Preprocessor.cpp revision 81b747b7fcc91c2fba9a3183d8fac80adbfc1d3e
1//===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===// 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 the Preprocessor interface. 11// 12//===----------------------------------------------------------------------===// 13// 14// Options to support: 15// -H - Print the name of each header file used. 16// -d[DNI] - Dump various things. 17// -fworking-directory - #line's with preprocessor's working dir. 18// -fpreprocessed 19// -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD 20// -W* 21// -w 22// 23// Messages to emit: 24// "Multiple include guards may be useful for:\n" 25// 26//===----------------------------------------------------------------------===// 27 28#include "clang/Lex/Preprocessor.h" 29#include "clang/Lex/HeaderSearch.h" 30#include "clang/Lex/MacroInfo.h" 31#include "clang/Lex/Pragma.h" 32#include "clang/Lex/ScratchBuffer.h" 33#include "clang/Lex/LexDiagnostic.h" 34#include "clang/Basic/SourceManager.h" 35#include "clang/Basic/FileManager.h" 36#include "clang/Basic/TargetInfo.h" 37#include "llvm/ADT/APFloat.h" 38#include "llvm/ADT/SmallVector.h" 39#include "llvm/Support/MemoryBuffer.h" 40#include "llvm/Support/raw_ostream.h" 41#include <cstdio> 42using namespace clang; 43 44//===----------------------------------------------------------------------===// 45 46PreprocessorFactory::~PreprocessorFactory() {} 47 48Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts, 49 TargetInfo &target, SourceManager &SM, 50 HeaderSearch &Headers, 51 IdentifierInfoLookup* IILookup) 52 : Diags(&diags), Features(opts), Target(target),FileMgr(Headers.getFileMgr()), 53 SourceMgr(SM), HeaderInfo(Headers), Identifiers(opts, IILookup), 54 BuiltinInfo(Target), CurPPLexer(0), CurDirLookup(0), Callbacks(0) { 55 ScratchBuf = new ScratchBuffer(SourceMgr); 56 CounterValue = 0; // __COUNTER__ starts at 0. 57 58 // Clear stats. 59 NumDirectives = NumDefined = NumUndefined = NumPragma = 0; 60 NumIf = NumElse = NumEndif = 0; 61 NumEnteredSourceFiles = 0; 62 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0; 63 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0; 64 MaxIncludeStackDepth = 0; 65 NumSkipped = 0; 66 67 // Default to discarding comments. 68 KeepComments = false; 69 KeepMacroComments = false; 70 71 // Macro expansion is enabled. 72 DisableMacroExpansion = false; 73 InMacroArgs = false; 74 IsMainFileEofCodeCompletion = false; 75 NumCachedTokenLexers = 0; 76 77 CachedLexPos = 0; 78 79 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 80 // This gets unpoisoned where it is allowed. 81 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned(); 82 83 // Initialize the pragma handlers. 84 PragmaHandlers = new PragmaNamespace(0); 85 RegisterBuiltinPragmas(); 86 87 // Initialize builtin macros like __LINE__ and friends. 88 RegisterBuiltinMacros(); 89} 90 91Preprocessor::~Preprocessor() { 92 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!"); 93 94 while (!IncludeMacroStack.empty()) { 95 delete IncludeMacroStack.back().TheLexer; 96 delete IncludeMacroStack.back().TheTokenLexer; 97 IncludeMacroStack.pop_back(); 98 } 99 100 // Free any macro definitions. 101 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I = 102 Macros.begin(), E = Macros.end(); I != E; ++I) { 103 // We don't need to free the MacroInfo objects directly. These 104 // will be released when the BumpPtrAllocator 'BP' object gets 105 // destroyed. We still need to run the dstor, however, to free 106 // memory alocated by MacroInfo. 107 I->second->Destroy(BP); 108 I->first->setHasMacroDefinition(false); 109 } 110 111 // Free any cached macro expanders. 112 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i) 113 delete TokenLexerCache[i]; 114 115 // Release pragma information. 116 delete PragmaHandlers; 117 118 // Delete the scratch buffer info. 119 delete ScratchBuf; 120 121 delete Callbacks; 122} 123 124void Preprocessor::setPTHManager(PTHManager* pm) { 125 PTH.reset(pm); 126 FileMgr.setStatCache(PTH->createStatCache()); 127} 128 129void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const { 130 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '" 131 << getSpelling(Tok) << "'"; 132 133 if (!DumpFlags) return; 134 135 llvm::errs() << "\t"; 136 if (Tok.isAtStartOfLine()) 137 llvm::errs() << " [StartOfLine]"; 138 if (Tok.hasLeadingSpace()) 139 llvm::errs() << " [LeadingSpace]"; 140 if (Tok.isExpandDisabled()) 141 llvm::errs() << " [ExpandDisabled]"; 142 if (Tok.needsCleaning()) { 143 const char *Start = SourceMgr.getCharacterData(Tok.getLocation()); 144 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength()) 145 << "']"; 146 } 147 148 llvm::errs() << "\tLoc=<"; 149 DumpLocation(Tok.getLocation()); 150 llvm::errs() << ">"; 151} 152 153void Preprocessor::DumpLocation(SourceLocation Loc) const { 154 Loc.dump(SourceMgr); 155} 156 157void Preprocessor::DumpMacro(const MacroInfo &MI) const { 158 llvm::errs() << "MACRO: "; 159 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) { 160 DumpToken(MI.getReplacementToken(i)); 161 llvm::errs() << " "; 162 } 163 llvm::errs() << "\n"; 164} 165 166void Preprocessor::PrintStats() { 167 llvm::errs() << "\n*** Preprocessor Stats:\n"; 168 llvm::errs() << NumDirectives << " directives found:\n"; 169 llvm::errs() << " " << NumDefined << " #define.\n"; 170 llvm::errs() << " " << NumUndefined << " #undef.\n"; 171 llvm::errs() << " #include/#include_next/#import:\n"; 172 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n"; 173 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n"; 174 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n"; 175 llvm::errs() << " " << NumElse << " #else/#elif.\n"; 176 llvm::errs() << " " << NumEndif << " #endif.\n"; 177 llvm::errs() << " " << NumPragma << " #pragma.\n"; 178 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n"; 179 180 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/" 181 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, " 182 << NumFastMacroExpanded << " on the fast path.\n"; 183 llvm::errs() << (NumFastTokenPaste+NumTokenPaste) 184 << " token paste (##) operations performed, " 185 << NumFastTokenPaste << " on the fast path.\n"; 186} 187 188//===----------------------------------------------------------------------===// 189// Token Spelling 190//===----------------------------------------------------------------------===// 191 192 193/// getSpelling() - Return the 'spelling' of this token. The spelling of a 194/// token are the characters used to represent the token in the source file 195/// after trigraph expansion and escaped-newline folding. In particular, this 196/// wants to get the true, uncanonicalized, spelling of things like digraphs 197/// UCNs, etc. 198std::string Preprocessor::getSpelling(const Token &Tok) const { 199 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 200 201 // If this token contains nothing interesting, return it directly. 202 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 203 if (!Tok.needsCleaning()) 204 return std::string(TokStart, TokStart+Tok.getLength()); 205 206 std::string Result; 207 Result.reserve(Tok.getLength()); 208 209 // Otherwise, hard case, relex the characters into the string. 210 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 211 Ptr != End; ) { 212 unsigned CharSize; 213 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features)); 214 Ptr += CharSize; 215 } 216 assert(Result.size() != unsigned(Tok.getLength()) && 217 "NeedsCleaning flag set on something that didn't need cleaning!"); 218 return Result; 219} 220 221/// getSpelling - This method is used to get the spelling of a token into a 222/// preallocated buffer, instead of as an std::string. The caller is required 223/// to allocate enough space for the token, which is guaranteed to be at least 224/// Tok.getLength() bytes long. The actual length of the token is returned. 225/// 226/// Note that this method may do two possible things: it may either fill in 227/// the buffer specified with characters, or it may *change the input pointer* 228/// to point to a constant buffer with the data already in it (avoiding a 229/// copy). The caller is not allowed to modify the returned buffer pointer 230/// if an internal buffer is returned. 231unsigned Preprocessor::getSpelling(const Token &Tok, 232 const char *&Buffer) const { 233 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 234 235 // If this token is an identifier, just return the string from the identifier 236 // table, which is very quick. 237 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) { 238 Buffer = II->getName(); 239 return II->getLength(); 240 } 241 242 // Otherwise, compute the start of the token in the input lexer buffer. 243 const char *TokStart = 0; 244 245 if (Tok.isLiteral()) 246 TokStart = Tok.getLiteralData(); 247 248 if (TokStart == 0) 249 TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 250 251 // If this token contains nothing interesting, return it directly. 252 if (!Tok.needsCleaning()) { 253 Buffer = TokStart; 254 return Tok.getLength(); 255 } 256 257 // Otherwise, hard case, relex the characters into the string. 258 char *OutBuf = const_cast<char*>(Buffer); 259 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 260 Ptr != End; ) { 261 unsigned CharSize; 262 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features); 263 Ptr += CharSize; 264 } 265 assert(unsigned(OutBuf-Buffer) != Tok.getLength() && 266 "NeedsCleaning flag set on something that didn't need cleaning!"); 267 268 return OutBuf-Buffer; 269} 270 271/// CreateString - Plop the specified string into a scratch buffer and return a 272/// location for it. If specified, the source location provides a source 273/// location for the token. 274void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok, 275 SourceLocation InstantiationLoc) { 276 Tok.setLength(Len); 277 278 const char *DestPtr; 279 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr); 280 281 if (InstantiationLoc.isValid()) 282 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc, 283 InstantiationLoc, Len); 284 Tok.setLocation(Loc); 285 286 // If this is a literal token, set the pointer data. 287 if (Tok.isLiteral()) 288 Tok.setLiteralData(DestPtr); 289} 290 291 292/// AdvanceToTokenCharacter - Given a location that specifies the start of a 293/// token, return a new location that specifies a character within the token. 294SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart, 295 unsigned CharNo) { 296 // Figure out how many physical characters away the specified instantiation 297 // character is. This needs to take into consideration newlines and 298 // trigraphs. 299 const char *TokPtr = SourceMgr.getCharacterData(TokStart); 300 301 // If they request the first char of the token, we're trivially done. 302 if (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)) 303 return TokStart; 304 305 unsigned PhysOffset = 0; 306 307 // The usual case is that tokens don't contain anything interesting. Skip 308 // over the uninteresting characters. If a token only consists of simple 309 // chars, this method is extremely fast. 310 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) { 311 if (CharNo == 0) 312 return TokStart.getFileLocWithOffset(PhysOffset); 313 ++TokPtr, --CharNo, ++PhysOffset; 314 } 315 316 // If we have a character that may be a trigraph or escaped newline, use a 317 // lexer to parse it correctly. 318 for (; CharNo; --CharNo) { 319 unsigned Size; 320 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features); 321 TokPtr += Size; 322 PhysOffset += Size; 323 } 324 325 // Final detail: if we end up on an escaped newline, we want to return the 326 // location of the actual byte of the token. For example foo\<newline>bar 327 // advanced by 3 should return the location of b, not of \\. One compounding 328 // detail of this is that the escape may be made by a trigraph. 329 if (!Lexer::isObviouslySimpleCharacter(*TokPtr)) 330 PhysOffset = Lexer::SkipEscapedNewLines(TokPtr)-TokPtr; 331 332 return TokStart.getFileLocWithOffset(PhysOffset); 333} 334 335/// \brief Computes the source location just past the end of the 336/// token at this source location. 337/// 338/// This routine can be used to produce a source location that 339/// points just past the end of the token referenced by \p Loc, and 340/// is generally used when a diagnostic needs to point just after a 341/// token where it expected something different that it received. If 342/// the returned source location would not be meaningful (e.g., if 343/// it points into a macro), this routine returns an invalid 344/// source location. 345SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc) { 346 if (Loc.isInvalid() || !Loc.isFileID()) 347 return SourceLocation(); 348 349 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features); 350 return AdvanceToTokenCharacter(Loc, Len); 351} 352 353 354 355//===----------------------------------------------------------------------===// 356// Preprocessor Initialization Methods 357//===----------------------------------------------------------------------===// 358 359 360/// EnterMainSourceFile - Enter the specified FileID as the main source file, 361/// which implicitly adds the builtin defines etc. 362void Preprocessor::EnterMainSourceFile() { 363 // We do not allow the preprocessor to reenter the main file. Doing so will 364 // cause FileID's to accumulate information from both runs (e.g. #line 365 // information) and predefined macros aren't guaranteed to be set properly. 366 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!"); 367 FileID MainFileID = SourceMgr.getMainFileID(); 368 369 // Enter the main file source buffer. 370 EnterSourceFile(MainFileID, 0); 371 372 if (IsMainFileEofCodeCompletion) { 373 // Tell our newly-created lexer that it should treat its end-of-file as 374 // a code-completion token. 375 IsMainFileEofCodeCompletion = false; 376 static_cast<Lexer *>(getCurrentFileLexer())->SetEofIsCodeCompletion(); 377 } 378 379 // Tell the header info that the main file was entered. If the file is later 380 // #imported, it won't be re-entered. 381 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID)) 382 HeaderInfo.IncrementIncludeCount(FE); 383 384 std::vector<char> PrologFile; 385 PrologFile.reserve(4080); 386 387 // FIXME: Don't make a copy. 388 PrologFile.insert(PrologFile.end(), Predefines.begin(), Predefines.end()); 389 390 // Memory buffer must end with a null byte! 391 PrologFile.push_back(0); 392 393 // Now that we have emitted the predefined macros, #includes, etc into 394 // PrologFile, preprocess it to populate the initial preprocessor state. 395 llvm::MemoryBuffer *SB = 396 llvm::MemoryBuffer::getMemBufferCopy(&PrologFile.front(),&PrologFile.back(), 397 "<built-in>"); 398 assert(SB && "Cannot fail to create predefined source buffer"); 399 FileID FID = SourceMgr.createFileIDForMemBuffer(SB); 400 assert(!FID.isInvalid() && "Could not create FileID for predefines?"); 401 402 // Start parsing the predefines. 403 EnterSourceFile(FID, 0); 404} 405 406 407//===----------------------------------------------------------------------===// 408// Lexer Event Handling. 409//===----------------------------------------------------------------------===// 410 411/// LookUpIdentifierInfo - Given a tok::identifier token, look up the 412/// identifier information for the token and install it into the token. 413IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier, 414 const char *BufPtr) { 415 assert(Identifier.is(tok::identifier) && "Not an identifier!"); 416 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!"); 417 418 // Look up this token, see if it is a macro, or if it is a language keyword. 419 IdentifierInfo *II; 420 if (BufPtr && !Identifier.needsCleaning()) { 421 // No cleaning needed, just use the characters from the lexed buffer. 422 II = getIdentifierInfo(BufPtr, BufPtr+Identifier.getLength()); 423 } else { 424 // Cleaning needed, alloca a buffer, clean into it, then use the buffer. 425 llvm::SmallVector<char, 64> IdentifierBuffer; 426 IdentifierBuffer.resize(Identifier.getLength()); 427 const char *TmpBuf = &IdentifierBuffer[0]; 428 unsigned Size = getSpelling(Identifier, TmpBuf); 429 II = getIdentifierInfo(TmpBuf, TmpBuf+Size); 430 } 431 Identifier.setIdentifierInfo(II); 432 return II; 433} 434 435 436/// HandleIdentifier - This callback is invoked when the lexer reads an 437/// identifier. This callback looks up the identifier in the map and/or 438/// potentially macro expands it or turns it into a named token (like 'for'). 439/// 440/// Note that callers of this method are guarded by checking the 441/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the 442/// IdentifierInfo methods that compute these properties will need to change to 443/// match. 444void Preprocessor::HandleIdentifier(Token &Identifier) { 445 assert(Identifier.getIdentifierInfo() && 446 "Can't handle identifiers without identifier info!"); 447 448 IdentifierInfo &II = *Identifier.getIdentifierInfo(); 449 450 // If this identifier was poisoned, and if it was not produced from a macro 451 // expansion, emit an error. 452 if (II.isPoisoned() && CurPPLexer) { 453 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning. 454 Diag(Identifier, diag::err_pp_used_poisoned_id); 455 else 456 Diag(Identifier, diag::ext_pp_bad_vaargs_use); 457 } 458 459 // If this is a macro to be expanded, do it. 460 if (MacroInfo *MI = getMacroInfo(&II)) { 461 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) { 462 if (MI->isEnabled()) { 463 if (!HandleMacroExpandedIdentifier(Identifier, MI)) 464 return; 465 } else { 466 // C99 6.10.3.4p2 says that a disabled macro may never again be 467 // expanded, even if it's in a context where it could be expanded in the 468 // future. 469 Identifier.setFlag(Token::DisableExpand); 470 } 471 } 472 } 473 474 // C++ 2.11p2: If this is an alternative representation of a C++ operator, 475 // then we act as if it is the actual operator and not the textual 476 // representation of it. 477 if (II.isCPlusPlusOperatorKeyword()) 478 Identifier.setIdentifierInfo(0); 479 480 // If this is an extension token, diagnose its use. 481 // We avoid diagnosing tokens that originate from macro definitions. 482 // FIXME: This warning is disabled in cases where it shouldn't be, 483 // like "#define TY typeof", "TY(1) x". 484 if (II.isExtensionToken() && !DisableMacroExpansion) 485 Diag(Identifier, diag::ext_token_used); 486} 487 488void Preprocessor::AddCommentHandler(CommentHandler *Handler) { 489 assert(Handler && "NULL comment handler"); 490 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) == 491 CommentHandlers.end() && "Comment handler already registered"); 492 CommentHandlers.push_back(Handler); 493} 494 495void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) { 496 std::vector<CommentHandler *>::iterator Pos 497 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler); 498 assert(Pos != CommentHandlers.end() && "Comment handler not registered"); 499 CommentHandlers.erase(Pos); 500} 501 502void Preprocessor::HandleComment(SourceRange Comment) { 503 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(), 504 HEnd = CommentHandlers.end(); 505 H != HEnd; ++H) 506 (*H)->HandleComment(*this, Comment); 507} 508 509CommentHandler::~CommentHandler() { } 510