Preprocessor.cpp revision 531cc8355608295caa01f31fa4e078d77a3d6470
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/Streams.h" 41#include <cstdio> 42using namespace clang; 43 44//===----------------------------------------------------------------------===// 45 46PreprocessorFactory::~PreprocessorFactory() {} 47 48bool PreprocessorFactory::FinishInitialization(Preprocessor *PP, bool UsesPCH) { 49 return false; 50} 51 52Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts, 53 TargetInfo &target, SourceManager &SM, 54 HeaderSearch &Headers, 55 IdentifierInfoLookup* IILookup) 56 : Diags(&diags), Features(opts), Target(target),FileMgr(Headers.getFileMgr()), 57 SourceMgr(SM), HeaderInfo(Headers), Identifiers(opts, IILookup), 58 CurPPLexer(0), CurDirLookup(0), Callbacks(0) { 59 ScratchBuf = new ScratchBuffer(SourceMgr); 60 61 // Clear stats. 62 NumDirectives = NumDefined = NumUndefined = NumPragma = 0; 63 NumIf = NumElse = NumEndif = 0; 64 NumEnteredSourceFiles = 0; 65 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0; 66 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0; 67 MaxIncludeStackDepth = 0; 68 NumSkipped = 0; 69 70 // Default to discarding comments. 71 KeepComments = false; 72 KeepMacroComments = false; 73 74 // Macro expansion is enabled. 75 DisableMacroExpansion = false; 76 InMacroArgs = false; 77 NumCachedTokenLexers = 0; 78 79 CachedLexPos = 0; 80 81 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 82 // This gets unpoisoned where it is allowed. 83 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned(); 84 85 // Initialize the pragma handlers. 86 PragmaHandlers = new PragmaNamespace(0); 87 RegisterBuiltinPragmas(); 88 89 // Initialize builtin macros like __LINE__ and friends. 90 RegisterBuiltinMacros(); 91} 92 93Preprocessor::~Preprocessor() { 94 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!"); 95 96 while (!IncludeMacroStack.empty()) { 97 delete IncludeMacroStack.back().TheLexer; 98 delete IncludeMacroStack.back().TheTokenLexer; 99 IncludeMacroStack.pop_back(); 100 } 101 102 // Free any macro definitions. 103 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I = 104 Macros.begin(), E = Macros.end(); I != E; ++I) { 105 // We don't need to free the MacroInfo objects directly. These 106 // will be released when the BumpPtrAllocator 'BP' object gets 107 // destroyed. We still need to run the dstor, however, to free 108 // memory alocated by MacroInfo. 109 I->second->Destroy(BP); 110 I->first->setHasMacroDefinition(false); 111 } 112 113 // Free any cached macro expanders. 114 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i) 115 delete TokenLexerCache[i]; 116 117 // Release pragma information. 118 delete PragmaHandlers; 119 120 // Delete the scratch buffer info. 121 delete ScratchBuf; 122 123 delete Callbacks; 124} 125 126void Preprocessor::setPTHManager(PTHManager* pm) { 127 PTH.reset(pm); 128 FileMgr.setStatCache(PTH->createStatCache()); 129} 130 131void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const { 132 llvm::cerr << tok::getTokenName(Tok.getKind()) << " '" 133 << getSpelling(Tok) << "'"; 134 135 if (!DumpFlags) return; 136 137 llvm::cerr << "\t"; 138 if (Tok.isAtStartOfLine()) 139 llvm::cerr << " [StartOfLine]"; 140 if (Tok.hasLeadingSpace()) 141 llvm::cerr << " [LeadingSpace]"; 142 if (Tok.isExpandDisabled()) 143 llvm::cerr << " [ExpandDisabled]"; 144 if (Tok.needsCleaning()) { 145 const char *Start = SourceMgr.getCharacterData(Tok.getLocation()); 146 llvm::cerr << " [UnClean='" << std::string(Start, Start+Tok.getLength()) 147 << "']"; 148 } 149 150 llvm::cerr << "\tLoc=<"; 151 DumpLocation(Tok.getLocation()); 152 llvm::cerr << ">"; 153} 154 155void Preprocessor::DumpLocation(SourceLocation Loc) const { 156 Loc.dump(SourceMgr); 157} 158 159void Preprocessor::DumpMacro(const MacroInfo &MI) const { 160 llvm::cerr << "MACRO: "; 161 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) { 162 DumpToken(MI.getReplacementToken(i)); 163 llvm::cerr << " "; 164 } 165 llvm::cerr << "\n"; 166} 167 168void Preprocessor::PrintStats() { 169 llvm::cerr << "\n*** Preprocessor Stats:\n"; 170 llvm::cerr << NumDirectives << " directives found:\n"; 171 llvm::cerr << " " << NumDefined << " #define.\n"; 172 llvm::cerr << " " << NumUndefined << " #undef.\n"; 173 llvm::cerr << " #include/#include_next/#import:\n"; 174 llvm::cerr << " " << NumEnteredSourceFiles << " source files entered.\n"; 175 llvm::cerr << " " << MaxIncludeStackDepth << " max include stack depth\n"; 176 llvm::cerr << " " << NumIf << " #if/#ifndef/#ifdef.\n"; 177 llvm::cerr << " " << NumElse << " #else/#elif.\n"; 178 llvm::cerr << " " << NumEndif << " #endif.\n"; 179 llvm::cerr << " " << NumPragma << " #pragma.\n"; 180 llvm::cerr << NumSkipped << " #if/#ifndef#ifdef regions skipped\n"; 181 182 llvm::cerr << NumMacroExpanded << "/" << NumFnMacroExpanded << "/" 183 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, " 184 << NumFastMacroExpanded << " on the fast path.\n"; 185 llvm::cerr << (NumFastTokenPaste+NumTokenPaste) 186 << " token paste (##) operations performed, " 187 << NumFastTokenPaste << " on the fast path.\n"; 188} 189 190//===----------------------------------------------------------------------===// 191// Token Spelling 192//===----------------------------------------------------------------------===// 193 194 195/// getSpelling() - Return the 'spelling' of this token. The spelling of a 196/// token are the characters used to represent the token in the source file 197/// after trigraph expansion and escaped-newline folding. In particular, this 198/// wants to get the true, uncanonicalized, spelling of things like digraphs 199/// UCNs, etc. 200std::string Preprocessor::getSpelling(const Token &Tok) const { 201 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 202 203 // If this token contains nothing interesting, return it directly. 204 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 205 if (!Tok.needsCleaning()) 206 return std::string(TokStart, TokStart+Tok.getLength()); 207 208 std::string Result; 209 Result.reserve(Tok.getLength()); 210 211 // Otherwise, hard case, relex the characters into the string. 212 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 213 Ptr != End; ) { 214 unsigned CharSize; 215 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features)); 216 Ptr += CharSize; 217 } 218 assert(Result.size() != unsigned(Tok.getLength()) && 219 "NeedsCleaning flag set on something that didn't need cleaning!"); 220 return Result; 221} 222 223/// getSpelling - This method is used to get the spelling of a token into a 224/// preallocated buffer, instead of as an std::string. The caller is required 225/// to allocate enough space for the token, which is guaranteed to be at least 226/// Tok.getLength() bytes long. The actual length of the token is returned. 227/// 228/// Note that this method may do two possible things: it may either fill in 229/// the buffer specified with characters, or it may *change the input pointer* 230/// to point to a constant buffer with the data already in it (avoiding a 231/// copy). The caller is not allowed to modify the returned buffer pointer 232/// if an internal buffer is returned. 233unsigned Preprocessor::getSpelling(const Token &Tok, 234 const char *&Buffer) const { 235 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 236 237 // If this token is an identifier, just return the string from the identifier 238 // table, which is very quick. 239 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) { 240 Buffer = II->getName(); 241 return II->getLength(); 242 } 243 244 // Otherwise, compute the start of the token in the input lexer buffer. 245 const char *TokStart = 0; 246 247 if (Tok.isLiteral()) 248 TokStart = Tok.getLiteralData(); 249 250 if (TokStart == 0) 251 TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 252 253 // If this token contains nothing interesting, return it directly. 254 if (!Tok.needsCleaning()) { 255 Buffer = TokStart; 256 return Tok.getLength(); 257 } 258 259 // Otherwise, hard case, relex the characters into the string. 260 char *OutBuf = const_cast<char*>(Buffer); 261 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 262 Ptr != End; ) { 263 unsigned CharSize; 264 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features); 265 Ptr += CharSize; 266 } 267 assert(unsigned(OutBuf-Buffer) != Tok.getLength() && 268 "NeedsCleaning flag set on something that didn't need cleaning!"); 269 270 return OutBuf-Buffer; 271} 272 273/// CreateString - Plop the specified string into a scratch buffer and return a 274/// location for it. If specified, the source location provides a source 275/// location for the token. 276void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok, 277 SourceLocation InstantiationLoc) { 278 Tok.setLength(Len); 279 280 const char *DestPtr; 281 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr); 282 283 if (InstantiationLoc.isValid()) 284 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc, 285 InstantiationLoc, Len); 286 Tok.setLocation(Loc); 287 288 // If this is a literal token, set the pointer data. 289 if (Tok.isLiteral()) 290 Tok.setLiteralData(DestPtr); 291} 292 293 294/// AdvanceToTokenCharacter - Given a location that specifies the start of a 295/// token, return a new location that specifies a character within the token. 296SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart, 297 unsigned CharNo) { 298 // If they request the first char of the token, we're trivially done. 299 if (CharNo == 0) return TokStart; 300 301 // Figure out how many physical characters away the specified instantiation 302 // character is. This needs to take into consideration newlines and 303 // trigraphs. 304 const char *TokPtr = SourceMgr.getCharacterData(TokStart); 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 (CharNo && Lexer::isObviouslySimpleCharacter(*TokPtr)) 311 ++TokPtr, --CharNo, ++PhysOffset; 312 313 // If we have a character that may be a trigraph or escaped newline, use a 314 // lexer to parse it correctly. 315 if (CharNo != 0) { 316 // Skip over characters the remaining characters. 317 for (; CharNo; --CharNo) { 318 unsigned Size; 319 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features); 320 TokPtr += Size; 321 PhysOffset += Size; 322 } 323 } 324 325 return TokStart.getFileLocWithOffset(PhysOffset); 326} 327 328/// \brief Computes the source location just past the end of the 329/// token at this source location. 330/// 331/// This routine can be used to produce a source location that 332/// points just past the end of the token referenced by \p Loc, and 333/// is generally used when a diagnostic needs to point just after a 334/// token where it expected something different that it received. If 335/// the returned source location would not be meaningful (e.g., if 336/// it points into a macro), this routine returns an invalid 337/// source location. 338SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc) { 339 if (Loc.isInvalid() || !Loc.isFileID()) 340 return SourceLocation(); 341 342 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager()); 343 return AdvanceToTokenCharacter(Loc, Len); 344} 345 346 347 348//===----------------------------------------------------------------------===// 349// Preprocessor Initialization Methods 350//===----------------------------------------------------------------------===// 351 352 353/// EnterMainSourceFile - Enter the specified FileID as the main source file, 354/// which implicitly adds the builtin defines etc. 355void Preprocessor::EnterMainSourceFile() { 356 // We do not allow the preprocessor to reenter the main file. Doing so will 357 // cause FileID's to accumulate information from both runs (e.g. #line 358 // information) and predefined macros aren't guaranteed to be set properly. 359 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!"); 360 FileID MainFileID = SourceMgr.getMainFileID(); 361 362 // Enter the main file source buffer. 363 EnterSourceFile(MainFileID, 0); 364 365 // Tell the header info that the main file was entered. If the file is later 366 // #imported, it won't be re-entered. 367 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID)) 368 HeaderInfo.IncrementIncludeCount(FE); 369 370 std::vector<char> PrologFile; 371 PrologFile.reserve(4080); 372 373 // FIXME: Don't make a copy. 374 PrologFile.insert(PrologFile.end(), Predefines.begin(), Predefines.end()); 375 376 // Memory buffer must end with a null byte! 377 PrologFile.push_back(0); 378 379 // Now that we have emitted the predefined macros, #includes, etc into 380 // PrologFile, preprocess it to populate the initial preprocessor state. 381 llvm::MemoryBuffer *SB = 382 llvm::MemoryBuffer::getMemBufferCopy(&PrologFile.front(),&PrologFile.back(), 383 "<built-in>"); 384 assert(SB && "Cannot fail to create predefined source buffer"); 385 FileID FID = SourceMgr.createFileIDForMemBuffer(SB); 386 assert(!FID.isInvalid() && "Could not create FileID for predefines?"); 387 388 // Start parsing the predefines. 389 EnterSourceFile(FID, 0); 390} 391 392 393//===----------------------------------------------------------------------===// 394// Lexer Event Handling. 395//===----------------------------------------------------------------------===// 396 397/// LookUpIdentifierInfo - Given a tok::identifier token, look up the 398/// identifier information for the token and install it into the token. 399IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier, 400 const char *BufPtr) { 401 assert(Identifier.is(tok::identifier) && "Not an identifier!"); 402 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!"); 403 404 // Look up this token, see if it is a macro, or if it is a language keyword. 405 IdentifierInfo *II; 406 if (BufPtr && !Identifier.needsCleaning()) { 407 // No cleaning needed, just use the characters from the lexed buffer. 408 II = getIdentifierInfo(BufPtr, BufPtr+Identifier.getLength()); 409 } else { 410 // Cleaning needed, alloca a buffer, clean into it, then use the buffer. 411 llvm::SmallVector<char, 64> IdentifierBuffer; 412 IdentifierBuffer.resize(Identifier.getLength()); 413 const char *TmpBuf = &IdentifierBuffer[0]; 414 unsigned Size = getSpelling(Identifier, TmpBuf); 415 II = getIdentifierInfo(TmpBuf, TmpBuf+Size); 416 } 417 Identifier.setIdentifierInfo(II); 418 return II; 419} 420 421 422/// HandleIdentifier - This callback is invoked when the lexer reads an 423/// identifier. This callback looks up the identifier in the map and/or 424/// potentially macro expands it or turns it into a named token (like 'for'). 425/// 426/// Note that callers of this method are guarded by checking the 427/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the 428/// IdentifierInfo methods that compute these properties will need to change to 429/// match. 430void Preprocessor::HandleIdentifier(Token &Identifier) { 431 assert(Identifier.getIdentifierInfo() && 432 "Can't handle identifiers without identifier info!"); 433 434 IdentifierInfo &II = *Identifier.getIdentifierInfo(); 435 436 // If this identifier was poisoned, and if it was not produced from a macro 437 // expansion, emit an error. 438 if (II.isPoisoned() && CurPPLexer) { 439 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning. 440 Diag(Identifier, diag::err_pp_used_poisoned_id); 441 else 442 Diag(Identifier, diag::ext_pp_bad_vaargs_use); 443 } 444 445 // If this is a macro to be expanded, do it. 446 if (MacroInfo *MI = getMacroInfo(&II)) { 447 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) { 448 if (MI->isEnabled()) { 449 if (!HandleMacroExpandedIdentifier(Identifier, MI)) 450 return; 451 } else { 452 // C99 6.10.3.4p2 says that a disabled macro may never again be 453 // expanded, even if it's in a context where it could be expanded in the 454 // future. 455 Identifier.setFlag(Token::DisableExpand); 456 } 457 } 458 } 459 460 // C++ 2.11p2: If this is an alternative representation of a C++ operator, 461 // then we act as if it is the actual operator and not the textual 462 // representation of it. 463 if (II.isCPlusPlusOperatorKeyword()) 464 Identifier.setIdentifierInfo(0); 465 466 // If this is an extension token, diagnose its use. 467 // We avoid diagnosing tokens that originate from macro definitions. 468 if (II.isExtensionToken() && Features.C99 && !DisableMacroExpansion) 469 Diag(Identifier, diag::ext_token_used); 470} 471