PPDirectives.cpp revision 7bfaaaecb3113f955db31e8d8a51acffd1bc0c27
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/Lex/PPCallbacks.h" 18#include "clang/Basic/Diagnostic.h" 19#include "clang/Basic/SourceManager.h" 20using namespace clang; 21 22//===----------------------------------------------------------------------===// 23// Utility Methods for Preprocessor Directive Handling. 24//===----------------------------------------------------------------------===// 25 26/// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 27/// current line until the tok::eom token is found. 28void Preprocessor::DiscardUntilEndOfDirective() { 29 Token Tmp; 30 do { 31 LexUnexpandedToken(Tmp); 32 } while (Tmp.isNot(tok::eom)); 33} 34 35/// isCXXNamedOperator - Returns "true" if the token is a named operator in C++. 36static bool isCXXNamedOperator(const std::string &Spelling) { 37 return Spelling == "and" || Spelling == "bitand" || Spelling == "bitor" || 38 Spelling == "compl" || Spelling == "not" || Spelling == "not_eq" || 39 Spelling == "or" || Spelling == "xor"; 40} 41 42/// ReadMacroName - Lex and validate a macro name, which occurs after a 43/// #define or #undef. This sets the token kind to eom and discards the rest 44/// of the macro line if the macro name is invalid. isDefineUndef is 1 if 45/// this is due to a a #define, 2 if #undef directive, 0 if it is something 46/// else (e.g. #ifdef). 47void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) { 48 // Read the token, don't allow macro expansion on it. 49 LexUnexpandedToken(MacroNameTok); 50 51 // Missing macro name? 52 if (MacroNameTok.is(tok::eom)) 53 return Diag(MacroNameTok, diag::err_pp_missing_macro_name); 54 55 IdentifierInfo *II = MacroNameTok.getIdentifierInfo(); 56 if (II == 0) { 57 std::string Spelling = getSpelling(MacroNameTok); 58 if (isCXXNamedOperator(Spelling)) 59 // C++ 2.5p2: Alternative tokens behave the same as its primary token 60 // except for their spellings. 61 Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name, Spelling); 62 else 63 Diag(MacroNameTok, diag::err_pp_macro_not_identifier); 64 // Fall through on error. 65 } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) { 66 // Error if defining "defined": C99 6.10.8.4. 67 Diag(MacroNameTok, diag::err_defined_macro_name); 68 } else if (isDefineUndef && II->hasMacroDefinition() && 69 getMacroInfo(II)->isBuiltinMacro()) { 70 // Error if defining "__LINE__" and other builtins: C99 6.10.8.4. 71 if (isDefineUndef == 1) 72 Diag(MacroNameTok, diag::pp_redef_builtin_macro); 73 else 74 Diag(MacroNameTok, diag::pp_undef_builtin_macro); 75 } else { 76 // Okay, we got a good identifier node. Return it. 77 return; 78 } 79 80 // Invalid macro name, read and discard the rest of the line. Then set the 81 // token kind to tok::eom. 82 MacroNameTok.setKind(tok::eom); 83 return DiscardUntilEndOfDirective(); 84} 85 86/// CheckEndOfDirective - Ensure that the next token is a tok::eom token. If 87/// not, emit a diagnostic and consume up until the eom. 88void Preprocessor::CheckEndOfDirective(const char *DirType) { 89 Token Tmp; 90 // Lex unexpanded tokens: macros might expand to zero tokens, causing us to 91 // miss diagnosing invalid lines. 92 LexUnexpandedToken(Tmp); 93 94 // There should be no tokens after the directive, but we allow them as an 95 // extension. 96 while (Tmp.is(tok::comment)) // Skip comments in -C mode. 97 LexUnexpandedToken(Tmp); 98 99 if (Tmp.isNot(tok::eom)) { 100 Diag(Tmp, diag::ext_pp_extra_tokens_at_eol, DirType); 101 DiscardUntilEndOfDirective(); 102 } 103} 104 105 106 107/// SkipExcludedConditionalBlock - We just read a #if or related directive and 108/// decided that the subsequent tokens are in the #if'd out portion of the 109/// file. Lex the rest of the file, until we see an #endif. If 110/// FoundNonSkipPortion is true, then we have already emitted code for part of 111/// this #if directive, so #else/#elif blocks should never be entered. If ElseOk 112/// is true, then #else directives are ok, if not, then we have already seen one 113/// so a #else directive is a duplicate. When this returns, the caller can lex 114/// the first valid token. 115void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 116 bool FoundNonSkipPortion, 117 bool FoundElse) { 118 ++NumSkipped; 119 assert(CurTokenLexer == 0 && CurLexer && 120 "Lexing a macro, not a file?"); 121 122 CurLexer->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 CurLexer->LexingRawMode = true; 128 Token Tok; 129 while (1) { 130 CurLexer->Lex(Tok); 131 132 // If this is the end of the buffer, we have an error. 133 if (Tok.is(tok::eof)) { 134 // Emit errors for each unterminated conditional on the stack, including 135 // the current one. 136 while (!CurLexer->ConditionalStack.empty()) { 137 Diag(CurLexer->ConditionalStack.back().IfLoc, 138 diag::err_pp_unterminated_conditional); 139 CurLexer->ConditionalStack.pop_back(); 140 } 141 142 // Just return and let the caller lex after this #include. 143 break; 144 } 145 146 // If this token is not a preprocessor directive, just skip it. 147 if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine()) 148 continue; 149 150 // We just parsed a # character at the start of a line, so we're in 151 // directive mode. Tell the lexer this so any newlines we see will be 152 // converted into an EOM token (this terminates the macro). 153 CurLexer->ParsingPreprocessorDirective = true; 154 CurLexer->KeepCommentMode = false; 155 156 157 // Read the next token, the directive flavor. 158 LexUnexpandedToken(Tok); 159 160 // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or 161 // something bogus), skip it. 162 if (Tok.isNot(tok::identifier)) { 163 CurLexer->ParsingPreprocessorDirective = false; 164 // Restore comment saving mode. 165 CurLexer->KeepCommentMode = KeepComments; 166 continue; 167 } 168 169 // If the first letter isn't i or e, it isn't intesting to us. We know that 170 // this is safe in the face of spelling differences, because there is no way 171 // to spell an i/e in a strange way that is another letter. Skipping this 172 // allows us to avoid looking up the identifier info for #define/#undef and 173 // other common directives. 174 const char *RawCharData = SourceMgr.getCharacterData(Tok.getLocation()); 175 char FirstChar = RawCharData[0]; 176 if (FirstChar >= 'a' && FirstChar <= 'z' && 177 FirstChar != 'i' && FirstChar != 'e') { 178 CurLexer->ParsingPreprocessorDirective = false; 179 // Restore comment saving mode. 180 CurLexer->KeepCommentMode = KeepComments; 181 continue; 182 } 183 184 // Get the identifier name without trigraphs or embedded newlines. Note 185 // that we can't use Tok.getIdentifierInfo() because its lookup is disabled 186 // when skipping. 187 // TODO: could do this with zero copies in the no-clean case by using 188 // strncmp below. 189 char Directive[20]; 190 unsigned IdLen; 191 if (!Tok.needsCleaning() && Tok.getLength() < 20) { 192 IdLen = Tok.getLength(); 193 memcpy(Directive, RawCharData, IdLen); 194 Directive[IdLen] = 0; 195 } else { 196 std::string DirectiveStr = getSpelling(Tok); 197 IdLen = DirectiveStr.size(); 198 if (IdLen >= 20) { 199 CurLexer->ParsingPreprocessorDirective = false; 200 // Restore comment saving mode. 201 CurLexer->KeepCommentMode = KeepComments; 202 continue; 203 } 204 memcpy(Directive, &DirectiveStr[0], IdLen); 205 Directive[IdLen] = 0; 206 } 207 208 if (FirstChar == 'i' && Directive[1] == 'f') { 209 if ((IdLen == 2) || // "if" 210 (IdLen == 5 && !strcmp(Directive+2, "def")) || // "ifdef" 211 (IdLen == 6 && !strcmp(Directive+2, "ndef"))) { // "ifndef" 212 // We know the entire #if/#ifdef/#ifndef block will be skipped, don't 213 // bother parsing the condition. 214 DiscardUntilEndOfDirective(); 215 CurLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true, 216 /*foundnonskip*/false, 217 /*fnddelse*/false); 218 } 219 } else if (FirstChar == 'e') { 220 if (IdLen == 5 && !strcmp(Directive+1, "ndif")) { // "endif" 221 CheckEndOfDirective("#endif"); 222 PPConditionalInfo CondInfo; 223 CondInfo.WasSkipping = true; // Silence bogus warning. 224 bool InCond = CurLexer->popConditionalLevel(CondInfo); 225 InCond = InCond; // Silence warning in no-asserts mode. 226 assert(!InCond && "Can't be skipping if not in a conditional!"); 227 228 // If we popped the outermost skipping block, we're done skipping! 229 if (!CondInfo.WasSkipping) 230 break; 231 } else if (IdLen == 4 && !strcmp(Directive+1, "lse")) { // "else". 232 // #else directive in a skipping conditional. If not in some other 233 // skipping conditional, and if #else hasn't already been seen, enter it 234 // as a non-skipping conditional. 235 CheckEndOfDirective("#else"); 236 PPConditionalInfo &CondInfo = CurLexer->peekConditionalLevel(); 237 238 // If this is a #else with a #else before it, report the error. 239 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else); 240 241 // Note that we've seen a #else in this conditional. 242 CondInfo.FoundElse = true; 243 244 // If the conditional is at the top level, and the #if block wasn't 245 // entered, enter the #else block now. 246 if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) { 247 CondInfo.FoundNonSkip = true; 248 break; 249 } 250 } else if (IdLen == 4 && !strcmp(Directive+1, "lif")) { // "elif". 251 PPConditionalInfo &CondInfo = CurLexer->peekConditionalLevel(); 252 253 bool ShouldEnter; 254 // If this is in a skipping block or if we're already handled this #if 255 // block, don't bother parsing the condition. 256 if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) { 257 DiscardUntilEndOfDirective(); 258 ShouldEnter = false; 259 } else { 260 // Restore the value of LexingRawMode so that identifiers are 261 // looked up, etc, inside the #elif expression. 262 assert(CurLexer->LexingRawMode && "We have to be skipping here!"); 263 CurLexer->LexingRawMode = false; 264 IdentifierInfo *IfNDefMacro = 0; 265 ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro); 266 CurLexer->LexingRawMode = true; 267 } 268 269 // If this is a #elif with a #else before it, report the error. 270 if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else); 271 272 // If this condition is true, enter it! 273 if (ShouldEnter) { 274 CondInfo.FoundNonSkip = true; 275 break; 276 } 277 } 278 } 279 280 CurLexer->ParsingPreprocessorDirective = false; 281 // Restore comment saving mode. 282 CurLexer->KeepCommentMode = KeepComments; 283 } 284 285 // Finally, if we are out of the conditional (saw an #endif or ran off the end 286 // of the file, just stop skipping and return to lexing whatever came after 287 // the #if block. 288 CurLexer->LexingRawMode = false; 289} 290 291/// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 292/// return null on failure. isAngled indicates whether the file reference is 293/// for system #include's or not (i.e. using <> instead of ""). 294const FileEntry *Preprocessor::LookupFile(const char *FilenameStart, 295 const char *FilenameEnd, 296 bool isAngled, 297 const DirectoryLookup *FromDir, 298 const DirectoryLookup *&CurDir) { 299 // If the header lookup mechanism may be relative to the current file, pass in 300 // info about where the current file is. 301 const FileEntry *CurFileEnt = 0; 302 if (!FromDir) { 303 SourceLocation FileLoc = getCurrentFileLexer()->getFileLoc(); 304 CurFileEnt = SourceMgr.getFileEntryForLoc(FileLoc); 305 } 306 307 // Do a standard file entry lookup. 308 CurDir = CurDirLookup; 309 const FileEntry *FE = 310 HeaderInfo.LookupFile(FilenameStart, FilenameEnd, 311 isAngled, FromDir, CurDir, CurFileEnt); 312 if (FE) return FE; 313 314 // Otherwise, see if this is a subframework header. If so, this is relative 315 // to one of the headers on the #include stack. Walk the list of the current 316 // headers on the #include stack and pass them to HeaderInfo. 317 if (CurLexer && !CurLexer->Is_PragmaLexer) { 318 if ((CurFileEnt = SourceMgr.getFileEntryForLoc(CurLexer->getFileLoc()))) 319 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, FilenameEnd, 320 CurFileEnt))) 321 return FE; 322 } 323 324 for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) { 325 IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1]; 326 if (ISEntry.TheLexer && !ISEntry.TheLexer->Is_PragmaLexer) { 327 if ((CurFileEnt = 328 SourceMgr.getFileEntryForLoc(ISEntry.TheLexer->getFileLoc()))) 329 if ((FE = HeaderInfo.LookupSubframeworkHeader(FilenameStart, 330 FilenameEnd, CurFileEnt))) 331 return FE; 332 } 333 } 334 335 // Otherwise, we really couldn't find the file. 336 return 0; 337} 338 339 340//===----------------------------------------------------------------------===// 341// Preprocessor Directive Handling. 342//===----------------------------------------------------------------------===// 343 344/// HandleDirective - This callback is invoked when the lexer sees a # token 345/// at the start of a line. This consumes the directive, modifies the 346/// lexer/preprocessor state, and advances the lexer(s) so that the next token 347/// read is the correct one. 348void Preprocessor::HandleDirective(Token &Result) { 349 // FIXME: Traditional: # with whitespace before it not recognized by K&R? 350 351 // We just parsed a # character at the start of a line, so we're in directive 352 // mode. Tell the lexer this so any newlines we see will be converted into an 353 // EOM token (which terminates the directive). 354 CurLexer->ParsingPreprocessorDirective = true; 355 356 ++NumDirectives; 357 358 // We are about to read a token. For the multiple-include optimization FA to 359 // work, we have to remember if we had read any tokens *before* this 360 // pp-directive. 361 bool ReadAnyTokensBeforeDirective = CurLexer->MIOpt.getHasReadAnyTokensVal(); 362 363 // Read the next token, the directive flavor. This isn't expanded due to 364 // C99 6.10.3p8. 365 LexUnexpandedToken(Result); 366 367 // C99 6.10.3p11: Is this preprocessor directive in macro invocation? e.g.: 368 // #define A(x) #x 369 // A(abc 370 // #warning blah 371 // def) 372 // If so, the user is relying on non-portable behavior, emit a diagnostic. 373 if (InMacroArgs) 374 Diag(Result, diag::ext_embedded_directive); 375 376TryAgain: 377 switch (Result.getKind()) { 378 case tok::eom: 379 return; // null directive. 380 case tok::comment: 381 // Handle stuff like "# /*foo*/ define X" in -E -C mode. 382 LexUnexpandedToken(Result); 383 goto TryAgain; 384 385 case tok::numeric_constant: 386 // FIXME: implement # 7 line numbers! 387 DiscardUntilEndOfDirective(); 388 return; 389 default: 390 IdentifierInfo *II = Result.getIdentifierInfo(); 391 if (II == 0) break; // Not an identifier. 392 393 // Ask what the preprocessor keyword ID is. 394 switch (II->getPPKeywordID()) { 395 default: break; 396 // C99 6.10.1 - Conditional Inclusion. 397 case tok::pp_if: 398 return HandleIfDirective(Result, ReadAnyTokensBeforeDirective); 399 case tok::pp_ifdef: 400 return HandleIfdefDirective(Result, false, true/*not valid for miopt*/); 401 case tok::pp_ifndef: 402 return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective); 403 case tok::pp_elif: 404 return HandleElifDirective(Result); 405 case tok::pp_else: 406 return HandleElseDirective(Result); 407 case tok::pp_endif: 408 return HandleEndifDirective(Result); 409 410 // C99 6.10.2 - Source File Inclusion. 411 case tok::pp_include: 412 return HandleIncludeDirective(Result); // Handle #include. 413 414 // C99 6.10.3 - Macro Replacement. 415 case tok::pp_define: 416 return HandleDefineDirective(Result); 417 case tok::pp_undef: 418 return HandleUndefDirective(Result); 419 420 // C99 6.10.4 - Line Control. 421 case tok::pp_line: 422 // FIXME: implement #line 423 DiscardUntilEndOfDirective(); 424 return; 425 426 // C99 6.10.5 - Error Directive. 427 case tok::pp_error: 428 return HandleUserDiagnosticDirective(Result, false); 429 430 // C99 6.10.6 - Pragma Directive. 431 case tok::pp_pragma: 432 return HandlePragmaDirective(); 433 434 // GNU Extensions. 435 case tok::pp_import: 436 return HandleImportDirective(Result); 437 case tok::pp_include_next: 438 return HandleIncludeNextDirective(Result); 439 440 case tok::pp_warning: 441 Diag(Result, diag::ext_pp_warning_directive); 442 return HandleUserDiagnosticDirective(Result, true); 443 case tok::pp_ident: 444 return HandleIdentSCCSDirective(Result); 445 case tok::pp_sccs: 446 return HandleIdentSCCSDirective(Result); 447 case tok::pp_assert: 448 //isExtension = true; // FIXME: implement #assert 449 break; 450 case tok::pp_unassert: 451 //isExtension = true; // FIXME: implement #unassert 452 break; 453 } 454 break; 455 } 456 457 // If we reached here, the preprocessing token is not valid! 458 Diag(Result, diag::err_pp_invalid_directive); 459 460 // Read the rest of the PP line. 461 DiscardUntilEndOfDirective(); 462 463 // Okay, we're done parsing the directive. 464} 465 466void Preprocessor::HandleUserDiagnosticDirective(Token &Tok, 467 bool isWarning) { 468 // Read the rest of the line raw. We do this because we don't want macros 469 // to be expanded and we don't require that the tokens be valid preprocessing 470 // tokens. For example, this is allowed: "#warning ` 'foo". GCC does 471 // collapse multiple consequtive white space between tokens, but this isn't 472 // specified by the standard. 473 std::string Message = CurLexer->ReadToEndOfLine(); 474 475 unsigned DiagID = isWarning ? diag::pp_hash_warning : diag::err_pp_hash_error; 476 return Diag(Tok, DiagID, Message); 477} 478 479/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive. 480/// 481void Preprocessor::HandleIdentSCCSDirective(Token &Tok) { 482 // Yes, this directive is an extension. 483 Diag(Tok, diag::ext_pp_ident_directive); 484 485 // Read the string argument. 486 Token StrTok; 487 Lex(StrTok); 488 489 // If the token kind isn't a string, it's a malformed directive. 490 if (StrTok.isNot(tok::string_literal) && 491 StrTok.isNot(tok::wide_string_literal)) 492 return Diag(StrTok, diag::err_pp_malformed_ident); 493 494 // Verify that there is nothing after the string, other than EOM. 495 CheckEndOfDirective("#ident"); 496 497 if (Callbacks) 498 Callbacks->Ident(Tok.getLocation(), getSpelling(StrTok)); 499} 500 501//===----------------------------------------------------------------------===// 502// Preprocessor Include Directive Handling. 503//===----------------------------------------------------------------------===// 504 505/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 506/// checked and spelled filename, e.g. as an operand of #include. This returns 507/// true if the input filename was in <>'s or false if it were in ""'s. The 508/// caller is expected to provide a buffer that is large enough to hold the 509/// spelling of the filename, but is also expected to handle the case when 510/// this method decides to use a different buffer. 511bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc, 512 const char *&BufStart, 513 const char *&BufEnd) { 514 // Get the text form of the filename. 515 assert(BufStart != BufEnd && "Can't have tokens with empty spellings!"); 516 517 // Make sure the filename is <x> or "x". 518 bool isAngled; 519 if (BufStart[0] == '<') { 520 if (BufEnd[-1] != '>') { 521 Diag(Loc, diag::err_pp_expects_filename); 522 BufStart = 0; 523 return true; 524 } 525 isAngled = true; 526 } else if (BufStart[0] == '"') { 527 if (BufEnd[-1] != '"') { 528 Diag(Loc, diag::err_pp_expects_filename); 529 BufStart = 0; 530 return true; 531 } 532 isAngled = false; 533 } else { 534 Diag(Loc, diag::err_pp_expects_filename); 535 BufStart = 0; 536 return true; 537 } 538 539 // Diagnose #include "" as invalid. 540 if (BufEnd-BufStart <= 2) { 541 Diag(Loc, diag::err_pp_empty_filename); 542 BufStart = 0; 543 return ""; 544 } 545 546 // Skip the brackets. 547 ++BufStart; 548 --BufEnd; 549 return isAngled; 550} 551 552/// ConcatenateIncludeName - Handle cases where the #include name is expanded 553/// from a macro as multiple tokens, which need to be glued together. This 554/// occurs for code like: 555/// #define FOO <a/b.h> 556/// #include FOO 557/// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 558/// 559/// This code concatenates and consumes tokens up to the '>' token. It returns 560/// false if the > was found, otherwise it returns true if it finds and consumes 561/// the EOM marker. 562static bool ConcatenateIncludeName(llvm::SmallVector<char, 128> &FilenameBuffer, 563 Preprocessor &PP) { 564 Token CurTok; 565 566 PP.Lex(CurTok); 567 while (CurTok.isNot(tok::eom)) { 568 // Append the spelling of this token to the buffer. If there was a space 569 // before it, add it now. 570 if (CurTok.hasLeadingSpace()) 571 FilenameBuffer.push_back(' '); 572 573 // Get the spelling of the token, directly into FilenameBuffer if possible. 574 unsigned PreAppendSize = FilenameBuffer.size(); 575 FilenameBuffer.resize(PreAppendSize+CurTok.getLength()); 576 577 const char *BufPtr = &FilenameBuffer[PreAppendSize]; 578 unsigned ActualLen = PP.getSpelling(CurTok, BufPtr); 579 580 // If the token was spelled somewhere else, copy it into FilenameBuffer. 581 if (BufPtr != &FilenameBuffer[PreAppendSize]) 582 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen); 583 584 // Resize FilenameBuffer to the correct size. 585 if (CurTok.getLength() != ActualLen) 586 FilenameBuffer.resize(PreAppendSize+ActualLen); 587 588 // If we found the '>' marker, return success. 589 if (CurTok.is(tok::greater)) 590 return false; 591 592 PP.Lex(CurTok); 593 } 594 595 // If we hit the eom marker, emit an error and return true so that the caller 596 // knows the EOM has been read. 597 PP.Diag(CurTok.getLocation(), diag::err_pp_expects_filename); 598 return true; 599} 600 601/// HandleIncludeDirective - The "#include" tokens have just been read, read the 602/// file to be included from the lexer, then include it! This is a common 603/// routine with functionality shared between #include, #include_next and 604/// #import. 605void Preprocessor::HandleIncludeDirective(Token &IncludeTok, 606 const DirectoryLookup *LookupFrom, 607 bool isImport) { 608 609 Token FilenameTok; 610 CurLexer->LexIncludeFilename(FilenameTok); 611 612 // Reserve a buffer to get the spelling. 613 llvm::SmallVector<char, 128> FilenameBuffer; 614 const char *FilenameStart, *FilenameEnd; 615 616 switch (FilenameTok.getKind()) { 617 case tok::eom: 618 // If the token kind is EOM, the error has already been diagnosed. 619 return; 620 621 case tok::angle_string_literal: 622 case tok::string_literal: { 623 FilenameBuffer.resize(FilenameTok.getLength()); 624 FilenameStart = &FilenameBuffer[0]; 625 unsigned Len = getSpelling(FilenameTok, FilenameStart); 626 FilenameEnd = FilenameStart+Len; 627 break; 628 } 629 630 case tok::less: 631 // This could be a <foo/bar.h> file coming from a macro expansion. In this 632 // case, glue the tokens together into FilenameBuffer and interpret those. 633 FilenameBuffer.push_back('<'); 634 if (ConcatenateIncludeName(FilenameBuffer, *this)) 635 return; // Found <eom> but no ">"? Diagnostic already emitted. 636 FilenameStart = &FilenameBuffer[0]; 637 FilenameEnd = &FilenameBuffer[FilenameBuffer.size()]; 638 break; 639 default: 640 Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename); 641 DiscardUntilEndOfDirective(); 642 return; 643 } 644 645 bool isAngled = GetIncludeFilenameSpelling(FilenameTok.getLocation(), 646 FilenameStart, FilenameEnd); 647 // If GetIncludeFilenameSpelling set the start ptr to null, there was an 648 // error. 649 if (FilenameStart == 0) { 650 DiscardUntilEndOfDirective(); 651 return; 652 } 653 654 // Verify that there is nothing after the filename, other than EOM. Use the 655 // preprocessor to lex this in case lexing the filename entered a macro. 656 CheckEndOfDirective("#include"); 657 658 // Check that we don't have infinite #include recursion. 659 if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) 660 return Diag(FilenameTok, diag::err_pp_include_too_deep); 661 662 // Search include directories. 663 const DirectoryLookup *CurDir; 664 const FileEntry *File = LookupFile(FilenameStart, FilenameEnd, 665 isAngled, LookupFrom, CurDir); 666 if (File == 0) 667 return Diag(FilenameTok, diag::err_pp_file_not_found, 668 std::string(FilenameStart, FilenameEnd)); 669 670 // Ask HeaderInfo if we should enter this #include file. 671 if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) { 672 // If it returns true, #including this file will have no effect. 673 return; 674 } 675 676 // Look up the file, create a File ID for it. 677 unsigned FileID = SourceMgr.createFileID(File, FilenameTok.getLocation(), 678 isSystemHeader(File)); 679 if (FileID == 0) 680 return Diag(FilenameTok, diag::err_pp_file_not_found, 681 std::string(FilenameStart, FilenameEnd)); 682 683 // Finally, if all is good, enter the new file! 684 EnterSourceFile(FileID, CurDir); 685} 686 687/// HandleIncludeNextDirective - Implements #include_next. 688/// 689void Preprocessor::HandleIncludeNextDirective(Token &IncludeNextTok) { 690 Diag(IncludeNextTok, diag::ext_pp_include_next_directive); 691 692 // #include_next is like #include, except that we start searching after 693 // the current found directory. If we can't do this, issue a 694 // diagnostic. 695 const DirectoryLookup *Lookup = CurDirLookup; 696 if (isInPrimaryFile()) { 697 Lookup = 0; 698 Diag(IncludeNextTok, diag::pp_include_next_in_primary); 699 } else if (Lookup == 0) { 700 Diag(IncludeNextTok, diag::pp_include_next_absolute_path); 701 } else { 702 // Start looking up in the next directory. 703 ++Lookup; 704 } 705 706 return HandleIncludeDirective(IncludeNextTok, Lookup); 707} 708 709/// HandleImportDirective - Implements #import. 710/// 711void Preprocessor::HandleImportDirective(Token &ImportTok) { 712 Diag(ImportTok, diag::ext_pp_import_directive); 713 714 return HandleIncludeDirective(ImportTok, 0, true); 715} 716 717//===----------------------------------------------------------------------===// 718// Preprocessor Macro Directive Handling. 719//===----------------------------------------------------------------------===// 720 721/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 722/// definition has just been read. Lex the rest of the arguments and the 723/// closing ), updating MI with what we learn. Return true if an error occurs 724/// parsing the arg list. 725bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI) { 726 llvm::SmallVector<IdentifierInfo*, 32> Arguments; 727 728 Token Tok; 729 while (1) { 730 LexUnexpandedToken(Tok); 731 switch (Tok.getKind()) { 732 case tok::r_paren: 733 // Found the end of the argument list. 734 if (Arguments.empty()) { // #define FOO() 735 MI->setArgumentList(Arguments.begin(), Arguments.end()); 736 return false; 737 } 738 // Otherwise we have #define FOO(A,) 739 Diag(Tok, diag::err_pp_expected_ident_in_arg_list); 740 return true; 741 case tok::ellipsis: // #define X(... -> C99 varargs 742 // Warn if use of C99 feature in non-C99 mode. 743 if (!Features.C99) Diag(Tok, diag::ext_variadic_macro); 744 745 // Lex the token after the identifier. 746 LexUnexpandedToken(Tok); 747 if (Tok.isNot(tok::r_paren)) { 748 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 749 return true; 750 } 751 // Add the __VA_ARGS__ identifier as an argument. 752 Arguments.push_back(Ident__VA_ARGS__); 753 MI->setIsC99Varargs(); 754 MI->setArgumentList(Arguments.begin(), Arguments.end()); 755 return false; 756 case tok::eom: // #define X( 757 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 758 return true; 759 default: 760 // Handle keywords and identifiers here to accept things like 761 // #define Foo(for) for. 762 IdentifierInfo *II = Tok.getIdentifierInfo(); 763 if (II == 0) { 764 // #define X(1 765 Diag(Tok, diag::err_pp_invalid_tok_in_arg_list); 766 return true; 767 } 768 769 // If this is already used as an argument, it is used multiple times (e.g. 770 // #define X(A,A. 771 if (std::find(Arguments.begin(), Arguments.end(), II) != 772 Arguments.end()) { // C99 6.10.3p6 773 Diag(Tok, diag::err_pp_duplicate_name_in_arg_list, II->getName()); 774 return true; 775 } 776 777 // Add the argument to the macro info. 778 Arguments.push_back(II); 779 780 // Lex the token after the identifier. 781 LexUnexpandedToken(Tok); 782 783 switch (Tok.getKind()) { 784 default: // #define X(A B 785 Diag(Tok, diag::err_pp_expected_comma_in_arg_list); 786 return true; 787 case tok::r_paren: // #define X(A) 788 MI->setArgumentList(Arguments.begin(), Arguments.end()); 789 return false; 790 case tok::comma: // #define X(A, 791 break; 792 case tok::ellipsis: // #define X(A... -> GCC extension 793 // Diagnose extension. 794 Diag(Tok, diag::ext_named_variadic_macro); 795 796 // Lex the token after the identifier. 797 LexUnexpandedToken(Tok); 798 if (Tok.isNot(tok::r_paren)) { 799 Diag(Tok, diag::err_pp_missing_rparen_in_macro_def); 800 return true; 801 } 802 803 MI->setIsGNUVarargs(); 804 MI->setArgumentList(Arguments.begin(), Arguments.end()); 805 return false; 806 } 807 } 808 } 809} 810 811/// HandleDefineDirective - Implements #define. This consumes the entire macro 812/// line then lets the caller lex the next real token. 813void Preprocessor::HandleDefineDirective(Token &DefineTok) { 814 ++NumDefined; 815 816 Token MacroNameTok; 817 ReadMacroName(MacroNameTok, 1); 818 819 // Error reading macro name? If so, diagnostic already issued. 820 if (MacroNameTok.is(tok::eom)) 821 return; 822 823 // If we are supposed to keep comments in #defines, reenable comment saving 824 // mode. 825 CurLexer->KeepCommentMode = KeepMacroComments; 826 827 // Create the new macro. 828 MacroInfo *MI = new MacroInfo(MacroNameTok.getLocation()); 829 830 Token Tok; 831 LexUnexpandedToken(Tok); 832 833 // If this is a function-like macro definition, parse the argument list, 834 // marking each of the identifiers as being used as macro arguments. Also, 835 // check other constraints on the first token of the macro body. 836 if (Tok.is(tok::eom)) { 837 // If there is no body to this macro, we have no special handling here. 838 } else if (Tok.is(tok::l_paren) && !Tok.hasLeadingSpace()) { 839 // This is a function-like macro definition. Read the argument list. 840 MI->setIsFunctionLike(); 841 if (ReadMacroDefinitionArgList(MI)) { 842 // Forget about MI. 843 delete MI; 844 // Throw away the rest of the line. 845 if (CurLexer->ParsingPreprocessorDirective) 846 DiscardUntilEndOfDirective(); 847 return; 848 } 849 850 // Read the first token after the arg list for down below. 851 LexUnexpandedToken(Tok); 852 } else if (!Tok.hasLeadingSpace()) { 853 // C99 requires whitespace between the macro definition and the body. Emit 854 // a diagnostic for something like "#define X+". 855 if (Features.C99) { 856 Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name); 857 } else { 858 // FIXME: C90/C++ do not get this diagnostic, but it does get a similar 859 // one in some cases! 860 } 861 } else { 862 // This is a normal token with leading space. Clear the leading space 863 // marker on the first token to get proper expansion. 864 Tok.clearFlag(Token::LeadingSpace); 865 } 866 867 // If this is a definition of a variadic C99 function-like macro, not using 868 // the GNU named varargs extension, enabled __VA_ARGS__. 869 870 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 871 // This gets unpoisoned where it is allowed. 872 assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!"); 873 if (MI->isC99Varargs()) 874 Ident__VA_ARGS__->setIsPoisoned(false); 875 876 // Read the rest of the macro body. 877 if (MI->isObjectLike()) { 878 // Object-like macros are very simple, just read their body. 879 while (Tok.isNot(tok::eom)) { 880 MI->AddTokenToBody(Tok); 881 // Get the next token of the macro. 882 LexUnexpandedToken(Tok); 883 } 884 885 } else { 886 // Otherwise, read the body of a function-like macro. This has to validate 887 // the # (stringize) operator. 888 while (Tok.isNot(tok::eom)) { 889 MI->AddTokenToBody(Tok); 890 891 // Check C99 6.10.3.2p1: ensure that # operators are followed by macro 892 // parameters in function-like macro expansions. 893 if (Tok.isNot(tok::hash)) { 894 // Get the next token of the macro. 895 LexUnexpandedToken(Tok); 896 continue; 897 } 898 899 // Get the next token of the macro. 900 LexUnexpandedToken(Tok); 901 902 // Not a macro arg identifier? 903 if (!Tok.getIdentifierInfo() || 904 MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) { 905 Diag(Tok, diag::err_pp_stringize_not_parameter); 906 delete MI; 907 908 // Disable __VA_ARGS__ again. 909 Ident__VA_ARGS__->setIsPoisoned(true); 910 return; 911 } 912 913 // Things look ok, add the param name token to the macro. 914 MI->AddTokenToBody(Tok); 915 916 // Get the next token of the macro. 917 LexUnexpandedToken(Tok); 918 } 919 } 920 921 922 // Disable __VA_ARGS__ again. 923 Ident__VA_ARGS__->setIsPoisoned(true); 924 925 // Check that there is no paste (##) operator at the begining or end of the 926 // replacement list. 927 unsigned NumTokens = MI->getNumTokens(); 928 if (NumTokens != 0) { 929 if (MI->getReplacementToken(0).is(tok::hashhash)) { 930 Diag(MI->getReplacementToken(0), diag::err_paste_at_start); 931 delete MI; 932 return; 933 } 934 if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) { 935 Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end); 936 delete MI; 937 return; 938 } 939 } 940 941 // If this is the primary source file, remember that this macro hasn't been 942 // used yet. 943 if (isInPrimaryFile()) 944 MI->setIsUsed(false); 945 946 // Finally, if this identifier already had a macro defined for it, verify that 947 // the macro bodies are identical and free the old definition. 948 if (MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo())) { 949 if (!OtherMI->isUsed()) 950 Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used); 951 952 // Macros must be identical. This means all tokes and whitespace separation 953 // must be the same. C99 6.10.3.2. 954 if (!MI->isIdenticalTo(*OtherMI, *this)) { 955 Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef, 956 MacroNameTok.getIdentifierInfo()->getName()); 957 Diag(OtherMI->getDefinitionLoc(), diag::ext_pp_macro_redef2); 958 } 959 delete OtherMI; 960 } 961 962 setMacroInfo(MacroNameTok.getIdentifierInfo(), MI); 963} 964 965/// HandleUndefDirective - Implements #undef. 966/// 967void Preprocessor::HandleUndefDirective(Token &UndefTok) { 968 ++NumUndefined; 969 970 Token MacroNameTok; 971 ReadMacroName(MacroNameTok, 2); 972 973 // Error reading macro name? If so, diagnostic already issued. 974 if (MacroNameTok.is(tok::eom)) 975 return; 976 977 // Check to see if this is the last token on the #undef line. 978 CheckEndOfDirective("#undef"); 979 980 // Okay, we finally have a valid identifier to undef. 981 MacroInfo *MI = getMacroInfo(MacroNameTok.getIdentifierInfo()); 982 983 // If the macro is not defined, this is a noop undef, just return. 984 if (MI == 0) return; 985 986 if (!MI->isUsed()) 987 Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used); 988 989 // Free macro definition. 990 delete MI; 991 setMacroInfo(MacroNameTok.getIdentifierInfo(), 0); 992} 993 994 995//===----------------------------------------------------------------------===// 996// Preprocessor Conditional Directive Handling. 997//===----------------------------------------------------------------------===// 998 999/// HandleIfdefDirective - Implements the #ifdef/#ifndef directive. isIfndef is 1000/// true when this is a #ifndef directive. ReadAnyTokensBeforeDirective is true 1001/// if any tokens have been returned or pp-directives activated before this 1002/// #ifndef has been lexed. 1003/// 1004void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef, 1005 bool ReadAnyTokensBeforeDirective) { 1006 ++NumIf; 1007 Token DirectiveTok = Result; 1008 1009 Token MacroNameTok; 1010 ReadMacroName(MacroNameTok); 1011 1012 // Error reading macro name? If so, diagnostic already issued. 1013 if (MacroNameTok.is(tok::eom)) { 1014 // Skip code until we get to #endif. This helps with recovery by not 1015 // emitting an error when the #endif is reached. 1016 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1017 /*Foundnonskip*/false, /*FoundElse*/false); 1018 return; 1019 } 1020 1021 // Check to see if this is the last token on the #if[n]def line. 1022 CheckEndOfDirective(isIfndef ? "#ifndef" : "#ifdef"); 1023 1024 if (CurLexer->getConditionalStackDepth() == 0) { 1025 // If the start of a top-level #ifdef, inform MIOpt. 1026 if (!ReadAnyTokensBeforeDirective) { 1027 assert(isIfndef && "#ifdef shouldn't reach here"); 1028 CurLexer->MIOpt.EnterTopLevelIFNDEF(MacroNameTok.getIdentifierInfo()); 1029 } else 1030 CurLexer->MIOpt.EnterTopLevelConditional(); 1031 } 1032 1033 IdentifierInfo *MII = MacroNameTok.getIdentifierInfo(); 1034 MacroInfo *MI = getMacroInfo(MII); 1035 1036 // If there is a macro, process it. 1037 if (MI) // Mark it used. 1038 MI->setIsUsed(true); 1039 1040 // Should we include the stuff contained by this directive? 1041 if (!MI == isIfndef) { 1042 // Yes, remember that we are inside a conditional, then lex the next token. 1043 CurLexer->pushConditionalLevel(DirectiveTok.getLocation(), /*wasskip*/false, 1044 /*foundnonskip*/true, /*foundelse*/false); 1045 } else { 1046 // No, skip the contents of this block and return the first token after it. 1047 SkipExcludedConditionalBlock(DirectiveTok.getLocation(), 1048 /*Foundnonskip*/false, 1049 /*FoundElse*/false); 1050 } 1051} 1052 1053/// HandleIfDirective - Implements the #if directive. 1054/// 1055void Preprocessor::HandleIfDirective(Token &IfToken, 1056 bool ReadAnyTokensBeforeDirective) { 1057 ++NumIf; 1058 1059 // Parse and evaluation the conditional expression. 1060 IdentifierInfo *IfNDefMacro = 0; 1061 bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro); 1062 1063 1064 // If this condition is equivalent to #ifndef X, and if this is the first 1065 // directive seen, handle it for the multiple-include optimization. 1066 if (CurLexer->getConditionalStackDepth() == 0) { 1067 if (!ReadAnyTokensBeforeDirective && IfNDefMacro) 1068 CurLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro); 1069 else 1070 CurLexer->MIOpt.EnterTopLevelConditional(); 1071 } 1072 1073 // Should we include the stuff contained by this directive? 1074 if (ConditionalTrue) { 1075 // Yes, remember that we are inside a conditional, then lex the next token. 1076 CurLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false, 1077 /*foundnonskip*/true, /*foundelse*/false); 1078 } else { 1079 // No, skip the contents of this block and return the first token after it. 1080 SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false, 1081 /*FoundElse*/false); 1082 } 1083} 1084 1085/// HandleEndifDirective - Implements the #endif directive. 1086/// 1087void Preprocessor::HandleEndifDirective(Token &EndifToken) { 1088 ++NumEndif; 1089 1090 // Check that this is the whole directive. 1091 CheckEndOfDirective("#endif"); 1092 1093 PPConditionalInfo CondInfo; 1094 if (CurLexer->popConditionalLevel(CondInfo)) { 1095 // No conditionals on the stack: this is an #endif without an #if. 1096 return Diag(EndifToken, diag::err_pp_endif_without_if); 1097 } 1098 1099 // If this the end of a top-level #endif, inform MIOpt. 1100 if (CurLexer->getConditionalStackDepth() == 0) 1101 CurLexer->MIOpt.ExitTopLevelConditional(); 1102 1103 assert(!CondInfo.WasSkipping && !CurLexer->LexingRawMode && 1104 "This code should only be reachable in the non-skipping case!"); 1105} 1106 1107 1108void Preprocessor::HandleElseDirective(Token &Result) { 1109 ++NumElse; 1110 1111 // #else directive in a non-skipping conditional... start skipping. 1112 CheckEndOfDirective("#else"); 1113 1114 PPConditionalInfo CI; 1115 if (CurLexer->popConditionalLevel(CI)) 1116 return Diag(Result, diag::pp_err_else_without_if); 1117 1118 // If this is a top-level #else, inform the MIOpt. 1119 if (CurLexer->getConditionalStackDepth() == 0) 1120 CurLexer->MIOpt.EnterTopLevelConditional(); 1121 1122 // If this is a #else with a #else before it, report the error. 1123 if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else); 1124 1125 // Finally, skip the rest of the contents of this block and return the first 1126 // token after it. 1127 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1128 /*FoundElse*/true); 1129} 1130 1131void Preprocessor::HandleElifDirective(Token &ElifToken) { 1132 ++NumElse; 1133 1134 // #elif directive in a non-skipping conditional... start skipping. 1135 // We don't care what the condition is, because we will always skip it (since 1136 // the block immediately before it was included). 1137 DiscardUntilEndOfDirective(); 1138 1139 PPConditionalInfo CI; 1140 if (CurLexer->popConditionalLevel(CI)) 1141 return Diag(ElifToken, diag::pp_err_elif_without_if); 1142 1143 // If this is a top-level #elif, inform the MIOpt. 1144 if (CurLexer->getConditionalStackDepth() == 0) 1145 CurLexer->MIOpt.EnterTopLevelConditional(); 1146 1147 // If this is a #elif with a #else before it, report the error. 1148 if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else); 1149 1150 // Finally, skip the rest of the contents of this block and return the first 1151 // token after it. 1152 return SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true, 1153 /*FoundElse*/CI.FoundElse); 1154} 1155 1156