PPDirectives.cpp revision f47724bf78299c7a50f008e0443c5f9f9f279ddc
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() { 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 return MI; 37} 38 39MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) { 40 MacroInfo *MI = AllocateMacroInfo(); 41 new (MI) MacroInfo(L); 42 return MI; 43} 44 45MacroInfo *Preprocessor::CloneMacroInfo(const MacroInfo &MacroToClone) { 46 MacroInfo *MI = AllocateMacroInfo(); 47 new (MI) MacroInfo(MacroToClone, BP); 48 return MI; 49} 50 51/// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 52/// be reused for allocating new MacroInfo objects. 53void Preprocessor::ReleaseMacroInfo(MacroInfo* MI) { 54 MICache.push_back(MI); 55 MI->FreeArgumentList(BP); 56} 57 58 59/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 60/// current line until the tok::eom token is found. 61void Preprocessor::DiscardUntilEndOfDirective() { 62 Token Tmp; 63 do { 64 LexUnexpandedToken(Tmp); 65 } while (Tmp.isNot(tok::eom)); 66} 67 68/// ReadMacroName - Lex and validate a macro name, which occurs after a 69/// #define or #undef. This sets the token kind to eom and discards the rest 70/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 71/// this is due to a a #define, 2 if #undef directive, 0 if it is something 72/// else (e.g. #ifdef). 73void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 74 // Read the token, don't allow macro expansion on it. 75 LexUnexpandedToken(MacroNameTok); 76 77 // Missing macro name? 78 if (MacroNameTok.is(tok::eom)) { 79 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 80 return; 81 } 82 83 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 84 if (II == 0) { 85 bool Invalid = false; 86 std::string Spelling = getSpelling(MacroNameTok, &Invalid); 87 if (Invalid) 88 return; 89 90 const IdentifierInfo &Info = Identifiers.get(Spelling); 91 if (Info.isCPlusPlusOperatorKeyword()) 92 // C++ 2.5p2: Alternative tokens behave the same as its primary token 93 // except for their spellings. 94 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 95 else 96 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 97 // Fall through on error. 98 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 99 // Error if defining "defined": C99 6.10.8.4. 100 Diag(MacroNameTok, diag::err_defined_macro_name); 101 } else if (isDefineUndef && II->hasMacroDefinition() && 102 getMacroInfo(II)->isBuiltinMacro()) { 103 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 104 if (isDefineUndef == 1) 105 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 106 else 107 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 108 } else { 109 // Okay, we got a good identifier node. Return it. 110 return; 111 } 112 113 // Invalid macro name, read and discard the rest of the line. Then set the 114 // token kind to tok::eom. 115 MacroNameTok.setKind(tok::eom); 116 return DiscardUntilEndOfDirective(); 117} 118 119/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 120/// not, emit a diagnostic and consume up until the eom. If EnableMacros is 121/// true, then we consider macros that expand to zero tokens as being ok. 122void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) { 123 Token Tmp; 124 // Lex unexpanded tokens for most directives: macros might expand to zero 125 // tokens, causing us to miss diagnosing invalid lines. Some directives (like 126 // #line) allow empty macros. 127 if (EnableMacros) 128 Lex(Tmp); 129 else 130 LexUnexpandedToken(Tmp); 131 132 // There should be no tokens after the directive, but we allow them as an 133 // extension. 134 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 135 LexUnexpandedToken(Tmp); 136 137 if (Tmp.isNot(tok::eom)) { 138 // Add a fixit in GNU/C99/C++ mode. Don't offer a fixit for strict-C89, 139 // because it is more trouble than it is worth to insert /**/ and check that 140 // there is no /**/ in the range also. 141 FixItHint Hint; 142 if (Features.GNUMode || Features.C99 || Features.CPlusPlus) 143 Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//"); 144 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint; 145 DiscardUntilEndOfDirective(); 146 } 147} 148 149 150 151/// SkipExcludedConditionalBlock - We just read a #if or related directive and 152/// decided that the subsequent tokens are in the #if'd out portion of the 153/// file. Lex the rest of the file, until we see an #endif. If 154/// FoundNonSkipPortion is true, then we have already emitted code for part of 155/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 156/// is true, then #else directives are ok, if not, then we have already seen one 157/// so a #else directive is a duplicate. When this returns, the caller can lex 158/// the first valid token. 159void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 160 bool FoundNonSkipPortion, 161 bool FoundElse) { 162 ++NumSkipped; 163 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 164 165 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 166 FoundNonSkipPortion, FoundElse); 167 168 if (CurPTHLexer) { 169 PTHSkipExcludedConditionalBlock(); 170 return; 171 } 172 173 // Enter raw mode to disable identifier lookup (and thus macro expansion), 174 // disabling warnings, etc. 175 CurPPLexer->LexingRawMode = true; 176 Token Tok; 177 while (1) { 178 CurLexer->Lex(Tok); 179 180 // If this is the end of the buffer, we have an error. 181 if (Tok.is(tok::eof)) { 182 // Emit errors for each unterminated conditional on the stack, including 183 // the current one. 184 while (!CurPPLexer->ConditionalStack.empty()) { 185 if (!isCodeCompletionFile(Tok.getLocation())) 186 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 187 diag::err_pp_unterminated_conditional); 188 CurPPLexer->ConditionalStack.pop_back(); 189 } 190 191 // Just return and let the caller lex after this #include. 192 break; 193 } 194 195 // If this token is not a preprocessor directive, just skip it. 196 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 197 continue; 198 199 // We just parsed a # character at the start of a line, so we're in 200 // directive mode. Tell the lexer this so any newlines we see will be 201 // converted into an EOM token (this terminates the macro). 202 CurPPLexer->ParsingPreprocessorDirective = true; 203 if (CurLexer) CurLexer->SetCommentRetentionState(false); 204 205 206 // Read the next token, the directive flavor. 207 LexUnexpandedToken(Tok); 208 209 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 210 // something bogus), skip it. 211 if (Tok.isNot(tok::identifier)) { 212 CurPPLexer->ParsingPreprocessorDirective = false; 213 // Restore comment saving mode. 214 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 215 continue; 216 } 217 218 // If the first letter isn't i or e, it isn't intesting to us. We know that 219 // this is safe in the face of spelling differences, because there is no way 220 // to spell an i/e in a strange way that is another letter. Skipping this 221 // allows us to avoid looking up the identifier info for #define/#undef and 222 // other common directives. 223 bool Invalid = false; 224 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation(), 225 &Invalid); 226 if (Invalid) 227 return; 228 229 char FirstChar = RawCharData[0]; 230 if (FirstChar >= 'a' && FirstChar <= 'z' && 231 FirstChar != 'i' && FirstChar != 'e') { 232 CurPPLexer->ParsingPreprocessorDirective = false; 233 // Restore comment saving mode. 234 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 235 continue; 236 } 237 238 // Get the identifier name without trigraphs or embedded newlines. Note 239 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 240 // when skipping. 241 char DirectiveBuf[20]; 242 llvm::StringRef Directive; 243 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 244 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 245 } else { 246 std::string DirectiveStr = getSpelling(Tok); 247 unsigned IdLen = DirectiveStr.size(); 248 if (IdLen >= 20) { 249 CurPPLexer->ParsingPreprocessorDirective = false; 250 // Restore comment saving mode. 251 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 252 continue; 253 } 254 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 255 Directive = llvm::StringRef(DirectiveBuf, IdLen); 256 } 257 258 if (Directive.startswith("if")) { 259 llvm::StringRef Sub = Directive.substr(2); 260 if (Sub.empty() || // "if" 261 Sub == "def" || // "ifdef" 262 Sub == "ndef") { // "ifndef" 263 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 264 // bother parsing the condition. 265 DiscardUntilEndOfDirective(); 266 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 267 /*foundnonskip*/false, 268 /*fnddelse*/false); 269 } 270 } else if (Directive[0] == 'e') { 271 llvm::StringRef Sub = Directive.substr(1); 272 if (Sub == "ndif") { // "endif" 273 CheckEndOfDirective("endif"); 274 PPConditionalInfo CondInfo; 275 CondInfo.WasSkipping = true; // Silence bogus warning. 276 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 277 InCond = InCond; // Silence warning in no-asserts mode. 278 assert(!InCond && "Can't be skipping if not in a conditional!"); 279 280 // If we popped the outermost skipping block, we're done skipping! 281 if (!CondInfo.WasSkipping) 282 break; 283 } else if (Sub == "lse") { // "else". 284 // #else directive in a skipping conditional. If not in some other 285 // skipping conditional, and if #else hasn't already been seen, enter it 286 // as a non-skipping conditional. 287 DiscardUntilEndOfDirective(); // C99 6.10p4. 288 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 289 290 // If this is a #else with a #else before it, report the error. 291 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 292 293 // Note that we've seen a #else in this conditional. 294 CondInfo.FoundElse = true; 295 296 // If the conditional is at the top level, and the #if block wasn't 297 // entered, enter the #else block now. 298 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 299 CondInfo.FoundNonSkip = true; 300 break; 301 } 302 } else if (Sub == "lif") { // "elif". 303 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 304 305 bool ShouldEnter; 306 // If this is in a skipping block or if we're already handled this #if 307 // block, don't bother parsing the condition. 308 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 309 DiscardUntilEndOfDirective(); 310 ShouldEnter = false; 311 } else { 312 // Restore the value of LexingRawMode so that identifiers are 313 // looked up, etc, inside the #elif expression. 314 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 315 CurPPLexer->LexingRawMode = false; 316 IdentifierInfo *IfNDefMacro = 0; 317 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 318 CurPPLexer->LexingRawMode = true; 319 } 320 321 // If this is a #elif with a #else before it, report the error. 322 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 323 324 // If this condition is true, enter it! 325 if (ShouldEnter) { 326 CondInfo.FoundNonSkip = true; 327 break; 328 } 329 } 330 } 331 332 CurPPLexer->ParsingPreprocessorDirective = false; 333 // Restore comment saving mode. 334 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 335 } 336 337 // Finally, if we are out of the conditional (saw an #endif or ran off the end 338 // of the file, just stop skipping and return to lexing whatever came after 339 // the #if block. 340 CurPPLexer->LexingRawMode = false; 341} 342 343void Preprocessor::PTHSkipExcludedConditionalBlock() { 344 345 while (1) { 346 assert(CurPTHLexer); 347 assert(CurPTHLexer->LexingRawMode == false); 348 349 // Skip to the next '#else', '#elif', or #endif. 350 if (CurPTHLexer->SkipBlock()) { 351 // We have reached an #endif. Both the '#' and 'endif' tokens 352 // have been consumed by the PTHLexer. Just pop off the condition level. 353 PPConditionalInfo CondInfo; 354 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 355 InCond = InCond; // Silence warning in no-asserts mode. 356 assert(!InCond && "Can't be skipping if not in a conditional!"); 357 break; 358 } 359 360 // We have reached a '#else' or '#elif'. Lex the next token to get 361 // the directive flavor. 362 Token Tok; 363 LexUnexpandedToken(Tok); 364 365 // We can actually look up the IdentifierInfo here since we aren't in 366 // raw mode. 367 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 368 369 if (K == tok::pp_else) { 370 // #else: Enter the else condition. We aren't in a nested condition 371 // since we skip those. We're always in the one matching the last 372 // blocked we skipped. 373 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 374 // Note that we've seen a #else in this conditional. 375 CondInfo.FoundElse = true; 376 377 // If the #if block wasn't entered then enter the #else block now. 378 if (!CondInfo.FoundNonSkip) { 379 CondInfo.FoundNonSkip = true; 380 381 // Scan until the eom token. 382 CurPTHLexer->ParsingPreprocessorDirective = true; 383 DiscardUntilEndOfDirective(); 384 CurPTHLexer->ParsingPreprocessorDirective = false; 385 386 break; 387 } 388 389 // Otherwise skip this block. 390 continue; 391 } 392 393 assert(K == tok::pp_elif); 394 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 395 396 // If this is a #elif with a #else before it, report the error. 397 if (CondInfo.FoundElse) 398 Diag(Tok, diag::pp_err_elif_after_else); 399 400 // If this is in a skipping block or if we're already handled this #if 401 // block, don't bother parsing the condition. We just skip this block. 402 if (CondInfo.FoundNonSkip) 403 continue; 404 405 // Evaluate the condition of the #elif. 406 IdentifierInfo *IfNDefMacro = 0; 407 CurPTHLexer->ParsingPreprocessorDirective = true; 408 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 409 CurPTHLexer->ParsingPreprocessorDirective = false; 410 411 // If this condition is true, enter it! 412 if (ShouldEnter) { 413 CondInfo.FoundNonSkip = true; 414 break; 415 } 416 417 // Otherwise, skip this block and go to the next one. 418 continue; 419 } 420} 421 422/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 423/// return null on failure. isAngled indicates whether the file reference is 424/// for system #include's or not (i.e. using <> instead of ""). 425const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 426 bool isAngled, 427 const DirectoryLookup *FromDir, 428 const DirectoryLookup *&CurDir) { 429 // If the header lookup mechanism may be relative to the current file, pass in 430 // info about where the current file is. 431 const FileEntry *CurFileEnt = 0; 432 if (!FromDir) { 433 FileID FID = getCurrentFileLexer()->getFileID(); 434 CurFileEnt = SourceMgr.getFileEntryForID(FID); 435 436 // If there is no file entry associated with this file, it must be the 437 // predefines buffer. Any other file is not lexed with a normal lexer, so 438 // it won't be scanned for preprocessor directives. If we have the 439 // predefines buffer, resolve #include references (which come from the 440 // -include command line argument) as if they came from the main file, this 441 // affects file lookup etc. 442 if (CurFileEnt == 0) { 443 FID = SourceMgr.getMainFileID(); 444 CurFileEnt = SourceMgr.getFileEntryForID(FID); 445 } 446 } 447 448 // Do a standard file entry lookup. 449 CurDir = CurDirLookup; 450 const FileEntry *FE = 451 HeaderInfo.LookupFile(Filename, isAngled, FromDir, CurDir, CurFileEnt); 452 if (FE) return FE; 453 454 // Otherwise, see if this is a subframework header. If so, this is relative 455 // to one of the headers on the #include stack. Walk the list of the current 456 // headers on the #include stack and pass them to HeaderInfo. 457 if (IsFileLexer()) { 458 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 459 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 460 return FE; 461 } 462 463 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 464 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 465 if (IsFileLexer(ISEntry)) { 466 if ((CurFileEnt = 467 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 468 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 469 return FE; 470 } 471 } 472 473 // Otherwise, we really couldn't find the file. 474 return 0; 475} 476 477 478//===----------------------------------------------------------------------===// 479// Preprocessor Directive Handling. 480//===----------------------------------------------------------------------===// 481 482/// HandleDirective - This callback is invoked when the lexer sees a # token 483/// at the start of a line. This consumes the directive, modifies the 484/// lexer/preprocessor state, and advances the lexer(s) so that the next token 485/// read is the correct one. 486void Preprocessor::HandleDirective(Token &Result) { 487 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 488 489 // We just parsed a # character at the start of a line, so we're in directive 490 // mode. Tell the lexer this so any newlines we see will be converted into an 491 // EOM token (which terminates the directive). 492 CurPPLexer->ParsingPreprocessorDirective = true; 493 494 ++NumDirectives; 495 496 // We are about to read a token. For the multiple-include optimization FA to 497 // work, we have to remember if we had read any tokens *before* this 498 // pp-directive. 499 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 500 501 // Save the '#' token in case we need to return it later. 502 Token SavedHash = Result; 503 504 // Read the next token, the directive flavor. This isn't expanded due to 505 // C99 6.10.3p8. 506 LexUnexpandedToken(Result); 507 508 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 509 // #define A(x) #x 510 // A(abc 511 // #warning blah 512 // def) 513 // If so, the user is relying on non-portable behavior, emit a diagnostic. 514 if (InMacroArgs) 515 Diag(Result, diag::ext_embedded_directive); 516 517TryAgain: 518 switch (Result.getKind()) { 519 case tok::eom: 520 return; // null directive. 521 case tok::comment: 522 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 523 LexUnexpandedToken(Result); 524 goto TryAgain; 525 526 case tok::numeric_constant: // # 7 GNU line marker directive. 527 if (getLangOptions().AsmPreprocessor) 528 break; // # 4 is not a preprocessor directive in .S files. 529 return HandleDigitDirective(Result); 530 default: 531 IdentifierInfo *II = Result.getIdentifierInfo(); 532 if (II == 0) break; // Not an identifier. 533 534 // Ask what the preprocessor keyword ID is. 535 switch (II->getPPKeywordID()) { 536 default: break; 537 // C99 6.10.1 - Conditional Inclusion. 538 case tok::pp_if: 539 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 540 case tok::pp_ifdef: 541 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 542 case tok::pp_ifndef: 543 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 544 case tok::pp_elif: 545 return HandleElifDirective(Result); 546 case tok::pp_else: 547 return HandleElseDirective(Result); 548 case tok::pp_endif: 549 return HandleEndifDirective(Result); 550 551 // C99 6.10.2 - Source File Inclusion. 552 case tok::pp_include: 553 return HandleIncludeDirective(Result); // Handle #include. 554 case tok::pp___include_macros: 555 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 556 557 // C99 6.10.3 - Macro Replacement. 558 case tok::pp_define: 559 return HandleDefineDirective(Result); 560 case tok::pp_undef: 561 return HandleUndefDirective(Result); 562 563 // C99 6.10.4 - Line Control. 564 case tok::pp_line: 565 return HandleLineDirective(Result); 566 567 // C99 6.10.5 - Error Directive. 568 case tok::pp_error: 569 return HandleUserDiagnosticDirective(Result, false); 570 571 // C99 6.10.6 - Pragma Directive. 572 case tok::pp_pragma: 573 return HandlePragmaDirective(); 574 575 // GNU Extensions. 576 case tok::pp_import: 577 return HandleImportDirective(Result); 578 case tok::pp_include_next: 579 return HandleIncludeNextDirective(Result); 580 581 case tok::pp_warning: 582 Diag(Result, diag::ext_pp_warning_directive); 583 return HandleUserDiagnosticDirective(Result, true); 584 case tok::pp_ident: 585 return HandleIdentSCCSDirective(Result); 586 case tok::pp_sccs: 587 return HandleIdentSCCSDirective(Result); 588 case tok::pp_assert: 589 //isExtension = true; // FIXME: implement #assert 590 break; 591 case tok::pp_unassert: 592 //isExtension = true; // FIXME: implement #unassert 593 break; 594 } 595 break; 596 } 597 598 // If this is a .S file, treat unknown # directives as non-preprocessor 599 // directives. This is important because # may be a comment or introduce 600 // various pseudo-ops. Just return the # token and push back the following 601 // token to be lexed next time. 602 if (getLangOptions().AsmPreprocessor) { 603 Token *Toks = new Token[2]; 604 // Return the # and the token after it. 605 Toks[0] = SavedHash; 606 Toks[1] = Result; 607 // Enter this token stream so that we re-lex the tokens. Make sure to 608 // enable macro expansion, in case the token after the # is an identifier 609 // that is expanded. 610 EnterTokenStream(Toks, 2, false, true); 611 return; 612 } 613 614 // If we reached here, the preprocessing token is not valid! 615 Diag(Result, diag::err_pp_invalid_directive); 616 617 // Read the rest of the PP line. 618 DiscardUntilEndOfDirective(); 619 620 // Okay, we're done parsing the directive. 621} 622 623/// GetLineValue - Convert a numeric token into an unsigned value, emitting 624/// Diagnostic DiagID if it is invalid, and returning the value in Val. 625static bool GetLineValue(Token &DigitTok, unsigned &Val, 626 unsigned DiagID, Preprocessor &PP) { 627 if (DigitTok.isNot(tok::numeric_constant)) { 628 PP.Diag(DigitTok, DiagID); 629 630 if (DigitTok.isNot(tok::eom)) 631 PP.DiscardUntilEndOfDirective(); 632 return true; 633 } 634 635 llvm::SmallString<64> IntegerBuffer; 636 IntegerBuffer.resize(DigitTok.getLength()); 637 const char *DigitTokBegin = &IntegerBuffer[0]; 638 bool Invalid = false; 639 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid); 640 if (Invalid) 641 return true; 642 643 // Verify that we have a simple digit-sequence, and compute the value. This 644 // is always a simple digit string computed in decimal, so we do this manually 645 // here. 646 Val = 0; 647 for (unsigned i = 0; i != ActualLength; ++i) { 648 if (!isdigit(DigitTokBegin[i])) { 649 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 650 diag::err_pp_line_digit_sequence); 651 PP.DiscardUntilEndOfDirective(); 652 return true; 653 } 654 655 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 656 if (NextVal < Val) { // overflow. 657 PP.Diag(DigitTok, DiagID); 658 PP.DiscardUntilEndOfDirective(); 659 return true; 660 } 661 Val = NextVal; 662 } 663 664 // Reject 0, this is needed both by #line numbers and flags. 665 if (Val == 0) { 666 PP.Diag(DigitTok, DiagID); 667 PP.DiscardUntilEndOfDirective(); 668 return true; 669 } 670 671 if (DigitTokBegin[0] == '0') 672 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 673 674 return false; 675} 676 677/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 678/// acceptable forms are: 679/// # line digit-sequence 680/// # line digit-sequence "s-char-sequence" 681void Preprocessor::HandleLineDirective(Token &Tok) { 682 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 683 // expanded. 684 Token DigitTok; 685 Lex(DigitTok); 686 687 // Validate the number and convert it to an unsigned. 688 unsigned LineNo; 689 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 690 return; 691 692 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 693 // number greater than 2147483647". C90 requires that the line # be <= 32767. 694 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 695 if (LineNo >= LineLimit) 696 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 697 698 int FilenameID = -1; 699 Token StrTok; 700 Lex(StrTok); 701 702 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 703 // string followed by eom. 704 if (StrTok.is(tok::eom)) 705 ; // ok 706 else if (StrTok.isNot(tok::string_literal)) { 707 Diag(StrTok, diag::err_pp_line_invalid_filename); 708 DiscardUntilEndOfDirective(); 709 return; 710 } else { 711 // Parse and validate the string, converting it into a unique ID. 712 StringLiteralParser Literal(&StrTok, 1, *this); 713 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 714 if (Literal.hadError) 715 return DiscardUntilEndOfDirective(); 716 if (Literal.Pascal) { 717 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 718 return DiscardUntilEndOfDirective(); 719 } 720 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 721 Literal.GetStringLength()); 722 723 // Verify that there is nothing after the string, other than EOM. Because 724 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 725 CheckEndOfDirective("line", true); 726 } 727 728 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 729 730 if (Callbacks) 731 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), 732 PPCallbacks::RenameFile, 733 SrcMgr::C_User); 734} 735 736/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 737/// marker directive. 738static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 739 bool &IsSystemHeader, bool &IsExternCHeader, 740 Preprocessor &PP) { 741 unsigned FlagVal; 742 Token FlagTok; 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; 747 748 if (FlagVal == 1) { 749 IsFileEntry = true; 750 751 PP.Lex(FlagTok); 752 if (FlagTok.is(tok::eom)) return false; 753 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 754 return true; 755 } else if (FlagVal == 2) { 756 IsFileExit = true; 757 758 SourceManager &SM = PP.getSourceManager(); 759 // If we are leaving the current presumed file, check to make sure the 760 // presumed include stack isn't empty! 761 FileID CurFileID = 762 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 763 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 764 765 // If there is no include loc (main file) or if the include loc is in a 766 // different physical file, then we aren't in a "1" line marker flag region. 767 SourceLocation IncLoc = PLoc.getIncludeLoc(); 768 if (IncLoc.isInvalid() || 769 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 770 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 771 PP.DiscardUntilEndOfDirective(); 772 return true; 773 } 774 775 PP.Lex(FlagTok); 776 if (FlagTok.is(tok::eom)) return false; 777 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 778 return true; 779 } 780 781 // We must have 3 if there are still flags. 782 if (FlagVal != 3) { 783 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 784 PP.DiscardUntilEndOfDirective(); 785 return true; 786 } 787 788 IsSystemHeader = true; 789 790 PP.Lex(FlagTok); 791 if (FlagTok.is(tok::eom)) return false; 792 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 793 return true; 794 795 // We must have 4 if there is yet another flag. 796 if (FlagVal != 4) { 797 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 798 PP.DiscardUntilEndOfDirective(); 799 return true; 800 } 801 802 IsExternCHeader = true; 803 804 PP.Lex(FlagTok); 805 if (FlagTok.is(tok::eom)) return false; 806 807 // There are no more valid flags here. 808 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 809 PP.DiscardUntilEndOfDirective(); 810 return true; 811} 812 813/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 814/// one of the following forms: 815/// 816/// # 42 817/// # 42 "file" ('1' | '2')? 818/// # 42 "file" ('1' | '2')? '3' '4'? 819/// 820void Preprocessor::HandleDigitDirective(Token &DigitTok) { 821 // Validate the number and convert it to an unsigned. GNU does not have a 822 // line # limit other than it fit in 32-bits. 823 unsigned LineNo; 824 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 825 *this)) 826 return; 827 828 Token StrTok; 829 Lex(StrTok); 830 831 bool IsFileEntry = false, IsFileExit = false; 832 bool IsSystemHeader = false, IsExternCHeader = false; 833 int FilenameID = -1; 834 835 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 836 // string followed by eom. 837 if (StrTok.is(tok::eom)) 838 ; // ok 839 else if (StrTok.isNot(tok::string_literal)) { 840 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 841 return DiscardUntilEndOfDirective(); 842 } else { 843 // Parse and validate the string, converting it into a unique ID. 844 StringLiteralParser Literal(&StrTok, 1, *this); 845 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 846 if (Literal.hadError) 847 return DiscardUntilEndOfDirective(); 848 if (Literal.Pascal) { 849 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 850 return DiscardUntilEndOfDirective(); 851 } 852 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 853 Literal.GetStringLength()); 854 855 // If a filename was present, read any flags that are present. 856 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 857 IsSystemHeader, IsExternCHeader, *this)) 858 return; 859 } 860 861 // Create a line note with this information. 862 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 863 IsFileEntry, IsFileExit, 864 IsSystemHeader, IsExternCHeader); 865 866 // If the preprocessor has callbacks installed, notify them of the #line 867 // change. This is used so that the line marker comes out in -E mode for 868 // example. 869 if (Callbacks) { 870 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 871 if (IsFileEntry) 872 Reason = PPCallbacks::EnterFile; 873 else if (IsFileExit) 874 Reason = PPCallbacks::ExitFile; 875 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 876 if (IsExternCHeader) 877 FileKind = SrcMgr::C_ExternCSystem; 878 else if (IsSystemHeader) 879 FileKind = SrcMgr::C_System; 880 881 Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind); 882 } 883} 884 885 886/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 887/// 888void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 889 bool isWarning) { 890 // PTH doesn't emit #warning or #error directives. 891 if (CurPTHLexer) 892 return CurPTHLexer->DiscardToEndOfLine(); 893 894 // Read the rest of the line raw. We do this because we don't want macros 895 // to be expanded and we don't require that the tokens be valid preprocessing 896 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 897 // collapse multiple consequtive white space between tokens, but this isn't 898 // specified by the standard. 899 std::string Message = CurLexer->ReadToEndOfLine(); 900 if (isWarning) 901 Diag(Tok, diag::pp_hash_warning) << Message; 902 else 903 Diag(Tok, diag::err_pp_hash_error) << Message; 904} 905 906/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 907/// 908void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 909 // Yes, this directive is an extension. 910 Diag(Tok, diag::ext_pp_ident_directive); 911 912 // Read the string argument. 913 Token StrTok; 914 Lex(StrTok); 915 916 // If the token kind isn't a string, it's a malformed directive. 917 if (StrTok.isNot(tok::string_literal) && 918 StrTok.isNot(tok::wide_string_literal)) { 919 Diag(StrTok, diag::err_pp_malformed_ident); 920 if (StrTok.isNot(tok::eom)) 921 DiscardUntilEndOfDirective(); 922 return; 923 } 924 925 // Verify that there is nothing after the string, other than EOM. 926 CheckEndOfDirective("ident"); 927 928 if (Callbacks) { 929 bool Invalid = false; 930 std::string Str = getSpelling(StrTok, &Invalid); 931 if (!Invalid) 932 Callbacks->Ident(Tok.getLocation(), Str); 933 } 934} 935 936//===----------------------------------------------------------------------===// 937// Preprocessor Include Directive Handling. 938//===----------------------------------------------------------------------===// 939 940/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 941/// checked and spelled filename, e.g. as an operand of #include. This returns 942/// true if the input filename was in <>'s or false if it were in ""'s. The 943/// caller is expected to provide a buffer that is large enough to hold the 944/// spelling of the filename, but is also expected to handle the case when 945/// this method decides to use a different buffer. 946bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 947 llvm::StringRef &Buffer) { 948 // Get the text form of the filename. 949 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 950 951 // Make sure the filename is <x> or "x". 952 bool isAngled; 953 if (Buffer[0] == '<') { 954 if (Buffer.back() != '>') { 955 Diag(Loc, diag::err_pp_expects_filename); 956 Buffer = llvm::StringRef(); 957 return true; 958 } 959 isAngled = true; 960 } else if (Buffer[0] == '"') { 961 if (Buffer.back() != '"') { 962 Diag(Loc, diag::err_pp_expects_filename); 963 Buffer = llvm::StringRef(); 964 return true; 965 } 966 isAngled = false; 967 } else { 968 Diag(Loc, diag::err_pp_expects_filename); 969 Buffer = llvm::StringRef(); 970 return true; 971 } 972 973 // Diagnose #include "" as invalid. 974 if (Buffer.size() <= 2) { 975 Diag(Loc, diag::err_pp_empty_filename); 976 Buffer = llvm::StringRef(); 977 return true; 978 } 979 980 // Skip the brackets. 981 Buffer = Buffer.substr(1, Buffer.size()-2); 982 return isAngled; 983} 984 985/// ConcatenateIncludeName - Handle cases where the #include name is expanded 986/// from a macro as multiple tokens, which need to be glued together. This 987/// occurs for code like: 988/// #define FOO <a/b.h> 989/// #include FOO 990/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 991/// 992/// This code concatenates and consumes tokens up to the '>' token. It returns 993/// false if the > was found, otherwise it returns true if it finds and consumes 994/// the EOM marker. 995bool Preprocessor::ConcatenateIncludeName( 996 llvm::SmallString<128> &FilenameBuffer) { 997 Token CurTok; 998 999 Lex(CurTok); 1000 while (CurTok.isNot(tok::eom)) { 1001 // Append the spelling of this token to the buffer. If there was a space 1002 // before it, add it now. 1003 if (CurTok.hasLeadingSpace()) 1004 FilenameBuffer.push_back(' '); 1005 1006 // Get the spelling of the token, directly into FilenameBuffer if possible. 1007 unsigned PreAppendSize = FilenameBuffer.size(); 1008 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 1009 1010 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 1011 unsigned ActualLen = getSpelling(CurTok, BufPtr); 1012 1013 // If the token was spelled somewhere else, copy it into FilenameBuffer. 1014 if (BufPtr != &FilenameBuffer[PreAppendSize]) 1015 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 1016 1017 // Resize FilenameBuffer to the correct size. 1018 if (CurTok.getLength() != ActualLen) 1019 FilenameBuffer.resize(PreAppendSize+ActualLen); 1020 1021 // If we found the '>' marker, return success. 1022 if (CurTok.is(tok::greater)) 1023 return false; 1024 1025 Lex(CurTok); 1026 } 1027 1028 // If we hit the eom marker, emit an error and return true so that the caller 1029 // knows the EOM has been read. 1030 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1031 return true; 1032} 1033 1034/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1035/// file to be included from the lexer, then include it! This is a common 1036/// routine with functionality shared between #include, #include_next and 1037/// #import. LookupFrom is set when this is a #include_next directive, it 1038/// specifies the file to start searching from. 1039void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1040 const DirectoryLookup *LookupFrom, 1041 bool isImport) { 1042 1043 Token FilenameTok; 1044 CurPPLexer->LexIncludeFilename(FilenameTok); 1045 1046 // Reserve a buffer to get the spelling. 1047 llvm::SmallString<128> FilenameBuffer; 1048 llvm::StringRef Filename; 1049 1050 switch (FilenameTok.getKind()) { 1051 case tok::eom: 1052 // If the token kind is EOM, the error has already been diagnosed. 1053 return; 1054 1055 case tok::angle_string_literal: 1056 case tok::string_literal: 1057 Filename = getSpelling(FilenameTok, FilenameBuffer); 1058 break; 1059 1060 case tok::less: 1061 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1062 // case, glue the tokens together into FilenameBuffer and interpret those. 1063 FilenameBuffer.push_back('<'); 1064 if (ConcatenateIncludeName(FilenameBuffer)) 1065 return; // Found <eom> but no ">"? Diagnostic already emitted. 1066 Filename = FilenameBuffer.str(); 1067 break; 1068 default: 1069 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1070 DiscardUntilEndOfDirective(); 1071 return; 1072 } 1073 1074 bool isAngled = 1075 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1076 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1077 // error. 1078 if (Filename.empty()) { 1079 DiscardUntilEndOfDirective(); 1080 return; 1081 } 1082 1083 // Verify that there is nothing after the filename, other than EOM. Note that 1084 // we allow macros that expand to nothing after the filename, because this 1085 // falls into the category of "#include pp-tokens new-line" specified in 1086 // C99 6.10.2p4. 1087 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1088 1089 // Check that we don't have infinite #include recursion. 1090 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1091 Diag(FilenameTok, diag::err_pp_include_too_deep); 1092 return; 1093 } 1094 1095 // Search include directories. 1096 const DirectoryLookup *CurDir; 1097 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1098 if (File == 0) { 1099 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1100 return; 1101 } 1102 1103 // The #included file will be considered to be a system header if either it is 1104 // in a system include directory, or if the #includer is a system include 1105 // header. 1106 SrcMgr::CharacteristicKind FileCharacter = 1107 std::max(HeaderInfo.getFileDirFlavor(File), 1108 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1109 1110 // Ask HeaderInfo if we should enter this #include file. If not, #including 1111 // this file will have no effect. 1112 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 1113 if (Callbacks) 1114 Callbacks->FileSkipped(*File, FilenameTok, FileCharacter); 1115 return; 1116 } 1117 1118 // Look up the file, create a File ID for it. 1119 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1120 FileCharacter); 1121 if (FID.isInvalid()) { 1122 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1123 return; 1124 } 1125 1126 // Finally, if all is good, enter the new file! 1127 EnterSourceFile(FID, CurDir, FilenameTok.getLocation()); 1128} 1129 1130/// HandleIncludeNextDirective - Implements #include_next. 1131/// 1132void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1133 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1134 1135 // #include_next is like #include, except that we start searching after 1136 // the current found directory. If we can't do this, issue a 1137 // diagnostic. 1138 const DirectoryLookup *Lookup = CurDirLookup; 1139 if (isInPrimaryFile()) { 1140 Lookup = 0; 1141 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1142 } else if (Lookup == 0) { 1143 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1144 } else { 1145 // Start looking up in the next directory. 1146 ++Lookup; 1147 } 1148 1149 return HandleIncludeDirective(IncludeNextTok, Lookup); 1150} 1151 1152/// HandleImportDirective - Implements #import. 1153/// 1154void Preprocessor::HandleImportDirective(Token &ImportTok) { 1155 if (!Features.ObjC1) // #import is standard for ObjC. 1156 Diag(ImportTok, diag::ext_pp_import_directive); 1157 1158 return HandleIncludeDirective(ImportTok, 0, true); 1159} 1160 1161/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1162/// pseudo directive in the predefines buffer. This handles it by sucking all 1163/// tokens through the preprocessor and discarding them (only keeping the side 1164/// effects on the preprocessor). 1165void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1166 // This directive should only occur in the predefines buffer. If not, emit an 1167 // error and reject it. 1168 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1169 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1170 Diag(IncludeMacrosTok.getLocation(), 1171 diag::pp_include_macros_out_of_predefines); 1172 DiscardUntilEndOfDirective(); 1173 return; 1174 } 1175 1176 // Treat this as a normal #include for checking purposes. If this is 1177 // successful, it will push a new lexer onto the include stack. 1178 HandleIncludeDirective(IncludeMacrosTok, 0, false); 1179 1180 Token TmpTok; 1181 do { 1182 Lex(TmpTok); 1183 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1184 } while (TmpTok.isNot(tok::hashhash)); 1185} 1186 1187//===----------------------------------------------------------------------===// 1188// Preprocessor Macro Directive Handling. 1189//===----------------------------------------------------------------------===// 1190 1191/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1192/// definition has just been read. Lex the rest of the arguments and the 1193/// closing ), updating MI with what we learn. Return true if an error occurs 1194/// parsing the arg list. 1195bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1196 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1197 1198 Token Tok; 1199 while (1) { 1200 LexUnexpandedToken(Tok); 1201 switch (Tok.getKind()) { 1202 case tok::r_paren: 1203 // Found the end of the argument list. 1204 if (Arguments.empty()) // #define FOO() 1205 return false; 1206 // Otherwise we have #define FOO(A,) 1207 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1208 return true; 1209 case tok::ellipsis: // #define X(... -> C99 varargs 1210 // Warn if use of C99 feature in non-C99 mode. 1211 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1212 1213 // Lex the token after the identifier. 1214 LexUnexpandedToken(Tok); 1215 if (Tok.isNot(tok::r_paren)) { 1216 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1217 return true; 1218 } 1219 // Add the __VA_ARGS__ identifier as an argument. 1220 Arguments.push_back(Ident__VA_ARGS__); 1221 MI->setIsC99Varargs(); 1222 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1223 return false; 1224 case tok::eom: // #define X( 1225 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1226 return true; 1227 default: 1228 // Handle keywords and identifiers here to accept things like 1229 // #define Foo(for) for. 1230 IdentifierInfo *II = Tok.getIdentifierInfo(); 1231 if (II == 0) { 1232 // #define X(1 1233 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1234 return true; 1235 } 1236 1237 // If this is already used as an argument, it is used multiple times (e.g. 1238 // #define X(A,A. 1239 if (std::find(Arguments.begin(), Arguments.end(), II) != 1240 Arguments.end()) { // C99 6.10.3p6 1241 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1242 return true; 1243 } 1244 1245 // Add the argument to the macro info. 1246 Arguments.push_back(II); 1247 1248 // Lex the token after the identifier. 1249 LexUnexpandedToken(Tok); 1250 1251 switch (Tok.getKind()) { 1252 default: // #define X(A B 1253 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1254 return true; 1255 case tok::r_paren: // #define X(A) 1256 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1257 return false; 1258 case tok::comma: // #define X(A, 1259 break; 1260 case tok::ellipsis: // #define X(A... -> GCC extension 1261 // Diagnose extension. 1262 Diag(Tok, diag::ext_named_variadic_macro); 1263 1264 // Lex the token after the identifier. 1265 LexUnexpandedToken(Tok); 1266 if (Tok.isNot(tok::r_paren)) { 1267 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1268 return true; 1269 } 1270 1271 MI->setIsGNUVarargs(); 1272 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1273 return false; 1274 } 1275 } 1276 } 1277} 1278 1279/// HandleDefineDirective - Implements #define. This consumes the entire macro 1280/// line then lets the caller lex the next real token. 1281void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1282 ++NumDefined; 1283 1284 Token MacroNameTok; 1285 ReadMacroName(MacroNameTok, 1); 1286 1287 // Error reading macro name? If so, diagnostic already issued. 1288 if (MacroNameTok.is(tok::eom)) 1289 return; 1290 1291 Token LastTok = MacroNameTok; 1292 1293 // If we are supposed to keep comments in #defines, reenable comment saving 1294 // mode. 1295 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1296 1297 // Create the new macro. 1298 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1299 1300 Token Tok; 1301 LexUnexpandedToken(Tok); 1302 1303 // If this is a function-like macro definition, parse the argument list, 1304 // marking each of the identifiers as being used as macro arguments. Also, 1305 // check other constraints on the first token of the macro body. 1306 if (Tok.is(tok::eom)) { 1307 // If there is no body to this macro, we have no special handling here. 1308 } else if (Tok.hasLeadingSpace()) { 1309 // This is a normal token with leading space. Clear the leading space 1310 // marker on the first token to get proper expansion. 1311 Tok.clearFlag(Token::LeadingSpace); 1312 } else if (Tok.is(tok::l_paren)) { 1313 // This is a function-like macro definition. Read the argument list. 1314 MI->setIsFunctionLike(); 1315 if (ReadMacroDefinitionArgList(MI)) { 1316 // Forget about MI. 1317 ReleaseMacroInfo(MI); 1318 // Throw away the rest of the line. 1319 if (CurPPLexer->ParsingPreprocessorDirective) 1320 DiscardUntilEndOfDirective(); 1321 return; 1322 } 1323 1324 // If this is a definition of a variadic C99 function-like macro, not using 1325 // the GNU named varargs extension, enabled __VA_ARGS__. 1326 1327 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1328 // This gets unpoisoned where it is allowed. 1329 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1330 if (MI->isC99Varargs()) 1331 Ident__VA_ARGS__->setIsPoisoned(false); 1332 1333 // Read the first token after the arg list for down below. 1334 LexUnexpandedToken(Tok); 1335 } else if (Features.C99) { 1336 // C99 requires whitespace between the macro definition and the body. Emit 1337 // a diagnostic for something like "#define X+". 1338 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1339 } else { 1340 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1341 // first character of a replacement list is not a character required by 1342 // subclause 5.2.1, then there shall be white-space separation between the 1343 // identifier and the replacement list.". 5.2.1 lists this set: 1344 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1345 // is irrelevant here. 1346 bool isInvalid = false; 1347 if (Tok.is(tok::at)) // @ is not in the list above. 1348 isInvalid = true; 1349 else if (Tok.is(tok::unknown)) { 1350 // If we have an unknown token, it is something strange like "`". Since 1351 // all of valid characters would have lexed into a single character 1352 // token of some sort, we know this is not a valid case. 1353 isInvalid = true; 1354 } 1355 if (isInvalid) 1356 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1357 else 1358 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1359 } 1360 1361 if (!Tok.is(tok::eom)) 1362 LastTok = Tok; 1363 1364 // Read the rest of the macro body. 1365 if (MI->isObjectLike()) { 1366 // Object-like macros are very simple, just read their body. 1367 while (Tok.isNot(tok::eom)) { 1368 LastTok = Tok; 1369 MI->AddTokenToBody(Tok); 1370 // Get the next token of the macro. 1371 LexUnexpandedToken(Tok); 1372 } 1373 1374 } else { 1375 // Otherwise, read the body of a function-like macro. While we are at it, 1376 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1377 // parameters in function-like macro expansions. 1378 while (Tok.isNot(tok::eom)) { 1379 LastTok = Tok; 1380 1381 if (Tok.isNot(tok::hash)) { 1382 MI->AddTokenToBody(Tok); 1383 1384 // Get the next token of the macro. 1385 LexUnexpandedToken(Tok); 1386 continue; 1387 } 1388 1389 // Get the next token of the macro. 1390 LexUnexpandedToken(Tok); 1391 1392 // Check for a valid macro arg identifier. 1393 if (Tok.getIdentifierInfo() == 0 || 1394 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1395 1396 // If this is assembler-with-cpp mode, we accept random gibberish after 1397 // the '#' because '#' is often a comment character. However, change 1398 // the kind of the token to tok::unknown so that the preprocessor isn't 1399 // confused. 1400 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1401 LastTok.setKind(tok::unknown); 1402 } else { 1403 Diag(Tok, diag::err_pp_stringize_not_parameter); 1404 ReleaseMacroInfo(MI); 1405 1406 // Disable __VA_ARGS__ again. 1407 Ident__VA_ARGS__->setIsPoisoned(true); 1408 return; 1409 } 1410 } 1411 1412 // Things look ok, add the '#' and param name tokens to the macro. 1413 MI->AddTokenToBody(LastTok); 1414 MI->AddTokenToBody(Tok); 1415 LastTok = Tok; 1416 1417 // Get the next token of the macro. 1418 LexUnexpandedToken(Tok); 1419 } 1420 } 1421 1422 1423 // Disable __VA_ARGS__ again. 1424 Ident__VA_ARGS__->setIsPoisoned(true); 1425 1426 // Check that there is no paste (##) operator at the begining or end of the 1427 // replacement list. 1428 unsigned NumTokens = MI->getNumTokens(); 1429 if (NumTokens != 0) { 1430 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1431 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1432 ReleaseMacroInfo(MI); 1433 return; 1434 } 1435 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1436 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1437 ReleaseMacroInfo(MI); 1438 return; 1439 } 1440 } 1441 1442 // If this is the primary source file, remember that this macro hasn't been 1443 // used yet. 1444 if (isInPrimaryFile()) 1445 MI->setIsUsed(false); 1446 1447 MI->setDefinitionEndLoc(LastTok.getLocation()); 1448 1449 // Finally, if this identifier already had a macro defined for it, verify that 1450 // the macro bodies are identical and free the old definition. 1451 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1452 // It is very common for system headers to have tons of macro redefinitions 1453 // and for warnings to be disabled in system headers. If this is the case, 1454 // then don't bother calling MacroInfo::isIdenticalTo. 1455 if (!getDiagnostics().getSuppressSystemWarnings() || 1456 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1457 if (!OtherMI->isUsed()) 1458 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1459 1460 // Macros must be identical. This means all tokens and whitespace 1461 // separation must be the same. C99 6.10.3.2. 1462 if (!OtherMI->isAllowRedefinitionsWithoutWarning() && 1463 !MI->isIdenticalTo(*OtherMI, *this)) { 1464 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1465 << MacroNameTok.getIdentifierInfo(); 1466 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1467 } 1468 } 1469 ReleaseMacroInfo(OtherMI); 1470 } 1471 1472 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1473 1474 // If the callbacks want to know, tell them about the macro definition. 1475 if (Callbacks) 1476 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1477} 1478 1479/// HandleUndefDirective - Implements #undef. 1480/// 1481void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1482 ++NumUndefined; 1483 1484 Token MacroNameTok; 1485 ReadMacroName(MacroNameTok, 2); 1486 1487 // Error reading macro name? If so, diagnostic already issued. 1488 if (MacroNameTok.is(tok::eom)) 1489 return; 1490 1491 // Check to see if this is the last token on the #undef line. 1492 CheckEndOfDirective("undef"); 1493 1494 // Okay, we finally have a valid identifier to undef. 1495 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1496 1497 // If the macro is not defined, this is a noop undef, just return. 1498 if (MI == 0) return; 1499 1500 if (!MI->isUsed()) 1501 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1502 1503 // If the callbacks want to know, tell them about the macro #undef. 1504 if (Callbacks) 1505 Callbacks->MacroUndefined(MacroNameTok.getLocation(), 1506 MacroNameTok.getIdentifierInfo(), MI); 1507 1508 // Free macro definition. 1509 ReleaseMacroInfo(MI); 1510 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1511} 1512 1513 1514//===----------------------------------------------------------------------===// 1515// Preprocessor Conditional Directive Handling. 1516//===----------------------------------------------------------------------===// 1517 1518/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1519/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1520/// if any tokens have been returned or pp-directives activated before this 1521/// #ifndef has been lexed. 1522/// 1523void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1524 bool ReadAnyTokensBeforeDirective) { 1525 ++NumIf; 1526 Token DirectiveTok = Result; 1527 1528 Token MacroNameTok; 1529 ReadMacroName(MacroNameTok); 1530 1531 // Error reading macro name? If so, diagnostic already issued. 1532 if (MacroNameTok.is(tok::eom)) { 1533 // Skip code until we get to #endif. This helps with recovery by not 1534 // emitting an error when the #endif is reached. 1535 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1536 /*Foundnonskip*/false, /*FoundElse*/false); 1537 return; 1538 } 1539 1540 // Check to see if this is the last token on the #if[n]def line. 1541 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1542 1543 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1544 MacroInfo *MI = getMacroInfo(MII); 1545 1546 if (CurPPLexer->getConditionalStackDepth() == 0) { 1547 // If the start of a top-level #ifdef and if the macro is not defined, 1548 // inform MIOpt that this might be the start of a proper include guard. 1549 // Otherwise it is some other form of unknown conditional which we can't 1550 // handle. 1551 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1552 assert(isIfndef && "#ifdef shouldn't reach here"); 1553 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1554 } else 1555 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1556 } 1557 1558 // If there is a macro, process it. 1559 if (MI) // Mark it used. 1560 MI->setIsUsed(true); 1561 1562 // Should we include the stuff contained by this directive? 1563 if (!MI == isIfndef) { 1564 // Yes, remember that we are inside a conditional, then lex the next token. 1565 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1566 /*wasskip*/false, /*foundnonskip*/true, 1567 /*foundelse*/false); 1568 } else { 1569 // No, skip the contents of this block and return the first token after it. 1570 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1571 /*Foundnonskip*/false, 1572 /*FoundElse*/false); 1573 } 1574} 1575 1576/// HandleIfDirective - Implements the #if directive. 1577/// 1578void Preprocessor::HandleIfDirective(Token &IfToken, 1579 bool ReadAnyTokensBeforeDirective) { 1580 ++NumIf; 1581 1582 // Parse and evaluation the conditional expression. 1583 IdentifierInfo *IfNDefMacro = 0; 1584 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1585 1586 1587 // If this condition is equivalent to #ifndef X, and if this is the first 1588 // directive seen, handle it for the multiple-include optimization. 1589 if (CurPPLexer->getConditionalStackDepth() == 0) { 1590 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1591 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1592 else 1593 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1594 } 1595 1596 // Should we include the stuff contained by this directive? 1597 if (ConditionalTrue) { 1598 // Yes, remember that we are inside a conditional, then lex the next token. 1599 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1600 /*foundnonskip*/true, /*foundelse*/false); 1601 } else { 1602 // No, skip the contents of this block and return the first token after it. 1603 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1604 /*FoundElse*/false); 1605 } 1606} 1607 1608/// HandleEndifDirective - Implements the #endif directive. 1609/// 1610void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1611 ++NumEndif; 1612 1613 // Check that this is the whole directive. 1614 CheckEndOfDirective("endif"); 1615 1616 PPConditionalInfo CondInfo; 1617 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1618 // No conditionals on the stack: this is an #endif without an #if. 1619 Diag(EndifToken, diag::err_pp_endif_without_if); 1620 return; 1621 } 1622 1623 // If this the end of a top-level #endif, inform MIOpt. 1624 if (CurPPLexer->getConditionalStackDepth() == 0) 1625 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1626 1627 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1628 "This code should only be reachable in the non-skipping case!"); 1629} 1630 1631 1632void Preprocessor::HandleElseDirective(Token &Result) { 1633 ++NumElse; 1634 1635 // #else directive in a non-skipping conditional... start skipping. 1636 CheckEndOfDirective("else"); 1637 1638 PPConditionalInfo CI; 1639 if (CurPPLexer->popConditionalLevel(CI)) { 1640 Diag(Result, diag::pp_err_else_without_if); 1641 return; 1642 } 1643 1644 // If this is a top-level #else, inform the MIOpt. 1645 if (CurPPLexer->getConditionalStackDepth() == 0) 1646 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1647 1648 // If this is a #else with a #else before it, report the error. 1649 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1650 1651 // Finally, skip the rest of the contents of this block and return the first 1652 // token after it. 1653 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1654 /*FoundElse*/true); 1655} 1656 1657void Preprocessor::HandleElifDirective(Token &ElifToken) { 1658 ++NumElse; 1659 1660 // #elif directive in a non-skipping conditional... start skipping. 1661 // We don't care what the condition is, because we will always skip it (since 1662 // the block immediately before it was included). 1663 DiscardUntilEndOfDirective(); 1664 1665 PPConditionalInfo CI; 1666 if (CurPPLexer->popConditionalLevel(CI)) { 1667 Diag(ElifToken, diag::pp_err_elif_without_if); 1668 return; 1669 } 1670 1671 // If this is a top-level #elif, inform the MIOpt. 1672 if (CurPPLexer->getConditionalStackDepth() == 0) 1673 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1674 1675 // If this is a #elif with a #else before it, report the error. 1676 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1677 1678 // Finally, skip the rest of the contents of this block and return the first 1679 // token after it. 1680 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1681 /*FoundElse*/CI.FoundElse); 1682} 1683