PPDirectives.cpp revision 81d24e135c60dcbe421864e6a4334ba99b8fe7ec
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/HeaderSearch.h" 16#include "clang/Lex/MacroInfo.h" 17#include "clang/Basic/Diagnostic.h" 18#include "clang/Basic/SourceManager.h" 19using namespace clang; 20 21//===----------------------------------------------------------------------===// 22// Utility Methods for Preprocessor Directive Handling. 23//===----------------------------------------------------------------------===// 24 25/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 26/// current line until the tok::eom token is found. 27void Preprocessor::DiscardUntilEndOfDirective() { 28 Token Tmp; 29 do { 30 LexUnexpandedToken(Tmp); 31 } while (Tmp.isNot(tok::eom)); 32} 33 34/// isCXXNamedOperator - Returns "true" if the token is a named operator in C++. 35static bool isCXXNamedOperator(const std::string &Spelling) { 36 return Spelling == "and" || Spelling == "bitand" || Spelling == "bitor" || 37 Spelling == "compl" || Spelling == "not" || Spelling == "not_eq" || 38 Spelling == "or" || Spelling == "xor"; 39} 40 41/// ReadMacroName - Lex and validate a macro name, which occurs after a 42/// #define or #undef. This sets the token kind to eom and discards the rest 43/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 44/// this is due to a a #define, 2 if #undef directive, 0 if it is something 45/// else (e.g. #ifdef). 46void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 47 // Read the token, don't allow macro expansion on it. 48 LexUnexpandedToken(MacroNameTok); 49 50 // Missing macro name? 51 if (MacroNameTok.is(tok::eom)) { 52 Diag(MacroNameTok, diag::err_pp_missing_macro_name); 53 return; 54 } 55 56 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 57 if (II == 0) { 58 std::string Spelling = getSpelling(MacroNameTok); 59 if (isCXXNamedOperator(Spelling)) 60 // C++ 2.5p2: Alternative tokens behave the same as its primary token 61 // except for their spellings. 62 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling; 63 else 64 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 65 // Fall through on error. 66 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 67 // Error if defining "defined": C99 6.10.8.4. 68 Diag(MacroNameTok, diag::err_defined_macro_name); 69 } else if (isDefineUndef && II->hasMacroDefinition() && 70 getMacroInfo(II)->isBuiltinMacro()) { 71 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 72 if (isDefineUndef == 1) 73 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 74 else 75 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 76 } else { 77 // Okay, we got a good identifier node. Return it. 78 return; 79 } 80 81 // Invalid macro name, read and discard the rest of the line. Then set the 82 // token kind to tok::eom. 83 MacroNameTok.setKind(tok::eom); 84 return DiscardUntilEndOfDirective(); 85} 86 87/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 88/// not, emit a diagnostic and consume up until the eom. 89void Preprocessor::CheckEndOfDirective(const char *DirType) { 90 Token Tmp; 91 // Lex unexpanded tokens: macros might expand to zero tokens, causing us to 92 // miss diagnosing invalid lines. 93 LexUnexpandedToken(Tmp); 94 95 // There should be no tokens after the directive, but we allow them as an 96 // extension. 97 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 98 LexUnexpandedToken(Tmp); 99 100 if (Tmp.isNot(tok::eom)) { 101 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType; 102 DiscardUntilEndOfDirective(); 103 } 104} 105 106 107 108/// SkipExcludedConditionalBlock - We just read a #if or related directive and 109/// decided that the subsequent tokens are in the #if'd out portion of the 110/// file. Lex the rest of the file, until we see an #endif. If 111/// FoundNonSkipPortion is true, then we have already emitted code for part of 112/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 113/// is true, then #else directives are ok, if not, then we have already seen one 114/// so a #else directive is a duplicate. When this returns, the caller can lex 115/// the first valid token. 116void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 117 bool FoundNonSkipPortion, 118 bool FoundElse) { 119 ++NumSkipped; 120 assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?"); 121 122 CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false, 123 FoundNonSkipPortion, FoundElse); 124 125 // Enter raw mode to disable identifier lookup (and thus macro expansion), 126 // disabling warnings, etc. 127 CurPPLexer->LexingRawMode = true; 128 Token Tok; 129 while (1) { 130 if (CurLexer) 131 CurLexer->Lex(Tok); 132 else 133 CurPTHLexer->Lex(Tok); 134 135 // If this is the end of the buffer, we have an error. 136 if (Tok.is(tok::eof)) { 137 // Emit errors for each unterminated conditional on the stack, including 138 // the current one. 139 while (!CurPPLexer->ConditionalStack.empty()) { 140 Diag(CurPPLexer->ConditionalStack.back().IfLoc, 141 diag::err_pp_unterminated_conditional); 142 CurPPLexer->ConditionalStack.pop_back(); 143 } 144 145 // Just return and let the caller lex after this #include. 146 break; 147 } 148 149 // If this token is not a preprocessor directive, just skip it. 150 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 151 continue; 152 153 // We just parsed a # character at the start of a line, so we're in 154 // directive mode. Tell the lexer this so any newlines we see will be 155 // converted into an EOM token (this terminates the macro). 156 CurPPLexer->ParsingPreprocessorDirective = true; 157 if (CurLexer) CurLexer->SetCommentRetentionState(false); 158 159 160 // Read the next token, the directive flavor. 161 LexUnexpandedToken(Tok); 162 163 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 164 // something bogus), skip it. 165 if (Tok.isNot(tok::identifier)) { 166 CurPPLexer->ParsingPreprocessorDirective = false; 167 // Restore comment saving mode. 168 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 169 continue; 170 } 171 172 // If the first letter isn't i or e, it isn't intesting to us. We know that 173 // this is safe in the face of spelling differences, because there is no way 174 // to spell an i/e in a strange way that is another letter. Skipping this 175 // allows us to avoid looking up the identifier info for #define/#undef and 176 // other common directives. 177 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation()); 178 char FirstChar = RawCharData[0]; 179 if (FirstChar >= 'a' && FirstChar <= 'z' && 180 FirstChar != 'i' && FirstChar != 'e') { 181 CurPPLexer->ParsingPreprocessorDirective = false; 182 // Restore comment saving mode. 183 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 184 continue; 185 } 186 187 // Get the identifier name without trigraphs or embedded newlines. Note 188 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 189 // when skipping. 190 // TODO: could do this with zero copies in the no-clean case by using 191 // strncmp below. 192 char Directive[20]; 193 unsigned IdLen; 194 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 195 IdLen = Tok.getLength(); 196 memcpy(Directive, RawCharData, IdLen); 197 Directive[IdLen] = 0; 198 } else { 199 std::string DirectiveStr = getSpelling(Tok); 200 IdLen = DirectiveStr.size(); 201 if (IdLen >= 20) { 202 CurPPLexer->ParsingPreprocessorDirective = false; 203 // Restore comment saving mode. 204 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 205 continue; 206 } 207 memcpy(Directive, &DirectiveStr[0], IdLen); 208 Directive[IdLen] = 0; 209 } 210 211 if (FirstChar == 'i' && Directive[1] == 'f') { 212 if ((IdLen == 2) || // "if" 213 (IdLen == 5 && !strcmp(Directive+2, "def")) || // "ifdef" 214 (IdLen == 6 && !strcmp(Directive+2, "ndef"))) { // "ifndef" 215 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 216 // bother parsing the condition. 217 DiscardUntilEndOfDirective(); 218 CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 219 /*foundnonskip*/false, 220 /*fnddelse*/false); 221 } 222 } else if (FirstChar == 'e') { 223 if (IdLen == 5 && !strcmp(Directive+1, "ndif")) { // "endif" 224 CheckEndOfDirective("#endif"); 225 PPConditionalInfo CondInfo; 226 CondInfo.WasSkipping = true; // Silence bogus warning. 227 bool InCond = CurPPLexer->popConditionalLevel(CondInfo); 228 InCond = InCond; // Silence warning in no-asserts mode. 229 assert(!InCond && "Can't be skipping if not in a conditional!"); 230 231 // If we popped the outermost skipping block, we're done skipping! 232 if (!CondInfo.WasSkipping) 233 break; 234 } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else". 235 // #else directive in a skipping conditional. If not in some other 236 // skipping conditional, and if #else hasn't already been seen, enter it 237 // as a non-skipping conditional. 238 CheckEndOfDirective("#else"); 239 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 240 241 // If this is a #else with a #else before it, report the error. 242 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 243 244 // Note that we've seen a #else in this conditional. 245 CondInfo.FoundElse = true; 246 247 // If the conditional is at the top level, and the #if block wasn't 248 // entered, enter the #else block now. 249 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 250 CondInfo.FoundNonSkip = true; 251 break; 252 } 253 } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) { // "elif". 254 PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel(); 255 256 bool ShouldEnter; 257 // If this is in a skipping block or if we're already handled this #if 258 // block, don't bother parsing the condition. 259 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 260 DiscardUntilEndOfDirective(); 261 ShouldEnter = false; 262 } else { 263 // Restore the value of LexingRawMode so that identifiers are 264 // looked up, etc, inside the #elif expression. 265 assert(CurPPLexer->LexingRawMode && "We have to be skipping here!"); 266 CurPPLexer->LexingRawMode = false; 267 IdentifierInfo *IfNDefMacro = 0; 268 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 269 CurPPLexer->LexingRawMode = true; 270 } 271 272 // If this is a #elif with a #else before it, report the error. 273 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 274 275 // If this condition is true, enter it! 276 if (ShouldEnter) { 277 CondInfo.FoundNonSkip = true; 278 break; 279 } 280 } 281 } 282 283 CurPPLexer->ParsingPreprocessorDirective = false; 284 // Restore comment saving mode. 285 if (CurLexer) CurLexer->SetCommentRetentionState(KeepComments); 286 } 287 288 // Finally, if we are out of the conditional (saw an #endif or ran off the end 289 // of the file, just stop skipping and return to lexing whatever came after 290 // the #if block. 291 CurPPLexer->LexingRawMode = false; 292} 293 294/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 295/// return null on failure. isAngled indicates whether the file reference is 296/// for system #include's or not (i.e. using <> instead of ""). 297const FileEntry *Preprocessor::LookupFile(const char *FilenameStart, 298 const char *FilenameEnd, 299 bool isAngled, 300 const DirectoryLookup *FromDir, 301 const DirectoryLookup *&CurDir) { 302 // If the header lookup mechanism may be relative to the current file, pass in 303 // info about where the current file is. 304 const FileEntry *CurFileEnt = 0; 305 if (!FromDir) { 306 unsigned FileID = getCurrentFileLexer()->getFileID(); 307 CurFileEnt = SourceMgr.getFileEntryForID(FileID); 308 } 309 310 // Do a standard file entry lookup. 311 CurDir = CurDirLookup; 312 const FileEntry *FE = 313 HeaderInfo.LookupFile(FilenameStart, FilenameEnd, 314 isAngled, FromDir, CurDir, CurFileEnt); 315 if (FE) return FE; 316 317 // Otherwise, see if this is a subframework header. If so, this is relative 318 // to one of the headers on the #include stack. Walk the list of the current 319 // headers on the #include stack and pass them to HeaderInfo. 320 if (IsFileLexer()) { 321 if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID()))) 322 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd, 323 CurFileEnt))) 324 return FE; 325 } 326 327 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 328 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 329 if (IsFileLexer(ISEntry)) { 330 if ((CurFileEnt = 331 SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID()))) 332 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, 333 FilenameEnd, CurFileEnt))) 334 return FE; 335 } 336 } 337 338 // Otherwise, we really couldn't find the file. 339 return 0; 340} 341 342 343//===----------------------------------------------------------------------===// 344// Preprocessor Directive Handling. 345//===----------------------------------------------------------------------===// 346 347/// HandleDirective - This callback is invoked when the lexer sees a # token 348/// at the start of a line. This consumes the directive, modifies the 349/// lexer/preprocessor state, and advances the lexer(s) so that the next token 350/// read is the correct one. 351void Preprocessor::HandleDirective(Token &Result) { 352 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 353 354 // We just parsed a # character at the start of a line, so we're in directive 355 // mode. Tell the lexer this so any newlines we see will be converted into an 356 // EOM token (which terminates the directive). 357 CurPPLexer->ParsingPreprocessorDirective = true; 358 359 ++NumDirectives; 360 361 // We are about to read a token. For the multiple-include optimization FA to 362 // work, we have to remember if we had read any tokens *before* this 363 // pp-directive. 364 bool ReadAnyTokensBeforeDirective = CurPPLexer->MIOpt.getHasReadAnyTokensVal(); 365 366 // Read the next token, the directive flavor. This isn't expanded due to 367 // C99 6.10.3p8. 368 LexUnexpandedToken(Result); 369 370 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 371 // #define A(x) #x 372 // A(abc 373 // #warning blah 374 // def) 375 // If so, the user is relying on non-portable behavior, emit a diagnostic. 376 if (InMacroArgs) 377 Diag(Result, diag::ext_embedded_directive); 378 379TryAgain: 380 switch (Result.getKind()) { 381 case tok::eom: 382 return; // null directive. 383 case tok::comment: 384 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 385 LexUnexpandedToken(Result); 386 goto TryAgain; 387 388 case tok::numeric_constant: 389 // FIXME: implement # 7 line numbers! 390 DiscardUntilEndOfDirective(); 391 return; 392 default: 393 IdentifierInfo *II = Result.getIdentifierInfo(); 394 if (II == 0) break; // Not an identifier. 395 396 // Ask what the preprocessor keyword ID is. 397 switch (II->getPPKeywordID()) { 398 default: break; 399 // C99 6.10.1 - Conditional Inclusion. 400 case tok::pp_if: 401 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 402 case tok::pp_ifdef: 403 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 404 case tok::pp_ifndef: 405 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 406 case tok::pp_elif: 407 return HandleElifDirective(Result); 408 case tok::pp_else: 409 return HandleElseDirective(Result); 410 case tok::pp_endif: 411 return HandleEndifDirective(Result); 412 413 // C99 6.10.2 - Source File Inclusion. 414 case tok::pp_include: 415 return HandleIncludeDirective(Result); // Handle #include. 416 417 // C99 6.10.3 - Macro Replacement. 418 case tok::pp_define: 419 return HandleDefineDirective(Result); 420 case tok::pp_undef: 421 return HandleUndefDirective(Result); 422 423 // C99 6.10.4 - Line Control. 424 case tok::pp_line: 425 // FIXME: implement #line 426 DiscardUntilEndOfDirective(); 427 return; 428 429 // C99 6.10.5 - Error Directive. 430 case tok::pp_error: 431 return HandleUserDiagnosticDirective(Result, false); 432 433 // C99 6.10.6 - Pragma Directive. 434 case tok::pp_pragma: 435 return HandlePragmaDirective(); 436 437 // GNU Extensions. 438 case tok::pp_import: 439 return HandleImportDirective(Result); 440 case tok::pp_include_next: 441 return HandleIncludeNextDirective(Result); 442 443 case tok::pp_warning: 444 Diag(Result, diag::ext_pp_warning_directive); 445 return HandleUserDiagnosticDirective(Result, true); 446 case tok::pp_ident: 447 return HandleIdentSCCSDirective(Result); 448 case tok::pp_sccs: 449 return HandleIdentSCCSDirective(Result); 450 case tok::pp_assert: 451 //isExtension = true; // FIXME: implement #assert 452 break; 453 case tok::pp_unassert: 454 //isExtension = true; // FIXME: implement #unassert 455 break; 456 } 457 break; 458 } 459 460 // If we reached here, the preprocessing token is not valid! 461 Diag(Result, diag::err_pp_invalid_directive); 462 463 // Read the rest of the PP line. 464 DiscardUntilEndOfDirective(); 465 466 // Okay, we're done parsing the directive. 467} 468 469void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 470 bool isWarning) { 471 // Read the rest of the line raw. We do this because we don't want macros 472 // to be expanded and we don't require that the tokens be valid preprocessing 473 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 474 // collapse multiple consequtive white space between tokens, but this isn't 475 // specified by the standard. 476 477 if (CurLexer) { 478 std::string Message = CurLexer->ReadToEndOfLine(); 479 unsigned DiagID = isWarning ? diag::pp_hash_warning : diag::err_pp_hash_error; 480 Diag(Tok, DiagID) << Message; 481 } 482 else { 483 CurPTHLexer->DiscardToEndOfLine(); 484 } 485} 486 487/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 488/// 489void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 490 // Yes, this directive is an extension. 491 Diag(Tok, diag::ext_pp_ident_directive); 492 493 // Read the string argument. 494 Token StrTok; 495 Lex(StrTok); 496 497 // If the token kind isn't a string, it's a malformed directive. 498 if (StrTok.isNot(tok::string_literal) && 499 StrTok.isNot(tok::wide_string_literal)) { 500 Diag(StrTok, diag::err_pp_malformed_ident); 501 return; 502 } 503 504 // Verify that there is nothing after the string, other than EOM. 505 CheckEndOfDirective("#ident"); 506 507 if (Callbacks) 508 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok)); 509} 510 511//===----------------------------------------------------------------------===// 512// Preprocessor Include Directive Handling. 513//===----------------------------------------------------------------------===// 514 515/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 516/// checked and spelled filename, e.g. as an operand of #include. This returns 517/// true if the input filename was in <>'s or false if it were in ""'s. The 518/// caller is expected to provide a buffer that is large enough to hold the 519/// spelling of the filename, but is also expected to handle the case when 520/// this method decides to use a different buffer. 521bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 522 const char *&BufStart, 523 const char *&BufEnd) { 524 // Get the text form of the filename. 525 assert(BufStart != BufEnd && "Can't have tokens with empty spellings!"); 526 527 // Make sure the filename is <x> or "x". 528 bool isAngled; 529 if (BufStart[0] == '<') { 530 if (BufEnd[-1] != '>') { 531 Diag(Loc, diag::err_pp_expects_filename); 532 BufStart = 0; 533 return true; 534 } 535 isAngled = true; 536 } else if (BufStart[0] == '"') { 537 if (BufEnd[-1] != '"') { 538 Diag(Loc, diag::err_pp_expects_filename); 539 BufStart = 0; 540 return true; 541 } 542 isAngled = false; 543 } else { 544 Diag(Loc, diag::err_pp_expects_filename); 545 BufStart = 0; 546 return true; 547 } 548 549 // Diagnose #include "" as invalid. 550 if (BufEnd-BufStart <= 2) { 551 Diag(Loc, diag::err_pp_empty_filename); 552 BufStart = 0; 553 return ""; 554 } 555 556 // Skip the brackets. 557 ++BufStart; 558 --BufEnd; 559 return isAngled; 560} 561 562/// ConcatenateIncludeName - Handle cases where the #include name is expanded 563/// from a macro as multiple tokens, which need to be glued together. This 564/// occurs for code like: 565/// #define FOO <a/b.h> 566/// #include FOO 567/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 568/// 569/// This code concatenates and consumes tokens up to the '>' token. It returns 570/// false if the > was found, otherwise it returns true if it finds and consumes 571/// the EOM marker. 572static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer, 573 Preprocessor &PP) { 574 Token CurTok; 575 576 PP.Lex(CurTok); 577 while (CurTok.isNot(tok::eom)) { 578 // Append the spelling of this token to the buffer. If there was a space 579 // before it, add it now. 580 if (CurTok.hasLeadingSpace()) 581 FilenameBuffer.push_back(' '); 582 583 // Get the spelling of the token, directly into FilenameBuffer if possible. 584 unsigned PreAppendSize = FilenameBuffer.size(); 585 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 586 587 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 588 unsigned ActualLen = PP.getSpelling(CurTok, BufPtr); 589 590 // If the token was spelled somewhere else, copy it into FilenameBuffer. 591 if (BufPtr != &FilenameBuffer[PreAppendSize]) 592 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 593 594 // Resize FilenameBuffer to the correct size. 595 if (CurTok.getLength() != ActualLen) 596 FilenameBuffer.resize(PreAppendSize+ActualLen); 597 598 // If we found the '>' marker, return success. 599 if (CurTok.is(tok::greater)) 600 return false; 601 602 PP.Lex(CurTok); 603 } 604 605 // If we hit the eom marker, emit an error and return true so that the caller 606 // knows the EOM has been read. 607 PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 608 return true; 609} 610 611/// HandleIncludeDirective - The "#include" tokens have just been read, read the 612/// file to be included from the lexer, then include it! This is a common 613/// routine with functionality shared between #include, #include_next and 614/// #import. LookupFrom is set when this is a #include_next directive, it 615/// specifies the file to start searching from. 616void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 617 const DirectoryLookup *LookupFrom, 618 bool isImport) { 619 620 Token FilenameTok; 621 CurPPLexer->LexIncludeFilename(FilenameTok); 622 623 // Reserve a buffer to get the spelling. 624 llvm::SmallVector<char, 128> FilenameBuffer; 625 const char *FilenameStart, *FilenameEnd; 626 627 switch (FilenameTok.getKind()) { 628 case tok::eom: 629 // If the token kind is EOM, the error has already been diagnosed. 630 return; 631 632 case tok::angle_string_literal: 633 case tok::string_literal: { 634 FilenameBuffer.resize(FilenameTok.getLength()); 635 FilenameStart = &FilenameBuffer[0]; 636 unsigned Len = getSpelling(FilenameTok, FilenameStart); 637 FilenameEnd = FilenameStart+Len; 638 break; 639 } 640 641 case tok::less: 642 // This could be a <foo/bar.h> file coming from a macro expansion. In this 643 // case, glue the tokens together into FilenameBuffer and interpret those. 644 FilenameBuffer.push_back('<'); 645 if (ConcatenateIncludeName(FilenameBuffer, *this)) 646 return; // Found <eom> but no ">"? Diagnostic already emitted. 647 FilenameStart = &FilenameBuffer[0]; 648 FilenameEnd = &FilenameBuffer[FilenameBuffer.size()]; 649 break; 650 default: 651 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 652 DiscardUntilEndOfDirective(); 653 return; 654 } 655 656 bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(), 657 FilenameStart, FilenameEnd); 658 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 659 // error. 660 if (FilenameStart == 0) { 661 DiscardUntilEndOfDirective(); 662 return; 663 } 664 665 // Verify that there is nothing after the filename, other than EOM. Use the 666 // preprocessor to lex this in case lexing the filename entered a macro. 667 CheckEndOfDirective("#include"); 668 669 // Check that we don't have infinite #include recursion. 670 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) { 671 Diag(FilenameTok, diag::err_pp_include_too_deep); 672 return; 673 } 674 675 // Search include directories. 676 const DirectoryLookup *CurDir; 677 const FileEntry *File = LookupFile(FilenameStart, FilenameEnd, 678 isAngled, LookupFrom, CurDir); 679 if (File == 0) { 680 Diag(FilenameTok, diag::err_pp_file_not_found) 681 << std::string(FilenameStart, FilenameEnd); 682 return; 683 } 684 685 // Ask HeaderInfo if we should enter this #include file. If not, #including 686 // this file will have no effect. 687 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) 688 return; 689 690 // The #included file will be considered to be a system header if either it is 691 // in a system include directory, or if the #includer is a system include 692 // header. 693 SrcMgr::CharacteristicKind FileCharacter = 694 std::max(HeaderInfo.getFileDirFlavor(File), 695 SourceMgr.getFileCharacteristic(getCurrentFileLexer()->getFileID())); 696 697 // Look up the file, create a File ID for it. 698 unsigned FileID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 699 FileCharacter); 700 if (FileID == 0) { 701 Diag(FilenameTok, diag::err_pp_file_not_found) 702 << std::string(FilenameStart, FilenameEnd); 703 return; 704 } 705 706 // Finally, if all is good, enter the new file! 707 EnterSourceFile(FileID, CurDir); 708} 709 710/// HandleIncludeNextDirective - Implements #include_next. 711/// 712void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 713 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 714 715 // #include_next is like #include, except that we start searching after 716 // the current found directory. If we can't do this, issue a 717 // diagnostic. 718 const DirectoryLookup *Lookup = CurDirLookup; 719 if (isInPrimaryFile()) { 720 Lookup = 0; 721 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 722 } else if (Lookup == 0) { 723 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 724 } else { 725 // Start looking up in the next directory. 726 ++Lookup; 727 } 728 729 return HandleIncludeDirective(IncludeNextTok, Lookup); 730} 731 732/// HandleImportDirective - Implements #import. 733/// 734void Preprocessor::HandleImportDirective(Token &ImportTok) { 735 Diag(ImportTok, diag::ext_pp_import_directive); 736 737 return HandleIncludeDirective(ImportTok, 0, true); 738} 739 740//===----------------------------------------------------------------------===// 741// Preprocessor Macro Directive Handling. 742//===----------------------------------------------------------------------===// 743 744/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 745/// definition has just been read. Lex the rest of the arguments and the 746/// closing ), updating MI with what we learn. Return true if an error occurs 747/// parsing the arg list. 748bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 749 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 750 751 Token Tok; 752 while (1) { 753 LexUnexpandedToken(Tok); 754 switch (Tok.getKind()) { 755 case tok::r_paren: 756 // Found the end of the argument list. 757 if (Arguments.empty()) { // #define FOO() 758 MI->setArgumentList(Arguments.begin(), Arguments.end()); 759 return false; 760 } 761 // Otherwise we have #define FOO(A,) 762 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 763 return true; 764 case tok::ellipsis: // #define X(... -> C99 varargs 765 // Warn if use of C99 feature in non-C99 mode. 766 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 767 768 // Lex the token after the identifier. 769 LexUnexpandedToken(Tok); 770 if (Tok.isNot(tok::r_paren)) { 771 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 772 return true; 773 } 774 // Add the __VA_ARGS__ identifier as an argument. 775 Arguments.push_back(Ident__VA_ARGS__); 776 MI->setIsC99Varargs(); 777 MI->setArgumentList(Arguments.begin(), Arguments.end()); 778 return false; 779 case tok::eom: // #define X( 780 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 781 return true; 782 default: 783 // Handle keywords and identifiers here to accept things like 784 // #define Foo(for) for. 785 IdentifierInfo *II = Tok.getIdentifierInfo(); 786 if (II == 0) { 787 // #define X(1 788 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 789 return true; 790 } 791 792 // If this is already used as an argument, it is used multiple times (e.g. 793 // #define X(A,A. 794 if (std::find(Arguments.begin(), Arguments.end(), II) != 795 Arguments.end()) { // C99 6.10.3p6 796 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II; 797 return true; 798 } 799 800 // Add the argument to the macro info. 801 Arguments.push_back(II); 802 803 // Lex the token after the identifier. 804 LexUnexpandedToken(Tok); 805 806 switch (Tok.getKind()) { 807 default: // #define X(A B 808 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 809 return true; 810 case tok::r_paren: // #define X(A) 811 MI->setArgumentList(Arguments.begin(), Arguments.end()); 812 return false; 813 case tok::comma: // #define X(A, 814 break; 815 case tok::ellipsis: // #define X(A... -> GCC extension 816 // Diagnose extension. 817 Diag(Tok, diag::ext_named_variadic_macro); 818 819 // Lex the token after the identifier. 820 LexUnexpandedToken(Tok); 821 if (Tok.isNot(tok::r_paren)) { 822 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 823 return true; 824 } 825 826 MI->setIsGNUVarargs(); 827 MI->setArgumentList(Arguments.begin(), Arguments.end()); 828 return false; 829 } 830 } 831 } 832} 833 834/// HandleDefineDirective - Implements #define. This consumes the entire macro 835/// line then lets the caller lex the next real token. 836void Preprocessor::HandleDefineDirective(Token &DefineTok) { 837 ++NumDefined; 838 839 Token MacroNameTok; 840 ReadMacroName(MacroNameTok, 1); 841 842 // Error reading macro name? If so, diagnostic already issued. 843 if (MacroNameTok.is(tok::eom)) 844 return; 845 846 // If we are supposed to keep comments in #defines, reenable comment saving 847 // mode. 848 if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments); 849 850 // Create the new macro. 851 MacroInfo *MI = new MacroInfo(MacroNameTok.getLocation()); 852 853 Token Tok; 854 LexUnexpandedToken(Tok); 855 856 // If this is a function-like macro definition, parse the argument list, 857 // marking each of the identifiers as being used as macro arguments. Also, 858 // check other constraints on the first token of the macro body. 859 if (Tok.is(tok::eom)) { 860 // If there is no body to this macro, we have no special handling here. 861 } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) { 862 // This is a function-like macro definition. Read the argument list. 863 MI->setIsFunctionLike(); 864 if (ReadMacroDefinitionArgList(MI)) { 865 // Forget about MI. 866 delete MI; 867 // Throw away the rest of the line. 868 if (CurPPLexer->ParsingPreprocessorDirective) 869 DiscardUntilEndOfDirective(); 870 return; 871 } 872 873 // Read the first token after the arg list for down below. 874 LexUnexpandedToken(Tok); 875 } else if (!Tok.hasLeadingSpace()) { 876 // C99 requires whitespace between the macro definition and the body. Emit 877 // a diagnostic for something like "#define X+". 878 if (Features.C99) { 879 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 880 } else { 881 // FIXME: C90/C++ do not get this diagnostic, but it does get a similar 882 // one in some cases! 883 } 884 } else { 885 // This is a normal token with leading space. Clear the leading space 886 // marker on the first token to get proper expansion. 887 Tok.clearFlag(Token::LeadingSpace); 888 } 889 890 // If this is a definition of a variadic C99 function-like macro, not using 891 // the GNU named varargs extension, enabled __VA_ARGS__. 892 893 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 894 // This gets unpoisoned where it is allowed. 895 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 896 if (MI->isC99Varargs()) 897 Ident__VA_ARGS__->setIsPoisoned(false); 898 899 // Read the rest of the macro body. 900 if (MI->isObjectLike()) { 901 // Object-like macros are very simple, just read their body. 902 while (Tok.isNot(tok::eom)) { 903 MI->AddTokenToBody(Tok); 904 // Get the next token of the macro. 905 LexUnexpandedToken(Tok); 906 } 907 908 } else { 909 // Otherwise, read the body of a function-like macro. This has to validate 910 // the # (stringize) operator. 911 while (Tok.isNot(tok::eom)) { 912 MI->AddTokenToBody(Tok); 913 914 // Check C99 6.10.3.2p1: ensure that # operators are followed by macro 915 // parameters in function-like macro expansions. 916 if (Tok.isNot(tok::hash)) { 917 // Get the next token of the macro. 918 LexUnexpandedToken(Tok); 919 continue; 920 } 921 922 // Get the next token of the macro. 923 LexUnexpandedToken(Tok); 924 925 // Not a macro arg identifier? 926 if (!Tok.getIdentifierInfo() || 927 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 928 Diag(Tok, diag::err_pp_stringize_not_parameter); 929 delete MI; 930 931 // Disable __VA_ARGS__ again. 932 Ident__VA_ARGS__->setIsPoisoned(true); 933 return; 934 } 935 936 // Things look ok, add the param name token to the macro. 937 MI->AddTokenToBody(Tok); 938 939 // Get the next token of the macro. 940 LexUnexpandedToken(Tok); 941 } 942 } 943 944 945 // Disable __VA_ARGS__ again. 946 Ident__VA_ARGS__->setIsPoisoned(true); 947 948 // Check that there is no paste (##) operator at the begining or end of the 949 // replacement list. 950 unsigned NumTokens = MI->getNumTokens(); 951 if (NumTokens != 0) { 952 if (MI->getReplacementToken(0).is(tok::hashhash)) { 953 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 954 delete MI; 955 return; 956 } 957 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 958 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 959 delete MI; 960 return; 961 } 962 } 963 964 // If this is the primary source file, remember that this macro hasn't been 965 // used yet. 966 if (isInPrimaryFile()) 967 MI->setIsUsed(false); 968 969 // Finally, if this identifier already had a macro defined for it, verify that 970 // the macro bodies are identical and free the old definition. 971 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 972 if (!OtherMI->isUsed()) 973 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 974 975 // Macros must be identical. This means all tokes and whitespace separation 976 // must be the same. C99 6.10.3.2. 977 if (!MI->isIdenticalTo(*OtherMI, *this)) { 978 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef) 979 << MacroNameTok.getIdentifierInfo(); 980 Diag(OtherMI->getDefinitionLoc(), diag::ext_pp_macro_redef2); 981 } 982 delete OtherMI; 983 } 984 985 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 986} 987 988/// HandleUndefDirective - Implements #undef. 989/// 990void Preprocessor::HandleUndefDirective(Token &UndefTok) { 991 ++NumUndefined; 992 993 Token MacroNameTok; 994 ReadMacroName(MacroNameTok, 2); 995 996 // Error reading macro name? If so, diagnostic already issued. 997 if (MacroNameTok.is(tok::eom)) 998 return; 999 1000 // Check to see if this is the last token on the #undef line. 1001 CheckEndOfDirective("#undef"); 1002 1003 // Okay, we finally have a valid identifier to undef. 1004 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 1005 1006 // If the macro is not defined, this is a noop undef, just return. 1007 if (MI == 0) return; 1008 1009 if (!MI->isUsed()) 1010 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 1011 1012 // Free macro definition. 1013 delete MI; 1014 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 1015} 1016 1017 1018//===----------------------------------------------------------------------===// 1019// Preprocessor Conditional Directive Handling. 1020//===----------------------------------------------------------------------===// 1021 1022/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1023/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1024/// if any tokens have been returned or pp-directives activated before this 1025/// #ifndef has been lexed. 1026/// 1027void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1028 bool ReadAnyTokensBeforeDirective) { 1029 ++NumIf; 1030 Token DirectiveTok = Result; 1031 1032 Token MacroNameTok; 1033 ReadMacroName(MacroNameTok); 1034 1035 // Error reading macro name? If so, diagnostic already issued. 1036 if (MacroNameTok.is(tok::eom)) { 1037 // Skip code until we get to #endif. This helps with recovery by not 1038 // emitting an error when the #endif is reached. 1039 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1040 /*Foundnonskip*/false, /*FoundElse*/false); 1041 return; 1042 } 1043 1044 // Check to see if this is the last token on the #if[n]def line. 1045 CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef"); 1046 1047 if (CurPPLexer->getConditionalStackDepth() == 0) { 1048 // If the start of a top-level #ifdef, inform MIOpt. 1049 if (!ReadAnyTokensBeforeDirective) { 1050 assert(isIfndef && "#ifdef shouldn't reach here"); 1051 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo()); 1052 } else 1053 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1054 } 1055 1056 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1057 MacroInfo *MI = getMacroInfo(MII); 1058 1059 // If there is a macro, process it. 1060 if (MI) // Mark it used. 1061 MI->setIsUsed(true); 1062 1063 // Should we include the stuff contained by this directive? 1064 if (!MI == isIfndef) { 1065 // Yes, remember that we are inside a conditional, then lex the next token. 1066 CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false, 1067 /*foundnonskip*/true, /*foundelse*/false); 1068 } else { 1069 // No, skip the contents of this block and return the first token after it. 1070 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1071 /*Foundnonskip*/false, 1072 /*FoundElse*/false); 1073 } 1074} 1075 1076/// HandleIfDirective - Implements the #if directive. 1077/// 1078void Preprocessor::HandleIfDirective(Token &IfToken, 1079 bool ReadAnyTokensBeforeDirective) { 1080 ++NumIf; 1081 1082 // Parse and evaluation the conditional expression. 1083 IdentifierInfo *IfNDefMacro = 0; 1084 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1085 1086 1087 // If this condition is equivalent to #ifndef X, and if this is the first 1088 // directive seen, handle it for the multiple-include optimization. 1089 if (CurPPLexer->getConditionalStackDepth() == 0) { 1090 if (!ReadAnyTokensBeforeDirective && IfNDefMacro) 1091 CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1092 else 1093 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1094 } 1095 1096 // Should we include the stuff contained by this directive? 1097 if (ConditionalTrue) { 1098 // Yes, remember that we are inside a conditional, then lex the next token. 1099 CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1100 /*foundnonskip*/true, /*foundelse*/false); 1101 } else { 1102 // No, skip the contents of this block and return the first token after it. 1103 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1104 /*FoundElse*/false); 1105 } 1106} 1107 1108/// HandleEndifDirective - Implements the #endif directive. 1109/// 1110void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1111 ++NumEndif; 1112 1113 // Check that this is the whole directive. 1114 CheckEndOfDirective("#endif"); 1115 1116 PPConditionalInfo CondInfo; 1117 if (CurPPLexer->popConditionalLevel(CondInfo)) { 1118 // No conditionals on the stack: this is an #endif without an #if. 1119 Diag(EndifToken, diag::err_pp_endif_without_if); 1120 return; 1121 } 1122 1123 // If this the end of a top-level #endif, inform MIOpt. 1124 if (CurPPLexer->getConditionalStackDepth() == 0) 1125 CurPPLexer->MIOpt.ExitTopLevelConditional(); 1126 1127 assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode && 1128 "This code should only be reachable in the non-skipping case!"); 1129} 1130 1131 1132void Preprocessor::HandleElseDirective(Token &Result) { 1133 ++NumElse; 1134 1135 // #else directive in a non-skipping conditional... start skipping. 1136 CheckEndOfDirective("#else"); 1137 1138 PPConditionalInfo CI; 1139 if (CurPPLexer->popConditionalLevel(CI)) { 1140 Diag(Result, diag::pp_err_else_without_if); 1141 return; 1142 } 1143 1144 // If this is a top-level #else, inform the MIOpt. 1145 if (CurPPLexer->getConditionalStackDepth() == 0) 1146 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1147 1148 // If this is a #else with a #else before it, report the error. 1149 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1150 1151 // Finally, skip the rest of the contents of this block and return the first 1152 // token after it. 1153 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1154 /*FoundElse*/true); 1155} 1156 1157void Preprocessor::HandleElifDirective(Token &ElifToken) { 1158 ++NumElse; 1159 1160 // #elif directive in a non-skipping conditional... start skipping. 1161 // We don't care what the condition is, because we will always skip it (since 1162 // the block immediately before it was included). 1163 DiscardUntilEndOfDirective(); 1164 1165 PPConditionalInfo CI; 1166 if (CurPPLexer->popConditionalLevel(CI)) { 1167 Diag(ElifToken, diag::pp_err_elif_without_if); 1168 return; 1169 } 1170 1171 // If this is a top-level #elif, inform the MIOpt. 1172 if (CurPPLexer->getConditionalStackDepth() == 0) 1173 CurPPLexer->MIOpt.EnterTopLevelConditional(); 1174 1175 // If this is a #elif with a #else before it, report the error. 1176 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1177 1178 // Finally, skip the rest of the contents of this block and return the first 1179 // token after it. 1180 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1181 /*FoundElse*/CI.FoundElse); 1182} 1183 1184