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/Basic/Builtins.h" 18 #include "clang/Basic/Diagnostic.h" 19 #include "clang/Basic/IdentifierTable.h" 20 #include "clang/Basic/SourceLocation.h" 21 #include "clang/Lex/Lexer.h" 22 #include "clang/Lex/MacroInfo.h" 23 #include "clang/Lex/PPCallbacks.h" 24 #include "clang/Lex/PPMutationListener.h" 25 #include "clang/Lex/PTHLexer.h" 26 #include "clang/Lex/PTHManager.h" 27 #include "clang/Lex/TokenLexer.h" 28 #include "llvm/ADT/ArrayRef.h" 29 #include "llvm/ADT/DenseMap.h" 30 #include "llvm/ADT/IntrusiveRefCntPtr.h" 31 #include "llvm/ADT/OwningPtr.h" 32 #include "llvm/ADT/SmallPtrSet.h" 33 #include "llvm/ADT/SmallVector.h" 34 #include "llvm/Support/Allocator.h" 35 #include <vector> 36 37 namespace llvm { 38 template<unsigned InternalLen> class SmallString; 39 } 40 41 namespace clang { 42 43 class SourceManager; 44 class ExternalPreprocessorSource; 45 class FileManager; 46 class FileEntry; 47 class HeaderSearch; 48 class PragmaNamespace; 49 class PragmaHandler; 50 class CommentHandler; 51 class ScratchBuffer; 52 class TargetInfo; 53 class PPCallbacks; 54 class CodeCompletionHandler; 55 class DirectoryLookup; 56 class PreprocessingRecord; 57 class ModuleLoader; 58 class PreprocessorOptions; 59 60 /// \brief Stores token information for comparing actual tokens with 61 /// predefined values. Only handles simple tokens and identifiers. 62 class TokenValue { 63 tok::TokenKind Kind; 64 IdentifierInfo *II; 65 66 public: 67 TokenValue(tok::TokenKind Kind) : Kind(Kind), II(0) { 68 assert(Kind != tok::raw_identifier && "Raw identifiers are not supported."); 69 assert(Kind != tok::identifier && 70 "Identifiers should be created by TokenValue(IdentifierInfo *)"); 71 assert(!tok::isLiteral(Kind) && "Literals are not supported."); 72 assert(!tok::isAnnotation(Kind) && "Annotations are not supported."); 73 } 74 TokenValue(IdentifierInfo *II) : Kind(tok::identifier), II(II) {} 75 bool operator==(const Token &Tok) const { 76 return Tok.getKind() == Kind && 77 (!II || II == Tok.getIdentifierInfo()); 78 } 79 }; 80 81 /// Preprocessor - This object engages in a tight little dance with the lexer to 82 /// efficiently preprocess tokens. Lexers know only about tokens within a 83 /// single source file, and don't know anything about preprocessor-level issues 84 /// like the \#include stack, token expansion, etc. 85 /// 86 class Preprocessor : public RefCountedBase<Preprocessor> { 87 IntrusiveRefCntPtr<PreprocessorOptions> PPOpts; 88 DiagnosticsEngine *Diags; 89 LangOptions &LangOpts; 90 const TargetInfo *Target; 91 FileManager &FileMgr; 92 SourceManager &SourceMgr; 93 ScratchBuffer *ScratchBuf; 94 HeaderSearch &HeaderInfo; 95 ModuleLoader &TheModuleLoader; 96 97 /// \brief External source of macros. 98 ExternalPreprocessorSource *ExternalSource; 99 100 101 /// PTH - An optional PTHManager object used for getting tokens from 102 /// a token cache rather than lexing the original source file. 103 OwningPtr<PTHManager> PTH; 104 105 /// BP - A BumpPtrAllocator object used to quickly allocate and release 106 /// objects internal to the Preprocessor. 107 llvm::BumpPtrAllocator BP; 108 109 /// Identifiers for builtin macros and other builtins. 110 IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__ 111 IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__ 112 IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__ 113 IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__ 114 IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__ 115 IdentifierInfo *Ident__COUNTER__; // __COUNTER__ 116 IdentifierInfo *Ident_Pragma, *Ident__pragma; // _Pragma, __pragma 117 IdentifierInfo *Ident__VA_ARGS__; // __VA_ARGS__ 118 IdentifierInfo *Ident__has_feature; // __has_feature 119 IdentifierInfo *Ident__has_extension; // __has_extension 120 IdentifierInfo *Ident__has_builtin; // __has_builtin 121 IdentifierInfo *Ident__has_attribute; // __has_attribute 122 IdentifierInfo *Ident__has_include; // __has_include 123 IdentifierInfo *Ident__has_include_next; // __has_include_next 124 IdentifierInfo *Ident__has_warning; // __has_warning 125 IdentifierInfo *Ident__building_module; // __building_module 126 IdentifierInfo *Ident__MODULE__; // __MODULE__ 127 128 SourceLocation DATELoc, TIMELoc; 129 unsigned CounterValue; // Next __COUNTER__ value. 130 131 enum { 132 /// MaxIncludeStackDepth - Maximum depth of \#includes. 133 MaxAllowedIncludeStackDepth = 200 134 }; 135 136 // State that is set before the preprocessor begins. 137 bool KeepComments : 1; 138 bool KeepMacroComments : 1; 139 bool SuppressIncludeNotFoundError : 1; 140 141 // State that changes while the preprocessor runs: 142 bool InMacroArgs : 1; // True if parsing fn macro invocation args. 143 144 /// Whether the preprocessor owns the header search object. 145 bool OwnsHeaderSearch : 1; 146 147 /// DisableMacroExpansion - True if macro expansion is disabled. 148 bool DisableMacroExpansion : 1; 149 150 /// MacroExpansionInDirectivesOverride - Temporarily disables 151 /// DisableMacroExpansion (i.e. enables expansion) when parsing preprocessor 152 /// directives. 153 bool MacroExpansionInDirectivesOverride : 1; 154 155 class ResetMacroExpansionHelper; 156 157 /// \brief Whether we have already loaded macros from the external source. 158 mutable bool ReadMacrosFromExternalSource : 1; 159 160 /// \brief True if pragmas are enabled. 161 bool PragmasEnabled : 1; 162 163 /// \brief True if the current build action is a preprocessing action. 164 bool PreprocessedOutput : 1; 165 166 /// \brief True if we are currently preprocessing a #if or #elif directive 167 bool ParsingIfOrElifDirective; 168 169 /// \brief True if we are pre-expanding macro arguments. 170 bool InMacroArgPreExpansion; 171 172 /// Identifiers - This is mapping/lookup information for all identifiers in 173 /// the program, including program keywords. 174 mutable IdentifierTable Identifiers; 175 176 /// Selectors - This table contains all the selectors in the program. Unlike 177 /// IdentifierTable above, this table *isn't* populated by the preprocessor. 178 /// It is declared/expanded here because it's role/lifetime is 179 /// conceptually similar the IdentifierTable. In addition, the current control 180 /// flow (in clang::ParseAST()), make it convenient to put here. 181 /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to 182 /// the lifetime of the preprocessor. 183 SelectorTable Selectors; 184 185 /// BuiltinInfo - Information about builtins. 186 Builtin::Context BuiltinInfo; 187 188 /// PragmaHandlers - This tracks all of the pragmas that the client registered 189 /// with this preprocessor. 190 PragmaNamespace *PragmaHandlers; 191 192 /// \brief Tracks all of the comment handlers that the client registered 193 /// with this preprocessor. 194 std::vector<CommentHandler *> CommentHandlers; 195 196 /// \brief True if we want to ignore EOF token and continue later on (thus 197 /// avoid tearing the Lexer and etc. down). 198 bool IncrementalProcessing; 199 200 /// \brief The code-completion handler. 201 CodeCompletionHandler *CodeComplete; 202 203 /// \brief The file that we're performing code-completion for, if any. 204 const FileEntry *CodeCompletionFile; 205 206 /// \brief The offset in file for the code-completion point. 207 unsigned CodeCompletionOffset; 208 209 /// \brief The location for the code-completion point. This gets instantiated 210 /// when the CodeCompletionFile gets \#include'ed for preprocessing. 211 SourceLocation CodeCompletionLoc; 212 213 /// \brief The start location for the file of the code-completion point. 214 /// 215 /// This gets instantiated when the CodeCompletionFile gets \#include'ed 216 /// for preprocessing. 217 SourceLocation CodeCompletionFileLoc; 218 219 /// \brief The source location of the 'import' contextual keyword we just 220 /// lexed, if any. 221 SourceLocation ModuleImportLoc; 222 223 /// \brief The module import path that we're currently processing. 224 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> ModuleImportPath; 225 226 /// \brief Whether the module import expectes an identifier next. Otherwise, 227 /// it expects a '.' or ';'. 228 bool ModuleImportExpectsIdentifier; 229 230 /// \brief The source location of the currently-active 231 /// #pragma clang arc_cf_code_audited begin. 232 SourceLocation PragmaARCCFCodeAuditedLoc; 233 234 /// \brief True if we hit the code-completion point. 235 bool CodeCompletionReached; 236 237 /// \brief The number of bytes that we will initially skip when entering the 238 /// main file, which is used when loading a precompiled preamble, along 239 /// with a flag that indicates whether skipping this number of bytes will 240 /// place the lexer at the start of a line. 241 std::pair<unsigned, bool> SkipMainFilePreamble; 242 243 /// CurLexer - This is the current top of the stack that we're lexing from if 244 /// not expanding a macro and we are lexing directly from source code. 245 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 246 OwningPtr<Lexer> CurLexer; 247 248 /// CurPTHLexer - This is the current top of stack that we're lexing from if 249 /// not expanding from a macro and we are lexing from a PTH cache. 250 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 251 OwningPtr<PTHLexer> CurPTHLexer; 252 253 /// CurPPLexer - This is the current top of the stack what we're lexing from 254 /// if not expanding a macro. This is an alias for either CurLexer or 255 /// CurPTHLexer. 256 PreprocessorLexer *CurPPLexer; 257 258 /// CurLookup - The DirectoryLookup structure used to find the current 259 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 260 /// implement \#include_next and find directory-specific properties. 261 const DirectoryLookup *CurDirLookup; 262 263 /// CurTokenLexer - This is the current macro we are expanding, if we are 264 /// expanding a macro. One of CurLexer and CurTokenLexer must be null. 265 OwningPtr<TokenLexer> CurTokenLexer; 266 267 /// \brief The kind of lexer we're currently working with. 268 enum CurLexerKind { 269 CLK_Lexer, 270 CLK_PTHLexer, 271 CLK_TokenLexer, 272 CLK_CachingLexer, 273 CLK_LexAfterModuleImport 274 } CurLexerKind; 275 276 /// IncludeMacroStack - This keeps track of the stack of files currently 277 /// \#included, and macros currently being expanded from, not counting 278 /// CurLexer/CurTokenLexer. 279 struct IncludeStackInfo { 280 enum CurLexerKind CurLexerKind; 281 Lexer *TheLexer; 282 PTHLexer *ThePTHLexer; 283 PreprocessorLexer *ThePPLexer; 284 TokenLexer *TheTokenLexer; 285 const DirectoryLookup *TheDirLookup; 286 287 IncludeStackInfo(enum CurLexerKind K, Lexer *L, PTHLexer* P, 288 PreprocessorLexer* PPL, 289 TokenLexer* TL, const DirectoryLookup *D) 290 : CurLexerKind(K), TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL), 291 TheTokenLexer(TL), TheDirLookup(D) {} 292 }; 293 std::vector<IncludeStackInfo> IncludeMacroStack; 294 295 /// Callbacks - These are actions invoked when some preprocessor activity is 296 /// encountered (e.g. a file is \#included, etc). 297 PPCallbacks *Callbacks; 298 299 /// \brief Listener whose actions are invoked when an entity in the 300 /// preprocessor (e.g., a macro) that was loaded from an AST file is 301 /// later mutated. 302 PPMutationListener *Listener; 303 304 struct MacroExpandsInfo { 305 Token Tok; 306 MacroDirective *MD; 307 SourceRange Range; 308 MacroExpandsInfo(Token Tok, MacroDirective *MD, SourceRange Range) 309 : Tok(Tok), MD(MD), Range(Range) { } 310 }; 311 SmallVector<MacroExpandsInfo, 2> DelayedMacroExpandsCallbacks; 312 313 /// Macros - For each IdentifierInfo that was associated with a macro, we 314 /// keep a mapping to the history of all macro definitions and #undefs in 315 /// the reverse order (the latest one is in the head of the list). 316 llvm::DenseMap<const IdentifierInfo*, MacroDirective*> Macros; 317 friend class ASTReader; 318 319 /// \brief Macros that we want to warn because they are not used at the end 320 /// of the translation unit; we store just their SourceLocations instead 321 /// something like MacroInfo*. The benefit of this is that when we are 322 /// deserializing from PCH, we don't need to deserialize identifier & macros 323 /// just so that we can report that they are unused, we just warn using 324 /// the SourceLocations of this set (that will be filled by the ASTReader). 325 /// We are using SmallPtrSet instead of a vector for faster removal. 326 typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy; 327 WarnUnusedMacroLocsTy WarnUnusedMacroLocs; 328 329 /// MacroArgCache - This is a "freelist" of MacroArg objects that can be 330 /// reused for quick allocation. 331 MacroArgs *MacroArgCache; 332 friend class MacroArgs; 333 334 /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma 335 /// push_macro directive, we keep a MacroInfo stack used to restore 336 /// previous macro value. 337 llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo; 338 339 // Various statistics we track for performance analysis. 340 unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma; 341 unsigned NumIf, NumElse, NumEndif; 342 unsigned NumEnteredSourceFiles, MaxIncludeStackDepth; 343 unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded; 344 unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste; 345 unsigned NumSkipped; 346 347 /// Predefines - This string is the predefined macros that preprocessor 348 /// should use from the command line etc. 349 std::string Predefines; 350 351 /// \brief The file ID for the preprocessor predefines. 352 FileID PredefinesFileID; 353 354 /// TokenLexerCache - Cache macro expanders to reduce malloc traffic. 355 enum { TokenLexerCacheSize = 8 }; 356 unsigned NumCachedTokenLexers; 357 TokenLexer *TokenLexerCache[TokenLexerCacheSize]; 358 359 /// \brief Keeps macro expanded tokens for TokenLexers. 360 // 361 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is 362 /// going to lex in the cache and when it finishes the tokens are removed 363 /// from the end of the cache. 364 SmallVector<Token, 16> MacroExpandedTokens; 365 std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack; 366 367 /// \brief A record of the macro definitions and expansions that 368 /// occurred during preprocessing. 369 /// 370 /// This is an optional side structure that can be enabled with 371 /// \c createPreprocessingRecord() prior to preprocessing. 372 PreprocessingRecord *Record; 373 374 private: // Cached tokens state. 375 typedef SmallVector<Token, 1> CachedTokensTy; 376 377 /// CachedTokens - Cached tokens are stored here when we do backtracking or 378 /// lookahead. They are "lexed" by the CachingLex() method. 379 CachedTokensTy CachedTokens; 380 381 /// CachedLexPos - The position of the cached token that CachingLex() should 382 /// "lex" next. If it points beyond the CachedTokens vector, it means that 383 /// a normal Lex() should be invoked. 384 CachedTokensTy::size_type CachedLexPos; 385 386 /// BacktrackPositions - Stack of backtrack positions, allowing nested 387 /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to 388 /// indicate where CachedLexPos should be set when the BackTrack() method is 389 /// invoked (at which point the last position is popped). 390 std::vector<CachedTokensTy::size_type> BacktrackPositions; 391 392 struct MacroInfoChain { 393 MacroInfo MI; 394 MacroInfoChain *Next; 395 MacroInfoChain *Prev; 396 }; 397 398 /// MacroInfos are managed as a chain for easy disposal. This is the head 399 /// of that list. 400 MacroInfoChain *MIChainHead; 401 402 /// MICache - A "freelist" of MacroInfo objects that can be reused for quick 403 /// allocation. 404 MacroInfoChain *MICache; 405 406 public: 407 Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts, 408 DiagnosticsEngine &diags, LangOptions &opts, 409 const TargetInfo *target, 410 SourceManager &SM, HeaderSearch &Headers, 411 ModuleLoader &TheModuleLoader, 412 IdentifierInfoLookup *IILookup = 0, 413 bool OwnsHeaderSearch = false, 414 bool DelayInitialization = false, 415 bool IncrProcessing = false); 416 417 ~Preprocessor(); 418 419 /// \brief Initialize the preprocessor, if the constructor did not already 420 /// perform the initialization. 421 /// 422 /// \param Target Information about the target. 423 void Initialize(const TargetInfo &Target); 424 425 /// \brief Retrieve the preprocessor options used to initialize this 426 /// preprocessor. 427 PreprocessorOptions &getPreprocessorOpts() const { return *PPOpts; } 428 429 DiagnosticsEngine &getDiagnostics() const { return *Diags; } 430 void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; } 431 432 const LangOptions &getLangOpts() const { return LangOpts; } 433 const TargetInfo &getTargetInfo() const { return *Target; } 434 FileManager &getFileManager() const { return FileMgr; } 435 SourceManager &getSourceManager() const { return SourceMgr; } 436 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; } 437 438 IdentifierTable &getIdentifierTable() { return Identifiers; } 439 SelectorTable &getSelectorTable() { return Selectors; } 440 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; } 441 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; } 442 443 void setPTHManager(PTHManager* pm); 444 445 PTHManager *getPTHManager() { return PTH.get(); } 446 447 void setExternalSource(ExternalPreprocessorSource *Source) { 448 ExternalSource = Source; 449 } 450 451 ExternalPreprocessorSource *getExternalSource() const { 452 return ExternalSource; 453 } 454 455 /// \brief Retrieve the module loader associated with this preprocessor. 456 ModuleLoader &getModuleLoader() const { return TheModuleLoader; } 457 458 /// \brief True if we are currently preprocessing a #if or #elif directive 459 bool isParsingIfOrElifDirective() const { 460 return ParsingIfOrElifDirective; 461 } 462 463 /// SetCommentRetentionState - Control whether or not the preprocessor retains 464 /// comments in output. 465 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) { 466 this->KeepComments = KeepComments | KeepMacroComments; 467 this->KeepMacroComments = KeepMacroComments; 468 } 469 470 bool getCommentRetentionState() const { return KeepComments; } 471 472 void setPragmasEnabled(bool Enabled) { PragmasEnabled = Enabled; } 473 bool getPragmasEnabled() const { return PragmasEnabled; } 474 475 void SetSuppressIncludeNotFoundError(bool Suppress) { 476 SuppressIncludeNotFoundError = Suppress; 477 } 478 479 bool GetSuppressIncludeNotFoundError() { 480 return SuppressIncludeNotFoundError; 481 } 482 483 /// Sets whether the preprocessor is responsible for producing output or if 484 /// it is producing tokens to be consumed by Parse and Sema. 485 void setPreprocessedOutput(bool IsPreprocessedOutput) { 486 PreprocessedOutput = IsPreprocessedOutput; 487 } 488 489 /// Returns true if the preprocessor is responsible for generating output, 490 /// false if it is producing tokens to be consumed by Parse and Sema. 491 bool isPreprocessedOutput() const { return PreprocessedOutput; } 492 493 /// isCurrentLexer - Return true if we are lexing directly from the specified 494 /// lexer. 495 bool isCurrentLexer(const PreprocessorLexer *L) const { 496 return CurPPLexer == L; 497 } 498 499 /// getCurrentLexer - Return the current lexer being lexed from. Note 500 /// that this ignores any potentially active macro expansions and _Pragma 501 /// expansions going on at the time. 502 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; } 503 504 /// getCurrentFileLexer - Return the current file lexer being lexed from. 505 /// Note that this ignores any potentially active macro expansions and _Pragma 506 /// expansions going on at the time. 507 PreprocessorLexer *getCurrentFileLexer() const; 508 509 /// \brief Returns the file ID for the preprocessor predefines. 510 FileID getPredefinesFileID() const { return PredefinesFileID; } 511 512 /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks. 513 /// Note that this class takes ownership of any PPCallbacks object given to 514 /// it. 515 PPCallbacks *getPPCallbacks() const { return Callbacks; } 516 void addPPCallbacks(PPCallbacks *C) { 517 if (Callbacks) 518 C = new PPChainedCallbacks(C, Callbacks); 519 Callbacks = C; 520 } 521 522 /// \brief Attach an preprocessor mutation listener to the preprocessor. 523 /// 524 /// The preprocessor mutation listener provides the ability to track 525 /// modifications to the preprocessor entities committed after they were 526 /// initially created. 527 void setPPMutationListener(PPMutationListener *Listener) { 528 this->Listener = Listener; 529 } 530 531 /// \brief Retrieve a pointer to the preprocessor mutation listener 532 /// associated with this preprocessor, if any. 533 PPMutationListener *getPPMutationListener() const { return Listener; } 534 535 /// \brief Given an identifier, return the MacroInfo it is \#defined to 536 /// or null if it isn't \#define'd. 537 MacroDirective *getMacroDirective(IdentifierInfo *II) const { 538 if (!II->hasMacroDefinition()) 539 return 0; 540 541 MacroDirective *MD = getMacroDirectiveHistory(II); 542 assert(MD->getUndefLoc().isInvalid() && "Macro is undefined!"); 543 return MD; 544 } 545 546 const MacroInfo *getMacroInfo(IdentifierInfo *II) const { 547 return const_cast<Preprocessor*>(this)->getMacroInfo(II); 548 } 549 550 MacroInfo *getMacroInfo(IdentifierInfo *II) { 551 if (MacroDirective *MD = getMacroDirective(II)) 552 return MD->getInfo(); 553 return 0; 554 } 555 556 /// \brief Given an identifier, return the (probably #undef'd) MacroInfo 557 /// representing the most recent macro definition. One can iterate over all 558 /// previous macro definitions from it. This method should only be called for 559 /// identifiers that hadMacroDefinition(). 560 MacroDirective *getMacroDirectiveHistory(const IdentifierInfo *II) const; 561 562 /// \brief Specify a macro for this identifier. 563 MacroDirective *setMacroDirective(IdentifierInfo *II, MacroInfo *MI, 564 SourceLocation Loc, bool isImported); 565 MacroDirective *setMacroDirective(IdentifierInfo *II, MacroInfo *MI) { 566 return setMacroDirective(II, MI, MI->getDefinitionLoc(), false); 567 } 568 /// \brief Add a MacroInfo that was loaded from an AST file. 569 void addLoadedMacroInfo(IdentifierInfo *II, MacroDirective *MD, 570 MacroDirective *Hint = 0); 571 /// \brief Make the given MacroInfo, that was loaded from an AST file and 572 /// previously hidden, visible. 573 void makeLoadedMacroInfoVisible(IdentifierInfo *II, MacroDirective *MD); 574 /// \brief Undefine a macro for this identifier. 575 void clearMacroInfo(IdentifierInfo *II); 576 577 /// macro_iterator/macro_begin/macro_end - This allows you to walk the macro 578 /// history table. Currently defined macros have 579 /// IdentifierInfo::hasMacroDefinition() set and an empty 580 /// MacroInfo::getUndefLoc() at the head of the list. 581 typedef llvm::DenseMap<const IdentifierInfo *, 582 MacroDirective*>::const_iterator macro_iterator; 583 macro_iterator macro_begin(bool IncludeExternalMacros = true) const; 584 macro_iterator macro_end(bool IncludeExternalMacros = true) const; 585 586 /// \brief Return the name of the macro defined before \p Loc that has 587 /// spelling \p Tokens. If there are multiple macros with same spelling, 588 /// return the last one defined. 589 StringRef getLastMacroWithSpelling(SourceLocation Loc, 590 ArrayRef<TokenValue> Tokens) const; 591 592 const std::string &getPredefines() const { return Predefines; } 593 /// setPredefines - Set the predefines for this Preprocessor. These 594 /// predefines are automatically injected when parsing the main file. 595 void setPredefines(const char *P) { Predefines = P; } 596 void setPredefines(const std::string &P) { Predefines = P; } 597 598 /// Return information about the specified preprocessor 599 /// identifier token. 600 IdentifierInfo *getIdentifierInfo(StringRef Name) const { 601 return &Identifiers.get(Name); 602 } 603 604 /// AddPragmaHandler - Add the specified pragma handler to the preprocessor. 605 /// If 'Namespace' is non-null, then it is a token required to exist on the 606 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC". 607 void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler); 608 void AddPragmaHandler(PragmaHandler *Handler) { 609 AddPragmaHandler(StringRef(), Handler); 610 } 611 612 /// RemovePragmaHandler - Remove the specific pragma handler from 613 /// the preprocessor. If \p Namespace is non-null, then it should 614 /// be the namespace that \p Handler was added to. It is an error 615 /// to remove a handler that has not been registered. 616 void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler); 617 void RemovePragmaHandler(PragmaHandler *Handler) { 618 RemovePragmaHandler(StringRef(), Handler); 619 } 620 621 /// \brief Add the specified comment handler to the preprocessor. 622 void addCommentHandler(CommentHandler *Handler); 623 624 /// \brief Remove the specified comment handler. 625 /// 626 /// It is an error to remove a handler that has not been registered. 627 void removeCommentHandler(CommentHandler *Handler); 628 629 /// \brief Set the code completion handler to the given object. 630 void setCodeCompletionHandler(CodeCompletionHandler &Handler) { 631 CodeComplete = &Handler; 632 } 633 634 /// \brief Retrieve the current code-completion handler. 635 CodeCompletionHandler *getCodeCompletionHandler() const { 636 return CodeComplete; 637 } 638 639 /// \brief Clear out the code completion handler. 640 void clearCodeCompletionHandler() { 641 CodeComplete = 0; 642 } 643 644 /// \brief Hook used by the lexer to invoke the "natural language" code 645 /// completion point. 646 void CodeCompleteNaturalLanguage(); 647 648 /// \brief Retrieve the preprocessing record, or NULL if there is no 649 /// preprocessing record. 650 PreprocessingRecord *getPreprocessingRecord() const { return Record; } 651 652 /// \brief Create a new preprocessing record, which will keep track of 653 /// all macro expansions, macro definitions, etc. 654 void createPreprocessingRecord(); 655 656 /// EnterMainSourceFile - Enter the specified FileID as the main source file, 657 /// which implicitly adds the builtin defines etc. 658 void EnterMainSourceFile(); 659 660 /// EndSourceFile - Inform the preprocessor callbacks that processing is 661 /// complete. 662 void EndSourceFile(); 663 664 /// EnterSourceFile - Add a source file to the top of the include stack and 665 /// start lexing tokens from it instead of the current buffer. Emit an error 666 /// and don't enter the file on error. 667 void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, 668 SourceLocation Loc); 669 670 /// EnterMacro - Add a Macro to the top of the include stack and start lexing 671 /// tokens from it instead of the current buffer. Args specifies the 672 /// tokens input to a function-like macro. 673 /// 674 /// ILEnd specifies the location of the ')' for a function-like macro or the 675 /// identifier for an object-like macro. 676 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroInfo *Macro, 677 MacroArgs *Args); 678 679 /// EnterTokenStream - Add a "macro" context to the top of the include stack, 680 /// which will cause the lexer to start returning the specified tokens. 681 /// 682 /// If DisableMacroExpansion is true, tokens lexed from the token stream will 683 /// not be subject to further macro expansion. Otherwise, these tokens will 684 /// be re-macro-expanded when/if expansion is enabled. 685 /// 686 /// If OwnsTokens is false, this method assumes that the specified stream of 687 /// tokens has a permanent owner somewhere, so they do not need to be copied. 688 /// If it is true, it assumes the array of tokens is allocated with new[] and 689 /// must be freed. 690 /// 691 void EnterTokenStream(const Token *Toks, unsigned NumToks, 692 bool DisableMacroExpansion, bool OwnsTokens); 693 694 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 695 /// lexer stack. This should only be used in situations where the current 696 /// state of the top-of-stack lexer is known. 697 void RemoveTopOfLexerStack(); 698 699 /// EnableBacktrackAtThisPos - From the point that this method is called, and 700 /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor 701 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will 702 /// make the Preprocessor re-lex the same tokens. 703 /// 704 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can 705 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will 706 /// be combined with the EnableBacktrackAtThisPos calls in reverse order. 707 /// 708 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack 709 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of 710 /// tokens will continue indefinitely. 711 /// 712 void EnableBacktrackAtThisPos(); 713 714 /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call. 715 void CommitBacktrackedTokens(); 716 717 /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since 718 /// EnableBacktrackAtThisPos() was previously called. 719 void Backtrack(); 720 721 /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and 722 /// caching of tokens is on. 723 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); } 724 725 /// Lex - To lex a token from the preprocessor, just pull a token from the 726 /// current lexer or macro object. 727 void Lex(Token &Result) { 728 switch (CurLexerKind) { 729 case CLK_Lexer: CurLexer->Lex(Result); break; 730 case CLK_PTHLexer: CurPTHLexer->Lex(Result); break; 731 case CLK_TokenLexer: CurTokenLexer->Lex(Result); break; 732 case CLK_CachingLexer: CachingLex(Result); break; 733 case CLK_LexAfterModuleImport: LexAfterModuleImport(Result); break; 734 } 735 } 736 737 void LexAfterModuleImport(Token &Result); 738 739 /// \brief Lex a string literal, which may be the concatenation of multiple 740 /// string literals and may even come from macro expansion. 741 /// \returns true on success, false if a error diagnostic has been generated. 742 bool LexStringLiteral(Token &Result, std::string &String, 743 const char *DiagnosticTag, bool AllowMacroExpansion) { 744 if (AllowMacroExpansion) 745 Lex(Result); 746 else 747 LexUnexpandedToken(Result); 748 return FinishLexStringLiteral(Result, String, DiagnosticTag, 749 AllowMacroExpansion); 750 } 751 752 /// \brief Complete the lexing of a string literal where the first token has 753 /// already been lexed (see LexStringLiteral). 754 bool FinishLexStringLiteral(Token &Result, std::string &String, 755 const char *DiagnosticTag, 756 bool AllowMacroExpansion); 757 758 /// LexNonComment - Lex a token. If it's a comment, keep lexing until we get 759 /// something not a comment. This is useful in -E -C mode where comments 760 /// would foul up preprocessor directive handling. 761 void LexNonComment(Token &Result) { 762 do 763 Lex(Result); 764 while (Result.getKind() == tok::comment); 765 } 766 767 /// LexUnexpandedToken - This is just like Lex, but this disables macro 768 /// expansion of identifier tokens. 769 void LexUnexpandedToken(Token &Result) { 770 // Disable macro expansion. 771 bool OldVal = DisableMacroExpansion; 772 DisableMacroExpansion = true; 773 // Lex the token. 774 Lex(Result); 775 776 // Reenable it. 777 DisableMacroExpansion = OldVal; 778 } 779 780 /// LexUnexpandedNonComment - Like LexNonComment, but this disables macro 781 /// expansion of identifier tokens. 782 void LexUnexpandedNonComment(Token &Result) { 783 do 784 LexUnexpandedToken(Result); 785 while (Result.getKind() == tok::comment); 786 } 787 788 /// Disables macro expansion everywhere except for preprocessor directives. 789 void SetMacroExpansionOnlyInDirectives() { 790 DisableMacroExpansion = true; 791 MacroExpansionInDirectivesOverride = true; 792 } 793 794 /// LookAhead - This peeks ahead N tokens and returns that token without 795 /// consuming any tokens. LookAhead(0) returns the next token that would be 796 /// returned by Lex(), LookAhead(1) returns the token after it, etc. This 797 /// returns normal tokens after phase 5. As such, it is equivalent to using 798 /// 'Lex', not 'LexUnexpandedToken'. 799 const Token &LookAhead(unsigned N) { 800 if (CachedLexPos + N < CachedTokens.size()) 801 return CachedTokens[CachedLexPos+N]; 802 else 803 return PeekAhead(N+1); 804 } 805 806 /// RevertCachedTokens - When backtracking is enabled and tokens are cached, 807 /// this allows to revert a specific number of tokens. 808 /// Note that the number of tokens being reverted should be up to the last 809 /// backtrack position, not more. 810 void RevertCachedTokens(unsigned N) { 811 assert(isBacktrackEnabled() && 812 "Should only be called when tokens are cached for backtracking"); 813 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back()) 814 && "Should revert tokens up to the last backtrack position, not more"); 815 assert(signed(CachedLexPos) - signed(N) >= 0 && 816 "Corrupted backtrack positions ?"); 817 CachedLexPos -= N; 818 } 819 820 /// EnterToken - Enters a token in the token stream to be lexed next. If 821 /// BackTrack() is called afterwards, the token will remain at the insertion 822 /// point. 823 void EnterToken(const Token &Tok) { 824 EnterCachingLexMode(); 825 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok); 826 } 827 828 /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching 829 /// tokens (because backtrack is enabled) it should replace the most recent 830 /// cached tokens with the given annotation token. This function has no effect 831 /// if backtracking is not enabled. 832 /// 833 /// Note that the use of this function is just for optimization; so that the 834 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is 835 /// invoked. 836 void AnnotateCachedTokens(const Token &Tok) { 837 assert(Tok.isAnnotation() && "Expected annotation token"); 838 if (CachedLexPos != 0 && isBacktrackEnabled()) 839 AnnotatePreviousCachedTokens(Tok); 840 } 841 842 /// \brief Replace the last token with an annotation token. 843 /// 844 /// Like AnnotateCachedTokens(), this routine replaces an 845 /// already-parsed (and resolved) token with an annotation 846 /// token. However, this routine only replaces the last token with 847 /// the annotation token; it does not affect any other cached 848 /// tokens. This function has no effect if backtracking is not 849 /// enabled. 850 void ReplaceLastTokenWithAnnotation(const Token &Tok) { 851 assert(Tok.isAnnotation() && "Expected annotation token"); 852 if (CachedLexPos != 0 && isBacktrackEnabled()) 853 CachedTokens[CachedLexPos-1] = Tok; 854 } 855 856 /// TypoCorrectToken - Update the current token to represent the provided 857 /// identifier, in order to cache an action performed by typo correction. 858 void TypoCorrectToken(const Token &Tok) { 859 assert(Tok.getIdentifierInfo() && "Expected identifier token"); 860 if (CachedLexPos != 0 && isBacktrackEnabled()) 861 CachedTokens[CachedLexPos-1] = Tok; 862 } 863 864 /// \brief Recompute the current lexer kind based on the CurLexer/CurPTHLexer/ 865 /// CurTokenLexer pointers. 866 void recomputeCurLexerKind(); 867 868 /// \brief Returns true if incremental processing is enabled 869 bool isIncrementalProcessingEnabled() const { return IncrementalProcessing; } 870 871 /// \brief Enables the incremental processing 872 void enableIncrementalProcessing(bool value = true) { 873 IncrementalProcessing = value; 874 } 875 876 /// \brief Specify the point at which code-completion will be performed. 877 /// 878 /// \param File the file in which code completion should occur. If 879 /// this file is included multiple times, code-completion will 880 /// perform completion the first time it is included. If NULL, this 881 /// function clears out the code-completion point. 882 /// 883 /// \param Line the line at which code completion should occur 884 /// (1-based). 885 /// 886 /// \param Column the column at which code completion should occur 887 /// (1-based). 888 /// 889 /// \returns true if an error occurred, false otherwise. 890 bool SetCodeCompletionPoint(const FileEntry *File, 891 unsigned Line, unsigned Column); 892 893 /// \brief Determine if we are performing code completion. 894 bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; } 895 896 /// \brief Returns the location of the code-completion point. 897 /// Returns an invalid location if code-completion is not enabled or the file 898 /// containing the code-completion point has not been lexed yet. 899 SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; } 900 901 /// \brief Returns the start location of the file of code-completion point. 902 /// Returns an invalid location if code-completion is not enabled or the file 903 /// containing the code-completion point has not been lexed yet. 904 SourceLocation getCodeCompletionFileLoc() const { 905 return CodeCompletionFileLoc; 906 } 907 908 /// \brief Returns true if code-completion is enabled and we have hit the 909 /// code-completion point. 910 bool isCodeCompletionReached() const { return CodeCompletionReached; } 911 912 /// \brief Note that we hit the code-completion point. 913 void setCodeCompletionReached() { 914 assert(isCodeCompletionEnabled() && "Code-completion not enabled!"); 915 CodeCompletionReached = true; 916 // Silence any diagnostics that occur after we hit the code-completion. 917 getDiagnostics().setSuppressAllDiagnostics(true); 918 } 919 920 /// \brief The location of the currently-active \#pragma clang 921 /// arc_cf_code_audited begin. Returns an invalid location if there 922 /// is no such pragma active. 923 SourceLocation getPragmaARCCFCodeAuditedLoc() const { 924 return PragmaARCCFCodeAuditedLoc; 925 } 926 927 /// \brief Set the location of the currently-active \#pragma clang 928 /// arc_cf_code_audited begin. An invalid location ends the pragma. 929 void setPragmaARCCFCodeAuditedLoc(SourceLocation Loc) { 930 PragmaARCCFCodeAuditedLoc = Loc; 931 } 932 933 /// \brief Instruct the preprocessor to skip part of the main source file. 934 /// 935 /// \param Bytes The number of bytes in the preamble to skip. 936 /// 937 /// \param StartOfLine Whether skipping these bytes puts the lexer at the 938 /// start of a line. 939 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) { 940 SkipMainFilePreamble.first = Bytes; 941 SkipMainFilePreamble.second = StartOfLine; 942 } 943 944 /// Diag - Forwarding function for diagnostics. This emits a diagnostic at 945 /// the specified Token's location, translating the token's start 946 /// position in the current buffer into a SourcePosition object for rendering. 947 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const { 948 return Diags->Report(Loc, DiagID); 949 } 950 951 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const { 952 return Diags->Report(Tok.getLocation(), DiagID); 953 } 954 955 /// getSpelling() - Return the 'spelling' of the token at the given 956 /// location; does not go up to the spelling location or down to the 957 /// expansion location. 958 /// 959 /// \param buffer A buffer which will be used only if the token requires 960 /// "cleaning", e.g. if it contains trigraphs or escaped newlines 961 /// \param invalid If non-null, will be set \c true if an error occurs. 962 StringRef getSpelling(SourceLocation loc, 963 SmallVectorImpl<char> &buffer, 964 bool *invalid = 0) const { 965 return Lexer::getSpelling(loc, buffer, SourceMgr, LangOpts, invalid); 966 } 967 968 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 969 /// token is the characters used to represent the token in the source file 970 /// after trigraph expansion and escaped-newline folding. In particular, this 971 /// wants to get the true, uncanonicalized, spelling of things like digraphs 972 /// UCNs, etc. 973 /// 974 /// \param Invalid If non-null, will be set \c true if an error occurs. 975 std::string getSpelling(const Token &Tok, bool *Invalid = 0) const { 976 return Lexer::getSpelling(Tok, SourceMgr, LangOpts, Invalid); 977 } 978 979 /// getSpelling - This method is used to get the spelling of a token into a 980 /// preallocated buffer, instead of as an std::string. The caller is required 981 /// to allocate enough space for the token, which is guaranteed to be at least 982 /// Tok.getLength() bytes long. The length of the actual result is returned. 983 /// 984 /// Note that this method may do two possible things: it may either fill in 985 /// the buffer specified with characters, or it may *change the input pointer* 986 /// to point to a constant buffer with the data already in it (avoiding a 987 /// copy). The caller is not allowed to modify the returned buffer pointer 988 /// if an internal buffer is returned. 989 unsigned getSpelling(const Token &Tok, const char *&Buffer, 990 bool *Invalid = 0) const { 991 return Lexer::getSpelling(Tok, Buffer, SourceMgr, LangOpts, Invalid); 992 } 993 994 /// getSpelling - This method is used to get the spelling of a token into a 995 /// SmallVector. Note that the returned StringRef may not point to the 996 /// supplied buffer if a copy can be avoided. 997 StringRef getSpelling(const Token &Tok, 998 SmallVectorImpl<char> &Buffer, 999 bool *Invalid = 0) const; 1000 1001 /// \brief Relex the token at the specified location. 1002 /// \returns true if there was a failure, false on success. 1003 bool getRawToken(SourceLocation Loc, Token &Result) { 1004 return Lexer::getRawToken(Loc, Result, SourceMgr, LangOpts); 1005 } 1006 1007 /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant 1008 /// with length 1, return the character. 1009 char getSpellingOfSingleCharacterNumericConstant(const Token &Tok, 1010 bool *Invalid = 0) const { 1011 assert(Tok.is(tok::numeric_constant) && 1012 Tok.getLength() == 1 && "Called on unsupported token"); 1013 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1"); 1014 1015 // If the token is carrying a literal data pointer, just use it. 1016 if (const char *D = Tok.getLiteralData()) 1017 return *D; 1018 1019 // Otherwise, fall back on getCharacterData, which is slower, but always 1020 // works. 1021 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid); 1022 } 1023 1024 /// \brief Retrieve the name of the immediate macro expansion. 1025 /// 1026 /// This routine starts from a source location, and finds the name of the macro 1027 /// responsible for its immediate expansion. It looks through any intervening 1028 /// macro argument expansions to compute this. It returns a StringRef which 1029 /// refers to the SourceManager-owned buffer of the source where that macro 1030 /// name is spelled. Thus, the result shouldn't out-live the SourceManager. 1031 StringRef getImmediateMacroName(SourceLocation Loc) { 1032 return Lexer::getImmediateMacroName(Loc, SourceMgr, getLangOpts()); 1033 } 1034 1035 /// CreateString - Plop the specified string into a scratch buffer and set the 1036 /// specified token's location and length to it. If specified, the source 1037 /// location provides a location of the expansion point of the token. 1038 void CreateString(StringRef Str, Token &Tok, 1039 SourceLocation ExpansionLocStart = SourceLocation(), 1040 SourceLocation ExpansionLocEnd = SourceLocation()); 1041 1042 /// \brief Computes the source location just past the end of the 1043 /// token at this source location. 1044 /// 1045 /// This routine can be used to produce a source location that 1046 /// points just past the end of the token referenced by \p Loc, and 1047 /// is generally used when a diagnostic needs to point just after a 1048 /// token where it expected something different that it received. If 1049 /// the returned source location would not be meaningful (e.g., if 1050 /// it points into a macro), this routine returns an invalid 1051 /// source location. 1052 /// 1053 /// \param Offset an offset from the end of the token, where the source 1054 /// location should refer to. The default offset (0) produces a source 1055 /// location pointing just past the end of the token; an offset of 1 produces 1056 /// a source location pointing to the last character in the token, etc. 1057 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) { 1058 return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, LangOpts); 1059 } 1060 1061 /// \brief Returns true if the given MacroID location points at the first 1062 /// token of the macro expansion. 1063 /// 1064 /// \param MacroBegin If non-null and function returns true, it is set to 1065 /// begin location of the macro. 1066 bool isAtStartOfMacroExpansion(SourceLocation loc, 1067 SourceLocation *MacroBegin = 0) const { 1068 return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, LangOpts, 1069 MacroBegin); 1070 } 1071 1072 /// \brief Returns true if the given MacroID location points at the last 1073 /// token of the macro expansion. 1074 /// 1075 /// \param MacroEnd If non-null and function returns true, it is set to 1076 /// end location of the macro. 1077 bool isAtEndOfMacroExpansion(SourceLocation loc, 1078 SourceLocation *MacroEnd = 0) const { 1079 return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, LangOpts, MacroEnd); 1080 } 1081 1082 /// DumpToken - Print the token to stderr, used for debugging. 1083 /// 1084 void DumpToken(const Token &Tok, bool DumpFlags = false) const; 1085 void DumpLocation(SourceLocation Loc) const; 1086 void DumpMacro(const MacroInfo &MI) const; 1087 1088 /// AdvanceToTokenCharacter - Given a location that specifies the start of a 1089 /// token, return a new location that specifies a character within the token. 1090 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, 1091 unsigned Char) const { 1092 return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, LangOpts); 1093 } 1094 1095 /// IncrementPasteCounter - Increment the counters for the number of token 1096 /// paste operations performed. If fast was specified, this is a 'fast paste' 1097 /// case we handled. 1098 /// 1099 void IncrementPasteCounter(bool isFast) { 1100 if (isFast) 1101 ++NumFastTokenPaste; 1102 else 1103 ++NumTokenPaste; 1104 } 1105 1106 void PrintStats(); 1107 1108 size_t getTotalMemory() const; 1109 1110 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a 1111 /// comment (/##/) in microsoft mode, this method handles updating the current 1112 /// state, returning the token on the next source line. 1113 void HandleMicrosoftCommentPaste(Token &Tok); 1114 1115 //===--------------------------------------------------------------------===// 1116 // Preprocessor callback methods. These are invoked by a lexer as various 1117 // directives and events are found. 1118 1119 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the 1120 /// identifier information for the token and install it into the token, 1121 /// updating the token kind accordingly. 1122 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const; 1123 1124 private: 1125 llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons; 1126 1127 public: 1128 1129 // SetPoisonReason - Call this function to indicate the reason for 1130 // poisoning an identifier. If that identifier is accessed while 1131 // poisoned, then this reason will be used instead of the default 1132 // "poisoned" diagnostic. 1133 void SetPoisonReason(IdentifierInfo *II, unsigned DiagID); 1134 1135 // HandlePoisonedIdentifier - Display reason for poisoned 1136 // identifier. 1137 void HandlePoisonedIdentifier(Token & Tok); 1138 1139 void MaybeHandlePoisonedIdentifier(Token & Identifier) { 1140 if(IdentifierInfo * II = Identifier.getIdentifierInfo()) { 1141 if(II->isPoisoned()) { 1142 HandlePoisonedIdentifier(Identifier); 1143 } 1144 } 1145 } 1146 1147 private: 1148 /// Identifiers used for SEH handling in Borland. These are only 1149 /// allowed in particular circumstances 1150 // __except block 1151 IdentifierInfo *Ident__exception_code, 1152 *Ident___exception_code, 1153 *Ident_GetExceptionCode; 1154 // __except filter expression 1155 IdentifierInfo *Ident__exception_info, 1156 *Ident___exception_info, 1157 *Ident_GetExceptionInfo; 1158 // __finally 1159 IdentifierInfo *Ident__abnormal_termination, 1160 *Ident___abnormal_termination, 1161 *Ident_AbnormalTermination; 1162 public: 1163 void PoisonSEHIdentifiers(bool Poison = true); // Borland 1164 1165 /// HandleIdentifier - This callback is invoked when the lexer reads an 1166 /// identifier and has filled in the tokens IdentifierInfo member. This 1167 /// callback potentially macro expands it or turns it into a named token (like 1168 /// 'for'). 1169 void HandleIdentifier(Token &Identifier); 1170 1171 1172 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of 1173 /// the current file. This either returns the EOF token and returns true, or 1174 /// pops a level off the include stack and returns false, at which point the 1175 /// client should call lex again. 1176 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false); 1177 1178 /// HandleEndOfTokenLexer - This callback is invoked when the current 1179 /// TokenLexer hits the end of its token stream. 1180 bool HandleEndOfTokenLexer(Token &Result); 1181 1182 /// HandleDirective - This callback is invoked when the lexer sees a # token 1183 /// at the start of a line. This consumes the directive, modifies the 1184 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 1185 /// read is the correct one. 1186 void HandleDirective(Token &Result); 1187 1188 /// CheckEndOfDirective - Ensure that the next token is a tok::eod token. If 1189 /// not, emit a diagnostic and consume up until the eod. If EnableMacros is 1190 /// true, then we consider macros that expand to zero tokens as being ok. 1191 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false); 1192 1193 /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 1194 /// current line until the tok::eod token is found. 1195 void DiscardUntilEndOfDirective(); 1196 1197 /// SawDateOrTime - This returns true if the preprocessor has seen a use of 1198 /// __DATE__ or __TIME__ in the file so far. 1199 bool SawDateOrTime() const { 1200 return DATELoc != SourceLocation() || TIMELoc != SourceLocation(); 1201 } 1202 unsigned getCounterValue() const { return CounterValue; } 1203 void setCounterValue(unsigned V) { CounterValue = V; } 1204 1205 /// \brief Retrieves the module that we're currently building, if any. 1206 Module *getCurrentModule(); 1207 1208 /// \brief Allocate a new MacroInfo object with the provided SourceLocation. 1209 MacroInfo *AllocateMacroInfo(SourceLocation L); 1210 1211 /// \brief Turn the specified lexer token into a fully checked and spelled 1212 /// filename, e.g. as an operand of \#include. 1213 /// 1214 /// The caller is expected to provide a buffer that is large enough to hold 1215 /// the spelling of the filename, but is also expected to handle the case 1216 /// when this method decides to use a different buffer. 1217 /// 1218 /// \returns true if the input filename was in <>'s or false if it was 1219 /// in ""'s. 1220 bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename); 1221 1222 /// \brief Given a "foo" or \<foo> reference, look up the indicated file. 1223 /// 1224 /// Returns null on failure. \p isAngled indicates whether the file 1225 /// reference is for system \#include's or not (i.e. using <> instead of ""). 1226 const FileEntry *LookupFile(StringRef Filename, 1227 bool isAngled, const DirectoryLookup *FromDir, 1228 const DirectoryLookup *&CurDir, 1229 SmallVectorImpl<char> *SearchPath, 1230 SmallVectorImpl<char> *RelativePath, 1231 Module **SuggestedModule, 1232 bool SkipCache = false); 1233 1234 /// GetCurLookup - The DirectoryLookup structure used to find the current 1235 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 1236 /// implement \#include_next and find directory-specific properties. 1237 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; } 1238 1239 /// \brief Return true if we're in the top-level file, not in a \#include. 1240 bool isInPrimaryFile() const; 1241 1242 /// ConcatenateIncludeName - Handle cases where the \#include name is expanded 1243 /// from a macro as multiple tokens, which need to be glued together. This 1244 /// occurs for code like: 1245 /// \code 1246 /// \#define FOO <x/y.h> 1247 /// \#include FOO 1248 /// \endcode 1249 /// because in this case, "<x/y.h>" is returned as 7 tokens, not one. 1250 /// 1251 /// This code concatenates and consumes tokens up to the '>' token. It 1252 /// returns false if the > was found, otherwise it returns true if it finds 1253 /// and consumes the EOD marker. 1254 bool ConcatenateIncludeName(SmallString<128> &FilenameBuffer, 1255 SourceLocation &End); 1256 1257 /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is 1258 /// followed by EOD. Return true if the token is not a valid on-off-switch. 1259 bool LexOnOffSwitch(tok::OnOffSwitch &OOS); 1260 1261 private: 1262 1263 void PushIncludeMacroStack() { 1264 IncludeMacroStack.push_back(IncludeStackInfo(CurLexerKind, 1265 CurLexer.take(), 1266 CurPTHLexer.take(), 1267 CurPPLexer, 1268 CurTokenLexer.take(), 1269 CurDirLookup)); 1270 CurPPLexer = 0; 1271 } 1272 1273 void PopIncludeMacroStack() { 1274 CurLexer.reset(IncludeMacroStack.back().TheLexer); 1275 CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer); 1276 CurPPLexer = IncludeMacroStack.back().ThePPLexer; 1277 CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer); 1278 CurDirLookup = IncludeMacroStack.back().TheDirLookup; 1279 CurLexerKind = IncludeMacroStack.back().CurLexerKind; 1280 IncludeMacroStack.pop_back(); 1281 } 1282 1283 /// \brief Allocate a new MacroInfo object. 1284 MacroInfo *AllocateMacroInfo(); 1285 1286 MacroDirective *AllocateMacroDirective(MacroInfo *MI, SourceLocation Loc, 1287 bool isImported); 1288 1289 /// \brief Release the specified MacroInfo for re-use. 1290 /// 1291 /// This memory will be reused for allocating new MacroInfo objects. 1292 void ReleaseMacroInfo(MacroInfo* MI); 1293 1294 /// ReadMacroName - Lex and validate a macro name, which occurs after a 1295 /// \#define or \#undef. This emits a diagnostic, sets the token kind to eod, 1296 /// and discards the rest of the macro line if the macro name is invalid. 1297 void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0); 1298 1299 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1300 /// definition has just been read. Lex the rest of the arguments and the 1301 /// closing ), updating MI with what we learn and saving in LastTok the 1302 /// last token read. 1303 /// Return true if an error occurs parsing the arg list. 1304 bool ReadMacroDefinitionArgList(MacroInfo *MI, Token& LastTok); 1305 1306 /// We just read a \#if or related directive and decided that the 1307 /// subsequent tokens are in the \#if'd out portion of the 1308 /// file. Lex the rest of the file, until we see an \#endif. If \p 1309 /// FoundNonSkipPortion is true, then we have already emitted code for part of 1310 /// this \#if directive, so \#else/\#elif blocks should never be entered. If 1311 /// \p FoundElse is false, then \#else directives are ok, if not, then we have 1312 /// already seen one so a \#else directive is a duplicate. When this returns, 1313 /// the caller can lex the first valid token. 1314 void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 1315 bool FoundNonSkipPortion, bool FoundElse, 1316 SourceLocation ElseLoc = SourceLocation()); 1317 1318 /// \brief A fast PTH version of SkipExcludedConditionalBlock. 1319 void PTHSkipExcludedConditionalBlock(); 1320 1321 /// EvaluateDirectiveExpression - Evaluate an integer constant expression that 1322 /// may occur after a #if or #elif directive and return it as a bool. If the 1323 /// expression is equivalent to "!defined(X)" return X in IfNDefMacro. 1324 bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro); 1325 1326 /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas: 1327 /// \#pragma GCC poison/system_header/dependency and \#pragma once. 1328 void RegisterBuiltinPragmas(); 1329 1330 /// \brief Register builtin macros such as __LINE__ with the identifier table. 1331 void RegisterBuiltinMacros(); 1332 1333 /// HandleMacroExpandedIdentifier - If an identifier token is read that is to 1334 /// be expanded as a macro, handle it and return the next token as 'Tok'. If 1335 /// the macro should not be expanded return true, otherwise return false. 1336 bool HandleMacroExpandedIdentifier(Token &Tok, MacroDirective *MD); 1337 1338 /// \brief Cache macro expanded tokens for TokenLexers. 1339 // 1340 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is 1341 /// going to lex in the cache and when it finishes the tokens are removed 1342 /// from the end of the cache. 1343 Token *cacheMacroExpandedTokens(TokenLexer *tokLexer, 1344 ArrayRef<Token> tokens); 1345 void removeCachedMacroExpandedTokensOfLastLexer(); 1346 friend void TokenLexer::ExpandFunctionArguments(); 1347 1348 /// isNextPPTokenLParen - Determine whether the next preprocessor token to be 1349 /// lexed is a '('. If so, consume the token and return true, if not, this 1350 /// method should have no observable side-effect on the lexed tokens. 1351 bool isNextPPTokenLParen(); 1352 1353 /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is 1354 /// invoked to read all of the formal arguments specified for the macro 1355 /// invocation. This returns null on error. 1356 MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI, 1357 SourceLocation &ExpansionEnd); 1358 1359 /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded 1360 /// as a builtin macro, handle it and return the next token as 'Tok'. 1361 void ExpandBuiltinMacro(Token &Tok); 1362 1363 /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then 1364 /// return the first token after the directive. The _Pragma token has just 1365 /// been read into 'Tok'. 1366 void Handle_Pragma(Token &Tok); 1367 1368 /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text 1369 /// is not enclosed within a string literal. 1370 void HandleMicrosoft__pragma(Token &Tok); 1371 1372 /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and 1373 /// start lexing tokens from it instead of the current buffer. 1374 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir); 1375 1376 /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and 1377 /// start getting tokens from it using the PTH cache. 1378 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir); 1379 1380 /// \brief Set the file ID for the preprocessor predefines. 1381 void setPredefinesFileID(FileID FID) { 1382 assert(PredefinesFileID.isInvalid() && "PredefinesFileID already set!"); 1383 PredefinesFileID = FID; 1384 } 1385 1386 /// IsFileLexer - Returns true if we are lexing from a file and not a 1387 /// pragma or a macro. 1388 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) { 1389 return L ? !L->isPragmaLexer() : P != 0; 1390 } 1391 1392 static bool IsFileLexer(const IncludeStackInfo& I) { 1393 return IsFileLexer(I.TheLexer, I.ThePPLexer); 1394 } 1395 1396 bool IsFileLexer() const { 1397 return IsFileLexer(CurLexer.get(), CurPPLexer); 1398 } 1399 1400 //===--------------------------------------------------------------------===// 1401 // Caching stuff. 1402 void CachingLex(Token &Result); 1403 bool InCachingLexMode() const { 1404 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means 1405 // that we are past EOF, not that we are in CachingLex mode. 1406 return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 && 1407 !IncludeMacroStack.empty(); 1408 } 1409 void EnterCachingLexMode(); 1410 void ExitCachingLexMode() { 1411 if (InCachingLexMode()) 1412 RemoveTopOfLexerStack(); 1413 } 1414 const Token &PeekAhead(unsigned N); 1415 void AnnotatePreviousCachedTokens(const Token &Tok); 1416 1417 //===--------------------------------------------------------------------===// 1418 /// Handle*Directive - implement the various preprocessor directives. These 1419 /// should side-effect the current preprocessor object so that the next call 1420 /// to Lex() will return the appropriate token next. 1421 void HandleLineDirective(Token &Tok); 1422 void HandleDigitDirective(Token &Tok); 1423 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning); 1424 void HandleIdentSCCSDirective(Token &Tok); 1425 void HandleMacroPublicDirective(Token &Tok); 1426 void HandleMacroPrivateDirective(Token &Tok); 1427 1428 // File inclusion. 1429 void HandleIncludeDirective(SourceLocation HashLoc, 1430 Token &Tok, 1431 const DirectoryLookup *LookupFrom = 0, 1432 bool isImport = false); 1433 void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok); 1434 void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok); 1435 void HandleImportDirective(SourceLocation HashLoc, Token &Tok); 1436 void HandleMicrosoftImportDirective(Token &Tok); 1437 1438 // Macro handling. 1439 void HandleDefineDirective(Token &Tok); 1440 void HandleUndefDirective(Token &Tok); 1441 void UndefineMacro(IdentifierInfo *II, MacroDirective *MD, 1442 SourceLocation UndefLoc); 1443 1444 // Conditional Inclusion. 1445 void HandleIfdefDirective(Token &Tok, bool isIfndef, 1446 bool ReadAnyTokensBeforeDirective); 1447 void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective); 1448 void HandleEndifDirective(Token &Tok); 1449 void HandleElseDirective(Token &Tok); 1450 void HandleElifDirective(Token &Tok); 1451 1452 // Pragmas. 1453 void HandlePragmaDirective(unsigned Introducer); 1454 public: 1455 void HandlePragmaOnce(Token &OnceTok); 1456 void HandlePragmaMark(); 1457 void HandlePragmaPoison(Token &PoisonTok); 1458 void HandlePragmaSystemHeader(Token &SysHeaderTok); 1459 void HandlePragmaDependency(Token &DependencyTok); 1460 void HandlePragmaComment(Token &CommentTok); 1461 void HandlePragmaMessage(Token &MessageTok); 1462 void HandlePragmaPushMacro(Token &Tok); 1463 void HandlePragmaPopMacro(Token &Tok); 1464 void HandlePragmaIncludeAlias(Token &Tok); 1465 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok); 1466 1467 // Return true and store the first token only if any CommentHandler 1468 // has inserted some tokens and getCommentRetentionState() is false. 1469 bool HandleComment(Token &Token, SourceRange Comment); 1470 1471 /// \brief A macro is used, update information about macros that need unused 1472 /// warnings. 1473 void markMacroAsUsed(MacroInfo *MI); 1474 }; 1475 1476 /// \brief Abstract base class that describes a handler that will receive 1477 /// source ranges for each of the comments encountered in the source file. 1478 class CommentHandler { 1479 public: 1480 virtual ~CommentHandler(); 1481 1482 // The handler shall return true if it has pushed any tokens 1483 // to be read using e.g. EnterToken or EnterTokenStream. 1484 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0; 1485 }; 1486 1487 } // end namespace clang 1488 1489 #endif 1490