Preprocessor.h revision 7412494982c8b50c90961302c3a718633b2c3ab7
1//===--- Preprocessor.h - C Language Family Preprocessor --------*- C++ -*-===// 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 defines the Preprocessor interface. 11// 12//===----------------------------------------------------------------------===// 13 14#ifndef LLVM_CLANG_LEX_PREPROCESSOR_H 15#define LLVM_CLANG_LEX_PREPROCESSOR_H 16 17#include "clang/Lex/MacroInfo.h" 18#include "clang/Lex/Lexer.h" 19#include "clang/Lex/PTHLexer.h" 20#include "clang/Lex/PPCallbacks.h" 21#include "clang/Lex/TokenLexer.h" 22#include "clang/Lex/PTHManager.h" 23#include "clang/Basic/Builtins.h" 24#include "clang/Basic/Diagnostic.h" 25#include "clang/Basic/IdentifierTable.h" 26#include "clang/Basic/SourceLocation.h" 27#include "llvm/ADT/DenseMap.h" 28#include "llvm/ADT/IntrusiveRefCntPtr.h" 29#include "llvm/ADT/SmallPtrSet.h" 30#include "llvm/ADT/OwningPtr.h" 31#include "llvm/ADT/SmallVector.h" 32#include "llvm/Support/Allocator.h" 33#include <vector> 34 35namespace clang { 36 37class SourceManager; 38class ExternalPreprocessorSource; 39class FileManager; 40class FileEntry; 41class HeaderSearch; 42class PragmaNamespace; 43class PragmaHandler; 44class CommentHandler; 45class ScratchBuffer; 46class TargetInfo; 47class PPCallbacks; 48class CodeCompletionHandler; 49class DirectoryLookup; 50class PreprocessingRecord; 51 52/// Preprocessor - This object engages in a tight little dance with the lexer to 53/// efficiently preprocess tokens. Lexers know only about tokens within a 54/// single source file, and don't know anything about preprocessor-level issues 55/// like the #include stack, token expansion, etc. 56/// 57class Preprocessor : public llvm::RefCountedBase<Preprocessor> { 58 Diagnostic *Diags; 59 LangOptions Features; 60 const TargetInfo &Target; 61 FileManager &FileMgr; 62 SourceManager &SourceMgr; 63 ScratchBuffer *ScratchBuf; 64 HeaderSearch &HeaderInfo; 65 66 /// \brief External source of macros. 67 ExternalPreprocessorSource *ExternalSource; 68 69 /// PTH - An optional PTHManager object used for getting tokens from 70 /// a token cache rather than lexing the original source file. 71 llvm::OwningPtr<PTHManager> PTH; 72 73 /// BP - A BumpPtrAllocator object used to quickly allocate and release 74 /// objects internal to the Preprocessor. 75 llvm::BumpPtrAllocator BP; 76 77 /// Identifiers for builtin macros and other builtins. 78 IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__ 79 IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__ 80 IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__ 81 IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__ 82 IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__ 83 IdentifierInfo *Ident__COUNTER__; // __COUNTER__ 84 IdentifierInfo *Ident_Pragma, *Ident__pragma; // _Pragma, __pragma 85 IdentifierInfo *Ident__VA_ARGS__; // __VA_ARGS__ 86 IdentifierInfo *Ident__has_feature; // __has_feature 87 IdentifierInfo *Ident__has_builtin; // __has_builtin 88 IdentifierInfo *Ident__has_attribute; // __has_attribute 89 IdentifierInfo *Ident__has_include; // __has_include 90 IdentifierInfo *Ident__has_include_next; // __has_include_next 91 92 SourceLocation DATELoc, TIMELoc; 93 unsigned CounterValue; // Next __COUNTER__ value. 94 95 enum { 96 /// MaxIncludeStackDepth - Maximum depth of #includes. 97 MaxAllowedIncludeStackDepth = 200 98 }; 99 100 // State that is set before the preprocessor begins. 101 bool KeepComments : 1; 102 bool KeepMacroComments : 1; 103 104 // State that changes while the preprocessor runs: 105 bool InMacroArgs : 1; // True if parsing fn macro invocation args. 106 107 /// Whether the preprocessor owns the header search object. 108 bool OwnsHeaderSearch : 1; 109 110 /// DisableMacroExpansion - True if macro expansion is disabled. 111 bool DisableMacroExpansion : 1; 112 113 /// \brief Whether we have already loaded macros from the external source. 114 mutable bool ReadMacrosFromExternalSource : 1; 115 116 /// Identifiers - This is mapping/lookup information for all identifiers in 117 /// the program, including program keywords. 118 mutable IdentifierTable Identifiers; 119 120 /// Selectors - This table contains all the selectors in the program. Unlike 121 /// IdentifierTable above, this table *isn't* populated by the preprocessor. 122 /// It is declared/instantiated here because it's role/lifetime is 123 /// conceptually similar the IdentifierTable. In addition, the current control 124 /// flow (in clang::ParseAST()), make it convenient to put here. 125 /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to 126 /// the lifetime of the preprocessor. 127 SelectorTable Selectors; 128 129 /// BuiltinInfo - Information about builtins. 130 Builtin::Context BuiltinInfo; 131 132 /// PragmaHandlers - This tracks all of the pragmas that the client registered 133 /// with this preprocessor. 134 PragmaNamespace *PragmaHandlers; 135 136 /// \brief Tracks all of the comment handlers that the client registered 137 /// with this preprocessor. 138 std::vector<CommentHandler *> CommentHandlers; 139 140 /// \brief The code-completion handler. 141 CodeCompletionHandler *CodeComplete; 142 143 /// \brief The file that we're performing code-completion for, if any. 144 const FileEntry *CodeCompletionFile; 145 146 /// \brief The number of bytes that we will initially skip when entering the 147 /// main file, which is used when loading a precompiled preamble, along 148 /// with a flag that indicates whether skipping this number of bytes will 149 /// place the lexer at the start of a line. 150 std::pair<unsigned, bool> SkipMainFilePreamble; 151 152 /// CurLexer - This is the current top of the stack that we're lexing from if 153 /// not expanding a macro and we are lexing directly from source code. 154 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 155 llvm::OwningPtr<Lexer> CurLexer; 156 157 /// CurPTHLexer - This is the current top of stack that we're lexing from if 158 /// not expanding from a macro and we are lexing from a PTH cache. 159 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 160 llvm::OwningPtr<PTHLexer> CurPTHLexer; 161 162 /// CurPPLexer - This is the current top of the stack what we're lexing from 163 /// if not expanding a macro. This is an alias for either CurLexer or 164 /// CurPTHLexer. 165 PreprocessorLexer *CurPPLexer; 166 167 /// CurLookup - The DirectoryLookup structure used to find the current 168 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 169 /// implement #include_next and find directory-specific properties. 170 const DirectoryLookup *CurDirLookup; 171 172 /// CurTokenLexer - This is the current macro we are expanding, if we are 173 /// expanding a macro. One of CurLexer and CurTokenLexer must be null. 174 llvm::OwningPtr<TokenLexer> CurTokenLexer; 175 176 /// IncludeMacroStack - This keeps track of the stack of files currently 177 /// #included, and macros currently being expanded from, not counting 178 /// CurLexer/CurTokenLexer. 179 struct IncludeStackInfo { 180 Lexer *TheLexer; 181 PTHLexer *ThePTHLexer; 182 PreprocessorLexer *ThePPLexer; 183 TokenLexer *TheTokenLexer; 184 const DirectoryLookup *TheDirLookup; 185 186 IncludeStackInfo(Lexer *L, PTHLexer* P, PreprocessorLexer* PPL, 187 TokenLexer* TL, const DirectoryLookup *D) 188 : TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL), TheTokenLexer(TL), 189 TheDirLookup(D) {} 190 }; 191 std::vector<IncludeStackInfo> IncludeMacroStack; 192 193 /// Callbacks - These are actions invoked when some preprocessor activity is 194 /// encountered (e.g. a file is #included, etc). 195 PPCallbacks *Callbacks; 196 197 /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping 198 /// to the actual definition of the macro. 199 llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros; 200 201 /// \brief Macros that we want to warn because they are not used at the end 202 /// of the translation unit; we store just their SourceLocations instead 203 /// something like MacroInfo*. The benefit of this is that when we are 204 /// deserializing from PCH, we don't need to deserialize identifier & macros 205 /// just so that we can report that they are unused, we just warn using 206 /// the SourceLocations of this set (that will be filled by the ASTReader). 207 /// We are using SmallPtrSet instead of a vector for faster removal. 208 typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy; 209 WarnUnusedMacroLocsTy WarnUnusedMacroLocs; 210 211 /// MacroArgCache - This is a "freelist" of MacroArg objects that can be 212 /// reused for quick allocation. 213 MacroArgs *MacroArgCache; 214 friend class MacroArgs; 215 216 /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma 217 /// push_macro directive, we keep a MacroInfo stack used to restore 218 /// previous macro value. 219 llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo; 220 221 // Various statistics we track for performance analysis. 222 unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma; 223 unsigned NumIf, NumElse, NumEndif; 224 unsigned NumEnteredSourceFiles, MaxIncludeStackDepth; 225 unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded; 226 unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste; 227 unsigned NumSkipped; 228 229 /// Predefines - This string is the predefined macros that preprocessor 230 /// should use from the command line etc. 231 std::string Predefines; 232 233 /// TokenLexerCache - Cache macro expanders to reduce malloc traffic. 234 enum { TokenLexerCacheSize = 8 }; 235 unsigned NumCachedTokenLexers; 236 TokenLexer *TokenLexerCache[TokenLexerCacheSize]; 237 238 /// \brief A record of the macro definitions and instantiations that 239 /// occurred during preprocessing. 240 /// 241 /// This is an optional side structure that can be enabled with 242 /// \c createPreprocessingRecord() prior to preprocessing. 243 PreprocessingRecord *Record; 244 245private: // Cached tokens state. 246 typedef llvm::SmallVector<Token, 1> CachedTokensTy; 247 248 /// CachedTokens - Cached tokens are stored here when we do backtracking or 249 /// lookahead. They are "lexed" by the CachingLex() method. 250 CachedTokensTy CachedTokens; 251 252 /// CachedLexPos - The position of the cached token that CachingLex() should 253 /// "lex" next. If it points beyond the CachedTokens vector, it means that 254 /// a normal Lex() should be invoked. 255 CachedTokensTy::size_type CachedLexPos; 256 257 /// BacktrackPositions - Stack of backtrack positions, allowing nested 258 /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to 259 /// indicate where CachedLexPos should be set when the BackTrack() method is 260 /// invoked (at which point the last position is popped). 261 std::vector<CachedTokensTy::size_type> BacktrackPositions; 262 263 struct MacroInfoChain { 264 MacroInfo MI; 265 MacroInfoChain *Next; 266 MacroInfoChain *Prev; 267 }; 268 269 /// MacroInfos are managed as a chain for easy disposal. This is the head 270 /// of that list. 271 MacroInfoChain *MIChainHead; 272 273 /// MICache - A "freelist" of MacroInfo objects that can be reused for quick 274 /// allocation. 275 MacroInfoChain *MICache; 276 277 MacroInfo *getInfoForMacro(IdentifierInfo *II) const; 278 279public: 280 Preprocessor(Diagnostic &diags, const LangOptions &opts, 281 const TargetInfo &target, 282 SourceManager &SM, HeaderSearch &Headers, 283 IdentifierInfoLookup *IILookup = 0, 284 bool OwnsHeaderSearch = false); 285 286 ~Preprocessor(); 287 288 Diagnostic &getDiagnostics() const { return *Diags; } 289 void setDiagnostics(Diagnostic &D) { Diags = &D; } 290 291 const LangOptions &getLangOptions() const { return Features; } 292 const TargetInfo &getTargetInfo() const { return Target; } 293 FileManager &getFileManager() const { return FileMgr; } 294 SourceManager &getSourceManager() const { return SourceMgr; } 295 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; } 296 297 IdentifierTable &getIdentifierTable() { return Identifiers; } 298 SelectorTable &getSelectorTable() { return Selectors; } 299 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; } 300 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; } 301 302 void setPTHManager(PTHManager* pm); 303 304 PTHManager *getPTHManager() { return PTH.get(); } 305 306 void setExternalSource(ExternalPreprocessorSource *Source) { 307 ExternalSource = Source; 308 } 309 310 ExternalPreprocessorSource *getExternalSource() const { 311 return ExternalSource; 312 } 313 314 /// SetCommentRetentionState - Control whether or not the preprocessor retains 315 /// comments in output. 316 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) { 317 this->KeepComments = KeepComments | KeepMacroComments; 318 this->KeepMacroComments = KeepMacroComments; 319 } 320 321 bool getCommentRetentionState() const { return KeepComments; } 322 323 /// isCurrentLexer - Return true if we are lexing directly from the specified 324 /// lexer. 325 bool isCurrentLexer(const PreprocessorLexer *L) const { 326 return CurPPLexer == L; 327 } 328 329 /// getCurrentLexer - Return the current lexer being lexed from. Note 330 /// that this ignores any potentially active macro expansions and _Pragma 331 /// expansions going on at the time. 332 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; } 333 334 /// getCurrentFileLexer - Return the current file lexer being lexed from. 335 /// Note that this ignores any potentially active macro expansions and _Pragma 336 /// expansions going on at the time. 337 PreprocessorLexer *getCurrentFileLexer() const; 338 339 /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks. 340 /// Note that this class takes ownership of any PPCallbacks object given to 341 /// it. 342 PPCallbacks *getPPCallbacks() const { return Callbacks; } 343 void addPPCallbacks(PPCallbacks *C) { 344 if (Callbacks) 345 C = new PPChainedCallbacks(C, Callbacks); 346 Callbacks = C; 347 } 348 349 /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to 350 /// or null if it isn't #define'd. 351 MacroInfo *getMacroInfo(IdentifierInfo *II) const { 352 if (!II->hasMacroDefinition()) 353 return 0; 354 355 return getInfoForMacro(II); 356 } 357 358 /// setMacroInfo - Specify a macro for this identifier. 359 /// 360 void setMacroInfo(IdentifierInfo *II, MacroInfo *MI); 361 362 /// macro_iterator/macro_begin/macro_end - This allows you to walk the current 363 /// state of the macro table. This visits every currently-defined macro. 364 typedef llvm::DenseMap<IdentifierInfo*, 365 MacroInfo*>::const_iterator macro_iterator; 366 macro_iterator macro_begin(bool IncludeExternalMacros = true) const; 367 macro_iterator macro_end(bool IncludeExternalMacros = true) const; 368 369 const std::string &getPredefines() const { return Predefines; } 370 /// setPredefines - Set the predefines for this Preprocessor. These 371 /// predefines are automatically injected when parsing the main file. 372 void setPredefines(const char *P) { Predefines = P; } 373 void setPredefines(const std::string &P) { Predefines = P; } 374 375 /// getIdentifierInfo - Return information about the specified preprocessor 376 /// identifier token. The version of this method that takes two character 377 /// pointers is preferred unless the identifier is already available as a 378 /// string (this avoids allocation and copying of memory to construct an 379 /// std::string). 380 IdentifierInfo *getIdentifierInfo(llvm::StringRef Name) const { 381 return &Identifiers.get(Name); 382 } 383 384 /// AddPragmaHandler - Add the specified pragma handler to the preprocessor. 385 /// If 'Namespace' is non-null, then it is a token required to exist on the 386 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC". 387 void AddPragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 388 void AddPragmaHandler(PragmaHandler *Handler) { 389 AddPragmaHandler(llvm::StringRef(), Handler); 390 } 391 392 /// RemovePragmaHandler - Remove the specific pragma handler from 393 /// the preprocessor. If \arg Namespace is non-null, then it should 394 /// be the namespace that \arg Handler was added to. It is an error 395 /// to remove a handler that has not been registered. 396 void RemovePragmaHandler(llvm::StringRef Namespace, PragmaHandler *Handler); 397 void RemovePragmaHandler(PragmaHandler *Handler) { 398 RemovePragmaHandler(llvm::StringRef(), Handler); 399 } 400 401 /// \brief Add the specified comment handler to the preprocessor. 402 void AddCommentHandler(CommentHandler *Handler); 403 404 /// \brief Remove the specified comment handler. 405 /// 406 /// It is an error to remove a handler that has not been registered. 407 void RemoveCommentHandler(CommentHandler *Handler); 408 409 /// \brief Set the code completion handler to the given object. 410 void setCodeCompletionHandler(CodeCompletionHandler &Handler) { 411 CodeComplete = &Handler; 412 } 413 414 /// \brief Retrieve the current code-completion handler. 415 CodeCompletionHandler *getCodeCompletionHandler() const { 416 return CodeComplete; 417 } 418 419 /// \brief Clear out the code completion handler. 420 void clearCodeCompletionHandler() { 421 CodeComplete = 0; 422 } 423 424 /// \brief Hook used by the lexer to invoke the "natural language" code 425 /// completion point. 426 void CodeCompleteNaturalLanguage(); 427 428 /// \brief Retrieve the preprocessing record, or NULL if there is no 429 /// preprocessing record. 430 PreprocessingRecord *getPreprocessingRecord() const { return Record; } 431 432 /// \brief Create a new preprocessing record, which will keep track of 433 /// all macro expansions, macro definitions, etc. 434 void createPreprocessingRecord(); 435 436 /// EnterMainSourceFile - Enter the specified FileID as the main source file, 437 /// which implicitly adds the builtin defines etc. 438 void EnterMainSourceFile(); 439 440 /// EndSourceFile - Inform the preprocessor callbacks that processing is 441 /// complete. 442 void EndSourceFile(); 443 444 /// EnterSourceFile - Add a source file to the top of the include stack and 445 /// start lexing tokens from it instead of the current buffer. Emit an error 446 /// and don't enter the file on error. 447 void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, 448 SourceLocation Loc); 449 450 /// EnterMacro - Add a Macro to the top of the include stack and start lexing 451 /// tokens from it instead of the current buffer. Args specifies the 452 /// tokens input to a function-like macro. 453 /// 454 /// ILEnd specifies the location of the ')' for a function-like macro or the 455 /// identifier for an object-like macro. 456 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args); 457 458 /// EnterTokenStream - Add a "macro" context to the top of the include stack, 459 /// which will cause the lexer to start returning the specified tokens. 460 /// 461 /// If DisableMacroExpansion is true, tokens lexed from the token stream will 462 /// not be subject to further macro expansion. Otherwise, these tokens will 463 /// be re-macro-expanded when/if expansion is enabled. 464 /// 465 /// If OwnsTokens is false, this method assumes that the specified stream of 466 /// tokens has a permanent owner somewhere, so they do not need to be copied. 467 /// If it is true, it assumes the array of tokens is allocated with new[] and 468 /// must be freed. 469 /// 470 void EnterTokenStream(const Token *Toks, unsigned NumToks, 471 bool DisableMacroExpansion, bool OwnsTokens); 472 473 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 474 /// lexer stack. This should only be used in situations where the current 475 /// state of the top-of-stack lexer is known. 476 void RemoveTopOfLexerStack(); 477 478 /// EnableBacktrackAtThisPos - From the point that this method is called, and 479 /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor 480 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will 481 /// make the Preprocessor re-lex the same tokens. 482 /// 483 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can 484 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will 485 /// be combined with the EnableBacktrackAtThisPos calls in reverse order. 486 /// 487 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack 488 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of 489 /// tokens will continue indefinitely. 490 /// 491 void EnableBacktrackAtThisPos(); 492 493 /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call. 494 void CommitBacktrackedTokens(); 495 496 /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since 497 /// EnableBacktrackAtThisPos() was previously called. 498 void Backtrack(); 499 500 /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and 501 /// caching of tokens is on. 502 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); } 503 504 /// Lex - To lex a token from the preprocessor, just pull a token from the 505 /// current lexer or macro object. 506 void Lex(Token &Result) { 507 if (CurLexer) 508 CurLexer->Lex(Result); 509 else if (CurPTHLexer) 510 CurPTHLexer->Lex(Result); 511 else if (CurTokenLexer) 512 CurTokenLexer->Lex(Result); 513 else 514 CachingLex(Result); 515 } 516 517 /// LexNonComment - Lex a token. If it's a comment, keep lexing until we get 518 /// something not a comment. This is useful in -E -C mode where comments 519 /// would foul up preprocessor directive handling. 520 void LexNonComment(Token &Result) { 521 do 522 Lex(Result); 523 while (Result.getKind() == tok::comment); 524 } 525 526 /// LexUnexpandedToken - This is just like Lex, but this disables macro 527 /// expansion of identifier tokens. 528 void LexUnexpandedToken(Token &Result) { 529 // Disable macro expansion. 530 bool OldVal = DisableMacroExpansion; 531 DisableMacroExpansion = true; 532 // Lex the token. 533 Lex(Result); 534 535 // Reenable it. 536 DisableMacroExpansion = OldVal; 537 } 538 539 /// LookAhead - This peeks ahead N tokens and returns that token without 540 /// consuming any tokens. LookAhead(0) returns the next token that would be 541 /// returned by Lex(), LookAhead(1) returns the token after it, etc. This 542 /// returns normal tokens after phase 5. As such, it is equivalent to using 543 /// 'Lex', not 'LexUnexpandedToken'. 544 const Token &LookAhead(unsigned N) { 545 if (CachedLexPos + N < CachedTokens.size()) 546 return CachedTokens[CachedLexPos+N]; 547 else 548 return PeekAhead(N+1); 549 } 550 551 /// RevertCachedTokens - When backtracking is enabled and tokens are cached, 552 /// this allows to revert a specific number of tokens. 553 /// Note that the number of tokens being reverted should be up to the last 554 /// backtrack position, not more. 555 void RevertCachedTokens(unsigned N) { 556 assert(isBacktrackEnabled() && 557 "Should only be called when tokens are cached for backtracking"); 558 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back()) 559 && "Should revert tokens up to the last backtrack position, not more"); 560 assert(signed(CachedLexPos) - signed(N) >= 0 && 561 "Corrupted backtrack positions ?"); 562 CachedLexPos -= N; 563 } 564 565 /// EnterToken - Enters a token in the token stream to be lexed next. If 566 /// BackTrack() is called afterwards, the token will remain at the insertion 567 /// point. 568 void EnterToken(const Token &Tok) { 569 EnterCachingLexMode(); 570 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok); 571 } 572 573 /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching 574 /// tokens (because backtrack is enabled) it should replace the most recent 575 /// cached tokens with the given annotation token. This function has no effect 576 /// if backtracking is not enabled. 577 /// 578 /// Note that the use of this function is just for optimization; so that the 579 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is 580 /// invoked. 581 void AnnotateCachedTokens(const Token &Tok) { 582 assert(Tok.isAnnotation() && "Expected annotation token"); 583 if (CachedLexPos != 0 && isBacktrackEnabled()) 584 AnnotatePreviousCachedTokens(Tok); 585 } 586 587 /// \brief Replace the last token with an annotation token. 588 /// 589 /// Like AnnotateCachedTokens(), this routine replaces an 590 /// already-parsed (and resolved) token with an annotation 591 /// token. However, this routine only replaces the last token with 592 /// the annotation token; it does not affect any other cached 593 /// tokens. This function has no effect if backtracking is not 594 /// enabled. 595 void ReplaceLastTokenWithAnnotation(const Token &Tok) { 596 assert(Tok.isAnnotation() && "Expected annotation token"); 597 if (CachedLexPos != 0 && isBacktrackEnabled()) 598 CachedTokens[CachedLexPos-1] = Tok; 599 } 600 601 /// \brief Specify the point at which code-completion will be performed. 602 /// 603 /// \param File the file in which code completion should occur. If 604 /// this file is included multiple times, code-completion will 605 /// perform completion the first time it is included. If NULL, this 606 /// function clears out the code-completion point. 607 /// 608 /// \param Line the line at which code completion should occur 609 /// (1-based). 610 /// 611 /// \param Column the column at which code completion should occur 612 /// (1-based). 613 /// 614 /// \returns true if an error occurred, false otherwise. 615 bool SetCodeCompletionPoint(const FileEntry *File, 616 unsigned Line, unsigned Column); 617 618 /// \brief Determine if this source location refers into the file 619 /// for which we are performing code completion. 620 bool isCodeCompletionFile(SourceLocation FileLoc) const; 621 622 /// \brief Determine if we are performing code completion. 623 bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; } 624 625 /// \brief Instruct the preprocessor to skip part of the main 626 /// the main source file. 627 /// 628 /// \brief Bytes The number of bytes in the preamble to skip. 629 /// 630 /// \brief StartOfLine Whether skipping these bytes puts the lexer at the 631 /// start of a line. 632 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) { 633 SkipMainFilePreamble.first = Bytes; 634 SkipMainFilePreamble.second = StartOfLine; 635 } 636 637 /// Diag - Forwarding function for diagnostics. This emits a diagnostic at 638 /// the specified Token's location, translating the token's start 639 /// position in the current buffer into a SourcePosition object for rendering. 640 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) { 641 return Diags->Report(Loc, DiagID); 642 } 643 644 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) { 645 return Diags->Report(Tok.getLocation(), DiagID); 646 } 647 648 /// getSpelling() - Return the 'spelling' of the token at the given 649 /// location; does not go up to the spelling location or down to the 650 /// instantiation location. 651 /// 652 /// \param buffer A buffer which will be used only if the token requires 653 /// "cleaning", e.g. if it contains trigraphs or escaped newlines 654 /// \param invalid If non-null, will be set \c true if an error occurs. 655 llvm::StringRef getSpelling(SourceLocation loc, 656 llvm::SmallVectorImpl<char> &buffer, 657 bool *invalid = 0) const { 658 return Lexer::getSpelling(loc, buffer, SourceMgr, Features, invalid); 659 } 660 661 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 662 /// token is the characters used to represent the token in the source file 663 /// after trigraph expansion and escaped-newline folding. In particular, this 664 /// wants to get the true, uncanonicalized, spelling of things like digraphs 665 /// UCNs, etc. 666 /// 667 /// \param Invalid If non-null, will be set \c true if an error occurs. 668 std::string getSpelling(const Token &Tok, bool *Invalid = 0) const { 669 return Lexer::getSpelling(Tok, SourceMgr, Features, Invalid); 670 } 671 672 /// getSpelling - This method is used to get the spelling of a token into a 673 /// preallocated buffer, instead of as an std::string. The caller is required 674 /// to allocate enough space for the token, which is guaranteed to be at least 675 /// Tok.getLength() bytes long. The length of the actual result is returned. 676 /// 677 /// Note that this method may do two possible things: it may either fill in 678 /// the buffer specified with characters, or it may *change the input pointer* 679 /// to point to a constant buffer with the data already in it (avoiding a 680 /// copy). The caller is not allowed to modify the returned buffer pointer 681 /// if an internal buffer is returned. 682 unsigned getSpelling(const Token &Tok, const char *&Buffer, 683 bool *Invalid = 0) const { 684 return Lexer::getSpelling(Tok, Buffer, SourceMgr, Features, Invalid); 685 } 686 687 /// getSpelling - This method is used to get the spelling of a token into a 688 /// SmallVector. Note that the returned StringRef may not point to the 689 /// supplied buffer if a copy can be avoided. 690 llvm::StringRef getSpelling(const Token &Tok, 691 llvm::SmallVectorImpl<char> &Buffer, 692 bool *Invalid = 0) const; 693 694 /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant 695 /// with length 1, return the character. 696 char getSpellingOfSingleCharacterNumericConstant(const Token &Tok, 697 bool *Invalid = 0) const { 698 assert(Tok.is(tok::numeric_constant) && 699 Tok.getLength() == 1 && "Called on unsupported token"); 700 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1"); 701 702 // If the token is carrying a literal data pointer, just use it. 703 if (const char *D = Tok.getLiteralData()) 704 return *D; 705 706 // Otherwise, fall back on getCharacterData, which is slower, but always 707 // works. 708 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid); 709 } 710 711 /// CreateString - Plop the specified string into a scratch buffer and set the 712 /// specified token's location and length to it. If specified, the source 713 /// location provides a location of the instantiation point of the token. 714 void CreateString(const char *Buf, unsigned Len, 715 Token &Tok, SourceLocation SourceLoc = SourceLocation()); 716 717 /// \brief Computes the source location just past the end of the 718 /// token at this source location. 719 /// 720 /// This routine can be used to produce a source location that 721 /// points just past the end of the token referenced by \p Loc, and 722 /// is generally used when a diagnostic needs to point just after a 723 /// token where it expected something different that it received. If 724 /// the returned source location would not be meaningful (e.g., if 725 /// it points into a macro), this routine returns an invalid 726 /// source location. 727 /// 728 /// \param Offset an offset from the end of the token, where the source 729 /// location should refer to. The default offset (0) produces a source 730 /// location pointing just past the end of the token; an offset of 1 produces 731 /// a source location pointing to the last character in the token, etc. 732 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) { 733 return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, Features); 734 } 735 736 /// DumpToken - Print the token to stderr, used for debugging. 737 /// 738 void DumpToken(const Token &Tok, bool DumpFlags = false) const; 739 void DumpLocation(SourceLocation Loc) const; 740 void DumpMacro(const MacroInfo &MI) const; 741 742 /// AdvanceToTokenCharacter - Given a location that specifies the start of a 743 /// token, return a new location that specifies a character within the token. 744 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, 745 unsigned Char) const { 746 return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, Features); 747 } 748 749 /// IncrementPasteCounter - Increment the counters for the number of token 750 /// paste operations performed. If fast was specified, this is a 'fast paste' 751 /// case we handled. 752 /// 753 void IncrementPasteCounter(bool isFast) { 754 if (isFast) 755 ++NumFastTokenPaste; 756 else 757 ++NumTokenPaste; 758 } 759 760 void PrintStats(); 761 762 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a 763 /// comment (/##/) in microsoft mode, this method handles updating the current 764 /// state, returning the token on the next source line. 765 void HandleMicrosoftCommentPaste(Token &Tok); 766 767 //===--------------------------------------------------------------------===// 768 // Preprocessor callback methods. These are invoked by a lexer as various 769 // directives and events are found. 770 771 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the 772 /// identifier information for the token and install it into the token, 773 /// updating the token kind accordingly. 774 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const; 775 776 /// HandleIdentifier - This callback is invoked when the lexer reads an 777 /// identifier and has filled in the tokens IdentifierInfo member. This 778 /// callback potentially macro expands it or turns it into a named token (like 779 /// 'for'). 780 void HandleIdentifier(Token &Identifier); 781 782 783 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of 784 /// the current file. This either returns the EOF token and returns true, or 785 /// pops a level off the include stack and returns false, at which point the 786 /// client should call lex again. 787 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false); 788 789 /// HandleEndOfTokenLexer - This callback is invoked when the current 790 /// TokenLexer hits the end of its token stream. 791 bool HandleEndOfTokenLexer(Token &Result); 792 793 /// HandleDirective - This callback is invoked when the lexer sees a # token 794 /// at the start of a line. This consumes the directive, modifies the 795 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 796 /// read is the correct one. 797 void HandleDirective(Token &Result); 798 799 /// CheckEndOfDirective - Ensure that the next token is a tok::eod token. If 800 /// not, emit a diagnostic and consume up until the eod. If EnableMacros is 801 /// true, then we consider macros that expand to zero tokens as being ok. 802 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false); 803 804 /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 805 /// current line until the tok::eod token is found. 806 void DiscardUntilEndOfDirective(); 807 808 /// SawDateOrTime - This returns true if the preprocessor has seen a use of 809 /// __DATE__ or __TIME__ in the file so far. 810 bool SawDateOrTime() const { 811 return DATELoc != SourceLocation() || TIMELoc != SourceLocation(); 812 } 813 unsigned getCounterValue() const { return CounterValue; } 814 void setCounterValue(unsigned V) { CounterValue = V; } 815 816 /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide 817 /// SourceLocation. 818 MacroInfo *AllocateMacroInfo(SourceLocation L); 819 820 /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI. 821 MacroInfo *CloneMacroInfo(const MacroInfo &MI); 822 823 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully 824 /// checked and spelled filename, e.g. as an operand of #include. This returns 825 /// true if the input filename was in <>'s or false if it were in ""'s. The 826 /// caller is expected to provide a buffer that is large enough to hold the 827 /// spelling of the filename, but is also expected to handle the case when 828 /// this method decides to use a different buffer. 829 bool GetIncludeFilenameSpelling(SourceLocation Loc,llvm::StringRef &Filename); 830 831 /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file, 832 /// return null on failure. isAngled indicates whether the file reference is 833 /// for system #include's or not (i.e. using <> instead of ""). 834 const FileEntry *LookupFile(llvm::StringRef Filename, 835 bool isAngled, const DirectoryLookup *FromDir, 836 const DirectoryLookup *&CurDir, 837 llvm::SmallVectorImpl<char> *SearchPath, 838 llvm::SmallVectorImpl<char> *RelativePath); 839 840 /// GetCurLookup - The DirectoryLookup structure used to find the current 841 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 842 /// implement #include_next and find directory-specific properties. 843 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; } 844 845 /// isInPrimaryFile - Return true if we're in the top-level file, not in a 846 /// #include. 847 bool isInPrimaryFile() const; 848 849 /// ConcatenateIncludeName - Handle cases where the #include name is expanded 850 /// from a macro as multiple tokens, which need to be glued together. This 851 /// occurs for code like: 852 /// #define FOO <a/b.h> 853 /// #include FOO 854 /// because in this case, "<a/b.h>" is returned as 7 tokens, not one. 855 /// 856 /// This code concatenates and consumes tokens up to the '>' token. It 857 /// returns false if the > was found, otherwise it returns true if it finds 858 /// and consumes the EOD marker. 859 bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer, 860 SourceLocation &End); 861 862 /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is 863 /// followed by EOD. Return true if the token is not a valid on-off-switch. 864 bool LexOnOffSwitch(tok::OnOffSwitch &OOS); 865 866private: 867 868 void PushIncludeMacroStack() { 869 IncludeMacroStack.push_back(IncludeStackInfo(CurLexer.take(), 870 CurPTHLexer.take(), 871 CurPPLexer, 872 CurTokenLexer.take(), 873 CurDirLookup)); 874 CurPPLexer = 0; 875 } 876 877 void PopIncludeMacroStack() { 878 CurLexer.reset(IncludeMacroStack.back().TheLexer); 879 CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer); 880 CurPPLexer = IncludeMacroStack.back().ThePPLexer; 881 CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer); 882 CurDirLookup = IncludeMacroStack.back().TheDirLookup; 883 IncludeMacroStack.pop_back(); 884 } 885 886 /// AllocateMacroInfo - Allocate a new MacroInfo object. 887 MacroInfo *AllocateMacroInfo(); 888 889 /// ReleaseMacroInfo - Release the specified MacroInfo. This memory will 890 /// be reused for allocating new MacroInfo objects. 891 void ReleaseMacroInfo(MacroInfo* MI); 892 893 /// ReadMacroName - Lex and validate a macro name, which occurs after a 894 /// #define or #undef. This emits a diagnostic, sets the token kind to eod, 895 /// and discards the rest of the macro line if the macro name is invalid. 896 void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0); 897 898 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 899 /// definition has just been read. Lex the rest of the arguments and the 900 /// closing ), updating MI with what we learn. Return true if an error occurs 901 /// parsing the arg list. 902 bool ReadMacroDefinitionArgList(MacroInfo *MI); 903 904 /// SkipExcludedConditionalBlock - We just read a #if or related directive and 905 /// decided that the subsequent tokens are in the #if'd out portion of the 906 /// file. Lex the rest of the file, until we see an #endif. If 907 /// FoundNonSkipPortion is true, then we have already emitted code for part of 908 /// this #if directive, so #else/#elif blocks should never be entered. If 909 /// FoundElse is false, then #else directives are ok, if not, then we have 910 /// already seen one so a #else directive is a duplicate. When this returns, 911 /// the caller can lex the first valid token. 912 void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 913 bool FoundNonSkipPortion, bool FoundElse); 914 915 /// PTHSkipExcludedConditionalBlock - A fast PTH version of 916 /// SkipExcludedConditionalBlock. 917 void PTHSkipExcludedConditionalBlock(); 918 919 /// EvaluateDirectiveExpression - Evaluate an integer constant expression that 920 /// may occur after a #if or #elif directive and return it as a bool. If the 921 /// expression is equivalent to "!defined(X)" return X in IfNDefMacro. 922 bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro); 923 924 /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas: 925 /// #pragma GCC poison/system_header/dependency and #pragma once. 926 void RegisterBuiltinPragmas(); 927 928 /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the 929 /// identifier table. 930 void RegisterBuiltinMacros(); 931 932 /// HandleMacroExpandedIdentifier - If an identifier token is read that is to 933 /// be expanded as a macro, handle it and return the next token as 'Tok'. If 934 /// the macro should not be expanded return true, otherwise return false. 935 bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI); 936 937 /// isNextPPTokenLParen - Determine whether the next preprocessor token to be 938 /// lexed is a '('. If so, consume the token and return true, if not, this 939 /// method should have no observable side-effect on the lexed tokens. 940 bool isNextPPTokenLParen(); 941 942 /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is 943 /// invoked to read all of the formal arguments specified for the macro 944 /// invocation. This returns null on error. 945 MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI, 946 SourceLocation &InstantiationEnd); 947 948 /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded 949 /// as a builtin macro, handle it and return the next token as 'Tok'. 950 void ExpandBuiltinMacro(Token &Tok); 951 952 /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then 953 /// return the first token after the directive. The _Pragma token has just 954 /// been read into 'Tok'. 955 void Handle_Pragma(Token &Tok); 956 957 /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text 958 /// is not enclosed within a string literal. 959 void HandleMicrosoft__pragma(Token &Tok); 960 961 /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and 962 /// start lexing tokens from it instead of the current buffer. 963 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir); 964 965 /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and 966 /// start getting tokens from it using the PTH cache. 967 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir); 968 969 /// IsFileLexer - Returns true if we are lexing from a file and not a 970 /// pragma or a macro. 971 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) { 972 return L ? !L->isPragmaLexer() : P != 0; 973 } 974 975 static bool IsFileLexer(const IncludeStackInfo& I) { 976 return IsFileLexer(I.TheLexer, I.ThePPLexer); 977 } 978 979 bool IsFileLexer() const { 980 return IsFileLexer(CurLexer.get(), CurPPLexer); 981 } 982 983 //===--------------------------------------------------------------------===// 984 // Caching stuff. 985 void CachingLex(Token &Result); 986 bool InCachingLexMode() const { 987 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means 988 // that we are past EOF, not that we are in CachingLex mode. 989 return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 && 990 !IncludeMacroStack.empty(); 991 } 992 void EnterCachingLexMode(); 993 void ExitCachingLexMode() { 994 if (InCachingLexMode()) 995 RemoveTopOfLexerStack(); 996 } 997 const Token &PeekAhead(unsigned N); 998 void AnnotatePreviousCachedTokens(const Token &Tok); 999 1000 //===--------------------------------------------------------------------===// 1001 /// Handle*Directive - implement the various preprocessor directives. These 1002 /// should side-effect the current preprocessor object so that the next call 1003 /// to Lex() will return the appropriate token next. 1004 void HandleLineDirective(Token &Tok); 1005 void HandleDigitDirective(Token &Tok); 1006 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning); 1007 void HandleIdentSCCSDirective(Token &Tok); 1008 1009 // File inclusion. 1010 void HandleIncludeDirective(SourceLocation HashLoc, 1011 Token &Tok, 1012 const DirectoryLookup *LookupFrom = 0, 1013 bool isImport = false); 1014 void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok); 1015 void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok); 1016 void HandleImportDirective(SourceLocation HashLoc, Token &Tok); 1017 1018 // Macro handling. 1019 void HandleDefineDirective(Token &Tok); 1020 void HandleUndefDirective(Token &Tok); 1021 1022 // Conditional Inclusion. 1023 void HandleIfdefDirective(Token &Tok, bool isIfndef, 1024 bool ReadAnyTokensBeforeDirective); 1025 void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective); 1026 void HandleEndifDirective(Token &Tok); 1027 void HandleElseDirective(Token &Tok); 1028 void HandleElifDirective(Token &Tok); 1029 1030 // Pragmas. 1031 void HandlePragmaDirective(unsigned Introducer); 1032public: 1033 void HandlePragmaOnce(Token &OnceTok); 1034 void HandlePragmaMark(); 1035 void HandlePragmaPoison(Token &PoisonTok); 1036 void HandlePragmaSystemHeader(Token &SysHeaderTok); 1037 void HandlePragmaDependency(Token &DependencyTok); 1038 void HandlePragmaComment(Token &CommentTok); 1039 void HandlePragmaMessage(Token &MessageTok); 1040 void HandlePragmaPushMacro(Token &Tok); 1041 void HandlePragmaPopMacro(Token &Tok); 1042 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok); 1043 1044 // Return true and store the first token only if any CommentHandler 1045 // has inserted some tokens and getCommentRetentionState() is false. 1046 bool HandleComment(Token &Token, SourceRange Comment); 1047 1048 /// \brief A macro is used, update information about macros that need unused 1049 /// warnings. 1050 void markMacroAsUsed(MacroInfo *MI); 1051}; 1052 1053/// \brief Abstract base class that describes a handler that will receive 1054/// source ranges for each of the comments encountered in the source file. 1055class CommentHandler { 1056public: 1057 virtual ~CommentHandler(); 1058 1059 // The handler shall return true if it has pushed any tokens 1060 // to be read using e.g. EnterToken or EnterTokenStream. 1061 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0; 1062}; 1063 1064} // end namespace clang 1065 1066#endif 1067