgain: 514 switch (Result.getKind()) { 515 case tok::eom: 516 return; // null directive. 517 case tok::comment: 518 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 519 LexUnexpandedToken(Result); 520 goto TryAgain; 521 522 case tok::numeric_constant: // # 7 GNU line marker directive. 523 if (getLangOptions().AsmPreprocessor) 524 break; // # 4 is not a preprocessor directive in .S files. 525 return HandleDigitDirective(Result); 526 default: 527 IdentifierInfo *II = Result.getIdentifierInfo(); 528 if (II == 0) break; // Not an identifier. 529 530 // Ask what the preprocessor keyword ID is. 531 switch (II->getPPKeywordID()) { 532 default: break; 533 // C99 6.10.1 - Conditional Inclusion. 534 case tok::pp_if: 535 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 536 case tok::pp_ifdef: 537 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 538 case tok::pp_ifndef: 539 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 540 case tok::pp_elif: 541 return HandleElifDirective(Result); 542 case tok::pp_else: 543 return HandleElseDirective(Result); 544 case tok::pp_endif: 545 return HandleEndifDirective(Result); 546 547 // C99 6.10.2 - Source File Inclusion. 548 case tok::pp_include: 549 return HandleIncludeDirective(Result); // Handle #include. 550 case tok::pp___include_macros: 551 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 552 553 // C99 6.10.3 - Macro Replacement. 554 case tok::pp_define: 555 return HandleDefineDirective(Result); 556 case tok::pp_undef: 557 return HandleUndefDirective(Result); 558 559 // C99 6.10.4 - Line Control. 560 case tok::pp_line: 561 return HandleLineDirective(Result); 562 563 // C99 6.10.5 - Error Directive. 564 case tok::pp_error: 565 return HandleUserDiagnosticDirective(Result, false); 566 567 // C99 6.10.6 - Pragma Directive. 568 case tok::pp_pragma: 569 return HandlePragmaDirective(); 570 571 // GNU Extensions. 572 case tok::pp_import: 573 return HandleImportDirective(Result); 574 case tok::pp_include_next: 575 return HandleIncludeNextDirective(Result); 576 577 case tok::pp_warning: 578 Diag(Result, diag::ext_pp_warning_directive); 579 return HandleUserDiagnosticDirective(Result, true); 580 case tok::pp_ident: 581 return HandleIdentSCCSDirective(Result); 582 case tok::pp_sccs: 583 return HandleIdentSCCSDirective(Result); 584 case tok::pp_assert: 585 //isExtension = true; // FIXME: implement #assert 586 break; 587 case tok::pp_unassert: 588 //isExtension = true; // FIXME: implement #unassert 589 break; 590 } 591 break; 592 } 593 594 // If this is a .S file, treat unknown # directives as non-preprocessor 595 // directives. This is important because # may be a comment or introduce 596 // various pseudo-ops. Just return the # token and push back the following 597 // token to be lexed next time. 598 if (getLangOptions().AsmPreprocessor) { 599 Token *Toks = new Token[2]; 600 // Return the # and the token after it. 601 Toks[0] = SavedHash; 602 Toks[1] = Result; 603 // Enter this token stream so that we re-lex the tokens. Make sure to 604 // enable macro expansion, in case the token after the # is an identifier 605 // that is expanded. 606 EnterTokenStream(Toks, 2, false, true); 607 return; 608 } 609 610 // If we reached here, the preprocessing token is not valid! 611 Diag(Result, diag::err_pp_invalid_directive); 612 613 // Read the rest of the PP line. 614 DiscardUntilEndOfDirective(); 615 616 // Okay, we're done parsing the directive. 617} 618 619/// GetLineValue - Convert a numeric token into an unsigned value, emitting 620/// Diagnostic DiagID if it is invalid, and returning the value in Val. 621static bool GetLineValue(Token &DigitTok, unsigned &Val, 622 unsigned DiagID, Preprocessor &PP) { 623 if (DigitTok.isNot(tok::numeric_constant)) { 624 PP.Diag(DigitTok, DiagID); 625 626 if (DigitTok.isNot(tok::eom)) 627 PP.DiscardUntilEndOfDirective(); 628 return true; 629 } 630 631 llvm::SmallString<64> IntegerBuffer; 632 IntegerBuffer.resize(DigitTok.getLength()); 633 const char *DigitTokBegin = &IntegerBuffer[0]; 634 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin); 635 636 // Verify that we have a simple digit-sequence, and compute the value. This 637 // is always a simple digit string computed in decimal, so we do this manually 638 // here. 639 Val = 0; 640 for (unsigned i = 0; i != ActualLength; ++i) { 641 if (!isdigit(DigitTokBegin[i])) { 642 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 643 diag::err_pp_line_digit_sequence); 644 PP.DiscardUntilEndOfDirective(); 645 return true; 646 } 647 648 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 649 if (NextVal < Val) { // overflow. 650 PP.Diag(DigitTok, DiagID); 651 PP.DiscardUntilEndOfDirective(); 652 return true; 653 } 654 Val = NextVal; 655 } 656 657 // Reject 0, this is needed both by #line numbers and flags. 658 if (Val == 0) { 659 PP.Diag(DigitTok, DiagID); 660 PP.DiscardUntilEndOfDirective(); 661 return true; 662 } 663 664 if (DigitTokBegin[0] == '0') 665 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 666 667 return false; 668} 669 670/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 671/// acceptable forms are: 672/// # line digit-sequence 673/// # line digit-sequence "s-char-sequence" 674void Preprocessor::HandleLineDirective(Token &Tok) { 675 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 676 // expanded. 677 Token DigitTok; 678 Lex(DigitTok); 679 680 // Validate the number and convert it to an unsigned. 681 unsigned LineNo; 682 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 683 return; 684 685 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 686 // number greater than 2147483647". C90 requires that the line # be <= 32767. 687 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 688 if (LineNo >= LineLimit) 689 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 690 691 int FilenameID = -1; 692 Token StrTok; 693 Lex(StrTok); 694 695 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 696 // string followed by eom. 697 if (StrTok.is(tok::eom)) 698 ; // ok 699 else if (StrTok.isNot(tok::string_literal)) { 700 Diag(StrTok, diag::err_pp_line_invalid_filename); 701 DiscardUntilEndOfDirective(); 702 return; 703 } else { 704 // Parse and validate the string, converting it into a unique ID. 705 StringLiteralParser Literal(&StrTok, 1, *this); 706 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 707 if (Literal.hadError) 708 return DiscardUntilEndOfDirective(); 709 if (Literal.Pascal) { 710 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 711 return DiscardUntilEndOfDirective(); 712 } 713 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 714 Literal.GetStringLength()); 715 716 // Verify that there is nothing after the string, other than EOM. Because 717 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 718 CheckEndOfDirective("line", true); 719 } 720 721 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 722 723 if (Callbacks) 724 Callbacks->FileChanged(DigitTok.getLocation(), PPCallbacks::RenameFile, 725 SrcMgr::C_User); 726} 727 728/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 729/// marker directive. 730static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 731 bool &IsSystemHeader, bool &IsExternCHeader, 732 Preprocessor &PP) { 733 unsigned FlagVal; 734 Token FlagTok; 735 PP.Lex(FlagTok); 736 if (FlagTok.is(tok::eom)) return false; 737 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 738 return true; 739 740 if (FlagVal == 1) { 741 IsFileEntry = true; 742 743 PP.Lex(FlagTok); 744 if (FlagTok.is(tok::eom)) return false; 745 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 746 return true; 747 } else if (FlagVal == 2) { 748 IsFileExit = true; 749 750 SourceManager &SM = PP.getSourceManager(); 751 // If we are leaving the current presumed file, check to make sure the 752 // presumed include stack isn't empty! 753 FileID CurFileID = 754 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 755 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 756 757 // If there is no include loc (main file) or if the include loc is in a 758 // different physical file, then we aren't in a "1" line marker flag region. 759 SourceLocation IncLoc = PLoc.getIncludeLoc(); 760 if (IncLoc.isInvalid() || 761 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 762 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 763 PP.DiscardUntilEndOfDirective(); 764 return true; 765 } 766 767 PP.Lex(FlagTok); 768 if (FlagTok.is(tok::eom)) return false; 769 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 770 return true; 771 } 772 773 // We must have 3 if there are still flags. 774 if (FlagVal != 3) { 775 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 776 PP.DiscardUntilEndOfDirective(); 777 return true; 778 } 779 780 IsSystemHeader = true; 781 782 PP.Lex(FlagTok); 783 if (FlagTok.is(tok::eom)) return false; 784 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 785 return true; 786 787 // We must have 4 if there is yet another flag. 788 if (FlagVal != 4) { 789 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 790 PP.DiscardUntilEndOfDirective(); 791 return true; 792 } 793 794 IsExternCHeader = true; 795 796 PP.Lex(FlagTok); 797 if (FlagTok.is(tok::eom)) return false; 798 799 // There are no more valid flags here. 800 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 801 PP.DiscardUntilEndOfDirective(); 802 return true; 803} 804 805/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 806/// one of the following forms: 807/// 808/// # 42 809/// # 42 "file" ('1' | '2')? 810/// # 42 "file" ('1' | '2')? '3' '4'? 811/// 812void Preprocessor::HandleDigitDirective(Token &DigitTok) { 813 // Validate the number and convert it to an unsigned. GNU does not have a 814 // line # limit other than it fit in 32-bits. 815 unsigned LineNo; 816 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 817 *this)) 818 return; 819 820 Token StrTok; 821 Lex(StrTok); 822 823 bool IsFileEntry = false, IsFileExit = false; 824 bool IsSystemHeader = false, IsExternCHeader = false; 825 int FilenameID = -1; 826 827 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 828 // string followed by eom. 829 if (StrTok.is(tok::eom)) 830 ; // ok 831 else if (StrTok.isNot(tok::string_literal)) { 832 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 833 return DiscardUntilEndOfDirective(); 834 } else { 835 // Parse and validate the string, converting it into a unique ID. 836 StringLiteralParser Literal(&StrTok, 1, *this); 837 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 838 if (Literal.hadError) 839 return DiscardUntilEndOfDirective(); 840 if (Literal.Pascal) { 841 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 842 return DiscardUntilEndOfDirective(); 843 } 844 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 845 Literal.GetStringLength()); 846 847 // If a filename was present, read any flags that are present. 848 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 849 IsSystemHeader, IsExternCHeader, *this)) 850 return; 851 } 852 853 // Create a line note with this information. 854 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 855 IsFileEntry, IsFileExit, 856 IsSystemHeader, IsExternCHeader); 857 858 // If the preprocessor has callbacks installed, notify them of the #line 859 // change. This is used so that the line marker comes out in -E mode for 860 // example. 861 if (Callbacks) { 862 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 863 if (IsFileEntry) 864 Reason = PPCallbacks::EnterFile; 865 else if (IsFileExit) 866 Reason = PPCallbacks::ExitFile; 867 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 868 if (IsExternCHeader) 869 FileKind = SrcMgr::C_ExternCSystem; 870 else if (IsSystemHeader) 871 FileKind = SrcMgr::C_System; 872 873 Callbacks->FileChanged(DigitTok.getLocation(), Reason, FileKind); 874 } 875} 876 877 878/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 879/// 880void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 881 bool isWarning) { 882 // PTH doesn't emit #warning or #error directives. 883 if (CurPTHLexer) 884 return CurPTHLexer->DiscardToEndOfLine(); 885 886 // Read the rest of the line raw. We do this because we don't want macros 887 // to be expanded and we don't require that the tokens be valid preprocessing 888 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 889 // collapse multiple consequtive white space between tokens, but this isn't 890 // specified by the standard. 891 std::string Message = CurLexer->ReadToEndOfLine(); 892 if (isWarning) 893 Diag(Tok, diag::pp_hash_warning) << Message; 894 else 895 Diag(Tok, diag::err_pp_hash_error) << Message; 896} 897 898/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 899/// 900void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 901 // Yes, this directive is an extension. 902 Diag(Tok, diag::ext_pp_ident_directive); 903 904 // Read the string argument. 905 Token StrTok; 906 Lex(StrTok); 907 908 // If the token kind isn't a string, it's a malformed directive. 909 if (StrTok.isNot(tok::string_literal) && 910 StrTok.isNot(tok::wide_string_literal)) { 911 Diag(StrTok, diag::err_pp_malformed_ident); 912 if (StrTok.isNot(tok::eom)) 913 DiscardUntilEndOfDirective(); 914 return; 915 } 916 917 // Verify that there is nothing after the string, other than EOM. 918 CheckEndOfDirective("ident"); 919 920 if (Callbacks) 921 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok)); 922} 923 924//===----------------------------------------------------------------------===// 925// Preprocessor Include Directive Handling. 926//===----------------------------------------------------------------------===// 927 928/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 929/// checked and spelled filename, e.g. as an operand of #include. This returns 930/// true if the input filename was in <>'s or false if it were in ""'s. The 931/// caller is expected to provide a buffer that is large enough to hold the 932/// spelling of the filename, but is also expected to handle the case when 933/// this method decides to use a different buffer. 934bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 935 const char *&BufStart, 936 const char *&BufEnd) { 937 // Get the text form of the filename. 938 assert(BufStart != BufEnd && "Can't have tokens with empty spellings!"); 939 940 // Make sure the filename is <x> or "x". 941 bool isAngled; 942 if (BufStart[0] == '<') { 943 if (BufEnd[-1] != '>') { 944 Diag(Loc, diag::err_pp_expects_filename); 945 BufStart = 0; 946 return true; 947 } 948 isAngled = true; 949 } else if (BufStart[0] == '"') { 950 if (BufEnd[-1] != '"') { 951 Diag(Loc, diag::err_pp_expects_filename); 952 BufStart = 0; 953 return true; 954 } 955 isAngled = false; 956 } else { 957 Diag(Loc, diag::err_pp_expects_filename); 958 BufStart = 0; 959 return true; 960 } 961 962 // Diagnose #include "" as invalid. 963 if (BufEnd-BufStart <= 2) { 964 Diag(Loc, diag::err_pp_empty_filename); 965 BufStart = 0; 966 return ""; 967 } 968 969 // Skip the brackets. 970 ++BufStart; 971 --BufEnd; 972 return isAngled; 973} 974 975/// ConcatenateIncludeName - Handle cases where the #include name is expanded 976/// from a macro as multiple tokens, which need to be glued together. This 977/// occurs for code like: 978/// #define FOO <a/b.h> 979/// #include FOO 980/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 981/// 982/// This code concatenates and consumes tokens up to the '>' token. It returns 983/// false if the > was found, otherwise it returns true if it finds and consumes 984/// the EOM marker. 985bool Preprocessor::ConcatenateIncludeName( 986 llvm::SmallVector<char, 128> &FilenameBuffer) { 987 Token CurTok; 988 989 Lex(CurTok); 990 while (CurTok.isNot(tok::eom)) { 991 // Append the spelling of this token to the buffer. If there was a space 992 // before it, add it now. 993 if (CurTok.hasLeadingSpace()) 994 FilenameBuffer.push_back(' '); 995 996 // Get the spelling of the token, directly into FilenameBuffer if possible. 997 unsigned PreAppendSize = FilenameBuffer.size(); 998 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 999 1000 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1001 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1002 1003 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1004 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1005 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1006 1007 // Resize FilenameBuffer to the correct size. 1008 if (CurTok.getLength() != ActualLen) 1009 FilenameBuffer.resize(PreAppendSize+ActualLen); 1010 1011 // If we found the '>' marker, return success. 1012 if (CurTok.is(tok::greater)) 1013 return false; 1014 1015 Lex(CurTok); 1016 } 1017 1018 // If we hit the eom marker, emit an error and return true so that the caller 1019 // knows the EOM has been read. 1020 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1021 return true; 1022} 1023 1024/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1025/// file to be included from the lexer, then include it! This is a common 1026/// routine with functionality shared between #include, #include_next and 1027/// #import. LookupFrom is set when this is a #include_next directive, it 1028/// specifies the file to start searching from. 1029void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1030 const DirectoryLookup *LookupFrom, 1031 bool isImport) { 1032 1033 Token FilenameTok; 1034 CurPPLexer->LexIncludeFilename(FilenameTok); 1035 1036 // Reserve a buffer to get the spelling. 1037 llvm::SmallVector<char, 128> FilenameBuffer; 1038 const char *FilenameStart, *FilenameEnd; 1039 1040 switch (FilenameTok.getKind()) { 1041 case tok::eom: 1042 // If the token kind is EOM, the error has already been diagnosed. 1043 return; 1044 1045 case tok::angle_string_literal: 1046 case tok::string_literal: { 1047 FilenameBuffer.resize(FilenameTok.getLength()); 1048 FilenameStart = &FilenameBuffer[0]; 1049 unsigned Len = getSpelling(FilenameTok, FilenameStart); 1050 FilenameEnd = FilenameStart+Len; 1051 break; 1052 } 1053 1054 case tok::less: 1055 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1056 // case, glue the tokens together into FilenameBuffer and interpret those. 1057 FilenameBuffer.push_back('<'); 1058 if (ConcatenateIncludeName(FilenameBuffer)) 1059 return; // Found <eom> but no ">"? Diagnostic already emitted. 1060 FilenameStart = FilenameBuffer.data(); 1061 FilenameEnd = FilenameStart + FilenameBuffer.size(); 1062 break; 1063 default: 1064 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1065 DiscardUntilEndOfDirective(); 1066 return; 1067 } 1068 1069 bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(), 1070 FilenameStart, FilenameEnd); 1071 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1072 // error. 1073 if (FilenameStart == 0) { 1074 DiscardUntilEndOfDirective(); 1075 return; 1076 } 1077 1078 // Verify that there is nothing after the filename, other than EOM. Note that 1079 // we allow macros that expand to nothing after the filename, because this 1080 // falls into the category of "#include pp-tokens new-line" specified in 1081 // C99 6.10.2p4. 1082 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1083 1084 // Check that we don't have infinite #include recursion. 1085 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1086 Diag(FilenameTok, diag::err_pp_include_too_deep); 1087 return; 1088 } 1089 1090 // Search include directories. 1091 const DirectoryLookup *CurDir; 1092 const FileEntry *File = LookupFile(FilenameStart, FilenameEnd, 1093 FilenameTok.getLocation(), 1094 isAngled, LookupFrom, CurDir); 1095 if (File == 0) { 1096 Diag(FilenameTok, diag::err_pp_file_not_found) 1097 << std::string(FilenameStart, FilenameEnd); 1098 return; 1099 } 1100 1101 // Ask HeaderInfo if we should enter this #include file. If not, #including 1102 // this file will have no effect. 1103 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) 1104 return; 1105 1106 // The #included file will be considered to be a system header if either it is 1107 // in a system include directory, or if the #includer is a system include 1108 // header. 1109 SrcMgr::CharacteristicKind FileCharacter = 1110 std::max(HeaderInfo.getFileDirFlavor(File), 1111 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1112 1113 // Look up the file, create a File ID for it. 1114 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1115 FileCharacter); 1116 if (FID.isInvalid()) { 1117 Diag(FilenameTok, diag::err_pp_file_not_found) 1118 << std::string(FilenameStart, FilenameEnd); 1119 return; 1120 } 1121 1122 // Finally, if all is good, enter the new file! 1123 std::string ErrorStr; 1124 if (EnterSourceFile(FID, CurDir, ErrorStr)) 1125 Diag(FilenameTok, diag::err_pp_error_opening_file) 1126 << std::string(SourceMgr.getFileEntryForID(FID)->getName()) << ErrorStr; 1127} 1128 1129/// HandleIncludeNextDirective - Implements #include_next. 1130/// 1131void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1132 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1133 1134 // #include_next is like #include, except that we start searching after 1135 // the current found directory. If we can't do this, issue a 1136 // diagnostic. 1137 const DirectoryLookup *Lookup = CurDirLookup; 1138 if (isInPrimaryFile()) { 1139 Lookup = 0; 1140 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1141 } else if (Lookup == 0) { 1142 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1143 } else { 1144 // Start looking up in the next directory. 1145 ++Lookup; 1146 } 1147 1148 return HandleIncludeDirective(IncludeNextTok, Lookup); 1149} 1150 1151/// HandleImportDirective - Implements #import. 1152/// 1153void Preprocessor::HandleImportDirective(Token &ImportTok) { 1154 if (!Features.ObjC1) // #import is standard for ObjC. 1155 Diag(ImportTok, diag::ext_pp_import_directive); 1156 1157 return HandleIncludeDirective(ImportTok, 0, true); 1158} 1159 1160/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1161/// pseudo directive in the predefines buffer. This handles it by sucking all 1162/// tokens through the preprocessor and discarding them (only keeping the side 1163/// effects on the preprocessor). 1164void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1165 // This directive should only occur in the predefines buffer. If not, emit an 1166 // error and reject it. 1167 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1168 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1169 Diag(IncludeMacrosTok.getLocation(), 1170 diag::pp_include_macros_out_of_predefines); 1171 DiscardUntilEndOfDirective(); 1172 return; 1173 } 1174 1175 // Treat this as a normal #include for checking purposes. If this is 1176 // successful, it will push a new lexer onto the include stack. 1177 HandleIncludeDirective(IncludeMacrosTok, 0, false); 1178 1179 Token TmpTok; 1180 do { 1181 Lex(TmpTok); 1182 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1183 } while (TmpTok.isNot(tok::hashhash)); 1184} 1185 1186//===----------------------------------------------------------------------===// 1187// Preprocessor Macro Directive Handling. 1188//===----------------------------------------------------------------------===// 1189 1190/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1191/// definition has just been read. Lex the rest of the arguments and the 1192/// closing ), updating MI with what we learn. Return true if an error occurs 1193/// parsing the arg list. 1194bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1195 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1196 1197 Token Tok; 1198 while (1) { 1199 LexUnexpandedToken(Tok); 1200 switch (Tok.getKind()) { 1201 case tok::r_paren: 1202 // Found the end of the argument list. 1203 if (Arguments.empty()) // #define FOO() 1204 return false; 1205 // Otherwise we have #define FOO(A,) 1206 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1207 return true; 1208 case tok::ellipsis: // #define X(... -> C99 varargs 1209 // Warn if use of C99 feature in non-C99 mode. 1210 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1211 1212 // Lex the token after the identifier. 1213 LexUnexpandedToken(Tok); 1214 if (Tok.isNot(tok::r_paren)) { 1215 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1216 return true; 1217 } 1218 // Add the __VA_ARGS__ identifier as an argument. 1219 Arguments.push_back(Ident__VA_ARGS__); 1220 MI->setIsC99Varargs(); 1221 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1222 return false; 1223 case tok::eom: // #define X( 1224 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1225 return true; 1226 default: 1227 // Handle keywords and identifiers here to accept things like 1228 // #define Foo(for) for. 1229 IdentifierInfo *II = Tok.getIdentifierInfo(); 1230 if (II == 0) { 1231 // #define X(1 1232 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1233 return true; 1234 } 1235 1236 // If this is already used as an argument, it is used multiple times (e.g. 1237 // #define X(A,A. 1238 if (std::find(Arguments.begin(), Arguments.end(), II) != 1239 Arguments.end()) { // C99 6.10.3p6 1240 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1241 return true; 1242 } 1243 1244 // Add the argument to the macro info. 1245 Arguments.push_back(II); 1246 1247 // Lex the token after the identifier. 1248 LexUnexpandedToken(Tok); 1249 1250 switch (Tok.getKind()) { 1251 default: // #define X(A B 1252 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1253 return true; 1254 case tok::r_paren: // #define X(A) 1255 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1256 return false; 1257 case tok::comma: // #define X(A, 1258 break; 1259 case tok::ellipsis: // #define X(A... -> GCC extension 1260 // Diagnose extension. 1261 Diag(Tok, diag::ext_named_variadic_macro); 1262 1263 // Lex the token after the identifier. 1264 LexUnexpandedToken(Tok); 1265 if (Tok.isNot(tok::r_paren)) { 1266 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1267 return true; 1268 } 1269 1270 MI->setIsGNUVarargs(); 1271 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1272 return false; 1273 } 1274 } 1275 } 1276} 1277 1278/// HandleDefineDirective - Implements #define. This consumes the entire macro 1279/// line then lets the caller lex the next real token. 1280void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1281 ++NumDefined; 1282 1283 Token MacroNameTok; 1284 ReadMacroName(MacroNameTok, 1); 1285 1286 // Error reading macro name? If so, diagnostic already issued. 1287 if (MacroNameTok.is(tok::eom)) 1288 return; 1289 1290 Token LastTok = MacroNameTok; 1291 1292 // If we are supposed to keep comments in #defines, reenable comment saving 1293 // mode. 1294 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1295 1296 // Create the new macro. 1297 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1298 1299 Token Tok; 1300 LexUnexpandedToken(Tok); 1301 1302 // If this is a function-like macro definition, parse the argument list, 1303 // marking each of the identifiers as being used as macro arguments. Also, 1304 // check other constraints on the first token of the macro body. 1305 if (Tok.is(tok::eom)) { 1306 // If there is no body to this macro, we have no special handling here. 1307 } else if (Tok.hasLeadingSpace()) { 1308 // This is a normal token with leading space. Clear the leading space 1309 // marker on the first token to get proper expansion. 1310 Tok.clearFlag(Token::LeadingSpace); 1311 } else if (Tok.is(tok::l_paren)) { 1312 // This is a function-like macro definition. Read the argument list. 1313 MI->setIsFunctionLike(); 1314 if (ReadMacroDefinitionArgList(MI)) { 1315 // Forget about MI. 1316 ReleaseMacroInfo(MI); 1317 // Throw away the rest of the line. 1318 if (CurPPLexer->ParsingPreprocessorDirective) 1319 DiscardUntilEndOfDirective(); 1320 return; 1321 } 1322 1323 // If this is a definition of a variadic C99 function-like macro, not using 1324 // the GNU named varargs extension, enabled __VA_ARGS__. 1325 1326 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1327 // This gets unpoisoned where it is allowed. 1328 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1329 if (MI->isC99Varargs()) 1330 Ident__VA_ARGS__->setIsPoisoned(false); 1331 1332 // Read the first token after the arg list for down below. 1333 LexUnexpandedToken(Tok); 1334 } else if (Features.C99) { 1335 // C99 requires whitespace between the macro definition and the body. Emit 1336 // a diagnostic for something like "#define X+". 1337 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1338 } else { 1339 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1340 // first character of a replacement list is not a character required by 1341 // subclause 5.2.1, then there shall be white-space separation between the 1342 // identifier and the replacement list.". 5.2.1 lists this set: 1343 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1344 // is irrelevant here. 1345 bool isInvalid = false; 1346 if (Tok.is(tok::at)) // @ is not in the list above. 1347 isInvalid = true; 1348 else if (Tok.is(tok::unknown)) { 1349 // If we have an unknown token, it is something strange like "`". Since 1350 // all of valid characters would have lexed into a single character 1351 // token of some sort, we know this is not a valid case. 1352 isInvalid = true; 1353 } 1354 if (isInvalid) 1355 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1356 else 1357 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1358 } 1359 1360 if (!Tok.is(tok::eom)) 1361 LastTok = Tok; 1362 1363 // Read the rest of the macro body. 1364 if (MI->isObjectLike()) { 1365 // Object-like macros are very simple, just read their body. 1366 while (Tok.isNot(tok::eom)) { 1367 LastTok = Tok; 1368 MI->AddTokenToBody(Tok); 1369 // Get the next token of the macro. 1370 LexUnexpandedToken(Tok); 1371 } 1372 1373 } else { 1374 // Otherwise, read the body of a function-like macro. While we are at it, 1375 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1376 // parameters in function-like macro expansions. 1377 while (Tok.isNot(tok::eom)) { 1378 LastTok = Tok; 1379 1380 if (Tok.isNot(tok::hash)) { 1381 MI->AddTokenToBody(Tok); 1382 1383 // Get the next token of the macro. 1384 LexUnexpandedToken(Tok); 1385 continue; 1386 } 1387 1388 // Get the next token of the macro. 1389 LexUnexpandedToken(Tok); 1390 1391 // Check for a valid macro arg identifier. 1392 if (Tok.getIdentifierInfo() == 0 || 1393 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1394 1395 // If this is assembler-with-cpp mode, we accept random gibberish after 1396 // the '#' because '#' is often a comment character. However, change 1397 // the kind of the token to tok::unknown so that the preprocessor isn't 1398 // confused. 1399 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1400 LastTok.setKind(tok::unknown); 1401 } else { 1402 Diag(Tok, diag::err_pp_stringize_not_parameter); 1403 ReleaseMacroInfo(MI); 1404 1405 // Disable __VA_ARGS__ again. 1406 Ident__VA_ARGS__->setIsPoisoned(true); 1407 return; 1408 } 1409 } 1410 1411 // Things look ok, add the '#' and param name tokens to the macro. 1412 MI->AddTokenToBody(LastTok); 1413 MI->AddTokenToBody(Tok); 1414 LastTok = Tok; 1415 1416 // Get the next token of the macro. 1417 LexUnexpandedToken(Tok); 1418 } 1419 } 1420 1421 1422 // Disable __VA_ARGS__ again. 1423 Ident__VA_ARGS__->setIsPoisoned(true); 1424 1425 // Check that there is no paste (##) operator at the begining or end of the 1426 // replacement list. 1427 unsigned NumTokens = MI->getNumTokens(); 1428 if (NumTokens != 0) { 1429 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1430 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1431 ReleaseMacroInfo(MI); 1432 return; 1433 } 1434 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1435 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1436 ReleaseMacroInfo(MI); 1437 return; 1438 } 1439 } 1440 1441 // If this is the primary source file, remember that this macro hasn't been 1442 // used yet. 1443 if (isInPrimaryFile()) 1444 MI->setIsUsed(false); 1445 1446 MI->setDefinitionEndLoc(LastTok.getLocation()); 1447 1448 // Finally, if this identifier already had a macro defined for it, verify that 1449 // the macro bodies are identical and free the old definition. 1450 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1451 // It is very common for system headers to have tons of macro redefinitions 1452 // and for warnings to be disabled in system headers. If this is the case, 1453 // then don't bother calling MacroInfo::isIdenticalTo. 1454 if (!getDiagnostics().getSuppressSystemWarnings() || 1455 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1456 if (!OtherMI->isUsed()) 1457 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1458 1459 // Macros must be identical. This means all tokes and whitespace 1460 // separation must be the same. C99 6.10.3.2. 1461 if (!MI->isIdenticalTo(*OtherMI, *this)) { 1462 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1463 << MacroNameTok.getIdentifierInfo(); 1464 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1465 } 1466 } 1467 1468 ReleaseMacroInfo(OtherMI); 1469 } 1470 1471 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1472 1473 // If the callbacks want to know, tell them about the macro definition. 1474 if (Callbacks) 1475 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1476} 1477 1478/// HandleUndefDirective - Implements #undef. 1479/// 1480void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1481 ++NumUndefined; 1482 1483 Token MacroNameTok; 1484 ReadMacroName(MacroNameTok, 2); 1485 1486 // Error reading macro name? If so, diagnostic already issued. 1487 if (MacroNameTok.is(tok::eom)) 1488 return; 1489 1490 // Check to see if this is the last token on the #undef line. 1491 CheckEndOfDirective("undef"); 1492 1493 // Okay, we finally have a valid identifier to undef. 1494 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1495 1496 // If the macro is not defined, this is a noop undef, just return. 1497 if (MI == 0) return; 1498 1499 if (!MI->isUsed()) 1500 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1501 1502 // If the callbacks want to know, tell them about the macro #undef. 1503 if (Callbacks) 1504 Callbacks->MacroUndefined(MacroNameTok.getIdentifierInfo(), MI); 1505 1506 // Free macro definition. 1507 ReleaseMacroInfo(MI); 1508 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1509} 1510 1511 1512//===----------------------------------------------------------------------===// 1513// Preprocessor Conditional Directive Handling. 1514//===----------------------------------------------------------------------===// 1515 1516/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1517/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1518/// if any tokens have been returned or pp-directives activated before this 1519/// #ifndef has been lexed. 1520/// 1521void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1522 bool ReadAnyTokensBeforeDirective) { 1523 ++NumIf; 1524 Token DirectiveTok = Result; 1525 1526 Token MacroNameTok; 1527 ReadMacroName(MacroNameTok); 1528 1529 // Error reading macro name? If so, diagnostic already issued. 1530 if (MacroNameTok.is(tok::eom)) { 1531 // Skip code until we get to #endif. This helps with recovery by not 1532 // emitting an error when the #endif is reached. 1533 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1534 /*Foundnonskip*/false, /*FoundElse*/false); 1535 return; 1536 } 1537 1538 // Check to see if this is the last token on the #if[n]def line. 1539 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1540 1541 if (CurPPLexer->getConditionalStackDepth() == 0) { 1542 // If the start of a top-level #ifdef, inform MIOpt. 1543 if (!ReadAnyTokensBeforeDirective) { 1544 assert(isIfndef && "#ifdef shouldn't reach here"); 1545 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo()); 1546 } else 1547 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1548 } 1549 1550 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1551 MacroInfo *MI = getMacroInfo(MII); 1552 1553 // If there is a macro, process it. 1554 if (MI) // Mark it used. 1555 MI->setIsUsed(true); 1556 1557 // Should we include the stuff contained by this directive? 1558 if (!MI == isIfndef) { 1559 // Yes, remember that we are inside a conditional, then lex the next token. 1560 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1561 /*wasskip*/false, /*foundnonskip*/true, 1562 /*foundelse*/false); 1563 } else { 1564 // No, skip the contents of this block and return the first token after it. 1565 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1566 /*Foundnonskip*/false, 1567 /*FoundElse*/false); 1568 } 1569} 1570 1571/// HandleIfDirective - Implements the #if directive. 1572/// 1573void Preprocessor::HandleIfDirective(Token &IfToken, 1574 bool ReadAnyTokensBeforeDirective) { 1575 ++NumIf; 1576 1577 // Parse and evaluation the conditional expression. 1578 IdentifierInfo *IfNDefMacro = 0; 1579 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1580 1581 1582 // If this condition is equivalent to #ifndef X, and if this is the first 1583 // directive seen, handle it for the multiple-include optimization. 1584 if (CurPPLexer->getConditionalStackDepth() == 0) { 1585 if (!ReadAnyTokensBeforeDirective && IfNDefMacro) 1586 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1587 else 1588 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1589 } 1590 1591 // Should we include the stuff contained by this directive? 1592 if (ConditionalTrue) { 1593 // Yes, remember that we are inside a conditional, then lex the next token. 1594 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1595 /*foundnonskip*/true, /*foundelse*/false); 1596 } else { 1597 // No, skip the contents of this block and return the first token after it. 1598 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1599 /*FoundElse*/false); 1600 } 1601} 1602 1603/// HandleEndifDirective - Implements the #endif directive. 1604/// 1605void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1606 ++NumEndif; 1607 1608 // Check that this is the whole directive. 1609 CheckEndOfDirective("endif"); 1610 1611 PPConditionalInfo CondInfo; 1612 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1613 // No conditionals on the stack: this is an #endif without an #if. 1614 Diag(EndifToken, diag::err_pp_endif_without_if); 1615 return; 1616 } 1617 1618 // If this the end of a top-level #endif, inform MIOpt. 1619 if (CurPPLexer->getConditionalStackDepth() == 0) 1620 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1621 1622 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1623 "This code should only be reachable in the non-skipping case!"); 1624} 1625 1626 1627void Preprocessor::HandleElseDirective(Token &Result) { 1628 ++NumElse; 1629 1630 // #else directive in a non-skipping conditional... start skipping. 1631 CheckEndOfDirective("else"); 1632 1633 PPConditionalInfo CI; 1634 if (CurPPLexer->popConditionalLevel(CI)) { 1635 Diag(Result, diag::pp_err_else_without_if); 1636 return; 1637 } 1638 1639 // If this is a top-level #else, inform the MIOpt. 1640 if (CurPPLexer->getConditionalStackDepth() == 0) 1641 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1642 1643 // If this is a #else with a #else before it, report the error. 1644 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1645 1646 // Finally, skip the rest of the contents of this block and return the first 1647 // token after it. 1648 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1649 /*FoundElse*/true); 1650} 1651 1652void Preprocessor::HandleElifDirective(Token &ElifToken) { 1653 ++NumElse; 1654 1655 // #elif directive in a non-skipping conditional... start skipping. 1656 // We don't care what the condition is, because we will always skip it (since 1657 // the block immediately before it was included). 1658 DiscardUntilEndOfDirective(); 1659 1660 PPConditionalInfo CI; 1661 if (CurPPLexer->popConditionalLevel(CI)) { 1662 Diag(ElifToken, diag::pp_err_elif_without_if); 1663 return; 1664 } 1665 1666 // If this is a top-level #elif, inform the MIOpt. 1667 if (CurPPLexer->getConditionalStackDepth() == 0) 1668 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1669 1670 // If this is a #elif with a #else before it, report the error. 1671 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1672 1673 // Finally, skip the rest of the contents of this block and return the first 1674 // token after it. 1675 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1676 /*FoundElse*/CI.FoundElse); 1677} 1678 1679