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/ModuleMap.h" 24 #include "clang/Lex/PPCallbacks.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 last token we lexed was an '@'. 227 bool LastTokenWasAt; 228 229 /// \brief Whether the module import expectes an identifier next. Otherwise, 230 /// it expects a '.' or ';'. 231 bool ModuleImportExpectsIdentifier; 232 233 /// \brief The source location of the currently-active 234 /// #pragma clang arc_cf_code_audited begin. 235 SourceLocation PragmaARCCFCodeAuditedLoc; 236 237 /// \brief True if we hit the code-completion point. 238 bool CodeCompletionReached; 239 240 /// \brief The number of bytes that we will initially skip when entering the 241 /// main file, which is used when loading a precompiled preamble, along 242 /// with a flag that indicates whether skipping this number of bytes will 243 /// place the lexer at the start of a line. 244 std::pair<unsigned, bool> SkipMainFilePreamble; 245 246 /// CurLexer - This is the current top of the stack that we're lexing from if 247 /// not expanding a macro and we are lexing directly from source code. 248 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 249 OwningPtr<Lexer> CurLexer; 250 251 /// CurPTHLexer - This is the current top of stack that we're lexing from if 252 /// not expanding from a macro and we are lexing from a PTH cache. 253 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null. 254 OwningPtr<PTHLexer> CurPTHLexer; 255 256 /// CurPPLexer - This is the current top of the stack what we're lexing from 257 /// if not expanding a macro. This is an alias for either CurLexer or 258 /// CurPTHLexer. 259 PreprocessorLexer *CurPPLexer; 260 261 /// CurLookup - The DirectoryLookup structure used to find the current 262 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 263 /// implement \#include_next and find directory-specific properties. 264 const DirectoryLookup *CurDirLookup; 265 266 /// CurTokenLexer - This is the current macro we are expanding, if we are 267 /// expanding a macro. One of CurLexer and CurTokenLexer must be null. 268 OwningPtr<TokenLexer> CurTokenLexer; 269 270 /// \brief The kind of lexer we're currently working with. 271 enum CurLexerKind { 272 CLK_Lexer, 273 CLK_PTHLexer, 274 CLK_TokenLexer, 275 CLK_CachingLexer, 276 CLK_LexAfterModuleImport 277 } CurLexerKind; 278 279 /// IncludeMacroStack - This keeps track of the stack of files currently 280 /// \#included, and macros currently being expanded from, not counting 281 /// CurLexer/CurTokenLexer. 282 struct IncludeStackInfo { 283 enum CurLexerKind CurLexerKind; 284 Lexer *TheLexer; 285 PTHLexer *ThePTHLexer; 286 PreprocessorLexer *ThePPLexer; 287 TokenLexer *TheTokenLexer; 288 const DirectoryLookup *TheDirLookup; 289 290 IncludeStackInfo(enum CurLexerKind K, Lexer *L, PTHLexer* P, 291 PreprocessorLexer* PPL, 292 TokenLexer* TL, const DirectoryLookup *D) 293 : CurLexerKind(K), TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL), 294 TheTokenLexer(TL), TheDirLookup(D) {} 295 }; 296 std::vector<IncludeStackInfo> IncludeMacroStack; 297 298 /// Callbacks - These are actions invoked when some preprocessor activity is 299 /// encountered (e.g. a file is \#included, etc). 300 PPCallbacks *Callbacks; 301 302 struct MacroExpandsInfo { 303 Token Tok; 304 MacroDirective *MD; 305 SourceRange Range; 306 MacroExpandsInfo(Token Tok, MacroDirective *MD, SourceRange Range) 307 : Tok(Tok), MD(MD), Range(Range) { } 308 }; 309 SmallVector<MacroExpandsInfo, 2> DelayedMacroExpandsCallbacks; 310 311 /// Macros - For each IdentifierInfo that was associated with a macro, we 312 /// keep a mapping to the history of all macro definitions and #undefs in 313 /// the reverse order (the latest one is in the head of the list). 314 llvm::DenseMap<const IdentifierInfo*, MacroDirective*> Macros; 315 friend class ASTReader; 316 317 /// \brief Macros that we want to warn because they are not used at the end 318 /// of the translation unit; we store just their SourceLocations instead 319 /// something like MacroInfo*. The benefit of this is that when we are 320 /// deserializing from PCH, we don't need to deserialize identifier & macros 321 /// just so that we can report that they are unused, we just warn using 322 /// the SourceLocations of this set (that will be filled by the ASTReader). 323 /// We are using SmallPtrSet instead of a vector for faster removal. 324 typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy; 325 WarnUnusedMacroLocsTy WarnUnusedMacroLocs; 326 327 /// MacroArgCache - This is a "freelist" of MacroArg objects that can be 328 /// reused for quick allocation. 329 MacroArgs *MacroArgCache; 330 friend class MacroArgs; 331 332 /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma 333 /// push_macro directive, we keep a MacroInfo stack used to restore 334 /// previous macro value. 335 llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo; 336 337 // Various statistics we track for performance analysis. 338 unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma; 339 unsigned NumIf, NumElse, NumEndif; 340 unsigned NumEnteredSourceFiles, MaxIncludeStackDepth; 341 unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded; 342 unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste; 343 unsigned NumSkipped; 344 345 /// Predefines - This string is the predefined macros that preprocessor 346 /// should use from the command line etc. 347 std::string Predefines; 348 349 /// \brief The file ID for the preprocessor predefines. 350 FileID PredefinesFileID; 351 352 /// TokenLexerCache - Cache macro expanders to reduce malloc traffic. 353 enum { TokenLexerCacheSize = 8 }; 354 unsigned NumCachedTokenLexers; 355 TokenLexer *TokenLexerCache[TokenLexerCacheSize]; 356 357 /// \brief Keeps macro expanded tokens for TokenLexers. 358 // 359 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is 360 /// going to lex in the cache and when it finishes the tokens are removed 361 /// from the end of the cache. 362 SmallVector<Token, 16> MacroExpandedTokens; 363 std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack; 364 365 /// \brief A record of the macro definitions and expansions that 366 /// occurred during preprocessing. 367 /// 368 /// This is an optional side structure that can be enabled with 369 /// \c createPreprocessingRecord() prior to preprocessing. 370 PreprocessingRecord *Record; 371 372 private: // Cached tokens state. 373 typedef SmallVector<Token, 1> CachedTokensTy; 374 375 /// CachedTokens - Cached tokens are stored here when we do backtracking or 376 /// lookahead. They are "lexed" by the CachingLex() method. 377 CachedTokensTy CachedTokens; 378 379 /// CachedLexPos - The position of the cached token that CachingLex() should 380 /// "lex" next. If it points beyond the CachedTokens vector, it means that 381 /// a normal Lex() should be invoked. 382 CachedTokensTy::size_type CachedLexPos; 383 384 /// BacktrackPositions - Stack of backtrack positions, allowing nested 385 /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to 386 /// indicate where CachedLexPos should be set when the BackTrack() method is 387 /// invoked (at which point the last position is popped). 388 std::vector<CachedTokensTy::size_type> BacktrackPositions; 389 390 struct MacroInfoChain { 391 MacroInfo MI; 392 MacroInfoChain *Next; 393 MacroInfoChain *Prev; 394 }; 395 396 /// MacroInfos are managed as a chain for easy disposal. This is the head 397 /// of that list. 398 MacroInfoChain *MIChainHead; 399 400 /// MICache - A "freelist" of MacroInfo objects that can be reused for quick 401 /// allocation. 402 MacroInfoChain *MICache; 403 404 struct DeserializedMacroInfoChain { 405 MacroInfo MI; 406 unsigned OwningModuleID; // MUST be immediately after the MacroInfo object 407 // so it can be accessed by MacroInfo::getOwningModuleID(). 408 DeserializedMacroInfoChain *Next; 409 }; 410 DeserializedMacroInfoChain *DeserialMIChainHead; 411 412 public: 413 Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts, 414 DiagnosticsEngine &diags, LangOptions &opts, 415 const TargetInfo *target, 416 SourceManager &SM, HeaderSearch &Headers, 417 ModuleLoader &TheModuleLoader, 418 IdentifierInfoLookup *IILookup = 0, 419 bool OwnsHeaderSearch = false, 420 bool DelayInitialization = false, 421 bool IncrProcessing = false); 422 423 ~Preprocessor(); 424 425 /// \brief Initialize the preprocessor, if the constructor did not already 426 /// perform the initialization. 427 /// 428 /// \param Target Information about the target. 429 void Initialize(const TargetInfo &Target); 430 431 /// \brief Retrieve the preprocessor options used to initialize this 432 /// preprocessor. 433 PreprocessorOptions &getPreprocessorOpts() const { return *PPOpts; } 434 435 DiagnosticsEngine &getDiagnostics() const { return *Diags; } 436 void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; } 437 438 const LangOptions &getLangOpts() const { return LangOpts; } 439 const TargetInfo &getTargetInfo() const { return *Target; } 440 FileManager &getFileManager() const { return FileMgr; } 441 SourceManager &getSourceManager() const { return SourceMgr; } 442 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; } 443 444 IdentifierTable &getIdentifierTable() { return Identifiers; } 445 SelectorTable &getSelectorTable() { return Selectors; } 446 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; } 447 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; } 448 449 void setPTHManager(PTHManager* pm); 450 451 PTHManager *getPTHManager() { return PTH.get(); } 452 453 void setExternalSource(ExternalPreprocessorSource *Source) { 454 ExternalSource = Source; 455 } 456 457 ExternalPreprocessorSource *getExternalSource() const { 458 return ExternalSource; 459 } 460 461 /// \brief Retrieve the module loader associated with this preprocessor. 462 ModuleLoader &getModuleLoader() const { return TheModuleLoader; } 463 464 bool hadModuleLoaderFatalFailure() const { 465 return TheModuleLoader.HadFatalFailure; 466 } 467 468 /// \brief True if we are currently preprocessing a #if or #elif directive 469 bool isParsingIfOrElifDirective() const { 470 return ParsingIfOrElifDirective; 471 } 472 473 /// SetCommentRetentionState - Control whether or not the preprocessor retains 474 /// comments in output. 475 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) { 476 this->KeepComments = KeepComments | KeepMacroComments; 477 this->KeepMacroComments = KeepMacroComments; 478 } 479 480 bool getCommentRetentionState() const { return KeepComments; } 481 482 void setPragmasEnabled(bool Enabled) { PragmasEnabled = Enabled; } 483 bool getPragmasEnabled() const { return PragmasEnabled; } 484 485 void SetSuppressIncludeNotFoundError(bool Suppress) { 486 SuppressIncludeNotFoundError = Suppress; 487 } 488 489 bool GetSuppressIncludeNotFoundError() { 490 return SuppressIncludeNotFoundError; 491 } 492 493 /// Sets whether the preprocessor is responsible for producing output or if 494 /// it is producing tokens to be consumed by Parse and Sema. 495 void setPreprocessedOutput(bool IsPreprocessedOutput) { 496 PreprocessedOutput = IsPreprocessedOutput; 497 } 498 499 /// Returns true if the preprocessor is responsible for generating output, 500 /// false if it is producing tokens to be consumed by Parse and Sema. 501 bool isPreprocessedOutput() const { return PreprocessedOutput; } 502 503 /// isCurrentLexer - Return true if we are lexing directly from the specified 504 /// lexer. 505 bool isCurrentLexer(const PreprocessorLexer *L) const { 506 return CurPPLexer == L; 507 } 508 509 /// getCurrentLexer - Return the current lexer being lexed from. Note 510 /// that this ignores any potentially active macro expansions and _Pragma 511 /// expansions going on at the time. 512 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; } 513 514 /// getCurrentFileLexer - Return the current file lexer being lexed from. 515 /// Note that this ignores any potentially active macro expansions and _Pragma 516 /// expansions going on at the time. 517 PreprocessorLexer *getCurrentFileLexer() const; 518 519 /// \brief Returns the file ID for the preprocessor predefines. 520 FileID getPredefinesFileID() const { return PredefinesFileID; } 521 522 /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks. 523 /// Note that this class takes ownership of any PPCallbacks object given to 524 /// it. 525 PPCallbacks *getPPCallbacks() const { return Callbacks; } 526 void addPPCallbacks(PPCallbacks *C) { 527 if (Callbacks) 528 C = new PPChainedCallbacks(C, Callbacks); 529 Callbacks = C; 530 } 531 532 /// \brief Given an identifier, return its latest MacroDirective if it is 533 // \#defined or null if it isn't \#define'd. 534 MacroDirective *getMacroDirective(IdentifierInfo *II) const { 535 if (!II->hasMacroDefinition()) 536 return 0; 537 538 MacroDirective *MD = getMacroDirectiveHistory(II); 539 assert(MD->isDefined() && "Macro is undefined!"); 540 return MD; 541 } 542 543 const MacroInfo *getMacroInfo(IdentifierInfo *II) const { 544 return const_cast<Preprocessor*>(this)->getMacroInfo(II); 545 } 546 547 MacroInfo *getMacroInfo(IdentifierInfo *II) { 548 if (MacroDirective *MD = getMacroDirective(II)) 549 return MD->getMacroInfo(); 550 return 0; 551 } 552 553 /// \brief Given an identifier, return the (probably #undef'd) MacroInfo 554 /// representing the most recent macro definition. One can iterate over all 555 /// previous macro definitions from it. This method should only be called for 556 /// identifiers that hadMacroDefinition(). 557 MacroDirective *getMacroDirectiveHistory(const IdentifierInfo *II) const; 558 559 /// \brief Add a directive to the macro directive history for this identifier. 560 void appendMacroDirective(IdentifierInfo *II, MacroDirective *MD); 561 DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II, MacroInfo *MI, 562 SourceLocation Loc, 563 bool isImported) { 564 DefMacroDirective *MD = AllocateDefMacroDirective(MI, Loc, isImported); 565 appendMacroDirective(II, MD); 566 return MD; 567 } 568 DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II, MacroInfo *MI){ 569 return appendDefMacroDirective(II, MI, MI->getDefinitionLoc(), false); 570 } 571 /// \brief Set a MacroDirective that was loaded from a PCH file. 572 void setLoadedMacroDirective(IdentifierInfo *II, MacroDirective *MD); 573 574 /// macro_iterator/macro_begin/macro_end - This allows you to walk the macro 575 /// history table. Currently defined macros have 576 /// IdentifierInfo::hasMacroDefinition() set and an empty 577 /// MacroInfo::getUndefLoc() at the head of the list. 578 typedef llvm::DenseMap<const IdentifierInfo *, 579 MacroDirective*>::const_iterator macro_iterator; 580 macro_iterator macro_begin(bool IncludeExternalMacros = true) const; 581 macro_iterator macro_end(bool IncludeExternalMacros = true) const; 582 583 /// \brief Return the name of the macro defined before \p Loc that has 584 /// spelling \p Tokens. If there are multiple macros with same spelling, 585 /// return the last one defined. 586 StringRef getLastMacroWithSpelling(SourceLocation Loc, 587 ArrayRef<TokenValue> Tokens) const; 588 589 const std::string &getPredefines() const { return Predefines; } 590 /// setPredefines - Set the predefines for this Preprocessor. These 591 /// predefines are automatically injected when parsing the main file. 592 void setPredefines(const char *P) { Predefines = P; } 593 void setPredefines(const std::string &P) { Predefines = P; } 594 595 /// Return information about the specified preprocessor 596 /// identifier token. 597 IdentifierInfo *getIdentifierInfo(StringRef Name) const { 598 return &Identifiers.get(Name); 599 } 600 601 /// AddPragmaHandler - Add the specified pragma handler to the preprocessor. 602 /// If 'Namespace' is non-null, then it is a token required to exist on the 603 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC". 604 void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler); 605 void AddPragmaHandler(PragmaHandler *Handler) { 606 AddPragmaHandler(StringRef(), Handler); 607 } 608 609 /// RemovePragmaHandler - Remove the specific pragma handler from 610 /// the preprocessor. If \p Namespace is non-null, then it should 611 /// be the namespace that \p Handler was added to. It is an error 612 /// to remove a handler that has not been registered. 613 void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler); 614 void RemovePragmaHandler(PragmaHandler *Handler) { 615 RemovePragmaHandler(StringRef(), Handler); 616 } 617 618 /// \brief Add the specified comment handler to the preprocessor. 619 void addCommentHandler(CommentHandler *Handler); 620 621 /// \brief Remove the specified comment handler. 622 /// 623 /// It is an error to remove a handler that has not been registered. 624 void removeCommentHandler(CommentHandler *Handler); 625 626 /// \brief Set the code completion handler to the given object. 627 void setCodeCompletionHandler(CodeCompletionHandler &Handler) { 628 CodeComplete = &Handler; 629 } 630 631 /// \brief Retrieve the current code-completion handler. 632 CodeCompletionHandler *getCodeCompletionHandler() const { 633 return CodeComplete; 634 } 635 636 /// \brief Clear out the code completion handler. 637 void clearCodeCompletionHandler() { 638 CodeComplete = 0; 639 } 640 641 /// \brief Hook used by the lexer to invoke the "natural language" code 642 /// completion point. 643 void CodeCompleteNaturalLanguage(); 644 645 /// \brief Retrieve the preprocessing record, or NULL if there is no 646 /// preprocessing record. 647 PreprocessingRecord *getPreprocessingRecord() const { return Record; } 648 649 /// \brief Create a new preprocessing record, which will keep track of 650 /// all macro expansions, macro definitions, etc. 651 void createPreprocessingRecord(); 652 653 /// EnterMainSourceFile - Enter the specified FileID as the main source file, 654 /// which implicitly adds the builtin defines etc. 655 void EnterMainSourceFile(); 656 657 /// EndSourceFile - Inform the preprocessor callbacks that processing is 658 /// complete. 659 void EndSourceFile(); 660 661 /// EnterSourceFile - Add a source file to the top of the include stack and 662 /// start lexing tokens from it instead of the current buffer. Emit an error 663 /// and don't enter the file on error. 664 void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir, 665 SourceLocation Loc); 666 667 /// EnterMacro - Add a Macro to the top of the include stack and start lexing 668 /// tokens from it instead of the current buffer. Args specifies the 669 /// tokens input to a function-like macro. 670 /// 671 /// ILEnd specifies the location of the ')' for a function-like macro or the 672 /// identifier for an object-like macro. 673 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroInfo *Macro, 674 MacroArgs *Args); 675 676 /// EnterTokenStream - Add a "macro" context to the top of the include stack, 677 /// which will cause the lexer to start returning the specified tokens. 678 /// 679 /// If DisableMacroExpansion is true, tokens lexed from the token stream will 680 /// not be subject to further macro expansion. Otherwise, these tokens will 681 /// be re-macro-expanded when/if expansion is enabled. 682 /// 683 /// If OwnsTokens is false, this method assumes that the specified stream of 684 /// tokens has a permanent owner somewhere, so they do not need to be copied. 685 /// If it is true, it assumes the array of tokens is allocated with new[] and 686 /// must be freed. 687 /// 688 void EnterTokenStream(const Token *Toks, unsigned NumToks, 689 bool DisableMacroExpansion, bool OwnsTokens); 690 691 /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the 692 /// lexer stack. This should only be used in situations where the current 693 /// state of the top-of-stack lexer is known. 694 void RemoveTopOfLexerStack(); 695 696 /// EnableBacktrackAtThisPos - From the point that this method is called, and 697 /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor 698 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will 699 /// make the Preprocessor re-lex the same tokens. 700 /// 701 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can 702 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will 703 /// be combined with the EnableBacktrackAtThisPos calls in reverse order. 704 /// 705 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack 706 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of 707 /// tokens will continue indefinitely. 708 /// 709 void EnableBacktrackAtThisPos(); 710 711 /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call. 712 void CommitBacktrackedTokens(); 713 714 /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since 715 /// EnableBacktrackAtThisPos() was previously called. 716 void Backtrack(); 717 718 /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and 719 /// caching of tokens is on. 720 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); } 721 722 /// Lex - Lex the next token for this preprocessor. 723 void Lex(Token &Result); 724 725 void LexAfterModuleImport(Token &Result); 726 727 /// \brief Lex a string literal, which may be the concatenation of multiple 728 /// string literals and may even come from macro expansion. 729 /// \returns true on success, false if a error diagnostic has been generated. 730 bool LexStringLiteral(Token &Result, std::string &String, 731 const char *DiagnosticTag, bool AllowMacroExpansion) { 732 if (AllowMacroExpansion) 733 Lex(Result); 734 else 735 LexUnexpandedToken(Result); 736 return FinishLexStringLiteral(Result, String, DiagnosticTag, 737 AllowMacroExpansion); 738 } 739 740 /// \brief Complete the lexing of a string literal where the first token has 741 /// already been lexed (see LexStringLiteral). 742 bool FinishLexStringLiteral(Token &Result, std::string &String, 743 const char *DiagnosticTag, 744 bool AllowMacroExpansion); 745 746 /// LexNonComment - Lex a token. If it's a comment, keep lexing until we get 747 /// something not a comment. This is useful in -E -C mode where comments 748 /// would foul up preprocessor directive handling. 749 void LexNonComment(Token &Result) { 750 do 751 Lex(Result); 752 while (Result.getKind() == tok::comment); 753 } 754 755 /// LexUnexpandedToken - This is just like Lex, but this disables macro 756 /// expansion of identifier tokens. 757 void LexUnexpandedToken(Token &Result) { 758 // Disable macro expansion. 759 bool OldVal = DisableMacroExpansion; 760 DisableMacroExpansion = true; 761 // Lex the token. 762 Lex(Result); 763 764 // Reenable it. 765 DisableMacroExpansion = OldVal; 766 } 767 768 /// LexUnexpandedNonComment - Like LexNonComment, but this disables macro 769 /// expansion of identifier tokens. 770 void LexUnexpandedNonComment(Token &Result) { 771 do 772 LexUnexpandedToken(Result); 773 while (Result.getKind() == tok::comment); 774 } 775 776 /// Disables macro expansion everywhere except for preprocessor directives. 777 void SetMacroExpansionOnlyInDirectives() { 778 DisableMacroExpansion = true; 779 MacroExpansionInDirectivesOverride = true; 780 } 781 782 /// LookAhead - This peeks ahead N tokens and returns that token without 783 /// consuming any tokens. LookAhead(0) returns the next token that would be 784 /// returned by Lex(), LookAhead(1) returns the token after it, etc. This 785 /// returns normal tokens after phase 5. As such, it is equivalent to using 786 /// 'Lex', not 'LexUnexpandedToken'. 787 const Token &LookAhead(unsigned N) { 788 if (CachedLexPos + N < CachedTokens.size()) 789 return CachedTokens[CachedLexPos+N]; 790 else 791 return PeekAhead(N+1); 792 } 793 794 /// RevertCachedTokens - When backtracking is enabled and tokens are cached, 795 /// this allows to revert a specific number of tokens. 796 /// Note that the number of tokens being reverted should be up to the last 797 /// backtrack position, not more. 798 void RevertCachedTokens(unsigned N) { 799 assert(isBacktrackEnabled() && 800 "Should only be called when tokens are cached for backtracking"); 801 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back()) 802 && "Should revert tokens up to the last backtrack position, not more"); 803 assert(signed(CachedLexPos) - signed(N) >= 0 && 804 "Corrupted backtrack positions ?"); 805 CachedLexPos -= N; 806 } 807 808 /// EnterToken - Enters a token in the token stream to be lexed next. If 809 /// BackTrack() is called afterwards, the token will remain at the insertion 810 /// point. 811 void EnterToken(const Token &Tok) { 812 EnterCachingLexMode(); 813 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok); 814 } 815 816 /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching 817 /// tokens (because backtrack is enabled) it should replace the most recent 818 /// cached tokens with the given annotation token. This function has no effect 819 /// if backtracking is not enabled. 820 /// 821 /// Note that the use of this function is just for optimization; so that the 822 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is 823 /// invoked. 824 void AnnotateCachedTokens(const Token &Tok) { 825 assert(Tok.isAnnotation() && "Expected annotation token"); 826 if (CachedLexPos != 0 && isBacktrackEnabled()) 827 AnnotatePreviousCachedTokens(Tok); 828 } 829 830 /// Get the location of the last cached token, suitable for setting the end 831 /// location of an annotation token. 832 SourceLocation getLastCachedTokenLocation() const { 833 assert(CachedLexPos != 0); 834 return CachedTokens[CachedLexPos-1].getLocation(); 835 } 836 837 /// \brief Replace the last token with an annotation token. 838 /// 839 /// Like AnnotateCachedTokens(), this routine replaces an 840 /// already-parsed (and resolved) token with an annotation 841 /// token. However, this routine only replaces the last token with 842 /// the annotation token; it does not affect any other cached 843 /// tokens. This function has no effect if backtracking is not 844 /// enabled. 845 void ReplaceLastTokenWithAnnotation(const Token &Tok) { 846 assert(Tok.isAnnotation() && "Expected annotation token"); 847 if (CachedLexPos != 0 && isBacktrackEnabled()) 848 CachedTokens[CachedLexPos-1] = Tok; 849 } 850 851 /// TypoCorrectToken - Update the current token to represent the provided 852 /// identifier, in order to cache an action performed by typo correction. 853 void TypoCorrectToken(const Token &Tok) { 854 assert(Tok.getIdentifierInfo() && "Expected identifier token"); 855 if (CachedLexPos != 0 && isBacktrackEnabled()) 856 CachedTokens[CachedLexPos-1] = Tok; 857 } 858 859 /// \brief Recompute the current lexer kind based on the CurLexer/CurPTHLexer/ 860 /// CurTokenLexer pointers. 861 void recomputeCurLexerKind(); 862 863 /// \brief Returns true if incremental processing is enabled 864 bool isIncrementalProcessingEnabled() const { return IncrementalProcessing; } 865 866 /// \brief Enables the incremental processing 867 void enableIncrementalProcessing(bool value = true) { 868 IncrementalProcessing = value; 869 } 870 871 /// \brief Specify the point at which code-completion will be performed. 872 /// 873 /// \param File the file in which code completion should occur. If 874 /// this file is included multiple times, code-completion will 875 /// perform completion the first time it is included. If NULL, this 876 /// function clears out the code-completion point. 877 /// 878 /// \param Line the line at which code completion should occur 879 /// (1-based). 880 /// 881 /// \param Column the column at which code completion should occur 882 /// (1-based). 883 /// 884 /// \returns true if an error occurred, false otherwise. 885 bool SetCodeCompletionPoint(const FileEntry *File, 886 unsigned Line, unsigned Column); 887 888 /// \brief Determine if we are performing code completion. 889 bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; } 890 891 /// \brief Returns the location of the code-completion point. 892 /// Returns an invalid location if code-completion is not enabled or the file 893 /// containing the code-completion point has not been lexed yet. 894 SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; } 895 896 /// \brief Returns the start location of the file of 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 getCodeCompletionFileLoc() const { 900 return CodeCompletionFileLoc; 901 } 902 903 /// \brief Returns true if code-completion is enabled and we have hit the 904 /// code-completion point. 905 bool isCodeCompletionReached() const { return CodeCompletionReached; } 906 907 /// \brief Note that we hit the code-completion point. 908 void setCodeCompletionReached() { 909 assert(isCodeCompletionEnabled() && "Code-completion not enabled!"); 910 CodeCompletionReached = true; 911 // Silence any diagnostics that occur after we hit the code-completion. 912 getDiagnostics().setSuppressAllDiagnostics(true); 913 } 914 915 /// \brief The location of the currently-active \#pragma clang 916 /// arc_cf_code_audited begin. Returns an invalid location if there 917 /// is no such pragma active. 918 SourceLocation getPragmaARCCFCodeAuditedLoc() const { 919 return PragmaARCCFCodeAuditedLoc; 920 } 921 922 /// \brief Set the location of the currently-active \#pragma clang 923 /// arc_cf_code_audited begin. An invalid location ends the pragma. 924 void setPragmaARCCFCodeAuditedLoc(SourceLocation Loc) { 925 PragmaARCCFCodeAuditedLoc = Loc; 926 } 927 928 /// \brief Instruct the preprocessor to skip part of the main source file. 929 /// 930 /// \param Bytes The number of bytes in the preamble to skip. 931 /// 932 /// \param StartOfLine Whether skipping these bytes puts the lexer at the 933 /// start of a line. 934 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) { 935 SkipMainFilePreamble.first = Bytes; 936 SkipMainFilePreamble.second = StartOfLine; 937 } 938 939 /// Diag - Forwarding function for diagnostics. This emits a diagnostic at 940 /// the specified Token's location, translating the token's start 941 /// position in the current buffer into a SourcePosition object for rendering. 942 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const { 943 return Diags->Report(Loc, DiagID); 944 } 945 946 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const { 947 return Diags->Report(Tok.getLocation(), DiagID); 948 } 949 950 /// getSpelling() - Return the 'spelling' of the token at the given 951 /// location; does not go up to the spelling location or down to the 952 /// expansion location. 953 /// 954 /// \param buffer A buffer which will be used only if the token requires 955 /// "cleaning", e.g. if it contains trigraphs or escaped newlines 956 /// \param invalid If non-null, will be set \c true if an error occurs. 957 StringRef getSpelling(SourceLocation loc, 958 SmallVectorImpl<char> &buffer, 959 bool *invalid = 0) const { 960 return Lexer::getSpelling(loc, buffer, SourceMgr, LangOpts, invalid); 961 } 962 963 /// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a 964 /// token is the characters used to represent the token in the source file 965 /// after trigraph expansion and escaped-newline folding. In particular, this 966 /// wants to get the true, uncanonicalized, spelling of things like digraphs 967 /// UCNs, etc. 968 /// 969 /// \param Invalid If non-null, will be set \c true if an error occurs. 970 std::string getSpelling(const Token &Tok, bool *Invalid = 0) const { 971 return Lexer::getSpelling(Tok, SourceMgr, LangOpts, Invalid); 972 } 973 974 /// getSpelling - This method is used to get the spelling of a token into a 975 /// preallocated buffer, instead of as an std::string. The caller is required 976 /// to allocate enough space for the token, which is guaranteed to be at least 977 /// Tok.getLength() bytes long. The length of the actual result is returned. 978 /// 979 /// Note that this method may do two possible things: it may either fill in 980 /// the buffer specified with characters, or it may *change the input pointer* 981 /// to point to a constant buffer with the data already in it (avoiding a 982 /// copy). The caller is not allowed to modify the returned buffer pointer 983 /// if an internal buffer is returned. 984 unsigned getSpelling(const Token &Tok, const char *&Buffer, 985 bool *Invalid = 0) const { 986 return Lexer::getSpelling(Tok, Buffer, SourceMgr, LangOpts, Invalid); 987 } 988 989 /// getSpelling - This method is used to get the spelling of a token into a 990 /// SmallVector. Note that the returned StringRef may not point to the 991 /// supplied buffer if a copy can be avoided. 992 StringRef getSpelling(const Token &Tok, 993 SmallVectorImpl<char> &Buffer, 994 bool *Invalid = 0) const; 995 996 /// \brief Relex the token at the specified location. 997 /// \returns true if there was a failure, false on success. 998 bool getRawToken(SourceLocation Loc, Token &Result, 999 bool IgnoreWhiteSpace = false) { 1000 return Lexer::getRawToken(Loc, Result, SourceMgr, LangOpts, IgnoreWhiteSpace); 1001 } 1002 1003 /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant 1004 /// with length 1, return the character. 1005 char getSpellingOfSingleCharacterNumericConstant(const Token &Tok, 1006 bool *Invalid = 0) const { 1007 assert(Tok.is(tok::numeric_constant) && 1008 Tok.getLength() == 1 && "Called on unsupported token"); 1009 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1"); 1010 1011 // If the token is carrying a literal data pointer, just use it. 1012 if (const char *D = Tok.getLiteralData()) 1013 return *D; 1014 1015 // Otherwise, fall back on getCharacterData, which is slower, but always 1016 // works. 1017 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid); 1018 } 1019 1020 /// \brief Retrieve the name of the immediate macro expansion. 1021 /// 1022 /// This routine starts from a source location, and finds the name of the macro 1023 /// responsible for its immediate expansion. It looks through any intervening 1024 /// macro argument expansions to compute this. It returns a StringRef which 1025 /// refers to the SourceManager-owned buffer of the source where that macro 1026 /// name is spelled. Thus, the result shouldn't out-live the SourceManager. 1027 StringRef getImmediateMacroName(SourceLocation Loc) { 1028 return Lexer::getImmediateMacroName(Loc, SourceMgr, getLangOpts()); 1029 } 1030 1031 /// CreateString - Plop the specified string into a scratch buffer and set the 1032 /// specified token's location and length to it. If specified, the source 1033 /// location provides a location of the expansion point of the token. 1034 void CreateString(StringRef Str, Token &Tok, 1035 SourceLocation ExpansionLocStart = SourceLocation(), 1036 SourceLocation ExpansionLocEnd = SourceLocation()); 1037 1038 /// \brief Computes the source location just past the end of the 1039 /// token at this source location. 1040 /// 1041 /// This routine can be used to produce a source location that 1042 /// points just past the end of the token referenced by \p Loc, and 1043 /// is generally used when a diagnostic needs to point just after a 1044 /// token where it expected something different that it received. If 1045 /// the returned source location would not be meaningful (e.g., if 1046 /// it points into a macro), this routine returns an invalid 1047 /// source location. 1048 /// 1049 /// \param Offset an offset from the end of the token, where the source 1050 /// location should refer to. The default offset (0) produces a source 1051 /// location pointing just past the end of the token; an offset of 1 produces 1052 /// a source location pointing to the last character in the token, etc. 1053 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) { 1054 return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, LangOpts); 1055 } 1056 1057 /// \brief Returns true if the given MacroID location points at the first 1058 /// token of the macro expansion. 1059 /// 1060 /// \param MacroBegin If non-null and function returns true, it is set to 1061 /// begin location of the macro. 1062 bool isAtStartOfMacroExpansion(SourceLocation loc, 1063 SourceLocation *MacroBegin = 0) const { 1064 return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, LangOpts, 1065 MacroBegin); 1066 } 1067 1068 /// \brief Returns true if the given MacroID location points at the last 1069 /// token of the macro expansion. 1070 /// 1071 /// \param MacroEnd If non-null and function returns true, it is set to 1072 /// end location of the macro. 1073 bool isAtEndOfMacroExpansion(SourceLocation loc, 1074 SourceLocation *MacroEnd = 0) const { 1075 return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, LangOpts, MacroEnd); 1076 } 1077 1078 /// DumpToken - Print the token to stderr, used for debugging. 1079 /// 1080 void DumpToken(const Token &Tok, bool DumpFlags = false) const; 1081 void DumpLocation(SourceLocation Loc) const; 1082 void DumpMacro(const MacroInfo &MI) const; 1083 1084 /// AdvanceToTokenCharacter - Given a location that specifies the start of a 1085 /// token, return a new location that specifies a character within the token. 1086 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart, 1087 unsigned Char) const { 1088 return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, LangOpts); 1089 } 1090 1091 /// IncrementPasteCounter - Increment the counters for the number of token 1092 /// paste operations performed. If fast was specified, this is a 'fast paste' 1093 /// case we handled. 1094 /// 1095 void IncrementPasteCounter(bool isFast) { 1096 if (isFast) 1097 ++NumFastTokenPaste; 1098 else 1099 ++NumTokenPaste; 1100 } 1101 1102 void PrintStats(); 1103 1104 size_t getTotalMemory() const; 1105 1106 /// HandleMicrosoftCommentPaste - When the macro expander pastes together a 1107 /// comment (/##/) in microsoft mode, this method handles updating the current 1108 /// state, returning the token on the next source line. 1109 void HandleMicrosoftCommentPaste(Token &Tok); 1110 1111 //===--------------------------------------------------------------------===// 1112 // Preprocessor callback methods. These are invoked by a lexer as various 1113 // directives and events are found. 1114 1115 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the 1116 /// identifier information for the token and install it into the token, 1117 /// updating the token kind accordingly. 1118 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const; 1119 1120 private: 1121 llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons; 1122 1123 public: 1124 1125 // SetPoisonReason - Call this function to indicate the reason for 1126 // poisoning an identifier. If that identifier is accessed while 1127 // poisoned, then this reason will be used instead of the default 1128 // "poisoned" diagnostic. 1129 void SetPoisonReason(IdentifierInfo *II, unsigned DiagID); 1130 1131 // HandlePoisonedIdentifier - Display reason for poisoned 1132 // identifier. 1133 void HandlePoisonedIdentifier(Token & Tok); 1134 1135 void MaybeHandlePoisonedIdentifier(Token & Identifier) { 1136 if(IdentifierInfo * II = Identifier.getIdentifierInfo()) { 1137 if(II->isPoisoned()) { 1138 HandlePoisonedIdentifier(Identifier); 1139 } 1140 } 1141 } 1142 1143 private: 1144 /// Identifiers used for SEH handling in Borland. These are only 1145 /// allowed in particular circumstances 1146 // __except block 1147 IdentifierInfo *Ident__exception_code, 1148 *Ident___exception_code, 1149 *Ident_GetExceptionCode; 1150 // __except filter expression 1151 IdentifierInfo *Ident__exception_info, 1152 *Ident___exception_info, 1153 *Ident_GetExceptionInfo; 1154 // __finally 1155 IdentifierInfo *Ident__abnormal_termination, 1156 *Ident___abnormal_termination, 1157 *Ident_AbnormalTermination; 1158 public: 1159 void PoisonSEHIdentifiers(bool Poison = true); // Borland 1160 1161 /// HandleIdentifier - This callback is invoked when the lexer reads an 1162 /// identifier and has filled in the tokens IdentifierInfo member. This 1163 /// callback potentially macro expands it or turns it into a named token (like 1164 /// 'for'). 1165 /// 1166 /// \returns true if we actually computed a token, false if we need to 1167 /// lex again. 1168 bool HandleIdentifier(Token &Identifier); 1169 1170 1171 /// HandleEndOfFile - This callback is invoked when the lexer hits the end of 1172 /// the current file. This either returns the EOF token and returns true, or 1173 /// pops a level off the include stack and returns false, at which point the 1174 /// client should call lex again. 1175 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false); 1176 1177 /// HandleEndOfTokenLexer - This callback is invoked when the current 1178 /// TokenLexer hits the end of its token stream. 1179 bool HandleEndOfTokenLexer(Token &Result); 1180 1181 /// HandleDirective - This callback is invoked when the lexer sees a # token 1182 /// at the start of a line. This consumes the directive, modifies the 1183 /// lexer/preprocessor state, and advances the lexer(s) so that the next token 1184 /// read is the correct one. 1185 void HandleDirective(Token &Result); 1186 1187 /// CheckEndOfDirective - Ensure that the next token is a tok::eod token. If 1188 /// not, emit a diagnostic and consume up until the eod. If EnableMacros is 1189 /// true, then we consider macros that expand to zero tokens as being ok. 1190 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false); 1191 1192 /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the 1193 /// current line until the tok::eod token is found. 1194 void DiscardUntilEndOfDirective(); 1195 1196 /// SawDateOrTime - This returns true if the preprocessor has seen a use of 1197 /// __DATE__ or __TIME__ in the file so far. 1198 bool SawDateOrTime() const { 1199 return DATELoc != SourceLocation() || TIMELoc != SourceLocation(); 1200 } 1201 unsigned getCounterValue() const { return CounterValue; } 1202 void setCounterValue(unsigned V) { CounterValue = V; } 1203 1204 /// \brief Retrieves the module that we're currently building, if any. 1205 Module *getCurrentModule(); 1206 1207 /// \brief Allocate a new MacroInfo object with the provided SourceLocation. 1208 MacroInfo *AllocateMacroInfo(SourceLocation L); 1209 1210 /// \brief Allocate a new MacroInfo object loaded from an AST file. 1211 MacroInfo *AllocateDeserializedMacroInfo(SourceLocation L, 1212 unsigned SubModuleID); 1213 1214 /// \brief Turn the specified lexer token into a fully checked and spelled 1215 /// filename, e.g. as an operand of \#include. 1216 /// 1217 /// The caller is expected to provide a buffer that is large enough to hold 1218 /// the spelling of the filename, but is also expected to handle the case 1219 /// when this method decides to use a different buffer. 1220 /// 1221 /// \returns true if the input filename was in <>'s or false if it was 1222 /// in ""'s. 1223 bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename); 1224 1225 /// \brief Given a "foo" or \<foo> reference, look up the indicated file. 1226 /// 1227 /// Returns null on failure. \p isAngled indicates whether the file 1228 /// reference is for system \#include's or not (i.e. using <> instead of ""). 1229 const FileEntry *LookupFile(SourceLocation FilenameLoc, StringRef Filename, 1230 bool isAngled, const DirectoryLookup *FromDir, 1231 const DirectoryLookup *&CurDir, 1232 SmallVectorImpl<char> *SearchPath, 1233 SmallVectorImpl<char> *RelativePath, 1234 ModuleMap::KnownHeader *SuggestedModule, 1235 bool SkipCache = false); 1236 1237 /// GetCurLookup - The DirectoryLookup structure used to find the current 1238 /// FileEntry, if CurLexer is non-null and if applicable. This allows us to 1239 /// implement \#include_next and find directory-specific properties. 1240 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; } 1241 1242 /// \brief Return true if we're in the top-level file, not in a \#include. 1243 bool isInPrimaryFile() const; 1244 1245 /// ConcatenateIncludeName - Handle cases where the \#include name is expanded 1246 /// from a macro as multiple tokens, which need to be glued together. This 1247 /// occurs for code like: 1248 /// \code 1249 /// \#define FOO <x/y.h> 1250 /// \#include FOO 1251 /// \endcode 1252 /// because in this case, "<x/y.h>" is returned as 7 tokens, not one. 1253 /// 1254 /// This code concatenates and consumes tokens up to the '>' token. It 1255 /// returns false if the > was found, otherwise it returns true if it finds 1256 /// and consumes the EOD marker. 1257 bool ConcatenateIncludeName(SmallString<128> &FilenameBuffer, 1258 SourceLocation &End); 1259 1260 /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is 1261 /// followed by EOD. Return true if the token is not a valid on-off-switch. 1262 bool LexOnOffSwitch(tok::OnOffSwitch &OOS); 1263 1264 private: 1265 1266 void PushIncludeMacroStack() { 1267 IncludeMacroStack.push_back(IncludeStackInfo(CurLexerKind, 1268 CurLexer.take(), 1269 CurPTHLexer.take(), 1270 CurPPLexer, 1271 CurTokenLexer.take(), 1272 CurDirLookup)); 1273 CurPPLexer = 0; 1274 } 1275 1276 void PopIncludeMacroStack() { 1277 CurLexer.reset(IncludeMacroStack.back().TheLexer); 1278 CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer); 1279 CurPPLexer = IncludeMacroStack.back().ThePPLexer; 1280 CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer); 1281 CurDirLookup = IncludeMacroStack.back().TheDirLookup; 1282 CurLexerKind = IncludeMacroStack.back().CurLexerKind; 1283 IncludeMacroStack.pop_back(); 1284 } 1285 1286 void PropagateLineStartLeadingSpaceInfo(Token &Result); 1287 1288 /// \brief Allocate a new MacroInfo object. 1289 MacroInfo *AllocateMacroInfo(); 1290 1291 DefMacroDirective *AllocateDefMacroDirective(MacroInfo *MI, 1292 SourceLocation Loc, 1293 bool isImported); 1294 UndefMacroDirective *AllocateUndefMacroDirective(SourceLocation UndefLoc); 1295 VisibilityMacroDirective *AllocateVisibilityMacroDirective(SourceLocation Loc, 1296 bool isPublic); 1297 1298 /// \brief Release the specified MacroInfo for re-use. 1299 /// 1300 /// This memory will be reused for allocating new MacroInfo objects. 1301 void ReleaseMacroInfo(MacroInfo* MI); 1302 1303 /// ReadMacroName - Lex and validate a macro name, which occurs after a 1304 /// \#define or \#undef. This emits a diagnostic, sets the token kind to eod, 1305 /// and discards the rest of the macro line if the macro name is invalid. 1306 void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0); 1307 1308 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro 1309 /// definition has just been read. Lex the rest of the arguments and the 1310 /// closing ), updating MI with what we learn and saving in LastTok the 1311 /// last token read. 1312 /// Return true if an error occurs parsing the arg list. 1313 bool ReadMacroDefinitionArgList(MacroInfo *MI, Token& LastTok); 1314 1315 /// We just read a \#if or related directive and decided that the 1316 /// subsequent tokens are in the \#if'd out portion of the 1317 /// file. Lex the rest of the file, until we see an \#endif. If \p 1318 /// FoundNonSkipPortion is true, then we have already emitted code for part of 1319 /// this \#if directive, so \#else/\#elif blocks should never be entered. If 1320 /// \p FoundElse is false, then \#else directives are ok, if not, then we have 1321 /// already seen one so a \#else directive is a duplicate. When this returns, 1322 /// the caller can lex the first valid token. 1323 void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc, 1324 bool FoundNonSkipPortion, bool FoundElse, 1325 SourceLocation ElseLoc = SourceLocation()); 1326 1327 /// \brief A fast PTH version of SkipExcludedConditionalBlock. 1328 void PTHSkipExcludedConditionalBlock(); 1329 1330 /// EvaluateDirectiveExpression - Evaluate an integer constant expression that 1331 /// may occur after a #if or #elif directive and return it as a bool. If the 1332 /// expression is equivalent to "!defined(X)" return X in IfNDefMacro. 1333 bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro); 1334 1335 /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas: 1336 /// \#pragma GCC poison/system_header/dependency and \#pragma once. 1337 void RegisterBuiltinPragmas(); 1338 1339 /// \brief Register builtin macros such as __LINE__ with the identifier table. 1340 void RegisterBuiltinMacros(); 1341 1342 /// HandleMacroExpandedIdentifier - If an identifier token is read that is to 1343 /// be expanded as a macro, handle it and return the next token as 'Tok'. If 1344 /// we lexed a token, return true; otherwise the caller should lex again. 1345 bool HandleMacroExpandedIdentifier(Token &Tok, MacroDirective *MD); 1346 1347 /// \brief Cache macro expanded tokens for TokenLexers. 1348 // 1349 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is 1350 /// going to lex in the cache and when it finishes the tokens are removed 1351 /// from the end of the cache. 1352 Token *cacheMacroExpandedTokens(TokenLexer *tokLexer, 1353 ArrayRef<Token> tokens); 1354 void removeCachedMacroExpandedTokensOfLastLexer(); 1355 friend void TokenLexer::ExpandFunctionArguments(); 1356 1357 /// isNextPPTokenLParen - Determine whether the next preprocessor token to be 1358 /// lexed is a '('. If so, consume the token and return true, if not, this 1359 /// method should have no observable side-effect on the lexed tokens. 1360 bool isNextPPTokenLParen(); 1361 1362 /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is 1363 /// invoked to read all of the formal arguments specified for the macro 1364 /// invocation. This returns null on error. 1365 MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI, 1366 SourceLocation &ExpansionEnd); 1367 1368 /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded 1369 /// as a builtin macro, handle it and return the next token as 'Tok'. 1370 void ExpandBuiltinMacro(Token &Tok); 1371 1372 /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then 1373 /// return the first token after the directive. The _Pragma token has just 1374 /// been read into 'Tok'. 1375 void Handle_Pragma(Token &Tok); 1376 1377 /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text 1378 /// is not enclosed within a string literal. 1379 void HandleMicrosoft__pragma(Token &Tok); 1380 1381 /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and 1382 /// start lexing tokens from it instead of the current buffer. 1383 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir); 1384 1385 /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and 1386 /// start getting tokens from it using the PTH cache. 1387 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir); 1388 1389 /// \brief Set the file ID for the preprocessor predefines. 1390 void setPredefinesFileID(FileID FID) { 1391 assert(PredefinesFileID.isInvalid() && "PredefinesFileID already set!"); 1392 PredefinesFileID = FID; 1393 } 1394 1395 /// IsFileLexer - Returns true if we are lexing from a file and not a 1396 /// pragma or a macro. 1397 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) { 1398 return L ? !L->isPragmaLexer() : P != 0; 1399 } 1400 1401 static bool IsFileLexer(const IncludeStackInfo& I) { 1402 return IsFileLexer(I.TheLexer, I.ThePPLexer); 1403 } 1404 1405 bool IsFileLexer() const { 1406 return IsFileLexer(CurLexer.get(), CurPPLexer); 1407 } 1408 1409 //===--------------------------------------------------------------------===// 1410 // Caching stuff. 1411 void CachingLex(Token &Result); 1412 bool InCachingLexMode() const { 1413 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means 1414 // that we are past EOF, not that we are in CachingLex mode. 1415 return !CurPPLexer && !CurTokenLexer && !CurPTHLexer && 1416 !IncludeMacroStack.empty(); 1417 } 1418 void EnterCachingLexMode(); 1419 void ExitCachingLexMode() { 1420 if (InCachingLexMode()) 1421 RemoveTopOfLexerStack(); 1422 } 1423 const Token &PeekAhead(unsigned N); 1424 void AnnotatePreviousCachedTokens(const Token &Tok); 1425 1426 //===--------------------------------------------------------------------===// 1427 /// Handle*Directive - implement the various preprocessor directives. These 1428 /// should side-effect the current preprocessor object so that the next call 1429 /// to Lex() will return the appropriate token next. 1430 void HandleLineDirective(Token &Tok); 1431 void HandleDigitDirective(Token &Tok); 1432 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning); 1433 void HandleIdentSCCSDirective(Token &Tok); 1434 void HandleMacroPublicDirective(Token &Tok); 1435 void HandleMacroPrivateDirective(Token &Tok); 1436 1437 // File inclusion. 1438 void HandleIncludeDirective(SourceLocation HashLoc, 1439 Token &Tok, 1440 const DirectoryLookup *LookupFrom = 0, 1441 bool isImport = false); 1442 void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok); 1443 void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok); 1444 void HandleImportDirective(SourceLocation HashLoc, Token &Tok); 1445 void HandleMicrosoftImportDirective(Token &Tok); 1446 1447 // Module inclusion testing. 1448 /// \brief Find the module for the source or header file that \p FilenameLoc 1449 /// points to. 1450 Module *getModuleForLocation(SourceLocation FilenameLoc); 1451 1452 /// \brief Verify that a private header is included only from within its 1453 /// module. 1454 bool violatesPrivateInclude(Module *RequestingModule, 1455 const FileEntry *IncFileEnt, 1456 ModuleMap::ModuleHeaderRole Role, 1457 Module *RequestedModule); 1458 1459 /// \brief Verify that a module includes headers only from modules that it 1460 /// has declared that it uses. 1461 bool violatesUseDeclarations(Module *RequestingModule, 1462 Module *RequestedModule); 1463 1464 /// \brief Verify that it is legal for the source file that \p FilenameLoc 1465 /// points to to include the file \p Filename. 1466 /// 1467 /// Tries to reuse \p IncFileEnt. 1468 void verifyModuleInclude(SourceLocation FilenameLoc, StringRef Filename, 1469 const FileEntry *IncFileEnt); 1470 1471 // Macro handling. 1472 void HandleDefineDirective(Token &Tok, bool ImmediatelyAfterTopLevelIfndef); 1473 void HandleUndefDirective(Token &Tok); 1474 1475 // Conditional Inclusion. 1476 void HandleIfdefDirective(Token &Tok, bool isIfndef, 1477 bool ReadAnyTokensBeforeDirective); 1478 void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective); 1479 void HandleEndifDirective(Token &Tok); 1480 void HandleElseDirective(Token &Tok); 1481 void HandleElifDirective(Token &Tok); 1482 1483 // Pragmas. 1484 void HandlePragmaDirective(SourceLocation IntroducerLoc, 1485 PragmaIntroducerKind Introducer); 1486 public: 1487 void HandlePragmaOnce(Token &OnceTok); 1488 void HandlePragmaMark(); 1489 void HandlePragmaPoison(Token &PoisonTok); 1490 void HandlePragmaSystemHeader(Token &SysHeaderTok); 1491 void HandlePragmaDependency(Token &DependencyTok); 1492 void HandlePragmaPushMacro(Token &Tok); 1493 void HandlePragmaPopMacro(Token &Tok); 1494 void HandlePragmaIncludeAlias(Token &Tok); 1495 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok); 1496 1497 // Return true and store the first token only if any CommentHandler 1498 // has inserted some tokens and getCommentRetentionState() is false. 1499 bool HandleComment(Token &Token, SourceRange Comment); 1500 1501 /// \brief A macro is used, update information about macros that need unused 1502 /// warnings. 1503 void markMacroAsUsed(MacroInfo *MI); 1504 }; 1505 1506 /// \brief Abstract base class that describes a handler that will receive 1507 /// source ranges for each of the comments encountered in the source file. 1508 class CommentHandler { 1509 public: 1510 virtual ~CommentHandler(); 1511 1512 // The handler shall return true if it has pushed any tokens 1513 // to be read using e.g. EnterToken or EnterTokenStream. 1514 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0; 1515 }; 1516 1517 } // end namespace clang 1518 1519 #endif 1520