PPDirectives.cpp revision a543016fe07030f695d6d56fd22c8c8da617e0d7
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 CurLexer->Lex(Tok); 164 165 // If this is the end of the buffer, we have an error. 166 if (Tok.is(tok::eof)) { 167 // Emit errors for each unterminated conditional on the stack, including 168 // the current one. 169 while (!CurPPLexer->ConditionalStack.empty()) { 170 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 171 diag::err_pp_unterminated_conditional); 172 CurPPLexer->ConditionalStack.pop_back(); 173 } 174 175 // Just return and let the caller lex after this #include. 176 break; 177 } 178 179 // If this token is not a preprocessor directive, just skip it. 180 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 181 continue; 182 183 // We just parsed a # character at the start of a line, so we're in 184 // directive mode. Tell the lexer this so any newlines we see will be 185 // converted into an EOM token (this terminates the macro). 186 CurPPLexer->ParsingPreprocessorDirective = true; 187 if (CurLexer) CurLexer->SetCommentRetentionState(false); 188 189 190 // Read the next token, the directive flavor. 191 LexUnexpandedToken(Tok); 192 193 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 194 // something bogus), skip it. 195 if (Tok.isNot(tok::identifier)) { 196 CurPPLexer->ParsingPreprocessorDirective = false; 197 // Restore comment saving mode. 198 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 199 continue; 200 } 201 202 // If the first letter isn't i or e, it isn't intesting to us. We know that 203 // this is safe in the face of spelling differences, because there is no way 204 // to spell an i/e in a strange way that is another letter. Skipping this 205 // allows us to avoid looking up the identifier info for #define/#undef and 206 // other common directives. 207 bool Invalid = false; 208 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation(), 209 &Invalid); 210 if (Invalid) 211 return; 212 213 char FirstChar = RawCharData[0]; 214 if (FirstChar >= 'a' && FirstChar <= 'z' && 215 FirstChar != 'i' && FirstChar != 'e') { 216 CurPPLexer->ParsingPreprocessorDirective = false; 217 // Restore comment saving mode. 218 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 219 continue; 220 } 221 222 // Get the identifier name without trigraphs or embedded newlines. Note 223 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 224 // when skipping. 225 char DirectiveBuf[20]; 226 llvm::StringRef Directive; 227 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 228 Directive = llvm::StringRef(RawCharData, Tok.getLength()); 229 } else { 230 std::string DirectiveStr = getSpelling(Tok); 231 unsigned IdLen = DirectiveStr.size(); 232 if (IdLen >= 20) { 233 CurPPLexer->ParsingPreprocessorDirective = false; 234 // Restore comment saving mode. 235 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 236 continue; 237 } 238 memcpy(DirectiveBuf, &DirectiveStr[0], IdLen); 239 Directive = llvm::StringRef(DirectiveBuf, IdLen); 240 } 241 242 if (Directive.startswith("if")) { 243 llvm::StringRef Sub = Directive.substr(2); 244 if (Sub.empty() || // "if" 245 Sub == "def" || // "ifdef" 246 Sub == "ndef") { // "ifndef" 247 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 248 // bother parsing the condition. 249 DiscardUntilEndOfDirective(); 250 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 251 /*foundnonskip*/false, 252 /*fnddelse*/false); 253 } 254 } else if (Directive[0] == 'e') { 255 llvm::StringRef Sub = Directive.substr(1); 256 if (Sub == "ndif") { // "endif" 257 CheckEndOfDirective("endif"); 258 PPConditionalInfo CondInfo; 259 CondInfo.WasSkipping = true; // Silence bogus warning. 260 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 261 InCond = InCond; // Silence warning in no-asserts mode. 262 assert(!InCond && "Can't be skipping if not in a conditional!"); 263 264 // If we popped the outermost skipping block, we're done skipping! 265 if (!CondInfo.WasSkipping) 266 break; 267 } else if (Sub == "lse") { // "else". 268 // #else directive in a skipping conditional. If not in some other 269 // skipping conditional, and if #else hasn't already been seen, enter it 270 // as a non-skipping conditional. 271 DiscardUntilEndOfDirective(); // C99 6.10p4. 272 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 273 274 // If this is a #else with a #else before it, report the error. 275 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 276 277 // Note that we've seen a #else in this conditional. 278 CondInfo.FoundElse = true; 279 280 // If the conditional is at the top level, and the #if block wasn't 281 // entered, enter the #else block now. 282 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 283 CondInfo.FoundNonSkip = true; 284 break; 285 } 286 } else if (Sub == "lif") { // "elif". 287 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 288 289 bool ShouldEnter; 290 // If this is in a skipping block or if we're already handled this #if 291 // block, don't bother parsing the condition. 292 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 293 DiscardUntilEndOfDirective(); 294 ShouldEnter = false; 295 } else { 296 // Restore the value of LexingRawMode so that identifiers are 297 // looked up, etc, inside the #elif expression. 298 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 299 CurPPLexer->LexingRawMode = false; 300 IdentifierInfo *IfNDefMacro = 0; 301 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 302 CurPPLexer->LexingRawMode = true; 303 } 304 305 // If this is a #elif with a #else before it, report the error. 306 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 307 308 // If this condition is true, enter it! 309 if (ShouldEnter) { 310 CondInfo.FoundNonSkip = true; 311 break; 312 } 313 } 314 } 315 316 CurPPLexer->ParsingPreprocessorDirective = false; 317 // Restore comment saving mode. 318 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 319 } 320 321 // Finally, if we are out of the conditional (saw an #endif or ran off the end 322 // of the file, just stop skipping and return to lexing whatever came after 323 // the #if block. 324 CurPPLexer->LexingRawMode = false; 325} 326 327void Preprocessor::PTHSkipExcludedConditionalBlock() { 328 329 while (1) { 330 assert(CurPTHLexer); 331 assert(CurPTHLexer->LexingRawMode == false); 332 333 // Skip to the next '#else', '#elif', or #endif. 334 if (CurPTHLexer->SkipBlock()) { 335 // We have reached an #endif. Both the '#' and 'endif' tokens 336 // have been consumed by the PTHLexer. Just pop off the condition level. 337 PPConditionalInfo CondInfo; 338 bool InCond = CurPTHLexer->popConditionalLevel(CondInfo); 339 InCond = InCond; // Silence warning in no-asserts mode. 340 assert(!InCond && "Can't be skipping if not in a conditional!"); 341 break; 342 } 343 344 // We have reached a '#else' or '#elif'. Lex the next token to get 345 // the directive flavor. 346 Token Tok; 347 LexUnexpandedToken(Tok); 348 349 // We can actually look up the IdentifierInfo here since we aren't in 350 // raw mode. 351 tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID(); 352 353 if (K == tok::pp_else) { 354 // #else: Enter the else condition. We aren't in a nested condition 355 // since we skip those. We're always in the one matching the last 356 // blocked we skipped. 357 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 358 // Note that we've seen a #else in this conditional. 359 CondInfo.FoundElse = true; 360 361 // If the #if block wasn't entered then enter the #else block now. 362 if (!CondInfo.FoundNonSkip) { 363 CondInfo.FoundNonSkip = true; 364 365 // Scan until the eom token. 366 CurPTHLexer->ParsingPreprocessorDirective = true; 367 DiscardUntilEndOfDirective(); 368 CurPTHLexer->ParsingPreprocessorDirective = false; 369 370 break; 371 } 372 373 // Otherwise skip this block. 374 continue; 375 } 376 377 assert(K == tok::pp_elif); 378 PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel(); 379 380 // If this is a #elif with a #else before it, report the error. 381 if (CondInfo.FoundElse) 382 Diag(Tok, diag::pp_err_elif_after_else); 383 384 // If this is in a skipping block or if we're already handled this #if 385 // block, don't bother parsing the condition. We just skip this block. 386 if (CondInfo.FoundNonSkip) 387 continue; 388 389 // Evaluate the condition of the #elif. 390 IdentifierInfo *IfNDefMacro = 0; 391 CurPTHLexer->ParsingPreprocessorDirective = true; 392 bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 393 CurPTHLexer->ParsingPreprocessorDirective = false; 394 395 // If this condition is true, enter it! 396 if (ShouldEnter) { 397 CondInfo.FoundNonSkip = true; 398 break; 399 } 400 401 // Otherwise, skip this block and go to the next one. 402 continue; 403 } 404} 405 406/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 407/// return null on failure. isAngled indicates whether the file reference is 408/// for system #include's or not (i.e. using <> instead of ""). 409const FileEntry *Preprocessor::LookupFile(llvm::StringRef Filename, 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(Filename, isAngled, FromDir, CurDir, CurFileEnt); 436 if (FE) return FE; 437 438 // Otherwise, see if this is a subframework header. If so, this is relative 439 // to one of the headers on the #include stack. Walk the list of the current 440 // headers on the #include stack and pass them to HeaderInfo. 441 if (IsFileLexer()) { 442 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 443 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 444 return FE; 445 } 446 447 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 448 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 449 if (IsFileLexer(ISEntry)) { 450 if ((CurFileEnt = 451 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 452 if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt))) 453 return FE; 454 } 455 } 456 457 // Otherwise, we really couldn't find the file. 458 return 0; 459} 460 461 462//===----------------------------------------------------------------------===// 463// Preprocessor Directive Handling. 464//===----------------------------------------------------------------------===// 465 466/// HandleDirective - This callback is invoked when the lexer sees a # token 467/// at the start of a line. This consumes the directive, modifies the 468/// lexer/preprocessor state, and advances the lexer(s) so that the next token 469/// read is the correct one. 470void Preprocessor::HandleDirective(Token &Result) { 471 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 472 473 // We just parsed a # character at the start of a line, so we're in directive 474 // mode. Tell the lexer this so any newlines we see will be converted into an 475 // EOM token (which terminates the directive). 476 CurPPLexer->ParsingPreprocessorDirective = true; 477 478 ++NumDirectives; 479 480 // We are about to read a token. For the multiple-include optimization FA to 481 // work, we have to remember if we had read any tokens *before* this 482 // pp-directive. 483 bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 484 485 // Save the '#' token in case we need to return it later. 486 Token SavedHash = Result; 487 488 // Read the next token, the directive flavor. This isn't expanded due to 489 // C99 6.10.3p8. 490 LexUnexpandedToken(Result); 491 492 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 493 // #define A(x) #x 494 // A(abc 495 // #warning blah 496 // def) 497 // If so, the user is relying on non-portable behavior, emit a diagnostic. 498 if (InMacroArgs) 499 Diag(Result, diag::ext_embedded_directive); 500 501TryAgain: 502 switch (Result.getKind()) { 503 case tok::eom: 504 return; // null directive. 505 case tok::comment: 506 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 507 LexUnexpandedToken(Result); 508 goto TryAgain; 509 510 case tok::numeric_constant: // # 7 GNU line marker directive. 511 if (getLangOptions().AsmPreprocessor) 512 break; // # 4 is not a preprocessor directive in .S files. 513 return HandleDigitDirective(Result); 514 default: 515 IdentifierInfo *II = Result.getIdentifierInfo(); 516 if (II == 0) break; // Not an identifier. 517 518 // Ask what the preprocessor keyword ID is. 519 switch (II->getPPKeywordID()) { 520 default: break; 521 // C99 6.10.1 - Conditional Inclusion. 522 case tok::pp_if: 523 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 524 case tok::pp_ifdef: 525 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 526 case tok::pp_ifndef: 527 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 528 case tok::pp_elif: 529 return HandleElifDirective(Result); 530 case tok::pp_else: 531 return HandleElseDirective(Result); 532 case tok::pp_endif: 533 return HandleEndifDirective(Result); 534 535 // C99 6.10.2 - Source File Inclusion. 536 case tok::pp_include: 537 return HandleIncludeDirective(Result); // Handle #include. 538 case tok::pp___include_macros: 539 return HandleIncludeMacrosDirective(Result); // Handle -imacros. 540 541 // C99 6.10.3 - Macro Replacement. 542 case tok::pp_define: 543 return HandleDefineDirective(Result); 544 case tok::pp_undef: 545 return HandleUndefDirective(Result); 546 547 // C99 6.10.4 - Line Control. 548 case tok::pp_line: 549 return HandleLineDirective(Result); 550 551 // C99 6.10.5 - Error Directive. 552 case tok::pp_error: 553 return HandleUserDiagnosticDirective(Result, false); 554 555 // C99 6.10.6 - Pragma Directive. 556 case tok::pp_pragma: 557 return HandlePragmaDirective(); 558 559 // GNU Extensions. 560 case tok::pp_import: 561 return HandleImportDirective(Result); 562 case tok::pp_include_next: 563 return HandleIncludeNextDirective(Result); 564 565 case tok::pp_warning: 566 Diag(Result, diag::ext_pp_warning_directive); 567 return HandleUserDiagnosticDirective(Result, true); 568 case tok::pp_ident: 569 return HandleIdentSCCSDirective(Result); 570 case tok::pp_sccs: 571 return HandleIdentSCCSDirective(Result); 572 case tok::pp_assert: 573 //isExtension = true; // FIXME: implement #assert 574 break; 575 case tok::pp_unassert: 576 //isExtension = true; // FIXME: implement #unassert 577 break; 578 } 579 break; 580 } 581 582 // If this is a .S file, treat unknown # directives as non-preprocessor 583 // directives. This is important because # may be a comment or introduce 584 // various pseudo-ops. Just return the # token and push back the following 585 // token to be lexed next time. 586 if (getLangOptions().AsmPreprocessor) { 587 Token *Toks = new Token[2]; 588 // Return the # and the token after it. 589 Toks[0] = SavedHash; 590 Toks[1] = Result; 591 // Enter this token stream so that we re-lex the tokens. Make sure to 592 // enable macro expansion, in case the token after the # is an identifier 593 // that is expanded. 594 EnterTokenStream(Toks, 2, false, true); 595 return; 596 } 597 598 // If we reached here, the preprocessing token is not valid! 599 Diag(Result, diag::err_pp_invalid_directive); 600 601 // Read the rest of the PP line. 602 DiscardUntilEndOfDirective(); 603 604 // Okay, we're done parsing the directive. 605} 606 607/// GetLineValue - Convert a numeric token into an unsigned value, emitting 608/// Diagnostic DiagID if it is invalid, and returning the value in Val. 609static bool GetLineValue(Token &DigitTok, unsigned &Val, 610 unsigned DiagID, Preprocessor &PP) { 611 if (DigitTok.isNot(tok::numeric_constant)) { 612 PP.Diag(DigitTok, DiagID); 613 614 if (DigitTok.isNot(tok::eom)) 615 PP.DiscardUntilEndOfDirective(); 616 return true; 617 } 618 619 llvm::SmallString<64> IntegerBuffer; 620 IntegerBuffer.resize(DigitTok.getLength()); 621 const char *DigitTokBegin = &IntegerBuffer[0]; 622 unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin); 623 624 // Verify that we have a simple digit-sequence, and compute the value. This 625 // is always a simple digit string computed in decimal, so we do this manually 626 // here. 627 Val = 0; 628 for (unsigned i = 0; i != ActualLength; ++i) { 629 if (!isdigit(DigitTokBegin[i])) { 630 PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i), 631 diag::err_pp_line_digit_sequence); 632 PP.DiscardUntilEndOfDirective(); 633 return true; 634 } 635 636 unsigned NextVal = Val*10+(DigitTokBegin[i]-'0'); 637 if (NextVal < Val) { // overflow. 638 PP.Diag(DigitTok, DiagID); 639 PP.DiscardUntilEndOfDirective(); 640 return true; 641 } 642 Val = NextVal; 643 } 644 645 // Reject 0, this is needed both by #line numbers and flags. 646 if (Val == 0) { 647 PP.Diag(DigitTok, DiagID); 648 PP.DiscardUntilEndOfDirective(); 649 return true; 650 } 651 652 if (DigitTokBegin[0] == '0') 653 PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal); 654 655 return false; 656} 657 658/// HandleLineDirective - Handle #line directive: C99 6.10.4. The two 659/// acceptable forms are: 660/// # line digit-sequence 661/// # line digit-sequence "s-char-sequence" 662void Preprocessor::HandleLineDirective(Token &Tok) { 663 // Read the line # and string argument. Per C99 6.10.4p5, these tokens are 664 // expanded. 665 Token DigitTok; 666 Lex(DigitTok); 667 668 // Validate the number and convert it to an unsigned. 669 unsigned LineNo; 670 if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this)) 671 return; 672 673 // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a 674 // number greater than 2147483647". C90 requires that the line # be <= 32767. 675 unsigned LineLimit = Features.C99 ? 2147483648U : 32768U; 676 if (LineNo >= LineLimit) 677 Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit; 678 679 int FilenameID = -1; 680 Token StrTok; 681 Lex(StrTok); 682 683 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 684 // string followed by eom. 685 if (StrTok.is(tok::eom)) 686 ; // ok 687 else if (StrTok.isNot(tok::string_literal)) { 688 Diag(StrTok, diag::err_pp_line_invalid_filename); 689 DiscardUntilEndOfDirective(); 690 return; 691 } else { 692 // Parse and validate the string, converting it into a unique ID. 693 StringLiteralParser Literal(&StrTok, 1, *this); 694 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 695 if (Literal.hadError) 696 return DiscardUntilEndOfDirective(); 697 if (Literal.Pascal) { 698 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 699 return DiscardUntilEndOfDirective(); 700 } 701 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 702 Literal.GetStringLength()); 703 704 // Verify that there is nothing after the string, other than EOM. Because 705 // of C99 6.10.4p5, macros that expand to empty tokens are ok. 706 CheckEndOfDirective("line", true); 707 } 708 709 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID); 710 711 if (Callbacks) 712 Callbacks->FileChanged(DigitTok.getLocation(), PPCallbacks::RenameFile, 713 SrcMgr::C_User); 714} 715 716/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line 717/// marker directive. 718static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit, 719 bool &IsSystemHeader, bool &IsExternCHeader, 720 Preprocessor &PP) { 721 unsigned FlagVal; 722 Token FlagTok; 723 PP.Lex(FlagTok); 724 if (FlagTok.is(tok::eom)) return false; 725 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 726 return true; 727 728 if (FlagVal == 1) { 729 IsFileEntry = true; 730 731 PP.Lex(FlagTok); 732 if (FlagTok.is(tok::eom)) return false; 733 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 734 return true; 735 } else if (FlagVal == 2) { 736 IsFileExit = true; 737 738 SourceManager &SM = PP.getSourceManager(); 739 // If we are leaving the current presumed file, check to make sure the 740 // presumed include stack isn't empty! 741 FileID CurFileID = 742 SM.getDecomposedInstantiationLoc(FlagTok.getLocation()).first; 743 PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation()); 744 745 // If there is no include loc (main file) or if the include loc is in a 746 // different physical file, then we aren't in a "1" line marker flag region. 747 SourceLocation IncLoc = PLoc.getIncludeLoc(); 748 if (IncLoc.isInvalid() || 749 SM.getDecomposedInstantiationLoc(IncLoc).first != CurFileID) { 750 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop); 751 PP.DiscardUntilEndOfDirective(); 752 return true; 753 } 754 755 PP.Lex(FlagTok); 756 if (FlagTok.is(tok::eom)) return false; 757 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP)) 758 return true; 759 } 760 761 // We must have 3 if there are still flags. 762 if (FlagVal != 3) { 763 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 764 PP.DiscardUntilEndOfDirective(); 765 return true; 766 } 767 768 IsSystemHeader = true; 769 770 PP.Lex(FlagTok); 771 if (FlagTok.is(tok::eom)) return false; 772 if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP)) 773 return true; 774 775 // We must have 4 if there is yet another flag. 776 if (FlagVal != 4) { 777 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 778 PP.DiscardUntilEndOfDirective(); 779 return true; 780 } 781 782 IsExternCHeader = true; 783 784 PP.Lex(FlagTok); 785 if (FlagTok.is(tok::eom)) return false; 786 787 // There are no more valid flags here. 788 PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag); 789 PP.DiscardUntilEndOfDirective(); 790 return true; 791} 792 793/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is 794/// one of the following forms: 795/// 796/// # 42 797/// # 42 "file" ('1' | '2')? 798/// # 42 "file" ('1' | '2')? '3' '4'? 799/// 800void Preprocessor::HandleDigitDirective(Token &DigitTok) { 801 // Validate the number and convert it to an unsigned. GNU does not have a 802 // line # limit other than it fit in 32-bits. 803 unsigned LineNo; 804 if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer, 805 *this)) 806 return; 807 808 Token StrTok; 809 Lex(StrTok); 810 811 bool IsFileEntry = false, IsFileExit = false; 812 bool IsSystemHeader = false, IsExternCHeader = false; 813 int FilenameID = -1; 814 815 // If the StrTok is "eom", then it wasn't present. Otherwise, it must be a 816 // string followed by eom. 817 if (StrTok.is(tok::eom)) 818 ; // ok 819 else if (StrTok.isNot(tok::string_literal)) { 820 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 821 return DiscardUntilEndOfDirective(); 822 } else { 823 // Parse and validate the string, converting it into a unique ID. 824 StringLiteralParser Literal(&StrTok, 1, *this); 825 assert(!Literal.AnyWide && "Didn't allow wide strings in"); 826 if (Literal.hadError) 827 return DiscardUntilEndOfDirective(); 828 if (Literal.Pascal) { 829 Diag(StrTok, diag::err_pp_linemarker_invalid_filename); 830 return DiscardUntilEndOfDirective(); 831 } 832 FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString(), 833 Literal.GetStringLength()); 834 835 // If a filename was present, read any flags that are present. 836 if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, 837 IsSystemHeader, IsExternCHeader, *this)) 838 return; 839 } 840 841 // Create a line note with this information. 842 SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, 843 IsFileEntry, IsFileExit, 844 IsSystemHeader, IsExternCHeader); 845 846 // If the preprocessor has callbacks installed, notify them of the #line 847 // change. This is used so that the line marker comes out in -E mode for 848 // example. 849 if (Callbacks) { 850 PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile; 851 if (IsFileEntry) 852 Reason = PPCallbacks::EnterFile; 853 else if (IsFileExit) 854 Reason = PPCallbacks::ExitFile; 855 SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User; 856 if (IsExternCHeader) 857 FileKind = SrcMgr::C_ExternCSystem; 858 else if (IsSystemHeader) 859 FileKind = SrcMgr::C_System; 860 861 Callbacks->FileChanged(DigitTok.getLocation(), Reason, FileKind); 862 } 863} 864 865 866/// HandleUserDiagnosticDirective - Handle a #warning or #error directive. 867/// 868void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 869 bool isWarning) { 870 // PTH doesn't emit #warning or #error directives. 871 if (CurPTHLexer) 872 return CurPTHLexer->DiscardToEndOfLine(); 873 874 // Read the rest of the line raw. We do this because we don't want macros 875 // to be expanded and we don't require that the tokens be valid preprocessing 876 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 877 // collapse multiple consequtive white space between tokens, but this isn't 878 // specified by the standard. 879 std::string Message = CurLexer->ReadToEndOfLine(); 880 if (isWarning) 881 Diag(Tok, diag::pp_hash_warning) << Message; 882 else 883 Diag(Tok, diag::err_pp_hash_error) << Message; 884} 885 886/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 887/// 888void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 889 // Yes, this directive is an extension. 890 Diag(Tok, diag::ext_pp_ident_directive); 891 892 // Read the string argument. 893 Token StrTok; 894 Lex(StrTok); 895 896 // If the token kind isn't a string, it's a malformed directive. 897 if (StrTok.isNot(tok::string_literal) && 898 StrTok.isNot(tok::wide_string_literal)) { 899 Diag(StrTok, diag::err_pp_malformed_ident); 900 if (StrTok.isNot(tok::eom)) 901 DiscardUntilEndOfDirective(); 902 return; 903 } 904 905 // Verify that there is nothing after the string, other than EOM. 906 CheckEndOfDirective("ident"); 907 908 if (Callbacks) 909 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok)); 910} 911 912//===----------------------------------------------------------------------===// 913// Preprocessor Include Directive Handling. 914//===----------------------------------------------------------------------===// 915 916/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 917/// checked and spelled filename, e.g. as an operand of #include. This returns 918/// true if the input filename was in <>'s or false if it were in ""'s. The 919/// caller is expected to provide a buffer that is large enough to hold the 920/// spelling of the filename, but is also expected to handle the case when 921/// this method decides to use a different buffer. 922bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 923 llvm::StringRef &Buffer) { 924 // Get the text form of the filename. 925 assert(!Buffer.empty() && "Can't have tokens with empty spellings!"); 926 927 // Make sure the filename is <x> or "x". 928 bool isAngled; 929 if (Buffer[0] == '<') { 930 if (Buffer.back() != '>') { 931 Diag(Loc, diag::err_pp_expects_filename); 932 Buffer = llvm::StringRef(); 933 return true; 934 } 935 isAngled = true; 936 } else if (Buffer[0] == '"') { 937 if (Buffer.back() != '"') { 938 Diag(Loc, diag::err_pp_expects_filename); 939 Buffer = llvm::StringRef(); 940 return true; 941 } 942 isAngled = false; 943 } else { 944 Diag(Loc, diag::err_pp_expects_filename); 945 Buffer = llvm::StringRef(); 946 return true; 947 } 948 949 // Diagnose #include "" as invalid. 950 if (Buffer.size() <= 2) { 951 Diag(Loc, diag::err_pp_empty_filename); 952 Buffer = llvm::StringRef(); 953 return true; 954 } 955 956 // Skip the brackets. 957 Buffer = Buffer.substr(1, Buffer.size()-2); 958 return isAngled; 959} 960 961/// ConcatenateIncludeName - Handle cases where the #include name is expanded 962/// from a macro as multiple tokens, which need to be glued together. This 963/// occurs for code like: 964/// #define FOO <a/b.h> 965/// #include FOO 966/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 967/// 968/// This code concatenates and consumes tokens up to the '>' token. It returns 969/// false if the > was found, otherwise it returns true if it finds and consumes 970/// the EOM marker. 971bool Preprocessor::ConcatenateIncludeName( 972 llvm::SmallString<128> &FilenameBuffer) { 973 Token CurTok; 974 975 Lex(CurTok); 976 while (CurTok.isNot(tok::eom)) { 977 // Append the spelling of this token to the buffer. If there was a space 978 // before it, add it now. 979 if (CurTok.hasLeadingSpace()) 980 FilenameBuffer.push_back(' '); 981 982 // Get the spelling of the token, directly into FilenameBuffer if possible. 983 unsigned PreAppendSize = FilenameBuffer.size(); 984 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 985 986 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 987 unsigned ActualLen = getSpelling(CurTok, BufPtr); 988 989 // If the token was spelled somewhere else, copy it into FilenameBuffer. 990 if (BufPtr != &FilenameBuffer[PreAppendSize]) 991 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 992 993 // Resize FilenameBuffer to the correct size. 994 if (CurTok.getLength() != ActualLen) 995 FilenameBuffer.resize(PreAppendSize+ActualLen); 996 997 // If we found the '>' marker, return success. 998 if (CurTok.is(tok::greater)) 999 return false; 1000 1001 Lex(CurTok); 1002 } 1003 1004 // If we hit the eom marker, emit an error and return true so that the caller 1005 // knows the EOM has been read. 1006 Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 1007 return true; 1008} 1009 1010/// HandleIncludeDirective - The "#include" tokens have just been read, read the 1011/// file to be included from the lexer, then include it! This is a common 1012/// routine with functionality shared between #include, #include_next and 1013/// #import. LookupFrom is set when this is a #include_next directive, it 1014/// specifies the file to start searching from. 1015void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 1016 const DirectoryLookup *LookupFrom, 1017 bool isImport) { 1018 1019 Token FilenameTok; 1020 CurPPLexer->LexIncludeFilename(FilenameTok); 1021 1022 // Reserve a buffer to get the spelling. 1023 llvm::SmallString<128> FilenameBuffer; 1024 llvm::StringRef Filename; 1025 1026 switch (FilenameTok.getKind()) { 1027 case tok::eom: 1028 // If the token kind is EOM, the error has already been diagnosed. 1029 return; 1030 1031 case tok::angle_string_literal: 1032 case tok::string_literal: 1033 Filename = getSpelling(FilenameTok, FilenameBuffer); 1034 break; 1035 1036 case tok::less: 1037 // This could be a <foo/bar.h> file coming from a macro expansion. In this 1038 // case, glue the tokens together into FilenameBuffer and interpret those. 1039 FilenameBuffer.push_back('<'); 1040 if (ConcatenateIncludeName(FilenameBuffer)) 1041 return; // Found <eom> but no ">"? Diagnostic already emitted. 1042 Filename = FilenameBuffer.str(); 1043 break; 1044 default: 1045 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 1046 DiscardUntilEndOfDirective(); 1047 return; 1048 } 1049 1050 bool isAngled = 1051 GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename); 1052 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 1053 // error. 1054 if (Filename.empty()) { 1055 DiscardUntilEndOfDirective(); 1056 return; 1057 } 1058 1059 // Verify that there is nothing after the filename, other than EOM. Note that 1060 // we allow macros that expand to nothing after the filename, because this 1061 // falls into the category of "#include pp-tokens new-line" specified in 1062 // C99 6.10.2p4. 1063 CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true); 1064 1065 // Check that we don't have infinite #include recursion. 1066 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 1067 Diag(FilenameTok, diag::err_pp_include_too_deep); 1068 return; 1069 } 1070 1071 // Search include directories. 1072 const DirectoryLookup *CurDir; 1073 const FileEntry *File = LookupFile(Filename, isAngled, LookupFrom, CurDir); 1074 if (File == 0) { 1075 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1076 return; 1077 } 1078 1079 // Ask HeaderInfo if we should enter this #include file. If not, #including 1080 // this file will have no effect. 1081 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) 1082 return; 1083 1084 // The #included file will be considered to be a system header if either it is 1085 // in a system include directory, or if the #includer is a system include 1086 // header. 1087 SrcMgr::CharacteristicKind FileCharacter = 1088 std::max(HeaderInfo.getFileDirFlavor(File), 1089 SourceMgr.getFileCharacteristic(FilenameTok.getLocation())); 1090 1091 // Look up the file, create a File ID for it. 1092 FileID FID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 1093 FileCharacter); 1094 if (FID.isInvalid()) { 1095 Diag(FilenameTok, diag::err_pp_file_not_found) << Filename; 1096 return; 1097 } 1098 1099 // Finally, if all is good, enter the new file! 1100 std::string ErrorStr; 1101 if (EnterSourceFile(FID, CurDir, ErrorStr)) 1102 Diag(FilenameTok, diag::err_pp_error_opening_file) 1103 << std::string(SourceMgr.getFileEntryForID(FID)->getName()) << ErrorStr; 1104} 1105 1106/// HandleIncludeNextDirective - Implements #include_next. 1107/// 1108void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 1109 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 1110 1111 // #include_next is like #include, except that we start searching after 1112 // the current found directory. If we can't do this, issue a 1113 // diagnostic. 1114 const DirectoryLookup *Lookup = CurDirLookup; 1115 if (isInPrimaryFile()) { 1116 Lookup = 0; 1117 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 1118 } else if (Lookup == 0) { 1119 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 1120 } else { 1121 // Start looking up in the next directory. 1122 ++Lookup; 1123 } 1124 1125 return HandleIncludeDirective(IncludeNextTok, Lookup); 1126} 1127 1128/// HandleImportDirective - Implements #import. 1129/// 1130void Preprocessor::HandleImportDirective(Token &ImportTok) { 1131 if (!Features.ObjC1) // #import is standard for ObjC. 1132 Diag(ImportTok, diag::ext_pp_import_directive); 1133 1134 return HandleIncludeDirective(ImportTok, 0, true); 1135} 1136 1137/// HandleIncludeMacrosDirective - The -imacros command line option turns into a 1138/// pseudo directive in the predefines buffer. This handles it by sucking all 1139/// tokens through the preprocessor and discarding them (only keeping the side 1140/// effects on the preprocessor). 1141void Preprocessor::HandleIncludeMacrosDirective(Token &IncludeMacrosTok) { 1142 // This directive should only occur in the predefines buffer. If not, emit an 1143 // error and reject it. 1144 SourceLocation Loc = IncludeMacrosTok.getLocation(); 1145 if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) { 1146 Diag(IncludeMacrosTok.getLocation(), 1147 diag::pp_include_macros_out_of_predefines); 1148 DiscardUntilEndOfDirective(); 1149 return; 1150 } 1151 1152 // Treat this as a normal #include for checking purposes. If this is 1153 // successful, it will push a new lexer onto the include stack. 1154 HandleIncludeDirective(IncludeMacrosTok, 0, false); 1155 1156 Token TmpTok; 1157 do { 1158 Lex(TmpTok); 1159 assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!"); 1160 } while (TmpTok.isNot(tok::hashhash)); 1161} 1162 1163//===----------------------------------------------------------------------===// 1164// Preprocessor Macro Directive Handling. 1165//===----------------------------------------------------------------------===// 1166 1167/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1168/// definition has just been read. Lex the rest of the arguments and the 1169/// closing ), updating MI with what we learn. Return true if an error occurs 1170/// parsing the arg list. 1171bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 1172 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 1173 1174 Token Tok; 1175 while (1) { 1176 LexUnexpandedToken(Tok); 1177 switch (Tok.getKind()) { 1178 case tok::r_paren: 1179 // Found the end of the argument list. 1180 if (Arguments.empty()) // #define FOO() 1181 return false; 1182 // Otherwise we have #define FOO(A,) 1183 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 1184 return true; 1185 case tok::ellipsis: // #define X(... -> C99 varargs 1186 // Warn if use of C99 feature in non-C99 mode. 1187 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 1188 1189 // Lex the token after the identifier. 1190 LexUnexpandedToken(Tok); 1191 if (Tok.isNot(tok::r_paren)) { 1192 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1193 return true; 1194 } 1195 // Add the __VA_ARGS__ identifier as an argument. 1196 Arguments.push_back(Ident__VA_ARGS__); 1197 MI->setIsC99Varargs(); 1198 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1199 return false; 1200 case tok::eom: // #define X( 1201 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1202 return true; 1203 default: 1204 // Handle keywords and identifiers here to accept things like 1205 // #define Foo(for) for. 1206 IdentifierInfo *II = Tok.getIdentifierInfo(); 1207 if (II == 0) { 1208 // #define X(1 1209 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 1210 return true; 1211 } 1212 1213 // If this is already used as an argument, it is used multiple times (e.g. 1214 // #define X(A,A. 1215 if (std::find(Arguments.begin(), Arguments.end(), II) != 1216 Arguments.end()) { // C99 6.10.3p6 1217 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 1218 return true; 1219 } 1220 1221 // Add the argument to the macro info. 1222 Arguments.push_back(II); 1223 1224 // Lex the token after the identifier. 1225 LexUnexpandedToken(Tok); 1226 1227 switch (Tok.getKind()) { 1228 default: // #define X(A B 1229 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 1230 return true; 1231 case tok::r_paren: // #define X(A) 1232 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1233 return false; 1234 case tok::comma: // #define X(A, 1235 break; 1236 case tok::ellipsis: // #define X(A... -> GCC extension 1237 // Diagnose extension. 1238 Diag(Tok, diag::ext_named_variadic_macro); 1239 1240 // Lex the token after the identifier. 1241 LexUnexpandedToken(Tok); 1242 if (Tok.isNot(tok::r_paren)) { 1243 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 1244 return true; 1245 } 1246 1247 MI->setIsGNUVarargs(); 1248 MI->setArgumentList(&Arguments[0], Arguments.size(), BP); 1249 return false; 1250 } 1251 } 1252 } 1253} 1254 1255/// HandleDefineDirective - Implements #define. This consumes the entire macro 1256/// line then lets the caller lex the next real token. 1257void Preprocessor::HandleDefineDirective(Token &DefineTok) { 1258 ++NumDefined; 1259 1260 Token MacroNameTok; 1261 ReadMacroName(MacroNameTok, 1); 1262 1263 // Error reading macro name? If so, diagnostic already issued. 1264 if (MacroNameTok.is(tok::eom)) 1265 return; 1266 1267 Token LastTok = MacroNameTok; 1268 1269 // If we are supposed to keep comments in #defines, reenable comment saving 1270 // mode. 1271 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 1272 1273 // Create the new macro. 1274 MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation()); 1275 1276 Token Tok; 1277 LexUnexpandedToken(Tok); 1278 1279 // If this is a function-like macro definition, parse the argument list, 1280 // marking each of the identifiers as being used as macro arguments. Also, 1281 // check other constraints on the first token of the macro body. 1282 if (Tok.is(tok::eom)) { 1283 // If there is no body to this macro, we have no special handling here. 1284 } else if (Tok.hasLeadingSpace()) { 1285 // This is a normal token with leading space. Clear the leading space 1286 // marker on the first token to get proper expansion. 1287 Tok.clearFlag(Token::LeadingSpace); 1288 } else if (Tok.is(tok::l_paren)) { 1289 // This is a function-like macro definition. Read the argument list. 1290 MI->setIsFunctionLike(); 1291 if (ReadMacroDefinitionArgList(MI)) { 1292 // Forget about MI. 1293 ReleaseMacroInfo(MI); 1294 // Throw away the rest of the line. 1295 if (CurPPLexer->ParsingPreprocessorDirective) 1296 DiscardUntilEndOfDirective(); 1297 return; 1298 } 1299 1300 // If this is a definition of a variadic C99 function-like macro, not using 1301 // the GNU named varargs extension, enabled __VA_ARGS__. 1302 1303 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 1304 // This gets unpoisoned where it is allowed. 1305 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 1306 if (MI->isC99Varargs()) 1307 Ident__VA_ARGS__->setIsPoisoned(false); 1308 1309 // Read the first token after the arg list for down below. 1310 LexUnexpandedToken(Tok); 1311 } else if (Features.C99) { 1312 // C99 requires whitespace between the macro definition and the body. Emit 1313 // a diagnostic for something like "#define X+". 1314 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 1315 } else { 1316 // C90 6.8 TC1 says: "In the definition of an object-like macro, if the 1317 // first character of a replacement list is not a character required by 1318 // subclause 5.2.1, then there shall be white-space separation between the 1319 // identifier and the replacement list.". 5.2.1 lists this set: 1320 // "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which 1321 // is irrelevant here. 1322 bool isInvalid = false; 1323 if (Tok.is(tok::at)) // @ is not in the list above. 1324 isInvalid = true; 1325 else if (Tok.is(tok::unknown)) { 1326 // If we have an unknown token, it is something strange like "`". Since 1327 // all of valid characters would have lexed into a single character 1328 // token of some sort, we know this is not a valid case. 1329 isInvalid = true; 1330 } 1331 if (isInvalid) 1332 Diag(Tok, diag::ext_missing_whitespace_after_macro_name); 1333 else 1334 Diag(Tok, diag::warn_missing_whitespace_after_macro_name); 1335 } 1336 1337 if (!Tok.is(tok::eom)) 1338 LastTok = Tok; 1339 1340 // Read the rest of the macro body. 1341 if (MI->isObjectLike()) { 1342 // Object-like macros are very simple, just read their body. 1343 while (Tok.isNot(tok::eom)) { 1344 LastTok = Tok; 1345 MI->AddTokenToBody(Tok); 1346 // Get the next token of the macro. 1347 LexUnexpandedToken(Tok); 1348 } 1349 1350 } else { 1351 // Otherwise, read the body of a function-like macro. While we are at it, 1352 // check C99 6.10.3.2p1: ensure that # operators are followed by macro 1353 // parameters in function-like macro expansions. 1354 while (Tok.isNot(tok::eom)) { 1355 LastTok = Tok; 1356 1357 if (Tok.isNot(tok::hash)) { 1358 MI->AddTokenToBody(Tok); 1359 1360 // Get the next token of the macro. 1361 LexUnexpandedToken(Tok); 1362 continue; 1363 } 1364 1365 // Get the next token of the macro. 1366 LexUnexpandedToken(Tok); 1367 1368 // Check for a valid macro arg identifier. 1369 if (Tok.getIdentifierInfo() == 0 || 1370 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 1371 1372 // If this is assembler-with-cpp mode, we accept random gibberish after 1373 // the '#' because '#' is often a comment character. However, change 1374 // the kind of the token to tok::unknown so that the preprocessor isn't 1375 // confused. 1376 if (getLangOptions().AsmPreprocessor && Tok.isNot(tok::eom)) { 1377 LastTok.setKind(tok::unknown); 1378 } else { 1379 Diag(Tok, diag::err_pp_stringize_not_parameter); 1380 ReleaseMacroInfo(MI); 1381 1382 // Disable __VA_ARGS__ again. 1383 Ident__VA_ARGS__->setIsPoisoned(true); 1384 return; 1385 } 1386 } 1387 1388 // Things look ok, add the '#' and param name tokens to the macro. 1389 MI->AddTokenToBody(LastTok); 1390 MI->AddTokenToBody(Tok); 1391 LastTok = Tok; 1392 1393 // Get the next token of the macro. 1394 LexUnexpandedToken(Tok); 1395 } 1396 } 1397 1398 1399 // Disable __VA_ARGS__ again. 1400 Ident__VA_ARGS__->setIsPoisoned(true); 1401 1402 // Check that there is no paste (##) operator at the begining or end of the 1403 // replacement list. 1404 unsigned NumTokens = MI->getNumTokens(); 1405 if (NumTokens != 0) { 1406 if (MI->getReplacementToken(0).is(tok::hashhash)) { 1407 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 1408 ReleaseMacroInfo(MI); 1409 return; 1410 } 1411 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 1412 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 1413 ReleaseMacroInfo(MI); 1414 return; 1415 } 1416 } 1417 1418 // If this is the primary source file, remember that this macro hasn't been 1419 // used yet. 1420 if (isInPrimaryFile()) 1421 MI->setIsUsed(false); 1422 1423 MI->setDefinitionEndLoc(LastTok.getLocation()); 1424 1425 // Finally, if this identifier already had a macro defined for it, verify that 1426 // the macro bodies are identical and free the old definition. 1427 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 1428 // It is very common for system headers to have tons of macro redefinitions 1429 // and for warnings to be disabled in system headers. If this is the case, 1430 // then don't bother calling MacroInfo::isIdenticalTo. 1431 if (!getDiagnostics().getSuppressSystemWarnings() || 1432 !SourceMgr.isInSystemHeader(DefineTok.getLocation())) { 1433 if (!OtherMI->isUsed()) 1434 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 1435 1436 // Macros must be identical. This means all tokes and whitespace 1437 // separation must be the same. C99 6.10.3.2. 1438 if (!MI->isIdenticalTo(*OtherMI, *this)) { 1439 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 1440 << MacroNameTok.getIdentifierInfo(); 1441 Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition); 1442 } 1443 } 1444 1445 ReleaseMacroInfo(OtherMI); 1446 } 1447 1448 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 1449 1450 // If the callbacks want to know, tell them about the macro definition. 1451 if (Callbacks) 1452 Callbacks->MacroDefined(MacroNameTok.getIdentifierInfo(), MI); 1453} 1454 1455/// HandleUndefDirective - Implements #undef. 1456/// 1457void Preprocessor::HandleUndefDirective(Token &UndefTok) { 1458 ++NumUndefined; 1459 1460 Token MacroNameTok; 1461 ReadMacroName(MacroNameTok, 2); 1462 1463 // Error reading macro name? If so, diagnostic already issued. 1464 if (MacroNameTok.is(tok::eom)) 1465 return; 1466 1467 // Check to see if this is the last token on the #undef line. 1468 CheckEndOfDirective("undef"); 1469 1470 // Okay, we finally have a valid identifier to undef. 1471 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1472 1473 // If the macro is not defined, this is a noop undef, just return. 1474 if (MI == 0) return; 1475 1476 if (!MI->isUsed()) 1477 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1478 1479 // If the callbacks want to know, tell them about the macro #undef. 1480 if (Callbacks) 1481 Callbacks->MacroUndefined(MacroNameTok.getIdentifierInfo(), MI); 1482 1483 // Free macro definition. 1484 ReleaseMacroInfo(MI); 1485 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1486} 1487 1488 1489//===----------------------------------------------------------------------===// 1490// Preprocessor Conditional Directive Handling. 1491//===----------------------------------------------------------------------===// 1492 1493/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1494/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1495/// if any tokens have been returned or pp-directives activated before this 1496/// #ifndef has been lexed. 1497/// 1498void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1499 bool ReadAnyTokensBeforeDirective) { 1500 ++NumIf; 1501 Token DirectiveTok = Result; 1502 1503 Token MacroNameTok; 1504 ReadMacroName(MacroNameTok); 1505 1506 // Error reading macro name? If so, diagnostic already issued. 1507 if (MacroNameTok.is(tok::eom)) { 1508 // Skip code until we get to #endif. This helps with recovery by not 1509 // emitting an error when the #endif is reached. 1510 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1511 /*Foundnonskip*/false, /*FoundElse*/false); 1512 return; 1513 } 1514 1515 // Check to see if this is the last token on the #if[n]def line. 1516 CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef"); 1517 1518 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1519 MacroInfo *MI = getMacroInfo(MII); 1520 1521 if (CurPPLexer->getConditionalStackDepth() == 0) { 1522 // If the start of a top-level #ifdef and if the macro is not defined, 1523 // inform MIOpt that this might be the start of a proper include guard. 1524 // Otherwise it is some other form of unknown conditional which we can't 1525 // handle. 1526 if (!ReadAnyTokensBeforeDirective && MI == 0) { 1527 assert(isIfndef && "#ifdef shouldn't reach here"); 1528 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII); 1529 } else 1530 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1531 } 1532 1533 // If there is a macro, process it. 1534 if (MI) // Mark it used. 1535 MI->setIsUsed(true); 1536 1537 // Should we include the stuff contained by this directive? 1538 if (!MI == isIfndef) { 1539 // Yes, remember that we are inside a conditional, then lex the next token. 1540 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), 1541 /*wasskip*/false, /*foundnonskip*/true, 1542 /*foundelse*/false); 1543 } else { 1544 // No, skip the contents of this block and return the first token after it. 1545 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1546 /*Foundnonskip*/false, 1547 /*FoundElse*/false); 1548 } 1549} 1550 1551/// HandleIfDirective - Implements the #if directive. 1552/// 1553void Preprocessor::HandleIfDirective(Token &IfToken, 1554 bool ReadAnyTokensBeforeDirective) { 1555 ++NumIf; 1556 1557 // Parse and evaluation the conditional expression. 1558 IdentifierInfo *IfNDefMacro = 0; 1559 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1560 1561 1562 // If this condition is equivalent to #ifndef X, and if this is the first 1563 // directive seen, handle it for the multiple-include optimization. 1564 if (CurPPLexer->getConditionalStackDepth() == 0) { 1565 if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue) 1566 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1567 else 1568 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1569 } 1570 1571 // Should we include the stuff contained by this directive? 1572 if (ConditionalTrue) { 1573 // Yes, remember that we are inside a conditional, then lex the next token. 1574 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1575 /*foundnonskip*/true, /*foundelse*/false); 1576 } else { 1577 // No, skip the contents of this block and return the first token after it. 1578 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1579 /*FoundElse*/false); 1580 } 1581} 1582 1583/// HandleEndifDirective - Implements the #endif directive. 1584/// 1585void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1586 ++NumEndif; 1587 1588 // Check that this is the whole directive. 1589 CheckEndOfDirective("endif"); 1590 1591 PPConditionalInfo CondInfo; 1592 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1593 // No conditionals on the stack: this is an #endif without an #if. 1594 Diag(EndifToken, diag::err_pp_endif_without_if); 1595 return; 1596 } 1597 1598 // If this the end of a top-level #endif, inform MIOpt. 1599 if (CurPPLexer->getConditionalStackDepth() == 0) 1600 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1601 1602 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1603 "This code should only be reachable in the non-skipping case!"); 1604} 1605 1606 1607void Preprocessor::HandleElseDirective(Token &Result) { 1608 ++NumElse; 1609 1610 // #else directive in a non-skipping conditional... start skipping. 1611 CheckEndOfDirective("else"); 1612 1613 PPConditionalInfo CI; 1614 if (CurPPLexer->popConditionalLevel(CI)) { 1615 Diag(Result, diag::pp_err_else_without_if); 1616 return; 1617 } 1618 1619 // If this is a top-level #else, inform the MIOpt. 1620 if (CurPPLexer->getConditionalStackDepth() == 0) 1621 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1622 1623 // If this is a #else with a #else before it, report the error. 1624 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1625 1626 // Finally, skip the rest of the contents of this block and return the first 1627 // token after it. 1628 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1629 /*FoundElse*/true); 1630} 1631 1632void Preprocessor::HandleElifDirective(Token &ElifToken) { 1633 ++NumElse; 1634 1635 // #elif directive in a non-skipping conditional... start skipping. 1636 // We don't care what the condition is, because we will always skip it (since 1637 // the block immediately before it was included). 1638 DiscardUntilEndOfDirective(); 1639 1640 PPConditionalInfo CI; 1641 if (CurPPLexer->popConditionalLevel(CI)) { 1642 Diag(ElifToken, diag::pp_err_elif_without_if); 1643 return; 1644 } 1645 1646 // If this is a top-level #elif, inform the MIOpt. 1647 if (CurPPLexer->getConditionalStackDepth() == 0) 1648 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1649 1650 // If this is a #elif with a #else before it, report the error. 1651 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1652 1653 // Finally, skip the rest of the contents of this block and return the first 1654 // token after it. 1655 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1656 /*FoundElse*/CI.FoundElse); 1657} 1658 1659